Article - News - RetroFox.gay

Clear sky
7.5°C

Thursday, April 9th 2026

On Query: REPLACE INTO news_data (id, data, timestamp) VALUES ('%s', '%s', %i);

MySQL Error Code: 1406

MySQL Error Msgs: Data too long for column 'data' at row 1

array(5) {
[0]=>
array(7) {
["file"]=>
string(25) "/app/sys/base/sql.inc.php"
["line"]=>
int(63)
["function"]=>
string(14) "ReportSqlError"
["class"]=>
string(3) "SQL"
["object"]=>
object(SQL)#1 (6) {
["CONNECTION":"SQL":private]=>
object(mysqli)#2 (18) {
["affected_rows"]=>
int(-1)
["client_info"]=>
string(14) "mysqlnd 8.2.30"
["client_version"]=>
int(80230)
["connect_errno"]=>
int(0)
["connect_error"]=>
NULL
["errno"]=>
int(1406)
["error"]=>
string(40) "Data too long for column 'data' at row 1"
["error_list"]=>
array(1) {
[0]=>
array(3) {
["errno"]=>
int(1406)
["sqlstate"]=>
string(5) "22001"
["error"]=>
string(40) "Data too long for column 'data' at row 1"
}
}
["field_count"]=>
int(2)
["host_info"]=>
string(18) "db:3306 via TCP/IP"
["info"]=>
NULL
["insert_id"]=>
int(0)
["server_info"]=>
string(37) "11.2.2-MariaDB-1:11.2.2+maria~ubu2204"
["server_version"]=>
int(110202)
["sqlstate"]=>
string(5) "22001"
["protocol_version"]=>
int(10)
["thread_id"]=>
int(832256)
["warning_count"]=>
int(0)
}
["URL":"SQL":private]=>
string(7) "db:3306"
["USER":"SQL":private]=>
string(8) "retrofox"
["PASS":"SQL":private]=>
string(16) "TreeakWqQPD9qnWR"
["DB":"SQL":private]=>
string(8) "retrofox"
["DROP_VALUES":"SQL":private]=>
array(1) {
[0]=>
string(0) ""
}
}
["type"]=>
string(2) "->"
["args"]=>
array(1) {
[0]=>
string(69) "REPLACE INTO news_data (id, data, timestamp) VALUES ('%s', '%s', %i);"
}
}
[1]=>
array(7) {
["file"]=>
string(25) "/app/sys/base/sql.inc.php"
["line"]=>
int(74)
["function"]=>
string(5) "Query"
["class"]=>
string(3) "SQL"
["object"]=>
object(SQL)#1 (6) {
["CONNECTION":"SQL":private]=>
object(mysqli)#2 (18) {
["affected_rows"]=>
int(-1)
["client_info"]=>
string(14) "mysqlnd 8.2.30"
["client_version"]=>
int(80230)
["connect_errno"]=>
int(0)
["connect_error"]=>
NULL
["errno"]=>
int(1406)
["error"]=>
string(40) "Data too long for column 'data' at row 1"
["error_list"]=>
array(1) {
[0]=>
array(3) {
["errno"]=>
int(1406)
["sqlstate"]=>
string(5) "22001"
["error"]=>
string(40) "Data too long for column 'data' at row 1"
}
}
["field_count"]=>
int(2)
["host_info"]=>
string(18) "db:3306 via TCP/IP"
["info"]=>
NULL
["insert_id"]=>
int(0)
["server_info"]=>
string(37) "11.2.2-MariaDB-1:11.2.2+maria~ubu2204"
["server_version"]=>
int(110202)
["sqlstate"]=>
string(5) "22001"
["protocol_version"]=>
int(10)
["thread_id"]=>
int(832256)
["warning_count"]=>
int(0)
}
["URL":"SQL":private]=>
string(7) "db:3306"
["USER":"SQL":private]=>
string(8) "retrofox"
["PASS":"SQL":private]=>
string(16) "TreeakWqQPD9qnWR"
["DB":"SQL":private]=>
string(8) "retrofox"
["DROP_VALUES":"SQL":private]=>
array(1) {
[0]=>
string(0) ""
}
}
["type"]=>
string(2) "->"
["args"]=>
array(4) {
[0]=>
string(69) "REPLACE INTO news_data (id, data, timestamp) VALUES ('%s', '%s', %i);"
[1]=>
string(85) "foxapi_article_SNgSFz1UYwUgnAye8Iq6AwimFnBOT9jdyRE0kXeBSMWDBytDmFUTOmTLbAFJl2KUpiOq4O"
[2]=>
string(135713) "{"id":"SNgSFz1UYwUgnAye8Iq6AwimFnBOT9jdyRE0kXeBSMWDBytDmFUTOmTLbAFJl2KUpiOq4O","title":"Synthetic super-enhancers enable precision viral immunotherapy","description":"Cell-type-specific promoters are used in gene therapy to restrict expression of the therapeutic payload. However, these promoters often have suboptimal strength, selectivity and size. Here, leveraging recent insights into the function of enhancers, we developed synthetic super-enhancers (SSEs) by assembling functionally validated enhancer fragments into multipart arrays. Focusing on the core SOX2-driven and SOX9-driven transcriptional regulatory network in glioblastoma stem cells (GSCs)1, we engineered SSEs with robust activity and high selectivity.\u00a0Single-cell profiling, biochemical analyses and genome-binding data indicated that SSEs integrate neurodevelopmental and signalling-state transcription factors to trigger the formation of large multimeric complexes of transcription factors. Moreover, GSC-selective expression of a combination of cytotoxic (HSV-TK and ganciclovir) and immunomodulatory (IL-12) payloads, delivered using adeno-associated virus vectors, as a single treatment led to curative outcomes in a mouse model of aggressive glioblastoma. Notably, IL-12 induced an immunological memory that prevented tumour recurrence. The activity and selectivity of the adeno-associated virus and SSE were validated using primary human glioblastoma tissue and normal cortex samples. In summary, SSEs harness the unique core transcriptional programs that define the GSC phenotype and enable precision immune activation. This approach may have broader applications in other contexts when precise control of transgene expression in specific cell states is necessary. Synthetic super-enhancers enable specific delivery of anticancer payloads, achieving tumour elimination after a single dose in a mouse model of aggressive glioblastoma.","author":"Pollard, Steven M.","source":"Nature","publish":"April 8, 2026, 11:28 pm","canonical":"https:\/\/www.nature.com\/articles\/s41586-026-10329-6","text":["Main\r\nGene therapies require the targeted expression of payloads in specific cell populations to achieve appropriate therapeutic dosing and to minimize off-target effects. Various strategies to attain this selectivity can be explored, including delivery routes, capsid engineering and regulatory elements (for example, promoters, enhancers, untranslated regions and termination signals)2. Enhancers determine cell-type-specific expression by recruiting complementary transcription factors (TFs) at high density via TF-binding motifs (TFBMs)3,4.","However, natural enhancers often have suboptimal features for application in gene therapy vectors, such as size, strength, selectivity and quality of the sequence (for example, GC-rich or repetitive sequences). Strategies for creating artificial promoters and enhancers have typically focused on screening small (around 10\u2009bp) synthetic TFBMs and assembling them into concatamers5,6,7.","However, this approach is inadequate owing to the lack of a natural TFBM grammar, that is, the spacing, order, orientation and affinity of TFBMs, which are necessary for cell-type selectivity8.\r\n\r\nCell-type-specific enhancers are often clustered in the genome to create super-enhancers9,10. Super-enhancers typically regulate transcriptional programs associated with cell identity through the local clustering of low-affinity motifs8. This leads to increased high-density binding of lineage and signalling TFs to promote RNA polymerase\u2009II recruitment and high transcriptional output11,12.","We propose that synthetic super-enhancers (SSEs) can be engineered by assembling natural enhancer fragments associated with different genes into multipart arrays and then functionally screening them to identify desired activity and selectivity. The resulting transgene regulatory elements would capture the fundamental grammar of the combinations of TFs that define a specific cell type and signalling state. Such SSEs could have substantial value in gene therapy applications.\r\n\r\nWe focused on glioblastoma (GBM) to test this platform. GBM is an incurable brain cancer that is driven by cells with fetal neural stem-like phenotypes termed GBM stem cells (GSCs).","These cells express high levels of master regulatory TFs associated with neural stem and progenitor cells, including SOX2 (ref. 13). SOX2 is a pioneer TF and reprogramming factor essential for inducing and maintaining the identity of neural stem cells (NSCs)14,15 and GSCs1,16,17,18. SOX2 is also a crucial master regulatory TF that is broadly expressed across diverse GSC subtypes18.\r\n\r\nHere we develop SSEs for the targeted expression of anticancer payloads in GSCs.","A single, locally delivered dose of adeno-associated virus (AAV)-SSE-719 was evaluated in an immunocompetent model of aggressive GBM. We demonstrate that tumours are safely cleared using this precise and controlled method to express a dual payload of a cytotoxic (herpes simplex virus thymidine kinase and ganciclovir (HSV-TK\/GCV)) and a cytokine (IL-12).\r\nGSC-specific enhancer fragment design\r\nTFs bind to thousands of sites in the genome, but only a small subset are functional enhancers20.","To identify candidate GSC-selective enhancers, we first re-analysed previously published datasets1 that mapped SOX2-binding sites and used chromatin immunoprecipitation with sequencing (ChIP\u2013seq) to define those specific to GSCs and lost in their differentiated progeny.","Of the 1,710 candidate GSC-specific peaks identified (range of 115\u20133,366\u2009bp, mean of 400\u2009bp), a library of 160-bp double-stranded DNA oligonucleotides, with overlapping 100\u2009bp segments, was synthesized (Extended Data Fig. 1). The 160\u2009bp length was chosen as this was deemed to be the optimum size to capture the core set of TFBMs with the necessary natural grammar and motif diversity that might confer selectivity.","Moreover, this length is compatible with nucleosome formation for pioneer factors such as SOX2 (ref. 21). Candidate fragments were amplified by PCR, cloned into a luciferase\u2013mNeonGreen reporter plasmid and assembled into an arrayed plasmid library (48\u2009\u00d7\u200996-well plates with 4,579 individual plasmids) (Fig. 1a).\r\nFig. 1: Functional enhancer screening of an arrayed plasmid library.\r\na, Strategy for enhancer fragment (160\u2009bp) functional screening using an arrayed library of 4,579 plasmids with luciferase (Luc;\u00a0yellow) and mNeonGreen (mNGreen; green) reporters (single mRNA linked by P2A; grey), with an mCMV promoter (light blue).","b, Results for the enhancer fragment screen in GSCs (G7 cells) (n\u2009=\u20091).","c, Validation of 135 hits from the primary screen using independent luciferase reporter assays (NanoLuc DLR assay) (n\u2009=\u20093 biological replicates, error bars represent the s.d. of the arithmetic mean). The inset shows the enriched SOX dimer motif identified using the MEME tool.","d,\u00a0Validation of the top 16 significant hits from c, using flow cytometry to determine the percentage of mNeonGreen reporter-positive cells in an independent GSC line (E28), and the MFI per cell for each fragment (right). mCMV and the full-length CMV promoter were used as negative and positive reference controls, respectively (each dot represents a biological replicate (n\u2009=\u20093), error bars represent the s.d. of the arithmetic mean). Ordinary one-way analysis of variance (ANOVA) with Dunnet\u2019s multiple comparison test against mCMV was performed. For the mNeonGreen (%) data, adjusted P\u00a0< 0.0001 for ID3146,\u00a0P\u2009=\u20090.0018 for ID1101\u00a0and\u00a0P\u00a0=\u00a00.0003 for ID2339.","In the MFI mNeonGreen panel, P\u00a0= 0.0171 for ID2706, P\u00a0= 0.0092 for ID3146,\u00a0P\u2009=\u20090.0038 for ID1101, P\u2009=\u20090.0034 for ID2339 and\u00a0P\u2009=\u20090.0037 for ID2904.","e, Negative control normal human dermal fibroblasts (huFib70) using the same assays as d (each dot represents a biological replicate (n\u2009=\u20093), error bars represent the s.d. of the arithmetic mean).\u00a0*P\u2009\u2264\u20090.05, **P\u2009\u2264\u20090.01, ***P\u2009\u2264\u20090.001, ****P\u2009\u2264\u20090.0001.\r\n\r\nSource data\r\nDiscovery of functional GSC enhancers\r\nPatient-derived GSCs can be routinely propagated in a monolayer\u00a0and can be efficiently transfected13. We screened the enhancer fragment plasmid library in GSCs (specifically GSC7 cells) using a luciferase assay to identify functional hits22. From this initial screen, 135 plasmids exhibited more than 10-fold activity over a minimal CMV (mCMV) promoter (Fig. 1b).","Several hits were identified recurrently, which provided confidence in the screening assay.\r\n\r\nSanger sequencing confirmed sequences of the 135 initial hits, and 32 unique sequences were validated in triplicate with >20-fold activity compared with mCMV (Fig.","1c, Extended Data Table 1 and Supplementary Table 1). The top 16 candidates were also validated using an independent flow cytometry assay to detect the mNeonGreen transcriptional reporter, which provides single-cell quantification of activity and mean fluorescence intensity (MFI). All 16 validated hits were active in an independent patient-derived GSC line (E28), but were inactive in fibroblasts and HEK293 cells, which were used as SOX2-negative controls (Fig.","1d,e and Extended Data Fig.","1c,d). As anticipated, these hit sequences contained TFBMs for many known neurodevelopmental or GBM-associated TF families, including motifs for SOX, bHLH, FOX, TCF and SMAD (Extended Data Fig. 1e).\r\n\r\nNotably, de novo motif searches using the MEME tool revealed that the 32 active fragments contained a specific SOX dimer motif that resembles a previously reported SOX9 motif23 (Fig. 1c). Although SOX2 typically operates as a monomer, SOX9\u2014a known NSC self-renewal factor24\u2014can function as a homodimer23.","This result suggests that the functional enhancer fragments potentially bind SOX9.\r\n\r\nTo functionally validate the importance of this motif in the enhancer fragments, we created a series of mutants in the dimer motif of ID2904 (which was chosen because it only has a single motif) with nucleotide substitutions, inversions or changes in spacing. Each of these alterations abolished the transcriptional activity of the fragment (Extended Data Fig. 2a\u2013f). Higher activity was observed when the distance between the half-sites was reduced by 2\u2009bp. This result suggests that dimer binding is a crucial feature of the activity of the motif and is probably important for SOX9 binding.","Indeed, using biochemical pull-down, we confirmed that SOX9 binds to ID2904 and that binding depends on the SOX dimer motif (Extended Data Fig. 2g).\r\n\r\nThe above findings led us to speculate that SOX2 and SOX9 might work together at shared target enhancers in GSCs. To find their putative shared target sites, we mapped genome-wide binding of SOX2 and SOX9 in seven independent GSC patient-derived cell lines using ChIP\u2013seq (Fig. 2a). Most SOX9 peaks co-bound with SOX2, and these overlapped with previously reported super-enhancers found across GSCs (Fig. 2b\u2013d). These co-bound peaks were located near genes involved in stem cell function (Fig.","2e), including essential regulators of neural stem and progenitor cells (for example, CDK6 and PTPRZ1; Fig. 2h). The expression of SOX2 and SOX9 was positively correlated, particularly in GBM (relative to most other cancers), which implied that these two factors have a disease-specific function and\/or regulatory interactions (Fig. 2f).","Notably, the SOX2 and SOX9 peaks in co-bound sites at enhancers were enriched with multiple SOX monomer and dimer motifs, respectively, with an optimal dimer motif spacing of 2\u20133\u2009bp (Fig.","2g,i). Altogether, these data indicate a specific requirement for SOX9 dimerization at loci in which SOX2 and SOX9 are co-bound.\r\nFig. 2: ChIP\u2013seq analysis of SOX2 and SOX9 binding in seven different patient-derived GSC cultures.\r\na, Schematic of SOX2 and SOX9 ChIP\u2013seq experimental design and analysis.\u00a0GO, gene ontology.","b, Venn diagram showing the overlap between SOX2 and SOX9 peaks for the sample E28 (left), and heatmaps of SOX2 and SOX9 binding intensity over SOX9 and SOX2 peaks, respectively (right), demonstrating co-binding events.","c, Bar plots showing the proportion of SOX9 peaks located in co-bound sites.","d, Venn diagrams showing the overlaps between consensus SOX2 and SOX9 peaks and co-bound SOX2\u2013SOX9 peaks with a consensus set of GSC super-enhancers (n\u2009=\u200944 publicly available GSC H3K27ac datasets;\u00a0Methods).","e, Gene ontology analysis of co-bound SOX2\u2013SOX9 peaks in super-enhancers.","f, Bar plot showing the distribution of SOX2\u2013SOX9 expression correlation values (y\u2009axis) across diverse cancer types from The Cancer Genome Atlas (x\u2009axis).","g, Bar plot showing the motif co-occurrences of inverted palindromic SOX motifs, grouped by the distances between the primary and secondary SOX motif pairings.","h, UCSC genome browser coverage tracks of SOX2 and SOX9 ChIP\u2013seq in the seven individual GSCs at PTPRZ1 and CDK6 loci, annotated by the presence of GSC super-enhancers, co-bound SOX2\u2013SOX9 peaks and the functional SOX dimer motif.\u00a0cCREs, candidate cis-regulatory elements; Cons 100 Verts, conservation across 100 vertebrates.","i, The distributions of centrally enriched de novo SOX monomer and dimer motifs at SOX2 and SOX9 peaks, respectively, restricted to those in co-bound SOX2\u2013SOX9 enhancers.\r\n\r\nIn summary, using functional screening in patient-derived GSCs, we identified candidate 160\u2009bp SOX2 and SOX9 co-bound enhancer fragments with high selectivity but low activity (relative to CMV). Together with previously published findings, such as genome-wide CRISPR screens18, these results highlight the importance of SOX2 and SOX9 as core components of a TF circuit that determines and sustains GSC identity.","Moreover, they share crucial transcriptional targets.\r\nSSEs for GBM\r\nOur goal was to create a GSC-selective transcriptional switch for cancer gene therapy that combines strong expression and high selectivity. Such a SSE would provide the necessary therapeutic window to target the tumour but spare surrounding normal CNS tissue. We proposed that combining functional and selective enhancer fragments into a larger artificial multipart assembly might increase activity by facilitating increased high-density recruitment of SOX2 and SOX9 alongside other signalling-associated transcriptional cofactors to mimic the mechanism of action of natural super-enhancers11,12.","We chose four-part assemblies, as this represents an optimal size compatible with AAV vector applications. These assemblies were constructed using sequences from the top\u200916 validated hits grouped into four sets: 1\u20134, 5\u20138, 9\u201312 and 13\u201316 (Fig.","3a,c).\r\nFig. 3: Combining SOX2 functional enhancer fragments results in synergistic increases in activity without compromising selectivity.\r\na, Activity of the 16 top hits relative to mCMV and CMV assessed by flow cytometry in GSCs (G7 cells) (each dot represents a biological replicate, error\u00a0bars represent the s.d. of the mean).","b, SSE activity detected using an mNeonGreen reporter across GSC lines (GSC7 and E28) and control fibroblasts. Error bars represent the s.d. of the mean. One-way ANOVA with Dunnett\u2019s multiple comparisons test against mCMV, adjusted P\u2009values for G7: P\u2009=\u20090.0001 for SSE-1, P\u2009=\u20090.0009 for SSE-3, P\u2009=\u20090.0003 for SSE-5 and P\u2009<\u20090.0001 for SSE-7. Adjusted P\u2009values for E28: P\u2009=\u20090.0099 for SSE-1, P\u2009=\u20090.0048 for SSE-3, P\u2009=\u20090.0075 for SSE-5 and P\u2009=\u20090.0008 for SSE-7.","UT, untreated.","c, Schematic of each SSE-1, SSE-3, SSE-5 and SSE-7 transgene design.","d, EMSA of SSE-7 DNA following incubation with GSC lysates compared with differentiated GSCs (representative image; n\u2009=\u20093).","e, EMSA showing SSE-7 DNA with fibroblast (hFib) lysate (left) and hFib with SOX2 and SOX9 expressed individually (middle) or in combination (right). Representative images shown (n\u2009=\u20093).","f, EMSA of SSE-7 DNA (left, representative images of n\u2009=\u20093) and its constituent fragments (right, n\u2009=\u20091), incubated with SOX2 and SOX9. Schematic of SOX motifs (thin black bar, partial SOX motif; thicker, SOX motif; thickest, dimer motif) is shown underneath the images.","g, SABER\u2013FISH analysis of transgenes. Representative images are shown (left), with quantification (right).","For SSE-7, mean copy number\u2009=\u2009167 (median, 119; Q1\u2013Q3, 76\u2013209; min\u2013max, 8\u2013806).","For CMV, mean copy number\u2009=\u2009115 (median, 87; Q1\u2013Q3, 57\u2013127; min\u2013max, 5\u2013761). Scale bars, 20\u00a0\u03bcm.","h, scRNA-seq analysis shows no correlation between SSE-7 activity transcriptional subtypes (AC, astrocyte; MES, mesenchymal; NPC, neural progenitor cell).\u00a0Neftel subtype, subtype as identified by Neftel et al.27. Distribution of SSE-7\u2013mCherryhigh cells using UMAP plots of scRNA-seq analysis in E21 cells (n\u2009=\u200923,384 cells) and E28 cells (n\u2009=\u200935,879).","i, SCENIC analysis to determine TF enrichment across replicate libraries. Combined Z\u2009scores from each library. *P\u2009\u2264\u20090.05, **P\u2009\u2264\u20090.01, ***P\u2009\u2264\u20090.001; NS, not significant. P\u2009values were calculated using repeated-measures ANOVA, followed by Bonferroni-adjusted pairwise comparisons of estimated marginal means. Boxes show the interquartile range, centre lines the median, and whiskers the minimum and maximum values.\r\n\r\nSource data\r\n\r\nEach variant displayed strong activity that matched or exceeded that of a full-length CMV promoter in GSCs. By contrast, the variants had low activity in both fibroblasts and HEK293 cells, such that levels were comparable to mCMV-like levels (Fig. 3b and Extended Data Fig. 4a).","Given the principles that underpin their design, their high activity (switch-like behaviour) and their high cell-type selectivity, we herein refer to these artificial elements as SSEs.\r\n\r\nTo initially explore selectivity in glial cell lineage differentiation, we evaluated SSE activity during GSC differentiation into astrocyte-like cells1,13. We observed a twofold reduction in MFI per cell within 24\u2009h, followed by an approximately tenfold decrease by day\u200910 of differentiation (Extended Data Fig.","4b,c). By contrast, the constitutive CMV promoter displayed high activity across all conditions. SSE activity is therefore highest in the immature stem cell-like state of GSCs and is rapidly reduced as cells differentiate.\r\n\r\nFocusing on SSE-7, we next tested a larger set of patient-derived GSCs that span diverse genetic and transcriptional subtypes25,26. SSE-7 was active across all seven patient-derived GSCs but had reduced levels in E37 (an atypical cell line with MYC amplifications and reduced SOX2 and SOX9 levels) (Extended Data Fig. 3).","However, there was variability across the cell lines in the percentage of positive cells in the population compared with CMV, which indicated heterogeneity in the cultures. This result was not explained by SOX2 or SOX9 protein levels, as these were similar across lines, which suggests that additional cooperating TFs might be necessary for SSE-7 activity.\r\n\r\nWe next explored the biochemical mechanism of the switch-like behaviour and selectivity of SSE-7 and the roles of SOX2 and SOX9 in this process.","To that end, we first used cell-free electrophoretic mobility shift assays (EMSAs) to analyse their interactions with SSE-7 alone and in combination using both cell lysates and cell-free recombinant SOX proteins.\r\n\r\nSSE-7 incubation with GSC cell extracts resulted in a major super-shift and the formation of large higher-order multimeric complexes, but this was not observed in GSCs that had differentiated into astrocytes (Fig. 3d). Using human fibroblasts with SOX2 and\/or SOX9 forced expression, we also observed a clear super-shifted band only when both factors were incubated together with SSE-7, but this effect was weaker compared with GSCs. This result highlights the importance of GSC-associated cofactors for triggering SSE activity (Fig.","3e).\r\n\r\nNext, to validate whether synergy exists with SOX2 and SOX9 co-binding, we used purified recombinant SOX2 and SOX9 proteins and tested their binding. When SOX2 alone was bound with SSE-7, it displayed a typical ladder-like banding pattern indicative of SOX2 binding individually at each motif in the 640-bp-long SSE-7 (Fig. 3f and Extended Data Fig. 4d\u2013f).","Notably, the binding of SOX9 led to the formation of the higher-order complexes, and this effect was enhanced in the presence of SOX2 (Fig. 3f). The analysis of each constituent 160\u2009bp enhancer fragment in isolation demonstrated that the formation of higher-order complexes depends on the architecture of the SSE (Fig. 3f). No higher-order complexes were formed when using a control 40\u2009bp FGF4 enhancer that contains a single SOX motif (Extended Data Fig. 4g).","We therefore conclude that a key feature of the mechanism of action of SSE-7 is its ability to induce higher-order TF complexes through high-density local recruitment of SOX2 and SOX9.\r\nMechanistic basis of SSE activity in GSCs\r\nWe next aimed to identify candidate SOX-cooperating TFs and the cofactors required for the SSE switch-like transcriptional activation seen in GSCs. To that end, we performed single-cell RNA sequencing (scRNA-seq) analysis of 2,711 SSE-7\u2013mCherry-positive cells across 8 different GSC lines, comparing these to matched mCherry-negative populations.","We first used the single-molecule imaging technique signal amplification by exchange reaction and fluorescence in situ hybridization (SABER\u2013FISH) and confirmed that transduction rates and nuclear transgene\u00a0copy number were similar among the different cell lines and that there was no correlation with mCherry levels (Fig. 3g).","Therefore, differential transduction or nuclear processing of the AAV does not explain the restricted SSE-7 activity observed in heterogeneous GSC cultures compared with CMV.\r\n\r\nUniform manifold approximation and projection (UMAP) analysis confirmed that these GSC cultures have, as expected, a diverse range of transcriptional subtypes or states27. This transcriptional heterogeneity mirrored that seen in patient tumour samples and reflected distinct differentiation states (Fig. 3h). We did not find a simple correlation of SSE-7high activity with previously reported transcriptional subtypes (Fig.","3h), SOX2 and SOX9 expression or cell cycle phase (Extended Data Figs. 5 and\u00a06a).\r\n\r\nTo perform an unbiased search for TF activities enriched in SSE-7high cells, we used the analysis tool SCENIC. The results revealed a clear correlation of multiple TF activities linked to cancer signalling pathways, including STAT, IRF, SMAD and FOS\u2013JUN (Fig. 3i). Such signalling end-point TFs are often highly activated in cancers, including GBM, and motifs for such TF families are present in SSE-7. This finding is consistent with the idea that these TFs\u00a0have a potential functional role (Fig. 3i and Extended Data Fig.","6b,c). These data suggest that SSE-7 is activated in cells that display a combination of activated signalling end-point TFs and an immature stem cell-like identity driven by SOX2 and SOX9.\r\n\r\nWe next explored the requirements needed for signalling pathway activation for SSE-7 function. We performed a screen of 160 different small-molecule kinase inhibitors, spanning 70 distinct families, to identify those that would reduce SSE-7 activity (Extended Data Fig. 7a). Given the diverse signalling states across GSCs and the results from the SCENIC analysis, we speculated that activity would be highly dependent on the cell line.","We therefore focused the screen on E55 and E31, two GSC lines with typical phenotypic heterogeneity and distinct\u00a0transcriptional states.\r\n\r\nInitial hits were validated, and a clear dose-dependent reduction in mCherry MFI was confirmed for four of these hits (Extended Data Fig.","7b,c). These four inhibitors all operate in the MAPK\u2013ERK signalling pathway: three MEK inhibitors (PD0325901, AS-703026 and PD184161) and one ERK inhibitor (SC-1) (Extended Data Fig. 7d). For these specific GSC lines (E55 and E31), SOX2 and SOX9 activity together with hyperactive MAPK signalling seemed to be necessary for SSE-7 activity.","Finally, a subpopulation of sorted SSE-7-negative GSCs could reactivate SSE-7 expression at later time points. This result is consistent with the finding that SSE activity depends on a specific signalling state (Extended Data Fig. 7e).\r\n\r\nAltogether, these functional genetic, biochemical and genomic data indicate that SSE-7 is a dynamic transcriptional regulatory element that is strongly activated in specific GSC states that have a combination of both SOX2 and SOX9 expression and hyperactive signalling-associated TFs. The specific signalling TFs involved are likely to vary among the heterogeneous GSC genetic and transcriptional subtypes.","However, our SSE design strategy (that is, mixing distinct enhancer fragments from diverse genes) enabled us to achieve robust and strong expression across GSCs\u00a0regardless of this variability.\r\nCell-state-specific activation of SSE-7\r\nOur functional enhancer fragments typically had high vertebrate evolutionary conservation, which suggested that SSEs may work in different species. To further explore selectivity across diverse tissues and organs, we first tested SSE activity in developing zebrafish (around 48\u2009h after fertilization) (Extended Data Fig. 8).","Overlapping expression domains for each SSE-driven eGFP reporter was observed for all four SSEs tested, with consistently high levels of expression in the optic placodes and a subset of forebrain and spinal cord neural progenitors (Extended Data Fig.","8a,b). This restricted developmental expression pattern was consistent with previously reported zebrafish Sox2 and Sox9 expression27,28, and confirms that not only are the SSEs highly active but they are also highly tissue restricted.","Notably, SSE-1, SSE-3 and SSE-5 had an additional weaker expression domain in more posterior CNS tissues, whereas for SSE-3, we detected expression in the posterior endoderm (Extended Data Fig. 8b). We therefore prioritized SSE-7 for subsequent experiments.\r\n\r\nWe created a stable transgenic zebrafish line and used this to look at later larval stages in the nervous system (Extended Data Fig. 8c). SSE-7 activity was nonoverlapping with mature neuronal markers, which confirmed that SSE-7 is not expressed in neurons and is only seen in a subset of fetal neural\u00a0progenitor cells (Extended Data Fig. 8d\u2013g).","Notably, when we forced the expression of a potent oncogenic form of Akt in zebrafish, we observed that SSE-7 activity increased in those cells. This result is consistent with the requirement of hyperactive signalling together with Sox expression to support SSE activation, as seen in human GSCs (Extended Data Fig. 8h). SSE-7 had low activity in oligodendrocyte progenitor cells (OPCs) derived from human induced pluripotent stem (iPS) cells relative to CMV (Extended Data Fig.","8i),\u00a0but was expressed in human fetal NSC cultures (Extended Data Fig. 8j).","Altogether, these data suggest that SSE-7 activity is highly restricted to specific\u00a0cell states and may be most active in a subset of fetal neural progenitors.\r\nSSE-7 activity in GBM core and margin\r\nAAV is an excellent gene therapy vector for use in GBM owing to the following characteristics: (1) its small physical size (for enhanced distribution); (2) persistence in quiescent cancer cells; (3) low immunogenicity; (4) diverse natural capsids that might enable improved transduction; (5) validated efficient brain delivery via convection-enhanced delivery; and (6) well-understood manufacturing and clinical safety profile.\r\n\r\nTo further explore SSE-7 selectivity using relevant adult human neural cell types and the effectiveness of AAV as the delivery vector, we generated human glutamatergic and GABAergic differentiated neurons (ioGlutamatergic and ioGABAergic, respectively), and microglia (ioMicroglia) from iPS cells, and tested AAV1-SSE-7\u2013mCherry activity (Fig.","4a). SSE-7 was strongly expressed in GSCs within a few days of transduction, with similar levels to the CMV positive control.","However, in postmitotic neurons, we did not observe SSE-7 activity (Fig. 4a) at either day\u20093 or day\u200910. For microglia, there was poor transduction with AAV1.\r\nFig. 4: GSC selectivity of SSE-7 confirmed in vitro and in tissue slice cultures.\r\na, AAV1-SSE-7\u2013mCherry reporter activity in GSCs and human iPS-cell-derived differentiated neurons and microglia assessed by flow cytometry at day\u20093 (left) and day\u200910 (right) following transduction. For day\u20093 for ioGABAergic neurons and ioMicroglia, n\u2009=\u20093 biological independent cell experiments; ioGlutamatergic neurons and GSCs, n\u2009=\u20094 biological independent cell experiments.","For day\u200910, n\u2009=\u20093 biological independent cell experiments excluding ioGABAergic neurons, for which n\u2009=\u20092. Data are the mean\u2009\u00b1\u2009s.e.m., and statistical analysis was performed using an unpaired t-test. **P\u2009\u2264\u20090.01, ***P\u2009\u2264\u20090.001.","b, Experimental schematic of AAV1-SSE-7 to test selectivity in fresh human GBM tumour mass and tumour margin tissue samples (infiltrating cortex) processed into slice cultures (Methods).","c, Immunostaining and confocal images of tissue for mCherry (red) and GSC markers (Nestin and SOX2) demonstrates GSC selectivity of SSE-7 relative to CMV. Tumour mass and adjacent matched normal brain tissue (tumour margin) are shown.","d, Percentage of mCherry-expressing cells co-expressing SOX2. Each point represents a biological replicate, matched tumour and tumour margin samples are indicated by matching symbols. n\u2009=\u20093 for margin and n\u2009=\u20094 for the tumour mass. Statistical analysis was performed using two-way ANOVA with multiple-comparisons correction (P\u2009=\u20090.0436). Scale bars, 10\u2009\u00b5m. Schematic in b created in BioRender.","Pollard,\u2009S.\u00a0M.\u00a0(2026) https:\/\/BioRender.com\/l774d5t.\r\n\r\nSource data\r\n\r\nTo directly test whether SSE-7 would have strong differential activity in GBM tumour tissue compared with the surrounding normal brain, we determined the transduction and expression of AAV1-SSE-7\u2013mCherry in ex vivo fresh tumour tissue slices obtained from patients with GBM (Fig. 4b) and compared these with the tumour margin (which is macroscopically normal tissue but contains infiltrating GBM cells). We first tested whether AAV1-SSE-7 would be active in the tissue, focusing on the mCherry reporter only.","After 7\u2009days of incubation with a single dose of the virus, we fixed and stained the tumour tissue and compared CMV and SSE-7 activity using the mCherry reporter.\r\n\r\nThere were significantly fewer mCherry-positive cells when using SSE-7 compared with CMV (Fig. 4c and Extended Data Fig. 9b) in the tumour, and this was even more significant in the tumour margin tissue. This result is consistent with SSE-7 labelling a subset of GBM cells but not normal healthy tissue. We performed this analysis for four independent patient samples\u2014grade\u20094 IDH mutant astrocytoma, IDH wild-type GBM, a grade\u20093 astrocytoma and recurrent GBM (Extended Data Fig. 9a)\u2014and obtained similar results.","By contrast, as expected, CMV induced widespread high expression across diverse normal CNS cell types, including cells with neuronal and endothelial morphology (Fig. 4c and Extended Data Fig. 9).\r\n\r\nTo next confirm whether mCherry is enriched in the tumour cell population, we performed and quantified SOX2 co-staining. This experiment revealed significant differences between CMV and SSE-7, with SSE-7 activated in around 90% of mCherry-positive tumour cells that co-expressed SOX2 (Fig. 4d). We conclude that SSE-7 transgene expression is differentially active in tumour versus normal tissue (tumour margin) compared with CMV.","Moreover, SSE-7 provides a strong and selective gene switch that can restrict transgene activity to GSCs in the tumour mass and the infiltrating tumour margin.\r\nComplete responses in a mouse model of GBM\r\nWe next determined whether AAV1-SSE-7 could be used to selectively kill GSCs using expression of the HSV-TK\/GCV enzyme prodrug system, which has previously been applied in the clinic as a gene-directed enzyme prodrug therapy29. SSE-7 retained activity and provided selective expression of the HSV-TK payload in GSCs (Extended Data Fig. 10a\u2013d). This enabled cytotoxic killing of GSCs, after GCV treatment, whereas fibroblasts were unaffected (Extended Data Fig. 10e\u2013g).","Screening of common natural AAV serotypes confirmed that AAV1 is the optimal serotype of choice, as it efficiently transduced all seven patient-derived GSCs with high efficiency (Extended Data Fig.","11a,b).\r\n\r\nDespite the promise of HSV-TK as a suicide gene therapy approach for oncology, there has been limited success in the clinic. This is partially explained by the challenges of efficient delivery to all cancer cells30. As immune payloads can act non-autonomously to trigger an immune response against residual disease, we reasoned that combining an immune payload with a cytotoxic payload might be required to achieve this goal.","Indeed, such a strategy could induce multiple pathways of the immune response to polarize local macrophages in the tumour microenvironment while enabling systemic adaptive responses by T\u2009cells. We proposed that this combination\u2014tumour cell killing, tumour-associated macrophage polarization and T\u2009cell education and activation\u2014would be potent and reduce chances of tumour regrowth from residual disease30.\r\n\r\nWe chose the cytokine IL-12 for evaluation as it triggers a potent inflammatory response (stimulating adaptive and innate immunity) and is effective in promoting anticancer immune responses31,32. We reasoned that by using the AAV1-SSE-7 platform, we could avoid the toxicities previously seen when using systemic delivery31.","This would enable controlled release of IL-12 alongside tumour killing and induce effective antitumour immune responses, with IL-12 levels being reduced as tumour cells are lost. We designed an AAV1 vector with SSE-7 driving the expression of a single mRNA encoding HSV-TK linked via the P2A sequence to the mouse IL-12 heterodimer fusion (Il12a and Il12b, which encode P40 and P35, respectively) (Fig. 5a). This dual payload design ensures that the cytokine and cytotoxic effects are synchronized and serves as a safety switch to prevent uncontrolled IL-12 levels.\r\nFig.","5: The combination of a cytotoxic and immunomodulatory payload is highly effective at clearing established GBM tumours.\r\na, Architecture of the AAV1 transgene used to treat NPE-IE GBM in mice.","ITR, inverted terminal repeat;\u00a0mCMV, minimal CMV promoter;\u00a0mIL-12, mouse IL-12.","b, Survival curves for the mice treated with AAV1-SSE-7\u2013HSV-TK\u2013IL-12 (n\u2009=\u200924) versus control (UT (PBS); n\u2009=\u200922). P\u2009<\u20090.0001 for pooled (n\u2009=\u20093) independent cohorts. Survival was analysed using the\u00a0Kaplan\u2013Meier method. Significance was determined using a\u00a0log-rank (Mantel\u2013Cox) test. Mice were selected at week\u20091 using bioluminescence imaging to ensure a similar starting size for tumours across the cohorts.","c, Longitudinal tracking of tumour growth in mice once a week using bioluminescence imaging (three examples shown for experimental and control; see Extended Data Fig. 12 for the larger cohort).","d,e, Bar plots of the abundance of total T\u2009(CD3)\u00a0cells, effector T (Teff) cells, CD8+ T\u2009cells, regulatory T (Treg) cells and natural killer (NK) cells (d), and myeloid-derived suppressor cells (MDSCs), monocytes, bone marrow-derived macrophages (BMDMs), dendritic cells (DCs) and microglia (e) out of total live CD45+ cells in the tumour microenvironment of untreated mice and mice treated with a single dose of AAV-SSE-7-HSV-TK\u2013IL-12 or with the AAV-SSE-7-HSV-TK\u2013IL-12\u2009+\u2009GCV combination, determined by flow cytometry.","f,g, Bar plots of the abundance of PD1\u2212Ki67\u2212, PD1\u2212Ki67+, PD1+Ki67\u2212 and PD1+Ki67+ cells from Teff cells (f) and from CD8+ T\u2009cells (g) in the tumour microenvironment of untreated mice and mice treated with a single dose of AAV-SSE-7-HSV-TK\u2013IL-12 or with the AAV-SSE-7-HSV-TK-IL-12\u2009+\u2009GCV combination, determined by flow cytometry. n\u2009=\u20097 biologically independent animals per group, examined in one experiment. Data are the mean\u2009\u00b1\u2009s.d. Significance was assessed using one-way ANOVA with Tukey\u2019s multiple comparison test.","*P\u2009\u2264\u20090.05, **P\u2009\u2264\u20090.01, ***P\u2009\u2264\u20090.001, ****P\u2009\u2264\u20090.0001.\r\n\r\nSource data\r\n\r\nFor in vivo preclinical efficacy studies, we used our previously reported disease-relevant syngeneic mouse model of GBM, termed NPE-IE25, which exhibits a myelosuppressive tumour microenvironment and transcriptional programs seen in human GBMs25. AAV1-SSE-7 transduction and SSE activity were first validated using mouse NPE-IE cells in vitro (Extended Data Fig.","11c,d).\r\n\r\nAAV1-SSE-7\u2013HSV-TK\u2013P2A-IL-12 was delivered to a cohort of mice with established orthotopic tumours (14\u2009days after tumour cell orthotopic engraftment) using direct intratumoral injection with a microinjection pump to deliver a single dose of the virus. We observed substantial tumour regression within 1\u20132\u2009weeks and complete clearance of tumours in 20\u201324 treated mice over the subsequent 2\u20133\u2009weeks, whereas control mice died within 1\u20132\u2009weeks (Fig.","5b,c).\r\n\r\nUsing bioluminescence imaging to track tumour burden in live mice, no further regrowth of tumours was observed after the initial tumours had cleared (Extended Data Fig. 12a).","Moreover, no prominent toxicity, scored by weight loss and neurological monitoring, was seen over the subsequent 11\u2009months (the experimental end point). We re-challenged treated mice (n\u2009=\u200910 that cleared the tumour) with fresh NPE-IE cell transplants into the striatum (5\u2009months after initiation of the first therapy).","Notably, there was no detectable tumour formation in any of the mice after the re-challenge, whereas parallel control mice formed tumours within 2\u20133\u2009weeks (Extended Data Fig. 12a).","Finally, antitumour activity was dose-dependent (Extended Data Fig. 12b).\r\nIL-12 with HSV-TK\/GCV drives antitumour immunity\r\nTo determine whether IL-12 alone would be sufficient for eradicating tumours without the concomitant HSV-TK\/GCV-induced killing (that is, single versus dual payloads, respectively), we treated mice with AAV1-SSE-7-HSV-TK\u2013P2A-IL-12 alone or with this virus and with GCV. The dual payload was the most effective and enhanced the effects of IL-12 alone (Extended Data Fig. 13a). To explore safety in normal CNS, we also assessed responses in mice without tumours to either CMV-driven or SSE-7-driven HSV-TK\u2013IL12.","As expected for CMV, there were rapid neurological issues within weeks, consistent with the expected toxicities with continuous expression of IL-12 in normal CNS. By contrast, we observed no toxicity with SSE-7, a result consistent with our earlier in vitro data confirming the selectivity of this SSE (Extended Data Fig. 13b).\r\n\r\nCharacterization of the immune cell composition after treatment using either IL-12 alone or IL-12 with HSV-TK\/GCV revealed that the latter performed better, with significantly increased levels of IFN\u03b3 and a trend towards increased TNF (Extended Data Fig. 13c\u2013i).","This finding is consistent with the observed increase in cytotoxic T\u2009cells and reduction in myelosuppressive cells that occurs through T helper\u20091-like responses induced by IL-12 (Fig. 5d\u2013g).\r\n\r\nThere was also an increase in proliferative Ki67+ T\u2009cells, with a shift towards increased CD44+, a marker of T\u2009cell activation and memory (Extended Data Fig.","13d,e). To assess the percentage of transduction in the tumour mass, we used the SSE-7\u2013mCherry reporter. Tumours were collected, dissociated and assessed using flow cytometry, which revealed around 20% positive tumour cells (GFP+mCherry+) (Extended Data Fig. 13j). This transduction efficiency was achieved using the same viral doses as the survival analysis experiments. This result confirms that only a fraction of the tumour mass needs to be transduced to achieve immune clearance of residual disease.\r\n\r\nIn conclusion, AAV1-SSE-7\u2013HSV-TK\u2013IL-12 is highly effective at eliminating tumours in an aggressive syngeneic\u00a0mouse GBM\u00a0model. Treated mice were resistant to subsequent re-challenge with fresh NPE-IE cells.","A single dose of AAV1-SSE-7\u2013HSV-TK\u2013IL-12 induced complete responses with durable immune memory without off-target toxicity.\r\nDiscussion\r\nThe development of effective gene therapies for oncology will require more effective payload combinations and more selective delivery mechanisms than has been possible so far30,33,34. We showed that SSEs, combined with delivery via AAV1, can address one of the significant challenges in oncology: namely, how to specifically deliver combinations of anticancer payloads at high doses without off-target toxicity.","Across genetically diverse GSC lines and high-grade glioma tissues, SSE-7 consistently read out a core transcriptional program characteristic of GSC identity: the immature NSC-like state.\r\n\r\nOur findings indicated that SOX2 and SOX9 share crucial target genes and may represent a core transcriptional module, an \u2018oncofetal\u2019 program, that defines aggressive tumour-initiating cells in GBM18,35. Using AAV-mediated delivery and SSE control of a transgene, we achieved specific elimination of active GSCs by targeting them from within, akin to a Trojan horse approach.","The persistence of AAV vectors, and our focus on the shared SOX circuits, meant that we were able to guard against plasticity, as transgene expression will be reactivated if cells transit through this specific cell state.\r\n\r\nImmunotherapy approaches have generally had limited success in GBM36,37,38. Our AAV\u2013SSE approach will now require careful safety studies to determine the potential for translation to patients.","The goal is to deliver cytotoxic payloads alongside innate and adaptive immune activation to achieve meaningful and durable suppression of high-grade gliomas without significant off-target toxicities.\r\n\r\nFuture studies will explore the molecular mechanisms that underlie SSE activity, which may involve transcriptional hubs or condensates. Our findings from exploring the biochemical binding of SOX2 and SOX9 suggests that there is clear synergy between these two factors.","However, additional cofactors are required for the creation of a multimeric complex. Detailed structural studies are needed to precisely determine how SOX2 and SOX9 operate and the signalling TF requirements and whether nucleosome context is critical39. In the future, improved computational methods and artificial-intelligence-based tools for the design of enhancers and synthetic associated motifs will complement the functional screening approaches used here40,41,42.","In conclusion, our study demonstrated the potential of SSEs and AAV vectors to deliver precision immune activation for the treatment of cancer.\r\nMethodsDestination vector cloning\r\nWe built a specific custom destination vector for efficient Golden Gate cloning of enhancer fragments. This included the reporter gene cassette NanoLuc-Ires-mNGreen-pA, downstream of a mCMV promoter. A bacterial suicide ccdB cassette spanning the enhancer position enabled assembly of a single enhancer for selection and efficient cloning. This was based on the EMMA Golden Gate cloning system43.","For the GSC long-term differentiation experiments, the same vector was built but included PiggyBac transposase recognition sites flanking the entire cassette.\r\nBioinformatics to design a SOX2 enhancer oligonucleotide pool\r\nWe re-analysed previously published1 SOX2 ChIP\u2013seq and H3K27ac GSC cell line data to identify GSC-specific SOX2 peaks that were overlapping with H3K27ac and absent in differentiated cells (cells in serum culture). The resulting shared peaks were then combined into one set and manually curated to remove centromeres. This resulted in 1,721 peaks with an average length of 402\u2009bp. These peaks were then used to design a set of 160\u2009bp sequences.","Next, 20\u2009bp adapters were included to flank the sequences and these sequences were synthesized as an oligonucleotide pool (Twist Bio).\r\nConstruction of an arrayed plasmid library of enhancer fragments\r\nThe oligonucleotide pool (9,523 unique sequences) was first amplified for 10 cycles with 0.25\u2009\u00b5l (2.5\u2009ng input) volume, and 0.5\u2009\u00b5l of this reaction was used for the subsequent 15 cycles of amplification to reduce PCR \u2018jackpot\u2019 amplification. The final products were cloned into the expression vector using an efficient Golden Gate reaction. We used KAPA HiFi Hotstart polymerase with GC buffer (Roche, KK2501), 68\u2009\u00b0C annealing temperature and 5\u2009s of extension time at 72\u2009\u00b0C.","A total of 4,579 individual plasmids were then randomly picked, miniprepped and plated as an arrayed plasmid DNA library on 96-well plates. This was deemed a practical number for arrayed library screening. A magnetic-bead-based SPRI purification method was used to minimize loss and to clean up the DNA before Golden Gate cloning. Sanger sequencing for quality checks of a sample of the plasmids confirmed that they were diverse, and all sequences could be mapped back to the original library design.\r\nScreening platform for 384-well plates\r\nCells were seeded in 384-well plates using a multidrop and transfected the following day using CyBio Felix (CyBio).","Two days later, a NanoGlo DLR assay was used to identify hits with a fold change to mCMV of >10. These hits were Sanger sequenced to determine the specific sequence and mapped back to the genome to validate overlap with the original SOX2 ChIP\u2013seq data. The GREAT tool predicted the target gene for hit sequences44.\r\nGeneral cell culture procedures\r\nThe GSC lines GCGR-E17 (E17), GCGR-E21 (E21), GCGR-E27 (E27), GCGR-E28 (E28), GCGR-E31 (E31), GCGR-E34 (E34), GCGR-E37 (E37) and GCGR-E55 (E55) and the human NSC lines NS9FB_B (NS9), NS12ST_A (NS12) and NS17ST_A (NS17) were generated in the Pollard Laboratory and are available upon request from the Glioma Cellular Genetics Resource. Informed consent was obtained for use of patient tissue.","All procedures on patient brain tissue received ethics approval from the NHS Health Research Authority, East of Scotland Research Ethics Service (REC reference 15\/ES\/0094), and all procedures on embryonic and fetal brain tissue received ethics approval from the Lothian NHS Board, South East Scotland Research Ethics Committee (REC reference 08\/S1101\/1).\r\n\r\nCells were grown as adherent monolayers under serum-free conditions as previously described13. All cells lines tested negative for mycoplasma at passage\u00a03 using a MycoAlert Mycoplasma Detection kit (Lonza, LT07-318). Cell line authentication by STR profiling was performed as a service by the European Collection of Authenticated Cell Cultures (ECACC).","Reports are available upon request from the Glioma Cellular Genetics Resource. Analysis was conducted using the Promega Fusion system (DC2402), analysing differences at 24 distinct hypervariable genetic loci, 16 of which are used for the final STR profile and report. The ioGlutamatergic (io1001S) and ioGABAergic (ioEA1003S) neurons and ioMicroglia (ioA021) were obtained from bit.bio and grown per the manufacturer\u2019s specifications. GSC7 was previously characterized22.\r\n\r\nFor lipofection, Plus reagent and Lipofectamine LTX (Life Technologies, 15338030) were each diluted in half the volume of Opti-MEM\u00a0I reduced-serum medium (hereafter referred to as Optimem) (Life Technologies, 31985062).","Following this step, the Plus reagent and Optimem premix was added to all DNA samples followed by the Lipofectamine LTX and Optimem premix. The transfection mix was incubated for 5\u2009min at room temperature and then carefully dropped onto the cells.","Generally, no change in medium was carried out. Cells were analysed 2\u00a0days after transfection.\r\n\r\nHEK293 cells were seeded at the specified density in the respective plate format. The next day GMEM medium, DNA and polyethylenimine were mixed and incubated for 15\u2009min at room temperature to facilitate formation of the complex. The transfection mix was subsequently dropped onto the cells, and analysis was carried out 2\u2009days later.\r\nScreening assay using the Nano-Glo Dual-Luciferase reporter assay system\r\nThis assay consists of two steps. Transfection was performed using a normalization plasmid PGK-Firefly-Luciferase (Promega, E5011) transfected in a 1:10 ratio with the plasmid of interest.","Therefore, in this assay, the firefly luciferase activity is first measured, which enables normalizing of the transfection efficiency. Cells were washed 3\u2009times with PBS and 20\u2009\u00b5l was left in a 96-well plate (25\u2009\u00b5l in a 384-well plate). Oneglo buffer was added, and the plate was shaken for 5\u2009min at 480\u2009rpm to allow cell lysis.","Next, 20\u2009\u00b5l cell lysate in a 96-well format (25\u2009\u00b5l cell lysate in a 384-well format) was transferred into an opaque white 96-well format (or 384-well format) and light was measured for 0.1\u2009s per well (Ensight Multimode Plate Reader, Perkin Elmer). For the subsequent NanoLuc reaction in a 96-well format, 2\u2009\u00b5l cell lysate was transferred into an opaque white plate containing 40\u2009\u00b5l PBS, and 20\u2009\u00b5l Stopglo buffer, supplemented with substrate, was added to each well. In a 384-well format, 20\u2009\u00b5l Stopglo buffer containing substrate was added on top of the undiluted cell lysate. The plate was shaken again for 5\u2009min at 480\u2009rpm to quench the firefly luciferase reaction and to ensure good mixing.","NanoLuc activity was measured for 0.1\u2009s per well. Data obtained by the NanoLuc reaction were normalized to the Firefly reaction to account for well-to-well variability of transfection efficiency. Normalized data were used to calculate the fold change compared with the empty vector control, which contained mCMV only and no enhancer fragments.\r\nImmunocytochemistry\r\nCells were washed twice with PBS and fixed in 4% paraformaldehyde (PFA) for 10\u2009min. Cells were permeabilized with 0.1% Triton-100 in PBS (PBST). Cells were blocked with blocking solution (1% BSA in PBST with 3% goat serum) and incubated with the primary antibody at 4\u2009\u00b0C overnight.","The next day, cells were washed 3\u2009times in PBST, and the respective secondary antibody was applied in blocking solution and incubated for 45\u201360\u2009min at room temperature. Cells were washed with PBS and incubated for 5\u2009min with a DAPI nuclear counterstain at 1\u2009\u00b5g\u2009ml\u20131 final concentration. Images were acquired using a Nikon TiE microscope and NIS elements software (Nikon).\r\nFlow cytometry\r\nFor flow cytometry, cells were detached, pelleted and resuspended in an appropriate volume of flow cytometry buffer (1% BSA in PBS, v\/v) or PBS.","Cells were stained with Draq7 (Abcam, ab109202, final concentration of 0.1\u2009\u00b5M) or DAPI (ThermoFisher Scientific, D3571) as a live\/dead stain and analysed using a BD LSRFortessa cell analyser (4 lasers, BD Bioscience). Analysis of flow cytometry data was carried out using FlowJo Analysis software (v.10.6.2) or FCS Express 7 flow cytometry software (v.7.22.x).\r\nRNA isolation, cDNA synthesis and RT\u2013qPCR\r\nRNA was extracted using a MasterPure RNA Purification kit (Epicentre). RNA was stored at \u221280\u2009\u00b0C, and concentration was determined using a NanoDrop spectrophotometer. cDNA synthesis was carried out using a SuperScript III Reverse Transcriptase kit (Life Technologies) according to the manufacturer\u2019s instructions.","Around 200\u2013500\u2009ng RNA was used for the cDNA reaction. The same amount of RNA was used for each experiment. After the reaction, cDNA was diluted to the required volume using nuclease-free water.\r\n\r\nFor RT\u2013qPCR, TaqMan Universal PCR master mix (Applied Biosystems) and TaqMan gene expression assays (Life Technologies) were used on a Quant Studio7 Flex Real-Time qPCR machine. RNA samples that did not undergo reverse transcription (to assess DNA contamination) and water controls were used on every plate. RT\u2013qPCR was carried out in technical duplicates. Data analysis was performed using the ddCt method, which assumes 100% PCR efficiency and is guaranteed with TaqMan assays.","In brief, the mean was calculated for technical replicates and normalized to the housekeeping gene GAPDH, which produces the dCt value.\r\nWestern immunoblotting\r\nCell lysates were prepared by resuspending cells in RIPA buffer (50\u2009mM HEPES pH\u20097.7, 150\u2009mM NaCl, 1% NP-40, 0.5% DOC and 0.1% SDS), incubating on ice for 5\u2009min, centrifuging at maximum speed for 10\u2009min in a tabletop centrifuge (5415D, Eppendorf) and collecting the supernatant (lysate). Protein extracts were quantified using a Pierce BCA Protein Assay kit (Thermo Scientific, 23225). SDS\u2013PAGE was performed using homemade 4\u201312% Bis-Tris gels.","Gels were transferred onto PVDF membranes (Millipore, IPVH00010), previously activated in methanol, by wet electroblotting or semi-dry blotting using a Bio-Rad Trans-blot turbo system. Western blots were revealed using HRP-conjugated antibodies, homemade ECL solutions or Clarity Western ECL (Bio-Rad), and imaged using X-ray films or a Bio-Rad ChemiDoc Imaging system.\r\nPrecipitation of interacting proteins\r\nPCR amplification with biotinylated primers and purification were used to generate templates that could pull down bound proteins.","Enhancer fragments were amplified with 5\u2032 biotinylated primers using PrimeStar Max (Takara) according to the manufacturer\u2019s instructions.The following primers were used: forward primer 595: TGATCCGTCTCGCCCTACTAGGTTACTGGTGCATGC; reverse primer 596: ACTAACGTCTCGGAGCACCCAAACTATTGGAGCGAG\u00a0(bold sequences, adaptors).\r\n\r\nPCR purification using Agencourt AMPure XP magnetic beads (Beckman Coulter) was carried out according to the manufacturer\u2019s instructions. PCR product quantification using Tapestation and reagents was carried out according to the manufacturer\u2019s instructions.\r\n\r\nAll the buffers were prepared the day before, passed through a 22\u2009\u00b5m filter (except the dialysis buffer) and left at 4\u2009\u00b0C overnight.","DTT and protease inhibitors were added immediately before use. Cell pellets were resuspended in 5\u2009ml per 40\u00a0million cells of ice-cold buffer\u2009A (10\u2009mM HEPES pH\u00a07.9, 1.5\u2009mM MgCl2, 10\u2009mM KCl, 0.5\u2009mM DTT and protease inhibitors (complete, Roche, 11697498001)) and incubated on ice for 10\u2009min. The cell suspension was transferred to a glass Dounce homogenizer and dounced 40 times on ice, transferred to a Falcon tube and centrifuged at 1,350\u2009rcf for 10\u2009min at 4\u2009\u00b0C.","The supernatant was discarded (cytosolic extract) and the pellet was resuspended in 100\u2009\u00b5l per 10\u2009million cells in ice-cold buffer\u2009B (20\u2009mM HEPES pH\u20097.9, 5% glycerol, 1\u2009M NaCl, 1.5\u2009mM MgCl2, 0.2\u2009mM EDTA pH\u20098.0, 0.5\u2009mM DTT and protease inhibitors (complete, Roche, 11697498001)). The suspension was rotated for 30\u2009min at 4\u2009\u00b0C and transferred into a dialysis membrane (SnakeSkin, Thermo Scientific, 68100). Dialysis was carried out in 500\u2009ml dialysis buffer (20\u2009mM HEPES pH\u20097.9, 5% glycerol, 100\u2009mM KCl, 0.83\u2009mM EDTA pH\u20098.0, 1.66\u2009mM DTT and protease inhibitors (complete, Roche, 11697498001)) for 2\u2009h at 4\u2009\u00b0C with rotation.","The dialysis buffer was changed (500\u2009ml), and dialysis was continued overnight at 4\u2009\u00b0C. The extract was collected in a 1.5\u2009ml tube and spun at maximum speed for 15\u2009min at 4\u2009\u00b0C (centrifuge 5415D Eppendorf). The supernatant was collected into new 1.5\u2009ml tubes and quantified (Pierce BCA Protein Assay kit, Thermo Scientific, 23225).","Next, 100\u2009\u00b5g aliquots were flash-frozen in liquid nitrogen and stored at \u221280\u2009\u00b0C. The following primary antibodies were used for western blotting: anti-SOX2 (rabbit) 1:1,000 Abcam (ab92494); anti-SOX9 (rabbit) 1:500 Millipore (ab5535); and anti-GAPDH (mouse) 1:10,000 Ambion (AM4300). The secondary antibodies used for western blotting were anti-rabbit 1:5,000 Novex (A16110) and anti-mouse 1:5,000 Novex (A16027).\r\n\r\nStreptavidin magnetic beads (NEB, S1420S) were used according to the manufacturer\u2019s instructions.","In brief, 10\u2009\u00b5l beads were aliquoted into low-bind tubes and washed 3\u2009times with binding buffer (20\u2009mM Tris-HCl pH\u20097.5, 0.5\u2009M NaCl and 1\u2009mM EDTA) on a magnetic stand.","Next, 20\u2009\u00b5l biotinylated DNA (around 20\u2009pmol) was mixed with 200\u2009\u00b5l binding buffer and was added to the beads. The suspension was rotated at room temperature for 2\u2009h. The beads were washed 3\u2009times with binding buffer, and 50\u2009\u00b5g nuclear extract was added (total volume of 50\u2009\u00b5l).","Beads, DNA and nuclear extract, to which 150\u2009\u00b5l dialysis buffer (20\u2009mM HEPES pH\u20097.9, 5% glycerol, 100\u2009mM KCl, 0.83\u2009mM EDTA pH\u20098.0, 1.66\u2009mM DTT and protease inhibitors (complete, Roche, 11697498001)) was added, were rotated overnight at 4\u2009\u00b0C. The following morning, beads were washed 3\u2009times with wash buffer (20\u2009mM HEPES pH\u20097.9, 5% glycerol, 250\u2009mM NaCl, 0.83\u2009mM EDTA pH\u20098.0 and 1.66\u2009mM DTT). For western blotting, proteins were eluted in 20\u2009\u00b5l loading buffer (lithium dodecyl sulfate buffer containing 50\u2009mM DTT) and boiled for 5\u2009min to denature the proteins.","For mass spectrometry, the beads were sent dry to the Institute for Genetics and Cancer core facility (University of Edinburgh) for analyses.\r\nSOX2 and SOX9 ChIP\u2013seq library preparation and analysis\r\nIn brief, GSCs cultured in 150\u2009mm dishes to 70% confluence were crosslinked in 1% formaldehyde for 10\u2009min at room temperature. Excess formaldehyde was quenched by incubating cells with 0.25\u2009M glycine for 5\u2009min at room temperature. The cell pellet was washed twice with ice-cold PBS before storage at \u221280\u2009\u00b0C. For each GSC line, cell pellets from eight 150\u2009mm dishes were combined before proceeding for cell lysis and sonication.","The cell pellet was lysed in buffer LB3 (10\u2009mM Tris-HCl pH\u20098.0, 100\u2009mM NaCl, 1\u2009mM EDTA pH\u20098.0, 0.5\u2009mM EGTA pH\u20098.0, 0.1% sodium deoxycholate, 0.5% N-lauroylsarcosine and 0.1% SDS) and chromatin was fragmented to 200\u2013600\u2009bp size on a Covaris M220 sonicator (total of 4 cycles, with the following treatment settings: 600\u2009s per cycle, peak power 75, duty factor 10, cycles\/burst 200; temperature: minimum 5\u00a0\u00b0C, set point 7\u00a0\u00b0C, maximum 9\u00a0\u00b0C). For each ChIP assay, 15\u201320\u2009\u00b5g fragmented chromatin was incubated overnight in a cold room with 10\u2009\u03bcg antibodies (R&D SOX2 AF2018, Millipore, SOX9 AB5535) and 30\u2009\u03bcl Protein G Dynabeads (10003D).","The next day, magnetic beads were washed 5 times with RIPA (50\u2009mM HEPES-KOH pH\u20097.5, 0.5\u2009M LiCl, 1\u2009mM EDTA pH\u20098.0, 1% NP-40 and 0.7% sodium deoxycholate), once with TE NaCl (10\u2009mM Tris-HCl pH\u20098.0, 1\u2009mM EDTA pH\u20098.0 and 50\u2009mM NaCl) and finally eluted in 200\u2009\u03bcl ChIP elution buffer (50\u2009mM Tris-HCl pH\u20098.0, 10\u2009mM EDTA pH\u20098.0 and 1% SDS). Immunoprecipitated chromatin was reverse crosslinked by incubating samples at 65\u2009\u00b0C for 10\u2009h and extracted using the phenol\u2013chloroform\u2013IAA (11896714) method. To account for intertumoral heterogeneity, ChIP\u2013seq was performed on seven independent primary GSC lines.","Libraries were prepared for two technical ChIP replicates and one input control using a NEBNext Ultra II DNA Library Prep kit for Illumina (E7645S). The libraries were barcoded using NEBNext Multiplex Oligos for Illumina (Dual Index Primers Set 1 kit, E7600S) and sequenced on an Illumina Novaseq (50\u2009bp read length).\r\n\r\nPaired-end reads were aligned to the hg38 genome using BWA, filtering out poor-quality (MAPQ\u2009<\u200910), duplicates, mitochondrial genome and blacklisted regions (https:\/\/www.encodeproject.org\/annotations\/ENCSR636HFF\/). Technical replicates were merged, and peaks were called using MACS2 with default settings.","Consensus SOX2 and SOX9 peak sets were derived by taking the overlap of peaks occurring in 5 out of 7 GSCs for each TF set. The overlap significance between the consensus SOX2 and SOX9 peak sets was determined using a circular permutation test in regioneR (ntimes\u2009=\u200910,000). Genomic regions near SOX2 and SOX9 peaks were annotated using HOMER45.\r\n\r\nTo perform overlapping analysis with GSC super-enhancers, we downloaded raw fastq files from publicly available H3K27ac datasets on GSCs (GSE119834, GSE74529, GSE121601 and GSE92458) and called super-enhancers using ROSE with default settings. A consensus set of GSC super-enhancers was derived by considering super-enhancers that occurred in at least two GSCs.","The overlap significance between the consensus co-bound SOX2\u2013SOX9 sites and consensus GSC super-enhancers was performed using a circular permutation test as described above. Gene ontology term association analysis was performed using GREAT. To find centrally enriched de novo motifs at SOX2 and SOX9 peaks in co-bound SOX2\u2013SOX9 enhancers, and to identify the spacing between the most significant de novo motifs from each peak set, we used CentriMo and SpaMo from the MEME-ChIP suite of tools46.\r\nscRNA-seq sample preparation\r\nOn the day of cell seeding, cells were detached as described above. Cells were counted using a haemocytometer and seeded into a 6-well Corning plate at 400,000 cells per well in 2\u2009ml.","Plates were incubated at 37\u2009\u00b0C with 5% CO2. The next day, AAV vector stocks were thawed at room temperature. The appropriate amount of virus stock was added to the culture medium to achieve a final multiplicity of infection (MOI) of 1,000,000, which ensured high transduction efficiency. Viral-containing medium was added to cells without replacing the existing medium. The cells were then returned to the incubator at 37\u2009\u00b0C with 5% CO2.\r\n\r\nAfter 3\u2009days of incubation with the viral particles, cells were detached as described above.","Next, 10% of the cells were prepared for flow cytometry using DAPI as a live\/dead stain (see\u00a0Supplementary Information for the gating strategy). The remaining 90% of the cells were processed for single-cell transcriptome sequencing.\r\n\r\nCells were fixed using a Parse Evercode v.2 Cell Fixation kit following the manufacturer\u2019s instructions, with an average of around 400,000 cells per sample. Fresh reagents were prepared for each fixation, and 7.5% Gibco BSA fraction\u2009V was added per the manufacturer\u2019s recommendations.","In brief, cells were washed with PBS and resuspended in cold Cell Prefixation buffer. Cells were filtered through a 40\u2009\u00b5m strainer after 5\u2009min of centrifugation at 300g at 4\u2009\u00b0C. Following this step, Cell Fixation solution and Cell Permeabilization solution were subsequently added, with incubation on ice for 10\u2009min and 3\u2009min, respectively. Cell Neutralization buffer was then added, and cells were centrifuged again at 300g for 5\u2009min at 4\u2009\u00b0C.","Finally, cells were resuspended in a volume of 50\u201390\u2009\u00b5l Cell Buffer containing 1% DMSO based on the live cell number before fixation. Cells were filtered through a 40\u2009\u00b5m strainer and stored at \u221280\u2009\u00b0C for up to 3\u2009months.\r\nscRNA-seq library preparation\r\nLibraries were generated using a Parse Evercode WT Mega v.2 kit according to the manufacturer\u2019s instructions.","In brief, 72 samples were loaded into a 96-well RT barcoding plate for reverse transcription, with each well containing a unique barcoded RT primer. Following cDNA synthesis, cells were pooled and subjected to two additional rounds of barcoding, achieving three rounds of barcoding. Barcoded cDNA was then amplified via template switching and pooled into 15 sublibraries, each containing around 62,500 cells. A fourth barcode was incorporated during PCR amplification of the sequencing libraries. Sublibraries were then fragmented, ligated with Illumina adapters and purified using Ampure XP beads. Library quality was assessed on an Agilent BioAnalyzer 2100 before sequencing.","Sublibraries, with 5% PhiX spike-in, were sequenced on an Illumina NovaSeq X using 300 cycles kits as paired-end, dual-index reads.\r\nscRNA-seq analysis\r\nSequencing reads were aligned to the human (hg38) genome, and UMI counting data were generated following the standard Parse Bioscience pipeline, \u2018splitpipe (v.1.1.2)\u2019, with default parameters. Three custom genes were added to the genome (mCherry, HSV-TK and bGHployA) to capture \u2018activated\u2019 SSE-7 cells.","In total, 345,495 cells were included in the initial dataset. Low-quality cells were filtered out based on the following thresholds: (1) the percentage of mitochondrial gene was >20%; (2) the number of genes was <300; (3) the number of uniquely aligned reads was <500; or (4) the number of uniquely aligned reads was >20,000. Genes detected in fewer than five cells were filtered out as low-quality genes. Potential doublets were identified using the R package \u2018DoubletFinder\u2019 (v.2.0.4), with an expected doublets rate of 3%, as guided by Parse. Doublets and clusters containing more than 20% doublets were removed. Ambient RNA-contaminated cells (cell score\u2009>\u20090.2) were removed using \u2018decontX\u2019 from the R package \u2018celda\u2019 (v.1.18.2).","After filtration, 270,842 cells remained for further analysis.\r\n\r\nThe data were normalized using \u2018LogNormalize\u2019, and principal component analysis was performed based on the top 2,000 variant genes using the R package \u2018Seurat\u2019 (v.5.0.3). The first 15 principal components were used as input for Louvain-based graphing. SSE-7-activated cells (2,711 cells) were labelled according to the following criteria: mCherry\u00a0>\u00a01 or HSV-TK\u2009>\u20091 or bGHpolyA\u2009>\u20090. Cells with a single count of mCherry or HSV-TK from the virus-treated group were labelled as false negative (9,583 cells), whereas the remaining cells were labelled as SSE-7 non-activated cells (125,525 cells).","Meanwhile, cells expressing mCherry or HSV-TK in the control group were labelled as false positive (1,998 cells). Transcriptional subtypes were predicted on the basis of a previously described gene signature27 using the function \u2018sigScores\u2019 and \u2018as_four_state_gbm\u2019 from the R package \u2018scalop\u2019 (v.1.1.0). Cell cycle was predicted using \u2018CellCycleScoring\u2019 from the R package \u2018Seurat\u2019.\r\nSCENIC analysis of scRNA-seq data\r\nFalse-negative and false-positive cells were removed, and the dataset was divided into 15 libraries as per the sequencing library preparation. The dataset was then analysed following the standard \u2018pyscenic (v.0.12.1)\u2019 pipeline.","Auxiliary human datasets (hg38 mc_v10_clust) were downloaded from the cisTarget resources website (https:\/\/resources.aertslab.org\/cistarget\/). Motif\u2013TF annotation was based on 10\u00a0kbp upstream and downstream around the transcription start sit (20\u00a0kbp in total). SSE-7-enriched TF modules were selected on the basis of the following criteria: (1) Z\u2009score\u2009>\u20090.7 and (2) Z\u2009score in the SSE-7-activated group was higher than in other groups, as determined by repeated-measures ANOVA test.","Enrichment for candidate TF modules across 15 libraries was as follows: SOX8 (found in 9 libraries), STAT1 (found in 15 libraries), IRF9 (found in 12 libraries), ETS1 (found in 13 libraries), MAF (found in 3 libraries), SMAD1 (found in 12 libraries), FOS (found in 11 libraries), HIF1A (found in 7 libraries), SOX9 (found in 10 libraries), STAT3 (found in 3 libraries), NEF2L2 (found in 13 libraries), JUN (found in 2 libraries) and SOX2 (found in 2 libraries).\r\nLentivirus production and titration\r\nTo produce SOX2, SOX9 and rtTA2M2 individual lentiviral supernatant samples, nearly 2.4\u2009million HEK293T cells per 15\u2009cm plate were cultured for 24\u2009h in GMEM medium supplemented with 10% fetal calf serum (FCS), 1\u2009mM sodium pyruvate, 1\u2009mM glutamine and non-essential amino acids (HEK medium) at 37\u2009\u00b0C and 5% CO2.","At 24\u2009h after seeding, HEK293T cells were transfected with a plasmid cocktail containing 7.5\u2009\u03bcg expression plasmid, 5.1\u2009\u03bcg psPAX2 packaging vector, 2.4\u2009\u03bcg envelope vector mixed in 45\u2009\u03bcl Fugene6 (Roche) and 855\u2009\u03bcl Opti-MEM medium (Invitrogen, Thermo Fisher Scientific). The cells were further cultured for 16\u2009h at 37\u2009\u00b0C and 5% CO2 before the medium was changed for HEK medium. At 65\u2009h after transfection, virus-containing supernatant was collected and cleared by centrifugation and filtered through a 0.45\u2009\u03bcm syringe filter (Merck Millipore).","Virus was concentrated (roughly 100-fold) by transferring to thin-walled ultracentrifuge tubes and pelleted by ultracentrifugation at 25,000\u2009rpm for 2.5\u2009h using a SW32-TI rotor in a Beckman Optima XPN ultracentrifuge (Beckman Coulter). The pelleted virus was resuspended in 300\u2009\u03bcl plain GMEM medium and incubated at 4\u2009\u00b0C for 16\u2009h before aliquoting and flash-freezing in liquid nitrogen for storage at \u221280\u2009\u00b0C.","The viral titre, calculated by flow cytometry, was 6\u2009\u00d7\u2009107 infectious units per ml.\r\nOverexpression of SOX2 and SOX9\r\nEarly passage (passage\u20094) human fibroblasts (hFibs) were seeded, 1.5\u2009million per 15\u2009cm plate, and cultured in hFib medium (GMEM medium supplemented with 10% FCS, 1\u2009mM sodium pyruvate, 1\u2009mM glutamine, MEM 1\u00d7 non-essential amino acids solution, 50\u2009\u03bcM 2-mercaptoethanol (Gibco, Thermo Fisher Scientific) and 2\u2009ml penicillin\u2013streptomycin (Invitrogen, 15140122)) 24\u2009h before transduction at 37\u2009\u00b0C and 5% CO2.","The cells were transduced by exchanging the growth culture medium with lentiviral medium, without antibiotics, containing 8\u2009\u03bcg\u2009ml\u20131 polybrene, and 65\u2009\u03bcl concentrated rtTA2M2 and either SOX2 or SOX9 virus, for single factor overexpression, or 75\u2009\u03bcl rtTA2M2, SOX2 and SOX9 viruses for double factor overexpression. At 48\u2009h after transduction, the cells were split 1:2 and cultured for 24\u2009h, after which the culture medium was replaced with induction medium containing 1\u2009\u03bcg\u2009ml\u20131 doxycycline (induction day\u20090). After 48\u2009h of induction, the cells were collected for lysis and nuclear protein extraction.\r\nNuclear extraction of GSC7 and hFib cells\r\nGSC7 cells were cultured as adherent monolayers13.","For differentiation, cells were cultured without the supplements EGF and FGF and in 5% serum for 2\u2009weeks. GSC7 and hFib nuclear lysates were prepared in the same way; cells were resuspended to 1\u00a0million cells per 100\u2009\u03bcl 10\u2009mM HEPES pH\u20097\u20138, 1.5\u2009mM MgCl2, 10\u2009mM KCl, 0.5\u2009mM DTT, supplemented with complete Ultra EDTA-free protease inhibitors (Roche, 05892970001) and homogenized using a dounce and tight-fit pestle with 1 stroke per 1\u2009million cells. Samples were centrifuged at 1,350\u2009rcf for 5\u2009min at 4\u2009\u00b0C.","The nucleus pellet was resuspended to 10\u2009\u03bcl per 1\u2009million starting cells in 20\u2009mM HEPES pH\u20097\u20138, 30% glycerol, 420\u2009mM NaCl, 1.5\u2009mM MgCl2, 0.2\u2009mM EDTA, 0.5\u2009mM DTT and complete Ultra EDTA-free protease inhibitors (Roche, 05892970001). Samples were dialysed for 4\u2009h into a minimum of 1\u20131,000 volume excess of 20\u2009mM HEPES pH\u20097\u20138, 30% glycerol, 100\u2009mM NaCl, 0.83\u2009mM EDTA, 1.66\u2009mM DTT, 0.2\u2009mM PMSF and complete Ultra EDTA-free protease inhibitors (Roche 05892953001) using Slide-A-lyzer mini dialysis units with 3,500 MWCO cassettes (Thermo).","Samples were divided into single-use aliquots (11\u2009\u03bcl) and flash-frozen in liquid nitrogen for storage at \u221280\u2009\u00b0C.\r\nWestern blot analysis of overexpression\r\nThe protein concentrations of the lysates were quantified using a Pierce BCA Protein Assay kit according to the manufacturer\u2019s 96-well format instructions (Thermo Fisher Scientific). The following overexpression (OE) values were used: SOX2 OE\u2009=\u200910.54\u2009mg\u2009ml\u20131, SOX9 OE\u2009=\u20095.56\u2009mg\u2009ml\u20131, SOX2 and SOX9 OE\u2009=\u20097.56\u2009mg\u2009ml\u20131, GSC7 GSCs\u2009=\u20091.02\u2009mg\u2009ml\u20131 and GSC7 differentiated cells\u2009=\u20090.43\u2009mg\u2009ml\u20131.","Verification of SOX protein overexpression, relative to untransfected hFib lysate, was resolved by SDS\u2013PAGE and electroblotting onto a PVDF membrane for western blotting. The primary antibody incubations with goat anti-human SOX2 antibody (1:363, AF2018, Abcam) and rabbit anti-mouse SOX9 (1:2,000 AB5535, Chemicon) were performed at 4\u2009\u00b0C for 16\u2009h. The secondary antibody incubations with donkey anti-goat IgG-HRP (1:2,000 dilution of ab97110, Abcam) and goat anti-rabbit (1:2,000 of 32460, Invitrogen) were performed for 1\u2009h at room temperature.","Blots were visualized using SuperSignal West Pico Chemiluminescent substrate (Thermo Fisher Scientific) and HRP was visualized using a Bio-Rad ChemiDoc MP imaging system.\r\nRecombinant SOX protein production\r\nHis-tagged human SOX2 and SOX9 were expressed from a pET28a backbone (cloned by G.\u00a0Roberts in the Soufi Laboratory) using Rosetta 2(DE3) pLys competent cells (Novagen).\r\n\r\nSOX2 was expressed as previously described47. Cells in LB medium (1% Bacto-tryptone, 0.5% yeast extract and 1% NaCl), supplemented with 30\u2009\u00b5g\u2009ml\u20131 kanamycin and chloramphenicol, were grown at 37\u2009\u00b0C overnight.","SOX9 was expressed in 2\u00d7\u00a0TY medium (1.6% Bacto-tryptone, 1% yeast extract and 0.5% NaCl), supplemented with 30\u2009\u00b5g\u2009ml\u20131 kanamycin and chloramphenicol, at 16\u2009\u00b0C overnight. After expression, bacteria cell pellets were lysed, 100\u2009ml buffer per 1\u2009l of culture, in GuHCl Denaturing buffer (6\u2009M guanidine-HCl, 50\u2009mM Tris-HCl (pH\u20098) and 500\u2009mM NaCl, 5% (v\/v) glycerol) overnight. Lysate was sonicated for 200\u2013300\u2009s at 25\u201330\u2009microns using a Soniprep 150 with intermittent chilling on ice. The supernatant was clarified by centrifugation at 18,000\u2009rpm in a FLA21-8x50y rotor for 30\u2009min.","The lysate was further sonicated for 120\u2009s, as above, then filtered through a 0.45\u2009\u00b5m filter before loading into a HisTrapHP column (Cytivia) for affinity purification of His-tagged proteins using an AKTA pure system. The column was equilibrated with 5\u00a0CV of denaturing 10\u00a0mM imidazole buffer (6\u2009M urea, 500\u2009mM NaCl, 50\u2009mM Tris-HCl (pH\u20098), 10\u2009mM imidazole and 5% (v\/v) glycerol), after sample loading was washed with 20\u2009CV of denaturing 50\u00a0mM imidazole buffer (6\u2009M urea, 500\u2009mM NaCl, 50\u2009mM Tris-HCl (pH\u00a08), 30\u2009mM imidazole and 10% (v\/v) glycerol), and bound proteins were eluted with denaturing 300\u2009mM imidazole (6\u2009M urea, 500\u2009mM NaCl, 50\u2009mM Tris-HCl (pH\u20098), 300\u2009mM imidazole and 10% glycerol).","Peak fractions of the elution were analysed by SDS\u2013PAGE, and samples with appropriately sized bands (40\u2009kDa for His\u2013SOX2, and 60\u2009kDa for His\u2013SOX9) were combined and desalted into 2\u2009M urea buffer using a HiPrep 26\/10 desalting column (GE Healthcare). The SOX2 desalting buffer comprised 50\u2009mM Tris-HCl (pH\u00a08), 240\u2009mM NaCl, 10\u2009mM KCl, 2\u2009mM MgCl2, 2\u2009mM CaCl2, 0.8\u2009M urea, 30% (v\/v) glycerol, 0.1% NP40 substitute, 0.05% Triton-X-100, 2\u2009mM EDTA and 5\u2009mM DTT resuspended in 1\u00d7 PBS. The SOX9 desalting buffer comprised 50\u2009mM HEPES NaOH (pH\u20097.5), 240\u2009mM NaCl, 10\u2009mM KCl, 2\u2009mM MgCl2, 2\u2009mM CaCl2, 2\u2009M urea, 30% (v\/v) glycerol, 0.1% NP40 substitute, 0.05% Triton-X-100, 2\u2009mM EDTA and 5\u2009mM DTT.","The fractions corresponding to a dip in the absorbance at A280 were collected and visualized by SDS\u2013PAGE before combining. Protein concentration was determined by SDS\u2013PAGE, comparing to rAlbumin standards. The gels were imaged by detecting Coomassie blue staining using a Bio-Rad ChemiDoc MP imaging system. The resulting images were visualized, and the bands quantified using ImageJ 2.8\/1.54i.\r\nSmall-molecule inhibitor screening\r\nE55 cells were screened using inhibitors included in the Kinase Screening Library (Cayman, 10505) as well as two YAP\u2013TEAD pathway inhibitors, MGH-CP1 (HY-139330) and TED-347 (HY-125269).","Around 18,000 cells were seeded per well in a 96-well plate and transduced with the corresponding AAVs at a MOI of 2,000,000 on the next day. Cells were incubated at 37\u2009\u00b0C, 5% CO2. In the primary screen, the cells were treated with a single concentration of 3\u2009\u03bcM per inhibitor at 24\u2009h after transduction. After 48\u2009h of inhibitor treatment at 37\u2009\u00b0C and 5% CO2, the cells were detached and analysed using an Agilent NovoCyte Penteon five laser flow cytometer and FCS Express 7 Research (v.7.24.0030). The results were expressed as the fold change in mCherry median fluorescence intensity in the inhibitor-treated samples relative to the DMSO control.","After performing the first round of screening, inhibitors that gave more than 20% changes were selected for the second round of screening.\r\n\r\nIn the second-round screen, the cells were seeded and transduced with AAVs using the same procedures as the primary screen. At 24\u2009h after transduction, the cells were treated with 3 different doses (0.1, 1 and 3\u2009\u03bcM), and incubated at 37\u2009\u00b0C and 5% CO2 for 48\u2009h. Cells were then detached and analysed using a Agilent NovoCyte Penteon five laser flow cytometer. Draq7 was used as the live\/dead stain in both rounds of screening.","The results of the second round screen were expressed as the fold change in mCherry MFI and as the percentages of mCherry-positive cells, each normalized to the DMSO control.\r\nEMSAs\r\nCy5-labelled SSE-7 bait for EMSAs was generated by PCR amplification of SSE-7 from the plasmid using Cy5-labelled primers (Sigma): 5\u2032-ACTAGGTTACTGGTGCATGCTTGTCCTGCCTTTGAGAACA-3\u2032 and 5\u2032-ACCCAAACTATTGGAGCGAGAGAAAGGAAAGAAAGAGGTC-3\u2032.","The following primers were used for Cy5-labelled fragments for EMSA: ID4310: 5\u2032-ACTAGGTTACTGGTGCATGCTTGTCCTGCCTTTGAGAACA-3\u2032 and 5\u2032-ACCCAAACTATTGGAGCGAGCTCATTTGAAGCAGAAGAAT-3\u2032; ID4428: 5\u2032-ACTAGGTTACTGGTGCATGCTCTAAATGGAGATCCTCCCA-3\u2032 and 5\u2032-ACCCAAACTATTGGAGCGAGACAGGAGGAAGTAGTAAATC-3\u2032; ID0785: 5\u2032-ACTAGGTTACTGGTGCATGCAAGAATGAACTGGGCCCAGC-3\u2032 and 5\u2032-ACCCAAACTATTGGAGCGAGTGAAAACCAGAGGCATCTCA-3\u2032; and ID3836: 5\u2032-ACTAGGTTACTGGTGCATGCTACCCTTCCAGGGGAGCAGT-3\u2032 and 5\u2032-ACCCAAACTATTGGAGCGAGAGAAAGGAAAGAAAGAGGTC-3\u2032.\r\n\r\nThe binding to Cy5-end-labelled SSE-7 was analysed in native 1% agarose gels (12\u00a0\u00d7\u00a013\u00a0\u00d7\u00a01\u00a0cm), which were prepared in 0.5\u00d7 TBE (45\u2009mM Tris-borate and 1\u2009mM EDTA).","Gels were stored overnight at 4\u2009\u00b0C before pre-running at 120\u2009V (approximately 10\u2009V\u2009cm\u20131) for 1\u2009h. For affinity analysis, a 10\u2009\u03bcl mixture typically containing 50\u2009nM Cy5-labelled full-length or fragmented SSE-7, 100\u2009ng\u2009\u00b5l\u20131 poly(dA:dT) (InvivoGen, tlrl-patn) and 0, 0.5, 1, 2 or 4\u2009\u03bcl of hFib lysates, with and without SOX overexpression (5.56\u2009mg\u2009ml for OE lysates and 3.47\u2009mg\u2009ml\u20131 for hFib control lysates), GSC7 GBM stem, or differentiated, cell lysates (at 0.43\u2009mg\u2009ml\u20131), or purified recombinant SOX protein (at 3.17\u2009\u00b5M). For SOX2 and SOX9 in combination, equimolar amounts of SOX2 and SOX9 were mixed for 0.5, 1, 2, 4 and 8\u2009\u00b5l of protein in total.","Next, 8\u2009\u00b5l SOX protein was added to 25\u2009nM SSE-7 DNA in a 20\u2009\u00b5l volume, and 20\u2009\u00b5l of sample was loaded for gels, which were prepared in 1\u00d7 binding buffer (10\u2009mM Tris HCl pH\u20097.5, 1\u2009mM MgCl2, 10\u2009\u03bcM ZnCl2, 10\u2009mM KCl, 1\u2009mM DTT, 5% (v\/v) glycerol and 0.5\u2009mg\u2009ml\u20131 BSA). The mixtures were incubated at 20\u2009\u00b1\u20091\u2009\u00b0C in the dark for 1\u2009h using DNA LoBind tubes (Eppendorf). The entirety of each sample was then loaded onto agarose gels and electrophoresis was performed at 120\u2009V for 3\u2009h at 4\u2009\u00b0C. The gels were imaged by detecting Cy5 fluorescence using a Bio-Rad ChemiDoc MP imaging system.","The resulting images were visualized, and the bands were quantified using ImageJ 2.8\/1.54i.\r\nAdult human cortex and brain tumour slice cultures\r\nTumour and non-tumour brain tissue were transferred to a 10-cm2 tissue culture dish with sterile PBS and placed on ice. Visibly damaged tissue was removed and the remaining tissue was transferred into a 35-mm2 dish with pre-warmed 3% SeaPlaque agarose (50100, Lonza). After cooling in ice, the block was removed and cut using a scalpel into an approximately 2\u2009cm cube around the brain. Before starting to cut, a 6-well plate was prepared.","In each well, we introduced one cell culture insert (PICMORG50, Millicell) and added below it 1\u2009ml culture medium in basal NSC medium, Dulbecco\u2019s modified Eagle medium: F12 supplemented with EGF and FGF (Life Technologies). The embedded brain was placed in a circular vibratome plate with glue. The vibratome (VT1000 S, Leica) plate was fixed in the platform and filled with PBS and penicillin\u2013streptomycin (15140-122, Gibco, 1:100).","Then, 300-\u00b5m-thick slices were cut, with vibrating frequency set at 10 and speed at 1. Each slice was transferred using a small brush onto the top of a Millipore culture insert. The platform was maintained cool at all times. The 6-well plate was placed in an incubator at 37\u2009\u00b0C with 5% CO2.\r\nAddition of virus\r\nA 5\u2009\u00b5l volume of virus was pipetted onto the centre of each tissue slice, using a 10\u2009\u00b5l tip without touching the tissue. Three repeat doses of 5\u2009\u00b5l virus was added to each slice at 5\u2009min intervals (20\u2009\u00b5l in total per slice). Slices were incubated for 7\u2009days.","The medium was replaced on day\u00a03 or 4.\r\nImmunostaining\r\nThe medium was removed and exchanged for 1\u2009ml freshly prepared 4% PFA; 1\u20132\u2009ml PFA was also placed gently on top to cover the slice. After 2\u2009h, PFA was removed, brain slices were washed 3\u2009times with PBS and transferred using a brush to a 24-well plate. Slices were incubated at room temperature for 1.5\u2009h in blocking solution (0.5% Triton X-100 and 3% goat serum; Sigma-Aldrich) followed by incubation with primary antibody for 2\u2009days at 4\u2009\u00b0C. Sections were washed five times with PBS, the respective secondary antibody was applied in blocking solution overnight.","The next day, slices were washed 5 times with PBS mounted and cleared in RapiClear 1.49 (RC149001, Sunjin Lab).\r\n\r\nImages were acquired using an Opera Phenix Plus high-content imaging system (Revvity) equipped with a \u00d740\/1.1 NA water-immersion objective. Image analysis was performed in Harmony software (v.5.2; Revvity). For the full pipeline see Supplementary Information.","In brief, nuclei were segmented using DAPI with the common thresholding method, and the nuclear mean intensity of channels SOX2 (488) and mCherry (555) were then calculated. Thresholds to define positive cells were set for each tissue slice individually based on 3\u20135 regions of interest. Nestin was segmented using \u2018find image region\u2019 with channel 647, and positive cells were defined as those with nuclei in these regions. Analysis was performed for a single z\u2009plane. Data were exported and analysed using GraphPad Prism.","Representative images of the segmentation can be found in the Supplementary Information.\r\nZebrafish experiments\r\nAll embryos were obtained by natural spawning and collected in conditioned aquarium water in 0.00001% methylene blue. Embryos were raised at 28.5\u2009\u00b0C in embryo medium (E3) on a 14\u2009h light\u201310\u2009h dark photoperiod and were treated with 200\u2009\u03bcM N-phenylthiourea (Sigma) from 6\u2009h post fertilization to inhibit pigmentation. Zygotes were injected at the one-cell stage of development. DNA constructs were created using the Tol2Kit system48,49.","Approximately 2\u2009nl of plasmid DNA (30\u2009ng\u2009\u03bcl\u20131) containing Tol2-capped mRNA (20\u2009ng\u2009\u03bcl\u20131), supplemented with 0.2% w\/v phenol red (Sigma) to facilitate visualization of injected volume, was injected. To induce Akt1 overexpression in neural progenitor cells, a combination of Tol2-pDEST-NBT:DlexPR-lexOP-pA (20\u2009ng\u2009\u03bcl\u20131) and Tol2-pDEST-lexOP:AKT1-lexOP:tagRFP (30\u2009ng\u2009\u03bcl\u20131) plasmids was injected as previously described50. The newly generated tgSSE-7:eGFP was outcrossed to tg(Xla.Tubb:DsRed) and tg(olig1:mScarlet). Live imaging was performed using either a Leica MS205 stereomicroscope or a Leica SP8 confocal microscope with a \u00d740\/NA 1.1 objective.","Animal experimentation was approved by the ethics review committee of the University of Edinburgh and the Home Office in accordance with the Scientific Procedure Act 1986.\r\nAAV medium-exchange transduction assay\r\nHEK cells were seeded into a 6-well plate a day before transfection, aiming for 60\u201370% confluence the next day. On the day of transfection, the culture medium was exchanged and HEK cells were transfected with RepCap, pHelper and AAV-ITR transgene plasmids using polyethylenimine. Two days later, conditioned HEK culture medium (now containing rAAV particles) was collected and centrifuged at 1,300\u2009rpm for 4\u2009min to remove any cells.","rAAV-conditioned medium was transferred to HEK or GSC7 cells seeded at low density (10\u201320%) in 6-well or 12-well plates. Two to three days later, cells were analysed for transduction. For AAV1 in vivo experiments, we purchased research-grade viral vectors from Vector Biolabs (iodixanol gradient ultracentrifugation purified).\r\nAAV transduction assay in iPS cell-derived neurons\r\nGSC7 and E55 cells were seeded in a 24-well plate format the day before transduction, whereas ioGlutamatergic neurons, ioGABAergic neurons and microglia (bit.bio) were seeded in a 48-well format before differentiation. After differentiation, cells were transduced with AAVs at a MOI of 5\u2009\u00d7\u2009105.","Cells were incubated at 37\u2009\u00b0C and 5% CO2 for 3 or 10\u00a0days.\r\nCell seeding and transduction for GCV killing assay\r\nOn the day of cell seeding, cells were detached using the method described above. Cells were counted using a haemocytometer and seeded into 96-well (1,000 cells per well in 50\u2009\u00b5l), 24-well (30,000 cells per well in 500\u2009\u00b5l) or 6-well (60,000 cells per well in 2\u2009ml) Corning plate and placed in an incubator at 37\u2009\u00b0C and 5% CO2. The next day, AAV stocks were thawed at room temperature, and an appropriate amount of virus stock was added to the required amount of culture medium to achieve the final MOI of 5\u2009\u00d7\u2009105. Culture medium containing viral particles was added to cells without the replacement of existing medium.","Cells were returned to a 37\u2009\u00b0C and 5% CO2 incubator.\r\nProdrug treatment\r\nLyophilized GCV was diluted in DMSO to achieve 100\u2009mM stock concentration. GCV stocks were aliquoted and were stored at \u221220\u2009\u00b0C for no longer than 1\u2009month. To make a working stock concentration, GCV was diluted 1:100 in appropriate culture medium, and 20\u2009\u00b5l was added to wells already containing 80\u2009\u00b5l culture medium for a final concentration of 200\u2009\u00b5M. As a negative control, DMSO was diluted 1:100 in appropriate culture medium to achieve a working stock.","Next, 20\u2009\u00b5l of working stock was added to wells already containing 80\u2009\u00b5l culture medium. As a positive control, 20\u2009\u00b5l DMSO was added to wells with 80\u2009\u00b5l culture medium to achieve a final concentration of 20% DMSO.\r\nIncucyte live-cell imaging\r\nTo track cell proliferation and morphological changes during treatments, cells were monitored using an Incucyte live-cell imaging system. Whole-well imaging of Corning 96-well\u00a0plates was performed every 4\u2009h. Basic confluence scoring analysis software (Incucyte) was used to estimate confluence.","Images at specific time points were extracted to verify cell confluence and morphology.\r\nMTT assay\r\nOn the day of assay, the culture medium was replaced with 0.3\u2009mg\u2009ml\u20131 MTT solution (diluted in cell line-appropriate culture medium). Cells were placed in incubator at 37\u2009\u00b0C and 5% CO2 for 3\u2009h. After incubation, the medium was removed and 70\u2009\u00b5l DMSO was added to each well. Each plate was kept in the dark at 37\u2009\u00b0C for 20\u2009min, shaking it occasionally. Before reading the plate, each well was visually inspected to make sure that all (formazan) crystals were dissolved.","Plates were read with plate reader at 560\u2009nm absorbance.\r\nStatistical analysis\r\nData analysis was performed using Microsoft Excel (v.16.23 for Mac), GraphPad Prism (v7) and RStudio (v1.1.456). Error bars are shown as the s.d. of the mean.","For illustrations, BioRender and Adobe Illustrator 22.0.1 were used.\r\n\r\nFurthermore, we used open-source programmes such as bedtools, GREAT, fastasplitter, HOMER and MEME for various analyses45,46.\r\nTumour initiation by transplantation and intratumoral AAV delivery\r\nAll animal procedures were approved by the University of Edinburgh Animal Welfare and Ethical Review Body (AWERB) and conducted under UK Home Office licence (PPL number: PP8631583) in accordance with the Animals (Scientific Procedures) Act 1986 and ARRIVE guidelines. Male and female C57BL\/6 mice (6\u20138\u2009weeks old) were obtained from Charles River.","Mice were housed in Individually ventilated cages with sterile bedding plus enrichment inside a pathogen-free facility on a 12-h light\u2013dark cycle, at a temperature range of 20\u201324\u2009\u00b0C and with relative humidity level of 45\u201365%. Transplantation experiments were performed as previously described25. We used 7-week-old male mice. Mice were transplanted with 200,000 NPE-IE cells in 2\u2009\u00b5l PBS per mouse. NPE-IE cells have been previously described25 and have Nf1 and Pten inactivating mutations, alongside EGFvIII overexpression and a GFP-Luc reporter construct.","At 2\u2009weeks, when tumours were visible by IVIS imaging (bioluminescence) and of similar size, AAV1-SSE-7\u2013HSV-TK\u2013IL-12 virus (1.2\u2009\u00d7\u20091013 viral genomes per ml, Vector Biolabs) was delivered directly into the mouse brain tumour (2.5\u2009\u00b5l per mouse) using a microsyringe linked to an injection pump, at a rate of 0.17\u2009\u00b5l\u2009min\u20131 and 2.3\u2009mm depth, higher than the previous depth of 2.4\u2009mm for tumour cells. Each mouse received 3.00\u2009\u00d7\u20091010 viral genomes. The following day after virus injections, GCV was intraperitoneally injected (500\u2009\u00b5l per mouse at 2\u2009mg\u2009ml\u20131) once daily for 20\u2009days. Mice were imaged by IVIS imaging (PerkinElmer) once a week for luciferase signals.","Animals were observed regularly for any neurological symptoms. Mice with tumours were culled based on the following criteria; (1) body weight loss of \u226520% relative to baseline (weights of mice were recorded once weekly); (2) abnormal neurological signs, including ataxia, head tilt, circling behaviour, seizures, paresis or inability to right itself within 3\u2009s; (3) hunched posture with reduced locomotor activity and lack of reach response; and (4) any extracranial tumours\u00a0\u2265\u00a015\u2009mm in diameter or ulcerated as per the Workman (2010) guidelines51. All mice reaching the humane end point were culled using the schedule\u00a01 method.","The end-point limits were not exceeded in any of the experiments.\r\nImmune characterization of mouse GBMs\r\nBrains were collected and processed 5\u2009days after treatment.","First, they were microdissected under a Leica stereo microscope to avoid contamination with healthy tissue. Tissues were mechanically disaggregated using scissors, followed by 30\u2009min of enzymatic digestion using a mix of 0.35\u00a0\u00b5g\u2009ml\u20131 Liberase TL (05401119001, Roche) and 0.23\u2009\u00b5g\u2009ml\u20131 DNAse (101041590001, Sigma) in plain RPMI. After digestion, tissues were dispersed through a 70\u2009nm cell strainer to obtain a single-cell suspension. After centrifugation, cells were resuspended in PBS for further flow cytometry staining. Cell suspensions were incubated for 10\u2009min at 4\u2009\u00b0C with a blocking mixture of mouse, rat and calf serum containing anti-CD16\/32 blocking antibodies (clone 2.4G2, BioXcell, BE0307).","Cells were washed with PBS and incubated for 20\u2009min in antibody-containing staining buffer plus Fixable Viability Dye eFluor-780 (eBioscience, 65-0865-14) to distinguish live and dead cells. Cells were then washed and resuspended in fixation\/permeabilization buffer (BD CytoFix\/Cytoperm; 554714 or FOXP3 Transcription Factor Staining Buffer set, eBioscience) followed by intracellular target staining.\r\nSABER\u2013FISH detection of transgene copy number\r\nSABER\u2013FISH was performed as previously described52,53. Non-overlapping 36\u2009bp probes were designed to target the HSV-TK transgene, and probes were extended using primer exchange reaction to approximately 500\u2009bp.","For imaging, cells were first stained with mCherry,then fixed with PFA and subsequently processed for immuno-SABER\u2013FISH54.\r\nAssessment of the fraction of the tumour cell population that activates AAV1-SSE-7\u2013mCherry\r\nAAV1-SSE-7-mCMV\u2013HSV-TK-V5\u2013mCherry was injected at around 3.58\u2009\u00d7\u20091014 viral genomes per ml and 7.2\u2009\u00d7\u20091013 viral genomes per ml (1:5 dilution) in PBS. AAV was injected into tumours as described above. After 7\u2009days, tumours were imaged using a stereomicroscope and surgically excised.\r\n\r\nEach tumour sample was added to 1\u2009ml digest mix (RPMI\u2009+\u20091\u2009mg\u2009ml\u20131 collagenase IV, 12.6\u2009\u00b5g\u2009ml\u20131 DNase\u2009I and 1% penicillin\u2013streptomycin) and incubated for 1\u2009h in a shaking incubator at 40\u2009rpm at 37\u2009\u00b0C.","Samples were filtered through a 100\u2009\u03bcm cell strainer, washed with 9\u2009ml FACS buffer and centrifuged at 250g for 15\u2009min at room temperature. The pellet was then resuspended in 1\u2009ml 37.5% Percoll in PBS and gradient centrifuged at 900g, room temperature, for 12\u2009min with no brake. The supernatant was discarded, and the bottom layer of cells resuspend in 500\u2009\u00b5l PBS and filtered into a FACS tube.","Cells were stained with DAPI as a live\/dead stain and analysed using a BD LSR Fortessa cell analyser (four lasers, BD Bioscience).\r\nReporting summary\r\nFurther information on research design is available in the\u00a0Nature Portfolio Reporting Summary linked to this article.\r\n\nData availability\r\nThe scRNA-seq data have been deposited at the European Nucleotide Archive (ENA) (accession number: PRJEB81816). The project number for ChIP\u2013seq data is PRJEB107008. For ChIP\u2013seq, raw (fastq) and processed (bigwig and narrowPeak) files have been deposited and made public (https:\/\/www.ebi.ac.uk\/ena\/browser\/view\/PRJEB107008).","The ENA does not take processed files; therefore, these have been uploaded to BioStudies, which links to the same ENA accession (PRJEB107008), otherwise they are at https:\/\/www.ebi.ac.uk\/biostudies\/studies\/S-BSST2733. All the codes and processing steps for scRNA-seq, SCENIC analysis and ChIP\u2013seq data are available at GitHub (https:\/\/github.com\/alhafidzhamdan\/sse_gene_therapy). The GitHub repository includes some processed files, including all the narrowPeak files and consensus sets; using these in conjunction with the .Rmd document for reproduction of Fig. 2. This also includes a link to the SOX2 and SOX9 tracks on the UCSC browser.","The linked and minted a repo is available at Zenodo (https:\/\/zenodo.org\/records\/18676096)55.\u00a0Source data is provided with this paper.\r\nReferences\r\nSuv\u00e0, M. L. et al. Reconstructing and reprogramming the tumor-propagating potential of glioblastoma stem-like cells.","Cell 157, 580\u2013594 (2014).\r\n\r\nArticle\nPubMed\nPubMed Central\nGoogle Scholar\r\n\r\nLi, C. & Samulski, R. J. Engineering adeno-associated virus vectors for gene therapy. Nat. Rev. Genet.","21, 255\u2013272 (2020).\r\n\r\nArticle\nCAS\nPubMed\nGoogle Scholar\r\n\r\nKribelbauer, J.","F.,","Rastogi, C., Bussemaker, H. J.","& Mann, R. S. Low-affinity binding sites and the transcription factor specificity paradox in eukaryotes. Annu. Rev. Cell Dev. Biol.","35, 357\u2013379 (2019).\r\n\r\nArticle\nCAS\nPubMed\nPubMed Central\nGoogle Scholar\r\n\r\nSpitz, F.","& Furlong, E. E. M. Transcription factors: from enhancer binding to developmental control. Nat. Rev. Genet.","13, 613\u2013626 (2012).\r\n\r\nArticle\nCAS\nPubMed\nGoogle Scholar\r\n\r\nSchlabach, M.","R.,","Hu, J.","K.,","Li, M.","& Elledge, S. J. Synthetic design of strong promoters. Proc. Natl Acad. Sci.","USA 107, 2538\u20132543 (2010).\r\n\r\nArticle\nADS\nCAS\nPubMed\nPubMed Central\nGoogle Scholar\r\n\r\nWu, M.-R. et al. A high-throughput screening and computation platform for identifying synthetic promoters with enhanced cell-state specificity (SPECS). Nat. Commun.","10, 2880 (2019).\r\n\r\nArticle\nADS\nPubMed\nPubMed Central\nGoogle Scholar\r\n\r\nCheng, J. K.","& Alper, H. S. Transcriptomics-guided design of synthetic promoters for a mammalian system. ACS Synth. Biol.","5, 1455\u20131465 (2016).\r\n\r\nArticle\nCAS\nPubMed\nGoogle Scholar\r\n\r\nJindal, G. A.","& Farley, E. K. Enhancer grammar in development, evolution, and disease: dependencies and interplay. Dev.","Cell 56, 575\u2013587 (2021).\r\n\r\nArticle\nCAS\nPubMed\nPubMed Central\nGoogle Scholar\r\n\r\nWhyte, W. A. et al. Master transcription factors and mediator establish super-enhancers at key cell identity genes.","Cell 153, 307\u2013319 (2013).\r\n\r\nArticle\nADS\nCAS\nPubMed\nPubMed Central\nGoogle Scholar\r\n\r\nZamudio, A. V. et al. Mediator condensates localize signaling factors to key cell identity genes. Mol.","Cell 76, 753\u2013766 (2019).\r\n\r\nArticle\nCAS\nPubMed\nPubMed Central\nGoogle Scholar\r\n\r\nHnisz, D.,","Shrinivas, K.,","Young, R.","A., Chakraborty, A. K.","& Sharp, P. A. A phase separation model for transcriptional control.","Cell 169, 13\u201323 (2017).\r\n\r\nArticle\nADS\nCAS\nPubMed\nPubMed Central\nGoogle Scholar\r\n\r\nHnisz, D. et al. Convergence of developmental and oncogenic signaling pathways at transcriptional super-enhancers. Mol.","Cell 58, 362\u2013370 (2015).\r\n\r\nArticle\nCAS\nPubMed\nPubMed Central\nGoogle Scholar\r\n\r\nPollard, S. M. et al. Glioma stem cell lines expanded in adherent culture have tumor-specific phenotypes and are suitable for chemical and genetic screens. Cell Stem Cell 4, 568\u2013580 (2009).\r\n\r\nArticle\nCAS\nPubMed\nGoogle Scholar\r\n\r\nGraham, V.,","Khudyakov, J.,","Ellis, P.","& Pevny, L. SOX2 functions to maintain neural progenitor identity.","Neuron 39, 749\u2013765 (2003).\r\n\r\nArticle\nCAS\nPubMed\nGoogle Scholar\r\n\r\nLujan, E.,","Chanda, S.,","Ahlenius, H.,","S\u00fcdhof, T. C.","& Wernig, M. Direct conversion of mouse fibroblasts to self-renewing, tripotent neural precursor cells. Proc. Natl Acad. Sci.","USA 109, 2527\u20132532 (2012).\r\n\r\nArticle\nADS\nCAS\nPubMed\nPubMed Central\nGoogle Scholar\r\n\r\nGangemi, R. M. R. et al. SOX2 silencing in glioblastoma tumor-initiating cells causes stop of proliferation and loss of tumorigenicity. Stem Cells 27, 40\u201348 (2009).\r\n\r\nArticle\nCAS\nPubMed\nGoogle Scholar\r\n\r\nSingh, D. K. et al. Oncogenes activate an autonomous transcriptional regulatory circuit that drives glioblastoma. Cell Rep.","18, 961\u2013976 (2017).\r\n\r\nArticle\nCAS\nPubMed\nPubMed Central\nGoogle Scholar\r\n\r\nMacLeod, G. et al. Genome-wide CRISPR\u2013Cas9 screens expose genetic vulnerabilities and mechanisms of temozolomide sensitivity in glioblastoma stem cells. Cell Rep.","27, 971\u2013986 (2019).\r\n\r\nArticle\nCAS\nPubMed\nGoogle Scholar\r\n\r\nKuzmin, D. A. et al. The clinical landscape for AAV gene therapies. Nat. Rev. Drug Discov.","20, 173\u2013174 (2021).\r\n\r\nArticle\nCAS\nPubMed\nGoogle Scholar\r\n\r\nFarnham, P. J. Insights from genomic profiling of transcription factors. Nat. Rev. Genet.","10, 605\u2013616 (2009).\r\n\r\nArticle\nCAS\nPubMed\nPubMed Central\nGoogle Scholar\r\n\r\nSoufi, A.,","Donahue, G.","& Zaret, K. S. Facilitators and impediments of the pluripotency reprogramming factors\u2019 initial engagement with the genome.","Cell 151, 994\u20131004 (2012).\r\n\r\nArticle\nCAS\nPubMed\nPubMed Central\nGoogle Scholar\r\n\r\nStricker, S. H. et al. Widespread resetting of DNA methylation in glioblastoma-initiating cells suppresses malignant cellular behavior in a lineage-dependent manner. Genes Dev.","27, 654\u2013669 (2013).\r\n\r\nArticle\nCAS\nPubMed\nPubMed Central\nGoogle Scholar\r\n\r\nHuang, Y.-H.,","Jankowski, A.,","Cheah, K. S.","E.,","Prabhakar, S.","& Jauch, R. SOXE transcription factors form selective dimers on non-compact DNA motifs through multifaceted interactions between dimerization and high-mobility group domains. Sci. Rep.","5, 10398 (2015).\r\n\r\nArticle\nADS\nPubMed\nPubMed Central\nGoogle Scholar\r\n\r\nScott, C. E. et al. SOX9 induces and maintains neural stem cells. Nat. Neurosci.","13, 1181\u20131189 (2010).\r\n\r\nArticle\nCAS\nPubMed\nGoogle Scholar\r\n\r\nGangoso, E. et al. Glioblastomas acquire myeloid-affiliated transcriptional programs via epigenetic immunoediting to elicit immune evasion.","Cell 184, 2454\u20132470 (2021).\r\n\r\nArticle\nCAS\nPubMed\nPubMed Central\nGoogle Scholar\r\n\r\nWang, Q. et al. Tumor evolution of glioma-intrinsic gene expression subtypes associates with immunological changes in the microenvironment. Cancer Cell 32, 42\u201356 (2017).\r\n\r\nArticle\nPubMed\nPubMed Central\nGoogle Scholar\r\n\r\nNeftel, C. et al. An integrative model of cellular states, plasticity, and genetics for glioblastoma.","Cell 178, 835\u2013849 (2019).\r\n\r\nArticle\nADS\nCAS\nPubMed\nPubMed Central\nGoogle Scholar\r\n\r\nYan, Y.-L. et al. A pair of Sox: distinct and overlapping functions of zebrafish sox9 co-orthologs in craniofacial and pectoral fin development. Development 132, 1069\u20131083 (2005).\r\n\r\nArticle\nCAS\nPubMed\nGoogle Scholar\r\n\r\nSantiago-Ortiz, J. L. & Schaffer, D. V. Adeno-associated virus (AAV) vectors in cancer gene therapy. J. Control. Release 240, 287\u2013301 (2016).\r\n\r\nArticle\nCAS\nPubMed\nPubMed Central\nGoogle Scholar\r\n\r\nSwift, E.","A.,","Pollard, S. M.","& Parker, A. L.","Engineering cancer selective virotherapies: are the pieces of the puzzle falling into place?","Hum. Gene Ther.","33, 1109\u20131120 (2022).\r\n\r\nArticle\nCAS\nPubMed\nPubMed Central\nGoogle Scholar\r\n\r\nLasek, W., Zago\u017cd\u017con, R. & Jakobisiak, M. Interleukin 12: still a promising candidate for tumor immunotherapy?\u00a0Cancer Immunol. Immunother.","63, 419\u2013435 (2014).\r\n\r\nArticle\nCAS\nPubMed\nPubMed Central\nGoogle Scholar\r\n\r\nAgliardi, G. et al. Intratumoral IL-12 delivery empowers CAR-T cell immunotherapy in a pre-clinical model of glioblastoma. Nat. Commun.","12, 444 (2021).\r\n\r\nArticle\nADS\nCAS\nPubMed\nPubMed Central\nGoogle Scholar\r\n\r\nHacker, U.","T.,","Bentler, M.,","Kaniowska, D.,","Morgan, M.","& B\u00fcning, H. Towards clinical implementation of adeno-associated virus (AAV) vectors for cancer gene therapy: current status and future perspectives. Cancers 12, 1889 (2020).\r\n\r\nArticle\nCAS\nPubMed\nPubMed Central\nGoogle Scholar\r\n\r\nAldape, K. et al. Challenges to curing primary brain tumours. Nat. Rev. Clin. Oncol.","16, 509\u2013520 (2019).\r\n\r\nArticle\nCAS\nPubMed\nPubMed Central\nGoogle Scholar\r\n\r\nDarmanis, S. et al. Single-cell RNA-seq analysis of infiltrating neoplastic cells at the migrating front of human glioblastoma. Cell Rep.","21, 1399\u20131410 (2017).\r\n\r\nArticle\nCAS\nPubMed\nPubMed Central\nGoogle Scholar\r\n\r\nArrieta, V. A. et al. Immune checkpoint blockade in glioblastoma: from tumor heterogeneity to personalized treatment. J. Clin. Invest.","133, e163447 (2023).\r\n\r\nArticle\nCAS\nPubMed\nPubMed Central\nGoogle Scholar\r\n\r\nLiau, L. M. et al. Association of autologous tumor lysate-loaded dendritic cell vaccination with extension of survival among patients with newly diagnosed and recurrent glioblastoma: a phase 3 prospective externally controlled cohort trial. JAMA Oncol.","9, 112\u2013121 (2022).\r\n\r\nArticle\nGoogle Scholar\r\n\r\nTodo, T. et al. Intratumoral oncolytic herpes virus G47\u2206 for residual or recurrent glioblastoma: a phase 2 trial. Nat. Med.","28, 1630\u20131639 (2022).\r\n\r\nArticle\nCAS\nPubMed\nPubMed Central\nGoogle Scholar\r\n\r\nO\u2019Dwyer, M. R. et al. Nucleosome fibre topology guides transcription factor binding to enhancers. Nature 638, 251\u2013260 (2025).\r\n\r\nArticle\nADS\nPubMed\nGoogle Scholar\r\n\r\nCompany, C. et al. Logical design of synthetic cis-regulatory DNA for genetic tracing of cell identities and state changes. Nat. Commun.","15, 897 (2024).\r\n\r\nArticle\nADS\nCAS\nPubMed\nPubMed Central\nGoogle Scholar\r\n\r\nTaskiran, I. I. et al. Cell-type-directed design of synthetic enhancers. Nature 626, 212\u2013220 (2022).\r\n\r\nArticle\nADS\nGoogle Scholar\r\n\r\nde Almeida, B. P. et al. Targeted design of synthetic enhancers for selected tissues in the Drosophila embryo. Nature 626, 207\u2013211 (2024).\r\n\r\nArticle\nADS\nPubMed\nGoogle Scholar\r\n\r\nMartella, A., Matjusaitis, M.,","Auxillos, J.,","Pollard, S. M.","& Cai, Y. EMMA: an extensible mammalian modular assembly toolkit for the rapid design and production of diverse expression vectors. ACS Synth. Biol.","6, 1380\u20131392 (2017).\r\n\r\nArticle\nCAS\nPubMed\nGoogle Scholar\r\n\r\nMcLean, C. Y. et al. GREAT improves functional interpretation of cis-regulatory regions. Nat. Biotechnol.","28, 495\u2013501 (2010).\r\n\r\nArticle\nADS\nCAS\nPubMed\nPubMed Central\nGoogle Scholar\r\n\r\nHeinz, S. et al. Simple combinations of lineage-determining transcription factors prime cis-regulatory elements required for macrophage and B cell identities. Mol.","Cell 38, 576\u2013589 (2010).\r\n\r\nArticle\nCAS\nPubMed\nPubMed Central\nGoogle Scholar\r\n\r\nMa, W.,","Noble, W. S.","& Bailey, T. L. Motif-based analysis of large nucleotide data sets using MEME-ChIP. Nat. Protoc.","9, 1428\u20131450 (2014).\r\n\r\nArticle\nCAS\nPubMed\nPubMed Central\nGoogle Scholar\r\n\r\nSoufi, A. et al. Pioneer transcription factors target partial DNA motifs on nucleosomes to initiate reprogramming.","Cell 161, 555\u2013568 (2015).\r\n\r\nArticle\nCAS\nPubMed\nPubMed Central\nGoogle Scholar\r\n\r\nKawakami, K. Tol2: a versatile gene transfer vector in vertebrates. Genome Biol.","8, S7 (2007).\r\n\r\nArticle\nPubMed\nPubMed Central\nGoogle Scholar\r\n\r\nKwan, K. M. et al. The Tol2kit: a multisite gateway-based construction kit for Tol2 transposon transgenesis constructs. Dev. Dyn.","236, 3088\u20133099 (2007).\r\n\r\nArticle\nCAS\nPubMed\nGoogle Scholar\r\n\r\nChia, K.,","Mazzolini, J.,","Mione, M.","& Sieger, D. Tumor initiating cells induce Cxcr4-mediated infiltration of pro-tumoral macrophages into the brain.","eLife 7, e31918 (2018).\r\n\r\nArticle\nPubMed\nPubMed Central\nGoogle Scholar\r\n\r\nWorkman, P. et al.\u00a0Guidelines for the welfare and use of animals in cancer research. Br. J. Cancer 102, 1555\u20131577 (2010).\r\n\r\nKishi, J. Y. et al. SABER amplifies FISH: enhanced multiplexed imaging of RNA and DNA in cells and tissues. Nat. Methods 16, 533\u2013544 (2019).\r\n\r\nArticle\nCAS\nPubMed\nPubMed Central\nGoogle Scholar\r\n\r\nWang, S.","K.,","Lapan, S.","W.,","Hong, C.","M.,","Krause, T. B.","& Cepko, C. L. In situ detection of adeno-associated viral vector genomes with SABER-FISH. Mol. Ther. Methods Clin. Dev.","19, 376\u2013386 (2020).\r\n\r\nArticle\nPubMed\nPubMed Central\nGoogle Scholar\r\n\r\nPurshouse, K. et al. Oncogene expression from extrachromosomal DNA is driven by copy number amplification and does not require spatial clustering in glioblastoma stem cells.","eLife 11, e80207 (2022).\r\n\r\nArticle\nCAS\nPubMed\nPubMed Central\nGoogle Scholar\r\n\r\nHamdan, A.","& Wang, Z. alhafidzhamdan\/sse_gene_therapy: For Koeber et al., Nature 2026 (1.0). Zenodo https:\/\/doi.org\/10.5281\/zenodo.18676096 (2026).\r\n\r\nDownload references\r\nAcknowledgements\r\nU.K.,","M.M., S.J.R. and S.M.P. are supported by the BBSRC UK Mammalian Synthetic Biology Centre (BB\/M018040\/1). The GSC and NSC lines were obtained from the GCGR. The GCGR was funded by a Cancer Research UK (CRUK) Accelerator Award (C157\/A21992). A.W. was supported by a BBSRC funded project grant. A.H. was supported by CRUK and the Edinburgh Clinical Academic Training (ECAT) programme (C157\/A29279). D.S. is supported by a CRUK programme foundation award. S.M.P. is supported by CRUK programme grant (DRCNPG-Nov21\\100002). S.A.Q. is funded by a CRUK Senior Cancer Research Fellowship (C36463\/A22246) and a CRUK Biotherapeutic Program grant (C36463\/A20764).","This work was undertaken at University College London with support from the CRUK\u00a0City of London Centre (C7893\/A26233). S.M.P. and S.A.Q. also received support from the CRUK Brain Tumour Award (A28592). We thank J.\u2009Mendrychowski and V.\u2009Kho for technical support; I.\u2009Chambers and J.\u2009M.\u2009Iglesias for their comments on the manuscript; and A.\u2009Corsinotti for technical advice on scRNA-seq.","For open access, the author has applied a Creative Commons Attribution (CC BY) licence to any Author Accepted Manuscript version arising from this submission.\r\nAuthor information\r\nAuthor notes\r\n\r\nUte Koeber\r\n\r\nPresent address: IQVIA RDS, Frankfurt am Main, Germany\r\n\r\nMantas Matjusaitis\r\n\r\nPresent address: Plurify, Cambridge, UK\r\n\r\nRachel White,\u00a0Michelle P.","Clark,\u00a0Carla Jacinto de Sousa\u00a0&\u00a0Ben Southgate\r\n\r\nPresent address: Trogenix, Edinburgh, UK\r\n\r\nFelipe Galvez Cancino\r\n\r\nPresent address: Laboratory of Immune Regulation, NDM Centre for Immuno-Oncology, University of Oxford, Oxford, UK\r\n\r\nThese authors contributed equally: Ute Koeber, Mantas Matjusaitis\r\nAuthors and Affiliations\r\nCentre for Regenerative Medicine, Institute for Regeneration and Repair, University of Edinburgh, Edinburgh, UK\r\n\r\nUte Koeber,\u00a0Mantas Matjusaitis,\u00a0Neza Alfazema,\u00a0Katharine Furlong,\u00a0Zeyu Wang,\u00a0Rachel White,\u00a0Alhafidz Hamdan,\u00a0Pooran Dewari,\u00a0Rosie Willis,\u00a0Michelle P.","Clark,\u00a0Carla Jacinto de Sousa,\u00a0Hei Ip Hong,\u00a0Shahida Sheraz,\u00a0Ben Southgate,\u00a0Justyna Cholewa-Waclaw,\u00a0Sabine Gogolok,\u00a0Gillian M.","Morrison,\u00a0Faye Robertson,\u00a0Anna Williams,\u00a0Paul M.","Brennan,\u00a0Abdenour Soufi\u00a0&\u00a0Steven M. Pollard\r\n\r\nCancer Research UK Scotland Centre, Edinburgh, UK\r\n\r\nUte Koeber,\u00a0Mantas Matjusaitis,\u00a0Neza Alfazema,\u00a0Zeyu Wang,\u00a0Rachel White,\u00a0Alhafidz Hamdan,\u00a0Pooran Dewari,\u00a0Gregoire Morisse,\u00a0Rosie Willis,\u00a0Jin Wang,\u00a0Michelle P.","Clark,\u00a0Carla Jacinto de Sousa,\u00a0Hei Ip Hong,\u00a0Shahida Sheraz,\u00a0Ben Southgate,\u00a0Justyna Cholewa-Waclaw,\u00a0Sabine Gogolok,\u00a0Gillian M.","Morrison,\u00a0Faye Robertson,\u00a0Paul M.","Brennan,\u00a0Dirk Sieger\u00a0&\u00a0Steven M. Pollard\r\n\r\nEdinburgh Pathology, Royal Infirmary Edinburgh, NHS Lothian, Edinburgh, UK\r\n\r\nAlhafidz Hamdan\r\n\r\nInstitute for Neuroscience and Cardiovascular Research, University of Edinburgh, Edinburgh, UK\r\n\r\nGregoire Morisse,\u00a0Jin Wang\u00a0&\u00a0Dirk Sieger\r\n\r\nImmune Regulation and Tumor Immunotherapy Group, Cancer Immunology Unit, Research Department of Haematology, UCL Cancer Institute, London, UK\r\n\r\nMariela Navarette,\u00a0Felipe Galvez Cancino\u00a0&\u00a0Sergio A. Quezada\r\n\r\nSchool of Biological Sciences, University of Edinburgh, Edinburgh, UK\r\n\r\nSusan J. Rosser\r\n\r\nCentre for Clinical Brain Sciences, University of Edinburgh, Edinburgh, UK\r\n\r\nPaul M. Brennan\r\nContributions\r\nS.M.P.","conceived the study, secured funding and supervised the project. U.K. designed and performed in vitro experiments to construct the SSEs, with bioinformatic support from B.S. M.M. designed and performed in vitro experiments related to the therapeutic transgene experiments and selectivity, and with N.A. in the animal efficacy studies. A.W. provided human OPCs. M.P.C. led the human in vitro selectivity experiments. The ChIP\u2013seq analysis was performed by P.D., and A.H. led the bioinformatic analyses. Z.W. designed and performed the scRNA-seq experiments. R.\u00a0Willis performed SABER\u2013FISH.\u00a0K.F. performed the biochemical analysis of SOX interactions, with supervision from A.S.","Key technical support for all the molecular biology was provided by R.\u00a0White and S.G. All zebrafish studies were performed by G.M.","and J.W., with design and supervision from D.S. Manufacture and quality control of all viral vectors was performed by M.M., R.\u00a0White and C.J.d.S. The inhibitor screen was designed and performed by H.I.H., with supervision from S.M.P. The human tissue slices were provided by P.M.B., with AAV testing performed by R.\u00a0Willis, S.S., M.P.C. and J.C.-W. The initial support for the transgene optimization and patient-derived GSC models was from S.G. and G.M.M. F.R. supported mouse model experimental designs and translational relevance. The plasmid library manufacture was supported by S.J.R. and staff at the Edinburgh Genome Foundry. All immune characterization was designed and performed by M.N. and F.G.C., with supervision from S.A.Q.","The manuscript and figures were drafted by S.M.P., with contributions from all authors. All authors reviewed, edited and approved the final version.\r\nCorresponding author\r\nCorrespondence to\nSteven M. Pollard.\r\nEthics declarations\nCompeting interests\r\nS.M.P., U.K. and M.M. are inventors of patents filed by the University of Edinburgh related to the discoveries in this study and to SSEs and associated sequences (WO2022162361A1). S.M.P. is founder of Trogenix, which is developing novel gene therapies for oncology, and is a paid consultant (part-time chief scientific officer, since February 2024) and shareholder.","B.S.,","R.\u00a0White, C.J.d.S. and M.P.C. have recently joined Trogenix as employees (since October 2025). S.A.Q. is a paid member of the scientific advisory board of Trogenix. P.M.B. and F.R. are paid external consultants for Trogenix related to ongoing clinical translation of these technologies. All other authors (N.A.,","K.F.,","Z.W.,","A.H.,","P.D.,","G.M.,","M.N., R.\u00a0Willis.,","J.W.,","H.I.H.,","S.S.,","J.C.-W.,","S.G.,","G.M.M.,","F.G.C.,","A.W.,","S.J.R., D.S. and A.S.) declare no competing interests.\r\n\nPeer review\nPeer review information\r\nNature thanks Ralf Jauch and the other, anonymous, reviewer(s) for their contribution to the peer review of this work.\r\n\nAdditional information\r\nPublisher\u2019s note Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.\r\nExtended data figures and tablesExtended Data Fig. 1 Design of candidate SOX2 GSC-specific peaks and their conversion into a library of candidate enhancer fragments.\r\n(a) Summary of\u00a0the workflow to design and build the enhancer fragment library (DGC: differentiated glioma cell). (b) The genomic annotation of the final library of candidate peaks.","(c) The top 16 fragment hits identified from the initial screen (Fig. 1) were validated in GSC7 GSCs for mNeongreen expression using flow cytometry (median fluorescence intensity shown (left). (d) The fragments have the same level of expression in HEK negative control cells as the minimal CMV (MFI and % of total cell population shown, left and right, respectively). (e) De novo motif analysis using HOMER for the Top32 hits of the validation (using shuffled sequences as background) from Fig.","1, confirms the repertoire of candidate TF motifs that are enriched in the functional enhancer fragments.\r\n\r\nSource data\r\nExtended Data Fig. 2 A dimeric SOX motif that binds SOX9 is a major contributor to activity.\r\n(a) A specific dimeric SOX dimer motif was identified using MEME for de novo motif discovery and is enriched in the set of functional enhancer hits (ACAAAGRGBVNHTKK) and was mutated as depicted (below) for 10 different variants of the ID2904 sequence to determine the functional impact. (b-e) Plasmid-based flow cytometry assay to determine the activity of each variant, in GSCs (GSC7 and E28, respectively).","Interestingly, increased activity was seen with 2\u2009bp reduction of the spacing consistent with the importance of this motif for activity. Each dot represents a biological replicate (2e) One-way Anova test with Dunnett\u2019s multiple comparisons test against ID2904 WT; ** represent p-value of 0.0040 (f) Examples of flow cytometry data used for panels c-f. M(f) Examples of flow cytometry data used for panels c-f. (g) SOX9 protein selectively interacts with ID1101 and ID2904 through the dimer motif.\r\n\r\nSource data\r\nExtended Data Fig.","3 Assessment of SSE-7 activity across a set of seven different patient-derived GSC lines.\r\n(a) Reporter reference controls were used to determine relative SSE activity by flow cytometry across each cell line, with minimal CMV (mCMV) and CMV as negative and positive controls, respectively.","(c, e, g, i, k, m and o) shows the % mCherry positive cells, whereas (d, f, h, j, l, n and p) shows the median fluorescence intensity (MFI) for each cell. E37 is an outlier line and has an atypical MYC amplification and lower levels of SOX2. Error bars represent the SD of the mean, each dot represents a biological replicate.\r\n\r\nSource data\r\nExtended Data Fig. 4 Characterisation of SSEs.\r\n(a) Comparison of activity from all seven SSE variants in GSC7 using flow cytometry to monitor expression of the mNeonGreen reporter. Each dot represents a biological replicate. (b) Time-course of GSC astrocyte differentiation to monitor SSE activity using flow cytometry.","(n\u2009=\u20092 biological replicates, each dot symbol represents the mean, the error bars represent the SD of the mean) (c) Live cell wide-field immunofluorescence imaging for the cells assessed in (b). (d-e) Quantitation of the band intensities from the EMSA experiments in Fig.","2d,e. Data points are mean from n\u2009=\u20093 with error bars of the SD. Solid line represents the Hill equation with the Kd values displayed above the graph. (f) EMSA of SSE-7 DNA using purified recombinant SOX2 and SOX9, alone or in combination (representative image of n\u2009=\u20093). (g) EMSA showing the individual and combinatorial binding of purified recombinant SOX2 and SOX9 to FGF4 enhancer element (40\u2009bp, n\u2009=\u20091) containing a typical SOX2 motif.\r\n\r\nSource data\r\nExtended Data Fig. 5 Single cell mRNA profiling of multiple patient derived GSC lines following transduction with AAV1-SSE-7-mCherry virus.\r\n(a) Distribution of transcriptional subtypes in scRNA data.","UMAP of transcriptional subtype in E17 (n\u2009=\u200929,829 cells), E20 (n\u2009=\u200923,301 cells), E31 (n\u2009=\u200939,823 cells), E34 (n\u2009=\u200927,766 cells), E43 (n\u2009=\u200931,538 cells), E55 (n\u2009=\u200947,291 cells). *P\u2009<\u20090.05, **P\u2009<\u20090.01, ***P\u2009<\u20090.001; NS, not significant. (b) SOX2 and SOX9 are broadly expressed across the diverse transcriptional subtype cell states in GSCs, consistent with their roles as core regulators. See also Extended Data Fig. 6 for cell cycle analysis.\r\nExtended Data Fig. 6 Single cell analysis of cell cycle and SCENIC analysis.\r\n(a) Distribution of cell cycle phases across GSC cell states in scRNA data (as per Extended Data Fig. 5).","(b) Full data for SCENIC analysis of transcription factor modules identifies signalling endpoint transcription factors alongside SOX as correlating with SSE-7 activity. (c) Heatmap across technical replicates of libraries from Fig. 3i.\r\nExtended Data Fig. 7 A screen of small molecule pharmacological inhibitors to identify signalling pathways that may impact the activity of SSE-7 in E55 GSCs.\r\n(a) The primary screen was performed in the E55 cells. Heatmap summarising the fold change of the normalised mCherry MFI in inhibitor-treated samples compared with DMSO. Initial hits identified with red font. The relative comparator is the CMV-SSE7-mCherry positive control.","Any reduction in mCherry selectively in the E55 is due to the different properties of the SSE7. (n\u2009=\u20092 biological replicate; independent screens) (b) Validation of the hit compounds using flow cytometry for mCherry detection across three different concentrations of compound (0.1, 0.3 or 3\u03bcM). The heatmap shows the fold change of the normalised mCherry MFI and percentages of mCherry positive cells of inhibitor-treated samples relative to the DMSO-treated control. (n\u2009=\u20092 biological replicates). Total % mCherry cells (top panel) and mean fluorescence intensity (MFI) bottom panel. Red font are MAPK signalling-related compounds validated with dose response and show selective effects on the SSE-7 compared to CMV promoter.","(c) Flow cytometry histograms for each validated hit compound. Typical example of flow cytometry overlay histograms for E55 GSCs treated with 3 different doses of MAPK signalling-related inhibitors (left panels) and E31 GSCs (right panels). E55 and E31 have distinct transcriptional subtype identities (mesenchymal and developmental\/proneural, respectively). (d) Summary of validated inhibitor hits (red font) and\u00a0their mechanism of action within the MAPK signalling pathway. (e) Flow cytometry analysis of positive and negative SSE-7mCherry cells sorted and reanalysed over a 72-hr time-course shows that SSE-7 activity can be activated in cells that are initially negative, indicative of a readout that is sensitive to signalling state.","n\u2009=\u20093 biological replicates. 2-way ANOVAs followed by two stage setup method were carried out to compare with mCMV neg. *P\u2009<\u20090.05, **P\u2009<\u20090.01, ***P\u2009<\u20090.001; NS, not significant. Note: SC-1 was toxic at the higher doses for E31 and so could not be assessed.\r\n\r\nSource data\r\nExtended Data Fig. 8 Selectivity in tissue expression in transgenic zebrafish embryos.\r\n(a) Reporter transgene for selective expression of SSEs in a zebrafish transgenic reporter model with a GFP reporter, with a transposase (Tol2) reporter designed track transgene insertion using mCherry expression in the eye from the Crystal promoter.","(b) SSE activity is tracked by eGFP levels in live cell imaging of the embryos at 48hpf and is tissue-restricted compared parallel CMV controls for each SSE tested. Note: the minCMV control exhibited GFP-expression in the eyes due to spurious transcription from the Crystal eye promoter. (c) A stable transgenic SSE-7:eGFP reporter zebrafish line confirms expression is restricted to the developing nervous system. Images were acquired using a Leica MS205 stereomicroscope. Scale bars represent 500\u2009\u00b5m for (b) and 728\u2009\u00b5m for (c). (d) Higher magnification image of the spinal cord and nototchord with neuronal reporter (Cyan, as per (e)).","(e) Outcross of tgSSE-7:eGFP to tg(Xla.Tubb:DsRed) reveals no expression in neurons; in (f) outcross of tgSSE-7:eGFP to tg(olig1:mScarlet) shows expression in a subset of oligodendrocyte precursor cells. Images represent maximum intensity projections of confocal stacks (optic tectum and cerebellum shown). Images were captured using a Leica SP8 confocal microscope with a 40X\/NA 1.1 objective. Scale bars represent 30\u2009\u00b5m. (g) Quantification of the OPC data from (f). (h)\u00a0Overexpression of human Akt1 in neuronal cells results in strong activity of\u00a0SSE-7. Images represent maximum intensity projections of confocal stacks. Images were captured using a Leica SP8 confocal microscope with a 40X\/NA 1.1 objective. Scale bars represent 30\u2009\u00b5m.","(i) Immunostaining and flow cytometry (top and bottom, respectively) of human ESC differentiated oligodendrocyte progenitor cells shows that SSE-7 is not active compared to CMV (n\u2009=\u20091). (j) SSE7 versus CMV across a set of three human foetal neural stem cell (NSC) cultures (derived from week of gestation 9, 12 and 17 tissue; NS9FB_B, NS12ST_A and NS17ST_A, respectively). Each dot represents a biological replicate. Error bars represent the SD of the mean.\r\n\r\nSource data\r\nExtended Data Fig. 9 Selective activity of AAV1-SSE7-mCherry in SOX2 expressing cells within the tumour tissue mass and infiltrating margin.\r\n(a)\u00a0Tissue slice-culture experiments were performed across diverse subtypes of high-grade glioma similarly to Fig. 2.","Immunostaining and confocal for mCherry across distinct tumour subtypes shows restricted activity of SSE-7-mCherry in\u00a0the tumour margin relative to CMV. (b) Quantitation of the % of mCherry expression in the tumour versus the margin shows SSE-7 is activated in a subset of GBM cells. (n\u2009=\u20094 for both tumour and margin with both CMV and SSE7). Matched tumour and margin samples were used across multiple tumour types; GBM IDH-WT (isocitrate dehydrogenase wild-type), Grade 4 Astrocytoma IDH-MUT (isocitrate dehydrogenase mutant), Grade 3 high-grade astrocytoma and denoted by distinct symbols, each point is a biological replicate. Error bars represent SD.","Statistical significance was assessed by two-way ANOVA with multiple-comparisons correction (p\u2009=\u20090.0182). n\u2009=\u20094. Image analysis and segmentation pipelines provided in the Supplementary Data. Scale bar = 100 micron. (c) Staining for mCherry alongside SOX2 and Nestin in a recurrent GBM\u00a0sample treated with the AAV1-SSE7-mCherry vector, confirming that SSE-7 is active in a subset of the tumour cells. Left panels, low magnification (no matched normal tissue was available for this recurrent sample).\r\n\r\nSource data\r\nExtended Data Fig. 10 In vitro killing of GSCs, but not fibroblasts, is achieved using SSE7-driven HSV-TK.\r\n(a) Schematic of the AAV-2 transgene encoding both HSV-tk and mCherry used in vitro to assess selective cytotoxicity.","(b and c) Flow cytometry was performed 4 days post-transduction. Each dot represents biological replicate (n\u2009=\u20093), error bars represent SD. (b) % mCherry positive cells and (c) median fluorescent intensity (MFI) of mCherry positive cells. (d) Western blot image developed against V5 and GAPDH in transduced and non-transduced GSC7 and huFB170 cells; the higher weaker band is the small fraction that was uncleaved. (e) Incucyte live cell imaging generate proliferation curves to determine the change in confluence over 10-day time-course for human GSC7 cells transduced with CMV or SSE-7-driven TKv5-P2A-mCherry in presence of 200\u2009\u00b5M GCV, 0.2% DMSO (vehicle control) or 20% DMSO (positive control).","(f; top panels) Live images of GSC7 cells at day 8 demonstrating cell number and morphological changes in cells transduced with constructs in the presence of 200\u2009\u00b5M GCV. (f; bottom panels) Live images of huFB170 cells at day 17 demonstrating cell number and morphological changes in cells transduced with constructs in the presence of 200\u2009\u00b5M GCV. (g) MTT assay for cell viability in GSC-7 and fibroblasts. Each dot represents an independent biological replicate. MTT values were normalized to vehicle-only control. Error bars represent SD (n\u2009=\u20093).\r\n\r\nSource data\r\nExtended Data Fig.","11 Evaluating SSE7-TKv5-P2A-mCherry activity from an AAV vector.\r\n(a) Example of flow cytometry data for GSCs (E17) cells following transduction at MOI of 5E5. (b) Heatmap summarizing mean AAV transduction efficiency in different patient-derived GSC cultures (n\u2009=\u20093, except E34 and E21, which are n\u2009=\u20092). (c) SSE-7 activity in mouse fibroblasts and GSCs. (c) Flow cytometry data comparing transcriptional activity between SSE-7- and CMV-driven TKv5-mCherry cassette in mouse NPE-IE cells (GSCs) and NIH3T3 cells when delivered using rAAV2. (d) MTT viability assay demonstrating GCV dose response kill curves in NPE-IE cells transduced with rAAV carrying specified constructs.\r\n\r\nSource data\r\nExtended Data Fig.","12 Treatment of mice with AAV-SSE7-TK-IL12 results in tumour clearance and long-term protection against rechallenge.\r\n(a) Bioluminescence imaging to track tumour growth with either control (PBS) injection (right panels) or virus plus 3-week ganciclovir treatment (shown are n\u2009=\u200910 mice from the first cohort of mice in Fig. 5b). Right panel are a cohort of parallel fresh controls using the same cells in new cohort of mice, confirming tumour initiation potential. Blue font indicates the new timeline after re-challenge with fresh tumour cells injected (male mice used). (b) Dose de-escalation using IVIS to track tumour growth following treatment with different doses of AAV1-SSE7-HSV-tk-IL12.","(male mice used).\r\n\r\nSource data\r\nExtended Data Fig. 13 Characterisation of immune cells with single or double payloads.\r\n(a) Survival curves of tumour-bearing animals for either single (IL-12) or combination payloads (IL-12 plus TK\/gcv), n\u2009=\u20097 male mice per\u00a0group. Survival was analysed using Kaplan-Meier method. Statistical significance was determined using log-rank (Mantel-Cox) test with p\u2009=\u20090.0009. (b) Non-tumour bearing mice treated with AAV-SSE7-TK-IL12 versus the CMV variant. n\u2009=\u20095 male mice per group. Survival was analysed using Kaplan-Meier method. Statistical significance was determined using log-rank (Mantel-Cox) test with p\u2009=\u20090.0034. The CMV driven payloads are toxic.","(c) Bar plots of the percentage of PD-L1+ cells in MDSCs, monocytes, BMDMs, DCs and microglia.","(d, e) Bar plots of the percentage of CD62L-CD44-, CD62L-CD44+, CD62L+CD44- and CD62L+CD44+ cells in T effector cells or CD8+\u2009T cells, respectively.","(f,g) Bar plots of the percentage of IFNg producing cells in T effector cells.","(h,i) Bar plots of the percentage of IFNg producing cells in CD8+\u2009T cells. (j) Flow cytometry analysis of tumor mass 7 days following AAV1-SSE7-mCherry transduction, reveals that ~20% of the NPE-IE-GFP tumor cell population has strong mCherry expression (left panel, gating of tumor cells; middle panel, mCherry histogram; right panel, quantitation n\u2009=\u20094 independent tumours); 2ul of 3.6E14\/ml (full dose) and 7.2E13\/ml (1:5 dose). (c-i) n\u2009=\u20097 biologically independent animals per group, examined in one experiment. Data are represented as mean values +\/\u2212 SD. Statistical significance was assessed using one-way ANOVA with Tukey\u2019s multiple comparison test.","*P\u2009\u2264\u20090.05; **P\u2009\u2264\u20090.01; ***P\u2009\u2264\u20090.001; ****P\u2009\u2264\u20090.0001.\r\n\r\nSource data\r\nExtended Data Table 1 Predicted target genes and distance from enhancer fragment to transcription start site (TSS)\r\nFull size table\r\nSupplementary informationSource dataRights and permissions\r\nOpen Access This article is licensed under a Creative Commons Attribution 4.0 International License, which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence, and indicate if changes were made.","The images or other third party material in this article are included in the article\u2019s Creative Commons licence, unless indicated otherwise in a credit line to the material. If material is not included in the article\u2019s Creative Commons licence and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this licence, visit http:\/\/creativecommons.org\/licenses\/by\/4.0\/.\r\n\r\nReprints and permissions\r\nAbout this article\r\n\r\nCite this article\r\nKoeber, U., Matjusaitis, M.,","Alfazema, N. et al. Synthetic super-enhancers enable precision viral immunotherapy.\nNature (2026). https:\/\/doi.org\/10.1038\/s41586-026-10329-6\r\n\r\nDownload citation\r\n\r\nReceived: 07 December 2023\r\n\r\nAccepted: 26 February 2026\r\n\r\nPublished: 08 April 2026\r\n\r\nVersion of record: 08 April 2026\r\n\r\nDOI: https:\/\/doi.org\/10.1038\/s41586-026-10329-6"],"images":[{"src":"\/news\/images\/TkbOmERl7ZELY4B6U8bwRFLp7W9qytfUVvStryoOOlUEGhhHL0eXn3oyzuSEao4y0AJ5ESqU9aly0hqeQlF1AuY6HehvMIasaugaW2HxU4xEjpq1kWjw2a5kegGpJxC3tseg50hmbqflgA3MX8A8XkYOCKH4f.jpg","width":0,"height":0,"source":"featured","size":0}]}"
[3]=>
int(1775731824)
}
}
[2]=>
array(7) {
["file"]=>
string(28) "/app/sys/data/newsdb.inc.php"
["line"]=>
int(14)
["function"]=>
string(6) "Insert"
["class"]=>
string(3) "SQL"
["object"]=>
object(SQL)#1 (6) {
["CONNECTION":"SQL":private]=>
object(mysqli)#2 (18) {
["affected_rows"]=>
int(-1)
["client_info"]=>
string(14) "mysqlnd 8.2.30"
["client_version"]=>
int(80230)
["connect_errno"]=>
int(0)
["connect_error"]=>
NULL
["errno"]=>
int(1406)
["error"]=>
string(40) "Data too long for column 'data' at row 1"
["error_list"]=>
array(1) {
[0]=>
array(3) {
["errno"]=>
int(1406)
["sqlstate"]=>
string(5) "22001"
["error"]=>
string(40) "Data too long for column 'data' at row 1"
}
}
["field_count"]=>
int(2)
["host_info"]=>
string(18) "db:3306 via TCP/IP"
["info"]=>
NULL
["insert_id"]=>
int(0)
["server_info"]=>
string(37) "11.2.2-MariaDB-1:11.2.2+maria~ubu2204"
["server_version"]=>
int(110202)
["sqlstate"]=>
string(5) "22001"
["protocol_version"]=>
int(10)
["thread_id"]=>
int(832256)
["warning_count"]=>
int(0)
}
["URL":"SQL":private]=>
string(7) "db:3306"
["USER":"SQL":private]=>
string(8) "retrofox"
["PASS":"SQL":private]=>
string(16) "TreeakWqQPD9qnWR"
["DB":"SQL":private]=>
string(8) "retrofox"
["DROP_VALUES":"SQL":private]=>
array(1) {
[0]=>
string(0) ""
}
}
["type"]=>
string(2) "->"
["args"]=>
array(4) {
[0]=>
string(69) "REPLACE INTO news_data (id, data, timestamp) VALUES ('%s', '%s', %i);"
[1]=>
string(85) "foxapi_article_SNgSFz1UYwUgnAye8Iq6AwimFnBOT9jdyRE0kXeBSMWDBytDmFUTOmTLbAFJl2KUpiOq4O"
[2]=>
string(135713) "{"id":"SNgSFz1UYwUgnAye8Iq6AwimFnBOT9jdyRE0kXeBSMWDBytDmFUTOmTLbAFJl2KUpiOq4O","title":"Synthetic super-enhancers enable precision viral immunotherapy","description":"Cell-type-specific promoters are used in gene therapy to restrict expression of the therapeutic payload. However, these promoters often have suboptimal strength, selectivity and size. Here, leveraging recent insights into the function of enhancers, we developed synthetic super-enhancers (SSEs) by assembling functionally validated enhancer fragments into multipart arrays. Focusing on the core SOX2-driven and SOX9-driven transcriptional regulatory network in glioblastoma stem cells (GSCs)1, we engineered SSEs with robust activity and high selectivity.\u00a0Single-cell profiling, biochemical analyses and genome-binding data indicated that SSEs integrate neurodevelopmental and signalling-state transcription factors to trigger the formation of large multimeric complexes of transcription factors. Moreover, GSC-selective expression of a combination of cytotoxic (HSV-TK and ganciclovir) and immunomodulatory (IL-12) payloads, delivered using adeno-associated virus vectors, as a single treatment led to curative outcomes in a mouse model of aggressive glioblastoma. Notably, IL-12 induced an immunological memory that prevented tumour recurrence. The activity and selectivity of the adeno-associated virus and SSE were validated using primary human glioblastoma tissue and normal cortex samples. In summary, SSEs harness the unique core transcriptional programs that define the GSC phenotype and enable precision immune activation. This approach may have broader applications in other contexts when precise control of transgene expression in specific cell states is necessary. Synthetic super-enhancers enable specific delivery of anticancer payloads, achieving tumour elimination after a single dose in a mouse model of aggressive glioblastoma.","author":"Pollard, Steven M.","source":"Nature","publish":"April 8, 2026, 11:28 pm","canonical":"https:\/\/www.nature.com\/articles\/s41586-026-10329-6","text":["Main\r\nGene therapies require the targeted expression of payloads in specific cell populations to achieve appropriate therapeutic dosing and to minimize off-target effects. Various strategies to attain this selectivity can be explored, including delivery routes, capsid engineering and regulatory elements (for example, promoters, enhancers, untranslated regions and termination signals)2. Enhancers determine cell-type-specific expression by recruiting complementary transcription factors (TFs) at high density via TF-binding motifs (TFBMs)3,4.","However, natural enhancers often have suboptimal features for application in gene therapy vectors, such as size, strength, selectivity and quality of the sequence (for example, GC-rich or repetitive sequences). Strategies for creating artificial promoters and enhancers have typically focused on screening small (around 10\u2009bp) synthetic TFBMs and assembling them into concatamers5,6,7.","However, this approach is inadequate owing to the lack of a natural TFBM grammar, that is, the spacing, order, orientation and affinity of TFBMs, which are necessary for cell-type selectivity8.\r\n\r\nCell-type-specific enhancers are often clustered in the genome to create super-enhancers9,10. Super-enhancers typically regulate transcriptional programs associated with cell identity through the local clustering of low-affinity motifs8. This leads to increased high-density binding of lineage and signalling TFs to promote RNA polymerase\u2009II recruitment and high transcriptional output11,12.","We propose that synthetic super-enhancers (SSEs) can be engineered by assembling natural enhancer fragments associated with different genes into multipart arrays and then functionally screening them to identify desired activity and selectivity. The resulting transgene regulatory elements would capture the fundamental grammar of the combinations of TFs that define a specific cell type and signalling state. Such SSEs could have substantial value in gene therapy applications.\r\n\r\nWe focused on glioblastoma (GBM) to test this platform. GBM is an incurable brain cancer that is driven by cells with fetal neural stem-like phenotypes termed GBM stem cells (GSCs).","These cells express high levels of master regulatory TFs associated with neural stem and progenitor cells, including SOX2 (ref. 13). SOX2 is a pioneer TF and reprogramming factor essential for inducing and maintaining the identity of neural stem cells (NSCs)14,15 and GSCs1,16,17,18. SOX2 is also a crucial master regulatory TF that is broadly expressed across diverse GSC subtypes18.\r\n\r\nHere we develop SSEs for the targeted expression of anticancer payloads in GSCs.","A single, locally delivered dose of adeno-associated virus (AAV)-SSE-719 was evaluated in an immunocompetent model of aggressive GBM. We demonstrate that tumours are safely cleared using this precise and controlled method to express a dual payload of a cytotoxic (herpes simplex virus thymidine kinase and ganciclovir (HSV-TK\/GCV)) and a cytokine (IL-12).\r\nGSC-specific enhancer fragment design\r\nTFs bind to thousands of sites in the genome, but only a small subset are functional enhancers20.","To identify candidate GSC-selective enhancers, we first re-analysed previously published datasets1 that mapped SOX2-binding sites and used chromatin immunoprecipitation with sequencing (ChIP\u2013seq) to define those specific to GSCs and lost in their differentiated progeny.","Of the 1,710 candidate GSC-specific peaks identified (range of 115\u20133,366\u2009bp, mean of 400\u2009bp), a library of 160-bp double-stranded DNA oligonucleotides, with overlapping 100\u2009bp segments, was synthesized (Extended Data Fig. 1). The 160\u2009bp length was chosen as this was deemed to be the optimum size to capture the core set of TFBMs with the necessary natural grammar and motif diversity that might confer selectivity.","Moreover, this length is compatible with nucleosome formation for pioneer factors such as SOX2 (ref. 21). Candidate fragments were amplified by PCR, cloned into a luciferase\u2013mNeonGreen reporter plasmid and assembled into an arrayed plasmid library (48\u2009\u00d7\u200996-well plates with 4,579 individual plasmids) (Fig. 1a).\r\nFig. 1: Functional enhancer screening of an arrayed plasmid library.\r\na, Strategy for enhancer fragment (160\u2009bp) functional screening using an arrayed library of 4,579 plasmids with luciferase (Luc;\u00a0yellow) and mNeonGreen (mNGreen; green) reporters (single mRNA linked by P2A; grey), with an mCMV promoter (light blue).","b, Results for the enhancer fragment screen in GSCs (G7 cells) (n\u2009=\u20091).","c, Validation of 135 hits from the primary screen using independent luciferase reporter assays (NanoLuc DLR assay) (n\u2009=\u20093 biological replicates, error bars represent the s.d. of the arithmetic mean). The inset shows the enriched SOX dimer motif identified using the MEME tool.","d,\u00a0Validation of the top 16 significant hits from c, using flow cytometry to determine the percentage of mNeonGreen reporter-positive cells in an independent GSC line (E28), and the MFI per cell for each fragment (right). mCMV and the full-length CMV promoter were used as negative and positive reference controls, respectively (each dot represents a biological replicate (n\u2009=\u20093), error bars represent the s.d. of the arithmetic mean). Ordinary one-way analysis of variance (ANOVA) with Dunnet\u2019s multiple comparison test against mCMV was performed. For the mNeonGreen (%) data, adjusted P\u00a0< 0.0001 for ID3146,\u00a0P\u2009=\u20090.0018 for ID1101\u00a0and\u00a0P\u00a0=\u00a00.0003 for ID2339.","In the MFI mNeonGreen panel, P\u00a0= 0.0171 for ID2706, P\u00a0= 0.0092 for ID3146,\u00a0P\u2009=\u20090.0038 for ID1101, P\u2009=\u20090.0034 for ID2339 and\u00a0P\u2009=\u20090.0037 for ID2904.","e, Negative control normal human dermal fibroblasts (huFib70) using the same assays as d (each dot represents a biological replicate (n\u2009=\u20093), error bars represent the s.d. of the arithmetic mean).\u00a0*P\u2009\u2264\u20090.05, **P\u2009\u2264\u20090.01, ***P\u2009\u2264\u20090.001, ****P\u2009\u2264\u20090.0001.\r\n\r\nSource data\r\nDiscovery of functional GSC enhancers\r\nPatient-derived GSCs can be routinely propagated in a monolayer\u00a0and can be efficiently transfected13. We screened the enhancer fragment plasmid library in GSCs (specifically GSC7 cells) using a luciferase assay to identify functional hits22. From this initial screen, 135 plasmids exhibited more than 10-fold activity over a minimal CMV (mCMV) promoter (Fig. 1b).","Several hits were identified recurrently, which provided confidence in the screening assay.\r\n\r\nSanger sequencing confirmed sequences of the 135 initial hits, and 32 unique sequences were validated in triplicate with >20-fold activity compared with mCMV (Fig.","1c, Extended Data Table 1 and Supplementary Table 1). The top 16 candidates were also validated using an independent flow cytometry assay to detect the mNeonGreen transcriptional reporter, which provides single-cell quantification of activity and mean fluorescence intensity (MFI). All 16 validated hits were active in an independent patient-derived GSC line (E28), but were inactive in fibroblasts and HEK293 cells, which were used as SOX2-negative controls (Fig.","1d,e and Extended Data Fig.","1c,d). As anticipated, these hit sequences contained TFBMs for many known neurodevelopmental or GBM-associated TF families, including motifs for SOX, bHLH, FOX, TCF and SMAD (Extended Data Fig. 1e).\r\n\r\nNotably, de novo motif searches using the MEME tool revealed that the 32 active fragments contained a specific SOX dimer motif that resembles a previously reported SOX9 motif23 (Fig. 1c). Although SOX2 typically operates as a monomer, SOX9\u2014a known NSC self-renewal factor24\u2014can function as a homodimer23.","This result suggests that the functional enhancer fragments potentially bind SOX9.\r\n\r\nTo functionally validate the importance of this motif in the enhancer fragments, we created a series of mutants in the dimer motif of ID2904 (which was chosen because it only has a single motif) with nucleotide substitutions, inversions or changes in spacing. Each of these alterations abolished the transcriptional activity of the fragment (Extended Data Fig. 2a\u2013f). Higher activity was observed when the distance between the half-sites was reduced by 2\u2009bp. This result suggests that dimer binding is a crucial feature of the activity of the motif and is probably important for SOX9 binding.","Indeed, using biochemical pull-down, we confirmed that SOX9 binds to ID2904 and that binding depends on the SOX dimer motif (Extended Data Fig. 2g).\r\n\r\nThe above findings led us to speculate that SOX2 and SOX9 might work together at shared target enhancers in GSCs. To find their putative shared target sites, we mapped genome-wide binding of SOX2 and SOX9 in seven independent GSC patient-derived cell lines using ChIP\u2013seq (Fig. 2a). Most SOX9 peaks co-bound with SOX2, and these overlapped with previously reported super-enhancers found across GSCs (Fig. 2b\u2013d). These co-bound peaks were located near genes involved in stem cell function (Fig.","2e), including essential regulators of neural stem and progenitor cells (for example, CDK6 and PTPRZ1; Fig. 2h). The expression of SOX2 and SOX9 was positively correlated, particularly in GBM (relative to most other cancers), which implied that these two factors have a disease-specific function and\/or regulatory interactions (Fig. 2f).","Notably, the SOX2 and SOX9 peaks in co-bound sites at enhancers were enriched with multiple SOX monomer and dimer motifs, respectively, with an optimal dimer motif spacing of 2\u20133\u2009bp (Fig.","2g,i). Altogether, these data indicate a specific requirement for SOX9 dimerization at loci in which SOX2 and SOX9 are co-bound.\r\nFig. 2: ChIP\u2013seq analysis of SOX2 and SOX9 binding in seven different patient-derived GSC cultures.\r\na, Schematic of SOX2 and SOX9 ChIP\u2013seq experimental design and analysis.\u00a0GO, gene ontology.","b, Venn diagram showing the overlap between SOX2 and SOX9 peaks for the sample E28 (left), and heatmaps of SOX2 and SOX9 binding intensity over SOX9 and SOX2 peaks, respectively (right), demonstrating co-binding events.","c, Bar plots showing the proportion of SOX9 peaks located in co-bound sites.","d, Venn diagrams showing the overlaps between consensus SOX2 and SOX9 peaks and co-bound SOX2\u2013SOX9 peaks with a consensus set of GSC super-enhancers (n\u2009=\u200944 publicly available GSC H3K27ac datasets;\u00a0Methods).","e, Gene ontology analysis of co-bound SOX2\u2013SOX9 peaks in super-enhancers.","f, Bar plot showing the distribution of SOX2\u2013SOX9 expression correlation values (y\u2009axis) across diverse cancer types from The Cancer Genome Atlas (x\u2009axis).","g, Bar plot showing the motif co-occurrences of inverted palindromic SOX motifs, grouped by the distances between the primary and secondary SOX motif pairings.","h, UCSC genome browser coverage tracks of SOX2 and SOX9 ChIP\u2013seq in the seven individual GSCs at PTPRZ1 and CDK6 loci, annotated by the presence of GSC super-enhancers, co-bound SOX2\u2013SOX9 peaks and the functional SOX dimer motif.\u00a0cCREs, candidate cis-regulatory elements; Cons 100 Verts, conservation across 100 vertebrates.","i, The distributions of centrally enriched de novo SOX monomer and dimer motifs at SOX2 and SOX9 peaks, respectively, restricted to those in co-bound SOX2\u2013SOX9 enhancers.\r\n\r\nIn summary, using functional screening in patient-derived GSCs, we identified candidate 160\u2009bp SOX2 and SOX9 co-bound enhancer fragments with high selectivity but low activity (relative to CMV). Together with previously published findings, such as genome-wide CRISPR screens18, these results highlight the importance of SOX2 and SOX9 as core components of a TF circuit that determines and sustains GSC identity.","Moreover, they share crucial transcriptional targets.\r\nSSEs for GBM\r\nOur goal was to create a GSC-selective transcriptional switch for cancer gene therapy that combines strong expression and high selectivity. Such a SSE would provide the necessary therapeutic window to target the tumour but spare surrounding normal CNS tissue. We proposed that combining functional and selective enhancer fragments into a larger artificial multipart assembly might increase activity by facilitating increased high-density recruitment of SOX2 and SOX9 alongside other signalling-associated transcriptional cofactors to mimic the mechanism of action of natural super-enhancers11,12.","We chose four-part assemblies, as this represents an optimal size compatible with AAV vector applications. These assemblies were constructed using sequences from the top\u200916 validated hits grouped into four sets: 1\u20134, 5\u20138, 9\u201312 and 13\u201316 (Fig.","3a,c).\r\nFig. 3: Combining SOX2 functional enhancer fragments results in synergistic increases in activity without compromising selectivity.\r\na, Activity of the 16 top hits relative to mCMV and CMV assessed by flow cytometry in GSCs (G7 cells) (each dot represents a biological replicate, error\u00a0bars represent the s.d. of the mean).","b, SSE activity detected using an mNeonGreen reporter across GSC lines (GSC7 and E28) and control fibroblasts. Error bars represent the s.d. of the mean. One-way ANOVA with Dunnett\u2019s multiple comparisons test against mCMV, adjusted P\u2009values for G7: P\u2009=\u20090.0001 for SSE-1, P\u2009=\u20090.0009 for SSE-3, P\u2009=\u20090.0003 for SSE-5 and P\u2009<\u20090.0001 for SSE-7. Adjusted P\u2009values for E28: P\u2009=\u20090.0099 for SSE-1, P\u2009=\u20090.0048 for SSE-3, P\u2009=\u20090.0075 for SSE-5 and P\u2009=\u20090.0008 for SSE-7.","UT, untreated.","c, Schematic of each SSE-1, SSE-3, SSE-5 and SSE-7 transgene design.","d, EMSA of SSE-7 DNA following incubation with GSC lysates compared with differentiated GSCs (representative image; n\u2009=\u20093).","e, EMSA showing SSE-7 DNA with fibroblast (hFib) lysate (left) and hFib with SOX2 and SOX9 expressed individually (middle) or in combination (right). Representative images shown (n\u2009=\u20093).","f, EMSA of SSE-7 DNA (left, representative images of n\u2009=\u20093) and its constituent fragments (right, n\u2009=\u20091), incubated with SOX2 and SOX9. Schematic of SOX motifs (thin black bar, partial SOX motif; thicker, SOX motif; thickest, dimer motif) is shown underneath the images.","g, SABER\u2013FISH analysis of transgenes. Representative images are shown (left), with quantification (right).","For SSE-7, mean copy number\u2009=\u2009167 (median, 119; Q1\u2013Q3, 76\u2013209; min\u2013max, 8\u2013806).","For CMV, mean copy number\u2009=\u2009115 (median, 87; Q1\u2013Q3, 57\u2013127; min\u2013max, 5\u2013761). Scale bars, 20\u00a0\u03bcm.","h, scRNA-seq analysis shows no correlation between SSE-7 activity transcriptional subtypes (AC, astrocyte; MES, mesenchymal; NPC, neural progenitor cell).\u00a0Neftel subtype, subtype as identified by Neftel et al.27. Distribution of SSE-7\u2013mCherryhigh cells using UMAP plots of scRNA-seq analysis in E21 cells (n\u2009=\u200923,384 cells) and E28 cells (n\u2009=\u200935,879).","i, SCENIC analysis to determine TF enrichment across replicate libraries. Combined Z\u2009scores from each library. *P\u2009\u2264\u20090.05, **P\u2009\u2264\u20090.01, ***P\u2009\u2264\u20090.001; NS, not significant. P\u2009values were calculated using repeated-measures ANOVA, followed by Bonferroni-adjusted pairwise comparisons of estimated marginal means. Boxes show the interquartile range, centre lines the median, and whiskers the minimum and maximum values.\r\n\r\nSource data\r\n\r\nEach variant displayed strong activity that matched or exceeded that of a full-length CMV promoter in GSCs. By contrast, the variants had low activity in both fibroblasts and HEK293 cells, such that levels were comparable to mCMV-like levels (Fig. 3b and Extended Data Fig. 4a).","Given the principles that underpin their design, their high activity (switch-like behaviour) and their high cell-type selectivity, we herein refer to these artificial elements as SSEs.\r\n\r\nTo initially explore selectivity in glial cell lineage differentiation, we evaluated SSE activity during GSC differentiation into astrocyte-like cells1,13. We observed a twofold reduction in MFI per cell within 24\u2009h, followed by an approximately tenfold decrease by day\u200910 of differentiation (Extended Data Fig.","4b,c). By contrast, the constitutive CMV promoter displayed high activity across all conditions. SSE activity is therefore highest in the immature stem cell-like state of GSCs and is rapidly reduced as cells differentiate.\r\n\r\nFocusing on SSE-7, we next tested a larger set of patient-derived GSCs that span diverse genetic and transcriptional subtypes25,26. SSE-7 was active across all seven patient-derived GSCs but had reduced levels in E37 (an atypical cell line with MYC amplifications and reduced SOX2 and SOX9 levels) (Extended Data Fig. 3).","However, there was variability across the cell lines in the percentage of positive cells in the population compared with CMV, which indicated heterogeneity in the cultures. This result was not explained by SOX2 or SOX9 protein levels, as these were similar across lines, which suggests that additional cooperating TFs might be necessary for SSE-7 activity.\r\n\r\nWe next explored the biochemical mechanism of the switch-like behaviour and selectivity of SSE-7 and the roles of SOX2 and SOX9 in this process.","To that end, we first used cell-free electrophoretic mobility shift assays (EMSAs) to analyse their interactions with SSE-7 alone and in combination using both cell lysates and cell-free recombinant SOX proteins.\r\n\r\nSSE-7 incubation with GSC cell extracts resulted in a major super-shift and the formation of large higher-order multimeric complexes, but this was not observed in GSCs that had differentiated into astrocytes (Fig. 3d). Using human fibroblasts with SOX2 and\/or SOX9 forced expression, we also observed a clear super-shifted band only when both factors were incubated together with SSE-7, but this effect was weaker compared with GSCs. This result highlights the importance of GSC-associated cofactors for triggering SSE activity (Fig.","3e).\r\n\r\nNext, to validate whether synergy exists with SOX2 and SOX9 co-binding, we used purified recombinant SOX2 and SOX9 proteins and tested their binding. When SOX2 alone was bound with SSE-7, it displayed a typical ladder-like banding pattern indicative of SOX2 binding individually at each motif in the 640-bp-long SSE-7 (Fig. 3f and Extended Data Fig. 4d\u2013f).","Notably, the binding of SOX9 led to the formation of the higher-order complexes, and this effect was enhanced in the presence of SOX2 (Fig. 3f). The analysis of each constituent 160\u2009bp enhancer fragment in isolation demonstrated that the formation of higher-order complexes depends on the architecture of the SSE (Fig. 3f). No higher-order complexes were formed when using a control 40\u2009bp FGF4 enhancer that contains a single SOX motif (Extended Data Fig. 4g).","We therefore conclude that a key feature of the mechanism of action of SSE-7 is its ability to induce higher-order TF complexes through high-density local recruitment of SOX2 and SOX9.\r\nMechanistic basis of SSE activity in GSCs\r\nWe next aimed to identify candidate SOX-cooperating TFs and the cofactors required for the SSE switch-like transcriptional activation seen in GSCs. To that end, we performed single-cell RNA sequencing (scRNA-seq) analysis of 2,711 SSE-7\u2013mCherry-positive cells across 8 different GSC lines, comparing these to matched mCherry-negative populations.","We first used the single-molecule imaging technique signal amplification by exchange reaction and fluorescence in situ hybridization (SABER\u2013FISH) and confirmed that transduction rates and nuclear transgene\u00a0copy number were similar among the different cell lines and that there was no correlation with mCherry levels (Fig. 3g).","Therefore, differential transduction or nuclear processing of the AAV does not explain the restricted SSE-7 activity observed in heterogeneous GSC cultures compared with CMV.\r\n\r\nUniform manifold approximation and projection (UMAP) analysis confirmed that these GSC cultures have, as expected, a diverse range of transcriptional subtypes or states27. This transcriptional heterogeneity mirrored that seen in patient tumour samples and reflected distinct differentiation states (Fig. 3h). We did not find a simple correlation of SSE-7high activity with previously reported transcriptional subtypes (Fig.","3h), SOX2 and SOX9 expression or cell cycle phase (Extended Data Figs. 5 and\u00a06a).\r\n\r\nTo perform an unbiased search for TF activities enriched in SSE-7high cells, we used the analysis tool SCENIC. The results revealed a clear correlation of multiple TF activities linked to cancer signalling pathways, including STAT, IRF, SMAD and FOS\u2013JUN (Fig. 3i). Such signalling end-point TFs are often highly activated in cancers, including GBM, and motifs for such TF families are present in SSE-7. This finding is consistent with the idea that these TFs\u00a0have a potential functional role (Fig. 3i and Extended Data Fig.","6b,c). These data suggest that SSE-7 is activated in cells that display a combination of activated signalling end-point TFs and an immature stem cell-like identity driven by SOX2 and SOX9.\r\n\r\nWe next explored the requirements needed for signalling pathway activation for SSE-7 function. We performed a screen of 160 different small-molecule kinase inhibitors, spanning 70 distinct families, to identify those that would reduce SSE-7 activity (Extended Data Fig. 7a). Given the diverse signalling states across GSCs and the results from the SCENIC analysis, we speculated that activity would be highly dependent on the cell line.","We therefore focused the screen on E55 and E31, two GSC lines with typical phenotypic heterogeneity and distinct\u00a0transcriptional states.\r\n\r\nInitial hits were validated, and a clear dose-dependent reduction in mCherry MFI was confirmed for four of these hits (Extended Data Fig.","7b,c). These four inhibitors all operate in the MAPK\u2013ERK signalling pathway: three MEK inhibitors (PD0325901, AS-703026 and PD184161) and one ERK inhibitor (SC-1) (Extended Data Fig. 7d). For these specific GSC lines (E55 and E31), SOX2 and SOX9 activity together with hyperactive MAPK signalling seemed to be necessary for SSE-7 activity.","Finally, a subpopulation of sorted SSE-7-negative GSCs could reactivate SSE-7 expression at later time points. This result is consistent with the finding that SSE activity depends on a specific signalling state (Extended Data Fig. 7e).\r\n\r\nAltogether, these functional genetic, biochemical and genomic data indicate that SSE-7 is a dynamic transcriptional regulatory element that is strongly activated in specific GSC states that have a combination of both SOX2 and SOX9 expression and hyperactive signalling-associated TFs. The specific signalling TFs involved are likely to vary among the heterogeneous GSC genetic and transcriptional subtypes.","However, our SSE design strategy (that is, mixing distinct enhancer fragments from diverse genes) enabled us to achieve robust and strong expression across GSCs\u00a0regardless of this variability.\r\nCell-state-specific activation of SSE-7\r\nOur functional enhancer fragments typically had high vertebrate evolutionary conservation, which suggested that SSEs may work in different species. To further explore selectivity across diverse tissues and organs, we first tested SSE activity in developing zebrafish (around 48\u2009h after fertilization) (Extended Data Fig. 8).","Overlapping expression domains for each SSE-driven eGFP reporter was observed for all four SSEs tested, with consistently high levels of expression in the optic placodes and a subset of forebrain and spinal cord neural progenitors (Extended Data Fig.","8a,b). This restricted developmental expression pattern was consistent with previously reported zebrafish Sox2 and Sox9 expression27,28, and confirms that not only are the SSEs highly active but they are also highly tissue restricted.","Notably, SSE-1, SSE-3 and SSE-5 had an additional weaker expression domain in more posterior CNS tissues, whereas for SSE-3, we detected expression in the posterior endoderm (Extended Data Fig. 8b). We therefore prioritized SSE-7 for subsequent experiments.\r\n\r\nWe created a stable transgenic zebrafish line and used this to look at later larval stages in the nervous system (Extended Data Fig. 8c). SSE-7 activity was nonoverlapping with mature neuronal markers, which confirmed that SSE-7 is not expressed in neurons and is only seen in a subset of fetal neural\u00a0progenitor cells (Extended Data Fig. 8d\u2013g).","Notably, when we forced the expression of a potent oncogenic form of Akt in zebrafish, we observed that SSE-7 activity increased in those cells. This result is consistent with the requirement of hyperactive signalling together with Sox expression to support SSE activation, as seen in human GSCs (Extended Data Fig. 8h). SSE-7 had low activity in oligodendrocyte progenitor cells (OPCs) derived from human induced pluripotent stem (iPS) cells relative to CMV (Extended Data Fig.","8i),\u00a0but was expressed in human fetal NSC cultures (Extended Data Fig. 8j).","Altogether, these data suggest that SSE-7 activity is highly restricted to specific\u00a0cell states and may be most active in a subset of fetal neural progenitors.\r\nSSE-7 activity in GBM core and margin\r\nAAV is an excellent gene therapy vector for use in GBM owing to the following characteristics: (1) its small physical size (for enhanced distribution); (2) persistence in quiescent cancer cells; (3) low immunogenicity; (4) diverse natural capsids that might enable improved transduction; (5) validated efficient brain delivery via convection-enhanced delivery; and (6) well-understood manufacturing and clinical safety profile.\r\n\r\nTo further explore SSE-7 selectivity using relevant adult human neural cell types and the effectiveness of AAV as the delivery vector, we generated human glutamatergic and GABAergic differentiated neurons (ioGlutamatergic and ioGABAergic, respectively), and microglia (ioMicroglia) from iPS cells, and tested AAV1-SSE-7\u2013mCherry activity (Fig.","4a). SSE-7 was strongly expressed in GSCs within a few days of transduction, with similar levels to the CMV positive control.","However, in postmitotic neurons, we did not observe SSE-7 activity (Fig. 4a) at either day\u20093 or day\u200910. For microglia, there was poor transduction with AAV1.\r\nFig. 4: GSC selectivity of SSE-7 confirmed in vitro and in tissue slice cultures.\r\na, AAV1-SSE-7\u2013mCherry reporter activity in GSCs and human iPS-cell-derived differentiated neurons and microglia assessed by flow cytometry at day\u20093 (left) and day\u200910 (right) following transduction. For day\u20093 for ioGABAergic neurons and ioMicroglia, n\u2009=\u20093 biological independent cell experiments; ioGlutamatergic neurons and GSCs, n\u2009=\u20094 biological independent cell experiments.","For day\u200910, n\u2009=\u20093 biological independent cell experiments excluding ioGABAergic neurons, for which n\u2009=\u20092. Data are the mean\u2009\u00b1\u2009s.e.m., and statistical analysis was performed using an unpaired t-test. **P\u2009\u2264\u20090.01, ***P\u2009\u2264\u20090.001.","b, Experimental schematic of AAV1-SSE-7 to test selectivity in fresh human GBM tumour mass and tumour margin tissue samples (infiltrating cortex) processed into slice cultures (Methods).","c, Immunostaining and confocal images of tissue for mCherry (red) and GSC markers (Nestin and SOX2) demonstrates GSC selectivity of SSE-7 relative to CMV. Tumour mass and adjacent matched normal brain tissue (tumour margin) are shown.","d, Percentage of mCherry-expressing cells co-expressing SOX2. Each point represents a biological replicate, matched tumour and tumour margin samples are indicated by matching symbols. n\u2009=\u20093 for margin and n\u2009=\u20094 for the tumour mass. Statistical analysis was performed using two-way ANOVA with multiple-comparisons correction (P\u2009=\u20090.0436). Scale bars, 10\u2009\u00b5m. Schematic in b created in BioRender.","Pollard,\u2009S.\u00a0M.\u00a0(2026) https:\/\/BioRender.com\/l774d5t.\r\n\r\nSource data\r\n\r\nTo directly test whether SSE-7 would have strong differential activity in GBM tumour tissue compared with the surrounding normal brain, we determined the transduction and expression of AAV1-SSE-7\u2013mCherry in ex vivo fresh tumour tissue slices obtained from patients with GBM (Fig. 4b) and compared these with the tumour margin (which is macroscopically normal tissue but contains infiltrating GBM cells). We first tested whether AAV1-SSE-7 would be active in the tissue, focusing on the mCherry reporter only.","After 7\u2009days of incubation with a single dose of the virus, we fixed and stained the tumour tissue and compared CMV and SSE-7 activity using the mCherry reporter.\r\n\r\nThere were significantly fewer mCherry-positive cells when using SSE-7 compared with CMV (Fig. 4c and Extended Data Fig. 9b) in the tumour, and this was even more significant in the tumour margin tissue. This result is consistent with SSE-7 labelling a subset of GBM cells but not normal healthy tissue. We performed this analysis for four independent patient samples\u2014grade\u20094 IDH mutant astrocytoma, IDH wild-type GBM, a grade\u20093 astrocytoma and recurrent GBM (Extended Data Fig. 9a)\u2014and obtained similar results.","By contrast, as expected, CMV induced widespread high expression across diverse normal CNS cell types, including cells with neuronal and endothelial morphology (Fig. 4c and Extended Data Fig. 9).\r\n\r\nTo next confirm whether mCherry is enriched in the tumour cell population, we performed and quantified SOX2 co-staining. This experiment revealed significant differences between CMV and SSE-7, with SSE-7 activated in around 90% of mCherry-positive tumour cells that co-expressed SOX2 (Fig. 4d). We conclude that SSE-7 transgene expression is differentially active in tumour versus normal tissue (tumour margin) compared with CMV.","Moreover, SSE-7 provides a strong and selective gene switch that can restrict transgene activity to GSCs in the tumour mass and the infiltrating tumour margin.\r\nComplete responses in a mouse model of GBM\r\nWe next determined whether AAV1-SSE-7 could be used to selectively kill GSCs using expression of the HSV-TK\/GCV enzyme prodrug system, which has previously been applied in the clinic as a gene-directed enzyme prodrug therapy29. SSE-7 retained activity and provided selective expression of the HSV-TK payload in GSCs (Extended Data Fig. 10a\u2013d). This enabled cytotoxic killing of GSCs, after GCV treatment, whereas fibroblasts were unaffected (Extended Data Fig. 10e\u2013g).","Screening of common natural AAV serotypes confirmed that AAV1 is the optimal serotype of choice, as it efficiently transduced all seven patient-derived GSCs with high efficiency (Extended Data Fig.","11a,b).\r\n\r\nDespite the promise of HSV-TK as a suicide gene therapy approach for oncology, there has been limited success in the clinic. This is partially explained by the challenges of efficient delivery to all cancer cells30. As immune payloads can act non-autonomously to trigger an immune response against residual disease, we reasoned that combining an immune payload with a cytotoxic payload might be required to achieve this goal.","Indeed, such a strategy could induce multiple pathways of the immune response to polarize local macrophages in the tumour microenvironment while enabling systemic adaptive responses by T\u2009cells. We proposed that this combination\u2014tumour cell killing, tumour-associated macrophage polarization and T\u2009cell education and activation\u2014would be potent and reduce chances of tumour regrowth from residual disease30.\r\n\r\nWe chose the cytokine IL-12 for evaluation as it triggers a potent inflammatory response (stimulating adaptive and innate immunity) and is effective in promoting anticancer immune responses31,32. We reasoned that by using the AAV1-SSE-7 platform, we could avoid the toxicities previously seen when using systemic delivery31.","This would enable controlled release of IL-12 alongside tumour killing and induce effective antitumour immune responses, with IL-12 levels being reduced as tumour cells are lost. We designed an AAV1 vector with SSE-7 driving the expression of a single mRNA encoding HSV-TK linked via the P2A sequence to the mouse IL-12 heterodimer fusion (Il12a and Il12b, which encode P40 and P35, respectively) (Fig. 5a). This dual payload design ensures that the cytokine and cytotoxic effects are synchronized and serves as a safety switch to prevent uncontrolled IL-12 levels.\r\nFig.","5: The combination of a cytotoxic and immunomodulatory payload is highly effective at clearing established GBM tumours.\r\na, Architecture of the AAV1 transgene used to treat NPE-IE GBM in mice.","ITR, inverted terminal repeat;\u00a0mCMV, minimal CMV promoter;\u00a0mIL-12, mouse IL-12.","b, Survival curves for the mice treated with AAV1-SSE-7\u2013HSV-TK\u2013IL-12 (n\u2009=\u200924) versus control (UT (PBS); n\u2009=\u200922). P\u2009<\u20090.0001 for pooled (n\u2009=\u20093) independent cohorts. Survival was analysed using the\u00a0Kaplan\u2013Meier method. Significance was determined using a\u00a0log-rank (Mantel\u2013Cox) test. Mice were selected at week\u20091 using bioluminescence imaging to ensure a similar starting size for tumours across the cohorts.","c, Longitudinal tracking of tumour growth in mice once a week using bioluminescence imaging (three examples shown for experimental and control; see Extended Data Fig. 12 for the larger cohort).","d,e, Bar plots of the abundance of total T\u2009(CD3)\u00a0cells, effector T (Teff) cells, CD8+ T\u2009cells, regulatory T (Treg) cells and natural killer (NK) cells (d), and myeloid-derived suppressor cells (MDSCs), monocytes, bone marrow-derived macrophages (BMDMs), dendritic cells (DCs) and microglia (e) out of total live CD45+ cells in the tumour microenvironment of untreated mice and mice treated with a single dose of AAV-SSE-7-HSV-TK\u2013IL-12 or with the AAV-SSE-7-HSV-TK\u2013IL-12\u2009+\u2009GCV combination, determined by flow cytometry.","f,g, Bar plots of the abundance of PD1\u2212Ki67\u2212, PD1\u2212Ki67+, PD1+Ki67\u2212 and PD1+Ki67+ cells from Teff cells (f) and from CD8+ T\u2009cells (g) in the tumour microenvironment of untreated mice and mice treated with a single dose of AAV-SSE-7-HSV-TK\u2013IL-12 or with the AAV-SSE-7-HSV-TK-IL-12\u2009+\u2009GCV combination, determined by flow cytometry. n\u2009=\u20097 biologically independent animals per group, examined in one experiment. Data are the mean\u2009\u00b1\u2009s.d. Significance was assessed using one-way ANOVA with Tukey\u2019s multiple comparison test.","*P\u2009\u2264\u20090.05, **P\u2009\u2264\u20090.01, ***P\u2009\u2264\u20090.001, ****P\u2009\u2264\u20090.0001.\r\n\r\nSource data\r\n\r\nFor in vivo preclinical efficacy studies, we used our previously reported disease-relevant syngeneic mouse model of GBM, termed NPE-IE25, which exhibits a myelosuppressive tumour microenvironment and transcriptional programs seen in human GBMs25. AAV1-SSE-7 transduction and SSE activity were first validated using mouse NPE-IE cells in vitro (Extended Data Fig.","11c,d).\r\n\r\nAAV1-SSE-7\u2013HSV-TK\u2013P2A-IL-12 was delivered to a cohort of mice with established orthotopic tumours (14\u2009days after tumour cell orthotopic engraftment) using direct intratumoral injection with a microinjection pump to deliver a single dose of the virus. We observed substantial tumour regression within 1\u20132\u2009weeks and complete clearance of tumours in 20\u201324 treated mice over the subsequent 2\u20133\u2009weeks, whereas control mice died within 1\u20132\u2009weeks (Fig.","5b,c).\r\n\r\nUsing bioluminescence imaging to track tumour burden in live mice, no further regrowth of tumours was observed after the initial tumours had cleared (Extended Data Fig. 12a).","Moreover, no prominent toxicity, scored by weight loss and neurological monitoring, was seen over the subsequent 11\u2009months (the experimental end point). We re-challenged treated mice (n\u2009=\u200910 that cleared the tumour) with fresh NPE-IE cell transplants into the striatum (5\u2009months after initiation of the first therapy).","Notably, there was no detectable tumour formation in any of the mice after the re-challenge, whereas parallel control mice formed tumours within 2\u20133\u2009weeks (Extended Data Fig. 12a).","Finally, antitumour activity was dose-dependent (Extended Data Fig. 12b).\r\nIL-12 with HSV-TK\/GCV drives antitumour immunity\r\nTo determine whether IL-12 alone would be sufficient for eradicating tumours without the concomitant HSV-TK\/GCV-induced killing (that is, single versus dual payloads, respectively), we treated mice with AAV1-SSE-7-HSV-TK\u2013P2A-IL-12 alone or with this virus and with GCV. The dual payload was the most effective and enhanced the effects of IL-12 alone (Extended Data Fig. 13a). To explore safety in normal CNS, we also assessed responses in mice without tumours to either CMV-driven or SSE-7-driven HSV-TK\u2013IL12.","As expected for CMV, there were rapid neurological issues within weeks, consistent with the expected toxicities with continuous expression of IL-12 in normal CNS. By contrast, we observed no toxicity with SSE-7, a result consistent with our earlier in vitro data confirming the selectivity of this SSE (Extended Data Fig. 13b).\r\n\r\nCharacterization of the immune cell composition after treatment using either IL-12 alone or IL-12 with HSV-TK\/GCV revealed that the latter performed better, with significantly increased levels of IFN\u03b3 and a trend towards increased TNF (Extended Data Fig. 13c\u2013i).","This finding is consistent with the observed increase in cytotoxic T\u2009cells and reduction in myelosuppressive cells that occurs through T helper\u20091-like responses induced by IL-12 (Fig. 5d\u2013g).\r\n\r\nThere was also an increase in proliferative Ki67+ T\u2009cells, with a shift towards increased CD44+, a marker of T\u2009cell activation and memory (Extended Data Fig.","13d,e). To assess the percentage of transduction in the tumour mass, we used the SSE-7\u2013mCherry reporter. Tumours were collected, dissociated and assessed using flow cytometry, which revealed around 20% positive tumour cells (GFP+mCherry+) (Extended Data Fig. 13j). This transduction efficiency was achieved using the same viral doses as the survival analysis experiments. This result confirms that only a fraction of the tumour mass needs to be transduced to achieve immune clearance of residual disease.\r\n\r\nIn conclusion, AAV1-SSE-7\u2013HSV-TK\u2013IL-12 is highly effective at eliminating tumours in an aggressive syngeneic\u00a0mouse GBM\u00a0model. Treated mice were resistant to subsequent re-challenge with fresh NPE-IE cells.","A single dose of AAV1-SSE-7\u2013HSV-TK\u2013IL-12 induced complete responses with durable immune memory without off-target toxicity.\r\nDiscussion\r\nThe development of effective gene therapies for oncology will require more effective payload combinations and more selective delivery mechanisms than has been possible so far30,33,34. We showed that SSEs, combined with delivery via AAV1, can address one of the significant challenges in oncology: namely, how to specifically deliver combinations of anticancer payloads at high doses without off-target toxicity.","Across genetically diverse GSC lines and high-grade glioma tissues, SSE-7 consistently read out a core transcriptional program characteristic of GSC identity: the immature NSC-like state.\r\n\r\nOur findings indicated that SOX2 and SOX9 share crucial target genes and may represent a core transcriptional module, an \u2018oncofetal\u2019 program, that defines aggressive tumour-initiating cells in GBM18,35. Using AAV-mediated delivery and SSE control of a transgene, we achieved specific elimination of active GSCs by targeting them from within, akin to a Trojan horse approach.","The persistence of AAV vectors, and our focus on the shared SOX circuits, meant that we were able to guard against plasticity, as transgene expression will be reactivated if cells transit through this specific cell state.\r\n\r\nImmunotherapy approaches have generally had limited success in GBM36,37,38. Our AAV\u2013SSE approach will now require careful safety studies to determine the potential for translation to patients.","The goal is to deliver cytotoxic payloads alongside innate and adaptive immune activation to achieve meaningful and durable suppression of high-grade gliomas without significant off-target toxicities.\r\n\r\nFuture studies will explore the molecular mechanisms that underlie SSE activity, which may involve transcriptional hubs or condensates. Our findings from exploring the biochemical binding of SOX2 and SOX9 suggests that there is clear synergy between these two factors.","However, additional cofactors are required for the creation of a multimeric complex. Detailed structural studies are needed to precisely determine how SOX2 and SOX9 operate and the signalling TF requirements and whether nucleosome context is critical39. In the future, improved computational methods and artificial-intelligence-based tools for the design of enhancers and synthetic associated motifs will complement the functional screening approaches used here40,41,42.","In conclusion, our study demonstrated the potential of SSEs and AAV vectors to deliver precision immune activation for the treatment of cancer.\r\nMethodsDestination vector cloning\r\nWe built a specific custom destination vector for efficient Golden Gate cloning of enhancer fragments. This included the reporter gene cassette NanoLuc-Ires-mNGreen-pA, downstream of a mCMV promoter. A bacterial suicide ccdB cassette spanning the enhancer position enabled assembly of a single enhancer for selection and efficient cloning. This was based on the EMMA Golden Gate cloning system43.","For the GSC long-term differentiation experiments, the same vector was built but included PiggyBac transposase recognition sites flanking the entire cassette.\r\nBioinformatics to design a SOX2 enhancer oligonucleotide pool\r\nWe re-analysed previously published1 SOX2 ChIP\u2013seq and H3K27ac GSC cell line data to identify GSC-specific SOX2 peaks that were overlapping with H3K27ac and absent in differentiated cells (cells in serum culture). The resulting shared peaks were then combined into one set and manually curated to remove centromeres. This resulted in 1,721 peaks with an average length of 402\u2009bp. These peaks were then used to design a set of 160\u2009bp sequences.","Next, 20\u2009bp adapters were included to flank the sequences and these sequences were synthesized as an oligonucleotide pool (Twist Bio).\r\nConstruction of an arrayed plasmid library of enhancer fragments\r\nThe oligonucleotide pool (9,523 unique sequences) was first amplified for 10 cycles with 0.25\u2009\u00b5l (2.5\u2009ng input) volume, and 0.5\u2009\u00b5l of this reaction was used for the subsequent 15 cycles of amplification to reduce PCR \u2018jackpot\u2019 amplification. The final products were cloned into the expression vector using an efficient Golden Gate reaction. We used KAPA HiFi Hotstart polymerase with GC buffer (Roche, KK2501), 68\u2009\u00b0C annealing temperature and 5\u2009s of extension time at 72\u2009\u00b0C.","A total of 4,579 individual plasmids were then randomly picked, miniprepped and plated as an arrayed plasmid DNA library on 96-well plates. This was deemed a practical number for arrayed library screening. A magnetic-bead-based SPRI purification method was used to minimize loss and to clean up the DNA before Golden Gate cloning. Sanger sequencing for quality checks of a sample of the plasmids confirmed that they were diverse, and all sequences could be mapped back to the original library design.\r\nScreening platform for 384-well plates\r\nCells were seeded in 384-well plates using a multidrop and transfected the following day using CyBio Felix (CyBio).","Two days later, a NanoGlo DLR assay was used to identify hits with a fold change to mCMV of >10. These hits were Sanger sequenced to determine the specific sequence and mapped back to the genome to validate overlap with the original SOX2 ChIP\u2013seq data. The GREAT tool predicted the target gene for hit sequences44.\r\nGeneral cell culture procedures\r\nThe GSC lines GCGR-E17 (E17), GCGR-E21 (E21), GCGR-E27 (E27), GCGR-E28 (E28), GCGR-E31 (E31), GCGR-E34 (E34), GCGR-E37 (E37) and GCGR-E55 (E55) and the human NSC lines NS9FB_B (NS9), NS12ST_A (NS12) and NS17ST_A (NS17) were generated in the Pollard Laboratory and are available upon request from the Glioma Cellular Genetics Resource. Informed consent was obtained for use of patient tissue.","All procedures on patient brain tissue received ethics approval from the NHS Health Research Authority, East of Scotland Research Ethics Service (REC reference 15\/ES\/0094), and all procedures on embryonic and fetal brain tissue received ethics approval from the Lothian NHS Board, South East Scotland Research Ethics Committee (REC reference 08\/S1101\/1).\r\n\r\nCells were grown as adherent monolayers under serum-free conditions as previously described13. All cells lines tested negative for mycoplasma at passage\u00a03 using a MycoAlert Mycoplasma Detection kit (Lonza, LT07-318). Cell line authentication by STR profiling was performed as a service by the European Collection of Authenticated Cell Cultures (ECACC).","Reports are available upon request from the Glioma Cellular Genetics Resource. Analysis was conducted using the Promega Fusion system (DC2402), analysing differences at 24 distinct hypervariable genetic loci, 16 of which are used for the final STR profile and report. The ioGlutamatergic (io1001S) and ioGABAergic (ioEA1003S) neurons and ioMicroglia (ioA021) were obtained from bit.bio and grown per the manufacturer\u2019s specifications. GSC7 was previously characterized22.\r\n\r\nFor lipofection, Plus reagent and Lipofectamine LTX (Life Technologies, 15338030) were each diluted in half the volume of Opti-MEM\u00a0I reduced-serum medium (hereafter referred to as Optimem) (Life Technologies, 31985062).","Following this step, the Plus reagent and Optimem premix was added to all DNA samples followed by the Lipofectamine LTX and Optimem premix. The transfection mix was incubated for 5\u2009min at room temperature and then carefully dropped onto the cells.","Generally, no change in medium was carried out. Cells were analysed 2\u00a0days after transfection.\r\n\r\nHEK293 cells were seeded at the specified density in the respective plate format. The next day GMEM medium, DNA and polyethylenimine were mixed and incubated for 15\u2009min at room temperature to facilitate formation of the complex. The transfection mix was subsequently dropped onto the cells, and analysis was carried out 2\u2009days later.\r\nScreening assay using the Nano-Glo Dual-Luciferase reporter assay system\r\nThis assay consists of two steps. Transfection was performed using a normalization plasmid PGK-Firefly-Luciferase (Promega, E5011) transfected in a 1:10 ratio with the plasmid of interest.","Therefore, in this assay, the firefly luciferase activity is first measured, which enables normalizing of the transfection efficiency. Cells were washed 3\u2009times with PBS and 20\u2009\u00b5l was left in a 96-well plate (25\u2009\u00b5l in a 384-well plate). Oneglo buffer was added, and the plate was shaken for 5\u2009min at 480\u2009rpm to allow cell lysis.","Next, 20\u2009\u00b5l cell lysate in a 96-well format (25\u2009\u00b5l cell lysate in a 384-well format) was transferred into an opaque white 96-well format (or 384-well format) and light was measured for 0.1\u2009s per well (Ensight Multimode Plate Reader, Perkin Elmer). For the subsequent NanoLuc reaction in a 96-well format, 2\u2009\u00b5l cell lysate was transferred into an opaque white plate containing 40\u2009\u00b5l PBS, and 20\u2009\u00b5l Stopglo buffer, supplemented with substrate, was added to each well. In a 384-well format, 20\u2009\u00b5l Stopglo buffer containing substrate was added on top of the undiluted cell lysate. The plate was shaken again for 5\u2009min at 480\u2009rpm to quench the firefly luciferase reaction and to ensure good mixing.","NanoLuc activity was measured for 0.1\u2009s per well. Data obtained by the NanoLuc reaction were normalized to the Firefly reaction to account for well-to-well variability of transfection efficiency. Normalized data were used to calculate the fold change compared with the empty vector control, which contained mCMV only and no enhancer fragments.\r\nImmunocytochemistry\r\nCells were washed twice with PBS and fixed in 4% paraformaldehyde (PFA) for 10\u2009min. Cells were permeabilized with 0.1% Triton-100 in PBS (PBST). Cells were blocked with blocking solution (1% BSA in PBST with 3% goat serum) and incubated with the primary antibody at 4\u2009\u00b0C overnight.","The next day, cells were washed 3\u2009times in PBST, and the respective secondary antibody was applied in blocking solution and incubated for 45\u201360\u2009min at room temperature. Cells were washed with PBS and incubated for 5\u2009min with a DAPI nuclear counterstain at 1\u2009\u00b5g\u2009ml\u20131 final concentration. Images were acquired using a Nikon TiE microscope and NIS elements software (Nikon).\r\nFlow cytometry\r\nFor flow cytometry, cells were detached, pelleted and resuspended in an appropriate volume of flow cytometry buffer (1% BSA in PBS, v\/v) or PBS.","Cells were stained with Draq7 (Abcam, ab109202, final concentration of 0.1\u2009\u00b5M) or DAPI (ThermoFisher Scientific, D3571) as a live\/dead stain and analysed using a BD LSRFortessa cell analyser (4 lasers, BD Bioscience). Analysis of flow cytometry data was carried out using FlowJo Analysis software (v.10.6.2) or FCS Express 7 flow cytometry software (v.7.22.x).\r\nRNA isolation, cDNA synthesis and RT\u2013qPCR\r\nRNA was extracted using a MasterPure RNA Purification kit (Epicentre). RNA was stored at \u221280\u2009\u00b0C, and concentration was determined using a NanoDrop spectrophotometer. cDNA synthesis was carried out using a SuperScript III Reverse Transcriptase kit (Life Technologies) according to the manufacturer\u2019s instructions.","Around 200\u2013500\u2009ng RNA was used for the cDNA reaction. The same amount of RNA was used for each experiment. After the reaction, cDNA was diluted to the required volume using nuclease-free water.\r\n\r\nFor RT\u2013qPCR, TaqMan Universal PCR master mix (Applied Biosystems) and TaqMan gene expression assays (Life Technologies) were used on a Quant Studio7 Flex Real-Time qPCR machine. RNA samples that did not undergo reverse transcription (to assess DNA contamination) and water controls were used on every plate. RT\u2013qPCR was carried out in technical duplicates. Data analysis was performed using the ddCt method, which assumes 100% PCR efficiency and is guaranteed with TaqMan assays.","In brief, the mean was calculated for technical replicates and normalized to the housekeeping gene GAPDH, which produces the dCt value.\r\nWestern immunoblotting\r\nCell lysates were prepared by resuspending cells in RIPA buffer (50\u2009mM HEPES pH\u20097.7, 150\u2009mM NaCl, 1% NP-40, 0.5% DOC and 0.1% SDS), incubating on ice for 5\u2009min, centrifuging at maximum speed for 10\u2009min in a tabletop centrifuge (5415D, Eppendorf) and collecting the supernatant (lysate). Protein extracts were quantified using a Pierce BCA Protein Assay kit (Thermo Scientific, 23225). SDS\u2013PAGE was performed using homemade 4\u201312% Bis-Tris gels.","Gels were transferred onto PVDF membranes (Millipore, IPVH00010), previously activated in methanol, by wet electroblotting or semi-dry blotting using a Bio-Rad Trans-blot turbo system. Western blots were revealed using HRP-conjugated antibodies, homemade ECL solutions or Clarity Western ECL (Bio-Rad), and imaged using X-ray films or a Bio-Rad ChemiDoc Imaging system.\r\nPrecipitation of interacting proteins\r\nPCR amplification with biotinylated primers and purification were used to generate templates that could pull down bound proteins.","Enhancer fragments were amplified with 5\u2032 biotinylated primers using PrimeStar Max (Takara) according to the manufacturer\u2019s instructions.The following primers were used: forward primer 595: TGATCCGTCTCGCCCTACTAGGTTACTGGTGCATGC; reverse primer 596: ACTAACGTCTCGGAGCACCCAAACTATTGGAGCGAG\u00a0(bold sequences, adaptors).\r\n\r\nPCR purification using Agencourt AMPure XP magnetic beads (Beckman Coulter) was carried out according to the manufacturer\u2019s instructions. PCR product quantification using Tapestation and reagents was carried out according to the manufacturer\u2019s instructions.\r\n\r\nAll the buffers were prepared the day before, passed through a 22\u2009\u00b5m filter (except the dialysis buffer) and left at 4\u2009\u00b0C overnight.","DTT and protease inhibitors were added immediately before use. Cell pellets were resuspended in 5\u2009ml per 40\u00a0million cells of ice-cold buffer\u2009A (10\u2009mM HEPES pH\u00a07.9, 1.5\u2009mM MgCl2, 10\u2009mM KCl, 0.5\u2009mM DTT and protease inhibitors (complete, Roche, 11697498001)) and incubated on ice for 10\u2009min. The cell suspension was transferred to a glass Dounce homogenizer and dounced 40 times on ice, transferred to a Falcon tube and centrifuged at 1,350\u2009rcf for 10\u2009min at 4\u2009\u00b0C.","The supernatant was discarded (cytosolic extract) and the pellet was resuspended in 100\u2009\u00b5l per 10\u2009million cells in ice-cold buffer\u2009B (20\u2009mM HEPES pH\u20097.9, 5% glycerol, 1\u2009M NaCl, 1.5\u2009mM MgCl2, 0.2\u2009mM EDTA pH\u20098.0, 0.5\u2009mM DTT and protease inhibitors (complete, Roche, 11697498001)). The suspension was rotated for 30\u2009min at 4\u2009\u00b0C and transferred into a dialysis membrane (SnakeSkin, Thermo Scientific, 68100). Dialysis was carried out in 500\u2009ml dialysis buffer (20\u2009mM HEPES pH\u20097.9, 5% glycerol, 100\u2009mM KCl, 0.83\u2009mM EDTA pH\u20098.0, 1.66\u2009mM DTT and protease inhibitors (complete, Roche, 11697498001)) for 2\u2009h at 4\u2009\u00b0C with rotation.","The dialysis buffer was changed (500\u2009ml), and dialysis was continued overnight at 4\u2009\u00b0C. The extract was collected in a 1.5\u2009ml tube and spun at maximum speed for 15\u2009min at 4\u2009\u00b0C (centrifuge 5415D Eppendorf). The supernatant was collected into new 1.5\u2009ml tubes and quantified (Pierce BCA Protein Assay kit, Thermo Scientific, 23225).","Next, 100\u2009\u00b5g aliquots were flash-frozen in liquid nitrogen and stored at \u221280\u2009\u00b0C. The following primary antibodies were used for western blotting: anti-SOX2 (rabbit) 1:1,000 Abcam (ab92494); anti-SOX9 (rabbit) 1:500 Millipore (ab5535); and anti-GAPDH (mouse) 1:10,000 Ambion (AM4300). The secondary antibodies used for western blotting were anti-rabbit 1:5,000 Novex (A16110) and anti-mouse 1:5,000 Novex (A16027).\r\n\r\nStreptavidin magnetic beads (NEB, S1420S) were used according to the manufacturer\u2019s instructions.","In brief, 10\u2009\u00b5l beads were aliquoted into low-bind tubes and washed 3\u2009times with binding buffer (20\u2009mM Tris-HCl pH\u20097.5, 0.5\u2009M NaCl and 1\u2009mM EDTA) on a magnetic stand.","Next, 20\u2009\u00b5l biotinylated DNA (around 20\u2009pmol) was mixed with 200\u2009\u00b5l binding buffer and was added to the beads. The suspension was rotated at room temperature for 2\u2009h. The beads were washed 3\u2009times with binding buffer, and 50\u2009\u00b5g nuclear extract was added (total volume of 50\u2009\u00b5l).","Beads, DNA and nuclear extract, to which 150\u2009\u00b5l dialysis buffer (20\u2009mM HEPES pH\u20097.9, 5% glycerol, 100\u2009mM KCl, 0.83\u2009mM EDTA pH\u20098.0, 1.66\u2009mM DTT and protease inhibitors (complete, Roche, 11697498001)) was added, were rotated overnight at 4\u2009\u00b0C. The following morning, beads were washed 3\u2009times with wash buffer (20\u2009mM HEPES pH\u20097.9, 5% glycerol, 250\u2009mM NaCl, 0.83\u2009mM EDTA pH\u20098.0 and 1.66\u2009mM DTT). For western blotting, proteins were eluted in 20\u2009\u00b5l loading buffer (lithium dodecyl sulfate buffer containing 50\u2009mM DTT) and boiled for 5\u2009min to denature the proteins.","For mass spectrometry, the beads were sent dry to the Institute for Genetics and Cancer core facility (University of Edinburgh) for analyses.\r\nSOX2 and SOX9 ChIP\u2013seq library preparation and analysis\r\nIn brief, GSCs cultured in 150\u2009mm dishes to 70% confluence were crosslinked in 1% formaldehyde for 10\u2009min at room temperature. Excess formaldehyde was quenched by incubating cells with 0.25\u2009M glycine for 5\u2009min at room temperature. The cell pellet was washed twice with ice-cold PBS before storage at \u221280\u2009\u00b0C. For each GSC line, cell pellets from eight 150\u2009mm dishes were combined before proceeding for cell lysis and sonication.","The cell pellet was lysed in buffer LB3 (10\u2009mM Tris-HCl pH\u20098.0, 100\u2009mM NaCl, 1\u2009mM EDTA pH\u20098.0, 0.5\u2009mM EGTA pH\u20098.0, 0.1% sodium deoxycholate, 0.5% N-lauroylsarcosine and 0.1% SDS) and chromatin was fragmented to 200\u2013600\u2009bp size on a Covaris M220 sonicator (total of 4 cycles, with the following treatment settings: 600\u2009s per cycle, peak power 75, duty factor 10, cycles\/burst 200; temperature: minimum 5\u00a0\u00b0C, set point 7\u00a0\u00b0C, maximum 9\u00a0\u00b0C). For each ChIP assay, 15\u201320\u2009\u00b5g fragmented chromatin was incubated overnight in a cold room with 10\u2009\u03bcg antibodies (R&D SOX2 AF2018, Millipore, SOX9 AB5535) and 30\u2009\u03bcl Protein G Dynabeads (10003D).","The next day, magnetic beads were washed 5 times with RIPA (50\u2009mM HEPES-KOH pH\u20097.5, 0.5\u2009M LiCl, 1\u2009mM EDTA pH\u20098.0, 1% NP-40 and 0.7% sodium deoxycholate), once with TE NaCl (10\u2009mM Tris-HCl pH\u20098.0, 1\u2009mM EDTA pH\u20098.0 and 50\u2009mM NaCl) and finally eluted in 200\u2009\u03bcl ChIP elution buffer (50\u2009mM Tris-HCl pH\u20098.0, 10\u2009mM EDTA pH\u20098.0 and 1% SDS). Immunoprecipitated chromatin was reverse crosslinked by incubating samples at 65\u2009\u00b0C for 10\u2009h and extracted using the phenol\u2013chloroform\u2013IAA (11896714) method. To account for intertumoral heterogeneity, ChIP\u2013seq was performed on seven independent primary GSC lines.","Libraries were prepared for two technical ChIP replicates and one input control using a NEBNext Ultra II DNA Library Prep kit for Illumina (E7645S). The libraries were barcoded using NEBNext Multiplex Oligos for Illumina (Dual Index Primers Set 1 kit, E7600S) and sequenced on an Illumina Novaseq (50\u2009bp read length).\r\n\r\nPaired-end reads were aligned to the hg38 genome using BWA, filtering out poor-quality (MAPQ\u2009<\u200910), duplicates, mitochondrial genome and blacklisted regions (https:\/\/www.encodeproject.org\/annotations\/ENCSR636HFF\/). Technical replicates were merged, and peaks were called using MACS2 with default settings.","Consensus SOX2 and SOX9 peak sets were derived by taking the overlap of peaks occurring in 5 out of 7 GSCs for each TF set. The overlap significance between the consensus SOX2 and SOX9 peak sets was determined using a circular permutation test in regioneR (ntimes\u2009=\u200910,000). Genomic regions near SOX2 and SOX9 peaks were annotated using HOMER45.\r\n\r\nTo perform overlapping analysis with GSC super-enhancers, we downloaded raw fastq files from publicly available H3K27ac datasets on GSCs (GSE119834, GSE74529, GSE121601 and GSE92458) and called super-enhancers using ROSE with default settings. A consensus set of GSC super-enhancers was derived by considering super-enhancers that occurred in at least two GSCs.","The overlap significance between the consensus co-bound SOX2\u2013SOX9 sites and consensus GSC super-enhancers was performed using a circular permutation test as described above. Gene ontology term association analysis was performed using GREAT. To find centrally enriched de novo motifs at SOX2 and SOX9 peaks in co-bound SOX2\u2013SOX9 enhancers, and to identify the spacing between the most significant de novo motifs from each peak set, we used CentriMo and SpaMo from the MEME-ChIP suite of tools46.\r\nscRNA-seq sample preparation\r\nOn the day of cell seeding, cells were detached as described above. Cells were counted using a haemocytometer and seeded into a 6-well Corning plate at 400,000 cells per well in 2\u2009ml.","Plates were incubated at 37\u2009\u00b0C with 5% CO2. The next day, AAV vector stocks were thawed at room temperature. The appropriate amount of virus stock was added to the culture medium to achieve a final multiplicity of infection (MOI) of 1,000,000, which ensured high transduction efficiency. Viral-containing medium was added to cells without replacing the existing medium. The cells were then returned to the incubator at 37\u2009\u00b0C with 5% CO2.\r\n\r\nAfter 3\u2009days of incubation with the viral particles, cells were detached as described above.","Next, 10% of the cells were prepared for flow cytometry using DAPI as a live\/dead stain (see\u00a0Supplementary Information for the gating strategy). The remaining 90% of the cells were processed for single-cell transcriptome sequencing.\r\n\r\nCells were fixed using a Parse Evercode v.2 Cell Fixation kit following the manufacturer\u2019s instructions, with an average of around 400,000 cells per sample. Fresh reagents were prepared for each fixation, and 7.5% Gibco BSA fraction\u2009V was added per the manufacturer\u2019s recommendations.","In brief, cells were washed with PBS and resuspended in cold Cell Prefixation buffer. Cells were filtered through a 40\u2009\u00b5m strainer after 5\u2009min of centrifugation at 300g at 4\u2009\u00b0C. Following this step, Cell Fixation solution and Cell Permeabilization solution were subsequently added, with incubation on ice for 10\u2009min and 3\u2009min, respectively. Cell Neutralization buffer was then added, and cells were centrifuged again at 300g for 5\u2009min at 4\u2009\u00b0C.","Finally, cells were resuspended in a volume of 50\u201390\u2009\u00b5l Cell Buffer containing 1% DMSO based on the live cell number before fixation. Cells were filtered through a 40\u2009\u00b5m strainer and stored at \u221280\u2009\u00b0C for up to 3\u2009months.\r\nscRNA-seq library preparation\r\nLibraries were generated using a Parse Evercode WT Mega v.2 kit according to the manufacturer\u2019s instructions.","In brief, 72 samples were loaded into a 96-well RT barcoding plate for reverse transcription, with each well containing a unique barcoded RT primer. Following cDNA synthesis, cells were pooled and subjected to two additional rounds of barcoding, achieving three rounds of barcoding. Barcoded cDNA was then amplified via template switching and pooled into 15 sublibraries, each containing around 62,500 cells. A fourth barcode was incorporated during PCR amplification of the sequencing libraries. Sublibraries were then fragmented, ligated with Illumina adapters and purified using Ampure XP beads. Library quality was assessed on an Agilent BioAnalyzer 2100 before sequencing.","Sublibraries, with 5% PhiX spike-in, were sequenced on an Illumina NovaSeq X using 300 cycles kits as paired-end, dual-index reads.\r\nscRNA-seq analysis\r\nSequencing reads were aligned to the human (hg38) genome, and UMI counting data were generated following the standard Parse Bioscience pipeline, \u2018splitpipe (v.1.1.2)\u2019, with default parameters. Three custom genes were added to the genome (mCherry, HSV-TK and bGHployA) to capture \u2018activated\u2019 SSE-7 cells.","In total, 345,495 cells were included in the initial dataset. Low-quality cells were filtered out based on the following thresholds: (1) the percentage of mitochondrial gene was >20%; (2) the number of genes was <300; (3) the number of uniquely aligned reads was <500; or (4) the number of uniquely aligned reads was >20,000. Genes detected in fewer than five cells were filtered out as low-quality genes. Potential doublets were identified using the R package \u2018DoubletFinder\u2019 (v.2.0.4), with an expected doublets rate of 3%, as guided by Parse. Doublets and clusters containing more than 20% doublets were removed. Ambient RNA-contaminated cells (cell score\u2009>\u20090.2) were removed using \u2018decontX\u2019 from the R package \u2018celda\u2019 (v.1.18.2).","After filtration, 270,842 cells remained for further analysis.\r\n\r\nThe data were normalized using \u2018LogNormalize\u2019, and principal component analysis was performed based on the top 2,000 variant genes using the R package \u2018Seurat\u2019 (v.5.0.3). The first 15 principal components were used as input for Louvain-based graphing. SSE-7-activated cells (2,711 cells) were labelled according to the following criteria: mCherry\u00a0>\u00a01 or HSV-TK\u2009>\u20091 or bGHpolyA\u2009>\u20090. Cells with a single count of mCherry or HSV-TK from the virus-treated group were labelled as false negative (9,583 cells), whereas the remaining cells were labelled as SSE-7 non-activated cells (125,525 cells).","Meanwhile, cells expressing mCherry or HSV-TK in the control group were labelled as false positive (1,998 cells). Transcriptional subtypes were predicted on the basis of a previously described gene signature27 using the function \u2018sigScores\u2019 and \u2018as_four_state_gbm\u2019 from the R package \u2018scalop\u2019 (v.1.1.0). Cell cycle was predicted using \u2018CellCycleScoring\u2019 from the R package \u2018Seurat\u2019.\r\nSCENIC analysis of scRNA-seq data\r\nFalse-negative and false-positive cells were removed, and the dataset was divided into 15 libraries as per the sequencing library preparation. The dataset was then analysed following the standard \u2018pyscenic (v.0.12.1)\u2019 pipeline.","Auxiliary human datasets (hg38 mc_v10_clust) were downloaded from the cisTarget resources website (https:\/\/resources.aertslab.org\/cistarget\/). Motif\u2013TF annotation was based on 10\u00a0kbp upstream and downstream around the transcription start sit (20\u00a0kbp in total). SSE-7-enriched TF modules were selected on the basis of the following criteria: (1) Z\u2009score\u2009>\u20090.7 and (2) Z\u2009score in the SSE-7-activated group was higher than in other groups, as determined by repeated-measures ANOVA test.","Enrichment for candidate TF modules across 15 libraries was as follows: SOX8 (found in 9 libraries), STAT1 (found in 15 libraries), IRF9 (found in 12 libraries), ETS1 (found in 13 libraries), MAF (found in 3 libraries), SMAD1 (found in 12 libraries), FOS (found in 11 libraries), HIF1A (found in 7 libraries), SOX9 (found in 10 libraries), STAT3 (found in 3 libraries), NEF2L2 (found in 13 libraries), JUN (found in 2 libraries) and SOX2 (found in 2 libraries).\r\nLentivirus production and titration\r\nTo produce SOX2, SOX9 and rtTA2M2 individual lentiviral supernatant samples, nearly 2.4\u2009million HEK293T cells per 15\u2009cm plate were cultured for 24\u2009h in GMEM medium supplemented with 10% fetal calf serum (FCS), 1\u2009mM sodium pyruvate, 1\u2009mM glutamine and non-essential amino acids (HEK medium) at 37\u2009\u00b0C and 5% CO2.","At 24\u2009h after seeding, HEK293T cells were transfected with a plasmid cocktail containing 7.5\u2009\u03bcg expression plasmid, 5.1\u2009\u03bcg psPAX2 packaging vector, 2.4\u2009\u03bcg envelope vector mixed in 45\u2009\u03bcl Fugene6 (Roche) and 855\u2009\u03bcl Opti-MEM medium (Invitrogen, Thermo Fisher Scientific). The cells were further cultured for 16\u2009h at 37\u2009\u00b0C and 5% CO2 before the medium was changed for HEK medium. At 65\u2009h after transfection, virus-containing supernatant was collected and cleared by centrifugation and filtered through a 0.45\u2009\u03bcm syringe filter (Merck Millipore).","Virus was concentrated (roughly 100-fold) by transferring to thin-walled ultracentrifuge tubes and pelleted by ultracentrifugation at 25,000\u2009rpm for 2.5\u2009h using a SW32-TI rotor in a Beckman Optima XPN ultracentrifuge (Beckman Coulter). The pelleted virus was resuspended in 300\u2009\u03bcl plain GMEM medium and incubated at 4\u2009\u00b0C for 16\u2009h before aliquoting and flash-freezing in liquid nitrogen for storage at \u221280\u2009\u00b0C.","The viral titre, calculated by flow cytometry, was 6\u2009\u00d7\u2009107 infectious units per ml.\r\nOverexpression of SOX2 and SOX9\r\nEarly passage (passage\u20094) human fibroblasts (hFibs) were seeded, 1.5\u2009million per 15\u2009cm plate, and cultured in hFib medium (GMEM medium supplemented with 10% FCS, 1\u2009mM sodium pyruvate, 1\u2009mM glutamine, MEM 1\u00d7 non-essential amino acids solution, 50\u2009\u03bcM 2-mercaptoethanol (Gibco, Thermo Fisher Scientific) and 2\u2009ml penicillin\u2013streptomycin (Invitrogen, 15140122)) 24\u2009h before transduction at 37\u2009\u00b0C and 5% CO2.","The cells were transduced by exchanging the growth culture medium with lentiviral medium, without antibiotics, containing 8\u2009\u03bcg\u2009ml\u20131 polybrene, and 65\u2009\u03bcl concentrated rtTA2M2 and either SOX2 or SOX9 virus, for single factor overexpression, or 75\u2009\u03bcl rtTA2M2, SOX2 and SOX9 viruses for double factor overexpression. At 48\u2009h after transduction, the cells were split 1:2 and cultured for 24\u2009h, after which the culture medium was replaced with induction medium containing 1\u2009\u03bcg\u2009ml\u20131 doxycycline (induction day\u20090). After 48\u2009h of induction, the cells were collected for lysis and nuclear protein extraction.\r\nNuclear extraction of GSC7 and hFib cells\r\nGSC7 cells were cultured as adherent monolayers13.","For differentiation, cells were cultured without the supplements EGF and FGF and in 5% serum for 2\u2009weeks. GSC7 and hFib nuclear lysates were prepared in the same way; cells were resuspended to 1\u00a0million cells per 100\u2009\u03bcl 10\u2009mM HEPES pH\u20097\u20138, 1.5\u2009mM MgCl2, 10\u2009mM KCl, 0.5\u2009mM DTT, supplemented with complete Ultra EDTA-free protease inhibitors (Roche, 05892970001) and homogenized using a dounce and tight-fit pestle with 1 stroke per 1\u2009million cells. Samples were centrifuged at 1,350\u2009rcf for 5\u2009min at 4\u2009\u00b0C.","The nucleus pellet was resuspended to 10\u2009\u03bcl per 1\u2009million starting cells in 20\u2009mM HEPES pH\u20097\u20138, 30% glycerol, 420\u2009mM NaCl, 1.5\u2009mM MgCl2, 0.2\u2009mM EDTA, 0.5\u2009mM DTT and complete Ultra EDTA-free protease inhibitors (Roche, 05892970001). Samples were dialysed for 4\u2009h into a minimum of 1\u20131,000 volume excess of 20\u2009mM HEPES pH\u20097\u20138, 30% glycerol, 100\u2009mM NaCl, 0.83\u2009mM EDTA, 1.66\u2009mM DTT, 0.2\u2009mM PMSF and complete Ultra EDTA-free protease inhibitors (Roche 05892953001) using Slide-A-lyzer mini dialysis units with 3,500 MWCO cassettes (Thermo).","Samples were divided into single-use aliquots (11\u2009\u03bcl) and flash-frozen in liquid nitrogen for storage at \u221280\u2009\u00b0C.\r\nWestern blot analysis of overexpression\r\nThe protein concentrations of the lysates were quantified using a Pierce BCA Protein Assay kit according to the manufacturer\u2019s 96-well format instructions (Thermo Fisher Scientific). The following overexpression (OE) values were used: SOX2 OE\u2009=\u200910.54\u2009mg\u2009ml\u20131, SOX9 OE\u2009=\u20095.56\u2009mg\u2009ml\u20131, SOX2 and SOX9 OE\u2009=\u20097.56\u2009mg\u2009ml\u20131, GSC7 GSCs\u2009=\u20091.02\u2009mg\u2009ml\u20131 and GSC7 differentiated cells\u2009=\u20090.43\u2009mg\u2009ml\u20131.","Verification of SOX protein overexpression, relative to untransfected hFib lysate, was resolved by SDS\u2013PAGE and electroblotting onto a PVDF membrane for western blotting. The primary antibody incubations with goat anti-human SOX2 antibody (1:363, AF2018, Abcam) and rabbit anti-mouse SOX9 (1:2,000 AB5535, Chemicon) were performed at 4\u2009\u00b0C for 16\u2009h. The secondary antibody incubations with donkey anti-goat IgG-HRP (1:2,000 dilution of ab97110, Abcam) and goat anti-rabbit (1:2,000 of 32460, Invitrogen) were performed for 1\u2009h at room temperature.","Blots were visualized using SuperSignal West Pico Chemiluminescent substrate (Thermo Fisher Scientific) and HRP was visualized using a Bio-Rad ChemiDoc MP imaging system.\r\nRecombinant SOX protein production\r\nHis-tagged human SOX2 and SOX9 were expressed from a pET28a backbone (cloned by G.\u00a0Roberts in the Soufi Laboratory) using Rosetta 2(DE3) pLys competent cells (Novagen).\r\n\r\nSOX2 was expressed as previously described47. Cells in LB medium (1% Bacto-tryptone, 0.5% yeast extract and 1% NaCl), supplemented with 30\u2009\u00b5g\u2009ml\u20131 kanamycin and chloramphenicol, were grown at 37\u2009\u00b0C overnight.","SOX9 was expressed in 2\u00d7\u00a0TY medium (1.6% Bacto-tryptone, 1% yeast extract and 0.5% NaCl), supplemented with 30\u2009\u00b5g\u2009ml\u20131 kanamycin and chloramphenicol, at 16\u2009\u00b0C overnight. After expression, bacteria cell pellets were lysed, 100\u2009ml buffer per 1\u2009l of culture, in GuHCl Denaturing buffer (6\u2009M guanidine-HCl, 50\u2009mM Tris-HCl (pH\u20098) and 500\u2009mM NaCl, 5% (v\/v) glycerol) overnight. Lysate was sonicated for 200\u2013300\u2009s at 25\u201330\u2009microns using a Soniprep 150 with intermittent chilling on ice. The supernatant was clarified by centrifugation at 18,000\u2009rpm in a FLA21-8x50y rotor for 30\u2009min.","The lysate was further sonicated for 120\u2009s, as above, then filtered through a 0.45\u2009\u00b5m filter before loading into a HisTrapHP column (Cytivia) for affinity purification of His-tagged proteins using an AKTA pure system. The column was equilibrated with 5\u00a0CV of denaturing 10\u00a0mM imidazole buffer (6\u2009M urea, 500\u2009mM NaCl, 50\u2009mM Tris-HCl (pH\u20098), 10\u2009mM imidazole and 5% (v\/v) glycerol), after sample loading was washed with 20\u2009CV of denaturing 50\u00a0mM imidazole buffer (6\u2009M urea, 500\u2009mM NaCl, 50\u2009mM Tris-HCl (pH\u00a08), 30\u2009mM imidazole and 10% (v\/v) glycerol), and bound proteins were eluted with denaturing 300\u2009mM imidazole (6\u2009M urea, 500\u2009mM NaCl, 50\u2009mM Tris-HCl (pH\u20098), 300\u2009mM imidazole and 10% glycerol).","Peak fractions of the elution were analysed by SDS\u2013PAGE, and samples with appropriately sized bands (40\u2009kDa for His\u2013SOX2, and 60\u2009kDa for His\u2013SOX9) were combined and desalted into 2\u2009M urea buffer using a HiPrep 26\/10 desalting column (GE Healthcare). The SOX2 desalting buffer comprised 50\u2009mM Tris-HCl (pH\u00a08), 240\u2009mM NaCl, 10\u2009mM KCl, 2\u2009mM MgCl2, 2\u2009mM CaCl2, 0.8\u2009M urea, 30% (v\/v) glycerol, 0.1% NP40 substitute, 0.05% Triton-X-100, 2\u2009mM EDTA and 5\u2009mM DTT resuspended in 1\u00d7 PBS. The SOX9 desalting buffer comprised 50\u2009mM HEPES NaOH (pH\u20097.5), 240\u2009mM NaCl, 10\u2009mM KCl, 2\u2009mM MgCl2, 2\u2009mM CaCl2, 2\u2009M urea, 30% (v\/v) glycerol, 0.1% NP40 substitute, 0.05% Triton-X-100, 2\u2009mM EDTA and 5\u2009mM DTT.","The fractions corresponding to a dip in the absorbance at A280 were collected and visualized by SDS\u2013PAGE before combining. Protein concentration was determined by SDS\u2013PAGE, comparing to rAlbumin standards. The gels were imaged by detecting Coomassie blue staining using a Bio-Rad ChemiDoc MP imaging system. The resulting images were visualized, and the bands quantified using ImageJ 2.8\/1.54i.\r\nSmall-molecule inhibitor screening\r\nE55 cells were screened using inhibitors included in the Kinase Screening Library (Cayman, 10505) as well as two YAP\u2013TEAD pathway inhibitors, MGH-CP1 (HY-139330) and TED-347 (HY-125269).","Around 18,000 cells were seeded per well in a 96-well plate and transduced with the corresponding AAVs at a MOI of 2,000,000 on the next day. Cells were incubated at 37\u2009\u00b0C, 5% CO2. In the primary screen, the cells were treated with a single concentration of 3\u2009\u03bcM per inhibitor at 24\u2009h after transduction. After 48\u2009h of inhibitor treatment at 37\u2009\u00b0C and 5% CO2, the cells were detached and analysed using an Agilent NovoCyte Penteon five laser flow cytometer and FCS Express 7 Research (v.7.24.0030). The results were expressed as the fold change in mCherry median fluorescence intensity in the inhibitor-treated samples relative to the DMSO control.","After performing the first round of screening, inhibitors that gave more than 20% changes were selected for the second round of screening.\r\n\r\nIn the second-round screen, the cells were seeded and transduced with AAVs using the same procedures as the primary screen. At 24\u2009h after transduction, the cells were treated with 3 different doses (0.1, 1 and 3\u2009\u03bcM), and incubated at 37\u2009\u00b0C and 5% CO2 for 48\u2009h. Cells were then detached and analysed using a Agilent NovoCyte Penteon five laser flow cytometer. Draq7 was used as the live\/dead stain in both rounds of screening.","The results of the second round screen were expressed as the fold change in mCherry MFI and as the percentages of mCherry-positive cells, each normalized to the DMSO control.\r\nEMSAs\r\nCy5-labelled SSE-7 bait for EMSAs was generated by PCR amplification of SSE-7 from the plasmid using Cy5-labelled primers (Sigma): 5\u2032-ACTAGGTTACTGGTGCATGCTTGTCCTGCCTTTGAGAACA-3\u2032 and 5\u2032-ACCCAAACTATTGGAGCGAGAGAAAGGAAAGAAAGAGGTC-3\u2032.","The following primers were used for Cy5-labelled fragments for EMSA: ID4310: 5\u2032-ACTAGGTTACTGGTGCATGCTTGTCCTGCCTTTGAGAACA-3\u2032 and 5\u2032-ACCCAAACTATTGGAGCGAGCTCATTTGAAGCAGAAGAAT-3\u2032; ID4428: 5\u2032-ACTAGGTTACTGGTGCATGCTCTAAATGGAGATCCTCCCA-3\u2032 and 5\u2032-ACCCAAACTATTGGAGCGAGACAGGAGGAAGTAGTAAATC-3\u2032; ID0785: 5\u2032-ACTAGGTTACTGGTGCATGCAAGAATGAACTGGGCCCAGC-3\u2032 and 5\u2032-ACCCAAACTATTGGAGCGAGTGAAAACCAGAGGCATCTCA-3\u2032; and ID3836: 5\u2032-ACTAGGTTACTGGTGCATGCTACCCTTCCAGGGGAGCAGT-3\u2032 and 5\u2032-ACCCAAACTATTGGAGCGAGAGAAAGGAAAGAAAGAGGTC-3\u2032.\r\n\r\nThe binding to Cy5-end-labelled SSE-7 was analysed in native 1% agarose gels (12\u00a0\u00d7\u00a013\u00a0\u00d7\u00a01\u00a0cm), which were prepared in 0.5\u00d7 TBE (45\u2009mM Tris-borate and 1\u2009mM EDTA).","Gels were stored overnight at 4\u2009\u00b0C before pre-running at 120\u2009V (approximately 10\u2009V\u2009cm\u20131) for 1\u2009h. For affinity analysis, a 10\u2009\u03bcl mixture typically containing 50\u2009nM Cy5-labelled full-length or fragmented SSE-7, 100\u2009ng\u2009\u00b5l\u20131 poly(dA:dT) (InvivoGen, tlrl-patn) and 0, 0.5, 1, 2 or 4\u2009\u03bcl of hFib lysates, with and without SOX overexpression (5.56\u2009mg\u2009ml for OE lysates and 3.47\u2009mg\u2009ml\u20131 for hFib control lysates), GSC7 GBM stem, or differentiated, cell lysates (at 0.43\u2009mg\u2009ml\u20131), or purified recombinant SOX protein (at 3.17\u2009\u00b5M). For SOX2 and SOX9 in combination, equimolar amounts of SOX2 and SOX9 were mixed for 0.5, 1, 2, 4 and 8\u2009\u00b5l of protein in total.","Next, 8\u2009\u00b5l SOX protein was added to 25\u2009nM SSE-7 DNA in a 20\u2009\u00b5l volume, and 20\u2009\u00b5l of sample was loaded for gels, which were prepared in 1\u00d7 binding buffer (10\u2009mM Tris HCl pH\u20097.5, 1\u2009mM MgCl2, 10\u2009\u03bcM ZnCl2, 10\u2009mM KCl, 1\u2009mM DTT, 5% (v\/v) glycerol and 0.5\u2009mg\u2009ml\u20131 BSA). The mixtures were incubated at 20\u2009\u00b1\u20091\u2009\u00b0C in the dark for 1\u2009h using DNA LoBind tubes (Eppendorf). The entirety of each sample was then loaded onto agarose gels and electrophoresis was performed at 120\u2009V for 3\u2009h at 4\u2009\u00b0C. The gels were imaged by detecting Cy5 fluorescence using a Bio-Rad ChemiDoc MP imaging system.","The resulting images were visualized, and the bands were quantified using ImageJ 2.8\/1.54i.\r\nAdult human cortex and brain tumour slice cultures\r\nTumour and non-tumour brain tissue were transferred to a 10-cm2 tissue culture dish with sterile PBS and placed on ice. Visibly damaged tissue was removed and the remaining tissue was transferred into a 35-mm2 dish with pre-warmed 3% SeaPlaque agarose (50100, Lonza). After cooling in ice, the block was removed and cut using a scalpel into an approximately 2\u2009cm cube around the brain. Before starting to cut, a 6-well plate was prepared.","In each well, we introduced one cell culture insert (PICMORG50, Millicell) and added below it 1\u2009ml culture medium in basal NSC medium, Dulbecco\u2019s modified Eagle medium: F12 supplemented with EGF and FGF (Life Technologies). The embedded brain was placed in a circular vibratome plate with glue. The vibratome (VT1000 S, Leica) plate was fixed in the platform and filled with PBS and penicillin\u2013streptomycin (15140-122, Gibco, 1:100).","Then, 300-\u00b5m-thick slices were cut, with vibrating frequency set at 10 and speed at 1. Each slice was transferred using a small brush onto the top of a Millipore culture insert. The platform was maintained cool at all times. The 6-well plate was placed in an incubator at 37\u2009\u00b0C with 5% CO2.\r\nAddition of virus\r\nA 5\u2009\u00b5l volume of virus was pipetted onto the centre of each tissue slice, using a 10\u2009\u00b5l tip without touching the tissue. Three repeat doses of 5\u2009\u00b5l virus was added to each slice at 5\u2009min intervals (20\u2009\u00b5l in total per slice). Slices were incubated for 7\u2009days.","The medium was replaced on day\u00a03 or 4.\r\nImmunostaining\r\nThe medium was removed and exchanged for 1\u2009ml freshly prepared 4% PFA; 1\u20132\u2009ml PFA was also placed gently on top to cover the slice. After 2\u2009h, PFA was removed, brain slices were washed 3\u2009times with PBS and transferred using a brush to a 24-well plate. Slices were incubated at room temperature for 1.5\u2009h in blocking solution (0.5% Triton X-100 and 3% goat serum; Sigma-Aldrich) followed by incubation with primary antibody for 2\u2009days at 4\u2009\u00b0C. Sections were washed five times with PBS, the respective secondary antibody was applied in blocking solution overnight.","The next day, slices were washed 5 times with PBS mounted and cleared in RapiClear 1.49 (RC149001, Sunjin Lab).\r\n\r\nImages were acquired using an Opera Phenix Plus high-content imaging system (Revvity) equipped with a \u00d740\/1.1 NA water-immersion objective. Image analysis was performed in Harmony software (v.5.2; Revvity). For the full pipeline see Supplementary Information.","In brief, nuclei were segmented using DAPI with the common thresholding method, and the nuclear mean intensity of channels SOX2 (488) and mCherry (555) were then calculated. Thresholds to define positive cells were set for each tissue slice individually based on 3\u20135 regions of interest. Nestin was segmented using \u2018find image region\u2019 with channel 647, and positive cells were defined as those with nuclei in these regions. Analysis was performed for a single z\u2009plane. Data were exported and analysed using GraphPad Prism.","Representative images of the segmentation can be found in the Supplementary Information.\r\nZebrafish experiments\r\nAll embryos were obtained by natural spawning and collected in conditioned aquarium water in 0.00001% methylene blue. Embryos were raised at 28.5\u2009\u00b0C in embryo medium (E3) on a 14\u2009h light\u201310\u2009h dark photoperiod and were treated with 200\u2009\u03bcM N-phenylthiourea (Sigma) from 6\u2009h post fertilization to inhibit pigmentation. Zygotes were injected at the one-cell stage of development. DNA constructs were created using the Tol2Kit system48,49.","Approximately 2\u2009nl of plasmid DNA (30\u2009ng\u2009\u03bcl\u20131) containing Tol2-capped mRNA (20\u2009ng\u2009\u03bcl\u20131), supplemented with 0.2% w\/v phenol red (Sigma) to facilitate visualization of injected volume, was injected. To induce Akt1 overexpression in neural progenitor cells, a combination of Tol2-pDEST-NBT:DlexPR-lexOP-pA (20\u2009ng\u2009\u03bcl\u20131) and Tol2-pDEST-lexOP:AKT1-lexOP:tagRFP (30\u2009ng\u2009\u03bcl\u20131) plasmids was injected as previously described50. The newly generated tgSSE-7:eGFP was outcrossed to tg(Xla.Tubb:DsRed) and tg(olig1:mScarlet). Live imaging was performed using either a Leica MS205 stereomicroscope or a Leica SP8 confocal microscope with a \u00d740\/NA 1.1 objective.","Animal experimentation was approved by the ethics review committee of the University of Edinburgh and the Home Office in accordance with the Scientific Procedure Act 1986.\r\nAAV medium-exchange transduction assay\r\nHEK cells were seeded into a 6-well plate a day before transfection, aiming for 60\u201370% confluence the next day. On the day of transfection, the culture medium was exchanged and HEK cells were transfected with RepCap, pHelper and AAV-ITR transgene plasmids using polyethylenimine. Two days later, conditioned HEK culture medium (now containing rAAV particles) was collected and centrifuged at 1,300\u2009rpm for 4\u2009min to remove any cells.","rAAV-conditioned medium was transferred to HEK or GSC7 cells seeded at low density (10\u201320%) in 6-well or 12-well plates. Two to three days later, cells were analysed for transduction. For AAV1 in vivo experiments, we purchased research-grade viral vectors from Vector Biolabs (iodixanol gradient ultracentrifugation purified).\r\nAAV transduction assay in iPS cell-derived neurons\r\nGSC7 and E55 cells were seeded in a 24-well plate format the day before transduction, whereas ioGlutamatergic neurons, ioGABAergic neurons and microglia (bit.bio) were seeded in a 48-well format before differentiation. After differentiation, cells were transduced with AAVs at a MOI of 5\u2009\u00d7\u2009105.","Cells were incubated at 37\u2009\u00b0C and 5% CO2 for 3 or 10\u00a0days.\r\nCell seeding and transduction for GCV killing assay\r\nOn the day of cell seeding, cells were detached using the method described above. Cells were counted using a haemocytometer and seeded into 96-well (1,000 cells per well in 50\u2009\u00b5l), 24-well (30,000 cells per well in 500\u2009\u00b5l) or 6-well (60,000 cells per well in 2\u2009ml) Corning plate and placed in an incubator at 37\u2009\u00b0C and 5% CO2. The next day, AAV stocks were thawed at room temperature, and an appropriate amount of virus stock was added to the required amount of culture medium to achieve the final MOI of 5\u2009\u00d7\u2009105. Culture medium containing viral particles was added to cells without the replacement of existing medium.","Cells were returned to a 37\u2009\u00b0C and 5% CO2 incubator.\r\nProdrug treatment\r\nLyophilized GCV was diluted in DMSO to achieve 100\u2009mM stock concentration. GCV stocks were aliquoted and were stored at \u221220\u2009\u00b0C for no longer than 1\u2009month. To make a working stock concentration, GCV was diluted 1:100 in appropriate culture medium, and 20\u2009\u00b5l was added to wells already containing 80\u2009\u00b5l culture medium for a final concentration of 200\u2009\u00b5M. As a negative control, DMSO was diluted 1:100 in appropriate culture medium to achieve a working stock.","Next, 20\u2009\u00b5l of working stock was added to wells already containing 80\u2009\u00b5l culture medium. As a positive control, 20\u2009\u00b5l DMSO was added to wells with 80\u2009\u00b5l culture medium to achieve a final concentration of 20% DMSO.\r\nIncucyte live-cell imaging\r\nTo track cell proliferation and morphological changes during treatments, cells were monitored using an Incucyte live-cell imaging system. Whole-well imaging of Corning 96-well\u00a0plates was performed every 4\u2009h. Basic confluence scoring analysis software (Incucyte) was used to estimate confluence.","Images at specific time points were extracted to verify cell confluence and morphology.\r\nMTT assay\r\nOn the day of assay, the culture medium was replaced with 0.3\u2009mg\u2009ml\u20131 MTT solution (diluted in cell line-appropriate culture medium). Cells were placed in incubator at 37\u2009\u00b0C and 5% CO2 for 3\u2009h. After incubation, the medium was removed and 70\u2009\u00b5l DMSO was added to each well. Each plate was kept in the dark at 37\u2009\u00b0C for 20\u2009min, shaking it occasionally. Before reading the plate, each well was visually inspected to make sure that all (formazan) crystals were dissolved.","Plates were read with plate reader at 560\u2009nm absorbance.\r\nStatistical analysis\r\nData analysis was performed using Microsoft Excel (v.16.23 for Mac), GraphPad Prism (v7) and RStudio (v1.1.456). Error bars are shown as the s.d. of the mean.","For illustrations, BioRender and Adobe Illustrator 22.0.1 were used.\r\n\r\nFurthermore, we used open-source programmes such as bedtools, GREAT, fastasplitter, HOMER and MEME for various analyses45,46.\r\nTumour initiation by transplantation and intratumoral AAV delivery\r\nAll animal procedures were approved by the University of Edinburgh Animal Welfare and Ethical Review Body (AWERB) and conducted under UK Home Office licence (PPL number: PP8631583) in accordance with the Animals (Scientific Procedures) Act 1986 and ARRIVE guidelines. Male and female C57BL\/6 mice (6\u20138\u2009weeks old) were obtained from Charles River.","Mice were housed in Individually ventilated cages with sterile bedding plus enrichment inside a pathogen-free facility on a 12-h light\u2013dark cycle, at a temperature range of 20\u201324\u2009\u00b0C and with relative humidity level of 45\u201365%. Transplantation experiments were performed as previously described25. We used 7-week-old male mice. Mice were transplanted with 200,000 NPE-IE cells in 2\u2009\u00b5l PBS per mouse. NPE-IE cells have been previously described25 and have Nf1 and Pten inactivating mutations, alongside EGFvIII overexpression and a GFP-Luc reporter construct.","At 2\u2009weeks, when tumours were visible by IVIS imaging (bioluminescence) and of similar size, AAV1-SSE-7\u2013HSV-TK\u2013IL-12 virus (1.2\u2009\u00d7\u20091013 viral genomes per ml, Vector Biolabs) was delivered directly into the mouse brain tumour (2.5\u2009\u00b5l per mouse) using a microsyringe linked to an injection pump, at a rate of 0.17\u2009\u00b5l\u2009min\u20131 and 2.3\u2009mm depth, higher than the previous depth of 2.4\u2009mm for tumour cells. Each mouse received 3.00\u2009\u00d7\u20091010 viral genomes. The following day after virus injections, GCV was intraperitoneally injected (500\u2009\u00b5l per mouse at 2\u2009mg\u2009ml\u20131) once daily for 20\u2009days. Mice were imaged by IVIS imaging (PerkinElmer) once a week for luciferase signals.","Animals were observed regularly for any neurological symptoms. Mice with tumours were culled based on the following criteria; (1) body weight loss of \u226520% relative to baseline (weights of mice were recorded once weekly); (2) abnormal neurological signs, including ataxia, head tilt, circling behaviour, seizures, paresis or inability to right itself within 3\u2009s; (3) hunched posture with reduced locomotor activity and lack of reach response; and (4) any extracranial tumours\u00a0\u2265\u00a015\u2009mm in diameter or ulcerated as per the Workman (2010) guidelines51. All mice reaching the humane end point were culled using the schedule\u00a01 method.","The end-point limits were not exceeded in any of the experiments.\r\nImmune characterization of mouse GBMs\r\nBrains were collected and processed 5\u2009days after treatment.","First, they were microdissected under a Leica stereo microscope to avoid contamination with healthy tissue. Tissues were mechanically disaggregated using scissors, followed by 30\u2009min of enzymatic digestion using a mix of 0.35\u00a0\u00b5g\u2009ml\u20131 Liberase TL (05401119001, Roche) and 0.23\u2009\u00b5g\u2009ml\u20131 DNAse (101041590001, Sigma) in plain RPMI. After digestion, tissues were dispersed through a 70\u2009nm cell strainer to obtain a single-cell suspension. After centrifugation, cells were resuspended in PBS for further flow cytometry staining. Cell suspensions were incubated for 10\u2009min at 4\u2009\u00b0C with a blocking mixture of mouse, rat and calf serum containing anti-CD16\/32 blocking antibodies (clone 2.4G2, BioXcell, BE0307).","Cells were washed with PBS and incubated for 20\u2009min in antibody-containing staining buffer plus Fixable Viability Dye eFluor-780 (eBioscience, 65-0865-14) to distinguish live and dead cells. Cells were then washed and resuspended in fixation\/permeabilization buffer (BD CytoFix\/Cytoperm; 554714 or FOXP3 Transcription Factor Staining Buffer set, eBioscience) followed by intracellular target staining.\r\nSABER\u2013FISH detection of transgene copy number\r\nSABER\u2013FISH was performed as previously described52,53. Non-overlapping 36\u2009bp probes were designed to target the HSV-TK transgene, and probes were extended using primer exchange reaction to approximately 500\u2009bp.","For imaging, cells were first stained with mCherry,then fixed with PFA and subsequently processed for immuno-SABER\u2013FISH54.\r\nAssessment of the fraction of the tumour cell population that activates AAV1-SSE-7\u2013mCherry\r\nAAV1-SSE-7-mCMV\u2013HSV-TK-V5\u2013mCherry was injected at around 3.58\u2009\u00d7\u20091014 viral genomes per ml and 7.2\u2009\u00d7\u20091013 viral genomes per ml (1:5 dilution) in PBS. AAV was injected into tumours as described above. After 7\u2009days, tumours were imaged using a stereomicroscope and surgically excised.\r\n\r\nEach tumour sample was added to 1\u2009ml digest mix (RPMI\u2009+\u20091\u2009mg\u2009ml\u20131 collagenase IV, 12.6\u2009\u00b5g\u2009ml\u20131 DNase\u2009I and 1% penicillin\u2013streptomycin) and incubated for 1\u2009h in a shaking incubator at 40\u2009rpm at 37\u2009\u00b0C.","Samples were filtered through a 100\u2009\u03bcm cell strainer, washed with 9\u2009ml FACS buffer and centrifuged at 250g for 15\u2009min at room temperature. The pellet was then resuspended in 1\u2009ml 37.5% Percoll in PBS and gradient centrifuged at 900g, room temperature, for 12\u2009min with no brake. The supernatant was discarded, and the bottom layer of cells resuspend in 500\u2009\u00b5l PBS and filtered into a FACS tube.","Cells were stained with DAPI as a live\/dead stain and analysed using a BD LSR Fortessa cell analyser (four lasers, BD Bioscience).\r\nReporting summary\r\nFurther information on research design is available in the\u00a0Nature Portfolio Reporting Summary linked to this article.\r\n\nData availability\r\nThe scRNA-seq data have been deposited at the European Nucleotide Archive (ENA) (accession number: PRJEB81816). The project number for ChIP\u2013seq data is PRJEB107008. For ChIP\u2013seq, raw (fastq) and processed (bigwig and narrowPeak) files have been deposited and made public (https:\/\/www.ebi.ac.uk\/ena\/browser\/view\/PRJEB107008).","The ENA does not take processed files; therefore, these have been uploaded to BioStudies, which links to the same ENA accession (PRJEB107008), otherwise they are at https:\/\/www.ebi.ac.uk\/biostudies\/studies\/S-BSST2733. All the codes and processing steps for scRNA-seq, SCENIC analysis and ChIP\u2013seq data are available at GitHub (https:\/\/github.com\/alhafidzhamdan\/sse_gene_therapy). The GitHub repository includes some processed files, including all the narrowPeak files and consensus sets; using these in conjunction with the .Rmd document for reproduction of Fig. 2. This also includes a link to the SOX2 and SOX9 tracks on the UCSC browser.","The linked and minted a repo is available at Zenodo (https:\/\/zenodo.org\/records\/18676096)55.\u00a0Source data is provided with this paper.\r\nReferences\r\nSuv\u00e0, M. L. et al. Reconstructing and reprogramming the tumor-propagating potential of glioblastoma stem-like cells.","Cell 157, 580\u2013594 (2014).\r\n\r\nArticle\nPubMed\nPubMed Central\nGoogle Scholar\r\n\r\nLi, C. & Samulski, R. J. Engineering adeno-associated virus vectors for gene therapy. Nat. Rev. Genet.","21, 255\u2013272 (2020).\r\n\r\nArticle\nCAS\nPubMed\nGoogle Scholar\r\n\r\nKribelbauer, J.","F.,","Rastogi, C., Bussemaker, H. J.","& Mann, R. S. Low-affinity binding sites and the transcription factor specificity paradox in eukaryotes. Annu. Rev. Cell Dev. Biol.","35, 357\u2013379 (2019).\r\n\r\nArticle\nCAS\nPubMed\nPubMed Central\nGoogle Scholar\r\n\r\nSpitz, F.","& Furlong, E. E. M. Transcription factors: from enhancer binding to developmental control. Nat. Rev. Genet.","13, 613\u2013626 (2012).\r\n\r\nArticle\nCAS\nPubMed\nGoogle Scholar\r\n\r\nSchlabach, M.","R.,","Hu, J.","K.,","Li, M.","& Elledge, S. J. Synthetic design of strong promoters. Proc. Natl Acad. Sci.","USA 107, 2538\u20132543 (2010).\r\n\r\nArticle\nADS\nCAS\nPubMed\nPubMed Central\nGoogle Scholar\r\n\r\nWu, M.-R. et al. A high-throughput screening and computation platform for identifying synthetic promoters with enhanced cell-state specificity (SPECS). Nat. Commun.","10, 2880 (2019).\r\n\r\nArticle\nADS\nPubMed\nPubMed Central\nGoogle Scholar\r\n\r\nCheng, J. K.","& Alper, H. S. Transcriptomics-guided design of synthetic promoters for a mammalian system. ACS Synth. Biol.","5, 1455\u20131465 (2016).\r\n\r\nArticle\nCAS\nPubMed\nGoogle Scholar\r\n\r\nJindal, G. A.","& Farley, E. K. Enhancer grammar in development, evolution, and disease: dependencies and interplay. Dev.","Cell 56, 575\u2013587 (2021).\r\n\r\nArticle\nCAS\nPubMed\nPubMed Central\nGoogle Scholar\r\n\r\nWhyte, W. A. et al. Master transcription factors and mediator establish super-enhancers at key cell identity genes.","Cell 153, 307\u2013319 (2013).\r\n\r\nArticle\nADS\nCAS\nPubMed\nPubMed Central\nGoogle Scholar\r\n\r\nZamudio, A. V. et al. Mediator condensates localize signaling factors to key cell identity genes. Mol.","Cell 76, 753\u2013766 (2019).\r\n\r\nArticle\nCAS\nPubMed\nPubMed Central\nGoogle Scholar\r\n\r\nHnisz, D.,","Shrinivas, K.,","Young, R.","A., Chakraborty, A. K.","& Sharp, P. A. A phase separation model for transcriptional control.","Cell 169, 13\u201323 (2017).\r\n\r\nArticle\nADS\nCAS\nPubMed\nPubMed Central\nGoogle Scholar\r\n\r\nHnisz, D. et al. Convergence of developmental and oncogenic signaling pathways at transcriptional super-enhancers. Mol.","Cell 58, 362\u2013370 (2015).\r\n\r\nArticle\nCAS\nPubMed\nPubMed Central\nGoogle Scholar\r\n\r\nPollard, S. M. et al. Glioma stem cell lines expanded in adherent culture have tumor-specific phenotypes and are suitable for chemical and genetic screens. Cell Stem Cell 4, 568\u2013580 (2009).\r\n\r\nArticle\nCAS\nPubMed\nGoogle Scholar\r\n\r\nGraham, V.,","Khudyakov, J.,","Ellis, P.","& Pevny, L. SOX2 functions to maintain neural progenitor identity.","Neuron 39, 749\u2013765 (2003).\r\n\r\nArticle\nCAS\nPubMed\nGoogle Scholar\r\n\r\nLujan, E.,","Chanda, S.,","Ahlenius, H.,","S\u00fcdhof, T. C.","& Wernig, M. Direct conversion of mouse fibroblasts to self-renewing, tripotent neural precursor cells. Proc. Natl Acad. Sci.","USA 109, 2527\u20132532 (2012).\r\n\r\nArticle\nADS\nCAS\nPubMed\nPubMed Central\nGoogle Scholar\r\n\r\nGangemi, R. M. R. et al. SOX2 silencing in glioblastoma tumor-initiating cells causes stop of proliferation and loss of tumorigenicity. Stem Cells 27, 40\u201348 (2009).\r\n\r\nArticle\nCAS\nPubMed\nGoogle Scholar\r\n\r\nSingh, D. K. et al. Oncogenes activate an autonomous transcriptional regulatory circuit that drives glioblastoma. Cell Rep.","18, 961\u2013976 (2017).\r\n\r\nArticle\nCAS\nPubMed\nPubMed Central\nGoogle Scholar\r\n\r\nMacLeod, G. et al. Genome-wide CRISPR\u2013Cas9 screens expose genetic vulnerabilities and mechanisms of temozolomide sensitivity in glioblastoma stem cells. Cell Rep.","27, 971\u2013986 (2019).\r\n\r\nArticle\nCAS\nPubMed\nGoogle Scholar\r\n\r\nKuzmin, D. A. et al. The clinical landscape for AAV gene therapies. Nat. Rev. Drug Discov.","20, 173\u2013174 (2021).\r\n\r\nArticle\nCAS\nPubMed\nGoogle Scholar\r\n\r\nFarnham, P. J. Insights from genomic profiling of transcription factors. Nat. Rev. Genet.","10, 605\u2013616 (2009).\r\n\r\nArticle\nCAS\nPubMed\nPubMed Central\nGoogle Scholar\r\n\r\nSoufi, A.,","Donahue, G.","& Zaret, K. S. Facilitators and impediments of the pluripotency reprogramming factors\u2019 initial engagement with the genome.","Cell 151, 994\u20131004 (2012).\r\n\r\nArticle\nCAS\nPubMed\nPubMed Central\nGoogle Scholar\r\n\r\nStricker, S. H. et al. Widespread resetting of DNA methylation in glioblastoma-initiating cells suppresses malignant cellular behavior in a lineage-dependent manner. Genes Dev.","27, 654\u2013669 (2013).\r\n\r\nArticle\nCAS\nPubMed\nPubMed Central\nGoogle Scholar\r\n\r\nHuang, Y.-H.,","Jankowski, A.,","Cheah, K. S.","E.,","Prabhakar, S.","& Jauch, R. SOXE transcription factors form selective dimers on non-compact DNA motifs through multifaceted interactions between dimerization and high-mobility group domains. Sci. Rep.","5, 10398 (2015).\r\n\r\nArticle\nADS\nPubMed\nPubMed Central\nGoogle Scholar\r\n\r\nScott, C. E. et al. SOX9 induces and maintains neural stem cells. Nat. Neurosci.","13, 1181\u20131189 (2010).\r\n\r\nArticle\nCAS\nPubMed\nGoogle Scholar\r\n\r\nGangoso, E. et al. Glioblastomas acquire myeloid-affiliated transcriptional programs via epigenetic immunoediting to elicit immune evasion.","Cell 184, 2454\u20132470 (2021).\r\n\r\nArticle\nCAS\nPubMed\nPubMed Central\nGoogle Scholar\r\n\r\nWang, Q. et al. Tumor evolution of glioma-intrinsic gene expression subtypes associates with immunological changes in the microenvironment. Cancer Cell 32, 42\u201356 (2017).\r\n\r\nArticle\nPubMed\nPubMed Central\nGoogle Scholar\r\n\r\nNeftel, C. et al. An integrative model of cellular states, plasticity, and genetics for glioblastoma.","Cell 178, 835\u2013849 (2019).\r\n\r\nArticle\nADS\nCAS\nPubMed\nPubMed Central\nGoogle Scholar\r\n\r\nYan, Y.-L. et al. A pair of Sox: distinct and overlapping functions of zebrafish sox9 co-orthologs in craniofacial and pectoral fin development. Development 132, 1069\u20131083 (2005).\r\n\r\nArticle\nCAS\nPubMed\nGoogle Scholar\r\n\r\nSantiago-Ortiz, J. L. & Schaffer, D. V. Adeno-associated virus (AAV) vectors in cancer gene therapy. J. Control. Release 240, 287\u2013301 (2016).\r\n\r\nArticle\nCAS\nPubMed\nPubMed Central\nGoogle Scholar\r\n\r\nSwift, E.","A.,","Pollard, S. M.","& Parker, A. L.","Engineering cancer selective virotherapies: are the pieces of the puzzle falling into place?","Hum. Gene Ther.","33, 1109\u20131120 (2022).\r\n\r\nArticle\nCAS\nPubMed\nPubMed Central\nGoogle Scholar\r\n\r\nLasek, W., Zago\u017cd\u017con, R. & Jakobisiak, M. Interleukin 12: still a promising candidate for tumor immunotherapy?\u00a0Cancer Immunol. Immunother.","63, 419\u2013435 (2014).\r\n\r\nArticle\nCAS\nPubMed\nPubMed Central\nGoogle Scholar\r\n\r\nAgliardi, G. et al. Intratumoral IL-12 delivery empowers CAR-T cell immunotherapy in a pre-clinical model of glioblastoma. Nat. Commun.","12, 444 (2021).\r\n\r\nArticle\nADS\nCAS\nPubMed\nPubMed Central\nGoogle Scholar\r\n\r\nHacker, U.","T.,","Bentler, M.,","Kaniowska, D.,","Morgan, M.","& B\u00fcning, H. Towards clinical implementation of adeno-associated virus (AAV) vectors for cancer gene therapy: current status and future perspectives. Cancers 12, 1889 (2020).\r\n\r\nArticle\nCAS\nPubMed\nPubMed Central\nGoogle Scholar\r\n\r\nAldape, K. et al. Challenges to curing primary brain tumours. Nat. Rev. Clin. Oncol.","16, 509\u2013520 (2019).\r\n\r\nArticle\nCAS\nPubMed\nPubMed Central\nGoogle Scholar\r\n\r\nDarmanis, S. et al. Single-cell RNA-seq analysis of infiltrating neoplastic cells at the migrating front of human glioblastoma. Cell Rep.","21, 1399\u20131410 (2017).\r\n\r\nArticle\nCAS\nPubMed\nPubMed Central\nGoogle Scholar\r\n\r\nArrieta, V. A. et al. Immune checkpoint blockade in glioblastoma: from tumor heterogeneity to personalized treatment. J. Clin. Invest.","133, e163447 (2023).\r\n\r\nArticle\nCAS\nPubMed\nPubMed Central\nGoogle Scholar\r\n\r\nLiau, L. M. et al. Association of autologous tumor lysate-loaded dendritic cell vaccination with extension of survival among patients with newly diagnosed and recurrent glioblastoma: a phase 3 prospective externally controlled cohort trial. JAMA Oncol.","9, 112\u2013121 (2022).\r\n\r\nArticle\nGoogle Scholar\r\n\r\nTodo, T. et al. Intratumoral oncolytic herpes virus G47\u2206 for residual or recurrent glioblastoma: a phase 2 trial. Nat. Med.","28, 1630\u20131639 (2022).\r\n\r\nArticle\nCAS\nPubMed\nPubMed Central\nGoogle Scholar\r\n\r\nO\u2019Dwyer, M. R. et al. Nucleosome fibre topology guides transcription factor binding to enhancers. Nature 638, 251\u2013260 (2025).\r\n\r\nArticle\nADS\nPubMed\nGoogle Scholar\r\n\r\nCompany, C. et al. Logical design of synthetic cis-regulatory DNA for genetic tracing of cell identities and state changes. Nat. Commun.","15, 897 (2024).\r\n\r\nArticle\nADS\nCAS\nPubMed\nPubMed Central\nGoogle Scholar\r\n\r\nTaskiran, I. I. et al. Cell-type-directed design of synthetic enhancers. Nature 626, 212\u2013220 (2022).\r\n\r\nArticle\nADS\nGoogle Scholar\r\n\r\nde Almeida, B. P. et al. Targeted design of synthetic enhancers for selected tissues in the Drosophila embryo. Nature 626, 207\u2013211 (2024).\r\n\r\nArticle\nADS\nPubMed\nGoogle Scholar\r\n\r\nMartella, A., Matjusaitis, M.,","Auxillos, J.,","Pollard, S. M.","& Cai, Y. EMMA: an extensible mammalian modular assembly toolkit for the rapid design and production of diverse expression vectors. ACS Synth. Biol.","6, 1380\u20131392 (2017).\r\n\r\nArticle\nCAS\nPubMed\nGoogle Scholar\r\n\r\nMcLean, C. Y. et al. GREAT improves functional interpretation of cis-regulatory regions. Nat. Biotechnol.","28, 495\u2013501 (2010).\r\n\r\nArticle\nADS\nCAS\nPubMed\nPubMed Central\nGoogle Scholar\r\n\r\nHeinz, S. et al. Simple combinations of lineage-determining transcription factors prime cis-regulatory elements required for macrophage and B cell identities. Mol.","Cell 38, 576\u2013589 (2010).\r\n\r\nArticle\nCAS\nPubMed\nPubMed Central\nGoogle Scholar\r\n\r\nMa, W.,","Noble, W. S.","& Bailey, T. L. Motif-based analysis of large nucleotide data sets using MEME-ChIP. Nat. Protoc.","9, 1428\u20131450 (2014).\r\n\r\nArticle\nCAS\nPubMed\nPubMed Central\nGoogle Scholar\r\n\r\nSoufi, A. et al. Pioneer transcription factors target partial DNA motifs on nucleosomes to initiate reprogramming.","Cell 161, 555\u2013568 (2015).\r\n\r\nArticle\nCAS\nPubMed\nPubMed Central\nGoogle Scholar\r\n\r\nKawakami, K. Tol2: a versatile gene transfer vector in vertebrates. Genome Biol.","8, S7 (2007).\r\n\r\nArticle\nPubMed\nPubMed Central\nGoogle Scholar\r\n\r\nKwan, K. M. et al. The Tol2kit: a multisite gateway-based construction kit for Tol2 transposon transgenesis constructs. Dev. Dyn.","236, 3088\u20133099 (2007).\r\n\r\nArticle\nCAS\nPubMed\nGoogle Scholar\r\n\r\nChia, K.,","Mazzolini, J.,","Mione, M.","& Sieger, D. Tumor initiating cells induce Cxcr4-mediated infiltration of pro-tumoral macrophages into the brain.","eLife 7, e31918 (2018).\r\n\r\nArticle\nPubMed\nPubMed Central\nGoogle Scholar\r\n\r\nWorkman, P. et al.\u00a0Guidelines for the welfare and use of animals in cancer research. Br. J. Cancer 102, 1555\u20131577 (2010).\r\n\r\nKishi, J. Y. et al. SABER amplifies FISH: enhanced multiplexed imaging of RNA and DNA in cells and tissues. Nat. Methods 16, 533\u2013544 (2019).\r\n\r\nArticle\nCAS\nPubMed\nPubMed Central\nGoogle Scholar\r\n\r\nWang, S.","K.,","Lapan, S.","W.,","Hong, C.","M.,","Krause, T. B.","& Cepko, C. L. In situ detection of adeno-associated viral vector genomes with SABER-FISH. Mol. Ther. Methods Clin. Dev.","19, 376\u2013386 (2020).\r\n\r\nArticle\nPubMed\nPubMed Central\nGoogle Scholar\r\n\r\nPurshouse, K. et al. Oncogene expression from extrachromosomal DNA is driven by copy number amplification and does not require spatial clustering in glioblastoma stem cells.","eLife 11, e80207 (2022).\r\n\r\nArticle\nCAS\nPubMed\nPubMed Central\nGoogle Scholar\r\n\r\nHamdan, A.","& Wang, Z. alhafidzhamdan\/sse_gene_therapy: For Koeber et al., Nature 2026 (1.0). Zenodo https:\/\/doi.org\/10.5281\/zenodo.18676096 (2026).\r\n\r\nDownload references\r\nAcknowledgements\r\nU.K.,","M.M., S.J.R. and S.M.P. are supported by the BBSRC UK Mammalian Synthetic Biology Centre (BB\/M018040\/1). The GSC and NSC lines were obtained from the GCGR. The GCGR was funded by a Cancer Research UK (CRUK) Accelerator Award (C157\/A21992). A.W. was supported by a BBSRC funded project grant. A.H. was supported by CRUK and the Edinburgh Clinical Academic Training (ECAT) programme (C157\/A29279). D.S. is supported by a CRUK programme foundation award. S.M.P. is supported by CRUK programme grant (DRCNPG-Nov21\\100002). S.A.Q. is funded by a CRUK Senior Cancer Research Fellowship (C36463\/A22246) and a CRUK Biotherapeutic Program grant (C36463\/A20764).","This work was undertaken at University College London with support from the CRUK\u00a0City of London Centre (C7893\/A26233). S.M.P. and S.A.Q. also received support from the CRUK Brain Tumour Award (A28592). We thank J.\u2009Mendrychowski and V.\u2009Kho for technical support; I.\u2009Chambers and J.\u2009M.\u2009Iglesias for their comments on the manuscript; and A.\u2009Corsinotti for technical advice on scRNA-seq.","For open access, the author has applied a Creative Commons Attribution (CC BY) licence to any Author Accepted Manuscript version arising from this submission.\r\nAuthor information\r\nAuthor notes\r\n\r\nUte Koeber\r\n\r\nPresent address: IQVIA RDS, Frankfurt am Main, Germany\r\n\r\nMantas Matjusaitis\r\n\r\nPresent address: Plurify, Cambridge, UK\r\n\r\nRachel White,\u00a0Michelle P.","Clark,\u00a0Carla Jacinto de Sousa\u00a0&\u00a0Ben Southgate\r\n\r\nPresent address: Trogenix, Edinburgh, UK\r\n\r\nFelipe Galvez Cancino\r\n\r\nPresent address: Laboratory of Immune Regulation, NDM Centre for Immuno-Oncology, University of Oxford, Oxford, UK\r\n\r\nThese authors contributed equally: Ute Koeber, Mantas Matjusaitis\r\nAuthors and Affiliations\r\nCentre for Regenerative Medicine, Institute for Regeneration and Repair, University of Edinburgh, Edinburgh, UK\r\n\r\nUte Koeber,\u00a0Mantas Matjusaitis,\u00a0Neza Alfazema,\u00a0Katharine Furlong,\u00a0Zeyu Wang,\u00a0Rachel White,\u00a0Alhafidz Hamdan,\u00a0Pooran Dewari,\u00a0Rosie Willis,\u00a0Michelle P.","Clark,\u00a0Carla Jacinto de Sousa,\u00a0Hei Ip Hong,\u00a0Shahida Sheraz,\u00a0Ben Southgate,\u00a0Justyna Cholewa-Waclaw,\u00a0Sabine Gogolok,\u00a0Gillian M.","Morrison,\u00a0Faye Robertson,\u00a0Anna Williams,\u00a0Paul M.","Brennan,\u00a0Abdenour Soufi\u00a0&\u00a0Steven M. Pollard\r\n\r\nCancer Research UK Scotland Centre, Edinburgh, UK\r\n\r\nUte Koeber,\u00a0Mantas Matjusaitis,\u00a0Neza Alfazema,\u00a0Zeyu Wang,\u00a0Rachel White,\u00a0Alhafidz Hamdan,\u00a0Pooran Dewari,\u00a0Gregoire Morisse,\u00a0Rosie Willis,\u00a0Jin Wang,\u00a0Michelle P.","Clark,\u00a0Carla Jacinto de Sousa,\u00a0Hei Ip Hong,\u00a0Shahida Sheraz,\u00a0Ben Southgate,\u00a0Justyna Cholewa-Waclaw,\u00a0Sabine Gogolok,\u00a0Gillian M.","Morrison,\u00a0Faye Robertson,\u00a0Paul M.","Brennan,\u00a0Dirk Sieger\u00a0&\u00a0Steven M. Pollard\r\n\r\nEdinburgh Pathology, Royal Infirmary Edinburgh, NHS Lothian, Edinburgh, UK\r\n\r\nAlhafidz Hamdan\r\n\r\nInstitute for Neuroscience and Cardiovascular Research, University of Edinburgh, Edinburgh, UK\r\n\r\nGregoire Morisse,\u00a0Jin Wang\u00a0&\u00a0Dirk Sieger\r\n\r\nImmune Regulation and Tumor Immunotherapy Group, Cancer Immunology Unit, Research Department of Haematology, UCL Cancer Institute, London, UK\r\n\r\nMariela Navarette,\u00a0Felipe Galvez Cancino\u00a0&\u00a0Sergio A. Quezada\r\n\r\nSchool of Biological Sciences, University of Edinburgh, Edinburgh, UK\r\n\r\nSusan J. Rosser\r\n\r\nCentre for Clinical Brain Sciences, University of Edinburgh, Edinburgh, UK\r\n\r\nPaul M. Brennan\r\nContributions\r\nS.M.P.","conceived the study, secured funding and supervised the project. U.K. designed and performed in vitro experiments to construct the SSEs, with bioinformatic support from B.S. M.M. designed and performed in vitro experiments related to the therapeutic transgene experiments and selectivity, and with N.A. in the animal efficacy studies. A.W. provided human OPCs. M.P.C. led the human in vitro selectivity experiments. The ChIP\u2013seq analysis was performed by P.D., and A.H. led the bioinformatic analyses. Z.W. designed and performed the scRNA-seq experiments. R.\u00a0Willis performed SABER\u2013FISH.\u00a0K.F. performed the biochemical analysis of SOX interactions, with supervision from A.S.","Key technical support for all the molecular biology was provided by R.\u00a0White and S.G. All zebrafish studies were performed by G.M.","and J.W., with design and supervision from D.S. Manufacture and quality control of all viral vectors was performed by M.M., R.\u00a0White and C.J.d.S. The inhibitor screen was designed and performed by H.I.H., with supervision from S.M.P. The human tissue slices were provided by P.M.B., with AAV testing performed by R.\u00a0Willis, S.S., M.P.C. and J.C.-W. The initial support for the transgene optimization and patient-derived GSC models was from S.G. and G.M.M. F.R. supported mouse model experimental designs and translational relevance. The plasmid library manufacture was supported by S.J.R. and staff at the Edinburgh Genome Foundry. All immune characterization was designed and performed by M.N. and F.G.C., with supervision from S.A.Q.","The manuscript and figures were drafted by S.M.P., with contributions from all authors. All authors reviewed, edited and approved the final version.\r\nCorresponding author\r\nCorrespondence to\nSteven M. Pollard.\r\nEthics declarations\nCompeting interests\r\nS.M.P., U.K. and M.M. are inventors of patents filed by the University of Edinburgh related to the discoveries in this study and to SSEs and associated sequences (WO2022162361A1). S.M.P. is founder of Trogenix, which is developing novel gene therapies for oncology, and is a paid consultant (part-time chief scientific officer, since February 2024) and shareholder.","B.S.,","R.\u00a0White, C.J.d.S. and M.P.C. have recently joined Trogenix as employees (since October 2025). S.A.Q. is a paid member of the scientific advisory board of Trogenix. P.M.B. and F.R. are paid external consultants for Trogenix related to ongoing clinical translation of these technologies. All other authors (N.A.,","K.F.,","Z.W.,","A.H.,","P.D.,","G.M.,","M.N., R.\u00a0Willis.,","J.W.,","H.I.H.,","S.S.,","J.C.-W.,","S.G.,","G.M.M.,","F.G.C.,","A.W.,","S.J.R., D.S. and A.S.) declare no competing interests.\r\n\nPeer review\nPeer review information\r\nNature thanks Ralf Jauch and the other, anonymous, reviewer(s) for their contribution to the peer review of this work.\r\n\nAdditional information\r\nPublisher\u2019s note Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.\r\nExtended data figures and tablesExtended Data Fig. 1 Design of candidate SOX2 GSC-specific peaks and their conversion into a library of candidate enhancer fragments.\r\n(a) Summary of\u00a0the workflow to design and build the enhancer fragment library (DGC: differentiated glioma cell). (b) The genomic annotation of the final library of candidate peaks.","(c) The top 16 fragment hits identified from the initial screen (Fig. 1) were validated in GSC7 GSCs for mNeongreen expression using flow cytometry (median fluorescence intensity shown (left). (d) The fragments have the same level of expression in HEK negative control cells as the minimal CMV (MFI and % of total cell population shown, left and right, respectively). (e) De novo motif analysis using HOMER for the Top32 hits of the validation (using shuffled sequences as background) from Fig.","1, confirms the repertoire of candidate TF motifs that are enriched in the functional enhancer fragments.\r\n\r\nSource data\r\nExtended Data Fig. 2 A dimeric SOX motif that binds SOX9 is a major contributor to activity.\r\n(a) A specific dimeric SOX dimer motif was identified using MEME for de novo motif discovery and is enriched in the set of functional enhancer hits (ACAAAGRGBVNHTKK) and was mutated as depicted (below) for 10 different variants of the ID2904 sequence to determine the functional impact. (b-e) Plasmid-based flow cytometry assay to determine the activity of each variant, in GSCs (GSC7 and E28, respectively).","Interestingly, increased activity was seen with 2\u2009bp reduction of the spacing consistent with the importance of this motif for activity. Each dot represents a biological replicate (2e) One-way Anova test with Dunnett\u2019s multiple comparisons test against ID2904 WT; ** represent p-value of 0.0040 (f) Examples of flow cytometry data used for panels c-f. M(f) Examples of flow cytometry data used for panels c-f. (g) SOX9 protein selectively interacts with ID1101 and ID2904 through the dimer motif.\r\n\r\nSource data\r\nExtended Data Fig.","3 Assessment of SSE-7 activity across a set of seven different patient-derived GSC lines.\r\n(a) Reporter reference controls were used to determine relative SSE activity by flow cytometry across each cell line, with minimal CMV (mCMV) and CMV as negative and positive controls, respectively.","(c, e, g, i, k, m and o) shows the % mCherry positive cells, whereas (d, f, h, j, l, n and p) shows the median fluorescence intensity (MFI) for each cell. E37 is an outlier line and has an atypical MYC amplification and lower levels of SOX2. Error bars represent the SD of the mean, each dot represents a biological replicate.\r\n\r\nSource data\r\nExtended Data Fig. 4 Characterisation of SSEs.\r\n(a) Comparison of activity from all seven SSE variants in GSC7 using flow cytometry to monitor expression of the mNeonGreen reporter. Each dot represents a biological replicate. (b) Time-course of GSC astrocyte differentiation to monitor SSE activity using flow cytometry.","(n\u2009=\u20092 biological replicates, each dot symbol represents the mean, the error bars represent the SD of the mean) (c) Live cell wide-field immunofluorescence imaging for the cells assessed in (b). (d-e) Quantitation of the band intensities from the EMSA experiments in Fig.","2d,e. Data points are mean from n\u2009=\u20093 with error bars of the SD. Solid line represents the Hill equation with the Kd values displayed above the graph. (f) EMSA of SSE-7 DNA using purified recombinant SOX2 and SOX9, alone or in combination (representative image of n\u2009=\u20093). (g) EMSA showing the individual and combinatorial binding of purified recombinant SOX2 and SOX9 to FGF4 enhancer element (40\u2009bp, n\u2009=\u20091) containing a typical SOX2 motif.\r\n\r\nSource data\r\nExtended Data Fig. 5 Single cell mRNA profiling of multiple patient derived GSC lines following transduction with AAV1-SSE-7-mCherry virus.\r\n(a) Distribution of transcriptional subtypes in scRNA data.","UMAP of transcriptional subtype in E17 (n\u2009=\u200929,829 cells), E20 (n\u2009=\u200923,301 cells), E31 (n\u2009=\u200939,823 cells), E34 (n\u2009=\u200927,766 cells), E43 (n\u2009=\u200931,538 cells), E55 (n\u2009=\u200947,291 cells). *P\u2009<\u20090.05, **P\u2009<\u20090.01, ***P\u2009<\u20090.001; NS, not significant. (b) SOX2 and SOX9 are broadly expressed across the diverse transcriptional subtype cell states in GSCs, consistent with their roles as core regulators. See also Extended Data Fig. 6 for cell cycle analysis.\r\nExtended Data Fig. 6 Single cell analysis of cell cycle and SCENIC analysis.\r\n(a) Distribution of cell cycle phases across GSC cell states in scRNA data (as per Extended Data Fig. 5).","(b) Full data for SCENIC analysis of transcription factor modules identifies signalling endpoint transcription factors alongside SOX as correlating with SSE-7 activity. (c) Heatmap across technical replicates of libraries from Fig. 3i.\r\nExtended Data Fig. 7 A screen of small molecule pharmacological inhibitors to identify signalling pathways that may impact the activity of SSE-7 in E55 GSCs.\r\n(a) The primary screen was performed in the E55 cells. Heatmap summarising the fold change of the normalised mCherry MFI in inhibitor-treated samples compared with DMSO. Initial hits identified with red font. The relative comparator is the CMV-SSE7-mCherry positive control.","Any reduction in mCherry selectively in the E55 is due to the different properties of the SSE7. (n\u2009=\u20092 biological replicate; independent screens) (b) Validation of the hit compounds using flow cytometry for mCherry detection across three different concentrations of compound (0.1, 0.3 or 3\u03bcM). The heatmap shows the fold change of the normalised mCherry MFI and percentages of mCherry positive cells of inhibitor-treated samples relative to the DMSO-treated control. (n\u2009=\u20092 biological replicates). Total % mCherry cells (top panel) and mean fluorescence intensity (MFI) bottom panel. Red font are MAPK signalling-related compounds validated with dose response and show selective effects on the SSE-7 compared to CMV promoter.","(c) Flow cytometry histograms for each validated hit compound. Typical example of flow cytometry overlay histograms for E55 GSCs treated with 3 different doses of MAPK signalling-related inhibitors (left panels) and E31 GSCs (right panels). E55 and E31 have distinct transcriptional subtype identities (mesenchymal and developmental\/proneural, respectively). (d) Summary of validated inhibitor hits (red font) and\u00a0their mechanism of action within the MAPK signalling pathway. (e) Flow cytometry analysis of positive and negative SSE-7mCherry cells sorted and reanalysed over a 72-hr time-course shows that SSE-7 activity can be activated in cells that are initially negative, indicative of a readout that is sensitive to signalling state.","n\u2009=\u20093 biological replicates. 2-way ANOVAs followed by two stage setup method were carried out to compare with mCMV neg. *P\u2009<\u20090.05, **P\u2009<\u20090.01, ***P\u2009<\u20090.001; NS, not significant. Note: SC-1 was toxic at the higher doses for E31 and so could not be assessed.\r\n\r\nSource data\r\nExtended Data Fig. 8 Selectivity in tissue expression in transgenic zebrafish embryos.\r\n(a) Reporter transgene for selective expression of SSEs in a zebrafish transgenic reporter model with a GFP reporter, with a transposase (Tol2) reporter designed track transgene insertion using mCherry expression in the eye from the Crystal promoter.","(b) SSE activity is tracked by eGFP levels in live cell imaging of the embryos at 48hpf and is tissue-restricted compared parallel CMV controls for each SSE tested. Note: the minCMV control exhibited GFP-expression in the eyes due to spurious transcription from the Crystal eye promoter. (c) A stable transgenic SSE-7:eGFP reporter zebrafish line confirms expression is restricted to the developing nervous system. Images were acquired using a Leica MS205 stereomicroscope. Scale bars represent 500\u2009\u00b5m for (b) and 728\u2009\u00b5m for (c). (d) Higher magnification image of the spinal cord and nototchord with neuronal reporter (Cyan, as per (e)).","(e) Outcross of tgSSE-7:eGFP to tg(Xla.Tubb:DsRed) reveals no expression in neurons; in (f) outcross of tgSSE-7:eGFP to tg(olig1:mScarlet) shows expression in a subset of oligodendrocyte precursor cells. Images represent maximum intensity projections of confocal stacks (optic tectum and cerebellum shown). Images were captured using a Leica SP8 confocal microscope with a 40X\/NA 1.1 objective. Scale bars represent 30\u2009\u00b5m. (g) Quantification of the OPC data from (f). (h)\u00a0Overexpression of human Akt1 in neuronal cells results in strong activity of\u00a0SSE-7. Images represent maximum intensity projections of confocal stacks. Images were captured using a Leica SP8 confocal microscope with a 40X\/NA 1.1 objective. Scale bars represent 30\u2009\u00b5m.","(i) Immunostaining and flow cytometry (top and bottom, respectively) of human ESC differentiated oligodendrocyte progenitor cells shows that SSE-7 is not active compared to CMV (n\u2009=\u20091). (j) SSE7 versus CMV across a set of three human foetal neural stem cell (NSC) cultures (derived from week of gestation 9, 12 and 17 tissue; NS9FB_B, NS12ST_A and NS17ST_A, respectively). Each dot represents a biological replicate. Error bars represent the SD of the mean.\r\n\r\nSource data\r\nExtended Data Fig. 9 Selective activity of AAV1-SSE7-mCherry in SOX2 expressing cells within the tumour tissue mass and infiltrating margin.\r\n(a)\u00a0Tissue slice-culture experiments were performed across diverse subtypes of high-grade glioma similarly to Fig. 2.","Immunostaining and confocal for mCherry across distinct tumour subtypes shows restricted activity of SSE-7-mCherry in\u00a0the tumour margin relative to CMV. (b) Quantitation of the % of mCherry expression in the tumour versus the margin shows SSE-7 is activated in a subset of GBM cells. (n\u2009=\u20094 for both tumour and margin with both CMV and SSE7). Matched tumour and margin samples were used across multiple tumour types; GBM IDH-WT (isocitrate dehydrogenase wild-type), Grade 4 Astrocytoma IDH-MUT (isocitrate dehydrogenase mutant), Grade 3 high-grade astrocytoma and denoted by distinct symbols, each point is a biological replicate. Error bars represent SD.","Statistical significance was assessed by two-way ANOVA with multiple-comparisons correction (p\u2009=\u20090.0182). n\u2009=\u20094. Image analysis and segmentation pipelines provided in the Supplementary Data. Scale bar = 100 micron. (c) Staining for mCherry alongside SOX2 and Nestin in a recurrent GBM\u00a0sample treated with the AAV1-SSE7-mCherry vector, confirming that SSE-7 is active in a subset of the tumour cells. Left panels, low magnification (no matched normal tissue was available for this recurrent sample).\r\n\r\nSource data\r\nExtended Data Fig. 10 In vitro killing of GSCs, but not fibroblasts, is achieved using SSE7-driven HSV-TK.\r\n(a) Schematic of the AAV-2 transgene encoding both HSV-tk and mCherry used in vitro to assess selective cytotoxicity.","(b and c) Flow cytometry was performed 4 days post-transduction. Each dot represents biological replicate (n\u2009=\u20093), error bars represent SD. (b) % mCherry positive cells and (c) median fluorescent intensity (MFI) of mCherry positive cells. (d) Western blot image developed against V5 and GAPDH in transduced and non-transduced GSC7 and huFB170 cells; the higher weaker band is the small fraction that was uncleaved. (e) Incucyte live cell imaging generate proliferation curves to determine the change in confluence over 10-day time-course for human GSC7 cells transduced with CMV or SSE-7-driven TKv5-P2A-mCherry in presence of 200\u2009\u00b5M GCV, 0.2% DMSO (vehicle control) or 20% DMSO (positive control).","(f; top panels) Live images of GSC7 cells at day 8 demonstrating cell number and morphological changes in cells transduced with constructs in the presence of 200\u2009\u00b5M GCV. (f; bottom panels) Live images of huFB170 cells at day 17 demonstrating cell number and morphological changes in cells transduced with constructs in the presence of 200\u2009\u00b5M GCV. (g) MTT assay for cell viability in GSC-7 and fibroblasts. Each dot represents an independent biological replicate. MTT values were normalized to vehicle-only control. Error bars represent SD (n\u2009=\u20093).\r\n\r\nSource data\r\nExtended Data Fig.","11 Evaluating SSE7-TKv5-P2A-mCherry activity from an AAV vector.\r\n(a) Example of flow cytometry data for GSCs (E17) cells following transduction at MOI of 5E5. (b) Heatmap summarizing mean AAV transduction efficiency in different patient-derived GSC cultures (n\u2009=\u20093, except E34 and E21, which are n\u2009=\u20092). (c) SSE-7 activity in mouse fibroblasts and GSCs. (c) Flow cytometry data comparing transcriptional activity between SSE-7- and CMV-driven TKv5-mCherry cassette in mouse NPE-IE cells (GSCs) and NIH3T3 cells when delivered using rAAV2. (d) MTT viability assay demonstrating GCV dose response kill curves in NPE-IE cells transduced with rAAV carrying specified constructs.\r\n\r\nSource data\r\nExtended Data Fig.","12 Treatment of mice with AAV-SSE7-TK-IL12 results in tumour clearance and long-term protection against rechallenge.\r\n(a) Bioluminescence imaging to track tumour growth with either control (PBS) injection (right panels) or virus plus 3-week ganciclovir treatment (shown are n\u2009=\u200910 mice from the first cohort of mice in Fig. 5b). Right panel are a cohort of parallel fresh controls using the same cells in new cohort of mice, confirming tumour initiation potential. Blue font indicates the new timeline after re-challenge with fresh tumour cells injected (male mice used). (b) Dose de-escalation using IVIS to track tumour growth following treatment with different doses of AAV1-SSE7-HSV-tk-IL12.","(male mice used).\r\n\r\nSource data\r\nExtended Data Fig. 13 Characterisation of immune cells with single or double payloads.\r\n(a) Survival curves of tumour-bearing animals for either single (IL-12) or combination payloads (IL-12 plus TK\/gcv), n\u2009=\u20097 male mice per\u00a0group. Survival was analysed using Kaplan-Meier method. Statistical significance was determined using log-rank (Mantel-Cox) test with p\u2009=\u20090.0009. (b) Non-tumour bearing mice treated with AAV-SSE7-TK-IL12 versus the CMV variant. n\u2009=\u20095 male mice per group. Survival was analysed using Kaplan-Meier method. Statistical significance was determined using log-rank (Mantel-Cox) test with p\u2009=\u20090.0034. The CMV driven payloads are toxic.","(c) Bar plots of the percentage of PD-L1+ cells in MDSCs, monocytes, BMDMs, DCs and microglia.","(d, e) Bar plots of the percentage of CD62L-CD44-, CD62L-CD44+, CD62L+CD44- and CD62L+CD44+ cells in T effector cells or CD8+\u2009T cells, respectively.","(f,g) Bar plots of the percentage of IFNg producing cells in T effector cells.","(h,i) Bar plots of the percentage of IFNg producing cells in CD8+\u2009T cells. (j) Flow cytometry analysis of tumor mass 7 days following AAV1-SSE7-mCherry transduction, reveals that ~20% of the NPE-IE-GFP tumor cell population has strong mCherry expression (left panel, gating of tumor cells; middle panel, mCherry histogram; right panel, quantitation n\u2009=\u20094 independent tumours); 2ul of 3.6E14\/ml (full dose) and 7.2E13\/ml (1:5 dose). (c-i) n\u2009=\u20097 biologically independent animals per group, examined in one experiment. Data are represented as mean values +\/\u2212 SD. Statistical significance was assessed using one-way ANOVA with Tukey\u2019s multiple comparison test.","*P\u2009\u2264\u20090.05; **P\u2009\u2264\u20090.01; ***P\u2009\u2264\u20090.001; ****P\u2009\u2264\u20090.0001.\r\n\r\nSource data\r\nExtended Data Table 1 Predicted target genes and distance from enhancer fragment to transcription start site (TSS)\r\nFull size table\r\nSupplementary informationSource dataRights and permissions\r\nOpen Access This article is licensed under a Creative Commons Attribution 4.0 International License, which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence, and indicate if changes were made.","The images or other third party material in this article are included in the article\u2019s Creative Commons licence, unless indicated otherwise in a credit line to the material. If material is not included in the article\u2019s Creative Commons licence and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this licence, visit http:\/\/creativecommons.org\/licenses\/by\/4.0\/.\r\n\r\nReprints and permissions\r\nAbout this article\r\n\r\nCite this article\r\nKoeber, U., Matjusaitis, M.,","Alfazema, N. et al. Synthetic super-enhancers enable precision viral immunotherapy.\nNature (2026). https:\/\/doi.org\/10.1038\/s41586-026-10329-6\r\n\r\nDownload citation\r\n\r\nReceived: 07 December 2023\r\n\r\nAccepted: 26 February 2026\r\n\r\nPublished: 08 April 2026\r\n\r\nVersion of record: 08 April 2026\r\n\r\nDOI: https:\/\/doi.org\/10.1038\/s41586-026-10329-6"],"images":[{"src":"\/news\/images\/TkbOmERl7ZELY4B6U8bwRFLp7W9qytfUVvStryoOOlUEGhhHL0eXn3oyzuSEao4y0AJ5ESqU9aly0hqeQlF1AuY6HehvMIasaugaW2HxU4xEjpq1kWjw2a5kegGpJxC3tseg50hmbqflgA3MX8A8XkYOCKH4f.jpg","width":0,"height":0,"source":"featured","size":0}]}"
[3]=>
int(1775731824)
}
}
[3]=>
array(7) {
["file"]=>
string(21) "/app/sys/news.inc.php"
["line"]=>
int(163)
["function"]=>
string(8) "saveData"
["class"]=>
string(6) "NewsDB"
["object"]=>
object(NewsDB)#4 (1) {
["db":"NewsDB":private]=>
object(SQL)#1 (6) {
["CONNECTION":"SQL":private]=>
object(mysqli)#2 (18) {
["affected_rows"]=>
int(-1)
["client_info"]=>
string(14) "mysqlnd 8.2.30"
["client_version"]=>
int(80230)
["connect_errno"]=>
int(0)
["connect_error"]=>
NULL
["errno"]=>
int(1406)
["error"]=>
string(40) "Data too long for column 'data' at row 1"
["error_list"]=>
array(1) {
[0]=>
array(3) {
["errno"]=>
int(1406)
["sqlstate"]=>
string(5) "22001"
["error"]=>
string(40) "Data too long for column 'data' at row 1"
}
}
["field_count"]=>
int(2)
["host_info"]=>
string(18) "db:3306 via TCP/IP"
["info"]=>
NULL
["insert_id"]=>
int(0)
["server_info"]=>
string(37) "11.2.2-MariaDB-1:11.2.2+maria~ubu2204"
["server_version"]=>
int(110202)
["sqlstate"]=>
string(5) "22001"
["protocol_version"]=>
int(10)
["thread_id"]=>
int(832256)
["warning_count"]=>
int(0)
}
["URL":"SQL":private]=>
string(7) "db:3306"
["USER":"SQL":private]=>
string(8) "retrofox"
["PASS":"SQL":private]=>
string(16) "TreeakWqQPD9qnWR"
["DB":"SQL":private]=>
string(8) "retrofox"
["DROP_VALUES":"SQL":private]=>
array(1) {
[0]=>
string(0) ""
}
}
}
["type"]=>
string(2) "->"
["args"]=>
array(2) {
[0]=>
string(85) "foxapi_article_SNgSFz1UYwUgnAye8Iq6AwimFnBOT9jdyRE0kXeBSMWDBytDmFUTOmTLbAFJl2KUpiOq4O"
[1]=>
string(135713) "{"id":"SNgSFz1UYwUgnAye8Iq6AwimFnBOT9jdyRE0kXeBSMWDBytDmFUTOmTLbAFJl2KUpiOq4O","title":"Synthetic super-enhancers enable precision viral immunotherapy","description":"Cell-type-specific promoters are used in gene therapy to restrict expression of the therapeutic payload. However, these promoters often have suboptimal strength, selectivity and size. Here, leveraging recent insights into the function of enhancers, we developed synthetic super-enhancers (SSEs) by assembling functionally validated enhancer fragments into multipart arrays. Focusing on the core SOX2-driven and SOX9-driven transcriptional regulatory network in glioblastoma stem cells (GSCs)1, we engineered SSEs with robust activity and high selectivity.\u00a0Single-cell profiling, biochemical analyses and genome-binding data indicated that SSEs integrate neurodevelopmental and signalling-state transcription factors to trigger the formation of large multimeric complexes of transcription factors. Moreover, GSC-selective expression of a combination of cytotoxic (HSV-TK and ganciclovir) and immunomodulatory (IL-12) payloads, delivered using adeno-associated virus vectors, as a single treatment led to curative outcomes in a mouse model of aggressive glioblastoma. Notably, IL-12 induced an immunological memory that prevented tumour recurrence. The activity and selectivity of the adeno-associated virus and SSE were validated using primary human glioblastoma tissue and normal cortex samples. In summary, SSEs harness the unique core transcriptional programs that define the GSC phenotype and enable precision immune activation. This approach may have broader applications in other contexts when precise control of transgene expression in specific cell states is necessary. Synthetic super-enhancers enable specific delivery of anticancer payloads, achieving tumour elimination after a single dose in a mouse model of aggressive glioblastoma.","author":"Pollard, Steven M.","source":"Nature","publish":"April 8, 2026, 11:28 pm","canonical":"https:\/\/www.nature.com\/articles\/s41586-026-10329-6","text":["Main\r\nGene therapies require the targeted expression of payloads in specific cell populations to achieve appropriate therapeutic dosing and to minimize off-target effects. Various strategies to attain this selectivity can be explored, including delivery routes, capsid engineering and regulatory elements (for example, promoters, enhancers, untranslated regions and termination signals)2. Enhancers determine cell-type-specific expression by recruiting complementary transcription factors (TFs) at high density via TF-binding motifs (TFBMs)3,4.","However, natural enhancers often have suboptimal features for application in gene therapy vectors, such as size, strength, selectivity and quality of the sequence (for example, GC-rich or repetitive sequences). Strategies for creating artificial promoters and enhancers have typically focused on screening small (around 10\u2009bp) synthetic TFBMs and assembling them into concatamers5,6,7.","However, this approach is inadequate owing to the lack of a natural TFBM grammar, that is, the spacing, order, orientation and affinity of TFBMs, which are necessary for cell-type selectivity8.\r\n\r\nCell-type-specific enhancers are often clustered in the genome to create super-enhancers9,10. Super-enhancers typically regulate transcriptional programs associated with cell identity through the local clustering of low-affinity motifs8. This leads to increased high-density binding of lineage and signalling TFs to promote RNA polymerase\u2009II recruitment and high transcriptional output11,12.","We propose that synthetic super-enhancers (SSEs) can be engineered by assembling natural enhancer fragments associated with different genes into multipart arrays and then functionally screening them to identify desired activity and selectivity. The resulting transgene regulatory elements would capture the fundamental grammar of the combinations of TFs that define a specific cell type and signalling state. Such SSEs could have substantial value in gene therapy applications.\r\n\r\nWe focused on glioblastoma (GBM) to test this platform. GBM is an incurable brain cancer that is driven by cells with fetal neural stem-like phenotypes termed GBM stem cells (GSCs).","These cells express high levels of master regulatory TFs associated with neural stem and progenitor cells, including SOX2 (ref. 13). SOX2 is a pioneer TF and reprogramming factor essential for inducing and maintaining the identity of neural stem cells (NSCs)14,15 and GSCs1,16,17,18. SOX2 is also a crucial master regulatory TF that is broadly expressed across diverse GSC subtypes18.\r\n\r\nHere we develop SSEs for the targeted expression of anticancer payloads in GSCs.","A single, locally delivered dose of adeno-associated virus (AAV)-SSE-719 was evaluated in an immunocompetent model of aggressive GBM. We demonstrate that tumours are safely cleared using this precise and controlled method to express a dual payload of a cytotoxic (herpes simplex virus thymidine kinase and ganciclovir (HSV-TK\/GCV)) and a cytokine (IL-12).\r\nGSC-specific enhancer fragment design\r\nTFs bind to thousands of sites in the genome, but only a small subset are functional enhancers20.","To identify candidate GSC-selective enhancers, we first re-analysed previously published datasets1 that mapped SOX2-binding sites and used chromatin immunoprecipitation with sequencing (ChIP\u2013seq) to define those specific to GSCs and lost in their differentiated progeny.","Of the 1,710 candidate GSC-specific peaks identified (range of 115\u20133,366\u2009bp, mean of 400\u2009bp), a library of 160-bp double-stranded DNA oligonucleotides, with overlapping 100\u2009bp segments, was synthesized (Extended Data Fig. 1). The 160\u2009bp length was chosen as this was deemed to be the optimum size to capture the core set of TFBMs with the necessary natural grammar and motif diversity that might confer selectivity.","Moreover, this length is compatible with nucleosome formation for pioneer factors such as SOX2 (ref. 21). Candidate fragments were amplified by PCR, cloned into a luciferase\u2013mNeonGreen reporter plasmid and assembled into an arrayed plasmid library (48\u2009\u00d7\u200996-well plates with 4,579 individual plasmids) (Fig. 1a).\r\nFig. 1: Functional enhancer screening of an arrayed plasmid library.\r\na, Strategy for enhancer fragment (160\u2009bp) functional screening using an arrayed library of 4,579 plasmids with luciferase (Luc;\u00a0yellow) and mNeonGreen (mNGreen; green) reporters (single mRNA linked by P2A; grey), with an mCMV promoter (light blue).","b, Results for the enhancer fragment screen in GSCs (G7 cells) (n\u2009=\u20091).","c, Validation of 135 hits from the primary screen using independent luciferase reporter assays (NanoLuc DLR assay) (n\u2009=\u20093 biological replicates, error bars represent the s.d. of the arithmetic mean). The inset shows the enriched SOX dimer motif identified using the MEME tool.","d,\u00a0Validation of the top 16 significant hits from c, using flow cytometry to determine the percentage of mNeonGreen reporter-positive cells in an independent GSC line (E28), and the MFI per cell for each fragment (right). mCMV and the full-length CMV promoter were used as negative and positive reference controls, respectively (each dot represents a biological replicate (n\u2009=\u20093), error bars represent the s.d. of the arithmetic mean). Ordinary one-way analysis of variance (ANOVA) with Dunnet\u2019s multiple comparison test against mCMV was performed. For the mNeonGreen (%) data, adjusted P\u00a0< 0.0001 for ID3146,\u00a0P\u2009=\u20090.0018 for ID1101\u00a0and\u00a0P\u00a0=\u00a00.0003 for ID2339.","In the MFI mNeonGreen panel, P\u00a0= 0.0171 for ID2706, P\u00a0= 0.0092 for ID3146,\u00a0P\u2009=\u20090.0038 for ID1101, P\u2009=\u20090.0034 for ID2339 and\u00a0P\u2009=\u20090.0037 for ID2904.","e, Negative control normal human dermal fibroblasts (huFib70) using the same assays as d (each dot represents a biological replicate (n\u2009=\u20093), error bars represent the s.d. of the arithmetic mean).\u00a0*P\u2009\u2264\u20090.05, **P\u2009\u2264\u20090.01, ***P\u2009\u2264\u20090.001, ****P\u2009\u2264\u20090.0001.\r\n\r\nSource data\r\nDiscovery of functional GSC enhancers\r\nPatient-derived GSCs can be routinely propagated in a monolayer\u00a0and can be efficiently transfected13. We screened the enhancer fragment plasmid library in GSCs (specifically GSC7 cells) using a luciferase assay to identify functional hits22. From this initial screen, 135 plasmids exhibited more than 10-fold activity over a minimal CMV (mCMV) promoter (Fig. 1b).","Several hits were identified recurrently, which provided confidence in the screening assay.\r\n\r\nSanger sequencing confirmed sequences of the 135 initial hits, and 32 unique sequences were validated in triplicate with >20-fold activity compared with mCMV (Fig.","1c, Extended Data Table 1 and Supplementary Table 1). The top 16 candidates were also validated using an independent flow cytometry assay to detect the mNeonGreen transcriptional reporter, which provides single-cell quantification of activity and mean fluorescence intensity (MFI). All 16 validated hits were active in an independent patient-derived GSC line (E28), but were inactive in fibroblasts and HEK293 cells, which were used as SOX2-negative controls (Fig.","1d,e and Extended Data Fig.","1c,d). As anticipated, these hit sequences contained TFBMs for many known neurodevelopmental or GBM-associated TF families, including motifs for SOX, bHLH, FOX, TCF and SMAD (Extended Data Fig. 1e).\r\n\r\nNotably, de novo motif searches using the MEME tool revealed that the 32 active fragments contained a specific SOX dimer motif that resembles a previously reported SOX9 motif23 (Fig. 1c). Although SOX2 typically operates as a monomer, SOX9\u2014a known NSC self-renewal factor24\u2014can function as a homodimer23.","This result suggests that the functional enhancer fragments potentially bind SOX9.\r\n\r\nTo functionally validate the importance of this motif in the enhancer fragments, we created a series of mutants in the dimer motif of ID2904 (which was chosen because it only has a single motif) with nucleotide substitutions, inversions or changes in spacing. Each of these alterations abolished the transcriptional activity of the fragment (Extended Data Fig. 2a\u2013f). Higher activity was observed when the distance between the half-sites was reduced by 2\u2009bp. This result suggests that dimer binding is a crucial feature of the activity of the motif and is probably important for SOX9 binding.","Indeed, using biochemical pull-down, we confirmed that SOX9 binds to ID2904 and that binding depends on the SOX dimer motif (Extended Data Fig. 2g).\r\n\r\nThe above findings led us to speculate that SOX2 and SOX9 might work together at shared target enhancers in GSCs. To find their putative shared target sites, we mapped genome-wide binding of SOX2 and SOX9 in seven independent GSC patient-derived cell lines using ChIP\u2013seq (Fig. 2a). Most SOX9 peaks co-bound with SOX2, and these overlapped with previously reported super-enhancers found across GSCs (Fig. 2b\u2013d). These co-bound peaks were located near genes involved in stem cell function (Fig.","2e), including essential regulators of neural stem and progenitor cells (for example, CDK6 and PTPRZ1; Fig. 2h). The expression of SOX2 and SOX9 was positively correlated, particularly in GBM (relative to most other cancers), which implied that these two factors have a disease-specific function and\/or regulatory interactions (Fig. 2f).","Notably, the SOX2 and SOX9 peaks in co-bound sites at enhancers were enriched with multiple SOX monomer and dimer motifs, respectively, with an optimal dimer motif spacing of 2\u20133\u2009bp (Fig.","2g,i). Altogether, these data indicate a specific requirement for SOX9 dimerization at loci in which SOX2 and SOX9 are co-bound.\r\nFig. 2: ChIP\u2013seq analysis of SOX2 and SOX9 binding in seven different patient-derived GSC cultures.\r\na, Schematic of SOX2 and SOX9 ChIP\u2013seq experimental design and analysis.\u00a0GO, gene ontology.","b, Venn diagram showing the overlap between SOX2 and SOX9 peaks for the sample E28 (left), and heatmaps of SOX2 and SOX9 binding intensity over SOX9 and SOX2 peaks, respectively (right), demonstrating co-binding events.","c, Bar plots showing the proportion of SOX9 peaks located in co-bound sites.","d, Venn diagrams showing the overlaps between consensus SOX2 and SOX9 peaks and co-bound SOX2\u2013SOX9 peaks with a consensus set of GSC super-enhancers (n\u2009=\u200944 publicly available GSC H3K27ac datasets;\u00a0Methods).","e, Gene ontology analysis of co-bound SOX2\u2013SOX9 peaks in super-enhancers.","f, Bar plot showing the distribution of SOX2\u2013SOX9 expression correlation values (y\u2009axis) across diverse cancer types from The Cancer Genome Atlas (x\u2009axis).","g, Bar plot showing the motif co-occurrences of inverted palindromic SOX motifs, grouped by the distances between the primary and secondary SOX motif pairings.","h, UCSC genome browser coverage tracks of SOX2 and SOX9 ChIP\u2013seq in the seven individual GSCs at PTPRZ1 and CDK6 loci, annotated by the presence of GSC super-enhancers, co-bound SOX2\u2013SOX9 peaks and the functional SOX dimer motif.\u00a0cCREs, candidate cis-regulatory elements; Cons 100 Verts, conservation across 100 vertebrates.","i, The distributions of centrally enriched de novo SOX monomer and dimer motifs at SOX2 and SOX9 peaks, respectively, restricted to those in co-bound SOX2\u2013SOX9 enhancers.\r\n\r\nIn summary, using functional screening in patient-derived GSCs, we identified candidate 160\u2009bp SOX2 and SOX9 co-bound enhancer fragments with high selectivity but low activity (relative to CMV). Together with previously published findings, such as genome-wide CRISPR screens18, these results highlight the importance of SOX2 and SOX9 as core components of a TF circuit that determines and sustains GSC identity.","Moreover, they share crucial transcriptional targets.\r\nSSEs for GBM\r\nOur goal was to create a GSC-selective transcriptional switch for cancer gene therapy that combines strong expression and high selectivity. Such a SSE would provide the necessary therapeutic window to target the tumour but spare surrounding normal CNS tissue. We proposed that combining functional and selective enhancer fragments into a larger artificial multipart assembly might increase activity by facilitating increased high-density recruitment of SOX2 and SOX9 alongside other signalling-associated transcriptional cofactors to mimic the mechanism of action of natural super-enhancers11,12.","We chose four-part assemblies, as this represents an optimal size compatible with AAV vector applications. These assemblies were constructed using sequences from the top\u200916 validated hits grouped into four sets: 1\u20134, 5\u20138, 9\u201312 and 13\u201316 (Fig.","3a,c).\r\nFig. 3: Combining SOX2 functional enhancer fragments results in synergistic increases in activity without compromising selectivity.\r\na, Activity of the 16 top hits relative to mCMV and CMV assessed by flow cytometry in GSCs (G7 cells) (each dot represents a biological replicate, error\u00a0bars represent the s.d. of the mean).","b, SSE activity detected using an mNeonGreen reporter across GSC lines (GSC7 and E28) and control fibroblasts. Error bars represent the s.d. of the mean. One-way ANOVA with Dunnett\u2019s multiple comparisons test against mCMV, adjusted P\u2009values for G7: P\u2009=\u20090.0001 for SSE-1, P\u2009=\u20090.0009 for SSE-3, P\u2009=\u20090.0003 for SSE-5 and P\u2009<\u20090.0001 for SSE-7. Adjusted P\u2009values for E28: P\u2009=\u20090.0099 for SSE-1, P\u2009=\u20090.0048 for SSE-3, P\u2009=\u20090.0075 for SSE-5 and P\u2009=\u20090.0008 for SSE-7.","UT, untreated.","c, Schematic of each SSE-1, SSE-3, SSE-5 and SSE-7 transgene design.","d, EMSA of SSE-7 DNA following incubation with GSC lysates compared with differentiated GSCs (representative image; n\u2009=\u20093).","e, EMSA showing SSE-7 DNA with fibroblast (hFib) lysate (left) and hFib with SOX2 and SOX9 expressed individually (middle) or in combination (right). Representative images shown (n\u2009=\u20093).","f, EMSA of SSE-7 DNA (left, representative images of n\u2009=\u20093) and its constituent fragments (right, n\u2009=\u20091), incubated with SOX2 and SOX9. Schematic of SOX motifs (thin black bar, partial SOX motif; thicker, SOX motif; thickest, dimer motif) is shown underneath the images.","g, SABER\u2013FISH analysis of transgenes. Representative images are shown (left), with quantification (right).","For SSE-7, mean copy number\u2009=\u2009167 (median, 119; Q1\u2013Q3, 76\u2013209; min\u2013max, 8\u2013806).","For CMV, mean copy number\u2009=\u2009115 (median, 87; Q1\u2013Q3, 57\u2013127; min\u2013max, 5\u2013761). Scale bars, 20\u00a0\u03bcm.","h, scRNA-seq analysis shows no correlation between SSE-7 activity transcriptional subtypes (AC, astrocyte; MES, mesenchymal; NPC, neural progenitor cell).\u00a0Neftel subtype, subtype as identified by Neftel et al.27. Distribution of SSE-7\u2013mCherryhigh cells using UMAP plots of scRNA-seq analysis in E21 cells (n\u2009=\u200923,384 cells) and E28 cells (n\u2009=\u200935,879).","i, SCENIC analysis to determine TF enrichment across replicate libraries. Combined Z\u2009scores from each library. *P\u2009\u2264\u20090.05, **P\u2009\u2264\u20090.01, ***P\u2009\u2264\u20090.001; NS, not significant. P\u2009values were calculated using repeated-measures ANOVA, followed by Bonferroni-adjusted pairwise comparisons of estimated marginal means. Boxes show the interquartile range, centre lines the median, and whiskers the minimum and maximum values.\r\n\r\nSource data\r\n\r\nEach variant displayed strong activity that matched or exceeded that of a full-length CMV promoter in GSCs. By contrast, the variants had low activity in both fibroblasts and HEK293 cells, such that levels were comparable to mCMV-like levels (Fig. 3b and Extended Data Fig. 4a).","Given the principles that underpin their design, their high activity (switch-like behaviour) and their high cell-type selectivity, we herein refer to these artificial elements as SSEs.\r\n\r\nTo initially explore selectivity in glial cell lineage differentiation, we evaluated SSE activity during GSC differentiation into astrocyte-like cells1,13. We observed a twofold reduction in MFI per cell within 24\u2009h, followed by an approximately tenfold decrease by day\u200910 of differentiation (Extended Data Fig.","4b,c). By contrast, the constitutive CMV promoter displayed high activity across all conditions. SSE activity is therefore highest in the immature stem cell-like state of GSCs and is rapidly reduced as cells differentiate.\r\n\r\nFocusing on SSE-7, we next tested a larger set of patient-derived GSCs that span diverse genetic and transcriptional subtypes25,26. SSE-7 was active across all seven patient-derived GSCs but had reduced levels in E37 (an atypical cell line with MYC amplifications and reduced SOX2 and SOX9 levels) (Extended Data Fig. 3).","However, there was variability across the cell lines in the percentage of positive cells in the population compared with CMV, which indicated heterogeneity in the cultures. This result was not explained by SOX2 or SOX9 protein levels, as these were similar across lines, which suggests that additional cooperating TFs might be necessary for SSE-7 activity.\r\n\r\nWe next explored the biochemical mechanism of the switch-like behaviour and selectivity of SSE-7 and the roles of SOX2 and SOX9 in this process.","To that end, we first used cell-free electrophoretic mobility shift assays (EMSAs) to analyse their interactions with SSE-7 alone and in combination using both cell lysates and cell-free recombinant SOX proteins.\r\n\r\nSSE-7 incubation with GSC cell extracts resulted in a major super-shift and the formation of large higher-order multimeric complexes, but this was not observed in GSCs that had differentiated into astrocytes (Fig. 3d). Using human fibroblasts with SOX2 and\/or SOX9 forced expression, we also observed a clear super-shifted band only when both factors were incubated together with SSE-7, but this effect was weaker compared with GSCs. This result highlights the importance of GSC-associated cofactors for triggering SSE activity (Fig.","3e).\r\n\r\nNext, to validate whether synergy exists with SOX2 and SOX9 co-binding, we used purified recombinant SOX2 and SOX9 proteins and tested their binding. When SOX2 alone was bound with SSE-7, it displayed a typical ladder-like banding pattern indicative of SOX2 binding individually at each motif in the 640-bp-long SSE-7 (Fig. 3f and Extended Data Fig. 4d\u2013f).","Notably, the binding of SOX9 led to the formation of the higher-order complexes, and this effect was enhanced in the presence of SOX2 (Fig. 3f). The analysis of each constituent 160\u2009bp enhancer fragment in isolation demonstrated that the formation of higher-order complexes depends on the architecture of the SSE (Fig. 3f). No higher-order complexes were formed when using a control 40\u2009bp FGF4 enhancer that contains a single SOX motif (Extended Data Fig. 4g).","We therefore conclude that a key feature of the mechanism of action of SSE-7 is its ability to induce higher-order TF complexes through high-density local recruitment of SOX2 and SOX9.\r\nMechanistic basis of SSE activity in GSCs\r\nWe next aimed to identify candidate SOX-cooperating TFs and the cofactors required for the SSE switch-like transcriptional activation seen in GSCs. To that end, we performed single-cell RNA sequencing (scRNA-seq) analysis of 2,711 SSE-7\u2013mCherry-positive cells across 8 different GSC lines, comparing these to matched mCherry-negative populations.","We first used the single-molecule imaging technique signal amplification by exchange reaction and fluorescence in situ hybridization (SABER\u2013FISH) and confirmed that transduction rates and nuclear transgene\u00a0copy number were similar among the different cell lines and that there was no correlation with mCherry levels (Fig. 3g).","Therefore, differential transduction or nuclear processing of the AAV does not explain the restricted SSE-7 activity observed in heterogeneous GSC cultures compared with CMV.\r\n\r\nUniform manifold approximation and projection (UMAP) analysis confirmed that these GSC cultures have, as expected, a diverse range of transcriptional subtypes or states27. This transcriptional heterogeneity mirrored that seen in patient tumour samples and reflected distinct differentiation states (Fig. 3h). We did not find a simple correlation of SSE-7high activity with previously reported transcriptional subtypes (Fig.","3h), SOX2 and SOX9 expression or cell cycle phase (Extended Data Figs. 5 and\u00a06a).\r\n\r\nTo perform an unbiased search for TF activities enriched in SSE-7high cells, we used the analysis tool SCENIC. The results revealed a clear correlation of multiple TF activities linked to cancer signalling pathways, including STAT, IRF, SMAD and FOS\u2013JUN (Fig. 3i). Such signalling end-point TFs are often highly activated in cancers, including GBM, and motifs for such TF families are present in SSE-7. This finding is consistent with the idea that these TFs\u00a0have a potential functional role (Fig. 3i and Extended Data Fig.","6b,c). These data suggest that SSE-7 is activated in cells that display a combination of activated signalling end-point TFs and an immature stem cell-like identity driven by SOX2 and SOX9.\r\n\r\nWe next explored the requirements needed for signalling pathway activation for SSE-7 function. We performed a screen of 160 different small-molecule kinase inhibitors, spanning 70 distinct families, to identify those that would reduce SSE-7 activity (Extended Data Fig. 7a). Given the diverse signalling states across GSCs and the results from the SCENIC analysis, we speculated that activity would be highly dependent on the cell line.","We therefore focused the screen on E55 and E31, two GSC lines with typical phenotypic heterogeneity and distinct\u00a0transcriptional states.\r\n\r\nInitial hits were validated, and a clear dose-dependent reduction in mCherry MFI was confirmed for four of these hits (Extended Data Fig.","7b,c). These four inhibitors all operate in the MAPK\u2013ERK signalling pathway: three MEK inhibitors (PD0325901, AS-703026 and PD184161) and one ERK inhibitor (SC-1) (Extended Data Fig. 7d). For these specific GSC lines (E55 and E31), SOX2 and SOX9 activity together with hyperactive MAPK signalling seemed to be necessary for SSE-7 activity.","Finally, a subpopulation of sorted SSE-7-negative GSCs could reactivate SSE-7 expression at later time points. This result is consistent with the finding that SSE activity depends on a specific signalling state (Extended Data Fig. 7e).\r\n\r\nAltogether, these functional genetic, biochemical and genomic data indicate that SSE-7 is a dynamic transcriptional regulatory element that is strongly activated in specific GSC states that have a combination of both SOX2 and SOX9 expression and hyperactive signalling-associated TFs. The specific signalling TFs involved are likely to vary among the heterogeneous GSC genetic and transcriptional subtypes.","However, our SSE design strategy (that is, mixing distinct enhancer fragments from diverse genes) enabled us to achieve robust and strong expression across GSCs\u00a0regardless of this variability.\r\nCell-state-specific activation of SSE-7\r\nOur functional enhancer fragments typically had high vertebrate evolutionary conservation, which suggested that SSEs may work in different species. To further explore selectivity across diverse tissues and organs, we first tested SSE activity in developing zebrafish (around 48\u2009h after fertilization) (Extended Data Fig. 8).","Overlapping expression domains for each SSE-driven eGFP reporter was observed for all four SSEs tested, with consistently high levels of expression in the optic placodes and a subset of forebrain and spinal cord neural progenitors (Extended Data Fig.","8a,b). This restricted developmental expression pattern was consistent with previously reported zebrafish Sox2 and Sox9 expression27,28, and confirms that not only are the SSEs highly active but they are also highly tissue restricted.","Notably, SSE-1, SSE-3 and SSE-5 had an additional weaker expression domain in more posterior CNS tissues, whereas for SSE-3, we detected expression in the posterior endoderm (Extended Data Fig. 8b). We therefore prioritized SSE-7 for subsequent experiments.\r\n\r\nWe created a stable transgenic zebrafish line and used this to look at later larval stages in the nervous system (Extended Data Fig. 8c). SSE-7 activity was nonoverlapping with mature neuronal markers, which confirmed that SSE-7 is not expressed in neurons and is only seen in a subset of fetal neural\u00a0progenitor cells (Extended Data Fig. 8d\u2013g).","Notably, when we forced the expression of a potent oncogenic form of Akt in zebrafish, we observed that SSE-7 activity increased in those cells. This result is consistent with the requirement of hyperactive signalling together with Sox expression to support SSE activation, as seen in human GSCs (Extended Data Fig. 8h). SSE-7 had low activity in oligodendrocyte progenitor cells (OPCs) derived from human induced pluripotent stem (iPS) cells relative to CMV (Extended Data Fig.","8i),\u00a0but was expressed in human fetal NSC cultures (Extended Data Fig. 8j).","Altogether, these data suggest that SSE-7 activity is highly restricted to specific\u00a0cell states and may be most active in a subset of fetal neural progenitors.\r\nSSE-7 activity in GBM core and margin\r\nAAV is an excellent gene therapy vector for use in GBM owing to the following characteristics: (1) its small physical size (for enhanced distribution); (2) persistence in quiescent cancer cells; (3) low immunogenicity; (4) diverse natural capsids that might enable improved transduction; (5) validated efficient brain delivery via convection-enhanced delivery; and (6) well-understood manufacturing and clinical safety profile.\r\n\r\nTo further explore SSE-7 selectivity using relevant adult human neural cell types and the effectiveness of AAV as the delivery vector, we generated human glutamatergic and GABAergic differentiated neurons (ioGlutamatergic and ioGABAergic, respectively), and microglia (ioMicroglia) from iPS cells, and tested AAV1-SSE-7\u2013mCherry activity (Fig.","4a). SSE-7 was strongly expressed in GSCs within a few days of transduction, with similar levels to the CMV positive control.","However, in postmitotic neurons, we did not observe SSE-7 activity (Fig. 4a) at either day\u20093 or day\u200910. For microglia, there was poor transduction with AAV1.\r\nFig. 4: GSC selectivity of SSE-7 confirmed in vitro and in tissue slice cultures.\r\na, AAV1-SSE-7\u2013mCherry reporter activity in GSCs and human iPS-cell-derived differentiated neurons and microglia assessed by flow cytometry at day\u20093 (left) and day\u200910 (right) following transduction. For day\u20093 for ioGABAergic neurons and ioMicroglia, n\u2009=\u20093 biological independent cell experiments; ioGlutamatergic neurons and GSCs, n\u2009=\u20094 biological independent cell experiments.","For day\u200910, n\u2009=\u20093 biological independent cell experiments excluding ioGABAergic neurons, for which n\u2009=\u20092. Data are the mean\u2009\u00b1\u2009s.e.m., and statistical analysis was performed using an unpaired t-test. **P\u2009\u2264\u20090.01, ***P\u2009\u2264\u20090.001.","b, Experimental schematic of AAV1-SSE-7 to test selectivity in fresh human GBM tumour mass and tumour margin tissue samples (infiltrating cortex) processed into slice cultures (Methods).","c, Immunostaining and confocal images of tissue for mCherry (red) and GSC markers (Nestin and SOX2) demonstrates GSC selectivity of SSE-7 relative to CMV. Tumour mass and adjacent matched normal brain tissue (tumour margin) are shown.","d, Percentage of mCherry-expressing cells co-expressing SOX2. Each point represents a biological replicate, matched tumour and tumour margin samples are indicated by matching symbols. n\u2009=\u20093 for margin and n\u2009=\u20094 for the tumour mass. Statistical analysis was performed using two-way ANOVA with multiple-comparisons correction (P\u2009=\u20090.0436). Scale bars, 10\u2009\u00b5m. Schematic in b created in BioRender.","Pollard,\u2009S.\u00a0M.\u00a0(2026) https:\/\/BioRender.com\/l774d5t.\r\n\r\nSource data\r\n\r\nTo directly test whether SSE-7 would have strong differential activity in GBM tumour tissue compared with the surrounding normal brain, we determined the transduction and expression of AAV1-SSE-7\u2013mCherry in ex vivo fresh tumour tissue slices obtained from patients with GBM (Fig. 4b) and compared these with the tumour margin (which is macroscopically normal tissue but contains infiltrating GBM cells). We first tested whether AAV1-SSE-7 would be active in the tissue, focusing on the mCherry reporter only.","After 7\u2009days of incubation with a single dose of the virus, we fixed and stained the tumour tissue and compared CMV and SSE-7 activity using the mCherry reporter.\r\n\r\nThere were significantly fewer mCherry-positive cells when using SSE-7 compared with CMV (Fig. 4c and Extended Data Fig. 9b) in the tumour, and this was even more significant in the tumour margin tissue. This result is consistent with SSE-7 labelling a subset of GBM cells but not normal healthy tissue. We performed this analysis for four independent patient samples\u2014grade\u20094 IDH mutant astrocytoma, IDH wild-type GBM, a grade\u20093 astrocytoma and recurrent GBM (Extended Data Fig. 9a)\u2014and obtained similar results.","By contrast, as expected, CMV induced widespread high expression across diverse normal CNS cell types, including cells with neuronal and endothelial morphology (Fig. 4c and Extended Data Fig. 9).\r\n\r\nTo next confirm whether mCherry is enriched in the tumour cell population, we performed and quantified SOX2 co-staining. This experiment revealed significant differences between CMV and SSE-7, with SSE-7 activated in around 90% of mCherry-positive tumour cells that co-expressed SOX2 (Fig. 4d). We conclude that SSE-7 transgene expression is differentially active in tumour versus normal tissue (tumour margin) compared with CMV.","Moreover, SSE-7 provides a strong and selective gene switch that can restrict transgene activity to GSCs in the tumour mass and the infiltrating tumour margin.\r\nComplete responses in a mouse model of GBM\r\nWe next determined whether AAV1-SSE-7 could be used to selectively kill GSCs using expression of the HSV-TK\/GCV enzyme prodrug system, which has previously been applied in the clinic as a gene-directed enzyme prodrug therapy29. SSE-7 retained activity and provided selective expression of the HSV-TK payload in GSCs (Extended Data Fig. 10a\u2013d). This enabled cytotoxic killing of GSCs, after GCV treatment, whereas fibroblasts were unaffected (Extended Data Fig. 10e\u2013g).","Screening of common natural AAV serotypes confirmed that AAV1 is the optimal serotype of choice, as it efficiently transduced all seven patient-derived GSCs with high efficiency (Extended Data Fig.","11a,b).\r\n\r\nDespite the promise of HSV-TK as a suicide gene therapy approach for oncology, there has been limited success in the clinic. This is partially explained by the challenges of efficient delivery to all cancer cells30. As immune payloads can act non-autonomously to trigger an immune response against residual disease, we reasoned that combining an immune payload with a cytotoxic payload might be required to achieve this goal.","Indeed, such a strategy could induce multiple pathways of the immune response to polarize local macrophages in the tumour microenvironment while enabling systemic adaptive responses by T\u2009cells. We proposed that this combination\u2014tumour cell killing, tumour-associated macrophage polarization and T\u2009cell education and activation\u2014would be potent and reduce chances of tumour regrowth from residual disease30.\r\n\r\nWe chose the cytokine IL-12 for evaluation as it triggers a potent inflammatory response (stimulating adaptive and innate immunity) and is effective in promoting anticancer immune responses31,32. We reasoned that by using the AAV1-SSE-7 platform, we could avoid the toxicities previously seen when using systemic delivery31.","This would enable controlled release of IL-12 alongside tumour killing and induce effective antitumour immune responses, with IL-12 levels being reduced as tumour cells are lost. We designed an AAV1 vector with SSE-7 driving the expression of a single mRNA encoding HSV-TK linked via the P2A sequence to the mouse IL-12 heterodimer fusion (Il12a and Il12b, which encode P40 and P35, respectively) (Fig. 5a). This dual payload design ensures that the cytokine and cytotoxic effects are synchronized and serves as a safety switch to prevent uncontrolled IL-12 levels.\r\nFig.","5: The combination of a cytotoxic and immunomodulatory payload is highly effective at clearing established GBM tumours.\r\na, Architecture of the AAV1 transgene used to treat NPE-IE GBM in mice.","ITR, inverted terminal repeat;\u00a0mCMV, minimal CMV promoter;\u00a0mIL-12, mouse IL-12.","b, Survival curves for the mice treated with AAV1-SSE-7\u2013HSV-TK\u2013IL-12 (n\u2009=\u200924) versus control (UT (PBS); n\u2009=\u200922). P\u2009<\u20090.0001 for pooled (n\u2009=\u20093) independent cohorts. Survival was analysed using the\u00a0Kaplan\u2013Meier method. Significance was determined using a\u00a0log-rank (Mantel\u2013Cox) test. Mice were selected at week\u20091 using bioluminescence imaging to ensure a similar starting size for tumours across the cohorts.","c, Longitudinal tracking of tumour growth in mice once a week using bioluminescence imaging (three examples shown for experimental and control; see Extended Data Fig. 12 for the larger cohort).","d,e, Bar plots of the abundance of total T\u2009(CD3)\u00a0cells, effector T (Teff) cells, CD8+ T\u2009cells, regulatory T (Treg) cells and natural killer (NK) cells (d), and myeloid-derived suppressor cells (MDSCs), monocytes, bone marrow-derived macrophages (BMDMs), dendritic cells (DCs) and microglia (e) out of total live CD45+ cells in the tumour microenvironment of untreated mice and mice treated with a single dose of AAV-SSE-7-HSV-TK\u2013IL-12 or with the AAV-SSE-7-HSV-TK\u2013IL-12\u2009+\u2009GCV combination, determined by flow cytometry.","f,g, Bar plots of the abundance of PD1\u2212Ki67\u2212, PD1\u2212Ki67+, PD1+Ki67\u2212 and PD1+Ki67+ cells from Teff cells (f) and from CD8+ T\u2009cells (g) in the tumour microenvironment of untreated mice and mice treated with a single dose of AAV-SSE-7-HSV-TK\u2013IL-12 or with the AAV-SSE-7-HSV-TK-IL-12\u2009+\u2009GCV combination, determined by flow cytometry. n\u2009=\u20097 biologically independent animals per group, examined in one experiment. Data are the mean\u2009\u00b1\u2009s.d. Significance was assessed using one-way ANOVA with Tukey\u2019s multiple comparison test.","*P\u2009\u2264\u20090.05, **P\u2009\u2264\u20090.01, ***P\u2009\u2264\u20090.001, ****P\u2009\u2264\u20090.0001.\r\n\r\nSource data\r\n\r\nFor in vivo preclinical efficacy studies, we used our previously reported disease-relevant syngeneic mouse model of GBM, termed NPE-IE25, which exhibits a myelosuppressive tumour microenvironment and transcriptional programs seen in human GBMs25. AAV1-SSE-7 transduction and SSE activity were first validated using mouse NPE-IE cells in vitro (Extended Data Fig.","11c,d).\r\n\r\nAAV1-SSE-7\u2013HSV-TK\u2013P2A-IL-12 was delivered to a cohort of mice with established orthotopic tumours (14\u2009days after tumour cell orthotopic engraftment) using direct intratumoral injection with a microinjection pump to deliver a single dose of the virus. We observed substantial tumour regression within 1\u20132\u2009weeks and complete clearance of tumours in 20\u201324 treated mice over the subsequent 2\u20133\u2009weeks, whereas control mice died within 1\u20132\u2009weeks (Fig.","5b,c).\r\n\r\nUsing bioluminescence imaging to track tumour burden in live mice, no further regrowth of tumours was observed after the initial tumours had cleared (Extended Data Fig. 12a).","Moreover, no prominent toxicity, scored by weight loss and neurological monitoring, was seen over the subsequent 11\u2009months (the experimental end point). We re-challenged treated mice (n\u2009=\u200910 that cleared the tumour) with fresh NPE-IE cell transplants into the striatum (5\u2009months after initiation of the first therapy).","Notably, there was no detectable tumour formation in any of the mice after the re-challenge, whereas parallel control mice formed tumours within 2\u20133\u2009weeks (Extended Data Fig. 12a).","Finally, antitumour activity was dose-dependent (Extended Data Fig. 12b).\r\nIL-12 with HSV-TK\/GCV drives antitumour immunity\r\nTo determine whether IL-12 alone would be sufficient for eradicating tumours without the concomitant HSV-TK\/GCV-induced killing (that is, single versus dual payloads, respectively), we treated mice with AAV1-SSE-7-HSV-TK\u2013P2A-IL-12 alone or with this virus and with GCV. The dual payload was the most effective and enhanced the effects of IL-12 alone (Extended Data Fig. 13a). To explore safety in normal CNS, we also assessed responses in mice without tumours to either CMV-driven or SSE-7-driven HSV-TK\u2013IL12.","As expected for CMV, there were rapid neurological issues within weeks, consistent with the expected toxicities with continuous expression of IL-12 in normal CNS. By contrast, we observed no toxicity with SSE-7, a result consistent with our earlier in vitro data confirming the selectivity of this SSE (Extended Data Fig. 13b).\r\n\r\nCharacterization of the immune cell composition after treatment using either IL-12 alone or IL-12 with HSV-TK\/GCV revealed that the latter performed better, with significantly increased levels of IFN\u03b3 and a trend towards increased TNF (Extended Data Fig. 13c\u2013i).","This finding is consistent with the observed increase in cytotoxic T\u2009cells and reduction in myelosuppressive cells that occurs through T helper\u20091-like responses induced by IL-12 (Fig. 5d\u2013g).\r\n\r\nThere was also an increase in proliferative Ki67+ T\u2009cells, with a shift towards increased CD44+, a marker of T\u2009cell activation and memory (Extended Data Fig.","13d,e). To assess the percentage of transduction in the tumour mass, we used the SSE-7\u2013mCherry reporter. Tumours were collected, dissociated and assessed using flow cytometry, which revealed around 20% positive tumour cells (GFP+mCherry+) (Extended Data Fig. 13j). This transduction efficiency was achieved using the same viral doses as the survival analysis experiments. This result confirms that only a fraction of the tumour mass needs to be transduced to achieve immune clearance of residual disease.\r\n\r\nIn conclusion, AAV1-SSE-7\u2013HSV-TK\u2013IL-12 is highly effective at eliminating tumours in an aggressive syngeneic\u00a0mouse GBM\u00a0model. Treated mice were resistant to subsequent re-challenge with fresh NPE-IE cells.","A single dose of AAV1-SSE-7\u2013HSV-TK\u2013IL-12 induced complete responses with durable immune memory without off-target toxicity.\r\nDiscussion\r\nThe development of effective gene therapies for oncology will require more effective payload combinations and more selective delivery mechanisms than has been possible so far30,33,34. We showed that SSEs, combined with delivery via AAV1, can address one of the significant challenges in oncology: namely, how to specifically deliver combinations of anticancer payloads at high doses without off-target toxicity.","Across genetically diverse GSC lines and high-grade glioma tissues, SSE-7 consistently read out a core transcriptional program characteristic of GSC identity: the immature NSC-like state.\r\n\r\nOur findings indicated that SOX2 and SOX9 share crucial target genes and may represent a core transcriptional module, an \u2018oncofetal\u2019 program, that defines aggressive tumour-initiating cells in GBM18,35. Using AAV-mediated delivery and SSE control of a transgene, we achieved specific elimination of active GSCs by targeting them from within, akin to a Trojan horse approach.","The persistence of AAV vectors, and our focus on the shared SOX circuits, meant that we were able to guard against plasticity, as transgene expression will be reactivated if cells transit through this specific cell state.\r\n\r\nImmunotherapy approaches have generally had limited success in GBM36,37,38. Our AAV\u2013SSE approach will now require careful safety studies to determine the potential for translation to patients.","The goal is to deliver cytotoxic payloads alongside innate and adaptive immune activation to achieve meaningful and durable suppression of high-grade gliomas without significant off-target toxicities.\r\n\r\nFuture studies will explore the molecular mechanisms that underlie SSE activity, which may involve transcriptional hubs or condensates. Our findings from exploring the biochemical binding of SOX2 and SOX9 suggests that there is clear synergy between these two factors.","However, additional cofactors are required for the creation of a multimeric complex. Detailed structural studies are needed to precisely determine how SOX2 and SOX9 operate and the signalling TF requirements and whether nucleosome context is critical39. In the future, improved computational methods and artificial-intelligence-based tools for the design of enhancers and synthetic associated motifs will complement the functional screening approaches used here40,41,42.","In conclusion, our study demonstrated the potential of SSEs and AAV vectors to deliver precision immune activation for the treatment of cancer.\r\nMethodsDestination vector cloning\r\nWe built a specific custom destination vector for efficient Golden Gate cloning of enhancer fragments. This included the reporter gene cassette NanoLuc-Ires-mNGreen-pA, downstream of a mCMV promoter. A bacterial suicide ccdB cassette spanning the enhancer position enabled assembly of a single enhancer for selection and efficient cloning. This was based on the EMMA Golden Gate cloning system43.","For the GSC long-term differentiation experiments, the same vector was built but included PiggyBac transposase recognition sites flanking the entire cassette.\r\nBioinformatics to design a SOX2 enhancer oligonucleotide pool\r\nWe re-analysed previously published1 SOX2 ChIP\u2013seq and H3K27ac GSC cell line data to identify GSC-specific SOX2 peaks that were overlapping with H3K27ac and absent in differentiated cells (cells in serum culture). The resulting shared peaks were then combined into one set and manually curated to remove centromeres. This resulted in 1,721 peaks with an average length of 402\u2009bp. These peaks were then used to design a set of 160\u2009bp sequences.","Next, 20\u2009bp adapters were included to flank the sequences and these sequences were synthesized as an oligonucleotide pool (Twist Bio).\r\nConstruction of an arrayed plasmid library of enhancer fragments\r\nThe oligonucleotide pool (9,523 unique sequences) was first amplified for 10 cycles with 0.25\u2009\u00b5l (2.5\u2009ng input) volume, and 0.5\u2009\u00b5l of this reaction was used for the subsequent 15 cycles of amplification to reduce PCR \u2018jackpot\u2019 amplification. The final products were cloned into the expression vector using an efficient Golden Gate reaction. We used KAPA HiFi Hotstart polymerase with GC buffer (Roche, KK2501), 68\u2009\u00b0C annealing temperature and 5\u2009s of extension time at 72\u2009\u00b0C.","A total of 4,579 individual plasmids were then randomly picked, miniprepped and plated as an arrayed plasmid DNA library on 96-well plates. This was deemed a practical number for arrayed library screening. A magnetic-bead-based SPRI purification method was used to minimize loss and to clean up the DNA before Golden Gate cloning. Sanger sequencing for quality checks of a sample of the plasmids confirmed that they were diverse, and all sequences could be mapped back to the original library design.\r\nScreening platform for 384-well plates\r\nCells were seeded in 384-well plates using a multidrop and transfected the following day using CyBio Felix (CyBio).","Two days later, a NanoGlo DLR assay was used to identify hits with a fold change to mCMV of >10. These hits were Sanger sequenced to determine the specific sequence and mapped back to the genome to validate overlap with the original SOX2 ChIP\u2013seq data. The GREAT tool predicted the target gene for hit sequences44.\r\nGeneral cell culture procedures\r\nThe GSC lines GCGR-E17 (E17), GCGR-E21 (E21), GCGR-E27 (E27), GCGR-E28 (E28), GCGR-E31 (E31), GCGR-E34 (E34), GCGR-E37 (E37) and GCGR-E55 (E55) and the human NSC lines NS9FB_B (NS9), NS12ST_A (NS12) and NS17ST_A (NS17) were generated in the Pollard Laboratory and are available upon request from the Glioma Cellular Genetics Resource. Informed consent was obtained for use of patient tissue.","All procedures on patient brain tissue received ethics approval from the NHS Health Research Authority, East of Scotland Research Ethics Service (REC reference 15\/ES\/0094), and all procedures on embryonic and fetal brain tissue received ethics approval from the Lothian NHS Board, South East Scotland Research Ethics Committee (REC reference 08\/S1101\/1).\r\n\r\nCells were grown as adherent monolayers under serum-free conditions as previously described13. All cells lines tested negative for mycoplasma at passage\u00a03 using a MycoAlert Mycoplasma Detection kit (Lonza, LT07-318). Cell line authentication by STR profiling was performed as a service by the European Collection of Authenticated Cell Cultures (ECACC).","Reports are available upon request from the Glioma Cellular Genetics Resource. Analysis was conducted using the Promega Fusion system (DC2402), analysing differences at 24 distinct hypervariable genetic loci, 16 of which are used for the final STR profile and report. The ioGlutamatergic (io1001S) and ioGABAergic (ioEA1003S) neurons and ioMicroglia (ioA021) were obtained from bit.bio and grown per the manufacturer\u2019s specifications. GSC7 was previously characterized22.\r\n\r\nFor lipofection, Plus reagent and Lipofectamine LTX (Life Technologies, 15338030) were each diluted in half the volume of Opti-MEM\u00a0I reduced-serum medium (hereafter referred to as Optimem) (Life Technologies, 31985062).","Following this step, the Plus reagent and Optimem premix was added to all DNA samples followed by the Lipofectamine LTX and Optimem premix. The transfection mix was incubated for 5\u2009min at room temperature and then carefully dropped onto the cells.","Generally, no change in medium was carried out. Cells were analysed 2\u00a0days after transfection.\r\n\r\nHEK293 cells were seeded at the specified density in the respective plate format. The next day GMEM medium, DNA and polyethylenimine were mixed and incubated for 15\u2009min at room temperature to facilitate formation of the complex. The transfection mix was subsequently dropped onto the cells, and analysis was carried out 2\u2009days later.\r\nScreening assay using the Nano-Glo Dual-Luciferase reporter assay system\r\nThis assay consists of two steps. Transfection was performed using a normalization plasmid PGK-Firefly-Luciferase (Promega, E5011) transfected in a 1:10 ratio with the plasmid of interest.","Therefore, in this assay, the firefly luciferase activity is first measured, which enables normalizing of the transfection efficiency. Cells were washed 3\u2009times with PBS and 20\u2009\u00b5l was left in a 96-well plate (25\u2009\u00b5l in a 384-well plate). Oneglo buffer was added, and the plate was shaken for 5\u2009min at 480\u2009rpm to allow cell lysis.","Next, 20\u2009\u00b5l cell lysate in a 96-well format (25\u2009\u00b5l cell lysate in a 384-well format) was transferred into an opaque white 96-well format (or 384-well format) and light was measured for 0.1\u2009s per well (Ensight Multimode Plate Reader, Perkin Elmer). For the subsequent NanoLuc reaction in a 96-well format, 2\u2009\u00b5l cell lysate was transferred into an opaque white plate containing 40\u2009\u00b5l PBS, and 20\u2009\u00b5l Stopglo buffer, supplemented with substrate, was added to each well. In a 384-well format, 20\u2009\u00b5l Stopglo buffer containing substrate was added on top of the undiluted cell lysate. The plate was shaken again for 5\u2009min at 480\u2009rpm to quench the firefly luciferase reaction and to ensure good mixing.","NanoLuc activity was measured for 0.1\u2009s per well. Data obtained by the NanoLuc reaction were normalized to the Firefly reaction to account for well-to-well variability of transfection efficiency. Normalized data were used to calculate the fold change compared with the empty vector control, which contained mCMV only and no enhancer fragments.\r\nImmunocytochemistry\r\nCells were washed twice with PBS and fixed in 4% paraformaldehyde (PFA) for 10\u2009min. Cells were permeabilized with 0.1% Triton-100 in PBS (PBST). Cells were blocked with blocking solution (1% BSA in PBST with 3% goat serum) and incubated with the primary antibody at 4\u2009\u00b0C overnight.","The next day, cells were washed 3\u2009times in PBST, and the respective secondary antibody was applied in blocking solution and incubated for 45\u201360\u2009min at room temperature. Cells were washed with PBS and incubated for 5\u2009min with a DAPI nuclear counterstain at 1\u2009\u00b5g\u2009ml\u20131 final concentration. Images were acquired using a Nikon TiE microscope and NIS elements software (Nikon).\r\nFlow cytometry\r\nFor flow cytometry, cells were detached, pelleted and resuspended in an appropriate volume of flow cytometry buffer (1% BSA in PBS, v\/v) or PBS.","Cells were stained with Draq7 (Abcam, ab109202, final concentration of 0.1\u2009\u00b5M) or DAPI (ThermoFisher Scientific, D3571) as a live\/dead stain and analysed using a BD LSRFortessa cell analyser (4 lasers, BD Bioscience). Analysis of flow cytometry data was carried out using FlowJo Analysis software (v.10.6.2) or FCS Express 7 flow cytometry software (v.7.22.x).\r\nRNA isolation, cDNA synthesis and RT\u2013qPCR\r\nRNA was extracted using a MasterPure RNA Purification kit (Epicentre). RNA was stored at \u221280\u2009\u00b0C, and concentration was determined using a NanoDrop spectrophotometer. cDNA synthesis was carried out using a SuperScript III Reverse Transcriptase kit (Life Technologies) according to the manufacturer\u2019s instructions.","Around 200\u2013500\u2009ng RNA was used for the cDNA reaction. The same amount of RNA was used for each experiment. After the reaction, cDNA was diluted to the required volume using nuclease-free water.\r\n\r\nFor RT\u2013qPCR, TaqMan Universal PCR master mix (Applied Biosystems) and TaqMan gene expression assays (Life Technologies) were used on a Quant Studio7 Flex Real-Time qPCR machine. RNA samples that did not undergo reverse transcription (to assess DNA contamination) and water controls were used on every plate. RT\u2013qPCR was carried out in technical duplicates. Data analysis was performed using the ddCt method, which assumes 100% PCR efficiency and is guaranteed with TaqMan assays.","In brief, the mean was calculated for technical replicates and normalized to the housekeeping gene GAPDH, which produces the dCt value.\r\nWestern immunoblotting\r\nCell lysates were prepared by resuspending cells in RIPA buffer (50\u2009mM HEPES pH\u20097.7, 150\u2009mM NaCl, 1% NP-40, 0.5% DOC and 0.1% SDS), incubating on ice for 5\u2009min, centrifuging at maximum speed for 10\u2009min in a tabletop centrifuge (5415D, Eppendorf) and collecting the supernatant (lysate). Protein extracts were quantified using a Pierce BCA Protein Assay kit (Thermo Scientific, 23225). SDS\u2013PAGE was performed using homemade 4\u201312% Bis-Tris gels.","Gels were transferred onto PVDF membranes (Millipore, IPVH00010), previously activated in methanol, by wet electroblotting or semi-dry blotting using a Bio-Rad Trans-blot turbo system. Western blots were revealed using HRP-conjugated antibodies, homemade ECL solutions or Clarity Western ECL (Bio-Rad), and imaged using X-ray films or a Bio-Rad ChemiDoc Imaging system.\r\nPrecipitation of interacting proteins\r\nPCR amplification with biotinylated primers and purification were used to generate templates that could pull down bound proteins.","Enhancer fragments were amplified with 5\u2032 biotinylated primers using PrimeStar Max (Takara) according to the manufacturer\u2019s instructions.The following primers were used: forward primer 595: TGATCCGTCTCGCCCTACTAGGTTACTGGTGCATGC; reverse primer 596: ACTAACGTCTCGGAGCACCCAAACTATTGGAGCGAG\u00a0(bold sequences, adaptors).\r\n\r\nPCR purification using Agencourt AMPure XP magnetic beads (Beckman Coulter) was carried out according to the manufacturer\u2019s instructions. PCR product quantification using Tapestation and reagents was carried out according to the manufacturer\u2019s instructions.\r\n\r\nAll the buffers were prepared the day before, passed through a 22\u2009\u00b5m filter (except the dialysis buffer) and left at 4\u2009\u00b0C overnight.","DTT and protease inhibitors were added immediately before use. Cell pellets were resuspended in 5\u2009ml per 40\u00a0million cells of ice-cold buffer\u2009A (10\u2009mM HEPES pH\u00a07.9, 1.5\u2009mM MgCl2, 10\u2009mM KCl, 0.5\u2009mM DTT and protease inhibitors (complete, Roche, 11697498001)) and incubated on ice for 10\u2009min. The cell suspension was transferred to a glass Dounce homogenizer and dounced 40 times on ice, transferred to a Falcon tube and centrifuged at 1,350\u2009rcf for 10\u2009min at 4\u2009\u00b0C.","The supernatant was discarded (cytosolic extract) and the pellet was resuspended in 100\u2009\u00b5l per 10\u2009million cells in ice-cold buffer\u2009B (20\u2009mM HEPES pH\u20097.9, 5% glycerol, 1\u2009M NaCl, 1.5\u2009mM MgCl2, 0.2\u2009mM EDTA pH\u20098.0, 0.5\u2009mM DTT and protease inhibitors (complete, Roche, 11697498001)). The suspension was rotated for 30\u2009min at 4\u2009\u00b0C and transferred into a dialysis membrane (SnakeSkin, Thermo Scientific, 68100). Dialysis was carried out in 500\u2009ml dialysis buffer (20\u2009mM HEPES pH\u20097.9, 5% glycerol, 100\u2009mM KCl, 0.83\u2009mM EDTA pH\u20098.0, 1.66\u2009mM DTT and protease inhibitors (complete, Roche, 11697498001)) for 2\u2009h at 4\u2009\u00b0C with rotation.","The dialysis buffer was changed (500\u2009ml), and dialysis was continued overnight at 4\u2009\u00b0C. The extract was collected in a 1.5\u2009ml tube and spun at maximum speed for 15\u2009min at 4\u2009\u00b0C (centrifuge 5415D Eppendorf). The supernatant was collected into new 1.5\u2009ml tubes and quantified (Pierce BCA Protein Assay kit, Thermo Scientific, 23225).","Next, 100\u2009\u00b5g aliquots were flash-frozen in liquid nitrogen and stored at \u221280\u2009\u00b0C. The following primary antibodies were used for western blotting: anti-SOX2 (rabbit) 1:1,000 Abcam (ab92494); anti-SOX9 (rabbit) 1:500 Millipore (ab5535); and anti-GAPDH (mouse) 1:10,000 Ambion (AM4300). The secondary antibodies used for western blotting were anti-rabbit 1:5,000 Novex (A16110) and anti-mouse 1:5,000 Novex (A16027).\r\n\r\nStreptavidin magnetic beads (NEB, S1420S) were used according to the manufacturer\u2019s instructions.","In brief, 10\u2009\u00b5l beads were aliquoted into low-bind tubes and washed 3\u2009times with binding buffer (20\u2009mM Tris-HCl pH\u20097.5, 0.5\u2009M NaCl and 1\u2009mM EDTA) on a magnetic stand.","Next, 20\u2009\u00b5l biotinylated DNA (around 20\u2009pmol) was mixed with 200\u2009\u00b5l binding buffer and was added to the beads. The suspension was rotated at room temperature for 2\u2009h. The beads were washed 3\u2009times with binding buffer, and 50\u2009\u00b5g nuclear extract was added (total volume of 50\u2009\u00b5l).","Beads, DNA and nuclear extract, to which 150\u2009\u00b5l dialysis buffer (20\u2009mM HEPES pH\u20097.9, 5% glycerol, 100\u2009mM KCl, 0.83\u2009mM EDTA pH\u20098.0, 1.66\u2009mM DTT and protease inhibitors (complete, Roche, 11697498001)) was added, were rotated overnight at 4\u2009\u00b0C. The following morning, beads were washed 3\u2009times with wash buffer (20\u2009mM HEPES pH\u20097.9, 5% glycerol, 250\u2009mM NaCl, 0.83\u2009mM EDTA pH\u20098.0 and 1.66\u2009mM DTT). For western blotting, proteins were eluted in 20\u2009\u00b5l loading buffer (lithium dodecyl sulfate buffer containing 50\u2009mM DTT) and boiled for 5\u2009min to denature the proteins.","For mass spectrometry, the beads were sent dry to the Institute for Genetics and Cancer core facility (University of Edinburgh) for analyses.\r\nSOX2 and SOX9 ChIP\u2013seq library preparation and analysis\r\nIn brief, GSCs cultured in 150\u2009mm dishes to 70% confluence were crosslinked in 1% formaldehyde for 10\u2009min at room temperature. Excess formaldehyde was quenched by incubating cells with 0.25\u2009M glycine for 5\u2009min at room temperature. The cell pellet was washed twice with ice-cold PBS before storage at \u221280\u2009\u00b0C. For each GSC line, cell pellets from eight 150\u2009mm dishes were combined before proceeding for cell lysis and sonication.","The cell pellet was lysed in buffer LB3 (10\u2009mM Tris-HCl pH\u20098.0, 100\u2009mM NaCl, 1\u2009mM EDTA pH\u20098.0, 0.5\u2009mM EGTA pH\u20098.0, 0.1% sodium deoxycholate, 0.5% N-lauroylsarcosine and 0.1% SDS) and chromatin was fragmented to 200\u2013600\u2009bp size on a Covaris M220 sonicator (total of 4 cycles, with the following treatment settings: 600\u2009s per cycle, peak power 75, duty factor 10, cycles\/burst 200; temperature: minimum 5\u00a0\u00b0C, set point 7\u00a0\u00b0C, maximum 9\u00a0\u00b0C). For each ChIP assay, 15\u201320\u2009\u00b5g fragmented chromatin was incubated overnight in a cold room with 10\u2009\u03bcg antibodies (R&D SOX2 AF2018, Millipore, SOX9 AB5535) and 30\u2009\u03bcl Protein G Dynabeads (10003D).","The next day, magnetic beads were washed 5 times with RIPA (50\u2009mM HEPES-KOH pH\u20097.5, 0.5\u2009M LiCl, 1\u2009mM EDTA pH\u20098.0, 1% NP-40 and 0.7% sodium deoxycholate), once with TE NaCl (10\u2009mM Tris-HCl pH\u20098.0, 1\u2009mM EDTA pH\u20098.0 and 50\u2009mM NaCl) and finally eluted in 200\u2009\u03bcl ChIP elution buffer (50\u2009mM Tris-HCl pH\u20098.0, 10\u2009mM EDTA pH\u20098.0 and 1% SDS). Immunoprecipitated chromatin was reverse crosslinked by incubating samples at 65\u2009\u00b0C for 10\u2009h and extracted using the phenol\u2013chloroform\u2013IAA (11896714) method. To account for intertumoral heterogeneity, ChIP\u2013seq was performed on seven independent primary GSC lines.","Libraries were prepared for two technical ChIP replicates and one input control using a NEBNext Ultra II DNA Library Prep kit for Illumina (E7645S). The libraries were barcoded using NEBNext Multiplex Oligos for Illumina (Dual Index Primers Set 1 kit, E7600S) and sequenced on an Illumina Novaseq (50\u2009bp read length).\r\n\r\nPaired-end reads were aligned to the hg38 genome using BWA, filtering out poor-quality (MAPQ\u2009<\u200910), duplicates, mitochondrial genome and blacklisted regions (https:\/\/www.encodeproject.org\/annotations\/ENCSR636HFF\/). Technical replicates were merged, and peaks were called using MACS2 with default settings.","Consensus SOX2 and SOX9 peak sets were derived by taking the overlap of peaks occurring in 5 out of 7 GSCs for each TF set. The overlap significance between the consensus SOX2 and SOX9 peak sets was determined using a circular permutation test in regioneR (ntimes\u2009=\u200910,000). Genomic regions near SOX2 and SOX9 peaks were annotated using HOMER45.\r\n\r\nTo perform overlapping analysis with GSC super-enhancers, we downloaded raw fastq files from publicly available H3K27ac datasets on GSCs (GSE119834, GSE74529, GSE121601 and GSE92458) and called super-enhancers using ROSE with default settings. A consensus set of GSC super-enhancers was derived by considering super-enhancers that occurred in at least two GSCs.","The overlap significance between the consensus co-bound SOX2\u2013SOX9 sites and consensus GSC super-enhancers was performed using a circular permutation test as described above. Gene ontology term association analysis was performed using GREAT. To find centrally enriched de novo motifs at SOX2 and SOX9 peaks in co-bound SOX2\u2013SOX9 enhancers, and to identify the spacing between the most significant de novo motifs from each peak set, we used CentriMo and SpaMo from the MEME-ChIP suite of tools46.\r\nscRNA-seq sample preparation\r\nOn the day of cell seeding, cells were detached as described above. Cells were counted using a haemocytometer and seeded into a 6-well Corning plate at 400,000 cells per well in 2\u2009ml.","Plates were incubated at 37\u2009\u00b0C with 5% CO2. The next day, AAV vector stocks were thawed at room temperature. The appropriate amount of virus stock was added to the culture medium to achieve a final multiplicity of infection (MOI) of 1,000,000, which ensured high transduction efficiency. Viral-containing medium was added to cells without replacing the existing medium. The cells were then returned to the incubator at 37\u2009\u00b0C with 5% CO2.\r\n\r\nAfter 3\u2009days of incubation with the viral particles, cells were detached as described above.","Next, 10% of the cells were prepared for flow cytometry using DAPI as a live\/dead stain (see\u00a0Supplementary Information for the gating strategy). The remaining 90% of the cells were processed for single-cell transcriptome sequencing.\r\n\r\nCells were fixed using a Parse Evercode v.2 Cell Fixation kit following the manufacturer\u2019s instructions, with an average of around 400,000 cells per sample. Fresh reagents were prepared for each fixation, and 7.5% Gibco BSA fraction\u2009V was added per the manufacturer\u2019s recommendations.","In brief, cells were washed with PBS and resuspended in cold Cell Prefixation buffer. Cells were filtered through a 40\u2009\u00b5m strainer after 5\u2009min of centrifugation at 300g at 4\u2009\u00b0C. Following this step, Cell Fixation solution and Cell Permeabilization solution were subsequently added, with incubation on ice for 10\u2009min and 3\u2009min, respectively. Cell Neutralization buffer was then added, and cells were centrifuged again at 300g for 5\u2009min at 4\u2009\u00b0C.","Finally, cells were resuspended in a volume of 50\u201390\u2009\u00b5l Cell Buffer containing 1% DMSO based on the live cell number before fixation. Cells were filtered through a 40\u2009\u00b5m strainer and stored at \u221280\u2009\u00b0C for up to 3\u2009months.\r\nscRNA-seq library preparation\r\nLibraries were generated using a Parse Evercode WT Mega v.2 kit according to the manufacturer\u2019s instructions.","In brief, 72 samples were loaded into a 96-well RT barcoding plate for reverse transcription, with each well containing a unique barcoded RT primer. Following cDNA synthesis, cells were pooled and subjected to two additional rounds of barcoding, achieving three rounds of barcoding. Barcoded cDNA was then amplified via template switching and pooled into 15 sublibraries, each containing around 62,500 cells. A fourth barcode was incorporated during PCR amplification of the sequencing libraries. Sublibraries were then fragmented, ligated with Illumina adapters and purified using Ampure XP beads. Library quality was assessed on an Agilent BioAnalyzer 2100 before sequencing.","Sublibraries, with 5% PhiX spike-in, were sequenced on an Illumina NovaSeq X using 300 cycles kits as paired-end, dual-index reads.\r\nscRNA-seq analysis\r\nSequencing reads were aligned to the human (hg38) genome, and UMI counting data were generated following the standard Parse Bioscience pipeline, \u2018splitpipe (v.1.1.2)\u2019, with default parameters. Three custom genes were added to the genome (mCherry, HSV-TK and bGHployA) to capture \u2018activated\u2019 SSE-7 cells.","In total, 345,495 cells were included in the initial dataset. Low-quality cells were filtered out based on the following thresholds: (1) the percentage of mitochondrial gene was >20%; (2) the number of genes was <300; (3) the number of uniquely aligned reads was <500; or (4) the number of uniquely aligned reads was >20,000. Genes detected in fewer than five cells were filtered out as low-quality genes. Potential doublets were identified using the R package \u2018DoubletFinder\u2019 (v.2.0.4), with an expected doublets rate of 3%, as guided by Parse. Doublets and clusters containing more than 20% doublets were removed. Ambient RNA-contaminated cells (cell score\u2009>\u20090.2) were removed using \u2018decontX\u2019 from the R package \u2018celda\u2019 (v.1.18.2).","After filtration, 270,842 cells remained for further analysis.\r\n\r\nThe data were normalized using \u2018LogNormalize\u2019, and principal component analysis was performed based on the top 2,000 variant genes using the R package \u2018Seurat\u2019 (v.5.0.3). The first 15 principal components were used as input for Louvain-based graphing. SSE-7-activated cells (2,711 cells) were labelled according to the following criteria: mCherry\u00a0>\u00a01 or HSV-TK\u2009>\u20091 or bGHpolyA\u2009>\u20090. Cells with a single count of mCherry or HSV-TK from the virus-treated group were labelled as false negative (9,583 cells), whereas the remaining cells were labelled as SSE-7 non-activated cells (125,525 cells).","Meanwhile, cells expressing mCherry or HSV-TK in the control group were labelled as false positive (1,998 cells). Transcriptional subtypes were predicted on the basis of a previously described gene signature27 using the function \u2018sigScores\u2019 and \u2018as_four_state_gbm\u2019 from the R package \u2018scalop\u2019 (v.1.1.0). Cell cycle was predicted using \u2018CellCycleScoring\u2019 from the R package \u2018Seurat\u2019.\r\nSCENIC analysis of scRNA-seq data\r\nFalse-negative and false-positive cells were removed, and the dataset was divided into 15 libraries as per the sequencing library preparation. The dataset was then analysed following the standard \u2018pyscenic (v.0.12.1)\u2019 pipeline.","Auxiliary human datasets (hg38 mc_v10_clust) were downloaded from the cisTarget resources website (https:\/\/resources.aertslab.org\/cistarget\/). Motif\u2013TF annotation was based on 10\u00a0kbp upstream and downstream around the transcription start sit (20\u00a0kbp in total). SSE-7-enriched TF modules were selected on the basis of the following criteria: (1) Z\u2009score\u2009>\u20090.7 and (2) Z\u2009score in the SSE-7-activated group was higher than in other groups, as determined by repeated-measures ANOVA test.","Enrichment for candidate TF modules across 15 libraries was as follows: SOX8 (found in 9 libraries), STAT1 (found in 15 libraries), IRF9 (found in 12 libraries), ETS1 (found in 13 libraries), MAF (found in 3 libraries), SMAD1 (found in 12 libraries), FOS (found in 11 libraries), HIF1A (found in 7 libraries), SOX9 (found in 10 libraries), STAT3 (found in 3 libraries), NEF2L2 (found in 13 libraries), JUN (found in 2 libraries) and SOX2 (found in 2 libraries).\r\nLentivirus production and titration\r\nTo produce SOX2, SOX9 and rtTA2M2 individual lentiviral supernatant samples, nearly 2.4\u2009million HEK293T cells per 15\u2009cm plate were cultured for 24\u2009h in GMEM medium supplemented with 10% fetal calf serum (FCS), 1\u2009mM sodium pyruvate, 1\u2009mM glutamine and non-essential amino acids (HEK medium) at 37\u2009\u00b0C and 5% CO2.","At 24\u2009h after seeding, HEK293T cells were transfected with a plasmid cocktail containing 7.5\u2009\u03bcg expression plasmid, 5.1\u2009\u03bcg psPAX2 packaging vector, 2.4\u2009\u03bcg envelope vector mixed in 45\u2009\u03bcl Fugene6 (Roche) and 855\u2009\u03bcl Opti-MEM medium (Invitrogen, Thermo Fisher Scientific). The cells were further cultured for 16\u2009h at 37\u2009\u00b0C and 5% CO2 before the medium was changed for HEK medium. At 65\u2009h after transfection, virus-containing supernatant was collected and cleared by centrifugation and filtered through a 0.45\u2009\u03bcm syringe filter (Merck Millipore).","Virus was concentrated (roughly 100-fold) by transferring to thin-walled ultracentrifuge tubes and pelleted by ultracentrifugation at 25,000\u2009rpm for 2.5\u2009h using a SW32-TI rotor in a Beckman Optima XPN ultracentrifuge (Beckman Coulter). The pelleted virus was resuspended in 300\u2009\u03bcl plain GMEM medium and incubated at 4\u2009\u00b0C for 16\u2009h before aliquoting and flash-freezing in liquid nitrogen for storage at \u221280\u2009\u00b0C.","The viral titre, calculated by flow cytometry, was 6\u2009\u00d7\u2009107 infectious units per ml.\r\nOverexpression of SOX2 and SOX9\r\nEarly passage (passage\u20094) human fibroblasts (hFibs) were seeded, 1.5\u2009million per 15\u2009cm plate, and cultured in hFib medium (GMEM medium supplemented with 10% FCS, 1\u2009mM sodium pyruvate, 1\u2009mM glutamine, MEM 1\u00d7 non-essential amino acids solution, 50\u2009\u03bcM 2-mercaptoethanol (Gibco, Thermo Fisher Scientific) and 2\u2009ml penicillin\u2013streptomycin (Invitrogen, 15140122)) 24\u2009h before transduction at 37\u2009\u00b0C and 5% CO2.","The cells were transduced by exchanging the growth culture medium with lentiviral medium, without antibiotics, containing 8\u2009\u03bcg\u2009ml\u20131 polybrene, and 65\u2009\u03bcl concentrated rtTA2M2 and either SOX2 or SOX9 virus, for single factor overexpression, or 75\u2009\u03bcl rtTA2M2, SOX2 and SOX9 viruses for double factor overexpression. At 48\u2009h after transduction, the cells were split 1:2 and cultured for 24\u2009h, after which the culture medium was replaced with induction medium containing 1\u2009\u03bcg\u2009ml\u20131 doxycycline (induction day\u20090). After 48\u2009h of induction, the cells were collected for lysis and nuclear protein extraction.\r\nNuclear extraction of GSC7 and hFib cells\r\nGSC7 cells were cultured as adherent monolayers13.","For differentiation, cells were cultured without the supplements EGF and FGF and in 5% serum for 2\u2009weeks. GSC7 and hFib nuclear lysates were prepared in the same way; cells were resuspended to 1\u00a0million cells per 100\u2009\u03bcl 10\u2009mM HEPES pH\u20097\u20138, 1.5\u2009mM MgCl2, 10\u2009mM KCl, 0.5\u2009mM DTT, supplemented with complete Ultra EDTA-free protease inhibitors (Roche, 05892970001) and homogenized using a dounce and tight-fit pestle with 1 stroke per 1\u2009million cells. Samples were centrifuged at 1,350\u2009rcf for 5\u2009min at 4\u2009\u00b0C.","The nucleus pellet was resuspended to 10\u2009\u03bcl per 1\u2009million starting cells in 20\u2009mM HEPES pH\u20097\u20138, 30% glycerol, 420\u2009mM NaCl, 1.5\u2009mM MgCl2, 0.2\u2009mM EDTA, 0.5\u2009mM DTT and complete Ultra EDTA-free protease inhibitors (Roche, 05892970001). Samples were dialysed for 4\u2009h into a minimum of 1\u20131,000 volume excess of 20\u2009mM HEPES pH\u20097\u20138, 30% glycerol, 100\u2009mM NaCl, 0.83\u2009mM EDTA, 1.66\u2009mM DTT, 0.2\u2009mM PMSF and complete Ultra EDTA-free protease inhibitors (Roche 05892953001) using Slide-A-lyzer mini dialysis units with 3,500 MWCO cassettes (Thermo).","Samples were divided into single-use aliquots (11\u2009\u03bcl) and flash-frozen in liquid nitrogen for storage at \u221280\u2009\u00b0C.\r\nWestern blot analysis of overexpression\r\nThe protein concentrations of the lysates were quantified using a Pierce BCA Protein Assay kit according to the manufacturer\u2019s 96-well format instructions (Thermo Fisher Scientific). The following overexpression (OE) values were used: SOX2 OE\u2009=\u200910.54\u2009mg\u2009ml\u20131, SOX9 OE\u2009=\u20095.56\u2009mg\u2009ml\u20131, SOX2 and SOX9 OE\u2009=\u20097.56\u2009mg\u2009ml\u20131, GSC7 GSCs\u2009=\u20091.02\u2009mg\u2009ml\u20131 and GSC7 differentiated cells\u2009=\u20090.43\u2009mg\u2009ml\u20131.","Verification of SOX protein overexpression, relative to untransfected hFib lysate, was resolved by SDS\u2013PAGE and electroblotting onto a PVDF membrane for western blotting. The primary antibody incubations with goat anti-human SOX2 antibody (1:363, AF2018, Abcam) and rabbit anti-mouse SOX9 (1:2,000 AB5535, Chemicon) were performed at 4\u2009\u00b0C for 16\u2009h. The secondary antibody incubations with donkey anti-goat IgG-HRP (1:2,000 dilution of ab97110, Abcam) and goat anti-rabbit (1:2,000 of 32460, Invitrogen) were performed for 1\u2009h at room temperature.","Blots were visualized using SuperSignal West Pico Chemiluminescent substrate (Thermo Fisher Scientific) and HRP was visualized using a Bio-Rad ChemiDoc MP imaging system.\r\nRecombinant SOX protein production\r\nHis-tagged human SOX2 and SOX9 were expressed from a pET28a backbone (cloned by G.\u00a0Roberts in the Soufi Laboratory) using Rosetta 2(DE3) pLys competent cells (Novagen).\r\n\r\nSOX2 was expressed as previously described47. Cells in LB medium (1% Bacto-tryptone, 0.5% yeast extract and 1% NaCl), supplemented with 30\u2009\u00b5g\u2009ml\u20131 kanamycin and chloramphenicol, were grown at 37\u2009\u00b0C overnight.","SOX9 was expressed in 2\u00d7\u00a0TY medium (1.6% Bacto-tryptone, 1% yeast extract and 0.5% NaCl), supplemented with 30\u2009\u00b5g\u2009ml\u20131 kanamycin and chloramphenicol, at 16\u2009\u00b0C overnight. After expression, bacteria cell pellets were lysed, 100\u2009ml buffer per 1\u2009l of culture, in GuHCl Denaturing buffer (6\u2009M guanidine-HCl, 50\u2009mM Tris-HCl (pH\u20098) and 500\u2009mM NaCl, 5% (v\/v) glycerol) overnight. Lysate was sonicated for 200\u2013300\u2009s at 25\u201330\u2009microns using a Soniprep 150 with intermittent chilling on ice. The supernatant was clarified by centrifugation at 18,000\u2009rpm in a FLA21-8x50y rotor for 30\u2009min.","The lysate was further sonicated for 120\u2009s, as above, then filtered through a 0.45\u2009\u00b5m filter before loading into a HisTrapHP column (Cytivia) for affinity purification of His-tagged proteins using an AKTA pure system. The column was equilibrated with 5\u00a0CV of denaturing 10\u00a0mM imidazole buffer (6\u2009M urea, 500\u2009mM NaCl, 50\u2009mM Tris-HCl (pH\u20098), 10\u2009mM imidazole and 5% (v\/v) glycerol), after sample loading was washed with 20\u2009CV of denaturing 50\u00a0mM imidazole buffer (6\u2009M urea, 500\u2009mM NaCl, 50\u2009mM Tris-HCl (pH\u00a08), 30\u2009mM imidazole and 10% (v\/v) glycerol), and bound proteins were eluted with denaturing 300\u2009mM imidazole (6\u2009M urea, 500\u2009mM NaCl, 50\u2009mM Tris-HCl (pH\u20098), 300\u2009mM imidazole and 10% glycerol).","Peak fractions of the elution were analysed by SDS\u2013PAGE, and samples with appropriately sized bands (40\u2009kDa for His\u2013SOX2, and 60\u2009kDa for His\u2013SOX9) were combined and desalted into 2\u2009M urea buffer using a HiPrep 26\/10 desalting column (GE Healthcare). The SOX2 desalting buffer comprised 50\u2009mM Tris-HCl (pH\u00a08), 240\u2009mM NaCl, 10\u2009mM KCl, 2\u2009mM MgCl2, 2\u2009mM CaCl2, 0.8\u2009M urea, 30% (v\/v) glycerol, 0.1% NP40 substitute, 0.05% Triton-X-100, 2\u2009mM EDTA and 5\u2009mM DTT resuspended in 1\u00d7 PBS. The SOX9 desalting buffer comprised 50\u2009mM HEPES NaOH (pH\u20097.5), 240\u2009mM NaCl, 10\u2009mM KCl, 2\u2009mM MgCl2, 2\u2009mM CaCl2, 2\u2009M urea, 30% (v\/v) glycerol, 0.1% NP40 substitute, 0.05% Triton-X-100, 2\u2009mM EDTA and 5\u2009mM DTT.","The fractions corresponding to a dip in the absorbance at A280 were collected and visualized by SDS\u2013PAGE before combining. Protein concentration was determined by SDS\u2013PAGE, comparing to rAlbumin standards. The gels were imaged by detecting Coomassie blue staining using a Bio-Rad ChemiDoc MP imaging system. The resulting images were visualized, and the bands quantified using ImageJ 2.8\/1.54i.\r\nSmall-molecule inhibitor screening\r\nE55 cells were screened using inhibitors included in the Kinase Screening Library (Cayman, 10505) as well as two YAP\u2013TEAD pathway inhibitors, MGH-CP1 (HY-139330) and TED-347 (HY-125269).","Around 18,000 cells were seeded per well in a 96-well plate and transduced with the corresponding AAVs at a MOI of 2,000,000 on the next day. Cells were incubated at 37\u2009\u00b0C, 5% CO2. In the primary screen, the cells were treated with a single concentration of 3\u2009\u03bcM per inhibitor at 24\u2009h after transduction. After 48\u2009h of inhibitor treatment at 37\u2009\u00b0C and 5% CO2, the cells were detached and analysed using an Agilent NovoCyte Penteon five laser flow cytometer and FCS Express 7 Research (v.7.24.0030). The results were expressed as the fold change in mCherry median fluorescence intensity in the inhibitor-treated samples relative to the DMSO control.","After performing the first round of screening, inhibitors that gave more than 20% changes were selected for the second round of screening.\r\n\r\nIn the second-round screen, the cells were seeded and transduced with AAVs using the same procedures as the primary screen. At 24\u2009h after transduction, the cells were treated with 3 different doses (0.1, 1 and 3\u2009\u03bcM), and incubated at 37\u2009\u00b0C and 5% CO2 for 48\u2009h. Cells were then detached and analysed using a Agilent NovoCyte Penteon five laser flow cytometer. Draq7 was used as the live\/dead stain in both rounds of screening.","The results of the second round screen were expressed as the fold change in mCherry MFI and as the percentages of mCherry-positive cells, each normalized to the DMSO control.\r\nEMSAs\r\nCy5-labelled SSE-7 bait for EMSAs was generated by PCR amplification of SSE-7 from the plasmid using Cy5-labelled primers (Sigma): 5\u2032-ACTAGGTTACTGGTGCATGCTTGTCCTGCCTTTGAGAACA-3\u2032 and 5\u2032-ACCCAAACTATTGGAGCGAGAGAAAGGAAAGAAAGAGGTC-3\u2032.","The following primers were used for Cy5-labelled fragments for EMSA: ID4310: 5\u2032-ACTAGGTTACTGGTGCATGCTTGTCCTGCCTTTGAGAACA-3\u2032 and 5\u2032-ACCCAAACTATTGGAGCGAGCTCATTTGAAGCAGAAGAAT-3\u2032; ID4428: 5\u2032-ACTAGGTTACTGGTGCATGCTCTAAATGGAGATCCTCCCA-3\u2032 and 5\u2032-ACCCAAACTATTGGAGCGAGACAGGAGGAAGTAGTAAATC-3\u2032; ID0785: 5\u2032-ACTAGGTTACTGGTGCATGCAAGAATGAACTGGGCCCAGC-3\u2032 and 5\u2032-ACCCAAACTATTGGAGCGAGTGAAAACCAGAGGCATCTCA-3\u2032; and ID3836: 5\u2032-ACTAGGTTACTGGTGCATGCTACCCTTCCAGGGGAGCAGT-3\u2032 and 5\u2032-ACCCAAACTATTGGAGCGAGAGAAAGGAAAGAAAGAGGTC-3\u2032.\r\n\r\nThe binding to Cy5-end-labelled SSE-7 was analysed in native 1% agarose gels (12\u00a0\u00d7\u00a013\u00a0\u00d7\u00a01\u00a0cm), which were prepared in 0.5\u00d7 TBE (45\u2009mM Tris-borate and 1\u2009mM EDTA).","Gels were stored overnight at 4\u2009\u00b0C before pre-running at 120\u2009V (approximately 10\u2009V\u2009cm\u20131) for 1\u2009h. For affinity analysis, a 10\u2009\u03bcl mixture typically containing 50\u2009nM Cy5-labelled full-length or fragmented SSE-7, 100\u2009ng\u2009\u00b5l\u20131 poly(dA:dT) (InvivoGen, tlrl-patn) and 0, 0.5, 1, 2 or 4\u2009\u03bcl of hFib lysates, with and without SOX overexpression (5.56\u2009mg\u2009ml for OE lysates and 3.47\u2009mg\u2009ml\u20131 for hFib control lysates), GSC7 GBM stem, or differentiated, cell lysates (at 0.43\u2009mg\u2009ml\u20131), or purified recombinant SOX protein (at 3.17\u2009\u00b5M). For SOX2 and SOX9 in combination, equimolar amounts of SOX2 and SOX9 were mixed for 0.5, 1, 2, 4 and 8\u2009\u00b5l of protein in total.","Next, 8\u2009\u00b5l SOX protein was added to 25\u2009nM SSE-7 DNA in a 20\u2009\u00b5l volume, and 20\u2009\u00b5l of sample was loaded for gels, which were prepared in 1\u00d7 binding buffer (10\u2009mM Tris HCl pH\u20097.5, 1\u2009mM MgCl2, 10\u2009\u03bcM ZnCl2, 10\u2009mM KCl, 1\u2009mM DTT, 5% (v\/v) glycerol and 0.5\u2009mg\u2009ml\u20131 BSA). The mixtures were incubated at 20\u2009\u00b1\u20091\u2009\u00b0C in the dark for 1\u2009h using DNA LoBind tubes (Eppendorf). The entirety of each sample was then loaded onto agarose gels and electrophoresis was performed at 120\u2009V for 3\u2009h at 4\u2009\u00b0C. The gels were imaged by detecting Cy5 fluorescence using a Bio-Rad ChemiDoc MP imaging system.","The resulting images were visualized, and the bands were quantified using ImageJ 2.8\/1.54i.\r\nAdult human cortex and brain tumour slice cultures\r\nTumour and non-tumour brain tissue were transferred to a 10-cm2 tissue culture dish with sterile PBS and placed on ice. Visibly damaged tissue was removed and the remaining tissue was transferred into a 35-mm2 dish with pre-warmed 3% SeaPlaque agarose (50100, Lonza). After cooling in ice, the block was removed and cut using a scalpel into an approximately 2\u2009cm cube around the brain. Before starting to cut, a 6-well plate was prepared.","In each well, we introduced one cell culture insert (PICMORG50, Millicell) and added below it 1\u2009ml culture medium in basal NSC medium, Dulbecco\u2019s modified Eagle medium: F12 supplemented with EGF and FGF (Life Technologies). The embedded brain was placed in a circular vibratome plate with glue. The vibratome (VT1000 S, Leica) plate was fixed in the platform and filled with PBS and penicillin\u2013streptomycin (15140-122, Gibco, 1:100).","Then, 300-\u00b5m-thick slices were cut, with vibrating frequency set at 10 and speed at 1. Each slice was transferred using a small brush onto the top of a Millipore culture insert. The platform was maintained cool at all times. The 6-well plate was placed in an incubator at 37\u2009\u00b0C with 5% CO2.\r\nAddition of virus\r\nA 5\u2009\u00b5l volume of virus was pipetted onto the centre of each tissue slice, using a 10\u2009\u00b5l tip without touching the tissue. Three repeat doses of 5\u2009\u00b5l virus was added to each slice at 5\u2009min intervals (20\u2009\u00b5l in total per slice). Slices were incubated for 7\u2009days.","The medium was replaced on day\u00a03 or 4.\r\nImmunostaining\r\nThe medium was removed and exchanged for 1\u2009ml freshly prepared 4% PFA; 1\u20132\u2009ml PFA was also placed gently on top to cover the slice. After 2\u2009h, PFA was removed, brain slices were washed 3\u2009times with PBS and transferred using a brush to a 24-well plate. Slices were incubated at room temperature for 1.5\u2009h in blocking solution (0.5% Triton X-100 and 3% goat serum; Sigma-Aldrich) followed by incubation with primary antibody for 2\u2009days at 4\u2009\u00b0C. Sections were washed five times with PBS, the respective secondary antibody was applied in blocking solution overnight.","The next day, slices were washed 5 times with PBS mounted and cleared in RapiClear 1.49 (RC149001, Sunjin Lab).\r\n\r\nImages were acquired using an Opera Phenix Plus high-content imaging system (Revvity) equipped with a \u00d740\/1.1 NA water-immersion objective. Image analysis was performed in Harmony software (v.5.2; Revvity). For the full pipeline see Supplementary Information.","In brief, nuclei were segmented using DAPI with the common thresholding method, and the nuclear mean intensity of channels SOX2 (488) and mCherry (555) were then calculated. Thresholds to define positive cells were set for each tissue slice individually based on 3\u20135 regions of interest. Nestin was segmented using \u2018find image region\u2019 with channel 647, and positive cells were defined as those with nuclei in these regions. Analysis was performed for a single z\u2009plane. Data were exported and analysed using GraphPad Prism.","Representative images of the segmentation can be found in the Supplementary Information.\r\nZebrafish experiments\r\nAll embryos were obtained by natural spawning and collected in conditioned aquarium water in 0.00001% methylene blue. Embryos were raised at 28.5\u2009\u00b0C in embryo medium (E3) on a 14\u2009h light\u201310\u2009h dark photoperiod and were treated with 200\u2009\u03bcM N-phenylthiourea (Sigma) from 6\u2009h post fertilization to inhibit pigmentation. Zygotes were injected at the one-cell stage of development. DNA constructs were created using the Tol2Kit system48,49.","Approximately 2\u2009nl of plasmid DNA (30\u2009ng\u2009\u03bcl\u20131) containing Tol2-capped mRNA (20\u2009ng\u2009\u03bcl\u20131), supplemented with 0.2% w\/v phenol red (Sigma) to facilitate visualization of injected volume, was injected. To induce Akt1 overexpression in neural progenitor cells, a combination of Tol2-pDEST-NBT:DlexPR-lexOP-pA (20\u2009ng\u2009\u03bcl\u20131) and Tol2-pDEST-lexOP:AKT1-lexOP:tagRFP (30\u2009ng\u2009\u03bcl\u20131) plasmids was injected as previously described50. The newly generated tgSSE-7:eGFP was outcrossed to tg(Xla.Tubb:DsRed) and tg(olig1:mScarlet). Live imaging was performed using either a Leica MS205 stereomicroscope or a Leica SP8 confocal microscope with a \u00d740\/NA 1.1 objective.","Animal experimentation was approved by the ethics review committee of the University of Edinburgh and the Home Office in accordance with the Scientific Procedure Act 1986.\r\nAAV medium-exchange transduction assay\r\nHEK cells were seeded into a 6-well plate a day before transfection, aiming for 60\u201370% confluence the next day. On the day of transfection, the culture medium was exchanged and HEK cells were transfected with RepCap, pHelper and AAV-ITR transgene plasmids using polyethylenimine. Two days later, conditioned HEK culture medium (now containing rAAV particles) was collected and centrifuged at 1,300\u2009rpm for 4\u2009min to remove any cells.","rAAV-conditioned medium was transferred to HEK or GSC7 cells seeded at low density (10\u201320%) in 6-well or 12-well plates. Two to three days later, cells were analysed for transduction. For AAV1 in vivo experiments, we purchased research-grade viral vectors from Vector Biolabs (iodixanol gradient ultracentrifugation purified).\r\nAAV transduction assay in iPS cell-derived neurons\r\nGSC7 and E55 cells were seeded in a 24-well plate format the day before transduction, whereas ioGlutamatergic neurons, ioGABAergic neurons and microglia (bit.bio) were seeded in a 48-well format before differentiation. After differentiation, cells were transduced with AAVs at a MOI of 5\u2009\u00d7\u2009105.","Cells were incubated at 37\u2009\u00b0C and 5% CO2 for 3 or 10\u00a0days.\r\nCell seeding and transduction for GCV killing assay\r\nOn the day of cell seeding, cells were detached using the method described above. Cells were counted using a haemocytometer and seeded into 96-well (1,000 cells per well in 50\u2009\u00b5l), 24-well (30,000 cells per well in 500\u2009\u00b5l) or 6-well (60,000 cells per well in 2\u2009ml) Corning plate and placed in an incubator at 37\u2009\u00b0C and 5% CO2. The next day, AAV stocks were thawed at room temperature, and an appropriate amount of virus stock was added to the required amount of culture medium to achieve the final MOI of 5\u2009\u00d7\u2009105. Culture medium containing viral particles was added to cells without the replacement of existing medium.","Cells were returned to a 37\u2009\u00b0C and 5% CO2 incubator.\r\nProdrug treatment\r\nLyophilized GCV was diluted in DMSO to achieve 100\u2009mM stock concentration. GCV stocks were aliquoted and were stored at \u221220\u2009\u00b0C for no longer than 1\u2009month. To make a working stock concentration, GCV was diluted 1:100 in appropriate culture medium, and 20\u2009\u00b5l was added to wells already containing 80\u2009\u00b5l culture medium for a final concentration of 200\u2009\u00b5M. As a negative control, DMSO was diluted 1:100 in appropriate culture medium to achieve a working stock.","Next, 20\u2009\u00b5l of working stock was added to wells already containing 80\u2009\u00b5l culture medium. As a positive control, 20\u2009\u00b5l DMSO was added to wells with 80\u2009\u00b5l culture medium to achieve a final concentration of 20% DMSO.\r\nIncucyte live-cell imaging\r\nTo track cell proliferation and morphological changes during treatments, cells were monitored using an Incucyte live-cell imaging system. Whole-well imaging of Corning 96-well\u00a0plates was performed every 4\u2009h. Basic confluence scoring analysis software (Incucyte) was used to estimate confluence.","Images at specific time points were extracted to verify cell confluence and morphology.\r\nMTT assay\r\nOn the day of assay, the culture medium was replaced with 0.3\u2009mg\u2009ml\u20131 MTT solution (diluted in cell line-appropriate culture medium). Cells were placed in incubator at 37\u2009\u00b0C and 5% CO2 for 3\u2009h. After incubation, the medium was removed and 70\u2009\u00b5l DMSO was added to each well. Each plate was kept in the dark at 37\u2009\u00b0C for 20\u2009min, shaking it occasionally. Before reading the plate, each well was visually inspected to make sure that all (formazan) crystals were dissolved.","Plates were read with plate reader at 560\u2009nm absorbance.\r\nStatistical analysis\r\nData analysis was performed using Microsoft Excel (v.16.23 for Mac), GraphPad Prism (v7) and RStudio (v1.1.456). Error bars are shown as the s.d. of the mean.","For illustrations, BioRender and Adobe Illustrator 22.0.1 were used.\r\n\r\nFurthermore, we used open-source programmes such as bedtools, GREAT, fastasplitter, HOMER and MEME for various analyses45,46.\r\nTumour initiation by transplantation and intratumoral AAV delivery\r\nAll animal procedures were approved by the University of Edinburgh Animal Welfare and Ethical Review Body (AWERB) and conducted under UK Home Office licence (PPL number: PP8631583) in accordance with the Animals (Scientific Procedures) Act 1986 and ARRIVE guidelines. Male and female C57BL\/6 mice (6\u20138\u2009weeks old) were obtained from Charles River.","Mice were housed in Individually ventilated cages with sterile bedding plus enrichment inside a pathogen-free facility on a 12-h light\u2013dark cycle, at a temperature range of 20\u201324\u2009\u00b0C and with relative humidity level of 45\u201365%. Transplantation experiments were performed as previously described25. We used 7-week-old male mice. Mice were transplanted with 200,000 NPE-IE cells in 2\u2009\u00b5l PBS per mouse. NPE-IE cells have been previously described25 and have Nf1 and Pten inactivating mutations, alongside EGFvIII overexpression and a GFP-Luc reporter construct.","At 2\u2009weeks, when tumours were visible by IVIS imaging (bioluminescence) and of similar size, AAV1-SSE-7\u2013HSV-TK\u2013IL-12 virus (1.2\u2009\u00d7\u20091013 viral genomes per ml, Vector Biolabs) was delivered directly into the mouse brain tumour (2.5\u2009\u00b5l per mouse) using a microsyringe linked to an injection pump, at a rate of 0.17\u2009\u00b5l\u2009min\u20131 and 2.3\u2009mm depth, higher than the previous depth of 2.4\u2009mm for tumour cells. Each mouse received 3.00\u2009\u00d7\u20091010 viral genomes. The following day after virus injections, GCV was intraperitoneally injected (500\u2009\u00b5l per mouse at 2\u2009mg\u2009ml\u20131) once daily for 20\u2009days. Mice were imaged by IVIS imaging (PerkinElmer) once a week for luciferase signals.","Animals were observed regularly for any neurological symptoms. Mice with tumours were culled based on the following criteria; (1) body weight loss of \u226520% relative to baseline (weights of mice were recorded once weekly); (2) abnormal neurological signs, including ataxia, head tilt, circling behaviour, seizures, paresis or inability to right itself within 3\u2009s; (3) hunched posture with reduced locomotor activity and lack of reach response; and (4) any extracranial tumours\u00a0\u2265\u00a015\u2009mm in diameter or ulcerated as per the Workman (2010) guidelines51. All mice reaching the humane end point were culled using the schedule\u00a01 method.","The end-point limits were not exceeded in any of the experiments.\r\nImmune characterization of mouse GBMs\r\nBrains were collected and processed 5\u2009days after treatment.","First, they were microdissected under a Leica stereo microscope to avoid contamination with healthy tissue. Tissues were mechanically disaggregated using scissors, followed by 30\u2009min of enzymatic digestion using a mix of 0.35\u00a0\u00b5g\u2009ml\u20131 Liberase TL (05401119001, Roche) and 0.23\u2009\u00b5g\u2009ml\u20131 DNAse (101041590001, Sigma) in plain RPMI. After digestion, tissues were dispersed through a 70\u2009nm cell strainer to obtain a single-cell suspension. After centrifugation, cells were resuspended in PBS for further flow cytometry staining. Cell suspensions were incubated for 10\u2009min at 4\u2009\u00b0C with a blocking mixture of mouse, rat and calf serum containing anti-CD16\/32 blocking antibodies (clone 2.4G2, BioXcell, BE0307).","Cells were washed with PBS and incubated for 20\u2009min in antibody-containing staining buffer plus Fixable Viability Dye eFluor-780 (eBioscience, 65-0865-14) to distinguish live and dead cells. Cells were then washed and resuspended in fixation\/permeabilization buffer (BD CytoFix\/Cytoperm; 554714 or FOXP3 Transcription Factor Staining Buffer set, eBioscience) followed by intracellular target staining.\r\nSABER\u2013FISH detection of transgene copy number\r\nSABER\u2013FISH was performed as previously described52,53. Non-overlapping 36\u2009bp probes were designed to target the HSV-TK transgene, and probes were extended using primer exchange reaction to approximately 500\u2009bp.","For imaging, cells were first stained with mCherry,then fixed with PFA and subsequently processed for immuno-SABER\u2013FISH54.\r\nAssessment of the fraction of the tumour cell population that activates AAV1-SSE-7\u2013mCherry\r\nAAV1-SSE-7-mCMV\u2013HSV-TK-V5\u2013mCherry was injected at around 3.58\u2009\u00d7\u20091014 viral genomes per ml and 7.2\u2009\u00d7\u20091013 viral genomes per ml (1:5 dilution) in PBS. AAV was injected into tumours as described above. After 7\u2009days, tumours were imaged using a stereomicroscope and surgically excised.\r\n\r\nEach tumour sample was added to 1\u2009ml digest mix (RPMI\u2009+\u20091\u2009mg\u2009ml\u20131 collagenase IV, 12.6\u2009\u00b5g\u2009ml\u20131 DNase\u2009I and 1% penicillin\u2013streptomycin) and incubated for 1\u2009h in a shaking incubator at 40\u2009rpm at 37\u2009\u00b0C.","Samples were filtered through a 100\u2009\u03bcm cell strainer, washed with 9\u2009ml FACS buffer and centrifuged at 250g for 15\u2009min at room temperature. The pellet was then resuspended in 1\u2009ml 37.5% Percoll in PBS and gradient centrifuged at 900g, room temperature, for 12\u2009min with no brake. The supernatant was discarded, and the bottom layer of cells resuspend in 500\u2009\u00b5l PBS and filtered into a FACS tube.","Cells were stained with DAPI as a live\/dead stain and analysed using a BD LSR Fortessa cell analyser (four lasers, BD Bioscience).\r\nReporting summary\r\nFurther information on research design is available in the\u00a0Nature Portfolio Reporting Summary linked to this article.\r\n\nData availability\r\nThe scRNA-seq data have been deposited at the European Nucleotide Archive (ENA) (accession number: PRJEB81816). The project number for ChIP\u2013seq data is PRJEB107008. For ChIP\u2013seq, raw (fastq) and processed (bigwig and narrowPeak) files have been deposited and made public (https:\/\/www.ebi.ac.uk\/ena\/browser\/view\/PRJEB107008).","The ENA does not take processed files; therefore, these have been uploaded to BioStudies, which links to the same ENA accession (PRJEB107008), otherwise they are at https:\/\/www.ebi.ac.uk\/biostudies\/studies\/S-BSST2733. All the codes and processing steps for scRNA-seq, SCENIC analysis and ChIP\u2013seq data are available at GitHub (https:\/\/github.com\/alhafidzhamdan\/sse_gene_therapy). The GitHub repository includes some processed files, including all the narrowPeak files and consensus sets; using these in conjunction with the .Rmd document for reproduction of Fig. 2. This also includes a link to the SOX2 and SOX9 tracks on the UCSC browser.","The linked and minted a repo is available at Zenodo (https:\/\/zenodo.org\/records\/18676096)55.\u00a0Source data is provided with this paper.\r\nReferences\r\nSuv\u00e0, M. L. et al. Reconstructing and reprogramming the tumor-propagating potential of glioblastoma stem-like cells.","Cell 157, 580\u2013594 (2014).\r\n\r\nArticle\nPubMed\nPubMed Central\nGoogle Scholar\r\n\r\nLi, C. & Samulski, R. J. Engineering adeno-associated virus vectors for gene therapy. Nat. Rev. Genet.","21, 255\u2013272 (2020).\r\n\r\nArticle\nCAS\nPubMed\nGoogle Scholar\r\n\r\nKribelbauer, J.","F.,","Rastogi, C., Bussemaker, H. J.","& Mann, R. S. Low-affinity binding sites and the transcription factor specificity paradox in eukaryotes. Annu. Rev. Cell Dev. Biol.","35, 357\u2013379 (2019).\r\n\r\nArticle\nCAS\nPubMed\nPubMed Central\nGoogle Scholar\r\n\r\nSpitz, F.","& Furlong, E. E. M. Transcription factors: from enhancer binding to developmental control. Nat. Rev. Genet.","13, 613\u2013626 (2012).\r\n\r\nArticle\nCAS\nPubMed\nGoogle Scholar\r\n\r\nSchlabach, M.","R.,","Hu, J.","K.,","Li, M.","& Elledge, S. J. Synthetic design of strong promoters. Proc. Natl Acad. Sci.","USA 107, 2538\u20132543 (2010).\r\n\r\nArticle\nADS\nCAS\nPubMed\nPubMed Central\nGoogle Scholar\r\n\r\nWu, M.-R. et al. A high-throughput screening and computation platform for identifying synthetic promoters with enhanced cell-state specificity (SPECS). Nat. Commun.","10, 2880 (2019).\r\n\r\nArticle\nADS\nPubMed\nPubMed Central\nGoogle Scholar\r\n\r\nCheng, J. K.","& Alper, H. S. Transcriptomics-guided design of synthetic promoters for a mammalian system. ACS Synth. Biol.","5, 1455\u20131465 (2016).\r\n\r\nArticle\nCAS\nPubMed\nGoogle Scholar\r\n\r\nJindal, G. A.","& Farley, E. K. Enhancer grammar in development, evolution, and disease: dependencies and interplay. Dev.","Cell 56, 575\u2013587 (2021).\r\n\r\nArticle\nCAS\nPubMed\nPubMed Central\nGoogle Scholar\r\n\r\nWhyte, W. A. et al. Master transcription factors and mediator establish super-enhancers at key cell identity genes.","Cell 153, 307\u2013319 (2013).\r\n\r\nArticle\nADS\nCAS\nPubMed\nPubMed Central\nGoogle Scholar\r\n\r\nZamudio, A. V. et al. Mediator condensates localize signaling factors to key cell identity genes. Mol.","Cell 76, 753\u2013766 (2019).\r\n\r\nArticle\nCAS\nPubMed\nPubMed Central\nGoogle Scholar\r\n\r\nHnisz, D.,","Shrinivas, K.,","Young, R.","A., Chakraborty, A. K.","& Sharp, P. A. A phase separation model for transcriptional control.","Cell 169, 13\u201323 (2017).\r\n\r\nArticle\nADS\nCAS\nPubMed\nPubMed Central\nGoogle Scholar\r\n\r\nHnisz, D. et al. Convergence of developmental and oncogenic signaling pathways at transcriptional super-enhancers. Mol.","Cell 58, 362\u2013370 (2015).\r\n\r\nArticle\nCAS\nPubMed\nPubMed Central\nGoogle Scholar\r\n\r\nPollard, S. M. et al. Glioma stem cell lines expanded in adherent culture have tumor-specific phenotypes and are suitable for chemical and genetic screens. Cell Stem Cell 4, 568\u2013580 (2009).\r\n\r\nArticle\nCAS\nPubMed\nGoogle Scholar\r\n\r\nGraham, V.,","Khudyakov, J.,","Ellis, P.","& Pevny, L. SOX2 functions to maintain neural progenitor identity.","Neuron 39, 749\u2013765 (2003).\r\n\r\nArticle\nCAS\nPubMed\nGoogle Scholar\r\n\r\nLujan, E.,","Chanda, S.,","Ahlenius, H.,","S\u00fcdhof, T. C.","& Wernig, M. Direct conversion of mouse fibroblasts to self-renewing, tripotent neural precursor cells. Proc. Natl Acad. Sci.","USA 109, 2527\u20132532 (2012).\r\n\r\nArticle\nADS\nCAS\nPubMed\nPubMed Central\nGoogle Scholar\r\n\r\nGangemi, R. M. R. et al. SOX2 silencing in glioblastoma tumor-initiating cells causes stop of proliferation and loss of tumorigenicity. Stem Cells 27, 40\u201348 (2009).\r\n\r\nArticle\nCAS\nPubMed\nGoogle Scholar\r\n\r\nSingh, D. K. et al. Oncogenes activate an autonomous transcriptional regulatory circuit that drives glioblastoma. Cell Rep.","18, 961\u2013976 (2017).\r\n\r\nArticle\nCAS\nPubMed\nPubMed Central\nGoogle Scholar\r\n\r\nMacLeod, G. et al. Genome-wide CRISPR\u2013Cas9 screens expose genetic vulnerabilities and mechanisms of temozolomide sensitivity in glioblastoma stem cells. Cell Rep.","27, 971\u2013986 (2019).\r\n\r\nArticle\nCAS\nPubMed\nGoogle Scholar\r\n\r\nKuzmin, D. A. et al. The clinical landscape for AAV gene therapies. Nat. Rev. Drug Discov.","20, 173\u2013174 (2021).\r\n\r\nArticle\nCAS\nPubMed\nGoogle Scholar\r\n\r\nFarnham, P. J. Insights from genomic profiling of transcription factors. Nat. Rev. Genet.","10, 605\u2013616 (2009).\r\n\r\nArticle\nCAS\nPubMed\nPubMed Central\nGoogle Scholar\r\n\r\nSoufi, A.,","Donahue, G.","& Zaret, K. S. Facilitators and impediments of the pluripotency reprogramming factors\u2019 initial engagement with the genome.","Cell 151, 994\u20131004 (2012).\r\n\r\nArticle\nCAS\nPubMed\nPubMed Central\nGoogle Scholar\r\n\r\nStricker, S. H. et al. Widespread resetting of DNA methylation in glioblastoma-initiating cells suppresses malignant cellular behavior in a lineage-dependent manner. Genes Dev.","27, 654\u2013669 (2013).\r\n\r\nArticle\nCAS\nPubMed\nPubMed Central\nGoogle Scholar\r\n\r\nHuang, Y.-H.,","Jankowski, A.,","Cheah, K. S.","E.,","Prabhakar, S.","& Jauch, R. SOXE transcription factors form selective dimers on non-compact DNA motifs through multifaceted interactions between dimerization and high-mobility group domains. Sci. Rep.","5, 10398 (2015).\r\n\r\nArticle\nADS\nPubMed\nPubMed Central\nGoogle Scholar\r\n\r\nScott, C. E. et al. SOX9 induces and maintains neural stem cells. Nat. Neurosci.","13, 1181\u20131189 (2010).\r\n\r\nArticle\nCAS\nPubMed\nGoogle Scholar\r\n\r\nGangoso, E. et al. Glioblastomas acquire myeloid-affiliated transcriptional programs via epigenetic immunoediting to elicit immune evasion.","Cell 184, 2454\u20132470 (2021).\r\n\r\nArticle\nCAS\nPubMed\nPubMed Central\nGoogle Scholar\r\n\r\nWang, Q. et al. Tumor evolution of glioma-intrinsic gene expression subtypes associates with immunological changes in the microenvironment. Cancer Cell 32, 42\u201356 (2017).\r\n\r\nArticle\nPubMed\nPubMed Central\nGoogle Scholar\r\n\r\nNeftel, C. et al. An integrative model of cellular states, plasticity, and genetics for glioblastoma.","Cell 178, 835\u2013849 (2019).\r\n\r\nArticle\nADS\nCAS\nPubMed\nPubMed Central\nGoogle Scholar\r\n\r\nYan, Y.-L. et al. A pair of Sox: distinct and overlapping functions of zebrafish sox9 co-orthologs in craniofacial and pectoral fin development. Development 132, 1069\u20131083 (2005).\r\n\r\nArticle\nCAS\nPubMed\nGoogle Scholar\r\n\r\nSantiago-Ortiz, J. L. & Schaffer, D. V. Adeno-associated virus (AAV) vectors in cancer gene therapy. J. Control. Release 240, 287\u2013301 (2016).\r\n\r\nArticle\nCAS\nPubMed\nPubMed Central\nGoogle Scholar\r\n\r\nSwift, E.","A.,","Pollard, S. M.","& Parker, A. L.","Engineering cancer selective virotherapies: are the pieces of the puzzle falling into place?","Hum. Gene Ther.","33, 1109\u20131120 (2022).\r\n\r\nArticle\nCAS\nPubMed\nPubMed Central\nGoogle Scholar\r\n\r\nLasek, W., Zago\u017cd\u017con, R. & Jakobisiak, M. Interleukin 12: still a promising candidate for tumor immunotherapy?\u00a0Cancer Immunol. Immunother.","63, 419\u2013435 (2014).\r\n\r\nArticle\nCAS\nPubMed\nPubMed Central\nGoogle Scholar\r\n\r\nAgliardi, G. et al. Intratumoral IL-12 delivery empowers CAR-T cell immunotherapy in a pre-clinical model of glioblastoma. Nat. Commun.","12, 444 (2021).\r\n\r\nArticle\nADS\nCAS\nPubMed\nPubMed Central\nGoogle Scholar\r\n\r\nHacker, U.","T.,","Bentler, M.,","Kaniowska, D.,","Morgan, M.","& B\u00fcning, H. Towards clinical implementation of adeno-associated virus (AAV) vectors for cancer gene therapy: current status and future perspectives. Cancers 12, 1889 (2020).\r\n\r\nArticle\nCAS\nPubMed\nPubMed Central\nGoogle Scholar\r\n\r\nAldape, K. et al. Challenges to curing primary brain tumours. Nat. Rev. Clin. Oncol.","16, 509\u2013520 (2019).\r\n\r\nArticle\nCAS\nPubMed\nPubMed Central\nGoogle Scholar\r\n\r\nDarmanis, S. et al. Single-cell RNA-seq analysis of infiltrating neoplastic cells at the migrating front of human glioblastoma. Cell Rep.","21, 1399\u20131410 (2017).\r\n\r\nArticle\nCAS\nPubMed\nPubMed Central\nGoogle Scholar\r\n\r\nArrieta, V. A. et al. Immune checkpoint blockade in glioblastoma: from tumor heterogeneity to personalized treatment. J. Clin. Invest.","133, e163447 (2023).\r\n\r\nArticle\nCAS\nPubMed\nPubMed Central\nGoogle Scholar\r\n\r\nLiau, L. M. et al. Association of autologous tumor lysate-loaded dendritic cell vaccination with extension of survival among patients with newly diagnosed and recurrent glioblastoma: a phase 3 prospective externally controlled cohort trial. JAMA Oncol.","9, 112\u2013121 (2022).\r\n\r\nArticle\nGoogle Scholar\r\n\r\nTodo, T. et al. Intratumoral oncolytic herpes virus G47\u2206 for residual or recurrent glioblastoma: a phase 2 trial. Nat. Med.","28, 1630\u20131639 (2022).\r\n\r\nArticle\nCAS\nPubMed\nPubMed Central\nGoogle Scholar\r\n\r\nO\u2019Dwyer, M. R. et al. Nucleosome fibre topology guides transcription factor binding to enhancers. Nature 638, 251\u2013260 (2025).\r\n\r\nArticle\nADS\nPubMed\nGoogle Scholar\r\n\r\nCompany, C. et al. Logical design of synthetic cis-regulatory DNA for genetic tracing of cell identities and state changes. Nat. Commun.","15, 897 (2024).\r\n\r\nArticle\nADS\nCAS\nPubMed\nPubMed Central\nGoogle Scholar\r\n\r\nTaskiran, I. I. et al. Cell-type-directed design of synthetic enhancers. Nature 626, 212\u2013220 (2022).\r\n\r\nArticle\nADS\nGoogle Scholar\r\n\r\nde Almeida, B. P. et al. Targeted design of synthetic enhancers for selected tissues in the Drosophila embryo. Nature 626, 207\u2013211 (2024).\r\n\r\nArticle\nADS\nPubMed\nGoogle Scholar\r\n\r\nMartella, A., Matjusaitis, M.,","Auxillos, J.,","Pollard, S. M.","& Cai, Y. EMMA: an extensible mammalian modular assembly toolkit for the rapid design and production of diverse expression vectors. ACS Synth. Biol.","6, 1380\u20131392 (2017).\r\n\r\nArticle\nCAS\nPubMed\nGoogle Scholar\r\n\r\nMcLean, C. Y. et al. GREAT improves functional interpretation of cis-regulatory regions. Nat. Biotechnol.","28, 495\u2013501 (2010).\r\n\r\nArticle\nADS\nCAS\nPubMed\nPubMed Central\nGoogle Scholar\r\n\r\nHeinz, S. et al. Simple combinations of lineage-determining transcription factors prime cis-regulatory elements required for macrophage and B cell identities. Mol.","Cell 38, 576\u2013589 (2010).\r\n\r\nArticle\nCAS\nPubMed\nPubMed Central\nGoogle Scholar\r\n\r\nMa, W.,","Noble, W. S.","& Bailey, T. L. Motif-based analysis of large nucleotide data sets using MEME-ChIP. Nat. Protoc.","9, 1428\u20131450 (2014).\r\n\r\nArticle\nCAS\nPubMed\nPubMed Central\nGoogle Scholar\r\n\r\nSoufi, A. et al. Pioneer transcription factors target partial DNA motifs on nucleosomes to initiate reprogramming.","Cell 161, 555\u2013568 (2015).\r\n\r\nArticle\nCAS\nPubMed\nPubMed Central\nGoogle Scholar\r\n\r\nKawakami, K. Tol2: a versatile gene transfer vector in vertebrates. Genome Biol.","8, S7 (2007).\r\n\r\nArticle\nPubMed\nPubMed Central\nGoogle Scholar\r\n\r\nKwan, K. M. et al. The Tol2kit: a multisite gateway-based construction kit for Tol2 transposon transgenesis constructs. Dev. Dyn.","236, 3088\u20133099 (2007).\r\n\r\nArticle\nCAS\nPubMed\nGoogle Scholar\r\n\r\nChia, K.,","Mazzolini, J.,","Mione, M.","& Sieger, D. Tumor initiating cells induce Cxcr4-mediated infiltration of pro-tumoral macrophages into the brain.","eLife 7, e31918 (2018).\r\n\r\nArticle\nPubMed\nPubMed Central\nGoogle Scholar\r\n\r\nWorkman, P. et al.\u00a0Guidelines for the welfare and use of animals in cancer research. Br. J. Cancer 102, 1555\u20131577 (2010).\r\n\r\nKishi, J. Y. et al. SABER amplifies FISH: enhanced multiplexed imaging of RNA and DNA in cells and tissues. Nat. Methods 16, 533\u2013544 (2019).\r\n\r\nArticle\nCAS\nPubMed\nPubMed Central\nGoogle Scholar\r\n\r\nWang, S.","K.,","Lapan, S.","W.,","Hong, C.","M.,","Krause, T. B.","& Cepko, C. L. In situ detection of adeno-associated viral vector genomes with SABER-FISH. Mol. Ther. Methods Clin. Dev.","19, 376\u2013386 (2020).\r\n\r\nArticle\nPubMed\nPubMed Central\nGoogle Scholar\r\n\r\nPurshouse, K. et al. Oncogene expression from extrachromosomal DNA is driven by copy number amplification and does not require spatial clustering in glioblastoma stem cells.","eLife 11, e80207 (2022).\r\n\r\nArticle\nCAS\nPubMed\nPubMed Central\nGoogle Scholar\r\n\r\nHamdan, A.","& Wang, Z. alhafidzhamdan\/sse_gene_therapy: For Koeber et al., Nature 2026 (1.0). Zenodo https:\/\/doi.org\/10.5281\/zenodo.18676096 (2026).\r\n\r\nDownload references\r\nAcknowledgements\r\nU.K.,","M.M., S.J.R. and S.M.P. are supported by the BBSRC UK Mammalian Synthetic Biology Centre (BB\/M018040\/1). The GSC and NSC lines were obtained from the GCGR. The GCGR was funded by a Cancer Research UK (CRUK) Accelerator Award (C157\/A21992). A.W. was supported by a BBSRC funded project grant. A.H. was supported by CRUK and the Edinburgh Clinical Academic Training (ECAT) programme (C157\/A29279). D.S. is supported by a CRUK programme foundation award. S.M.P. is supported by CRUK programme grant (DRCNPG-Nov21\\100002). S.A.Q. is funded by a CRUK Senior Cancer Research Fellowship (C36463\/A22246) and a CRUK Biotherapeutic Program grant (C36463\/A20764).","This work was undertaken at University College London with support from the CRUK\u00a0City of London Centre (C7893\/A26233). S.M.P. and S.A.Q. also received support from the CRUK Brain Tumour Award (A28592). We thank J.\u2009Mendrychowski and V.\u2009Kho for technical support; I.\u2009Chambers and J.\u2009M.\u2009Iglesias for their comments on the manuscript; and A.\u2009Corsinotti for technical advice on scRNA-seq.","For open access, the author has applied a Creative Commons Attribution (CC BY) licence to any Author Accepted Manuscript version arising from this submission.\r\nAuthor information\r\nAuthor notes\r\n\r\nUte Koeber\r\n\r\nPresent address: IQVIA RDS, Frankfurt am Main, Germany\r\n\r\nMantas Matjusaitis\r\n\r\nPresent address: Plurify, Cambridge, UK\r\n\r\nRachel White,\u00a0Michelle P.","Clark,\u00a0Carla Jacinto de Sousa\u00a0&\u00a0Ben Southgate\r\n\r\nPresent address: Trogenix, Edinburgh, UK\r\n\r\nFelipe Galvez Cancino\r\n\r\nPresent address: Laboratory of Immune Regulation, NDM Centre for Immuno-Oncology, University of Oxford, Oxford, UK\r\n\r\nThese authors contributed equally: Ute Koeber, Mantas Matjusaitis\r\nAuthors and Affiliations\r\nCentre for Regenerative Medicine, Institute for Regeneration and Repair, University of Edinburgh, Edinburgh, UK\r\n\r\nUte Koeber,\u00a0Mantas Matjusaitis,\u00a0Neza Alfazema,\u00a0Katharine Furlong,\u00a0Zeyu Wang,\u00a0Rachel White,\u00a0Alhafidz Hamdan,\u00a0Pooran Dewari,\u00a0Rosie Willis,\u00a0Michelle P.","Clark,\u00a0Carla Jacinto de Sousa,\u00a0Hei Ip Hong,\u00a0Shahida Sheraz,\u00a0Ben Southgate,\u00a0Justyna Cholewa-Waclaw,\u00a0Sabine Gogolok,\u00a0Gillian M.","Morrison,\u00a0Faye Robertson,\u00a0Anna Williams,\u00a0Paul M.","Brennan,\u00a0Abdenour Soufi\u00a0&\u00a0Steven M. Pollard\r\n\r\nCancer Research UK Scotland Centre, Edinburgh, UK\r\n\r\nUte Koeber,\u00a0Mantas Matjusaitis,\u00a0Neza Alfazema,\u00a0Zeyu Wang,\u00a0Rachel White,\u00a0Alhafidz Hamdan,\u00a0Pooran Dewari,\u00a0Gregoire Morisse,\u00a0Rosie Willis,\u00a0Jin Wang,\u00a0Michelle P.","Clark,\u00a0Carla Jacinto de Sousa,\u00a0Hei Ip Hong,\u00a0Shahida Sheraz,\u00a0Ben Southgate,\u00a0Justyna Cholewa-Waclaw,\u00a0Sabine Gogolok,\u00a0Gillian M.","Morrison,\u00a0Faye Robertson,\u00a0Paul M.","Brennan,\u00a0Dirk Sieger\u00a0&\u00a0Steven M. Pollard\r\n\r\nEdinburgh Pathology, Royal Infirmary Edinburgh, NHS Lothian, Edinburgh, UK\r\n\r\nAlhafidz Hamdan\r\n\r\nInstitute for Neuroscience and Cardiovascular Research, University of Edinburgh, Edinburgh, UK\r\n\r\nGregoire Morisse,\u00a0Jin Wang\u00a0&\u00a0Dirk Sieger\r\n\r\nImmune Regulation and Tumor Immunotherapy Group, Cancer Immunology Unit, Research Department of Haematology, UCL Cancer Institute, London, UK\r\n\r\nMariela Navarette,\u00a0Felipe Galvez Cancino\u00a0&\u00a0Sergio A. Quezada\r\n\r\nSchool of Biological Sciences, University of Edinburgh, Edinburgh, UK\r\n\r\nSusan J. Rosser\r\n\r\nCentre for Clinical Brain Sciences, University of Edinburgh, Edinburgh, UK\r\n\r\nPaul M. Brennan\r\nContributions\r\nS.M.P.","conceived the study, secured funding and supervised the project. U.K. designed and performed in vitro experiments to construct the SSEs, with bioinformatic support from B.S. M.M. designed and performed in vitro experiments related to the therapeutic transgene experiments and selectivity, and with N.A. in the animal efficacy studies. A.W. provided human OPCs. M.P.C. led the human in vitro selectivity experiments. The ChIP\u2013seq analysis was performed by P.D., and A.H. led the bioinformatic analyses. Z.W. designed and performed the scRNA-seq experiments. R.\u00a0Willis performed SABER\u2013FISH.\u00a0K.F. performed the biochemical analysis of SOX interactions, with supervision from A.S.","Key technical support for all the molecular biology was provided by R.\u00a0White and S.G. All zebrafish studies were performed by G.M.","and J.W., with design and supervision from D.S. Manufacture and quality control of all viral vectors was performed by M.M., R.\u00a0White and C.J.d.S. The inhibitor screen was designed and performed by H.I.H., with supervision from S.M.P. The human tissue slices were provided by P.M.B., with AAV testing performed by R.\u00a0Willis, S.S., M.P.C. and J.C.-W. The initial support for the transgene optimization and patient-derived GSC models was from S.G. and G.M.M. F.R. supported mouse model experimental designs and translational relevance. The plasmid library manufacture was supported by S.J.R. and staff at the Edinburgh Genome Foundry. All immune characterization was designed and performed by M.N. and F.G.C., with supervision from S.A.Q.","The manuscript and figures were drafted by S.M.P., with contributions from all authors. All authors reviewed, edited and approved the final version.\r\nCorresponding author\r\nCorrespondence to\nSteven M. Pollard.\r\nEthics declarations\nCompeting interests\r\nS.M.P., U.K. and M.M. are inventors of patents filed by the University of Edinburgh related to the discoveries in this study and to SSEs and associated sequences (WO2022162361A1). S.M.P. is founder of Trogenix, which is developing novel gene therapies for oncology, and is a paid consultant (part-time chief scientific officer, since February 2024) and shareholder.","B.S.,","R.\u00a0White, C.J.d.S. and M.P.C. have recently joined Trogenix as employees (since October 2025). S.A.Q. is a paid member of the scientific advisory board of Trogenix. P.M.B. and F.R. are paid external consultants for Trogenix related to ongoing clinical translation of these technologies. All other authors (N.A.,","K.F.,","Z.W.,","A.H.,","P.D.,","G.M.,","M.N., R.\u00a0Willis.,","J.W.,","H.I.H.,","S.S.,","J.C.-W.,","S.G.,","G.M.M.,","F.G.C.,","A.W.,","S.J.R., D.S. and A.S.) declare no competing interests.\r\n\nPeer review\nPeer review information\r\nNature thanks Ralf Jauch and the other, anonymous, reviewer(s) for their contribution to the peer review of this work.\r\n\nAdditional information\r\nPublisher\u2019s note Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.\r\nExtended data figures and tablesExtended Data Fig. 1 Design of candidate SOX2 GSC-specific peaks and their conversion into a library of candidate enhancer fragments.\r\n(a) Summary of\u00a0the workflow to design and build the enhancer fragment library (DGC: differentiated glioma cell). (b) The genomic annotation of the final library of candidate peaks.","(c) The top 16 fragment hits identified from the initial screen (Fig. 1) were validated in GSC7 GSCs for mNeongreen expression using flow cytometry (median fluorescence intensity shown (left). (d) The fragments have the same level of expression in HEK negative control cells as the minimal CMV (MFI and % of total cell population shown, left and right, respectively). (e) De novo motif analysis using HOMER for the Top32 hits of the validation (using shuffled sequences as background) from Fig.","1, confirms the repertoire of candidate TF motifs that are enriched in the functional enhancer fragments.\r\n\r\nSource data\r\nExtended Data Fig. 2 A dimeric SOX motif that binds SOX9 is a major contributor to activity.\r\n(a) A specific dimeric SOX dimer motif was identified using MEME for de novo motif discovery and is enriched in the set of functional enhancer hits (ACAAAGRGBVNHTKK) and was mutated as depicted (below) for 10 different variants of the ID2904 sequence to determine the functional impact. (b-e) Plasmid-based flow cytometry assay to determine the activity of each variant, in GSCs (GSC7 and E28, respectively).","Interestingly, increased activity was seen with 2\u2009bp reduction of the spacing consistent with the importance of this motif for activity. Each dot represents a biological replicate (2e) One-way Anova test with Dunnett\u2019s multiple comparisons test against ID2904 WT; ** represent p-value of 0.0040 (f) Examples of flow cytometry data used for panels c-f. M(f) Examples of flow cytometry data used for panels c-f. (g) SOX9 protein selectively interacts with ID1101 and ID2904 through the dimer motif.\r\n\r\nSource data\r\nExtended Data Fig.","3 Assessment of SSE-7 activity across a set of seven different patient-derived GSC lines.\r\n(a) Reporter reference controls were used to determine relative SSE activity by flow cytometry across each cell line, with minimal CMV (mCMV) and CMV as negative and positive controls, respectively.","(c, e, g, i, k, m and o) shows the % mCherry positive cells, whereas (d, f, h, j, l, n and p) shows the median fluorescence intensity (MFI) for each cell. E37 is an outlier line and has an atypical MYC amplification and lower levels of SOX2. Error bars represent the SD of the mean, each dot represents a biological replicate.\r\n\r\nSource data\r\nExtended Data Fig. 4 Characterisation of SSEs.\r\n(a) Comparison of activity from all seven SSE variants in GSC7 using flow cytometry to monitor expression of the mNeonGreen reporter. Each dot represents a biological replicate. (b) Time-course of GSC astrocyte differentiation to monitor SSE activity using flow cytometry.","(n\u2009=\u20092 biological replicates, each dot symbol represents the mean, the error bars represent the SD of the mean) (c) Live cell wide-field immunofluorescence imaging for the cells assessed in (b). (d-e) Quantitation of the band intensities from the EMSA experiments in Fig.","2d,e. Data points are mean from n\u2009=\u20093 with error bars of the SD. Solid line represents the Hill equation with the Kd values displayed above the graph. (f) EMSA of SSE-7 DNA using purified recombinant SOX2 and SOX9, alone or in combination (representative image of n\u2009=\u20093). (g) EMSA showing the individual and combinatorial binding of purified recombinant SOX2 and SOX9 to FGF4 enhancer element (40\u2009bp, n\u2009=\u20091) containing a typical SOX2 motif.\r\n\r\nSource data\r\nExtended Data Fig. 5 Single cell mRNA profiling of multiple patient derived GSC lines following transduction with AAV1-SSE-7-mCherry virus.\r\n(a) Distribution of transcriptional subtypes in scRNA data.","UMAP of transcriptional subtype in E17 (n\u2009=\u200929,829 cells), E20 (n\u2009=\u200923,301 cells), E31 (n\u2009=\u200939,823 cells), E34 (n\u2009=\u200927,766 cells), E43 (n\u2009=\u200931,538 cells), E55 (n\u2009=\u200947,291 cells). *P\u2009<\u20090.05, **P\u2009<\u20090.01, ***P\u2009<\u20090.001; NS, not significant. (b) SOX2 and SOX9 are broadly expressed across the diverse transcriptional subtype cell states in GSCs, consistent with their roles as core regulators. See also Extended Data Fig. 6 for cell cycle analysis.\r\nExtended Data Fig. 6 Single cell analysis of cell cycle and SCENIC analysis.\r\n(a) Distribution of cell cycle phases across GSC cell states in scRNA data (as per Extended Data Fig. 5).","(b) Full data for SCENIC analysis of transcription factor modules identifies signalling endpoint transcription factors alongside SOX as correlating with SSE-7 activity. (c) Heatmap across technical replicates of libraries from Fig. 3i.\r\nExtended Data Fig. 7 A screen of small molecule pharmacological inhibitors to identify signalling pathways that may impact the activity of SSE-7 in E55 GSCs.\r\n(a) The primary screen was performed in the E55 cells. Heatmap summarising the fold change of the normalised mCherry MFI in inhibitor-treated samples compared with DMSO. Initial hits identified with red font. The relative comparator is the CMV-SSE7-mCherry positive control.","Any reduction in mCherry selectively in the E55 is due to the different properties of the SSE7. (n\u2009=\u20092 biological replicate; independent screens) (b) Validation of the hit compounds using flow cytometry for mCherry detection across three different concentrations of compound (0.1, 0.3 or 3\u03bcM). The heatmap shows the fold change of the normalised mCherry MFI and percentages of mCherry positive cells of inhibitor-treated samples relative to the DMSO-treated control. (n\u2009=\u20092 biological replicates). Total % mCherry cells (top panel) and mean fluorescence intensity (MFI) bottom panel. Red font are MAPK signalling-related compounds validated with dose response and show selective effects on the SSE-7 compared to CMV promoter.","(c) Flow cytometry histograms for each validated hit compound. Typical example of flow cytometry overlay histograms for E55 GSCs treated with 3 different doses of MAPK signalling-related inhibitors (left panels) and E31 GSCs (right panels). E55 and E31 have distinct transcriptional subtype identities (mesenchymal and developmental\/proneural, respectively). (d) Summary of validated inhibitor hits (red font) and\u00a0their mechanism of action within the MAPK signalling pathway. (e) Flow cytometry analysis of positive and negative SSE-7mCherry cells sorted and reanalysed over a 72-hr time-course shows that SSE-7 activity can be activated in cells that are initially negative, indicative of a readout that is sensitive to signalling state.","n\u2009=\u20093 biological replicates. 2-way ANOVAs followed by two stage setup method were carried out to compare with mCMV neg. *P\u2009<\u20090.05, **P\u2009<\u20090.01, ***P\u2009<\u20090.001; NS, not significant. Note: SC-1 was toxic at the higher doses for E31 and so could not be assessed.\r\n\r\nSource data\r\nExtended Data Fig. 8 Selectivity in tissue expression in transgenic zebrafish embryos.\r\n(a) Reporter transgene for selective expression of SSEs in a zebrafish transgenic reporter model with a GFP reporter, with a transposase (Tol2) reporter designed track transgene insertion using mCherry expression in the eye from the Crystal promoter.","(b) SSE activity is tracked by eGFP levels in live cell imaging of the embryos at 48hpf and is tissue-restricted compared parallel CMV controls for each SSE tested. Note: the minCMV control exhibited GFP-expression in the eyes due to spurious transcription from the Crystal eye promoter. (c) A stable transgenic SSE-7:eGFP reporter zebrafish line confirms expression is restricted to the developing nervous system. Images were acquired using a Leica MS205 stereomicroscope. Scale bars represent 500\u2009\u00b5m for (b) and 728\u2009\u00b5m for (c). (d) Higher magnification image of the spinal cord and nototchord with neuronal reporter (Cyan, as per (e)).","(e) Outcross of tgSSE-7:eGFP to tg(Xla.Tubb:DsRed) reveals no expression in neurons; in (f) outcross of tgSSE-7:eGFP to tg(olig1:mScarlet) shows expression in a subset of oligodendrocyte precursor cells. Images represent maximum intensity projections of confocal stacks (optic tectum and cerebellum shown). Images were captured using a Leica SP8 confocal microscope with a 40X\/NA 1.1 objective. Scale bars represent 30\u2009\u00b5m. (g) Quantification of the OPC data from (f). (h)\u00a0Overexpression of human Akt1 in neuronal cells results in strong activity of\u00a0SSE-7. Images represent maximum intensity projections of confocal stacks. Images were captured using a Leica SP8 confocal microscope with a 40X\/NA 1.1 objective. Scale bars represent 30\u2009\u00b5m.","(i) Immunostaining and flow cytometry (top and bottom, respectively) of human ESC differentiated oligodendrocyte progenitor cells shows that SSE-7 is not active compared to CMV (n\u2009=\u20091). (j) SSE7 versus CMV across a set of three human foetal neural stem cell (NSC) cultures (derived from week of gestation 9, 12 and 17 tissue; NS9FB_B, NS12ST_A and NS17ST_A, respectively). Each dot represents a biological replicate. Error bars represent the SD of the mean.\r\n\r\nSource data\r\nExtended Data Fig. 9 Selective activity of AAV1-SSE7-mCherry in SOX2 expressing cells within the tumour tissue mass and infiltrating margin.\r\n(a)\u00a0Tissue slice-culture experiments were performed across diverse subtypes of high-grade glioma similarly to Fig. 2.","Immunostaining and confocal for mCherry across distinct tumour subtypes shows restricted activity of SSE-7-mCherry in\u00a0the tumour margin relative to CMV. (b) Quantitation of the % of mCherry expression in the tumour versus the margin shows SSE-7 is activated in a subset of GBM cells. (n\u2009=\u20094 for both tumour and margin with both CMV and SSE7). Matched tumour and margin samples were used across multiple tumour types; GBM IDH-WT (isocitrate dehydrogenase wild-type), Grade 4 Astrocytoma IDH-MUT (isocitrate dehydrogenase mutant), Grade 3 high-grade astrocytoma and denoted by distinct symbols, each point is a biological replicate. Error bars represent SD.","Statistical significance was assessed by two-way ANOVA with multiple-comparisons correction (p\u2009=\u20090.0182). n\u2009=\u20094. Image analysis and segmentation pipelines provided in the Supplementary Data. Scale bar = 100 micron. (c) Staining for mCherry alongside SOX2 and Nestin in a recurrent GBM\u00a0sample treated with the AAV1-SSE7-mCherry vector, confirming that SSE-7 is active in a subset of the tumour cells. Left panels, low magnification (no matched normal tissue was available for this recurrent sample).\r\n\r\nSource data\r\nExtended Data Fig. 10 In vitro killing of GSCs, but not fibroblasts, is achieved using SSE7-driven HSV-TK.\r\n(a) Schematic of the AAV-2 transgene encoding both HSV-tk and mCherry used in vitro to assess selective cytotoxicity.","(b and c) Flow cytometry was performed 4 days post-transduction. Each dot represents biological replicate (n\u2009=\u20093), error bars represent SD. (b) % mCherry positive cells and (c) median fluorescent intensity (MFI) of mCherry positive cells. (d) Western blot image developed against V5 and GAPDH in transduced and non-transduced GSC7 and huFB170 cells; the higher weaker band is the small fraction that was uncleaved. (e) Incucyte live cell imaging generate proliferation curves to determine the change in confluence over 10-day time-course for human GSC7 cells transduced with CMV or SSE-7-driven TKv5-P2A-mCherry in presence of 200\u2009\u00b5M GCV, 0.2% DMSO (vehicle control) or 20% DMSO (positive control).","(f; top panels) Live images of GSC7 cells at day 8 demonstrating cell number and morphological changes in cells transduced with constructs in the presence of 200\u2009\u00b5M GCV. (f; bottom panels) Live images of huFB170 cells at day 17 demonstrating cell number and morphological changes in cells transduced with constructs in the presence of 200\u2009\u00b5M GCV. (g) MTT assay for cell viability in GSC-7 and fibroblasts. Each dot represents an independent biological replicate. MTT values were normalized to vehicle-only control. Error bars represent SD (n\u2009=\u20093).\r\n\r\nSource data\r\nExtended Data Fig.","11 Evaluating SSE7-TKv5-P2A-mCherry activity from an AAV vector.\r\n(a) Example of flow cytometry data for GSCs (E17) cells following transduction at MOI of 5E5. (b) Heatmap summarizing mean AAV transduction efficiency in different patient-derived GSC cultures (n\u2009=\u20093, except E34 and E21, which are n\u2009=\u20092). (c) SSE-7 activity in mouse fibroblasts and GSCs. (c) Flow cytometry data comparing transcriptional activity between SSE-7- and CMV-driven TKv5-mCherry cassette in mouse NPE-IE cells (GSCs) and NIH3T3 cells when delivered using rAAV2. (d) MTT viability assay demonstrating GCV dose response kill curves in NPE-IE cells transduced with rAAV carrying specified constructs.\r\n\r\nSource data\r\nExtended Data Fig.","12 Treatment of mice with AAV-SSE7-TK-IL12 results in tumour clearance and long-term protection against rechallenge.\r\n(a) Bioluminescence imaging to track tumour growth with either control (PBS) injection (right panels) or virus plus 3-week ganciclovir treatment (shown are n\u2009=\u200910 mice from the first cohort of mice in Fig. 5b). Right panel are a cohort of parallel fresh controls using the same cells in new cohort of mice, confirming tumour initiation potential. Blue font indicates the new timeline after re-challenge with fresh tumour cells injected (male mice used). (b) Dose de-escalation using IVIS to track tumour growth following treatment with different doses of AAV1-SSE7-HSV-tk-IL12.","(male mice used).\r\n\r\nSource data\r\nExtended Data Fig. 13 Characterisation of immune cells with single or double payloads.\r\n(a) Survival curves of tumour-bearing animals for either single (IL-12) or combination payloads (IL-12 plus TK\/gcv), n\u2009=\u20097 male mice per\u00a0group. Survival was analysed using Kaplan-Meier method. Statistical significance was determined using log-rank (Mantel-Cox) test with p\u2009=\u20090.0009. (b) Non-tumour bearing mice treated with AAV-SSE7-TK-IL12 versus the CMV variant. n\u2009=\u20095 male mice per group. Survival was analysed using Kaplan-Meier method. Statistical significance was determined using log-rank (Mantel-Cox) test with p\u2009=\u20090.0034. The CMV driven payloads are toxic.","(c) Bar plots of the percentage of PD-L1+ cells in MDSCs, monocytes, BMDMs, DCs and microglia.","(d, e) Bar plots of the percentage of CD62L-CD44-, CD62L-CD44+, CD62L+CD44- and CD62L+CD44+ cells in T effector cells or CD8+\u2009T cells, respectively.","(f,g) Bar plots of the percentage of IFNg producing cells in T effector cells.","(h,i) Bar plots of the percentage of IFNg producing cells in CD8+\u2009T cells. (j) Flow cytometry analysis of tumor mass 7 days following AAV1-SSE7-mCherry transduction, reveals that ~20% of the NPE-IE-GFP tumor cell population has strong mCherry expression (left panel, gating of tumor cells; middle panel, mCherry histogram; right panel, quantitation n\u2009=\u20094 independent tumours); 2ul of 3.6E14\/ml (full dose) and 7.2E13\/ml (1:5 dose). (c-i) n\u2009=\u20097 biologically independent animals per group, examined in one experiment. Data are represented as mean values +\/\u2212 SD. Statistical significance was assessed using one-way ANOVA with Tukey\u2019s multiple comparison test.","*P\u2009\u2264\u20090.05; **P\u2009\u2264\u20090.01; ***P\u2009\u2264\u20090.001; ****P\u2009\u2264\u20090.0001.\r\n\r\nSource data\r\nExtended Data Table 1 Predicted target genes and distance from enhancer fragment to transcription start site (TSS)\r\nFull size table\r\nSupplementary informationSource dataRights and permissions\r\nOpen Access This article is licensed under a Creative Commons Attribution 4.0 International License, which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence, and indicate if changes were made.","The images or other third party material in this article are included in the article\u2019s Creative Commons licence, unless indicated otherwise in a credit line to the material. If material is not included in the article\u2019s Creative Commons licence and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this licence, visit http:\/\/creativecommons.org\/licenses\/by\/4.0\/.\r\n\r\nReprints and permissions\r\nAbout this article\r\n\r\nCite this article\r\nKoeber, U., Matjusaitis, M.,","Alfazema, N. et al. Synthetic super-enhancers enable precision viral immunotherapy.\nNature (2026). https:\/\/doi.org\/10.1038\/s41586-026-10329-6\r\n\r\nDownload citation\r\n\r\nReceived: 07 December 2023\r\n\r\nAccepted: 26 February 2026\r\n\r\nPublished: 08 April 2026\r\n\r\nVersion of record: 08 April 2026\r\n\r\nDOI: https:\/\/doi.org\/10.1038\/s41586-026-10329-6"],"images":[{"src":"\/news\/images\/TkbOmERl7ZELY4B6U8bwRFLp7W9qytfUVvStryoOOlUEGhhHL0eXn3oyzuSEao4y0AJ5ESqU9aly0hqeQlF1AuY6HehvMIasaugaW2HxU4xEjpq1kWjw2a5kegGpJxC3tseg50hmbqflgA3MX8A8XkYOCKH4f.jpg","width":0,"height":0,"source":"featured","size":0}]}"
}
}
[4]=>
array(4) {
["file"]=>
string(22) "/app/news/article.html"
["line"]=>
int(11)
["function"]=>
string(18) "news_article_fetch"
["args"]=>
array(1) {
[0]=>
string(70) "SNgSFz1UYwUgnAye8Iq6AwimFnBOT9jdyRE0kXeBSMWDBytDmFUTOmTLbAFJl2KUpiOq4O"
}
}
}