Synthetic super-enhancers enable precision viral immunotherapy | Nature

## Summary
To identify candidate GSC-selective enhancers, we first re-analysed previously published datasets 1 that mapped SOX2-binding sites and used chromatin immunoprecipitation with sequencing (ChIP–seq) to define those specific to GSCs and lost in their differentiated progeny. The inset shows the enriched SOX dimer motif identified using the MEME tool. d , Validation 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 = 3), error bars represent the s.d. of the arithmetic mean). GO, 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–SOX9 peaks with a consensus set of GSC super-enhancers ( n = 44 publicly available GSC H3K27ac datasets; Methods ). e , Gene ontology analysis of co-bound SOX2–SOX9 peaks in super-enhancers. f , Bar plot showing the distribution of SOX2–SOX9 expression correlation values ( y axis) across diverse cancer types from The Cancer Genome Atlas ( x axis). 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–seq in the seven individual GSCs at PTPRZ1 and CDK6 loci, annotated by the presence of GSC super-enhancers, co-bound SOX2–SOX9 peaks and the functional SOX dimer motif. cCREs, 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–SOX9 enhancers. Full size image a , 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 bars represent the s.d. of the mean). b , SSE activity detected using an mNeonGreen reporter across GSC lines (GSC7 and E28) and control fibroblasts.

## Article Content
Download PDF
Subjects
Cancer immunotherapy
Targeted gene repair
Abstract
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. Single-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.
Main
Gene 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 bp) synthetic TFBMs and assembling them into concatamers
5
,
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 selectivity
8
.
Cell-type-specific enhancers are often clustered in the genome to create super-enhancers
9
,
10
. Super-enhancers typically regulate transcriptional programs associated with cell identity through the local clustering of low-affinity motifs
8
. This leads to increased high-density binding of lineage and signalling TFs to promote RNA polymerase II recruitment and high transcriptional output
11
,
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.
We 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 GSCs
1
,
16
,
17
,
18
. SOX2 is also a crucial master regulatory TF that is broadly expressed across diverse GSC subtypes
18
.
Here we develop SSEs for the targeted expression of anticancer payloads in GSCs. A single, locally delivered dose of adeno-associated virus (AAV)-SSE-7
19
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).
GSC-specific enhancer fragment design
TFs bind to thousands of sites in the genome, but only a small subset are functional enhancers
20
. To identify candidate GSC-selective enhancers, we first re-analysed previously published datasets
1
that mapped SOX2-binding sites and used chromatin immunoprecipitation wit

---

## Expert Analysis

### Merits
- 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.
- The inset shows the enriched SOX dimer motif identified using the MEME tool. d , Validation 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 = 3), error bars represent the s.d. of the arithmetic mean).
- This result suggests that dimer binding is a crucial feature of the activity of the motif and is probably important for SOX9 binding.
- SSEs for GBM Our goal was to create a GSC-selective transcriptional switch for cancer gene therapy that combines strong expression and high selectivity.

### Areas for Consideration
N/A

### Implications
- This approach may have broader applications in other contexts when precise control of transgene expression in specific cell states is necessary.
- Such SSEs could have substantial value in gene therapy applications.
- The 160 bp 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.
- This result suggests that the functional enhancer fragments potentially bind SOX9.

### Expert Commentary
This article covers enhancers, fig, gsc topics. Notable strengths include discussion of enhancers. Readability: Flesch-Kincaid grade 0.0. Word count: 2313.