Super-enhancer driven SOX2 promotes tumor formation by chromatin re-organization in nasopharyngeal carcinoma

Abstract

Background: Nasopharyngeal carcinoma (NPC) is a malignant head and neck cancer with a high incidence in Southern China and Southeast Asia. Patients with remote metastasis and recurrent NPC have poor prognosis. Thus, a better understanding of NPC pathogenesis may identify novel therapies to address the unmet clinical needs.

Methods: H3K27ac ChIP-seq and HiChIP was applied to understand the enhancer landscapes and the chromosome interactions. Whole genome sequencing was conducted to analyze the relationship between genomic variations and epigenetic dysregulation. CRISPRi and JQ1 treatment were used to evaluate the transcriptional regulation of SOX2 SEs. Colony formation assay, survival analysis and in vivo subcutaneous patient-derived xenograft assays were applied to explore the function and clinical relevance of SOX2 in NPC.

Findings: We globally mapped the enhancer landscapes and generated NPC enhancer connectomes, linking NPC specific enhancers and SEs. We found five overlapped genes, including SOX2, among super-enhancer regulated genes, survival related genes and NPC essential genes. The mRNA expression of SOX2 was repressed when applying CRISPRi targeting different SOX2 SEs or JQ1 treatment. Next, we identified a genetic variation (Chr3:181422197, G > A) in SOX2 SE which is correlated with higher expression of SOX2 and poor survival. In addition, SOX2 was highly expressed in NPC and is correlated with short survival in patients with NPC. Knock-down of SOX2 suppressed tumor growth in vitro and in vivo.

Interpretation: Our study demonstrated the super-enhancer landscape with chromosome interactions and identified super-enhancer driven SOX2 promotes tumorigenesis, suggesting that SOX2 is a potential therapeutic target for patients with NPC.

Funding: A full list of funding bodies that contributed to this study can be found in the Acknowledgements section.

Keywords: Chromatin re-organization; Enhancer connectomes; Genetic variation; Nasopharyngeal carcinoma; Super-enhancers.

Fig. 1

Landscape of H3K27ac signals and essential regulatory genes in NPC. (a) H3K27ac ChIP-seq (upper panel) and HiChIP tracks (lower panel) of essential regulatory genes SLC2A1, MYB, and IGF1R from NPC and NPN tissues and cell lines are shown. In the upper panel, tracks in green, blue and red indicate samples from NPC PDX, NPC and NPN tissues or cell lines, respectively. Curves indicate significant HiChIP link in the lower panel. Enhancer regions were indicated by yellow bars. (b) Patients with high expression level of essential regulated genes SLC2A1, MYB, and IGF1R have poorer overall survival in NPC (log-rank test). Data were collected from GSE102349 and HRA00087 datasets. (c) H3K27ac ChIP-seq heatmap and signal curves at NPC-specific, NPN-specific and shared chromatin regions from NPC PDX, NPC and NPN samples. (d) Unique TF binding motifs enriched in NPC-specific and NPN-specific chromatin regions.

Fig. 2

Atlas of NPC super-enhancers and their biological functions. (a) H3K27ac signals within 12.5-kb window at significant peaks were ranked for the representative NPC PDX, NPC and NPN samples. The inflection point on the plotted curve was selected as the cutoff to separate SEs from typical enhancers. Genes important in oncogenesis are indicated by arrows. (b) GO and KEGG pathway enrichment analysis for super-enhancer regulated genes. Important pathways were labeled as red. (c) H3K27ac ChIP-seq (upper panel) and HiChIP tracks (lower panel) of essential super-enhancer regulated genes ATL2 and SYNPO2 from NPC and NPN are shown. In the upper panel, tracks in green, blue and red indicate samples from NPC PDX, NPC and NPN tissues or cell lines, respectively. Curves indicate significant HiChIP link in the lower panel. Super-enhancer regions were indicated by orange bars.

Fig. 3

NPC super-enhancers connectome of essential driver genes. (a) NPC essential driver genes identified by intersection of SE-linked genes from ChIP-seq and HiChIP dataset and differentially expressed genes from RNA-seq dataset. (b) KEGG pathway enrichment analysis for NPC essential driver genes using ShinyGO (http://bioinformatics.sdstate.edu/go/). (c) Connectome of NPC essential driver genes were analyzed using STRING (https://string-db.org/) with default settings.

Fig. 4

Identification of SOX2 as a major NPC-specific super-enhancer driven oncogene. (a) Identification of SOX2 by intersection of NPC essential driver genes, prognostic-related genes and NPC essential genes from CRISPR screen dataset. (b) Read counts of SOX2 sgRNAs from genome-wide CRISPR screen assay in C666-1 and NP69 cells (n = 2; Sidak multiple comparisons test). (c) H3K27ac ChIP-seq (upper panel) and HiChIP tracks (lower panel) at the SOX2 locus are shown. In the upper panel, tracks in green, blue and red indicate samples from NPC PDX, NPC and NPN tissues or cell lines, respectively. Curves indicate significant HiChIP link in the lower panel. SOX2 enhancers were indicated by red arrows. Super-enhancer regions were indicated by yellow bars. (d) qRT-PCR detection of SOX2 expression level after JQ1 treatment for 48 or 96 h in NP69, NPC43, C17 and C666-1 cells (n = 3; student’s t-test). (e) qRT-PCR detection of SOX2 expression level following SOX2 enhancer repression by CRISPRi in NPC43 cells (n = 3; student’s t-test).

Fig. 5

Pathogenic genetic variation involved in SOX2 SE formation. (a) Comparison of H3K27ac signals and interaction patterns between NPC tissues with or without target SOX2 SNV in enhancer element 1. H3K27ac ChIP-seq (upper panel) and HiChIP (lower panel) tracks at the SOX2 locus are presented. In the upper panel, tracks in red and blue indicate samples from NPC and NPN tissues, respectively. Curves indicate significant HiChIP link in the lower panel. SOX2 enhancer regions with significant signals in NPC samples were indicated by yellow bars. (b) Validation of target SOX2 SNV in a large NPC FFPE cohort by sanger sequencing. Target SOX2 SNV site was indicated by orange bar. (c) Patients with target SOX2 SNV have poorer overall survival in a large NPC FFPE cohort (log-rank test). Patients of SOX2 SNV group includes G/A and A/A genotypes. (d) IHC score of SOX2 in patient groups with or without target SOX2 SNV (Mann–Whitney test). (e) Detailed TF binding pattern alterations caused by target SOX2 SNV. Red dots represented TFs with higher binding affinity in target SOX2 SNV region and blue dots represented TFs with higher binding affinity in corresponding WT region. Motif of TFs specific to target SOX2 SNV region, with detailed SNV positions pointed out by red arrows in their binding sites, are shown in the form of sequence logos. (f) Enrichment of target transcription factors on SOX2 SE1 SNV region (−7436 ∼ −7336), upstream (Enhancer 1-1, −12,222 ∼ −12,045) and downstream (Enhancer 1–2, +15,203 ∼ +15,322) of SE1, and SE5 (+205,873 ∼ 206,059) in NPC PDX samples (n = 3; student’s t-test).

Fig. 6

SOX2 accelerated tumor progression and correlated with poor survival in NPC. (a) SOX2 expression level comparison between NPC tissues and non-cancerous tissues (student’s t-test for GSE118719; Mann–Whitney test for GSE53819). (b) Immunoblot analysis of SOX2 expression in NPC and NPN cell lines. (c) Cell growth was inhibited after SOX2 knock-out in three NPC cell lines. Two-tailed student t-test was applied to analyze the statistical results at the endpoint of the experiments (n = 4). (d) Colony formation assay showed impaired colony formation ability in C666-1 cells after SOX2 knock-down by siRNA treatment (n = 3; student’s t-test). (e) Representative pictures and growth curve of PDX03 with SOX2 knock-down (siSOX2) or vector control (siNC) treatment was plotted by measuring the relative tumor volume at indicated day (n = 5). Two-way analysis of variance (ANOVA) was used for analysis. (f) Tumor weight of PDX03 with SOX2 knock-down (siSOX2) or vector control (siNC) treatment were measured at the endpoint of the experiment (n = 5; student’s t-test). (g) qRT-PCR detection of SOX2 expression level was performed in PDX samples from (e) (n = 5). Each point represented the average SOX2 expression of one mouse (student’s t-test). (h) Patients with high expression level of SOX2 have poorer overall survival in the patient cohort of NPC (log-rank test).