Inhibition of human-HPV hybrid ecDNA enhancers reduces oncogene expression and tumor growth in oropharyngeal cancer

Abstract

Extrachromosomal circular DNA (ecDNA) has been found in most types of human cancers, and ecDNA incorporating viral genomes has recently been described, specifically in human papillomavirus (HPV)-mediated oropharyngeal cancer (OPC). However, the molecular mechanisms of human-viral hybrid ecDNA (hybrid ecDNA) for carcinogenesis remains elusive. We characterize the epigenetic status of hybrid ecDNA using HPVOPC cell lines and patient-derived tumor xenografts, identifying HPV oncogenes E6/E7 in hybrid ecDNA are flanked by previously unrecognized somatic DNA enhancers and HPV L1 enhancers, with strong cis-interactions. Targeting of these enhancers by clustered regularly interspaced short palindromic repeats interference or hybrid ecDNA by bromodomain and extra-terminal inhibitor reduces E6/E7 expression, and significantly inhibites in vitro and/or in vivo growth only in ecDNA(+) models. HPV DNA in hybrid ecDNA structures are associated with previously unrecognized somatic and HPV enhancers in hybrid ecDNA that drive HPV ongogene expression and carcinogenesis, and can be targeted with ecDNA disrupting therapeutics.

Fig. 1. Identification of hybrid ecDNA in HPVOPC using AA and FastViFI and validation of hybrid ecDNA by multi-FISH.

Examples of circular hybrid ecDNA suggested by AA results in HMS001 (A and C) and PDX tumor PDX_A (B and D). Multi-FISH using each ecDNA specific probe (green) and HPV specific probe (red) for metaphase spread cells showed overlapping of each probe signal in the same place only in hybrid ecDNA(+) samples (allow in E and F). Hybrid ecDNA(−) cell line SCC154 only showed the red integrated HPV signal in chromosome (arrowhead in G). Control cell line NOKSI did not show any signal (H). Cell nuclei were counterstained with DAPI (blue). Scale bar shows 10 μm. An expanded view of white box in each is shown upper right corner. Each experiment was at least repeated twice with similar results (E–H).

Fig. 2. Detecting active enhancers using ChIP-seq and identifying HPV integration mechanisms in hybrid ecDNA.

ChIP-seq results of Input, K4me3 (promoter), K4me1 (enhancer), and K27ac (activation mark) for NOKSI (normal control), SCC154 (hybrid ecDNA(−) HPVOPC) and HMS001 (hybrid ecDNA(+) HPVOPC) were shown. Promoter was indicated by the light green box, enhancers by the yellow box, and the sequences included in hybrid ecDNA of HMS001 by the light blue box. The active enhancer was marked by the red oval (top). Expanded hybrid ecDNA sequence with only chimeric reads of each ChIP-seq data were shown, and HPV integration occurred in the exact center of the active enhancer mark (pink line) (bottom) (A). Hybrid ecDNA in HMS001with ATAC-seq (top), ChIP-seq (middle), and RNA-seq (bottom) were shown in the CycleViz plot (B). The same analysis using NOKSI (normal control), PDX_C (hybrid ecDNA(−)HPVOPC), and PDX_A (hybrid ecDNA(+) HPVOPC) were shown. The active enhancer that did not exist in other cell lines made a complex with 2 promoters (bottom) (C and D).

Fig. 3. Human and viral genomes on hybrid ecDNA interacted directly with each other.

Cis-interactions between enhancer and HPV in hybrid ecDNA were analyzed by HiC-seq. Human genome regions on hybrid ecDNA were divided into 2 segments (S1 and S2). The S1 enhancer region interacted with the HPV L1 region (black arrow in A and B), and the S2 enhancer region significantly interacted with the HPV E6/E7 regions in HMS001 (Right green arrow in A and B). Blue dots indicate significant interactions. On the other hand, there was no such interaction in SCC154 that lacked hybrid ecDNA (C). This phenomenon was confirmed in PDX tumors (D–F). In PDX_A, the enhancer existed only in the S1 segment, and the S1 enhancer region significantly interacted with the HPV L1 and E6/E7 regions in PDX_A (black and yellow arrows in D and E). Blue dots indicate significant interactions. On the other hand, there was no such interaction in PDX_C that lacked hybrid ecDNA (F). Although each hybrid ecDNA structure was unique, each of the human enhancer regions closely interacted with HPV, confirming the direct interaction of the human and viral genomes on the hybrid ecDNAs.

Fig. 4. CRISPR interference targeting enhancers on hybrid ecDNA blocks HPV oncogene expression.

CRISPR interference, using dCas9-KRAB to target enhancers on the hybrid ecDNA of HMS001, was performed. gRNAs targeting S1: the long part (gRNA#1) and S2: the short part (gRNA#2) of the enhancer on hybrid ecDNA of HMS001 and nontarget controls were used (A). The expression of dCas9 after doxycycline induction was confirmed by qPCR (B) and Western blotting (C). Two biological replicates were used and the median with SD was shown in qPCR results (B). MYC and GAPDH were used as controls (C). Western blotting results of E6 and E7 of each CRISPRi condition indicate E6 and E7 expression were reduced by CRISPRi targeting the S2 enhancer (D). Proliferation assay results targeting the nontarget control (nonT), S1 enhancer, and S2 enhancer in HMS001, SCC154, and NOKSI were shown. Twelve biological replicates were used and median with SD were shown. Dox induction significantly inhibited the proliferation only in targeting S2 in HMS001 (**P = 0.004, two-tailed student’s t-test) (E), but not in SCC154 or NOKSI (F and G, two-tailed student’s t-test). Source Data is available for panels B and E–G.

Fig. 5. Hybrid ecDNAs interference with each other and reduced after JQ1 treatment.

Multi-probe FISH, using an EYA2 probe and an HPV probe on hybrid ecDNA on super-resolution DMSO and JQ1 treatment of HMS001 cells, is shown (A and B). Hybrid ecDNAs were observed nearby in the nucleus in the “no treatment” condition (A). Each signal (green: EYA2, red: HPV, and blue: DAPI) was also shown separately (bottom) (A). FISH signals of hybrid ecDNAs were reduced after JQ1 treatment (B). Scale bar shows 5 μm (0.2 μm in the expanded picture). Cartoon illustrating how disruption of interferences between hybrid ecDNAs may decrease transcription is illustrated (right in A and B). Each experiment was repeated twice with similar results.

Fig. 6. JQ1 treatment on hybrid ecDNA significantly blocks HPV oncogene expression and proliferation.

JQ1 treatment of HMS001 and SCC154 was performed. The schema of the experiments was shown (A). qPCR results of MYC and E6/E7 using HMS001 were shown. ACTB was used for internal control. Four biological replicates were used. MYC and E6/E7 expressions were reduced in a JQ1 concentration-dependent manner in qPCR using HMS001 (****P < 1 × 10⁻⁴, ****P < 1 × 10⁻⁴, respectively, two-tailed Student’s t-test) at 24 h. Error bars represent SD (B). MYC and E6/E7 expressions were reduced in a JQ1 concentration-dependent manner in western blotting using HMS001 at 6 and 24 h after JQ1 treatment (C). Proliferation assay using JQ1 for HMS001(left) and SCC154 (right) were shown. Five biological replicates were used and median with SD were shown. JQ1 treatment significantly inhibited tumor growth only in HMS001 in 1uM, but not in SCC154 (*P = 0.03, P = 0.12, respectively, two-tailed student’s t-test) (D). Source Data is available for panels B and D.

Fig. 7. JQ1 inhibited proliferation only in hybrid ecDNA(+) HPVOPC PDX tumors.

JQ1 treatment was performed in HPVOPC PDX models. Six PDX_A (hybrid ecDNA+) mice were divided into a vehicle control group and a JQ1 treatment group. Each mouse possessed tumors in both flanks (A). Tumors were harvested after 2 weeks of vehicle control or JQ1 treatment (B). Tumors were harvested after 2 weeks of JQ1 treatment or vehicle control. Tumor volumes of each condition are shown. Six replicates for each group were used. Median with SD (C) and each replicate (D) were shown. In the PDX_A JQ1 treatment group, tumor growth was significantly inhibited compared to the vehicle control group (tumor volume: P = 2 × 10⁻⁵, tumor weight: P = 1 × 10⁻⁴, respectively, two-tailed Student’s t-test). Six replicates for each group were used and the median with SD was shown (C and E). qPCR of E6/E7 is shown between JQ1 treatment and vehicle control. Ten replicates for each group were used and median with SD were shown. E6/E7 expression was also reduced after JQ1 treatment (P < 1 × 10⁻⁴, two-tailed Sstudent’s t-test) (F). JQ1 treatment for hybrid ecDNA(−) PDX (PDX_C and PDX004) was also performed. Six replicates for each group were used and the median with SD was shown (G–L). Neither tumor volume nor tumor weight was inhibited significantly compared to the vehicle control group in PDX_C (G–I) and PDX004 (J–L) (two-tailed Student’s t-test). Source Data is available for panels C–F, H, I, and K, L.