HPV Integration in HNSCC Correlates with Survival Outcomes, Immune Response Signatures, and Candidate Drivers

Abstract

The incidence of human papillomavirus (HPV)-related oropharynx cancer has steadily increased over the past two decades and now represents a majority of oropharyngeal cancer cases. Integration of the HPV genome into the host genome is a common event during carcinogenesis that has clinically relevant effects if the viral early genes are transcribed. Understanding the impact of HPV integration on clinical outcomes of head and neck squamous cell carcinoma (HNSCC) is critical for implementing deescalated treatment approaches for HPV⁺ HNSCC patients. RNA sequencing (RNA-seq) data from HNSCC tumors (n = 84) were used to identify and characterize expressed integration events, which were overrepresented near known head and neck, lung, and urogenital cancer genes. Five genes were recurrent, including CD274 (PD-L1) A significant number of genes detected to have integration events were found to interact with Tp63, ETS, and/or FOX1A. Patients with no detected integration had better survival than integration-positive and HPV^- patients. Furthermore, integration-negative tumors were characterized by strongly heightened signatures for immune cells, including CD4⁺, CD3⁺, regulatory, CD8⁺ T cells, NK cells, and B cells, compared with integration-positive tumors. Finally, genes with elevated expression in integration-negative specimens were strongly enriched with immune-related gene ontology terms, while upregulated genes in integration-positive tumors were enriched for keratinization, RNA metabolism, and translation.Implications: These findings demonstrate the clinical relevancy of expressed HPV integration, which is characterized by a change in immune response and/or aberrant expression of the integration-harboring cancer-related genes, and suggest strong natural selection for tumor cells with expressed integration events in key carcinogenic genes. Mol Cancer Res; 16(1); 90-102. ©2017 AACR.

Figure 1. Identified HPV integration sites in the human genome

Viral-host fusional breakpoints found in HPV-positive head and neck cancer samples from TCGA (42 out of 66 samples are integration-positive) and University of Michigan (UM; 9 out of 18 samples are integration-positive). The integration sites are broadly distributed across the genome and are annotated within or near 89 unique human genes. CFSs = common fragile sites.

Figure 2. Assessment of HPV-host fusional breakpoints in fragile and repetitive regions of the human genome

A. Number of breakpoints in common fragile sites (CFS), rare fragile sites (RFS) and non-fragile regions (NFR), compared to what is expected by chance. HPV is not more prone to integrate into fragile sites in the human genome in HNSCC tumors. B. Number of breakpoints in repetitive regions compared to what is expected by chance. Several repetitive element types (LINE, DNA, LTR and all repeats combined) are determined to be significantly enriched for HPV integrations (Chi-squared test p-values with FDR adjustment = 7.42E-05, 0.015, 8.40E-07, and 4.41E-10 correspondingly). (See also Supplemental Table S6 for a further breakdown by repeat family).

Figure 3. Genes associated with a detected integration are enriched with HNSCC, urogenital and lung neoplasm related genes, and are often up-regulated

A. Protein interaction network constructed from the 89 human genes associated with integration event(s); displayed is the highly-connected subnetwork consisting of the 65 (of the 89) host genes that had direct interactions. Genes ETS, TP63, RUNX1, HNF3, KLF5 and CTGF are hubs, indicated with green rectangles, in this network. A legend for the shapes used for the nodes is provided in MetaCore Quick Refernce Guide https://ftp.genego.com/files/MC_legend.pdf. B. Genes in the network are statistically enriched for relevant human diseases. The –log₁₀(p-values) for enrichment were calculated using MetaCore GeneGO with the hypergeometric distribution. The enrichment was tested for the 89 genes and shows the unadjusted significance levels of enrichment. C. Venn diagram of the genes harboring virus-host fusional breakpoints (HNSCC integration) with genes having mutations for: head and neck SCC (HNSCC mut) [16], lung SCC (Lung SCC mut) [36], and cervical SCC and endocervical adenocarcinoma (Cervical mut) (TCGA, Provisional [23, 24]). Gene symbols in bold were significantly up-regulated in the samples where integration occurred compared to all other samples. No genes were significantly down-regulated. D. Distribution of human gene expression levels categorized by the genic or intergenic region where the integration occurred (exons, introns, upstream of the TSS, or downstream of the TES). The left box for each region represents expression in the samples where the integrations occurred (“integr”); the right boxes represent the average expression of the same genes in all other integration-positive samples (“other”). *** Significant difference (paired t-test p-value = 1.60E-09) in expression when insertional breakpoints occur in a gene exons, ** p-value <0.01 and * p-value <0.05. E. The bars represent the number of genes harboring the insertional breakpoints in each type of region (exons, introns, upstream of TSS, or downstream of the TES). Dark blue bars represent number of genes harboring the insertional breakpoints with significantly elevated expression in samples with integration. Numbers above the bars represent the two-sided Fisher’s Exact Test p-values (unadjusted) calculated from testing whether there are more or fewer samples with elevated expression of the gene harboring the integration (in sample with integration) than expected by chance. Integrations into exons of the genes tend to result in upregulation of these genes in the samples with the integration. Conversely, introns and upstream regions have a trend toward fewer than expected by chance.

Figure 4. Association of HPV integration events with survival and immune response

A. Overall survival for TCGA patients with HPV(+) tumors by integration status and HPV(−) patients; p-value was calculated using a univariate Kaplan–Meier log-rank test. B. Gene Ontology terms for genes differentially expressed by HPV integration status. Shown are the top 10 enriched gene sets by integration status after filtering out functional redundancies in the list of gene sets. Genes with elevated expression in integration-negative samples were most strongly enriched for immune related terms; up-regulated genes in integration-positive tumors were enriched for cell-cell adhesion terms related to RNA metabolism and keratinization. C. HPV gene expression (the log₂(E6E7/E1E2) ratio) and immune cell type specific signatures of analyzed HNSCC tumors. Waterfall plot of E6E7/E1E2 ratio values demonstrate properties of samples whose integration-status was reclassified using RNA-seq versus WGS data (in blue and yellow). The lower panel shows cell type specific expression signatures (only cell types significantly discriminating integration-positive and integration-negative tumors are shown). Samples that were defined as integration-positive from WGS by TCGA and reclassified as integration-negative from RNA-seq (light blue in waterfall plot) are closer to other integration-negative tumors (green) by their E6E7/E1E2 ratio and by expression of cell type specific signatures. These patients (with ID TCGA-CN-5374, TCGA-CR-7404, TCGA-CR-5243, and TCGA-6482) have survival status alive with follow-up of 9.5, 48.4, 84.2, and 11.3 months respectively. The sample defined as integration-negative from WGS by TCGA and reclassified as integration-positive from RNA-seq (yellow in waterfall plot) has characteristics closer to other integration-positive samples by its E6E7/E1E2 ratio and cell type signatures. This patient (TCGA-HD-7832) did not have any follow-up.