Comprehensive Analysis of the Carcinogenic Process, Tumor Microenvironment, and Drug Response in HPV-Positive Cancers

Abstract

Human papillomavirus (HPV) is a common virus, and about 5% of all cancers worldwide is caused by persistent high-risk HPV infections. Here, we reported a comprehensive analysis of the molecular features for HPV-related cancer types using TCGA (The Cancer Genome Atlas) data with HPV status. We found that the HPV-positive cancer patients had a unique oncogenic process, tumor microenvironment, and drug response compared with HPV-negative patients. In addition, HPV improved overall survival for the four cancer types, namely, cervical squamous cell carcinoma (CESC), head and neck squamous cell carcinoma (HNSC), stomach adenocarcinoma (STAD), and uterine corpus endometrial carcinoma (UCEC). The stronger activity of cell-cycle pathways and lower driver gene mutation rates were observed in HPV-positive patients, which implied the different carcinogenic processes between HPV-positive and HPV-negative groups. The increased activities of immune cells and differences in metabolic pathways helped explain the heterogeneity of prognosis between the two groups. Furthermore, we constructed HPV prediction models for different cancers by the virus infection score (VIS) which was linearly correlated with HPV load and found that VIS was associated with drug response. Altogether, our study reveals that HPV-positive cancer patients have unique molecular characteristics which help the development of precision medicine in HPV-positive cancers.

Keywords: HPV; carcinogenic process; drug response; prediction model; tumor microenvironment.

Figure 1

Survival analysis for HPV-related cancer. (A) The HPV-positive group had significantly higher survival rates compared with the HPV-negative group in four cancer types (CESC, cervical squamous cell carcinoma; HNSC, head and neck squamous cell carcinoma; STAD, stomach adenocarcinoma; UCEC, uterine corpus endometrial carcinoma). (B) Forest plot of the multivariate Cox regression analysis of HPV and clinical indicators with stepwise selection. The red horizontal lines correspond to the 95% CI, on which the dot reflects the hazard ratio. (Nx, regional lymph nodes could not be evaluated; N1, lymph node metastases with a maximum diameter of less than 3 cm; N2, lymph node metastases with a maximum diameter of less than 6 cm and greater than 3 cm; N3, the maximum diameter of metastatic lymph nodes is greater than 6 cm).

Figure 2

The different carcinogenic processes between HPV-positive group and HPV-negative group. (A) A waterfall plot showing the significant differences in driver gene mutations between the HPV-positive group and HPV-negative group in three cancer types (CESC, cervical squamous cell carcinoma; HNSC, head and neck squamous cell carcinoma; UCEC, uterine corpus endometrial carcinoma). The top panel shows the HPV status of each sample. Red boxes represent gene mutation, and while gray boxes represent wild-type. (B) A bubble plot shows the significant differences of DDR (DNA damage repair) pathway activity, genomic instability indicators, and mitotic oncogenic pathway activity between the HPV-positive group and HPV-negative group. The size of the bubble represents FDR, and the color represents upregulation or downregulation. The color of the label on the Y-axis represents the different carcinogenic processes. (C) The NRPM (normalized reads per million) of HPV in different cancer types. The threshold of HPV infection was signed by a black horizontal line.

Figure 3

The difference of immune and stroma cell types between the HPV-positive group and HPV-negative group. The red bubbles represent the significantly upregulated abundance of immune cell infiltration, and the blue ones represent that significantly downregulated. The sizes of the bubbles represent FDR, and different cell types are marked by different colors on the Y-axis label (CESC, cervical squamous cell carcinoma; HNSC, head and neck squamous cell carcinoma; UCEC, uterine corpus endometrial carcinoma; COAD, colon adenocarcinoma; GBM, glioblastoma multiforme; OV, ovarian serous cystadenocarcinoma; ESCA, esophageal carcinoma; STAD, stomach adenocarcinoma).

Figure 4

The HPV impact on the tumor microenvironment. (A, B) show the significant changes in immune indicators and metabolic pathway activity, respectively. The size of bubble represents FDR, and the color represents upregulation or downregulation. All the p values of the non-parametric test have been corrected by FDR. (C) The figure shows significantly different expression of HPV-integrated protein-coding genes between the HPV-positive group (yellow) and the HPV-negative group (blue). The red dots in the box diagram are the expression of HPV-integrated genes in HNSC.

Figure 5

Construction of HPV the prediction model. After UMAP (Uniform Manifold Approximation and Projection) dimensionality reduction, all samples are projected to a two-dimensional coordinate system. The colors of the points in (A) and (B) represent the different clusters after DBSCAN (Density-Based Spatial Clustering of Applications with Noise) clustering and HPV status in different cancer types, respectively. (C) The Spearman correlation between VIS and NRPM in each cancer type. (D) A heatmap shows the AUC values of prediction models which were trained in one cancer type (rows) and applied to the others (columns). (E) Two GEO data sets were used for external verification of HPV status prediction models derived from TGCA HNSC and CESC data.

Figure 6

The connection between VIS and drug response. (A) Distribution of scaled VIS levels for TCGA stage III and stage IV samples according to the RECIST standard. (B) Proportion of chemotherapy response in different groups segmented by scaled VIS. (C) Differences in drug IC50 values between high- and low-scaled VIS groups in the GDSC cell line data. (D–F) Show the significant differences in HLA family expression, immune cell infiltration, and immune checkpoint gene expression between high- and low-scaled VIS groups. All the p values of nonparametric test have been corrected by FDR (* represent fdr < 0.05, ** represent fdr < 0.01 and *** represent fdr < 0.001).