Viruses have been found in nearly 20 percent of cancers and affect a wide range of cancer types. The proportion of patients affected by a given virus varies depending on the cancer and each virus can have unique effects on a cancer's biology. Our recent study published in Cancer Research has shown that each of these viruses elicits distinct immune responses, regardless of the tissues they infect. Preliminary findings suggest that these immune responses predispose some patients to respond to immunotherapies (2018;78:6413-6423).
To date, seven oncogenic viruses have been identified, including hepatitis B and C virus, human papillomavirus, Epstein-Barr virus, human T-cell lymphotropic virus type I, Merkel cell polyomavirus, and Kaposi's sarcoma virus. These viruses have been shown to promote cancer formation through two distinct mechanisms. Human papillomavirus, human T-cell lymphotropic virus type I, Epstein-Barr virus, Merkel cell polyomavirus, and Kaposi's sarcoma virus express oncogenic proteins that directly contribute to oncogenesis, while hepatitis viruses induce chronic inflammation that can indirectly lead to cancer.
Immunotherapies are treatments that harness a patient's own immune system to fight cancer. Immune checkpoint inhibitors, a class of immunotherapy, have represented a paradigm shift in cancer treatment. These therapies function by blocking proteins that prevent the immune system from working properly and have induced tumor remission and substantially prolonged the survival of patients with several types of cancer. Despite the success of these therapies, only a fraction of patients has proven to be responsive to them, creating a need to identify biomarkers that predict treatment response.
The most commonly used biomarker to predict immune checkpoint inhibitor response is tumor mutation burden. Mutations in cancer can lead to altered peptide products, known as neoantigens, that are recognized as foreign by the immune system. Thus, it is hypothesized that tumors with high mutation burdens elicit stronger immune reactions that can facilitate tumor clearance when coupled with a checkpoint inhibitor. Like neoantigens, the proteins expressed by viruses are also recognized as non-self by the immune system, making virus infections another source of antigen that may help predict immunotherapy response.
Clinical trials testing an immune checkpoint inhibitor (anti-PD-1; nivolumab) in two virus-associated cancers, head and neck squamous cell carcinoma and Merkel cell carcinoma, have found that virus-positive patients exhibit higher response rates than virus-negative patients (Lancet Oncol 2016;17:956-965, N Engl J Med 2016;374:2542-2552). Additionally, a trial in gastric cancer identified an EBV-positive patient that exhibited a strong response to avelumab, a checkpoint inhibitor targeting the protein anti-PD-L1 (J Natl Cancer Inst 2018;110:316-320).
However, a study in virus-positive cervical cancers receiving adoptive T-cell therapy revealed that the therapy-induced immune response in these patients was directed at their neoantigens and germline antigens instead of their viral antigens (Science 2017;356:200-205). These disparate results have created a need to investigate how the immune system responds to different cancers, as well as elucidate how these responses vary across different cancer types.
Large-scale multi-dimensional datasets such as The Cancer Genome Atlas (TCGA) have made it possible to perform pan-cancer analyses that compare and contrast different aspects of various cancer types. DNA and RNA sequencing data available in TCGA have recently been used to infer whether a tumor was infected with a specific virus, making this dataset a rich resource to study how virus infections change the biology of different cancer types.
Our study, published in Cancer Research, has integrated this data with immune response information inferred from tumor gene expression profiles to study how different viruses affect the immune response across cancer. In total, this study spanned over 2,000 samples from six different virus-associated cancer types: bladder urothelial carcinoma, cervical squamous cell carcinoma, colorectal adenocarcinoma, head and neck squamous cell carcinoma, liver hepatocellular carcinoma, and gastric cancer (stomach adenocarcinoma and esophageal carcinoma).
Eliciting Distinct Immune Responses
To characterize the altered immune responses associated with a specific virus infection in each cancer type, we used a tumor's gene expression profile to infer the level at which it was infiltrated by four immune cell types: CD8+ T cells, B cells, natural killer cells, and macrophages. We then stratified each cancer type into a virus-positive and virus-negative subset based on whether a sample was infected with a given virus.
This analysis showed that specific viruses were associated with altered immune response profiles. Specifically, infection by different types of human papillomavirus was associated with increased levels of CD8+ T cells, B cells, and natural killer cells in cervical cancer and head and neck cancers, suggesting that human papillomavirus antigens can elicit a strong immune response in these cancer types. Similarly, Epstein-Barr virus associated with increased levels of CD8+ T cells and natural killer cells in gastric cancer. Conversely, hepatitis B infection in liver hepatocellular carcinoma was associated with decreased levels of CD8+ T cells and natural killer cells, indicating that this virus was associated with an immunosuppressed microenvironment.
While infection from these viruses was associated with altered immune response dynamics, the antigens causing these responses remained unclear. To determine whether the immune response was directed at virus-associated antigens, we used RNAseq data to profile the infiltrating T-cell receptor (TCR) repertoires in each TCGA tumor sample. We hypothesized that virus-positive samples would have more clonal T-cell responses, suggesting that the T cells in these tumors expanded in response to a viral antigen.
To determine this association, we compared how TCR diversity, as measured using the number of clonotypes per thousand TCR reads and TCR evenness, differed between samples that were positive for a given virus and those that were negative for all viruses. Interestingly, we found that in most cases, samples that were infected by a given virus did not exhibit differing TCR diversity levels compared to those that were virus-negative. However, samples that were positive for Epstein-Barr virus consistently demonstrated less diverse TCR repertoires, regardless of cancer type. This intriguing result suggests that Epstein-Barr virus antigens are capable of eliciting clonal T-cell responses.
Epstein-Barr as Potential Biomarker
The finding that Epstein-Barr virus infection is associated with elevated immune infiltration and a clonal T-cell response is significant, as both of these factors have been associated with response to immunotherapy. Supporting this, a patient with Epstein-Barr virus-positive, metastatic gastric cancer experienced meaningful clinical benefit after being treated with the anti-PD-L1 antibody, avelumab (J Natl Cancer Inst 2018;110:316-320). Histological analysis of the pre-treatment tumor revealed intense lymphocytic infiltration in the tumor and associated stroma. Interestingly, this tumor did not show evidence of high mutation burden, suggesting that the elevated immune response in this patient was directed toward the viral antigen. This result, combined with our findings using TCGA data, indicate that Epstein-Barr virus-positive gastric cancers represent a good target population for immune checkpoint inhibitors and other immunotherapies.
Beyond gastric cancer, our results suggest that Epstein-Barr virus infection may also predict immunotherapy response in patients with colorectal adenocarcinoma, liver hepatocellular carcinoma, and head and neck squamous cell carcinoma. While infection from this virus is rare in these cancers, the TCR repertoires of the Epstein-Barr virus-positive tumors in each cancer type were consistently the least diverse of all samples. This suggests that Epstein-Barr virus is capable of eliciting a clonal T-cell response in these cancers. Additional studies are now needed in larger cohorts to determine whether these associations are robust.
While our results suggest that patients with Epstein-Barr virus-positive cancers are more likely to respond to immunotherapy, follow-up studies are now needed to assess how effective this marker is at predicting response. Performing these analyses will require additional information from patients who have been treated with immune checkpoint inhibitors and other immunotherapeutic approaches.
Future studies are also needed to determine the antigens the immune system is targeting in other virus-positive cancers. The results from our study indicated that T cells were not primarily targeting viral antigens in tumors infected with viruses that were not Epstein-Barr virus. It is thus possible that immune responses in these cancers are directed toward a patient's neoantigens or tumor-associated antigens.
While there is still much to learn about how viruses affect the tumor immune response, there appear to be several features of virus-positive cancers that make them uniquely treatable. Going forward, it will be exciting to see how these features are leveraged in the clinic to refine therapies and improve patient outcomes.
FREDERICK S. VARN, PHD, is from the Jackson Laboratory for Genomic Medicine, Farmington, Conn. CHAO CHENG, PHD, is Assistant Professor in the Department of Medicine, Institute for Clinical and Translation Research, and Dan L Duncan Comprehensive Cancer Center at the Baylor College of Medicine, Houston.
Copyright © 2019 Wolters Kluwer Health, Inc. All rights reserved.
More on ONCOLOGY-TIMES.com...