The tumor virus landscape of AIDS-related lymphomas

Abstract

Immunodeficiency dramatically increases susceptibility to cancer as a result of reduced immune surveillance and enhanced opportunities for virus-mediated oncogenesis. Although AIDS-related lymphomas (ARLs) are frequently associated with known oncogenic viruses, many cases contain no known transforming virus. To discover novel transforming viruses, we profiled a set of ARL samples using whole transcriptome sequencing. We determined that Epstein-Barr virus (EBV) was the only virus detected in the tumor samples of this cohort, suggesting that if unidentified pathogens exist in this disease, they are present in <10% of cases or undetectable by our methods. To evaluate the role of EBV in ARL pathogenesis, we analyzed viral gene expression and found highly heterogeneous patterns of viral transcription across samples. We also found significant heterogeneity of viral antigen expression across a large cohort, with many patient samples presenting with restricted type I viral latency, indicating that EBV latency proteins are under increased immunosurveillance in the post-combined antiretroviral therapies era. Furthermore, EBV infection of lymphoma cells in HIV-positive individuals was associated with a distinct host gene expression program. These findings provide insight into the joint host-virus regulatory network of primary ARL tumor samples and expand our understanding of virus-associated oncogenesis. Our findings may also have therapeutic implications, as treatment may be personalized to target specific viral and virus-associated host processes that are only present in a subset of patients.

Figure 1

Transcriptome sequencing coupled with the PathSeq computational analysis pipeline is highly sensitive and specific for detecting presence of novel and known viruses in primary AIDS-related lymphoma samples. (A) The PathSeq pipeline profiles total RNA by random hexamer priming and high-throughput sequencing. Reads that do not map to the human genome are aligned against known viral and bacterial genomes. The remaining unmappable reads are assembled into contigs to detect transcripts originating from novel viruses. (B) PathSeq is sensitive and specific for detecting the EBV in tumor cells. Viral read count fraction (y-axis) is shown for all ARL samples (x-axis). Using a threshold of 0.01%, we were able to uniquely identify all EBV-positive cases, which were confirmed by in situ hybridization for the EBV encoded RNA 1 (EBER1) transcript. A complete listing of read counts, viral ISH detection, and case identifiers can be found in Table 1. (C) Samples with <0.01% of reads aligning to EBV were profiled by ISH for EBER1 to determine if any of these cases were mislabeled as EBV negative by PathSeq. While scattered EBV⁺ cells infiltrated the tumor, tumor cells were in fact negative for EBV. This is exemplified by cases LY09, LY20, and LY24 with morphologically nontransformed tumor infiltrating lymphocytes highlighted. LY18 is shown to demonstrate EBER1 staining of EBV⁺ samples with EBV RNA in tumor cells.

Figure 2

Transcriptional regulation of EBV gene expression demonstrates diversity of latency types. (A) Gene expression is shown as read counts per million reads aligned (y-axis) at canonical type I latency genes EBER1, EBER2, and EBNA1. Rows show individual samples and are colored alternating red/blue to emphasize sample-track pairing. (B) Immunogenic viral genes associated with type II and III latency are also expressed, such as LMP1. (C) EBV gene expression is highly diverse across ARL samples. The heatmap shows gene expression profiles of latency-associated genes (y-axis) across samples (x-axis). Gene expression is in units of RPKM normalized by fraction of reads aligning to EBV (“Methods”). (D) Genes associated with lytic reactivation, such as transcripts at the BamHI H locus, were expressed in primary ARL tumor samples. (E) Immunohistochemical staining for viral proteins EBNA2 and LMP1 was used to classify viral gene expression latency in ARL cases. A BL sample with latency I (upper) and a DLBCL sample with latency III (lower) are representative examples. A complete analysis can be found in Table 2.

Figure 3

Viral antigen expression variation is comparable in immunodeficient and immunocompetent hosts. (A) Analysis of polyadenylated RNA-seq data confirms that viral expression is detectable in HIV-negative sBL in immunocompetent hosts. Expression of latency I associated EBNA1 is robustly detected in primary samples (prefixed by SLN) and cell lines. (B) LMP1 and LMP2 RNA were detected in most primary HIV-negative sBL tumor samples. SLN2521 is excluded in the right panel due to low read counts. (C) Viral RNA is translated into immunogenic protein expression in a subset of cases (additional analysis in Table 3). IHC stains in 4 representative cases demonstrate that ARL (left) and sBL (right) cases contain highly immunogenic viral proteins. All samples are positive for the respective stains.

Figure 4

EBV-positive ARL cases have a distinct host gene expression signature. (A) HSPA7 is an example of a gene that is more highly expressed in EBV⁺ samples (above) compared with EBV⁻ samples (below). (B) Genes consistently up- and downregulated in the presence of EBV are shown in DLBCL cases. Axes show log₂(x + 1), where x is the gene expression level in units of RPKM as quantified by cufflinks. (C) EBV-associated host gene expression changes are significantly similar in AIDS-related BL and DLBCL. Significantly differentially expressed upregulated (upper) and downregulated (lower) genes in EBV-positive samples are part of the EBV signature in both BL and DLBCL (hypergeometric test). (D) EBV-associated host gene expression changes cluster into pathways, including plasma cell differentiation, maintenance of hematopoietic stem cells, and genes dysregulated in primary effusion lymphoma. Heatmap shows pathways (y-axis) enriched for host expression signatures associated with EBV in a variety of cellular and host settings (x-axis), including HIV-negative sBL, AR BL, AR DLBCL, and LCLs derived from whole peripheral blood infected with high titers of EBV. Enrichment significance is determined by 1-sided Kolmogorov-Smirnov tests.