Geographic EBV variants confound disease-specific variant interpretation and predict variable immune therapy responses

Abstract

Epstein-Barr virus (EBV) is a potent carcinogen linked to hematologic and solid malignancies and causes significant global morbidity and mortality. Therapy using allogeneic EBV-specific lymphocytes shows promise in certain populations, but the impact of EBV genome variation on these strategies remains unexplored. To address this, we sequenced 217 EBV genomes, including hematologic malignancies from Guatemala, Peru, Malawi, and Taiwan, and analyzed them alongside 1307 publicly available EBV genomes from cancer, nonmalignant diseases, and healthy individuals across Africa, Asia, Europe, North America, and South America. These included, to our knowledge, the first natural killer (NK)/T-cell lymphoma (NKTCL) EBV genomes reported outside of East Asia. Our findings indicate that previously proposed EBV genome variants specific to certain cancer types are more closely tied to geographic origin than to cancer histology. This included variants previously reported to be specific to NKTCL but were prevalent in EBV genomes from other cancer types and healthy individuals in East Asia. After controlling for geographic region, we did identify multiple NKTCL-specific variants associated with a 7.8-fold to 21.9-fold increased risk. We also observed frequent variations in EBV genomes that affected peptide sequences previously reported to bind common major histocompatibility complex alleles. Finally, we found several nonsynonymous variants spanning the coding sequences of current vaccine targets BALF4, BKRF2, BLLF1, BXLF2, BZLF1, and BZLF2. These results highlight the need to consider geographic variation in EBV genomes when devising strategies for exploiting adaptive immune responses against EBV-related cancers, ensuring greater global effectiveness and equity in prevention and treatment.

Graphical abstract

Figure 1.

Selection of novel and publicly available EBV genomes for analysis. (A) Schematic illustrating the source, histologic type, and number of EBV⁺ hematologic malignancies from Guatemala, Peru, Malawi, Taiwan, and the United States that underwent EBV-genome sequencing and the source of publicly available reference and cancer-associated EBV genomes that were utilized for this study. (B) Maps demonstrating the country of origin and number of EBV genomes after filtering based on sequence quality. Publicly available EBV genomes are indicated in black, and novel EBV genomes generated by this study are indicated in red. CAEBV, chronic active EBV disease; GC, gastric cancer; IM, infectious mononucleosis; misc, miscellaneous; PTLD, posttransplant lymphoproliferative disorder; QC, quality control.

Figure 2.

Phylogenetic analysis of reference, noncancer-associated, and cancer- associated EBV genomes. The inner circle represents the histology for cancer- associated genomes or other phenotype, the middle circle represents the country of origin, and the outer circle represents the region of origin. Genomes are aligned to the type I reference genome NC_007605 assembled from data for B95-8 (V01555) and Raji (M35547) genomes. CAEBV, chronic active EBV disease; GC, gastric cancer; IM, infectious mononucleosis; misc, miscellaneous; pLELC, primary lymphoepithelioma-like carcinoma; PTLD, posttransplant lymphoproliferative disorder.

Figure 3.

Map of variant frequency across EBV genomes. (A) Identification and frequency by region and histology of all variants identified in 1376 EBV genomes meeting predetermined quality metrics. (B) Comparison of variant detection in standard NGS and duplex sequencing. Bar colors represent the nonsynonymous variant type detected. (C) Comparison of frequency of previously identified disease–associated nonsynonymous mutations in NKTCL and NPC. Variants previously associated with NPC are starred, all others were previously associated with NKTCL.

Figure 4.

Specific EBV genome variants are enriched in or depleted from NKTCL. (A) Volcano plot demonstrating all variants identified. Variants above the horizontal red line indicate variants with a Bonferroni corrected P value <.05. Significant variants with log(OR) less than −1.5 are shown in red (lower risk), and log(OR) >1.5 in blue (higher risk). (B) Heat map demonstrating a correlation between variants significantly enriched in NKTCL–derived EBV genomes with log(OR) >1.5.

Figure 5.

EBV genome variation in intervals encoding prototypic T-cell epitopes is predicted to generate variant peptides with altered binding to class I MHC molecules. (A) Heat map of predicted binding affinity (BA) of 30 prototypic EBV- encoded CD8⁺ T-cell epitopes (rows) to 16 class I MHC alleles (columns) and of the corresponding peptides encoded by EBV genome variants. The far-right column contains a heat map of the frequency of the EBV genome variant that encodes each peptide in the 1376 EBV genomes meeting predetermined quality metrics in this study. (B) Heat map indicating the presence (blue) or absence (white) of the class I MHC alleles known to present 6 prototypic EBV-encoded peptides to CD8⁺ T cells in the 50 most frequent class I MHC haplotypes in Guatemala (left) and China (right). The associated class I MHC allele is shown in bold beneath each peptide. NB, no binding; SB, strong binder; WB, weak binder.

Figure 6.

Frequency of nonsynonymous single nucleotide variants in 6 canonical protein-coding genes in EBV genomes from 6 global regions. All 6 genes have been the focus of EBV vaccine development efforts. The frequency (%) is calculated as the (number of EBV genomes from each region carrying the indicated sequence variant)/(total number of EBV genomes analyzed from that region) × 100.