Genome‐Wide Analysis of Epstein‐Barr Virus Isolated from Extranodal NK/T‐Cell Lymphoma, Nasal Type

Extranodal natural killer cell/T cell lymphoma (NKTCL) has been associated with Epstein‐Barr virus (EBV) infection. This article describes the distribution of genetic variation in EBV genomes isolated from NKTCL biopsy samples and evaluates genomic diversity in the EBV strains.

Abstract

Background.

Extranodal natural killer (NK) cell/T‐cell lymphoma (NKTCL), a rare type of non‐Hodgkin's lymphoma, has strongly been associated with Epstein‐Barr virus (EBV) infection. However, there are no EBV genomes isolated from NKTCL, and the roles the variations of EBV strains play in the pathogenesis of NKTCL are still unclear.

Materials and Methods.

In this study, whole EBV genomes from eight primary NKTCL biopsy specimens were obtained using next‐generation sequencing, designated NKTCL‐EBV1 to NKTCL‐EBV8.

Results.

Compared with the six mostly referenced EBV strains, NKTCL‐EBVs closely resemble the GD1 strain but still harbor 2,072 variations, including 1,938 substitutions, 58 insertions, and 76 deletions. The majority of nonsynonymous mutations were located in latent and tegument genes. Moreover, the results from phylogenetic analysis of whole NKTCL genomes and specific genes demonstrated that all the NKTCL‐EBVs were related to Asian EBV strains. Based on the amino acid changes in certain residues of latent membrane protein 1 (LMP1) and EBV‐determined nuclear antigen 1 (EBNA1), all the NKTCL‐EBVs were sorted to China 1 and V‐val subtype, respectively. Furthermore, changes in CD4+ and CD8+ T‐cell epitopes of EBNA1 and LMP1 may affect the efficacy for a cytotoxic T lymphocyte (CTL)‐based therapy.

Conclusion.

This is the first large study to our knowledge to obtain EBV genomes isolated from NKTCL and show the diversity of EBV genomes in a whole genome level by phylogenetic analysis.

Implications for Practice.

In this study, the full‐length sequence of Epstein‐Barr virus (EBV) isolated from eight patients with nasal natural killer/T‐cell lymphoma (NKTCL) was determined and further compared with the sequences previously reported isolated from other malignancies. Phylogenetic analysis showed that NKTCL‐EBV strains are close to other Asian subtypes instead of non‐Asian ones, leading to the conclusion that EBV infections are more likely affected by different geographic regions rather than particular EBV‐associated malignancies. Therefore, these data have implications for the development of effective prophylactic and therapeutic vaccine approaches targeting the personalized or geographic‐specific EBV antigens in these aggressive diseases.

Introduction

Epstein‐Barr virus infection has been shown to be associated with the pathogenesis of several malignant diseases, including lymphoma, nasopharyngeal carcinoma (NPC), and a subset of gastric carcinoma [1]. Among them, extranodal natural killer (NK) cell/T‐cell lymphoma (NKTCL), a rare type of non‐Hodgkin's lymphoma, is characterized by the presence of Epstein‐Barr virus (EBV) in virtually all cases, irrespective of their ethnicity or geographical origin. NKTCL is an aggressive malignancy, predominantly occurs in the nasal, paranasal, and oropharyngeal sites [2], and is much more prevalent in East Asia [3], [4], [5], [6], [7] and Latin America [3], [8] than in Western countries. Although the association of this B lymphotropic virus with malignancies of T and NK cell origin was quite unexpected, both the presence of virus sequences in tumor cells and the virus's oncogenic potency have led to the hypothesis that whether particular EBV strains are preferentially selected in NKTCL. Moreover, understanding more about EBV sequence variations isolated from NKTCL is essential for developing cancer vaccine for immunotherapy and for monitoring dynamically in different stages of the diseases using the circulating EBV‐DNA, which is derived from apoptotic and necrotic cells [9]. However, few studies have evaluated diversity in more than a single region of the EBV genome in NKTCL, and the full spectrum of diversity existent within the EBV genome from NKTCL tissue has not been adequately studied.

Since the first complete EBV genome sequence, B95‐8, was published in 1984, more than 100 EBV genomes have been reported isolated from NPC, lymphoma, gastric cancer, and lung cancer, as well as from healthy carriers [10]. Recently, we have described an EBV genome capture method to assay EBV genomes isolated from gastric cancer [11] and lung cancer [12], as well as a combined analysis of these alongside previously published sequences. Genome‐wide study of EBV provides the first opportunity to test the general validity of the EBV genetic map and explore recombination, geographic variation, and the major features of variation in this virus.

In the present study, we obtained eight EBV genomes isolated from NKTCL biopsy samples, designated NKTCL‐EBV1–8, and then illustrated the distribution of genetic variation in these EBV strains and further phylogenetic analysis to evaluate their genomic diversity.

Materials and Methods

Patients with NKTCL

The tissue specimens were biopsied and then collected after obtaining informed written consent from patients diagnosed with NKTCL at Peking University Cancer Hospital, Beijing, China, from 2014 to 2015.

Collection of the tumor samples from the patients with NKTCL was approved by the institutional review board of the hospital for the purpose of EBV genomic sequencing. The clinical information for the patients with NKTCL included in this study is summarized in supplemental online Table 1. The extraction of DNA was performed within 1 hour after obtaining biopsy specimens and temporarily stored in the −4°C.

Sample DNA Preparation

The DNA of the NKTCL biopsy specimens was extracted by using a Qiagen blood and tissue kit according to the manufacture's protocol (Qiagen, Hilden, Germany). A NanoDrop spectrophotometer (Thermo Fish Scientific, Wilmington, DE) was used to determine the concentration of the DNA samples. Nondegraded and purified DNA with A260/A280 ratio between 1.8 and 2.0 was applied to the subsequent experiments.

Complete Workflow for Sequencing of EBV Genomes

The workflow, including library preparation, target capture, next‐generation sequencing, de novo assembly, and scaffolding and joining contigs to subsequent analysis, is accounted for in detail below. A flow chart for this process is depicted in supplemental online Figure 1.

Library Preparation

The initial amount of DNA of each sample was 3 μg, following DNA shearing process performed by Covaris S2 (Covaris, Woburn, MA). Consequently, the total DNA was divided into about 150 base pair (bp). Then, these obtained DNA fragments were used to undergo end repair, addition of the adenine‐tailing, and adaptor ligation reaction based on the standard protocol steps. Finally, the fragments were enriched by the polymerase chain reaction (PCR) to ensure the sufficient volume. The captured DNA library should be quantified using NanoDrop 2000 and verified in 1% agarose gel electrophoresis. The results show that there is a regional band in the gel but not a single band. The final construed DNA fragments were tested around 300–500 bp.

Target Regions Capture

The biotinylated single‐strand DNA capture probe, developed independently by the MyGenostics (MyGenostics, Beijing, China), was applied to hybridized with the built DNA library in light of the prescriptive protocol. The probe contains six subtype EBV genomes as the reference sequences, which can test almost all of the known virus types, as well as the classification of the various mutations at the same time. The uncombined parts of the products were washed, thereby amplifying the target DNA fragments. Finally, the corresponding washing buffer was used to remove the combined parts, which were applied to the follow‐up sequencing.

Next‐Generation Sequencing and Bioinformation Analysis

The sequence results were gained with the pair‐end sequencing reactions by the Illumina HiSeq 2500 sequencer (Illumina Inc., San Diego, CA). All untrimmed paired‐end reads with fragment length of about 150 bp were subjected to a filtering process to remove the containing adaptor, low quality, and short sequences (<40 bp), using the Cutadapt program and Trim Galore program, respectively. These reads covering human genomes and the EBV genomes are called clean reads. Then we aligned the clean reads to the human (National Center for Biotechnology Information [NCBI] build 37, HG19) and EBV reference genomes, and clipped the parts that are uniquely mapped to the human reference genome (hg19). After that, the remaining parts are known as EBV clean reads, which are the raw EBV sequences collected to assemble relative genomes.

De Novo Assembly of EBV Genomes

All high‐quality EBV clean reads were assembled using the Velvet 1.2.10, with the optimized settings for each sample using Kmer length of 59∼73 and minimum Kmer coverage of 35–70, and its regular parameters are ‐ins_length 180 and ‐exp_cov auto. The reads were merged the overlapping regions based on de Bruijn graph algorithm to generate contigs. The contigs shorter than 200 bp were filtered out. Then, we realigned the contigs to NCBI nucleotide databases aiming at removing the contigs mapped to the human genome and linked them into scaffolds; subsequently, the constructed scaffolds were arranged based on the EBV GD1 reference genomes.

PCR Amplification of EBV Fragments

The gaps between scaffolds, which was mostly located in the repeat regions do make the sequence difficult to assemble. Internal repeats, terminal repeats, and a family of repeats were excluded in PCR. First, in the nonrepetitive genome region, PCR primers were designed based on breakpoints or reference genomes. Second, we used HotStarTaq (Qiagen) to perform the reaction, amplifying the missing regions and verifying the ambiguous sites. 100 ng of the tumor DNA was performed using 50 μl reaction mixture, containing 1 μl each of 10 uM forward and reverse primers, 1 μl of 10 uM deoxynucleotide triphosphate, and 0.3 μl HotStarTaq enzyme, and the rest is replenished by ddH₂O (if the products contained high G + C%, 10 μl Q buffer was needed to add into the mixture). The setting of the automated thermal cycler was denaturation (94°C or 45 seconds), annealing (56°C for 45 seconds), and extension (72°C for 2–6 min), where the optimal cycle number and extension time was adjusted according to the length of fragment. Finally, we use Sanger sequencing to detection the exact bases. If the above‐mentioned ways do not get the base information, we can supply the gaps with “Ns.” Here, eight EBV genomes derived from NKTCL biopsy were successfully retrieved, designated NKTCL‐EBV 1–8.

Detection of SNVs and Mutation Analysis

All output EBV reads were first aligned to the EBV reference genome by Burrows‐Wheeler Aligner software, removing the interferential duplicates by Sequence Alignment/Map tools (SAMtools) 3 to gain the basic classified subtype, applying the following criteria: (a) the average sequencing depth is not less than 10; (b) the faction of EBV covered with at least 4 times is more than 50%. Variations, including single nucleotide variants (SNVs) and insertion and deletions (INDEL), were generated from the comparison of the EBV‐DNA genomes against the reference GD1 sequence. In this study, GATK HaplotypeCaller was used to recognize the SNVs and INDEL, and GATK VariantFiltration was used to filter the raw results. After that, ANNOVAR was used to do annotation and classification. In addition, we defined a position to be homozygous if the variant frequency is =1.00 and a position to be heterozygous if the variant frequency is <1.00 (both read depth of 5 or above).

What should be noticed is that the major repeats, involving internal repeats 1–4 and terminal repeats, were not taken into account. Positions without sequence data or those marked with “Ns” were also not mentioned in the variations analysis. Furthermore, the effects caused by SNVs are nonsynonymous mutations and synonymous mutations. The nonsynonymous mutations were noted according to the nine categories of EBV‐encoded proteins with known and putative functions, as defined by Kwok et al. [13]. NKTCL was categorized as type II latency pattern (positive for latent membrane protein 1 [LMP1] in some cases and positive for EBV nuclear antigen 1 [EBNA1] in all cases [14]. Hence, these amino acids or epitopes changes, which may affect the corresponding immunological recognition processes mediated by CD4⁺ and CD8⁺ T cell, should be given the special attention.

Phylogenetic Analysis

The MAFFT software was used to multiple sequence alignments, comparing the NKTCL‐EBV genomes with the other 19 strains [10], [13], [15], [16], [17], [18], [19], [20], [21], [22], [23] reported previously. Molecular Evolutionary Genetics Analysis software, version 7, was used to perform the phylogenetic analyses by neighbor‐joining algorithm in this study. EBNA1, EBNA2, LMP1, and LMP2A genes were selected. Bootstrap analysis of 1,000 replicates was performed on each tree to determine the confidence.

Results

Summary of the Sequencing Data

In this study, we performed the EBV genome capture and deep sequencing (average depth >10) in eight NKTCL samples. A total of 1,083,561 75‐base pair end reads were generated, which were equivalent to 21.48 MB in the whole samples. Moreover, aiming at knowing the detail of subtype, the obtained DNA sequences were compared with six mostly referenced sequences, including AG878, B95‐8, EBV wild type, GD1, GD2, and HKNPC1. The GD1 coverage percentages are higher than the rest, ranging from 86.63% in NKTCL‐EBV5 to 94.94% in NKTCL‐EBV8. The mean depth of the eight cases was 107‐fold, varying from 50.29–161.41‐fold. The fraction of effective bases on EBV, which is mapped to reference GD1‐EBV, varied from 16.05% (31.53/418.38 Mb, NKTCL‐EBV2) to 72.77% (340.12/467.39, NKTCL‐EBV6). Details of the reads statistics and subtype results are listed in online supplemental Table 2.

Assembly of NTKCL‐EBV1–8 Genomes

The filtered reads were utilized to perform de novo assembly, generating scaffoldings using Velvet. The number of the contigs were from 20 (NKTCL‐EBV5) to 32 (NKTCL‐EBV7). N50 size of contigs ranged from 8,342 bp (NKTCL‐EBV2) to 21,323 bp (NKTCL‐EBV8). The longest contigs were about 43 kb in length for the whole samples. The summary of these contigs are provided in supplemental online Table 3. As illustrated above, the gaps would be supplemented with the tracts “N” based on the EBV reference GD1 if the PCR and Sanger sequencing did not link up the contigs effectively. Finally, eight EBV genomes were successfully obtained, named NKTCL‐EBV1 to NKTCL‐EBV8. The genome sizes estimated based on the reference EBV sequence were as follows: NKTCL‐EBV1 (172,059 bp), NKTCL‐EBV2 (171,730 bp), NKTCL‐EBV3 (171,639 bp), NKTCL‐EBV4 (171,613 bp), NKTCL‐EBV5 (171,663 bp), NKTCL‐EBV6 (171,727 bp), NKTCL‐EBV7 (171,706 bp), and NKTCL‐EBV8 (171,590 bp), with guanine and cytosine (GC) contents of approximately 55%.

Mutation Analysis of the NKTCL‐EBV Genomes

Whole‐genome sequence alignments revealed extensive nucleotide variation in all of the eight NKTCL‐EBV genomes. In comparison with the most similar reference EBV genome GD1, the NKTCL‐EBV1 to 8 harbored 2,072 variations in total, including 1,938 substitutions, 58 insertions, and 76 deletions. Among them, 1,218 substitutions, 26 deletions, and 15 insertions were located in the coding regions, whereas the 720 substitutions, 43 insertions, and 50 deletions were located in the noncoding regions. The summary of the mutation data is shown in the supplemental online Table 4. The number of the variations and the corresponding variability, which represents the variations account for the proportion in the total number of bases of the NKTCL‐EBV genomes, were as follows: NKTCL‐EBV1 (323, 0.188%), NKTCL‐EBV2 (446, 0.260%), NKTCL‐EBV3 (267, 0.156%), NKTCL‐EBV4 (514, 0.299%), NKTCL‐EBV5 (472, 0.275%), NKTCL‐EBV6 (479, 0.279%), NKTCL‐EBV7 (445,0.260%), and NKTCL‐EBV8 (423, 0.247%). Figure 1 depicts the variation of all available NKTCL‐EBV genomes compared with the reference GD1 sequence. Remarkably, the density of variations was substantially higher when the genomes were compared with B95‐8, the prototype of EBV genome originally isolated from a patient with infectious mononucleosis from America (supplemental online Fig. 2).

Figure 1.

Genetic variations of Epstein‐Barr virus (EBV) genome in all of the natural killer cell/T‐cell lymphoma (NKTCL)‐EBVs compared with GD1. Tracks from outer to inner represent NKT‐EBV1 to 8 genomes. Mutation internal repeats and terminal repeats are disregarded. The outmost circle shows the positive‐strand ORFs (orange), repeat regions (brown), and negative‐strand ORFs (blue) in the reference GD1 strain (AY961628.3). The red, green, and blue points in the inner circles show the distributions of SNVs, deletions, and insertions, respectively. The size of the point represents the length of the deletion or insertion.

Abbreviations: ORF, open reading frame; SNV, single nucleotide variation.

The nine categories, discriminated and defined by the different protein functions [24], are summarized in the supplemental online Table 5. In total, the mutation number of genes coding latent and tegument were higher than any other genes, on average accounting for 56.2% of the whole nonsynonymous mutations (419 in latent, 345 in tegument, respectively), followed by 146 mutations (accounting for 10.7%) located in genes expressing membrane (glyco) protein. There were 44–68 nonsynonymous mutations located in the latent genes in NKTCL‐EBV1, 2, 4, 5, 6, 7, and 8, whereas NKTCL‐EBV3 had a minimum quantity of 20 mutations. These latent gene mutations accounted for 43.3% (NKTCL‐EBV3) to 60.6% (NKTCL‐EBV5) of all nonsynonymous mutation detected for each genome. Given the genes encoding tegument proteins, the statistics of mutations are in stark contrast between the NKTCL‐EBV1, 3, 4, and 8 (the corresponding numbers are 27, 25, 29, 28) and NKTCL‐EBV2, 5, 6, and 7 (the numbers are 61, 59, 59, 57, respectively). Genes encoding tegument proteins contained 25 (24.0%, NKTCL‐EBV3) to 61 (35.1%, NKTCL‐EBV2). When observing the mutations in each case, we detected the maximum quantity in NKTCL‐EBV6, which harbored 194 nonsynonymous mutations, followed by the NKTCL‐EBV5(188), NKTCL‐EBV4(184), NKTCL‐EBV7(183), NKTCL‐EBV8(183), NKTCL‐EBV2(174), NKTCL‐EBV1(149), and NKTCL‐EBV3(104) (Fig. 2).

Figure 2.

Number of nonsynonymous mutations including the nine categories of EBV‐encoded proteins. The majority of the amino acid changes are located in the latent proteins (blue) and tegument proteins (red) in the total of the NKTCL‐EBVs.

Abbreviations: EBV, Epstein‐Barr virus; NKTCL, natural killer cell/T‐cell lymphoma.

Sequence Variations in the EBNA1 and LMP1 Genes

The total of 467 nonsynonymous mutations was located in the protein‐coding genes, accounting for 38.3% of the SNVs. The EBNA1 and LMP1 genes were summarized in this study because these two regions are the most frequently studied regions to date.

EBNA1 is the only latent protein that is consistently expressed in all EBV‐positive malignancies and is a transcriptional activator of the EBV expression. EBNA1 have been defined as P (prototype, B95‐8) and V (variant) EBNA1, which differed from B95‐8 by 15 amino acid (AA) substitutions [25]. P and V subtypes were further classified based on the signature changes at AA residue 487 in the C terminus of EBNA1: P‐alanine (P‐ala), P‐threonine, V‐proline, V‐leucine, and V‐Valine (V‐val) [26], [27]. Compared with EBNA1 sequence of EBV wild type, 19 nonsynonymous mutations were observed in total: 6 in the N‐terminal regions and 13 in the C‐terminal regions. All of the eight strains of NKTCL‐EBV were arranged into V‐val subtype. As Do et al. [28] and Wang et al. [27] had mentioned that the subvariants of V‐val have AA substitution at 528 and 585, we observed that EBNA1 strain of NKTCL‐EBV 2 was identified as a V‐val subvariant.

LMP1 is the major transforming protein of EBV and contains a higher degree of polymorphism than most EBV genes [29]. Compared with the EBV prototypic type (B95‐8), 45 nonsynonymous mutations were detected in the LMP1 strains of the eight specimens: 4 in the N terminus (AA: 1–25), 26 in the transmembrane domains (AA:26–196), and 15 in the C terminus (AA: 197–386). Using the Edwards classification system, based on the signature AA changes relative to the prototypic LMP‐1 (B95‐8) in the C terminus (AA 189–377), six sequence variants of LMP1, termed Alaskan, China 1, China 2, Mediterranean+ (Med+), Med‐, and North Carolina (NC), have been identified [30]. All eight NKTCL‐EBV strains were identified as China 1 type [31]. The other two commonly classification system used to date characterized variants LMP‐1 gene polymorphisms include a 30‐bp deletion in the C terminus and the loss of restriction site Xho I in the N terminus of the gene. The LMP1 strain in NKTCL‐EBV1–7, but not NKTCL‐EBV8, harbored the 30‐bp deletion, causing the loss of 10 AA (343–352). In addition, all eight NKTCL‐EBV strains had Xho I restriction site loss at exon 1 of the LMP1 gene as B95‐8 (169423–169428; GAGCTC→GATCTC).

Amino Acid Changes in CD4+ and CD8+ T‐Cell Epitopes of EBNA1, LMP1, and LMP2A

Almost all of the latent proteins and lytic proteins harbored nonsynonymous mutations in epitopes induced by both the CD4⁺ and CD8⁺ T cells responses. Many epitopes were defined and were mapped in EBV antigens and correlated with major histocompatibility complex type in previous studies. Compared with B95‐8, amino acids changes were found in 3 CD8+ epitopes of EBNA1, 8 epitopes of LMP1, and 12 epitopes of LMP2A. Eleven CD4+ epitopes of EBNA1, 13 in LMP1, and 9 in LMP2A contained amino acids. Some of the nonsynonymous mutations were affecting multiple epitopes. For instance, a C‐to‐T substitution at coordinate 97121 (NC_007605) resulted in the change of residue 487 (A → V) of EBNA1 in all of the eight NKTCL specimens, where CD4+ epitopes SNP, NPK, ENI, IAE, and LRA were located. The C‐to‐T and C‐to‐G substitution at coordinate 168229 induced LMP1 epitope changes in residue 212 G → S of NKTCL‐EBV1–6 and 8 and in residue 212 G → R in NKTCL‐EBV7. Meanwhile, in the NKTCL‐EBV2 strains, another two substitutions, C‐to‐G at coordinate 168223 and A‐to‐C at coordinate 168224, caused the epitope changes in residue 213 H → Q and 214 E → Q, respectively, where CD4+ epitopes QAT, SSH, and SGH were located. The positions of nonsynonymous mutation changes located in the epitopes are illustrated in Figure 3 and supplemental online Table 6.

Figure 3.

Amino acid changes in CD8+ and CD4+ T‐cell epitopes in EBNA1, LMP1, and LMP2A. The total changes of epitopes in the natural killer cell/T‐cell lymphoma (NKTCL) 1–8 are marked with hallow (CD8+) and solid arrows (CD4+), respectively.

Abbreviations: AA, amino acid; EBNA, Epstein‐Barr virus‐determined nuclear antigen; LMP, latent membrane protein.

Phylogenetic Analysis of the NKTCL‐EBV Genomes and EBNA1 Interstrain Recombinant

The phylogenetic trees were conducted based on alignment of edited full‐length of eight NKTCL‐EBVs and previously published 28 strains, as illustrated in Figure 4A. Of note, all NKTCL‐EBVs genomes clearly sort into type 1, based on differences in whole genome and especially EBNA2 (supplemental online Fig. 3). Principal component analysis, as well as whole genomes analysis, supported that the type 1/type 2 classification is quantitatively the greatest variation in EBV strains worldwide. The phylogenetic analysis from the whole genome showed that all the NKTCL‐EBVs are related to other Asian EBV strains, including EBV‐associated GC (EBVaGC) 1–9, HKNPC1–9, C666‐1, GD1, and GD2 obtained from China, and Akata from Japan, whereas none of the specimens were clustered in a branch of non‐Asian strains AG876, B95‐8, Mutu, K4413‐Mi, and K4123‐Mi. Phylogenetic analysis of EBNA1 (Fig. 4B) and LMP1 (Fig. 4C) showed similar phylogenetic clustering, as we found using full‐length whole EBV genomes, albeit with minor variations. Of interest, EBNA1 of NKTCL‐EBV3 sequence showed clustered away from the other seven NKTCL‐EBV strains. Analysis of amino acid sequences of EBNA1 supported that EBNA1 of NKTCL‐EBV3 may arise from recombination of GD1 and B95‐8 (Fig. 5, supplemental online Table 7).

Figure 4.

Phylogenetic trees of the whole EBV genomes and nucleotide sequences of EBNA1 and LMP1 sequences. The evolutionary history was inferred using the neighbor‐joining method, and evolutionary analyses were conducted in MEGA7. The percentage of replicate trees in which the associated taxa clustered together in the bootstrap test (1,000 replicates) is shown next to the branches. NKTCL‐EBV1–8 in the phylogenetic trees are marked with an asterisk.

Abbreviations: EBV, Epstein‐Barr virus; EBNA, EBV nuclear antigen; LMP, latent membrane protein; NKTCL, natural killer cell/T‐cell lymphoma.

Figure 5.

Interstrain recombination analysis in the EBV‐determined nuclear antigen (EBNA) 1 gene of NKTCL‐EBV3. The proteins of EBNA1 that are identical to B95‐8 are in normal text, and those identical to GD1 are in bold.

Abbreviations: EBV, Epstein‐Barr virus; NKTCL, natural killer cell/T‐cell lymphoma.

Discussion

To date, pathogenesis and genotype analyses of NKTCL have focused on genetic variations in a small fraction of EBV genes, which is limited to define the spectrum of diversity within the whole genome of EBV. The genome‐wide characteristics of EBV are essential to understand the diversity of strains isolated from NKTCL. Here, we directly sequenced EBV‐captured DNA from eight primary NKTCL biopsy samples from China using Illumina HiSeq 2500 sequencer platform and presented the eight EBV sequences obtained from primary NKTCL tumors.

The variability of NKTCL‐EBV1–8 ranged from 0.156% (NKTCL‐EBV3) to 0.299% (NKTCL‐EBV4), which is lower than the interstrain variability of ∼0.5% for virus of the same type. Hence, it excludes the variability raised by coinfection of different viral strains [11], [13]. In accordance with the previous reports [16], the number of the nonsynonymous mutations is highest in the gene regions encoding latent proteins in each of the NKTCL‐EBV genomes, followed by genes encoding the tegument protein and membrane glycoproteins. One of the basic goals in EBV‐associated study is aimed at understanding EBV geographic and disease associations. The phylogenetic tree constructed based on whole genome alignment of all strains illustrated that the genomes are actually closer to the other Asian variants than non‐Asian strains, indicating that a phylogenetic relationship corresponding to the geographical origin of the viral genomes instead of the type 1 and 2 dichotomy. In order to investigate how genomic variations of EBV contribute to pathogenesis of NPC, it is essential to sequence the EBV genomes of a large number of normal and diseased individuals in the same geographical region and conduct a case‐control comparison.

EBNA1 and LMP1 are the most frequently studied regions to date. EBNA1 is critical in maintaining EBV in the infected cells and facilitating episomal replication, and it is also a transcriptional activator. The codon 487 is a hot spot variation in the C terminus, classifying the P and V subtypes based on the AA present at residue 487. The EBNA1 strain in each NKTCL‐EBV case was the V‐val category. Interestingly, V‐val was detected in both cases and controls almost exclusively in China according to previously reported studies. Chang et al. [32] had summarized the prevalence of EBNA1 variants by clinical condition and geographic region. They showed that 95.2% of the NPC and 100% of the Hodgkin's lymphoma in China are V‐val, whereas in the healthy or nonmalignant people, about 48.6% is the V‐val subtype, and 32.4% are V‐val and P‐ala subtypes. LMP1 is thought to play a principal role in tumorigenesis. Based on the signature AA changes relative to the prototypic LMP‐1 in the C terminus (AA 189–377), LMP1 variants have previously been classified into seven groups named NC, Med, Alaskan, B95‐8, China 1, China 2, and China 3. All of the NKTCL‐EBV were classified China 1, consistent with results obtained from the lung cancers, HKNPCs, and EBVaGCs from China. This showed a phylogenetic relationship corresponding to the geographical origin of the viral genomes, supporting the conclusion that LMP1 can serve as a geographical marker [31]. Moreover, the 30‐bp deletion, resulting in the proteins loss between the AA 343 and AA 352, had been detected in the NKTCL‐EBV1–7, a marked predominance of LMP1 deletion (del‐LMP1) over wild‐type LMP1 variants was observed in NKTCL as previously reported [33]. Recently, Cabrera et al. [34] suggested that EBV with Xho I restriction site loss at exon 1 of the LMP1 gene was associated with the risk of NKTCL in Chile. And we found that all the eight NKTCL specimens were Xho I loss; although the available data suggest the Xho I loss in NPC and HL is common in China, additional studies are needed to fully evaluate geography and disease of Xho I loss in the NKTCL.

NKTCL is associated with type II EBV latency, in which only restricted EBV antigens, namely EBNA1, and LMP 1 and 2, are expressed [35], [36]. These viral proteins offer the potential targets for a T cell‐based immunotherapy. Adoptive transfer of cytotoxic T cells (CTLs) specific for EBV antigens has proved safe and effective as prophylaxis and treatment for EBV‐associated lymphoproliferative disease. Some patients with advanced stage or relapsed EBV‐associated malignancies, including NKTCL, achieved complete remission after treatment with autologous LMP1/2‐ and EBNA1‐specific cytotoxic T lymphocytes (CTLs) or activated by peptides derived from LMP1/2 [37], [38]. Nonetheless, some cases still did not respond to LMP‐CTL therapy, and this failure was usually attributed to immune escapes by antigen loss. It is worth noting that all these previous studies used prototype EBV sequence, B95‐8, to design full‐length LMP epitopes. Therefore, our work gives an alternative explanation for the lack of tumor response. As shown in Figure 3, sequence analysis identified LMP and EBNA1 variations containing epitope changes in comparing to the prototype B95–8 sequence in all cases of NKTCLs. Highly polymorphic LMP and EBNA1 genes lead to amino acid changes in CD4+ and CD8+ specific T‐cell epitopes. As an example, YFL, a human leukocyte antigen A2‐restricted epitope of LMP1, was a variant peptide of the wild‐type YLL. The changes in residue 2(L → F) and residue 5(M → I) were found in all of the NKTCL‐EBV genomes, which are also found prevalently in NPC specimens [16]. The CTL recognition of YFL was abrogated compared with the wild‐type epitopes [39]. Thus, this LMP1 variant epitopes might contribute to the evasion of the EBV‐infected cells from T‐cell surveillance [16]. Whether changes in such epitopes confer immune evasion of the NKTCL cells may constitute another hypothesis for future testing. Our data provide optimization proposal for selecting EBV genome for treatment from individual patients or at least predominant strains prevalent in geographical regions instead of commonly used B95‐8 genome.

Conclusion

We reported eight EBV genomes isolated from primary NKTCL biopsy specimens based on a complete large‐scale sequencing workflow. To the best of our knowledge, this is the first report analyzing the sequence diversity on a whole‐genome level of EBV obtained from primary NKTCL tumors, although their pathogenesis remains to be clarified. Furthermore, phylogenetic analysis showed that all NKTCL‐EBV strains isolated in extranodal NKTL are close to other Asian subtypes, leading to the conclusion that EBV infections are more likely affected by different geographic regions rather than particular EBV‐associated malignancies. Therefore, our data have implications for the development of effective prophylactic and therapeutic vaccine approaches targeting the personalized EBV antigens in these aggressive diseases.

See http://www.TheOncologist.com for supplemental material available online.

Author Contributions

Conception/design: Jun Zhu, Zhemming Lu

Provision of study material or patients: Ningjing Lin, Yuqin Song, Jun Zhu

Collection and/or assembly of data: Ningjing Lin, Wenjing Ku

Data analysis and interpretation: Ningjing Lin, Wenjing Ku

Manuscript writing: Ningjing Lin, Wenjing Ku, Zheming Lu

Final approval of manuscript: Ningjing Lin, Wenjing Ku, Yuqin Song, Jun Zhu, Zheming Lu

Disclosures

The authors indicated no financial relationships.