Patterns of Interindividual Variability in the Antibody Repertoire Targeting Proteins Across the Epstein-Barr Virus Proteome

Patterns of Interindividual Variability in the Antibody Repertoire Targeting Proteins Across the... Abstract Background Little is known about variation in antibody responses targeting the full spectrum of Epstein-Barr virus (EBV) proteins and how such patterns inform disease risk. Methods We used a microarray to measure immunoglobulin G (IgG) and immunoglobulin A (IgA) antibody responses against 199 EBV protein sequences from 5 EBV strains recovered from 289 healthy adults from Taiwan. We described positivity patterns, estimated the correlation between antibodies, and investigated the associations between environmental and genetic risk factors and variations in antibody responses. Results Healthy adults were more likely to mount IgG antibody responses to EBV proteins (median positivity frequency, 46.5% for IgG and 17.3% for IgA; P = 1.6 × 10–46, by the Wilcoxon rank sum test). Responses against glycoproteins were particularly prevalent. The correlations between antibody responses of the same class were higher than correlations across classes. The mucosal exposure to proteins involved in EBV reactivation (as determined by the IgA response) was associated with smoking (P = .002, by the sequence kernel association test–combined), and approximately one quarter of adults displayed antibody responses associated with EBV-related cancer risk. Conclusions These data comprehensively define the variability in human IgG and IgA antibody responses to the EBV proteome. Patterns observed can serve as the foundation for elucidating which individuals are at highest risk of EBV-associated clinical conditions and for identifying targets for effective immunodiagnostic tests. Antibody microarray, antibody response, Epstein-Barr virus, Taiwan, IgA, IgG Epstein-Barr virus (EBV) is a ubiquitous herpesvirus that infects >90% of individuals worldwide [1]. Primary infection with the virus typically occurs in early childhood [2, 3], after which EBV establishes lifelong latency in human B cells. During the course of lifelong infection, EBV periodically reactivates in B cells or epithelial cells, expressing a broader set of lytic proteins required for viral replication [4]. This ongoing replication in epithelial cells of the oral cavity contributes to shedding of the virus into saliva, facilitating person-to-person transmission of EBV. Although established EBV infection is generally asymptomatic, infection manifests as clinical disease in a subset of infected individuals; such diseases include hematopoietic malignancies (eg, Burkitt and Hodgkin lymphomas) and epithelial malignancies (eg, gastric cancer and nasopharyngeal carcinoma [NPC]) [4]. EBV is etiologically involved in approximately 200000 cancers worldwide each year [5]. Despite its ubiquity and association with various chronic conditions, there are currently no proven ways to prevent EBV infection from occurring or to predict with sufficient accuracy which infected individuals will develop serious chronic diseases associated with this virus [5, 6]. One way to monitor patterns of EBV exposure in the general population is to measure levels of circulating anti-EBV antibodies that reflect responses to viral proteins. For example, immunoglobulin G (IgG) antibodies appear within several weeks after primary infection and remain in circulation for many years [7]. Prior studies aimed at defining rates of EBV infection by age across populations have used positivity for IgG antibodies targeting EBV proteins such as the viral capsid antigen (VCA) and EBV nuclear antigen 1 (EBNA1) [8] to classify individuals as EBV positive (ie, as ever exposed to EBV). IgG antibodies can also be used to better understand the nature of an individual’s EBV infection. For example, a subset of long-term EBV carriers displays elevated levels of IgG antibody targeting the early antigen protein (EA) [9]. This EA complex is expressed as part of the EBV lytic cycle; therefore, elevated levels of EA indicate ongoing viral lytic activity (reactivation) and a potential lack of control over the virus [10]. In contrast to IgG, immunoglobulin A (IgA) responses to viral exposure are believed to reflect recent reactivation at mucosal sites (eg, the oral epithelium) [11]. This type of immune response is of interest because of the oral route of EBV transmission. During primary infection, IgA responses against EBV in the oral cavity can be strong but are short lived [12]. The durability and spectrum of EBV-directed IgA responses in healthy adults are largely uncharacterized. However, it has been shown that individuals at greatest risk of developing NPC, an EBV-associated cancer, have elevated anti-EBV IgA responses to VCA and EBNA1 years before a cancer diagnosis [8, 13, 14]. Although informative, prior studies of humoral responses to EBV infection [15, 16] have largely focused on measuring levels of IgG and IgA antibodies to a handful (<10) of the approximately 90 EBV protein targets [8]. With the advent of protein array systems capable of screening for antibody responses against the full complement of EBV proteins [17–19], it is now feasible to evaluate humoral responses to EBV infection in a more comprehensive manner. Thus, we evaluated IgG and IgA anti-EBV antibody responses against 199 predicted sequences and 3 synthetic antigens from 86 EBV proteins involved in various stages of the EBV life cycle (eg, latent and lytic cycles) among 289 healthy individuals from Taiwan, a region with high rates of EBV-associated cancer. METHODS A total of 289 healthy, asymptomatic adults who served as controls for 2 previously published Taiwanese studies were selected for this report [20, 21]. Details are described in the Supplementary Materials. A previously described EBV protein microarray targeting IgG and IgA antibodies against 199 predicted EBV protein sequences (86 EBV proteins) from 5 EBV strains (AG876, Akata, B95-8, Mutu, and Raji) was used [17, 18, 22, 23]. Briefly, these 199 sequences represent nonredundant open reading frames from 5 EBV strains representing both African (Mutu, Raji, and AG876) and Asian (Akata) variants, as well as known splice variants identified in the literature. Eighty-five percent of 199 predicted sequences represented complete transcripts of EBV genes. The remaining 30 predicted sequences represented linear segments from 8 EBV genes >1000 bp long. Each of the protein sequences cloned into the pXT7 expression vector prior to printing onto the microarray included N-terminal 10x histidine (His) and C-terminal hemagglutinin (HA) tags to confirm expression on the microarray. After printing, anti-polyHis probing confirmed the presence of 99.5% of sequences; anti-HA probes confirmed the presence of 98.5% of sequences. High coverage was achieved across the 5 EBV strains, with 97% of the predicted sequences from each strain represented on the microarray at ≥99% homology. A positive antibody response in a healthy adult was defined as a response greater than that individual’s person-specific background (mean of 4 no-DNA-control spots plus 1.5 standard deviations) [17]. As previously reported [23], we included blinded quality control replicates during testing and observed good reproducibility for antibodies measured using this custom protein microarray. The average percentage agreement for classifying an individual’s antibody response as positive versus negative was high for both IgA (84%; range, 52%–100%) and IgG (79%; range, 48%–100%), respectively. The antibody level (determined on the basis of the standardized signal intensity output from the array) was further grouped into quartiles (4 categories) based on the population distribution of the response for each given antibody. Based on coefficients of variation reported from the previous study [23], we excluded from analysis the 2 IgG array spots out of 199 evaluated that had coefficients of variation of >20% (BZLF1/IgG [array protein sequence CAA24861.1-103155-102655] and BXRF1/IgG [array protein sequence CAA24798.1-144860-145606]). Finally, we included 3 synthetic EBV peptides that are putative cancer biomarkers (VCAp18, EBNA1, and EAdp47) on the microarray, bringing the total number of anti-EBV antibodies for analysis to 402 (200 IgG and 202 IgA antibodies). Ethics Statement All human subjects included were adults, and all studies were approved by appropriate human subjects committees in Taiwan and the United States. This study was reviewed/approved by the National Cancer Institute Special Studies Institutional Review Board and the National Taiwan University Institutional Review Board)\. Written informed consent was obtained for all participants. Statistical Analysis Visualization and statistical analyses were conducted using R statistical software. Details on the statistical analyses are described in the Supplementary Materials. Briefly, to pool data across the 2 study populations, we used population-specific cut points to generate similar underlying antibody distributions and then compared individual response rates (ie, percentage positivity values) across immunoglobulin class and EBV life cycle. We estimated the correlation between antibodies across immunoglobulin classes (ie, IgG against protein X versus IgA against protein X) and within each immunoglobulin class (ie, IgG against protein X versus IgG against protein Y), using Spearman correlation coefficients. All individuals were clustered according to the positive responses across the full EBV proteome, using unsupervised hierarchical clustering with binary distance and complete linkage [24]. To examine the association between environmental and genetic risk factors (ie, smoking, alcohol drinking, and HLA alleles) and biological groupings of antibody response (eg, responses directed against a select EBV life cycle), we used the sequence kernel association test–combined (SKAT-C) from the single-nucleotide polymorphism set [25], using binary variables (with a positive antibody response assigned a value of 1 and a negative response assigned a value of 0). RESULTS Anti-EBV IgG and IgA Antibody Responses Mounted by Healthy Individuals All individuals had measurable (ie, positive) antibody responses against EBV (Figures 1A–C; Supplementary Figures 1–10). The individual with the narrowest pattern responded to 38 EBV targets on the array (28 IgG antibodies and 10 IgA antibodies), representing 21 EBV proteins (Figure 1A). The individual with the broadest response had a positive response against 329 EBV targets on the array (163 IgG antibodies and 166 IgA antibodies), representing 84 of 86 EBV proteins evaluated. Positive IgG (ie, long-term) responses against EBV proteins were markedly more frequent than positive IgA (ie, more-recent) responses (Table 1). On average, individuals mounted IgG antibody responses against 93 (46.5%) of 200 EBV targets on the array and IgA antibody responses against 35 (17.3%) of 202 EBV targets. Figure 1. View largeDownload slide Anti–Epstein-Barr virus (EBV) immunoglobulin G (IgG) and immunoglobulin A (IgA) antibody responses among healthy individuals in Taiwan. A, Number of positive antibody responses in 289 healthy individuals. Blue indicates IgG antibody positivity, and red indicates IgA positivity. The individual with the narrowest response had antibodies against 38 EBV targets on the array (28 IgG antibodies and 10 IgA antibodies), representing 21 EBV proteins. In contrast, the individual with the broadest response had antibodies against 329 EBV targets (163 IgG antibodies and 166 IgA antibodies), representing 84 of 86 tested EBV proteins. B and C, The 20 IgG (B) and IgA (C) antibodies with the highest antibody positivity rates and the 20 with the lowest antibody positivity rates. Percentages indicate the antibody positivity rate for the given target antigen in this population. Figure 1. View largeDownload slide Anti–Epstein-Barr virus (EBV) immunoglobulin G (IgG) and immunoglobulin A (IgA) antibody responses among healthy individuals in Taiwan. A, Number of positive antibody responses in 289 healthy individuals. Blue indicates IgG antibody positivity, and red indicates IgA positivity. The individual with the narrowest response had antibodies against 38 EBV targets on the array (28 IgG antibodies and 10 IgA antibodies), representing 21 EBV proteins. In contrast, the individual with the broadest response had antibodies against 329 EBV targets (163 IgG antibodies and 166 IgA antibodies), representing 84 of 86 tested EBV proteins. B and C, The 20 IgG (B) and IgA (C) antibodies with the highest antibody positivity rates and the 20 with the lowest antibody positivity rates. Percentages indicate the antibody positivity rate for the given target antigen in this population. Table 1. Number of Positive Antibody Responses to Epstein-Barr Virus (EBV) Proteins Among Healthy Individuals in Taiwan, by Antibody Type and Stage of EBV Life Cycle Targeted Antibody, Target Protein Antibodies, No. Median (%a) IQR Range IgG  Overall 200 93 (46.5) 67–132 22–188  Latent 58 30 (51.7) 21–40 21–58  Immediate early lytic 11 4 (36.4) 2–7 0–11  Early lytic 44 20 (45.5) 13–28 5–42  Late lytic 47 20 (42.6) 14–30 7–45  Glycoprotein 19 13 (68.4) 11–16 4–19  Other/unknown 21 7 (33.3) 4–12 0–20 IgA  Overall 202 35 (17.3) 18–85 4–166  Latent 58 10 (17.2) 4–20 0–48  Immediate early lytic 12 0 (0.0) 0–1 0–9  Early lytic 45 7 (15.6) 3–15 0–35  Late lytic 47 8 (17.0) 6–14 0–39  Glycoprotein 19 8 (42.1) 3–15 0–19  Other/unknown 21 1 (4.8) 0–6 0–17 Antibody, Target Protein Antibodies, No. Median (%a) IQR Range IgG  Overall 200 93 (46.5) 67–132 22–188  Latent 58 30 (51.7) 21–40 21–58  Immediate early lytic 11 4 (36.4) 2–7 0–11  Early lytic 44 20 (45.5) 13–28 5–42  Late lytic 47 20 (42.6) 14–30 7–45  Glycoprotein 19 13 (68.4) 11–16 4–19  Other/unknown 21 7 (33.3) 4–12 0–20 IgA  Overall 202 35 (17.3) 18–85 4–166  Latent 58 10 (17.2) 4–20 0–48  Immediate early lytic 12 0 (0.0) 0–1 0–9  Early lytic 45 7 (15.6) 3–15 0–35  Late lytic 47 8 (17.0) 6–14 0–39  Glycoprotein 19 8 (42.1) 3–15 0–19  Other/unknown 21 1 (4.8) 0–6 0–17 Abbreviations: IgA, immunoglobulin A; IgG, immunoglobulin G; IQR, interquartile range. aCalculated on the basis of the median number of positive antibodies among the total number of antibodies evaluated in each group. View Large Table 1. Number of Positive Antibody Responses to Epstein-Barr Virus (EBV) Proteins Among Healthy Individuals in Taiwan, by Antibody Type and Stage of EBV Life Cycle Targeted Antibody, Target Protein Antibodies, No. Median (%a) IQR Range IgG  Overall 200 93 (46.5) 67–132 22–188  Latent 58 30 (51.7) 21–40 21–58  Immediate early lytic 11 4 (36.4) 2–7 0–11  Early lytic 44 20 (45.5) 13–28 5–42  Late lytic 47 20 (42.6) 14–30 7–45  Glycoprotein 19 13 (68.4) 11–16 4–19  Other/unknown 21 7 (33.3) 4–12 0–20 IgA  Overall 202 35 (17.3) 18–85 4–166  Latent 58 10 (17.2) 4–20 0–48  Immediate early lytic 12 0 (0.0) 0–1 0–9  Early lytic 45 7 (15.6) 3–15 0–35  Late lytic 47 8 (17.0) 6–14 0–39  Glycoprotein 19 8 (42.1) 3–15 0–19  Other/unknown 21 1 (4.8) 0–6 0–17 Antibody, Target Protein Antibodies, No. Median (%a) IQR Range IgG  Overall 200 93 (46.5) 67–132 22–188  Latent 58 30 (51.7) 21–40 21–58  Immediate early lytic 11 4 (36.4) 2–7 0–11  Early lytic 44 20 (45.5) 13–28 5–42  Late lytic 47 20 (42.6) 14–30 7–45  Glycoprotein 19 13 (68.4) 11–16 4–19  Other/unknown 21 7 (33.3) 4–12 0–20 IgA  Overall 202 35 (17.3) 18–85 4–166  Latent 58 10 (17.2) 4–20 0–48  Immediate early lytic 12 0 (0.0) 0–1 0–9  Early lytic 45 7 (15.6) 3–15 0–35  Late lytic 47 8 (17.0) 6–14 0–39  Glycoprotein 19 8 (42.1) 3–15 0–19  Other/unknown 21 1 (4.8) 0–6 0–17 Abbreviations: IgA, immunoglobulin A; IgG, immunoglobulin G; IQR, interquartile range. aCalculated on the basis of the median number of positive antibodies among the total number of antibodies evaluated in each group. View Large Additional evidence of broad anti-EBV IgG reactivity was provided by the observation that almost all individuals (≥95%) in our study mounted a positive IgG response against 14 different EBV proteins, including 2 latent proteins (EBNA3B and EBNA1), 4 early lytic proteins (BILF1, BDLF4, BLRF2, and BGLF5), 4 late lytic proteins (BRRF2, BFRF3, VCA-p18, and BPFL1), and 4 glycoproteins (BZLF2, BDLF3, BBRF3, and BMRF2; Figure 1B and Supplementary Table 1). This IgG pattern stands in contrast to that of IgA, for which only 1 target on the array (representing BZLF2) elicited a uniformly positive response (Figure 1C; Supplementary Table 1), with a plurality of study participants (40.5% [117]) positive for <25 array targets. Only 8 EBV targets were weakly immunogenic at eliciting IgG responses (ie, <5% of individuals tested positive), including 1 targeting a latent protein (EBNA3C), 1 early lytic protein (EAD-p47: BMRF1), and 4 late lytic proteins (BGRF1/BDRF1, BFLF2, BRFR1A, and BcLF1; Figure 1B and Supplementary Table 2). This contrasts with 33 EBV proteins that were weakly immunogenic for IgA (Figure 1C; Supplementary Table 2). Antibody Responses Against EBV Proteins Expressed at Different Stages of the EBV Life Cycle Although IgG responses were more common than IgA responses, the biological function of the target protein was also an important determinant of the host antibody response (Table 1). IgG against EBV glycoproteins (median IgG positivity frequency, 68.4%) was more frequently observed than IgG targeting proteins from other stages of the EBV life cycle. Robust IgG responses to EBV latent proteins and proteins involved in the switch from latent to lytic infection were also observed (median positivity frequencies, 51.7% and 45.5%, respectively). Responses to EBV proteins involved in immediate early lytic proteins (median positivity frequency, 36.4%) were less widespread but still frequent, evidence of broad, long-term exposure to proteins from all stages in the EBV life cycle among healthy adults. IgA antibody responses against EBV glycoproteins (median IgA positivity frequency, 42.1%; Table 1) was also more frequently observed than responses against proteins from other stages of the EBV life cycle. The median positivity frequencies observed for IgA antibodies across proteins from latent and lytic phases of the EBV life cycle were similar (range, 15.6%–17.2%; Table 1), with 1 notable exception: IgA antibody responses against EBV proteins involved in the switch from latency to lytic phases were particularly infrequent, with a median positivity frequency of 0% (0 of 12; range, 0%–75%). This is particularly interesting because IgA responses to some of these proteins have been shown to be predictive of the risk of the EBV-associated malignancy NPC [8]. Clustering of Antibody Responses by Immunoglobulin Class Consistent with the observation that IgG responses to EBV were more frequently observed than IgA responses, we noted that IgG responses to a particular antigenic target had, on average, a higher correlation with IgG responses to other antigenic targets, rather than with IgA responses against the same target (median Spearman correlation, 0.495 [range, −0.093–0.978]; 85.9% of comparisons had a Bonferroni-corrected P value of < .05; Figures 2A and B and Supplementary Table 3). Practically, this meant that an individual with elevated IgG titers against a specific EBV protein (eg, protein X) was more likely to have elevated IgG titers against another EBV protein (eg, protein Y), rather than elevated IgA responses to protein X. The same phenomenon was evident for IgA responses, with IgA responses against a particular target being more strongly correlated with IgA responses to other antigenic targets (median Spearman correlation, 0.366 [range, −0.293–0.904]; 64.5% of comparisons had a Bonferroni-corrected P value of < .05), rather than with IgG responses to the same antigen. Figure 2. View largeDownload slide Spearman correlation between the average immunoglobulin G (IgG) and immunoglobulin A (IgA) antibody responses in this population. A, Unsupervised hierarchical clustering of IgG and IgA antibody responses, based on the correlation between the average antibody output for any 2 given Epstein-Barr virus (EBV) targets (Spearman correlation coefficient). Red depicts a strong positive correlation, and blue indicates a negative correlation. B, Distribution of Spearman correlation coefficients according to the type of comparison being made (eg, between 2 IgA antibodies or between IgG and IgA antibodies targeting a selected antigen on the array). aEach row/column represents the average antibody response to a given EBV target in the study population. bEach histogram plots the set of average antibody responses illustrated in panel A. Figure 2. View largeDownload slide Spearman correlation between the average immunoglobulin G (IgG) and immunoglobulin A (IgA) antibody responses in this population. A, Unsupervised hierarchical clustering of IgG and IgA antibody responses, based on the correlation between the average antibody output for any 2 given Epstein-Barr virus (EBV) targets (Spearman correlation coefficient). Red depicts a strong positive correlation, and blue indicates a negative correlation. B, Distribution of Spearman correlation coefficients according to the type of comparison being made (eg, between 2 IgA antibodies or between IgG and IgA antibodies targeting a selected antigen on the array). aEach row/column represents the average antibody response to a given EBV target in the study population. bEach histogram plots the set of average antibody responses illustrated in panel A. While correlations tended to be higher within rather than across antibody class, interimmunoglobulin correlations (ie, IgG vs IgA correlations) were higher when considering the same protein target (median Spearman correlation, 0.226; 23.0% of correlations had a Bonferroni-corrected P value of < .05), rather than different EBV proteins (median Spearman correlation, 0.127 [range, −0.334–0.563]; 1.9% of the correlations had a Bonferroni-corrected P value of < .05). Similar patterns were observed across all EBV life cycle stages (Supplementary Table 3 and Supplementary Figures 11A–E). Clustering of Healthy, Asymptomatic Adults According to Anti-EBV Antibody Responses In addition to understanding the breadth of the antibody response marking EBV exposure in healthy adults, hierarchical clustering identified 4 distinct patterns among study participants that could be distinguished by their average number of positive IgG and IgA antibody responses (Figures 3A and B). We observed 4 groups: group 1 comprised participants with low IgG and IgA responses, group 2 comprised those with a low IgG response and an elevated IgA response, group 3 comprised those with an intermediate IgG response and a low IgA response, and group 4 comprised those with high IgG and IgA responses. Group 1 participants (9.0% [26]) mounted 46 IgG and 15 IgA antibody responses, on average. Group 4 participants (26.0% [75]) mounted 152 IgG and 69 IgA responses, on average. Interestingly, although IgA responses were less frequent than IgG responses overall, group 2 and 4 participants (39% [114]) displayed a high level of IgA reactivity, indicating that ongoing EBV exposure at mucosal surfaces is not an infrequent event in healthy, Taiwanese adults previously exposed to EBV. Figure 3. View largeDownload slide Clustering of individuals defined according to their immunoglobulin G (IgG) and immunoglobulin A (IgA) antibody response. A, Dendrogram showing clusters of individuals defined according to the number of positive antibody responses (with a positive antibody response assigned a value of 1 and a negative response assigned a value of 0) across the full spectrum of evaluated IgG and IgA antibody responses, using unsupervised hierarchical clustering with binary distance and complete linkage. B, Box plot showing number of positive IgG/IgA responses according to 4 groups/clusters based on antibody response, compared with individuals with nasopharyngeal carcinoma (NPC) who had a blood sample collected at the time of cancer diagnosis. Group 1 comprised participants with low IgG and IgA responses, group 2 comprised those with a low IgG response and an elevated IgA response, group 3 comprised those with an intermediate IgG response and a low IgA response, and group 4 comprised those with high IgG and IgA responses. See Methods and Results for additional details. Figure 3. View largeDownload slide Clustering of individuals defined according to their immunoglobulin G (IgG) and immunoglobulin A (IgA) antibody response. A, Dendrogram showing clusters of individuals defined according to the number of positive antibody responses (with a positive antibody response assigned a value of 1 and a negative response assigned a value of 0) across the full spectrum of evaluated IgG and IgA antibody responses, using unsupervised hierarchical clustering with binary distance and complete linkage. B, Box plot showing number of positive IgG/IgA responses according to 4 groups/clusters based on antibody response, compared with individuals with nasopharyngeal carcinoma (NPC) who had a blood sample collected at the time of cancer diagnosis. Group 1 comprised participants with low IgG and IgA responses, group 2 comprised those with a low IgG response and an elevated IgA response, group 3 comprised those with an intermediate IgG response and a low IgA response, and group 4 comprised those with high IgG and IgA responses. See Methods and Results for additional details. We further examined whether group 4 participants were more likely to respond to targets from a specific stage of the EBV life cycle and observed that these subjects had elevated responses to all stages (Supplementary Figures 12A and B). Notably, this high response pattern included higher IgA antibody reactivity against immediate early lytic proteins (median number of positivity among individuals in this group, 1; interquartile range, 0–2.5) compared to the other groups (median number of positivity, 0; interquartile range, 0–0; P = 1.9 × 10-7, by the rank test). We further compared the antibody pattern in group 4 to that previously described among patients with NPC (n = 175) [23]. IgG and IgA responses observed among individuals in group 4 were broadly comparable to those observed among NPC cases (Figure 3B). When we estimated a cancer risk score for asymptomatic individuals in our study by using our previously reported 14-antibody NPC risk stratification signature [23], 79% of individuals in group 4 had scores that ranked in the highest quartile of cancer risk, compared with only 7% of individuals in groups 1–3 (P = 2.2 × 10-14, by the χ2 test; Table 2). These findings suggest that healthy individuals can be classified as having distinct patterns of EBV responses and that a proportion of healthy individuals are characterized by high IgG and IgA responses typically observed among individuals at highest risk of NPC development. Table 2. Nasopharyngeal Carcinoma (NPC) Risk Scores Among Controls From a Cancer Screening Population Cohort in Taiwan, by Immunoglobulin A (IgA) and Immunoglobulin G (IgG) Responses Risk Scorea Control Group,b Participants, No. (%) Group 1 (n = 16) Group 2 (n = 21) Group 3 (n = 49) Group 4 (n = 28) Quartile 1 11 (68.7) 2 (9.5) 14 (28.6) 2 (7.1) Quartile 2 2 (12.5) 5 (23.8) 19 (38.8) 2 (7.1) Quartile 3 2 (12.5) 11 (52.4) 14 (28.6) 2 (7.1) Quartile 4 1 (6.2) 3 (14.3) 2 (4.1) 22 (78.6) Risk Scorea Control Group,b Participants, No. (%) Group 1 (n = 16) Group 2 (n = 21) Group 3 (n = 49) Group 4 (n = 28) Quartile 1 11 (68.7) 2 (9.5) 14 (28.6) 2 (7.1) Quartile 2 2 (12.5) 5 (23.8) 19 (38.8) 2 (7.1) Quartile 3 2 (12.5) 11 (52.4) 14 (28.6) 2 (7.1) Quartile 4 1 (6.2) 3 (14.3) 2 (4.1) 22 (78.6) aEstimated using our previously reported 14-antibody NPC risk stratification signature [23]. Quartile 1 denotes the lowest NPC risk, and quartile 4 denotes the highest NPC risk. bGroup 1 comprised participants with low IgG and IgA responses, group 2 comprised those with a low IgG response and an elevated IgA response, group 3 comprised those with an intermediate IgG response and a low IgA response, and group 4 comprised those with high IgG and IgA responses. See Methods and Results for additional details. View Large Table 2. Nasopharyngeal Carcinoma (NPC) Risk Scores Among Controls From a Cancer Screening Population Cohort in Taiwan, by Immunoglobulin A (IgA) and Immunoglobulin G (IgG) Responses Risk Scorea Control Group,b Participants, No. (%) Group 1 (n = 16) Group 2 (n = 21) Group 3 (n = 49) Group 4 (n = 28) Quartile 1 11 (68.7) 2 (9.5) 14 (28.6) 2 (7.1) Quartile 2 2 (12.5) 5 (23.8) 19 (38.8) 2 (7.1) Quartile 3 2 (12.5) 11 (52.4) 14 (28.6) 2 (7.1) Quartile 4 1 (6.2) 3 (14.3) 2 (4.1) 22 (78.6) Risk Scorea Control Group,b Participants, No. (%) Group 1 (n = 16) Group 2 (n = 21) Group 3 (n = 49) Group 4 (n = 28) Quartile 1 11 (68.7) 2 (9.5) 14 (28.6) 2 (7.1) Quartile 2 2 (12.5) 5 (23.8) 19 (38.8) 2 (7.1) Quartile 3 2 (12.5) 11 (52.4) 14 (28.6) 2 (7.1) Quartile 4 1 (6.2) 3 (14.3) 2 (4.1) 22 (78.6) aEstimated using our previously reported 14-antibody NPC risk stratification signature [23]. Quartile 1 denotes the lowest NPC risk, and quartile 4 denotes the highest NPC risk. bGroup 1 comprised participants with low IgG and IgA responses, group 2 comprised those with a low IgG response and an elevated IgA response, group 3 comprised those with an intermediate IgG response and a low IgA response, and group 4 comprised those with high IgG and IgA responses. See Methods and Results for additional details. View Large Smoking Is Associated With Differential EBV-Directed IgA Antibody Responses The extent to which the effect of known NPC risk factors is mediated through their ability to modulate immune responses to EBV is not fully understood. We observed that individuals from group 4 also had the highest proportion of current smokers, compared with other groups (72% vs 45%; P = 0.02, by the χ2 test). In addition, when considering the proteome-wide antibody response, the total level of the IgA response but not the IgG response was significantly different between smokers and nonsmokers in our population (P = .002, by the SKAT-C; Figure 4A). Notably, when evaluating anti-EBV antibody responses targeting proteins across different stages of the EBV life cycle, the EBV-directed IgA response to latent and early lytic proteins differed significantly by smoking behavior (Bonferroni-corrected P value of < .05, by the SKAT-C). Figure 4. View largeDownload slide Associations between anti–Epstein-Barr virus (EBV) antibody responses and smoking status among males. A, Association between anti-EBV immunoglobulin A (IgA) antibodies against proteins expressed during different EBV life cycles and smoking status (current vs former/never). P values were obtained from sequence kernel association test–C tests. The dashed line represents the Bonferroni-corrected statistically significant P value threshold of .005. B, Box plot showing the association between LMP1/IgA reactivity and smoking status. The P value was obtained from linear regression, adjusted for age. Figure 4. View largeDownload slide Associations between anti–Epstein-Barr virus (EBV) antibody responses and smoking status among males. A, Association between anti-EBV immunoglobulin A (IgA) antibodies against proteins expressed during different EBV life cycles and smoking status (current vs former/never). P values were obtained from sequence kernel association test–C tests. The dashed line represents the Bonferroni-corrected statistically significant P value threshold of .005. B, Box plot showing the association between LMP1/IgA reactivity and smoking status. The P value was obtained from linear regression, adjusted for age. Further, in logistic regression models that evaluated the association between smoking and each marker individually, we observed suggestive associations (P < .05) for multiple anti-EBV IgA antibodies from all stages of the EBV life cycle (37.9% [22 of 58] were associated with latent proteins, 8.3% [1 of 12] were associated with immediate early proteins, 20.0% [9 of 45] were associated with early lytic proteins, 6.4% [3 of 47] were associated with late lytic proteins, and 21.0% [4 of 19] were associated with glycoproteins; Supplementary Tables 4–8). The strongest individual association was observed between smoking and IgA targeting LMP1(odds ratio, 4.64; P = .001; Figure 4B), an EBV oncoprotein expressed in NPC tumors that is normally considered a weakly immunogenic protein [26, 27]. Taken together, these findings suggest that smoking may alter the anti-EBV mucosal antibody response. We also evaluated the association between anti-EBV antibody responses, alcohol drinking, and HLA alleles that modulate NPC risk (ie, HLA-A*0207 and HLA-A*1101; Supplementary Figures 13A–C). HLA-A*0207 was nominally associated with IgA responses against immediate early lytic proteins (P = .02). However, no statistically significant associations persisted after correction for multiple comparisons. DISCUSSION Our findings represent the first comprehensive evaluation of natural variation in the host antibody response to the full spectrum of EBV proteins. This unique characterization was made possible through the use of protein microarray technology in a population of healthy, asymptomatic adults from Taiwan. Because the antibody repertoire can shed light on exposure to viral proteins, this virus-wide description of host response to the EBV proteome represents an important step forward in understanding population-level variation that might have implications for multiple EBV-related diseases. We found that all adults evaluated in our study had evidence of exposure to EBV proteins, that these EBV-positive adults were more likely to mount IgG rather than IgA antibody responses to EBV, and that responses against glycoproteins were particularly prevalent. If the IgG response targeting 1 EBV protein (eg, protein X) was elevated in an individual, they were more likely to mount an IgG response targeting other antigens, rather than an IgA response against protein X, likely reflecting differential timing of exposure (longer-term exposure for IgG and more-recent exposure for IgA) and separate triggering mechanisms in different compartments (eg, oral mucosa vs systemic circulation). Approximately one quarter of the subjects in this study displayed antibody response patterns similar to NPC cases, including positive responses reflecting exposure to immediate early antigenic targets, such as Zta and Rta. Importantly, our data indicate that smoking is associated with altered IgA antibody patterns that could reflect poor control over EBV lytic activity, suggesting a possible mechanism to explain the role of smoking in the etiology of NPC. Previous studies have evaluated responses against a handful of anti-EBV antibodies by using techniques including immunoblot and microarray chips targeting 10–15 EBV proteins [28, 29]. Zheng et al used a microarray to test 82 EBV open-reading frames in a study that included 10 nondiseased adults and showed that these adults mounted IgG responses against an average of 17 proteins (at 1:100 serum dilution) and IgA antibodies against an average of 13 proteins (at 1:1000 serum dilution) [19]. This is in line with our findings of more frequent IgG reactivity, rather than IgA reactivity. In the present study, we expanded testing to 199 EBV open-reading frames among 289 asymptomatic adults. Because our array was not designed to detect antibodies to conformational epitopes that require glycosylation, future studies will be required to better understand patterns of response to those epitopes and how they are related to those reported herein. Our finding that antibody responses to some glycoproteins were the most prevalent is biologically plausible because these proteins are expressed on the viral surface and therefore readily accessible as targets for the immune system [30, 31]. Given that IgA responses are thought to reflect recent pathogen exposure, our finding that IgA responses to glycoproteins were also more frequently observed than responses against proteins from other stages of the EBV life cycle suggests that the presence of EBV in the oral cavity is not a rare event [32]. It was notable that individuals in groups 2 and 4 (39% [114]) mounted a substantial, proteome-wide IgA response. This supports data from studies reporting EBV viral load shedding in the saliva of the majority of healthy adults measured [32, 33], as well as EBV reactivation rates that are much more frequent than those for other herpesviruses, such as cytomegalovirus [34]. An important public health implication of our data is that smoking, a modifiable behavior, was associated with elevated IgA responses against EBV. An association between smoking and EBV reactivation has been posited by one study, which evaluated VCA/IgA in individuals from southern China [35]. In the present study, we expand beyond VCA/IgA alone to demonstrate large-scale differences in the EBV-directed IgA repertoire in smokers. The most significant smoking association was observed for IgA responses against LMP1, an important NPC oncogene that has cell-transforming ability in rodent fibroblasts [36]. In addition to these individual antibody associations, we observed that individuals in group 4 also had the highest proportion of current smokers. Taken together, these findings suggest that smoking may negatively impacts an individual’s ability to control EBV infection, leading to an increased frequency of viral reactivation/exposure in the host. However, we did not assess the degree to which the established association between smoking and NPC is mediated by interindividual variation in the immune response to this ubiquitous virus. The underlying biological mechanisms merit further investigation. In conclusion, we described the pattern of the IgG and IgA responses to EBV among asymptomatic, healthy adults, using a protein microarray chip targeting 199 EBV protein sequences. Given that the generation of antibodies is a complex process requiring interplay between viral antigens and numerous compartments of the immune system [37], our study represents an initial but important step in understanding the immune response to EBV infection. In the future, our approach could be applied to other populations to characterize the similarities and differences observed in the humoral response to EBV in individuals predisposed to different EBV-associated conditions (eg, sub-Saharan Africans at higher risk of Burkitt lymphoma). Such studies could elucidate subsets of EBV-infected individuals at highest risk of EBV-associated clinical conditions and help identify targets for the development of effective immunodiagnostics. Supplementary Data Supplementary materials are available at The Journal of Infectious Diseases online. Consisting of data provided by the authors to benefit the reader, the posted materials are not copyedited and are the sole responsibility of the authors, so questions or comments should be addressed to the corresponding author. Notes Financial support. This work was supported by the National Cancer Institute Intramural Research Program, the National Health and Medical Research Council of Australia, and the National Science Council of Taiwan. Potential conflicts of interest. J. M. M. received payments as owner and chief executive officer of Cyto-Barr. All other authors report no potential conflicts. All authors have submitted the ICMJE Form for Disclosure of Potential Conflicts of Interest. Conflicts that the editors consider relevant to the content of the manuscript have been disclosed. References 1. Kieff E , Rickinson AB . Epstein-Barr virus and its replication . In Fields virology . 5th ed . ed. Philadelphia, PA : Lippincott Williams & Wilkins , 2007 . 2. Kangro HO , Osman HK , Lau YL , Heath RB , Yeung CY , Ng MH . Seroprevalence of antibodies to human herpesviruses in England and Hong Kong . J Med Virol 1994 ; 43 : 91 – 6 . Google Scholar CrossRef Search ADS PubMed 3. Chen CY , Huang KY , Shen JH , Tsao KC , Huang YC . A large-scale seroprevalence of Epstein-Barr virus in Taiwan . PLoS One 2015 ; 10 : e0115836 . Google Scholar CrossRef Search ADS PubMed 4. Young LS , Rickinson AB . Epstein-Barr virus: 40 years on . Nat Rev Cancer 2004 ; 4 : 757 – 68 . Google Scholar CrossRef Search ADS PubMed 5. Cohen JI , Fauci AS , Varmus H , Nabel GJ . Epstein-Barr virus: an important vaccine target for cancer prevention . Sci Transl Med 2011 ; 3 : 107fs7 . Google Scholar CrossRef Search ADS PubMed 6. Kanekiyo M , Bu W , Joyce MG , et al. Rational design of an Epstein-Barr virus vaccine targeting the receptor-binding site . Cell 2015 ; 162 : 1090 – 100 . Google Scholar CrossRef Search ADS PubMed 7. Hjalgrim H , Friborg J , Melbye M . The epidemiology of EBV and its association with malignant disease . In: Arvin A , Campadelli-Fiume G , Mocarski E et al. , eds. Human herpesviruses: biology, therapy, and immunoprophylaxis . Cambridge: Cambridge University Press , 2007 . Google Scholar CrossRef Search ADS 8. Coghill AE , Hildesheim A . Epstein-Barr virus antibodies and the risk of associated malignancies: review of the literature . Am J Epidemiol 2014 ; 180 : 687 – 95 . Google Scholar CrossRef Search ADS PubMed 9. Melbye M , Ebbesen P , Levine PH , Bennike T . Early primary infection and high Epstein-Barr virus antibody titers in Greenland Eskimos at high risk for nasopharyngeal carcinoma . Int J Cancer 1984 ; 34 : 619 – 23 . Google Scholar CrossRef Search ADS PubMed 10. Henle W , Henle G . Epstein-Barr virus-specific serology in immunologically compromised individuals . Cancer Res 1981 ; 41 : 4222 – 5 . Google Scholar PubMed 11. Brandtzaeg P . Potential of nasopharynx-associated lymphoid tissue for vaccine responses in the airways . Am J Respir Crit Care Med 2011 ; 183 : 1595 – 604 . Google Scholar CrossRef Search ADS PubMed 12. Bhaduri-McIntosh S , Landry ML , Nikiforow S , Rotenberg M , El-Guindy A , Miller G . Serum IgA antibodies to Epstein-Barr virus (EBV) early lytic antigens are present in primary EBV infection . J Infect Dis 2007 ; 195 : 483 – 92 . Google Scholar CrossRef Search ADS PubMed 13. Chien YC , Chen JY , Liu MY , et al. Serologic markers of Epstein-Barr virus infection and nasopharyngeal carcinoma in Taiwanese men . N Engl J Med 2001 ; 345 : 1877 – 82 . Google Scholar CrossRef Search ADS PubMed 14. Cao SM , Liu Z , Jia WH , et al. Fluctuations of Epstein-Barr virus serological antibodies and risk for nasopharyngeal carcinoma: a prospective screening study with a 20-year follow-up . PLoS One 2011 ; 6 : e19100 . Google Scholar CrossRef Search ADS PubMed 15. Henle W , Henle G , Zajac BA , Pearson G , Waubke R , Scriba M . Differential reactivity of human serums with early antigens induced by Epstein-Barr virus . Science 1970 ; 169 : 188 – 90 . Google Scholar CrossRef Search ADS PubMed 16. Ginsburg CM , Henle W , Henle G , Horwitz CA . Infectious mononucleosis in children. Evaluation of Epstein-Barr virus-specific serological data . JAMA 1977 ; 237 : 781 – 5 . Google Scholar CrossRef Search ADS PubMed 17. Doolan DL . Plasmodium immunomics . Int J Parasitol 2011 ; 41 : 3 – 20 . Google Scholar CrossRef Search ADS PubMed 18. Davies DH , Duffy P , Bodmer JL , Felgner PL , Doolan DL . Large screen approaches to identify novel malaria vaccine candidates . Vaccine 2015 ; 33 : 7496 – 505 . Google Scholar CrossRef Search ADS PubMed 19. Zheng D , Wan J , Cho YG , et al. Comparison of humoral immune responses to Epstein-Barr virus and Kaposi’s sarcoma-associated herpesvirus using a viral proteome microarray . J Infect Dis 2011 ; 204 : 1683 – 91 . Google Scholar CrossRef Search ADS PubMed 20. Chen CJ , Yang HI , Su J , et al. ; REVEAL-HBV Study Group . Risk of hepatocellular carcinoma across a biological gradient of serum hepatitis B virus DNA level . JAMA 2006 ; 295 : 65 – 73 . Google Scholar CrossRef Search ADS PubMed 21. Hildesheim A , Apple RJ , Chen CJ , et al. Association of HLA class I and II alleles and extended haplotypes with nasopharyngeal carcinoma in Taiwan . J Natl Cancer Inst 2002 ; 94 : 1780 – 9 . Google Scholar CrossRef Search ADS PubMed 22. Vigil A , Davies DH , Felgner PL . Defining the humoral immune response to infectious agents using high-density protein microarrays . Future Microbiol 2010 ; 5 : 241 – 51 . Google Scholar CrossRef Search ADS PubMed 23. Coghill AE , Pfeiffer RM , Proietti C , et al. Identification of a novel, EBV-based antibody risk stratification signature for early detection of nasopharyngeal carcinoma in Taiwan . Clin Cancer Res 2018 ; 24 : 1305 – 14 . Google Scholar CrossRef Search ADS PubMed 24. Galili T . dendextend: an R package for visualizing, adjusting and comparing trees of hierarchical clustering . Bioinformatics 2015 ; 31 : 3718 – 20 . Google Scholar CrossRef Search ADS PubMed 25. Ionita-Laza I , Lee S , Makarov V , Buxbaum JD , Lin X . Sequence kernel association tests for the combined effect of rare and common variants . Am J Hum Genet 2013 ; 92 : 841 – 53 . Google Scholar CrossRef Search ADS PubMed 26. Dawson CW , Port RJ , Young LS . The role of the EBV-encoded latent membrane proteins LMP1 and LMP2 in the pathogenesis of nasopharyngeal carcinoma (NPC) . Semin Cancer Biol 2012 ; 22 : 144 – 53 . Google Scholar CrossRef Search ADS PubMed 27. Meij P , Vervoort MB , Aarbiou J , et al. Restricted low-level human antibody responses against Epstein-Barr virus (EBV)-encoded latent membrane protein 1 in a subgroup of patients with EBV-associated diseases . J Infect Dis 1999 ; 179 : 1108 – 15 . Google Scholar CrossRef Search ADS PubMed 28. Fachiroh J , Schouten T , Hariwiyanto B , et al. Molecular diversity of Epstein-Barr virus IgG and IgA antibody responses in nasopharyngeal carcinoma: a comparison of Indonesian, Chinese, and European subjects . J Infect Dis 2004 ; 190 : 53 – 62 . Google Scholar CrossRef Search ADS PubMed 29. Loebel M , Eckey M , Sotzny F , et al. Serological profiling of the EBV immune response in Chronic Fatigue Syndrome using a peptide microarray . PLoS One 2017 ; 12 : e0179124 . Google Scholar CrossRef Search ADS PubMed 30. Hutt-Fletcher LM . EBV glycoproteins: where are we now ? Future Virol 2015 ; 10 : 1155 – 62 . Google Scholar CrossRef Search ADS PubMed 31. Li Q , Turk SM , Hutt-Fletcher LM . The Epstein-Barr virus (EBV) BZLF2 gene product associates with the gH and gL homologs of EBV and carries an epitope critical to infection of B cells but not of epithelial cells . J Virol 1995 ; 69 : 3987 – 94 . Google Scholar PubMed 32. Hadinoto V , Shapiro M , Sun CC , Thorley-Lawson DA . The dynamics of EBV shedding implicate a central role for epithelial cells in amplifying viral output . PLoS Pathog 2009 ; 5 : e1000496 . Google Scholar CrossRef Search ADS PubMed 33. Yao QY , Rickinson AB , Epstein MA . A re-examination of the Epstein-Barr virus carrier state in healthy seropositive individuals . Int J Cancer 1985 ; 35 : 35 – 42 . Google Scholar CrossRef Search ADS PubMed 34. Ling PD , Lednicky JA , Keitel WA , et al. The dynamics of herpesvirus and polyomavirus reactivation and shedding in healthy adults: a 14-month longitudinal study . J Infect Dis 2003 ; 187 : 1571 – 80 . Google Scholar CrossRef Search ADS PubMed 35. Xu FH , Xiong D , Xu YF , et al. An epidemiological and molecular study of the relationship between smoking, risk of nasopharyngeal carcinoma, and Epstein-Barr virus activation . J Natl Cancer Inst 2012 ; 104 : 1396 – 410 . Google Scholar CrossRef Search ADS PubMed 36. Wang D , Liebowitz D , Kieff E . An EBV membrane protein expressed in immortalized lymphocytes transforms established rodent cells . Cell 1985 ; 43 : 831 – 40 . Google Scholar CrossRef Search ADS PubMed 37. Nutt SL , Hodgkin PD , Tarlinton DM , Corcoran LM . The generation of antibody-secreting plasma cells . Nat Rev Immunol 2015 ; 15 : 160 – 71 . Google Scholar CrossRef Search ADS PubMed Published by Oxford University Press for the Infectious Diseases Society of America 2018. This work is written by (a) US Government employee(s) and is in the public domain in the US. http://www.deepdyve.com/assets/images/DeepDyve-Logo-lg.png The Journal of Infectious Diseases Oxford University Press

Patterns of Interindividual Variability in the Antibody Repertoire Targeting Proteins Across the Epstein-Barr Virus Proteome

Loading next page...
 
/lp/ou_press/patterns-of-interindividual-variability-in-the-antibody-repertoire-TudQld1lyP
Publisher
Oxford University Press
Copyright
Published by Oxford University Press for the Infectious Diseases Society of America 2018.
ISSN
0022-1899
eISSN
1537-6613
D.O.I.
10.1093/infdis/jiy122
Publisher site
See Article on Publisher Site

Abstract

Abstract Background Little is known about variation in antibody responses targeting the full spectrum of Epstein-Barr virus (EBV) proteins and how such patterns inform disease risk. Methods We used a microarray to measure immunoglobulin G (IgG) and immunoglobulin A (IgA) antibody responses against 199 EBV protein sequences from 5 EBV strains recovered from 289 healthy adults from Taiwan. We described positivity patterns, estimated the correlation between antibodies, and investigated the associations between environmental and genetic risk factors and variations in antibody responses. Results Healthy adults were more likely to mount IgG antibody responses to EBV proteins (median positivity frequency, 46.5% for IgG and 17.3% for IgA; P = 1.6 × 10–46, by the Wilcoxon rank sum test). Responses against glycoproteins were particularly prevalent. The correlations between antibody responses of the same class were higher than correlations across classes. The mucosal exposure to proteins involved in EBV reactivation (as determined by the IgA response) was associated with smoking (P = .002, by the sequence kernel association test–combined), and approximately one quarter of adults displayed antibody responses associated with EBV-related cancer risk. Conclusions These data comprehensively define the variability in human IgG and IgA antibody responses to the EBV proteome. Patterns observed can serve as the foundation for elucidating which individuals are at highest risk of EBV-associated clinical conditions and for identifying targets for effective immunodiagnostic tests. Antibody microarray, antibody response, Epstein-Barr virus, Taiwan, IgA, IgG Epstein-Barr virus (EBV) is a ubiquitous herpesvirus that infects >90% of individuals worldwide [1]. Primary infection with the virus typically occurs in early childhood [2, 3], after which EBV establishes lifelong latency in human B cells. During the course of lifelong infection, EBV periodically reactivates in B cells or epithelial cells, expressing a broader set of lytic proteins required for viral replication [4]. This ongoing replication in epithelial cells of the oral cavity contributes to shedding of the virus into saliva, facilitating person-to-person transmission of EBV. Although established EBV infection is generally asymptomatic, infection manifests as clinical disease in a subset of infected individuals; such diseases include hematopoietic malignancies (eg, Burkitt and Hodgkin lymphomas) and epithelial malignancies (eg, gastric cancer and nasopharyngeal carcinoma [NPC]) [4]. EBV is etiologically involved in approximately 200000 cancers worldwide each year [5]. Despite its ubiquity and association with various chronic conditions, there are currently no proven ways to prevent EBV infection from occurring or to predict with sufficient accuracy which infected individuals will develop serious chronic diseases associated with this virus [5, 6]. One way to monitor patterns of EBV exposure in the general population is to measure levels of circulating anti-EBV antibodies that reflect responses to viral proteins. For example, immunoglobulin G (IgG) antibodies appear within several weeks after primary infection and remain in circulation for many years [7]. Prior studies aimed at defining rates of EBV infection by age across populations have used positivity for IgG antibodies targeting EBV proteins such as the viral capsid antigen (VCA) and EBV nuclear antigen 1 (EBNA1) [8] to classify individuals as EBV positive (ie, as ever exposed to EBV). IgG antibodies can also be used to better understand the nature of an individual’s EBV infection. For example, a subset of long-term EBV carriers displays elevated levels of IgG antibody targeting the early antigen protein (EA) [9]. This EA complex is expressed as part of the EBV lytic cycle; therefore, elevated levels of EA indicate ongoing viral lytic activity (reactivation) and a potential lack of control over the virus [10]. In contrast to IgG, immunoglobulin A (IgA) responses to viral exposure are believed to reflect recent reactivation at mucosal sites (eg, the oral epithelium) [11]. This type of immune response is of interest because of the oral route of EBV transmission. During primary infection, IgA responses against EBV in the oral cavity can be strong but are short lived [12]. The durability and spectrum of EBV-directed IgA responses in healthy adults are largely uncharacterized. However, it has been shown that individuals at greatest risk of developing NPC, an EBV-associated cancer, have elevated anti-EBV IgA responses to VCA and EBNA1 years before a cancer diagnosis [8, 13, 14]. Although informative, prior studies of humoral responses to EBV infection [15, 16] have largely focused on measuring levels of IgG and IgA antibodies to a handful (<10) of the approximately 90 EBV protein targets [8]. With the advent of protein array systems capable of screening for antibody responses against the full complement of EBV proteins [17–19], it is now feasible to evaluate humoral responses to EBV infection in a more comprehensive manner. Thus, we evaluated IgG and IgA anti-EBV antibody responses against 199 predicted sequences and 3 synthetic antigens from 86 EBV proteins involved in various stages of the EBV life cycle (eg, latent and lytic cycles) among 289 healthy individuals from Taiwan, a region with high rates of EBV-associated cancer. METHODS A total of 289 healthy, asymptomatic adults who served as controls for 2 previously published Taiwanese studies were selected for this report [20, 21]. Details are described in the Supplementary Materials. A previously described EBV protein microarray targeting IgG and IgA antibodies against 199 predicted EBV protein sequences (86 EBV proteins) from 5 EBV strains (AG876, Akata, B95-8, Mutu, and Raji) was used [17, 18, 22, 23]. Briefly, these 199 sequences represent nonredundant open reading frames from 5 EBV strains representing both African (Mutu, Raji, and AG876) and Asian (Akata) variants, as well as known splice variants identified in the literature. Eighty-five percent of 199 predicted sequences represented complete transcripts of EBV genes. The remaining 30 predicted sequences represented linear segments from 8 EBV genes >1000 bp long. Each of the protein sequences cloned into the pXT7 expression vector prior to printing onto the microarray included N-terminal 10x histidine (His) and C-terminal hemagglutinin (HA) tags to confirm expression on the microarray. After printing, anti-polyHis probing confirmed the presence of 99.5% of sequences; anti-HA probes confirmed the presence of 98.5% of sequences. High coverage was achieved across the 5 EBV strains, with 97% of the predicted sequences from each strain represented on the microarray at ≥99% homology. A positive antibody response in a healthy adult was defined as a response greater than that individual’s person-specific background (mean of 4 no-DNA-control spots plus 1.5 standard deviations) [17]. As previously reported [23], we included blinded quality control replicates during testing and observed good reproducibility for antibodies measured using this custom protein microarray. The average percentage agreement for classifying an individual’s antibody response as positive versus negative was high for both IgA (84%; range, 52%–100%) and IgG (79%; range, 48%–100%), respectively. The antibody level (determined on the basis of the standardized signal intensity output from the array) was further grouped into quartiles (4 categories) based on the population distribution of the response for each given antibody. Based on coefficients of variation reported from the previous study [23], we excluded from analysis the 2 IgG array spots out of 199 evaluated that had coefficients of variation of >20% (BZLF1/IgG [array protein sequence CAA24861.1-103155-102655] and BXRF1/IgG [array protein sequence CAA24798.1-144860-145606]). Finally, we included 3 synthetic EBV peptides that are putative cancer biomarkers (VCAp18, EBNA1, and EAdp47) on the microarray, bringing the total number of anti-EBV antibodies for analysis to 402 (200 IgG and 202 IgA antibodies). Ethics Statement All human subjects included were adults, and all studies were approved by appropriate human subjects committees in Taiwan and the United States. This study was reviewed/approved by the National Cancer Institute Special Studies Institutional Review Board and the National Taiwan University Institutional Review Board)\. Written informed consent was obtained for all participants. Statistical Analysis Visualization and statistical analyses were conducted using R statistical software. Details on the statistical analyses are described in the Supplementary Materials. Briefly, to pool data across the 2 study populations, we used population-specific cut points to generate similar underlying antibody distributions and then compared individual response rates (ie, percentage positivity values) across immunoglobulin class and EBV life cycle. We estimated the correlation between antibodies across immunoglobulin classes (ie, IgG against protein X versus IgA against protein X) and within each immunoglobulin class (ie, IgG against protein X versus IgG against protein Y), using Spearman correlation coefficients. All individuals were clustered according to the positive responses across the full EBV proteome, using unsupervised hierarchical clustering with binary distance and complete linkage [24]. To examine the association between environmental and genetic risk factors (ie, smoking, alcohol drinking, and HLA alleles) and biological groupings of antibody response (eg, responses directed against a select EBV life cycle), we used the sequence kernel association test–combined (SKAT-C) from the single-nucleotide polymorphism set [25], using binary variables (with a positive antibody response assigned a value of 1 and a negative response assigned a value of 0). RESULTS Anti-EBV IgG and IgA Antibody Responses Mounted by Healthy Individuals All individuals had measurable (ie, positive) antibody responses against EBV (Figures 1A–C; Supplementary Figures 1–10). The individual with the narrowest pattern responded to 38 EBV targets on the array (28 IgG antibodies and 10 IgA antibodies), representing 21 EBV proteins (Figure 1A). The individual with the broadest response had a positive response against 329 EBV targets on the array (163 IgG antibodies and 166 IgA antibodies), representing 84 of 86 EBV proteins evaluated. Positive IgG (ie, long-term) responses against EBV proteins were markedly more frequent than positive IgA (ie, more-recent) responses (Table 1). On average, individuals mounted IgG antibody responses against 93 (46.5%) of 200 EBV targets on the array and IgA antibody responses against 35 (17.3%) of 202 EBV targets. Figure 1. View largeDownload slide Anti–Epstein-Barr virus (EBV) immunoglobulin G (IgG) and immunoglobulin A (IgA) antibody responses among healthy individuals in Taiwan. A, Number of positive antibody responses in 289 healthy individuals. Blue indicates IgG antibody positivity, and red indicates IgA positivity. The individual with the narrowest response had antibodies against 38 EBV targets on the array (28 IgG antibodies and 10 IgA antibodies), representing 21 EBV proteins. In contrast, the individual with the broadest response had antibodies against 329 EBV targets (163 IgG antibodies and 166 IgA antibodies), representing 84 of 86 tested EBV proteins. B and C, The 20 IgG (B) and IgA (C) antibodies with the highest antibody positivity rates and the 20 with the lowest antibody positivity rates. Percentages indicate the antibody positivity rate for the given target antigen in this population. Figure 1. View largeDownload slide Anti–Epstein-Barr virus (EBV) immunoglobulin G (IgG) and immunoglobulin A (IgA) antibody responses among healthy individuals in Taiwan. A, Number of positive antibody responses in 289 healthy individuals. Blue indicates IgG antibody positivity, and red indicates IgA positivity. The individual with the narrowest response had antibodies against 38 EBV targets on the array (28 IgG antibodies and 10 IgA antibodies), representing 21 EBV proteins. In contrast, the individual with the broadest response had antibodies against 329 EBV targets (163 IgG antibodies and 166 IgA antibodies), representing 84 of 86 tested EBV proteins. B and C, The 20 IgG (B) and IgA (C) antibodies with the highest antibody positivity rates and the 20 with the lowest antibody positivity rates. Percentages indicate the antibody positivity rate for the given target antigen in this population. Table 1. Number of Positive Antibody Responses to Epstein-Barr Virus (EBV) Proteins Among Healthy Individuals in Taiwan, by Antibody Type and Stage of EBV Life Cycle Targeted Antibody, Target Protein Antibodies, No. Median (%a) IQR Range IgG  Overall 200 93 (46.5) 67–132 22–188  Latent 58 30 (51.7) 21–40 21–58  Immediate early lytic 11 4 (36.4) 2–7 0–11  Early lytic 44 20 (45.5) 13–28 5–42  Late lytic 47 20 (42.6) 14–30 7–45  Glycoprotein 19 13 (68.4) 11–16 4–19  Other/unknown 21 7 (33.3) 4–12 0–20 IgA  Overall 202 35 (17.3) 18–85 4–166  Latent 58 10 (17.2) 4–20 0–48  Immediate early lytic 12 0 (0.0) 0–1 0–9  Early lytic 45 7 (15.6) 3–15 0–35  Late lytic 47 8 (17.0) 6–14 0–39  Glycoprotein 19 8 (42.1) 3–15 0–19  Other/unknown 21 1 (4.8) 0–6 0–17 Antibody, Target Protein Antibodies, No. Median (%a) IQR Range IgG  Overall 200 93 (46.5) 67–132 22–188  Latent 58 30 (51.7) 21–40 21–58  Immediate early lytic 11 4 (36.4) 2–7 0–11  Early lytic 44 20 (45.5) 13–28 5–42  Late lytic 47 20 (42.6) 14–30 7–45  Glycoprotein 19 13 (68.4) 11–16 4–19  Other/unknown 21 7 (33.3) 4–12 0–20 IgA  Overall 202 35 (17.3) 18–85 4–166  Latent 58 10 (17.2) 4–20 0–48  Immediate early lytic 12 0 (0.0) 0–1 0–9  Early lytic 45 7 (15.6) 3–15 0–35  Late lytic 47 8 (17.0) 6–14 0–39  Glycoprotein 19 8 (42.1) 3–15 0–19  Other/unknown 21 1 (4.8) 0–6 0–17 Abbreviations: IgA, immunoglobulin A; IgG, immunoglobulin G; IQR, interquartile range. aCalculated on the basis of the median number of positive antibodies among the total number of antibodies evaluated in each group. View Large Table 1. Number of Positive Antibody Responses to Epstein-Barr Virus (EBV) Proteins Among Healthy Individuals in Taiwan, by Antibody Type and Stage of EBV Life Cycle Targeted Antibody, Target Protein Antibodies, No. Median (%a) IQR Range IgG  Overall 200 93 (46.5) 67–132 22–188  Latent 58 30 (51.7) 21–40 21–58  Immediate early lytic 11 4 (36.4) 2–7 0–11  Early lytic 44 20 (45.5) 13–28 5–42  Late lytic 47 20 (42.6) 14–30 7–45  Glycoprotein 19 13 (68.4) 11–16 4–19  Other/unknown 21 7 (33.3) 4–12 0–20 IgA  Overall 202 35 (17.3) 18–85 4–166  Latent 58 10 (17.2) 4–20 0–48  Immediate early lytic 12 0 (0.0) 0–1 0–9  Early lytic 45 7 (15.6) 3–15 0–35  Late lytic 47 8 (17.0) 6–14 0–39  Glycoprotein 19 8 (42.1) 3–15 0–19  Other/unknown 21 1 (4.8) 0–6 0–17 Antibody, Target Protein Antibodies, No. Median (%a) IQR Range IgG  Overall 200 93 (46.5) 67–132 22–188  Latent 58 30 (51.7) 21–40 21–58  Immediate early lytic 11 4 (36.4) 2–7 0–11  Early lytic 44 20 (45.5) 13–28 5–42  Late lytic 47 20 (42.6) 14–30 7–45  Glycoprotein 19 13 (68.4) 11–16 4–19  Other/unknown 21 7 (33.3) 4–12 0–20 IgA  Overall 202 35 (17.3) 18–85 4–166  Latent 58 10 (17.2) 4–20 0–48  Immediate early lytic 12 0 (0.0) 0–1 0–9  Early lytic 45 7 (15.6) 3–15 0–35  Late lytic 47 8 (17.0) 6–14 0–39  Glycoprotein 19 8 (42.1) 3–15 0–19  Other/unknown 21 1 (4.8) 0–6 0–17 Abbreviations: IgA, immunoglobulin A; IgG, immunoglobulin G; IQR, interquartile range. aCalculated on the basis of the median number of positive antibodies among the total number of antibodies evaluated in each group. View Large Additional evidence of broad anti-EBV IgG reactivity was provided by the observation that almost all individuals (≥95%) in our study mounted a positive IgG response against 14 different EBV proteins, including 2 latent proteins (EBNA3B and EBNA1), 4 early lytic proteins (BILF1, BDLF4, BLRF2, and BGLF5), 4 late lytic proteins (BRRF2, BFRF3, VCA-p18, and BPFL1), and 4 glycoproteins (BZLF2, BDLF3, BBRF3, and BMRF2; Figure 1B and Supplementary Table 1). This IgG pattern stands in contrast to that of IgA, for which only 1 target on the array (representing BZLF2) elicited a uniformly positive response (Figure 1C; Supplementary Table 1), with a plurality of study participants (40.5% [117]) positive for <25 array targets. Only 8 EBV targets were weakly immunogenic at eliciting IgG responses (ie, <5% of individuals tested positive), including 1 targeting a latent protein (EBNA3C), 1 early lytic protein (EAD-p47: BMRF1), and 4 late lytic proteins (BGRF1/BDRF1, BFLF2, BRFR1A, and BcLF1; Figure 1B and Supplementary Table 2). This contrasts with 33 EBV proteins that were weakly immunogenic for IgA (Figure 1C; Supplementary Table 2). Antibody Responses Against EBV Proteins Expressed at Different Stages of the EBV Life Cycle Although IgG responses were more common than IgA responses, the biological function of the target protein was also an important determinant of the host antibody response (Table 1). IgG against EBV glycoproteins (median IgG positivity frequency, 68.4%) was more frequently observed than IgG targeting proteins from other stages of the EBV life cycle. Robust IgG responses to EBV latent proteins and proteins involved in the switch from latent to lytic infection were also observed (median positivity frequencies, 51.7% and 45.5%, respectively). Responses to EBV proteins involved in immediate early lytic proteins (median positivity frequency, 36.4%) were less widespread but still frequent, evidence of broad, long-term exposure to proteins from all stages in the EBV life cycle among healthy adults. IgA antibody responses against EBV glycoproteins (median IgA positivity frequency, 42.1%; Table 1) was also more frequently observed than responses against proteins from other stages of the EBV life cycle. The median positivity frequencies observed for IgA antibodies across proteins from latent and lytic phases of the EBV life cycle were similar (range, 15.6%–17.2%; Table 1), with 1 notable exception: IgA antibody responses against EBV proteins involved in the switch from latency to lytic phases were particularly infrequent, with a median positivity frequency of 0% (0 of 12; range, 0%–75%). This is particularly interesting because IgA responses to some of these proteins have been shown to be predictive of the risk of the EBV-associated malignancy NPC [8]. Clustering of Antibody Responses by Immunoglobulin Class Consistent with the observation that IgG responses to EBV were more frequently observed than IgA responses, we noted that IgG responses to a particular antigenic target had, on average, a higher correlation with IgG responses to other antigenic targets, rather than with IgA responses against the same target (median Spearman correlation, 0.495 [range, −0.093–0.978]; 85.9% of comparisons had a Bonferroni-corrected P value of < .05; Figures 2A and B and Supplementary Table 3). Practically, this meant that an individual with elevated IgG titers against a specific EBV protein (eg, protein X) was more likely to have elevated IgG titers against another EBV protein (eg, protein Y), rather than elevated IgA responses to protein X. The same phenomenon was evident for IgA responses, with IgA responses against a particular target being more strongly correlated with IgA responses to other antigenic targets (median Spearman correlation, 0.366 [range, −0.293–0.904]; 64.5% of comparisons had a Bonferroni-corrected P value of < .05), rather than with IgG responses to the same antigen. Figure 2. View largeDownload slide Spearman correlation between the average immunoglobulin G (IgG) and immunoglobulin A (IgA) antibody responses in this population. A, Unsupervised hierarchical clustering of IgG and IgA antibody responses, based on the correlation between the average antibody output for any 2 given Epstein-Barr virus (EBV) targets (Spearman correlation coefficient). Red depicts a strong positive correlation, and blue indicates a negative correlation. B, Distribution of Spearman correlation coefficients according to the type of comparison being made (eg, between 2 IgA antibodies or between IgG and IgA antibodies targeting a selected antigen on the array). aEach row/column represents the average antibody response to a given EBV target in the study population. bEach histogram plots the set of average antibody responses illustrated in panel A. Figure 2. View largeDownload slide Spearman correlation between the average immunoglobulin G (IgG) and immunoglobulin A (IgA) antibody responses in this population. A, Unsupervised hierarchical clustering of IgG and IgA antibody responses, based on the correlation between the average antibody output for any 2 given Epstein-Barr virus (EBV) targets (Spearman correlation coefficient). Red depicts a strong positive correlation, and blue indicates a negative correlation. B, Distribution of Spearman correlation coefficients according to the type of comparison being made (eg, between 2 IgA antibodies or between IgG and IgA antibodies targeting a selected antigen on the array). aEach row/column represents the average antibody response to a given EBV target in the study population. bEach histogram plots the set of average antibody responses illustrated in panel A. While correlations tended to be higher within rather than across antibody class, interimmunoglobulin correlations (ie, IgG vs IgA correlations) were higher when considering the same protein target (median Spearman correlation, 0.226; 23.0% of correlations had a Bonferroni-corrected P value of < .05), rather than different EBV proteins (median Spearman correlation, 0.127 [range, −0.334–0.563]; 1.9% of the correlations had a Bonferroni-corrected P value of < .05). Similar patterns were observed across all EBV life cycle stages (Supplementary Table 3 and Supplementary Figures 11A–E). Clustering of Healthy, Asymptomatic Adults According to Anti-EBV Antibody Responses In addition to understanding the breadth of the antibody response marking EBV exposure in healthy adults, hierarchical clustering identified 4 distinct patterns among study participants that could be distinguished by their average number of positive IgG and IgA antibody responses (Figures 3A and B). We observed 4 groups: group 1 comprised participants with low IgG and IgA responses, group 2 comprised those with a low IgG response and an elevated IgA response, group 3 comprised those with an intermediate IgG response and a low IgA response, and group 4 comprised those with high IgG and IgA responses. Group 1 participants (9.0% [26]) mounted 46 IgG and 15 IgA antibody responses, on average. Group 4 participants (26.0% [75]) mounted 152 IgG and 69 IgA responses, on average. Interestingly, although IgA responses were less frequent than IgG responses overall, group 2 and 4 participants (39% [114]) displayed a high level of IgA reactivity, indicating that ongoing EBV exposure at mucosal surfaces is not an infrequent event in healthy, Taiwanese adults previously exposed to EBV. Figure 3. View largeDownload slide Clustering of individuals defined according to their immunoglobulin G (IgG) and immunoglobulin A (IgA) antibody response. A, Dendrogram showing clusters of individuals defined according to the number of positive antibody responses (with a positive antibody response assigned a value of 1 and a negative response assigned a value of 0) across the full spectrum of evaluated IgG and IgA antibody responses, using unsupervised hierarchical clustering with binary distance and complete linkage. B, Box plot showing number of positive IgG/IgA responses according to 4 groups/clusters based on antibody response, compared with individuals with nasopharyngeal carcinoma (NPC) who had a blood sample collected at the time of cancer diagnosis. Group 1 comprised participants with low IgG and IgA responses, group 2 comprised those with a low IgG response and an elevated IgA response, group 3 comprised those with an intermediate IgG response and a low IgA response, and group 4 comprised those with high IgG and IgA responses. See Methods and Results for additional details. Figure 3. View largeDownload slide Clustering of individuals defined according to their immunoglobulin G (IgG) and immunoglobulin A (IgA) antibody response. A, Dendrogram showing clusters of individuals defined according to the number of positive antibody responses (with a positive antibody response assigned a value of 1 and a negative response assigned a value of 0) across the full spectrum of evaluated IgG and IgA antibody responses, using unsupervised hierarchical clustering with binary distance and complete linkage. B, Box plot showing number of positive IgG/IgA responses according to 4 groups/clusters based on antibody response, compared with individuals with nasopharyngeal carcinoma (NPC) who had a blood sample collected at the time of cancer diagnosis. Group 1 comprised participants with low IgG and IgA responses, group 2 comprised those with a low IgG response and an elevated IgA response, group 3 comprised those with an intermediate IgG response and a low IgA response, and group 4 comprised those with high IgG and IgA responses. See Methods and Results for additional details. We further examined whether group 4 participants were more likely to respond to targets from a specific stage of the EBV life cycle and observed that these subjects had elevated responses to all stages (Supplementary Figures 12A and B). Notably, this high response pattern included higher IgA antibody reactivity against immediate early lytic proteins (median number of positivity among individuals in this group, 1; interquartile range, 0–2.5) compared to the other groups (median number of positivity, 0; interquartile range, 0–0; P = 1.9 × 10-7, by the rank test). We further compared the antibody pattern in group 4 to that previously described among patients with NPC (n = 175) [23]. IgG and IgA responses observed among individuals in group 4 were broadly comparable to those observed among NPC cases (Figure 3B). When we estimated a cancer risk score for asymptomatic individuals in our study by using our previously reported 14-antibody NPC risk stratification signature [23], 79% of individuals in group 4 had scores that ranked in the highest quartile of cancer risk, compared with only 7% of individuals in groups 1–3 (P = 2.2 × 10-14, by the χ2 test; Table 2). These findings suggest that healthy individuals can be classified as having distinct patterns of EBV responses and that a proportion of healthy individuals are characterized by high IgG and IgA responses typically observed among individuals at highest risk of NPC development. Table 2. Nasopharyngeal Carcinoma (NPC) Risk Scores Among Controls From a Cancer Screening Population Cohort in Taiwan, by Immunoglobulin A (IgA) and Immunoglobulin G (IgG) Responses Risk Scorea Control Group,b Participants, No. (%) Group 1 (n = 16) Group 2 (n = 21) Group 3 (n = 49) Group 4 (n = 28) Quartile 1 11 (68.7) 2 (9.5) 14 (28.6) 2 (7.1) Quartile 2 2 (12.5) 5 (23.8) 19 (38.8) 2 (7.1) Quartile 3 2 (12.5) 11 (52.4) 14 (28.6) 2 (7.1) Quartile 4 1 (6.2) 3 (14.3) 2 (4.1) 22 (78.6) Risk Scorea Control Group,b Participants, No. (%) Group 1 (n = 16) Group 2 (n = 21) Group 3 (n = 49) Group 4 (n = 28) Quartile 1 11 (68.7) 2 (9.5) 14 (28.6) 2 (7.1) Quartile 2 2 (12.5) 5 (23.8) 19 (38.8) 2 (7.1) Quartile 3 2 (12.5) 11 (52.4) 14 (28.6) 2 (7.1) Quartile 4 1 (6.2) 3 (14.3) 2 (4.1) 22 (78.6) aEstimated using our previously reported 14-antibody NPC risk stratification signature [23]. Quartile 1 denotes the lowest NPC risk, and quartile 4 denotes the highest NPC risk. bGroup 1 comprised participants with low IgG and IgA responses, group 2 comprised those with a low IgG response and an elevated IgA response, group 3 comprised those with an intermediate IgG response and a low IgA response, and group 4 comprised those with high IgG and IgA responses. See Methods and Results for additional details. View Large Table 2. Nasopharyngeal Carcinoma (NPC) Risk Scores Among Controls From a Cancer Screening Population Cohort in Taiwan, by Immunoglobulin A (IgA) and Immunoglobulin G (IgG) Responses Risk Scorea Control Group,b Participants, No. (%) Group 1 (n = 16) Group 2 (n = 21) Group 3 (n = 49) Group 4 (n = 28) Quartile 1 11 (68.7) 2 (9.5) 14 (28.6) 2 (7.1) Quartile 2 2 (12.5) 5 (23.8) 19 (38.8) 2 (7.1) Quartile 3 2 (12.5) 11 (52.4) 14 (28.6) 2 (7.1) Quartile 4 1 (6.2) 3 (14.3) 2 (4.1) 22 (78.6) Risk Scorea Control Group,b Participants, No. (%) Group 1 (n = 16) Group 2 (n = 21) Group 3 (n = 49) Group 4 (n = 28) Quartile 1 11 (68.7) 2 (9.5) 14 (28.6) 2 (7.1) Quartile 2 2 (12.5) 5 (23.8) 19 (38.8) 2 (7.1) Quartile 3 2 (12.5) 11 (52.4) 14 (28.6) 2 (7.1) Quartile 4 1 (6.2) 3 (14.3) 2 (4.1) 22 (78.6) aEstimated using our previously reported 14-antibody NPC risk stratification signature [23]. Quartile 1 denotes the lowest NPC risk, and quartile 4 denotes the highest NPC risk. bGroup 1 comprised participants with low IgG and IgA responses, group 2 comprised those with a low IgG response and an elevated IgA response, group 3 comprised those with an intermediate IgG response and a low IgA response, and group 4 comprised those with high IgG and IgA responses. See Methods and Results for additional details. View Large Smoking Is Associated With Differential EBV-Directed IgA Antibody Responses The extent to which the effect of known NPC risk factors is mediated through their ability to modulate immune responses to EBV is not fully understood. We observed that individuals from group 4 also had the highest proportion of current smokers, compared with other groups (72% vs 45%; P = 0.02, by the χ2 test). In addition, when considering the proteome-wide antibody response, the total level of the IgA response but not the IgG response was significantly different between smokers and nonsmokers in our population (P = .002, by the SKAT-C; Figure 4A). Notably, when evaluating anti-EBV antibody responses targeting proteins across different stages of the EBV life cycle, the EBV-directed IgA response to latent and early lytic proteins differed significantly by smoking behavior (Bonferroni-corrected P value of < .05, by the SKAT-C). Figure 4. View largeDownload slide Associations between anti–Epstein-Barr virus (EBV) antibody responses and smoking status among males. A, Association between anti-EBV immunoglobulin A (IgA) antibodies against proteins expressed during different EBV life cycles and smoking status (current vs former/never). P values were obtained from sequence kernel association test–C tests. The dashed line represents the Bonferroni-corrected statistically significant P value threshold of .005. B, Box plot showing the association between LMP1/IgA reactivity and smoking status. The P value was obtained from linear regression, adjusted for age. Figure 4. View largeDownload slide Associations between anti–Epstein-Barr virus (EBV) antibody responses and smoking status among males. A, Association between anti-EBV immunoglobulin A (IgA) antibodies against proteins expressed during different EBV life cycles and smoking status (current vs former/never). P values were obtained from sequence kernel association test–C tests. The dashed line represents the Bonferroni-corrected statistically significant P value threshold of .005. B, Box plot showing the association between LMP1/IgA reactivity and smoking status. The P value was obtained from linear regression, adjusted for age. Further, in logistic regression models that evaluated the association between smoking and each marker individually, we observed suggestive associations (P < .05) for multiple anti-EBV IgA antibodies from all stages of the EBV life cycle (37.9% [22 of 58] were associated with latent proteins, 8.3% [1 of 12] were associated with immediate early proteins, 20.0% [9 of 45] were associated with early lytic proteins, 6.4% [3 of 47] were associated with late lytic proteins, and 21.0% [4 of 19] were associated with glycoproteins; Supplementary Tables 4–8). The strongest individual association was observed between smoking and IgA targeting LMP1(odds ratio, 4.64; P = .001; Figure 4B), an EBV oncoprotein expressed in NPC tumors that is normally considered a weakly immunogenic protein [26, 27]. Taken together, these findings suggest that smoking may alter the anti-EBV mucosal antibody response. We also evaluated the association between anti-EBV antibody responses, alcohol drinking, and HLA alleles that modulate NPC risk (ie, HLA-A*0207 and HLA-A*1101; Supplementary Figures 13A–C). HLA-A*0207 was nominally associated with IgA responses against immediate early lytic proteins (P = .02). However, no statistically significant associations persisted after correction for multiple comparisons. DISCUSSION Our findings represent the first comprehensive evaluation of natural variation in the host antibody response to the full spectrum of EBV proteins. This unique characterization was made possible through the use of protein microarray technology in a population of healthy, asymptomatic adults from Taiwan. Because the antibody repertoire can shed light on exposure to viral proteins, this virus-wide description of host response to the EBV proteome represents an important step forward in understanding population-level variation that might have implications for multiple EBV-related diseases. We found that all adults evaluated in our study had evidence of exposure to EBV proteins, that these EBV-positive adults were more likely to mount IgG rather than IgA antibody responses to EBV, and that responses against glycoproteins were particularly prevalent. If the IgG response targeting 1 EBV protein (eg, protein X) was elevated in an individual, they were more likely to mount an IgG response targeting other antigens, rather than an IgA response against protein X, likely reflecting differential timing of exposure (longer-term exposure for IgG and more-recent exposure for IgA) and separate triggering mechanisms in different compartments (eg, oral mucosa vs systemic circulation). Approximately one quarter of the subjects in this study displayed antibody response patterns similar to NPC cases, including positive responses reflecting exposure to immediate early antigenic targets, such as Zta and Rta. Importantly, our data indicate that smoking is associated with altered IgA antibody patterns that could reflect poor control over EBV lytic activity, suggesting a possible mechanism to explain the role of smoking in the etiology of NPC. Previous studies have evaluated responses against a handful of anti-EBV antibodies by using techniques including immunoblot and microarray chips targeting 10–15 EBV proteins [28, 29]. Zheng et al used a microarray to test 82 EBV open-reading frames in a study that included 10 nondiseased adults and showed that these adults mounted IgG responses against an average of 17 proteins (at 1:100 serum dilution) and IgA antibodies against an average of 13 proteins (at 1:1000 serum dilution) [19]. This is in line with our findings of more frequent IgG reactivity, rather than IgA reactivity. In the present study, we expanded testing to 199 EBV open-reading frames among 289 asymptomatic adults. Because our array was not designed to detect antibodies to conformational epitopes that require glycosylation, future studies will be required to better understand patterns of response to those epitopes and how they are related to those reported herein. Our finding that antibody responses to some glycoproteins were the most prevalent is biologically plausible because these proteins are expressed on the viral surface and therefore readily accessible as targets for the immune system [30, 31]. Given that IgA responses are thought to reflect recent pathogen exposure, our finding that IgA responses to glycoproteins were also more frequently observed than responses against proteins from other stages of the EBV life cycle suggests that the presence of EBV in the oral cavity is not a rare event [32]. It was notable that individuals in groups 2 and 4 (39% [114]) mounted a substantial, proteome-wide IgA response. This supports data from studies reporting EBV viral load shedding in the saliva of the majority of healthy adults measured [32, 33], as well as EBV reactivation rates that are much more frequent than those for other herpesviruses, such as cytomegalovirus [34]. An important public health implication of our data is that smoking, a modifiable behavior, was associated with elevated IgA responses against EBV. An association between smoking and EBV reactivation has been posited by one study, which evaluated VCA/IgA in individuals from southern China [35]. In the present study, we expand beyond VCA/IgA alone to demonstrate large-scale differences in the EBV-directed IgA repertoire in smokers. The most significant smoking association was observed for IgA responses against LMP1, an important NPC oncogene that has cell-transforming ability in rodent fibroblasts [36]. In addition to these individual antibody associations, we observed that individuals in group 4 also had the highest proportion of current smokers. Taken together, these findings suggest that smoking may negatively impacts an individual’s ability to control EBV infection, leading to an increased frequency of viral reactivation/exposure in the host. However, we did not assess the degree to which the established association between smoking and NPC is mediated by interindividual variation in the immune response to this ubiquitous virus. The underlying biological mechanisms merit further investigation. In conclusion, we described the pattern of the IgG and IgA responses to EBV among asymptomatic, healthy adults, using a protein microarray chip targeting 199 EBV protein sequences. Given that the generation of antibodies is a complex process requiring interplay between viral antigens and numerous compartments of the immune system [37], our study represents an initial but important step in understanding the immune response to EBV infection. In the future, our approach could be applied to other populations to characterize the similarities and differences observed in the humoral response to EBV in individuals predisposed to different EBV-associated conditions (eg, sub-Saharan Africans at higher risk of Burkitt lymphoma). Such studies could elucidate subsets of EBV-infected individuals at highest risk of EBV-associated clinical conditions and help identify targets for the development of effective immunodiagnostics. Supplementary Data Supplementary materials are available at The Journal of Infectious Diseases online. Consisting of data provided by the authors to benefit the reader, the posted materials are not copyedited and are the sole responsibility of the authors, so questions or comments should be addressed to the corresponding author. Notes Financial support. This work was supported by the National Cancer Institute Intramural Research Program, the National Health and Medical Research Council of Australia, and the National Science Council of Taiwan. Potential conflicts of interest. J. M. M. received payments as owner and chief executive officer of Cyto-Barr. All other authors report no potential conflicts. All authors have submitted the ICMJE Form for Disclosure of Potential Conflicts of Interest. Conflicts that the editors consider relevant to the content of the manuscript have been disclosed. References 1. Kieff E , Rickinson AB . Epstein-Barr virus and its replication . In Fields virology . 5th ed . ed. Philadelphia, PA : Lippincott Williams & Wilkins , 2007 . 2. Kangro HO , Osman HK , Lau YL , Heath RB , Yeung CY , Ng MH . Seroprevalence of antibodies to human herpesviruses in England and Hong Kong . J Med Virol 1994 ; 43 : 91 – 6 . Google Scholar CrossRef Search ADS PubMed 3. Chen CY , Huang KY , Shen JH , Tsao KC , Huang YC . A large-scale seroprevalence of Epstein-Barr virus in Taiwan . PLoS One 2015 ; 10 : e0115836 . Google Scholar CrossRef Search ADS PubMed 4. Young LS , Rickinson AB . Epstein-Barr virus: 40 years on . Nat Rev Cancer 2004 ; 4 : 757 – 68 . Google Scholar CrossRef Search ADS PubMed 5. Cohen JI , Fauci AS , Varmus H , Nabel GJ . Epstein-Barr virus: an important vaccine target for cancer prevention . Sci Transl Med 2011 ; 3 : 107fs7 . Google Scholar CrossRef Search ADS PubMed 6. Kanekiyo M , Bu W , Joyce MG , et al. Rational design of an Epstein-Barr virus vaccine targeting the receptor-binding site . Cell 2015 ; 162 : 1090 – 100 . Google Scholar CrossRef Search ADS PubMed 7. Hjalgrim H , Friborg J , Melbye M . The epidemiology of EBV and its association with malignant disease . In: Arvin A , Campadelli-Fiume G , Mocarski E et al. , eds. Human herpesviruses: biology, therapy, and immunoprophylaxis . Cambridge: Cambridge University Press , 2007 . Google Scholar CrossRef Search ADS 8. Coghill AE , Hildesheim A . Epstein-Barr virus antibodies and the risk of associated malignancies: review of the literature . Am J Epidemiol 2014 ; 180 : 687 – 95 . Google Scholar CrossRef Search ADS PubMed 9. Melbye M , Ebbesen P , Levine PH , Bennike T . Early primary infection and high Epstein-Barr virus antibody titers in Greenland Eskimos at high risk for nasopharyngeal carcinoma . Int J Cancer 1984 ; 34 : 619 – 23 . Google Scholar CrossRef Search ADS PubMed 10. Henle W , Henle G . Epstein-Barr virus-specific serology in immunologically compromised individuals . Cancer Res 1981 ; 41 : 4222 – 5 . Google Scholar PubMed 11. Brandtzaeg P . Potential of nasopharynx-associated lymphoid tissue for vaccine responses in the airways . Am J Respir Crit Care Med 2011 ; 183 : 1595 – 604 . Google Scholar CrossRef Search ADS PubMed 12. Bhaduri-McIntosh S , Landry ML , Nikiforow S , Rotenberg M , El-Guindy A , Miller G . Serum IgA antibodies to Epstein-Barr virus (EBV) early lytic antigens are present in primary EBV infection . J Infect Dis 2007 ; 195 : 483 – 92 . Google Scholar CrossRef Search ADS PubMed 13. Chien YC , Chen JY , Liu MY , et al. Serologic markers of Epstein-Barr virus infection and nasopharyngeal carcinoma in Taiwanese men . N Engl J Med 2001 ; 345 : 1877 – 82 . Google Scholar CrossRef Search ADS PubMed 14. Cao SM , Liu Z , Jia WH , et al. Fluctuations of Epstein-Barr virus serological antibodies and risk for nasopharyngeal carcinoma: a prospective screening study with a 20-year follow-up . PLoS One 2011 ; 6 : e19100 . Google Scholar CrossRef Search ADS PubMed 15. Henle W , Henle G , Zajac BA , Pearson G , Waubke R , Scriba M . Differential reactivity of human serums with early antigens induced by Epstein-Barr virus . Science 1970 ; 169 : 188 – 90 . Google Scholar CrossRef Search ADS PubMed 16. Ginsburg CM , Henle W , Henle G , Horwitz CA . Infectious mononucleosis in children. Evaluation of Epstein-Barr virus-specific serological data . JAMA 1977 ; 237 : 781 – 5 . Google Scholar CrossRef Search ADS PubMed 17. Doolan DL . Plasmodium immunomics . Int J Parasitol 2011 ; 41 : 3 – 20 . Google Scholar CrossRef Search ADS PubMed 18. Davies DH , Duffy P , Bodmer JL , Felgner PL , Doolan DL . Large screen approaches to identify novel malaria vaccine candidates . Vaccine 2015 ; 33 : 7496 – 505 . Google Scholar CrossRef Search ADS PubMed 19. Zheng D , Wan J , Cho YG , et al. Comparison of humoral immune responses to Epstein-Barr virus and Kaposi’s sarcoma-associated herpesvirus using a viral proteome microarray . J Infect Dis 2011 ; 204 : 1683 – 91 . Google Scholar CrossRef Search ADS PubMed 20. Chen CJ , Yang HI , Su J , et al. ; REVEAL-HBV Study Group . Risk of hepatocellular carcinoma across a biological gradient of serum hepatitis B virus DNA level . JAMA 2006 ; 295 : 65 – 73 . Google Scholar CrossRef Search ADS PubMed 21. Hildesheim A , Apple RJ , Chen CJ , et al. Association of HLA class I and II alleles and extended haplotypes with nasopharyngeal carcinoma in Taiwan . J Natl Cancer Inst 2002 ; 94 : 1780 – 9 . Google Scholar CrossRef Search ADS PubMed 22. Vigil A , Davies DH , Felgner PL . Defining the humoral immune response to infectious agents using high-density protein microarrays . Future Microbiol 2010 ; 5 : 241 – 51 . Google Scholar CrossRef Search ADS PubMed 23. Coghill AE , Pfeiffer RM , Proietti C , et al. Identification of a novel, EBV-based antibody risk stratification signature for early detection of nasopharyngeal carcinoma in Taiwan . Clin Cancer Res 2018 ; 24 : 1305 – 14 . Google Scholar CrossRef Search ADS PubMed 24. Galili T . dendextend: an R package for visualizing, adjusting and comparing trees of hierarchical clustering . Bioinformatics 2015 ; 31 : 3718 – 20 . Google Scholar CrossRef Search ADS PubMed 25. Ionita-Laza I , Lee S , Makarov V , Buxbaum JD , Lin X . Sequence kernel association tests for the combined effect of rare and common variants . Am J Hum Genet 2013 ; 92 : 841 – 53 . Google Scholar CrossRef Search ADS PubMed 26. Dawson CW , Port RJ , Young LS . The role of the EBV-encoded latent membrane proteins LMP1 and LMP2 in the pathogenesis of nasopharyngeal carcinoma (NPC) . Semin Cancer Biol 2012 ; 22 : 144 – 53 . Google Scholar CrossRef Search ADS PubMed 27. Meij P , Vervoort MB , Aarbiou J , et al. Restricted low-level human antibody responses against Epstein-Barr virus (EBV)-encoded latent membrane protein 1 in a subgroup of patients with EBV-associated diseases . J Infect Dis 1999 ; 179 : 1108 – 15 . Google Scholar CrossRef Search ADS PubMed 28. Fachiroh J , Schouten T , Hariwiyanto B , et al. Molecular diversity of Epstein-Barr virus IgG and IgA antibody responses in nasopharyngeal carcinoma: a comparison of Indonesian, Chinese, and European subjects . J Infect Dis 2004 ; 190 : 53 – 62 . Google Scholar CrossRef Search ADS PubMed 29. Loebel M , Eckey M , Sotzny F , et al. Serological profiling of the EBV immune response in Chronic Fatigue Syndrome using a peptide microarray . PLoS One 2017 ; 12 : e0179124 . Google Scholar CrossRef Search ADS PubMed 30. Hutt-Fletcher LM . EBV glycoproteins: where are we now ? Future Virol 2015 ; 10 : 1155 – 62 . Google Scholar CrossRef Search ADS PubMed 31. Li Q , Turk SM , Hutt-Fletcher LM . The Epstein-Barr virus (EBV) BZLF2 gene product associates with the gH and gL homologs of EBV and carries an epitope critical to infection of B cells but not of epithelial cells . J Virol 1995 ; 69 : 3987 – 94 . Google Scholar PubMed 32. Hadinoto V , Shapiro M , Sun CC , Thorley-Lawson DA . The dynamics of EBV shedding implicate a central role for epithelial cells in amplifying viral output . PLoS Pathog 2009 ; 5 : e1000496 . Google Scholar CrossRef Search ADS PubMed 33. Yao QY , Rickinson AB , Epstein MA . A re-examination of the Epstein-Barr virus carrier state in healthy seropositive individuals . Int J Cancer 1985 ; 35 : 35 – 42 . Google Scholar CrossRef Search ADS PubMed 34. Ling PD , Lednicky JA , Keitel WA , et al. The dynamics of herpesvirus and polyomavirus reactivation and shedding in healthy adults: a 14-month longitudinal study . J Infect Dis 2003 ; 187 : 1571 – 80 . Google Scholar CrossRef Search ADS PubMed 35. Xu FH , Xiong D , Xu YF , et al. An epidemiological and molecular study of the relationship between smoking, risk of nasopharyngeal carcinoma, and Epstein-Barr virus activation . J Natl Cancer Inst 2012 ; 104 : 1396 – 410 . Google Scholar CrossRef Search ADS PubMed 36. Wang D , Liebowitz D , Kieff E . An EBV membrane protein expressed in immortalized lymphocytes transforms established rodent cells . Cell 1985 ; 43 : 831 – 40 . Google Scholar CrossRef Search ADS PubMed 37. Nutt SL , Hodgkin PD , Tarlinton DM , Corcoran LM . The generation of antibody-secreting plasma cells . Nat Rev Immunol 2015 ; 15 : 160 – 71 . Google Scholar CrossRef Search ADS PubMed Published by Oxford University Press for the Infectious Diseases Society of America 2018. This work is written by (a) US Government employee(s) and is in the public domain in the US.

Journal

The Journal of Infectious DiseasesOxford University Press

Published: Mar 2, 2018

There are no references for this article.

You’re reading a free preview. Subscribe to read the entire article.


DeepDyve is your
personal research library

It’s your single place to instantly
discover and read the research
that matters to you.

Enjoy affordable access to
over 18 million articles from more than
15,000 peer-reviewed journals.

All for just $49/month

Explore the DeepDyve Library

Search

Query the DeepDyve database, plus search all of PubMed and Google Scholar seamlessly

Organize

Save any article or search result from DeepDyve, PubMed, and Google Scholar... all in one place.

Access

Get unlimited, online access to over 18 million full-text articles from more than 15,000 scientific journals.

Your journals are on DeepDyve

Read from thousands of the leading scholarly journals from SpringerNature, Elsevier, Wiley-Blackwell, Oxford University Press and more.

All the latest content is available, no embargo periods.

See the journals in your area

DeepDyve

Freelancer

DeepDyve

Pro

Price

FREE

$49/month
$360/year

Save searches from
Google Scholar,
PubMed

Create lists to
organize your research

Export lists, citations

Read DeepDyve articles

Abstract access only

Unlimited access to over
18 million full-text articles

Print

20 pages / month

PDF Discount

20% off