Epstein-Barr virus and human papillomavirus serum antibodies define the viral status of nasopharyngeal carcinoma in a low endemic country

Epstein-Barr virus (EBV) causes nasopharyngeal carcinoma (NPC) in endemic regions, where almost every tumor is EBV-positive. In Western populations, NPC is rare, and human papillomavirus infection (HPV) has been suggested as another viral cause. We validated multiplex serology with molecular tumor markers, to define EBV-positive, HPV-positive and EBV-/HPV-negative NPCs in the United Kingdom, and analyzed survival differences between those groups. Sera from NPC cases (n = 98) and age-and sex-matched controls (n = 142) from the Head and Neck 5000 clinical cohort study were analyzed. IgA and IgG serum antibodies against 13 EBV antigens were measured and compared with EBER in situ hybridization (EBER-ISH) data of 41 NPC tumors (29 EBER-ISH positive, 12 negative). IgG antibodies to EBV LF2 correctly diagnosed EBV-positive NPCs in 28 of 29 cases, while all EBER-ISH negative NPCs were seronegative to LF2 IgG (specificity = 100%, sensitivity = 97%). HPV early antigen serology was compared to HPV molecular markers (p16 expression, HPV DNA and RNA) available for 41 NPCs (13 positive, 28 negative). Serology matched molecular HPV markers in all but one case (specificity = 100%, sensitivity = 92%). EBV and HPV infections were mutually exclusive. Overall, 67% of the analyzed NPCs were defined as EBV-positive, 18% as HPV-positive and 14% as EBV/HPV-negative. There was no statistical evidence of a difference in survival between the three groups. These data provide evidence that both, EBV-positive and HPV-positive NPCs are present in a low incidence country, and that EBV and HPV serum antibodies correlate with the viral status of the tumor. and HPV-positive NPCs are present in a low-incidence country, and that EBV and HPV serum antibodies correlate with the viral status of the tumor.


| INTRODUCTION
Nasopharyngeal carcinoma (NPC) is a squamous cell cancer arising from the lining of the nasopharynx. The incidence of nasopharyngeal cancer varies widely across the world. Globally, the incidence rate is below one per 100 000 person-years. 1 However, incidence rates in NPC endemic regions, including Southeast Asia, North Africa, China and the Arctic, are up to 30 times higher. 2 In high-incidence or endemic regions, Epstein-Barr-virus (EBV) is uniformly associated with NPC development. 3 Elevated antibody levels to EBV proteins in NPC patients have been described since 1966 3 in many case-control [4][5][6] and prospective studies 7,8 and are a useful tool for diagnosis and disease prediction in endemic populations. Until recently, these serological analyses were limited to IgA antibody levels against the viral capsid antigen (VCA), EBV nuclear antigen 1 (EBNA1) and early antigen diffuse (EA-D). In 2018, Coghill et al described a novel EBV antibody risk stratification signature, consisting of 14 IgA and IgG antibodies against diverse EBV proteins for the prediction of NPC development in Taiwan, a high-incidence region. 9 This model for NPC prediction showed an accuracy of 93% for detecting NPCs, compared with an accuracy of 82% for VCAp18/ EBNA1 IgA biomarkers alone (P < .01).
In low-incidence regions, other risk factors besides EBV infection have been described to be associated with NPC development. 10 The role of smoking as a risk factor has been examined in several studies in both, low incidence and endemic regions, and the risk associated with smoking has been shown to be higher in low incidence countries. 11 More recently, several studies also described human papillomavirus (HPV) as a risk factor for NPCs in low incidence regions, although the causal link remains under debate. [12][13][14][15] HPV, especially type 16, is established as a causal agent for the development of oropharyngeal carcinoma (OPC), 16,17 while the prevalence of HPV-driven head and neck tumors outside the oropharynx is low. 18 However, NPCs and OPCs share properties that distinguish them from other head and neck cancers, like their predominantly viral etiology, early age at disease onset, and similar clinical characteristics, especially early lymph node involvement. 19 In OPCs, early antigen HPV serology (especially antibodies against the HPV16 oncoprotein E6) has been shown to be very strongly associated with molecularly defined HPV-positive OPCs, both at diagnosis and prospectively. 17,20,21 HPV detection in NPCs has been limited to p16 immunohistochemistry (IHC) and HPV DNA detection by PCR or in situ hybridization (ISH). [12][13][14][15] Currently, no serological assay has been validated for detecting HPV-positive NPCs.
The aim of our study was thus to examine whether EBV and HPV serum antibodies or antibody pattern can differentiate NPCs associated with EBV, with HPV, and those not associated with either virus, and to describe the roles of EBV and HPV infection in the development of NPCs in a low incidence region. These data have the potential to make important contributions to the clinical management of NPCs.
While screening approaches are not reasonable in the United Kingdom, due to low NPC incidence, the viral status of the tumor could reveal potential survival differences and underline the need for specific treatment regimens. HPV-driven OPCs have been shown to have a much better survival than HPV-negative OPCs, and can be identified by HPV16-specific antibodies. 22 Existing analyses of survival of EBV-positive, HPV-positive and EBV/HPV-negative NPCs are sparse and contradictory 13,23 and need to be further investigated.
To this end, we examined the molecular and serological EBV and HPV status of NPC cases from the United Kingdom, a country with low NPC incidence (0.27 per 100 000 24 ). The analysis was based on the Head and Neck 5000 clinical cohort study, 25 and included 98 incident NPC cases and 142 age-and sex-matched laryngeal squamous cell carcinoma cases (LSCCs) as controls. We used molecular analyses to validate serology for both, EBV and HPV, and calculated sensitivities and specificities for NPC diagnosis. Moreover, we defined three groups, NPCs associated with either EBV infection or HPV infection, and EBV/HPV-negative NPCs, and compared risk factors, histological subtypes and survival.

| Study population
The Head and Neck 5000 clinical cohort study has been described in detail elsewhere. 25

| EBV serology
Serological testing of blood taken at diagnosis was performed with multiplex serology, a high-throughput assay for simultaneous detection of serum antibodies against a large number of antigens. 32 Testing for EBV antibodies included separate IgA and IgG detection, as indicated by a recently published NPC risk stratification signature. 9 Sera were preincubated at 1:50 dilution for IgA testing (final dilution 1:100) and at 1:5000 dilution for IgG testing (final dilution 1:10 000) in a serum preincubation buffer based on PBS with 2 mg/mL casein and additionally containing 2 g/L of lysate proteins of Escherichia coli overexpressing glutathione-S-transferase (GST)-tag, 5 g/L polyvinyl alcohol and 8 g/L polyvinyl-pyrrolidone. 33 EBV serology was based on 13 antigens. VCAp18, EBNA1 peptide (pep), EBNA1 truncated (trunc), ZEBRA and EA-D were previously validated for multiplex serology. 34 The remaining eight antigens (BXLF1,  Overall, 12 HPV16-positive cases were defined with the standard definition, and five cases were defined as HPV18-positive using the extended definition.

| Statistics
Statistical tests for categorical analyses included chi-square test, and Fisher's exact test for small (n < 6) cell counts. t-Test was used to compare the mean age of the case and control groups. Differences in MFI values of cases and controls were calculated by Mann-Whitney test.
Smoking was categorized as never, former or current, and alcohol consumption was analyzed in the categories nondrinker, moderate, hazardous or harmful. Detailed information on smoking and alcohol history was obtained at baseline via a self-reported questionnaire as described before. 35 Since there was only one harmful drinker among the NPC cases, the "harmful" and "hazardous" categories were combined in subsequent analyses (Table 1).
Socioeconomic status was categorized in five categories from "1-least deprived" to "5-most deprived" based on the English Index of Multiple Deprivation (IMD) 2010 quintiles using participants' home postcode. 36 This area-based index of deprivation is derived from measures of income, education, crime and barriers to housing.
We used receiver operating characteristic (ROC) analysis of 41 cases with EBER-ISH status to define antigen-specific cut-offs for the EBV antigens based on ≥90% specificity. Resulting cut-off values are listed in Table S1. The technical minimum cut-off above assay background is 30 MFI for IgA at 1:100 dilution and IgG at 1:10 000 dilution, and 50 MFI for the triple isotype-specific (IgG/IgM/IgA) antibody at 1:100 dilution.
Odds ratios for the association of EBV antibodies with NPC were calculated using unconditional logistic regression with 95% confidence intervals (CI) adjusted for age, sex, smoking and alcohol consumption.
All NPC cases and controls were included in regression models, and EBV antigens were treated as binary variables, based on the cut-offs described above. 3 | RESULTS

| Participant characteristics
The nasopharyngeal carcinoma cases (n = 98) and the laryngeal squamous cell carcinoma cases used as a control group (n = 142) did not differ in mean age, sex and socioeconomic status (Table 1). However, the proportion of never smokers and nondrinkers was significantly higher in NPC cases than among LSCCs (33% vs 5%, P < .01, and 41% vs 30%, P = .01, respectively).

| EBV serology validation
All case and control sera were tested for the presence of 26 EBV antibody markers, including IgA and IgG antibodies for 13 antigens. IgA antibody responses to all 13 antigens and IgG antibody responses to all antigens except one (VCAp18) were significantly higher among cases than controls ( Figure S1).  (Figure 1 and Table S2).  (Table S1). Comparing IgA and IgG antibody responses, a higher sensitivity of IgG antibodies was observed for 9 out of 13 antigens, whereas IgA antibodies were more sensitive for four antigens; the latter was particularly evident for VCAp18 (66% for IgA, 7% for IgG).
Examining individual antibody performances, LF2 IgG antibodies were able to differentiate EBV-positive NPCs from EBV-negative NPCs with 97% sensitivity and 100% specificity ( Figure 2 IgA. Seroprevalence of LF2 IgG in control LSCCs was only 3% ( Figure 2). The second and third best-performing antibodies were BGLF2 IgG with a sensitivity of 97% at 92% specificity, and BXLF1 F I G U R E 1 Odds ratios (OR) and 95% confidence intervals (CI) adjusted for age, gender, smoking and alcohol consumption for the association of EBV IgA and IgG antibodies with NPC F I G U R E 2 Median fluorescence intensities (MFI) for antibody responses against LF2 IgG, stratified by EBER-ISH status for 98 NPC cases (n = 29 positive, n = 12 negative, n = 57 missing) and control laryngeal squamous cell cancer (LSCC; n = 142). The dotted line indicates the seropositivity cut-off (30 MFI) based on receiver operator characteristic (ROC) analysis of NPC cases with available EBER-ISH data F I G U R E 3 Heatmap showing EBV and HPV serostatus as well as molecular markers (EBER-ISH and HPV DNA, HPV RNA and p16 IHC) of all 98 NPC cases. Positivity is shown in green, negativity in red, and black indicates nonavailable data. EBV serology is defined as seropositivity to LF2 IgG antibodies. HPV serology is defined as either HPV16 E6 > 1000 MFI, or positivity to three out of four early antigens of HPV16 or of HPV18, or positivity to two type-concordant early antigens (E6 and E7) of HPV 31, 33, 35, 45, 52 or 58 IgG (93% and 92%, respectively). However, 10% and 17% of the control LSCCs were seropositive for BGLF2 and BXLF1 IgG, respectively.
BGLF2 IgA antibodies showed the strongest association with NPC among all cases (Table S2), based on high specificity (only 1% of control LSCCs were seropositive) but with low sensitivity (only 47% of NPC cases were seropositive; Table S2). Even when restricting the analysis to EBER-ISH-positive cases, the sensitivity of BGLF2 IgA antibodies was much lower (76%) than for BGLF2 IgG and LF2 IgG (both 97%).
Combining several biomarkers did not improve classification of EBV-positive NPCs compared to single markers in this dataset. The combination of the two best stand-alone markers (positivity for at least one of the two), LF2 IgG and BGLF2 IgG, yielded a slightly higher  sensitivity of 100% (vs 97% for the best marker LF2 IgG alone) but also a lower specificity of 92% compared to 100% for LF2 IgG alone.
Combining LF2 IgG and BGLF2 IgG, the positivity for control LSCCs was 12% and thus higher than for LF2 IgG (3%) or BGLF2 IgG (10%) alone. The gain of sensitivity of the biomarker combination is thus lower than the loss of specificity for both, EBV negative NPCs and control LSCCs in our study.
Based on these data, IgG antibodies to LF2 were the best standalone marker for EBV-positive NPCs in the United Kingdom as a nonendemic region within the Head and Neck 5000 study. Thus, LF2 IgG serology was used further to define EBV-positive NPCs by serology in the subsequent analyses described below.

| HPV serology validation
All case and control sera were tested for the presence of antibodies against high-risk HPV types. No significant differences were observed for HPV antibody levels between cases and controls (data not shown).
Comparing survival stratified by WHO histology classification, there was no statistical evidence of a difference in survival between WHO type I, II and III tumors (P = .51; Figure S3).

| DISCUSSION
Our analysis of 98 NPC cases from the United Kingdom, a low incidence region, showed that 67% of all NPCs were EBV-positive, 18% were HPV-positive and 14% were not associated with either of these viral infections. This finding is in accordance with the existing, yet sparse literature, showing that 60% to 76% of NPCs in low-incidence regions are EBV-positive and 9% to 16% are HPV-positive. [12][13][14][15] For the first time, we have adapted a previously developed large EBV antigen panel for multiplex serology, and showed that IgG antibodies against LF2 are sufficient to define EBV-positive NPCs. In addition, we provided evidence that the comprehensive HPV multiplex serology panel, established to determine molecular HPV status in OPCs, also reliably identifies HPV tumor status of NPCs.
Our study was based on the Head and Neck 5000 prospective clinical cohort study, including more than 5000 participants with head and neck cancer. Although the number of NPC cases was limited to 98 due to the low incidence of NPCs in this nonendemic region, this likely represents one of the largest NPC case series from a Western country. However, to validate LF2 IgG antibodies as a reliable stand-alone biomarker and to investigate survival differences between EBV/HPV-positive NPCs and those not associated with either viral infection, larger and prospective studies are needed. The inclusion of other markers for defining EBV-positive NPCs, for example, BGLF2 IgG, has to be re-evaluated in other studies with larger case numbers.
One limitation of our study is the use of LSCCs as a control group.
Tobacco smoking and alcohol consumption are the largest risk factors for LSCCs, and create an imbalance between cases and controls, which we have adjusted for in our analyses. Since we only used the laryngeal cases to check whether antibodies distinguish between NPCs and non-NPCs, they display a suitable control group, as LSCCs are not known to be associated with EBV infection. 27 LSCC is rarely (<5%) 18 associated with HPV infection, however, we have not calculated risk estimates for HPV serology that could have been affected by HPV-positive LSCCs. Among people with head and neck cancer, LSCCs are the most anatomically different and distant to NPCs, which rules out misclassification bias. There is potential misclassification between LSCCs and OPCs, which however does not affect our analysis, as we expect both, LSCCs and OPCs, to be EBV-negative.
As a strength of our study, we requested extended histopathology reports and cross-checked information between those against the clinical data, to confirm tumor origin and rule out anatomical site misclassification as far as possible, which has been discussed as a potential explanation for HPV-positive NPCs in the past. 14 The extended antigen panel we used for EBV serology is based on a previous study that described differing IgA and IgG antibody responses in NPC cases and controls from Taiwan, thus reflecting antibody pattern in a high incidence region. 9 All antigens performed well in differentiating NPC cases and controls from the United Kingdom, where NPC incidence is low. Although a lower proportion of NPCs in low incidence regions are associated with EBV infection, the EBVpositive NPCs seem to be broadly serologically similar to NPCs from high incidence regions. However, IgG antibodies to one antigen, LF2, were sufficient to differentiate between EBV-positive and EBVnegative NPCs in our study, rather than a panel of 14 IgA and IgG antibodies. 9 These findings need to be confirmed with larger case numbers.  Little is known about the biological role of LF2 in NPC development. The EBV protein interaction map shows that LF2 binds exclusively to one other EBV protein, Rta (ie, BRLF1). 40  As suggested by Lo et al, 49 we provide additional evidence that HPV-positive NPCs are associated with keratinizing (WHO type I) NPCs. Of four keratinizing NPCs, three were positive for HPV, and the remaining one was negative for both EBV and HPV.
Keratinizing NPCs have also been associated with smoking and alcohol use, a shared characteristic with other head and neck squamous cell carcinomas. 49 Of the four keratinizing NPCs we report, two are former smokers and reported hazardous alcohol consumption; one never smoked, but reported hazardous alcohol consumption; and one was a former smoker, but nondrinker; in summary, tobacco and alcohol exposure seems very high in this population.
Comparing the survival of the WHO type I, II and III cases, we did not observe a difference in survival between these histological subtypes ( Figure S3). Larger studies are needed for a reliable comparison of survival data.
There is no statistical evidence for a difference in smoking between EBV-positive, HPV-positive and EBV/HPV-negative NPCs.
However, it is noticeable that there is no current smoker in the EBV/HPV-negative NPC group. Since EBV infection and smoking have been described as independent risk factors, 50 we expected more current or former smokers in the group of EBV/HPV-negative NPCs.
Instead, we observed 55% of former smokers in this group (compared to 50% for HPV-positive and 44% for EBV-positive NPC) and not a single current smoker, while 23% of EBV-positive and 30% of HPVpositive NPC cases were smoking at the time of diagnosis. In consequence, NPCs which are neither associated with EBV, HPV and smoking may be caused by other risk factors, which may include other viral infections not considered in our study. The absence of any smoker in the group of EBV/HPV-negative NPCs could also be based on reporting bias. However, in our study, data on smoking is incomplete and our case groups are too small to draw final conclusions.
In summary, both EBV and HPV serology were included in our analysis and validated with molecular tumor markers, and we have shown that EBV and HPV serum antibodies correlate with the viral status of NPC tumors. Individual serum antibodies or antibody patterns represent an attractive, little invasive diagnostic marker that does not require tumor tissue. The methods we presented here should be applied in further case/control and prospective studies to confirm results with larger case numbers (especially for HPVpositive NPCs), healthy control groups, and prospectively collected serum samples.