Autoantibody Signatures Combined with Epstein–Barr Virus Capsid Antigen-IgA as a Biomarker Panel for the Detection of Nasopharyngeal Carcinoma

Nasopharyngeal carcinoma (NPC) is one of the most common head and neck cancers in Southern China and Northern Africa, and the incidence reaches 20 to 30 per 100,000 men and 15 to 20 per 100,000 women (1). Because patients with early-stage disease are often lacking clinically specific symptoms, NPC tends to present at an advanced stage at diagnosis. For patients with early-stage diseases, the 5-year survival probability is as great as up to 95%; however, this proportion declines to approximately 70% in patients with stage III and IV disease (2, 3). Therefore, it highlights the need to improve the early diagnosis rate of NPC through a screening protocol.

Epstein–Barr virus (EBV) infection is closely related to NPC development. Detection of EBV-DNA and viral capsid antigen immunoglobulin A (VCA-IgA) has been the most commonly used markers for screening the disease; however, specificity and sensitivity of these methods are considered unsatisfactory (4–8). Therefore, NPC screening by EBV-related detection cannot completely meet the requirements of a diagnostic marker, and other markers that may improve diagnostic accuracy should be taken into consideration. Autoantibodies have attracted more attention as possible biomarkers in recent years. A large number of studies describe the presence of autoantibodies to tumor-associated antigens (TAA) in serum samples from patients with a variety of cancers (9–15). Interestingly, autoantibodies seem to appear months to years before the clinical diagnosis of a tumor, which makes measurement of autoantibodies suitable for early-stage cancer diagnosis (9, 14, 15).

In a previous study, we showed that measurement of autoantibodies to at least one of six antigens [p53, NY-ESO-1, matrix metalloproteinase-7 (MMP-7), heat shock protein 70 (Hsp70), peroxiredoxin VI (Prx VI), and BMI1 polycomb ring finger oncogene (Bmi-1)] could differentiate early-stage esophageal squamous cell carcinoma patients from normal controls, with a sensitivity/specificity of at least 45%/95% (16). Recently, we reported that autoantibodies against NY-ESO-1 might be used as a supplement to the traditional EBV markers for screening and diagnosis of NPC; however, the number of the NPC patients in this investigation was relatively small (17). Our results support the viewpoint that measurement of a single autoantibody will not provide the adequate sensitivity or specificity needed for early detection or screening (18). Here, we enlarged the number of participants and assessed the diagnostic accuracy of autoantibodies to the same panel of the six TAAs for NPC. In addition, assay of VCA-IgA was conducted.

We screened 321 patients with NPC consecutively from the Department of Radiation Oncology, the Cancer Hospital of Shantou University Medical College between December 2012 and April 2014. Participants (11 patients) who had a history of other tumors were excluded from the study. Finally, 310 patients with NPC were recruited. A total of 218 healthy volunteers as a control group were enrolled. The patient group and the control group for age and sex were matched as much as possible (Supplementary Table S1). NPC was defined as previously described (17) and was biopsy proven in all poorly differentiated squamous carcinoma type. Tumor stage was defined according to the seventh edition of the UICC/AJCC staging system for NPC (19). The 310 NPC patients and 218 normal controls were randomly assigned to a training set (220 NPC samples vs. 150 normal controls) or to a validation set (90 NPC samples vs. 68 normal controls) on the basis of a computer-generated allocation sequence. We aimed to identify a clinically significant diagnostic autoantibody signature from the training set and tested it in the validation set. Our study size allows for a power of 90% at a significance level of 0.05.

Patients were all newly diagnosed. Vacuum blood collection tubes without anticoagulant (RICH Science) were used to collect peripheral blood from normal controls and NPC patients before treatment. The serum was removed from the tube and stored in 1.7 ml SafeSeal Microcentrifuge Tubes (Sorenson BioScience) at −80°C until use. Prior to the use of these clinical materials for investigation, approval for the study from the institutional ethics review committee and informed consent of patients were obtained. This work was complied with the principles laid down in the Declaration of Helsinki. Data collection and analyses were undertaken by two researchers (L.-S. Huang and T.-T. Zhai).

The recombinant proteins were expressed, purified, and analyzed as previously described (16). Briefly, the coding sequence (CDS) regions for TP53 (NM_001276760.1), CTAG1B (NM_001327.2), PRDX6 (NM_004905.2), BMI1 (NM_005180.8), MMP7 (NM_002423.3), and HSPA1A (NM_005345.5) were subcloned into the pDEST17 expression vector (Invitrogen, cat. no. 11803-012) with a small His tag. The resulting recombinant plasmids were verified by sequencing prior to expression trials. To obtain the recombinant proteins, the expression host E. coli Rosetta (DE3; Novagen, cat. no. 71402-3) was transformed with the recombinant plasmid. Transformed colonies were inoculated in 5 mL LB medium supplemented with 100 μg/mL ampicillin, and cultured overnight at 37°C. Then the cell culture was transferred to 2 L of fresh LB medium. When the optical density (OD) at 600 nm reached 0.4 to 0.6, IPTG (Merck, cat. no. 420322) was added to a final concentration of 0.4 mmol/L to induce expression of recombinant protein at 30°C. After about 3 hours, the cells were harvested and resuspended in PBS buffer supplemented with 8 mol/L urea and 1 mmol/L PMSF. Cell debris was cleared by centrifugation, and the supernatants were incubated on a Ni²⁺-NTA-Sepharose column (Novagen, cat. no. 70666-3). The column was washed with wash buffer A (PBS, 8 mol/L urea, pH 8.0) followed by wash buffer B (PBS). The proteins of interest were eluted with elution buffer (PBS, 500 mmol/L imidazole, pH 7.4) and dialyzed twice against 4 L of 50% glycerol in PBS. A negative control peptide (His tag alone, with the sequence of MSYYHHHHHHLESTSLYKKAG) was synthesized and purified. Protein concentrations were determined using a BCA protein assay (Thermo, cat. no. 23225), and bovine serum albumin was used as a standard. The purity of the recombinant protein was determined by SDS–PAGE and Coomassie Blue staining (Imperial Protein Stain; Thermo, cat. no. 24615). Only proteins that were >95% pure were used in the assays.

Enzyme-linked immunosorbent assay (ELISA) for serum autoantibodies was performed by two researchers (Y.-W. Xu and Y.-H. Peng) who had no access to patient clinical information as previously described (16). Briefly, the optimal antigen-coating concentration and the serum dilution for the ELISA of each autoantibody test were determined using a checkerboard titration in preliminary studies. Purified recombinant antigens, p53, NY-ESO-1, MMP-7, Hsp70, Prx VI, and Bmi-1, were diluted to a final protein concentration of 0.1, 0.1, 0.6, 0.8, 1.5, and 0.6 μg/mL, respectively. The antigen dilutions were dispensed in 100 μL per well volumes into 96-well microtiter plates (Biohaotian, cat. no. HT081) and incubated overnight at 4°C. Negative control antigens consisting of the purified His tags were also included to validate antibody-binding specificity. The plates were washed with PBST (PBS containing 0.05% Tween 20), and then blocked with a blocking buffer (PBST containing 1% BSA). After washing, 100 μL of serum samples and quality control samples (QCS, a pooled plasma sample collected randomly from 50 patients with NPC) were diluted 1 of 110 in blocking buffer, then were added to the plates and incubated at 37°C for 1 hour, as well as appropriate control rabbit polyclonal antibodies specific for capture proteins (rabbit anti-His tag antibody, Sigma, SAB1306084; rabbit anti-p53 polyclonal antibody, Immunosoft, IS0001; rabbit anti-NY-ESO-1 polyclonal antibody, Immunosoft, IS0022; rabbit anti-MMP-7 polyclonal antibody, Immunosoft, IS0125; rabbit anti-Hsp70 polyclonal antibody, Immunosoft, IS0033; rabbit anti-Prx VI polyclonal antibody, Immunosoft, IS0272; rabbit anti-Bmi-1 polyclonal antibody, Immunosoft, IS0079). One hundred microliters of blocking buffer as negative control was also added. After washing, horseradish peroxidase (HRP)-conjugated goat anti-human IgG (Santa Cruz Biotechnology, sc-2907) or anti-rabbit IgG (Santa Cruz Biotechnology, sc-2054) were used as secondary antibodies at the dilution recommended by the manufacturer (1:10,000). After incubation, the plates were washed and ready prepared 3,3′,5,5′-tetramethylbenzidine (TMB; InTec PRODUCTS) and hydrogen peroxide (InTec PRODUCTS) were added. Color formation was allowed to proceed at 37°C for 15 minutes and then stopped with 50 μL of 0.5 mol/L H₂SO₄. The absorbance of each well was read at 450 nm and referenced to 630 nm within 5 minutes by a plate microplate reader (Multiskan MK3; Thermo Fisher Scientific).

All cancer and normal samples were interspersed on the plates and run in duplicate. The intraassay coefficients of variation (CV) for autoantibodies against p53, NY-ESO-1, Prx VI, Bmi-1, MMP7, and Hsp70 were 7.3%, 7.6%, 5.9%, 9.2%, 7.1%, and 8.9%, respectively, and the interassay CVs were 9.1%, 9.0%, 7.4%, 9.8%, 8.9%, and 9.7%, respectively. QCSs were run to ensure quality control monitoring of the assay runs by using Levey–Jennings plots. With the purpose of minimizing an intraassay deviation, the ratio of the difference between duplicated sample OD values to their sum was used to assess precision of the assay. If the ratio was >10%, the test of this sample was treated as being invalid and the sample was repeated.

Concentrations of VCA-IgA in all samples were determined in duplicate by ELISA using commercial kits (Berer Bioengineering, cat. no. 3400638) as previous described (17). The experiments were conducted by the manufacturer's instructions. Briefly, negative control, positive control, and each serum sample at a 1:10 dilution were added to the plates. After washing, 100 μL of HRP-conjugated anti-human IgA was added into each well. Color development and the measurement of the absorbance were performed according to the above method of autoantibody detection.

All analyses were done with SPSS (version 17.0) or GraphPad Prism software. All statistical data are expressed as mean ± SE. The comparison of the different markers between two independent groups was done with the use of the independent samples t test. Where differences were identified by F test, independent samples t tests were adjusted for unequal variances (Mann–Whitney U test). The cutoff value of VCA-IgA assay was calculated by following the manufacturer's suggestion. For individual autoantibodies and combined marker performance, receiver operating characteristic (ROC) analysis was performed to work out optimum cutoff value, sensitivity, specificity, and area under the ROC curve (AUC) with 95% confidence interval (CI). The optimum cutoff value for positive reactivity was determined by achieving the maximum sensitivity when the specificity was >90%, and by minimizing the distance of the cutoff value to the top-left corner of the ROC curve. A specificity of >90% was selected in order to produce a test which could be suitable for the aim of early detection and which would be health economically viable (10). To test the diagnostic accuracy when the different markers were combined, we estimated functions of the combined markers by binary logistic regression, and the values of these functions were used as one marker and subjected to ROC analysis (20). For example, to estimate the diagnostic accuracy of the combined use of the six autoantibody markers, a variable predicted probability (p) for NPC was created on the basis of an equation obtained by binary logistic regression with “Enter” method (all NPC vs. all controls in the test cohort): ln[p/(1 − p)] = 12.625 × (p53) + 10.248 × (NY-ESO-1) − 2.986 × (Prx-6) + 7.123 × (Bmi-1) − 4.136 × (MMP7) + 13.850 × (Hsp70) − 2.785. The regression equations and optimum predicted probabilities (i.e., optimum cutoff values) for the combinations of different markers are provided in the Supplementary Table S2. The positive predictive value (PPV), negative predictive value (NPV), positive likelihood ratio (PLR), and negative likelihood ratio (NLR) were presented to improve clinical interpretation. In all tests, we considered P values lower than 0.05 (two-sided) to be significant.

The presence of autoantibodies to all TAAs in sera samples obtained from NPC patients and normal controls is shown for one concentration of antigen in the scatter plots in Fig. 1. The levels of autoantibodies for all six TAAs were significantly higher in patients with NPC compared with those in normal controls (Fig. 1, all P < 0.0001). The ROC curves for individual autoantibodies that discriminate the cancer groups from normal controls are shown in Fig. 2. In the training set, ROC curves showed the optimum cutoff values for autoantibodies against p53, NY-ESO-1, Prx VI, Bmi-1, MMP7, and Hsp70 were 0.107, 0.189, 0.170, 0.143, 0.230, and 0.118, respectively, and the AUC values for individual autoantibody assays ranged between 0.574 (MMP-7) and 0.781 (p53), with the sensitivity values in individual autoantibody assays ranging between 21.4% (MMP-7) and 46.8% (p53; Table 1). To verify the diagnostic accuracy of these autoantibodies, we further confirmed the results in the validation set (Table 1).

We observed similar results in the early-stage NPC patients to those in all of the NPC patients (Fig. 2, Table 1). Interestingly, whether in the training set or in the validation set, the discriminatory power of autoantibodies to p53 and NY-ESO-1 in early-stage NPC gave larger AUC values with better sensitivity compared with those in all NPC patients (Table 1).

The ability of the autoantibody panel to correctly identify NPC is shown in Table 2 and graphically constructed by ROC curve in Fig. 3. In the training set, use of these six autoantibody assays provided an AUC value of 0.855 (95% CI, 0.818–0.892), with a sensitivity/specificity of 68.2%/90.0%. Moreover, we examined whether different combinations of these autoantibodies would gain similar diagnostic capacity with the panel of six autoantibody assays, and we found that a restricted panel consisting of the four antigens p53, NY-ESO-1, Bmi-1, and Hsp70 exhibited only a slightly reduced sensitivity of 65.0% with the same specificity (AUC = 0.848; 95% CI, 0.810–0.886). Testing the autoantibody panel of four derived from the training set on the validation sample set gave an AUC value of 0.862 (95% CI, 0.807–0.916), a sensitivity/specificity of 61.1%/91.2%. Therefore, the simplified four-autoantibody panel also had considerable discriminatory power.

According to the manufacturer's instructions, the recommended clinical cutoff value of VCA-IgA was 0.150. The sensitivity/specificity of VCA-IgA in the training set and validation set were 45.0%/96.7% and 52.2%/92.6%, respectively. ROC analysis illustrated that measurement of both autoantibody panel and VCA-IgA increased the diagnostic accuracy for NPC, compared with the test of the autoantibody panel alone (Fig. 3). In the training set, this combined testing yielded a sensitivity and specificity of 81.4% and 90.0%, respectively, with an enhanced AUC of 0.911 (95% CI, 0.881–0.940; Table 2). When the restricted panel (p53, NY-ESO-1, Bmi-1, and Hsp70) was combined with VCA-IgA, the diagnostic capacity for NPC was similar to the combination of the autoantibody panel and VCA-IgA (Table 2, Fig. 3). These results were further confirmed in the validation set (Table 2, Fig. 3). Moreover, in the training set, 82 (67.8%) of 121 VCA-IgA–negative patients with NPC had positive antoantibody panel results, and similar results were obtained in the validation set (Table 3). ROC curve analysis of the autoantibody panel indicated a diagnosis of NPC regardless of VCA-IgA status (Table 3).

In patients with early-stage NPC, the AUC for the autoantibody panel was 0.893 (95% CI, 0.825–0.961) in the training set and 0.897 (95% CI, 0.793–1.000) in the validation set. The combination of the autoantibody panel and VCA-IgA could discriminate early-stage NPC from normal controls with greater AUC, sensitivity, and specificity compared with the autoantibody panel used alone (Table 2). The predictive values and likelihood ratios for the combination were also better than those for the autoantibody panel alone (Table 2). Interestingly, restriction of the panel to the presence of p53, NY-ESO-1, Bim-1, and Hsp70 autoantibodies exhibited almost the same diagnostic performance for early-stage NPC (Fig. 3, Table 2).

We next assessed the correlation of individual autoantibody assays and VCA-IgA with clinical variables in NPC patients in both sets. We found that individual autoantibodies or VCA-IgA did not significantly differ with age, gender, smoking status, T stage, N stage, or overall stage (Supplementary Tables S3 and S4).

Although the EBV serology tests (such as VCA-IgA) are used for NPC for many years, their usefulness as general screening test for NPC remains unsatisfactory because of problems with either low specificity or low sensitivity (4–6). In this study, we have demonstrated that the detection of autoantibodies to selected TAAs in the peripheral blood has diagnostic potential for NPC, especially for patients with VCA-IgA–negative status and early-stage patients, and its value in the training set was confirmed in a validation set. The diagnostic value of the autoantibody panel for NPC was similar with our published data from the same panel in esophageal squamous cell carcinoma (16). In patients with early-stage NPC, use of the autoantibody panel provided a better diagnostic efficacy (Table 3). This indicates that the autoantibody panel may be a promising marker for the early detection of NPC and that the induction of autoantibodies occurs early in the process of carcinogenesis, which is consistent with previous studies (9–17). Analysis of VCA-IgA presented here is also shown to be a useful diagnostic biomarker of NPC, which is in agreement with a study by Ai and colleagues (21), but the high false-negative rates will prevent the timely diagnosis of NPC patients in endemic areas, particularly the symptomless, early-stage patients. Thus, measurement of the autoantibody panel could improve results in VCA-IgA–negative patients. Indeed, we found that 67.8% and 67.4% VCA-IgA–negative patients with NPC in the training set and the validation set, respectively, had positive autoantibody panel results.

Compelling evidence has emerged in recent years, suggesting that effective and accurate detection of cancer, particularly early-stage cancer, will likely depend on the combination of a number of biomarkers generated from different mechanism that have greater sensitivity and specificity than each biomarker alone (22–25). Xie and colleagues (25) found that a multiplex assay combining autoantibodies plus PSA in differentiating prostate cancer from nonmalignant conditions achieved an AUC of 0.910, with a sensitivity of 79%, a specificity of 84%, whereas the AUC for PSA alone was decreased to 0.66, with a reduced sensitivity/specificity of 52%/79%, respectively. A previous study by ours showed that the combination of autoantibodies against NY-ESO-1 and VCA-IgA yielded an enhanced sensitivity (17). These publications have highlighted the potential value of the combination of autoantibodies with a conventional cancer biomarker for the diagnosis of cancer. In this study, when we evaluated the diagnostic accuracy of the combination of the autoantibody panel and VCA-IgA, we observed that this combination allowed us to significantly distinguish NPC from normal controls, with a larger AUC compared with the autoantibody panel used alone (Fig. 3, Table 2). The sensitivity and specificity were high, and the predictive values and likelihood ratios were satisfactory for the diagnosis of NPC, including for early-stage disease (Table 2). Therefore, combined analysis of our autoantibody panel and VCA-IgA improves the diagnostic performance, especially for early-stage NPC.

The potential clinical significance of autoantibodies for screening and early detection seems to rely on searching an optimized panel of autoantibodies (16, 26, 27). Previous study, investigating the presence of autoantibodies to these six antigens in esophageal cancer, demonstrates that there was some overlap of reactivity between the various markers and that measurement of the autoantibody responses to p53 and NY-ESO-1 antigens was integral to the panel assays as reported here (16). This emphasizes that the combination of an optimized autoantibody panel has to be selective for high sensitivity and specificity. Thus, in this study, we intended to find whether some markers could be removed without greatly affecting the diagnostic sensitivity of NPC. By performing logistic regressions, we produced different models using different combinations of the potential autoantibody biomarkers and found that the restricted panel (p53, NY-ESO-1, Bmi-1, and Hsp70) remained highly sensitive and specific in detecting samples from all NPC patients and early-stage patients, as well as in the case of combination of VCA-IgA (Fig. 3, Table 2). Taking into account the cost–benefit and patients' financial burden, the restricted panel combined with VCA-IgA is acceptable for screening of NPC. Our data presented here further underline that it is important to select the right combinations of markers to achieve maximal sensitivity and specificity. Increased sensitivity is likely to depend on replacing the antigens used in the panel (such as MMP7 or Prx VI) with other NPC-specific antigens in a panel assay. In the future, we would search new autoantibody markers that could enhance the diagnostic efficiency of our present panel, with the use of proteomics-based technologies, which allow for the identification of autoantibodies to TAAs (28–30).

We expect that our TAAs panel combined with VCA-IgA may be used as a novel noninvasive approach to screen NPC in regions with high incidence rates. However, we observed that some TAAs, such as p53 and NY-ESO-1 used in the current panel assay, were also detected in several types of cancer, including lung, breast, and esophageal cancer (9, 10, 14, 16). This reveals autoantibody detection may not be suitable for distinguishing one type of cancer from another. Hence, if measurement of our autoantibody panel is used as a screen for NPC, persons who are identified without NPC but with positive results should be considered to have the possibility of suffering other types of cancer. In addition, the HRP-conjugated goat anti-human IgG used in this study was raised against the whole IgG molecule (not only the Fc region), which means that the antibody might cross-react with other heavy chain classes (such as IgA and IgM antibodies). Therefore, it is possible that assay signals obtained in the ELISA could stem from patient IgG or other classes of antibody. In the future, we will consider using peroxidase-conjugated antibodies specific to the heavy chain constant regions as secondary antibody to increase assay specificity by eliminating assay signals relevant to different antibody classes.

In conclusion, this study shows an autoantibody assay using a panel of specific autoantibodies is able to help to resolve the deficiencies of VCA-IgA in the detection of VCA-IgA–negative NPC patients. More importantly, we provide a novel platform that integrates autoantibody signatures with a traditional EBV marker (i.e., VCA-IgA), which may aid early detection of NPC. However, the diagnostic values of autoantibody signatures with or without the combination of VCA-IgA in early-stage NPC were obviously different in the training set and the validation set, which may be due to the relatively small number of the early-stage NPC patients in this study, especially in the validation set (only 10 early patients). Therefore, new studies using larger cohorts of early patient need to be performed across multiple populations before clinical development to determine whether NPC could be prospectively detected with the use of this platform.

No potential conflicts of interest were disclosed.

Conception and design: Y.-H. Peng, Y.-W. Xu, E.-M. Li, L.-Y. Xu

Development of methodology: Y.-H. Peng, Y.-W. Xu

Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): Y.-H. Peng, Y.-W. Xu, T.-T. Zhai, L.-H. Dai, S.-Q. Qiu, Y.-S. Yang, W.-Z. Chen

Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): Y.-H. Peng, Y.-W. Xu, L.-S. Huang, T.-T. Zhai, L.-Q. Zhang

Writing, review, and/or revision of the manuscript: Y.-H. Peng, Y.-W. Xu

Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): L.-H. Dai

Study supervision: L.-H. Dai, E.-M. Li, L.-Y. Xu

This work was supported by grants from the National High Technology Research and Development Program of China (No. 2012AA02A503, to E.M. Li; No. 2012AA02A209, to E.M. Li), the Natural Science Foundation of China-GuangDong Joint Fund (No. U0932001, to E.M. Li; No. U1301227, to L.Y. Xu), the Natural Science Foundation of China (No. 81172264, to L.Y. Xu; No. 81472613, to E.M. Li), and the Science and Technology Program of Guangdong (No. 2013B021800250, to Y.H. Peng, Y.W. Xu, S.Q. Qiu, and Y.S. Yang).

The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.