Preoperative serum immunoglobulin G and A antibodies to Porphyromonas gingivalis are potential serum biomarkers for the diagnosis and prognosis of esophageal squamous cell carcinoma

Background The key-stone-pathogen, Porphyromonas gingivalis associates not only with periodontal diseases but with a variety of other chronic diseases such as cancer. We previously reported an association between the presence of Porphyromonas gingivalis in esophageal squamous cell carcinoma (ESCC) and its progression. We now report the diagnostic and prognostic potential of serum immunoglobulin G and A antibodies (IgG/A) against Porphyromonas gingivalis for ESCC. Methods An enzyme-linked immunosorbent assay (ELISA) was used to determine the serum levels of Porphyromonas gingivalis IgG and IgA in 96 cases with ESCC, 50 cases with esophagitis and 80 healthy controls. Results The median serum levels of IgG and IgA for P. gingivalis were significantly higher in ESCC patients than non-ESCC controls. P. gingivalis IgG and IgA in serum demonstrated sensitivities/specificities of 29.17%/96.90% and 52.10%/70.81%, respectively, and combination of IgG and IgA produced a sensitivity/specificity of 68.75%/68.46%. The diagnostic performance of serum P. gingivalis IgA for early ESCC was superior to that of IgG (54.54% vs. 20.45%). Furthermore, high serum levels of P. gingivalis IgG or IgA were associated with worse prognosis of ESCC patients, in particular for patients with stage 0-IIor negative lymphnode metastasis, and ESCC patients with high levels of both IgG and IgA had the worst prognosis. Multivariate analysis revealed that lymph node status, IgG and IgA were independent prognostic factors. Conclusions The IgG and IgA for P. gingivalis are potential serum biomarkers for ESCC and combination of IgG and IgA improves the diagnostic and prognostic performance. Furthermore, serum P. gingivalis IgG and IgA can detect early stage ESCC. Electronic supplementary material The online version of this article (10.1186/s12885-017-3905-1) contains supplementary material, which is available to authorized users.


Background
Esophageal squamous cell carcinoma (ESCC) remains the predominant histological subtype of esophageal carcinoma and ranks as the fourth most common cancer in terms of both incidence and mortality in China [1,2]. Although significant advances in diagnostic and therapeutic modalities have improved the prognosis of ESCC patients, the overall 5-year survival rate still ranges from 25% to 30%, mainly due to advanced stage at initial presentation [1,[3][4][5][6][7]. On the other hand, accurate staging and prognosis is difficult to assess at diagnosis, which hampers ESCC tailoring therapy, treatment efficiency and recurrence monitoring. It is, therefore, imperative to identify novel biomarkers for early detection, metastasis and recurrence to reduce ESCC-related morbidity and mortality.
A number of epidemiological and clinical studies have reported a positive association between the conditions of oral microbiome, periodontal disease or tooth loss and the progression of multiple cancers [8][9][10][11][12][13][14][15][16][17][18][19][20][21][22][23][24][25], and even gastric precancerous lesions [26,27]. The oral microbiome inhabiting the oral cavity contains multiple species in a complex community that generally exist in a balanced immunoinflammatory state with the host [28]. Disruption of this equilibrium has deleterious effects on the mucosal lining, surrounding tissues and even distant organs and systems of human body through the combined effects of a dysbiotic microbial community and a dysregulated immune response [12,13,29]. Porphyromonas gingivalis has become regarded as a key-stone pathogen and is closely associated with periodontal diseases, a variety of presumably unrelated chronic diseases and multiple cancers [30,31]. Although the self-reported tooth loss may have a microbial basis in the case of esophageal cancer [16,17], there is no convincing evidence of direct and specific microbial etiologic agents until our recent findings, which revealed a higher frequency (61%) of P. gingivalis presence in ESCC [18].
As P. gingivalisis is an important periodontal pathogen in various types of periodontal disease, numerous studies have reported that antibody responses to P. gingivalis correlate with severity and progression of periodontitis, extent of attachment loss and treatment effects [32][33][34][35][36]. In a cohort study of NHANES III, not only the increasing severity of periodontitis but the higher serum IgG for P. gingivalis was associated with increased orodigestive cancer mortality [25]. In another European prospective cohort study, high levels of antibodies to P. gingivalis rendered a > 2-fold increased risk to pancreatic cancer [21]. In clinical settings, serum tumor biomarkers take priority over other measures for screening, diagnosis and clinical management of cancer. However, conventional serum markers for ESCC, such as squamous cell carcinoma antigen (SCCA), carcinoembryonic antigen (CEA), CYFRA21-1 and carbohydrate antigen (CA)19-9, lack sufficient sensitivity and specificity for the early detection and progression of ESCC [37][38][39][40][41].
On the grounds of our recent study establishing the association between the infection of P. gingivalis in esophageal epithelium and progression of ESCC, herein we investigate the serum levels of immunoglobulin G and A (IgG and IgA) for P. gingivalis and their clinical significance for the diagnosis and postoperative prognosis of ESCC.

Patients
The first cohort of 96 preoperative serum samples were recruited from ESCC patients, who underwent curative esophagectomy at the First Affiliated Hospital of Henan University of Science & Technology and Anyang people's hospital. None of ESCC patients received preoperative neoadjuvant chemoradiotherapy. The clinical stage of ESCC was classified in accordance with the seventh edition of AJCC and early stage was defined as AJCC stage 0 + I + IIA. Another cohort of 50 serum samples were collected from patients with esophagitis, who underwent gastroscopy. In addition, 80 healthy individuals without evidence of comorbid disease were recruited as healthy controls from the physical examination center of our hospital.
Enzyme-linked immunosorbent assay P. gingivalis ATCC 33277, used as the antigen in our experiment, was cultured and prepared as previously described. For enzyme-linked immunosorbent assay (ELISA), 100 ul of reconstituted protein extracts of P. gingivalis (10 μg/ml) was used to coat microtiter plates followed by incubation with 1:200 diluted serum incubation, 1: 1000 biotin-conjugated anti-human IgG and IgA, and 1:400 avidin-conjugated peroxidase. Antibodies levels were expressed as ELISA units (EUs) with the use of a reference serum pool [42].

Statistical analysis
The statistical analyses were performed using SPSS 19.0 software package (SPSS, Chicago, IL, USA). Data are expressed as mean ± standard deviation (SD). Comparisons between groups were performed using t tests. The receiver operating characteristic (ROC) was used to determine the optimal cut-off value of IgG and IgA. The accuracy, sensitivity, specificity, false negative rate (FNR), false positive rate (FPR) and area under the ROC (AUC) were used to assess the classification efficiency. Overall survival (OS) was defined as the interval between the date of surgery and the date of death or the date of last follow-up. Follow-up data was available for 80 ESCC patients with a median follow-up interval of 10.5 months (3.0-42.6 months). Clinical stage and lymph node metastasis were available for 78 ESCC patients. Survival curves were plotted using the Kaplan-Meier method and differences between curves were tested by log-rank tests. The significance of prognostic factors on survival was studied by Cox regression model.

Levels of serum IgG and IgA for P. gingivalis in ESCC
The details of ESCC characteristics are presented in Table 1. Figure 1 shows the frequency distributions of IgG and IgA for P. gingivalis across the three cohorts. As there were no significant differences between healthy controls and non-ESCC patients with esophagitis with regards to serum levels of P. gingivalis IgG or IgA, we combined these two cohorts as non-ESCC controls hereafter. The median serum levels of IgG and IgA for P. gingivalis were significantly higher in ESCC patients than in non-ESCC controls (150.69 EU vs. 109.13 EU, P < 0.001 for IgG; 33.16 EU vs. 19.14 EU, P < 0.01 for IgA). However, no significant correlation was found between serum levels of P. gingivalis IgG and IgA (r 2 = 0.03, P > 0.05, data not shown).
Seeking to determine the diagnostic potential of P. gingivalis IgG and IgA, ROC curves were plotted to distinguish 96 patients of ESCC from 130 non-ESCC controls. As shown in Fig. 2a, AUCs of IgG and IgA for P. gingivalis were 0.612 and 0.632, with optimal cut-off values of 189.17 EU and 21.25 EU, respectively. The specificity for IgG was higher (96.90%) than that of IgA (70.81%) but not the sensitivity (29.17% vs. 52.10%, Fig. 2b). Combination of IgG and IgA, i.e. seropositivity for at least one subtype of IgG or IgA antibody, produced an AUC of 0.686 with a sensitivity of 68.75% and a specificity of 68.46%, respectively (Fig. 2a). Figure 2b shows the diagnostic performance of IgG, IgA, and combination of IgG and IgA in terms of accuracy, sensitivity, specificity, FNR and FPR.
Diagnostic value of IgG and IgA for P. gingivalis in early stage of ESCC.
There were 44 patients with early stage disease in our cohort of ESCC. The mean value of P. gingivalis IgA in early stage ESCC was lower (32.08 EU) than that of late stage ESCC (41.76 EU) without statistical significance (P = 0.29), whereas the mean IgG value was marginally higher in early stage ESCC (114.35 EU vs. 113.62 EU, P = 0.058). The sensitivity of P. gingivalis IgA for detection of early stage ESCC was 54.54% (24/44) with a specificity of 70.82%, and was far better than that of IgG (20.45%, (9/44)).

Associations between P. gingivalis IgG and IgA with clinicopathological features and overall survival of ESCC
The associations between clinicopathological features of ESCC and serum levels of IgG or IgA for P. gingivalis were determined by t test. No significant associations were observed between any clinicopathological features with IgG or IgA serum levels. Likewise, ROCs were plotted to predict the 3-year OS rate of ESCC. Figure 2c shows the time-dependent ROC  (Fig. 2d). Figure 3a shows the postoperative survival of 80 ESCC patients with a median survival time of 31.58 months, 61 surviving patients and 19 ESCC-related deaths at the last clinical follow-up (Fig. 3a). Using the optimal cut-off value of 138.23 EU, Kaplan-Meier survival analysis revealed that ESCC patients with higher serum level of P. gingivalis IgG had a significantly worse prognosis than The prognostic values of histopathological features were also evaluated by Kaplan-Meier method and logrank test. With regards to clinical TNM stage, stage I-II ESCC patients (stage I-II, 63.75%, n = 51) survived longer than stage III-IV ESCC cases (Stage III-IV, 36.25%, n = 27, Additional file 1: Figure S1A). For the subgroup ESCC patients with early clinical stage, a significant benefit in OS was observed in patients with low serum level of P. gingivalis IgA but non-significant for IgG than in patients with high level (log-rank test, x 2 = 9.141, P = 0.003, Additional file 1: Figure S1B & D), and neither IgG nor IgA was associated with OS of late stage ESCC (Additional file 1: Figure S1C &E). In addition, lymph node metastasis was significantly associated with shorter OS ((log-rank test, x 2 = 5.61, P = 0.018, Additional file 2: Figure S2A). In ESCC patients with negative lymph node metastasis, those with high levels of P. gingivalis IgG or IgA had worse OS than patients with low IgG or IgA serum level (log-rank test, x2 = 6.097/ 6.097, P = 0.014/0.011, Additional file 2: Figure S2B & D), whereas no significant differences were observed between P. gingivalis IgG or IgA and OS in positive lymph node metastasis (Additional file 2: Figure S2C & E).

Discussion
Early diagnosis remains one of the key determinants to improve the long-term survival of patients with ESCC. The majority of patients with ESCC present at  an advanced stage and have limited treatment options, resulting in dismal prognosis [1,[3][4][5][6][7]. Although gastroscopy with biopsy offers an efficient method for diagnosis of patients with ESCC, poor compliance of gastroscopy in asymptomatic patients precludes early detection. Compared with gastroscopy, blood testing is less invasive and cost-effective. Therefore, serum biomarkers have the priority over other measures for clinical application to detect ESCC at an early stage [43]. First and foremost, the present study demonstrates that serum antibody levels against P. gingivalis have the potential for diagnosis of ESCC. Although inflammation plays a key role in esophageal carcinogenesis, our results revealed that morphological esophagitis harboring inflammatory cells without transformed cells in esophageal mucosa failed to show increased IgG and IgA antibody response to P. gingivalis. This finding indicates that P. gingivalis may not be involved in the process of esophagitis, but do not rule out the possibility that P. gingivalis or host responses against P. gingivalis contribute to the development and progression of ESCC. In sharp contrast, titers of IgG and IgA against P. gingivalis in serum of patients with ESCC increased remarkably compared to patients with esophagitis and healthy controls, which provides direct evidence that P. gingivalis is implicated in the pathogenesis of ESCC. Using an optimal diagnostic cut off value of 425 EU, individual IgA had the highest sensitivity (52.1%) for discrimination of ESCC from non-ESCC controls compared with conventional serum markers for ESCC, such as SCCA, CYFRA21-1, CEA, CA19-9 [37][38][39][40][41], whereas the specificity was low (70.8%). However, ELISA results of SCCA1, SCCA2, CYFRA21-1 and CEA did not show diagnostic value in our cohort (data not shown). Growing evidence indicates that combination of several individual biomarkers is superior to any single biomarker [44]. Combination of IgG and IgA for P. gingivalis had an increased AUC (0.671) compared with an individual IgG or IgA.
For detection of early stage ESCC, conventional serum biomarkers of ESCC have little diagnostic benefit. For instance, the positive frequencies of both CYFRA21-1 and SCCA in patients with early stage ESCC (stage 0-II) varied from 4.7% to 24% [37,40]. In contrast, the diagnostic performance of serum P. gingivalis IgA for early ESCC was superior as evidenced by a sensitivity of 54.54% in our study. Although the specificity of single IgA was not sufficient, combination of IgG and IgA produced a specificity of 91.5%.
Mounting clinical evidence indicates a positive association between P. gingivalis or periodontal disease and an increased risk for a variety of cancers and even poor prognosis [11,12,18,21,25]. In normal distal esophagus, bacterial colonization was not uncommon [45]. Furthermore, the global esophageal microbiome in both esophagitis and Barrett's esophagus altered from typeI bacteria in normal esophageal mucosa to typeIIbacteria, many of which are Gram-negative anaerobes/microaerophiles and putative pathogens of periodontal disease [46]. Our previous study demonstrated that P. gingivalis infection in ESCC was prevalent (61%) and negatively correlated with OS of ESCC [18]. In the present study, we looked into the prognostic potential of human immune response to P. gingivalis in terms of IgG and IgA. In line with the presence of P. gingivalis in ESCC, higher serum levels of P. gingivalis IgG and IgA were associated with worse prognosis of patients with ESCC. In particular for early stage ESCC, i.e. ESCC with stage 0-II or negative lymphnode metastasis, patients with high level of P. gingivalis IgG or IgA had a significantly lower OS relative to ESCC patients with low level, and patients with high level of both IgG and IgA had the worst prognosis. Multivariate analysis identified lymph node status, IgG and IgA as independent prognostic factors. Therefore, IgG and IgA were combined and we found that the combination produced higher predictive accuracy than an individual IgG or IgA.

Additional files
Additional file 1: Figure S1. Additional file 2: Figure S2. Kaplan-Meier survival curves of ESCC patients with regards to lymph node stage. A The 3-year OS rates in ESCC patients without lymph node metastasis (n = 44) and patients with lymph node metastasis (n = 34) were 63.87% and 27.85%, respectively (P = 0.018