Serum DSG2 as a potential biomarker for diagnosis of esophageal squamous cell carcinoma and esophagogastric junction adenocarcinoma

Abstract Background: Exploration of serum biomarkers for early detection of upper gastrointestinal cancer is required. Here, we aimed to evaluate the diagnostic potential of serum desmoglein-2 (DSG2) in patients with esophageal squamous cell carcinoma (ESCC) and esophagogastric junction adenocarcinoma (EJA). Methods: Serum DSG2 levels were measured by enzyme-linked immunosorbent assay (ELISA) in 459 participants including 151 patients with ESCC, 96 with EJA, and 212 healthy controls. Receiver operating characteristic (ROC) curves were used to evaluate diagnostic accuracy. Results: Levels of serum DSG2 were significantly higher in patients with ESCC and EJA than those in healthy controls (P<0.001). Detection of serum DSG2 demonstrated an area under the ROC curve (AUC) value of 0.724, sensitivity of 38.1%, and specificity of 84.8% for the diagnosis of ESCC in the training cohort, and AUC 0.736, sensitivity 58.2%, and specificity 84.7% in the validation cohort. For diagnosis of EJA, measurement of DSG2 provided a sensitivity of 29.2%, a specificity of 90.2%, and AUC of 0.698. Similar results were observed for the diagnosis of early-stage ESCC (AUC 0.715 and 0.722, sensitivity 36.3 and 50%, and specificity 84.8 and 84.7%, for training and validation cohorts, respectively) and early-stage EJA (AUC 0.704, sensitivity 44.4%, and specificity 86.9%). Analysis of clinical data indicated that DSG2 levels were significantly associated with patient age and histological grade in ESCC (P<0.05). Conclusion: Serum DSG2 may be a diagnostic biomarker for ESCC and EJA.


Introduction
Esophageal cancer ranks seventh in terms of incidence and sixth for cancer-related deaths worldwide [1], and is among the most invasive and metastatic malignancies, representing a serious global health problem [2]. There are two main histologic types of esophageal cancer: esophageal squamous cell carcinoma (ESCC) and esophageal adenocarcinoma (EAC) [3]. ESCC is the most prevalent form of esophageal cancer, accounting for 70% of cases. Notably, the prevalence of ESCC has decreased in recent years, while the incidence of esophagogastric junction adenocarcinoma (EJA) is increasing significantly worldwide [4][5][6]. In China, EJA is also prevalent in areas with high incidence rates of ESCC [7]. Smoking and alcohol consumption account for more than 90% of ESCC in the Western world [8], but are not important contributing factors in ESCC occurrence in China [7]. Nevertheless, geographical differences in incidence rates in China strongly suggest that there are major etiological environmental or lifestyle factors influencing the development of ESCC and EJA [9]. values were obtained from Youden's index values of the ROC curves, which yield maximum values of sensitivity plus (100%-specificity). Positive predictive value (PPV), negative predictive value (NPV), false positive rate (FPR), false negative rate (FNR), positive likelihood ratio (PLR), and negative likelihood ratio (NLR) were obtained using the optimal cut-off values. Associations between clinical characteristics and levels of DSG2 were evaluated using the chi-square test. In all statistical tests, P-values (two sided) <0.05 were considered statistically significant.

Serum DSG2 levels in patients with ESCC and EJA and healthy controls
To evaluate the level of DSG2 in serum samples, 459 participants were selected, including 97 patients with ESCC and 92 healthy volunteers in the training cohort; 54 patients with ESCC and 59 healthy volunteers in the validation cohort; and 96 patients with EJA and 61 healthy volunteers ( Figure 1). The clinical features of patients and healthy controls are summarized in Table 1.
Initial analysis of serum DSG2 levels indicated a difference in distributions between the healthy control and patient groups, where the patient group accounted for more histogram volume at higher DSG2 concentrations and the healthy control group accounted for more histogram volume at lower concentrations ( Figure 2). In patients with ESCC in the training cohort, mean + − SD serum DSG2 concentration was 0.168 + − 0.135 ng/ml, while values in the early-stage disease and healthy control groups were 0.156 + − 0.123 and 0.093 + − 0.069 ng/ml, respectively (Table 2). To better visualize the distribution and degree of dispersion, scatter plots of serum DSG2 levels in each group were generated ( Figure 3). Patients with ESCC had significantly higher serum DSG2 levels than healthy controls (P<0.001) ( Figure  3). As shown in Figure 3 and Table 2, the difference between patients with early-stage ESCC and healthy controls was also significant (P=0.019). In the validation and combined cohorts, serum DSG2 levels were also higher in patients with ESCC than controls ( Figure 3 and Table 2). Similar results were observed in EJA ( Figure 3 and Table 2).
In the validation and joint cohorts, we found similar diagnostic performance to those determined using training cohort data when the same cut-off value was used (    75.2-93.8%) ( Table 3). To improve clinical interpretation, we also present predictive values and likelihood ratios for use of DSG2 for the diagnosis of ESCC and EJA (Table 3).

Associations between serum DSG2 concentration and clinicopathological features
The data presented in Tables 4 and 5 demonstrate the relationships between levels of serum DSG2 and clinicopathological features of ESCC and EJA, respectively. Levels of DSG2 in the joint ESCC cohort were significantly associated with patient age and histological grade (P<0.05), but not with other analyzed factors, including sex, smoking, depth of invasion, metastasis, and TNM stage (Table 4). Further, levels of DSG2 were not significantly associated with any clinical features in the ESCC training cohort, ESCC validation cohort (Table 4), or EJA (Table 5).

Discussion
The application of serum biomarkers in cancer screening, monitoring response to treatment, and diagnosis is increasingly being investigated by researchers, mainly due to their greater acceptance and wider accessibility [30]. In recent years, some tumor-specific proteins have been identified as biomarkers for various cancers to assist and guide final diagnosis; for example, cancer antigen 125 (CA125) for ovarian cancer [31], carbohydrate antigen 19-9 [32,33] for pancreatic cancer, α-fetoprotein (AFP) [34,35] for hepatocellular cancer, carcinoembryonic antigen (CEA) [36] for colorectal cancer, and prostate-specific antigen (PSA) [37] for prostate cancer. CEA, cytokeratin 19 fragment (CYFRA21-1), and squamous cell carcinoma antigen (SCCA) are the most commonly used biomarkers for detection of ESCC [38]; however, their accuracy is not ideal [39,40]. Therefore, identification of sensitive and specific serum or plasma biomarkers for non-invasive diagnosis of ESCC remains a clinical challenge. Our study demonstrates that DSG2 is a potential candidate biomarker for ESCC. In the current study, we measured the expression levels of DSG2 in serum from patients with ESCC and EJA, and normalized the results using a standard reference, to minimize between-plate variation. Serum DSG2 levels in patients with ESCC and EJA were significantly higher than those in healthy controls (Figure 1), indicating that DSG2 is a potential serological markers for detection of ESCC and EJA. Our findings that DSG2 is up-regulated in ESCC and EJA are consistent with reports on other cancers. Proteomics analysis of tissues from patients with pancreatic ductal adenocarcinoma identified DSG2 as among the top four candidate up-regulated proteins, and serum validation showed significant elevation of DSG2 levels in cancer patient samples [41]. Moreover, patients with ovarian cancer with high serum levels of shed DSG2 had significantly shorter progression-free and overall survival than those with lower DSG2 levels [16]. Klessner et al. found that MMP and ADAM10 caused shedding of the extracellular domain of DSG2 [42,43], and shed DSG2 can be detected in the serum of xenograft tumor models [44]. In addition, cleaved DSG2 enhances the proliferation of intestinal epithelial cells by interacting with HER2 or HER3, to activate the Akt/mTOR and MAPK signaling pathways [45]. These results suggest that DSG2 or cleaved DSG2 may be involved in ESCC progression; however, the role and regulation of cleaved DSG2 fragments in ESCC cells require further investigation.
Additionally, in our current study, serum DSG2 levels were demonstrated to serve as a diagnostic marker for patients with early ESCC and EJA. Measurement of DSG2 exhibited an AUC value of 0.724 with a sensitivity of 38.1% and specificity of 84.8% for diagnosis of ESCC in the training cohort, and these data were further verified in the validation and joint cohorts. Regarding EJA, serum DSG2 expression levels showed an AUC value of 0.698, associated with 29.2% sensitivity and 90.2% specificity. Notably, similar results were obtained for early-stage ESCC and  In the present study, although we demonstrated that DSG2 is a useful biomarker for the diagnosis of ESCC and EJA, its sensitivity requires improvement; however, to our knowledge, this is the first report on the diagnostic value of serum DSG2 for early-stage cancer. Furthermore, most ESCC and EJA patients are already in advanced stages, which makes it difficult to recruit more patients in the early stages. Further investigation with large sample size can be conducted in multiple institutions, which is helpful to better evaluate the diagnostic value of DSG2 as a biomarker. In addition, previous studies have shown that the combined detection of multiple serum proteins as a single panel can increase the sensitivity or specificity of a single biomarker [29,46,47]. Therefore, one of the limitations of our study was that DSG2 was not used in combination with common markers (such as CEA, CYFRA21-1, and SCCA) for the diagnosis of ESCC and EJA. The high sensitivity of DSG2 is expected to be useful for early diagnosis of ESCC and EJA and to improve the prognosis of patients with these diseases. We hope that DSG2 can serve as a complement to specific tumor-markers for the diagnosis of ESCC and EJA, such as CEA.

Conclusion
Taken together, our data provide evidence that serum DSG2 is higher in patients with ESCC and EJA and that analysis of serum DSG2 concentration generates novel and useful information for diagnosis of ESCC and EJA; however, the sample size of patients with early-stage ESCC and EJA was relatively small in the present study. Further verification of the diagnostic value of DSG2 in a larger sample set is warranted.