DCBLD2 was associated with unfavorable survival in PDAC
Fig. 1 shows the research workflow of this study. The univariate Cox regression analysis screened hundreds of survival-related genes in each cohort, while only DCBLD2 was identified as a robust prognostic gene by Venn diagram. In all available PDAC cohorts with clinical data, DCBLD2 was significantly associated with overall survival (OS) of patients (GSE21501: HR = 1.21, 95% CI = 1.06–1.39, P = 0.0065; GSE28735: HR = 1.98, 95% CI = 1.24–3.14, P = 0.004; GSE57495: HR = 1.53, 95% CI = 1.16–2.01, P = 0.0029; GSE62452: HR = 1.68, 95% CI = 1.24–2.29, P = 0.0009; GSE71729: HR = 1.28, 95% CI = 1.00–1.64, P = 0.0457; GSE85916: HR = 1.89, 95% CI = 1.43–2.51, P < 0.0001; MTAB-6134: HR = 1.39, 95% CI = 1.19–1.56, P < 0.0001; PACA-AU: HR = 1.52, 95% CI = 1.13–2.05, P = 0.0056; PACA-CA: HR = 1.14, 95% CI = 1.03–1.25, P = 0.0098; TCGA: HR = 1.24, 95% CI = 1.03–1.49, P = 0.0199).
DCBLD2 is upregulated in PDAC with diagnostic potential
We first investigated the expression pattern of DCBLD2 in PDAC patients. As illustrated in Fig. 2a, DCBLD2 was remarkably overexpressed in PDAC tissues compared with normal tissues in seven GEO datasets. According to the expression data from Gene Expression Profiling Interactive Analysis (GEPIA), DCBLD2 was significantly elevated in PDAC tissues (Fig. 2b). Fig. 2c-f demonstrated that DCBLD2 expression was also increased in PDAC tissues compared with paired normal tissues. Furthermore, we explored the diagnostic potential of DCBLD2 in four GEO cohorts (GSE32676, GSE60979, GSE62165 and GSE71729). The area under the curve (AUC) value of DCBLD2 was 0.800, 0.874, 0.984 and 0.818, respectively (Fig. 2g-j), which was no less than that of an established diagnostic marker, CA19-9, whose AUC value was approximately 0.84 . In addition, the expression of DCBLD2 was remarkably increased in patients with high grade (P<0.05), suggesting that DCBLD2 was related to high tumor malignancy (Additional file 2: Fig. S1).
Prognostic performance of DCBLD2
We next assessed the prognostic efficiency of DCBLD2 in ten independent PDAC cohorts. K-M survival curves illustrated that DCBLD2 could precisely capture the survival differences between low- and high-expression patients (Fig. 3). The calibration curves revealed that the clinical outcomes predicted by DCBLD2 were in good accordance with the actual observations (Fig. 4).
We further compared the robustness of DCBLD2 with clinical indicators, including histological grade, N stage and T stage, in MTAB-6134, PACA-AU and TCGA cohorts. The AUC value of DCBLD2 was 0.708, 0.753 and 0.690, respectively, which is greater than that of clinical factors in all three cohorts (Fig. 5). This finding suggested that DCBLD2 outperformed traditional indicators in predicting PDAC survival. Moreover, patients in the high-expression group had a significantly decreased disease-free survival (DFS) compared with low-expression group in MTAB-6134 and TCGA cohorts, indicating that DCBLD2 may serve as a prognostic indicator of DFS (Additional file 2: Fig. S2).
Relationship between immune cell infiltration and DCBLD2 expression
We subsequently determined the relationship between DCBLD2 expression and immune cell infiltration in PDAC samples from MTAB-6134 and TCGA cohorts by the software CIBERSORT. The abundance of macrophage M0 and M2 was positively related to DCBLD2 expression, while CD8 + T cells had negatively correlation with DCBLD2 expression in MTAB-6134 cohort (P<0.05, Fig. 6a-c). Similar trends were observed in TCGA cohort (P<0.05, Fig. 6d-f).
Biological function and pathway of DCBLD2
In order to clarify the function mechanism of DCBLD2, we performed biological process and KEGG pathway enrichment analyses on top 1000 positively co-expressed genes of DCBLD2. For biological process, DCBLD2 was found to be primarily involved in angiogenesis, cell adhesion, cell motility and cell migration in both cohorts (Fig. 7a-b). For pathway enrichment, DCBLD2 was mainly associated with PI3K-AKT signaling pathway, Hippo signaling pathway, Rap1 signaling pathway and pancreatic cancer in both cohorts (Fig. 7c-d).
Expression of DCBLD2 in extracellular vesicles from human plasma samples
Early diagnosis of PDAC remains challengeable, and extracellular vesicles have emerged as attractive diagnostic biomarkers for early detection of PDAC. Since DCBLD2 was upregulated in PDAC tissues, we wondered whether DCBLD2 was also highly expressed in extracellular vesicles from plasma samples of PDAC patients. As Fig. 8a illustrated, expression of DCBLD2 in extracellular vesicles from plasma samples of PDAC patients was higher than that of normal donors (P=0.0029) or CP patients (P<0.0001). In addition, DCBLD2 in extracellular vesicles could serve as a moderate diagnostic biomarker for PDAC as the AUC value was 0.627 (Fig. 8b).
Validation of DCBLD2 expression in extracellular vesicles from human serum samples
Extracellular vesicles can be isolated from both plasma and serum of whole blood, and we had proved that DCBLD2 in extracellular vesicles from plasma samples had diagnostic value based on the public data. We next analyzed our own data to evaluate the diagnostic value of DCBLD2 in extracellular vesicles from serum samples. The results of NTA analysis and TEM demonstrated typical characteristics of isolated extracellular vesicles (Fig. 9a-b). Fig. 9c showed that the expression of DCBLD2 in extracellular vesicles from serum samples was markedly elevated in PDAC patients compared with normal donors (P<0.0001) or CP patients (P=0.0018). Similarly, DCBLD2 in extracellular vesicles from serum samples also bore moderate diagnostic value for PDAC as the AUC value was 0. 756 (Fig. 9d).