Expression of TM4SF1 and DDR1 in cancers
The biological functions and roles of TM4SF1 and DDR1 in cancer are still unclear. To reveal their roles in tumorigenesis and development, differentially expressed genes were identified. The results showed that TM4SF1 and DDR1 were highly expressed in a variety of common malignant tumours (Figure 1). GEPIA analysis showed that TM4SF1 and DDR1 were significantly higher in ovarian cancer than in normal ovaries (both P＜0.05). A meta-analysis of the Oncomine database (including 10 analyses of 7 ovarian cancer datasets) also showed that TM4SF1 and DDR1 were significantly more highly expressed in ovarian cancer than in normal ovarian tissue (TM4SF1: P=0.004; DDR1: P=3.22E- 4) as shown in Figure 2.
Enrichment analysis of TM4SF1- and DDR1-interacting proteins
The top 50 proteins that directly interact with TM4SF1 and DDR1 were queried by the STRING online tool, and 28 and 25 proteins were found to interact with TM4SF1 and DDR1, respectively. Among these interacting proteins, both CYSTM1 and NT5M interacted with TM4SF1 and DDR1, and TM4SF1 also interacted with DDR1 (Figure 3).
GO enrichment analysis of the interacting proteins showed that they were mainly involved in the composition of ECM, collagen, integrin and their complexes as well as adhesion spots, endoplasmic reticulum and other intracellular and extracellular components. GO enrichment analysis also indicated that these interacting proteins had the molecular function of binding with integrin, cell adhesion molecules, growth factors, proteoglycans, neurotrophic factor receptors and γ -catenin. GO analysis also showed that they were mainly involved in cell-matrix adhesion and cell response to amino acid stimulation, and it indicated that they played biological functions by activating integrin and collagen-mediated signalling pathways (Figure 4A, Supplementary Table 1). KEGG enrichment analysis showed that these interacting proteins were mainly involved in adhesion spot-related signal transmission, ECM-receptor interaction, actin cytoskeleton regulation, small RNA signalling pathways related to cancer, proteoglycan signalling pathways related to cancer, the PI3K/Akt signalling pathway, neurotrophic factor signalling pathways and a variety of tumour-related pathways, such as small cell lung cancer-, chronic myeloid leukaemia-, thyroid cancer- and bladder cancer-related pathways (Figure 4B, Supplementary Table 2).
Expression of TM4SF1 and DDR1 proteins in ovarian cancer and the relation with clinicopathologic features
A total of 94 patients with epithelial ovarian cancer were included, and the patient age ranged from 28 to 83 years old with a median age of 51 years. A total of 56 patients received neoadjuvant chemotherapy. The results showed that 46 patients (48.94%) were positive for TM4SF1 protein, and 52 patients (55.32%) were positive for DDR1 protein. In addition, 26 patients (27.66%) were positive for both TM4SF1 and DDR1 protein. The staining of TM4SF1 and DDR1 proteins in ovarian cancer tissues is shown in Figure 5. Correlation analysis showed that only the positive expression of DDR1 protein was significantly different in different histological grades, FIGO stages and intraperitoneal metastases (both P＜0.05, Table 1).
Relationship of TM4SF1 and DDR1 expression with the prognosis of ovarian cancer
Biological database analysis showed that ovarian cancer patients with high expression of TM4SF1 had significantly lower disease-free survival(DFS) or PFS than those with low expression (DFS: HR=1.3, P=0.046, n=424, Figure 6A; PFS: HR=1.17, P =0.019, N=1435, Figure 6E). In addition, the expression of DDR1 had no correlation with patient DFS or PFS, and the expression of TM4SF1 and DDR1 had no significant correlation with patient OS (Figure 6B-D, Figure 6F-H). However, KM plotter multigene analysis showed that higher expression of TM4SF1 and DDR1 was significantly associated with shorter PFS of patients with ovarian cancer (HR = 1.15, P =0.039, n=1435, Figure 6I) but not with OS (Figure 6J).
Analysis of clinical data showed that 94 patients had ovarian cancer with a median follow-up of 33 months, and there was no significant difference in the median OS between TM4SF1-positive and TM4SF1-negative patients (29 vs. 47 months, P>0.05). In contrast, the median OS of DDR1-positive patients was significantly shorter than that of DDR1-negative patients (31 vs. >73 months, P<0.05). The median OS of patients with TM4SF1 and DDR1 coexpression was significantly shorter than that of patients lacking TM4SF1 and DDR1 coexpression (21 vs. 49 months, P<0.05). In addition, the median PFS of TM4SF1-positive patients was significantly shorter than that of TM4SF1-negative patients (18 vs. 26 months, P<0.05). The median PFS of patients with TM4SF1 and DDR1 coexpression was significantly shorter than that of patients lacking TM4SF1 and DDR1 coexpression (12 vs. 26 months, P<0.05), while the expression of DDR1 was not related to the median PFS (P>0.05) (Figure 7). Thus, the TM4SF1 and DDR1 coexpression indicated that patients with ovarian cancer had shorter PFS and OS.
Analysis of prognostic factors in ovarian cancer
Regarding the clinicopathological factors, univariate analysis showed that FIGO stage, intraperitoneal invasion, DDR1 expression, TM4SF1 expression and coexpression of DDR1 and TM4SF1 were factors affecting the OS of patients with ovarian cancer, while FIGO stage, lymph node resection and coexpression of TM4SF1 and DDR1 were factors affecting the PFS of ovarian cancer patients (Table 2).
Cox multivariate analysis showed that TM4SF1 and DDR1 coexpression was the only independent risk factor affecting OS and PFS in ovarian cancer patients (FIG. 8A, B), suggesting that the expression of TM4SF1 and DDR1 may be synergistically involved in the development of ovarian cancer.