Abnormal expression of TRAIL in ESCC is negatively correlated with patient clinical outcomes
To reveal the expression patterns of inflammatory factors and the corresponding receptors in ESCC tumor and normal tissues, we analyzed TCGA RNA-seq database. We found that CXCL14, TGFβ1, and TRAIL (TNFSF10) were highly expressed in tumor tissues compared with adjacent normal samples (Fig. 1a). Next, we analyzed the relationship between these three genes and patient clinical parameters. CXCL14 and TGFβ1 were not correlated with lymph node metastasis, staging, or overall survival (Fig. 1b, c). Only TRAIL (TNFSF10) had a significant positive relationship with lymph node metastasis and staging, and patients with high TRAIL (TNFSF10) levels exhibited poor prognosis (Fig. 1d). To further verify this, we analyzed TRAIL expression in 83 patients with ESCC at both the mRNA and protein levels. A similar trend was observed, and TRAIL was highly expressed in tumor tissues compared with adjacent normal samples (Fig. 1e, f). In addition, TRAIL was significantly increased in advanced stages (IIb–IV) compared with the early stages (I–IIa; Fig. 1f) and was negatively correlated with overall survival of patients with ESCC (Fig. 1g). Together, these data suggest that TRAIL accumulates in ESCC tumor sites and is negatively correlated with patient survival.
TRAIL is expressed by ESCC cells and knocking it down reduces ESCC cell migration in vitro
Initially, TRAIL was reported to be expressed by immune cells and that it defends against pathogens and self-antigens. However, TRAIL was found expressed by colon and lung cancers and a contributor to tumor progression in recent years[9]. Herein, TRAIL was found negatively correlated with patient prognosis, and immunohistochemistry data revealed that TRAIL was mainly expressed around the tumor cells. Thus, we hypothesized that ESCC cells may be the source of TRAIL. To determine whether ESCC cells express TRAIL, we examined TRAIL expression in seven human ESCC cell lines at different stages of differentiation (EC9706, TE7, EC1, KYSE150, KYSE70, TE-1, and EC109) at the mRNA and protein levels (Fig. 2a). TRAIL expression was found heterogeneous in ESCC cell lines, with KYSE150/KYSE70 cells exhibiting the highest expression levels (Fig. 2a).
To evaluate the possible biological function of TRAIL in ESCC, we silenced TRAIL in KYSE70 and KYSE150 cells using siRNAs. After verifying the silencing efficiency by RT-qPCR (Supplementary Fig. 1a), sphere formation, Transwell, and CCK8 assays were performed. The data showed that knocking down TRAIL significantly reduced the ability of cells to form spheroids, migrate, and proliferate (Supplementary Fig. 1d–f). Moreover, we analyzed stemness-related gene expression (Bmi1, Oct4, and Klf4) and found that TRAIL silencing decreased their gene expression levels (Supplementary Fig. 1g). These results suggest that TRAIL may enhance the stem properties of ESCC. To investigate this, we knocked down TRAIL using short hairpin RNAs (shRNAs); knockout efficiency was verified at both the mRNA and protein levels (Fig. 2b, c), after which sphere formation, Transwell, and CCK8 assays were performed. Similarly, knocking down TRAIL reduced ESCC cell migration, spheroid formation, and proliferation ability (Fig. 2d–f). Moreover, expression of the stemness-related genes Bmi1, Oct4, Cd44, and Klf4 significantly declined in TRAIL-knockdown KESE70 and KYSE150 cells; the change in Sox2 expression levels were not significant (Fig. 2g). Together, these results suggest that ESCC cells express TRAIL and that TRAIL may promote tumor stemness.
Overexpression of TRAIL promotes ESCC cell migration, invasion, and proliferation in vitro
We next stably overexpressed TRAIL in EC1 and TE1 cells, which showed the lowest TRAIL levels. After verifying the efficacy at both the mRNA and protein levels (Fig. 3a, b) we performed sphere formation, Transwell, and CCK8 assays. In contrast to TRAIL knockdown, overexpression significantly enhanced EC1/TE1 cell sphere formation (Fig. 3c), migration (Fig. 3d), and proliferation (Fig. 3e) abilities. As for the stemness-related genes, their expression was significantly increased in TRAIL overexpressed EC1 and TE1 cells compared with the control group (Fig. 3f). We further we used recombinant human TRAIL (rh-TRAIL) to verify this phenomenon. Similar to the overexpression system, rh-TRAIL treatment significantly increased the expression of the stemness-related markers CD271, CXCR4, and Bmi1 (Supplementary Fig. 2a–c). Additionally, rh-TRAIL significantly promoted cellular migration and sphere formation (Supplementary Fig. 2d, e). Taken together, our data indicate that TRAIL facilitates the migration, invasion, proliferation, and stemness of ESCC cells.
TRAIL promotes EMT of ESCC cells
Previous studies have shown that tumor distant metastasis and migration mainly result from EMT[20] and that TRAIL regulates EMT in breast cancer[21]. Given that TRAIL was found expressed by ESCC cells, which promoted their migration and stemness, we investigated whether TRAIL promotes ESCC stemness and migration in an EMT-dependent manner. We examined the main markers associated with EMT and found that knocking out TRAIL significantly increased E-cadherin expression and markedly reduced N-cadherin, vimentin, MMP2, and MMP9 levels (Fig. 4a, b). Of note, MMP3 was significantly elevated in shTRAIL-KYSE150 cells, but not in shTRAIL-KYSE70 cells (Fig. 4a, b). During the initial EMT step, cancer cells secrete MMPs to degrade collagen and fibrin[22]. MMP2/3/9 can degrade collagen and fibrin, which may explain why these two cell lines showed different MMP expression patterns. The TRAIL-induced modulation of EMT-related genes (E-cadherin, N-cadherin, and vimentin) was further validated at the protein level by western blotting (Fig. 4c). Similar results were obtained after using siRNA to knockdown TRAIL (Supplementary Fig. 3a–d).
On the other hand, when TRAIL was overexpressed in EC1 and TE1 cells, we observed the opposite trends at both the mRNA (Fig. 4d, e) and protein levels (Fig. 4e–g). Adding human recombinant TRAIL increased PD-L1, vimentin and N-cadherin expression levels, reduced E-cadherin expression (Supplementary Fig. 3e–h). Taken together, these results indicate that TRAIL promotes EMT of ESCC cells, which may underly the promotion of ESCC progression by TRAIL.
TRAIL activates the ERK/STAT3 signaling pathway to induce PD-L1 expression in ESCC
Next, we investigated the mechanism of TRAIL-induced EMT. Programmed death-ligand 1 (PD-L1), as an immunosuppressive molecule, is highly expressed on the surface of a variety of cancer cells[23]. Recently, it was reported that PD-L1 promotes EMT, and thus induces tumor progression in lung and breast cancer[24–26]. To investigate whether PD-L1 is involved in TRAIL-induced EMT, we first analyzed the expression of immunosuppressive molecules (CD274, TIM3, LAG3, CTLA4, CD38, CD101, MKI67, ICOS, EOMES, CD44, CD28, KLRG1, CD5, TIGIT, and CD47) in the TRAIL high and low groups in TCGA database[27]. We found that PD-L1 (CD274) and CTLA4 show significantly higher expression in the TRAILhigh group than in the TRAILlow group (Fig. 5a); however, PD-L1 (CD274) was predominantly expressed by tumors[28] while CTLA4 and LAG3 were mostly expressed by T cells in the tumor microenvironment[29]. Notably, a markedly positive correlation was observed between PD-L1 (CD274) and TRAIL (TNFSF10) in ESCC tissues (Fig. 5b). To confirm this, we performed immunohistochemistry of ESCC samples collected from the patients. Similar expression results were obtained, and a significant correlation between PD-L1 and TRAIL was observed (Fig. 5c, d). Furthermore, we examined TRAIL-knockout and overexpressed cells and found that PD-L1 expression was lower in the knockdown group and higher in the overexpression group than in control (Fig. 5e). Thus, these results suggest that TRAIL regulates PD-L1 expression in ESCC.
We then assessed the mechanism underlying TRAIL regulation of PD-L1 expression. As a key molecule in tumor immune escape, the main mechanisms involved in PD-L1 regulation have been reported. We first examined the KEGG enrichment analysis in TCGA and found that only the MAPK/STAT3 signaling pathway was significantly up-regulated in tissues with high TRAIL expression (Fig. 5f). To reveal whether TRAIL-induced PD-L1 expression depends on MAPK/STAT3, we performed western blotting of TRAIL-knockout and overexpressed ESCC cells. Knocking down TRAIL in KYSE150 cells markedly reduced p-ERK and p-STAT3 expression, while overexpression in EC1 cells significantly increased p-ERK and p-STAT3 levels (Fig. 5g). Furthermore, treatment of EC1 cells with ERK or STAT3 inhibitors reversed the TRAIL overexpression-induced increase in PD-L1 levels (Fig. 5h, i).
When PD-L1 antibody was added to TRAIL-overexpression EC1 cells, only N-cadherin was significantly downregulated, while E-cadherin and vimentin showed no difference in mRNA levels (Supplementary Fig. 4a). However, after silencing PD-L1 in TRAIL-overexpression EC1 cells, both N-cadherin and vimentin were markedly downregulated, while E-cadherin expression did not change (Supplementary Fig. 4b). The aggressive and proliferation abilities of tumor cells were also inhibited (Supplementary Fig. 4c, d). These findings indicate that cytoplasmic PD-L1 promotes EMT, whereas PD-L1 expressed on the cell membrane cannot promote cell metastasis and invasion. Altogether, the results demonstrate that TRAIL-induced EMT is dependent on ERK/STAT3-activated PD-L1 expression in ESCC.
TRAIL activates the ERK/STAT3 pathway, inducing PD-L1 and promoting EMT of ESCC cells in vivo
We next examined whether the TRAIL-induced promotion of EMT through ERK/STAT3-activated PD-L1 expression in vitro occurs in vivo using mouse models. Knocking down TRAIL significantly delayed the tumor growth of KYSE150 cells (Fig. 6a, b). Moreover RT-qPCR and western blotting showed a significant increase in E-cadherin expression and a marked decrease in the expression of N-cadherin, vimentin, PD-L1, MMP2, MMP3, MMP9, and stemness-related genes (Sox2, Bmi1, Oct4, and Cd44) in shTRAIL-KYSE150 cells compared with control (Fig. 6c, d). Meanwhile, immunohistochemistry data showed a significant positive correlation between TRAIL and PD-L1 (Fig. 6e, f). In contrast, overexpression of TRAIL in EC1 cells significantly accelerated tumor growth (Supplementary Fig. 5a, b) and promoted EMT- and stemness-related gene expression at the mRNA and protein levels (Supplementary Fig. 5c, d). Moreover, overexpression of TRAIL upregulated p-ERK/p-STAT3/PD-L1 levels (Supplementary Fig. 5d). The correlation between TRAIL and PD-L1 was also demonstrated (Supplementary Fig. 5e–g). Taken together, these data indicate that TRAIL promotes EMT-induced cell metastasis through ERK/STAT3 signaling pathway activation and inducing PD-L1 expression in ESCC in vivo.