DJ-1 is upregulated in ESCC samples and positively correlated to tumor progression
The TCGA database was used to analyze the mRNA expression profiles of DJ-1/PARK7 in normal esophageal tissues (n = 10), ESCC tissues (n = 81), and EA tissues (n = 78), and it was found that the expression of DJ-1/PARK7 was significantly upregulated in ESCC tissues (Fig. S2A). In addition, the mRNA levels of DJ-1/PARK7 in patients with T3 stage ESCC were higher than those in patients with T1 stage ESCC (Fig. S2B). Consistent with the TCGA data analysis results, GEO data (GSE53625) analysis also showed that DJ-1/PARK7 was highly expressed in ESCC tumor tissues (Fig. S2C). Kaplan–Meier survival analysis indicated that DJ-1/PARK7 expression was negatively correlated with survival in patients with ESCC (Fig. S2D, n = 179, p = 0.0084). Furthermore, the expression of DJ-1 was detected in the serum of 177 ESCC patients and 50 normal donors and verified the conclusion of its high expression (Fig. 1A). Moreover, the expression of DJ-1 in serum with T3 stage was higher than that in T1 and T2 stage (1.43 times and 1.21 times) (Fig. 1B), and DJ-1 in serum with N2 stage was also higher than that in N0 and N1 stage (1.80 times and 1.59 times) (Fig. 1C). These results suggested that DJ-1 expression is positively correlated with ESCC development.
Irradiation induces secretion of DJ-1 from ESCC cells
In 46 pairs of serum samples from ESCC patients before and after cumulative 40 Gy dose radiotherapy, the expression of DJ-1 in the serum of patients after radiotherapy was much higher than that before radiotherapy (Fig. 1D). To explore whether the elevated DJ-1 in serum was secreted by ESCC cells, we examined the DJ-1 levels by ELISA in the cell supernatants of six ESCC cell lines (KYSE-30, KYSE-150, KYSE-450, ECA-109, TE-1 and TE-10) after irradiation. The results suggested that, at the point of 2 h after 6 Gy-irradiation, DJ-1 was detected in the supernatants of all 6 cell lines and the two cell lines with the highest DJ-1 content in the cell supernatant were KYSE-150 and ECA-109 (Fig. 1E and F). We then selected these two cell lines for detailed content-dose-time analysis. These two ESCC cell lines (KYSE-150 and ECA-109) were irradiated with 0, 2, 4, 6, 8, and 10 Gy doses, and DJ-1 levels in the medium were measured at 0.5, 1, 2, 4, 8, 12, and 24 h after irradiation. The results indicated that the DJ-1 content in the medium reached a peak at 2 h after 8 Gy irradiation, which was approximately 798 and 727 pg/ml (Fig. S3A and B). At 2 h, with the increase in radiation dose, DJ-1 levels in the medium gradually increased, and plateaued after the dose increased to 8–10 Gy (Fig. 1G and H). The dose-time gradient after 8 Gy irradiation showed that the level of DJ-1 in the medium increased within 0–2 h, maintained a certain level at 2–4 h, and decreased rapidly to the initial level after 8 h (Fig. 1I and J). Meanwhile, Western blotting was used to detect the expression levels of DJ-1 in cells receiving 8 Gy irradiation at the corresponding time points. We found that the relative levels of DJ-1 reached their minimum at 2 h after irradiation (8 Gy) in both KYSE-150 and ECA-109 cell lines (Fig. S3C-F). These results demonstrated that irradiation induces the transport of DJ-1 from the intracellular to the extracellular regions of ESCC cells.
Exogenous DJ-1 enhances the metastasis of ESCC cells in vitro and vivo
Since DJ-1 is an important signal transduction molecule, we speculated that DJ-1 secreted by ESCC cells may be associated with RIBE. First, in vitro experiments were performed. To simulate the occurrence process of RIBE, ESCC cells irradiated with 8 Gy were co-cultured with unirradiated bystander cells for 24 h. Transwell and wound healing assays demonstrated that co-culture with irradiated ESCC cells enhanced the migration and invasion of bystander K150 and E109 cells (Fig. 2A-F). Interestingly, when we used exogenous DJ-1 to stimulate cells, the migration and invasion of bystander cells were also strengthened (Fig. 2A-F). In addition, we obtained the same result through a 3D tumor spheroid invasion assay (Fig. 2G and H). Next, to generate a pulmonary metastasis model, K150 and E109 cells stably transfected with luciferase were injected into the tail vein of nude mice to determine the effect of DJ-1 on ESCC tumor metastasis. As expected, exogenous DJ-1 through tail vein injection greatly promoted the invasion of ESCC cells and aggravated pulmonary metastasis compared to the control group (Fig. 2I-K). Altogether, in vitro and in vivo experiments confirmed that exogenous DJ-1 reinforces the metastasis of ESCC cells.
Exogenous DJ-1 promotes metastasis of bystander ESCC cells by activating TGF-β1 pathway
To explore the mechanism by which exogenous DJ-1 promotes metastasis, we first determined the localization of DJ-1 in ESCC cells. We labeled the recombinant protein DJ-1 with a His-tag and added it to the culture medium. Using immunofluorescence staining, we found that the His-tag DJ-1 was highly localized in the cytoplasm and partly in the nucleus (Fig. 3A). Fluorescence co-localization analysis showed that the site of His and DJ-1 double staining could be regarded as a recombinant DJ-1 protein, which strongly proved that DJ-1 can be freely transferred between ESCC cells (Fig. S4A).
Having determined the intracellular localization of DJ-1, we inferred that DJ-1 might play a role in signal transmission. To gain insight into the molecular mechanism of exogenous DJ-1 action on bystander cells, we analyzed the transcriptomes of bystander cells with and without exogenous DJ-1. The whole results of RNA-seq are shown in Table S6. According to the screening criteria of logFC > 0.7 and FDR < 0.05, 434 differentially expressed genes (DEGs) were discovered (Figs. 3B and S4B). KEGG enrichment was performed to demonstrate the top overrepresented canonical pathways sorted by p-value annotation for DEGs. The results indicated that the TGF-β1 pathway related to tumor metastasis was activated, which could be the reason for the metastasis of bystander ESCC cells (Fig. 3C). Based on the prediction of the protein interaction network of DEGs enriched in the TGF-β1 pathway, we found that Smad3 was the core of the network (Fig. 3D).
To verify the results of RNA-seq analysis, western blot was conducted to detect the expression of TGF-β1 pathway proteins and its downstream factors. The results showed that the ratio of p-Smad3/Smad3 increased, and the expression of TSP1, TGF-β1, and MMPs proteins increased to varying degrees (Fig. 3E). To investigate whether the TGF-β1 signaling pathway participates in DJ-1-induced metastasis, an inhibitor of the TGF-β1 signaling pathway (LY2109761) was used. As shown in Fig. 3F and G, LY2109761 reduced DJ-1-induced metastasis of bystander ESCC cells. Similar results were obtained in the 3D tumor spheroid invasion assay (Fig. S4C and D).
Exogenous DJ-1 induces metastasis by binding to HSC70 in ESCC cells
To further determine the regulatory mechanism of DJ-1 in the TGF-β1 pathway, we performed liquid chromatography-tandem mass spectrometry (LC–MS/MS) in K150 cell samples co-immunoprecipitated with the DJ-1 antibody. Figure S5A shows the co-immunoprecipitation samples stained with Coomassie Brilliant Blue after electrophoresis. A total of 2392 proteins were identified according to the iBAQ values (Fig. 4A and Table S7). To increase the accuracy of identification of interacting proteins by MS, we combined the protein data with the transcriptome data from RNA-seq for a comprehensive analysis. Three proteins (HSPA8, ACTA2 and MYH14) were identified by overlapping the identified proteins and RNA-seq DEGs (Fig. 4B).
Previous studies have reported that HSPA8/HSC70 promotes TGF-β-induced Smad2/3 activation in fibroblast NRK-49F cells [15]. Additionally, it has been shown that the upregulation of HSPA8/HSC70 is associated with poor prognosis in patients with multiple cancers and stimulates tumor immune responses or drug resistance [16,17,18,19]. We inferred that HSC70 may be an effective protein in DJ-1 regulating the TGF-β1 pathway and promoted metastasis of ESCC cells. To verify the interaction between DJ-1 and HSC70, we found that DJ-1 and HSC70 were co-localized in the cytoplasm of K150 and E109 cells using immunofluorescence (Fig. 4C, Fig. S5C and D). Co-immunoprecipitation and western blotting further demonstrated a protein interaction between DJ-1 and HSC70 (Fig. 4D).
To further investigate the role of HSC70 in DJ-1-induced metastasis, we transfected siRNA targeting HSPA8/HSC70 into K150 and E109 cells. The transfection efficiency was verified by western blotting (Fig. S5B). Surprisingly, HSPA8/HSC70 siRNA transfection drastically decreased the migration and invasion of K150 and E109 cells induced by exogenous DJ-1 in transwell and 3D tumor spheroid invasion assays (Fig. 4E-F and Fig. S5E-F). These results indicated that exogenous DJ-1 promotes metastasis by binding to HSC70 in ESCC cells.
Exogenous DJ-1 acting on HSC70 contributes to Smad3 phosphorylation and nuclear aggregation in ESCC cells
After identifying HSC70 as a downstream target of DJ-1, we speculated that the DJ-1/HSC70 complex may be concerned in the activation of Smad3. First, we transfected HSC70 siRNA or control siRNA into K150 and E109 cells. After stimulation with exogenous DJ-1, the phosphorylation status of Smad3 (p-Smad3) was examined by western blotting from 0 to 120 min. The ratio of p-Smad3/Smad3 reached a peak at the time point of 30 min in the control group, while the upward trend disappeared in cells transfected with HSC70 siRNA (Fig. 5A).
Next, the effect of HSC70 on the nuclear translocation of p-Smad3 was examined by immunofluorescence microscopy. As shown in Fig. 5B, both K150 and E109, in cells transfected with control siRNA, nuclear localization of p-Smad3 (green) was increased by DJ-1 stimulation. In contrast, in cells transfected with HSC70 siRNA, the level of HSC70 was completely suppressed (red), and the nuclear accumulation of p-Smad3 was also decreased (green) compared with the control group. The fluorescence intensity was significantly different between the two groups (Fig. S6A and B).
To explore the involvement of HSC70 in DJ-1-induced Smad3 signaling, the interaction between HSC70 and Smad3 was examined in K150 and E109 cells. The cells were treated for 30 min with or without exogenous DJ-1 and cell lysates were prepared with lysis buffer. HSC70 and Smad3 were immunoprecipitated from the lysate using specific antibodies. The immunoprecipitate was examined by western blotting using antibodies against HSC70 and Smad3. As shown in Fig. 5C, a small amount of Smad3 was immunoprecipitated with anti-HSC70 antibody in untreated cells. Upon stimulation with DJ-1, the levels of Smad3 increased in cells. Meanwhile, a low level of HSC70 was immunoprecipitated with an anti-Smad3 antibody in untreated cells. HSC70 expression also increased in cells stimulated by exogenous DJ-1. Collectively, these results suggested that the interaction between HSC70 and Smad3 is enhanced upon stimulation with exogenous DJ-1.
We also examined the effects of Smad3 on DJ-1-induced metastasis. We used siRNA to knock down the expression of Smad3 in ESCC cells and verified the knockdown efficiency by western blotting (Fig. S6C). As expected, knockdown expression of Smad3 significantly reduced DJ-1-induced migration and invasion of K150 and E109 cells in transwell and 3D tumor spheroid invasion assays (Fig. 5D-E and Fig. S6D-E). In summary, the above results demonstrated that DJ-1 binding to HSC70 accelerates the phosphorylation and nuclear aggregation of Smad3, which enhances the metastasis of ESCC cells.
Exogenous TGF-β1 reactivates Smad3 without the presence of HSC70
TGF-β1 is a secretory growth factor that leads to the activation of Smad or non-Smad pathways [20, 21]. Ikezaki M et al. demonstrated that HSC70 facilitated TGF-β-induced activation of Smad2/3 [15]. In our study, to further elucidate the regulatory effect of HSC70 on Smad3 and thoroughly investigate its relationship with TGF-β1, we used the recombinant protein TGF-β1 to stimulate ESCC cells together with DJ-1 while transfecting cells with HSC70 siRNA. Surprisingly, the results of transwell and 3D tumor spheroid invasion assays indicated that the migration and invasion of K150 and E109 cells were strengthened again after being induced by TGF-β1 (Fig. 6A-D). We then examined the activation of Smad3 in TGF-β1-stimulated ESCC cells. As shown in Fig. 6E-F, the immunofluorescence-stained images vividly demonstrated the re-aggregation of nuclear p-Smad3 in HSPA8-knockdown ESCC cells after stimulated by TGF-β1. Meanwhile, the western blot results also reflected the increased phosphorylation level of Smad3 in ESCC cells after TGF-β1 stimulation (Fig. 6G). These phenomena of Smad3-reactivation indicated that the effect DJ-1/HSC70 acting on Smad3 is separate and concurrent with that of TGF-β1 on Smad3.
Smad3 transcribes THBS1 to accelerate the metastasis of K150 cells
Considering the restricted regulatory effect of limited exogenous DJ-1, we inferred that there may be a positive feedback loop and DJ-1 probably acts as a catalyst. Smad3 has been widely recognized as an important transcription factor, especially transducing signals from TGF-β superfamily ligands, thus entering the nucleus and activating the transcription of a series of oncogenes [22]. In our study, western blot results suggested that with the passage of time (0–240 min), the ratio of p-Smad3/Smad3 gradually stabilized, and the expression of TSP1, TGF-β1, and MMPs gradually increased in K150 cells (Fig. 7A).
Since phosphorylation and nuclear aggregation of Smad3 are the reasons for DJ-1-induced metastasis, we studied the transcriptional role of Smad3 in K150 cells. Smad3 and DEGs enriched in the TGF-β1 pathway from the RNA-seq results were included in this study. We employed bioinformatics analysis using JASPAR (https://jaspar.genereg.net/) [23] to investigate Smad3 binding sites on the promoter region of above DEGs. We found that there was a Smad3 binding site in the THBS1 promoter region (-1956 ~ -1947), which ranked first in the prediction list (Fig. 7B).
To acquire a deeper understanding of THBS1 transcriptional regulation, we mutated the putative SMAD3-binding sites in the THBS1 promoter and transduced K150 cells with these constructs. Luciferase experiments showed that THBS1 promoter activity was accelerated when stimulated by DJ-1 but inhibited when the SMAD3-binding site was mutated (Fig. 7C). Moreover, we found that transfection with a plasmid overexpressing Smad3 promoted the promoter activity of THBS1 (Fig. 7D). Additionally, western blot analysis revealed that Smad3 silencing by siRNA prominently reduced the expression of THBS1 (Fig. 7E).
After identifying THBS1 as a transcriptional target of Smad3, we investigated its biological function in the metastasis of K150 cells. First, THBS1 siRNA was used to inhibit THBS1 expression in K150 cells. The transfection efficiency is shown in Fig. S7A. Importantly, the migration and invasion stimulated by DJ-1 were decreased in THBS1-knockdown K150 cells compared to control cells in transwell and 3D tumor spheroid invasion assays (Fig. 7F-G and Fig. S7B-C). LSKL, an inhibitor of TSP1, is a latency-associated protein (LAP)-TGFβ-derived tetrapeptide and competitive TGF-β1 antagonist [24]. Therefore, we tested the effect of LSKL on DJ-1-induced metastasis. Transwell and 3D tumor spheroid invasion assays demonstrated that the use of LSKL significantly reduced the migration and invasion induced by DJ-1 in K150 cells (Fig. 7H-I and Fig. S7D-E). In general, these results indicate that after activation by exogenous DJ-1, Smad3 transcribes THBS1 to promote the metastasis of K150 cells.
Blocking TSP1 weakens the phosphorylation and nucleation of Smad3 and reduces the metastasis of K150 cells in vivo
Thrombospondin 1 (TSP1), encoded by THBS1, has been reported as an activator on latent TGF-β. Previous studies have reported that TSP1 interacts with the LSKL sequence of the N-terminal domain of the LAP of latent TGF-β and induces a conformational change at this site to improve the accessibility of TGF-β to its receptor [25, 26]. In addition, our previous experiments demonstrated that TGF-β1 reactivates Smad3 and enhances the migration and invasion of HSC70-knockdown ESCC cells. Therefore, we speculated that Smad3, TSP1, and TGF-β1 might constitute a positive cycle altogether.
To verify our hypothesis, we first transfected siRNA targeting THBS1 into K150 cells. Results of western blotting indicated that compared to the control group, the ratio of p-Smad3/Smad3 in THBS1-knockdown cells suffered a significant decline. At 240 min after DJ-1 stimulation, the p-Smad3/Smad3 ratio was even lower than the initial level. Furthermore, the expression of TGF-β1 and MMPs in the THBS1-knockdown group was markedly decreased compared to that in the control group (Fig. 8A). The translocation of p-Smad3 was examined by immunofluorescence microscopy. As presented in Fig. 8B-C, the level of nuclear aggregation of p-Smad3 greatly decreased in the group that cells transfected with THBS1-knockdown siRNA. Next, we proved that LSKL reduced the phosphorylation and nuclear aggregation of Smad3 using the same methods (Fig. 8D-F). These data revealed that TSP1 responsively activated Smad3 by interacting with TGF-β1, which formed a positive feedback loop.
In addition, to confirm that DJ-1 was a regulator rather than an effector involved in the positive feedback loop, we examined the mRNA levels of PARK7 in ESCC cells by qPCR. The results suggested that exogenous DJ-1 did not strengthen the transcription level of DJ-1 (Fig. S7F).
Having validated the existence of the Smad3/TSP1/ TGF-β1 positive cycle pathway, we predicted that it may be the core effective signaling of DJ-1-induced metastasis and blocking the function of TSP1 or TGF-β1 could interrupt this effect. We established mouse model of pulmonary metastasis to test the therapeutic effects of LY2109761 and LSKL. The results revealed that while DJ-1 was injected through the tail vein equivalently, intraperitoneal injection injected LY2109761 or LSKL greatly suppressed the invasion of K150 cells and alleviated pulmonary metastasis compared with the control group (Fig. 8G-I).
DJ-1 regulatory pathway proteins are correlated mutually in patient samples and serum DJ-1 has prognostic value in ESCC patients receiving radiotherapy
To further verify the rationality of the DJ-1 regulatory pathways, we collected 36 ESCC tissues for histological analysis. First, we performed immunohistochemical staining for DJ-1, HSC70, p-Smad3, TSP1 and TGF-β1. We calculated the sum of the optical density per unit area and analyzed the correlation of the integrated optical density (IOD) between the detected targets. The results indicated that there were close correlations between the IOD of DJ-1-HSC70, DJ-1-p-Smad3, DJ-1-TGF-β1, and DJ-1-TSP1 (R2 = 0.281, p = 0.0009; R2 = 0.2401, p = 0.0024; R2 = 0.1913, p = 0.0076; and R2 = 0.2858, p = 0.0008, respectively) (Fig. S8A-B). Moreover, immunofluorescence staining for DJ-1, HSC70, p-Smad3, TSP1 and TGF-β1 was performed, and the IOD per area was used as the statistical indicator. Consistent with the IHC results, close correlations of IOD for DJ-1-HSC70, DJ-1-p-Smad3, DJ-1-TGF-β1, and DJ-1-TSP1 (R2 = 0.331, p = 0.0002; R2 = 0.1955, p = 0.0069; R2 = 0.274, p = 0.0011; and R2 = 0.1741, p = 0.0113, respectively) were observed (Fig. 9A-B). Additionally, to validate the expression results of histological staining, we used qRT-PCR to determine the expression of the five targets at the mRNA level. The -log(2^-Δct) score was used as a statistical indicator. As presented in Fig. 9C-D, we observed positive correlations for DJ-1-HSC70, DJ-1-p-Smad3, DJ-1-TGF-β1, and DJ-1-TSP1 (R2 = 0.4556, p < 0.0001; R2 = 0.2405, p = 0.0024; R2 = 0.2643, p = 0.0013; and R2 = 0.1656, p = 0.0138, respectively). Taken together, these data provide evidence for the existence of DJ-1 regulatory pathways at the human tissue level.
Having determined that DJ-1 induced metastasis in ESCC cells and illustrated its mechanism, we sought to evaluate the predictive value of DJ-1 in survival after radiation in ESCC patients. Follow-up information was collected from 46 patients who had serum samples collected after cumulative 40 Gy dose radiotherapy. Kaplan–Meier survival analysis revealed that high DJ-1 expression in serum after radiotherapy predicted poor overall survival (OS) in patients with ESCC (Fig. S8C, n = 46, p = 0.0074), while no significant difference was observed in disease free survival (DFS) between the two groups of ESCC patients (Fig. S8D, n = 46, p = 0.1228). To further evaluate the predictive value of DJ-1 regulatory pathways in the prognosis of patients with ESCC, we calculated the hazard ratios of PARK7, HSPA8, Smad3, THBS1, and TGFB1 in the GSE53625 dataset. However, we did not find any statistically significant differences (Fig. S8E). In summary, although we did not find any value for DJ-1 regulatory pathways in predicting the prognosis of patients with ESCC, serum DJ-1 level could be a useful biomarker to predict survival in patients with ESCC receiving radiotherapy, which still needs to be validated using a large sample of clinical data.