SEC23A confers ER stress resistance in gastric cancer by forming the ER stress-SEC23A-autophagy negative feedback loop

A total of 82 paired GC and adjacent normal tissues were obtained at the First Affiliated Hospital of Nanjing Medical University for comparisons of SEC23A expression between paired tumor and paratumor tissues. Of these, 30 pairs of tissues were used for immunohistochemical staining, 12 pairs for western blotting, and 82 pairs were used for qRT-PCR. None of these patients have received chemotherapy or radiotherapy. The research has been endorsed by the Medical Ethics Committee of the First Affiliated Hospital of Nanjing Medical University.

MKN45 and HGC27 human GC cell lines were purchased from Cell Bank of the Chinese Academy of Sciences (Shanghai, China) and were cultured in RPMI-1640 (Wisent, 350–000-CL) supplemented with 10% fetal bovine serum (FBS) (Gibco, 10270–106) in an incubator supplemented with 5% CO₂ and operating at 37 °C.

To establish stable SEC23A-knockdown and SEC23A-overexpressing GC cells, lentivirus vectors containing short-hairpin RNA (shRNA) sequences against SEC23A (shSEC23A-1 and shSEC23A-2), SEC23A sequence plasmid (oeSEC23A) and their control vectors (shNC and vector) were obtained from GEMA (Shanghai, China).

For the silencing of STAT3, ATF6, ATF4, CHOP, XBP1 and CREB3L2, MKN45 and HGC27 cells were transiently transfected with small interfering RNAs (siRNAs) (GEMA, Shanghai, China) targeting these mRNAs and their negative controls by using the Lipofectamine 3000 Reagent (Invitrogen, L3000015) according to the protocol.

All the sequences are listed in Supplementary Table S1.

Total RNA was extracted from cultured cells as described in a previous study [27]. The cDNA synthesis and quantitative PCR assays were performed using HiScript III All-in-one RT SuperMix Perfect for qPCR (Vazyme, R333-01) and AceQ qPCR SYBR Green Master Mix (Vazyme, Q141-02) according to the kit protocols. All samples were measured with QuantStudio 7 system (Thermo Fisher Scientific, USA) in 96-well plate. The relative expression levels were standardized to the internal control GAPDH using the ?Ct method.

The primers used for qPCR were listed in Supplementary Table S2.

Cell protein was extracted with RIPA Lysis Buffer (Beyotime, P0013B) and the concentration was detected by Enhanced BCA Protein Assay Kit (Beyotime, P0010). Western blotting was performed as described previously [27]. The results were measured by ImageJ software.

The antibodies were used for western blotting are listed in Supplementary Table S3.

The animal procedures were approved by the Committee on the Ethics of Animal Experiments of Nanjing Medical University. All the mice were purchased from the Animal Center of Nanjing Medical University. For the subcutaneous tumor formation models, five mice per group, 2 × 10⁶ shNC and shSEC23A MKN45 cells were diluted with PBS to 100 µl and then injected into the axilla of the forelimbs of 4- to 5-week-old nude mice. 12 days after inoculation, mice were intraperitoneally injected with drugs respectively (100 µl/d PBS for the control group, 0.5 mg/kg/d TM or 0.5 mg/kg/d 5-FU for the treatment groups) every four days. The tumor volume was monitored every 2 days postinjection and calculated with the following formula: V = p /6 (Length × Width²). At the 4th week postinoculation, animals were sacrificed and tumors were collected and weighed.

Expression differences and survival analyses of SEC23A in TCGA dataset were obtained from GEPIA2 database. Survival analyses in GEO dataset and 82 GC patients from our center, univariate and multivariate Cox regression analysis, GSEA, and Spearman correlation analysis were performed with RStudio (1.4.1717).

All experiments were preformed three times independently and the results are reported as mean ± SD.Comparisons of SEC23A expression between paired tumor and paratumor tissues were performed by paired Student’s t-test. The relationship between SEC23A expression and clinicopathological characteristics was analyzed with chi-square test. Differences between experimental and control groups were analyzed with unpaired Student’s t-test for parametric tests and Wilcoxon signed-rank test for nonparametric tests. Values of p < 0.05 were considered statistically significant. *p < 0.05; **p < 0.01; ***p < 0.001; ****p < 0.0001.

According to the TCGA and GTEx datasets, we found that SEC23A was highly expressed in GC tissue (Fig. 1A). In addition, the GEPIA2 database revealed that GC patients with a higher SEC23A expression exhibited shortened overall survival and disease-free survival (Fig. 1B). These results suggested that SEC23A may take an essential part in the development of GC. According to qRT-PCR and western blot data, SEC23A was considerably upregulated in GC tissues when compared to adjacent normal tissues (Fig. 1C and D). IHC staining and the H-Score also confirmed the upregulation of SEC23A in GC tissues (Fig. 1E and F). The clinicopathologic and survival information for the 82 GC patients from our center showed that high SEC23A expression was associated with more advanced tumors and worse prognosis (Table 1 and Fig. 1G). Consistent survival analysis results were obtained from the GSE62254 and GSE15459 datasets (Fig. 1H). Univariate and multivariate Cox regression analyses with the TCGA dataset revealed high SEC23A expression was an independent indicator for predicting the overall survival of GC patients (Fig. 1I and J). Collectively, these results demonstrated that SEC23A was highly expressed in GC tissues and was associated with a poor prognosis.

To develop a basic understanding of the roles and mechanisms of SEC23A on GC, we conducted GSEA using TCGA dataset. Results indicated a significant correlation between SEC23A and UPR (Fig. 2A). Since UPR is the symbolic pathway of ER stress, we initially performed immunofluorescence staining of SEC23A in human GC tissues with GRP78/BiP, an ER stress marker in the UPR pathway [28]. We discovered that BiP-positive regions displayed significantly higher SEC23A staining level, suggesting that SEC23A was associated with ER stress (Fig. 2B and C). To determine the relationship between SEC23A and ER stress, we first tested whether ER stress influences SEC23A expression in GC cells. MKN45 and HGC27 cells were treated with the ER stress inducers dithiothreitol (DTT) and tunicamycin (TM). qRT-PCR and western blotting results revealed that both DTT and TM treatments induced significant increases in SEC23A expression (Fig. 2D-G; Fig. S1A and B). Moreover, treatment with 4-PBA, an ER stress inhibitor, apparently reduced SEC23A expression (Fig. 2H and I). Comparable outcomes were obtained when we used different ER stress inducers, such as 5-FU and H₂O₂, to mimic the ER stressors that GC cells typically encounter (Fig. 2J-L; Fig. S1C and D). Then we further conducted immunofluorescence staining of SEC23A in human normal stomach tissues with BiP, found the same spatial expression association between SEC23A and BiP (Fig. S1E and F). In the BiP-positive region, SEC23A and BiP fluorescence intensity showed a significant positive correlation (Fig S1G). Moreover, the normal stomach tissues displayed lower expression of SEC23A and BiP than GC tissues (Fig. S1H and I). There was a positive correlation between SEC23A and BiP expression (Fig. S1J). Based on the results, we concluded that SEC23A was upregulated in response to ER stress and the increased ER stress may contributed to the high expression of SEC23A in GC tissues. Then we constructed stable SEC23A overexpressing or SEC23A knockdown GC cells using lentiviruses to explore whether SEC23A regulates ER stress (Fig. S1K and L). According to western blotting data, SEC23A knockdown increased BiP expression (Fig. S1M), whereas SEC23A overexpression decreased BiP expression, indicating that SEC23A may reduce the severity of ER stress (Fig. S1N). The expression relationship between SEC23A and BiP demonstrated by western blotting contradicts the results described above. We hypothesized that ER stress increases SEC23A expression, which serves in turn as a protective mechanism to relieve ER stress and promotes GC cells ER stress resistance.

Previous studies have shown that cells activate the UPR in response to ER stress, including IRE1a, ATF6a and PERK pathways [29]. Due to the crosstalk between these UPR pathways, it is inappropriate to investigate the mechanism by which ER stress promotes SEC23A expression by only blocking only one of them. TM treatments increased SEC23A expression at both mRNA and protein levels, indicating that ER stress may stimulate SEC23A transcription. We first verified whether the several critical transcription factors in the UPR were upstream of SEC23A, including ATF4, ATF6, XBP1, CHOP, and CREB3L2. The findings demonstrated that suppression of these transcription factors had no discernible effect on SEC23A expression during ER stress (Fig. S2A-E). Then we explored the transcription factors that could potentially associate with the promoter region of SEC23A using the UCSC genome browser and JASPAR database. We found STAT3 may bind to the SEC23A promoter, and there was a significantly positive correlation between STAT3 and SEC23A mRNA expression in GC using GEPIA2 database (Fig. 3A). ER stress has been reported to be associated with the nuclear accumulation of pY705-STAT3 via the phosphorylation of PERK and subsequent activation of the JAK2-STAT3 pathway [30, 31]. To verify the involvement of this pathway in ER stress-induced SEC23A expression, we inhibited STAT3 using siRNA in GC cells and detected SEC23A expression with TM treatment. As expected, suppression of STAT3 expression attenuated ER stress-induced SEC23A upregulation, STAT3 inhibitor S3I-201 treatment produced the same outcomes (Fig. 3B, C). According to JASPAR database predicted results, three classical STAT3 binding motifs were identified in the promoter of SEC23A (2000 bp upstream of the transcription start site). Therefore, we constructed three luciferase reporter plasmids named BS1, BS2 and BS3, containing these motifs to further determine whether STAT3 transcriptionally regulates SEC23A (Fig. 3D). Only BS2 exhibited increased luciferase activity upon TM treatment, and this increase was diminished by STAT3 inhibition (Fig. 3D and E). The regulation of BS2 luciferase activity following TM treatment was blocked in the mutant plasmid, indicating that the second binding site was a positive STAT3 binding site on the SEC23A promoter (Fig. 3E). ChIP-qPCR assays showed that STAT3 could directly bind to the SEC23A promoter region, and the occupation of STAT3 on the SEC23A promoter was increased following TM treatment (Fig. 3F). Taken together, our data indicated that STAT3 was an upstream transcriptional factor of SEC23A during ER stress. Consistent with reported results, TM treatment promoted STAT3 phosphorylation at Tyr705 in both the cytoplasm and the nucleus, as shown by western blotting (Fig. 3G) and immunofluorescence confocal microscopy (Fig. 3H). Furthermore, increased SEC23A expression resulting from ER stress was diminished by JAK2 inhibitor AG490 and STAT3 inhibitor S3I-201 treatments (Fig. 3I). Taken together, these data indicated that activation of the JAK2-STAT3 pathway contributes to ER stress-induced SEC23A upregulation in GC cells.

We conducted additional functional experiments to ascertain whether increased SEC23A expression is a factor that protects GC cells from ER stress-induced apoptosis. As determined by CCK-8 and colony formation assays, SEC23A suppression significantly decreased cell viability in GC cells treated with TM (Fig. 4A-D), whereas SEC23A overexpression significantly increased GC cell resistance to ER stress (Fig. S3A-C). The effect of SEC23A on the apoptosis of TM-treated cells was further verified by flow cytometry using annexin V-PI labeling and western blotting against c-caspase3 and c-PARP (Fig. 4E-G; Fig. S3D-H). Then we subcutaneously inoculated the shNC and shSEC23A-1 MKN45 cells into nude mice to verify the roles of SEC23A in GC growth and ER stress adaptation in vivo. Four experimental groups were established: PBS + shNC, PBS + shSEC23A-1, TM + shNC, and TM + shSEC23A-1. The tumor volume and weight showed that SEC23A knockdown could inhibit tumor growth and enhance MKN45 cells sensitivity to ER stress in vivo (Fig. 4H-J). C-caspase3 and BiP staining and Tunel assays showed that the shSEC23A-1 group with or without TM treatments exhibited more apoptotic cells and ER stress severity, suggesting that shSEC23A-1 cells were more susceptible to apoptosis and more sensitive to ER stress than shNC cells in vivo (Fig. 4K-M and O). The immunohistochemistry results of the shSEC23A-1 group revealed less Ki-67 staining in tumor tissues than the shNC group, with or without TM treatment (Fig. 4K and N). In conclusion, SEC23A promoted GC cells’ survival advantage during ER stress, whether from baseline or extra stimuli.

Previous studies have reported that SEC23A is involved in elevating autophagy [26]. Autophagy has been linked to promotion of ER stress adaptation in cancer cells [32, 33]. Therefore, we explored whether SEC23A enhanced autophagy to protect GC cells from ER stress-induced cell death. Western blotting revealed that the shSEC23A-1 and shSEC23A-2 cells displayed a lower expression of LC3B-II and ATG5, and a higher expression of SQSTM2/p62 than the shNC cells (Fig. 5A). Consistently, the oeSEC23A GC cells exhibited higher LC3B-II and ATG5, lower p62 expression compared to the vector GC cells (Fig. 5B). Confocal microscopy observations demonstrated that SEC23A knockdown decreased autophagosomes and autolysosomes numbers (Fig. 5C and D). SEC23A overexpression elevated the accumulation of autophagosomes and autolysosomes (Fig. S4A and B). These results suggested that SEC23A enhanced basal autophagic flux in GC cells. Furthermore, TM treatments induced significant increases in the number of autophagosomes and autolysosomes, whereas SEC23A suppression attenuated the accumulation of autophagosomes and autolysosomes induced by TM treatment (Fig. 5C and D). Moreover, western blotting results showed the increased LC3B-II and ATG5, decreased p62 following TM treatments were reversed by SEC23A suppression (Fig. 5E). The same results were obtained through electron microscopy observation (Fig. 5F). Accordingly, it was concluded that SEC23A contributed to the autophagy activation in GC cells under ER stress.

To explore the mechanisms by which SEC23A induces autophagy in GC cells undergoing ER stress, we preformed IP assay and mass spectrometry to identify the proteins which SEC23A interacts in MKN45 cells under ER stress (Fig. 6A). According to the results, Annexin A2 (ANXA2) was bound to SEC23A with the highest score (Fig. 6B). Co-IP and immunofluorescence assays further validated the binding and colocalization of SEC23A and ANXA2 (Fig. 6C and D; Fig. S5A). ANXA2 is an annexin family protein involved in multiple biological processes, including epithelial–mesenchymal transition, chemoresistance and autophagy [34‐36]. We hypothesized that ANXA2 contributes to SEC23A induced autophagy and ER stress relief. Western blotting results showed that knockdown or overexpression of SEC23A had no effect on ANXA2 expression during ER stress (Fig. S5B). Previous research has demonstrated that the spatial localization of ANXA2 affects how it functions. Cytoplasmic ANXA2 prevents autophagy by combining with TFEB, a dominant transcription factor for autophagy, and rendering it inactive [34]. While ANXA2, which is located at the plasma membrane, promotes autophagy by boosting the production of plasma membrane-derived ATG16L⁺ autophagosomes [35]. We speculated that SEC23A regulated the subcellular localization of ANXA2. Subcellular fractionation western blotting and cellular immunofluorescence performed on TM-treated GC cells, we first demonstrated that ANXA2 was predominantly localized in the cytoplasm when SEC23A was suppressed, whereas ANXA2 cell membrane localization was dramatically increased when SEC23A was overexpressed, indicating that SEC23A facilitated localization of ANXA2 on cell membranes during ER stress (Fig. 6E-G; Fig. S5C-E). The interaction between ANXA2 and TFEB was decreased following SEC23A knockdown (Fig. 6H; Fig. S5F). SEC23A knockdown reduced the colocalization between ANXA2 and ATG16L1, indicating SEC23A knockdown may inhibit the production of ATG16L⁺ autophagosomes by regulating ANXA2 plasma membrane trafficking (Fig. 6I and J; Fig. S5G and H). Then we treated oeSEC23A GC cells with PY-60, a chemical that liberates ANXA2 from the plasma membrane by inhibiting its binding phosphoinositide [37]. The protection during ER stress induced by SEC23A was attenuated by PY-60 treatment (Fig. S5I). PY-60 treatment decreased LC3B-II and increased p62 expression in TM treated GC cells, and PY-60 reversed the SEC23A-induced decreases in BiP, c-caspase3 and c-PARP expression (Fig. 6K; Fig. S5J and K). Collectively, these findings suggested that SEC23A promotes autophagy and ER stress relief by regulating the localization of ANXA2 at the plasma membrane.

To confirm the contribution of autophagy to ER stress survival advantage conferred by SEC23A, we severally applied the autophagy activator RAPA on shSEC23A GC cells and the autophagy inhibitor CQ on oeSEC23A GC cells. Results from the CCK8 assay revealed that shSEC23A-1 + RAPA group performed with greater cell viability than the shSEC23A group following TM treatment (Fig. 7A). After receiving CQ treatment, the cell viability of oeSEC23A group was dramatically decreased (Fig. 7B). Colony formation assays produced comparable findings (Fig. 7C and D). We next conducted flow-cytometric apoptosis assays and western blotting against c-caspase3 and c-PARP to detect cell apoptosis. RAPA treatments reversed the increased cell apoptosis observed in shSEC23A GC cells (Fig. 7E-G, I and J). Moreover, CQ treatments decreased ER stress resistance in oeSEC23A GC cells (Fig. 7E, F, H, K and L). These results further verified that autophagy contributes to ER stress adaptation induced by SEC23A in GC cells.

Anticancer drugs, such as sorafenib, bortezomib, cisplatin, 5-FU, and paclitaxel have been shown to cause cancer cell death by triggering ER stress-mediated apoptosis [38‐42]. Various tumors acquire chemoresistance as a result of ER stress adaptation [9, 43, 44]. Resistance to chemotherapy can be lessened by treatments that target ER stress [45‐50]. Inhibition of cytoprotective autophagy could enhance chemotherapeutic efficacy [51‐54]. Considering the roles that SEC23A played in ER stress relief and autophagy, we next investigated whether SEC23A influences the effectiveness of chemotherapy. We treated GC cells with 5-FU, a first-line chemotherapeutic agent against GC [1]. Cell transmission electron microscopy and western blotting experiments indicated that 5-FU triggered ER stress, and knockdown of SEC23A further increased the severity of ER stress induced by 5-FU (Fig. 8A and B). To unveil the possible link connecting SEC23A and chemotherapy effect, we conducted functional experiments in vitro and vivo. SEC23A knockdown significantly sensitized GC cells to 5-FU and overexpressed SEC23A decreased 5-FU-induced cell death, as evidenced by flow cytometry and cell viability (Fig. 8C and D; Fig. S6A-C). Compared to shSEC23A-1 group, autophagy activator RAPA reversed the decreased cell viability and increased cell apoptosis rate (Fig. 8C and D; Fig. S6A). Furthermore, autophagy inhibitor CQ apparently weakened cell viability and increased cell apoptotic rate in SEC23A overexpressed GC cells treated with 5-FU (Fig. S6A-C). The colony formation assays and western blotting against c-caspase3 and c-PARP obtained similar results (Fig. 8E and F; Fig. S6D-H). Under 5-FU treatments, SEC23A knockdown inhibited tumor growth, as demonstrated by the volume and weight of tumors (Fig. 8G and H). These results were also validated by Ki67, c-caspase3, BiP and tunel staining (Fig. 8I-L). Combined with the previous findings, we concluded that SEC23A attenuated 5-FU efficacy through autophagy-induced ER stress relief in GC cells.

GEPIA2: http://​gepia2.​cancer-pku.​cn/​

JASPAR database: https://​jaspar.​genereg.​net/​

UCSC genome browser: http://​genome.​ucsc.​edu/​