CST1 is up-regulated in primary and metastatic GC and is associated with a poor prognosis

To identify the differential genes associated with the peritoneal metastasis of GC, we first performed RNA-seq analysis to compare their expression levels between 4 paired primary GC tissues and adjacent normal tissues. Genes with a p value <0.05 were only considered. The RNA-seq results revealed 3904 different genes, and the heatmap showed the highest and lowest 20 expression genes (Fig. 1A). Among differentially expressed genes, CST1 was significantly up-regulated in primary GC tissues. Volcano plots showed consistent results (Fig. 1B).

Fig. 1: CST1 is up-regulated in primary and metastatic GC tissues and is associated with a poor prognosis.

Heatmap mainly showing expression levels of 20 up-regulated and 20 down-regulated different expressed genes in primary GC (A) and metastatic GC (C) vs. adjacent normal tissues (>1.5-fold). Volcano plots with differentially expressed genes in primary GC (B) and metastatic GC (D) vs. adjacent normal tissues. E Venn diagram representation of 768 overlapped up-regulated genes. F Histogram was shown the overlapped highest 20 up-regulated genes in these two clusters. G, H Higher expression of CST1 was found in GC samples than the matched normal tissues (based on GSE54129 and GSE66229 database). I, J Kaplan–Meier plots of overall survival and progression-free survival for GC samples from the KM Ploter database. K RT-qPCR analysis of CST1 mRNA expression in 100 pairs of GC patient samples. Data are shown as the mean ± SD of triplicate independent sets of experiments; statistical significance was assessed by paired t-test, *<0.05, n = 100. L Western blot analysis was performed using an antibody against CST1 in 5 pairs of GC patients’ samples (upper panel); protein band intensities were measured by ImageJ software and normalized to GAPDH (lower panel). M IHC staining was performed using an antibody against CST1 and representative photographs of CST1 in GC patients. Scale bar: 100 μm. N IHC stain scoring of CST1 in 185 GC tissues and 185 normal tissues, statistical significance was assessed by unpaired t-test, ****<0.0001.

Then, we performed RNA-seq analysis again in 4 paired peritoneal metastasized GC tissues. The corresponding controls, heatmap, and volcano plots showed that CST1 was also up-regulated in this cohort (Fig. 1C, D). Then, the Venn diagram of the up-regulated genes showed the numbers of common and specific genes in two clusters, screening 768 common up-regulated genes. CST1 was up-regulated in GC and peritoneal metastasized GC tissues (Fig. 1E, F).

To validate our results, we evaluated the expression levels of CST1 in the matched pairs of GC and normal tissues samples using the GEO RNA-seq database, including GEO GSE54129, and GSE66229 database (Fig. 1G, H), GEO GSE79973, GSE26899, GSE13911, GSE19826 database (Fig. S1A), and TCGA database (Fig. S1B), consistent with our observation, the result from these datasets showed significantly up-regulated CST1 expression in GC tissue samples. Additionally, the expression of CST1 was positively correlated with the clinical stage of gastric cancer, but the difference was not statistically significant (Fig. S1C). More interestingly, overall survival and progression-free survival were lower in patients with higher CST1 levels in the Kaplan-Meier Plotter online database (Fig. 1I, J).

To further test the expression of CST1 in GC tissues, RT-qPCR was performed on 100 paired GC and adjacent normal tissues, showing that CST1 was up-regulated in GC tissues (Fig. 1K). In addition, Western blot analysis of 5 paired GC and adjacent normal tissues revealed that CST1 expression was higher in GC than in paired normal tissues (Fig. 1L). At the same time, the immunohistochemical analysis result showed that the CST1 was highly expressed in the GC tissues regardless of their differentiation grade (Fig. 1M); the upper panel was poorly differentiated while the down panel was well differentiated. In addition, IHC of 185 paired GC and normal tissues revealed that CST1 was highly expressed in GC tissues (Fig. 1N).

CST1 promotes GC cell metastasis but has no effect on cell proliferation in vitro

Next, we examined the CST1 mRNA and protein expression levels in GC cell lines. Higher expression of CST1 mRNA was found in AGS, BGC823, and MKN45 cell lines than in HGC-27 and SNU-1 cell lines (Fig. 2A, B). We then used HGC-27 cells to stably overexpress CST1, AGS and MKN45 cells to knockdown CST1 (Fig. 2C and S2A) and to examine the role of CST1 on cell behavior in vitro. CCK8 and plate clone formation assays suggested that the overexpression of CST1 had no effect on HGC-27 cells compared to control cells; similar results were obtained for MKN45-sh1-CST1 and MKN45-sh2-CST1 vs. control cells and CST1 knockdown AGS stable cell lines (Fig. 2D, E and S2B, C). However, the wound healing assay indicated that CST1 could promote cell migration. HGC-27-CST1 cells showed faster migration while MKN45-sh1-CST1/MKN45-sh2-CST1 cells and CST1 knockdown AGS stable cell lines showed slower migration compared to the negative control group (Fig. 2F and S2D).

Fig. 2: CST1 promotes GC cell metastasis but not proliferation in vitro.

A, B RT-qPCR and Western blot analysis showing the expression of CST1 in different GC cell lines. Total GAPDH was used as a loading control. C Western blot analysis of HGC-27 and MKN45 stably transfected with CST1 overexpression/knockdown lentiviruses and control lentiviruses. Total GAPDH was used as a loading control. D CCK8 assay analyzed the proliferation of HGC-27-Vector/HGC-27-CST1 and MKN45-shNC/MKN45-sh1-CST1/MKN45-sh2-CST1 stable cell lines. Data are shown as the mean ± SD of triplicate independent sets of experiments; statistical significance was assessed by paired t-test. E A colony formation assay. Left panel: representative images, right panel: quantification analysis. Data from independent experiments are presented as the mean ± SD. Statistical was assessed by unpaired t-test, ns means no significance. F Wound healing analysis for assessing migration of HGC-27-Vector/HGC-27-CST1 and MKN45-shNC/MKN45-sh1-CST1/MKN45-sh2-CST1 at 0, 24, and 48 h. Representative images (left panel) and quantification (right panel) are shown as indicated. Data from independent experiments are presented as the mean ± SD. Statistical significance was assessed by an unpaired t-test. ***p < 0.001. Scale bar: 100 μm. G Transwell migration and Matrigel invasion assays were performed to assess migration and invasion ability of CST1-overexpression and knockdown stable cell lines. Representative images (left panel) and quantification (right panel) are shown as indicated. Data from independent experiments are presented as the mean ± SD. Statistical significance was assessed by an unpaired t-test. ***p < 0.001. Scale bar: 100 μm.

To further assess the contribution of CST1 to the development of migratory and invasive phenotypes of GC cells, migration and invasion experiments were conducted using both HGC-27 and MKN45 and AGS cells. Data indicated that the overexpression of CST1 significantly increased the migration and invasion of HGC-27 cells compared with its control, while the migration of MKN45 and AGS cells was reduced (Fig. 2G and S2E). Collectively, these observations suggest that CST1 is a positive regulator of migration and invasion in GC cells.

CST1 interacts with GPX4 to improve the stability of the GPX4 protein

To investigate the mechanism underlying the CST1 in GC cells, we analyzed MKN45 cells by immunoprecipitation (IP) assay using CST1-specific antibody (IgG antibody was used as a control group). The samples were then separated using SDS-PAGE and analyzed by a liquid chromatograph-mass spectrometer (LC-MS); the experiment was performed three times (Fig. 3A). Silver staining of the IP cell lysates revealed that CST1 was successfully pulled down and further to LC-MS (Fig. 3B). Venn diagram revealed that 63 proteins were enriched by CST1 antibody compared to those in the IgG control samples (Supplementary File 4). The data of the three time-mass spectrometry eliminated the contamination proteins (Fig. 3C). Kyoto Encyclopedia of Genes and Genomes (KEGG) pathways analysis of these 63 proteins further revealed that the ferroptosis pathway was highly enriched (Fig. 3D) and GPX4 was a key protein in the ferroptosis pathway.

Fig. 3: CST1 interacts with GPX4 to improve the stability of GPX4 protein.

A Schematic of CST1 interactor discovery. The potential interactors were presented at higher levels in the CST1 experimental group than in the IgG control group. Three replicates of IP-MS were conducted. B Silver staining of IP cell lysates. C Venn diagram of three times of LC-MS results showing 63 co-upregulated proteins. D KEGG pathways analysis of 63 co-upregulated proteins. E Immunoprecipitation of the CST1 protein by an anti-GPX4 antibody in MKN45 cells. IgG was used as a negative control. F Immunoprecipitation of the GPX4 protein by an anti-Flag antibody in HEK293 cells transfected with pcDNA3.1-FLAG-CST1. PcDNA3.1-vector was used as a negative control. G Western blot showing GPX4 expression in HGC-27 and MKN45 stable cell lines; total GAPDH was used as a loading control. H, I Degradation of the GPX4 protein was measured after the treatment of 200 µg/ml CHX at the indicated time points in HEK293T, which transfected with Flag-tagged CST1 expression plasmids and Myc-tagged GPX4 expression plasmids. J, K Western blot analysis of GPX4 expression after treatment with 20 µM MG132 for 4 h in MKN45 stable cell lines. Data were expressed as a fold-change relative to control. L Analysis of GPX4 ubiquitination was performed by immunoprecipitation using an anti-Myc antibody, followed by immunoblot with anti-HA antibody and anti-Myc antibody in HEK293T cells transfected with the indicated constructs.

To further demonstrate the protein-protein interaction between CST1 and GPX4, we first performed Co-IP experiments using the MKN45 cells overexpressing CST1 (Fig. 3E). Consequently, a strong binding between CST1 and GPX4 was found. Next, the HEK293T cells were transiently transfected with Flag-tagged CST1 expression plasmids, followed by Co-IP assays with an anti-Flag antibody. The result showed that CST1 and GPX4 proteins bound to each other in these cells (Fig. 3F). Collectively, these observations suggested that CST1 interacting with GPX4 may induce the malignant progression of GC.

Next, we examined whether CST1 modulates the mRNA or protein levels of GPX4 through their interaction with each other. To test this hypothesis, HGC-27-Vector/HGC-27-CST1 and MKN45-shNC/MKN45-sh1-CST1/MKN45-sh2-CST1 cells were established. The levels of CST1 and GPX4 expression were confirmed by RT-qPCR analysis and Western blot. The results showed that CST1 overexpression increased GPX4 protein levels (Fig. 3G) without altering GPX4 mRNA level (Fig. S3A). Moreover, in the GEO database GSE54129 and GSE66229 data sets, the Pearson correlation line analysis showed no significant correlation between CST1 and GPX4 gene expression (the p values were all >0.05) (Fig. S3B, C).

Next, we examined whether CST1 enhances the stability of GPX4 protein. Cycloheximide (CHX), a protein synthesis inhibitor, has been used to determine the effect of CST1 on GPX4 stability. The HEK293T cells were transiently transfected with Flag-tagged CST1 expression plasmid and Myc-tagged GPX4 expression plasmid and treated with 200 µg/ml CHX for different times to block protein synthesis. The degradation rates of the existing GPX4 protein were measured by Western blot. The results showed that the overexpression of CST1 weakened GPX4 degradation compared with the control group (Fig. 3H, I). To further evaluate the relationship between GPX4 protein stability mediated by CST1 and the proteasome system, we used MG132, a 26S proteasome inhibitor. Notably, treatment with MG132 allowed an accumulation of GPX4 protein, as shown in MKN45-shNC/MKN45-sh1-CST1/MKN45-sh2-CST1 cells. Also, GPX4 protein levels induced by proteasome inhibition could not be further decreased by CST1 knockdown (Fig. 3J, K).

Taken together, our data indicate that CST1 upregulates GPX4 expression by increasing the stability of the GPX4 protein through the proteasome pathway.

CST1 relieves GPX4 ubiquitination through deubiquitinase OTUB1

To study whether CST1 affects GPX4 ubiquitination, we conducted ubiquitination assays. We performed ubiquitination assays in HEK293T that induced exogenous Myc-GPX4 and HA-Ub in the presence or the absence of Flag-CST1 and/or MG132. The cell extracts were immunoprecipitated using an anti-Myc antibody, followed by immunoblot analysis with an anti-HA antibody. As expected, GPX4-ubiquitination was high without CST1 (Fig. 3L, lane 2); however, the existence of CST1 obviously decreased the ubiquitination of GPX4 (Fig. 3L, lane 3–4); and GPX4-ubiquitination was lower in the presence of CST1 without MG132 (Fig. 3L, lane 3). To explore the effect of CST1 on the ubiquitination of endogenous GPX4 in gastric cancer cells, we used the CST1 knockdown stable cell line MKN45-shNC/MKN45-sh1-CST1/MKN45-sh2-CST1, immunoprecipitated GPX4, respectively, and then detected the Ub level by WB. The results showed that in the negative control group, the level of ubiquitination of GPX4 was low, whereas the level of ubiquitination of GPX4 was increased in CST1-knockdown MKN45-sh1-CST1/MKN45-sh2-CST1 cells (Fig. S3D). These findings suggest that CST1 inhibits the ubiquitination of GPX4 to regulate GC progression.

Next, we investigated the deubiquitinase or E3 interacting with GPX4. First, we performed Gene Set Enrichment Analysis (GSEA) on the pre-GC tissue transcriptome sequencing data and GEO dataset (GSE66229), according to the CST1 gene expression, the two groups of data were divided into CST1 high and low expression groups respectively. The results showed that in our transcriptome sequencing data, the genes in the CST1 high expression group were more involved in the deubiquitination modification pathway (Fig. 4A left); Similarly, in the GSE66229 dataset, genes in the CST1 high expression group are also mainly involved in the deubiquitination pathway (Fig. 4A right); the above bioinformatics analysis results confirm that CST1 is involved in protein deubiquitination modification process.

Fig. 4: CST1 relieves GPX4 ubiquitination through deubiquitinase OTUB1.

A GSEA analysis of our previous transcriptome sequencing data and GEO data set (GSE66229) enriched the deubiquitination pathway. B Predicting the deubiquitinase DUBs that interact with GPX4 through the online database (BioGRID, IntAct) and Venn analysis with all known DUBs. Intersection proteins include OTUB1 and OTUD5. C Co-IP assay on HEK293T cells transfected with Flag-tagged OTUB1 and OTUD5 plasmids. D WB detection of Flag, HA, Myc tagged proteins after IP in cells co-transfected with Myc-tagged CST1 and Flag-tagged OTUB1. E Endogenous Co-IP of GPX4 in MKN45-shNC/MKN45-sh1-CST1/MKN45-sh2-CST1. WB detection of IP proteins; OTUB1 binding to GPX4 was significantly reduced. F HGC-27-Vector/HGC-27-CST1 cells were transiently transfected with OTUB1 siRNA. The ubiquitination assay showed that the level of ubiquitin bound by GPX4 increased with the decrease of OTUB1, and the ubiquitination level of GPX4 was more obvious in HGC27 cells overexpressing CST1. G HGC-27-Vector/HGC-27-CST1 cells were transiently transfected with Flag-OTUB1-Con/WT/D88A and HA-GPX4 plasmids, WB detection showed that the GPX4 protein in the OTUB1-WT group was more stable, while the GPX4 protein in the OTUB1-D88A group was reduced. H Predicted binding complex models of CST1, OTUB1 and GPX4 by using the Cluspro online protein docking tool. I Schematic diagram of the construction of full-length and truncated CST1 proteins. J HEK293T cells were transfected with CST1-FL/N/C-Myc plasmids and OTUB1-Flag, GPX4-HA plasmid, respectively. Myc protein was immunoprecipitated, WB showed that CST1-FL and CST1-N truncated protein can bind to OTUB1, while CST1-FL and CST1-C truncated proteins can bind to GPX4. The SDS-PAGE used for separating CST fragments was 15%. (GSEA Gene set enrichment analysis, Con control/empty plasmid, WT Wild-type plasmid, D88A the D88 site of OTUB1 is point mutated to A plasmid).

By predicting the deubiquitinase DUBs that interact with GPX4 through the online database (BioGRID, IntAct) and performing Venn analysis with all known DUBs, we found that OTUB1 and OTUD5 proteins may potentially interact with GPX4 (Fig. 4B). To verify which deubiquitinase binds to GPX4, we transfected Flag-tagged OTUB1 and OTUD5 plasmids into HEK293T cells, respectively, and performed Co-IP experiments with HA-GPX4. The results indicated that only OTUB1 could bind to GPX4 (Fig. 4C). Next, we co-transfected Myc-tagged CST1 and Flag-tagged OTUB1, the results showed that in the absence of OTUB1, CST1 could bind to GPX4 (Fig. 4D, lane 2); in the absence of CST1, OTUB1 could also bind to GPX4 (Fig. 4D, lane 3); and when OTUB1-Flag was transfected at the same time, OTUB1 bound to GPX4 was significantly increased upon CST1-Myc (Fig. 4D, lane 4).These results implied that CST1 mediates the deubiquitination of GPX4 by OTUB1. To further verify this, we first performed endogenous Co-IP in a cell line in which CST1 was stably knocked down by MKN45. We found that with the decline of CST1, OTUB1 decreased the binding to GPX4 (Fig. 4E). HGC-27 cells were transiently transfected with OTUB1 siRNA, and the results showed that the level of ubiquitin bound by GPX4 increased with the decrease of OTUB1 and the ubiquitination level of GPX4 was more obvious in HGC-27 cells even overexpressing CST1 (Fig. 4F).

OTUB1 is an important deubiquitinase, which mainly depends on the inhibition of E2-conjugating enzymes and can stabilize the target proteins. When the D88 site of OTUB1 is point mutated to A(D88A), the deubiquitinase activity of OTUB1 can be significantly attenuated [27, 29]. To address whether CST1-stabilizing GPX4 protein dependent on OTUB1 deubiquitinase activity. We constructed the D88A mutant of OTUB1, and then transfected to HGC-27-Vector/CST1 stable cells, the results showed that the GPX4 protein in the OTUB1-WT group was more stable, while the GPX4 protein in the OTUB1-D88A group was significantly reduced, and this change was more pronounced after CST1 overexpression (Fig. 4G). The results confirmed that CST1 mediates the effect of OTUB1 to stabilize GPX4 protein, and depends on the inhibition of E2-conjugating enzymes by OTUB1 to exert its deubiquitin function.

Subsequently, we used the method of protein-protein docking to predict the 3D complex model and potential interaction domains of CST1, OTUB1 and GPX4. The 3D spatial structures of these proteins were obtained from SWISS-MODE and PDB databases (Fig. S4A), and the most likely complex model of CST1, OTUB1 and GPX4 binding was predicted using the Cluspro online protein docking tool (Fig. 4H). Then PDBePISA was used online, the interaction surface in the protein complex was analyzed, and it was found that the free energy of binding between CST1, OTUB1 and GPX4 was low (Fig. S4B), indicating that a stable complex can be formed between them. LIGPLOT was used to map the “eyelash figure” of protein-protein interactions (Fig. S4C), showing the potential binding domains between CST1, OTUB1 and GPX4.

Further, we constructed different truncated proteins of CST1. The full length of CST1(CST1-FL) protein contains 141 amino acids, of which amino acids 1-73 truncation name as CST1-N, and amino acids 74-141 truncation name as CST1-C (Fig. 4I), which were added Myc tags respectively. Subsequently, HEK293T cells were transfected with CST1-FL/N/C-Myc plasmids and OTUB1-Flag, GPX4-HA plasmid, respectively. Then Myc-protein was immunoprecipitated, we found that CST1-FL and CST1-N truncated protein can bind to OTUB1, while CST1-FL and CST1-C truncated proteins can bind to GPX4 (Fig. 4J). The above results indicate that OTUB1 and GPX4 can combine with different domains of CST1 to form a stable complex structure.

CST1 reduces intracellular ROS and inhibits ferroptosis through GPX4

Knowing that CST1 can regulate the stability of GPX4 protein, and GPX4 is a key molecule in regulating ferroptosis, we investigated whether CST1 affects the occurrence of ferroptosis through GPX4. Treatment of HGC-27-Vector/ HGC-27-CST1 cells with the ferroptosis inducer erastin (10 μM) inhibited the decrease of HGC-27-CST1 cell viability relative to the DMSO-treated group (Fig. 5A). Moreover, treating MKN45-shNC/MKN45-sh1-CST1/MKN45-sh2-CST1 cells with erastin (10 μM) decreased the viability of cells more significantly than the DMSO-treated group. At the same time, the ferroptosis inhibitor liproxstatin-1 (1 μM) was used for the recovery experiment, and the results showed that the viability of MKN45-sh cells was increased (Fig. 5B). The results suggested that CST1 might be associated with erastin-induced ferroptosis.

Fig. 5: CST1 reduces intracellular ROS and inhibits ferroptosis through GPX4.

A Treatment of HGC-27-Vector/HGC-27-CST1 with the ferroptosis inducer erastin (10 μM) inhibited the decrease of HGC-27-CST1 cell viability relative to the DMSO-treated group (****<0.0001). B Treating MKN45-shNC/MKN45-sh1-CST1/MKN45-sh2-CST1 cells with erastin (10 μM) decreased the viability more significantly than the DMSO-treated group (****<0.0001). After ferroptosis inhibitor liproxstatin-1 treatment, the viability of MKN45-sh cells increased. C, D The level of ROS in HGC-27-Vector/HGC-27-CST1cells detected by fluorescence microscopy. The results showed that in the erastin-treated group, CST1 reduced the level of intracellular ROS (****<0.0001). E, F In MKN45-shNC/MKN45-sh1-CST1/MKN45-sh2-CST1 cells, intracellular ROS was significantly increased after erastin treatment (****<0.0001), while ferroptosis inhibitor liproxstatin-1 reversed this process. G MDA content in HGC-27-CST1 cells decreased, and this difference was more significant after erastin treatment (****<0.0001). H MDA content in MKN45-sh cells significantly increased, and the difference was more significant after erastin treatment. After liproxstain-1 treatment, the MDA content decreased again. I The GSH content in HGC-27-CST1 cells increased. J GSH content in MKN45-sh cells decreased, and the difference was more significant after erastin treatment and increased after liproxstatin-1 treatment (****<0.0001).

The level of ROS in HGC-27-Vector/HGC-27-CST1 cells was further detected by fluorescence microscopy. The results showed that in the erastin-treated group, CST1 reduced the level of intracellular ROS (Fig. 5C, D). In MKN45-shNC/MKN45-sh1-CST1/MKN45-sh2-CST1 cells, intracellular ROS was significantly increased after erastin treatment while ferroptosis inhibitor liproxstatin-1 rescued the above effect (Fig. 5E, F).

Next, we detected the content of glutathione (GSH), an important substrate of GPX4, and malondialdehyde (MDA), a lipid peroxide product. The results showed that the MDA content in HGC-27 cells decreased with the up-regulation of CST1 expression, and this difference was more significant after erastin treatment (Fig. 5G). However, after CST1 was down-regulated, the MDA content in MKN45 cells significantly increased, and the difference was more significant after erastin treatment. After liproxstain-1 treatment, the MDA content decreased again (Fig. 5H). The GSH content in HGC-27 cells increased with the up-regulation of CST1 expression, while the GSH content in MKN45 cells significantly decreased after CST1 down-regulation, and the difference was more significant after erastin treatment and increased after liproxstatin-1 treatment (Fig. 5I, J).

Intracellular iron overload is also one of the important signs of ferroptosis. Thus, we detected the intracellular iron level by Prussian blue staining. The results showed that the up-regulation or down-regulation of CST1 did not significantly affect the intracellular iron level (Fig. S5A, B).

The above experimental results confirmed that CST1 could reduce intracellular ROS, and thus inhibit the occurrence of ferroptosis.

CST1 mediate GPX4 protein stability to promote migration and invasion in epithelial-mesenchymal transition manner in GC cells

To assess whether GPX4 is an effective target of CST1, knockdown of GPX4 in HGC-27-Vector and HGC-27-CST1 cells were used and analyzed by Western blot, which allowed for the determination of the transient transfection efficiency of GPX4-siRNA1 and GPX4-siRNA2. On the contrary, exogenous overexpression of GPX4 was found in MKN45-shNC, MKN45-sh1-CST1, and MKN45-sh2-CST1 cells, and Western blot revealed the transfection efficiency (Fig. 6A). Transwell assays were performed using HGC-27 and MKN45 cells to investigate the effect of GPX4 on cell migration and invasion abilities. The results showed that compared to HGC-27-CST1, GPX4-siRNA1 and GPX4-siRNA2 cohort significantly decreased the migration ability of HGC-27-CST1 cells. However, overexpression of GPX4 in MKN45-sh1-CST1 and sh2-CST1 cells increased the migration and invasion ability (Fig. 6B, D). The migration/invasion rate in relevant HGC-27 and MKN45 groups further confirmed this data (Fig. 6C, E).

Fig. 6: CST1 mediates GPX4 protein stability to promote migration and invasion in epithelial-mesenchymal transition manner in GC cells.

A Western blot analysis of GPX4 in HCG27 and MKN45 stable cell lines transfected with GPX4-siRNA or exogenous overexpressing GPX4. B, C Transwell migration and invasion assays were performed in HGC-27-Vector/HGC-27-CST1 treated with GPX4-siRNA and control as indicated. D, E Transwell migration and invasion assays were performed in MKN45-shNC/MKN45-sh1-CST1/MKN45-sh2-CST1 treated with transient transfection of GPX4 exogenous overexpression or vector control as indicated. Data are presented as the mean ± SD. Statistical significance was assessed by an unpaired t-test. ****p < 0.0001. F Gene set enrichment analysis of CST1 related to invasiveness and degradation of the extracellular matrix. G, H Correlation analysis of CST1 and FN1/Snail in GSE54129 and GSE66229 as illustrated in the dot plot (Person’s correlation test). I Western blot analysis of E-cadherin, Vimetin, FN1 expression in the indicated cells.

Epithelial-mesenchymal transitions (EMT) have an important role in conferring the ability of tumor cells to migrate, invade, and metastasize [30]. CST1 can promote EMT in thyroid cancer and liver cancer [31, 32]; however, its role in promoting EMT in gastric cancer has not yet been investigated. To explore the signaling pathways by which CST1-GPX4 exerted its migration and invasion effects, we performed RT-qPCR assays to measure diverse pathways. CST1 overexpression promoted HGC-27 cell mesenchymal marker (N-Cadherin) and Snail mRNA expression and reduced the epithelial marker (E-Cadherin) mRNA expression. After down-regulation of CST1 expression, the mRNA expression of mesenchymal markers (N-Cadherin) and Snail of MKN45 decreased, while the mRNA expression of epithelial markers (E-Cadherin) increased (Fig. S6A). We also found that CST1 was positively associated with the EMT marker.

To verify this mechanism, we performed Gene set enrichment analysis showed that the gene sets related to invasiveness and degradation of the extracellular matrix were enriched in samples with high CST1 expression (Figs. 6F and S6B). To further define the role of CST1 and EMT in GC cells, we evaluated their expression in GEO GSE54129, and GSE66229. Statistical analysis revealed a significant positive correlation when the expression level of CST1 in these two GEO datasets was plotted against that of FN1 (p < 0.001), Snail (p < 0.01), MMP9 (p < 0.001); however, there was no significant correlation between CST1 and E-cadherin, Vimentin (Figs. 6G, H and S6C). This observation was further confirmed by alterations in the protein expression patterns of epithelial and mesenchymal markers, which was consistent with the mRNA expression in the GEO dataset. Briefly, CST1 overexpression promoted EMT, as evidenced by suppression of E-cadherin (epithelial marker) and the upregulation of Vimentin/FN1 (mesenchymal marker) in HGC-27 cells. Conversely, MKN45-sh1/2-CST1 cells reverted to an epithelial phenotype compared with control cells (Fig. 6I). Together, these observations demonstrated that CST1 is a positive regulator of EMT in GC cells.

CST1 promotes gastric cancer cell migration and invasion by regulating GPX4-K11 site ubiquitination

It is a key experiment to explore the regulation of GPX4 stabilization by CST1 by mutating the ubiquitination site of GPX4 and promoting gastric cancer migration and invasion. To search for GPX4 ubiquitination sites, we firstly adopted bioinformatics methods, respectively, in GPS-Uber (http://gpsuber.biocuckoo.cn/index.php) and BDM-PUB (http://bdmpub.biocuckoo.org/prediction.php) online website predicted the possible ubiquitination site of GPX4, and found that the K11 site could be predicted in both sites (Fig. S7A), so we assumed the K11 amino acid of GPX4 as a potential ubiquitination site of transformation. Next, we constructed a GPX4-K11 site mutant, transfected HGC-27-Vector/HGC-27-CST1 stable gastric cancer cells with GPX4-WT plasmid, empty control plasmid and Ub-HA plasmid, respectively, and immunoprecipitated Myc tag. The ubiquitination level of the precipitated protein was detected by WB. The results showed that in HGC-27-Vector cells with low CST1 protein, the ubiquitination level of GPX4 protein was significantly decreased after the GPX4-K11 site mutation (Fig. S7B lane 3). In HGC-27-CST1 cells with up-regulated CST1, the ubiquitination level of GPX4 protein decreased more significantly after GPX4-K11 site mutation (Fig. S7B lane 6). The above results confirm that the GPX4-K11 site is its ubiquitination site, but it is worth noting that after mutating the K11 site, GPX4 still undergoes a small amount of ubiquitination modification, indicating that there are other possible ubiquitination sites, further research is needed.

Further, we investigated whether GPX4-K11 ubiquitination site mutation affects the migration and invasion ability of gastric cancer cells, and found that in HGC-27-Vector cells with low CST1 protein, after GPX4-K11 site mutation, the migration and invasion abilities of HGC-27 cells were enhanced; in HGC-27-CST1 cells with up-regulated CST1, after GPX4-K11 site mutation, more HGC-27 cells migrated and invaded, and the difference was statistically significant (Fig. S7C, D).

CST1 promotes distant metastasis in vivo

To validate the biological function of CST1 in GC metastasis in vivo, we injected CST1-overexpressing cells (HGC-27-CST1), their corresponding controls (HGC-27-Vector), and HGC-27-GPX4#sh cells into the abdominal cavity of nude mice. After 60 days, metastasis in the peritoneum was analyzed. As expected, ectopic expression of CST1 significantly increased the number of peritoneum xenograft tumors and the ascites volume in the abdominal cavity. However, sh-GPX4 effectively rescued the role of CST1 in the peritoneum metastasis model (Fig. 7A). Conversely, CST1-silenced MKN45 peritoneum xenograft tumors and ascites volume in the abdominal cavity were lower than in control groups (Fig. 7B).

Fig. 7: CST1 promotes distant metastasis in vivo.

A Left: Images of peritoneal metastasis in nude BALB/c mice after injection of HGC-27-Vector/HGC-27-CST1/HGC-27-GPX4#sh cells into their abdominal cavity; images taken 60 days after injection. Right: statistical significance of the peritoneal nodules number and the ascites volume assessed by paired t-test, *p < 0.05, **p < 0.01. B Left: CST1-silenced MKN45 peritoneum xenograft tumors 30 days after injection. Right: statistical significance of the peritoneal nodules number and the ascites volume assessed by paired t-test, **p < 0.01, ****p < 0.0001. C Left: corresponding images of the lungs after injection of HGC-27-Vector/HGC-27-CST1/HGC-27-GPX4#sh cells by tail vein; images taken 3 months after injection. Right: statistical significance of the metastasis nodules number assessed by paired t-test, **p < 0.01. D Left: lung metastasis model of MKN45-shNC/MKN45-sh1-CST1/MKN45-sh2-CST1 8 weeks after tail vain injection. Right: statistical significance of the metastasis nodules number assessed by paired t-test, ***p < 0.001. E, F Left: HGC-27-Vector/HGC-27-CST1 and MKN45-shNC/MKN45-sh1-CST1/MKN45-sh2-CST1 cells were injected into tail vein; metastatic tumors in the livers were assessed 2 and 3 months after injection. Right: statistical significance of the metastasis nodules number assessed by paired t-test, **p < 0.01, ***p < 0.001. G Western blot analysis indicated metastasis tumors for GPX4 and CST1. H The content of MDA in the peritoneal metastatic tumor tissue of nude mice was detected, and the difference was statistically significant (***p < 0.001, ****p < 0.0001). HE-stained sections were magnified ×0.66 and ×5, respectively.

To further explore whether CST1 could promote GC cell metastasis in vivo, HGC-27-Vector/HGC-27-CST1/HGC-27-GPX4#sh were injected into the lateral tail veins of nude mice. Metastasis formation was measured by continuous pathological sections and HE staining 3 months after cell injection. The number of lung metastasis tumors and liver metastasis was significantly increased in the CST1 overexpression group compared with the control group, and the GPX4#sh group could rescue the role of the CST1 (Fig. 7C, E). Contrary, mice injected with MKN45-sh1-CST1/MKN45-sh2-CST1 showed less lung metastasis and liver metastasis compared to control (Fig. 7D, F).

Finally, the role of CST1 was further examined in ex vivo. Immunoblot showed that in tumor grinding cells of HGC-27-Vector/HGC-27-CST1/HGC-27-GPX4#sh, CST1 overexpression increased the expression of the GPX4. The knockdown of CST1 had the opposite effect (Fig. 7G). In addition, the expression of MDA in tumor tissue decreased after overexpression of CST1, while it increased again after the downregulation of GPX4; opposite effects were seen after down-regulation of CST1 (Fig. 7H).

To sum up, this data suggests that CST1 can promote gastric cancer metastasis in vivo through GPX4 inhibition of ferroptosis.

High levels of CST1 and GPX4 expression correlate with tumor aggressiveness and poor clinical outcome in GC patients

To further define the role of CST1 and GPX4 in GC patients, we evaluated their expression in 95 GC patients’ tissue by IHC. Statistical analysis revealed a significant positive correlation when the expression level of CST1 in 95 tissues was plotted against that of GPX4 (p < 0.0001) (Fig. 8A). Also, the positive correlation between CST1 and GPX4 expression levels remained unchanged regardless of the degree of gastric cancer tissue differentiation (Fig. 8B). These findings suggested that CST1 and GPX4 might be coregulated in GC.

Fig. 8: High CST1 and GPX4 expression levels correlate with tumor aggressiveness and poor clinical outcome in GC patients.

A IHC for 95 GC patients’ tissue. Pearson correlation analysis of the expression of CST1 and GPX4. B Representative graph of CST1 and GPX4 expression in IHC according to the degree of differentiation of gastric cancer tissues. C Survival analysis; the survival time of 95 gastric cancer patients was analyzed; patients with high CST1 expression had a shorter survival (p = 0.0036). D ROC curve analysis of the sensitivity and specificity of CST1 in the diagnosis of gastric cancer patients (AUC = 0.9311, p < 0.0001). E Serum ELISA assay in 50 normal people, 45 stage I–II GC patients, and 30 stage III–IV GC patients revealed that CST1 was highly expressed in GC patients’ serum compared with normal people and highly expressed consistent with patient’s clinicopathological stage (p < 0.0001). F, G Western blot detecting intracellular and extracellular CST1 in HEK293T cells with the FLAG-CST1 expression vector. H Peritoneal lavage fluid of 50 GC patients without metastasis and 30 GC patients with malignant ascites by ELISA. Expression of CST1 in malignant ascites was significantly higher than without metastasis patients (p < 0.0001). I Multivariate Cox regression model analysis of the relationship between CST1 expression and prognosis of gastric cancer. J Mechanistic diagram showing the role of CST1 in gastric cancer cells.

We performed WB detection of GPX4 protein in 5 paired gastric cancer tissues and adjacent normal gastric mucosa tissue samples, the results showed that GPX4 protein levels were higher in gastric cancer tumor tissues than in adjacent normal gastric mucosa tissues (Fig. S8A), which is in good agreement with the expression of CST1 in Fig. 1L. At the same time, we detected the expression of GPX4 protein in normal gastric mucosa epithelial cells GES-1 and gastric cancer cell lines AGS etc., the results showed that the level of GPX4 protein in gastric cancer cells increased (Fig. S8B), which was a good correlation with the expression of CST1 protein in Fig. 2B.

In order to explore whether the expressions of CST1 and GPX4 are also related in other tumor tissues, we selected colon cancer tissues and cells for experiments. We performed WB detection of CST1 and GPX4 proteins in 5 matched colon cancer tissues and adjacent normal intestinal mucosal tissue samples. The results showed that compared with adjacent normal intestinal mucosal tissues, the levels of CST1 and GPX4 proteins in colon cancer tumor tissues were higher (Fig. S8C), and the expression of the two molecules were correlated. At the same time, we detected the expression of CST1 and GPX4 protein