2.1 STIM1 is temporally regulated during autophagy-triggered EMT in HCC cells

Autophagy has been reported to be involved in multiple stages of cancer, particularly tumorigenesis and metastasis.^{23, 24} In order to investigate the correlation between autophagy and HCC progression, we conducted Gene Set Variation Analysis (GSVA) using an autophagy-related gene set (MAP1LC3B, ATG10, ATG12, ATG13, ATG3, ULK1) obtained from The Cancer Genome Atlas (TCGA) HCC patient database. The HCC patients were divided into a high autophagy score (AutS) group and a low AutS group based on the mean AutS. Patients with high AutS exhibited better overall survival (OS), longer progression-free survival and disease-free interval (Figure 1A). Furthermore, AutS was significantly higher in HCC tissues compared to normal para-cancerous tissues (Figure 1B). Using bioinformatics, we analyzed the correlation between cancer-related pathways and autophagy in HCC. Our findings revealed that EMT was the most closely associated pathway with autophagy (Figure 1C). Additionally, we observed a significantly positive correlation between the EMT score and AutS in HCC patients from the TCGA database (Figure 1D). Furthermore, we conducted experiments where autophagy was induced by serum-free starvation using Earle's Balanced Salt Solution (EBSS). In this experiment, EBSS-triggered autophagy facilitated EMT by increasing N-cadherin expression and decreasing E-cadherin expression in HCC cells after 24 and 36 h (Figure 1E). These results strongly suggest that autophagy plays a significant role in HCC progression, particularly in relation to EMT.

In our previous research, we discovered that STIM1 plays a role in regulating the metastasis of HCC cells as a metabolic checkpoint. Interestingly, we observed that the expression of STIM1 is upregulated during HCC proliferation in a hypoxic microenvironment, but downregulated when EMT begins.²¹ Additionally, we found that after inducing autophagy, the expression of STIM1 increased at 24 h but decreased at 36 h (Figure 1E). Furthermore, STIM1 expression was decreased during autophagy and the EMT process induced by carbonyl cyanide 4-(trifluoromethoxy) phenylhydrazone (FCCP), but was restored in the presence of the autophagy inhibitor Bafilomycin A1 (BafA1) (Figure 1F). These findings suggest that STIM1 may be involved in autophagic degradation during autophagy-triggered EMT in HCC cells. In addition, our findings indicate that STIM1 is significantly upregulated in tumor tissue compared to normal tissue in HCC patients (Figure 1G), which is consistent with our previous study.²⁰ Furthermore, correlation analysis demonstrated a positive association between STIM1 and autophagy-related genes as well as EMT-related genes in the database of HCC patients (Figure 1H). Consistently, patients with both lower STIM1 and AutS have better OS (Figure S1), indicating that the combined inhibition of STIM1 and autophagy may benefit survival of HCC patients. These results provide evidence that STIM1 is temporally regulated during autophagy-induced EMT in HCC cells.

2.2 STIM1 mediates autophagy and EMT of HCC cells

To investigate the role of STIM1 in autophagy and EMT, HepG2 and Hep3B cells, which have high levels of STIM1 protein, were selected to generate STIM1 knockout (KO) cells using the CRISPR/Cas9 system (Figures 2A and S2). Following induction of autophagy by FCCP, a significant decrease in the LC3BII/I ratio and an increase in sequestosome 1 (SQSTM1/p62) expression in the STIM1-KO cells were observed, indicating a reduction in autophagy levels (Figure 2A). Additionally, electron microscopy analysis revealed a marked decrease in the number of autophagosomes in HepG2 cells upon STIM1 depletion (Figure 2B). Next, mRFP-GFP-LC3 lentiviral was transfected into HCC cells to assess autophagic flux. The STIM1-KO cells exhibited a reduced number of LC3 puncta, indicating a decrease in autophagic flux after FCCP treatment (Figure 2C). Consistent with previous studies,²¹ the STIM1-KO cells displayed lower EMT features compared to the control groups (Figure 2D), as well as a decreased migration rate and invasion capacity (Figures S3A–C). Furthermore, a higher coexpression of STIM1 and LC3B in the metastasis edge of HCC was observed than that of adjacent normal tissue (Figure 2E). These findings highlight the critical role of STIM1 in both autophagy and EMT processes in HCC cells.

2.3 STIM1 regulates autophagy in both SOCE-dependent and independent manners

We then speculated whether SOCE is solely responsible for the regulation of autophagy by STIM1. In contrast to STIM1-KO, the SOCE inhibitor SKF96365 only showed a minimal impact on reducing the autophagy level induced by FCCP in HCC cells (Figure 3A). This implies that there is an alternative mechanism, independent of SOCE but involving STIM1, that is necessary for autophagy. To further investigate the underlying mechanism, we generated a lentivirus containing a functional mutant of STIM1 called STIM1-dCTD. This mutant lacks SOCE activation activity because it has a deleted C-terminal domain (aa441-685). We then introduced STIM1-dCTD into the STIM1-KO cells (Figure 3B). Interestingly, the overexpression of STIM1-dCTD significantly enhanced the formation of autophagosomes, and increased the ratio of LC3BII/I in STIM1-KO HepG2 cells (Figures 3C and D). In addition, the overexpression of STIM1-dCTD in cells significantly augmented the EMT features. This was characterized by increased expressions of zinc-finger-enhancer protein 1 (ZEB1) and N-cadherin, decreased expression of E-cadherin, and enhanced migration and invasion capabilities (Figures 3E–G). These findings suggest that STIM1 plays a critical role in regulating autophagy through both SOCE-dependent and independent mechanisms in HCC cells.

2.4 STIM1 directly interacts with LC3B through SAM domain

In light of STIM1 being an ER structural protein that is located in the same area as autophagy initiation,⁸ we next investigated whether STIM1 facilitates autophagy by interacting with autophagy-related proteins. LC3B is a crucial autophagic ubiquitin-like protein involved in the formation of autophagosomes. We observed a strong interaction between STIM1 and LC3B after FCCP treatment (Figures 4A and B), suggesting that STIM1 may influence autophagosome formation mediated by LC3B. Furthermore, this interaction between STIM1 and LC3B gradually increased within 3 h after FCCP treatment, which was consistent with the LC3BII/I ratio. However, between 3 and 6 h after FCCP administration, the combination of STIM1 and LC3B gradually decreased (Figure 4C). These findings indicate that STIM1 binds with LC3B during the process of autophagosome formation and is subsequently degraded in autophagosomes or autolysosomes in HCC cells.

After establishing docking models to predict the binding region of STIM1 and LC3B, we identified five possible binding sites located on the SAM domain of STIM1 (Figure 4D). This prediction was supported by the observation that STIM1-dCTD could also form a complex with LC3B (Figure 4A). To further investigate their interaction domains, we performed coimmunoprecipitation experiments using lysates of STIM1-KO HCC cells overexpressing full-length STIM1 or five different deletion mutations of STIM1 along with full-length LC3B. Our results showed that the absence of the SAM domain (d127-200) resulted in a decreased ratio of LC3BII/I, which significantly reduced the interaction between STIM1 and LC3B (Figure 4E). These data suggest that STIM1 directly interacts with LC3B through SAM domain during autophagy process.

2.5 Deletion of SAM domain in STIM1 diminishes its autophagy-promoting ability in HCC cells

The SAM domain is known to be essential for the oligomerization of STIM1, which is required for the activation of Orai1.¹⁶ To further investigate the role of the SAM domain of STIM1 in autophagy and EMT, we transfected lentivirus with the STIM1 variant of deletion SAM domain (STIM1-dSAM) into STIM1-KO HCC cells. Similar to STIM1-dCTD, STIM1-dSAM also failed to trigger SOCE in STIM1-KO cells (Figure 5A). However, the introduction of STIM1-dSAM resulted in a lower level of autophagosomes compared to STIM1-dCTD (Figure 5B). Additionally, cells expressing STIM1-dCTD exhibited a higher number of LC3B puncta compared to cells expressing STIM1-dSAM, indicating the crucial role of the SAM domain of STIM1 in inducing autophagy (Figure 5B). Moreover, immunofluorescence staining revealed that STIM1-dCTD colocalized more with LC3B in the cytosol, while STIM1-dSAM did not show any colocalization with LC3B in HCC cells. This further supports the significance of the SAM domain in mediating protein-binding between STIM1 and LC3B (Figures 5C and D). Taken together, these findings suggest that STIM1 forms a complex with LC3B through its SAM domain, promoting the formation of autophagosomes and enhancing autophagy in HCC cells.

2.6 STIM1/LC3B complex contributes to HCC metastasis through activating autophagy

Given the significant role of autophagy in HCC metastasis, we proceeded to investigate the involvement of the STIM1/LC3 complex in this process. We found that STIM1-dCTD, rather than STIM1-dSAM, notably enhanced the migration ability of HepG2 STIM1-KO cells (Figure 6A). To further investigate the role of autophagy in STIM1 modulation of EMT, BafA1 was used to block autophagy. In the presence of BafA1, the migration and invasion capability enhanced by STIM1-dCTD was significantly reduced in STIM1-KO HepG2 cells. Additionally, FCCP failed to induce autophagy and accelerate the migration and invasion rate in STIM1-dSAM expressing STIM1-KO HepG2 cells (Figures 6B and C). These findings suggest that the SAM domain of STIM1 plays a predominant role in promoting autophagy and EMT in HCC cells.

Next, we examined the impact of STIM1 on the metastasis of HCC cells in vivo. Using an orthotopic liver injection model, we observed that the absence of STIM1 significantly reduced the metastatic ability of HepG2 cells. Interestingly, after ectopically expressing STIM1-dCTD (but not STIM1-dSAM), a dramatic enhancement in intrahepatic metastasis of STIM1-KO cells was observed (Figure 6D). Furthermore, mouse models injected with STIM1-KO cells or ectopically expressing STIM1-dSAM cells exhibited a significantly improved survival rate (Figure 6E). We also utilized a splenic vein injection model and obtained consistent results (Figures 6F and G). Compared with NC groups, STIM1-KO led to significantly decreased LC3B level in metastatic lesions. Introduction of STIM1-dCTD, other than STIM1-dSAM, possesses the capability of restoring LC3B expression (Figure S4). Collectively, our findings suggest that STIM1 promotes autophagy by interacting with LC3B through its SAM domain, forming a complex that leads to EMT and metastasis in HCC cells.

4.1 Human HCC tissue specimens

Human HCC tissue specimens were obtained from the Chongqing University Cancer Hospital. The use of clinical specimens in this study was approved by Ethical Review Board of the Chongqing University Cancer Hospital ethics committee, and informed written consent was obtained from all participants.

4.2 Cell lines

HepG2, HuH-7, MHCC97-H, HCCLM3, BEL-7402, PLC/PRF/5, SMMC-7721, and Hep3B were purchased from the American Type Culture Collection (Rockville, MD, USA). All cell lines used in this study were authenticated and tested for mycoplasma. Passages with a value of less than 10 were utilized.

4.3 Animals

All male BALB/c nude mice were purchased from Vital River (Beijing, China). For orthotopic tumor models, 5×10⁴ cells were resuspended in 50 μL of PBS and injected into the left lateral lobe of the liver. After a period of 4 weeks, the mice were sacrificed in order to observe the metastasis. In the case of splenic vein injection tumor models, 1×10⁷ cells were resuspended in 200 μL of PBS and injected into the splenic vein. After a period of 6 weeks, the mice were sacrificed and the metastasis lesions were counted by dissecting the liver. All experimental protocols were approved by the Army Medical University Animal Ethics Committee and strictly adhered to the guidelines of the National Institutes of Health regarding animal welfare.

4.4 Chemicals and reagents

SKF96365 (HY-100001), puromycin (HY-B1743), BafA1 (HY-100558), cyclopiazonic acid (CPA; HY-N6771), and FCCP (HY-100410) were obtained from MedChemExpres (Monmouth Junction, NJ, USA). EBSS (14155063), Fura2-AM (F1221), and Lipofectamine 3000 (L3000015) were obtained from Thermo Fisher Scientific (Waltham, MA, USA). Dimethysulfoxide (34869) and CPA (239805) were purchased from Sigma–Aldrich (St. Louis, MO, USA).

4.5 Plasmids and lentiviral

To generate truncated fragments of STIM1 with different deleted domains, including d127-200aa (deletion of SAM domain), d201-234aa (deletion of TM domain), d235-340aa (deletion of CC1 domain), d341-389aa (deletion of CC2 domain), and 390−440aa (deletion of CC3 domain), a His-tag was added to the C-terminus of each fragment. These fragments were then subcloned into the pCMV3-C-His vector (Genscript, Wuhan, China). Lentivirus expressing sgRNAs for deleting STIM1 and lentivirus over-expressing the STIM1-dCTD mutant have been previously described.²¹ Lentivirus carrying the STIM1-dSAM expression element (lacking 127−200 aa) and mRFP-GFP-LC3 lentivirus were obtained from Genechem (Shanghai, China). The procedures for lentivirus infection, plasmid transfection, and selection of stable expressing STIM1-mutant cells were previously described.²¹

4.6 Western blotting and coimmunoprecipitation

LC3B antibody (3868S), N-cadherin antibody (13116S), E-cadherin antibody (3195S), STIM1 antibody (5668S), SQSTM1/p62 antibody (23214S), ZEB antibody (83243SF), MMP-2 antibody (40994S), Vimentin antibody (5741S), and His-tag antibody (12698S) were purchased from Cell Signaling Technology (Danvers, MA, USA). β-Actin antibody (81115-1-RR) was purchased from Proteintech (Wuhan, China). The western blot was performed according to previously described methods.²⁰ For coimmunoprecipitation, cells were seeded and grown on 10 cm dishes overnight. After treatment, cells were harvested in lysis buffer with PMSF. The supernatants collected from centrifugation were immunoprecipitated with His-tag antibody at 4°C overnight. Then, the mixture was blended with protein A/G PLUS-agarose and incubated at 4°C for 2 h. Next, the mixture was washed by centrifuging five times with cold lysis buffer containing PMSF and the sediment was heated for denaturation with SDS buffer. Finally, immunoblotting analysis was performed.

4.7 Observation of autophagosomes by transmission electron microscope

Cell resuspension solution was centrifuged at 200 g. The sediment was immediately fixed with 2.5% glutaraldehyde in 0.1 mol/L sodium cacodylate and stored at 4°C until embedding. Afterwards, samples were postfixed with 1% osmium tetroxide and dehydrated using an increasing concentration gradient of ethanol and propylene oxide. Subsequently, the samples were embedded and ultrathin (50–60 nm) sections were cut using an ultramicrotome. Finally, the images were examined using a JEM-1200 electron microscope (JEOL, Japan) at 80 kV after staining with 3% uranyl acetate and lead citrate.

4.8 Calcium imaging

Calcium imaging was performed following the established protocol.²¹ Briefly, cells were seeded on circular glass slides coated with poly-d-lysine (Nest, Wuxi, China) and incubated overnight. Fura2-AM was used to label intracellular Ca²⁺, and the fluorescence signal was measured using an inverted fluorescence microscope (Nikon, Tokyo, Japan). The Ca²⁺ signal was collected at wavelengths of 340 and 380 nm. To investigate the impact of different solutions, we replaced the initial solution with either HBSS containing CPA or HBSS containing 2 mM CaCl₂. The fluorescence signals at 340 and 380 nm were recorded and the ratios were calculated.

4.9 Immunofluorescence

Cells were seeded on circular glass coated with poly-d-lysine and cultured overnight. Afterward, cells were fixed using 4% Polyoxymethylene and washed three times. Subsequently, cells were blocked with BSA and incubated with the primary antibody overnight at 4°C. Following this, the glass was washed and incubated with the second antibody for 2 h. Finally, the slides were covered with antifade mounting medium containing DAPI and observed under a ZEISS LSM 880 confocal microscope (ZEISS, Germany).

4.10 Docking prediction

The 3D structure of the STIM1/LC3B interaction was constructed using the SWISS-MODEL server (http://swissmodel.expasy.org/). The structure of STIM1 (PDB ID: 2K60) and LC3B (PDB ID: 2ZJD 1−125) was utilized. Molecular docking calculations were performed using the ZDOCK 3.0/3.0.2 Scoring Function.³⁰

4.11 Cell invasion and migration assay

The wound-healing assay, transwell assay, and immunofluorescence were performed as previously described.²¹

4.12 Quantification and statistical analysis

GSVA analysis was conducted by using the GSCA website (http://bioinfo.life.hust.edu.cn/GSCA) with data obtained from the TCGA database. Correlation analysis was performed by using the GEPIA website (http://gepia.cancer-pku.cn/) with data also obtained from the TCGA database. Statistical analysis was carried out using GraphPad Prism7 (version 7; La Jolla, CA). All data were presented as mean ± SEM and analyzed using Student's t-test or one-way ANOVA. Survival analysis was analyzed using Mantel–Cox test. Statistical significance was defined as p values < 0.05.