STOML2 potentiates metastasis of hepatocellular carcinoma by promoting PINK1-mediated mitophagy and regulates sensitivity to lenvatinib

Two sets of HCC samples were used in our study. The first set containing 48 HCC samples was used to analyze STOML2 expression at mRNA and protein level. The second set containing 227 HCC samples was used to analyze STOML2 protein expression and evaluate the correlation with clinicopathological features. The details are described in Additional file 1: supplementary materials and methods.

IHC staining was performed using EnVisiontm system as previously described [21]. Antibodies applied in this experiment are listed in Additional file 2: Table S1. The detail procedures were presented in Additional file 1: supplementary materials and methods.

All transfections were performed using Lipofectamine™ 3000 (Invitrogen, L3000015) according to the manufacturer’s instructions. The respective primers sequence for STOML2 knock-down are shown in Additional file 2: Table S2. The details are described in Additional file 1: supplementary materials and methods.

All the HCC cells used were seeded on cover slides in 24-well plates, incubated overnight and then fixed in 4% paraformaldehyde for 15 min, permeabilized with 1% Triton X-100 for 5 min, blocked in 1% bovine serum albumin (BSA) for 60 min, and incubated with primary antibodies for 60 min at RT, followed by secondary antibodies for 60 min at RT. Nuclei were stained with 4′,6-diamidino-2-phenylindole dihydrochloride (DAPI, Cell Signaling Technology, #4083) at RT for 10 min. Photographs were captured with a laser confocal microscopy (Leica Microsystems AG). Antibodies applied in this experiment are listed in Additional file 2: Table S1.

Three days after infection with lentiviruses containing control, STOML2, or STOML2-specific shRNAs, proteosomal degradation was blocked by treating the cells with 20 μM MG132 for 6 h. Cells were lysed with 150 μL of denaturing lysis buffer (50 mM Tris–Cl at pH 6.8, 1.5% SDS), and then collected by scraping followed by boiling for 15 min. Ninety microliters of the denatured protein samples was added to 1 mL of EBC/BSA buffer (50 mM Tris–Cl at pH 6.8, 180 mM NaCl, 0.5% NP40, 0.5% BSA) and incubated with anti-PINK1 antibody or anti-FLAG antibody overnight and with protein A/G beads for 1 h at 4 °C. Ubiquitin antibody was used to detect poly-ubiquitinated PINK1 in the IP samples.

All experimental procedures involving animals were approved by The Animal Care and Use Committee of Fudan University, China. Five-week-old nude mice (BALB/c) were randomly divided into indicated groups (n = 5 per group) before inoculation or injection. HCCLM3-shNC and HCCLM3-shSTOML2 cells were subcutaneously injected into the mice (1.0 × 10⁷ cells/mouse) to form the subcutaneous model. For the xenograft model, subcutaneous tumors were removed and dissected into 1 mm³ sections, which were incubated into the liver parenchyma of nude mice. For drug treated groups, mice were injected intraperitoneally with hydroxychloroquine (HCQ, 50 mg/kg), lenvatinib (LV, 5 mg/kg or 10 mg/kg), combination with lenvatinib (5 mg/kg) and HCQ (50 mg/kg) or saline as control. The mice were sacrificed at a time-defined endpoint and tumor volume, weight, and number of metastatic nodules were assessed by double-blinded evaluation. The details are presented in Additional file 1: supplementary materials and methods.

Statistical analysis was performed with SPSS 22.0. All the data were expressed as mean ± standard deviation (SD). Kaplan–Meier analysis was used for survival analysis, and the log-rank test was chosen to compare the difference. Cox proportional hazards regression analyses were adopted for multivariate analysis. Pearson test or Fisher’s exact test were employed to compare qualitative variables, while Student t test or One-way ANOVA were used for quantitative variables. P < 0.05 was considered statistically significant.

For more details and other methods, see Additional file 1: supplementary materials and methods.

To further demonstrate what we found in previous gene expression profiles analysis that STOML2 was highly expressed in HCC, especially with metastasis [17], quantitative real-time PCR (qRT-PCR), immunoblotting assay, and immunohistochemistry in 48 HCC tissues and corresponding peri-tumor liver tissues were carried out. The data showed that the expression of STOML2 in HCC was much higher than the adjacent non-tumor liver tissues on both mRNA and protein levels (Fig. 1a–e), which is identical to the statistical results of evaluated analyzing the mRNA expression level of STOML2 by the GEPIA database (Additional file 3: Figure S1A) [22]. Notably, in comparison with non-metastatic HCC tissues (n = 37), the expression of STOML2 was significantly higher (P = 0.0003) in metastatic HCC tissues (n = 11), which showed intrahepatic spreading or tumor invasion into blood vessel or bile ducts (Fig. 1d). These data showed a close association of STOML2 upregulation with HCC.

Then, we evaluated the above association of STOML2 and HCC as well as the prognostic value of STOML2 in HCC. Survival analysis based on the RNA sequencing expression data of the HCC patients (n = 369) from the TCGA database was performed first and the results indicated that patients with higher STOML2 expression had remarkably worse overall survival (OS, P = 0.011) and disease-free survival (DFS, P = 0.029) (Additional file 3: Figure S1B-C) [23]. Second, the immunohistochemistry study of STOML2 in consecutive tissue microarray with 227 HCC tissues was performed. These patients, among which 186 cases (81.9%) showed liver cirrhosis with hepatitis B background, were divided into high or low STOML2 expression groups according to the immunostaining scores (Fig. 1f). Patients with microvascular invasion and advanced TNM stage appeared to possess high STOML2 levels in primary HCC tissues. More importantly, when compared with low STOML2 expression group, the significantly worse OS (median OS: 38.9 months versus more than 55 months, log-rank = 10.3, P = 0.001) (Fig. 1g) and shortened time to tumor recurrence (TTR) (median TTR: 22 months versus 53 months, log-rank = 8.7, P = 0.003) (Fig. 1h) were found in STOML2 high expression group. Hepatitis B e antigen positive, high AFP level, large tumor size, microvascular invasion, multiple tumors, and advanced TNM stage were also found associated with worse OS and shortened TTR in univariate analysis (Table 1, Additional file 2: Table S3). To assess the correlation between high STOML2 level and other risk factors, a Cox proportional hazards analysis was performed. The results showed that high STOML2 level is an independent risk factor for worse OS (hazard ratio = 1.596, P = 0.026) and shortened TTR (hazard ratio = 1. 638, P = 0.011). Taken together, these data imply that upregulated STOML2 has a significant correlation with poor prognosis of HCC and it may contribute to HCC progression.

Concerning the higher expression of STOML2 in HCC with metastasis, we evaluated the functions of STOML2 in HCC cells using both in vitro and in vivo assays. First, we examined the expression of STOML2 in HCC cell lines with different metastasis potential (Additional file 4: Figure S2A-B). The results of qRT-PCR and Western blot showed that high metastatic potential HCCLM3 and MHCC-97H cells expressed high levels of STOML2 versus low expression of STOML2 in low malignant potential SMMC-7721, Bel-7402 cells. Modulating the expression of STOML2 in HCC cells with lentivirus-mediated specific short hairpin (sh) RNAs or STOML^Flag (Fig. 2a), we found that over-expression of STOML2 in SMMC-7721 cells resulted in significant promotion of colony formation (Fig. 2b) and cell proliferation (Fig. 2c). Accordingly, the percentage of total (both early and late) apoptotic cells were obviously decreased (Fig. 2d, Additional file 4: Figure S2D).

To specifically address the important role of STOML2 in the migration and invasion of HCC cells, transwell assays were performed. The results suggested that the promoting migration and invasion role of STOML2 (P < 0.001) is much more significant than that on the proliferation (Fig. 2e). We then verified the influence of STOML2 on malignant potential in HCCLM3 with STOML2 specific shRNA#1 and shRNA#2 (shS#1, shS#2). Identical to upregulation of STOML2, STOML2-knockdown (KD) decreased the proliferation and colony formation, especially decreased migration and invasion significantly (Fig. 2c-e, Additional file 4: Figure S2C).

To confirm in vitro findings, in vivo experiments were carried out. In subcutaneous implantation nude mice models, tumor growth was monitored every 5 days. 20 days later, the nude mice injected with HCCLM3 cells transfected with negative control shRNA (shNC) were found to have much bigger tumor sizes than those injected with HCCLM3-shS#1 and HCCLM3-shS#2 cells (P < 0.001, Fig. 2f, left and middle panel). Moreover, we observed a significant reduction of Ki-67 expression by immunohistochemistry analysis and induction of apoptosis confirmed by TUNEL staining when STOML2 was knockdown (Fig. 2g–h). In tail vein injection lung metastasis models, one mouse in shS#1 group was found lung metastasis and none in shS#2 group, whereas the mean number of lung metastasis in shNC group was 2.2 per lung (Fig. 2f, right panel). In brief, these in vitro and in vivo gain- and loss-of-functional studies suggested the key role of STOML2 in promoting HCC growth and metastasis.

STOML2 has been identified as a mitochondrial inner membrane protein, belonging to a superfamily of putative scaffolding proteins, including Prohibitin 2, an inner membrane mitophagy receptor [19]. Thus, we speculated that STOML2 might be involved in autophagy in HCC, especially in mitophagy. To validate this hypothesis, we first assessed the transcriptomes of TCGA HCC patients with varying STOML2 expression and analyzed the top 500 differentially expressed genes in STOML2 high-expression versus STOML2 low-expression patients. The results showed that these genes were significantly enriched in biological processes including regulation of protein targeting to mitochondrial and autophagy (Fig. 3a). Next, monodansylcadaverine (MDC)-labeled autophagic vacuoles were detected by flow cytometry in vitro. The fluorescence intensity of SMMC-7721 with upregulation of STOML2 was significantly stronger than control cells while it dropped drastically in HCCLM3 with STOML2-KD (Fig. 3b). More bilayer membrane-bound autophagosomes were also observed by transmission electron microscopy (TEM) in those HCC cells with high STOML2 expression, which further supported the strong correlation between STOML2 and autophagy (Fig. 3c). The analysis of Western blot ulteriorly demonstrated that, in HCC cells with high STOML2 expression, autophagy marker cytosolic microtubule-associated protein 1A/1B-light chain 3 B (LC3B I, an important subtype of LC3 I) became conjugated phosphatidylethanolamine to form LC3B II. Consistently, STOML2-KD in HCC cells resulted in weakened LC3B lipidation under the treatment of carbonyl cyanide mchlorophenylhydrazone (CCCP) (Fig. 3d, left panel). More diffuse fluorescence of LC3 B were detected using immunofluorescence in HCC cells with high STOML2 expression (Fig. 3e).

An increase of mitochondrial protein TOMM20, COX IV, Tim23, and VDAC1 in HCC cells with STOML2-KD was demonstrated by Western blot or confocal laser microscope, indicating the accumulation of mitochondrial proteins. Furthermore, we purified mitochondria from HCC cells and examined the protein expression of p62 and LC3B II in mitochondria and cytoprotein, respectively. Much different to the weak change of them in cytoprotein, p62 and LC3B II increased significantly in mitochondria of SMMC-7721 with STOML2-overexpresion while decreased in HCCLM3 with STOML2-KD, respectively (Fig. 3d, middle and right panels). With confocal co-localization analysis, we found a substantial amount of TOMM20 predominantly co-localized with lysosome degradation marker LAMP1 in HCC cells with STOML2 high expression (Fig. 3f). Collectively, these results indicated the important roles of STOML2 on mitophagy, which might tightly connect with HCC proliferation and invasion.

To uncover the underlying molecular mechanisms of STOML2 regulating HCC mitophagy, immunoprecipitation/mass spectrometry (IP/MS) was conducted to identify key molecules. STOML2-Flag produced in SMMC-7721 cells was immunoprecipitated by Flag mAb and coprecipitated proteins were visualized by sliver staining after electrophoresis and recognized by MS. One coprecipitated molecule turned out to be PINK1 (Fig. 4a), a key regulator of mitophagy. We next performed co-immunoprecipitation to verify whether STOML2 interacts with PINK1. PINK1 was precipitated by Flag-STOML2-tagged beads but not control beads in SMMC-7721 upon CCCP treatment, and the interaction between endogenous STOML2 with PINK1 was also confirmed in HCCLM3 (Fig. 4b). Furthermore, we observed the intracellular distribution of STOML2 and PINK1 in the presence of CCCP stimulation by confocal laser microscopy, which indicated that STOML2 co-localized with PINK1 in SMMC-7721-STOML2 and HCCLM3 (Fig. 4c). The similar co-localization was also found in STOML2 high expression HCC tissues (Fig. 4d). Furthermore, the expression and co-localization between PINK1 and COXIV were strong in SMMC-7721-STOML2 and HCCLM3 versus sparse expression of PINK1 and weak co-localization with COXIV in SMMC-7721, HCCLM3-shS#1 and shS#2 (Additional file 5: FigureS3A-B). Then, Western blot was carried out to detect the expression of STOML2, PINK1, and Parkin. The result showed that variation tendency of PINK1 and Parkin was consistent with STOML2, increased when STOML2 was upregulated, and vice versa, especially in the mitochondrial level (Fig. 4e). Notably, we detected and compared the expression of PINK1, Parkin, and LC3B I/II in HCC tissues and peri-tumor liver tissues (Fig. 4f). Identical to what have been found, the expression of PINK1, Parkin, and LC3B II in HCC tissues was much higher than that in peri-tumor liver tissues, and the statistical results showed positive correlation between STOML2 and PINK1 (Fig. 4f, right panel).

To determine the mechanism of the phenomenon that PINK1 could be upregulated by STOML2, qPCR analysis was carried out. We found that neither overexpression nor knockdown of STOML2 had little effect on transcription level of PINK1 (Additional file 5: Figure S3C), whereas the protein levels kept the similar variation tendency (Fig. 4e). The above results indicated that STOML2 upregulates PINK1 at the protein level rather than the mRNA level. Subsequently, we investigated the effect of STOML2 on the protein stability of PINK1. MG132, the proteasome-specific inhibitor, could protect PINK1 protein from degradation in STOML2-low expression HCC cells (Additional file 5: Figure S3D). A significant decrease of polyubiquitinated PINK1 protein was observed in STOML2-overexpression SMMC-7721 while STOML2-KD in HCCLM3 had the converse phenomenon (Fig. 4g). Furthermore, we tested the effect of STOML2 on PINK1 degradation rate by cycloheximide (CHX) pulse-chase assay, under the treatment of CCCP, STOML2-overexpression could dramatically prolong the half-life of PINK1 protein after adding CHX. In contrast, STOML2-KD remarkably decreased the half-life of PINK1 protein (Fig. 4h). These results indicated that STOML2 could inhibit the degradation of PINK1.

It is well known that facing all kinds of stresses, such as antiangiogenesis, tumor has to change accordingly to develop drug insensitivity to survival. Increasing evidences suggest mitophagy is involved in the process. We postulated that lenvatinib, a first line antiangiogenesis drug for HCC, could cause mitochondrial dysfunction and induce mitophagy, which may be close correlated with the widespread drug insensitivity of antiangiogenesis. First, the mitochondrial membrane potential (MMP), a key inducing factor of mitophagy, was evaluated in SMMC-7721 and HCCLM3 with lenvatinib treated. JC-1 staining was applied to monitor MMP of cells by cytometry [24]. The measurement demonstrated that MMP of SMMC-7721 and HCCLM3 with lenvatinib treated decreased significantly compared with controls (Additional file 6: Figure S4A). Lenvatinib also promoted the transformation of LC3B I to LC3B II, with increased expression of PINK1, Parkin, and p62 in a time-dependent manner (Fig. 5a). Confocal co-localization analysis supported what have been demonstrated by Western blot. Compared with DMSO treatment, the co-localization between PINK1 and COXIV increased sharply as lenvatinib treated in HCCLM3-shNC cells (Additional file 6: Figure S4B). More LC3B, LAMP1 were detected and co-localized with mitochondria marker for the most part in lenvatinib treated cells (Fig. 5b, Additional file 6: Figure S4C). Furthermore, STOML2-KD in HCCLM3 partly counteracted the enhancement of PINK1 and LC3B with lenvatinib treatment. The co-localization between LC3B, PINK1, LAMP1, and mitochondrial marker also significantly decreased, respectively (Fig. 5b, Additional file 6: Figure S4B-C). In brief, these data indicated that mitophagy could be activated by lenvatinib, and STOML2 played important an role in mitophagy activation.

To explore the role of STOML2 on malignant potential in HCC cells following lenvatinib treatment, we compared the different responses to lenvatinib treatment in SMMC-7221, HCCLM3, and derived cells. STOML2-KD in HCCLM3 cells enhanced the cellular response to lenvatinib (Fig. 5c), inhibiting cell colony formation (Fig. 5d, Additional file 7: Figure S5A), migration (Fig. 5e, Additional file 7: Figure S5B) and inducing total apoptosis (Fig. 5f) to a greater extent, whereas overexpression of STOML2 in SMMC-7221 had the opposite effects (Fig. 5c–f). Notably, blockade of mitophagy by chloroquine (CQ) or PINK1 siRNA (Additional file 7: Figure S5C-D) reversed STOML2-related protection against lenvatinib (Fig. 5c–f, Additional file 7: Figure S5A-B). In conclusion, STOML2-induced mitophagy served as a protective function in HCC cells treated with lenvatinib, and blocking mitophagy enhanced the inhibition efficacy of lenvatinib to HCC cells.

Hypoxia is one of the main contributors to the acquisition of drug resistance and closely related to a worse prognosis [25, 26]. As one of the multi-kinase inhibitors, lenvatinib inhibits HCC growth and metastasis significantly, but accentuates hypoxia environment in tumor tissue. So we wonder whether hypoxia-inducible factor 1α (HIF-1α), a master regulator of oxygen homeostasis [27], could be regulated by lenvatinib. Thus, we investigated the expression of HIF-1α under the treatment of lenvatinib (10 μM) and Western blot analysis indicated that HIF-1α expression increased gradually with the time extending (Fig. 6a). Interestingly, we also found that STOML2 was increased simultaneously (Fig. 6a). To further determine whether increase of STOML2 in lenvatinib treatment is HIF-1α dependent, we examined STOML2 expression in HCC cells with HIF-1α downregulation and found that lenvatinib-induced STOML2 upregulation was abrogated upon silencing HIF-1α in both SMMC-7721 and HCCLM3 (Fig. 6b). An increase of STOML2 expression might result from activated STOML2 transcription and/or elevated STOML2 protein stability. Indeed, we found STOML2 mRNA levels were significantly increased upon lenvatinib treatment (Fig. 6c), suggesting that STOML2 might be transcriptionally regulated by HIF-1α. To confirm whether STOML2 is a direct transcriptional target of HIF-1α, STOML2 promoter region from − 2000 bp to the first ATG was cloned and putative hypoxia-responsive element (HRE) containing HIF-1α-binding consensus sequence 5′-A/GCGTG-3′ was located (Additional file 8: Figure S6A-B). To identify whether HIF-1α could bind at this site, chromatin immunoprecipitation (ChIP) assay was carried out and the results showed that the chromatin fragments from HCCLM3 cells immunoprecipitated by anti-HIF-1α mAb specifically enriched compared to IgG group (Fig. 6d). The sequences containing the putative 5′-promoter region of STOML2 (WT) or the region with mutated binding site (Mut) were cloned into luciferase promoter reporter vectors. The activity of wild type STOML2 promoter (WT) was significantly increased in hypoxic SMMC-7721 cells, while it was decreased in the HRE mutation group which further suggested the predicted HRE in the STOML2 promoter region is functional (Fig. 6e, Additional file 8: Figure S6B). In addition, over-expression of HIF-1α stimulated the STOML2 promoter reporter activity in SMMC-7721 cells while knockdown of HIF-1α expression downregulated the STOML2 promoter reporter activity in the HCCLM3 cells (Fig. 6f). Taken together, these data suggested that STOML2 was a direct transcriptional target of HIF-1α.

To confirm these in vitro findings, we further examine the efficacy of lenvatinib (5 mg/kg or 10 mg/kg, respectively) and hydroxychloroquine (50 mg/kg) alone, or combination (lenvatinib 5 mg/kg + hydroxychloroquine 50 mg/kg) on HCC growth and metastasis in immunodeficient mice bearing orthotopic HCCLM3 xenograft tumors compared to normal saline control. As shown in Fig. 7, statistical differences were found among control, lenvatinib low and high dose groups. When compared with the tumor weight in control group (2.52 ± 0.42 g), lenvatinib reduced tumor weight in a dose-dependent mode, 1.74 ± 0.41 g for 5 mg/kg and 1.26 ± 0.33 g for 10 mg/kg, respectively (Fig. 7a, b), whereas the tumor inhibition effects were limited in hydroxychloroquine treatment group (2.06 ± 0.587 g). The similar effects on lung metastasis inhibition and prolonged survival were also found in those groups received lenvatinib treatment. Notably, the highest suppression of primary tumor growth was found in mice treated with the combination therapy. The average tumor size was 0.68 ± 0.37 g, with tumor growth inhibition rate 73%. The lowest lung metastasis was also achieved in this group (Fig. 7c, d). Additionally, the median overall survival of mice increased from 36 days in control group to over 63 days (P < 0.001) in combination therapy group. The effect of the combination treatment group is even better than high dose of lenvatinib treatment group (P = 0.006) (Fig. 7e).

Taken together, these in vitro and in vivo gain- and loss-of-functional studies and in vivo combination therapy demonstrated that STOML2 played important roles in mitophagy, which resulting in HCC growth and invasion. Lenvatinib and chloroquine/ hydroxychloroquine combination treatment inhibited the growth and metastasis of HCC and increased median OS. Schematic representation of the major molecular mechanism was shown in Fig. 7f.