Supervillin promotes epithelial-mesenchymal transition and metastasis of hepatocellular carcinoma in hypoxia via activation of the RhoA/ROCK-ERK/p38 pathway

The primary antibodies described in this article include anti-supervillin (H340 [35]), anti-ERK1/2 (#4695; Cell Signaling Technology; MA, USA), anti-p-ERK1/2 (#4370; Cell Signaling Technology), anti-p38 (#8690; Cell Signaling Technology), anti-p-p38 (#4511; Cell Signaling Technology), anti-c-Jun N-terminal kinase (JNK)1/2 (#9252; Cell Signaling Technology), anti-p-JNK1/2 (#4668; Cell Signaling Technology), anti-E-cadherin (#sc7870; Santa Cruz Biotechnology, Inc.; CA, USA), anti-Vimentin (#sc73258; Santa Cruz Biotechnology, Inc.; CA, USA), anti-Snail1 (#sc393172; Santa Cruz Biotechnology, Inc), anti-β-actin (#3700; Cell Signaling Technology), and anti-β-tubulin (#TA506805; Origene; China).

HCC tissue microarrays were obtained from US Biomax, Inc. (Rockville, MD, USA). The immunohistochemical analyses of HCC tissue microarrays were carried out as previously described [42]. The KF-PRO Digital Slide Scanning System (Kongfong Biotech International Co., LTD; Ningbo, China) was used to visualize the signal.

HCC cell lines MHCC-97H, Huh-7, and HepG2 were a gift from Pro. Z.Y. Tang (Liver Cancer Institute, Fudan University, Shanghai) and were used in a previous study [42]. All the cell lines were kept at low passages for experimental use, and revived every 3 to 4 months. All cell lines used in this study were regularly authenticated by morphologic observation and tested for the absence of mycoplasma contamination. They were maintained in Dulbecco’s Modified Eagle’s Medium or DMEM (Hyclone; Logan, UT, USA) supplemented with 10% fetal bovine serum or FBS (Gibco; Grand Island, NY, USA) and 1% penicillin/streptomycin (Hyclone). Cells were exposed to hypoxia (1.0% O₂) in a hypoxic chamber (Thermo Fisher Scientific, Inc., Waltham, MA, USA) for the indicated period.

Cells were transfected with supervillin Stealth siRNA #1, #2, #3, and #4 and negative control siRNA (Invitrogen; Carlsbad, CA, USA) at 40 nM using Lipofectamine® RNAiMAX (Invitrogen), or with GFP-tagged SV1, SV4, and SV5 plasmids with the BTX ECM® 830 Electroporation System (Harvard Apparatus; Holliston, MA, USA), according to the manufacturer’s instructions. The RNAi targeting sequences and their corresponding target exons in the supervillin gene are shown in Additional file 1: Table S1.

MHCC-97H supervillin knockdown stable clones were constructed using pLVshRNA-EGFP (2A) Puro (Invivogen, Shanghai, China) lentiviral shRNA vector targeting supervillin (5′-AGGTGATGAAGCCAGATGA-3′).

To determine the functional crosstalk between RhoA/ROCK and ERK/p38 signaling, HCC cells were pretreated with the specific mitogen-activated protein kinase kinase (MEK) inhibitor (PD0325901, 10 μM; MedChemExpress, Monmouth Junction, NJ, USA), p38 MAPK inhibitor (SB239063, 10 μM; MedChemExpress), or ROCK inhibitor (Y27632 2HCl, 10 μM; MedChemExpress) for 1 h. To investigate the functional crosstalk between supervillin expression and HIF1α, we used the HIF inhibitor (2-Methoxyestradiol; 2-MeOE2, 10 μM; MedChemExpress) for 8 h.

A 100-μl cell suspension (MHCC-97/SVIL-shRNA, containing 5 × 10⁶ cells) in PBS was subcutaneously injected into the flanks of male BALB/c nude mice (n = 5) at the age of 6 weeks. Mice were sacrificed 2 weeks later and the tumors were cut into approximately 1 mm³ for successive orthotopic implantation into 6-week-old male BALB/c nude mice. To restrict blood flow into the liver and mimic hypoxia, a HAL was carried out by tying a fine thread around the main branch of the hepatic artery. For tumors that were formed from GFP-labeled cell lines, the growth of the tumor was monitored by bioluminescence detection using IVIS 100 Imaging System (Xenogen; Princeton, NJ, USA). Six weeks after the orthotopic implantation, mice were sacrificed, and their lungs and livers were excised. The fluorescence intensity from tumor foci was taken as the tumor metastasis potential for comparison among transplanted nude mice. Animal experiments were performed according to the guidelines of the Animal Use and Care Committees at Hefei Institutes of Physical Science, CAS.

Transwell chambers with 8-μm pores (Corning; NY, USA) were used for cell migration assay. For the tumor cell invasion assay, the Transwell membrane was pre-coated with 30 μl of Matrigel matrix (1:3 mixed with PBS; BD Biosciences; Heidelberg, Germany). Cell suspension in serum-free medium was added to the upper chamber, and then incubated in normoxia or hypoxia for 18 h. After incubation, migrated or invaded cells were fixed with methanol for 30 min, stained with 0.1% crystal violet, and counted under a light microscope at 100× magnification (Leica, DMI 4000 B; Germany). Then, the crystal violet was eluted by methanol and the OD at 546 nm was examined using a spectrophotometer.

Cell lysate preparation and immunoprecipitation was conducted as previously described [43], using 30 μl Anti-HA Magnetic Beads (Pierce Biotechnology, Rockford, IL, USA). Twenty microliters of beads were resuspended in 20 μl of 2× protein sample buffer before analysis.

Cells were washed, fixed on glass slides, and stained with primary antibodies: anti-E-cadherin (1:200), anti-Vimentin (1:250), and anti-Snail1 (1:100). The secondary antibodies conjugated with fluorescein (FITC; Invitrogen; Carlsbad, CA, USA) were then incubated for 1 h at room temperature and stained with DAPI. Stained cells were then visualized under a fluorescence microscope (Leica, DMI 4000 B; Germany).

GTP loading assays were carried out as described [41]. Briefly, cells were extracted in lysis buffer at 4 °C and centrifuged for 10 min at 14000 g. Supernatants were incubated with GST-RBD or GST-PBD pre-bound to glutathione-Sepharose beads for 1 h at 4 °C. GST-peptide beads were centrifuged, washed thrice in washing buffer A (25 mM Tris, 40 mM NaCl, 30 mM MgCl₂, 10 μg/ml leupeptin, 10 μg/ml aprotinin, 1 mM 4-(2-aminoethyl)-benzene sulfonyl fluoride, pH 7.5), boiled in sample buffer, and resolved by SDS-PAGE.

Statistical calculations were performed using Prism 5 software (GraphPad). Data were expressed as mean ± SD, unless otherwise indicated. Continuous variables were evaluated using an unpaired Student’s t-test for comparisons between two groups. The Pearson’s chi-square test was used to analyze the association of supervillin expression with clinicopathological parameters. Multivariate analysis was performed using forward stepwise logistic regression analysis. Two-sided tests were performed with a p < 0.05 indicating statistical significance: ns, non significant, * - p < 0.05, ** - p < 0.01, *** - p < 0.001.

To investigate the prognostic value of supervillin expression levels in HCCs, we performed microarray-based immunohistochemical analyses of 173 HCC tissues from two independent cohorts with comparable clinicopathological features and complete follow-up data. In 124 HCC specimens, supervillin expression levels were dramatically upregulated, compared to those in adjacent non-tumor specimens or normal liver tissues (Fig. 1a). Notably, a higher expression of supervillin was significantly correlated with the presence of PVTT (χ2 = 15.44, p < 0.01), occurrence of serosal infiltration (χ2 = 12.64, p < 0.01), and distant metastasis (χ2 = 9.79, p < 0.01), which indicated the potential important roles of supervillin played in HCC metastasis (Table 1).

PVTT is a significant risk factor for metastasis in HCC patients and frequently accompanies substantial hypoxia within the tumor microenvironment [44]. Supervillin was positively expressed in hypoxic PVTT (Fig. 1b). Thus, we first examined the expression profiles of supervillin during hypoxia. We quantified the levels of supervillin splicing variants, the hypoxia-response factor, HIF1α, and the downstream target of HIF1α, CA9, in lysates from Huh-7 (low metastatic HCC cell line) and MHCC-97H (high metastatic HCC cell line) cells as a function of exposure time to a reduced (1%) oxygen environment (Fig. 1c). As expected, HIF1α and CA9 proteins gradually accumulated due to a hypoxia-induced increase in protein stability. Strikingly, the levels of the supervillin isoforms SV4 and SV5 reached a peak after 16 h (SV4, p < 0.001; SV5, p < 0.001) and then gradually declined during hypoxia, as did HIF1α. By contrast, the level of SV1 did not change significantly during hypoxia. Our previous study showed that a reduction in SV4 and SV5 levels decreased the rates of cell migration and invasion of human A549 lung carcinoma cells during normoxia [36]. This suggests that increased SV4 and SV5 expression during hypoxia may play a role in the metastatic potential of HCC.

To determine the potential role of supervillin, especially the different supervillin splicing isoforms on HCC cell migration and invasion under hypoxic conditions, we used Stealth RNAi™ dsRNAs that target sequences within SVIL coding exon 4 (RNAi #1; specific for SV4), coding exon 5 (RNAi #2; targets both SV4 and SV5), coding exon 10 (RNAi #3; targets all three isoforms), and the 3’-UTR (RNAi #4; targets all three isoforms). As described previously [36], each Stealth siRNA that targeted supervillin splice isoforms (SV1, SV4, and SV5) in HCC cells reduced the level of each isoform by ≥75% (Additional file 1: Figure S1A-C). Indeed, hypoxia caused 17 and 41% increases in the rates of cell migration of Huh-7 cells (Fig. 2a and b, and Additional file 1: Figure S2A, 282 ± 4.42 μm in hypoxia vs. 208 ± 2.85 μm in normoxia; p < 0.01) and MHCC-97H cells (Fig. 2c and d, and Additional file 1: Figure S2B, 431 ± 6.15 μm in hypoxia vs. 382 ± 4.15 μm in normoxia; p < 0.01), respectively. Transfection of supervillin-specific RNAi #2 or #3 resulted in an 81 and 83% reduction in cell migration in Huh-7 and MHCC-97H cells after 18 h in normoxia, respectively (Fig. 2a and c, and Additional file 1: Figure S2A and B). In RNAi #2 or #3-transfected cells, cell migration increased by only 23 and 22% after 18 h of hypoxia, respectively (Fig. 2b and d, and Additional file 1: Figure S2A and B). These results suggest that a reduction in the levels of supervillin isoforms, especially SV4 and SV5, blocks hypoxia-induced increases in cell migration rates, as well as reducing cell migration in normoxia.

To strengthen our findings from the wound-healing model, we further studied cell migration and invasion using the Boyden chamber method. The results showed that cell migration increased by 36 and 86% (Fig. 2e and g, and Additional file 1: Figure S2C and D) and cell invasion by 58 and 26% (Fig. 2f and h, and Additional file 1: Figure S2E and F) in Huh-7 and MHCC-97H cells under hypoxic stress, respectively. Supervillin knockdown with RNAi #2 or #3 resulted in a 65 and 72% reduction in cell migration (Fig. 2e and g, and Additional file 1: Figure S2C and D) and a 73 and 59% reduction in cell invasion (Fig. 2f and h, and Additional file 1: Figure S2E and F) for Huh-7 and MHCC-97H cells after 18 h in normoxia, respectively, and almost completely abolished hypoxia-induced increases in HCC cell migration and invasion. Consistent with these findings, overexpression of SV1, SV4, or SV5 in supervillin knockdown cells rescued cell migration rates to different degrees (Fig. 2i, and Additional file 1: Figure S2G, 21% increase after SV1 overexpression; p < 0.01, 43% increase after SV4 overexpression; p < 0.01, and 39% increase after SV5 overexpression; p < 0.01). These results support the involvement of supervillin during the migration and matrix invasion of HCC cells under hypoxic stress.

Given that the EMT program is critical for matrix invasion and the dissemination of most and possibly all carcinoma types, we examined whether supervillin expression induces EMT in hypoxic HCC cell lines. After exposure to 1% oxygen for 16 h, Huh-7 and HepG2 cells became elongated (Fig. 3a). Expression of the epithelial biomarker E-cadherin was significantly down-regulated, with simultaneous rapid increase in the expression of the mesenchymal biomarkers Vimentin and Snail1 in both Huh-7 and MHCC-97H cells (Fig. 3b and Additional file 1: Figure S3A-C). During hypoxia, the amount of E-cadherin and its localization at the plasma membrane were reduced, whereas the signals for Vimentin and Snail1 became stronger than those in normoxia (Fig. 3c and Additional file 1: Figure S3D). These results are consistent with hypoxia-stimulated EMT in HCC cells [44, 45]. RNAi-mediated down-regulation of supervillin in Huh-7 and MHCC-97H cells largely prevented the hypoxia-induced decrease in E-cadherin levels and the increase in Vimentin and Snail1 signaling (Fig. 3b and c, and Additional file 1: Figure S3A-F).

During epithelial to mesenchymal transition (EMT), cancer cells gain migratory and invasive properties due to the dramatic reorganization of the actin cytoskeleton [10‐12]. As shown in Additional file 1: Figure S3G-H, there was a significant increase in F-actin during hypoxia, as compared with that in normoxia. However, the reduction of supervillin decreased the number and thickness of stress fibers during hypoxia (Additional file 1: Figure S3I). These results suggest that supervillin plays an important role in the hypoxia-induced EMT in HCC, and is responsible for the rearrangement of actin cytoskeleton under hypoxia.

During hypoxia, the small G proteins of the Rho family are involved in the reorganization of the actin cytoskeleton and cell migration [18, 46]. To explore the mechanism of supervillin action during HCC migration and invasion, we first examined the crosstalk with Rho GTPases known to regulate cell-substrate adhesion, cell polarization, and the rates of cell spreading and translocation. We screened for functional crosstalk between supervillin and Rac1, Cdc42, and RhoA (Fig. 4a, and Additional file 1: Figure S4A and B). As has been described previously in HeLa cervical carcinoma cells, supervillin knockdown with RNAi #2 or #3 decreased Rac1 loading in normoxia [41]. During hypoxia, knockdown of supervillin with RNAi #2 or RNAi #3 was accompanied by a significant reduction in the amount of GTP-loaded (active) RhoA, as assessed by pulldown assays with GST-tagged Rho-binding domain (RBD). In contrast, no significant changes were observed in GTP loading of Cdc42. When supervillin expression was allowed to recover to initial levels, GTP-RhoA levels almost returned to control values (Additional file 1: Figure S5A). Next, we identified if RhoA-GTP is a novel direct or indirect interaction partner of supervillin (Fig. 4b and Additional file 1: Figure S5B). Notably, the N-terminus of supervillin co-precipitated with RhoA, especially the region containing the three documented F-actin binding domains [37]. Supervillin preferentially bound to active RhoA (WT and the constitutively active V14 mutant; Fig. 4c and d), suggesting that this association might play an important role during cytoskeletal rearrangements in hypoxia. Consistent with this hypothesis, expression of the RhoA(V14) mutant reversed the inhibitory effects of supervillin knockdown on cell migration (Fig. 4e, and Additional file 1: Figure S4C and D, p < 0.01 in Huh-7 and MHCC-97H cells) and invasion (Fig. 4f, and Additional file 1: Figure S4C and D, p < 0.01 in Huh-7 and MHCC-97H cells) during hypoxia. When HCC cells were pretreated with the ROCK inhibitor (Y27632 2HCl), RhoA(V14)-induced rescues of cell migration and invasion were blocked (Fig. 4e and f, and Additional file 1: Figure S4c and d, p < 0.01 in Huh-7 and MHCC-97H cells), implying that supervillin-induced increase in HCC cell migration and invasion during hypoxia require RhoA/ROCK signal transduction.

In migrating cells, supervillin and associated proteins coordinate a rapid, basolateral membrane recycling pathway that contributes to ERK signaling and actin-based cell motility [35]; the supervillin isoform SV3 has been proposed to scaffold ERK with its upstream kinase MEK in smooth muscle cells [47]. To further elucidate supervillin-associated mechanisms in hypoxia, we monitored changes in the MAPK/ERK1/2 and associated signal molecules in Huh-7 and MHCC-97H HCC cells (Fig. 5). Inhibition of supervillin expression induced a decrease in phosphorylated p38 and ERK1/2 to different degrees under hypoxic stress. Expression of supervillin isoforms rescued these decreases after supervillin knockdown in Huh-7 cells (Additional file 1: Figure S5C). In addition, overexpression of individual supervillin isoforms during hypoxia increased the levels of both total and phosphorylated ERK1/2 (Fig. 6a). The increases in both total and phosphorylated ERK1/2 were reversed with the MEK inhibitor PD0325901 (Fig. 6a). In addition, phosphorylation of p38, Vimentin, and Snail1 were also inhibited by PD0325901 (Fig. 6a). HCC cell migration and invasion ability was blocked after the addition of 10 μM PD0325901 for 18 h during hypoxia (Fig. 6b and c). These data show that the MAPK/ERK/p38 signaling cascade participates in the supervillin-driven HCC cell migration and invasion within the hypoxic microenvironment.

Both RhoA/ROCK and ERK/p38 signaling are indispensable for cell migration and invasion; therefore, we investigated potential crosstalk between the RhoA/ROCK and ERK/p38 signaling pathways. We found that phosphorylation of ERK and p38 depended on the RhoA/ROCK pathway because the levels of ERK/p38 activation were affected by the presence of active RhoA and various inhibitors (Fig. 6d and Additional file 1: Figure S6). The expression of active RhoA(V14) increased ERK/p38 phosphorylation relative to control values, even in the presence of RNAi #2. The ROCK inhibitor (Y27632 2HCl) abolished the RhoA-induced increase in ERK and p38 phosphorylation in MHCC-97H and Huh-7 cells under hypoxic conditions. However, the MEK inhibitor (PD0325901) and p38 inhibitor (SB239063) had no effect on RhoA/ROCK activation (Fig. 4a, and Additional file 1: Figure S4A and B), indicating that ERK/p38 phosphorylation is downstream of RhoA/ROCK activation. Collectively, our results suggest that supervillin promotes cell mobility through the RhoA/ROCK and ERK/p38 pathways within hypoxic microenvironments.

To explore a potential role for supervillin in HCC metastasis in vivo, we carried out orthotopic liver implantation experiments. Tumor seeds derived from MHCC-97H with non-target control or supervillin-knockdown cells were implanted into the livers of mice, after either HAL or a mock surgery (Mock). HAL increased the stability of the HIF1α protein and the amount of supervillin, especially in the SV4 and SV5 isoforms (Fig. 7a and b), similar to our results using a hypoxic incubator (Fig. 1d). HAL treatment enhanced tumor growth and lung metastasis in vivo (Fig. 7c-f). The largest number of metastatic tumor foci was observed in the lungs of animals subjected to HAL (Fig. 7e and f). The metastatic capability was suppressed in mice implanted with supervillin-knockdown cells (Fig. 7e and f). The suppressed metastatic capability was consistent with a decrease in phosphorylated p38 and ERK1/2 to different degrees under hypoxic stress (Additional file 1: Figure S7). Indeed, knockdown of supervillin was accompanied by a significant reduction in the amount of GTP-loaded (active) RhoA (Additional file 1: Figure S7). These results suggest that supervillin contributes significantly to cell motility and metastasis of HCC cells under hypoxic stress in vivo, as well as in vitro.