PMPCB Silencing Sensitizes HCC Tumor Cells to Sorafenib Therapy

Abstract

Hepatocellular carcinoma (HCC) tumors invariably develop resistance to cytotoxic and targeted agents, resulting in failed treatment and tumor recurrence. Previous in vivo short hairpin RNA (shRNA) screening evidence revealed mitochondrial-processing peptidase (PMPC) as a leading gene contributing to tumor cell resistance against sorafenib, a multikinase inhibitor used to treat advanced HCC. Here, we investigated the contributory role of the β subunit of PMPC (PMPCB) in sorafenib resistance. Silencing PMPCB increased HCC tumor cell susceptibility to sorafenib therapy, decreased liver tumor burden, and improved survival of tumor-bearing mice receiving sorafenib. Moreover, sorafenib + PMPCB shRNA combination therapy led to attenuated liver tumor burden and improved survival outcome for tumor-bearing mice, and it reduced colony formation in murine and human HCC cell lines in vitro. Additionally, PMPCB silencing enhanced PINK1-Parkin signaling and downregulated the anti-apoptotic protein MCL-1 in sorafenib-treated HCC cells, which is indicative of a healthier pro-apoptotic phenotype. Higher pre-treatment MCL-1 expression was associated with inferior survival outcomes in sorafenib-treated HCC patients. Elevated MCL-1 expression was present in sorafenib-resistant murine HCC cells, while MCL-1 knockdown sensitized these cells to sorafenib. In conclusion, our findings advocate combination regimens employing sorafenib with PMPCB knockdown or MCL-1 knockdown to circumvent sorafenib resistance in HCC patients.

Zheng et al. demonstrate that silencing the β subunit of PMPC (PMPCB) increases HCC tumor cell susceptibility to sorafenib therapy, decreases liver tumor burden, and improves survival of tumor-bearing mice receiving sorafenib. Moreover, PMPCB silencing enhances PINK1-Parkin signaling and downregulates the anti-apoptotic protein MCL-1 in sorafenib-treated HCC cells, indicative of a healthier, pro-apoptotic phenotype.

Introduction

The genomic landscape of tumors is complex and dynamic, and mutations accumulate and evolve over time.¹ It remains challenging to differentiate oncogenic tumor drivers from passenger mutations that arise from genomic instabilities, and, consequently, from identifying the most central and vulnerable mutations for targeted therapy.¹ Solid tumors invariably acquire resistance to targeted therapies, resulting in poorer patient response and prognosis.2, 3

Hepatocellular carcinoma (HCC) is an archetype solid tumor that develops chemoresistance and constitutes a significant health burden, with over 700,000 deaths globally per year.⁴ HCCs are intrinsically resistant to cytotoxic agents,⁵ and even U.S. Food and Drug Administration (FDA)-approved sorafenib, a multikinase inhibitor, confers only a modest gain in survival of ∼3 months.⁶ Although sorafenib is known to inhibit wild-type (WT) RAF proto-oncogene serine/threonine-protein kinase (Raf1), vascular endothelial growth factor receptor (VEGFR) subtypes 2 and 3, and both WT and mutant proto-oncogene Braf,⁷ it is not precisely known by what mechanism(s) HCC tumors develop chemoresistance to sorafenib at the molecular level.

To better elucidate these mechanism(s), Rudalska et al.⁸ established a system of negative-selection screens of short hairpin RNAs (shRNAs) performed in situ within murine HCC cells by leveraging transposon-mediated integration of shRNAs into the HCC genome. The shRNA screen was performed in the presence of sorafenib to identify transcripts that conferred greater susceptibility to sorafenib.⁸ This screening process uncovered mitochondrial-processing peptidase (Mpp, Pmpc) as a possible candidate.⁸ That being said, the mitochondrial-processing peptidase (PMPC)’s contributory role to sorafenib chemoresistance in HCC (if any) remains unexplored.

Therefore, we investigated PMPC’s inhibition of pro-apoptotic signaling mediated by PTEN-induced putative kinase 1 (PINK1) and Parkin ligase as a central pathway in sorafenib chemoresistance in HCC cultures and in a sorafenib-resistant murine model of HCC. Combination treatment of sorafenib with an shRNA against the β subunit of PMPC (PMPCB) attenuated the growth of HCCs in vitro and improved the survival outcome of mice with sorafenib-resistant HCC tumors. Our findings implicate the silencing of PMPCB expression as a potential approach to overcoming sorafenib chemoresistance and improving the therapeutic benefit of sorafenib therapy.

Results

Generation of Sorafenib-Resistant Murine HCCs Using a Nras^G12V Transposon-Based Model

To model sorafenib-resistant HCC in mice, we employed a well-established murine model where oncogenic Nras^G12V transposon is delivered into p19^Arf-deficient mouse livers via tail vein injection8, 9, 10 (Figure S1). The resulting Nras^G12V; p19^Arf−/− mouse model reliably triggers the growth of aggressive, multifocal HCCs that are resistant to sorafenib therapy.⁸ Rudalska et al.’s⁸ previously published shRNA screen in p19^Arf-deficient livers under sorafenib therapy uncovered the alpha subunit of the mitochondrial-processing peptidase Pmpc (Pmpca) as a possible gene conferring resistance to sorafenib. Therefore, we first investigated whether Pmpca and Pmpcb expressions are differentially regulated in various HCC cells with different driving oncogenes in response to sorafenib.

We subjected murine Nras^G12V; p19^Arf−/− HCC cells, murine Nras^G12V/Akt-1; p19^Arf−/− HCC cells (which also possess a constitutively active form of Akt-1), and the human HCC cell lines Huh7 and Hep3B to sorafenib to determine sorafenib’s effects on Pmpca and Pmpcb expressions in vitro. With the exception of Hep3B cells, sorafenib significantly upregulated Pmpca mRNA and protein expressions in a time-dependent manner, but it had no such effect on Pmpcb expression in vitro (Figures 1A–1D). We then subjected Nras^G12V; p19^Arf−/− mice to sorafenib therapy to determine sorafenib’s effects on Pmpca and Pmpcb expression in vivo. Similarly, we found that sorafenib therapy significantly upregulated Pmpca mRNA and protein expressions in a time-dependent manner, but it had no such effect on Pmpcb expression in vivo (Figures 1E and 1F).

Figure 1.

PMPCA Upregulation in HCC Cells Occurs between 1 and 4 Weeks following Sorafenib Initiation

(A–D) qRT-PCR and western blot (WB) of lysates from cultured murine Nras^G12V; p19^Arf−/− (NG12V) cells, murine Nras^G12V/Akt-1; p19^Arf−/− (NG12V/Akt-1) cells, human Hep3B cells, and human Huh7 cells analyzing (A) PMPCA mRNA levels, (B) PMPCA protein levels, (C) PMPCB mRNA levels, and (D) PMPCB protein levels over a duration of 3 days to 4 weeks following sorafenib treatment. Sorafenib was added to cells 1 day after plating and maintained at the following concentrations during the culture period: NG12V and NG12V/Akt-1 cells (8 μM), Hep3B (2 μM), and Huh7 (4 μM). (E) qRT-PCR and (F) WB of murine Nras^G12V; p19^Arf−/− tumors for Pmpca and Pmpcb levels following 15 and 30 days of treatment with sorafenib or vehicle. Mice (n = 9) were orally administered sorafenib (100 mg/kg) every other day. Liver tumor specimens were collected for analysis after anesthesia. All in vitro experiments: n = 3 biological replicates × 3 technical replicates. Error bars express the means ± SEMs.

Pmpcb Knockdown by shRNA Sensitizes Murine HCCs to Sorafenib

Pmpca and Pmpcb are subunits of Pmpc (Mpp), a protein that belongs to the family of mitoproteases that modulate several biological activities necessary for proper mitochondrial functioning, including apoptosis.¹¹ To further examine the role of Pmpc in HCC sorafenib resistance, we engineered the well-established pCaNIG transposon construct8, 12 carrying Nras^G12V/GFP and a non-coding shRNA (pCaNIG-shNC) or shPmpca (pCaNIG-shPmpca) or shPmpcb (pCaNIG-shPmpcb) (Figure S2A) to knock down Pmpca subunit and Pmpcb subunit expressions in p19^Arf−/− knockout (KO) murine cells, respectively. The three Pmpca shRNAs and the three Pmpcb shRNAs caused efficient knockdown (KD) of their respective proteins (Figures S2B and S2C); we chose the most potent shRNAs, shPmpca.3 and shPmpcb.3 (hereinafter termed shPmpca and shPmpcb), for all subsequent experiments.

Stable KD of Pmpca or Pmpcb was constructed in p19^Arf−/− KO mice by hydrodynamic injection of pCaNIG-shPmpca, pCaNIG-shPmpcb, or pCaNIG-shNC, followed by sorafenib or vehicle administration (Figure 2A). Notably, KD of Pmpca or Pmpcb itself did not influence the survival outcome or liver weights of untreated mice (Figures 2B and 2C; Figure S3). However, KD of Pmpcb did increase the susceptibility of autochthonous HCC tumors to sorafenib, manifested both as better survival outcome and a decrease in liver tumor burden for animals receiving sorafenib administration (p < 0.05); KD of Pmpca did have similar effects, but they were not significant (p > 0.05; Figures 2B and 2C; Figure S3). TUNEL (terminal deoxynucleotidyl transferase dUTP nick end labeling) and Ki67 analyses revealed a significant impact of Pmpcb KD on apoptosis and proliferation under sorafenib therapy (p < 0.05); KD of Pmpca had a similar effect, but it was not significant (p > 0.05; Figures 2D–2F; Figure S4). Overall, the results implicate that the susceptibility to sorafenib is stronger via the silencing of Pmpcb expression, which is primarily mediated by inducing apoptosis. Therefore, we chose to continue our investigation on the Pmpcb subunit of Pmpc as opposed to its Pmpca subunit.

Figure 2.

Pmpcb KD by Transposon-Vectored shRNA Sensitizes Murine HCCs to Sorafenib

(A) Experimental setup to assess the influence of Pmpca and Pmpcb KD by shRNAs in NrasG12V; p19^Arf−/− HCC mouse models. (B) Kaplan-Meier survival curves for p19^Arf−/− mice receiving pCaNIG-shPmpca, pCaNIG-shPmpcb, or pCaNIG-shNC transposon vector injections with the administration of either sorafenib or vehicle beginning 1 week following injection. The p values were obtained by a log rank test (n = 12 mice per group); *p < 0.05, **p < 0.01 versus shNC, carrier; †p < 0.05, ††p < 0.01 versus shPmpcb, carrier. (C) Liver weights in p19^Arf−/− animals 5 weeks following injection of pCaNIG-shNC, pCaNIG-shPmpca, or pCaNIG-shPmpcb vector transposons with concurrent sorafenib or vehicle treatment (n = 12 mice per group). (D) Typical images and (E) quantification of cells positive for TUNEL staining as well as (F) quantification of Ki67 staining in p19^Arf−/− tumor puncta 5 weeks following injection of pCaNIG-shNC, pCaNIG-shPmpcb, or pCaNIG-shPmpca vector transposons with concurrent sorafenib or vehicle treatment (n = 12 mice per group). p values were obtained by a two-tailed Student’s t test. (G) Experimental setup for generating liver tumors driven by NrasG12V with inducible expression of shRNAs. (H) Kaplan-Meier survival curves for wild-type (WT) mice following orthotopic implantation of NrasG12V; p19^Arf−/− HCC cells with stable expression of MSCV-rtTA3 with TtGMP-shNC or TtGMP-shPmpcb and administered sorafenib and doxycycline (dox) beginning 1 week following injection (shPmpcb – dox, shNC + dox, shPmpcb + dox). The p values were obtained by a log rank test (n = 12 mice per group); *p < 0.05, **p < 0.01 versus shPmpcb (−dox); †p < 0.05, ††p < 0.01 versus shNC (+dox). (I) Typical camera images of tumor liver burden 6 weeks following orthotopic implantation of NrasG12V; p19^Arf−/− HCC cells with expression of TtGMP-shNC or TtGMP-shPmpcb and sorafenib and dox administration (scale bar, 1 cm). Error bars express the means ± SEMs.

We next employed an HCC mouse model of doxycycline-induced conditional RNAi to determine if concurrent administration of sorafenib with Pmpcb KD would be effective against aggressive, late-stage HCC. Immunocompetent, WT C57BL/6 mice were orthotopically implanted with Nras^G12V; p19^Arf−/− HCCs, which had stable expression of Pmpcb shRNA or scrambled shRNA under the control of a doxycycline-inducible promoter (pTtGMP and rtTA3 vectors; Figure 2G). Mice received sorafenib or vehicle concurrently with doxycycline 1 week following orthotopic implantation of HCCs (Figure 2G). Remarkably, even in late-stage HCC tumors, sorafenib with doxycycline-induced Pmpcb KD led to lower tumor burden (p > 0.05; Figures 2H and 2I) and improved survival (p > 0.05; shNC versus shPmpcb, +doxycycline). Interestingly, shPmpcb tumors induced with doxycycline expressed little or no GFP, which is co-expressed with shRNAs and implies a low level of Pmpcb KD (Figure 2I), suggesting the actual therapeutic effect of sorafenib with Pmpcb KD may even be greater.

Silencing PMPCB Expression by Oral Nanoparticle-Delivered shRNA Improves the Therapeutic Effectiveness of Sorafenib

To validate our experiments of genetic Pmpcb KD, we next ventured to test whether oral nanoparticle-delivered Pmpcb shRNA would have the same effect against aggressive, late-stage HCC. We selected the pluronic P85-polyethyleneimine/d-α-tocopheryl polyethylene glycol 1,000 succinate nanocomplex (SSN) shRNA delivery system specifically developed and validated by Shen et al.¹³ for drug-resistant HCC tumors. For initial in vitro validation, we found that shPmpcb-loaded SSNs effectively lowered the levels of Pmpcb in HCC cells, with a shPmpcb/polymer mass ratio of 8.0 providing the optimal suppression (Figure S5).

Prior to our sorafenib experiments, the effects of shPmpcb-loaded SSNs on proliferation, cell cycle phase distribution, and apoptosis levels in various types of mouse and human HCC cell lines were analyzed to determine the effects of Pmpcb silencing in the absence of sorafenib. We found that Pmpcb silencing alone did not affect these characteristics in Nras^G12V; p19^Arf−/−, Nras^G12V/Akt-1; p19^Arf−/−, and c-myc/Akt-1; p19^Arf−/− cells, but it did affect these characteristics in Hep3B and Huh7 cells (Figure S6). These findings are consistent with previous research demonstrating that Pmpcb silencing promotes apoptosis in EpCAM+ HCC cells (such as Hep3B and Huh7 cells¹⁴), but not in EpCAM– cells.¹⁵

Next, we assessed sorafenib and shPmpcb-loaded SSN combination therapy in p19^Arf−/− mice with de novo autochthonous liver tumors generated by oncogenic Nras^G12V. At 1 week after tumors were induced, treatment was started with sorafenib, shPmpcb-loaded SSNs, sorafenib + shPmpcb-loaded SSN combination therapy, or vehicle (empty SSNs). Combination sorafenib + shPmpcb-loaded SSN treatment was the most effective, and it led to attenuated liver tumor burden and improved survival outcome for mice compared to sorafenib monotherapy (Figures 3A–3C).

Figure 3.

Silencing PMPCB Expression by Oral Nanoparticle-Delivered shRNA Improves the Therapeutic Effectiveness of Sorafenib

(A) Typical images and (B) liver weights in p19^Arf−/− animals 4 weeks following stable integration of Nras^G12V (pCaN) with administration of vehicle, sorafenib, shPmpcb-loaded SSNs (shPmpcb), or sorafenib + shPmpcb combination (typical camera images scale bar, 1 cm). (C) Kaplan-Meier survival curves for p19^Arf−/− mice following Nras^G12V injection with administration of treatment as outlined in Figure 2A. The p values were obtained by a log rank test; *p < 0.05, **p < 0.01 versus Carrier; †p < 0.05, ††p < 0.01 versus shPmpcb; ‡p < 0.05, ‡‡p < 0.01 versus sorafenib. (D–H) Crystal violet-stained colony formation assay of (D) Nras^G12V; p19^Arf−/− cells, (E) Nras^G12V/Akt-1; p19^Arf−/− cells, (F) c-myc/Akt-1; p19^Arf−/− cells, (G) Hep3B cells, and (H) Huh7 cells following 4 days of incubation with DMSO, sorafenib, shPmpcb, or sorafenib + shPmpcb combination (typical camera images). All in vitro experiments: n = 3 biological replicates × 3 technical replicates. Error bars express the means ± SEMs.

We corroborated our results of sorafenib + shPmpcb-loaded SSN combination treatment on various types of mouse and human HCC cell lines in vitro. Sorafenib + shPmpcb-loaded SSN combination treatment suppressed colony formation in murine Nras^G12V; p19^Arf−/− cells (Figure 3D), Nras^G12V/Akt-1; p19^Arf−/− cells (Figure 3E), and c-myc/Akt-1; p19^Arf−/− cells (Figure 3F) to a greater extent than either treatment alone, which only induced a modest decrease in the number of colonies. Combination sorafenib + shPmpcb-loaded SSN treatment also reduced colony formation in the human HCC cell line Hep3B and Huh7 cells (Figures 3G and 3H). Consistent with our previous findings (Figure S6), shPmpcb alone also reduced colony formation in Hep3B and Huh7 cells (Figures 3G and 3H). Cumulatively, our findings advocate sorafenib with concurrent silencing of PMPCB to treat genetically diverse human and mouse HCCs.

Pmpcb Silencing and Mapk14 Silencing Show Similar Efficacies in Improving the Therapeutic Effectiveness of Sorafenib

Rudalska et al.’s⁸ previously published work in the Nras^G12V; p19^Arf−/− mouse model uncovered mitogen-activated protein kinase 14 (Mapk14) as a protein conferring resistance to sorafenib. Therefore, here we subjected shMapk14-treated and shPmpcb-treated Nras^G12V; p19^Arf−/− mice and HCC cell lines to sorafenib therapy to compare the effects of Mapk14 silencing versus Pmpcb silencing on the susceptibility to sorafenib. For initial in vitro validation, we found that shMapk14-loaded SSNs effectively lowered the levels of Mapk14 in HCC cells, with a shMapk14/polymer mass ratio of 8.0 providing the optimal suppression (Figure S7A).

At 1 week after de novo autochthonous liver tumors were induced in p19^Arf−/− mice, treatment was started with carrier (empty SSNs), sorafenib, sorafenib + shMapk14-loaded SSN combination therapy, or sorafenib + shPmpcb-loaded SSN combination therapy. Sorafenib + shMapk14-loaded SSN therapy and sorafenib + shPmpcb-loaded SSN therapy were similarly effective, and they led to attenuated liver tumor burden and improved survival outcome for mice compared to sorafenib monotherapy (Figures S7B and S7C). We corroborated our results on mouse and human HCC cell lines in vitro. Sorafenib + shMapk14-loaded SSN therapy and sorafenib + shPmpcb-loaded SSN therapy were similarly effective in suppressing colony formation in murine Nras^G12V; p19^Arf−/− cells (Figure S7D) as well as the human HCC cell line Hep3B and Huh7 cells (Figures S7E and S7F) to a greater extent than sorafenib alone. Cumulatively, our findings suggest that PMPCB silencing is as effective as MAPK14 silencing in sensitizing HCC cells to sorafenib therapy.

PMPCB Silencing Enhances Pink1-Parkin Signaling and Downregulates Mcl-1 in HCC Cells under Sorafenib Treatment

The majority of human HCC tumors display upregulated RAF-MEK-ERK signaling, which contributes strong oncogenic activation to liver cell proliferation.16, 17 Consistent with sorafenib’s strong inhibition of RAF proteins, sorafenib-treated cells display a significantly lower level of RAF-, MEK-, and ERK-dependent signaling events.7, 8 Notably, the MEK-ERK pathway also upregulates transcription of the downstream pro-malignant, anti-apoptotic protein MCL-1,¹⁸ which is also upregulated in human HCC tumors.19, 20 Interestingly, MCL-1 degradation is promoted by PINK1-Parkin signaling,²¹ which is inhibited by PMPC.²²

These combined findings suggest that PMPCB silencing would enhance PINK1-Parkin signaling and downregulate MCL-1 in HCC tumors. To verify this hypothesis, we performed western blotting on HCCs expressing shPmpcb versus shNC under untreated and sorafenib-treated conditions. We noted a significant increase in Pink1 expression and Parkin phosphorylation coupled with downregulation of Mcl-1 in HCCs with Pmpcb KD under sorafenib-treated conditions, indicative of a healthier, pro-apoptotic phenotype (Figure 4A).

Figure 4.

MCL-1 Expression Positively Correlates with Poorer Response to Sorafenib Therapy

(A) WB of Nras^G12V; p19^Arf−/− HCCs expressing shPmpcb or shNC following 3-day treatment with sorafenib or vehicle, with the indicated primary antibodies. *p < 0.05, **p < 0.01 versus shNC, DMSO; †p < 0.05, ††p < 0.01 versus shPmpcb, DMSO; ‡p < 0.05, ‡‡p < 0.01 versus shNC, sorafenib. (B and C) WBs of (B) Nras^G12V; p19^Arf−/− HCC cells and (C) Hep3B HCC cells following 3 days or 8 weeks of treatment with sorafenib. *p < 0.05, **p < 0.01 versus 3 days of DMSO; †p < 0.05, ††p < 0.01 versus 8 weeks of DMSO. (D) Clonogenic assay of long-term-treated Hep3B HCC cells (8 weeks of sorafenib exposure) that were exposed to an additional 4 days of treatment with sorafenib or vehicle. (E) Typical PINK1 IHC images of patient-derived HCC sections prior to treatment with sorafenib. High PINK1 (above median expression, panel on left) and low PINK1 (below median expression, panel on right) are shown (scale bar, 50 μm). Kaplan-Meier survival curves for HCC patients receiving sorafenib correlated to low versus high PINK1 protein levels. The p values were obtained by a log rank test (n = 40 patients per group). (F) Typical MCL-1 IHC images of patient-derived HCC sections prior to treatment with sorafenib. High MCL-1 (above median expression, panel on left) and low MCL-1 (below median expression, panel on right) are shown (scale bar, 50 μm). Kaplan-Meier survival curves for HCC patients receiving sorafenib correlated to low versus high pre-treatment MCL-1 protein levels. The p values were obtained by log rank test (n = 40 patients per group). (G) Cell count quantification and (H) WB of Pink1 and Mcl-1 expression for murine Nras^G12V; p19^Arf−/−, Nras^G12V/Akt-1; p19^Arf−/−, Nras^G12V/c-myc; p19^Arf−/−, and c-myc/Akt-1; p19^Arf−/− HCC cells treated for 3 days with sorafenib or vehicle. All in vitro experiments: n = 3 biological replicates × 3 technical replicates. Error bars express the means ± SEMs.

MCL-1 Expression Positively Correlates with Poorer Response to Sorafenib Therapy

To test whether long-term sorafenib therapy affects PINK1-Parkin-MCL-1 signaling in HCC cells, we treated human and mouse HCCs with sorafenib for a short duration (3 days) or long duration (8 weeks), and we compared PINK1, p-Parkin, and MCL-1 protein levels. Following long-term sorafenib treatment, we observed downregulated PINK1 expression and Parkin phosphorylation coupled with MCL-1 upregulation in HCC cells (Figures 4B and 4C), with a concurrent loss of susceptibility to sorafenib (Figure 4D).

To corroborate these observations in patients, we observed various levels of MCL-1 upregulation in human HCC tumor specimens prior to initiating sorafenib therapy (n = 40 patients) (Figure S8). Since repeated biopsies could not be obtained from the same HCC patient to examine the time course of PINK1-Parkin signaling during sorafenib therapy, we opted to correlate PINK1 and MCL-1 levels with HCC patient survival. This study revealed an inverse relationship between PINK1 and MCL-1 biopsy levels just prior to sorafenib treatment commencement with poorer prognosis after starting sorafenib treatment (Figures 4E and 4F). Of the two biomarkers, high (above median) MCL-1 staining correlated more significantly with poorer outcome (p < 0.001). The correlation between pre-treatment human HCC tumor MCL-1 expression to sorafenib susceptibility was further validated in various murine HCC cell lines, which demonstrated that elevated Mcl-1 expression was present in HCC cells resistant to sorafenib (Figures 4G and 4H).

Mcl-1 Silencing Sensitizes Sorafenib-Resistant Murine HCC Cells to Sorafenib

Consequently, we determined whether the attenuation of Mcl-1 signaling could enhance the susceptibility to sorafenib. Mcl-1 KD by shMcl-1.1 and shMcl-1.2 resulted in stable attenuation of Mcl-1 expression in mouse HCC cells (Figure 5A). Mcl-1 KD combined with sorafenib lowered cellular proliferation more significantly than treatment with only sorafenib (Figures 5B and 5C). Since Mcl-1 KD in the absence of sorafenib treatment only marginally lowered proliferation (Figures 5B and 5C), the impact of Mcl-1 KD with sorafenib on proliferation was attributed to Mcl-1-mediated sensitization to sorafenib.

Figure 5.

Mcl-1 Silencing Sensitizes Sorafenib-Resistant Murine HCC Cells to Sorafenib

(A) Quantitative real-time PCR analysis of Mcl-1 transcript expression in Nras^G12V; p19^Arf−/− HCC cells stably expressing shMcl-1.1 and shMcl-1.2 versus a scrambled control shRNA. (B) Doubling times of Nras^G12V-shMcl-1.1, Nras^G12V-shMcl-1.2, and Nras^G12V-shNC HCCs upon sorafenib or vehicle treatment for 4 days. The p values were obtained by two-tailed Student’s t test. (C) Clonogenic assay of Nras^G12V-shMcl-1.2 and Nras^G12V-shNC cells stained with crystal violet following 8 days of treatment with sorafenib or vehicle (typical camera images shown). (D) Crystal violet-stained clonogenic assay of Nras^G12V; p19^Arf−/− HCC cells following 4 days of the outlined treatment (typical camera images shown). (E) Crystal violet-stained clonogenic assay of Nras^G12V; p19^Arf−/− HCC cells following 4 days of the outlined treatment (typical camera images shown). (F) Schematic representation of Raf-Mek-Erk and Pmpc-Pink1-Mcl-1 signaling in relation to resistance and susceptibility to sorafenib. (G) Clonogenic assay stained with crystal violet of sorafenib-resistant Hep3B human HCC cells following 4 days of treatment with sorafenib, MCL-1 inhibitor S63845, sorafenib + S63845 combination, or vehicle (typical camera images shown). All in vitro experiments: n = 3 biological replicates × 3 technical replicates. Error bars express the means ± SEMs.

Raf targets other proteins downstream in addition to Mek-Erk signaling,23, 24 which we tested by combining the inhibition of Raf by vemurafenib with Mcl-1 inhibition by S63845 or concurrently with Mek1/2 inhibition by PD325901. Inhibition of both Raf and Mcl-1 produced the same phenotype as combination sorafenib + Mcl-1 inhibition therapy (Figure 5D). In contrast, however, the inhibition of both Mek1/2 and Mcl-1 only induced a modest decrease in proliferation (Figure 5E). The anti-proliferative effect was restored upon combined inhibition of Mek1/2, Mcl-1, and Raf (Figure 5E). Similarly, combined Mek1/2 and Raf inhibition with shRNA-induced Mcl-1 KD produced strong anti-proliferative action (Figure 5E). Cumulatively, these findings imply a combinatorial strategy where the greatest therapeutic effect against sorafenib resistance in HCC tumors depends on targeting three points simultaneously: (1) inhibiting Mek directly, (2) inhibiting targets other than Mek-Erk signaling downstream of Raf by inhibiting Raf directly, and (3) knocking down Pmpcb or inhibiting Mcl-1 (resistance pathway) (Figure 5F). We partially validated our model by showing that sorafenib-resistant Hep3B cells become susceptible to combination sorafenib + MCL-1 inhibition therapy (Figure 5G).

Discussion

Mitoproteases identify and cleave mitochondrial importing signals necessary for the proper couriering of proteins into the mitochondria, and they assist in their proper folding and activation.¹¹ The heterodimer mitoprotease PMPC is composed of a proteolytic beta subunit PMPCB and a non-proteolytic alpha subunit PMPCA.¹¹ PMPCA binds to the presequence signals of the imported proteins, thereby enabling proteolytic processing by PMPCB.¹¹ On account of disruptions in mitochondrial metabolism commonly found in cancer cells, mounting research has unveiled dysregulated expression of several mitoproteases in human cancers, such as the mitoproteases LONP, HTRA2, and METAP1D.¹¹ Indeed, PMPCB has also been identified to be upregulated in breast cancer cells,²⁵ but its role in HCC remains largely uncharacterized. Our current work identified silencing of PMPCB as a potential strategy to sensitize HCC tumors to sorafenib treatment. The Nras^G12V; p19^Arf−/− murine model of sorafenib-resistant HCC used here is highly flexible, and it can be employed for investigating other potential sorafenib-sensitive or -resistant genes in HCC.

Our findings advocate sorafenib with concurrent silencing of PMPCB to treat genetically diverse human and mouse HCCs. Interestingly, we did find that PMPCB silencing alone (in the absence of sorafenib therapy) promotes apoptosis in EpCAM+ HCC cells (i.e., Hep3B and Huh7 cells¹⁴), but not in EpCAM– cells.¹⁵ PMPCB blockade alone has been shown to produce mitochondrial dysfunction in both EpCAM+ HCC cells and EpCAM– HCC cells.²⁶ However, only EpCAM+ HCC cells undergo apoptosis in response to PMPCB blockade alone,²⁶ suggesting that EpCAM+ HCC cells are more dependent on PMPCB for survival. These findings suggest that combination sorafenib + Pmpcb KD therapies may be especially effective in HCC patients with EpCAM+ tumors.

Our current results can be applied for translational purposes, as combination sorafenib + Pmpcb KD therapies can be developed and tested in other animal models and, eventually, HCC patients. To our knowledge, there are no commercially available PMPCB-specific inhibitors that have been tested in vitro or in vivo. Only the non-specific metal-ion chelators O-phenanthroline and EDTA have been demonstrated to have efficacy against PMPC activity in vitro.27, 28 Unfortunately, the administration of metal-ion chelators does not result in changes to PINK1 expression.²⁹ Only PMPCB KD (as opposed to chemical inhibition) has been shown to upregulate PINK1 levels,²² suggesting that lowering PMPCB expression is necessary for affecting the sorafenib-sensitizing PINK1-MCL-1 pathway. Therefore, rather than chemical inhibition of PMPC activity, the nanoparticle-based shPMPCB delivery system used here¹³ may be of more clinical interest. Indeed, there are several phase I clinical trials underway that are testing similar nanoparticle-based shRNA and small interfering RNA (siRNA) delivery systems targeting specific oncogenes in advanced cancer patients (e.g., ClinicalTrials.gov: NCT01505153 delivering pbi-shRNA-STMN1-loaded lipid nanoparticles and ClinicalTrials.gov: NCT01591356 delivering siRNA-EphA2-DOPC-loaded liposomes).³⁰

Our current findings are not limited to the effects of PMPCB KD, as we show that KD of MCL-1 also promotes sensitization to sorafenib in sorafenib-resistant HCC cells. MCL-1 is a potent anti-apoptotic protein in the B cell lymphoma 2 (BCL-2) family consisting of other potent anti-apoptotic proteins, such as BCL-2, BCL-xL, and BCL-w;³¹ notably, MCL-1 upregulation contributes to oncogenesis and drug resistance in a variety of solid tumors,32, 33, 34, 35, 36 including sorafenib resistance in HCC cells.37, 38 Therefore, based on previous and current findings, a combinatorial strategy involving RAF, MEK, and MCL-1 inhibition could show promise in sorafenib-resistant HCC patients (Figure 5F). Several MCL-1-specific small molecule inhibitors are now under development and testing, which could be used in combination with sorafenib; specifically, AMG 176 is the first MCL-1 inhibitor to undergo phase I testing (ClinicalTrials.gov: NCT02675452), while several alternative MCL-1 inhibitors (e.g., Les Laboratories Servier & Vernalis’s thienopyrimidine scaffold-based compounds and Amgen’s tetrahydronaphthalene scaffold-based compounds) are still undergoing preclinical in vivo testing.³⁹ The important question remains as to whether the use of multiple agents will have negligible liver toxicity effects in human subjects. As the majority of HCC cases occur in patients with cirrhotic liver disease and compromised liver function, a combinatorial strategy using nanoparticle-mediated shRNA delivery may have a better chance of moving forward into clinical trials.

With respect to the future design of clinical trials to investigate the efficacy of combining sorafenib therapy with PMPCB KD or MCL-1 KD in HCC patients, it should be mentioned that PMPCB KD or MCL-1 KD can rescue sorafenib sensitivity in HCC tumors that have developed sorafenib resistance from long-term treatment. This information suggests that trials could be structured such that HCC patients initially undergo sorafenib monotherapy followed by an additional trial investigating sorafenib with PMPCB KD or MCL-1 KD. As our data suggest that MCL-1 may be a biomarker in predicting poor sorafenib response, future clinical research should also investigate whether HCC patients with higher pre-therapeutic tumor MCL-1 protein expression are better candidates for sorafenib combination regimens.

Materials and Methods

Design of Vectors and shRNAs

Vectors possessing Sleeping beauty 13 (SB13) transposase and the Nras^G12V, Nras^G12V/Akt-1, Nras^G12V/c-myc, and c-myc/Akt-1 transposon plasmids were constructed according to previously published methods.⁸ Briefly, the sequence for Nras^G12V in the vector was substituted by a polylinker, then sequences for Nras^G12V, Akt-1, and/or c-myc were amplified by PCR with primers containing MluI, AscI, and either AgeI or NotI sites, which were then cloned into the plasmid to form Nras^G12V, Nras^G12V/Akt-1, Nras^G12V/c-myc, and c-myc/Akt-1 plasmids. The sequence for IRES-GFP was inserted into the vector between NotI and AgeI sites. Then, a miR-30-based shRNA construct⁴⁰ was inserted using AgeI and NheI 3′ from IRES-GFP. This miR-30-based shRNA construct consists of the following (5′ to 3′): a U6 promoter, a small nuclear RNA (snRNA) leader sequence, the 5′ end of the miR-30-flanking region, the shRNA sequence, the 3′ end of the miR-30-flanking region, and a pol III termination signal.

The individual shRNAs against murine p19^Arf (p16^Ink4a, Cdkn2a, Mts1, and Pctr1), murine Pmpca, murine Pmpcb, murine Mcl-1, human PMPCB, and non-coding scrambled controls were designed and synthesized by Shanghai Genechem (Shanghai, China). The sequences are detailed in Table S1. These shRNAs were inserted between XhoI and EcoRI sites within a murine stem cell virus (MSCV) plasmid, which was then cloned into the miR-30-based shRNA construct described above between the XhoI, MluI, and AscI restriction sites. The shRNAs were subcloned to preserve a 1,000-fold excess of shRNA colonies.

HCC Patient Samples

Liver tumor biopsy specimens were obtained from 40 HCC patients at Shanghai Pudong Hospital (Fudan University Pudong Medical Center) between June 2013 and June 2015. All tissue donors provided written informed consent prior to the procedure. Eligible participants possessed (1) a radiographic, histologic, or cytologic diagnosis of unresectable HCC with (2) a life expectancy of at least 8 weeks and were (3) scheduled to begin sorafenib treatment within the next 2 weeks. Exclusion criteria were the following: (1) a previous history of sorafenib therapy, or (2) the presence of any contraindication(s) according to the locally approved sorafenib protocol. Patients with previous locoregional interventions for HCC, including transarterial chemoembolization (TACE), percutaneous radiofrequency ablation, and/or percutaneous ethanol injection, were allowed to participate. The clinicopathological characteristics of these HCC patients are detailed in Table S2. The biopsy procedure was executed using a coaxial needle technique under ultrasound guidance that enabled repetitive sampling of core tumor tissue and adjacent normal liver tissue. Samples were routinely processed for diagnostic histopathology and immunohistochemical analysis. Participants were followed for a minimum period of 24 months after initiating sorafenib therapy.

Generation of Mouse Strains and Treatment Protocols

Mice used in all studies possessed a C57BL/6 genetic background, and randomized cohorts comprised an equal gender distribution. Sample sizes were not statistically predetermined. C57BL/6 WT animals were procured from Janvier (Saint Berthevin, France) or Harlan (Rossdorf, Germany). Mice with p19^Arf tumor suppressor KO (p19^Arf−/−) were created by the Sherr lab (St. Jude Children’s Research Hospital, Memphis, TN, USA) and procured as a C57BL/6 strain from the Lowe lab (Memorial Sloan Kettering Cancer Center, New York, NY, USA).

Hydrodynamic injections in the tail vein were performed used plasmids purified with an EndoFree Plasmid Maxi Kit (QIAGEN, Hilden, Germany). Plasmids were solubilized in 0.9% NaCl solution, and the total volume constituted 10% of body weight. Mice (aged 6 weeks) were injected with transposase (5 μg) and transposon plasmids (25 μg) within 10 s. Animals that did not receive an efficient injection were not included in the studies.

To generate the orthotopically implanted HCC models, mice (aged 10 weeks) underwent anesthesia (xylazine and ketamine) followed by laparotomy. One million (1.0 × 10⁶) murine HCC cells were injected through the incision into the left liver lobe, and the peritoneum was rinsed with sterile distilled water to prevent seeding of cancer cells within the abdominal cavity. Doxycycline (2 mg/mL; Sigma-Aldrich, St. Louis, MO, USA) and saccharose (10 mg/mL) were included in the drinking water.

Mice were administered sorafenib (100 mg/kg; Nexavar) in 50:50 4× cremophor EL:95% ethanol, which were administered by oral gavage every other day. The dose of sorafenib was based on previous studies in murine models of HCC.8, 41, 42 Livers were collected and weighed by two blinded investigators from mice (aged 10 weeks) that were anesthetized with xylazine and ketamine.⁴³ Collected liver samples were imaged by camera. Imaging of GFP distribution within liver tissue was performed on an imaging system (Hamamatsu Photonics).

Nanoparticle-Based Delivery of shPmpcb

P85-PEI/D-α-tocopheryl polyethylene glycol succinate (TPGS) nanomicelles (NMs) were constructed as previously described using thin-film hydration.¹³ Briefly, P85-PEI and TPGS (50 mg each) were dissolved in 10 mL methanol, which was subjected to rotary evaporation for 1 h (35°C). The residual methanol solvent was removed by vacuum at room temperature overnight, and the thin film was hydrated with ultra-pure H₂O. shPmpcb-loaded P85-PEI/TPGS complex nanoparticles (shPmpcb-SSNs) were constructed by combining various concentrations of shPmpcb and NMs to achieve a 6.0–10.0 shPmpcb/polymer mass ratio. The mixture was vortexed and incubated for 30 min at room temperature. Vehicle SSNs were prepared in an identical way with no addition of shPmpcb.

In vitro transfection was conducted with the indicated HCC cell lines using shPmpcb-SSNs, with vehicle SSNs as a negative control, as previously described.¹³ Briefly, HCC cells were seeded at 1.0 × 10⁵ cells/well in 500 mL RPMI-1640. shPmpcb-SSNs displaying a shPmpcb concentration of 4 mg/mL (or an equivalent volume of vehicle SSNs) were incubated with the HCC cells for 48 h. The Pmpcb silencing efficiency was visualized via western blotting. For in vivo delivery of SSNs, murine tail vein injection was performed two times every other day with shPmpcb-SSNs (at a shPmpcb concentration of 2 mg/kg) or an equivalent volume of vehicle SSNs.

Tissue Histopathology (H&E), Immunohistochemistry, and In Situ Hybridization

Tissue samples were embedded in paraffin, and sections were subjected to H&E staining and assessed by qualified pathologists. Immunohistochemistry (IHC) for proliferation (Ki67 antibody, 1:200; ab15580, Abcam, Cambridge, UK) and in situ hybridization (ISH) for apoptosis by TUNEL (Roche, Basel, Switzerland) were performed on sectioned flash-frozen tissue. Sampling was performed on 10 random images (400× magnification), which were scored by blinded investigators.

IHC staining of deidentified, numbered patient-derived HCC tissue sections was performed by an automated immunostaining instrument (Ventana BenchMark XT, Ventana Medical Systems, Oro Valley, AZ, USA), according to the manufacturer’s instructions. Antigen retrieval was performed in citrate buffered at pH 6, and staining was accomplished with primary antibodies against PINK1 (1:50; PA5-13402) and MCL-1 (1:50; PA5-11389), both from Thermo Fisher Scientific (Waltham, MA, USA). Protein levels were semiquantitatively analyzed employing an immunoreactive score, as reported in the literature.⁴⁴

Cell Culturing and Viral Production

Cells were grown in DMEM (with 10% fetal calf serum [FCS]) and incubated at 37°C in an atmosphere of 7% CO₂. Cultures were routinely tested for mycoplasma contamination by a PCR protocol. The human HCC cell lines Hep3B and Huh7 were procured from ATCC. Primary murine HCC cells were derived from mouse tumor tissue in sterile conditions by dissociating tumors with collagenase (0.1 U/mL) and dispase (1,000 U/mL; both from Roche) in DMEM buffered with HEPES for a duration 30 min (37°C, gentle rocking). Cells were strained by a nylon mesh filter (100 μm), pelleted (80 × g, 10 min), and rinsed twice with media before being deposited in fresh media in a culture dish (coated with 0.1% gelatin). Retroviruses were produced in Phoenix packaging cells, which were transfected using calcium phosphate with retroviral DNA. Target cells were then infected directly with viral supernatant supplemented with polybrene (1–10 μg/mL) to increase infectivity, and they underwent selection on puromycin (1–10 μg/mL) or hygromycin (300–1,000 μg/mL).

Cell Proliferation and Colony Formation Assays

Drugs were tested on cells grown in vitro in plates. Proliferation was assessed by a doubling assay, represented as a fold change compared to the starting point, using a Guava flow cytometer (Merck, Darmstadt, Germany). Tested drugs were formulated in DMSO and added to cells 1 day after plating for the following drugs: (1) sorafenib to primary mouse HCC cells (8 μM), Hep3B (2 μM), and Huh7 (4 μM); (2) S63845 (1 μM); (3) vemurafenib (PLX4032; 16 μM); (4) PD325901 (10 μM); or (5) appropriate combination treatment. These drug doses were directly based on Rudalska et al.’s⁸ dosing procedure. Colony formation (clonogenic) assays were performed on 10,000 plated cells stained with crystal violet solution (0.07%), and visualization and scoring to compare treatment with drug versus DMSO were performed by known methods.

Western Blot

Cells were rinsed in buffer (PBS, pH 7.4) followed by disruption in lysis buffer (50 mM Tris buffered at pH 7.5 with 150 mM NaCl and 0.5% NP-40 detergent, supplemented with protease and phosphatase inhibitors). Wells in polyacrylamide gels were loaded with lysate containing 50–80 μg protein, which were resolved by electrophoresis under denaturing conditions (SDS-PAGE). Proteins were then transferred by a semidry blotting method to nitrocellulose membranes (Hybond ECL; GE Healthcare, Chicago, IL, USA). Blots were exposed to the following primary antibodies: α-tubulin (1:3,000; 11H10; 2125, Cell Signaling Technology), PMPCA/Pmpca (1:200; sc-390471, Santa Cruz Biotechnology), PMPCB/Pmpcb (1:1,000; PA5-24912, Thermo Fisher Scientific), PINK1/Pink1 (1:1,000; PA5-13402, Thermo Fisher Scientific), phospho-Parkin (Ser65) (1:1,000; ab154995, Abcam), total Parkin (1:1,000; PA5-13399, Thermo Fisher Scientific), MCL-1/Mcl-1 (1:500; PA5-27597, Thermo Fisher Scientific), pro-caspase-3 (1:10,000; ab32499, Abcam), and cleaved PARP (1:1,000; ab32064, Abcam). Blots were then incubated with peroxidase-conjugated secondary antibodies, and exposure was captured by a ChemiDoc MP Imaging System (Bio-Rad, Hercules, CA, USA). Band intensities were measured via densitometric analysis with ImageJ software (NIH, Bethesda, MD, USA).

Quantitative Real-Time PCR

mRNA purification was performed from cells using TRIzol (Invitrogen, Thermo Fisher Scientific, Waltham, MA, USA) and a RNeasy Mini Kit from QIAGEN. cDNA was synthesized by a TaqMan cDNA Synthesis Kit (Applied Biosystems, Thermo Fisher Scientific) employing random hexamer primers and followed by quantitative real-time PCR with SYBR Green Master Mix (Applied Biosystems). Data were normalized to β-actin transcripts.

Statistical Analysis

GraphPad Prism (GraphPad, La Jolla, CA, USA) was used in all statistical analyses, and p < 0.05 was taken to be significant. Survival for mice receiving hydrodynamic tail vein injection was assessed from the time of injection until the time of death, assuming treatment would increase survival. Survival comparisons between vehicle and treatment groups were performed on Kaplan-Meier plots and calculated by log rank tests. Survival was measured starting from the time sorafenib therapy began for human participants. Survival time was censored for patients who did not experience the investigated event, i.e., were alive at last contact. Protein expression levels were assigned as high or low relative to the median level of expression. A Kaplan-Meier plot was employed to display the correlation between survival and protein expression, and comparisons were made by log rank tests. Other comparisons were made by a two-tailed Student’s t test or one-way ANOVA where appropriate.

Ethics Statement and Animal Care

This project was reviewed and approved by the Ethics Committee at Baoan Central Hospital of Shenzhen and Shanghai Pudong Hospital (Fudan University Pudong Medical Center). Human participants that provided HCC tumor tissue samples for the study signed written informed consent forms before their enrollment. All animal experiments were conducted according to the NIH Guide for the Care and Use of Laboratory Animals (eighth edition). Mice were housed in pathogen-free facilities.

Author Contributions

Design, J.Z. and G.Q.; Experiments and Analyses, S.H., Z.Z., X.G., and L.C.; Drafting, J.Z.

Conflicts of Interest

The authors declare no competing interests.