Mitophagy promotes sorafenib resistance through hypoxia-inducible ATAD3A dependent Axis

 Liver cancer tissues and paried normal tissues were collected form the First Affiliated Hospital of Third Military Medical University and Shanghai Biochip Company Ltd (Shanghai, China). The informed consent was obtained from all patients. 

Human hepatoma cell lines Huh7 and LM3 were obtained from the American Type Culture Collection (ATCC) and authenticated by the Cell Bank of Type Culture Collection of Chinese Academy of Science. The cells were maintained under recommended conditions. For hypoxia treatment, the cells were either cultured in a sealed hypoxia chamber (Thermo fisher, Inc.) containing 1% O₂, 5% CO₂ and 94% N₂, or treated with 100 mM CoCl₂ for 24 or 48 h. Sorafenib-resistant Huh7 (Huh7-SR) and LM3 (LM3-SR) cells were enriched by steadily increasing the drug dose, while hypoxia-induced sorafenib resistant cell lines (Huh7-H-SR and LM3-H-SR) were established by culturing under 1% O₂ with increasing drug doses.

The pLVX-CMV-EGFP-3FLAG-PGK-Puro lentiviral vector expressing full-length human ATAD3A was purchased from SunBio (Shanghai. China). The pMAGic2.1-CMV-HygroR-U6 shRNA lentivirus vector purchased from SunBio (Shanghai. China), and the shRNA sequences are shown in Additional file 1. Human HEK293T cells (American Type Culture Collection) were cultured in 6-well plates till ~ 70% confluent, and co-transfected with 2 μg overexpression or knockdown virus vector, 1 μg pMD2.G and 1 μg psPAX2 lentivirus packaging vectors using Lipofectamine 2000 (Invitrogen) according to the manufacturer’s protocol. The cells were maintained in high-glucose DMEM containing10% FBS, 2 mM glutamine, and100 units/ml penicillin and streptomycin. The supernatants with virus were harvested twice at 48 h and 72 h, filtered through a 0.45 μm syringe filter and frozen in liquid nitrogen.

HCC cells were cultured in complete DMEM, and 1.5 × 10⁵ cells were mixed with 450 μl virus-containing supernatant in the presence of 4 μg/ml polybrene (Sigma). The cells were seeded into a 6-well plate, and the medium was changed 12 h post-infection. After 48 h, the infected cells were trypsinized and seeded into 10 cm culture dish with 4 μg/ml puromycin (Thermo Fisher Scientific). The stably transduced cells were selected over 48 h, and harvested 6 days post-infection to determine knockdown efficiency.

Total protein was extracted from cancer cells using Mammalian Protein Extraction Buffer (P0013, Beyotime, Beijing, China) supplemented with protease inhibitor cocktail (87786, ThermoFisher, USA). The mitochondrial fractions were separated using Mitochondria Isolation Kit (number: SM0020, Solarbio, Beijing, China) according to the manufacturer’s protocol, and the protein was extracted as above. Equal amounts of protein lysates were separated by SDS-PAGE gel and electro-transferred to PVDF membrane (Millipore, USA). After blocking in 5% milk-PBST for 2 h at 37 °C, the membranes were incubated overnight with primary antibodies (Additional file 2) at 4 °C and with secondary antibodies at 37 °C for 2 h. The positive bands were visualized with Immobilon Western Chemiluminescent HRP Substrate detection reagent (Millipore, USA), and acquired using a ChemiDoc™ imaging System (Bio-Rad, USA).

Total RNA was extracted from cancer cells using Trizol Reagent (Invitrogen, USA), and qRT-PCR was performed using SYBR Prime Script RT-PCR kit (TaKaRa, Japan) on a Rotor-Gene 6000 real-time genetic analyzer (Corbett Life Science, USA). The primer sequences and the product sizes are listed in Additional file 3. Glyceraldehyde-3-phosphate dehydrogenase (GAPDH) was used as the internal control. The PCR conditions were as follows: denaturation at 95 °C for 2 mins, followed by 40 cycles of amplification and quantification (95 °C for 5 s, 55 °C–57 °C for 30 s), and melting curve (55 °C–95 °C, with 0.5 °C increment each cycle). Each sample was tested in triplicates.

For apoptosis assays, the suitably treated cells were washed twice with cold PBS, and resuspended in binding buffer at the density of 1 × 10⁶ cells/ml, and distributed into 500 μl aliquots (1 × 10⁵) in 2 ml tubes. 5 μL of Annexin V-FITC were added to each tube. Cells were incubated at room temperature (25 °C) for 15 min in a dark environment, and then analyzed on a FACS Calibur (BD Biosciences, USA), and results were calculated using Cell Quest software (BD Biosciences, USA).

HCC cells were washed, fixed with 5% paraformaldehyde (PFA) and permeabilized in 0.1% Triton X-100. After incubating overnight with monoclonal mouse anti-HIF-1α antibody (Cell signaling technology, #79233) or monoclonal rabbit anti-ATAD3A antibody (Invitrogen, PA5–03671) at 4 °C, the cells were probed with fluorescein isothiocyanate (FITC)-labeled anti-mouse IgG or Cy3-labeled anti-rabbit IgG secondary antibodies (Santa Cruz Biotechnology, Santa Cruz, USA). The nuclei were counterstained with Hochest 33258 and observed under laser confocal scanning microscopy (Leica TCS-SP5, Germany).

The 3′-UTR region of human ATAD3A was amplified by PCR from genomic DNA and cloned downstream to the firefly luciferase coding region in the pMIR-REPORTTM plasmid. The 293 T cells were seeded in 96-well plates, and co-transfected with 100 ng/ml reporter plasmid and 50 nM miR-210-5P or NC mimics using Lipofectamine 2000 24h late. Luciferase activity were measured after 72 h using the Dual-Glo Luciferase Assay System (Promega, Madison, WI, USA). All experiments were performed in triplicates.

Cells were seeded in 96-well plates at the density of 5000 cells/well. After treating with 5.13 μM (Huh7) or 7.92 μM (LM3) sorafenib for varying durations (0 h,12 h,24 h,48 h and 72 h), the cells were incubated for 2 h with 0.5 mg/ml 3-(4, 5-dimethyl-thiazol-2-yl)-2,5-diphenyl-tetrazolium bromide (MTT) (Sigma-Aldrich) in serum free medium. To measure the IC50 of different cell lines, 10^− 3,10^− 2,10^− 1,10⁰,10¹ and 10²μM sorafenib were added in the 96-well plates for 72 h, then treated same as above. The optical densities were measured at 450 nM spectral wavelength using the microplate reader (ELx800; Bio-Tek Instruments, Inc.).

For colony formation assay, 5 × 10² cells were seeded in 6-well plates with sorafenib 5.13 μM (Huh7) or 7.92 μM (LM3) as appropriate. The conditioned medium was removed 3 days later and replaced by medium containing 10% serum for 2 weeks. The ensuing colonies were stained by crystal violet and counted. The experiment was performed thrice.

Western blotting was used to test LC3 mobility shift by analyzing the expression levels of LC3 I/II, and mitochondrial marker protein TOMM20 and TOMM70. Subcellular localization of mitophagy bodies was tracked by transmission-electron-microscopy (TEM, JEM-1230, Japan). The cells were fixed in cooling 2% glutaraldehyde and 0.1 M cacodylate buffer at 4 °C overnight, then exposed to phosphate buffer containing 1% osmium tetroxide for 1 h. Following dehydrated in different concentrations of acetone, these cells were infiltrated and embedded into Epon. The embedded cells were sectioned and stained with 3% uranyl acetate and lead citrate. Mitochondrial content was analyzed by Mito-Tracker staining (number: C1049, Beyotime Biotechnology, Shanghai, China) according to the manufacturer’s protocol and then analysed by FACS (BD Biosciences, USA), and results were calculated using Flow Jowo software (BD Biosciences, USA).

The suitably treated cells were seeded in 6-well plates in triplicate at the density of 5*10⁴ cells/well. After culturing for 2 days, the number of viable cells was counted, and the culture medium was collected. Lactate levels were detected in the latter using a specific analytical kit (Nanjing Jian cheng Bioengineering Institute, China), and normalized to cell number and calculated as relative units per cell.

IHC was performed on tissue array slides as previously described. The tissues were probed with rabbit anti-human ATAD3A (Invitrogen, PA5–03671 1:200) and rabbit anti-human HIF-1α (abcam, ab51608, 1:200) antibodies, and counterstained with hematoxylin (Sigma). The staining intensity (negative = 0, weak = 1, moderate = 2, or strong = 3) and the percentage of positively stained cells (<5%=0, 5% to < 25% = 1, 25 to 50% = 2, > 50 to < 75% = 3, ≥75% = 4) were scored independently by two pathologists. The staining index was calculated by multiplying the intensity score with percentage score, and the samples were classified as negative/low expression or positive/high expression accordingly.

LM3-shControl and LM3-shATAD3A cells were generated as before and total RNA was extracted. RNA libraries were constructed using a TruSeq Stranded mRNA LT Sample Prep Kit (Illumina, San Diego, CA, USA) according to the manufacturer’s instructions, and sequenced on an Illumina platform (HiSeqTM 2500 or Illumina HiSeq X Ten) into 125 bp/150 bp paired-end reads by OE Biotech (Shanghai, China). The differentially expressed genes (DEGs) were identified using Cuff-diff with p value < 0.05 and fold change > 2 as the criteria. Gene set enrichment analysis was performed using software.​broadinstitute.​org/​gsea. Heatmaps were generated with the heatmap package of R program.

Female BALB/c (nu/nu) athymic nude mice, 5 weeks of age, were purchased from hfkbio Inc. Mice were maintained in specific pathogen-free conditions: 20–24 °C, 12/12 h of dark/light cycle, 60 ± 5% of humidity, and plastic cage (3–4 mice/cage). Bedding materials were changed every week, and environmental enrichment was done with sterile materials. All animal experiments were approved by the Institutional Animal Care and Use Committee of the Medicine University of Electronic Science and Technology of China’s. Control and miR-210-5P antagomir-transfected Huh7-shControl and Huh7-shATAD3A cells were inoculated subcutaneously (n = 5 each, 2 × 10⁶ cells per mouse) in the right flank of 4–6 week old BALB/c athymic female nude mice. Once the tumors grew to approximately 100mm³ (around 5 days), the mice were intraperitoneally injected with sorafenib tosylate (10 mg/kg) daily for a week, and monitored every 4 days for the appearance of subcutaneous tumors. We conducted the sacrifice of mice at 28 days, placed the mice in the chamber and introduced 100% carbon dioxide. After we removed each tumor, we maintained the carbon dioxide flow for a minimum of 1 min after respiration ceases. The tumor weight and tumor volume (TV; mm3) were calculated. Tumor volume was calculated as d2 × D/2 (d and D represent the shortest and the longest diameters, respectively). All animal experiments were performed in accordance with the guidelines of the ARRIVE reporting guidelines.

As shown in Supplementary Fig. S1a-1d, HCC cells cultured under hypoxia showed significantly lower apoptosis rates and better cell viability response to sorafenib treatment in a time-dependent manner. To further analyze the role of hypoxia in sorafenib resistance, we established sorafenib resistant Huh7 (Huh7-SR) and LM3 (LM3-SR) cell lines, as well as the corresponding hypoxia-induced sorafenib resistant cell lines (Huh7-H-SR and LM3-H-SR) in the presence of increasing doses of sorafenib under 1% O₂ culture condition (Fig. 1a-c). Compared to the respective control, Huh7-H-SR and LM3-H-SR cells showed increased levels of HIF-1α and drug resistance-related genes, including ABC20, ABCB1, ABCC1 and LRP (Supplementary Fig. S1e). Interestingly, extensive mitophagy was observed in Huh7-H-SR cells compared to the untreated controls, hypoxia treated Huh7 and Huh7-SR cells (Fig. 1d). Consistent with this, the Huh7-H-SR cells showed highest level of PINK1 expression, along with accumulation of PINK1, Parkin and LC3-II in the mitochondrial fraction (Fig. 1e). We also characterized a higher ubiquitination of multiple mitochondrial proteins via Parkin in Huh7-H-SR cells (Fig. 1e). Hypoxia-induced sorafenib resistant Huh7 and LM3 cells lines significantly upregulated PINK1, LC3-II, HIF-1α, FUNDC1 expression, and downregulated TOMM20 and ATAD3A expression (Fig. 1f). Furthermore, TCGA analysis results showed that PINK1 levels were positively correlated with that of ABCB1 and ABCG2 in HCC patients (Fig. 1g-h), and HIF-1α was positively correlated with ABCC1 and negatively with TOMM20 and TOMM70 (Supplementary Fig. S1f-1 h). These results indicated that activated mitophagy is the likely cause of hypoxia-induced sorafenib resistance. In agreement with this, the mitophagy inhibitor 3MA sensitized Huh7-H-SR cells to sorafenib, resulting in a significant decrease in colony yield and proliferation rates (Fig. 1i-j). Furthermore, PINK1 knockdown in Huh7-H-SR cells also sensitized the cells to sorafenib (Fig. 1k-l). Taken together, hyperactivated mitophagy plays a pro-survival role in the hypoxia-induced sorafenib resistant HCC cells.

The mitochondria transmembrane protein ATAD3A was significantly downregulated in the Huh7-H-SR and LM3-H-SR cells compared to the respective controls (Supplementary Fig. S2a). To determine the biological relevance of ATAD3A in HCC, we established stable ATAD3A-overexpressing and knockdown Huh7 and LM3 cell lines (Fig. 2c and Supplementary Fig.S2b-2c). Mitophagy was hyperactivated in the shATAD3A LM3 cells compared to the control (Fig. 2a), which paralleled a marked increase in the polyubiquitination and accumulation of global mitochondrial proteins as seen in the hypoxia-induced sorafenib resistant cells (Fig. 2b). In addition, PINK1 and LC3-II were significantly upregulated, and TOMM20, TOMM70 were downregulated following ATAD3A knockdown (Fig. 2c). Transcriptome sequencing analysis of the control and shATAD3A LM3 cells identified 319 upregulated and 302 downregulated genes in the latter (Supplementary Fig. S2i-2j). Gene set enrichment analysis (GSEA) (Fig. 2f) showed that the shATAD3A cells were commonly enriched with drug resistance genes (NES = -1.939, P = 0.00828 for GEFITINIB resistance; NES = 1.8435; P = 0.02 for TAMOXIFEN resistance) (Fig. 2d, e). The expression levels of these genes were validated by qRT-PCR (Fig. 2g, h). We also confirmed that the expression level of these genes in LM3-H-SR were similar to that in sh-ATAD3A cells compared to shControl (Supplementary Fig. S2f). In addition, shATAD3A and hypoxia-induced sorafenib resistantLM3 cells desensitized the HCC cells to sorafenib, resulting in a significant decrease in apoptotic index (Fig. 2i and Supplementary Fig. S2g-2 h), whereas ATAD3A overexpression augmented the sorafenib-induced cell death (Fig. 2j, k and Supplementary Fig. S2d, 2e). Loss of ATAD3A downregulated the pro-apoptotic PARP, Bax and Cleaved-Caspase 3 expression (Fig. 2l), decreased mitochondrial mass (Fig. 2m), and increased lactate production (Supplementary Fig. S3a-3b) in Huh7 and LM3 cells. Taken together, ATAD3A plays a vital role in mitophagy and sorafenib insensitivity in HCC cells.

To further determine the biological significance of hypoxia and ATAD3A expression in HCC cells, we introduced ectopic ATAD3A in the hypoxia cultured cells. As shown in Fig. 3a, hypoxia-induced mitophagy was abrogated by ATAD3A overexpression, which also upregulated TOMM20 and TOMM70 in Huh7 and LM3 cells (Fig. 3b). Ectopic ATAD3A also restored the mitochondrial mass (Fig. 3c, d) and decreased lactate concentration (Supplementary Fig. S3c, 3d) upon hypoxia-treated LM3 and Huh7 cells. Under hypoxia condition, cancer cells were insensitive to sorafenib treatment, while, gain of ATAD3A expression mitigated the insensitivity as indicated by the fewer colonies (Fig. 3e, f) and increased apoptotic index in LM3 cells (Fig. 3g, h). Western blot confirmed that PARP, Bax and Cleaved-Caspase3 were downregulated in hypoxia cells which were restored in ATAD3A overexpression cell under sorafenib treatment. Furthermore, we confirmed that ATAD3A expression restored sorafenib sensitivity in hypoxia treated Huh7 and LM3 cells via the apoptotic related proteins (Fig. 3i).

To determine the mechanism underlying hypoxia-mediated regulation of ATAD3A expression, we performed a systematic bioinformatics analysis of gene-gene interaction networks based on mutations, copy number alterations, mRNA expression profiles, and protein expression profiles using the STRING database (http://​string-db.​org/​cgi/​network). ATAD3A and the hypoxia-responsive transcriptional factor HIF-1α are located in a network containing 40 nodes (Fig. 4a), suggesting a potential link between HIF-1α signaling and ATAD3A. Furthermore, in situ HIF-1α protein expression in 85 HCC tissues was negatively correlated with that of ATAD3A (Fig. 4b). As shown in Fig. 4c and Supplementary Fig. S4a, hypoxia treatment significantly down regulated ATAD3A expression in hepatoma cells. In addition, both 1% O₂ and CoCl₂ induced HIF-1α expression and downregulated ATAD3A by western blot in Huh7 and LM3 cells (Fig. 4d), and decreased ATAD3A expression in Huh7 cells as a time-dependent manner (Supplementary Fig. S4b). In addition, we found a loss of ATAD3A gene expression in 1% O₂ and CoCl₂ treated Huh7 and LM3 cells (Fig. 4e-f and Supplementary Fig. S4c). Bioinformatics analysis identified hsa-miR-210 and hsa-miR-193b as the most hypoxia-responsive miRNAs (Supplementary Fig. S4d), and ATAD3A as the potential target gene of miR-210-5P (Fig. 4g). Consistent with this, 1% O₂ and CoCl₂ increased miR-210-5p expression after 24 h (Fig. 4h and Supplementary Fig. S4e). Furthermore, miR-210-5P suppressed the expression of reporter gene carrying the wild-type but not mutant 3′-UTR of human ATAD3A gene (Fig. 4j, k). Hepatoma cells transfected with miR-210-5P mimic significantly inhibited ATAD3A expression, whereas miR-210-5p inhibitor had the opposite effect (Fig. 4i, l and Supplementary Fig. S4f). These findings indicated that ATAD3A is a direct target of the hypoxia induced miR-210-5P. This was confirmed by treating hepatoma cells with CoCl₂ and/or the miR-210-5P inhibitor, which showed that hypoxia-induced downregulation of ATAD3A was abrogated by blocking miR-210-5P (Fig. 4m-o). Taken together, inhibition of ATAD3A under hypoxia is dependent on miR-210-5P.

MiR-210 is a classic hypoxia-induced miRNA that can be activated by HIF-1α. Here, we found miR-210-5P mimic induced mitophagy in HCC cells that were similar to CoCl₂ treatment. The hyper-activated mitophagy induced by CoCl₂ could be reversed by miR-210-5P inhibitor transfection (Fig. 5a). Furthermore, miR-210-5P mimic downregulated TOMM20 and TOMM70 expression (Fig. 5b), decreased mitochondrial mass by flow cytometry (Fig. 5c), upregulated PINK1 expression level (Fig. 5e), and increased lactate production (Supplementary Fig. S3e-h) in the Huh7 and LM3 cells. Nevertheless, these could be neutralized by the miR-210-5P inhibitor in CoCl₂-treated cells. As shown in Fig. 5d-h, depletion of miR-210-5P significantly increased the percentage of apoptotic cells in the presence of sorafenib, while the miR-210-5P mimic decreased apoptosis in the sorafenib-treated cells. The expression levels of the pro-apoptotic proteins differed accordingly in the miR-210-5P mimic/inhibitor-transfected cells (Fig. 5i). In conclusion, hypoxia-induced mitophagy is partially mediated by miR-210-5P.

To investigate the expression level of ATAD3A in HCC tissues and their paired normal tissues. We generated a tumor sample tissue array containing 135 hepatocellular carcinoma samples with survival information and 85 cases with paired normal tissues to determine ATAD3A expression by IHC. Results showed that ATAD3A expression in HCC patients was heterogeneous, ranging from completely absent to strongly expressed and the representative images of hepatocellular carcinoma samples with strong, moderate or low ATAD3A expression were shown in Fig. 6a. ATAD3A expression level was lower in cancerous liver tissues compared to the adjacent normal tissues (Fig. 6b-c), and lower expression of ATAD3A was associated with higher-grade HCC patients (Fig. 6d). 14% of hepatocellular carcinoma samples had strong ATAD3A expression, whereas percentages of moderate and no ATAD3A expression were 21% and 65%, respectively (Fig. 6e). Kaplan–Meier analysis showed that lower ATAD3A expression in HCC was correlated closely with poor overall survival (OS) of the patients (P = 0.0184; Fig. 6f). Western blot analysis also showed that ATAD3A had a significantly lower expression in 6 HCC tissues compared to their paired normal tissues (Fig. 6g). Taken together, these data indicated that ATAD3A is down regulated in HCCs and indicative of poor clinical outcome of HCC patients. To further determine the tumorigenicity of sh-ATAD3A and whether miR-210-5P antagomir could attenuate the tumor progression role of sh-ATAD3A under sorafenib treatment in vivo. Huh7-shcontrol and Huh7-shATAD3A cells that have been transfected with or without antagomir of miR-210-5p construct were subcutaneously implanted into the right flank of nude mice (Fig. 6h). Results showed that sh-ATAD3A in Huh7 cells significantly increased the tumor weight and tumor volume compared to the control group in the presence of sorafenib treatment, and miR-210-5P antagomir reduced the tumor size and tumor volume both in control and sh-ATAD3A groups (Fig. 6i-l). Collectively, these data provide evidence for an important inverse regulating relationship between ATAD3A and miR-210-5P in vitro and in vivo. Taken together, our findings point to an inverse regulatory relationship between ATAD3A and miR-210-5P in mediating mitophagy in the sorafenib-resistant HCC cells under hypoxia (Fig. 6m).