ABCC5 facilitates the acquired resistance of sorafenib through the inhibition of SLC7A11-induced ferroptosis in hepatocellular carcinoma

Antibodies and reagents

Antibodies against ABCC1 (A11153), ABCC3 (A9849), ABCC5 (A3028), SLC7A11 (A2413, A13685), NRF2 (A11159) and GPx4 (A11243) were obtained from Abclonal (Wuhan, China). PI3K (60225), p-PI3K (55311), AKT (10176), p-AKT (66444) and β-tubulin (66240) were bought from Proteintech (Wuhan, China). β-actin (4970L) was obtained from Cell Signal Technology (Danvers, MA). MK571 (S8126), Sorafenib (S7397), SB216763 (S1075) were purchased from Selleck Chemicals (Houston, USA). Deferoxamine (DFO, 1166003), Erastin (E7781) were obtained from Sigma-Aldrich (Missouri, USA). Brusatol (HY-19543), LY294002 (HY-10108), SC79 (HY-18749), Ferrostatin-1 (Fer-1, HY-100579), RSL3 (HY-100218A), ML385 (HY-100523), MG132 and Chloroquine (CQ) were obtained from MedChemExpress (New Jersey, USA).

Cell lines and treatments

Human hepatocellular carcinoma cell lines HuH7, HepG2 and Sk-Hep-1 were obtained from the ATCC (Manassas, USA). HuH7 cells are well characterized and commonly used for subcutaneous transplantation into nude mice. All cell lines were maintained in Dulbecco's modified Eagle's medium (DMEM: Selleck Chemicals, Houston, USA) supplemented with 10% foetal bovine serum (FBS: Gibco, Carlsbad, USA) at 37 °C in a humidified incubator containing 5% CO₂.

In terms of treatment with inhibitors or drugs, sorafenib (3 μM or 10 μM), deferoxamine (100 μM), ferrostatin-1 (1 μM), SB216763 (10 μM), SC79 (5 μg/ml), LY294002 (10 μM), ML385 (5 μM), brusatol (50 nM), MK571 (10 μM), erastin (10 μM), RSL3 (3 μM) , MG132 (10 mmol/L) and Chloroquine (10 mmol/L) were first dissolved in DMSO and then added to the cultured HuH7 or Sk-Hep-1 cells or HuH7-SR cells.

Generation of sorafenib-resistant cells

We subjected two HCC cell lines to long-term exposure to sorafenib at a dose of 0.5 μM and escalating through incremental doses to 10 μM, the highest clinically achievable concentration in which the cells survived, thereby establishing HuH7 and Sk-Hep-1 sorafenib-resistant cells (HuH7-SR and Sk-Hep-1-SR). The control cells were named HuH7-WT and Sk-Hep-1-WT. We measured sorafenib sensitivity by CCK-8 assays (24 h), and the results displayed that the sorafenib-resistant cells had lower sorafenib sensitivity than did the WT control cells.

Determination of sorafenib and its metabolites by UPLC8-MS/MS

The UPLC conditions: system, Waters Acquity^TM; column, Acquity UPLC HSS T3 column (100 × 2.1 mm, 1.8 μm, Waters,Milford, MA, USA); mobile phase A: 0.1% (v/v) formic acid in water; mobile phase B: 100% acetonitrile; gradient, 0 min to 4 min at 50 to 70% B, 4 min to 5 min at 70 to 50% B; flow rate, 0.3 mL/min; column temperature, 50 °C; and injection volume, 10 μL. The MS/MS detector was a quadrupole tandem mass spectrometer (Waters, USA). Samples were analyzed using electrospray ionization in the positive model. The main working parameters were set as follows: capillary voltage, 3 kV; cone voltage, 38 V; collision voltage, 35 V; source temperature, 120 °C; desolvation temperature, 400 °C; desolvation gas flow, 600 L/H; cone gas flow, 50 L/H; and collision gas glow, 0.20 mL/min. Data were collected and analyzed by Waters Quantify software (Masslynx 4.1, Waters, USA). Sorafenib was monitored at m/z 465.3>252.4. Metabolites including M3, M4 and M5 were monitored at m/z 481.3>268.3, m/z 451.0>406.3 and m/z 467.0>202.0 transitions, respectively. Testosterone was used as an internal standard (IS) and was monitored at m/z 289.5>97.4.

Flow cytometry assay

Changes in intracellular lipid ROS levels were measured by the oxidative conversion of cell permeable 2′,7′-dichlorofluorescein diacetate (DCFH-DA, Sigma) with a fluorescence spectrophotometer. Cells were plated in 6 cm dishes (1 × 10⁶ cells/well) and allowed to attach overnight. The cells were incubated with control medium with or without sorafenib for 24 h. The cells were washed with D-Hank's solution and incubated with DCFH-DA at 37 °C for 20 min. Then, the DCF fluorescence distribution of 20,000 cells was determined by a FACSCalibur flow cytometer (Beckman Coulter, CytoFLEX S, United States) at an excitation wavelength of 488 nm and at an emission wavelength of 535 nm [14].

The mitochondrial membrane potential (MMP) assay is a cytofluorimetric method that is both qualitative and quantitative, and the results are validated by analysing the MMP at the level of a single mitochondrion. JC-1 in an aggregated form emits a red or orange emission (590 ± 17.5 nm) in the matrix of the mitochondria, indicating a normal MMP; with the loss of MMP, JC-1 is converted to the monomeric form and emits green fluorescence with an emission of 530 ± 15 nm. Therefore, a decrease in the red/green fluorescence intensity ratio indicates a reduction in MMP [15].

Construction of lentiviral vectors

To generate cells with ABCC5 knockdown and overexpression, lentiviral vectors were amplified with the insert (full-length human ABCC5; NR_135125.2) by qRT-PCR, compared to human reference cDNA. Lentiviruses were produced by transient transfection of 293T cells with pSPAX2, pMD2G, and pHB-U6-MCS-zsgreen-puro (empty) (Hanbio, Shanghai, China) plasmid DNA (2439 -BamHI and 2456 -EcoRI sites) plus LipoFiter (Hanbio Biotechnology, ShangHai, China), following the manufacturer's protocol.

Quantitative real time polymerase chain reaction

Total RNA isolation and quantitative RT-PCR (qRT-PCR) were carried out according to the manufacturer's protocols. Briefly, first-strand cDNA synthesis was carried out by using a Reverse Transcription System Kit (#11801-025, OriGene Technologies, Rockville, MD, USA). cDNA from various cell samples was amplified with specific primers (Table S2).

Statistical analysis

Data were analyzed using SPSS version 20.0 software (SPSS; Chicago, USA). Student's t-tests and one-way ANOVAs were carried out on the western blot analysis. CCK-8 analyzes were applied to generate the cell growth curve and calculate cell viability. The significance of the correlation between the expression of ABCC5 and various histopathological factors was determined using Pearson's chi square (χ2) test. Multivariate analysis was performed by applying the Cox proportional hazards test. Statistical significance was established at P < 0.05.

Sorafenib induces ABCC5 expression in human HCC cells

To gain a insight in the possible mechanisms of sorafenib resistanc, a heatmap of differentially expressed genes between sorafenib-resistant HCC cells and the untreated HCC cells was conducted. Compared with the untreated HCC cells, ATP-binding cassette subfamily C (ABCC) members were up-regulated (P < 0.05) in sorafenib-resistant HCC cells (GSE96793; Fig. 1A). To determine whether sorafenib induces ABCC expression, HuH7 (a human HCC cell line) cells were treated with sorafenib for 24 h and the mRNA levels of ABCC family were analyzed via qRT-PCR. As shown in Fig. 1B, the expression levels of ABCC1, ABCC3 and ABCC5 were the highest following sorafenib treatment. In addition, the Western blot analysis revealed that compared with ABCC1 or ABCC3, the protein expression of ABCC5 was significantly increased in HuH7 cells treated with sorafenib (Fig. 1C). Moreover, qRT-PCR and Western blot analysis demonstrated that sorafenib-induced ABCC5 expression was observed in other human HCC cell lines (HepG2 and Sk-Hep-1 cells) (Fig. 1D, E). Intermittent drug induction method was conducted to induce resistance to sorafenib in HuH7 and Sk-Hep-1 cells, named HuH7-SR and Sk-Hep-1-SR, respectively. Cells were cultured in a complete medium containing sorafenib (Fig. 1F). The IC50 values of the two resistant cell lines were calculated as 9.804 ± 0.995 μM and 9.564 ± 1.152 μM, which were significantly higher than HuH7-WT cells (4.851 ± 0.293 μM) and Sk-Hep-1-WT (3.732 ± 0.262 μM) (Fig. 1G). To further verify the reliability of sorafenib-resistant cell lines, flow cytometry assay was performed to detect cell apoptosis. The results displayed that compared with WT cells, SR cells showed significant drug resistance (Fig. 1H). As indicated by qRT-PCR and Western blot analysis, the expression of ABCC5 in both SR cells was higher than that in WT cells (Fig. 1I, J). In vivo, HuH7-WT and HuH7-SR cells were implanted into the subcutaneous space of the left and right flanks of nude mice. Beginning at day seven, these mice were treated with sorafenib. Compared with the control HuH7-WT group, sorafenib treatment did not reduce the size or the weight of tumors in the HuH7-SR group (Fig. 1K, L). Moreover, the histological observation of H&E and immunohistochemical staining indicated that the expression of ABCC5 in the HuH7-SR group was significantly higher than that of the HuH7-WT group (Fig. 1M). These findings indicate that sorafenib remarkably induces ABCC5 expression in human HCC cells.

PI3K/AKT/NRF2 pathway activation is required for sorafenib-induced ABCC5 expression

Long-term treatment with sorafenib can activate the phosphatidylinositol 3-kinase (PI3K)/AKT pathway in HCC [16,17]. Consistently, our data revealed that sorafenib specifically induced PI3K/AKT pathway activation in HuH7 cells and Sk-Hep-1 cells (Fig. 2A). The nuclear factor erythroid 2-related factor 2 (NRF2), a downstream target of the PI3K/AKT pathway, can protect cells against multiple injuries and mediate antioxidation [18]. Additionally, it's reported that sorafenib can regulate the activity of NRF2 to affect the death and survival of HCC cells [19,20]. To determine if short-term sorafenib treatment could activate PI3K/AKT or increase NRF2 expression in HuH7 cells, the levels of them were recovered when the drug was withdrawn (Fig. 2B). The Western blot analysis showed that NRF2 and ABCC5 were significantly up-regulated by the PI3K/AKT pathway agonists SC79 and SB216763. After treated with the PI3K/AKT pathway inhibitor LY294002, the expression of NRF2 was reduced, indicating that the AKT pathway positively regulated NRF2 in HuH7 cells (Fig. 2C). Furthermore, the Western blot analysis revealed that compared with HuH7-WT cells, the expression of p-PI3K, p-AKT, and NRF2 was increased in HuH7-SR cells. However, LY294002 treatment decreased the expression of NRF2 in HuH7-SR cells, suggesting that the AKT pathway positively regulated NRF2 in sorafenib-resistant HCC cells (Fig. 2D, E). Consistently, immunofluorescence staining showed that the fluorescence intensity of NRF2 was enhanced in sorafenib-resistant HCC cells (HuH7-SR), especially in the nucleus, but LY294002 significantly decreased the expression of NRF2, further indicating that long-term sorafenib exposure activated the PI3K/AKT/NRF2 pathway (Fig. 2F). qRT-PCR analysis and bioinformatics data (GEPIA2 websites) showed that the correlation between ABCC5 and NRF2 was significant in HCC (Fig. S1). Moreover, ABCC5 expression was significantly reduced in WT and SR cells after treated with target-specific siRNAs against NRF2 (Fig. 2G, H) or NRF2 inhibitors (ML385 and Brusatol) (Fig. 2I, J), suggesting that NRF2 was required for the sorafenib-induced ABCC5 expression. The inhibitors of the AKT/NRF2 pathway as previously mentioned significantly enhances sorafenib-induced cell death at 24 h as shown by cell viability assay (Fig. 2K). Besides, the expression of p-PI3K, NRF2 and ABCC5 were evaluated by immunohistochemistry (IHC) in the hepatocellular carcinoma tissue of each group. Immunohistochemical staining showed that compared with the control HuH7-WT group, the p-PI3K, NRF2 and ABCC5 expression levels were significantly increased in HuH7-SR (Fig. 2L). Taken together, these data indicate that activation of PI3K/AKT/NRF2 pathway is required for sorafenib-induced ABCC5 expression.

ABCC5 mediates acquired resistance to sorafenib in vitro

To determine whether ABCC5 expression regulates the anti-cancer activity of sorafenib, sgRNA targeting ABCC5 and a lentivirus vector of ABCC5 were transfected into Sk-Hep-1 and HuH7 cells (Fig. 3A, B). First, to evaluate the effect of ABCC5 on the transport of sorafenib in cells, HuH7-SR and ABCC5-overexpressing HuH7 cells were treated with sorafenib and detected by LC-MS. We found that overexpression of ABCC5 had no significant effect on the intercellular concentration of sorafenib or the intracellular concentration of its metabolites M3, M4 and M5 in HuH7 cells (P > 0.05) (Fig.3C). However, intercullar sorafenib and its metabolites were increased in drug-resistant strain HuH-SR compared with HuH-WT (Fig.3D). The data suggested that the efflux effect of ABCC5 on sorafenib was not the main mechanism causing sorafenib resistance. Overexpression of ABCC5 in HuH7 cells significantly inhibited sorafenib-induced cytotoxicity, which was effectively recovered by the ABCC inhibitor MK571. Additionally, knockdown and inhibition of ABCC5 expression can further intensify sorafenib-induced cytotoxicity in Sk-Hep-1cells (Fig. 3E). Furthermore, sgRNA targeting ABCC5 was used to neutralize the increased expression of ABCC5 in HuH7-SR cells (Fig. 3F). Colony formation assays also indicated that the suppression of ABCC5 expression inhibited long-term HuH7-SR cell proliferation following sorafenib treatment (Fig. 3G). Thus, these findings suggest that ABCC5 mediates acquired resistance to sorafenib in HCC cells in vitro.

ABCC5 is a negative regulator of ferroptosis and contributes to sorafenib resistance

Sorafenib is a strong inducer of ferroptosis but not of apoptosis of HCC cells [21]. Sorafenib induced lipid peroxidation accumulation and mitochondrial damage (Fig. S2A, B) and suppressed the expression of SLC7A11 and GPx4 (Fig. S2C) in HCC cells. Ferroptosis inhibitors Fer-1 and DFO significantly restored HCC cell viability inhibited by sorafenib (Fig. S2D). Ferroptosis is a new type of regulated cell death, which is mainly mediated by iron-dependent lipid peroxidation [22]. Thus, we further investigated whether ABCC5 is involved in the regulation of lipid peroxidation and mitochondrial damage contributing to sorafenib resistance. Overexpression of ABCC5 significantly offset the lipid peroxidation accumulation and the mitochondrial membrane potential (MMP) reduction induced by sorafenib in HuH7 cells. Consistently, ferroptosis inhibitors Fer-1 and DFO partially ameliorated lipid peroxidation and mitochondrial damage in HuH7 cells treated with sorafenib (Fig. 4A). Moreover, suppression of ABCC5 expression by sgRNA attenuated lipid peroxidation and enhanced MMP in Sk-Hep-1 cells and HuH7-SR cells treated with sorafenib. As expected, ferroptosis inhibitors Fer-1 and DFO partially attenuated lipid peroxidation and mitochondrial damage in HuH7-SR cells and ABCC5-knockdown Sk-Hep-1 cells treated with sorafenib (Fig. 4B, C). In summary, these results suggest that ABCC5 negatively regulates ferroptosis via suppressing lipid peroxidation and increasing MMP, which contributes to sorafenib resistance in HCC cells.

ABCC5 negatively regulates ferroptosis by stabilizing SLC7A11

Sorafenib increases intracellular lipid peroxidation levels by inhibiting SLC7A11, a specific light-chain subunit of the cystine/glutamate antiporter [23]. Next, we explored whether ABCC5 could interact with SCL7A11 to regulate ferroptosis. The bioinformatics analysis based on the starBase database showed that ABCC5 was significantly correlated with SLC7A11 (Fig. S3). IF assays showed that ABCC5 colocalized with SLC7A11, and knockdown of ABCC5 decreased the expression of SLC7A11 in HuH7, Sk-Hep-1 and HuH7-SR cells (Fig. 4D). Furthermore, the interaction of ABCC5 with SLC7A11 was detected by co-IP assays in WT and SR cells (Fig. 4E). Western blot analysis revealed that SLC7A11 was highly expressed in sorafenib-resistant cells (Fig. 5A). In addition, knockdown of ABCC5 suppressed SLC7A11 protein expression in HuH7-SR cells (Fig.5B). Conversely, the suppression or overexpression of SLC7A11 decreased or increased the expression of ABCC5 (Fig. 5C). In addition, we used the autophagy inhibitor chloroquine (CQ) and the proteasome degradation inhibitor MG132 to confirm that ABCC5 protected SLC7A11 from being degraded by autophagy (Fig. S4). NRF2 negatively regulates ferroptosis by limiting ROS production and reducing cellular iron uptake [24]. Analysis using the STRING database indicated that NRF2 may interact with ABCC5 and SLC7A11, possibly as the intermediary (Fig. S5). Treatment with sorafenib in combination with ML385, brusatol, or MK571, but not with the SLC7A11 inhibitor erastin [25], reversed the overexpression of ABCC5 and SLC7A11 induced by sorafenib in HuH7-SR cells (Fig. 5D,E). Similarly, knockdown of ABCC5 increased GSH depletion, mitigated lipid peroxidation and GPx4 expression in HuH7-SR cells treated with sorafenib and/or erastin (Fig. 5F,G). In addition, intracellular cystine content indicated that ABCC5 could further enhance the uptake of Cystine by SLC7A11 protein and generation of GSH (Fig. S6). To test whether sonifenib resistance is mediated through NRF2/ABCC5/SLC7A11 axis, the expression of ABCC5 and ferroptosis-related proteins in HuH7-SR cells were examined by Western blot assays. Our data demonstrated that NRF2/ABCC5/SLC7A11 axis was inhibited in HuH7-SR cells treated with inhibitors LY294002, Brusatol, MK571, Erastin or RSL3, respectively (Fig. 5H). Thus, after being transcriptionally activated by NRF2, ABCC5 interacts with SCL7A11 to negatively regulate ferroptosis.

Knockdown of ABCC5 enhances the anti-cancer activity of sorafenib in resistant HCC cells in vivo

To determine whether inhibition of ABCC5 expression led to the resistance to sorafenib in vivo, the control Huh7-SR cells (i) and ABCC5-knockdown Huh7-SR cells (ii and iii) were implanted into the subcutaneous space of the right flank of nude mice in different groups. Beginning at day seven, the three groups were treated with sorafenib, and the iii group was also treated with erastin. Compared with the control group, sorafenib treatment effectively reduced the size and weight of tumors formed by ABCC5-knockdown cells in the ii group. Moreover, compared with the ii group, both the size and weight of tumors were reduced in iii group, which was additionally treated with erastin (Fig. 6A, B, C). IHC assays showed decreased expression of SLC7A11 and GPx4 in tumors of the ii and iii group mice (Fig. 6D).

Clinically, the TCGA analysis showed that ABCC5 expression was higher in HCC tissues than in normal liver tissues (Fig. S7, P < 0.05). Consistently, IHC assays confirmed higher ABCC5 expression in HCC tissues compared to normal liver tissues (Fig. 6E, P < 0.05). In addition, IHC analysis indicated that NRF2, ABCC5 and SLC7A11 expression were significantly correlated in HCC tissues (Fig. 6F, P < 0.05). Furthermore, Kaplan–Meier survival curves based on the TCGA data demonstrated that the OS of HCC patients with low ABCC5 expression was better than that of HCC patients with high ABCC5 expression (Fig. S8).

Ethics approval and consent to participate

All experiments performed are endorsed by the Ethics Committee of Southern Medical University and complied with the Declaration of Helsinki. All animal experiments were carried out with the approval of the Southern Medical University Animal Care and Use Committee in accordance with the guidelines for the ethical treatment of animals. All animal experiments involved ethical and humane treatment under a license from the Guangdong Provincial Bureau of Science.

Consent for publication

No informed consent was required because data were analyzed anonymously.

Availability of data and material

Not applicable.