Tumor tissue specimens from patients were obtained in strict accordance with the regulations of the research ethics committee of Qingdao Municipal Hospital. In addition, all immunohistochemistry experiments were performed under the supervision of personnel of the research ethics committee. Animal experiment protocols were designed in strict accordance with the ARRIVE guidelines and reviewed and approved by the research ethics committee of Qingdao Municipal Hospital. All experimental procedures were recorded, reviewed, and completed under the supervision and guidance of the research ethics committee (ethics approval number: 104).

Immunohistochemistry was performed as previously described (19). Tumor and adjacent noncancerous (normal) liver tissues (60 pairs in total) were obtained from patients with liver cancer, fixed with 4.0% paraformaldehyde, paraffin-embedded, and cut into 4-μm-thick sections. The sections were treated with 3.0% hydrogen peroxide and blocked with 5.0% bovine serum, and then incubated with FNDC5 (1:200, ab181884; Abcam, Cambridge, MA, USA) and Nrf2 (1:200, 16396-1-AP, Proteintech, Beijing, China) antibodies overnight at 4°C. After washing with PBS, the sections were incubated with secondary antibodies conjugated with species-specific horseradish peroxidase (HRP) for 1 h at 25°C, and finally stained with diaminobenzidine (DAB) and hematoxylin. Images were taken using an AxioVision Rel.4.6 computerized image analysis system (Carl Zeiss). The degree of staining was assessed based on the staining area and intensity, using to the following formula: degree of staining = area × intensity. Staining area was scored as follows: 0 (staining area close to 0%), 1 (staining area less than 10%), 2 (staining area between 10% and 35%), 3 (staining area between 35% and 75%), and 4 (staining area between 76% and 100%). Staining intensity was scored as follows: 0 (no staining), 1 (weak staining), 3 (moderate staining), and 4 (strong staining). The final expression level of the target protein is expressed as the SI score, which was calculated using the following formula: SI = staining area score × staining intensity score, with SI ≥ 8 indicating high expression, whereas SI < 8 indicating low expression.

HepG-2 and Huh7 human HCC cell lines were provided by the Chinese Academy of Sciences Cell Bank (Shanghai, China). The cells were cultured in Dulbecco’s modified Eagle’s medium (DMEM; HyClone, USA) containing 10% fetal bovine serum (FBS; Excellbio, USA) and 1% penicillin–streptomycin (HyClone, USA). All cells were incubated at 37°C under 5% CO₂.

The HCC cells were incubated in 96-well plates with different concentrations of sorafenib, and the half-inhibitory concentration (IC₅₀) was determined. The HCC cells were then cultured in six-well plates at a density of 10⁴ cells per well and treated with sorafenib at concentrations slightly below the IC₅₀ value. The concentration of sorafenib was increased by 0.25 µmol/L every week until the maximum tolerated concentration of 10 µM was reached; at this point, the HepG2-SR- and Huh7-SR-resistant cell lines were obtained and cultured continuously in a medium with sorafenib at 1 µM to ensure the persistence of drug resistance.

The HCC cells were inoculated in 96-well plates at a density of 5 × 10³ cells/well and incubated for 12 h, and then treatment with different concentrations (1, 2, 4, 8, 16, 32, 64, and 128 µM) of sorafenib for 24 h. The medium was replaced with fresh medium; there were six replicates per group. The Cell Counting Kit-8 (CCK8) reagent (APExBIO, K1018, USA) was added into the wells according to the manufacturer’s instructions. After incubating the plates for 1 h in the dark, the absorbance of the samples was measured at 450 nm using a microplate reader (Bio-Tek, VT, USA). Cell survival was calculated using the following formula: Cell survival (%) = (test group OD − negative control value)/(control group OD − negative control OD) × 100. The IC₅₀ was calculated using GraphPad Prism 7.0 software.

The cells cultured in six-well plates were collected, washed twice with PBS, and resuspended using the binding buffer available in the Annexin-V-FITC/PI kit; then, 5 μL of Annexin V-FITC and 5 μL of PI were successively added. The cells were stained in the dark for 15 min, and then immediately analyzed using flow cytometry.

Cell proliferation was determined using the BeyoClick EdU-594 kit (Beyotime, Shanghai, China). The cultured HCC cells were washed with PBS, and then 10 µM EdU and fresh medium were added; the cells were incubated at 37°C for 2 h. Subsequently, the cells were washed twice with PBS, fixed with 4% paraformaldehyde for 15 min, washed with PBS, and finally stained with DAPI for 5 min before washing with PBS. Images were acquired under a fluorescence microscope (IX70; Olympus, Tokyo, Japan).

Human FNDC5 cDNA was constructed by Gene Chem (Shanghai, China), packaged in a lentiviral plasmid vector, and used for cell transfection. The experimental group overexpressing FNDC5 and negative control were named FNDC5-OE and FNDC5-NC2, respectively. siRNAs for knocking down FNDC5 were constructed by RIBOBIO with the following sequences (5’-3’): 1. GGAGGATACGGAGTACATA; 2. AGAAGATGGCCTCCAAGAA; 3. GCTTCATCCAGGAGGTGAA. The first and second siRNAs exhibiting better knockdown effects were selected, and plasmids were constructed and packaged into lentiviral plasmid vectors by Gene Chem, using the component sequence hU6-MCS-CBh-gcGFP-IRES-puromycin. The experimental groups with FNDC5 knockdown and negative control were named FNDC5-KD1, FNDC5-KD2, and FNDC5-NC1, respectively.

Following the culture of HepG-2 and Huh7 HCC cells and drug-resistant cell lines in six-well plates to reach 40% fusion, 10 µL of viral solution and 40 µL of infection enhancing solution were added, and the cells were transfected for 24 h. The complete medium was then replaced, and the culture was continued for 24 h. Finally, successfully transfected cells were screened using 1 µg/mL puromycin and used in the subsequent experiments.

The HCC cells were fixed with 2.5% glutaraldehyde for 1 h, washed three times with PBS, and fixed in 2% sodium tetroxide for 1 h. Thereafter, ethanol dehydration was performed, followed by sectioning and staining after embedding of the samples. Finally, the sections were observed using a Hitachi-7500 transmission electron microscope (Hitachi, Tokyo, Japan).

JC-1 is a fluorescent probe that detects the ΔΨm mitochondrial membrane potential. When mitochondrial membrane potential decreases, JC-1 is converted from red fluorescent polymers (J-aggregates) to green fluorescent monomers (J-monomers). The HCC cells were incubated with JC-1 (Beyotime, Shanghai, China) in the dark at 37°C for 20 min, and then washed twice with staining buffer before analysis using a flow cytometer (Accuri C6, BD Biosciences, USA).

The levels of reactive oxygen species (ROS) in HCC cells cultured in six-well plates were measured using the oxidation-sensitive fluorescent probe DCFH-DA (#D6883; Sigma) and analyzed using flow cytometry. After cell lysis, the concentration of malondialdehyde (MDA) was determined using the lipid peroxidation malondialdehyde assay kit (S0131S; Beyotime), according to the manufacturer’s instructions.

Total protein was extracted from HCC cells using the RIPA lysis buffer (Beyotime). Nuclei and cytoplasmic proteins were extracted using the nuclei and cytoplasmic protein extraction kit (Beyotime). Protein concentration was determined using a BCA protein quantification kit (Beyotime). The samples were then boiled for 10 min, followed by electrophoresis on 10% SDS-polyacrylamide gels, and the separated proteins were transferred on to polyvinylidene fluoride (PVDF) membranes (Millipore, Bedford, MA, USA). The membranes were then blocked with 5% skimmed milk and incubated at 25°C for 2 h. The membranes were then incubated with primary antibodies (1:1000, Proteintech) at 4°C overnight. Antibodies used included SLC7A11 (26864-1-AP), GPX4 (67763-1-Ig), DMT1 (20507-1-AP), PI3K (20584-1-AP), Akt (60203-2-Ig), p-Akt (66535-1-Ig), and Nrf2 (16396-1-AP), whereas β-actin (1:5000, #4967, Cell Signaling Technology, USA), and lamin B1 (1:2000, 12987-1-AP, Proteintech) were used as the internal references. The membranes were then washed thrice with TBST buffer and incubated with diluted HRP-coupled secondary antibodies for 2 h under 25°C. Finally, protein blots were detected using an ECL reagent (Millipore). Three independent replicate experiments were performed, and the intensity of protein blots was analyzed using ImageJ software.

NOD-SCID (NOD CB17-Prkdcscid/NcrCrl) 5-week-old male mice were purchased from Beijing Life River Laboratory Animal Technology Company (Beijing, China). The mice were housed in a 12-h light/dark cycling environment. The temperature was 25°C ± 1°C, humidity was maintained at 56%, and the mice were provided adequate food and water. Mice weighing between 20 and 23 g were selected and 5 × 10⁶ Hep-G2 cells were subcutaneously injected into their backs. The mice were subsequently divided into the following four groups: control (n = 5), FNDC5 overexpressing (n = 5), FNDC5 overexpressing followed by treatment with the PI3K inhibitor LY294002 (MCE, China), and FNDC5 knockdown (n = 5). Seven days after cell injection, sorafenib (30 mg/kg) was administered to all mice via intraperitoneal injection every alternate day for 4 weeks. The mice in the third group were intraperitoneally injected with LY294002 (25 mg/kg) diluted with DMSO twice a week for 4 weeks. Tumor volumes were measured every 4 days, and the mice were sacrificed on day 28 using the cervical dislocation method. Tumor tissues were dissected out and processed into homogenates, lysed, and centrifuged, and the expression level of the target proteins was detected using western blotting. The levels of MDA in tumor tissues were detected using the MDA kit.

GraphPad 7 (GraphPad Software Inc., CA, USA) was used for data analysis. All experimental results represent data from at least three independent replicate experiments, and the results are expressed as mean ± SD. A t-test was used to compare the differences between groups. A one-way ANOVA was used to compare differences among multiple experimental groups. Results with p value < 0.05 were considered statistically significant.

In our previous study, we found that FNDC5 was highly expressed in HCC tissues, promoting cancer progression (18). Our present immunohistochemical results also showed that among the 60 pairs of HCC and paraneoplastic tissues, the expression of FNDC5 was high in 48 HCC tissues, whereas its expression was low in 12 HCC and all paracancerous tissues ( Figure 1A ). To investigate whether the level of expression of FNDC5 was altered in sorafenib-resistant HCC cells, we treated HCC cell lines with increasing concentrations of sorafenib for 24 h. Following the CCK8 assay, we calculated the 24 h half-inhibitory concentration (IC₅₀) of sorafenib in HepG2 and Huh7 cell lines, and by gradually increasing the concentration of sorafenib, we established sorafenib-resistant cell lines. We also examined the IC₅₀ of the obtained resistant cell lines. We found that the IC₅₀ of sorafenib for HepG2-SR and Huh7-SR cells was 2–3 times higher than that for sorafenib-sensitive HCC cells ( Figure 1B ). Following treatment of HCC cells with different concentrations of sorafenib for 24 h, we found that the viability of drug-resistant cells was significantly higher than that of the sorafenib-sensitive HCC cells ( Figure 1B ). After confirming the successful development of sorafenib-resistant HCC cells, we examined the expression of FNDC5 in HCC sorafenib-resistant and sorafenib-sensitive HCC cells using western blotting. We found that the expression of FNDC5 was significantly higher in resistant cells than in sorafenib-sensitive HCC cells ( Figure 1C ).

To further investigate whether the resistance of HepG2-SR and Huh7-SR cells to sorafenib was caused by changes in the expression of FNDC5, we knocked down FNDC5 in HCC resistant cells ( Figure 2A ). The cells were then treated sorafenib-sensitive, drug-resistant, and FNDC5-knockdown drug-resistant cell lines using the same concentration of sorafenib, followed by staining with EdU and Annexin-V-FITC/PI to detect the inhibitory effect of sorafenib on each group ( Figures 2B , C ). We observed a higher rate of apoptosis in sorafenib-sensitive HCC cells, whereas drug-resistant HCC cells were obviously insensitive to sorafenib. However, we noticed that resistance to sorafenib was reversed after knocking down FNDC5 in drug-resistant HCC cells. We then examined the alterations in the level of ferroptosis in each group of HCC cells through the evaluation of the levels of ROS and MDA ( Figures 3A , B ), JC-1 detection of mitochondrial membrane potential ( Figure 3C ), electron microscopic observation of mitochondria ( Figure 3D ), and detection of the changes in ferroptosis markers in each subgroup using western blotting ( Figure 3E ). We found that the sorafenib-induced level of ferroptosis was lower in sorafenib-resistant cells with high expression of FNDC5 than in sorafenib-sensitive HCC cells, whereas the level of ferroptosis was elevated in FNDC5-knockdown sorafenib-resistant cells.

Based on previous findings, we hypothesized that FNDC5 affects the expression of Nrf2 in HCC tissues. After selecting HCC tissues with different levels of expression of FNDC5 via immunohistochemical detection, we found that the expression of Nrf2 was also elevated in HCC tissues with high expression of FNDC5 ( Figure 4A ). As the PI3K/Akt pathway regulates the expression of Nrf2, and as we previously found that FNDC5 affects the PI3K/Akt pathway, we evaluated the changes in the expression of PI3K, Akt, and Nrf2 in sorafenib-sensitive HCC, sorafenib-resistant, and FNDC5-knockdown resistant cell lines following treatment with the same concentration of sorafenib ( Figure 4B ). Our western blotting results showed that the levels of PI3K, pAkt, and Nrf2 in the nucleus were elevated in the drug-resistant cells compared with sorafenib-sensitive HCC, whereas the activation of the PI3K pathway was diminished and the level of Nrf2 in the nucleus was reduced after the knockdown of FNDC5. To further verify the effect of FNDC5 on the levels of Nrf2 via the PI3K/Akt pathway, we treated HCC sorafenib-resistant cells with a PI3K inhibitor for 1 h, and then added sorafenib to detect the levels of Nrf2 and ferroptosis marker proteins. Compared with those in the resistant cells not treated with the PI3K inhibitor, the level of Nrf2 in the nucleus of HCC sorafenib-resistant cells decreased, whereas the level of ferroptosis increased after the administration of the PI3K inhibitor ( Figure 5A ). We further confirmed that the addition of the PI3K inhibitor led to increased levels of ferroptosis in sorafenib-resistant cells as indicated by the levels of ROS and MDA and mitochondrial membrane potential ( Figures 5B –D ). Furthermore, we demonstrated that the PI3K inhibitor reversed the resistance of sorafenib-resistant cells as indicated by the flow cytometry detection of apoptotic rate ( Figure 5E ).

To further demonstrate that high expression of FNDC5 leads to the resistance of HCC cells to sorafenib, we overexpressed FNDC5 in sorafenib-sensitive HCC cell lines ( Figure 6A ). Annexin-V-FITC/PI staining indicated that the overexpression of FNDC5 in HCC cells led to increased resistance to sorafenib ( Figure 6B ). Furthermore, analyses of the levels of ROS and MDA and mitochondrial membrane potential further confirmed the level of ferroptosis was decreased in sorafenib-sensitive HCC cells with overexpression of FNDC5 compared with sorafenib-sensitive HCC cells ( Figures 6C , D ). Finally, western blotting showed that after the overexpression of FNDC5, the PI3K/Akt pathway was activated, the level of Nrf2 in the nucleus was increased, the levels of GPX4 and SLC7721 were increased, whereas the levels of DMT1 and ferroptosis were reduced ( Figure 6E ).

We subcutaneously injected HCC cells with different levels of expression of FNDC5 into mice to further investigate whether the changes in the levels of expression of FNDC5 in HCC cells continued to affect the efficacy of sorafenib in vivo ( Figures 7A –C ). At the same concentration of sorafenib, HCC cells with high expression of FNDC5 exhibited an increased rate of proliferation under the skin of mice and low levels of ferroptosis, whereas FNDC5-knockdown HCC cells grew slowly and presented elevated levels of ferroptosis as indicated by western blotting and MDA assays. In addition, we found that treatment with a combination of sorafenib and PI3K inhibitors slowed the growth of HCC cells with high expression of FNDC5 and increased the level of ferroptosis ( Figures 7D , E ).