AKR1C3 Binds β‐Trcp to Promote the Degradation of TFRC to Protect Hepatocellular Carcinoma From Ferroptosis Authors: Lei Qi, Jingyi Hua, Di Pan, Wanwan Yang, Geng Tian, Fangyu Ye, Lingxiang Liu, Yuhan Mao, Qinglong Guo, Haopeng Sun, Li Zhao Published: 2026 ABSTRACT ABSTRACTFerroptosis, an iron‐dependent form of programmed cell death driven by lipid peroxidation, represents a new potential therapeutic target in cancer. However, emerging evidence indicates that hepatocellular carcinoma (HCC) frequently exhibits resistance to ferroptosis induction, while the underlying molecular mechanism is poorly understood. Here, we found that aldo‐keto reductase family 1 member C3 (AKR1C3), a protein highly expressed in ferroptosis‐resistant HCC cells, negatively regulates ferroptosis in an enzyme‐independent manner. Mechanistically, AKR1C3 promotes ubiquitin‐proteasomal degradation of the transferrin receptor (TFRC), which is indispensable for cellular iron uptake. AKR1C3 knockdown restores TFRC expression, increases the level of labile iron pool, and sensitizes HCC cells to ferroptosis. Furthermore, AKR1C3 acts as a scaffolding protein to promote the degradation of TFRC and reduce iron uptake by promoting nuclear export of Beta‐transducin repeats‐containing proteins (β‐TrCP) and its binding to TFRC. Notably, AKR1C3 is upregulated in NRF2‐driven sorafenib‐resistant HCC, and its inhibition reversed ferroptosis and sorafenib resistance. Our work uncovers AKR1C3 suppresses ferroptosis in HCC by promoting β‐TrCP‐mediated TFRC degradation, positioning AKR1C3 as a promising therapeutic target to enhance ferroptosis‐based anticancer strategies.Keywords: Aldo‐keto reductase, ferroptosis, hepatocellular carcinoma, sorafenib resistance, ubiquitination Abbreviations Aldo‐keto reductase family 1 member C3 Cancer Cell Line Encyclopedia Cyclohexane hepatocellular carcinoma immunoprecipitation‐mass spectrometry Lipocalin 2 Leukemia inhibitory factor receptor labile iron pool V‐maf musculoaponeurotic fibrosarcoma oncogene homolog G Nuclear factor kappa‐B Nuclear factor (erythroid‐derived 2)‐like 2 Patient‐derived xenograft Solute carrier family 7 member 11 Transferrin receptor protein Yes‐associated protein Beta‐transducin repeats‐containing proteins Introduction Hepatocellular carcinoma (HCC), the predominant form of primary liver cancer, is often diagnosed at an advanced stage [1], where drug resistance severely limits the efficacy of first‐line therapies and contributes to poor prognosis [2]. Thus, elucidating drug resistance mechanisms and developing novel therapies are critical to reducing HCC mortality. Ferroptosis, an iron‐dependent form of regulated cell death driven by lipid peroxidation [3], has emerged as a potential therapeutic vulnerability in cancers [4]. In HCC, ferroptosis is often weakened due to multiple signaling pathways activation, like the Hippo‐YAP signaling pathway [5], LIFR‐NF‐κB‐LCN2 axis [6], and the Keap1‐NRF2 system [7]. Notably, the transferrin receptor (TFRC), a key iron transporter, is a critical positive regulator of ferroptosis [8]. Hence, a deeper understanding of the mechanisms regulating TFRC involved in ferroptosis represents a promising avenue for disclosing the new strategies in HCC. The aldo‐keto reductase AKR1C3 is implicated in HCC progression [9, 10]. Beyond its enzymatic role, we previously reported that AKR1C3 promotes HCC cell proliferation through non‐catalytic functions [11]. However, to date, the exact role and underlying mechanism of AKR1C3 in remodeling ferroptosis susceptibility remain to be elucidated. Here we uncovered an unexpected role of AKR1C3 as a scaffolding protein in remodeling ferroptosis susceptibility in HCC cells. In the present study, we reveal that AKR1C3‐β‐TrCP‐TFRC axis as a pivotal central pathway that governs resistance to ferroptosis in HCC. AKR1C3 functions as an adaptor protein, decreases the level of labile iron pool (LIP) via β‐TrCP‐mediated TFRC ubiquitination and destabilization, whereas genetics and pharmacological suppression of AKR1C3 restores TFRC protein levels and sensitizes HCC cells to ferroptosis. Moreover, we found the critical role of AKR1C3 in HCC sorafenib resistance using clinical data and patient‐derived xenograft (PDX) sorafenib resistance modeling. Loss of AKR1C3 restores TFRC‐induced ferroptosis and renders cells sensitive to sorafenib. Our study provides a deeper insight into the mechanism of LIP‐mediated ferroptosis in HCC and suggest that targeting AKR1C3 may represent a viable approach to overcome resistance of ferroptosis‐targeted therapy in HCC. Material and Methods Patients and Clinical Specimens All human tissue studies were performed with approval from the Institutional Review Board of Nanjing Drum Tower Hospital and in accordance with the Declaration of Helsinki. This study utilized 17 pairs of paraffin‐embedded HCC and adjacent paracancerous tissues, as well as nine recurrent tumor samples from patients treated with sorafenib, all obtained from Nanjing Drum Tower Hospital. The patients' information is presented in Table S1. Informed consent was obtained from all participants. Lentivirus‐Mediated Overexpression and Knockdown Lentivirus Infections The lentivirus infections were performed as previously reported for stable overexpression and shRNA‐mediated knockdown. The target sequences are listed in Table S2. Lentivirus was generated by transfecting HEK‐293 T cells in six‐well plates with a Lentiviral packaging Kit (Yeasen Biotech, 41102ES10). HCC cells were infected with recombinant lentivirus transduction units in the presence of 6 μg/mL polybrene (Sigma‐Aldrich, TR‐1003), and the lentivirus‐infected cells were further selected by 1 μg/mL puromycin (Solarbio Life Science, P8230) or 1 μg/mL Blasticidin S (Solarbio Life Science, B9300) for 3 days. Quantitative Real‐Time PCR (qRT‐PCR) Total RNA was extracted from cells using Trizol reagent (Vazyme, R711‐01) according to the manufacturer's instructions. The quantification and quality analysis of RNA was performed on NanoDrop OneC spectrophotometer (Thermo Fisher Scientific, 840–317,500). Reverse transcription was conducted using SuperScript IV Reverse Transcriptase (Vazyme, R211). qPCR reaction was performed using Hieff qPCR SYBR Green Master Mix (Yeasen Biotech, 11200ES10). The primers were synthesized from Sango Biotech, and the sequence information was listed in Table S3. The fold‐changes were calculated using the 2−△△Ct method, with beta‐actin serving as a reference gene. Statistical Analysis All experiments were independently repeated at least three times. For IHC analysis, a minimum sample size of four was used. Statistical analyses were performed using GraphPad Prism 8 software. Outliers were identified using the ROUT test. All data were normally distributed as confirmed by the Anderson‐Darling test; therefore, parametric tests were applied. Comparisons were made using Student's paired or unpaired t‐test (for single comparisons) or one‐way ANOVA (for multiple comparisons). Kaplan–Meier survival curves were analyzed with the Mantel‐Cox test. Post hoc tests were conducted only when the ANOVA F‐value was statistically significant (p < 0.05) and variance homogeneity was not violated. A p < 0.05 was considered statistically significant (*p < 0.05, **p < 0.01, ***p < 0.001). Cell lines and cell culture, reagents, mycoplasma detection, animal experiments, cell viability, lactate dehydrogenase release assay, western blot, mass spectrometry analysis, immunohistochemistry, colony formation assay, EdU staining, measurement of Malondialdehyde and Glutathione Content, measurement of LIP and ROS Levels, transmission electron microscopy, co‐immunoprecipitation assay, confocal immunofluorescence analysis and public data availability. The above experiments and detailed information were detailed in the Supporting Information and Methods Doc S1. The relevant antibody sources and dilution ratios are shown in Table S4. Results AKR1C3 Is Positively Associated With Ferroptosis Resistance in HCC Emerging evidence indicates that HCC exhibits intrinsic resistance to ferroptosis [12]. To investigate mechanisms underlying the intrinsic ferroptosis resistance of HCC, we conducted a multi‐omics analysis. Transcriptomic data identified 772 differentially expressed genes in HCC versus normal tissue (Figure 1A). Cross‐referencing these with a ferroptosis regulator database revealed 22 ferroptosis‐associated DEGs (Figure 1B). Subsequent validation across multiple HCC datasets consistently highlighted AKR1C3, AURKA, and FOXM1 as significantly overexpressed (Figure 1C). Pharmacogenomic analysis further identified AKR1C3 as the sole gene whose expression correlated significantly with resistance to four classic ferroptosis inducers (RSL3, Erastin, ML162, ML210) (Figure 1D). Complementary validation through ROC Plotter demonstrated elevated AKR1C3 expression in HCC cells unresponsive to the above four ferroptosis inducers (Figure 1E). Experimentally, we assessed basal AKR1C3 protein levels and ferroptosis sensitivity across eight HCC cell lines. Cells with high AKR1C3 expression exhibited significantly higher Erastin IC50 values (indicating greater resistance) compared to low‐expressing cells (Figure 1F,G). Multivariate regression confirmed a strong negative correlation between AKR1C3 protein levels and Erastin sensitivity (R2 = 0.661, p = 0.026) (Figure 1H). In summary, these integrated bioinformatic and experimental data establish a strong association between high AKR1C3 expression and ferroptosis resistance in HCC, prompting our focused investigation into its functional role and underlying mechanisms. AKR1C3 is positively associated with ferroptosis resistance in HCC. (A) Analysis of HCCDB reveals 772 differentially expressed genes in HCC. (B) Venn diagram analysis reveals 22 ferroptosis‐related differentially expressed genes in HCC. (C) Analysis of the HCCDB database revealed a clear trend of overexpression of AKR1C3, AURKA and FOXM1 in HCC. (D) Analysis of CTRP database reveals the correlation of the expression levels of AKR1C3 and ferroptosis sensitivity. Plotted values are z‐scored Pearson's correlation coefficients. Line, median; box, 10th–90th percentiles. (E) Analysis by ROC Plotter showed the expression level of AKR1C3 in 4 ferroptosis inducers non‐responsive cells. (F) Western blot analysis of AKR1C3 expression in a panel of HCC cell lines. (G–H) The ferroptotic sensitivity of HCC cell lines to various Erastin concentrations and its correlation with AKR1C3 expression, analyzed by Spearman's correlation. n = 3 biological replicates. Statistical analysis was done with two‐tailed Student's t test. AKR1C3 Inhibited Ferroptosis Independently of Enzyme Activity To study AKR1C3's role in ferroptosis resistance in hepatocellular carcinoma (HCC), we knocked down AKR1C3 in HCC‐LM3 and MHCC‐97 L cells, which have high endogenous AKR1C3 expression (Figure 2A). AKR1C3 loss sensitized cells to Erastin‐induced ferroptosis, an effect reversed by ferroptosis inhibitor ferrostatin‐1 (Figures 2B and S1B,C). Re‐expression of wild‐type AKR1C3 restored ferroptosis resistance (Figures 2B and S1B,C). Given that AKR1C3 typically functions as an aldo‐keto reductase with both physiological and tumor‐promoting roles, we generated a catalytically inactive AKR1C3 mutant to examine whether its enzymatic activity is required for ferroptosis resistance. Such catalytic mutation was achieved by mutating Tyr55 and His117 near NADP+ to Ala (Figure S1A). AKR1C3 inhibited ferroptosis independently of enzyme activity. (A) Western blot analysis of MHCC‐97 L and HCC‐LM3 cells knocking down AKR1C3 (shA1C3) and ectopic expression of wild type AKR1C3 (OEA1C3) or mutant AKR1C3 (OEA1C3▲) after knockdown in HCC‐LM3 cells. (B) Cell death of MHCC‐97 L or HCC‐LM3 cells derived from A treated with Erastin and fer‐1. (C) Western blot analysis of ectopic expression of AKR1C3 in SK‐Hep‐1 cells. (D) Cell death of SK‐Hep‐1/OEAKR1C3 cells treated with Erastin and fer‐1. (E–F) Intracellular MDA and GSH content of MHCC‐97 L, HCC‐LM3 or SK‐Hep‐1 cells derived from A or C after treatment with Erastin. (G–I) Intracellular Reactive oxygen species (ROS) and Fe2+ content of MHCC‐97 L, HCC‐LM3 or SK‐Hep‐1 cells derived from A or C after treatment with Erastin. Scale bars, 100 μm(G), 50 μm(H). (J) Cellular mitochondrial morphology of MHCC‐97L or SK‐Hep‐1 cells after treatment with Erastin by transmission electron microscopy. Scale bars, 1 or 2 μm. n = 3 biological replicates. *p < 0.05; **p < 0.01; ***p < 0.001 and n.s. p > 0.05 between indicated groups. Interestingly, a catalytically inactive AKR1C3 mutant similarly restored resistance (Figures 2B and S1B,C), indicating a reductase‐independent mechanism. Since SK‐Hep‐1 cells lack detectable endogenous AKR1C3 expression (Figure 2C). In SK‐Hep‐1 cells, which less endogenous AKR1C3, ectopic AKR1C3 expression also conferred ferroptosis resistance (Figures 2D and S1B,C). AKR1C3 potently suppresses ferroptosis in HCC cells. Knockdown of AKR1C3 increased levels of the lipid peroxidation marker MDA and depleted glutathione (GSH) upon Erastin treatment, effects reversed by re‐expressing either wild‐type or catalytically inactive AKR1C3. Conversely, AKR1C3 overexpression in SK‐Hep‐1 cells reduced Erastin‐induced MDA and restored GSH (Figure 2E,F). Similarly, AKR1C3 knockdown elevated Erastin‐induced reactive oxygen species (ROS) and labile iron pool (LIP), which were attenuated by AKR1C3 re‐expression; overexpression yielded the opposite effect (Figure 2G–I). Transmission electron microscopy revealed that AKR1C3 overexpression preserved mitochondrial structure under Erastin stress, while its knockdown exacerbated mitochondrial damage (Figure 2J). Taken together, AKR1C3 promotes the resistance of HCC cells to ferroptosis in an enzyme‐independent pathway. AKR1C3 Alter Ferroptosis Susceptibility via Downregulating TFRC To elucidate the molecular mechanism by which AKR1C3 confers ferroptosis resistance, we performed immunoprecipitation‐mass spectrometry (IP‐MS) to identify potential AKR1C3‐interacting proteins. Among the candidates, several proteins associated with ferroptosis sensitivity or resistance were identified, including TFRC, SLC3A2, PHGDH, PCBP1, PCBP2, GCLC, and ACSL3 (Figures 3A and S2A). We next examined whether AKR1C3 influences the expression of these proteins. Subsequent analysis revealed that AKR1C3 specifically and negatively regulates TFRC protein levels. In contrast, none of the other ferroptosis‐related proteins identified by IP‐MS exhibited significant changes (Figures 3B and S2B). A strong negative correlation was observed between AKR1C3 and TFRC protein levels (R2 = 0.6920, p = 0.0008) using the Cancer Cell Line Encyclopedia (CCLE) database (Figure 3C,D), whereas no significant correlation was found between AKR1C3 and other candidate ferroptosis‐related proteins (Figure S2C). Consistent with this, western blot analysis confirmed a negative correlation between AKR1C3 and TFRC expression (R2 = 0.3478, p = 0.0438) (Figure 3E,F). Moreover, AKR1C3 reduced TFRC protein levels in a dose‐dependent manner (Figure 3G). As shown in Figure 3H, neither wild‐type nor catalytically inactive AKR1C3 affected TFRC mRNA levels. Endogenous and exogenous co‐immunoprecipitation assays confirmed a robust interaction between AKR1C3 and TFRC (Figures 3I and S2D) and confocal microscopy showed their co‐localization primarily at the cell membrane (Figure 3J). AKR1C3 alter ferroptosis susceptibility via downregulating TFRC. (A) IP‐MS strategies used to find binding proteins for AKR1C3. (B) Western blot analysis of TFRC protein in different groups of MHCC‐97 L, HCC‐LM3 and SK‐Hep‐1 cells. (C–D) The expression of AKR1C3 and TFRC and their correlation were assessed in a panel of HCC cell lines from the CCLE database. (E–F) Western blot analysis of AKR1C3 and TFRC protein in a panel of HCC cell lines and the correlation of AKR1C3 and TFRC protein expression. (G) Western blot analysis in HEK293T cells transfected with indicated doses of AKR1C3 for detection of TFRC expression. (H) qPCR analysis of TFRC mRNA in the indicated cells. (I) Immunoprecipitation analysis of the binding of endogenous AKR1C3 and TFRC in MHCC‐97 L cells. (J) Confocal microscopy assay of the colocalization of AKR1C3 and TFRC in MHCC‐97 L, HCC‐LM3 and SK‐Hep‐1 cells. Scale bars, 10 μm. (K) Western blot analysis of different groups of MHCC‐97 L and SK‐Hep‐1 treated. (L) Cell death measurements of different groups of MHCC‐97 L and SK‐Hep‐1 treated with Erastin. (M) The level of intracellular Fe2+, in different groups of MHCC‐97 L and SK‐Hep‐1 treated with Erastin or DFO, analyzed by ImageJ. Scale Bar, 20 μm. n = 3 biological replicates. *p < 0.05; **p < 0.01; ***p < 0.001 and n.s. p > 0.05 between indicated groups. To assess the functional relevance of TFRC in AKR1C3‐mediated ferroptosis resistance, genetic manipulation was performed and validated at the protein level (Figure 3K). Knocking out TFRC in AKR1C3‐deficient cells largely prevented Erastin‐induced ferroptosis. Conversely, overexpressing TFRC in AKR1C3‐overexpressing cells restored ferroptosis sensitivity (Figures 3L and S2E). Mechanistically, the elevated Fe2+, MDA, and depleted GSH caused by AKR1C3 loss were reversed by TFRC knockout, and these effects were blocked by the iron chelator Deferoxamine mesylate (DFO). Accordingly, TFRC overexpression in AKR1C3‐overexpressing SK‐Hep‐1 cells enhanced Erastin‐induced Fe2+ and MDA accumulation and reduced GSH levels, effects that were also blocked by DFO (Figures 3M and S2F,G). Similarly, TFRC knockdown in AKR1C3‐deficient MHCC‐97 L cells markedly attenuated the anti‐proliferative effect of Erastin (Figure S2H). Together, these data clearly demonstrate that AKR1C3 confers ferroptosis resistance in HCC cells by downregulating TFRC. AKR1C3 Aggravates TFRC Ubiquitination by Binding to E3 Ubiquitin Ligase β‐TrCP To investigate how AKR1C3 downregulates TFRC, we assessed TFRC protein stability. AKR1C3 overexpression accelerated TFRC degradation, while its knockdown stabilized TFRC (Figure 4A). This suggests that AKR1C3 promotes TFRC protein turnover. Supporting this finding, exogenous expression of Flag‐AKR1C3 in HEK293T cells enhanced the degradation rate of co‐transfected His‐TFRC (Figure S3A). Notably, the proteasome inhibitor MG132 completely rescued TFRC protein levels in AKR1C3‐overexpressing cells, while the lysosomal inhibitor chloroquine (CQ) had no such effect (Figures 4B and S3B). Furthermore, AKR1C3 knockdown markedly reduced polyubiquitinated TFRC, which was restored upon re‐expression of either wild‐type or catalytically inactive AKR1C3 (Figure 4C). Correspondingly, AKR1C3 overexpression enhanced TFRC ubiquitination in both endogenous settings and HEK293T reconstitution assays (Figure S3C). Collectively, these data establish that AKR1C3 promotes TFRC degradation via the ubiquitin–proteasome system. AKR1C3 directs the E3 ubiquitin ligase β‐TrCP to TFRC and subsequent ubiquitination. (A) Western blot analysis in SK‐Hep‐1/OEAKR1C3 cells and MHCC‐97 L/shAKR1C3 cells for detection of TFRC after the treatment with CHX. Quantitative data of band gray density of TFRC are shown. (B) Western blot analysis in SK‐Hep‐1/OE AKR1C3 cells treated with CQ (20 μM, 4 h) and MG132 (10 μM, 4 h). (C) Immunoprecipitation analysis to detect polyubiquitination of exogenous TFRC in different groups of HCC‐LM3 under MG132 treatment. (D) Western blot analysis in SK‐Hep‐1/OE AKR1C3 cells for detection of TFRC after silencing STUB1 or β‐TrCP. (E) Immunoprecipitation analysis to detect polyubiquitination of exogenous TFRC in SK‐Hep‐1 overexpressing AKR1C3 cells after silencing β‐TrCP. (F) Western blot analysis of β‐TrCP protein in different groups of HCC cells. (G) Exogenous and endogenous immunoprecipitation assay of the interaction between TFRC and β‐TrCP. (H) Confocal microscopy assay of the colocalization of TFRC andβ‐TrCP in different groups of HCC‐LM3 cells. Scale bars, 5 μm. (I) Schematic diagram of the construction of full‐length and truncated AKR1C3 proteins. (J) Exogenous immunoprecipitation analysis of the interaction between truncated AKR1C3 and TFRC. (K) Exogenous immunoprecipitation analysis of the interaction between truncated AKR1C3 and β‐TrCP. (L) Exogenous immunoprecipitation assay of the interaction between TFRC and β‐TrCP in HEK293T transfected with truncated AKR1C3. (M) Immunoprecipitation analysis to detect polyubiquitination of exogenous TFRC in HEK293T transfected with truncated AKR1C3. n = 3 biological replicates. *p < 0.05; **p < 0.01; ***p < 0.001 and n.s. p > 0.05 between indicated groups. To identify the E3 ubiquitin ligase responsible for AKR1C3‐mediated TFRC degradation, we tested candidates previously implicated in TFRC regulation. Knockdown experiments revealed that specifically silencing β‐TrCP, but not STUB1, restored TFRC protein levels in AKR1C3‐overexpressing cells (Figure 4D). Furthermore, silencing of Siah2, an E3 ligase previously linked to AKR1C3, also failed to attenuate TFRC degradation (Figure S3D) [13], confirming the specificity of β‐TrCP in this pathway. Moreover, β‐TrCP depletion substantially impaired AKR1C3‐enhanced TFRC ubiquitination (Figure 4E). Functionally, knocking down β‐TrCP in AKR1C3‐overexpressing cells increased sensitivity to Erastin‐induced ferroptosis. This loss of resistance was only rescued by re‐expressing wild‐type β‐TrCP, not a mutant version defective in forming the SCF complex (Figure S3E,F). Consistently, β‐TrCP knockdown restored the LIP in AKR1C3‐overexpressing cells, and this effect was reversed only by wild‐type β‐TrCP (Figure S3G). These results underscore the essential role of β‐TrCP E3 ligase activity in AKR1C3‐mediated ferroptosis resistance. Co‐immunoprecipitation assays confirmed a robust interaction between AKR1C3 and β‐TrCP (Figure S3H). However, AKR1C3 did not alter β‐TrCP protein levels (Figure 4F), nor did β‐TrCP influence AKR1C3 stability (Figure S3I). This led us to hypothesize that AKR1C3 may function as a scaffold to facilitate the TFRC–β‐TrCP interaction. Indeed, in both exogenous and endogenous co‐immunoprecipitation assays, the presence of AKR1C3 significantly enhanced the binding between TFRC and β‐TrCP (Figure 4G). Confocal microscopy further revealed that AKR1C3 knockdown reduced β‐TrCP cytoplasmic localization and its co‐localization with TFRC, which was rescued by re‐expression of wild‐type or catalytically inactive AKR1C3 (Figure 4H). Conversely, forced expression of AKR1C3 promoted β‐TrCP nuclear export and enhanced its co‐localization with TFRC (Figure S3J). The results of nucleoplasmic separation also supported the above findings (Figure S3K). To molecularly define the regions of AKR1C3 responsible for its scaffolding function, we generated a series of truncated mutants. AKR1C3, a member of the aldo‐keto reductase superfamily, adopts a conserved (α/β)8‐barrel (TIM barrel) structure that comprises its catalytic core. As no established functional domains outside this core are available to guide truncation, we divided the 323‐amino‐acid full‐length protein into three structural segments surrounding the central catalytic channel: an N‐terminal region, a central segment, and a C‐terminal region (Figure S3L). Accordingly, we generated AKR1C3M1 (amino acids 106–323), AKR1C3M2 (a deletion mutant lacking residue 106–217), and AKR1C3M3 (amino acids 1–217) (Figure 4I). Co‐immunoprecipitation assays using Flag‐tagged AKR1C3 constructs and His‐TFRC demonstrated that AKR1C3M2 failed to bind TFRC, whereas both AKR1C3M1 and AKR1C3M3 retained this interaction (Figure 4J), identifying residues 106–217 as the critical region for TFRC binding. Conversely, parallel Co‐IP experiments using Myc‐tagged AKR1C3 constructs and Flag‐β‐TrCP revealed that AKR1C3M1 lost β‐TrCP binding capacity, while AKR1C3M2 and AKR1C3M3 maintained the interaction (Figure 4K), indicating that the N‐terminal region (residues 1–105) is essential for β‐TrCP recruitment. Consistent with these findings, neither AKR1C3M1 nor AKR1C3M2 could facilitate β‐TrCP‐TFRC complex formation or promote TFRC ubiquitination (Figure 4L,M). These results demonstrate that β‐TrCP and TFRC bind to distinct domains of AKR1C3, enabling the formation of a stable ternary complex. We next sought to identify the specific region within β‐TrCP that mediates its interaction with AKR1C3. Since AKR1C3 is not a degradation substrate of β‐TrCP, we reasoned that it might not bind to the canonical substrate‐binding WD40 domain (residues 301–605). We therefore generated a truncated β‐TrCP mutant lacking the F‐box domain (β‐TrCPΔF‐box, Δ191–228). Co‐IP assays showed that β‐TrCPΔF‐box still interacted with AKR1C3 (Figure S4A), indicating that F‐box domains are not required for this interaction. However, a β‐TrCP mutant lacking the N‐terminal region (β‐TrCPΔN, residues 1–190 deleted) completely lost its ability to bind AKR1C3 (Figure S4B), establishing the N‐terminal domain as essential for the interaction. Functionally, AKR1C3 could not promote β‐TrCPΔN‐TFRC complex formation and the subsequent TFRC ubiquitination (Figure S4C,D). Confocal microscopy further confirmed that AKR1C3 failed to facilitate the cytoplasmic redistribution of β‐TrCPΔN or its co‐localization with TFRC (Figure S4E). To assess the physiological relevance of this interaction, we reconstituted β‐TrCP‐depleted SK‐Hep‐1 cells with either wild‐type β‐TrCP or β‐TrCPΔN. While AKR1C3 overexpression conferred strong ferroptosis resistance in cells expressing wild‐type β‐TrCP, this effect was significantly attenuated in cells expressing β‐TrCPΔN (Figure S4F,G). Correspondingly, the characteristic ferroptosis phenotypes—elevated MDA and LIP levels, and GSH depletion—were markedly less pronounced in β‐TrCPΔN‐reconstituted cells compared to wild‐type controls upon AKR1C3 overexpression (Figure S4H–J). Taken together, these findings suggest that AKR1C3 binds to the N‐terminal region of β‐TrCP via its N‐terminal region, thereby promoting nuclear export of β‐TrCP and hence formation of the TFRC‐β‐TrCP complex, which promotes TFRC degradation and confers ferroptotic resistance. AKR1C3 Involved in HCC Sorafenib Resistance Sorafenib, a first‐line therapy for advanced HCC, kills tumor cells in part by inducing ferroptosis [14, 15, 16]. We first confirmed that the ferroptosis inhibitor ferrostatin‐1 blocks sorafenib‐induced cell death (Figure S5A). We then investigated whether the AKR1C3‐TFRC axis contributes to sorafenib resistance. Bioinformatic analysis of GEO datasets showed that AKR1C3 expression is significantly higher in sorafenib‐resistant patient tissues and cell lines (Figure 5A). Data from CCLE and GDSC2 databases further revealed a positive correlation between AKR1C3 protein levels and sorafenib IC50 values (R 2 = 0.4646, p = 0.043) (Figure 5B). In vitro validation across eight HCC cell lines confirmed this relationship, showing a strong negative correlation between AKR1C3 expression and sorafenib sensitivity (R 2 = 0.661, p = 0.026) (Figure S5B,C). Clinically, AKR1C3 was markedly upregulated in recurrent HCC tissues compared to primary tumors, while TFRC expression decreased (Figures 5C and S5D). In vitro, sorafenib treatment induced AKR1C3 expression in a time‐ and dose‐dependent manner (Figures 5D,E and S5E,F). To model clinical resistance dynamics, we established patient‐derived xenografts (PDX) and derived sorafenib‐resistant F0‐F9 sublines (Figure 5F). In this process, we extracted tumor cells from different generations of tissues and determined the resistance indices of F3, F5, and F9 (RI F3 = 1.8, RI F5 = 3.7, and RI F9 = 7.2), respectively (Figure 5G,H). AKR1C3 expression progressively increased across resistant generations, alongside rising resistance indices, whereas TFRC showed the opposite trend (Figures 5J,K and S5G,H). Survival analysis indicated that high AKR1C3 expression, or a high AKR1C3/TFRC ratio, predicted poor overall survival in HCC patients and was associated with reduced effectiveness of sorafenib treatment (Figure 5L–O), suggesting that the AKR1C3‐TFRC axis may have a preliminary link with sorafenib resistance in HCC. AKR1C3 involved in HCC sorafenib resistance. (A) AKR1C3 gene expression analysis from the GEO database containing HCC Sorafenib‐resistance and Sorafenib‐sensitive tissues and cells. (B) Correlation between AKR1C3 protein expression (CCLE) and sorafenib IC₅₀ (GDSC) in HCC cells (Spearman's correlation). (C) AKR1C3, NRF2 and TFRC expression from primary and recurrent after sorafenib treatment HCC patients, Scale Bar, 100 μm. (D–E) Western blot and qPCR analyses of AKR1C3 and NRF2 expression in MHCC‐97 L and SK‐Hep‐1 cells treated with 10 μM Sorafenib for the indicated times. (D) Western blot analysis of AKR1C3 and NRF2 in MHCC‐97 L and SK‐Hep‐1 cells treated with 10 μM Sorafenib for the indicated times. (E) qPCR analysis in MHCC‐97 L and SK‐Hep‐1 cells treated with 10 μM Sorafenib for the indicated times. (F) Patient‐derived xenograft (PDX) and derived sorafenib‐resistant F0–F9 sublines. (G) Resistance indexes of F0, F3, F5, and F9. (H) Sensitivity of cells from F0, F3, F5, and F9 to sorafenib. (I) Western blot analysis of NRF2 and AKR1C3 proteins in MHCC‐97 L and SK‐Hep‐1 cells with indicated transfection under Sorafenib treatment. (J–K) Protein levels of AKR1C3, NRF2, and TFRC in serially passaged cells (F0 to F9) and in a Sorafenib‐resistant PDX model. (L–M) The relationship between AKR1C3 expression levels and the overall survival time of liver cancer patients with or without sorafenib treatment by Kaplan–Meier Plotter. (N–O) The relationship between the ratio of AKR1C3 to TFRC and the overall survival time of liver cancer patients with or without sorafenib treatment by Kaplan–Meier Plotter. n = 3 biological replicates. *p < 0.05; **p < 0.01; ***p < 0.001 and n.s. p > 0.05 between indicated groups. Since the transcription factor NRF2 is known to regulate AKR1C3 and can be activated by ferroptosis inducers like sorafenib, we examined its role [7, 17]. NRF2 was also elevated in relapsed HCC samples (Figures 5C and S5D), and in our sorafenib‐resistant PDX models (Figures 5J,K and S5G,H). In vitro, sorafenib induced NRF2 alongside AKR1C3 (Figure 5D and Figure S5E,I,J), and NRF2 knockdown significantly attenuated sorafenib‐induced AKR1C3 expression (Figure 5I). Collectively, these findings establish the NRF2‐AKR1C3‐TFRC axis as a central mediator of ferroptosis resistance in HCC. AKR1C3 Inhibition Reverses HCC Sorafenib Resistance To determine whether inhibiting AKR1C3 could reverse sorafenib resistance, we modulated its expression in primary cells derived from our sorafenib‐resistant PDX model (Figure 6A). Knockdown of AKR1C3 in resistant cells restored sorafenib sensitivity and enhanced its anti‐proliferative effect, while its overexpression in sensitive cells conferred resistance (Figure 6B,C). Similar results were observed in MHCC‐97 L and SK‐Hep‐1 cells (Figure S6A,B). Mechanistically, AKR1C3 knockdown in resistant cells augmented sorafenib‐induced ferroptosis, evidenced by increased MDA, ROS, and LIP levels and decreased GSH (Figures 6D–G and S6C,D). AKR1C3 inhibition reverses HCC sorafenib resistance. (A) Western blot analysis of XB‐1/OEAKR1C3 and XB‐1 R/shAKR1C3 cells. (B) IC50 assay of XB‐1 R cells knocking down AKR1C3 treated with sorafenib. (C) Colony formation assay of XB‐1 R/shAKR1C3 cells treated with sorafenib. (D–G) Intracellular MDA, GSH, ROS and Fe2+ content of XB‐1 R/shAKR1C3 cells treated with sorafenib. Scale bars, 100 μm (F), 10 μm (G). (H–I) Xenograft tumors of XB‐1 R or XB‐1 R/shAKR1C3 cells in mice supplemented with Sorafenib. Tumor volume and nodule sizes in the liver 1 month after cell injection, analyzed by Halo Digital Pathology Image Platform software, Scale Bar, 1 mm or 500 μm. (J) The weight of tumors stripped from the liver in the above group. (K) Representative image of 4‐HNE and Ki67 staining in above mentioned xenograft tumors. Scale bars, 100 μm. (L) The image of the isolated tumors was presented. (M–N) Xenograft tumors of XB‐1 or XB‐1/OEAKR1C3 cells in mice supplemented with sorafenib and indomethacin. The growth kinetics of the formed tumors were monitored and analyzed, along with the weights of the isolated tumors. (O) Representative image of TFRC, 4‐HNE and Ki67 staining in above mentioned xenograft tumors. Scale bars, 100 μm for 4‐HNE, 50 μm for Ki67 and TFRC. (P‐Q) Xenograft tumors of different group of XB‐1 cells in mice supplemented with sorafenib. Growth kinetics of the formed tumors were monitored and analyzed. Tumor weights of the isolated tumors were analyzed and compared. (R) Representative image of 4‐HNE and Ki67 staining in above mentioned xenograft tumors. n = 5 or 6 biological replicates. Scale bars, 100 μm for 4‐HNE, 50 μm for Ki67. *p < 0.05; **p < 0.01; ***p < 0.001 and n.s. p > 0.05 between indicated groups. In vivo, AKR1C3 knockdown in orthotopic liver tumors derived from resistant cells potentiated the antitumor efficacy of sorafenib, resulting in reduced tumor growth, decreased proliferation (Ki67), and increased lipid peroxidation (4‐HNE) (Figures 6H–K and S6E). IHC confirmed the potent knockdown of AKR1C3 and the increase in TFRC protein levels (Figure S6F). In contrast, overexpression of AKR1C3 in subcutaneous xenograft tumors promoted tumor growth and significantly attenuated the inhibitory effect of sorafenib on tumor proliferation and lipid peroxidation (Figure 6L–O). Finally, we tested if restoring TFRC could counteract AKR1C3‐driven resistance. In subcutaneous models, the accelerated tumor growth induced by AKR1C3 overexpression under sorafenib treatment was significantly reversed by co‐overexpression of TFRC (Figures 6P,Q, and S6G). The efficiency of AKR1C3 and TFRC overexpression was demonstrated by IHC (Figure S6H). TFRC restoration also enhanced lipid peroxidation and proliferation inhibition in these tumors (Figure 6R). In summary, AKR1C3 promotes sorafenib resistance in HCC by suppressing ferroptosis, an effect mediated through TFRC downregulation. Targeting AKR1C3 represents a promising strategy to overcome sorafenib resistance. Discussion Our study identifies AKR1C3 as a key metabolic regulator that suppresses ferroptosis by facilitating the β‐TrCP–TFRC interaction, thereby reducing the labile iron pool (LIP). We demonstrate that AKR1C3, a ferroptosis‐related gene upregulated in HCC, is significantly elevated in ferroptosis‐resistant cell lines, highlighting its central role in mediating this resistance (Figure 7). Sorafenib, a first‐line HCC treatment and known ferroptosis inducer, often faces resistance [18, 19, 20]. Our data link this resistance to AKR1C3: its expression is increased in relapsed HCC patients after sorafenib therapy and progressively rises in a sorafenib‐resistant PDX model. Genetic inhibition of AKR1C3 resensitizes HCC to sorafenib, confirming its functional role. Furthermore, the observed upregulation of AKR1C3 following long‐term sorafenib exposure appears to be driven by the transcription factor NRF2. Graphical summary illustrating the mechanism by which AKR1C3 reduces the sensitivity of HCC to ferroptosis. AKR1C3 acts as an enzyme‐independent scaffold protein by promoting the nuclear export of β‐TrCP and its binding to TFRC in an enzyme‐independent manner. Promotes ubiquitin‐proteasome degradation of TFRC and reduces iron uptake. Meanwhile, AKR1C3 was up‐regulated in NrF2‐driven sorafenib resistance HCC, and inhibition of AKR1C3 expression reversed ferroptosis and sorafenib resistance. The aldo‐keto reductase superfamily member AKR1C3 plays a well‐established role in steroid metabolism, and its dysregulation is implicated in various cancers, including HCC [21, 22, 23]. Beyond its enzymatic functions, our prior work showed that AKR1C3 promotes HCC progression through non‐enzymatic mechanisms [11]. While recent studies link AKR1C3 to ferroptosis resistance in several cancers [24, 25, 26], the reported mechanism in HCC involved modulation of the YAP/SLC7A11 axis [27]. In contrast, our study reveals a distinct, enzyme‐independent scaffold function: AKR1C3 physically bridges the E3 ligase β‐TrCP to its substrate TFRC, thereby promoting TFRC ubiquitination and degradation to suppress ferroptosis. This role as a positive adaptor driving specific substrate degradation in iron metabolism is a novel finding for AKR1C3. Together with emerging evidence of its scaffolding function in other contexts, these findings solidify AKR1C3 disruption as a promising strategy to overcome ferroptosis resistance in cancer. Notably, our previous work showed that AKR1C3 stabilizes PARP1 independently of its catalytic activity to promote HCC cisplatin resistance [11]. In the present study, further analyses indicated that β‐TrCP does not interact with PARP1, and the AKR1C3‐PARP1 and AKR1C3‐β‐TrCP‐TFRC axes function as two independent parallel pathways in HCC. These findings highlight that AKR1C3 employs distinct mechanisms to regulate multiple targets simultaneously, thereby coordinately driving HCC progression and drug resistance. TFRC, a key mediator of cellular iron uptake, is essential for supplying the LIP that drives ferroptosis [28]. While elevated TFRC increases ferroptosis sensitivity, its targeted degradation is a known resistance mechanism in cancer [8, 29]. The study by Guo et al. established the TRIB2‐β‐TrCP‐TFRC axis as a central pathway to desensitize HCC cells to ferroptosis by reducing labile iron [30]. Our work positions AKR1C3 as a novel, stress‐inducible regulatory node on the established β‐TrCP‐TFRC degradation axis. Specifically, AKR1C3 promotes nuclear export of β‐TrCP by binding to its non‐functional N‐terminal sequence, and AKR1C3 functions as a scaffold: its central domain binds TFRC, while its N‐terminus recruits β‐TrCP via the E3 ligase's N‐terminal region. This ternary complex promotes TFRC polyubiquitination and degradation, thereby depleting LIP and conferring ferroptosis resistance. Thus, our work identifies a novel AKR1C3‐TFRC regulatory pathway to mediate ferroptosis susceptibility in HCC. Inhibition of the AKR1C3‐TFRC axis is positioned as a promising sensitization strategy to overcome underlying resistance mechanisms and thereby improve the efficacy of a variety of anticancer therapies. However, TFRC plays a context‐dependent, dual role in HCC. Under normal conditions, it supports proliferation by supplying iron [31], Under ferroptotic stress, however, it becomes a liability, as excess iron uptake catalyzes lethal lipid peroxidation [32]. We propose that the AKR1C3‐mediated degradation of TFRC represents an adaptive survival strategy. Our model proposes that when faced with oxidative stress and ferroptosis, cancer cells up‐regulate AKR1C3 to temporarily reduce iron uptake. The immediate goal of this action is not to inhibit tumor growth, sacrificing some iron uptake capacity to avoid ferroptotic death. This could explain why this mechanism may be enriched in advanced or resistant tumors. While our study provides mechanistic insights into AKR1C3‐mediated ferroptosis regulation, several questions remain. First, in vivo validation in patient‐derived xenografts or genetically engineered mouse models is warranted to confirm the results of AKR1C3 promoting the interaction between TFRC and β‐TrCP. Also, structural dissection of AKR1C3‐TFRC‐β‐TrCP interfaces—particularly identifying critical binding residues—would enable targeted inhibition of this complex. In addition, due to the limited cohort size, the clinical analysis is preliminary and should be subsequently validated in larger clinical samples. Author Contributions Lei Qi: writing – original draft, methodology, investigation. Jingyi Hua: writing – original draft, investigation, methodology. Di Pan: writing – review and editing, investigation, resources, methodology. Wanwan Yang: investigation, methodology. Geng Tian: supervision, writing – review and editing. Fangyu Ye: conceptualization, writing – review and editing. Lingxiang Liu: methodology. Yuhan Mao: investigation, methodology. Qinglong Guo: methodology. Haopeng Sun: software, methodology. Li Zhao: conceptualization, funding acquisition, resources. Funding This work was supported by the General Program of State Key Laboratory of Neurology and Oncology Drug Development (SKLSIM‐2024150); the program of the National Science Foundation of China (82273962, 82304556, 82473962, 82460488); and the Normal Project of Guizhou Province Department of Science and Technology (ZK[2024]142). Disclosure The authors have nothing to report. Ethics Statement Informed consent was obtained from all participants. This study was performed in line with the principles of the Declaration of Helsinki. Approval was granted by the Animal Care and Use Committee of China Pharmaceutical University (YSL‐202509013) and the Nanjing Drum Tower Hospital Institutional Review Board (IRB) (No AF/SC‐08/03.0). Conflicts of Interest The authors declare no conflicts of interest. Supporting information Figure S1: (A) Visualization of AKR1C3 enzyme active mutation sites. (B) IC50 of different groups of MHCC‐97 L, HCC‐LM3 and SK‐Hep‐1 cells for Erastin. (C) LDH release assay of different groups of MHCC‐97 L, HCC‐LM3 and SK‐Hep‐1 cells treated with erastin (10 μM) and fer‐1 (2 μM). Data and error bars are mean ± s.d. n = 3 biological independent experiments. Statistical analysis was done with two‐tailed student's t test. NC: nontargeting control. Figure S2: (A) Seven ferroptosis‐associated proteins with potential binding to Aldo‐keto reductase family 1 member C3 (AKR1C3) identified by IP‐MS. (B) Western blot analysis of ferroptosis‐related protein potentially interacting with AKR1C3 in different groups of MHCC‐97 L, HCC‐LM3 and SK‐Hep‐1 cells. (C) The expression of ferroptosis‐related protein potentially interacting with AKR1C3 in a panel of cell lines from Cancer Cell Line Encyclopedia (CCLE) database. (D) Co‐immunoprecipitation was used to analyze the binding of endogenous AKR1C3 and TFRC in HCC‐LM3 and exogenous Flag‐AKR1C3 and His‐TFRC in HEK‐293 T cells. (E) LDH release assays of different groups of MHCC‐97 L and SK‐Hep‐1 treated with erastin (10 μM). (F–G) The level of intracellular GSH and MDA, in different groups of MHCC‐97 L and SK‐Hep‐1 treated with erastin (10 μM). (H) 5‐ethynyl‐2‐deoxyuridine (EdU) analysis of MHCC‐97 L cells silencing AKR1C3 or Transferrin receptor protein 1 (TFRC) treated with erastin (10 μM). The nucleus was stained with Hoechst 33342 (blue). Scale bars, 250 μm. The images were analyzed by ImageJ. The Western blot experiments were repeated three times independently. n = 3 biological independent experiments. Data and error bars are mean ± s.d. Statistical analysis was done with two‐tailed student's t test. NC: nontargeting control. Figure S3: (A) Western blot analysis in HEK‐293 T transfected Flag‐Aldo‐keto reductase family 1 member C3 (AKR1C3) and His‐Transferrin receptor protein 1 (TFRC) for detection of TFRC after the treatment with Cycloheximide (CHX) (10 μg/mL). Quantitative data of band gray density of His are shown. (B) Quantitative data of band gray density of TFRC. (C) Immunoprecipitation analysis to detect polyubiquitination of TFRC in SK‐Hep‐1 cells overexpressing AKR1C3 and HEK293T transfected Flag‐AKR1C3 under MG132 treatment. (D) Western blot analysis in SK‐Hep‐1 overexpressing AKR1C3 cells for detection of TFRC after silencing siah E3 ubiquitin protein ligase 2 (Siah2). (E–F) Cell death and LDH release assays of different groups of SK‐Hep‐1 cells treated with erastin (10 μM). (G) Intracellular Fe2+ content of different groups of SK‐Hep‐1 cells treated with erastin (10 μM) or DFO (100μM). The images were analyzed by ImageJ, Scale Bar, 50 μm. (H) Immunoprecipitation analysis of the binding of endogenous AKR1C3 and β‐transducing repeat protein (β‐TrCP) in MHCC‐97 L and HCC‐LM3 cells. (I) Western blot analysis of β‐TrCP protein in different groups of MHCC‐97 L, HCC‐LM3 and SK‐Hep‐1 cells. (J) Confocal microscopy assay of the colocalization of TFRC and β‐TrCP in different groups of SK‐Hep‐1 cells. The nucleus was stained with 4′,6‐diamidino‐2‐phenylindole (blue), and the colocalization of β‐TrCP (green) and TFRC (red) was visualized as yellow fluorescence in the merged panel. Scale bars, 5 μm. (K) Western blot analysis of β‐TrCP proteins isolated from nuclei and cytoplasm of different groups of MHCC‐97 L and SK‐Hep‐1 cells. (L) Different domains in the 3D structure of AKR1C3. The Western blot experiments were repeated three times independently. n = 3 biological independent experiments. Data and error bars are mean ± s.d. Statistical analysis was done with two‐tailed student's t test. UB: Ubiquitin. Figure S4: (A–B) Schematic diagram of the construction of full‐length and truncated β‐transducing repeat protein (β‐TrCP) proteins, and exogenous immunoprecipitation analysis of the interaction between truncated HA‐β‐TrCP and Flag‐Aldo‐keto reductase family 1 member C3 (AKR1C3). (C) Exogenous immunoprecipitation assay of the interaction between His‐Transferrin receptor protein 1 (TFRC) and HA‐β‐TrCP or HA‐β‐TrCPΔN in HEK293T transfected with Flag‐AKR1C3. (D) Immunoprecipitation analysis to detect polyubiquitination of exogenous TFRC in HEK‐293 T transfected with truncated β‐TrCP. (E) Confocal microscopy assay of the colocalization of TFRC and HA‐β‐TrCP or HA‐β‐TrCPΔN in SK‐Hep‐1 transfected with Flag‐AKR1C3 and truncated HA‐β‐TrCP. The nucleus was stained with 4′,6‐diamidino‐2‐phenylindole (blue), and the colocalization of HA (green) and TFRC (red) was visualized as yellow fluorescence in the merged panel. Scale bars, 5 μm. (F–G) Cell death and LDH release assays of different groups of siβ‐TrCP SK‐Hep‐1 cells treated with erastin (10 μM). (H–J) Intracellular MDA, GSH and Fe2+ content of different groups of siβ‐TrCP SK‐Hep‐1 cells treated with erastin (10 μM) or DFO (100μM). The images were analyzed by ImageJ, Scale Bar, 50 μm. The Western blot experiments were repeated three times independently. n = 3 biological independent experiments. Data and error bars are mean ± s.d. Statistical analysis was done with two‐tailed student's t test. UB: Ubiquitin. Figure S5: (A) Cell viability measurement of MHCC‐97 L and SK‐Hep‐1 treated with sorafenib (20 μM) and ferrostatin‐1 (2 μM) for 24 h. (B) The ferroptotic sensitivity assessment of a panel of hepatocellular carcinoma (HCC) cell lines after the treatment of indicated concentrations of sorafenib. (C) The correlation of Aldo‐keto reductase family 1 member C3 (AKR1C3) expression and half maximal inhibitory concentration (IC50) of sorafenib. Bivariate relationship was calculated by Spearman's rank correlation coefficient. (D) Immunochemistry intensity scoring of AKR1C3, Nuclear factor E2‐related factor 2 (NRF2) and Transferrin receptor protein 1 (TFRC) staining in tumors primary and recurrent after sorafenib treatment HCC patients. The H‐score analysis using Halo Digital Pathology Image Platform software. (E) Western blot analysis in MHCC‐97 L and SK‐Hep‐1 cells with different concentrations of sorafenib for 24 h for detection of AKR1C3 and NRF2. (F) qPCR analysis in MHCC‐97 L and SK‐Hep‐1 cells with different concentrations of sorafenib for 24 h for detection of AKR1C3 mRNA level. (G) Immunochemistry intensity scoring of AKR1C3, NRF2 and TFRC staining in sorafenib‐resistance Patient‐derived tumor xenograft (PDX) model. The H‐score analysis using Halo Digital Pathology Image Platform software. (H) The mRNA levels of AKR1C3 and Nrf2 in cells from F0, F3, F5, and F9. (I) real time PCR (qPCR) analysis in MHCC‐97 L and SK‐Hep‐1 cells treated with 10 μM sorafenib for the indicated times for detection of Nrf2 mRNA level. (J) qPCR analysis in MHCC‐97 L and SK‐Hep‐1 cells with different concentrations of sorafenib for 24 h for detection of Nrf2 mRNA level. The Western blot experiments were repeated three times independently. n = 3 biological independent experiments. Data and error bars are mean ± s.d. Statistical analysis was done with two‐tailed student's t test. Fer‐1: ferrostatin‐1. Figure S6: (A) Half maximal inhibitory concentration (IC50) assay of MHCC‐97 L knocking down Aldo‐keto reductase family 1 member C3 (AKR1C3) cells and SK‐Hep‐1 overexpressing AKR1C3 cells treated with sorafenib (10 μM). (B) Colony formation assay of MHCC‐97 L knocking down AKR1C3 cells and SK‐Hep‐1 overexpressing AKR1C3 cells treated with sorafenib (10 μM). (C–D) Quantification of intracellular Reactive oxygen species (ROS) and Fe2+ content of XB‐1 R cells knocking down AKR1C3 after treatment with sorafenib (10 μM). The images were analyzed by ImageJ. (E) Images showing liver cancer tumors in situ. (F) Representative image of AKR1C3 and Transferrin receptor protein 1 (TFRC) staining in tumors of XB‐1 R or XB‐1 R AKR1C3 knockdown cells in mice treated with sorafenib (120 mg/kg, i.g. once 2 days). Scale bars, 100 μm. (G) The image of the isolated tumors was presented. (H) Representative image of AKR1C3 and Transferrin receptor protein 1 (TFRC) staining in xenograft tumors of different group of XB‐1 cells in mice supplemented with sorafenib (120 mg/kg, i.g. once 2 days). Scale bars, 50 μm. n = 3 or 6 biological independent experiments. Data and error bars are mean ± s.d. Statistical analysis was done with two‐tailed student's t test. AU: Arbitrary Unit; NC: nontargeting control; DMSO: Dimethyl sulfoxide. Table S1: Patients' information of 9 recurrence LIHCs and 17 primary LIHCs. Table S2: The sequence for gene‐specific shRNA and siRNA. Table S3: Primers for qPCR. Table S4: Antibodies used in this study. Data S1: Supporting Information. Acknowledgments We thank Liyuan Zhang from Public Experimental Platform, China Pharmaceutical University for her technical guidance. Also, we wish to thank the help of the following professors of China Pharmaceutical University, Dr. Lei Qiang for gifting plasmids, and Dr. Xiaosheng Wang for the bio‐information direction. Further, we acknowledge pathologists Hongyan Wu (Nanjing Drum Tower Hospital) and Zhiwen Li (Nanjing Drum Tower Hospital) for acquiring HCC tissue. In addition, we would like to thank Associate Professor Wei Liu (Nanjing Medical University) for her guidance in electron microscopy photography. Contributor Information Qinglong Guo, Email: anticancer_drug@163.com. Haopeng Sun, Email: sunhaopeng@163.com. Li Zhao, Email: zhaoli@cpu.edu.cn. Data Availability Statement Data sharing not applicable to this article as no datasets were generated or analysed during the current study. References Associated Data Supplementary Materials Figure S1: (A) Visualization of AKR1C3 enzyme active mutation sites. (B) IC50 of different groups of MHCC‐97 L, HCC‐LM3 and SK‐Hep‐1 cells for Erastin. (C) LDH release assay of different groups of MHCC‐97 L, HCC‐LM3 and SK‐Hep‐1 cells treated with erastin (10 μM) and fer‐1 (2 μM). Data and error bars are mean ± s.d. n = 3 biological independent experiments. Statistical analysis was done with two‐tailed student's t test. NC: nontargeting control. Figure S2: (A) Seven ferroptosis‐associated proteins with potential binding to Aldo‐keto reductase family 1 member C3 (AKR1C3) identified by IP‐MS. (B) Western blot analysis of ferroptosis‐related protein potentially interacting with AKR1C3 in different groups of MHCC‐97 L, HCC‐LM3 and SK‐Hep‐1 cells. (C) The expression of ferroptosis‐related protein potentially interacting with AKR1C3 in a panel of cell lines from Cancer Cell Line Encyclopedia (CCLE) database. (D) Co‐immunoprecipitation was used to analyze the binding of endogenous AKR1C3 and TFRC in HCC‐LM3 and exogenous Flag‐AKR1C3 and His‐TFRC in HEK‐293 T cells. (E) LDH release assays of different groups of MHCC‐97 L and SK‐Hep‐1 treated with erastin (10 μM). (F–G) The level of intracellular GSH and MDA, in different groups of MHCC‐97 L and SK‐Hep‐1 treated with erastin (10 μM). (H) 5‐ethynyl‐2‐deoxyuridine (EdU) analysis of MHCC‐97 L cells silencing AKR1C3 or Transferrin receptor protein 1 (TFRC) treated with erastin (10 μM). The nucleus was stained with Hoechst 33342 (blue). Scale bars, 250 μm. The images were analyzed by ImageJ. The Western blot experiments were repeated three times independently. n = 3 biological independent experiments. Data and error bars are mean ± s.d. Statistical analysis was done with two‐tailed student's t test. NC: nontargeting control. Figure S3: (A) Western blot analysis in HEK‐293 T transfected Flag‐Aldo‐keto reductase family 1 member C3 (AKR1C3) and His‐Transferrin receptor protein 1 (TFRC) for detection of TFRC after the treatment with Cycloheximide (CHX) (10 μg/mL). Quantitative data of band gray density of His are shown. (B) Quantitative data of band gray density of TFRC. (C) Immunoprecipitation analysis to detect polyubiquitination of TFRC in SK‐Hep‐1 cells overexpressing AKR1C3 and HEK293T transfected Flag‐AKR1C3 under MG132 treatment. (D) Western blot analysis in SK‐Hep‐1 overexpressing AKR1C3 cells for detection of TFRC after silencing siah E3 ubiquitin protein ligase 2 (Siah2). (E–F) Cell death and LDH release assays of different groups of SK‐Hep‐1 cells treated with erastin (10 μM). (G) Intracellular Fe2+ content of different groups of SK‐Hep‐1 cells treated with erastin (10 μM) or DFO (100μM). The images were analyzed by ImageJ, Scale Bar, 50 μm. (H) Immunoprecipitation analysis of the binding of endogenous AKR1C3 and β‐transducing repeat protein (β‐TrCP) in MHCC‐97 L and HCC‐LM3 cells. (I) Western blot analysis of β‐TrCP protein in different groups of MHCC‐97 L, HCC‐LM3 and SK‐Hep‐1 cells. (J) Confocal microscopy assay of the colocalization of TFRC and β‐TrCP in different groups of SK‐Hep‐1 cells. The nucleus was stained with 4′,6‐diamidino‐2‐phenylindole (blue), and the colocalization of β‐TrCP (green) and TFRC (red) was visualized as yellow fluorescence in the merged panel. Scale bars, 5 μm. (K) Western blot analysis of β‐TrCP proteins isolated from nuclei and cytoplasm of different groups of MHCC‐97 L and SK‐Hep‐1 cells. (L) Different domains in the 3D structure of AKR1C3. The Western blot experiments were repeated three times independently. n = 3 biological independent experiments. Data and error bars are mean ± s.d. Statistical analysis was done with two‐tailed student's t test. UB: Ubiquitin. Figure S4: (A–B) Schematic diagram of the construction of full‐length and truncated β‐transducing repeat protein (β‐TrCP) proteins, and exogenous immunoprecipitation analysis of the interaction between truncated HA‐β‐TrCP and Flag‐Aldo‐keto reductase family 1 member C3 (AKR1C3). (C) Exogenous immunoprecipitation assay of the interaction between His‐Transferrin receptor protein 1 (TFRC) and HA‐β‐TrCP or HA‐β‐TrCPΔN in HEK293T transfected with Flag‐AKR1C3. (D) Immunoprecipitation analysis to detect polyubiquitination of exogenous TFRC in HEK‐293 T transfected with truncated β‐TrCP. (E) Confocal microscopy assay of the colocalization of TFRC and HA‐β‐TrCP or HA‐β‐TrCPΔN in SK‐Hep‐1 transfected with Flag‐AKR1C3 and truncated HA‐β‐TrCP. The nucleus was stained with 4′,6‐diamidino‐2‐phenylindole (blue), and the colocalization of HA (green) and TFRC (red) was visualized as yellow fluorescence in the merged panel. Scale bars, 5 μm. (F–G) Cell death and LDH release assays of different groups of siβ‐TrCP SK‐Hep‐1 cells treated with erastin (10 μM). (H–J) Intracellular MDA, GSH and Fe2+ content of different groups of siβ‐TrCP SK‐Hep‐1 cells treated with erastin (10 μM) or DFO (100μM). The images were analyzed by ImageJ, Scale Bar, 50 μm. The Western blot experiments were repeated three times independently. n = 3 biological independent experiments. Data and error bars are mean ± s.d. Statistical analysis was done with two‐tailed student's t test. UB: Ubiquitin. Figure S5: (A) Cell viability measurement of MHCC‐97 L and SK‐Hep‐1 treated with sorafenib (20 μM) and ferrostatin‐1 (2 μM) for 24 h. (B) The ferroptotic sensitivity assessment of a panel of hepatocellular carcinoma (HCC) cell lines after the treatment of indicated concentrations of sorafenib. (C) The correlation of Aldo‐keto reductase family 1 member C3 (AKR1C3) expression and half maximal inhibitory concentration (IC50) of sorafenib. Bivariate relationship was calculated by Spearman's rank correlation coefficient. (D) Immunochemistry intensity scoring of AKR1C3, Nuclear factor E2‐related factor 2 (NRF2) and Transferrin receptor protein 1 (TFRC) staining in tumors primary and recurrent after sorafenib treatment HCC patients. The H‐score analysis using Halo Digital Pathology Image Platform software. (E) Western blot analysis in MHCC‐97 L and SK‐Hep‐1 cells with different concentrations of sorafenib for 24 h for detection of AKR1C3 and NRF2. (F) qPCR analysis in MHCC‐97 L and SK‐Hep‐1 cells with different concentrations of sorafenib for 24 h for detection of AKR1C3 mRNA level. (G) Immunochemistry intensity scoring of AKR1C3, NRF2 and TFRC staining in sorafenib‐resistance Patient‐derived tumor xenograft (PDX) model. The H‐score analysis using Halo Digital Pathology Image Platform software. (H) The mRNA levels of AKR1C3 and Nrf2 in cells from F0, F3, F5, and F9. (I) real time PCR (qPCR) analysis in MHCC‐97 L and SK‐Hep‐1 cells treated with 10 μM sorafenib for the indicated times for detection of Nrf2 mRNA level. (J) qPCR analysis in MHCC‐97 L and SK‐Hep‐1 cells with different concentrations of sorafenib for 24 h for detection of Nrf2 mRNA level. The Western blot experiments were repeated three times independently. n = 3 biological independent experiments. Data and error bars are mean ± s.d. Statistical analysis was done with two‐tailed student's t test. Fer‐1: ferrostatin‐1. Figure S6: (A) Half maximal inhibitory concentration (IC50) assay of MHCC‐97 L knocking down Aldo‐keto reductase family 1 member C3 (AKR1C3) cells and SK‐Hep‐1 overexpressing AKR1C3 cells treated with sorafenib (10 μM). (B) Colony formation assay of MHCC‐97 L knocking down AKR1C3 cells and SK‐Hep‐1 overexpressing AKR1C3 cells treated with sorafenib (10 μM). (C–D) Quantification of intracellular Reactive oxygen species (ROS) and Fe2+ content of XB‐1 R cells knocking down AKR1C3 after treatment with sorafenib (10 μM). The images were analyzed by ImageJ. (E) Images showing liver cancer tumors in situ. (F) Representative image of AKR1C3 and Transferrin receptor protein 1 (TFRC) staining in tumors of XB‐1 R or XB‐1 R AKR1C3 knockdown cells in mice treated with sorafenib (120 mg/kg, i.g. once 2 days). Scale bars, 100 μm. (G) The image of the isolated tumors was presented. (H) Representative image of AKR1C3 and Transferrin receptor protein 1 (TFRC) staining in xenograft tumors of different group of XB‐1 cells in mice supplemented with sorafenib (120 mg/kg, i.g. once 2 days). Scale bars, 50 μm. n = 3 or 6 biological independent experiments. Data and error bars are mean ± s.d. Statistical analysis was done with two‐tailed student's t test. AU: Arbitrary Unit; NC: nontargeting control; DMSO: Dimethyl sulfoxide. Table S1: Patients' information of 9 recurrence LIHCs and 17 primary LIHCs. Table S2: The sequence for gene‐specific shRNA and siRNA. Table S3: Primers for qPCR. Table S4: Antibodies used in this study. Data S1: Supporting Information. Data Availability Statement Data sharing not applicable to this article as no datasets were generated or analysed during the current study.