Fatty Acid Oxidation Mediated by Malonyl-CoA Decarboxylase Represses Renal Cell Carcinoma Progression

Reprogramming cellular metabolism is a hallmark of cancer (1). Lipid metabolism reprogramming is regarded as the most prominent feature in some tumors, such as renal cell carcinoma (RCC) and prostate cancer. RCC is the most common pathologic type of kidney cancer and is one of the most prevalent malignancies of the urinary system (2). Morphologically, approximately 80% of RCC are clear cell renal cell carcinomas (ccRCC), characterized by aberrant accumulation of lipid droplets (LD) in the cytoplasm (3). Lipid storage is an essential tumor adaptation rather than a bystander effect in tumor progression. Our previous studies reported that inhibiting intratumoral LDs deposition represses cancer progression (4–5). LDs, as subcellular structures, are composed of surface-coated proteins and neutral lipid cores containing triglycerides and cholesterol esters, whose precursors are fatty acids (FA). De novo FA synthesis and suppression of FA oxidation have been confirmed to contribute to lipid storage. Recent studies have reported that HIF represses the transcription of carnitine palmitoyltransferase (CPT1A), driving lipid accumulation and tumor progression in ccRCC by inhibiting FAs transport into the mitochondria for oxidation (6). De novo FA synthesis is a complicated process through which acetyl-coenzyme A (acetyl-CoA) derived from glucose and glutamine is converted into FAs, catalyzed by a variety of FA synthetases, including FA synthase (FASN) and stearoyl-CoA desaturase 1 (SCD1; ref. 7). FASN expression is positively correlated with the progression and poor prognosis of ccRCC, whereas inhibition of SCD1 or acetyl-CoA carboxylase (ACC) decreases FA synthesis and induces cell death in ccRCC (8–10). FA anabolism and catabolism are related to FA synthesis and FA oxidation, respectively. However, how to regulate FA anabolism to prevent de novo FA synthesis while simultaneously increasing FA consumption and completely blocking LD accumulation remains elusive.

It has been demonstrated that, except for FA synthetases, acetyl-CoA, as a central metabolic intermediate, can not only act as a substrate for de novo FA synthesis, but also affects the activities of various FA synthetases and plays a pivotal role in the process of FA synthesis in cells (11). In addition, as an intermediate product of FA synthesis, fatty acyl CoA affects intracellular FA synthetase activities and participates in cellular activities, such as FA oxidation and signal transduction (12). Intracellular fatty acyl CoA plays an important role in FA metabolism. Therefore, understanding the underlying mechanisms and the importance of FA anabolism may provide clues for the development of novel therapeutic strategies for RCC.

Here, we identified malonyl-CoA decarboxylase (MLYCD) as a key gene involved in FA anabolism. Our results demonstrated that MLYCD is downregulated and predicts poor outcomes in RCC. MLYCD regulates FA anabolism by mediating the intracellular content of malonyl-CoA to induce the depletion of LDs, increasing multistress and sensitivity to ferroptosis in RCC. Overexpression of MLYCD induced striking RCC cell death, reduced tumor growth, and reversed sunitinib resistance in vitro and in vivo. Mechanistically, our data demonstrated that HIF inhibits MLYCD translation by upregulating eIF4G3 microexons. Taken together, our results demonstrate a previously unrecognized molecular mechanism regulating FA anabolism in RCC.

Human RCC cell lines (OS-RC-2, 786-O, A498, Caki-1, ACHN, and 769-P) and renal proximal tubular epithelial cell lines (HK-2 and RPTEC) were obtained from the ATCC, which were all authenticated periodically by short tandem repeat profiling (IDEXX BioResearch). Using the high-glucose DMEM that was supplemented with 10% FBS (Gibco) and 1% streptomycin–penicillin (Servicebio) and culturing all the cell lines at 37°C in 5% CO₂. Sunitinib, lysosome inhibitor chloroquine, proteasome inhibitor MG132, ferroptosis inductor erastin, ferroptosis inhibitor ferrostatin-1 (Fer-1), CPT1A inhibitor etomoxir (ETO), and ROS scavenger N-acetyl-cysteine (NAC) were purchased from MedChemExpress (MCE). Malonyl-CoA and palmitate were obtained from Sigma-Aldrich. Lipofectamine 3000 reagent (Thermo Fisher Scientific) was used for transfection.

Animal experiments were approved by the Institutional Animal Care and Use Committee of First Affiliated Hospital of Zhengzhou University (Henan Province, China). C57BL/6J-Mlycd^Loxp/Loxp mice and Ggt1 promoter–driven Cre mice (Ggt1-Cre⁺; ref. 13) mice were purchased from Cyagen Bio-Technique Co. Ltd. F1 generation (Ggt1-Cre⁺; Mlycd^Loxp/+) mice were obtained by crossbreeding Mlycd^Loxp/Loxp mice with Ggt1-Cre⁺ mice. Selected the F1 mice to mate with Mlycd^Loxp/Loxp mice and identified the genotypes of the F2 mice after birth 2 weeks later to obtain kidney proximal tubule–specific Mlycd-knockout mice (Ggt1-Cre⁺; Mlycd^Loxp/Loxp). C57BL/6J-Mlycd^Loxp/LoxpGgt1^Cre+mice were purchased from Shanghai Model Organisms Center, Inc. All the mice were fed under standard conditions of food, water, light, and temperature.

Male BALB/c nude mice were purchased from the Vital River Laboratory Animal Technology Co., Ltd. Sunitinib-resistant RCC cells were obtained as previously described (14). Briefly, 5  × 10⁶ OS-RC-2 (RCB catalog no. RCB0735, RRID:CVCL_1626) or 786-O (CLS catalog no. 300107/p747_786-O, RRID:CVCL_1051) cells were subcutaneously injected into the upper back of the mice. Tumor size was measured every 3 days, and the volume of the xenograft was calculated on the basis of a²b/2 (a > b). After the volume reached 200 mm³, the mice were intragastrically administered vehicle or sunitinib (40 mg/kg/day) according to standard therapy (4 weeks on and 2 weeks off treatment). When one treatment course was completed, the tumors were excised and isolated as single cells (as previously described; ref.15), which were then transplanted into nude mice again and treated with vehicle or sunitinib. The cells isolated from the third-generation tumor were identified as sunitinib-resistant RCC cells (OS-RC-2-Su3rd and 786-O-Su3rd).

To evaluate the function of MLYCD in RCC, the nod mice were randomized into five per group. A total of 5 × 10⁶ 786-O-Su3rd cells stably overexpressing MLYCD or the control (NC) were subcutaneously injected into the upper back of nude mice. Then treated the mice with intragastric administration of vehicle or sunitinib (40 mg/kg/day) from days 15 to 40. The size of the xenografts was measured every three days. After euthanizing the mice, xenografts were excised and measured. All animal experiments were approved by the Institutional Animal Care and Use Committee of First Affiliated Hospital of Zhengzhou University.

Paired RCC and normal kidney tissues and RCC tissue microarrays were obtained from the First Affiliated Hospital of Zhengzhou University and the Huazhong University of Science and Technology Affiliated Union Hospital (Wuhan, Hubei, China). We obtained written informed consent from the patients and permission from Institutional Research Ethics Committee to use these clinical materials.

We performed IHC according to standard methods, as described previously (5). The primary antibodies included MLYCD (Abcam, ab234879, 1:100); HIF1α (ProteinTech, catalog no. 66730–1-Ig, RRID:AB_2882080); HIF2α (Abcam, ab109616, 1:100); VHL (ProteinTech, 24756–1-AP, 1:200); Ki67 (ProteinTech, catalog no. 27309–1-AP, RRID:AB_2756525); PLIN2 (Abcam, catalog no. ab52356, RRID:AB_2223599); PERK (Abcam, catalog no. ab192591, RRID:AB_2728666); PTGS2 (Abcam, ab179800, 1:1,000); caspase-3 (Cell Signaling Technology, #9664S, 1:1,000); RCC (Abcam, ab196022, 1:2,000); CA-IX (Abcam, ab108351, 1:200); vimentin (Abcam, catalog no. ab92547, RRID:AB_10562134); and CK-7 (Abcam, catalog no. ab68459, RRID:AB_1139824). The staining intensity of protein was graded on the basis of the immunoreactivity score (IRS): 0–1 indicate negative; 2 to 3 indicate mild; 4 to 8 indicate moderate; 9 to 12 indicate strongly positive.

Tissues and cells were lysed to extract proteins by RIPA buffer containing phosphatase and protease inhibitors. Thirty micrograms of proteins added to Bis-Tris gel (4%–12%) were separated by the Bio-Rad Transblot system for 90 minutes and then transferred onto a polyvinylidene difluoride (PVDF) membrane. After blocking the membrane in 5% nonfat milk diluted by TBST, we overnight incubated the membrane at 4°C with the primary antibodies: MLYCD (Abcam, ab234879, 1:1,000); HIF1α (ProteinTech, 66730–1-Ig, 1:1,000); HIF2α (Abcam, ab109616, 1:1,000); VHL (ProteinTech, 24756–1-AP,1:1,000); eIF4G3 (ProteinTech, 11281–1-AP, 1:2,000); p-PERK(Cell Signaling Technology, #3179, 1:1,000); ATF6 (Abcam, ab227830, 1:1,000); p-IER1α (Abcam, ab124945, 1:1,000); Bip (Cell Signaling Technology, #3177S, 1:1,000); p-eIF2α (Cell Signaling Technology, #3398S, 1:1,000); ATF4 (Cell Signaling Technology, # 11815S, 1:1,000); CHOP (Cell Signaling Technology, #2895S, 1:1,000); caspase-3 (Cell Signaling Technology, #9664S, 1:1,000); SCD (Abcam, ab236868, 1:1,000); ALB(Abcam, ab207327, 1:2,000); AQP3 (Abcam catalog no. ab125219, RRID:AB_11000698); PTGS2 (Abcam catalog no. ab179800, RRID:AB_2894871); PLIN2 (Abcam, ab52356, 1:2,000); β-actin (ProteinTech, 81115–1-RR, 1:5,000); GAPDH ((ProteinTech, catalog no. 60004–1-Ig, RRID:AB_2107436). After incubation with the secondary antibody for 1 hour at room temperature, membranes were exposed to an enhanced chemiluminescence substrate.

Total RNA was isolated using TRIzol reagent (Thermo Fisher Scientific) as described previously (5). The RevertAid First Strand cDNA Synthesis Kit (TaKaRa) was used according to the manufacturer's protocol for the reverse transcription of cDNA. Quantitative real-time PCR was performed using SYBR Green mix (TaKaRa) in the StepOne Plus Real-Time PCR System. The gene primers were as follows: MLYCD (forward, 5′-ACCTAGAACGGGTTACCTGG-3′; reverse, 5′-CAGGGGTCGAACAGTGAGAA-3′); eIF4G3-microexons (forward, 5′- CAACCTCAAACCCGTTCTCC-3′; reverse, 5′- TGGTAGCTCTAGGAGGCTG-3′); GAPDH (forward, 5′- ACAACTTTGGTATCGTGGAAGG-3′; reverse, 5′- GCCATCACGCCACAGTTTC-3′). Gene expression was calculated using the comparative method (2^−ΔΔC_t).

Overexpression lentivirus for MLYCD, sgRNA for MLYCD, VHL and eIF4G3 microexons, and the paired control vector were provided by Genechem. Expression plasmids for HIF1α and Flag-HIF2α, knockdown plasmids (shRNA) for HIF1α and Flag-HIF2α, and the paired control vector were purchased from GeneChem. Transduction was performed according to the manufacturer's protocol. Lentivirus-transduced RCC cells were treated with 2 μg/mL puromycin to obtain stably overexpressed or knockout cell lines, followed by Western blotting to validate the overexpression or knockout efficiency.

Cell proliferation was evaluated using the CCK-8 and colony formation assays. For CCK-8 assay, 100 μL of 3 × 10³ cells per well was seeded in 96-well plates and cultured in 5% CO₂ at 37°C. Cell counts were measured from day 2 to day 5 using a CCK-8 kit (DOJINDO) according to the manufacturer's instructed protocol. For colony formation assay, 2 mL of 1 × 10³ cells per well was seeded in 6-well plates and cultured in a humidified atmosphere of 5% CO₂ at 37 °C for 15 days, during which the medium was changed every 3 days. Then, pictures were taken and the colonies were counted.

To evaluate the migration and invasion of the cells, we carried out the Transwell assay as described previously (5). After randomly selecting five fields, we counted the number of cells for analysis.

We collected the cells and stored at −80°C for the absolute quantitative lipidomics. The detection and analysis of absolute quantitative lipidomics were performed by Shanghai Applied Protein Technology Co., Ltd. Cell samples were separated using a UHPLC Nexera LC-30A Ultra-high Performance Liquid Chromatography System, followed by mass spectrometry analysis using a Q Exactive series mass spectrometer (Thermo Fisher Scientific). Lipid identification and data processing were performed to obtain the absolute content of lipid species in the samples using LipidSearch software (Thermo Fisher Scientific) and an isotope internal standard of 13 lipid molecules (16).

The cells with stably overexpressing MLYCD and the control were collected and stored at −80°C for the RNA sequencing. RNA sequencing was executed by Shanghai Applied Protein Technology Co., Ltd. Then used the DESeq2 (DESeq, RRID:SCR_000154) R package to differential expression analysis, followed by identification of differentially expressed genes (DEG) based on |LogFC| > 1 and P < 0.05.

Malonyl-CoA and acetyl-CoA content was measured using the malonyl coenzyme A ELISA Kit (CUSABIO) and acetyl-CoA assay kit (Solarbio BC0980) according to the manufacturer's instructions. The FFA, T-CHO or TG content was measured using the nonesterified Free Fatty Acid Assay Kit (Nanjing Bioengineering), the Total Cholesterol Assay Kit (Nanjing Bioengineering) or the TG assay kit (Nanjing Bioengineering), according to the manufacturer's protocol. We also measured the lipogenesis by using D₂O as previously described (17).

CPT1A activity was measured using a previously described protocol (18). Briefly, we extracted 20 μg of mitochondrial protein from the cell fragments. L-[methyl-3H] carnitine and palmitoyl-CoA act as substrates. The measurement assay of enzyme activity was carried out in a volume of 200 mL for 4 minutes at 30°C.

LDs were detected by Oil Red O staining or BODIPY staining. Oil Red O staining was performed using the Oil Red O staining kit (Beyotime Biotechnology) according to the manufacturer's instructions, followed by measuring the size and counting the number of LDs using ImageJ (ImageJ, RRID:SCR_003070). For BODIPY staining, the cells or tissues were fixed in 4% paraformaldehyde solution for 10 minutes, and then incubated with BODIPY 493/503 (MedChemExpress) for 30 minutes and DAPI for 10 minutes in the dark. The corrected total cell fluorescence (CTCF) of BODIPY was quantified using ImageJ (RRID:SCR_003070).

After incubation with 10 μmol/L fluorescent probes 2′,7′-dichlorodihydrofluorescein diacetate (H2DCFDA; Beyotime Biotechnology) in 1 mL of DMEM without FBS at 37°C for 30 minutes, the cells were collected and detected the level of total intracellular ROS by a flow cytometer (Beckman CytoFLEX S). To test the level of lipid-derived ROS, the cells were treated with 1 μmol/L BODIPY 665/676 (Thermo Fisher Scientific) in 1 mL DMEM without FBS at 37°C for 30 minutes. After costaining with Hoechst or MitoTracker Green probes for 30 minutes, confocal imaging was performed.

The oxygen consumption rate (OCR) was determined using a Seahorse XFe 24 Extracellular Flux Bioanalyzer (Agilent) according to the manufacturer's protocol.

After washing with cold PBS, the cells were fixed overnight in 2.5% phosphate-buffered glutaraldehyde (Servicebio) at 4°C. Cells were postfixed in 1% phosphate-buffered osmium tetroxide for 2 hours at room temperature before being embedded in EMbed-812. Cells were stained with lead citrate and uranyl acetate, followed by scanning using an H-7650 transmission electron microscope (Hitachi).

We performed ER tracker staining to evaluate the level of ER stress using an ER-Tracker Red Kit (Beyotime) according to the manufacturer's instructions. The Mito-Tracker Deep Red FM probe (Beyotime) or Mito-Tracker Green probe (Beyotime) was used to detect the morphology of mitochondria according to the manufacturer's instructions.

Chromatin immunoprecipitation (ChIP) was performed as previously described (19). We used the HIF2α antibody for immunoprecipitation and IgG as a negative control.

We executed the luciferase reporter assays according to the previous methods (20). The HIF2α overexpression and blank vector plasmids, eIF4G3 promoter (eIF4G3-WT), and mutant (eIF4G3-MUT) vector plasmids were purchased from GeneChem. Luciferase activity was detected using a Dual-Luciferase Assay Kit (Promega), according to the manufacturer's protocol. Renilla luciferase was used for normalization.

The data of the genes expression and the patient's clinical data in The Cancer Genome Atlas (TCGA)-Kidney Renal Clear Cell Carcinoma (KIRC) cohort were downloaded from TCGA database (https://www.cbioportal.org/). Based on |LogFC| > 1 and P < 0.05, we used the DESeq (RRID:SCR_000154) to identify DEGs. Gene Ontology (GO) and Kyoto Encyclopedia of Genes and Genomes (KEGG) analysis were executed via the R “clusterProfiler” (Version3.14.3). Gene set enrichment analysis (RRID:SCR_003199) was implemented to identify related pathways, according to P < 0.05, and FDR value < 0.25. Integration analysis of transcriptome and lipidomics was performed using Spearman correlation analysis to calculate the correlation coefficients between significant DGEs and significant differentially differentiated lipids. All statistical analyses were performed using GraphPad Prism (RRID:SCR_002798) or R software (Version 4.1.2). Data on continuous variables between the two subgroups were analyzed using independent t tests or one-way ANOVA. The Kaplan–Meier plot was used to show the overall survival of patients with RCC and was analyzed using the log-rank test. On the basis of P values <0.05, we identified the difference as statistically significant.

RNA sequencing data generated in this study are publicly available in the SRA database at PRJNA1014202 (https://www.ncbi.nlm.nih.gov/sra/?term=PRJNA1014202) and PRJNA1014200 (https://www.ncbi.nlm.nih.gov/sra/?term=PRJNA1014200). The lipidomic date generated in this article are available in Zenodo (https://www.zenodo.org/record/8347199). The publicly available TCGA datasets analyzed in this study were obtained from cBioPortal (http://www.cbioportal.org/). All other raw data are available upon request from the corresponding author.

To assess the importance of FA anabolism in tumor progression, we first performed single-sample GSEA (ssGSEA) using key gene signatures of FA catabolism and FA biosynthesis in the TCGA-KIRC cohort. Compared with normal tissues, the FA catabolic scores in ccRCC tissues were lower, while the FA synthesis scores were the opposite (Fig. 1A), suggesting that the FA anabolism pattern may be the increased FA synthesis and decreased catabolism in renal cancer tumors, which is consistent with previous research (3). The ssGESA score was used to categorize the patient into a low (below the 50th percentile) or high (above the 50th percentile) subgroup. Kaplan–Meier analysis demonstrated that patients with a lower FA catabolism score had worse overall survival (OS) time, while a higher FA biosynthesis score indicated shorter OS times (Supplementary Fig. S1A). Moreover, the worst outcomes were observed in tumors with low FA catabolism and high FA synthesis score (Q3 as reference; Fig. 1B), indicating that remodeling the FA anabolism, which means preventing de novo FA synthesis while simultaneously increasing FA consumption, may be a promising therapeutic strategy for patients with ccRCC.

Next, we identified the key genes involved in FA anabolism. By the intersection of FA biosynthesis, acetyl-CoA metabolism, and acyl-CoA metabolism–related gene sets, one gene (MLYCD) was selected (Fig. 1C). We found that patients with low MLYCD expression (below the 50th percentile) had shorter OS than those with high MLYCD expression (above the 50th percentile) in the TCGA_KIRC cohort and TCGA_KIPC cohort, while there was no statistical significance between the OS time of patients with low MLYCD expression (below the 50th percentile) and the OS time of patients with high MLYCD expression (above the 50th percentile) in the TCGA_KIHC cohort (Supplementary Fig. S1B). Meanwhile, according to the Human Protein Atlas (https://www.proteinatlas.org/), the protein level of MLYCD expression was significantly lower in RCC than in normal kidney tissues (Supplementary Fig. S1C). To investigate the role of MLYCD, we used ccRCC tissues to examine MLYCD protein expression. As expected, MLYCD protein levels were conspicuously lower in ccRCC tissues than in normal tissues (Fig. 1D and E). In addition, the MLYCD^low group of patients with ccRCC showed significantly worse survival than the MLYCD^high group of patients with RCC (Fig. 1E). We also observed that all RCC cell lines (ACHN, 786-O, 769-P, CAKI-1, A498, and OS-RC-2) exhibited lower MLYCD expression compared with the normal renal cell line (HK-2; Fig. 1F). Together, these results indicate that lower MLYCD expression could predict worse prognosis in patients with ccRCC.

To prove our hypothesis that MLYCD plays an important role in ccRCC progression, we first performed GSEA and found that MLYCD expression was negatively correlated with cell-cycle and metastasis pathways (Supplementary Fig. S1D). To investigate the effect of MLYCD on ccRCC proliferation and metastasis, we restored and knocked out MLYCD expression in ccRCC cell lines (OS-RC-2 and 786-O) using a lentivirus and sgRNA (CRISPR-Cas9), respectively (Supplementary Fig. 1E). The results of CCK-8, colony formation, and transwell assays showed that OS-RC-2 and 786-O cells stably overexpressing MLYCD visibly restrained cell proliferation, the percentage of S-phase cells, invasion, and migration (Fig. 1G–J; Supplementary Fig. S1F), while reduced expression of MLYCD was the opposite (Supplementary Fig. S1H–S1K), suggesting a progression-inhibiting role of MLYCD in ccRCC progression. Moreover, to evaluate the toxicity of MLYCD in normal cells, we used the overexpression lentivirus to construct the normal cell line HK-2 with stably overexpression of MLYCD, and the result of CCK-8 assay showed that MLYCD overexpression has little effect on the cell activity of HK-2 (Supplementary Fig. 1G); using the sgRNA (CRISPR-Cas9) of MLYCD, we knocked out the MLYCD expression in normal kidney cell line RPTEC and found that reduced expression of MLYCD promoted cell proliferation (Supplementary Fig. S1L).

Next, we determined the function of MLYCD in regulating biological processes in RCC cells. GSEA demonstrated that the pathways involved in lipid raft and lipid metabolism (especially FA metabolism) were significantly associated with MLYCD expression in the TCGA_KIRC cohort (Supplementary Fig. S2A). We observed that restoring the expression of MLYCD dramatically decreased the size of LDs, whereas knockout of MLYCD in 786-O and OS-RC-2 cells increased the size of LDs by Oil Red O staining (Fig. 2A), indicating that MLYCD inhibited intracellular LDs accumulation in ccRCC. Meanwhile, the BODIPY staining assay showed that knockout of MLYCD promoted the LD accumulation in RPTEC cells (Supplementary Fig. S1M).

We then performed lipidomic analysis in RCC cells overexpressing MLYCD and in the control groups to examine the effect of MLYCD on lipid homeostasis. The representative lipid profile of OC-RC-2 cells with altered MLYCD expression is shown in Fig. 2B. We found that, as expected, triglycerides (TG) and diglycerides (DG), which were synthesized from FAs as substrates, were significantly reduced in cells stably overexpressing MLYCD (Fig. 2C). Meanwhile, acylcarnitine CoA (ACCA), which was transformed from FAs by carnitine palmitoyltransferase 1 (CPT1) catalysis and transported into the mitochondria for oxidation (6), was obviously elevated (Fig. 2D). Besides, we observed that ceramides (Cer) were increased, along with sphingomyelin (SM) reduction, in MLYCD restoration (Fig. 2E). The levels of some structural lipids, such as phosphatidylinositol (PI), lysophosphatidylinositol (LPI), phosphatidylglycerol (PG), and lysophosphatidylglycerol (LPG), declined upon MLYCD overexpression. In contrast, the intracellular content of lipid species in phosphatidylcholine (PC), phosphatidylethanolamine (PE), phosphatidylserine (PS), lysophosphatidylcholine (LPC), lysophosphatidylethanolamine (LPE), and lysophosphatidylserine (LPS) was significantly elevated, accompanied by MLYCD overexpression (Fig. 2F and G).

To investigate the regulatory mechanisms of MLYCD function in RCC, RNA sequencing was carried out in OS-RC-2 cells with stable MLYCD overexpression and control cells. We identified 844 DEGs associated with MLYCD expression in RCC (Fig. 2H). Gene Ontology (GO) analysis demonstrated that the DEGs were mostly enriched in multiple biological processes, especially FA biosynthesis, FA catabolism, lipid metabolism, and lipid homeostasis (Fig. 2I). Combined transcriptomic and lipidomic analyses showed that MLYCD affected the content of various intracellular lipid species by mediating the expression of multiple genes (such as CYP1A1 and GRIA2, etc.) expression (Fig. 2J).

Altogether, these results demonstrated that restoring MLYCD expression blocked TG and LD formation and shifted FAs to mitochondria and structural lipids, suggesting disruption of lipid homeostasis in RCC cells.

It is known that MLYCD regulates malonyl-CoA content through decarboxylation of malonyl-CoA back to acetyl-CoA (21). Malonyl-CoA is not only the substrate of FA synthesis but also inhibits the enzymatic activity of CPT1A, which is the rate-limiting step in FA oxidation. As shown in Fig. 3A, MLYCD acts as a metabolic switch between de novo FA biosynthesis and catabolism.

Here, we examined the influence of MLYCD on de novo FA synthesis. As expected, we found that compared with the control cells, the RCC cells stably overexpressing MLYCD had significantly lower levels of malonyl-CoA, free FAs (FFA), and TG, while the opposite was observed in cells with MLYCD knockout (Fig. 3B and C). We also observed that restored MLYCD expression increased the content of Acetyl CoA, but there was a slight statistical difference in the content of cholesterol in cells (Supplementary Fig. S2B). We next demonstrated the inhibition of MLYCD in de novo FA synthesis and undergo lipogenesis, by measuring [¹⁴C]-acetate incorporation into total lipids (Fig. 3D). Next, we treated stably overexpressing MLYCD cells with malonyl-CoA. The percentage of [¹⁴C]-acetate incorporation into total lipids was relatively decreased in RCC cells with MLYCD restoration, whereas the percentage was not significant in the control/MLYCD-overexpressing RCC cells with malonyl CoA treatment (Fig. 3E), indicating that malonyl CoA treatment abolished the effects of MLYCD on the repression of de novo FA biosynthesis in RCC cells. Bodipy staining showed a similar result, which was performed to characterize LD accumulation (Fig. 3F; Supplementary Fig. S2C). Moreover, the results of the CCK-8 assay indicated that malonyl-CoA almost completely abolished the inhibition of MLYCD on the proliferation of OS-RC-2 and 786-O cells (Fig. 3G; Supplementary Fig. S2D). We also found that palmitate, as the exogenous lipid supplementation, could rescue the inhibitory effect of overexpression of MLYCD (Supplementary Fig. 2E).

Given that LDs have physiologic and functional connections with the endoplasmic reticulum (ER), lipid storage is necessary for ER homeostasis. Once lipid accumulation is destroyed, ER stress is inevitable, which triggers cell death by the lethal unfolded protein response (UPR, ref. 22). As expected, we observed ER expansion in MLYCD-overexpressing cells using ER Track imaging (Fig. 3H; Supplementary Fig. S2F). Western blotting demonstrated that restoration of MLYCD increased the expression of ER stress markers, while knockout of MLYCD had the opposite effect (Fig. 3I; Supplementary S2G and S2H). In detail, PERK/ATF4/CHOP/Caspase 3 pathway, as the transcriptional program of ER pathway, was activated by MLYCD restoration. To determine whether aggrandizing ER stress contributes to the MLYCD-induced repression of ccRCC cell activity, we blocked ER stress by treatment with 4-phenylbutyric acid (4-PBA), an inhibitor of ER stress. Western blot assay was performed to detect the expression of ER stress markers after 4-PBA treatment. As shown in Supplementary Fig. S2I and S2J, the results showed that 4-PBA represses all ER stress branches, which is consistent with previous literature (23). The results of the CCK-8 assay showed that this inhibitor partially rescued the repression of cell activity induced by MLYCD overexpression (Fig. 3J).

Considering that ER stress inhibitors only partially alleviated the inhibitory effect of MLYCD on cell viability, we next investigated additional mechanisms of MLYCD in ccRCC. The results of deeply GSEA demonstrated that MLYCD was significantly associated with the mitochondrial membrane, FA oxidation, and biosynthesis of unsaturated FAs and peroxisomal lipid metabolism–related pathways (Supplementary Fig. S3A). Malonyl-CoA acts as an inhibitor of CPT1A, which plays an important role in the transport of long-chain FAs, in the form of acylcarnitine, across the mitochondrial membrane for FA oxidation. Therefore, we hypothesized that, in addition to FA synthesis, MLYCD affects FA catabolism by mediating CPT1A activity.

To test our hypothesis, we first examined CPT1A activity in cells with stably overexpressed/knockout MLYCD and found that restored MLYCD expression strongly increased CPT1A protein activity, while knockout MLYCD expression had the opposite effect (Fig. 4A). Moreover, culturing cells with malonyl-CoA suppressed the activity of CPT1A in MLYCD-overexpressing cells (Supplementary Fig. S3B). Next, we measured the OCR of cells to examine whether MLYCD levels affected mitochondrial activity. The results showed that elevation of MLYCD expression dramatically reduced the OCR compared with the control cells, while downregulation of MLYCD expression showed an opposite trend (Fig. 4B; Supplementary Fig. S3C). We used transmission electron microscopy (TEM) to observe ECC cellular morphology and found that the mitochondria of the MLYCD-overexpressing RCC cells lost cristae and became fragmented and round, unlike the long tubular mitochondria in the control cells (Fig. 4C). The fluorescence imaging of MitoTracker Red staining was in accordance with the TEM findings, while there was no distinct change in mitochondrial morphology in the control or MLYCD-overexpressing cells treated with malonyl CoA (Fig. 4D). In addition, restoration of MLYCD expression markedly decreased mitochondrial membrane potential (Supplementary Fig. S3D). But treatment with palmitate treatment, we found that lipids supplementation could partially rescue mitochondrial morphology crisis (Supplementary Fig. S3E). Previous study has reported that the disruption of lipid homeostasis and oxidative stress impaired the mitochondrial structure (24). These data demonstrate that the seriously destroyed mitochondrial function and structure are due to disruption of lipid homeostasis and oxidative stress by malonyl-CoA/CPT1A regulation.

Next, we investigated the influence of MLYCD on the production of reactive oxygen species (ROS) in ccRCC cells. Total intracellular ROS were detected using fluorescent probes 2′,7′-dichlorodihydrofluorescein diacetate (DCFH-DA). The results demonstrated that ROS was significantly increased upon MLYCD restoration in ccRCC cells but decreased upon MLYCD knockout (Fig. 4E; Supplementary Fig. S3F), indicating that MLYCD regulated oxidative stress. BODIPY 665/676 probes were used to detect lipid-derived ROS. Imaging showed that lipid-derived ROS levels were dramatically increased in MLYCD-overexpressing cells (Fig. 4F). Meanwhile, we observed that the raised lipid-derived ROS was localized to the mitochondria (Fig. 4F), indicating that the elevated ROS in MLYCD restoration was produced in the mitochondria, compared with the control cells. After treatment with ETO, a CPT1A inhibitor, the effect of MLYCD restoration on ROS elevation disappeared (Fig. 4F). The results of the CCK-8 assay demonstrated that ETO significantly recovered ccRCC cell activity, which was suppressed by MLYCD overexpression (Fig. 4G).

Ferroptosis is a form of cell death, the key mechanism of which is the iron-dependent accumulation of lipid peroxides and lipid-derived ROS (25). Polyunsaturated FAs (PUFA) are highly susceptible to peroxidation (25). Given the effect of MLYCD on peroxisomal lipid metabolism, biosynthesis of unsaturated FAs (Supplementary Fig. S3A), and lipid-derived ROS production (Fig. 4F), we hypothesized that restoration of MLYCD expression increased the sensitization of ccRCC cells to ferroptosis. Lipidomic analysis showed that the content of PUFAs were dramatically increased in MLYCD-overexpressing cells compared with the control cells, although there was no significant difference in monounsaturated FAs (MUFA; Fig. 4H). We also observed that the content of malondialdehyde, an indicator of the degree of lipid peroxidation, was obviously increased with MLYCD restoration, which was the opposite in cells with MLYCD knockout, compared with the control cells (Fig. 4H; Supplementary Fig. S3G). We observed that knockout of MLYCD expression moderately relieved cell death induced by erastin (a ferroptosis inducer), while Ferr-1, a ferroptosis inhibitor, partly restored RCC cell activity, which was suppressed by MLYCD overexpression (Fig. 4I and J). RNA sequencing and Western blotting demonstrated that MLYCD restoration strongly upregulated the expression of the ferroptosis-marker genes (PTGS2, ALB, AQP3) and downregulated the expression of the ferroptosis-suppressor gene (SCD1); however, there was an opposite trend in cells with MLYCD knockout, compared with the control cells (Fig. 4K; Supplementary Fig. S3H and S3I), indicating that MLYCD induced ferroptosis in RCC cells. Furthermore, after treatment with NAC, a ROS scavenger, we found that the effect of MLYCD restoration on the elevation of lipid-derived ROS disappeared (Fig. 4L). The results of the CCK-8 assay demonstrated that NAC significantly restored ccRCC cell activity, which was suppressed by MLYCD overexpression (Fig. 4M). Together, these data strongly demonstrated that MLYCD induces mitochondrial damage and ROS elevation to inhibit ccRCC cell activity.

It is well known that the typical alteration in ccRCC is the von Hippel-Lindau (VHL; an E3 ubiquitin ligase) inactivation, which lead to improve the activity of HIF1 and HIF2 by the stabilization of the HIFα subunits (26). It has been reported that HIF is the major mediator of VHL-dependent functions in metabolic and energy homeostasis (7). Therefore, we were justified in wondering that the MLYCD expression was mediated by HIF. The VHL and HIF status were detected in the RCC tissues and the RCC cell lines. We observed that ACHN and Caki-1 were VHL wild-type, while A498, OS-RC-2, 786-O and 769-P were VHL mutant; the expression of HIF1α and/or HIF2α protein in RCC cell lines was higher than the normal renal cell line HK-2 (Supplementary Fig. S4A). Consistent with previous reports (26–28), VHL absence existed in most of ccRCC, but the HIF1α expression was slightly lower in the RCC tissues than the normal kidney tissues; the HIF2α expression was significantly higher in the RCC tissues than the normal tissues (Supplementary Fig. S4B). We upregulated or downregulated HIF1α and HIF2α, respectively, in ccRCC cells. We found that knockdown of HIF1α or HIF2α expression increased the protein level of MLYCD in RCC cells (Fig. 5A and B), while both HIF1α and HIF2α overexpression significantly decreased the protein level of MLYCD (Supplementary Fig. S4C and S4D). However, the regulation of HIF2α in MLYCD expression was more effective than HIF1α. To delineate the regulatory mechanism of HIF in MLYCD expression, HIF2α was selected as a representative. Unexpectedly, compared with the control cells, there was no statistical difference in the mRNA level of MLYCD after downregulation of HIF2α in 786-O and OS-RC-2 or upregulation of HIF2α in 769-P (where HIF is not expressed) cells (Fig. 5C; Supplementary Fig. S4E), indicating that HIF2α regulated MLYCD expression via posttranscriptional pathway. Posttranscriptional regulation is primarily involved in translation and protein stability. Accordingly, to explore whether HIF2α regulates MLYCD expression by mediating MLYCD protein stability, we cultured HIF2α-overexpressed cells with a lysosome inhibitor (chloroquine) or a proteasome inhibitor (MG132). The results of Western blotting showed that there was no change in the expression of MLYCD induced by HIF2α (Fig. 5D), suggesting that HIF2α did not affect the stability of MLYCD protein. These results led us to hypothesize that HIF reduced MLYCD protein levels by inhibiting mRNA translation. Using RNA sequencing of HIF2α stable knockdown cells and three pairs of ccRCC tissues with paired normal tissues from our previous studies (19), we screened and obtained a translation-related gene, eIF4G3, which was correlated with HIF2α and differently expressed in ccRCC tissues (Fig. 5E; Supplementary Table S1), indicating that HIF2α mediates gene translation in RCC by regulating the expression of eIF4G3. We found that the expression of HIF2α was significantly positively correlated with that of eIF4G3 at the mRNA level (Fig. 5F).

Recent research has demonstrated that eIF4G microexons, as translational brakes, induce ribosome stalling and inhibit the expression of critical synaptic proteins (29). We reasoned that HIF2α inhibits the translation of MLYCD by promoting eIF4G3 microexon expression. Real-time PCR assays demonstrated that eIF4G3 microexons expression was higher in the RCC cell lines than the normal kidney cell line HIK-2 (Supplementary Fig. S4F), and the eIF4G3 microexons are dynamically regulated along with HIF2α expression levels (Fig. 5G; Supplementary Fig. S4G). We then predicted the position of the putative HIF2α binding motif in the −2,000-bp human eIF4G3 promoter (Fig. 5H). We performed a ChIP assay and confirmed that there was strong enrichment between the HIF2α antibody and position E2 (Fig. 5I). Luciferase reporter assays demonstrated that the luciferase activity of the vector with the wild-type (WT) eIF4G3 promoter was promoted by HIF2α overexpression in cells (Fig. 5J; Supplementary Fig. S4H). Furthermore, the Western blot assay demonstrated that compared with the control cells, HIF2α overexpression increased eIF4G3 expression at the protein level, and knockdown of HIF2α expression decreased eIF4G3 expression (Fig. 5K; Supplementary Fig. S4I). In addition, we stably knocked down HIF2α with simultaneous depletion of eIF4G3 microexons in ccRCC cells and found that eIF4G3 microexon depletion eliminated the effect of HIF2α on MLYCD expression (Fig. 5K). We also observed that reduced expression of eIF4G3 microexons inhibited the cell proliferation, and decreased the size of LDs by CCK-8 and BODIPY staining assay (Supplementary Fig. S4J and S4K), indicating that deletion the expression of eIF4G3 microexons could mimic the effect of MLYCD overexpression. These data suggest that HIF2α suppresses MLYCD protein levels by mediating eIF4G microexons.

Recent studies have reported that HIF drives lipid accumulation in RCC by regulating LDs surface protein PLIN2-mediated lipid storage and CPT1A-mediated FA oxidation (6, 22). Given these results, we speculated that HIF2α regulates FA anabolism in ccRCC via MLYCD, mediating de novo FA synthesis and FA oxidation. To test our hypothesis, we constructed RCC cells with stable knockdown of HIF2α and/or MLYCD (Fig. 6A; Supplementary Fig. S5A). We found that reduced expression of HIF2α led to significantly decreased content of malonyl-CoA and rate of de novo FA biosynthesis, but increased the activity of CPT1a in ccRCC. However, compared with the cells with stably knocked out MLYCD expression, the content of malonyl-CoA, the rate of de novo FA synthesis, and the activity of CPT1a were scarcely significant in RCC cells with stable knockdown of HIF2α expression and knockout MLYCD expression (Fig. 6B). To rule out that shHIF2a may also decrease ¹⁴C-acetate-incorporated lipogenesis via inactivating ACSS2, we measured the lipogenesis in shHIF2α cells with or without sgMLYCD by using D₂O as previously described (17). The result showed that compared with the control RCC cells, the rate of de novo FA synthesis was decreased in the shHIF2α cells, while knocking out MLYCD expression in the shHIF2α cells recused this effect of knocking down HIF2α (Supplementary Fig. S5B). We also observed that the contents of FFA and TG were distinctly attenuated in the cells with stable knockdown of HIF2α expression compared with the control cells, while there was a slight decrease in the content of FFA and TG in RCC cells with stable knockdown of HIF2α expression and knockout MLYCD expression compared with the RCC cells with stably knocked out MLYCD expression (Fig. 6C). These results indicated that HIF2α mediated FA anabolism in an MLYCD-dependent manner. Compared with the control cells, the fluorescence intensity of BODIPY was lower in HIF2α-knockdown RCC cells, while MLYCD-knockout RCC cells with or without knockdown HIF2α expression showed little difference in the fluorescence intensity of BODIPY (Fig. 6D; Supplementary Fig. S5C), indicating that knockout MLYCD expression significantly eliminated the effects of HIF2α on LD storage in RCC cells. Similar results were found by CCK-8 and Transwell assays (Fig. 6E and F; Supplementary Fig. S5D), indicating that HIF2α regulated RCC cell activity by mediating MLYCD expression.

In terms of morphology, ultrastructural analysis by TEM showed the presence of irregularly shaped rough and dilated ER, which was consistent with ER stress, and fragmented mitochondria that lost cristae in HIF2α-knockdown RCC cells, while knocked-out MLYCD expression inhibited these effects of HIF2α (Fig. 6G). As expected, downregulation of HIF2α expression was significantly induced in activating the ER stress pathway, while there was a slight elevation in the expression of ER stress biomarkers in RCC cells with stably knocked down HIF2α expression and knockout MLYCD expression compared with RCC cells with stably knocked out MLYCD expression (Supplementary Fig. S5E). Moreover, we observed that the RCC cells with downregulation of HIF2α expression increased the OCR compared with the control cells, but the change in downregulating HIF2α on OCR was modestly eliminated by MLYCD knockout (Fig. 6H), indicating that HIF2α mediates mitochondrial oxidative stress in an MLYCD-dependent manner. HIF2α knockdown resulted in a significant increase in lipid-derived ROS, lipid peroxidation levels, and sensitivity to erastin-induced ferroptosis (Fig. 6I–J; Supplementary Fig. S5F). However, these changes were attenuated by MLYCD knockout in HIF2α knockdown cells (Fig. 6I–J; Supplementary Fig. S5F). Similar results were obtained for the expression of ferroptosis markers (Fig. 6K; Supplementary Fig. S5G). Besides, we found that the level of PLIN2 expression was lower in the cell with HIF2α knockdown than in the control cells, while there was little difference in the MLYCD-knockout RCC cells with or without knockdown HIF2α expression (Fig. 6K and Supplementary Fig. S5G). Taken together, these results demonstrate that MLYCD is required for HIF2α-regulated FA anabolism.

In recent decades, multiple agents against VEGF, the mTOR pathway, and AXL tyrosine-protein kinase receptors (TKI), such as sunitinib and sorafenib, have been approved according to their significant activity in RCC; however, due to drug toxicity and tolerability, this disease is still not curative (30). A recent study showed that HIF2a and VEGFA levels were significantly higher in the tissues of patients with RCC treated with sunitinib than in those not receiving adjuvant treatment, and high HIF2α protein levels were associated with resistance to sunitinib (31). Iwamoto and colleagues demonstrated that lipid metabolism in cancer cells confers antiangiogenic drug resistance in colon cancer (32). Thus, we assumed that HIF2α-mediating MLYCD expression regulation of FA anabolism by MLYCD may be a potential mechanism for resistance to TKIs, which were represented by sunitinib in RCC. To test our hypothesis, we obtained sunitinib-resistant RCC cell lines according to a previous study (14). Immunofluorescence (IF) assays showed that during the process of sunitinib resistance, the blood vessel level in the tumor was significantly reduced along with LD accumulation (Fig. 7A). Further research found that the development of sunitinib resistance was accompanied by a decrease in MLYCD expression, malonyl CoA content, and CPT1A enzyme activity and an increase in de novo FA synthesis (Fig. 7B and C; Supplementary Fig. S6A). We then examined the sensitivity of cells at different stages to sunitinib and found that the lower the expression of MLYCD in the RCC cell line, the lower the sensitivity to sunitinib (Supplementary Fig. S6B), indicating an alternative mechanism of resistance to TKIs by regulating FA anabolism. We performed the CCK-8 assay and observed that compared with DMSO treatment, restoration of MLYCD expression increased the sensitivity of 786-O-Su3rd and OS-RC-2-Su3rd cells to sunitinib and reversed drug tolerance (Fig. 7D and E). Moreover, compared with control cells, reduced MLYCD expression increased the tolerance of RCC cells to sunitinib (Fig. 7F). Moreover, Western blot assay was carried out to detect the RTKs expression upon MLYCD overexpression. The result showed that the expression of PDGFRβ was signification decrease in MLYCD overexpression compared with the control cells (Supplementary Fig. S6C). Therefore, we concluded that MLYCD restoration reversed sunitinib resistance in RCC in vitro.

Next, we constructed a subcutaneous xenograft model to evaluate the function of MLYCD in RCC. The same number of OS-RC-2-Su3rd/786-O-Su3rd cells, with or without stable overexpression of MLYCD, were injected subcutaneously into each mouse. The mice were fed sunitinib or vehicle daily from days 15 to 40. There was no significant difference in tumor growth and weight between the control subgroup and sunitinib treatment alone, whereas restoration of MLYCD expression dramatically suppressed tumor growth, and both the volume and weight of tumors in the overexpressed MLYCD subgroup mice fed sunitinib were the lowest, demonstrating that MLYCD restored RCC sensitivity to sunitinib (Fig. 7G–I; Supplementary Fig. S6D–S6F). Next, we performed IHC and IF staining assays, and the results showed that compared with the control subgroup, there was no statistic difference in sunitinib treatment alone on Ki67 expression, while overexpression of MLYCD decreased the expression of KI67, and the level of Ki67 expression was lowest in the overexpressed MLYCD subgroup with sunitinib treatment (Fig. 7J; Supplementary Fig. S6G).We also observed that enhanced MLYCD expression blocked lipid deposition and increased the expression of PERK, PTGS2 and caspase-3 (Fig. 7J; Supplementary Fig. S6G). Altogether, these results demonstrate that MLYCD restores the sensitivity of RCC to sunitinib in vivo.

To evaluate the role of endogenous MLYCD, we generated kidney proximal tubule–specific Mlycd-knockout mice (Ggt1-Cre⁺; Mlycd^Loxp/Loxp) by crossbreeding the gamma-glutamine transferase (Ggt1) promoter–driven Cre mice with Mlycd^Loxp/Loxp mice to delete Mlycd in the proximal tubular cells. IHC staining showed that the expression of Mlycd in the kidneys of conditional Mlycd-knockout mice was approximately 50% lower than that in the kidneys of wild-type mice (Ggt1-Cre⁻; Mlycd^Loxp/Loxp; Fig. 8A). Although there was no microscopic malignant RCC, we observed occasional areas of tubular epithelial disorganization (Fig. 8B). IHC staining showed that Mlycd knockout significantly increased the levels of Ki67 and RCC biomarkers (Vimentin, Ca-ix, and Rcc; Fig. 8C; Supplementary Fig. S7A). To understand the implications of endogenous Mlycd, we performed RNA sequencing followed by DEG analysis using kidneys from conditional Mlycd-knockout mice and wild-type mice. We identified 520 DEGs associated with endogenous Mlycd knockout (Fig. 8D). GSEA and GO analysis for the 520 DEGs demonstrated that several oncogenic pathways, including immune response, lipid metabolic process, and carboxylic acid catabolic process, were significantly regulated in the kidneys of Mlycd-CKO mice compared with the control mice, which were also regulated in RCC tissues compared with normal kidney tissues of TCGA-KIRC cohorts (Fig. 8E), suggesting that kidney proximal tubule–specific Mlycd-knockout mice phenocopied the status of some pathways associated with ccRCC. Moreover, we observed that detection Mlycd expression increased LDs accumulation and the content of malonyl-CoA, FA, and TG, but decreased the activity of Cpt1a (Fig. 8F and G), indicating that Mlycd detection was responsible for the regulation of FA anabolism in the kidney. IHC staining showed that Mlycd knockout significantly decreased the levels of Plin2, p-Perk, Ptgs2, Scd, and Caspase-3 (Fig. 8H). Moreover, we generated kidney proximal tubule–specific Mlycd-knockout mice (Ggt1-Cre+; Vhl^Loxp/Loxp) by crossbreeding the Ggt1 promoter–driven Cre mice with Vhl^Loxp/Loxp mice to knockout Vhl expression in the kidney. IHC staining showed that the expression of Vhl and Mlycd in the kidneys of conditional Vhl-knockout mice was observably lower than that in the kidneys of wild-type mice (Ggt1-Cre⁻; Vhl^Loxp/Loxp; Supplementary Fig. S7B). We also found that Vhl loss increased LD accumulation and the content of malonyl-CoA, FA, and TG (Supplementary Fig. S7C and S7D), indicating that VHL loss could mimic MLYCD CKO phenotype. We knocked out VHL expression in ACHN (VHL-null) cells, and constructed VHO KO with/without MLYCD-overexpressing cell lines. We found that VHL loss promoted the cell proliferation, increased the content of FFA and lipid accumulation and tumor growth, but MLYCD overexpression partially eliminated this effect (Supplementary Fig. S7E–S7J). These results demonstrate that the deletion of endogenous Mlycd-mediated FA anabolism may be responsible for ccRCC.

Metabolic reprogramming, a hallmark of cancer, enables cancer cell growth, proliferation, and survival. Intracellular LD accumulation is characteristic of abnormal metabolism in RCC, especially ccRCC. Previous studies have reported that activated FA synthesis, maintenance of LDs homeostasis, or FA catabolic blockage can lead to LDs accumulation and facilitate tumor progression in RCC (6, 22, 33). We previously identified that reducing abnormal lipid accumulation, a phenomenon called tumor cell “slimming,” repressed tumor progression (4), but how to block abnormal lipid accumulation thoroughly has rarely been investigated. In this study, we systematically and comprehensively evaluated the relationship between FA anabolism and RCC prognosis and found that the subtype with high synthesis and low catabolism had the worst prognosis, suggesting that reducing FA synthesis and increasing FA consumption might be a new strategy for RCC treatment. We then identified the role of MLYCD in FA anabolic regulation by mediating malonyl-CoA content, which hindered de novo FA synthesis and boosted FA oxidation. Restoration of MLYCD significantly retarded tumor growth and reversed RCC resistance to sunitinib in vitro and in vivo. We demonstrated that MLYCD induced multistress (including ER stress and oxidative stress) and ferroptosis, ultimately suppressing cancer cell activity. These results provide a strong rationale for the use of MLYCD-mediated FA anabolism in RCC therapy.

Malonyl-CoA decarboxylase (MLYCD, 55 kDa), converts malonyl-CoA to acetyl-CoA and CO₂. Malonyl-CoA is not only a precursor to de novo FA biosynthesis but also a negative regulator of FA oxidation through the inhibition of CPT1 (34). The switch between the synthesis and catabolism of FAs depends on the malonyl-CoA levels (35). It has been recognized that MLYCD, which mediates malonyl-CoA content, participates in many physiologic and pathologic functions, such as appetite control in the hypothalamus, energy selection in muscle tissues based on carbohydrates or fat, and ketogenic responses to fasting and diabetes in the liver (32–36). However, the role of the MLYCD–malonyl-CoA axis in cancer development has rarely been reported. Some studies have reported that inhibition of MLYCD expression induces cytotoxicity and exerts antitumor effects in breast cancer and rhabdomyosarcoma (37–38). Bruning and colleagues demonstrated that FASN silencing increased malonyl-CoA levels, ultimately reducing angiogenesis by inducing malonylation of mTOR (39). Here, we observed that MLYCD deficiency was prevalent in RCC and associated with poor prognosis, indicating that MLYCD may serve as a suppressor gene in RCC. Restoration of MLYCD unambiguously hindered LD deposition and disrupted lipid homeostasis by blocking intratumoral de novo FA synthesis and increasing FA catabolism in RCC. Synthesizing de novo FAs and storing excess them into LDs are frequently advantageous strategy developed by malignant tumors (3). Our results strongly suggest that MLYCD reduces the malonyl-CoA content, the substrate for de novo FAs synthesis, completely blocking this strategy from the source. Besides, our research found that MLYCD has little effects on cholesterol in ccRCC cells, although that restored MLYCD expression promoted the content of Acetyl CoA in cells. Simon and colleagues have demonstrated that ccRCC mainly dependent on exogenous cholesterol while genes encoding cholesterol biosynthetic enzymes are repressed (40). Of note, our data showed that overexpression of MLYCD triggered multiple stresses, including ER stress and oxidative stress. Lipid homeostasis has been demonstrated to be essential for maintaining ER and energy homeostasis (22). We observed that MLYCD restoration induced in increasing ER stress and ceramide content. Although the ER stress inhibitor did not completely reverse the MLYCD-induced decrease in cell viability, we reasonably believe that MLYCD overexpression triggered ER stress–induced cell death. Moreover, our data showed that restoring MLYCD expression significantly upregulated the activity of CPT1A, which can promote the oxidation of FA into the mitochondria, resulting in a dramatically elevating in ROS level. The raised ROS impaired the structure and oxidation functions of the mitochondria, leading to acyl-CoA accumulation and future increased oxidative stress, eventually leading to inevitable cell death. Lipid metabolomics revealed that MLYCD was associated with the rearrangement of a variety of membrane structural phospholipids, suggesting that MLYCD-mediated FA anabolism led to FA redistribution and disruption of lipid homeostasis. Interestingly, we found that MLYCD restoration was accompanied by an obvious increase in unsaturated FAs content, lipid peroxidation levels, and lipid-derived ROS. Ferroptosis, an iron-dependent cell death, is characterized by the lethal accumulation of lipid-derived ROS associated with enhanced lipid peroxidation and PUFA content. Targeting ferroptosis is a hot topic in cancer therapeutic (41). Several recent studies were supported that RCC is sensitive to ferroptosis due to the enrichment of PUFAs (42). Nevertheless, maintaining ROS/lipid–derived ROS homeostasis is essential (43). It is clear from our data that MLYCD restoration represses cell activity by increasing sensitivity to ferroptosis. In addition, growing evidence has emerged to demonstrate that ER stress and excessive accumulation of ROS/oxidative stress usually interact with each other and deeply determine the state and fate of cell (44–45). Therefore, we proposed that only through a certain signaling pathway is not sufficient, MLYCD-mediated FA anabolism dysregulated lipid homeostasis and caused multistress work jointly to synergize with ferroptosis to inhibit RCC cell activity. Moreover, our study highlights the crucial role of endogenous MLYCD in FA anabolism and the development of kidney cancer.

Another interesting finding of our study was that MLYCD expression is regulated in an HIF-dependent, noncanonical manner. HIFs, including HIF1α and HIF2α, transcriptionally upregulate tumorigenic hypoxia-responsive genes (22, 46). Some studies have reported that HIF, acting as a nontypical transcriptional repressor, directly inhibits the transcription of CPT1A, or activates transcriptional repressors, including ZEB1/2, DEC1/2, REST, and Snail, downregulating the expression of downstream genes (6, 47), but our data show that HIF obviously downregulated the expression of MLYCD in protein level, but not in mRNA level. We also excluded the effect of HIF on the stability of MLYCD protein. Besides, our results showed that the regulation of HIF2α in MLYCD expression was more effective than HIF1α, so we selected HIF2α for further investigation. We have the same opinion that ccRCC lipid turnover has been described HIF2α-dependent and not HIF1α. Notably, we identified that HIF repressed MLYCD translation by mediating eIF4G microexons, which have been reported to cause ribosomes to stall by acting as a translational brake (29). Our studies are in no way exhaustive, but this study does provide some insight that the binding of HIF to a translation regulatory factor strongly suggests that the function of protein translation is not nonexistent. Moreover, the usual way that hypoxia/HIF promotes lipid accumulation is through induction of LD storage and catabolism, our study replenished the role of HIF in FA anabolism.

In addition, our data suggest that MLYCD-mediated FA anabolism is involved in the resistance to targeted therapy. It is well known that targeted therapy inhibits angiogenesis via inhibiting VEGF. However, because that vascular density often reduces to an awfully low level, hypoxic environments are created and exacerbated, which can increase the HIF and growth factor levels, change the composition of the cell types, and reprogram metabolic forms, eventually leading to drug resistance by circumventing drug targets in pancreatic cancer and colorectal cancer (32, 48–49). Bensaad and colleagues reported that in response to anti-VEGF therapy and tissue hypoxia, lipid transport and storage were increased in glioblastoma cells (50). Our data showed that, in the process of RCC resistance to TKIs, there were obvious dynamic changes in FA anabolism, accompanied by decreased angiogenesis; resistant RCC cells regained sensitization to sunitinib by restoring MLYCD expression in vitro and in vivo. This study provides new strategies for the treatment of RCC postresistance to targeted therapy.

Our study has several limitations. Although we observed FAs/lipid redistribution and the regulation of lipid-related gene expression, we did not identify the exact effect of certain FAs/lipid changes and the regulated mechanism of lipid-related genes mediated by MLYCD overexpressed in RCC, which will be a focus of future investigation. Clinically, sunitinib-resistant patients are not suitable for surgery to remove the tumor. We could not obtain clinical specimens that were resistant to sunitinib. Therefore, the absolute levels of MLYCD to predict sunitinib responsiveness cannot be assessed at present. In addition, only a few of the conditional Mlycd-knockout mice exhibited tubular morphological changes, and no tumor was formed. Because most tumors cannot be caused by nonsingle gene changes, the role of MLYCD deletion in the origin of renal carcinoma requires further investigation. In addition, mRNA sequencing using conditional Mlycd-knockout and wild-type mice revealed that the absence of endogenous MLYCD might cause changes in the immune response in the microenvironment. The effect of MLYCD-mediated FA anabolism on the TME is another focus of future research. Another limitation of this study is that the development of specific inducers of MLYCD expression is needed, which may be an innovative treatment strategy for RCC.

In summary, our study highlights the crucial role of MLYCD-mediated FA anabolism in RCC development. This study demonstrated that targeting MLYCD-mediated FA anabolism is an effective way to repress RCC tumor growth and revert resistance to TKIs both in vitro and in vivo. These findings suggest the concept of combination therapy including TKIs and regulation of FA anabolism, which would be promising for the effective treatment of RCC.

No disclosures were reported.

L. Zhou: Conceptualization, software, formal analysis, supervision, funding acquisition, validation, investigation, visualization, methodology, writing–original draft, writing–review and editing. Y. Luo: Conceptualization, funding acquisition, investigation, visualization, methodology, writing–original draft. Y. Liu: Conceptualization, funding acquisition, validation, investigation, visualization, methodology, writing–original draft. Y. Zeng: Validation, investigation, visualization, methodology, writing–original draft. J. Tong: Validation, investigation, visualization, methodology, Writing–original draft. M. Li: Software, visualization, methodology, writing–original draft. Y. Hou: Software, visualization, methodology. K. Du: Software, visualization, methodology. Y. Qi: Software, visualization, methodology. W. Pan: Visualization, methodology. Y. Liu: Software, visualization, methodology. R. Wang: Software, validation, visualization. F. Tian: Software, funding acquisition, validation, writing–review and editing. C. Gu: Conceptualization, supervision, funding acquisition, validation, writing–original draft, project administration, writing–review and editing. K. Chen: Conceptualization, supervision, funding acquisition, validation, writing–original draft, Project administration, writing–review and editing.

This work was supported by grants from the National Natural Sciences Foundation of China (no. 82203099 to L. Zhou; no. 82173294 to C. Gu; and no. 82103000 to Y. Liu), the Joint Construction Project between Medical Science and Technology Research Project of Henan Province (no. LHGJ20220335 to L. Zhou), funding for Scientific Research and Innovation Team of The First Affiliated Hospital of Zhengzhou University (no. QNCXTD2023023 to C. Gu), the Training Program for Middle-aged and Young Discipline Leaders of Health of Henan Province (no. HNSWJW-2021004 to C. Gu).

The publication costs of this article were defrayed in part by the payment of publication fees. Therefore, and solely to indicate this fact, this article is hereby marked “advertisement” in accordance with 18 USC section 1734.