SLC13A2-transported citrate remodels transcriptional regulation through protein acetylation to suppress tumor growth

Abstract

Metabolic reprogramming is a hallmark of cancer, while tricarboxylic acid cycle is increasingly recognized as a multifaceted hub driving tumor metabolism and progression. Integrated analysis of solute carrier (SLC) transporters revealed consistent down-regulation of SLC13A2 in hepatocellular carcinoma (HCC) cells and liver tissues from human patients and mouse models. Adeno-associated virus–mediated liver-specific knockout or overexpression of SLC13A2 (SLC13A2-OE) promoted or ameliorated HCC progression, indicating its protective role. SLC13A2 inhibited HCC proliferation by decreasing mitochondrial function via suppressed glycolysis, respiration, and adenosine 5′-triphosphate production. Flux analysis showed that SLC13A2 imported citrate to generate acetyl–coenzyme A for pyruvate kinase isozyme type M2 acetylation, triggering its degradation. Reduced pyruvate kinase activity limited pyruvate supply, impairing amino acid synthesis and nucleotide metabolism. Moreover, SLC13A2-imported citrate induced intracellular protein acetylation, particularly histone proteins, which provided an epigenetic basis for transcriptional regulation and contributed to tumor suppression. Thus, SLC13A2 perturbs metabolic and transcriptional programs to suppress tumor growth, highlighting potential drug targets for HCC therapy.

SLC13A2-imported citrate induces protein acetylation that connects metabolism and epigenetics to suppress tumor growth.

INTRODUCTION

Cancer is characterized by metabolic reprogramming to fulfill its increased demands for uncontrolled cell proliferation. Malignant cells, including hepatocellular carcinoma (HCC), rapidly uptake and catabolize nutrients for cancer progression (1), making metabolic dysregulation an attractive therapeutic target (2).

Tricarboxylic acid (TCA) cycle integrates nutrient metabolism and provides precursors for macromolecule synthesis (3). Although normal cells rely primarily on glucose-derived pyruvate for oxidative phosphorylation, cancer cells preferentially use aerobic glycolysis and glutamine metabolism, thereby rewiring TCA cycle flux to support rapid proliferation (4). While the Warburg effect is a well-established hallmark of cancer, accumulating evidence reveals that mitochondria and TCA cycle are recognized to play indispensable and multifaceted roles during tumor initiation and progression. Aberrant TCA activity, such as isocitrate dehydrogenase 2 mutation (mIDH2) or increased citrate synthase (CS) activity, reprograms central carbon metabolism and promotes tumorigenesis (5). Altered expression of transporters that modulate TCA intermediates also rewires cancer metabolism status (6). TCA cycle function requires precise regulation of intermediates via transporters across the plasma and mitochondrial membranes (7). Solute carriers (SLCs) mediate the transmembrane transport of various nutrients and metabolites, such as glucose, amino acids, lipids, and nucleotides, and thus serve as metabolic switches (8). As most SLCs transport small molecules, they are increasingly considered promising targets for drug development (9), as exemplified by the successful targeting of SLC1A5 in cancer therapy (10). Six main SLC families with members have been identified as carriers that regulate TCA cycle flux by transporting TCA cycle intermediates such as citrate, succinate, and malate through the plasma membrane or mitochondrial matrix according to an updated version of the guide to transporters (11). However, despite the observations, the precise roles of these SLC transporters in regulating TCA cycle function and tumor growth remain largely elusive. Based on previous findings, these SLC family members likely modulate anaplerotic or cataplerotic flux in response to metabolic adaptation of cancer cells.

Here, we identified SLC13A2, a citrate transporter, as down-regulated in both human/mouse HCC liver tissues and cells. Liver-specific knockout (LKO) and overexpression models confirmed its tumor-suppressive role in restraining HCC progression. Given its role as a key transporter for TCA cycle intermediates, elucidating the precise role and underlying mechanisms of SLC13A2 in HCC pathogenesis is crucial. Our findings reveal SLC13A2 as a critical node connecting metabolic signaling to transcriptional regulation for tumor growth, suggesting its potential as an attractive therapeutic target for HCC.

RESULTS

SLC13A2 is a down-regulated transporter for TCA metabolites in HCC

Initially, primary HCC mouse models were established by hydrodynamic tail vein injection (HTVi) of oncogene plasmids [mesenchymal-epithelial transition factor (c-MET)/ΔN90-β-catenin] in combination with Sleeping Beauty (SB) transposase (fig. S1A) or tail vein injection of adeno-associated virus (AAV) encoding oncogenic cMYC and nRAS (AAV-cMYC/nRAS) as described previously (fig. S1F) (12, 13). Nontargeted metabolomics of serum as well as liver tumor tissues from animal models and clinical patients consistently revealed TCA cycle metabolites as the most prominently altered ones in HCC mice (Fig. 1A and fig. S1, D and G), highlighting the pivotal role of TCA cycle during HCC pathogenesis. Therefore, we focused on transporters of TCA cycle metabolites, including SLC5, SLC13, SLC16, SLC25, SLC33, and SLC54, which have been identified to transport TCA intermediates such as citrate, succinate, α-ketoglutaric acid (α-KG), malate, lactate, and pyruvate, according to an updated transporters guide (Fig. 1B) (11). We performed Cox survival analysis and revealed that several TCA-related transporters were negatively correlated with the hazard ratio in HCC patients, including SLC5A8, SLC13A2, SLC13A5, SLC25A1, SLC25A10, SLC25A11, and SLC54A1 (Fig. 1C). Then, RNA sequencing (RNA-seq) data from heterogeneous liver tumors induced by diverse cancer driver genes selected by mutational frequency from human HCC cohorts with different etiological factors and ethnic origins were analyzed (14). The results show that SLC13A2 was potently and consistently down-regulated in all liver tumors, showing an extraordinary pattern compared with other detectable SLC transporters (Fig. 1D).

Fig. 1. SLC13A2 is a transporter for TCA cycle intermediates deregulated in HCC.

(A) Liver tumor tissues were collected from the HTVi model and clinical patients and subjected to untargeted metabolomic profiling for pathway enrichment. (B) Summarized SLC transporters for TCA cycle intermediates. (C) Forest plot showing the hazard ratios (HR) for multiple genes encoding SLC transporters of TCA cycle intermediates in HCC. The horizontal line represents the 95% confidence interval. (D) Heatmap of detectable SLC transporters for TCA cycle intermediates in heterogeneous primary liver cancer models generated by genome editing of cancer driver genes selected by mutational frequency from human HCC cohorts (PRJNA674008). (E) qPCR analysis to screen candidate SLC transporters for HCC progression based on the relative expression of SLC transporters in the HTVi model (n = 4 for the control group and n = 5 for the model group), mouse HCC Hepa1-6 cells, AML12 cells, and MPHs (n = 3 independent experiments). (F) Venn diagram showing the overlap of significantly differentially expressed SLC transporters in the HTVi model, mouse HCC Hepa1-6 cells, and normal hepatocytes (AML12 and MPHs). (G and H) SLC13A2 expression in liver tissues from HTVi, AAV-cMYC/nRAS, and STZ-HFD HCC model. The data are presented as means ± SEM. *P < 0.05; **P < 0.01, two-tailed unpaired Student’s t test. Ctrl, control; n.d., not determined; NS, not significant.

Based on the above data, we used the HTVi model to evaluate the changes in the mentioned SLC family members (fig. S1, A and B). Obvious tumor nodules were observed in the livers that were transfected by the oncogene plasmids (fig. S1B, middle), confirmed by histological changes (fig. S1B, right) and increased liver/body weight ratio (fig. S1B, left). In addition, gene expression involved in glucose and fatty acid metabolism, as well as proteins and cell signaling that promote cell proliferation, further supports the success in generating the primary HCC mouse model (fig. S1, C and E). Next, we detected the expression of SLC transporters of interest and found that the expression of Slc16a3 and Slc16a8 significantly increased, while Slc13a2, Slc25a1, Slc25a10, Slc25a11, Slc25a21, Slc54a1, and Slc54a2 significantly decreased in liver tissues from mice with HCC (Fig. 1E, top). Among these genes, Slc13a2 decreased 10 to 100 times and attracted our attention because it exhibited the most abundant variation. Moreover, expression of Slc25a21 and Slc54a3 significantly increased in mouse HCC Hepa1-6 cells compared with primary hepatocytes, while other SLC transporters of interest decreased, except for Slc16a1 (Fig. 1E, middle). Similarly, we detected SLC transporter expression in mouse HCC Hepa1-6 cells compared with immortalized mouse hepatocyte alpha mouse liver 12 (AML12) cells. The gene expression of Slc25a1, Slc25a10, and Slc54a1 increased, while that of Slc13a2, Slc13a3, Slc13a5, and Slc54a3 decreased in the HCC cell line (Fig. 1E, bottom). Integrated analysis revealed that Slc13a2 was the only gene consistently decreased both in vivo and in vitro (Fig. 1F). In addition, reduced protein and RNA expression of SLC13A2 was further confirmed in HCC liver tissues from HTVi, AAV-cMYC/nRAS, and streptozotocin–high-fat diet (STZ-HFD) primary models (Fig. 1, G and H), in which HCC mice also exhibited significantly increased liver/body weight ratios, prominent tumor nodules, and increased levels of proliferation-associated proteins (fig. S1, F and H to K). These results indicate that SLC13A2 is down-regulated in HCC and stands out from the other SLC transporters for TCA cycle intermediates.

By analyzing a published Gene Expression Omnibus (GEO) dataset (GSE193567), we identified 4247 differentially expressed genes (DEGs) between cancerous and paracancerous tissues of HCC patients. Gene enrichment analysis revealed that the DEGs were enriched in macromolecular biosynthesis, membrane trafficking, and transmembrane transporters (Fig. 2A, left). Among them, SLC13A2 was one of the three significantly altered SLC transporters for TCA cycle intermediates (Fig. 2A, right). As shown for all DEGs, SLC13A2 was significantly reduced in tumor tissues (Fig. 2B). Spatial transcriptomic data from the SpatialTME portal revealed that SLC13A2 expression was markedly reduced in tumor regions, as shown by spatial mapping in a representative HCC sample from clinical patients (Fig. 2C, top). Consistently, single-cell transcriptomic profiling from Integrative Molecular Database of Hepatocellular Carcinoma (HCCDB) confirmed that SLC13A2 expression was decreased in tumor cell populations compared with normal counterparts (Fig. 2C, bottom). Moreover, differential expression analysis of The Cancer Genome Atlas Liver Hepatocellular Carcinoma (TCGA-LIHC) dataset revealed that SLC13A2 expression was significantly down-regulated in HCC samples compared with normal tissues (Fig. 2D). GEO datasets (GSE14323, GSE60502, and GSE14520) also revealed a significant down-regulation of SLC13A2 in HCC samples compared with normal liver tissues, with similar reductions also observed in cirrhotic liver tissues and cirrhosis-associated HCC tissues (Fig. 2E). Moreover, Kaplan-Meier survival curves revealed that male patients with high SLC13A2 expression had a better prognosis and a greater overall survival rate (Fig. 2F). In addition, SLC13A2 mRNA was significantly lower in human HCC cell lines than in the normal human liver cell line transformed human liver epithelial-2 (THLE-2) cells (Fig. 2G). Furthermore, we collected liver tissues from clinical patients with HCC and analyzed the expression of SLC13A2. It was confirmed that all HCC patients had lower SLC13A2 expression in cancerous tissues than in paracancerous tissues (Fig. 2, H and I). Immunohistochemical (IHC) staining also revealed that SLC13A2 expression was down-regulated in HCC samples compared with adjacent normal tissues (Fig. 2J). The above data suggest that reduced expression of SLC13A2 is conserved in mouse and human HCC and is positively associated with the clinical prognosis of this disease.

Fig. 2. SLC13A2 expression is decreased in human HCC specimens and cell lines, which is consistent with the findings in animal models.

(A and B) RNA-seq dataset analysis of human HCC and paracancerous specimens (GSE193567). Shown are ontology enrichment [(A), left], heatmap of SLC transporters for TCA intermediates [(A), right], volcano plot highlighting SLC13A2 [(B), left], and SLC13A2 expression in tumor versus para-tumor tissues [(B), right]. (C) Spatial transcriptomics data from the SpatialTME portal and single-cell RNA-seq data from HCCDB showing SLC13A2 expression and distribution in tumor and nontumor tissues, with purple intensity indicating expression level. Tumor regions are outlined by white dashed lines. (D) SLC13A2 expression in the dataset of TCGA-LIHC. (E) SLC13A2 expression in GEO datasets (GSE14323, GSE60502 and GSE14520). (F) Kaplan-Meier survival analysis of male patients with HCC stratified by high or low SLC13A2 expression. (G) SLC13A2 expression in human HCC cell lines versus normal hepatocytes THLE-2 cells after being split and cultured for 48 hours. (H and I) SLC13A2 expression in eight clinical HCC patients. C, cancerous; PC, paracancerous. (J) Representative IHC staining of SLC13A2 in human HCC samples and adjacent nontumor tissue. The data are presented as means ± SEM. *P < 0.05; **P < 0.01, two-tailed unpaired Student’s t test. FPKM, fragments per kilobase of transcript per million mapped reads; OS, overall survival.

SLC13A2 inhibits HCC progression induced by hydrodynamic transfection of oncogenes

The function of SLC13A2, a previously described transporter in renal tubules and intestinal cells, in physiological homeostasis in the liver is still obscure. In our published data (15), neither hepatocyte-specific knockout (KO) of SLC13A2 nor AAV-mediated overexpression resulted in consistent differences in blood glucose levels, body weight, liver weight, histological staining, or the expression of genes involved in glycolysis, TCA cycle, or lipogenesis. These data imply that hepatocyte-specific KO or overexpression of SLC13A2 (SLC13A2-OE) primes the cell metabolic status for HCC tumorigenesis but is not sufficient to directly induce the initiation of HCC.

Then, CRISPR-Cas9–mediated LKO mice were established and injected with oncogene plasmids using HTVi after 2 weeks of AAV8 injection and euthanized after 8 more weeks (Fig. 3A). When SLC13A2 was efficiently knocked out (Fig. 3B), liver/body weight ratio was significantly increased (Fig. 3D). Neither body weight nor blood glucose levels showed notable differences between the groups (fig. S2, A and B). The number of surface tumor nodules was significantly greater in LKO group (Fig. 3C), with remarkably more mice with larger maximal tumor sizes than the control group (Fig. 3E). Tumor burden was further assessed by grading tumors into light, medium, and severe categories. The LKO group exhibited fewer light-grade tumors and markedly more severe-grade tumors compared with controls, as shown by the distribution of tumor numbers in each grade (fig. S2E). As indicators of liver tumor burden (16), the enzymatic activity of serum and liver alanine transaminase (ALT)/aspartate transaminase (AST) also increased with SLC13A2 ablation (Fig. 3F). Larger nodules with more irregularly shaped nuclei were observed in the liver hematoxylin and eosin (H&E)–stained slides from the SLC13A2 LKO group (Fig. 3G). Ki67 staining also revealed 16.6% nuclear positivity in the control group and ~22.9% nuclear positivity in the SLC13A2 LKO group, demonstrating a significant increase in cell proliferation induced by the KO of SLC13A2 in vivo (Fig. 3H). Consistently, the phosphorylation of mammalian target of rapamycin (mTOR), protein kinase B (AKT), and glycogen synthase kinase 3β (GSK3β), which are involved in cellular signaling related to cell proliferation and metabolism, was significantly elevated, consistent with the promotion of tumor growth in LKO mice (Fig. 3I).

Fig. 3. SLC13A2 inhibits HCC progression in the mouse model induced by HTVi.

(A) Diagram showing LKO strategy and study design. (B) SLC13A2 KO efficiency in mouse liver tissues (n = 6). (C and D) Representative liver morphology and liver/body weight ratios (n = 16). Scale bar, 5 mm. (E) Tumor incidence and maximal tumor size (n = 13) in the two groups. (F) Liver and serum ALT/AST activity (n = 10). (G and H) Representative H&E staining and Ki67 staining showing tissue morphology and proliferation (n = 6). (I) Immunoblots of signaling pathways associated with tumor growth (n = 6). (J) Diagram of study design for the overexpression experiment. (K) SLC13A2 overexpression efficiency in mouse HCC tissues. (L and M) Representative liver morphology and liver/body weight ratios (n = 11 for the GFP group and n = 12 for the SLC13A2-OE group). Scale bar, 5 mm. (N) Tumor incidence (left) and the maximal tumor size (right, n = 11 for the GFP group and n = 10 for the SLC13A2-OE group) in the two groups. (O) Liver and serum ALT/AST activity (n = 11 for the GFP group and n = 10 for the SLC13A2-OE group). (P and Q). Representative H&E staining and Ki67 staining showing tissue morphology and proliferation (n = 6). (R) Immunoblots of signaling pathways associated with tumor growth (n = 6). The data are presented as means ± SEM. *P < 0.05; **P < 0.01, LKO versus Ctrl group, SLC13A2-OE versus GFP group, two-tailed unpaired Student’s t test.

Likewise, liver-specific overexpression of SLC13A2 (OE) was established using AAV8 under the control of the thyroid hormone–binding globulin (TBG) promoter. After 2 weeks of AAV injection, the mice were injected with oncogene plasmids via HTVi and were euthanized after 9 more weeks, which was a longer period than that in the LKO experiments, to better demonstrate tumor suppression by SLC13A2 (Fig. 3J). When SLC13A2 was efficiently overexpressed (Fig. 3K), body weight or blood glucose levels remained unchanged (fig. S2, C and D). The number of surface tumor nodules and liver/body weight ratio were significantly lower in the SLC13A2-OE group (Fig. 3, L and M), and maximal tumor size was also significantly reduced, with a higher proportion of light-grade tumors compared to the green fluorescent protein (GFP) group (Fig. 3N and fig. S2F). The enzymatic activity of ALT and AST was consistently inhibited in the blood and liver tissues from SLC13A2-OE mice (Fig. 3O). Hepatocytes were smaller and more separated in the SLC13A2-OE group, as indicated by a decreased nucleoplasmic ratio (Fig. 3P). Consistent with the decrease in tumor growth, Ki67 staining revealed 8.0% nuclear positivity in the control group and ~2.8% nuclear positivity in the SLC13A2-OE group, demonstrating a significant decrease in cell proliferation induced by SLC13A2 (Fig. 3Q). Consistently, immunoblotting showed a clear reduction in the phosphorylation of mTOR and GSK3β, whereas p-AKT displayed a downward trend without reaching statistical significance (Fig. 3R). Similarly, RNA expression of cell proliferation–associated genes, such as Ccnb1, Foxm1, Fgf21, and Mcm2, decreased in liver tissues overexpressing SLC13A2 (fig. S2G). Moreover, metabolic genes, such as Hk2 and Mdh2, also significantly decreased with the overexpression of SLC13A2 (fig. S2G). These findings indicate that SLC13A2 overexpression disrupts tumor-favoring metabolic programs in HCC.

SLC13A2 restricts proliferation by depleting TCA intermediates for oxidative respiration

Next, a mouse HCC cell line, Hepa1-6, and a human HCC cell line, HepG2, were chosen to confirm the capacity of SLC13A2 to prevent the proliferation of HCC cells. To further investigate the effect of SLC13A2 on the proliferation of HCC cells in vitro, we used RNA interference to knock down (KD) and plasmid/adenovirus transfection to induce SLC13A2 overexpression. Notably, RNA interference effectively reduced SLC13A2 protein levels in Hepa1-6 cells, while plasmid or adenoviral transfection increased its expression in both Hepa1-6 and HepG2 cells (fig. S3A). Notably, SLC13A2 KD significantly enhanced cell viability and colony formation, whereas its overexpression suppressed cell proliferation in both cell lines (fig. S3, B and C). Consistently, the ratio of 5-ethynyl-2′-deoxyuridine (EdU)–positive cells was significantly greater in SLC13A2-KD cells, while attenuated in SLC13A2-OE cells (fig. S3, D and E). Consistently, a cell cycle distribution assay showed that SLC13A2 KD decreased the proportion of cells in S phase, while SLC13A2 overexpression increased S phase arrest (fig. S3F).

Since classical substrates of SLC13A2 are TCA cycle intermediates that control cellular function and cell fate (3), we conducted ultraperformance liquid chromatography (UPLC)–quadrupole orthogonal acceleration–time-of-flight (Q-TOF)–mass spectrometry (MS)–based untargeted metabolomics to investigate its effect on metabolism in HCC Hepa1-6 cells. The levels of intermediates of the TCA cycle and glycolysis, including succinate, citrate, and glucose-6-phosphate (G-6-P); nonessential amino acids, such as glutamine and aspartate; and metabolites involved in nucleotide metabolism, such as cytidine and orotidine, decreased after SLC13A2 overexpression (Fig. 4A and fig. S5A). The differentially abundant metabolites were largely enriched in alanine-aspartate-glutamate metabolism, d-glutamine and d-glutamate metabolism, and the aminoacyl-tRNA biosynthesis pathway (fig. S5B). Seahorse analysis showed that SLC13A2 overexpression impaired respiratory capacity and glycolysis rate in cancerous Hepa1-6 cells (Fig. 4, B and C, and fig. S5, C and D), while showing no effect in normal hepatocytes AML12 cells (fig. S5E). Stably labeled citrate tracing study revealed no change of labeled TCA cycle metabolites in Hepa1-6 cells, but increased ones in mouse primary hepatocytes (MPHs), confirming the disparate roles of SLC13A2 in cancerous and normal hepatocytes (fig. S4A). Consistently, adenosine 5′-triphosphate (ATP) production decreased with SLC13A2 overexpression and increased with SLC13A2 KD (fig. S5, F and G). Moreover, mitochondrial membrane potential was assessed using JC-1 staining. SLC13A2 overexpression reduced the aggregate-to-monomer ratio, indicating depolarization, while KD increased the ratio (fig. S5, H and I). The findings suggested that SLC13A2 decreases cellular mitochondrial capability by increasing mitochondrial permeability.

Fig. 4. SLC13A2 limits the TCA cycle and respiration to suppress tumor growth via citrate-mediated PKM2 acetylation and degradation.

(A) Relative abundance of TCA cycle and glycolysis metabolites in Hepa1-6 cells overexpressing SLC13A2 for 60 hours (n = 3). (B and C) Mitochondrial respiration (OCR) and glycolysis (ECAR) measured by Seahorse XF analysis. (D) Untargeted metabolomics of vector- or SLC13A2-transfected Hepa1-6 cells cultured with [U-¹³C₆] glucose for 1 or 2 hours (n = 3). (E) Relative abundance of pyruvate (n = 3). (F) Cell viability/colony formation with the treatment of 500 μM pyruvate. *P < 0.05 versus vector without pyruvate; ^#P < 0.05 versus SLC13A2 without pyruvate. (G) Intracellular fractional labeling of citrate (m + 6) and acetyl-CoA (m + 2) in vector- or SLC13A2-transfected Hepa1-6 after 1 hour [U-¹³C₆] citrate tracing (n = 3). (H and I) Relative abundance of oxaloacetate and extracellular/mitochondrial citrate levels (n = 3). (J) Cell viability after 72 hours ACLY inhibitor (ACLYi) treatment (50 μM). *P < 0.05 versus vector without ACLY inhibitor; ^#P < 0.05 versus SLC13A2 without ACLY inhibitor. (n = 4). (K) PKM2 subcellular protein levels after 60 hours transfection. (L) Co-IP assay showing the acetylation of PKM2. (M) PKM2 protein levels in SLC13A2-KD Hepa1-6 cells expressing WT/mutant PKM2, cultured for 60 hours. (N) Pyruvate kinase activity after 60 hours transfection (n = 3). (O) SLC13A2 and/or PKM2 overexpression efficiency (left) and cell growth curves measured at 0, 24, 72, and 120 hours (right). *P < 0.05 SLC13A2 versus vector; ^#P < 0.05 SLC13A2 + PKM2 versus SLC13A2. (P and Q) Pyruvate kinase activity (n = 10), pyruvate content (n = 10), and PKM2 protein levels (n = 6) in HTVi liver tissues with SLC13A2 LKO or OE. (R) NADH/NAD⁺ ratio in HTVi liver tissues with SLC13A2 overexpression (n = 6). Data are presented as means ± SEM. *P < 0.05, two-tailed unpaired Student’s t test [(A), (D), (E), (G) to (I), (N), and (P) to (R)]; two-way analysis of variance (ANOVA), followed by Bonferroni multiple comparisons [(F), (J), and (O)]. h, hour; IB, immunoblot; NC, Negative Control.

SLC13A2-imported citrate induces PKM2 acetylation and degradation, restricting TCA cycle

The mechanisms by which SLC13A2, a recently described hepatic citrate inward transporter as described by our group previously (15), restricts the TCA cycle to suppress HCC tumor growth are intriguing. Therefore, we performed [U-¹³C₆] glucose metabolic flux analysis to assess the impact of SLC13A2 on glycolysis. Cell samples were collected at 1 hour for the measurement of upstream glycolytic intermediates and at both 1 and 2 hours for pyruvate. Unexpectedly, the levels of glycolysis upstream intermediates G-6-P, glyceraldehyde-3-phosphate/dihydroxyacetone phosphate (GAP/DHAP), 3-phosphoglycerate (3-PG), and phosphoenolpyruvate (PEP) significantly increased at 1 hour, while the levels of the downstream metabolite pyruvate significantly decreased at both 1 and 2 hours with SLC13A2 overexpression (Fig. 4D). Furthermore, targeted metabolomic analysis confirmed that intracellular pyruvate content decreased significantly with SLC13A2 overexpression (Fig. 4E), which was consistent with the restoration of cell proliferation by an additional supply of pyruvate (500 μM) (Fig. 4F). Compatibly, [U-¹³C₆] glucose tracing study revealed suppressed production of TCA cycle metabolites, including citrate, fumarate, malate, succinate, and α-KG (fig. S4B). The suppression of TCA cycle was also confirmed using [U-¹³C₅] glutamine metabolic flux study, as revealed by decreased production of malate, fumarate, and α-KG from glutamine (fig. S4C). These findings suggested that SLC13A2 controls the entrance of the TCA cycle, which was consistent with its ability to suppress tumor growth and attenuate cell proliferation.

Then, [U-¹³C₆] citrate tracing study was applied to confirm that SLC13A2 transports citrate in HCC cells, leading to the increased synthesis of acetyl–coenzyme A (CoA) (Fig. 4G). It was thus hypothesized that citrate imported by SLC13A2 was catalyzed by ATP–citrate lyase (ACLY) to produce oxaloacetate and acetyl-CoA. We used targeted metabolomic analysis to detect intracellular oxaloacetate and found that it was significantly increased by SLC13A2 overexpression (Fig. 4H), while extracellular and mitochondrial citrate were significantly decreased (Fig. 4I). These results indicated that SLC13A2 imported extracellular citrate, which produced oxaloacetate. The treatment of an ACLY inhibitor (50 μM) significantly reversed SLC13A2-inhibited cell growth (Fig. 4J). We hypothesized that acetyl-CoA released by citrate catalysis may act as a precursor for the acetylation of intracellular proteins, as there is no significant synthesis of TCA cycle metabolites from citrate (fig. S4A). Pyruvate kinase isozyme type M2 (PKM2), a major enzyme that catalyzes the transition of PEP to pyruvate in HCC cells (17), may be a target of acetylation, leading to its protein degradation (18). Therefore, we performed subcellular fractionation and found that SLC13A2 significantly decreased PKM2 protein levels in the cytosol and nucleus (Fig. 4K), without altering Pkm2 mRNA levels (fig. S6A), indicating a posttranslational modification of regulation. Coimmunoprecipitation (co-IP) experiments revealed that the acetylation of PKM2 increased significantly with SLC13A2 overexpression (Fig. 4L). Furthermore, protein degradation was the primary pathway involved in the SLC13A2-mediated decrease in the PKM2 protein level (fig. S6B). Since Lys³⁰⁵ has been reported to be acetylated for PKM2 protein degradation (18), we generated an acetylation-mimic mutant, PKM2 Lys³⁰⁵→Gln (K305Q), and transfected it with SLC13A2 KD. It confirmed that the level of the PKM2 K305Q mutant protein significantly decreased compared with that of the wild-type (WT) protein. The KD of SLC13A2 increased the levels of both the endogenous protein and the transfected WT PKM2 protein but did not increase the levels of the K305Q mutant protein (Fig. 4M). Consistently, pyruvate kinase activity was strongly attenuated by SLC13A2 overexpression (Fig. 4N). Furthermore, PKM2 overexpression partially reversed the inhibitory effect of SLC13A2 on HCC cell proliferation (Fig. 4O). Compatibly, overexpression of SLC13A2 decreases the protein level of PKM2, pyruvate kinase activity, and pyruvate content in the tumor tissues of HTVi model. The LKO increases PKM2 protein and pyruvate kinase activity, without reaching a significant value in pyruvate content (Fig. 4, P and Q). Consistent with the in vitro data, nicotinamide adenine dinucleotide (reduced form)/oxidized form of NAD⁺ (NADH/NAD⁺) ratio was also increased in the SLC13A2-OE group (Fig. 4R and fig. S6C), confirming the suppression of TCA cycle in the in vivo model. The results indicate that SLC13A2 suppresses HCC tumor growth at least partially through the degradation of PKM2.

SLC13A2 induces citrate-dependent protein acetylation and reprograms transcriptional regulation

Since PKM2 degradation only partially mediated SLC13A2 effects, we hypothesized that SLC13A2-transported citrate generated acetyl-CoA, serving as a universal donor for protein acetylation. Consistently, pan-lysine acetylation was markedly increased in HTVi liver tissues with SLC13A2 overexpression, but reduced in SLC13A2 LKO mice (Fig. 5A). The increase of lysine-acetylated proteins was distributed in both cytosolic and nuclear compartments (Fig. 5B). Citrate supplementation further enhanced global acetylation in an SLC13A2-dependent manner (fig. S7, A to C).

Fig. 5. SLC13A2 regulates global protein acetylation and transcriptional programs in HCC.

(A) Total protein acetylation in liver tissues from HTVi mice with liver-specific overexpression or KO of SLC13A2 (n = 3). (B) Immunoblot of subcellular fractions from Hepa1-6 cells overexpressing SLC13A2, probed with pan-acetyl-lysine antibody. (C) Kyoto Encyclopedia of Genes and Genomes pathway enrichment analysis of proteins lacking detectable acetylation sites identified by acetyl-proteomics in SLC13A2-OE Hepa1-6 cells. (D) GO molecular function enrichment analysis of proteins with detectable acetylation sites identified by acetyl-proteomics in SLC13A2-OE Hepa1-6 cells. (E) Histone modification levels in HTVi liver tumor tissues of control and SLC13A2 LKO mice detected by immunoblotting. Data are presented as means ± SEM (n = 3). *P < 0.05; **P < 0.01, versus controls, two-tailed unpaired Student’s t test. (F) Immunoblot analysis of histone acetylation marks in Hepa1-6 cells overexpressing SLC13A2. The data are representative from three independent experiments. (G) Immunoblot analysis of histone acetylation marks in Hepa1-6 cells transfected and treated with increasing concentrations of citrate (0, 250, and 500 μM) for 48 hours. (H) Volcano plot of hepatic genes in SLC13A2-OE mice. (I) Heatmap of representative genes involved in cell proliferation and metabolism. (J) GSEA of RNA-seq showing gene enrichment in SLC13A2-OE mice. (K) Heatmap of key TFs in SLC13A2-OE mice. NES, Normalized Enrichment Score.

Acetyl-proteomic analysis was thus applied to systematically investigate acetylation reprogramming. Differential proteins were grouped based on the presence of detectable acetylation sites. Functional enrichment analysis revealed that proteins lacking acetylation sites were mainly involved in chromatin remodeling and DNA replication and repair (Fig. 5C), whereas proteins with acetylation sites were enriched in transcription-related functions, including transcriptional activation, transcription coactivator activity, and histone acetyltransferase activity (Fig. 5D). Given that histone acetylation is a classical epigenetic mark that alters chromatin structure and facilitates transcription factor (TF) binding, we examined its regulation by SLC13A2. Immunoblotting with histone-specific acetylation antibodies confirmed that SLC13A2 overexpression markedly increased acetylation at H3K9ac, H3K27ac, H4K5ac, and H4K12ac, whereas SLC13A2 LKO led to a significant reduction at H3K9ac, H3K27ac, and H4K12ac (Fig. 5, E and F). To further investigate the role of SLC13A2 in histone acetylation, we performed KD and overexpression experiments with citrate supplementation. In SLC13A2-KD cells, citrate supplementation failed to restore acetylation, with multiple histone marks remaining reduced (Fig. 5G). In contrast, in SLC13A2-OE cells, citrate supplementation further increased acetylation of the same histone marks (fig. S7D, left). Given that succinate is also a substrate of SLC13A2, we supplemented the cells with succinate. However, histone acetylation levels did not change in response to succinate (fig. S7D, right). These results indicate that citrate-mediated enhancement of histone acetylation is largely dependent on SLC13A2. As H3K9ac, H3K27ac, and H4K12ac are well-characterized activation-associated modifications that facilitate transcriptional initiation and elongation (19–21), SLC13A2-mediated acetylation reprogramming plays a critical role in promoting transcriptionally active chromatin states in HCC.

Therefore, RNA-seq was performed on HTVi tumors with overexpression of SLC13A2 (Fig. 3J), to link histone acetylation to transcriptional regulation, and a total of 2205 hepatic DEGs were identified (false discovery rate < 0.05, fold change > 2, adjusted P < 0.05) (Fig. 5H). Down-regulated genes were enriched in cell proliferation and nuclear division, whereas up-regulated genes were mainly in small-molecule catabolism and nutrient metabolism (Fig. 5I). Gene set enrichment analysis (GSEA) revealed 102 gene sets significantly differentially regulated by SLC13A2, with organic acid and amino acid catabolic process most enriched (Fig. 5J). TF annotation of the DEGs was performed using the JASPAR database to identify potential regulators underlying the observed transcriptional changes. This analysis revealed that key TFs, including Cebpa, Bhlhe41, Mafb, Snai2, Rorc, Tcf21, and Klf1, were significantly up-regulated upon SLC13A2 overexpression (Fig. 5K), indicating that SLC13A2 drives transcriptional reprogramming in HCC cells through activation of specific transcriptional regulators.

To determine whether SLC13A2-induced transcriptional regulation was linked to chromatin acetylation, we performed genome-wide chromatin immunoprecipitation sequencing (ChIP-seq) using pan-acetylation antibodies. This revealed increased acetylation signals near transcription start sites (TSSs) of genes up-regulated by SLC13A2 (up gene, purple line). Heatmaps also confirmed this pattern, displaying strong acetylation enrichment at TSSs of up-regulated genes, but weak, diffuse signals for down-regulated genes (Fig. 6A). These data indicate that SLC13A2 selectively promotes acetylation modifications in transcriptionally active regions. Acetylation-enriched genes were associated with metabolic regulation, transcriptional activation, and mitochondrial function (Fig. 6B). In parallel, gene ontology (GO) enrichment analysis of DEGs identified by RNA-seq showed that up-regulated genes were enriched in metabolic pathways, indicating enhanced metabolic activity, whereas down-regulated genes were involved in cell growth, revealing suppression on proliferative programs (Fig. 6C). Integrative Genomics Viewer (IGV) tracks further illustrated these changes: Apoa2 displayed increased expression with elevated acetylation around the TSS, whereas Igfbp6 showed reduced expression without acetylation enrichment (Fig. 6D). Together, these findings suggest that SLC13A2 drives transcriptional reprogramming through acetylation, linking chromatin acetylation profile with transcriptional outputs.

Fig. 6. SLC13A2-mediated histone acetylation promotes transcriptional activation of key TFs.

(A) Metagene and heatmap analyses showing the distribution of acetylated lysine (Acetyl-Lys) peaks around TSS (±2 kb) from ChIP-seq of Acetyl-Lys in SLC13A2-OE Hepa1-6 cells. (B) Circular chord diagram displaying GO enrichment of DEGs categorized by functional terms. ATPase, adenosine triphosphatase. (C) Bar plots of GO enrichment analysis for up-regulated (orange) and down-regulated (blue) genes from RNA-seq of SLC13A2-OE mice. (D) Representative IGV tracks showing ChIP-seq (Acetyl-Lys) signals in SLC13A2-OE Hepa1-6 cells and RNA-seq signals from liver tumor tissues of SLC13A2-OE mice at selected loci (Igfbp6 and Apoa2). (E) IGV tracks of Acetyl-Lys ChIP-seq signals at Cebpa, Tcf21, and Bhlhe41 loci, showing increased acetylation. (F) Public ChIP-seq tracks (H3K9ac, H3K27ac, and H4K12ac) from H1 human embryonic stem cells and HepG2 cells (ENCODE database) showing promoter acetylation at Cebpa, Tcf21, and Bhlhe41. (G) GO enrichment analysis of down-regulated DEGs from RNA-seq that are potential downstream targets of TFs Cebpa, Tcf21, and Bhlhe41. (H) Heatmap of representative downstream genes of Cebpa, Tcf21, and Bhlhe41 in SLC13A2-OE tumor tissues.

TFs up-regulated in the RNA-seq dataset were analyzed in the ChIP-seq dataset. It revealed that three of seven previously found TFs (Fig. 5K)—CCAAT enhancer-binding protein alpha (CEBPA), basic helix-loop-helix family member e41 (BHLHE41), and transcription factor 21 (TCF21)—exhibited increased acetylation signals around their TSS and gene bodies, suggesting that SLC13A2 preferentially enhances chromatin acetylation of specific TFs, which directly promote their transcriptional activation (Fig. 6E and fig. S7E). Furthermore, public ChIP-seq data from the ENCODE Project revealed strong enrichment of histone acetylation marks (H3K9ac, H3K27ac, and H4K12ac) at the promoters of the up-regulated TFs CEBPA, BHLHE41, and TCF21 (Fig. 6F). These findings suggest that SLC13A2 facilitates transcriptional activation of CEBPA, BHLHE41, and TCF21 TFs through histone acetylation remodeling.

To investigate whether SLC13A2 suppresses proliferation-related genes via up-regulated TFs, genome-wide targets of CEBPA, BHLHE41, and TCF21 from ChIP-Atlas were overlapped with RNA-seq down-regulated genes upon SLC13A2 overexpression (fig. S7F). The intersecting genes, considered putative direct TF targets, were enriched in pathways related to cell proliferation, nuclear division, and DNA replication. These data indicate that SLC13A2 activates inhibitory TFs to mediate coordinated antitumor programs, including cell cycle arrest, metabolic reprogramming, and epithelial differentiation (Fig. 6G). A heatmap illustrates representative target genes with marked expression changes induced by SLC13A2 (Fig. 6H). The coordinated activation of this transcriptional network provides mechanistic evidence for the tumor-suppressive role of SLC13A2 through epigenetic regulation.

SLC13A2 suppresses tumor growth in patient-derived xenograft model

Given the established potential of AAV as a gene therapy vector, we developed a patient-derived xenograft (PDX) model by subcutaneously implanting HCC tumor fragments into immunodeficient mice. SLC13A2 was overexpressed via intratumoral injection of AAV9 vectors driven by the CAG promoter to assess its tumor-suppressive function (Fig. 7A). Tumors in the SLC13A2-OE group were visibly smaller than those in control (Fig. 7B), with significantly reduced tumor weight and volume but no change in body weight (Fig. 7, C to E). Histological staining revealed smaller and more separated hepatocytes in the SLC13A2 group (Fig. 7F). Consistent with the decrease in tumor growth, Ki67 staining showed markedly decreased nuclear positivity in the SLC13A2-OE group, demonstrating a significant decrease in cell proliferation induced by SLC13A2 in vivo (Fig. 7G). Similarly, phosphorylation of mTOR, AKT, and GSK3β, as well as expression of the proliferation-associated proteins CDK2 and CCND1, were significantly down-regulated (Fig. 7H). The decrease of PKM2 protein and increased histone acetylation verified the mechanisms of SLC13A2-mediated metabolic and epigenetic rewiring underlying the suppression of tumor growth (Fig. 7, I and J). The present study establishes a model in which citrate imported by SLC13A2 generates acetyl-CoA for protein acetylation, including PKM2 and histone proteins, and functions as a fundamental hub connecting metabolic rewiring and transcriptional regulation to constrain HCC development (Fig. 7K). The data identify a promising translational target for clinical HCC intervention in the future.

Fig. 7. SLC13A2 overexpression suppresses tumor growth in PDX model.

(A) hSLC13A2 overexpression efficiency in PDX HCC tissues (n = 4). (B) PDX HCC morphology upon dissection. Scale bar, 10 mm. (C and D) Body weight and tumor weight in the two groups (n = 4). (E) Tumor volume in the two groups (n = 4). (F) Representative images of H&E staining. (G) IHC staining of Ki67 (n = 3). (H to J) Immunoblot analysis of tumor-growth associated signaling pathways, histone acetylation marks, and PKM2. (K) Schematic illustration of SLC13A2-mediated metabolic and epigenetic regulation in tumor cells. OAA, oxaloacetate. The data are presented as means ± SEM. *P < 0.05; **P < 0.01, hSLC13A2-OE versus GFP group, two-tailed unpaired Student’s t test.

DISCUSSION

As the biggest metabolic organ, how liver rewires metabolism to adapt to liver tumor growth builds up the foundation for cancer biology. Liver tumor actively secretes into or uptakes metabolites from circulation (22–24) and also influences the metabolic status of peripheral tissues including adipose tissue and skeletal muscle (25, 26). Therefore, untargeted metabolomic profiling of serum and liver tumor tissues from multiple HCC mouse models and clinical patients was used to identify metabolites involved in HCC progression (Fig. 1A and fig. S1, D and G). This analysis revealed consistent alterations in key TCA cycle intermediates in circulation and tumor tissues, which highlights the finding that fluctuations in serum metabolites reliably reflect the metabolic state within liver tumors. The findings underscore the utility of integrating serum and tissue metabolomics to capture tumor-specific metabolic remodeling, providing a foundation for investigating targeted metabolic interventions in HCC.

Through integrated transcriptomic analysis, SLC13A2 is identified as a consistently down-regulated transporter for TCA cycle intermediates in HCC tissues and cells (Figs. 1 and 2). SLC13A2 has high specificity for dicarboxylates with different affinities for substrates [0.35 mM Michaelis constant (K_m) for succinate and 0.6 mM K_m for citrate] (27). Although SLC13A2 has a greater affinity for succinate than citrate, the latter is much more abundant in HCC cells than the former (the content ratio is more than 2800-fold) (fig. S1L). There is also less intracellular α-KG than citrate (~14-fold). In contrast to the general belief that SLC13A2 is expressed at very low levels in the liver, we found that SLC13A2 mRNA is normally expressed in healthy liver tissue as a membrane transporter (table S1), indicating its role in the physiological homeostasis of liver function. Using [U-¹³C]-labeled citrate, it is confirmed that SLC13A2 transports citrate in HCC cells (Fig. 4G). These considered TCA cycle metabolites also have functions beyond mitochondria (28, 29). The activities of enzymes involved in glycolysis, lipogenesis, and gluconeogenesis are regulated by binding to citrate, for example, leading to the activation of 1,6-bisphosphatase (FBP1) and acetyl-CoA carboxylase alpha (ACACA) and the inhibition of phosphofructokinase (PFK) (30, 31). We landscape the metabolic profile to exclude these cytosolic functions of citrate in SLC13A2-mediated anti-HCC effect, as it is shown that blood glucose as well as the products of PFK and ACACA, fructose 2,6-bisphosphate (F2,6-BP) and malonyl-CoA, are not changed (figs. S2, B and D, and S6D). In addition, the synthesis of TCA cycle intermediates from SLC13A2-imported citrate is not changed (fig. S4A). Besides being a metabolic substrate and an allosteric modulator, citrate is also considered as a source of acetyl-CoA for lipogenesis and epigenetic modification (32, 33). While excluding fatty acid synthesis (fig. S6E), we provide evidence that SLC13A2 imports citrate, instead of succinate, which induces histone protein acetylation for global regulation of gene expression to suppress tumor growth (Fig. 5G and fig. S7, A and D). This aligns with previous reports showing that low intracellular citrate levels facilitate tumor growth (34).

The transported citrate by SLC13A2 results in increased intracellular acetyl-CoA (Fig. 4G, bottom) for protein acetylation including PKM2 and histone proteins (Figs. 4L and 5G). To strengthen physiological relevance, we treated cells with 250 and 500 μM citrate, equivalent to or about two- to fivefold of normal human serum citrate levels reported in the Human Metabolome Database (35), which allowed us to assess transporter-dependent effects under physiological or pathologically elevated conditions. Notably, citrate supplementation in various concentrations enhanced histone acetylation, which is further elevated with the overexpression of SLC13A2 (fig. S7D). It excludes the possibility that SLC13A2-imported citrate disrupts gradient across cell membrane and suppresses its transporting activity in HCC cells. Considering the high extracellular sodium concentration, SLC13A2-driven citrate transport is directing inward hepatocytes, distinguishing this active transport from passive diffusion.

SLC13A2 inhibits glycolysis and oxidative respiration in cancer cells, although it does not affect normal hepatocytes (Fig. 4, B and C, and fig. S5E). This disparate role is in accordance with our previous findings that SLC13A2-transported citrate secretes acetyl-CoA, which serves as a precursor for the cholesterol synthesis to endorse liver regenerative capacity (15). SLC13A2-imported citrate cannot be used for lipid synthesis (fig. S6E), but suppresses TCA cycle through the acetylation and degradation of PKM2 (Fig. 4, K and L, and fig. S6B). TCA cycle intermediates synthesis is decreased significantly by SLC13A2 in HCC cells (fig. S4, B and C). Conversely, TCA cycle intermediates from imported citrate are increased by SLC13A2 in normal hepatocyte (fig. S4A), which reflects the functionality of mitochondria and electron transport chain driving the hepatic regenerative process (36). Our findings point out that subtly tuning metabolism is a dynamic regulation lever, which represents an alternative strategy to target the differential metabolic contexts to combat cancer as well as promote organ regeneration in the meanwhile.

SLC13A2 overexpression reduces pyruvate levels while markedly decreases both PKM2 protein and pyruvate kinase activity (Fig. 4, P and Q), consistent with a blockade of glycolysis (Fig. 4C). Pyruvate levels are only modestly decreased in the KO model (Fig. 4P), suggesting that PKM2-mediated glycolytic regulation may contribute only partially to the tumor-suppressive effect of SLC13A2. In parallel, SLC13A2 markedly enhanced histone acetylation, including H3K9ac, H3K27ac, and H4K12ac (Fig. 5E and fig. S7D), indicating widespread epigenetic reprogramming (Fig. 6A). Integrative transcriptomic analysis showed that these acetylation changes were preferentially associated with genes involved in metabolism and tumor suppression, highlighting the coordinated interplay between metabolic and epigenetic programs. To date, few studies have linked SLC13A2 with cancer, and no studies have investigated its roles and mechanisms in cancer pathogenesis. According to the Human Protein Atlas (37), the expression of SLC13A2 has low cancer specificity. A high expression level of SLC13A2 is positively associated with a better overall survival rate in patients with renal cancer, suggesting its role in improving the prognosis of cancer patients. Transcriptome analysis of primary colorectal cancer tissues from patients revealed that SLC13A2 is a representative DEG that is responsible for liver metastasis (38). Another study conducted a genome-wide association analysis with familial hepatitis B virus (HBV)–related HCC in comparison with non-HCC controls with chronic HBV to identify susceptibility loci (39). The results revealed a single-nucleotide polymorphism cluster located at the 3′ end of SLC13A2, which is among the genes most strongly associated with familial HBV-associated HCC (39). This finding is in accordance with our findings suggesting that SLC13A2 is a hereditary genetic factor related to the pathogenesis of HCC.

MATERIALS AND METHODS

Mice

Six- to eight-week-old male C57BL/6J, Rosa26-flox-STOP-flox-Cas9 knockin, and NZG (NOD-Prkdc^−/−Il2rg^−/−) mice were purchased from the University of Yangzhou, the Jackson Laboratory, and Hangzhou Ziyuan Laboratory Animal Technology Co., Ltd. or bred in our animal facility. All animal experiments were approved by the Institutional Animal Care and Use Committee of China Pharmaceutical University (approval number: 2020-08-006).

Human specimens

Human HCC and adjacent liver tissues were obtained from patients undergoing liver resection at Nanjing First Hospital and Jiangsu Taizhou People’s Hospital, with informed consent and approval from the respective Ethics Committees (approval nos. KY20241223-KS-04 and 2022-03-017). Tumor tissues from eight patients (mean age of 68.1 years, both sexes) were collected. For PDX model generation, fresh tumors were cut into 2- to 3-mm³ fragments and subcutaneously implanted into immunodeficient mice under sterile conditions.

Cell culture

All cell lines were characterized by short tandem repeat profiling and checked free from mycoplasma. The human HCC cell lines HepG2 and Hep3B were obtained from the Institute of Biochemistry and Cell Biology of the Chinese Academy of Sciences (Shanghai, China). The mouse HCC cell line Hepa1-6 and mouse normal hepatocyte cell line AML12 were obtained from the American Type Culture Collection as described previously (40). Mouse primary hepatocytes were isolated from WT male C57BL/6J mice and cultured as described previously (40).

HCC model establishment

Mice were injected with oncogene plasmids through the tail vein for 4 to 7 s via a technique called hydrodynamic transfection with oncogene delivery and SB-mediated somatic integration for stable and long-term gene expression in hepatocytes, as previously described (12). The oncogenic plasmids used were pT3-EF1aH-c-MET (20 μg) and pT3-EF1aH-ΔN90-β-catenin (20 μg), along with SB transposase (5 μg). The oncogene plasmids were diluted with saline to 10% body weight (2 ml/20 g). For liver-specific SLC13A2 KO experiments, single-guide RNA (sgRNA) sequences were designed based on a CRISPR design tool targeting the first coding exon (table S2) and constructed together with TBG-Cre into AAV8 (5 × 10¹¹ genome copies per mouse), which was injected into the tail vein of Rosa26-flox-STOP-flox-Cas9 knockin mice (Rosa26-LSL-Cas9, the Jackson Laboratory, no. 024857) as previously described (41). For liver-specific SLC13A2 overexpression experiments, adult male mice (6 to 8 weeks of age, ~20 g body weight) were injected with AAV8 (1 × 10¹¹ genome copies per mouse) under the control of the liver-specific TBG promoter via the tail vein (AAV-TBG-GFP versus AAV-TBG-SLC13A2). To establish the AAV-cMYC/nRAS induced HCC model, 6- to 8-week-old male C57BL/6J mice were administered a combination of AAV8-TBG-cMYC and AAV8-TBG-nRAS viral vectors (2 × 10¹⁰ viral genome copies per mouse) via tail vein injection, as previously described (13).

To generate the STZ-HFD–induced HCC model, newborn C57BL/6J mice received a single subcutaneous injection of STZ (200 μg per mouse) on postnatal day 2. Beginning at 4 weeks of age, the mice were fed an HFD (60% kcal from fat) continuously for 28 weeks to induce nonalcoholic steatohepatitis (NASH) HCC. For the PDX model, fresh HCC tissues were cut into 2- to 3-mm³ fragments and subcutaneously implanted into the flanks of 6- to 8-week-old male NZG mice under sterile conditions. Tumor-bearing mice were monitored until the tumors became palpable, at which point they were randomly divided into experimental and control groups. To achieve SLC13A2 overexpression, mice were intratumorally injected with AAV9 vectors encoding SLC13A2 under the control of the CAG promoter (AAV9-CAG-SLC13A2, 5 × 10¹⁰ genome copies per tumor), or with control AAV9-CAG-GFP, in a total volume of 30 μl of phosphate-buffered saline (PBS). Tumor size was measured using calipers, and tumor volume was calculated using the formula volume = (length × width²) / 2. The plasmids were constructed using the primers shown in table S2 and used for the AAV package. All animal suffering was minimized during the experiments.

Reverse transcription quantitative polymerase chain reaction

Total RNA was extracted from cells or liver tissues using TRIzol reagent (Vazyme, Nanjing, China). cDNA was synthesized using the HiScript IV RT SuperMix for quantitative polymerase chain reaction (qPCR) (Vazyme, Nanjing, China). Reverse transcription qPCR was performed with the qPCR SYBR Green Master Mix (High ROX Premixed) kit (AG Bio, Hunan, China). The primers designed with National Center for Biotechnology Information (NCBI) primer BLAST are shown in table S2.

Immunoblotting and immunoprecipitation

Protein samples were mixed with 3× SDS loading buffer, separated using SDS–polyacrylamide gel electrophoresis, and then transferred onto a polyvinylidene fluoride membrane. The membranes were incubated with primary antibodies at 4°C for 12 to 16 hours and then with secondary antibodies for 1 hour at room temperature. Enhanced chemiluminescence reagents were used to visualize the target proteins. For the immunoprecipitation assay, cell extracts were incubated with primary antibodies or control immunoglobulin G in a rotating incubator overnight at 4°C, followed by incubation with protein A agarose beads (Beyotime, Shanghai, China) for another 4 hours. The immunoprecipitates were washed five times with lysis buffer and analyzed by immunoblotting.

Histological analysis and IHC staining

Formalin-fixed and paraffin-embedded mouse liver sections were stained with H&E. For IHC staining, the samples were dewaxed and hydrated. The antigen was retrieved with citric acid buffer (pH 6.0), blocked with horse serum, and incubated with Ki67 and SLC13A2 antibody at 4°C for 12 to 16 hours and then with a secondary antibody conjugated with horseradish peroxidase for 1 hour at room temperature. Images were taken with a BX53 light microscope (Olympus, Japan).

Plasmid construction

The plasmids PLVX-GFP-N1 and PLVX-GFP-SLC13A2 were constructed using the primers listed in table S2. An empty control vector for pReceiver-M02 (EX-NEG-M02) and a mouse PKM2 expression plasmid (EX-Mm04490-M02) were purchased from Guangzhou FulenGen Co., Ltd. (Guangzhou, China). The PKM2 K305Q mutant plasmid was generated by site-directed mutagenesis using the Mut Express II Fast Mutagenesis Kit V2 (Vazyme, China) according to the manufacturer’s protocol. All constructs were verified by Sanger sequencing.

Small interfering RNA and plasmid transfection

For small interfering RNA (siRNA) or plasmid transfection, Hepa1-6 cells were seeded at 40 or 80% confluence the next day and transfected with siRNAs or plasmids using Lipo8000 (Beyotime, Shanghai, China) according to the manufacturer’s instructions (siRNA sequences are shown in table S2). HepG2 cells were infected with adenovirus (pAdeno-MCMV-3×FLAG versus pAdeno-MCMV-SLC13A2-3×FLAG, OBiO, Shanghai, China) at a multiplicity of infection of 25. After 12 to 18 hours of incubation, the medium was changed to complete medium, and the cells were then cultured for the indicated period before the experiments.

Generation of CRISPR-Cas9 KO cell lines

For CRISPR-based gene KO, Hep3B cells were seeded at 60 to 70% confluence 1 day prior to transduction and infected with lentivirus carrying either control vector (Plenti-CRISPR v2-CTRL) or sgRNA targeting SLC13A2 (Plenti-CRISPR v2-sgSLC13A2). Viral supernatants were prepared by cotransfecting human embryonic kidney–293T cells with Plenti-CRISPR v2-sgRNA, psPAX2 (Addgene no. 12260), and pMD2.G (Addgene no. 12259) using polyethylenimine according to the manufacturer’s protocol. After 48 hours of infection, puromycin (3 μg/ml) was added for 5 to 7 days to select stably transduced cells. The sgRNA sequences used are listed in table S2.

Assays for cell proliferation in vitro

For the KD experiment, Hepa1-6 cells were transfected with siRNAs for 24 hours, digested for seeding in six-well plates at a density of 2.5 × 10³ cells per well, and then cultured for 7 days. For the overexpression experiment, Hepa1-6 and HepG2 cells were transfected with the overexpression plasmid or adenovirus, respectively, for 24 hours, digested for seeding onto a six-well plate at a density of 1 × 10³ cells per well, and cultured for 14 days. Cell colonies were fixed with 4% formalin for 15 min and stained with 0.4% crystal violet for 10 min. For the EdU incorporation assay, Hepa1-6 and HepG2 cells were incubated with EdU for 4 hours, fixed for 30 min, and analyzed using a commercial kit (Beyotime, Shanghai, China). Similarly, cell cycle arrest and cell viability were analyzed using a cell cycle analysis kit (Beyotime, Shanghai, China) and a CCK8 assay kit (Vazyme, Nanjing, China), respectively, according to the manufacturer’s instructions.

Measurement of the oxygen consumption rate

The cells were seeded into the wells of a Seahorse XF 96-well culture plate (Agilent, USA) at a density of 2 × 10³ cells per well for Hepa1-6 and AML12, and cultured overnight at 37°C and 5% CO₂. The oxygen consumption rate (OCR) was measured with the assurance that the cells were even and ~90% confluent in a monolayer. The sensor cartridge was incubated in a non-CO₂ incubator at 37°C overnight. Cells in growth media were replaced with fresh and prewarmed XF media (10 mM glucose, 1 mM sodium pyruvate, and 1 mM glutamine) and then incubated in a non-CO₂ incubator at 37°C for 45 min. Compounds were diluted with fresh XF media and injected for measurement as follows (1.5 μM oligomycin, 1 μM carbonyl cyanide p-trifluoromethoxyphenylhydrazone, and 0.5 μM rotenone/antimycin). The data were analyzed directly by Seahorse software. The basal and maximal OCRs were obtained for statistical analysis.

For extracellular acidification rate (ECAR) measurements, Hepa1-6 and AML12 cells were seeded at 2 × 10³ cells per well and incubated overnight at 37°C with 5% CO₂. Prior to the assay, growth media were replaced with prewarmed Seahorse XF Glycolysis Stress Test medium containing 2 mM glutamine, and cells were incubated in a non-CO₂ incubator at 37°C for 45 min. The assay was performed according to the manufacturer’s instructions. Glucose (10 mM), oligomycin (1 μM), and 2-deoxy-d-glucose (50 mM) were sequentially injected to assess glycolytic function, including basal glycolysis, glycolytic capacity, and glycolytic reserve. ECAR values were recorded and analyzed using Seahorse software.

Mitochondrial membrane potential assay

Hepa1-6 cells were seeded at a density of 5 × 10⁴ cells per well in a 24-well plate. After transfection with the plasmid for 48 hours, the cells were stained with JC-1 according to the manufacturer’s instructions (Beyotime, Shanghai, China). Images were taken using an inverted fluorescence microscope (IX71, Olympus, Japan) with the imaging software Olympus cellSens Standard software.

Isolation of subcellular fractions

The procedures were conducted on ice with prechilled buffers and equipment. Hepa1-6 cells in 10-cm dishes were transfected with adenovirus and harvested after 72 hours. After being washed with PBS three times, the cell pellets were suspended in homogenization buffer (225 mM D-mannitol, 75 mM sucrose, 0.1 mM EDTA, and 30 mM tris-HCl, pH 7.5) and then transferred to a 2-ml Dounce homogenizer to disrupt the cells with several strokes of the pestle. For each stroke, the pestle was pressed straight down the tube to maintain a firm and steady pressure. When checking the degree of homogenization with a phase-contrast microscope, eight to nine naked nuclei for every whole cell indicated a good result. The homogenate was transferred to a centrifuge tube and spun down at 500g for 5 min. While the supernatant was used for preparing the cytoplasm at 7000g for 10 min, the pellet was gently suspended in homogenization buffer followed by centrifugation at 700g for 10 min to collect the nuclei. The extracted subcellular fractions were confirmed by immunoblotting for cytosolic and nuclear tubulin and proliferating cell nuclear antigen (PCNA), respectively.

ATP measurement

Transfected cells were seeded into 96-well culture plates at a density of 2 × 10³ cells per well. After 72 hours, ATP production was detected with a CellCounting-Lite 2.0 Luminescent Cell Viability Assay Kit (Vazyme Biotech, Nanjing, China) by a plate reader (Beckman Coulter, Krefeld, Germany) according to the manufacturer’s instructions. To account for differences in cell number, ATP levels were normalized to cell viability as determined by a 3-(4,5-dimethylthiazol-2-yl)-2,5-diphenyltetrazolium bromide assay performed in parallel wells, and ATP levels were expressed as relative ATP production per viable cell.

Metabolomic analysis

Hepa1-6 cells were collected, and metabolites were extracted after adenovirus transfection for 72 hours. Cell metabolites were dissolved in 0.5 ml of ice-cold 80% methanol solution containing 4-chloro-DL-phenylalanine (1 μg/ml; Sigma-Aldrich, C6506). The samples were then incubated for 30 min in a −80°C freezer, followed by centrifugation at 13,000g for 15 min at 4°C, after which the supernatant was dried using a vacuum centrifugal concentrator (Thermo Fisher Scientific, Massachusetts, USA). Before analysis, the samples were reconstituted with methanol, followed by centrifugation at 18,000g for 10 min at 4°C twice, and untargeted metabolomics was conducted using a 6546 liquid chromatography (LC)/Q-TOF (Agilent, USA) instrument equipped with an electrospray ionization (ESI) source. Separation was achieved on an XBridge Premier BEH Amide VanGuard FIT column (2.5 μm, 4.6 mm by 150 mm, Waters, USA). MS data were acquired in negative ESI mode over a range of mass/charge ratio (m/z) 50 to 850. The mobile phase consisting of water (containing 0.3% ammonium acetate and 0.3% ammonia in phase A, v/v) and acetonitrile (phase B) was delivered at a flow rate of 0.4 ml/min. The total elution time was 36 min for the gradient program. The data were analyzed by Progenesis QI with conditions of P < 0.05, fold change > 2, and variable importance in projection > 1 to screen for differentially abundant metabolites. Metabolic pathway analysis was performed based on the MetaboAnalyst 5.0 platform.

Mitochondria were isolated from cultured cells for nontargeted metabolomics analysis using a differential centrifugation procedure from previously described methods (15). The mitochondrial pellets were washed once with isolation buffer and immediately quenched with 80% methanol (−80°C) containing internal standards prior to LC-MS processing. All steps were performed on ice or at 4°C to preserve metabolite integrity.

For in vivo samples, serum was collected from the AAV-cMYC/nRAS and HTVi models at the endpoint. After centrifugation at 5000g for 10 min at 4°C, 100 μl of serum was mixed with 400 μl of 80% methanol (precooled to −20°C) containing 4-chloro-dl-phenylalanine (1 μg/ml). After vortexing and incubation at −80°C for 6 hours, samples were centrifuged at 15,000g for 10 min. For tumor tissue samples, ~15 mg of liver tumor tissue was collected from AAV-cMYC/nRAS model, HTVi model, or from human HCC patients undergoing surgical resection with informed consent, weighed, and homogenized in 500 μl of precooled 80% methanol containing 4-chloro-dl-phenylalanine (1 μg/ml). The homogenates were incubated at −80°C for 4 hours and then centrifuged at 15,000g for 10 min at 4°C. The supernatant was dried and processed following the same LC-MS protocol as described above.

For the quantitative determination of metabolites, cell samples and mixed standards of gradient concentrations were measured on a TripleTOF 5600 system (Sciex, USA) by direct infusion. The instrument was set to acquire data over an m/z range of 100 to 2000 for the TOF-MS/MS scan. Aliquots of reconstituted supernatants (5 μl) were analyzed with an XBridge BEH Amide column (3.5 μm, 4.6 mm by 100 mm, Waters, USA) on an LC-30 high-performance liquid chromatography system (Shimadzu). The mobile phase consisting of water/acetonitrile (95:5, v/v) (containing 0.1% ammonium acetate and 0.3% ammonia in phase A, v/v) and acetonitrile (phase B) was delivered at a flow rate of 0.4 ml/min. The total elution time was 36 min for the gradient program. The concentrations of the tested substances were calculated by comparing the parent ion and daughter ion as well as the retention time with the standard curve formed by the gradient standards.

Stable isotope resolved metabolomics analysis

The culture medium was replaced with [U-¹³C₆] glucose (Cambridge) in glucose-free Dulbecco’s modified Eagle’s medium (DMEM) (Sigma-Aldrich), [U-¹³C₅] glutamine (MCE) in glutamine-free DMEM (Sigma-Aldrich), or 125 μM [U-¹³C₆] citric acid (Cambridge) in DMEM (Keygen) after the transfection of Hepa1-6 cells for 60 hours. The cells were labeled at the indicated time points (1 or 2 hours). An unlabeled culture was prepared in parallel by adding equal concentrations of unlabeled glucose, glutamine, or citrate to the media to identify unlabeled metabolites.

Cell metabolites were extracted with 80% methanol and completely concentrated to dryness using a Speedvac (Thermo Fisher Scientific, USA). The samples were redissolved in pure methanol and used for LC-MS/MS analysis. Standard metabolites were analyzed, and the MS/MS spectra of these compounds were manually confirmed according to METLIN Database or Human Metabolome Database (www.hmdb.ca).

RNA-seq and bioinformatics analysis

Liver RNA-seq was performed on DNBSEQ-T7 from GFP- or SLC13A2-OE mice tumor tissues in the HTVi model. The FASTQ files were aligned to the mouse reference genome (mm10) using the aligner STAR. RNA-seq FASTQ data have been deposited in the Sequence Read Archive (PRJNA1083485). The published RNA-seq raw data for heterogeneous liver cancer induced by diverse cancer driver genes were downloaded from the NCBI Sequence Read Archive with accession code PRJNA674008 (14). The published RNA-seq raw data of human HCC cancerous and paracancerous specimens were downloaded from the GEO database (GSE193567, GSE14323, GSE60502, and GSE14520) (42–45). TCGA HCC RNA-seq data were downloaded from the National Institutes of Health Genomic Data Commons repository (https://portal.gdc.cancer.gov/), which included 371 primary HCC patients in level 3. Differential gene expression analysis was performed using the Deseq2 package in the R program. The ENSEMBL numbers in the analysis results were converted to gene names using probes in the human gene annotation package, and genes with a log₂ fold change greater than 1 or −1 and a P value less than 0.05 were selected for plotting using the ggplot2 package. Heatmaps and volcano plots were generated by R. Gene enrichment analysis was performed using MetaScape (www.metascape.org). To annotate TFs among the DEGs identified in RNA-seq, a curated human TF list was downloaded from the JASPAR database (https://jaspar.genereg.net/), which provides high-confidence TF annotations based on DNA binding motif profiles. The gene symbols of all up-regulated and down-regulated DEGs were intersected with the JASPAR TF list using R, and the resulting TF subset was used for subsequent heatmap visualization and enrichment analysis. Public ChIP-seq datasets for histone acetylation modifications (H3K9ac, H3K27ac, and H4K12ac) in human H1 embryonic stem cells and HepG2 HCC cells were downloaded from the ENCODE project (www.encodeproject.org). These datasets included bigWig files representing genome-wide signal intensity and peak annotation files (narrowPeak or broadPeak). The data were aligned to the human reference genome (hg38). University of California, Santa Cruz (UCSC) Genome Browser (https://genome.ucsc.edu/) was used for visualization of histone acetylation signals across target gene loci, including CEBPA, BHLHE41, and TCF21. Custom tracks were uploaded through the UCSC “My Data” portal using the bigWig files to compare enrichment of histone acetylation marks at gene promoters. Representative signal peaks at TSSs were shown and used to confirm epigenetic activation associated with transcriptional induction.

ChIP-seq library preparation and sequencing

Hepa1-6 cells were infected with adenovirus expressing SLC13A2 or vector for 72 hours. Cells were then cross-linked with 1% formaldehyde at 4°C for 15 min, followed by quenching with 125 mM glycine. Chromatin was sheared into 200– to 500–base pair (bp) fragments. Immunoprecipitation was performed overnight at 4°C using 5 μg of pan-acetyl-lysine antibody and protein A/G magnetic beads (ABclonal, Wuhan, China). Immunoprecipitated chromatin was washed, reverse cross-linked, and purified.

ChIP DNA libraries were prepared using the DNA Lib Prep Kit (Wuhan, ABclonal, no. RK20250) according to the manufacturer’s protocol. Library quality was assessed by Qubit fluorometry (Thermo Fisher Scientific) and Bioanalyzer 2100 (Agilent). Sequencing was performed on the Illumina NovaSeq X Plus platform with paired-end 150-bp reads. Raw reads were subjected to quality control using FastQC and aligned to the mouse reference genome (mm10) using Bowtie2. Peaks were called using MACS2, and differential analysis was conducted using the edgeR package.

Proteomic analysis

Hepa1-6 cells were infected with adenovirus expressing SLC13A2 or vector and cultured for 60 hours. Cells were then harvested and lysed in immunoprecipitation lysis buffer supplemented with protease inhibitors. Total protein concentrations were quantified using a bicinchoninic acid (BCA) assay. Equal amounts of protein lysates were incubated overnight at 4°C with anti–acetyl-lysine antibody-conjugated agarose beads, followed by extensive washing to remove nonspecific proteins. The enriched acetylated proteins were then eluted, reduced, alkylated, and digested with trypsin for LC-MS/MS analysis. Raw MS data were analyzed using Proteome Discoverer 3.0 (Thermo Fisher Scientific) and searched against the mouse proteome database with lysine acetylation (+42.01 Da) set as a variable modification.

Enzymatic activity assay

Approximately 100 mg of liver tissue was ground with a 1:9 (w/v) ratio of PBS and centrifuged at 2500g at 4°C for 10 min. Hepa1-6 cells were transfected in six-well plates with plasmids for 72 hours. The collected cells were mixed with 500 μl of PBS and homogenized manually. The enzymatic activity of the homogenate was detected by ALT and AST kits (Jiancheng, Nanjing, China). Protein concentrations were quantified using the Bradford protein assay and used for the normalization of enzymatic activity.

Pyruvate kinase enzymatic activity assay

Pyruvate kinase activity was measured using a commercial assay kit (Solarbio, China). Approximately 20 to 30 mg of liver tissue was taken from each sample and homogenized on ice in the kit-provided extraction buffer. The homogenates were centrifuged at 8000g for 10 min at 4°C, and the clear supernatant was collected for activity measurement. Protein concentrations were quantified using the BCA protein assay and used for the normalization of enzymatic activity.

NADH/NAD⁺ quantification assay

NADH and NAD⁺ levels were measured using a commercial NAD⁺/NADH quantification kit (Elabscience, China). Approximately 30 mg of liver tissue was taken from each sample and homogenized on ice in the extraction buffer provided in the kit. The tissue lysates were centrifuged at 10,000g for 10 min at 4°C, and the resulting supernatants were transferred to 10-kDa ultrafiltration tubes for clarification. After ultrafiltration, samples were centrifuged again at 12,000g for 10 min at 4°C, and the final supernatants were collected for the determination of total NAD (NAD⁺ + NADH) and NADH according to the manufacturer’s instructions. NAD⁺ levels were calculated by subtracting NADH from total NAD. Protein concentrations were quantified using the BCA protein assay and used for normalization.

Pyruvate content assay

Pyruvate content was measured using an Amplex Red Pyruvate Assay Kit (Beyotime, China). Approximately 15 mg of liver tissue was homogenized on ice in the provided extraction buffer. Liver tissue samples were homogenized at 4°C using a tissue grinder in BeyoLysis Buffer A for metabolic assay at a ratio of 100 μl of buffer per 10 mg of tissue. The homogenates were centrifuged at 12,000g for 5 min at 4°C, and the resulting supernatants were collected for subsequent measurement. Fluorescence intensity was measured using a microplate reader, and pyruvate content was normalized to protein concentrations determined by the BCA protein assay.

Cycloheximide chase assay

To assess the half-life of the proteins, Hepa1-6 cells overexpressing SLC13A2 were treated with cycloheximide (30 μg/ml), a protein synthesis inhibitor, and harvested at 0, 3, 6, or 9 hours for immunoblot analysis (46). The densitometry of the bands was quantified using ImageJ.

Statistical analysis

The data are expressed as the means ± SEM. The differences between two groups were statistically analyzed using a two-tailed Student’s t test. Differences from the growth curve study were analyzed by two-way analysis of variance (ANOVA), followed by Bonferroni multiple comparisons. P < 0.05 was considered to indicate statistical significance.

Supplementary Materials

This PDF file includes:

Figs. S1 to S7

Tables S1 to S3