Chemicals and reagents were acquired from designated suppliers: SC79 (Cat No. T2274; TargetMol, Shanghai, China), DMSO (Cat No. D4540; Sigma-Aldrich, Darmstadt, Germany), Sorafenib (Cat No. HY-10201; MCE, Monmouth Junction, NJ, USA), Lenvatinib (Cat No. HY-10981; MCE, Monmouth Junction, NJ, USA), LY345899 (Cat No. T15827; TargetMol, Shanghai, China), FBS (Cat. No. 10099141C; Gibco, Grand Island, NY, USA), Lipofectamine RNAiMAX Transfection Reagent (Cat No. 13778500; Thermo Fisher Scientific, Waltham, MA, USA), CCK-8 Kit (Cat No. C008; 7Sea, Shanghai, China), Detergent Compatible Bradford Protein Assay Kit (Cat No.P0006C; Beyotime, Hangzhou, China), MycoBlue Mycoplasma Detector (Cat No. D101-01; Vazyme, Nanjing, China), Cell-Light EdU In Vitro Kit (Cat No. C10310; RiboBio, Guangzhou, China), TRIzol reagent (Cat No. 15596026; Thermo Fisher Scientific, Waltham, MA, USA), SYBR Premix Ex Taq Kit (Cat No. AKA1105; TaKaRa, Tokyo, Japan), PrimeScript II 1st Strand cDNA Synthesis Kit (Cat No. 6210A; TaKaRa, Tokyo, Japan), 1X SDS-PAGE Sample Loading Buffer (Cat No. P0015A; Beyotime, Hangzhou, China), SuperSignal West Femto (Cat No. 34094; Thermo Fisher Scientific, Waltham, MA, USA), BSA (Cat No. 4240GR500; Saiguo, Guangzhou, China).

Antibodies were obtained from the following sources: p-AKT (Ser473) (1:1000, Cat No. 4060; CST, Danvers, MA, USA), AKT (1:1000, Cat No. 4691; CST, Danvers, MA, USA), MTHFD2 (1:1000, Cat. No. H00010797-M01; Abnova, Taipei, China), Ki-67 (1:500, Cat No. ET1609-34; HUABIO, Hangzhou, China), and $\beta$-actin (1:3000, Cat No. A5441; Sigma-Aldrich, Darmstadt, Germany).

The HCC tissue microarray (Cat No. HLiv-HCC180Sur-05) was obtained from Outdo BioTech (Shanghai, China). The HLiv-HCC180Sur-05 contained 95 HCC tissues and 85 adjacent tissues, 85 of which were paired tissues. Detailed clinicopathological features of the cohorts were also provided by Outdo BioTech (Shanghai, China). And other 35 pairs of matched HCC tissue samples were obtained from Zhoushan Hospitital (Zhoushan, China). An Ethics Committee approved by the appropriate hospital obtained written informed consent from each subject or subject’s guardian.

Tumor tissues and adjacent normal tissues were fixed with 10% neutral formalin for a duration of 48 h, following standard procedures for paraffin embedding. Subsequently, the tissues were sliced into sections measuring 4 µm. The tissue sections were then subjected to treatment with 1 mM EDTA (pH = 8.0) and 3% H${}_2$O${}_2$ for a duration of 10 min. Following antigen retrieval using citrate buffer (pH = 6.0), the sections were blocked with 1% BSA for 30 min and subsequently incubated with anti-MTHFD2 antibody and anti-Ki-67 antibody at a temperature of 4 °C overnight. The immunostaining procedure employed the streptavidin-biotin peroxidase complex method (K060911, Dako). After rinsing in PBS, the peroxidase reaction was visualized through incubation with 3, 3${}^{\prime }$-Diaminobenzidine tetrahydrochloride (DAB). The results were assessed by grading the proportion of MTHFD2 positively stained cells was graded as follows: less than 5%–grade 0; 5% to 25%–grade 1; 26% to 50%–grade 2; 51% to 75%–grade 3; more than 75%–grade 4. The staining grades were categorized as follows: 0 denoted no staining; 1 indicated weakly positive staining; 2 represented medium staining; 3 denoted strong staining. Based on the sum of the two results: a total score of 0–1 indicated no staining (–), 2–3 denoted weakly positive (+), 4–5 represented medium positive (++), and 6–7 indicated strongly positive (+++). Additionally, the combination of (–, +) was considered as low expression, while (++, +++) indicated high expression [15]. The Ki-67 labeling index (LI) was evaluated by determining the percentage of cells exhibiting positively staining. With the threshold value of 10%, HCC lesions were categorized into the low Ki-67 group (Ki-67 LI $\le$10%) and the high Ki-67 group (Ki-67 LI $>$10%) [16]. The quantification of positively stained cells was performed by two pathologists independently, following a blinded manner, and a consensus was reached to establish the final outcome.

The human HCC cell lines, SMMC-7721 and HepG2, were obtained from the National Collection of Authenticated Cell Cultures. Cells were cultured in Dulbecco’s Modified Eagle Medium (DMEM) supplemented with 10% fetal bovine serum (FBS) at 37 °C in a humidified atmosphere of 5% CO${}_2$ in an incubator. All cell lines were validated by STR profiling and tested negative for mycoplasma (MycoBlue Mycoplasma Detection Kit). The transfection of siRNA was carried out utilizing the Lipofectamine™ RNAiMAX Transfection Reagent. All experimental procedures were performed according to the manufacturer’s protocols.

In accordance with the manufacturer’s protocol, the cells were transfected with 100 nM of siRNAs targeting MTHFD2 or the control. The target sequences of siRNAs were shown in the Supplementary Material.

A quantity of 5 µg of RNA was extracted from HCC tumor tissues and matched adjacent tissues using TRIzol reagent. Subsequently, the entire RNA was transcribed into cDNA using the PrimeScript II 1st Strand cDNA Synthesis Kit. The SYBR Premix Ex Taq Kit was employed for conducting quantitative PCR with the Applied Biosystems 7500 Real-Time PCR System (Applied Biosystems, Hammonton, NJ, USA). The relative expression was obtained using the 2${}^{-△△\text{Ct}}$ method by using GAPDH as the reference gene. The primer sequences of MTHFD2 and GAPDH were shown in Supplementary Material.

Cells were washed twice using ice-cold PBS, and protein extracts were prepared using RIPA reagent. The concentration was quantified using the Bradford assay. Subsequently, 30 µg of proteins were subjected to SDS-PAGE and transferred onto PVDF membranes. Protein analysis was conducted using antibodies against MTHFD2, p-AKT (Ser473), AKT, and $\beta$-actin were used for protein analysis. Bands were identified using enhanced chemiluminescence.

A total of 3 $\times$ 10${}^3$ cells were seeded onto 96-well plates before experiments. After incubating at 37 °C for 24 h, cells were treated with sorafenib or lenvatinib for 72 h. Cell proliferation activity was conducted using the CCK-8 Kit. Specifically, 10 µL of cell counting solution was added to each well, followed by an incubation at 37 °C for 2 h. Subsequently, the absorbance of the solution was measured at a wavelength of 450 nm.

The EdU proliferation test was performed according to the manufacturer’s protocol. After incubation at 37 °C, cells were added with 50 µM EdU and incubated for 2 h. Cells were then treated with 4% paraformaldehyde and then subjected to staining using Apollo Dye Solution. Nucleus was stained with hoechst33342. The images were taken using Leica Application Suite (version 3.0.0, Leica Microsystems, Wetzlar, Germany).

The bioinformatics analysis software and website were presented in the Supplementary Material.

All data were analyzed with GraphPad Prism 6.0 (GraphPad Software, Inc., San Diego, CA, USA). The bar graphs depict the fold changes or percentage relative to the control, along with a standard deviation of three independent experiments. Student’s t-test was employed for the analysis of normally distributed data, while one-way ANOVA with Tukey’s multiple comparisons was utilized for the analysis of more than two groups. Survival was estimated by the Kaplan–Meier method and compared by the log-rank test. Statistical significance was determined as a p value below 0.05. Levels of significance were denoted as *p 0.05, **p 0.01, ***p 0.001, ****p 0.0001.

To investigate the role of folate metabolism-related genes in liver cancer, we first analyzed the changes in mRNA expression profiles and kyoto encyclopedia of genes and genomes (KEGG)/gene ontology (GO) pathways using the gene expression omnibus (GEO) database (GSE112790) (Fig. 1A; Supplementary Fig. 1). Then, we focused our initial efforts on 5 differentially-expressed genes (DEGs) with overlapping functions with folate metabolism-related genes (Fig. 1B). Of these, two genes (TYMS and MTHFD2) were upregulated and three genes (ALDH1L1, MTHFD2L, and FTCD) were reduced in HCC tumors relative to normal liver tissue (Fig. 1C,D).

Fig. 1.

Analysis of folate metabolism-related genes in liver cancer. (A) Heat map of differentially expressed mRNAs in GEO112790. (B) Venn diagram of the differentially expressed genes (DEGs) in GSE112790 and genes associated with the folate cycle (C) Volcano plots of the DEGs. (D) Differential expression analysis of five genes in GSE112790. (E) The LASSO partial likelihood deviance plot and the coefficient profiles of these 5 genes. *p 0.05, **p 0.01, ****p 0.0001.

Next, we performed LASSO regression analysis for the 5 genes related to folate metabolism to establish a prognostic model (Fig. 1E). Specifically, the model is a RiskScore formula containing multiple genes where each gene is assigned a numeric weight, a negative number represents a protective gene (FTCD), and the positive number represents risk genes (TYMS, MTHFD2, ALDH1L1, and MTHFD2L). The calculated Riskscore = (0.0129) $\times$ MTHFD2 + (–0.146) $\times$ FTCD + (0.0449) $\times$ ALDH1L1 + (0.1948) $\times$ MTHFD2L + (0.1475) $\times$ TYMS.

These 5 DEGs were subsequently used to divided HCC patients into high-risk and low-risk groups (Fig. 2A). Their survival status is shown in Fig. 2B. There was a highly significant difference in survival between the high and low-risk groups (p = 1.86 $\times$ 10${}^{-5}$). Prognostic survival prediction was performed at 1, 3, and 5 year intervals, and a prognostic model, specifically ROC-AUC score for 1 year, showed a notable separability (Fig. 2C).

Fig. 2.

MTHFD2 is overexpressed in hepatocellular carcinoma (HCC) tissues. (A) Distribution of risk scores for the 5-genes model. (B) Kaplan-Meier survival curve of the patients in the high-risk and low-risk groups for OS. (C) Time-related ROC analysis proved the prognostic performance of the risk score in the training set. (D) The methylenetetrahydrofolate dehydrogenase 2 (MTHFD2) mRNA expression between tumor and normal tissues was detected by qRT-PCR (n = 35). (E) Immunohistochemical staining images of the tumor and normal tissues. (F) Kaplan-Meier analysis according to different combinations of MTHFD2 and Ki67 expression. ***p 0.001. ROC, receiver operating characteristic; OS, overall survival; qRT-PCR, quantitative real-time reverse transcription polymerase chain reaction.

MTHFD2 is a key enzyme in the folate-mediated metabolism pathway that is often hyperactivated in human cancers [17, 18, 19]. Thus, we examined its expression in 35 pairs of matched HCC/normal tissue samples using qRT-PCR. As shown in Fig. 2D, MTHFD2 mRNA abundance was significantly higher in HCC tissues compared with their adjacent counterparts in 18 out of 35 (51.4%) cases (p 0.01). In addition, IHC analysis of 85 pairs of patient-matched HCC/normal tissue samples included in a HCC tissue microarray indicated that higher staining intensity was observed in 63 (74.1%) of HCC tissues compared to matched normal tissues (Fig. 2E). The expression of MTHFD2 was positively related to histologic grade and Ki-67 expression through clinic-pathological correlation analysis. Here we also noted a significant (p 0.0001) correlation between the expression of Ki-67 and histologic grade (Table 1). Thus, data obtained indicate that MTHFD2 is overexpressed in HCC tissues and potentially involved in disease progression.

Ki-67 is a traditional proliferation marker widely used in studies of various cancer types, including HCC. Furthermore, tumor histologic grade has also been found to be tightly associated with tumor prognosis [20, 21]. Consequently, we examined the influence of MTHFD2 and Ki-67 expression on the survival of patients using Kaplan-Meier analysis with the long-rank test. Findings obtained indicated that the poorer overall survival was associated with the combination of high MTHFD2 and high Ki-67 levels (p 0.05), while the expression of MTHFD2 or Ki-67 alone showed no statistically significant difference in HCC patient survival (Fig. 2F). These findings suggest MTHFD2 is a potential prognostic biomarker for HCC.

Next, we explored the biological significance of MTHFD2 expression on various signaling pathways using gene set enrichment analysis (GSEA) analysis, and found that comparing the MTHFD2-high versus the MTHFD2-low groups indicated a significantly enrichment in multiple proliferation-related pathways. It is generally believed that the $|$NES$|$ $>$1, p 0.05 pathway is significantly enriched. As shown in our GSEA analysis (Fig. 3A), compared with the MTHFD2-low expression group, the MTHFD2-high group was significantly associated with the proliferation-related pathways, including VEGF signaling, ERBB signaling and cell adhesion.

Fig. 3.

Downregulation of MTHFD2 inhibits the proliferation of HCC. (A) Gene set enrichment analysis (GSEA) showed that several Proliferation-related signal pathways were inhibited in the MTHFD2-low group. (B) Western blotting analysis showed the MTHFD2 knockdown efficiency with two unique siRNAs (#1, #2) in SMMC-7721 and HepG2 cells. (C) Effects of MTHFD2 knockdown on cell proliferation were assessed by CCK-8 cell proliferation assay. (D) DNA synthesis in SMMC-7721 and HepG2 cells was measured using the EdU assay. Photographs were taken under an optical microscope with a magnification of 400$\times$. Experiments were performed three times independently and the data were represented as the means $\pm$ SD. **p 0.01, ***p 0.001.

Therefore, to evaluate the effect of MTHFD2 on HCC cell proliferation, MTHFD2 was knocked down in SMMC-7721 and HepG2 HCC cell lines (Fig. 3B). As shown in Fig. 3C,D, knockdown of MTHFD2 considerably decreased cell proliferation ability, suggesting a significant role in modulating HCC cell activity. Additionally, the MTHFD2 inhibitor LY345899 also reduced the proliferation of HCC (Supplementary Fig. 3).

To investigate the mechanism underlying the effect of MTHFD2 on cell proliferation, bioinformatic analysis was utilized to analyze the correlation between signaling pathways and MTHFD2 expression. Firstly, the expression profile data and corresponding clinical information from liver cancer patients in the cancer genome atlas (TCGA) dataset were downloaded. Second, bioinformatic analysis was applied and results show that PI3K/AKT signaling is the most enriched pathway associated with MTHFD2 expression with 90 up-regulated genes participating in this signaling axis, and its GeneRatio score is the highest among all signals (Fig. 4A,B; Supplementary Fig. 2).

Fig. 4.

MTHFD2 regulates HCC cell proliferation via PI3K/AKT signaling pathway. (A) Volcano plots of the DEGs between MTHFD2-high and MTHFD2-low groups. (B) Results of KEGG enrichment analysis of DEGs between high-expression and low-expression groups of MTHFD2. (C) Correlation analysis between MTHFD2 and signal pathways. (D,E) Protein levels of p-AKT, AKT, and MTHFD2 were analyzed using Western blotting in cells transfected with MTHFD2 siRNA, as well as cells treated with or without the AKT pathway activator SC79 (15 µM, 24 h). (F) CCK-8 assay was used to assess cell proliferation in cells transfected with MTHFD2 siRNAs, with or without the AKT activator SC79 (15 µM, 24 h). (G) CCK-8 assay was used to assess cell proliferation in cells incubated with LY345899 (50 µM, 72 h), with or without SC79 (15 µM, 24 h). Experiments were performed three times independently and the data were represented as the means $\pm$ SD. **p 0.01, ***p 0.001.

The correlations between MTHFD2 and pathway scores were examined using Spearman analysis. We selected four pathways that were significantly correlated with tumor malignancy and analyzed their correlation with MTHFD2 expression. Results indicated that all four pathways showed a significant positive correlation with MTHFD2 (r $>$ 0), and the PI3K/AKT signaling axis displayed the most significant correlation with MTHFD2 expression (r = 0.61) (Fig. 4C). As shown in Fig. 4D, the levels of AKT phosphorylation at Ser473 in MTHFD2-silenced HCC cells were significantly lower than those in siRNA control cells. By incubating HCC cells with the novel AKT activator SC79, inhibitory effects on proliferation displayed by MTHFD2 knockdown or inhibited cells were notably reversed (Fig. 4E–G). Thus, knockdown of MTHFD2 suppresses cell growth signaling mediated by PI3K/AKT signaling in HCC cells.

To confirm whether MTHFD2 expression impacts HCC cell chemosensitivity, MTHFD2 was knocked down in SMMC-7721 and HepG2 cell lines followed by treatment with sorafenib or lenvatinib. As shown in Fig. 5 and Supplementary Fig. 4, MTHFD2 knockdown resulted in decreased proliferative ability compared with control groups, indicating that knockdown of MTHFD2 effectively sensitizes HCC cells to sorafenib and lenvatinib.

Fig. 5.

Knockdown of MTHFD2 sensitized HCC cells to sorafenib and lenvatinib treatment. (A) Proliferation of SMMC-7721 and HepG2 cells treated with MTHFD2 siRNAs in the presence of control, different concentrations of sorafenib or lenvatinib. (B) DNA synthesis in SMMC-7721 and HepG2 cells treated with MTHFD2 siRNAs in the presence of DMSO, sorafenib, or lenvatinib. Photographs were taken under an optical microscope with a magnification of 400$\times$. Experiments were performed three times independently and the data were represented as the means $\pm$ SD. *p 0.05, **p 0.01, ***p 0.001.