The Protein Kinase Activity of NME7 Activates Wnt/β-Catenin Signaling to Promote One-Carbon Metabolism in Hepatocellular Carcinoma

Hepatocellular carcinoma (HCC) is one of the most common malignancies (1). Aberrant activation of Wnt/β-catenin signaling plays important roles in the occurrence and progression of HCC (2–5). Therefore, revealing the activation mechanism and the downstream biological events of Wnt/β-catenin signaling is highly important for treatment.

β-Catenin is the core molecule of the Wnt/β-catenin signaling pathway (6). In the resting state, β-catenin is phosphorylated by the degradation complex (composed of Axin, APC, GSK3β, CK1α, etc.), subsequently recognized by the E3 ligase β-TrCP, and degraded through the ubiquitination pathway (6, 7). When the Wnt ligand binds to its receptors (Frizzled and LRP6) on the cell membrane, the carboxyl terminus (C-terminus) of Frizzled recruits Dvl2, and Dvl2 recruits Axin, resulting in dissociation of the degradation complex. Therefore, β-catenin accumulates in the cytoplasm and then enters the nucleus, where it forms a complex with TCF4 and activates the transcription of downstream genes (6). The aberrant activation of Wnt/β-catenin signaling in 60% to 70% of HCC clinical samples is mainly due to upregulation of Wnt ligand expression and Axin-inactivating mutations or β-catenin constitutive activation mutations (8–12).

In addition to regulating tumor cell growth and invasion, Wnt/β-catenin signaling pathway also affects the metabolism of tumor cells (13, 14). In HCC, the glutamine synthetase gene is a downstream target of constitutively activated β-catenin (15), and tumors that carry constitutively activated mutations of β-catenin are very sensitive to mTOR inhibitors (15). In addition, HMGCR, the rate-limiting enzyme in the mevalonate pathway, is a target gene of Wnt/β-catenin signaling in pancreatic cancer (16), and statin, an inhibitor of HMGCR, inhibits tumorigenesis in mouse models of pancreatic cancer (16). These observations indicate that the metabolic targets of Wnt/β-catenin signaling are helpful for cancer therapy. In HCC, the expression of enzymes involved in one-carbon metabolism, including SHMT1, SHMT2, GLDC, and methylenetetrahydrofolate dehydrogenase (MTHFD2), is dysregulated (17–21). These enzymes catabolize serine, tryptophan, glycine, and histidine to produce one-carbon units (22, 23), enhancing pyrimidine and purine synthesis, and promoting the rapid division of tumor cells (22). However, the regulatory mechanism of one-carbon metabolism in HCC remains largely unknown.

Kinases are among the most common antitumor therapeutic targets. GSK3β and CK1α in the Wnt/β-catenin signaling pathway were first identified in the context of developmental biology (24). Due to their toxicity and side effects, inhibitors of GSK3β and CK1α are unusable in cancer treatment. Therefore, in the context of tumor biology, identifying the abnormally expressed kinases in tumor tissues that activate the Wnt/β-catenin signaling pathway will provide novel therapeutic targets. By analyzing data from our previous screening of the kinome (25), we found that nucleoside diphosphate kinase 7 (nonmetastatic cells 7, NME7) is a potential regulator of the Wnt/β-catenin signaling pathway. This study aimed to explore the mechanism by which the protein kinase activity of NME7 regulates the Wnt/β-catenin signaling pathway and one-carbon metabolism.

HCC cell lines (HepG2, Huh7, QGY-7701, 7404, 7721, and MHCC97H), normal hepatocytes (LO2 and HHL-5), HEK293 and HEK293T cells were obtained from the Cell Bank of the Chinese Academy of Sciences. PVTT cells were established as described previously (26). QGY-7701 cells were cultured in RPMI1640 medium, whereas other cells were cultured in DMEM. FBS (10%) and antibiotics (100 U/mL penicillin and 100 μg/mL streptomycin) were added to all media. All cells were cultured in a constant temperature incubator (5% CO₂, 37°C). Cell transfection was performed using Lipofectamine 8000 according to the instruction manual. The following reagents were used for cell experiments: the MTHFD2 inhibitor LY345899 (MedChemExpress, HY-101943) and doxycycline hydrochloride (dox; Sangon, A600889). Wnt3a conditioned medium was prepared from L-Wnt3a cells according to the method described by the ATCC. All cells were freshly thawed before experiments and cells older than 8 weeks were not used in the study. All cell lines were frequently assayed for Mycoplasma using Mycoplasma Stain Assay Kit (Beyotime) to ensure they were free of Mycoplasma contamination.

HCC tissues and matched adjacent nontumor tissues were collected from the Shanghai Eastern Hepatobiliary Surgery Hospital, Second Military Medical University, with informed consent from the patients. This study was approved by the Ethics Committee of the Second Military Medical University.

The coding sequence (CDS) of NME7 was inserted into the pLVX-IRES-puro vector, whereas the coding sequence of MTHFD2 was inserted into the pLenti-CMV-GFP-Hygro vector. The shRNAs targeting NME7 and MTHFD2 were designed with the help of the Sigma website and cloned into the pLKO.1-puro vector. To construct shRNA-resistant NME7 (NME7^R) and shRNA-resistant MTHFD2 (MTHFD2^R), the nucleotide sequence GGAGTAGTGACCGAATATC in NME7 CDS targeted by shRNA-NME7–2# was synonymously mutated into GGTGTAGTAACTGAATACC, the nucleotide sequence CGAGAAGTGCTGAAGTCTAAA in MTHFD2 CDS targeted by shRNA-MTHFD2–2# was synonymously mutated into AGAGAAGTCCTGAAATCTAAG, respectively. Then, the coding sequence of NME7^R and MTHFD2^R was inserted into the pLenti-CMV-GFP-Hygro vector. The list of shRNA sequences is shown in Supplementary Table S1. Lentivirus was packaged in HEK293T cells with psPAX2 and pMD2.G as the packaging plasmids. After being concentrated in PEG8000, the collected virus-containing solution was centrifuged for 1 hour (4°C, 1,600 × g). After removing the supernatant, the precipitated virus was dissolved in 2 mL of DMEM. Cells were seeded into a 6-well plate at a density of 50% to 60%. The next day, 400 μL of the lentiviral suspension was added to the cells and placed in a constant temperature incubator for incubation overnight. Two days later, the cells were cultured with puromycin (1 μg/mL) for 4 days. Then, the resistant cells were pooled, and Western blotting was used to detect the expression of NME7 and MTHFD2.

TRIzol (Invitrogen) was used to extract RNA, and 1 μg of RNA was then reverse transcribed into cDNA using a PrimeScript RT Kit (Takara) according to the instructions. A SYBR Green Kit and CFX96 real-time fluorescent qPCR detection system (Bio-Rad) were used for qPCR with 18S rRNA as the internal reference. The 2^−ΔΔC_t method was used to calculate the relative expression levels of the target genes. See Supplementary Table S1 for the details of the primer sequences used in the experiment.

Nude mice ages 4 to 6 weeks (4 mice/group) were used. Control and PVTT cells (2.5 × 10⁶) with NME7 knockdown were injected subcutaneously at each site. The tumor size was measured weekly with a Vernier caliper. The tumor volume was calculated according to the following formula: volume = (length × width²)/2. Mice were sacrificed in the 7th week after the start of the experiment to harvest tumors. The study was approved by the Animal Ethics Committee of Central South University.

Hydrodynamic tail vein injection was performed as described previously (27). The plasmids used in each mouse (male, C57BL/6J background) were diluted with 1 mL of Ringer's solution (each control mouse was injected with 20 μg of PT3 plasmid, 20 μg of sleeping beauty transposase plasmid, and 20 μg of PT3-Myc plasmid, whereas each experimental mouse was injected with 20 μg of PT3-NME7 plasmid, 20 μg of sleeping beauty transposase plasmid, and 20 μg of PT3-Myc plasmid). Plasmids were injected into 6- to 8-week-old mice through the tail vein within 5 to 7 seconds. Twelve weeks later, mouse livers were harvested to assess tumorigenicity.

DEN-induced liver cancer model was performed as described previously (28). To induce liver tumors, a single dose of DEN (50 mg/kg of body weight, Sigma-Aldrich) was injected into 2-week-old male mice (129 SV). Ten months later, mice were killed for examination of the livers, which were fixed with 4% formalin for histologic analysis.

Details about H11-Myc mice were described previously (29). The H11-Myc mice (C57BL/6J background) harboring a loxp-stop-loxp (LSL) cassette upstream of the Myc coding sequence were obtained from Shanghai Biomodel Organism Science & Technology Development Co., Ltd. By crossing these mice with Alb-Cre mice, the LSL cassette was removed, and liver cancer was induced at the age of 6 weeks. The study was approved by the Animal Ethics Committee of Central South University.

Data were expressed as the mean ± SD and were analyzed using the t test. The chi-squared test was used to analyze the relationship between clinicopathologic data and the NME7 and MTHFD2 scores. Survival curves were plotted with the Kaplan–Meier method, whereas the log-rank test was used for analysis. GraphPad Prism 8 and SPSS 17.0 were used for statistical analysis.

The protocol for the collection of HCC tissues was approved by the Ethics Committee of Shanghai Eastern Hepatobiliary Surgery Hospital (Second Military Medical University), and the patients' written informed consents were obtained. All animal experiments were approved by the Ethics Committee of Xiangya Hospital and were in accordance with the National Policy on Humane Care and Use of Laboratory Animals.

Data were generated by the authors and deposited in a repository.

Details about “IHC, Western blot analysis, nuclear protein extraction, immunoprecipitation (IP), luciferase reporter assay, in vitro kinase activity assay, ubiquitination assay, ELISA, chromatin immunoprecipitation (ChIP) assay, soft agar colony formation assay, sphere formation assay, CCK8 assay, crystal violet staining assay, EdU assay, organoids, RNA sequencing” are provided in the Supplementary Materials and Methods.

Kyoto Encyclopedia of Genes and Genomes pathway analysis of the potential regulatory kinases in the Wnt/β-catenin signaling pathway was performed (those with a change in the Topflash reporter activity of greater than 1.5-fold or less than 0.75-fold; ref. 25), and showed that most of these changed kinases were enriched in metabolic pathways (Supplementary Table S2; Supplementary Fig. S1A). Further analysis using the BIOCYC database revealed significant enrichment of the kinases (NME2, NME3, NME5, NME6, and NME7) in nucleotide metabolism (Supplementary Table S3; Supplementary Fig. S1B). Mining the UALCAN database revealed that the expression of NME2, NME6, and NME7 was inversely correlated with the survival of patients with HCC (Supplementary Figs. S1C–S1G). Furthermore, NME7 activated the Topflash reporter more strongly than did NME2 and NME6 (Supplementary Figs. S1H–S1I). Therefore, NME7 was further studied.

In GSE76427, GSE63898, and GSE25097, three datasets of the Gene Expression Omnibus (GEO) database, the expression of NME7 was significantly upregulated in HCC samples (Supplementary Fig. S2A). Analysis of the UALCAN database also suggested that the expression of NME7 is upregulated in liver cancer (Supplementary Fig. S2B). In addition, higher NME7 protein levels were observed in HCC cell lines (Fig. 1A).

Next, the NME7 mRNA expression level in 90 HCC tissues and matched adjacent tissues was examined and analyzed. The mRNA level of NME7 in HCC tissues was increased (Fig. 1B), and the mRNA level of NME7 in 81.1% of HCC tissues was higher than that in matched adjacent tissues (Fig. 1C). Moreover, the mRNA level of NME7 in HCC tissue was positively correlated with tumor size (Supplementary Table S4). Subsequently, the NME7 protein level in an HCC tissue array was examined and analyzed. Consistently, the NME7 protein level in the HCC tissues was increased (Fig. 1D), and the NME7 protein level in 63% of the HCC tissues samples was higher than that in the matched normal liver tissue samples (Fig. 1E and F). Furthermore, the NME7 protein level in HCC tissues was positively correlated with tumor size but was not correlated with the levels of AFP, ALP, ALT, AST, CA199, CEA, GGT, or AFU in the patients (Supplementary Table S5; Supplementary Fig. S2C).

Western blot analysis of the NME7 protein level in 24 HCC tissues and matched adjacent tissues also showed that the NME7 protein level was significantly increased in HCC tissues (Fig. 1G–I). In addition, the expression of NME7 was upregulated in both the c-Myc-driven and DEN-induced mouse models of liver cancer (Fig. 1J and K). Moreover, analysis via the Kaplan–Meier Plotter databases indicated that the expression level of NME7 was negatively correlated with the overall survival and recurrence-free survival of patients (Fig. 1L and M). These findings suggest that the upregulation of NME7 in HCC is very likely to promote the progression of this malignancy.

To study the functions of NME7 in HCC, NME7 was overexpressed in HHL-5, MHCC97H, QGY-7701, and Huh7 cells (Supplementary Fig. S3A), and overexpression of NME7 promoted cell growth (Supplementary Fig. S3B). Furthermore, overexpression of NME7 enhanced colony formation (Supplementary Figs. S3C and S3D), and promoted the anchorage-independent growth of HCC cells (Supplementary Figs. S3E and S3F).

Then, we induced knockdown of NME7 expression with dox by using two independent interference sequences (Fig. 2A). The results of the CCK8 (Supplementary Figs. S4A and S4B), EdU (Supplementary Figs. S4C and S4D), colony formation (Fig. 2B and C; Supplementary Fig. S4E), and soft agar anchorage-independent growth assays (Fig. 2D; Supplementary Fig. S4F–S4H) showed that knocking down NME7 inhibited the proliferation, colony formation, and anchorage-independent growth of HCC cells. Moreover, knockdown of NME7 inhibited sphere formation (Fig. 2E). In the subcutaneous tumorigenesis assay, knockdown of NME7 expression significantly inhibited the tumorigenicity of PVTT cells in nude mice (Fig. 2F and G).

To confirm the specificity of the shRNA targeting NME7, we restored the expression of NME7 using shRNA-resistant NME7 expression vector (NME7^R) in QGY-7701 and Huh7 cells with endogenous NME7 knockdown (Supplementary Fig. S5A). The expression of NME7^R abolished the inhibition of cell growth (Supplementary Fig. S5B), and colony formation (Supplementary Fig. S5C and S5D) in liquid culture and anchorage-independent growth in soft agar (Supplementary Figs. S5E and S5F) caused by knockdown of NME7, suggesting the specificity of the shRNA.

To explore the molecular mechanism through which NME7 activates Wnt/β-catenin signaling, we first examined whether HCC cell lines could respond to Wnt3a. Treatment with Wnt3a induced the expression of Axin2, the target gene of Wnt/β-catenin signaling, in QGY-7701, Huh7, PVTT, MHCC97H, and LO2 cells (Supplementary Fig. S6A), which was consistent with previous reports (30–32). These observations indicated that the Wnt/β-catenin pathway was intact in these cell lines. However, the expression of Axin2 in HepG2 cells did not change in response to Wnt3a treatment, which was consistent with the previous finding that HepG2 cells harbor mutant β-catenin with a deleted N-terminus (Supplementary Fig. S6A; ref. 32).

In HEK293 cells, overexpression of NME7 coordinated with Wnt3a to activate Topflash reporter activity (Fig. 3A) and induce the expression of Axin2 (Fig. 3B). In addition, knockdown of NME7 inhibited the Wnt3a-induced accumulation of β-catenin (Fig. 3C; Supplementary Figs. S6B and S6C) and the nuclear localization of β-catenin (Fig. 3D; Supplementary Fig. S6D). Moreover, knockdown of NME7 upregulated the phosphorylation of the N-terminal serine and threonine residues of β-catenin (Fig. 3E) and promoted the ubiquitination of β-catenin (Fig. 3F; Supplementary Fig. S6E), suggesting that NME7 regulates the degradation of β-catenin. Furthermore, the inhibitory effects of NME7 knockdown on the anchorage-independent growth of QGY-7701 and Huh7 cells were reversed by overexpression of Flag-β-catenin (Fig. 3G and H), suggesting that knockdown of NME7 suppresses the malignant phenotype of HCC cells by inhibiting Wnt/β-catenin signaling. In the 24 pairs of HCC tissues and adjacent noncancerous tissues where the expression of NME7 was evaluated by Western blotting (Fig. 1G), the level of the β-catenin protein was significantly positively correlated with that of NME7 (Fig. 3I). Both cytoplasmic and nuclear β-catenin are recognized as active forms (6). The NME7 and β-catenin protein levels in the HCC tissue array were evaluated and scored by IHC analysis, which showed that the NME7 protein level was positively correlated with the β-catenin protein level in the cytoplasm and/or nucleus (Fig. 3J and K). In conclusion, these results show that knockdown of NME7 inhibits Wnt/β-catenin signaling.

To explore how NME7 regulates β-catenin protein stability, we first investigated the interaction between NME7 and key components of Wnt/β-catenin pathway. NME7 did not interact with Axin, CK1α, DVL2, LRP6, β-catenin, or RNF43 (Supplementary Figs. S7A–S7F). However, NME7 interacted with GSK3β (Fig. 4A). In particular, Wnt3a induced the interaction between them (Fig. 4B). Consistently, knockdown of NME7 downregulated the phosphorylation of GSK3β (p-GSK3β-S9; Fig. 4C; Supplementary Fig. S7G). However, neither Wnt3a treatment nor knockdown of NME7 affected the phosphorylation of GSK3α (p-GSK3α-S21) or AKT (pAKT-S473; Supplementary Fig. S7G). These observations indicated that NME7 phosphorylated GSK3β. In vitro kinase assays showed that NME7 phosphorylated serine 9 (Ser9, S9) of wild-type GSK3β, but serine 9 of GSK3β could not be phosphorylated by NME7 after mutation (S9A). In addition, mutation of the NME7 active site (mutation of histidine 206 to alanine; H206A) abolished the phosphorylation of GSK3β at serine 9 (Fig. 4D), indicating that GSK3β is a substrate of NME7 and that the metabolic kinase NME7 can exert nonmetabolic function. Consistently, overexpression of NME7 relieved the inhibition of β-catenin protein levels induced by GSK3β but not by GSK3β^S9A (Fig. 4E). Moreover, the NDPK7B domain of NME7 and the kinase domain of GSK3β were found to mediate the interaction between these two proteins (Fig. 4F and G). In the functional study, mutation of NME7 (H206A) abolished the promoting effect of NME7 on the colony formation of HCC cells (Fig. 4H), and overexpression of NME7 relieved the inhibition of colony formation induced by GSK3β but not by GSK3β^S9A (Fig. 4I). In addition, the protein levels of NME7, phosphorylated GSK3β (p-GSK3β-S9), and β-catenin (in the cytoplasm and/or nucleus) were positively correlated in the HCC tissue array analysis (Fig. 4J and K). Therefore, NME7 functions as a protein kinase, phosphorylating serine 9 of GSK3β and thereby activating Wnt/β-catenin signaling.

To identify the molecular events that mediate the biological functions downstream of NME7/Wnt/β-catenin signaling, we performed RNA sequencing (RNA-seq). After knockdown of NME7 in PVTT cells, the expression levels of many metabolic enzymes were altered (Supplementary Table S6), which was verified by qPCR (Fig. 5A). Among the altered metabolic enzymes, MTHFD2 was downregulated after knockdown of NME7 (Fig. 5A).

UALCAN database analysis showed the upregulation of MTHFD2 in HCC tissues (Supplementary Fig. S8A). In addition, the mRNA level of MTHFD2 in HCC tissues was higher than that in adjacent tissues (Supplementary Fig. S8B). Moreover, the mRNA level of MTHFD2 in 81.1% of the HCC tissues was higher than that in the matched adjacent tissues (Supplementary Fig. S8C), and was positively correlated with tumor size (Supplementary Table S7). The MTHFD2 protein level was elevated in the HCC tissue array. In 66% of the HCC tissues, the MTHFD2 protein level was higher than that in the matched adjacent tissues (Supplementary Figs. S8D and S8E), and was significantly correlated with tumor size and micrometastasis (Supplementary Table S8), but was not correlated with the levels of serum markers of HCC (AFP, ALP, ALT, AST, CA199, CEA, GGT, and AFU; Supplementary Fig. S8F). The tissue array analysis also showed that the expression of MTHFD2 predicted poor prognosis (Supplementary Fig. S8G). Similarly, analysis via the Human Protein Atlas database indicated that the expression of MTHFD2 was significantly negatively correlated with the overall survival of patients (Supplementary Fig. S8H).

MTHFD2 is a key metabolic enzyme in one-carbon metabolism (33), and the downstream metabolites hypoxanthine (IMP) and NADH are important for the synthesis of nucleotides (34, 35). In HCC tissues, the mRNA level of NME7 was significantly positively correlated with the mRNA level of MTHFD2 (Fig. 5B). Knockdown of NME7 in PVTT, Huh7, and QGY-7701 cells decreased the mRNA and protein levels of MTHFD2 (Fig. 5C and D), the IMP content and the NADH/NAD⁺ ratio (Fig. 5E), suggesting that knockdown of NME7 inhibits the synthesis of nucleotides. In MHCC97H cells, knockdown of MTHFD2 abolished the effect of NME7 overexpression on the colony formation ability of HCC cells (Supplementary Figs. S9A–S9C). In addition, the promoting effect of NME7 overexpression on the anchorage-independent growth of MHCC97H cells was eliminated by knocking down the expression of MTHFD2 (Fig. 5F and G) or adding an MTHFD2 inhibitor (Fig. 5H and I). This finding indicates that MTHFD2 mediates the biological function of NME7. Expression of MTHFD2^R abolished the inhibition of cell growth caused by knockdown of MTHFD2, suggesting the specificity of the shRNA targeting MTHFD2 (Supplementary Figs. S9D and S9E). On the other hand, in PVTT, Huh7 and QGY-7701 cells with NME7 knockdown, overexpression of MTHFD2 (Supplementary Fig. S9F) restored cell growth (Supplementary Fig. S9G), colony formation ability (Supplementary Figs. S9H and S9I), and anchorage-independent growth (Fig. 5J and K), as well as IMP content and the NADH/NAD+ ratio (Fig. 5L; Supplementary Fig. S9J). Similarly, the addition of sodium formate, a downstream product of MTHFD2, effectively restored the inhibitory effects of NME7 knockdown on the anchorage-independent growth of Huh7 and QGY-7701 cells (Fig. 5M and N).

Next, we investigated whether the expression of MTHFD2 is regulated by the Wnt/β-catenin signaling pathway. Wnt3a stimulation significantly increased the MTHFD2 protein level (Fig. 6A). Promoter analysis revealed two TCF binding element (TBE) sites (TBE1 and TBE2) located at nucleotides −1362bp and −1240bp in the promoter of MTHFD2 (Fig. 6B). In HEK293T cells and HCC cells, the promoter of MTHFD2 could be activated by LiCl treatment or β-catenin overexpression (Fig. 6C). In the promoter mutagenesis analysis, the promoter of MTHFD2 did not respond to LiCl when TBE1 was deleted (Fig. 6D), suggesting that TBE1 is the β-catenin binding site. The ChIP results also showed that β-catenin formed a complex with TCF4 at TBE1 of the MTHFD2 promoter (Fig. 6E and F). In addition, in the HCC tissue array, the MTHFD2 protein levels were significantly positively correlated with the β-catenin protein levels in the cytoplasm and/or nucleus. (Fig. 6G).

To clarify whether NME7 regulates the expression of MTHFD2 through the Wnt/β-catenin signaling pathway, we used Wnt3a and LiCl to stimulate PVTT cells with NME7 knockdown. After knockdown of NME7 expression, the expression of MTHFD2 in PVTT cells no longer responded to the treatment of Wnt3a or LiCl (Fig. 6H and I). Similarly, knockdown of NME7 via an inducible system (through dox addition) abolished the effect of Wnt3a on the induced MTHFD2 protein expression (Fig. 6J). Consistent with these observations, knockdown of NME7 decreased the binding of β-catenin to the promoter of MTHFD2 in the ChIP assay (Fig. 6K and L).

To study the roles of NME7 in HCC tumorigenesis, we used the sleeping beauty transposase system and introduced NME7 and c-Myc expression plasmids into mice by hydrodynamic injection via the caudal vein. Twelve weeks later, the numbers of HCC tumors in mice injected with NME7 and c-Myc were higher than those in control mice (Fig. 7A and B). Hematoxylin and eosin staining showed that the tumor area was increased in the liver with NME7 overexpression (Fig. 7C and D); moreover, the weight of the liver was increased with NME7 overexpression (Fig. 7E). Furthermore, the expression of Ki67, MTHFD2, and β-catenin was significantly upregulated in tumor tissues with NME7 overexpression (Fig. 7C and F). In addition, the organoids formed from tumor tissues with NME7 overexpression had larger diameters and grew faster than the organoids formed from control tumor tissues (Fig. 7G,–I). Moreover, the MTHFD2 inhibitor LY345899 inhibited the growth of organoids (Fig. 7J), and organoids overexpressing NME7 were more sensitive to the MTHFD2 inhibitor (Fig. 7K). These findings reveal that NME7 promotes hepatocarcinogenesis in vivo.

To explore the significance of NME7, β-catenin and MTHFD2 expression in the prognosis of HCC, we first evaluated their expression in the mouse models of HCC. Compared with that in control mice, the expression of NME7, β-catenin and MTHFD2 in mice with Myc-driven HCC (Alb-Cre; Lsl-Myc) was upregulated (Fig. 8A and B). A similar result was observed in the DEN-induced liver cancer model (Fig. 8C). This finding indicates that the expression of NME7, β-catenin and MTHFD2 is upregulated during the initiation and progression of liver cancer.

We then evaluated the expression of NME7, β-catenin and MTHFD2 in 12 pairs of HCC tissues and matched adjacent tissues by Western blot analysis (Supplementary Fig. S9 S9K), and found that the protein levels of NME7, β-catenin and MTHFD2 were positively correlated (Fig. 8D), which were also observed in the IHC analysis (Fig. 8E). Furthermore, we analyzed the correlations between the expression levels of these three proteins and patient survival. The overall survival and disease-free survival times of the patients with high expression of both NME7 and MTHFD2 were significantly shorter than those of patients with low expression of both NME7 and MTHFD2 (Fig. 8F). Similarly, the overall survival and disease-free survival times of patients with high expression of both NME7 and β-catenin were significantly shorter than those of patients with low expression of both NME7 and β-catenin (Fig. 8G), and the overall survival and disease-free survival times of patients with high expression of both MTHFD2 and β-catenin were significantly shorter than those of patients with low expression of both MTHFD2 and β-catenin (Fig. 8H). These findings support the notion that the NME7/Wnt/β-catenin/MTHFD2 signaling cascade may play critical roles in the progression of liver cancer.

Recently, HCC has been divided into three subtypes based on integration of transcriptome and metabolome data in some studies. The survival characteristics of these three subtypes differ owing to their considerable differences in metabolic features, such as kynurenic acid metabolism, Wnt/β-catenin-related lipid metabolism and PI3K/mTOR signaling pathway activity (20). Here, we reveal the cooperation between driving signals and cancer metabolism during the progression of HCC. In the present study, NME7 activated the Wnt/β-catenin signaling pathway to upregulate the expression of MTHFD2, an enzyme critical for one-carbon metabolism, thereby promoting the progression of liver cancer. Inhibitors of MTHFD2 significantly suppressed the growth and colony formation of HCC cells, suggesting that MTHFD2 may be a potential therapeutic target for liver cancer (Fig. 8I).

One of the most interesting findings of this study is that the Wnt/β-catenin signaling pathway regulates the expression of MTHFD2. Wnt/β-catenin signaling is aberrantly activated in approximately 70% of HCC cases (36–38), and our study provided a reasonable explanation for the high expression level of MTHFD2 in HCC. In this study, inhibitors of MTHFD2 significantly suppressed the malignant phenotype of tumor cells induced by the activation of NME7/Wnt/β-catenin signaling. Therefore, targeting MTHFD2 is likely to be a strategy for treating HCCs with constitutive activation of Wnt/β-catenin signaling.

Another interesting finding from this study is that NME7 phosphorylates GSK3β. Wnt3a stimulation both inhibits GSK3β kinase activity and leads to an increase in the phosphorylation level of GSK3β at serine 9 (39). Although this phenomenon is generally well accepted, the mechanism by which Wnt3a stimulation leads to an increase in the phosphorylation level of GSK3β at serine 9 remains to be clarified. Our research findings provide a rational explanation for the inhibitory effect of Wnt3a stimulation on GSK3β kinase activity.

NME7 is a member of the NDPK family. Members of this family catalyze the transfer of a phosphate group from a triphosphate nucleoside to a diphosphate nucleotide (40, 41). Currently, little is known about the functions of NME7, and whether it has metabolic kinase activity or even protein kinase activity is unclear. Deregulation of NME7 and other purine metabolism genes in Parkinson disease has been reported (42). Although the application of inhibitors showed that NME7 facilitates microtubule nucleation in a kinase-dependent manner (43), its substrate and whether this process depends on its metabolic kinase activity or protein kinase activity are unknown. Under pathological conditions (including cancers), some metabolic enzymes have been reported to perform nonclassical or even non-metabolism-related alternative functions. These metabolic enzymes mediate DNA repair (44), cell proliferation and tumor microenvironment remodeling through these alternative functions (45), thereby promoting disease progression. In this study, we identified GSK3β as the direct substrate of the NME7 kinase in vitro and in vivo, thus demonstrating the protein kinase activity of NME7.

Several studies have demonstrated that phosphorylation on serine-9 of GSK3β affected the stability of β-catenin (46–48), which is further confirmed by our observations in this study (Fig. 4). However, Alessi's group reported that mice in which serine-9 in GSK3β and serine-21 in GSK3α were mutated to Alanines had a normal Wnt signaling and no apparent Wnt-associated phenotypes (49). The discrepancy between our data and Alessi's work might be attributed to the different biological context. Alessi's study was performed using mouse fibroblast or mouse embryonic stem cell under the normal physiological conditions, whereas our study was performed under the HCC pathological conditions. Thus, it is possible that the stability of β-catenin may be regulated in different ways in a context-dependent manner.

Although Akt-mediated phosphorylation of Serine-9 has been widely studied (50), our study showed that activation of AKT did not response to the treatment of Wnt3a in HCC cells (Supplementary Fig. S7). However, the phosphorylation of GSK3β (S9) was upregulated upon the treatment of Wnt3a and was downregulated after knockdown of NME7. These observations suggested that AKT did not involve in the regulation of Wnt/β-catenin signaling by NME7 in the HCC context. In addition, our RNA-seq data showed that knockdown NME7 modulated inflammation, IFNγ signaling, IL2–Stat5 signaling, and IL6–Stat3 signaling. Therefore, it is probably that NME7–GSK3β signaling might phosphorylate other substrates besides β-catenin.

The context of tumorigenesis and tumor progression is significantly different from that in the biological process of embryonic development. Therefore, we speculated that in addition to GSK3β and CK1α, there might be new important kinases that regulate Wnt/β-catenin signaling during tumorigenesis. The kinase library screening provided many other potential regulators of Wnt/β-catenin signaling (25).

In summary, our study reveals both the function of NME7 as a protein kinase and the important regulatory effect of Wnt/β-catenin signaling on one-carbon metabolism. NME7 activates Wnt/β-catenin signaling by phosphorylating GSK3β, thus promoting the expression of MTHFD2, which in turn activates one-carbon metabolism promoting the rapid growth of HCC cells. Therefore, NME7 and MTHFD2 are potential therapeutic targets for HCC.

No disclosures were reported.

X. Ren: Data curation, formal analysis, validation, investigation. Z. Rong: Investigation. X. Liu: Investigation. J. Gao: Investigation. X. Xu: Investigation. Y. Zi: Investigation. Y. Mu: Investigation. Y. Guan: Investigation. Z. Cao: Investigation. Y. Zhang: Investigation. Z. Zeng: Investigation. Q. Fan: Investigation. X. Wang: Investigation. Q. Pei: Investigation. X. Wang: Investigation. H. Xin: Investigation. Z. Li: Investigation. Y. Nie: Investigation. Z. Qiu: Investigation. N. Li: Conceptualization, funding acquisition. L. Sun: Conceptualization, funding acquisition, writing–original draft, writing–review and editing. Y. Deng: Conceptualization, funding acquisition, writing–original draft, project administration, writing–review and editing.

This work was supported by National Natural Science Foundation of China 82172980, 81874200, 82030087, 82060308, 82060308, and 81772955; fund 2018RS3028 and 2021JJ30039 from Hunan Provincial Science and Technology Department, China postdoctoral Science Foundation for COVID-19 Prevention and Control (2020M670103ZX), the non-profit Central Research Institute Fund of Chinese Academy of Medical Sciences (2019PT320003), research program from Department of Science and Technology of Guizhou (2018-5801 and 2017-5724). The authors thank the talent base program from Guizhou Talent Office.

The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.