MTHFD2 promotes tumorigenesis and metastasis in lung adenocarcinoma by regulating AKT/GSK‐3β/β‐catenin signalling

Abstract Recent studies have demonstrated that one‐carbon metabolism plays a significant role in cancer development. Methylenetetrahydrofolate dehydrogenase 2 (MTHFD2), a mitochondrial enzyme of one‐carbon metabolism, has been reported to be dysregulated in many cancers. However, the specific role and mechanism of MTHFD2 in lung adenocarcinoma (LUAD) still remains unclear. In this study, we evaluated the clinicopathological and prognostic values of MTHFD2 in LUAD patients. We conducted a series of functional experiments in vivo and in vitro to explore novel mechanism of MTHFD2 in LUAD. The results showed that MTHFD2 was significantly up‐regulated in LUAD tissues and predicted poor prognosis of LUAD patients. Knockdown of MTHFD2 dramatically inhibited cell proliferation and migration by blocking the cell cycle and inducing the epithelial‐mesenchymal transition (EMT). In addition, MTHFD2 knockdown suppressed LUAD growth and metastasis in cell‐derived xenografts. Mechanically, we found that MTHFD2 promoted LUAD cell growth and metastasis via AKT/GSK‐3β/β‐catenin signalling. Finally, we identified miR‐30a‐3p as a novel regulator of MTHFD2 in LUAD. Collectively, MTHFD2 plays an oncogenic role in LUAD progression and is a promising target for LUAD diagnosis and therapy.

prognosis is mainly ascribed to cancer recurrence and metastasis.
Consequently, it is urgent to identify more effective targets for diagnosis and therapy in LUAD.
One-carbon metabolism, as a part of metabolic reprogramming, has attracted researchers' attention to cancer in recent years. 5,6 In addition to its primary role in nucleotide synthesis, one-carbon metabolism also plays a crucial role in the methionine cycle, methylation reactions, ATP production and redox defence. [6][7][8] The reactions in one-carbon metabolism comprise two parallel pathways in cytosolic and mitochondrial compartments. Although most of the reactions are reversible, those enzymes catalyse to the direction of thymidylate and purine synthesis during cell proliferation especially in cancer cells. 7,8 Moreover, the mitochondrial enzymes are reported to be up-regulated in numerous cancers. 7,[9][10][11] Methylenetetrahydrofolate dehydrogenase 2 (MTHFD2), which is known as a mitochondrial one-carbon metabolism enzyme, 7 is reported to be up-regulated in various tumours. 10,[12][13][14] Overexpression of MTHFD2 predicts poor prognosis and promotes colorectal cancer cell proliferation and metastasis. 14 In breast cancer, MTHFD2 is identified to regulate the migration and invasion abilities of tumour cells. 10 MTHFD2 overexpression not only facilitates cancer cell migration and invasion, but also predicts poor clinical outcomes in hepatocellular carcinoma and renal cell carcinoma. 12,13 In addition to its canonical role to facilitate one-carbon metabolism for purine synthesis, recent studies characterize MTHFD2 with redox defence, epigenetic modification, RNA translation and DNA repair. [14][15][16][17] Knockdown of MTHFD2 in colorectal cancer hinders NADPH production and makes cancer cells more sensitive to oxidative stress. 14 In renal cell carcinoma, MTHFD2 promotes the mRNA methylation of HIF-2α via metabolic reprogramming. 16 Moreover, MTHFD2 is found to be involved in RNA metabolism and translation by interacting with nuclear proteins in cancer cells. 17 Besides its role in mitochondria, MTHFD2 located in the nucleus stabilizes the phosphorylation of EXO1 to promote homologous recombination repair, which ensures the pluripotency of mouse stem cells. 15 Since MTHFD2 is present in embryonic development but barely expressed in most adult tissues, 18 it is a promising therapeutic target. Although MTHFD2 has been reported to promote NSCLC proliferation, 19 the underlying mechanism of MTHFD2 in LUAD is largely unknown.
miRNAs are a kind of highly conserved small non-coding RNAs, which regulate gene expression by affecting the translation and stability of mRNAs, which modulate cell differentiation, survival and tumorigenesis. 20 In many diseases, miRNAs show powerful regulatory ability to modulate one-carbon metabolism. 21-23 miR-370 and miR-373 are reported to alter one-carbon metabolism via SHMT2 and MECP2 to alleviate osteoarthritis. 21 In gastric cancer, miR-6778-5p strengthens stem cell signature by regulating cytosolic enzyme SHMT1. 22 In breast cancer, MTHFD2 is identified as a target of miR-9 that inhibits cell proliferation. 23 Nevertheless, the relationship between miRNAs and MTHFD2 in LUAD remains to be established.
In the current study, we explored the expression, clinical characteristics and prognostic value of MTHFD2 in LUAD patients. Then, we elucidated the effect of MTHFD2 on cell proliferation, apoptosis and migration of LUAD cells in vitro and in vivo. Mechanically, we revealed that MTHFD2 was involved in AKT/GSK-3β/β-catenin signalling. Furthermore, we found that miR-30a-3p negatively regulated MTHFD2 in LUAD. Therefore, this study exhibited the clinical values of MTHFD2 in LUAD patients and determined the biological effects, potential molecular mechanism and regulation of MTHFD2 in LUAD cells.

| Patient tissue collection
This study was permitted by the ethical committee of Sir Run Run Shaw Hospital. A total of 21 paired primary cancer tissues and adjacent normal lung tissues were collected from LUAD patients in Sir Run Run Shaw Hospital from January 2019 to January 2020. Written informed consents were obtained from each patient, who had not received any radiotherapy or chemotherapy before surgery. The normal lung tissues were cut at least 5 cm away from cancer tissues.
All samples were histologically confirmed by pathologists. The fresh tissues were instantly frozen in liquid nitrogen and then stored at −80℃, or fixed and paraffin-embedded.
Human bronchial epithelial cell (HBEC) was a gift from School of Medicine, Zhejiang University. The LUAD cell lines, NSCLC cell lines and HBEC were maintained in RPMI 1640 medium (Solarbio).

| Real-time quantitative PCR (RT-qPCR)
RNA was extracted using RNA extract reagent according to the

| Western blot
Total protein was extracted using RIPA lysis buffer (Beyotime) containing phosphatase and protease inhibitors (Solarbio). Nuclear protein was extracted using Nuclear Protein Extraction Kit (Solarbio, R0050). Protein concentration was detected using BCA protein assay kit (Beyotime). Equal amount of protein (10 μg) was loaded onto 10% SDS-PAGE and then transferred to PVDF membranes (Bio-Rad, USA). All membranes were then blocked for 1 hour with 5% BSA and incubated at 4℃ overnight with primary antibodies.
Absorbance (OD) at 450 nm was used to access the cell viability.

| Cell cycle and cell apoptosis
For cell cycle analysis, 75% cold ethanol was used to fix cells and cells were kept at 4℃ over night. Next day, propidium iodide (PI)/ RNase Staining Buffer (550825, BD Biosciences) was applied to stain the cells. Then, FACS Calibur (BD Biosciences) was used for analysis.
Cell cycle distributions were quantified by ModFit LT software.
For cell apoptosis, Annexin V-FITC/PI binding buffer (C1062L, Beyotime) was applied to stain the cells. Then, FACS Calibur (BD Biosciences) was used for analysis. The percentages of apoptotic cells were quantified by FlowJo software.

| Wound healing assay and transwell assay
For wound healing assay, 24-well plates were applied and cells were incubated until 90% confluence. Scratch wounds were made with 1-mL sterile pipette tips across cell surface. Images of wound closure were collected at 0 hour and 24 hours using a microscope.
For transwell migration assay, 200 μL FBS-free medium with 1 × 10 5 cells was added into upper chambers of Transwell chambers (3422, Corning). Lower chambers were filled with 600 μL 10% FBS medium. After incubating for 24-48 hours, methanol and 0.25% crystal violet (Solarbio) were applied to fix and stain the cells. Microscope was used to count the stained cells, and the mean numbers of cells in three random fields were calculated as the results.

| Tumorigenesis and metastasis in vivo
The Institutional Animal Care and Use Committee of Zhejiang University approved all experimental procedures. Male BALB/cnude mice (4 weeks old) were housed in specific pathogen-free environment and randomly divided into 5 mice per group.
Tumour sizes were measured using vernier calliper every three days.
The tumour volumes were calculated on the basis of the formula: volume = length × width 2 /2. After injection, Mice were killed for weighing the tumour and IHC staining after 3 weeks.
For tail vein metastatic tumour model, stably transfected cells (2 × 10 6 /mouse) were resuspended in PBS and injected into the nude mice through tail veins. Six weeks after injection, mice were killed for removing the lungs and HE staining. The metastatic tumours were counted using a microscope.

| Statistical analysis
All experiments had three individual times repetition. All statistical analyses were conducted using GraphPad Prism 7 software and SPSS 20.0. The difference between two groups was compared using Student's t test. One-way ANOVA or two-way ANOVA was used to compare multiple groups. The results were shown as mean ± standard deviation (SD). Overall survival was compared using log-rank test and presented as Kaplan-Meier survival curve.
GSEA analysis was performed based on MTHFD2 expression of TCGA LUAD patients. The correlation between two groups was analysed using Pearson's correlation test. P < .05 was set as statistical significance.

| MTHFD2 is up-regulated in LUAD patients and predicts poor prognosis
Since one-carbon metabolism enzymes are reported to be dysreg-

| MTHFD2 knockdown inhibits LUAD cell proliferation and induces LUAD cell apoptosis
Given the MTHFD2 overexpression found in LUAD tissues, we tested its expression in LUAD cell lines. As shown by Western blot, MTHFD2 protein level was basically higher in LUAD and two other NSCLC cell lines than in HBEC (Figure 2A). To clarify the role of MTHFD2 in LUAD, MTHFD2 was knocked down using small interfering RNAs (siRNAs) in PC-9 and H1975 cells and overexpressed using plasmid in SPC-A-1 cells. Then, Western blot con-

| MTHFD2 knockdown suppresses LUAD cell migration
We next assessed the impact of MTHFD2 on the migration capac-

| MTHFD2 knockdown suppresses the tumorigenesis and metastasis of LUAD cells in vivo
To further confirm the impact of MTHFD2 on LUAD tumorigenesis in vivo, PC-9 cells with stable MTHFD2 knockdown (shMTHFD2) were injected into nude mice subcutaneously. In accordance with the results in vitro, the shMTHFD2 PC-9 group grew much slower than the shNC PC-9 group ( Figure 4A). The shMTHFD2 PC-9 group showed decreased tumour volume and weight compared with the shNC PC-9 group ( Figure 4B,C). In addition, the shMTHFD2 PC-9 group exhibited a notably reduced cell proliferation index and increased cell apoptosis, as determined by Ki-67 and TUNEL staining ( Figure 4D-G).
The tail vein metastatic tumour assay showed that mice injected with shMTHFD2 PC-9 cells had significantly fewer metastatic nodules in the lung than those injected with shNC PC-9 cells ( Figure 4H).

| MTHFD2 knockdown inhibits the AKT/ GSK-3β/β-catenin signalling to suppress LUAD proliferation and metastasis
To further elucidate the mechanism of MTHFD2 in LUAD development, we performed gene set enrichment analysis (GSEA) of MTHFD2 on TCGA LUAD samples. We noticed that MTHFD2 could affect potential protein serine/threonine kinase activity ( Figure 5A). Western blot analysis showed that MTHFD2 knockdown had no effect on ERK1/2 signalling, but significantly de-

| MTHFD2 overexpression is mediated by down-regulation of miR-30a-3p in LUAD cells
We further explored why MTHFD2 is up-regulated in LUAD.
Since miRNAs play important roles in tumorigenesis mainly through regulating genes including one-carbon metabolism enzymes via post-transcriptional processes, we next explored potential miRNAs targeting MTHFD2. Twenty common candidate miRNAs were screened from the miRWalk, miRDB, miRanda and TargetScan databases ( Figure 7A, Table S1). MiR-30a-3p F I G U R E 2 Knockdown of MTHFD2 inhibits proliferation and promotes apoptosis in LUAD cells. A, Western blot analysis for MTHFD2 expression in NSCLC cell lines and human bronchial epithelial cells (HBECs). B and C, Western blot confirmed the knockdown efficiency of MTHFD2 in PC-9 and H1975 cells with siRNA (#1, #2). siNC, negative control siRNA. D, CCK-8 assay assessed the proliferation of PC-9 and H1975 cells after transfection. E and F, The effect of MTHFD2 knockdown on cell cycle in PC-9 and H1975 cells as indicated treatment for 48 h. G and H, Cell apoptosis (Q2 + Q4) of PC-9 and H1975 were measured using Annexin Ⅴ-FITC/PI staining after transfection with siNC or siMTHFD2 (si#1, si#2) for 48 h. *P < .05, **P < .01, ***P < .001 F I G U R E 3 Knockdown of MTHFD2 inhibits migration ability of LUAD cells. A and B, Wound healing assay showed the migration ability of H1299 and H1975 cells after transfection. Scale bar, 500 μm. C and D, Representative images of transwell assay in H1299 and H1975 cells after transfection. Scale bar, 200 μm. E and F, Western blot analysis for N-cadherin and vimentin expression of H1299 cells, E-cadherin and vimentin expression of H1975 cells as indicated treatment for 48 h. *P < .05, **P < .01, ***P < .001 was identified from them because it was reported to be downregulated in many cancers, including NSCLC. [31][32][33] The results derived from TCGA LUAD samples using starBase 34 showed that LUAD tissues had much lower miR-30a-3p expression level compared with normal lung tissues ( Figure 7B), and miR-30a-3p exhibited a negative relationship to MTHFD2 ( Figure 7C). In line with the results from the database, we confirmed LUAD cells also had lower miR-30a-3p expression levels than in HBECs ( Figure 7D). In addition, transfection with miR-30a-3p mimics significantly inhibited MTHFD2 at both mRNA and protein levels in LUAD cells ( Figure 7E-G). Dual luciferase reporter assay validated that miR-30a-3p interacted with MTHFD2 at sites 91-97 of the 3′UTR (Figure 7H,I). Moreover, miR-30a-3p showed an inhibitory effect on PC-9 and H1975 cell growth, which could be partially reversed by MTHFD2 ( Figure 7J). Collectively, the down-regulation of miR-30a-3p is found to be responsible for MTHFD2 overexpression in LUAD.

| D ISCUSS I ON
Metabolic reprogramming is a fundamental property of tumorigenesis. 35 MTHFD2, a mitochondrial one-carbon metabolism enzyme, is dysregulated in a variety of cancers. 10,12,14 This study displayed the gene expression levels of one-carbon metabolism enzymes and found that MTHFD2 was the most up-regulated enzyme in LUAD.
It has been established that NSCLC patients who smoke have a higher chance of Kras mutation than those who do not smoke, 36 and Kras mutation is known to be associated with MTHFD2 expression via MYC transcriptional regulation. 14,37,38 Both of these layers of evidence substantiate our result that male sex and smoking habit F I G U R E 5 MTHFD2 knockdown inhibits the AKT/GSK-3β/β-catenin signalling pathway. A, GSEA revealed high MTHFD2 expression was associated with protein serine threonine kinase activity gene set based on TCGA LUAD samples. NES, normalized enrichment score. B and C, Western blot analysis for p-AKT (S473), AKT, p-ERK1/2, ERK1/2 and nuclear β-catenin expression of PC-9 and H1975 cells after transfection. D and E, Western blot analysis for p-AKT (S473), AKT, p-GSK-3β (S9), GSK-3β, β-catenin and MTHFD2 expression of PC-9 and H1975 cells after transfection with siNC or si#2 (48 h), vector or MTHFD2 plasmid (24 h) in the presence of DMSO or MK-2206 (10 μM, 24 h) as indicated. *P < .05, **P < .01, ***P < .001 F I G U R E 7 MTHFD2 overexpression is mediated by down-regulation of miR-30a-3p in LUAD cells. A, The common miRNAs targeting MTHFD2 were predicted by the miRWalk, miRDB, miRanda and TargetScan Databases. B, The relative expression levels of miR-30a-3p in LUAD samples compared with normal lung tissues from TCGA database using starBase. C, Correlation analysis of relationship between MTHFD2 and miR-30a-3p mRNA expression levels in LUAD samples from TCGA database using starBase. D, RT-qPCR analysis of miR-30a-3p mRNA expression levels in LUAD cell lines compared with HBECs. E, F and G, RT-qPCR and western blot analysis for MTHFD2 mRNA and protein levels in PC-9 and H1975 cells after transfection with NC mimics or miR-30a-3p mimics. H, Schematic of miR-30a-3p putative binding site in the wild-type (WT) and mutant (MUT) 3′UTR of MTHFD2. I, The dual luciferase reporter assay determined the interaction between miR-30a-3p and 3′UTR of MTHFD2. J, CCK-8 assay assessed the viability of PC-9 and H1975 cells after co-transfected with NC mimics or miR-30a-3p mimics, and vector or MTHFD2 plasmid. *P < .05, **P < .01, ***P < .001 knockdown blocked the cell cycle in LUAD cells at G0/G1 phase. In line with our results, MTHFD2 has been shown to co-express with cell cycle-related proteins, 19 as well as to promote tumour growth in hepatocellular carcinoma and renal cell carcinoma. 12,13 Metastasis is another major element responsible for cancer deaths. Accordingly, we investigated the role of MTHFD2 in LUAD metastasis. MTHFD2 suppression significantly inhibited H1299 and H1975 cell migration.
Besides, MTHFD2 knockdown diminished N-cadherin and vimentin expression in LUAD cells since EMT is a fundamental property for cancer metastasis. Since PC-9 cells barely showed N-cadherin and vimentin expression in our experiment conditions (data not shown), we selected H1299 and H1975 cells for migration assays. Consistently, MTHFD2 has also been reported to affect cancer cell migration and invasion abilities by modulating vimentin expression in breast cancer and renal cell carcinoma. 10,13 Furthermore, we confirmed that AKT, a serine/threonine kinase, is involved in diverse biological processes, including cell survival, differentiation and motility. 39,40 AKT can inactivate GSK-3β, a glycogen synthase and serine/threonine kinase, via phosphorylation at Ser9, which contributes to tumorigenesis. 41 Wnt/β-catenin signalling is known to be associated with EMT, metastasis and CSC property maintenance. 42,43 GSK-3β-mediated phosphorylation of β-catenin leads to its ubiquitination and degradation. 44 Activation of AKT/GSK-3β signalling can lead to the accumulation and translocation of β-catenin, which facilitates cancer malignancy. 43,45 We speculated that MTHFD2 affected protein serine/threonine kinase activity given our GSEA results. Intriguingly, we found a positive relationship between MTHFD2 and AKT signalling. Subsequently, we first revealed that AKT/GSK-3β/β-catenin signalling induced by MTHFD2 was involved in LUAD progression. However, AKT was reported to regulate MTHFD2 expression via c-Myc in colorectal cancer. 14 Based on both evidence, there may be a close connection between MTHFD2 and AKT signalling.
Accumulating evidence defines non-coding RNAs' pivotal role in carcinogenesis. miRNAs, a kind of small endogenous single-stranded RNAs, consist of about 20 nucleotides, which mostly serve as negative regulators by targeting the 3′UTR of target genes at the posttranscriptional level. 46,47 Recent studies show miRNAs can affect cancer progression via manipulating MTHFD2. Through microarray profiling, MTHFD2 is identified as a target of miR-9 which inhibits cell proliferation in breast cancer. 23 In acute myeloid leukaemia, miR-92a and miR-504-3p serve as tumour suppressors by regulating MTHFD2. 48,49 Moreover, miR-940 and miR-33a-5p are reported to inhibit cancer progression through down-regulating MTHFD2 in glioma and colorectal cancer, respectively. 50,51 Herein, we determined that miR-30a-3p was down-regulated and acted as a tumour suppressor in LUAD. Besides, we first validated that miR-30a-3p was a negative regulator of MTHFD2. Therefore, miR-30a-3p downregulation accounts for the overexpression and effect of MTHFD2 in LUAD.
Accordingly, MTHFD2 will become a promising target for LUAD diagnosis and therapy in the future.

ACK N OWLED G EM ENTS
This study was supported by National Natural Science Foundation of China (31627801, 81970049).

CO N FLI C T O F I NTE R E S T
All authors declare no conflicts of interest.

DATA AVA I L A B I L I T Y S TAT E M E N T
All data used in the current study are available from the corresponding author on reasonable request.