Metformin Is a Pyridoxal-5'-phosphate (PLP)-Competitive Inhibitor of SHMT2

Abstract

The anticancer actions of the biguanide metformin involve the functioning of the serine/glycine one-carbon metabolic network. We report that metformin directly and specifically targets the enzymatic activity of mitochondrial serine hydroxymethyltransferase (SHMT2). In vitro competitive binding assays with human recombinant SHMT1 and SHMT2 isoforms revealed that metformin preferentially inhibits SHMT2 activity by a non-catalytic mechanism. Computational docking coupled with molecular dynamics simulation predicted that metformin could occupy the cofactor pyridoxal-5'-phosphate (PLP) cavity and destabilize the formation of catalytically active SHMT2 oligomers. Differential scanning fluorimetry-based biophysical screening confirmed that metformin diminishes the capacity of PLP to promote the conversion of SHMT2 from an inactive, open state to a highly ordered, catalytically competent closed state. CRISPR/Cas9-based disruption of SHMT2, but not of SHMT1, prevented metformin from inhibiting total SHMT activity in cancer cell lines. Isotope tracing studies in SHMT1 knock-out cells confirmed that metformin decreased the SHMT2-channeled serine-to-formate flux and restricted the formate utilization in thymidylate synthesis upon overexpression of the metformin-unresponsive yeast equivalent of mitochondrial complex I (mCI). While maintaining its capacity to inhibit mitochondrial oxidative phosphorylation, metformin lost its cytotoxic and antiproliferative activity in SHMT2-null cancer cells unable to produce energy-rich NADH or FADH₂ molecules from tricarboxylic acid cycle (TCA) metabolites. As currently available SHMT2 inhibitors have not yet reached the clinic, our current data establishing the structural and mechanistic bases of metformin as a small-molecule, PLP-competitive inhibitor of the SHMT2 activating oligomerization should benefit future discovery of biguanide skeleton-based novel SHMT2 inhibitors in cancer prevention and treatment.

Keywords: folate; glycine; one-carbon metabolism; serine; serine hydroxymethyltransferase.

Figure 1

Metformin inhibits the catalytic activity of human purified SHMT2. (A) Top. The quinonoid intermediate formed upon binding of glycine to SHMTs accumulates only when a folate ligand is simultaneously bound to the enzyme, thereby forming an enzyme–glycine–folate ternary complex. In the scheme, the pyridine ring of the cofactor is indicated by Py. Bottom. The activity of purified (5 μmol/L) recombinant SHMT1 (black symbols) and SHMT2 (red symbols) was measured in the presence of 5 μmol/L formyl-THF, a saturating concentration of glycine (3 mmol/L) and increasing concentrations of metformin (MET) (0–30 mmol/L). Values are the mean ± standard deviation (SD) of three independent experiments. * p < 0.05 (B) Top panels. Absorbance changes at 502 nm were measured upon the addition of graded concentrations of metformin (0–200 mmol/L) to a solution of 5 μmol/L SHMT, glycine, and formyl-THF. In parallel experiments (left panels), assays were performed varying the concentration of formyl-THF (2.5–160 μmol/L) at a fixed glycine concentration (10 mmol/L) or varying the concentration of glycine (0.04–10 mmol/L) at fixed formyl-THF concentration (160 μmol/L). Bottom panels. Double reciprocal plots of absorbance changes at 502 nm (plotted as 1/ΔAbs_{502 nm}). Kinetic mechanisms of inhibition by metformin are shown in which the SHMT2 enzyme (either the SHMT2–glycine or the SHMT2–formyl–THF complex) forms a complex with metformin according to a variety of dissociation rate constants (K_d = K for glycine or formyl-THF binding; K_i = K_d for metformin binding to the SHMT2 binary complex; K_i1 = Kd for metformin binding to the SHMT2 ternary complex).

Figure 2

Metformin is computationally predicted to target the PLP-binding cavity of SHMT.Top panels. Models of incorporation of the predicted PLP binding cavity-targeting metformin behavior in human SHMT1 and SHMT2. Figures show the composition of a closed apo-SHMT (monomer A, pale green) and holo-SHMT (monomer B, light blue) before (0 ns) and after (100 ns) a molecular dynamics simulation in which a docked metformin (MET) molecule (represented as spheres) is included in the PLP binding cavity of both subunits. Insets show the detailed interactions of the best poses of MET docked to the PLP-binding site of SHMTs using the PLIP algorithm, indicating the participating amino acids involved in the interaction and the type of interaction (hydrogen bonds, hydrophilic interactions, salt bridges, Π-stacking, etc.). Insets in the left panel include also the Gibbs free energy variation (ΔG, kcal/mol) values calculated from molecular docking experiments (pH 7.4) using YASARA v19.9.17 software. Bottom panels. Surface representation of closed SHMT dimers before (0 ns) and after (100 ns) a molecular dynamics simulation. Proteins are represented as a function of the hydrophobicity of its surface amino acids and the Na⁺ and Cl⁻ ions have been eliminated to facilitate visualization. Insets show a detailed view of the PLP binding cavity in SHMT1/2 enzymes. A MET molecule (sphere) is shown when visible from outside the cavity. Figures were prepared using PyMol 2.3 software.

Figure 3

Metformin regulates SHMT2 dimer–tetramer transition. (A) Top. Differential scanning fluorimetry (DSF) analysis of apo-SHMT1 and apo-SHMT2 (0.5 μmol/L each) in the absence and presence of PLP (10 μmol/L) and/or MET (200 mmol/L). Bottom. Melting temperatures (T_m) for each of the conditions shown in left panels. Changes in melting temperatures (ΔT_m) were calculated by subtracting the T_m of SHMT1/2 with buffer and without PLP from the T_m of SHMT1/2 with added PLP in the absence and presence of MET. (B) Top. T_m of apo-SHMT1/2 with increasing concentrations of PLP (0–10 μmol/L) in the absence and presence of MET. Bottom. Computational- and DSF-based conceptual model of the ability of metformin (MET) to differentially impact the SHMT2 dimer–tetramer equilibrium in response to PLP while sparing the SHMT1 PLP-independent constitutive tetramer.

Figure 4

Metformin selectively inhibits SHMT2 activity in cancer cells. Suspensions of SHMT1⁺/SHMT2⁺ HAP1 parental cells and SHMT1/SHMT2 KO-HAP1 derivatives were incubated with graded concentrations of metformin (MET, 0–150 mmol/L) for 1.5 h before adding the glycine radioisotope glycine-2-³H. Left. The results describe the dose-dependent effect of MET on the ability of SHMT to catalyze the exchange of glycine 2-pro-S proton with the solvent. Right. The inhibitory effect of MET on SHMT activity occurred in a dose-dependent manner but the IC₃₀, IC₅₀, and IC₈₀ curves were shifted to the right (less potent) in SHMT2 KO-HAP1 cells.

Figure 5

Metformin targets the mitochondrial serine catabolism via mitochondrial complex I and SHMT2. (A,B) Models summarizing mammalian SGOC metabolism in which the oxidation of the third carbon of serine to formate relies on functional mitochondria in a complex I-dependent manner [17,32,52]. Formate produced via mitochondrial SHMT2 is released into the cytosol where it supplies the 1C demand for nucleotide synthesis [53]. Formate can be also recycled back to re-synthesize serine via cytosolic 1C metabolism [47]. Thus, in cells with defective mitochondrial 1C metabolism, the cytosolic pathway is reverted, compensating for the loss of mitochondrial formate production. The metformin-resistant Saccharomyces cerevisiae protein NDI1 is a single-subunit NADH dehydrogenase that oxidizes NADH similar to the multi-subunit mammalian complex I without proton pumping or reactive oxygen species generation [16,54]. Graphs show formate release (A) and exchange rates of formate (B) M + 0 and M + 1 isotopologues in the extracellular milieu of SHMT1⁺/SHMT2⁺ HAP1, SHMT1 KO-HAP1 cells, and isogenic derivatives engineered to overexpress NDI1, all cultured for 24 h in the absence or presence of graded concentrations of metformin (MET). (C) Changes in dTTP, AICAR, and SAM following treatment of SHMT1⁺/SHMT2⁺ HAP1, SHMT1 KO-HAP1 cells, and isogenic derivatives engineered to overexpress NDI1 with 1 and 5 mmol/L MET. Data are presented as means (columns) ± SD (bars) (n = 3 cultures representative of at least two independent experiments). Comparisons of means were performed by ANOVA. p values < 0.05 and < 0.005 (versus untreated controls) were considered to be statistically significant (denoted as * and **, respectively; n.s. not significant).

Figure 6

SHMT2 depletion prevents the cytotoxic and anti-proliferative effects of metformin. (A) Left. Representative immunoblot for SHMT2 or β-actin from cell lysates of HAP1 and SHMT2-KO HAP1 cell lines engineered to overexpress SHMT2 cDNA. Right. Bar graph showing MTT-based IC₅₀ values for metformin in HAP1 and SHMT2-KO HAP1 cells before and after re-expression of SHMT2. Columns are means and error bars are SD. n = 3 biological replicates in technical triplicates. (B) RTCA profiles and normalized cell indexes of HAP1 and SHMT2-KO HAP1 cells before and after re-expression of SHMT2 cultured in the absence or presence of graded concentrations of metformin (MET). Columns are means and error bars are SD. n = 3 biological replicates in technical triplicates. Comparisons of means were performed by ANOVA. p values < 0.05 and < 0.005 were considered to be statistically significant (denoted as * and **, respectively; n.s. not significant). EV: empty-vector; Rec: reconstituted.

Figure 7

Depletion of SHMTs alters the mitochondrial bioenergetic profile in HAP1 cells.Top. Mitochondrial phenotyping of HAP1 cells upon SHMT1 and SHMT2 loss using the Seahorse XFp Cell Mito Stress Test Assay. Bottom. OCR profiling. Seahorse OCR bioenergetic profiles before and after injections of oligomycin, FCCP, and antimycin A in HAP1 parental, SHMT1-KO HAP1, and SHMT2-KO HAP1 counterparts pre-incubated in the absence or presence of 5 mmol/L metformin (MET) for 24 h. **** p <0.0001; *** p < 0.0005; ** p < 0.005; * p < 0.05.

Figure 8

Depletion of SHMTs alters the mitochondrial functioning in HAP1 cellsTop. Mitochondrial phenotyping of HAP1 cells upon SHMT1 and SHMT2 loss using Mitochondrial Function Assays with Biolog Mitoplates. Mitochondrial function in HAP1 parental and SHMT2-KO HAP1 counterparts was assayed by measuring the rates of electron flow into and through the electron transport chain (ETC) for cytoplasmic, TCA cycle, and other metabolic substrates producing NAD(P)H (pyruvate, isocitrate, α-ketoglutarate, malate, fatty acids) or FADH₂ (succinate, glycerol-3-phosphate, fatty acids). The electrons travel from the beginning (complex I or II) to the distal portion of the ETC where a tetrazolium redox dye (MC) acts as a terminal electron acceptor that turns purple upon reduction. Bottom. Figure shows the heatmap of the metabolic substrate consumption of fatty acids, glycolysis, amino acids, and TCA cycle metabolites in HAP1 and SHMT2-KO HAP1 cells. Changes in pathway-specific consumption reveal that SHMT2-KO HAP1 cells consumed notably less TCA substrates than HAP1 parental cells.

Figure 9

Metformin and SHMT2 functioning: a working model. The pharmacological capacity of metformin to operate as a PLP-competitive inhibitor of the oligomeric state of mitochondrial SHMT2 might offer a new perspective for understanding the anti-cancer mechanism of action of metformin. First, mitochondrial SHMT2 is the key channeler of serine-derived carbon into the cytoplasmic and mitochondrial 1C pools of THF-based cofactors that are required not only for nucleotide synthesis but also for mitochondrial respiration, including assembly of the mCI, tRNA formylation-driven translation of mitochondrially encoded proteins, and methylation transfer reactions in cancer cells. Second, the selective advantage provided by the (SHMT2) mitochondrial 1C metabolism versus the (SHMT1) cytosolic pathway is based on the ability of the SHMT2-dependent rate of mitochondrial serine-to-formate catabolism to regulate a master metabolic switch in nucleotide and AMPK-sensed energy metabolism. Third, the inactive apo-SHMT2 dimer is recognized as an unexpected driver of immune signaling capable of promoting inflammatory cytokine signaling via interaction with the deubiquitylating BRCC36 isopeptidase complex (BRISC). The ability of metformin to directly target the enzymatic activity of SHMT2 may improve our understanding of the SGOC metabolism-related anti-tumor actions of metformin, which should not exclusively be viewed as a downstream consequence of mCI targeting, but also as a causal upstream inhibitory effect against SHMT2-driven mitochondrial serine catabolism. Although weak, the PLP-competitive behavior of metformin hindering the formation of active holo-SHMT2 tetramers might synergize with unfavorable catalytic conditions characteristic of the folate trap imposed by vitamin B12 deficiency (e.g., decreased formation of the SHMT2 substrate THF and augmented levels of folate-trapping intermediates that inhibit SHMT2 activity) occurring upon metformin treatment.