Regulation of folate and methionine metabolism by multisite phosphorylation of human methylenetetrahydrofolate reductase

Methylenetetrahydrofolate reductase (MTHFR) catalyzes the irreversible conversion of 5,10-methylene-tetrahydrofolate (THF) to 5-methyl-THF, thereby committing one-carbon units to the methionine cycle. While MTHFR has long been known to be allosterically inhibited by S-adenosylmethionine (SAM), only relatively recently has N-terminal multisite phosphorylation been shown to provide an additional layer of regulation. In vitro, the multiply phosphorylated form of MTHFR is more sensitive to allosteric inhibition by SAM. Here we sought to investigate the kinases responsible for MTHFR multisite phosphorylation and the physiological function of MTHFR phosphorylation in cells. We identified DYRK1A/2 and GSK3A/B among the kinases that phosphorylate MTHFR. In addition, we found that MTHFR phosphorylation is maintained by adequate cellular SAM levels, which are sensed through the C-terminal SAM binding domain of MTHFR. To understand the function of MTHFR phosphorylation in cells, we generated MTHFR CRISPR knockin mutant lines that effectively abolished MTHFR phosphorylation and compared them with the parental cell lines. Whereas the parental cell lines showed increased 5-methyl-THF production in response to homocysteine treatment, the knockin cell lines had high basal levels of 5-methyl-THF and did not respond to homocysteine treatment. Overall, our results suggest that MTHFR multisite phosphorylation coordinates with SAM binding to inhibit MTHFR activity in cells.

. (c) Peptide library result for DYRK1A. Recombinant DYRK1A was distributed across a 384-well plate, mixed with the peptide substrate library, Kinase Assay Buffer I (SignalChem) and ATP[γ32P], and incubated for 90 minutes at 30 °C. Each well contains a mixture of peptides with a centralized phospho-acceptor (serine and threonine at a 1:1 ratio) and one fixed amino acid, in a randomized background. All 20 amino acids are sampled across the range of −5 to +4, relative to the centralized serine/threonine, to determine the individual contributions of amino acids along the substrate peptide, as shown in the heatmap where the x-axis represents fixed amino acid and the y-axis represents relative position. The resulting densitometry values are used to generate substrate logos for the kinase. (d) Sequence logo of the substrate consensus motif of DYRK1A as determined in (c). The central residue "0" on the X axis denotes the phosphoacceptor serine or threonine. "−1" and "1", for example, denote the immediate N-terminal and C-terminal residues respectively relative to the serine or threonine phosphorylation site. Taking into account how kinases are selective of the surrounding amino acid sequence, the letter height in the logos is proportional to the favorability of the corresponding amino acid at each position.

Mutational analysis of the N-terminal serine-rich region of MTHFR.
Our proposed scheme has three predictions. First, T34 phosphorylation by DYRKs primes all the subsequent phosphorylation events. Indeed, the T34A mutant migrated faster than the wild-type on both regular Bis-Tris and phos-tag gels, and was resistant to calf intestine alkaline phosphatase treatment (Fig. 2a-c), consistent with previous studies 7,11 . Second, R31 is a strong determinant for the priming phosphorylation at T34 by DYRKs. As expected, the R31A mutant was phosphorylated to a much lesser extent than wild-type MTHFR (Fig. 2a,b). Third, the S30A mutation abolishes GSK3 phosphorylation, resulting in a singly phosphorylated species that can be recognized by the S(pT) P motif antibody (Fig. 1b), which importantly does not react with (pS)(pT)P. This was observed (Fig. 2a-c), although it was apparent on phos-tag gel that only ~1/3 of the S30A MTHFR protein was phosphorylated at T34, with the remainder being unphosphorylated (Fig. 2c).
Evidence for DYRK1A and DYRK2 as priming kinases for MTHFR T34 phosphorylation. To test if DYRKs are the priming kinases, we treated both wild-type and S30A-MTHFR overexpressing MDA-MB-468 and The same as (A), but the ectopic V5-tag was located at the C-terminus instead. (c) Top, the wild-type (WT), S30A, and T34A MTHFR immunoprecipitates in (a) were further analyzed on Phos-tag gels, with or without Calf Intestinal phosphatase (CIP) treatment. Bottom, S(pT)P motif antibody showed reactivity towards the upper, but not the lower, band of S30A MTHFR on phos-tag gel.
www.nature.com/scientificreports www.nature.com/scientificreports/ 293FT cells with a relatively specific DYRK inhibitor, harmine 12 . Although wild-type MTHFR was largely unaffected, S30A-MTHFR was rapidly dephosphorylated, as indicated by the decrease in intensity of the band recognized by the S(pT)P motif antibody (Fig. 3a). The reason why wild-type MTHFR did not respond to harmine will be discussed in the "Discussion".
Furthermore, we treated MDA-MB-468 cells with siRNAs against DYRK1A and/or DYRK2. The treatment with siRNAs against both DYRK1A and DYRK2 led to a visible increase in the lower, presumably unphosphorylated band in both cell lines (Fig. 3b). Together, the results suggest that DYRK1A and DYRK2 are responsible for MTHFR T34 phosphorylation. Nevertheless, since the reduction in phosphorylation was incomplete, it is likely that additional kinases also work on this site.

Evidence for subsequent involvement of GSK3 in MTHFR phosphorylation.
To test if GSK3 is responsible for subsequent phosphorylations, we treated Flag-tagged MTHFR overexpressing MDA-MB-468, 293FT, MCF7, and HelaS3 cells with the GSK3 inhibitor CHIR99021. Although dephosphorylation was slow, the appearance of a lower Flag-tagged MTHFR band was apparent by 3 h and was accompanied by an increase in intensity of the band recognized by the S(pT)P motif antibody (Fig. 4a). Thus, GSK3 inhibition mimics the effect of S30A mutation (Fig. 2a,b), indicating that S30 is one of the GSK3 phosphorylation sites (Fig. 1b). To rule out overexpression artifacts, using CRISPR we introduced a C-terminal Flag tag into the endogenous MTHFR locus in MDA-MB-468 cells, and the same result was obtained with GSK3 inhibition (Fig. 4b).
To complement pharmacology with genetics, we also tested the effect of inducible knockdown of GSK3A in MDA-MB-468 parental and GSK3B CRISPR knockout cells. Whereas knocking down GSK3A in the parental background had no effect, in the GSK3B-knockout cells this led to marked reduction in MTHFR phosphorylation, as indicated by the prominence of the lower MTHFR band (Fig. 4c). Together, these results point to the role of both GSK3A and GSK3B in MTHFR phosphorylation.

MTHFR N-terminal phosphorylation is maintained by SAM binding to the C-terminal regulatory domain.
Previous studies have suggested that MTHFR phosphorylation can be reduced upon methionine deprivation or homocysteine addition, conditions associated with a decreased cellular SAM/SAH ratio 7,13 . www.nature.com/scientificreports www.nature.com/scientificreports/ We have confirmed these findings in both HCT116 and MDA-MB-468 cells (Fig. 5a,b). Moreover, we found that a gradual decrease in medium methionine concentration from supraphysiological to subphysiological levels also reduced MTHFR phosphorylation (Fig. 5c). The point at which significant reduction in MTHFR phosphorylation started to occur (at 50 μM medium methionine) coincided with a sudden drop in intracellular SAM levels ( Fig. 5d), suggesting that adequate SAM levels are crucial for maintaining MHTFR phosphorylation.
Because human MTHFR has a C-terminal regulatory domain that binds SAM, we questioned whether this domain is essential for sensing SAM levels and sustaining MTHFR phosphorylation. To this end, we engineered a chimeric MTHFR where the human N-terminal domain is fused to the Arabidopsis C-terminal domain that is known to be insensitive to SAM regulation 14 . Unlike human MTHFR, this chimeric MTHFR and its various point mutants did not display differences in mobility on SDS-PAGE (cf. Figs 5e with 2a,b), suggesting lack of phosphorylation.
To further confirm the lack of phosphorylation for the chimeric protein, we performed an in vitro kinase assay where purified DYRK1A was allowed to react with immunoprecipitated human or chimeric MTHFR in the presence of ATP. For human MTHFR, there was little reaction because the majority of the protein was already phosphorylated. By contrast, the chimeric MTHFR with the wild-type, but not T34A, serine-rich region was efficiently phosphorylated by DYRK1A, as indicated by the s(pT)P motif band and by the band shift on the phos-tag gel (Fig. 5f). Together, these results suggest that maintenance of mammalian MTHFR N-terminal phosphorylation is critically dependent on SAM binding to the C-terminal regulatory domain, presumably due to the protection of these sites from dephosphorylation when bound to the C-terminal regulatory domain that is occupied by SAM.

Wild-type and non-phosphorylatable mutant MTHFR respond to homocysteine differently.
Phosphorylated MTHFR is less active than unphosphorylated MTHFR at physiologically relevant SAM concentrations (1-3 μM) in vitro 8 . To confirm this in cells, using CRISPR we knocked in non-phosphorylatable mutations at the endogenous MTHFR locus in MDA-MB-468 and HCT116 cell lines (Fig. 6a,b), and compared their folate response to homocysteine treatment. Most of these knockin cells proliferate at rates comparable to the parental lines (Fig. S2). For both parental lines, homocysteine addition caused a dramatic increase in 5-CH 3 -THF production (Fig. 6c,d). By contrast, the knockin cell lines with non-phosphorylatable MTHFR had high basal levels www.nature.com/scientificreports www.nature.com/scientificreports/ of 5-CH 3 -THF and did not respond to homocysteine stimulation (Fig. 6c,d). For MDA-MB-468 S30A knockin (which partially retains T34 phosphorylation, Fig. 2), homocysteine addition caused an increase in 5-CH 3 -THF, but the magnitude of increase was less than that of MDA-MB-468 parental cells (Fig. 6c). Thus, wild-type MTHFR became more active as the SAM/SAH ratio decreased and as its phosphorylation was reduced, both as a result of homocysteine treatment. These results suggest that MTHFR can be switched on through dephosphorylation, to control the level of homocysteine by providing the methyl group for homocysteine remethylation. www.nature.com/scientificreports www.nature.com/scientificreports/

Discussion
In this study, we identified DYRK1A/2 and GSK3A/B among the kinases that phosphorylate the N-terminal serine-rich region of MTHFR (Fig. 7a). Both of these kinases are known to be constitutively active in resting cells 15,16 , thus explaining why the majority of MTHFR is phosphorylated under basal conditions. However, these kinases could be inhibited under certain conditions, thereby altering the degree of MTHFR phosphorylation. For example, GSK3 is known to be inhibited by PI3K-Akt signaling 17 , so the level of MTHFR phosphorylation might be reduced under conditions of PI3K hyperactivation, which occurs often in cancers. By contrast, the regulation of DYRK1A/2 is poorly understood 15 . DYRK1A is overexpressed in Down Syndrome, and DYRK1A overexpression in mice led to perturbations of the methionine cycle, specifically a decrease in plasma homocysteine level 18 . Such an effect of DYRK1A overexpression is not predicted by our model, where DYRK1A-primed MTHFR phosphorylation would lead to a decrease in MTHFR activity, hence an increase in homocysteine.
We confirmed previous findings that MTHFR N-terminal phosphorylation is also regulated by the cellular SAM/SAH ratio (Fig. 7a) 7,13 . Furthermore, we provided evidence that this regulation requires the C-terminal SAM binding domain. Such interaction between MTHFR N-terminal phosphorylation region and the C-terminal SAM-binding domain is well-supported by the recently solved crystal structure of human MTHFR, which suggests proximity of these two regions 8 . Presumably, when SAM is bound, the N-terminal phosphorylation region interacts with the C-terminal domain, such that the multiple phosphorylation sites are buried and therefore shielded from phosphatases (Fig. 7b). The strength of this interaction might depend on the number of phosphates. If so, this could explain why the DYRK1A inhibitor harmine failed to dephosphorylate wild-type MTHFR but worked efficiently on the singly phosphorylated S30A species (Fig. 3a). On the other hand, in the absence of SAM or when SAH instead is bound, this interaction is weak or absent, so the phosphorylation sites are exposed to phosphatases, leading to dephosphorylation.
The model presented in Fig. 7b can also explain why the multisite phosphorylation increases the sensitivity to SAM inhibition. There are five distinct conformation states: the top two are active, the middle two are inactive, and the bottom one is also inactive. SAM preferentially binds to and stabilizes the middle two inactive

P a r e n t a l S 2 6 A ,T 3 4 A / T 3 4 A ( A 3 ) S 3 0 A ( A 8 ) P r o 3 5 -d e l P a r e n t a l T 3 4 A ( 1 A 9 ) T 3 4 A ( 2 G 6 ) T 3 4 A -d e l
c d www.nature.com/scientificreports www.nature.com/scientificreports/ conformations. When MTHFR is multiply phosphorylated, however, SAM binding further shifts the middle inactive state towards the bottom inactive state. Thus, in the presence of SAM, the proportion of inactive states is higher than without the multisite phosphorylation.
What's the physiological function of MTHFR multisite phosphorylation? We surmise that under high methionine conditions, the constitutive, multisite phosphorylation ensures that MTHFR is maximally inhibited by SAM, so that one-carbon units can be spared for crucial processes such as purine and dTMP synthesis. Under low methionine conditions, MTHFR is dephosphorylated and becomes more active (Fig. 5c), so that more one-carbon units are diverted to produce methionine and SAM. In addition, our results with homocysteine-treated cells (Fig. 6c,d) suggest that MTHFR can be dephosphorylated and disinhibited to provide 5-CH 3 -THF for homocysteine remethylation. High concentrations of homocysteine are toxic, posing risk for cardiovascular diseases 19 . Thus, MTHFR dephosphorylation and disinhibition may also represent a cellular response to ameliorate hyperhomocysteinemia. Figure 7. Models. (a) MTHFR multisite phosphorylation integrates various signals, such as nutrient inputs that influence the cellular SAM:SAH ratio, and signaling pathways that modulates DYRK1A and GSK3 kinase activities. (b) A molecular model explaining why MTHFR phosphorylation responds to changes in the SAM/ SAH ratio, and why multiply phosphorylated MTHFR is more sensitive to SAM inhibition. In this model, multisite phosphorylation of the N-terminal tail of MTHFR allows a new inactive conformation (B3), where the N-terminal tail is not accessible for phosphatases. When SAM level is high, SAM can preferentially bind to and stabilize the inactive states A2, B2, and B3, thereby shifting the equilibrium to these inactive states including B3 that cannot be dephosphorylated. Conversely, when SAH level is high, the action of SAM is reversed, thereby favoring conformations (B1 and B2) that can be readily dephosphorylated. Multiply phosphorylated MTHFR is more sensitive to SAM inhibition because, unlike unphosphorylated MTHFR, it can be further trapped in the inactive, B3 state.
All these cell lines were cultured in DMEM with 10% FBS.
To determine the phosphorylation motifs for the kinases, in vitro kinase assays were performed with recombinant kinases on the kinase substrate library in the presence of ATP[γ-32P]. There reactions were carried out in Kinase Buffer I (SignalChem) at 30 °C for 90 minutes. The peptides, which contain C-terminal biotinyl groups, were blotted onto streptavidin-conjugated membranes and imaged with a Typhoon FLA 7000. Detailed information of the experimental protocol is provided elsewhere 20 .
The spot densitometry values were used to generate sequence logos. The densitometry matrix was normalized by each row, in order to create a probability matrix, was converted to a probability sequence logo. All the sequence logos were generated in R 3.5.1 using the 'ggseqlogo' package 21 .
CRISPR/Cas9-mediated genome-editing in cell lines. Single guide RNA sequence was cloned into PX459_V2.0 (Addgene) according to a published protocol 22 . Cell lines with transfected with guide using Lipofectamine 3000 or Lipofectamine LTX (Invitrogen). 24 hr post transfection, puromycin (typically at 1 μg/ml) was added to enrich positively transfected cells. The puromycin selection typically lasts for 2 days, until mock-transfected, control cells were completely eliminated by puromycin. Single cell cloning was performed by serial dilution in 96-well plates. After 1.5-3 weeks, 15-20 clones were picked for knockout verification.
If antibodies were available, knockout clones were first screened by western blotting. The genotypes of selected clones were determined by PCR amplification of a genomic region encompassing the editing site, TOPO cloning of the individual alleles into PCR4-TOPO vector (Invitrogen), and Sanger sequencing. At least 20 bacterial colonies were picked for sequencing to ensure approximately equal representation of both alleles. We were unable to find a specific DYRK2 antibody, so DYRK2 knockout status was determined by TOPO cloning followed by Sanger sequencing.
MTHFR CRISPR Knockin cells was generated using nucleofection of sgRNAs together with ssODN repair templates. The PAM sequence was mutated and a restriction site was introduced into the ssODN repair templates. Correct clones were initially screened using restriction digest, and further verified using Sanger sequencing of the genomic DNA PCR amplicon. siRNA and inducible shRNA knockdown. Pooled siGENOME siRNA (30 nM) against DYRK1A and DYRK2 were purchased from Dharmacon and transfected using Lipofectamine RNAiMAX reagent.
DOX-inducible GSK3A knockdown target sequence was predicted using the Splash algorithm (http:// splashrna.mskcc.org/) and cloned into the inducible lentiviral vector according to a previously published protocol 23 . Lentivirus-infected, GFP-positive cells were sorted prior to analysis.
Cloning of expression constructs. MGC cDNA clones of human MTHFR was purchased from Open Biosystems.
The cDNA for Human/Arabidopsis MTHFR chimera was synthesized by IDT. The coding region contains the first 357 amino acids of human MTHFR, followed by KNEE, followed by the last 266 amino acids of Arabidopsis MTHFR. The cDNA was then subcloned into pLEX_306 using gateway technology. www.nature.com/scientificreports www.nature.com/scientificreports/ Thermo). Lysates were cleared by centrifugation at 13,000 g at 4 °C for 10 min. Protein concentration was determined by BCA assay. Proteins were denatured in NuPAGE LDS 4X sample buffer (Invitrogen) supplemented with β-mercaptoethanol. Typically, 20-50 μg of total protein was loaded per lane on an 8% Bis-Tris Bolt gel (Invitrogen), and the apparatus was kept on ice, for a 2-3 hour run at 120 volts. 7.5% SuperSep Phos-tag gels were run according manufacturer's protocols (Wako Cat# 198-17981); in particular, 10 mM MnCl 2 was included in each sample, and EDTA washes were performed before proceeding to the transfer step. SDS-PAGE was followed by standard immunoblotting steps. Immunoblot signals were detected by enhanced chemiluminescence.

DYRK1A protein purification and
In vitro kinase assay. For DYRK1A protein expression, One-shot E.coli BL21(DE3) (Invitrogen) was transformed with DYRK1A-pNIC28 vector (Addgene, Cat# 38913). The bacteria were grown at 37 °C in 2xYT medium containing 100 μg/ml Carbenicillin until A600nm reached ~0.6, cooled down to room temperature, followed by 0.5 mM IPTG induction of protein expression at room temperature for 16 hours. The DYRK1A protein was purified using the standard protocol for Ni-NTA purification under native conditions as described in "The QIAexpressionist Handbook".

LC-MS analysis of SAM and SAH.
Pre-cooled 80% methanol with 0.1% formic acid (1 mL) was added to each 6-cm dish of 70% confluent cells. The lysate was scraped into an Eppendorf tube and then centrifuged at 4 °C for 10 minutes at 14,000 rpm. The supernatant was evaporated using a Speed-Vac. Targeted LC/MS analyses were performed on a Q Exactive Orbitrap mass spectrometer (Thermo Scientific) coupled to a Vanquish UPLC system (Thermo Scientific). The Q Exactive operated in positive mode. An XBridge amide column (2.1 mm i.d. × 150 mm, Waters) was used for separation of metabolites. The flow rate was set at 150 μL/min. Mobile phases consisted of 0.1% formic acid in acetonitrile for line A, and 0.1% formic acid in water for line B. Gradient ran from 70% to 0% A in 13 min followed by a wash with 0% A and re-equilibration at 70% A. Metabolites were identified on the basis of exact mass within 5 ppm and retention time. Relative metabolite quantitation was performed based on peak area for each metabolite. Data analysis was done using in-house written scripts. SAM and SAH metabolites were also manually checked using Thermo X-Caliber. Normalization was based on protein content.
HpLC analysis for intracellular folates. HPLC analysis of intracellular folates was performed as previously reported 24 . Briefly, cells were cultured in 50 nM 3 H-5-formyl-tetrahydrofolate for 48 hr. Then cellular folates were liberated from proteins and protected from air oxidation by boiling in 50 mM Tris-HCl pH7.4 buffer with 3% (w/v) ascorbate and 200 μM β-mercaptoethanol. The polyglutamate tails of folates were then cleaved by recombinant γ-glutamyl hydrolase (GGH), and the resulting folate monoglutamates were resolved by reversed phase ion-pair HPLC with radioactivity detection.