Folate-dependent methylation of septins governs ciliogenesis during neural tube closure

Abstract

Periconception maternal folic acid (vitamin B₉) supplementation can reduce the prevalence of neural tube defects (NTDs), although just how folates benefit the developing embryo and promote closing of the neural tube and other morphologic processes during development remains unknown. Folate contributes to a 1-carbon metabolism, which is essential for purine biosynthesis and methionine recycling and affects methylation of DNA, histones, and nonhistone proteins. Herein, we used animal models and cultured mammalian cells to demonstrate that disruption of the methylation pathway mediated by folate compromises normal neural tube closure (NTC) and ciliogenesis. We demonstrate that the embryos with NTD failed to adequately methylate septin2, a key regulator of cilium structure and function. We report that methylation of septin2 affected its GTP binding activity and formation of the septin2-6-7 complex. We propose that folic acid promotes normal NTC in some embryos by regulating the methylation of septin2, which is critical for normal cilium formation during early embryonic development.—Toriyama, M., Toriyama, M., Wallingford, J. B., Finnell, R. H. Folate-dependent methylation of septins governs ciliogenesis during neural tube closure.

The neural tube serves as the precursor structure of the entire central nervous system, including the brain and spinal cord, forming early in embryonic development. During the process of neural tube closure (NTC), the neural plate dynamically changes its shape through a complex series of highly regulated morphogenetic processes. The earliest recognition of a developing nervous system is the establishment of the neural plate, which is a thickening of pseudostratified epithelium covering the midline structures, such as the notochord and the paraxial mesoderm (1). Where the neural plate meets the surface ectoderm, the neural folds form, and biomechanical forces enable them to elevate, thus establishing the neural groove running along the rostral–caudal embryonic axis. At the initial point of neural fold epithelial cell fusion, the neural tube closes bidirectionally, reaching the hindbrain rostrally and posteriorly to the caudal neuropore (2). Discontinuous closure occurs between the rostral aspect of the forebrain, proceeding caudally to meet the closure moving cranially from the hindbrain. NTC progresses in a zipper-like but discontinuous fashion in both directions (rostrocaudal and caudorostral) (3–5). Rostral NTC is completed before posterior closure, which involves canalization of a mesenchymal core of progenitor cells, by ∼2 d (6).

Failure of NTC results in a variety of possible neural tube defects (NTDs), including the 2 most common forms, anencephaly and myelomeningocele (spina bifida). NTDs are the second most common structural malformation in humans, and recent estimates indicate that they have a prevalence in the United States of 1 in 3000 and a worldwide prevalence per 1000 live births from 1 (in Europe and the Middle East) to 3 (in 2014 in northern China after folate supplementation campaigns, down from 10 in 2000–2004) (7–9).

For the past 30 yr, it has been recognized that folic acid supplementation markedly decreased the prevalence of NTDs, resulting in a United States government–mandated folic acid fortification of certain foods, a program that has now been implemented in more than 80 countries worldwide. Despite the wide acceptance of folic acid as an important cofactor to prevent NTDs, the mechanism by which folate reduces the risk for these devastating congenital malformations remain unknown.

Our laboratory previously showed that inactivation of the folate receptor 1 (Folr1) gene resulted in a highly penetrant NTD phenotype in the mouse (10). Folate contributes to 1-carbon metabolism, which is used for the de novo synthesis of purines, is involved in methionine recycling pathways, and contributes to DNA and protein methylation (11). S-adenosyl methionine (SAM), a derivative form of methionine, serves as a methyl group donor for the methylation of DNA, histones, and nonhistone proteins. It is widely accepted that folate depletion decreases SAM concentrations, which subsequently decreases DNA methylation in humans (12, 13). Methyl groups are added to cytosine-guanine (CpG) dinucleotide islands, which are inherited by daughter cells. DNA or histone methylation affects gene transcriptional activity, which, if adversely altered, can compromise normal NTC. For example, it is known that disruption of DNA methyltransferases and histone demethylase in mice results in NTDs (14, 15). These studies demonstrate the importance of maintaining methylation homeostasis such that the regulation of genes critical for NTC is not compromised by low folate status.

Protein methylation, including histone methylation, is a post-translational modification whereby a methyl group is added to an arginine or lysine residue. On the arginine residue, 1 to 2 methyl groups are added, forming mono- or dimethylated arginine. On the lysine residue, 1–3 methyl groups are added, forming a mono-, di-, or trimethylated lysine. Methylation controls protein–protein interactions and protein functions (16). As such, it is important to better understand just what proteins are being methylated and what the functions of nonhistone protein methylation are during the period of NTC, if we ever hope to understand the myriad of events that occur leading up to an NTD.

Herein, we provide evidence that a functional methylation pathway is necessary for proper NTC, acting at the level of cilia, which are essential organelles for cell–cell signaling (17, 18). We further demonstrate that folate availability affects the methylation status of septin2, a cytoskeletal protein required for normal ciliogenesis (19). We found that septin2 is methylated on lysine and arginine residues in human cells and that this methylation is essential for the normal regulation of septin2 functions, including GTP binding activity and septin2-6-7 complex formation. Finally, we demonstrated that defective septin2 methylation correlates positively with ciliogenesis and Hedgehog signaling defects.

MATERIALS AND METHODS

Mouse strains

All procedures were performed in accordance with the Guide for the Care and Use of Laboratory Animals [National Institutes of Health (NIH), Bethesda, MD, USA] and all protocols were approved by the Institutional Animal Care and Use Committee of The University of Texas at Austin. Mouse embryonic fibroblasts (MEFs) were derived from mice carrying the floxed conditional-knockout cassette between exon2 of slc19a1, were generated by targeted recombination.

Xenopus embryo manipulations and drug treatment

Female adult Xenopus laevis were ovulated by injection of human chorionic gonadotropin, and the eggs were fertilized in vitro and incubated in 0.3× Marc’s modified Ringer’s (MMR) solution. At stage 13, the vitelline membrane was removed by forceps, and the embryos were treated with adox or DMSO as a control, after which they were allowed to incubate at room temperature. Embryos were fixed with minimum essential medium without folic acid (MEM-FA) for 2 h at room temperature, and bright-field images of the neural tube were captured on a stereomicroscope (Axiozoom V16; Zeiss, Thornwood, NY, USA). The distance of between neural folds was measured with Fiji software (NIH).

Immunostaining of primary cilia

NIH3T3 or MEFs were cultured on 24-well plates, fixed. and immunostained. Transverse sections of Xenopus embryos were made on a Vibratome series 1000 (Leica Biosystems, Wetzlar, Germany) and blocked with 10% fetal bovine serum (FBS) in 1× PBS containing 0.05% Triton X-100 (PBST). Primary cilia were stained with anti-acetylated-α tubulin (1:2000, clone 6-11B-1; Sigma-Aldrich, St. Louis, MO, USA), or anti-Arl13B (1:2000, 17711-1-AP; Proteintech, Chicago, IL, USA). Cell membranes were stained with anti-ZO-1 antibody. Primary antibody was detected with Alexa Fluor 555 or Alexa Fluor 488 conjugated goat anti-mouse or rabbit antibodies (Thermo Fisher Scientific, Waltham, MA, USA). Xenopus images were obtained with an LSM700 confocal microscope (Zeiss). The other images were obtained using the Operetta high-content screening microscope (Perkin Elmer, Waltham, MA, USA).

Antibodies and reagents

Antibodies against slc19a1 (ab62302) and methylated lysine (ab23366) were purchased from Abcam (Cambridge, MA, USA). Acetylated tubulin (clone 6-11B-1, T6793) and FLAG (clone M2, F1804) antibodies were acquired from Sigma-Aldrich. Gli3 (H280, sc-20688) antibody was purchased from Santa Cruz Biotechnology (Dallas, TX, USA). Myc (clone 9E10, 631206) antibody was obtained from Takara Bio USA (Mountain View, CA, USA). mCherry (AP32117PU-S) was purchased from Acris Antibody (Rockville, MD, USA). Green fluorescent protein (GFP; A11122) antibody was obtained from Thermo Fisher Scientific. Arl13b (17711-1-AP) antibody was bought from Proteintech. Septin6 (clone 9E7, 05-1566) and septin7 (ABT354) were obtained from EMD-Millipore (Billerica, MA, USA). α-Tubulin (clone E7) was obtained from the Developmental Studies Hybridoma Bank (University of Iowa, Iowa City, IA, USA).

Fluorescent secondary antibodies for immunofluorescence (Alexa Fluor 680 goat anti-mouse IgG; A21057, Alexa Fluor 488 goat anti-mouse IgG; A11001, Alexa Fluor 647 goat anti-rabbit IgG; A21244) were all obtained from Thermo Fisher Scientific. IRDye 700CW and IRDye 800CW secondary antibodies for immunoblot analyses were purchased from Li-Cor Biosciences (Lincoln, NE, USA). FLAG peptide (F3290) was obtained from Sigma-Aldrich. Folinic acid calcium salt hydrate (F7878) and adox (adenosine, periodate oxidized, A7154) were obtained from Sigma-Aldrich.

Plasmids

pMK782-septin6-GFP (38296), pMSCV-IRES-Puro-GFP (21654), pMSCV-IRES-Puro-GFP-cre (50935), and pCMV-VSV-G (8454) plasmids were purchased from Addgene (Cambridge, MA, USA). pLX304-septin2 (HsCD00444076), pANT-7cGST-septin7 (HsCD00403339), and pDONR221-Slc19a1 (HsCD00043914) were obtained from DNASU (20).

FLAG-tagged septin2 was generated by PCR and inserted into the pCAG vector. Full-length septin2 cDNA was inserted into pEGFP-C1 (Takara Bio USA) and pCS-mCherry vectors. Full-length septin7 and Slc19a1 cDNA was inserted into pCMV-myc vector.

Point mutants of septin2 at Lys 183 (K183) and Arg (R300) were generated by overlap extension PCR, and using the following primers: K183A: 5′-TGTCATTGCAGCAGCAGCTGACACTC-3′; K183Q: 5′-TGTCATTGCACAAGCTGACACTC-3′; K183R: 5′-TGTCATTGCAAGAGCTGACACTC-3′; R300A: 5′-GAAAACTTCGCTTCTGAGAGAC-3′, R300Q: 5′-GAAAACTTCCAATCTGAGAGAC-3′; and R300K: 5′-GAAAACTTCAAATCTGAGAGAC-3′.

Point mutants of slc19a1 at 500G, 532G, and 535C were generated by overlap extension PCR, then inserted into a pCMV-Myc vector by using the following primers: 500GA, 5′-GCGTGTTCACCAACTCCGTG-3′; 532G, 5′-GTCACTGTGAGCCGAGTCTC-3′; and 535CT, 5′-GTCACTGTGGGCTGAGTCTC-3′.

Cell culture, transfection, and retrovirus infection

HEK293T, MEF, and NIH3T3 cells were maintained as previously described (21). Twenty-four hours before the experiments, the cells were cultured in RPMI1640, folic-acid–free medium containing 10% dialyzed FBS. Plasmids were transfected into these cells with Polyethylenimine Max (Polyscience, Niles, IL, USA). Primary MEFs were isolated from E9.5 slc19a1 flox/flox mouse embryos and cultured in RPMI1640, folic-acid–free medium containing 200 µg/ml folinic acid and 10% dialyzed FBS. Retrovirus was produced in Phoenix GP cells (American Type Culture Collection, Manassas, VA, USA) by transfecting pCMV-VSV-G and pMSCV with retroviral vectors. Two days after transfection, medium was collected, and retrovirus was centrifuged at 4000 g for 30 min in polyethylene glycol (PEG) 6000 solution (8% polyethylene glycol 6000 (Santa Cruz Biotechnology), 100 mM NaCl, 10 mM HEPES (pH 7.8, adjusted to pH 7.8 with NaOH). MEFs were infected with retrovirus, and 4 d later, they were used for assay procedures.

GTP binding assay

HEK293T cells were transfected with FLAG-septin2. Sixteen hours after transfection, the medium was changed to RPMI1640 medium containing 300 µg/ml folinic acid and 10% dialyzed FBS (Thermo Fisher Scientific). Adox (40 µM) was added to the medium for 16 h before cells were collected. Cells were lysed with lysis buffer [25 mM Tris-HCl (pH 7.5), 150 mM NaCl, 1% NP-40, and 50 mM EDTA], then FLAG-septin2 was immunoprecipitated with FLAG antibody (Sigma-Aldrich). After elution of FLAG-septin2 with 50 µg/ml FLAG peptide in 60 µl lysis buffer (Sigma-Aldrich), a GTP binding assay was performed. Fifty microliters of eluate was used for each assay. ActivX desthiobiotin-GTP probe (10 µM; Thermo Fisher Scientific) and MgCl₂ (10 mM) was added and incubated at room temperature for 10 min. To stop the reaction, 400 µl of lysis buffer containing 5 M urea was added, and then GTP binding septin2 was precipitated by streptavidin agarose beads (Thermo Fisher Scientific). GTP binding septin2 was detected by Western blot analysis.

Mass spectrometry

HEK293T cells were transfected with FLAG-septin2 and cultured in RPMI1640 medium containing 200 µg/ml folinic acid and 10% dialyzed FBS for 2 d. Cells were lysed with lysis buffer [25 mM Tris-HCl (pH 7.5), 150 mM NaCl, 1% NP-40, and 50 mM EDTA], and then septin2 was immunoprecipitated by FLAG antibody. Septin2 protein was eluted by boiling Protein G beads with Laemmli buffer. After this step, the septin2 was separated by SDS-PAGE and stained with Coomassie brilliant blue solution. After the band was cut from the gel, in-gel tryptic digestion was performed. The gel was destained overnight with 50% methanol and 5% acetic acid and dehydrated with acetonitrile. The protein in the gel was reduced by 10 mM DTT for 1 h, and alkylated with 50 mM iodoacetamide in 100 mM ammonium bicarbonate. After the gels were washed several times, the proteins in the gel were digested by 20 ng/l trypsin, overnight. Finally, digested peptides were extracted with 5% formic acid in 50% acetonitrile. The Proteomics Facility at The University of Texas at Austin performed the MS/MS analysis.

Septins complex formation analysis

HEK293T cells were transfected with FLAG-septin2 and GFP-septin6. Two days after transfection, cell lysate were prepared with lysis buffer [25 mM Tris-HCl (pH7.5), 150 mM NaCl, 1% NP-40, and 50 mM EDTA], then FLAG-septin2 was immunoprecipitated with FLAG antibody. Precipitated proteins were eluted by boiling protein resin. Coprecipitated septins were immunoblotted by these antibodies.

RESULTS

Slc19a1 deletion in neural crest and in yolk sac exhibit neural tube defects

Slc19a1, previously called the reduced folate carrier (22, 23), is a ubiquitously expressed protein that mediates intake of reduced folate into cells (24). The slc19a1 conventional-knockout embryos were embryolethal at embryonic day (E)6.5, making it difficult to investigate the molecular function of the slc19a1 protein during NTC (25). To solve this problem and also to better understand the function of slc19a1 in specific tissues, we generated slc19a1 flox/flox mice, with the slc19a1 gene being inactivated after crossing with a Cre line mouse, such that the mice were viable, and we could carefully observe their phenotype during NTC. Exon2 containing the first codon, is flanked with LoxP sites, so that recombination at LoxP sites disrupted slc19a1 gene expression (Fig. 1A). The genotype was determined by PCR performed on tissues extracted from the yolk sac or tail (Fig. 1B). To obtain embryos ubiquitously lacking a functional slc19a1 gene, slc19a1^+/−;E2A-Cre^+/− mice were crossed with slc19a1 flox/flox mice (Fig. 1C) (26), and 71.4% of the slc19a1-null embryos exhibited an open neural tube (Fig. 1D, E and Table 1). None of the slc19a1-null embryos was observed at E10.5 because of in utero lethality (data not shown), indicating that this gene is essential for early embryonic development.

Figure 1.

slc19a1 conditional-knockout mice show neural tube defects. A) The endogenous slc19a1 gene comprising exon 6 (top). After homologous recombination, the neomycin phosphotransferase (Neo) cassette was initially flanked by the FRT site and subsequently crossed with Cre line mice (bottom). B) PCR genotyping was performed using primer 1 (P1, A) and primer 2 (P2, A). Genomic DNA was extracted from the yolk sac or tail. C) Scheme of crossbreeding. Slc19a1 flox/flox mice were crossed with RFC1^+/−;E2A-Cre mice. Embryos were isolated and observed at E9.5. D, E) RFC1 flox/⁻;E2A-cre^+/− embryo (E) exhibited an open neural tube (asterisk) compared with wild-type (D). F) Scheme of crossbreeding. Slc19a1 flox/flox mice were crossed with slc19a1 ^+/−;Wnt1-Cre^+/− mice. Embryos were isolated and observed at E10.5, E13.5, E16.5, and E18.5. G) The comparison image between wild-type (right) and slc19a1 flox/⁻;Wnt1-cre^+/− (left) embryo. H) A magnified image of a slc19a1-knockout embryo; asterisk: open neural tube. I) Scheme of crossbreeding. Slc19a1 flox/flox mice were crossed with slc19a1^+/−;Ttr-Cre^+/− mice. Embryos were isolated and observed at E9.5, E13.5, and E18.5. J, K) The comparison image with wild-type (left) and slc19a1 flox/⁻;Ttr1-cre^+/− (right) embryo at E13.5; asterisk: open neural tube.

TABLE 1.

Phenotype of the Cre promoter E2A-driven knockout embryos

				NTD [n (%)]
Dam	E	Total embryos	Resorption	flox/+;cre−/−	flox/−;cre+/−a
				flox/−;cre−/−
				flox/+;cre+/−
3	9.5	34	2	0/25 (0)	5/7 (71.4)

^aConditional knockout.

To further explore the function of slc19a1 specifically in neural crest cells, matings were established using slc19a1^+/−;wnt1-cre^+/− mice, in that the Cre protein in this case is specifically expressed in neural crest cells (Fig. 1F) (27, 28). We observed that 34.6% of slc19a1 deletion mutant embryos exhibited open NTDs at E9.5. However, no NTD-affected fetuses were observed when examined at E13.5, E16.5, or E18.5 (Fig. 1G, H and Table 2). This result suggests that slc19a1 gene function in Wnt-1 promoter–activated tissues is essential for NTC.

TABLE 2.

Phenotype of the Cre promoter Wnt-1-driven knockout embryos

				NTD [n (%)]
Dam	E	Total embryos	Resorption	flox/+;cre−/−	flox/−;cre+/−a
				flox/−;cre−/−
				flox/+;cre+/−
8	10.5	72	3	0/46 (0)	9/26 (71.4)
4	13.5	34	5	0/29 (0)	0/5 (0)
4	16.5	32	7	0/28 (0)	0/4 (0)
4	18.5	30	5	0/24 (0)	0/6 (0)

^aConditional knockout.

We also evaluated embryos in which the slc19a1 gene was specifically deleted in the yolk sac by mating with the TTR-Cre strain mice (29). At E9.5, 33.3% of the conditional-knockout embryos had an open neural tube, and this phenotype was similar to that of the neural crest-specific knockout embryos (Fig. 1G, H). One embryo had an open neural tube at E13.5, although no NTDs were observed at E16.5 and E18.5 (Table 3). These results also suggest that a folate pathway in the neural crest and yolk sac is necessary for proper NTC and embryonic development.

TABLE 3.

Phenotype of Cre promotor Ttr-driven knockout embryos

				NTD [n (%)]
Dam	E	Total embryos	Resorption	flox/+;cre−/−	flox/−;cre+/−a
				flox/−;cre−/−
				flox/+;cre+/−
8	9.5	66	6		4/21 (33.3)
12	13.5	100	7	0/75 (0)	2/25 (8.0)
3	18.5	25	3	0/21 (0)	0/4 (0.0)

^aConditional knockout.

Folates regulates Shh signaling via methylation

Questioning whether the NTD phenotypes observed in the mutant mice may be related to alterations in methylation homeostasis, we used the Xenopus embryo, which is a powerful platform for studies of embryonic morphogenesis. Xenopus embryos were cultured in the presence of adenosine dialdehyde (adox), a global methyltransferase inhibitor that acts by inhibiting AdoHcy hydrolase, leading to the elevation of intracellular AdoHcy levels. This increase in AdoHcy concentrations results in a feedback inhibition of most methylation reactions (30). Adox treatment significantly delayed NTC in a concentration-dependent manner (Fig. 2A). When examining the impact of the adox treatment of the Xenopus embryos, the average distance between the neural folds was markedly greater in adox-treated embryos compared to controls, suggesting that these embryos would fail to complete NTC (Fig. 2B). Previous studies showed that NTD is often associated with the misregulation of Hedgehog signaling and that the Hedgehog signaling requires processing via the primary cilium (17). Accordingly, NTD is a commonly observed abnormal phenotype in mice with defective ciliogenesis. We therefore questioned whether our manipulations of the methylation pathway may compromise normal ciliogenesis. Specifically, we immunostained acetylated tubulin after adox treatment and observed a significant reduction in the number of cilia in the neural tube of Xenopus embryos (Fig. 2C).

Figure 2.

The methylation pathway is essential for neural tube closure in Xenopus. A) Dorsal view of stage 19 Xenopus embryos. Dashed lines: the medial side of each neural fold. Neural folds were closely opposed in the control embryo. After adox treatment, the neural tube remained open. B) The average distance between neural folds was 8.06 ± 2.63 mm (n = 25) in control embryos and 22.25 ± 4.27 (n = 33), 47.96 ± 6.74 (n = 29), and 39.5 ± 4.67 mm (n = 31) in 50, 100, and 200 µM adox-treated embryos, respectively. Mean values with bars representing sem are shown. **P < 0.01, ***P < 0.001 (Student’s t test). C) Primary cilia in Xenopus embryos. Arrows: cilia stained by acetylated tubulin.

We subsequently evaluated ciliogenesis under folate-deficient conditions in MEFs derived from wild-type C57BL/6J/129 mice. We found that a reduction in folate concentrations significantly reduced cilium formation (Fig. 3A, B). Furthermore, treatment with the methylation inhibitor adox markedly inhibited cilium formation in these cells (Fig. 3C, D). Finally, because the slc19a1 gene is essential for normal folate transport into cells, we investigated ciliogenesis in MEFs isolated from slc19a1 flox/flox embryos infected with a retrovirus, to induce expression of Cre recombinase. Inactivation of slc19a1 reduced the number of cilia formed, comparable to the effects of either the induced folate deficiency in the mouse or the adox treatment in the Xenopus embryos (Fig. 3E–G). Taken together (Figs. 1–3), these findings raise the possibility that the normal methylation pathway, mediated by folate, is necessary for NTC and also for proper cilium formation in vertebrates.

Figure 3.

Folate promotes ciliogenesis via methylation pathway. A) MEFs were grown in RPMI1640 folic-acid–free medium containing 200 µg/ml (top) or 0 µg/ml folinic acid, a 5-formyl derivative of tetrahydrofolic acid (bottom) for 2 d. Cells were cultured in serum-starved RPMI1640 medium containing 200 µg/ml (top) or 0 µg/ml folinic acid (bottom), for 16 h before fixing. Subsequently, the cells were immunostained with anti-acetylated tubulin (acetyl tub, ciliary marker; green) antibody. Nuclei were stained with DAPI (blue). B) Ciliated cells (A) were counted and graphed. Six different experiments were performed. C) MEFs were treated with DMSO as the control (top) or 10 µM adox, methylation inhibitor (bottom). Cilia were immunostained with anti-acetylated tubulin antibody (green). Nuclei were stained with DAPI (blue). D) The number of ciliated cells (C) was counted and graphed. Three different experiments were performed. E, F) Primary MEFs were isolated from E9.5 slc19a1 flox/flox embryos, then infected for 3 d with a retrovirus harboring GFP as control (E) or GFP with Cre recombinase (F). Cells were grown with RPMI1640, folic-acid–free medium containing 200 µg/ml folinic acid and 10% dialyzed FBS for 2 d. Sixteen hours before fixation, medium was changed to FBS-free RPMI1640 medium containing 200 µg/ml folinic acid. DMSO (top) or 40 µM adox (bottom) treatment was administered for 16 h before fixing, at which time the cells were immunostained with anti-arl13b (green) and DAPI (blue). G) Number of ciliated cells (E, F) were counted and graphed. *P < 0.05, **P < 0.01 (Student’s t test). Error bars ± se.

Folate regulates Shh signaling via methylation

The sonic hedgehog (Shh) signaling pathway is critical to many of the morphogenetic events occurring during embryonic development, including NTC. Shh signaling components are highly abundant in primary cilia, and genetic ablation of cilia is known to produce a range of embryonic structural defects, including NTD (17). Gli3 is a transcription factor regulated by Shh signaling, and plays a crucial role in neural development (31). Having demonstrated that folate and methylation pathways are required for cilium formation (Figs. 1 and 2), we performed experiments to determine whether the folate and methylation pathways regulate the abundance of Gli3R, a truncated form of Gli3. In these experiments, we observed that folate depletion decreased the amount of Gli3R protein in HEK293T cells (Fig. 4A, B). Furthermore, adox treatment markedly reduced Gli3 processing in an adox concentration-dependent manner (Fig. 4C, D). To examine Shh signaling activity in vivo, we isolated MEFs from E9.5 mutant slc19a1^−/− mouse embryos and assayed for Gli3 protein expression by Western blot. In slc19a1^−/− embryos, the amount of Gli3R decreased significantly, compared with wild-type or heterozygous embryos (Fig. 4E). In summary, these data raised the interesting possibility that the folate pathway regulates Shh signaling by controlling ciliogenesis during NTC.

Figure 4.

Folate regulates Shh signaling via methylation pathway. A) Western blot analysis for Gli3 (top band; full-length Gli3, bottom band; truncated Gli3) and β-tubulin in HEK293T cell lysate. HEK293T cells were grown in RPMI1640 medium containing 200 µg/ml folinic acid for 48 h. Adox (20 µM) was added 16 h before the cell lysate was prepared. DMSO was added as a control. B) Relative amount of truncated Gli3 (A) was calculated by normalizing to Gli3 full-length expression and then graphed. Three different experiments were performed. C) HEK293T cells were treated with 0–80 µM adox for 16 h before the cell lysate was prepared. Western blot analysis for Gli3 or β-tubulin was shown. D) Relative amount of truncated Gli3 (C) was calculated and graphed. Three different experiments were performed. E) Slc19a1-knockout mouse embryos were isolated at E9.5. After genotyping using PCR, Western blot analysis for Gli3 and β-tubulin was performed, as in A. *P < 0.05, **P < 0.01, ***P < 0.001 (Student’s t test). Error bars ± se.

Human septin2 is methylated via the folate pathway

Having demonstrated that the methylation pathway is critical for NTC by regulating ciliogenesis and Shh signaling (Figs. 1–4), we had yet to identity the specific targets of methylation. When evaluating a proteomic survey of methylated proteins in yeast, we noted that septin2, known as CDC11 in yeast, was dimethylated at Arg 35, raising the possibility that human septin2 is also methylated (32). This finding was significant, because septin2 is also necessary for generating and maintaining functional cilia, acting as a barrier preventing the diffusion of ciliary membrane proteins (19, 33, 34).

To determine whether septin2 is methylated in human cells, HEK293T cells were transfected with human FLAG-septin2 and immunoprecipitated using FLAG antibody. Methylated lysine residues on the precipitated septin2 proteins were detected by Western blot with an anti-methylated lysine (meK) antibody. In folinic-acid–treated cells, an anti-methylated lysine antibody detected some proteins. In contrast, depletion of folate decreased the intensity—hence, the quantity of the protein signal (Supplemental Fig. 1). Moreover, we found that both the 20 µM adox treatment or culturing HEK cells in a folate-deficient medium significantly reduced meK septin2 levels (Fig. 5A, B). Finally, using liquid chromatography–tandem mass spectrometry (LC-MS/MS), we determined the presence of 10 methylation sites in human FLAG-septin2 (Supplemental Table 1). It is interesting to note that the acetylated lysine residues were not detected by LC-MS/MS. Among these 10 methylated residues, we focused on 2 specific residues that are highly conserved from yeast to humans, a monomethylation at Lys 183 (K183) and a dimethylation at Arg 300 (R300) (Fig. 5C–F). K183 is located within the GTP binding domain, and R300 is exposed at the dimer interface, respectively (35).

Figure 5.

Human septin2 is methylated via folate pathway. A) Western blot analysis for methylated lysine (meK) and FLAG. HEK293T cells were transfected with FLAG-septin2. Thirty-six hours after transfection, the cells were treated with 40 µM adox for 16 h. After preparation of cell lysates, FLAG-septin2 was immunoprecipitated using FLAG antibody. Mouse IgG antibody was used as the control for the immunoprecipitation. B) HEK293T cells expressing FLAG-septin2 were grown in RPMI1640 medium containing either 200 µg/ml folinic acid or folate depletion medium for 2 d. After immunoprecipitation of FLAG-septin2 from cell lysate, Western blot analysis was performed to detect meK and FLAG. C, D) Spectrum data for septin2 peptides containing Lys 183 (C) and Arg 300 (D) by LC-MS/MS. FLAG-septin2 purified from HEK293T cells were digested by trypsin, and mass spectrometry was performed. Molecular weight increase secondary to methylation (MW 14) at both residues was identified. E, F) Sequence alignment of septin2 from human to yeast. Methylated lysine (E) and dimethylated arginine (F) are highlighted in yellow. Highly conserved residues from human to yeast are highlighted in blue. Nonconcordance with human residues is highlighted in green.

Methylation of septin2 controls its GTP binding and hetero-oligomerization activities

Septins are GTP binding cytoskeletal proteins that form hetero-oligomeric complexes and assemble into filaments or ring structures during cell division. Our working hypothesis was that methylation at the K183 and R300 residues may regulate septin2 functions, including GTP binding activity and septin-complex formation. Both the adox treatment and culturing the HEK293T cells in folate-deficient medium increased the GTP binding affinity of septin2 (Fig. 6A, B), which suggests that methylation of septin2 attenuates GTP binding. Unexpectedly, magnesium was not necessary for GTP binding. As a further test, using site-directed mutagenesis, we generated K183 and R300 point mutants (Lys to Gln or Arg or Ala and Arg to Gln or Lys or Ala, respectively). Mutations in K183 to Gln (K183Q) reduced the GTP binding affinity compared with the wild-type, K183A, K183R, and R300 mutants (Fig. 6C, D). These findings suggest that methylation at K183 is important for regulating GTP binding and suggests that the K183Q mutation works as a methylation mimic protein.

Figure 6.

Methylation of septin2 regulates GTP binding affinity and formation of septin complex. A) HEK293T cells were transfected with FLAG-septin2 and cultured in RPMI1640 medium containing 200 µg/ml folinic acid for 2 d. The cells were treated with 40 µM adox for 16 h before they were collected. After immunoprecipitation of FLAG-septin2, GTP binding assays were performed. The biotinylated GTP binding form of FLAG-septin2 was specifically precipitated using streptavidin beads. Western blot analysis of FLAG-septin2 in precipitant or 5% input was shown. B) Relative amount of GTP binding form of septin2 (A) was calculated and graphed. Band intensity was normalized to input. C) HEK293T cells were transfected with FLAG-septin2 mutants, and septin2 proteins were purified by immunoprecipitation. GTP binding assays were performed, and GTP binding septin2 was detected by FLAG antibody. D) Relative amount of GTP binding form of septin2 shown in C. Band intensity was normalized to input. Three different experiments were performed. E) HEK293T cells were cotransfected with FLAG-septin2 and GFP-septin6. Sixteen hours before cell lysates were prepared, 40 µM adox was added to the medium. FLAG-septin2 was continuously immunoprecipitated by FLAG antibody, and coprecipitated proteins were detected by Western blot. F) Binding affinity of GFP-septin6 and endogenous septin7 with FLAG-septin2 (E) was graphed. G) FLAG-septin2 mutants in HEK293T cells were immunoprecipitated with FLAG antibody and coprecipitated GFP-septin6 or endogenous septin7 was detected. H) Binding affinity of GFP-septin6 and endogenous septin7 with FLAG-septin2 (G) was graphed. *P < 0.05, **P < 0.01, ***P < 0.001 (Student’s t test). Error bars ± se.

We next investigated whether methylation of septin2 affects its ability to assemble into septin2-septin6- septin7 hetero-oligomers. We observed that adox treatment significantly increased the binding affinity of septin2 with endogenous septin7. On the other hand, adox decreased the coprecipitation of GFP-septin6 (Fig. 6E, F). K183R also showed a low binding affinity with septin6 and high binding affinity with septin7, which were similar to the septin2 results obtained after treatment with adox, so that the K183R mutation appears to be capable of working as a nonmethylated protein (Fig. 6G, H). The K183Q mutation interacts strongly with both septin6 and septin7. These results strongly suggest that methylation at K183 of septin2 promotes the septin2-6-7 complex formation, while regulating the GTP binding affinity of septin2.

Demethylation at Lys 183 of septin2 promotes septin2 aggregation and suppresses ciliogenesis

We investigated whether methylation regulates the intracellular localization of septin2 in NIH3T3 cells. When these cells were exposed to a 20 µM adox treatment, we observed a significant aggregation of septin2 in the NIH3T3 cells (Fig. 7A, B). As an important control, we noted that GFP itself never aggregated in the presence of adox. In addition, only a small percentage of cells expressing the K183Q mutation exhibited aggregated septin2, whereas nearly 40% of the cells expressing the K183R mutant had aggregated septin2 (Fig. 7C, D). Furthermore, adox treatment did not promote the aggregation of K183Q mutants (Supplemental Fig. 2). Results from the K183A mutants were similar to those in wild-type cells, indicating that the methylation of septin2 at K183 inhibits aggregation, and that the K183R mutant acts as a nonmethylated septin2. Overexpression of K183R mutant significantly decreased the number of cilia and also Gli3 proteolysis, when compared with the K183Q mutant, suggesting that the K183R functions as a dominant–negative mutation of septin2, which inactivates primary cilia and the Shh signaling pathway (Fig. 7E–G). Based on our novel findings, we propose that one mechanism by which folic acid influences NTC is the methylation of septin2, which plays important roles in the regulation of both cilium formation and the Shh signaling pathway. These findings suggest that methylation status of septin2 is altered in patients who have slc19a1 mutations, thereby increasing their susceptibility to NTDs and ciliopathies (Fig. 8).

Figure 7.

Methylation at Lys 183 of septin2 prevents aggregation thereby promoting ciliogenesis. A) Cell localization of GFP or GFP-septin2. NIH3T3 cells were transfected with GFP or GFP-septin2. Two days after transfection, cells were cultured with serum-starved medium containing 20 µM adox for 16 h. DMSO was used as the control. Primary cilia were detected with anti-acetylated tubulin (acetyl tub, red) antibody. Nuclei were detected with DAPI (blue). Arrows: primary cilia; arrowheads: aggregated septin2. B) Number of aggregated GFP-septin2 shown in A were counted and graphed. C) Cell localization of GFP-septin2 mutants. Primary cilia and nuclei were detected with anti-acetylated tubulin antibody (red) or DAPI (blue), respectively. Arrows: primary cilia; arrowheads: aggregated septin2. D) Aggregated septin2 (C) was counted and its aggregation rate was calculated. Three different experiments were performed. E) Number of ciliated cells (C) were counted and graphed. Three different experiments were performed. F) HEK293T cells were transfected with FLAG-septin2. Gli3, FLAG, and β-tubulin in cell lysate were detected by Western blot. G) Relative amount of truncated Gli3 (F) was calculated by normalizing to Gli3 full-length and graphed. *P < 0.05, **P < 0.01 (Student’s t test). Error bars ± se.

Figure 8.

Folate-methylation pathway for ciliogenesis during embryogenesis. Folate signaling contributes to methylation of septin2, which regulates formation of the septin2-6-7 complex and ciliogenesis.

DISCUSSION

Neural tube defect stands out as a preventable birth defect, yet it remains a persistent public health problem, even nearly 20 yr after mandatory folic acid fortification of the U. S. food supply. Research spanning decades, including randomized and community-based clinical trials, demonstrates that maternal, periconceptional supplementation with folic acid can reduce the risk of NTD in offspring (36). The maintenance of intracellular folate homeostasis is critical to essential 1-carbon metabolism, which is critically involved in numerous cellular reactions. These include biosynthesis of purines and thymidylate (dTMP), which are essential for DNA and RNA synthesis, and production of the methyl donor S-adenosyl-methionine, which is essential in the methylation of DNA, histones, proteins, and lipids. These metabolic processes play important roles during development, strongly influencing cell differentiation (37, 38) and cell migration (39). Mammalian cells harvest folate from circulating blood via folate receptors (FRs), also known as folate binding proteins (Folrs) in the mouse, via the reduced folate carrier (rfc1, slc19a1) and the proton coupled folate transporter (Pcft). Folrs are membrane-bound GPC receptors that take up folic acid with relatively high affinity (K_m = ∼1 nM) into the cytosol by an endocytotic process (40, 41). The murine Folr1 is the homolog of human FR-α, which is expressed primarily in epithelial cells of the choroid plexus, kidney, ovaries, and placenta and a limited number of other organs and tissues. Inactivation of Folr1 in the mouse led to embryonic lethality by E10 and the embryos expressed a high prevalence of NTD and other congenital malformations (10). In addition to having a metabolic impact on embryogenesis, folate homeostasis is known to be critical for epigenetic modifications. These are particularly dynamic morphogenetic processes, with extensive reprogramming of DNA and chromatin methylation occurring during early embryogenesis (42), and defective histone methylation has been implicated in several NTD mouse models (15, 43, 44). Thus, one of the key challenges in demystifying the complex etiology of NTD involves understanding the contribution of folic acid and 1-carbon metabolism in the normal and pathologic development of the neural tube, an question that remains poorly understood despite the early success of folate supplementation (36).

Folate has been widely thought to act predominantly at the level of gene expression, because methylation of both DNA and histones is tightly regulated during development. It has been shown that a folate deficiency increases both mutation rates and generalized genome instability. Thus, it may be that folic acid protects developing embryos by reducing the rate of spontaneous detrimental mutations. MacFarlane and coworkers (45) recently demonstrated that mice maintained on a low-folate diet resulted in increased mutagenesis at the Pig-a locus, and expressed elevated micronuclei numbers in their red blood cells. The underlying problems caused by folate insufficiency could also be its impact on de novo thymidylate synthesis in the nucleus. Using an NTD mouse model system in which the scaffolding enzyme serine hydroxymethyltransferase 1, coded for by the Shmt1 gene, has been inactivated, Martiniova and colleagues (46) evaluated the impact of maternal supplementation with the pyrimidine nucleosides uridine, thymidine, or deoxyuridine, in the presence of either high or low maternal folate on NTD response frequencies. They observed that maternal deoxyuridine supplementation prevented NTDs among the progeny of Shmt1 heterozygous dams maintained on a reduced folate diet, whereas maternal uridine supplementation increased the rate of NTD, irrespective of maternal folate status or embryonic genotype. These studies underscore the importance of 1-carbon metabolism in promoting DNA synthesis to maintain essential cellular proliferation during critical periods of neural development.

Although methylation of proteins plays a key regulatory role in many biologic systems, links between folate and protein methylation have not been extensively studied. Our data therefore suggests a role for folate in the methylation of important cytoskeletal proteins, as we have shown that folate-dependent methylation of septin2 is necessary for proper septin function, ciliogenesis, and perhaps ultimately for neural tube closure. Septins are crucial regulators of both the actin and microtubule cytoskeletons (33). Our data highlight the importance of septins in ciliogenesis, where they form a diffusion barrier regulating trafficking of ciliary membrane proteins (19, 34, 37, 48). Septins are also essential for convergent extension cell movements (34), which in turn, contribute to NTC (49). Our data highlights the key role that folate plays in controlling the cytoskeletal events regulating neural tube closure.

Glossary

AdoHcy

S-adenosylhomocysteine

Adox

adenosine dialdehyde

dTMP

thymidine monophosphate

E

embryonic day

FBS

fetal bovine serum

Folr1

folate receptor 1

FR

folate receptor

GFP

green fluorescent protein

LC-MS/MS

liquid chromatography-tandem mass spectrometry

MEF

mouse embryonic fibroblasts

meK

methylated lysine

MEM-FA

minimum essential medium-folic acid

MMR solution

Marc’s modified Ringer’s solution

NTC

neural tube closure

NTD

neural tube defects

PCFT

proton coupled folate transporter

RFC1

reduced folate carrier1

SAM

S-adenosyl methionine

Shh

sonic hedgehog

AUTHOR CONTRIBUTIONS

Ma. Toriyama and R. H. Finnell conceived and designed the experiments and wrote the manuscript; Ma. Toriyama conducted the mouse experiments; Mi. Toriyama conducted the Xenopus experiments; and J. B. Wallingford contributed to the writing and editing of the manuscript.