The Folate Metabolic Enzyme ALDH1L1 Is Restricted to the Midline of the Early CNS, Suggesting a Role in Human Neural Tube Defects

THE JOURNAL OF COMPARATIVE NEUROLOGY 500:368–383 (2007)

† TODD E. ANTHONY AND NATHANIEL HEINTZ* Laboratory of Molecular Biology, Howard Hughes Medical Institute, Rockefeller University, New York, New York

ABSTRACT Folate supplementation prevents up to 70% of human neural tube defects (NTDs), although the precise cellular and metabolic sites of action remain undefined. One possibility is that folate modulates the function of metabolic enzymes expressed in cellular populations involved in neural tube closure. Here we show that the folate metabolic enzyme ALDH1L1 is cell-specifically expressed in PAX3-negative radial glia at the midline of the neural tube during early murine embryogenesis. Midline restriction is not a general property of this branch of folate metabolism, as MTHFD1 displays broad and apparently ubiquitous expression throughout the neural tube. Consistent with previous work showing antiproliferative effects in vitro, ALDH1L1 upregulation during central nervous system (CNS) development correlates with reduced proliferation and most midline ALDH1L1ϩ cells are quiescent. These data provide the first evidence for localized differences in folate metabolism within the early neural tube and suggest that folate might modulate proliferation via effects on midline Aldh1l1ϩ cells. To begin addressing its role in neurulation, we analyzed a microdeletion mouse strain lacking Aldh1l1 and observed neither increased failure of neural tube closure nor detectable proliferation defects. Although these results indicate that loss-of-function Aldh1l1 mutations do not impair these processes in mice, the specific midline expression of ALDH1L1 and its ability to dominantly suppress proliferation in a folate responsive manner may suggest that mutations contributing to disease are gain-of- function, rather than loss-of-function. Moreover, a role for loss-of-function mutations in human NTDs remains possible, as Mthfr null mice do not develop NTDs even though MTHFR mutations increase human NTD risk. J. Comp. Neurol. 500:368–383, 2007. © 2006 Wiley-Liss, Inc.

Indexing terms: radial glial cell; folic acid; glia; BAC transgenic; Chromosome 6; CNS injury; reactive astrocyte; NEUT2; Pax3

Neural tube defects (NTDs) are among the most com- fects of folate metabolism. This idea is supported by stud- mon human birth defects, occurring in ϳ1 in 1,000 preg- ies using the mouse mutant splotch (Sp2H), which has a nancies in the United States (Cragan et al., 1995) and an estimated 300,000 or more newborns worldwide each year (Shibuya and Murray, 1998). NTDs are polygenic, with no Grant sponsor: Howard Hughes Medical Institute (to N.H.); Grant spon- single dominant or recessive causative gene yet identified sor: National Institutes of Health Genetics Predoctoral Training Grants; (Juriloff and Harris, 2000); accordingly, the risk of recur- Grant numbers: GM0848504, GM0848505, GM0798217S1, GM0798218S1 ϳ (to T.E.A.); Grant sponsor: National Institutes of Health Molecular Biology rence in sibships is 2–5% (McBride, 1979; Campbell et Predoctoral Training Grant; Grant number: GM0798219 (to T.E.A.). al., 1986). Although folate taken during the periconcep- *Correspondence to: N. Heintz, Laboratory of Molecular Biology, tual period has been shown to prevent the recurrence of up Howard Hughes Medical Institute, Rockefeller University, 1230 York Ave., to 70% of human NTDs (Smithells et al., 1981; Wald et al., New York, NY. E-mail: [email protected] 1991; Czeizel and Dudas, 1992), the mechanisms under- †Current address: California Institute of Technology, Div. of Biology MC lying these protective effects have remained unclear. 216-76, 1200 E. California Blvd., Pasadena, CA 91125 Received 8 March 2006; Revised 5 July 2006; Accepted 25 August 2006 As folate deficiency alone is insufficient to cause NTDs DOI 10.1002/cne.21179 in mice (Heid et al., 1992), one possible mode of action is Published online in Wiley InterScience (www.interscience.wiley. that folate supplementation acts to correct inherited de- com).

NEURAL TUBE MIDLINE RESTRICTION 369 deletion within the paired homeodomain of the transcrip- during neural tube closure has not been explored, we tion factor Pax3 (Epstein et al., 1991). Homozygous Sp2H investigated ALDH1L1 expression and function further. embryos develop exencephaly and/or spina bifida (Epstein et al., 1991; Fleming and Copp, 1998), and ϳ40% of these can be prevented when folate is delivered in utero (Flem- MATERIALS AND METHODS ing and Copp, 1998). Moreover, analysis of cultured em- Mice and BAC transgenesis bryos demonstrated that folate cycling is compromised in Sp2H/Sp2H mice but can be restored to levels indistin- Mice were maintained and utilized according to proto- guishable from wildtype mice when exogenous folic acid is cols approved by the Institutional Animal Care and Use added (Fleming and Copp, 1998). The normalization of Committee at Rockefeller University. Wildtype C57B6/J both neural tube closure and folate cycling suggests that mice were used for expression analysis. NEUT2 mice were folate might exert its protective effects in splotch mice by previously generated as described (Champion et al., 1994) supplementing impaired metabolic pathways. Although and were obtained from Charles River Laboratories (Wil- this mechanism may explain NTD prevention in some mington, MA) with the permission of Dr. Carol Giometti. cases, there is evidence that folate does more than just NEUT2 mice were bred to C57B6/J for three generations correct biochemical deficiencies. In particular, mice lack- and then the F3s intercrossed for maintenance. To restore ing the Cited2 gene develop exencephaly and die at late Aldh1l1 expression onto a NEUT2 background, the BAC gestation due to failed neural tube closure in the midbrain clone RP23-7M9 (see text and Fig, 5) was injected into Ϫ Ϫ ϫ and hindbrain (Bamforth et al., 2001; Barbera et al., oocytes generated from a NEUT2 / B6CBA F1 cross using standard transgenic techniques (Hogan et al., 1994). 2002); despite normal folate cycling in these mutants, the Positive founders were then bred to NEUT2 homozygous NTDs can be prevented by exogenous folate (Barbera et mice to generate BAC transgene positive NEUT2 homozy- al., 2002). These data suggest that, in addition to rescuing gotes; these are denoted Ϫ/Ϫ;TgN(RP23-7M9). All quan- impaired metabolism, folate supplementation could also titative analysis was done using mice from these crosses, act by modulating folate-dependent reactions in popula- and therefore have a mixed genetic background of C57B6, tions of cells that are involved in neural tube closure CBA, and BALB/C. (Barbera et al., 2002; Copp et al., 2003). Currently, it is not known upon which cellular population(s) folate acts, Immunostaining which reactions might be modulated, or how such effects can prevent NTDs. Tissues were either immersion-fixed (embryos) or perfused (postnatal/adult) with 4% paraformaldehyde (PFA) One approach toward gaining insight into these issues and postfixed overnight at 4°C. Tissues were then washed is to study the expression and function of folate metabolic in phosphate-buffered saline (PBS, pH 7.4) and cut either enzymes present in the developing nervous system. We on a cryostat (20 ␮m) or vibratome (75 ␮m); tissues to be previously found that the expression of aldehyde dehydro- cryosectioned were first cryoprotected in 15% sucrose genase 1 family, member L1 (Aldh1l1) is developmentally overnight at 4°C. Details on source and specificity of the regulated in the cerebellum, with high mRNA levels post- antibodies used are as follows: Rabbit ␣-ALDH1L1 was natally and decreased levels in the adult (Kuhar et al., used at 1:1,000 and was a gift of Dr. Robert Cook. This 1993); this observation suggested that Aldh1l1 might ful- antiserum was generated against full-length recombinant fill a unique developmental role. ALDH1L1 (EC 1.5.1.6; rat ALDH1L1; specificity was confirmed by a complete a.k.a. 10-formyltetrahydrofolate dehydrogenase, FDH, lack of staining in NEUT2 homozygotes (Aldh1l1 null FTHFD) is a cytosolic folate binding protein that irrevers- mice, Fig. 5). Rat ␣-BrdU (1:100, Oxford Biotechnology, ibly converts 10-formyltetrahydrofolate (10-FTHF) to tet- Kidlington, UK) was generated against bromodeoxyuri- rahydrofolate (THF) and CO2 (Kutzbach and Stokstad, dine (BrdU) coupled to keyhole limpet hemocyanin (KLH), 1971; Cook, 2001a), and which may also be responsible for and yielded no staining in mice not administered BrdU. regulating cellular THF levels (Kim et al., 1996). As the Rabbit ␣-GFAP (1:500, Dako, Carpinteria, CA) was gen- reaction catalyzed by ALDH1L1 results in permanent re- erated against glial fibrillary acidic protein (GFAP) iso- moval of one-carbon units from the cellular pool, it has lated from bovine spinal cord; mouse ␣-GFAP (1:200, been suggested that ALDH1L1 may function primarily as Sternberger Monoclonals, Baltimore, MD) was generated a regulatory molecule (Krebs et al., 1976; Scrutton and against human GFAP. Both of these antibodies labeled Beis, 1979). Moreover, as 10-FTHF is required during two only cells with the morphology of fibrillary astrocytes. steps of purine synthesis, it has been proposed that Rabbit ␣-Ki67 (1:500, Novocastra, Newcastle upon Tyne, ALDH1L1 might restrict proliferation. This idea is sup- UK) was generated against a 1,086-bp Ki67 motif- ported by the findings that overexpression of ALDH1L1 containing cDNA fragment, gave the expected nuclear suppresses proliferation of certain cell lines, and that staining pattern and was restricted to areas of active ALDH1L1 is significantly downregulated in several differ- proliferation. Rabbit ␣-MTHFD1 (1:2,000, gift of Dr. Dean ent tumor tissues (Krupenko and Oleinik, 2002; Oleinik Appling) was generated against full-length enzyme puri- and Krupenko, 2003). However, the precise physiological fied from rat liver; Western analysis of crude liver extracts role played by ALDH1L1 in vivo is unknown. demonstrated that the antiserum labeled a single band of In the course of analyzing the embryonic expression ϳ116 kDa that comigrated with purified MTHFD1 (Cheek pattern of ALDH1L1, we observed its specific localization and Appling, 1989). Mouse ␣-Nestin/rat-401 (1:4, Devel- in radial glia at the ventral and dorsal midlines of the E12 opmental Studies Hybridoma Bank, Iowa City, IA) was spinal cord. As no other enzyme directly involved in folate generated against homogenized Sprague-Dawley rat spi- metabolism has previously been shown to have such re- nal cord (Hockfield and McKay, 1985) and gave character- stricted expression within the developing neural tube, and istic filamentous staining of cells in germinal zones. Mono- as a role for localized differences of folate metabolism clonal mouse ␣-PAX3 (1:2,000 and TSA-amplified, The Journal of Comparative Neurology. DOI 10.1002/cne

370 T.E. ANTHONY AND N. HEINTZ

Developmental Studies Hybridoma Bank) was generated TABLE 1. List of Genes Deleted in NEUT2 Homozygous Mice against amino acids 298–481 of quail PAX3. This anti- 1. Vomeronasal 1 receptor, B7 body has been used previously used to immunolabel PAX3 2. Vomeronasal 1 receptor, B9 3. Vomeronasal 1 receptor, pseudogene 2 (Baker et al., 1999; Venters et al., 2004), and in our hands 4. Vomeronasal 1 receptor, pseudogene 3 yielded staining in the murine spinal cord that was nuclear 5. Similar to guanylate binding protein 5a 6. Vomeronasal 1 receptor, A6 and in a pattern essentially identical to that reported for 7. Vomeronasal 1 receptor, A5 Pax3 mRNA and protein (Li et al., 1999; Mansouri et al., 8. Vomeronasal 1 receptor, B4 2001; Milewski et al., 2004). Rabbit ␣-phosphorylated his- 9. Vomeronasal 1 receptor, pseudogene 4 10. Vomeronasal 1 receptor, A2 tone H3 (Ser10) (PH3, 1:250, Upstate Biotechnology, Lake 11. Vomeronasal 1 receptor, B8 Placid, NY) was generated against a peptide (ARKpSTG- 12. Vomeronasal 1 receptor, A4 13. Vomeronasal 1 receptor, A3 GKAPRKQLC) corresponding to amino acids 7–20 of 14. Vomeronasal 1 receptor, B2 human histone H3, and gave completely nuclear stain- 15. Vomeronasal 1 receptor, B1 16. Vomeronasal 1 receptor, A1 ing present only in actively proliferating regions. All 17. Vomeronasal 1 receptor, A7 secondary antibodies (Cy2 and Cy3) were from Jackson 18. Vomeronasal 1 receptor, A8 Immunoresearch (West Grove, PA) and used at 1:500. 19. Vomeronasal 1 receptor, B3 20. Vomeronasal 1 receptor, pseudogene 5 Sections were preblocked in 5% donkey serum, 0.1% 21. Vomeronasal 1 receptor, A9 Triton X-100 in PBS, and incubated with primary anti- 22. cDNA sequence BC048671 23. Similar to carbohydrate (chondroitin 4) sulfotransferase 13 bodies overnight at 4°C. After washing, sections were 24. Urocanase domain containing 1 incubated with secondary antibodies generated in don- 25. cDNA sequence BC003332 26. RIKEN cDNA C230069K22 keys for 2 hours in the same solution at room tempera- 27. Kruppel-like factor 15 ture. BrdU immunostaining was done as follows: sec- 28. *Aldehyde dehydrogenase 1 family, member L1 (Aldh1l1) tions were incubated in 0.3% H2O2 in methanol for 15 minutes, washed in PBS, treated in 1 mg/ml sodium borohydride in PBS for 20 minutes, washed again, de- natured in 2 N HCl for 40 minutes, then washed exten- sively in PBS before being used for immunostaining. some; generation of product indicates that the sample is Sections to be used for PAX3 immunostaining were either wildtype or heterozygous: 1_forward 5Ј-CCACC- microwaved at full power for 10 minutes in 10 mM TATAGGGTTGCAG-3 and 1_reverse 5Ј-GGATGGACCA- sodium citrate, pH 6.0, prior to application of primary TCTAGAGAC-3Ј, 241 bp product; 2_forward 5Ј-ACAGC- antibody. Confocal imaging was done on an LSM 510 ACCCTGGAACACAGAG-3Ј and 2_reverse 5Ј-TTGTA- Axioplan (Zeiss, Thornwood, NY) in the Rockefeller Uni- TGTGGCACAGTGCGTC-3Ј, 252 bp product. Both of these versity Bioimaging Facility, and full resolution images primer sets amplify a deleted region of NEUT2 mice that were exported directly into Adobe Photoshop (San Jose, is not present in BAC RP23-7M9; thus, they can be also be CA) for figure construction. used for typing Ϫ/Ϫ;TgN(RP23-7M9) mice. The second Mapping, complementation, and genotyping two primer sets amplify across the deletion breakpoint; the NEUT2 microdeletion generation of product indicates that the sample is either heterozygous or homozygous: 3_forward 5Ј-CTGCACT- Since the full-length Aldh1l1 cDNA hybridizes to ATAGACCCAAGC-3Ј and 3 _reverse 5Ј-TGAGTGTTG- NEUT2 homozygous genomic DNA (Champion et al., GGATCTCTGC-3Ј, 237 bp product; 4_forward 5Ј-CCCA- 1994), one end of the deletion must be within the Aldh1l1 ACGCATGTCAGAGG-3Ј and 4_reverse 5Ј-CCCACG- gene. PCR against the first and last exons demonstrated TCTGCATGATCC-3Ј, 295 bp product. BAC transgenic that the deletion was in the 5Ј part of the gene. By design- mice were identified by PCR against the backbone of the ing PCRs against the most 3Ј end of the gene and walking RP23 library vector, pBACe3.6 using the following two progressively 5Ј, the 3Ј breakpoint could be mapped to Ј within 200 bp, in the intron between exons 10 and 11. primers: e3.6 for 5 -CAAGGTGCTGATGCCGCTGGCG- Next, NEUT2 Ϫ/Ϫ DNA was digested separately with ATTCAGG-3Ј and e3.6 rev 5Ј-AGAGCAATATAGTCCT- three different restriction enzymes and nested ligation ACAATGTCAAGCTC-3Ј. mediated PCR was performed using the following primers located in the 3Ј breakpoint region: 5Ј-AAGCCTGTAG- Cortical injury model GGAGACAGTGG-3Ј and 5Ј-CCCAACGCATGTCAGAG- Injury experiments were done using the method previ- GTG-3Ј as the 3Ј primers and oligos provided in the Vec- ously described (Koguchi et al., 2002). Briefly, adult mice Ј torette II Kit (Sigma Genosys, St. Louis, MO) as the 5 were deeply anesthetized with avertin and a small hole primers. Products were sequenced from all three reactions was drilled in the skull over the cortex using a 21G needle. and each yielded the same result of an 841,250 bp dele- Cortical wounds were inflicted by inserting a 27G needle tion. The 5Ј breakpoint is: 5Ј-TAACCCAACTAATGCT- GCTTATAACT-3Ј_deletion. The 3Ј breakpoint is: into the cortex to a depth of 2 mm, and animals then deletion_5Ј-GAGTTTCCTAGAACCCAGTCAAGAG-3Ј. This allowed to recover on a heating pad. Proliferating astro- deletion removes all but the last two exons of the Aldh1l1 cytes were marked 3 days after injury by administration of gene, the entire 5Ј flanking genomic sequence (putative pro- a single pulse of BrdU at 0.5 mg/10 grams body weight to moter region), as well as 27 other upstream genes (see text label cells in S-phase. Two and a half hours after receiving and Table 1). Neither Aldh1l1 mRNA (Champion et al., BrdU, mice were deeply anesthetized with nembutal at 60 1994) nor protein (Fig. 5) is detectable in NEUT2 Ϫ/Ϫ mice. mg/kg, followed by transcardial perfusion with 4% PFA. Once the deletion breakpoints were determined, mice The percentage of GFAP-positive astrocytes that were were typed by PCR with four sets of primers. The first two also-BrdU positive (labeling index for astrocytes) was then primer sets are against portions of the deleted chromo- determined by immunostaining as described above. The Journal of Comparative Neurology. DOI 10.1002/cne

NEURAL TUBE MIDLINE RESTRICTION 371

Fig. 1. ALDH1L1 is specifically expressed in radial glia at the identity confirmed by double labeling with nestin; F,G show higher midline of the developing neural tube. Immunofluorescent staining in magnification of the center of panel E. Note that expression of nestin the embryonic CNS for ALDH1L1 (red in A–F,H–L,O) and nestin (which is induced earlier than ALDH1L1 in the CNS) appears to be (green in G,H). Regions shown are the E9.5 hindbrain (A,B), E10.5 higher in the ALDH1L1Ϫ radial glia in the dorsal spinal cord, sug- hindbrain (C,D), E12.5 cervical spinal cord (E–H), E14.5 cervical gesting that these cells may be developmentally less mature. Upregu- spinal cord (I), tectum (J), hypothalamus (K), and the E16.5 hypothal- lation of ALDH1L1 in more dorsal regions of the spinal cord occurs amus (L). Analysis of the onset of ALDH1L1 expression in the devel- over the next 2 days, until all radial glia are positive by E14.5 (I). This oping nervous system shows that it is initially upregulated at the same developmental progression of expression was also observed in ventral midline (A–D), but by E12.5 is detected also at the dorsal other CNS regions. At E14.5, ALDH1L1 is largely restricted to the midline (arrow in E). The arrowhead in E denotes the boundary midline of the tectum (J) and hypothalamus, but is expressed separating the dorsal and ventral portions of the spinal cord, shown at throughout these structures by E16.5; staining for the hypothalamus higher magnification in F. At E12.5, no ALDH1L1ϩ cells are detect- is shown in L. Scale bar ϭ 100 ␮m for A,C,E,I–L; 35 ␮m for D,F–H; 25 able above this point other than those at the dorsal midline. ␮m for B. ALDH1L1ϩ cells have the distinctive morphology of radial glia, an

RESULTS hindbrain and cervical spinal cord (Fig. 1A,B); the ventral ALDH1L1 is restricted to midline radial glia midline in this region of the neural tube is also known as in the early neural tube and is the median hinge point (MHP), which is one of two major sites where the neural plate bends during neurulation developmentally regulated in the central (Copp et al., 2003). Expression was further upregulated in nervous system (CNS) these ventral midline cells by E10.5 and could be detected To determine the timing and location of ALDH1L1 onset using standard immunofluorescent staining (Fig. 1C,D). in the neural tube, we immunostained cryosections from Progressive dorsal expansion of staining was observed embryonic day (E)8.5 to E12.5. Using TSA-amplified im- over the next few days in adjacent ventral regions, and by munofluorescence, ALDH1L1 was not detected at E8.5 E12.5 cells flanking the dorsal midline (the site of neural (prior to neural tube closure) but was present by E9.5 in fold fusion; Copp et al., 2003) were also labeled (arrow in cells flanking the ventral midline of the neural tube in the Fig. 1E). Interestingly, all ALDH1L1ϩ cells have the dis- The Journal of Comparative Neurology. DOI 10.1002/cne

372 T.E. ANTHONY AND N. HEINTZ tinctive morphology of radial glia, cells that play both Given the restricted expression of ALDH1L1 at the mid- structural and precursor roles in the developing nervous line and association of an MTHFD1 allele with human system (Fishell and Kriegstein, 2003; Rakic, 2003; Ever NTDs, we compared their expression patterns to deter- and Gaiano, 2005). To confirm their identity, we double- mine how they might interact to regulate folate metabo- labeled with nestin, a gene expressed in all radial glia but lism and neural tube closure. In contrast to ALDH1L1, absent from neurons; as expected, all ALDH1L1ϩ cells MTHFD1 does not show regionally restricted expression contain high levels of nestin, demonstrating cell-specific in the embryonic CNS (Fig. 2). Broad Mthfd staining is expression (Fig. 1F–H). ALDH1L1 expression continues to observed in neuroepithelial cells at all rostrocaudal levels be upregulated in more dorsally situated cells, and by of the CNS, both prior to (E8.5, Fig. 2A) and after (E9.5, E14.5 essentially all spinal cord radial glia at all rostro- Fig. 2B) neural tube closure. Moreover, MTHFD1 is not caudal levels of the CNS are ALDH1L1ϩ (Fig. 1I). cell-specifically expressed; examination after the start of Similar spatiotemporal gradients of staining were ob- neurogenesis demonstrates that MTHFD1 is present in served in other brain regions. For example, high levels of both neurons (Fig. 2D–G,L–N) and radial glia (Fig. 2D,H– ALDH1L1 protein are detected in midline radial glia of K). One additional difference between ALDH1L1 and the E14.5 tectum and hypothalamus (Fig. 1J,K). As in the MTHFD that was particularly obvious in neocortex at spinal cord, ALDH1L1 becomes induced in lateral regions later stages of embryogenesis is their subcellular distri- over the next couple of days, and extensive expression in bution. ALDH1L1 protein is detectable in cell bodies and these areas is seen by E16.5 (Fig. 1L and data not shown). throughout radial glial processes (Fig. 2O,Q), whereas Given that broad and high levels of ALDH1L1 are not seen MTHFD1 is largely restricted to cell bodies in both neu- until relatively late stages of embryogenesis, these data rons and radial glia (Fig. 2P,R). This raises the interesting suggest that ALDH1L1ϩ radial glia might be developmen- possibility that 10-FTHF concentrations may vary within tally more mature than ALDH1L1Ϫ cells, and that radial different subcellular compartments of radial glia express- glia at the ventral and dorsal midlines may differentiate ing both enzymes. earlier than intervening populations. ALDH1L1 upregulation correlates with Given its role in folate metabolism, the above-described restricted expression near sites involved in neurulation decreased proliferation suggests that ALDH1L1 might play a role in neural tube The above expression data indicate that MTHFD1 ac- closure. Regarding the timing of ALDH1L1 onset, it tivity fulfills general metabolic requirements for 10-FTHF should be noted that two different mouse mutants have throughout the embryonic CNS, while ALDH1L1 modu- increased frequencies of NTDs after initial neural tube lates 10-FTHF-dependent processes in a spatiotemporally closure. In Tc/tw5 compound mutants, neural folds in the dynamic manner. Two lines of evidence suggest that pro- caudal CNS fuse but then subsequently rupture and reopen liferation may be one of the processes affected by (Park et al., 1989). Furthermore, a substantially larger per- ALDH1L1 activity. First, ALDH1L1 overexpression has centage of Cited2 null embryos display obvious exencephaly been shown to repress proliferation in vitro (Krupenko at E10.5 (ϳ80% of Ϫ/Ϫ embryos) than at E9.5 (ϳ50%) (Bar- and Oleinik, 2002; Oleinik and Krupenko, 2003). Second, bera et al., 2002). These examples demonstrate that success- radial glial maturation is accompanied by an increase in ful neural tube closure requires processes that are active cell cycle length and a decrease in the percentage of mi- both before and after neural fold fusion. Accordingly, the totic cells (Takahashi et al., 1995; Noctor et al., 2002), and onset of ALDH1L1 expression after neural fold fusion does ALDH1L1 becomes broadly expressed at relatively late not rule out a potential role in neurulation. stages of radial glial development (Fig. 1). Therefore, we MTHFD1 is broadly expressed throughout next examined the proliferative status of regions that have upregulated ALDH1L1. To do this, adjacent sections the developing CNS of different CNS regions were immunostained for ALDH1L1 is the first folate metabolic enzyme shown to ALDH1L1 and either Ki67 (expressed by proliferating display restricted expression at sites known to be involved cells during all stages of the cell cycle) or PH3 (a marker in neurulation. Although correlative, potential involve- for cells in mitosis). ment of ALDH1L1 in neural tube closure is also suggested In the E12.5 spinal cord, ALDH1L1 is restricted to by the fact that a single nucleotide polymorphism (R653Q) radial glia flanking the ventral and dorsal midlines (Fig. in the methylenetetrahydrofolate dehydrogenase 1 gene 3A). Strikingly, these regions are largely devoid of prolif- (MTHFD1) has been shown to associate with increased erating cells: a gradient of Ki67ϩ staining in the spinal risk of NTDs and decreased embryo survival in humans cord is observed, with virtually no cycling cells at the most (Brody et al., 2002). MTHFD1 is a trifunctional enzyme ventral portion of the spinal cord, increasing numbers in that possesses two activities that generate 10-FTHF: 1) more dorsally situated cells, but a sharp decrease again at 10-formyl-THF synthetase (EC 6.3.4.3), which converts the most dorsal point (Fig. 3A’). This relative lack of di- formate and THF to 10-FTHF; and 2) 5,10-methenylTHF viding cells at the dorsal midline is also clearly observed in cyclohydrolase (EC 3.5.4.9), which converts 5,10- the midbrain at the tectal midline; whereas ALDH1L1 is methenylTHF to 10-FTHF (Cheek and Appling, 1989; expressed specifically in midline radial glia (Fig. 3B), es- Cook, 2001a). ALDH1L1 and MTHFD1 therefore have sentially all proliferating cells lie outside this region (Fig. opposing effects on cellular levels of the same folate deriv- 3B’). To determine how proliferation changes as ative and likely affect a similar set of metabolic reactions. ALDH1L1 is upregulated, we compared the number of Although the functional consequences of the R653Q poly- mitotic cells in the caudal ganglionic eminence (CGE) at morphism are unknown, one could predict that reduced E12.5 (prior to ALDH1L1 expression) and E14.5 (after MTHFD1 activity (loss-of-function) might have similar ALDH1L1 upregulation). As shown in Figure 3C,D, up- phenotypic effects as increased ALDH1L1 activity (gain- regulation of ALDH1L1 correlates with a sharp decrease of-function), and vice versa. in the number of mitotic cells in the ventricular zone, the The Journal of Comparative Neurology. DOI 10.1002/cne

NEURAL TUBE MIDLINE RESTRICTION 373

Fig. 2. MTHFD1 is broadly expressed in both radial glia and (K,P); double-labeling with ␤-III Tubulin (E–G,L–N) and nestin (H–J) neurons throughout the embryonic CNS. Immunofluorescent staining confirms this. This contrasts sharply with ALDH1L1 expression, for MTHFD1 (A–E,G,H,J–L,N,P,R), ␤-III Tubulin (F,G,M,N), Nestin which is confined to radial glial cells (O,Q and Fig. 1). The subcellular (I,J), and ALDH1L1 (O,Q). High-power double-labeling is shown for distributions of ALDH1L1 and MTHFD1 are also distinct, and were regions in D with star (E–G) and box (H–J). Broad and apparently particularly noticeable during late stages of cortical development ubiquitous expression of MTHFD1 is observed prior to neural tube (E17.5 is shown in O–R). Whereas ALDH1L1 is present throughout closure at E8.5 (A) and is similar at E9.5 (B), a timepoint when radial glial cell bodies and processes (O,Q), MTHFD1 is contained ALDH1L1 (in contrast) displays highly localized expression in ventral only in neuronal and radial glial cell bodies (P,R). SVZ, subventricular midline radial glia (Fig. 1A,B). Arrowhead points to region shown at zone (location of radial glial cell bodies); CP, cortical plate (location of higher power in box at lower right; note the characteristic cytoplasmic neuronal cell bodies). Scale bar ϭ 100 ␮m for C,O,P; 90 ␮m for B; 50 staining. MTHFD1 is present in both neurons and radial glia, as is ␮m for A; 40 ␮m for K,Q,R; 25 ␮m for B inset, D,L–N; 15 ␮m for E–J. evident from its distribution in E12.5 spinal cord (C,D) and neocortex The Journal of Comparative Neurology. DOI 10.1002/cne

374 T.E. ANTHONY AND N. HEINTZ

Fig. 3. ALDH1L1 upregulation correlates with decreased prolifer- zone (VZ), which is the location of radial glial cell bodies. Regions ation in the developing CNS. Immunoﬂuorescent staining of adjacent expressing high levels of ALDH1L1 were largely devoid of proliferat- sections for ALDH1L1 (A–D), Ki67 (A’,B’), and PH3 (C’,D’). Regions ing cells, as shown in A–B’. Moreover, examination of proliferation shown are the E12.5 spinal cord (A,A’) and tectal dorsal midline before (C) and after (D) the onset of ALDH1L1 expression shows a (B,B’), and the caudal ganglionic eminences (CGE) at E12.5 (C,C’) and sharp decrease in the number of mitotic cells following ALDH1L1 E14.5 (D,D’). Photographs in C’ and D’ are high-magniﬁcation images upregulation; compare PH3 staining in the VZ of C’ and D’. Scale of the CGE; the brackets in these panels demarcate the ventricular bar ϭ 100 ␮m for A,A’,C,D; 50 ␮m for B,B’,C’D’.

location of ALDH1L1ϩ radial glial cell bodies. It should be mice (Keller-Peck and Mullen, 1997), suggesting that noted that subpallial expression is not unique to the CGE; Pax3 might positively regulate proliferation in the early ALDH1L1 is highly expressed in all ventral telencephalic neural tube. The relationship between ALDH1L1 and radial glia at E14.5 (e.g., the lateral and medial ganglionic PAX3 is of potential interest, as Sp2H embryos (Pax3 mu- eminences, Fig. 5E,G). Moreover, as observed in the spinal tants) develop NTDs that can be rescued by exogenous cord, expression is progressively upregulated in more dor- folate, and this folate-responsiveness is due at least in sal structures and is detectable in pallial (neocortical) part to disrupted folate metabolism (Fleming and Copp, radial glia by E17.5 (Fig. 2O). 1998). However, the link between PAX3 function and folic These data demonstrate that cell division is minimal or acid metabolism remains unknown. We therefore directly absent in ALDH1L1ϩ midline radial glia, and that up- examined the relationship between the expression of regulation of ALDH1L1 correlates temporally with a de- PAX3 and ALDH1L1 using double immunofluorescence. crease in the percentage of mitotic radial glia. Further- At E11.5, ALDH1L1 was detected only in a small popula- more, the inverse relationship between ALDH1L1 tion of radial glia flanking the floor plate (Fig. 4A), expression and proliferation suggests that a shift in folate whereas PAX3 was highly expressed in progenitors metabolism could be involved in regulating proliferation throughout the dorsal spinal cord (Fig. 4B). At E12.5, in ALDH1L1ϩ midline radial glia. ALDH1L1 expression increased considerably in the ven- ALDH1L1 and PAX3 expression are tral midline radial glia and had begun to spread to more dorsally located radial glia (Fig. 4C). However, a sharp inversely correlated in the developing boundary was observed between the regions containing spinal cord ALDH1L1 immunopositive and negative cells that ap- ALDH1L1 expression is restricted to radial glia in the peared to correlate tightly with PAX3 expression (Fig. ventral portion of the spinal cord from E9.5–12.5 (Fig. 1), 4C,D). Examination of this boundary at higher magnifica- and its inverse correlation with proliferation (Fig. 3) is tion showed that ALDH1L1ϩ cells were essentially PAX3Ϫ consistent with its known antiproliferative activity (Kru- and vice versa (Fig. 4E–G). Moreover, ALDH1L1 expres- penko and Oleinik, 2002; Oleinik and Krupenko, 2003). sion was upregulated in radial glia throughout the dorsal– Interestingly, these characteristics are opposite those of ventral extent of the spinal cord by E14.5 (Fig. 4H); by this the transcription factor PAX3. In the spinal cord, PAX3 age, PAX3 can no longer be detected (Fig. 4I). These data expression is restricted to mitotic dorsal progenitors demonstrate a strong inverse relationship between PAX3 (Goulding et al., 1991). Moreover, significantly fewer cells and ALDH1L1 expression, and raise the possibility that are born in the dorsal spinal cord at E10 in Pax3 mutant PAX3 might negatively regulate Aldh1l1 transcription. The Journal of Comparative Neurology. DOI 10.1002/cne

NEURAL TUBE MIDLINE RESTRICTION 375

Fig. 4. ALDH1L1 and PAX3 expression are inversely correlated in mutually exclusive expression, PAX3 could no longer be detected in the developing spinal cord. Double immunoﬂuorescent staining of the spinal cord at E14.5 (I), a time when ALDH1L1 has been upregu- embryonic spinal cord for ALDH1L1 (red; A,C,E,G,H), PAX3 (green; lated in radial glia throughout the dorsoventral axis (H). GLAST B,D,F,G,I,K,M), and GLAST (red; J,L–N). Ages shown are E11.5 expression at E12.5 is also grossly distinct from PAX3 expression (A,B), E12.5 (C–G,J–M), and E14.5 (H,I,N). At all timepoints exam- (J,K), although some GLASTϩ cells at the dorsoventral border contain ined a strong inverse relationship was observed between ALDH1L1 high levels of PAX3 protein (L,M). Similar to ALDH1L1, GLAST is and PAX3 expression. Moreover, high-magniﬁcation examination of highly expressed in all spinal cord radial glia by E14.5 (N). Scale the dorsoventral border showed that ALDH1L1 was essentially un- bar ϭ 100 ␮m for A–D,J,K; 80 ␮m for H,I,N; 25 ␮m for E–G; 10 ␮m for detectable in PAX3ϩ cells and vice versa (E–G). Consistent with their L,A.

Another radial glial gene previously reported to have re- gain-of-function in the absence of known human or mouse stricted ventral expression in the developing spinal cord is mutations is complicated, as it is difficult to predict what the glutamate transporter, GLAST (Shibata et al., 1997). To relevant alleles might look like. Therefore, we chose to determine if PAX3 might also be absent from GLASTϩ cells, assess the effects of Aldh1l1 loss-of-function by analyzing we double-immunolabeled for these proteins. At E12.5, low- NEUT2 mice; this strain contains a microdeletion in the magnification images showed that GLASTϩ regions are region containing the Aldh1l1 gene. These mice were gen- largely devoid of PAX3ϩ cells and vice versa (Fig. 4J,K). erated by exposing mice to fission spectrum neutrons However, examination of the boundary region at higher and then breeding the mutations to homozygosity. Two- magnification clearly showed that strong PAX3 immunore- dimensional electrophoresis of liver proteins in one of activity was present in some GLASTϩ cells (Fig. 4L,M). the homozygous lines (called NEUT2) demonstrated the These results could suggest that while PAX3 negatively reg- absence of a protein that sequencing established was ulates both ALDH1L1 and GLAST in dorsal radial glia, it ALDH1L1; Northern blotting later confirmed the lack of exerts tighter suppression over the Aldh1l1 locus. Aldh1l1 mRNA (Champion et al., 1994; Giometti et al., Characterization and complementation of 1994). Viable NEUT2 homozygotes could be generated and biochemical studies showed that hepatic folate an Aldh1l1 null mouse strain pools are distorted, with a threefold decrease of THF Either loss- or gain-of-function mutations in Aldh1l1 and similar increase of 10-FTHF levels (Champion et could potentially affect neural tube closure. Modeling al., 1994). In addition, hepatic histidine catabolism via The Journal of Comparative Neurology. DOI 10.1002/cne

376 T.E. ANTHONY AND N. HEINTZ

Fig. 5. Characterization and complementation of NEUT2 homozy- (C,F,I), and NEUT2 Ϫ/Ϫ;TgN(RP23-7M9) (D,G,J) mice. Clone RP23- gous mice. A: A diagram (not to scale) depicting the NEUT2 microde- 7M9 was sufﬁcient to restore ALDH1L1 expression in the wildtype letion on mouse chromosome 6D1. The breakpoints are denoted by pattern to NEUT2 Ϫ/Ϫ mice throughout the CNS; shown here are arrows, and each deleted gene is shown as a shaded circle. Of the 28 radial glia in the E14.5 tectum (B–D) and forebrain (E–G), and adult genes deleted, 21 code for vomeronasal receptors (gray circles); all cerebellar Bergmann glia and internal granular layer (IGL) astroglia genes contained within the deletion are listed in Table 1. BAC clone (H–J). Note the restricted expression in both wildtype and BAC trans- RP23-7M9 is aligned below; note that this clone contains the entire genic mice at the tectal midline (arrows in B,D). Scale bar ϭ 100 ␮m Aldh1l1 gene but none of the other genes deleted in NEUT2 mice. for E–G; 40 ␮m for B–D; 20 ␮m for H–J. B–J: Immunostaining for ALDH1L1 in wildtype (B,E,H), NEUT2 Ϫ/Ϫ The Journal of Comparative Neurology. DOI 10.1002/cne

NEURAL TUBE MIDLINE RESTRICTION 377

Fig. 6. Folate deﬁciency induced malformations resembling NTDs B is representative of the types of defects observed). The embryo in B at low frequency on the NEUT2 background. Whole mount normal developed a malformation resembling rachischisis (failure of neural (A) and malformed (B) E12.5 embryos derived from dams maintained tube closure over a large portion of the spinal cord), but appeared on an amino acid diet with 0.4 mg/kg folate (mutant embryo shown in otherwise normal. Scale bar ϭ 750 ␮m.

folate-dependent deamination is abnormal in NEUT2 previously described distortions in folate pools are most homozygotes (Cook, 2001b). Thus, NEUT2 mice have likely due to the absence of ALDH1L1, making NEUT2 disruptions of folate metabolism, the consequences of mice a potentially useful model for studying the physio- which have not yet been determined in the developing logical effects of altered folate metabolism in tissues ex- CNS. pressing ALDH1L1. As the extent of the chromosomal deletion (and hence In order to identify which phenotypes present in the total number of deleted genes) had not been previously NEUT2 mice are due to loss of Aldh1l1, it was necessary determined, we mapped the deletion breakpoints using to restore its expression onto a NEUT2 homozygous back- ligation-mediated PCR as described in Materials and ground. To accomplish this, BAC transgenic mice were Methods. This analysis demonstrated an 841,250 bp dele- generated using the clone RP23-7M9. Importantly, this tion present on mouse chromosome 6D1 in NEUT2 mice; clone spans only the most telomeric extent of the NEUT2 examination of the deleted region using public databases deletion, containing the entire Aldh1l1 gene and ϳ40 kb of demonstrated the loss of 28 genes (Fig. 5, Table 1). Of 5Ј flanking genomic sequence, but none of the other genes these, 21 code for vomeronasal receptors, genes that are present in the deletion (Fig. 5A). This results only in the expressed in the vomeronasal organ and that appear to be complementation of the Aldh1l1 mutation and rescue of primarily involved in detecting pheromones (Dulac, 2000; any phenotype in NEUT2 mice carrying the BAC trans- Del Punta et al., 2002). Of the remaining seven genes, the gene can be attributed to restoration of Aldh1l1. As shown only deleted gene other than Aldh1l1 that has any known in Figure 5B–J, RP23-7M9 is sufficient to drive ALDH1L1 effect on folate metabolism is urocanase domain contain- expression in NEUT2 homozygotes in the same pattern as ing 1 (Uroc1), which is required for folate-dependent his- in wildtype mice throughout the embryonic and postnatal tidine catabolism. However, the absence of Uroc1 in CNS. Thus, NEUT2 homozygotes carrying the RP23-7M9 NEUT2 homozygotes has been proposed to actually lessen BAC transgene (Ϫ/Ϫ;TgN(RP23-7M9)) serve as an appro- the disturbance to folate metabolism by reducing the load priate control to assess the specific effects of ALDH1L1 on the cellular THF pool (Cook, 2001b). Therefore, the deficiency in NEUT2 mice. The Journal of Comparative Neurology. DOI 10.1002/cne

378 T.E. ANTHONY AND N. HEINTZ

Fig. 7. No morphological or proliferation defects are observed in ﬁbers appeared unaffected by loss of ALDH1L1 function (A–C). The NEUT2 Ϫ/Ϫ embryos. Immunostaining for nestin (A–C), Ki67 relative absence of proliferation at midline structures is preserved (D,E), and PH3 (F,G). Regions shown are the E14.5 lateral ganglionic (arrows in D,E), and no apparent increase in the number of mitotic eminence VZ (A–C,F,G) and tectal dorsal midline (D,E); genotypes are cells was observed in regions that would normally have upregulated stated in each panel. Staining for nestin revealed no obvious defects in ALDH1L1 (F,G). Scale bar ϭ 50 ␮m. the radial glial scaffolding at any CNS site, and density of radial glial

Aldh1l1 null mice do not have increased resemble NTDs (Fig. 6), but there was no significant dif- incidence of NTDs ference in occurrence among the three genotypes (#malformations/#embryos: wildtype, 1/54 ϳ1.8%; heterozy- Mendelian ratios of mice generated from heterozygous gous, 6/184 ϳ3.3%; homozygous, 4/116 ϳ3.5%). Although crosses did not statistically differ from the expected 1:2:1 the trend appears to be toward Aldh1l1 deficiency increas- wildtype:heterozygous:homozygous proportions (n ϭ 58: 2 ing risk of malformations on a low folate diet, the chi- 91:48) when typed at 10 days of age (␹ Ͼ0.5). As NTD square test statistic cannot be applied to such a low pen- penetrance is variable in both humans and mice (Juriloff and Harris, 2000), we screened E12 embryos to see if low penetrant NTDs might occur in NEUT2 homozygotes. No NTDs were found for any genotype (n ϭ 58:98:119 wt:het: homo), indicating that Aldh1l1 is not critical for neural Fig. 8. ALDH1L1 is induced in reactive astrocytes but is not tube closure in mice under normal circumstances. essential for regulating their proliferation. Immunostaining for ALDH1L1 (A–D,F,G,I), BrdU (E,F,K,L), and GFAP (H–J,L) in the Although low folate status alone is not sufficient to injured adult neocortex 3 days postlesion. Following neuronal injury, induce NTDs (Heid et al., 1992), the finding that some ALDH1L1 is upregulated in astrocytes specifically in the ipsilateral mothers of affected pregnancies have low serum folate hemisphere (A,B; the asterisk in B marks the wound site). The mor- suggests that on some genetic backgrounds maternal fo- phology of ALDH1L1ϩ reactive astrocytes is shown in C. Note that late levels could be an important determinant of disease one of these cells (arrow) has nuclear immunoreactivity; this was (Daly et al., 1995). To determine if NEUT2 homozygotes observed in many cells surrounding the wound site, but rarely in normal brains and not at all in Aldh1l1 null mice or when preimmune might be susceptible to NTDs in the presence of low ma- antisera was used (data not shown). Immunostaining for BrdU ad- ternal serum folates, dams were placed on a defined amino ministered 2 and 1⁄2 hours prior to perfusion demonstrates the pres- acid diet containing folic acid at 0.4 mg/kg (906 nmol/kg); ence of ALDH1L1ϩBrdUϩ cells (D–F; arrows in F). As expected, some this is ϳ7.5-fold less than normal mouse chow, and has BrdUϩ cells are not positive for ALDH1L1 (arrowhead in F); the small size of this nucleus suggests it is a microglial cell. Double labeling for previously been shown to significantly reduce maternal ϩ ϩ folate profiles without effects on litter size, resorption ALDH1L1 and GFAP confirms that all ALDH1L1 cells are GFAP astrocytes and vice versa (G–I); this allowed GFAP/BrdU double rate, or embryonic weight (Heid et al., 1992). E12.5 em- staining (J–L) to be used in NEUT2 Ϫ/Ϫ mice to quantitate prolifer- bryos examined from dams fed this diet developed malfor- ation in Aldh1l1 null reactive astrocytes. Scale bar ϭ 100 ␮m for A,B; mations at low frequency that superficially appeared to 25 ␮m for J–L; 20 ␮m for D–I; 10 ␮m for C. The Journal of Comparative Neurology. DOI 10.1002/cne

NEURAL TUBE MIDLINE RESTRICTION 379

Figure 8 The Journal of Comparative Neurology. DOI 10.1002/cne

380 T.E. ANTHONY AND N. HEINTZ etrant phenotype without a significantly larger number of part of the repair process (Amat et al., 1996; Stoll et al., animals. Thus, our results suggest that, at best, Aldh1l1 1998; Raivich et al., 1999). The expression of ALDH1L1 in deficiency could cause only a modest increase in NTDs. mature astrocytes (Neymeyer et al., 1997) suggested that No morphological defects due to absence of it might be part of a pathway that limits proliferation of reactive astroglia following neuronal injury. To test this Aldh1l1 are detectable idea, we used an established injury model in which mice Grossly, NEUT2 homozygous embryos derived from receive a cortical stab wound (Amat et al., 1996; Koguchi dams on both normal and folate-deficient diets appeared et al., 2002). Staining of injured adult wildtype mice dem- normal. Immunostaining for nestin revealed no apparent onstrated that ALDH1L1 protein is upregulated in astro- defects in the radial glial scaffold in any CNS region (Fig. cytes surrounding the wound site, but not on the con- 7A–C). Postnatally, hydrocephalus of varying severity was tralateral side of the cortex (Fig. 8A–C). Upregulation observed specifically in homozygotes (#hydrocephalic/#an- could be detected as early as 1 day postlesion, and strong imals: wildtype, 0/10; heterozygous, 0/14; homozygous, 21/ ALDH1L1 expression was seen by 3 days postlesion (data 21), which caused death by 1 month of age in ϳ18.5% of not shown). To determine the percentage of reactive as- homozygous mice (27/146). However, hydrocephaly was trocytes in S-phase, BrdU was administered to injured not rescued in Ϫ/Ϫ;TgN(RP23-7M9) mice (16/16 displayed mice at 3 days postlesion, and animals were anesthetized hydrocephalus), demonstrating that it is not due to loss of and perfused 2 and 1⁄2 hours later; this time point was Aldh1l1. Detailed analysis (including hematology, clinical chosen since astroglial proliferation has previously been chemistry, urinalysis, and histological examination of all shown to be maximal 3 days after injury (Amat et al., tissues) of 3-month-old mice maintained for 6 weeks on a 1996). Double immunofluorescence for ALDH1L1 and folate-deficient diet (0.4 mg/kg amino acid diet) revealed BrdU identified numerous ALDH1L1ϩBrdUϩ cells, dem- no significant differences between Ϫ/Ϫ and Ϫ/Ϫ; onstrating that upregulation of ALDH1L1 is not incom- TgN(RP23-7M9) mice (two mice each genotype analyzed). patible with proliferation. To determine if ALDH1L1 No proliferation defects are detectable in might be involved in regulating the rate of reactive astroglial division, we compared the percentage of BrdUϩ as- Aldh1l1 null mice trocytes in Ϫ/Ϫ and Ϫ/Ϫ;TgN(RP23-7M9) adult mice at 3 Initial analysis of radial glial proliferation in embryos days postlesion (three mice per genotype). Since all taken from dams fed a normal diet did not show any ALDH1L1ϩ astrocytes surrounding the wound site were obvious difference in the absence of Aldh1l1. However, it found to be GFAPϩ and vice versa (Fig. 8G–I), double remained a possibility that ALDH1L1 function might only labeling for GFAP and BrdU was used for quantitation be rate-limiting under certain conditions, i.e., reduced (Fig. 8J–L). No significant differences in the percentage availability of folic acid. To test this idea, dams were of BrdUϩ astrocytes were observed between the two placed on low folate diets as described above and embryos genotypes [BrdUϩGFAPϩ cells/GFAPϩ cells Ϯ SD: Ϫ/Ϫ were analyzed for proliferation. As was the case for em- 16.5% Ϯ 1 (n ϭ 317 GFAPϩ cells), Ϫ/Ϫ;TgN(RP23-7M9) bryos on normal diets, immunostaining revealed no obvi- 15.1% Ϯ 1.8 (n ϭ 425 GFAPϩ cells)]. These results ous expansion of proliferation into the midlines of either demonstrate that loss of ALDH1L1 activity is not re- the tectum (Fig. 7D,E) or spinal cord (data not shown) in quired to regulate reactive astroglial proliferation. low-folate embryos. To obtain quantitative data on prolif- Ventral midline radial glia are labeled in eration in low-folate embryos, we looked at the lateral ganglionic eminences (LGE). Similar to the CGE (Fig. Aldh1l1::eGFP mice 3C,D), ALDH1L1 is absent from the LGE at E12.5 but Aldh1l1ϩ midline radial glia are a potential cellular strongly upregulated by E14.5, correlating with a decrease target upon which supplemental folate may act. Because in the number of mitotic cells in this region. To determine RP23-7M9 (the BAC used to restore Aldh1l1 expression if ALDH1L1 is required for this decrease, we counted the onto NEUT2 homozygotes) drove correct transgenic ex- number of PH3ϩ cells (mitotic cells) present in the ven- pression, ALDH1L1ϩ midline radial glia can now be ge- tricular zone at E14.5 and E15.5 (Fig. 7F,G). Importantly, netically manipulated to test their role in neural tube we compared Ϫ/Ϫ and Ϫ/Ϫ;TgN(RP23-7M9) littermates, closure. In addition, these cells can be isolated for tran- sectioned coronally and counted only in the most rostral scriptional profiling in order to determine what other com- portion of the LGE (up until the MGE) to avoid ambigu- ponents of folate metabolism might be specifically ex- ities due to possible regional proliferative differences; pressed at the midline in normal or mutant mice. To three embryos per genotype for each timepoint were facilitate this, we generated BAC transgenic mice that stained. No differences were observed at either timepoint express eGFP from the Aldh1l1 genomic locus. Although [PH3ϩ cells/section Ϯ standard deviation, SD: E14.5 Ϫ/Ϫ the full characterization of these mice will be presented 61.5 Ϯ 5.7 (n ϭ 10 sections), E14.5 Ϫ/Ϫ;TgN(RP23-7M9) elsewhere, we report here that eGFP expression in 64.5 Ϯ 7.9 (n ϭ 7 sections), E15.5 Ϫ/Ϫ 46.3 Ϯ 4.8 (n ϭ 9 Aldh1l1::eGFP mice is detected specifically in midline ra- sections), E15.5 Ϫ/Ϫ;TgN(RP23-7M9) 47.7 Ϯ 7.0 (n ϭ 12 dial glia in the early neural tube (Fig. 9). Use of sections)]. These data indicate that Aldh1l1 is not essen- fluorescence-activated cell sorting to purify Aldh1l1ϩ mid- tial for the normal reduction in proliferation that occurs as line cells will enable comparisons with Aldh1l1Ϫ cells radial glia mature. within the neural tube both in normal mice as well as in An alternative possibility that existed was that folate-responsive NTD mouse models. These mice have ALDH1L1 might be required to prevent excessive prolif- been transferred to the publicly available repository main- eration in response to mitogenic signals after most devel- tained by the Mutant Mouse Regional Resource Centers opmental processes were finished. One situation where (MMRRC), and should provide a valuable tool for the such signaling takes place is following neuronal injury, dissection of metabolic pathways involved in neurulation where damaged tissues induce astroglial proliferation as as well as those that prevent disease. The Journal of Comparative Neurology. DOI 10.1002/cne

NEURAL TUBE MIDLINE RESTRICTION 381 DISCUSSION In this study we provide the first evidence that an enzyme involved in folate metabolism has restricted expression in the developing neural tube. We have shown that ALDH1L1 is expressed in radial glia flanking the ventral midline from as early as E9.5, and becomes upregulated at the dorsal midline soon after. Comparison with markers of dividing cells showed that regions expressing ALDH1L1 were largely nonproliferative and vice versa, demonstrating that proliferation is repressed in specific midline populations and raising the possibility that restricted expression of folate metabolic enzymes could be a mechanism controlling proliferation in subdo- mains of the neural tube. Finally, we demonstrated that Aldh1l1 null mice are not prone to NTDs and do not have detectable proliferation defects, indicating that ALDH1L1 activity is not essential for these processes. It is important to note that although NEUT2 homozygous mice do not develop NTDs, Aldh1l1 remains a candidate for involvement in human NTDs. One obvious reason for this is that Aldh1l1 mutations that contrib- ute to disease could be gain-of-function. Support for this idea comes from the metabolic effects of the teratogen valproic acid (VPA), which is an anticonvulsant that increases risk of human NTDs (Lammer et al., 1987). VPA is also teratogenic in mice and causes NTDs that can be prevented with folate supplementation (Padma- nabhan and Shafiullah, 2003). When VPA-treated embryos were biochemically analyzed, they were found to contain reduced levels of 10-FTHF and increased levels of THF (Greene and Copp, 2005). These changes to cellular folate pools are the opposite of those observed in NEUT2 homozygous livers (Champion et al., 1994), and the same as what would be predicted if ALDH1L1 were overexpressed or overactive. Considering that we did not detect proliferation defects in Aldh1l1 null mice, whereas overexpressing ALDH1L1 suppresses proliferation (Krupenko and Oleinik, 2002; Oleinik and Kru- penko, 2003), the available data suggest that if Aldh1l1 does affect neurulation it is gain-of-function mutations that would be most likely to cause disease. If loss-of-function mutations in Aldh1l1 are relevant, there are several reasons why significant phenotypes were not observed in NEUT2 mice. First, the NEUT2 deletion elimi- nates both Aldh1l1 and Uroc1 (described above). Loss of Uroc1 is believed to mitigate the disruption of folate metabolism in the absence of Aldh1l1 by reducing the load on the cellular THF pool (Cook, 2001b). Therefore, it remains possible that a specific targeted mutation in the Aldh1l1 gene could reveal a role in neural tube closure that cannot be readily observed in NEUT2 mice. With regard to this latter point, it should be noted that our analysis of NTD incidence in NEUT2 mice maintained on low folate included only 354 embryos; as the NTD penetrance was very low, this number was insufficient to provide a definitive conclusion on the potential role of Aldhd1l1. As the mitigating effects of the Uroc1 deletion may have contributed to the observed low

Fig. 9. Aldh1l1::eGFP BAC transgenic mice drive expression in midline spinal cord radial glia. Immunostaining for the radial glial marker BLBP (A,C) and eGFP (B,C) in the E12.5 spinal cord of Aldh1l1::eGFP BAC transgenic embryos. Note that, whereas BLBP is expressed in all radial glia, eGFP is restricted to cells in the most ventral and dorsal areas; this is identical to endogenous ALDH1L1 expression (compare with Fig. 1). Scale bar ϭ 50 ␮m. The Journal of Comparative Neurology. DOI 10.1002/cne

382 T.E. ANTHONY AND N. HEINTZ penetrance, either targeted mutation of Aldh1l1 or analysis Thus, these data argue that folate rescues NTDs in Cited2 of a significantly larger number of NEUT2 embryos will be null mice by driving compensatory processes in or near required to resolve this issue. It should also be noted that the ALDH1L1ϩ midline cells. Moreover, exogenous folic acid etiology of human NTDs is complex and depends on both has been shown to be capable of overcoming ALDH1L1- genetic and environmental factors (Botto et al., 1999; Juriloff mediated inhibition in vitro: overexpression of ALDH1L1 and Harris, 2000). Accordingly, it is possible that the mixed represses growth of several cell types, but addition of NEUT2 genetic background studied here is not susceptible folinic acid (5-FTHF) abolishes this repression (Krupenko to Aldh1l1 deficiency under the conditions tested. This issue and Oleinik, 2002). A direct genetic test for an inhibitory is clearly relevant to modeling folate-preventable NTDs in role of ALDH1L1 would be to determine if the NEUT2 mice, as genetic background has been observed to alter the microdeletion can rescue NTDs in folate-responsive mouse frequency of cranial NTDs in Sp2H/Sp2H embryos (Fleming models, particularly the Cited2 knockout, which has no and Copp, 2000). Another factor to consider is the difference inherent defects in folate cycling. Similarly, it would be in the amount of time neurulation takes in mice compared interesting to know whether ALDH1L1 overexpression with humans. Whereas neurulation in the mouse is complete can increase incidence of disease in mouse strains that in less than 2 days (Harris and Juriloff, 1999), humans develop folate-responsive NTDs (e.g., by placing the RP23- require nearly 2 weeks to complete the process (Nolte, 1993). 7M9 BAC transgene onto a Cited2 null background). Efforts to model human neurodegenerative diseases in mice have often required drastic increases in the toxicity of disease-causing molecules (above that already present in the ACKNOWLEDGMENTS forms that are pathogenic in humans) in order to see a We thank Dr. Robert J. Cook for providing antisera phenotype in the short lifespan of a mouse (Watase and against ALDH1L1 and for numerous extremely helpful Zoghbi, 2003). Similarly, multiple folate-dependent path- discussions on folate biochemistry, Dr. Donald Horne for ways may need to be disrupted in mice in order to mimic the doing folate profiles on NEUT2 embryos, Dr. Dean Ap- effects of prolonged deficiency of a single mutant gene in pling for the MTHFD1 antiserum, and Dr. Carol Giometti humans. Additionally, and as in other cases of mouse mu- for providing the NEUT2 mice. We also thank Drs. Robert tants without detectable phenotypes, compensation by or Darnell, Michael Young, and Mary-Beth Hatten for help- alteration of the activity in other components of folate me- ful discussions during the course of this work, and Dr. tabolism may have obscured a role for Aldh1l1 in neurula- Hatten for generous provision of reagents. Antibodies tion. Finally, we note that other folate metabolic enzymes against nestin and PAX3 were obtained from the Devel- implicated in human NTDs have been knocked out in mice opmental Studies Hybridoma Bank (Iowa City, IA). with no apparent detrimental effect on neural tube closure. For example, numerous studies have shown that a thermo- labile allele of the 5,10-methylenetetrahydrofolate reduc- LITERATURE CITED tase gene (Mthfr) is associated with increased risk of Amat JA, Ishiguro H, Nakamura K, Norton WT. 1996. Phenotypic diversity NTDs in multiple human populations (reviewed in Boyles and kinetics of proliferating microglia and astrocytes following cortical et al., 2005); however, Mthfr null mice do not develop stab wounds. Glia 16:368–382. NTDs (Chen et al., 2001). This example in particular Baker CV, Stark MR, Marcelle C, Bronner-Fraser M. 1999. Competence, makes it clear that negative results in mouse models must specification and induction of Pax-3 in the trigeminal placode. Devel- opment 126:147–156. be interpreted with caution. Bamforth SD, Braganca J, Eloranta JJ, Murdoch JN, Marques FI, Kranc KR, Although we did not detect increases in radial or astro- Farza H, Henderson DJ, Hurst HC, Bhattacharya S. 2001. Cardiac mal- glial proliferation in NEUT2 homozygotes, overexpression formations, adrenal agenesis, neural crest defects and exencephaly in of Aldh1l1 has previously been shown to suppress prolif- mice lacking Cited2, a new Tfap2 co-activator. Nat Genet 29:469–474. eration in vitro (Krupenko and Oleinik, 2002; Oleinik and Barbera JP, Rodriguez TA, Greene ND, Weninger WJ, Simeone A, Copp Krupenko, 2003). Moreover, multiple tumor tissues were AJ, Beddington RS, Dunwoodie S. 2002. Folic acid prevents exencephaly in Cited2 deficient mice. Hum Mol Genet 11:283–293. shown to downregulate Aldh1l1 expression (Krupenko Botto LD, Moore CA, Khoury MJ, Erickson JD. 1999. Neural-tube defects. and Oleinik, 2002). These data indicate that while N Engl J Med 341:1509–1519. Aldh1l1 is not an essential regulator of proliferation, its Boyles AL, Hammock P, Speer MC. 2005. Candidate gene analysis in human presence may be restrictive. If correct, this would suggest neural tube defects. Am J Med Genet C Semin Med Genet 135:9–23. a testable mechanism by which folic acid might prevent Brody LC, Conley M, Cox C, Kirke PN, McKeever MP, Mills JL, Molloy NTDs in the absence of defective folate metabolism. By AM, O’Leary VB, Parle-McDermott A, Scott JM, Swanson DA. 2002. A restricting proliferation in and around the midline, polymorphism, R653Q, in the trifunctional enzyme methylenetetrahydrofolate dehydrogenase/methenyltetrahydrofolate cyclohydrolase/ ALDH1L1 may limit critical compensatory proliferative formyltetrahydrofolate synthetase is a maternal genetic risk factor for adjustments necessary in the presence of NTD-inducing neural tube defects: report of the Birth Defects Research Group. Am J mutations. However, the presence of high levels of exoge- Hum Genet 71:1207–1215. nous folate could overcome the repressive effects of Campbell LR, Dayton DH, Sohal GS. 1986. Neural tube defects: a review of ALDH1L1, allowing compensatory reactions to proceed. human and animal studies on the etiology of neural tube defects. Teratology 34:171–187. Experimental support for such a model comes from sev- Champion KM, Cook RJ, Tollaksen SL, Giometti CS. 1994. Identification of eral sources. First, excessive apoptotic cell death was ob- a heritable deficiency of the folate-dependent enzyme 10- served in Cited2 null mice specifically at the midline of the formyltetrahydrofolate dehydrogenase in mice. Proc Natl Acad Sci U S tectum (Barbera et al., 2002), a region in which we have A 91:11338–11342. shown both restricted expression of ALDH1L1 and mini- Cheek WD, Appling DR. 1989. Purification, immunoassay, and tissue mal proliferation. Although folate does not prevent this distribution of rat C1-tetrahydrofolate synthase. Arch Biochem Bio- phys 270:504–512. excessive cell death (Barbera et al., 2002), apoptosis is Chen Z, Karaplis AC, Ackerman SL, Pogribny IP, Melnyk S, Lussier-Cacan likely causative, as crossing splotch mice onto a p53 null S, Chen MF, Pai A, John SW, Smith RS, Bottiglieri T, Bagley P, Selhub background completely prevents NTDs (Pani et al., 2002). J, Rudnicki MA, James SJ, Rozen R. 2001. Mice deficient in methyl- The Journal of Comparative Neurology. DOI 10.1002/cne

NEURAL TUBE MIDLINE RESTRICTION 383

enetetrahydrofolate reductase exhibit hyperhomocysteinemia and de- Kutzbach C, Stokstad EL. 1971. 10-Formyltetrahydrofolate: NADP oxire- creased methylation capacity, with neuropathology and aortic lipid ductase. Methods Enzymol 18B:793–798. deposition. Hum Mol Genet 10:433–443. Lammer EJ, Sever LE, Oakley GP Jr. 1987. Teratogen update: valproic Cook R. 2001a. Folate metabolism. In: Carmel R, Jacobsen D, editors. acid. Teratology 35:465–473. Homocysteine in health and disease. Cambridge, UK: Cambridge Uni- Li J, Liu KC, Jin F, Lu MM, Epstein JA. 1999. Transgenic rescue of versity Press. congenital heart disease and spina bifida in Splotch mice. Development Cook RJ. 2001b. Disruption of histidine catabolism in NEUT2 mice. Arch 126:2495–2503. Biochem Biophys 392:226–232. Mansouri A, Pla P, Larue L, Gruss P. 2001. Pax3 acts cell autonomously in Copp AJ, Greene ND, Murdoch JN. 2003. The genetic basis of mammalian the neural tube and somites by controlling cell surface properties. neurulation. Nat Rev Genet 4:784–793. Development 128:1995–2005. Cragan JD, Roberts HE, Edmonds LD, Khoury MJ, Kirby RS, Shaw GM, Velie McBride ML. 1979. Sib risk of anencephaly and spina bifida in British EM, Merz RD, Forrester MB, Williamson RA, Krishnamurti DS, Stevenson Columbia. Am J Med Genet 3:377–387. RE, Dean JH. 1995. Surveillance for anencephaly and spina bifida and the impact of prenatal diagnosis—United States, 1985-1994. MMWR CDC Milewski RC, Chi NC, Li J, Brown C, Lu MM, Epstein JA. 2004. Identifi- Surveill Summ 44:1–13. cation of minimal enhancer elements sufficient for Pax3 expression in neural crest and implication of Tead2 as a regulator of Pax3. Develop- Czeizel AE, Dudas I. 1992. Prevention of the first occurrence of neural-tube ment 131:829–837. defects by periconceptional vitamin supplementation. N Engl J Med 327:1832–1835. Neymeyer V, Tephly TR, Miller MW. 1997. Folate and 10-formyltetrahydrofolate dehydrogenase (FDH) expression in the central ner- Daly LE, Kirke PN, Molloy A, Weir DG, Scott JM. 1995. Folate levels and vous system of the mature rat. Brain Res 766:195–204. neural tube defects. Implications for prevention. JAMA 274:1698–1702. Noctor SC, Flint AC, Weissman TA, Wong WS, Clinton BK, Kriegstein AR. Del Punta K, Leinders-Zufall T, Rodriguez I, Jukam D, Wysocki CJ, Ogawa S, Zufall F, Mombaerts P. 2002. Deficient pheromone responses in mice 2002. Dividing precursor cells of the embryonic cortical ventricular lacking a cluster of vomeronasal receptor genes. Nature 419:70–74. zone have morphological and molecular characteristics of radial glia. J Neurosci 22:3161–3173. Dulac C. 2000. Sensory coding of pheromone signals in mammals. Curr Opin Neurobiol 10:511–518. Nolte J. 1993. The human brain. St. Louis: Mosby Year-Book. Epstein DJ, Vekemans M, Gros P. 1991. Splotch (Sp2H), a mutation Oleinik NV, Krupenko SA. 2003. Ectopic expression of 10-formyl- affecting development of the mouse neural tube, shows a deletion tetrahydrofolate dehydrogenase in A549 cells induces G1 cell cycle within the paired homeodomain of Pax-3. Cell 67:767–774. arrest and apoptosis. Mol Cancer Res 1:577–588. Ever L, Gaiano N. 2005. Radial ‘glial’ progenitors: neurogenesis and sig- Padmanabhan R, Shafiullah MM. 2003. Amelioration of sodium valproate- naling. Curr Opin Neurobiol 15:29–33. induced neural tube defects in mouse fetuses by maternal folic acid Fishell G, Kriegstein AR. 2003. Neurons from radial glia: the consequences supplementation during gestation. Congenit Anom (Kyoto) 43:29–40. of asymmetric inheritance. Curr Opin Neurobiol 13:34–41. Pani L, Horal M, Loeken MR. 2002. Rescue of neural tube defects in Fleming A, Copp AJ. 1998. Embryonic folate metabolism and mouse neural Pax-3-deficient embryos by p53 loss of function: implications for Pax- tube defects. Science 280:2107–2109. 3-dependent development and tumorigenesis. Genes Dev 16:676–680. Fleming A, Copp AJ. 2000. A genetic risk factor for mouse neural tube Park CH, Pruitt JH, Bennett D. 1989. A mouse model for neural tube defects: defining the embryonic basis. Hum Mol Genet 9:575–581. defects: the curtailed (Tc) mutation produces spina bifida occulta in Tc/ϩ animals and spina bifida with meningomyelocele in Tc/t. Teratol- Giometti CS, Tollaksen SL, Grahn D. 1994. Altered protein expression ogy 39:303–312. detected in the F1 offspring of male mice exposed to fission neutrons. Mutat Res 320:75–85. Raivich G, Bohatschek M, Kloss CU, Werner A, Jones LL, Kreutzberg GW. Goulding MD, Chalepakis G, Deutsch U, Erselius JR, Gruss P. 1991. 1999. Neuroglial activation repertoire in the injured brain: graded Pax-3, a novel murine DNA binding protein expressed during early response, molecular mechanisms and cues to physiological function. neurogenesis. EMBO J 10:1135–1147. Brain Res Brain Res Rev 30:77–105. Greene ND, Copp AJ. 2005. Mouse models of neural tube defects: investi- Rakic P. 2003. Developmental and evolutionary adaptations of cortical gating preventive mechanisms. Am J Med Genet C Semin Med Genet radial glia. Cereb Cortex 13:541–549. 135:31–41. Scrutton MC, Beis I. 1979. Inhibitory effects of histidine and their reversal. Harris MJ, Juriloff DM. 1999. Mini-review: toward understanding mech- The roles of pyruvate carboxylase and N10-formyltetrahydrofolate de- anisms of genetic neural tube defects in mice. Teratology 60:292–305. hydrogenase. Biochem J 177:833–846. Heid MK, Bills ND, Hinrichs SH, Clifford AJ. 1992. Folate deficiency alone Shibata T, Yamada K, Watanabe M, Ikenaka K, Wada K, Tanaka K, Inoue does not produce neural tube defects in mice. J Nutr 122:888–894. Y. 1997. Glutamate transporter GLAST is expressed in the radial Hockfield S, McKay RD. 1985. Identification of major cell classes in the glia-astrocyte lineage of developing mouse spinal cord. J Neurosci developing mammalian nervous system. J Neurosci 5:3310–3328. 17:9212–9219. Hogan B, Beddington R, Costantini F, Lacy E. 1994. Manipulating the mouse Shibuya K, Murray CJL. 1998. Congenital anomalies. In: Murray CJL, embryo: a laboratory manual. Plainview, NY: Cold Spring Harbor Press. Lopez AD, editors. Health dimensions of sex and reproduction: the global burden of sexually transmitted diseases, HIV, maternal condi- Juriloff DM, Harris MJ. 2000. Mouse models for neural tube closure tions, perinatal disorders, and congenital anomalies. Boston: Harvard defects. Hum Mol Genet 9:993–1000. University Press. p 455–512. Keller-Peck CR, Mullen RJ. 1997. Altered cell proliferation in the spinal Smithells RW, Sheppard S, Schorah CJ, Seller MJ, Nevin NC, Harris R, Read cord of mouse neural tube mutants curly tail and Pax3 splotch-delayed. Brain Res Dev Brain Res 102:177–188. AP, Fielding DW. 1981. Apparent prevention of neural tube defects by periconceptional vitamin supplementation. Arch Dis Child 56:911–918. Kim DW, Huang T, Schirch D, Schirch V. 1996. Properties of tetrahydrop- teroylpentaglutamate bound to 10-formyltetrahydrofolate dehydroge- Stoll G, Jander S, Schroeter M. 1998. Inflammation and glial responses in nase. Biochemistry 35:15772–15783. ischemic brain lesions. Prog Neurobiol 56:149–171. Koguchi K, Nakatsuji Y, Nakayama K, Sakoda S. 2002. Modulation of Takahashi T, Nowakowski RS, Caviness VS Jr. 1995. The cell cycle of the astrocyte proliferation by cyclin-dependent kinase inhibitor p27(Kip1). pseudostratified ventricular epithelium of the embryonic murine cere- Glia 37:93–104. bral wall. J Neurosci 15:6046–6057. Krebs HA, Hems R, Tyler B. 1976. The regulation of folate and methionine Venters SJ, Argent RE, Deegan FM, Perez-Baron G, Wong TS, Tidyman metabolism. Biochem J 158:341–353. WE, Denetclaw WF Jr, Marcelle C, Bronner-Fraser M, Ordahl CP. Krupenko SA, Oleinik NV. 2002. 10-Formyltetrahydrofolate dehydroge- 2004. Precocious terminal differentiation of premigratory limb muscle nase, one of the major folate enzymes, is down-regulated in tumor precursor cells requires positive signalling. Dev Dyn 229:591–599. tissues and possesses suppressor effects on cancer cells. Cell Growth Wald N, Sneddon J, Densem J, Frost C, Stone R. 1991. Prevention of Differ 13:227–236. neural tube defects: results of the Medical Research Council Vitamin Kuhar SG, Feng L, Vidan S, Ross ME, Hatten ME, Heintz N. 1993. Study. MRC Vitamin Study Research Group. Lancet 338:131–137. Changing patterns of gene expression define four stages of cerebellar Watase K, Zoghbi HY. 2003. Modelling brain diseases in mice: the chal- granule neuron differentiation. Development 117:97–104. lenges of design and analysis. Nat Rev Genet 4:296–307.