Glycosylphosphatidylinositol Biosynthesis and Remodeling are Required for Neural Crest Cell, Cardiac and Neural Development

The glycosylphosphatidylinositol (GPI) anchor attaches nearly 150 proteins to the cell surface. Patients with pathogenic variants in GPI biosynthetic pathway genes display an array of phenotypes including seizures, developmental delay, dysmorphic facial features and cleft palate. There is virtually no mechanism to explain these phenotypes. we identified a novel mouse mutant (cleft lip/palate, edema and exencephaly; Clpex) with a hypomorphic mutation in Post-Glycophosphatidylinositol Attachment to Proteins-2 (Pgap2). Pgap2 is one of the final proteins in the GPI biosynthesis pathway and is required for anchor maturation. We found the Clpex mutation results in a global decrease in surface GPI expression. Surprisingly, Pgap2 showed tissue specific expression with enrichment in the affected tissues of the Clpex mutant. We found the phenotype in Clpex mutants is due to apoptosis of neural crest cells (NCCs) and the cranial neuroepithelium, as is observed in the GPI anchored Folate Receptor 1-/- mouse. We showed folinic acid supplementation in utero can rescue the cleft lip phenotype in Clpex. Finally, we generated a novel mouse model of NCC-specific total GPI deficiency in the Wnt1-Cre lineage. These mutants developed median cleft lip and palate demonstrating a cell autonomous role for GPI biosynthesis in NCC development.


Introduction
Inherited glycophosphatidylinositol deficiency (IGD) disorders are a class of congenital disorders of glycosylation that affect the biosynthesis of the glycosylphosphatidylinositol (GPI) anchor. The clinical spectrum of IGDs is broad and includes epilepsy, developmental delay, structural brain malformations, cleft lip/palate, skeletal hypoplasias, deafness, ophthalmological abnormalities, gastrointestinal defects, genitourinary defects, heart defects, Hirschsprung's disease, hyperphosphatasia, and nephrogenic defects [1][2][3][4]. However, across all GPI deficiency disorders, the most penetrant defects affect the central nervous system and the craniofacial complex [1][2][3][4]. Indeed, automated image analysis was able to predict the IGD gene mutated in each patient from facial gestalt [2].
Interestingly, facial gestalt was a better predictor of patient mutation than analysis of the degree of GPI biosynthesis by flow cytometry. Little is known about the mechanism(s) that causes these phenotypes or why disparate tissues are differentially affected [4] [4].
We sought to determine the mechanism responsible for these phenotypes using a novel mouse model of reduced enzymatic function within the GPI biosynthesis pathway.
The GPI anchor is a glycolipid added post-translationally to nearly 150 proteins which anchors them to the outer leaflet of the plasma membrane and traffics them to lipid rafts [1]. The biosynthesis and remodeling of the GPI anchor is extensive and requires nearly 30 genes [1]. Once the glycolipid is formed and transferred to the C-terminus of the target protein by a variety of Phosphophatidylinositol Glycan (PIG proteins), it is transferred to the Golgi Apparatus for remodeling by Post-GPI Attachment to Proteins (PGAP proteins).
One of these PGAP proteins involved in remodeling the GPI anchor is Post-Glycosylphosphatidylinositol Attachment to Proteins 2 (PGAP2). PGAP2 is a transmembrane protein that catalyzes the addition of stearic acid to the lipid portion of the GPI anchor and cells deficient in Pgap2 lack stable surface expression of a variety of GPIanchored proteins (GPI-APs) [1,5]. Autosomal recessive mutations in PGAP2 cause Hyperphosphatasia with Mental Retardation 3 (HPMRS3 OMIM # 614207), an IGD that presents with variably penetrant hyperphosphatasia, developmental delay, seizures, microcephaly, heart defects, and a variety of neurocristopathies including Hirschsprung's disease, cleft lip, cleft palate, and facial dysmorphia [5][6][7][8]. Currently, there is no known molecular mechanism to explain the cause of these phenotypes or therapies for these patients.
In a forward genetic ENU mutagenesis screen, we previously identified the Clpex mouse mutant with Cleft Lip, Cleft Palate, Edema, and Exencephaly (Clpex) [9]. Here we present evidence that this mutant phenotype is caused by a hypo-morphic allele of Pgap2. To date, embryonic phenotypes of GPI biosynthesis mutants have been difficult to study due to the early lethal phenotypes associated with germline knockout of GPI biosynthesis genes [10][11][12][13]. The International Mouse Phenotyping Consortium labelled Pgap2 homozygote knockout mice preweaning lethal and the Deciphering the Mechanisms of Developmental Disorders initiative labelled Pgap2 homozygotes early lethal as no embryos were recovered at embryonic day E9. 5. In this study, we took advantage of the Clpex hypomorphic mutant to determine the mechanism of the various phenotypes and tested the hypothesis that GPI-anchored Folate Receptor 1 (FOLR1) is responsible for the phenotypes observed.
As we observed tissue specific defects in NCCs in the Clpex germline mutant, we sought to determine the cell autonomous requirement for GPI biosynthesis pathway generally in NCC development. To do this, we took a conditional approach to abolish GPI biosynthesis in NCCs using the Wnt1-Cre transgene and a floxed allele of a critical initiator of GPI biosynthesis. The Clpex mutant and our NCC conditional mutant serve as models for studying the effect of GPI biosynthesis defects in development of the craniofacial complex and testing potential therapeutics in utero for patients.

Results
The Clpex mutant phenotype is caused by a missense mutation in Pgap2.
We previously identified the Clpex (cleft lip and palate, edema, and exencephaly) mutant in a mouse N-ethyl-N-nitrosourea (ENU) mutagenesis screen for recessive alleles leading to organogenesis phenotypes [9]. Clpex homozygous mutants displayed multiple partially penetrant phenotypes. In a subset of 70 mutants from late organogenesis stages (~E16.5-  [9]. We then took a whole exome sequencing approach and sequenced 3 Clpex homozygous mutants. Analysis of single base pair variants which were homozygous in all 3 mutants with predicted high impact and not already known strain polymorphisms in dbSNP left only one candidate variant (Table 1). This was a homozygous missense mutation in the initiating methionine (c.A1G, p.M1V) in exon 3 of post-GPI attachment to proteins 2 (Pgap2). We confirmed the whole exome sequencing result by Sanger Sequencing (Fig. 1R). This mutation abolishes the canonical translation start codon for Pgap2. However, there are alternative start sites of Pgap2 and multiple alternatively spliced transcripts that may lead to production of variant forms of Pgap2.
To determine whether the Clpex phenotype was caused by the missense mutation in Pgap2, we performed a genetic complementation test using the Pgap2 tm1a(EUCOMM)Wtsi (hereafter referred to as Pgap2 null ) conditional gene trap allele. We crossed Pgap2 Clpex/+ heterozygotes with Pgap2 null/+ heterozygotes to generate Pgap2 Clpex/null embryos.
Pgap2 Clpex/null embryos displayed neural tube defects, bilateral cleft lip, and edema similar to the Pgap2 Clpex/Clpex embryos at E13.5 ( Fig. 2 A-H). Pgap2 Clpex/null embryos also displayed micro-opthalmia (Fig. 2F) and more penetrant cleft lip and edema phenotypes than observed in Pgap2 Clpex/Clpex homozygotes (Fig. 2I). Pgap2 Clpex/null embryo viability was decreased in this line with lethality at approximately E13.5-E14.5, precluding analysis of palatal development in these mutants. Histological analysis of the heart in Pgap2 Clpex/null E13.5 embryos showed pericardial effusion, a reduction in thickness of the myocardium, and an underdeveloped ventricular septum and valves ( Fig. 2J-Q). As the Clpex allele failed to complement a null allele of Pgap2, we concluded the Clpex phenotype is caused by a hypomorphic allele of Pgap2. The Deciphering Mechanisms of Developmental Disorders (DMDD) initiative has generated Pgap2 null/null homozygotes and classified them as "early lethal" abnormal chorion and trophoblast layer morphology [12]. These data argue Pgap2 is required for early embryogenesis and our Clpex allele must be hypomorphic, as the embryos survive to E18.5.

Pgap2 is dynamically expressed throughout development.
Based on the tissues affected in the Clpex mutant, we hypothesized Pgap2 is expressed in the neural folds and facial primordia of the developing mouse embryo at early stages.
We used the lacZ expression cassette within the Pgap2 null allele to perform detailed expression analysis of Pgap2 throughout development (n= 21 litters at multiple developmental stages). We performed RNA in situ hybridization in parallel for some stages to test the fidelity of the lacZ expression and found high concordance. Pgap2 was expressed relatively uniformly and ubiquitously at neurulation stages in the mouse from E7.5-E8.5 ( Fig. 3A-D). We also noted extraembryonic expression at E7.5, consistent with the abnormal placental development observed in Pgap2 null/null embryos (Fig. 3A,B) [12].
At E9.5-11.5, there was clear enrichment of Pgap2 expression in the first branchial arch showed lower expression in the liver (Fig. 3S) and most of the brain except for a thin layer of the cortex and the choroid plexus at E16.5 (Fig. 3S, U-V, X). We also noted expression in the genital tubercle (Fig. 3W). We conclude Pgap2 shows tissue specific regions of increased expression which may help to explain why certain tissues such as the craniofacial complex, central nervous system, and heart are differentially affected in GPI biosynthesis mutants. These data are in contrast to previous reports where some GPI biosynthesis genes are shown to be ubiquitously and uniformly expressed, including Pign in the mouse and pigu in zebrafish [10,14]. Our Pgap2 expression is more consistent with the expression of Pigv which is enriched in C. elegans epidermal tissues [15] Pgap2 is required for the proper anchoring of GPI-APs, including FOLR1.
PGAP2 is the final protein in the GPI biosynthesis pathway and catalyzes the addition of stearic acid to the GPI anchor [16]. GPI is a glycolipid added post-translationally to nearly 150 proteins in the endoplasmic reticulum (ER) and remodeled in the Golgi Apparatus.
GPI anchored proteins (GPI-APs) require the GPI anchor for their presentation in the outer leaflet of the plasma membrane and association with lipid rafts [1,16]. In the absence of Pgap2, cells lack a variety of GPI-APs on the cell surface leading to a functional GPI deficiency [5][6][7][8]16]. To determine the effect of the Clpex mutation on Pgap2 function, we performed Fluorescein-labeled proaerolysin (FLAER) flow cytometry staining to quantify the overall amount of the GPI anchor on the cell surface. FLAER is a bacterial toxin conjugated to fluorescein that binds directly to the GPI anchor in the plasma membrane. We hypothesized Pgap2 function is impaired in Clpex mutants due to the ENU mutation in the initiating methionine. We found mouse embryonic fibroblasts (MEFs) from Clpex mutants displayed a significantly decreased expression of FLAER compared to wildtype MEFs, consistent with a defect in GPI biosynthesis (n=3 separate experiments with 4 WT and 4 Clpex cell lines) (Fig. 4A,B).
Our genetic complementation analysis results suggested the Clpex allele might be a hypomorphic allele of Pgap2. To test this hypothesis, we generated human embryonic kidney (HEK) 293T clones with a 121bp deletion in exon 3 of PGAP2 with CRISPR/Cas9 (termed PGAP2 -/cells; Fig. S1). In parallel, we recapitulated the Clpex mutation in 3 independent clones of HEK293T cells by CRISPR/Cas9 mediated homologous directed repair (termed Clpex KI Clones 1, 4, and 7; Fig.S1). We found there was a statistically significant decrease in FLAER staining between WT and 2 of 3 KI clones and the PGAP2 -/cells. However, we observed no statistical difference in FLAER staining in PGAP2 -/cells when compared to Clpex KI cells (Fig. 4C,D). Therefore, we conclude the Clpex missense mutation severely affects PGAP2 function similar to the effect seen upon total depletion of PGAP2. As a positive control, we used CRISPR/Cas9 to delete phosphatidylinositol glycan anchor biosynthesis, class A (Piga; Fig.S1). Piga is the first gene in the GPI biosynthesis pathway and is absolutely required for GPI biosynthesis [6,17]. We utilized CRISPR/Cas9 to generate a 29bp out-of-frame deletion in exon 3 of PIGA. These PIGA -/cells showed an even further decrease in FLAER staining compared to PGAP2 -/cells, confirming our staining accurately reflects GPI anchor levels (n=4 separate experiments; Current estimates suggest nearly 150 genes encode proteins which are GPI anchored [18]. Our manual review of the MGI database found 102 GPI-APs have been genetically manipulated and phenotyped in mice [19]. Of these, the null allele of Folr1 has a phenotype most similar to the Clpex mutant with cranial neural tube defects, cleft lip/palate, and heart outflow tract phenotypes [20]. Tashima et. al. previously showed PGAP2 is required for stable cell surface expression of FOLR1 in CHO cells [16]. To confirm this finding, we overexpressed a myc-tagged FOLR1 construct in WT and PGAP2 -/-293T cells and assessed the presentation on the plasma membrane by immunocytochemistry for wheat germ agglutinin (WGA). We observed a decrease in colocalization of FOLR1 with WGA in PGAP2 -/cells compared to controls (Fig.   4E,F,H,I,K,L). In the absence of PIGA, cells lack the surface expression of any GPI-APs However, both PIGA -/and PGAP2 -/cells produced similar amounts of FOLR1 protein by western blot indicating that the defect is in trafficking, and not protein production ( Fig.   4N).
We observed no significant deficit in cNCC migration in the mutant embryos as compared to littermate controls at either stage (representative images of n=2 E9.5 mutants and n=5 E11.5 mutants) ( Fig. 5A-F). However, we observed hypoplasia of the medial and lateral nasal processes at E11.5, suggesting the Clpex phenotype is due to earlier defects in NCC survival (Fig 5E-F). As Pgap2 was highly expressed in the epithelium and epithelial barrier defects are a known cause of cleft palate, we next sought to determine whether the epidermis was compromised in the Clpex mutant [35]. We performed a Toluidine Blue exclusion assay but found no significant defects in barrier formation in the mutant (Fig.   S2). We also observed apoptosis in the cranial neuroepithelium at the dorsolateral hinge points ( Fig. 5S-T). The dorsolateral hingepoints constrict bilaterally in order to close the neural tube. This apoptosis was exclusively confined to the cranial aspects of the neural tube at the midbrain-hindbrain boundary

Dietary folinic acid supplementation partially rescues cleft lip in Clpex mutants.
Dietary folinic acid supplementation has been shown to rescue the early embryonic lethal phenotype of Folr1 -/mice and these mice can then survive to adulthood [26,28,37]. Our data suggest FOLR1 receptor trafficking is impaired in the Clpex mutant ( Fig. 4), leading us to hypothesize folinic acid supplementation in utero may rescue the Clpex phenotype.
We further hypothesized the folinic acid diet would have a greater beneficial effect as folinic acid (reduced folate) has a higher affinity for other folate receptors including solute carrier family 19 (folate transporter), member 1 (Slc19a1) and solute carrier family 46, member1 (Slc46a1),) which are not GPI anchored [38]. In comparison, folic acid has a higher affinity for the GPI anchored folate receptors FOLR1 and FOLR2 [39,40]. We supplemented pregnant Clpex dams from E0-E9.5 or E16.5 with a 25 parts per million (ppm) folinic acid diet, 25 ppm folic acid diet or control diet, and collected Clpex mutants for phenotypic analysis and Nile Blue staining for cell viability. We first found folinic acid supplementation decreased the Nile Blue staining for non-viable cells in the facial primordia of E9.5 mutants compared to controls (n=2 mutants; Fig. 6A,B). We also found the folinic acid diet had an effect on Clpex phenotypes at E16.5. The treatment group had a significantly smaller proportion of mutants with cleft lip (p=0.02) but there was no effect on the incidence of NTD or cleft palate (Fig. 6C,D). We did note a decrease in mutants with edema, however, this decrease was not statistically significant given our sample size (p=0.06; Fig. 6C,D). Consistent with our hypothesis, we found the folinic acid reduced the number of mutants with cleft lip by 23% (2/25 mutant vs. 22/70 control), which is more significant than the 10% reduction observed in folic acid treated mice (3/14, Fig. 6C,D).
Therefore, we conclude folinic acid treatment increased the viability of Clpex facial primordia and decreased the incidence of cleft lip among Clpex mutants.

RNA sequencing reveals changes in patterning genes in Clpex mutants
As folinic acid supplementation could not rescue all phenotypes observed in the Clpex  (Table 3). 13 Anterior patterning genes were statistically significantly downregulated and 10 posterior patterning genes were statistically significantly upregulated (Table 2). We confirmed changes in expression of three of these A/P patterning defects by RNA in situ hybridization at E9.5 (Fig. S3). We identified a decrease in Alx3 in Clpex mutants which is both only an anterior patterning gene with a prominent role in frontonasal development and a genetic target of folate signaling [50].
We investigated Lhx8 because it is expressed in the head at E9.5 and Lhx8 -/mice develop cleft palate [51]. We found Lhx8 was decreased in Clpex mutant heads. Finally, the master posterior patterning gene Tbxt (brachyury) is critical for determining tail length and posterior somite identity [45,46]. We found Tbxt was slightly increased in Clpex mutants compared to their overall body size.
Upon closer inspection, the genes in these categories were largely genes expressed in the mesendoderm including Alpha fetal protein and the Apolipoprotein gene family. We concluded the decreased expression of these genes in the Clpex mutant embryos is consistent with a defect in mesendoderm induction, rather than specific cholesterol and apolipoprotein activities.
Collectively, these findings from our transcriptomic analysis suggest other GPI-Aps

NCC-specific deletion of Piga completely abolishes GPI biosynthesis and leads to median cleft lip, cleft palate, and craniofacial skeletal hypoplasia
We observed cell type specific apoptosis in the cNCCs in the Clpex mutant and to further our understanding we sought to determine the cell autonomous role of GPI biosynthesis more generally in these cells. Phosphatidylinositol glycan anchor biosynthesis, class A (Piga) is part of the GPI-N-acetylglucosaminyltransferase complex that initiates GPI biosynthesis from phosphatidylinositol and N-acetylglucosamine [1]. Piga is totally required for the biosynthesis of all GPI anchors and Piga deletion totally abolishes GPI biosynthesis [11,58,59]. We first performed RNA whole mount in situ hybridization for Piga and showed it has a similar regionalized expression as we observed in the Pgap2 expression experiments. Piga expression at E11.5 is enriched in the first branchial arch, heart, limb, and CNS (representative images from n=8 antisense and 2 sense controls over 3 separate experiments) (Fig. 7A-F). However, Piga showed a unique enrichment in the medial aspect of both medial nasal processes as opposed to the Pgap2 expression which appeared to line the nasal pit epithelium (Fig. 7C). Other GPI biosynthesis genes showed a similar regionalization pattern of expression (Fig. S4).
To determine the NCC specific role for GPI biosynthesis we generated a novel model of These data are consistent with the hypothesis that GPI biosynthesis is involved in the survival of early cNCCs as we observed in the Clpex homozygous mutants and not in the later patterning or differentiation of cNCCs. The critical requirement for GPI biosynthesis appears to be at early stages of cNCC survival just after they have migrated from the dorsal neural tube, and before they have committed to differentiation to bone or cartilage.

Discussion
In this study we aimed to determine the role of GPI biosynthesis in craniofacial development with two novel models of GPI deficiency. First, we characterized the Contrary to previous studies of other GPI biosynthesis pathway genes, we found Pgap2 clearly shows enriched expression in certain tissues during certain stages of development. We observed a similar pattern in Piga RNA expression suggesting GPI biosynthesis genes share similar gene enrichment domains. These tissues are the most affected in GPI biosynthesis mouse mutants and include the craniofacial complex, CNS, limb, and heart. This may mean Pgap2 and other GPI biosynthesis genes are required in certain tissues for anchoring GPI-APs critical to that tissue. Alternatively, these areas may be particularly "GPI-rich." A variety of mutants have been described in the GPI biosynthesis pathway with a wide array of phenotypes [1,4]. While germline mutants in this pathway remain poorly understood, recent research in Paroxysmal Nocturnal Hemoglobinuria (PNH) caused by somatic mutations in PIGA has revolutionized our understanding of GPI deficiency related pathology. In PNH, clones of GPI deficient hematopoietic stem cells proliferate in the bone marrow and give rise to blood cells that lack GPI-anchored CD55/59 which are required to prevent complement-mediated lysis of red blood cells. PNH patients suffer from episodes of hemolysis and thrombosis which can be deadly [60]. Blockade of complement in these patients via eculizumab, a monoclonal antibody that inhibits the conversion of C5 to C5a, has been shown to greatly improve survival [61][62][63][64]. Thus, a single GPI-AP seems to be largely responsible for the disease observed in these patients.
In this study we aimed to identify a single GPI-AP that could be responsible for all the phenotypes observed in our germline GPI biosynthesis Clpex mutant. Of the known GPI-AP knockout models, Clpex shares the most phenotypic overlap with the Folr1 -/mouse.
We directly tested the hypothesis that FOLR1 deficiency is solely responsible for the Clpex phenotype by dietary supplementation of folinic acid during embryonic development. To our surprise, folinic acid supplementation could partially rescue the cleft lip phenotype but not the NTD or cleft palate (Fig. 6C,D). These data also argue high Clpex mutants [75]. Efna5 -/mice appear to form DLHPs though the neural folds do not fuse in the midline which is less severe than the defect we observe in Clpex mutants [22].
Therefore, we find it unlikely the loss of these GPI-APs are primarily responsible for the defects observed in the Clpex mutant although contributions to the phenotype may come from abnormal presentation of one or several of these GPI-Aps on the cellular membranes.
It has been known for decades that treatment of embryos with phospholipase C to release GPI-APs from the cell surface causes NTD in utero [76]. To investigate the cause of the NTD in Clpex mutants, we performed histological and immunohistochemical analysis of the mutant at neurulation stages. We found the Clpex mutant fails to form dorsolateral hinge points and the cranial neuroepithelium is apoptotic in the region of the developing DLHP. Neuroepithelial apoptosis was restricted to the midbrain/hindbrain boundary and likely explains why Clpex mutants develop cranial NTDs as opposed to caudal NTDs such as spina bifida. These cellular defects likely underlie the NTD but the cause of the neuroepithelial apoptosis remains unclear as the NTD did not respond to folinic acid supplementation. It remains controversial, but the NTD in Folr1 -/mice may be related to an expansion of the Shh signaling domain that patterns the neural tube [36,77]. Indeed, many Shh gain-of-function mutants develop NTD as Shh expansion impairs the formation of DLHPs and closure of the neural tube [77]. Our RNA sequencing analysis did not identify a dysregulation in the Shh signaling pathway so there are likely differences in the mechanism responsible for the NTD in Folr1 -/mice and Clpex mutants.
To determine alternative mechanisms responsible for the Clpex phenotype, we performed RNA sequencing from E9.5 WT and Clpex mutants. We found the largest differences in gene expression were in A/P patterning genes and mesendoderm induction genes. The A/P axis and induction of mesendoderm has been shown to require GPI-anchored CRIPTO, a Tgfβ superfamily member co-receptor of NODAL. A variety of studies have shown CRIPTO/Tgfβsuper family members pathway function is impaired in GPI biosynthesis mutants because CRIPTO is GPI-anchored and cleavage of the anchor affects CRIPTO function [10,52].
While GPI deficiency has been studied in the context of A/P patterning, this is the first study to implicate GPI biosynthesis in the survival of neural crest cells in a cell autonomous fashion. Indeed, the enrichment of Piga in the developing medial nasal process and the median cleft lip/cleft palate and craniofacial hypoplasia in our Piga cKOs confirms a unique cell autonomous role for GPI biosynthesis in these structures.
Interestingly, these mutants do not show a complete loss of the craniofacial skeleton, rather a general, mild hypoplasia consistent with a role for GPI biosynthesis in early NCC survival, but not later patterning or differentiation.
Our study provides potential mechanistic explanations for the developmental defects observed in a GPI biosynthesis mutant model. We propose GPI biosynthesis is involved in anchoring critical survival factors for NCCs and the neuroepithelium. In GPI deficient states, NCCs undergo apoptosis leading to hypoplastic nasal processes and palatal shelves. As we reduced the degree of GPI biosynthesis from the germline Clpex mutant hypomorph to our totally GPI deficient NCC cKO model, we observed a worsening of the  Table S1. Sample2Snp custom Taqman probes were designed by Thermo-Fisher and used to genotype the point mutation in the Clpex line.

Mapping and Sequencing
Mapping of the Clpex mutation was previously described [9]. Whole exome sequencing was done at the CCHMC DNA Sequencing and Genotyping Core. The Pgap2 exon 3 variant was Sanger sequenced using the Zymo DNA clean & Concentrator kit (Zymo Research Corporation, Irvine, CA).

Whole Mount In Situ Hybridization
RNA in situ hybridization was performed as previously described [78]. Briefly, whole E8- Mesenchymal Cells (MPNMCs) were generated from E13.5-E14.5 microdissected embryo heads in a protocol similar to that used for MEPMS [79]. The lower jaw, eyes and brain were removed and the remaining upper jaw and nasal mesenchyme were lysed in 0.25% trypsin for 10 min at 37°C, passaged through a P1000 pipette several times to create a single cell suspension, and cultured in 12 well plates. These cells displayed a stellate mesenchymal cell appearance after culture overnight. They were then stained after 72 hours from isolation with FLAER.

CRISPR Knockout/Knock-in gene editing
We utilized a double guide approach to generate knockout clones with deletions in PGAP2 and PIGA in HEK293T cells. Two small guide RNAs targeting exon 3 of either PGAP2 or PIGA were designed using Benchling software (Benchling, San Francisco, CA) and 5' overhangs were added for cloning into CRISPR/Cas9 PX459M2 puromycinresistance vector [80]. We also generated a single gRNA and donor oligonucleotide for homologous recombination to recapitulate the Clpex mutation in 293T cells (Integrated DNA Technologies ultramer). We cloned these guides into the PX459M2 plasmid using the one-step digestion-ligation with BbsI enzyme as described by Ran et. al. [81]. Two guides per gene were transfected in WT 293T cells using Lipofectamine 3000 and cells  Table S1. Sequencing of clones is presented in Fig. S1.

Histology
Whole embryos E8-E16.5 were fixed in formalin and embedded in paraffin for coronal sectioning and stained with hematoxylin and eosin using standard methods.

NCC lineage trace and Xgal staining
Clpex heterozygous females were crossed to Wnt1-Cre R26R transgenic mice as described in results. Whole embryos were fixed in 4% PFA for 15 minutes at RT, washed in lacZ buffer, and stained in a solution containing 1mg/mL X-gal (Sigma #B4252) [82].
They were washed 3 times in PBS-T and imaged after several hours in X-gal stain at room temperature.

Barrier Function assay
E18.5 embryos were dehydrated through a methanol series and then rehydrated. Next, they were placed in 0.1% Toludine Blue (Sigma #89640) in water for 2 minutes on ice.
They were destained in PBS on ice and imaged.

Statistical Analysis
Statistical analysis was performed using Graphpad Prism (GraphPad Software, San Diego, CA). All tests but diet results were unpaired, two-tailed t-tests and significance was M.L. and R.W.S. conceived, designed, and wrote the study.