Snx3 is important for mammalian neural tube closure via its role in canonical and non-canonical WNT signaling

Neural tube (NT) closure is the process by which the embryonic precursor of the mammalian brain and spinal cord is formed. It is a highly dynamic morphogenetic process that starts with the specification of the neural epithelium, which thickens to become the flat neural plate. Molecular signals from the nonneural ectoderm, surrounding mesoderm and ventral notochord, as well as forces generated by cell and tissue movements, enable the flat neuroepithelial plate to intercalate toward the midline and elongate along the anterior-posterior axis as well as form medial and lateral hinge points in the neural plate to ‘roll up’ and ultimately allow the neural folds to fuse together, forming a closed tube along the length of the rostral-caudal axis (Ray and Niswander, 2012; Wilde et al., 2014; Juriloff and Harris, 2018). This neuroepithelial tube continues to grow and differentiate into the cell types of the brain and spinal cord.

Neural tube defects (NTDs) affect 1/1000 live births worldwide, making it the second most common birth defect (Zaganjor et al., 2016). The most common types of NTDs are the cranial NTD exencephaly that ultimately results in anencephaly and the caudal NTD spina bifida or myelomeningocele (Zaganjor et al., 2016). Current treatment options for NTDs are often inefficacious and identifying methods to prevent NTDs have been the main focus of researchers. This led to public health campaigns of folic acid supplementation to reduce the risk for NTDs. However, folic acid supplementation has not eliminated NTDs, indicating our opacity in understanding the full etiology of NTDs and how to prevent them (Au et al., 2010; Wilde et al., 2014). The mouse has been invaluable in defining the genetic contributions to NT closure and, currently, over 200 mutations have been associated with NTDs in mice (a list of genes identified as required for mouse NT closure is available at sbseqconsortium.org/mouse-ntd-model). Cranial NTD is more common in mouse, whereas in the human population caudal NTDs are predominant (Harris and Juriloff, 2007; Au et al., 2010). Despite this discrepancy, investigating NTDs in the mouse model is advantageous to understand the molecular mechanisms of NT closure and genetic data from mouse has greatly informed studies to uncover the etiology of NTDs in humans.

Specifically, genes that facilitate cell-to-cell communication via intracellular signaling pathways are required for neural tube closure through bone morphogenetic protein (BMP) signaling, sonic hedgehog (SHH) signaling and wingless integrated 1 (WNT) signaling (Kanzler et al., 2000; Ybot-Gonzalez et al., 2002; Castranio and Mishina, 2009; Misra and Matise, 2010; Murdoch and Copp, 2010; Zhao et al., 2014; Allache et al., 2015; Anderson et al., 2016). WNT signaling in particular is important for neural patterning, growth, cell intercalation and neural tube closure. WNT signaling acts to induce both the canonical WNT signaling pathway for transcription of WNT target genes, such as Axin2, Lef1 and Pax3 in the neural epithelium, and the non-canonical WNT signaling pathway for cell intercalation of the neural epithelium and elongation of the body axis in a process called convergent extension (Perry et al., 1995; Carter et al., 2005; Qian et al., 2007; Gao, 2012). Many non-canonical WNTs act through planar cell polarity proteins, such as VANGL2, to initiate cellular polarity changes to induce convergent extension in the neural epithelium (Greene et al., 1998; Juriloff and Harris, 2012). Canonical WNTs bind their co-receptors [frizzled (FZD) proteins and LRP5/6], which inactivates the β-catenin destruction complex to induce WNT target gene expression for maintaining anterior-posterior and dorsal-ventral identity of the neural tube, and for neural tube closure, brain development and craniofacial development (Song et al., 2009; Wang et al., 2010; Zhao et al., 2014). Briefly, WNT signaling is the process by which post-translationally modified proteins called WNTs are secreted from cells to be received by surface receptors of the receiving cell. This results in either the specific transcription of WNT target genes (canonical WNT signaling) or activation of cytoskeletal reorganization (non-canonical WNT signaling) in a spatially and temporally regulated manner to control tissue and organ development (Nusse et al., 1991). WNT signaling is an ancient and conserved signaling system that, when disrupted, is implicated in many developmental disorders (Logan and Nusse, 2004). Both canonical and non-canonical WNT proteins require an indispensable binding protein, Wntless (WLS), in order to be trafficked to the cell surface for proper secretion and signaling to neighboring cells. The role of WLS is evolutionary conserved and in both C. elegans and D. melanogaster, the trafficking of WLS is dependent on sorting nexin (SNX) proteins (Zhou et al., 2003; Belenkaya et al., 2008; Harterink et al., 2011; Zhang et al., 2011; Yu et al., 2014; Sun et al., 2017; McGough et al., 2018).

Sorting nexins are a class of intracellular trafficking proteins characterized by the presence of the Phox homology domain (PX domain) (Haft et al., 1998; Johannes and Wunder, 2011). The PX domain binds to phospholipid bilayers within the cell such as membranes of endosomes and lysosomes, then binds specific cargo for intracellular transport to carry out specific functions (Worby and Dixon, 2002; Lenoir et al., 2018; Cui et al., 2019). This article investigates the role of Snx3 for specific transport of the WNT ligand-binding protein WLS to enable NT closure. A specific role for SNX3 in mammalian development has not been studied.

The intracellular trafficking machinery that facilitates cellular signaling during NT closure has been explored in a few cases. For example, SEC24B is a cargo-sorting protein that specifically sorts VANGL2 into COPII-coated vesicles to be trafficked from the endoplasmic reticulum to the Golgi body (Merte et al., 2010; Yang et al., 2013). Mutations in SEC24B are present in human cohorts with NTDs, highlighting the importance of this cargo-sorting protein for correct NT closure. Although disruption of intracellular trafficking can disrupt embryonic development, whether specific cargo and specific trafficking events are required for NT closure and development is not fully understood. Thus, understanding the biology of NT closure at the cellular level, as investigated here, is crucial to adopting a prevention-based method of treatment to decrease the risk and incidence of NTDs.

Recently, the Knockout Mouse Phenotyping Program (KOMP2) created a null allele of Snx3, a sorting nexin gene that is crucial for intracellular retrograde trafficking and recycling (Haft et al., 1998). Snx3^−/− KO mice show cranial NTDs, brain and craniofacial deformities, and embryonic lethality (mousephenotype.org/data/genes/MGI:1860188). NTDs are currently not associated with retrograde trafficking defects, which makes it interesting to explore the function of SNX3 in the process of neural tube closure.

Here, we investigate the role of SNX3 in the pathogenesis of NTDs. Our data show that Snx3^−/− embryos display both neural patterning and convergent-extension defects, suggesting disruption in WNT signaling. Using both in vivo and in vitro approaches, we show that SNX3 is essential for NT closure via recycling of the WNT ligand-binding protein WLS to promote both canonical and non-canonical WNT signaling. In the absence of SNX3, WLS is targeted for degradation in the lysosome and the level of WNT signaling is reduced. Moreover, we find that a SNX3 mutation identified in an NTD-affected individual disrupts WLS recycling. Rescue of the NTD phenotype in mouse embryos treated with a WNT agonist further highlights the importance of SNX3-WLS trafficking to maintain a proper balance in WNT signaling during NT closure.

The KOMP2 program at The Jackson Laboratory generated the Snx3tm1.1(KOMP)Vlcg (MGI:1860188) allele following microinjection of embryonic stem (ES) cells targeted by the Velocigene program at Regeneron Pharmaceuticals as part of the KOMP program (clone 155528-B8). This deletion generated a null allele of Snx3. Preliminary studies by the KOMP2 program showed that heterozygous Snx3^−/+ mice are phenotypically normal; however, intercrosses to produce animals for the adult and embryonic phenotyping pipeline (PMID:28650483, PMID:27626380) showed Snx3^−/− embryos at normal Mendelian ratios with death by birth (Table 1). Snx3^−/− null embryos displayed multiple severe phenotypes, including cranial NTD. We characterized the NTD phenotype starting at embryonic day 8.5 (E8.5) at the beginning of neurulation, and at E9.5, E10.5 and E15.5 (Fig. 1). Wild-type and mutant embryos were somite matched, and we observed consistent developmental disruptions in the Snx3^−/− embryos beginning with E8.5. Although the Snx3 KO embryos were much shorter than the wild-type littermates, they were not developmentally delayed and somite number did not vary significantly within a litter. At E9.5 and E10.5, 100% of Snx3^−/− embryos exhibited cranial NTD, ultimately leading to anencephaly with the brain tissue exposed at E15.5 (n=129 Snx3 mutants observed). At all stages, Snx3^−/− embryos were reduced in size compared with somite-matched wild type (Fig. 1). All Snx3^−/− embryos also show severe craniofacial defects. Staining of E9.5 and E12.5 heterozygous embryos to detect LacZ driven by Snx3 regulatory elements showed strongest expression at E9.5 in the neural epithelium, craniofacial region and heart (Fig. S1A,C,D), with ubiquitous expression at E12.5 (Fig. S1B). In terms of neurulation, only cranial NTD is observed: the reason for this restricted phenotype is unclear, as LacZ expression was detected in the neural epithelium along the rostral-caudal axis. This could reflect a more-complex morphogenesis of the cranial neural folds or perhaps alternative pathways that can compensate for Snx3 loss during spinal neural tube closure. These data indicate that Snx3 is crucial for cranial neural tube closure and embryonic development in mice.

The shortened length of Snx3^−/− embryos resembled mice with mutations that disrupt the non-canonical WNT planar cell polarity pathway (Ikeya et al., 1997; Wang, 2006; Qian et al., 2007). This led us to ask whether WNT signaling was disrupted in Snx3^−/− embryos. To address this issue, we first performed RNA in situ hybridization to assess the expression of common canonical WNT target genes, Axin2, Lef1 and Pax3. These genes are expressed in the anterior cranial neural folds of E8.5 wild-type embryos (Fig. 2A,B,E,F,I,J). However, in the somite-matched Snx3^−/− E8.5 whole embryos, expression of Axin2, Lef1 and Pax3 was greatly decreased in the anterior cranial neural folds (Fig. 2C,D,G,H,K,L). Furthermore, we used a TCF/Lef:H2B-GFP transgenic reporter mouse to visualize WNT signaling activity in the cranial neural folds and observed a significant decrease in GFP in the neural epithelium of Snx3^−/− embryos compared with their somite-matched wild-type littermates (Fig. 2M-P, Fig. S2A-C). These data together indicate an overall significant decrease in WNT signaling and target canonical WNT gene expression in the cranial neural folds of Snx3^−/− embryos, including expression of Pax3, which is required for NT closure (Zhao et al., 2014).

There was also a striking difference in the overall shape of the E8.5 Snx3^−/− embryos compared with wild type. The Snx3^−/− embryos appeared much shorter and wider than their somite-matched wild-type littermates. This phenotype is similar to a disruption in non-canonical WNT signaling (planar cell polarity), which leads to defective convergent extension of the neural epithelium and NTDs. A quantitative assessment for a convergent extension defect at early embryonic stages is to calculate the length-to-width ratio of the mouse embryo (Ybot-Gonzalez et al., 2007). Therefore, we quantified length-to-width ratios on all somite-matched wild-type and Snx3^−/− E8.5 embryos and found a statistically significant smaller length-to-width ratio of Snx3^−/− compared with wild-type embryos (Fig. 3). Taken together, the lack of expression of canonical WNT target genes Axin2, Lef1 and Pax3, decreased GFP expression using the TCF/Lef:H2B-GFP transgenic reporter, and the decreased length-to-width ratio in the Snx3^−/− embryos indicate that both canonical WNT signaling and non-canonical WNT signaling are disrupted in embryos that lack Snx3 function.

As Snx3 mutant embryos display the hallmarks of disrupted canonical and non-canonical WNT signaling, we next asked whether loss of Snx3 results in reduced WNT protein secretion. To test this, we compared wild-type and Snx3^−/− mouse embryonic fibroblasts (MEFs) using enzyme-linked immunosorbent assays (ELISA) with cell culture media collected from identically seeded and processed in parallel wild-type or Snx3^−/− MEFs transfected with canonical Wnt1 or non-canonical Wnt5a expression constructs. This showed lower WNT1 and WNT5A protein secretion into the media from the Snx3^−/− cells compared with wild type (Fig. 4). To directly test that the secretion defect was solely due to the lack of Snx3, we added a plasmid encoding wild-type SNX3 to the Snx3^−/− cells. This showed a restoration of WNT1 and WNT5A secretion into the media (Fig. 4), indicating the importance of SNX3 in maintaining proper levels of WNT secretion. These data indicate a defect in secretion of both canonical and non-canonical WNTs in the absence of Snx3 and aligns with our in vivo results.

The data presented so far provide evidence that Snx3 is important for NT closure via a role in regulation of canonical and non-canonical WNT signaling. The mechanism of action of the SNX3 homolog in C. elegans and D. melanogaster is in the intracellular retrograde trafficking of WLS, the WNT ligand binding protein (Harterink et al., 2011; Zhang et al., 2011; McGough et al., 2018). In these contexts, SNX3 transports the endocytosed ligand-free WLS from the early endosome of the cell to the trans-Golgi network, where WLS binds WNT to facilitate the process of normal WNT secretion (Johannes and Wunder, 2011). Without SNX3, WNT secretion decreases in both C. elegans and D. melanogaster (Belenkaya et al., 2008; Harterink et al., 2011; Johannes and Wunder, 2011; Zhang et al., 2011). To determine whether SNX3 assumes a similar role in trafficking WLS in mammals, we evaluated the colocalization and relationship between WLS and SNX3. We performed live cell imaging of wild-type MEFs transfected with plasmids to express SNX3-GFP and WLS-RFP fusion proteins. This showed that SNX3 colocalizes with WLS in mouse cells (Fig. 5A-B). Furthermore, SNX3 and WLS traffic together in real time. Trafficking of SNX3-GFP and WLS-RFP was captured over 10 s intervals, showing both proteins budded off the SNX3-GFP labeled early endosome together (Fig. 5C). In the wild-type setting we would predict WLS to localize with the early endosome for recycling, rather than being trafficked to the lysosome. In the absence of SNX3, we hypothesized that WLS would be mis-trafficked to the lysosomal pathway for degradation, thus disrupting the recycling of WLS and leading to dysregulation of the tightly regulated WNT secretion cycle. To test whether the lack of SNX3 results in WLS being mis-trafficked to the lysosome, plasmids coding for fusion proteins to tag the early endosome (RAB5-BFP) or the lysosome (LAMP1-GFP), as well as WLS-RFP, were transfected into wild-type and Snx3^−/− cells. We then calculated the colocalization of WLS with the early endosome or the lysosome. We observed that WLS predominantly colocalizes with the early endosome in wild-type cells, whereas WLS predominantly colocalizes with the lysosome in Snx3^−/− cells (Fig. 5D-G″,J,J′,K,K′,M). To confirm that this mis-sorting of WLS to the lysosome was due to the absence of SNX3, we introduced exogenous SNX3 into Snx3^−/− cells and saw that the localization of WLS was restored to the early endosome (Fig. 5H-I,L,L′,M). Higher magnification of the endosomes (RAB5-BFP labeled) and lysosomes (LAMP1-BFP) showed that WLS predominantly colocalizes to the early endosome in wild-type cells, the lysosome in Snx3^−/− cells, and endosome localization is rescued in Snx3^−/− cells transfected with exogenous Snx3 plasmid (Fig. 5J-L′).

SNX3 specifically localizes and binds to phosphatidyl inositol-3-phosphate (PI3P) predominantly at early endosomal membranes (Lenoir et al., 2018). The early endosome is the site of protein sorting for recycling either to the plasma membrane or trans-Golgi network, or for lysosomal degradation (Delevoye et al., 2014; Cullen and Steinberg, 2018; Wang et al., 2018; Singla et al., 2019). Thus, if lack of Snx3 results in WLS being mis-sorted to the lysosome followed by lysosomal degradation, we would predict to observe less WLS-RFP in the cells lacking Snx3. Consistent with this hypothesis, our live-cell imaging data show a reduced level of WLS-RFP protein in Snx3^−/− cells compared with wild-type cells (Figs 5D,E,F,G and 6A,B,E). To evaluate further whether WLS is being degraded without Snx3, we treated the cells with bafilomycin, a lysosomal degradation inhibitor. Quantification of WLS in wild-type and Snx3^−/− cells treated with bafilomycin showed an increase in the WLS levels when lysosomal degradation was inhibited (Fig. 6C-E). There was a small, but not significant, increase in the treated wild-type cells compared with untreated wild-type cells (Fig. 6C-C′,E), consistent with the idea that this recycling pathway is in a steady state of retrograde trafficking regulation from the early endosome to the trans-Golgi or the lysosome. In the Snx3^−/− cells, we observed a striking increase in WLS levels in the bafilomycin-treated group compared with the untreated group, and the WLS-RFP was largely in LAMP1-positive lysosomal vesicles (Fig. 6D-E). Furthermore, we observed statistically similar WLS levels in bafilomycin-treated Snx3^−/− cells as in both treated and untreated wild-type cells (Fig. 6E). This indicates that the WLS levels are restored to wild-type levels when lysosomal degradation is halted. Taken together, these data are consistent with the defective recycling of WLS in Snx3-deficient cells and suggest that SNX3 is crucial in the trafficking of WLS from the early endosome for recycling in order to bind another WNT protein and facilitate another cycle of WNT secretion. Furthermore, in the absence of SNX3, WLS is trafficked to the lysosome for degradation, thus reducing WNT signaling on a cellular level.

The protein sequence of SNX3 is 100% conserved from mice to humans (NP_059500.2 and NP_00376.1). This and the strong NTD phenotype in mice led to the question of whether sequence variation in the SNX3 gene may genetically contribute to NTDs in humans. Whole-exome sequencing of a cohort of 511 myelomeningocele-affected sporadic individuals identified a missense point mutation (GRCh37/hg19; 6:108582021:G:T) in the SNX3 gene in one individual with a NTD. Specifically, the subject is heterozygous for the SNX3 p.Asn35Lys variant. The overall dataset cannot be released due to HIPAA Privacy Rules; however, the subject exhibited myelomeningocele and was heterozygous for the SNX3 variant as well as another heterozygous variant in the CITED2 gene (chr6:139695005:T:C; p.His31Arg), which is also associated with NTD. The SNX3 p.Asn35Lys variant in this subject resulted in a G-to-T transversion in exon 1 (ORF pos. 105), creating an asparagine-to-lysine amino acid change (Fig. 7A). This is a novel point mutation that has not been reported in the literature or public variant databases. The Combined Annotation Dependent Depletion (CADD) score for this point mutation is 25.8, predicting the change to be deleterious to SNX3 function. To test whether this specific point mutation results in a functionally impaired SNX3, we performed site-directed mutagenesis to generate SNX3-p.Asn35Lys-GFP mutant plasmid, then transfected it into Snx3^−/− MEFs along with Wls-RFP. Colocalization analysis indicated that there was significantly less SNX3-p.Asn35Lys-GFP colocalized with WLS-RFP compared with wild-type SNX3 (Fig. 7B-D). Furthermore, our live-cell imaging showed reduced WLS fluorescence, consistent with our results above that loss of SNX3 function results in aberrant trafficking to the lysosome and degradation (Fig. 7D′). Thus, we treated the SNX3-p.Asn35Lys-GFP cells with the lysosomal inhibitor bafilomycin to determine whether WLS was being degraded in the context of the human mutation. We observed that WLS fluorescence was significantly increased upon treatment with bafilomycin in the SNX3-p.Asn35Lys-GFP cells compared with the untreated SNX3-p.Asn35Lys-GFP cells (Fig. 7E,F). These data indicate that this novel genetic variant identified in an NTD-affected individual could functionally impair SNX3 protein and result in WLS degradation and reduced WNT signaling.

Our data show that the absence of Snx3 results in disrupted WLS recycling and reduced WNT secretion and signaling, and leads to NTD in mice. This suggests that it may be possible to rescue the NTD in the Snx3^−/− mice by pharmacologically stimulating the WNT signaling pathway downstream of WNT reception. To do so, we used lithium chloride (LiCl), which inhibits the phosphorylation and degradation of β-catenin, allowing β-catenin to translocate to the nucleus and bind TCF/LEF transcription factors to induce WNT target gene expression, in the absence of WNT signaling (Klein and Melton, 1996). As LiCl stimulates the canonical WNT signaling pathway, we predicted that the NTD would be only partially rescued. However, as the canonical and non-canonical WNT signaling pathways frequently overlap and interact, it is possible that both canonical and non-canonical WNT signaling may be stimulated by LiCl (Komiya and Habas, 2008). Thus, we used a LiCl treatment regime similar to others (Wang et al., 2010) in which pregnant dams received intraperitoneal injections of 300 µl of LiCl on E7.5 and E8.5, and embryos were dissected at E10.5. These injection time points were chosen to stimulate the WNT pathway at the time of neural induction and at the start of cranial neurulation. LiCl treatment did not affect wild-type embryos (Fig. 8A, Fig. S3A-C) but strikingly rescued the NTD in 50% of the Snx3^−/− embryos dissected (Fig. 8B, Fig. S3D-F). The other 50% of the Snx3^−/− embryos dissected from these litters with LiCl treatment still showed cranial NTD (Fig. 8C, Fig. S3G-I). These data indicate that SNX3 is crucial for NT closure and that the key function of SNX3 at this time of development is in the regulation of WNT signaling.

Here, we show that SNX3 is crucial for mammalian neural tube closure. SNX3 functions to recycle the WNT ligand-binding protein WLS to regulate and facilitate additional cycles of canonical WNT secretion for the expression of downstream WNT target genes, as well as non-canonical WNT secretion for convergent extension in the neural epithelium for NT closure. In the absence of SNX3, WLS is mis-trafficked to the lysosomal pathway and is degraded, disrupting the tight regulation of WNT secretion. Furthermore, we show that the NTD can be fully rescued in 50% of Snx3^−/− embryos by pharmacologically stimulating the WNT signaling pathway. Our studies demonstrate that intracellular trafficking of WLS via SNX3 is essential to maintain proper levels of WNT signaling during mammalian development.

Snx3^−/− embryos show 100% penetrant cranial NTD, explored here, as well as brain and craniofacial phenotypes, reduced size, and embryonic lethality. However, as we have shown that the NTD phenotype is due to reduced WNT signaling, it is curious that embryonic development is not more extensively disrupted given the fact that WNT signaling is important for so many aspects. A complete knockout of Wls in the mouse results in embryonic lethality at gastrulation because it stops all WNT secretion (Fu et al., 2011). It is important to note that SNX3 regulates WLS recycling. As WLS and WNT are still present and able to interact for at least one round of secretion and signaling, loss of SNX3 would not abolish WNT signaling, as in the Wls KO, but would be expected to decrease the overall level of WNT signaling. Moreover, at the time of neurulation, Snx3 is predominantly expressed in the neural epithelium and craniofacial tissues. These data indicate that SNX3 is crucial for some aspects of WLS function but in other tissues Snx3 is not essential or is perhaps compensated for by another SNX protein, as there are 25 mammalian sorting nexins, each with specific cargo they traffic and a specific cellular localization (Worby and Dixon, 2002). Relative to NT closure, it is possible that the cranial region of the embryo requires higher levels of WNT signaling facilitated by Snx3 to produce specific developmental outcomes, whereas the caudal region of the embryo might use lower levels of WNT or is better able to compensate for the decrease in WNT regardless of the lack of Snx3, perhaps by simply upregulating Wls. Furthermore, WNTs are tightly spatially and temporally regulated, and it is possible that the cranial region of the embryo at the early developmental time points is particularly sensitive to alterations in WNT signaling compared with the caudal region of the embryo. Additionally, SNX3 in the context of non-canonical WNT signaling in particular has not been previously reported. The Snx3^−/− phenotype has provided a new viewpoint into WLS function during NT closure.

The tyrosine residues in the PX domain of SNX3 serve to localize SNX3 to the early endosome through binding of the glycerol moiety in PI3P, which is enriched at the early endosomal membrane (Xu et al., 2001; Lucas et al., 2016). WLS binds the retromer subunit VPS26, which, in turn, is bound to SNX3 through the N-terminal tail of VPS26 (Lucas et al., 2016). SNX3 undergoes conformational changes upon binding the retromer complex (VPS26-VPS35-VPS29) that then allows binding of cargo such as WLS to the retromer complex (Lucas et al., 2016). Given the fact that SNX3 undergoes conformational changes based on its binding partners, we hypothesize that the asparagine to lysine amino acid mutation at position 35 in the PX domain of the human protein affects SNX3 conformation and disrupts WLS binding to the retromer subunit VPS26. We show the SNX3 p.Asn35Lys mutation results in disrupted colocalization with WLS, such that WLS is mis-trafficked to the lysosome, similar to what we observed in Snx3 loss-of-function MEFs, instead of the normal recycling of WLS back to the trans-Golgi network for further binding of WNT. Loss of Snx3 in mouse cells and the patient SNX3 p.Asn35Lys allele both lead to abnormal WLS degradation. In mice, we show that loss of Snx3 results in decreased WNT signaling during neurulation and leads to NTD. Thus, the SNX3 p.Asn35Lys mutation could contribute to the NTD phenotype in humans. Furthermore, asparagine at position 35 is a weak nucleophile and would not be expected to be post translationally modified, whereas the point mutation introduces a lysine, which often can be post-translationally modified. Possible post-translational modifications that could be added to the lysine may affect the conformation of SNX3 and also decrease its functionality. Further studies are needed to understand how this point mutation affects the function of SNX3 and distribution of WLS; however, these data provide further evidence to support our hypothesis that Snx3 is crucial for mammalian neural tube closure because of its colocalization and trafficking roles with WLS.

The fact that the NTD in Snx3^−/− embryos can be rescued with LiCl shows that WNT is the key mechanistic link between SNX3 and neural tube closure in the mouse model. It is important to note, however, that the NTD was not rescued in all Snx3^−/− embryos by LiCl treatment. This could suggest the importance of the timing and dosage of treatment, given the gradients of WNT as a morphogen, resulting in rescue rates that differ among litters. In addition, the varying rescue rates could be reflective of the inability to sufficiently rescue non-canonical WNT signaling required for NT closure. Alternatively, it is likely that there are other factors involved in NT closure that SNX3 regulates apart from WNT. As trafficking proteins rarely traffic only one cargo, it is possible that there are other factors involved in the NT closure process that are disrupted in the absence of Snx3. For example, in erythroid cells, SNX3 also traffics the transferrin receptor (TFRC), which is needed for iron import (Chen et al., 2013). Furthermore, epidemiological studies and studies in mouse models indicate that iron is a crucial micronutrient for embryonic development and neural tube closure (Milunsky et al., 1989; Zohn et al., 2007; Stokes et al., 2016; Li et al., 2018). Perhaps the Snx3^−/− NTD phenotype could be rescued more effectively with combined WNT stimulation and iron supplementation.

Overall, we have discovered a previously unreported mechanism by which NT closure is regulated in the mammalian model. The SNX3 p.Asn35Lys mutation significantly affects SNX3 function and, although this was a heterozygous mutation in the NTD-affected individual, the exquisite sensitivity of NT closure to the levels of WNT signaling suggest this could be a contributing factor to NTD risk. Further studies are required to understand the full role of SNX3 in NT closure and the genetic intersection between Snx3 and other genes that contribute to this.

All mouse work was carried out in accordance with University of Colorado Boulder Institutional Animal Care and Use Committee (IACUC) policies and procedures, and animals were maintained under standard light and ambient conditions. The genetic background of all mice was C57BL/6J. One male Snx3^−/+ and one female Snx3^−/+ mouse was obtained from the KOMP2 program at The Jackson Laboratory, which generated the Snx3tm1.1(KOMP)Vlcg (MGI:1860188) allele following microinjection of embryonic stem (ES) cells targeted by the Velocigene program at Regeneron Pharmaceuticals as part of the KOMP program (clone 155528-B8). Following germline transmission of the targeted allele, the mice were bred to a B6N congenic Sox2-cre line [Edil3^{Tg(Sox2−cre)1Amc}/J; JAX STOCK# 14094] to remove the neo selection cassette. The allele consists of 32,453 bp deletion in the Snx3 gene, replaced by a LacZ cassette, resulting in a complete knockout [allele is B6N(Cg)-Snx3^{tm1.1(KOMP)Vlcg}/J, MGI ID is MGI:1860188]. The colony was expanded by crossing littermates and heterozygous males and females were used to generate Snx3^−-/− (KO) embryos. Vaginal plugs were checked in the morning for timed mating, with noon on the day of positive identification of a plug considered to be embryonic day 0.5. Genotyping was performed using a tail clip or yolk sac digested in lysis buffer and proteinase K overnight at 56°C, and heat inactivated for 10 min at 95°C. Samples were diluted 2:100 and combined with DMSO (New England Biosciences), Amplitaq Gold (ThermoFisher Scientific, N8080241), PCR6 buffer, and forward and reverse primers (IDT) for wild-type or Snx3 KO, and run on the thermal cycler program 95°C for 5 min followed by 35 cycles of 95°C for 30 s, 60°C for 30 s and 72°C for 1.5 min, the 72°C for 5 min. Samples were run on a 1.5% agarose gel and imaged on the Bio-Rad gel imager system. Bands at 359 base pairs indicated wild type, bands at 576 bp indicated KO, and the presence of both bands indicated heterozygous genotype. Primers were as follows: wild-type Snx3 forward primer sequence, 5′-TTG GTC TCT TGT CCC ACC TT-3′; wild-type Snx3 reverse primer sequence: 5′-CTG TAG AGG ACC TGG GTT CG-3′; Snx3 KO forward primer sequence: 5′-TGT CAA CTG CCA AGA CAA GC-3′; and Snx3 KO reverse primer sequence: 5′-CGG TCG CTA CCA TTA CCA GT-3′.

Dissections were carried out according to NIH guidelines and IACUC protocol at various time points. To observe the Snx3^−/− phenotype, heterozygous parents were crossed and embryos were dissected at E8.5, E9.5, E10.5 and E15.5, and fixed in 4% paraformaldehyde (PFA) at 4°C overnight then washed in 1× phosphate-buffered saline (PBS Thermo Fisher, 10010049). Embryos were imaged on the Olympus IX83 Fluorescence Microscope with a 4× objective (Figs 1 and 7) or the Nikon SMZ18 Stereoscope with Nikon DS-Ri1 CCD-cooled camera with a 3× objective (Fig. 2) and analyzed using Image J.

Embryos were dissected at E8.5, and embryo phenotypes were recorded at the time of dissection while within the yolk sac, ensuring no mechanical forces affected the shape of the embryos apart from the Snx3-related phenotype. Embryos were fixed in 4% PFA prior to processing for whole-mount in situ hybridization on somite-matched embryos as previously described (Holmes and Niswander, 2001). Probes used were: mouse Axin2, mouse Lef1 and mouse Pax3 (all probes generously provided by L.A.N.’s lab). The RNA in situ hybridization probes used were digoxigenin-labeled probes recognized by anti-digoxigenin antibodies conjugated with alkaline phosphatase (Sigma-Aldrich, 11093724910). Upon incubation with the chromogenic substrate, purple color develops that in turn indicates the expression of the specific gene in that tissue (BM-purple, Sigma-Aldrich, 11442074001). Photos were taken using the Nikon SMZ18 Stereoscope with Nikon DS-Ri1 CCD-cooled camera with the 3× objective.

β-Galactosidase staining of whole-mount embryos was performed as previously described (Heffner et al., 2012). Briefly, following fixation, E9.5 and E12.5 embryos were washed and permeabilized in detergent and incubated overnight at 37°C in 1 mg/ml X-gal staining solution, then fixed in 4% PFA in 1× PBS at 4°C overnight. X-gal-stained Snx3^+/− embryos were imaged on a Leica M125 stereoscope using a 1× Planapo objective. E9.5 embryos were then embedded in OCT, sectioned at 20 µm and re-imaged using the Olympus IX83 Fluorescence Microscope.

B6N(Cg)-Snx3^{tm1.1(KOMP)Vlcg}/mice were crossed with the STOCK Tg(TCF/Lef1-HIST1H2BB/EGFP)61Hadj/J (Stock #013752) WNT reporter strain to produce heterozygous Snx3 knockout animals carrying the H2B-EGFP transgene. Female Snx3^+/− animals were crossed with Snx3^+/−/H2B-EGFP^Tg/0 males. Embryos were dissected at E8.5 and fixed in 4% PFA for 30 min, rinsed in 1× PBS and genotyped. Snx3^+/+/H2B-EGFP^Tg/0 and Snx3^−/−/H2B-EGFP^Tg/0 embryos were imaged using a Leica M165 FC Fluorescence Microscope with a 1× Planapo objective. A 350 ms exposure time and gain setting of 1 were used for image capture. E8.5 embryos were then embedded in OCT, sectioned at 20 µm and imaged on the Nikon Structured Illumination Super-resolution and A1 Laser Scanning Confocal Microscope with a 20× objective.

Embryos were carefully dissected at E8.5, fixed in 4% PFA overnight, rinsed in 1× PBS, imaged on Olympus IX83 Fluorescence Microscope with the 4× objective or the Nikon SMZ18 Stereoscope with Nikon DS-Ri1 CCD-cooled camera with the 3× objective, and measured using ImageJ.

Heterozygous B6N(Cg)-Snx3tm1.1(KOMP)Vlcg/J were housed together and observance of the vaginal plug was considered to be 0.5 days post conception. Embryos were dissected from the dam at embryonic day 11.5 and 12.5 to make mouse embryonic fibroblast (MEF) primary culture cells. The head and internal organs were discarded, and the remaining embryonic tissue was dissociated in 3 ml 0.25% Trypsin/EDTA 1× (Gibco Life Technologies, 25200-056) at 37°C for 5 min. 5 ml of the total mixture of DMEM/10% FBS/1% GlutaMAX/1% Pen Strep (Gibco Life Technologies, 11995-065, A3160601, 35050061, 15140163) was added to each sample and spun down at 1968 g for 5 min. Supernatant was aspirated off and the pellet was resuspended in 3 ml of a mixture of DMEM/10% FBS/1% GlutaMAX/1% Pen Strep and plated into a 12-well dish or a 6-well dish (Costar, 3513 or 3516), plating one embryo per well. Cells were incubated overnight and media was changed the following day. Cells were expanded into larger dishes until they reached passage 2 or 3. Yolk sacs from each embryo were genotyped as described above and cells were stored frozen in DMEM/10%FBS/1%GlutaMAX/1%Pen Strep with 10% DMSO (Sigma-Aldrich, 472301) in liquid nitrogen.

Snx3-gfp was made using mouse-tagged ORF clone (NM_017472, Origene, MR215963) and pCMV6-AC-GFP (Origene, PS100010) both separately digested with SgfI and Mlul (New England Biosciences) to extract the Snx3 reading frame and linearize, respectively. Digests were run on a 1% agarose gel and the Snx3 insert (495 bp) and linearized pCMV6-AC-GFP (6551 base pairs) were gel extracted using Zymoclean Gel DNA Recovery Kit #D4008) and ligated together in a 3:1 ratio using Takara Clontech DNA ligation kit version 2.1 (6022). The ligation reaction was transformed using Takara Stellar competent cells (636736), plated on ampicillin-resistant plates, and single clones were picked to inoculate cultures overnight and then midi-prepped using Zyppy Plasmid Midiprep kit (D4200). Plasmids were sequenced by Quintarabio using in house VP1.5 and XL39 primers. Snx3-rfp was made following the same protocol, except using pCMV6-AC-RFP (Origene, PS100034).

SNX3-p.Asn35Lys-GFP was made via site-directed mutagenesis of the previously described Snx3-gfp plasmid using QuickChange Lightning Multi Site-Directed Mutagenesis Kit (Agilent Technologies #2105-15-5) according to the manufacturer's protocol. Plasmids were sequenced by Quintarabio using in house VP1.5 and XL39 primers to confirm successful mutagenesis: mutagenesis forward primer, 5′-TCG CGA TCG ACG TGA GCA AAC CGC AGA CCG-3′; mutagenesis reverse primer, 5′-CGG TCT GCG GTT TGC TCA CGT CGA TCT CGA-3′.

Wls-rfp was made using mouse tagged ORF clone (NM_026582, Origene, M208639) and pCMV6-AC-RFP (Origene, PS100034) both separately digested with SgfI and Mlul (New England Biosciences) to extract the Wls reading frame and linearize, respectively. Digests were run on a 1% agarose gel and the Wls insert (1623 bp) and linearized pCMV6-AC-RFP (6600 bp) were gel extracted, ligated and clones selected as above. Plasmids were sequenced by Quintarabio using in house VP1.5 and XL39 primers.

Rab5-BFP, Lamp1-BFP and Lamp1-GFP were generously gifts from the lab of Dr Gia Voeltz (University of Colorado at Boulder, CO, USA). Mouse Wnt1-gfp ORF tagged clone was from Origene (MG205702). Mouse Wnt5a-rfp was generously provided by Bin Li (University of Colorado at Boulder, CO, USA).

50,000 MEFs were plated in 24-well plates (Costar #3526). Wild-type and Snx3^−/− cells were transfected with 0.5 µg Wnt1-gfp plasmid and 0.5 µg pCMV6-AC-RFP as a transfection control. For the rescue, Snx3^−/− cells were transfected with 0.5 µg Snx3-rfp and 0.5 µg Wnt1-gfp. Wild type represents three biological replicates containing respective technical replicates of n=6, n=3 and n=2, for a total of n=11 data points. Snx3^−/− represents three biological replicates containing respective technical replicates of n=2, n=3 and n=5, for a total of 10 data points. Snx3^−/−+wild-type Snx3 (rescue) represents three biological replicates containing respective technical replicates of n=2, n=3 and n=6, for a total of 11 data points.

50,000 MEFs were plated in 24-well plates (Costar, 3526). Wild-type and Snx3^−/− cells were transfected with 0.5 µg Wnt5a-rfp plasmid and 0.5 µg pCMV6-AC-GFP as a transfection control. For the rescue, Snx3^−/− cells were transfected with 0.5 µg Snx3-gfp and 0.5 µg Wnt5a-rfp. Wild type represents three biological replicates containing respective technical replicates of n=8, n=2 and n=2, for a total of n=12 data points. Snx3^−/− represents four biological replicates containing respective technical replicates of n=8, n=2, n=2 and n=3 for a total of 15 data points. Snx3^−/−+wild-type Snx3 represents three biological replicates containing respective technical replicates of n=1, n=2 and n=3, for a total of six data points.

All cells were seeded at identical cell densities and transfected with equal amounts of plasmid from master mix and performed in parallel. All transfections were performed using Xfect transfection kit according to manufacturer's protocol (Takara Clontech, 631317). Media were collected 24 h post-transfection using Eppendorf LoBind Microcentrifuge Tubes (22431081). Media were spun down for 10 min at 27,670 g and the supernatant was transferred to a new LoBind tube and diluted 1:3 with the sample diluent provided in the respective kit. The media from each condition was subject to the ELISA kit according to the manufacturer's protocol (mouse WNT1 ELISA, Antibodies Online ABIN1153029; mouse WNT5A ELISA, Antibodies Online, ABIN837310). The absorbance was read at 450 nm using the Synergy 2 Multi-Mode Microplate Reader. Graphs show three independent experiments and all individual data points.

Cells from embryos were isolated as described above, then counted using a hemocytometer, 300,000 cells seeded into 35 mm glass-bottom plates (MatTek, P35G-1.5-14-C) and incubated for 24 h. Wild-type MEFs were transfected with Snx3-gfp+Wls-rfp, Rab5-bfp+Wls-rfp or Lamp1-bfp+Wls-rfp. Snx3^−/− MEFs were transfected with Rab5-bfp+Wls-rfp or Lamp1-bfp+Wls-rfp. For the rescue condition, Snx3^−/− MEFs were transfected with Snx3-gfp+Rab5-bfp+Wls-rfp or Snx3-gfp+Lamp1-bfp+Wls-rfp. For the transfection of the plasmid containing the human mutation, Snx3^−/− MEFs were transfected with SNX3-p.Asn35Lys-GFP+Wls-rfp. Up to 3 µg total plasmid was transfected for each condition. All cells treated with bafilomycin were seeded and processed identically up to the drug treatment step. All transfections were carried out using the Xfect transfection kit according to manufacturer's protocol (Takara Clontech, 631317). For the Bafilomycin treatment, cells were cultured in 100 nm bafilomycin A1 (Fisher Scientific, 501031771) for 3 h and immediately imaged. All cells were imaged 24 h post-transfection, in a humidified chamber with 5% CO₂ on the Cell Voyager CV1000 scanning disk confocal system at the University of Colorado Boulder microscopy core. Cells were imaged using the 40× oil immersion lens (0.1915 μm/pixel), with image acquisition every 10 s, up to 100 times for each cell. The excitation, exposures and gain were set within the same range of values across biological samples. Statistics represent three independent experiments of three biological replicates containing three technical replicates for each condition. Colocalization analysis and total fluorescence quantification was performed using ImageJ.

Whole exome sequencing of a cohort of 511 myelomeningocele affected sporadic individuals recruited from spina bifida clinics from multiple sites in Texas, California, Kentucky and Ontario underwent whole-exome sequencing in order to identify potential mutations that could contribute to the NTD. Whole-exome sequencing was performed using the Ion Proton platform following the manufacturer's standard protocols. The total read depth was 62 with 37 Ts and 25 Cs. Torrent Suite software was used for sequence mapping and sequence analysis with reference to GRCh37/hg19 (Life Technologies/Fisher Scientifics). Variants that passed standard Variant Quality Score Recalibration filters following the Genome Analysis Toolkit HaplotypeCaller best practices workflows were retained for further analyses (Van der Auwera et al., 2013). Additional hard filters (alternate allele depth ≥25% and genotype quality ≥20, excess heterozygosity ≤56) were applied to minimize false-positive variant calls. A missense point mutation identified was in the SNX3 gene in one individual resulting in a G-to-T transversion in exon 1 (GRCh37/hg19 6:108582021:G:T with G allele read depth of 49 and T allele read depth of 40), creating an asparagine-to-lysine amino acid change (p.Asn35Lys).

Heterozygous male and females were housed together, and observance of the vaginal plug was considered 0.5 days post-conception. Pregnant females were intraperitoneally injected with 300 µl of 150 mM lithium chloride at E7.5 and E8.5. Embryos were dissected at E10.5 and washed in 4% paraformaldehyde at 4°C overnight. The embryos were genotyped and imaged using the Olympus IX83 Fluorescence Microscope. Three independent litters were dissected.

Total GFP fluorescence in the TCF/Lef:H2B-GFP reporter wild-type and Snx3^−/− E8.5 embryos was measured using ImageJ and background was subtracted. An unpaired t-test with Welch's correction was performed to test for significance. Graphs show individual data points and mean±s.e.m. plotted using Graphpad Prism software.

Length-to-width ratios were calculated using ImageJ with the ‘line draw’ tool to trace the length and the width of each wild-type (n=10) or Snx3^−/− (n=9) E8.5 embryo. The means for each genotype were calculated and a two-tailed t-test was performed to test for statistical significance using Graphpad Prism software. Data are mean±s.e.m. and were plotted using Graphpad Prism software.

Quantification of the ELISAs was performed using Graphpad Prism software to calculate the means of the absorbance reading at 450 nm for each experimental condition, then performing a two tailed t-test to test for statistical significance. Data are mean±s.e.m. and were plotted using Graphpad Prism software.

Colocalization analysis for live-cell imaging was quantified using the Colocalization plug-in by ImageJ. Briefly, the images were loaded to ImageJ, channels were merged to create an 8-bit TIFF, then fluorescent channels were split and auto-thresholded for pixel intensity. Then the Colocalization plug-in was employed to calculate the percent overlap between pixels of each channel, indicating overlap of fluorescent proteins within the frame (imagej.net). Statistical analysis was performed using Graphpad Prism Software for two-tailed t-tests to test for statistical differences in pixel overlap of each experimental condition. Data are mean±s.e.m. and were plotted using Graphpad Prism software.

Quantification of total WLS fluorescence in all conditions was performed using ImageJ with the background subtracted and normalized for area. Statistical analysis was performed using Graphpad Prism software using two-tailed t-tests or non-parametric t-tests for data without normal distribution, to test for statistically significant differences in total WLS fluorescence. Individual data points and mean±s.e.m. were plotted using Graphpad Prism software.

Fishers exact test was performed using Graphpad Prism to evaluate the statistical significance of the rescued Snx3^−/− embryos treated with LiCl compared with the Snx3^−/− embryos not rescued by LiCl. Power calculations were performed to identify the smallest number of animals required for adequate detection of effect size. The B6N(Cg)-Snx3tm1.1(KOMP)Vlcg/J mouse shows a classic Mendelian ratio for all genotypes and the Snx3^−/− mice show a 100% penetrant NTD phenotype. All datasets were tested for normality using Graphpad Prism and P<0.05 was considered significant for all experiments. The pre-established exclusion criteria were that any single data point three standard deviations from the mean was considered an outlier and not included in analysis.

We thank the members of L.A.N.’s lab for the help and feedback throughout the course of these experiments, especially Lori Bulwith who helped with mouse husbandry. We also thank CU Boulder undergraduate student Megan Hupka for help with mouse dissections, MEF production and transfections, as well as CU Boulder undergraduate student Caitlin Ryan for help with genotyping and mouse dissections. We thank the CU Boulder Office for Animal Resources staff for help with colony care and training on intraperitoneal injections. We thank the University of Colorado Boulder Light Microscopy Core Facility and Dr James Orth, PhD for assistance with the CellVoyager CV1000 Yokogawa spinning disk confocal microscope for all live-cell imaging, University of Colorado Boulder Stem Cell Research and Technology Core and Dr Teisha Rowland, PhD for the use of the Olympus IX83 Fluorescence Microscope for whole embryo imaging, and the University of Colorado Anschutz Medical Campus laboratory of Dr Julie Siegenthaler, PhD for the use of the Nikon SMZ18 Stereoscope with Nikon DS-Ri1 CCD-cooled camera to image whole-mount RNA in situ hybridization embryos.