TMEM132A regulates mouse hindgut morphogenesis and caudal development Free

Around one to five per 100,000 human infants are born with a wide spectrum of structural defects in caudal structures including the sacrococcygeal bone (tail bone), the pelvic bones and lower limbs (Andrish et al., 1979). These conditions, described with partially overlapping terms such as caudal agenesis, caudal dysgenesis or caudal regression, are frequently associated with additional developmental defects including posterior neural tube closure defects and genitourinary and/or gastrointestinal tract abnormalities (Warner et al., 2020; Lee et al., 2021). In the most severe cases, the pelvic and leg bones are fused to form a single leg in the ventral midline, a condition known as sirenomelia or mermaid syndrome (Boer et al., 2017). Multiple hypotheses have been proposed to explain the etiology of these caudal developmental defects, ranging from mesodermal cell migration defects and lateral compression by amniotic folds to insufficient blood supply (Boer et al., 2017). However, none of these hypotheses explains the whole spectrum of genitourinary and gastrointestinal tract defects seen in these patients.

Parts of the genitourinary tract, including the urinary bladder, urethra and external genitalia, and the posterior-most part of the gastrointestinal tract, the rectum, were originally derived from the cloaca, an enlarged area at the posterior-most part of the hindgut, and the mesenchyme around it (Roberts, 2000; Liaw et al., 2018; Hashimoto et al., 2019). Therefore, the gastrointestinal and genitourinary tract abnormalities frequently associated with the caudal structural defects may share the same developmental origin in abnormal hindgut development. This possibility has not been discussed or experimentally explored in any model systems.

During gastrulation and neural tube closure, cells are rearranged such that the embryos become narrower along the mediolateral axis and longer along the anteroposterior axis, a process known as convergent extension (CE) (Nikolopoulou et al., 2017; Wang et al., 2019). Mutations in genes encoding the core planar cell polarity (PCP) proteins, including cadherin EGF LAG seven-pass G-type receptor-1 (Celsr1) (Curtin et al., 2003), frizzled class receptor 3/6 (Fzd3/6) (Wang et al., 2006) and Van Gogh-like 2 (Vangl2) (Kibar et al., 2001; Murdoch et al., 2001), disrupt CE and neural tube closure, suggesting that CE is important for neural tube morphogenesis. Biochemically, homophilic interactions between CELSR1 on neighboring cells leads to the subsequent recruitment of FZD3/6 and VANGL2 to the opposite sides of the cell/cell junction, which in turn recruits downstream cytoplasmic effectors of the PCP signaling pathway, leading to the cytoskeletal changes needed for CE and other cellular behavior (Nikolopoulou et al., 2017; Wang et al., 2019).

Transmembrane protein 132A (TMEM132A) is a member of the novel TMEM132 family of single-pass transmembrane proteins (Sanchez-Pulido and Ponting, 2018). Biochemistry, cell biology and bioinformatics studies have linked this protein family to multiple processes, including WNT protein release, interaction with extracellular matrix, cell adhesion and actin cytoskeleton regulation (Sanchez-Pulido and Ponting, 2018; Li and Niswander, 2020; Wang et al., 2021; Li et al., 2022). A mutant analysis in Caenorhabditis elegans suggests that the nematode Tmem132 homolog is an important regulator of neuronal morphogenesis (Wang et al., 2021). The Drosophila Tmem132 homolog has been implicated in reproduction and thermal nociception (Honjo et al., 2016; Chon et al., 2021). Loss of Tmem132a in mouse leads to preweaning lethality with neural tube closure defects and abnormal limb development (Dickinson et al., 2016; Li et al., 2022).

Here, we show that Tmem132a mutant mice exhibit defects in caudal development, including fused pelvic bones, approximation of hindlimbs, and short and abnormal tails. These mutants also exhibit abnormal development of the kidneys, ureters, bladders, external genitalia and rectum, a phenotype frequently associated with caudal development defects. We show that hindgut fails to extend into the tail bud of Tmem132a mutant embryos, likely resulting from an earlier failure in the exclusion of visceral endoderm from the hindgut. Finally, we show that TMEM132A physically interacts with PCP proteins CELSR1 and FZD6, and synergizes with VANGL2 in the regulation of CE and neural tube closure.

To study the in vivo function of TMEM132A, we acquired the Tmem132a^{tm1a(KOMP)Wtsi} mutant mouse strain from the Knockout Mouse Project (KOMP; The Jackson Laboratory). In this strain, a lacZ reporter cassette was inserted into the first intron of Tmem132a to disrupt its transcription and splicing (Figs S1 and S2). For brevity, we will call this allele Tmem132a^lacKI for the rest of this paper. We then crossed Tmem132a^lacKI with germline flippase (ROSA26-FLPe) mice to produce Tmem132a^flox allele (a.k.a. Tmem132a^{tm1c(KOMP)Wtsi}), which has loxP sites in introns 1 and 3 (Fig. S1). Tmem132a^flox homozygous mice (n=117) were healthy and fertile, suggesting that the developmental defects observed in Tmem132a^lacKI mutants (described below) resulted specifically from the insertion of the lacZ cassette in the Tmem132a locus. Finally, we crossed Tmem132a^flox mice with germline Cre (E2a-Cre) mice to remove exons 2 and 3 (Fig. S1). In this Tmem132a^ΔEx2,3 allele, the deletion of two exons was predicted to result in a frameshift shortly after the translation start codon, which we confirmed by sequencing the resulting transcripts (Fig. S2). The Tmem132a^lacKI and Tmem132a^ΔEx2,3 alleles exhibited an identical phenotype as described below, hence we will refer to both alleles as Tmem132a^–, unless otherwise noted.

Only three out of 235 Tmem132a^lacKI and three out of 217 Tmem132a^ΔEx2,3 pups weaned from crossings between corresponding heterozygotes were homozygous mutants, suggesting that both alleles were homozygous pre-weaning lethal with high penetrance. All surviving homozygous mutant pups exhibited oligosyndactyly and kinked tails (Fig. 1A). Homozygous mutants were recovered at Mendelian ratio at embryonic stages up to embryonic day (E) 18 (467 out of 1737, 27%). In addition to the limb and tail phenotype observed in the surviving mutant pups, the homozygous mutant embryos frequently exhibited spina bifida (158 out of 256, 62%) and approximation of the hindlimbs (21 out of 38, 55%) (Fig. 1B). To better appreciate the structural defects of the caudal region of Tmem132a mutant embryos, we performed skeletal preparation at E18.5. This analysis confirmed the defects observed in wholemount embryos, including short tails and severe loss of digits in the hindlimbs (Fig. 1C). It also showed the loss of fibula (Fig. 1C). Strikingly, in Tmem132a mutants we observed the ventral shift and partial fusion of the ilia, part of the pelvis that is normally located to the lateral sides of the vertebrae (Fig. 1C). This pelvic defect brought the two femurs together to the ventral midline, leading to hindlimb approximation in Tmem132a mutants.

It has been reported previously that a slightly different allele, Tmem132a^{tm1b(KOMP)Wtsi}, exhibited kidney defects, but no detailed information was provided (Fig. S1 and Dickinson et al., 2016). We found a range of defects affecting kidney and genitourinary tracts in Tmem132a mutants. The most frequent defects were hydronephrosis and hydroureters (Fig. 1D,E; n=32/76). Less frequent kidney defects in Tmem132a mutants included renal agenesis (Fig. 1D; n=18/76) and horseshoe kidneys (Fig. S3A,B; n=5/76). We also observed bladder agenesis in some Tmem132a mutants (Fig. 1D; n=7/23), which could underlie hydronephrosis and hydroureter as no outlet existed for urine. However, some embryos exhibited hydronephrosis and hydroureter despite the presence of bladders, suggesting defective urine drainage downstream of the ureters. To test this, we injected India ink into the E18.5 kidneys. In wild type, the ink flew into the bladder via the ureter (Fig. 1F; n=3). In contrast, the injected ink entered the ureter but failed to reach the bladder in Tmem132a mutants, suggesting that the junction between the ureters and the bladder, the ureterovesical junction, was not properly established (Fig. 1F; n=3).

The bladder is derived from the caudal-most part of the hindgut known as cloaca, which also gives rise to the genital tubercle (GT; precursor of the external genitalia) and rectum (Roberts, 2000; Liaw et al., 2018; Hashimoto et al., 2019). If the cloaca development was defective, we would expect developmental defects in the GT and rectum in Tmem132a mutants. Indeed, after inspecting 26 mutants ranging from E14.5 to newborn, we found that GTs failed to form in seven mutants, whereas GTs in 11 mutants were apparently smaller than their littermates and were malformed (Fig. 1G and Fig. S3C,D). Interestingly, imperforate anuses were found in 10 Tmem132a mutants, suggesting defects in rectum morphogenesis. The malformation of bladder, genitalia and rectum strongly suggests a common origin of developmental defects in cloaca morphogenesis.

To understand the origin of the defects in caudal neural tube closure and cloaca morphogenesis, we focused on E9.5 embryos. At this stage, the neural tube is closed along its entire length except for the caudal-most part, where the posterior neural pore (PNP) remains open. Interestingly, we observed clear morphological differences between the wild-type and Tmem132a mutant PNPs. In wild-type embryos, the PNP did not reach the tip of the nascent tail bud, and has a rhombic shape when viewed from a dorsolateral angle (Fig. 2A). In contrast, the posterior-most neural epithelium extended beyond the tip of the tail bud and reached ventrally in Tmem132a mutants (Fig. 2A). To better illustrate this phenotype, we made two measurements. First, we compared the PNP length in E9.5 wild-type (n=20) and Tmem132a mutant (n=7) embryos (double arrow lines in Fig. 2A), and found that the mutant PNP was three times longer than that in wild type (Fig. 2B). We then measured the tail curvatures, defined as the angle between one line tangential to the last somite, and another line connecting the anterior and posterior ends of the PNP (Fig. 2C). We found that the tail curvature was significantly greater in Tmem132a mutants (n=9) than in wild-type (n=26) embryos (Fig. 2C).

We next examined the expression of a neural epithelium marker, Sox2, in E9.5 embryos (Fig. 2Da,Db; n=3 for each genotype). We found that the neural ridges were more elevated in wild-type than in Tmem132a PNPs (Fig. 2Da′,Db′). We also found ventrally located Sox2-expressing neural epithelial cells on the sections near the tip of the Tmem132a mutant tail bud, consistent with PNP reaching ventrally due to increased curvature (Fig. 2Db″). In contrast, Sox2 expression is absent on wild-type sections near the tip of tail buds (Fig. 2Da″). We conclude that failure in closing the PNP underlies spina bifida at later stages.

We next investigated whether abnormal dorsal/ventral (D/V) patterning of the neural tube could contribute to the failure in neural tube closure. We found that despite regionalized expression of the D/V marker genes such as Foxa2, Olig2 and Nkx6.1 in more anterior regions of both wild-type and Tmem132a mutant E9.5 neural tube, none of these genes was expressed at the level of PNP (Fig. S4; n=3 for each genotype). These results suggest that PNP closure occurs before the establishment of the neural tube D/V pattern, thus the neural tube defects in Tmem132a mutants do not result from abnormal neural patterning.

The presence of various tail morphogenetic defects in Tmem132a mutants raised the possibility that the neural tube closure defects may be secondary to defective tail bud growth. To test this hypothesis, we examined the expression of Wnt3a, brachyury (Bra; also known as T) and Fgf8 (Kispert and Herrmann, 1994; Crossley and Martin, 1995; Greco et al., 1996). All three genes were expressed at high levels in both the wild-type and Tmem132a mutant tail buds, suggesting that the signals promoting tail bud growth were not severely affected (Fig. 3A; n=3 for each genotype).

Interestingly, although the expression levels of Fgf8 in the tail buds were comparable between wild type and Tmem132a mutants, the expression patterns were clearly different. In the wild-type tail buds, Fgf8 expression was present in the dorsal mesoderm as well as the dorsal part of the hindgut (indicated by an arrow in Fig. 3A; Gofflot et al., 1997). The hindgut expression of Fgf8 was not present in Tmem132a mutants, despite normal expression of Fgf8 in the mesoderm (Fig. 3A). Shh was also known to be expressed in the ventral hindgut at this stage (Gofflot et al., 1997), hence we examined whether Shh expression was abnormal in Tmem132a mutant embryos. The hindgut expression of Shh extends all the way to the tip of the tail bud in wild-type embryos, whereas its expression domain was significantly shorter in Tmem132a mutants (Fig. 3B; n=3 for each genotype). In contrast, the notochord expression of Shh appears to be unchanged between wild-type and Tmem132a mutant embryos (Fig. 3B). To better characterize the potential hindgut morphogenesis defects, we labeled the hindgut by crossing Ai9, a TdTomato-based fluorescent Cre reporter line, with Sox17GFPCre, a knock-in mouse line expressing a GFP-Cre fusion protein in the endodermal lineage (Madisen et al., 2010; Choi et al., 2012). At E9.5, the TdTomato-labeled hindgut reached the tip of the wild-type tail buds (Fig. 3C, n=10). In contrast, the hindgut did not extend into the Tmem132a mutant tail buds (Fig. 3C, n=4). This defect in hindgut extension into the tail bud likely results in the significant increase in the curvature of the tail bud, leading to spina bifida in Tmem132a mutants.

To investigate the cellular mechanism underlying the hindgut extension defects in Tmem132a mutants, we visualized hindgut formation at E8.25 through both optical and electron microscopy. The first sign of hindgut morphogenesis was the formation of caudal intestinal portal (CIP), through which the hindgut invaginated and extended caudally (Fig. 4A,B) (Nowotschin and Hadjantonakis, 2020). Interestingly, we frequently found piles of extra cells near the CIP of Tmem132a mutant embryos, suggesting an early onset of the hindgut morphogenetic defects (Fig. 4A,B; n=43 out of 54 mutant embryos).

Before gastrulation, the exterior layer of the wild-type mouse embryo is composed of visceral endoderm (VE). During gastrulation, epiblast-derived definitive endoderm (DE) cells migrate toward, and integrate themselves into, the distal region of the endodermal layer (Viotti et al., 2014; Nowotschin and Hadjantonakis, 2020). By E8.0, the distal DE cells and proximal VE cells could be easily distinguished under a scanning electron microscope because the apical surface of DE cells was flat whereas that of the VE cells was convex (Fig. 4B). The boundary between DE and VE formed a perfect arc in the posterior part of wild-type embryos at this stage, indicating the exclusion of VE by DE in the posterior/medial region (Fig. 4B; n=5). In contrast, the VE was not efficiently excluded by DE in the posterior-medial region of Tmem132a mutant embryos, leaving a narrow streak of VE cells along the midline of the embryo, flanked on both sides by lateral DE cells (Fig. 4B; n=5).

To confirm these morphology-based observations, we sought to label VE cells specifically. DBA-lectin associated with the surface of VE cells, but not DE cells (Fig. 4C; n=10; Kimber, 1986). This assay confirmed the VE exclusion defects in E7.5 and E8.0 Tmem132a mutant hindguts, as DBA-lectin-binding VE cells in the posterior-medial region were not excluded efficiently by DE cells that do not bind DBA-lectin (Fig. 4C; n=5 at each stage). Interestingly, the DBA-lectin staining allowed us to detect small groups of lectin-binding VE cells surrounded by DE cells in Tmem132a mutant hindguts (Fig. 4C). These ectopic VE cells probably delaminate from surrounding DE cells, forming the piles of cells at the CIP of E8.25 Tmem132a mutant embryos (Fig. 4A,B).

Transthyretin (Ttr) is specifically expressed in VE, but not DE, cells during gastrulation and early hindgut morphogenesis (Cereghini et al., 1992). Using RNA in situ hybridization, we found that Ttr-expressing VE cells were not excluded from the posterior/medial region of E7.5 Tmem132a mutant as were those in control wild-type embryos, consistent with the observation from the DBA-lectin binding assay (Fig. 4C,D; n=5 for each genotype). At E8.0, small groups of Ttr-expressing VE cells were present among Ttr-negative DE cells in posterior hindguts of Tmem132a mutant embryos (Fig. 4D; n=5 for each genotype). To confirm that the piles of cells observed near the CIP in E8.25 Tmem132a mutant embryos were indeed misplaced VE cells, we examined Ttr expression at this stage, and found that these cells did indeed express Ttr (Fig. 4D; n=5 for each genotype).

The failure in efficient VE exclusion in the medial hindgut region suggests that DE cells may have difficulty entering this region. To test whether this is the case, we crossed Tmem132a mutants to Sox17GFPCre knock-in mice, in which a GFP-CRE fusion protein is produced from the Sox17 locus to label the DE cells (Choi et al., 2012). We found that in E8.0 wild-type littermate hindgut, the GFPCre-positive DE region expanded normally, forming an arc-shaped boundary with the GFPCre-negative VE region (Fig. 4E; n=5). In striking contrast, GFPCre-expressing DE cells were excluded from the medial region of the hindgut in Tmem132a mutant embryos, consistent with a failure of DE cells to exclude VE cells in the absence of Tmem132a (Fig. 4E; n=5). Consistent with direct involvement of Tmem132a in gut morphogenesis, we found Tmem132a expression in the E8.0 ectoderm, mesoderm and embryonic endoderm, but not in the flanking VE-derived yolk sac (Fig. 4F; n=3).

In summary, these analyses suggest that the hindgut morphogenetic defects observed at E9.5 result from an earlier defect during gastrulation, when DE cells failed to efficiently exclude VE cells in the posterior medial region. Later, the expansion of the lateral DE leads to the delayed formation of CIP, but the hindgut failed to extend into the tail bud as in wild type. Isolated groups of VE cells failed to retract posteriorly and formed piles of cells in the hindgut, likely exacerbating the hindgut morphogenetic defects.

The TMEM132 family of proteins shares large extracellular domains comprising multiple bacterial Ig domains, which are known to mediate cell adhesion (Sanchez-Pulido and Ponting, 2018). We set out to test whether TMEM132A exhibits homophilic interaction across cellular junctions by first examining the subcellular localization of TMEM132A. We found that overexpressed TMEM132A-GFP colocalized with cadherin 1 (CDH1) at the adherens junction in MDCK cells (Fig. 5A). Comparison with the tight junction protein ZO-1 (TJP1) indicated that TMEM132A was not localized to the apical domain of MDCK cells (Fig. 5B).

We next examined whether homophilic interaction exists between TMEM132A proteins. We co-expressed FLAG and GFP tagged TMEM132A in HEK293 T cells, and immunoprecipitated TMEM132A-FLAG. As expected, TMEM132A-GFP co-immunoprecipitated with TMEM132A-FLAG, suggesting homphilic interaction between TMEM132A proteins (Fig. 5C).

The homophilic interaction may indicate dimerization and/or oligomerization between TMEM132A proteins expressed in the same cells. Alternatively, or additionally, TMEM132A proteins may interact in trans with ones from neighboring cells across cellular junctions. To determine whether this intercellular interaction exists, we performed a cell aggregation assay using K-562, a leukemia cell line (Lozzio and Lozzio, 1975). Control, GFP-expressing K-562 cells were overwhelmingly in single cells after overnight culture on a shaker (Fig. 5D). In contrast, TMEM132A-GFP-expressing K-562 cells formed many small multicellular aggregates (Fig. 5D). As a positive control, CDH1-GFP-expression led to the formation of large aggregates of K-562 cells (Fig. 5D). These results indicate that the TMEM132A proteins interact across cellular junction with moderate affinity.

To further understand the molecular function of TMEM132A, we sought to identify additional proteins it interacts with. Hein et al. (2015) surveyed the interaction partners of more than 1100 proteins, including TMEM132A, through affinity-based enrichment analyses (Hein et al., 2015). One interesting candidate TMEM132A-interacting protein identified in this study is a WNT receptor and core PCP protein, FZD6. To confirm this interaction, we co-expressed TMEM132A-GFP and FZD6-FLAG in HEK293 T cells. We found that these two proteins co-immunoprecipitated, suggesting that they did interact with each other (Fig. 6A). We next investigated the subcellular localization of these two proteins in HeLa cells, and found that they colocalized at the cell surface (Fig. 6B).

Previous studies suggested that CELSR1 physically interacts with FZD6 and recruits it to the cell junction (Devenport and Fuchs, 2008; Stahley et al., 2021). To investigate whether TMEM132A also interacts with CELSR1, we performed a co-immunoprecipitation analysis between overexpressed CELSR1-MYC and TMEM132A-FLAG. We found that immunoprecipitating TMEM132A-FLAG greatly enriched CELSR1-MYC, suggesting physical interaction between these two proteins (Fig. 6C). Consistent with this, we found that co-expressed CELSR1-GFP and TMEM132A-MYC in HeLa cells were co-localized at the cell junctions (Fig. 6D).

Taken together, we found that TMEM132A physically interacted with CELSR1 and FZD6 and colocalized with them at the cell junction, suggesting that it may be directly involved in PCP signaling.

The WNT family of secreted proteins regulates numerous morphogenetic processes through a few distinct pathways (Vuong and Mlodzik, 2022). The canonical pathway involves the stabilization and nuclear translocation of β-catenin, and the conversion of the TCF/LEF transcriptional repressor into an activator. Using a TCF/Lef:H2BGFP reporter mouse line (Ferrer-Vaquer et al., 2010), we found no significant decrease in canonical WNT activities in Tmem132a mutant embryos (Fig. S5). This is consistent with FZD6 as a major mediator for the non-canonical PCP pathway, rather than the β-catenin-dependent canonical pathway (Wang et al., 2019).

The loop tail (Lp) mice, carrying a dominant/negative mutation in the core PCP gene Vangl2, fail to close the neural tube along the entire anterior/posterior axis, a condition known as craniorachischisis (Kibar et al., 2001; Murdoch et al., 2001). Interestingly, Vangl2 exhibits strong synergistic interaction with other genes involved in PCP and neural tube closure such that compound mutants with one copy of Lp mutation and one or both copies of mutations in another locus exhibit neural tube defects ranging from severe spina bifida to craniorachischisis (e.g. Lu et al., 2004; Yu et al., 2010). The potential involvement of Tmem132a in PCP prompted us to examine the synergistic interaction between Tmem132a and Vangl2 in Tmem132a;Lp compound mutants. Consistent with previous reports, we observed kinky or looped tails in most E16.5 heterozygous Lp mutant embryos, but no defects in neural tube closure (n=68; Fig. 6E). Tmem132a homozygous mutant embryos exhibited frequent spina bifida (21 out of 24; Fig. 6E), and tail defects similar to those in Lp heterozygotes (n=24; Fig. 6E). Interestingly, seven out of 24 Tmem132a^–/–;Lp/+ compound mutants exhibited craniorachischisis, suggesting synergistic genetic interaction between Tmem132a and Vangl2 during neurulation (Fig. 6E). This is consistent with the physical interactions between TMEM132A and some PCP core proteins, suggesting an important role for TMEM132A in regulating PCP in mammals.

Defects in convergent extension during neurulation are known to underlie neural tube closure defects in many mouse mutants (Nikolopoulou et al., 2017; Wang et al., 2019). To determine whether convergent extension was affected in Tmem132a^–/– mutant and Tmem132a^–/–;Lp/+ compound mutant embryos, we measured their length-to-width ratio (LWR) at the three- to five-somite stage (Fig. 6F). We found that the LWRs of three- to five-somite-stage Lp/+ (n=11) and Tmem132a^–/– (n=4) embryos were not significantly different than those of wild type (n=20) embryos, but the LWR of Tmem132a^–/–;Lp/+ compound mutants (n=7) was significantly reduced compared with wild type (Fig. 6F), suggesting that convergent extension is defective in these embryos.

In this study, we find that Tmem132a mutant mice exhibit spina bifida, pelvic fusion and tail malformation, as well as genitourinary and gastrointestinal tract malformations frequently seen in caudal regression and sirenomelia. The characterization of earlier developmental defects suggests that most, if not all, of these defects can be explained by a failure in the exclusion of VE by DE in the hindgut. Our study indicates that TMEM132A mediates intercellular interaction through homophilic interaction, interacts physically with and colocalizes with PCP proteins CELSR1 and FZD6 at the cell surface. Consistent with its involvement in the PCP pathway, we find that Tmem132a genetically interacts with Vangl2 in the regulation of convergent extension and neural tube closure.

Caudal developmental defects, including caudal regression, caudal dysgenesis, caudal agenesis and sirenomelia, are rare but devastating birth defects with unclear etiology (Boer et al., 2017; Warner et al., 2020; Lee et al., 2021). Existing hypotheses point to defects in mesodermal migration or insufficient blood supply as potential origins of the caudal developmental defects, but they have failed to explain the genitourinary and gastrointestinal tract malformations frequently associated with caudal developmental defects. Our study suggests that abnormal hindgut extension could account for many caudal developmental defects (Fig. 7). First, bladder, external genitalia and rectum are either derived from the cloaca, the caudal-most part of the hindgut, or mesenchyme closely associated with the cloaca. Therefore, the loss or malformation of the bladder and external genitalia, as well as imperforate anus in Tmem132a mutants, are the direct consequences of earlier failure of the hindgut extension. Hydroureter and hydronephrosis in Tmem132a mutants are indirect results of cloacal anomaly, as our ink injection assay indicates failed connection between ureters and the bladder. The defects in hindgut extension may also account for the fusion of the pelvic bones, approximation of hindlimbs and horseshoe kidneys, because these lateral mesodermal tissues are no longer separated by the medially located endoderm (Fig. 7A).

A recent report showed accumulation of paraxial mesoderm in the caudal-most region of Tmem132a mutants (Li et al., 2022). This observation is consistent with ours on the medial shift of mesodermal structures, which we believe is secondary to the absence or reduction of cloaca-derived medial structures due to an earlier hindgut extension defect. Nevertheless, our conclusion does not exclude an additional potential role of Tmem132a in promoting mesodermal migration as suggested by the previous report (Li et al., 2022). It will be interesting to reveal tissue-specific roles of Tmem132a in endoderm and paraxial mesoderm through more functional studies.

The closure of the posterior neural tube depends on the underlying mesoderm and hindgut, a mechanism best illustrated by the study of curly-tail (ct) mutant mice (van Straaten and Copp, 2001). In ct mutant embryos, compromised proliferation in the hindgut leads to increased ventral curvature, preventing the closure of PNP (Copp et al., 1988a). Correcting this defect by reducing proliferation in the neural tube, or by preventing ventral curvature through mechanical splinting, effectively prevents neural tube defects in ct mutant mice (Copp et al., 1988b; Brook et al., 1991). On the other hand, mechanically forcing an increased tail curvature results in caudal neural tube defects in wild-type embryos, suggesting that a proper curvature is a general requirement for closing the posterior neural tube (Peeters et al., 1996). Interestingly, the failure of hindgut extension into the tail buds of E9.5 Tmem132a mutants similarly results in increased tail curvature and PNP length (Fig. 2B,C). Therefore, the posterior neural tube closure defects in Tmem132a mutant embryos are likely secondary to the hindgut extension defects (Fig. 7B).

The fact that TMEM132A physically interacts with, and is colocalized with, both CELSR1 and FZD6 suggests that TMEM132A is directly involved in non-canonical WNT/PCP signaling. Combined with our K-562 cell aggregation assay suggesting cross-junctional homophilic interaction of TMEM132A, we hypothesize that TMEM132A and CELSR1 belong to the same signaling complex triggering downstream PCP response. The strong genetic interaction between Tmem132a and Vangl2, another core PCP gene, in convergent extension and neural tube closure, further confirms a role of Tmem132a in PCP signaling.

The convergent extension defects and craniorachischisis in Tmem132a^–/–;Lp/+ compound mutants raise an interesting question: does a convergent extension defect also contribute to spina bifida in Tmem132a mutants? Although we cannot rule out the contribution of a mild convergent extension defect, the drastic changes in the morphology of the hindgut and tail bud of Tmem132a mutant embryos is more likely the main underlying cause of the localized PNP closure defect.

There have been two models explaining how VE is excluded from the distal part of the embryo where gut forms. A traditional model posits that VE cells are displaced physically by DE cells to the proximal part of the embryo (Lawson et al., 1986). In contrast, a new model based on recent live imaging and lineage tracing studies posits that instead of being physically displaced, the distal VE cells switch their fate by turning off VE-specific genes (e.g. Ttr) and turning on DE-specific genes (e.g. Sox17) (Kwon et al., 2008; Viotti et al., 2014). Unfortunately, little is known about the molecular regulation of VE exclusion due to the lack of genetic resources such as mutants affecting this process. To our best knowledge, we are the first to report here a mutant with specific defects in VE exclusion in the hindgut, providing an invaluable new resource for the study of this important step of endodermal development.

In conclusion, our study reveals an important role of Tmem132a in the PCP pathway, convergent extension and exclusion of VE by DE in the hindgut. Our results implicate endodermal defects as a previously unsuspected origin of various caudal developmental syndromes, providing explanations for many features of these diseases such as spina bifida, pelvic, hindlimb, genitourinary and gastrointestinal tract defects. It will be interesting to investigate whether the Tmem132a locus is genetically associated with human caudal developmental defects, and how Tmem132a regulates the interaction between visceral and definitive endodermal cells.

Tmem132a^{tm1a(KOMP)Wtsi} mutant mice were purchased from Mutant Mouse Resource and Research Centers (MMRRC) at University of California, Davis. The following primers were used to genotype wild-type (279 bps) and this mutant (332 bps) allele: forward 1, ACCCTTAGGGAGTGCCAGT; reverse 1, CCACAACGGGTTCTTCTGTT; reverse 2, GACACAGCTCCTCCTGCATT. The following two primers were used to genotype the wild-type (348 bps) and Tmem132a^flox (380 bps) allele: forward, CTTCCAGGTGAGTGGGTTGT; reverse, AAGGAACGGCTCTTTCCAGT. Finally, the following primers were used to genotype Tmem132a^ΔEx2,3 allele (400 bps): forward, ACCCTTAGGGAGTGCCAGTT; reverse, AAGGAACGGCTCTTTCCAGT. Sox17^{tm2(EGFP/cre)Mgn} mice were kind gifts from Dr Magnuson at Vanderbilt University and were genotyped as previously described (Choi et al., 2012). Vangl2^Lp-m1Jus mice were purchased from MMRRC and genotyped as previously described (Kibar et al., 2001; Heydeck and Liu, 2011). TCF/Lef:H2BGFP, E2A-Cre and ROSA26-FLPe transgenic mice were purchased from The Jackson Laboratory (stock #032577, #003724 and #016226, respectively), and genotyped as previously described (Lakso et al., 1996; Farley et al., 2000; Ferrer-Vaquer et al., 2010). All animal studies in this work have been approved by the Pennsylvania State University Institutional Animal Care and Use Committee.

The kidneys were fixed in 4% paraformaldehyde (PFA) overnight and embedded in paraffin, sectioned at 5µm using a Leica CM1850 cryostat and subsequently stained with Hematoxylin and Eosin (H&E) as described previously (Cai and Liu, 2016). Photos were taken using a Nikon E600 microscope and a QImaging Micropublisher camera.

Embryos were fixed in 95% ethanol, skinned and eviscerated and then stained with Alcian Blue and Alizarin Red as described previously (Chang et al., 2015). Photos were taken in 80% glycerol/1% KOH using a Zeiss Discovery microscope and a QImaging micropublisher camera.

E18.5 mouse embryos were euthanized and partially eviscerated to expose the kidneys, ureters and bladders. The specimens were pinned onto a black Sylgard plate, and black India Ink was injected into one of the kidneys with a gauge 30 needle. Photos were taken on a Zeiss Discovery microscope and a QImaging micropublisher camera.

MDCK (ATCC, #CCL-34) or HeLa (ATCC, #CCL-2) cells were seeded on glass coverslips and transfected using jetPRIME (Polyplus, 114-07) or Lipofectamine 3000 (Invitrogen, L3000-015) the next day. Then, 48 h after transfection, cells were fixed with 4% PFA for 10 min followed by blocking and permeabilization with 10% fetal bovine serum (FBS)+0.1% Triton X-100 in PBS for 10 min at room temperature. The cells were incubated with primary antibodies at 4°C overnight, washed three times with PBS, then incubated with secondary antibodies and Hoechst 33258 for 1-2 h. The manufacturer's recommended protocol was followed when using the anti-FLAG antibody (Sigma-Aldrich, F1804, 1:100). Images were taken using an Olympus Fv1000 confocal microscope and processed in FIJI. CDH1 (Sigma-Aldrich, U3254, 1:250) served as an adherens junction marker, and ZO-1 [R26.4C, 1:200, deposited to the Developmental Studies Hybridoma Bank (DSHB) by D.A. Goodenough, Harvard Medical School, MA, USA] was used to label tight junctions. Additional antibodies used include: GFP (Invitrogen, A11122, 1:1000), Myc (Sigma-Aldrich, M4439, 1:500), AlexaFluor488-coupled anti-IgG (Invitrogen, A11001 and A11008, 1:250) and Cy3-coupled anti-IgG (Jackson ImmunoResearch, 111-165-003 and 115-165-003, 1:250). All antibodies were validated by comparing the staining patterns to ones provided by the manufacturers.

The caudal region (hindlimb buds to tail bud) of E9.5 embryos were fixed in 4% PFA for 1 h and processed for transverse cryosections at 10 µm. The sections were incubated with primary antibodies at 4°C overnight after permeabilization and blocking in PBS plus 0.1% Triton X-100 (PBST) and 1% goat serum. The sections were then washed in PBST and incubated in secondary antibodies and Hoechst 33258. Following a final wash, the sections were mounted in DABCO (Sigma-Aldrich) and imaged with an ECHO Revolve microscope. Antibodies used in this assay were: FOXA2 (4C7, 1:25, deposited to the DSHB by T.M. Jessell and S. Brenner-Morton, Columbia University, NY, USA), OLIG2 (Millipore, AB9610; 1:1000), NKX6.1 (F55A12, deposited to the DSHB by O.D. Madsen, Beta Cell Biology Consortium, 1:500), AlexaFluor488-coupled anti-IgG (Invitrogen, A11001 and A11008, 1:250) and Cy3-coupled anti-IgG (Jackson ImmunoResearch, 111-165-003 and 115-165-003, 1:250). All antibodies have been successfully used in our previous studies (Hoover et al., 2008; Jia et al., 2009; Zeng et al., 2010; Liu et al., 2012, 2015; Cai and Liu, 2017).

K-562 cells (CCL-243, ATCC) were cultured in Iscove's Modified Dulbecco's Medium supplemented with 10% FBS at 37°C with 5% CO₂. A stable K-562 cell pool expressing TMEM132A-GFP was generated by FACS after transfection and G418 selection. Stable cell lines were generated by seeding individual transfected cells in 96-well plates. For aggregation assays, K-562 cells were passed at 1:2, and were counted and resuspended in fresh culture medium at 2.5∼5.0×10⁵ cells/ml the next day. The cells were then cultured for 24 h on a rocker (Boekel Scientific Rocker II, 260350, 25 rpm), and were imaged immediately afterwards on an ECHO Revolve microscope.

RNA in situ hybridization on cryosections was performed as described previously (Liu and Liu, 2014, 2020). Briefly, the sections were treated with proteinase K and acetylated with acetic anhydride before being incubated with the Digoxigenin (DIG)-labeled probes. After hybridization, the sections were extensively washed and incubated with AP-coupled anti-DIG antibody (Sigma-Aldrich, 11207733910; 1:2500). RNA in situ hybridization on wholemount embryos was performed as described previously (Hoover et al., 2008). The anti-sense RNA probes for Bra, Sox2, Tmem132a, Wnt3a and Ttr were produced by in vitro transcription using the 3′ UTRs as templates.

HEK293T cells were transfected with expression constructs using PEI (Polysciences, 24765). Then, 48 h after transfection, cells were lysed with buffer containing 20 mM Hepes (pH 7.5), 150 mM NaCl, 1 mM EDTA, 1% Triton X-100, 10% glycerol and protease inhibitors. The concentrations of the lysates were determined by BCA protein assay (Pierce, 23225). For immunoprecipitation, the same amount of lysate from each sample was incubated with anti-FLAG M2-Agarose Affinity Gel (Sigma-Aldrich, A2220) at 4°C overnight. Antibodies used for western blots were: GFP (Life Technologies, A11122; 1:2500), FLAG (Sigma-Aldrich, F7425; 1:4000) and β-tubulin (Sigma-Aldrich, T4026; 1:5000). All western blots were scanned using a LI-COR Odyssey DLx Imaging System and quantified using LI-COR Image Studio Lite.

Embryos were dissected in cold PBS and fixed in 4% PFA overnight. They were then washed with PBS and dehydrated in ethanol. Subsequently, the samples were critical point dried in a Leica EM CPD300 dryer and sputter coated in a Bal-tec SCD-050 sputter-coater. They were visualized using a Zeiss SIGMA VP-FESEM scanning microscope.

Embryos were fixed briefly in 4% PFA and washed in PBST. They were then blocked in PBST plus 1% bovine serum albumin for 1 h, and incubated with 2 mg/ml FITC-coupled DBA-lectin (Life Technologies) at 4°C overnight. Finally, the embryos were washed three times in PBST and visualized using a Zeiss Discovery microscope and a QImaging Micropublisher camera.

Embryos at the three- to five-somite stage were briefly fixed in 4% PFA followed by the removal of the extra-embryonic membranes for genotyping. The embryos were then flat-mounted, and photos were taken using a Zeiss Discovery microscope and a QImaging Micropublisher camera. The length (L) and area (A) of the embryos were measured in FIJI. LWR was calculated as LWR=L²/A. Statistical analyses were performed in Prism.

We thank Drs Zhi-chun Lai and Yingwei Mao for critically reading the manuscript, Drs McMahon and Ornitz for Shh and Fgf8 probes for RNA in situ hybridization and Dr Magnuson for providing the Sox17GFPCre mice. We thank Penn State Huck Microscopy Facility for providing histology and electron microscopy services, Genomics Facility for DNA sequencing service and Flow Cytometry Facility for sorting transfected K-562 cells. We thank the Animal Resource Program at Penn State for animal care.