Introduction

Failure of neural tube closure (NTC) leads to severe birth defects, such as anencephaly and spina bifida, more commonly known as neural tube defects (NTDs). The prevalence statistics indicate that NTDs occur in 5 per 10,000 live births in the United States (Wallingford et al. 2013) and at a considerably higher rate in China (138.7/10,000) (Li et al. 2006). In higher vertebrates, the neural tube proceeds initially with the formation of the neural plate, followed by the elevation, bending and fusion of the neural folds to become a closed tube. These complicated processes are regulated by multiple mechanisms, including coordination of the activity of Rho kinase (ROCK) and the planar cell polarity (PCP) pathways. More than 400 different mouse genes with diverse cellular functions have been implicated as genetic risk factors for NTDs (Harris and Juriloff 2010). At the present time, none of the mouse NTD candidate genes adequately informs our understanding of the genetic basis underlying the etiology of human NTDs.

Two morphogenetic processes, neural plate apical constriction and convergent extension, are required for neural tube closure (Nishimura et al. 2012). PCP molecules have been recognized as linking these two morphogenetic processes in the neural tube (Copp and Greene 2010; Nishimura et al. 2012). Genetic modification of PCP core genes in mouse models, such as Vangl2 and Celsr1, lead to severe NTDs, indicating that there is a strong connection between PCP signaling and risk for NTDs (Murdoch et al. 2014). PCP-dependent convergent extension is crucial for NTC, and loss of function (LoF) alleles of Vangl2 (Greene et al. 1998), Celsr1 (Curtin et al. 2003), Frizzled3 (Fz3) (Wang et al. 2006), Frizzled6 (Fz6) (Wang et al. 2006), Dvl1 and Dvl2 (Etheridge et al. 2008) cause craniorachischisis. Recently, SHROOM3 was demonstrated to function downstream of the PCP pathway to regulate myosin II distribution and cellular behavior to promote NTC in mouse models (McGreevy et al. 2015). SHROOM3 is a cytoskeleton regulating protein that contains three evolutionarily conserved domains. The ASD1 and ASD2 domains lie in the N- and C-terminal, while the PDZ domain is located in the central region. Previous studies showed that neuroepithelial cells in homozygous SHROOM knockout mouse embryos failed to convergence in the dorsal midline resulting in exencephaly and spina bifida phenotypes (Hildebrand and Soriano 1999). In Xenopus embryos, SHROOM induces apical constriction to regulate the formation of hinge points during NTC (Haigo et al. 2003). In zebrafish embryos, knock down of SHROOM3 blocks apical constriction (Ernst et al. 2012). In polarized epithelia, induced expression of SHROOM elicits apical constriction and reorganization of the actomyosin network (Hildebrand 2005). Rho/ROCK signaling promotes actomyosin contractility (Rath and Olson 2012), which is utilized for controlling cell shape and behavior. Exposure to inhibitors of Myosin II or ROCK resulted in NTDs due to failed apical constriction and convergent extension (Kinoshita et al. 2008). SHROOM3 can directly interact with Rho kinase (ROCK) through the ASD2 domain (Nishimura and Takeichi 2008), which is highly conserved in the SHROOM gene family (Hildebrand and Soriano 1999). Finally, the interaction between SHROOM3 and ROCK plays a pivotal role during neural tube morphogenesis (Das et al. 2014).

Animal models and human NTD cohort studies have made it clear that SHROOM3 is involved in NTD etiology (Haigo et al. 2003; Hildebrand and Soriano 1999; Lemay et al. 2015). The importance of SHROOM3 in NTC is beyond dispute, and a recent study identified two protein truncating de novo mutations in SHROOM3 in human NTD patients (Lemay et al. 2015). However, the genetic contributions of other SHROOM family members in NTDs have remained elusive. The excess of damaging variants, including LoF (loss of function, including splice acceptor/donor, stop gained/lost, initiator codon and frameshift indels) and D-mis (deleterious missense) variants, were used in identifying causal genes in birth defect diseases, such as congenital heart disease (Homsy et al. 2015) and craniosynostosis (Timberlake et al. 2016). To better understand human NTD risk-associated damaging variants of SHROOM family genes and their underlying biological mechanisms, we systematically investigated the mutation patterns of SHROOM1-4 in a Chinese population composed of 343 NTD and 206 control-genomic DNA samples.

Materials and methods

The Chinese NTD study population

We implemented a two-stage design for this study. Subjects in the capture sequencing stage, consisting of 343 aborted fetuses with NTDs and 206 blood samples from healthy controls, were all ethnically Han Chinese, as previously described (Lei et al. 2010; Shi et al. 2012; Yang et al. 2013). This study was approved by the Ethics Committee of the School of Life Sciences, Fudan University. All of these samples were received with informed parental consent.

DNA sequencing, genotyping and data analysis

The genomic structure of human SHROOM1-4 genes were determined using NCBI GenBank (NM_133456, NM_001649, NM_020859 and NM_020717). The 5′-UTR, 3′-UTR, and coding regions in SHROOM1-4 were detected by targeted next-generation sequencing. Sequencing was performed on Illumina GAII (Illumina, San Diego, CA, USA) platform using the paired-end program. Variant calling and annotation was performed using previously established methods (Qiao et al. 2016). Coding variants were classified as synonymous, missense, LoF (loss of function, including splice acceptor/donor, stop gained/lost, initiator codon and frameshift indels) and others using Variant Effect Predictor (VEP) (McLaren et al. 2010). PolyPhen-2 (Adzhubei et al. 2010) and SIFT (Kumar et al. 2009; Sim et al. 2012) were utilized to predict the genomic consequences of missense mutations via VEP tools. The missense variants that were predicted to be damaging by PolyPhen-2 or deleterious by SIFT were annotated as deleterious missense variants (D-mis). Two variants, identified by targeted capture sequencing and considered likely to be damaging, were selected to be further validated by Sanger sequencing.

Binomial testing as previously used in congenital heart disease study (Homsy et al. 2015), was performed for burden analysis of rare coding variants by comparisons of NTD cases and controls. Statistical analyses were performed by R (http://www.R-project.org).

Plasmid construction

Myc-DDK tagged human SHROOM2 (NM_001649) cDNA was purchased from OriGene Technologies (Beijing, China). Human ROCK1 (NM_005406) was purchased from YouBio (Hunan, China). The ASD2 domain (aa1317-1611) was PCR amplified from SHROOM2 cDNA and inserted into the pCMV6-Entry empty vector at the SgfI-MluI restriction sites. All mutations in SHROOM2 were generated using a QuickChange Site-Directed Mutagenesis Kit (Stratagene). The SHROOM binding domain (SBD) of ROCK1 was inserted into pCMV6-HA empty vector at the SgfI–MluI restriction sites. For the CheckMate™ Mammalian Two-Hybrid system, the bait constructs were PCR amplified from SHROOM2 and ROCK1 cDNAs. The ASD2 domains (aa1317-1611) of SHROOM2 wildtype and mutant sequences were subcloned into pBIND vectors and fused in-frame with GAL4 at SalI and XbaI restriction sites. The SBD located in ROCK1 was cloned into the pACT vector and fused with VP16 in-frame at the same restriction sites as the pBIND vector.

All plasmids used in this study were confirmed by DNA sequencing, and the primers used for plasmid construction are listed in Table S1.

Western blot and Co-immunoprecipitation

Myc-DDK tagged SHROOM2-ASD2 wildtype or mutant was co-transfected into HEK293T cells with pCMV6-GFP empty vector. Twenty-four hours later, cells were harvested with lysis buffer [50 mM of Tris (pH7.4), 150 mM of NaCl, 1% NP-40, 0.25% sodium deoxycholate, protease inhibitors]. Samples were loaded and separated on a 10% SDS-PAGE gel and were then transferred to a polyvinylidenedifluoride (PVDF) membrane (Millipore). After blocking for 1 h with 5% non-fat milk, the membrane was incubated with the primary antibody at 4 °C overnight. ECL was used to visualize the protein bands after incubating with secondary antibody for 1 h at room temperature. GFP empty vector was used as a transfection efficiency control.

A co-immunoprecipitation assay was further performed to evaluate whether there is an interaction between SHROOM2 and ROCK1. Myc-DDK tagged SHROOM2-ASD2 wildtype or mutant was co-transfected into HEK293T cells with HA-tagged ROCK1-SBD. Lysates were produced in lysis buffer and incubated with anti-Myc agarose beads (Abmart) at 4 °C for 2 h. After washing five times with lysis buffer, the beads were collected by centrifugation, eluted in 1× SDS loading buffer, boiled at 100 °C for 10 min, and analyzed by western blotting. Band density of target proteins was quantitatively analyzed using Image J. The band density ratios of IB:HA (upper) relative to IP:Myc were used to measure the binding ability to ROCK1 between wildtype and mutants. Four independent experiments were performed and representative results were shown.

CheckMate™ Mammalian Two-Hybrid system

The CheckMate™ Mammalian Two-Hybrid System (Promega), which contains three plasmids, was used for detecting protein–protein interactions. The pGL4.3luc vector is a firefly luciferase reporter plasmid containing five GAL4 binding sites. The pACT vector has a VP16 domain that can activate the firefly luciferase expression when it binds to GAL4. The pBIND vector, which has a GAL4 DNA-binding domain, is used to express Renilla luciferase and normalize the transfection efficiency. When two proteins of interest are fused with GAL4 and VP16 separately, firefly luciferase activity will increase compared to the negative control, if such a positive interaction exists. To test the interaction between SHROOM2 and ROCK1, HEK293T cells were co-transfected with the above-indicated plasmids using Lipofectamine 2000 (Invitrogen). Twenty-four hours post-transfection, cells were harvested for a luciferase assay, and the ratios of firefly and Renilla luciferase activities were determined.

Results

Enrichment of rare coding variants in SHROOM genes

In this study, we sequenced the coding regions of SHROOM1-4 from 343 NTD and 206 normal control samples by targeted next-generation sequencing, and detected 161 and 62 rare coding variants in the NTD samples and controls, respectively (MAF < 0.01). To interrogate the rare coding variants of SHROOM family genes, a burden test was performed based on a binomial test. The results indicated that rare coding variants were markedly increased in NTDs compared with controls (enrichment = 1.56, p = 2.9 × 10−3) (Table 1). We also observed significant enrichment of rare missense variants of 1.65 (p = 6.8 × 10−3) in SHROOM1-4. To evaluate the genomic effects of the mutations, we grouped coding variants into synonymous, missense, D-mis, LoF, damaging (LoF and D-mis) and other mutation types. One LoF variant, frameshift variant p.N594fs in SHROOM3, was identified in NTD cases, but LoF variants were not significantly enriched in the NTD cohort. We subsequently analyzed the total burden of rare damaging variants in SHROOM1-4. No significant enrichments of D-Mis variants were found in SHROOM1, SHROOM3 and SHROOM4. However, we detected significant 4.5-fold enrichment of D-Mis variants in SHROOM2 (p = 0.04).

Table 1 Burden analysis by comparisons of NTD cases vs. controls
Full size table

Rare damaging variants of SHROOM2 with diverse NTD phenotypes

Among all 42 detected SHROOM2 variants in the 343 NTD samples, 15 rare damaging mutants were identified from 15 NTD samples (Table 2). The 15 mutants were distributed among eight loci (Fig. 1). All of these damaging mutants are very rare in the ExAC database, with MAF < 0.001, and four of the variants were novel and did not exist in ExAC (Table 2). Samples with rare damaging variants were observed to have different NTD phenotypes (Table 2). Sample D56 with p.S1219R, and samples D42 and D44 with p.R1557H were from fetuses with anencephaly. Sample D20 with p.A119T, T9 with p.G1349V and D117 with p.R1557H were spina bifida cases. Sample SY3180 and SZNT6 with p.A1331S, D125 with p.A1470T and D135 with p.R1557H displayed encephalocele phenotypes. Five other patients were observed to have mixed or multiple NTD phenotypes.

Table 2 Rare damaging mutations in SHROOM2 with diverse NTD phenotypes
Full size table
Fig. 1
figure 1

Mapping of rare damaging mutations affecting SHROOM2

Full size image

Damaging variants of SHROOM2 significantly decreased interactions with ROCK1

SHROOM proteins interact with ROCK1 through its conserved ASD2 domain to direct its subcellular localization and participate in regulating cytoskeletal dynamics (Zalewski et al. 2016). The two damaging variants (p.A1331S and p.R1557H) with the highest frequencies (4/15 and 5/15, respectively) were located in the ASD2 domain (Fig. 1). These two variants were confirmed by Sanger sequencing (Fig. 2a, b). We further evaluated whether these D-Mis variants of SHROOM2 were functional variants using western blot and the CheckMate™ Mammalian Two-Hybrid system. These two investigated SHROOM2 D-Mis variants (p.A1331S and p.R1557H) have no significant effects on protein expression (Fig. 2c). However, the interactions of these two SHROOM2 D-Mis variants with ROCK1 were significantly decreased using the CheckMate™ Mammalian Two-Hybrid system (Fig. 2d). Additionally, the interaction between SHROOM2 ASD2 domain and ROCK1 SBD was validated by co-immunoprecipitation (Fig. 2e), and both SHROOM2 mutations significantly decreased their interaction with ROCK1 SBD (Fig. 2f). This suggests that deleterious variants of SHROOM2 attenuate their interaction with ROCK1 which can have significant consequences during NTC.

Fig. 2
figure 2

Deleterious variants of SHROOM2 attenuate their interactions with ROCK1. a, b Sanger sequencing results showed heterozygous missense mutants SHROOM2 c.3991G > T (p.A1331S) and SHROOM2 c.4670G > A (p.R1557H). c Western blot analysis of HEK293T cells transfected with SHROOM2-ASD2 wildtype, mutant or empty vector. GFP was co-transfected and used to normalize transfection efficiency. d The interactions of these two SHROOM2 D-Mis variants with ROCK1 were significantly decreased using the CheckMate™ Mammalian Two-Hybrid system. Various combinations of plasmids were transfected into HEK293T cells as indicated. e Co-immunoprecipitation (Co-IP) assay also validated the reduced interaction between SHROOM2 ASD2 domain and ROCK1 SBD. f Quantitative analysis of Co-IP results in e. Band density ratios were calculated and analyzed between wildtype and indicated SHROOM2 variants. Four independent experiments were performed and the asterisk indicates a statistical difference (*p < 0.05, **p < 0.01, ***p < 0.001) by Student’s t test

Full size image

Discussion

SHROOM3 mediates the PCP pathway (McGreevy et al. 2015) and plays an essential role in regulating cytoskeletal dynamics (Zalewski et al. 2016), which are critically important during the closure of the neural tube. Recently, de novo LoF mutations in SHROOM3 were experimentally associated with the development of NTDs, suggesting that these variants may also contribute to the genetic etiology of severe human NTDs (Lemay et al. 2015). However, the genetic contributions of other SHROOM family members to human NTDs have not been previously reported. In this study, we systematically investigated the mutation patterns of SHROOM1-4 in a Han Chinese NTD cohort. Our results suggest that rare coding variants in the SHROOM gene family were markedly increased in NTD patients, and that rare damaging variants of SHROOM2 might also contribute to the risk for human NTDs by adversely affecting interactions of SHROOM2-ROCK1, which are involved in both the PCP pathway and cytoskeletal regulation.

PCP molecules control the morphogenetic processes of apical constriction and convergent extension, which are required for NTC (Copp and Greene 2010; Nishimura et al. 2012). The association between rare mutations of PCP-related genes and human NTDs has been well studied (Allache et al. 2012; Bosoi et al. 2011; De Marco et al. 2012; Kibar et al. 2007; Robinson et al. 2012; Seo et al. 2015). Previous animal models have demonstrated that homozygous Shroom3 mutant embryos exhibited exencephaly (Hildebrand and Soriano 1999). Shroom3 interacts genetically with the PCP components Vangl2 and Wnt5a, and depletion of Shroom3 and Vangl2 or Wnt5a can drastically increase the penetrance and severity of NTDs (McGreevy et al. 2015). The PCP pathway is highly dosage sensitive, and over- or under-expression of PCP core genes in zebrafish and Xenopus embryos results in convergent extension defects (Roszko et al. 2009; Wallingford 2005, 2013). Our results identified a LoF variant p.N594fs in SHROOM3. Despite the dysregulation of the PCP pathway induced by rare LoF mutants of SHROOM3 might contribute the risk of human NTDs; there is no significant enrichment of D-Mis variants in SHROOM3.

SHROOM2 is significantly enriched in D-Mis variants in our study, suggesting that it is likely to be pathogenic for NTDs. ROCK1 is a serine/threonine kinase whose activity drives cytoskeletal regulation associated with apical constriction (Das et al. 2014; Haigo et al. 2003; Hildebrand 2005). SHROOM3 can regulate NTC through interactions with the SBD domain of ROCK1 in vertebrates (Das et al. 2014; Hildebrand 2005), indicating that SHROOM2 may have similar molecular activities. Moreover, we verified that these two damaging variants (p.A1331S and p.R1557H) in SHROOM2 significantly decreased the interaction between SHROOM2 and ROCK1 (Fig. 2). Collectively, our study provided strong direct evidence that rare damaging mutations in SHROOM2 impaired SHROOM2-ROCK1 binding, which would suggest that rare damaging variants in SHROOM2 result in the misregulation of ROCK1, leading to an alteration of cytoskeletal remodeling.

Despite the importance of demonstrating SHROOM-mediated remodeling in the etiology of NTDs using a mouse model, the genetic contributions and the underlying molecular mechanisms of SHROOM gene family members in human NTD patients were lacking. We systematically investigated the mutation patterns of SHROOM1-4 via case–control burden analysis in a Han Chinese population, and further performed in vitro functional confirmation studies. The results suggested that damaging variants in SHROOM2 are likely to play major roles in contributing to NTD risk. We provided molecular insight into the effects of rare damaging variants in SHROOM2-ROCK1 binding, indicating that rare damaging variants of SHROOM2 might lead to the misregulation of PCP signaling and alterations in cytoskeletal remodeling that contribute to the risk of human NTDs.