Golgb1 regulates protein glycosylation and is crucial for mammalian palate development

In mammals, palatal shelves arise from the oral side of the maxillary prominences during embryogenesis and initially grow downward, flanking the developing tongue (Bush and Jiang, 2012; Lan et al., 2015). At a specific developmental stage, the palatal shelves re-orient to the horizontal position above the tongue, grow towards the midline and subsequently fuse with each other to form the roof of the oral cavity. Concurrent to palate development, the surrounding craniofacial structures undergo significant growth and morphogenesis. Cleft palate could result from intrinsic defects in palatal shelf growth, elevation or fusion, or from failure of the palatal shelves to contact each other at the right developmental stage due to malformation of surrounding structures (Bush and Jiang, 2012; Gritli-Linde, 2007; Hilliard et al., 2005; Lan et al., 2015). Although extensive gene knockout studies in mice in the last 25 years have shown that all major signaling pathways, including Bmp, Fgf, Notch, Shh, Tgfβ and Wnt pathways, play crucial roles in palate development, mutations in known palate development regulators have been found in <30% of cleft palate patients (Bush and Jiang, 2012; Gritli-Linde, 2007; Hilliard et al., 2005; Lan et al., 2015; Moore and Persaud, 1998). Many more genes with crucial roles in palate development remain to be discovered.

The Golgi apparatus is an intracellular organelle composed of a series of flattened membrane structures and it functions as a carbohydrate synthesis and glycosylation factory at the crossroads of the secretory pathway (Mellman and Warren, 2000). Almost all signaling molecules and extracellular matrix (ECM) components are processed through the Golgi prior to arriving at the cell surface (Mellman and Warren, 2000; Zhong, 2011). In addition to containing a large number of enzymes that are responsible for catalyzing protein and lipid modifications, the Golgi houses many proteins that are involved in packaging, tethering and transporting secretory vesicles (Linstedt and Hauri, 1993; Munro, 2011; Puthenveedu and Linstedt, 2001; Short et al., 2005; Zhong, 2011). Dozens of human developmental disorders and diseases have been associated with mutations in Golgi-associated proteins (Freeze and Ng, 2011; Smits et al., 2010). Remarkably, several mutations in genes encoding vesicle transport proteins cause craniofacial and skeletal disorders with distinct phenotypes (Follit et al., 2008; Freeze and Ng, 2011; Kim et al., 2012; Smits et al., 2010), which highlights the importance of understanding Golgi-mediated cell biological processes in the context of development (Freeze and Ng, 2011; Kim et al., 2012; Smits et al., 2010).

Golgb1, also known as giantin, is a Golgi-associated large transmembrane protein (Linstedt and Hauri, 1993). Golgb1 has been found to interact with the transcytosis-associated protein p115 (also known as Uso1), a cytosolic protein involved in tethering endoplasmic reticulum (ER)-derived vesicles to the Golgi in mammalian cells as well as in yeast (Cao et al., 1998; Koreishi et al., 2013; Puthenveedu and Linstedt, 2001). Mouse embryos deficient in p115 exhibit early embryonic lethality and Golgi fragmentation in trophoblast cells (Kim et al., 2012). Although no Golgb1 mutant mice have been reported, Katayama et al. recently identified a 10-bp insertion in exon 13 of the Golgb1 gene in rats carrying a spontaneous osteochondrodysplasia (ocd) mutation (Katayama et al., 2011). The ocd homozygous mutant rats exhibit disproportionate dwarfism, systematic edema, cleft palate and neonatal lethality (Katayama et al., 2011), but whether these developmental defects are caused by loss of Golgb1 function has not been validated and the cellular functions of Golgb1 in vivo remain to be characterized. In this study, we identify a splice donor site mutation in the Golgb1 gene that completely co-segregates with a cleft palate phenotype in a new chemically induced mutant mouse line. We validate the hypothesis that loss of Golgb1 function causes cleft palate in mice through generation and analysis of mice carrying additional Golgb1 loss-of-function alleles by using CRISPR/Cas9-mediated genome editing. Furthermore, we demonstrate that Golgb1 plays a crucial role in protein glycosylation in the developing palatal mesenchyme cells. Our results provide novel insights into Golgi biology and the molecular and cellular mechanisms of palate development.

We performed an N-ethyl-N-nitrosourea (ENU) mutagenesis screen in mice to identify new genes that are crucial for mammalian organogenesis. Using a three-generation screening strategy designed to recover autosomal recessive mutations (Fig. S1A), we screened embryonic day (E) 16.5-E18.5 embryos for structural defects in 12 G1 families. The ENU1483 line produced embryos with cleft palate at high frequency (Fig. S1B). Single nucleotide polymorphism (SNP) linkage analysis mapped the causal mutation to proximal chromosome 16, between 29 Mb and 50 Mb (Table S1). Loss-of-function mutations in two genes, Gsk3b and Dlg1, inside this chromosomal region have been associated with cleft palate in mice (He et al., 2010; Rivera et al., 2009), but targeted sequencing of all exons and exon-intron boundaries of these two genes did not find any mutation in the ENU1483 line.

Analysis of whole-exome sequencing results revealed that the cleft palate pups were homozygous for novel single nucleotide changes in three closely linked genes within the chromosome 16 region: a G-to-A transition at the intron-9 splice donor site in the Golgb1 gene (Golgb1^ivs9+1G>A), a C-to-T transition that causes a S793F missense mutation in the Itgb5 gene (Itgb5^S793F), and a G-to-A transition that causes a G179S missense mutation in the Adcy5 gene (Adcy5^G179S) (Table S2). Genomic PCR followed by Sanger sequencing confirmed that all three mutations were ENU induced and absent in the parental A/J and FVB/NJ strains. Because neither Itgb5 null mice nor Adcy5 null mice have palate defects (Huang et al., 2000; Iwamoto et al., 2003) and since ocd/ocd rats exhibit syndromic cleft palate associated with a Golgb1 disruption (Katayama et al., 2011), the Golgb1^ivs9+1G>A mutation became a prime candidate for causing cleft palate in the ENU1483 mice.

The Golgb1 gene contains 22 exons, with the ATG start codon located in exon 2 (Fig. 1A). We performed RT-PCR using two different pairs of primers flanking exon 9 and exon 10 sequences and found that the ENU1483 mutant tissues yielded larger PCR fragments than the wild-type samples (Fig. 1B). Sequencing the RT-PCR products revealed that the mutant Golgb1 mRNAs contained an extra 435 nucleotides from the 5′ region of intron 9, indicating splicing of the mutant Golgb1 pre-mRNAs from a cryptic splice donor site inside intron 9 (Fig. 1A).

The mutant Golgb1 mRNAs contain an in-frame stop codon just 20 nucleotides 3′ to the disrupted splice site (Fig. 1A). Western blot analysis using a polyclonal antibody against the N-terminal 469 residues of human GOLGB1 protein (Abcam ab24586) detected Golgb1 protein specifically in the wild-type samples but no product in the homozygous mutant embryos (Fig. 1C). Immunofluorescence staining also revealed an absence of Golgb1 protein in the homozygous mutant embryos (Fig. 2A-F) although the mutant cells appear to have intact Golgi apparatus, as marked by localized staining of another well-known Golgi marker GM130 (also known as Golga2) (Nakamura et al., 1995) (Fig. 2A-F).

We then performed transmission electron microscopy (TEM) analysis and found that the Golgb1^ivs9+1G>A homozygous mutant embryonic palate cells had Golgi structures with a normal appearance (Fig. 2G,H). These data are consistent with previous reports from Golgb1 (giantin) knockdown studies in tissue culture cells showing that Golgb1 is not required for Golgi structural integrity (Petrosyan et al., 2012, 2014; Puthenveedu and Linstedt, 2001; Wong and Munro, 2014).

To validate the hypothesis that the cleft palate defect was due to loss of Golgb1 function, we first purchased target-verified mouse embryonic stem cell (ESC) clones containing the ‘knockout-first and conditional ready’ Golgb1^tm1a(EUCOMM) gene-targeted allele from the International Knockout Mouse Consortium and generated high-percentage chimeric founder male mice through blastocyst microinjection. However, none of the chimeric mice gave germline transmission of the correctly targeted Golgb1^tm1a(EUCOMM) allele. Recently, an RNA-guided, nuclease-mediated genome-editing technology, based on the Archaeal type II CRISPR (clustered regularly interspaced short palindromic repeat)/Cas9 system, has been successfully adapted for genome editing in mammals (Jinek et al., 2012; Mali et al., 2013; Ran et al., 2013; Wang et al., 2013). We thus used the CRISPR/Cas9 system to create independent Golgb1 mutant alleles in mice. We designed synthetic guide RNAs (sgRNAs) to target exon-10 and exon-13 regions, respectively, of the Golgb1 gene (Fig. 3A) and tested the sgRNA/Cas9 constructs by transfection of mouse kidney MK4 cells. SURVEYOR assay of the transfected cells showed that the E13-1 and E10 sgRNAs resulted in highly efficient editing of the target sites (Fig. 3B). The E13-1 and E10 sgRNAs were subsequently co-injected with Cas9 mRNAs into FVB/NJ zygotes for generation of Golgb1 mutant mice. Of 34 founder mice, nine carried various indel mutations that cause a frame shift in either exon 10 or exon 13 (Fig. 3C). In addition, one founder mouse carrying an 11 kb deletion from exon 10 to exon 13, resulting in an in-frame deletion of 137 amino acid residues in the protein product, was obtained (Fig. 3C). We crossed the founder mice to FVB/NJ wild-type mice and obtained germline transmission of five distinct Golgb1 alleles. F1 male mice carrying four distinct alleles (one exon-10 indel and three exon-13 indel mutations), respectively, were crossed to Golgb1^ivs9+1G>A heterozygous female mice and embryos harvested at E16.5 to E18.5 for phenotypic analysis. All trans-heterozygous embryos had cleft palate (Fig. 4B,D).

We next generated and analyzed embryos homozygous for a frame-shift indel mutation in exon 10 as well as those homozygous for each of two exon-13 mutations that deleted 41 bp and 25 bp, respectively. Immunofluorescence staining of E13.5 homozygous embryos confirmed the absence of the Golgb1 protein in these mutants (Fig. 4F). Histological analysis at E15.5 and E16.5 showed that the homozygous mutants had cleft palate (Fig. 5F,H). These results confirm that loss of Golgb1 function causes cleft palate in mice. Mice homozygous for the 11 kb deletion from exon 10 to exon 13 survived postnatally and appeared grossly normal, indicating that the in-frame deletion of 137 amino acid residues in the cytoplasmic domain of the Golgb1 protein did not significantly disrupt its cellular function.

To investigate the mechanism of cleft palate pathogenesis in Golgb1 mutant mice, we performed histological analysis of Golgb1^ivs9+1G>A, Golgb1^E13-25bp and Golgb1^E13-41bp homozygous mutant embryos and their control littermates from E12.5 to postnatal day (P) 0. The palatal shelves appeared similar in shape and size in the Golgb1 mutant and control littermates up to E13 (Fig. 5A,B). By E14.5, however, when the palatal shelves had already elevated and initiated fusion at the midline in wild-type and heterozygous embryos, the palatal shelves in homozygous mutant embryos remained vertically oriented (Fig. 5C,D). From E15.5 to E16.5, wild-type and heterozygous embryos showed extensive fusion of the palatal shelves throughout the anterior-posterior axis (Fig. 5E,G), but homozygous mutant embryos still had separate, vertically oriented, palatal shelves (Fig. 5F,H). In addition, the tongue did not descend to the floor of the oral cavity in the mutant embryos (Fig. 5D,F,H).

Because Golgb1 is ubiquitously expressed, we examined whether the Golgb1 mutant embryos have defects in early neural crest development. We carried out lineage tracing of migrating neural crest cells using the Wnt1Cre:R26R genetic lineage labeling system (Chai et al., 2000; Jiang et al., 2000) and examined the distribution of neural crest-derived cells from E9.5 to E12.5 in Golgb1^ivs9+1G>A homozygous mutant and control littermates. No differences in the patterns of labeled neural crest cells were detected in Golgb1^ivs9+1G>A homozygous mutant and control littermates (Fig. S2).

We analyzed whether Golgb1 mutant embryos had defects in palatal cell proliferation. No significant differences in the percentage of bromodeoxyuridine (BrdU)-labeled cells in the palatal epithelium or mesenchyme were detected in Golgb1^ivs9+1G>A homozygous mutant and control littermates at E12.5 and E13.5, respectively (Fig. S3).

Although palatal cell proliferation index was not significantly different between mutant and control embryos, careful examination of histological sections of embryos after E13.5, but prior to palatal shelf elevation, showed that the mutant palatal shelves became consistently smaller, but with higher cellular density, than those in control littermates. Quantification of DAPI-stained nuclei on serial sections of the embryos revealed that mutant palatal shelves had significantly increased cellular density throughout the anterior-posterior axis prior to palatal shelf elevation (Fig. 6C). Including Alcian Blue staining in the histology procedure showed that the intensity of Alcian Blue staining is obviously reduced in the mutant palatal mesenchyme in comparison with wild-type littermates (Fig. 6A,B), suggesting that the mutant palatal mesenchyme cells have a defect in the synthesis or accumulation of polysaccharides such as glycosaminoglycans (GAGs).

Hyaluronan (HA) is a high molecular mass GAG in the palatal mesenchyme ECM implicated in playing a role in palatal shelf elevation (Brinkley and Bookstein, 1986; Ferguson, 1988). Detection of HA using biotin-labeled HA-binding peptide (HABP) revealed that Golgb1 mutant embryos have significantly reduced HA accumulation throughout the anterior-posterior axis of the palatal mesenchyme, in comparison with control littermates, at E13.5 (Fig. 7). Similar to the Alcian Blue staining pattern, HA appears to accumulate at higher levels in the nasal side and in the hinge region of the palatal mesenchyme in control embryos, and HA accumulation in both regions was reduced in the mutant embryos. Since HA is a major component of the palatal mesenchyme ECM, the dramatic reduction in HA accumulation is likely to be at least part of the reason for the increased cell density and slower volume expansion of the palatal mesenchyme in the Golgb1 mutant embryos.

Although the reduced HA accumulation in the palatal mesenchyme indicates an intrinsic defect in the palatal shelf, it is possible that the failure of palatal shelf elevation in Golgb1 mutant embryos might be secondary to failure of the tongue to descend. To investigate this possibility, we isolated E13.5 embryos from intercrosses of Golgb1^E13-25bp and Golgb1^E13-41bp heterozygous mice, and cultured freshly microdissected maxillary explants in suspension culture in modified BGJb medium for 3 days as described previously (Abbott, 2000; Shiota et al., 1990; Takahara et al., 2004). Of eight litters of embryos, eight embryos were homozygous mutants and all eight mutant explants showed vertically oriented palatal shelves at the end of the 3-day culture (n=8; Fig. 8B,D). By contrast, of the 43 wild-type and heterozygous explants from these litters, all showed horizontally oriented palatal shelves and 33 of them (77%) showed at least partial fusion of the palatal shelves at the midline (Fig. 8A,C). These results indicate that Golgb1 plays an intrinsic role in the palatal mesenchyme in regulating palatal shelf elevation.

Although Golgi is the major site for GAG synthesis, HA is a unique GAG that is primarily synthesized at the plasma membrane (Fraser et al., 1997). However, HA synthases are integral membrane proteins synthesized in the ER and transported to the plasma membrane through the Golgi. We thus examined whether Golgb1^ivs9+1G>A homozygous mutant palatal mesenchyme (MEPM) cells have defects in vesicle trafficking through the ER-Golgi secretory pathway using adenovirus expressing a GFP-tagged temperature-sensitive mutant of the vesicular stomatitis protein VSVG^ts045-GFP. This mutant protein is unfolded and retained in the ER at 40°C but it can fold properly and transit through the Golgi network at 32°C. After incubating the infected cells at 40.5°C for 20 h, all VSVG^ts045-GFP molecules were retained in the ER in both control and Golgb1^ivs9+1G>A homozygous mutant MEPM cells (Fig. S4). After transferring the infected cells to 32°C for 20 min, the VSVG^ts045-GFP proteins were localized in the Golgi network (Fig. S4). No differences in the patterns of VSVG^ts045-GFP protein localization were detected in between control and Golgb1^ivs9+1G>A homozygous mutant cells, suggesting that Golgb1 function is not required for anterograde trafficking from ER through the Golgi.

It has been shown that HA synthase activity is regulated by protein glycosylation (Vigetti et al., 2009). To investigate whether protein glycosylation is affected during palate development in Golgb1 mutant embryos, we examined patterns of binding by GSII Lectin from Griffonia simplicifolia, which preferentially binds to terminal non-reducing N-acetyl-D-glucosamine, a glycan structure expected to be rare in properly modified mature glycoproteins (Puthenveedu et al., 2006; Smits et al., 2010). As shown in Fig. 9, most palatal mesenchyme cells in the control embryos did not have much GSII Lectin binding (Fig. 9A,C,E), but Golgb1 mutant palatal mesenchyme cells, particularly in the middle and posterior regions of the palatal shelves, exhibited increased GSII Lectin binding (Fig. 9D,F,G). The palatal regions that exhibit increased GSII binding in the mutant embryos appear to partly overlap with the regions that exhibited reduced HA accumulation, including proximal palatal mesenchyme at the nasal side and the hinge regions of the palatal shelves (compare Fig. 7C-G with Fig. 9C-G).

To investigate further the effect of loss of Golgb1 function on protein glycosylation during palate development, we used the Vector Lectin Kit III, which contains multiple lectins with selectivity for distinct glycosylation products, to test for differential binding to the palatal tissues in the Golgb1 mutant and control littermates. We found that Golgb1 mutant palatal mesenchyme had significantly increased binding to both peanut agglutinin (PNA) and Vicia villosa agglutinin (VVA). PNA, isolated from Arachis hypogaea (peanut), has high specificity for terminal β-galactose residues of glycoproteins (Pereira et al., 1976; Soga et al., 2015). In most mature glycoproteins, the β-galactose residues are further modified by sialylation (Gillespie et al., 1993). Thus, PNA binds to a broad range of cell surface glycoproteins following treatment with neuraminidase (sialidase) to remove terminal sialic acids (Novogrodsky et al., 1975). We found that neuraminidase treatment of sections of E13.5 wild-type embryos resulted in strong binding of the palatal mesenchyme to PNA (Fig. 10A,C,E), indicating that palatal mesenchyme cells normally produce a large amount of sialylated glycoproteins. The neuraminidase-treated sections of E13.5 mutant palatal shelves exhibited increased PNA binding compared with that in control littermates, particularly in the middle and posterior regions (Fig. 10B,D,F,G). The untreated mutant palatal mesenchyme also exhibited significant binding to PNA whereas little PNA binding was detected in the untreated wild-type palatal mesenchyme (Fig. 10A′-F′). Quantification of the mean fluorescence intensity in the palatal mesenchyme on serial sections showed that PNA binding is significantly increased throughout the anterior-posterior axis of the developing palatal shelves of the mutant embryos in comparison with control littermates (Fig. 10G′). VVA, isolated from Vicia villosa (hairy vetch) seeds, binds highly selectively to the Tn antigen, a simple monosaccharide structure composed of N-acetyl-D-galactosamine with a glycosidic α linkage to serine/threonine residues in glycoproteins (Ju et al., 2011; Tollefsen and Kornfeld, 1983). Tn antigen is produced at the first step of mucin-type O-glycosylation in the cis-Golgi and is usually further modified by sialylation or addition of additional sugar residues (Ju et al., 2011). We found that Golgb1 mutant embryos exhibit significantly increased VVA binding throughout the anterior-posterior axis of the palatal mesenchyme at E13.5, in comparison with control littermates (Fig. 11A-G). Following neuraminidase treatment, whereas sections of the wild-type E13.5 embryonic heads exhibited low VVA binding to the palatal mesenchyme, sections of the mutant palatal shelves exhibited significantly stronger VVA binding to palatal mesenchyme (Fig. 11A′-G′). Together, these results indicate that Golgb1 mutant embryos have defects at multiple steps in protein glycosylation during palate development. Remarkably, both PNA and VVA showed preferential binding to the nasal side of the palatal mesenchyme in the mutant embryos (Figs 10, 11), whereas PNA binding to the neuraminidase-treated samples showed a relatively uniform distribution throughout the palatal mesenchyme in both control and mutant embryos (Fig. 10), suggesting that Golgb1 is required for proper glycosylation of only a subset of glycoproteins that are preferentially synthesized in the nasal side palatal mesenchyme.

Most of our current knowledge regarding the cellular and molecular mechanisms of mammalian organogenesis has arisen through ‘reverse genetic’ approaches in which the gene of interest is mutated and the phenotype analyzed. However, reverse genetic approaches are inevitably biased towards studying genes with some prior information that suggests a detectable mutant phenotype. To circumvent the limitations of these approaches, several large-scale random mutagenesis screens have been performed in mice using both chemical and genetic mutagens (Araki et al., 2009; Barbaric et al., 2007; Boles et al., 2009; Hrabé de Angelis and Strivens, 2001; Sakuraba et al., 2005). Recent advances in genomic sequencing technologies have enabled rapid identification of mutations by using exome or whole-genome sequencing (Caruana et al., 2013). Because of the high rate of ENU-induced mutations throughout the genome, however, there are almost always multiple linked nucleotide changes around each candidate gene and it is vital to validate candidate causal genes by generation and analysis of independent mutant mouse lines. In this study, we used ENU mutagenesis to isolate a new cleft palate mutation, and SNP mapping and exome sequencing to identify the candidate causal mutation, followed by validating the requirement for Golgb1 function in palate development through generation of mice carrying independent mutant alleles using CRISPR/Cas9 genome editing. Despite several other ENU mutagenesis screens designed to identify recessive mutations affecting embryogenesis, no Golgb1 mutant mice have been reported. Thus, the combination of phenotype-driven mutation discovery with validation of causality using CRISPR/Cas9-mediated genome editing is an effective approach for uncovering novel molecular mechanisms underlying structural birth defects and other biological processes.

Golgb1 belongs to the golgin family of large coiled-coil proteins located at the cytoplasmic surface of the Golgi apparatus (Munro, 2011). Of all golgins, Golgb1 is unique because it contains a transmembrane domain at its C terminus, which anchors the protein to the Golgi membrane or COPI vesicles, and a p115-binding domain at its N terminus (Alvarez et al., 2001; Puthenveedu and Linstedt, 2001; Sapperstein et al., 1995; Sönnichsen et al., 1998). This unique structure suggested that it might play a role in vesicle tethering to the Golgi membranes in conjunction with p115 and GM130, another p115-interacting golgin (Alvarez et al., 2001; Lesa et al., 2000; Linstedt et al., 2000; Sönnichsen et al., 1998). Whereas several other golgins, as well as p115, have been shown to be essential for Golgi structural integrity in siRNA knockdown studies in cell culture assays, knockdown of Golgb1 mRNAs or protein did not cause disruption of Golgi structures or reassembly following mitosis (Koreishi et al., 2013; Puthenveedu and Linstedt, 2001). Recently, Kim et al. generated mice heterozygous for Uso1, encoding the p115 protein, and confirmed genetically that p115 function is required for Golgi structural integrity and cell survival (Kim et al., 2012). By contrast, rats homozygous for a frame-shift mutation in Golgb1 survive to birth (Katayama et al., 2011), indicating that Golgb1 is not essential for cell survival in vivo. Our finding that mice homozygous for either the ENU-induced splicing mutation or the CRISPR/Cas9-induced frame-shift mutations also survive to birth further validates this conclusion.

We found that mouse embryos deficient in Golgb1 consistently exhibit failure of palatal shelf elevation. Although failure of palatal shelf elevation has been described in several other mutant mouse strains (Jiang et al., 1998; Jin et al., 2010; Lan et al., 2004; Matsumura et al., 2011; Snyder-Warwick et al., 2010; Yu et al., 2010), the molecular and cellular mechanisms regulating palate shelf elevation are still not well understood (Bush and Jiang, 2012; Ferguson, 1988). It has been proposed that an intrinsic force is progressively generated to drive palatal shelf elevation and the chief component of this intrinsic force appears to be regionally specific accumulation of extracellular GAGs, predominantly HA (Ferguson, 1988). HA is capable of binding a large amount of water, and HA accumulation results in swelling of the ECM and a corresponding decrease in cell density (Brinkley and Bookstein, 1986; Ferguson, 1988). Brinkley and Vickerman showed that reducing HA in palatal mesenchyme through pharmacological enhancement of its degradation inhibited palatal shelf elevation in organ culture as well as in vivo (Brinkley and Vickerman, 1982). A palatal shelf elevation defect in mice homozygous for the Fgfr2^C342Y missense mutation was also associated with reduced HA accumulation in the palatal mesenchyme (Snyder-Warwick et al., 2010). We found that Golgb1 mutant embryos exhibit reduced HA accumulation and concomitantly increased cell density in the palatal mesenchyme by E13.5. Kikukawa et al. showed that the ocd/ocd neonatal rats had drastically reduced amounts of HA in the cartilage than control littermates (Kikukawa et al., 1990). Thus, Golgb1 plays a crucial role in HA accumulation in multiple tissues and reduction in HA accumulation in the palatal mesenchyme is likely to be an important factor in pathogenesis in Golgb1 mutant mice.

Recent cell culture studies suggest that Golgb1 mediates Golgi-targeting of specific glycosyltransferases. In particular, siRNA knockdown studies suggest that the C2GnT-M and C2GnT-L synthases in the O-glycosylation pathway use Golgb1 exclusively to target to Golgi membrane, whereas C1GalT1 and ST3Gal1 synthases use Golgb1 and GM130-Grasp65 (also known as Gorasp1) as alternative docking sites (Petrosyan et al., 2012, 2014; Wong and Munro, 2014). O-glycosylation in the Golgi starts with synthesis of the glycoconjugate GalNAc-α1-O-Ser/Thr, known as the Tn antigen. The most common modification of the Tn antigen is catalyzed by the core 1 β1,3-galatosyltransferase (C1GalT1) to produce the Core 1 disaccharide known as the T antigen (Ju et al., 2011). Remarkably, we found that Golgb1 mutant palatal mesenchyme cells exhibit significantly increased binding to both Tn- and T antigen-binding lectins, VVA and PNA, respectively. These results are consistent with the cell biological data showing that Golgb1 is involved in Golgi targeting of the Core 1 and Core 2 glycosyltransferases and identify an important role for Golgb1 in regulating the early steps of O-glycosylation during mammalian organogenesis.

Mice lacking another ubiquitously expressed member of the golgin family, Trip11, also known as GMAP-210, die neonatally with skeletal dysplasia (Smits et al., 2010). Human patients homozygous for loss-of-function mutations in TRIP11 also exhibit neonatal lethal skeletal dysplasia (Smits et al., 2010). Although Trip11 mutant mouse skin fibroblasts and chondrocytes showed impairment in protein glycosylation, no cleft palate phenotype has been noted in Trip11 null mice, indicating that Trip11 and Golgb1 have distinct cellular functions. Further comparative investigation of the distinct and complementary functions of Golgb1, Trip11 and other members of the golgin family in protein glycosylation and tissue morphogenesis in mutant mouse models will provide novel insights into the cellular and molecular mechanisms involving Golgi-associated proteins in human diseases.

All animal studies were performed in accordance with the guidelines set by the Institutional Animal Care and Use Committee (IACUC) at the Cincinnati Children's Hospital Medical Center. ENU mutagenesis screening was conducted as described previously (Coles and Ackerman, 2013). One line exhibiting autosomal recessive inheritance of isolated cleft palate, ENU1483, was established and maintained by backcrossing the carrier male to FVB/NJ female mice. Genomic DNA from 17 cleft palate pups from this line was used for linkage mapping using the mouse universal genotyping array (MUGA) SNP microarray (GeneSeek, Lincoln, NE, USA), which resulted in mapping the cleft palate mutation to proximal chromosome 16.

Whole exome sequencing and variant calling were conducted at the Cincinnati Children's Hospital Medical Center DNA Core Facility. Novel variants in the mapped chromosome 16 region were verified by sequencing genomic PCR products at the Cincinnati Children's Hospital DNA Core.

The PCR primers Golgb1RT-1F (5′-GTAGGATCCTGCTGCAGAAGGAGCTAACG-3′) and Golgb1RT-2R (5′-GATGGATCCAGAGCTAAGGAGCAAGG-3′) were used to amplify a fragment of 704 bp from exon 8 to exon 11. Another forward primer Golgb1RT-3F (5′-AAGGTCGACCTGGAGACACAGTACAGT-3′) was used in combination with Golgb1RT-2R to amplify a fragment of 405 bp from exon 9 to exon 11. RT-PCR products were gel-purified and sequenced at the Cincinnati Children's Hospital DNA Core.

We constructed one guide RNA (sgRNA) targeting exon 10 (target sequence 5′-GTCGCCCTCATTATTAACATC-3′) and three sgRNAs targeting exon 13 (target sequences exon13-1: 5′-GCTGCACTTCTCTACGGTCG-3′, exon13-2: 5′-CACCGCTGGTACTACTCTGCTCAAC-3′, and exon13-3: 5′-CACCGCTGCAGGGCAAGAGGCCCGA-3′) using the online CRISPR Design Tool (http://crispr.mit.edu). The target sequences were subcloned into the PX459 mammalian Cas9-expression vector (Addgene, #48139) and tested by transfection into cultured immortalized mouse embryonic kidney (MK4) cells (Ran et al., 2013; Valerius et al., 2002). MK4 cells were provided by Dr Steven Potter at Cincinnati Children's Hospital Medical Center from the original stock. A previously reported sgRNA for targeting the Tet2 gene was used as positive control (Wang et al., 2013). The targeting efficiencies were estimated using SURVEYOR assays (Transgenomics). Subsequently, sgRNAs targeting exon 10 and exon 13, respectively, were synthesized in vitro and co-injected with Cas9 mRNAs into zygotes from FVB/NJ inbred mice to generate gene-targeted mice in the CCHMC Transgenic Animal and Genome Editing Facility. Transgenic founder mice were identified by sequencing PCR products and were bred to FVB/NJ inbred mice to test for germline transmission. G1 mice heterozygous for indel mutations that cause a frame shift in exon 10 or exon 13 of the Golgb1 gene were crossed to the Golgb1^ivs9+1G>A heterozygous mice to generate trans-heterozygotes or were intercrossed in order to analyze phenotypes of homozygotes.

Immunofluorescence staining was performed using paraffin or frozen sections following standard protocols (Xu et al., 2014). Primary antibodies used were: rabbit anti-giantin (Abcam, ab24586; lot # GR6988-3; 1:300), mouse anti-GM130 (BD610822; lot # 29000; 1:500), rabbit anti-ERGIC-53 (Sigma, E1031; 1:250). For detection of HA, biotin-labeled HABP (Millipore, 385911-50UG; lot# D00175939; 1:200) staining was detected using Texas Red-conjugated streptavidin (Vector, SA-5006; 1:200). For lectin-binding assays, Alexa Fluor 594-conjugated Lectin GSII (Life Technologies, L-21416; 1:200), biotin-conjugated Lectin VVA (Vector, B-1235; 1:200), or fluorescein-conjugated PNA (Vector, FL-1071; 1:200) were used. Slides were mounted with ProLong Diamond Antifade Mountant (ThermoFisher). Nuclear density and fluorescence intensity over the entire palatal shelf area were recorded and analyzed separately for the anterior, middle and posterior regions of the palatal shelves using ImageJ (http://fiji.sc/Fiji). The statistical significance of the results was assessed using Student's t-test.

For TEM, embryos were dissected at E13.5. Palatal shelves and limbs of each embryo were fixed with 3% glutaraldehyde, washed with sodium cacodylate buffer, post-fixed with 1% OsO₄, processed through an ethanol gradient, washed in serial dilutions of plastic (Ladd Research, LX-112)/propylene oxide mixture, and moved to 60°C for 3 days. The tissue samples were sectioned at 100 nm thickness using a Leica EM UC7 ultramicrotome and put onto 200-mesh copper grids for examination under a Hitachi Model 7650 transmission electron microscope (EM Core of Cincinnati Children's Hospital Medical Center).

Embryos were dissected at the desired stages (noon of date of vaginal plug detection was counted as E0.5) and processed for histology analysis as described (Baek et al., 2011; Xu et al., 2015).

For western blot assays, three pairs of microdissected palatal shelves (E14.0) were pooled and lysed in RIPA buffer containing proteinase inhibitors (Santa Cruz, SC-24948) and experiments were repeated three times. The supernatant was mixed with Laemmli buffer and separated by SDS-PAGE on a 4-12% Mini-PROTEAN gel (Bio-Rad). Western blots were incubated with rabbit anti-giantin (Abcam, ab24586; 1:3000) or anti-β-actin (Santa Cruz, SC-47778, lot # 10710; 1:3000) antibody followed by washing and incubation with HRP-conjugated secondary antibody (Santa Cruz, SC-3020). The signals were detected using Immobilon Western HRP peroxidase solution and Luminol reagent (EMD Millipore) and exposed to X-ray film (Associated Metals, Ann Arbor, MI, USA).

Palatal cell proliferation was analyzed as described previously (Lan et al., 2004; Xu et al., 2015; Zhou et al., 2013). The numbers of BrdU-labeled nuclei and total nuclei were recorded for five independent control and mutant littermate pairs. Student's t-test was used to analyze the differences in the datasets and P<0.05 was considered statistically significant.

Skeletal preparations of E18 and P0 pups were performed as described previously (Liu et al., 2013; Zhou et al., 2013).

Freshly dissected maxillary explants of E13.5 embryos were cultured as described previously (Abbott, 2000; Shiota et al., 1990; Takahara et al., 2004). Briefly, three to five explants were placed in a 50 ml penicillin bottle with 8 ml BGJb medium (Invitrogen) supplemented with 2.8 mg/ml L-glutamine, 6 mg/ml BSA and 1% penicillin and streptomycin. The bottle was flushed for 2 min each day with a gas mixture (50% O₂, 45% N₂ and 5% CO₂), sealed air-tight, and incubated at 37°C on a Wheaton Mini Bench-Top roller bottle system at a speed of 25 rpm for three days. A small piece of tissue was taken from each explant for genotyping and explants were fixed individually for histological analyses.

Palatal shelves were dissected from each E13.5 mouse embryo, rinsed in cold PBS, transferred into a 12-well tissue culture plate, and treated with 0.5% trypsin and 2.5% pancreatin in ice-cold PBS for 30 min to remove the epithelium. The palatal mesenchyme tissues were then treated with 0.5% trypsin/EDTA for 10 min at 37°C, with pipetting up and down several times every 3-4 min to break into cell suspension, followed by culturing at 37°C in DMEM supplemented with 10% FBS (Invitrogen) and 1% penicillin and streptomycin. Two hours later, the culture wells were rinsed with PBS to remove unattached cells and tissue debris, and replaced with fresh medium. Upon ∼70% confluence, cells were trypsinized and split at 1:3 into 6-well tissue culture plates. Cells were cryopreserved upon 80% confluence.

VSVG^ts045-GFP adenovirus was provided by Dr Yanzhuang Wang at University of Michigan (Xiang et al., 2013). Wild-type and mutant primary MEPM cells were plated, respectively, in slide chambers at a density of 2×10⁶ cells/chamber. VSVG^ts045-GFP adenovirus was added at 0, 2.5×10⁸, 5×10⁸ and 10×10⁸ particles per chamber and incubated at 40.5°C (5% CO₂) for 20 h and shifted to 32°C (5% CO₂) for fixation at 0, 20 and 60 min after the temperature shift. The cells were co-immunostained for ERGIC-53 (also known as Lman1) (Sigma E1031, 1:250) and GM130 (BD610822, lot # 29000, 1:500).

We thank Dr Yanzhuang Wang for providing the VSVG^ts045-GFP adenovirus. We thank Cai Zhang and Takahisa Nakamura for help with adenovirus preparation, and Julie Serr and Lynessa McGee for laboratory assistance.