Introduction

BMPs are secreted molecules that were initially discovered as bone inducing factors and were subsequently shown to play numerous roles during embryogenesis (Bragdon et al., 2011). Bmp2, 4, 5, 6 and 7 are broadly expressed throughout embryogenesis, often in overlapping patterns (Danesh et al., 2009). Bmp2 and Bmp4 play important and non-redundant developmental roles, as mice homozygous for null mutations in either gene null die during early development (Winnier et al., 1995; Zhang and Bradley, 1996). Although Bmp5, Bmp6 or Bmp7 null homozygotes survive until birth or beyond, Bmp5;Bmp7 and Bmp6;Bmp7 double mutants die during embryogenesis, revealing functional redundancy within this subgroup (Kim et al., 2001; Solloway and Robertson, 1999).

BMPs are grouped into subfamilies based on sequence similarity and can signal as either homodimers or as heterodimers. The class I BMPs, BMP2 and BMP4, can heterodimerize with class II BMPs, consisting of BMPs 5-8 (Guo and Wu, 2012). Heterodimers composed of class I and class II BMPs show a higher specific activity than homodimers. For example, BMP2/7, BMP4/7 and BMP2/6 heterodimers are significantly more potent than any homodimer in multiple assays (Aono et al., 1995; Chauhan et al., 2024; Kaito et al., 2018; Valera et al., 2010). In vivo, endogenous Bmp2/7 heterodimers are essential to establish the dorsoventral axis in fish (Little and Mullins, 2009) and Drosophila (Shimmi et al., 2005). More recent studies have shown that endogenous BMP4/BMP7 and BMP2/7 heterodimers are required to generate full BMP activity in most or all tissues of mouse embryos (Kim et al., 2019).

The choice of whether a given BMP will form a homodimer or a heterodimer is made within the biosynthetic pathway. BMPs are made as inactive precursor proteins that dimerize and fold within the endoplasmic reticulum (ER) and are cleaved by members of the proprotein convertase (PC) family, such as furin, to generate the active, disulfide-bonded ligand and two prodomain monomers (Bragdon et al., 2011). Prodomains lack canonical signaling activity but are essential for ligand folding and dimerization, and can regulate subcellular trafficking, localization and bioavailability of mature ligands (Constam, 2014; Cui et al., 2001; Goldman et al., 2006; Gray and Mason, 1990; Harrison et al., 2011; Sengle et al., 2008; Tilak et al., 2014). Structural studies show that some dimerized TGF-² superfamily precursors adopt a conformation whereby the prodomain of one monomer contacts the ligand domain of the opposite monomer (Wang et al., 2016; Zhao et al., 2018), suggesting that prodomains may play a key role in heterodimer formation and function.

Bmp4 and Bmp7 preferentially form heterodimers rather than either homodimer when co- expressed in Xenopus, and the Bmp4 prodomain is necessary and sufficient to generate properly folded, functional BMP4 homodimers and Bmp4/7 heterodimers (Neugebauer et al., 2015). Bmp4 is sequentially cleaved at two sites within the prodomain to generate an active ligand (Cui et al., 2001). An initial cleavage occurs adjacent to the ligand domain, and this generates a non- covalently associated prodomain/ligand complex that is subsequently dissociated by cleavage at an upstream site (Figure 1A) (Degnin et al., 2004). Sequential cleavage at both sites, and formation of the transient prodomain/ligand complex is essential to generate a stable, fully active ligand (Goldman et al., 2006; Tilak et al., 2014). By contrast, BMP7 is cleaved at a single site and the prodomains remain transiently and noncovalently associated with the ligand following cleavage. The Type II BMP receptor then competes with the N-terminus of the prodomain for ligand binding and displaces the prodomain from the complex to allow for downstream signaling (Sengle et al., 2008). The same may be true for Bmp4/7 heterodimers, since both prodomains remain non-covalently attached following cleavage (Figure 1B) (Neugebauer et al., 2015).The factors that drive formation of fully functional heterodimers or homodimers are unknown.

Figure 1.

In humans, heterozygous mutations within the prodomain of BMP4 are associated with a spectrum of ocular, brain, kidney, dental and palate abnormalities (Bakrania et al., 2008; Chen et al., 2007; Nixon et al., 2019; Reis et al., 2011; Schild et al., 2013; Suzuki et al., 2009; Weber et al., 2008; Yu et al., 2019; Zhang et al., 2009). Two of the BMP4 prodomain missense mutations found in humans (c.271A >T, p.S91C and c.278A>G, p.E93G) disrupt a highly conserved phosphorylation motif (S-X-E/pS) (Figure 1C) that is recognized by the secretory pathway kinase Family with sequence similarity to 20C (FAM20C). FAM20C phosphorylates endogenous BMP4 at S91 within this motif (Tagliabracci et al., 2015) and BMP activity is decreased in ameloblasts and dental epithelium from Fam20c mutant mice (Liu et al., 2018; Liu et al., 2020). Mutations in human FAM20C cause the often-lethal disorder, Raine syndrome (Simpson et al., 2007). Some of the dental and bone defects observed in Raine patients can be accounted for by loss of phosphorylation of FAM20C substrates involved in biomineralization (Faundes et al., 2014) and of FGF23 (Tagliabracci et al., 2014). The etiology of other defects in these patients, such as cleft palate and craniofacial abnormalities, is unclear. Whether and how phosphorylation of the prodomain impacts BMP ligand formation or activity is unknown.

In the current studies, we analyzed the impact of p.S91C and p.E93G prodomain mutations on ligand function. We found that these mutations disrupt the formation of functional BMP4 homodimers, but not BMP4/7 heterodimers in ectopic expression assays. Mice carrying Bmp4^S91Cor Bmp4^E93G knock in mutations show early lethality, reduced BMP activity and accumulation of uncleaved homodimeric BMP4 precursor proteins with a concomitant loss of cleaved ligand. Our results suggest that phosphorylation of the BMP4 prodomain is required for proteolytic activation of BMP4 homodimers, but not heterodimers.

Results

Point mutations predicted to interfere with phosphorylation of the BMP4 prodomain selectively interfere with BMP4 homodimer but not BMP4/7 heterodimer activity

Humans heterozygous for missense mutations p.S91C or p.E93G within the BMP4 prodomain (Figure 1A,C) display enhanced predisposition to colorectal cancer, micropthalmia, skeletal defects, brain abnormalities, cleft lip and/or kidney dysgenesis (Bakrania et al., 2008; Lubbe et al., 2011; Suzuki et al., 2009; Weber et al., 2008), suggesting that these prodomain mutations interfere with the activity of BMP4 homodimers or BMP4/7 heterodimers. To test this possibility, we injected RNA encoding murine BMP4^HAMyc, BMP4^HAMycS91C or BMP4^HAMycE93G into Xenopus embryos near the dorsal marginal zone (DMZ) of four cell embryos. We have previously shown that these epitope tags do not interfere with the activity of BMP4 in vivo (Tilak et al., 2014). The DMZ was explanted from embryos at the early gastrula stage and immunoblots of DMZ extracts were probed for phosphoSmad1,5,8 (hereafter shortened to pSmad1), which provides a direct read out for BMP activity (assay illustrated in Figure 1D). Levels of pSmad1 are high in the VMZ, where endogenous Bmp ligands are expressed, but low in the DMZ, where BMP inhibitors are expressed. Ectopic BMP4^S91C or BMP4^E93G induced significantly less pSmad1 than did wild type BMP4 (Figure 1D, E), demonstrating that these point mutations interfere with BMP4 homodimer activity. We also tested whether a putative phosphomimetic mutant (BMP4^S91D) would interfere with ligand activity. pSmad1 activity was lower in DMZ explants expressing BMP4^S91D than in those expressing wild type BMP4 but was significantly higher than in those expressing BMP4^S91C (Fig. S1A, B). We used a second ectopic expression assay to verify results. RNA encoding wild type or mutant BMP4 was injected near the animal pole of 2-cell Xenopus embryos, ectoderm was explanted at the early gastrula stage and expression of the BMP target gene, tbxt, was analyzed by semi-quantitative (semi-q) RT-PCR. tbxt levels were significantly lower in explants from embryos expressing BMP4^S91C or BMP4^E93G relative to those expressing wild type BMP4 (Fig. 1G, H). These findings suggest that phosphorylation of S91 is required to generate fully active BMP4 homodimers.

We then repeated these assays in embryos co-injected with RNA encoding wild type or mutant BMP4 together with BMP7. We have previously shown that BMP4 and BMP7 preferentially generate heterodimers over either homodimer when co-expressed and that heterodimers generate significantly more BMP activity than either homodimer (Neugebauer et al., 2015). There was no significant difference in the levels of pSmad1 (Fig. 1D, F) or tbxt (Fig. 1G, I) induced in embryos co-injected with BMP7 together with wild type or either point mutant form of BMP4. This suggests that the S91C and E93G prodomain point mutations selectively interfere with BMP4 homodimer but not BMP4/7 heterodimer activity.

Bmp4^S91C homozygotes die during mid-embryogenesis and show reduced BMP activity in multiple tissues

To ask whether S91C and E93G prodomain missense mutations interfere with endogenous BMP4 activity in vivo, we introduced nucleotide mutations to encode single amino acid changes (p.S91C or p.E93G) along with sequence encoding an HA epitope tag into the prodomain of the Bmp4 allele in mice (Bmp4^S91C and Bmp4^E93G). We have previously generated control mice that carry an HA- epitope tag at the same position in the wild type Bmp4 locus, as well as a myc tag in the ligand domain (Bmp4^HAMyc). These mice are adult viable and show no visible defects (Tilak et al., 2014). Bmp4^S91/+ mice were intercrossed to determine whether homozygotes were adult viable.

Bmp4^S91C homozygotes were not recovered at weaning (Table 1A). We then established timed matings to determine when during embryogenesis death occurred. Bmp4^S91C/S91C mutants were recovered at the predicted Mendelian frequency at embryonic day (E)9.5 and E10.5 (Table 2A, Table S1C, D) but were slightly smaller than littermates at E10.5 (Fig. 2A-C) and were absent or resorbing by E11.5 (Fig. 2D-F, Table 2A, Table S1A, B).

Figure 2.

Table 1.

Table 2.

To determine whether and where BMP activity is reduced in Bmp4^S91Chomozygotes, we analyzed BMP activity in BRE:LacZ transgenic embryos at E10.5. This transgene contains a BMP-responsive element coupled to LacZ, which serves as an in vivo reporter of BMP signaling downstream of all endogenous BMP ligands (Monteiro et al., 2004). X-GAL staining of Bmp4^+/+;BRE:LacZ embryos revealed strong endogenous BMP activity in the brain and spinal cord, eye, branchial arches (BA), limb buds, heart, somitic mesoderm (SM) and ventroposterior mesoderm (VPM) (Fig. 2G, J). Bmp4^S91C/S91C;BRE:LacZ mutants were smaller and exhibited a severe (Fig. 2H, K, 1 out of 6 embryos) or modest (Fig. 2I,L; 5 out of 6 embryos) reduction in BMP activity in all tissues except the eye. Notably, LacZ staining was also reduced in the heart and VPM of Bmp4^S91C/S91C;BRE:LacZ embryos relative to wild type littermates at E9.5 (Fig. S2).

We also examined expression of the BMP target gene Nkx2-5 in the heart of wild type and Bmp4^S91C mutant littermates using whole mount in situ hybridization. At E10.5 expression of Nkx2-5 was reduced (1/3) or nearly absent (2/3) in hearts of Bmp4^S91C/S91Cembryos relative to littermates (Fig. 2M-O). Although mutant hearts were smaller than littermates, they showed grossly normal patterning of atria, ventricles and outflow tract (Fig. 2M-O).

Bmp4^E93G/E93G female mice are underrepresented at weaning and Bmp4^-/E93Gmutants die during late gestation with defects in ventral body wall closure, small eyes and heart defects

Bmp4^E93/+ mice were intercrossed to determine viability. Bmp4^E93G homozygotes were underrepresented at weaning, and this could be entirely accounted for by an underrepresentation of female but not male Bmp4^E93G/E93G mice (Table 1B). Bmp4^E93G/E93G male mice appeared grossly normal at weaning, although they weighed slightly less than wild type littermates (Fig. S3A). Unilateral or bilateral microphthalmia, anopthalmia and/or craniofacial defects were observed with low frequency in Bmp4^E93G heterozygotes at late gestation (Fig. S3B-D) (n=1/12) or at weaning (n=6/74). These defects have been previously reported in Bmp4 hypomorphic mice (Bonilla-Claudio et al., 2012; Furuta and Hogan, 1998; Goldman et al., 2006) and in humans heterozygous for the BMP4^E93 prodomain mutation (Bakrania et al., 2008).

To ask whether a single allele of Bmp4^E93G is sufficient to support viability, we intercrossed mice heterozygous for a null allele of Bmp4 (Bmp4^-/+) with Bmp4^E93G heterozygotes. Bmp4^-/E93G compound mutants were not recovered at weaning (Table 1C) but were recovered at the predicted Mendelian frequency through E14.5 (Table 2B, Table S2). At E13.5-14.5, all Bmp4^-/E93G mutants were slightly runted relative to littermates and had defects in ventral body wall closure in which the liver was partially or fully externalized (n=16/16) (Fig. 3A-G). Most Bmp4^-/E93G embryos also had small or absent eyes (n= 15/16) (Fig. 3C, G). A subset of mice heterozygous for the null allele of Bmp4 also had small or absent eyes (2/19) as previously reported (Dunn et al., 1997). Histological examination of hearts dissected from Bmp4^-/E93G mutants at E14.5 revealed highly trabeculated ventricular walls that failed to undergo compaction (3/3) and ventricular septal defects (2/3) (VSDs) (Fig. 3I) that were never observed in wild type embryos (Fig. 3H). A similar spectrum of abnormalities including small or absent eyes, failure to close the ventral body wall and VSDs have previously been reported in mice with reduced dosage of Bmp4 (Furuta and Hogan, 1998; Goldman et al., 2009; Kim et al., 2019; Uchimura et al., 2009).

Figure 3.

Bmp4^S91C/E93G female compound heterozygotes are underrepresented at weaning

Bmp4^S91/+ and Bmp4^E93G/+ mice were intercrossed to determine whether compound heterozygotes were adult viable. Bmp4^S91C/E93G mice were underrepresented at weaning, and this could be entirely accounted for by an underrepresentation of female but not male compound mutants (Table 1D). Bmp4^S91C/E93Gmice appeared grossly normal at weaning, although they weighed slightly, but not significantly, less than littermates (Fig. S4). The late lethality of female, but not male Bmp4^S91C/E93G compound heterozygotes and Bmp4^E93G homozygotes, along with their grossly normal phenotype contrasts with the early (E11) lethality and more severe phenotypic defects of Bmp4^S91Chomozygotes. These results raise the possibility that loss of FAM20C- mediated phosphorylation of the BMP4 prodomain fully accounts for the late lethality of Bmp4^E93G/E93G and Bmp4^S91C/E93G mutants whereas the p.S91C mutation may cause additional, FAM20C-independent defects in BMP4 function when encoded on both alleles of Bmp4.

E93G and S91C mutations lead to accumulation of BMP4 precursor protein and reduced levels of cleaved ligand and pSmad1 in vivo

To ask if BMP activity and/or protein levels are reduced in Bmp4^S91C or Bmp4^E93G mutants, we used immunoblot analysis to compare levels of pSMAD1 and BMP4 in E10.5 embryonic lysates isolated from wild type or mutant littermates. Levels of BMP4 precursor protein were significantly higher, and levels of pSMAD1 and of cleaved BMP4 ligand were significantly lower in E10.5 lysates from Bmp4^S91C homozygotes compared to wild type controls (Fig. 4A-D). No significant differences in levels of pSMAD1, BMP4 precursor or BMP4 ligand were observed in E10.5 lysates from Bmp4^E93G mutants compared to wild type controls (Fig. 4E-H), or in wild type embryos carrying HA and Myc epitope tags compared to untagged controls (Fig. S5A-D). When we repeated this analysis in E13.5 mouse embryonic fibroblasts (MEFs) isolated from wild type or mutant littermates, we found that pSMAD1 levels were significantly reduced in Bmp4^S91Cheterozygotes (Fig. 4I, J) and in Bmp4^E93G homozygotes (Fig. 4M, N) compared to wild type controls. In addition, levels of BMP4 precursor protein were increased and levels of cleaved ligand were reduced in Bmp4^S91Cheterozygotes (Fig. 4I, K, L) and in Bmp4^E93G heterozygotes or homozygotes (Fig. 4M, O, P) compared to wild type controls. Thus, p.E93G and p.S91C mutations in the prodomain of BMP4 interfere with proteolytic maturation of the BMP4 precursor, leading to accumulation of unprocessed precursor protein and reduced levels of cleaved ligand and BMP activity in vivo.

Fig. 4.

BMP4^E93G and BMP4^S91C precursor proteins are O-glycosylated and exit the ER

FAM20C-mediated phosphorylation of FGF23, like that of BMP4, is required for furin mediated cleavage (Wang et al., 2012). Mechanistically, phosphorylation of FGF23 prevents O- glycosylation of a nearby residue, and O-glycosylation sterically blocks the furin cleavage site (Tagliabracci et al., 2014). To test if a similar mechanism accounts for lack of cleavage of BMP4^E93G or BMP4^S91C, MEFS from E13.5 wild type, Bmp4^E93G/E93G or Bmp4^S91C/+ mice were incubated in the presence or absence of O-Glycosidase and Neuraminidase and BMP4 precursor protein was analyzed on immunoblots. Treatment with these deglycosylases led to a more rapid migration of the BMP4 precursor protein, indicating that BMP4 is O-glycosylated, but there was no difference in the migration of wild type or point mutation precursors in the presence or absence of deglycosylases (Fig. 5A), suggesting that phosphorylation does not alter O-glycosylation.

Fig. 5.

Another mechanism by which FAM20C-mediated phosphorylation might regulate cleavage of the BMP4 is if it is required for the BMP4 precursor protein to adopt its native folded conformation so that it is able to exit the ER. We analyzed N-linked glycosylation of endogenous wild type and mutant precursor proteins present in cultured MEFs. N-linked carbohydrate residues that are transferred onto proteins in the ER are sensitive to digestion with Endoglycosidase H (Endo H), but when further modified in the Golgi they become Endo H resistant but remain sensitive to digestion with Peptide N-Glycosidase F (PNGase). Thus, Endo H resistance/PNGase sensitivity is a hallmark of proteins that are properly folded and able to traffic from the ER into the Golgi. As shown in Figure 5B, Endo H sensitive (asterisks) and Endo H resistant/PNGase F sensitive (arrowheads) forms of wild type and mutant precursors were detected in MEFs. We have previously shown that high mannose, Endo H sensitive carbohydrates are retained at one or more glycosylation site(s) on BMP4 ligand homodimers even after BMP4 has folded, trafficked through the trans-Golgi network (TGN) and been cleaved (Degnin et al., 2004). This explains why partial EndoH sensitivity is observed in both wild type and mutant precursors that have exited the ER. Collectively, our findings are most consistent with a model in which FAM20C-mediated phosphorylation of the BMP4 prodomain is not required for folding or exit of the precursor protein from the ER, but is required for PC recognition and/or for trafficking to post-TGN compartment(s) where BMP4 is cleaved (Tilak et al., 2014).

Discussion

Humans heterozygous for p.S91C or p.E93G point mutations within the prodomain of BMP4 show congenital birth defects and/or enhanced predisposition to colorectal cancer (Bakrania et al., 2008; Lubbe et al., 2011; Suzuki et al., 2009; Weber et al., 2008). These mutations disrupt a conserved -S-X-E- motif required for FAM20C-mediated phosphorylation of Ser91 suggesting that phosphorylation of the prodomain, which itself lacks biologic activity, is required for ligand activity. Here we show that these mutations lead to reduced activity of BMP4 homodimers but not BMP4/7 heterodimers in Xenopus assays. Furthermore, in mice carrying either of these mutations proteolytic maturation of endogenous BMP4 is impaired, leading to reduced BMP activity. Our results suggest that phosphorylation of the BMP4 prodomain by FAM20C is essential for furin to cleave the precursor protein to generate active BMP4 homodimers.

How might phosphorylation facilitate proteolytic maturation of BMP4 homodimers? We have previously shown that cleavage of BMP4 occurs in membrane proximal vesicles, and this is proposed to protect the ligand from degradation by passing it off to extracellular matrix binding partners as cleavage occurs (Tilak et al., 2014). This raises the question of how BMP4 can traffic through the TGN, where furin is first active (Anderson et al., 2002; Thomas, 2002), yet escape cleavage until it reaches the cell surface. We propose two possible mechanisms by which FAM20C-mediated phosphorylation facilitates cleavage of BMP4 and also ensures that this occurs in a membrane proximal subcellular compartment. One model proposes that phosphorylation occurs in the TGN and directs trafficking of BMP4 out of the golgi to the cell surface via a route that is distinct from the pathway taken by furin (Thomas, 2002) (Fig. 6A). In this model, furin would first encounter and cleave BMP4 in membrane-proximal intracellular compartments, similar to what has been observed for other substrates (Thomas, 2002). This model predicts that BMP4^E93G and BMP4^S91C precursor proteins are trapped in the TGN or other subcellular compartments that furin cannot access. An alternate model proposes that furin, BMP4 and FAM20C traffic together out of the TGN but that the kinase is not active until it exits the TGN. In this model, phosphorylation in a membrane proximal compartment is required for furin to cleave BMP4 (Fig. 6B). In support of this model, FAM20C is synthesized as a basally active kinase that is tethered to the golgi by its propeptide. Propeptide cleavage releases Fam20C from the Golgi, and greatly enhances its kinase activity as it to traffics to the cell surface (Chen et al., 2021). Furthermore, AlphaFold structure predictions position the phosphorylated serine of BMP4 (Ser91) in close proximity to the -R²⁸⁷-R-R-A-K-R²⁹² sequence motif recognized by furin (Fig. 6C-E) (Jumper et al., 2021; Varadi and Velankar, 2023), consistent with the possibility that phosphorylation is required for furin to access the cleavage site. This model predicts that BMP4^E93G and BMP4^S91C precursor proteins can traffic to membrane proximal locations but remain uncleaved.

Fig. 6.

Our results suggest that phosphorylation of the BMP4 prodomain is required for cleavage of BMP4 homodimers, but not BMP4/7 heterodimers. Interestingly, alphafold predicts that the furin cleavage motif on each monomeric chain of BMP4 homodimers (Figure 6F, purple) is situated on opposite sides of the dimeric molecule, each in close proximity to S91 (Fig. 6C, D). By contrast, the two furin cleavage motifs (Fig. 6G, purple) on heterodimers composed BMP4 (light shading) and BMP7 (dark shading) are predicted to be located on the same face of the dimeric molecule, in close proximity to each other (Fig. S6C) (Jumper et al., 2021; Varadi and Velankar, 2023). This raises the possibility that furin can be recruited to the cleavage motif on BMP7, enabling it to access the cleavage motif on BMP4 independent of phosphorylation when present as a BMP4/7 heterodimer.

The late-stage lethality of Bmp4^E93G/E93G and Bmp4^S91C/E93Gmice is consistent with a selective loss of BMP4 homodimers rather than BMP4/7 heterodimers since mice homozygous for a mutation (Bmp7^R-GFlag) that disrupts the formation of functional BMP2/7 and BMP4/7 heterodimers, or mice compound heterozygous for the Bmp7^R-GFlag and Bmp4 null alleles die by E11.5 (Kim et al., 2019). Furthermore, whereas Bmp7^{R-GFlag/R-GFlag}and Bmp7^R-GFlag/+;Bmp4^-/+ mutants show stereotypical heart defects including thinner myocardial walls, a common atrium, and a small, malformed OFT relative to wild type littermates, Bmp4^S91C/S91Cmutant hearts show grossly normal patterning of atria, ventricles and outflow tract while Bmp4^-/E93G mutant hearts show reduced myocardial trabeculation and VSDs. The latter heart defects phenocopy those observed in Ngly1 mutant mice, which also show VSDs, poorly elaborated myocardial trabeculation and late stage (P0) lethality (Fujihira et al., 2017; Galeone et al., 2020). NGLY1 is a deglycosylase that is required to clear misfolded BMP4 precursor monomers from the ER. This, in turn, promotes the formation of properly folded BMP4 homodimers that can be transported out of the ER to be cleaved to generate the functional ligand (Galeone et al., 2020). The Drosophila ortholog of NGLY1 (Pngl1) is required for formation of functional Dpp (the fly ortholog of BMP2/4) homodimers, but not heterodimers (Galeone et al., 2017). In pngl1 mutant flies, and in MEFs from Ngly1 mutant mice, Dpp/BMP4 precursor accumulates in the ER and very little cleaved ligand is observed (Galeone et al., 2020). By contrast, in Bmp4^E93G/E93G mutant MEFs BMP4 precursor accumulates in post-ER compartments but very little cleaved ligand is observed. The almost complete loss of mature BMP4 secreted from Bmp4^E93G/E93Gmutant MEFs is consistent with a selective loss of homodimers since Bmp7 is not expressed in MEFs (Lienert et al., 2011) and thus BMP4 is predicted to exist primarily as a homodimer in these cells. Thus, mutations that prevent phosphorylation of the BMP4 prodomain, or mutations in Ngly1 both lead to reduced levels of mature BMP4 homodimers and similar phenotypic consequences, although the mechanisms underlying the loss of proteolytic maturation are distinct.

Our findings demonstrate that Bmp4^S91C and Bmp4^E93Gare hypomorphic alleles, although Bmp4^S91C causes a more severe loss of function. Although both mutations are predicted to prevent FAM20C mediated phosphorylation of the BMP4 prodomain, it is likely that the p.S91C mutation leads to additional defects in protein function due to deleterious effects of introducing an ectopic cysteine residue. BMP4 folds into a cystine knot containing three stereotypical intermolecular disulfide bonds that stabilize ligand monomers and a single intramolecular disulfide bond that stabilizes the dimer (Schwarz, 2017). The addition of an ectopic cysteine residue may lead to aberrant intramolecular or intermolecular disulfide bonds in precursor proteins, as suggested by biochemical analysis of BMP4^S91C homodimers expressed in cultured mammalian cells (Tabatabaeifar et al., 2009). This interpretation is consistent with our finding that Bmp4^S91C/E93Gcompound heterozygotes have a less severe phenotype than Bmp4^S91C/S91Cmice, and with our observation that ectopically expressed BMP4^S91D generates significantly higher activity in Xenopus embryos than does BMP4^S91C.

In summary, our data suggest that phosphorylation of the BMP4 prodomain is required for proteolytic maturation of BMP4 homodimers but not heterodimers, although further studies will be required to validate this in vivo and to understand the process mechanistically. Previous studies have demonstrated the importance of BMP heterodimers as endogenous ligands in multiple organisms (Bauer et al., 2023; Kim et al., 2019; Little and Mullins, 2009; Shimmi et al., 2005). The current findings show that BMP4 homodimers are also functionally relevant. The existence and relative role of heterodimers versus homodimers is likely to vary widely among different organisms, tissues and developmental stages. Additional studies are required to understand where and how a given BMP is directed to form one species of the other in vivo.

Materials and Methods

Xenopus embryo culture and manipulation

Animal procedures followed protocols approved by the University of Utah Institutional Animal Care and Use Committee. Embryos were obtained, microinjected, and cultured as described (Mimoto and Christian, 2011). Embryo explants were performed as described (Mimoto et al., 2015; Neugebauer et al., 2015).

Mouse strains

Animal procedures followed protocols approved by the University of Utah Institutional Animal Care and Use Committees. Bmp4^LacZ/+ (RRID:MGI:3811252) and BRE-LacZ mice were obtained from Dr. B. Hogan (Duke University) and Dr. C. Mummery (Leiden University), respectively. Bmp4^S91C and Bmp4^E93G mice were generated by personnel in the Mutation Generation & Detection and the Transgenic & Gene Targeting Mouse Core Facilities at the University of Utah using CRISPR-Cas9 technology. sgRNA RNAs (5’-TGAGCTCCTGCGGGACTTCG-3’) were injected into C57BL/6J zygotes together with long single stranded donor DNA repair templates (E93G: 5’- GGAAGAAAAAAGTCGCCGAGATTCAGGGCCACGCGGGAGGACGCCGCTCAGGGCAGAGCCATG AGCTCCTGCGGGACTTCGAaGCGACACTTTATCCATACGACGTGCCAGACTATGCACTACAGATGT TTGGGCTGCGCCGCCGTCCGCAGCCTAGCAAGAGCGCCGTCATTCCGGATTACATGAGGGATCT TTACCGGCTCCAGTCTGGaGgGGAGGAGGAGGAAGAGCAGAGCCAGGGAACCGGGCTTGAGTAC CCGGAGCGTCCCGCCAGCCGAGCCAACACTG-3’; S91C: 5’- GGAAGAAAAAAGTCGCCGAGATTCAGGGCCACGCGGGAGGACGCCGCTCAGGGCAGAGCCATG AGCTCCTGCGGGACTTCGAaGCGACACTTTATCCATACGACGTGCCAGACTATGCACTACAGATGT TTGGGCTGCGCCGCCGTCCGCAGCCTAGCAAGAGCGCCGTCATTCCGGATTACATGAGGGATCT TTACCGGCTCCAGTgTGGcGAGGAGGAGGAGGAAGAGCAGAGCCAGGGAACCGGGCTTGAGTAC CCGGAGCGTCCCGCCAGCCGAGCCAACACTG-3’; sequence encoding HA epitope underlined; nucleotide changes bold and small case) and Cas9 protein. G0 founders were crossed to C57BL/6J females to obtain heterozygotes. DNA fragments PCR-amplified from genomic DNA were sequenced to verify the presence of the epitope tag and absence of other sequence changes. Genotypes were determined by PCR amplification of tail DNA using primers that anneal to sequence immediately surrounding the HA epitope tag (5’ primer: 5’-TATGCCAAGTCCTGCTAG-3’ and 3’ primer: 5’- GATCCCTCATGTAATCCG-3’) under the following conditions: 94°C for 30 seconds, 60°C for 30 seconds, 72°C for 30 seconds, 35 cycles.

cDNA constructs

cDNAs encoding mouse BMP4^HAMyc and BMP7^Flag have been described previously (Kim et al., 2019; Tilak et al., 2014). cDNAs encoding BMP4^HAMycS91C, BMP4^HAMycS91D and BMP4^HAMycE93G were generated using a QuickChange II XL site-directed mutagenesis kit (Agilent Technologies).

Analysis of RNA

Total RNA was isolated using TRIzol (Invitrogen). Semi quantitative RT-PCR was performed as described (Nakayama et al., 1998) using an annealing temperature of 58°C.

In situ hybridization and ß-galactosidase staining

Embryos were processed for in situ hybridization with digoxigenin-labeled Nkx2.5 riboprobes as described previously (Wilkinson and Nieto, 1993). ß-galactosidase staining of BRE-LacZ embryos was performed as described (Lawson et al., 1999) and color was developed using RedGal. Investigators were blinded to genotype until after morphology and/or staining intensity had been documented.

Immunoblot analysis of Xenopus extracts

Proteins were harvested from 10 pooled DMZ explants by freon extraction as described previously (Mimoto and Christian, 2011). Proteins were resolved by SDS-PAGE under reducing and non-reducing conditions and transferred onto PVDF membranes. Membranes were probed with anti-actin (1:10000, Sigma, RRID:AB_476693) and anti-pSmad1/5 (1:1000, Cell signaling RRID:AB_491015) antibodies. Immunoreactive proteins were detected using Enhanced Chemiluminescence reagent (Pierce) and light emissions captured with x-ray film. Images were scanned and relative band intensity was quantified using ImageJ software.

Immunoblot analysis of mouse embryo lysates and MEFs

Mouse embryos were dissected from pregnant females at E10.5, homogenized in lysis buffer (150 mM NaCl, 20 mM Tris-Cl pH 7.5, 1mM EDTA, 1% Sodium deoxycholate, 1% NP40, 1X protease inhibitor (cOmplete Mini, EDTA-free, Roche), 1X phosphatase inhibitor cocktail (Sigma) and protein concentration was measured by BCA kit (Thermo Scientific). 80 ug, 10 ug and 5 ug of embryo lysates were used for detection of BMP4 mature ligand, BMP4 precursor and pSmad1, respectively. MEFs were isolated at E13.5 as described (Durkin et al., 2013) and cultured in 10% FBS, 1X pen/strep, 1x Glutamate in DMEM. Following the second passage, MEFs were cultured in serum containing media until they reached 80% confluency and were then cultured in serum free media for 24 hours before collecting cells and conditioned media. 800 ul of conditioned media was used for TCA precipitation. Proteins were deglycosylated according to manufacturer’s instructions using O-Glycosidase, Neuramidase, EndoH and PNGase purchased from NEB. Proteins were separated by electrophoresis on 10% or 12% gels and transferred to PVDF membranes that were probed with anti-Bmp4 (1:1000, Santa Cruz RRID: AB_2063534), anti- pSmad1/5 (1:1000, Cell Signaling, RRID:AB_491015) or anti-actin (1:10,000, Sigma, RRID:AB_476693) antibodies followed by HRP-conjugated anti-rabbit IgG or anti-mouse IgG2b (Jackson ImmunoResearch) secondary antibodies. Immunoreactive proteins were detected using Enhanced Chemiluminescence reagent (Pierce) and light emissions captured with x-ray film. Images were scanned and relative band intensity was quantified using ImageJ software.

Structure Prediction

Full length (starting after processed signal sequence) Bmp4 homodimer and Bmp4/7 heterodimer sequences were input into https://alphafoldserver.com/ running AlphaFold version 3. Protein structures were analyzed and figures made using PyMol (Schrodinger, 2020) ( Retrieved from http://www.pymol.org/pymol). Quality of prediction as measured by the pLDDT can be observed per residue in Fig. S6 where very low (<50), low (50-70), confident (70-90) and very high (>90) rankings are shown indicating that the furin binding loops are of low predictive confidence.

Statistics

NIH Image J software was used to quantify band intensities. A student’s t-test was used to compare differences in gene expression or protein levels between two groups. Differences with P < 0.05 were considered statistically significant. All results were reproduced in three biological replicates.

Data and resource availability

All relevant data and resource can be found within the article and its supplementary information.

Acknowledgements

We thank Dr. Eyad Marashly for generating the BMP4^E93G cDNA and synthetic RNAs and Dr. Diana Lim for education in the use of Adobe illustrator. We would like to acknowledge the University of Utah Mutation Generation and Detection Core and the Transgenic and Gene Targeting Core for generating the Bmp4^S91C and Bmp4^E93G mouse lines. This work utilized DNA and peptide shared resources supported by the Huntsman Cancer Foundation and the National Cancer Institute of the NIH (grant P30CA042014). The content is solely the responsibility of the authors and does not represent the official views of the NIH.

Additional information

Competing interests

There are no actual or perceived conflicts on the part of any author.

Funding

This work was supported by the Eunice Kennedy Shriver National Institute of Child Health and Human Development, Grant/Award Numbers: R21HD102668; R21HD102668-W1 to JLC; National Institute of Diabetes and Digestive and Kidney Diseases, Grant/Award Number: R01DK128068 to JLC.

Additional files

Supp Figs and tables