ZC4H2 stabilizes Smads to enhance BMP signalling, which is involved in neural development in Xenopus

Bone morphogenetic proteins (BMPs) play vital roles in regulating stem cell maintenance, differentiation and embryonic development. Intracellularly, BMP signalling is mediated by Smad proteins, which are regulated post-transcriptionally through reversible phosphorylation and ubiquitination. ZC4H2 is a small nuclear protein associated with intellectual disability and neural development in humans. Here, we report that ZC4H2 is highly expressed in the developing neural system and is involved in neural patterning and BMP signalling in Xenopus. Knockdown of ZC4H2 led to expansion of the expression of the pan neural plate marker Sox2 in Xenopus embryos. In mammalian cells, ZC4H2 promotes BMP signalling and is involved in BMP regulated myogenic and osteogenic differentiation of mouse myoblast cells. Mechanistically, ZC4H2 binds and stabilizes Smad1 and Smad5 proteins through reducing their association with the Smurf ubiquitin ligases and thus their ubiquitination. We also found that a group of ZC4H2 mutations, which have been isolated in patients with intellectual disorders, showed weaker Smad-stabilizing activity, suggesting that the ZC4H2–Smad interaction might contribute to proper neural development in humans.


Introduction
Bone morphogenetic proteins (BMPs) belong to the transforming growth factor b (TGFb) superfamily. They are involved in a broad range of cellular responses in metazoa, including cell proliferation, differentiation and embryonic development [1][2][3][4]. In vertebrate embryos, the BMP pathway provides essential signals that induce and pattern the primary germ layers, regulate tissue morphogenesis and left-right asymmetry, and affect cellular pluripotency, differentiation, growth and death [4,5]. In Xenopus embryos in particular, mesoderm and endoderm are induced by BMP ligands, acting together with FGFs, and early tissue patterning is achieved by BMPs alongside Wnt and FGF signals. In the ectoderm, different levels of BMP signalling trigger differentiation of the epidermis, neural crest, sensory placodes and neural tissues [5]. To maintain homeostasis in adults, the BMP signal also participates in tissue remodelling and regeneration. BMPs can determine the fate of mesenchymal stem cells by stimulating their differentiation into the chondro-osteoblastic lineage and meanwhile blocking their differentiation into the myoblastic lineage. In response to BMP signals, critical osteogenic transcription factors, such as Runx2 and Ostrix, are induced and drive efficient osteogenic differentiation. On the other hand, BMPs can inhibit myogenic differentiation by suppressing the expression of myogenic basic helix-loop-helix (bHLH) transcriptional factors, such as MyoD and Myf5, and/ or inducing the expression of Id (inhibitor of differentiation or inhibitor of DNA binding) proteins that block the DNA-binding ability of bHLH transcriptional factors [2,[6][7][8][9][10].
BMP signalling is canonically transferred by the Smad (signalling mothers against decapentaplegic) cascade. BMP ligands can bind to type I and type II receptors on the cell surface. The type II receptors phosphorylate and activate the type I receptors, which in turn phosphorylate downstream receptor-regulated Smads (R-Smads), i.e. Smad1 and Smad5 (or Smad1/5). The activated phospho-R-Smads form complexes with Smad4 and translocate into the nucleus. The Smad complex acts as a transcriptional activator or repressor to regulate target genes expression [2,3,[11][12][13]. In addition to phosphorylation by BMP receptors for their activation, the cytoplasm-nuclear shutting of Smads is also tightly controlled by the phosphorylation and de-phosphorylation cycling [14][15][16][17][18]. Alongside phosphorylation, Smad1/5 are also regulated by the ubiquitination and de-ubiquitination system. The HECT (homologous to the E6-AP carboxyl terminus) type ubiquitin ligases Smurf1 and Smurf2 were first reported to target Smad1/5 for polyubiquitination and proteasomal degradation through direct interaction with their PY motifs in the linker region [1,8,11,[19][20][21][22]. Several other ubiquitin ligases that target Smads have also been isolated, including SCF/Roc1, NEDD42, WWP1 and HsN3 [20,[22][23][24]. Moreover, non-proteolytic ubiquitination through atypical ubiquitin linkages and monoubiquitination have also been noted in Smad activity regulation [25]. A number of deubiquitination enzymes were also reported to regulate the stability or activity of Smads, such as USP15 and USP9X/FAM [25,26].
ZC4H2 is a zinc-finger protein belonging to the family of proteins with a C-terminal zinc-finger domain characterized by four cysteine residues and two histidine residues. Additionally, ZC4H2 has a coiled-coil domain. Association between ZC4H2 mutation and human intellectual disability syndrome has been reported by two independent studies recently. They both provided evidence that ZC4H2 is specifically expressed in the developing neural system and required for neural development in zebrafish [27,28]. Although the ZC4H2 mutations found in patients are predicted to destabilize the protein and show impaired restoration activity in ZC4H2 mutant or morphant zebrafish embryos, little is known about the underlying molecular mechanism for ZC4H2 in neural development and the related human diseases. In this study, we report that ZC4H2 is a new modulator of BMP signalling, through which ZC4H2 is involved in BMP-mediated neural patterning in Xenopus and stem cell differentiation in mammalian cells. Mechanistically, ZC4H2 stabilizes Smads by attenuating their ubiquitination via inhibiting the association between Smurfs and Smads. Furthermore, we also provide evidence to suggest that the impaired Smads-stabilizing activity of ZC4H2 mutants found in patients with intellectual disability might contribute to the related disease progression.

ZC4H2 is involved in Xenopus neural development
Semi-quantitative RT-PCR was used to investigate the temporal expression patterns of ZC4H2 during Xenopus embryonic development. ZC4H2 mRNA was expressed from the early cleavage stages to the swimming tadpole stages that we examined (figure 1a). Whole mount in situ hybridization was carried out to determine its spatial expression pattern. ZC4H2 transcripts were detected at the animal hemispheres of the cleavage embryos (figure 1b,c) and at the dorsal marginal zone of the gastrula embryos (figure 1d). As the embryos develop, ZC4H2 transcripts were mainly localized in the developing neural tissues, i.e. neural plate at neural stages and neural tube at tadpode stages (figure 1e-g). To determine the potential function of endogenous ZC4H2 during Xenopus development, knockdown experiments were carried out with specific morpholinos (MO) against ZC4H2, which efficiently blocked the expression of a GFP reporter carrying its targeted sequences when co-injected into Xenopus embryos (data not shown). Interestingly, knockdown of ZC4H2 led to reduction of head and eye structures with shortened trunks and less pigments, an effect that was largely rescued by co-injection of ZC4H2 mRNA (figure 1h,i). At the neural stage, the expression of Sox2, a pan neural plate marker, expanded while Slug, a neural crest marker, showed reduced and marginalized expression pattern, and these effects were also well rescued by co-injection of ZC4H2 mRNA (figure 1j,k). In Xenopus animal caps, Sox2 and Pax3, two neural plate markers, were upregulated while Keratin, an epidermis marker and Slug, a neural crest marker, were downregulated when ZC4H2 was knocked-down, which could all be largely rescued by co-injection with ZC4H2 mRNA (figure 1l ). These results imply that ZC4H2 is involved in neural development in Xenopus.

ZC4H2 regulates BMP signalling in Xenopus embryos and C2C12 cells
As Xenopus ectoderm is patterned by BMP signalling gradient, we tested whether ZC4H2 is involved in BMP signalling. In BMP luciferase reporter assays in Xenopus animal caps, knockdown of ZC4H2 downregulated while its overexpession upregulated the expression level of the BMP reporter Id1-Luc [29] (figure 2a,b). Next, we examined the effects of ZC4H2 overexpression or knockdown on BMP signalling in Hep3B and C2C12 cells, two BMP responsive mammalian cell lines. Firstly, we examined ZC4H2 level by real-time RT-PCR in the ZC4H2 overexpressing or knocking-down Hep3B or C2C12 cells and found that all the siRNAs used for targeting human ZC4H2 in Hep3B and mouse ZC4H2 in C2C12 cells showed significant knockingdown efficiency for endogenous ZC4H2 and that both cells received a large amount of exogenous ZC4H2 transcripts when the ZC4H2 plasmids were transfected into them. Also, BMP treatment shows no effects on either the endogenous or exogenous ZC4H2 expression in both the cell lines (figure 2c,e,g,i). In both Hep3B and C2C12 cells, the transactivation of Id1-Luc reporter by BMP2 was enhanced by overexpression of ZC4H2, and reduced by three independent siRNA-mediated knockdowns of endogenous ZC4H2 (figure 2c-j). The mRNA levels of Id1 and Id2, two established BMP target genes, were decreased when ZC4H2 was knocked-down in C2C12 cells (figure 2k,l). The mouse myoblast cell line C2C12 has been a model system to analyse BMP signalling, in which BMP inhibits its rsob.royalsocietypublishing.org Open Biol. 7: 170122 myogenic differentiation while promoting its osteogenic differentiation [6,8]. When cultured in myogenesis induction medium, C2C12 cells typically exhibited induced expression of myogenic indicators, including Myf5 and MyHC ( figure 3). Under such a condition, their expression levels were markedly decreased when ZC4H2 was over-expressed while increased when the cells were transfected with ZC4H2 siRNAs (figure 3). When treated with recombinant human BMP2 protein, C2C12 cells quickly undergo osteogenic differentiation and express osteoblast markers such as alkaline phosphatase (ALP), Osteocalcin, Osterix and Type I Collagen (figure 4). Overexpression of ZC4H2 enhanced the expression of the rhBMP2-mediated induction of osteoblast markers (figure 4b-e) and increased the ALP activity induced by rhBMP2 treatment (figure 4f,g). In contrast, knockdown of ZC4H2 expression by siRNAs reduced BMP-induced expression of the osteoblast markers and ALP activity ( figure 5). These observations support a positive role of ZC4H2 in BMP signalling during BMP-regulated myoblast differentiation. Note that the endogenous ZC4H2 level remained the same during myogenic differentiation or BMP-induced osteogenic differentiation of C2C12 cells (figures 3a,d, 4a and 5a).

ZC4H2 stabilizes Smad1 and Smad5
Because ZC4H2 is predicted as a nuclear protein with a highly conserved nuclear location signal at its C terminus and Smad, the key component of the BMP signalling pathway, can shuttle between cytoplasm and nucleus, we hypothesized that ZC4H2 regulates BMP signalling by directly interacting with Smads. To test this hypothesis, co-immunoprecipitation stage: rsob.royalsocietypublishing.org Open Biol. 7: 170122 (co-IP) assays were carried out. In HEK293 cells transiently cotransfected with FLAG-tagged Smad1 or Smad5 and myctagged ZC4H2, Smad1 and Smad5 co-immunoprecipitated with ZC4H2 (figure 6a,b). In the reverse experiment, ZC4H2 also co-immunoprecipitated with both Smad1 and Smad5 (figure 6c,d). To map the domains of Smad proteins that are responsible for the observed interaction, a series of Smad1 and Smad5 truncated constructs were tested via IP assays in HEK293 cells. Smad proteins are approximately 465 amino acids in length and consist of two globular domains (MH1 and MH2 domains) linked by a linker region in between [12]. Our results showed that it is the MH2 domain of Smad1 or Smad5 that is responsible for their interaction with ZC4H2, as all the truncates containing the MH2 domain could pull-down ZC4H2 while the MH2 domain deleted truncates could not (figure 6e-g).
Interestingly, we noted that co-transfection of ZC4H2 increased the protein levels of both Smad1 and Smad5 in HEK293 cells, but not the level of the DLinker-Smad1 (figure 7a,b). We inhibited protein synthesis by cycloheximide and assessed the stability of Smad 1 or Smad5 with or without ZC4H2 co-expression in HEK293 cells. The data showed that both the Smad1 and Smad5 proteins were clearly stabilized in the presence of ZC4H2 (figure 7c-f). Consistently, ZC4H2 increased the expression of a Gal4 reporter when co-expressed with Gal4-Smad1or Smad5 in a heterologous luciferase reporter assay (figure 7g,h).
As ubiquitination plays a key role in Smad stability regulation, we examined whether ZC4H2 may affect the ubiquitination status. The results showed that co-expression of ZC4H2 strongly reduced the ubiquitination of both Smad1 and Smad5 proteins, but not that of the DLinker-Smad1, which is relatively weakly ubiquitinated (figure 7i,j). The above data imply that ZC4H2 stabilizes Smad1 and Smad5 by attenuating their polyubiquitination.
The most common ubiquitin E3 ligases responsible for Smads ubiquitination are Smurf1 and Smurf2, two typical HECT type E3 ligases. The PY motif in the Linker domain of Smad1 or Smad5 is responsible for the recruitment of Smurfs for the following ubiquitination and degradation [1,8,19 -22,30]. A series of in vitro ubiquitination assays were carried out to examine the effect of ZC4H2 on Smurfmediated Smad ubiquitination. Indeed, Smurf1 or Smurf2 dramatically increased the levels of polyubiquitinated Smad1 and Smad5, and this activity was largely inhibited when ZC4H2 was co-expressed (figure 8a-d). The ubiquitination level of both Smad1 and Smad5 were not changed when the ligase-dead mutant of Smurf1 (Sumrf1 CA) or Smurf2 (Smurf2 CG) was used, whether ZC4H2 was co-expressed or not (figure 8a-d). In co-immunoprecipitation assays, ZC4H2 co-expression reduced the levels of Smurf proteins associated with Smads, while the DLinker Smad1 failed to pull-down Smurf1 (figure 8e-h), as reported before [1,21]. These data suggest that ZC4H2 reduced the interaction between Smads and Smurfs, and thus their ubiquitination.

ZC4H2 is required for Smad stabilization in vivo
Next we examined whether ZC4H2 is involved in the stabilization of endogenous Smads in mammalian cells. In C2C12  n.s.
si-hZC4H2 (c-f) and C2C12 cells (g-j). Note that BMP treatment has no effect on either exogenous and endogenous ZC4H2 mRNA levels in both the cell lines (c,e,g,i and HeCaT cells, the protein levels of both total Smad1 and phosphorylated Smad1/5 were increased when ZC4H2 was over-expressed and were reduced when ZC4H2 was knocked-down by three independent siRNAs, either in the presence or the absence of rhBMP2 treatment (figure 9). These data imply that ZC4H2 is required for Smads stabilization in vivo.
In humans, a group ofZC4H2 substitutional mutations have been reported to be associated with intellectual disorders, including V63L, A200V, P201S, R198Q and R213 W, which all failed to rescue the ZC4H2 morphants in zebrafish [27,28]. We tested the Smad-stabilizing effects of mutants. Interestingly, we found that all these mutated ZC4H2 proteins showed weaker Smad-stabilizing activity compared to wild-type ZC4H2 proteins (figure 10a,b). In the BMP reporter assays, all the ZC4H2 mutants except R198Q mutant proteins showed weaker activity to enhance Smad transactivity (figure 10c,d). In addition, we also found that all these mutated ZC4H2 proteins showed weaker interaction with Smad1 and Smad5 in co-IP assays (figure 10e,f ) and their abilities to prevent Smad1/5 ubiquitination were impaired (figure 10g,h). In Xenopus embryos, these mutated ZC4H2 constructs also failed to rescue the expansion of Sox2 expression in the ZC4H2 morphants (figure 10i,j). We suggest that the weakened Smads-stabilizing activity of the mutated ZC4H2 forms might contribute to the impaired neural development in those patients.

Discussion
BMP/Smads signalling plays a critical role in ectodermal patterning and counteracts an intrinsic neural induction programme in the nascent ectoderm [4,5]. In this study, we found that ZC4H2 was expressed in the animal hemispheres and ectoderm tissues before/in gastrulation and in the neural specific ectoderm when the embryos established their body axes after gastrulation in developing Xenopus embryos. Morpholino-mediated knockdown of ZC4H2 mis-patterned the early ectoderm in Xenopus embryos, which was also reproduced in animal caps. The expansion of the neural plate also fits with a potential reduced BMP signalling in the ZC4H2 morphants. We showed that ZC4H2 was involved in BMP signalling in Xenopus animal caps, as indicated by changes of expression levels of the Id1-Luc reporter in ZC4H2 knockdown or overexpression experiments. Our work suggests that ZC4H2 might be involved in the BMP signalling mediated neural induction in Xenopus embryos. As we provided strong evidence that ZC4H2 stabilizes Smad1/5 and promotes BMP signalling cell autonomously in BMP receiving cells, it is possible that ZC4H2 cell autonomously enhances BMP signalling to facilitate ectoderm differentiation while inhibiting neural induction in Xenopus early developing embryos with a broad ectoderm-specific expression pattern.
ZC4H2 is highly conserved across vertebrates. In the ZC4H2 mutant zebrafish larvae, a specific loss of V2 interneurons in the ventral spinal cord and brain was observed [27,28], which implies an important role for ZC4H2 in dorsal -ventral patterning of the spinal cord. Therefore, we hypothesized that ZC4H2 plays two different roles in early and late development embryogenesis stages: it functions as an ectoderm inducer and neural inhibitor through cell autonomous BMP enhancement at early stages with a broader ectoderm expression pattern, and plays an important role in dorsal -ventral patterning of the neural tube at late stages when its expression is mainly restricted to the neural ectoderm tissues. As BMP signalling is also required for dorsal -ventral patterning of spinal cord, we could not rule out the possibility that the cell autonomous BMP enhancing activity of ZC4H2 also contributes to its functions in this late development process ZC4H2 is involved in. And also,    rsob.royalsocietypublishing.org Open Biol. 7: 170122 ubiquitination of Smad1/5. It is less likely that BMP membrane receptors are also stabilized by ZC4H2 as: (i) ZC4H2 is predicted as a nuclear protein with a strong nuclear location signal at its C terminus and, in fact, ZC4H2 showed clear nuclear localization in the cells we tested here (data not shown); and (ii) ZC4H2 itself cannot directly interact with Smurfs (data not shown) and only in the presence of wild-type Smads proteins, ZC4H2 and Smurfs existed in the same protein complex and showed Smad stabilizing activity (figure 8). Smad1/5 interact with ZC4H2 and Smurfs through MH2 and PY motifs in Linker domains, respectively (figures 6 and 8). We showed that ZC4H2 reduced the interaction between Smad1/5 and Smurfs. It is possible that interaction with ZC4H2 blocked the surface or destroyed the conformation of Smads needed for Smurfs interaction. And also, it is possible that the lysine residues in Smads targeted by Smurfs were masked by ZC4H2 interaction. We found that a serial ZC4H2 mutation isolated in patients with intellectual disorders showed weaker Smad-stabilizing activity and they all failed to rescue ZC4H2 morphants in Xenopus, implying   that the ZC4H2-Smad interaction may also contribute to proper neural development in humans [27,28]. BMPs regulate mesenchymal stem cell differentiation through promoting osteogenic differentiation and meanwhile blocking myogenic differentiation. This process is tightly controlled by a range of extracellular, intracellular and nuclear modulators [2,[7][8][9]17,18,34,35]. Using a mouse myoblast C2C12 cell differentiation model, we found that ZC4H2, as a Smads stabilizer, increased the BMP-induced osteogenic and decreased the myogenic potential of C2C12 cells by both gain-of-and loss-of-function approaches. Whether ZC4H2 is involved in muscle/bone development in vivo remains to be explored.

Experimental procedures
4.1. Xenopus embryos, whole mount in situ hybridization, semi-quantitative RT-PCR assays, microinjection and animal cap assays Adult Xenopus laevis frogs were obtained from Nasco. Xenopus embryos, whole mount in situ hybridization, microinjection and animal cap assays were carried out as described previously [36,37]. The sequence of the antisense MO for ZC4H2 which was obtained from Gene Tools is: ZC4H2 MO: 5 0 -GAATGAAATAAAATGCCAGCAAAAC-3 0 . MO and mRNA were injected into the dorsal region of 2-to 4-cell stage embryos. In total, 25 ng MO was injected per stitch, 0.6 ng ZC4H2 wild-type or its mutant mRNAs were injected for overexpression or rescue experiments. For in situ hybridization, the full-length of ZC4H2 cDNA sequence was used for probe preparation and the probes of Sox2 and Slug were used as described previously. The following primers were used for semi-quantitative RT-PCR assay: Xenopus ZC 4H2: forward, 5 0 -ATCAGAAACAAGACCCTCCAAAT-3 0 and reverse, 5 0 -CCTCTGTAGACCCAATGTCATCC-3 0 ; Pax3: forward, 5 0 -CAGCCGAATTTTGAGGAGCAAAT-3 0 and reverse,

ALP activity assay and ALP staining
Cells were cultured in 96-well plates, transfected and administered with 50 ng ml 21 rhBMP2 for 3 days. ALP activity was examined by using an Alkaline Phosphatase Assay Kit (Fluorometric) (ab83371, Abcam) following the manufacturer's protocols. ALP staining was performed by using BCIP/NBT Alkaline Phosphatase Color Development Kit (C3206, Beyotime) following the manufacturer's instructions.

Statistics
GraphPad and IMAGE J software was used for statistical analysis. Comparisons were performed using the two-tailed Student's t-test. p-Values of less than 0.05 were considered statistically significant and less than 0.01 were considered statistically very significant. All experiments were carried out at least three times and samples were analysed in triplicate.
Ethics. The care of Xenopus laevis (Nasco), in vitro fertilization procedure and manipulation of embryos were performed according to standard protocols. All animal protocols were approved by the Ethics Committee of Kunming Institute of Zoology, Chinese Academy of Sciences ( permit number: SYDW-2006-006).