ZFP423 Coordinates Notch and Bone Morphogenetic Protein Signaling, Selectively Up-regulating Hes5 Gene Expression*

Zinc finger protein 423 encodes a 30 Zn-finger transcription factor involved in cerebellar and olfactory development. ZFP423 is a known interactor of SMAD1-SMAD4 and of Collier/Olf-1/EBF proteins, and acts as a modifier of retinoic acid-induced differentiation. In the present article, we show that ZFP423 interacts with the Notch1 intracellular domain in mammalian cell lines and in Xenopus neurula embryos, to activate the expression of the Notch1 target Hes5/ESR1. This effect is antagonized by EBF transcription factors, both in cultured cells and in Xenopus embryos, and amplified in vitro by BMP4, suggesting that ZFP423 acts to integrate BMP and Notch signaling, selectively promoting their convergence onto the Hes5 gene promoter.


Zinc finger protein 423 encodes a 30 Zn-finger transcription factor involved in cerebellar and olfactory development. ZFP423 is a known interactor of SMAD1-SMAD4 and of Collier/Olf-1/EBF proteins, and acts as a modifier of retinoic acidinduced differentiation. In the present article, we show that ZFP423 interacts with the Notch1 intracellular domain in mammalian cell lines and in Xenopus neurula embryos, to activate the expression of the Notch1 target Hes5/ESR1. This effect is antagonized by EBF transcription factors, both in cultured cells and in
Xenopus embryos, and amplified in vitro by BMP4, suggesting that ZFP423 acts to integrate BMP and Notch signaling, selectively promoting their convergence onto the Hes5 gene promoter.
A small set of regulatory factors, mostly secreted or surface molecules, modulates neural development, from neural induction through synaptogenesis. Morphogens, acting instructively, permissively, or through inhibitory interactions, control various aspects of neurogenesis, eliciting different responses in target cells, dependent upon their evolving windows of competence. The integrated effects of various morphogens regulate a range of developmental switches, controlling, among other aspects, regional identity, fate determination, and the timing of neuronal commitment and differentiation. Although the activities of extracellular factors have been intensively studied, many of their cell-intrinsic effectors have yet to be discovered and characterized.
The Notch pathway exerts regulatory activities in a diverse array of developmental contexts (reviewed in Refs. 1,2). In neurogenesis, Notch signaling suppresses the neurogenetic cascade, which is promoted and sustained by proneural basic helix-loop-helix (bHLH) transcription factors. When the Notch single pass receptor is bound by the Delta or Jagged/ Serrate family of transmembrane ligands expressed by adjacent cells, the Notch intracellular domain (NICD) 4 is cleaved proteolytically (3,4) and translocates into the nucleus. There, it interacts with the DNA-binding protein CSL CBF1/Su(H)/LAG-1 (5), displacing a co-repressor complex (6) and recruiting a transcriptional activation complex (7)(8)(9). This leads to the transcription of various immediate target genes, including Drosophila enhancer of Split, and its vertebrate homologs Hairy, Hes1, and Hes5 (10 -12). These genes encode transcriptional repressors of the basic helix-loop-helix family, which act as inhibitors of neuronal differentiation. In the developing nervous system, Hes1, Hes5 double mutants feature a loss of mitotic progenitors and a massive premature differentiation, particularly in the dorsal neural tube (13).
Whereas this pathway effectively delays neuronal commitment and differentiation, it also promotes diversity in neuronal development by actively preserving a pool of uncommitted and mitotic neural progenitors, sustaining the birth of successive waves of distinct neuronal and glial types (reviewed in Ref. 14). To perform this broad array of modulatory effects at different developmental stages and in distinct morphogenetic domains, the Notch signaling cascade is tightly regulated, from the cell surface to the nucleus (reviewed in Refs. 15,16).
The Zfp423 gene is expressed alongside the dorsal midline of the embryonic mouse neural tube, at the border with the roof plate, particularly in the hindbrain and cerebellum (17). The roof plate provides critical signals for cerebellar development (18). The gene encodes a 30 Zn-finger domain nuclear protein involved in cerebellar and olfactory development. Interestingly, Zfp423 null mice develop a profound hypoplasia of the cerebellar vermis (19 -21), reminiscent of the Dandy-Walker malformation (reviewed in Ref. 22), and a premature differentiation of olfactory neuron progenitors (23), although the underlying molecular mechanisms remain unclarified.
ZFP423 is known to interact with the SMAD1-SMAD4 complex, which transduces bone morphogenetic protein (BMP2/ 4/7) signaling (reviewed in Ref. 24) into the nucleus, up-regulating Xvent2 transcription in Xenopus laevis gastrulae and mammalian cells (25). However, no information is available to date as to the functional significance (if any) of the interaction of ZFP423 and receptor-dependent SMADs in mammalian neural development.
ZFP423 has also been found to complex with EBF COE proteins (26,27). EBF COE TFs are important players in the context of neuronal differentiation and migration (28 -31), olfactory neurogenesis (32)(33)(34), cerebellar PC migration, and survival (35) (68) and cerebellar cortical patterning (35,36). Finally, a recent article described the role of ZFP423 as a modifier of retinoic acid-induced differentiation (37). Thus, ZFP423 is poised to interact with multiple signaling pathways and transcriptional effectors, likely integrating their function during development.
In the present work, we analyze some of the functional and molecular interactions established by ZFP423 in vitro and in vivo. We demonstrate that ZFP423 interacts functionally and molecularly with the NICD in mammalian cell lines and in Xenopus neurula embryos, to activate the expression of the Notch target Hes5/ESR1. A small proximal region of the Hes5 promoter is sufficient to reproduce this cooperation in vitro. This effect is enhanced in BMP4 treated cells. By triggering Hes5 expression and by modulating BMP signaling cell autonomously, ZFP423 may help maintain a pool of Hes5 positive neurogenic progenitors in the developing neural tube.

EXPERIMENTAL PROCEDURES
Animal Care-All experiments described in this report were conducted in agreement with the stipulations of the San Raffaele Scientific Institute Animal Care and Use Committee, and the University of Utah Institutional Animal Care and Use Committee guidelines.
Tissue Preparation-Pregnant mice were anesthetized with Avertin (Sigma). For in situ hybridization on sagittal sections, embryos were fixed overnight by immersion with 4% PFA, cryoprotected in 30% sucrose overnight, embedded in OCT (Bioptica), and stored at Ϫ80°C, before sectioning on a cryotome (20 m). For whole mount in situ hydridization and whole mount LacZ staining embryos were fixed with 4% PFA 6 h or 10 min, respectively. Zfp423 expression at embryonic day 10.5 was reexamined by LacZ staining using a transgenic line obtained from the German Genetrap Consortium (ID: W008G09, 38) carrying a LacZ gene inserted by gene trapping within the Zfp423 gene. LacZ staining was performed as described (35).
Xenopus Embryo Microinjection-Mouse Zfp423 from pCDNA3-Zfp423 was subcloned into the pCS2ϩ expression vector and used to make capped mRNA in vitro using the Message mMachine kit (Ambion). Also, the following constructs were used as DNA templates to make capped mRNA : pCS2ϩX-Delta stu (67), pCS2ϩMT-Xotch⌬E (referred to here as N act ) (39), pCS2ϩXebf2 (29), pCS2ϩXebf3 (29), pCS2ϩn␤gal (40), and pCS2ϩGFP (41). The full length Xenopus Zfp423 (XZfp423) cDNA including part of 5Ј-UTR was acquired by 5Ј-RACE (Roche) using the sequence of image clone 6636947, and then by RT-PCR with Superscript II Reverse Transcriptase (Invitrogen) and PfuUltraII fusion HS DNA polymerase (Stratagene) (GenBank TM Accession No. GQ421283). Control morpholino: the sequence of our XZfp423 morpholino (Gene Tools) is TCCACTGTACCTCAAAACTA-ACCCC, which is complementary to nucleotides Ϫ26 to Ϫ2. mRNA and morpholino were injected into one blastomere of 2-cell stage embryos in the following amounts: Zfp423 (1 ng for single injection and co-injection with Delta stu , 600 pg for coinjection with Xotch⌬E), Delta stu (400 pg), Xotch⌬E (100 pg), Xebf2 (100 pg), Xebf3 (100 pg), n␤gal (50 pg), and XZfp423 morpholino (30 ng). mRNA was injected into one dorsal blastomere of 16-cell stage embryos in the following amounts : Zfp423 (300 pg) and Gfp (200 pg). mRNA for n␤gal or Gfp was co-injected into all embryos as a tracer. Embryos were grown until neural plate stages (42) and fixed in MEMFA for 30 min (43). X-Gal staining was performed on the embryos injected with n␤gal as described (44). GFP-expressing embryos were sorted under fluorescence based on which side was the injected side and fixed for 30 more minutes.
RT-PCR-Total RNA was extracted with RNeasy MicroKit (Qiagen), according to the manufacturer's instructions. 1-1.5 g of total RNA was retrotranscribed using first strand cDNA MMLV-Retrotranscriptase (Invitrogen) and random primers. Each cDNA was diluted 1:10, and 3 l was used for each real-time reaction. mRNA quantitation was performed with LightCycler480 SYBR Green I Master Mix (Roche) on a LightCycler480 instrument (Roche) following the manufacturer's protocol.
RNA Interference-Single-stranded DNA oligos encoding the pre-miRNAs were annealed according to the manufacturer's instructions (Invitrogen). Pre-miRNA double-stranded oligos were cloned in pCDNA TM 6.2-GW/-EmGFP-miR vector (Invitrogen). A pre-miRNA containing a non-targeting sequence (Invitrogen) was used as a negative control. The pre-miRNAs were transfected via Lipofectamine2000 (Invitrogen) according to the manufacturer's instructions. To test the efficacy of the pre-miRNA, HEK293 cells were transfected with pcDNA3-6Myc-ZFP423 and either a non-targeting pre-miRNA or ZFP423-specific pre-miRNA vector, and cell lysates were analyzed via Western blot. Among the specific pre-miRNA tested, we selected the most effective one in abolishing protein expression. Transfected P19 cells were sorted for GFP expression and lysed for RNA extraction with RNeasy MicroKit (Qiagen), according to the manufacturer's instructions.
Coating Assay-To generate a soluble form of Notch1 ligand, 293T cells grown in a 10-cm Petri dish were transfected using Fugene HD (Roche), according to the manufacturer's instructions, with either 5 g of Fc-TRAIL-Receptor 4 (Fc-control) or 5 g of Fc-Jagged1 expression plasmids (Courtesy of Tom Kadesch, 51) After 48 h, the growth medium was collected and filtered through a 0.45-m syringe filter. Fc-TRAIL-R4 or Fc-Delta4 fusion protein was immobilized by incubating polystyrene plates for 2 h at room temperature with 10 g/ml rabbit anti-human IgG Fc antibody (Jackson ImmunoResearch Laboratories), and then incubated for 2 h with respective filtered supernatant. The pre-miRNAs were transfected in P19 cells via Lipofectamine 2000. After 48 h, the cells were splitted and grown for 24 h on plates coated with Fc-TRAIL-R4 (control) or Fc-Jagged1. Transfected P19 cells were sorted for GFP expression, and RNA was extracted with RNeasy MiniKit, according to the manufacturer's instructions.
Promoter Reporter Assays-The day before transfection, C2C12 cells were plated in 12-well plates and grown in DMEM supplemented with 10% FBS. Luciferase assays were carried out 24 h after transfection using Dual Luciferase Assay kit according to the manufacturer's instructions (Promega). Each result is the average of three independent measurements, and each experiment was repeated at least three times.
Immunoprecipitation and Western Blots-All biochemical procedures were conducted as described (52). In particular, COS7 and HEK293 cells were harvested 24 h after transfection and centrifuged; pellets were frozen at Ϫ80°C. For co-immunoprecipitation experiments, cell pellets were thawed at room temperature and lysed in 5 volumes of extraction buffer (10 mM Hepes pH 7.9; 400 mM NaCl; 5% glycerol, PMSF 1 mM, leupeptin 0.5 mM, NaF 50 mM, pepstatin 1 mM). Samples were centrifuged at 34,000 rpm for 30 min at 4°C, and supernatants were collected. Protein concentration was determined by the BCA assay (Pierce). Part of the lysate (20%) was kept as a positive input control. Lysates were incubated overnight with 10 g of the indicated antibodies; 30 l of protein G-Sepharose (GE Healthcare) were added for 4 h at 4°C. The resin was washed five times with extraction buffer. Protein complexes were eluted by addition of sample buffer (Tris-Cl, 125 mM pH 6.8; 0.1 M 2-mercaptoethanol; 2% SDS; 20% glycerol; 25 mg/ml bromphenol blue), boiled for 15 min and separated on an 8% SDSpolyacrylamide gel. Proteins were transferred on a PVDF membrane (Millipore). Western blots were performed with the following antibodies: mouse anti-Myc (9E10, Santa Cruz Biotechnology), rabbit anti-Notch (C-20, Santa Cruz Biotechnology), rabbit anti-OAZ (ZFP423) (H-105 Santa Cruz Biotechnology), mouse anti-␤-actin (Sigma). As secondary antibodies, a goat anti-rabbit HRP-conjugated antibody (Bio-Rad), and a sheep anti-mouse HRP-conjugated antibody (Amersham Biosciences) were used. Blots were developed with the LiteAblot substrate (EuroClone).
Statistical Analysis-Statistical significance was determined by the Student t test with a threshold for significance set to p ϭ 0.05. All results are plotted as the mean Ϯ S.D.

Zfp423 Expression in the Cerebellar Primordium-Because
Zfp423 mutants feature a severe cerebellar midline deletion (17,19,20), we re-examined the expression of Zfp423 at the onset of cerebellar neurogenesis. To this end, we used a transgenic line (ID: W008G09, Ref. 38) carrying a LacZ gene inserted by gene trapping within the Zfp423 gene. Zfp423 expression in the neural plate is detectable by whole mount in situ hybridization as early as E7.5 (not shown). Zfp423 strongly labels the rhombencephalon and mesencephalon at E8.5 and E9.5 (not shown, Ref. 17), and the hindbrain and cerebellum thereafter (Fig. 1A). In situ hybridization of embryonic tissue sections (Fig. 1, B-E) revealed that the gene is expressed in both germinative epithelia of the cerebellar primordium, i.e. the RL (arrow), that will give rise to all glutamatergic progenitors, and the cerebellar VZ (solid arrowhead), that harbors all GABAergic progenitors. Because Zfp423 mutants feature a severe cerebellar midline deletion (17,19,20), we asked if the gene is expressed at this site ( Fig. 1, B and D). Interestingly, in the E12.5 cerebellar primordium, Zfp423 is expressed at high levels flanking the midline, where it colocalizes with Hes5 (Fig. 1, BЈ-DЈ), an immediate transcriptional target of the Notch1 signaling pathway expressed in the VZ and RL. In contrast, Hes1 is more restricted at this stage, labeling the isthmic organizer (empty arrowhead) and rhombic lip (arrow), adjacent to the cerebellar roof plate, while it is considerably down-regulated in most of the VZ (Fig. 1, AЉ-EЉ).

Functional and Molecular Interactions between ZFP423 and
Notch Signaling-Based on the above observations, we sought to determine whether ZFP423 acts in the context of the Notch signaling pathway. To this end, we overexpressed the corresponding gene in neuralized P19 cells (see "Experimental Procedures") together with a cDNA encoding NICD (53). Shown in Fig. 2A is a sample semi-quantitative RT-PCR analysis of P19 cells transfected with a fixed amount of Nicd and increasing concentrations of Zfp423, illustrating the conditions achieved in subsequent experiments.
We measured the levels of four known direct targets of Notch signaling: Hes5 and Hes1 (12), Nrarp (54) and Blbp (55). As regards Hes5 (Fig. 2B), whereas overexpression of Nicd alone produces a significant increase in Hes5 gene expression over mock-transfected cells, the overexpression of Zfp423 alone has no effect. However, combined overexpression of Nicd and Zfp423 elicits a strong cooperative interaction, leading to a significant increase in Hes5 transcript levels that is dependent on Zfp423 DNA dosage. Likewise, Blbp gene expression (Fig. 2C) is cooperatively up-regulated by NICD and ZFP423, whereas Hes1 and Nrarp transcript levels (Fig. 2, D and E, respectively) are unaffected by Zfp423 overexpression in addition of Nicd. Because neuralized P19 cells express endogenous Hes5 and Zfp423 (Fig. 1A and Ref. 25), we moved to a system in which neither gene is expressed: the C2C12 myoblastoid cell line (25). In this system, overexpression of Nicd up-regulates Hes1 and Hey1 (56), but activates Hes5 transcription very weakly (our observation). Our results indicate that in C2C12 cells Hes5 expression is strictly dependent upon co-expression of Nicd and Zfp423, and that the levels of Hes5 expression are again Zfp423 dose-dependent (Fig. 2F). Next, we asked whether ZFP423 contributes to Hes5 gene regulation even at physiological levels of expression. To address this point, we used RNA interference (RNAi). The efficiency of Zfp423 knockdown was tested and validated at the protein level (supplemental Fig. S1). RNAs extracted from P19 cell lysates were analyzed by RT-qPCR for Hes5 gene expression. In unstimulated P19 cells, transfected with a Zfp423 shRNA, Hes5 transcription was clearly down-regulated (Fig. 2G). Next, we activated endogenous Notch signaling by growing P19 cells, transfected with the Zfp423 shRNA, onto plates coated with a secreted form of the Notch ligand Jagged (Fc-Jagged) (51) or with an inactive control (Fc-TRAIL-R4). Our results indicated that in Jagged-activated P19 cells, Hes5 is up-regulated (Fig. 2H), and that Zfp423 RNAi causes a significant down-regulation of the same gene.
Next, we asked if the functional cooperation found to occur between ZFP423 and NICD also takes place in vivo. To address this question, we used Xenopus laevis neurula embryos. We identified a Xenopus expressed sequence tag clone (Image clone 6636947, GenBank TM accession: BU911031) very similar (83% identity at the nucleotide level) to Gallus gallus ZFP423 (see "Experimental Procedures"). This clone was used as an in situ probe to analyze Zfp423 distribution in neurula and tailbud stage embryos. The gene is expressed in the head and spinal cord (Fig. 3, A and B; see also stages 28 and 34 in supplemental Fig. S2). Subsequently, two-cell Xenopus embryos were injected unilaterally with mouse Zfp423 and/or mRNA for Xenopus Notch ⌬E (N act ); ␤-galactosidase  OCTOBER 1, 2010 • VOLUME 285 • NUMBER 40 JOURNAL OF BIOLOGICAL CHEMISTRY 30817 mRNA was included to mark the injected side. N act encodes an N-terminally deleted, constitutively active Xenopus NOTCH protein (39). Both N act and Zfp423 mRNA concentrations were titrated so that either construct would produce moderate changes in target gene expression when overexpressed alone. As a first target, we analyzed the Hes5 ortholog ESR1 (57,58). Our results (Fig. 3, C-E) indicate that embryos injected with Zfp423 (600 pg) alone show a moderate activation of ESR1 expression (Fig. 3C); likewise, injection of low amounts of N act alone (100 pg) promoted a low-level expansion of the ESR1 domain (Fig. 3D); however, the simultaneous overexpression of N act (100 pg) and Zfp423 (600 pg) sharply increased ESR1 expression on the injected side (Fig. 3E). Furthermore, we asked whether the effect on ESR1 expression produced by the injection of Zfp423 alone (600 pg, Fig. 3C; 1 ng, Fig. 3F) requires endogenous Notch pathway activation. To address this question, Zfp423 (1 ng) was coinjected with a dominant negative Delta ligand (DeltaStu, 40). Delta Stu coinjection ablated the expansion of ESR1 consequent to Zfp423 overexpression (Fig.  3G), indicating that this response is strictly dependent upon endogenous Notch signaling activation. Finally, we asked whether the endogenous expression of ESR1 depends upon the presence of endogenous ZFP423. We injected 2-cell embryos with a ZFP423-specific morpholino antisense oligonucleotide and found a significant down-regulation of ESR1 on the injected side (Fig. 3H) while the control morpholino had no effects on ESR1 expression (Fig. 3I).

ZFP423 Integrates Notch and BMP Signaling
Because of the previously reported role of ZFP423 as a SMAD cofactor in the context of mesodermal patterning (25), the overexpression experiments were repeated by targeting unilaterally a dorsal blastomere at the 16-cell stage to prevent a possible interaction of exogenous ZFP423 with the BMP effector complex p-SMAD1-SMAD4 during gastrulation. The results of this experiment further corroborated the notion that Zfp423 overexpression promotes ESR1/Hes5 up-regulation on the injected side (Fig. 3I). In the same experiment, the Hes1 ortholog Hairy1 (10, 46) was either unchanged or slightly down-regulated on the injected side (Fig. 3J) and Nrarp (47,54) transcript levels were also either left unchanged or slightly down-regulated (Fig. 3K) in response to Zfp423 overexpression, indicating that Zfp423 promotes a dissociation in the response to NICD in vivo, favoring the expression of Hes5 over Hes1 or other targets. Taken together, our in vivo results are consistent with those previously obtained in cell lines and point to a role for Zfp423 as a cell-autonomous modifier of Notch signal transduction.
Hes5 Promoter Analysis-To identify and map Hes5 promoter sequences mediating the transcriptional response of Hes5 to ZFP423 and NICD, we performed luciferase assays in C2C12 cells, using different extents of the Hes5 5Ј-flanking region, as described (61,62). NICD binds to the Hes5 promoter by forming a complex with CBF1/RBPj/CSL on a CBF1 recognition site (5), located 153 bp upstream of the murine Hes5 ATG. At first, we used a luciferase (luc) reporter containing 1051 bp of the Hes5 5Ј-flanking sequence ϩ 5Ј UTR (sketched in Fig. 4A). Our results show that coexpression of Nicd and Zfp423 leads to a significant (2-fold) increase in luc expression compared with Nicd alone (Fig. 4A). Similar results were obtained using a reporter construct containing a shorter stretch of the Hes5 promoter/5Ј UTR (267 bp) (Sketched in Fig. 4B), although the absolute activity levels were lower (Fig. 4B). These results indicate that while distal promoter regions contain elements that are important for Hes5 gene activation, those regions are however not essential for the cooperation between NICD and ZFP423 to occur.
Next, we set out to establish whether ZFP423 interacts in vivo with the Hes5 promoter. To this end, we performed a ChIP experiment using neuralized P19 cells (Fig. 4C) (see "Experimental Procedures"). P19 cell chromatin was immunoprecipitated using a ZFP423 Ab. For PCR amplification, we developed Hes5 promoter-specific primers amplifying a 94-bp product spanning the BMP-responsive element (BRE) located within the Hes5 promoter region depicted in Fig. 4B, an extremely GC-rich, PCR unfriendly sequence. We performed a qPCR using the Hes5 primer pair and different control primers corresponding to an upstream 5Ј-flanking sequence (Ϫ2500 bp) and to two downstream sequences (ϩ1600 bp and ϩ2500 bp). In addition, we analyzed a syntenic gene (Mrps15) located megabases away from Hes5 (63). The histogram in Fig. 4C shows a massive fold-enrichment for the Hes5-promoter-specific product with respect to flanking Hes5 sequences and to the syntenic Mrps15 gene.
Molecular Interactions between ZFP423 and NICD-Because our results indicated that Zfp423 and Nicd cooperate functionally in vitro and in vivo, we investigated whether the corresponding proteins interact at the molecular level. First, COS7 cells were cotransfected with constructs encoding 6xmycZFP423 and flagNICD. Lysates were immunoprecipitated with an irrelevant anti-GFP, or with an anti-Myc monoclonal antibody. Immunoprecipitates were analyzed by WB using Abs for ZFP423 and for NICD, revealing an NICD-specific band only in the Myc-immunoprecipitated lane (Fig. 5A). To exclude the possibility that NICD might bind nonspecifically to the beads or the Myc antibody, the experiment was repeated, and lysates from singletransfected cells were immunoprecipitated as negative controls (Fig. 5C). NICD coimmunoprecipitated only in the cell lysates containing both factors.
EBF TF Overexpression Abolishes the Cooperation of ZFP423 with NICD-In previously published work (29), we showed that Nicd overexpression was capable of reducing or abolishing the ability of XEbf2 to induce Nfm gene expression in Xenopus neurulas. This result suggested that Notch and EBF2 might act antagonistically in neuronal differentiation, through a mechanism independent of Ebf2 gene transcription. Because ZFP423 is a molecular interactor of EBF TFs, we repeated the experiment described in Fig. 2B and added Ebf1, Ebf2, or Ebf3 to the transfection mix. Transfected cells were harvested, lysed, and analyzed by RT-qPCR for Hes5 mRNA levels. Our results indicate that Ebf1-3 overexpression alone has no detectable effect on Hes5 transcription. However, cotransfecting cells with Nicd,  OCTOBER 1, 2010 • VOLUME 285 • NUMBER 40

JOURNAL OF BIOLOGICAL CHEMISTRY 30819
Zfp423 and either Ebf1, Ebf2, or Ebf3 antagonized the cooperative effect of ZFP423 and NICD on Hes5 gene transcription (Fig. 6A). The experiment was repeated in vivo, by injecting 2-cell Xenopus embryos unilaterally with the same combination of mRNAs. Again, our in vivo results faithfully recapitulated those obtained in P19 cells: coinjection of either XEbf2 (C) or XEbf3 (Fig. 6D) reduced ESR1 activation induced by exogenous Zfp423 (Fig. 6B). XEbf2 or XEbf3 injection led to a low level of ectopic ESR1 expression (Fig. 6, E and F). This might be caused by increased X-Delta-1 expression, driven by XEBF2 (30) or to expression in differentiating neurons outside of the neural plate (28). Because our data indicate that EBFs interfere with the cooperation of ZFP423 and NICD, we asked if EBF overexpression can antagonize the assembly of the ZFP423-NICD molecular complex. To address this point, COS7 cells were cotransfected with 6xmycZFP423, flagNICD, and flagEBf3. Lysates were immunoprecipitated with anti-Myc and analyzed by WB using Abs for ZFP423, NICD and EBF3. Both NICD-specific and EBF specific bands were present in the anti-Myc immunoprecipitates (Fig. 5C). Again, lysates from single-transfected cells were used as negative controls. These results suggest that the three proteins form a complex exhibiting a reduced ability to activate Hes5 gene expression.
Zfp423 Enhances the Cooperation between NICD and BMP/ SMAD Signaling-Previous studies (25) showed that ZFP423 interacts with the SMAD1/SMAD4 complex in the nucleus to activate the Xvent2 promoter in response to BMP signaling activation. Moreover, other authors have described the existence of interactions between BMP and Notch signaling pathways, resulting in repression of neurogenesis (61) and myogen-  . Molecular interaction between ZFP423 and NICD. In A, COS7 cells transfected with the indicated constructs, were subjected to immunoprecipitation using anti-GFP as an irrelevant antibody, or anti-Myc to precipitate 6xmycZFP423. Filters were cut and stained for ZFP423, NICD, and actin (unrelated protein). Only NICD coprecipitated in the fraction immunoprecipitated with anti-Myc. B, lysates of COS7 cells, transfected with the indicated constructs, were blotted and immunostained as shown. C, lysates shown in B were immunoprecipitated with the anti-Myc antibody. Filters were immunostained for ZFP423, NICD, and EBF3. Notably, NICD coprecipitated equally with ZFP423 in the presence or absence of EBF3. esis (56), while others have proposed BMP as a rhombic lip (RL)-inducing signal in cerebellar development, antagonized by Notch signaling, which sets the RL ventral limit (64). Results obtained by other authors indicate that ZFP423 is a functional and molecular interactor of the SMAD1-SMAD4 complex. That interaction occurs thanks to a domain constituted by Zn fingers 9 -20, encoded by exon 4 (25,65). We asked if Zfp423 uses the same domain to interact with NICD. To address this question we generated a deleted construct missing Zn-fingers 9 -20 (sketched in Fig. 7A). First, we overexpressed the construct in COS7 cells to determine the subcellular localization of the corresponding protein. ⌬9 -20 ZFP423 is expressed and displays a prevalent nuclear localization (Fig. 7B). Subsequently, we performed a promoter-reporter assay in C2C12 cells to compare the ability of wt and ⌬9 -20 ZFP423 to regulate Hes5 gene expression. Our results indicate that ⌬9 -20 ZFP423 fails to cooperate with NICD in Hes5 gene activation (Fig. 7C).
Because ZFP423 uses the same domain to interact functionally with both NICD and the SMAD complex, we asked whether and in which way ZFP423 might modulate the response of Hes5 to BMP signaling activation (61). In vivo, Zfp423 is expressed in the cerebellar primordium flanking the roof plate (Fig. 1A), a territory that expresses high levels of various BMP family molecules. To determine whether ZFP423 coordinates Notch-and BMP signaling recruiting both pathways in the activation of Hes5 transcription, we performed an RT-qPCR experiment on mRNA extracted from C2C12 cells, measuring Hes5 transcript levels in response to NICD and ZFP423, and in the presence or absence of purified BMP4 in the culture medium (Fig. 7D). To exclude the possibility that Hes5 activation could stem from unequal expression levels of NICD in cells co-transfected with NICD and ZFP423 versus NICD alone, the levels of Nicd were analyzed by RT-qPCR and found to be virtually unchanged (Fig.  7E). Our results reproducibly indicate that, in the absence of ZFP423, a 2-h BMP treatment does up-regulate Hes5 compared with the transcript levels achieved by transfecting the cells with Nicd alone. The interaction between BMP4 and NICD observed in C2C12 cells confirms the results reported by other authors (61). However, cotransfecting C2C12 with ZFP423 significantly potentiates (over 10-fold) the co-activating effect of BMP4 on NICD-induced Hes5 gene expression (Fig. 7D), suggesting that this protein may play an important role integrating BMP and Notch signaling in dorsal territories of the neural tube.

ZFP423 Interacts Functionally and Molecularly with NICD to
Activate Hes5 Gene Expression-In this report, we show that ZFP423 interacts functionally with the Notch intracellular domain to activate cooperatively and selectively the expression of one direct Notch target: Hes5 (working model in Fig. 8). This effect occurs on a small stretch of Hes5 proximal promoter, containing both a CBF1 binding site and a BMP responsive element. This interaction occurs in vivo, as shown by the results of ChIP experiments. However, ZFP423 has no noticeable effect on the expression of other NICD targets, such as Hes1 or Nrarp. This conclusion is supported by experiments conducted both in cell lines and Xenopus embryos. The results of both gain-of-function and loss-of-function experiments, conducted in vivo and in vitro, support the notion that ZFP423 and NICD cooperate in Hes5 regulation. Strikingly, in C2C12 cells, that are negative for both the Hes5 and Zfp423 transcripts, the expression of Hes5 is strictly dependent upon the addition of exogenous Zfp423, as Nicd overexpression alone is not sufficient to activate it significantly, while it activates other Notch targets. This result suggests that ZFP423 acts selectively to recruit NICD onto the Hes5 promoter.
ZFP423 May Coordinate BMP4 and Notch Signaling to Activate Hes5 Gene Expression Flanking the Dorsal Midline-Our results indicate that ZFP423, a known nuclear interactor of the BMP-dependent SMAD complex (25), acts to modulate the function of NICD. The corresponding transcript, Zfp423, is expressed in two symmetric stripes of cells flanking the roof plate in the cerebellar primordium (present work, Fig. 1, and Ref. 21). Thus, we wondered if ZFP423 might integrate Notch signaling and roof plate signals cell autonomously to promote Hes5 gene expression. Functional interactions between Notch and BMP signaling have been observed in urogenital and endothelial cells (59,66), and BMP has been known to activate Hes5 transcription mildly in mouse neuroepithelial cells (61), but the role of ZFP423 in this context was not examined. In C2C12 cells, we find that ZFP423 triggers a cooperative interaction between NICD and the SMAD complex, leading to a strong activation of Hes5 gene expression.
While this result could not be replicated in Xenopus embryos, because BMP activation in early embryos interferes with neural induction, in vitro studies reproducibly reveal that the cooperative interaction existing between NICD and the BMP pathway is enhanced by ZFP423 and results in Hes5 up-regulation. For its cooperation with NICD (present work) and SMAD1/4 (25), ZFP423 uses the same domain, containing Zn fingers 9 -20. This is a very large domain that likely accommodates both proteins permitting their cooperative rather than antagonistic interaction.
Previous reports have shown that Zfp423 null mutants feature a disorganized cerebellar ventricular zone with disassembled radial glia. In neuralized P19 cells, ZFP423 regulates both Hes5 and Blbp transcription in response to NICD. In vivo, Hes5 transcription is strictly dependent upon Notch signaling activation and Hes5 is expressed in asymmetrically dividing radial glia (12,60). In the cerebellar primordium, Hes5 labels the cerebellar VZ and rhombic lip, whereas Hes1 is highly expressed in the rhombic lip and isthmic organizer, but is downregulated in most of the VZ (Fig. 1). We speculate that in dorsal territories of the cerebellar primordium the interaction occurring between Notch and BMP signaling, and enhanced by ZFP423, could maintain a pool of Hes5-positive radial glial progenitors supporting several rounds of asymmetric, neurogenic cell division, and prevent the premature occurrence of terminal differentiation. In keeping with this interpretation, expression of the radial glia marker BLBP has been found reduced in Zfp423 mutants (20).
EBF TFs Antagonize the ZFP423-NICD-mediated Activation of Hes5 Gene Expression-The cooperative interaction established by ZFP423 and NICD can be quenched by cotransfecting cells with cDNAs encoding TFs of the EBF COE family. EBF TFs also block the cooperative interaction between ZFP423 and N act in Xenopus embryos. EBF TFs have been implicated in neuronal differentiation, and ZFP423 acts as an EBF antagonist both in promoter-reporter assays (26) and, in vivo, in olfactory neurogenesis (23). The reported ability of ZFP423 to block differentiation when expressed ectopically in postmitotic precur- ⌬9 -20 ZFP423 localizes partially in the cell nucleus but fails to activate Hes5 gene expression in cooperation with NICD. A, scheme of the in-frame deletion of exon 4 producing a protein devoid of Zn fingers 9 -20, implicated in the interaction with SMAD proteins and the BMP-responsive element. B, immunofluorescence analysis of the subcellular localization of a Myc-tagged wt and ⌬9 -20 construct in COS7 cells. Note that the mutant protein has a nuclear localization, although in some cells it is also distributed in the cytoplasm. DAPI labels DNA. ovl, overlay. C, histogram illustrating the results of a promoter-reporter assay revealing the lack of a cooperative interaction between Zfp423 ⌬9 -20 and NICD in Hes5 gene activation. D, cooperative activation of Hes5 gene expression by BMP4 and Notch mediated by ZFP423. Real-time RT-qPCR analysis of RNA from C2C12 cells, mock transfected or transfected with either Nicd, Zfp423, or both. Cells were either left untreated or treated with BMP4 for 2 h. Transfection with Zfp423 strongly enhances the cooperative effect of NICD and BMP signaling on Hes5 gene expression. E, RNAs from the C2C12 cell lysates analyzed in D were subjected to RT-qPCR using primers specific for flagNICD. FlagNICD expression levels are comparable in the presence and absence of Zfp423. This excludes the possibility that differences in Hes5 expression depend on different levels of flag-Nicd in Zfp423-transfected versus Zfp423-untransfected samples. *, p Ͻ 0.05; **, p Ͻ 0.005. sors, and to revert them to an undifferentiated state, could be explained partially by the ability, shown here, of ZFP423 to promote Hes5 gene expression. Conversely, the ability of EBF TFs to couple cell cycle exit to the onset of differentiation may stem in part from recruiting ZFP423 into molecular processes other than its cooperation with Notch signaling. In other words, EBF TFs may compete with NICD for ZFP423 in differentiating radial progenitors, and recruit ZFP423 in neuronal differentiation and migration (28). Our co-immunoprecipitation results indicate that EBF3 does not interfere significantly with the assembly of a ZFP423-NICD complex, thus suggesting that the presence of EBF TFs may recruit NICD and ZFP423 into a distinct regulatory network, sequestering it from the Hes5 promoter.
In summary, ZFP423 synergizes with NICD and cell-autonomously integrates BMP signaling with the Notch pathway, leading to the selective up-regulation of one direct target of Notch: Hes5. By doing so, Zfp423 may promote the maintenance of an undifferentiated radial glial progenitor pool. In vivo gain-of-function studies are now required to fully elucidate the roles of Zfp423 in the context of cerebellar neurogenesis.