|Home | About | Journals | Submit | Contact Us | Français|
The online version of this article has been published under an open access model. Users are entitled to use, reproduce, disseminate, or display the open access version of this article for non-commercial purposes provided that: the original authorship is properly and fully attributed; the Journal and Oxford University Press are attributed as the original place of publication with the correct citation details given; if an article is subsequently reproduced or disseminated not in its entirety but only in part or as a derivative work this must be clearly indicated. For commercial re-use, please contact journals.permissions/at/oupjournals.org
The gata2 gene encodes a transcription factor implicated in regulating early patterning of ectoderm and mesoderm, and later in numerous cell-specific gene expression programs. Activation of the gata2 gene during embryogenesis is dependent on the bone morphogenetic protein (BMP) signaling pathway, but the mechanism for how signaling controls gene activity has not been defined. We developed an assay in Xenopus embryos to analyze regulatory sequences of the zebrafish gata2 promoter that are necessary to mediate the response to BMP signaling during embryogenesis. We show that activation is Smad dependent, since it is blocked by expression of the inhibitory Smad6. Deletion analysis identified an octamer binding site that is necessary for BMP-mediated induction, and that interacts with the POU homeodomain protein Oct-1. However, this element is not sufficient to transfer a BMP response to a heterologous promoter, requiring an additional more proximal cooperating element. Based on recent studies with other BMP-dependent promoters (Drosophila vestigial and Xenopus Xvent-2), our studies of the gata2 gene suggest that POU-domain proteins comprise a common component of the BMP signaling pathway, cooperating with Smad proteins and other transcriptional activators.
Bone morphogenetic proteins (BMPs) are members of the TGF-β superfamily of extracellular ligands, which signal by binding to heterodimeric serine/threonine kinase receptor complexes. Much progress has been made identifying the molecular components that regulate transduction of a BMP signal from the plasma membrane to the nucleus (1,2). Ligand binding activates the receptor complex, which relays the signal by recruiting downstream mediators. A commonly used subset of mediators are known collectively as Smads, which are categorized into three functional groups: (i) receptor regulated Smads (R-Smads) that are directly phosphorylated by the activated receptor; (ii) Smad4, which cooperates with R-Smads to form an active signaling complex; and (iii) inhibitory Smads (I-Smads), which serve as negative regulators of the pathway (3). Although there may be exceptions (4,5), generally R-Smads 1, 5 and 8 are phosphorylated by the BMP pathway, whereas R-Smads 2 and 3 function instead to transduce a TGF-β, nodal or activin signal. The R-Smad/Smad4 complex interacts with specific nuclear transcription factors to activate gene transcription. There are also Smad-independent mechanisms to mediate BMP signaling, e.g. by p38 MAPK (6).
Defining the mechanisms by which BMP-induced Smads activate specific target genes is an important goal (7) given the wide range of developmental and physiological responses that are under the control of the pathway. BMP signaling patterns early germ layers to establish a dorsal/ventral mesoderm axis, the anterior/posterior endoderm character and the distinction of neural/epidermal ectoderm (8–12). BMP signaling also regulates lineage and morphogenetic programs relevant to bone, cartilage, kidney, heart and reproductive organ development. Thus, a wide range of highly specific gene expression programs is coordinated by the action of this common signaling pathway, presumably by the presence or absence of intersecting signaling pathways and specific nuclear co-factors.
An R-Smad/Smad4 complex binds DNA weakly on its own, relying on interaction with other nuclear partners to achieve stable and functional binding (13). A paradigm of Smad–cofactor interaction was established for TGF-β/activin signaling by the identification of winged helix proteins, such as FAST-1 as DNA-binding partners that interact with Smad2/3 at target promoters (14,15). With respect to the BMP pathway, an analogous example is the 30-zinc finger protein OAZ, shown to interact with BMP-induced Smads to activate the promoter for the homeobox gene Xvent2 (16). Most, but not all, known BMP response elements (BREs) are also Smad binding elements (SBEs), including those in the promoters for the genes encoding Vestigial (17), xVent2B (16,18), Smad6 (19), Id1 (20,21), Dlx3 (22) and Hex (23). A BMP-4 syn-expression motif was identified (TGGCGCC) as a conserved BRE (7), and SBEs have in some cases (17,24) been associated with GC-rich elements (GCCGnCGC) or as other defined short (GTCT or GCAT) motifs (18,25). However, there is no single consensus sequence that can readily predict a functional Smad1 binding site. Given that Smads have weak affinity for DNA, it is necessary to define the DNA-binding cofactor(s) to understand how Smads are targeted to any particular BRE.
One of the best characterized BMP-regulated pathways relates to the induction of ventral/posterior cell fate during Xenopus development. The homeobox transcription factor Xvent2 was the first target gene of BMP signaling to be investigated mechanistically, and several elements that contribute to BMP responsiveness have been defined, including an SBE/OAZ BRE, GCAT motifs, Vent2 auto-regulatory binding sites, and most recently, binding sites for the POU-domain protein Oct-25 (26). Another gene important to the ventral pathway is gata2, encoding a zinc finger transcription factor that is expressed throughout ventral ectoderm and ventral and lateral mesoderm (27–29). The gata2 gene is induced by BMP-4 (30,31) and forced expression of an engrailed-Gata2 fusion protein is sufficient to dorsalize Xenopus embryos (32), providing strong evidence that Gata factor activity is essential for ventral cell fate. Gata2 and xVent2 cooperate to activate the xVent1 gene, placing these two transcription factors within a common ventral network (32,33). Here, we investigate the mechanism by which the gata2 gene is activated by BMP signaling. We establish a reporter assay to define sequences of the gata2 promoter that mediate induction by BMP-4 during embryonic development. We find that an octamer binding site, interacting with Oct-1, is necessary for induction, but only functions in the context of a separate more proximal DNA sequence. The results provide an independent example of a POU-homeodomain protein mediating induction by BMP signaling, and implicate a common network controlling regulatory genes of the ventral pathway.
RNA was isolated from zebrafish embryos (75% epiboly) with Tri-reagent (MRC, Inc.). Superscript Reverse Transcriptase (BRL) was used to generate cDNA, followed by treatment with RNAse. Adapters were ligated to the cDNA ends using the Invitrogen 5′ RACE kit as per the manufacturer's instructions. A 3′ gene-specific reverse primer, corresponding to position +92 (relative to the translation start site) of the gata2 transcript, was used to amplify cDNA by PCR. The primer sequence was 5′-GTAAGCCCGTGATGATGTGACTCT (TE872). A single PCR product was identified in this analysis and subcloned in pCR-2.1 TOPO (Invitrogen). Sequence analysis confirmed that the clone included the previously characterized gata2 cDNA, with additional 5′-untranslated region. The start site and exon/intron boundaries were determined by comparing this new cDNA sequence with the genomic sequence.
A genomic region including ~7.1 kb of sequence upstream of the initiation methionine (6.5 kb upstream of transcriptional start site) was provided by Shuo Lin (UCLA), and transferred to the Kpn1 site of the pGL3basic luciferase reporter plasmid (Promega). The deletion to −1070 was facilitated by a unique MluI restriction site at that position. The other 5′ deletion constructs were generated by PCR, incorporating in the upstream primer a KpnI site and using a reverse primer corresponding to position +58 of the gata2 promoter. PCR products were cloned into the pCR-2.1 TOPO vector (Invitrogen). Subsequent cloning steps took advantage of an internal BssH2 site located at position −97. Therefore, each PCR-derived clone was digested with KpnI/BssH2 and the fragment used to substitute the full-length gata2 promoter sequence in the context of the pGL3basic luciferase reporter. PCR primers used to generate the progressive deletions were (KpnI sites underlined): −890 Kpn I forward, 5′-AAAGGTACCGCCTGCTGCTTGTTTTGCC (TE851); −819 KpnI forward, 5′-AAAGGTACCGGCTTTATTTGGACCTGC (TE876); −751 KpnI forward, 5′-AAA GGTACCAAGTGACGTTTGATCGCT (TE878); −728 KpnI forward, 5′-AAAGGTACCCGGCATACAAAGACAGTG (TE852); +58 reverse, 5′-CCCGATTGTTAAAGTCTCCC (TE837).
To transfer a ‘minimal’ BRE region upstream of the SV40 promoter, a ClaI site at position −443 was used with promoter constructs truncated at −819 or −751. Plasmids were digested with KpnI/ClaI, transferred first into pBluescript KS digested with KpnI/ClaI, re-excised with KpnI and SmaI, and finally transferred into similarly digested SV40 pGL3promoter luciferase plasmid (Promega). This resulted in reporter plasmids that transfer from −819 to −443 (377 bp) or −751 to −443 (308 bp) of the gata2 promoter, respectively. Additional minimal promoter constructs (which transfer 233, 184 or 134 bp) were generated by PCR, first cloning products into pCR-2.1 TOPO, followed by re-excision with KpnI and SacI, and transfer into the KpnI/SacI sites of the SV40 pGL3promoter luciferase plasmid (Promega). This resulted in the transfer of regions −819 to −586 (233 bp), −819 to −635 (184 bp), or −819 to −685 (134 bp), respectively. The PCR primers used for this purpose were (restriction sites underlined): −819 KpnI forward, 5′-AAAGGTACCGGCTTTATTTGGACCTGC (TE 876); −685 SacI reverse, 5′-AAAGAGCTCGTCACCACATTCTTCTTG (TE 1127); −635 SacI reverse 5′-AAAGAGCTC CTCTAAATTGGGGGCTATTG (TE1161); −586 SacI reverse, 5′-AAAGAGCTCCTCAACTCCTAGCACCTC (TE1128).
For site-directed mutagenesis, the GeneEditor in vitro mutagenesis system (Invitrogen) was used with the following oligonucleotides (mutations underlined): mutation #1, 5′-TCGATAGGTACCACTTTTATTTGGACC (TE1009); mutation #2, 5′-ACCGGCTTTATTGCTACCTGCCCATGC (TE1012); mutation #3, 5′-GGCTTTATTTGGGAGTGCCCATGCGAC (TE1013); mutation #4, 5′-TGGACCTGCCCAAAGGACCTGTCGGCAC (TE1016); mutation #5, 5′-CACCTCCAAGAGACTAGTTCGCTATTAATATGTAAAGTG (TE1134); mutation #6, 5′-AGAGACGGGCTCGCTCTAGATATGTAAAGTGACG (TE1135); mutation #7, 5′-GGGCTCGCTATTAAGCTTTAAAGTGACGTTTGATCG (TE1147).
RNA used for micro-injection was obtained by in vitro transcription using linearized pCS2 vectors. The vectors were pCS2-lacZ, pCS2-xSmad1 (from J. Thomsen), pCS2-zSmad5 (from M. Mullins), pCS2-xSmad6 (from J. Christian), pCS2-xSmad8 (34) and pCS2-BMP4 (from C. Wright) were used as templates for in vitro transcription. One microgram of linearized template was used to generate capped mRNAs with the mMessage mMachine kit (Ambion), followed by precipitation with LiCl. RNA was quantified by optical density, and integrity was confirmed by gel electrophoresis.
Eggs were obtained from female Xenopus by standard gonadotropin induction protocols. Eggs were fertilized in vitro and dejellied in a solution of 2% cysteine (pH 8.0). Microinjection was performed with embryos in 0.1× MBS and 5% Ficoll. Each injection mix contained 62.5 pg of RNA encoding either BMP-4 or an irrelevant control RNA (such as lacZ), 25 pg of luciferase reporter DNA and 0.10 pg of RNA encoding Renilla luciferase (to normalize reporter activity). In some experiments, 250 pg of RNA encoding either xSmad1, zSmad5 or xSmad8 was included in place of BMP-4 RNA. In a standard assay, mixes were injected into the two dorsal blastomeres of a four-cell stage embryo. To generate lysates for gel mobility shift experiments, RNA was injected at the two-cell stage.
Embryos were staged according to Nieuwkoop and Faber (35). Following microinjection, embryos were cultured in 0.1× MBS until stage 13, at which point embryos were collected and lysed in 1× Passive Lysis Buffer (Promega). Luciferase assays were performed according to the manufacturer's instructions. For each assay, four embryos were lysed in 100 μl of buffer, and typically 3–5 sets were used to generate independent data points for each independent experiment, using 5 μl of lysate per 50 μl of luciferase substrate, measured in a Turner TD-20e luminometer. The firefly luciferase values were divided by corresponding values for the control Renilla luciferase. Normalized data from embryos injected with BMP-4 was divided by normalized numbers obtained from embryos injected with control RNA to determine ‘BMP-4 fold induction’. All luciferase data represent averages of at least three independent microinjection experiments. Error bars in the figures indicate standard error of the mean.
Total cell extracts were prepared by freezing embryos in liquid N2, followed by homogenization in total cell extract buffer (50 mM Tris–HCl, pH 8.0, 400 mM KCl, 1 mM DTT, 25% glycerol, 50 mM NaF and protease inhibitors). Extracts were cleared by centrifugation three times for 10 min at 8160g 10–20 μg per reaction was used in gel mobility shift reactions. To generate nuclear extracts, a nuclear pellet was obtained from stage 13 embryos as described (36). The pellet was resuspended in Buffer C (20 mM HEPES–KOH, 25% glycerol, 420 mM NaCl, 1.5 mM MgCl2, 0.2 mM EDTA, 0.5 mM DTT and protease inhibitors) on ice and centrifuged at 10000 g in a microfuge for 1 min at 4°C to clear contaminating pigment granules: 1–5 μg was used for each gel mobility shift assay. To obtain lysates from embryos expressing BMP-4, embryos were injected at the two-cell stage with 500 pg of RNA and harvested at stage 13. Total protein concentration for all experiments was obtained using a modified Bradford assay (BioRad).
Oligomers were gel-purified prior to use as probes or cold competitors. For probes, the top strand was end-labeled with [γ-32P]ATP using T4 polynucleotide kinase, followed by purification thorough a G-25 Sepharose column. The radiolabeled top-strand was annealed with the complementary bottom strand by heating to 95°C for 5 min in 0.1 M NaCl, and then gradually cooling the reaction to room temperature. Cold competitors were prepared identically, without labeling. For super-shift experiments, reactions included 0.5 μg of anti-Oct-1 antibody (sc–8024, Santa Cruz) or 0.5 μg of control isotype-matched antibody (various, but the one shown is anti-Chicken Gata2, sc-267). Gels were run and analyzed as described (37).
The sequences of the top strand of the oligomers are as follows: −763 to −737, 5′-AGCTTATTAATATGTAAAGTGACGTTTGATA (TE1175); −763 to −737 (mutated), 5′-AGCTTATTAAGCATTAAAGTGACGTTTGATA (TE1191); Octamer consensus, 5′-CGAATATGCAAATAAGGC (TE1189); CCAAT, 5′-ACTGGCTGGCGGAGGCTTGTGATTGGCTGGCCCGG (TE1207).
The gata2 gene is activated at the transcriptional level by BMP signaling, but the mechanism for induction is not known. Previous transgenic reporter experiments demonstrated that 7.3 kb of sequences upstream of the zebrafish gata2 initiation ATG are sufficient to direct appropriate expression of a reporter to the early ventral domain (38). We sought to define the specific DNA regulatory elements that mediate transduction of the BMP signal to the gata2 gene. The gata2 gene structure has been studied in a variety of organisms, including human (39,40), mouse (41), chicken (42) and Xenopus (39,43). Both the mouse and the chicken genes contain two alternative first exons (proximal and distal), and the different transcriptional start sites are used depending on the cell type (41,42). Alternative first exons are not described for either the Xenopus or human genes, but it remains possible that multiple promoters are a common feature of the gata2 locus, and this could complicate any mapping experiments. Therefore, before attempting to map BMP responsive sequences, we first identified the transcriptional start site for the zgata2 gene in order to define the proximal promoter.
Sequence comparison of the zebrafish gata2 putative promoter region with DNA sequences from exon 1 of the gata2 gene from a variety of species failed to identify regions of homology (data not shown). Therefore, to identify the zebrafish gata2 transcriptional start site, 5′ rapid amplification of cDNA ends (RACE) was performed on RNA isolated from embryos at 75% epiboly, the first stage at which gata2 transcripts are readily detected (44). A single PCR product was obtained from this analysis, and subsequent sequencing and comparison with the genomic sequence determined that the transcription start site is located 595 bp upstream of the initiation methionine (Figure 1A). The analysis identifies a short non-coding first exon and places the ATG initiation codon within the second exon (Figure 1A and B). The cDNA isolated by RACE includes sequences just upstream of an independently isolated gata2 cDNA clone (45). No alternative 5′ sequences that would suggest an alternative exon were identified in the expressed sequence tag database. The exon 1 sequence does not show homology with gata2 genomic sequences from other species. The result does not rule out that additional promoters might be used in specific cell types or later stages of development, but indicates that a proximal promoter functions during early embryogenesis, initiating transcription at ~595 bp upstream of the initiation ATG.
Therefore, we considered the ~6.5 kb of sequences upstream of this region as putative regulatory sequences and developed an assay to determine if they are able to mediate a BMP response. Although it should be possible to do this experiment in the homologous zebrafish system, Xenopus embryos provide a unique advantage in that the dorsal/ventral axis is discernable as early as the four-cell stage. Furthermore, the primary components of the early embryonic BMP signaling pathway are thought to be highly conserved. Therefore, the genomic sequence including the ATG was cloned upstream of a firefly luciferase reporter gene (Figure 1C) and this plasmid was injected into developing Xenopus embryos. This approach allows activity of the reporter to be analyzed in vivo during the early stages of embryogenesis when gata2 expression is known to be dependent on BMP signaling. In the Xenopus system, blastomeres committed to dorsal-anterior or ventral-posterior fates can be distinguished already at the four-cell stage, by virtue of pigment differences that arise following cortical rotation. A reporter sensitive to the endogenous BMP program is expected to be relatively more active when injected at the four-cell stage into presumptive ventral-posterior blastomeres compared to dorsal-anterior blastomeres. The reporter was co-injected with control RNA expressing Renilla luciferase. Preliminary experiments demonstrated that the control luciferase is equally active regardless of the injected blastomere (data not shown). The approach of using Renilla luciferase RNA (rather than a Renilla luciferase expression plasmid) avoids any potential concern that BMP-4 might also influence a control promoter. Embryos were harvested at stage 12–13 (during early neurulation) and protein lysates were tested for the relative amount of firefly luciferase, normalized to Renilla luciferase.
We found that the gata2:luciferase reporter is more active when injected into the two ventral-posterior blastomeres of a four-cell embryo compared to activity when injected into the two dorsal-anterior blastomeres (Figure 2A). The difference is ~2-fold, consistent with the gradient of BMP signaling across the dorsal/ventral axis. Importantly, co-injection of RNA encoding Xenopus Smad6 (xSmad6), which block BMP signaling (46–48), reduces the activity of the reporter when injected into ventral blastomeres to a level below what is found when the reporter is injected into dorsal-anterior blastomeres (Figure 2A). Thus, the reporter responds to the embryonic BMP program in a manner consistent with the endogenous pathway. To demonstrate directly that the reporter is activated by the BMP pathway, the gata2 and control reporter constructs were co-injected with RNA encoding BMP-4 into the two dorsal blastomeres. In this assay, the activity of the injected BMP-4 is confirmed by the subsequent ventralized phenotype of injected embryos (data not shown). As shown in Figure 2B, BMP-4 expression results in an average 5-fold induction of reporter activity compared to activity of the reporter when coinjected with control RNA. Co-injection of Smad6 is sufficient to block this activation to the basal level activity of the promoter. We conclude that 6.5 kb of upstream gata2 genomic sequence is sufficient to respond to BMP signaling, and that the developing Xenopus embryo can be used as a model system to define the relevant cis elements.
To define the BREs, a series of progressive 5′ deletion constructs were each injected into developing Xenopus embryos with either control RNA or RNA encoding BMP-4. In each case, the activity of the reporter was measured in luciferase assays following injection into the two presumptive dorsal blastomeres of a four-cell embryo, and normalized to Renilla luciferase activity from co-injected Renilla luciferase RNA. As shown in Figure 3, deletion of sequences from −6520 to −819 has no effect on the levels of induction caused by BMP-4. However, removing a 68 bp region located between −819 and −751 (relative to the start of transcription) results in a significant loss of induction, indicating that this region mediates a response to BMP-4 and might contain one or more BREs.
We showed that the BMP response on the gata2 promoter is sensitive to Smad6 (the inhibitory Smad), so it is expected that BMP-dependent R-Smads should be at least part of the mechanism that mediates the induction between −819 and −751, although this could be direct or indirect. We therefore tested whether injection of RNA encoding BMP regulated Smads 1, 5 or 8 could function through the gata2 BRE to induce reporter activity. The reporter either contained the BRE (−819) or lacked the BRE (−751), as illustrated in the diagram of Figure 3. As shown in Figure 4, injection of mRNA encoding Xenopus Smad1 (xSmad1) or zebrafish Smad5 (zSmad5) was sufficient to activate the promoter containing the 68 bp BRE (−819). Injection of Xenopus Smad8 (xSmad8) failed to activate the same reporter, indicating that not all BMP regulated Smads can mediate this effect and that there are functional distinctions between Smads 1, 5 and 8 [shown also in other studies (34,49)]. Importantly, Smad1 failed to activate a reporter construct deleted of the 68 bp BRE (−751), indicating that this region is required for both BMP-4 signaling and Smad1 activity to induce the gata2 promoter. The experiment does not distinguish whether Smads function directly by binding this promoter, but only that their activity is mediated by the 68 bp sequence.
We first considered whether sequences similar to previously defined SBEs exist within the 68 bp gata2 BRE identified by the deletion analysis. Alignment of the region to the syntenic region of the Fugu rubripes gata2 gene showed that the sequence is 69% conserved (data not shown), but with no obvious clustering, and with no sequences that match perfectly a known SBE. However, we tested several potential candidate sequences by site-specific mutagenesis (Figure 5A). The sequence 5′-TGGAGC is important for mediating BMP-4 activation of the xVent2 promoter by directing the assembly of a Smad1/Smad4/OAZ transcriptional complex (16). The zebrafish gata2 BRE contains a related sequence: 5′-TGGACC (Figure 5A, #2 and 3). Also, Drosophila Mad binds to the GC-rich sequence 5′-GCCGnCGC (17), and a related sequence 5′-GGGCTCGC is present in the zebrafish gata2 promoter (Figure 5A, #5). Both these elements and several other sequences with some similarity to previously identified SBEs (e.g. GGCT, Figure 5A, #1), or A/T-rich sequences (Figure 5A, #6) were altered using site-directed mutagenesis. However, none of these mutations (#1–6) had any effect on the ability of BMP-4 to induce the promoter (Figure 5B).
Gel mobility shift assays using the 68 bp BRE as a probe failed to reveal any specific protein complexes using whole cell or nuclear extracts from Xenopus embryos injected with xSmad1 or mSmad4 (data not shown). Neither GST–Smad1 nor GST–Smad4 fusion proteins were able to bind the 68 bp sequence (data not shown). Although these are negative results, they indicate that Smad proteins might not function directly by binding sequences in this region. Therefore, we considered other potential binding sites, and identified a strong match with the consensus binding site (5′-ATGCAAAT) for the POU homeodomain transcription factor Oct-1 (50–52). The related sequence 5′-ATGTAAAG is truncated precisely at the junction of the 68 bp deletion. Interestingly, Oct-1 binding is maximal if the flanking sequences are A/T rich (51), which is the case in the gata2 sequence. Mutation of this putative Oct-1 binding site in the gata2 promoter nearly eliminated induction of the promoter by BMP-4, equivalent to truncation of the promoter at −751 (Figure 5B, mutation #7). It is important to point out that the mutation does not alter the basal activity of the promoter, but only the activation by BMP. Thus, the Oct-1 site mediates BMP responsiveness in this system.
In gel mobility-shift assays, nuclear extracts derived from stage 13 Xenopus embryos generate a single predominant complex with a probe that is centered on the putative Oct-1 binding site (arrow in Figure 6A). This complex was not detected using probes from the original 68 bp BRE, since the octamer site is truncated on those sequences. This binding activity is competed in the presence of excess unlabeled probe corresponding to the wild-type sequence, but not with an unlabeled oligomer probe corresponding to the mutated sequence that fails to support a BMP response. Furthermore, this complex is fully competed by an oligomer probe containing an Oct-1 consensus binding site (52). Finally, the complex that forms with the wild-type probe derived from the gata2 BRE is super-shifted specifically by addition of a monoclonal antibody raised against the C-terminus of the Xenopus Oct-1 protein (Figure 6B). Incubation with control isotype-matched antibodies does not alter migration of the complex. Thus, Oct-1 binds to the gata2 BRE (designated BRE1), and a mutation that abolishes Oct-1 binding abrogates the ability of BMP-4 to induce the gata2 promoter.
We tested if BMP signaling modifies the interaction of Oct-1 with the BRE1 site. For this purpose, fertilized eggs were injected with BMP-4 RNA or control RNA. Nuclear extracts were subsequently prepared from the embryos at stage 13 and compared for binding activities. Although extracts from BMP-4 injected embryos appeared to have a modest relative decrease in Oct-1 binding activity, this difference is essentially the same difference found for a distinct complex formed with a control oligomer containing a CCAAT binding site (data not shown). Therefore, the amount and activity of normalized Oct-1 complex formation is not obviously altered by BMP-4 signaling.
Since BMP signaling does not appear to directly alter the Oct-1 interaction with the BRE1, we next considered whether Oct-1 cooperates with one or more additional complexes, which themselves might be direct targets of BMP signaling. Note also that the deletion of the BRE1 or mutation of the Oct-1 site does not eliminate entirely the BMP response, typically reducing the effect from 5- to 2-fold, indicating that additional proximal sequences might contribute to activity. We first tested whether the BRE1 was sufficient to function autonomously upstream of a heterologous promoter. Because the Oct-1 binding site spanned the 3′ junction of the 68 bp region, a larger region of the gata2 promoter including the Oct-1 site and flanking sequences (134 bp, from −819 to −685) was tested for the ability to mediate BMP responsiveness when placed as a single copy or in triplicate upstream of the SV40 promoter. These sequences were unable to mediate a BMP response with the SV40 promoter, indicating that the Oct-1 dependent BRE1 is necessary but not sufficient for mediating induction (Figure 7A).
When a larger region including sequences downstream of the BRE was tested (377 bp, from −819 to −443), BMP-4 induction was conferred to the SV40 promoter. This result is consistent with the presence of one or more cooperating cis-elements located within the more proximal sequences. Therefore, progressive 3′ deletions were tested to define the 3′ boundary of this activity. As shown in Figure 7A, sequences between positions −666 and −646 (BRE2) are required for conferring a BMP-4 response onto the SV40 promoter. To confirm that this sequence cooperates with the original BRE1 defined by the 5′ deletion series (the Oct-1 site), the upstream sequences were removed from the context of the BRE2. Deletion of 68 bp from the larger 377 bp region (308 bp, from −751 to −443) removes the Oct-1 site and eliminates induction of the SV40 promoter by BMP-4 (Figure 7B). We conclude that BMP induction requires both the distal Oct-1 site (BRE1) and sequences within or flanking a 20 bp region of a proximal element (BRE2).
Gata2 is a transcription factor expressed in multiple tissues including those of the hematopoietic (31,53–55) and nervous systems (56–58). The gene is likely to be regulated by distinct pathways in diverse tissues, but the early activation in ventral and lateral embryonic mesoderm and ectoderm is fully dependent on the BMP signaling pathway. The gata2 gene is therefore a component of the BMP syn-expression network. Our results show that BMP signals activate the gata2 promoter, dependent on the participation of R-Smad proteins, and requiring a binding site for the POU homeodomain transcription factor Oct-1, which functions in conjunction with a more proximal regulatory region. It is important to point out that the gata2 gene is expressed broadly throughout the BMP syn-expression domain, at the stages of development we are investigating. The Oct-1 binding site is therefore presumably involved in mediating this broad activation of the gata2 promoter, and our experiments do not address whether it continues to be important for the regulatory mechanisms that later restrict expression of the gata2 gene to defined cell lineages, e.g. in the hematopoietic system.
Although originally considered a ubiquitous transcription factor (50), Xenopus Oct-1 is expressed during embryogenesis in a restricted pattern (59) in regions of both ectoderm and mesoderm coincident with gata2 expression. Furthermore, mis-expression of Oct-1 in developing Xenopus embryos causes axis defects (59), suggesting that Oct-1 might be a component of early dorsal–ventral patterning. Axis defects are similarly caused by expression of a dominant-negative Gata2 isoform (32). There is precedence for mediation of the BMP pathway by POU domain proteins, provided first by studies of the Dpp-regulated vestigial gene in Drosophila (60). A sequence from the vestigial locus, known as the vgQ enhancer, is directly regulated by Dpp signaling through binding of Mad to this element (17). However, over-expression of the POU domain protein Drifter (DFR) expands the domain of LacZ expression in transgenic flies that carry a vgQ-LacZ reporter (60). Subsequent promoter deletion analysis identified a DFR binding element in close proximity to the already characterized Mad binding element, consistent with the idea that full Dpp-mediated transcriptional activity requires the cooperation of both DFR and Mad proteins at the vestigial promoter (60). Ectopic expression of Dpp or a constitutively activated form of the Dpp receptor Thick Veins fails to activate DFR expression, implying that DFR itself is not a direct target of the Dpp signaling pathway (60).
Recently, Cao et al. (26) found using a one-hybrid assay that Oct-25 binds to and stimulates the activity of the Xvent-2B promoter, and that Oct-25 can also bind to components of the BMP signaling pathway, including Smad1, Smad4 and Xvent-2. Interestingly, Oct-25 expression is not sufficient for activation of the Xvent-2B promoter, but requires the cis-elements that interact with Smad1/4 and OAZ. This is analogous to the situation we describe with the gata2 promoter, in that Oct-1 binding is not sufficient but requires in addition the more proximal BRE2 sequences, which we propose are direct or indirect targets for Smads and/or other activators. Our preliminary attempts at further defining the key sequences of BRE2 by site-directed mutagenesis have not yielded consistent results, perhaps indicating the presence of overlapping or redundant elements that will require alternative approaches to unravel.
There are two adjacent octamer sequences present in the Xvent-2 promoter, neither of which matches exactly the single site of the gata2 promoter. POU transcription factors bind DNA via their unique POU domain, consisting of two sub-domains: a POU-type homeodomain (POUH) and a POU-specific domain (POUS), tethered to each other by a linker of variable length and sequence (61). In the case of Oct-1, both these domains are indispensable for DNA binding to the octamer consensus sequence 5′-ATGCAAAT-3′ (62). The Oct-1 binding site in the gata2 BRE1, 5′-ATGTAAAG, retains the four (underlined) essential bases, and overall is identical at six of the eight nucleotides. Furthermore, supershift experiments using an antibody specific to Oct-1 suggest that Oct-1 is the protein in Xenopus nuclear extracts that binds to the gata2 BRE1. Mutations that abolish this binding also block BMP-4 mediated induction of the gata2 promoter. However, at this point it seems feasible that Oct-25 could also bind to the gata2 BRE1 (and also that Oct-1 might be able to interact at the Xvent-2 promoter). Until specificity is further documented, it seems reasonable to consider that POU-domain octamer binding proteins represent functional components that mediate BMP signaling.
Based on the studies of the vestigial and Xvent-2 promoters, and our data showing that Smad1 can substitute for BMP4, while Smad6 blocks the BRE response, it seems likely that Oct-1 functions with Smad1/4 and perhaps Vent homeodomain proteins for activation of the gata2 promoter. Our attempts to demonstrate that Smad1/4 or Vent2 bind directly to sequences in the proximal BRE2 element have so far yielded negative results, but this may be due to relatively low binding affinities or the requirement of additional co-factors. Although the cooperating elements in the vestigial locus were located within 25 bp of each other (60), it is conceivable that proteins bound to the gata2 elements could interact over larger distances, presumably facilitated by DNA bending, since the Oct-1 POU domain can bend DNA in vitro (63).
The 20 bp region defined as the BRE2 has sequences that might be capable of binding Smad (5′-AGAC) and/or Vent (5′-TAAT) proteins (Figure 7C). The core sequence 5′-TAAT-3′ has been previously shown to be absolutely required for binding and activity of Xvent-2 (64,65). Xvent-2 is an excellent candidate as a co-modulator of the gata2 response to BMP signaling. First, it is a direct target of BMP-4 (66). Second, zebrafish embryos in which the Vent-related homeobox genes vox and vent are inactivated by mutation have reduced expression of gata2 at mid-gastrula stages (67). Third, Xvent-2 can associate with Smad1 to regulate the transcription of downstream target genes (68). Fourth, Xvent-2 and gata2 co-regulate the expression of Xvent-1, placing both these genes in the same ventralizing pathway (33). Collectively, these data are consistent with the hypothesis that gata2 expression is induced by a combination of Smad1/4, Oct-1 and Xvent-2 (and perhaps other co-factors), followed by cooperation of these transcription factors to activate other proteins important for ventral fate and the specification of ventrally derived tissues.
The full-length zebrafish gata2 promoter was provided by Shuo Lin (UCLA) and the initial set of deletions was generated based on the sequences analyzed previously in his laboratory in transgenic animals. We are also grateful to J. Thomsen (SUNY, Stony Brook), J. Christian (Oregon Health Sciences University), M. Mullins (University of Pennsylvania) and C. Wright (Vanderbilt University) for providing expression constructs. These studies were supported by grants to T.E. from the National Institutes of Health (HL056182) and the Irma T. Hirschl Trust. Funding to pay the Open Access publication charges for this article was provided by the National Institutes of Health.
Conflict of interest statement. None declared.