Bone morphogenetic protein-4-induced activation of Xretpos is mediated by Smads and Olf-1/EBF associated zinc finger (OAZ)

Abstract

We have previously isolated Xretpos, a novel family of long terminal repeat (LTR)-retrotransposons in Xenopus laevis, whose transcript is restricted to ventro-posterior-specific regions and induced by bone morphogenetic protein-4 (BMP-4) signaling. To explore the molecular mechanism of the transcriptional regulation, we identified and characterized Xretpos promoter regions consisting of LTRs and a 5′-untranslated region. We demonstrated that this promoter region contains all the necessary regulatory elements for the spatial and temporal expression of Xretpos. Sequence analysis of the Xretpos promoter revealed multiple Smad-binding elements and Olf-1/EBF-associated zinc finger (OAZ) binding sites similar to BMP-4 response element, which were identified and proved to be required for BMP-4 induction in the Xvent2 promoter. We further demonstrated that Smads and OAZ proteins bind to their response elements in the promoter and these bindings are essential for the BMP-4-induced activation of the Xretpos promoter. Furthermore, we showed that the endogenous expression of Xretpos protein indeed occurred and was temporally regulated and BMP-4-inducible during the early Xenopus development. Finally, overexpression and partial loss-of-function study revealed that Xretpos has a posterio-ventralizing activity. Together, our results place Xretpos downstream of BMP-4 and provide evidence for the conserved mechanism of transcriptional regulation of the BMP-4 target genes.

INTRODUCTION

The transforming growth factor-β (TGF-β) superfamily is divided into two general branches, the bone morphogenetic proteins (BMPs) and TGF-β/activin/Nodal branches (1,2). In Xenopus, whereas TGF-β/activin/Nodal signals have been implicated in meso-endoderm formation and dorsal mesoderm patterning, BMP (BMP2/4/7) has been shown to be a key regulator in the ventral mesoderm patterning, tail formation and ectodermal cell fate specification (2,3). Dorsal–ventral and neural–epidermal cell fates are determined by the gradient of BMP activity that is established by the negative signal from the Spemann’s organizer.

These distinct biological effects of the TGF-β superfamily are exerted by the transcriptional regulation of target genes in a cell-type-specific manner. In Xenopus, the BMP signaling activates homeobox genes, Xvent-1 (4,5), Xvent-2 (6–10) and Msx-1 (11,12) and erythroid transcription factors, GATA-1 (13) and GATA-2 (14). In contrast, Xenopus homeobox gene goosecoid (15,16), Mix.2 (17,18) and Xnr1 (19) are activated by activin/Nodal signals. These TGF-β signals from the cell surface to the nucleus are mediated through two different types of serine/threonine kinase receptors, type I and type II, and intracellular transducer Smad proteins (1–3,20,21). Smad1, Smad5 and Smad8 are involved in BMP-related signals, and Smad2 and Smad3 are involved in TGF-β/activin/Nodal-related signals. In addition, a common partner Smad (Co-Smad), Smad4, is required for the Smad’s complex formation with receptor-regulated Smads (R-Smads) and transcriptional activation.

Our understanding of the basic principles of the molecular mechanism by which Smads activate their target genes has been accomplished mainly through the identification and characterization of TGF-β/activin target genes. The responsive regions of TGF-β/activin target genes contain Smad binding elements (SBEs), consisting of a consensus sequence of CAGAC (or minimally AGAC) (22,23). But Smad binding to SBEs is very weak and lacks selectivity, therefore requiring additional DNA-binding factors for high-affinity, specific recruitment of the Smad complex to a distinct target promoter (20,21,24). The FAST family and Mix family (Mixer and Milk) have been characterized to be such DNA-binding factors, recruiting Smad2–Smad4 complexes to the activin responsive element of the Mix.2 (25–27), mouse goosecoid (28) and Xenopus goosecoid (29), respectively.

Little progress, however, had been made on the BMP responsive genes until recent characterization of Xvent2 promoter regions. The Xvent2 family is a direct target of BMP signaling and mimics the expression patterns and the ventralizing effects of BMP-4 (6–10). Analysis of the Xvent2 promoter identified the BMP response element (BRE), which contains the SBEs for Smad1/Smad4 binding and the 3′ flanking box (3′ box) for Olf-1/EBF associated zinc finger (OAZ) binding (30,31). The Smad1/Smad4/OAZ complex is required for the cooperative binding to the BRE and subsequent activation of the Xvent2 promoter (31). It is therefore suggested that the cooperative interaction of the Smads and a DNA-binding cofactor is a conserved mechanism in both BMP and TGF-β/activin signaling (32).

In a previous study (33), we isolated and added Xretpos to the limited pool of BMP-4 target genes. Xretpos is a novel family of long terminal repeat (LTR)-retrotransposons in Xenopus laevis, whose transcripts are restricted to ventro-posterior-specific regions and induced by BMP-4 signal (33). In this report, we have analyzed the Xretpos promoter region responsible for its spatial and temporal expression. The Xretpos promoter contains multiple SBEs and an OAZ binding site similar to the BRE of the Xvent2 promoter. We further demonstrated that Smads and OAZ proteins bind to their response elements and these bindings are essential for the BMP-4-induced activation of the Xretpos promoter. Furthermore, the presence of endogenous Xretpos protein in the embryos and overexpression and partial loss-of-function study revealed that Xretpos has a posterio-ventralizing activity. These results would place Xretpos downstream of BMP-4 and provide evidence for the conserved mechanism of transcriptional regulation of the BMP-4 target genes.

MATERIALS AND METHODS

Reporter constructs

Full-sized Xretpos promoter, pGL2-Xretpos (LTR-UTR), which contains 792 bp of LTRs and a 5′-untranslated region (5′-UTR), were polymerase chain reaction (PCR)-amplified using Xretpos genomic clone λ-OLT (33) as template and subcloned into the BglII–HindIII site of the pGL2-enhancer vector (Promega). The deletion constructs were created by PCR using sequence-specific primers. Constructs containing linker-substitution mutations, which replace 10 bp of wild-type sequence composed of 6 bp of restriction site and four additional base changes, were generated by PCR-directed technique (34) and are shown in Figures 4A, 5A and 7C. All constructs were verified by sequencing.68

Figure 4.

Bacterially expressed Smad proteins bind directly to SBE I and SBE III in EMSA. (A) The position and nucleotide sequences of wild-type (U3:1–207 and 31 bp SBE III) and mutants (mt SBE I and mt SBE III) used for EMSA are shown. mt SBE I was completely mutated by the linker- substitution method, while two nucleotides were mutated in mt SBE III. The SBE sequence is upper case. (B) Smad4, but not Smad3 or Smad1, binds directly to SBE I. Smad1, Smad3 and Smad4 proteins were expressed and purified as full-length His-fusion proteins in E.coli and incubated with ³²P-labeled U3:1–207 containing SBE I. The concentration of His-Smad fusion proteins was 0.5 or 1 µg. Lane 1, no protein control. In competition experiments, 100-fold excess of unlabeled wild-type or mutant (mt SBE I) duplex was added in the reaction. The positions of free probe and DNA–protein complex are indicated by arrows. (C) Smad1, Smad3 and Smad4 bind directly to SBE III. EMSA was performed in the same procedures as described above except that a 32 bp double-stranded oligonucleotide encompassing SBE III was used as probe and mt SBE III as competitor.

Figure 5.

Bacterially expressed Smad proteins bind directly to Xretpos BRE-like region and the GCAT motif in UTR. (A) The position and nucleotide sequences of wild-type and mutants used for EMSA are shown. SBE and 3′ box are upper case and underlined. In GCAT motif, GCAT sequences are upper case and AT-rich flanking sequences are underlined. (B) Smad1 bound this fragment with higher affinity than Smad4 and Smad3. EMSA was performed with ³²P-labeled 32 bp Xretpos BRE-like region containing SBE II and 3′ box. Lanes 1 and 14, no protein control. In competition experiments, 100-fold excess of unlabeled wild-type or mutant duplex as indicated above the lanes was added in the reaction. (C) Smad1 binds directly to the GCAT motif (and Smad4 binds weakly). EMSA was performed with ³²P-labeled UTR encompassing GCAT sequences and AT-rich flanking sequences. Lane 1, no protein control.

Figure 7.

Smads and OAZ binding to their response element are essential for the BMP-4-induced activation of Xretpos promoter. Xenopus embryos were co-injected into one dorsal blastomere at the four-cell stage with the indicated RNAs and reporter constructs and allowed to develop until stage 13 for luciferase activity measurement. (A) Smad1/Smad4 and OAZ functionally cooperated in BMP-4-induced activation of Xretpos promoter. RNAs for Smad1, Smad4 and OAZ were synthesized in vitro from pCS2-Flag-Smad1, pCS2-Smad4 and pCS2-6Myc-OAZ-ZF(9–19), respectively. RNA concentrations were 500 pg/blastomere. (B) OAZ is required for the BMP-4-induced activation of Xretpos. Dominant-negative OAZ mutant RNA was synthesized in vitro from pCS2-Flag-OAZ-ZF(1–13). RNA concentrations were 500 pg/blastomere for BMP-4, and 1 ng/blastomere for ZF(9–19) and ZF(1–13), respectively. (C) SBE II, 3′ box and GCAT motif contribute to the BMP-4/Smad1/4-induced transcriptional activation of Xretpos. RNA concentrations were the same as described in (A). Linker-substitution mutations were introduced into SBE II, 3′ box and GCAT motif (left); the wild-type SBE II GAGCAGACAT sequences were replaced by mutated cgGaAttCcg sequences (pGL2-Xretpos-LS-SBE II), the wild-type 3′ box GGTGGGGCAG sequences were replaced by GcTctaGagc sequences (pGL2-Xretpos-LS-3′ box), and the wild-type GCAT motif GCATGGCATT sequences were replaced by mutated cggaattccg sequences (pGL2-Xretpos-LS-2XGCAT), respectively.

Figure 6.

OAZ, Smad1 and Smad4 expressed and isolated from the embryos bind to Xretpos-BRE. (A and B) Aliquots of 10 µg of nuclear extracts were prepared from stage 13 embryos injected with RNAs encoding 6Myc-OAZ-ZF(9–19), 6Myc-Smad4 or Flag-Smad1 and incubated with the 32 bp Xretpos BRE-like probe. Addition of 2 µl of anti-Myc (0.2 µg/µl) or anti-Flag antibodies (1 µg/µl) as well as unlabeled probe (100-fold) disrupted the binding complex. Lanes 1 and 7, no protein control; lanes 2 and 3, uninjected control embryo extracts. (C) Immunoblot analysis was performed to visualize the expression and localization of epitope-tagged proteins. Aliquots of 10 µg of nuclear extracts or cytoplasmic extracts from stage 13 embryos injected with epitope-tagged RNAs were separated by 10% SDS–PAGE and blotted with an anti-epitope antibody.

Figure 8.

Detection of endogenous Xretpos protein by western blot analysis. Aliquots of 50 µg of total cell extracts (A) or 20 µg of nuclear or cytoplasmic extracts (B) from the indicated stage embryos were separated by 10% SDS–PAGE, Ponceau stained, electroblotted onto nitrocellulose and probed with anti-Xretpos antibody. The position of Xretpos protein is indicated by an arrow. Note that Xretpos protein expression is temporally regulated (A) and BMP-4-incucible (B), and localized only to cytoplasmic fractions (B). Ponceau staining was performed as a loading control.

Microinjection and luciferase assay

Xenopus embryos were obtained and fertilized as described previously (33) and staged according to Nieuwkoop and Faber (35). Four-cell stage embryos were injected with luciferase reporter constructs (20 pg/blastomere) and collected for luciferase activity measurement. The pRL-TK vector [herpes simplex virus thymidine kinase (HSV-TK) promoter in front of the Renilla luciferase; Promega] and Dual-Luciferase assay system (Promega) were used to normalize the effect of overexpression of RNA on the transcription of the reporter DNA. All assays were carried out in at least three independent experiments.

GFP constructs and whole-mount in situ hybridization

pXretpos-GFP was created by cloning the XhoI–HindIII fragment from the full-sized Xretpos promoter into SalI– HindIII-digested pCS-GFP3 (a gift from Dr Ken Cho). pgsc-GFP3, which contains 1500 bp of Xenopus goosecoid promoter (15), was kindly provided by Dr Ken Cho for the control experiment. Whole-mount in situ hybridization was performed as described previously (33). A probe specific for green fluorescent protein (GFP) was generated by cloning a BamHI–XhoI GFP fragment into the pGEM7 vector (Promega) and in vitro transcription by T7 RNA polymerase.

Preparation and microinjection of RNA

RNAs used for microinjection experiments were in vitro transcribed using mMESSAGE mMACHINE™ (Ambion) with the following templates and linearization: pT7TS-BMP-4 (33) (Xenopus BMP-4; digested with EcoRI, transcribed with T7), pCS2-Flag-Smad1 (Xenopus Smad1; Asp718, SP6), pCS2-Smad4 (human Smad4; NotI, SP6), pCS2-6Myc-Smad4 (human Smad4, NotI, SP6), pCS2-6Myc-OAZ-ZF(9–19) (human OAZ amino acids 381–772; NotI, SP6), pCS2-Flag-OAZ-ZF(1–13) (human OAZ amino acids 1–587; Asp718, SP6), pT7TS-Xretpos(L) (33) (sense RNA, EcoRI, T7; antisense RNA, HindIII, SP6). Flag-tagged or 6Myc-tagged expression constructs for Smads and OAZ were generated by PCR, subcloned into pCS2 vector and verified by sequencing. Smads and OAZ cDNAs used as PCR templates to generate epitope-tagged expression constructs were generously provided by Dr Kristine Henningfeld (30) and Dr Ali Hemmati-Brivanlou (31), respectively.

Preparation of embryonic extracts

For the preparation of total embryonic extracts, Xenopus embryos were homogenized in lysate buffer containing 300 µl of 300 mM NaCl, 10 mM Tris–HCl pH 7.5, 1 mM dithiothreitol, 1 mM EDTA and 1 mM phenylmethylsulfonyl fluoride solution, and cleared by centrifugation for 5 min at 12 000 g at 4°C. Nuclear extracts and cytoplasmic extracts were prepared from the embryos using the sucrose gradient as described previously (36). The protein concentration was measured by the BCA protein assay (Pierce).

Preparation of His-Smad fusion proteins

His₆-tagged expression constructs for Smad1, Smad3 and Smad4 were kindly provided by Dr Kristine Henningfeld at Dr Walter Knöchel’s laboratory (30). His-Smad fusion proteins were expressed in Escherichia coli BL21 (DE3) and partially purified under native conditions using Ni-NTA agarose (Qiagen) according to the manufacturer’s protocol. The protein concentration was measured by the BCA protein assay (Pierce) and Coomassie (Bio-Rad) staining after polyacrylamide gel electrophoresis (PAGE).

Electrophoretic mobility shift assays

Electrophoretic mobility shift assays (EMSAs) were performed in a total volume of 20 µl containing 25 mM Tris–HCl pH 8.0, 50 mM KCl, 5 mM MgCl₂, 1 mM dithiothreitol, 10% glycerol and 1 µg of poly(dI–dC) (Sigma) as described previously (30). Each binding reaction contained ∼2 × 10⁴ c.p.m. of ³²P-labeled probe and 10 µg of nuclear extracts or 0.5–1 µg of His-Smad fusion proteins. The reactions were allowed to proceed for 30 min on ice and analyzed on a 5% native polyacrylamide gel containing 0.5× Tris–borate– EDTA (run at 150 V at room temperature). For supershift experiments, 2 µl of M2 anti-Flag monoclonal antibody (Sigma) or anti-Myc polyclonal antibody (Santa Cruz Biotechnology) was added to a reaction mixture 30 min before the addition of ³²P-labeled probes. The gel was visualized by a fluorescent image analyzer FLA-2000 (Fuji Film).

Anti-Xretpos antisera and western blot analysis

For production of polyclonal antibodies recognizing Xretpos protein, C-terminal truncated Xretpos fragment was expressed in E.coli BL21 (DE3) from pRSET-Xretpos-(1–272), a plasmid with Xretpos cDNA encoding amino acids 1–272 inserted into EcoRI-digested pRSET (Invitrogen). The His₆-Xretpos fusion proteins were electrophoretically purified and used as antigen for immunization of rabbits.

For western blot analysis, embryonic extracts were separated by 10% SDS–PAGE, electrophoretically transferred to nitrocellulose, Ponceau (Sigma) stained and probed with antibodies. The anti-Xretpos antisera were used at 1:1000 dilution to detect endogenous Xretpos, and M2 anti-Flag monoclonal antibody (Sigma) and anti-Myc polyclonal antibody (Santa Cruz Biotechnology) were used to visualize the Flag- or Myc-tagged proteins in the embryo extracts, respectively. Detection was performed using an enhanced chemiluminescence kit (Amersham Pharmacia) according to the manufacturer’s protocol.

RESULTS

Promoter activity of Xretpos resides in LTR and 5′-UTR regions

Xretpos is a novel family of LTR-retrotransposons in Xenopus laevis and its transcripts are induced by BMP-4 signaling (33). As a first step to understand the molecular mechanism of transcriptional regulation of Xretpos, we identified and characterized the Xretpos promoter. Since promoter activities of most retrovirus and retrotransposons were shown to be present in the LTRs and 5′-UTR, we generated the full-sized Xretpos promoter, pGL2-Xretpos (LTR-UTR) (Fig. 1A, top) and measured the promoter activity of this construct (or the promoterless vector pGL2-enhancer) during the early Xenopus development. As shown in Figure 1A, the luciferase activity was detected after midblastula transition, increased during neurulation and declined afterwards, consistent with the result of a developmental northern blot analysis (33). No significant luciferase activity was observed in pGL2- enhancer-injected embryos.

Figure 1.

Stage- and tissue-specific expression of reporter constructs containing the LTR and 5′-UTR of Xretpos promoter. The LTR is divided into three regions, U3, R and U5. (A) Activities of luciferase reporter constructs were measured from the extracts of embryos injected with pGL2-Xretpos (LTR-UTR) or pGL2-enhancer into all four blastomeres of the four-cell stage embryos. One representative experiment is shown for this figure. (B) Xretpos reporter gene expression (left) was compared with endogenous Xretpos RNA (center). pgsc-GFP expression is shown as control (right). pXretpos-GFP was injected equatorially into all four blastomeres of four-cell stage embryos, and the expression of GFP transcripts was visualized by whole-mount in situ hybridization. The dorsal blastopore lip is indicated by arrowheads.

Next, the ability of the promoter sequences to drive the regional specific expression was examined using pXretpos-GFP, in which LTRs and 5′-UTR were inserted upstream of GFP (Fig. 1B, top). Four-cell stage embryos were injected into all the blastomeres with a total of 100 pg of reporter DNA, and GFP expression was analyzed and compared with the endogenous Xretpos expression by whole-mount in situ hybridization. The expression pattern of pXretpos-GFP was localized to the marginal zone at the gastrula stage (Fig. 1B, reporter-Xretpos), enriched in the dorsal-posterior region around blastopore, but was absent from the future head structure (data not shown). This expression pattern recapitulated that of the endogenous Xretpos (33) (Fig. 1B, Xretpos RNA), although expression was weaker and mosaic due to uneven segregation of the injected reporter plasmids during cleavage. This expression pattern was specific to Xretpos promoter sequences, because the goosecoid promoter directed organizer-specific mosaic expression (Fig. 1B, reporter-gsc). These results indicate that the cis-elements in the LTRs and 5′-UTR are sufficient for the stage- and tissue-specific expression of Xretpos.

Xretpos promoter is composed of multiple regulatory elements

To delineate the regions in the Xretpos promoter that are important for the transcriptional regulation, we constructed various deletion mutants and compared their luciferase activities (Fig. 2). The promoter ‘strength’of the full-sized Xretpos promoter was ∼16% of the SV40 promoter in embryos [compare pGL2-Xretpos (LTR-UTR) and pGL2-SV40]. The deletion analysis showed that the U3 region, in which a putative TATA box and CAAT motif are located (33), provides basal promoter activity since promoter activity was totally abolished by removing the U3 region [compare pGL2-Xretpos (LTR-UTR) and pGL2-Xretpos (R-U5-UTR)]. We also assume that 110 bp of the 5′ R region and UTR contains a positive regulatory element since promoter activities were considerably (∼5- and 2-fold, respectively) reduced by removing these regions [compare pGL2-Xretpos (U3-R111) and pGL2-Xretpos (U3), and pGL2-Xretpos (LTR-UTR) and pGL2-Xretpos (LTR)]. In addition to the positive regulatory elements, a negative regulatory element was defined in the 3′ R region and U5 region since elimination of these regions was found to permit ∼11-fold greater basal luciferase activity [compare pGL2-Xretpos (LTR-UTR) and pGL2-Xretpos (U3-R111)]. This may be due to a repressive effect of the 3′ R region and U5 region on transcription or translation. A similar repressive effect of sequences in the U5 region was described for other retroelements such as bovine immunodefienciency virus (37), human foamy virus (38) and HERV-K (39). In summary, the Xretpos promoter is composed of positive and negative regulatory elements, suggesting that multiple signaling may be involved in Xretpos regulation.

Figure 2.

The Xretpos promoter is composed of positive and negative regulatory elements. Schematic representations of reporter constructs and their transcriptional activities are shown. Various deletion constructs of the Xretpos promoter were generated, injected into the two non-adjacent blastomeres of the four-cell stage (20 pg/blastomere) and allowed to develop to stage 11 for luciferase activity measurement. Fold induction was calculated as the ratio between pGL2-Xretpos (LTR-UTR) and other reporter constructs and represents the mean values from at least three independent experiments using different preparations of the plasmids. Positions of positive and negative regulatory elements in the Xretpos promoter are indicated at the top.

Identification of the SBEs in the Xretpos promoter

A careful analysis of Xretpos promoter sequences predicted a number of putative binding sites for transcription factors. These sites included three copies of consensus SBEs (CAGAC sequences), two (SBE I and SBE II) in U3 region and one (SBE III) in U5 region (Fig. 3A). CAGAC (or minimally AGAC) sequences are known to be optimal SBEs (22,23) and are found in the responsive regions of several TGF-β, activin or BMP target genes (Fig. 3B). After sequence comparison between Xretpos and Xvents promoters, we further observed that SBE II constitutes a region (BRE-like) of similarity to the Xvents BRE which is composed of an SBE and 3′ box (Fig. 3A and C) (31).

Figure 3.

SBEs and the BRE-like region in the Xretpos promoter. (A) The Xretpos promoter contains three copies of consensus SBEs (SBE I–III) and Xvent BRE-like region consisting of SBE II and the 3′ box. (B) Xretpos SBE CAGAC sequences are compared with other SBEs from the responsive regions of several TGF-β, activin or BMP target genes. (C) The Xretpos BRE-like region is compared with Xvents BRE. Consensus sequences in SBE and the 3′ motif are in bold type and shaded.

Smad proteins bind directly to SBEs

The presence of SBEs in the Xretpos promoter suggests the possible binding of Smads to Xretpos SBEs. We performed EMSAs with a double-stranded DNA (dsDNA) probe encompassing SBEs and bacterially expressed full-length Smad1, Smad3 and Smad4. As for SBE I, a 207 bp dsDNA probe (U3:1–207) containing SBE I was used as a substrate for EMSA (Fig. 4A). As shown in Figure 4B, Smad4 was capable of binding to this region in a concentration-dependent manner (lanes 1–3). However, Smad1 and Smad3 had no detectable binding (lanes 6–12). The addition of excess (100×), unlabeled wild-type probe reduced Smad4 binding (lane 4), indicating the specificity of binding. In contrast, mt SBE I, in which consensus CAGAC SBE was entirely mutated by the linker-substitution method did not effectively compete for Smad4 binding, suggesting that SBE I is required for Smad4 binding (compare lanes 4 and 5).

For the in vitro binding study with SBE III, we applied a 31 bp oligonucleotide duplex that encompasses SBE III (31 bp SBE III; Fig. 4A). We observed concentration-dependent DNA–protein complex formation with Smad1 (Fig. 4C, lanes 6 and 7), but Smad4 and Smad3 resulted in low-level binding activity (lanes 1–3 and 10–12). The binding was SBE III sequence-dependent since mt SBE III inhibited the binding less effectively than the wild-type competitor in competition experiments (compare lanes 8 and 9).

To determine the binding profile of the Xretpos BRE-like region, we constructed a 33 bp oligonucleotide duplex which encodes SBE II and the 3′ box (Fig. 5A). Figure 5B shows that Smad1 bound this fragment much stronger than Smad4 and Smad3. In competition EMSA, point mutations introduced into SBE II, 3′ box or both elements resulted in less effective competition than with the unlabeled wild-type (Fig. 5B, compare lanes 4–7), suggesting that these two elements are necessary for optimal binding.

Smad1 binding is mediated by a GCAT motif

Further EMSA with Smads identified the other Smad1 binding region in the UTR of the Xretpos promoter (Fig. 5C). Smad1 bound to this region in a concentration-dependent (Fig. 5C, compare lanes 1–3) and sequence-specific manner (compare lane 3 and 4) but Smad4 binding to this region was very weakly detectable (lanes 6–9). The identified UTR contains two tandem copies of GCAT sequences flanked by AT-rich sequences (Fig. 5A). It is of note that the same GCAT sequence and flanking AT-rich regions were identified and characterized as critical for Smad1 binding in the Xvent-2B promoter (30). We performed competition EMSA with mt GCAT in which two tandem GCAT motifs were linker-substitution mutated and found that competition by mt GCAT was not effective as compared with the wild-type competitor (Fig. 5C, compare lanes 4 and 5). These results suggest that Smad1 binding to the Xretpos UTR is dependent on GCAT sequence.

In the above competition EMSA, it is of note that the Smad binding was not only in competition with wild-type probe but also with mutant probes (Fig. 4C, lanes 8 and 9; Fig. 5B, lanes 4–7; Fig. 5C, lanes 4 and 5). This suggests that other core sequences or flanking sequences outside the core SBE that were not mutated within the probes are also important for binding. Nevertheless, it is obvious that mutant probes were less effective competitors than wild-type probes in all competition EMSAs and this observation was further supported by the independent EMSA in which Smad binding to mutant probes [mt SBE II, mt 3′ box and mt (SBE II-3′ box)] was clearly shown to be less efficient than to the wild-type probe (BRE-like) (data not shown).

OAZ, Smad1 and Smad4 isolated from the embryos bind to the BRE-like region

The Xretpos promoter contains an Xvent-2 BRE-like region consisting of a CAGAC SBE and 3′ box, which work as binding sites for Smads and OAZ, respectively (31). To examine whether OAZ can indeed bind to the Xretpos BRE-like region, an EMSA was performed with 33 bp BRE-like oligonucleotide duplex and nuclear extracts prepared from stage 13 embryos injected with 6Myc-OAZ-ZF(9–19) RNA. Nuclear extracts were used as a protein substrate for EMSA because most OAZ proteins were localized in nuclear fractions as shown in Figure 6C. Zinc finger 9–19 of human OAZ [OAZ-ZF(9–19)] was used because it contains sufficient functional domains for BRE binding and Smad1 interaction and mimics the full-length human OAZ (31). As shown in Figure 6A, we could identify the induction of a DNA–protein complex. This complex was also present at a low level in the reaction from the uninjected control extracts (Fig. 6A, compare lanes 2 and 4) and was competed by unlabeled probe in both reactions (lanes 3 and 6). The specificity of OAZ in the binding was confirmed by the disappearance of the complex by incubation with anti-Myc antibody (Fig. 6A, compare lanes 4 and 5).

We next tested whether Smad1 and Smad4, which were expressed and isolated from the embryo, also interact with the Xretpos BRE-like region, by EMSA using nuclear extracts from stage 13 embryos in which Flag-Smad1 or 6Myc-Smad4 were overexpressed. As with bacterially expressed Smads, a specific DNA–protein complex was formed (Fig. 6B). The incubation with a respective anti-epitope antibody also disrupted the complex (compare lanes 8 and 9, and lanes 11 and 12), indicating the specificity of Smads in the complex. The nuclear localization of epitope-tagged OAZ and Smads was visualized by western blot analysis (Fig. 6C). The presence of a single DNA–protein complex which is specific against competitor and antibody incubation suggests that the OAZ, Smad1 and Smad4 are components of the Xretpos BRE-binding complex.

Smads and OAZ binding to their response element are essential for the BMP-4-induced activation of Xretpos promoter

Having demonstrated that Smads and OAZ were capable of binding to the specific binding sites in the Xretpos promoter, we next examined the functional role of Smads and OAZ in the Xretpos regulation by BMP-4. As shown in Figure 7A, Smad1/Smad4 and OAZ functionally cooperated in BMP-4-induced activation of the Xretpos promoter. To further confirm the requirement of OAZ in BMP-4-induced activation of Xretpos, we used a dominant-negative OAZ mutant [OAZ-ZF(1–13)]. This mutant contains zinc finger domains responsible for BRE binding but lacks a Smad1 binding domain, therefore working in a dominant negative manner (31). Overexpression of the dominant-negative OAZ mutant reduced the OAZ-ZF(9–19)-induced Xretpos promoter activation (Fig. 7B).

To examine the importance of SBE II, 3′ box and GCAT motif in BMP-4-induced activation of Xretpos, we introduced linker-substitution mutations into these sites and analyzed the functional consequences. In contrast to the potent activation of the wild-type reporter construct by BMP-4/Smad1/Smad4, luciferase activities were considerably reduced in all three mutant constructs (Fig. 7C), suggesting that all these sites are important. Together, these results suggest that the binding of Smads and OAZ to their response elements is essential for the BMP-4-induced activation of Xretpos.

Xretpos protein expression was temporally regulated and BMP-4 inducible

Together with the ventro-posterior expression and BMP-4-induction of Xretpos in our previous work (33), the above results from the Xretpos promoter study indicate that Xretpos is a downstream target gene in BMP-4 signaling and suggest a potential role in embryonic development. Since Xretpos is a family of LTR-retrotransposon-like elements, we first tested whether Xretpos protein is indeed synthesized in the embryo. Anti-Xretpos polyclonal antibody was obtained by using bacterially expressed His₆-Xretpos (amino acids 1–272) fusion proteins as an antigen and applied for western blot analysis. As shown in Figure 8A, a single band of the predicted size for the endogenous Xretpos, as well as an antigen (lane 1), was detected in embryo extracts, and antibody staining was temporally regulated during the early development. Moreover, Xretpos protein was activated by BMP-4 and restricted only to the cytoplasmic fraction (Fig. 8B). This temporal regulation and BMP-4 induction of Xretpos protein clearly correlates with the expression pattern of Xretpos RNA (33).

Xretpos has a posterio-ventralizing activity

Next we investigated the potential embryonic activity of Xretpos in early development by microinjection experiment. We injected 4 ng of Xretpos RNA into dorsal animal or marginal regions at the four-cell stage embryos and assessed the resulting phenotype. As shown in Figure 9A, we obtained posterio-ventralized embryos with reduced anterior head structures and no eyes [27%, dorso-anterior index (DAI) = 0–2, n = 88]. Control injection of preprolactin (PPL) RNA with the same concentration had no effect on development.

Figure 9.

Phenotypic effects of overexpression (A) and partial loss (B and C) of Xretpos function. (A) Overexpression of Xretpos RNA resulted in reduced anterior head structures. Four-cell stage embryos were injected into two dorsal animal or marginal regions at the four-cell stage embryos with 4 ng of Xretpos RNA and allowed to develop until stage 43. Control embryos injected with 4 ng of PPL RNA produced normal tadpoles. (B) Rescue of axial structures in UV-irradiated embryos by antisense Xretpos RNA. Partial twinned axis structures were seen in UV-irradiated embryos injected into two non-adjacent blastomeres at the four-cell stage with antisense Xretpos RNA (1 ng/blastomere), but not in uninjected UV-irradiated controls. (C) Phenotypic effects of antisense Xretpos RNA injection in wild-type embryos. Antero-dorsalized structures were observed in embryos which were injected diagonally at the four-cell stage with antisense Xretpos RNA (1 ng/blastomere). Uninjected control embryos developed normally. (D) Reduction of endogenous Xretpos protein by antisense Xretpos RNA injection. The position of Xretpos protein is indicated by an arrow and Ponceau staining was performed as a loading control.

Posterio-ventralizing activity of Xretpos was further assessed by the partial loss-of-function experiments by antisense Xretpos RNA injection. First, we tested whether antisense Xretpos RNA was able to rescue axis structures in UV-ventralized embryos. Xenopus zygotes were UV-irradiated and two non-adjacent blastomeres were injected with antisense Xretpos RNA (1 ng/blastomere) at the four-cell stage. As shown in Figure 9B, these embryos produced partial twinned axes with high frequency (65%, n = 23), whereas uninjected, UV-treated control embryos were strongly ventralized (average DAI = 0.9, n = 38). The induction of twinned axes in the embryos injected into two opposing sites at the four-cell stage may exclude the possibility that these embryos escaped the UV irradiation. We next analyzed the phenotypic effects of loss of Xretpos function in the wild-type Xenopus embryo. When the wild-type four-cell stage embryos were diagonally injected with antisense Xretpos RNA (1 ng/blastomere), antero-dorsalized embryos were produced (32%; DAI = 6–8, n = 38), whereas uninjected control embryos developed normally (Fig. 9C). These antero-dorsalized embryos resembled LiCl-treated embryos with a DAI of 7–8 (44) or embryos in which BMP-4 function was partially blocked by antisense BMP-4 RNA injections (45). We found that antisense Xretpos RNA injection indeed reduced the amount of Xretpos protein to ∼60% of wild-type levels in embryo extracts in western blot analysis (Fig. 9D). Together, these results strongly demonstrate an antero-dorsalizing activity of antisense Xretpos RNA, and in turn support the evidence that Xretpos indeed has a posterio-ventralizing activity.

DISCUSSION

The present study describes Xretpos as a novel BMP-4 downstream gene with posterio-ventralizing activity, whose activation by BMP-4 is mediated by the association between Smads/OAZ and their consensus binding sites in the Xretpos promoter. Xretpos was previously identified as a novel LTR-retrotransposon-like element in X.laevis and it was shown that its transcripts are restricted to ventro-posterior-specific regions and induced by BMP-4 overexpression in animal cap and whole embryos (33). In this paper, we studied Xretpos promoter regions consisting of LTRs and 5′-UTR in order to understand the mechanism of transcriptional activation of Xretpos by BMP-4. We showed, through luciferase assay and whole-mount in situ hybridization detection of reporter constructs, that this Xretpos promoter was sufficient to provide the stage- and tissue-specific expression of Xretpos during early Xenopus development. The Xretpos promoter contained three copies of consensus SBEs (SBE I–III) and a BRE-like region that is composed of SBE II and 3′ box and similar to the Xvents BRE. Additionally, we demonstrated that Smads and OAZ proteins actually bind to these sites and cooperatively activate the transcription. Mutations in the consensus sites of the SBE and 3′ box motif abolished the BMP-4-induced activation of Xretpos promoter. These findings support the idea that the assembly of OAZ and BMP-activated Smads to promoters containing the Smad/OAZ consensus BRE would be the conserved mechanisms for BMP target gene activation.

Independent study on Xvent2 promoter by Henningfeld et al. (30) identified a novel Smad1 binding site consisting of GCAT and flanking AT-rich sequences and showed that this site is required for full transcriptional activation. The similar GCAT motif containing two tandem copies of GCAT sequence and flanking AT-rich sequences were found in the Xretpos UTR. Through competition EMSA and mutational luciferase assay, this motif was shown to be required for Smad1 binding and promoter activation. The importance of flanking AT-rich sequences in this motif in Smad1 binding was shown previously (30) and further confirmed when a probe containing only the GCAT sequence without flanking AT-rich sequences present in the U3 region could not be bound by Smad1 (data not shown). Therefore, the GCAT motif may be the conserved cis-element responsible for Smad1 binding and activity.

The Xretpos promoter contains three putative CAGAC SBEs, which showed differential affinity and specificity for Smads. Through EMSA with bacterially purified Smads, it was shown that Smad1 strongly bound to SBE II and SBE III but did not bind to SBE I, whereas Smad4 bound strongly to SBE I but weakly to SBE II and SBE III. Smad3 had no binding affinity to SBE I and bound only weakly to SBE II and SBE III. Since the core CAGAC sequences in three SBEs are identical, it seems that the surrounding sequences outside the core SBE may be additional determinants for the specificity of Smad binding. This interpretation may be further supported by the finding that, in addition to the core SBE II sequence, the 3′ box in the Xretpos BRE-like region was important for Smad1 binding as revealed by competition EMSA experiments (Fig. 5B, lanes 4–7). Moreover, according to the previous DNase I protection experiment (30), the region protected by Smad4 and Smad1 was ∼20 bp long, extending the core motifs in the Xvent2 promoter.

There have been some controversies as to the exact DNA sequences that are recognized by Smads. Since first identified as optimal DNA-binding sequences for Smad3 and Smad4 by PCR-based selection from a pool of degenerate oligonucleotides (22), one or two copies of SBEs began to be identified in TGF-β-responsive genes (Fig. 3B). Whereas these SBEs were bound and transactivated by Smad3 and Smad4 (40–42), they were neither bound by Smad1 nor responded to BMP signaling (41). Additionally, SBEs in the Xvent2 BRE were directly bound and protected by Smad4 but not by Smad1 (30). There fore, it was suggested that the CAGAC sequence represents the binding site for Smad3 and Smad4 but not BMP-regulated Smads. However, mounting evidence suggests the alternative hypothesis that BMP-regulated Smads (Smad1/5/8) recognize the same CAGAC sequence, as do Smad3 and Smad4. First, the crystal structure analysis revealed that both the core DNA-contacting residues and the surrounding region of DNA-binding β hairpin are highly conserved invariantly among Smads (23). Secondly, it was shown that the MH1 domain of the Smad1 binds specifically to SBE with a similar affinity to Smad4 (23). Thirdly, multimerized SBE sequences from the JunB promoter were responsive to BMP as well as activin (40), and Smad1 was able to activate artificial tandem copies of SBE synergistically in combination with Smad4 (46). Additionally, Smad5, a BMP-regulated Smad, was shown to be capable of binding AGAC-containing oligonucleotides and 8 bp palindromic SBE (GTCTAGAC) in the Smad7 promoter (47). Therefore, our present findings that SBE CAGAC sequences in the Xretpos promoter are recognized not only by Smad3 and Smad4 but also by Smad1 (Figs 4–6) provide additional evidence that the same SBE CAGAC sequence can be recognized both by BMP-regulated Smads and Smad3 and Smad4, but it may depend on the flanking sequences.

Xretpos is not the only retroviral element that is regulated by TGF-β family members. Human endogenous retrovirus HERV-K (48) was shown to be induced by BMP and retinoic acid, and Xenopus LTR-retrotransposon-like element BIG3/1A11 (49,50) was shown to be induced by activin, FGF and the T-box gene Xbra. Moreover, multiple SBE CAGAC sequences were found in their respective LTR sequences (the presence of SBEs in BIG3/1A11 LTRs were unpublished observations by S.Shim), although the potential Smad binding activity was not verified. Zebrafish LTR-retroelement bhikhari (bic) was also shown to be induced by activin and the induction was mediated by FAST protein binding to an activin response element in bic LTR (42). Addition ally, it is interesting to find that BORG, an immediate early response gene of BMPs without any extensive ORFs, contained, in its middle part, a cluster of multiple interspersed repetitive sequences including one region related to mammalian short interspersed element and three regions related to mammalian apparent LTR-retrotransposons (51). The evolutionary and functional implications of this apparently general mechanism of the transcriptional regulation of the retroviral genome by the TGF-β superfamily and other growth factors remains to be determined.

The Smad binding and transcription assays in this study suggest that Xretpos may behave like an immediate early gene via BMP-4. According to our previous study, however, BMP-4 induction of Xretpos was greatly reduced by the translational inhibitor cycloheximide treatment before midblastula transition (33). It was also shown that BMP-4 activation of Xretpos transcription was detected at neurula stage but not gastrula stage from northern blot analysis. Therefore, it seems likely that de novo protein synthesis of newly activated transcripts after late gastrulation or neurulation is required for the full activation of Xretpos by BMP-4. OAZ seems unlikely to be such a candidate because, although OAZ transcripts increase from the neurula stage, they are also present before the midblastula transition (31).

Our previous studies showed that Xretpos has only limited homology to the gag gene and no homology to the pol gene such as protease, reverse transcriptase and integrase that are required for its retrotransposition, suggesting that it is a defective transposable element (33). However, our previous findings of the presence of multiple copies of Xretpos-related elements in the Xenopus genome implied that the Xretpos element might once have been active in the early days of its evolution (33). Therefore, Xretpos probably constitutes a rare case of a cellular gene that lost its capability of retrotransposition with the loss of pol gene but instead acquired a novel gene function through transduction by an LTR element. To our knowledge, Xretpos is the first example of a protein that was proven to be expressed in vivo from an LTR-retrotransposon-like element, as revealed by the western blot analysis with anti-Xretpos antibody. Only ORF1 of L1Hs (52), a non-LTR retrotransposon in human, has so far been detected as an endogenous product in human cells, and coding potential was only presumed for other LTR- or non-LTR retrotransposons such as Bs1 in maize (53) and ORF2 of L1Hs (54). More importantly, Xretpos may be the first retrotransposon element whose biological function was manifested in the embryo besides its evolutionary and gene regulatory effects (55–57).

The most interesting and prospective structural features by which Xretpos may exert its biological effects are a ‘CCHC’ zinc finger motif and a coiled-coil leucine zipper structure (33). The ‘CCHC’ motif is found in the nucleocapsid proteins, products of the retroviral gag gene (58) and is also included in a variety of eukaryotic proteins postulated to regulate transcription, RNA splicing or translation such as SLU7 in yeast (59), Xpo (60) and X-cat2 (61) in Xenopus, glh (62) and lin-28 (63) in Caenorhabditis elegans, and cellular nucleic acid binding proteins in Xenopus (64) or mammals (65,66). Despite the high level of conservation of its structure, a direct role for the ‘CCHC’ motif still needs further investigation.

Xretpos protein also encodes a leucine zipper structure at its C-terminus (amino acids 328–370), which is embedded in a region of high amphipathic α-helical structure and can mediate specific dimeric complex formation between homologous or heterologous proteins (67). This observation was obtained from both the coiled-coil algorithm prediction [Paircoils (68); COIL (69)] and PSI-BLAST homology search program (70; data not shown). The additional findings that Xretpos proteins are cytoplasmic and lack a basic domain required for DNA binding suggest that it may play an Id-like role in negative regulation of transcription factors; Id family proteins that have a helix–loop–helix (HLH) domain but lack a basic domain act in a dominant-negative manner by sequestering basic helix–loop–helix (bHLH) proteins (71). In this respect it is interesting to find that Id expression is activated by BMP through an SBE-containing BMP-responsive region in the promoter (72,73) and Id proteins contribute to the anti-myogenic and anti-neurogenic effect of BMP by negatively regulating myogenic and proneural bHLH factors such as MyoD and neurogenin (73–75). Additionally, leucine zipper-containing JunB is activated by TGF-β and BMP putatively through SBE in the promoter (40,76) and involved in the BMP-2-induced inhibition of myogenic differentiation. Direct physical interaction of the leucine zipper domain of Jun with the HLH region of MyoD was implicated in repressing bHLH function (77). Therefore, our putative hypothesis is that the posterio-ventralizing effect of Xretpos protein may be accomplished by negatively regulating basic leucine zipper or bHLH proteins that are involved in anterior–posterior or dorso–ventral specification in which myogenesis and neurogenesis are its essential components. Identification of the proteins with which Xretpos interacts will allow us to verify our hypothesis and clarify the mechanism of how Xretpos acts in the axis specification.