Reversal of Xenopus Oct25 Function by Disruption of the POU Domain Structure*

Xenopus Oct25 is a POU family subclass V (POU-V) transcription factor with a distinct domain structure. To investigate the contribution of different domains to the function of Oct25, we have performed gain of function analyses. Deletions of the N- or C-terminal regions and of the Hox domain (except its nuclear localization signal) result in mutants being indistinguishable from the wild type protein in the suppression of genes promoting germ layer formation. Deletion of the complete POU domain generates a mutant that has no effect on embryogenesis. However, disruption of the α-helical structures in the POU domain, even by a single amino acid mutation, causes reversal of protein function. Overexpression of such mutants leads to dorsalization of embryos and formation of secondary axial structures. The underlying mechanism is an enhanced transcription of genes coding for antagonists of the ligands for ventralizing bone morphogenetic protein and Wnt pathways. Corresponding deletion mutants of Xenopus Oct60, Oct91, or mouse Oct4 also exhibit such a dominant-negative effect. Therefore, our results reveal that the integrity of the POU domain is crucial for the function of POU-V transcription factors in the regulation of genes that promote germ layer formation.

During Xenopus embryogenesis, formation of germ layers and body plan is primarily induced by two maternal factors, VegT and ␤-catenin. VegT induces the nodal related genes (Xnrs) that encode ligands for the nodal signaling pathway, which is the major signal involved in mesoderm and endoderm induction (1)(2)(3). ␤-Catenin, in collaboration with VegT, establishes the dorsalizing signals in the Spemann organizer, and the organizer harbors quite a few secreted proteins, such as Chordin, Noggin, Cerberus, and Dkk1, that antagonize the BMP 2 and the Wnt signaling pathways (4 -9). To guarantee the correct formation of germ layers and body plan, these signals are finely tuned so that they can function at the right level and time and in the right locations. In Xenopus laevis, three Oct4 homologous factors, Oct60, Oct25, and Oct91, are expressed during oogenesis and early embryogenesis (10 -12). We have found that Xenopus Oct factors play an important role in the regulation of the activities of VegT, ␤-catenin, and nodal and BMP4 signaling pathways and prevent premature and incorrect differentiation of embryonic cells to ensure correct formation of germ layers and of body axes (13)(14)(15)(16).
Oct4 is a central player in embryonic stem (ES) cells. On the one hand, it maintains the self-renewal and pluripotency of ES cells (17)(18)(19), and on the other hand, it has the capability to introduce pluripotency into somatic cells (20 -25). However, the molecular mechanisms underlying the functions of Oct4 have not been clearly understood. Because of the functional homology between mammalian Oct4 and Xenopus Oct proteins (14,26), the analysis of these proteins might provide important insights into the molecular mechanisms by which Oct4 performs its functions.
Oct4 and its relatives are members of the POU family transcription factors of subclass V (POU-V). This protein family is characterized by a unique POU-specific domain (POU) located at the N-terminal region and a POU homeobox (Hox) domain at the C-terminal region. These two domains are joined by a variable linker region. Subclasses of this family are divided by the features of the POU and the linker sequences (27). In the classical point of view, POU factors regulate transcription of target genes via interaction between the two conserved domains, POU and Hox, and the octamer motif, ATGCAAAT (18), or certain variants (28). To achieve a higher specificity, Oct4 may form protein complexes with other transcriptional regulators. One well known example is the Oct4-Sox2 complex on the Oct4 and fgf4 promoters in ES cells (29,30). In addition to the POU and Hox domains that are responsible for DNA binding, both the N-and C-terminal regions contain gene transactivation domains (31)(32)(33). Therefore, each region seems to play its part in the function of the Oct4 protein.
In this study, we have investigated each region of the Oct4related protein Oct25 for its relevance in Xenopus embryogenesis. A series of deletion or point mutations were analyzed for their effects on embryonic development and gene transcription. Interestingly, disturbance of the POU domain structure but not of the Hox domain created a dominant-negative effect.
Overexpression of corresponding mutants in Xenopus embryos led to a strongly dorsalized phenotype. Accordingly, the genes that promote mesoderm and endoderm germ layer differentiation and specify the dorso-ventral body axis, like Xnrs, Siamois, Chordin, Goosecoid, Dkk1, and cerberus, are strongly up-regu-lated. The dorsalizing effect is dependent upon functional VegT and ␤-catenin/TCF signaling and overrides the ventralizing activity of BMP signaling. Although DNA binding of corresponding Oct25 mutants to an Oct target sequence is abolished, they still exhibit protein/protein interactions with signal transducers and transcriptional regulators. We also demonstrate that a similar reversal of function as described for Oct25 can be obtained with the homologous proteins Oct60, Oct91, and mouse Oct4 by disturbance of the POU domain at corresponding positions. We suggest that differential binding of co-repressors and/or co-activators is responsible for the observed functional reversal.

EXPERIMENTAL PROCEDURES
Embryos and Explants-Embryos were obtained with in vitro fertilization and cultured in 0.1ϫ MBSH (1ϫ MBSH: 88 mM NaCl, 2.4 mM NaHCO 3 , 1 mM KCl, 0.82 mM MgSO 4 , 0.41 mM CaCl 2 , 0.33 mM Ca(NO 3 ) 2 , 10 mM HEPES, pH 7.4). Animal cap explants were cut from uninjected or injected embryos at stage 8.5. Control and injected embryos or animal cap explants were cultured to desired stages and collected for further analyses.
Whole Mount in Situ Hybridizations-Standard procedures for whole mount in situ hybridization were used (34).
Plasmid Construction, in Vitro Transcription, and Microinjection-Plasmid construction was made by using a PCR-based strategy. All the mutants of Oct25, Oct60, Oct91, and mouse Oct4 (supplemental Fig. S1) that were used for RNA microinjection were ligated into a pCS2 ϩ vector. GFP fusions of Oct25, Oct25⌬NLS, and Oct25⌬POU(273-301) used for cell transfection were subcloned into a pCS2 ϩ eGFPmcs vector. The GSTtagged fusions of Oct25 and its mutants used for EMSAs were subcloned into pGEX-4T1 (Amersham Biosciences). For luciferase assays, the promoter region Ϫ257/ϩ24 of Xnr3 (the first nucleotide of transcription start site defined as position ϩ1) (35) was amplified from X. laevis genomic DNA and subcloned into pGL3-basic vector (Promega) to generate Xnr3Luc. All constructs were confirmed by sequencing.
Different doses (see below) of RNAs coding for mutants of Oct25, Oct60, Oct91, and Oct4 were injected into the equatorial region of either two dorsal blastomeres, two ventral blastomeres or all four blastomeres of four-cell stage embryos. 400 pg of VegT RNA, 100 pg of BMP4 RNA, 800 pg of dnXAR1 RNA, and 400 pg of dnTCF3 RNA were injected per embryo. 50 ng of an antisense morpholino-oligonucleotide (Gene Tools) against VegT (VegTMO, 5Ј-TTCCCGACAGCAGTTTCT-CATTCCA-3Ј) were injected per embryo to knock down the VegT activity in Xenopus embryos.
DNA Transfection and Imaging-HEK293 cells (ATCC CRL 1573) were grown at 37°C under 5% CO 2 in Dulbecco's modified Eagle's medium (Invitrogen) supplemented with 10% fetal calf serum (Biochrom, Berlin, Germany). To analyze the subcellular localization of Oct25 proteins, the cells were plated on chambered cover glasses (Nunc, Rochester, NY) at a density of 80,000 cells per chamber in 2 ml of medium. After 16 h of incubation at 37°C, cells were transfected with 200 ng of expression plasmids for enhanced GFP or Oct25-EGFP fusions using the Nanofectine transfection reagent (PAA, Pasching, Austria) according to the manufacturer's instructions. 24 h after transfection, the living cells were analyzed using an IX71 fluorescence microscope (Olympus, Hamburg, Germany) equipped with a digital camera (C4742, Hamamatsu, Hamamatsu, Japan), a 100-watt mercury lamp, and a standard fluorescein isothiocyanate (excitation, HQ470/40; emission, HQ525/50) filter set.
Quantitative RT-PCR-RNAs were extracted from embryos or animal cap explants using QiaZol (Qiagen), treated with DNase I, and purified with RNeasy kit (Qiagen). cDNAs were prepared from total RNAs using RevertAid TM first strand cDNA synthesis kit (Fermentas). The method and primers for quantitative RT-PCR (qRT-PCR) were exactly performed as described previously (14,15).
EMSA and GST Pulldown Assays-EMSAs and GST pulldown assays using bacterially expressed proteins were performed essentially as described previously (13,15).

Nuclear Localization Signal Is Essential for the Activity of
Oct25-Oct25 and other Oct4 homologous proteins reveal a similar domain structure, including the diverse N-and C-terminal regions and the well conserved POU-specific (aa 237-301) and Hox-specific (aa 322-380) domains. These two domains are spaced by a linker region (aa 302-321). Moreover, a nuclear localization signal (RKRKR, aa 320 -324) exists at the beginning of the Hox domain ( Fig. 1 and supplemental Fig. S1). We investigated the contribution of each region to the function of Oct25 by generating deletion constructs (supplemental Fig.  S1) that were overexpressed in Xenopus embryos. qRT-PCR showed that deletion mutants either lacking the C-terminal region (Oct25⌬C), N-terminal region (Oct25⌬N), or lacking both the C-and N-terminal regions (Oct25PH) inhibited mesodermal and endodermal marker gene expression (Fig. 1A), which is reminiscent of the wild type protein effect (14 -16, 26). Therefore, these data suggested that Oct25 lacking these regions does not alter its function in repression of mesendo-derm formation. However, when both the POU and the Hox domains, which are responsible for DNA binding, were removed from Oct25, the resulting mutant (Oct25⌬PH) did not generate any significant effect on embryogenesis after RNA injection (Table 1 and supplemental Fig. S2A).
Oct4 has a nuclear localization signal (NLS) responsible for its residence in the nucleus (36). The NLS motif is also present at the beginning of the Oct25 Hox domain ( Fig.  1 and supplemental Fig. S1). A mutant Oct25 without the NLS motif (Oct25⌬NLS) fused to GFP revealed a diffuse distribution throughout the cells, similar to GFP alone (Fig. 1B). In contrast, the wild type Oct25 was localized exclusively in the nuclei, suggesting that the motif is indeed required for the nuclear localization of the wild type protein. Consequently, Oct25⌬NLS had no strong effect on mesendodermal gene expression when injected into embryos (Fig. 1C). This indicates that the NLS is essential for the activity of Oct25.
Removal of aa 273-301 in the POU Domain of Oct25 Revealed a Dorso-anteriorizing Activity-Overexpression of further deletion mutants of Oct25 were examined regarding their effects on embryogenesis. The effect of the mutant Oct25⌬(330 -381) lacking the complete Hox domain but retaining the NLS was still similar to the wild type protein (  Fig. 2B). This effect is contrary to the overexpression of wild type protein, which caused differentiation defects and malformations of posterior structures (Fig. 2C). Moreover, deletion of this region did not alter the nuclear localization in HEK293 cells transfected with Oct25⌬POU(273-301)-GFP plasmid (Fig. 1B). The production of partial secondary axis was inhibited when wild type Oct25 RNA was co-injected (supplemental Fig. S3A; supplemental Table S1). To characterize this phenotype in more detail, we performed whole mount in situ hybridization to analyze marker gene expression in Oct25⌬POU(273-301)-injected embryos. During gastrulation, the pan-mesodermal marker Xbra was not significantly altered compared with uninjected control embryos (Fig. 2D). However, the endodermal gene, Xsox17␣, revealed ectopic expression, which extended to the mesodermal and even ectodermal areas (Fig. 2E). Expression of Chordin (Chd) and Goosecoid (Gsc) was dramatically enhanced when Oct25⌬POU(273-301) RNA was injected dorsally. Interestingly, ventral injection led to ectopic activation of Chd but not of Gsc (Fig. 2, F and G). Injection of wild type Oct25 has been shown to lead to a down-regulation of Xbra, Xsox17␣, and Gsc (14,16). During neurulation, both dorsal and ventral injections resulted in a severe reduction of epidermal keratin (keratin) expression (Fig. 2H). Xsox2, a neural plate marker gene, was significantly increased at the dorsal side and ectopically activated at the ventral side (Fig. 2I). Therefore, the injected embryos developed extra neural tissue at the cost of the epidermis. During the tailbud stage, injected embryos formed extra tissue that was derived from all three germ layers as indicated by ectopic expression of XAG2 (Fig. 2J), a gene that marks the cement gland as the most anterior structure, the somite gene XMyoD (Fig. 2K), the neural gene NCAM (Fig. 2L), and the anterior endodermal marker gene Xhex (Fig. 2M). The supplemental Table S2 lists the numbers of embryos in these whole mount experiments showing significant changes in gene expression. We further analyzed the expression of genes that are involved in germ layer formation by qRT-PCR. In whole embryos, the organizer genes that specify dorsal structures, Chd, Gsc, noggin, cerberus, Dkk1, and Xnr3, were all strongly up-regulated (Fig. 3A). Noteworthy, all these genes were significantly down-regulated by injection of wild type Oct25 RNA. The nodally related genes Xnr1, Xnr2, Xnr4, Xnr5, Xnr6, and Siamois (Sia), a target gene of maternal ␤-catenin pathway, as well as the neural plate genes Xsox2 and Xsox3 were also dramatically augmented. However, the genes in the BMP signaling pathway (BMP4, Xvent1, and Xvent2), which specify ventro/ posterior structures, were not affected or slightly down-regulated. Next, we figured out how these genes respond to injection of Oct25⌬POU(273-301) in animal caps. Indeed, the genes responsible for mesendoderm formation and neural induction, for instance Chd, Gsc, noggin, cerberus, Dkk1, Xnr1-3, Xsox2, and Xsox3, were also induced or enhanced in injected animal caps (Fig. 3B).

Stimulation of Germ Layer Formation by Oct25⌬POU(273-301) Is Dependent on the Activities of VegT, Nodal/Activin, and ␤-Catenin Signaling-Maternal
VegT and ␤-catenin induce mesendoderm formation via transcriptional activation of ligands of the nodal signaling pathway and dorsalizing genes, like Sia and Xnr3. Therefore, we asked whether the induction of partial secondary axis formation in embryos by Oct25⌬POU(273-301) relies on the activities of VegT, ␤-catenin, or nodal/activin signaling. We used an antisense morpholino-oligonucleotide against VegT (VegTMO) to knock down the function of VegT in embryos. First, we tested the efficiency of VegTMO. Vegetal injection of VegTMO led to severely reduced body axis, strongly pigmented belly side, and gradual death after stage 30. VegTMO-injected embryos were rescued by VegT RNA injection to form a nearly normal body plan (supplemental Fig. S4; supplemental Table S1). We then injected Oct25⌬POU(273-301) RNA or VegTMO alone and in combination and analyzed gene expression (Fig. 3D). Again, genes like Gsc, Chd, Xnr1, Xnr2, and Sia were up-regulated by Oct25⌬POU(273-301) alone. In contrast, injection of VegTMO resulted in a reduction of these genes. When Oct25⌬POU(273-301) RNA was co-injected with VegTMO, we did not observe transcriptional stimulation of Gsc, Chd, Xnr1, Xnr2, and Sia. The results suggest that stimulation of these genes by Oct25⌬POU(273-301) is dependent on VegT activity. Furthermore, we demonstrated that up-regulation was also dependent on the nodal/activin signaling pathway, because blocking with a dominant-negative activin receptor, dnXAR (38), prevented the Oct25⌬POU(273-301) mutant from stimulating these genes to high levels anymore (Fig. 3E). Finally, when the ␤-catenin signaling was blocked by dominant-negative TCF3 (dnTCF3 (39)), transcription of these genes, including Xnr3, was almost completely abolished. Inhibition of ␤-catenin signaling also caused a failure in stimulation of these genes by co-injection of Oct25⌬POU(273-301) RNA (Fig. 3F). This is especially obvious for the dorsalizing genes Xnr3 and Sia, which are direct targets of the Wnt pathway, but somewhat less obvious for Gsc, Chd, and Xnr1, which are also activated by the nodal/activin pathway. In summary, these experiments reveal that Oct25⌬POU(273-301) alone was not able to stimulate transcription of genes responsible for germ layer formation without endogenous inducing activities of VegT, nodal/ activin, and especially ␤-catenin signaling.
Oct25⌬POU(273-301) Rescues BMP4-ventralized Embryos-We further explored whether Oct25⌬POU(273-301) could rescue embryos that were ventralized by the BMP signaling pathway. Dorsal injection of BMP4 RNA caused formation of "belly pieces" without dorsal structures (Fig. 4B). Co-injection of a low dose of Oct25⌬POU(273-301) RNA (400 pg) together with BMP4 RNA already led to formation of discernible body axes (data not shown). A better rescue effect was achieved with a higher dose (1 ng) of Oct25⌬POU(273-301) RNA. In this case, the embryos displayed clear dorsal-ventral and anterior-posterior body axes, although normal embryos did not form ( Fig. 4C; supplemental Table S1). As expected, gene expression analysis after dorsal injection of BMP4 RNA showed an inhibition of dorsal genes, like Chd, Gsc, Dkk1, Xsox2, and Xsox3, whereas the ventral gene Xwnt8 was up-regulated. However, transcript levels of these genes were reversed by co-injection of Oct25⌬POU(273-301) RNA, as shown by up-regulation of dorsal genes and down-regulation of the ventral marker gene Xwnt8 (Fig. 4D). Therefore, we conclude that Oct25⌬POU(273-301) can rescue the ventralized phenotype caused by dorsal activation of BMP signaling and thus reveals an anti-ventralizing activity.
To investigate how Oct25⌬POU(273-301) interacts with the BMP signaling pathway, we co-injected Oct25⌬POU(273-301) RNA with constitutively active type I BMP receptor RNA, hAlk-6 (40), and analyzed the expression pattern of the BMP4 target gene Xvent2. We found that the ectopic activation of the Xvent2 gene by hALK-6 RNA injection could not be rescued by co-injection of Oct25⌬POU(273-301) (Fig. 4E). In contrast, the effect of BMP4 RNA injection was antagonized by Oct25⌬POU(273-301) co-injection (Fig. 4F). These experiments indicate that the BMP-antagonizing and dorsalizing effect of Oct25⌬POU(273-301) is caused at the level of the ligand and not downstream via the receptor or components of the transduction pathway.

Reversal of Oct25 Function by Single Amino Acid Deletions or Mutations-As
Oct25 with loss of the C-terminal half of the POU-specific domain showed a dominant-negative effect, we have asked whether the depletion of the N-terminal half in the POU domain also led to a similar effect or if we could pinpoint the amino acids that were responsible for such an effect. Actually, deletion of the N-terminal half of the POU domain from aa 237 to 272 (Oct25⌬POU(237-272)) also caused an up-regulation of dorsalizing genes. Deletions of smaller stretches in this region from aa 268 to 272 (Oct25⌬POU(268 -272)) or even only two amino acids 271 to 272 (Oct25⌬POU(271-272)) resulted in increased transcription of these genes (Fig. 6A). Such an effect was also observed when a longer stretch of aa 268 -301 (Oct25⌬POU(268 -301)) was depleted (Fig. 6B). When most part of the POU-specific domain from aa 250 to 301 (Oct25⌬POU(250 -301)) was removed, we could still clearly observe the dorsalized phenotype in embryos upon RNA injection (Table 1; supplemental Fig. S3B) and up-regulation of the dorsalizing genes (Fig. 6B). However, when a region from aa 283 to 301 of the C-terminal end of POU domain (Oct25⌬POU(283-301)) was missing, there was no significant dorsalizing effect anymore. If only aa 293-301 (Oct25⌬POU(293-301)) were deleted, this mutant showed a gene repression effect resembling that of wild type Oct25 (Fig.  6B). The results suggest that the middle region of the POU domain is critical for repression of gene transcription by Oct25. As a matter of fact, deletion of aa 273-276 (Oct25⌬POU(273-276)), 273-274 (Oct25⌬POU(273-274)), or only a single amino acid 273 (Oct25⌬POU(273)) indeed caused an up-regulation of dorsalizing genes (Fig. 6C).
To gain additional support whether the sequence of the middle part of the POU domain would change the function of Oct25, we made a mutant in which the amino acids TTIC were reversed to ICTT (Oct25(TTIC 3 ICTT)). Moreover, another two mutants were generated, in which the amino acid cysteine 274 (Cys-274) was changed to proline (Oct25(C274P)) or serine (Oct25(C274S)), respectively, because proline is structurally different from and serine is similar to cysteine. Injection of Oct25(TTIC 3 ICTT) RNA and Oct25(C274P) RNA caused a significant up-regulation of Chd and Gsc expression ( Fig. 6D; supplemental Table S3). However, the mutant Oct25(C274S) did not reveal such an effect but instead repressed these genes severely ( Fig. 6D; supplemental Table S2). Hence, the results clearly show that disruption of the POU domain structure leads to a reversal of Oct25 function.

Alteration of the POU Domain Structure Leads to the Loss of DNA Binding Activity but Retains Protein Interaction
Properties-As Oct25 binds DNA via its POU-and Hox-specific domains, we explored whether the mutants described above show any changes in their DNA binding activity. We made use of the Xnr1 promoter that contains an Oct25-binding motif (15) to test the DNA-binding properties of the mutants. EMSAs demonstrated again that wild type Oct25 bound to a double-stranded oligonucleotide containing the Oct25-binding motif of the Xnr1 promoter. The mutant Oct25PH containing only the POU and Hox domains showed strong affinity for this target. When the Hox domain was deleted (Oct25⌬Hox(330 -381)), the mutant did not bind the promoter anymore (Fig. 7A). Moreover, all the mutants in which the POU-specific domain was completely or partially deleted (Oct25⌬POU-(237-301), Oct25⌬POU(273-301), Oct25⌬POU(268 -272), Oct25⌬POU(273-281), and Oct25⌬POU (273)) exhibited a loss of DNA binding activity. The mutants Oct25(TTIC 3 ICTT) and Oct25(C274P) were also not able to bind the DNA fragment (Fig. 7A). In contrast, the mutant Oct25(C274S) showed DNA binding activity. This situation is not unexpected, because Oct-1, a POU family protein binding to the octamer motif, has a Ser instead of Cys in the corresponding position of Cys-274 in Oct25 (supplemental Fig. S6).
Oct25 represses transcription of target genes of VegT, ␤-catenin, or nodal signaling via formation of protein complexes with signal transducers like VegT, TCF3, FAST1, and the Smad (derived from Caenorhabditis elegans small and Drosophila MAD genes) transducers on the target gene promoters (15,16). We therefore explored whether the mutants have lost the binding capacities to these proteins. However, in GST pulldown assays, both the mutants Oct25(C274P) and Oct25⌬POU(273-301) displayed similar interaction activities compared with that of wild type Oct25 (Fig. 7B). As expected, the mutant Oct25⌬POU(273-301) stimulated transcription from the wild type Xnr1Luc(Ϫ279) luciferase reporter (Fig.  7C) (15). We found that mutant reporters lacking the Octbinding site (Xnr1Luc(Ϫ279-Oct)), the TCF/LEF-binding site (Xnr1Luc(Ϫ279-TCF)), or the Tbox-binding site (Xnr1Luc(Ϫ279-Tbox)) (15) were also strongly stimulated by co-injection of Oct25⌬POU(273-301) RNA (Fig. 7C). This means that up-regulation of the promoter does not require each motif by itself. Also, a promoter mutant lacking both, the TCF/LEF and the Tbox sites (Xnr1Luc(Ϫ279-TCF-Tbox)), displays a strong activation by Oct25⌬POU(273-301) (Fig. 7D). However, this stimulation is severely reduced by co-injection of dnXAR1. Therefore, we suggest that the activation of this Xnr1 promoter fragment does not only involve complex formation between dorsalizing Oct25 mutants and the transcription factors VegT and TCF but that it also involves interaction with additional transcriptional regulators, which are provided by the nodal/activin pathway. To exclude an artificial binding of the Oct25 mutant to the Ϫ279/Ϫ5 Xnr1 promoter fragment, we performed EMSAs with the wild type and the mutant protein. Although Oct25 was bound, no interaction occurred with the Oct25⌬POU(273-301) mutant (Fig. 7E).

DISCUSSION
Here, we report a detailed study on the functional domains of the Oct4 homologous protein Oct25 in X. laevis. Although quite a few studies have been carried out on the functions of the mammalian Oct4 protein domains, we discovered a previously undefined phenomenon, i.e. when the POU-specific domain structure is disturbed, either by deletion of a few amino acids or by mutation of a single amino acid, the resulting mutated proteins exhibit dorsalizing activity in Xenopus embryos. This indicates a reversal of protein function, because the wild type protein was previously shown to repress mesodermal and endodermal germ layer formation (14 -15, 26).
The Oct4 homologous proteins are composed of the N-terminal region, the POU-specific domain, the linker region, the nuclear localization signal, the Hox-specific domain, and the C-terminal region. In this study, we show that Oct25 without nuclear localization signal (Oct25⌬NLS) has apparently no effect on embryonic development and gene expression. This is not surprising, because the mutant protein distributes primarily in cytoplasm instead of the nucleus. In a study of an Oct4 mutant without NLS (Oct4⌬NLS), Oct4 and Oct4⌬NLS form a dimer and, consequently, the endogenous Oct4 protein in ES cells is retained in the cytoplasm by Oct4⌬NLS and prevents binding to target gene promoters (36). Such a mechanism may also apply to Oct25⌬NLS.
The mutants with complete removal of one or two regions or domains display two types of effects. Either the transcription of genes for mesendoderm formation is repressed or these genes are not significantly affected. Oct25 deletion mutants lacking the N-terminal region (Oct25⌬N) or the C-terminal region (Oct25⌬C) inhibit transcription of genes responsible for mesendoderm formation. We have previously shown that this effect is also caused by overexpression of wild type Oct25 protein (14 -16), suggesting that the N-and C-terminal regions are functionally redundant to regulate at least a subset of target genes. A similar conclusion was reported in ES cell-based complementation assays (41). Oct4 proteins lacking either the N-or C-terminal region were able to substitute the wild type Oct4 function in rescuing the ES cell phenotype in the ZHBTc4 ES cell line. This does not automatically mean that the two deletion mutants are identical, because it seems that at least Oct4 lacking the C-terminal region can support the expression of Ebaf/Lefty1, although the mutant without the N-terminal region cannot (41). Like the Oct25⌬N and Oct25⌬C mutants, the Oct25 mutant lacking both N-and C-terminal regions (Oct25PH) also exerts an inhibitory effect on mesendodermal gene transcription. By analogy, it would be expected that the equivalent mutant of Oct4 could rescue the phenotype in ES cell complementation assays as well. However, unlike Oct4 lacking the N-or C-terminal region, Oct4 lacking both regions cannot rescue this ES cell phenotype (41). It is known that Oct4 may play dual roles for gene transcription. On the one hand, it stimulates transcription of genes supporting self-renewal and pluripotency, and on the other hand it represses genes promoting differentiation (42)(43). It may be that when both transactivation domains in the N-and C-terminal region are missing, the remaining part of the protein is still able to exert its repression effect on gene transcription, but the capacity for gene activation is somehow lost.
The difference in the activities of the Oct25 mutant lacking the complete POU domain (Oct25⌬POU(237-301)) and the mutant lacking the complete Hox domain but retaining the NLS (Oct25⌬Hox(331-381)) suggests that the POU domain but not the Hox domain is a prerequisite for delivering the regulatory effect on genes examined in this study. This notion is supported by the effect of a mutant in which the C-terminal half of the POU domain (Oct25⌬POU(273-301)) is missing. Contrary to the wild type protein that inhibits germ layer formation, Oct25⌬POU(273-301) induces ectopic germ layer formation in embryos as shown by the presence of partial secondary axis. Additionally, a series of analyses clearly revealed enhanced or ectopic transcription of genes that induce germ layer formation, like Xnrs and Sia, which are target genes of maternal VegT and ␤-catenin, and the organizer genes, like Chd, noggin, cerberus, and Dkk1, which code for antagonists against BMPs and Wnts. Consequently, we were able to rescue BMP4-ventralized embryos to form dorsal structures by using the Oct25⌬POU(273-301) mutant. Noteworthy, overexpression of the mutant in embryos did not generate any significant quantitative effect on the transcription of BMP4 and its downstream targets Xvent2 and Xvent1. This observation suggests that Oct25⌬POU(273-301) does not achieve its dorsalizing effect via transcriptional suppression of BMP. Rather, it induces or enhances transcription of genes coding for proteins such as Xnr3, Chd, Cerberus, and Dkk1, which are antagonists of BMPs and Wnts via direct interaction. As a matter of fact, Oct25⌬POU(273-301) RNA injection antagonized the effects of BMP4 on ectopic expression of Xvent2 but not the effects of the constitutively active BMP receptor hAlk-6. Therefore, we conclude that the underlying mechanism for the dorsalization of embryos by the mutant is due to the inhibition of ligand activities of BMP and Wnt signaling pathways. Although the wild type Oct25 inhibits the activities of VegT, ␤-catenin, as well as the nodal pathways, but promotes BMP4 signaling pathway (13)(14)(15)(16), we show here that the Oct25⌬POU(273-301) mutant has an opposite activity. These results also substantiate our previous conclusion that the endogenous Oct proteins serve as general inhibitors of differentiation signals in a way that germ layer formation can precisely occur.
The dominant-negative effect was not only observed in the mutant missing the C-terminal half of the POU domain but also in the mutant missing a stretch of amino acids located toward the N terminus of the POU domain (Oct25⌬POU(237-272), Oct25⌬POU(268 -272), and Oct25⌬POU(271-272)). However, complete removal of the POU domain leads to a mutant protein that has essentially lost both DNA binding activity and its capability of gene regulation. Therefore, it seems that the middle part of the POU domain is critical for gene regulation. This postulation is confirmed by the deletion of four amino FIGURE 7. DNA binding, protein interaction, and luciferase assays using Oct25 mutants. A, EMSAs (8% PAGE) show that the GST-Oct25 fusion protein binds to a canonical octamer motif (underlined) within the Xnr1 promoter. Binding is also observed for the mutant containing only the POU and Hox domains (Oct25PH). Deletion of the Hox-specific domain results in loss of DNA binding activity. GST alone was used as control. The mutants with a truncation of either the complete POU-specific domain (aa 237-301) or different regions of the POU-specific domain lose their DNA interaction capacity. DNA binding is also lost, when the amino acids TTIC are changed to ICTT (TTIC 3 ICTT) or Cys-274 is mutated to Pro (C274P) but is retained in the (C274S) mutant. B, GST pulldown assays show that Oct25(C274P) and Oct25(⌬POU(273-301) still interact with Smad4, Smad2, Smad3, Smad1, FAST1, VegT, or TCF3. C, luciferase reporter activities driven by the wild type Xnr1 promoter (Xnr1Luc(Ϫ279) or the indicated deletions (15) in the absence or presence of Oct25⌬POU(273-301). D, luciferase reporter assays driven by the Ϫ279(ϪTCF-Tbox) Xnr1 promoter upon co-injection with Oct25⌬POU(Ϫ273-301) and dnXAR1. E, EMSA of the Ϫ279/Ϫ5 Xnr1 promoter fragment (6% PAGE) reveals binding of Oct25 but no interaction with Oct25⌬POU(Ϫ273-301) GST fusion proteins. acids, two amino acids, or a single amino acid, suggesting that even a minor disturbance will cause a functional reversal. As a matter of fact, a change in the order of the four amino acids TTIC to ICTT or a mutation of the amino acid cysteine at position 274 to a structurally different proline again results in a dominant-negative effect. When the same amino acid is mutated to the structurally similar amino acid serine, the function of the mutant as compared with wild type Oct25 is not significantly altered. This confirms that the configuration of the POU domain is essential for the regulation of gene transcription. Sequence comparison between Oct4 homologous proteins (supplemental Fig. S6) reveals that the POU domain is well conserved and especially that amino acids within the middle part of POU domain are identical. It is therefore rational that partial deletions of the POU domain in Xenopus Oct60 (Oct60⌬POU(248 -276)), Oct91 (Oct91⌬POU(264 -292)), or mouse Oct4 (Oct4⌬POU(177-205)) all exhibit a similar effect to that of Oct25⌬POU(273-301). These data also imply that Oct4⌬POU(177-205) may exert a dominant-negative effect on gene transcription in ES cells. Therefore, it would be interesting to know whether similar mutants that reverse the function of POU-V factors do naturally occur. A search for sequence variants within an 11.3-kb region of the human OCT3/4 gene revealed a high degree of polymorphism, but no sequence variation was detected in exons 2 and 3 encoding the POU-specific domain (44). However, a novel Oct3/4 transcript retaining the complete 225-bp intron 2 sequence as a putative novel exon has recently been described (45). This insertion disturbs the coding sequence starting at the center of the POU domain. The biological significance of this novel splice variant is not yet known, but it is expressed in human ES cells and is down-regulated following the onset of differentiation.
The strong activation of the differentiation genes by Oct25⌬POU(273-301) and Oct25(C274P) raises the question for the underlying molecular mechanisms. Formation of a partial secondary axis lacking heads but including cement gland structures is reminiscent of the blockage of the BMP4 pathway by a dominant-negative type II receptor (46). Also, suppression of epidermal keratin and expansion of the neural field as shown by Xsox2 after overexpression of Oct25⌬POU(273-301) are characteristic of an inhibition of the BMP4 signaling pathway. In line with that, we were able to rescue the BMP4 overexpression phenotype by co-injection of Oct25⌬POU(273-301).
How the dorsalizing mutant can stimulate transcription of genes that promote germ layer formation and dorso-anteriorization of embryos is still not clear. A canonical viewpoint for the transcriptional regulation of genes by POU family proteins is that they can interact with the enhancers via the octamer motif, with the POU domain binding the ATGC half-site and the Hox domain binding the AAAT half-site. The POU domain is composed of a cluster of four ␣-helices, and the Hox domain also contains three ␣-helices. In each domain, helices ␣2 and ␣3 form a typical helix-turn-helix motif for DNA interaction. In the POU protein Pit-1, amino acids Ser, Gln, Thr, and Arg in helix ␣3 of the POU domain and the amino acids Arg, Val, Cys, Asn, and Gln in helix ␣3 of the Hox domain directly contact DNA base pairs (47). Interestingly, these ␣3 helices are identical in all POU family proteins, including those in subclass V (sup-plemental Fig. S6). This reflects that the helix structure is crucial for DNA binding, and its destruction will essentially lead to the loss of DNA interaction. Consequently, all mutants with changes in the POU or Hox domain with the exception of C274S cannot bind the promoter via the octamer sequence anymore. However, only those mutants with alterations in the POU domain demonstrate a reversed activity in the regulation of genes involved in mesendoderm formation. Hence, the observed dominant-negative effect cannot be solely interpreted by the loss of DNA interaction. However, there might exist DNA sequences other than the canonical octamer motif for protein binding, and mutants with disruption of the POU domain might bind artificially via their Hox domain to other target sites. But when the Hox domain except for the NLS is removed from Oct25(C274P), the resulting mutant Oct25(C274P)⌬Hox(330 -381) still exhibits dorsalizing activity (supplemental Fig. S3C). Thus, it is excluded that disruption of the POU domain leads to artificial binding of the Hox domain and that the dorsalized phenotype is due to alternative DNA binding.
Stimulation of differentiation genes by Oct25⌬POU(273-301) is severely compromised in response to functional knockdown of VegT or blocking of the ␤-catenin and nodal/activin signaling pathways. We have previously reported that Oct25 forms complexes with VegT, TCF3, FoxH1, or Smad transducers to block transcription of VegT, ␤-catenin, or nodal/activin target genes (15,16). Interestingly, the dorsalizing mutants Oct25(C274P) and Oct25⌬POU(273-301) can still interact with these proteins like the wild type Oct25. Moreover, the dorsalizing mutant Oct25⌬POU(273-301) stimulates the Xnr1 promoter constructs in which the Oct-binding site, the TCFbinding site, or the Tbox binding site are missing. Even removal of both the TCF-and Tbox-binding sites does not abolish promoter activity. However, this promoter is also subject to regulation by the nodal/activin signaling pathway. Indeed, we could show that the stimulation obtained with the dorsalizing mutant is severely reduced by dnXAR1. We assume that the activity of the dorsalizing mutants is not due to direct interaction with DNA but must be mediated by protein/protein interactions. This assumption is further supported by previous observations that Oct25 represses gene transcription also in the absence of the octamer motif (15,16). Co-repressors or co-activators that are differentially recruited to the regulatory complexes of Oct25 or of the dorsalizing mutants could lead to an opposite effect on the regulation of corresponding genes. It will therefore be interesting to investigate further components of the regulatory complex containing Oct25 mutants that promote dorsalization and formation of a partial secondary axis.