Klf4 is required for germ-layer differentiation and body axis patterning during Xenopus embryogenesis

During Xenopus early embryogenesis, Nodal/Activin, Wnt, BMP and FGF signaling pathways play key roles in promoting germ-layer formation. Nodal/Activin is the primary signal to induce mesoderm and endoderm in a dose-dependent fashion. FGF signaling also participates in mesoderm formation (Amaya et al., 1991; Amaya et al., 1993), mainly through providing competence for the embryonic cells to Nodal/Activin. BMP and Wnt pathways are active at the ventral side of embryo (Christian et al., 1991; Dale and Wardle, 1999) and are principally responsible for ventro-posteriorization of germ layers (Maéno et al., 1994; Suzuki et al., 1994; Schmidt et al., 1995; Dale and Wardle, 1999). At the dorsal side, they are blocked by antagonists secreted from the Spemann organizer: notably Noggin, Chordin, Cerberus, Dkk1, Xnr3, etc. (De Robertis et al., 2000). Thus the two groups of signals establish a balance for patterning body plan.

In Xenopus, the Nodal ligand genes, Xnr1-6, are induced by the maternal transcription factor VegT in vegetal cells (Clements et al., 1999; Hyde and Old, 2000; Takahashi et al., 2000; Hilton et al., 2003). Upon ligand gene transcription, Nodal signal is transmitted downstream and induces transcription of mesoderm- and endoderm-specific genes: Xbra, Mix1, Mix2, Goosecoid, Milk, Mix.1, Mixer, Sox17 and GATA4-6, for example (Xanthos et al., 2001; Shivdasani, 2002; Zorn and Wells, 2007). Endoderm-specific genes, meanwhile, inhibit mesoderm genes such that mesoserm and endoderm formation is restricted within correct locations. Maternal β-catenin signaling is enriched in dorsal-vegetal cells and induces Siamois transcription in the Nieuwkoop centre (Wodarz and Nusse, 1998), which subsequently induces gene transcription in the Spemann organizer (Wessely et al., 2001) to antagonize ventral signals. β-Catenin also works in synergism with VegT to enhance transcription of Nodal-related genes (Agius et al., 2000; Takahashi et al., 2000), hence establishing a gradient of Nodal signal, with higher activity dorsally and lower activity ventrally. In addition, complex autoregulatory loops play important roles in the regulation of the activity of Nodal signaling (Schier, 2003).

Differentiation of early embryonic cells into germ layers is accompanied by the loss of pluripotency, which is maintained by pluripotency factors. In mammals, these factors are typically Oct4, Sox2, Nanog, cMyc and Klf4 (Niwa et al., 2000; Zaehres et al., 2005; Avilion et al., 2003; Fong et al., 2008; Nakatake et al., 2006). Xenopus Oct4 homologous factors Oct60, Oct25 and Oct91 inhibit mesendoderm germ-layer formation via inhibition of the activities of VegT, β-catenin and Nodal (Cao et al., 2006; Cao et al., 2007; Cao et al., 2008). Sox2 is well known for its role in neural fate specification. Although these factors are crucial for the maintenance of pluripotency and self-renewal of embryonic stem (ES) cells, they exhibit distinct functions in ES cell differentiation assays and in embryonic development. Here, we report the identification and characterization of Kruppel-like factor 4 (Klf4) during Xenopus early embryogenesis. It promotes endoderm differentiation in both Nodal/Activin-dependent and -independent mechanisms. Moreover, it is involved in body axis patterning via activation of a subset of Spemann organizer genes, which code for Nodal/Activin, Wnt and BMP antagonists. In addition, loss of Klf4 function leads to failure of germ-layer differentiation. Thus we propose that Klf4 confers the competence of early embryonic cells to the activities of inducing signals such as Nodal/Activin so that embryonic cells can differentiate properly. Our results gain novel insights into the functions of Klf4 and the regulatory network for germ-layer differentiation and axis patterning in Xenopus embryos.

Xenopus laevis embryos and embryonic explants were obtained and cultured using conventional methods. To block endogenous Nodal activity, uninjected or injected embryos were incubated in culture medium containing 100 μM SB431542 (Sigma) from the four-cell stage until gastrulation. To block protein translation, uninjected or injected embryos were incubated in medium containing cycloheximide (CHX) at 25 μg/ml from stage 7 until stage 10.5. Embryos were staged according to Nieuwkoop and Faber (Nieuwkoop and Faber, 1975).

By searching databases with mouse Klf4, we found the Klf4 homologue in Xenopus tropicalis (XtKlf4). Further search of databases with XtKlf4 cDNA revealed two Xenopus laevis expressed sequence tags (ESTs) (GenBank accession numbers: BI445569 and CB196881) that encode two peptides sharing highest identities with XtKlf4. One EST contains the translational start site and the other contains the stop site. The cDNA containing the complete open reading frame (ORF) was amplified from a pool of cDNAs derived from stage 1 to stage 26 embryos. To make expression plasmid of Xenopus laevis Klf4, the ORF was subcloned to pCS2+ to generate pCS2+Klf4. The N-terminal region aa 1-304 with C-terminal zinc fingers missing was PCR amplified to make construct pCS2+Klf4ΔZF. The C-terminal DNA-binding domain (DBD) aa 270-404 was subcloned to make pCS2+Klf4(DBD). For the test of efficiency of the antisense morpholino against Klf4, the ORF including the morpholino binding site was ligated to pCS2+eGFPmcs and pCS2+6MTmcs vectors to make pCS2+Klf4-eGFP and pCS2+Klf4-MT, respectively. The repression and activation form of Klf4 were made by ligating Klf4 DNA binding domain to pCS2+evemcs and pCS2+VP16mcs (Cao et al., 2008), thus resulting in plasmids pCS2+eve-Klf4(DBD) and pCS2+VP16-Klf4(DBD). A plasmid containing complete cDNA of mouse Klf4 (mKlf4) was purchased from IMAGE Consortium (Berlin) and the coding region was subcloned to make pCS2+mKlf4.

Whole-mount in situ hybridization on whole embryos or animal caps was carried out essentially as described (Harland, 1991).

Antisense RNA probes for whole-mount in situ hybridization and mRNAs for microinjection were prepared as described (Cao et al., 2006). To prepare antisense RNA probes for whole-mount in situ hybridization, plasmids for Cerberus, Chordin, Dkk1, Gsc, Klf4, Mix2, Mixer, Noggin, Siamois, Sox17α, Sox2, XAG2, Xbra, Xnr1, Xnr5 and Xvent2 were linearized and transcribed with T7 RNA polymerase. To prepare mRNAs for microinjection, plasmids pCS2+Klf4, pCS2+Klf4-eGFP, pCS2+Klf4-MT, pCS2+mKlf4, pCS2+Klf4ΔZF, pCS2+Klf4(DBD), pCS2+VP16-Klf4(DBD), pCS2+eve-Klf4(DBD), pCS2+dnTCF3, pCS2+NLS-LacZ, pSP64T-activinβB, pSP64T-dnXAR1, tBR-64T, pXFD (dnFGFR) and pSP64T-Xnr2 were linearized and transcribed with Sp6 mMessage mMachine kits (Ambion). All probes and mRNAs were cleaned up with an RNeasy Kit (Qiagen). An antisense morpholino oligonucleotide (MO), K4MO: TTCCCTCCACCTCTCATTAATCTGG – which targets 36/−12bp of 5′UTR – was designed to knock down endogenous Klf4 in Xenopus laevis. A six-base mismatched MO, K4MO6mis: TTCtCTCgACCTaTCATgAATaTGc (mismatched bases are in lowercase), and the standard control MO (ctrlMO), CCTCTTACCTCAGTTACAATTTATA, were used as controls. All MOs were purchased from GeneTools. Injected doses of mRNAs or MOs are described in the text.

Total RNAs and cDNAs were prepared using exactly the same procedure as described (Cao et al., 2006). Quantitative RT-PCR (qPCR) was performed on an ABI 7300 system and primers are listed in supplementary material Table S1. Amplification parameters were as follows: one cycle of predenaturation at 95°C for 10 seconds, followed by 40 cycles of denaturation at 95°C for 5 seconds, annealing and extension at 60°C for 31 seconds and an additional cycle for the melting curve. Crosspoints were calculated using ABI 7300 system SDS software. Final results were presented as histograms with relative units.

Luciferase assays were carried out with embryos or cells. In embryos, promoter reporter plasmid DNAs and mRNAs were injected into the equatorial region of all blastomeres at the two- or four-cell stage. Embryos were collected at gastrula stage and the method for measuring luciferase activity was as described (Cao et al., 2007). HEK 293T cells were grown in 24-well plates and cells in each well were transfected with 100 ng of reporter plasmid together with 100 ng of various expression plasmids. In each well, 1 ng of Renilla luciferase reporter plasmid was co-transfected as internal control and the total amounts of transfected plasmids were normalized using pCS2+ empty vector. Luciferase activity was measured using the Dual-Luciferase Assay System (Promega). Each measurement was repeated with at least four independent transfections.

Uninjected and injected embryos were collected at stage 10.5, homogenized in cold lysis buffer (50 mM Tris-HCl pH 7.5, 150 mM NaCl, 1% NP-40) supplemented with protease inhibitors (Roche). Homogenates were incubated on ice for 10 minutes and centrifuged at 12,000 rpm for 10 minutes, and supernatants were transferred to fresh tubes, boiled in 1× Laemlli buffer, and centrifuged again at 12,000 rpm for 5 minutes. Supernatants were collected and 10 μl of each sample were loaded into SDS-PAGE for electrophoresis. Western blotting was performed using the conventional method. A myc-tag antibody was used to detect the expression of Klf4-MT and an α-actin antibody was used for detection of actin.

Embryos injected with lacZ mRNA were fixed in HEMFA and subjected to X-gal staining (Coffman et al., 1990). After staining, embryos were washed in PBS, fixed again in HEMFA and stored in 100% ethanol at −20°C, until processed for whole-mount in situ hybridization.

The identified cDNA encodes a protein of 404 amino acids. The sequence has the highest similarities to Klf4 in other species: for instance, 94% in Xenopus tropicalis, 51% in zebrafish and 55% in mouse (supplementary material Fig. S1A,B). Three classical zinc-finger motifs are present at the carboxyl terminus, which are typical for Kruppel-like factors (Pearson et al., 2008) and nearly identical among Klf4 proteins in different species (supplementary material Fig. S1A). There is a record for Xenopus laevis Klf4 under accession number NM_001086359 in GenBank; however, this gene product shares the highest identity to Xenopus Klf17 [or Neptune (NM_001088664)] and mouse Klf2, but not Klf4. In the genome of Xenopus tropicalis, Klf4 gene locates upstream sequentially to rad23b, znf462 and tmem38b. When the order of these genes is reversed, the arrangement is identical to that in both zebrafish and mouse (Zfp462 is synonymous with znf462) (supplementary material Fig. S1C). These comparisons suggested that the sequence we identified is orthologous to Klf4 in other species.

Klf4 is maternally transcribed as it is present in the animal region of early cleavage stages, e.g. stages 3 and 6.5 (supplementary material Fig. S2A,B). During midblastula, Klf4 was detected ubiquitously in embryos but slightly enriched at one side of the embryos (supplementary material Fig. S2C). Later, the enrichment was found in the dorsal marginal zone in gastrula embryos (supplementary material Fig. S2D,E). Bisection of a gastrula embryo showed that Klf4 was present in ectoderm and the marginal zone, but enriched slightly in the dorsal margin of the organizer, prechordal mesoderm and endomesoderm (supplementary material Fig. S2F). During neurulation, Klf4 localizes to two narrow lines within the neural folds (supplementary material Fig. S2G) and the anlage of cement gland (supplementary material Fig. S2H). Neural expression of Klf4 soon disappears but the cement gland expression persists until the tadpoles hatch (supplementary material Fig. S2I,J). Klf4 is also specifically present in trigeminal nerve and lung primordium at stage 34 (supplementary material Fig. S2J), and the prospective duodenum/stomach at stage 43 (supplementary material Fig. S2K). During embryogenesis, maternal Klf4 is more abundant than zygotic Klf4 in gastrulae and neurulae. During the tailbud stages, the expression level rises up again (supplementary material Fig. S2L). Klf4 transcript is present in both animal and vegetal blastomeres at the eight-cell embryo stage (supplementary material Fig. S2M). At stage 8.5 when zygotic transcription and germ-layer differentiation starts, transcript was detected in animal, equatorial and vegetal regions (supplementary material Fig. S2N). Therefore, Klf4 transcription is ubiquitous in early embryos. In summary, spatiotemporal expression patterns of Klf4 suggest that it might be involved in early embryonic development.

The blastopore formed normally in uninjected control embryos and tended to close at stage 11.5. In embryos injected with Klf4 mRNA, gastrulation was severely interrupted, as there was no clear blastopore formation (Fig. 1A,C). At stage 32, the majority of these embryos showed severely reduced anteroposterior body axis, pronounced belly protrusion with heavy pigmentation and seemingly exaggerated cement glands (Fig. 1B,C). In injected embryos, expression of the pan-mesoderm marker Xbra was strongly inhibited, suggesting that mesoderm formation was blocked (Fig. 1D,G). The endoderm gene Sox17α was detected only in the vegetal area of normal embryos, but it was ectopically activated in equatorial and animal regions in injected embryos (Fig. 1E,G). The neuroectoderm gene Sox2 was expressed at the dorsal side of control gastrula embryos; however, Klf4 RNA injection led to expansion of the Sox2 expression domain to the ventral side, thus suggesting an increment in neuroectoderm (Fig. 1F,G). Furthermore, we injected one ventral-animal blastomere at the eight-cell stage with lacZ RNA alone or lacZ and Klf4 RNAs together. Klf4-induced ectopic expression of Sox17α or Sox2 occurred within the lacZ-labeling regions (Fig. 1H-J), implying an autonomous effect of Klf4.

Isolated Xenopus blastula ectoderm, i.e. the animal caps, differentiates into epidermis. It can be induced to adopt different cell fates by inducers. At the gastrula stage, animal caps without Klf4 injection did not exhibit any discernible Xbra and Sox17α expression (Fig. 2A,B). Caps injected with Klf4 showed no difference from uninjected caps with respect to Xbra expression. However, there was strong activation of Sox17α in caps injected with Klf4 (Fig. 2B). We observed repeatedly weak Sox2 expression in uninjected caps, but Klf4 overexpression clearly led to an increase (Fig. 2C). These results are in agreement with the data observed in whole embryos. In addition, Mixer, another gene that is required for endoderm induction (Henry and Melton, 1998), was also strongly stimulated by Klf4 overexpression in both whole embryos and animal caps (Fig. 2D). Therefore, Klf4 is capable of inhibiting mesoderm while promoting endoderm and neuroectoderm formation. At neurula stage, Klf4-injected caps still showed higher levels of genes that specify neural precursors, e.g. Sox2, Sox3 and SoxD, but no neural tissue differentiation was observed, as revealed by NCAM expression (Fig. 2E). Epidermal differentiation was nearly completely blocked in Klf4 caps (Fig. 2E). Genes marking mesodermal tissues, α-globin and α-actin, were detected only in background levels in both control and Klf4 caps (Fig. 2E). Instead, significant increases in expression of the liver marker genes Xhex and XPTB (Chen et al., 2003) demonstrated that endodermal tissue differentiation occurred in Klf4 caps (Fig. 2E). The result suggested that Klf4 is capable of promoting the formation of neural precursor cells, but is not able to induce neural tissue differentiation on its own.

We designed an antisense morpholino oligonucleotide (K4MO) to knock down Xenopus laevis Klf4 by targeting the 5′UTR of its mRNA. K4MO could efficiently inhibit translation of the mRNA for the fusion protein Klf4-GFP (Fig. 3A,B) and mRNA for Klf4-MT fusion protein in embryos (Fig. 3C). By contrast, both the six-base mismatched control MO (K4MO6mis) and the standard control MO (ctrlMO) did not inhibit protein translation (Fig. 3A-C), showing the specificity of K4MO.

At the tailbud stage, the Klf4 morphant displayed a severely reduced anteroposterior body axis and head size (Fig. 3D,E). This phenotype was rescued by co-injection of 10, 20, 30 or 40 pg Klf4 mRNA, as co-injection of the mRNA reversed the shortening of the body axis to different degrees, with a better rescuing effect at higher doses (Fig. 3D,E). The rescued embryos were obviously better in body axis formation than the Klf4 morphant. Moreover, co-injection of 10 or 30 pg mouse Klf4 RNA (mKlf4) also resulted in a similar rescuing effect (Fig. 3D,E), suggesting a conserved function of Xenopus and mouse Klf4.

Injection of ctrlMO or K4MO6mis in embryos didn't affect expression of Xbra, Sox17α and Sox2. However, they were inhibited in the Klf4 morphant (Fig. 3F,G). In addition, the mesoderm genes Chordin, Xvent1, Xvent2 and Wnt8, the endoderm genes Mixer, FoxA2, GATA4, GATA5 and GATA6, and the ectoderm genes Sox2, Sox3, SoxD and XEMA were all repressed (Fig. 3H). The repression effect was specific for these genes because other genes such as Goosecoid (Gsc), Oct25, Oct60, Oct91, KMT5C, Cbx4 and the germ cell genes Nanos and Xpat were not significantly altered or even upregulated (Fig. 3H). These results implied that Klf4 is a prerequisite for the differentiation of early embryonic cells to germ layers.

In the classic point of view, the Nodal/Activin pathway is the primary inducing signal for endoderm specification. Thus we have investigated whether Klf4 has any influence on Nodal/Activin activity. Overexpression of Klf4 induced ectopic transcription of Xnr1 and Xnr5 (Fig. 4A,B,D). Accordingly, the Nodal/Activin direct target gene, Mix2, was stimulated (Fig. 4C,D). In HEK293T cells, transfection of the Xnr5 expression plasmid slightly stimulated ARELuc, the Nodal/Activin responsive luciferase reporter (Pierreux et al., 2000; Germain et al., 2000). Co-transfection of the Klf4 expression plasmid stimulated the reporter somewhat more strongly. FAST1, a key nuclear signal transducer for the Nodal/Activin pathway, could stimulate ARELuc significantly. Addition of Klf4 plasmid to the cells led to much stronger stimulation of the reporter (Fig. 4E). Stimulation of ARELuc reporter by Klf4 was recapitulated nicely in embryos via injection of the Nodal ligand mRNA alone or together with Klf4 mRNA (Fig. 4F). Injection of these mRNAs had no significant effect on the pGL3-basic plasmid (Fig. 4G) that was used for constructing the ARELuc reporter, excluding the unspecific stimulation of ARELuc by Klf4. Therefore, Klf4 promoted Nodal/Activin activity. In uninjected embryos, blocking protein translation with CHX treatment or blocking Nodal/Activin with SB431542, a specific chemical inhibitor of the Nodal/Activin type I receptor, led to dramatic inhibition of Mix2 expression. It was completely eradicated in embryos treated with both CHX and SB431542 (Fig. 4H,I). Vegetal injection of Klf4 RNA enhanced Mix2 transcription compared with uninjected embryos. Treatment of injected embryos with CHX, SB431542 or both resulted in significant reduction in Mix2; however, the expression levels were much higher than in uninjected embryos, respectively (Fig. 4H,I). Therefore, even in the absence of protein translation and/or the Nodal feedback loop, Klf4 was still able to stimulate Nodal target gene expression. Klf4 exhibited the same effect on Mix2 promoter activity. A luciferase reporter, Mix2Luc(−712), which contains −712/+13 fragment of Xenopus Mix2 promoter (Cao et al., 2008), was stimulated in embryos by Klf4 or Xnr2 overexpression. The stimulation grew much stronger when Klf4 and Xnr2 were simultaneously overexpressed (Fig. 4J). SB431542 treatment dampened the stimulation, however, the promoter activity in embryos with Klf4 overexpression was still much higher than the background level in treated embryos without Klf4 overexpression (Fig. 4J). Both gene expression and promoter analyses demonstrated that Klf4 directly regulates Nodal/Activin target gene expression. As a support, Klf4 could rescue the decrease of Mix2 expression resulting from injection of dnXAR1 (Fig. 4K,L), the dominant-negative Xenopus Activin receptor I (Hemmati-Brivanlou and Melton, 1992).

We tested the effect of Klf4 on mesoderm and endoderm differentiation induced by Nodal/Activin in animal caps. At the neurula stage, control animal caps differentiated into epidermis and showed no elongation; however, those injected with Activin mRNA showed obvious elongation (Fig. 5A,B). By contrast, caps injected with both Activin and Klf4 mRNAs did not elongate (Fig. 5A,B). qPCR revealed that, at the gastrula stage, animal caps injected with Activin mRNA showed high levels of mesoderm genes such as Xbra, Chordin, Gsc, Xvent2 and XmenF. Meanwhile, the Nodal target and endoderm genes such as Mix1, Mix2, Sox17α, Gata4-6, FoxA2 and FoxA4a were also induced in these caps (Fig. 5C). When Klf4 mRNA was co-injected, the mesoderm genes were dramatically inhibited, whereas the Nodal target genes and endoderm genes were enhanced (Fig. 5C). In summary, these analyses showed that Klf4 promotes endoderm differentiation while inhibiting mesoderm differentiation.

Activation of Nodal via injection of Xnr2 mRNA or inhibition of Nodal activity via injection of dnXAR1 mRNA led to different developmental defects (Fig. 6A,B), as previously reported (Hemmati-Brivanlou and Melton, 1992). Ventral injection of Klf4 resulted in belly protrusion with heavy pigmentation, resembling an anteriorized phenotype. K4MO injection generated embryos with a significantly shortened anteroposterior axis (Fig. 6A,B). Interestingly, co-injection of Xnr2 RNA and K4MO led to nearly normal embryos. By contrast, co-injection of Klf4 and dnXAR1 brought about an extremely anteriorized phenotype, which showed a severely decreased anteroposterior body axis but with hugely exaggerated cement glands (Fig. 6A,B). Finally, when dnXAR1 RNA and K4MO were injected together ventrally, embryos bent towards the ventral side, probably owing to a lack of tissue differentiation (Fig. 6A,B).

In congruence with the phenotypes above, injection of Klf4 or dnXAR1 RNA alone resulted in ectopic expression of XAG2, an anterior marker gene. When they were injected together, much stronger XAG2 expression was observed (Fig. 6C,D). This confirms the idea that overexpression of Klf4 in the absence of Nodal/Activin leads to extreme anteriorization of the body axis. The effect was also observed in embryos with Klf4 overexpression and inhibition of either BMP, FGF or Wnt (supplementary material Fig. S3A,B). Therefore, Klf4 anteriorizes body axis and enhances anteriorization in response to inhibition of posteriorization signals.

We examined how the crosstalk between Klf4 and Nodal/Activin affected germ-layer differentiation. First, injection of dnXAR1 mRNA resulted in decreased Sox17α in the vegetal region. Overexpression of Klf4 again activated ectopic Sox17α. When both Klf4 and dnXAR1 were injected together, Sox17α was still ectopically stimulated in the animal region (Fig. 7A,B). Such an effect could be reproduced in embryos in which Nodal/Activin was blocked with SB431542 (Fig. 7A,B). Mixer, another gene involved in endoderm differentiation, could also be induced by overexpression of Klf4 in embryos with SB431542 treatment (supplementary material Fig. S4A,B). These effects are similar in that Klf4 still activates Mix2 expression in the absence of Nodal/Activin. By contrast, Klf4 knockdown compromised the upregulation of mesendoderm genes Xbra, Chordin and Sox17α resulting from Xnr2 RNA injection (Fig. 7C,D). This result was in agreement with the fact that Klf4 knockdown was able to rescue the Xnr2 overexpression phenotype (Fig. 6A). Finally, blocking either Nodal/Activin or Klf4 led to a decrease in Xbra, Sox17α and Cerberus expression. When both signals were inhibited, gene expression was nearly totally lost (Fig. 7E,F). This means that there is a synergistic effect between the two signals and both are a prerequisite for germ-layer formation.

Injection of Klf4 alone or meanwhile in the absence of Nodal/Activin activity led to anteriorization of embryos. Accordingly, Klf4 overexpression stimulated transcription of Siamois (Fig. 8A,B) at midblastula, which is required for induction of the organizer precursor (Nieuwkoop centre) and organizer gene transcription. Organizer genes such as Cerberus, Dkk1 and Noggin were activated prematurely at midblastula, when there were no or very weak transcription of these genes (supplementary material Fig. S5A,B). During gastrulation, these genes were expressed in much broader or ectopic regions in Klf4-injected embryos (Fig. 7A,B). Activation of the organizer genes was a direct effect. In uninjected embryos, CHX treatment totally eliminated gene expression. By contrast, Klf4 overexpression was able to activate the genes in both DMSO- and CHX-treated embryos (supplementary material Fig. S6A,B).

Klf4 did not stimulate all organizer genes, as Chordin and Gsc were inhibited in response to Klf4 overexpression (Fig. 8C,D). Moreover, the ventral gene Xvent2 was also significantly inhibited in Klf4-injected embryos (Fig. 8C,D). Similar to Xenopus Klf4, mouse Klf4 (mKlf4) likewise induced Dkk1 activation in Xenopus embryos, supporting the idea that they are functionally homologous (Fig. 8C,D). Opposite to the effect of overexpression, Klf4 knockdown resulted in downregulation of Dkk1, Cerberus and Noggin (Fig. 8E,F). Upregulation of ventral genes such as Xvent2 was not detected in Klf4 morphant (Fig. 8E,F), this was possibly due to the failure of germ-layer differentiation (Fig. 7E). Hence, Klf4 is sufficient and necessary for the transcription of the subset of organizer genes, which are known as anterior fate inducers. Overexpression of Klf4 enhanced anteriorization when Nodal, BMP or Wnt was blocked. Accordingly, expression of Dkk1 was also strongly augmented in such embryos during gastrulation (Fig. 8G,H).

Regulation of genes in germ-layer differentiation and axis patterning by Klf4 gave rise to the question of whether it acts as a repressor or an activator. Klf4 consists of the transcriptional regulation domain at the N-terminus and the DNA-binding domain (DBD) of three zinc fingers at the C-terminus (supplementary material Fig. S7A). Injection of mRNA for the N-terminal region without zinc fingers (Klf4ΔZF) led to no significant change in embryogenesis. Injection of mRNA for the C-terminal region [Klf4(DBD)] had a minor effect on embryonic development. The affected embryos showed decreased tail region and slight belly protrusion (supplementary material Fig. S7B,C). Therefore, both regions are required for Klf4 function. As the C-terminus is responsible for DNA binding, we replaced the N-terminus of Klf4 with VP16 transcriptional activation domain or Even-skipped transcriptional repression domain to make VP16-Klf4(DBD) and eve-Klf4(DBD) fusion constructs, respectively (Fig. 9A). Injection of mRNA for VP16-Klf4(DBD) led to stimulation of Dkk1 and XAG2, similar to the wild-type Klf4. However, eve-Klf4(DBD) caused an opposite effect (Fig. 9B,C). Luciferase assays in cells exhibited that VP16-Klf4(DBD) acted similarly to Klf4 to enhance FAST1 stimulated ARELuc activity, but eve-Klf4(DBD) could not exert such an effect (Fig. 9D). As CHX treatment showed that Mix2 is a direct target of Klf4, a fragment of Mix2 promoter −221/+13 that contains the FoxH1 and Smad binding sites (the Activin response element) and a fragment of −142/+13 that does not contain the binding sites (Cao et al., 2008) were also used for testing the activity of the fusion constructs in embryos. VP16-Klf4(DBD) showed strong stimulating effect on both reporters; by contrast, eve-Klf4(DBD) was repressive (Fig. 9E,F). These experiments demonstrated that Klf4 is an activator for transcription of these genes.

We have observed for the first time in the present study that during Xenopus embryogenesis: (1) Klf4 promotes endoderm differentiation; (2) Klf4 functions in pattern formation of the body axis; and (3) Klf4 acts as a competence factor for early embryonic cells to differentiate.

We identified the cDNA for the mammalian Klf4 orthologue in Xenopus laevis. This is supported not only by sequence comparison and synteny analysis, but also by the fact that mouse Klf4 can rescue the Xenopus Klf4 morphant and that overexpression of Xenopus or mouse Klf4 generates a similar effect on gene expression. Xenopus Klf4 is maternally expressed. In mammals, there is little maternal transcript of Klf4 according to microarray data (Hamatani et al., 2008); however, significant maternal inheritance of Klf4 is observed in medaka and zebrafish (Wang et al., 2011; Li et al., 2011; Luo et al., 2011). This might be because of the divergence of gene regulation between lower and high vertebrates during evolution. Zygotic transcription of Klf4 is detected in both lower vertebrate and mammalian early embryos, but the abundance of zygotic transcripts is lower than maternal ones, as revealed in Xenopus and medaka fish (Wang et al., 2011). Ubiquitous transcription of Klf4 in early cleavage, blastula and gastrula embryos fits well with its function in germ-layer formation described in the present study. Besides, the localized expression of Klf4 in later stages suggests that Klf4 might also function in tissue differentiation or organ formation.

Overexpression and knockdown experiments in whole embryos and animal caps demonstrate that Klf4 plays critical roles for germ-layer differentiation and body axis patterning. In animal caps, Klf4 promotes neuroectoderm and endoderm; however, only endodermal tissue differentiation and not neural tissue differentiation occurs. Thus, Klf4 can drive endodermal differentiation and maintain the identity of neural precursors. The fate choice should be context-dependent. Although Klf4 induces transcription of Dkk1, Cerberus and Noggin, which neuralize ectoderm, the promotion of neuroectoderm by Klf4 seems to be cell-autonomous. In dissociated animal caps with Klf4 knockdown, neuralization of cap cells is not affected significantly (data not shown). In ES cells, Klf4 and Sox2 are components of the core regulatory circuitry for the maintenance of pluripotency via protein-protein interaction, thus promoting transcription of each other (Orkin et al., 2008; Wei et al., 2009). By analogy, Klf4 may directly interact with Sox2 in Xenopus embryos to promote neuroectoderm formation.

Klf4 enhances the activity of the Nodal pathway. The effect is reflected by the fact that Klf4 activates transcription Nodal ligand genes and target genes, and enhances promoter/reporter activity. This provides at least in part a mechanism for how Klf4 promotes endoderm differentiation. Luciferase assay with ARELuc in cells and embryos suggests that Klf4 might regulate the Nodal pathway indirectly, possibly via a Nodal feedback loop. Blocking the activity of Nodal receptor leads to a decrease in Klf4-induced Mix2 expression and promoter activity. Therefore, the feedback loop plays a role in Klf4-regulated Nodal activity. Meanwhile, Mix2 expression and promoter activity in response to CHX or/and SB431542 treatment also imply that Klf4 regulates Nodal target genes directly. Hence, both direct and indirect effects play a role in the regulation of Nodal/Activin by Klf4. This is consistent with the idea that Klf4 still induces ectopic endoderm gene expression, but the endogenous endoderm genes are repressed when Nodal receptor is blocked. As endoderm genes, e.g. Sox17α and Mixer, also inhibit mesoderm gene expression, it is plausible that endoderm differentiation dominates over mesoderm in both animal caps and whole embryos with overexpression of Klf4.

Overexpression of Klf4 caused an anteriorized phenotype resembling the effect of the simultaneous blocking of BMP, Wnt and Nodal (Piccolo et al., 1999). The effect can be explained by the idea that Klf4 is sufficient and necessary for the activation of a subset of organizer genes, e.g. Dkk1, Cerberus and Noggin, which code for antagonists for BMP4, Xnrs and Wnt8. However, Klf4 is not able to induce a complete secondary body axis. This is probably due to the fact that Klf4 induces some organizer genes but at the same time represses others; hence it is insufficient to drive the formation of a complete secondary axis. This can be supported by the idea that the anteriorized phenotype grows much stronger in embryos with Klf4 overexpression and Nodal, Wnt or BMP inhibition.

The mechanism of differential regulatory effects of Klf4 on organizer genes remains to be elucidated. Previous studies demonstrated that Klf4 can function as both a transcriptional activator and a repressor, depending on the transcriptional corepressors or coactivators it recruits (Ai et al., 2007; Evans et al., 2007; Evans et al., 2010). Thus Klf4 might regulate different organizer genes in cooperation with different transcriptional corepressors or coactivators. In addition, organizer genes and Nodal, BMP or Wnt pathways regulate each other and consist of a regulatory network for axis patterning. Disturbance of one or more signals in the network by Klf4 will inevitably lead to changes in other signals. According to the present knowledge, there are more than a dozen of genes expressed in the organizer. It will be interesting to investigate thoroughly the differential regulatory effects of Klf4 on these genes, which will give us more insights into the molecular mechanisms that control body axis patterning.

Loss of Klf4 function results in failure of the differentiation of germ layers, suggesting that Klf4 is required for the initiation of a differentiation program. This is because, in the absence of Klf4, Nodal/Activin activity is impaired. Therefore, target gene expression and mesendoderm differentiation is inhibited. Blocking both Klf4 and Nodal/Activin activity results in more severe inhibition. By contrast, Klf4 activates Nodal/Activin target gene expression and promoter activity. Thus, a synergistic effect exists between Klf4 and Nodal/Activin to induce target gene transcription. This dual regulation might exemplify a model for the correlation between Klf4 and other signaling, especially the Wnt pathway. Dkk1 is a known Wnt target, and Klf4 is also required for Dkk1 transcription: likewise Wnt and Klf4 might cooperate to regulate Dkk1. The model remains an intriguing topic and the detailed mechanisms need further investigation. In summary, we propose that Klf4 serves as a competence factor and enables early embryonic cells to be responsive to inducing signals for germ-layer differentiation and body axis patterning.

We thank Drs W. Wu, Y. L. Chen, M. Taira and D. Melton for gifts of plasmids.