Pou5f1/oct4 in pluripotency control: Insights from zebrafish



Gastrulation in vertebrates is a conserved process, which involves transition from cellular pluripotency to early precursors of ectoderm, mesoderm, and endoderm. Pluripotency control during this stage is far from being understood. Recent genetic and transcriptomic studies in zebrafish suggest that the core pluripotency transcription factors (TFs) Pou5f1 and TFs of the SoxB1 group are critically involved in large-scale temporal coordination of gene expression during gastrulation. A significant number of evolutionary conserved target genes of Pou5f1 in zebrafish are also involved in stem-cell circuit in mammalian ES cell cultures. Here, I will review the roles of Pou5f1 in development and discuss the evolutionary conservation of Pou5f1 functions and their relation to pluripotency control. genesis 50:75–85, 2012. © 2011 Wiley Periodicals, Inc.


All developing vertebrate embryos pass through a transient pluripotency period. A pluripotent cell is initially developmentally unrestricted and capable of realizing multiple cell fates. Loss of pluripotency by a cell is associated with the acquisition of specific developmental programs, which are dictated by position of the cell in the developing embryo. Cell transplantation studies in mice, chick, zebrafish, and frogs have demonstrated that most cells remain pluripotent during morphogenetic movements in gastrulation (Domingo and Keller,2000; Garcia-Martinez and Schoenwolf,1992; Ho and Kimmel,1993; Lawson et al.,1991; Okabayashi and Asashima,2003).

The pluripotency phenomenon is intensively studied in mammals, where the pluripotent cells of the embryo can be used to generate embryonic stem cells (ESC). The “Core transcription factors (TFs)” Oct4, Sox2, and Nanog control the expression of several thousand genes in ESC (Boyer et al.,2005; Chen et al.,2008a; Kim et al.,2008). The pivotal roles played by TF control in the establishment and maintenance of undifferentiated ESC is reinforced by the discovery that exogenous expression of the TFs, Sox2, Oct4, Klf4, and c-Myc, are capable of reprogramming mouse and human somatic cells into iPS cells (Takahashi and Yamanaka,2006).

In spite of a growing number of identified interactions of core TFs, the general principles of pluripotency network organization still remain a mystery. There is still very limited data about the dynamics of changes of transcriptional and protein–protein interaction landscapes of pluripotent cells during the phase when they differentiate to embryonic tissues. In part, this is due to poor accessibility of mammalian embryos at pregastrulation and gastrulation developmental stages. In this respect, the potential of lower vertebrate model organisms, where these stages are easily available, remains largely underemployed.

Genome-wide comparisons of pluripotency gene regulatory networks between different mammals show relatively high variability (Kunarso et al.,2010; Xie et al.,2010). Functional and genomic studies, comparing pluripotency gene regulatory networks of mouse and chick, point out to intensive restructuring of the regulatory links, co-option of new components and poor conservation of genomic regulatory elements between orthologous genes (Fernandez-Tresguerres et al., 2010; Canon et al., 2011). From the other hand, the similarity of chromatin signatures of ESCs and zebrafish early embryos (Vastenhouw et al.,2010), the critical role of core pluripotency factor Pou5f1 in the development of nonmammalian vertebrates, and successful cross-species rescue experiments argue for functional conservation of some mechanisms controlling pluripotency in vertebrates (Lavial et al.,2007; Morrison and Brickman,2006; Onichtchouk et al.,2010).

This review will cover the function of the core pluripotency TF Pou5f1/Oct4 in the embryonic development, with a focus on comparison of anamniotes, in particular, zebrafish, and mammals. The role of Pou5f1/Oct4 TF in the ES cells is not discussed in detail; therefore, we refer to several recent reviews, covering this subject (Chambers and Tomlinson,2009; Chen et al.,2008a; Chenoweth et al.,2010; Niwa,2010; Niwa,2007a,b).


The POU-domain gene family is divided into six classes of TFs, POUI-POUVI, which are involved in cellular decisions in multiple-cell lineages (Phillips and Luisi,2000; Veenstra et al.,1997). Pou5f1 is the single member of the POUV family. The closest relatives of the POUV family are POUIII family members, which are involved in the development of the neural system in vertebrates and invertebrates (Sasai,2001; Takemoto et al.,2006). Although POUIII classes are present in all metazoans (Hemmrich and Bosch,2008), the Pou5f1 gene is an evolutionary novelty. Pou5f1 is absent from invertebrates and primitive vertebrates, such as lancelets (Takatori et al.,2008), and first appears in the fish lineage; genomes of all five sequenced teleost species have one copy of the gene.

Pou5f1 homologues were characterized in birds (cPouV; Lavial et al.,2007), Xenopus (XlPou91, XlPou25, and XlPou60; Hinkley et al.,1992; Morrison and Brickman,2006), axolotl (Axoct4; Bachvarova et al.,2004), zebrafish (pou2; Takeda et al.,1994), and medaka [Oct4, (Sanchez-Sanchez et al.,2010)].

According to the current view, a common ancestor of jawed vertebrates had a single POUV class gene, syntenic to zebrafish pou5f1/pou2. This pou5f1/pou2-type gene was duplicated later in the evolution to give rise to co-orthologues called pou2 and Pou5f1/Oct4. Evolutionary time of duplication of POUV class gene in vertebrates was discussed in two recent works. Based on the analysis of platypus and opossum genomes, which contain both pou5f1 and pou2, Niwa et al. (2008) assigned the duplication event to early mammalian evolution and hypothesized that Pou5f1 is an adaptation to placental development in mammals. However, in depth analysis of synteny and sequence of all pou2/pou5f1-related genes in vertebrates revealed that the genomic duplication to pou2 and pou5f1 occurred at least in the early tetrapod lineage, long before establishment of mammals (Frankenberg et al.,2009, Fig. 1). Surprisingly, all vertebrates, except monotremes (Platypus, echidna), and marsupials (opossum, tammar wallaby) retained only one POUV class gene paralog in the genome, while the second one was selectively eliminated. This elimination occurred several times in the evolution: pou2 copy is present in anuran amphibians (all three copies of Xenopus POUV class gene XlPou91, XlPou25, and XlPou60) and birds, while only pou5f1 is present in urodelian amphibians (Axolotl), lizards and eutherian mammals (Frankenberg et al.,2009, Fig. 1).

Figure 1.

Evolutionary relationships of Pou5f1 homologues in vertebrates. Model depicting the evolution of class V POU domain genes in vertebrates [slightly modified from Frankenberg et al., (2009), with permission from Elsiever]. The ancestral Pou5f1/Pou2 gene (blue) duplicated to produce Pou5f1/Oct4 (red) after the divergence of the lineages giving rise to teleosts and tetrapods, respectively. Pou5f1/Pou2 became extinct (broken lines) independently in the lineages giving rise to lizards, urodeles, and eutherian mammals, respectively, while Pou5f1/Oct4 became extinct in lineages that led to birds and anurans, respectively.

In both evolutionary scenarios (Frankenberg et al.,2009; Niwa et al.,2008), the gene duplication occurred later than the separation of the fish lineage from a common ancestor. Therefore, the pou2 gene of fish should be considered as an orthologue of POUV class genes of all higher vertebrates, by definition of orthologous genes, as “genes originating from a single ancestral gene in the last common ancestor of the compared genomes” (Koonin,2005). In line with this strict definition, the zebrafish nomenclature committee has approved a single POUV class zebrafish gene with the name pou5f1 (RefSeq-ID NM_131112.1), which was previously called pou2.

Pou5f1 Expression and Loss-of-Function Phenotypes in Vertebrates

In vertebrates, Pou5f1 gene transcripts are already deposited during oogenesis in the egg and show broad ubiquitous expression during gastrulation stages until the end of gastrulation (Bachvarova et al.,2004; Belting et al., 2001; Burgess et al.,2002; Downs,2008; Lavial et al.,2007; Lunde et al.,2004; Morrison and Brickman,2006). Thus, in all organisms, Pou5f1 expression is associated with the cells at the pluripotency stages and during loss of pluripotency. Other expression domains of Pou5f1 genes were described for separate lineages. Pou5f1 in fish and mouse is also expressed in the neural plate until midsomitogenesis (Downs,2008; Reim and Brand,2002; Takeda et al.,1994). Expression in primordial germ cells is present in mouse (Kehler et al.,2004), where it is critically required for germ cell survival and also in chick (Lavial et al.,2007) and medaka (Sanchez-Sanchez et al.,2010), but not in zebrafish (Reim and Brand,2006) or Xenopus.

Pou5f1 loss-of-function studies in these vertebrate model organisms demonstrate that the gene is critically required for early embryonic development, although the loss-of-function phenotypes are different in zebrafish, Xenopus, and mouse. Zebrafish maternal and zygotic null Pou5f1 mutants, MZspg (spiel-ohne-grenzen; Burgess et al.,2002; Lunde et al.,2004; Reim et al.,2004) survive through gastrulation stages, although different aspects of development are severely affected (Fig. 2A). MZspg do not develop endodermal germ layer (Lunde et al.,2004; Reim et al.,2004), have and excess of dorsal tissues due to reduced BMP signaling (Belting et al.,2011; Reim and Brand,2006), and have cell motility and cytoskeletal defects leading to abnormal gastrulation (Lachnit et al.,2008; Reim and Brand,2006). Combination of these defects results in disorganized development and death of most of the embryos within the first 2 days after fertilization.

Figure 2.

Mouse Pou5f1 overexpression rescues the phenotype of zebrafish maternal and zygotic Pou5f1 mutant embryos (MZspg). (a) Embryonic phenotype (MZspg) at 24 h postfertilization (hpf). Note the absence of body axes and recognizable organs. (b) MZspg embryos that were injected with 25 pg mouse Pou5f1 mRNA at one cell stage at 24 hpf. Note that body axes, head, eyes, and tail formation are restored. (c) Wild-type control embryos are 24 hpf.

In Xenopus, three are three co-orthologues of Pou5f1 (XlPou91, XlPou25, and XlPou60, Hinkley et al.,1992), which arose by frog-specific duplication event. Loss-of function for all three co-orthologues of Pou5f1 (Morrison and Brickman,2006) or double knockdown of XlPou25 and XlPou60 (Cao et al.,2006) produces excess of endoderm, neutralization of the ectoderm, and gastrulation arrest. Effects on the mesodermal germ layer differ depending on knockout degree: block of the single gene, XlPou91 (Snir et al.,2006) or double block of XlPou25 and XlPou60 (Cao et al.,2006) results in expansion of the mesodermal marker Xbra to the ectodermal territories. In contrast, in triple knockdown embryos (Morrison and Brickman,2006) and upon dominant-negative XlPou25 overexpression, Xbra expression is downregulated.

Mouse Oct4-deficient embryos (Nichols et al.,1998) fail to accomplish the first differentiation event in development: separation of inner cell mass and trophectoderm. They develop to the blastocyst stage, but the inner cell mass cells abnormally differentiate along the extraembryonic trophoblast lineage. Niwa et al. (2002) has shown that reciprocal inhibition between lineage-specific TFs Oct4 and Cdx2 is necessary for proper trophectoderm specification. Similarly, mouse ES cells differentiate into trophectoderm upon inducible knockout of Oct4 (Niwa et al.,2002). Pou5f1 in mammalian development is critically required for primordial germ cell survival (Kehler et al.,2004) and controls mesendoderm differentiation through eomesodermin (Teo et al.,2011). Ubiquitous expression of Pou5f1 the epiblast until the day 5, five postcoitum (Downs,2008) suggests additional roles of Pou5f1 in the later embryonic stages.

Cross-species Rescue Assays of Pou5f1 Deficiency Suggest Conserved Functions at Gastrulation

Most families of TFs and signaling molecules that are critically involved in embryogenesis (“developmental genes”) have ancient evolutionary origin and were already present in the last common ancestor of bilaterians. The subsequent evolution of novel functions for these developmental genes involved addition of novel expression domains, so that proteins regulating development participate in multiple, independent developmental processes. Changes in the regulatory regions of developmental genes (changes in cis) occur much more often than changes in the amino acid sequence [changes in trans, see Carroll (2008) for the review]. The necessity to maintain the gene's original function is expected to constitute a strong constraint on the evolution of the coding regions: new protein domains may only evolve in parts of the amino acid sequence not required for the current function (Carroll,2008; True and Carroll,2002). Pou5f1 is different from typical developmental gene because of relatively late evolutionary origin and little sequence conservation outside of DNA-binding domain. TFs controlling development usually exhibit functionally equivalent activities in vivo when substituted for homologous proteins in divergent taxa (Carroll,2008; references therein). Dissimilarities of Pou5f1 loss-of-function phenotypes in different animals pose the question, if there is a functional equivalence between Pou5f1 homologues. This question was addressed by cross-species rescue assays in frogs, zebrafish, and mice.

Injection of mouse Pou5f1/Oct4 RNA at one cell stage embryos rescues severe phenotypic defects of zebrafish Pou5f1 null-mutant MZspg, which is devoid of both maternal and zygotic Pou5f1 protein (Onichtchouk et al.,2010; see Fig. 2A–C). Mouse Pou5f1 can also functionally replace Xenopus Pou5f1 genes (Cao et al.,2006; Morrison and Brickman,2006), compatible with the view on conserved functions at gastrulation stages.

Niwa et al. (2002) established ZHBTc4 complementation system in mouse ESCs. These ESCs undergo inducible knockout of the endogenous Oct4 gene upon addition of Tet and differentiate into trophectoderm. Transfection of exogenous mouse Oct4 (Pou5f1), but not Oct6 (Pou3f1), rescues the Tet-induced knockout (Niwa et al.,2002): transfected cells form pluripotent ES cell colonies. This functional complementation assay was used to evaluate the ability of different POU factors to rescue and maintain self-renewal and pluripotency in the absence of the mouse Pou5f1 gene. Human and Platypus Pou5f1 (Niwa et al.,2008) and Xenopus XlPou91 (Morrison and Brickman,2006) complement the system with high efficiency, while axolotl Pou5f1, Xenopus XlPou60, and XlPou25 (Morrison and Brickman,2006) and chick PouV (Lavial et al.,2007) produce a reduced number of ES cell colonies. As reported by Morrison and Brickman (2006), zebrafish Pou5f1 could not replace mouse Oct4 in this assay. More recently, Niwa et al. (2008) reported that, although zebrafish Pou5f1 worked poorly and produced only a small number of very small stem cell colonies, some of them expressed Nanog and Sox2 and propagated. Thus, limited, but significant functional overlap between mouse and zebrafish orthologues in this assay exists, considering that the nonclass V POU factors such as Pou1f1, Pou2f1, and Pou3f1 were unable to generate stem cell colonies in this rescue system (Niwa et al.,2008).

Mouse ES-cell complementation system tests for the ability to suppress trophectoderm differentiation, the novel function first appearing in mammalian lineage. Indeed, the first days of mouse development embryo are occupied by building extraembryonic layers, while after implantation, the processes of building the embryo proper start. Therefore, mouse preimplantation stages, used to derive ES cells, represent mammalian-specific adaptations to uterine development and have no direct evolutionary equivalent stages in lower vertebrate species (Nichols and Smith,2009). Thus, poor performance of zebrafish Pou5f1 in this system is rather expected, while the ability of XlPou91 to rescue self-renewal of stem cells is rather surprising. XlPou91, which is the paralog to mouse Pou5f1/Oct4, performs in ESC rescue assay better than Axolotl Pou5f1, direct ortholog of mouse Pou5f1/Oct4. As suggested by Niwa et al., (2008), this could be due to convergent amino acid changes in mouse and Xenopus Pou5f1, which, for example, create an interface for the novel protein-protein interaction, or ability to bind novel type of DNA sequences.

Molecular events involved in embryonic patterning and axis specification at pregastrulation stages show significant level of conservation of key regulatory processes between mammals and lower vertebrates (Arnold and Robertson,2009; Beddington and Robertson,1998,1999) and can be directly compared between zebrafish, Xenopus, and mice. The ability of mouse Pou5f1 to complement the Pou5f1 deficiency in the lower vertebrates argues that, in addition to well-known roles in ES cells and inhibiting trophectoderm differentiation, mouse Pou5f1 also retains the common ancestral functions in development.

Pou5f1 Genes, Embryonic Signaling Pathways, and Embryonic Competence

Embryonic induction provides a signal for multi- or pluripotent potent cells to develop toward one of several developmental pathways. Competence is the ability to respond to such a specific inductive signal (Waddington,1940). Competence for inductive signals is a temporal phenomenon in development: if the cell is not induced during a certain period of time, it will loose the competence and self-differentiate according to default mechanisms. Competence of pluripotent embryonic cells will be gradually restricted by acquirement of subsequent developmental choices and finally lost when the cells undergo terminal differentiation. In spite of major breakthroughs of molecular embryology, our understanding of the molecular basis of embryonic competence is still superficial. Pou5f1 genes were suggested to regulate competence to different signaling pathways in several embryonic contexts.

  • 1In zebrafish, Pou5f1 (spg/pou2) is expressed in the forming mid-hindbrain boundary and mediates the competence to respond to Fgf8 inductive signaling in this region. Mid-hindbrain boundary is missing in the zygotic spg mutants (Burgess et al.,2002; Reim and Brand,2002). Mouse Pou5f1/Oct4 is expressed at E8-8.5 throughout the neural plate (Downs,2008; Reim and Brand,2002), though the expression is not restricted to the midbrain–hindbrain domain as in zebrafish. Injection of synthetic mRNA for mouse Pou5f1/Oct4 to the one cell stage zygotic spg mutant zebrafish rescues the expression of TF pax2.1, which is required for mid-hindbrain boundary formation and is normally severely reduced in zygotic spg mutants (Burgess et al.,2002; Reim and Brand,2002). These results suggest that Pou5f1/Oct4 may function in activation of Pax2 also in normal mouse development.
  • 2In Xenopus development, two competence windows of response to FGF growth factor follow each other: at blastula stages, FGF induces mesoderm, but at gastrula stages FGF regulates neuroectoderm formation. Snir et al. (2006) investigated the role of XlPou91 in regulation of this transition. Although overexpression of XlPou91 inhibited FGF-mediated induction of mesodermal markers in the ectoderm, in ectoderm explants, devoid of XlPou91, mesoderm responsiveness to FGF was extended in time (from blastula to gastrula stages), while neural induction was abolished. Thus, competence switch from mesodermal to neural induction did not occur in XlPou91 morphants. Snir et al. (2006) further demonstrated that XlPou91 is necessary for the expression of Churchill (Chch) and Sip1, two genes regulating neural competence, and ectopic Sip1 or Chch expression rescues the XlPou91 morphant phenotype. Thus, XlPou91 epistatically lies upstream of Chch/Sip1 gene expression, regulating the competence transition that is critical for neural induction.
  • 3The BMP signaling pathway is required for specifying ventral fates during gastrulation stages of all vertebrates. Expression of BMP ligands and establishment of a BMP patterning gradient is positively regulated by Pou5f1 homologs in Xenopus (Cao et al.,2004; Morrison and Brickman,2006) and by Pou5f1 in zebrafish (Reim and Brand,2006). The mechanisms include conserved direct transcriptional activation; in Xenopus, XlPou25 induces ventral transcriptional repressor Xvent2 (Cao et al.,2004), in zebrafish, Pou5f1 directly induces the closely related ventral repressor vox (Belting et al.,2011). However, slightly later in development, XOct-25 (XlPou25) was suggested to play a role in the formation of neural tissue of Xenopus by inhibiting competence to BMP-mediated induction of epidermal markers (Takebayashi-Suzuki et al.,2007).
  • 4Endoderm formation: so far shown for zebrafish only, Pou5f1 is required for endoderm formation from mesendoderm; maternal-zygotic Pou5f1 mutants do not have endodermal precursors. Together with the zebrafish-specific transcription partner Casanova (Sox32), Pou5f1 is required to activate the endoderm-specification gene sox17. Epistatic analysis suggests a direct action at the level of sox17 promoter (Lunde et al.,2004; Reim et al.,2004). In contrast, in Xenopus, endoderm formation is increased upon knockout of all POUV factors (Morrison and Brickman,2006). Most likely, this difference between zebrafish and Xenopus may be explained not by species-specific differences in properties of Pou5f1 proteins, but by difference in structure of the endoderm-specifying networks.
  • 5Cao and colleagues (2006,2007,2008) demonstrated that Xenopus PouV class genes are able to antagonize beta-catenin (Wnt signaling), VegT, and nodal/activin embryonic signaling in Xenopus by forming repressive transcriptional complexes with downstream signal transducers at the promoters of the target genes. The authors conclude that Xenopus POUV proteins are required to control the levels of embryonic signaling pathways, thereby ensuring the correct specification of germ layers. Recently, another group confirmed the ability of Xenopus POUV proteins to downregulate endogenous Wnt (Abu-Remaileh et al.,2010), but the explanation for this effect is different from that of Cao et al. (2007). Abu-Remaileh et al., (2010) demonstrated that mouse Oct4 negatively regulates Wnt signaling by specifically interacting with nuclear beta-catenin and facilitating its proteasomal degradation, and suggested that Xenopus POUV proteins act by the same mechanism during development.

Although it is difficult to find a unifying principle of Pou5f1 gene action from these examples, it is clear that Pou5f1 regulates multiple targets, and, in the absence of Pou5f1 activity, the cues driving proper embryonic cell fates are lost. A bird's eye view onto the Pou5f1-regulated transcriptome of several species is needed to understand the complex circuitry of interacting pathways.

Zebrafish Pou5f1 Coordinates Developmental Programs in Time

Recently published time-resolved transcriptome analyses (Onichtchouk et al.,2010) provided an initial view into the Pou5f1-regulated transcriptome in zebrafish. Comparison of gene expression in zebrafish wild type and Pou5f1-deficient MZspg mutant embryos during the first 8 h of zebrafish development revealed time shifts in expression onset for more than 7,000 genes in MZspg mutants. Strikingly, Pou5f1 deficiency leads to premature expression of differentiation genes. Multiple genes, which function during organogenesis stages in the wild-type development, are already expressed during gastrula stage in MZspg embryos. This class of prematurely expressed genes, collectively termed “promoters of differentiation” (PODs), include orphan nuclear receptors (nr2f1, nr2f2, and nr2f5), Hox genes (hoxa2b and hoxb8a), neuron differentiation marker genes (her2, ascl1b, elavl3, and dlb), and genes encoding TFs involved in patterning of various tissues at post gastrulation stages, such as tbx2, gbx2, sox21, and pax6a. As revealed by experiments using the suppression of translation of primary target gene mRNAs, Pou5f1 delays the expression of “PODs” indirectly, activating sets of transcriptional repressors (repressors of differentiation, RODs; Fig. 3A,B). Transcriptional factor genes with repressory function, induced by Pou5f1, include foxD3, klf2b, klf4, her3, and hesx1, which are evolutionary conserved targets of Pou5f1 from fish to mammals.

Specificity of RODs toward their downstream targets needs further investigation. On one hand, forced overexpression of Her3 or Hesx1 transcriptional repressors could suppress the premature expression of neural PODs nr2f1, pax6a, and sox21b in the MZspg background (Onichtchouk et al.,2010). On the other hand, injections of her3 or hesx1 morpholinos or their combined loss of function did not cause detectable premature expression of any of these PODs (Leichsenring, Onichtchouk, Driever, unpublished). Most likely, multiple Pou5f1-dependent repressors are involved in keeping PODs silent during the gastrulation time window.

Pou5f1 Activates Transcriptional Repressors With or Without Partnering with SoxB1 Genes

Together with Pou5f1, Sox2, a Sox B1 class TF, plays a central role in maintenance of pluripotency in mammalian ES cells (Masui et al.,2007). SoxB1 class members have been implicated in various processes of zebrafish embryogenesis (Okuda et al.,2006), but their role in pluripotency stages was until recently masked by redundancy of their functions. Quadruple knockdown of the four B1 sox genes expressed in early embryogenesis, sox2/3/19a/19b, results in a pleiotropic phenotype, with abnormal dorso-ventral patterning, regionalization of the nervous system, and gastrulation movements (Okuda et al.,2010). The list of SoxB1 target genes with reduced or absent expression in quadruple knockdown embryos (Okuda et al.,2010) highly overlaps with the list of genes with reduced or absent expression in Pou5f1 MZ mutant embryos (Onichtchouk et al.,2010), suggesting transcriptional partnering between two Pou5f1 and SoxB1 genes during activation of a wide range of targets.

The loss-of function phenotypes of MZspg embryos and SoxB1 knockout are not identical, pointing out that these genes have both separate and overlapping functions in early embryogenesis. The levels of Pou5f1 at 0–8 hpf are relatively constant and not regulated by SoxB1 activity. In contrast, most of SoxB1 activity present in the zebrafish embryo is under control of Pou5f1 (Onichtchouk et al.,2010). SoxB1 activity is present already during the maternally controlled period of development and dynamically grows between 3 and 8 hpf (Fig. 4B, SoxB1).

Figure 3.

Pou5f1 activates repressors of differentiation to suppress premature expression of PODs. (a) Wild-type embryo (WT). Scheme at the left: Pou5f1 activates repressor of differentiation (ROD) to prevent premature expression of pax6a. Schematic temporal expression profiles for Pou5f1 (red), ROD (green), and pax6a (violet) shading. Below the scale: pax6a expression visualized by whole-mount in situ hybridization at the indicated time points (8 and 10 hpf). (b) MZspg embryo (maternal and zygotic Pou5f1 mutant). Scheme at the left: at the absence of Pou5f1 ROD is not activated and pax6a is expressed prematurely. Schematic temporal expression profile for pax6a (violet shading), which shows a temporal shift, indicated by arrow. Below the time scale: pax6a expression visualized by whole-mount in situ hybridization at the indicated time points (8 and 10 hpf). Embryos are shown in anterior view, hpf- hours postfertilization. Note that the spatial position of the pax6a premature expression domain at 8 hpf is roughly similar to its localization at 10 hpf in MZspg and in the WT.

Figure 4.

Pou5f1 partners with SoxB1 on direct transcriptional target genes—repressors of differentiation (RODs). (a) Schematic drawing of midgastrula (shield stage) embryo, with expression domains of selected Pou5f1—only target RODI in the mesoderm (foxD3), and double Pou5f1 and SoxB1 targets RODII in the non-neural (klf2b, gata2) and neural (hesx1, her3) parts of the ectoderm. Simplified interaction chart of Pou5f1 and SoxB1 TFs; arrow, direct transcriptional activation. (b) SoxB1 activity levels define expression onset of RODs II, but not RODI. Schematic drawing of mRNA expression profiles for RODI foxD3 and RODII hesx1 and her3. Growing SoxB1 activity levels are shown by orange shadow. Pou5f1 activity is present and constant (not shown). Dark and light blue arrows show the time points, when SoxB1 activity reaches the threshold levels necessary to switch on hesx1 and her3 activation, respectively. FoxD3 expression does not depend on SoxB1 activity and is switched on immediately after MBT (3 hpf). Hpf- hours postfertilization.

Although comparisons of Pou5f1 and SoxB1-dependent transcriptomes revealed large number of common targets, this overlap is not complete. Majority of double Pou5f1 and SoxB1 transcriptional targets, with putative transcriptional repressory activity, are located in prospective neural (her3 and hesx1) or non-neural ectoderm (klf2b and gata2a) in the midgastrula embryo (RODII, Fig. 4A). For mesendodermal target genes of Pou5f1 (RODI in Fig. 4A), activation by SoxB1 is indirect or it was not detected (Onichtchouk et al.,2010).

Analyzing the temporal behavior of Pou5f1 transcriptional targets in development, we could build a correlation between the shape of temporal profile and dependency on SoxB1. Most SoxB1-independent direct Pou5f1 targets reached half-maximal expression levels soon after the start of zygotic transcription (foxD3 in Fig. 4B). Double direct SoxB1- and Pou5f1-dependent genes tend to have biphasic or delayed expression and reach maximum levels at 6–7 hpf [hesx1 and her3 in Fig. 4B (Onichtchouk et al.,2010)]. A dynamic mathematical model, based on an interaction chart schematically shown in Figure 4A, was validated in vivo. Analysis of the model parameters led to the conclusion that the level of SoxB1 activity exerts the main temporal control of biphasic targets [blue and light blue-dotted lines in Fig. 4B show SoxB1 levels, which are necessary to activate hesx1 (shortly before 5 hpf) and high level expression of her3 (shortly after 6 hpf)]. Timing of the expression of double Pou5f1 and SoxB1-dependent targets is relatively independent of the amount of Pou5f1, as long as it is present above certain threshold. Indeed, most of maternal mutant (M) spg embryos and MZ embryos rescued by Pou5f1 mRNA injection develop normally, despite the vastly different Pou5f1 activity levels in these embryos. This situation apparently differs from the mode of Pou5f1 action in ES cells, where the levels of Pou5f1/Oct4 need to be precisely tuned (Niwa et al.,2000,2002). However, the situation in the later stages of mammalian development, that is, in the pluripotent epiblast cells after implantation, may be different and it would be interesting to test, whether the zebrafish model regulatory scenario fits to the expression of double Pou5f1 and SoxB1 targets in these systems.

Sox-Oct Motives in ES Cells and Development

High correlation between Pou5f1 and SoxB1 targets in the developing embryo suggests that SoxB1 and Pou5f1 TFs may cooperate during large-scale coordination between various early developmental processes. SoxB1 and Pou5f1 proteins activate transcription by cooperatively binding to DNA on Sox-Oct composite elements. These elements consist of adjacent Sox2- (5′-CATTGTT-3′) and POU-binding sequences (5-′ATGCAAAT-3′; Kondoh and Kamachi,2010). Sox-Oct motives, matching the mammalian consensus sequence (Loh et al.,2006), are significantly enriched in the proximity of Pou5f1 direct target genes of zebrafish (Onichtchouk et al.,2010). Pou5f1-dependent conserved Sox-Oct elements were found to be critical for early expression of sox2 (Iwafuchi-Doi et al.,2011) and her3 (Onichtchouk et al.,2010) and are also bound by Pou5f1 in the promoters of the vox (Belting et al.,2011) and hesx1 (Okuda et al.,2010).

In mouse ES cells, profiling multiple TF-binding sites on the whole-genome scale revealed extensive colocalization of multiple TFs around Sox-Oct motives. These sites bind core TFs Nanog, Oct4, Sox2, and Klfs and recruit the transcriptional co-activator p300, which is strong evidence for active enhancers in ES cells (Chen et al.,2008a; Kim et al.,2008). Furthermore, Sox-Oct consensus sites are co-occupied with downstream components of the LIF, BMP (Chen et al.,2008b), and Wnt (Cole et al.,2008) signaling pathways. Thus, in ES cells, the Sox-Oct consensus module is suggested to integrate the input of multiple TFs and signaling pathways, adding developmental signals to the core regulatory circuitry of ES cells to directly influence the balance between pluripotency and differentiation.

Do Sox-Oct modules play a similar widespread role in pregastrula and gastrula stages of lower vertebrates? Are the inputs of developmental signaling pathways integrated on Sox-Oct sites? Answer to these question wait for whole-genome scale chromatin immunoprecipitation (Chip-Chip or CHIP-seq) for SoxB1 and Pou5f1.

Pou5f1 Activates Tissue-Specific Repressors of Differentiation: Is This an Evolutionary Basal Feature?

Pou5f1 is expressed throughout the whole embryo during early zebrafish developmental stages. However, the majority of its direct transcriptional targets are compartmentalized at midgastrulation. Evolutionary conserved RODs are expressed in defined territories: foxD3 in mesendoderm, her3 and hesx1 in neural ectoderm, and klf2ab and klf4 in non-neural ectoderm (Fig. 4A). We hypothesized that Pou5f1 downstream transcriptional network is subdivided to smaller sub networks; in each of those, tissue-specific repressor sets prevent premature expression of PODs (Onichtchouk et al.,2010).

Significant overlap between zebrafish and mammalian Pou5f1 transcriptional targets, cross-species rescue assays, and conservation of RODs (Onichtchouk et al.,2010) makes it possible to speculate that the mammalian stem cell network may have evolved from a basal situation similar to the one observed for teleosts. This would mean that in addition to the later evolved functions in naive pluripotency stages, the Pou5f1 network in mammals may control tissue-specific modules, repressing premature differentiation, at pregastrulation and gastrulation stages.

If this tissue-specific principle underlies the elaborated network in ES cells, loss-of function of conserved RODs in ES cells may result in the tissue specificity of differentiation. FoxD3 loss would, for example, result in abnormal differentiation into mesodermal derivatives, while loss of Klfs would promote differentiation towards ectoderm. Two recent publications support this idea. (1) In FoxD3, knockout ES cells expression levels of meso- and endodermal markers increased, indicating selective differentiation toward this lineage, while neuroectoderm markers either decreased or remained the same (Liu and Labosky,2008). (2) Triple knockdown of Klf2, Klf4, and Klf5 in ES cells results in the upregulation of ectodermal differentiation markers in the earliest time point, suggesting that Klfs suppress ectoderm-specific differentiation in ES cells (Jiang et al.,2008).

Although the zebrafish network supports short-lasting pluripotency in development, mammalian network can be brought to the stable-state, allowing indefinite pluripotency maintenance in ES cells. This important difference between mammalian and zebrafish networks may be explained biologically, as an adaptation to much slower intrauterine development in mice, where the cells need to be kept pluripotent for at least 5 days. Mechanistically, the stability of the mammalian system may be supported by higher interconnectivity between the components, including multiple positive feedback loops (Som et al.,2010), which are absent in zebrafish.


Transcriptional landscapes of different pluripotent stem-cell populations appear widely diverse, as shown by comparisons of mouse and human ES cells (Kunarso et al.,2010; Loh et al.,2006), comparisons of preimplantation developmental networks between three mammalian species (Xie et al.,2010), and comparisons of early developmental stages of different amniotes (Fernandez-Tresguerres et al., 2010; Canon et al., 2011). The alternative designs of the networks may be attributes to little evolutionary conservation during the early developmental stages of amniotes, which are occupied by generation of the extraembryonic cell lineages. At gastrulation stages, when essential aspects of developmental control are driven by similar mechanisms in all vertebrates, the transition from pluripotency to defined cell states may also use evolutionary conserved architectures.

Mechanisms controlling pluripotency are intensively studied using steady-state mammalian cell-culture systems, which make it difficult to understand the integration of pluripotency control mechanisms in the context of developmental programs acting in the embryo. In contrast to relatively inaccessible embryos of placental mammals, in zebrafish, early stages of development can be easily analyzed and efficiently manipulated experimentally. Zebrafish data suggest regulatory linkages of Pou5f1 transcriptional network with the establishment of the earliest cell fate choices, which divide the embryo to mesendodermal, ectodermal, and neural precursors. Gain- or loss-of-function studies for “repressors of differentiation” complemented by transcriptomic experiments will allow characterizing the network further. Perturbing the critical components of Pou5f1 downstream transcriptional network would be expected to affect timing or reversibility of cell commitment to particular embryonic cell fate and can be tested by isochronic and heterochronic cell transplantations (Ho and Kimmel,1993).

Comparisons of transient pluripotency networks between the vertebrate organisms, and understanding their evolution, should highlight the conserved determinants, which can be difficult to distinguish using other methods and reveal general principles of pluripotency network organization.


I am grateful to Wolfgang Driever, Miguel Manzanares, Corinne Vannier, Kay Kottkamp, Stephanie Eckerle, Sungmin Song, Rainer Duden, and anonymous referees for reading and commenting on this review.