Pou5f1 Transcription Factor Controls Zygotic Gene Activation In Vertebrates

See allHide authors and affiliations

Science  30 Aug 2013:
Vol. 341, Issue 6149, pp. 1005-1009
DOI: 10.1126/science.1242527

Pluripotency Control

The transcription factors Pou5f1/Oct4, Sox2, and Nanog play central roles in pluripotency control in mammalian embryonic stem (ES) cells. The evolution of the pluripotency regulatory network and its roles during early development of nonmammalian vertebrates is unknown. Leichsenring et al. (p. 1005, published online 15 August) show that in zebrafish embryos, Pou5f1 controls priming and transcriptional activation of the first zygotically expressed genes. This mechanism for transition from the transcriptionally silent cleavage stage to the transcriptionally active blastula stage may have evolved to control the prolonged cell pluripotency state in mammalian early development and ES cells, establishing a link between zygotic gene activation and pluripotency control.


The development of multicellular animals is initially controlled by maternal gene products deposited in the oocyte. During the maternal-to-zygotic transition, transcription of zygotic genes commences, and developmental control starts to be regulated by zygotic gene products. In Drosophila, the transcription factor Zelda specifically binds to promoters of the earliest zygotic genes and primes them for activation. It is unknown whether a similar regulation exists in other animals. We found that zebrafish Pou5f1, a homolog of the mammalian pluripotency transcription factor Oct4, occupies SOX-POU binding sites before the onset of zygotic transcription and activates the earliest zygotic genes. Our data position Pou5f1 and SOX-POU sites at the center of the zygotic gene activation network of vertebrates and provide a link between zygotic gene activation and pluripotency control.

In early metazoan development, the zygotic genome is not immediately transcribed; instead, factors expressed during oogenesis from the maternal genome control development. The controlled synchronous onset of expression of the earliest large wave of zygotic transcripts is termed zygotic genome activation (ZGA). In vertebrates, ZGA is triggered by the change of nuclear-to-cytoplasmic ratio during early cleavage stages, as well as by other genome-level mechanisms (1). ZGA is one component of the maternal-to-zygotic transition (MZT) period, which also includes the gradual degradation of maternal transcripts (2). In Drosophila, the transcription factor Zelda (Zld) selectively primes the earliest zygotic genes for activation at MZT (38) and controls ZGA. Because no homologs of Zld have been reported outside the insect clade, it is unclear whether similar specific control of transcription onset exists in early vertebrate development.

In zebrafish, zygotic transcription starts during the midblastula transition (MBT), after the 10th cell division (9). Pou5f1 (also named Pou2 or Pou5f3;, Nanog, and the functionally redundant SoxB1 group of transcription factors (Sox2, Sox3, Sox19a, and Sox19b) are ubiquitously present in the zebrafish egg and early embryo (1013). Their homologs are considered core transcription factors of the mammalian embryonic stem (ES) cell state. In mammalian ES cells, the Pou5f1-Sox2 complex cooperatively assembles to target gene promoters containing bipartite SOX-POU binding sites (1416). In zebrafish, loss of Pou5f1 function leads to severe disturbance of developmental progress immediately after MBT (1720). Time-resolved expression analysis of MZspg mutants, devoid of both maternal and zygotic Pou5f1 function, revealed delays and desynchronization in expression of hundreds of genes (21), suggesting a role for Pou5f1 in temporal control of development. Here, we show that Pou5f1 in zebrafish selectively primes the earliest zygotic genes for activation, providing functionality similar to Drosophila Zelda (38).

To detect in vivo Pou5f1 and Sox2 chromatin binding events, we performed chromatin immunoprecipitation followed by parallel sequencing (ChIP-seq) and identified 7747 Pou5f1-bound regions at the pre-MBT 512-cell stage [2.75 hours postfertilization (hpf)], as well as 6670 Pou5f1-bound and 5924 Sox2-bound regions at the post-MBT stage (5 hpf; figs. S1 and S2 and table S1). Post- and pre-MBT Pou5f1 and Sox2 binding sites tended to colocalize (Fig. 1, A and D). To determine whether Pou5f1- or Sox2-bound genes correspond to targets of Pou5f1 and SoxB1 transcriptional regulation, we compared the list of genes bound by Pou5f1 or Sox2 within 20 kb upstream of their transcription start site (TSS) with lists of genes differentially expressed in embryos deficient for the respective transcription factors (13, 21). More than 100 genes activated by Pou5f1 and SoxB1 were directly linked to Pou5f1 and Sox2 binding regions, whereas no correlation with repressed genes was observed (fig. S3, A and B, table S2, and supplementary text).

Fig. 1 Characterization of Pou5f1 and Sox2 binding regions and target genes.

(A) Region map shows colocalization of Pou5f1 post-MBT ChIP-Seq peaks with pre-MBT Pou5f1, post-MBT Sox2, and post-MBT Nanog peaks; * from (10). Analyzed 5-kb regions are centered on Pou5f1 peaks 320 base pairs long. (B and C) Top-scoring Pou5f1 (B) and Sox2 (C) zebrafish motifs compared with the respective mammalian motifs. (D) Colocalization of Pou5f1 and Sox2 at the regulatory regions of pou5f1, sox2, and nanog; the y axis indicates ChIP-Seq reads per base.

We identified the top-scoring DNA binding motifs de novo from Pou5f1 and Sox2 ChIP-seq data. The calculated binding position weight matrices (PWMs) were bipartite, with SOX and POU binding halfsites. The post-MBT PWM closely resembled mammalian Oct4-Sox2 PWMs (Fig. 1, B and C, and table S3). However, the SOX halfsite in the PWM derived from pre-MBT Pou5f1 ChIP-seq was less pronounced (Fig. 1B and table S3), which suggests that Pou5f1 may use a broader range of transcriptional partners before MBT. Gene Ontology and expression annotations of genes located within 20 kb of pre- and post-MBT Pou5f1 and Sox2 ChIP-seq peaks were enriched for developmental genes, signaling pathway components, and factors with early regionalized expression (fig. S3C). Pou5f1 and Sox2 were bound to the regulatory regions of many orthologs of known targets of mammalian Pou5f1 and Sox2 in ES cells (1416) including pou5f1, nanog, sox2, and sox3 (Fig. 1D, fig. S4, fig. S5, A and B, table S4, and supplementary text).

Simultaneous and steeply increasing transcription of the earliest zygotic gene set is characteristic for ZGA (22, 23). To test whether Pou5f1 or SoxB1 may control zebrafish ZGA, we investigated whether post-MBT Pou5f1 and Sox2 sites are specifically enriched in the regulatory regions of the earliest-expressed zygotic genes. To define the earliest zygotic gene group, we clustered our previously obtained expression data (21) into 30 groups defined by changes of expression during normal development: five 1-hour time windows (3 to 8 hpf), each further divided according to six criteria for relative change in expression (up-regulated by a factor of >5, 2 to 5, or <2; down-regulated by a factor of <2, 2 to 5, or >5; fig. S6A). We defined 162 genes that were up-regulated by a factor of >5 between 3 and 4 hpf as belonging to the “ZGA group” (Fig. 2A and table S5). We compared the fraction of genes close to Pou5f1 and Sox2 sites in each of the 30 above-defined expression groups. Pou5f1 and Sox2 binding to ZGA group genes exceeded the average of all groups by a factor of 4 (Fig. 2B). Using the MBT expression gene group independently defined by Aanes et al. (23), we confirmed our results obtained for the ZGA group genes (fig. S7). Thus, Pou5f1 and Sox2 preferentially bind to regulatory regions of genes most steeply activated at ZGA.

Fig. 2 Pou5f1 activates the earliest zygotic genes.

(A) Explanatory chart visualizing the classification of transcripts by expression change per 1-hour time window (the y axis shows arbitrary units). Six exemplified expression profiles are based on level and direction of regulation in the 3- to 4-hpf time window; 162 genes up-regulated by a factor of >5 were defined as ZGA group. (B) Percentage of Pou5f1 and Sox2 targets among the 30 gene groups clustered by expression change per time window. (C to E) Relative expression levels of ZGA group genes (top) and 6- to 7-hpf group up-regulated by a factor of 5 (bottom) in microarray experiments. Each row represents one microarray probe, each column one experimental condition. Genes were sorted according to ChIP-seq Pou5f1 or Sox2 binding, as shown by blue (post-MBT Pou5f1), red (Sox2), and green (pre-MBT Pou5f1) vertical lines. Maximal expression level was set to 10 (color code at right). (C) Pou5f1, but not Sox2, induces expression of ZGA group genes in MZspg. (D) ZGA genes are up-regulated in Mspg relative to MZspg. (E) ZGA genes are up-regulated in wild type (WT) relative to Mspg. (F and G) SoxB1 QKD phenotype (13) in WT (F) and MZspg (G). Scale bars, 0.1 mm. Leftmost column shows whole-mount in situ hybridization for klf2a at 4.7 hpf, animal view. Other columns show live embryos at indicated ages; lateral views, animal pole up. White dotted line, epiboly border; white arrow, free yolk; *, position of tailbud.

To address the functional importance of Pou5f1 and SoxB1 for ZGA, we analyzed expression changes in ZGA group genes in response to Pou5f1 and Sox2 overexpression. For identification of directly regulated targets, post-MBT translation was inhibited with cycloheximide (CHX). In MZspg mutants, Pou5f1 overexpression strongly activated most ZGA group genes, whereas Sox2 overexpression alone did not (Fig. 2C). However, in wild-type embryos with endogenous Pou5f1 present, Sox2 overexpression hyperactivated most ZGA group genes (fig. S6). In the other 29 expression groups, no global expression changes were observed upon Pou5f1 overexpression (Fig. 2C). To investigate whether endogenous Pou5f1 is necessary for ZGA, we analyzed expression changes in maternal-only (Mspg) and maternal-zygotic (MZspg) Pou5f1 mutants. Mspg mutants fertilized with wild-type sperm started endogenous Pou5f1 expression at 3 hpf. Most of the ZGA group genes at 4 hpf already had higher expression levels in Mspg than in MZspg embryos (Fig. 2D). However, Mspg embryos showed lower levels of expression of ZGA group genes than wild-type embryos (Fig. 2E), revealing that maternally derived Pou5f1 available pre-MBT contributes to this activation. Similar increases were not observed in the other 29 groups (Fig. 2, D and E, bottom), which demonstrates that Pou5f1 is selectively required pre- and post-MBT for fast activation of most ZGA genes.

To test for overlapping functions of Pou5f1 and SoxB1, we compared phenotypes of SoxB1 quadruple morpholino knockdown (QKD) (13) in wild-type, MZspg (Fig. 2, F and G, and fig. S8, A to D), and Mspg (fig. S8, E and F, and supplementary text) mutants. Expression onset of the SoxB1 and Pou5f1 combined ZGA group target klf2a (table S5) was retarded in MZspg SoxB1-QKD relative to MZspg or wild-type SoxB1-QKD embryos (Fig. 2, F and G). Thus, SoxB1 and Pou5f1 both are required for steep up-regulation of klf2a expression. However, SoxB1-QKD did not worsen the MZspg or Mspg embryonic phenotype (Fig. 2G and fig. S8, E and F) at stages preceding gastrulation. These findings indicate that although SoxB1 transcription factors are necessary for fine-tuning expression onset of certain genes, they have no Pou5f1-independent functions essential for embryonic survival during the first several hours after MBT.

The steep ZGA gene group up-regulation suggests involvement of additional mechanisms preparing for robust activation. Recent studies have revealed several candidate mechanisms, including recruitment of RNA polymerase II (Pol II) and modification of the chromatin landscape at regulatory regions before the actual start of transcription [reviewed in (24)]. To investigate whether these features occur at Pou5f1 and Sox2 binding regions, we compared positions of pre- and post-MBT Pou5f1 and post-MBT Sox2 binding with pre-MBT, MBT, and post-MBT epigenetic chromatin modification data (25) (Fig. 3A, fig. S9, tables S6 to S8, and supplementary text).Trimethylated histone H3 Lys4 (H3K4me3) active chromatin marks and Pol II are enriched at Pou5f1 and Sox2 binding regions before MBT and during later stages (Fig. 3 and fig. S10). Genes premarked with Pol II before MBT at Pou5f1 and Sox2 binding regions (table S9) encode zygotic proteins with key regulatory functions in the early embryo (table S10), including components of major signaling pathways such as bmp2b, dkk1, sdf2, lefty1, and lefty2. Because Pol II does not have its own DNA-binding specificity, it may be recruited to SOX-POU binding sites by maternal Pou5f1 protein.

Fig. 3 Pou5f1 and Sox2 binding regions are already enriched in H3K4me3 and Pol II before MBT.

(A) Pre- and post-MBT Pou5f1, post-MBT Sox2, Pol II binding, and H3K4me3 modifications at four loci. ChIP-seq data for Pou5f1 and Sox2 are shown in reads per base. For H3K4me3 and Pol II, log2 Ma2C signal is shown for stages indicated; data from (25). E, enhancer; P, promoter; x axis indicates distance (kb) from TSS. (B to D) Distribution of Pol II across genomic regions centered on the Pou5f1 pre-MBT peaks (green), Pou5f1 post-MBT peaks (blue), Sox2 peaks (red), or background (gray), at indicated stages (analysis in 50-bp steps; shading denotes SD of eight neighboring regions).

High co-occupancy with multiple transcription factors is a characteristic feature of Zelda binding sites in Drosophila (6, 8). Therefore, we investigated whether binding of other transcription factors may also be facilitated at SOX-POU binding regions. We examined co-occupancy of post-MBT Pou5f1 and Sox2 sites with Nanog (10) and three germ layer–specific developmental transcription factors (Mxtx2, Smad2, and Ntla) using published zebrafish ChIP-seq or ChIP-on-chip post-MBT data (10, 26, 27). Similar to their orthologs in ES cells, Pou5f1, Sox2, and Nanog tended to colocalize at the same sites (Fig. 1A, fig. S5, C to F, fig. S11, A and B, and supplementary text). Binding regions of Mxtx2, Smad2, and Ntla also colocalized with those of Pou5f1 and Sox2 (fig. S11, C to E, and table S1). To analyze the contribution of Pou5f1 and SoxB1 to transcription of Ntla target genes, we combined the published Ntla gene regulation network (26) with Pou5f1 and Sox2 binding and loss-of-function data (13, 21). Pou5f1 or Sox2 bound to promoters or enhancers of most Ntla targets (fig. S11 and table S1). The input of Pou5f1 expression was essential for 11 of 28 Ntla target genes in the network; similarly, the input of SoxB1 expression was essential for 6 of the 28 Ntla target genes (fig. S12).

The emerging zebrafish ZGA mechanisms are remarkably similar to those in Drosophila. Zelda binds to specific motifs in the enhancers of early-expressed developmental control genes (35) to facilitate binding of other transcription factors to these sites by keeping chromatin in the open state (6, 7). Zelda in combination with patterning factors coordinates temporal and spatial expression of early gene batteries (68). Homologs of Zelda or transcription factors with similar function have not been reported outside the insect clade. Our results reveal that Pou5f1 functions in a similar fashion as a key ZGA activator involved in priming and activation of the earliest zygotic genes (Fig. 4). At early post-ZGA stages, SoxB1 transcription factors co-occupy SOX-POU binding regions together with Pou5f1 and feed into regulatory cascades involved in regional specification of the embryos, ensuring that multiple genes have precisely timed transcription starts and sufficient levels of activation (21).

Fig. 4 Priming-activation-timing model of ZGA activators during early development of zebrafish.

Before ZGA, Pou5f1 binds to the SOX-POU “priming sites” and attracts Pol II to ensure robust activation at ZGA. After ZGA, SoxB1 proteins co-occupy SOX-POU sites; both factors cooperate with developmental regulators to ensure precise transcriptional timing and proper level of expression.

The ancestral roles of the mammalian pluripotency control factors Pou5f1 and Sox2 during the early development of nonmammalian vertebrates have long been a mystery (28). Our results show that the composition of post-MBT Pou5f1 and Sox2 binding sites, co-occupancy with Nanog, their chromatin state, and Pol II binding are similar in zebrafish embryos and mammalian ES cells (fig. S13 and supplementary text). Thus, the ancestral function of the pluripotency factors is zygotic gene activation and developmental timing control in the early vertebrate embryo. In a sense, it is a first major reprogramming event from transcriptionally silent cleavage-stage cells to pluripotent post-MBT blastomers. In this context, Pou5f1 and Sox2 contribute to all main embryonic regulatory pathways. Considered together with the orthology of target gene sets, the zygotic priming-activation-timing mechanism may have evolved to control the cell pluripotency state in mammalian development.

Supplementary Materials

Materials and Methods

Supplementary Text

Figs. S1 to S13

Tables S1 to S10

References (3056)

References and Notes

  1. Acknowledgments: We thank S. Shvartsman for scientific discussion; S. Arnold, G. Pyrowolakis, V. Taylor, A. Fuchs, S. Eckerle, and M. Fernandes for comments; B. Wendik and J. Paddeken for participating in initial stages of this work; Y. Kamachi for Sox3/19a/19b antibody; L. Yampolsky for help with statistics; and S. Götter for fish care. Supported by Deutsche Forschungsgemeinschaft grants DFG-EXC294 and DFG-SFB592 (W.D. and D.O.) and Bundesministerium für Bildung und Forschung grant BMBF-0313921FRISYS (W.D.). Microarray and ChIP-Seq data GEO ( accession numbers: CHIP-seq data, GSE39780; microarray data, GSE41107, GSE17655, and GSE17656.
View Abstract

Stay Connected to Science

Navigate This Article