|Home | About | Journals | Submit | Contact Us | Français|
An emerging concept in development is that transcriptional poising pre-sets patterns of gene expression in a manner that reflects a cell’s developmental potential. However, it is not known how certain loci are specified in the embryo to establish poised chromatin architecture as the developmental program unfolds. We find that, in the context of transcriptional quiescence prior to the midblastula transition in Xenopus, dorsal specification by the Wnt/β-catenin pathway is temporally uncoupled from the onset of dorsal target gene expression, and that β-catenin establishes poised chromatin architecture at target promoters. β-catenin recruits the arginine methyltransferase Prmt2 to target promoters, thereby establishing asymmetrically dimethylated H3 arginine 8 (R8). Recruitment of Prmt2 to β-catenin target genes is necessary and sufficient to establish the dorsal developmental program, indicating that Prmt2-mediated histone H3R8 methylation plays a critical role downstream of β-catenin in establishing poised chromatin architecture and marking key organizer genes for later expression.
Transcriptional poising represents a widespread mechanism of post-initiation control of gene expression that is observed in metazoan biological model systems [see (Margaritis and Holstege, 2008; Saunders et al., 2006) for reviews]. Establishment of poised chromatin architecture at genetic loci allows for a rapid and synchronous transcriptional response to environmental and biological stimuli (Baugh et al., 2009; Boettiger and Levine, 2009; Hargreaves et al., 2009; Muse et al., 2007; Radonjic et al., 2005; Rougvie and Lis, 1988). Poised loci have undergone successful pre-initiation complex formation, yet are stalled at the transition from transcriptional initiation to elongation (Saunders et al., 2006). Thus, they are marked by covalent histone modifications (acetylation of lysine 9 and 14, and trimethylation of lysine 4 on Histone H3, H3K9/14ac and H3K4me3, respectively) and a phosphorylated form of the large subunit of the RNA Polymerase holoenzyme (Pol II CTDpSer5) that correlate with transcriptional initiation prior to the onset of mRNA expression (Guenther et al., 2007; Margaritis and Holstege, 2008). Remarkably, in the context of embryonic development, poised chromatin architecture is established within multipotent precursor cells in a manner that reflects the developmental potential of the lineage (Bernstein et al., 2006; Guenther et al., 2007; Hammoud et al., 2009; Vastenhouw et al., 2010; Zeitlinger et al., 2007). However, it is not well understood how certain loci are specified to establish poised chromatin architecture as the developmental program unfolds.
The earliest events in embryogenesis are controlled by maternal factors until the activation of the zygotic genome. In Xenopus, Drosophila, and Zebrafish, zygotic genome activation occurs several hours and cell divisions after fertilization, at the midblastula transition (MBT) (Edgar and Schubiger, 1986; Kane and Kimmel, 1993; Newport and Kirschner, 1982). However, while zygotic transcription is constrained before the MBT, essential steps in embryonic patterning are accomplished before the MBT and embryos emerge from this period having begun the process of regional specification. In particular, the Wnt/β-catenin pathway mediates the earliest cell fate decision in amphibian (and teleost) embryogenesis, the establishment of the dorso-ventral axis. Dorsal specification by the Wnt/β-catenin pathway takes place under conditions of global transcriptional repression, prior to the MBT (Heasman et al., 2000; Kao et al., 1986; Yamaguchi and Shinagawa, 1989; Yang et al., 2002b). While β-catenin is required for the transcription of a small set of genes that are expressed before the MBT (Takahashi et al., 2000; Yang et al., 2002b), the critical Wnt target genes that direct dorsal development are silent until the MBT. Notably, β-catenin can interact with numerous factors that direct both chromatin modification and RNA Pol II recruitment to promoters [reviewed in (Mosimann et al., 2009)], including factors that establish both H3K9/14ac and H3K4me3. These observations raise the possibility that β-catenin functions during the preMBT period to establish a heritable, transcriptionally poised state that results in the later expression of dorsal determinants such as siamois and xnr3.
We have investigated the chromatin architecture of β-catenin target genes before the MBT, and report that β-catenin contributes to the establishment of poised chromatin architecture, thus priming target promoters for activation at the onset of zygotic gene expression. Before the MBT, β-catenin target promoters associate with RNA Pol II (CTDpSer5) and are marked by H3K9/14ac and H3K4me3, independently of their level of mRNA expression. Deposition of H3K4me3, in particular, requires both preMBT β-catenin and RNA Pol II function. Importantly, during dorsal specification, β-catenin recruits the arginine methyltransferase Prmt2 to target gene promoters, which results in the asymmetric dimethylation of Histone H3 arginine 8. Recruitment of Prmt2 to β-catenin target gene promoters is both necessary and sufficient to establish the dorsal gene expression program. We therefore provide direct evidence for a complex pre-transcriptional mechanism at work in early embryos to pre-set patterns of gene expression, and provide an initial analysis of chromatin architecture during this critical period of development.
The maternal Wnt/β-catenin pathway in Xenopus (and zebrafish) specifies dorsal cell fates before the MBT under conditions of global transcriptional repression. Two classes of dorsal genes are expressed in response to maternal β-catenin (Yang et al., 2002b): genes such as siamois and xnr3 are expressed at the MBT (Figure 1A), whereas genes exemplified by xnr5 and xnr6 are transcribed as early as the 256-cell stage, bypassing preMBT global transcriptional repression (Figure 1A).
Maternal β-catenin is required to activate the dorsal gene expression program (Heasman et al., 2000), but it is not yet clear which β-catenin targets are required for dorsal development. Siamois induces complete secondary axes and rescues dorsal development in ventralized embryos (Lemaire et al., 1995), whereas combined loss of siamois and the closely related gene twin blocks dorsal development (Ishibashi et al., 2008; Laurent et al., 1997), indicating that siamois and twin play essential, instructive roles in dorsal induction. However, these experiments leave open the possibility that dorsal development depends on additional targets of maternal Wnt/β-catenin signaling. To test whether siamois expression is sufficient for dorsal development in β-catenin-deficient embryos, β-catenin was depleted by injection of a morpholino oligonucleotide (βMO), which blocked dorsal development (Figure 1B, panel ii), as described (Heasman et al., 2000); expression of siamois in these embryos rescued dorsal development to the same extent as β-catenin itself (Figure 1B, panels iii and iv). Therefore, establishment of zygotic siamois expression represents an essential patterning event driven by the maternal Wnt/β-catenin pathway.
It is paradoxical, then, to consider how a transcription factor such as β-catenin functions during a period of global transcriptional repression. One possibility is that, while β-catenin is present throughout cleavage stages, it only functions at the MBT to activate target genes such as siamois and xnr3. This seems unlikely, as β-catenin is required for preMBT transcription of xnr5 and xnr6 and dorsal specification by β-catenin is complete by the 32-cell stage (Yang et al., 2002b). However, those experiments did not directly address the requirement for early β-catenin function in the induction of siamois and xnr3 at the MBT. We therefore inhibited Wnt pathway activity at discrete times before the MBT using a dexamethasone-inducible form of Tcf3 that is unable to bind β-catenin (ΔNTcf3-GR) and thereby inhibits endogenous β-catenin/Tcf3 transactivation. While inhibition of β-catenin/Tcf3 transactivation at the 4-cell stage significantly reduces siamois and xnr3 expression after the MBT, inhibition at the 32-cell stage has little or no effect (Figure 1C). This indicates that, as early as the 32-cell stage, dorsal cell fates have been specified and “locked-in”, as the embryo is no longer sensitive to inhibition of β-catenin/Tcf3 transactivation. Notably, six cell divisions separate the 32-cell stage from the MBT, indicating that the information imparted to promoters must be inherited through multiple cell divisions following dorsal specification. We thus conclude that dorsal specification by β-catenin is temporally uncoupled from the onset of dorsal target gene expression.
We next hypothesized that β-catenin binds to target promoters and poises them for activation before the MBT so that they may be expressed following the large-scale activation of the zygotic genome. In general, transcriptionally poised loci are bound by initiating (CTD pSer5) RNA Pol II and marked by H3K9/14ac and H3K4me3 prior to the onset of transcription (Guenther et al., 2007). To test whether these promoters bear marks of poised chromatin architecture before the onset of mRNA expression, we performed chromatin immunoprecipitation (ChIP) assays (Blythe et al., 2009) for these marks of poised loci and for β-catenin in 1000-cell stage embryos, the earliest preMBT time-point (prior to the onset of siamois and xnr3 expression) where it is feasible, in our hands, to perform ChIP (Figure 1D-F).
Although transcription is constrained before the MBT, phosphorylated forms of RNA Pol II corresponding to both initiating (CTD pSer5) and elongating (CTD pSer2) are detected before the MBT in whole embryo lysates (Figure S1), albeit at lower levels than after the MBT. To assess whether initiating or elongating RNA Pol II associates with β-catenin target genes, we performed ChIP against CTD pSer5- and CTD pSer2-RNA Pol II (Figure 1D), comparing samples collected before the onset of siamois and xnr3 expression (1000-cell) with those collected after the onset of expression (Stage 9, blastula, see Figure 1A). As a positive control, we compared RNA Pol II occupancy at the xnr6 locus, which is transcribed during both the 1000-cell stage and at Stage 9 (see Figure 1A). As expected, similar levels of both forms of RNA Pol II are found associated with the xnr6 locus before and after the MBT (Figure 1D). Consistent with our hypothesis, the siamois and xnr3 loci are bound by initiating RNA Pol II before the onset of their expression, at levels similar to those found after the MBT (Figure 1D, left panel); however the amount of elongating RNA Pol II associated with siamois and xnr3 increases after the MBT (Figure 1D, right panel). These observations indicate that the siamois and xnr3 loci are bound by initiating RNA Pol II prior to the onset of their expression, as expected for poised loci.
In addition to binding the promoters of preMBT expressed genes, xnr5 and xnr6, β-catenin also binds the siamois and xnr3 promoters before the MBT (Figure 1E and F). We also tested whether β-catenin could bind later (zygotic Wnt) targets during preMBT stages. Before the MBT, β-catenin does not bind a cluster of Tcf/Lef binding sites flanking the zygotic Wnt target gene myf5 (Yang et al., 2002a), although it binds these sites after the MBT (Figure 1E and data not shown), indicating that β-catenin binding is restricted to maternal Wnt target genes before the MBT. The siamois and xnr3 promoters are also marked by H3K9/14ac before the onset of their expression (Figure 1E). However, while β-catenin binding is restricted to maternal Wnt target genes, H3K9/14ac binding is more widepread: the promoters of all genes tested—maternal and zygotic β-catenin target genes and a negative control locus, myosin light chain 2 (mlc2)—contain H3K9/14ac before the MBT. Finally, maternal Wnt target promoters are marked before the MBT with another marker of poised loci, H3K4me3 (Figure 1F). Taken together, these data demonstrate that the β-catenin target genes siamois and xnr3 arrive at the MBT poised for expression.
These observations raise the possibility that β-catenin plays an instructive role in establishing poised chromatin architecture at the siamois and xnr3 loci. In support of this hypothesis, depletion of β-catenin reduces H3K4me3 at the siamois and xnr3 promoters (Figure 2A). In addition, inhibition of β-catenin/Tcf3 transactivation by ΔNTcf3 reduces β-catenin binding to the siamois and xnr3 promoters and blocks H3K4me3 (Figure 2B). In contrast, ΔNTcf3 has less of an effect on H3K9/14ac at the poised siamois and xnr3 promoters, suggesting that preMBT acetylation at siamois and xnr3 is established, at least in part, independently of β-catenin.
Dorsal specification by β-catenin prior to the 32-cell stage is sensitive to inhibition of RNA Pol II (Yang et al., 2002b). Deposition of H3K4me3 at promoters also correlates with occupancy of initiating polymerase (Ng et al., 2003), and RNA Pol II is also required for β-catenin-mediated H3K4me3 in Drosophila (Parker et al., 2008). We therefore tested whether preMBT RNA Pol II function is necessary for the establishment of other marks of poised loci. Knocking down β-catenin has no effect on the occupancy of initiating RNA Pol II at the poised siamois and xnr3 promoters (Figure 2C), indicating that initiating RNA Pol II is established at poised loci independently of β-catenin. In addition, RNA Pol II function is not required for binding of β-catenin; however, inhibition of Pol II greatly reduces H3K4me3 at the siamois and xnr3 promoters, but does not affect H3K9/14ac (Figure 2D).
In summary, we find that the maternal Wnt target genes siamois and xnr3 arrive at the MBT poised for activation. β-catenin functions, in collaboration with RNA Pol II, to establish the H3K4me3 mark in particular, whereas these promoters are bound by initiating RNA Pol II and H3K9/14ac even when Wnt/β-catenin signaling is inhibited. We conclude that β-catenin is required for the establishment of poised chromatin architecture at these loci, thereby priming them for activation at the MBT.
We next hypothesized that, during dorsal specification (between the 4- and 32-cell stage), β-catenin interacts with a chromatin-modifying activity that functions to establish poised chromatin architecture at target promoters. To identify such a factor, we immunoprecipitated β-catenin from 8- to 32-cell stage embryos and performed in vitro histone acetyl- or methyl-transferase (HAT or HMT) assays. During dorsal specification, β-catenin interacts with a HMT activity that specifically methylates Histone H3 but not H4 (Figure 3A). Under these assay conditions, we were unable to detect an associated HAT activity (Figure S2A). β-catenin is predicted to interact with several functionally different macromolecular complexes (Gottardi and Gumbiner, 2001). We determined that β-catenin interacts with this HMT activity in a high molecular weight complex that is unique to a subset of cellular β-catenin (Figure S2B-D).
To identify the residue on Histone H3 targeted by the β-cat/HMT complex, we performed β-catenin IP/HMT assays using as the substrate recombinant H3.3 (rH3.3) with alanine point mutations at candidate target residues (Figure 3B) on the H3 N-terminal tail previously shown to be methylated: arginines (R) 2, 8, 17, 26 and lysines (K) 4 and 9 (Bedford and Clarke, 2009; Kouzarides, 2007). As with H3, the β-cat/HMT significantly methylates rH3.3(WT) over background. Importantly, mutation of K4 has no effect on H3 methylation, suggesting that the H3K4me3 observed at the poised siamois and xnr3 promoters is indirectly established by β-catenin (Figure 1F, 2A and 2B, see Discussion). These observations also rule out R2, 17, and 26 as the major methyl acceptor sites for the β-cat/HMT. In contrast, mutation of either R8 or K9 prevents H3 methylation by the β-cat/HMT. Similarly, while the β-cat/HMT methylates an unmodified H3 (1-15) peptide (Figure 3C, lane 2) to a similar level as full-length H3 (data not shown), modification of H3 peptides pre-modified at either R8 (asymmetric dimethyl) or K9 (acetyl and trimethyl) prevents methylation by the β-cat/HMT (Figure 3C, lanes 3, 6, and 7). Thus, in addition to targeting either position R8 or K9, the β-cat/HMT is also sensitive to the modification status of these residues.
H3K9 methylation is generally associated with heterochromatin and transcriptional repression (Kouzarides, 2007), and is therefore an unlikely target residue for the β-cat/HMT. Several ChIP experiments failed to detect H3K9me2 and –me3 at the siamois and xnr3 promoters between the 1000-cell stage and MBT, and H3K9me1 levels were not sensitive to stabilization of β-catenin by LiCl treatment (Figure 3D and data not shown). On the other hand, the effect of H3R8 methylation on transcriptional control is poorly understood. To test whether β-catenin activity regulates H3R8 methylation at target promoters, we generated an antibody that specifically detects asymmetrically dimethylated H3R8 (H3R8me2a) (Figure S2E and F), and confirmed that the H3R8me2a modification occurs in vivo. Subsequently, we measured H3R8 methylation at the siamois promoter by ChIP. H3R8me2a associates with the siamois promoter at the MBT and symmetric H3R8 dimethylation (H3R8me2s) is not detected (Figure 3E). This result was confirmed with an independent source of anti-H3R8me2a antibody (Figure 3F and Figure S2G). To test whether association of H3R8me2a correlates with β-catenin activity, we exposed preMBT embryos to a pulse of LiCl, which stabilizes β-catenin throughout the embryo. One hour after the LiCl pulse, H3R8me2a increased dramatically at the siamois and xnr3 promoters (Figure 3E and data not shown), indicating that asymmetric H3R8 methylation correlates with preMBT β-catenin activity. Furthermore, knockdown of β-catenin reduced H3R8me2a at the siamois and xnr3 promoters before the MBT (Figure 3F). Thus, the β-cat/HMT asymmetrically dimethylates H3R8 and is sensitive to the modification state of H3K9. Our results also indicate that β-catenin interacts with a type I (asymmetric) arginine HMT in early Xenopus embryos.
We undertook a candidate-based approach to identify the β-catenin-associated arginine HMT. Of the ten members of the protein arginine methyltransferase (Prmt) family, three have been shown to methylate Histone H3 in vitro: Carm1, Prmt5, and Prmt6 (Guccione et al., 2007; Hyllus et al., 2007; Pal et al., 2004; Schurter et al., 2001). Of these, only the type II, symmetric HMT Prmt5 specifically targets R8 (Pal et al., 2004). By expressing myc-tagged Prmts in Xenopus embryos and immunoprecipitating β-catenin, we found that none of the Prmts known to methylate H3 (Carm1, Prmt5, and Prmt6) co-purified with β-catenin (data not shown). However, Prmt2, which is most closely related to Carm1 and Prmt6, co-immunoprecipitates with β-catenin (Figure 4A). Although recombinant β-catenin and Prmt2 do not interact directly in vitro (data not shown), GST-tagged Prmt2 interacts with β-catenin in Xenopus embryo lysates in a temperature and ATP-dependent manner (Figure 4B). Coupled with our observation that the β-cat/HMT complex is large (Figure S2B-D), we propose that an unknown catalytic activity is required for Prmt2 and β-catenin to interact within a large macromolecular complex.
By sequence, Prmt2 is most closely related to the type I HMTs Carm1 and Prmt6, but its substrate preference has not been determined because recombinant Prmt2 has little to no activity in vitro (Lakowski and Frankel, 2009; Scott et al., 1998). Also, to our knowledge, HMT activity associated with endogenous Prmt2 has not been reported. Interestingly, endogenous Prmt2 immunoprecipitated from mouse embryonic stem cells methylates Histone H3 (Figure 4C). Likewise, myc-Prmt2 expressed in Xenopus embryos methylates Histone H3, and this activity requires H3R8 (Figure 4D). In these experiments, purified Prmt2 has HMT activity towards Histone H3, reflecting either the activity of Prmt2 itself or the activity of another HMT that co-immunoprecipitates with Prmt2. Further investigation will be required to determine the factors that regulate endogenous Prmt2 catalysis.
Importantly, Prmt2 binds the siamois promoter in preMBT embryos (Figure 4E) as determined by ChIP. Furthermore, β-catenin knockdown reduced Prmt2 occupancy at the siamois promoter, suggesting that β-catenin recruits Prmt2 to target promoters. We therefore conclude that Prmt2 represents the HMT activity that associates with β-catenin in early Xenopus embryos based on the observations that β-catenin interacts with Prmt2, Prmt2 has a HMT activity directed at H3 that is sensitive to the R8A mutation, and β-catenin recruits Prmt2 to target genes during the preMBT period.
Based on the above observations, we predicted that recruitment of Prmt2 by β-catenin to dorsal target genes would be essential for specifying the dorsal developmental program. Therefore, we tested whether loss of Prmt2 function in early embryos would affect dorsal specification. Xenopus Prmt2 is expressed maternally, and knockdown of Prmt2 in fertilized embryos only weakly affects expression of siamois and xnr3 (data not shown). Therefore, we knocked down Prmt2 in oocytes and generated maternally depleted Prmt2 (prmt2-) embryos by the host transfer method (Mir and Heasman, 2008). Prmt2- embryos generated by antisense DNA are impaired in dorsal development (Figure 5A). These embryos develop with a range of dorso-ventral morphologies, displaying completely ventralized, anterior-truncated (partially ventralized), and normal phenotypes within single clutches (Figure 5A and B, N=66 embryos from six independent experiments). Importantly, depletion of Prmt2 reduces the expression of both siamois and xnr3 following the MBT (Figure 5C and Figure S3A and B), indicating that Prmt2 function is necessary for the activation of maternal Wnt/β-catenin dependent transcription. These ventralized phenotypes are specific to loss of Prmt2 function, as expression of mouse Prmt2 mRNA in prmt2- embryos rescues siamois and xnr3 expression and partially rescues the morphological phenotype (Figure 5C and B, N=26 from three independent experiments). We also replicated the knockdown of siamois and xnr3 with a translational blocking morpholino oligonucleotide against Prmt2 (Figure S3A). However, while the morpholino generated a more severe knockdown of siamois and xnr3, it was more difficult to generate viable MBT-stage embryos, possibly due to a general requirement for Prmt2 in the regulation of chromatin structure beyond its proposed role in the Wnt pathway. Alternatively, the antisense DNA yielded a partial knockdown of maternal prmt2 message that allowed for recovery of viable embryos at the expense of the severity of the phenotype (Figure 5 and Figure S3). Nonetheless, these observations strongly support the conclusion that recruitment of Prmt2 to dorsal gene promoters is a necessary step in establishing the dorsal gene expression program.
β-catenin interacts with chromatin via the Tcf/Lef family of DNA-binding factors (Behrens et al., 1996; Molenaar et al., 1996), and previous investigations have exploited this interaction to target factors of interest to Tcf/Lef binding sites and test their effects on target gene expression (Vleminckx et al., 1999). Therefore, we generated chimeric proteins (Figure 6A) between Prmt2 and the DNA binding domain of Lef-1 (ΔNLef1) to direct Prmt2 to target genes and evaluate its effect on dorsal specification. Chimeras between ΔNLef1 and a SAM-binding mutant of Prmt2 (Prmt2GG)(Qi et al., 2002), or wild type Carm1, Prmt5, and Prmt6 were also generated as controls. All chimeric proteins were expressed to similar levels in blastula stage embryos (Figure 6B and data not shown).
Recruitment of Prmt2 to β-catenin target genes is sufficient to specify the dorsal developmental program, as expression of the Prmt2:ΔNLef1 chimera in ventral blastomeres converts them to dorsal progenitors (N=172, Figure 6B), similar to activation of Wnt/β-catenin signaling. Ventral expression of the ΔNLef1 DNA binding domain alone did not induce secondary axes (N=31, data not shown). To determine whether Prmt2-induced dorsal cell fates are dependent on its catalytic activity, we tested the ability of the Prmt2GG mutant to induce secondary axes. At similar levels of expression, Prmt2GG:ΔNLef1 induced significantly fewer secondary axes (N=81, Figure 6B, histogram), and these axes were typically truncated (Figure 6B, photo) compared to those observed with Prmt2:ΔNLef1, indicating that dorsal specification by Prmt2 is dependent on its HMT activity. However, Prmt2GG nonetheless conferred some dorsal axis inducing activity; therefore we cannot rule out that additional factors that interact with Prmt2 contribute to Prmt2-dependent dorsal specification.
In a complementary approach, we tested whether Prmt2:ΔNLef1 could rescue β-catenin loss of function. Embryos depleted for β-catenin develop with a ventralized phenotype (Figure 6C ii; also figure 1B ii), whereas expression of Prmt2:ΔNLef1 in β-catenin-depleted embryos restores the full range of dorsal and anterior structures (Figure 6C iii), remarkably similar to control embryos (Figure 6C i), albeit typically with a single eye (86% rescue, N=172). The anterior defects in Prmt2:ΔNLef1-rescued β-MO embryos could result from activation of the posteriorizing zygotic Wnt target genes, as expression of Prmt2:ΔNLef1 alone in dorsal blastomeres did not affect dorsal specification, but did induce a weak posteriorized phenotype (data not shown, N=29). Importantly, expression of Prmt2:ΔNLef1 in β-MO embryos rescues organizer gene expression, whereas the ΔNLef1 DNA binding domain alone has minimal activity (Figure 6D).
To assess the specificity of Prmt2:ΔNLef1, we also tested whether other Prmts could rescue dorsal specification in β-catenin depleted embryos. Prmt5 shares target residue specificity with Prmt2, but symmetrically dimethylates H3R8 (Pal et al., 2004), and Prmt5:ΔNLef1 is unable to rescue dorsal specification in β-catenin depleted embryos (Figure 6B iv—0% Rescue, N=31)). On the other hand, the robust transcriptional activator Carm1 asymmetrically methylates H3 R2, R17, and R26, with only a weak activity towards R8 (Chen et al., 2000; Schurter et al., 2001). Expression of Carm1:ΔNLef1 is toxic to embryos shortly after gastrulation, so phenotypic rescue could not be scored. However, at gastrula stages, Carm1:ΔNLef1 does not rescue siamois and xnr3 expression (Figure 6D). Finally, Prmt6, which targets H3R2 (Guccione et al., 2007; Hyllus et al., 2007), also did not rescue dorsal specification (not shown, 0% rescue, N=30). Thus, of the Prmts tested, only the recruitment of Prmt2 to β-catenin target promoters is sufficient to rescue dorsal specification, demonstrating the unique role of Prmt2 in the regulatory events that establish the transcriptional network driving dorsal development in Xenopus embryogenesis.
In this work, we demonstrate that dorsal specification by the Wnt/β-catenin pathway drives transcriptional poising at Wnt regulated dorsal target genes. This accounts for the uncoupling of dorsal specification (prior to the 32-cell stage) from the onset of target gene expression (MBT for siamois and xnr3). We have investigated this from two perspectives. First, ChIP analysis demonstrates that the β-catenin target genes siamois and xnr3 arrive at the MBT poised for activation, marked by initiating RNA Pol II, H3K4me3, and H3K9/14ac. β-catenin activity, in particular, is required for the ultimate establishment of H3K4me3. Second, biochemical and functional analyses show that β-catenin interacts with the H3R8 methyltransferase Prmt2 during dorsal specification and that Prmt2 activity is both necessary and sufficient to drive dorsal development downstream of β-catenin. Transcriptional poising occurs in eukaryotes from yeast to mammals, including in several embryonic contexts (Akkers et al., 2009; Baugh et al., 2009; Bernstein et al., 2006; Guenther et al., 2007; Hargreaves et al., 2009; Muse et al., 2007; Radonjic et al., 2005; Rougvie and Lis, 1988; Vastenhouw et al., 2010; Zeitlinger et al., 2007); here we illustrate how a critical developmental signaling pathway, acting through a gene-specific transcription factor, poises genetic loci for expression at a later time in development.
Drosophila heat shock genes are the most well understood examples of poised genetic loci. Initially, “pioneer factors” (such as trithorax-like/GAGA factor) function to open chromatin and nucleate the assembly of the initiated (but stalled) RNA Pol II holoenzyme [(Lee et al., 2008; Lee et al., 1992; Shopland et al., 1995) and references therein]. Following binding of “release factors” (such as Heat Shock Factor) that recruit elongation-promoting factors such as CyclinT/Cdk9, RNA Pol II enters into productive elongation [(Lis et al., 2000) and references therein]. Our work does not suggest that β-catenin functions as a pioneer factor: in the absence of β-catenin function, target gene promoters are nonetheless maintained in an “incompletely poised” state containing stalled RNA Pol II and H3K9/14ac but not H3K4me3. Rather, our work suggests that β-catenin functions as part of the mechanism that establishes a “fully poised” state, recruiting additional chromatin modifying activities (such as Prmt2), ultimately resulting in the downstream establishment of H3K4me3 in collaboration with RNA Pol II. Upon establishment of a fully poised promoter, β-catenin target genes thereby become receptive to transcriptional activation at the MBT. We further speculate that the incompletely poised state of dorsal target genes seen in the absence of β-catenin reflects the embryonic competency to activate the dorsal gene expression program throughout the embryo.
The targets of the maternal Wnt/β-catenin pathway demonstrate different latencies between dorsal specification and the onset of gene expression. The preMBT genes xnr5 and xnr6 require additional input from the maternal transcription factor VegT (Takahashi et al., 2000). Thus, VegT could function as a “release factor” for these genes, recruiting elongation-promoting factors to the xnr5 and xnr6 loci downstream of β-catenin. On the other hand, factors that regulate the global activation of the zygotic genome at the MBT could be responsible for the release of the later responding siamois and xnr3 genes. Indeed, transcriptional poising is an attractive mechanism to account for the synchronous activation of large-scale zygotic gene expression at the MBT. While such global activating factors have yet to be identified in Xenopus, in Drosophila, both Smaug and Zelda have been shown to be essential factors for zygotic genome activation (Benoit et al., 2009; Liang et al., 2008). Smaug activity is essential for the establishment of elongating (CTD pSer2) RNA Pol II at the MBT, and could thereby promote “release” of such preMBT poised loci, albeit indirectly (Benoit et al., 2009). The mechanism of action for Zelda is unknown, but it binds DNA sequences present in the majority of immediate-early zygotic transcripts (Liang et al., 2008). Further investigation is needed to determine the extent of transcriptional poising at immediate-early zygotic loci and the mechanism of action for such global zygotic gene activators at the MBT.
While preMBT Xenopus embryos are transcriptionally competent (Prioleau et al., 1994; Toyoda and Wolffe, 1992), several overlapping mechanisms dominantly suppress zygotic gene expression (Veenstra, 2002). Interfering with these repressive activities can reveal a suppressed pro-transcriptional activity. Depleting embryos of the DNA methyltransferase Dnmt1 causes precocious expression of many genes, suggesting that these genes are poised for activation prior to the MBT but are repressed by Dnmt1-mediated DNA methylation (Stancheva and Meehan, 2000). Also, embryos generated from transplantation of transcriptionally active nuclei will display preMBT expression of genes that were active in the original donor cells (Ng and Gurdon, 2005). This transcriptional memory is linked to chromatin modifications that correlate with active transcription, particularly the incorporation of the histone variant H3.3 (Ng and Gurdon, 2008). These observations demonstrate the competency of preMBT embryos to establish and maintain active-but-repressed chromatin. We further speculate that transcriptional poising is a major mechanism underlying the activation of the zygotic genome at the MBT.
β-catenin can interact with a number of chromatin modifying enzymes and may recruit additional factors to poised loci, including H3K4 methyltransferases or, at the MBT, elongation promoting factors (Mosimann et al., 2006; Sierra et al., 2006). As hypothesized by Mosimann and Basler (2009), “cofactor switching” of β-catenin is an attractive model to account for the numerous possible interactions between β-catenin and its coactivators, yet has proven difficult to dissect at the molecular level. A recent genome-wide screen for β-catenin interactors in colorectal carcinoma cell lines (Major et al., 2008) confirmed several of these interactions and identified novel complex members, including the H4 arginine HMT Prmt1. Carm1 also interacts with β-catenin in the context of androgen receptor signaling (Koh et al., 2002). In preMBT embryos, β-catenin interacts with Prmt2, but not with Carm1 or Prmt1, suggesting that the β-catenin/chromatin remodeling complex associates with different Prmt family members in a context-dependent manner. Additionally, in colon cancer cell lines, the β-catenin/chromatin remodeling complex contains the MLL family of H3K4 HMTs (Major et al., 2008; Sierra et al., 2006). Our analysis indicates that, in early Xenopus embryos, the primary HMT activity associated with β-catenin is directed at H3R8, but the possibility remains that β-catenin interacts with MLL family members in later developmental stages. A systematic comparison of these complexes from diverse developmental stages will be necessary to investigate this possibility further.
Notably, after the 32-cell stage, siamois and xnr3 expression becomes “locked in” and resistant to inhibition of β-catenin/Tcf transactivation. Establishment of poised chromatin architecture is an attractive model to account for this observation. ΔNTcf3 functions by inhibiting Tcf3 binding to chromatin (Tutter et al., 2001), thereby excluding β-catenin from its genomic loci. However, another co-factor of β-catenin, Pygopus, binds methylated H3K4 via its PHD domain (Fiedler et al., 2008) and is also essential for dorsal specification (Belenkaya et al., 2002). This suggests an attractive model where, following establishment of poised chromatin architecture prior to the 32-cell stage, Pygopus could function as a Tcf3-independent tether for β-catenin at poised target genes (de la Roche and Bienz, 2007; Fiedler et al., 2008). Additionally, since poised loci can also behave as chromatin insulators (Chopra et al., 2009), the poised chromatin architecture of the siamois and xnr3 loci could serve the dual purpose of marking these genes for activation at the MBT and insulating them from down-regulation by transcriptional repressors, such as non-liganded Tcf3.
A remaining question is what role asymmetric H3R8 methylation plays in either the establishment or maintenance of poised chromatin architecture, in particular H3K4 methylation. The Histone Code Hypothesis postulates that histone modifications function combinatorially to regulate events such as transcriptional activation (Strahl and Allis, 2000). Thus, H3R8 methylation could recruit downstream complexes containing an H3K4 HMT, resulting in the subsequent methylation of H3K4. Although factors that “read” H3R8me2a have yet to be identified, histone arginine methylation influences the patterns of histone tail modifications (Bedford and Clarke, 2009). Alternatively, H3R8 methylation could inhibit the activity of a repressive factor, such as an H3K4 demethylase, which would otherwise maintain transcriptional repression. In support of this model, H3R8 methylation inhibits demethylation of K4 by LSD-1 (Forneris et al., 2006). In addition, we observe that β-catenin mediated H3R8 methylation is sensitive to the modification state of K9, as also observed for the symmetric H3R8 HMT Prmt5 (Pal et al., 2004). Similarly, Prmt1, whose methylation of Histone H4R3 potentiates downstream H4 acetylation, will not methylate a pre-acetylated H4 tail (Wang et al., 2001). Conversely, H3K9 methylation by the HMT G9a is inhibited by methylation of H3R8 (Rathert et al., 2008). These observations suggest that H3 R8 and K9 antagonism represents a critical regulatory node underlying the interpretation of the Histone Code, where asymmetric methylation of R8 activates and K9 represses gene expression. Antagonistic H3K9 methylation could thus represent a mechanism to restrict the expression domains of dorsal determinants in early embryos.
In conclusion, we have demonstrated how a developmental signaling pathway plays an instructive role in the establishment of poised chromatin architecture, linking with chromatin regulatory mechanisms to pre-set gene expression programs in the embryo. Future investigations will focus on how the constellation of chromatin marks established at promoters before the MBT function to regulate the initial patterns of embryonic gene expression.
Extensive details are provided as a supplement
Embryos were obtained, cultured, and microinjected as described (Sive et al., 2000). Capped mRNAs were produced by in vitro transcription using the SP6 mMessage mMachine Kit (Ambion). β-catenin morpholino (Heasman et al., 2000) was injected separately from mRNA to prevent precipitation.
cDNAs for mouse Prmt2, 5, 6, and Carm1 were PCR-amplified from I.M.A.G.E. Consortium [LLNL] clones (Open Biosystems) (Lennon et al., 1996) and subcloned into pCS2+MT.
Radiolabeled RT-PCR was performed as described (Yang et al., 2002b) using GoTaq Flexi polymerase (Promega). Purified RNA (5 embryos/sample) was DNase treated and re-purified by RNeasy (Qiagen). 2.5μg total RNA was used per cDNA synthesis reaction.
ChIP was performed as described (Blythe et al., 2009) with 50 embryos per IP and either radiolabeled PCR or SYBR green QPCR. Anti N-β-catenin was described in (Blythe et al., 2009). Purchased antibodies included anti-H3acK9/14 and -H3K4me3 (Millipore #06-599 & #07-473), anti-Pol II p-Ser5 (Active Motif #39233), anti-Pol II p-Ser2 (Abcam #Ab5095), H3K9me1 and H3K9me3 (Millipore 07-450 & 07-442), H3R8me2a (Active Motif #39651), and anti-Myc epitope (Sigma #C3956). The anti-H3R8me2s antibody was kindly provided by Dr. Said Sif (Ohio State University) (Pal et al., 2004). For primer information, see Supplemental Methods.
Unmodified and R8me2a H3 (1-15+C) peptides were synthesized by the Proteomics Resource Center of the Rockefeller University. Unmodified H3 1-21, H3K9ac 1-20, and H3K9me3 1-24 peptides were from AnaSpec.
H3R8me2a antiserum was generated by immunizing rabbits with H3R8me2a (1-15+C) peptide coupled to KLH (Cocalico Biologicals). This antiserum was depleted of antibodies to unmodified H3 and affinity purified with the R8me2a peptide. The limit of detection was ~8pmol H3R8me2a (1-15) by dot blot. Results were duplicated with H3R8me2a antiserum from Active Motif (39651) (see Figure S2G).
Embryos were homogenized in HM 0.1M buffer (Sierra et al., 2006), supplemented with Protease Inhibitor Cocktail (Sigma P8340) and Phosphatase Inhibitor Cocktail I and II (Sigma P2850 & P5726). Insoluble material was pelleted by centrifugation (14k × g) for 10 minutes at 4°C. Supernatants were adjusted to 0.15mM KCl and 0.2% NP-40 and filtered through a 0.45mm cellulose acetate syringe filter (Millipore) prior to addition of antibodies. Immune complexes precipitated with recombinant protein G agarose were washed 5x in wash buffer (HMZ 0.15M) (Sierra et al., 2006) prior to IP/HMT analysis or western blot.
HM 0.1M lysates of 8- to 32-cell embyos were immunoprecipitated and processed for IP/HMT assays as described (Sierra et al., 2006)(see supplement) using 50mM Tris-HCl (pH8.8) in the reaction buffer. 5μg Calf Thymus Histones, 2.5μg recombinant HA-H3.3, or 5μg synthetic peptide was used as substrates. IP/HMT reactions typically generated <1pmol of in vitro labeled substrate.
Maternal depletions were performed as described (Mir and Heasman, 2008). Antisense DNA (1ng/nl) against Prmt2 mRNA (Genbank HM205111) was injected vegetally into defolliculated oocytes at the indicated doses. Control and antisense injected oocytes were incubated for 24 hours prior to in vitro maturation (2μM Progesterone, 10 hours) and transfer to an egg-laying host. Transferred eggs were collected in high-salt buffer for 4 hours following transplantation prior to in vitro fertilization. For mRNA rescue experiments, 10nl of mouse Prmt2 mRNA (100pg/nl) was injected vegetally shortly before oocyte maturation. The sequence of the Prmt2 antisense oligo is: 5’- TCC GTT CTG TAT CTC TCC —3’.
We thank Daniel Kessler, Steve DiNardo, Marisa Bartolomei, Gerd Blobel, Tom Kadesch, Chris Wylie, Matt Kofron and the members of the Klein Lab for helpful discussions and encouragement. We also thank Dr. Said Sif (Ohio State University) for his gift of the H3R8me2s antiserum, and Scott Paschke (Active Motif) for providing the H3R8me2a antiserum. This work was supported by National Institutes of Health grants (T32-GM007229) and (T32-HD007516) to SAB and (R01-GM76621) to PSK.
Publisher's Disclaimer: This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to our customers we are providing this early version of the manuscript. The manuscript will undergo copyediting, typesetting, and review of the resulting proof before it is published in its final citable form. Please note that during the production process errors may be discovered which could affect the content, and all legal disclaimers that apply to the journal pertain.