|Home | About | Journals | Submit | Contact Us | Français|
Hox proteins form complexes with Pbx and Meis cofactors to control gene expression, but the role of Meis is unclear. We demonstrate that Hoxb1-regulated promoters are highly acetylated on histone H4 (AcH4) and occupied by Hoxb1, Pbx and Meis in zebrafish tissues where these promoters are active. Inhibition of Meis blocks gene expression and reduces AcH4 levels at these promoters, suggesting a role for Meis in maintaining AcH4. Within Hox transcription complexes, Meis binds directly to Pbx and we find that this binding displaces histone deacetylases (HDACs) from Hoxb1-regulated promoters in zebrafish embryos. Accordingly, Pbx mutants that cannot bind Meis act as repressors by recruiting HDACs and reducing AcH4 levels, while Pbx mutants that bind neither HDAC nor Meis are constitutively active and recruit CBP to increase AcH4 levels. We conclude that Meis acts, at least in part, by controlling access of HDAC and CBP to Hox-regulated promoters.
Hox proteins are transcription factors that control anteroposterior body axis formation in animal embryos by regulating gene expression in discrete domains (reviewed in (McGinnis and Krumlauf, 1992). In many situations, Hox proteins form complexes with Pbx and/or Meis cofactors - TALE class transcription factors that are broadly expressed in most tissues and at most stages of embryogenesis - to control transcription of target genes (reviewed in (Mann and Affolter, 1998). There is great variability among Hox-regulated promoters both in their constellation of Meis, Pbx and Hox binding sites and in the transcriptional outcome (activation versus suppression) (Mann and Affolter, 1998). No general rules have emerged for how particular combinations of cofactors control transcriptional outcome, but it appears that the presence of Meis cofactors correlates with active transcription of Hox-regulated genes in many instances (Ferretti et al., 2005; Ferretti et al., 2000; Jacobs et al., 1999; Ryoo et al., 1999), although there are exceptions (Gebelein et al., 2002). In Drosphila, Meis proteins may act in part by facilitating nuclear translocation of Pbx proteins (Rieckhof et al., 1997), but some Meis activities require its DNA binding domain (Noro et al., 2006) and Meis proteins may not control Pbx nuclear translocation in other organisms (e.g. zebrafish; (Choe et al., 2002; Vlachakis et al., 2001)), suggesting that Meis cofactors have additional functions in activating transcription.
An evolutionarily conserved Hox-regulated cascade is required for activation of paralog group 1 (PG1) and PG2 hox gene expression in rhombomere 4 (r4) of the vertebrate hindbrain and this cascade requires Meis function (Choe et al., 2002; Ferretti et al., 2005; Ferretti et al., 2000; Jacobs et al., 1999; McClintock et al., 2002; Popperl et al., 1995; Pöpperl et al., 2000; Waskiewicz et al., 2001). In zebrafish (Fig. 1A), this cascade is initiated by Hoxb1b, which activates hoxb1a expression. Once Hoxb1a is expressed, it can maintain its own expression in r4. In addition, Hoxb1b and Hoxb1a activate hoxb2a expression in r4. Importantly, Meis and Pbx cofactors are required for Hoxb1b and Hoxb1a to drive PG1 and PG2 gene expression in zebrafish r4 (Choe et al., 2002; Choe and Sagerstrom, 2005; Pöpperl et al., 2000; Vlachakis et al., 2001; Waskiewicz et al., 2001; Waskiewicz et al., 2002). We have used this simple Hox-regulated cascade to examine the function of Meis proteins and we find that Meis cofactors control accessibility of HDAC and CBP histone modification enzymes to Hox-regulated promoters during zebrafish development.
We initially dissected 12–14hpf embryos (Fig. 1B) into posterior pieces (that express hoxb1a) and anterior pieces (that do not express hoxb1a). ChIP analysis using antibodies raised to endogenous proteins revealed that Pbx, Meis and Hoxb1a/b occupy the hoxb1a promoter in tissues expressing hoxb1a, but not in non-expressing tissues (Fig. 1C). The hoxb2a promoter is similarly occupied by Pbx, Meis and Hoxb1a/b in posterior, but not anterior, tissues (Supp. Fig. 1). We next examined acetylation of histone H4 (AcH4), a marker of transcriptionally active chromatin, at the hoxb1a and hoxb2a promoters. As expected, ChIP analysis revealed a higher level of AcH4 at the hoxb1a and hoxb2a promoters in posterior tissues (where both genes are expressed) than in anterior (non-expressing) tissues (Fig. 1D). In contrast, the pax2 promoter, which is expressed in both anterior and posterior tissues independently of Hox proteins, shows similar levels of AcH4 in both tissues (Fig. 1D).
Various dominant-negative constructs have been used to interfere with Meis function in zebrafish and Xenopus embryos (Choe et al., 2002; Dibner et al., 2001; Waskiewicz et al., 2001). In particular, the PBCAB dominant-negative construct blocks the function of all known zebrafish Meis/Prep proteins by preventing their nuclear translocation (without affecting the nuclear localization of Pbx or Hox proteins (Choe et al., 2002; Choe and Sagerstrom, 2004)) and interferes with transcription of hoxb1a and hoxb2a in r4 (Choe et al., 2002) (Fig. 1E, Supp. Fig. 1C). Using the PBCAB construct, we find that embryos with reduced Meis function have reduced AcH4 levels at both the hoxb1a and the hoxb2a promoter (Fig. 1F), while total H4 levels are unaffected at both promoters (Supp. Fig 1D). As expected, Meis proteins are not detected at the promoter in embryos expressing the PBCAB dominant negative construct (Supp. Fig. 1E). We note that Pbx, and perhaps Hoxb1a/b, also are not detected, likely due to chromatin compaction resulting from the loss of AcH4. These results indicate that Meis proteins may be required for transcription of Hox-regulated genes by acting at, or upstream of, the histone H4 acetylation step.
While Meis proteins have not been shown to regulate histone acetylation, Pbx proteins reportedly interact with histone deacetylases (HDACs) (Saleh et al., 2000). To explore this further, we identified a zebrafish hoxb1a promoter fragment that contains all Meis, Pbx and Hox binding sites previously identified as necessary for expression of mouse hoxb1 in r4 (Ferretti et al., 2005; Ferretti et al., 2000; Jacobs et al., 1999). This fragment is sufficient to recapitulate the hoxb1a expression pattern in zebrafish embryos (Supp. Fig. 2A, B) and we used it for reporter assays in HEK293 cells (experiments in HeLa cells yielded similar results; Supp. Fig. 3). We find that transfecting Meis3 or Pbx4 has only limited effects on expression of the reporter, whereas transfecting Hoxb1b induces expression ~6-fold (Fig. 2A columns 1–4). This Hoxb1b-mediated activation is dependent on endogenous Pbx, since BMHoxb1b (a point-mutant that cannot bind Pbx (Vlachakis et al., 2001)) does not activate the reporter (Fig. 2A, column 5). Co-transfecting Pbx4 with Hoxb1b leads to a dose-dependent reduction in reporter activation (Fig. 2A, compare columns 7 and 8 to column 4). This effect has been reported previously as being caused by Pbx recruiting HDACs to repress transcription of Hox-regulated genes (Saleh et al., 2000). Accordingly, we observe that treatment with the HDAC inhibitor Trichostatin A (TSA) restores reporter expression to cells transfected with Pbx4 and Hoxb1b (Fig. 2A, compare columns 7, 8 to column 9). We hypothesized that Meis proteins might act to overcome this HDAC-mediated repression. Indeed, co-transfecting Meis3 together with Pbx4 and Hoxb1b restores expression of the reporter similar to treatment with TSA (Fig. 2A, column 11). In contrast, BMMeis3 (a Meis3 mutant that cannot bind Pbx (Vlachakis et al., 2001)) does not restore expression (Fig. 2A, column 12), indicating that Meis3 must bind Pbx4 to overcome HDAC-mediated repression. Accordingly, blocking endogenous Meis activity by the PBCAB dominant negative construct blocks reporter activation by Hoxb1b (Fig. 2A, compare column 6 to column 4).
We next generated a series of Pbx4 deletion constructs (Fig. 2B, top panel) and tested their ability to bind Meis and/or HDAC (Fig. 2B, bottom panel). We find that ΔPBCA2, ΔPBCA3 and ΔPBCB5 do not bind Meis3 (Fig. 2B, lanes 13–15), but vary in their ability to bind HDAC1 such that ΔPBCA2 binds HDAC1 at wild type levels (Fig. 2B, lane 18), ΔPBCA3 partially binds HDAC1 (Fig. 2B, lane 19) and ΔPBCB5 binds HDAC1 only very weakly (Fig. 2B, lane 20). Testing these deletion constructs in the reporter assay revealed that ΔPBCA2 acts similarly to wild type Pbx4 in its ability to repress the reporter when co-transfected with Hoxb1b (Fig. 2C, compare columns 5 and 8 to column 4), consistent with ΔPBCA2 retaining HDAC1 binding. However, while Meis3 substantially enhances reporter activation in cells co-transfected with Pbx4 and Hoxb1b (Fig. 2C, column 6), it has minimal effect in cells transfected with ΔPBCA2 and Hoxb1b (Fig. 2C, column 9), consistent with Meis only overcoming HDAC-mediated repression if it can bind Pbx. In contrast, co-transfection of ΔPBCA3 or ΔPBCB5 with Hoxb1b leads to robust activation of the reporter even in the absence of co-transfected Meis3 (Fig. 2C, columns 11 and 13), indicating that if Pbx4 cannot bind HDACs, Meis is no longer needed to induce expression. Interestingly, this suggests that various Meis domains, including C-terminal activation domains identified in some studies (e.g. (Huang et al., 2005)) may not be required for Hox-mediated transcription, at least under our conditions. We also note that ΔPBCB5 is about three times more active than ΔPBCA3, consistent with ΔPBCA3 retaining partial HDAC binding (Fig. 2B). Together, these findings indicate that Meis binds Pbx4 to overcome HDAC-mediated repression.
We next tested if the ΔPBCA2 and ΔPBCB5 constructs have the predicted effect on histone H4 acetylation and expression of Hox-regulated promoters in zebrafish embryos. Specifically, since ΔPBCA2 binds HDAC, but not Meis, we expect it to recruit HDACs, promote histone deacetylation and repress expression of endogenous Hox-target genes. Indeed, expressing ΔPBCA2 in wild type embryos reduces AcH4 at the hoxb1a and hoxb2a promoters (Fig. 3B), similar to the effect observed using PBCAB to block endogenous Meis function (Fig. 1F), while the otx1 promoter is unaffected. Analysis of hoxb1a expression (Fig. 3A) revealed only weak repression by ΔPBCA2 in wild type embryos. We reasoned that this mild effect might be due to competition from endogenous Pbx for binding to the hoxb1a promoter and therefore assayed lazarus (lzr) mutant embryos that retain maternal Pbx4, but lack zygotic Pbx4 (Pöpperl et al., 2000). We find that ΔPBCA2 leads to near complete repression of hoxb1a in lzr embryos (Fig. 3A). Notably, a mutant form of ΔPBCA2 that cannot bind DNA has no effect on hoxb1a expression in lzr embryos, demonstrating that ΔPBCA2 must bind DNA to mediate its effect.
Based on its activity in the reporter assay, the ΔPBCB5 construct, which does not bind HDACs or Meis, is expected to promote histone acetylation and expression of endogenous Hox target genes. Indeed, expression of ΔPBCB5 increases AcH4 at the hoxb1a and hoxb2a promoters 2- to 2.5-fold (Fig. 3C) and induces weak ectopic hoxb1a expression posteriorly (bracket in Fig. 3D) in wild type embryos. Notably, hoxb1b is expressed in this posterior domain during gastrula and early segmentation stages (when hoxb1a expression is initiated), consistent with ΔPBCB5 acting together with endogenous Hoxb1b to drive this ectopic hoxb1a expression. Dissections allowed us to focus specifically on this posterior tail domain and, using quantitative RT-PCR, we find that ΔPBCB5 induces hoxb1a expression 1.5- to 2-fold in this domain (Fig. 3E). ChIP analysis of dissected tail regions also reveals a 3- to 4-fold increase in AcH4 at the hoxb1a promoter in this posterior domain (Fig. 3F). The modest induction of ectopic hoxb1a expression in this posterior domain is likely due to the presence of factors that repress hoxb1a expression posterior to r4 (e.g. vhnf1 (Hernandez et al., 2004; Sun and Hopkins, 2001; Wiellette and Sive, 2003)). We reasoned that the hoxb1a(β–globin):eGFPum8 transgenic line (that contains only a 1.0kb fragment from the hoxb1a promoter; Supp. Fig. 2A, B) might be less susceptible to such repression. Indeed, ΔPBCB5 induces robust ectopic expression of the transgenic promoter (Supp. Fig. 2C). We also examined the effect of ΔPBCB5 on krox20 expression. krox20 was recently shown to contain a Meis:Pbx:Hox-regulated element in its enhancer (Wassef et al., 2008) and we confirmed that this element is occupied by Meis, Pbx and Hox proteins (Supp. Fig. 4). krox20 is normally expressed in rhombomeres 3 and 5, but we find that krox20 expression is strongly up-regulated posteriorly in ΔPBCB5-injected embryos (Fig. 3D, E). This up-regulation is accompanied by a substantial increase in AcH4 at the krox20 promoter (Fig. 3F), confirming that an ectopic gene expression program is initiated posteriorly in ΔPBCB5-injected embryos. We conclude that ΔPBCA2 (which binds HDAC, but not Meis) promotes histone deacetylation and represses transcription, while ΔPBCB5 (which binds neither HDAC nor Meis) promotes histone acetylation and activates transcription, consistent with our conclusion from figure 2 that Meis binds Pbx to overcome HDAC-mediated repression of Hox target genes.
We note that our observations can be explained if Meis proteins compete with HDACs for binding to Pbx. Indeed, our deletion analyses (Fig. 2B) suggest that the sites required for Meis and HDAC binding reside near one another in Pbx4. Consistent with this hypothesis, co-immunoprecipitation revealed that HDAC1 interacts with Pbx4 following co-transfection into HEK293 cells (Fig. 4A, lane 2). When Meis3 is co-transfected with Pbx4 and HDAC1, it replaces HDAC1 as the Pbx4 interaction partner (Fig. 4A, lane 3), demonstrating that Meis3 and HDAC1 compete for binding to Pbx4. Furthermore, ΔNMeis3 (a Meis3 binding mutant that cannot interact with Pbx proteins (Vlachakis et al., 2001)) cannot displace HDAC1 from Pbx4 (Fig. 4A, lane 4). This finding suggests that Meis proteins may act to displace HDACs from Pbx proteins bound to Hox-regulated promoters in vivo. To test this possibility directly, we again made use of ChIP analysis in zebrafish embryos. Although AcH4 is already present at the hoxb1a and hoxb2a promoters in control embryos (Fig. 1), we find that injection of Meis3 can further increase AcH4 at the hoxb1a and hoxb2a, but not the pax2, promoter (Fig. 4B, left panel), consistent with Meis displacing HDACs also in vivo. In contrast, the ΔNMeis3 construct, which does not bind Pbx, has no effect on AcH4 levels (Fig. 4B, right panel). As observed for ΔPBCB5 injection in figure 3, Meis3 injection does not induce ectopic hoxb1a expression (not shown), but does induce ectopic expression from the hoxb1a(β–globin):eGFPum8 transgene (Supp. Fig. 2C). We next assayed HDAC1 occupancy at the hoxb1a promoter in zebrafish embryos and find that over-expression of Meis3 or ΔPBCB5 decreases HDAC1 occupancy at the hoxb1a promoter (Fig. 4C). Furthermore, expression of ΔPBCA2 increases HDAC1 occupancy at the hoxb1a promoter, likely because endogenous Meis proteins are unable to displace HDACs from ΔPBCA2 occupying the hoxb1a promoter, and the ΔNMeis3 construct has no effect. We conclude that Meis proteins bind Pbx to displace HDACs from the hoxb1a promoter in vivo.
Lastly, we considered that histone acetyl transferase (HAT) enzymes might need to be recruited to Meis:Pbx:Hox complexes in order to maintain high AcH4 levels and active transcription. In particular, several Hox proteins reportedly bind the CBP/p300 HAT enzyme (Chariot et al., 1999; Saleh et al., 2000). We therefore examined hoxb1a promoter occupancy by CBP. We find that expression of Meis3 or ΔPBCB5 increases CBP occupancy at the hoxb1a promoter (Fig. 4D). ΔPBCA2 reproducibly reduces CBP occupancy to a small extent, consistent with the mild effect of ΔPBCA2 on gene expression in wild type embryos in figure 3, while ΔNMeis3 has no effect on CBP occupancy (Fig. 4D). While this finding explains how Meis proteins promote histone H4 acetylation at the hoxb1a promoter, it also suggests that Meis proteins do not recruit CBP directly, since ΔPBCB5 (that cannot bind Meis) is sufficient to increase CBP occupancy. Accordingly, we find that CBP does not bind Meis3, but binds both Pbx4 and PG1 Hox proteins (Fig. 4E). We postulate that Meis proteins promote CBP recruitment indirectly, possibly by displacing HDACs to permit CBP binding. We conclude that Meis proteins are required as Hox cofactors because they modulate HDAC and CBP accessibility at Hox-regulated promoters.
All DNA constructs were generated using standard molecular biology techniques. See Supplemental Materials for detailed description.
HEK 293 and HeLa cells were transfected with 0.5 ug of each expression plasmid using FuGENE 6 (Roche) and harvested after 36 hours. Luciferase activity was normalized using co-transfected Renilla luciferase. Trichostatin A (TSA) treatments were for 12 hours starting 24 hours after transfection. See Supplemental Materials for detailed description of methods used.
ΔPBCA2 (700pg), ΔPBCA2* (700pg), ΔPBCB5 (500pg), ΔPBCB5* (500pg), PBCAB (500pg), meis3 (500pg), ΔNMeis3 (500pg) or myc-HDAC1 (500pg) mRNA were microinjected into 1- to 2-cell stage zebrafish embryos and raised. For qRT-PCR, cDNA was prepared from dissected embryos and subjected to quantitative PCR with gene specific primers (Supp. Table 1). In situ hybridizations were carried out as described previously (Choe et al., 2002; Choe and Sagerstrom, 2004). See Supplemental Materials for detailed description of methods used.
Polyclonal rabbit antisera were raised to full-length Meis3, Pbx4 and Hoxb1b. The Meis3 antiserum does not recognize Prep1, Pbx2, Pbx4, Hoxb1b or Hoxb1a, but cross-reacts weakly with Meis1, 2 and 4 (referred to as ‘Meis antiserum’ in text). The Hoxb1b antiserum does not recognize Pbx2, Pbx4, Prep1, Meis1, Meis2, Meis3 or Meis4, but cross-reacts weakly with Hoxb1a (referred to as ‘Hoxb1a/b antiserum’ in text). The Pbx4 antiserum cross-reacts with Pbx2, but does not recognize Prep1, Meis1, Meis2, Meis3, Meis4, Hoxb1b or Hoxb1a (referred to as ‘Pbx antiserum’ in text). ChIPs were performed based on protocols published previously (Salma et al., 2004). Zebrafish embryos were dissociated and cross-linked in 1% formaldehyde. Genomic DNA was sheared to 200–1000 bp DNA fragments by sonication and 1% of sample volume (Input) was set aside for normalization. Samples were incubated with the appropriate antibody overnight, immune complexes were collected and washed followed by reversal of cross-links. Quantitative PCR was performed using promoter specific primers (Supp. Table 1). PCR amplification was quantified and normalized to the corresponding input sample (1% of total input). Control amplifications using primers to the hoxb1a ORF, did not yield signals above background. Control amplification from ChIPs using preimmune serum or no antibody was subtracted (except in Figures 1 and and33 where background was less than 1% of signal). Data is presented as the average of a minimum of three experiments with error bars indicating standard deviation. Statistical significance was determined using the Student’s t-test in Microsoft Excel. See Supplemental Materials for detailed description of methods used.
We are grateful to Letitiah Etheridge for assistance with experiments, to Drs Y. Ohkawa and A. Imbalzano for assistance with the ChIP protocol and to Drs Kouzarides and Roeder for providing plasmids. This work was supported by NIH grant NS038183.
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.