A new coactivator complex required for retinoic acid-dependent regulation of embryonic symmetry

Bilateral symmetry is a striking feature of the vertebrate body plan organization. Vertebral precursors, called somites, provide one of the best illustrations of embryonic symmetry. Maintenance of somitogenesis symmetry requires Retinoic acid (RA) and its coactivator Rere/Atrophin2. Here, using a proteomic approach we identify a protein complex, containing Wdr5, Hdac1, Hdac2 and Rere (named WHHERE), which regulates RA signalling and controls embryonic symmetry. We demonstrate that Wdr5, Hdac1 and Hdac2 are required for RA signalling in vitro and in vivo. Mouse mutants for Wdr5 and Hdac1 exhibit asymmetrical somite formation characteristic of RA-deficiency. We also identify the Rere-binding histone methyltransferase Ehmt2/G9a, as a RA coactivator controlling somite symmetry. Upon RA treatment, WHHERE and Ehmt2 become enriched at RA target genes to promote RNA Polymerase II recruitment. Our work identifies a novel protein complex linking key epigenetic regulators acting in the molecular control of embryonic bilateral symmetry.


Bilateral symmetry is a striking feature of the vertebrate body plan organization.
Vertebral precursors, called somites, provide one of the best illustrations of embryonic symmetry. Maintenance of somitogenesis symmetry requires Retinoic acid (RA) and its coactivator Rere/Atrophin2. Here, using a proteomic approach we identify a protein complex, containing Wdr5, Hdac1, Hdac2 and Rere (named WHHERE), which regulates RA signalling and controls embryonic symmetry. We demonstrate that Wdr5, The development of bilaterally symmetrical structures such as limbs or somites takes place concomitantly with the asymmetric formation of internal organs such as heart, gut and liver.
Whereas the pathway responsible for establishing left-right identity in the embryo begins to be well understood 1 , little is known about the mechanisms controlling embryonic symmetry.
Retinoic acid (RA) is a derivative of vitamin A, signalling via a heterodimeric RAR/RXR nuclear receptor transcription factor [2][3][4] . In the absence of RA, the heterodimer binds target genes together with the SMRT and NCoR corepressor complexes and histone deacetylases such as Hdac3 to silence gene expression. When the RA ligand binds to RAR, the corepressors are replaced by a set of coactivators including histone acetyltransferases, contributing to active transcription of RA target genes 5,6 . In the absence of RA signalling in the mouse embryo, somite formation becomes asymmetrical, showing a significant delay on the right side 7 . A similar somite desynchronization phenotype is also observed in mutants for the protein Rere (or Atrophin2) which acts as a coactivator for RA signalling 8 .
In order to understand the mechanism of action of Rere in the RA pathway controlling somite symmetry, we first set out to identify Rere-interacting proteins in the mouse mesoderm. To that end, we generated a transgenic mouse line, which allows the conditional expression of a tagged version of Rere containing two HA epitopes at the C-terminal end of the protein (Rere-HA). A Rere-HA construct preceded by a LoxP-STOP-LoxP cassette was introduced into the Rosa26 locus by homologous recombination in mouse embryonic stem (ES) cells.
We then used these cells to generate a Rosa26-LoxP-STOP-LoxP-Rere-HA mouse line (RS-Rere-HA line). Whereas Rere mutants (Rere om/om ) die around E9.5 with defects in forebrain, heart and a right-side specific delay in somite formation 8,9 , expression of at least one Rere-HA allele in the mutant Rere om/om background led to morphologically normal embryos ( Supplementary Fig. 1a-c). Therefore, the tagged Rere-HA protein is functional in vivo.
To direct expression of Rere-HA to the mesoderm, RS-Rere-HA mice were crossed to the T-Cre mouse line 10 . We prepared whole cell protein extracts from approximately 600 RS-Rere-HA;T-Cre embryos, and performed affinity purification using anti-HA antibodies under high or low salt conditions (Fig. 1a). To identify the immunoprecipitated proteins, eluted fractions were submitted to mass spectrometry analysis using the multidimensional protein identification technology (MudPIT) 11 . A set of 105 common proteins was found between the different immunopurification conditions (Supplementary Fig. 1d and Supplementary Table   1). Hierarchical clustering analysis of the 105 proteins based on the normalized spectral abundance factor (NSAF) 12 in the different immunoprecipitation conditions identified three abundant proteins tightly clustering with Rere ( Supplementary Fig. 1e,f). These include Rere's known binding partners Hdac1 and Hdac2 as well as a novel partner, Wdr5 13-16 ( Fig.  1b and Supplementary Fig. 2a). To estimate relative protein levels, we compared the NSAF values for each protein 12 . NSAF for Rere, Hdac1 and Hdac2 were similar while it was three times higher for Wdr5 (Fig. 1b). These results suggest that the proteins Rere, Hdac1, Hdac2 and Wdr5 can interact in the mesoderm.
To validate the identification of Wdr5 as a novel interacting partner of Rere, Hdac1 and Hdac2, we co-expressed tagged versions of the four proteins (Rere-Flag, Flag-Hdac1, Flag-Hdac2 and HA-Wdr5) using a baculovirus-insect cell expression system. After Flag immunoprecipitation of Rere, Hdac1 and Hdac2, HA-Wdr5 was detected in the eluates, as confirmed by LC-MS/MS analysis of the Coomassie stained gel bands ( Fig. 1c and Supplementary Fig. 2b). The four proteins still co-purified together at high salt concentration (500 mM KCl) suggesting the existence of a stable protein interaction network between Rere, Hdac1, Hdac2 and Wdr5 (Fig. 1c). By co-immunoprecipitation of baculovirus-expressed Rere-Flag and HA-Wdr5, we could demonstrate that Rere binds directly to Wdr5 (Fig. 1d).
However Wdr5 does not bind directly to Hdac1 or Hdac2 in high or low salt wash ( Supplementary Fig. 2c), suggesting that Rere acts as a scaffolding component binding Hdac1/Hdac2 and Wdr5. To analyse whether the four co-immunopurified proteins form a stable protein complex, we carried out gel filtration chromatography followed by western blot and mass spectrometry analysis. All four proteins co-eluted together in a fraction corresponding to a high molecular weight complex of 0.5-0.6 MDa (Fig. 1e,f and Supplementary Fig. 2d,e). This molecular weight is consistent with the abundance predicted by the NSAF values of the proteomic analysis. Additionally, Hdac1, Hdac2 and Wdr5 coimmunoprecipitated with the endogenous Rere in NIH3T3 cells further supporting the existence of such a protein complex (Fig. 1g)  Treatment with Hdac2 siRNA led to an increase of RARE-Luciferase reporter activity ( Fig.   2c and Supplementary Fig. 3d), which might be explained by a stabilization of Hdac1 (due to the decrease of Hdac2), potentially resulting in an increase in RA signalling 18,19 . Consistent with this possibility, the double knockdown of Hdac1 and Hdac2 further decreased RA signalling compared to Hdac1 siRNA alone (Fig. 2d). Furthermore, Hdac1 or Hdac1/Hdac2 depletion reduced RA activation mediated by Rere and Wdr5 (Fig. 2e,f). Inhibition of deacetylase enzymatic activity with a range of chemical inhibitors decreased RA signalling ( Fig. 2g) 20 . In line with this, overexpression of Rere or Wdr5 did not lead to a significant increase in RA signalling when cells were treated with the HDAC inhibitors Trichostatin A (TSA) or Sodium Butyrate (SB) (Fig. 2h,i). Hdac1 and Hdac2 have been shown to bind Rere N-terminal region 9,13,14 . In the presence of RA, overexpression of the N-terminal region of Rere (N-Rere) strongly increased RA signalling whereas no activation could be seen with Rere C-terminal region (Rere-C) (Fig. 2j). The activation by N-Rere was dependent on deacetylase activity since TSA or SB treatment strongly decreased N-Rere dependent RA signalling (Fig. 2k). Overall, these results show that the WHHERE complex proteins Rere, Wdr5, Hdac1 and Hdac2 act to trigger the RA pathway activation in NIH3T3 cells.
Moreover, activation of RA signalling by the WHHERE complex depends on Hdac1 and Hdac2 deacetylase activity.
To analyze WHHERE-dependent RA regulation in vivo, we characterized the phenotype of Next we analyzed the binding of the WHHERE complex to RA regulated genes by chromatin immunoprecipitation (ChIP). To that end, we first checked the expression kinetics of the RARE-Luciferase reporter and of the endogenous Rarβ gene in NIH3T3 cells treated with RA for 1, 2 and 6 hours. At 2 and 6 hours, strong transcriptional activation was observed for both RA targets (Fig. 4a,b). In mouse embryos, we observed by ChIP analysis that the Retinoic Acid Receptor alpha (Rarα) and all the WHHERE complex components were present at the Rarβ promoter and at the RARE-containing reporter RARE-LacZ (Fig. 4c-f). In ChIP experiments performed 1 hour after RA treatment of NIH3T3 cells, Rere, Wdr5, Hdac1 and Hdac2 but not Rarα increased at the Rarβ promoter (Fig. 4g,h). This effect was specific, as no significant enrichment of WHHERE complex members could be observed in regions upstream of the Rarβ promoter in similar conditions (Fig. 4h, bottom graph). Recruitment of RNA Polymerase II (Pol II) increased following RA treatment paralleling the WHHERE complex recruitment (Fig. 4i). Then we investigated the requirement of Hdac1 deacetylase activity in transcription activation and Pol II recruitment during RA signalling. Decreased  (Fig. 4l-n). Together, these data support a role of the WHHERE complex in the recruitment of Pol II necessary for early activation of RA regulated genes.
This role depends on the deacetylase activity of Hdac1 and Hdac2.
The histone methyltransferase Ehmt2 (G9a) was shown to bind the N-terminal SANT domain of Rere, and together with Hdac1/Hdac2 to regulate the methylation of H3K9 at specific loci leading to the formation of compact heterochromatin and gene silencing 14 . In the proteomic experiment, Ehmt2 (and the related protein Ehmt1) were detected with low NSAF values compared to the members of the WHHERE complex suggesting that its binding to Rere might be transient (Supplementary Fig. 5a). In mouse embryos deficient for Ehmt2 23 crossed to the RARE-LacZ reporter 22 , LacZ expression is downregulated suggesting that Ehmt2 is also implicated in positive regulation of RA signalling (Fig. 5a,b). Furthermore, half of the Ehmt2 mutant embryos presented a desynchronization of somite formation on the right side resembling mutants of members of the WHHERE complex (Fig. 5c-e; Supplementary Fig.   4f). In NIH3T3 cultures, siRNA-mediated knockdown of Ehmt2 led to a downregulation of the RARE-Luciferase reporter activity ( Fig. 5f and Supplementary Fig. 3e), whereas overexpressing Ehmt2 (or Ehmt1) stimulates RA signalling ( Fig. 5g and Supplementary Fig.   5b). Co-transfection of Ehmt2 with Rere increases the RA response more than transfection of either one alone (Fig. 5h). siRNA-mediated knockdown of Ehmt2 inhibited Rere and Hdac1dependent activation of the RA pathway (Fig. 5i). In NIH3T3 cells, Ehmt2 was recruited at the RARE element of the Rarβ promoter after 1 hour of RA treatment while no such enrichment was observed in upstream regions ( Fig. 5j and Supplementary Fig. 5c). In mouse embryos Ehmt2 could also be detected at the RARE element present in both the Rarβ promoter and the RARE-LacZ reporter (Fig. 4o,p). Knockdown of Ehmt2 in NIH3T3 cells reduced the occupancy levels of Rere, Wdr5, Hdac1, Hdac2 and Pol II at the Rarβ promoter ( Fig. 5k,l and Supplementary Fig. 5d). In NIH3T3 cells, inhibition of Ehmt2 methyltransferase activity with UNC0638 (U38) or UNC0646 (U46) 24, 25 did not alter RARE-Luciferase reporter activity or Rarβ mRNA expression ( Supplementary Fig. 5e,f). Also no difference in H3K9me1 nor H3K9me2 levels was observed by ChIP after 1 hour of RA treatment suggesting that Ehmt2 function is independent from its methyltransferase activity ( Supplementary Fig. 5g,h). This suggests that Ehmt2 could function as a scaffold protein to stabilize the WHHERE complex at RA regulated genes to allow Pol II loading. Thus, these results indicate that Ehmt2 acts together with the WHHERE coactivator complex in the RAdependent control of somite bilateral symmetry.
The WHHERE complex contains Wdr5, which is part of complexes such as ATAC, MOF and MLL, that are involved in histone acetylation and H3K4 methylation, which are chromatin modifications associated with transcriptional activation 26 . Despite the requirements of HDAC enzymatic activity for WHHERE function, we observed an increase of the levels of acetylated H3 and H4 as well as H3K27ac on the Rarβ promoter upon 1 hour of RA treatment of NIH3T3 cells (Fig. 6a-c). This suggests that Hdac1 and Hdac2 might act on nonhistone substrates or both could participate to the stability of the WHHERE complex. The H3K36me3 mark, which is associated to transcription elongation, was also increased (Fig.   6d). We also found that in the same conditions, H3K27me3 is absent from the Rarβ promoter ( Fig. 6e), whereas H3K4me1 increases while H3K4me2 and H3K4me3 decrease (Fig. 6f-h).
MLL3 and MLL4 are complexes which contain Wdr5 and regulate the deposition of the H3K4me1 mark 27 . The MLL3 and MLL4 complexes have been shown to be involved in RAdependent transcription and Wdr5 might provide a link with the WHHERE complex 28 . Whether these complexes are also required for the WHHERE-dependent Pol II recruitment to the promoter of RA targets remains to be investigated.
The histone acetyltransferase Ep300 (p300) has been shown to acetylate Hdac1 leading to a decrease of its deacetylase activity 29 . In the presence of RA, overexpression of Ep300 in NIH3T3 cells decreased RA signalling and inhibited Hdac1-dependent RA activation (Fig.   7a). In line with this, transfection of an Hdac1 mutant form resistant to Ep300 acetylation (H1-6R) 29 activated more strongly the RA pathway than wild-type Hdac1 (H1-WT) (Fig. 7b).
Moreover, Rarα-and Rere-dependent RA activation was inhibited following Ep300 overexpression whereas transfection of Kat2a (Gcn5) increased RA signalling more than Rarα or Rere alone (Fig. 7c,d). While treatment of fibroblast cultures with siRNAs against Rere or Kat2a in the presence of RA decreased RA reporter activity, knockdown of Ep300 did not affect the RA pathway (Fig.7e). Similarly in Ep300 mutant embryos (Ep300 -/-) 30 , the RARE-LacZ reporter 22 expression appeared normal (Fig. 7f,g) and somitogenesis progressed symmetrically (Fig. 7h,i). Altogether these results demonstrated that Ep300 negatively regulates Hdac1 activation of RA signalling, and Kat2a can participate together with the WHHERE complex in the activation of the RA pathway. Strikingly most of the members of the WHHERE complex including Rere, Hdac1, Hdac2 and also Ehmt2 have been generally associated to transcriptional repression [34][35][36] . Studies like those reported here may be necessary to fully understand the functions of coregulators, which may act as coactivators or corepressors depending on the developmental context and genes involved. Rere was shown to act as a coactivator for RA signalling 8 and Hdac1/Hdac2 are required for transcriptional activation of the MMTV promoter downstream of the glucocorticoid receptor 29 . Ehmt2 has also been shown to participate in the positive regulation of a subset of glucocorticoid receptor-regulated genes 37,38 . Our data shows that in the context of RA signalling these transcriptional repressors can form a complex, which acts as a coactivator of the pathway. The positive role of these proteins in signalling mediated by other nuclear receptors suggests that the WHHERE complex could also have a broader role as a coactivator downstream of nuclear receptor signalling. The positive regulation of RA signalling by Hdac1 and Hdac2 identified in this work might have significant implications for the potential use of HDAC inhibitors for the treatment of RA-sensitive cancers 39 .

METHODS
Methods and any associated references are available in the online version of the paper.   In all graphs data represent mean ± s.e.m.. NS -not significant, *P < 0.05 and **P < 0.01.  In all graphs data represent mean ± s.e.m. unless otherwise specified. NS -not significant, *P < 0.05 and **P < 0.01. hours.
(j) ChIP of the Rarβ promoter with a specific antibody for Ehmt2 using NIH3T3 cells treated or not with 1 µM RA during 1 hour (n = 3).
(k, l) ChIP analysis of the Rarβ promoter in NIH3T3 cells transfected with siRNA for Ehmt2 and treated with 1 µM RA during 1 hour. ChIP was performed with antibodies specific to Rere, Wdr5, Hdac1 and Hdac2 (k) or Pol II (l) (n = 3).