Retinoic acid (RA) plays key roles in cell differentiation and growth arrest by activating nuclear RA receptors (RARs) (α, β and γ), which are ligand-dependent transcription factors. RARs are also phosphorylated in response to RA. Here, we investigated the in vivo relevance of the phosphorylation of RARs during RA-induced neuronal differentiation of mouse embryonic stem cells (mESCs). Using ESCs where the genes encoding each RAR subtype had been inactivated, and stable rescue lines expressing RARs mutated in phospho-acceptor sites, we show that RA-induced neuronal differentiation involves RARγ2 and requires RARγ2 phosphorylation. By gene expression profiling, we found that the phosphorylated form of RARγ2 regulates a small subset of genes through binding an unusual RA response element consisting of two direct repeats with a seven-base-pair spacer. These new findings suggest an important role for RARγ phosphorylation during cell differentiation and pave the way for further investigations during embryonic development.
Retinoic acid (RA), the active metabolite of vitamin A, regulates multiple biological processes and plays key roles in embryonic development through the regulation of cell growth and differentiation (Samarut and Rochette-Egly, 2012). RA exerts its effects through nuclear RA receptors (RARs), which are ligand-regulated transcription factors with a well-defined domain organization and structure, consisting of a variable N-terminal domain (NTD), and two well-structured and conserved domains, a central DNA-binding domain (DBD) and a C-terminal ligand-binding domain (LBD) (Bastien and Rochette-Egly, 2004; Rochette-Egly and Germain, 2009).
The RA signalling system is highly complex as it comprises three subtypes (RARα, RARβ and RARγ), and for each subtype there are at least two isoforms, differing in their NTD (Chambon, 1996). Moreover, RARs act as heterodimers with another family of nuclear receptors, the retinoid X receptors (RXRs) (Germain et al., 2006). According to the canonical model, RXR–RAR heterodimers control the expression of their target genes involved in cell growth and differentiation through binding to specific and polymorphic RA response elements (RAREs) located in their regulatory regions (Moutier et al., 2012) and through the dynamic association and dissociation of coregulator complexes (Rosenfeld et al., 2006).
During the past decade, this scenario became more complicated with the discovery that RA also has non-genomic effects and induces the rapid activation of the MAPK pathway (Piskunov and Rochette-Egly, 2012; Stavridis et al., 2010), and that RARs are phosphoproteins (Al Tanoury et al., 2013b; Rochette-Egly, 2003). Indeed, we have shown that the RA-activated kinases induce RARs phosphorylation at a serine residue located in the NTD (Bastien et al., 2000; Rochette-Egly et al., 1997). Interestingly, this serine residue is conserved between RARs and during evolution (Samarut et al., 2011), emphasizing that phosphorylation of RARs might be important in RA signalling. However, the in vivo relevance of the RAR phosphorylation still remains to be defined, especially in the context of development. As cell differentiation is one of the most crucial steps during development, mouse embryonic stem cells (mESCs) (Gudas and Wagner, 2011) provide an interesting model to study the influence of RAR phosphorylation. Indeed, ESCs are pluripotent cells that self-renew indefinitely and can differentiate in vitro into a large variety of cell types (Wilson et al., 2009), such as neuronal cells, in response to RA (Bibel et al., 2007).
Here, by using mESCs specifically lacking each RAR subtype and stable rescue cell lines expressing RAR phosphomutants, we demonstrate that their neuronal differentiation crucially involves the phosphorylated form of RARγ2. Finally, genome-wide RNA-seq experiments identify direct target genes whose expression is controlled by phosphorylated RARγ2. These results suggest that there is an important role for RAR phosphorylation in RA signalling during cell differentiation, and pave the way for further investigations during embryonic and tissue development.
RARγ2 mediates the RA-induced neuronal differentiation of mESCs
When cultured as cellular aggregates, treated with RA and then dissociated and plated according to the protocol of Bibel et al. (Bibel et al., 2007; Bibel et al., 2004), pluripotent mESCs first adopted a spindle-shape morphology characteristic of neural progenitors (Fig. 1A, panel b) (Bibel et al., 2004). Then within 2 days these progenitors gave rise to glutamergic neuronal cells (Fig. 1A, panels c and d) and formed a dense axon network expressing class III β-tubulin (Tuj1) by 7–9 days (Fig. 1B). Concomitantly, the markers for pluripotency [Oct4 (also known as Pou5f1) and Nanog] were downregulated, whereas those for neuronal differentiation (Pax6) were increased (Fig. 1C,D).
As this entire process is initiated by activation of RARs, we investigated which specific RAR subtype (α, β or γ) is required to drive differentiation. ESCs constitutively expressed the RARα and RARγ proteins and their levels were not significantly affected upon RA treatment (Fig. 2A). In contrast, the RARβ protein was hardly detectable but was induced in response to RA (Fig. 2A). Such data suggest that the RARα and/or RARγ subtypes might be the primary RA targets for mediating the initial phase of neuronal differentiation. To investigate the role of these receptors in the RA-driven neuronal differentiation of mESCs, we used cell lines in which the Rara or Rarg genes (encoding RARα and RARγ, respectively) were disrupted by homologous recombination (Lohnes et al., 1993; Lufkin et al., 1993) (Fig. 2B). Our results show that, like wild-type (WT) ESCs, cells lacking RARα became neuronal progenitors (Fig. 2C) that gave rise to neurons forming a dense axon network (Fig. 2D). In contrast, cells lacking RARγ did not differentiate into cells of the neuronal lineage. They did not become neuronal progenitors (Fig. 2C) and axon structures did not appear even up to 15 days after RA addition (Fig. 2D). These results indicate that RARγ is essential for the RA-induced commitment of mESCs into neuronal precursors that give rise to neurons. They also indicate that RARα and RARβ cannot functionally compensate for loss of RARγ during this process.
Seven murine RARγ isoforms (mRARγ1 to mRARγ7) have been characterized so far, generated by alternative splicing of at least seven exons and differing in their NTD (Kastner et al., 1990). In mESCs, RARγ1 and RARγ2 are the predominant isoforms, whereas the others (RARγ3 to RARγ7) have not been detected (Kastner et al., 1990). Moreover, both RARγ1 and RARγ2 can be detected at the protein level in mESCs as assessed by immunoblotting after immunoprecipitation with antibodies recognizing specifically each isoform (Bastien et al., 2000) (Fig. 2E).
To investigate the role played by each isoform in the RA-induced neuronal differentiation of mESCs, stable rescue lines expressing the RARγ2 or RARγ1 proteins were established from Rarg−/− cells. Several clones were obtained for each ‘rescue’ transgene, and these expressed RARγ1 or RARγ2 at levels similar to the endogenous receptors (Fig. 2F; supplementary material Fig. S1A). We investigated the ability of the RARγ1 and RARγ2 rescue lines to differentiate in response to RA. Remarkably, the RARγ2 rescue lines became neuronal progenitors (Fig. 2G) that gave rise to neurons forming a dense axon network (Fig. 2H). In contrast, the RARγ1-expressing lines did not give spindle-shaped neuronal progenitors (Fig. 2G) nor any axon structures (Fig. 2H). These results suggest that the RARγ2 isoform mediates the effects of RA for the commitment of mESCs into the neuronal lineage.
RARγ2 phosphorylation is required for the neural differentiation of mESCs
RARs are phosphoproteins and RARγ2 comprises two serine (S) residues (S66 and S68) that are located in a proline-rich motif of the NTD and are substrates for phosphorylation (Bastien et al., 2000) (Fig. 3A). We have previously shown that RARs become rapidly phosphorylated at one of these serine residues (S68 in RARγ2) subsequent to a kinase cascade initiated by MAPKs after RA treatment (Bruck et al., 2009; Lalevée et al., 2010). In mESCs, Erk1/2 (MAPK3 and MAPK1, respectively) were rapidly (within minutes) activated after RA addition (Fig. 3B). The amount of RARγ2 phosphorylated at S68 also rapidly increased, as assessed by phosphoprotein affinity purification followed by immunoblotting with antibodies specifically recognizing RARγ phosphorylated at this residue (Fig. 3C).
To investigate whether RARγ2 phosphorylation plays a role in mESC neuronal differentiation, we generated additional stable rescue lines expressing RARγ2 with S68 replaced with an alanine (RARγ2S68A) in the Rarg−/− background. Rescue lines expressing RARγ2 with the other serine residue replaced with an alanine were also established (RARγ2S66A) as a control. Biophysical studies (nuclear magnetic resonance, circular dichroism and small-angle X-ray scattering) coupled to molecular dynamics simulations performed with synthetic model peptides indicated that replacement of S68 (and S66) by an alanine residue did not affect the global hydrodynamic behaviour of the peptide (B. Kieffer, A. Dejaegere and C.R.E, unpublished data).
Several clones expressing the mutated receptors at levels similar to the endogenous were obtained (Fig. 3D; supplementary material Fig. S1B) and analysed for their differentiation potential. Remarkably, the RARγ2S68A rescue cell lines did not differentiate into neurons. No axon structures were obtained by 7–9 days and even up to 17 days (Fig. 3E) after RA addition. In contrast, the RARγ2S66A rescue cell line was able to form a dense axon network comparable to the RARγ2WT rescue line (Fig. 3E). Collectively, these results indicate that RARγ2 phosphorylation at S68 plays an important role in mediating RA signalling.
Genome-wide analysis of the early RA target genes that are regulated by RARγ2
Then the question was whether the RA-induced neuronal differentiation of mESCs correlates with the activation and/or repression of a subset of genes. Therefore, we profiled the genes that were regulated by RARγ2 by comparing the WT mESCs to the Rarg−/− and RARγ2WT cell lines in the absence and presence of RA. RNA-seq was performed on cellular aggregates obtained from each cell line with or without a RA treatment of 2 hours, in order to identify the primary (early) RARγ2 target genes involved in the commitment to the neuronal lineage. Transcripts whose expression was induced or repressed by RA (experiments designated RA/Ctrl) were then assessed (supplementary material Table S1).
The list of genes that were upregulated by RA in the WT mESCs was in agreement with other studies (Mahony et al., 2011; Moutier et al., 2012; Simandi et al., 2010) and included the gene encoding the RARβ2 isoform, the canonical RA target genes involved in RA metabolism (Cyp26a1, Cyp26b1, Cyp26c1 and Dhrs3), patterning genes exemplified by the Hox genes (Hoxa1, Hoxa3, Hoxa5, Hoxb1 and Hoxb4), other genes encoding homeobox proteins (Meis2, Cdx1, Gbx2, Dlx3 and Hnf1b) and genes with a wide variety of functions such as Lefty1, Arg1 and Stra8. Only a few genes were downregulated, and included the Otx2 gene, also in agreement with other studies.
To assess which of these genes are specifically regulated by RARγ2 in the presence of RA, the results obtained with the WT cells were cross-referenced with those obtained with the Rarg−/− cell line and the RARγ2 WT rescue cell line. The Venn diagram (Fig. 4A) shows that 37% of the genes upregulated in RA-treated WT cells were attenuated in the Rarg−/− cells and fully restored in the RARγ2WT rescue line, showing that they are RARγ2 target genes. These genes are exemplified by the Meis2, Lefty1, Hnf1b, Arg1, Hoxa3 and Hoxa5 genes, and their regulation by RARγ2 was corroborated in RT-qPCR experiments (Fig. 5). The Gbx2 gene can also be added to this list because its activation was significantly decreased (but not below the selected cut-off of the RNA-seq analysis) in the Rarg−/− cell line and fully restored in the RARγ2WT rescue cell line (supplementary material Table S1; Fig. 5), indicating that it is a RARγ2 target.
It is worth noting that 13% of the upregulated genes were also attenuated in the Rarg−/− cells but not restored in the RARγ2 rescue line, suggesting that these genes might be regulated by the RARγ1 isoform. Given that RARγ1 does not restore differentiation of mESCs, these genes were considered not to be essential in our model. Finally, 50% of the upregulated genes were not grossly affected in absence of RARγ, suggesting that they are also not essential for the differentiation of mESCs. Among these genes are the canonical RA target genes involved in RA metabolism (Cyp26a1, Cyp26b1 and Dhrs3). There is also the gene encoding the RARβ2 isoform, corroborating our first hypothesis that its expression is not essential for neuronal differentiation of ESCs. Remarkably, additional studies performed with mESCs lacking RARα indicated that the activation of RARβ2 is mediated by the RARα subtype (supplementary material Fig. S2). In contrast, activation of Cyp26a1 involves the two receptors (supplementary material Fig. S2) and, hence, RARα might functionally compensate for RARγ loss. Interestingly, expression of several genes was upregulated in the Rarg−/− cells compared to WT cells, suggesting a repressive effect of RARγ. Several others were activated only in the RARγ2 rescue cell line, suggesting that in the WT cells, the function of RARγ2 might be antagonized by RARγ1.
In conclusion, these experiments revealed a repertoire of genes that are specifically regulated by RARγ2 in mESCs. It also highlighted the complexity of the regulation of the other RA target genes.
The phosphorylated form of RARγ2 regulates a subset of genes
Given that RARγ2 phosphorylation is required for the neuronal differentiation of mESCs, we investigated the genes that are controlled by the phosphorylated form of this receptor. Cellular aggregates obtained from mESCs expressing RARγ2S68A were treated with RA for 2 hours as above and gene expression assessed by RNA-seq (supplementary material Table S1). The list of genes with altered expression was compared with the list of RARγ2 target genes generated above. The Venn diagram (Fig. 4B) shows that ∼50% of RARγ2-target genes are not expressed in the RARγ2S68A recue cell line, suggesting that their activation is controlled by the phosphorylation of RARγ2. These genes include Lefty1 (left right determination factor 1) and the homeobox gene Hnf1b (hepatocyte nuclear factor 1 homeobox B) and their phospho-RARγ2 dependency was corroborated in RT-qPCR experiments (Fig. 5). The homeobox genes Gbx2 (gastrulation brain homeobox 2) and Meis2 can be added to this list. Indeed Gbx2 activation, which was decreased in the Rarg−/− cells (but not abolished), was well restored in the RARγ2WT rescue line but not in the RARγ2S68A line. Concerning Meis2, the values obtained with the RARγ2S68A rescue line were at the limit of the threshold, suggesting that this gene might be also an interesting candidate. RT-qPCR experiments corroborated that both Gbx2 and Meis2 are regulated by the phosphorylation of RARγ2 (Fig. 5). Taken together, these results suggest that the phosphoRARγ2 programme would trigger the neuronal differentiation of mESCs via the activation of a particular set of key genes including at least Lefty1, Gbx2, Meis2 and Hnf1b.
Phosphorylation controls the recruitment of RARγ2 to gene promoters
Next, we asked whether phosphorylation controls the recruitment of RARγ2 to the promoters of the selected genes. We analysed the RAR-binding sites mapped in the Lefty1, Meis2, Gbx2 and Hnf1b genes in ChIP-seq experiments performed with cellular aggregates treated or not with RA for 2 hours and with a pan-RAR antibody (Moutier et al., 2012). We identified one or several major RAR-binding sites that were enriched after RA addition (supplementary material Table S2; Fig. 6). The sequences under these peaks were analysed to detect consensus 5′-RGKTSA-3′ half sites (with R = A/G, K = G/T and S = C/G, according to the IUPAC convention) with either no (DR0) or 1- to 8-base-pair (DR1–DR8) spacers (Moutier et al., 2012). Then primers specifically surrounding the RAREs were designed and used in ChIP-qPCR experiments performed with cellular aggregates from the RARγ2WT and RARγ2S68A rescue cell lines and with our purified specific RARγ antibodies (Lalevée et al., 2010; Mendoza-Parra et al., 2011).
One major RA-regulated RAR-binding site was found in the promoter of the Lefty1 gene in the −10 kb region flanking the transcription start site (TSS). This site comprises a consensus DR7 element and two DR2s with one mismatch (supplementary material Table S2). Our ChIP experiments revealed that, after RA addition, occupancy of the DR7 sequence was markedly enriched with RARγ2WT, but not with the phosphomutant (Fig. 6A). Occupancy of the DR2 sequences was also enriched, but less markedly and without significant difference between RARγ2WT and RARγ2S68A. ChIP experiments were also performed with RARα antibodies, using the same chromatin preparations and the same primers. However, none of the sequences were significantly occupied in both the RARγ2WT and RARγ2S68A cell lines (Fig. 6A). Collectively, these results indicate that the activation of Lefty1 requires the binding of the phosphorylated form of RARγ2 to a DR7 element.
A major RAR-binding site was also identified in the −10 kb region flanking the TSS of the Gbx2 gene. It contains a DR7, a DR5 and a DR2 element, all with one mismatch (supplementary material Table S2). As for the Lefty1 gene, occupancy of the DR7 element was enriched with RARγ2WT, but not with the phosphomutant after RA addition (Fig. 6B). Similar results were obtained with the sequence encompassing the DR2 and DR5 elements. ChIP experiments performed with RARα antibodies showed that, in both the RARγ2WT and RARγ2S68A cell lines, all elements were also enriched with RARα, but less markedly (Fig. 6B), explaining at least in part why the activation of the Gbx2 gene was not fully abrogated upon RARγ invalidation in our RNA-seq experiments.
Concerning the Hnf1b gene, three RAR-binding sites were found, one upstream of the TSS with a consensus DR7 sequence and two downstream of the TSS (one intragenic with a DR5 and one downstream of the gene end with a DR2) (supplementary material Table S2; Fig. 6C). Occupancy of the DR2 element was not significantly enriched with RARγ2 either WT or S68A. However, occupancy of the DR7 and DR5 elements was enriched by RARγ2WT but not by the phosphomutant after RA addition (Fig. 6C). It is worth noting that, compared to the DR7 element, the DR5 was more markedly enriched by RARγ2 and was not occupied by RARα, suggesting that the regulation of the Hnf1b gene by the phosphorylated form of RARγ2 concerns this DR5 RARE. In contrast, the Meis2 locus comprises two major RAR-binding sites located downstream of the TSS in intragenic regions. One peak contains a consensus DR5 and the other a consensus DR2 element (supplementary material Table S2; Fig. 6D). Both elements were enriched with RARγ2WT after RA addition, but only the DR5 element was not enriched with the RARγ2 phosphomutant (Fig. 6D).
Taken together, these results highlight novel RARγ2-binding elements with 5- or 7-base-pair spacers, which specifically recruit the phosphorylated form of RARγ2 in response to RA.
RA-regulated differentiation, proliferation, survival or death is at the basis of complex physiological processes such as development. During the past decade, numerous studies have provided an enormous gain of knowledge into the molecular and structural features of RARs. However, the complexity of the scenarios increased with the discovery that RA also has non-genomic effects and activates several signalling pathways that are integrated in the nucleus and target several proteins, including RARs, for phosphorylation processes (Al Tanoury et al., 2013b). Thus, we are far from understanding how the non-genomic and genomic effects of RA coordinate in embryonic development and organogenesis.
As cell differentiation is one of the most crucial steps during development, we have used mESCs, which can differentiate into a large variety of cell types, such as neuronal cells, in response to RA. However, whether a particular RAR subtype was involved in the differentiation of these cells was still undefined. Moreover, the relevance of RAR phosphorylation had also not been addressed. Therefore, we investigated which RAR subtype is involved in neuronal differentiation of mESCs and whether their phosphorylation is required. With that aim, we took advantage of mESCs in which the different RARs have been inactivated and used them to generate rescue lines expressing RAR phosphomutants.
The first novel finding of this study is that RARγ, and more precisely the RARγ2 isoform, which is the predominant RARγ isoform in embryonic stem cells (Kastner et al., 1990), is required for the RA-induced generation of neuronal progenitors, which then give rise to neurons forming a dense axon network. An additional important observation is the identification of a subset of early RA target genes involved in development whose regulation is lost in the RARγ-knockout cells, but well re-established in the RARγ2 rescue line. Because these genes are activated as early as 2 hours after RA addition to the cell aggregates, one can hypothesize that they are involved in the commitment of ESC to the neuronal lineage. A recent study conducted by Kashyap et al. (Kashyap et al., 2013) has also shown that inactivation of RARγ is associated with a reduced expression of several genes. However, the cells were not analysed under the same conditions as here and there was no correlation with neuronal differentiation.
The second novel finding is the role of the phosphorylation state of RARγ2 (Fig. 7). We show that ERKs are rapidly activated after RA addition in mESCs, corroborating other studies (Stavridis et al., 2010). Today, it is known that the rapid activation of the kinase signalling pathways by RA is mediated by non-genomic effects, and that the effects of these kinases are integrated in the nucleus where they phosphorylate several factors involved in the expression of the RA target genes (Al Tanoury et al., 2013b). Here, the important point is that RARγ2 becomes phosphorylated at its N-terminal serine residue, and that this phosphorylation process is required for the neuronal differentiation of mESCs. Indeed our present data demonstrate that mESCs expressing RARγ2 mutated at the N-terminal serine residue are not able to enter the neuronal lineage. They also reveal a small set of target genes involved in neuronal development and exemplified by Lefty1, Gbx2, Meis2 and Hnf1b, which require the phosphorylation of RARγ2 to be activated. This is the first report showing that RA target genes involved in development are regulated by the phosphorylation of RARγ2.
The third novel finding is the characterization of RAREs with specific spacings, which recruit the phosphorylated form of RARγ2 in response to RA. Indeed, we found that the promoters of the above genes targeted by phosphoRARγ2 depict not only canonical DR2 and DR5 response elements but also atypical DR7 elements, and that only the DR5 and DR7 elements recruit specifically the phosphorylated form of RARγ2 in response to RA. The next challenge will be to determine how phosphorylation of the N-terminal serine residue, which is located in the vicinity of the DBD, regulates the binding of RARγ2 only to response elements with DR5 or DR7 spacing. However, it is worth pointing out that the spacing of the response elements directs the architecture of the DNA-bound RXR–RAR heterodimers (Brélivet et al., 2012; Rochel et al., 2011). Because phosphorylation of the N-terminal serine residue induces the dissociation of partners with SH3 domains (Lalevée et al., 2010), preliminary studies have suggested that the interaction of the non-phosphorylated form with SH3 proteins would impede the binding of RARγ2 only to DRs with specific spacing (DR5 and DR7), whereas the phosphorylated form (without SH3 partners) would be recruited to these DRs in response to RA (B. Kieffer and C.R.-E., unpublished results).
It is important to note that the DR7 elements show a rather low frequency (Moutier et al., 2012) among the RAR-occupied sites, suggesting that they would control the expression of a small set of genes involved in the commitment of pluripotent cells to specific lineages. Here, we have defined some of these genes, i.e. the Lefty1, Gbx2 and Hnfb1 genes, and we show that they belong to an early phosphoRARγ2-regulated gene programme and are crucial for the loss of pluripotency and for triggering the neuronal differentiation of mESCs. Remarkably, comparison of our findings with those reported for other cell types highlights that Lefty1 and Gbx2, which are also known as the ‘stimulated by RA’ Stra3 and Stra7 genes, respectively (Chazaud et al., 1996; Oulad-Abdelghani et al., 1998), are involved in early neural development (Li et al., 2009; Smith et al., 2008; Sunmonu et al., 2011) and, thus, are also activated in the P19 cell line that differentiates into neurons in response to RA. However, they are absent from the lists of RA-activated genes in cells with other features, such as mouse embryo carcinoma cells (F9 cell line) (Lalevée et al., 2011; Mendoza-Parra et al., 2011; Su and Gudas, 2008), MCF7 cells (Hua et al., 2009) and mouse embryonic fibroblasts (Al Tanoury et al., 2013a). Collectively such data highlight the importance of these genes for commitment to the neuronal lineage. Concerning the Hnf1b and Meis2 (also known as Stra10; Oulad-Abdelghani et al., 1997) genes, they are also involved in neuronal differentiation, but are not restricted to this cell type because they are expressed in F9 cells and other developing tissues (Mendoza-Parra et al., 2011; Oulad-Abdelghani et al., 1997). Nevertheless, all these data suggest that RARγ2 phosphorylation would control neuronal differentiation via the activation of target genes with specific DR7 and/or DR5 element combinations.
In attempts to validate the role of the early phosphoRARγ2-regulated genes in the neuronal differentiation of mESCs, we monitored the effects of shRNA-mediated Gbx2 and Meis2 knockdown on the appearance of axons after RA treatment. In fact, despite the efficiency of the knockdown, the neuronal differentiation of the cells was not significantly affected. Thus knockdown of one of these transcription factors alone is not sufficient to abolish the differentiation programme. Therefore, either the residual levels of expression are sufficient to sustain differentiation, or the genes act together to establish a sub-programme of interconnected regulatory networks that are not strongly perturbed by the loss of only one of these genes. Further studies involving simultaneous knockdown of several or all of these genes will be required to answer this question. It is also important to note that almost 50% of the RA-induced genes are still normally activated in the RARγ-knockout cells although these cells do not differentiate into neurons. This again highlights the crucial role played by the small subset of genes specifically induced by the RARγ2 isoform.
In conclusion, we provide several lines of evidence for a function of RARγ2 phosphorylation in positively regulating neuronal differentiation (Fig. 7). It would be interesting to extrapolate this work to other RA responsive systems. Ultimately, one can predict specific sets of genes with DR5 and/or DR7 elements that could be controlled by phosphorylation of RARs, depending on the feature of the cells.
MATERIALS AND METHODS
All constructs containing the RARγ receptors were cloned into the pCA vector, which is driven by a CAG early promoter, coupled to hygromycin/neomycin resistance. mRARγ2 and mRARγ1 were isolated as XhoI/BamH1 fragments from the pSG5 constructs (Bastien et al., 2000) and subcloned in the same sites of pCA. mRARγ2S66A and mRARγ2S68A in pCA were constructed by PCR amplification reactions. Internal oligonucleotides used in the PCR reaction encoded alanine (A) instead of serine (S) residues at positions 66 or 68.
Rabbit polyclonal antibodies recognizing RARα [RPα(F)], RARβ [RPβ(F) and RARγ [RPγ(F)] were as described previously (Bastien et al., 2000; Bruck et al., 2009; Rochette-Egly et al., 1992). RPγ(F) was purified on sulfoLink gel columns (Pierce Chemical) coupled to the corresponding immunizing peptide (Lalevée et al., 2010). Mouse monoclonal antibodies specifically recognizing RARγ phosphorylated at position S68 were as previously described (Lalevée et al., 2010), as were mouse monoclonal antibodies recognizing specifically the RARγ1 [Ab1γ(A1)] and RARγ2 [Ab10γ(A2)] isoforms (Bastien et al., 2000). Rabbit polyclonal antibodies against GAPDH were from Sigma-Aldrich and those against Pax6, Oct4 and Nanog were from Abcam Ltd Mouse monoclonal antibodies recognizing neuronal class III β-tubulin (Tuj1) were from Eurogentec France. The antibodies against ERKs and their active phosphorylated forms were from Santa Cruz Biotechnology.
Mouse ESC lines and culture conditions
Mouse ESCs (clone D4), derived from the 129 sub-strain were as previously described, as were the Rarg−/− (clone AA71) and Rara−/− (clone KC25) cell lines (Lohnes et al., 1993; Lufkin et al., 1993). To establish the rescue lines, the RARγ2WT, RARγ2S66A or RARγ2S68A constructs were introduced into the Rarg−/− cells by Lipofectamine® 2000 transfection (Invitrogen). The stable rescue lines were selected with G418 (350 µg/ml) or hygromycin B (100 µg/ml) for 1 week and analysed for the presence of the transgene by qPCR and western blotting.
All cell lines were grown on inactivated mouse embryonic fibroblast feeder cells in the presence of LIF under standard conditions [DMEM supplemented with Glut Amax™-I (Fischer Scientific SAS), 15% FCS, non essential amino acids and β-mercaptoethanol]. Then cells were differentiated into neurons according to the protocol of Bibel et al. (Bibel et al., 2007). In brief, cells were trypsinized, plated on non-adhesive bacteriological Greiner Petri dishes (4×106 cells) in 15 ml CA (cellular aggregates) medium (DMEM supplemented with GlutaMAX™-I, 10% FCS, non essential amino acids and β-mercaptoethanol) and aggregated for 8 days with a medium change every 2 days. At day four, all-trans RA (2 µM) (Sigma-Aldrich Chimie SARL) was added and 4 days later, the cellular aggregates were washed with PBS, dissociated with trypsin, suspended in N2 medium [DMEM/Ham-F12, BSA and N2® (Fischer Scientific SAS)], and plated on culture dishes precoated with PORN (Poly-DL-ornithine hydrobromide, Sigma-Aldrich, Chimie SARL) and laminin (Roche Diagnostics). After 1 day, the N2 medium was changed and 2 days later replaced by neurobasal medium supplemented with B27® (Fisher Scientific).
Immunoblotting, immunoprecipitation and immunofluorescence assays
Extract preparation and immunoblotting were as described previously (Bour et al., 2005). Immunoprecipitation was performed with mouse monoclonal antibodies immobilized on Dynabeads® Protein A/G (Invitrogen). For immunofluorescence assays, cells were grown on Lab-Tek® glass chamber slides (Thermoscientific), fixed in 4% formaldehyde (PFA)-PBS (20 min), permeabilized with 0.1% Triton X-100 (15 min) and blocked with 3% non-immune serum in PBS (30 min). Then, cells were incubated with primary antibodies, followed by Alexa-Fluor-448- or Alexa-Fluor-555-conjugated secondary antibodies (Invitrogen). Nuclei were counterstained with DAPI (Sigma-Aldrich). Cells were analysed by fluorescence microscopy using a LEICA DMRX microscope equipped with a LEICA True Confocal Scanner TCS SP.
Detection of phosphorylated RARγ
Cell extracts were applied to phosphoprotein purification columns (Qiagen). Column eluates containing protein peaks were concentrated and analysed by immunoblotting as previously described (Bruck et al., 2009).
RNA extraction and RT-qPCR
Total RNA was extracted from cellular aggregates grown for 4 days in the absence of RA and then treated with RA for 2–6 hours. Aliquots were subjected to RT-qPCR as described (Bruck et al., 2009). Transcripts were normalized according to the housekeeping gene GAPDH. Primer amplification and specificity were verified on DNA serial dilutions. Primer sequences are available from the corresponding author upon request.
High-throughput mRNA sequencing
After isolation of total RNA, a library of template molecules suitable for high-throughput DNA sequencing (RNA-Seq) was created following the Illumina ‘Truseq RNA sample prep v2’ protocol with some modifications. Briefly, mRNA was purified from 2 µg total RNA using oligo-dT magnetic beads and fragmented using divalent cations (94°C, 8 minutes). The cleaved mRNA fragments were reverse-transcribed to cDNA using random primers and then the second cDNA strand was synthesized using Polymerase I and RNase H. The next steps of RNA-Seq library preparation were performed in a fully automated system using SPRIworks Fragment Library System I kit (ref A84803, Beckman Coulter, Inc.) with the SPRI-TE instrument (Beckman Coulter, Inc.). Briefly, in this system, double-stranded cDNA fragments were blunted, phosphorylated and ligated to indexed adapter dimers, and fragments in the range of ∼200–400 bp were selected. Finally, the library was amplified by PCR [30 s at 98°C (10 s at 98°C, 30 s at 60°C, 30 s at 72°C)×12 cycles; 5 min at 72°C] and the surplus PCR primers were removed using AMPure beads (Agencourt Biosciences Corporation) with the Biomek 3000 instrument (Beckman Coulter, Inc.). DNA libraries were checked for quality and quantified using 2100 Bioanalyzer (Agilent). The libraries were loaded in the flow cell at 11 pM concentration and clusters were generated and sequenced in the Illumina Hiseq2000 as single-end 50-base reads.
Image analysis and base calling were performed using CASAVA v1.8.2 sequence reads were mapped onto the mm9 assembly of the mouse genome by using Tophat v1.4.1 (Trapnell et al., 2009) and the bowtie v0.12.7 aligner. Only uniquely aligned reads have been retained for further analyses. Gene expression was quantified using HTSeq v0.5.3p3 (Anders and Huber, 2010) and gene annotations from Ensembl release 66. Read counts were normalized across libraries with the method proposed by Anders and Huber (Anders and Huber, 2010). Comparisons of interest were performed using the statistical method proposed by Anders and Huber (Anders and Huber, 2010) implemented in the DESeq v.1.6.1 Bioconductor package. P-values are adjusted for multiple testing by using the Benjamini and Hochberg (Benjamini and Hochberg, 1995) method. Only genes with |log2 fold-change|>1 or <−1 and an adjusted P<0.05 were considered. Functional analyses of these genes were performed using the Manteia program (http://manteia.igbmc.fr).
The gene regions located ±10 kb from gene limits (Ensembl release 63) were analysed using regular expression search to detect perfect consensus 5′-RGKTSA-3′ half sites with different spacing. The potential RAR-binding elements were aligned on the same strand to ensure the sense and antisense matches gave homogeneous positions.
Chromatin immunoprecipitation experiments
Cellular aggregates were treated with RA for 45 minutes and chromatin immunoprecipitation (ChIP) experiments were performed as previously described (Bruck et al., 2009). Control ChIP were performed without antibodies, and RARγ was immunoprecipitated with purified RPγ(F) immobilized on Dynabeads® Protein A (Invitrogen). RARα was also immunoprecipitated as previously described (Bruck et al., 2009). Immunoprecipitated DNA was amplified by PCR primers designed using Primer3 software (Rozen and Skaletsky, 2000), which are available upon request. Occupancy of the promoters was calculated by normalizing the PCR signals from the immunoprecipitated samples to the signals obtained from the input DNA.
We are grateful to Jean-Marie Garnier for the RARγ constructs and to M. Oulad Abdelghani (IGBMC) for the mouse monoclonal antibodies. Special thanks to Marie Hestin, Regis Lutzing and the cell culture facilities for help.
Z.A.T. and A.P. devised and undertook all experimental work and analyzed the data. A.D. generated the KO cell lines. S.G. performed the RT-qPCR and ChIP experiments. S.U. and I.D. performed and analyzed the ChIP-seq experiments. B.J. performed the RNA-seq experiments; T.Y. and C.K. performed the bioinformatic analysis of the results. C.R.E. analyzed the data and wrote the paper.
This work was supported by the Centre national de la recherche scientifique (CNRS); the Institut national de la santé et de la recherche médicale (INSERM); the Agence Nationale pour la Recherche [grant numbers ANR-05-BLAN-0390-02, ANR-09-BLAN-0297-01, ANR-SVS8-11-Rarescales]; the Association pour la recherche sur le Cancer [grant numbers ARC-07-1-3169 and SL220110603474]; the Fondation pour la Recherche Médicale (FRM) [grant number DEQ20090515423], and the Institut National du Cancer [grant numbers INCa-PL09-194, PL07-96099]. A.P. was supported by FRM and the Lady TATA Memorial Trust, Z.A.T. by INCA, and S.U. by the Ministère de la Recherche. I.D. is an ‘équipe labélisée” of the Ligue Nationale contre le Cancer. The Institut de Génétique et de Biologie Moléculaire et Cellulaire (IGBMC) sequencing facility is a member of the ‘France Génomique’ consortium [grant number ANR10-INBS-09-08].
The authors declare no competing interests.