Fgfr3 Is a Transcriptional Target of Ap2δ and Ash2l-Containing Histone Methyltransferase Complexes

Polycomb (PcG) and trithorax (trxG) proteins play important roles in establishing lineage-specific genetic programs through induction of chromatin modifications that lead to gene silencing or activation. Previously, we described an association between the MLL/SET1 complexes and a highly restricted, gene-specific DNA-binding protein Ap2δ that is required for recruitment of the MLL/SET1 complex to target Hoxc8 specifically. Here, we reduced levels of Ap2δ and Ash2l in the neuroblastoma cell line, Neuro2A, and analyzed their gene expression profiles using whole-genome mouse cDNA microarrays. This analysis yielded 42 genes that are potentially co-regulated by Ap2δ and Ash2l, and we have identified evolutionarily conserved Ap2-binding sites in 20 of them. To determine whether some of these were direct targets of the Ap2δ-Ash2l complex, we analyzed several promoters for the presence of Ap2δ and Ash2l by chromatin immunoprecipitation (ChIP). Among the targets we screened, we identified Fgfr3 as a direct transcriptional target of the Ap2δ-Ash2l complex. Additionally, we found that Ap2δ is necessary for the recruitment of Ash2l-containing complexes to this promoter and that this recruitment leads to trimethylation of lysine 4 of histone H3 (H3K4me3). Thus, we have identified several candidate targets of complexes containing Ap2δ and Ash2l that can be used to further elucidate their roles during development and showed that Fgfr3 is a novel direct target of these complexes.


Introduction
PcG and trxG proteins act antagonistically to maintain heritable patterns of gene expression, with the former marking genes for repression and the latter for activation. PcG complexes are associated with trimethylation of histone H3 at lysine 27 (H3K27me3) whereas trxG complexes are linked to H3K4me3 [1,2]. This relationship embodies the characteristic of cellular memory to establish the identity in each cell type during development. Previously, these marks were considered to be static; recent evidence, however, has shown that these marks are involved in dynamic gene regulation through active recruitment of PcG and trxG complexes during cellular differentiation [2,3]. Studies using embryonic stem (ES) cells and neural and muscle progenitors reveal that these marks vary depending on the cell type and that the majority of these marks is present at the promoters of key developmental genes [3,4]. Furthermore, experiments that are based on chromatin immunoprecipitation coupled to DNA microarray analysis (ChIPchip) and the more recent ChIP-seq, in which enriched DNA is directly sequenced, reveal an association between the intensity of the H3K4me3 epigenetic mark at the promoter and active transcription [3]. Conversely, the presence of the H3K27me3 mark is associated with gene repression [3]. These data suggest that PcG and trxG proteins play a role in establishing lineage-specific genetic programs through induction of chromatin modifications.
The trxG protein group includes members of the MLL/SET1 family of histone lysine methyltransferases (HKMTs) and their associated proteins. The MLL/SET1 family consists of six members, Mixed Lineage Leukemia 1 (MLL1), MLL2 (ALR), MLL3 (HALR), MLL4, SET1A and SET1B, which share a catalytic SET domain that has been shown to have H3K4 methyltransferase activity [5,6,7,8]. MLL/SET1 proteins exist in multimeric complexes that contain three highly conserved subunits: Ash2l, RbBP5 and WDR5 [9]. Recently, it had been reported that these subunits are important for regulating the enzymatic activity of the SET domain-containing factor. Ash2l, in particular, was shown to be critical for H3K4me3 as downregulation of Ash2l leads to a genome-wide decrease in this epigenetic mark [10].
We recently reported that the gene-specific transcription factor Activating protein 2d (Ap2d) is important for the recruitment of MLL2 to the Hoxc8 locus during embryogenesis and that this recruitment leads to H3K4me3 and subsequent gene activation [11]. Ap2d is a member of the Ap2 family of sequence-specific DNA-binding proteins, which consists of Ap2a, -b, -c, -d and -e. Ap2 proteins bind a GC-rich consensus sequence that is found on a variety of cellular and viral enhancers. Ap2d is considered the most divergent family member, as it has a unique transactivation domain (TAD) that has been shown to specifically bind Ash2l. Additionally, Ap2d's gene, Tcfap2d, has a highly restricted expression pattern and is found in the developing myocardium, central nervous system and retina [12].
To search systematically for Ap2d-and Ash2l-regulated targets, we assessed the transcriptome using cDNA microarrays to identify genes whose expression was significantly decreased when either protein was diminished and then filtered those results using the criterion of an evolutionarily conserved Ap2-binding site in the promoter. To identify true targets of Ap2d and Ash2l-containing complexes among several candidates, we tested for the presence of Ap2d and Ash2l at their promoters. Among the genes we tested, we identified Fgfr3 as a direct target of Ap2d and Ash2l given that both proteins were present at the Fgfr3 promoter and that downregulation of either Ap2d or Ash2l resulted in a decrease of Fgfr3 expression. Thus, we provide evidence suggesting that Ap2d plays an important role in altering the epigenetic landscape of a set of developmentally regulated targets through recruitment of Ash2l-containing HKMT complexes.

Identification of Ap2d and Ash2l Target Genes by cDNA Microarray Analysis
To identify targets of Ap2d and Ash2l, we performed whole genome analysis of cDNA expression levels obtained from Neuro2a cells treated with either Tcfap2dor Ash2l-specific RNAi or a scrambled control. As shown in Fig. 1, treatment with either Tcfap2dor Ash2l-specific RNAi resulted in a significant decrease of their respective transcripts only. Previously, we had shown that treatment with Ash2l-specific RNAi resulted in a decrease in Ash2l protein only [11].
To systematically identify genes regulated by both Ap2d and Ash2l at the genome-wide scale, we obtained cDNAs from RNAitreated Neuro2a cells and performed microarray analysis using the GeneChipH Mouse Genome 430 2.0 Array. Triplicate microarray experiments were performed comparing signals obtained from cells treated with either Tcfap2dor Ash2l-specific RNAi to those of cells treated with a scrambled control. Signal values were calculated using the MAS5 and PLIER statistical algorithms. Genes that had a significance level of p,0.05 and a fold change greater than 1.1 were selected for analysis. Using signal values obtained from the MAS5 probe summarization algorithm, we identified 917 and 806 genes that were differentially expressed when Ap2d or Ash2l, respectively, was downregulated. Comparison of these two groups yielded 76 genes whose expression was significantly altered when Tcfap2d and Ash2l were knocked down individually ( Fig. 2A). Given that Ap2d and Ash2l form a complex that is involved in H3K4me3, we assumed that a reduction in either Ap2d or Ash2l would lead to decreased expression of their direct targets. Hence, to identify candidate targets of the Ap2d-Ash2l complex, we focused solely on genes that were downregulated when they were reduced. Of the 76 genes whose expressed was significantly changed when Ap2d and Ash2l were decreased, 33 genes with known function were downregulated. To determine whether we could identify additional targets that had not been previously identified by MAS5, we applied an alternative method using the PLIER probe summarization algorithm to obtain signal values. Through this method, we identified 9 additional genes that were downregulated when Ap2d and Ash2l levels were decreased. Altogether, 42 genes were identified as candidate targets of the Ap2d-Ash2l complex (Fig. 2B).
Functional annotation was performed based on gene ontology (GO) for the 42 genes that were downregulated in RNAi-treated cells (Table 1). A majority of the genes encodes for proteins that are involved in particular developmental functions, such as transcriptional regulation and signal transduction (Fig. 2C). Indeed, a significant enrichment of transcriptional regulators was identified in our analysis as only a small percentage (,4%) of genes encode for this class of genes in the mouse genome [14]. Additionally, these gene functions are consistent with the role of both Ap2d and Ash2l in development. We therefore concluded that these genes were probable targets of the Ap2d-Ash2l complex. Altogether, these candidate targets may shed some insight into the role of the Ap2d-Ash2l complex during development.

Prediction of Evolutionarily Conserved Ap2-Binding Sites in the Promoter Region of Putative Targets
To identify direct targets of the Ap2d-Ash2l complex, we searched the promoters of the 42 candidate genes for evolutionarily conserved Ap2-binding sites using rVista 2.0 [15]. We analyzed genomic sequences up to 5 kilobases (kb) upstream and 2 kb downstream of the transcriptional start site (TSS). Of the 42 candidate targets tested, we found highly conserved Ap2-binding sites in 21 of 42 genes we assessed ( Table 1). Given that a number of the candidate genes had Ap2-binding sites within 5 kb of the TSS, we concluded that these genes with Ap2-binding sites might be direct targets of the Ap2d-Ash2l complex.

Knockdown of Tcfap2d and Ash2l Leads to Decreased Expression of Candidate Ap2d and Ash2l Targets
To determine whether candidate targets with evolutionarily conserved Ap2-binding sites were indeed regulated by Ap2d and Ash2l, we investigated the expression level of these targets upon downregulation of Ap2d or Ash2l. We selected a number of genes that had roles in transcriptional regulation, development and signal transduction and tested the expression level of these genes using quantitative RT-PCR. Steady-state transcript levels for Plexin A3 (Plxna3), Fibroblast growth factor receptor 3 (Fgfr3) and Dickkopf homolog 3 (Dkk3) were significantly downregulated in Neuro2a cells after treatment with either Tcfap2dor Ash2l-specific RNAi (Fig. 3). Plxna3, Fgfr3 and Dkk3 encode proteins that play important roles in neuronal development [16,17,18]. Ap2d, in turn, has been implicated in neuronal development due to its highly restricted expression pattern in this tissue during embryogenesis [12]. Furthermore, MLL complexes have been implicated in neuronal differentiation, as MLL recruitment leads to increased H3K4me3 and activation of neuronal-specific genes [19]. Given that the candidate genes have overlapping roles in neuronal development with Ap2d and Ash2l-containing complexes, we predicted that these candidate genes were likely to be direct targets of Ap2d and Ash2l.

Ap2d Recruits Ash2l to the Fgfr3 Locus and Promotes H3K4me3
To identify direct targets of Ap2d and Ash2l, we determined whether these proteins were present on the Fgfr3, Plxna3 and Dkk3 promoters. We hypothesized that Ap2d and Ash2l would bind the promoters of these genes through highly conserved Ap2-binding sites that were previously identified in silico. To test this hypothesis, we performed chromatin immunoprecipitation (ChIP) using antibodies against V5/Ap2d and Ash2l and analyzed the bound DNA by quantitative PCR. We found that Ap2d and Ash2l were present only at the Fgfr3 promoter. Additionally, these proteins colocalized at various regions of the promoter that were highly enriched in evolutionarily conserved Ap2-binding sites (Fig. 4A). These regions include the sites ,1. To determine whether Ap2d recruits Ash2l-containing histone methyltransferases to the Fgfr3 promoter, we performed ChIP analysis with anti-Ash2l antibodies and Neuro2a cells treated with Tcfap2d-specific siRNA. Downregulation of Ap2d significantly decreased the association of Ash2l with the 21.2 kb, TSS and +1.6 kb sites (Fig. 4B). Having demonstrated an Ap2d-dependent recruitment of Ash2l, we next determined whether this recruitment altered H3K4 trimethylation at the Fgfr3 locus, as this epigenetic status marks transcriptional initiation [20,21]. We performed ChIP experiments using anti-trimethylated H3K4 antibodies with chromatin fragments obtained from Neuro2a cells  Fgfr3 Is a Target of Ap2d PLoS ONE | www.plosone.org treated with either Tcfap2d-specific siRNA or a scrambled control. We found significantly reduced levels of H3K4 trimethylation at the Fgfr3 locus when Ap2d was downregulated (Fig. 4C). These results indicate that Ap2d recruits Ash2l to the Fgfr3 promoter resulting in H3K4me3 and increased gene activation. It should be noted that downregulation of Ap2d did not alter Ash2l and Alr protein levels similar to previous results [11].

Discussion
In this study, we identified a number of candidate targets of Ap2d and Ash2l by comparing the gene expression profiles of Neuro2a cells treated with Tcfap2dor Ash2l-specific RNAi. Functional classification of probable Ap2d-and Ash2l-regulated targets revealed an overrepresentation of genes involved in transcriptional regulation, signal transduction and development. We also identified a number of genes that were involved in the Wnt/b-catenin pathway (Dixdc1, Dkk3, Lrp4, AA407331/Smad2) and the small GTPase-mediated signaling pathway (Arhgef1, Gapvd1) implying a probable role of Ap2d and Ash2l-containing complexes in these developmental pathways. Moreover, a significant portion of these candidate genes (21 out of 42) contained evolutionarily conserved Ap2-binding sites. We dem-onstrated that one of the three candidate genes we assessed, Fgfr3, was indeed a direct target of Ap2d and Ash2l given that both proteins co-localized at the promoter and that downregulation of Ap2d or Ash2l resulted in decreased Fgfr3 expression. Given that we had only assessed a limited window for Ap2-binding sites and only those sites for Ap2d-Ash2l binding, it should be noted that the genes for which we did not find any Ap2 binding sites might have binding sites further up-and downstream of the regions that we had assessed. As such, genes that were downregulated but for which we failed to find an Ap2-binding site are not necessarily false positives or indirectly regulated.
A growing body of evidence has shown that Fgfr1, -2 and -3 play important roles in the proliferation and differentiation of neural stem cells (NSCs). Immunocytochemical studies with NSCs derived from E15 rat striatum showed that expression of these receptors is developmentally regulated and cell lineage-specific. During the first day of culture, 50% and 70% of the NSCs or early precursors expressed Fgfr1 and -2, respectively, while a restricted population expressed Fgfr3 [22]. After 10 days in culture, the number of cells expressing Fgfr1 and -2 was significantly decreased to 15% of the total cell number whereas those expressing Fgfr3 comprised a significant portion of the population suggesting that Fgfr3 is increased during the process of cellular differentiation [22].  These results indicate that Fgfr3 may play a critical role in terminal differentiation while Fgfr1 and -2 may be important for earlier events, such as cell specification. Additionally, these Fgfrpositive cells were probed with various cell-lineage markers to determine whether there was an enrichment of a specific Fgfr in a particular cell lineage. Fgfr1 and -2 were detected in early oligodendroglial precursors whereas Fgfr3 was detected in early oligodendroglial precursors, oligodendrocytes and astrocytes [22]. These data suggest that Fgfr1-3 play specific roles in the differentiation of NSCs into neurons, oligodendrocytes and astroctyes. To demonstrate that Fgfr1, 2 and 3 have roles in neuronal differentiation, NSCs were treated with basic FGF (bFGF), which is a ligand for these receptors. As predicted, treatment with exogenous bFGF resulted in increased proliferation of NSCs and an increased number of oligodendrocytes after seven days in culture [22]. Recently, mice were generated carrying various combinations of Fgfr mutant alleles to establish the role of Fgfr in vivo [23]. As predicted, mice with mutations in two or three Fgfr genes demonstrated patterning defects and increased apoptosis in the CNS, supporting the notion that Fgfr's are important for cell survival and identity [23]. Altogether, these data imply that Fgfr1, -2 and -3 play important roles in the ability of NSCs to self-renew and differentiate into distinct neuronal cell types. Moreover, their expressions in the developing CNS overlap with that of Tcfap2d confirming the role of Ap2d in Ffgr3 regulation during development [24]. Indeed, these roles are consistent with those of trxG proteins whose functions have been linked to cell differentiation and memory. Although we identified several Ap2d-and Ash2l-regulated genes, a vast majority of differentially expressed genes did not overlap when the gene expression profiles were compared between cells treated with Tcfap2d-specific RNAi and those treated with Ash2l-specific RNAi. Given that Ash2l and its associated proteins are expressed ubiquitously, we hypothesized that Ash2l achieved its specificity through interactions with developmentally regulated transcription factors, such as Ap2d. This would imply that Ash2l would have functions independent of those attributed to Ap2d. Indeed, we found that only 76 out of 806 differentially expressed genes in Ash2l RNAi-treated cells overlapped with those in Ap2d RNAi-treated cells. Similarly, Ap2d may also interact with other co-activators, such as histone acetyltransferases and lysine demethylases, to activate its downstream targets. Previously, it had been reported that Ap2 proteins interacted with Cited2 and CBP to activate their targets indicating that Ap2d may also associate with these factors in activating its downstream targets. This hypothesis is further supported by the observation that interactions with Cited2 and CBP occur in regions of the Ap2 protein that are nearly identical among Ap2 family members including Ap2d. As such, Ap2d's interaction with Ash2l may occur independently or in addition to its interaction with Cited2 and CBP. These interactions would, in turn, result in a repertoire of genes that are regulated by Ap2d independently of Ash2l. Our results are consistent with this hypothesis, as only 76 out of 917 differentially expressed genes in Ap2d RNAi-treated cells overlapped with those in Ash2l RNAi-treated cells.
It had been suggested that Ash2l and its associated proteins, including the MLL/SET1 subunits, are global regulators of gene expression given their expression patterns and developmental functions. As such, deletion of Ash2l or any of its associated proteins may result in either embryonic lethality or a pleiotropic defect that could potentially mask a variety of distinct developmental phenotypes. To circumvent this issue, one could potentially analyze the role of Ash2l through analysis of its various regulators, such as Ap2d. Our studies are, therefore, an initial step in elucidating the function of Ash2l in vivo, providing a library of genes and pathways that are potentially regulated by Ash2l when it interacts with Ap2d. Additionally, our studies in Neuro2a cells using endogenous proteins may reflect to a limited extent conditions similar to that of neural progenitors in vivo. Given that Neuro2a cells have oncogenic properties, these targets will need to be validated in vivo.
In conclusion, we have identified a library of genes that are regulated by both Ap2d and Ash2l. A significant portion of these candidate target genes contains evolutionarily conserved Ap2binding sites implying that several of them are direct targets of the Ap2d and Ash2l-containing HMT complexes. Among the targets we screened, we identified Fgfr3 as a novel target of both Ap2d and Ash2l. Thus, we provide evidence that these candidate genes will be useful in elucidating the developmental roles of Ap2d and Ash2l.

RNA Analysis
For siRNA knockdown experiments, Neuro2a cells were transfected with Tcfap2dor Ash2l-specific siRNA or a scrambled control using Dharmafect 1 (Dharmacon, Lafayette, CO), and total RNA isolated 72 h post transfection. Total RNA was extracted using Trizol reagent according to the manufacturer's protocol (Invitrogen) and reverse transcribed using Superscript TM III reverse transcriptase and oligo-dT primers (Invitrogen). Transcript levels were determined by real-time PCR using Gapdh as an internal control.

Microarray Analysis
Total RNA was extracted from Neuro2a cells transfected with either Tcfap2dor Ash2l-specific siRNA using the RNeasy Kit (Qiagen, Valencia, CA). Total RNA was reverse transcribed using a T7-oligo d(T) primer (Affymetrix, Santa Clara, CA), and cDNA was used as template for in vitro transcription using biotin-modified ribonucleotides. Biotinylated cRNA targets were fragmented and hybridized to Affymetrix GeneChip Mouse Genome 430 2.0 Arrays. Arrays were subsequently washed, stained and scanned using an Affymetrix GeneChipH-related software. ArrayAssist (Stratagene) was used to determine statistical significance among probe sets that were differentially expressed between gene-specific siRNA-and NTCtreated samples. Probe sets that were changed at least 1.1-fold with a p-value#0.05 were used for further studies. Triplicate arrays were used for each sample to obtain statistical significance.