To characterize the molecular signature of ash1 and ash2 mutant wing discs, we compared whole genome expression profiles in wing imaginal discs from bromophenol blue-staged third instar larvae for the alleles ash2^I1, ash2¹¹²⁴¹¹ and ash1²² with those from an isogenic line used as a common reference. The ash2^I1 allele is lethal in late larval or early pupal stages whereas ash2¹¹²⁴¹¹ is a weaker allele that, although lethal at the pharate stage, presents a low percentage of adults 1112. The null ash1²² allele is a late larval/early pupal lethal 34. The wing disc phenotype associated with ash2^I1 discs involves abnormal shape and severe reduction in size, while ash2¹¹²⁴¹¹ and ash1²² discs are less affected (Figure 1a).

We performed a series of two-color microarray experiments with these alleles and obtained, for each of them, lists of genes showing greater than 1.5- or 2.0-fold changes in expression (Additional data files 2-4). The number of genes that were misregulated in each experimental condition is summarized in Additional data file 1 and all data are deposited in the Gene Expression Omnibus (GEO) database 3536 with accession number GSE4923. In accordance with the phenotypes of these alleles, there were many less misregulated genes in ash1 than in ash2 mutants.

To estimate the reproducibility of the results across platforms, two Affymetrix GeneChips were hybridized in a different laboratory with independently extracted ash2^I1 RNA and reference RNA, respectively. We calculated Pearson's correlation coefficients for all the microarray results, including Affymetrix (Figure 1b). The highest similarity observed was between the two-color and the Affymetrix microarrays hybridized with ash2^I1 and reference RNA, highlighting the quality of the data obtained.

Comparison of genes showing significant levels of misexpression between ash2 and ash1 alleles revealed an extremely high and significant overlap between genes regulated in the same direction (Figure 1c) and the correlation coefficients were consistent with the wing disc phenotypes and lethality phases observed for the different alleles. Since we also intended to assess the similarity between the three alleles in more general terms, automatic functional annotation of all results was performed based on GO (Additional data files 5-16 and 22 and 23). Although this type of annotation is very useful to obtain a description of the transcriptome, it can be misleading if the absolute number of genes that fall into each class is used as the only factor for selection of relevant classes. Consequently, we identified which GO terms were significantly over-represented in each mutant allele under study. If ash2 and ash1 did not have specific functions, we would expect misregulated genes to fall into various GO categories and that the proportion of the genes in each category would be the same as the proportion of all Drosophila genes found in that category. However, as shown in the examples of GO terms given to misregulated genes in Figure 2a, more genes than expected by chance were involved in certain processes or functions. The bars (grey for downregulated and black for upregulated genes) indicate whether there are more (positive values) or less (negative values) genes with a given GO term than expected with a random distribution. For instance, out of the 526 underexpressed genes identified in ash2^I1 mutants, 440 were annotated in the GO database. If these genes were randomly distributed, we would expect 6 of them to have the GO term 'wing disc development', since only 1.3% of all Drosophila genes with GO annotations have that term. However, 18 of the 440 genes held that descriptor (4.1%), representing a 3-fold enrichment in relation to the number expected by chance. This change corresponds to the size of the grey bar (downregulated genes) in the row 'wing disc development' for the ash2^I1 allele (see also Additional data file 5).

While the two ash2 alleles affected the same processes almost identically, the ash1 mutant displayed some slight differences (Figure 2a,b). Even though the significantly misexpressed genes were not identical between the three alleles, the expression of most genes tended to change in the same direction. The great similarity observed between the transcriptomes correlates with the similarities between the mutant phenotypes, indicating that the data obtained is reliable and that the functional relationship between ash2 and ash1 found by genetic interactions 13 is maintained at the transcriptional level.

We found more underexpressed and fewer overexpressed genes with the terms 'development' and/or 'organ development' than expected by chance, supporting the view that trxG proteins are usually involved in maintaining the activated state of genes involved in such processes. For some of those genes (ash2, rpr, Act57B, wbl and ImpE2) we confirmed their change of expression by performing semi-quantitative RT-PCR (Figure 2c). For others, like en or vg, we went one step further and analyzed their protein levels by immunohistochemistry and clonal analysis. In those genes, a correlation was observed between RNA expression and protein levels (Figure 3a,b). Some of the genes identified in this analysis may help to understand the mutant phenotypes observed, thereby establishing a link between the transcriptome and the phenotype.

To gain insight into the specific functions of ash2 and ash1, we compared their expression profiles with those of other proteins that also act at the chromatin level. In a recent study, Affymetrix GeneChips were used to establish a relationship between dmyc and Polycomb (Pc) in Drosophila embryos 47. Although PcG and trxG genes act upon some identical target genes in an antagonistic manner, we did not detect a significant overlap between the ash2 or ash1 and Pc sets of misregulated genes. We also compared our results with those obtained in larval null mutants of the Nurf301 subunit of the nucleosome remodeling factor complex NURF 48, and cells with reduced levels of Sin3A, a component of a multimeric histone deacetylase complex that includes Rpd3 49. While comparison of ash2 and ash1 transcriptomes with those from Nurf301 mutants did not reveal any significant overlap, intriguing coincidences were found when comparing ash2- and Sin3A-regulated genes.

The genes upregulated in S2 and Kc Drosophila cell lines with a Sin3A deficiency generated by RNA interference are enriched with GO annotations related to the mitochondrion 49 (Additional data file 19), a trend also observed in genes upregulated in ash2 mutants (Figure 2a; Additional data files 6 and 10). The upregulation of genes related to mitochondria and/or ribosomes is not a general effect, since this is not observed in either the Pc and Nurf301 sets of misregulated genes 4748 or in other mutants analyzed with our microarrays (data not shown). In addition, a gene to gene comparison between the misregulated genes for Sin3A and those for ash2 indicated that many of them tend to change in the same direction (Figure 4a; Additional data file 20). One possibility could be that Sin3A or another subunit of the deacetylase complex is a target of ASH2. However, from the two core members of the complex (Sin3A and Rpd3), only a slight reduction in the levels of Sin3A is observed in the ash2¹¹²⁴¹¹ mutant allele. Moreover, RT-PCR analysis in both ash2 mutants confirms that there are no major differences in Sin3A expression levels (data not shown), suggesting that Sin3A and Rpd3 are unlikely to be ASH2 target genes. Taking into account the important differences between these arrays (wing disc versus tissue culture cells and two-color microarray platform versus Affymetrix GeneChips), the large number of commonly regulated genes between the two analyses suggests a putative functional relationship between ASH2 and Sin3A.

In recent years, several studies in mammals have reported the presence of ASH2 in multimeric complexes homologous to the yeast Set1 complex that can methylate H3K4 21222324252650. Strikingly, one variant of this activation-related complex is tethered together by HCF-1 with the Sin3/histone deacetylase (HDAC) complex. The latter is associated with repression activities and includes, amongst others, Sin3, RbAp48 and the Rpd3 homologues HDAC-1 and HDAC-2 25.

To assess a putative physical interaction between ASH2 and Sin3A in Drosophila, we performed co-immunoprecipitation experiments in S2 cells and embryos and immunocytochemical detection of ASH2, Sin3A and HCF on salivary gland polytene chromosomes. As expected for proteins that regulate transcription, all three were located in the nucleus (Additional data file 24) and co-immunoprecipitation experiments demonstrated that HCF binds both Sin3A and ASH2 (Figure 4b-d). The interaction between HCF and ASH2 is easily detected but the interaction between HCF and Sin3A is only visible by a faint band in the embryo extract. Moreover, the HCF-ASH2 interaction seems to be more stable than that of HCF-Sin3A, since the former resists more stringent conditions in terms of lysis buffer composition (see Materials and methods). Although we were not able to co-immunoprecipitate Sin3A and ASH2, we cannot rule out the possibility that they are tethered together by HCF, as occurs in their human counterparts 25. Although this putative association of ASH2 and Sin3A through HCF might only happen in certain cell contexts, as previously suggested in humans 25, it is supported by the results of immunohistochemistry on polytene chromosomes showing that these proteins colocalize at several loci (Figure 5a,b). ASH2 binding patterns are highly coincident with those of Sin3A and HCF and all of them are non-overlapping with DAPI counterstaining, with the exception of HCF at the chromocenter. Comparison of bands bound by Sin3A, HCF and ASH2 revealed that some sites are shared between all three (Figure 5b), suggesting the existence of a complex containing these proteins. Although it might seem surprising that proteins with opposite transcriptional outcomes might be found in the same complex, this is not an exception. For example, the trxG human Brahma nucleosome remodelling complex (hBrm) contains Sin3, HDAC-1, HDAC-2 and RpAb48, all of them proteins from the human Sin3 histone deacetylase complex 51. Furthermore, these four proteins and hBrm are also found in the ALL-1 (acute lymphoblastic leukemia - 1) complex, which exhibits chromatin remodeling and histone acetylation, deacetylation and methylation activities 52.

Consistent with the H3K4 methyltransferase activity reported for the Set1/ASH2 complex in humans 25, we found that ASH2 colocalizes on polytene chromosomes with trimethylated H3K4 (Figure 6a). To gain insight into the molecular function of ASH2, we compared the levels of H3K4 trimethylation in histones purified from wt and ash2^I1 mutant flies, respectively, by western blot (Figure 6b). The lack of ASH2 results in a dramatic reduction of H3K4 trimethylation, indicating that ASH2 is involved in this modification. Analysis of H3K4 trimethylation in vivo confirmed the results, since this epigenetic mark was below detectable levels in ash2^I1 mutant clones (Figure 6c). The results presented here are the first obtained in Drosophila that show that ASH2 plays a role in the chromatin modification machinery. Although the nature of this regulation remains unclear, the fact that ASH2 does not have any known domain with histone methyltransferase activity suggests that it could be acting in a complex as a structural platform to facilitate the interaction between nucleosomes and proteins with enzymatic activity. This hypothesis is supported by the presence in ASH2 of a PHD domain, which has the ability to bind methylated lysines 53. However, other proteins might act to recognize and present H3 in a similar way to human WDR5 5455.

All Drosophila strains and crosses were kept on standard media with 0.025% bromophenol blue at 25°C. The reference line used in all cases was the w¹¹¹⁸;+;+ isogenic line from the DrosDel Collection 68. To reduce the differences in the biological background between the alleles under study and the reference strain, we transferred chromosomes X, Y and 2 from the isogenic line to the TM6c-balanced ash2^I1, ash2¹¹²⁴¹¹ and ash1²² alleles. The latter was provided by A Shearn and was studied in homozygosis by combining it with ash1²²F2A, a kind gift from J Muller 29.

To generate UAS-ash2 transgenic flies, full-length ash2-RC and three copies of the HA (hemagglutinin) epitope were cloned between the EcoRI and NotI sites of the pUAST plasmid 69. To generate UAS-HCF transgenic flies, full-length HCF-RC amplified from the pACXT-T7-HCF-Flag (kindly provided by A Wilson 70) was introduced between the NotI and KpnI sites of the pUAST plasmid 69. MS1096-Gal4 and nub-Gal4 drivers were used for rescue experiments, sgs3-Gal4 and Act5C-Gal4 for polytene chromosome immunolocalization and da-Gal4 for co-immunoprecipitation experiments in embryos.

Microarrays were printed on Corning UltraGAPS slides (Corning, NY, USA) at the Plataforma de Transcriptòmica (SCT-PCB, Universitat de Barcelona, Spain) using the Drosophila Genome Oligo Set version 1.1 (Operon Biotechnologies Inc., Huntsville, AL, USA), a collection of 14,593 probes representing 13,577 Drosophila genes with Flybase ID. The 70mer Arabidopsis sequences from TIGR 71 and spots with no material or with buffer were also printed to be used as spike-in and negative controls, respectively. The microarray annotation is deposited in the Gene Expression Omnibus (GEO) database 3536 with number GPL3797.

Wandering blue-gut staged Tb+ early third instar larvae were selected in all cases to extract total RNA from wing discs using the RNeasy Protect Mini Kit (Qiagen Inc., Valencia, CA, USA). At least two independent total RNA extractions were carried out from wing imaginal discs for w¹¹¹⁸;+;ash2^I1, w¹¹¹⁸;+;ash2¹¹²⁴¹¹, w¹¹¹⁸;+;ash1²²/ash1²²F2A and w¹¹¹⁸;+;+ strains. Quality was assessed in all samples using the Eukaryote Total RNA Nano Assay on a 2100 Bioanalyzer (Agilent Technologies Inc., Santa Clara, CA, USA). Total RNA from w¹¹¹⁸;+;+ wing imaginal discs was pooled and used as a common reference against ash2^I1, ash2¹¹²⁴¹¹ and ash1²²/ash1²²F2A. Four microarrays were hybridized for each experiment in biological replicate pairs, such that the total RNA from the sample used as a starting material came from different extractions. Both arrays from each pair were hybridized with the same amplified RNA from sample and common reference (obtained using the Amino-Allyl Messageamp II aRNA Amplification Kit from Ambion Inc., Austin, TX, USA) but with dyes Cy3 and Cy5 (GE Healthcare UK Ltd, Buckinghamshire, UK) swapped to take dye-bias into account.

GenePix Results (GPR) data files were obtained for each microarray with an Axon 4000B scanner and GenePix Pro 6 (Molecular Devices Corp., Sunnyvale, CA, USA). All GPR files were analyzed with the Limma package from BioConductor 7273 using the same criteria. The spots not fulfilling the quality thresholds (based on spot size, foreground versus background signals, saturation, coincidence between differently calculated ratio measures and R² of regression ratio) were eliminated from the analysis, and the data were background corrected with the normexp method and normalized using OLIN 74. Between-array normalization was carried out independently for each set of four arrays using the mad method from OLIN, and a linear model was fitted and corrected with False Discovery Rate (FDR) 72. We obtained lists of genes that were differentially expressed (had an FDR-corrected p value of less than 0.05 and at least two spots from the four replicate arrays that passed the quality filters) over 1.5- or 2.0-fold in the mutants compared to the reference strain.

A script was written to collect all GO terms for each gene on the array and to assess whether they are enriched with regulated genes by using a hypergeometric distribution. P values of each term are adjusted with the FDR procedure 75 to correct for multiple testing. The output from that script can be used as input for another one that draws pie charts (available at 76). Calculations of the statistical significance of the number of overlapping genes between different sets were also corrected for multiple testing in a similar way.

Two Affymetrix GeneChip Drosophila Genome 2.0 arrays (Affymetrix Inc.) were hybridized at the Unitat de Genòmica (IDIBAPS, Barcelona, Spain) with amplified RNA from wing imaginal discs of either ash2^I1 or the common reference. Analysis was performed with the Affymetrix GeneChip Operating Software (GCOS) and only spots called 'Present' (detectable expression) were taken into account for calculation of Pearson's correlation coefficients. The number of spots involved in each pairwise comparison in Figure 1b (from left to right and top to bottom) were 7,817, 7,170, 6,147, 9,051 and 6,436.

Reverse transcription reactions with independently extracted RNA from all mutant alleles and the reference were used to synthesize cDNA with Superscript III (Invitrogen Corp., Carlsbad, CA, USA) according to the manufacturer's instructions. One microliter of each RT reaction was used in each PCR reaction, in which the test and the reference gene (mRpL9) were amplified together. Samples from 22, 26, 30 and 34 cycles were run on 2% agarose gels to estimate saturating conditions. Samples with two discrete non-saturated bands were quantified using a DNA 1000 Assay in a 2100 Bioanalyzer (Agilent Technologies Inc., Santa Clara, CA, USA). Expression values of test genes were normalized using the reference gene. Four measures were taken for each gene to evaluate standard error.

Clones were generated at 52 ± 6 hours after egg lay by FLP (Flipase)-mediated mitotic recombination 77 with a 30 minute heat shock at 37°C. Immunohistochemistry was performed according to standard protocols with the primary antibodies mouse 4D9 anti-En/Inv (Developmental Studies Hybridoma Bank, University of Iowa, IA, USA), rabbit anti-Vg (kindly provided by S Carroll), rabbit anti-cleaved Caspase 3 (Cell Signalling Technology Inc., Danvers, MA, USA) and rabbit anti-trimethyl H3K4 (Abcam Inc., Cambridge, MA, USA) used at concentrations of 1:10, 1:20,1:100 and 1:200, respectively. Rhodamine Red-conjugated secondary antibodies were from Jackson ImmunoResearch Laboratories (West Grove, PA, USA). Dying cells were detected by TUNEL assay, labeling DNA 3'-OH ends with Chromatide BODIPY Texas Red-14-dUTP (Invitrogen Corp.) for 1.5 hours at 37°C with terminal deoxynucleotidyl transferase (Roche Diagnostics, Indianapolis, IN, USA). In all cases, the discs were mounted in Slowfade Light Antifade (Invitrogen Corp.) and images collected using a Leica TCS 4D confocal laser scanning microscope.

S2 cells were maintained at 25°C in Drosophila Schneider's medium containing 10% fetal bovine serum and 1% penicillin/streptomycin (Invitrogen Corp.). Cells were transfected with Effectene (Qiagen Inc.), according to the manufacturer's protocol and collected 48 hours later. Three different constructs were used containing full-length coding regions of either ash2, Sin3A or HCF. pAc5.1-ash2-V5-His A was produced by inserting ash2, obtained from the Berkeley Drosophila Genome Project clone LD31680, into pAc5.1-V5-His A (Invitrogen Corp.). pAc5.1-Sin3A-HA-His B was produced by inserting Sin3A, obtained from the p3NB-Sin3A plasmid provided by D Pauli 78, into a pAc5.1-V5-His B in which V5 was substituted for HA.

For immunofluorescence analysis, transfected S2 cells were plated onto coverslips previously coated with 0.5 mg/ml concanavalin A for 1 hour and fixed with 4% paraformaldehyde. The antibodies used were as follows: rabbit anti-V5 (1:1,000; Sigma-Aldrich Corp., St Louis, MO, USA), mouse anti-HA (1:200; Roche Diagnostics), rabbit anti-Flag (1:100; Sigma-Aldrich Corp.), donkey anti-rabbit FITC (1:400; Jackson ImmunoResearch Laboratories), and goat anti-mouse FITC (1:400; Jackson ImmunoResearch Laboratories). Rhodamine phalloidin (Invitrogen Corp.) was used at a dilution of 1:40. Cell nuclei were stained with DAPI. Cells were mounted in Mowiol-Dabco and visualized under a Leica spectral microscope.

For co-immunoprecipitation assays, cells were washed twice with cold phosphate-buffered saline and lysed and disrupted by freezing and thawing cycles. For anti-Flag immunoprecipitation, the lysis buffer was 50 mM Tris HCl, pH 7.4, 150 mM NaCl, 1 mM EDTA, 1% Triton X-100 + complete protease inhibitors (Roche Diagnostics). The whole-cell extract was pre-cleared for 1 hour at 4°C with protein A-Sepharose and incubated with 40 μl of equilibrated EZview Red anti-Flag M2 AffinityGel beads (Sigma-Aldrich Corp.) for at least 3 hours at 4°C. The immunoprecipitates were washed three times for 15 minutes in lysis buffer (in the second wash the concentration of NaCl was increased to 300 mM) and eluted by boiling in loading buffer containing SDS. For anti-HA immunoprecipitations, we used the protocol described above with the following modifications: cells were lysed and washed in 20 mM HEPES, pH 7.9, 20% glycerol, 100 mM NaCl, 0.2 mM EDTA, 0.1% NP-40 + complete protease inhibitors and cell extracts were incubated without pre-clearing with 75 μl of anti-HA affinity matrix (Roche Diagnostics). For co-immunoprecipitation experiments from embryo extracts, 0-16 h embryos were dechorionated and rinsed extensively, incubated with 50 mM Tris pH 8, 200 mM NaCl, 5 mM EDTA, 0.5% NP40 + complete protease inhibitors, homogenized and centrifuged. The supernatant was collected, sonicated and mixed with 40 μl of EZview Red anti-Flag M2 AffinityGel beads or anti-HA affinity matrix to perform the co-immunoprecipitation experiments as described above. In all cases, samples were run on 7.5-10% SDS-PAGE gels and transferred to PVDF membranes. The following antibodies from Sigma-Aldrich Corp. were used for western blots: rabbit anti-V5 (1:2,000), rabbit anti-Flag (1:400), sheep anti-mouse peroxidase (1:3,000) and goat anti-rabbit peroxidase (1:3,000). The mouse anti-HA (1:1,000) was from Roche Diagnostics and rabbit anti-Sin3A (1:500) and rabbit anti-HCF (1:5,000) were kindly provided by D Wassarman and J Workman, respectively. Proteins were detected with the EZL system (Biological Industries Ltd., Kibbutz Beit Haemek, Israel).

To compare the levels of trimethylated H3K4 between wt and ash2^I1 larvae, histones were obtained by acidic extraction, dialyzed and separated on 15% SDS-PAGE gels. Immunoblots were performed with anti-H3 (1:2,000) and anti-trimethyl H3K4 (1:5,000) antibodies (Abcam Inc.).

Salivary glands of wt and UAS-ash2 transgenic flies were dissected in Gohen buffer, fixed for 2 minutes and transferred to a solution with acetic acid and formaldehyde for 3 minutes before squashing to spread the polytene chromosomes. Staining was performed by O/N incubation at 4°C with the following antibodies: mouse anti-HA (1:200), rabbit anti-Sin3A (1:100), rat anti-HCF (1:50) and rabbit anti-trimethyl H3K4 (1:200). FITC- or rhodamine red-conjugated secondary antibodies (1:200) were obtained from Jackson ImmunoResearch Laboratories and preparations were mounted with Mowiol-DAPI to stain the DNA. Chromosomes were visualized under a Leica microscope.