Chromatin immunoprecipitation and sequencing

ChIP experiments were performed according to [35], using stage 13–14 col 2.6_0.9^ColCONS-lacZ embryos, with a mix of three different monoclonal antibodies recognizing separate epitopes of the Col protein. The mock was monoclonal HA antibody (HA.11 Clone 16B13-Covance, Dedham, Massachusetts, USA). To improve the purification yield, we used a competitive elution with purified recombinant Col protein (Kn-RB isoform; http://flybase.org); this step contaminated the ChIP-DNA with plasmid DNA, requiring bio-informatics elimination of the contaminating sequences. The precipitated DNA was quantified using Qubit dsDNA H Assay Kits (invitrogen#Q32851). Real time quantitative PCR was performed on a MyiQ single color real time PCR detection system (Biorad). CT values were collected and analysis was performed using the 2(-Delta Delta C(T)) method [36], using cg11964 to normalize calculations of relative expression. For comparison, we substituted the consensus EBF/Col binding motif established in vitro (TCCCNNGGGA; [11] for the endogenous TGTCNNGGGA site in the reporter construct col 2.6_0.9^ColCONS-lacZ. Primers sequences for col and col 2.6_0.9^ColCONS are available on request. All qRT-PCR data are representative of three independent experiments. DNA from two independent Col and Mock iPs was pooled before ChiP fragment amplification and High-throughput sequencing (Genome analyzer IIx, Illumina; Microarray and sequencing platform, IGBMC, Illkirch). Reads were aligned to the D. melanogaster genome (BDGP5) using Bowtie v0.12.7 [37]. Data are available at Gene expression Omnibus, Accession number GSE67805.

Identification of Col binding peaks

Col peaks were searched using SISSRs v1.4 [38]; [39], run with default parameters except for the following options: pValue threshold = 0.1, eValue threshold = 1500, one read per genomic coordinate, average fragment length = 191. 559 Col binding peaks retrieved by SISSRs were further analysed.

de novo motif discovery

MEME-ChIP v4.10.0 [40] was used to search 200 bp of DNA, centred on the summit of each of the 559 Col peaks. Motif discovery was performed by scanning both DNA strands for 4 to 25 nucleotides long motifs, with a distribution probability of zero or one occurrence per sequence and a 4-order Markov model as background reference. Results from SISSRS and MEME analyses are provided as S1 File and dataset GSE 67805.

Genome annotation and GO enrichment of Col binding peaks

Peak Analyzer [41] was used to associate each Col ChiP-Seq peaks with one gene in the Drosophila genome (BDGP 5.74). Peaks were associated with the overlapping gene when in introns or the nearest gene transcription start. Enrichment of Col target genes for GO biological processes was using GeneCodis (http://genecodis.cnb.csic.es/)

Reporter constructs and transgenic lines

pDEST-moeGFP and pDEST-moeRFP were made by replacing Gal4 sequences from pbGUw (addgene plasmid #17575) by moeGFP and moeRFP sequences, respectively. For candidate Col targets (Table 1), 1kb long DNA fragments centered on the Col binding site were cloned upstream of moeGFP by Gateway recombination (invitrogen–life technologies) into pDEST-moeGFP. Site-directed mutagenesis of the Col binding site(s) (S2 Table) was done by PCR. The ap_Col^mut and eya_Col^mut.1 et eya_Col^mut.2 were made in pDEST-moeRFP. All moe-GFP/RFP reporters were inserted at position 68A4 on the third chromosome by injection into nos-phiC31-NLS and made homozygous. Other fly stocks are available at the Bloomington Stock Center.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 1. Candidate Col direct targets and linked CRMs.

https://doi.org/10.1371/journal.pone.0133387.t001

Immunohistochemistry and in situ hybridization

Immunostaining and in situ hybridization procedures were as in [33]. The following primary antibodies were used: mouse anti-β-galactosidase (Promega) 1/800; anti-Col 1/50; rabbit anti-GFP (Torrey) 1/500; anti-RFP (Rockland 1/500); anti-Nau 1/100 (B. Paterson, Bethesda, USA), anti-β3-Tubulin 1/5000 (R. Renkawitz-Pohl, Marburg, Germany). Secondary antibodies were Alexa Fluor -488, -647, -555 conjugated antibodies (1:300; Molecular Probes). Mounting samples for confocal microscopy (Leica SP2, SP5 and SPE microscopes, Wetzlar, Germany; 20x and 63x objectives) was in Vectashield medium (Vector Laboratories). The Vectastain ABC Kit PK-401, from Vector, was used for DAB (3, 3’-diaminobenzidine) immunostaining; phosphatase-conjugated antibodies for BCIP/NBT detection of ISH transcripts were from Roche. DAB immunostaining experiments were repeated at least 3 times and NBT/BCIP ISH at least twice with large collections of embryos. At least 10 randomly selected embryos at one given developmental stage were recorded for each experiment, using a 20x objective. Effect of mutating the Col binding site(s) was considered as significant when >80% of embryos showed a change in expression pattern between the intact and mutated reporter constructs, except when otherwise indicated in the text. Images shown in figures and supporting figures are representative examples. Colocalisation of signals from 2 different fluorochromes used Image J colocalization highlighter plugin (Pierre Bourdoncle, Institut Cochin, Paris, France).

Genome-wide mapping of Col binding sites to chromatin in stage 13–14 embryos

In order to identify Col direct targets, we used chromatin from 10 to 12h old Drosophila embryos (stages 13–14). During stages 11–13, Col is expressed the muscle PCs and FCs at the origin of the dorso-lateral (DL)—DA3, DO3, DO4, DO5, LL1 and DT1- somatic muscles, and starts to be expressed in the ventral nerve cord (VNC); it is also expressed in the hypopharyngeal lobe (HL) (Fig 1A). At stage 14 and later, it is expressed in the differentiating DA3 muscle, about 50 VNC neurons and 2 or 3 multidendritic (md) neurons per hemisegment, and the developing LG [5]; [24]; [25]; [31] (Fig 1A). To immuno-precipitate Col-bound chromatin fragments, we opted for a mix of monoclonal anti-Col antibodies. As an internal control for IP specificity, we used a reporter transgene carrying a modified col CRM, col 2.6_0.9^ColCONS, whose activity in the DA3 muscle depends upon direct Col binding [33]. Quantitative real time PCR [36] of DNA fragments covering the endogenous and transgenic col CRMs was performed on DNA samples from two independent IPs. It showed a significant enrichment of both fragments, of around 2 (2.08 and 2.19) and 4 (3.90 and 4.26) folds, respectively, compared to the intergenic region of cg11964, a control housekeeping gene [35]. This differential enrichment both confirmed the efficiency of Col antibodies and indicated that the nucleotide sequence of the Col binding motif could influence the occurrence and/or stability of contextual in vivo Col binding.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 1. Genome-wide mapping of Col binding sites.

(A) Col expression in stage 12 and stage 14 embryos, lateral view. In this and all subsequent figures, embryos are oriented anterior to the left. HL: hypopharyngeal lobe; PCs: dorso-lateral muscle progenitors; vnc: ventral nerve chord; lg: lymph gland; md: class IV multidendritic neuron. (B) Fold enrichment distribution of the 559 Col ChIP peaks selected using SISSRs. (C) Single most enriched sequence motif identified by MEME analysis in the 559 Col binding peaks. (D) Graphical representation of the position of Col-binding motifs relative to the center of Col binding peaks. The axis gives the number of motifs in each cluster. (E) GO clusters enriched in putative Col direct target genes. The p-Value of the top two clusters is given in brackets.

https://doi.org/10.1371/journal.pone.0133387.g001

Low amounts of chromatin were obtained for each IP, correlating well with the small number of Col-expressing cells. We therefore pooled IP samples before sequencing. 18.3x10⁶ and 14.6x10⁶ immuno-precipitated fragments of 190 bp average size were sequenced for the Col-IP and mock (HA-IP) samples, respectively (dataset GSE 67805). The sequences were aligned to the D. melanogaster genome (BDGP release 5) using Bowtie v0.12.7 [37]. 16.1x10⁶ and 13.1x10⁶ unique reads for the Col and mock-IP, corresponding to 25 and 22 times the Drosophila genome size, respectively, were kept for analysis. Peak calling using the SISSRs software [42] detected 559 Col binding peaks (pvalue threshold = 0.1, eValue threshold = 1500), with a fold enrichment ranging from 1.94 to 16.1 (Fig 1B and S1 Table). The peaks were located either in introns or intergenic regions, consistent with Col binding in vivo to cis-regulatory regions.

de novo motif discovery was then performed on the entire set of 559 Col peaks, using the MEME suite software [43]. 200 nucleotides long windows centered on each peak’s summit were considered for this analysis (S1 File). It revealed that 97% of the peaks (542/559) contain one motif of consensus sequence CCCnnGGGA (Fig 1C). This consensus site is similar to the consensus in vivo binding site determined for mouse EBF in cultured lymphomas [44]. Significant enrichment of positions 13 and 14 for A and T nucleotides, respectively, was also consistent with the ATTCCCNNGGGAAT sequence of the in vitro EBF binding site defined by selex [11]. The calculated E-value: 2,1e-381 and predominant position of the CCCnnGGGA motif close to the center of the ChIP peak (Fig 1D) supported the conclusion that this motif is bound by Col in vivo. MEME analysis failed, however, to reveal other significantly enriched motifs, which could have represented binding sites for other sequence-specific TFs acting synergistically with Col in the different cell types where Col is expressed.

Direct Col-target genes are enriched in transcription factors

PeakAnalyzer [41] associated the 559 Col ChIP-seq peaks to 415 genes (dm3/FlyBase R5.74). Several peaks (between 2 and 9) were associated with the same gene in 95 cases (S1 Table). 150 peaks (27%) were located 5’ and 100 (18%), 3’ of the nearest transcription start site (TSS), while most other peaks (309; 55%) were located in introns. Of course, this association did not exclude that, in some cases, the Col-bound region could act as remote enhancer for other/and additional genes [45], especially for those peaks found further than 10 kb from the nearest TSS (72 peaks; 13%). In 43 cases, the automatic peak association could be manually curated, to account for gene association issued from compared expression of Vienna Tiles (VT) reporter constructs [45] or other CRMs described in REDFly3.3., with that of nearby genes. For 7 peaks, of which 4 were located far away from the nearest transcription start, this analysis modified the associated gene (S1 Table). The described embryonic expression patterns of 176 of the 415 Col targets (http://insitu.fruitfly.org/) indicated that 100 (57%) are expressed in the CNS and 45 (25%) in the mesoderm, consistent with the Col expression pattern in stage 12–14 embryos. In order to further associate Col target genes to biological processes we used GeneCodis [46]; [47]; [48] to perform a Gene Ontology (GO) analysis (Fig 1E). 397 out of 415 Col-bound genes had GO annotations for Biological Processes and were considered by GeneCodis. 156 were represented in at least one cluster (>5 genes) of p-value<10⁻⁵. Highly enriched GO terms identified several developmental processes, correlating with the diverse functions played by Col during mid-embryogenesis (Fig 1E). GO terms further revealed a statistically significant enrichment in the category “regulation of transcription”, with 64 of 397 (16%, P-value 9x10^-30) annotated Col targets present in this category. This could, in part, be due to larger than average regulatory regions of developmental control genes [49]. Yet, several of these TF genes were already shown to act genetically downstream of Col in different cell types, thus validating our dataset. One example was cap n’ collar (cnc), which acts downstream of Col in head patterning [5]; [26]. In addition to cnc, the dataset included pox-meso (poxm), slouch/S59 and col itself which are regulated by Col in specific muscle lineages [33]; [32], and apterous (ap), even-skipped (eve), and eyes-absent (eya) which are regulated by Col in specific subsets of VNC neurons [28]; [31]. eve expression depends upon Col in a specific subset of neurons, the Eve-lateral (EL) neurons [31]. Previous dissection of eve cis-regulatory elements identified a 0.7kb genomic region, EveEL, specifically required for eve expression in EL neurons [50]; [51]. We found that Col binding to eve precisely mapped within EveEL (S1 Fig). The congruence between col/eve epistatic interaction, and matching genomic positions of EveEL and in vivo Col binding, led us to propose that Col binding reflected direct eve regulation in EL neurons. By extension, GO analysis suggested that Col directly regulates the transcription of various other TFs.

In summary, our ChIP-seq analysis identified 415 potential direct Col targets, among which 64 transcription regulators, suggesting that Col occupies a hierarchical position in a diversity of transcription regulatory networks. More than half of putative Col targets are “unknown” genes for which Col binding provides an entry site for studying their developmental expression and biological role.

Selection of candidate Col direct targets

Identification of Col chromatin binding sites provided an opportunity to identify genes regulated in different Col-expressing cells and the corresponding CRMs. Here, we focused on a small set of 10 candidate Col targets. Selection was based on a number of criteria, including a range of peak heights (from 2.39 to 8.03); genes displaying from 1 to 6 Col peaks; documented expression, especially in the nervous system and/or mesoderm (http://insitu.fruitfly.org/); GO analysis (Fig 1E) and previous literature on Col embryonic functions. It included 8 TFs or co-factors—ap, eya, jing, Myocardin-related transcription factor (Mrtf), nervous fingers-1 (nerfin-1), O/E-associated zinc finger protein (Oaz), and sine oculis (so)—and two transmembrane proteins, Amalgam (Ama), and thickveins (tkv) (Table 1). Among TFs, ap and eya were chosen because of prior evidence for their regulation by Col in specific neurons, with no evidence that this control was direct [28]; [52]. eya encodes a protein with tyrosine phosphatase activity which is a partner of SIX homeodomain TFs (D-Six4, Optix and So in Drosophila) [53,54]. so is the only SIX gene bound by Col in vivo. Oaz is the Drosophila ortholog of vertebrate Oaz/Zfp243, the only TF reported to physically interact with EBF [55]; [56].

For each gene in our selection, we cloned 1 kb fragments centered on the Col peak summit, upstream of a moe-GFP reporter construct and followed GFP expression in transgenic Drosophila lines. Each CRM construct was given the name of the gene, followed by _Col (Table 1 and S2 Table). For jing and Oaz which display 5 and 4 Col peaks, respectively, we analyzed one peak per gene, where the nucleotide sequence of the predicted Col recognition site was well conserved between 12 different Drosophila species ([57]; [58]; Dataset GSE 67805). For eya, which displays 2 nearby peaks, they were analysed together as well as separately. To precisely determine the role of Col binding for CRM activity, we made a parallel series of constructs in which the predicted Col binding site(s) was mutated at 3 nucleotide positions by substituting CCCNNCCC for CCCNNGGG (S2 Table). This substitution was shown to abolish EBF binding in vitro [11]. As an in vivo control, we mutated the Col binding site in the col2.3–0.9 CRM (col2.3–0.9^mut). col-2.3–0.9^mut activation in the head was not modified, while expression was severely reduced in the DA3 muscle where col2.3–0.9 is subject to autoregulation [33], showing that our reporter strategy could identify context-dependent direct regulation by Col (S2 Fig).

Analysis of the 10 candidate CRMs revealed overlap with Col expression in various tissues (Table 1), including the head hypopharyngeal lobe (HL) (Ama, cnc; Fig 2), the LG (jing and tkv, S3 Fig), specific VNC neurons (ap, eya, nerfin-1; Figs 3 and 4 and S4 Fig), body wall muscles (eya, Mrtf, oaz, so, Fig 5 and S5–S7 Figs). tkv_Col and jing_Col were expressed in numerous cell types, including the LG. Since LG development is a dynamic process [27], we compared tkv_Col with tkv_Col^mut, and jing_Col with jing_Col^mut expression by ISH at two developmental stages, in order to detect subtle transcriptional regulation potentially hindered by the time lag between moeGFP transcription and protein accumulation (S3 Fig). Upon mutation of the Col binding site, expression of both reporters was specifically decreased in the LG, indicating that Col positively regulates tkv_Col and jing_Col activity in this tissue. However, since the decreased expression was only indubitable in about half of the embryos, these targets were not investigated in more detail.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 2. Col control of cnc and Ama expression in the head.

(A) Annotation of the Col peak in cnc, adapted from Gene Browser (GEO submission GSE67805). 39,5 kb of cnc genomic region are shown (Chr3R: 19.009.000–19.048.500) with the Flybase gene annotation indicated by bars (transcribed regions) and intervening blue lines (introns). Black arrows indicate the direction of transcription of cnc and fuzzy onions (fzo), inwardly rectifying potassium channel 1 (Irk1). The cnc transcripts coding for the protein isoforms CncA and CncB are indicated. ChIP-seq data for Col (green) substracted from HA (mock) data (red) are shown on the bottom. The Col Dam-ID binding regions [59] are indicated by yellow bars, top line. The summit of the ChIP-Col peak identified by SISSRs and position of the Col binding site(s) identified by MEME are indicated by blue and violet lines, respectively; the position of cnc_Col is represented by a black box; scale is indicated. (B-D) Ventral anterior views of stage 11 embryos. (B) Overlap between cnc_Col (GFP, green) and Col (red) expression in the HL (white arrow). (C) cnc_Col and (D) cnc_Col^mut mRNA expression, showing down-regulation of cnc_Col^mut in the mandibular segment (open arrow). (E,F) Ama_Col, (G,H) Ama_Col^mut expression in stage 10 (E,G) and 11 (F,H) embryos. HL Ama_Col expression (arrow) is lost in Ama_Col^mut (open arrow). (I, J) Overlap between Ama (red), Col (blue), and col2.6–0.9moeGFP (green) expression in the HL (white arrow) in stage 11 wt (I), and col¹ mutant embryos (J). Separate signals for Ama, GFP (I, J) and Col (inset in I, left panel) are shown in black and white. Ama expression is specifically lost in the HL in col¹ mutants. The asterisk in C, D, F, H, J indicates Ama and cnc expression independent on Col.

https://doi.org/10.1371/journal.pone.0133387.g002

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 3. Col direct control of ap expression in Ap neurons.

(A) Annotation of the Col peak in ap, same representation as in Fig 2A; 35.8 kb of the ap genomic region are shown (Chr2R: 1.593.000–1.628.800); the previously described apC enhancer is represented by a blue box. (B) ap_Col (GFP) expression in the dAp (yellow arrow) and Tv1-Tv4 neurons (white arrow) in stage 15 embryos, ventral view. (C) ap_Col^mut expression is severely reduced in dAP neurons and Tv neurons. (D,D’) Close up view of 4 segments of stage 16 embryos, showing the specific overlap between Col (red) and ap_Col (green) in the Tv1 and dAp neurons. (E) all Tv neurons express ap_Col and Eya. (F,G) ap_Col^mut expression is lost in dAp and strongly reduced in Tv neurons.

https://doi.org/10.1371/journal.pone.0133387.g003

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 4. Col direct control of eya expression in Ap neurons.

(A) Annotation of the Col peaks in eya, same representation as in Fig 2A; 24 kb of the eya genomic region are shown (Chr2L: 6.524.500–6.548.500); the summits of the two ChIP-Col peaks are numbered 1 and 2. (B, C) eya_Col (GFP) expression in the dAp (yellow arrow) and Tv1-Tv4 neurons (white arrow) in stage 16 embryos, ventral view. (C) Close up view of abdominal segments, showing the specific overlap between Col (red) and eya_Col (green) in the dAp (yellow arrow) and Tv1 neurons (white arrow). (D, E) eya_Col^mut expression is lost in dAP neurons and reduced in TV neurons. (F, G) Mutation of the Col binding site 2 (G), but not site 1 (F) eliminates eya_Col (RFP) expression in dAp neurons (yellow arrow).

https://doi.org/10.1371/journal.pone.0133387.g004

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 5. Cross-regulation between eya and col is required to specify dorso-lateral muscles.

(A) eya_Col (GFP) expression in the DA3 muscle (arrows) in stage 15 embryos, is not detected for eya_Col^mut (B). (C-E) Triple staining of st11 wt (C), +;col^eCRM-lacZ (D) and col¹;col^eCRM-lacZ (E) embryos for Nau (blue), eya transcripts (red), and either Col (C) or β-galactosidase (D,E) (green), shows co-expression of Col, col^e-CRM-LacZ, Nau and eya in DL muscle PCs (white arrow). (C’-E’) only Nau and eya stainings are shown. eya transcription is specifically lost in DL PCs in col¹ mutant embryos (E,E’); dorso-lateral view of the T2 and T3 segments is shown. (F) Schematic drawing of the dorsal, dorso-lateral and lateral transverse muscles in a stage 16 wt embryo, with the DA3 muscle in red and the LL1 muscle indicated by an arrow. (G, H) Staining of stage 16 wt (G) and (H), eya^cIi-IID/Df(2L)BSC354 (null) mutant embryos for Col (red) and β3-tubuliin (green). Lateral view of 3 segments. In absence of eya, Col expression is lost in most segments, the LL1 muscle is missing (white arrow) and the DA3 muscle (asterisk) malformed. White brackets indicate dorsal, unaffected muscles while the lateral transverse and ventral muscles (yellow brackets) are moderately affected. (I-J) Col immunostaining of st.16 wt (I), and (J) eya mutant embryos, showing the loss of Col muscle expression in most segments (see also G, H), in eya mutants. (K, L) promuscular Col expression, early stage 11, is reduced in eya mutant embryos (L), compared to wt (K).

https://doi.org/10.1371/journal.pone.0133387.g005

Col targets in the head

Col is required for head patterning in the embryo, downstream of gap genes [23], more specifically in parasegment 0 (posterior intercalary and anterior mandibular segment; also described as HL from stage 11), where it regulates expression of the segment polarity gene hh and the segment identity gene cnc. Other than col itself [33], hh was the only direct Col target previously characterized [34]. Col ChIP-seq failed to detect binding to hh, perhaps because it is regulated at earlier developmental stages than used for our ChIP-seq [23]; [34]. Concerning cnc, cnc_Col expression reproduced cnc expression in the HL, i.e. partly overlapping that of Col (Fig 2B) [26]. Since rapid morphogenetic movements occur in the head between embryonic stages 8 and 12, we compared cnc_Col and cnc_Col^mut expression by ISH. Lower cnc_Col^mut expression was observed in the HL, compared to cnc_Col, suggesting that Col regulation of cnc via cnc_Col contributes to robustness of cnc expression (Fig 2C and 2D). Ama_Col expression also overlaped that of Col in the HL (Fig 2E and 2F). Ama_Col^mut expression was not detected in the HL, suggesting a direct control of Ama activation by Col (Fig 2G and 2H). We then examined endogenous Ama transcription and found that it was specifically lost in the HL, in col mutant embryos, confirming that this expression is under direct Col control (Fig 2I and 2J). Together, analysis of Ama_Col and cnc_Col indicated that Ama and cnc expression is under direct Col regulation. Beyond, it showed that Col binding to embryonic chromatin could identify new genes involved in head development.

Col targets in neurons

Col is expressed in various subsets of interneurons displaying diverse molecular identities and neurotransmitter phenotypes [31]. nerfin-1, a TF essential for the expression of a subset of axon guidance genes in nascent neurons and expressed in most neuroblasts and nascent neurons [60], was found among Col targets. We found that nerfin-1_Col overlaps with the neuronal nerfin-1-6 enhancer [61] and is widely active in the VNC at stages 14–15, raising the question of a modular regulation in Col expressing-neurons. We noticed, however, that, beyond stage 15, nerfin-1_Col^mut expression was lost in a few lateral neurons (S4 Fig), which we identified as EL neurons by Eve antibody staining (data not shown). As mentionned above, Col controls eve expression in EL neurons [31], likely directly (S1 Fig). Whether nerfin-1 regulation by Col is only direct or also contributed by Eve in a feed-forward process is an open question. Several Col peaks were found to overlap with DNA fragments driving reporters active in EL neurons in stage 16 embryos (http://www.janelia.org/gal4-gen1) [62]. It will be interesting to see in the future which of the associated genes are direcly regulated by Col in these neurons.

Col targets in dorso-lateral muscles

Col is transiently expressed in muscle PCs at the origin of DL muscles, before being uniquely maintained in the elongating DA3 muscle. Time-series datasets for Mef2 and RNA Pol II binding to chromatin and marks of open chromatin and/or active enhancers, H3K79me3 and H3K27ac [64]; [65]; [66] were suggestive of eya_Col, Mrtf_Col, Oaz_Col and so_Col activity in the mesoderm (S2 Table). We found that eya_Col was stochastically expressed in the DA3 muscle, mainly in thoracic segments (see below, Fig 5A). so_Col and Oaz_Col were expressed in the DA3 and other DL muscles. The expression pattern of so_Col^mut did not, however, significantly differ from that of so_Col (S5 Fig) while Oaz_Col^mut expression was lost (S6 Fig). Lastly, Mrtf_Col expression was observed in virtually all muscles, but not altered upon mutation of the Col binding site (S7 Fig). In two cases, eya_Col and Oaz_Col, their expression was dependent upon the predicted Col binding site. In 2 other cases, so_Col and Mrtf_Col, a direct regulatory role for Col binding was left uncertain.

Cross-regulation between col and eya reveals new complexity in the muscle specification process

eya_Col expression was also observed in the DA3 muscle. Furthermore, this expression was lost with eya_Col^mut (Fig 5A and 5B) and down-regulated by mutating either Col site 1 or 2 (data not shown). eya was previously reported to be broadly expressed in the somatic mesoderm at stage 9, and assigned a rather general function in modulating somatic myogenesis [67], while col was only required in DL muscles [32]. The eya and col mutant phenotypes and eya_Col expression evoked the possibility of eya/col cross-regulation in DL muscle lineages. Immunostaining stage 11 embryos for both Col and Nautilus (Nau), to identify muscle PCs and FCs [68], together with eya ISH, revealed that eya and Col are co-expressed in DL PCs, including the PC giving rise to the DA3 muscle (Fig 5C and 5C’). We therefore examined eya expression in col¹ mutants. To circumvent the lack of detectable Col protein in these mutants, we used the early mesodermal col CRM (col^ECRM-LacZ; [69] to follow col-expressing myoblasts (Fig 5D and 5D’). Triple staining for β-gal, Nau and eya showed that eya expression was strongly decreased, when not lost, in col mutant DL PCs, showing that Col regulates eya transcription at that stage (Fig 5E and 5E’). Since a DA3 phenotype has not been previously noted, we re-examined the eya DL muscle phenotype [67]. Loss of LL1, and malformation of DA3 muscles was observed with high frequency in eya^cli-IID/Df(2L)BSC354 embryos completely lacking eya function (Fig 5F–5H). This was correlated with the loss of muscle Col expression (Fig 5I and 5J), which could be traced back to stage 11, during the process of selection of PCs from the Col-expressing promuscular cluster (Fig 5K and 5L). These results indicated that eya activity was required for high level col expression at that step. Together, the reciprocal changes of eya and col expression in col and eya mutants and related muscle defects, led us to conclude that regulation of eya by Col in muscle PCs is direct, and that cross-regulation between col and eya genes is critical for establishing the Drosophila muscle pattern.

Col appears to directly regulate eya in specific muscle and neuronal lineages, via binding to the same genomic sites (Fig 6). This provides the first example of diversification of Col regulatory functions, through differential regulation of the same gene in different cell contexts.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 6. Scheme for Col direct regulation of ap and eya in specific neuron and muscle lineages.

Col directly controls ap_Col and eya_Col CRM activity in the Tv and dAP neurons and eya_Col and col_Col CRM in the DA3 muscle lineage. The two Col binding sites in eya_Col CRM are not functionally equivalent (double arrow). eya and col cross-regulate each other in the DA3 PC. Control of DopR and Nplp1 expression in Tv and dAP neurons (Baumgardt et al., 2007) could be indirect.

https://doi.org/10.1371/journal.pone.0133387.g006

Col direct gene targets

ChIP-Seq analysis of chromatin from 10–12h old embryos identified 559 Col binding peaks, corresponding to 415 potential Col direct target genes. Studies of 10 targets revealed regulation by Col in different cell-types, confirming the versatility of Col regulatory functions. Of note, our selection of Col targets included genes with known expression profiles and was therefore biased towards positive regulation. It does not exclude that Col could repress some of its targets. 792 regions binding Col/Kn were independently identified in 0 to 12h embryos by the modENCODE cis-regulatory annotation project [70]. The modENCODE genomic regions did not significantly overlap with the positions of Col peak summits mapped here, however, and we could not detect an enrichment of the EBF/Col motif in the modENCODE peak collection. One possibility to explain this lack of overlap is the different source of antibodies used for ChIP. Another genome-wide search for Col/Kn chromatin binding sites was performed in 3^rd instar larvae by the DamID method [59]. 99 genes were found to be both bound and regulated by Col in class IV md neurons, with a broad range of molecular functions possibly contributing to dendritic arbor formation [59]. 15 were found among our 415 genes set (S1 Table), suggesting that they are bound by Col already in embryos. Another set of 34 “neuronal” genes bound by Col in larvae, though not regulated by Col in md neurons, is bound by Col in embryos. (S1 Table) [59]. 9 of them code for TFs, including nerfin-1, for which we obtained evidence of direct regulation by Col in EL neurons. It will be interesting to determine in which neuron subtypes, TFs bound by Col in embryos and larvae are expressed, as potential entry sites to new networks controlling neuron identity from embryos to larvae. A study in the nematode C. elegans concluded that the COE factor UNC-3 was a master regulator of « cholinergic genes » in motor neurons [21]. Our ChIP-seq data failed to detect Col binding to either the Acetylcholine esterase, or Acetylcholine receptor genes, or the choline acetyl transferase upstream region driving expression in cholinergic neurons [71], correlating with the observation that Col is expressed in one cholinergic interneuron per hemisegment, but not in embryonic motor neurons. Among the Col-expressing neurons, the best characterized are Tv1 and dAp. CRM activity demonstrated that both ap and eya expression is under independent, direct Col control in these neurons, with ap_Col being active earlier than eya_Col. Col requirement for ap and eya expression in the Tv1 neuron persists to the adult stage [52]. It will be interesting to see whether the same CRMs remain active from embryo to adult.

Context-dependent Col binding to chromatin and gene regulation

Over-expression experiments showed that Col ability to auto-regulate or activate target genes is highly cell-context dependent, indicating a need to co-operate with other TFs [23]; [33]; [25]. The consensus CCCnnGGGA motif was found close to the center of a large majority of Col ChIP-seq peaks, showing that Col binds in vivo preferentially, if not exclusively, to this motif in the chromatin of different cell types. No other significantly enriched motif was found by MEME analysis of the whole set of 559 Col peaks. This can be related to the diversity of expression profiles and therefore cis-regulatory signatures of Col-regulated genes. Mutating the Col binding site resulted into either loss (e.g., ap, Ama, eya) or decrease (e.g., cnc) of CRM activity, suggesting that Col either activates or up-regulates/maintains gene transcription, depending upon the gene and cell type/developmental process. In other cases, such as so and Mrtf, CRM activity failed to reveal a requirement for Col binding, raising the possibility that Col could prime some genes in the chromatin for their later regulation by other lineage-specific transcription factors, as proposed for EBF [44]; [72]. The genome distribution of Col peaks shows cases of several peaks associated with the same gene (S1 Table), for example eya. Independent mutation of each Col binding site in eya_Col showed that one site is strictly required for CRM activity in dAp neurons, while the other provides robustness to this activity. Both sites contribute to CRM activity in Tv neurons and the DA3 muscle. The presence of several Col peaks can therefore reflect redundance of cis-elements required for robustness/precision of Col regulation in a given cell, or/and differential regulation in different cell types, with possible evolutionary implications.

New loops in transcriptional regulatory networks controlling skeletal muscle identity

eya_Col activity revealed that eya is a direct target of Col in DL muscle PCs. Previously, eya was assigned a rather general role in modulating somatic myogenesis, downstream of the myogenic factor Tinman (Tin)/Nkx2.5 [67], while col was shown to be negatively regulated by Tin and Tail-up/Islet1 in dorsal muscles and cooperate with Nautilus/D-MyoD in DL muscles [68]; [32]. Our discovery of col/eya cross regulation in DL muscle PCs therefore brings to light a new layer of intricacy in the transcriptional control of muscle identity, with evolutionary implication. Indeed, cooperation between Eya1/2, Six1/4 and MyoD (Mrf), and Ebfs and MyoD has been shown to operate during vertebrate myogenesis [73]; [74]; [75]; [76]; [18]. Future investigation of epistatic interactions between muscle iTFs such as Col and myogenic regulators such as eya, Six, and Mrf(s) in patterning the Drosophila somatic musculature, should provide deeper insight into which aspects of the myogenic regulatory network have been conserved and/or diversified during evolution. The question of ancestral functions of COE proteins, which appeared with metazoans [7], is highly speculative, in view of COE pleiotropic functions in extant phyla ([8]; [31]; [77], and references in the above text). Col direct regulation of eya, both in specific muscle and neuronal lineages, may illustrate one mechanism behind the evolutionary diversification of Col biological roles.