Identification of cis-regulatory modules encoding temporal dynamics during development

Abstract

Background: Developmental transcriptional regulatory networks are circuits of transcription factors (TFs) and cis-acting DNA elements (Cis Regulatory Modules, CRMs) that dynamically control expression of downstream genes. Comprehensive knowledge of these networks is an essential step towards our understanding of developmental processes. However, this knowledge is mostly based on genome-wide mapping of transcription factor binding sites, and therefore requires prior knowledge regarding the TFs involved in the network.

Results: Focusing on how temporal control of gene expression is integrated within a developmental network, we applied an in silico approach to discover regulatory motifs and CRMs of co-expressed genes, with no prior knowledge about the involved TFs. Our aim was to identify regulatory motifs and potential trans-acting factors which regulate the temporal expression of co-expressed gene sets during a particular process of organogenesis, namely adult heart formation in Drosophila. Starting from whole genome tissue specific expression dynamics, we used an in silico method, cisTargetX, to predict TF binding motifs and CRMs. Potential Nuclear Receptor (NR) binding motifs were predicted to control the temporal expression profile of a gene set with increased expression levels during mid metamorphosis. The predicted CRMs and NR motifs were validated in vivo by reporter gene essays. In addition, we provide evidence that three NRs modulate CRM activity and behave as temporal regulators of target enhancers.

Conclusions: Our approach was successful in identifying CRMs and potential TFs acting on the temporal regulation of target genes. In addition, our results suggest a modular architecture of the regulatory machinery, in which the temporal and spatial regulation can be uncoupled and encoded by distinct CRMs.

Figure 1

In silico predictions of transcriptional regulatory motifs within gene sets defined from cardiac remodelling transcriptome dynamics. PWM matrices logos are provided in Additional file 1: Figure S1. The temporal expression pattern of the 13 co-expressed gene sets are shown (as defined in [16]; y axis: expression level, x axis: time). Cardiac expression dynamics was analyzed at 21, 24, 27, 30, 33, 36, 42 and 48 hours APF. The PWMs that rank corresponding gene sets above the automatic enrichment score (ES) threshold are displayed, together with their enrichment score and IUPAC sequence (PWMs converted to IUPAC sequence with RSATools “convert matrix” using Drosophila melanogaster as background model [18]). Color code corresponds to stamp clustering of all retrieved PWM across all clusters. Blue: Nuclear receptor R type motifs; green: bZIP motifs; purple: Mef2 like motifs; Pale purple: GATA-like motifs. Note that no PWMs were recovered above the automatic threshold in cluster 1 and 7.

Figure 2

Enrichment of the nuclear receptor – type GCNF motifs in cluster 12 gene set. A) ROC curve showing significant enrichment in putative NR binding sites (GCNF position weight matrix) among the 42 genes constituting cluster 12 (y axis) compared to a randomized set of 1000 Drosophila genes (x axis) using cisTargetX. The blue curve shows the detection of cluster 12 genes, the red line a random distribution, and the green curve shows a 2 sigma interval from random. B) Expression profile of the 10 best ranked genes. All genes display marked expression increase after 42 h after puparium formation (APF). C) The 10 genes are listed together with the size of the 6 tested CRMs (see Additional file 1: Table S2 and Additional file 1: Figure S2 for details).

Figure 3

Gene reporter assays reveal the temporal dynamics driven by tested CRM. Right: Each CRM predicted by cisTargetX is schematically represented (see Additional file 1: Figure S2 for a detailed description). Vertical black lines represent evolutionarily conserved GCNF clusters as predicted by cisTargetX. Horizontal boxes summarize regions tested by transgenic assays. Left: Reporter activity was examined at different times after puparium formation (APF, indicated on top). A ventral and a dorsal view are shown in each case. Top: CG15545 was also analyzed using a GFP reporter (Figure  4) which serves as a control for βGal staining. No βGal expression is visible, except on pharate adults (96 h APF) in pericardiac cells, which indicate endogenous βGal activity in this tissue. 24 h APF: No βGal activity was observed, except in a few discrete tissues in CG3902-LacZ and CG10175-LacZ flies (respectively in the developing eye and in discrete spots at the basis of the head). 48 h APF: In all LacZ transgenic lines, reporter construct induce βGal expression in a variety of tissues (weak staining was occasionally observed at 42 h, not shown). a: antennae, e: eye, h: head, l: legs, m: muscles, w: wings. 72 h APF: Maximum βGal activity was observed around this time point. 96 h APF: X-Gal staining in pharate adults is due to stability of βGal, since no expression at this stage was observed on GFP-reporter constructs (see Figure 4).

Figure 4

In vivo GFP reporter activities of predicted NR target enhancers. Enhancer GFP reporter assays for CG15545 (A) CG17298 (B) and CG4998 (C) at increasing time points during metamorphosis. Top: timing of pupal development in hours APF (After Puparium Formation). Expression driven by all 3 enhancers start to be detected at 48 h APF (arrow heads) and their activity increases up to 72 h APF. At 96 h APF, GFP signal is almost not detected. All CRMs drive expression in specific regions of the w: wings, l: legs, a: arista, p: proboscis and in different c: cuticle part.

Figure 5

NR motifs are functional within CRMsNR motifs are functional within CRMs. A) Schematic representation of mutations performed in NR motifs of CG4998 (top) and CG7298 (bottom) CRMs. The location of CRMs (black rectangles) with respect to corresponding genes is indicated on top, and putative NR motifs are represented as grey boxes. NR motifs sequences are indicated below, together with the corresponding mutated sequences. Note that in CG4998-CRM both putative motifs are comprised of two inverted overlapping motifs. Motifs in light grey are of lesser quality and were not analyzed in this study. B) GFP expression pattern at 72 h APF in WT (left) and mutated (right) individuals for CG4998 (top) and CG17298 (bottom) CRMs. A ventral and a dorsal view are shown in each case. NR motif mutations lead to a marked increase of GFP signal. C) Time course of GFP expression between 24 and 96 h APF driven by CG4998 and CG4998-mutated CRMs analyzed by qRT-PCR. GFP expression was normalized to RP49 expression levels and expression ratio relative to GFP expression at 48 h in non-mutated CRM are represented (*p < 0.05; **p < 0.001, Student t-test). C’) The dynamics of GFP expression in the mutated CG4998 CRM is significantly different from the one observed with wild type CRM (*p < 0.05 χ2 test).

Figure 6

Eip75B impacts on CG15545-GFP expression. A) Expression dynamics of NR in the cardiac tube during its remodeling from Zeitouni et al. 2007 [16] time course microarray data. Only those whose expression level is above the detection limit at at least one time point were considered. Expression values after within-array and between-array normalization are plotted. B) Expression of these NRs in whole pupae according to flybase modENCODE_mRNA-Seq_U data (http://flybase.org/reports/FBlc0000085.html). Note the particular feature of Eip75B and Hr46 expression which peak at 48 h APF. C) Effect of Eip75B ubiquitous knock down on CG15545-GFP dynamic expression. White prepupae were selected and grown at 25°C (permissive temperature) for 25 hours and shifted at 29°C (restrictive temperature) for 18 h (t0) to 24 h (t + 6 h). Left: expression of GFP in Tub-Gal4, Gal80ts; CG15545-GFP control individuals (WT). Right: expression of GFP in UAS > Eip75BRNAi Tub-Gal4, Gal80ts; CG15545-GFP individuals. Eip75B knock down induces a precocious activation of CG15545-GFP expression noticeable at t0 and lead to increased GFP expression at t3 and t6. Representative individuals. All (30) animals examined displayed this precocious activation of GFP (see Material and Methods).