Global analysis of target genes of 21 members of the ZAD transcription factor family in Drosophila melanogaster

Abstract

The zinc-finger associated domain (ZAD) family is the largest transcription factor family in dipteran insects. Still, their functional significance is barely recognized in the literature due in part to their resistance to mutagenesis screens in genetic studies. Therefore, we employed in vitro techniques to identify the DNA-binding characteristics of several members of the Drosophila melanogaster ZAD family in an effort to study their target genes. In this comprehensive investigation, we constructed a panel of GST-Zinc finger (ZnF) array chimera from 21 selected ZAD proteins and used them to select binding sites from an oligonucleotide library by employing electrophoretic mobility shift assays (EMSA). Samples of the binding population were sequenced and used to derive DNA-binding consensus sequence for each member. These consensus sequences were tested for complex formation with their respective protein chimera and the specificity of binding ascertained by competition EMSA. Bioinformatics tools were used to identify potential genetic targets. The identified consensus sequences were distinct for each member and the putative genomic targets were clustered in the regulatory regions of specific genes. This appears to be consistent with a conservation of function between members and also suggests that the overlapping functions of ZAD proteins are the result of positive selection to maintain redundancy and not simply artifacts of recent expansion. Putative target genes suggest a major role of the ZAD family members in the regulation of several early developmental genes including homeobox transcription factors.

Fig. 1

ZAD-ZFPs and GST-ZnF recombinant proteins: (A) schematics of the natural architecture of 21 ZAD family members (Section 1). (B) Diagrammatic representation of 22 GST-ZnF recombinant plasmids constructed from 21 ZAD members. Purified GST-ZnF recombinant proteins are shown for the first four (C), an additional ten (D), and another eleven (E) ZAD family members along with broad range protein marker (M), and purified GST affinity tag as control (indicated by black arrows) (panels D and E also contain GST-ZnF proteins of CG8319, CG9797 and CG31365 that were not considered further as mentioned in Section 3.2). (F) Binding site selection and EMSA. Purified GST-ZnF recombinant proteins were bound to an end labeled 49-mer oligonucleotide library. Representative binding for four ZAD members are shown. Red arrows indicate the shifted DNA–protein complexes.

Fig. 1

ZAD-ZFPs and GST-ZnF recombinant proteins: (A) schematics of the natural architecture of 21 ZAD family members (Section 1). (B) Diagrammatic representation of 22 GST-ZnF recombinant plasmids constructed from 21 ZAD members. Purified GST-ZnF recombinant proteins are shown for the first four (C), an additional ten (D), and another eleven (E) ZAD family members along with broad range protein marker (M), and purified GST affinity tag as control (indicated by black arrows) (panels D and E also contain GST-ZnF proteins of CG8319, CG9797 and CG31365 that were not considered further as mentioned in Section 3.2). (F) Binding site selection and EMSA. Purified GST-ZnF recombinant proteins were bound to an end labeled 49-mer oligonucleotide library. Representative binding for four ZAD members are shown. Red arrows indicate the shifted DNA–protein complexes.

Fig. 2

Derivation of binding site consensus by MEME analysis: for each member, several binding site selected clone’s inserts were sequenced and used to build consensus binding sites with the MEME alignment tool. Position weight matrices for each of the 22 GST-ZnF constructs are shown along with their associated ‘E’ values.

Fig. 3

Competition EMSA: oligonucleotides containing the wild-type consensus (wt) and scrambled (mutant) sequences were used in a series of competition gel-shift assays. Wild-type consensus sequences efficiently competed as identified by the loss of the DNA–protein complex. The mutant sequences were much less effective in dislodging the complex, even at 40×, indicating an essential nature of conserved positions. Competition EMSA results are presented for CG12219 (full gel) (A) and regions containing the DNA–protein complexes for CG7938, CG30020 and CG17958 (B). ^$For CG7938, the lane with wt-shifted complex is from a replicate gel and for *CG17958, the 20× wild type competition lane had a background radioactive smudge that obscures the complex.

Fig. 4

Functional annotation cluster enrichment scores from DAVID analysis of the predicted ZAD target genes: the target genes are grouped into each cluster based on related gene ontology terms. Each cluster was assigned an enrichment value based on the geometric mean of the Fisher exact test ‘p’ values for each of its GO terms. Shown in the heat map are the values for the most enriched clusters within the ten most prevalent gene ontology categories (rows) for each of the GST-ZnF constructs (columns). If no cluster that can fit the category is identified, then a value of 0.0 is assigned. Shaded in gray are values greater than 0.0 and less than 1.0; in yellow are values between 1.0 (inclusive) and 2.0; in red are values 2.0 and above. Clusters indicated in * contained gene ontology terms associated with multiple categories.

Fig. 5

Graphical representation of highly enriched gene ontology terms: the fold enrichment (Y axis) for a selection of gene ontology terms (X axis) describing the target gene sets from ZAD proteins CG7928 (A), CG14710 (B), CG10309 (C), and CG8145 (D) are shown.

Fig. 6

Comparisons of serendipity binding site sequences: (A) 1. our EMSA-derived binding site consensus for sry β; 2. reported in vivo binding site for sry β; 3. reported nuclease protection assay derived binding site consensus for sry β; 4. reported nuclease protection assay derived binding site consensus for sry δ; and 5. our EMSA derived binding site consensus for sry δ. Also shown are side-by-side comparisons of the previously reported sequences for sry β and sry δ (B), and of our consensus sequences for sry β and sry δ (C).