|Home | About | Journals | Submit | Contact Us | Français|
There is a growing interest in the evolution of transcription factor binding sites and corresponding functional change of transcriptional regulation. In this context, we have examined the structural changes of the ADF-1 binding sites at the Adh promoters of Drosophila funebris and D. virilis. We detected an expanded footprinted region in D. funebris that contains various adjacent binding sites with different binding affinities. ADF-1 was described to direct sequence-specific DNA binding to sites consisting of the multiple trinucleotide repeat . The ADF-1 recognition sites with high binding affinity differ from this trinucleotide repeat consensus sequence and a new consensus sequence is proposed for the high-affinity ADF-1 binding sites. In vitro transcription experiments with the D. funebris and D. virilis ADF-1 binding regions revealed that stronger ADF-1 binding to the expanded D. funebris ADF-1 binding region only moderately lead to increased transcriptional activity of the Adh gene. The potential of this regional expansion is discussed in the context of different ADF-1 cellular concentrations and maintenance of the ADF-1 stimulus. Altogether, evolutionary change of ADF-1 binding regions involves both, rearrangements of complex binding site cluster and also nucleotide substitutions within sites that lead to different binding affinities.
There is increasing interest in the understanding of sequence evolution of non-coding DNA. It has long been claimed that phenotype diversification among species does not only involve alterations of biochemical properties of translated gene products, but also depends much more on differentiations of spatiotemporal expression of genes within a tissue or throughout the whole organism (1). Numerous studies have supported that mutations within cis-regulatory regions underlie a variety of phenotypic differences in morphology and physiology (2,3).
A problem in the understanding of cis-regulatory sequence evolution is that few general rules have been elucidated yet to relate it to functional evolution of gene expression (4,5). Inter-specific sequence comparisons of non-coding DNA reveal conserved regions and many of them likely are cis-regulatory elements (6–11). But despite obvious indications of selective constraint the structure and sequences of cis-regulatory modules change over time, even in cases where expression patterns are conserved. For example, the stripe pattern expression of the pair rule gene ‘even-skipped’ (eve) is highly conserved in Drosophila. However, the functional analysis of the eve stripe 2 enhancer revealed functional differences between closely related species and functional convergence between distantly related species (5,12,13) and it was proposed that stabilizing selection has not only maintained phenotypic constancy of eve gene expression, but also allowed mutational turnover of functionally important sites within the stripe 2 enhancer.
A future approach to comprehensively understand the relationship between cis-regulatory DNA sequences and gene transcription might therefore be the establishment of a framework of quantitative probabilistic models (14,15). However, crucial information is still needed to be implemented for a realistic prediction of transcriptional outputs.
Although high-throughput approaches more rapidly gain data of protein–DNA interactions and binding preferences of transcription factors, the investigation of case studies is justified because high-throughput methods are often not sensitive enough for special characteristics of selected components. For example, transcription factors often recognize multiple distinct sequence motifs (16) that might be better resolved in more meticulous small-scale experimental approaches. Overall, the understanding of transcriptional regulation and its evolutionary diversification among species requires a detailed knowledge of DNA–protein interactions of every transcription factor involved. It is therefore desirable to focus on the analysis of regulatory changes of well-studied genes for which experimental data exist.
The Adh gene expression and transcriptional regulation of Drosophila melanogaster have been intensively studied and several regulatory mechanisms have been proposed to account for differential Adh transcription in a characteristic spatiotemporal pattern (17–29).
The transcription factor ADF-1 binds among other genes at the distal and proximal regulatory promoters of the Adh gene of D. melanogaster. While ADF-1 activates Adh transcription through binding at the distal promoter, the function of the interaction at the proximal promoter has remained unclear (27,30). The Adh proximal promoter region is partially conserved in a wide range of Drosophilidae species and putative ADF-1 binding sites are detected. In D. funebris, the Adh gene lacks the distal promoter and its regulatory promoter is diverged compared with other species of the subgenus Drosophila such as D. virilis (31,32). We have studied the interaction of ADF-1 with the D. funebris Adh regulatory promoter, its binding preferences, site diversification and functional contribution to Adh transcription. We show that ADF-1 binds to multiple adjacent recognition sites within an expanded ADF-1 binding region at the D. funebris Adh regulatory promoter. ADF-1 contains a Myb/SANT-like DNA binding domain of approximately 80 amino acids that was described to direct sequence-specific DNA binding to a site consisting of multiple trinucleotide repeats. The ADF-1 binding consensus was described as a repeat sequence (33). However, we found that high-affinity ADF-1 binding sites do not match this consensus and have proposed a new binding consensus. Although regions with more adjacent binding sites revealed also increased ADF-1 binding in vitro, only moderatedly higher activation of Adh transcription was observed. We speculate that different regional expansions of ADF-1 binding site clusters might result in differential response to varying cellular ADF-1 concentrations, similarly to what had been suggested for homotypic binding site clusters, scattered over larger genomic regions (34,35). Therefore, not only the nucleotide substitutions but also gains and losses of recognition sites are important in the evolution of the ADF-1 binding regions.
ADF-1 was expressed in BL21 cells (Novagen) with Adf-1 coding sequences of D. melanogaster (GenBank accession number NM206028), D. funebris (GQ922007) and Scaptodrosophila lebanonensis (AJ538295) cloned into the pASKIBA37p expression vector (IBA) and N-terminal His-tagged ADF-1 was purified from inclusion bodies from 1-l liquid culture after 4h induction at RT, following Gaul et al. (36) with some modifications. Inclusion body precipitate was dissolved in 20-ml denaturation buffer (250mM HEPES, 500mM NaCl, 1mM EDTA, 8M urea, pH 8) and dialyzed twice for 2h in 1l of dialysis buffer (50mM HEPES, 0.5M NaCl, 40mM imidazole, 10% glycerol, pH 8.0). ADF-1 was purified from renaturaded protein solutions with a 1-ml His-trap FF column (GE Healthcare).
Labeling of DNA fragments for electromobility shift assay (EMSA) was obtained by PCR either by α32P dCTP incorporation or with end-labeled oligonucleotides (37). Labeling of DNA fragments of the EMSA shown in Figure 1 was carried out by PCR with 2.5mM dATP, dTTP, dGTP, 0.25mM dCTP and variable amounts of α32P dCTP. For equal labeling of the different fragments, the amounts of α32P dCTP were adjusted to a ratio α32P dCTP/cold dCTP that in average led to one incorporation of α32P dCTP per double-stranded DNA molecule. PCR products were purified by excision from 7% acrylamide gels (1× TBE: 89mM Tris, 89mM boric acid, 10mM EDTA; pH 8.3) followed by electroelution in dialysis tubes and phenol:chloroform extraction.
EMSA binding reactions were performed with 2–70fmol labeled DNA mixed with 12.5ng sheared salmon sperm (unspecific competitor DNA) per microliter reaction volume, 0.5 volumes of 2× binding buffer I (65mM HEPES pH 7.6, 0.1mM EDTA, 12.5mM MgCl2, 1mM DTT, 10% glycerol, 100mM KCl) or 2× binding buffer II (20mM Tris pH 7.9, 0.2mM EDTA, 8mM MgCl2, 100mM NaCl, 9% Ficoll 400, 100µg/ml BSA, 0.4% NP-40, 2mM DTT) and purified ADF-1 (0–200ng). Binding reactions were incubated for 15min at 23°C. Samples were loaded on a 7% acrylamide gel in 0.5× TBE and 0.05% NP-40 and electrophoresis was carried out at 400V for 2min and 200V for 1–2h in 0.5× TBE buffer (44.5mM Tris, 44.5mM boric acid, 5mM EDTA; pH 8.3).
DNase I footprinting binding reactions were carried out in binding buffer I with 6fmol end-labeled DNA 0–150ng ADF-1, for 20min at 23°C. Reactions were mixed with 20µl of diluted DNase I (10mM MgCl2, 20mM CaCl2, 1.5–6ng DNase I) and incubation was stopped after 30s at 23°C with addition of 40µl stop solution (0.3% SDS, 15mM EDTA, 600mM Na acetate, 250µg/ml yeast tRNA, 15µg/ml proteinase K). Phenol:chloroform extracted samples were run on 5% sequencing gels. G+A chemical sequencing DNA ladder was applied to each gel.
For the methylation interference assay, end-labeled DNA fragments were partially methylated (approximately one methylated guanine per DNA molecule) with dimethylsulfate (DMS) as described in Carey and Smale (38). Preparative EMSA experiments were carried out with 300–400fmol of end-labeled and partially methylated DNA and 1.5–2µg ADF-1. After electrophoresis the bands of shifted and free DNA were excised and purified by electroelution and phenol:chloroform extraction. Precipitated DNA was resuspended in 1M piperidine and was incubated for cleavage for 30min at 90°C, followed by four rounds of freeze drying with dry ice and a speed vac (Heto), and resuspension in 10µl ddH2O. Equal amounts of each sample (10000–15000 cpm) were run on a 7% sequencing gel.
An anti-ADF-1 antibody was purified from rabbits, immunized with four boosts of 400µg ADF-1. The antibody was purified by affinity chromatography from serum using recombinant ADF-1, with Gln residues 129–136 substituted by three Ala residues in order to remove a redundant epitope in the trans-activation domain. A total of 5-mg modified ADF-1 was coupled to 2-ml NHS-activated sepharose (GE Healthcare) and anti-ADF-1 was purified from 45ml of 1:3 diluted anti-ADF-1 serum (in PBS). The column was washed with 20ml PBS and the antibody was eluted with 100mM glycine, pH 2.3 into 1/10th volume 2M Tris–Cl, 1.5M NaCl, pH 8.0.
Chromatin crosslinking and chromatin extract preparation of embryos was performed as described at FlyChIP (www.flychip.org) but scaled down 5-fold.
Crosslink and homogenization of 250-mg adult tissue was carried out as described in Nègre et al. (39). Samples were centrifuged and rehomogenized twice, coarse matter was sedimented from homogenates by centrifugation at 400g for 1min at 4°C and supernatants were centrifuged at 1100g for 10min at 4°C. Pellets were resuspended in 5ml cell lysis buffer (5mM PIPES, pH 8.0, 85mM KCl, 0.5% NP-40, protease inhibitors). The sonication and lysis steps were identical to the FlyChIP protocol.
Immunoprecipitations (IPs) were carried out as described by Perez-Lluch et al. (40) with 100µl of chromatin extract, 1µg of anti- ADF-1 polyclonal antibody or 1µg unspecific rabbit immunoglobuline G (IgG) (Sigma). Immunoprecipitated and input DNA (sample prior to IP) was quantified by quantitative PCR with Power SYBR® Green PCR Master Mix (Applied Biosystems) and oligonucleotides (Supplementary Table S1) in an ABI PRISM 7700® Sequence Detection System by absolute standard curve quantifications (41) with serial 4-fold dilutions of specific plasmid DNA (Supplementary Figure S1) and three amplification replicas of each sample.
Embryonic nuclear extracts and in vitro transcription experiments were performed according to Kamakaka and Kadonaga (42). The preparation of embryonic extracts was scaled down 10-fold with an approximate yield of 1.8ml of soluble nuclear fraction from 10g dechorionized Drosophila embryos.
Plasmid clones were prepared by standard cloning techniques and contained the Adh genes of D. funebris and D. melanogaster used as the internal control. Plasmids with partial deletions of the D. funebris Adh promoter region were prepared by PCR cloning using specific oligonucleotides (Supplementary Table S1). Plasmid DNA was prepared with Miniprep columns, linearized with BamHI (GE-Healthcare) in order to unlink the two Adh loci and DNA was quantified with a micro-volume spectrophotometer (Nanodrop).
Sequencing and non-denaturating gels were exposed to imaging screens (Fuji) and the screens were scanned with a TyphoonTM 9400 (GE-Healthcare). Band intensities from EMSA, primer extension products, methylation interference and DNase I footprints were determined with the Quantity One software (BioRad) from non-saturated scans of the entire gel using the level option to maximize visualization of band intensity differences.
Statistical analysis was carried out with the Statgraphics Centurion XVI software.
Putative ADF-1 binding sites were previously identified in species of the genus Drosophila (32,43,44). However, the sequence divergence relative to the ADF-1 binding site of D. melanogaster was quite high and we were wondering how those changes at the nucleotide level could potentially translate into differential ADF-1 binding and Adh transcriptional regulation. The ADF-1 binding was preliminarily tested in vitro for D. funebris and D. virilis using EMSAs (Figure 1B) with the ADF-1 protein of the species Scaptodrosophila lebanonensis. This species is an outgroup to the genus Drosophila (45,46). The ADF-1 protein is highly conserved among Drosophilidae species (Lang, M. et al., manuscript in preparation). The 100 amino acids of the N-terminus that contains the DNA binding domain (47) shows 99% of identity between D. melanogaster and S. lebanonensis and 98% between S. lebanonensis and D. funebris. The strongest binding was observed for the fragment of D. funebris. This fragment includes the two predicted ADF-1 binding sites (32) and as a consequence it is expanded compared with other species. Moderate binding was observed with the D. virilis proximal promoter region. No specific ADF-1 binding was observed at the D. virilis distal promoter region, in accordance with the previous predictions (43).
The strong binding of ADF-1 observed in the D. funebris Adh promoter region was interpreted as the result of binding at the two distinct predicted sites (32). DNase I footprinting was carried out with the ADF-1 protein of D. funebris. A large ADF-1 protected region of ~100bp was detected between positions −97 to −196 relative to the Adh transcriptional start site (Figure 2). At low amounts of ADF-1, three subregions were protected that were interspersed with hypersensitive sites (Figure 2). Increasing amounts of ADF-1 progressively expanded the protected regions so that they merged. In comparison, ADF-1 protects a region 40-bp upstream of the Adh distal promoter in D. melanogaster. (27,33). Our result reveals that the ADF-1 footprint is expanded at the D. funebris Adh promoter and that this is one of the explanations for the stronger binding observed in EMSA.
Next, we tested if ADF-1 binds at the D. funebris Adh promoter in vivo by Chromatin IP (ChIP). ChIPs were performed with chromatin extracts prepared from embryos or 4-day adult flies and with the purified anti-ADF-1 antibody. In vivo binding was verified by the enrichment of immunoprecipitated ADF-1 binding regions with respect to control experiments carried out with unspecific IgG (Figure 3). As ADF-1 binding at the D. melanogaster Adh distal promoter has been intensely studied in vitro (27,30), we performed ChIP experiments with this known interaction as a positive control. Immunoprecipitated DNA and DNA in the chromatin extracts (input sample) were determined by quantitative PCR using the absolute quantification method (41).
A 35–75-fold enrichment of ADF-1 binding regions was detected compared with the control experiments with unspecific IgG carried out in parallel (Figure 3B). The fraction of immunoprecipitated D. funebris ADF-1 binding regions with respect to the amounts in each chromatin extracts (input sample) was 0.74% in adult females, 0.9% in adult males and 0.03% in embryos. Comparing those fractions between IPs carried out with anti-ADF-1 or with unspecific IgG revealed highly significant enrichments of anti-ADF-1 immunoprecipitated DNA for embryo, adult female and adult male (Wilcoxon rank sum tests: P<0.00042 each). The total fractions of immunoprecipitated DNA were at similar orders of magnitude compared with experiments with the D. melanogaster Adh distal promoter; although stage- and sex-specific differences were observed. The fractions observed with embryonic extracts were always much smaller (Figure 3C). This was probably due to the much higher cell content of embryonic tissues and more genomic DNA in chromatin extracts. Overall, the results confirmed that ADF-1 interacts with the D. funebris Adh promoter in vivo.
Methylation interference assays were carried out with D. funebris ADF-1 protein in order to determine the guanine nucleotides that contact with ADF-1 in the major groove. However, only a slight interference was observed in the coding strand at positions −163 and −177 (Supplementary Figure S2–1) when we used a DNA fragment with the whole DNase I footprinted region. ADF-1 binding at those positions, however, could not account for the large DNase I protected region of ~100bp. One possible interpretation for this result was the presence of multiple adjacent ADF-1 binding sites within the DNase I protected region. In the methylation interference procedure, EMSA is used to selectively enrich for unmethylated binding sites. In the ideal case of a single binding site per DNA molecule, the bound fraction should get depleted of binding sites with a methylated guanine because it would interfere with the binding of the protein. However, in case of multiple binding sites ADF-1 can bind to unmethylated sites of DNA molecules that carry methylations at adjacent sites. This would prevent the selective enrichment of one particular unmethylated site.
We tested this interpretation by methylation interference assays using PCR products amplified from plasmid clones with partial deletions of the ADF-1 binding region (Figure 4A). In EMSA experiments, the complete deletion of the ADF-1 binding region (p97M) abolished the binding of ADF-1 (Supplementary Figure S3) and longer subregions (p125M, p154M) favored the formation of supershifts. In the methylation interference assays, the partial ADF-1 binding region of clone p126M was the smallest and revealed strongest depletion of cleavage fragments between positions −101 to −110 relative to the Adh transcriptional start site. The complete region (p5-P) resulted in low discrimination of methylation interference at positions −163 to −177, similar to what had been observed initially (Supplementary Figure S2–1). However, the combination of these different experiments enabled us to identify two elements at the 5′ and 3′ extremes of the ADF-1 binding region that showed the highest interference (Figure 4B). Interestingly, the guanine residues within the subregion of clone p126M that matched the previously published ADF-1 binding consensus (33) did only reveal weak contacts with ADF-1. A methylation interference experiment with this clone and purified D. melanogaster ADF-1 was repeated and produced similar results (Supplementary Figure S2–2B).
To test whether the established binding consensus represents well ADF-1 binding preferences, we determined the ADF-1 binding sites within the DNase I protected region at the D. melanogaster Adh distal promoter since the results of a previous MPE·FE(II) footprinting (33) lacked of sufficient resolution for a clear determination of the interacting nucleotides. Therefore, we carried out methylation interference experiments with purified D. melanogaster ADF-1 and the corresponding promoter region (Supplementary Figure S2–3). ADF-1 interacts much more strongly with the 3′ half of the DNase I protected region although the 5′ half contains the ADF-1 binding consensus established by England et al. (33).
Altogether, these results demonstrate that the previously established ADF-1 binding consensus does not represent well the ADF-1 binding preferences although it appears to be conserved among species.
We further wanted to compare the binding affinities of ADF-1 to the different subregions. EMSA experiments were carried out with double-stranded oligonucleotides of partial sequences of the ADF-1 binding region (Figure 5A).
Oligo 3 contained the subregion with the England’s ADF-1 binding consensus. Oligos 1 and 4 contained subregions with most abundant methylation interference sites and Oligo 5 overlapped Oligos 3 and 4. EMSA experiments with these set of oligonucleotides and the protein of D. funebris (Figure 5B) or D. melanogaster (Supplementary Figure S4) were carried out in three replicates for each oligonucleotide and protein concentration. The low amounts of labeled oligonucleotide (2 ftmol) far below the apparent Kd allowed direct comparisons of the different EMSA experiments performed with each end-labeled oligonucleotide. The apparent dissociation constants (Kd) were estimated for each oligonucleotide (Table 1).
The binding affinities differed among oligonucleotides (Kruskal–Wallis test, D. funebris ADF-1: 9.43, P=0.024; D. melanogaster ADF-1: 10.3846, P=0.016). The lowest binding affinity of ADF-1 was observed with Oligo 3 that contains the England’s consensus. The highest affinity was observed either with Oligo 1 or 4 that contain most of the methylation interference sites. A supershift was observed in experiments with Oligo 5. Since Oligos 3 and 4 only contain one of the two sites, the supershift with Oligo 5 indicates that ADF-1 can simultaneously bind at the two sites. The binding affinites of D. funebris and D. melanogaster ADF-1 for the oligonucleotides tested were extremely similar with the exception of Oligo 2. However, no significant differences could be detected for any of the four oligonucleotides (t-test comparisons, Table 1).
We also wanted to compare the ADF-1 binding region of D. funebris with the shorter orthologous region of D. virilis. Although the D. virilis ADF-1 binding region has not been characterized by DNase I footprinting, it could be deduced from DNA sequence comparisons using our data. The D. funebris ADF-1 binding region in the plasmid p5 was replaced by a 41-bp double-stranded oligonucleotide (DFDV) that contained the putative binding region of D. virilis (Supplementary Figure S2–4A). The cloned D. virilis ADF-1 binding region was amplified by PCR and characterized by EMSA and methylation interference assays (Supplementary Figure S2–4). Using D. funebris ADF-1, we identified the guanines that when methylated interfere with the binding in a region of 23bp. This region was much shorter than the complex ADF-1 binding region of the D. funebris Adh promoter. The ADF-1 binding region of D. funebris (p5-P) was also amplified and analyzed by EMSA with ADF-1 of this species in order to compare ADF-1 binding with both regions (Figure 6A). Furthermore, ADF-1 binding to the double-stranded oligonucleotide DFDV was compared with the binding to Oligos 1 and 5 (Figure 6B). The ADF-1 quantities required for 50% fractional occupancy (Supplementary Table S2) were determined by regression in order to compare overall binding of ADF-1 to the different binding targets. This simple analysis was preferred to a binding model fit to the data because of the different numbers of binding sites present in each tested binding region.
Comparisons of the entire binding regions revealed stronger binding of ADF-1 to the larger D. funebris binding region (t-test: t=16.834, P<0.00008; F-test for similar standard deviations: F=0.08858, P=0.16274). On the contrary, we observed a 10% higher fractional occupancy of the D. virilis oligonucleotide DFDV compared with those of Oligos 1 and 5 (multiple range test: three significantly distinct groups). This result shows that the increased overall binding of ADF-1 to the D. funebris promoter region is the consequence of the regional expansion of the ADF-1 binding region and it is not due to higher binding affinities at specific ADF-1 binding sites.
The DNA binding experiments with ADF-1 to the Adh regulatory promoter of different Drosophilidae species demonstrate that the England’s ADF-1 binding consensus does not match ADF-1 binding preferences. Also, previous work with the Adh distal promoter already supported that ADF-1 binds with highest affinity to sites that do not follow England’s consensus (27).
The binding site preferences and affinities resulted to be extremely similar for the different orthologous ADF-1 proteins of D. melanogaster, D. funebris and S. lebanonensis (Supplementary Figures S2–2, S2–5 and S6). This prompted us to combine the results of the reported methylation interference experiments and to determine an ADF-1 binding consensus sequence that better describes the sequence motifs with high ADF-1 binding affinity. Most of the binding sites with high ADF-1 binding affinities contained the motif (Figure 7A). These sequences were aligned to establish the high-affinity consensus (Figure 7B). The sites Dmel 1, Dfun 2 and Dvir 1 were excluded from the comparison because those either contain only a few methylation interference sites and/or have lower affinity values. The obtained sequence logo with the core sequence of 9bp differs from England’s consensus by being 3bp shorter. Furthermore, it differs at the positions 5 and 6 with and , respectively, instead of . These differences apparently account for the difference between high and low ADF-1 binding affinties.
The transcriptional activity of ADF-1 binding sites in the Adh regulatory promoter of D. funebris and D. virilis was tested by in vitro transcription using embryonic nuclear extracts of D. melanogaster.
The DNA templates for in vitro transcription were the plasmid clones with a mutated D. funebris Adh regulatory promoter (Figure 4B) and the D. melanogaster Adh distal promoter as the internal control (Supplementary Figure S2). The transcription of the D. funebris Adh gene was first tested in the presence or absence of the ADF-1 binding region or the element (Figure 8A).
The band intensities of primer extension products of each in vitro transcribed D. funebris Adh template were compared with the in vitro transcription of the D. melanogaster Adh distal promoter template used as internal control. We first compared the amount of D. melanogaster Adh distal transcript obtained from each clone in order to verify independent transcription from the different promoters within each experiment. Transcription with clone p7 which contains only the Adh gene of D. melanogaster was similar to that of clones p5-PS, p5-PGM and p97S that additionally contained the D. funebris Adh insert (Figure 8A). Decreased transcription levels of clone p97GM were due to erroneous plasmid quantifications.
Removal of either the element (p5-PGM) or the ADF-1 binding region (p97S) or both (p97GM) resulted in a progressive decrease of D. funebris Adh transcription from 92.5% to 59% relative to clone p5-PS (analysis of variance (ANOVA): F = 246.34, P<0.00001; Figure 8B). A reduction of 25% of D. funebris Adh transcription was observed with removal of the ADF-1 binding region. This value was three times higher than that observed with the lack of the element, which caused a reduction of 8% only.
Finally, the partial deletions of the ADF-1 binding region of the clones p126GM, p154GM, p125GM and pDVGM with the mutated element were tested for differential D. funebris Adh transcription (Figure 8C). Decreased D. funebris Adh transcription was observed with all deletion clones compared with clone p5-PGM (ANOVA: F=135.921, P<0.00001). Using a multiple range test, we determined five distinct homogenous groups of the six different experiments with similar expression levels detected for clones p154GM and pDVGM. A maximal reduction of 36% was obtained with clone p97GM relative to clone p5-PGM (Figure 8D).
The magnitude of expression differences was comparable with the results of previous in vitro transcription studies with the Adh proximal promoter of D. melanogaster and D. affinidisjuncta (27,48). An advantage to the previous studies was the use of the D. melanogaster Adh distal promoter transcription as internal control experiment. This enabled us to discriminate between small expression differences because of the robustness to the experimental error. A reduction of 8% of transcription was observed with clone p125GM that contained a deletion of the 3′ ADF-1 binding region. For the complementary deletion of clone p126GM, a 20% reduction of transcription was observed.
Clones p154GM and pDVGM showed 13% reduction of transcription. Thus, the D. virilis ADF-1 binding region in the context of the D. funebris Adh promoter partially rescues the loss of D. funebris ADF-1 binding region.
The clones p154GM and p125GM contained binding regions of equal length, but lacked either the 5′or the 3′ region for which ADF-1 shows the highest affinity. The Adh transcription with clone p125GM was 5% higher, thus the 5′ ADF-1 binding region enhanced Adh transcription more efficiently. Overall, regional extensions of the ADF-1 binding region as well as sequence-specific effects seem to enhance Adh transcription in an additive manner.
ADF-1 binds with different affinities to at least four DNA binding sites within a region of ~100-bp upstream of the D. funebris Adh promoter. Our data imply that at low concentrations, ADF-1 binds to the most affine sites and increasing concentrations successively lead to the coverage of the entire binding region. The ADF-1 binding sites with the highest binding affinities are located at the 5′ and at the 3′-end of this region. The orthologous binding region at the Adh proximal promoter of D. virilis is shorter and consists of two adjacent sites Dvir1 and Dvir2 that give rise to two mobility shift complexes in EMSA experiments. This shows that one of the mechanisms of diversification of ADF-1 binding regions is the gain and loss of entire recognition sites. Multiple adjacent binding sites were also reported for other transcription factors in Drosophila, such as bcd, Ubx or Zeste (49–51).
One aspect in the variation of binding region expansion is the probable response to different ADF-1 cellular concentrations. The more sites are encountered in the binding region, the higher is the probability that at least one of them is bound by ADF-1. This is observed in vitro when we compare the ADF-1 binding regions of D. funebris and D. virilis. D. virilis contains two ADF-1 recognition sites in the region. However, the EMSA experiments show a higher fractional occupancy of the ADF-1 binding region of D. funebris which does not contain binding sites with higher affinity than those of D. virilis but with more binding sites. In this case, it is the gain of sites that increases ADF-1 binding. Altogether, the extension of ADF-1 binding regions could potentially compensate for lower ADF-1 expression levels in vivo.
The EMSA experiments with double-stranded oligonucleotides, demonstrate that the binding site Dfun 3 (), which matches the England’s binding consensus is a weak binding site. Isolated from the regional context in Oligo 3, we observed a >10-fold reduced affinity of ADF-1 to this site compared with its affinity for the sites Dfun 1 and Dfun 4. However, this site appears to be functional in the presence of the adjacent site Dfun 4 within Oligo 5, detectable by the formation of the supershifts at protein concentrations equal or higher than 1.003µM. This observation prompted us to consider that ADF-1 binding to the site Dfun 3 might require cooperative interactions with ADF-1 bound at the adjacent site Dfun 4.
Cooperative interactions can be evaluated with titrations of transcription factor concentrations in series of EMSA experiments (52). Using this approach, cooperative binding was determined for a number of transcription factors such as LacI, CAP, Ubx bicoid or λ-repressor (49,52–54). Although we have not performed such a fine analysis of ADF-1 binding kinetics, we used our experimental data to fit the constants to a simple two binding site model, following Senear and Brenowitz (52) (Supplementary Figure S6 and Table S3). The fit of the model to our experimentally determined fractional occupancies strongly indicated cooperative interactions between ADF-1 proteins bound to the two adjacent sites Dfun 3 and Dfun 4 (Supplementary Table S3). Similar results were obtained for the transcription factor Ubx that was found to bind to clusters of multiple individual binding site sequences downstream of its own promoter and close to the transcriptional start site of the Antp P1 promoter (55). Interactions at the Ubx locus with other distant UBX DNA complexes further leads to the formation of DNA loops (49). UBX binding interactions are cooperative and the multiple binding sites have different binding affinities. Cooperative binding was also reported for the Adf-1 structural homolog Zeste (50). The ability to form cooperative interactions could probably be mediated by the C-terminus of ADF-1. This functional domain was also responsible for other protein–protein interactions, such as the formation of homodimers and the protein–protein contacts with TAF 110 and TAF 250 (47).
Assuming cooperative interactions between ADF-1 molecules that bind at adjacent or near binding sites, the functional conservation of low-affinity sites would therefore not require high sequence conservation. However, during evolution the high-affinity site should remain conserved, while adjacent sites could accumulate changes to some extent. In addition, new binding sites could evolve with few changes in proximity to the high-affinity sites. This molecular mechanism would further explain the origin of the different expansions of the two analyzed orthologous ADF-1 binding regions of D. funebris and D. virilis.
We have demonstrated with different experiments that the England’s binding consensus for ADF-1 is not fully compatible with high-affinity binding sites but is fully compatible with low-affinity sites. The analysis of ADF-1 binding regions at other D. melanogaster gene promoters, as Adh proximal, Ddc and Antp P1, (33) encountered several motifs within ADF-1 DNase I protected regions that follow well this low-affinity consensus (Supplementary Figure S5A). However, the footprints were obtained with 5 to 10 times more protein than for the distal Adh promoter that has the high-affinity consensus. Also the protected region in the dpp gene obtained with nuclear extracts showed the low-affinity consensus (56). The re-examination of the ADF-1 binding regions of those genes revealed the lack of high-affinity binding sites (Supplementary Figure S5A). This shows that the motif described by England et al. (33) is an abundant sequence within ADF-1 binding regions and it is obviously the reason why it was proposed as the binding consensus although previous data did not support this interpretation. Footprinting experiments with partial deletions of the D. melanogaster Adh distal promoter showed that the important portion for binding did not contain England’s consensus (27). The re-examination of the deleted sequence that reduced 70% of DNase I protection has revealed that it has the high-affinity consensus, described in this study. The presence of both, high- and low-affinity recognition sites within ADF-1 binding regions may constitute structural features that act as complex binding targets.
We have also shown that the binding site preferences are conserved among the different ADF-1 orthologus proteins of D. melanogaster, D. funebris and S. lebanonensis (Figure 4, Supplementary Figures S2–2 and S2–5) as expected from the high amino acid sequence conservation in the DNA binding domain.
The experimental design of the in vitro transcription experiments allowed us to detect minor changes of transcriptional activity based on the use of an internal control in the same plasmid that reduced the experimental error. The ADF-1 binding region enhances Adh transcription as well as a conserved transcription factor binding site for the Drosophila GATA homolog ABF-1 that was shown to be a functional element at the orthologous promoters of D. melanogaster, D. mulleri (57) and D. affinidisjuncta (48). ABF-1 is expressed in the embryonic fat body (57) and was shown to be present in embryonic nuclear extracts of D. melanogaster (48). Thus, we assume that ABF-1 also binds to this site in D. funebris and activates Adh transcription, as demonstrated for the other species.
The enhancement of Adh transcription by ADF-1 and ABF-1 is probably additive but not synergic. Synergic effects are observed among factors that bind close to the transcriptional start site by simultaneous interactions with different parts of the transcriptional machinery (58,59). The moderately accumulative activation of Adh transcription in response to those elements suggests that in D. funebris ABF-1 and ADF-1 enhance Adh transcription independently and are involved in different molecular mechanisms.
Removal of the GATA element in the presence of the ADF-1 binding region caused a reduction of 8% of the Adh transcription. Similarly, removal of the ADF-1 binding region reduced 25% Adh transcription in the presence of the GATA element and 41% in its absence instead of the 33% expected if it was strictly additive. This shows that the effect of the deletion of one element on Adh transcription is greater in the absence of another and that one element could partially compensate for the absence of the other. The ADF-1 binding region at the Adh proximal promoter of D. melanogaster is less expanded than in D. funebris and does not have any high-affinity binding site. According to previous data, this region did not appear to be critical for Adh transcription (27). Possibly, this smaller ADF-1 binding region is embedded in between other activating elements. Those might enhance transcription and could have compensated the loss of ADF-1 mediated Adh transcriptional activation similar to what we observed in our experiments with the D. funebris Adh promoter.
The orthologous ADF-1 binding region of D. virilis in the context of the D. funebris Adh promoter enhances Adh transcription, but 13% less efficiently than the endogenous D. funebris ADF-1 binding region. This suggests that the expansion of the ADF-1 binding region in D. funebris has increased the effect of ADF-1 on Adh transcription. However, the absolute difference in Adh transcriptional activation is quite small. Therefore, the probable biological function of this expanded region might be better interpreted as an adaptation to compensate for lower absolute expression levels of ADF-1 in certain tissues rather than enhancing Adh transcription. Similar findings were discussed for other transcription factors, such as BCD (35). Clusters of high- and low-affinity binding sites were evaluated in silico in known regulatory regions and it was concluded that the composition of such site clusters scattered over genomic regions of ~500bp potentially mediates differential transcriptional responses to a given transcription factor concentration. A stochastic enrichment of binding interactions does occur in vitro at the characterized ADF-1 binding region as revealed by EMSA (Figure 6B).
Overall, the expanded ADF-1 binding region exemplifies the structural change of the Adh promoter region in D. funebris. We detected conservation of this expansion in the related species D. kuntzei and D. immigrans. At least for D. kuntzei it is known that the Adh gene is also transcribed from a single promoter (44). Molecular phylogenetic analysis of Drosophilidae (46,60) demonstrates that these three species form a monophyletic group within the subgenus Drosophila. The expansion of the ADF-1 binding region at the Adh proximal promoter in this lineage can further be associated with the loss of the Adh distal promoter, either as compensation or as a predating regulatory change that might have allowed its disappearance.
Supplementary Data are available at NAR Online.
Ministerio de Educación y Ciencia (MEC) (grant BFU2004-00668 to E.J.); the Comissió Interdepartamental de Recerca I Innovació Tecnològica (CIRIT) (grant 2005SGR 0166 to M. Aguadé); Generalitat de Catalunya [pre-doctoral fellowship (FI-IQUC 2004-2007) to M.L.]. Funding for open access charge: Comissió Interdepartamental de Recerca i Innovació Tecnològica Generalitat de Catalunya.
Conflict of interest statement. None declared.
We thank A. Pérez-Farrerons for his valuable help in the preparation of embryonic nuclear extracts, methylation interference and in vitro transcription experiments. We are grateful to P. Librado for his help and the estabishment of the phastCons conservation scores. We also thank M.L. Espinàs and S. Lluch-Pérez for their helpful comments on ChIP analysis. The radioactive experiments were performed in the Servei de Radioctivitat of the Facultat de Biologia (UB) under the supervision of C. Benito. The real-time PCR and automated sequencing of clones was carried out in the Serveis Cientifc Tècnics (SCT-UB) under the supervision of A. Amador and R. Seminago.