Functional regulatory evolution outside of the minimal even-skipped stripe 2 enhancer

Developmental enhancers contain multiple binding sites for transcription factors, together specifying the precise time, level and location of gene expression. Classically, minimal enhancers have been identified as the smallest DNA fragments that are sufficient to direct reporter-gene expression in a particular tissue or domain of normal gene expression (Crocker and Stern, 2013). These studies have provided mechanistic insight into how transcriptional logic is encoded in individual enhancers (Spitz and Furlong, 2012; Levo and Segal, 2014). However, in even the earliest studies, it was clear that minimal enhancers are insufficient to define the normal gene expression pattern with complete fidelity (Small et al., 1992). More recently, genomic studies have provided evidence that minimal enhancers are embedded within larger regions containing additional transcription factor-binding sites that may be required for normal enhancer function (Ludwig et al., 2011; Adkins et al., 2016).

Phenotypic evolution results largely from sequence changes in enhancers (Levine, 2010; Prescott et al., 2015), even between closely related species (MacArthur and Brookfield, 2004; Ludwig and Kreitman, 1995; Nord et al., 2013; Crocker et al., 2008, 2010, 2015b; Rebeiz et al., 2009). It is not clear, however, how often functional evolution includes changes within minimal enhancers versus outside of these regions. We have explored this problem through studies of the even-skipped (eve) gene.

The Drosophila melanogaster eve gene is expressed in seven transverse stripes along the anterior-posterior axis in the blastoderm embryo (Macdonald et al., 1986; Ilsley et al., 2013). Minimal enhancers have been identified that each direct expression in either one or two stripes and that together drive expression in all seven stripes (Fujioka et al., 1999). Of all these enhancers, the minimal element for stripe 2 has been studied in the greatest detail (Stanojevic et al., 1991; Arnosti et al., 1996; Small et al., 1992). This enhancer contains multiple binding sites for transcriptional activators (Bicoid and Hunchback) and repressors (Giant, Krüppel, and Sloppy-paired). The collective activity of transcription factor binding to these sites drives eve expression specifically in stripe 2 (Small et al., 1992; Ilsley et al., 2013; Arnosti et al., 1996).

Previously, reporter gene assays were used to investigate the functional evolution of eveS2 from three divergent Drosophila species with transgenic assays in D. melanogaster (Ludwig et al., 2000, 2005; Ludwig et al., 1998). These studies revealed that the eveS2 enhancers from D. yakuba and D. pseudoobscura, which diverged ∼10 and ∼40 million years ago, respectively, from D. melanogaster, drove apparently normal expression in stripe 2. However, eveS2 from D. erecta, which is closely related to D. yakuba, failed to drive appreciable levels of expression.

There are several possible reasons for why the D. erecta eveS2 element does not drive appreciable expression in D. melanogaster (Ludwig et al., 2005). First, the ‘minimal’ functional D. erecta eveS2 element might have been replaced elsewhere in the D. erecta eve locus with a functionally equivalent enhancer. Second, the D. erecta enhancer may contain all the required information to drive native expression, but the enhancer might have evolved to accommodate differences in the D. erecta embryonic environment (Barrière et al., 2012; Fowlkes et al., 2011) – for example, differences in transcription factor concentrations (Crocker et al., 2008; Chahda et al., 2013). Third, the true functional eveS2 enhancer might be larger than defined by the D. melanogaster minimal element and shifts in the locations of key transcription factor-binding sites may have rendered the D. erecta region corresponding to the D. melanogaster ‘minimal’ element unable to drive appropriate expression.

Here, we tested these hypotheses with a functional dissection of the D. erecta eve locus in transgenic D. melanogaster. We found that a large transgene carrying the entire D. erecta eve locus drove normal expression in all seven stripes. This D. erecta transgene rescued downstream eve targets and larval segmentation defects caused by an eve null mutation. We found that regulatory information required for D. erecta stripe 2 expression is located outside of the minimal stripe 2 enhancer region. Finally, we found that the D. erecta minimal eveS2 region lacks multiple Zelda transcription factor-binding sites that are found in D. melanogaster and we demonstrate that normal function of the D. erecta eveS2 enhancer requires Zelda transcription factor-binding motifs located outside of the minimal enhancer element. Zelda is apparently not required for patterning, but instead for making enhancers accessible for regulation (Xu et al., 2014; Foo et al., 2014; Liang et al., 2008; Harrison et al. 2011; Nien et al., 2011; Li et al., 2014; Sun et al., 2015; Schulz et al., 2015). Many studies have suggested that the transcription factor-binding sites required for normal enhancer function are often distributed over regions larger than experimentally determined minimal enhancer elements and our results demonstrate that crucial transcription factor-binding sites can shift during evolution between locations within and outside of ‘minimal’ regions.

D. melanogaster and D. erecta both express eve in seven embryonic stripes at similar levels and locations (Fig. 1A,B). A transgene of the D. melanogaster eveS2 minimal enhancer, when re-introduced into D. melanogaster, drives robust expression in approximately the same region as the native expression of eve stripe 2 (Fig. 1C). However, the orthologous eveS2 fragment from D. erecta does not drive expression in transgenic D. melanogaster (Fig. 1D) (see also Ludwig et al., 2005). A 20 kb region of the D. melanogaster genome surrounding the eve transcription unit is sufficient to drive apparently normal eve expression and rescues transcription of genes that are normally regulated by Eve (Fujioka et al., 1999). To maximize the likelihood that we would capture the entire eve locus from D. erecta, we tested the ability of a ∼47 kb D. erecta fosmid to drive expression in D. melanogaster embryos deficient for native eve function. This fosmid contains D. erecta sequences orthologous to all of the D. melanogaster stripe enhancers (Fig. 1E).

We found that this D. erecta fosmid drove apparently normal expression of all seven eve stripes (Fig. 1F-K). To determine whether the timing, levels and spatial distribution of Eve expression driven by the D. erecta fosmid were correct in D. melanogaster, we tested the ability of the D. erecta locus to functionally rescue gene regulatory networks that are normally regulated by Eve. We found that the D. erecta eve fosmid rescued expression of the segment polarity gene engrailed (Fig. 2A-C). Additionally, the fosmid rescued all segmentation defects in the cuticles of larvae deficient for eve (Fig. 2D-F). Finally, we found that the D. erecta fosmid rescued most flies to adult viability (86%, ±2.5 s.d., n=3), which is similar to rescue levels generated by the 20 kb D. melanogaster eve rescue construct (Ludwig et al., 2011). Together, these results suggest that the 47 kb D. erecta fosmid contains all of the regulatory information required for eve expression in stripe 2 and all other eve stripes.

The regulatory information for D. melanogaster eve stripe 2 and stripes 3/7 is located upstream of the eve promoter (Arnosti et al., 1996; Small et al., 1991, 1992, 1996). Increasing the size of the DNA regions tested for these enhancers increases levels of transcription (Arnosti et al., 1996; Small et al., 1991, 1992, 1996). Additionally, although the minimal D. melanogaster enhancer is sufficient to drive approximately normal expression of eve stripe 2, sequences surrounding the minimal eveS2 element contribute to the robustness of the minimal enhancer (Ludwig et al., 2011). Therefore, we suspected that sequences near the D. erecta eveS2 might be required for normal expression.

To test whether regulatory information 5′ of the D. erecta eveS2 region is required for normal stripe 2 expression, we tested a series of constructs that included the minimal stripe 2 enhancer plus progressively more 5′ DNA (Fig. 3A). Whereas the 855 bp D. erecta eveS2 construct drove very little expression in D. melanogaster (Fig. 3B,C), a fragment containing an additional 832 bp upstream of the minimal enhancer drove weak expression in the stripe 2 domain (Fig. 3D,E). We found that the expression levels driven by this larger D. erecta construct were similar to expression levels driven by the minimal D. melanogaster eveS2 element (Fig. S1). Increasing the D. erecta eveS2 construct size with an additional 1609 bp 5′ of the enhancer, up to the boundary with the minimal stripe 3/7 enhancer, further increased levels of stripe 2 expression and drove weak expression in stripes 3 and 7 (Fig. 3F,G). This indicates both that information critical for stripe 2 expression resides outside the minimal stripe 2 region in D. erecta, and that patterning information for stripes 3 and 7 resides outside of the minimal stripe 3/7 enhancer. Increasing the size of the element further, so that it encompassed the minimal eve3/7 enhancer, did not further increase expression levels in stripe 2, but did increase expression in stripes 3 and 7 (Fig. 3H,I). Together, these results demonstrate that up to 1609 bp 5′ of the minimal D. erecta eve stripe-two enhancer contains regulatory information required for eve expression in transgenic D. melanogaster.

We next performed a computational search for binding sites of four transcription factors that regulate the spatial pattern of the D. melanogaster eveS2 enhancer – Bicoid, Giant, Hunchback and Krüppel – across the regulatory region upstream of the minimal enhancer (Fig. S2). We observed no obvious turnover of these binding sites in D. erecta that would explain the loss of expression from the minimal eveS2 enhancer, consistent with previous studies (Ludwig et al., 2005).

We next examined the spatial distribution of putative binding sites for the Zelda protein. Zelda is expressed ubiquitously in the blastoderm embryo and Zelda protein binds to many enhancers that drive transcription in the early blastoderm embryo (Xu et al., 2014; Foo et al., 2014; Liang et al., 2008; Harrison et al., 2011; Nien et al., 2011; Li et al., 2014). Zelda activity correlates with chromatin accessibility (Foo et al., 2014; Sun et al., 2015; Schulz et al., 2015) and Zelda appears to make enhancers accessible to transcription factors that drive specific patterns of gene expression (Xu et al., 2014; Foo et al., 2014; Li et al., 2014; Schulz et al., 2015; Crocker et al., 2017).

We searched for putative Zelda motifs in the eveS2 minimal element in D. melanogaster, D. sechellia, D. yakuba and D. erecta. We found four sites that are perfectly conserved between these species, and six sites that are present in D. melanogaster but absent in D. erecta. It is possible that the loss of these motifs has led to the loss of activity of the D. erecta eveS2 minimal element. Because the larger D. erecta reporter constructs we tested drove stripe 2 expression, we searched for additional Zelda motifs 5′ of the minimal element. We identified three Zelda motifs in this region in D. erecta, two of which are conserved in D. melanogaster (Fig. 4A,B; Fig. S3). We tested the activity of these motifs in vivo by deleting all three motifs from the large D. erecta eveS2 construct (Fig. 4C,D). Removal of these upstream Zelda motifs abrogated reporter gene expression (Fig. 4D). These Zelda motifs are therefore essential for stripe 2 expression in D. erecta.

To test whether these Zelda motifs were sufficient to drive expression when appended to the minimal D. erecta eveS2, we placed the three Zelda motifs upstream of the minimal element. Strikingly, we found that this construct drove expression in stripe 2 (Fig. 4E,F). Addition of these Zelda sites is therefore sufficient to allow properly patterned expression of the minimal D. erecta stripe 2 element. Furthermore, the patterns of Zelda motif gain and loss suggest that loss of Zelda motifs within the minimal elements prevents the D. erecta minimal enhancer element from driving expression and that gain of a new Zelda motif upstream of the minimal element may be required for normal D. erecta stripe 2 expression.

Our results suggest that functional regulatory evolution has occurred outside the minimal eveS2 enhancer between D. melanogaster and D. erecta. The D. erecta eveS2 region contains binding sites necessary for patterning of expression in stripe 2, but lacks sufficient Zelda binding sites to initiate expression. These results support the hypothesis that Zelda contributes to determining the regulatory state of the eveS2 enhancer in the blastoderm embryo – ON versus OFF – and can be decoupled, at least in part, from the patterns and levels of expression driven by the enhancer (Xu et al., 2014; Foo et al., 2014; Li et al., 2014; Schulz et al., 2015). The potential independence of enhancer state and patterning information has important evolutionary implications, as selection could act on each feature independently. Furthermore, Zelda binding sites are typically located near transcriptional activator-binding sites (Sun et al., 2015), suggesting there might be a constraint on enhancer size, which has widespread consequences for genome evolution.

We have not explored the 777 bp upstream of these Zelda sites, corresponding to the extra DNA in construct D. ere cta eve 2+inter 3\7, for the sequences that drive stronger expression in stripe 2. This is a poorly conserved genomic region and it is therefore surprising that it contains information both for bolstering stripe 2 expression and for patterning stripes 3 and 7. Interestingly, there is an additional Bcd site in the extended region of D. erecta. Although we have not tested this Bcd site, it is possible that there are multiple ways to increase the activity of the D. erecta eveS2. This is consistent with a large body of data demonstrating the rapid evolution, and variable architectures, of transcriptional enhancers with conserved functions (Ilsley et al., 2013; Rastegar et al., 2008; Jin et al., 2013; Brown et al., 2007; Menoret et al., 2013; Lusk and Eisen, 2010; Hare et al., 2008; Junion et al., 2012; Crocker et al., 2016). The function of this upstream DNA region would benefit from further exploration.

There are several inherent caveats to cross-species enhancer analysis. For example, basal promoter interactions (Ohtsuki et al., 1998), changes in regulatory binding specificities (True and Haag, 2001; Gasch et al., 2004), and systems drift (True and Haag, 2001) can make it difficult to interpret results. There is also the possibility that additional regulatory sequences exist outside of the tested enhancer regions, including cryptic shadow enhancers (Frankel et al., 2010; Perry et al., 2010) and autoregulatory elements (Jiang et al., 1991). Interestingly, the minimal enhancers from D. yakuba and D. ananassae also contain only three Zelda sites, but they are both active in D. melanogaster (Ludwig et al., 2005). It is therefore possible that there are different elements in these enhancers that function to increase DNA accessibility. Alternatively, it is possible that the structure of Zelda sites, for example how tightly consolidated they are in D. melanogaster versus D. erecta, could contribute to nuanced differences in expression. In the future, it may be possible to take a synthetic biology approach to understand how the position of Zelda sites within enhancers influences regulatory function and evolution (Crocker et al., 2017).

This study highlights the importance of understanding gene regulatory evolution at the level of the entire locus. Although studies of minimal enhancer elements have provided enormous insight into the mechanisms of gene regulation, the broader DNA regions around minimal enhancers also contribute to enhancer function and evolution (de Laat and Duboule, 2013; Symmons et al., 2016; Gierman et al., 2007; Akhtar et al., 2013; Chen et al., 2013; Dey et al., 2015; Chen and Zhang, 2016; Dunipace et al., 2011). Our results encourage caution in comparative studies of enhancer evolution when functional studies are performed on enhancers from only one species. One goal for future research is to understand how gene regulation evolves at the level of the entire locus, including information such as binding-site composition, the arrangement of enhancers, the activities of intervening sequences, and the distance between functional sequences.

The D. erecta locus was tested by cloning the BDERF01-4213 fosmid, obtained from BACPAC resources (https://bacpacresources.org/), corresponding to D. melanogaster chr2R:5,840,325-5,887,364, into vector TKBL-w+ (Kondo et al., 2009). All enhancer constructs were cloned into the placZattB expression construct with a hsp70 promoter (Crocker et al., 2015b). See supplementary Materials and Methods for complete construct sequences.

D. melanogaster strains were maintained under standard laboratory conditions. Transgenic enhancer constructs were injected into flies by Rainbow Transgenic Flies (CA, USA) and were integrated at the attP2 landing site. The eveR13 mutation was used to test for complementation using the D. erecta fosmid. The lethal mutations were balanced over marked balancer chromosome CyO P(hb-lacZ) to allow identification of mutant embryos by immunostaining for β-galactosidase (Promega, Z3781; 1:250). Relative viability of flies with the D. erecta fosmid, in the eveR13 background, were determined based on the number of adult survivors, as previously described (Ludwig et al., 2011).

Embryos were raised at 25°C and fixed and stained according to standard protocols (Crocker and Stern, 2013). Briefly, primary antibodies obtained from the Developmental Studies Hybridoma Bank were used to detect Eve (3C10; 1:20) and En (4D9; 1:20) proteins, which was followed by detection of primary antibodies using secondary antibodies labeled with Alexa Fluor dyes (1:500, Invitrogen, A31570). Cuticle preps were performed using standard protocols (Stern and Sucena, 2011). Details of the immunohistochemistry protocols can be found in the supplementary Materials and Methods.

Each series of experiments to measure transcript levels was performed entirely in parallel. Embryo collection, fixation and hybridization, and image acquisition and processing were performed side-by-side under identical conditions. Confocal exposures were identical for each series and were set to not exceed the 255 maximum level. Confocal images were obtained on a Leica DM5500 Q Microscope with an ACS APO 20×/0.60 IMM CORR lens and Leica Microsystems LAS AP software. Sum projections of confocal stacks were assembled, embryos were scaled to match sizes, background was subtracted using a 50-pixel rolling-ball radius and plot profiles of fluorescence intensity were analyzed using ImageJ software (http://rsb.info.nih.gov/ij/). Data from the plot profiles were further analyzed in Matlab.

We thank the entire Stern lab for discussion and comments, Claire Standley for comments on the manuscript, and Jessica Cande for the Bourbon fruit cake that proved important for manuscript preparation.