Phenotypic robustness conferred by apparently redundant transcriptional enhancers.

Genes include cis-regulatory regions that contain transcriptional enhancers. Recent reports have shown that developmental genes often possess multiple discrete enhancer modules that drive transcription in similar spatio-temporal patterns: primary enhancers located near the basal promoter and secondary, or 'shadow', enhancers located at more remote positions. It has been proposed that the seemingly redundant activity of primary and secondary enhancers contributes to phenotypic robustness. We tested this hypothesis by generating a deficiency that removes two newly discovered enhancers of shavenbaby (svb, a transcript of the ovo locus), a gene encoding a transcription factor that directs development of Drosophila larval trichomes. At optimal temperatures for embryonic development, this deficiency causes minor defects in trichome patterning. In embryos that develop at both low and high extreme temperatures, however, absence of these secondary enhancers leads to extensive loss of trichomes. These temperature-dependent defects can be rescued by a transgene carrying a secondary enhancer driving transcription of the svb cDNA. Finally, removal of one copy of wingless, a gene required for normal trichome patterning, causes a similar loss of trichomes only in flies lacking the secondary enhancers. These results support the hypothesis that secondary enhancers contribute to phenotypic robustness in the face of environmental and genetic variability.

The cis-regulatory region of the svb gene integrates inputs from multiple gene regulatory networks to generate a complex pattern of transcription in the embryonic epidermis of insect species6,8. Svb protein then activates many downstream genes, ultimately resulting in trichome morphogenesis9,10. Three enhancer modules located in a 50 Kb region upstream of the svb transcription start site (called 7, E, and A) together recapitulate the complete svb epidermal expression pattern11. Partial loss of function of all three enhancers led to the evolutionary loss of the long, thin quaternary trichomes (indicated in Fig. 1a and 2a) on first-instar larvae of D. sechellia, a species that is closely related to D. melanogaster11. Evolution of svb expression patterns has probably also contributed to parallel loss of quaternary trichomes in the D. virilis group, species of which are distantly related to D. melanogaster12.

Figure 1

The svb cis-regulatory region in D. melanogaster

a, Drawing from the lateral perspective of a D. melanogaster first instar larva. The domain producing quaternary trichomes on the fifth abdominal segment is enclosed in a red outline. b, Diagram of the region upstream of the svb first exon, showing the positions of the five enhancers for this locus: DG2, Z, A, E, and 7. The expression driven by these enhancers in quaternary cells is shown in purple in the diagrams below each enhancer. The piggyBac elements used to generate Df(X)svb are shown as blue triangles. c,f, Expression pattern driven by D. melanogaster Z::lacZ (c) and DG2::lacZ (f) in the 5th and 6th abdominal segments of a stage-15 embryo (purple). An anti-Dusky-like antibody was used to stain developing trichomes (green). d,g, Expression pattern driven by D. simulans Z::lacZ (d) and DG2::lacZ (g). e,h, Expression pattern driven by D. sechellia Z::lacZ (e) and DG2::lacZ (h). β-galactosidase protein produced by D. melanogaster Z::lacZ is expressed in the cytoplasm; β-galactosidase from all other constructs is localized to the nucleus.

Figure 2

Effect of Df(X)svb on the number of quaternary trichomes

a, The lateral patch (green) and dorsal region (blue) in which trichomes were counted. The green and blue boxes correspond to the regions where the Z and DG2 enhancers are expressed strongly. The primary, secondary, tertiary, and quaternary cell types are indicated with horizontal lines above the photograph. The arrow marks the spiracle that was used to set the lower boundary for the green box. The blue box was positioned directly above the green box. The red box identifies the stout tertiary trichomes, which were excluded from the counts. b, Number of trichomes in the lateral plus dorsal region (blue and green boxes) of the fifth abdominal segment of the larva. Open circles give trichome numbers for each individual (n=10); the black symbols and lines show the mean ±1SD. Embryos from each of the two genotypes (C108 and Df(X)svb) were reared at three different temperatures: 17°C, 25°C, and 32°C. c, Cuticle images showing the quaternary trichomes in the lateral patch (below) and dorsal region (above) of Df(X)svb first-instar larvae that developed at the three different temperatures. The genotype by temperature interaction term of a two-way ANOVA was highly significant (F = 27.57, P<0.0001).

We noticed that a 41 kb region upstream of the three known svb enhancers displays high conservation among drosophilids, but contains only one small gene named SIP3 (Fig. 1b and S1). To test whether this region contained additional svb enhancers, we assayed reporter constructs encompassing the entire region (Fig. S1). Two constructs drove expression in the dorso-lateral epidermis in patterns that reproduced part of the native svb expression pattern (Fig. 1c, f, and S2). To characterize the precise expression domains driven by these newly-discovered enhancers, we performed co-immunodetection of the β-galactosidase reporter and of the Dusky-like protein, an early component of developing trichomes10.

The Z enhancer drove expression in many cells that produce quaternary trichomes (Fig. 1c). This expression overlaps the patterns driven by the three previously identified enhancers: 7, E, and A (Fig. 1b). The DG2 enhancer drove expression in a more restricted region (Fig. 1f) that overlaps the domain of expression driven by the E enhancer. Both Z and DG2 drive expression starting at stage 14 of embryogenesis (Fig. S2), which is similar to the time when svb mRNA can be detected in epidermal cells.

Given the redundant expression patterns of Z and DG2 with the three previously identified enhancers, we sought further evidence that Z and DG2 encode functional svb enhancers. We reasoned that if the Z and DG2 enhancers contribute to trichome patterning, then they should have evolved in a similar way to the previously discovered 7, E, and A enhancers; they should retain expression in species that also produce quaternary trichomes (such as D. simulans), and show reduced expression in D. sechellia, which has lost quaternary trichomes. We therefore assayed Z and DG2 enhancer constructs made with orthologous regions from D. simulans and D. sechellia. These regions were straightforward to identify because the genomes of these species are 3-5% divergent from D. melanogaster. The D. simulans Z and DG2 enhancers drove an expression pattern similar to that of the orthologous D. melanogaster enhancers (Fig.1c, d, f, and g), which suggests that Z and DG2 contribute to the production of quaternary trichomes both in D. melanogaster and in D. simulans. In contrast, the Z and DG2 enhancers from D. sechellia drove low levels of expression in only a few cells (Fig. 1e and h). The weak expression driven by the D. sechellia Z and DG2 constructs is consistent with the partial loss of expression driven by the D. sechellia A, E, and 7 enhancers and with the loss of quaternary trichomes in this species11.

To further assess the functional importance of the Z and DG2 enhancers, we generated a 32 kbp chromosomal deficiency on the X chromosome that removes both enhancers, called Df(X)svb (Fig. 1b). As a control, we used strain C108, which carries both of the parental transposable elements that were used to generate the deletion. Df(X)svb flies are viable and display no gross abnormalities. We examined first-instar larvae in detail and found that, when Df(X)svb embryos developed at the optimal temperature for development (25°C), larvae exhibited slightly fewer quaternary trichomes (Fig. 2b) and a reduction in the size of the lateral sensory bristles (Fig. S3). These results suggest that, under optimal conditions, Z and DG2 are functional enhancers of the svb gene that contribute to fine details of trichome patterning and perhaps to bristle morphogenesis. Despite this evidence that the Z and DG2 enhancers contribute to svb activity, their loss-of-function phenotype was considerably weaker than one would have expected, given the strong expression driven by these enhancers. We reasoned that this resulted from the fact that the Z and DG2 enhancers drive overlapping expression with the enhancers 7, E, and A, and that the latter three enhancers drive expression levels that are sufficient to generate most larval trichomes when embryos develop under optimal conditions11.

We therefore considered the hypothesis that Z and DG2 contribute to phenotypic robustness. Natural populations experience repeated stresses over evolutionary time, including variable temperatures. Temperature influences membrane fluidity, enzymatic activity, protein folding, protein-protein interactions, and protein-DNA interactions13,14. Organisms have evolved developmental mechanisms to buffer the phenotype in the face of temperature-induced cellular changes. We reasoned that sub-optimal temperatures might destabilize the transcriptional output of genes during embryogenesis and that secondary enhancers may confer a selective advantage by maintaining transcription above a required minimum threshold. We therefore tested the effect of Df(X)svb in embryos that had developed at 17°C and 32°C, temperatures close to the extremes at which Drosophila embryos survive15. We counted the number of quaternary trichomes in the regions where Z and DG2 are expressed strongly (Fig. 2a). The svb gene is an ideal target for this analysis, because quantitative changes in svb transcription influence trichome density, size and shape16.

Control embryos reared at all temperatures produced similar numbers of trichomes, implying that the number of trichomes is canalized against temperature variation17. The number of trichomes on Df(X)svb larvae reared at 25°C was similar to the number on control C108 larvae at all temperatures (Fig. 2b). In contrast, Df(X)svb larvae displayed a highly significant decrease in trichome numbers when reared at extreme temperatures (Fig. 2b). The primary and tertiary trichomes look normal on Df(X)svb larvae at all temperatures (data not shown), which is expected, because the Z and DG2 enhancers do not drive expression in cells producing primary and tertiary trichomes.

In principle, the loss of trichomes observed on Df(X)svb larvae reared at extreme temperatures may have resulted from mechanisms acting independently of the Z and DG2 enhancers. If the effects observed with Df(X)svb resulted from loss of the Z and DG2 enhancers, then reintroducing a functional Z or DG2 enhancer into a Df(X)svb background should rescue some trichomes. We tested this hypothesis for the Z enhancer. We generated a transgene carrying the svb cDNA under the transcriptional control of the Z enhancer and introduced it onto the third chromosome of Df(X)svb flies. At extreme temperatures, the Z::svb cDNA transgene completely rescued wild type trichome numbers in the lateral patch (Fig. 3a and S3). However, in the region dorsal to the lateral patch, the rescue is very weak or absent (Fig. 3b and S3). This is consistent with the fact that Z drives expression at high levels in the lateral region, where rescue is observed, and only weakly in a small number of cells of the dorsal region (Fig. 1). The loss of canalization in the dorsal region of Df(X)svb larvae may be caused by loss of DG2, which drives expression mainly in this dorsal region. These results demonstrate that Z contributes to phenotypic robustness. Moreover, the rescue of trichome numbers by a transgene introduced onto a different chromosome from the svb locus suggests that Z does not need to be in intimate contact with other svb enhancers or with the svb basal promoter to buffer svb function. Instead, we hypothesize that Z contributes to phenotypic robustness simply by boosting levels of svb transcription in the cells in which Z drives expression.

Figure 3

Rescue of the temperature-dependent trichome loss in the lateral patch by a Z::svb transgene

a,b, Trichome number in the lateral patch (a) and dorsal region (b) of the 5th abdominal segment of larvae with the genotypes C108, Df(X)svb, and Df(X)svb::svb. Open circles represent trichome numbers for each individual (n=10); the black symbols and lines show the mean ±1SD.

Given this evidence that the Z enhancer, and possibly also DG2, contribute to robustness against environmental perturbations, we asked whether these enhancers also buffer against genetic perturbations. For example, Boettiger & Levine18 have reported that two Dorsal target genes that possess “shadow” enhancers maintain synchronous transcriptional activation across Dorsal+/- embryos, whereas two Dorsal target genes that seem to lack such “shadow” enhancers display less synchrony in Dorsal+/- embryos. Therefore, we tested the effect of reducing Wingless signaling, which is required for normal development of quaternary trichomes7, by crossing the Df(X)svb allele and the C108 control allele into a background heterozygous for a wingless null allele. At 25°C, the Df(X)svb-/+ embryos produced significantly fewer trichomes than C108;wg-/+ embryos, Df(X)svb embryos, and C108 embryos (Fig. 4). The combined results suggest that the Z and DG2 enhancers buffer against both environmental and genetic perturbations.

Figure 4

Effect of Df(X)svb-/+ on the number of quaternary trichomes

C108 and Df(X)svb embryos that were heterozygous for a null allele of wingless were reared at 25°C. Quaternary trichomes were counted as described in the legend to Fig. 2. A two-way ANOVA reveals a highly significant genotype by temperature interaction (F=7.79, p=0.0084), which is caused by a large reduction in the number of trichomes on Df(X)svb-/+ larvae relative to all other genotypes.

These results indicate that the production of larval trichomes is normally canalized and that this is accomplished, at least in part, through transcriptional activation mediated by the svb secondary enhancers that are removed in Df(X)svb.

The svb locus contains multiple enhancers with overlapping expression patterns. Similar patterns of overlapping enhancer activity have been found for the cis-regulatory regions of the Drosophila genes sog1, vnd3, and brinker1 and for the cis-regulatory regions of the mouse genes sonic hedgehog4 and sox102. Moreover, it has been estimated that 50% of the target genes of the transcription factor Dorsal contain shadow enhancers5. Therefore, the presence of additional enhancers in cis-regulatory regions may be a common signature of developmental regulators. This may explain why, in previous reports, animals carrying deletions of highly conserved enhancers have not displayed observable phenotypic defects when reared in standard laboratory conditions19,20.

Developmental buffering is likely to result from many molecular mechanisms. For example, deletion of the conserved miRNA miR7 in D. melanogaster has no obvious phenotypic effect in normal laboratory conditions, but it is required to canalize the expression of the gene atonal under fluctuating temperatures21. Similarly, our results indicate that svb secondary enhancers have a minimal role at optimal conditions for development, but that they are essential to buffer the trichome phenotype under genetic or environmental variability. Secondary enhancers are likely to be evolutionarily maintained by selection for robustness against temperature fluctuation, genetic background effects22, and expression noise23.

Methods summary

The target regions were PCR-amplified from genomic DNA from D. melanogaster, D. simulans, and D. sechellia. These PCR fragments were cloned into pCaSpeR-hs43-lacZ or placZattB and integrated into the D.melanogaster genome to test their enhancer activity. The precise expression domains of the enhancer constructs were determined by double staining with a mouse anti-βGal antibody (Promega) and a rabbit anti-Dusky-like antibody10 and then by examining stained embryos with a confocal microscope. Df(X)svb was generated via flipase-induced deletion of the DNA between two FRTs present in C108. We made 0-3 hour embryo collections and reared embryos to hatching at different temperatures. First-instar larvae were mounted in 1:1 Hoyer's:lactic acid mixture and cuticles were imaged with phase-contrast microscopy. Trichomes were counted using ImageJ. A null allele of wingless (wg2)24 was used to obtain males with the genotypes Df(X)svb /Y, +/wgIG22 and C108/Y, +/wgIG22.

Region name	Forward primer	Reverse primer
D.melanogaster DGO	TGGCCTGTGCCATGTGTGCGAGTACG	TGGGTGCGCAATTATGCCGCCAGAGC
D.melanogaster DG1	CTGGGTGTGTGTGCAATATGTGAGC	GTGAGGGTACAAGGCGAAATCGAAA
D.melanogaster DG2	AATTGTTCGCACGCTTCGCTCTAA	GATTGGTGCCGAGAGGTGAAAGTG
D.melanogaster DG3	GGCCACAACTCAATGGCAAAAATG	CAGCAGCGAATCAAGACGAAAGGT
D.melanogaster DG4	CCCCCGTCTTTGTCTGTTTGTCTG	GGAACACAATCTGCCTGCCTGACT
D.melanogaster DG5	TATCCTTTTACGACGCCCCTGTGTC	GATTCGGTTCCTTGGGATTGGATTT
D.melanogaster Z	ATTGCTTCGGCTCTCCCGTTA	TTGTGTGGCTCACTTGGCAC
D.simulans Zprox	GTGAAAGATCGGATCCGTCT	GTTCGTATCGCCCACTTGAAT
D.simulans Z	ATTGCTTCGGCTCTCCCGTTA	TTATGTGGCTCACTTGGCAC
D.sechellia Z	ATTGCTTCGGCTCTCCCGTTA	TTGTGTGGCTCACTTGGCAC
D.simulans DG2	TGCTTTTCCAACCCCTCAGTT	GGGGGTGCAGGCTATTTTGTTC
D.sechellia DG2	TGCTTTTCCAACCCCTCAGTT	GAGGGTGCAGGCTATTTTGTTC