The evolution of new morphological features is due predominantly to modifications of spatial patterns of gene expression. Changes in the expression of a particular gene can result from alterations either in its cis-regulatory sequences or in the deployment and function of the trans-acting transcription factors that control it, or both. Understanding the evolution of new morphological traits thus requires both the identification of genes that control trait formation and the elucidation of the cis- and trans-modifications that account for gene expression differences.
Evolution of cis-regulatory elements has been proposed to be a major source of morphological diversification because mutations in regulatory elements can produce discrete tissue-specific expression pattern changes while avoiding deleterious pleiotropic effects1, 2, 3. In the best-studied cases of gene expression changes underlying morphological divergence, cis-regulatory modifications have been proposed4, 5, 6, occasionally suggested by genetic evidence7, 8, 9, 10, but have only rarely been formally demonstrated11 or analysed at the molecular level12, 13. It is currently not known whether the evolution of new morphological traits occurs largely through the modification of pre-existing cis-regulatory elements or from the generation of new elements; neither is it understood how many or what kinds of modifications are required for a regulatory element to drive a novel pattern.
To address these issues, we have analysed the evolution of a conspicuous male-specific wing pigmentation pattern in Drosophila biarmipes, a species closely related to Drosophila melanogaster 14 (Fig. 1). Wing pigmentation patterns in insects are highly diversified and have various biological functions including mimicry, camouflage, thermoregulation, and mate selection15. In D. biarmipes, the sexually dimorphic wing pattern is associated with a courtship behaviour in which males display their wings conspicuously to the females, suggesting a function for this spot in mate choice16, 17. This wing spot has evolved recently in some species of the D. melanogaster group, such as D. biarmipes, and it is absent from close outgroup species such as D. pseudoobscura 17 (Fig. 1).
Figure 1: Expression of the Yellow protein prefigures adult wing pigmentation.
The conspicuous spot of dark pigmentation present at the tip of the male wing of Drosophila biarmipes (left) is a new trait evolved among species of the Drosophila melanogaster group14,46 (about 15 Myr of divergence; divergence time is 60–80 Myr for the family Drosophilidae30), superimposed on the ancestral pattern of uniform grey shading and darker veins found both in D. melanogaster and in D. pseudoobscura, a species from the sister D. obscura group (25 Myr of divergence29,30). In all three species the male pupal distribution of Yellow in the wing, revealed by a specific antibody (right), foreshadows the adult pigmentation.
High resolution image and legend (55K)Formation of wing pigmentation results from the conversion of melanin precursors diffusing from the veins into pigment deposits at specific positions along the wing, wherever converting proteins are present18. The product of the yellow (y) gene is required for the production of black pigments, and the distribution of its product prefigures adult pigmentation patterns11, 19. The Yellow protein is expressed uniformly at low levels throughout the developing wings of D. melanogaster and D. pseudoobscura, where it imparts a low overall level of melanic pigmentation. In contrast, in D. biarmipes, in addition to the low, uniform expression, Yellow protein is highly expressed in an anterior distal spot19 (Fig. 1). This tight correlation between a novel Yellow expression pattern and a novel pigmentation pattern prompted us to ask whether regulatory evolution at the y locus underlies the novel distribution of the Yellow protein in D. biarmipes, or whether this is due to changes in trans-acting regulators of y.
Regulatory changes in cis to the yellow locus
To test whether the observed differences in Yellow expression between D. biarmipes and D. melanogaster (Fig. 1) are due to changes at the y locus, we transformed D. melanogaster with green fluorescent protein (GFP)-reporter constructs containing non-coding DNA from the D. biarmipes y (y bia) locus. If relevant evolutionary changes have occurred in cis, then the reporter gene might be regulated in D. melanogaster in a manner similar to the native y gene in D. biarmipes. If, however, the changes have occurred in trans, the D. biarmipes y regulatory element might drive reporter expression similar to that of the y gene in D. melanogaster (that is, uniformly). We found that D. melanogaster transgenic flies carrying the entire 5' region (8 kilobases; Fig. 2a) of the y bia gene (5' y bia) express GFP in the pupal wings in a pattern similar to the native D. biarmipes Yellow expression (Fig. 2b). Low levels of GFP are uniformly distributed across the wing, and higher levels of GFP are confined to the distal part of the anterior compartment. This result shows that the transcription factors deployed in the developing wing of D. melanogaster recognize y bia cis-regulatory sequences. Furthermore, the D. biarmipes-like expression pattern in a D. melanogaster trans-regulatory context shows that evolutionary changes in Yellow expression involve primarily cis-regulatory modifications at the y locus, which presumably entail the gain (or loss) of binding sites for transcription factors.
Figure 2: Cis-regulatory changes at the yellow locus are responsible for species-specific differences in Yellow distribution.
a, The organization of the y locus is similar in Drosophila melanogaster, Drosophila biarmipes and D. pseudoobscura. b, The entire 5' region of D. biarmipes y, comprising sequences between the coding sequences of y and the closest predicted gene (CG3777), is sufficient to drive reporter GFP expression in D. melanogaster at a time and in a pattern similar to those of y expression in native D. biarmipes. The y bia intron does not drive wing expression other than in the marginal sensory bristles, and the partial locus drives expression in a pattern similar to the entire 5' region of y bia (not shown). Black boxes, coding sequence; grey boxes, fragments analysed in transgenic constructs.
High resolution image and legend (70K)The 5' y bia element does not recapitulate the precise restriction of the native spot of Yellow expression; higher levels of reporter protein expression extend along the proximal–distal axis, indicating that additional regulatory differences exist between D. biarmipes and D. melanogaster. Additional reporter constructs suggest that these differences are trans effects or are due to cis-acting elements located outside the region we have tested. The unique intron of the D. biarmipes y gene does contain another cis-regulatory element (for all developing sensory bristles) but has no activity in the wing other than in these sense organs. Furthermore, a transgene containing the 5' non-coding, 5' untranslated region, first exon, intron and second exon sequences (partial locus, Fig. 2a) is expressed in a similar pattern to that of the 5' y bia element, indicating that the differences in y expression are not due to any of these sequences (data not shown).
Having localized major regulatory differences to the y bia 5' region, we next investigated whether the novel cis-regulatory activity of the y bia region arose in a pre-existing regulatory element or evolved de novo in the D. biarmipes lineage.
Evolution of a wing-specific cis-regulatory element
In D. melanogaster, analysis of the y regulatory region has revealed that an 800-base-pair (bp) element located 1 kilobase upstream of the transcription start site, named the wing mel element (Fig. 2a), is sufficient to drive gene expression throughout the pupal wing (Fig. 3b) and is necessary for adult wing pigmentation11, 20, 21. We hypothesized that functional modifications of this element might account for differences in Yellow expression between the wings of D. melanogaster and D. biarmipes. There is strong sequence conservation of this portion of the y locus between the two species (Fig. 3a and Supplementary Fig. 1). We transformed D. melanogaster with a GFP-reporter construct containing a 920-bp fragment from D. biarmipes orthologous to the D. melanogaster wing element (wing mel), termed wing bia (Fig. 2a). This fragment drives a reporter pattern resembling that driven by the 5' y bia element, with slightly less contrast between the levels of overall expression in the wing and in the anterior distal area (not shown). A larger fragment encompassing wing bia, named wing bia large (1,542 bp; Fig. 2a) drives a reporter pattern identical to the 5' y bia element (Fig. 3b). These results indicate that the sequences required for the strong anterior-distal activation of Yellow expression in D. biarmipes pupal wings are located within and immediately adjacent to a wing-specific cis-regulatory element that is orthologous to the wing mel element.
Figure 3: The cis-regulatory sequences governing spot formation evolved in the context of an ancestral wing enhancer.
a, Conservation of the wing element sequence between D. biarmipes (bia) and D. melanogaster (mel) or D. pseudoobscura (pse) determined by Vista47 with a 10-base-pair window length; only conservation above 75% is shown as solid boxes. Arrows show the boundaries of the left and right fragments. b, Reporter expression driven by the orthologous wing elements and its subfragments left and right (columns) of D. melanogaster (top), D. biarmipes (middle; the wing bia large element is shown) and D. pseudoobscura (bottom), all expressed in D. melanogaster. The ubiquitous expression driven by the outgroup species wing pse element (expression is present in vein cells at a lower levels comparable to those in left pse) shows that the sequences responsible for the spot pattern in D. biarmipes have evolved in the context of an ancestral wing regulatory element. The sequences controlling the spot pattern are separable from those controlling general expression in D. biarmipes (left and right). Note that the posterior boundary of activity of the left bia construct lies near or at the anterior–posterior compartment boundary.
High resolution image and legend (57K)To determine whether the novel wing bia sequences evolved within an ancestral wing cis-regulatory element, we examined D. pseudoobscura, an outgroup species that belongs to a clade generally devoid of wing pigmentation patterns other than the grey (light black) homogeneous shading (Fig. 1). Phylogenetic character reconstruction suggests that the pigmentation spot was present in the common ancestor of D. biarmipes and D. melanogaster and has been lost in the D. melanogaster lineage22, 23. There is substantial sequence conservation at the y gene between D. biarmipes and D. pseudoobscura (Fig. 3a and Supplementary Fig. 1), which allowed us to identify a region in D. pseudoobscura that is orthologous to the wing bia element, named wing pse (724 bp). This wing pse element drives ubiquitous wing expression (Fig. 3b), demonstrating that a functional wing element is ancestral to the D. melanogaster/D. biarmipes lineage and that sequences within and/or adjacent to this element were modified to control high levels of expression in the anterior distal part of the wing in D. biarmipes.
To understand the organization of the wing bia element and to localize its novel functional sequences, we further dissected the wing bia element. We found that the sequences necessary for the anterior distal expression are separable from those controlling the general wing expression in D. biarmipes. Two complementary, non-overlapping sequences of the wing bia element, right bia and left bia (Fig. 2a), drive respectively ubiquitous expression throughout the wing blade and strong activation in the anterior distal area of the wing (Fig. 3b) (as do the complementary subfragments from the wing bia large element, right bia large (not shown) and left bia large; Fig. 4b). A similar dissection clearly separates two wing-specific complementary functions in D. pseudoobscura (ubiquitous expression, and expression around the veins) but yields non-functional elements in D. melanogaster (Fig. 3b). These results indicate that sites in both regions of the wing element are required for its function in D. melanogaster and D. pseudoobscura, and that some or all of the novel sequences in D. biarmipes responsible for the specific anterior distal wing expression of Yellow are located in the left bia large element, hereafter referred to as the spot element. The distinct and robust activities of the two parts of the wing bia element raise the possibility that the wing element has been subfunctionalized in D. biarmipes into two elements controlling expression throughout the wing and in the spot, respectively.
Figure 4: The spot element evolved through the acquisition of sites for both activators and repressors.
a, Schematic of the 675-base-pair spot element showing the boundaries of deletion constructs and the location of identified binding sites. b, d–f, Expression of GFP driven by the spot element (b) and related constructs. c, The anterior border of expression of the selector gene engrailed abuts the spot element expression domain. d, A 196-base-pair element drives the spot pattern but is derepressed in the posterior compartment. e, Deletion of bp 425–453 abolishes activity of the spot element, indicating that sites required for activation lie within this region. f, Disruption of two characterized Engrailed binding sites from the spot element derepresses reporter expression in the posterior compartment. g, The two candidate sites are bound specifically by the Engrailed protein in vitro. Increasing amounts of Engrailed homeodomain–GST fusion protein (0.25–5 nM) specifically shift labelled DNA oligonucleotides representing native sequences containing putative binding sites (left part of each gel) but not sequences in which Engrailed sites have been mutated (underlined in the sequences, right part of each gel). Addition of anti-GST antibody supershifts complexes. Addition of specific (spe.) or non-specific (non-spe.) unlabelled competitor DNA (+ , 50 ng; + + , 500 ng) reveals the specificity of the formation of complexes. Supershift and competition experiments were performed in the presence of 5.0 nM protein.
High resolution image and legend (93K)Multiple sites evolved in the wing spot element
In principle, the evolution of the spot pattern could arise through gaining binding sites for a single transcription factor that is expressed precisely in the cells that form the spot pattern. Alternatively, the spot pattern could result from the evolution of a combination of binding sites for multiple activators, as well as potential repressors that might restrict expression to this area. Resolution of these possibilities bears on the general question of the number of steps involved in the evolution of new patterns of gene expression and cis-regulatory element function.
To distinguish between these possibilities, we derived a series of reporter gene constructs with smaller portions of the 675-bp spot element. A 196-bp construct (335–530; Fig. 4a) retained activity in the anterior distal region of the wing, although we noted that reporter expression now extended into the posterior compartment (Fig. 4d). This suggested that one or more sites critical for activation resided in this 196-bp element and that one or more sites necessary for the restriction of expression from the posterior compartment resided outside it.
To localize further sequences required for activation of the spot element, we constructed a series of small deletions spanning the length of the 196-bp element (Fig. 4a, e, and data not shown). We found that a fragment lacking the internal sequences from bp 425 to 453 completely lacked reporter expression (Fig. 4e). This indicates that sequences required for activation in the spot are located within or overlap with bp 425–453. Together, these results indicate that sites for both at least one activator and one repressor have evolved in the spot element.
Direct regulation of the spot element by Engrailed
We next sought to identify potential trans-acting factors that regulate the spot element in D. biarmipes. The conspicuous posterior boundary of gene expression observed with the left bia and spot elements (Figs 3b and 4b) is reminiscent of the compartment boundary of the wing24 defined by the anterior border of expression of the selector transcription factor Engrailed25 (Fig. 4c). The posterior expansion of GFP expression in the deletion constructs shown in Fig. 4d would be consistent with posterior repression of the spot element by Engrailed.
To test whether Engrailed might be a direct regulator of the wing bia element, we searched the spot sequence for putative Engrailed binding sites26 and identified several candidate sites. Two of these sites, clustered within 43 bp, are specific to the D. biarmipes spot element (absent from D. melanogaster and D. pseudoobscura elements; Supplementary Fig. 1) and one site is located outside the 196-bp construct that exhibits some reporter expression in the posterior compartment (Fig. 4d). Gel-shift experiments on the native and mutated versions of these two sites showed that Engrailed binds specifically to them in vitro (Fig. 4g). Disruptions of these two Engrailed binding sites in the context of the spot element result in the specific derepression of reporter gene expression in the posterior compartment (Fig. 4f). These results show that the selector protein Engrailed directly represses the expression of the y gene in the posterior compartment of the wing and is one of the inputs that shapes the contours of the wing spot in D. biarmipes.
Multistep and multigenic evolution of the spot
Although multiple cis-regulatory modifications at the y locus have produced a profound evolutionary change in Yellow protein expression, it is important to ascertain whether changes in this one gene are sufficient for the evolution of the physical trait or whether additional evolutionary events are required. We have found that changes at y alone are not sufficient to create a pigmentation spot.
D. melanogaster y mutants carrying the D. biarmipes y gene (Fig. 2a, partial locus) recover only their species-specific pigment patterns; no wing spot is generated (not shown). Additional loci must therefore be involved.
The formation of pigment patterns is a multigenic process, and evolution at other pigmentation loci could also contribute to pattern evolution18, 27, 28. The male spot of D. biarmipes is also associated with the localized downregulation of the melanin-inhibiting product of the ebony (e) gene during wing development19 (Fig. 5c), in a pattern that is approximately the inverse of Yellow expression. This suggests that, at least, both the repression of e and the activation of y are necessary for the formation of a dark spot. Consistent with this hypothesis is the observation that in D. melanogaster e mutants carrying the y bia partial locus transgene (Fig. 2a), a slight darkening is observed specifically in the anterior area of the wing where yellow is strongly expressed (data not shown). However, this darkening is not comparable to the intense pigmentation spot of D. biarmipes. Changes in the expression of other pigmentation genes must also be involved. Furthermore, we have not been able to test whether changes in the trans-acting regulatory network of D. biarmipes might also contribute to the unique patterns of gene expression in the area of the wing spot. Taken together, these results indicate that the evolution of the novel pigmentation pattern of D. biarmipes required changes at multiple loci.
Figure 5: Concerted changes in the expression of Yellow and Ebony underlie the evolution of novel wing patterns.
a, The distant species D. guttifera (a member of the D. quinaria group48) has evolved a complex pattern of dark spots located at the intersection of wing veins and where campaniform sensilla form. The grey shading is also reinforced in some interveins. b, The pupal distribution of Yellow also prefigures this adult pattern. c, In Drosophila biarmipes, the spatial repression of Ebony is also associated with the formation of the adult male spot of pigmentation19. d, This repression of Ebony associated with pigmentation patterns seems to be general, because it is also seen in D. guttifera, where the adult spots will form.
High resolution image and legend (137K)To determine whether the inverse regulation of expression of y and e is a general mechanism for the evolution of novel wing pigmentation patterns, we examined the expression of these proteins in D. guttifera, a species that separated from the D. melanogaster lineage about 40 Myr ago29. This species has independently evolved a strikingly different and more complex wing pigmentation pattern (Fig. 5a). We found that the pattern of expression of the two proteins also exhibits an inverse relationship with higher levels of Yellow (Fig. 5b) and lower levels of Ebony (Fig. 5d) in the pupal wing where the eventual adult pigmentation spots will form. This indicates that the evolution of both y and e expression is involved in the formation and evolution of novel wing pigmentation patterns in drosophilids.
Chance caught on the wing: novelty by co-option
In drosophilid flies, the shape of the wings and the pattern of venation have not changed much over 60–80 Myr of evolution30, 31. Their development and patterning are largely understood in D. melanogaster and the regulatory proteins involved are conserved32. One such protein, the selector protein Engrailed, is a deeply conserved feature of the compartmental organization of arthropod segments and appendages. In the Drosophila wing, Engrailed is part of the regulatory circuit that sequentially organizes the patterning of the anterior–posterior axis33. Here we have shown that the activity of this transcription factor has been co-opted to control a feature of the novel wing pigmentation pattern in D. biarmipes through the evolution of specific binding sites within, or in the immediate vicinity of, a wing-specific regulatory element of the y gene. Because the expression driven by the spot element is also spatially modulated in D. melanogaster, this indicates that other conserved components of the wing trans-regulatory landscape (that is, one or more activators) have similarly been co-opted by the evolution of binding sites within the y wing element.
These findings suggest a general means by which novel expression patterns and characters can arise (Fig. 6a). Specifically, the random mutation of ancestral cis-regulatory elements (including point and insertional mutations) generates potential binding sites. If and when these sites can be recognized by transcription factors expressed in cells in which the ancestral element is active, the pattern or level of gene expression may be modified (Fig. 6a), in a manner similar to the mechanism of gene co-option demonstrated by the vertebrate crystallin genes34. The patterns of expression of the eligible transcription factors are initially cryptic with respect to the target gene or trait, but these cryptic 'prepatterns' are revealed once functional binding sites have evolved in target genes. In this sense, and in this example, evolution is precisely a matter, as Jacques Monod put it, of 'chance caught on the wing'35, 36.
Figure 6: Cryptic prepatterns and the evolution of novel gene expression patterns through the evolution of cis-regulatory sequences.
a, The upper panel shows a model of the conserved landscape of transcriptional regulators that pattern and shape the Drosophila wing (green and pink represent repressor and activator, respectively). The evolution of binding sites for a subset of these regulators in the yellow wing cis-regulatory element (coloured stars) co-opts them to modify yellow expression (lower panel). Combined with other regulatory changes at other loci, the changes at the y locus result in a novel pigmentation spot. b, Wing pigmentation patterns similar to D. biarmipes (left) or D. guttifera (right) evolved independently in other fly families (here Otitidae and Lauxaniidae).
High resolution image and legend (73K)This model has two specific implications for the evolution of novel wing patterns. First, it explains how the observed diversity of wing pigmentation patterns might result from combinations of the numerous transcription factors expressed in the wing. Each of these combinations might constitute a distinct prepattern for pigmentation genes such as y or e, provided that the corresponding binding sites evolve in the proper cis-regulatory context. For instance, some of the spots on the wing of D. guttifera surround the sensory organs located on the veins, which form at similar positions in most drosophilids. This raises the possibility that transcription factors involved in the positioning of these landmark organs have been co-opted to change y or e regulation in D. guttifera. Second, this model might explain the widespread repeated evolution of strikingly similar pigmentation patterns observed in distantly related species (for instance, pigmentation patterns similar to those studied here have evolved independently in other dipterans; Fig. 6b). The evolutionary stability of the trans-regulatory landscape in drosophilid wings, reflected by the strong conservation of the wing shape and venation pattern in the family, suggests that similar pigmentation patterns might arise in parallel through the repeated evolution of binding sites for the same transcription factors in cis-regulatory regions of pigmentation genes.
Methods
Fly stock and maintenance
Flies were bred at 25 °C on Wheeler–Clayton37 or cornmeal38 medium. Constructs were transformed into D. melanogaster yw mutants as described previously39, 40. The CantonS strain was used as wild-type D. melanogaster. Drosophila pseudoobscura, Drosophila biarmipes and Drosophila guttifera stocks were obtained from the Tucson stock centre (stock numbers 14011-0121.94, 14023-0361.01 and 15130-1971.10, respectively). All mature D. biarmipes males of this stock exhibited the wing spot. The en-Gal4 and UAS–GFP stocks were obtained from the Bloomington Drosophila stock centre.
Immunochemistry
Pupal wings (70 h after puparium formation), still attached to the fly, were allowed to unfold in water after removal of the pupal cuticle. Flies were transferred to phosphate-buffered saline (PBS), in which the wings were cut off with a razor blade. Wings were fixed flat for 15 min between a slide and a coverslip in 4% formaldehyde PBT (PBS containing 0.03% Triton X-100), transferred on ice to a scintillation vial in the fixing solution for a further 15 min, sonicated briefly in the fixative with a Branson 200 ultrasonic cleaner, fixed for a further 30 min, washed with PBT, blocked for 1 h in PBT containing 1% bovine serum albumin, stained with a rat anti-yellow or a rabbit anti-ebony primary antibody19 and revealed respectively with a fluorescein isothiocyanate (FITC)-conjugated anti-rat antibody or FITC-conjugated anti-rabbit IgG antibody (Jackson Immunoresearch).
Cloning
The D. biarmipes y locus sequence was amplified by direct and inverse polymerase chain reaction (PCR; details are available from the authors on request). The entire 5' region was amplified by PCR with primers designed in the coding sequences of y and the closest gene upstream of y in D. melanogaster (CG3777; ref. 41). All y fragments for reporter constructs were cloned into a customized version of the P-based transformation vector42 from which one of the two gypsy insulators had been removed and a new polylinker had been added. Fragments from D. melanogaster and D. pseudoobscura were amplified by PCR from genomic DNA and specific primers designed using available genome sequences43, 44 (see Supplementary Table 1 for primer sequences).
Biochemistry
The D. melanogaster Engrailed homeodomain sequence was cloned into the glutathione S-transferase (GST) gene fusion vector pGEX-3X (Amersham Bioscience). The GST fusion protein was purified by affinity chromatography45. DNA probes for electrophoretic mobility-shift assays were double-stranded oligonucleotides labelled with 32P by end-filling in at both ends with the Klenow fragment of DNA polymerase I. Single-stranded oligonucleotides were annealed at a final concentration of 0.1 
M in 10 mM Tris-HCl pH 7.5 containing 0.1 M NaCl and 1 mM EDTA. Sequences of the oligonucleotide pairs were as follows: native sequences, 5'-TTTCCGCCTAATTGATG-3' and 5'-TTTCATCAATTAGGCGG-3', 5'-TTTTGCCAATCATTTTT-3' and 5'-TTTAAAAATGATTGGCA-3'; mutated versions, 5'-TTTCCGCCTcccTGATG-3' and 5'-TTTCATCAgggAGGCGG-3', TTTTGCCgggCATTTTT-3' and 5'-TTTAAAAATGcccGGCA-3'. Labelled probes were purified with G50 Sephadex beads (Sigma) on chromatography columns (Bio-Rad). DNA-binding assays, competition experiments and gel migrations were performed with 10–15 fmol of labelled probes (about 104 c.p.m.) following a published protocol26; they were pre-run for 0.5 h and run for 1.5 h at 4 °C on 8% native polyacrylamide minigels in 0.5 
Tris/borate/EDTA buffer pH 8.3. Non-specific competitor consisted of herring sperm DNA (Sigma) and the specific competitor was as used elsewhere26.
Wing imaging
Adult wings were mounted flat in Hoyer's medium38 and processed for bright-field imaging with a 4 or 10 dry lens on a Zeiss
Axiophot microscope equipped with a Kontron
charge-coupled device camera. For all reporter lines, pupal wings 70–90 h after puparium formation were mounted flat between a slide and a coverslip in PBT, without fixation, and imaged immediately with an Optiphot confocal microscope (Nikon) equipped with a 4 dry lens and a BioRad
1024 system. Antibody-stained preparations were mounted in glycerol and imaged.
-galactosidase transformation vectors for promoter/enhancer analysis in Drosophila. Biotechniques 29, 726−732 (2000) | 
