Rapid Evolution of Sex Pheromone-Producing Enzyme Expression in Drosophila

Rapid evolution of gene expression patterns responsible for pheromone production in 24 species of Drosophila was mapped to simple mutations within the regulatory domain of the desatF gene.


Introduction
Chemical communication is widespread in the animal world [1]. Pheromones can mediate aggregation, signal danger, attract mates, and elicit a variety of other behaviors. The speciesspecificity of pheromonal signals is of decisive importance in kin recognition and in sexual reproduction [2]. Evolutionary transitions in sexual communication have been suggested to govern the early stages of speciation [3]. Moreover, reproductively relevant traits such as sex pheromones are thought to evolve rapidly under sexual selection [4]. These observations raise the possibility that the genes that control how individuals communicate with each other are evolutionarily labile and may display a large degree of functional divergence across taxa.
However, the molecular mechanisms that govern how pheromone signals evolve are not understood. In principle, pheromone signaling could evolve via changes in pheromone production, chemical structure, or reception. In order to decipher how pheromone signaling may evolve, the genes that contribute to pheromone signaling and the mutations that alter it must be identified.
Drosophila male courtship behavior is triggered in part by female pheromones that act either by direct contact and/or by transmission over short distances. Sex pheromones in Drosophila are largely fatty-acid derived hydrocarbons that are present on the fly's cuticle [5]. Among Dipteran fly species, females exhibit considerable divergence in the number or position of double bonds in cuticular hydrocarbons [6]. In the Sophophora subgenus, females of some species, such as D. melanogaster [6][7][8][9][10], D. sechellia [6,9,11], and D. erecta [6] specifically produce long-chain dienes, which are hydrocarbons that contain two double bonds [6]. Alternatively, other species, such as D. serrata, D. pseudoobscura, and D. persimilis [12][13][14] produce these compounds in both sexes. Moreover, other species such as D. simulans, D. mauritiana, D. yakuba, D. teissieri, D. orena, and D. santomea [6,7,9,15,16] do not produce dienes. Cuticular hydrocarbons are suggested to have multiple roles (e.g., protection against dessication and cold resistance [17,18]), and evolutionary transitions in the production of these compounds have been linked to reproductive isolation [9,19]. Males from species that produce dienes dimorphically preferentially court heterospecific females that carry dienes over heterospecific females that do not harbor these compounds [9,20].
In D. melanogaster, the synthesis of dienes depends in part on the product of the desatF gene (also known as Fad2) [21], a desaturase that catalyzes the addition of a second double bond in cuticular hydrocarbons. This gene is transcribed female specifically in adults [21]. The loss of desatF activity causes both a decrease in the amount of dienes, and a decrease in the attractiveness of females to males during courtship [21], suggesting a crucial role in mate recognition. Others have hypothesized that differences at the desatF locus may contribute to the difference in diene production between D. simulans and D. sechellia [22] and other Drosophila species [23].
Here, we investigate the evolution and regulation of the desatF gene across the subgenus Sophophora. First, we show that the desatF locus and its expression are extremely rapidly evolving across the subgenus Drosophila. Second, we demonstrate that the femalespecific isoform of the protein encoded by the sex differentiation gene, doublesex (dsx), directly activates desatF expression in species that express desatF female-specifically. Third, we reveal that one species evolved monomorphic expression of desatF by functional inactivation of an ancestral DSX-binding site in the desatF regulatory region. And finally, we uncover an apparent case of recent stabilizing selection on desatF expression and show that in D. melanogaster, new cis-regulatory inputs in the desatF enhancer have evolved by a series of small deletions. We suggest that rapid evolution in the expression desatF underlies changes in the synthesis of cuticular hydrocarbons, which are likely to alter chemical communication between and within Drosophila species.

desatF Is Expressed Female Specifically in Oenocytes
It has been suggested that the biosynthesis of cuticular hydrocarbons takes place in specialized cells called oenocytes [21,24,25], which are present underneath the dorsal and ventral abdominal cuticle. Histological, confocal, and electron microscopy studies have characterized adult oenocyte cells as being organized in metameric, transverse ribbon-like stripes, that do not cross the midline and are positioned just anterior to the intersegmental region of each segment in the dorsal abdomen [26]. Adult oenocytes cells are also present in each segment of the ventral abdomen [26]. While desatF is known to be transcribed female-specifically, spatially restricted expression in oenocytes has not been demonstrated since all other analyses relied on reverse transcription (RT)-PCR on whole flies [21,23]. We therefore developed an in situ hybridization protocol to visualize mRNA transcripts in adult abdomens. Consistent with previous studies, desatF expression is female-specific in D. melanogaster. Our in situ analysis revealed desatF expression in a pattern that is entirely consistent with previous histological descriptions [26] of adult oenocyte cells ( Figure 1A, purple stripes). Moreover, desatF expression in the adult abdomen is identical to the pattern revealed by a previously characterized GAL4 driver that is active in oenocyte tissue (see ''desatF Expression in Female Oenocytes Is Directly Activated by the Female-Specific DOUBLESEX Isoform,'' below; Figure 4A and 4B, [21]). desatF expression in oenocyte cells is consistent with its role in diene biosynthesis. Our in situ analysis was limited to oenocyte cells and we cannot rule out expression of desatF in other tissues.

Dynamic Evolutionary Changes in desatF Expression Correlate with Diene Production in the Sophophora Subgenus
The profile of cuticular diene production in the Sophophora subgenus exhibits several states, depending on the species (Figure 2, middle column) (unpublished data; [6][7][8][9][10][11][12][13][14][15][16]). Across this group, diene production displays an apparent transition from sexual monomorphism to dimorphism in an ancestor of the D. melanogaster species subgroup ( Figure 2, left column, green arrowhead) and several transitions to a state of no production ( Figure 2, left column, black arrow and not shown). Since desatF has been shown to be crucial for the production of dienes in D. melanogaster [21], we asked whether evolutionary changes at this locus could provide an explanation for these differences.
In order to do so, we tested whether desatF expression correlated with diene production. We cloned the desatF coding region and several kb of its upstream putative regulatory sequence from 24 species within the subgenus Sophophora. desatF expression was assessed by in situ hybridization in species where the gene lacked interruption in its reading frame. In 15 out of 24 species, desatF

Author Summary
Mate selection is a complex process involving communication between potential partners on many levels, such as visual, aural, and olfactory cues. Many animals use chemical signals in the form of pheromones to communicate and correctly recognize individuals of the appropriate species and sex during reproduction. Evolutionary changes in the production of these chemicals have been suggested to contribute to speciation. Yet, the molecular mechanisms governing these transitions have seldom been addressed. Here, we show that expression of the gene desatF, which encodes an enzyme involved in the production of the Drosophila pheromones known as dienes, is highly variable and rapidly evolving across Drosophila species. Changes in desatF gene expression correlate with changes in sex-and species-specific production of dienes. Further, these changes in diene production can be explained by simple modifications in the regulatory regions of the desatF gene, providing a molecular level understanding of the evolution of pheromone production in Drosophila.
appeared to be intact (Figure 2, right column and summarized in Figure 7). desatF expression correlated with diene production (Figure 2, compare the middle and right columns). In species where diene production is monomorphic (e.g., D. pseudoobscura, D. persimilis, D. serrata), desatF is expressed in oenocytes of both sexes. In species that do not produce dienes, desatF was either not expressed (D.  Figure 7). Finally, in species where diene production is strongly female-biased, desatF was expressed only in female oenocytes (D. melanogaster, D. sechellia, and D. erecta).
Our survey revealed that the expression of desatF has evolved with extraordinary rapidity (summarized in Figure 7). Of the 24 Figure 2. desatF expression correlates with evolutionary transitions in diene production from sexual monomorphism to dimorphism. Left column: phylogenetic relationships of the species used in this study. The D. melanogaster species subgroup is highlighted in red. Transitions in diene production from sexual monomorphism to dimorphism, and from dimorphism to amorphism, are indicated by a green and a black arrowhead, respectively. Middle column: diene production in males and females of these species. Diene production across the Sophophora subgenus shows several transitions: from monomorphism to dimorphism in the most recent common ancestor of the D. melanogaster species subgroup, and from dimorphism to amorphism in D. simulans (and other species not shown in the D. melanogaster species subgroup including D. mauritiana, D. yakuba, D. santomea, D. teissieri, and D. orena; summarized in Figure 7). We refer to dienes as any long-chain hydrocarbon (longer than 20 carbons) with two double bonds. Published reports of hydrocarbon profiles in these species are indicated in the column titled Ref. We have also independently validated the hydrocarbon profiles of D. melanogaster and D. simulans (unpublished data). Our analysis of D. takahashii males and females showed that they both produce a C23 diene (unpublished data). *, Dienes are present on D. sechellia males, but account for less than 2% of the total amount of cuticular hydrocarbons present on the fly [6]. The putative expression of desatF in D. sechellia males is likely to be below the detection capabilities of in situ hybridization analyses. Right column: In situ hybridization of desatF expression performed on 4-d-old adults. In all species studied, we observed expression of desatF in abdominal oenocytes (purple stripes) in accordance with their status of diene production. doi:10.1371/journal.pbio.1000168.g002 species analyzed, spanning approximately 40 million years of evolution, we uncovered ten independent evolutionary transitions (not including an additional sex-specific transition discussed below) (summarized in Figure 7). On the basis of the phylogenetic tree (adapted from [27]), desatF was disrupted six times by deletions and insertions of repetitive DNA (summarized in Figure 7, red bars). The expression of an intact desatF was lost independently at least three times (summarized in Figure 7, black bars). Based on our expression data, and in agreement with the inferences of others [23], it appears that female-specific expression was gained once at the base of the D. melanogaster species subgroup ( Figure 2, right column).
These ten evolutionary transitions in the state of desatF expression among such recently diverged species mark the fastest evolving pattern of gene utilization that we know of. We were particularly interested in understanding the mechanism by which female-specific expression of desatF had evolved in the D. melanogaster species subgroup.

The Transition from Monomorphic to a Dimorphic State of desatF Expression Occurred via a cis-Regulatory Mechanism
Transitions of diene production from monomorphic to dimorphic states can be simply explained by a modification of desatF expression. In order to investigate the molecular basis of the regulatory changes underlying transitions in desatF expression, we first identified the location of a cis-regulatory element (CRE) within the desatF locus that governs gene expression in oenocytes. A screen of the 15-kb upstream intergenic region was conducted to find CREs driving enhanced green fluorescent protein (eGFP expression) in 4-d-old D. melanogaster oenocytes. This reporter gene assay revealed a 638-bp CRE immediately upstream of desatF in D. melanogaster, which drove eGFP expression in a manner identical to endogenous gene expression ( Figures 3B and 3C, and S1, region 4).
We next tested whether differences in desatF expression evolved by changes in its identified CRE, or by changes in trans-factors that regulate its expression. We distinguished among these possibilities by analyzing the activity of orthologous CREs of desatF from seven species in reporter gene assays in adult D. melanogaster. If cisregulatory sequence evolution accounts for differences in desatF expression, then reporter activity driven by the CREs should recapitulate the expression pattern of desatF in the species from which the CRE was derived. All transgenes tested (except D. sechellia; Figures 3D and 3E; discussed below) recapitulated endogenous temporal-, spatial-, and sex-specific-expression of desatF of the species from which it was derived ( Figures 3F-3M). Thus, the changes underlying the transition between monomorphic and dimorphic expression of desatF largely occurred in the cisregulatory regions of desatF. In order to elucidate how and when dimorphic desatF expression evolved, we next sought to dissect the molecular mechanisms regulating its expression in dimorphic species.
desatF Expression in Female Oenocytes Is Directly Activated by the Female-Specific DOUBLESEX Isoform In order to elucidate factors regulating dimorphic expression of desatF, we first delineated the minimal D. melanogaster and D. erecta CREs capable of recapitulating sex-specific expression. In D. melanogaster, we identified a 271-bp element (mel-oe2; oe is an abbreviation for oenocyte element), upstream of desatF, which drives reporter activity in a pattern identical to the larger element ( Figure S1). Further subdivisions led us to isolate a smaller CRE (mel-oe1), which also confers full reporter activity ( Figures 4J and   4K). In D. erecta, a 710-bp CRE upstream of desatF (ere-oe) was found to recapitulate endogenous expression in transgenic reporter assays in D. melanogaster adults ( Figures 4N, 4O, and S2).
To identify putative transcription factor binding sites within the desatF enhancer, we compared the sequence of mel-oe1 with ere-oe. This comparison revealed a single orthologous putative binding site in both elements for the sex-specific transcription factors encoded by the doublesex gene (dsx) [28]. Female-and male-specific isoforms of the DSX transcription factor specify sexual development of soma [29]. The Yolk protein coding genes (yp1 and yp2) and the bric-a-brac locus (bab) are the only known direct targets of DSX regulation [30][31][32]. In both cases, DSX proteins regulate target genes sex-specifically by modulating the function of tissuespecific activators; DSX-F enhances target gene expression in females, and DSX-M represses expression in males.
The presence of a putative DSX-binding site in the desatF CREs from D. melanogaster and D. erecta raised the possibility that desatF is a direct target of DSX regulation. This prediction was confirmed by a variety of genetic, biochemical, and transgenic reporter experiments. If DSX directly regulates desatF, then flies lacking DSX function should be altered in desatF expression. Indeed, in situ hybridization analyses indicated that females lacking DSX activity completely lacked the expression of desatF in adult oenocyte cells ( Figure S3). Furthermore, depletion of dsx transcripts in oenocytes using a UAS-dsx-RNAi transgene driven by an oenocyte-specific Gal4 transgene (OK72-Gal4; Figure 4A and 4B) caused the loss of desatF expression in adult oenocyte cells (compare controls in Figures 4C and 4E with Figure 4G).
We next tested whether the DSX protein bound to the putative DSX binding sites in the desatF CREs from D. melanogaster and D. erecta. Electrophoretic mobility shift assays (EMSAs) demonstrated that the DSX-DNA-binding domain (DSX-DBD) specifically and efficiently bound the wild-type site ( Figure 4I, lanes 1-5), and this binding was abolished when the DSX-binding site was mutated ( Figure 4I, lanes 6-10). Furthermore, mutations in the DSX binding site of an otherwise wild-type mel-oe1 ( Figures 4J and 4K) and ere-oe ( Figures 4N and 4O) caused a complete loss of reporter activity in vivo ( Figure 4L and 4M and 4P and 4Q, respectively). Taken together, these data demonstrate that DSX-F directly activates desatF expression in adult oenocyte tissue. Note that while DSX-F is directly required for female-specific expression of desatF, it is not sufficient, and additional cis-regulatory inputs are also necessary for gene expression (see ''cis-Regulatory Sites in desatF Were Gained by a Series of Small Deletions during D. melanogaster Evolution'' below).
Our experiments did not indicate a repressive function for DSX-M in regulating desatF expression. The loss of DSX function in males did not lead to an upregulation of desatF in oenocyte cells (compare controls in Figure 4D and 4F with 4H), and mutations in the DSX binding site of mel-oe1 and ere-oe did not lead to a gain of reporter expression in males ( Figure 4M and 4Q). Thus, while DSX-F is required directly to activate desatF expression in females (Figure 4), unlike other known targets of DSX proteins [30][31][32], DSX-M apparently does not regulate desatF in males. Our studies thus demonstrate an additional mode of DSX target gene regulation.
Our phylogenetic analysis of desatF expression in species within the Sophophora subgenus led us to infer that sexually dimorphic expression arose in the ancestor of the D. melanogaster species subgroup ( Figure 2, left column, green arrowhead). Given that DSX-F directly activates desatF expression in female oenocytes of D. melanogaster and D. erecta, we posited that the origin of femalespecific expression arose with the DSX-binding site. In order to test this hypothesis, we investigated the ancestry of this site.

The DSX-Binding Site at desatF Predates the D. melanogaster Species Subgroup
If the DSX-binding site at desatF evolved concomitantly with the origin of dimorphic expression, then all outgroup species to the D. melanogaster species subgroup analyzed in our study, which display either monomorphic expression, or no expression of desatF, should lack an orthologous DSX-binding site in the oenocyte CRE of desatF. While we did not find an orthologous DSX-binding site in the outgroup species D. pseudoobscura, D. persimilis, or D serrata, which display monomorphic expression of desatF, we were surprised to find several outgroup species that contained an orthologous sequence similar to the DSX-binding site consensus [28] within the upstream regulatory region of desatF ( Figure 5A). In D. prostipennis, D. paralutea, and D. eugracilis, all of which lack desatF expression in adult oenocytes, and in D. takahashii, which expresses desatF in both sexes, there is an orthologous sequence that matches the D. melanogaster DSX-binding site for at least ten out of 13 base pairs ( Figure 5A, right panel). In D. paralutea, the orthologous site matches the D. melanogaster site perfectly. These data indicate that  the origin of the DSX-binding site most likely predated the origin of the D. melanogaster species subgroup ( Figure 5A, red). If true, then the origin of the DSX-binding site would be deeper in the phylogeny ( Figure 5A, green arrowhead) than the origin of dimorphic expression originally inferred from our phylogenetic expression analyses ( Figure 5A, black arrowhead), and those from others [23].
Monomorphic Expression of desatF in D. takahashii Evolved by Functional Inactivation of the DSX-Binding Site One interpretation of the observations above is that dimorphic expression of desatF actually arose with the acquisition of the DSXbinding site in the desatF CRE and was subsequently lost in the D. takahashii subgroup and in D. eugracilis. If this were true, then the state of desatF expression in D. takahashii, D. prostipennis, D. paralutea, and D. eugracilis evolved from an ancestor that expressed desatF female-specifically by direct DSX regulation. Moreover, monomorphic expression of desatF in D. takahashii would then be predicted to have evolved at least in part by loss of direct DSX regulation.
In order to test this possibility, we examined the D. takahashii desatF CRE and putative DSX-binding site in greater detail. We observed that the putative DSX-binding site present in the desatF CRE of D. takahashii diverges from the consensus sequence in two positions, one of which is located in the core of the site, where an A is found instead of the consensus C ( Figure 5A). We note that this difference is likely to be derived in D. takahashii as all other species with the orthologous site contain a consensus C at this position ( Figure 5A). Given that this change is located in the core of the sequence, it suggested to us that DSX proteins might not bind the D. takahashii site. EMSAs revealed that, in fact, the DSX-DBD failed to bind the orthologous site in D. takahashii desatF ( Figure 5C, lanes 1-5). Furthermore, introducing the identical C-to-A mutation in the core of the D. melanogaster DSX-binding site greatly diminished binding of the DSX-DBD relative to wild type ( Figure 5B, compare lanes 1-5 with lanes 6-10), and when introduced in an otherwise wild-type mel-oe1 ( Figures 5D and 5E), this C-to-A mutation caused a complete loss of reporter activity ( Figure 5F). These results show that the putative D. takahashii DSXbinding site is nonfunctional.
Our data are consistent with an evolutionary scenario in which monomorphic expression of desatF in D. takahashii evolved from a dimorphic ancestor by a loss-of-function mutation in the ancestral DSX-binding site. We tested this scenario by assessing whether the restoration of a functional DSX-site by an A-to-C transition in the D. takahashii oenocyte CRE would result in dimorphic reporter expression. A 296-bp CRE upstream of desatF from D. takahashii is fully sufficient to recapitulate monomorphic expression of desatF in transgenic reporter assays in D. melanogaster ( Figure 5H and 5I). EMSAs confirmed that an A-to-C mutation in the core of the D. takahashii DSX-like binding site, which converts it to the consensus sequence, is sufficient to restore binding by the DSX-DBD ( Figure 5C, compare lanes 1-5 with lanes 6-10). Remarkably, this change was also sufficient to increase reporter expression in females and to decrease reporter expression in males relative to wild-type constructs ( Figure 5J and 5K). This modified CRE is thus functionally dimorphic in contrast to the wild-type functionally monomorphic construct (compare Figure 5H and 5I with 5J and 5K). These results are consistent with the monomorphic expression of desatF in D. takahashii having evolved, at least in part, by the inactivation of the DSX-binding site that was present in an ancestor that expressed desatF female-specifically. Furthermore, this brings us to a total of two transitions in the sex-specificity of desatF expression (summarized in Figure 7, pink bars): a gain of dimorphism and a subsequent transition to monomorphism.

Female-Specific Expression of desatF Is Consistent with Stabilizing Selection
In the course of our studies of desatF regulation, we were surprised to discover that while D. melanogaster, D. sechellia, and D. erecta express desatF similarly in female oenocytes (Figure 2, right column), their respective oenocyte CREs were significantly different in structure. We found that sequences orthologous to the mel-oe2 CRE from D. erecta (i.e., ere-oe2; Figure S2) and D. sechellia (i.e., sec-oe2; Figure 3, compare 3B with 3D) failed to drive expression in transgenic reporter assays in D. melanogaster. These results suggested that female-specific desatF expression in D. erecta and D. sechellia rely at least in part on different cis-regulatory sites than those characterized in D. melanogaster.
Indeed, for D. erecta, additional sequences outside the ere-oe2 region are required for reporter activity. By extending the 59-end of ere-oe2 by 190 bp (ere-oe3; Figure S2), we obtained full reporter activity in D. melanogaster female oenocytes. Importantly, this 190bp region, by itself, is not sufficient for reporter function. Furthermore, the orthologous region from D. melanogaster is clearly not required for CRE function, as mel-oe2 is a fully functional CRE despite lacking the orthologous 190-bp region ( Figure S1). Thus, the functional D. melanogaster and D. erecta CREs share common necessary features (e.g., the DSX-binding site), but also exhibit critical differences (e.g., the 190-bp region required for D. erecta CRE activity).
For D. sechellia, an exhaustive search of all intergenic sequences upstream and downstream of desatF failed to identify a region that drove reporter expression in the D. melanogaster genetic background (unpublished data). This may indicate that the D. sechellia CRE for oenocyte expression is located outside of the regions searched and/or that there are trans-acting regulatory differences between the species, which are key for desatF expression in D. sechellia females. We note there is a putative DSX-binding site in the D. sechellia desatF CRE that matches the consensus site described for D. melanogaster [28]. Thus the absence of reporter expression from the D. sechellia desatF CRE in the D. melanogaster molecular background is not likely to be due to a failure of DSX to bind the site. Together, our data indicate that D. melanogaster, D. sechellia, and D. erecta express desatF female-specifically in part by distinct cisregulatory mechanisms.
One explanation for these results is that stabilizing selection has maintained phenotypic constancy for desatF expression while mutational turnover of functionally important sites has taken place. This phenomenon has been previously reported [33][34][35][36][37][38], but described for species that have diverged over relatively long periods of evolutionary time (40 millions y or more). It is surprising that drastic alterations in the cis-regulatory mechanisms at desatF occurred in a short period of time (D. melanogaster and D. sechellia diverged only 2-3 million y ago). More detailed investigation of the D. melanogaster CRE uncovered cis-regulatory sites specific to this species and important for desatF female-specific expression.

cis-Regulatory Sites in desatF Were Gained by a Series of Small Deletions during D. melanogaster Evolution
While investigating the cis-regulatory differences at desatF between the three dimorphic species, we uncovered an unusual feature specific to D. melanogaster. We found eight copies of the hexamer AATTTG in its upstream regulatory region ( Figure 6A), i.e., within mel-oe1 and -oe2, which was overrepresented with high statistical significance (occurrence p = 0.0000051; from the Oligo- analysis program [39]). The motif does not match any binding site consensus to our knowledge. In order to test whether these motifs were functionally relevant, we introduced point mutations in six of these motifs in an otherwise wild-type mel-oe2 CRE. This led to a complete loss of reporter activity in female oenocytes (unpublished data), indicating that these motifs are indeed necessary for CRE function.
Most of these hexamer motifs were absent in the orthologous region from D. erecta, D. sechellia ( Figure 6A), and all other species examined in our study (unpublished data), suggesting new cisregulatory sites evolved recently in the D. melanogaster lineage. The evolution of these hexamer motifs could have occurred through a variety of mechanisms [40]. Rearrangement events such as transposition and duplication, and binding site formation by point mutation, are the two main modes by which new cis-regulatory content has been suggested to evolve [41][42][43]. In order to understand the mutational path that produced these hexamer motifs at desatF in D. melanogaster, we compared the mel-oe2 sequence to its ortholog in closely related species, D. simulans and D. erecta. Of particular interest is the cluster of three motifs in the forward direction in mel-oe2 ( Figure 6A). A sequence alignment of mel-oe2 and its ortholog from D. erecta and D. simulans revealed that, except for the hexamers, this region is largely conserved, excluding a transposition event ( Figure 6D; see below). Closer scrutiny of the alignment in the hexamer region revealed in D. simulans and D. erecta the presence of common insertions/deletions disrupting each of the three hexamer motifs in the cluster ( Figure 6D). On the basis of the phylogenetic relationships among these three species, we infer that the three hexamer motifs were gained during D. melanogaster evolution by a series of nonidentical small deletions.
We tested if these three particular motifs were required for CRE function by introducing point mutations in the mel-oe1 CRE. We found that they caused a complete loss of reporter activity ( Figure 6B and 6C). To test whether the AT content of these motifs, instead of their sequence, could explain their functional relevance, we mutated the three sites in mel-oe1 without altering their AT percentage. This construct also failed to produce reporter activity ( Figure S4). We suggest that these hexamer sequences are binding sites for a transcription factor and that they evolved via a series of small deletions.

Discussion
Pheromone differences between closely related species of Lepidoptera and Diptera suggest that their production is rapidly evolving [6,44,45]. While rapidly evolving traits have been characterized many times, especially regarding sexually related traits [46,47], few studies, to our knowledge, have identified the genes and mutations that give rise to these evolutionary changes [32]. We have investigated the evolution of the mechanisms that govern the production of pheromonal signals between males and females during Drosophila courtship. Our studies provide several insights into the molecular mechanisms of pheromone signal evolution.
We have shown that the desatF gene is rapidly evolving in the subgenus Sophophora. We found that changes in the desatF expression have evolved numerous independent times within 40 million y of Drosophila evolution, including six independent instances of gene loss, two modifications in sex-specific expression, and three independent losses of expression without gene loss. Altogether, among 24 species surveyed, we observed 11 transitions in the state of desatF expression (summarized in Figure 7). desatF has also been duplicated in some Drosophila species [23]. Our results reveal that the evolution of the desatF gene is extraordinarily dynamic, and displays, to our knowledge, the fastest evolving pattern of gene utilization observed to date.
We identified the CRE regulating desatF expression in D. melanogaster oenocytes and characterized DSX-F as a necessary and direct input for its female-specific expression. We found that transitions from sexual monomorphism to dimorphism, and the reverse, rely at least in part on the gain and loss of direct DSX regulation. Remarkably, simple evolutionary changes in cisregulatory sequences were sufficient to explain the transition of desatF expression from dimorphism to monomorphism, as seen in D. takahashii.
Because desatF and dienes contribute strongly to D. melanogaster mating behaviors, it is likely that dimorphic expression of desatF is under sexual selection. Since the gain of female-specific expression, we count five losses of sexual dimorphism, which includes one transition to monomorphism and four transitions to amorphism (Figure 7, black dots). What does this pattern of evolutionary change suggest? The loss of sexually selected traits is widespread [48]. This pattern of frequent trait loss may be an indication of relaxed selection resulting from rapidly shifting regimes of sexual selection. For example, since fly courtship is regulated by multiple sensory cues (visual, chemosensory, auditory, etc), the modality that is under sexual selection may change, leading to trait loss.

Regulatory Evolution as a Mechanism to Evolve Sex-Specificity
A long-standing question in evolutionary biology is how sexually dimorphic traits evolve [49]. For example, monomorphic patterns can evolve from dimorphic patterns and vice versa, however, the molecular mechanisms that govern these transitions have seldom been addressed. In Lepidoptera and Diptera, duplication or structural changes of genes encoding desaturases have been suggested to [23] or shown to contribute to evolutionary alterations in pheromone signals [50][51][52], however, none of these phenomena alone could account for evolutionary transitions in sex-specificity of pheromone production. Here, we have provided evidence that cis-regulatory sequence evolution led to transitions from monomorphic to dimorphic expression of desatF, and its reversion, and concomitant changes in diene production. the ability of the DSX-DBD protein to bind annealed radiolabeled oligonucleotide probes containing the D. melanogaster DSX-binding site (lanes 1-5) and a mutated version of this site (lanes 6-10) containing a C to A point mutation (in red), as found in D. takahashii. This mutation greatly reduced binding of the DSX-DBD. (C) EMSAs comparing the ability of the DSX-DBD protein to bind annealed radiolabeled oligonucleotide probes containing the D. takahashii putative DSX-binding site and a mutated version of this site containing a C in the core of the putative site instead of an A, which is found in the consensus and in the D. melanogaster site. In contrast to the wild-type D. takahashii site (lanes 1-5), where no significant binding is observed, the DSX-DBD protein binds the mutant site relatively efficiently (lanes 6-10). The arrow and arrowhead point to a single and pair of DSX-DBD monomers bound to the probe. The asterisk marks the position of the free probe. (D-K) eGFP reporter expression in abdomens of 4-d-old D. melanogaster females (D, F, H, and J) and males (E, G, I, and K) and carrying two copies of the transgenes indicated at the top of the columns. Introducing a C to A point mutation in the DSX-binding site of mel-oe1 abolishes eGFP reporter expression in females (compare D and F), while leaving the absence of expression in males unchanged (compare E and G). Introducing an A to C point mutation in the putative DSX-binding site of tak-oe produces sexually dimorphic eGFP expression (J, K), whereas a wild-type tak-oe drives monomorphic expression (H, I). By pinpointing one of these transitions at the level of individual base-pairs, we propose that monomorphic expression of desatF in D. takahashii evolved from a dimorphic ancestor through a derived mutation in a single critical residue inactivating the orthologous DSX-binding site (Figure 4). A simple model for the origin of monomorphic gene expression is that a mutation in the DSX-binding site abrogated repression by DSX-M, in turn, upregulating desatF expression in males. Furthermore, the loss of regulation by DSX-F would lead to a decrease in desatF expression in females. These alterations, together, would produce monomorphic expression of desatF in D. takahashii. However, we note that this model is at odds with our finding that DSX-M appears to not regulate desatF in D. melanogaster. This suggests that the ancestor of D. melanogaster and D. takahashii regulated desatF dimorphically by either a D. melanogaster-like mechanism, or by a mode that involved repression by DSX-M. While we currently cannot polarize these possibilities, both models implicate the inactivation of the ancestral DSX-binding site as a necessary step in the transition to monomorphic expression.

desatF and Speciation
In order to understand the mechanisms that drive speciation, the genetic changes that lead to reproductive isolation must be elucidated. It has been suggested that, ''speciation genes are those that contribute to reproductive isolation, often in the form of hybrid inviability, sterility or behavioral aberration'' [53]. While progress has been made in identifying genes that contribute to postzygotic isolation (such as Xmrk2 [54,55], OdsH [53], Nup96 [56], see review [57]), little is known of genes that contribute to prezygotic isolation. We suggest that desatF could be one such gene.
There is evidence that diene production contributes to reproductive isolation. For example, it has been documented that dienes inhibit D. simulans male courtship behavior [58]. Moreover, D. simulans/D. melanogaster hybrid females lacking desatF expression elicit greater levels of courtship activity from D. simulans males, relative to hybrids expressing desatF female-specifically [59]. Taken together, these data indicate that expression of desatF and the production of dienes in D. melanogaster females contribute to the reproductive isolation between these sibling species.
desatF expression has evolved numerous times during Drosophila evolution. If, as others have suggested [12], transitions in dienes contribute to sexual behavior in species other than D. melanogaster, then the contribution of desatF to speciation may be widespread.

Fly Stocks
Wild-type stocks were obtained from the University of California, San Diego (UCSD) stock center (see Table S1). Gal4-UAS analyses were performed using the following lines: OK72-Gal4 was obtained from the Bloomington Stock Center; dsx 1 p p , UAS-dsxRNAi, and UAS-lacZ were provided by M. McKeown (Brown University).

Imaging of Fly Abdomens
Images of in situ hybridizations and X-Gal stained adult abdomens were taken using an Olympus SZX16 Stereo Microscope equipped with an Olympus DP71 microscope digital camera. Adult transgenic eGFP-reporter line samples were imaged using an Olympus Fluoview FV 1000 confocal microscope and software. Wings and head were removed from 4-d-old adults, which were then mounted in Halocarbon 700 oil for confocal analysis.

Sequence Analysis of Orthologous desatF Loci
Sequences for D. melanogaster, D. simulans, D. sechellia, D. yakuba, D. erecta, D. ananassae, and D. pseudoobscura were obtained from their respective genome databases. All other sequences were obtained by cloning and sequencing of orthologous sequences using genomic DNA prepared from species stocks obtained from the UCSD Drosophila stock center (see Table S1). Sequences were PCR amplified using different sets of degenerate primers and then fused to give rise to the sequence of the whole locus. Details are available upon request to the authors. Novel sequences have been deposited in GenBank (http://www.ncbi.nlm.nih.gov/Genbank, submission numbers are listed in Table S1). Orthologous sequences were aligned using ClustalW2 [60] with subsequent manual alignment in problematic regions. We used the GenePalette program to analyze our sequences (www.genepalette.org). We used Oligoanalysis to look for overrepresented motifs in our sequence [39]. This program calculates the probability that the analyzed sequence contains an oligonucleotide sequence at a frequency greater than that expected at random.

In Situ Hybridization on Adult Abdomens
In situ hybridization was performed as previously described [61] with minor modifications. The complete adult abdominal in situ protocol is available at http://www.molbio.wisc.edu/carroll/. Primers used to amplify probes are listed in Table S2.

DNA-Binding Analyses
EMSAs were performed as previously described [27,32]. PAGE-purified oligos used in EMSAs are listed in Table S3.

Transgenic Fly Production
All transgenic lines were produced by using the Phage QC31 Integrase system. Embryos from flies containing the X-chromosome attP docking site VK00046 [62] were injected as previously described [32]. Primers used to clone the constructs are listed in Table S4. Figure S1 Screen of the D. melanogaster desatF locus for CREs. Numbers and adjacent bars indicate the desatF locus region surveyed in a given reporter construct. Green bars represent the regions that drove eGFP reporter activity in oenocytes in 4-d-old D. melanogaster flies, black bars represent regions that did not drive activity. Scale in base pairs is boxed. Additional information on the numbered eGFP-reporter gene constructs is in Table S4. Found at: doi:10.1371/journal.pbio.1000168.s001 (0.78 MB EPS) Figure S2 Screen of the D. erecta desatF locus for active CREs. Numbers and adjacent bars indicate the desatF locus region surveyed in a given reporter construct. Green bars represent the regions that drove eGFP reporter activity in oenocytes in 4-d-old D. melanogaster flies, black bars represent regions that did not show activity. Scale is boxed. Additional information on the numbered eGFP-reporter gene constructs can be found in Table S4.  Table S2 Primers used to amplify probes for in situ hybridizations. The close proximity of some species allowed cross-hybridization. The D. melanogaster probe was hence also used on D. mauritiana, D. simulans, and D. sechelllia. The D. santomea probe was also used on D. teissieri. The D. pseudoobscura probe was also used on D. persimilis.