Evolution of sexually dimorphic pheromone profiles coincides with increased number of male‐specific chemosensory organs in Drosophila prolongata

Abstract Binary communication systems that involve sex‐specific signaling and sex‐specific signal perception play a key role in sexual selection and in the evolution of sexually dimorphic traits. The driving forces and genetic changes underlying such traits can be investigated in systems where sex‐specific signaling and perception have emerged recently and show evidence of potential coevolution. A promising model is found in Drosophila prolongata, which exhibits a species‐specific increase in the number of male chemosensory bristles. We show that this transition coincides with recent evolutionary changes in cuticular hydrocarbon (CHC) profiles. Long‐chain CHCs that are sexually monomorphic in the closest relatives of D. prolongata (D. rhopaloa, D. carrolli, D. kurseongensis, and D. fuyamai) are strongly male‐biased in this species. We also identify an intraspecific female‐limited polymorphism, where some females have male‐like CHC profiles. Both the origin of sexually dimorphic CHC profiles and the female‐limited polymorphism in D. prolongata involve changes in the relative amounts of three mono‐alkene homologs, 9‐tricosene, 9‐pentacosene, and 9‐heptacosene, all of which share a common biosynthetic origin and point to a potentially simple genetic change underlying these traits. Our results suggest that pheromone synthesis may have coevolved with chemosensory perception and open the way for reconstructing the origin of sexual dimorphism in this communication system.

a genetic correlation between the emitter and receiver components, leading to a phenomenon of reciprocal and self-reinforcing selection between signaling trait(s) and signal preference(s). These mechanisms often work in concert, and as a result, correlated evolutionary changes are generally expected between the expression and perception of secondary sexual traits (Andersson, 1994). For example, in three-spined stickleback, males with the brightest red breeding coloration are preferred by females (Milinski & Bakker, 1990), whose visual sensitivity in the red spectrum increases during the reproductive season (Cronly-Dillon & Sharma, 1968).
Insects rely on chemical communication to locate, identify, and select mates. Chemical cues and their corresponding receptors offer an excellent opportunity to study the evolution of sexual dimorphism (Steiger & Stökl, 2014). Insect pheromones, which include both volatile and contact (cuticular) hydrocarbons (CHCs), are critical for courtship and mating behavior (Blomquist & Bagnères, 2010;Stanley & Nelson, 1993). Sexually dimorphic pheromones have been discovered across many insect taxa (Howard & Blomquist, 2005).
Similar to the pattern often observed in visual communication systems, sex-specific pheromones have coevolved with their cognate receptors. Many examples of correlated evolutionary changes between CHC production and perception have been documented, with evolutionary gains and losses of sex-specific CHCs mirroring evolutionary gains and losses of sensory response to these CHCs (Choe & Crespi, 1997;Dekker et al., 2015;Ng et al., 2014;Sappington & Taylor, 1990).
The best models for understanding evolutionary innovations in male-female communication are those where sexual dimorphism in both signaling and perception has evolved recently from an ancestral monomorphic state. However, among the closest relatives of D. melanogaster (the melanogaster species subgroup), sexually dimorphic expression of 7,11-HD appears to be ancestral, with most F I G U R E 1 Evolution of a sexually dimorphic chemosensory system in Drosophila prolongata. (a) Phylogenetic relationships among the species used in this study (Based on Barmina & Kopp, 2007). Drosophila elegans is included as an outgroup to the rhopaloa species subgroup, which encompasses the remaining species. Species for which leg images are shown are indicated in black, the rest are in light gray. Red bar indicates the inferred timing of male-specific expansion of foreleg gustatory organs. (b-d) Proximal foreleg tarsi of D. elegans (b), D. carrolli (c), and D. prolongata (d), with females on the left and males on the right. The two types of sensory organs can be distinguished by external morphology: mechanosensory bristles are straight, pointed, and bear triangular bracts at the base, whereas gustatory bristles are curved, bractless, and have rounded tips. Note the dramatic increase in the number of gustatory organs in D. prolongata males evolutionary changes due to secondary losses of sex-biased dienes (Jallon & David, 1987;Shirangi, Dufour, Williams, & Carroll, 2009).
Drosophila prolongata, a member of the melanogaster species group Toda, 1991), is a promising model to study both the selective forces and the genetic mechanisms behind sexual dimorphism. This species displays several striking sexually dimorphic traits, including the exaggerated male forelegs that distinguish D. prolongata from closely related species  and are essential for courtship and mating success (Setoguchi et al., 2014). This sex-specific increase in leg size has been accompanied by an equally recent male-specific expansion of the chemosensory system ( Figure 1, see Section 3). Because foreleg chemoreceptors are directly involved in pheromone perception (Fan et al., 2013;Inoshita, Martin, Marion-Poll, & Ferveur, 2011;Stocker, 1994), we hypothesized that pheromone profiles may also display a strong degree of sexual dimorphism in D. prolongata. To test this hypothesis, we characterized the male and female CHC profiles of D. prolongata and its relatives. We found that, indeed, the CHC profiles of D. prolongata are highly distinct from all other species, due primarily to a strong male-specific increase in the amount of long-chain CHCs. We also observe an intriguing female-limited CHC polymorphism in this species, where some females have male-like CHC profiles. These results set the stage for investigating potential coevolution between sex-specific pheromone production and perception on a recent evolutionary time scale.

| Specimen preparation and microscopic imaging
Prothoracic legs were dissected from adult female and male flies under CO 2 anesthesia and mounted in Hoyer's media between two coverslips. After overnight clearing, the legs were imaged under bright field illumination with a 20× lens on a Leica DM500B microscope with a Leica DC500 camera. Stacks of images were merged into single extended depth-of-field images and processed further using Adobe Photoshop. and stored at −20°C. Prior to analysis, samples were resolubilized by adding 15 μl hexane with two spiked-in alkanes: n-hexacosane (Sigma-Aldrich) and n-triacotane (Sigma-Aldrich), 10 ng/μl each.

| Cuticular hydrocarbon (CHC) extraction
These compounds were absent in both sexes across all studied species and thus were selected as external standards, referred to as ES-1 and ES-2 hereafter.

| Gas chromatography (GC) and mass spectrometry (MS) analyses
The cuticular hydrocarbon extracts were analyzed on an Agilent 7890B GC fitted with a 30 m × 0.25 mm × 0.25 μm HP-5 Ultra Inert column and coupled to an Agilent 5977A mass spectrometer (Agilent Technologies). One microliter sample was introduced to the injection port using an Agilent 7683B autosampler in split-less mode. The oven temperature was programmed as follows: ramped from 160 to 280°C at a rate of 2.5°C/min, held at 280°C for 1 min, and increased to 315°C at 15°C/min followed by 1 min final hold. The injector and transfer line temperature were kept constant at 275 and 280°C, respectively. Helium was used as the carrier gas with a constant flow rate at 2 ml/min. in the NIST 05 reference library and those in previous Drosophila pheromone publications (Dembeck et al., 2015;Everaerts, Farine, Cobb, & Ferveur, 2010;Howard, Jackson, Banse, & Blows, 2003).
DMDS-derivatization reactions were not conducted to confirm the position of the double bonds.
Representative CHC profiles in both sexes of all studied species are presented in Figure 2 and Figure S1 (also see nomenclature in Table 1). For quantification, individual chromatographic peaks were first called using the built-in ChemStation integrator with initial peak width 0.045 and initial threshold 16, followed by manual adjustment to include minor peaks and deconvolute overlapping peaks.
Consensus peaks were first constructed within groups defined by species and sex, by aligning orthologous peaks among biological replicates using retention time. The final consensus was obtained by merging group consensuses based on inferred chemical identities and/or Kovats indices (KI; Carlson, Bernier, & Sutton, 1998).
Cuticular hydrocarbons were initially quantified by measuring individual peak areas and then scaled by external standards to obtain absolute amounts (in nanograms, summarized in  PCA was also performed on log transformed data ( Figure S3A) and on the data that included cVA ( Figure S3B). In addition to PCA, nonmetric multidimensional scaling (nMDS) analysis was conducted as an alternative noneigenvector approach to examine the spatial organization pattern among individuals ( Figure S3C). Two-dimensional plot and stress value were obtained using the "metaMDS" function in the "vegan" package (Oksanen et al., 2019), with Bray-Curtis dissimilarity as input and 25 random starting configurations. All three approaches yielded qualitatively similar results ( Figure S3).

| Statistical analysis
Following 2D ordination by PCA, clustering was performed on pairwise Euclidean distances of CHC composition between individuals to further characterize the spatial heterogeneity. Using the function "hopkins" in the "clustertend" package (Luo & Zeng, 2015), the clustering tendency was first examined by the Hopkins statistics (ranging from 0 to 0.5), with smaller values indicating the presence of spatial patchiness. The clustering tendency was cross-validated by a visual approach ( Figure S5A), implemented in the function "fviz_dist" from the "factoextra" package (Kassambara & Mundt, 2017). The optimal number of clusters was adopted using the majority rule, after comparing 30 results (Table S3) obtained by the function "fviz_nbclust" and "NbClust" in the package in the "factoextra" (Kassambara & Mundt, 2017) and "NbClust" package (Niknafs, Ghazzali, Boiteau, & Niknafs, 2015), respectively. Partition clustering by K-means algorithm was performed on the CHC proportions, using the function "eclust" in the "factoextra" package (Kassambara & Mundt, 2017) with 50 random starting configurations. The performance of clustering was examined by silhouette width (Figure S5B), which captured both the within-cluster compactness and betweencluster separation. The silhouette indices ranged from −1 to 1, with increasing reliability of cluster assignment.
F I G U R E 2 Sexually dimorphic CHC profiles in Drosophila prolongata. The graphs show representative GC-MS chromatograms of cuticular CHCs of a single 7-day-old virgin male and female from the Sapa strain, with the female at the top (red) and male in mirror image at the bottom (blue). Compounds corresponding to each numbered peak are listed in Table 1. Compounds that are shared between sexes bear the same number. Arrows indicate peaks that were not perfectly resolved, with the minor component shown by dashed lines. Unit-less abundances are direct measurements from the mass selective detector. ES-1 (n-hexacosane) and ES-2 (n-triacotane) are external standards used to calculate the absolute amounts of each compound in females (1,953 ± 164 ng; n = 24) and males (3,348 ± 263 ng; n = 22). Sexually dimorphic compounds include the female-biased n-heneicosane (nC21), 9-tricosene (9T), and 7-tricosene (7T) and the male-biased 9-pentacosene (9P) and 9-heptacosene (9H) To identify the candidate CHCs that explained most of the between-sample variation, variation retained by the principal components was partitioned using the "fviz_contrib" function in the "factoextra" package (Kassambara & Mundt, 2017). Briefly, the contribution of individual CHC to a given principal component was calculated as proportion of squared loading coefficients to the sum of squares. Using the function "fviz_pca_var" in the "factoextra" package (Kassambara & Mundt, 2017), variable map was constructed to visualize the correlation between candidate CHCs and (a) principal components, (b) discrete chemical clusters, (c) species, and (d) sex within a species. Metric-based plots were produced using the "gg-plot2" package (Wickham et al., 2018), and all statistical analyses were conducted in R studio (R Core Team, 2018).

| The chemosensory system of D. prolongata shows a recent increase in sexual dimorphism
In  Linear alkane 408 2900 a Peak numbers corresponded to the elution order of each analyte and to the peaks shown in Figure 1 and Figure S1.

| The CHC profile of D. prolongata shows increased sexual dimorphism
We used GC-MS analysis to identify CHC compounds in male and female D. fuyamai, D. rhopaloa, D. carrolli, D. kurseongensis, and two strains of D. prolongata. Across all samples, we identified 27 distinct chromatogram peaks (Figure 2 and Figure S1) and determined the chemical identities of their corresponding compounds (Table 1). All these compounds have been identified previously in D. melanogaster.
With the exception of the well-known male-specific cis-vaccenyl acetate (cVA; Butterworth, 1969), we did not detect any qualitative sexual dimorphism, that is, the remaining 26 peaks were observed in both males and females (Table S1). These CHCs ranged from 21 to 29 carbons and fell into three chemical classes-linear alkanes, methyl-branched alkanes, and mono-unsaturated alkenes (monoenes).
Monoenes with the same carbon number could be further divided into three positional isomers based on the location of the double bond (Table 1).
We quantified the relative abundance of each of 26 identified CHCs as a proportion of the total CHC blend of an individual. The compositional representation of CHCs was highly uneven for all species ( within which monoenes were the major constituents across all groups. Notably, for monoenes with the same carbon number, we discovered that the 9 isomers, especially 9-tricosene and 9-pentacosene (23 C and 25 C, respectively), were the major structural isomers in all species except in D. fuyamai. In

| Sexual dimorphism is variable in D. prolongata
In contrast to other species and sexes, the CHC profiles of D. prolongata Sapa females were highly variable, spanning the area from the top-left to the top-right cluster (Figure 3a). Drosophila prolongata Sapa males clustered together with the D. prolongata Bavi males. This pattern indicated that some Sapa females were similar to D. prolongata Bavi females, some others were similar to conspecific males, and many had CHC profiles that were intermediate between Bavi males and females. This polymorphism will be discussed below.

| Dimorphic CHC profiles are caused by sexbiased proportions of long and short monoenes
In D. prolongata, CHC composition is sexually dimorphic due to the unique CHC profile of males. To identify specific CHCs driving this difference, we partitioned the variance explained by the first 2 PCs.
Five individual CHCs-n-heneicosane (nC21), 9-tricosene (9T), 7tricosene (7T), 9-pentacosene (9P), and 9-heptacosene (9H)-had higher than expected contributions to this variance (Figure 3b). To infer the relationship between the abundance of these CHCs and the clustering of individuals by species and sex on PCA plots (Figure 3a), we mapped the candidate CHCs to the coordinates defined by the first two PCs (Figure 3c). 7T was highly and negatively correlated with PC2, suggesting it was the major discriminator distinguishing D. fuyamai from the other species, where 9-monoene isomers prevailed. Indeed, we found that 7T had much higher abundance in D. fuyamai than in the other species ( Figure S4). Remarkably, PC1 was negatively correlated with short-chain CHCs (9T, 7T, and nC21), To further assess whether sexual dimorphism within species and divergence between species could be attributed to specific short and long monoenes, we compared the relative proportions of the three candidates that constitute a homolog series (9T, 9P, and 9H), out of the total CHC blend, across all groups ( Figure 4a). As expected, relative levels of these candidate CHCs were roughly equal between sexes in the three chemically monomorphic species: D. fuyamai, D. rhopaloa, and D. carrolli. In chemically dimorphic species, however, the relative proportions of 9T, 9P, and 9H were clearly different between sexes. In D. prolongata, especially in the Bavi strain, the longer-chain 9P was much more abundant in males than in females, while the shorter-chain 9T showed the opposite pattern (Figure 4a).
This dimorphism was reversed in D. kurseongensis, where males had a higher proportion of 9T and females a higher proportion of 9P. carbon backbone (Figure 3d). This structural similarity points to a single biosynthetic pathway that could explain both the sexual dimorphism and the intraspecific variation in D. prolongata, as we discuss below.

| D ISCUSS I ON
Sexual selection theories have extensively discussed how interactions between signals and preferences can drive the evolution of sexual dimorphism (Andersson, 1994). These theories can best be tested in models that show recent and simultaneous emergence of sexual dimorphism in signal production and perception. In this respect, D. prolongata has the potential to be an excellent model. Its exaggerated sexual dimorphism in foreleg morphology is clearly of very recent origin and contributes directly to courtship behavior and mating success (Setoguchi, Kudo, Takanashi, Ishikawa, & Matsuo, 2015;Setoguchi et al., 2014). This evolutionary change is accompanied by a greatly increased number of chemosensory bristles exclusively in males and, as we show here, by an equally recent change in the male-specific CHC profile, suggesting potential coevolution between pheromone production and perception.
Reconstructing the origin of sexual dimorphism requires a mechanistic understanding of the pathways that produce the dimorphic traits. The biosynthetic pathway responsible for CHC synthesis in Drosophila is the ultimate source of all sex-specific and sexually monomorphic compounds (Howard & Blomquist, 2005). In most populations of D. melanogaster, sexually dimorphic expression of elongase and desaturase enzymes (eloF and desatF) leads to sex-biased abundance of several CHCs including the female-enriched 7,11-HD and the male-enriched 7-Tricosene (7T; Antony & Jallon, 1982;Chertemps et al., 2007;Chertemps et al., 2006). Interestingly, we find that 7T is enriched in D. fuyamai, compared with other species, but this enrichment is sexually monomorphic (Figure 3a,c and Figure   S4). Female-enriched alkadienes are also found in most relatives of D. melanogaster (Jallon & David, 1987), which show female-biased expression of desatF (Shirangi et al., 2009) and eloF (Combs et al., 2018). This sexual dimorphism has an important functional role, for example, evolutionary changes in eloF expression contribute to strong behavioral isolation between the sibling species D. simulans and D. sechellia (Combs et al., 2018). However, sexually dimorphic alkadiene production in D. melanogaster and its relatives appears to have evolved at the base of the melanogaster species subgroup, with most species differences caused by secondary losses (Shirangi et al., 2009). In contrast, the male-biased expression of long-chain monoenes 9P and 9H in D. prolongata, at the expense of the shorter female-biased 9T, has evolved more recently and provides a good alternative model for understanding the emergence of sexually dimorphic communication.
Based on the structure of the CHC synthesis pathway, we hypothesize that both the male-biased abundance of 9P and 9H and the female-biased abundance of 9T in D. prolongata could in principle be attributable to a simple genetic change. Namely, we hypothesize that male-biased expression of an elongase enzyme that catalyzes the conversion of 9-C24:1-CoA into 9-C26:1-CoA and 9-C28:1-CoA could explain both the higher abundance of 9P and 9H and the lower abundance of 9T in males. Conversely, lower expression of this enzyme in females would lead to higher abundance of 9T and lower abundance of 9P and 9H. Under this model, the key evolutionary change would be a transition from sexually monomorphic to malebiased carbon chain elongation in D. prolongata following its divergence from D. carrolli and D. rhopaloa. This change could be due to the gain of a new elongase gene, changes in the chemical activity of an existing enzyme, or, most simply, to a regulatory mutation that leads to male-biased expression of an elongase that was expressed monomorphically in the ancestral condition. The female-limited polymorphism in the relative proportions of 9P and 9T might also be explained by the same mechanism. Artyom Kopp https://orcid.org/0000-0001-5224-0741

DATA AVA I L A B I L I T Y S TAT E M E N T
All raw data will be accessible on Dryad with the accession https :// doi.org/10.25338/ B86K7V.