Abstract
Sexually dimorphic morphological traits are among the fastest evolving animal features. Similar sex-specific structures have sometimes evolved independently in multiple lineages, presumably as targets of parallel sexual selection. In such cases, comparing the cellular mechanisms that generate these structures in different species can elucidate the interplay between selection and developmental constraint in evolution. In Drosophilidae, male-specific tarsal brushes on the front legs are found in at least four separate lineages. In this study, we combine phylogenetic reconstruction with developmental analyses and behavioral observations to investigate the evolutionary origin of these structures. We show that the sex brush has evolved independently at least three times from sexually monomorphic ancestral morphology. However, all sex brushes have very similar fine structure and develop through indistinguishable cellular processes, providing a striking example of developmental convergence. In all examined species, males use their sex brushes to grasp the female abdomen prior to copulation. We discuss potential reasons why convergent evolution of novel structures is rare even in the face of similar functional demands.
Introduction
Most animals are sexually dimorphic. Perhaps the most fascinating feature of sexual dimorphism is the rapid evolutionary turnover of sex-specific traits. Even among close relatives, the characters that distinguish males from females vary greatly from species to species. This simple observation implies that new sexual characters are gained, and ancestral ones are often lost, during the evolution of many if not most animal lineages. Understanding the genetic and developmental basis of this turnover is necessary to shed light on one of the most important drivers of biological diversity. Examples where similar traits have evolved multiple times are of particular value, as they often provide insights into the general patterns and mechanisms of evolution (Kopp 2009).
Most higher Diptera mate with the male on top of the female, and the male front (T1) legs are often involved in grasping or stimulating the female (Huber et al., 2007; McAlpine, 1981). Perhaps for this reason, male-specific ornaments or grasping structures are found on the T1 legs of many dipteran species (Daugeron et al., 2011; Eberhard, 2001; Ingram et al., 2008; Sivinski, 1997). In Drosophilidae, the most obvious male-specific leg modifications include the sex combs found in the Drosophila melanogaster and obscura species groups and in the genus Lordiphosa (Katoh et al., 2018; Kopp, 2011); branched or spoon-shaped tarsi of some Hawaiian Drosophila species (Hardy, 1965; Stark and O’Grady, 2009); and tarsal brushes that are the focus of this study (Figure 1). Sex brushes are found in at least four separate groups within the Drosophilidae: the Drosophila immigrans species group, the loiciana species complex, D. repletoides, and the genus Zaprionus. Due to uncertain relationships among these lineages, the evolutionary origin(s) of this structure remain unclear.
In the immigrans group, male sex brushes are found in some but not all of the species; the most likely scenario is that the brush was present in the last common ancestor of this clade, but was secondarily lost in the nasuta subgroup and greatly reduced in several other species (Rice et al., 2018). In Zaprionus, male brushes are present in most African species, with the exception of Z. neglectus, Z. spineus, and Z. spinosus (Tsacas and Chassagnard, 1990; Yassin et al., 2008; Yassin and David, 2010). The Zaprionus phylogeny is not fully resolved, but the distant relationship between the first species and the last two suggests that their lack of brushes is likely to reflect independent secondary losses. The situation is more complicated among species assigned to the Oriental Anaprionus subgenus of Zaprionus. Many of its members, including Z. lineosus, Z. spinilineosus, Z. orissaensis, Z. multistriatus, Z. grandis, and Z. aungsani, lack leg brushes (Gupta, 1972; Kikkawa and Peng, 1938; Okada and Carson, 1983; Wynn and Toda, 1988). However, Anaprionus is now thought to be polyphyletic (Yassin, 2007; Yassin et al., 2010), and these species appear to be more closely related to the genus Xenophorticella than to Zaprionus sensu stricto (M. Toda, pers. comm.). Other Anaprionus species such as Z. bogoriensis, Z. obscuricornis, and Z. pyinoolwinensis have leg brushes (Mainx, 1958; Okada, 1964; Wynn and Toda, 1988) and likely form a clade with the Afrotropical Zaprionus (M. Toda, pers. comm.). Thus, the leg brush has evolved either at or near the base of Zaprionus.
D. pruinosa belongs to the loiciana species complex, which also includes D. loiciana, D. allochroa, D. pachneissa, D. semipruinosa, and D. xanthochroa. All of these species have male leg brushes of different sizes (Tsacas, 2002; Tsacas and Chassagnard, 2000). The fourth lineage where a male leg brush is found consists of a single species, D. repletoides, which does not have any known close relatives; a species described originally as D. tumiditarsus (Tan et al., 1949) was later synonymized with D. repletoides (Hsu, 1943; Wheeler, 1981). Yassin (Yassin, 2007) suggested that some species currently classified as Zaprionus (Z. multistriatus, Z. flavofasciatus, and Z. cercociliaris) could in fact be more closely related to D. repletoides than to Zaprionus; unfortunately, these species have not been included in any molecular phylogenies.
The four clades of interest – Zaprionus, the immigrans species group, D. repletoides, and the loiciana complex – have never been included together in the same molecular phylogeny. Different combinations of these taxa have been examined in several phylogenetic studies, which were based on a small number of loci and produced different results. Da Lage et al (Da Lage et al., 2007) and Yassin et al (Yassin et al., 2010) provided some evidence for a distant relationship among Zaprionus, D. repletoides, and the immigrans species group. Russo et al (Russo et al., 2013) placed D. pruinosa as sister to D. sternopleuralis (a member of the histrio species group that lacks a sex brush), and the resulting clade as sister to the immigrans species group. The phylogenies of Da Lage et al (Da Lage et al., 2007) and Izumitani et al (Izumitani et al., 2016) did not include D. pruinosa, but did not support a sister-group relationship between D. sternopleuralis and the immigrans species group.
In this study, we used a larger multilocus dataset to test whether the male leg brush evolved independently in each of these four clades, or whether its distribution could be better explained by shared origin in some of these lineages. To facilitate this analysis, we included one or more representatives of each clade, as well as several brush-less species that have been suggested by previous studies to be closely related to the brush-bearing clades. In parallel, we compared the cellular mechanisms that produce the male leg brushes in different species, as well as the role of these ornaments in mating behavior. Our results strongly suggest that the sex brushes evolved convergently in several distantly related lineages, but develop through virtually identical mechanisms.
Materials and Methods
Leg imaging
For brightfield imaging, male front legs were dissected, mounted in Hoyers media between two coverslips, and photographed on a Leica DM500B microscope with a Leica DC500 camera. For electron microscopy, adult legs were dehydrated in 100% ethanol, critical point dried, and coated with gold. Scanning electron micrographs were taken on Thermo Fisher Quattro S and Philips XL30 TMP.
Sequence data collection
The sources of live Drosophila strains, fixed specimens, and unpublished genome assemblies used in this study are listed in Table S1. DNA was extracted from live or alcohol-fixed flies using an affinity resin based protocol (Hi Yield® Genomic DNA Mini Kit, Süd-Laborbedarf Gauting, Germany). PCR was carried out using DreamTaq polymerase (Thermo Fisher) and the following cycling conditions: 95° 5’ => (95° 30” => 55° 30” => 72° 80”)x35 => 72° 5’ =>12°; the loci and primer sequences are listed in Table S2. In some cases, two rounds of PCR with nested primers were needed to obtain amplicons from fixed specimens. Amplified fragments were gel-purified and sequenced from both ends using amplification primers. Sequence chromatograms were trimmed in SnapGene Viewer, and the two end reads were aligned and edited in Geneious. Heterozygous nucleotide positions, if present, were represented by IUPAC ambiguity codes. All new sequences were deposited in Genbank under accession number listed in Table S1. Additional sequences were obtained from Genbank or extracted from whole-genome assemblies using Blast v2.2.23 (Table S1).
Sequence analysis
The sequences of each locus were aligned using the MUSCLE algorithm (Edgar, 2004) implemented in Geneious (Kearse et al., 2012). The alignments were trimmed at the ends, and poorly aligning intronic regions were removed. The alignments of all eight loci were then concatenated for combined analysis. Combined Bayesian analysis was carried out in MrBayes v3.2.6 (Ronquist et al., 2012). Two sets of analyses were conducted. In the first, the dataset was partitioned by gene, and each locus was allowed to follow a different nucleotide substitution model with empirically estimated parameters. In the second, we partitioned the dataset by gene and by codon position, and used PartitionFinder with the PhyML algorithm (Guindon et al., 2010; Lanfear et al., 2012; Lanfear et al., 2017) to identify the appropriate partitioning scheme; this resulted in a total of 21 character subsets, each of which was subsequently allowed to follow its own substitution model with empirically estimated parameters. For each analysis, two parallel runs of 1,500,000 generations, each starting from a different random tree, were carried out, and convergence was confirmed by comparing tree likelihoods and model parameters between the two runs. D. melanogaster was used as outgroup. Trees were sampled every 1000 generations and summarized after a 20% relative burn-in. Each analysis was also repeated after excluding D. quadrilineata (see Results). Samples of probable trees were extracted from the .tprobs file, and a strict consensus of most probable trees with combined posterior probabilities of 95% or 99% was constructed from these sets of trees in Geneious. Consensus trees were then formatted using FigTree v1.3.1 (http://tree.bio.ed.ac.uk/software/figtree/).
Immunocytochemistry and microscopy
Fly cultures were raised on standard Drosophila media at room temperature. Since each species develops at a different rate, the timing of pupal stages was determined empirically based on the morphology of transverse bristle rows (TBRs). Each species was imaged at an early stage when TBR bristles of the tibia and the first tarsal segment are separated by one or more intervening epithelial cells; this stage corresponds to 16 - 21 hours after pupariation (AP) in D. melanogaster. Each species was also imaged at a late stage, after the packing of TBRs is completed and bristle shaft differentiation is underway (corresponding to 36+ hrs AP in D. melanogaster). Pupal legs were dissected, processed and immunostained as in Tanaka et al (2009). The primary antibodies used were rat anti-E-cadherin (DCAD2, from the Developmental Studies Hybridoma Bank, at 1:20) for D. pruinosa and Z. tuberculatus, and mouse anti-Armadillo (N2 7A1, DSHB; 1:30) for D. immigrans and D. repletoides. AlexaFluor 488 secondary antibodies (Invitrogen) were used at 1:400. Fluorescent images were taken on an Olympus 1000 confocal microscope and processed using Image J and Adobe Photoshop.
Analysis of mating behavior
High-speed videos of mating behavior were recorded using a Fascam Photron SA4 mounted with a 105 mm AF Micro Nikkor lens. In brief, individual virgin males were isolated upon eclosion in food vials and aged for up to two weeks. Virgin females were isolated upon eclosion and housed in groups of 20-30. Pairs of males and females were then gently aspirated into single wells of a 96 well culture plate (Corning 05-539-200) filled halfway with a hardened 2% agarose solution and sealed using a glass microscope slide and tape. Video clips were captured at 1000 frames per second (fps) using Photron Fastcam Viewer software.
Results
Sex brush morphology is highly similar in distantly related species
In all examined species of Zaprionus, the immigrans species group, loiciana species complex, and D. repletoides, the male-specific leg brush is found on the anterior-ventral surface of the first (ta1) and sometimes also the second (ta2) tarsal segment of the prothoracic leg. In females and in other species, this area is occupied by transverse bristle rows (TBRs), which the flies use to clean their head and eyes (Tokunaga 1962; Kopp 2011). The male brush replaces the distal TBRs, with a few TBRs remaining at the proximal end of ta1 (Figure 1). In the immigrans group (Rice et al. 2018) and in the other species (Figure 1), the brush shows several major differences from the TBRs. The bristles of each TBR are aligned into a straight, tightly packed row that is nearly perpendicular to the proximo-distal (PD) leg axis, while the consecutive TBRs along the PD axis are separated by many cell diameters. In contrast, the modified bristles (“hairs”) that make up the brush are packed closely together in all directions and do not show any regular arrangement. The shafts of the TBR bristles are robust and straight, with ridges and grooves running their length and a triangular bract at the base; the brush hairs are thin and wavy with a smooth surface, and lack bracts (Figure 1).
Although the morphology of the male leg brush is similar across species, detailed scanning electron microscopy (SEM) analysis reveals a small but consistent difference. In D. immigrans and D. pruinosa, the tips of brush hairs are thin and flat, taper to a point, and form hooks that curve toward the base of the leg (Figure 1B, D). In D. repletoides and in Zaprionus, the hair tips are noticeably thicker and curve away from the leg base (Figure 1 F, H, J, L); in Zaprionus, they are also flattened into paddle-like shapes (Figure 1 H, J, L). Thus, while the spatial arrangement of bristles appears to be similar in all species, some differences exist in the morphology of bristle shafts. These differences may reflect the phylogenetic relationships among these taxa, especially the close relationship between D. pruinosa and the immigrans species group (see next section).
Male leg brushes evolved independently at least three times
We sequenced partial coding sequences of eight nuclear, protein-coding loci: acon, eno, glyp, Amyrel, Ddc, Gpdh, Pepck, and Pgm. Separate analyses of each locus produced very poorly resolved trees. We therefore combined the data from all loci (up to 9060 nucleotides per species) for a partitioned Bayesian analysis where each locus was allowed to follow its own, empirically estimated substitution model but all loci were constrained to the same tree topology. The resulting tree (Figure 2A; brush-bearing clades labeled in blue) suggests a close relationship of D. pruinosa to the immigrans species group, with the (D. sternopleuralis + D. trisetosa) clade, which belongs to the histrio species group, as the next outgroup. In contrast, D. repletoides is placed as sister group to the (D. busckii + D. brachytarsa) clade, well away from the immigrans-pruinosa lineage. Finally, the Zaprionus genus is distantly related to both D. repletoides and the immigrans-pruinosa lineage, and is placed near the base of the tree, separately from the Drosophila and Sophophora subgenera. We also note that D. curviceps and D. annulipes appear as sister groups with 100% support, while there is no support for clustering the immigrans species group either with the (D. curviceps + D. annulipes) clade or with D. quadrilineata (see Supplementary text). A strict consensus of 11 trees with the cumulative posterior probability of 95% is not resolved near the base, but does not support a close relationship among the brush-bearing lineages: D. repletoides, the immigrans-pruinosa clade, and Zaprionus (Figure S1).
We then examined 27 most probable trees, with the cumulative posterior probability of 99%, to determine the probability of each bipartition of interest to our study. The Zaprionus genus (occasionally together with D. quadrilineata; see below) was placed at the base of the tree, well separated from D. repletoides and the immigrans-pruinosa lineage (Figure 2A, Table S3). D. pruinosa clustered with the immigrans species group, to the exclusion of the (D. sternopleuralis + D. trisetosa) clade, with ∼87% probability; the alternative grouping, of D. pruinosa with the (D. sternopleuralis + D. trisetosa) clade to the exclusion of the immigrans species group, was observed with ∼13% probability. D. repletoides was grouped with the (D. busckii + D. brachytarsa) clade with 95% probability, and this group was separated from the immigrans-pruinosa lineage by multiple internal branches. In contrast, potential groupings of brush-bearing lineages – for example, of Zaprionus with either D. repletoides or the immigrans-pruinosa clade, or of the repletoides-busckii-brachytarsa clade with the immigrans-pruinosa clade – were never observed among this set of probable trees.
We noticed that among these 27 probable trees, the position of D. quadrilineata was by far the most unstable. We therefore repeated the analysis after excluding this species. The resulting tree (Figure 2B, Table S3) shows the same relationships as the full analysis (Figure 2A), but with stronger support for most basal nodes. This tree, as well as the strict consensus of the 11 most probable trees with the cumulative posterior probability of 99% (Figure S2) places Zaprionus near the base of the tree, separately from the Drosophila and Sophophora subgenera; D. pruinosa together with the immigrans species group and the (D. sternopleuralis + D. trisetosa) clade; and D. repletoides in a clade that is clearly separated both from Zaprionus and from the immigrans-pruinosa lineage. The probability of sister-group relationship between D. pruinosa and the immigrans species groups is estimated at 93.8% ± 1.32% across these trees.
Finally, we carried out an analysis with a more complicated partitioning scheme, where each locus and each codon position was allowed to follow its own substitution model; this analysis was also performed with and without D. quadrilineata. In both cases, it produced a tree with the same topology as in the simpler partitioning scheme, but with slightly different levels of node support (Figure S3, S4 and Table S3). In summary, we find substantial though not overwhelming support for a close relationship between D. pruinosa, and by implication the loiciana species complex, and the immigrans species group. However, based on our data, the probability of a close relationship among the different brush-bearing clades – D. repletoides, Zaprionus, and the immigrans-pruinosa lineage – is very low for each of the possible pairwise relationships.
The same cellular processes underlie sex brush development in all species
In D. melanogaster, TBR bristle precursors are specified between 6 and 12 hours after pupariation (AP) (Joshi and Orenic 2006, Schroff 2007). Initially, these cells are specified in sparse, loosely organized rows, and separated from one another by several epithelial cells. By 20-21 hrs AP, the bristle cells that are destined to make each TBR migrate toward each other to form a straight, contiguous row, while the intervening epithelial cells are expelled distally and proximally from the TBRs (Atallah et al. 2009; Tanaka et al. 2009). This mechanism suggests two potential explanations for the tight packing of brush hairs. One possibility is that the hair progenitor cells are specified with minimal spacing; in this case, the initial spacing between hair cells is expected to be much denser than the spacing of TBR progenitors. Alternatively, hair progenitors may first be specified similarly to TBR bristles, with wide separation by epithelial cells, but expel the epithelial cells (either laterally or basally) at later stages to form a densely packed brush. In principle, independently evolved brushes in different species could utilize different cellular mechanisms to produce adult structures that are essentially indistinguishable.
In order to characterize and compare brush development in different species, we used antibodies against membrane-localized proteins to visualize cell arrangement in pupal legs. When labeled with antibodies against the beta-catenin Armadillo (Arm) or the E-cadherin Shotgun (DE-cad), bristle cells can be distinguished from the surrounding epithelial cells by their unique membrane shape (Figure 3). We examined brush development at two timepoints: an early stage roughly corresponding to ∼16-21 hr AP in D. melanogaster, when the bristle cells of the future TBRs begin to migrate toward each other and expel the intervening epithelial cells, and a later stage when cell migration is completed. We found that at the early stage, most bristle cells in the developing brush are each surrounded by four to six epithelial cells in all species. In effect, the bristle cells are separated from one another by one to two epithelial cells (Figure 3). At the late stage, this spacing remains virtually invariant, although the cells appear more organized compared to the early stage (Figure 3). We did not see evidence of cell migration in any of the examined species. These observations indicate that despite the packed appearance similar to TBRs, the brush cells are not directly adjacent to each other, are specified at high density, and undergo minimal if any migration during development. Importantly, the cellular mechanism of brush development is very similar in all species.
Leg brushes are used in male mating behavior
High-speed video recordings show that the proximal tarsal segments of the male T1 legs, including the sex brushes, are used to grab the female abdomen and resist the female’s efforts to dislodge the male during copulation attempts in Z. tuberculatus, D. immigrans, and D. repletoides (Supplementary movies 1-3); unfortunately, no mating attempts were observed in D. pruinosa. In the Drosophila melanogaster species group, males use their sex combs, which are also located on the T1 tarsus, to grab the female genitalia as in D. melanogaster, or the middle abdominal segments as in D. kikkawai, D. ananassae, and D. bipectinata (Massey et al., 2019). In these species, male T2 and T3 legs remain on the substrate and are not involved in mating. Typically, these males proceed very quickly from mounting to attempted copulation; females may resist by walking away or using their hind legs to kick the male off. Interestingly, similar behavior is observed in D. willistoni, one of the closest relatives of the melanogaster species group that lacks sex combs (Massey et al., 2019), suggesting that the origin of the sex comb did not radically change this aspect of mating. In contrast, D. immigrans, Z. tuberculatus, and D. repletoides males use their sex brushes to grab females more anteriorly, near the constriction between the thorax and abdomen (Supplementary movies 1-3). In the former two species, males also use their T2 legs to grab the female mid-abdomen, while in D. repletoides T1, T2 and T3 legs are all used to grab the female so that the male “rides” on the female and is not in contact with the substrate. In all these species, females appear to resist mating attempts more vigorously than in the melanogaster group, using side-to-side bucking and wing vibrations in apparent efforts to dislodge the male, while the males use their legs to resist these efforts. The delay between mounting and attempted copulation is longer in the brush-bearing species, especially in Z. tuberculatus, than in the comb-bearing species; most mountings result in the male being eventually dislodged and do not lead to copulation attempts.
A more systematic analysis, including many lineages that lack male-specific leg modifications, will be needed to test whether morphological evolution correlates with the evolution of behavior. At this point, we can only speculate that male leg brushes, which consist of hundreds of thin hairs that are hooked at the tips and have a very large combined surface area, may have evolved to provide a more secure grip of the female abdomen, especially if stronger grip is needed to counteract the female attempts to dislodge the male.
Discussion
Repeated evolution of leg brushes
Our results suggest that the male-specific brush on the tarsal segments of front legs has evolved independently at least three times in Drosophilidae: in Zaprionus, in D. repletoides, and in the common ancestor of the immigrans species group and the loiciana complex, which includes D. pruinosa. Although our study does not provide unequivocal support for a sister-group relationship between D. pruinosa and the immigrans group (a closer relationship of D. pruinosa to the sternopleuralis-trisetosa clade cannot be completely ruled out, and of course there may exist other closely related species that we have not sampled), independent evolution of brushes in the immigrans species group and the loiciana species complex appears less likely than a single origin in a common ancestor of these clades. The proximally curving tips of brush hairs in D. pruinosa and D. immigrans, distinct from the distally curving tips in Zaprionus and D. repletoides (Fig. 1), are consistent with a close relationship between the immigrans group and the loiciana species complex.
Convergent origin of leg brushes in Zaprionus, D. repletoides, and the immigrans-loiciana clade is remarkable given the strong structural similarities of these brushes, and especially the fact that the cellular mechanisms that produce them in different species are essentially identical. In all species examined, the bristles that make up the sex brush are specified with only one or two intervening epithelial cells between them. Bristle specification in Drosophila and other insects is governed by a lateral inhibition mechanism, which is based on contact signaling between adjacent cells, and prevents two adjacent cells from both assuming the fate of bristle precursors (Simpson, 1990). Later in development, a tighter packing of bristles can be achieved through cell migration, as observed for example in the transverse bristle rows of the front legs (Atallah et al., 2009; Tanaka et al., 2009). However, no cell migration is observed during sex brush development; instead, the future brush hairs are always specified at the maximum density allowed by lateral inhibition. Independent evolution of this highly derived spatial pattern in multiple lineages suggests that similar selective pressures may elicit not only similar structures, but also similar changes in the underlying developmental mechanisms.
A hierarchy of convergence, from structures to genes
Convergence is a pervasive phenomenon in evolution. Pictures comparing bats and birds, fish and dolphins, or the vertebrate and cephalopod eyes are textbook clichés. Less obvious but equally striking examples of convergent forms shaped by a common function include the asymmetrical ears of owls (Nishikawa, 2002), Mullerian mimicry in butterflies (Brower, 1994; Naisbit et al., 2003), trophic morphology in cichlid fishes (Ruber and Adams, 2001), and many others (Moore and Willmer, 1997). A fundamental question raised by the widespread occurrence of convergent traits is to what extent phenotypic convergence reflects an underlying similarity of the molecular and cellular mechanisms that control individual development.
The mechanisms of convergence can be viewed in either genetic or developmental terms. From the genetic standpoint, convergent phenotypes are controlled by the same mechanism if they are the result of changes in the same loci (“genetic convergence”) (Gompel and Prud’homme, 2009; Martin and Orgogozo, 2013). Alternatively, convergent changes may be considered to share a similar mechanism if they are produced by the same changes in development, regardless of the ultimate genetic causes (“developmental convergence”) (Wake, 1991; Wake et al., 2011; Wray, 2002). We can describe convergence in hierarchical terms by applying three increasingly stringent criteria: phenotypic, developmental, and genetic. Phenotypic convergence can occur without developmental convergence: for example, the disproportionally elongated bodies of burrowing salamanders can be caused either by an increase in the number of vertebrae, or by increased length of the individual vertebrae – a superficial phenotypic resemblance produced by entirely different developmental mechanisms (Parra-Olea and Wake, 2001). Similarly, larval abdominal prolegs of butterflies and sawflies, while superficially similar and playing similar roles in locomotion, develop from non-homologous leg segments (Suzuki and Palopoli, 2001). In D. melanogaster, latitudinal clines in wing size have evolved on several continents, but the differences in wing size are due largely to differences in cell number in some populations, while in other populations differences in the size of individual cells make a larger contribution (James et al., 1997; Zwaan et al., 2000).
Developmental convergence, in turn, does not necessarily reflect genetic convergence. In Drosophila, similar color patterns can evolve in different species through changes at different loci (Signor et al., 2016; Wittkopp et al., 2003b; Yassin et al., 2016), but in the end all of these genetic changes are translated into adult phenotypes through the same set of enzymatic reactions in the catecholamine metabolism pathway (Kronforst et al., 2012; Wittkopp and Beldade, 2009; Wittkopp et al., 2003a), presenting an example of developmental convergence in the absence of genetic convergence. In other cases, the evolution of similar phenotypes in distantly related species is clearly caused by independent changes in the same genes – as, for example, in the loss of larval cuticular hairs through regulatory mutations in the shavenbaby gene in two Drosophila species separated by >40 million years (Sucena et al., 2003), or in the evolution of mimetic wing color patterns in different species of Heliconius butterflies through parallel changes in the optix and WntA genes (Gallant et al., 2014; Hines et al., 2011; Martin et al., 2012; Reed et al., 2011; Supple et al., 2013).
Going even deeper from the genetic to the molecular level, we can ask whether recurrent involvement of the same gene in multiple instances of convergent evolution reflects repeated appearance of the same mutations, or whether different mutations can alter the function of the causative gene in similar ways, leading to the evolution of similar phenotypes. Again, we can find examples in support of either scenario. A variety of proteins have evolved similar amino acid residues and active sites repeatedly in different taxa (Gasparini et al., 1998; Hill et al., 2019; Kriener et al., 2000; Lopreato et al., 2001; Yokoyama, 2002). Although non-coding DNA sequences are clearly less constrained in their function than proteins, similar regulatory mutations can nevertheless appear and be fixed independently in different populations or species in response to similar selective pressures (Chan et al., 2010; Loehlin et al., 2019; Xie et al., 2019). Conversely, mutations that affect different amino acids, or different cis-regulatory elements, can produce similar phenotypic outcomes in different populations or species (Gross et al., 2009; Kingsley et al., 2009; Manceau et al., 2010; Protas et al., 2006; Rosenblum et al., 2009; Yassin et al., 2016).
The sex brush represents a clear case of developmental convergence, but the genetic control of its development remains to be determined. Specification and positioning of the sex brush are likely to involve the HOX gene Scr and the sex determination gene doublesex (Rice et al., 2018; Tanaka et al., 2011). However, the identity of the downstream genes that translate the regulatory information provided by dsx and Scr into the final 3-dimensional structure is unknown. In principle, these genes need not be the same in different lineages. Future work may show whether the convergence we observe at the level of cell behavior extends to the genetic and molecular level.
Convergent innovation and the failure to innovate
The evolution of similar structures in response to similar functional demands is perhaps not surprising; should we instead be surprised when organisms faced with similar demands fail to evolve convergent phenotypes (Blount et al., 2018)? Males grasp female abdomen with their front legs in many Drosophila species, including those that primitively lack any male-specific leg ornaments (Massey et al., 2019; Spieth, 1952). Why don’t all Drosophila species evolve sex combs, brushes, or other grasping structures? Development does not provide any clues. At the level of cell behavior, both the ancestral/female condition (cell migration that produces tightly packed bristle rows from sparsely spaced precursors) and the derived/male condition (specification of bristle precursors at the maximum density permitted by lateral inhibition) are the same in all lineages where the sex brush is present. There is no a priori reason to think that the transition between these modes of development is easier in some species than others.
The answer may lie instead in either behavior or population genetics. Although males of different species use their sex brushes in at least superficially similar ways, we don’t know the female side of the story. If females of different species vary in their responses to male grasping, the evolution of specialized leg structures in males may not be universally favored. This may also explain why both the sex brushes and the sex combs (Kopp, 2011) have been secondarily lost multiple times. Moreover, it is difficult to know whether the female preferences observed today are the same as they were in the distant past when the male-specific structures evolved (Watts et al., 2019).
Alternatively, the origin of a new trait such as the leg brush may require such an unlikely series of genetic changes that it may often fail to occur even in response to strong selective pressure. For example, it is possible that while a single mutation is sufficient to modify or eliminate an existing morphological structure, the origin of a new structure may require simultaneous changes in multiple genes. From the population-genetic perspective, this would mean that functionally novel and positively interacting alleles at multiple loci must segregate in the same population at the same time in order for selection in favor of a new structure to be effective. Naturally, this would greatly reduce the probability of evolutionary innovations compared to other types of phenotypic change. This is of course pure speculation; we do not know why convergent innovations evolve in some lineages but fail to evolve in others. We hope that research models where both the functional roles and the genetic basis of novel traits can be studied in parallel will shed some light on this intriguing question in the future.
Supplemental text - The immigrans species group
The immigrans species group and the wider immigrans-tripunctata radiation have long played a pivotal role in Drosophila systematics (Throckmorton, 1975; Yassin, 2013). However, the composition of the immigrans group has not been entirely clear. Historically, this group was proposed to include five subgroups: immigrans, hypocausta, nasuta, quadrilineata, and curviceps (Huang et al., 2002; Zhang and Toda, 1988, 1992; Zhang et al., 1995). Recent studies have shown, however, that both the hypocausta and the immigrans subgroups are likely polyphyletic, as some species assigned to these subgroups are actually closer to the nasuta subgroup (Da Lage et al., 2007; Katoh et al., 2007; Rice et al., 2018). The quadrilineata and curviceps subgroups have been the source of even greater complications. Previous phylogenetic studies have provided evidence against a sister-group relationship between D. quadrilineata and the rest of the immigrans species group (Katoh et al., 2007; Morales-Hojas and Vieira, 2012; Yassin, 2013). Moreover, D. annulipes, traditionally assigned to the quadrilineata subgroup (Lin and Tseng, 1973), was found to be distantly related to D. quadrilineata, and closer to the virilis-repleta radiation (Katoh et al., 2007). The phylogeny of Huang et al (Huang et al., 2002) showed a close relationship between D. annulipes and the clade composed of D. curviceps and D. oritisa, another species assigned to the curviceps subgroup. Morales-Hojas and Vieira (Morales-Hojas and Vieira, 2012) confirmed the distant relationship between D. annulipes and D. quadrilineata, as well as between D. annulipes and the immigrans species group sensu stricto. Yassin (Yassin, 2013) pointed out important differences between D. quadrilineata and the curviceps subgroup and the immigrans group s.s. Finally, Pradhan et al (Pradhan et al., 2015) removed the curviceps subgroup from the immigrans species group, elevating it to the status of a separate species group.
Our analysis, based on a larger amount of sequence data than previous studies, confirms these observations, namely that (1) D. quadrilineata is not closely related either to the immigrans species group s.s. or to D. annulipes, (2) D. curviceps is not closely related to the immigrans species group s.s., and (3) D. annulipes is related to D. curviceps. Based on these results, we support the conclusions on other authors (Katoh et al., 2007; Pradhan et al., 2015; Yassin, 2013) that the definition of the immigrans species group should be restricted to the immigrans, hypocausta, and nasuta subgroups. Further work, with a more extensive sampling of the immigrans group s.s., is needed to revise its internal taxonomy and re-establish monophyletic subgroups. Similarly, better taxon sampling will be needed to determine whether D. annulipes could be reassigned to the curviceps species group (Pradhan et al., 2015), or whether it is possible to establish a monophyletic quadrilineata species group.
Figure Legends
Supplement Figure 1. Strict consensus of 11 trees with the cumulative posterior probability of 95% for the complete taxon sample, labeled as in Figure 1. Numbers at each node indicate the posterior probabilities of the respective taxon bipartitions.
Supplement Figure 2. Strict consensus of 11 trees with the cumulative posterior probability of 99% for the taxon sample without D. quadrilineata.
Supplement Figure 3. Bayesian phylogenetic tree for the complete taxon sample, with the dataset partitioned by gene and codon.
Supplement Figure 4. Phylogenetic tree for the taxon sample without D. quadrilineata, with the dataset partitioned by gene and codon.
Acknowledgements
We thank the US Drosophila species stock center, Ehime University Drosophila stock center, J.-R. David, S. Prigent, and M. Watada for Drosophila strains and specimens; Developmental Studies Hybridoma Bank for antibodies; A. Comeault, A. Gloss, M. Lang, D. Matute, D. Miller, V. Orgogozo, and N. Whiteman for genome assemblies; M. Toda for advice on Zaprionus phylogeny; G. Rice for D. immigrans images; and A. Yassin for comments on the manuscript. We also thank the MCB Light Microscopy Imaging Facility, Department of Cell Biology and Human Anatomy EM Lab, and Materials Science and Engineering AMCaT Lab at UC Davis for equipment and imaging assistance, and Gwyneth Card and W. Ryan Williamson for use of the Photron camera. The Olympus FV1000 confocal used in this study was purchased using NIH Shared Instrumentation Grant 1S10RR019266-01. AK and OB are grateful to N. Gompel for allowing some of this work to be conducted in his lab while on a sabbatical visit, and to members of the Gompel lab for treating them as their labmates. This work was supported by NIH grant 5R35GM122592 to AK and by funds from Ludwig-Maximilians-Universität München to Nicolas Gompel.