Aedeagus morphology is a diagnostic trait that along with chromosomal inversions provides a guide for species identification in the genus Drosophila and particularly in the D. repleta species group [6,19,20,32]. Several studies have, recently, turned the attention to the D. buzzatii cluster, a guild of cactophilic flies, in active cladogenesis, that inhabit the arid regions of Southern South America . However, recent molecular phylogenetic studies  cast doubts on the reliability of male genital (aedeagus) morphology to infer the evolutionary relationships in the D. buzzatii cluster . For instance, mitochondrial DNA sequence data place D. koepferae as the sister species of D. buzzatii, albeit the comparative analysis of aedeagus morphology indicates a close relationship with D. serido and allied species . These results indicate that rates of evolution of male genitalia may be heterogeneous among branches of the clusters' phylogenetic tree. In this context, our study, motivated by such inconsistency between molecular and morphological data, provides valuable information that could help to distinguish the relevant factors involved in morphological variation of male genitalia in D. buzzatii and D. koepferae, a pair of species in the heart of an evolutionary conflict.
The first issue raised by our study is that aedeagus size and shape vary substantially in both species, and, that a significant portion of variation is genetically determined. Moreover, the inclusion, in our experimental design, of semi-natural rearing media prepared with different cactus hosts permitted the characterization of morphological variation in terms of phenotypic plasticity in both species. In this sense, phenotypic plasticity in genital morphology was evident in D. buzzatii: flies emerged in alternative cactus hosts showed significant differences in aedeagus size and shape. Moreover, the analysis of the sources of variation also revealed a significant Line by Cactus interaction in D. buzzatii, i.e. isofemale lines did not respond in the same way to the two environments tested, suggesting that the plastic response of genital morphology has a genetic basis . Such a plastic expression of genital morphology constitutes the first evidence of a rearing substrate affecting genital morphology in Drosophila, and is in line with a recent study showing a plastic response of aedeagus morphology in relation to rearing temperature in D. mediopunctata . In contrast, comparisons between males of D. koepferae emerged in different cactus media did not reveal any sign of phenotypic plasticity, suggesting a more canalized (non plastic) development of male genitalia.
Another relevant point is that patterns of allometry within (aedeagus size and shape) and between organs (aedeagus and wing size) also differed between D. buzzatii and D. koepferae. In the latter, shape and size of male genitalia, as well as aedeagus size and wing length, appeared to be largely uncoupled as suggested by the low level of within and between organ allometries. These results suggest that differences among flies in wing length are not expected to be accompanied by changes in the size of the genitalia, indicating that the factors involved in development of wings and male genitalia are largely independent in D. koepferae. In contrast, aedeagus shape and size were significantly correlated in D. buzzatii, suggesting that factors affecting size (for instance the type of cactus host) may also indirectly affect the shape of the organ.
In D. buzzatii, body size related traits (such as wing length), known to be under natural selection [36-38], are affected by the nature of the breeding substrate [21,39-41]. Furthermore, a positive correlation between body size and male reproductive success is well known in this species [38,42-44]. If the phenotypic correlation observed between aedeagus size and wing length has a genetic basis, any directional selective pressure affecting body size (wing length) would indirectly affect the evolution of male genitalia.
Several features of the mating system, such as female remating frequency, premating time, copulation duration, interval between successive matings, and progeny numbers, have been shown to be genetically variable in D. buzzatii . However, the connection between these traits and male genital morphology has not been explored. Indeed, the implications of our present results in genital evolution and speciation would obviously depend on the kind of relationship between genital morphology and mating success, as it was reported to occur in other insect taxa such as Heteroptera  and Coleoptera .
To this point we have presented basic features of the patterns of variation of aedeagus morphology in D. buzzatii and D. koepferae, now, we would like to examine whether our data allow a critical evaluation of the plausibility of the three main hypotheses proposed to explain genital evolution. Though our results are not entirely conclusive in this respect, the extensive phenotypic and genetic variation in aedeagus morphology are strong evidence against the "lock and key" hypothesis, that predicts low levels of variation (both phenotypic and genetic) in genital structures. However, the correlation between aedeagus morphology and wing length, along with the condition dependence (phenotypic plasticity in relation to cactus hosts) are in agreement, at least in D. buzzatii, with predictions of the hypothesis of pleiotropy. Concerning the third hypothesis, sexual selection, we must await for the results of experiments testing the relationship, if any, between genital morphology and reproductive performance.
The final issue we would like to address is whether our study allows us to envisage the genetic architecture underlying differences in male genitalia between D. buzzatii and D. koepferae. In this sense, our results seem in conflict with the single available work comparing genital morphology in hybrids and parental species [17,47]. Genital size and shape of hybrid males were not intermediate and the morphological resemblance of hybrids to either D. buzzatii or D. koepferae varied among crosses. In fact, hybrid's morphological distance to D. buzzatii and D. koepferae depended on the parental strains employed in the crosses. In none of the 4 crosses hybrid morphology was phenotypically intermediate (Figure 4). Hybrids from crosses 4853 and 8832 tended to be more similar to the D. koepferae parental strain, while in crosses 3512 and 4855 the morphology of the hybrid genitalia resembled closer that of D. buzzatii male parent. These results seem to be incompatible with the idea that interspecific differences in the morphology of male genitalia are caused by polygenes with small additive effects as claimed by Coyne and Orr for D. simulans and D. mauritiana  (however, it should be noted that the authors in  acknowledged some degree of dominance and epistasis). Actually, our results suggest a complex genetic architecture probably involving a certain degree of dominance and the involvement of genetic factors with large effect.
However, there are certain differences between our study and Liu et al's  that are worth mentioning. The first relates to the part of the genitalia examined in each case, the intromittent organ in D. buzzatii and D. koepferae and the posterior lobe (a particular element of male genitalia in the D. melanogaster group ) in D. simulans and D. mauritiana. These organs perform different functions during copulation  and therefore their evolution might be governed by different processes. The second is methodological and can be avoided by applying our methodology to Liu et al's dataset. To this end, we captured the outlines available in the digital version of , and tested for shape differences between F1 hybrids and parental species. This reanalysis confirmed that F1 hybrids have an intermediate morphology between D. simulans and D. mauritiana (the morphological distances between hybrids to both parental phenotypes were not significantly different: F1, 18 =, 008, p = 0, 93). Another non trivial point, that may complicate our interpretation is the difference in the time of divergence between the members of the two pairs of species, since development in interspecific hybrids is a result of a balance between the effects of the degree of heterozygosity and the degree of genomic coadaptation (expected to increase/decrease as a function of divergence, respectively) and the outcome of the past selection pressures on the species studied (see ). D. simulans and D. mauritiana are two recently derived species that shared their last common ancestor 0.6 – 0.9 million years ago , while D. koepferae and D. buzzatii are older species that diverged 5 million years ago . Finally, D. simulans and D. mauritiana are homosequential species, i.e. their basic polytene chromosome banding patterns are identical , whereas two inversions became fixed since divergence in D. buzzatii and D. koepferae. In addition, rich second chromosome inversion polymorphisms have evolved independently in the latter pair of species [20,23]. Our knowledge of inversion polymorphisms is mostly restricted to D. buzzatii, in which polymorphic inversions are known to affect morphological and fitness related traits (see  and references therein). Although there is no direct evidence linking inversions and genital morphology, inversions may affect aedeagus morphology via its effect on general body size (recall the allometric relationship detected between aedeagus size and wing length in D. buzzatii). In this context, the idea that morphology (size and shape) of an organ potentially involved in species recognition (such as aedeagus morphology), might be associated to polymorphic inversions is consistent with recent theories linking chromosomal rearrangements and reproductive isolation [53,54], and deserves further investigation.