The role of courtship song in female mate choice in South American Cactophilic Drosophila

Courtship songs have undergone a spectacular diversification in the Drosophila buzzatii cluster. Accordingly, it has been suggested that sexual selection has played a significant role in promoting rapid diversification, reproductive isolation and speciation. However, there is no direct evidence (i.e., song playback experiments with wingless males) supporting this tenet. Moreover, several studies have showed that the courtship song in the genus Drosophila is not always used in female mate choice decisions, nor plays the same role when it is taken into account. In this vein, we use an approach that combines manipulative and playback experiments to explore the importance and the role of courtship song in female mate choice in four species of the D. buzzatii cluster and one species of the closely related D. martensis cluster for outgroup comparison. We also investigate the importance of courtship song in sexual isolation in sympatry between the only semi-cosmopolitan species, D. buzzatii, and the other species of the D. buzzatii cluster. Our study revealed that the courtship song is used by females of the D. buzzatii cluster as a criterion for male acceptance or influences the speed with which males are chosen. In contrast, we showed that this characteristic is not shared by D. venezolana, the representative species of the D. martensis cluster. We also found that the studied species of the D. buzzatii cluster differ in the role that conspecific and heterospecific songs have in female mate choice and in sexual isolation. Our findings support the hypothesis that divergence in female preferences for courtship songs has played a significant role in promoting rapid diversification and reproductive isolation in the D. buzzatii cluster. However, evidence from D. venezolana suggests that the use of the courtship song in female mate choice is not a conserved feature in the D. buzzatii complex.


Introduction
Males display multiple signals of different sensory modalities during courtship and most of these signals have undergone spectacular diversification even in groups of closely related species [1,2]. However, it is not clear whether all of these signals are important in mate choice, or whether females emphasize one type of signal over others [3][4][5].
In flies of the genus Drosophila, males produce an innate complex courtship behavior in which acoustic and chemical signals are thought to be the more conspicuous targets of female mate choice [6]. Among these signals, the courtship song produced by wing vibrations has been the most extensively studied (e.g., [7][8][9][10][11][12][13][14]). Previous works evaluating the role of mating song in sexual selection and sexual isolation showed wide variation across species. For instance, it has been suggested that the courtship song of D. heteroneura males (D. planitibia group) does not influence male mating success and, therefore, cannot be considered a sexual signal [15]. On the other hand, females of D. montana (D. virilis group) reject the courting male if they do not hear the song [16]. In some species of the D. melanogaster complex, courtship song does not seem to be necessary for mating, although it decreases the time to copulation [17,18]. In addition, there is evidence that song characters do not need to be speciesspecific since heterospecific songs can also stimulate female sexual behavior in some species of the D. melanogaster complex [17,18]. Nevertheless, females of D. sechellia (D. simulans complex) and D. biauraria (D. auraria complex) reject males when song characters are heterospecific [19,20]. Overall, these studies show that although courtship songs vary extensively across species, it does not mean that they always play an important role in female mate choice or mate recognition [17].
The D. buzzatii complex consists of three species clusters, namely D. buzzatii, D. martensis and D. stalkeri. The distribution area comprises Argentina, Brazil, Bolivia and Paraguay (D. buzzatii cluster), Colombia and Venezuela (D. martensis cluster), and the Caribbean Islands and Florida (D. stalkeri cluster) [21]. Flies of the D. buzzatii complex inhabit arid and/ or semiarid environments where they breed and feed on fermenting cacti, either prickly pears (genus Opuntia) or columnar cacti [21,22]. Some species are host specialists, whereas others use predominantly one type of host (primary host) and less frequently alternative hosts (secondary hosts), indicating some degree of niche overlap. The evolutionary relationships within the D. buzzatii complex are poorly known since nodes have low support in all phylogenetic hypotheses proposed. Chromosomal [21] and molecular sequence data (including the nuclear gene Xdh and the mitochondrial genes COI, COII and COIII) [23] suggest that the D. stalkeri cluster is the oldest lineage and is paraphyletic with respect to the D. buzzatii and D. martensis clusters that are reciprocally monophyletic sister clades. However, recent work based on more extensive molecular data (four mitochondrial and six nuclear markers) recovered D. buzzatii and D. stalkeri clusters as sister clades [22].
Recent studies of courtship songs and cuticular hydrocarbons (CHC) in the D. buzzatii cluster, revealed rapid evolution of male courtship songs [24] and a certain degree of conservatism in CHC profiles [25]. In fact, despite quantitative differences among species, CHC profiles can be predicted by species ancestry. Therefore, it was suggested that courtship songs may encode information that mediates mating discrimination between species [24]. However, there is no direct evidence (i.e., song playback experiments with wingless males) supporting this tenet.
The evolution of male courtship songs in cactophilic flies of the D. buzzatii cluster is characterized by a pattern of losses and/or elaboration of their components [24]. Songs are composed of low-frequency, polycyclic pulses arranged into pulse trains that have different structural and temporal features [24]. Thus, some species like D. borborema and D. gouveai perform one type of song whereas others like D. antonietae, D. buzzatii, D. koepferae, D. serido and D. seriema perform two. Moreover, the courtship song of D. buzzatii males is the only one with a bimodal distribution of interpulse intervals (IPIs) in the primary or A song ( [11]; P.P. Iglesias pers obs). Surprisingly, Oliveira et al. [24] did not report this tendency to alternate pulses with long and short IPI in the primary or A song of D. buzzatii males. However, a bimodal distribution of IPIs in the primary song can be detected by measuring the IPIs in the sonogram of Fig 20C of Oliveira et al. [24] (see Discussion for more details). In addition, among species with two types of songs, only D. buzzatii males produced significantly different primary and secondary songs, i.e., short IPI in the primary song and long IPI in the secondary song ( [24]; P.P. Iglesias pers obs). IPI differences are interesting since insects usually have low spectral resolution and it is the temporal component that carries most information [26]. The importance of IPI in species recognition has been demonstrated in several species of Drosophila (reviewed in [27]).
Besides having a divergent courtship song, D. buzzatii also differs from the other six species of the cluster (the so-called "D. serido sibling set"; [28]) in ecological, morphological and biogeographic characteristics. First, it is the only species that prefers decaying tissues of prickly pears of the genus Opuntia as primary hosts, although it can also use columnar cacti as secondary hosts. Second, It has the most divergent aedeagus morphology (male genitalia), in both size and shape, as compared to the other species. Finally, it is the only widespread species that cooccurs with all other species of the cluster in different areas of its vast geographic range in South America ( [28,29]; Fig 1). These features make D. buzzatii an interesting target for evolutionary studies of traits involved in precopulatory sexual isolation because the strengthening of the mechanisms involved in mate discrimination can avoid the potential harmful hybridization [30]. In this sense, the courtship song of D. buzzatii males is a salient precopulatory sexual trait since it is complex enough to be easily discriminated by females of the species of the D. serido sibling set and to be easily recognized by conspecific females.
In the present study we use an approach that combines manipulative and playback experiments to explore the importance and the role of courtship song in female mate choice in four species of the D. buzzatii cluster and one species of the closely related D. martensis cluster for outgroup comparison. In addition, we focus on the importance of courtship song in sexual isolation in sympatry between D. buzzatii and the other species of the D. buzzatii cluster. We studied female mate choice in terms of variation in two aspects: which acoustic conditions favor male acceptance and how fast are males accepted in each acoustic condition.

Drosophila stocks
We used 13 isofemale lines of four species of the D. buzzatii cluster namely: D. antonietae (4 lines), D. borborema (1), D. buzzatii (4), and D. koepferae (4 [32]. We also included an isofemale line of D. venezolana of the D. martensis cluster for outgroup comparison. This species is also a cactophilic fly that is distributed in northern South America and is allopatric to all D. buzzatii cluster species ( [21]; Fig 1). This line was kindly provided by Antonio Fontdevila and derives from the progeny of a wild inseminated female collected in 1984 in Isla Los Roques (Venezuela) [31]).

Experimental conditions
Two types of no choice experiments were performed (Fig 2). In Experiment 1 we investigated the importance and the role of courtship song in female mate choice when unmanipulated  [25], and P.P. Iglesias, J. Hurtado, and E. Hasson, unpublished field data. The map was created using SimpleMappr http://www.simplemappr.net/ and edited in CorelDRAW X7.
https://doi.org/10.1371/journal.pone.0176119.g001 females were exposed to conspecific males without wings (i.e., without the structures involved in the emission of songs) in three different acoustic conditions: with no playback (NP), with conspecific song playback (CP), and with heterospecific song playback (HP). We also included a positive control (Ct+) that consisted of pairs of conspecific winged males and females. Since the song of D. buzzatii is the most divergent and all females of the D. buzzatii cluster derive from collections in localities where they coexist with D. buzzatii, we used playbacks of D. buzzatii song as heterospecific playbacks in the HP condition. In trials involving D. buzzatii females, we played back D. koepferae songs in the HP condition because lines of both species derive from the same locality. Experiment 2 aimed to explore the role of courtship song in sexual isolation between D. buzzatii and the other species of the cluster included in this study (Fig 2). To this end, we confronted unmanipulated females of D. borborema, D. antonietae or D. koepferae with wingless D. buzzatii males in two different acoustic conditions: with no playback (NP) and with conspecific song playback (CP). In this experiment we included two types of controls: a positive control (Ct+) identical to the control group of Experiment 1, using winged conspecific males, and a negative control (Ct-), using winged heterospecific D. buzzatii males, to corroborate the existence of precopulatory sexual isolation between the species involved. As in Experiment 1, in trials involving D. buzzatii females, we use wingless D. koepferae males as heterospecifics because lines of both species come from an area where both species coexist.
Experiments were carried out between 8:00 AM and 11:00 AM when flies showed highest mating activity, at 25 ± 1˚C. Flies were sexed under light CO 2 anesthesia within 6 h of eclosion, when wings were already extended. At this moment, since wingless males were required for some assays, both wings of some males were removed with a pair of microsurgical scissors from the base of the wing whereas other males remained unmanipulated (winged males). Thus, both winged and wingless males were exposed to light CO 2 anesthesia for the same time span. Both types of males and females were maintained separately for 5 to 7 days in food vials until their use in experiments.

Behavioral variables measured
To evaluate the effects of the different acoustic conditions, we measured two aspects of female mate choice: mate acceptance (MA) and copulation latency (CL). Mate acceptance was computed as a binary measure, 1 when copulation occurred and 0 otherwise. Copulation latency was measured as the time elapsed since the male began to follow the female until copulation. The observation period was limited to 10 minutes (= min) since female latency time started to be measured.
Measuring mating behavioral variables since males initiated courtship excludes variation due to male preferences, especially in Experiment 2 in which we assessed heterospecific pairs. Therefore, CL can be compared between conditions involving conspecific males (Ct+) and heterospecific males (Ct-, NP and CP) without bias. Also, it is important to take into account that D. buzzatii males (or D. koepferae when appropiate) are very promiscuous and, once courtship begins, they persist even when the female kicks, raises and lowers her abdomen, or move around.
We used copulation duration to define MA in heterospecific pairs on Experiment 2. Since copulation duration could be shortened in heterospecific pairs by processes involving postcopulatory preferences or discrimination [33][34][35], considering only prolonged copulations would underestimate the female response to the male precopulatory behavior. Accordingly, we recorded as acceptance when the female remained still for at least 10 seconds (= s) after the pair were in-copula without exhibiting rejection behavior (i.e., kicking or moving around), which minimized pseudo-copulations. We recorded copulation duration in all couples in Experiment 2 (see Table 1) and our criterion is discussed.
We assessed 30 pairs for each combination of species, isofemale line and acoustic condition in each experiment, except in the case of D. borborema in Experiment 2 in which two replicates were discarded since flies took too long to initiate courtship. All isofemale lines were used in Experiment 1; however, the long time required by the males to start courting in heterospecific pairs (whose effect our experimental design is intended to rule out) and our interest in recording the duration of copulation, prevented us to evaluate a large number of pairs. Thus, one line per species was used in Experiment 2.

Song recording and playback
Song playbacks used in this study were recorded individually from ten males for each combination of isofemale line and species with a SONY ICD-SX712 in an uncompressed Linear PCM wav file, and using a sampling rate of 44.1 kHz and 16-bit precision. Recordings, in the same channel, were carried out using the SONY ICD-SX712 built-in microphone in an acoustically isolated room at 25±0.3˚C (measured with a digital thermometer) controlled by a heating/cooling system. Mating arenas consisted of polymethyl methacrylate cylinders 0.5 mm height and 1.2 mm in internal diameter, the same diameter as the microphone grid. The floor of the chamber was the microphone itself and had a foam cap in the opposite extreme where flies were introduced. Males were exposed to females whose wings had been removed to avoid interferences during recording and songs were recorded until copulation. Before being used in playback experiments, songs were edited by eliminating the period of time elapsed before the male began to sing. We employed the software Sound Forge 9.0 (Sony Creative Software Inc., 2007) for song editing. Representative sonograms of the courtship songs of the five species recorded in this work are illustrated in Fig 3. For playback assays we used the same SONY ICD-SX712 equipment coupled to SONY MDR-E708 speakers. In these assays, mating arenas were 1.2 mm height and we replaced the microphone by a speaker. The ten songs recorded for each isofemale line were randomly used in each test by line (each song was used three times in each acoustic condition by line) and the playback started when the male began to follow the female. In HP condition, we chose 10 random songs of each species used as heterospecifics (i.e., D. buzzatii or D. koepferae when appropriate) and we used it in the manner described above.
The whole set of replicates by species was randomized into different assay sessions to avoid confounding effects.
Statistical analysis. Since both variables measured in this study describe two different aspects of mate choice [36], "what" is preferred (mate acceptance) and "by how much" (mating latency time), we decided to analyze them separately.
All models included acoustic condition as a fixed factor, and, for D. antonietae, D. buzzatii and D. koepferae in Experiment 1, isofemale line was also included as a random factor (only in these cases we used mixed models). For mixed models, we used a random slope model, which allows for the relationship between the acoustic conditions and the female responses to be Table 1 Ct+  257  191  1559  190  105  290  119  78  229  202  141  307   NP  ------------CP  356  89  1974  48  21  187  68  37 148 different for each isofemale line [37]. Since this model failed to converge in D. koepferae, we fitted a random slope with fixed intercept model in this species. Therefore, to test if mate acceptance differed among acoustic conditions, a generalized linear model (GLM) or a generalized linear mixed model (GLMM; lme4 package [38]) both with a binomial distribution of errors and a logit link was fitted for each species as appropriate. Tukey contrasts failed when we compared some acoustic conditions in which all females mated (mate acceptance status = 1) versus acoustic conditions in which all females refused to mate (mate acceptance status = 0); therefore, we changed the mate acceptance status of a single individual in some 'acoustic condition × isofemale line' groups.

D. antonietae D. borborema D. buzzatii D. koepferae
To investigate differences in copulation latency among acoustic conditions with more than 50 percent of mate acceptance, we used a linear model (LM) or a linear mixed model (LMM; lme4 package). The time variable was log-transformed to better meet the assumptions of homoscedasticity and normally distributed errors. Type III sums-of-squares (SS) was used to account for the unbalanced design ('car' package [39]).
Tukey single-step tests were carried out using the 'multcomp' package [40]. All statistical analyses were run in R software version 3.2.0 [41]. We used Kaplan-Meier plots (survival package, [42]) to graphically summarize all experimental results. The analysis of the data revealed a significant effect of the acoustic stimulus on mate acceptance (MA) in all species of the D. buzzatii cluster assayed (D. antonietae, D. borborema, D. buzzatii, and D. koepferae). In effect, MA was significantly reduced (relative to controls) when wingless males court in the absence of song playback (Fig 4A-4D, NP condition; S1 Table). This implies that wings are important for male mating success, but does not establish the importance of the courtship song. Although at least one female accepted the male in the absence of playback in all species, the reduction in MA was much more marked in D. borborema ( Fig  4B) and D. buzzatii (Fig 4C), intermediate in D. antonietae (Fig 4A), whereas D. koepferae females were the least affected by the absence of male wings (Fig 4D). The high acceptance of wingless males in the absence of playback by D. koepferae females was accompanied by a significant increase in the mean time to achieve copulation (NP: 87±106 s vs. Ct+: 25±43 s, Tukey test: Z = 6.77, P < 0.001). These results show that there are interspecific differences in relation to the effect that the absence of male wings exerts on female mate choice. The presence or absence of male wings determined the acceptance or rejection of males in most D. antonietae, D. borborema and D. buzzatii females, whereas male wings in D. koepferae primarily reduced the necessary time to be accepted by females.

Experiment 1: Significance of the courtship song in conspecific pairs
To determine the role of courtship song on female mate choice, we tested wingless males either with the conspecific or a heterospecific song playback.
Similar to the results obtained in assays without any playback (NP condition), MA was significantly reduced (relative to the Ct+ group) when wingless males courted with the heterospecific song playback (Fig 4A-4D; HP condition; S1 Table). However, the conspecific song playback (CP condition) restored female mating behavior in terms of both mate acceptance and copulation latency in all species (Fig 4A-4D; S1 and S2 Tables). These results indicate that conspecific song playback is enough to compensate for the absence of wings. Therefore, the reduction in MA and the increased in CL in assays involving wingless males with no playback can be explained only by the absence of the conspecific song. Moreover, no significant differences were found between heterospecific song playback (HP) and no playback (NP) conditions in D. antonietae, D. borborema and D. buzzatii (Fig 4A-4C; S1 Table). These results suggest  Consequently, only in the case of D. koepferae we found evidence that support a reduction in MA by the heterospecific song playback.
Finally, to explore whether the patterns observed in species of the D. buzzatii cluster regarding the importance of acoustic stimulus in female mate choice are conserved in the D. buzzatii complex or, on the contrary, there is evidence of greater diversity, we included D. venezolana, a species of the closely related D. martensis cluster. Our results revealed that the different acoustic conditions assessed in D. venezolana (Fig 4E)  To demonstrate the existence of precopulatory sexual isolation we exposed females of each species of the D. buzzatii cluster to winged D. buzzatii males (or winged D. koepferae males in assays involving D. buzzatii females). In all species, MA was completely suppressed in these couples (Ct-condition; Fig 5). Therefore, we can conclude that all heterospecific pairs exhibit strong precopulatory sexual isolation in the time interval considered. Furthermore, MA was also completely abolished when wingless heterospecific males were used with no playback (NP condition; Fig 5). These results suggest that heterospecific song is not responsible for the rejection of winged heterospecific males by females in Ct-condition.
Playing back the conspecific song restored both mate acceptance (CP vs Ct+ conditions, Tukey test, Z = 0.46, P = 0.97) and copulation latency (CP vs Ct+ conditions, LM: F 1,53 = 0.99, P = 0.32) to levels comparable to the Ct+ group in pairs involving D. buzzatii females and wingless D. koepferae males (Fig 5C). Therefore, precopulatory sexual isolation is completely eliminated in these pairs by playing back the conspecific song.
Females of D. antonietae also accepted wingless D. buzzatii males when conspecific song was played back without significant differences in mate acceptance as compared to the Ct+ group (CP vs Ct+ conditions, Tukey test: Z = -0.581, P = 0.937). However, there was a significant effect on copulation latency since time to copulation was longer in the CP condition than in the Ct+ group (CP: 27±30 s vs. Ct+: 12±8 s, LM: F 1, 56 = 8.9841, P < 0.01). This implies that, although conspecific song playback was sufficient to breakdown precopulatory sexual isolation between D. antonietae females and wingless D. buzzatii males, other non-acoustic cues seemed to influence female mate choice by lengthening the time females took to accept wingless heterospecific males.
In trials involving D. borborema females and wingless D. buzzatii males, playing back the conspecific song failed to fully restore mate acceptance relative to the Ct+ group [CP: 0.61 vs Ct+: 0.93, Z = -2.57, P < 0.05(0.0477)]. Nevertheless, more than a half of the females accepted wingless D. buzzatii males when conspecific song was played back. In this case, although conspecific song playback reduced precopulatory sexual isolation between D. borborema females and wingless D. buzzatii males, it was not enough to eliminate it completely.
In contrast, MA was strongly reduced in D. koepferae regardless of the acoustic condition when females were exposed to D. buzzatii males (NP, CP and Ct-conditions in Fig 5). The fact that winged and wingless with no playback D. buzzatii males were not accepted by D. koepferae females (Fig 5D, Ct-and NP conditions, respectively) suggests that, despite heterospecific song playback has the potential to reduce AM in D. koepferae (Fig 4D, HP condition), the rejection of wingless D. buzzatii males would be a consequence of non acoustic cues (Fig 5D, NP condition). In addition, females also did not accept wingless D. buzzatii males with the conspecific song playback (Fig 5D, CP condition). Therefore, we can conclude that the conspecific song cannot eliminate or reduce precopulatory sexual isolation between D. koepferae females and D. buzzatii males.

Experiment 2: Copulation duration
There was considerable variation in the duration of conspecific copula among Drosophila species assessed (Table 1). In heterospecific pairs, the minimum duration of copulation registered was twice the time considered as a criterion to define male acceptance (Table 1). In assays using D. antonietae females, the average length of heterospecific copulations with conspecific playback (CP) exceeded those of the Ct+ group (Table 1). On the contrary, in trials with females of D. borborema and D. buzzatii, the average length of copulation in pairs involving heterospecific males (CP) was shorter than in the Ct+ group (Table 1).

Discussion
Male courtship songs have undergone a spectacular diversification in the Drosophila buzzatii cluster ( [24], Fig 3). Accordingly, it has been suggested that sexual selection has played a significant role in promoting rapid diversification, reproductive isolation and speciation [24]. However, there is evidence that the courtship song in Drosophila is not always used in female mate choice decisions, as in the case of the Hawaiian D. heteroneura [15], nor plays the same role when it is taken into account (reviewed in [43] and in [27]). In this vein, our study investigated the role that the courtship song plays in female mate choice and in sexual isolation by evaluating two different aspects: which acoustic conditions favor the acceptance of males (MA) and how fast are they accepted (CL).

Importance of the courtship song in the D. buzzatii cluster
Our study revealed that acoustic cues are used by females of the studied species of the D. buzzatii cluster as a criterion for male acceptance or influence the speed with which males are chosen. Therefore, our results support the hypothesis that sexual selection has played an important role in the diversification of courtship songs in this group of closely related species.
The courtship song of D. buzzatii is the most divergent within the D. buzzatii cluster. It is the only species that has "doublets of pulses" in the primary or A song ( [11]; Fig 20C in [24]; P.P. Iglesias pers obs) (Fig 6) and, in addition, the IPIs of these pulses clearly differ from those of the secondary or B song ( [24]; P.P. Iglesias pers obs). These particular features of the D. buzzatii courtship song are an inherent part of the species' acoustic signal at least in Argentinean and Brazilian populations (Fig 20C in [24]; P.P. Iglesias & E. Hasson unpublished results). Therefore, these characteristics have the potential to make the D. buzzatii courtship song easily discriminable by females of the remaining species of the cluster and also easily recognizable by conspecific females.
The fact that male acceptance was reduced when wingless conspecific males courted in silence (NP condition) or with the heterospecific song playback (HP condition; Fig 4A-4D) and that the conspecific song playback restored MA and CL to levels observed in the Ct+ group, suggests that there are song parameters that stimulate females in a species-specific manner. In this regard, three song parameters have been reported as cues used by Drosophila females for species recognition or sexual selection. IPI is the most widely studied song parameter and has been reported as relevant for species recognition (reviewed in [27]). In addition, studies in D. montana showed that pulse duration and carrier frequency are also predictors of mating success [44,45]. In this respect, Oliveira et al. [24] found significant overlap for these parameters among species of the D. buzzatii cluster and suggested that females may use more than one song parameter in mate recognition. Furthermore, it is clear that D. koepferae females, unlike the other species of the cluster, discriminate conspecific males with the heterospecific song playback more intensely than conspecific males with no playback (Fig 4D; HP and NP conditions).
Taken together, these findings highlight females of D. koepferae as the most different in terms of the use of the courtship song in mate choice. Females accepted males when the playback was conspecific, intensely reduced male acceptance when playback was heterospecific and exhibited an intermediate response when courtship occurred in the absence of acoustic stimulus (Fig 4D). Since high MA in the absence of playback was accompanied by an increase in CL, the difference in MA between conditions with no playback and with conspecific song playback may be explained by the decrease in the courtship time necessary for males to be accepted by females, which causes an increase in the number of males accepted in the time interval defined in this experiment (i.e., 10 min). This suggests that D. koepferae females use other cues (e.g., hydrocarbons) to discriminate compatible males and song parameters to rank compatible males. Thus, it may be hypothesized that the courtship song alters initial male attractiveness, determined by other cues (as can be seen in the NP condition), and enhances or reduces male mating success as was also reported for D. biauraria and D. sechellia [18,19].
In summary, our results indicate that the importance and the role of conspecific and heterospecific songs differ among species. In D. koepferae, the evidence suggests that the heterospecific song plays an important role in reducing the number of non compatible matings and that conspecific song is not particularly necessary to achieve copulation, though it reduces the courting time needed for males to be accepted (Fig 4). In the remaining species, the presence or absence of the conspecific song determines male acceptance in most of the females, whereas the heterospecific song does not seem to play a critical role in mate choice because we did not find significant differences between conditions with no playback and with the heterospecific song playback (S1 Table). However, it is important to note that at least one female accepted the conspecific male in the absence of the conspecific song in D. antonietae, D. borborema and D. buzzatii. These observations suggest that there is intraspecific variation among females in relation to the sensory modalities used in mate choice or the degree of female choosiness [4], and that variation is greater in D. antonietae (Fig 4A-4C). Nevertheless, it is worth mentioning that male acceptance always implied a longer CL in the absence of the conspecific song playback (Fig 4A-4C).
In addition, this study provides evidence that the courtship song emitted either by wing vibration or through a speaker is equivalently processed by the female nervous system. This finding contrasts with results in D. melanogaster, in which the reproduction of natural or artificially generated conspecific songs consistently reduces MA relative to controls [17,[46][47][48][49]. To account for these observations, it has been proposed that wing vibration may also contribute to the spreading of pheromones that would stimulate female receptivity [48]. However, our results show that wings do not provide or facilitate any visual or chemical signal in these species.

Importance of the courtship song in the D. buzzatii complex
In order to explore the generality of these results outside the D. buzzatii cluster, we included D. venezolana as representative of the closely related D. martensis cluster. Females of this species accepted conspecific males independently of the acoustic condition (Fig 4E), suggesting that the acoustic stimulus is not relevant during courtship since we did not even find differences in CL. In this sense, our results point towards the idea that other signals influence mate choice in D. venezolana females.
Despite the fact that we only studied a single species outside the D. buzzatii cluster, the latter results show that the use of courtship song in female mate choice is not a conserved feature in the D. buzzatii complex. The study of more species of the D. martensis cluster is necessary to establish whether the loss of the use of courtship song in female mate choice is a characteristic shared by all or some species of the D. martensis cluster or it is an autapomorphy of D. venezolana.
Thus, the evolution of female preferences in the D. buzzatii complex (composed of the D. buzzatii, D. martensis and D. stalkeri clusters) involved at least one change in the sensory channel used by females to sense males. Females switched from sensing the quality and/or compatibility of males using non-acoustic cues (e.g., hydrocarbons) to using song parameters, or vice versa. However, to accurately define the branch of the phylogenetic tree in which this evolutionary change arose, and the polarity (direction) of this transformation, it is necessary to study the role of courtship song in female mate choice in more representatives of these clusters and interpret the results in a well sampled phylogenetic hypothesis. In this regard, the D. buzzatii complex may provide a model system for understanding the reduction or loss of female preferences in species in which the male trait (i.e., courtship song) is still present [5].

Courtship song and its role in sexual isolation
Drosophila buzzatii stands out as the most divergent species within the D. buzzatii cluster, not only due to the specific features of its courtship song, but also in ecological, morphological and biogeographical aspects. Given that D. buzzatii is also the only species that co-occurs with all other species in the cluster (Fig 1), differentiation in courtship songs, and/or in the sensory channels used by females for mate choice, have the potential to reduce or avoid incompatible matings in sympatry. Using heterospecific pairs, we evaluated the role of courtship song in precopulatory sexual isolation between D. buzzatii and the remaining species of the cluster.
On one hand, we found that females exhibited strong precopulatory sexual isolation when they were exposed to winged heterospecific males (Fig 5A-5D, Ct-acoustic condition) and that isolation was maintained when using heterospecific males without wings, that is, that cannot produce the heterospecific courtship song (Fig 5A-5D, NP condition). Accordingly, we can rule out that the heterospecific song plays a primordial role in sexual isolation. The only species capable of discriminating against the heterospecific song was D. koepferae (Fig 4, HP  condition). However, 40% of females copulated when the D. buzzatii song was reproduced. Our findings suggest that the ability of most D. koepferae females to discriminate the heterospecific song when exposed to conspecific males (Fig 4, HP condition) would reinforce precopulatory sexual isolation, but it does not determine it, since isolation was not reduced by eliminating the heterospecific courtship song when using heterospecific males (Fig 5, NP  condition).
On the other hand, we found that the importance of the conspecific song in sexual isolation varied among species. Females of D. antonietae, D. borborema and D. buzzatii accepted wingless heterospecific males when we played back the female conspecific song (Fig 5A-5C; CP condition). In contrast, females of D. koepferae rejected D. buzzatii males in all acoustic conditions ( Fig 5D). However, conspecific song playback did not have the same effect among females of the species that could be deceived and mated with the heterospecific male. Playing back the conspecific song was sufficient to completely break down precopulatory sexual isolation between D. buzzatii females and D. koepferae males since no significant differences were found in MA and CL when compared to the Ct+ group (Fig 5C, CP and Ct+ conditions). Courtship song as the main signal responsible for sexual isolation between sympatric species has already been reported in D. montana [50] and in the D. athabasca species complex [51]. Nevertheless, our results provide indirect evidence that other cues affect MA and CL in D. antonietae and D. borborema. Females of D. antonietae in the CP condition did not differ from the Ct+ group in MA; however, they took longer to accept D. buzzatii males than conspecific males (S2 Table; conditions CP and Ct+, respectively). Therefore, we can infer that another type of signal transmitted by D. buzzatii males through a sensory channel different from the acoustic one is less attractive for D. antonietae females than those emitted by conspecific males. In D. borborema assays, on the contrary, MA decreased significantly (Fig 5B; CP condition), yet a large proportion of females were deceived and accepted the heterospecific male. This suggests that other signals are also influencing mate choice in D. borborema. Sexual signaling in Drosophila takes place through visual, auditory and chemosensory cues [6]. Since the species studied in this paper are cryptic [28,52] and also courtship occurs behind the female (P.P. Iglesias pers obs), males cannot transmit visual information [6]. Thus, we expect that differences in hydrocarbons (volatile and/or not volatile) may explain the observed differences in CL and MA. Hydrocarbons have been shown to play an important role in mate choice in many species of several groups of Drosophila such as the D. melanogaster complex [53], the D. serrata species group [54], the D. subquinaria complex [55] and the D. yakuba species group [56]. However, we cannot rule out the possibility that D. buzzatii males may differ in their courtship intensity toward D. borborema females, since D. buzzatii males not only took longer to start courtship when were exposed to these females (P.P. Iglesias pers obs) but also some males never started to court.
It is important to note that only D. buzzatii and D. koepferae lines used in this study come from the same area and, therefore, are truly sympatric populations. On the contrary, females of D. antonietae and D. borborema derive from populations sympatric with D. buzzatii, but allopatric with respect to the D. buzzatii males used as heterospecifics in this study. There are no previous studies investigating whether courtship song varies among populations in the D. buzzatii cluster species and whether character displacement (reinforcement) has occurred. However, if character displacement in courtship song has occurred in the areas of sympatry we expect female discrimination against sympatric heterospecific songs to be more drastic than against allopatric ones. Since MA was almost completely reduced in the assays with D. antonietae and D. borborema females (Fig 3, HP condition) we do not expect our broad conclusions to be substantially altered by the use of sympatric D. buzzatii males in Experiment 1. However, the use of sympatric D. buzzatii males in Experiment 2 might reduce the effect of the conspecific song playback in the assays with D. antonietae and D. borborema females if there is character displacement in signals perceived by sensory modalities other than acoustics. Nevertheless, our results from Experiment 1 suggest that non acoustic cues would not be as important for D. antonietae and D. borborema females as they are for D. koepferae females given the strong reduction in MA observed in assays with D. antonietae and D. borborema females in the no playback condition relative to what was observed in D. koepferae (Fig 4A, 4B and 4D; NP condition).

Copulation duration
To determine the acceptance of the male in heterospecific pairs of Experiment 2, we used the criterion of considering as acceptance the cases in which the pair remained in the copulation position and the female did no exhibit rejection behavior for at least 10 s. Since D. buzzatii has the most divergent morphology of the intromittent organ (aedeagus) in comparison to the other species of the cluster [28], one could speculate that the ability to remain in copulation position could be prevented in heterospecific pairs by some kind of mechanical incompatibility as predicted by the key and lock hypothesis [57]. However, according to this hypothesis, reproductive isolation could operate through two different mechanisms (reviewed in [58]). The first mechanism is called classical or structural lock-and-key and corresponds to the existence of mechanical incompatibilities that reduce or prevent copulation and/or insemination [59,60]. The second mechanism is called sensory lock-and-key and proposes that morphological differences are perceived by one or both sexes and evoke behavioral or physiological responses that result in the premature termination of copulation or a reduction of postcopulatory reproductive fitness [61]. An increasing body of experimental evidence does not support the structural mechanism. However, it does not rule out that the sensory mechanism can be widely extended, limiting gene flow among species (reviewed in [58]).
Our results show that there were significant differences in MA among acoustic conditions considering the 10 s time interval in the assays involving D. antonietae, D. borborema and D. buzzatii females, suggesting that the ability to remain in copulation position is not prevented in these heterospecific pairs (Fig 5). Only in the assays involving D. koepferae females we did not observe pairs in copula (Fig 5). The aedeagus size and shape of D. koepferae males is similar to that of D. antonietae males [28]. However, all D. antonietae females did copulate with wingless D. buzzatii males in the CP condition. In fact, when D. antonietae females were used, the minimum time of permanence in copulation position was 89 s (Table 1) and the mean time of copulation duration was even higher than in the corresponding Ct + pairs (Table 1). This provides indirect evidence that the divergent size and shape of the aedeagus of D. buzzatii may not be responsible for the results found in D. koepferae trials. The lowest time in copula was recorded in heterospecific pairs involving D. borborema females, even though males of D. borborema have intermediate aedeagus morphology between the aedeagus of D. buzzatii and those of D. antonietae and D. koepferae [28]. Nevertheless, the minimum time recorded doubled the time used as acceptance criterion. These results suggest that differences in aedeagus size and shape are not fair predictors of both MA and duration of copulation in the heterospecific pairs assessed.
Our findings do not support the mechanical lock and key mechanism hypothesis but do not contradict the sensory lock-and-key mechanism expectations. Therefore, these results indicate that the criterion used to define male acceptance in heterospecific pairs of Experiment 2 is appropriate to avoid confusing the effect of precopulatory (e.g., courtship songs) and postcopulatory (e.g., aedeagus morphology) traits in female mate choice.

Conclusions
Our study revealed that acoustic cues are used by females of the studied species of the D. buzzatii cluster as a criterion for male acceptance or influence the time females take to accept the male. Since our study only included four species of the D. buzzatii cluster (D. antonietae, D. borborema, D. buzzatii and D. koepferae), we cannot ensure that our observations can be extended to the other three species of the cluster (i.e., D. gouveai, D. serido and D. seriema). Furthermore, the fact that this characteristic is not shared by D. venezolana, a species of the D. martensis cluster that we use as external group, indicates that there is a higher diversity in the sensory modalities used by females during courtship in the D. buzzatii complex. We also found that the studied species of the D. buzzatii cluster differ in the role that conspecific and heterospecific songs have in female mate choice and in sexual isolation in relation to the only semi-cosmopolitan species, D. buzzatii. Thus, our findings indicate that divergence in female preferences for courtship songs and the differential use of alternative sensory modalities during courtship are drivers of acoustic divergence in the D. buzzatii cluster. In this sense, the results suggest that courtship songs may have played a predominant role in the evolution of this species cluster as it successfully diversified in the xeric areas of South America. However, more comprehensive studies would help to shed light on whether these particular signals were more important early (when there is little genetic and no morphological divergence) or later (during secondary contact in sympatry) in the speciation process. The well-known ecology of some of these species provides the opportunity to disentangle the relationship among courtship songs, hydrocarbons and host cacti in the evolution of sexual isolation in a group of species with varying divergence times.
Finally, our study also revealed intraspecific variation in female preferences or female choosiness since a small number of females behaved differently in some acoustic conditions. Therefore, future studies should focus on investigating the causes that underlie variation in female behavior and on studying patterns of intraspecific variation in each species.
Supporting information S1