Chemical signals and social structures strengthen sexual isolation in Drosophila pseudoobscura

Jezovit, Jacob A.; Levine, Joel D.

doi:10.1038/s42003-025-07483-4

Download PDF

Article
Open access
Published: 17 January 2025

Chemical signals and social structures strengthen sexual isolation in Drosophila pseudoobscura

Communications Biology volume 8, Article number: 76 (2025) Cite this article

1468 Accesses
1 Altmetric
Metrics details

Subjects

Abstract

Species that coexist in hybrid zones sexually isolate through reproductive character displacement, a mechanism that favours divergence between species. In Drosophila, behavioural and physiological traits discourage heterospecific mating between species. Recently, social network analysis revealed flies produce strain-specific and species-specific social structures. A gene, degrees of kevin bacon (dokb) has also been discovered that accounts for differences in social structures between flies. Why differences in social structures exist between drosophilids is currently unknown. Here we show through an experimental evolution study that six generations of selection in experimental sympatry led to the divergence of social structures measured in Drosophila pseudoobscura and Drosophila persimilis flies. We found that the frequency of hybrid offspring decreased within a few generations, suggesting social structures are associated with the sexual isolation of species. We also report increased species’ differences in the concentration of the cuticular hydrocarbon 5, 9-pentacosadiene after six generations of selection. The mean concentration of this compound converged in female flies of both species and diverged in male flies of both species, suggesting a quantitative link between increased sexual dimorphism and sexual isolation. Our results suggest that chemical signals, together with social structures, increase the sexual isolation between species in hybrid zones.

Phenotypic disruption of cuticular hydrocarbon production in hybrids between sympatric species of Hawaiian picture-wing Drosophila

Article Open access 22 March 2022

Large-scale characterization of sex pheromone communication systems in Drosophila

Article Open access 06 July 2021

Ancestral neural circuits potentiate the origin of a female sexual behavior in Drosophila

Article Open access 28 October 2024

Introduction

Reproductive character displacement refers to phenotypic differences that are elevated between species which coexist in sympatry and reduced where the same species are geographically isolated in allopatry¹. For species living in sympatry, it is expected that natural selection favours the evolution of prezygotic barriers to strengthen sexual isolation and reduce the birth of sterile hybrid offspring, a mechanism known as reinforcement. This phenomenon is well studied in the sister species Drosophila pseudoobscura and Drosophila persimilis. These two species are thought to share a recent speciation event due to their indistinguishable morphology, their geographic overlap, and their ability to produce sterile hybrid offspring^2,3. A precise genetic mechanism to their divergence is not fully understood, although large chromosomal inversions spanning the X chromosome and the second autosome distinguishes these species⁴. Comparative studies highlight that D. pseudoobscura and D. persimilis display large differences in their cuticular hydrocarbon (CHC) profile⁵, in their courtship song patterns⁶, and in their sperm quality⁷, especially in sympatric regions, highlighting the relationship between reproductive character displacement and prezygotic barriers produced at the physiological and behavioural levels.

Several experimental evolution approaches have been applied to study the sexual isolation between D. pseudoobscura and D. persimilis. One approach was the “kill all hybrids” design where two species are mixed in a common environment and strong selection is applied against hybrid offspring. Consistent evidence was gathered decades ago that mixing these species in population cages and discarding hybrid offspring leads to increased sexual isolation within a few generations^8,9. Similar experiments using other sister species pairs found support for the rapid evolution of sexual isolation, including when gene flow was enabled between populations, and when selection against hybrids was relaxed^10,11. Increased sexual isolation was also observed in experiments that maintained different Drosophila melanogaster mutants for generations in experimental sympatry^12,13. Although the effects of reinforcement have been replicated in artificial sympatric environments, whether reinforcement acts on natural populations remains unresolved. Some studies have reported increased sexual isolation between species captured from sympatry^14,15, while others have reported no large differences in the sexual isolation measured between allopatric and sympatric flies^16,17.

Although studies on D. pseudoobscura and D. persimilis have advanced ideas on sexual isolation, these studies have not examined the influence of social group structure. Most of our knowledge about the social dynamics of flies stems from the literature on D. melanogaster¹⁸. This work has demonstrated that flies aggregate on food sources in large densities¹⁹; that flies collectively communicate using chemical signals²⁰, acoustic signals²¹, and haptic signals²²; that males attack one another as seen in aggressive displays²³; and that females lay eggs in close proximity²⁴. These behaviours point out the intricacies of the Drosophila social environment where group composition can affect individual gene expression, CHC profiles, and mating frequency in a complex feedback loop²⁵. For example, drosophilid species, including D. pseudoobscura, alter their courtship techniques in the face of rivals, highlighting how life history and the social environment of flies affects reproductive behaviours^26,27. To better study the social environments of flies, behavioural assays have been developed that involve filming social groups, tracking videos through machine vision systems, and using tracking data to study spatiotemporal aspects of social interactions^{28,29,30,31,32,33,34}. Social network analysis in D. melanogaster revealed that flies form social structures that reflect allelic differences in a gene called degrees of kevin bacon³⁵. These social networks are dependent on both visual and chemical communication between group members^28,32. Quantifying these structures across various drosophilids further revealed that closely related species produce divergent networks³⁶, suggesting social structures show patterns of reproductive character displacement. Given the genetic link to Drosophila social networks, we were curious if social structures maintain sexual isolation by segregating species into different social strata. Presumably, this segregation may reduce encounters between heterospecifics, lowering the probability of producing hybrid offspring. This process can be captured by measures, such as clustering coefficient and betweenness centrality, which detect the cohesion of social interactions within the network³⁷. To our knowledge, this idea has never been experimentally tested.

Here, we present the results of an experimental evolution study designed to test the hypothesis that social structures, measured through social networks, diverge over time to enhance the sexual isolation between species. This experiment was conducted using D. pseudoobscura and D. persimilis flies. We applied a “kill all hybrids” approach where replicate populations of flies in experimental sympatry and experimental allopatry were maintained for six generations. Hybrid offspring were counted in all the sympatric populations as a proxy for the strength of sexual isolation. We hypothesized that in early generations, the sympatric flies would display evidence of weak sexual isolation. Over subsequent generations within experimental sympatry, sexual isolation would strengthen, leading to a decrease in the number of hybrid offspring. If social networks are associated with sexual isolation, we expected social networks to differ between groups of flies raised in sympatry compared to groups of flies raised in allopatry after six generations of selection. We predicted CHCs would play a role connecting sexual isolation to social structures, given their sensitivity to changes in social contexts²⁵. If CHCs serve this role, we also expected CHCs to diverge between sympatric and allopatric flies.

Results

Sexual isolation in sympatry

For six generations, we counted all progeny from three sympatric populations and calculated the frequency of hybrid offspring (Fig. 1A). Hybrid offspring were born at a relatively high mean frequency at the first generation (9.71 ± 1.26%), followed by a net decrease at the sixth generation (5.89 ± 1.19). We found that generation was a significant predictor variable for this negative relationship (Wald χ² = 7.0, df = 1, p = 0.008), offering support for the observation that hybrid frequency significantly decreased over time. This relationship was consistent in all three populations when examined separately, although hybrid frequency spiked upward in the middle generations in two populations (Supplementary Fig. 1). This data suggests sexual isolation increases between sympatric D. pseudoobscura and D. persimilis within six generations.

**Fig. 1: Frequency of offspring pooled across three populations for six generations.**

We also monitored the frequency of D. pseudoobscura and D. persimilis offspring across the generations (Fig. 1B). On average the mean frequency of D. pseudoobscura and D. persimilis offspring was equal for the first generation (Dpse: 45.03 ± 2.98; Dper: 44.66 ± 3.17). By the final generation, D. pseudoobscura offspring displayed a net increase in their mean frequency (56.92 ± 5.37%) while D. persimilis offspring displayed a net decrease (37.66 ± 5.43). Generation was found to be a significant predictor variable for the positive relationship in D. pseudoobscura offspring (Wald χ² = 68.6, df = 1, p < 0.0001), and for the negative relationship in D. persimilis offspring (Wald χ² = 59.7, df = 1, p < 0.0001). These trends were consistent in all three populations when examined separately (Supplementary Fig. 2). Together, this data suggests that the increased sexual isolation we observed (Fig. 1A) is associated with increased conspecific mating of D. pseudoobscura flies. However, this is speculative, and we cannot rule out inbreeding depression and differences in fecundity as alternative explanations for the decrease in D. persimilis offspring over time.

No strong evidence of sexual isolation within allopatric flies

To examine whether the reduction in hybrid offspring we observed in experimental sympatry was a result of inbreeding depression or drift, we compared the sexual isolation within mixed-species and mixed-sex groups of flies that were all reared either in sympatry or allopatry. These mixed groups of flies were used in all behavioural experiments (see below), and allowed us to test the hypothesis that allopatric flies do not develop sexual isolation. Upon the completion of each behavioural trial, we transferred all flies from their arenas into vials with standard cornmeal media and left them at 16 °C for 3–5 days to copulate and lay eggs. All adults were removed, and all offspring were counted to calculate the frequency of hybrid offspring. We found that 36% of the vials with allopatric adults from generation 1 produced hybrid offspring and this remained consistent at generations 4–6 (Fig. 2A). The proportion of vials with sympatric adults that produced hybrid offspring was like the allopatric vials at generation 1, but generations 4-6 displayed a notable decrease (Fig. 2A, Supplementary Fig. 3), although the differences were not significant following chi-square tests (Generation 4: p = 0.06; Generation 6: p = 0.22). Out of the small subset of vials that produced hybrids each generation, we found no significant differences in the frequency of hybrid offspring between the treatments (p = 0.65, aligned rank transformation) and between the generations (p = 0.14, aligned rank transformation), although the variance decreased at later generations (Fig. 2B). While these results suggest drift may play a role in decreasing the sexual isolation of flies in both environments over time, the sympatric flies still appear to develop stronger sexual isolation compared to allopatric flies.

**Fig. 2: Frequency of offspring from groups of 12 allopatric flies (blue) and sympatric flies (red).**

We also calculated the frequency of D. pseudoobscura and D. persimilis offspring across the generations from both treatments. The results were consistent with the offspring trends seen in the sympatric bottles (see Fig. 1B). Drosophila pseudoobscura offspring descending from sympatric flies significantly increased across the generations relative to the offspring from allopatric flies (generation effect: p < 0.0001), treatment effect: p < 0.0001, two-way ANOVA). Drosophila persimilis offspring descending from sympatric flies significantly decreased across the generations relative to offspring from allopatric flies (generation effect: p < 0.0001), treatment effect: p < 0.0001, two-way ANOVA). In fact, by generation 6 the frequency of D. pseudoobscura and D. persimilis was divergent between allopatric and sympatric flies (Fig. 2C, D), and this was consistent in two out of three populations (Supplementary Fig. 4). Overall, the offspring descending from vials of 12 flies complements the offspring frequency data from the bottles of 80 flies, and we suspect the lower incidence of hybrid offspring from these vials (Fig. 2, Supplementary Fig. 3) could arise from the small group sizes. Like in the bottles (Fig. 1B) we observe D. pseudoobscura offspring increasing in experimental sympatry compared to D. persimilis offspring.

Plasticity of cuticular hydrocarbons in Drosophila pseudoobscura

We predicted that CHCs would provide insight into strengthened sexual isolation, considering these compounds vary between species and support reproductive barriers^{38,39,40,41,42}. To test this, we extracted CHCs from individual sympatric and allopatric flies for both species and sex. The concentration of 5,9-pentacosadiene we measured in D. pseudoobscura flies was far greater than our expectations, given that this compound is characteristic of D. persimilis flies^5,39 (Supplementary Fig. 5-7). To confirm whether this arose from the cool rearing temperature throughout this study, considering phenotypic plasticity has been reported in the profile of D. pseudoobscura⁴³, CHCs were also extracted from inbred D. pseudoobscura and D. persimilis flies after the experimental evolution study was completed. A total of 13 chromatogram peaks were compared between the species (Supplementary Figs. 5–7).

First, we hypothesized that the rearing temperature would affect the whole CHC profile of both species. We predicted the profiles of D. pseudoobscura and D. persimilis flies would be similar at 16 °C since this temperature enhances the birth of hybrid offspring⁹. To address this, we conducted a principal component analysis where the concentration of all 13 compounds we measured were fit into two principal components which accounted for 73.77% of the variation within the data. Visualizing these two principal components revealed that the rearing temperature affects the whole CHC profile of male and female flies for both species. However, the degree of plasticity differs between the species since the two D. pseudoobscura treatments cluster further apart than the D. persimilis treatments for both sexes (Fig. 3A, B). The principal component analysis also revealed that at 16 °C, the CHC profile for male D. pseudoobscura closely resembled the profile for D. persimilis males (Fig. 3A). This finding explains the heightened interbreeding at 16 °C⁹—male flies are less distinguishable to female flies, likely confusing female mate choice.

**Fig. 3: Phenotypic plasticity measured in cuticular hydrocarbons extracted from wild-type *D. pseudoobscura* and *D. persimilis flies*.**

Next, we analyzed two compounds: 5,9-pentacosadiene (5,9-PD) and 5,9-heptacosadiene (5,9-HD). These compounds are the most concentrated in the profiles of D. persimilis and D. pseudoobscura, respectively^5,39. 5,9-PD and 5,9-HD were also large contributors to the variance within the first principal component (Supplementary Table 1). In male and female flies, we found significant species × temperature interaction effects for 5,9-PD (male: p < 0.0001, female: p = 0.002, two-way ANOVA) and 5,9-HD (male: p < 0.0001, female: p < 0.0001, two-way ANOVA). A trend was observed in D. pseudoobscura male (p < 0.0001, two-way ANOVA) and female flies (p < 0.0001, two-way ANOVA) where the mean 5,9-PD concentration was high at 16 °C (male: 1200.78 ± 23.92 ng; female: 371.92 ± 63.46 ng) and low at 24 °C (male: 157.11 ± 12.87 ng; female: 64.32 ± 1.92 ng) (Fig. 3C). The opposite trend was seen in the mean concentration of 5,9-HD for D. pseudoobscura male and female flies. Flies reared at 16 °C displayed a low mean concentration of 5,9-HD (male: 231.47 ± 10.88 ng; female: 925.88 ± 52.87 ng), while flies reared at 24 °C displayed a high mean concentration of 5,9-HD (male: 1647.52 ± 50.41 ng; female: 1906.10 ± 48.57 ng) (Fig. 3D). For the average concentration of 5,9-PD, male D. pseudoobscura flies reared at 16 °C resembled D. persimilis flies, while species differences remained distinct for 5,9-HD at both temperatures. Rearing temperature had no effect on the concentration of 5,9-PD and 5,9-HD, on average, in male and female D. persimilis flies, suggesting D. pseudoobscura alone possesses temperature-dependent plasticity for these two compounds. This finding suggests that 5,9-PD is a strong candidate for a single CHC compound that has profound effects on the sexual isolation between these species. At cool temperatures male flies of both species may be indistinguishable to females, increasing heterospecific mating.

Divergence of cuticular hydrocarbons between species across the generations

Since rearing both species at 16 °C leads to a similar CHC profile for male flies, we hypothesized the CHC profiles would shift in the species as sexual isolation increases within sympatry. To test this hypothesis, a principal component analysis was also conducted on the CHC data extracted from both species raised in sympatry and allopatry across the generations. We found that differences in CHC profiles were minimal between sympatric and allopatric flies across the generations (Fig. 4A, B). In male flies, species differences in CHC profiles increased by generation 6, while in female flies, they decreased. Although this offers little support to our hypothesis that CHC profiles change in sympatry, species’ profiles still changed over time.

**Fig. 4: Shifts in CHC profiles in *D. pseudoobscura* and *D. persimilis* flies across generations of selection.**

The high concentration of 5,9-PD, on average, was consistently measured within the male and female D. pseudoobscura flies across the generations. In fact, the concentration of 5,9-PD within the male D. pseudoobscura flies was higher, on average, compared to that of D. persimilis. This may be the result from using outbred D. pseudoobscura flies, while using inbred D. persimilis flies in this experiment (see “Methods”). Given the high concentration of this compound in both species at 16 °C, and a report of this compound being highly expressed in male flies adapted to polygamy³⁹, we suspected 5,9-PD would shift in concentration after generations of selection within sympatry and restore the sexual isolation between the species. Analyzing the concentration of this compound revealed significant species differences throughout six generations in both environments (Fig. 5A, B). In female sympatric flies, both species began to converge in their 5,9-PD concentration after four generations of selection and this trend is seen in each population (Supplementary Fig. 8). After six generations of selection, there was little difference in the mean concentration of 5,9-PD between sympatric female flies (Dpse: 889.20 ± 55.28 ng; Dper: 1011 ± 44.39 ng), resulting in no significant differences between species (p = 0.23, Tukey–Kramer multiple comparison test) (Fig. 5B). The opposite effect was observed in the male flies for each population (Supplementary Fig. 9). Combining the population data together displayed a consensus where male D. pseudoobscura and D. persimilis display significant differences in the concentration of 5,9-PD in both environments (p < 0.0001 at all generations, two-way ANOVA). However, the mean difference in the 5,9-PD concentration was greater between species raised in sympatry compared to those raised in allopatry, especially at generation 6 (Fig. 5A).

**Fig. 5: The concentration of 5,9-PD shifts in *D. pseudoobscura* and *D. persimilis* flies across the six generations of selection.**

To further test the relationship of 5,9-PD between species, environment, and through the generations, we calculated a species differential score by subtracting the D. persimilis 5,9-PD scores from the pseudoobscura 5,9-PD scores for all datapoints (see “Methods”). For the male flies, the species differential marginally differed across the generations when accounting for multiple tests (p = 0.039, two-way ANOVA), but significantly differed between environments (p = 0.001, two-way ANOVA). On average, the differential was higher for sympatric flies, and a clear divergence was observed at generations 4 and 6 (Generation 4: p = 0.05, Generation 6: p = 0.1, Tukey-Kramer post-hoc test; Fig. 5C), suggesting species’ differences in the concentration of 5,9-PD increased in experimental sympatry compared to allopatry. This trend was consistent when visualized in each population separately, although Population A showed less evidence of divergence between the environments (Supplementary Fig. 10). For the female flies, the differential scores are interpreted differently since D. persimilis flies have a higher 5,9-PD concentration than D. pseudoobscura flies, opposite of what was observed in males. Marginal differences were found across the generations in the female data (p = 0.083, two-way ANOVA) and significant differences were found between environments (p = 0.025, two-way ANOVA), although Tukey-Kramer post hoc tests revealed no significant differences between environments at specific generations (Generation 1: p = 0.93; Generation 4: p = 0.65; Generation 6: p = 0.66). On average, the differential was closer to zero for sympatric flies, supporting the previous observation that females raised in experimental sympatry converged in their 5,9-PD concentration (Fig. 5B). This trend was also consistent in Populations B-C when visualized separately (Supplementary Fig. 11). The differential scores also showed evidence of divergence at generations 4 and 6, but not to the same extent as males. Taken together, the female convergence and male divergence in the concentration of 5,9-PD between species in experimental sympatry suggests sexual dimorphism increased in the sympatric flies. This supports our hypothesis that CHC profiles shift to strengthen sexual isolation and that phenotypic differences between species vary across the two environments.

Social networks diverge between allopatric and sympatric flies

In total, we computed five measurements to investigate the social structure of each social group, and we corrected for multiple tests. The five social structure measures were assortativity, clustering coefficient, betweenness centrality, global efficiency, and reciprocation (see Methods for definitions). We found no evidence that allopatric and sympatric groups differed in their assortativity, clustering coefficient, and betweenness centrality scores (Supplementary Figs. 12–16). We did find evidence that global efficiency, a measure which scores the average length of pathways within a network, was higher in groups of sympatric flies by the final generation (Fig. 6A), however this relationship was not considered significant after adjusting for multiple tests (adjusted p = 0.087, two-way ANOVA). This marginal effect stems from the divergence between the treatments at generation 6, confirmed through post-hoc tests (p = 0.1, Tukey-Kramer multiple comparison test). Examining the global efficiency data for each population separately revealed the trend in generation 6 was consistent in two out of three populations (Supplementary Fig. 16). In fact, the average global efficiency was higher in the sympatric treatment compared to the allopatric treatment for Population A flies and Population C flies (Supplementary Fig. 16). This confirms the marginal effect seen in the pooled population data (Fig. 6A) results from the divergence observed in the Population A flies (Supplementary Fig. 16). Despite the weak statistical support, these results suggest that six generations of sympatry lead to flies forming social structures with shorter paths of connection, implying group cohesion increased in sympatric flies.

**Fig. 6: Behavioural measures analyzed from tracked videos containing mixed-sex and mixed-species social groups consisting of flies raised in allopatry (blue) and flies raised in sympatry (red).**

We also found evidence that the tendency to reciprocate social interactions significantly diverged between the treatments (adjusted p < 0.0001, two-way ANOVA) beginning at generation 4 (p = 0.03, Tukey-Kramer multiple comparison test) and continuing at generation 6 (p < 0.0001, Tukey-Kramer multiple comparison test) (Fig. 6B). Examining reciprocation across the separate populations revealed only one population displayed this divergence at generation 4, but the divergence at generation 6 was shared in all populations (Supplementary Fig. 17). This result suggests reciprocation and global efficiency are associated since both measures qualitatively diverge between sympatric and allopatric flies at the final generation in most populations. The reduction in the mean reciprocation for sympatric flies suggests a higher proportion of one-way social interactions in these networks, which may contradict the interpretation that sympatric groups are more cohesive. We speculate that this indicates segregation in the social network – cohesion is increasing within two subnetworks disconnected by a lack of reciprocity. Also, lower reciprocation may indicate higher frequencies of courtship since the male’s head would approach the female’s rear at an angle of 180°. Our social interaction criteria would not count these as reciprocal interactions because the angle parameters never exceeded 140^o for all treatments (Supplementary Table 1). The reduced reciprocation observed in groups of sympatric flies may indicate higher levels of courtship, potentially through increased conspecific preference.

Subgroup networks reveal divergence between allopatric and sympatric D. pseudoobscura

Our data above suggests an association between physiology, behaviour, and sexual isolation between species. The behavioural divergence observed in the social networks of allopatric and sympatric flies at generation 6 (Fig. 6) coincides with the divergence in 5,9-PD (Fig. 5) and the reduction in hybrid offspring (Fig. 1A). We speculated that the differences in social networks between allopatric and sympatric flies could indicate differences in the social organization patterns of D. pseudoobscura flies and D. persimilis flies separately. The two species raised in sympatry may form connections with conspecifics differently than those raised in allopatry. This would suggest the species organize their social interactions in a way that favours segregation between species. The social network analysis of the mixed-sex and mixed-species groups (Fig. 6) did not examine the nuances between social interactions within a single species (conspecific interactions) and social interactions between species (heterospecific interactions). Instead that analysis applied general social criteria that was estimated to be species’ specific (Supplementary Table 3). Investigating the social networks at this deeper level was critical to testing the hypothesis that species in sympatry diverge in their social structures.

To analyze the dynamics between conspecific interactions, two separate social networks were generated by isolating social interactions within species (1: D. pseudoobscura males and females; 2: D. persimilis males and females). Additionally, to analyze the dynamics between heterospecific interactions, two additional social networks were generated by isolating the social interactions between species in both combinations (3: D. pseudoobscura males and D. persimilis females; 4: D. persimilis males and D. pseudoobscura females). These four “subgroup social networks” were generated for the allopatric and sympatric social treatments within each population. All subgroups were assigned unique social criteria before calculating social networks (Supplementary Tables 4–6). We conducted this subgroup analysis on all five behavioural measures. Since the three populations displayed variation in their behavioural trends when analyzed separately (Supplementary Figs. 13–17), we decided to conduct this subgroup analysis separately for each population. In total, we conducted 45 statistical tests to compare differences between subgroup networks (see Methods). We hypothesized that if subgroup social structures are associated with sexual isolation between species, then differences between allopatric and sympatric subgroup networks would be observed at the sixth generation. Additionally, by the sixth generation we expected differences to be observed between social networks formed by conspecific D. persimilis social interactions compared to networks formed by conspecific D. pseudoobscura social interactions.

Consistent differences between allopatric and sympatric subgroups were evident in three behavioural measures at generation 6 – betweenness centrality (Supplementary Fig. 18), global efficiency (Fig. 7), and reciprocation (Fig. 8). Assortativity displayed no significant differences between subgroups in all populations (Supplementary Fig. 19), and clustering coefficient displayed significant differences between subgroups in an inconsistent pattern within two populations (Supplementary Fig. S20). For betweenness centrality, global efficiency, and reciprocation, we observed consistent differences between allopatric and sympatric flies in the conspecific social networks formed by D. pseudoobscura at generation 6 (Supplementary Fig. 18, Figs. 7, 8). Global efficiency displayed significant differences between allopatric and sympatric flies within this subgroup in all populations (p < 0.0001 for all populations; t-test), and this trend was present in reciprocation (Population A: p < 0.0001; Population B: p < 0.0001; Population C: p = 0.003; t-test), and betweenness centrality (Population A: p = 0.0039; Population B: p = 0.0004; Population C: p = 0.004; t-test). We observed no consistent patterns in the heterospecific social networks, nor the conspecific social networks formed by D. persimilis flies across the populations for all social network measures.

Fig. 7: Global efficiency analyzed for each population separately after splitting the treatment groups into two conspecific subgroups (Dper/Dpse) and two heterospecific subgroups (DperF+DpseM and DperM+DpseF).

Fig. 8: Reciprocation analyzed for each population separately after splitting the treatment groups into two conspecific subgroups (Dper/Dpse) and two heterospecific subgroups (DperF+DpseM and DperM+DpseF).

The global efficiency and reciprocation measures also displayed consistent differences between the networks formed by conspecific D. persimilis interactions and the conspecific D. pseudoobscura interactions (Figs. 7–8). By the fourth generation, the sympatric D. pseudoobscura subgroup generated networks with a global efficiency that significantly differed from the sympatric D. persimilis subgroup in two out of three populations (Population A: p = 0.006; Population B: p = 0.007; Population C: 0.038; t-test), and this trend continued into the sixth generation (Population B: p = 0.005; Population C: p = 0.006; t-test). A similar trend was observed for reciprocation, with remarkable differences between the sympatric D. pseudoobscura and D. persimilis subgroups at the fourth generation (Fig. 8), but stronger effects were found at the sixth generation (Population A: p = 0.024; Population B: p < 0.0001; Population C: p = 0.002; t-test). A divergence was also observed at the fourth and sixth generations for both measures between the allopatric D. pseudoobscura and D. persimilis subgroups (Figs. 7 and 8). However, in most cases the divergence between the allopatric subgroups was opposite to the divergence observed between the sympatric subgroups (Figs. 7 and 8), suggesting the two environments influence the development of different social structures. Together, this data supports our hypothesis that species adapted to sympatry produce divergent social structures, but we did not expect to see such differences emerge between the species adapted to allopatry (Figs. 7 and 8).

While differences were found in conspecific networks formed by allopatric versus sympatric D. pseudoobscura flies, the interpretation of these differences is complex. In population A we observed a higher mean betweenness centrality, higher mean global efficiency, and lower mean reciprocation in sympatric flies compared to allopatric flies at generation 6 (Supplementary Fig. 18, Figs. 7 and 8), suggesting networks formed by conspecific D. pseudoobscura social interactions resulted in higher group cohesion, shorter paths between nodes, and more courtship activity for sympatric flies. Also, we measured a higher global efficiency and a lower reciprocation in the D. pseudoobscura sympatric subgroup relative to the D. persimilis sympatric subgroup for this population, suggesting D. pseudoobscura flies in sympatric environments engage in conspecific interactions that leads to higher group cohesion compared to that in D. persimilis flies. Populations B-C display the exact opposite trend at the same generation (Supplementary Fig. 18, Figs. 7 and 8), suggesting these sympatric networks are less cohesive and emerge from less courtship interactions compared to allopatric networks. This trend in populations B-C does not support our hypothesis, and instead suggests allopatric D. pseudoobscura flies display higher conspecific preference at generation 6. However, in generation 4 we also observed divergent trends between sympatric and allopatric D. pseudoobscura subgroup networks. In all populations, sympatric D. pseudoobscura flies formed conspecific networks with a higher global efficiency and lower reciprocation compared to its allopatric counterparts, and compared to sympatric D. persimilis flies and allopatric D. persimilis flies (Figs. 7 and 8), supporting our hypothesis that conspecific preference increased for D. pseudoobscura flies raised in sympatry. Taken together, this data suggests either separate behavioural strategies evolved in the flies throughout the six generations, or strong drift occurred between the fourth and sixth generation that radically shifted behavioural phenotypes between the populations. The consistent divergence between allopatric and sympatric flies within networks formed by D. pseudoobscura conspecific interactions, suggests that emergent social structures may be associated with sexual isolation that evolved in a single species.

Discussion

In this study we observed that D. pseudoobscura flies increase their sexual isolation with D. persimilis flies following their placement in an artificial sympatric environment for six generations (Fig. 1, Supplementary Fig. 1). This parallels other “kill all hybrids” experiments despite the different methods used in this report^9,10,11. Our experimental design used fewer flies and fewer generations compared to similar experiments, yet we arrived at similar conclusions. With strong pressure against hybrids, flies with heightened prezygotic barriers are selected, driving the evolution of sexual isolation¹⁵. Alternatively, another mechanism for the increased sexual isolation in experimental sympatry could relate to overcoming sexual conflict since the presence of non-hybridizing species can enhance heterospecific mating, likely through confusion from an excess of communication signals within the environment⁴⁴. This may explain why hybrid offspring were not consistently found when offspring were counted in vials of 12 flies (Fig. 2A); in small groups D. pseudoobscura and D. persimilis can better distinguish conspecifics. Regardless, both mechanisms involve the selection of traits that discriminate conspecifics from heterospecifics, favouring divergence between species.

CHCs are well studied in insects because they play a dual role in social communication and in tolerating environmental pressures, linking behaviour to environment^45,46. The compound 5,9-PD, known to be concentrated in D. persimilis⁵, was measured at surprisingly high concentrations in D. pseudoobscura flies (Fig. 3C). Koopman reported that D. pseudoobscura and D. persimilis maximize heterospecific copulations at 16 °C⁹, and here our results suggest this may be associated with the phenotypic plasticity of 5,9-PD in male D. pseudoobscura flies (Fig. 3). Other researchers have observed an increase in the concentration of 5,9-PD within D. pseudoobscura flies that were reared in cool environments, suggesting this is a species-wide phenomenon^38,43,47. D. persimilis flies do not share the plasticity in this compound, as seen by the similar concentration of 5,9-PD from both rearing temperatures (Fig. 3C). The plasticity in D. pseudoobscura CHCs may increase this species’ tolerance to thermal and desiccation stress, explaining why D. pseudoobscura thrives in a wider geographic range compared to D. persimilis⁴⁸. More broadly, differences in phenotypic plasticity may be an important factor promoting speciation, since a recent experiment reported a temperature-dependent effect on genital morphology in D. santomea that was not shared by it’s sister species D. yakuba⁴⁹. Further efforts are necessary to confirm this hypothesis.

Whether 5,9-PD increases mating success in D. pseudoobscura is not understood. An experiment spanning 80 generations found that D. pseudoobscura reared within a polygamous social environment expressed higher concentrations of 5,9-PD compared to monogamous flies³⁹. Another experiment reported a difference in the concentration of 5,9-PD between American and Colombian D. pseudoobscura populations, suggesting this compound plays a role in the sexual isolation of those populations³⁸. In this experiment, 5,9-PD diverged in male and female flies after four generations of selection in sympatry. In fact, the concentration of 5,9-PD in D. pseudoobscura males surpassed that of D. persimilis males, while species differences in female flies decreased (Fig. 5, Supplementary Figs. 8 and 9). It is widely reported that males of both species court females indiscriminately, while female flies are selective in their mate choice^50,51. Therefore, our results suggest that the divergence of 5,9-PD in males enhanced the sexual dimorphism of flies in sympatry, enabling females to better discriminate the males. This assortative mating could have led to the decreased frequency of hybrids throughout the six generations. Since the increased sexual dimorphism for this compound was detected at generation 4, it is plausible that physiological changes from cuticular hydrocarbons first evolved in sympatry, leading to increased sexual isolation and changes to social structures detected at generation 6. Although our data offers an association between CHCs and sexual isolation, it is important to note that CHCs can also shift in the laboratory through drift, or through selection from desiccation stress⁴⁷, and we cannot rule out these confounds. Additionally, our system was designed to artificially weaken sexual isolation by maintaining flies at a temperature conducive to increased heterospecific mating. Although 5,9-PD is a strong candidate for the sexual isolation observed in this system, more work is necessary to confirm if this compound reinforces sexual isolation within natural populations.

A fundamental question addressed in this study was whether increased sexual isolation was associated with changes in social structures. We hypothesized that flies from the sympatric environments would produce more segregated social structures than flies from the allopatric environments. When studying large mixed-sex and mixed-species groups, there was marginal evidence suggesting social structures were associated with sexual isolation. Sympatric flies increased their social cohesion at the sixth generation, as seen by differences in the social network property global efficiency (Fig. 6A). Also, sympatric flies appeared to elevate courtship frequency at the sixth generation compared to allopatric flies as seen by the reciprocation measure (Fig. 6B). These trends offered support to the idea that sympatric flies produce segregated social structures with increased conspecific preference. However, this simple interpretation became uncertain when examining the subgroup networks. Instead, we found conflicting trends in each population. Despite this conflict, networks formed through conspecific social interactions of D. pseudoobscura flies displayed some degree of divergence between sympatric and allopatric flies by the sixth generation and this difference was common in all populations (Figs. 7 and 8, Supplementary Fig. 18). Additionally, networks formed by D. pseudoobscura conspecific interactions differ from the networks formed by D. persimilis conspecific interactions in both sympatry and allopatry. Although the interpretation of these differences in social networks was varied, the difference between treatments was consistent and suggests social structures diverged in alternative ways unique to each population. No consistent trends were observed in the conspecific networks compared between allopatric and sympatric D. persimilis, nor in the heterospecific social networks. Together this implies that selection acted solely on the behaviour of D. pseudoobscura in our evolutionary experiment. The divergence in the networks between sympatric and allopatric D. pseudoobscura at the fourth generation coincides with the divergence of 5,9-PD beginning at generation 4, offering further support that social structures evolve alongside physiological traits to enhance the sexual isolation between species.

In conclusion, our study represents an artificial examination of the physiological and behavioural dynamics between species in hybrid zones. The consistency of these findings to other “kill all hybrids” experiments^9,10,11 (Fig. 1) strengthens the evidence that increased sexual isolation evolves within a few generations inside a closed sympatric environment. The changes to the CHCs (Fig. 4), and the changes to social behaviours detected within subnetworks (Figs. 7–8, Supplementary Fig. 18) supports our hypothesis that flies coexisting in sympatric hybrid zones form a social environment that differs in fundamental ways to flies found in geographic isolation. The composition of social groups is known to impact individual expression of circadian rhythms, CHC profiles and reproductive behaviours of insects through indirect genetic effects^25,52. For example, the aggression observed in a male fly depends not only on his genetic predisposition and the abiotic factors in his surrounding environment, but also on social cues such as the concentration of chemical signals and prior social encounters⁵³. At a larger scale, sympatric versus allopatric zones in nature represent two social contexts that exert indirect genetic effects on the physiology and behaviour of individuals within the population, ultimately influencing the sexual isolation between species. One mechanism involves the concentration of CHCs shifting in flies first, leading to the increased sexual isolation and divergence of social structure between allopatric and sympatric flies. Considering the known sensitivity of CHCs to group composition²⁵, an alternative mechanism is that the social structure drives the segregation of species into two modular social environments, influencing CHC expression and increased sexual isolation. Of course there are other mechanisms we did not account for in this experiment, such as sperm competition⁷, that may have interacted with changes in CHC expression to enhance the sexual isolation we observed. Nonetheless, we favour the interpretation that the data presented in this report suggest that social network properties contribute to sexual isolation between species.

Methods

Fly stocks

Since D. persimilis and D. pseudoobscura have no reliable morphological markers to distinguish the species, a D. persimilis orange (or) mutant line was obtained from the Drosophila Species Stock Center (stock # 14011-0111.60) and a D. pseudoobscura glass (gl) mutant line was generated from an upturned-glass double mutant line (Drosophila Species Stock Center, 14011-0121.16). Since gl and or are on separate autosomes, hybrid offspring appear as wild type (Supplementary Fig. 21).

Wild D. pseudoobscura stocks captured from the Okanagan Lake area of British Columbia in July 2017 were also used in this study. These 68 isofemale lines were all confirmed as D. pseudoobscura through the amplification of a reliable species identification marker DPSX002⁵⁴. D. persimilis isofemale lines were not captured in this location, suggesting the D. pseudoobscura flies were captured in an allopatric environment. However, this is speculative and does not rule out the possibility that the D. pseudoobscura lines were temporally isolated from D. persimilis in an otherwise sympatric area.

To produce a genetically diverse progenitor population for the selection treatments, three generations of backcrossing were completed. At the first step, 150 virgin gl females and 3-4 males from each of the 68 isofemale lines were mixed in a 9 inch by 14.4 inch population cage (Genesee Scientific, Cat# 59-116). Four 177 mL square bottom stock bottles (VWR, International) containing 40 mL of standard cornmeal media were placed inside the population cage and changed every 2–3 days. At the second step, virgin heterozygote offspring were collected from the bottles and randomly mixed into fresh bottles. Each bottle was flipped every 3–4 days to maximize the collection of gl virgin female offspring to be used for the next generation of crossing. This two-step cross was repeated three times.

For the experiment measuring the effects of temperature on CHC profiles, the or mutant line was used to represent D. persimilis. Since the gl mutation has a far more noticeable effect on morphology, an inbred wild-type stock of D. pseudoobscura was used (14011-0121.148, Drosophila Species Stock Centre). This stock originates from California, a location in sympatry withD. persimilis. The CHC profile of the wild type stock and the glass mutants were similar (Supplementary Fig. 7).

Selection treatments

Three independent populations were organized into the following selection treatments using 177 mL, square bottom Drosophila stock bottles (VWR, International) to represent two environments: (1) experimental sympatry where 40 D. pseudoobscura flies and 40 D. persimilis flies were maintained in stock bottles with an equal number of males and females; 2) experimental allopatry where either 80 D. pseudoobscura flies or 80 D. persimilis flies were maintained in stock bottles with an equal number of males and females. All bottles contained 40 mL of standard cornmeal media. For each population, 8 sympatric bottles and 8 allopatric bottles (4 per species) were seeded each generation with virgin flies that were aged 5–14 days at 16 °C. To avoid inbreeding, virgin male and female flies were randomly collected from all bottles and evenly divided and shuffled into their respective sympatric and allopatric bottles to seed the next generation. All bottles were maintained at 16 °C on a 12-h L:D cycle and the flies were transferred every 3-5 days into fresh bottles to maximize the collection of offspring. The adult flies were maintained at 16 °C to enhance the frequency of heterospecific mating as stated by Koopman⁹. After each transfer, the bottles containing larvae were maintained at 24 °C on the same 12-h L:D cycle to accelerate their development. For all populations, generations 5 and 6 were seeded with 6 sympatric and 6 allopatric bottles. This was due to difficulties in collecting healthy offspring and from D. persimilis flies being less abundant at these generations.

Estimating Sexual Isolation in Sympatry

For each of the six generations of selection, the number of D. pseudoobscura, D. persimilis and hybrid offspring were counted from the first 8 sympatric bottles (first 6 sympatric bottles in generations 5 & 6) to estimate the strength of sexual isolation for each population. The percentage of D. pseudoobscura, D. persimilis, and hybrid offspring was calculated for each bottle since the total number of offspring varied. All hybrid offspring were discarded after counting.

Extraction and characterization of cuticular hydrocarbons

Male and female D. pseudoobscura and D. persimilis flies were collected from the sympatric and allopatric bottles for cuticular hydrocarbon (CHC) extractions following: (i) one generation of selection; (ii) four generations of selection; iii) six generations of selection. Previous examinations of CHCs in these species demonstrated that D. persimilis flies have higher concentrations of the alkadiene 5, 9-pentacosadiene (5, 9-PD) while D. pseudoobscura flies have higher concentrations of 5, 9-heptacosadiene (5, 9-HD)^5,39. By extracting and characterizing the CHC profiles of these species after multiple generations of selection, the hypothesis was tested that CHC profiles play a role in strengthening sexual isolation between the species.

For the CHC extractions, virgin male and female flies of both species were collected from sympatric and allopatric bottles and aged for three days in 25 × 95 mm Drosophila vials (VWR, International) containing 8 mL of standard cornmeal media. A total of 12–16 flies were housed in these vials at 16 °C on a 12 h light:dark cycle. All extractions were started 9.5–10.5 hours after the initiation of the light phase in the flies’ light:dark cycle. Extractions were done by anaesthetizing the flies on ice and submerging a single fly in 50 µL of hexane containing 10 ng/µL octadecane (C18) and 10 ng/µL hexacosane (C26) as internal standards. The samples were vortexed for 5 minutes before gently removing the flies. All samples were numbered by order and time of extraction. A total of 15 flies of each species, sex, and from each selection treatment (sympatric or allopatric) were extracted for each population, totalling 360 samples per generation. However, CHCs were not extracted from Population C D. persimilis flies at the sixth generation due to these flies eclosing in limited supply. All samples were kept in a −20 °C chest freezer for a maximum of two weeks prior to characterizing the samples through gas chromatography. All CHC extracts were loaded onto the gas chromatograph in a pattern that alternated between sex, species and environment (allopatric vs sympatric). This was to ensure different treatment samples were characterized at a similar time.

CHC extracts were analyzed using an Agilent 7890 A gas chromatograph system with a flame ionization detector (GC/FID) and PTV injector (cool-on-column mode) and outfitted with a DB-1 20 m × 0.18 mm Agilent 121-1022 fused silica capillary column (Agilent Technologies, Inc. Santa Clara, CA, USA). The samples were first heated to 70 °C for 1 min, followed by a ramp-up to 180 °C at a rate of 20 °C/min, then 4 °C/min to 220 °C, and finally 15 °C/min to 320 °C where it was held for 2 min, as outlined by Hunt et al.³⁹. The resulting chromatograms for D. pseudoobscura were identical to those illustrated by Hunt et al.³⁹, offering the identification of key compounds (Supplementary Figs. 3–5). A total of 11 compounds were identified, but the compounds 5, 9-PD and 5, 9-HD were emphasized because of their high concentration in each species. Two additional compounds were included in downstream analyses that were not confidently identified but were present in both species at relatively high concentration: (1) a C25 compound following 8-Pentacosene on the chromatogram (Supplementary Figs. 5 and 6); (2) a C27 compound following 5, 9-Heptacosadiene on the chromatogram (Supplementary Figs. 5 and 6).

Effects of temperature on cuticular hydrocarbon profile

Upon analyzing the CHC profiles, we noticed D. pseudoobscura flies had concentrations of 5, 9-PD comparable to D. persimilis, contradicting Hunt et al. that posits D. pseudoobscura have relatively low concentrations of 5, 9-PD³⁹. To confirm whether there are species differences in CHC plasticity, an experiment was conducted where virgin male and female flies of both species were reared at 16 °C and 24 °C for three days in vials containing 8 mL of standard cornmeal. CHCs were extracted and analyzed as outlined above for a total of 28 D. pseudoobscura male flies (14 raised at 16 °C, 14 raised at 24 °C), 28 D. persimilis male flies (14 raised at 16 °C, 14 raised at 24 °C), 28 D. pseudoobscura females flies (14 raised at 16 °C, 14 raised at 24 °C), and 16 D. persimilis female flies (8 raised at 16 °C, 8 raised at 24 °C).

Measuring social structures

To test our hypothesis that social structures diverge between flies raised in allopatry versus sympatry, independent social groups of flies were confined within arenas and filmed to gather videos for the social network assay³². For all three populations, male and female flies of both species were collected for use in this assay after: (i) one generation of selection; (ii) four generations of selection; (iii) six generations of selection. All flies were collected as virgins and housed for three days at 16 °C in vials containing 8 mL of standard cornmeal media. During that rearing period, each vial contained 12–16 flies of a single species and sex at the same light:dark cycle described above. All flies were housed in homogenous groups and had no social experience with heterospecifics prior to experiments.

Two mixed-sex and mixed-species social treatments were monitored in this behavioural assay: (1) D. persimilis flies from sympatry mixed with D. pseudoobscura flies from sympatry (sympatric treatment); (2) D. persimilis flies from allopatry mixed with D. pseudoobscura flies from allopatry (allopatric treatment). Both treatments consisted of 12 flies with an equal number of male and female flies of both species. All behavioural experiments were conducted within a chamber maintained at 16 °C and 65% relative humidity (Biochambers, USA). Flies were gently aspirated into circular arenas made of plexiglass with a diameter of 60 mm and a depth of 3 mm. Glass lids were fit over the arena to confine the flies. Additionally, a 0.5 mm strand of spring steel was used to keep male and female flies separated when loading the arenas. Once the arenas were stationed beneath the cameras, the flies were left to acclimate for 10 min with the dividers in place. After acclimation, the dividers were gently removed, and thirty minutes of video footage was filmed at 22.8 fps using Grasshopper3 USB 3.0 cameras (GS3-U3-51S5C-C, Teledyne FLIR LLC, USA). These cameras were outfitted with Fujinon 1.5 MP 12.5 mm lenses (Fujifilm, Global) to enhance video capture. Flies were discarded after the filming was complete and each video was acquired with new flies. All videos were tracked through Ctrax and the quality of each tracking file was assessed through the fixerrors script in MATLAB.

For each population, 15–20 videos were obtained for each treatment. The following social space parameters were estimated for each treatment in all populations and generations: (1) the distance (body lengths) between the centres of mass for interacting flies; (2) the angle (degrees) of the line connecting the centres of mass from interacting flies and the line extending from the head region of the fly initiating the interaction; (3) the time (seconds) required for the distance and angle criteria to fulfil a social interaction. This algorithm, published by Schneider and Levine⁵⁵, was modified to generate separate social space criteria in the mixed species groups. Thus, all networks were generated from social interactions that accounted for variation in each species’ social space (Supplementary Table 3).

Five measures were generated from the social network data. (1) Assortativity: defined as the probability of an individual interacting with another individual with a similar number of social connections⁵⁶. Networks with high assortativity, on average, implies nodes are connected to other similar nodes, suggesting segregation between group members. (2) Clustering coefficient: a measure of how interconnected neighbours are to one another⁵⁶. Networks with high clustering coefficient, on average, indicates a higher cliquishness among nodes, suggesting segregation between group members. (3) Betweenness centrality: the relative importance of a given individual for communication relay and maintaining group cohesion⁵⁶. Networks with low betweenness centrality, on average, have fewer individuals in the centre of the network, suggesting segregation between group members. (4) Global efficiency: a measure of redundant pathways, offering insight into the average distance between individuals in a network⁵⁷. Networks with low global efficiency, on average, implies larger paths between nodes, suggesting segregation between group members. 5) Reciprocation: the average frequency of social interactions reciprocated as a percentage of all social interactions observed⁵⁶. Networks with low reciprocation, on average, implies weaker cohesion between nodes, suggesting segregation between group members. These measurements have been reported to capture social structure dynamics in various social contexts across Drosophila strains and species, highlighting their relevance in measuring social properties at the group level⁵⁸.

Estimating sexual isolation in sympatric and allopatric social groups

Following the social network assay, the groups of 12 flies were not immediately discarded. Instead, the flies were gently anaesthetized with CO₂, carefully removed from the arenas, and transferred to vials with 8 mL of standard cornmeal media. These vials were kept for 3–5 days at 16 °C on a 12-h light: dark cycle. Then, adults were removed and the vials containing eggs were transferred to an incubator set to 24 °C with the same light: dark cycle. The number of D. pseudoobscura, D. persimilis,and hybrid offspring was counted for each vial and the percentage of offspring was calculated since the total number of offspring varied. All offspring were discarded from these vials and not used for any experiments.

Measuring social structures within subgroups

An additional social network analysis was conducted that investigated social interactions between males and females that were species-specific (conspecific) and non-species-specific (heterospecific) between allopatric and sympatric flies. Since all tracked videos contained data for 12 fly trajectories, tracks were removed from each video so that social interactions and social networks could be analyzed within subgroups of 6 flies. Thus, each video was measured repeatedly in various combinations of subgroups. The following subgroups were analyzed: (1) D. persimilis male and female flies (Dper subgroup); (2) D. persimilis male and D. pseudoobscura female flies (DperM+DpseF subgroup); (3) D. persimilis female and D. pseudoobscura male flies (DperF+DpseM subgroup); (4) D. pseudoobscura male and female flies (Dpse subgroup). Social networks were generated for each subgroup as described above. To emphasize, this data did not result from pruning down each original 12-node social network. Instead, social space criteria (Supplementary Tables 4–6) were generated for all subgroups of 6 flies and from these criteria new social networks, which we call subnetworks, were generated and analyzed separately. This analysis was completed in all populations separately for generations 1, 4, and 6.

Statistics and reproducibility

We tested whether the generation variable significantly predicts differences in offspring frequency over time through a Wald chi-square test. First, models were generated by fitting the generation variable as a single predictor to each of the three offspring frequency response variables (hybrid, Dpse, Dper) separately using the glm() function in R. The Poisson distribution was selected to model the error distribution. Then, Wald chi-square tests were performed using the wald.test() function was used in R where the variance-covariance matrix, the coefficients, and the single term (generation) were implemented from each of the three models. To account for multiple tests, p-values less than 0.016 from the Wald chi-square test were considered statistically significant. Prior to conducting these tests, all data from the three populations was pooled together (n = 24 for generations 1–4; n = 18 for generations 5-6). A datapoint was considered an outlier if it was greater than the 75th quartile + (1.5*IQR) or lower than the 25th quartile-(1.5*IQR). A total of seven outliers were removed when analyzing the hybrid offspring frequency, and nine outliers were removed when analyzing the D. pseudoobscura and D. persimilis offspring frequency.

To compare the differences in the frequency of hybrid offspring between the allopatric and sympatric treatments following the social network assay, an aligned rank-transformation test was conducted. This test was selected since the sample sizes were relatively small for each treatment due to a low incidence of hybrid offspring born from the vials (Gen 1 allopatric: n = 20; Gen 1 sympatric: n = 19; Gen 4 allopatric: n = 18; Gen 4 sympatric: n = 11; Gen 6 allopatric: n = 20; Gen 6 sympatric: n = 12). In these tests, the frequency of hybrid offspring was the response variable, while generation and treatment were the factors. This test was implemented in R using the art() function in R (ARTTool package). Additionally, the frequency of D. pseudoobscura and D. persimilis offspring were compared across the generations and treatments through a two-way ANOVA in R using the aov() function. A Tukey multiple comparisons of means post-hoc test was performed using the TukeyHSD() function in R. The data from all populations was pooled together prior to statistical analysis (Gen 1 allopatric: n = 55; Gen 1 sympatric: n = 56; Gen 4 allopatric: n = 53; Gen 4 sympatric: n = 54; Gen 6 allopatric: n = 56; Gen 6 sympatric: n = 46). Outliers were also removed before conducting these statistical tests and a datapoint was considered an outlier if it was greater than the 75th quartile + (1.5*IQR) or lower than the 25th quartile-(1.5*IQR). Finally, the proportion of vials that produced hybrid offspring were compared between allopatric and sympatric treatments. A chi-square test was conducted for each generation separately using the chisq.test() function in Microsoft Excel. The observed values was the percentage of vials that produced hybrids over the total number of vials. The expected values were calculated by summing the total number of vials that produced hybrid offspring (both allopatric and sympatric) and dividing it by the total number of vials (both allopatric and sympatric).

To test the effect of temperature on CHC profiles, a principal component analysis (PCA) was conducted using the prcomp() function in R. A total of 12 CHC compounds (plus cVA) were implemented into the PCA. Prior to conducting the test, all 13 measurements were standardized into z-scores. The results were visualized by plotting the first two principal components using the autoplot() function in R. Given that 5, 9-PD and 5, 9-HD are the most concentrated compounds in D. persimilis and D. pseudoobscura, respectively, the concentrations of these compounds were compared between species and temperature regimes. A two-way ANOVA was conducted, where 5, 9-PD concentration and 5, 9-HD concentration were the response variables, and the two factors were temperature (16 °C/24 °C) and species (Dper/Dpse). A p-value less than 0.0125 (α = 0.05/4) was considered statistically significant to account for the repeated measures. A Tukey multiple comparisons of means post-hoc test was performed using the TukeyHSD() function in R. For these temperature-plasticity experiments, 28 CHC samples were extracted from D. persimilis males (16 °C: n = 14, 24 °C: n = 14), 27 were extracted from D. pseudoobscura males (16 °C: n = 14, 24 °C: n = 13), 16 were extracted from D. persimilis females (16 °C: n = 8, 24 °C: n = 8), and 28 were extracted from D. pseudoobscura females (16 °C: n = 14, 24 °C: n = 14).

To test the effect of selection on CHC profiles, a PCA was also conducted on the CHC data extracted from each sympatric and allopatric population across the generations. For male flies, CHCs were extracted from: 178 flies after the first generation (Dper Sym: n = 45; Dper Allo: n = 45; Dpse Sym: n = 45; Dpse Allo: n = 43); 177 flies after the fourth generation (Dper Sym: n = 43; Dper Allo: n = 45; Dpse Sym: n = 44; Dpse Allo: n = 45); and 134 flies after the sixth generation (Dper Sym: n = 27; Dper Allo: n = 30; Dpse Sym: n = 38; Dpse Allo: n = 39). For female flies, CHCs were extracted from: 182 flies after the first generation (Dper Sym: n = 46; Dper Allo: n = 45; Dpse Sym: n = 45; Dpse Allo: n = 46); 178 flies after the fourth generation (Dper Sym: n = 45; Dper Allo: n = 45; Dpse Sym: n = 44; Dpse Allo: n = 44); and 138 flies after the sixth generation (Dper Sym: n = 30; Dper Allo: n = 30; Dpse Sym: n = 39; Dpse Allo: n = 39). This represented the total number of samples extracted and pooled across the three populations. The PCA was conducted as described above, except the male and female data was analyzed separately. Each generation was analyzed separately, and data from all populations was combined. Additionally, the effect of selection on the concentration of 5, 9-PD was explored for each population and generation, because this compound was found to be expressed at high concentration in both species at 16 ^oC. This was done through a two-way ANOVA where the concentration of 5, 9-PD was the response variable, and environment (allopatric vs sympatric) and species were the two factors. This was completed in R using the aov() function. This analysis was conducted for each generation in male and female flies separately. Due to the repeated testing, an effect was considered statistically significant if the p-value was less than 0.016 (α = 0.05/3). A Tukey multiple comparisons of means post-hoc test was performed using the TukeyHSD() function in R. Prior to conducting the statistical test, all outliers were removed from each treatment in all populations. A datapoint was considered an outlier if it was greater than the 75th quartile + (1.5*IQR) or lower than the 25th quartile-(1.5*IQR). The data from all populations was pooled together prior to statistical analysis.

The species differential of 5,9-PD was calculated as follows. For the male and female data separately, D. pseudoobscura and D. persimilis datapoints were paired based on a number tag used to label samples during the CHC extraction. This sample number tag also corresponded with the order that samples were run on the gas chromatography machine. For example, the datapoint corresponding to the first sympatric D. pseudoobscura male sample that was processed on the gas chromatograph was paired with the first sympatric D. persimilis male sample that was run shortly afterwards. Therefore, the paired datapoints represent two measurements processed by the gas chromatograph on the same day during the same hour. The differential for 5,9-PD was calculated as D. pseudoobscura 5,9-PD concentration (ng) – D. persimilis 5,9-PD concentration (ng). A positive number indicates the D. pseudoobscura concentration was higher than D. persimilis, while a negative number indicates the opposite. Values close to 0 indicate the concentrations were equal. This species differential was analyzed through a two-way ANOVA where generation and environment (sympatric or allopatric) were the two factors. This was analyzed in the male and female data separately. A p value equal to or less than 0.025 was considered significant to account for multiple testing. This was completed in R using the aov() function in R and a Tukey multiple comparison of means post-hoc test was performed using the TukeyHSD() function in R. A datapoint was considered an outlier if it was greater than the 75th quartile + (1.5*IQR) or lower than the 25th quartile-(1.5*IQR). The data from all populations was pooled together prior to statistical analysis.

To test the effect of selection on the social structures of the flies, social networks were compared between groups of sympatric and allopatric flies across the generations. A total of 106 networks were compared at generation 1 (Sym: n = 53; Allo: n = 53), a total of 94 networks were compared at generation 4 (Sym: n = 44; Allo: n = 50), and a total of 95 networks were compared at generation 6 (Sym: n = 42; Allo: n = 53). A two-way ANOVA test was implemented using the aov() function in R where social network measures were the response variables, while generation and treatment (allopatric/sympatric) were the two factors. Since five behavioural measures were investigated, the p-values were adjusted using the false discovery rate method to correct for multiple testing using the p.adjust() function in R. A Tukey multiple comparisons of means post-hoc test was performed using the TukeyHSD() function in R. Prior to conducting the statistical test, all outliers were removed from each treatment in all populations. A datapoint was considered an outlier if it was greater than the 75th quartile + (1.5*IQR) or lower than the 25th quartile-(1.5*IQR). The data from all populations was pooled together prior to statistical analysis.

For the 6-fly social network subgroup analysis, measures of the four subgroups and two social treatments were compared through a two-way ANOVA using the aov() function in R. This resulted in comparing the subgroups between allopatric and sympatric flies across three generation points, in three populations, using five behavioural measures, resulting in 45 two-way ANOVA tests. Therefore the p-values were adjusted using the false discovery rate method to correct for multiple testing using the p.adjust() function in R. For the two-way ANOVA tests that resulted in at least marginal effects or interactions (adjusted p-value ≤ 0.1), a series of four t-tests were performed to compare between the treatments (allopatric/sympatric) for each subgroup, and two additional t-tests compared between the D. pseudoobscura and D. persimilis subgroups. The post-hoc tests were considered significant if the p-value was less than 0.0083 (0.05/6) All outliers were removed from the data as described above. Each population was statistically analyzed separately due to large differences in trends between populations.

Reporting summary

Further information on research design is available in the Nature Portfolio Reporting Summary linked to this article.

Data availability

All source data can be obtained from the Supplementary Data file. Data is also available to download from Zenodo. https://doi.org/10.5281/zenodo.14269474

Code availability

Computer code is available from Zenodo https://doi.org/10.5281/zenodo.14269474.

References

Fry, J. D. Laboratory experiments on speciation. In Experimental Evolution: Concepts, Methods, and Applications of Selection Experiments. 631–656 (University of California Press, 2009).
Dobzhansky, T. Is there gene exchange between Drosophila pseudoobsura and Drosophila persimilis in their natural habitats? Am Natur 107, 312–314 (1973).
Article Google Scholar
Markow, T. A. & O’Grady, P. M. Drosophila: A Guide to Species Identification and Use (Elsevier, Inc, 2006).
Noor, M. A., Grams, K. L., Bertucci, L. A. & Reiland, J. Chromosomal inversions and the reproductive isolation of species. Proc Natl Acad Sci USA 98, 12084–12088 (2001).
Article CAS PubMed PubMed Central Google Scholar
Noor, M. A. & Coyne, J. A. Genetics of a difference in cuticular hydrocarbons between Drosophila pseudoobscura and D. persimilis. Genet Res 68, 117–123 (1996).
Article CAS PubMed Google Scholar
Noor, M. A. & Aquadro, C. F. Courtship songs of Drosophila pseudoobscura and D. persimilis: analysis of variation. Anim Behav 56, 115–125 (1998).
Article CAS PubMed Google Scholar
Castillo, D. M. & Moyle, L. C. Conspecific sperm precedence is reinforced, but postcopulatory sexual selection weakened, in sympatric populations of Drosophila. Proc Biol Sci 286, 20182535 (2019).
PubMed PubMed Central Google Scholar
Dobzhansky, T. Genetic structure of natural populations. Carnegie Institution of Washington Year Book 44, 127–134 (1945).
Google Scholar
Koopman, K. F. Natural selection for reproductive isolation between Drosophila pseudoobscura and Drosophila persimilis. Evolution 4, 135–148 (1950).
Matute, D. R. Reinforcement can overcome gene flow during speciation in Drosophila. Curr Biol 20, 2229–2233 (2010).
Article CAS PubMed PubMed Central Google Scholar
Jarvis, W. M. C., Arthur, N. J., Rundle, H. D. & Dyer, K. A. An experimental test of the evolutionary consequences of sympatry in Drosophila subquinaria. Evolution. https://doi.org/10.1093/evolut/qpad236 (2023)
Knight, G., Robertson, A. & Waddington, C. Selection for sexual isolation within a species. Evolution 10, 14–22 (1956).
Crossley, S. A. Changes in mating behavior produced by selection for ethological isolation between ebony and vestigial mutants of Drosophila melanogaster. Evolution 28, 631–647 (1974).
Comeault, A. A., Venkat, A. & Matute, D. R. Correlated evolution of male and female reproductive traits drive a cascading effect of reinforcement in Drosophila yakuba. Proc. R Soc. B Biol. Sci. 283, 20160730 (2016).
Article Google Scholar
Noor, M. A. Speciation driven by natural selection in Drosophila. Nature 375, 674–675 (1995).
Article CAS PubMed Google Scholar
Anderson, W. W. & Kim, Y.-K. Sexual isolation between sympatric and allopatric populations of Drosophila pseudoobscura and D. persimilis. Behav. Genet. 35, 305–312 (2005).
Article PubMed Google Scholar
Davis, J. S., Castillo, D. M. & Moyle, L. C. Remating responses are consistent with male postcopulatory manipulation but not reinforcement in D. pseudoobscura. Ecol. Evol. 7, 507–515 (2017).
Article PubMed Google Scholar
Billeter, J.-C., Bailly, T. & Kohlmeier, P. The social life of Drosophila melanogaster. Insectes Sociaux, 1–14 https://doi.org/10.1007/s00040-024-00990-3 (2024).
Schneider, J., Atallah, J. & Levine, J. D. One, two, and many–a perspective on what groups of Drosophila melanogaster can tell us about social dynamics. Adv. Genet. 77, 59–78 (2012).
Article PubMed Google Scholar
Billeter, J. C., Atallah, J., Krupp, J. J., Millar, J. G. & Levine, J. D. Specialized cells tag sexual and species identity in Drosophila melanogaster. Nature 461, 987–991 (2009).
Article CAS PubMed Google Scholar
Ewing, A. W. & Bennet-Clark, H. The courtship songs of Drosophila. Behaviour 31, 288–301 (1968).
Article Google Scholar
Ramdya, P. et al. Mechanosensory interactions drive collective behaviour in Drosophila. Nature 519, 233–236 (2015).
Article CAS PubMed Google Scholar
Kravitz, E. A. & Fernandez, M. P. Aggression in Drosophila. Behav. Neurosci. 129, 549 (2015).
Article PubMed Google Scholar
Duménil, C. et al. Pheromonal cues deposited by mated females convey social information about egg-laying sites in Drosophila melanogaster. J. Chem. Ecol. 42, 259–269 (2016).
Article PubMed PubMed Central Google Scholar
Krupp, J. J. et al. Social experience modifies pheromone expression and mating behavior in male Drosophila melanogaster. Curr. Biol. 18, 1373–1383 (2008).
Article CAS PubMed Google Scholar
Maguire, C. P., Lizé, A. & Price, T. A. Assessment of rival males through the use of multiple sensory cues in the fruitfly Drosophila pseudoobscura. PLoS ONE 10, e0123058 (2015).
Article PubMed PubMed Central Google Scholar
Bretman, A., Westmancoat, J. D. & Chapman, T. Male control of mating duration following exposure to rivals in fruitflies. J. Insect Physiol. 59, 824–827 (2013).
Article CAS PubMed PubMed Central Google Scholar
Bentzur, A. et al. Early life experience shapes male behavior and social networks in Drosophila. Curr. Biol. 31, 670 (2021).
Article CAS PubMed Google Scholar
Jiang, L. et al. Emergence of social cluster by collective pairwise encounters in. Elife 9, e51921 (2020).
Article CAS PubMed PubMed Central Google Scholar
Liu, G. et al. A simple computer vision pipeline reveals the effects of isolation on social interaction dynamics in Drosophila. PLoS Comput. Biol. 14, e1006410 (2018).
Article PubMed PubMed Central Google Scholar
Pasquaretta, C. et al. How social network structure affects decision-making in Drosophila melanogaster. Proc Biol Sci 283, 20152954 (2016).
PubMed PubMed Central Google Scholar
Schneider, J., Dickinson, M. H. & Levine, J. D. Social structures depend on innate determinants and chemosensory processing in Drosophila. Proc. Natl Acad. Sci. USA 109, 17174–17179 (2012).
Article CAS PubMed PubMed Central Google Scholar
Simon, A. F. et al. A simple assay to study social behavior in Drosophila: measurement of social space within a group. Genes Brain Behav 11, 243–252 (2012).
Article CAS PubMed Google Scholar
Sun, Y. et al. Social attraction in Drosophila is regulated by the mushroom body and serotonergic system. Nat. Commun. 11, 5350 (2020).
Article CAS PubMed PubMed Central Google Scholar
Rooke, R. et al. The gene “degrees of kevin bacon” (dokb) regulates a social network behaviour in Drosophila melanogaster. Nat. Commun. 15, 3339 (2024).
Article CAS PubMed PubMed Central Google Scholar
Jezovit, J. A., Rooke, R., Schneider, J. & Levine, J. D. Behavioral and environmental contributions to drosophilid social networks. Proc Natl Acad Sci USA 117, 11573–11583 (2020).
Article CAS PubMed PubMed Central Google Scholar
Rodriguez-Moral, A. & Vorsatz, M. An overview of the measurement of segregation: classical approaches and social network analysis. Complex networks and dynamics: Social and economic interactions. 93–119 (Springer International Publishing, Switzerland, 2016).
Kim, Y. K., Ruiz-García, M., Alvarez, D., Phillips, D. R. & Anderson, W. W. Sexual isolation between North American and Bogota strains of Drosophila pseudoobscura. Behav. Genet. 42, 472–482 (2012).
Article PubMed Google Scholar
Hunt, J., Snook, R. R., Mitchell, C., Crudgington, H. S. & Moore, A. J. Sexual selection and experimental evolution of chemical signals in Drosophila pseudoobscura. J. Evol. Biol. 25, 2232–2241 (2012).
Article CAS PubMed Google Scholar
Rundle, H. D., Chenoweth, S. F., Doughty, P. & Blows, M. W. Divergent selection and the evolution of signal traits and mating preferences. PLoS Biol. 3, e368 (2005).
Article PubMed PubMed Central Google Scholar
Poikela, N. et al. Strength of sexual and postmating prezygotic barriers varies between sympatric populations with different histories and species abundances. Evolution 73, 1182–1199 (2019).
Article PubMed Google Scholar
Serrato-Capuchina, A. et al. Pure species discriminate against hybrids in the Drosophila melanogaster species subgroup. Evolution 75, 1753–1774 (2021).
Article PubMed Google Scholar
Toolson, E. C. Effects of rearing temperature on cuticle permeability and epicuticular lipid composition in Drosophila pseudoobscura. J. Exp. Zool. 222, 249–253 (1982).
Article CAS Google Scholar
Matute, D. R. Noisy neighbors can hamper the evolution of reproductive isolation by reinforcing selection. Am. Nat. 185, 253–269 (2015).
Article PubMed Google Scholar
Bontonou, G. & Wicker-Thomas, C. Sexual communication in the Drosophila genus. Insects 5, 439–458 (2014).
Article PubMed PubMed Central Google Scholar
Krupp, J. J., Nayal, K., Wong, A., Millar, J. G. & Levine, J. D. Desiccation resistance is an adaptive life-history trait dependent upon cuticular hydrocarbons, and influenced by mating status and temperature in D. melanogaster. J. Insect Physiol. 121, 103990 (2020).
Article CAS PubMed Google Scholar
Toolson, E. C. & Kuper-Simbron, R. Laboratory evolution of epicuticular hydrocarbon composition and cuticular permeability in Drosophila pseudoobscura: effects on sexual dimorphism and thermal-acclimation ability. Evolution 43, 468–473 (1989).
PubMed Google Scholar
Machado, C. A., Kliman, R. M., Markert, J. A. & Hey, J. Inferring the history of speciation from multilocus DNA sequence data: the case of Drosophila pseudoobscura and close relatives. Mol. Biol. Evol. 19, 472–488 (2002).
Article CAS PubMed Google Scholar
Peluffo, A. E. et al. A morphological trait involved in reproductive isolation between Drosophila sister species is sensitive to temperature. Ecol. Evol. 11, 7492–7506 (2021).
Article PubMed PubMed Central Google Scholar
Dukas, R. Dynamics of learning in the context of courtship in Drosophila persimilis and D. pseudoobscura. Anim. Behav. 77, 253–259 (2009).
Article Google Scholar
Noor, M. A. Absence of species discrimination in Drosophila pseudoobscura and D. persimilismales. Anim. Behav. 52, 1205–1210 (1996).
Article Google Scholar
Levine, J. D., Funes, P., Dowse, H. B. & Hall, J. C. Resetting the circadian clock by social experience in Drosophila melanogaster. Science 298, 2010–2012 (2002).
Article CAS PubMed Google Scholar
Schneider, J., Atallah, J. & Levine, J. D. Social structure and indirect genetic effects: genetics of social behaviour. Biol. Rev. 92, 1027–1038 (2017).
Article PubMed Google Scholar
Noor, M. A., Schug, M. D. & Aquadro, C. F. Microsatellite variation in populations of Drosophila pseudoobscura and Drosophila persimilis. Genet. Res. 75, 25–35 (2000).
Article CAS PubMed Google Scholar
Schneider, J. & Levine, J. D. Automated identification of social interaction criteria in Drosophila melanogaster. Biol. Lett. 10, 20140749 (2014).
Article CAS PubMed PubMed Central Google Scholar
Newman, M. Networks: An Introduction (Oxford University Press, 2010).
Latora, V. & Marchiori, M. Efficient behavior of small-world networks. Phys. Rev. Lett. 87, 198701 (2001).
Article CAS PubMed Google Scholar
Jezovit, J. A., Alwash, N. & Levine, J. D. Using flies to understand social networks. Front. Neural Circuits 15, 755093 (2021).
Article PubMed PubMed Central Google Scholar

Download references

Acknowledgements

The authors acknowledge Asher Cutter and Marla Sokolowski for their advice on experimental and statistical logistics of this study. The authors would also like to thank Joshua Krupp for his expertise in gas chromatography and Mireille Golemiec for her assistance in collecting the isofemale lines used in these experiments. This work was funded by Natural Sciences and Engineering Research Council, Canada Research Chairs Program (Tier-1), CIFAR Program on Child and Brain Development.

Author information

Authors and Affiliations

Department of Ecology and Evolutionary Biology, University of Toronto, Toronto, ON, Canada
Jacob A. Jezovit & Joel D. Levine

Authors

Jacob A. Jezovit
View author publications
Search author on:PubMed Google Scholar
Joel D. Levine
View author publications
Search author on:PubMed Google Scholar

Contributions

J.A.J. and J.D.L. conceived the project. J.A.J. performed all preparations, data collection and analysis. J.A.J. wrote the manuscript with input from J.D.L.

Corresponding author

Correspondence to Joel D. Levine.

Ethics declarations

Competing interests

The authors declare no competing interests.

Peer review

Peer review information

Communications Biology thanks Allen Moore and the other, anonymous, reviewer(s) for their contribution to the peer review of this work. Primary Handling Editor: David Favero. A peer review file is available.

Additional information

Publisher’s note Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.

Supplementary information

Supplementary Material

Description of Additional Supplementary File

Supplementary Data

Reporting Summary

Transparent Peer Review file

Rights and permissions

Open Access This article is licensed under a Creative Commons Attribution-NonCommercial-NoDerivatives 4.0 International License, which permits any non-commercial use, sharing, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence, and indicate if you modified the licensed material. You do not have permission under this licence to share adapted material derived from this article or parts of it. The images or other third party material in this article are included in the article’s Creative Commons licence, unless indicated otherwise in a credit line to the material. If material is not included in the article’s Creative Commons licence and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this licence, visit http://creativecommons.org/licenses/by-nc-nd/4.0/.

Reprints and permissions

About this article

Cite this article

Jezovit, J.A., Levine, J.D. Chemical signals and social structures strengthen sexual isolation in Drosophila pseudoobscura. Commun Biol 8, 76 (2025). https://doi.org/10.1038/s42003-025-07483-4

Download citation

Received: 15 February 2024
Accepted: 07 January 2025
Published: 17 January 2025
DOI: https://doi.org/10.1038/s42003-025-07483-4