All experiments were performed using three different fly species (Figure 1A; Siomava et al., 2016). We used the highly inbred laboratory strain Drosophila melanogaster w1118, which was kept at 18°C on standard food (400 g of malt extract, 400 g of corn flour, 50 g of soy flour, 110 g of sugar beet syrup, 51 g of agar, 90 g of yeast extract, 31.5 ml of propionic acid, and 7.5 g of Nipagin dissolved in 40 ml of Ethanol, water up to 5 l). The other two flies were Musca domestica wild type ITA1 collected in Italy, Altavilla Silentia in 2013 (Y. Wu and L. Beukeboom, GELIFES, Netherlands) and Ceratitis capitata wild type Egypt II (IAEA). The M. domestica strain was reared at room temperature (RT) (22 ± 2°C) on food composed by 500 g of wheat bran, 75 g of wheat flour, 60 g of milk powder, 25 g of yeast extract, 872 ml of water, and 18.85 ml of Nipagin (2.86 g of Nipagin in 10 ml of Ethanol). Adult M. domestica flies were provided with sugar water. C. capitata were kept at 28°C, 55 ± 5% RH on a diet composed by 52.5 g of yeast extract, 52.5 g of carrot powder, 2 g of Sodium benzoate, 1.75 g of agar, 2.25 ml of 32% HCl, and 5 ml of Nipagin (2.86 g of Nipagin in 10 ml of Ethanol), water up to 500 ml for larvae. For adult flies, we used a 1:3 mixture of sugar and yeast extract.

To generate a range of sizes for each species, we varied two environmental factors known to influence overall body size – temperature and density. Prior to the experiment, D. melanogaster flies were placed at 25°C for two days. On the third day, flies were moved from vials into egg-collection chambers and provided with apple-agar plates. After several hours, we started egg collection by removing apple-agar plates with laid eggs once per hour. Collected plates were kept at 25°C for 24 h to allow embryonic development to complete. Freshly hatched first-instar larvae were transferred into 50 ml vials with 15 ml of fly food. Three vials containing 25 larvae each (low density) and three vials with 300 larvae each (high density) were moved to 18°C; the second set of six vials with the same densities was left at 25°C.

Ceratitis capitata flies were kept at 28°C and allowed to lay eggs through a net into water. Every hour, eggs were collected and transferred to the larval food. After 22 h, first-instar larvae were transferred into small Petri dishes (diameter 55 mm) with 15 ml of the larval food in three densities: 25 (low density), 100 (medium density), or 300 (high density) larvae per plate. Two plates of each density were moved to 18°C. The second set of six plates was left at 28°C for further development.

Eggs of M. domestica were collected in the wet larval food at RT and after 24 h, all hatched larvae were removed from food. Only larvae hatched within the next hour were transferred into 50 ml vials with 5 g of food. Collection of larvae was repeated several times to obtain two experimental sets with three replicates of three experimental densities 10 (low density), 20 (medium density), or 40 (high density) larvae. One set of nine vials was moved to 18°C, the other was left at RT.

After pupation, individuals of C. capitata and M. domestica were collected from the food and kept until eclosion in vials with a wet sponge, which was refreshed every other day.

The experimental temperature regimes were chosen for the following reasons. D. melanogaster is known to survive in the range 10 to 33°C, but flies remain fertile at 12 to 30°C with the optimum at 25°C (Hoffmann, 2010). Reproduction temperatures in C. capitata range from 14°C to 30°C with the optimum at 28°C (Duyck and Quilici, 2002; Navarro-Campos et al., 2011). M. domestica flies survive at 10 to 35°C (Hewitt, 1914) with the optimum between 24 and 27°C (Hafez, 1948; Chun-Hsung, 2012). The low temperature for our experiment was chose as the one above the survival and fertile minimums for all three species – 18°C. The warm temperature was aimed to be optimal for each species.

For each combination of temperature and density, we randomly picked 8 to 24 (median: 12.5) flies for C. capitata, 20 to 31 (median: 23) flies for D. melanogaster and 5 to 15 (median: 8) flies for M. domestica. Both wings were dissected, embedded in Roti^®-Histokitt II (Roth, Buchenau) on a microscope slide, and photographed under the Leica MZ16 FA stereo microscope with the Q Imaging Micro Publisher 5.0 RTV Camera.

Wing shape was analyzed using landmark-based geometric morphometric methods (Rohlf, 1990; Bookstein, 1991). We digitized 10 anatomically homologous landmarks on wings of the three species (Figure 1A). The landmarks were the following [nomenclature is given after (Colless and McAlpine, 1991)]: 1, branching point of veins R₁ and R_S (base of R₂₊₃ and R₄₊₅); 2, branching point of veins R₂₊₃ and R₄₊₅; 3, intersection of veins C and R₁; 4, intersection of vein R₄₊₅ and crossvein r-m (anterior crossvein); 5, intersection of crossvein r-m and vein M₁₊₂; 6, intersection of vein M₁₊₂ and crossvein i-m (posterior crossvein); 7, intersection of crossvein i-m and vein M₃₊₄; 8, intersection of M₃₊₄ and the wing margin; 9, intersection of veins C and R₂₊₃; 10, intersection of veins C and R₄₊₅.

Wing images were digitized using tpsUtil and tpsDig2 (Rohlf, 2015) to obtain raw x and y landmark coordinates. Using superimposition methods, it is possible to register landmarks of a sample to a common coordinate system in three steps: translating all landmark configurations to the same centroid, scaling all configurations to the same centroid size, and rotating all configurations until the summed squared distances between the landmarks and their corresponding sample average is a minimum scaling (Slice, 2005; Mitteroecker et al., 2013). To follow these three steps, we applied the generalized Procrustes analysis (GPA) (Dryden and Mardia, 1998; Slice, 2005) as implemented in MorphoJ (version 1.06d) (Klingenberg, 2011). The wings were aligned by principal axes, the mean configuration of landmarks was computed, and each wing was projected to a linear shape tangent space. The coordinates of the aligned wings were the Procrustes coordinates. To evaluate wing asymmetry, we used the R package Geomorph (v. 3.3.1) (Adams et al., 2021) to partition shape variation by individuals, side (directional asymmetry) and the interaction between individual and side (fluctuating asymmetry). Procrustes ANOVA was used to evaluate statistical significance. Since we obtained no evidence for fluctuating asymmetry for all species (Supplementary Table 1), we averaged the coordinates for the right and left wings for further analyses. Next, we used Type III Procrustes ANOVA (with Randomization of null model residuals and 1,000 permutations) as implemented in Geomorph (v. 3.3.1) (Adams et al., 2021) to determine the effect of the species, sex, temperature, and density as well as potential interactions between these explanatory variables and shape (Table 1). Temperature and density values were re-labeled as low, high, or intermediate to allow the comparison between species. To visualize the main components of shape variation, we performed a principal component analysis (PCA) and generated wireframes using MorphoJ (version 1.06d) (Figure 1B).

Since most of the variation in shape was explained by differences between species (see section “Results”), we split the analysis to further evaluate the effects of sex, rearing temperature, and density as well as potential interactions on wing shape within each species using Procrustes ANOVA in Geomorph (v. 3.3.1) (Supplementary Table 2). Magnitudes of sexual shape dimorphism were estimated using the discriminant function analysis (DFA) and expressed in units of Procrustes distance using MorphoJ (version 1.06d). DFA identifies shape features that differ the most between groups relative to within groups and it can only be applied to contrast two experimental groups. Therefore, we used this method to define sexual shape dimorphism (males and females), as well as effects of the rearing temperature (high and low) and density (high and low) in each species. Wing shape changes were visualized using wireframe graphs. To test for the significance of the observed differences, we ran a permutation test with 1,000 random permutations (Good, 1994) for each test using MorphoJ (version 1.06d).

To evaluate the impact of wing size on shape variation, we estimated wing centroid size (WCS) that was computed from raw data of landmarks and measured as the square root of the sum of squared deviations of landmarks around their centroid (Bookstein, 1996; Dryden and Mardia, 1998; Slice, 2005) in MorphoJ (version 1.06d). The wing images were taken at different magnifications to account for the different body sizes of the three species. C. capitata wings were takes at a 1,280 × 960 resolution, M. domestica at a 2,560 × 1,920 resolution and for D. melanogaster images with both resolutions were obtained. Therefore, the WCS values were corrected for differences in magnification and resolution among photos before being used as covariate. The effects of sex, temperature, density, and their interactions on WCS was tested using Type III ANOVA for each species.

Since a Type III Procrustes ANOVA for all three species including WCS as covariate revealed significant interaction between species and WCS (Supplementary Table 3, sheet “All Species_total”), we estimated the non-allometric shape component for each species independently. The non-allometric Procrustes coordinates were obtained as the residuals of the multiple regression of WCS onto the Procrustes coordinates pooling the samples by sex, temperature, and density, as implemented in MorphoJ (version 1.06d) (Klingenberg, 2011, 2016). To evaluate the success of the size correction, we ran a Type II ANOVA testing for additive effects of explanatory variables prior to and after the size correction. This analysis revealed that WCS had no impact on wing shape after size correction (Supplementary Table 3, compare sheets “^∗_total” to “^∗_non-allometric”), suggesting that the approach accounted for most of the impact of size on shape. Note that species-specific Type III Procrustes ANOVAs including WCS as covariate revealed significant interactions for C. capitata and D. melanogaster (Supplementary Table 3, sheets “C. capitata_total,” “D. melanogaster_total”), suggesting complex relationships between wing size and shape. Since the regression approach to estimate allometry across all explanatory variables (i.e., sex, temperature, and density) assumes similar slopes of within-group regressions (Gidaszewski et al., 2009; Klingenberg, 2016), we analyzed those regressions in more detail. Although the directions of within-group regressions (i.e., regression scores of the regression of Procrustes coordinates onto WCS pooled by sex, temperature, and density) were comparable (Figure 3), type III ANOVA testing for interactions of explanatory variables partially resulted in significant interactions in C. capitata and D. melanogaster (Supplementary Table 4), suggesting some significant differences in slopes. Therefore, the distinction between allometric and non-allometric shape differences may not be fully accurate. However, due to a lack of a better method, we decided to proceed with the analysis and to interpret the results cautiously. The amount of shape variation explained by wing centroid size was estimated by obtaining R² for the linear model describing the pooled regression (i.e., the unique curve after pooling the data by groups: centered_reg_scores ∼centered_WCS).

Magnitudes of non-allometric shape differences for each explanatory variable and for sex-specific effects of different temperature and density conditions were estimated using DFA as described before. Since wing size and shape of intermediate densities for C. capitata and M. domestica overlapped with the high- and low-density conditions, respectively (Figure 3), we restricted the detailed shape analysis to the extreme values.

To characterize wing shape variation among Ceratitis capitata, Drosophila melanogaster, and Musca domestica, we applied geometric morphometrics based on 10 landmarks (Figure 1A). Procrustes ANOVA revealed that most of the shape variance was caused by differences among species (Table 1). This result was also reflected by the principal component analysis (PCA) which clearly distinguished the three species with the first two principal components (PCs) accounting for almost 98% of the variation (Figure 1B). The main shape difference along PC1 (77.3% of the variation) reflected the ratio between the proximal and distal parts of the wing as well as wing length. The wireframe graphs showed that C. capitata wings were broad in the proximal part (landmarks 1–5) and narrow in the distal part (landmarks 6–10), while D. melanogaster wings showed the opposite pattern with the proximal part being heavily compressed along the anterior-posterior axis. M. domestica had an intermediate morphology being, however, more similar to C. capitata (Figure 1B, wireframe graphs along PC1). PC2 explained 20.5% of the variation mainly accounting for the ratio between the length and width of the whole wing. C. capitata and D. melanogaster showed wider and shorter wings, while M. domestica showed more elongated wings (Figure 1B, wireframe graphs along PC2). Together, we found evidence for extensive wing shape variation among C. capitata, D. melanogaster, and M. domestica.

Our interspecific shape analysis revealed a clear sexual shape dimorphism which was most pronounced for C. capitata (Figure 1B and Table 1). The extreme sexual shape dimorphism in C. capitata and the observation that males and females occupy different relative positions along PC1 compared to D. melanogaster and M. domestica likely explain the significant interaction between species and sex (Table 1). Species-specific Procrustes ANOVA for all three species revealed that within-species shape variation was affected independently by sex and rearing conditions (i.e., temperature and density, see below) (Supplementary Table 2). Therefore, we split the analysis by species and ran a DFA to determine the total shape differences caused by sex (Figures 2A–C). For all three species we found a highly significant sexual shape dimorphism (Figures 2A–C). Male wings were broader than those of females in C. capitata (Figure 2A) and D. melanogaster (Figure 2B), while the opposite trend was observed in M. domestica (Figure 2C). Wings of C. capitata females were slightly longer than male wings (Figure 2A) and in D. melanogaster and M. domestica male wings were longer (Figures 2B,C). The radio-medial crossvein (r-m) defined by landmarks 4 and 5 was different between males and females in C. capitata (Figure 2A). In D. melanogaster we observed clear sexual differences in the radial vein R2 + 3 defined by landmarks 2 and 9 and in the basal-madial-cubital crossvein (bm-cu) defined by landmarks 6 and 7 (Figure 2B). In summary, we found a clear total sexual shape dimorphism in all three studied species.

Since sex had a significant influence on wing size in all three species (Table 2, see also Figure 3) (see Siomava et al., 2016 for independent wing size measures) we next asked how much of the shape differences between sexes was explained by differences in wing size. Wing size explained 16% and 14% of wing shape variation in C. capitata and M. domestica, respectively (Figure 3). Wing shape was more clearly associated with differences in wing size in D. melanogaster (34%; Figure 3). To analyze the non-allometric shape differences between sexes in more detail, we accounted for wing size (Klingenberg, 2016; see section “Materials and Methods” for details). A species-specific DFA using the non-allometric shape component revealed significant sexual shape dimorphism for all three species (Figures 2D–F). This observation was confirmed by a Procrustes ANOVA (Supplementary Table 2). In line with the minor impact of wing size on wing shape, the extant of the total compared to the non-allometric sexual shape dimorphism was very similar in C. capitata and M. domestica (Figures 2D,F). Accordingly, the differences in shape decreased after size correction in D. melanogaster (Figure 2E). Specifically, differences in the length and the width of the wings were largely reduced in the non-allometric shape component. These results suggest that major difference between sexes observed in C. capitata and M. domestica can be explained by the non-allometric shape component, while wing size differences contribute to the sexual shape dimorphism in D. melanogaster.

To evaluate the effect of wing size on shape in the three species in more detail, we raised flies at different temperatures and densities. The Procrustes ANOVA for all three species revealed significant interactions between species and density and temperature, respectively (Table 1), suggesting species-specific effects of the rearing conditions on wing shape. Density had significant effects on wing size in all three species, while temperature affected only D. melanogaster and C. capitata (Table 2). Since temperature and density contributed additively to wing shape differences (Supplementary Table 2), we therefore split the analyses by species and performed species-specific DFA to study the total shape differences caused by different rearing temperatures (Figures 4A–C) and densities (Figure 5A), respectively. Additionally, we calculated the non-allometric component of shape (see section “Materials and Methods” for details) and analyzed the differences using DFA (Figures 4D–F for temperature and Figure 5B for density).

In accordance with the species-specific Procrustes ANOVA testing for effects of rearing conditions on total wing shape variation (Supplementary Table 2), the DFA clearly assigned wings to one of the two rearing temperatures (Figures 4A–C). The most obvious effect of different rearing temperatures on wing shape was observed in D. melanogaster (Figure 4B). Wings of D. melanogaster flies raised at higher temperatures were wider in the distal-central region defined by landmarks 7–9 and distally shortened (i.e., displacement of landmark 10) (Figure 4B). The distal contraction remained after size correction, while the width was much less affected (Figure 4E). This observation suggests that temperature-dependent plasticity in wing width is predominantly caused by differences in wing size. Different rearing temperatures also affected distal-central wing width in C. capitata (Figure 4A) and M. domestica (Figure 4C). Narrower wings were observed at higher temperatures in M. domestica, while wings were narrower at lower temperatures in C. capitata (Figures 4A,C). M. domestica wings were longer at higher rearing temperatures (i.e., displacement of landmark 10) (Figure 4C). Only minor changes were present between total and non-allometric shape differences in C. capitata, suggesting that temperature-dependent size differences had small effects on wing shape in this species. In line with non-significant effects of temperature on wing size (Table 2), we observed no differences between the allometric and non-allometric shape in M. domestica (Figures 4D,F), In all three species, we observed temperature-dependent plasticity in the placement of the radio-medial (r-m) crossveins (defined by landmarks 4 and 5), the basal-medial-cubital (bm-cu) crossveins (defined by landmarks 6 and 7), the radial vein R2 + 3 (defined by landmarks 2 and 9) and the anterior cubital (CuA1) veins (defined by landmarks 7 and 8) (Figure 4).

As suggested by the Procrustes ANOVA (Supplementary Table 2), the DFA showed that different rearing densities only significantly affected total wing shape in D. melanogaster (Figure 5A and Supplementary Table 5). Wings of flies raised at high densities were wider in the central-distal region (defined by landmarks 7–9) and elongated (i.e., displaced landmark 10) (Figure 5A). We observed plasticity in the placement of the basal-medial-cubital (bm-cu) crossveins (defined by landmarks 6 and 7), the anterior cubital (CuA1) veins (defined by landmarks 7 and 8), the radial vein R2 + 3 (defined by landmarks 2 and 9) and the costal (C) vein (defined by landmarks 3 and 9) (Figure 5A). All these differences were gone after size correction (Figure 5B), suggesting that wing shape differences in response to rearing densities were predominantly associated with variation in size. The displacement of landmark 3 was consistent in both analyses (compare Figures 5A,B), implying that this region of the wing may be affected by different rearing densities independent of wing size. Despite a weak effect of density on C. capitata wing shape in our Procrustes ANOVA (Supplementary Table 2), the DFA did not reveal significant shape differences for flies raised at high and low densities (Supplementary Table 5). The same was true for M. domestica, supporting the Procrustes ANOVA results (Supplementary Table 2). Note that we observed significant differences in wing shape at different rearing densities after size correction in C. capitata and M. domestica in our Procrustes ANOVA (Supplementary Table 2) and in the DFA (Supplementary Table 5).

In summary, we showed that different rearing temperatures had a profound effect on wing shape D. melanogaster and minor effects were observed in C. capitata and M. domestica. Variation in rearing densities only clearly affected wing shape in D. melanogaster.

Since the non-allometric component of shape was consistently affected by different rearing conditions (Supplementary Tables 2, 5), we next asked for each species, whether a sexual dimorphism in response to different rearing temperatures and densities exists. In all three species, we did not observe significant interactions between sex and rearing conditions in our Procrustes ANOVA of size corrected shape variables (Supplementary Table 2). Accordingly, in C. capitata and D. melanogaster a DFA revealed very similar non-allometric shape differences between high and low temperatures (Supplementary Figure 1 and Supplementary Table 5) and densities (Supplementary Figure 2 and Supplementary Table 5), respectively, for both sexes. Despite insignificant interaction terms in the Procrustes ANOVA the effect of interactions between sex and rearing conditions was up to 9 times higher in M. domestica compared to the other two species (see sum of squares in Supplementary Table 2; “^∗_non-allometric”-sheets). Interestingly, the DFA revealed a significant effect of temperature (Figures 6A,B and Supplementary Table 5) and density (Figures 6C,D and Supplementary Table 5) on non-allometric shape differences in males, while females were unaffected.

In summary, we found no evidence for a sexual dimorphism in the response to different rearing conditions in C. capitata and D. melanogaster. In contrast, we found indications for a sexual dimorphism in the non-allometric shape differences between extreme rearing temperatures and densities in M. domestica.

We raised flies at different temperature and density regimes and analyzed the impact on wing size, wing shape and the relationship of both. Overall, we revealed a few common, but mostly species-specific patterns of plastic responses.

In accordance with previous observations (Bitner-Mathé and Klaczko, 1999; Gilchrist et al., 2000; Gidaszewski et al., 2009; de Camargo et al., 2015; Siomava et al., 2016; Pieterse et al., 2017; Lemic et al., 2020; Rohner, 2020; Cortés-Suarez et al., 2021), we detected a clear sexual dimorphism in wing size and shape in all three species, which was most pronounced in C. capitata. Sexual dimorphism in wing shape is most likely functionally relevant because it is widespread in insects (Cowley et al., 1986; Pretorius, 2005; Bogdanović et al., 2009; Gidaszewski et al., 2009; Ribak et al., 2009; Allen et al., 2011; Benítez et al., 2011; de Camargo et al., 2015; Gallesi et al., 2015; Virginio et al., 2015; Lorenz et al., 2017; Rodríguez and Liria, 2017; Pajač Živković et al., 2018). In C. capitata and D. melanogaster, males and females differed mostly in wing width. Males of both species had wider wings in favor of more efficient buzzing during courtship (Burk and Webb, 1983; Webb et al., 1983; Wheeler et al., 1988; Talyn and Dowse, 2004; de Souza et al., 2015). A correlation between male wing shape and mating success has been established in C. capitata because males with wider wings copulated more often successfully (de Souza et al., 2015). Interestingly, males with elongated wings have a higher mating success in D. melanogaster (Menezes et al., 2013), showing that intra-sex variation in wing shape are caused by species-specific processes. Indeed, both species differ in the frequency and quality of their courtship song (Cowling and Burnet, 1981; Burk and Webb, 1983; Briceño et al., 2002; Rybak et al., 2002). In contrast to C. capitata and D. melanogaster, we found the most obvious sexual differences in M. domestica in wing length. The slightly elongated wings in males may increase flying efficiency and could be linked to the mating strike (Murvosh et al., 1964; Alves and Bélo, 2002).

Among the three studied species we observed different effects of wing size on sex-specific wing shape. Exclusion of the allometric coefficient clearly decreased the sexual shape dimorphism in D. melanogaster, suggesting that most of the observed shape differences could be explained by differences in wing size. In contrast, sex had only a minor effect on wing size in C. capitata and M. domestica (Siomava et al., 2016) and accordingly, the impact of the allometric component on wing shape was weak. Species-specific sexual dimorphism in scaling relationships were also observed in five schizophoran Diptera (Rohner, 2020) and in Sphingidae moths, where up to 60% of the shape variation could be explained by wing size differences (de Camargo et al., 2015). An extreme example has been observed in the aphid parasitoid Ephedrus persicae, where male wings are twice as large as female wings although females have generally larger body sizes. In this species, only 5% of the sexual shape differences can be explained by wing size (Bogdanović et al., 2009), suggesting that exaggerated male wing size may be more advantageous than sex-specific wing shape. All this data implies that sexual dimorphisms in wing shape scaling may reflect species-specific adaptations rather than conserved patterns across insects.

Since sexually dimorphic organs have been shown to grow disproportionally with body size in one sex in response to environmental cues (Teder and Tammaru, 2005; Bonduriansky, 2007; Lavine et al., 2015; Siomava et al., 2016), we tested for sexually dimorphic patterns of plasticity. We did not find evidence for a sexual dimorphism in the response to different rearing temperatures and densities in all three studied species. It is important to note that our experimental design is particularly limited with respect to this question because the number of individuals in each relevant group (i.e., split by sex, density and temperature) is very low and we might not have the statistical power to detect any general trends. However, our DFA implies that non-allometric shape differences between extreme rearing temperatures and densities observed in M. domestica may be restricted to males. This trend contradicts a previous observation in this species that showed a significant positive correlation between wing size and wing width with latitude specifically in females of different populations from South America (Alves and Bélo, 2002). This discrepancy could be explained by the use of an Italian strain in our experiments. Future experiments addressing these specific questions should be performed to obtain more conclusive results.