Introduction

Sexual reproduction is characterized by the production of genetically diversified individuals, resulting from genetic shuffling that occurs during meiosis through recombination, ploidy reduction, and effective fusion of two different pronuclei1. This mode of reproduction has been a successful adaptation already present in the last eukaryotic common ancestor2 and provided populations with an efficient mechanism to cope with fluctuating environments3,4. On the opposite, many organisms with fully sexual ancestors showed the reappearance of asexual reproduction as either a facultative or obligatory process5. This mode of reproduction is achieved by a wide variety of processes across taxa, ranging from mitotic divisions (as in fission, fragmentation, budding) to different modified meiotic pathways (as in parthenogenesis or gynogenesis)6,7. This flexibility of processes is especially obvious in many hybrid complexes, where the accumulation of subtle incompatibilities between parental haplomes can cause aberrant gametogenesis, leading to a continuum of forms between sexual and asexual organisms5,8.

Although asexual reproduction appears to have many advantages despite the low or absent genetic variation among individuals, sexual reproduction still dominates in most eukaryotes, suggesting that benefits must outweigh costs9, a counterintuitive fact described as the paradox of sex. Considerations of the costs and benefits of sexuality and asexuality have often been of demographics, population genetics, or genetic nature10,11,12,13. However, another underappreciated component of this paradox is the developmental cost of each reproductive strategy to the individual itself. For example, meiotic reduction in animals results in eggs that are smaller than their asexual counterparts14,15,16. This either reduces the cost of producing a given number of eggs or results in a higher number of eggs for the same energetic cost17. Thus, asexual females are expected to make a trade-off in their energy allocation, potentially leading to inherent changes in their development. Furthermore, asexual reproduction leads to variations in or around the meiotic process, which in turn result from differences in gene expression compared to their sexual counterparts18,19,20,21, which may consequently affect their development.

In addition to the reproductive structures themselves, does asexuality inherently alter the shape of somatic traits in animals? While this question has been the subject of many studies, none has provided a satisfactory answer due to multiple confounding factors. For example, studies comparing sexual and asexual organisms exclusively in different species18, lineages22, or ploidies23 leave the effects of genetic variation intertwined with those of reproductive strategies. Sexual selection has also been shown to confound results. Because parthenogenetic females do not require a mate to reproduce, sexual selection is relieved and may explain differences with closely related sexual species or lineages24,25,26.

Thus, the inherent shape of reproductive strategies can be elusive and difficult to isolate. To our knowledge, no published study in animals has the potential to distinguish whether the observed divergence between asexual and sexual females is due to their reproductive strategy or to confounding factors. In this paper, we aim to determine whether sexual and asexual gametogenesis are associated with an inherent shape.

To address this objective, we performed geometric morphometric analyses in the Chrosomus eos-neogaeus fish complex, which is characterized by the presence of all-female hybrids with distinct reproductive pathways (Fig. 1). We focused on two groups of triploid hybrids that exhibit both sexual and asexual reproduction, and in which the known confounding factors are absent.

Fig. 1: Chrosomus eos-neogaeus fish complex.
figure 1

Hybridization events between C. eos (EE) and C. neogaeus (NN) resulted in diploid hybrids (EN) that reproduce asexually by gynogenesis: unreduced eggs requiring sperm from males of a close species, such as C. eos, to initiate development66. The sperm haplome (E’) may or may not be incorporated into the embryo (paternal leakage), resulting in either triploid individuals (E’EN) or diploid clones identical to their mother (EN)66,67,68. The first generation triploids reproduce asexually: they reject the sperm haplome (E’) before meiosis and produce eggs with the hybrid genome (EN), in a process called ameiotic hybridogenesis 27,28. Again, the sperm haplome (E*) may or may not be incorporated into the embryo, leading to triploid (E*EN) or diploid individuals. These second-generation triploids reproduce sexually: they reject the C. neogaeus haplome (N) before meiosis and produce reduced C. eos haploid eggs in a process called meiotic hybridogenesis 27,28. Fertilization by a C. eos sperm results in the production of C. eos individuals. See28 for a more extensive description of the complex and27 for a more detailed description of the gametogeneses employed by the members of this complex. Legend: Solid lines indicate the fusion of the paternal genome with the maternal genome in the embryo; dashed lines indicate the exclusion of the paternal genome from the embryo. Figure adapted from28.

Full size image

The sexually and asexually reproducing triploids of this complex are a prime model for the purpose of this study. They both occurred following the integration of a C. eos haplome into a clonal hybrid genome. Thus, in a given hybrid lineage, triploid individuals differ from each other by one haplome inherited from their fathers. However, since the paternal genome comes from the same population of C. eos, there is no difference in the genetic composition of the two groups of triploid hybrids. Both triploid groups show modified gametogenesis starting with the rejection of one haplome to form diploid cells, followed by meiosis leading to the production of gametes with or without genetic reduction. All triploid hybrids are also expected to be under similar sexual selection since they are all females and must compete for the sperm of the same males to trigger the development of their eggs. They reproduce at the same time and location27, and spermatozoids from males of a close species (often C. eos) are required for either cytosolic components (asexual triploids) or as additional genetic material (sexual triploids). Thus, sexual selection is expected to be homogeneous among all-female hybrids. Finally, such triploid lineages cannot have developed specific adaptations to their asexual or sexual mode of reproduction since neither group persists more than a single generation in nature28. Therefore, if triploids exhibit shape differences, these are expected to spontaneously result from factors associated with the reproductive strategies.

In this work, we used geometric morphometrics analysis to show a clear shape differentiation between sexually and asexually reproducing individuals. This difference between reproductive strategies is consistently detected in genetically different hybrid lineages, under different environmental conditions, and when compared to genetically different members of this complex encompassing different ploidies (C. eos and diploid hybrids).

Results

Shape differs between triploids

We first tested the hypothesis of morphological differences resulting from reproductive strategy by analyzing the shape of these genetically similar triploids in their natural environment. We compared the morphometric portrait based on 9 landmarks and 40 semi-landmarks of the anterior part of sexual and asexual adult triploid hybrids (Fig. 2a). To increase the reliability of the results, we analyzed two genetically distinct lineages from different habitats: A-11 (creek) and B-01 (lake)29.

Fig. 2: Shape variation of sexual and asexual triploid females.
figure 2

a Landmarks position. Red dots and blue lines represent the landmarks (1–9, numbered in the white circles) and the semi-landmarks, respectively, used in this study. For the semi-landmarks, 15 were used on the chin curve and 25 on the operculum curve. b, c Shape variation. Principal Component Analysis (PCA) of the two hybrid lineages (A-11 and B-01, shape) with individuals originating from two natural habitats (creek and lake, respectively) (b) or from common garden conditions (c), and according to their reproductive strategy (color). In both (b) and (c), small and semitransparent data points represent individuals and plain and large data points with error bars represent group (reproduction mode and lineage) centroids and their standard deviation. Error bars were computed from all individuals within a given lineage, reproduction mode, and environmental condition: (b) A-11/Asexual: n = 9; A-11/Sexual: n = 14; B-01/Asexual: n = 26; B-01/Sexual: n = 33 and (c) A-11/Asexual: n = 8; A-11/Sexual: n = 4; B-01/Asexual: n = 5; B-01/Sexual: n = 4 (Supplementary Table II). While a single PCA was performed among all samples, two panels according to environments were used to represent the ordination, for more clarity. Convex hulls and extreme shapes of the PCA are shown in Supplementary Fig. 1. Source data for this figure can be found in Supplementary Data 1 and Supplementary Code 1.

Full size image

We first determined how triploids clustered, according to their shape, by partitioning individuals of each lineage into two groups (expected to discriminate sexual and asexual individuals) using a k-means clustering analysis. This partitioning resulted in a high proportion (65.2% for A-11, n = 23 and 81.4% for B-01, n = 59) of triploid individuals successfully assigned to their own reproductive strategies considering only the analyzed morphological traits. Even though individuals belonged to distinct lineages (genetically different) and originated from different habitats (lake/creek), a principal component analysis (PCA) showed that the main axis of variance (PC1 = 30.93%) in the measured shape was related to the reproductive strategies, while genetic differences between lineages were observed only along the second axis (PC2 = 23.54%; Fig. 2b). A multivariate linear regression by redundancy analysis (RDA) confirmed reproductive strategy as the most important factor explaining shape variation (Adjusted R2 = 10.20%, p ≤ 0.0001, Supplementary Table I) compared to lineages (AdjR2 = 4.81%, p ≤ 0.0001) and size (AdjR2 = 2.13%, p = 0.0055). These results suggested a specific shape associated with sexual and asexual reproduction, that could be discriminated from the effects of lineages/habitats.

Differential survival does not drive shape

To determine whether differences between hybrids resulted from the effects of reproductive strategies on the development of these traits, we first had to rule out the possibility that morphological differences were driven by differential survival due to the influence of environmental conditions. To this end, we used individuals of the same two lineages reared together in a common garden setting from a previous study30. Nearly all individuals transferred from natural conditions at the larval stage survived the five months of the experiment, suggesting no selection due to the artificial environment31. The next analyses were, therefore, to determine whether the morphological differences observed in natural conditions persisted in such a common garden.

All individuals reared in controlled conditions were successfully assigned to their reproductive strategies using the k-means clustering analysis (A-11 n = 12; B-01 n = 9), and a clear difference in the shape of sexual and asexual individuals for both lineages can be observed along the first axis of the PCA (Fig. 2c), though the sample size was smaller compared to naturally reared individuals. Again, an RDA confirmed the determining effect of the reproductive strategies (AdjR2 = 22.17%, p ≤ 0.0001, Supplementary Table I), whereas genetic differences between lineages had only a limited effect on shape (AdjR2 = 3.96%, p = 0.0337). In the absence of mortality during the experiment, such discrimination between sexual and asexual organisms ruled out the possibility that morphological differences in triploids were the result of differential survival. This suggested that the effects of reproductive strategies on development were responsible for the difference in their traits.

Effects of reproductive mode on shape

We then assessed which traits among all the landmarks were specifically affected by sexual and asexual reproductive strategies. To disentangle the effect of reproductive strategies from the effect of the environment on shape variation, we compared individuals of both lineages from natural and common garden conditions. First, we used a model selection based on constrained ordination to objectively select a subset of explanatory variables to explain the measured morphological variation. We confirmed the determining effect of reproductive strategies (AdjR2 = 8.83%, p ≤ 0.0001, Supplementary Table I), in addition to size (AdjR2 = 1.70%, p = 0.0033) and environment × lineage interaction (AdjR2 = 1.95%, p = 0.0007). The low but significant environment × lineage interaction suggested a different natural to common garden morphological change for individuals of the B-01 lineage, compared to those of the A-11 lineage (Fig. 2b, c).

Using a partial RDA to assess the marginal effect of reproductive strategy, but controlling for other covariates, we detected a significant divergence between sexual and asexual organisms at 4 landmarks and 20 semi-landmarks (Fig. 3a). Using sexual individuals as a reference, the eye position (#1) and the top of the operculum (#7) of asexual individuals were located more dorsally, while the pectoral fin (#6) and the end of the operculum (#9) were displaced ventrally. This resulted in the development of a dorsoventrally larger operculum in the asexual individuals. In addition, semi-landmarks describing the curve of the operculum (from 25 to 49) were significantly different between reproductive strategies (except #39-43 in the middle of the structure), highlighting the global shift of the operculum curve toward the posterior region of the asexual females. These differences were large enough to be detectable on specimens to the naked eye (Fig. 3b).

Fig. 3: Shape changes according to reproductive strategy.
figure 3

a Marginal effect of reproductive strategy on shape variation. Landmark deviation from sexual to asexual forms (arrows, not magnified), when controlling for size and environment × lineage interaction. Green and red arrows represent respectively significant (p ≤ 0.05) and nonsignificant (p > 0.05) differences between sexual and asexual organisms for each landmark. Landmarks were analyzed by a partial RDA, conditioned for the environmental condition (natural/common garden), lineage, and size (allometry). P-values were computed by ANOVA and adjusted by a Bonferroni correction for multiple tests. b Asexual and sexual individual head photographs. Typical lineage B-01 asexual triploid (left) and sexual triploid (right), from the natural habitat are shown to highlight the difference at the level of the operculum. The graph paper behind the subjects shows 1 mm2 squares. Source data for this figure can be found in Supplementary Data 1 and Supplementary Code 1.

Full size image

The structures in the anterior part of the head (nostril, mouth, and chin) did not change significantly according to the reproductive strategies. However, they are strongly influenced by genetic composition and environmental conditions, as observed previously30,32. These results underline the heterogeneous response of the analyzed structures due to genetic and environmental components, while the difference between sexual and asexual organisms remained the same.

Gametogenesis changes shape

The morphological differences detected between sexual and asexual triploids could arise from two distinct mechanisms: the mechanism of targeting and excluding one of the haplomes (i.e., the C. neogaeus haplome in sexual hybrids; the C. eos haplome in asexual hybrids), or the mechanism related to gametogenesis itself (i.e., reduction or non-reduction of gametes content). To disentangle these mechanisms, we included in the analyses two diploid members of the complex as references for reproductive strategies: the sexual parental species C. eos and asexual diploid hybrids (B-01 lineage only), from both natural and common garden conditions. Since these additional forms did not reject a haplome prior to meiosis27, our prediction is that if the factors responsible for the differences in shape are inherent to the production of unreduced gametes, both diploids and triploids will exhibit the same morphology for a given reproductive strategy.

Maternal effects are also known to determine the emergence of either sexual or asexual triploids in this fish complex28. Maternal ploidy, either diploid or triploid, leads respectively to asexual and sexual triploid females. If maternal ploidy has an influence on the morphology of the progeny, diploid hybrids, which can also arise from a diploid or triploid mother28, would be expected to split into two groups clustering with sexual and asexual triploids. Adding diploid hybrids to the study then also allows the discrimination of these effects.

Using the landmarks significantly correlated with reproductive strategy, we determined whether reproductive strategies explained shape variation across these members of the fish complex. Variation partitioning indicated that differences between sexual and asexual members explained the higher proportion of the total variation (AdjR2 = 8.21%, p ≤ 0.0001, Supplementary Table I). Environmental conditions (AdjR2 = 5.79%, p ≤ 0.0001), lineages (AdjR2 = 2.53%, p ≤ 0.0001) and size (AdjR2 = 2.29%, p ≤ 0.0001) were also retained. Such a major effect of the reproductive strategy was illustrated by sexual individuals occupying a different morphospace than asexuals, sexually reproducing individuals being located to the right of asexual organisms in the PCA ordination (Fig. 4).

Fig. 4: Shape variation according to reproductive strategies in members of C. eos-neogaeus fish complex.
figure 4

Principal Component Analysis (PCA) from all individuals, including the two hybrid lineages (A-11 and B-01, shape) with individuals originating from two natural habitats (creek and lake, respectively) or from common garden conditions, and according to their reproductive strategy (color). Small and semitransparent data points represent individuals and plain and large data points with error bars represent group (reproduction mode and lineage) centroids and their standard deviation. Error bars were computed from all individuals within a given ploidy, lineage, reproductive mode, and environmental condition (natural/common garden): Natural/Triploids/A-11/Asexual: n = 9; Natural/Triploids/A-11/Sexual: n = 14; Natural/Triploids/B-01/Asexual: n = 26; Natural/Triploids/B-01/Sexual: n = 33; Natural/Diploids/B-01/Asexual: n = 32; Natural/Diploids/C. eos/Sexual: n = 17; Common Garden/Triploids/A-11/Asexual: n = 8; Common Garden/Triploids/A-11/Sexual: n = 4; Common Garden/Triploids/B-01/Asexual: n = 5; Common Garden/Triploids/B-01/Sexual: n = 4; Common Garden/Diploids/B-01/Asexual: n = 14; Common Garden/Diploids/C. eos/Sexual: n = 27 (Supplementary Table II). While a single PCA was performed among all samples, four panels (according to environments and ploidy) were used to represent the ordination, for more clarity. Source data for this figure can be found in Supplementary Data 1 and Supplementary Code 1.

Full size image

Following our prediction, the ploidy variable was not retained by the multivariate model selection (nor its interaction with other components), indicating that differences between diploids and triploids do not affect the shape, especially in asexual individuals. In fact, for a given condition, asexually reproducing hybrids (both diploid and triploid) could not be morphologically distinguished (pairwise analyses p > 0.8984, Supplementary Table I). Sexual triploid hybrids were different from C. eos in natural conditions (p = 0.0004), but not in the common garden (p = 0.2360). Both sexual forms thus have the same morphology, except that it is also affected by the effects of plasticity associated with distinct habitat use in natural conditions. Overall, these results showed a clear morphological convergence of traits according to reproductive strategy, demonstrating that trait differences are not unique to triploid hybrids.

Discussion

Using geometric morphometrics analyses, we have shown that either sexually or asexually reproductive mode shape individual morphology, regardless of environmental conditions and individual ploidy. Notwithstanding a smaller sample size in the common garden compared to naturally reared individuals, we ruled out differential survival as a cause of shape differences between sexual and asexual organisms, since individuals reared in a common garden show the same patterns. Furthermore, diploid and triploid asexually reproducing hybrids reared in a given environment could not be morphologically distinguished, indicating that the additional genome of triploid hybrids, as well as the mechanism leading to the exclusion of the paternal haplome prior to meiosis, have no detectable effect on the measured morphological traits. The absence of two distinct groups of diploid hybrids also ruled out maternal ploidy as responsible for these morphological differences. Therefore, while maternal effects are responsible for the different reproductive forms in triploids28, they do not seem to directly influence their respective morphology. Taken together, these results indicate that the morphological differences are correlated to the altered gametogenesis in the asexual hybrids.

How these shape changes occurred in Chrosomus eos-neogaeus hybrids and whether they are direct modifications or a by-product of the modification of another trait has not been addressed and remains unclear, opening the doors to further research in developmental biology. While morphological differences between asexual and sexual individuals are expected to result from specific epigenetic profiles and gene expression, the interactions between C. eos and C. neogaeus genomes in a hybrid genomic context result in admixture, additivity, and transgressivity30,33. The comparison made with diploid sexual C. eos and asexual hybrids of the same fish complex strongly supported the effect of the reproduction mode on the head morphology, as trait differences are not unique to triploid hybrids, but are found globally as an inherent shape of reproductive strategy in females harboring at least one C. eos haplome. It would be relevant to parallelize the morphological differences observed here with a system of females harboring at least one C. neogaeus haplome (C. neogaeus and C. neogaeus × eos-neogaeus triploids). Another potential avenue regards the ova production. Indeed, in order to produce clonal eggs, modification of gametogenesis is required, often including mechanisms that allow duplication of the genome of the reproductive cells prior to meiosis (premeiotic genome endoreplication)34. The resulting unreduced (or partially reduced) oocytes are then larger and less abundant than those of sexually reproducing females28. This fundamental difference with sexual organisms is expected to lead to a trade-off between either less abundant, larger eggs, or more abundant, smaller eggs. The consequence of such a dichotomy in gametogenesis may lead to a domino effect, influenced by hormones and gene expression, to the observed variation in development. Furthermore, different gametogenic processes can lead to a different number of surviving ova, as observed in the hybrids of Squalius35, Cobitis34,36, and Darevskia37 complexes. Different reproduction modes could thus lead to differences in the number of gametes growing to a mature stage, and thus to similar consequences as exposed before.

Morphological differences between sexual and asexual individuals mostly affect operculum shape, eyes, and pectoral fin position. Asexually reproducing females showed a proportionally larger operculum than their sexually reproducing counterparts (with the landmarks of the top and bottom of the operculum being positioned respectively more dorsally and ventrally compared to the other group), a more dorsal position of the center of the eye, and a more ventral position of the pectoral fin. The shape of the operculum is known to be ecologically relevant in trophic and habitat specialization of fishes, affecting, for example, their attractiveness to mates, swimming performance, niche adaptations, detection by prey, and susceptibility to predation38,39,40. Similarly, eye and pectoral fin positions could act on size/distance perception41, and thus on diverse interactions with the environment, and swimming speed and feeding habits38, respectively.

The results of this paper have thus far reached implications well outside the complex studied. The inherent developmental differences between sexually and asexually reproducing individuals suggest that they may occupy different ecological niches. A niche shift resulting from the development of a specific morphology according to reproductive strategies in genetically similar groups of hybrids is expected to reduce competition between sexual and asexual organisms, which in turn may alleviate the costs of asexual and sperm-dependent organisms on the sexual populations. Indeed, in addition to negatively impacting the demography of the sexual species by stealing the gametes of the males, they also compete with them for other resources (Janko, et al. 42 and references therein). In both cases, niche shifting may facilitate the coexistence between sexual and asexual organisms by reducing competition, but also by limiting the demography of asexual individuals according to the productivity of that particular niche. The findings of this paper suggest that asexually reproducing hybrids may actually have a lesser impact on sexual females than previously thought, by occupying a somewhat different ecological niche than the sexual species.

Morphological changes may also be a potential cost in asexual organisms. Changes in traits associated with gametogenesis could impair the ability of individuals to interact with the environment, thereby reducing fitness as a result. This may, in part explain why asexual lineages that emerge in a sexually reproducing species have a hard time establishing themselves43.

These findings have broad implications for how sex and its costs are viewed. The paradox of sex focuses mostly on how offspring are produced, what Maynard Smith10 called the “cost of males”, and its genetic diversity to face environmental changes. The results of this study suggest an additional potential cost of asexual reproduction: an ecological niche shift or a reduction in fitness in the niche of sexual organisms, adding more layers to the paradox of the sex concept.

Methods

Samples

This research was performed under institutional animal care guidelines (permit #13-084, #19-021, #20-022, and #21-023 issued by the Université de Montréal) and conforms to the mandatory guidelines of the Canadian Council on Animal Care. The Quebec Ministry of Natural Resources and Wildlife (MRNF) provided sampling permits.

All individuals used in this study were already part of the laboratory’s ethanol-preserved collection and were sampled according to institutional animal care guidelines and in accordance with the mandatory guidelines of the Canadian Council on Animal Care. Hybrid and C. eos fish were collected from Desjardins Lake (B-01 lineage; Quebec, Canada; 45.91693, -74.07292) and Halfway Creek (A-11 lineage; Quebec, Canada; 45.04337, -72.36199)30. The sample totaled 193 females (Supplementary Table II). Within a given sampling site, all hybrids belong to the same lineage29; therefore, all diploids are clonal, and triploid individuals only differ by their one paternal C. eos haplome.

Specimens from natural environments were sexually mature and collected during their reproductive season, while those from the common garden were younger (~ 5 months old)30. To select only true diploid and triploid females, ploidy was confirmed using six microsatellite markers: four on the C. eos genome and two on the C. neogaeus genome29. Suspected mosaic individuals (with a mixture of diploid and triploid cells44) were excluded from the study. The reproductive phenotype of triploid individuals (sexual or asexual) was assessed according to the egg size following the procedure in Lafond and Angers28.

Morphological data acquisition

High-resolution digital photographs were taken on the left side of the anterior part, with specimen orientation following the horizontal axis. Landmarks were placed and samples scaled using the tpsDIG2 program45 (Supplementary Data 1), by a single observer to minimize measurement error/variation. Samples were randomly organized to reduce bias in landmark positioning.

The shape was assessed with 9 homologous “fixed” landmarks and 40 sliding semi-landmarks46,47,48,49 (Fig. 2a). Position of the landmarks is the following: (1) Eye center, identified at the intersection of the maximum dorsal and ventral diameters of the eye, (2) Top, posterior part of the nostril (3) Intersection of the upper lip and snout, (4) Intersection of the upper and lower lips, (5) Most dorsal mouth corner, (6) Intersection of the most dorsal pelvic spin with the body, (7) Dorsal start of the operculum, (8) Vertical projection of the mouth corner on the ventral margin (described in the study as “chin”), and (9) Horizontal projection of the pectoral fin origin on the operculum (described in the study as “bottom of the operculum”). Landmarks were chosen for their adequacy to describe both the trophic niche (mouth shape) and the individual interaction with the environment (eye and pectoral fin position)30, and were based on similar study cases30,50,51. Semi-landmarks, used to describe the chin and operculum curves, were used because they have been suggested to increase statistical power in a similar setting52. While the relationship between the chin and trophic niche is more obvious, the operculum curve was used as its relationship to the trophic niche has been previously established in cichlids53. The two curves were delimited respectively by landmarks 4–6 and 7–9 and resampled to be equidistant. The number of semi-landmarks in both curves was chosen in a way to precisely define the shape without losing details in more complex parts of each anatomical part. While semi-landmarks may not always add much new information in statistical analyses beyond a certain number, they typically add anatomical resolution, which can be helpful in interpretation and shape reconstruction, and exceeding this threshold does not affect the results of nonparametric statistical tests46,52.

To perform downstream analyses and compare individuals purely on their shape, landmark coordinates were adjusted by a General Procrustes Analysis (GPA)54 so that size, orientation, and position were normalized among all photographs. Minimum bending energy was used to slide landmarks during GPA55. Measurement error due to 2D parallax and landmark positioning was assessed by digitizing twice 34 randomly selected individuals, and thus placing landmarks on a total of 68 pictures. Then, a linear model was produced with Individual Identity as the predictor. This analysis revealed that 96.1% (p ≤ 0.0001) of the variation was detected between individuals, leaving a small amount of within individual variation56.

Statistical analyses

To assess whether triploids were different according to their reproductive strategies, k-means clustering analyses57 were performed with two groups reflecting sexual/asexual reproduction. The success of each analysis was then calculated as the proportion of individuals correctly assigned to their own reproductive strategy. Principal Component Analyses (PCA) were performed to visualize shape variation across the first principal components. We used multivariate linear models to extract the relative effects of the main variables. First, we used a model selection analysis to highlight the factors that best explained shape variation. This function used an automatic stepwise model for constrained ordination methods, with permutation tests in both forward and backward selection. We then assessed the proportion of variation explained by the factors selected in the previous step, expressed as the adjusted R2 (AdjR2) and whose significance was assessed by permutation tests using 9999 randomizations58. We used centroid size (square root of the summed squared distances between landmarks and the centroid of the individual, as a proxy for individual size to control for potential allometry59), ploidy (diploid/triploid), environmental conditions (natural/common garden), lineages (A-11, B-01, C. eos), reproductive strategies (sexual/asexual) as well as their interactions as explanatory variables, and according to the tested hypothesis (Supplementary Table II), and shape coordinates after GPA as the response variable.

To visualize shape changes, we plotted the mean shape of both reproductive strategies of all triploid individuals, with each landmark connected by an arrow showing the trajectory of change. To assess the significance of differences in landmark positioning between sexual and asexual individuals, we performed a partial RDA with reproductive strategies as the main fixed factor and controlled for other factors selected by the previous automatic stepwise model (i.e., centroid size and environment × lineage interaction included as conditional variation in the analysis). Adjustment for multiple testing was performed using a Bonferroni correction.

Statistical analyses were performed in RStudio (version 2023.6.1.524)60, running on R4.1.3 R Core61. GPA and mean shapes were computed using the geomorph package (V4.0.3)62,63. Model selection and RDA analysis were performed using ordistep and rda functions from the vegan package64. For other analyses (ANOVA, p.adjust, and pr.comp), visualization, and figures, packages stats61 and ggplot2 65 were used.

Reporting summary

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