Evolutionary consequences of loss of sexual reproduction on male‐related traits in parthenogenetic lineages of the pea aphid

Transition from sexual reproduction to parthenogenesis constitutes a major life‐history change with deep evolutionary consequences for sex‐related traits, which are expected to decay. The pea aphid Acyrthosiphon pisum shows intraspecific reproductive polymorphism, with cold‐resistant cyclically parthenogenetic (CP) lineages that alternate sexual and asexual generations and cold‐sensitive obligately parthenogenetic (OP) lineages that produce only asexual females but still males. Here, the genotyping of 219 pea aphid lineages collected in cold‐winter and mild‐winter regions revealed contrasting population structures. Samples from cold‐winter regions consisted mostly of distinct multilocus genotypes (MLGs) usually represented by a single sample (101 different MLGs for 111 samples) and were all phenotyped as CP. In contrast, fewer MLGs were found in mild‐winter regions (28 MLGs for 108 samples), all but one being OP. Since the males produced by OP lineages are unlikely to pass on their genes (sexual females being rare in mild‐winter regions), we tested the hypothesis that their traits could degenerate due to lack of selection by comparing male production and male reproductive success between OP and CP lineages. Male production was indeed reduced in OP lineages, but a less clear pattern was observed for male reproductive success: females mated with OP males laid fewer eggs (fertilized or not) but OP and CP males fertilized the same proportion of eggs. These differences may stem from the type of selective forces: male production may be counter‐selected whereas male performances may evolve under the slower process of relaxed selection. The overall effective reproductive capacity of OP males could result from recent sex loss in OP lineages or underestimated reproductive opportunities.


| INTRODUC TI ON
Sexual reproduction is thought to have evolved in the common ancestor of eukaryotes, as it is highly frequent in this domain of life (Goodenough & Heitman, 2014). However, sexual reproduction was secondarily and independently lost in many groups, with loss of sex being occasional in protists, plants and fungi and rare in vertebrates (Bell, 1982;Simon et al., 2003;Vrijenhoek et al., 1989).
The transition from obligate sexual reproduction to parthenogenesis has many evolutionary consequences for phenotypic and genotypic characteristics. In particular, traits that evolved under sexual reproduction are expected to decay through selection or drift, depending on the cost of expressing the trait. As such, males produced by parthenogenetic females represent an energetic cost without any associated benefit when they are isolated from sexually reproducing populations. We thus expect natural selection to favour parthenogenetic females allocating fewer resources to male production. This has indeed been documented in several parthenogenetic species, including snails (Neiman et al., 2012) and Daphnia (Innes et al., 2000). Similarly, allocation into the male function in hermaphrodite mussel species is lower in populations with less outcrossing opportunities (Johnston et al., 1998).
When males still occur in exclusively parthenogenetic populations isolated from sexual counterparts, they are no longer under selection. No matter the potential reproductive performance of the males, they will not pass on their genes. Therefore, in the absence of pleiotropy, deleterious mutations are expected to accumulate through the slow process of drift leading to male sexual trait decay (Kraaijeveld et al., 2016). Evidence of degeneration is found in sperm production and morphology in several parthenogenetic species, including the snail Potamopyrgus antipodarum (Jalinsky et al., 2020), the parasitoid wasp Muscidifurax uniraptor (Gottlieb & Zchori-Fein, 2001) and the springtail Folsomia candida (Kampfraath et al., 2020). The ability to fertilize sexual females is altered at different degrees in males of several thelytokous parasitoid wasps (Adachi-Hagimori & Miura, 2020;Pannebakker et al., 2005;Zchori-Fein et al., 1992) as well as in asexual stick insects (Schwander et al., 2013).
Despite this evidence of degeneration, males produced by parthenogenetic species remain globally fertile (van der Kooi & Schwander, 2014). This may be explained by a recent loss of sex in these lineages, leaving insufficient time for relaxed selection to operate. Furthermore, the power of these hypotheses as explanations for the decay of male traits depends on the extent to which sons of parthenogenetic mothers have mating opportunities. Even a low rate of sex may be sufficient to prevent mutational degeneration of male genes as demonstrated by modelling in a hermaphrodite nematode (Chasnov & Chow, 2002), and rare opportunities for reproduction with sexual females may be particularly difficult to detect.
Aphids are appropriate organisms to test for sexual trait decay following sex loss as these insects present various modes of reproduction. The ancestral reproductive mode of this group, which evolved around 280 million years ago, is cyclical parthenogenesis (Grimaldi & Engel, 2005). This peculiar life-cycle consists of an alternation between many generations of parthenogenesis by viviparous parthenogenetic females in spring and summer and a sexual generation in autumn with males and oviparous sexual females (Dixon, 1998;Hales et al., 1997). The production of the sexual morphs is triggered by the autumnal decrease in daylight length . An XX/X0 sex determining system allows the production of males by parthenogenesis, with the elimination of one of the X chromosomes to produce a male (Blackman, 1975). Then, as males produce only X-bearing sperms, the fusion of male and female gametes leads to a diploid individual at the X, which develops into a parthenogenetic female.
Interestingly, several aphid species-including the pea aphid Acyrthosiphon pisum (Harris, 1776)-show genetically determined reproductive polymorphism, with lineages that are cyclically parthenogenetic (CP lineages) and others that have lost the ability to reproduce sexually (Dedryver et al., 2013;Jaquiéry et al., 2014;Simon et al., 2010). These obligately parthenogenetic lineages (OP) are not able to produce sexual females though they often retain the ability to produce males in response to day length shortening. Only CP lineages can survive cold winters because they produce oviparous sexual females, which lay frost-resistant eggs and remain in diapause for three to four months (Simon et al., 2002). In contrast, OP lineages are cold-sensitive and prevail in warmer areas, owing to their ability to maintain parthenogenetic development even in winter, which gives them a demographic advantage over CP lineages in such conditions (Simon et al., 2002). This divergent selection on the mode of reproduction by climate thus generally induces a geographical separation of the sexual (CP) and asexual (OP) lineages (Dedryver et al., 2001;Papura et al., 2003;Simon et al., 1999;Vorburger et al., 2003). Local adaptation to climate probably facilitated the long-term coexistence of these two reproductive modes .
Gene flow between OP and CP lineages is nevertheless likely to occur in different aphid species. First, the two types of lineages are expected to be sympatric in regions with intermediate climates (Dedryver et al., 2001;.
The high dispersal ability of aphids could further favour the cooccurrence of OP and CP lineages even in less suitable habitats.
Indeed, parthenogenetic females can develop wings under certain environmental conditions, and thus may be passively winddispersed (Loxdale et al., 1993). Second, although it is difficult to demonstrate by direct observation that OP males are also produced in nature (but see Halkett et al., 2008), we know that OP aphid lineages can produce males in laboratory conditions in response to sexinducing cues. OP males from the laboratory have been successfully mated with sexual females (Blackman, 1972;Dedryver et al., 2012;Jaquiéry et al., 2014), suggesting that crossings in the wild may also be achieved. Indirect evidence of gene flow between OP and CP lineages is supported by population genetic data: in the alfalfa host race of the pea aphid complex, OP and CP populations more than 450 km apart show low genome-wide genetic differentiation (F ST = 2.4%), except at an 840 kb genomic region associated with the reproductive phenotype Rimbault et al., 2022).
This suggests that gene flow between OP and CP lineages almost counterbalances genetic drift-which is probably moderate given the large sizes of aphid populations living on legume crops-and thus prevents genetic differentiation in genomic regions not linked with reproductive polymorphism.
Here, we first aimed at characterizing the genotypic diversity and genetic structure of clover-adapted pea aphid populations sampled in two regions of France with contrasting winter climates. By characterizing the reproductive phenotype of nearly a 100 lineages, we demonstrated extensive geographic segregation of the two reproductive modes, suggesting that males from OP lineages-which do not produce sexual females-may rarely encounter sexual females. In such a situation, male-related traits could be expected to degenerate as a result of relaxed selection or counter-selection. Our second aim was thus to investigate the fate of male traits-including male production and male mating success-in OP compared to CP lineages. The reproductive success of OP and CP males was measured with and without male-male competition to explore a range of conditions that might accentuate potential differences in performance. Finally, if OP lineages represent independent transitions to obligate parthenogenesis, their respective time since sex loss may differ, which could affect the intensity of degeneration of male traits. We thus examined the genetic relatedness between the OP and CP lineages to assess whether OP lineages derived from single or multiple sex loss events were closely related to CP lineages, which would suggest recent sex loss. We then tried to relate the genetic relationships between CP and OP lineages and the male trait performances of the different OP lineages.

| Aphid lineages and genetic analyses
A total of 563 A. pisum individuals were collected in clover fields lo-  Caillaud et al., 2004, Jaquiéry et al., 2012 following DNA extraction and PCR protocols described in Peccoud et al. (2008). Lineages that had the same genotype at all 15 loci or differed by at most one allele were identified with the R package RClone (Bailleul et al., 2016) and were assumed to be clonal copies of the same multilocus genotype (MLG). Only 28 different MLGs were found in the west (out of the 108 genotyped samples), while there were 101 different MLGs in the east (out of 111 genotyped samples). This dataset is available in Table S1. We kept alive only one representative for each MLG for all subsequent analyses. A subsample of 75 lineages (out of the 101 different MLGs) from the east and 24 lineages from the west (4 lineages died prior to this point) were then characterized in the laboratory for their mode of reproduction (OP versus CP, see below for detailed protocols). Finally, we assessed the genetic differentiation (F ST ) between eastern and western samples (keeping only one copy of each MLG, that is, 24 from the west and 75 from the east) on the 15 microsatellite markers using the hierfstat package (Goudet & Jombart, 2022) in R (version 4.0.3, R Core Team, 2019).
Expected heterozygosity (H e ) within each geographical region (using only one copy of each MLG) was computed using the adegenet package (Jombart, 2008) and the difference between regions tested with the function Hs.test (with 999 permutations). A principal component analysis (PCA) was also performed on lineages using the ade4 R package (Bougeard & Dray, 2018). Finally, we assessed the genetic relationship between OP and CP lineages using the shared allele distance D AS (Chakraborty & Jin, 1993). A neighbour-joining tree F I G U R E 1 Genotypic diversity of pea aphid (Acyrthosiphon pisum) samples collected in western (blue crosses) and eastern France (red crosses) in clover fields. Each slice of the pie charts represents one distinct MLG (multilocus genotype at 15 microsatellite markers), its size being proportional to the number of individuals sampled in the wild (and genotyped) having this MLG. The reproductive mode of each MLG is shown in light grey for OP (obligate parthenogenesis) and dark grey for CP (cyclical parthenogenesis). based on D AS was built using the ggtree R package (Yu et al., 2017),

Western
and bootstrap values were computed over 1000 replications with the aboot function (Kamvar et al., 2015) from the R package poppr (Kamvar et al., 2014).

| Characterization of the reproductive mode
To determine the reproductive mode (OP or CP) of the 99 lineages and to measure their investment in the production of the different morphs (i.e. males, sexual females and parthenogenetic females), lineages were reared under sex-inducing conditions (as described in Le  (Miyazaki, 1987) once they were adults. For each of the 99 lineages, three replicates of this experiment were performed. These analyses revealed that all the 75 lineages from the east were CP, and that all but one (i.e. 23) from the west were OP (relevant data is fully available in Table S2).
To assess whether the proportion of males produced by a lineage depends on its reproductive mode, a generalized linear mixed model (GLMM) assuming a binomial error and a logit-link function was fitted using the glmer function from the lme4 package (Bates et al., 2015). We considered the reproductive mode of the lineage as a two-level fixed factor and the lineage as a nested random factor within the reproductive mode (Equation 1). This random factor extracts the influence of aphid genotype on the dependent variables and removes any confounding effect between reproductive mode and aphid genotype, while accounting for variability within the reproductive mode. We added an observation-level random effect (OLRE, Harrison, 2014, 2015 to deal with overdispersion. The best model was selected based on term significance, which was determined by a Wald Chi-square test as a type III analysis of variance using the car package Anova function (Fox & Weisberg, 2018).
Traditional R 2 measures are not well suited to measuring the goodness of fit of mixed-effects models, which include random-effects variance components. We therefore calculated two appropriate measures of R 2 for mixed-effects models on the final model including only significant variables: the marginal R 2 (the variance explained by the fixed effects) and the conditional R 2 (the variance explained by both fixed and random effects, Nakagawa & Schielzeth, 2013).
This was done using the r 2 _nakagawa function from the R package performance (Lüdecke et al., 2021).

| Mating experiments without male-male competition
Mating experiments were conducted on the 24 lineages (12 OP and 12 CP lineages) that were still maintained in the laboratory at the start of the experiment (early 2020, Tables S1 and S3). These 24 lineages had been kept alive since 2018 because they were representative of the diversity of the 99 lineages. To produce a large number of males required for mating experiments, sex-inducing conditions similar to those described above were used. However, to optimize the production of males, tightly control their age and minimize handling, four individuals of the second-generation (those producing sexual morphs) were maintained on the same plant (instead of one per plant previously). The four individuals were transferred onto a new plant every one or two days (depending on the number of larvae laid) to avoid crowding. Two CP lineages (CP4 and CP10) were used to produce sexual females for mating experiments using the same protocol. We chose to work with two female genotypes to account for a possible female effect, but we could not test more due to handling constraints. By isolating males and females prior to sexual maturation (at the fourth larval stage), we ensured that sexual morphs were virgin at the start of experiments.
We first assessed the individual reproductive success of OP and CP males when alone with five sexual female partners (1:5 maleto-female ratio). We used 13-to 14-day-old virgin wingless males and females (hence sexually mature, Sack & Stern, 2007). First, we placed five females of the same lineage (either CP4 or CP10) and one male of each of the 24 lineages (either a CP or an OP lineage) on a faba bean plant grown in a tube filled with soil (2 cm diameter, 9.5 cm long). Plants were cut at the apical bud (leaving only the first two leaves) so that they all had similar leaf area and morphology allowing more contacts between male and females. The soil was covered with parafilm, to facilitate egg recovery at the end of the experiment. The plant was covered with a small drilled bag and this experimental setup was left untouched for 7 days under long day conditions (16 h of light, 18°C). After 7 days, the six individuals were removed from the plant. Fertilized and viable eggs laid by A. pisum females turn black through melanization after a few days whereas unfertilized or unviable eggs stay green (Blackman, 1987). aphid females' ovaries contain 14 ovarioles (Wieczorek et al., 2019) and one ovariole contains 32 germ cells, 11 of them developing into oocytes (Blackman, 1987;Miura et al., 2003). Sexual females can therefore lay a large number of eggs throughout their adult life, indicating that the number of eggs per female is not a limiting factor in our study. We observed that sexual females laid more eggs when they were in contact with more males (see results), which suggests that males (or mating) stimulate egg laying. Therefore, the total num-  Table S3. The variance explained by the different types of variables was measured through marginal and conditional R 2 as previously described.

| Mating experiments under male-male competition
To test the reproductive competitiveness of CP and OP males, we compared their respective reproductive success in a situation of strong male-male competition for females (10:5 male-to-female ratio). This competition experiment involved five males from different OP lineages (lineages OP1, OP2, OP3, OP4 and OP5), five males from four different CP lineages (CP1, CP2, CP3 and two males from CP4) and five females of the CP10 lineage. We only used CP10 as the female-producing lineage in this mating experiment as it produced the highest number of sexual females and could therefore be more easily synchronized with male production. The lineages used to produce males for this experiment were chosen because a sufficient number of same-aged males (13-to 14-day-old) was simultaneously available to perform six replicates. The even number of CP and OP males leads to the same competitive pressure between OP and CP males. However, two males from the same CP4 lineages (instead of one from CP4 and one from CP9) were mistakenly included in the six replicates of this mating experiment. This was considered in the statistical analysis. The 10 males were placed altogether with the five females on a faba bean plant cut at the apical bud, covered with a drilled bag and with soil covered with parafilm. After 7 days, all aphids were removed from the plant and preserved in 96% ethanol for microsatellite genotyping, in order to confirm their genotype.
Among the six replicates, only two males and two females were found dead at the end of the experiment (4.4%). The counting of the total number of eggs and melanized eggs was done 5 days later. Eggs were stored individually in Eppendorf tubes at −20°C before being manually crushed with a plastic pestle in a proteinase K lysis solution. DNA extraction was then performed as in Peccoud et al. (2008).
Each egg and adult were genotyped at 7 polymorphic microsatellite loci (AlA09M, AlB07M, AlB08M, AlB12M, ApF08M, ApH08M and ApH10M, Caillaud et al., 2004) following the conditions described in Peccoud et al. (2008). Each genotype is characterized by a specific combination of alleles at these highly polymorphic microsatellite loci, allowing paternity inference (Tables S1 and S4). Six replicates of this competition experiment, with exactly the same lineages, were carried out. Four of them were conducted in the laboratory while the two remaining ones were done outside the laboratory, in a room with less controlled conditions (12:12 L:D regime, ~18-20°C) due to the COVID-19 lockdown. Since two males from CP4 genotype were mistakenly put in each replicate, the number of eggs fertilized by these males was theoretically twice the number of eggs one male of this lineage could really fertilize. Therefore, half of the number of eggs fertilized by males from CP4 lineage were assigned to this lineage in the following analysis, to give the same weight to each male lineage. Relevant data is fully available in Table S4. The replicate number was included as a random effect, the twolevel factor location as fixed effect, and an observation-level random effect (OLRE, Harrison, 2014Harrison, , 2015 was included to deal with overdispersion.
The variance explained by the different types of variables was measured through marginal and conditional R 2 as previously described.

| Comparison of reproductive success with and without competition
To examine whether the fitness of males from the different lineages was consistent regardless of competition level, we calcu-

| Male production in OP and CP lineages
We then analysed the investment of the 23 OP and 76 CP lineages into the production of males, sexual and parthenogenetic females in more detail. All three types of morph could be produced by CP lineages, whereas OP lineages never produced sexual females and produced less males and more parthenogenetic females than CP lineages ( Table 1, Figures S1). The total fecundity (i.e. the sum of all morphs produced) was slightly higher for CP compared to OP lineages (GLMM, p = .03, Table 1). The frequency of lineages that produced males was higher in CP lineages than in OP lineages, but this was not significant (chi 2 test, χ 2 = 1.369, df = 1, p = .24, Table 1).
Furthermore, the averaged proportion of males produced by OP lineage was smaller than for CP lineages (GLMM, p = .015, Tables 1 and 2, Figure 3a). The contribution of both fixed and random effects to the variance of this GLMM model was 39.9% (given by conditional R 2 ). This value partitioned into 3.6% due to fixed effects (i.e. the reproductive mode, given by marginal R 2 ) and 36.3% to random effects (i.e. the male lineage and observation-level random effects, obtained by subtracting the marginal R 2 from the conditional R 2 ).

| Male reproductive success without competition
The individual reproductive success of males from 12 OP and 12 CP lineages was then assessed. After seven days of contact between the male and the five females, eggs were found in 98.4% of the 246 replicates. In total, 6056 eggs were identified, of which 77% were fertilized. Fertilized eggs were found in half of the replicates involving OP lineages and two-third of those involving CP lineages ( Table 3, Table S3).
Among the 246 replicates, from zero to 103 eggs were found per tube. The best-fitting GLMM explaining the number of eggs laid revealed a significant effect of the reproductive mode of the male lineage (p = .005), with fewer eggs laid by females sired by OP males (4) logit(proportion of fertilized eggs) ∼ reproductive mode + location + (1|reproductive mode : male lineage) + (1|replicate) + (1|OLRE) than by CP males (Tables 2 and 3, Figure 3b). The effect of the reproductive mode of the male lineage on the number of eggs laid also varied with the female lineage (p = .048): CP4 females were more fecund than CP10 females, but both laid less eggs when paired with OP males than with CP males ( Table 3). No effect of the location where the experiments took place was observed (p = .14). The contribution of both fixed and random effects to the variance explained by the model was 45.4% (given by conditional R 2 ). This value partitioned into 14.8% due to fixed effects kept in the final model (i.e. reproductive mode, female lineage and the interaction between these two variables, given by marginal R 2 ) and 30.6% to random effects

| Male reproductive success under male-male competition
After 7 days of contact between the 10 males and the five females, 483 eggs were found in the six replicates (corresponding to an aver-  were performed on 426 eggs for which the father was identified (Table S4). From six up to nine male lineages (7.5 in average) reproduced among the six replicates (that each involved 9 different male lineages). No significant difference was observed in the proportion of eggs fertilized by males according to the reproductive mode of their lineage (GLMM, p = .36, Table 2): CP males sired on average 11.95% of the fertilized eggs and OP males 8.05% of them ( Figure 3d). We also found no effect of the location (in the laboratory vs outside the laboratory) of the mating experiments on the proportion of fertilized eggs (p = .78, Table 2). The contribution of random effects (i.e. male lineage and observation-level random effects) to the variance explained by the model was 18.8%.

| Male reproductive success with and without competition
Male mating success (measured here as the number of fertilized eggs) with and without male-male competition correlated positively but not significantly for the nine male lineages that were used in both experiments (Spearman rank correlation r S = 0.51, n = 9, p = .16, Figure S2). The average number of fertilized eggs per tube was 71.7 in the experiments with male-male competition (hence each male fertilized an average of 7.17 eggs under competition) and 19.1 without competition (which corresponded to the number of eggs fertilized by single males). The increased female reproductive success in the presence of many males (the number of females was constant between the two experiments) indicated that male resource was a limiting factor for female reproduction.

| DISCUSS ION
Here, we showed that sexual (CP) and asexual (OP) lineages of the clover host race of the pea aphid complex are geographically well separated. We then investigated the fate of sexual traits in a subset of these obligately asexual lineages, which retain male function despite low chance of encountering sexual females. OP lineages

Males of CP lineages (n = 12) a
Average number of eggs laid per female b 3.41 6.00

TA B L E 3
Reproductive success of males produced by obligately parthenogenetic lineages (OP) and cyclically parthenogenetic lineages (CP) of Acyrthosiphon pisum in the absence of competition.
showed a significant reduction of the production of males compared to CP lineages but the males produced were still largely functional.
We detected, however, a small decline of OP male fitness that was significant only for the number of eggs laid.

| OP and CP lineages are well separated geographically
Under first, the phenotype of western and eastern populations drastically differs in this host race of the pea aphid. Such climate-dependent segregation of aphid reproductive phenotypes has also been found in other species (Dedryver et al., 2001;Simon et al., 1999;Vorburger et al., 2003), and could affect the communities of predators and parasitoids as a result of the difference in availability of aphids as prey or hosts during winter (e.g. Le Ralec et al., 2010).  Frantz et al., 2010), though parthenogenetic females can be winged. Therefore, this system allows the hypothesis that a large reduction of reproductive opportunities could lead to a degeneration of male sexual traits in asexual lineages to be tested.

| OP lineages produce fewer males than CP lineages
We first showed that OP lineages produced significantly fewer males than CP lineages in response to sex-inducing conditions. This reduction in male production may be the result of selection against a trait that a shift to obligate parthenogenesis renders useless and costly in the absence of sexual partners. Indeed, the production of offspring is energy-constrained, with trade-offs between the number of sexual and asexual morphs produced in aphids (Nespolo et al., 2009). In OP lineages, we did observe a reallocation of investment in males to parthenogenetic females, as male production decreased while the number of parthenogenetic females increased compared to CP lineages. Maternal control of sex determination in aphids (based on the elimination of an X chromosome to produce a male) makes this system particularly prone to rapid sex ratio evolution because there are no longer Mendelian constraints to bypass as in XX/XY or ZZ/ZW species. Nevertheless, OP lineages produced slightly fewer offspring than the CP lineages. Whether this is due to different costs in producing male or female embryos is unknown (we do not know if weight at birth differs between males and females since larvae cannot be sexed at birth). Alternatively, OP lineages could suffer from an overall reduction in fitness because the lack of recombination prevents the purging of deleterious mutations (e.g. Lynch et al., 1993).

| Males produced by OP lineages maintain good reproductive performance
In contrast, we found very little evidence for reduced functionality of males. First, a large majority of OP males fertilized eggs, demonstrating that male fertility is maintained. Second, only the number of eggs laid by a female in the presence of OP males was significantly reduced, with a small effect size. All other measures showed no difference in reproductive performance between OP and CP males, even under strong male-male competition where one would expect an exacerbation of differences, if any (e.g. Schwander et al., 2013).
The reduction of the propensity of the females to lay eggs in the presence of OP males is not straightforward to interpret. As we observed that contact/mating with males seems to trigger egg laying, the number of eggs laid might be indicative of the propensity of a female to lay her eggs in the presence of a male. It can be expected that a male with altered reproductive functions will not be as stimulating to the female as a male whose same functions are under strong selection because of their importance on fitness.

| Different evolutionary trajectories for costly and neutral traits following loss of sex
The reduction in male production by OP lineages while maintaining their overall reproductive capacity is probably related to the selective mechanisms at play. Male production by OP lineages represents a net cost in the absence of female partners and therefore should be rapidly counter-selected, whereas male degeneration occurs only through relaxed selection. This supports predictions and observations that counter-selected sexual traits degenerate more rapidly than traits evolving under relaxed selection (van der Kooi & Schwander, 2014).

| Origin and number of sex loss events
The low branch resolution prevents assertion of whether the three main clusters of OP lineages in the D AS tree represent independent or shared sex loss events. Thus, we cannot rule out that the majority of our lineages derive from the same transition to obligate asexuality, which would imply that each OP lineage should not be considered a statistically independent observation. Yet, another lineage (OP2) clearly clustered with CP lineages suggesting that transitions to obligate parthenogenesis could have occurred at least twice in this host race. This finding relies heavily on the OP2 lineage, so it is crucial to be certain of its phenotype. This lineage was determined as OP under sex-inducing conditions in two independent experiments and a total of 6 individual copies of this clone were found in clover fields 2 km apart, a pattern that also supports asexual reproduction. Further sampling of wild individuals is necessary to get a more comprehensive picture of the frequency of such sex loss events. Also, our data does not inform on how sex is lost. New OP aphid lineages have been shown to originate from multiple genetic mechanisms depending on species: (i) de novo mutation(s) at the gene(s) controlling the production of sexual morphs (Delmotte et al., 2001), (ii) contagious asexuality resulting from crosses between OP males and CP sexual females (Halkett et al., 2008;Jaquiéry et al., 2014), (iii) hybridization between closely related species . The ancestor of OP2 could have lost the ability to produce sexual females by spontaneous mutation or by contagious asexuality, the resolution of our tree being insufficient to disentangle the two hypotheses, especially if the op allele(s) is (are) recessive as in the alfalfa host race of the pea aphid complex . The third mechanism is unlikely in our system as the OP and CP lineages do not differ in heterozygosity, as would have been the case if OP lineages had been generated by interspecific crosses.

| CON CLUS ION
This study shows that the transition from cyclical to obligate parthenogenesis results in changes in selective regimes between CP and OP populations with various evolutionary consequences: they manifest as (1) a reduction in the proportion of males produced by OP lineages, supposedly resulting from counter-selection, and (2) a maintenance of egg fertilizing capacity of OP males, supposedly resulting from slow drift-driven decay or underestimated reproductive opportunities. Sampling in more intermediate habitats would help to determine the extent of distribution overlap between the OP and CP lineages and better characterize potential gene flow. Other major issues in this study are whether and how quickly a trait will decay when selection exerted on it is removed or relaxed, and how this will relate to the trait functions and the mechanisms involved. Our study contributes to show that species adaptation following a change in reproductive system also involves the decay of useless or maladaptive traits. To expand this analysis to other functional traits, future studies should consider the evolutionary consequences of sex loss at the transcriptomic and genomic scale. This would enable us to better apprehend and quantify the decay or maintenance of sexual traits under asexuality, and especially those related to sexual females, that are not exposed to selection in OP aphids.

ACK N O WLE D G E M ENTS
We warmly thank Editor Tatiana Giraud and three anonymous referees for constructive comments on previous drafts, as well as Chris Bass for his critical reading.

CO N FLI C T O F I NTER E S T S TATEM ENT
The authors declare no conflicting interest.

DATA AVA I L A B I L I T Y S TAT E M E N T
The authors confirm that all relevant data underlying the findings are fully available without restriction, within the paper and its supporting information files. Relevant data are fully available in Tables S1 to S4.