The evolution of separate sexes from hermaphroditism is thought to have occurred independently many times, and may be linked to the evolution of sex chromosomes. Even though we have a good understanding of the theoretical steps in the evolution of sex chromosomes from a hermaphrodite ancestor, the initial stages are still hard to study in animals because many well-studied animal sex chromosome systems are old. We addressed this problem by experimentally selecting a hermaphrodite via sex-limited experimental evolution for several generations, simulating the early stages in the evolution of a sex chromosome. After 14 generations, a fitness assay revealed evidence of incipient sex role specialization in the female-selected lines, presumably reflecting the release from constraints usually imposed by selection on the other sex role. Importantly, however, this was not simply explained by morphology because testis and ovary sizes did not diverge among treatments. There was no evidence of a change in the male-selected lines. Our study shows that sex role specialization can occur rapidly as a result of sex-limited selection, which is consistent with genetic constraints between sex roles, and in line with the first predicted steps toward the evolution of a new sex chromosome system.

Experimental evolution protocol

The experimental evolution lines consist of four replicate populations (denoted 1-4) within each selection regime (female-limited, male-limited, and control), resulting in 12 populations in total (denoted F1-4, M1-4, and C1-4). Populations with the same replicate number are not more closely genetically related to each other than to other replicates, but they are related in terms of handling (i.e. culturing was during the same time of the day, and worms with the same replicate number were placed in the same area of the incubator).

A new generation starts with 48 mature, GFP(+) individuals from each replicate population being crossed with two worms each from the matched “source” population. Source populations are maintained in the same way as the other laboratory lines described above, and have the same genetic origin as the GFP(+) individuals from the same combination of treatment and replicate. They are used to provide mating partners for the GFP(+) individuals that neither carry the marker nor are exposed to the sex-limited selection. The trios of worms (1 GFP(+) and 2 GFP(-) individuals) are held in individual wells of 24-well tissue culture plates for one week, to provide opportunities for sperm competition and mate choice (figure 1). Therefore, the effective population size for each experimental line was approximately N_e ≈ N + 0.5 = 144.5 (i.e. 48 GFP(+) individuals and 96 GFP(-) individuals; Caballero 1994, Falconer and Mackay 1996), which is similar to several previous experimental evolution studies (e.g. Prasad et al. 2007, Michalczyk et al. 2011, Innocenti et al. 2014, Buechel et al. 2016). Only 48 individuals per population experience sex-specific selection each generation, but we expected that this selection would be sufficient, given the success of other experimental evolution studies similar to our design (e.g. Morrow et al. 2008).

After 7 days for mating, worms are isolated in new wells to lay eggs for one week. In the female-limited selection regime, the GFP(+) focal worms are isolated so that the GFP-marker is inherited via the female sex role, i.e. eggs. In the male-limited selection regime, the two GFP(-) mating partners are isolated for egg-laying, so that the GFP-marker is inherited via the male sex role, i.e. sperm. Finally, in the control lines, half of the selected worms are treated in the same way as the female-limited selection regime, and half are treated in the same way as the male-limited selection regime. Since the mating partners are not selected for one sex role or the other, the female-limited selection regime essentially mimics a gynodioecious system, while the male-limited selection regime mimics an androdioecious system. Egg-laying worms are discarded after one week, and eggs are left to hatch in the wells. After one week of growth, offspring in each well are moved to petri dishes to mature. If an experimental line does not produce sufficient numbers of GFP(+) offspring, backup offspring from the previous generation are used, so that generations are mainly non-overlapping, but not completely so.

During the first ten generations, each generation lasted for four weeks, but due to poor production of offspring within the female selection regime, we subsequently extended the maturation period to two weeks instead of one, which was successful in increasing offspring numbers. We based this decision on prior knowledge that in juveniles, testes mature slightly earlier than ovaries, so that female-limited selection could have resulted in longer ovary maturation times (Vizoso and Schärer 2007).

Fitness assay

Worms from the 14th experimental generation were collected and isolated in wells, directly after the completion of egg-laying to produce generation 15. More specifically, mating partners not used in the egg-laying were held in new wells instead of being discarded, and after egg-laying for generation 15 was complete, egg-laying worms were combined with the same partners again to provide 24 trios per selection regime and replicate. These trios of worms were again allowed to interact for 7 days. Next, all individuals were isolated in new wells to lay eggs for 7 days. Adult worms were then discarded, eggs were left to hatch, and juveniles to grow. The whole procedure (mating for 7 days followed by egg laying for 7 days, and growth of juveniles) was then repeated using the same trios of individuals, in order to increase total offspring production and thereby decrease the error in the individual fitness measurements. Fitness was measured as the number of GFP(+) and GFP(-) offspring per well (i.e. per GFP(+) focal individual). The total number of offspring produced via eggs was used as a measure of female fitness, and the proportion GFP(+) offspring produced by both GFP(-) mating partners was used as a measure of male fitness. This fitness assay builds on a standard fitness assay protocol commonly used in Drosophila (e.g. Abbott et al. 2013, Lund-Hansen et al. 2020) and is essentially the same as the sex-limited selection protocol, except for the fact that all worms are given the opportunity to lay eggs.

One possible source of error in our experimental set-up compared to the Drosophila studies mentioned above is the fact that the GFP(+) focal individuals used in the fitness assays are expected to be heterozygous for the marker, meaning that only half of all offspring produced by the GFP(+) focal individual are expected to inherit the GFP(+) phenotype. For female fitness, this is not an issue, since female fitness was measured as total number of offspring produced, regardless of GFP phenotype. However, it could result in an additional error in measures of male fitness, since in this case, the presence of the marker is essential for identifying the progeny of the GFP(+) focal individual. We have previously investigated rates of transmission of the GFP marker via eggs and sperm and found them to be highly correlated, though individually variable (Nordén and Abbott 2017), presumably due to variation in copy number at the GFP locus (Wudarski et al. 2017). An analysis of gene expression in these lines suggests that we have not inadvertently selected on GFP(+) expression as a byproduct of the selection protocol (Cīrulis 2022), so we therefore believe that neither variable expression of the GFP gene nor biased transmission depending on gamete type is likely to have important confounding effects on our measurements of male fitness. We measured sex-specific fitness for 267 individuals in total (91 from the control regime, 91 from the male-limited regime, and 85 from the female-limited regime).

Phenotypic measurements

Body size, relative testis size, and relative ovary size were estimated with a standard method used in this species (Schärer and Ladurner 2003). Briefly, worms were first isolated in wells containing f/2 solution and starved overnight. Each worm was then anesthetized for ten minutes in a well containing a mix of 600 μl artificial seawater (ASW) and 1 ml MgCl₂ –solution (conc 7.14 mg/ml). It was then slightly squeezed between a microscope slide and a cover slip, with pieces of plastic film of a standard thickness used as spacers in between (Schärer and Ladurner 2003). A picture of the whole body was then taken at x40 magnification, and of the ovaries and testes at x200 magnification. Photos were processed and analyzed in the program ImageJ (version 1.51), where the area of the body and area of the ovaries (sum of left and right) and the testes (sum of left and right) were calculated, respectively. Worms were photographed in random order, and the observer was blind with respect to the selection regime. We determined the repeatability of the morphological measurements by measuring 20 pictures three times for testes size, ovary size, and body area.

This assay was carried out at generation 25. After discarding poor-quality images, our final sample size was 121 for body area (40 from the control regime, 40 from the male-limited regime, and 41 from the female-limited regime), 116 for testes area (39 from the control regime, 37 from the male-limited regime, and 40 from the female-limited regime), and 115 for ovary area (40 from the control regime, 36 from the male-limited regime, and 39 from the female-limited regime). The repeatability of the morphological measurements was high (body: intraclass correlation coefficient, ICC = 0.998, p = < 0.001, testes: ICC = 0.940, p = < 0.001 and ovary: ICC = 0.882, p= < 0.001; Koo and Li, 2016, Vaz et al 2013).

Statistical analysis

All statistical analysis was carried out in R Version 4.1.0 (R Core Team 2021). For the fitness assay, we excluded data from worms that had no offspring in either sex role (which could for example occur if the worm was injured during handling) and individuals that had undefined male fitness (i.e. no offspring from either mating partner at all). The sex-specific fitness measures were standardized to make them directly comparable. Specifically, within each sex role we subtracted the mean within the control regime from all measurements, then divided by the pooled standard deviation. This standardization procedure results in the overall control mean being equal to zero, so that the difference in mean between the control and the female- or male-selected regimes reflects the magnitude of the change in sex-specific fitness measured in standard deviations. Changes in sex-specific fitness between selection regimes were analysed using a mixed model approach implemented in lme4 (Bates et al. 2015) with selection regime, sex role (male or female), and their interaction as fixed effects, and individual ID and replicate population nested within selection regime and sex role as random effects (Lund-Hansen et al. 2020, Manat et al. 2021). Since we have two types of sex-specific fitness measures per individual, the random effect of ID was to control for variance arising from overall fitness differences between individuals, and the nested effect of replicate population was included in order to avoid pseudoreplication (Arnqvist 2020). Posthoc comparisons were carried out using the pairs() function from emmeans (Lenth 2021).

Because gonad size is correlated with overall body size, we used relative gonad areas as a measure of investment in testes and ovaries (i.e relative testes area was calculated as testes area/body area, and relative ovary area was calculated as ovary area/body area). Differences in body area, relative testes area, and relative ovary area were each analysed with mixed models where selection regime was a fixed effect, and replicate population nested within selection regime was a random effect. Although here we report results from analyses using relative gonad sizes, results were qualitatively similar if body area was instead included as a covariate (see Results).

To examine whether there was a trade-off between investment in testes and ovaries, and if the magnitude of this trade-off differed between selection regimes, we first calculated the correlation between relative testes area and relative ovary area within each combination of selection regime and replicate population. We then used these 12 correlation coefficients as the dependent variable in a one-way anova analysis, with selection regime as the independent variable. The aim of this analysis was to test for consistent differences in the correlation coefficient between selection regimes. Since there was no evidence of any differences between treatments (see Results), we also tested if the mean correlation coefficient across all replicate populations was significantly different from zero using a one-sample t-test.

Finally, we tested whether differences in morphology could explain differences in sex-specific fitness. Since morphological data and sex-specific fitness data were collected at different times, this analysis could not be carried out on the individual level, so instead, we used replicate population mean values. We used a regression approach with sex-specific fitness as the dependent variable, and morphological variable (body area, relative testes area, or relative ovary area) as the predictor variable. The effect of selection regime was not included since replication was so limited in this analysis. However, results were qualitatively similar when carrying out an ancova analysis including both the morphological variable and selection regime as predictors (data not shown).