Genetic variation in age-dependent attractiveness in a fish with a mixed mating system

Abstract

Reproductive senescence is common across taxa and females often show a predictable decline in fecundity after maturity. Attending to these age-dependent cues could help males make optimal mate choice decisions. Here, we examined reproductive senescence and male mate choice in the androdioecious mangrove rivulus (Kryptolebias marmoratus), where self-fertilizing hermaphrodites exist with rare males. Hermaphrodites showed a strong decline in fecundity as they aged and genetic lineages varied in their fecundity at both young and old ages. Surprisingly, when given a simultaneous choice between genetically identical old and young hermaphrodites, males did not simply prefer younger hermaphrodites. Instead, male preference for younger versus older partners depended on the genetic lineage of the partners, resulting in a strong genotype × age interaction. For some genetic lineages, hermaphrodites were more attractive to males when younger, but for other genetic lineages, hermaphrodites were more attractive when older. Our results suggest that the genetic identity of the partner is key to how males weigh age-dependent changes in fecundity and that males are able to assess genetic variation in attractiveness over a partner’s reproductive lifespan. Exploring how gamete viability and outcrossing are affected by age across genetic lineages could help us further understand these male preferences.

1. Introduction

In many species, females show a predictable decline in fecundity with age beyond maturity, i.e. reproductive senescence [1–3]. Older females often produce fewer offspring [4–8] and lower-quality offspring [9,10]. Offspring of older mothers can also have lower survival and reproduction later in life [11–14]. Thus, female age can be particularly important in determining the reproductive potential and attractiveness of a particular mate.

Although less studied than female mate choice, there is evidence that male mate choice is common across taxa [15–19] and is often based on female traits that are correlated with fecundity [17–21]. Despite reproductive senescence causing female fecundity to decline with age, there is mixed evidence of how male mate choice is affected by female age. In some species, males prefer younger, more fecund females over older females [6–8,22]. Additionally, males often prefer virgin females [16], and in nature, these females will usually be younger than previously mated females. However, in other species, males prefer older females or show no age preferences, despite age-related declines in fecundity [5,23].

Understanding how fecundity varies with age and how this might affect male mate choice can be particularly complicated in mixed mating systems (i.e. androdioecious species), where self-fertilizing hermaphrodites exist with rare males [24]. In these species, self-fertilization is the predominant reproductive mode, but outcrossing with males can provide genetic benefits by increasing genetic diversity in offspring [25,26]. Male mate choice could be favoured in these systems because males are rare and likely to encounter multiple hermaphrodites simultaneously [27]. Males may be able to target mates that are more fecund [17–21], are genetically dissimilar to themselves [27] or are most receptive to outcrossing [28]. Reproductive senescence could be important in male mate choice because ageing in hermaphrodites could affect the viability and/or supply of both male and female gametes and their ability to self-fertilize (and tendency to outcross) [3,29,30]. For example, in the androdioecious nematode Caenorhabditis elegans, hermaphrodites release particular compounds that attract males [29]. However, some of these chemicals are only released once the hermaphrodite’s sperm levels are depleted [3,30], and thus these chemicals are indicative of an increased likelihood of outcrossing. Correspondingly, males strongly prefer older (sperm-depleted) hermaphrodites [30].

The mangrove rivulus (Kryptolebias marmoratus) also consists of self-fertilizing hermaphrodites and rare males (electronic supplementary material, figure S1) and represents a unique androdioecious vertebrate system [24] with which to examine male mate choice and genetic variation in fecundity and senescence. First, males are at low frequency (ranging from 1% to 20% depending on the population [31]) and their production can be triggered by cool temperatures during embryo development [32,33]. Second, hermaphrodites predominantly self-fertilize their eggs internally, resulting in distinct genetic lineages (i.e. isogenic lines) and the production of genetically identical offspring over time [31,34], but hermaphrodites can also release unfertilized eggs that are fertilized externally by males [35–37]. In the laboratory, unfertilized eggs are rarely released by hermaphrodites [38] even when males are present [39], and outcrossing with males has been difficult to observe [37]. However, genetic data from the field suggest that outcrossing does occur, although rates vary among populations [34,40,41]. Third, outcrossing in rivulus has genetic benefits and increased heterozygosity is associated with lower parasite loads in the field [25]. In line with this, mate choice studies with visual-only cues and olfactory-only cues show that males prefer hermaphrodites that are genetically dissimilar to themselves (i.e. from a different genetic lineage than themselves) [27]. Additionally, hermaphrodites prefer to associate with males over other hermaphrodites [27,42] but do not seem to prefer males from different genetic lineages [27]. Finally, hermaphrodite age might be particularly important in the likelihood of outcrossing. In fishes, the quality of male gametes declines with age [1,43], and hermaphrodites might switch to outcrossing as they age due to their own sperm becoming less viable (similar to older sperm-depleted C. elegans [30]). However, old hermaphrodites can also transition to secondary males as they age, particularly if they are exposed to shortened day length [44], suggesting that ovarian tissue and egg production might be affected negatively by age. Thus, although males direct their mating efforts towards unrelated hermaphrodites from different genetic lineages than their own [27], it is unclear whether their mate choices are affected by partner fecundity and age.

In this study, we were interested in (i) how fecundity varies with respect to age in hermaphrodites and (ii) whether this age-dependent fecundity helps explain male mate choice in the mangrove rivulus. We also took advantage of distinct genetic lineages arising naturally from many generations of self-fertilization [31] to examine whether any age-dependent fecundity and male preference patterns vary with genetic lineage.

2. Methods

Mangrove rivulus hermaphrodites are generally reproductively mature at approximately three months and can live for more than 5 years in the laboratory [45]. Although variable among genetic lineages, hermaphrodites can produce embryos at high levels until approximately 340 days old, after which production declines and plateaus [38], but they can produce viable self-fertilized embryos for over 3 years [32]. Because we were interested in how age affects hermaphrodite fecundity and attractiveness in reproductively mature fish, the terms ‘younger’ versus ‘older’ used here are relative to one another. The ‘younger’ age refers to hermaphrodites that are fully mature (approx. 250 days old), and the ‘older’ age refers to hermaphrodites that are roughly twice as old as the ‘younger’ ones (approx. 500 days old). Previous studies of male mate choice in rivulus have used hermaphrodites of approximately 240 days old [27], which correspond to our ‘younger’ aged hermaphrodites.

We used hermaphrodites from four distinct genetic lineages (BP16, BP21, CROC27 and CROC31), and because males prefer hermaphrodites from dissimilar genetic lineages [27], we used males from five different genetic lineages (BP11, BP13, BWN3, BWS34 and BWS38). Males had all developed as primary males [32,33] and were identified by their distinct orange colouration (electronic supplementary material, figure S1) [32,46], in combination with never having embryos in their containers. All genetic lineages had been propagated via selfing for 6−9 generations prior to these experiments.

Upon hatching, all fish (males and hermaphrodites) were housed individually in small, transparent plastic cups for approximately 2–3 weeks filled with approximately 85 ml of 15 parts per thousand (ppt) water made with Instant Ocean^® and tap water treated with CloramX^®. All fish were then transferred to individual 750 ml Rubbermaid^® TakeAlongs^© Deep Square containers containing a polyester floss nest (Poly-Fil^®) and filled with approximately 700 ml of 15 ppt water. Fish were fed more than 1 ml of live Artemia nauplii every 1−2 days. Individuals were on a 12 h light cycle between 07.00 and 19.00 hours, and the temperature was maintained at approximately 23–25°C.

(a). Part 1: genetic and age-dependent differences in fecundity

We compared the embryo production of younger hermaphrodites (mean ± s.e.; 255 ± 2 days old; n = 41) and older hermaphrodites (502 ± 1 days old; n = 47) over three weeks in April–May. On day 0, we replaced the floss nest and added clean 15 ppt water. We then counted and removed all embryos over three weeks (with collection blind to age and genetic lineage). Only one unfertilized/inviable egg was found during these counts, and it was excluded.

In our analysis, we examined how total embryo production was affected by hermaphrodite age (younger versus older), genetic lineage and their interaction. Due to some hermaphrodites producing zero embryos, we used a negative binomial general linear model with a log link and specified Type III SS for the ANOVA function.

(b). Part 2: genetic and age-dependent differences in attractiveness

In a separate experiment with different individuals from those used in part 1, we examined how male preference was related to partner age. We compared the preference of male rivulus for a younger hermaphrodite partner (274 ± 2 days old; n = 40) and older hermaphrodite partner (484 ± 11 days old; n = 40) when given a simultaneous choice. We controlled for genetic lineage by pairing younger and older hermaphrodites within a genetic lineage (e.g. a BP16 pair had both a younger and older BP16 hermaphrodite), and there were 10 pairs (each with two partners) within each of the four genetic lineages. Note that younger hermaphrodites are also smaller than older hermaphrodites due to indeterminate growth (younger body size: 26.6 ± 0.2 mm, n = 40; older body size: 29.3 ± 0.4 mm, n = 40). Males were always from a different genetic lineage from that of the hermaphrodites with whom they were partnered [27]. At the start of the choice tests, these males were 300 ± 17 days old (n = 20) and had never interacted or mated with a hermaphrodite previously.

Male mate choice assays were conducted in 10-gallon glass tanks (50 × 25 cm²; length × width) filled to 8 cm with 16 l of 15 ppt water (electronic supplementary material, figure S2). These tanks were covered in opaque black plastic on three sides with observers positioned at the front long side of the tank. One small plastic plant was positioned in the centre area at the back of the tank. Hermaphrodites were netted from home containers and released into mesh breeding boxes (16 × 12.5 × 13 cm³; length × width × height; Marina^© fish net breeders) that allowed passage of visual and chemical cues and were positioned against the wall at either end of the tank (electronic supplementary material, figure S2). The side of the younger/older partner was determined by a coin toss before each trial.

The focal male was netted from his home container and gently put into a transparent plastic acclimation cup positioned in the centre of the tank. This allowed the male to see both hermaphrodites during his acclimation period. The male remained in the cup for 1 min, after which he was gently released, and the cup was removed.

The focal male was observed for 10 min. To measure association time, we recorded each time the male entered the partner area (his snout crossed a line drawn on the tank) and the time spent in that area (electronic supplementary material, figure S2). After a trial, each hermaphrodite was measured for body size using a laminated ruler. Test tanks were drained and rinsed with hot water and refilled with clean 15 ppt water before the next trial.

Choice tests were performed over a period of three weeks. Hermaphrodites were only used once (n = 80 fish, 40 pairs). Males were assigned randomly to hermaphrodite pairs and were used twice (due to their limited availability, n = 20). After their first choice test, males were returned to their individual home containers. After 11−16 days, males were reused in a second choice test, but with a pair from a different genetic lineage from that of their first encounter. Fish were not fed before testing that day. Partners and focal males were given random numbers and letters for assignment. Observers were blinded to the age of the partners and the genetic lineage of the partners and males.

First, we determined whether there was any preference or avoidance of particular genetic lineages. For each trial, we summed the total time the male spent with both the younger and older partners to get a total association time. We used a mixed model to examine whether this association time differed among genetic lineages, including male identity as a random factor because each male was used twice and male genetic lineage as a random factor because there were five different male lineages. We specified Type III SS and degrees of freedom were estimated using the Satterthwaite method.

Second, we examined whether male mate preference was affected by partner age and whether this varied among genetic lineages. Based on the residual plots, we square-root-transformed male preference time. We then compared the time spent with each partner (younger versus older) for each hermaphrodite genetic lineage using a linear mixed model, including trial as a random factor because the two partner hermaphrodites were not independent within a trial, and including both male identity and male genetic lineages as additional random factors. We did not include hermaphrodite body size as a covariate in the models for two reasons. First, body size is confounded with age (older hermaphrodites are larger than younger hermaphrodites) and genetic lineage (some lineages are larger than others). Second, in some cases, there is no overlap in body sizes between younger and older hermaphrodites within a genetic lineage, and in some cases, there is no overlap among genetic lineages within a particular age (electronic supplementary material, figure S3). For an ANCOVA to be useful, there must be enough data overlap among groups in the covariate to be able to compare the groups in their main effects at the same level of the covariate. We specified Type III SS and degrees of freedom were estimated using the Satterthwaite method. When the model was singular due to the random effect(s) accounting for zero variance, we reran the model after removing those random effects to confirm that patterns remained the same. We also used a linear model to examine whether the body size of the hermaphrodites varied with their age (younger versus older), genetic lineage and their interaction in a similar analysis. Although we could not include partner body size as a covariate when age and genetic lineage were considered factors, we did examine the relationship between partner body size and male preference time using Spearman correlations.

Finally, to examine whether any male traits might be playing a role, we used Spearman correlations to explore consistency between a male’s preference (as a proportion of association time spent with the older partner) in his first and his second choice assay, as well as the relationships between male age (and size) and preference (as a proportion of association time spent with the older partner).

All analyses were conducted in RStudio v. 2024.09.0+5375 [47] using the packages dplyr [48], lme4 [49], lmerTest [50], effectsize [51], car [52], MASS [53] and emmeans [54] with figures made in ggplot2 [55].

3. Results

(a). Part 1: genetic and age-dependent differences in fecundity

As expected, we found an age-associated decline in fecundity, with younger hermaphrodites producing four times more embryos than older hermaphrodites (figure 1; age: X² = 26.8, p < 0.0001). However, embryo production also varied significantly among genetic lineages (X² = 22.9, p < 0.0001), with some lineages producing more embryos than others in both younger and older hermaphrodites. There was not a significant age × lineage interaction (X² = 6.0, p = 0.1112).

Figure 1.

The total number of embryos produced by younger and older hermaphrodites from different genetic lineages over 3 weeks (n = 88 hermaphrodites with 8–13 individuals of each age–genetic lineage combination). Symbols indicate mean ± s.e. for each lineage.

(b). Part 2: genetic and age-dependent differences in attractiveness

Overall, we found that males spent over 73% of their time associating with a partner, and this was true across all of the genetic lineages (electronic supplementary material, figure S4; F_3,40 = 0.2767, p = 0.8419). Interestingly, when examining preferences for younger versus older partners, we did not find a consistent preference for partners of particular ages (F_1,80 = 2.742, p = 0.1016; η² = 0.03), nor consistent differences among genetic lineages (F_3,80 = 0.083, p = 0.9693; η² < 0.01). Instead, we found a strong ‘partner age’ by ‘lineage’ interaction, and preference for younger versus older partners depending on the genetic lineage of the partners (figure 2a; F_3,80 = 10.466, p < 0.00001; η² = 0.28). For some genetic lineages, males strongly preferred the younger hermaphrodite over the older one (BP16, CROC31), but for other genetic lineages, males showed the opposite preference (CROC27) or equal preference for both ages (BP21). None of the random effects included (male identity, male genetic lineage and hermaphrodite pair) in any analysis accounted for significant variation and their removal from the models did not change the results substantially.

Figure 2.

(a) The total amount of time males spent associating with either their younger and older partner from different genetic lineages and (b) the body sizes of those younger and older partners from different genetic lineages. Symbols indicate mean ± s.e. for each lineage (n = 40 trials/males, with 80 hermaphrodites).

As expected, older hermaphrodites were significantly larger than younger hermaphrodites (F_1,72 = 112.5, p < 0.0001; η² = 0.61), and overall size differed among lineages (figure 2b; F_3,72 = 49.9, p < 0.0001; η² = 0.68). Furthermore, some lineages grew more as they aged than others (F_3,72 = 10.9, p < 0.0001; η² = 0.31). However, it is clear that the ranking of male preferences does not simply reflect the ranking of partner sizes (figure 2). Consistent with this, there was not a significant relationship between male preference time and partner body size for younger hermaphrodites (r_S = 0.02, p = 0.908, n = 40), older hermaphrodites (r_S = −0.04, p = 0.816, n = 40) or across all partners (r_S = −0.04, p = 0.721, n = 80; electronic supplementary material, figure S5).

Finally, how males behaved during their first choice assay was unrelated to how they behaved at their second choice assay (proportion of association time spent with older partner at assay 1 versus assay 2: r_S = 0.02, p = 0.926, n = 20). Male age and body size were also unrelated to their preference (proportion of association time spent with older partner and male age: r_S < 0.01, p = 0.979, n = 40; male size: r_S = 0.07, p = 0.650, n = 40).

4. Discussion

Central questions in sexual selection focus on why individuals should be choosy about mates and the traits they use to make these choices [15–19,56]. Less is known about how mate choice and the traits used in these choices vary as individuals age or change in condition, and even less is known about the genetic variation underlying these changes. Here, we found that despite continued indeterminate growth, mangrove rivulus hermaphrodites showed a strong decline in fecundity as they aged, consistent with what we know about senescence in many other taxa [1,2,4]. Genetic lineages also varied in their fecundity, consistent with other studies finding genetic variation in fecundity [57]. Thus, males could benefit from being choosy and distinguishing among mates of different ages and from different genetic lineages. Surprisingly, however, males incorporated cues beyond these age-related and genetic differences in fecundity into their mate choices. Specifically, males showed strong preferences for partners of particular ages, but this depended on the genetic lineage of the partner.

We found that attractiveness to males varies over the hermaphrodite lifespan, and furthermore, that these patterns depend on hermaphrodite genotype, resulting in a genotype × age (i.e. genotype × condition) interaction. Genes coding for attractiveness (or fecundity) might be differentially expressed over the reproductive lifespan, or different genes might code for attractiveness (or fecundity) at different ages [56], and both of these could vary with the genetic lineage. For example, genetic lineages might differ in the age at which genes that underlie behavioural or chemical ‘attractiveness’ cues are upregulated (or downregulated). Interestingly, in fruit flies, there is compelling evidence that the loci affecting fecundity early in life are different from those that affect fecundity later in life [57]. Genotype by environment interactions can help maintain genetic variation in sexually selected traits [56,58–60]. Indeed, our results suggest that every lineage has attractive hermaphrodites at some point—some when they are young and others when they are old. Males outcrossing with a mix of genetic lineages and ages could potentially increase their reproductive success and maintain genetic variation in fecundity and reproductive senescence.

The age-related decline in hermaphrodite fecundity shown here could be due to a loss of male function, female function or both [9]. In fishes, there is evidence that males can show stronger reproductive senescence in terms of gamete quality than females [1,43]. Males might prefer older hermaphrodites in lineages where there is a strong age-related decline in sperm quality as hermaphrodites age, and thus an increased likelihood of releasing an unfertilized egg and outcrossing with a male with higher-quality sperm, as in nematodes [3,29,30]. Similarly, males might avoid older hermaphrodites in lineages where there is a strong age-related decline in egg quality and loss of ovarian tissue and an increased likelihood of these old hermaphrodites transitioning to secondary males [44,46,61]. Hermaphrodites (including those transitioning to secondary males) might produce behavioural and/or chemical ‘attractiveness’ cues that indicate to males the viability of their male and female gametes and their likelihood of outcrossing with them, driving the male preference patterns we see here. Chemical cues are necessary (and sufficient) for males to distinguish genetically similar and dissimilar hermaphrodites [27]. Interestingly, when males are presented with visual cues only, they show no preference for hermaphrodites of different ages, suggesting that chemical cues are important [62] (although that study did not examine genotype × age interactions). Comparing how age affects the quality and number of both male and female gametes in hermaphrodites across genetic lineages would shed light on the mechanisms underlying these fecundity and mate choice patterns.

Our results suggest that genotype by condition (i.e. age of the partner) interactions can play a strong role in male mate choice. Specifically, the genetic identity of the partner is key to how males might weigh age-dependent changes in fecundity. These types of interactions could be particularly important when there is genetic variation in fecundity and in androdioecious species where outcrossing is controlled by hermaphrodites, potentially in an age-dependent way.

Ethics

All assays and maintenance were approved by the Animal Care and Use Committee at the University of the South (Protocols: McGhee 1-2021 and 13-2022) and adhered carefully to guidelines for ethical animal treatment [63]. Fish that were not used in additional studies were euthanized with an overdose of sodium-bicarbonate buffered MS-222 anaesthetic (Finquel^®) in 15 ppt water, as approved by ACUC protocol.

Data accessibility

Data, README files, and R code to conduct all analyses are available from the Dryad Digital Repository [64]. Additional figures (S1–S5) are available as electronic supplementary material [65].

Declaration of AI use

We have not used AI-assisted technologies in creating this article.

Authors’ contributions

J.O.G.: data curation, investigation, methodology, writing—review and editing; M.C.N.: data curation, investigation, methodology, writing—review and editing; C.S.N.: data curation, investigation, methodology, writing—review and editing; K.E.M.: conceptualization, data curation, formal analysis, investigation, methodology, project administration, resources, supervision, validation, visualization, writing—original draft, writing—review and editing.

All authors gave final approval for publication and agreed to be held accountable for the work performed therein.

Conflict of interest declaration

We declare we have no competing interests.

Funding

No funding has been received for this article.