Heat stress reveals a fertility debt owing to postcopulatory sexual selection

Abstract Climates are changing rapidly, demanding equally rapid adaptation of natural populations. Whether sexual selection can aid such adaptation is under debate; while sexual selection should promote adaptation when individuals with high mating success are also best adapted to their local surroundings, the expression of sexually selected traits can incur costs. Here we asked what the demographic consequences of such costs may be once climates change to become harsher and the strength of natural selection increases. We first adopted a classic life history theory framework, incorporating a trade-off between reproduction and maintenance, and applied it to the male germline to generate formalized predictions for how an evolutionary history of strong postcopulatory sexual selection (sperm competition) may affect male fertility under acute adult heat stress. We then tested these predictions by assessing the thermal sensitivity of fertility (TSF) in replicated lineages of seed beetles maintained for 68 generations under three alternative mating regimes manipulating the opportunity for sexual and natural selection. In line with the theoretical predictions, we find that males evolving under strong sexual selection suffer from increased TSF. Interestingly, females from the regime under strong sexual selection, who experienced relaxed selection on their own reproductive effort, had high fertility in benign settings but suffered increased TSF, like their brothers. This implies that female fertility and TSF evolved through genetic correlation with reproductive traits sexually selected in males. Paternal but not maternal heat stress reduced offspring fertility with no evidence for adaptive transgenerational plasticity among heat-exposed offspring, indicating that the observed effects may compound over generations. Our results suggest that trade-offs between fertility and traits increasing success in postcopulatory sexual selection can be revealed in harsh environments. This can put polyandrous species under immediate risk during extreme heat waves expected under future climate change.

Individual condition (C) was assumed to determine the amount of resources that can be allocated to germline maintenance (M) in form of anti-oxidative defence and repair needed to maintain gamete viability 3,4 , or reproductive effort (R) in form of production of gametes and ejaculatory components that increase a male's success in sperm competition 5,6 , such that: Eq. 1 where k is the proportion of resources allocated to reproduction.Gamete viability, , is assumed to be dependent on the amount of maintenance per reproductive effort (i.e. per gamete or ejaculate volume): ≈ [ where  describes environmentally dependent consequences of sub-maximal germline maintenance, such that some environmental conditions will impair fertility more than others (e.g.hot temperature or high salinity increases ).The addition of +1 to the denominator of Eq. 2 assures that viability ranges between 0 and 1, but we note that this choice was arbitrary and that other expressions for Eq. 2 resulted in the same qualitative results.While we here express the effect of germline maintenance (M) on gamete viability, we note that gamete quality is directly related to the survivorship and quality of offspring in species with limited parental care, and hence, our results relate directly to measures of offspring survival (as measured in the empirical data).If reproductive success follows a power function of reproductive effort, and if fitness, , is the product of sperm competition success and gamete viability, then:

Eq.3
where parameter  describes how reproductive effort translates into postcopulatory reproductive success.When b = 1, success in sperm competition is directly proportional to the amount of resources invested, as envisioned in "fair raffle" models of sperm allocation under risk of sperm competition 7 .When  < 1, reproductive success is less than proportional to investment and postcopulatory sexual selection is relatively weak as expected when risk of sperm competition is low, whereas  > 1 gives a disproportionate advantage to individuals investing more in reproductive effort and sexual selection is very strong.We note that this last scenario implies some sort of threshold mechanism at play, where male allocation need to exceed some particular value to achieve fertilization.Such mechanisms could, for example, be at play through female choice of male sperm, but we note that empirical evidence for this hypothesis is scarce 8 .Individuals in a population experiencing strong sexual selection (b>1) need to invest much more in reproductive effort to secure a significant share of paternity relative to individuals from a population where sexual selection is weak (b<1).Given a tradeoff between investment in reproductive effort and germline maintenance (Eq.1), such excessive germline allocation is predicted to result in reduced fertility, an effect that is particularly pronounced under harsh environmental conditions (Fig. S1.1).

A B C
The optimal germline allocation strategy (kopt) for different strengths of sexual selection (b) and viability selection (a) is given by differentiation of equation ( 3) with respect to k: which shows that optimal allocation to reproduction,   , increases with the strength of sexual selection, , and decreases with environmentally dependent viability selection, .
Unsurprisingly, increased sexual selection does indeed lead to decreased gamete viability as follows from the trade-off scenario described by equation ( 1) (Supplementary Fig. S1.1 and Fig. S1.2A).Optimal allocation (and resulting fertility) is independent of condition, , when reproductive success is a power function of investment, but we note that reproductive effort and gamete viability can either be increasing or decreasing functions of condition, depending on the fitness functions used (results not shown, but see: [8][9][10] ).
What consequences do differences in mating system and a history of intense sperm competition (high ) have for fertility responses to increased environmental stress (increases in )?We illustrate these effects by first replacing  in equation ( 4) with   , representing viability selection in a relatively benign ancestral environment, making it possible to solve for   for different values of .We then replace k in equation ( 2) with this expression for   and differentiate with respect to  to show how gamete viability, , is affected by increasing environmental stress for allocation strategies that have evolved under different scenarios of sexual selection and viability selection in the ancestral environment: Eq. 5 Predictions from equation ( 5) are presented in Supplementary Fig. S1.2B (see also Fig. 1B in main text) and show that, for any strength of viability selection in the ancestral environment, populations that have evolved under a history of strong sexual selection are predicted to suffer a greater fertility loss following increased environmental stress.

Extended discussion of model assumptions and the relation between predictions and data
Our aim with the described model was to generate predictions for how previous differences in mating system and adaptations to postcopulatory sexual selection would affect immediate responses to environmental stress, qualitative predictions that could be tested with our empirical data.In our simple model we therefore described viability selection and sexual selection by power functions with exponents a and b without discussing mechanisms behind exactly what determines a and b.Indeed, we made the very simple assumption that sperm competition should be stronger (b greater) in a population where there is multiple mating and mate choice (the N+S and S regimes) compared to a population where there is no sexual selection (the N regime).This simple logic is supported by both theoretical models on sperm and ejaculate allocation 8 and a rich empirical literature 5,8,11,12 .We also made the reasonable assumption that viability selection increases (via exponent a) at more stressful conditions, but we note that the use of power functions to describe both these relationships were arbitrarily chosen.
Indeed, sperm competition is the results of multifaceted frequency-dependent pre-and postcopulatory processes 8,12 .Nevertheless, even though we did not explicitly model frequency-dependence here but rather manipulated (i.e.fixed) the strength of sexual selection at different values of b to evaluate effects on fertility under environmental change, ESS models on sperm competition reassuringly produce the same qualitative results as we obtained here with reference to how the risk of sperm competition benefits increased allocation to ejaculate traits that increase competitive reproductive success 8 (parameter k in our model).For example, in Parker's (1990) simple "fair raffle" scenario 7 , k is directly proportional to the risk of sperm competition.Further note that the risk of sperm competition in our N regime is = 0, and essentially ~1 in our N+S and S lines (based on remating rates estimated in 13 ).Hence, both our model and ESS models (reviewed in: 8 ) make the qualitative prediction that allocation to k should be greater in N+S and S lines relative to N lines.Our model then increased environmental stress (a) to show that this allocation decision results in compromised fertility in the N+S and S regimes under the assumption of a trade-off between gamete viability and sperm competition success.Moreover, because the optimal allocation decision is determined by a balance between viability selection and sexual selection, the model predicts that the S regime, where females unconditionally contributed only two offspring in each generation (and viability selection therefore was weaker compared to the N and N+S regime), should show the greatest environmental sensitivity.We did indeed find evidence for these qualitative predictions in our data (see main text).
The model predicts that there should still be differences in fertility between our evolution regimes in benign settings as long as there is some viability selection (a > 0) , although these differences should be much smaller than at high environmental stress (a increases).We do not see a difference in fertility at benign temperature in our empirical data.We believe a very likely reason for this is that these lines also evolve other components related to fertility.For example, we have found evidence that N+S lines are overall in better condition and have higher fecundity, suggesting that the overall stronger selection in this regime may have more efficiently purged deleterious alleles 14 .Another (not mutually exclusive) explanation is that environmental harshness in our experiment (exponent a in the model) is close to 0 in our benign conditions (corresponding to the scenario with a = 0.02 in figures S1.1 and S1.2).Given the ad libitum egg-laying substrate and high fertility of C. maculatus in laboratory conditions (egg-to-adult survival typically at 95%, where the 5% mortality could be due to handling), suggest that a is indeed close to 0. If so, the model predicts very small (statistically undetectable) difference in fertility between regimes when held at benign conditions.Moreover, even though our experimental evolution lines have been propagated for a relatively long time compared to other similar laboratory evolution experiments manipulating mating system, we expect them to not have reached their new evolutionary optima (allocation decisions maximizing fitness in their respective lab environments).Hence, there should also be some caution when comparing empirical data and model predictions.
Finally, because (postcopulatory) sexual selection is inherently a frequency dependent process, it is likely that abrupt changes in the environment (increases in a) could themselves modulate parameter b, by for example changing the density and quality of rivalling males at mating sites 7,8,14-16 .If we are concerned with predicting how germline plasticity and future evolution of reproductive strategies in changing environments affect fertility and population health, such dynamics would ultimately need to be considered in more sophisticated mechanistic models incorporating frequency-dependent selection.Here, however, we kept b constant to describe the strength of sexual selection in the ancestral environment as we were interested in generating predictions for immediate fertility responses under abrupt environmental change attributed to the organism's evolutionary history of natural and sexual selection.We then could compare these qualitative predictions directly with our empirical data (see main text).

Experimental design: main experiment
Design of the main experiment, including experimental evolution under the three alternative selection regimes, and two generations of common garden prior to the experimental generation.In the first experimental generation (parental), virgin beetles of all three regimes were picked within 24 hours after eclosion and males were either kept isolated or kept together with other males.Subsequently, half of the beetles of each regime and treatment combination were exposed to a heat shock.The thermal sensitivity of fertility was assessed by comparing heat shocked and control beetles within regime and treatment according to: TSF = 1 -(offspringheat shocked / offspringcontrol).F1 offspring of heat shocked and control N+S males and females were used for the assessment of transgenerational effects of heat shock (see supplementary figure S4).

Details of the experimental design used to manipulate mating system
The selection regimes used here were designed to manipulate the relative strength of selection on males and females.However, the way the regimes were implemented also caused small differences in the timing of propagation, the amount of egg-laying substrate per female, and container size during propagation, which we discuss below.

Timing of reproduction:
The regimes impose slightly different reproductive schedules.In N + S lines males and females can mate and lay eggs for 48 hours after their emergence, in N lines males and females can mate for five hours after which females get 48 hours for egg-laying.Hence, the 48-hour egg-laying period is delayed by 5 hours in N compared to N+ S females.It has been shown that female C. maculatus age substantially faster once egg-laying starts (e.g., Tatar et al. 1993, Wilson 1994,  Maklakov & Bonduriansky 2009).We therefore argue that this 5-hour shift is negligible, even in a short-lived species such as C. Maculatus, since the time period is very short and females have no access to beans, are not ovipositing during that time, and are kept at very low density (alone together with a single male).For females evolving under the S regime egg-laying is delayed by 48 hours compared to the N + S regime.However, in the S regime females are under relaxed selection (females contribute one female and one male offspring to the next generation) and females typically live around seven days while reproducing under crowded conditions.Thus, rather than imposing selection on the reproductive schedule of females, this regime removes selection on females all together.

Juvenile density:
The number of beans per female varies slightly between the experimental evolution regimes (N + S: ~32 beans/female, N: ~39 beans/female, S: ~30 beans/female) (see Martinossi-Allibert et al. 2019, for details regarding population size under the various regimes).However, a single bean can sustain the development of more than 10 beetles and a female rarely lays more than 100 eggs.
If these differences in rearing conditions were of importance, we would expect the regimes to evolve differences in longevity and body size.We have found no such differences in adult body mass at generation 16 (Martinossi-Allibert et al., 2019), or at generation 40 (White S, Bolund E and Berger D, unpublished data).The last experiment was designed to test for the evolution of sex-specific genetic (co)variances in life history, and sample size were therefore sizeable.While there was a slight tendency for differences in longevity between the three regimes in this experiment, it was polygamy that tended to have longer life (although marginally non-significant).Importantly, the polygamy regime was intermediate in our current experiment on thermal sensitivity (TSF), and any potential differences in longevity thus seem unlinked to our current results.

Use of different containers:
To apply different selection pressures in the various experimental evolution regimes, we kept beetles in different containers and environments during the reproduction and egg-laying (N + S: mating and egg-laying in a large jar with beans, N: mating in a 60mm Petri-dish and egg-laying in a large jar with beans, S: mating in a large jar with a cardboard structure and egglaying in a 60mm Petri-dish).These environments inevitably cause the beetles to receive different environmental cues depending on their selection regime.We do, however, believe that we designed a selection protocol that minimises these differences.The Petri-dish used for mating in the N regime was applied for a short time (5h) and the main objective with this treatment was to allow females to mate while removing sexual selection and conflict.When the Petri-dish treatment is applied in the S regime, it is after mating interactions, to females that only contribute 2 offspring to the next generation (effectively removing fecundity selection), so the applied Petri-dish treatment is extremely unlikely to have enforced any selection in the S regime.What is important in this species is the density and complexity of the environment during mating interactions, hence the card-board structure placed in the S-treatment.

Figure S1. 1 .
The relationship between competitive reproductive success (i.e.success in sperm competition) and gamete viability for different strengths of sexual selection (b = 0.1; 1; 3 from panel A-C) and viability selection (a; salmon = 0.02, red = 0.10, brown = 0.20).To facilitate interpretation, competitive reproductive success is expressed relative to the success of an individual with the same condition (C) investing all resources in reproduction (kmax = 1).Optimal allocation in each scenario is found by maximizing the product of gamete viability and competitive fertilization success, corresponding to the area under each curve's inflection point.Individuals experiencing strong sexual selection (panel C) need to invest more in reproduction to get the same share of paternity compared to individuals experiencing weak sexual selection (panel A) and pay a fertility cost in terms of reduced gamete viability.This cost becomes more pronounced in harsh environments (brown lines).
A) the optimal reproductive effort (  ; hatched lines) and resulting gamete viability (; full lines) for different levels of sexual selection (b) and viability selection (a; salmon = 0.02, red = 0.10, brown = 0.20).In B) the change in gamete viability as environmental stress (a) changes from ancestral conditions (  ; salmon = 0.02, brown = 0.20) for populations that have evolved optimal allocation under either weak (b = 0.1, thin lines), strong (b = 1, intermediate lines) or very strong (b = 3, thick lines) postcopulatory sexual selection.