Exposure to high male density causes maternal stress and female-biased sex ratios in a mammal

Abstract

A shift from the traditional perspective that maternal stress is invariably costly has instigated recent interest into its adaptive role in offspring sex allocation. Stress generated by social instability has been linked to offspring sex ratio biases that favour the production of female offspring, which converges with the theoretical prediction that mothers in the poor condition are better off investing in daughters rather than sons. However, previous research has failed to disentangle two different processes: the passive consequence of maternal stress on sex-specific mortality and the adaptive effect of maternal stress at the time of conception. Here, I show that exposure to high male density social conditions leads to elevated stress hormone levels and female-biased in utero offspring sex ratios in house mice (Mus musculus domesticus), and identify that sex-specific offspring production—not sex-specific mortality—is the mechanism accounting for these sex ratio skews. This outcome reflects the optimal fitness scenario for mothers in a male-dominated environment: the production of daughters, who are guaranteed high mate availability, minimizes male–male competition for their sons. Overall, this study supports the idea that maternal stress has the potential to be adaptive and advances our understanding of how exposure to different social conditions can influence sex allocation in mammals.

1. Introduction

Theory predicts that parents will be advantaged by adjusting offspring sex ratios (n_males/n_{total offspring}) in relation to their own ability to bear the costs of producing sons or daughters [1] and in relation to the future reproductive success of sons versus daughters [2–4]. Studies of sex allocation have traditionally addressed two central topics—the adaptive significance of producing offspring of either sex and the mechanism(s) that allow for facultative sex allocation—but integrating both topics has proved to be challenging. For example, in the absence of any evidential support of the underlying mechanism(s), correlations between social stressors and an elevation in female human births [5] are usually attributed to high rates of prenatal male death [6,7]. Human literature has debated whether differential embryonic mortality is an adaptive consequence of maternal manipulation (the ‘culled’ cohort hypothesis) or a by-product of maternal stress (the ‘damaged’ cohort hypothesis) [8], a debate that has also been addressed more generally in evolutionary biology [9]. From an evolutionary standpoint, the most cost-effective mechanisms are those that operate close to conception [10–12]. Surprisingly, sex-allocating mechanisms that occur early on in the reproductive sequence (e.g. sperm selection and differential implantation) are poorly studied, most likely owing to the requirement to prenatally quantify offspring sex which, in mammals, can be technically difficult. Because offspring sex ratio skews can compromise population viability by persisting in the adult cohort and distorting the ratio of males to females in the mating pool [13,14], understanding both the underlying mechanisms as well as the proximate causes of biased offspring sex ratios is critical to multiple facets of science, including biodiversity conservation [12,15,16].

The developmental gestation period that follows internal fertilization provides an inherent link between maternal condition and offspring sex allocation in mammals. The Trivers–Willard hypothesis (TWH) predicts that mothers in the good condition favour the production of male offspring, because sons—when competitive and successful—should generate more grandoffspring than daughters [4]. Typically, however, sons are also more ‘expensive’ to produce as they require greater maternal energy requirement during late gestation and lactation than do daughters [17,18]. Consequently, mothers in the poor condition are expected to have daughters [4] (note that this prediction also adheres to the cost of reproduction hypothesis; reviewed [1]). Despite its intuitive appeal, studies testing the TWH have produced equivocal results with inconsistencies being prevalent among and even within species [19]. One contributing factor to this ambiguity has been the use of a number of different measures of the maternal condition and variation in the level (i.e. population versus individual) and time point (e.g. birth versus conception) at which they are quantified [19,20]. The most commonly applied indices have included those that are associated with maternal age and reproductive experience, nutritional status (e.g. body mass and glucose level) and social rank (e.g. testosterone level) [11,21–25].

The degree of chronic stress that an individual has or is experiencing can be applied as an index of condition [26,27]. The TWH predicts that highly stressed, poor condition mothers should produce daughters [4]. In humans, exposure to distressing situations, such as severe life events, economic disruption and natural disasters, is often correlated with a reduction in male births [7]. A bias towards female offspring during times of stress is also often observed in non-human mammals [28], although patterns appear to vary based on the reproductive stage at which sex ratios are measured [29]. Stress hormones, which transduce stressful stimuli from the environment into internal physiological signals, have been identified to be likely candidates that initiate mechanisms of sex allocation by triggering physiological processes [29]. However, very few mammal studies have directly assessed the relationship between hormonal measures of maternal stress and variation in offspring sex ratios (reviewed [19,28,29]). Specifically, this relationship remains unexplored at the prenatal stage (i.e. the primary or implantation sex ratio) because studies, to date, have focused exclusively on birth sex ratios (i.e. the secondary sex ratio) [29].

The type and frequency of interactions that individuals share with conspecifics in the social environment can affect hormonal responses and strategies of sex allocation [29]. Sex-biased social interaction between individuals in the local environment, such as competition (local mate and resource competition; [2,30]) and cooperation (local resource enhancement; [31]) among relatives and other residents, can alter the costs and benefits associated with producing each sex. Accordingly, individuals are expected to produce female offspring under conditions of intensified local competition among males [2]. Female haplodiploid invertebrates, which have direct control over offspring sex by choosing whether (daughters) or not (sons) to release sperm and fertilize eggs, have been shown to maximize the fitness of their sons by mediating sex ratios based on local male density [32]. There is also convincing evidence that local conditions influence sex ratio variation in birds, where, here too, mothers govern the sex of ovulated eggs [28,33,34]. In mammals, sex allocation is more of a level playing field: the paternal gamete determines offspring sex, but further development is under maternal control [35]. Consequently, there are multiple reproductive stages at which the social environment can influence offspring sex, from mediating sperm sex ratios prior to mating [36] to differential postnatal maternal care [37]. Social factors, such as hierarchical rank, have been shown to shape stress levels and linked to adaptive sex allocation in mammals, albeit with inconsistencies in the direction of the response [29]. The key challenge lies in being able to disentangle multiple, interacting and potentially opposing sex allocation strategies that operate simultaneously within natural populations [1]. Experimental manipulation under controlled conditions is an effective way to address this complexity.

Here, I performed an experiment on wild-sourced individuals of a tractable mammal species to investigate the role of socially induced maternal stress in sex allocation during early development. Social instability has been identified to be an effective model of stress for many mammals, especially rodents [38]. Specifically, female house mice (Mus musculus domesticus) elicit different stress responses when exposed to male and female conspecifics [39]. Thus, I exposed female house mice to one of two different social environments during sexual development: low-stress, high female density conditions and high stress, high male density conditions [36]. I measured maternal blood corticosterone (CORT) (i.e. the predominant glucocorticoid and stress-associated hormone in rodents) and glucose levels, and genetically sexed both viable and non-viable 14-day developing embryos. In line with theory, I hypothesized that the male-dominated environment would favour the production of female offspring and that this sex ratio bias would be linked to elevated maternal stress [2,4]. I also measured embryo number, size and in utero position (i.e. in relation to the sex of neighbouring embryos) to further explore fitness implications of socially induced maternal stress.

2. Material and methods

(a). Experimental animals

Wild house mice (M. m. domesticus; n = 100) were sourced from Rat Island (28° 42′ S, 113° 47′ E), which is located off the coast of Western Australia, and maintained in the laboratory at the University of Western Australia. The animals were held in constant temperature rooms (24°C) on a reverse 10 L : 14 D cycle and provided with water and food ad libitum. Mice were outbred under common garden conditions for two generations. For this, male and female pairs were housed together for a maximum of 14 days, following which pregnant females were housed alone and left undisturbed to give birth and rear their litter. Female and male offspring were separated at three weeks of age; female offspring were housed in sister groups and male offspring were housed individually. A total of 24 experimental female sister pairs, 30 non-focal females and 30 non-focal males were used in the social manipulation. Another 24 sexually experienced males (i.e. those that mated with a sister pair) were used in the experimental matings (see below).

(b). Social environment manipulation

Social instability has been identified to be an effective model of stress in rodents [38]. For example, previous research has shown that perceiving the presence of a male, but not mating or being solicited to mate, generates a stress response in adult females [39]. It is believed that this result is owing to a mismatch between sexual anticipation and reward, with exposure to males and their scents generating an anticipatory increase in the CORT level, but the lack of mating meaning these levels remain consistently high [39]. Here, I manipulated social cues that had previously been shown to elicit responses in reproductive traits in house mice [36,40–42], with the expectation that exposure to sex-specific cues during development would result in varied levels of stress [39]. Thus, females were exposed to either high male (n = 24) or high female density (n = 24) conditions from the time of independence from their mother until sexual maturity (figure 1). The two social environments were created by housing females in standard cages (28 × 46 × 13 cm) on metal racks in two separate constant temperature rooms. In the high male density environment, females matured within close proximity to males and were exposed to males and their scents on a regular basis (for more detail, see the electronic supplementary material, figure S1a). By contrast, females in the high female density conditions matured within close proximity to females and were occasionally exposed to the scent of a single male (for more details, see the electronic supplementary material, figure S1b). Sisters were used across treatments to account for family-derived variation.

Figure 1.

The experimental design. Wild-sourced female mice were reared under standard conditions from birth to weaning and then exposed to social environments that were anticipated to induce variation in the maternal stress level: high male or high female density conditions. The different sized mice represent a family (mother and offspring) to demonstrate that sisters were used across treatments, which controlled for family-derived variation that might influence offspring sex allocation. This was reinforced by ensuring that sisters reproduced with the same male. CORT concentration, blood glucose level, prenatal offspring sex ratios and embryo traits were measured at 14 days gestation. (Online version in colour.)

(c). Experimental matings

Once sexually mature, females reproduced with males that had been reared under standard conditions (figure 1). The experimental matings were performed during the dark phase of the 10 L : 14 D cycle in a third constant temperature room (24°C) as described previously [43]. When oestrus was apparent [44], females were placed in a cage with a male. The cage was then placed inside a large, opaque plastic tub (49 × 74 × 41 cm) to eliminate female exposure to other conspecifics during the mating period. Females were checked every half an hour for the presence of a mating plug, which is indicative of a complete and successful mating [43]. After mating, females were weighed, placed in their own cage and provided with the nesting material. Mated females were returned to their designated social environment, but were no longer exposed to the scents of conspecifics. Each sister pair reproduced with the same sexually experienced male; the order of first and second females to mate was alternated between treatments. Pregnant females were housed alone and left undisturbed until 14 days after mating, at which time they were sacrificed.

(d). Serum corticosterone assays

Maternal CORT concentration—the predominant rodent glucocorticoid—and glucose level were measured in blood extracted via postmortem cardiac puncture. Females were sacrificed between 9.00 and 10.00 am on gestation day 14. Blood was collected via cardiac puncture, refrigerated overnight at 4°C and then centrifuged at 3250 rpm for 20 min. The serum was removed, aliquoted and stored at −80°C. CORT concentration in undiluted serum samples was assayed in duplicate using a commercially available enzyme-linked immunosorbent assay (ELISA) according to the manufacturer's instructions (Mouse Corticosterone ELISA Kit no. 80556, Crystal Chem, Elk Grove Village, IL, USA).

(e). Blood glucose measurement

Maternal blood glucose level, a major energy source that is indicative of nutritional status [45], has been linked to variation in offspring sex ratio (e.g. see [46]). Consequently, blood glucose concentration was quantified using a commercially available Accu-Chek Performa Nano glucometer from the serum samples described above. It was not possible to obtain glucose concentration for one sample (female treatment) owing to insufficient serum quantity.

(f). Embryo measurements

Embryo viability status was scored as appearing either normal in size and shape (viable) or as a large haematoma that was discernible in shape and being resorbed (non-viable) [47,48]. The uterine position of each embryo was recorded, following which they were resected and stored in 100% ethanol. The prenatal sex ratio was quantified by genetically sexing both viable and non-viable embryos [49]. Each embryo was sexed using a rapid polymerase chain reaction (PCR)-based method, which included a Y-chromosome (Sry)-specific primer and a myogenin (Myog) control primer [49,50]. The 10 µl reactions were performed with 1 ng of DNA template, 1× PCR buffer, 0.3 µl MgCl₂ (50 mM), 2 µl dNTPs (200 mM), 0.1 µl Taq (0.5 units reaction⁻¹) and 0.5 µl each primer (0.25 µM). PCR products were denatured at 94°C for 5 min, 30 cycles of 5 s at 94°C and 40 s at 67°C, and 1 min extension at 72°C. The PCR products were electrophoresed on a 1.5% agarose gel for 30 min at 100 V, visualized with gel red (Biotium, Inc.) under ultraviolet illumination and scored as either male (two bands) or female (single band). Two people blind to treatment independently scored the bands. Embryo length was measured to the nearest 0.01 mm using digital callipers. Of the total 339 viable embryos, seven were unable to be measured owing to failed storage and desiccation (n_{female treatment} = 3; n_{male treatment} = 4).

(g). Statistical analyses

Linear mixed models (LMMs) and generalized linear mixed models (GLMMs) were conducted in R v.3.5.1 using the lmer or glmer functions implemented within the package lme4 [51]. LMMs were used to analyse treatment differences in (i) maternal and embryo anatomical traits, (ii) embryo number and (iii) concentrations of maternal blood constituents (glucose and CORT). GLMMs were used to analyse treatment differences in (i) embryo number (family = Poisson), (ii) the proportion of male offspring that mothers produced (i.e. the sex ratio; n_{male embryos}/n_{total embryo number}; family = binomial), (iii) the proportion of viable embryos that mothers produced (n_{viable embryos}/n_{total embryo number}; family = binomial), and (iv) the proportion of 0M (embryos positioned between two female embryos) and 2M (embryos positioned between two male embryos) sons and daughters that mothers produced (n_{positioned embryos}/n_{total embryo number−4}; note that peripherally located embryos were not included in this analysis; family = binomial). A GLMM with the intercept removed was applied to test for embryo sex ratio parity within each treatment (i.e. against 0.5). Binomial GLMMs were performed using the cbind function except for (iii) the proportion of viable embryos that mothers produced, as this approach restricts testing for embryo-level differences (i.e. sex-specific variation). In the Poisson GLMM (embryo number), penalized quasi-likelihood was used by running the function glmmPQL from the MASS package to account for underdispersion [52]. In the embryo viability GLMM and the embryo length LMM, both female and family identities were included as random effects. In all other cases, family identity only was included as a random effect in the model.

A single female failed to become pregnant. Subsequently, this sister pair was excluded from the dataset, which resulted in a total sample size of 46 females (n_{female treatment} = 23, n_{male treatment} = 23). An initial analysis confirmed that there was no relationship between the embryo number and the prenatal sex ratio (electronic supplementary material, figure S2). Female body mass was included as a covariate in the LMMs testing glucose and CORT concentration, as well as embryo number and length. Embryo sex was included as an additional covariate in the LMM testing embryo length. One glucose concentration was identified as an outlier (+2 s.d. from the mean) and therefore removed from further analysis (male treatment). The significance of the fixed effects was calculated using maximum-likelihood and Wald tests using the function Anova (car package). All non-significant two- and three-way interaction terms were removed from the models (full models: electronic supplementary material, tables S1 and S2; reduced models: electronic supplementary material, tables S3 and S4). The biological relevance of a borderline significant interaction term in the embryo viability GLMM (CORT × sex ratio bias; p = 0.047; electronic supplementary material, table S2) was unclear and therefore removed from the final model (electronic supplementary material, table S4).

Coefficients of variation (CVs) for embryo number and embryo length were calculated using the formula: CV = (standard deviation/mean) × 100, and then tested for differences among females reared in the different social environments using two-tailed z-tests compared with critical values from the t-distribution where the degrees of freedom are infinite [53].

3. Results and discussion

(a). Maternal stress and the prenatal sex ratio

The analyses revealed that there was no difference in maternal body weight or blood glucose level among females reared under different social conditions (electronic supplementary material, table S3 and figure S3). As expected, mothers exposed to the high male density conditions had elevated CORT concentrations compared to mothers subjected to high female density conditions (χ² = 14.031, p = 0.001; figure 2). While mothers exposed to non-stressful social conditions produced sons and daughters in equal proportions, mothers exposed to stressful social conditions produced female-biased sex ratios (between treatments: χ² = 4.711, p = 0.030; parity within treatments: high female density conditions: z-value = −0.666, p = 0.505, high male density conditions: z-value = −3.625, p = <0.001; electronic supplementary material, tables S4 and S5; figure 2). Although it cannot be determined unequivocally, the analyses do support a causal relationship between CORT concentration and offspring sex ratio. Overall, across both treatments, there was a significant negative relationship between CORT concentration and the proportion of male offspring produced (figure 2; electronic supplementary material, table S4). Further, when CORT concentration and treatment were included in the analysis together, CORT concentration accounted for a greater amount of the variation in offspring sex ratio than the social environment, which as a result dropped out of the analysis as non-significant (electronic supplementary material, table S4). This result makes it seem unlikely that exposure to high male conditions leads to sex allocation shifts via an alternative, non-stress-related pathway. An association between maternal CORT concentration and sex ratio could arise if the ‘cost’ of producing male versus female offspring differs during early development, for example if gestating female embryos impose greater maternal stress than male embryos. In turn, this could lead to different patterns of embryo viability among mothers carrying female- versus male-biased litters, with the expectation that a high proportion of female embryos leads to a general increase in reproductive failure owing to elevated maternal stress. Here, female embryo mortality was higher than male mortality (χ² = 13.406, p = <0.001; electronic supplementary material, table S4; figure 3a); however, embryo viability was not linked to the direction of sex ratio bias nor maternal CORT concentration (electronic supplementary material, table S4). The idea that female offspring impose greater maternal stress during gestation actually contradicts what is known of differences in the energetic requirements of male and female embryos [54]. For example, a study on the bank vole (Myodes glareolus) found that maternal daily energy expenditure was greatest among females that were concurrently gestating and lactating male-biased litters, pointing to a higher cost of producing sons than daughters [55]. Elevated maternal stress therefore appears to be the cause, and not a consequence, of producing high numbers of female offspring.

Figure 2.

Social stress leads to high CORT levels and female-biased sex ratios. CORT levels and offspring sex ratios were higher among mothers reared in a high male density environment (a) compared to mothers reared in a high female density environment (b). The figure shows the observations (pink and blue circles), the prediction from the model in the scale of response variable (pink and blue solid lines) and the 95% confidence interval from the bootstrapped sample (shadowed area). The model prediction lines calculated with bootstrapping (n = 10 000) are just visible beneath the solid lines (pink and blue dashed lines). Sex ratios falling below the grey dashed line are female-biased. The inserts display the CORT concentration data as a boxplot for each treatment (ng ml⁻¹; data points have been jittered). (Online version in colour.)

Figure 3.

Sex-biased offspring production, not mortality, as the explanatory mechanism (a). Boxplots displaying the proportion of viable son and daughter embryos for mothers reared in different social environments (n = 340 embryos from 46 females). Son viability was consistently high (electronic supplementary material, table S2). Extended fitness implications of the sex ratio skew for sons (b,c). Boxplots displaying the proportion of sons and daughters that were positioned in utero between two female (0M) or two male embryos (2M) for mothers reared in different social environments (n = 248 embryos from 46 females; asterisk (*) indicates significance at p < 0.05; electronic supplementary material, table S2). The in utero sex ratio skew did not influence the intrauterine position of daughter embryos (b). Higher proportions of 0M sons were produced in the high male density environment and higher proportions of 2M sons were produced in the high female density environment (c). Data points have been jittered. (Online version in colour.)

(b). Evidence of a low-cost sex allocation mechanism

Previous studies of mammals have produced qualitatively similar results to what was uncovered here—evidence which suggests that female-biased offspring production is a response to maternal stress that escalates around the time of conception [56–59] (although see [60,61])—but little is known about the underlying mechanism(s) responsible for stress-related offspring sex ratio manipulation. Mechanisms of sex ratio adjustment that operate late in development can be energetically costly and maladaptive for female mammals (e.g. differential embryo resorption; [62]). Instead, the most beneficial maternal strategy is to establish a bias in the primary or implantation sex ratio, which precedes any significant energetic investment in the offspring [63]. The current investigation provides novel insight into sex allocation mechanism(s) in operation, which can be isolated to operating either prior to or during implantation. Importantly, because both viable and non-viable in utero embryos were genetically sexed, it was established that skewed sex ratios produced under stressful social conditions were not attributable to sex-specific mortality during development (figure 3a). Further, there was no difference in overall offspring number between stressed and non-stressed mothers (electronic supplementary material, table S4), suggesting that there was no difference in egg/zygote death among mothers exposed to high male and high female density conditions during the earliest stages of reproduction. Rather, if mothers from both treatments ovulated an excess of eggs, which is typically a trait of female house mice (i.e. often up to 10 or more eggs are ovulated than what is implanted; see [64]), there would be opportunity for prenatal sex ratio biases to be established during implantation.

Maternal stress may influence the viability or competitiveness of X- and Y-chromosome-bearing sperm (CBS) within the reproductive tract [63]. For example, differences between the in utero survival of X- and Y-CBS subpopulations may be influenced by physical and/or biochemical changes in the female tract that are linked to maternal stress, and which result in sex-skewing during fertilization [65]. Alternatively, different immune responses to X- and Y-CBS elicited by specific regions of the female tract may lead to variation in the sex ratio of implanted embryos [10]. Sexual dimorphism and sex-specific signalling of conceptuses reinforce the possibility that offspring sex ratios can be facultatively adjusted during implantation [66]. The reproductive window during which the offspring sex ratio bias was laid down in this experiment (i.e. after insemination but no later than during implantation) supports the hypothesis for the evolution of prenatal sex-biasing mechanisms and builds on accumulating evidence that during challenging times mammalian mothers exert control over the sex of their offspring.

(c). Fitness implications of female-biased sex ratios in a male-dominated environment

The social conditions that parents experience at the time of reproduction may be reflective of the environment that their offspring will face. In order to offset the significant cost of reproduction and maximize their inclusive fitness, mothers especially have a vested interest in ensuring that offspring phenotype aligns with local conditions. It could be argued that offspring sex is the phenotypic trait that will have the greatest influence on the success of an individual in different social contexts. For example, the production of male offspring under high male density conditions would be maladaptive as sons will be forced to compete with their non-dispersing brothers, as well as unrelated local resident males, for access to females [2]. Male–male competition becomes more intense when females mate with multiple males and the sperm of those rivals are forced to compete for fertilizations [67]. Under such conditions, however, competition among female offspring is negligible. Rather, when male density is high and daughters are guaranteed of high mate availability, the production of female offspring will provide the best assurance for grand-parentage and consequently the most favourable fitness outcome [68]. The biased production of daughters under high male density conditions observed here in house mice aligns with theory which predicts that mothers will manipulate offspring sex ratios to avoid reproductive competition among their sons [2]. Although this result is seemingly adaptive, it must be noted that relationships between maternal condition and offspring sex ratios can change under different environmental conditions (e.g. see [69,70] on the red deer, Cervus elaphus). Further, changes in the demography of the local environment over time, for example owing to patterns of (sex-specific) dispersal or differences in the extent to which generations overlap, mean that the maternal environment would not necessarily reflect the environment that their sons experience and there is a high degree of complexity when it comes to the influence of prevailing local conditions (both social and ecological) on strategies of sex allocation [1]. Here, in this experiment, I focused on sex allocation in relation to maternal stress in isolation and discovered that social stress may have the potential to generate beneficial fitness outcomes by leading to an offspring sex bias that could offset future male–male competition. Of course, in nature, a suite of other factors may influence the direction and strength of this result.

The identification of a bias in the prenatal sex ratio can provide insights into subtle fitness implications that would otherwise go undetected if birth sex ratio was assessed. In litter-bearing mammals, the prenatal phase is a critical developmental period during which offspring are exposed to a milieu of hormones owing to (i) the close intrauterine proximity to neighbouring embryos and (ii) placental transfer from the mother [71]. In utero sex ratio skews will affect the probability that an embryo will neighbour embryos of the same- or opposite-sex. In this experiment, female embryos were positioned in equal proportions between two male (2M) and two female (0M) embryos across both treatments (electronic supplementary material, table S4 and figure 2b). However, the sons of stressed mothers (female-biased sex ratio) were more likely to develop between two female embryos (0M) (χ² = 4.570, p = 0.033), while sons of non-stressed mothers (equal sex ratio) were more likely to be positioned between two male embryos (2M) (χ² = 5.653, p = 0.017) (electronic supplementary material, table S4; figure 3c). The intrauterine position has been shown to affect the male adult phenotype in relation to traits that influence competitive ability (e.g. testosterone level, body size, testes size, scent marking frequency and territory size); 2M males ultimately becoming more masculinized, and 0M males more feminized, in adulthood (reviewed [72]). As a consequence, females prefer to mate with 2M males over 0M males [73]. Therefore, a prospective cost of producing female-biased offspring sex ratios is the making of non-competitive sons that have low reproductive potential. Following birth, 0M males receive less maternal attention compared to 2M males [74], suggesting that discriminative postnatal care may be an effective maternal strategy for minimizing expenditure on sons that are destined to be evolutionary dead ends.

(d). Further fitness implications of socially induced stress

In mice, different social environments experienced during postnatal development can affect the adult phenotype in terms of altered body size (e.g. high male social conditions lead to increased male body size; see [36,42]). In the current study, although there was no overall difference in body size among females maturing under high male (relatively stressful) and high female (relatively non-stressful) conditions, the relationship between maternal body weight and CORT concentration interacted in different ways: under high male density conditions, ‘more stressed’ females were heavier while under high female density conditions ‘less stressed’ females were heavier (χ² = 5.591, p = 0.018; electronic supplementary material, table S3 and figure S4). These differences may reflect variation in activity levels under different social conditions, for example if generally high stress levels are associated with greater activity levels and lower body weights, but above some critical threshold the opposite is observed. In support, it has been shown that mice experiencing elevated baseline CORT levels exhibit high locomotor activity, but significant elevations lead to affective disorders (e.g. depression) characterized by reduced activity levels [75]. Further investigation is required to fully understand the difference in the direction of the relationship between body weight and stress level among female mice exposed to different social conditions.

In addition, prenatal maternal stress can affect in utero offspring development in terms of altered growth rates [76]. In many mammals, including humans, prenatal exposure to elevated maternal glucocorticoids have been shown to cause a wide range of stress effects, including both accelerated (adaptive developmental plasticity hypothesis) and deaccelerated (developmental constraint hypothesis) growth [77]. Here, the potential for flow-on fitness implications of maternal stress becomes evident in an analysis of offspring size, which revealed that the embryos of mothers exposed to a stressful environment are, on average, smaller than the embryos of mothers exposed to a non-stressful environment (χ² = 5.908, p = 0.015; non-sex-specific effect; electronic supplementary material, table S3; figure 4). Stunted in utero growth owing to lowered maternal investment during stressful periods can result in delayed sexual maturation and/or small adult size, which in turn may reduce reproductive rate or lifespan [78]. Yet, a meta-analysis has demonstrated that stunted offspring growth trajectories imposed by maternal stress during early development are often accelerated during juvenility [77]. This suggests that small in utero size may not necessarily persist in the adult phenotype. Further, although there was no overall difference in offspring number (electronic supplementary material, table S4), variation in the embryo number was significantly reduced in mothers exposed to a stressful environment compared to mothers exposed to a non-stressful environment (CV: z-value = 2.813, p = 0.005; electronic supplementary material, table S6; figure 4). This result highlights the possibility that during periods of stress, mothers are strategically ‘selecting’ offspring number before or at the time of implantation. In line with offspring sex ratio adjustments being adaptive, a mechanism that selects for the optimal offspring number at the earliest reproductive stage represents the most cost-effective strategy for mothers in the poor condition. In this sense, prenatal maternal stress effects on both offspring number and in utero growth (i.e. when growth trajectories are accelerated later in life) allow both mothers and offspring to make the best of a bad situation [77].

Figure 4.

Reduced offspring size and optimized offspring number under stressful social conditions. There was no relationship between average embryo number and embryo size (r = 0.109, p = 0.223); however, there was less variation in the offspring number among mothers reared in the high male density environment (blue circles, solid line) compared to mothers reared in the high female density environment (pink circles, dashed line) (electronic supplementary material, table S4). These mothers also produced, on average, smaller embryos compared to mothers reared in the high female density environment (electronic supplementary material, table S1). Lines represent best fit. Data points have been jittered. (Online version in colour.)

4. Conclusion

The impact of maternal stress on juvenile development has intrigued scientists for many years. There has been a strong focus on the adverse effects of prenatal stress in terms of offspring cognitive ability, which relates to social behaviour [79]. In comparison, very little is known on the influence of different social conditions on maternal stress, and even less on how, in turn, socially induced maternal stress shapes offspring phenotype. The adaptive significance of maternal stress remains an ongoing core debate in diverse research fields, including animal ecology and human health. Despite the significance of this debate, many important offspring traits—including sex—have been ignored [29]. Here, using house mice, it was shown that exposure to high male density conditions during development leads to elevated maternal stress hormone levels and a bias in the prenatal sex ratio that aligns with theoretical predictions [2,4]. It was identified that the sex ratio bias was cost-effective (i.e. biased female production), and therefore seemingly an adaptive strategy, rather than a passive consequence (i.e. biased male mortality) of maternal stress. To unequivocally demonstrate adaptive significance, the next step is to show that the fitness of females that adjust offspring sex ratios is, in fact, greater than those that do not. In terms of evolutionary fitness, females will retain greater capacity for future reproductive events by mitigating wasted investment in the unwanted sex. Thus, a sex-biasing mechanism(s) that operates around the time of conception offers mothers the most cost-effective means of influencing offspring sex ratios. Finally, this investigation demonstrates that glucocorticoids are effective in translating information from the social environment into sex allocation responses that have the potential to be adaptive [29]. This outcome is relevant to a broad range of research areas, including species conservation and the management of threatened or fragmented populations [12,15] as well as captive breeding programmes [16].

Ethics

The work reported in this article followed the guidelines for the ethical treatment of animals in research under UWA Ethics Committee approval (03/100/1456).

Data accessibility

The data file is provided in the online electronic supplementary material.

Authors' contributions

R.C.F. designed the study, performed the research, analysed the data and wrote the manuscript.

Competing interests

I declare I have no competing interests.

Funding

This research was funded by the Australian Research Council (FT180100625).