Neglected patterns of variation in transgenerational plasticity: The importance of different sources of environmental variation differs across ages and sexes in a cyprinid fish
Abstract
- Adaptive transgenerational plasticity (TGP) requires individuals to integrate environmental experience across multiple sources. However, few empirical studies have considered that the relative relevance of certain sources might vary across ontogeny and sexes.
- Here, we address this knowledge gap by studying inducible antipredator defences, one of the most convincing examples of TGP. We assessed individual and combined effects of perceived high predation risk in mothers, fathers, caring males and personal environments on the morphology of juvenile, adult male and adult female cyprinids Pimephales promelas.
- Parental rather than personal environmental experience determined morphological defence expression across ages and sexes, likely because parents had a longer sampling period.
- In juveniles and adult males, egg-mediated environmental experience outweighed sperm-mediated environmental experience in the induction of body shape differences, likely because eggs can transmit information beyond epigenomes. However, in adult females, where body shape responses can be interpreted as life-history plasticity, information from egg and sperm were equally important, likely resulting from different integration mechanisms between morphological and life-history plasticity.
- The importance of care-mediated relative to gamete-mediated variation changed between juveniles and adult males, likely because they represent short- and long-term environmental experience, respectively. Instead, in adult females, both sources were again equally important, potentially owing to lag-times of life-history plasticity. Parental care intensity only contributed marginally to defence formation.
- These results highlight age- and sex-specific prioritization of different environmental experiences so as to generate optimal phenotypes.
Read the free Plain Language Summary for this article on the Journal blog.
1 INTRODUCTION
Phenotypic plasticity is the ability of individual genotypes to respond to environmental change within one generation by producing different phenotypes (West-Eberhard, 2003). During plastic adaptation, not only personal experience with the environment but also environmental experience from past generations contribute to the resulting phenotype through mechanics of non-genetic inheritance, which is referred to as transgenerational plasticity (TGP; Salinas et al., 2013; Meuthen, 2022). Numerous studies revealed not only the relevance of an individual's personal environment (West-Eberhard, 2003) but also its maternal environment (Badyaev, 2008), paternal environment (Curley et al., 2011) as well as the environment of the caring parent (Head et al., 2012; Weaver et al., 2004). Since environmental experience can differ between these different sources, individuals face the challenge of adaptively integrating this variation to generate a fine-tuned phenotype. However, we still lack a good understanding of the relative importance of these different environments on phenotype formation during TGP, and this is what contributes to controversial meta-analysis results regarding its adaptiveness (Sánchez-Tójar et al., 2020).
Recently, it was suggested that we may better understand variation in plasticity by measuring responses across development (Dupont et al., 2023). This may also be true for the integration during TGP as individuals of different ages rarely share the same ecological niches and environments, which should logically cause variation in the relative importance of different sources of environmental experience for phenotype expression. We now develop a set of not always mutually exclusive hypotheses that may explain why variation across ontogeny occurs.
First, theory predicts that the relative priority of parental and personal environmental cues crucially depends on the ontogenetic stage of individuals (Stamps & Bell, 2021). That is because the relative duration of the exposure period to an environment differs between parents and offspring (Stamps & Bell, 2021). The generation that samples environmental variation over a longer absolute time period is expected to be the predominant force in shaping offspring phenotypes. Thus, during early life, parental environmental experience should dominate whereas later in development, personal environmental experience should take over (McNamara et al., 2016; Stamps & Bell, 2021 but see Moore et al., 2019). We call this empirically supported theory (Chang et al., 2021; Shama et al., 2014) the “sampling-duration-degradation” hypothesis (Figure 1a).
Second, there are two mutually exclusive theories on whether maternal or paternal environmental experience is more relevant for offspring phenotypes. First, the theory that we call the “shared-niche” hypothesis, postulates that the parental sex that shares the same ecological niche as their offspring and therefore better samples the related environmental variation differs between offspring ontogenetic stages and sexes (Bell & Hellmann, 2019). In other words, in early life, the parental sex that shares their environment with offspring should constitute the predominant environmental experience whereas in adulthood, the parental sex which matches the offspring sex should become a more reliable predictor of future environmental conditions (Figure 1b). This theory is empirically supported (Hellmann et al., 2020; Herman et al., 2014). Second, in the absence of parental care, mothers can not only transmit epigenomes but also embed nutrients, hormones, antibodies and enzymes within their large eggs (Boulinier & Staszewski, 2008; Giesing et al., 2011). This “maternal access” hypothesis (Figure 1c) may explain why maternal effects often outweigh paternal ones across ontogeny (Chang et al., 2021; Hellmann et al., 2020).
Third, irrespective of the previous environment, environmental change during the relatively short parental care period alone is already sufficient in inducing clear phenotypic changes in offspring (Stein & Bell, 2014). At the same time, to affect gametes, environmental change needs to happen during embryonic gonadal development or during gametogenesis (Tariel-Adam et al., 2023). We thus hypothesize that caring parent information represents short-term environmental experience with a short lag-time between generations whereas gametes contain more long-term environmental experience. A switch between parental care and gamete-mediated environmental experience across ontogeny is also observable across the results of two empirical studies (Hellmann et al., 2021; Meuthen et al., 2021). Thus, in a theory that we term the “gametic-care-ontogenetic-switch” hypothesis, we postulate that in early ontogeny, care-mediated information should dominate, whereas in late ontogeny, gamete-mediated information should take over (Figure 1d).
Here, we comprehensively study these four theories by taking advantage of cue-induced antipredator defences, which are one of the first described cases of adaptive phenotypic plasticity (Brönmark & Miner, 1992) and one of the most convincing examples of adaptive TGP (Uller et al., 2013). As our model system, we use the fathead minnow Pimephales promelas, an established model system for antipredator TGP (Crane et al., 2021; Meuthen et al., 2021, 2023). This small (6.0–7.4 cm) sexually dimorphic common prey cyprinid fish is widespread across Northern American freshwaters (Scott & Crossman, 1998). Ecological niches differ between ages and sexes: juveniles and females form shoals in the open water whereas adult males establish benthic territories. Here, they provide alloparental care to clutches (Unger & Sargent, 1988) by continuously mouthing and rubbing the eggs until they hatch 4 days later (Scott & Crossman, 1998). Predation risk varies strongly across different P. promelas populations (Mathis et al., 1993) but as common predators of this species live longer than multiple P. promelas generations (Clark & Steinbach, 1959), within-population risk is relatively stable, which allows for reliable predictions of future risk levels across generations. In the presence of predator-related cues, P. promelas develops typical fish morphological defences (Meuthen et al., 2019), that is deeper bodies, which increase escape performance (Domenici et al., 2008) and handling times for gape-limited predators (Nilsson et al., 1995). Here, we study variation in this morphological defence at two ontogenetic stages and in both sexes after manipulating perceived predation risk (high or low) from birth onwards and then crossing these treatments across paternal, maternal, parental care and personal environments.
2 MATERIALS AND METHODS
All procedures followed the Canadian Council on Animal Care guidelines for humane animal use and were approved by the University of Saskatchewan's Animal Research Ethics Board (Animal Use Protocol: 20170089).
2.1 Experimental fish
We derived P. promelas from a laboratory population (details in the Supporting Information S1) and adopted a split-clutch approach to control for genetic effects (Figure 2a). Offspring were lifelong exposed to either conspecific alarm cues (high-risk, concentration of injected liquid: 0.0067 cm2 conspecific skin/L, resulting in 3.302 × 10−6 cm2 skin/L within tanks) or a distilled water treatment (low-risk, likewise injected in 1 mL aliquots). Alarm cues are a reliable signal of high predation risk (Chivers & Smith, 1998) that induce morphological defences across taxa, including P. promelas (Meuthen et al., 2019). We crossed risk treatments across maternal, paternal and personal environments in the absence of parental care (2 × 2 × 2 design, Figure 2b). We also crossed biparental risk with caring parent risk (2 × 2 design, Figure 2c). As not many pairs reproduced concurrently, we could not construct a fully factorial cross-fostering set-up as we initially planned. Instead, we switched remaining clutches between different parents only when the risk treatment differed between genetic parents and caring parents. Otherwise, when genetic parents and caring parents were supposed to be of the same risk treatment, genetic parents cared for their own eggs. Altogether, from 41 families, we generated 2920 offspring that we set up in 292 tanks containing 10 fish each (range 20–44 tanks per treatment). Throughout the study, tank volumes, water change frequency and food regimens were increased sequentially to conform to the requirements of growing fish (details in Supporting Information S1). To assess individual morphology, we photographed individuals at 18 days of age (completion of larval development, see Devlin et al., 1996), and at 180days of age (first signs of sexual maturity, see Scott & Crossman, 1998). For 18-day-old juveniles, we always sampled fish from one replicate tank per family and treatment; in total we sampled from 158 tanks (range 10–24 tanks per treatment, see Figure 2). For 180-day-old adults, due to individual mortality and unequal distribution of sexes between tanks, we instead sampled from all tanks that remained with living fish, that is from 158 tanks (range 9–27 tanks per treatment, see Figure 2).
2.2 Parental care intensity
To assess parental care intensity, we videotaped (C922x Pro Stream, Logitech, Suzhou, China) the 20 males caring for 40 clutches (each caretaker cared for 1–3 clutches) every day for a period of 10 min each at a randomly selected time between 1600 and 1900 h, before feeding. We did so in total four times, until eggs hatched.
2.3 Photography procedure
Using the same procedures as described in Meuthen et al. (2019), we photographed 1560 juveniles and 1431 adult fish along with size-standards (E-3, Olympus, Japan). Deformed individuals (38 fish) and adults that could not be clearly assigned to a sex (31 fish) were not used; in total we analysed 1555 photographs from juveniles and 1367 photographs from adults (Figure 2).
2.4 Replication statement
| Scale of inference | Scale at which the factor of interest is applied | Number of replicates at the appropriate scale |
|---|---|---|
| Individuals | Families (full-sibling groups) | Biparental low-risk: 12 families, biparental high-risk: 11 families, maternal high-risk and paternal low-risk: 10 families; maternal low-risk and paternal high-risk: 10 families; more detail in Figure 2 |
2.5 Data analysis
2.5.1 Parental care intensity
We analysed parental care behaviour only during the last 5 min of each video, a common time period for minnow parental care studies (Sargent, 1988). As a proxy for parental care intensity, we measured the average proportion of time that caring males spent within one standard length of the clutch, inside the breeding tile to which the eggs were attached, as established by Meuthen et al. (2021). In total, we analysed 160 videos; for each clutch we calculated average parental care intensity over its four videos. Meuthen et al. (2021) analysed the same dataset and found that instead of treatment, parental care intensity was shaped by the day of care (it consistently increased over time until eggs hatched), the proportional change in clutch size that results from taking out eggs or switching clutches (less care was provided when clutches became smaller and more care when they got larger), and by whether own or adopted eggs were taken care of (on average 22% less care was provided for adopted eggs). Here we focus on analysing the effect of parental care intensity on offspring defence expression.
2.5.2 Photographs
We converted raw photographs to PNG format using Stepok RAW Importer (Stepok Image Lab, New Zealand). With tpsDig v.2.30, we set the scale in accordance with the size standard present in every image and assessed individual body shape by placing the same thirteen landmarks that have been established for P. promelas geometric morphometrics in Meuthen et al. (2019). We digitized and analysed body shapes from juveniles, adult males and adult females separately.
2.6 Statistical analysis
As we do not have information on juvenile sex available, because ontogenetic variation far outweighs treatment-related variation, and because ontogenetic and sex differences in the plasticity of our model system are already well-established (Meuthen et al., 2019), we produced three different datasets (juveniles, adult males, adult females) for analysis.
All analyses were conducted with R v.4.2.1 (R Core Team, 2021). Following Procrustes superimposition using the function “gpagen” from the “geomorph” package v.4.0.4 (Adams et al., 2021), we evaluated treatment effects on individual body shape using the function “lm.rrpp” from the “RRPP” package v.1.3.1, which is a new-generation non-parametric permutation-based ANOVA that generalizes the statistics used in univariate ANOVA to multivariate data (Collyer & Adams, 2018). Each model was run with 10,000 permutations. As this model does not require normally distributed data and can analyse variation across all shape landmarks in one analysis, which avoids the issue of inflated Type I errors that arises from analysing principal components separately, its application is widespread across modern plasticity studies (de Jong et al., 2022; Machida et al., 2021; Petrosino et al., 2023). Unfortunately, it is not possible to fit random effects using this new-generation approach. Furthermore, the inclusion of fixed effects that are not shared across all other fixed effects leads to truncated and incomplete models. Thus, we cannot enter family effects (which are not shared across different parents) in full models. However, we control for family effects by entering family identity as a fixed effect when investigating the isolated effects of personal risk or caring parent risk, where the same family could be shared across treatments. Controlling for tank effects with this new-generation approach is unfortunately impossible across any of the models as individual tanks were never shared across different treatments.
First, for each ontogenetic stage and sex, we analysed variation in body shape in the absence of parental care (Figure 2b) by entering maternal, paternal and personal risk as well as all possible two-way and three-way interactions as fixed effects.
Second, for each ontogenetic stage and sex, we analysed variation in body shape in the presence of parental care (Figure 2c) by entering biparental (gametic) risk, caring parent risk, as well as the biparental risk × caring parent risk interaction as fixed effects. We also entered parental care intensity as a covariate. Interactions involving parental care intensity were not added to the model as our number of samples below 50% parental care intensity was too low to reliably estimate such interactions (see Figure S8).
As the calculation of an AIC score for high-dimensional multivariate model selection is still in strong need of further theoretical development (M. Collyer, personal communication), here we drop non-significant (p > 0.05) interactions from full models to generate final models with reliable estimates (Engqvist, 2005). However, we avoided the removal of non-significant fixed effects as all entered effects were considered meaningful and this practice also avoids pseudoreplication associated with test-qualified pooling (Colegrave & Ruxton, 2018). In the presence of significant interactions, we split the dataset to estimate main effects for each level separately. We applied Type III ANOVAs in the presence of interactions, Type II ANOVAs in the presence of multiple fixed effects and otherwise Type I ANOVAs. Effect sizes are given as phenotypic (Euclidean) distances (d) along with 95% confidence intervals in square brackets. Because distances are without direction, all confidence intervals are one-tailed.
To elucidate the direction of body shape changes reflected by our analysis, from each of our final models, instead of extracting principal components according to their eigenvalues, which has been criticized (Morton & Altschul, 2019), we extracted all principal components that each explained more than 2% of phenotypic variation (Figures S1–S6), as established in other studies of morphological plasticity (Frommen et al., 2011; Meuthen et al., 2019). Then, we selected the components corresponding to known typical antipredator phenotypes for discussion (juveniles without parental care: PC1, juveniles with parental care: PC1-2, adult males without parental care: PC1-3, adult males with parental care: PC1-2, adult females without parental care: PC1, adult females with parental care: PC1-2). The detailed selection process and shape changes across all other principal components each explaining more than 2% of phenotypic variation are shown in the Supporting Information S2 (Figures S9–S16).
Preliminary analyses revealed that centroid size (which corresponds to surface area size) effects largely mirror the observed treatment effects on body shape (as shape changes largely manifest in an increase in body depth, which is synonymous with an increase in surface area). Unsurprisingly, including centroid size as a covariate to control for allometry eliminates most other effects. Thus, we did not analyse centroid size separately and, following practices established in Meuthen et al. (2019), also did not include it as a covariate in models. Instead, to control for allometry, we analysed variation in standard length (i.e. the distance from the tip of the snout to the base of the tail fin) using linear mixed-effect models with the same fixed effects along with family identity nested in tank identity as a random intercept (see Supporting Information S3). The only outcome of this analysis was that in juveniles that were raised in the absence of parental care, personal risk accelerated growth in a statistically significant way only when both paternal and maternal risk were low. Otherwise, body size did not respond to the risk treatments or parental care intensity in a statistically significant way, suggesting that our results are not confounded by plastically altered growth speed.
3 RESULTS
3.1 Juveniles
In juvenile fathead minnows that were raised in the absence of parental care, we observed a statistically significant three-way interaction between maternal, paternal and personal risk environments in the multidimensional model (Table 1). We then split the dataset to disentangle this interaction. This revealed first that, paternal risk, as evidenced by consistently smaller-than-average effect sizes, had a lower impact on the formation of antipredator phenotypes than personal and maternal risk (Table 2). Exposure to high paternal risk consistently generated bodies with average depth (Figure 3a). Second, maternal risk effect magnitudes were more often above average compared to personal risk (Table 2). Maternal risk outweighs personal risk effects in the formation of deep-bodied antipredator phenotypes (Figure 3a).
| Fixed effects | Estimated distance [95% CI] | R 2 | df | F | p |
|---|---|---|---|---|---|
| Juveniles | |||||
| Maternal risk | 0.004 [0, 0.005] | 0.001 | 1 | 1.381 | 0.188 |
| Paternal risk | 0.006 [0, 0.005] | 0.002 | 1 | 2.690 | 0.021 |
| Personal risk | 0.004 [0, 0.005] | 0.002 | 1 | 1.940 | 0.064 |
| Maternal risk × paternal risk | 0.007 [0, 0.008] | 0.002 | 1 | 2.045 | 0.055 |
| Maternal risk × personal risk | 0.007 [0, 0.007] | 0.002 | 1 | 1.791 | 0.087 |
| Paternal risk × personal risk | 0.004 [0, 0.007] | 0.001 | 1 | 0.859 | 0.502 |
| Maternal risk × paternal risk × personal risk | 0.012 [0, 0.010] | 0.003 | 1 | 3.006 | 0.012 |
| Residuals | 0.978 | 1141 | |||
| Adult males (full model) | |||||
| Maternal risk | 0.007 [0, 0.008] | 0.003 | 1 | 1.624 | 0.108 |
| Paternal risk | 0.008 [0, 0.007] | 0.005 | 1 | 2.225 | 0.028 |
| Personal risk | 0.007 [0, 0.006] | 0.005 | 1 | 2.339 | 0.023 |
| Maternal risk × paternal risk | 0.010 [0, 0.012] | 0.003 | 1 | 1.329 | 0.209 |
| Maternal risk × personal risk | 0.008 [0, 0.010] | 0.003 | 1 | 1.238 | 0.259 |
| Paternal risk × personal risk | 0.008 [0, 0.010] | 0.003 | 1 | 1.253 | 0.241 |
| Maternal risk × paternal risk × personal risk | 0.012 [0, 0.016] | 0.002 | 1 | 1.090 | 0.344 |
| Residuals | 0.975 | 465 | |||
| Adult males (final model) | |||||
| Maternal risk | 0.004 [0, 0.004] | 0.002 | 1 | 2.757 | 0.010 |
| Paternal risk | 0.003 [0, 0.004] | 0.001 | 1 | 1.358 | 0.193 |
| Personal risk | 0.005 [0, 0.004] | 0.002 | 1 | 2.929 | 0.008 |
| Residuals | 0.984 | 469 | |||
| Adult females (full model) | |||||
| Maternal risk | 0.008 [0, 0.007] | 0.004 | 1 | 2.335 | 0.022 |
| Paternal risk | 0.008 [0, 0.008] | 0.004 | 1 | 2.220 | 0.030 |
| Personal risk | 0.004 [0, 0.006] | 0.001 | 1 | 0.761 | 0.623 |
| Maternal risk × paternal risk | 0.015 [0, 0.011] | 0.007 | 1 | 3.724 | 0.001 |
| Maternal risk × personal risk | 0.006 [0, 0.010] | 0.001 | 1 | 0.817 | 0.575 |
| Paternal risk × personal risk | 0.006 [0, 0.010] | 0.001 | 1 | 0.638 | 0.749 |
| Maternal risk × paternal risk × personal risk | 0.010 [0, 0.015] | 0.002 | 1 | 0.921 | 0.471 |
| Residuals | 0.975 | 551 | |||
| Adult females (final model) | |||||
| Maternal risk | 0.009 [0, 0.005] | 0.005 | 1 | 6.183 | <0.001 |
| Paternal risk | 0.006 [0, 0.005] | 0.002 | 1 | 2.685 | 0.012 |
| Personal risk | 0.003 [0, 0.004] | 0.001 | 1 | 1.002 | 0.412 |
| Maternal risk × paternal risk | 0.011 [0, 0.007] | 0.004 | 1 | 4.795 | <0.001 |
| Residuals | 0.981 | 554 | |||
- Note: Bold p-values indicate p < 0.05 in the final models.
| Estimated distance [95% CI] | Effect magnitude | R 2 | F | p | |
|---|---|---|---|---|---|
| Juveniles | |||||
| Maternal low-risk | |||||
| Personal × paternal risk | 0.009 [0, 0.007] | 0.006 | 3.278 | 0.006 | |
| Maternal low-risk, paternal low-risk | |||||
| Personal risk | 0.006 [0, 0.004] | ↑ | 0.003 | 0.010 | <0.001 |
| Family | 0.204 | 7.097 | <0.001 | ||
| Maternal low-risk, paternal high-risk | |||||
| Personal risk | 0.004 [0, 0.005] | ↓ | 0.006 | 1.432 | 0.169 |
| Family | 0.182 | 5.191 | <0.001 | ||
| Maternal low-risk, personal low-risk | |||||
| Paternal risk | 0.006 [0, 0.005] | ↓ | 0.002 | 0.008 | 0.013 |
| Maternal low-risk, personal high-risk | |||||
| Paternal risk | 0.005 [0, 0.005] | ↓ | 0.001 | 0.009 | 0.054 |
| Maternal high-risk | |||||
| Personal × paternal risk | 0.004 [0, 0.005] | 0.001 | 0.859 | 0.500 | |
| Paternal risk | 0.006 [0, 0.003] | ↓ | 0.114 | 6.707 | <0.001 |
| Personal risk | 0.003 [0, 0.003] | ↓ | 0.003 | 1.941 | 0.075 |
| Paternal low-risk | |||||
| Personal × maternal risk | 0.007 [0, 0.007] | 0.004 | 2.199 | 0.039 | |
| Paternal low-risk, personal low-risk | |||||
| Maternal risk | 0.009 [0, 0.005] | ↑ | 0.023 | 3.998 | <0.001 |
| Paternal low-risk, personal high-risk | |||||
| Maternal risk | 0.006 [0, 0.005] | ↑ | 0.015 | 2.523 | 0.006 |
| Paternal low-risk, maternal low-risk | |||||
| Personal risk | 0.006 [0, 0.004] | ↑ | 0.010 | 4.332 | <0.001 |
| Family | 0.204 | 7.097 | <0.001 | ||
| Paternal low-risk, maternal high-risk | |||||
| Personal risk | 0.003 [0, 0.004] | ↓ | 0.004 | 1.215 | 0.262 |
| Family | 0.295 | 10.681 | <0.001 | ||
| Paternal high-risk | |||||
| Personal × maternal risk | 0.007 [0, 0.007] | 0.003 | 1.813 | 0.082 | |
| Personal risk | 0.003 [0, 0.004] | ↓ | 0.002 | 1.260 | 0.236 |
| Maternal risk | 0.004 [0, 0.003] | ↓ | 0.005 | 3.096 | 0.009 |
| Personal low-risk | |||||
| Maternal × paternal risk | 0.012 [0, 0.007] | 0.009 | 6.503 | <0.001 | |
| Personal low-risk, maternal low-risk | |||||
| Paternal risk | 0.006 [0, 0.005] | ↓ | 0.008 | 2.868 | 0.013 |
| Personal low-risk, maternal high-risk | |||||
| Paternal risk | 0.007 [0, 0.005] | ↑ | 0.013 | 4.771 | <0.001 |
| Personal low-risk, paternal low-risk | |||||
| Maternal risk | 0.009 [0, 0.005] | ↑ | 0.023 | 8.232 | <0.001 |
| Personal low-risk, paternal high-risk | |||||
| Maternal risk | 0.006 [0, 0.005] | ↓ | 0.010 | 3.376 | 0.006 |
| Personal high-risk | |||||
| Maternal × paternal risk | 0.007 [0, 0.007] | 0.005 | 2.195 | 0.041 | |
| Personal high-risk, maternal low-risk | |||||
| Paternal risk | 0.005 [0, 0.005] | ↓ | 0.009 | 2.005 | 0.054 |
| Personal high-risk, maternal high-risk | |||||
| Paternal risk | 0.006 [0, 0.005] | ↓ | 0.012 | 2.784 | 0.017 |
| Personal high-risk, paternal low-risk | |||||
| Maternal risk | 0.006 [0, 0.005] | ↑ | 0.015 | 3.372 | 0.006 |
| Personal high-risk, paternal high-risk | |||||
| Maternal risk | 0.004 [0, 0.005] | ↓ | 0.007 | 1.496 | 0.147 |
| Adult females | |||||
| Maternal low-risk | |||||
| Paternal risk | 0.006 [0, 0.005] | ↓ | 0.011 | 2.823 | 0.008 |
| Maternal high-risk | |||||
| Paternal risk | 0.006 [0, 0.005] | ↓ | 0.009 | 2.732 | 0.010 |
| Paternal low-risk | |||||
| Maternal risk | 0.006 [0, 0.005] | ↓ | 0.009 | 2.203 | 0.031 |
| Paternal high-risk | |||||
| Maternal risk | 0.008 [0, 0.005] | ↑ | 0.018 | 5.788 | <0.001 |
- Note: Bold p-values indicate p < 0.05 in the final models.
In the presence of parental care, we observed a statistically significant interaction between parental and caring parent risk (Table 3). Additionally, parental care intensity had an overall small effect on offspring body shape, with higher parental care intensities inducing marginally deeper bodies as showcased mainly by the second principal component (Table 3, Figure S8a). Splitting datasets to disentangle this interaction revealed that biparental and caring parent risk were similar in their phenotypic impact, which in turn, was greatest when risk was high in the respective other treatment as well (Table 4). When risk was high in both, deep bodies were generated according to the first principal component (Figure 4a). While this coincided with a shift towards a shallower body in the second principal component (Figure S10a), it does explain a substantially lower percentage of phenotypic variation (PC1: 51.62% vs. PC2: 35.05%).
| Fixed effects | Estimated distance [95% CI] | R 2 | df | F | p |
|---|---|---|---|---|---|
| Juveniles | |||||
| Parental risk | 0.010 [0, 0.006] | 0.004 | 1 | 5.576 | <0.001 |
| Caring parent risk | 0.007 [0, 0.006] | 0.002 | 1 | 2.677 | 0.024 |
| Parental care intensity | 0.016 [0, 0.012] | 0.003 | 1 | 4.143 | 0.003 |
| Parental risk × caring parent risk | 0.013 [0, 0.009] | 0.004 | 1 | 4.846 | 0.001 |
| Residuals | 0.969 | 401 | |||
| Adult males (full model) | |||||
| Parental risk | 0.012 [0, 0.01] | 0.003 | 1 | 3.061 | 0.003 |
| Caring parent risk | 0.009 [0, 0.01] | 0.001 | 1 | 1.729 | 0.087 |
| Parental care intensity | 0.024 [0, 0.021] | 0.002 | 1 | 2.433 | 0.021 |
| Parental risk × caring parent risk | 0.013 [0, 0.013] | 0.001 | 1 | 1.713 | 0.087 |
| Residuals | 0.957 | 151 | |||
| Adult males (final model) | |||||
| Parental risk | 0.007 [0, 0.006] | 0.002 | 1 | 2.152 | 0.031 |
| Caring parent risk | 0.005 [0, 0.006] | 0.001 | 1 | 1.133 | 0.322 |
| Parental care intensity | 0.020 [0, 0.020] | 0.002 | 1 | 1.861 | 0.069 |
| Residuals | 0.968 | 152 | |||
| Adult females | |||||
| Parental risk | 0.013 [0, 0.010] | 0.003 | 1 | 3.498 | 0.001 |
| Caring parent risk | 0.007 [0, 0.009] | 0.001 | 1 | 1.234 | 0.262 |
| Parental care intensity | 0.013 [0, 0.015] | 0.001 | 1 | 1.379 | 0.184 |
| Parental risk × caring parent risk | 0.016 [0, 0.014] | 0.002 | 1 | 2.733 | 0.007 |
| Residuals | 0.969 | 174 | |||
- Note: Bold p-values indicate p < 0.05 in the final models.
| Estimated distance [95% CI] | Effect magnitude | R 2 | F | p | |
|---|---|---|---|---|---|
| Juveniles | |||||
| Low-risk care | |||||
| Parental risk | 0.007 [0, 0.006] | ↓ | 0.012 | 2.624 | 0.030 |
| Parental care intensity | 0.035 [0, 0.021] | 0.033 | 6.942 | <0.001 | |
| High-risk care | |||||
| Parental risk | 0.008 [0, 0.006] | ↑ | 0.018 | 3.683 | 0.003 |
| Parental care intensity | 0.017 [0, 0.014] | 0.015 | 3.147 | 0.008 | |
| Parental low-risk | |||||
| Caring parent risk | 0.006 [0, 0.005] | ↓ | 0.010 | 2.543 | 0.016 |
| Parental care intensity | 0.016 [0, 0.016] | 0.007 | 1.943 | 0.058 | |
| Family | 0.222 | 5.283 | <0.001 | ||
| Parental high-risk | |||||
| Caring parent risk | 0.009 [0, 0.007] | ↑ | 0.017 | 4.616 | <0.001 |
| Parental care intensity | 0.025 [0, 0.024] | 0.005 | 1.331 | 0.210 | |
| Family | 0.278 | 7.363 | <0.001 | ||
| Adult females | |||||
| Low-risk care | |||||
| Parental risk | 0.008 [0, 0.009] | ↓ | 0.014 | 1.376 | 0.188 |
| Parental care intensity | 0.035 [0, 0.027] | 0.032 | 3.155 | 0.001 | |
| High-risk care | |||||
| Parental risk | 0.011 [0, 0.01] | ↑ | 0.029 | 2.375 | 0.017 |
| Parental care intensity | 0.013 [0, 0.019] | 0.011 | 0.876 | 0.533 | |
| Parental low-risk | |||||
| Caring parent risk | 0.007 [0, 0.009] | ↓ | 0.008 | 0.901 | 0.503 |
| Parental care intensity | 0.024 [0, 0.023] | 0.016 | 1.877 | 0.056 | |
| Family | 0.295 | 7.833 | <0.001 | ||
| Parental high-risk | |||||
| Caring parent risk | 0.011 [0, 0.01] | ↑ | 0.018 | 1.849 | 0.062 |
| Parental care intensity | 0.024 [0, 0.033] | 0.013 | 1.343 | 0.199 | |
| Family | 0.293 | 3.431 | <0.001 | ||
- Note: Bold p-values indicate p < 0.05 in the final models.
3.2 Males
In adult male P. promelas that did not experience parental care, we did not find any evidence for interactions between maternal, paternal and personal risk; instead the final reduced model revealed that paternal risk did not impact body shape in a statistically significant way whereas both high maternal and high personal risk did on a similar level (Table 1). However, closer inspection reveals that this effect differs across principal components: In the first component, high personal risk alone was insufficient to induce deeper bodies, leading effects of high maternal risk to slightly outweigh personal risk (Figure 3b). In the second component, high personal risk was associated with shallower bodies and high maternal risk with deeper ones except for when personal risk was high as well (Figure S9b). In the third component, neither personal nor maternal risk impacted body depth (Figure S11). Taken together, maternal effects outweigh personal effects by a small amount in terms of their impact on body depth.
In the presence of parental care, adult males did not show a statistically significant interaction between gametic (biparental) and caring parent risk (Table 3). Parental care intensity has only a small effect on male body shape with greater levels of care being associated with shallower bodies and larger heads according to the first and second principal components (Table 3, Figure S8b). While caring parent risk did not have a statistically significant effect (Table 3), gametic high risk induced significantly deeper bodies (Figure 4b) with larger heads (Figure S10b).
3.3 Females
In adult female fathead minnows that were raised without parental care, we did not observe evidence for a three-way interaction, interactions involving personal risk, or an effect of personal risk, but found a statistically significant interaction between paternal and maternal risk (Table 1). Splitting the dataset revealed that maternal and paternal effects were similar in phenotypic impact (Table 4). Only the combination of high maternal and paternal risk caused females to develop deep bodies (Figure 3c).
In females that were raised in the presence of parental care, we found a statistically significant interaction between parental risk and caring parent risk while we did not observe the same for parental care intensity (Table 3, Figure S8c). Splitting the data reveals similar patterns as in juveniles as statistically significant phenotypic effects were generated only when both biparental and caring parent risk were high (Table 4). While this effect was observable mainly in the first principal component (Figure 4c) and not in the second (Figure S10c), we have to consider that the impact of this component on female body depth is comparatively marginal (Figure S6b).
4 DISCUSSION
As predicted, the relative importance of the different sources of environmental experience in inducing typical fish morphological defences differed across ontogeny. However, the observed patterns did not match all of our hypotheses. First, on average, maternal risk outweighed personal risk effects throughout assessed ages and sexes, which does not give much support to the “sampling-duration” hypothesis. Second, maternal effects were a main non-additive driver of inducible defences in juveniles and adult males whereas paternal effects were generally of lower relevance, which gives little support to the “shared-niche” hypothesis. Instead, this result strongly supports our ontogeny-independent “maternal-access” hypothesis, which predicts this effect to arise due to mothers having more mechanisms available to transmit environmental experience in the absence of parental care. The only exception was adult females, where maternal and paternal risk were of similar importance. Third, we found support for the “gametic-care-ontogenetic-switch” hypothesis only for the switch from juveniles (where gamete-mediated and care-mediated environmental experience were of similar importance) towards adult males, where gamete-mediated information dominated. Instead, in adult females, gametic and caring parent risk remained to be of similar importance. Here, caring parent risk was a better predictor of offspring phenotypes than parental care intensity, suggesting that environmental experience is not fully communicated through care intensity.
Our first observation that maternal environmental experience consistently outweighed personal environmental experience across ontogeny and sexes is in accordance with previous research on earlier developmental stages of the same species (Meuthen et al., 2021, 2023) and with a meta-analysis suggesting the same to be true for maternal effects in behaviour and physiology across taxa (Moore et al., 2019). At the same time, this result appears to be in opposition to the “sampling-duration” theory. However, we have to consider that while we sampled individuals at 180 days age, where they reach sexual maturity and are expected to show their second peak of plasticity (Meuthen et al., 2018, 2019), their parents deposited clutches at an age of between 192 and 349 days (median ± IQR, range; mothers: 277 ± 118, 195–349 days; fathers: 277 ± 122, 192–348 days). This means that parents had their environmental experience over a longer time period. If at the same time, the theoretically predicted degradation of parental information (Stamps & Bell, 2021) occurs at a very low rate, our results make sense even under the “sampling-duration” theory. We also have to note that empirical tests of this theory are challenging (Stamps & Bell, 2021), making it difficult to predict the exact point during which this switch occurs. Follow-up research should explore the possibility that parental information only degrades quickly when their environmental experience is not consistent from birth onwards.
Our second observation was that effects of gamete-transmitted maternal environmental experience outweighed paternal ones in juveniles and adult males whereas they were of similar importance for adult females. Under the “shared-niche” hypothesis, we would expect maternal effects to be the predominant force for juveniles and adult females whereas paternal effects should impact adult males to a greater extent. However, our findings of strong maternal effects in adult males, and some paternal effects in adult females do not follow the “shared-niche” hypothesis (Bell & Hellmann, 2019). Instead, it gives more support to the “maternal-access” hypothesis. Our finding is also consistent with the many empirical studies that found maternal effects to dominate across various taxa (Chang et al., 2021; Hellmann et al., 2020; Shama et al., 2014 but see Tariel et al., 2020). An explanation for the relevance of paternal effects in adult females would be that adult female body depth may also be indicative of sexual maturity (i.e. life-history plasticity), which may be more sensitive to paternal information as discussed in one of the following paragraphs. These results may also be confounded by the lack of parental care, whose importance is highlighted by our other results. Thus, future studies need to investigate the relative impact of maternal vs. paternal environments under the presence of parental care.
Third, both gamete-mediated and care-mediated environmental experience were equally important for the formation of morphological defences in juveniles, in accordance with previous P. promelas research (Meuthen et al., 2021, 2023). In adult males, gamete-mediated information emerged to be more relevant instead, which is in accordance with Hellmann et al. (2021), and supports to our “gametic-care-ontogenetic-switch” hypothesis. Interestingly, in adult females, gamete-mediated and care-mediated environmental experience was of similar importance as in juveniles, which suggests again that females may have responded with life-history plasticity, which has a long lag-time and thus has to be induced during the juvenile period as discussed below. While in our study, caring parents were sometimes genetic parents, which increases care levels (Meuthen et al., 2021), parental care intensity had only a relatively small impact on individual morphology (Tables 1 and 2, Figure S8). Thus, follow-up studies need to explore the mechanics of how the environmental experience of caring parents impact offspring beyond parental care intensity.
In adult female P. promelas, deep-bodied phenotypes can also be interpreted as accelerated sexual maturity as body depth is related to the presence of ova. As an earlier onset of egg production allows reproduction before being depredated, it is a common antipredator defence across taxa (Corbel & Carazo, 2022; Reznick & Endler, 1982). The long lag-times of such life-history plasticity (Auld et al., 2010) may explain the observed patterns of information prioritization. That is because if life-history plasticity is induced in early life, it makes sense that care-mediated environmental experience persisted into adulthood for females. Similarly, paternal effects were relevant for both juveniles and adult females, further supporting our theory. Follow-up research should aim to confirm whether deeper body shapes in adult female P. promelas indeed represent life-history plasticity by investigating their gonads.
Observed effect sizes and coefficients of determination were often small, in line with other studies that suggest small plasticity effect sizes in general (van Heerwaarden et al., 2016) and for fish morphological antipredator plasticity in specific (Frommen et al., 2011; Meuthen et al., 2018). Because predator preferences are driven by relative differences between prey items (Stige et al., 2019), even small phenotypic changes can still contribute to enhanced survival. Moreover, cue-induced TGP manifests not only in P. promelas morphology but also in behaviour (Meuthen et al., 2021, 2023). As behavioural and morphological defences can interact additively (DeWitt et al., 1999), the combination of altered traits is likely to impact survival to a greater extent than any of these changes in isolation (see also Meuthen et al., 2018).
AUTHOR CONTRIBUTIONS
DM, DPC and MCOF conceived the study and designed the experiments. DM conducted the experiment, collected the data, analysed it and wrote the manuscript. All authors improved the manuscript and agreed to the final content.
ACKNOWLEDGEMENTS
We are grateful to Adam Crane for helping with fish maintenance and to Markus Herberg Hovd for statistical advice. This research was funded by the Deutsche Forschungsgemeinschaft (DFG) (ME 4974/1-1, ME 4974/2-1), by a Bielefeld Young Researcher's Fund scholarship of Bielefeld University, by a Freigeist fellowship of the VolkswagenStiftung (both awarded to Denis Meuthen) and by Natural Sciences and Engineering Research Council of Canada (NSERC) grants to Douglas P. Chivers and Maud C. O. Ferrari. Open Access funding enabled and organized by Projekt DEAL.
CONFLICT OF INTEREST STATEMENT
The authors declare no competing interests.
STATEMENT OF INCLUSION
Our study brings together authors from a number of different countries, including scientists based in the country where the study was carried out. All authors were engaged early on with the research and study design to ensure that the diverse sets of perspectives they represent was considered from the onset. Whenever relevant, literature published by scientists from the region was cited; efforts were made to consider relevant work published in the local language.
Open Research
DATA AVAILABILITY STATEMENT
Data available from the Dryad Digital Repository: https://doi.org/10.5061/dryad.1jwstqk2w (Meuthen et al., 2024a). Code available from Zenodo https://doi.org/10.5281/zenodo.10849573 (Meuthen et al., 2024b).
