Evolutionary consequences of pesticide exposure include transgenerational plasticity and potential terminal investment transgenerational effects
Castaño-Sanz, Verónica; Gomez-Mestre, Ivan; Garcia-Gonzalez, Francisco (2022), Evolutionary consequences of pesticide exposure include transgenerational plasticity and potential terminal investment transgenerational effects, Dryad, Dataset, https://doi.org/10.5061/dryad.tdz08kq2s
Transgenerational plasticity, the influence of the environment experienced by parents on the phenotype and fitness of subsequent generations, is being increasingly recognised. Human-altered environments, such as those resulting from the increasing use of pesticides, may be major drivers of such cross-generational influences, which in turn may have profound evolutionary and ecological repercussions. Most of these consequences are, however, unknown. Whether transgenerational plasticity elicited by pesticide exposure is common, and the consequences of its potential carry-over effects on fitness and population dynamics remains to be determined. Here we investigate whether exposure of parents to a common pesticide elicits intra-, inter- and transgenerational responses (in F0, F1 and F2 generations) in life-history (fecundity, longevity, and lifetime reproductive success-LRS-), in an insect model system, the seed beetle Callosobruchus maculatus. We also assessed sex-specificity of the effects. We found sex-specific and hormetic intergenerational and transgenerational effects on longevity and lifetime reproductive success, manifested both in the form of maternal and paternal effects. In addition, the transgenerational effects via mothers detected in this study are consistent with a new concept: terminal investment transgenerational effects. Such effects could underlie cross-generational responses to environmental perturbation. Our results indicate that pesticide exposure leads to unanticipated effects on population dynamics and have far-reaching ecological and evolutionary implications.
We individually exposed male and female beetles from the stock to either water (controls) or sub-lethal concentrations of pesticide. We applied the experimental treatments only to the parental generation, and recorded longevity, fecundity, lifetime reproductive success and body size (assessed as dry body weight; see below) of individuals up to the second generation (F2), after mating the focal parent to standardized mates (see below). Sample size doubled with each subsequent generation as we randomly selected one son and one daughter from each pair of parents to start each subsequent generation to test for sex-related differences in transgenerational effects (Fig. 1). The total number of individuals scored was expected to be 1050 (75 F0 mothers, 75 F0 fathers, 150 F1 daughters, 150 F1 sons, 300 F2 granddaughters, 300 F2 grandsons), but due to some individuals escaping and accidental mishaps, the final sample sizes were slightly lower and varied for the different traits (sample size for each analysis is indicated in the tables in the Results section).
We placed inoculated beans individually in Eppendorf tubes to ensure virginity of the emerging adult beetles and to keep track of individual age. Bean isolation was done 11 days after female oviposition, to minimize potential effects of mechanical manipulation on freshly laid eggs or early instar larvae inside the beans. The age of individuals at the time of exposure to the experimental treatments (toxicant/control) ranged between 1 and 4 days. We exposed each individual beetle to either one of two non-lethal concentrations of the pesticide (1 g/L and 2 g/L COMBO Deltamethrin, 2.5% w/v, Sarabia) or to carbon filtered dechlorinated tap water (control group) inside a flow cabinet at room temperature for 24 h. We previously conducted pilot tests to identify non-lethal concentrations of the pesticide for this species (see Figure S1). We aimed to expose parental beetles to sub-lethal concentrations so as not to confound potential transgenerational effects (influences on offspring's phenotype via non-genetic inheritance) with genetic changes due to differential survival, i.e. selection. Each beetle was randomly allocated to a single treatment (control, low pesticide, high pesticide). We exposed 25 F0 individuals from each sex to each treatment, for a total of 150 initial individuals (25 individuals x 2 sexes x 3 treatments; Fig. 1).
To expose each parental beetle to its corresponding treatment, we impregnated the ends of cotton swabs with 30 µL of either pesticide or water and placed these in 2 mL Eppendorf tubes that we kept open. We then introduced the target beetle in another tube with the bottom cut out and fit this tube into the one containing the impregnated swab, placing a mesh in between both tubes. This setup allowed exposure of beetles to airborne pesticide but prevented direct contact with it. After the exposure phase, focal beetles were individually placed into 26 mL plastic vials with pinholes in the cap to allow airflow, and ad libitum beans to be used as oviposition substrate. Each individual shared the vial with a tester non-exposed mate (male or female). These tester individuals were sourced from a standardized heterozygote tester line that was generated by crossing two near-isogenic lines generated after 17 generations following a brother-sister mating protocol. The use of these genetically homogeneous tester individuals minimizes sampling variance arising from random sampling of mates (see for rationale and application Garcia-Gonzalez and Evans 2011; Garcia-Gonzalez and Dowling 2015; Travers et al. 2015), and also ensures that transgenerational effects would be only minimally obscured by genetic or non-genetic variation introduced by individuals used as mates (Garcia-Gonzalez and Dowling 2015). We allowed beetles to mate and lay eggs in the vial for 48 h. Afterwards, we extracted males from the vials and individually kept them in Eppendorf tubes, to track their longevity. We then transferred mated females (focal females mated to tester males, or tester females mated to focal males) to a second 26 mL container with ad libitum beans where they laid eggs for the remainder of their lifespan. We counted the lifetime number of eggs (fecundity) and number of adult offspring (Lifetime Reproductive Success-LRS) produced by each female. LRS is a good estimate of fitness, as it measures the lifetime production of sexually mature progeny.
We did not expose the two subsequent generations (F1 and F2) to the pesticide, but kept housing and mating conditions identical to those of the parental generation. To obtain virgin F1 individuals we waited until day 11 after oviposition by parental females (see above) and individually isolated beans from the mating vials in Eppendorf tubes, as described above. We followed the same procedure in the F2 generation.
We tracked longevity by checking the survival of isolated individuals daily. We determined fecundity by carefully inspecting all beans and counting the total number of eggs laid by each female, both in the mating vial and the subsequent oviposition vial (only for F0 and F1 generations). We measured LRS as number of adults produced by each female during her lifetime. To obtain these data, we froze vials on day 28 since the last possible day of oviposition (in the second oviposition vial, the day the female was found dead), and later counted the number of adult individuals present in the vial. This procedure ensures that our LRS measurement excludes individuals from a subsequent generation (i.e., produced by crossings between siblings), because we take into account female egg laying span and the timing of egg-to-adult transition. Reproducing females only live a few days (15 days maximum; mean longevity ± SE = 8.38 ± 0.069; 99% of females lived less than 13 days) and their fecundity is maximal the first few days after mating (Zajitschek et al. 2018), hardly laying any eggs after the first week post-mating (unless they are housed with males and remate during this period). Moreover, new adults take ca. 23 days to emerge from newly laid eggs under our experimental conditions and virtually all adults emerge from the beans within 28 days (Rodriguez-Exposito 2018; Rodriguez-Exposito and Garcia-Gonzalez, unpublished). We therefore preserved the vials after all adults have emerged, but before any of their subsequent offspring could have emerged. Lifetime reproductive success has two components. First, an intragenerational component, because a female’s total number of adult offspring produced over its lifespan is largely determined by fecundity and effects of mothers on egg-to-adult viability (including maternal effects such as egg provisioning). Second, an intergenerational component, because egg-to-adult viability is expected to be largely dependent on properties of the individual offspring. This consideration also affects the interpretation of changes in LRS in the F1 and F2 generations discussed below, and generally, it implies that our conclusions on the extent of cross-generational influences of pesticide exposure are conservative.
We measured the dry weight of focal beetles with a Sartorius Cubis MSA6.6S microbalance (readability 0.001 mg, Sartorius, Goettingen, Germany). Beetles were previously frozen at -20 ºC (after death, in order to be able to track longevity), and afterwards thawed and dried to constant weight on an oven at 40 ºC for ~1 week, hence removing variation in body weight due to variation in time elapsed between death and freezing.
We used linear models (LMs; F0 and F1 generations) and general linear mixed models (GLMMs; F2 generation) to test whether experimental treatments applied to parental individuals affected not only their phenotype and life-history traits, but also those of subsequent generations. Our response variables were longevity, fecundity and LRS, all of which were normally or approximately normally distributed. Models included the experimental treatment as a fixed factor. Longevity was included as covariate in models testing effects on fecundity and LRS. Body size of the focal individual was included as covariate for longevity of those focal individuals, whereas body size of the female (focal or tester) was included as covariate for fecundity and LRS. Individuals' age at mating (1-4 days old) was also entered as a covariate. Longevity was also analysed by means of survival analysis using Cox models, and Cox mixed models for the F2 generation. Results from these proportional hazards models are consistent with those from the LMs and the GLMMs (Table S4).
To ease the interpretation of results and also because in this species males and females have distinct life-histories (Hallsson and Björklund 2012), we conducted the analyses separately according to sex of the focal individuals and by the sex of the parent/grandparent (for F1 and F2, respectively) initially exposed to treatment. We fitted LMs to the data from the F0 and F1 generations using the function “lm” in R (R Core Team, 2016). Since we only took one male and one female per cross in each generation, and because we conducted the analyses separately by sex, F1 data analyses did not require inclusion of a random effect to correct for parental ID. For the F2 generation, we fitted GLMMs using the “lmer” function in the “lme4” package (Bates et al. 2014) including grandparent ID as a random factor. P-values were calculated on maximum likelihood models, while REML models were run to obtain parameter estimates. The package pbkrtest (Halekoh and Højsgaard 2014) was used to calculate the degrees of freedom for the estimates in the case of GLMMs. A few instances of F0 LRS data <10 were excluded from the analyses a priori because such low fertility data are likely outliers due to intrinsic individual health problems or to infertility problems (Rodriguez-Exposito 2018). The few cases of LRS < 10 were distributed across groups, including the controls, which supports the notion that these cases are unrelated to the treatment (number of cases in the female-treated dataset: 2 in the control group, 1 in the low group, 1 in the high group; number of cases in the male-treated dataset: 1 in the control group, 2 in the low group). Due to the large number of P-values estimated throughout our study, we controlled the false discovery rate using the p.adjust function in R applying the Benjamini-Hochberg method (Benjamini and Hochberg 1995). We corrected P values within each generation and show adjusted P values in the Results' section and the tables.
We also calculated cross-generational multiplicative fitness for females and males exposed to treatment and we did so both via daughters and via sons (Fig. S2). In these calculations a 1:1 sex ratio was assumed. Multiplicative fitness via daughters for each F0 individual was calculated by multiplying its LRS (halved, as half of the number of adults produced would be daughters, assuming equal sex ratio) by the LRS of his/her daughters (F1), and by the LRS of his/her grandoffspring (F2) produced via daughters. We followed a similar procedure to calculate multiplicative fitness via sons. Due to the multiplicative nature of these calculations, sample sizes in these analyses are reduced compared to the analyses above because missing values for any family in any given generation implies a missing value for the net fitness of a particular lineage across the three generations (see final sample sizes for the lineages in the Results section). As before, F0 LRS data < 10 were excluded a priori from the analyses. Effect sizes (Cohen's d, that is, the mean difference divided by the pooled standard deviation) and their associated 95% CIs for the difference in multiplicative fitness between groups were calculated using the package compute.es (Re 2013). Cohen's d for intragenerational effects of the pesticide on female fitness (see below) was calculated from raw data using compute.es, but also using marginal means from the model, with the package emmeans (Lenth et al. 2018).
For simplicity, we do not state or discuss specific influences of covariables on the response variables. Their effects are nonetheless reflected on the tables. Notably, transgenerational effects were detected even after controlling for these influences. Graphical representation of data for response variables that were not affected by treatment can be found in Figures S3-S7 in the Supplementary Information (SI). Tukey post-hoc tests for all models can be found in Table S1 in the SI.
txt reader: .txt
European Regional Development Fund, Award: PID2019-105547GB-I00/AEI/10.13039/501100011033
Ministerio de Ciencia e Innovación, Award: CGL2016-76173-P
Ministerio de Ciencia e Innovación, Award: FPU17/02623