Wild adult Physa acuta snails were collected in February 2017 in a lentic backwater of the Rhône River in Lyon, France (N 45°48′06″ E 04°55′33″). P. acuta is a globally invasive freshwater snail from North America (Lydeard et al., 2016). These snails constituted the F0 generation (see Figure 1 for number of individuals; see Appendix 1 for the schematic experimental design). In the laboratory, F0 snails were pooled in a 10 L plastic box filled with dechlorinated tap water (control water hereafter) and interbreed overnight. Then, they were isolated in 80 mL plastic boxes (these rearing boxes were used until the end of the experiment; 4.5 × 6 cm) filled with control water to ensure that a box contained the progeny of only one F0 snail (i.e., one F1 family per box). They laid eggs during 24 h, which was enough for most snails to lay an egg mass, and then were removed. F1 eggs developed until hatching at 25°C (ca. 7 days). After hatching and until the end of snail development, they were reared in a room at 25°C with a 12 h/12 h photoperiod. Snails were fed ad libitum with boiled and mixed lettuce. The water and food were renewed twice a week. Siblings of each F1 family developed together for 10 days in control water without being manipulated (P. acuta newborns are very small and easily damaged). Then, siblings of each F1 family were split into two developmental environments: six siblings remained in control (C) water and six siblings moved to predator-cue (P) water (Figure 1). F1 families with less than 12 siblings were discarded (see Figure 1 for number of families). After 7 days, F1 snails were isolated in the same environment (control or predator-cue water) and developed individually until 31 days old to reach a size large enough to be sexually mature (see Figure 1 for number of individuals). The predator-cue water was obtained by (1) mixing the rearing water of six Orconectes limosus crayfish (reared individually in 4 L box of dechlorinated tap water) and (2) “infusing” several smashed P. acuta adult snails (one snail for every 4 L) for 1 h before using the predator-cue water. These smashed snails were then used as crayfish food.

To generate the F2 generation, we performed three types of pairs with F1 snails (each partner in a pair generated a progeny as P. acuta is a hermaphrodite; Figure 1):

We paired partners from different families, and we were careful to associate the same families in the three types of pairs. To keep the partner identity during copulation, we painted shells (Henry and Jarne, 2007). We put the two partners into a rearing box for 24 h to copulate (P. acuta prefers outcross; Tsitrone et al., 2003). We then submitted snails of the F2 generation to the two treatments (control or predator-cue) following the same protocol as for the F1 generation (see Figure 1 for number of individuals). As F2 snails developed slower, they developed until 61 days old to reach a size large enough for their weight to be accurately measured.

Finally, our full factorial design resulted in four combinations of maternal and paternal environments (CxC, PxC, CxP, PxP) and in two F2 offspring developmental environments (C, P) (Figure 1).

We assessed two anti-predator behaviors: escape behavior by the time taken to crawl-out the water (sec) and activity by the total covered distance (cm). These two behaviors were assessed on two different subsamples of F2 snails (Figure 1). The sub-samples were composed of as many different F2 families as possible.

For each trial of escape behavior, the snail first acclimated in a chamber at the center of a behavioral arena (4.5 × 6 cm box with 2.3 cm water and 0.7 cm polystyrene at the bottom). A minute later, we opened the acclimation chamber and recorded time taken by the snail to crawl to the water surface with JWatcher (Blumstein and Daniel, 2007). After 5 min, we put back the snail in its rearing box for at least 1 h before the next trial and we changed arena water. For each snail, the escape behavior trial was repeated two times in each immediate environment (two times in control water first and two times in predator-cue water after).

For each trial of activity, the snail first acclimated 1 min in the entire behavioral arena (11.5 × 17.5 cm box with 0.6 cm water). We then recorded the snail position every 30 s on a grid (0.5 cm tiles) drawn at the bottom of the arena. After 5 min, we put back the snail in its rearing box for at least 1 h before the next trial and we changed arena water. The activity trial was repeated two times only in control water. As snails escape (crawl-out the water) in predator-cue water, the activity could not be assessed in this immediate environment.

After assessing behavior, we gently dried snails with a paper towel and measured total wet mass (body and shell) with an electronic scale at the nearest 0.0001 g.

We tested with two separate linear mixed models (LMMs) whether maternal, paternal and developmental environments influenced F2 snails’ escape behavior and activity. For the escape behavior, a LMM was used instead of a survival model as just a few trials were censored (censoring means that the snail did not crawl-out the water within the five min trial; 32 trials out of 640 trials). The time to crawl-out was transformed with a log10 transformation to achieve normality. We also multiplied by −1 the time to crawl-out to better reflect escape behavior, as a short time to crawl-out is associated with high escape behavior. The fixed effects of the two LMMs included maternal (E1m), paternal (E1p) and developmental (E2) environments and all interactions. For the LMM of escape behavior, we added the immediate (I) environment and all interactions. For all LMMs, we also added the scaled snail mass to control for size. We reduced the number of fixed effects using the model selection explained in Zuur et al. (2009). The random effect of the two LMMs included individual identity to control for intrinsic differences in behavior between individuals. We fitted LMMs with restricted maximum likelihood estimation and Kenward and Roger’s approximation for degrees of freedom using the lmer() function (Bates et al., 2015). Type 2 F-tests were used for significance of fixed effects with the package lmerTest (Kuznetsova et al., 2017). Post hoc comparisons were realized with the package emmeans (Lenth, 2019) to compare the “control” treatment (both parents not exposed: CxC) with the other treatments (only exposed mothers: PxC; only exposed fathers: CxP; and both parents exposed: PxP). These post hoc comparisons are reported in Appendix 2. All statistical analyses were performed with R 3.6.0 (R Development Core Team, 2019).

Maternal (E1m), paternal (E1p), developmental (E2) and immediate (I) environments all influenced escape behavior and their effects interacted with each other (significant E1m × E1p × E2 interaction and E1m × E1p × I interaction in Table 1 and Figure 2). In the immediate environment without predator cues, neither maternal nor paternal environments influenced escape behavior (Figure 2A and Appendix 2). In the immediate environment with predator cues, both maternal and paternal environments influenced escape behavior. The single maternal exposure to predator cues (PxC) resulted in a 49 s (61%) slower escape than unexposed parents (CxC; black solid arrows on Figure 2B) while the single paternal exposure (CxP) did not influence escape behavior (Figure 2B and Appendix 2). The combination of maternal and paternal exposures (PxP) resulted in a similar time to escape than unexposed parents (CxC) when offspring developed without predator cues (Figure 2Bi and Appendix 2). Conversely, the combination of maternal and paternal exposures (PxP) resulted in a 28 s (31%) slower escape than unexposed parents (CxC) when offspring developed with predator cues (Figure 2Bii and Appendix 2).

Differences in escape behavior among snails were consistent across the different behavioral trials (significant individual random effect in Table 1).

Maternal (E1m), paternal (E1p) and developmental (E2) environments all influenced activity and their effects interacted with each other (significant E1m × E1p × E2 interaction in Table 2 and Figure 3). However, post hoc comparisons did not reveal significant differences between the activity of offspring from unexposed parents (CxC) to the activity of offspring from exposed mothers, exposed fathers or both exposed parents (PxC, CxP, or PxP).

Differences in activity among snails were consistent across the different behavioral trials (significant individual random effect in Table 2).

The maternal, paternal, and developmental environments had an influence on escape behavior but only in the immediate context of predation risk. This means that snails integrated the cues from parental and developmental environments only in the light of the immediate environment (Figure 2B). The immediate environment being the most accurate information about the predation risk, the snails did not exhibit differences in escape behavior when the immediate environment was safe (without predator cues), even if they or their parents were exposed to predator cues. Indeed, the maternal exposure decreased the escape behavior of offspring only in the immediate predation environment whatever the offspring developmental environment (Figure 2B). In the immediate safe environment (Figure 2A), this negative maternal effect is likely masked by an overall absence of behavioral response. Indeed, the time to crawl-out in immediate safe environment was ca. 150 s for all snails (Figure 2A) while in the immediate predation environment (Figure 2B), only the snails from exposed mothers took 150 s to escape, the others snails escaping faster (ca. 90 s). This negative maternal effect is likely revealed only when offspring must express behavioral defenses against predators (faster escape in immediate predation environment).

For activity of offspring, we found consistently that maternal, paternal, and developmental environments interacted. However, we did not identify significant patterns resulting from this triple interaction. As for escape behavior, where we did not observe significant differences among snails in the immediate environment without predator cues, the immediate environment is likely the first to be integrated by offspring. In absence of immediate predator cues, offspring did not alter their activity according to parental and developmental environments while in the presence of predator cues, activity cannot be measured as all snails exhibited an escape behavior.

In our predator-prey system, we did not expect sex-specific TGP as P. acuta is a simultaneous hermaphrodite with an internal fertilization and does not provide any parental care. The mating partners therefore have the same morphology, behavior, ecology, and gamete dispersal; they just differ in the reproduction function used at the time of mating (female: eggs and mucus; male: sperm and seminal fluid). Hence, they experience the same predation pressures and may have the same phenotypic optima in response to the current predation risk. This suggests that (1) both sexes would exert similar effects on offspring traits and (2) offspring would similarly benefit to process the maternal or paternal cues (Kamel et al., 2010; Bell and Hellmann, 2019). Surprisingly, we found sex-specific TGP on escape behavior in the immediate context of predation risk, meaning that female and male functions have distinct effects on this trait depending on the immediate environment. For activity (only measured without predator cues), no distinct effects of maternal or paternal environments were found. Investigations of such sex-specific TGP in simultaneous hermaphrodites is restricted to plant species. However, in plants, huge differences in gamete and seed dispersals between male and female functions may promote sex-specific TGP (Galloway, 2001; but see Akkerman et al., 2016). In animals, evidence of sex-specific TGP often used species with sex-specific selection pressures or species providing uniparental care (e.g., Shama et al., 2014; Emborski and Mikheyev, 2019; Lehto and Tinghitella, 2020). Our results suggest that sex-specific TGP may be driven by other factors than sex-specific selection pressures and uniparental care.

Firstly, sex-specific TGP in P. acuta may arise from differences in inheritance mechanisms between female and male functions (Akkerman et al., 2016; Guillaume et al., 2016; Bell and Hellmann, 2019). Female function has a narrow relationship with the subsequent generation and can add informative molecules (e.g., hormones, non-coding RNA) or nutrients in eggs during the production of oocytes or during the offspring very early development (after fertilization and before laying). By contrast, male function can only influence offspring via sperm cells and seminal fluid and therefore has a limited number of molecular pathways to transmit information to the subsequent generation (e.g., epigenetic mechanisms). Although we do not know the specific inheritance mechanisms acting in our study, we can suppose that female and male functions transmit environmental cues via different bearers of information that would explain this sex-specific TGP for escape behavior. For example, in yellow monkeyflower (Mimulus guttatus), Akkerman et al. (2016) have shown that demethylation erased the effect of maternal but not paternal environment, suggesting that only maternal information is transmitted via DNA methylation.

Secondly, sex-specific TGP on escape behavior may result from stronger effects of stress on female function than on male function. The single maternal exposure to predator cues induced a decrease in escape behavior (offspring took longer to crawl-out the water) while the single paternal exposure to predator cues did not influence the escape behavior. These different effects between female and male parental functions on offspring behavior may result from differences in reproductive investment. As in numerous organisms, the female function is likely more costly than the male one in P. acuta, although it is difficult to evaluate it in a simultaneous hermaphrodite. Therefore, stressful environments (such as with predation risk) that negatively influence the snail’s state (DeWitt, 1998) are more likely to be passed on by the female function than by the male function. This hypothesis is corroborated by the decrease in escape behavior induced by the maternal exposure. Although in this study we did not test offspring survival in the face of a real predator, our results (Appendix 3) and other studies (e.g., DeWitt et al., 1999; Beaty et al., 2016) have shown that a faster escape behavior of snails exposed to predator-cues in their immediate environment is adaptive. Consequently, the decrease in snail escape behavior induced by maternal exposure may have deleterious fitness effects on offspring in response to a predator. The negative direction of the maternal effect therefore suggests that the maternal effect is a state-based effect rather than an anticipatory one (see also Coslovsky and Richner, 2011; McGhee et al., 2012). Finally, the combination of stressful environments and difference in reproductive investment between sexes may be sufficient to generate a state-based TGP that is sex-specific.

In the immediate predation environment (Figure 2B), the effect of paternal cues depended on whether they were integrated alone or in combination with maternal and developmental cues (see also Galloway, 2001). This results in a complex interaction between paternal, maternal and offspring environments that we can dissect as follows: (1) The single paternal exposure never affected the offspring escape behavior whatever the offspring developmental environment; (2) When both mother and father were exposed, paternal cues were integrated by offspring according to their own developmental environment. In a developmental environment without predator cues (Figure 2Bi), the paternal exposure mitigated the reduction in escape behavior due to the maternal exposure. Consequently, offspring from both exposed parents escaped as fast as offspring from unexposed parents. This mitigation suggests that offspring can process the cues from exposed fathers to buffer the maternal state-based effect, which we can interpret as an information-based TGP of the paternal cues. Similar results have been observed in sticklebacks (Gasterosteus aculeatus) where the negative effect of paternal exposure on offspring survival in the face of a real predator was mitigated by maternal exposure (Hellmann et al., 2019). In this species, as males are the sole providers of paternal care, a paternal exposure to predators leads to negative stress-mediated effects. In a developmental environment with predator cues (Figure 2Bii), the paternal exposure combined with the maternal one did not compensate the negative maternal effect resulting in a slow behavioral response of snails, similar as observed from exposed mother alone. The exposure to predator cues during offspring development might influence the offspring state (DeWitt, 1998) that might be then no longer able to process the paternal cues to mitigate the negative effect of their exposed mothers.

The non-additive patterns we observed here do not fit with the responses already observed in the literature (threshold hypothesis: Stein et al., 2018; Bell and Hellmann, 2019; Lehto and Tinghitella, 2019, 2020; back-up hypothesis: Bell and Hellmann, 2019; Lehto and Tinghitella, 2019; bayesien hypothesis: Stamps and Krishnan, 2014). We showed that the non-additive responses of offspring for escape behavior might result from a hierarchical integration of cues (i.e., information-based) constrained by state-based effects. Immediate and paternal exposures to predator cues might led to information-based effects on the escape behavior of offspring while developmental and maternal effects might have negative state-based effects. Hence, offspring might first integrate the cues from their immediate environment (the most accurate predictor of predation risk) before to integrate then the cues from their fathers (mitigation of the negative maternal effect) if their state is not constrained by their own developmental environment (without predator cues).