The two main questions of the present study are: Is an ecologically relevant exposure to EE2 toxic to embryos and larvae of a population of river-spawning grayling, and is there additive genetic variance for the tolerance to EE2 in this population, i.e. does the population currently have a potential to rapidly adapt to this type of pollution? The first question is relevant even if the toxicity of EE2 has been demonstrated in many other fish taxa [e.g. 30, 31-33], because the chemical pollution of freshwater habitats that has happened since the market launch of the contraceptive pill, i.e. during more than 5 decades, could have led to adaptation and hence to reduced toxicity in some fish. The answer to the second question may help us to better understand if pollution by EE2 has induced rapid evolution because, in our study population, the period of exposure is likely to span on average more than 10 or even 15 generations, i.e. there could have been enough time for evolution to deplete any genetic variance for tolerance to EE2 that the population could have had at the beginning of the exposure. Moreover, both questions are of ecotoxicological relevance [10, 34, 35] because standard ecotoxicological testing typically ignores potential taxon-specific toxicities .
Regarding our first main question: Our study adds the grayling to the list of salmonids whose embryos and larvae are susceptible to ecologically relevant concentrations of EE2, like whitefish , Atlantic salmon (Salmo salar) [37, 38], rainbow trout (Oncorhynchus mykiss) , and brown trout . With the present study, at least one species of each subfamily of the Salmonidae (Coregoninae, Salmoninae, and Thymallinae) has now been tested on the same ecologically relevant concentration of EE2, applied in a one-dose exposure of 2 pg to embryos developing in 2 mL wells [11, 17], and at least two further species (the lake char Salvelinus umbla from Lake Geneva and the whitefish C. suidteri from Lake Hallwil, Switzerland) will soon be added to this list (Garaud et al., experiments ongoing). While EE2 induced significant mortality in whitefish and brown trout [11, 17], no significant effects on mortality could be found in grayling. The EE2-induced increase in embryos mortality was around 3 percent points (pp) in C. palaea and around 13 pp in C. albellus . The analogous increase in mortality was 0.9 pp in brown trout and only significantly different from zero because of an extra-ordinary large sample size (N = 7,302 singly embryos) . In the present study, the analogous and statistically non-significant increase in mortality was 1.5 pp for embryos and 0.4 pp for larvae, i.e. the effect sizes seem comparable to the ones observed in brown trout.
While EE2- and sham-exposed grayling embryos hatched at similar size, exposure to EE2 reduced larval growth and consumption of yolk sac after hatching by about 4% each during the first 8 days after hatching. We therefore conclude that EE2 is toxic to grayling at early developmental stages. Such a reduction in growth was predicted from recent analyses of physiological reactions to EE2 in Atlantic salmon [37, 38] but was not observed in brown trout . One possible explanation for this apparent discrepancy between brown trout and grayling larvae is that hatching was not induced in the study on brown trout  but induced by an increase in temperature in the present study on grayling. Under the given conditions, EE2-exposed brown trout embryos hatched later and at smaller size than sham-exposed ones, while, in the present study on grayling, no treatment-related difference in the timing of hatching nor on hatchling size could be observed. If growth rate after hatching is dependent on larval size and developmental stage, such differences in the experimental protocols could be responsible for the apparent differences in treatment effects on growth rates. However, in both cases, the combined effects of EE2 on embryo and larval development would be expected to delay the emergence from gravel at the end of the yolk sac stage and could even lead to smaller body sizes at emergence. Time to emergence, and body size at emergence, is likely to be linked to fitness in salmonids because larvae that emerge earlier and larger than others may face less competition for resources (e.g. feeding territory) and are more prone to outcompete their late emerging counterparts [40, 41].
In a parallel study on another sample taken from 5 of the 40 sibgroups we studied here, Selmoni et al.  found strong effects of exposure to EE2 on gene expression in embryos, hatchlings, and juveniles around the end of the yolk-sac stage. Their analyses confirmed that EE2 disturbs normal physiological processes at critical developmental stages. Therefore, single-factor laboratory studies like ours are likely to underestimate the ecotoxicological relevance of EE2 in the wild. It is possible that other types of environmental stressors such as microbes , temperature variations, or other micropollutants  interact with the effects of EE2 and thereby amplify its toxicity [42, 43]. If such interaction effects exist, doses even lower than 2 pg per embryo could induce higher toxicity in the wild than what we observed in the laboratory. Interestingly, the effects of EE2 on gene expression that Selmoni et al.  observed were strongly sex-specific (i.e. dependent on sex genotype, gonad formation starts at later stages ). At hatching, for example, over 20,000 genes were differentially expressed in genetic females while hardly any effects of EE2 on gene expression could be observed in freshly hatched genetic males. Sex-specific effects can eventually lead to sex-specific survival and hence skew population sex ratios in the wild. This could lead to detrimental consequences for population fitness . Indeed, the study population suffers from a skewed sex ratio (far more males than females, ) that seems not be due to environmental sex reversal [25, 26] but rather be caused by sex-specific mortality . This sex-specific mortality still needs to be explained.
As far as we know, there exist no measurements of estrogenic pollution around the spawning ground of our study population. However, this spawning ground is located in the river Aare within a city of more than 40,000 inhabitants, a large sewage treatment plant about 4 km downstream, and several nearby villages (with several thousand inhabitants each) upstream. Marques da Cunha et al.  tested whether variation in estrogenic pollution creates population differences in toxicity of EE2, and genetic variation for tolerance to EE2, in brown trout. They sampled 7 different streams that varied in their ecology and in the level of estrogenic pollution and found population differences in various embryo and larval traits, but none in the reaction to EE2. They argued that very low concentrations on EE2 can already cause selection and hence induce rapid evolution. The hypothesis is supported by the observation that the 2 pg EE2 in the aqueous exposure seemed continuously taken up by the embryo (about 80% within 4 weeks) while the concentration remained constant in empty plates . This suggests that salmonid eggs take up EE2 at concentrations that are far lower than the 1 ng/L that are sometimes even found in groundwater . On the other side of the scale: when Brazzola et al.  exposed whitefish embryos to 1 ng/L, 10 ng/L, or 100 ng/L EE2, increasing concentration seemed only weakly linked to increased toxicity. Similar observations were made by Duffy et al.  who exposed Atlantic salmon to 0.004 nM, 0.04 nM, and 0.4 nM EE2 that correspond to 1.2 ng/L, 11.9 ng/L, and 118.6 ng/L, respectively. We therefore argue that our one-dose aqueous exposure to 2 pg EE2 was ecologically relevant for grayling embryos and larvae.
Regarding our second main question: Because grayling males do not provide any parental care, significant sire effects on offspring traits reveal additive genetic variance in full-factorial breeding experiments . The dam effect then represents a combination of additive genetic variance and maternal environmental effects . In salmonids, maternal environmental effects comprises characteristics such as egg size  and compounds that females allocate to their eggs [e.g. 47, 48-51]. We found strong direct maternal effects on every offspring trait that we measured, and a dam x EE2 interaction on the timing of hatching. We conclude that maternal sib groups reacted differently to exposure to EE2. However, these maternal effects seem to be mainly due to maternal environmental effects, because we found no significant additive genetic variance for tolerance to EE2 pollution in any of the analysed traits.
No significant additive genetic variance could potentially be due to a type II error (false negative). However, such an error is unlikely here because (i) our analysis is based on a large sample size (1,555 singly-reared embryos) and 40 sib groups, (ii) our sample revealed overall additive genetic variance (i.e. significant sire effects) on embryo mortality and the timing of hatching, (iii) a parallel study on other samples of the same 40 families revealed genetic variation in the tolerance to infection by a bacterium , and (iv) singly-reared salmonid embryos are sensitive indicators of environmental stress, and studies based on comparable breeding designs have demonstrated additive genetic variance for the tolerance to other types of stressors, including other types of pollutants [52, 53], pathogens  or even water-borne cues linked to infection .
The finding of significant additive genetic variance for tolerance to EE2 pollution in grayling is in sharp contrast to the findings of Brazzola et al.  and Luca et al. (in preparation) on lake-spawning whitefish. However, our findings correspond well with the ones of Marques da Cunha et al.  who used a similar experimental protocol to test for this type of genetic variation in 7 genetically distinct populations of river-spawning brown trout and found none (in a total sample size of 7,302 singly embryos, i.e. a type II error was also unlikely in their case). Taken together, these observations support the view that the appearance of the novel stressor EE2 has induced evolution and thereby used up the corresponding additive genetic variance in river-spawning salmonid that are exposed to the pollutant, while lake-spawning salmonids who are less exposed still have a strong potential to evolve rapidly to EE2. However, alternative explanations are possible. Future studies could therefore compare exposed and non-exposed populations of the same species (if possible, given human population density and the finding that very low doses of EE2 can induce selection), add analogous tests on further river-or lake-spawning salmonids, or test for signatures of selection in the EE2 response pathways [56, 57].