Long-term effects of widespread pharmaceutical pollution on trade-offs between behavioural, life-history and reproductive traits in fish

Abstract

In our rapidly changing world, understanding how species respond to shifting conditions is of paramount importance. Pharmaceutical pollutants are widespread in aquatic ecosystems globally, yet their impacts on animal behaviour, life-history and reproductive allocation remain poorly understood, especially in the context of intraspecific variation in ecologically important traits that facilitate species' adaptive capacities. We test whether a widespread pharmaceutical pollutant, fluoxetine (Prozac), disrupts the trade-off between individual-level (co)variation in behavioural, life-history and reproductive traits of freshwater fish. We exposed the progeny of wild-caught guppies (Poecilia reticulata) to three field-relevant levels of fluoxetine (mean measured concentrations: 0, 31.5 and 316 ng/L) for 5 years, across multiple generations. We used 12 independent laboratory populations and repeatedly quantified activity and risk-taking behaviour of male guppies, capturing both mean behaviours and variation within and between individuals across exposure treatments. We also measured key life-history traits (body condition, coloration and gonopodium size) and assessed post-copulatory sperm traits (sperm vitality, number and velocity) that are known to be under strong sexual selection in polyandrous species. Intraspecific (co)variation of these traits was analysed using a comprehensive, multivariate statistical approach. Fluoxetine had a dose-specific (mean) effect on the life-history and sperm trait of guppies: low pollutant exposure altered male body condition and increased gonopodium size, but reduced sperm velocity. At the individual level, fluoxetine reduced the behavioural plasticity of guppies by eroding their within-individual variation in both activity and risk-taking behaviour. Fluoxetine also altered between-individual correlations in pace-of-life syndrome traits: it triggered the emergence of correlations between behavioural and life-history traits (e.g. activity and body condition) and between life-history and sperm traits (e.g. gonopodium size and sperm vitality), but collapsed other between-individual correlations (e.g. activity and gonopodium size). Our results reveal that chronic exposure to global pollutants can affect phenotypic traits at both population and individual levels, and even alter individual-level correlations among such traits in a dose-specific manner. We discuss the need to integrate individual-level analyses and test behaviour in association with life-history and reproductive traits to fully understand how animals respond to human-induced environmental change.

Keywords: Bayesian statistics; animal personality; antidepressants; behavioural syndrome; chemical pollution; ecotoxicology; ejaculate traits; multivariate regression analysis.

FIGURE 1

Schematic of the exposure protocol and experimental design. (a) Adult guppies with an equal sex ratio were collected from the wild and exposed to control, low and high fluoxetine concentrations across 12 replicated mesocosm populations (180 × 60 × 60 cm) for five consecutive years. (b) Following exposure, males from each treatment (four independent mesocosms per treatment) were assayed for their behavioural, life‐history and sperm traits.

FIGURE 2

Mean‐level effect of fluoxetine on male (a) body condition (scaled mass index, in g), (b) gonopodium length (mm) and (c) sperm velocity (VCL, μm/s). The letters denote significant differences between treatments. In each plot, filled‐black circles represent the mean estimates, vertical error bars denote 89% credible intervals and coloured‐dotted values represent the probability densities of conditional effects.

FIGURE 3

Between‐individual variance in (a) activity (distance moved, in centimetre) and (b) refuge use (in seconds) and within‐individual variance in (c) activity (distance moved, in centimetre) and (d) refuge use (in seconds) of males from the three exposure treatments (control, low fluoxetine and high fluoxetine). In each plot, filled‐black circles represent the mean–variance estimates, vertical error bars denote 89% credible intervals and coloured‐dotted values represent probability densities.

FIGURE 4

Between‐individual variance in (a) body condition (scaled mass index) and (b) sperm count of males from the three exposure treatments (control, low fluoxetine and high fluoxetine). In each plot, filled‐black circles represent the mean–variance estimates, vertical error bars denote 89% credible intervals and coloured‐dotted values represent probability densities.

FIGURE 5

Between‐individual correlation in behaviour, life‐history, and sperm traits (i.e. pace‐of‐life syndrome) of males. In each plot, filled circles with horizontal error bars denote correlational values 89% credible intervals, and the colours represent three exposure treatments (control, low fluoxetine and high fluoxetine). The black dashed line indicates no correlation effects, while the values on the right side of the line show a positive relationship, and the ones on the left side indicate a negative relationship between traits.