Rapid evolution of diet choice in an introduced population of Trinidadian guppies

Abstract

Eco-evolutionary theory has brought an interest in the rapid evolution of functional traits. Among them, diet is an important determinant of ecosystem structure, affecting food web dynamics and nutrient cycling. However, it is largely unknown whether diet, or diet preference, has a hereditary basis and can evolve on contemporary timescales. Here, we study the diet preferences of Trinidadian guppies Poecilia reticulata collected from directly below an introduction site of fish transplanted from a high-predation environment into a low predation site where their densities and competition increased. Behavioural assays on F2 common garden descendants of the ancestral and derived populations showed that diet preference has rapidly evolved in the introduced population in only 12 years (approx. 36 generations). Specifically, we show that the preference for high-quality food generally found in high-predation guppies is lost in the newly derived low-predation population, who show an inertia toward the first encountered food. This result is predicted by theory stating that organisms should evolve less selective diets under higher competition. Demonstrating that diet preference can show rapid and adaptive evolution is important to our understanding of eco-evolutionary feedbacks and the role of evolution in ecosystem dynamics.

1. Introduction

Eco-evolutionary dynamics occur when organismal traits evolve in response to their environment and this in turn feeds back to affect their environment [1,2]. The study of feedback loops between ecology and evolution has received much attention in the past decade [3]. Of particular interest has been to understand how evolution can change the ecosystem function of organisms over time to drive such feedbacks [4,5]. Empirical and theoretical studies have shown how changes in trophic traits can have important effects on ecosystem structure and processes. For example, the feeding specialization of threespine sticklebacks Gasterosteus aculeatus (benthic versus limnetic) affects invertebrate community structure and dissolved organic carbon dynamics [6], while divergence in the degree of omnivory of guppies Poecilia reticulata affects primary production and nutrient cycling [7,8]. Another study showed that primary productivity in lakes is drastically reduced by Daphnia pulicaria's evolution of cyanobacterial toxin tolerance (and consumption) [9].

Functional traits related to trophic interactions, such as prey palatability or consumer diet preference, are important determinants of eco-evolutionary dynamics. For example, theoretical and empirical studies show that the trade-off between resource uptake and palatability drives the nature of predator–prey cycles [10,11]. Also, divergence in the diet of organisms can change ecosystems' structure and processes through population regulation, trophic cascades [6,7] or nutrient cycling [8]. Despite the central importance of diet divergence in ecosystem processes and the importance of rapid evolution for eco-evolutionary feedbacks [2,3,12], little is known regarding the ability of organisms to rapidly evolve their diet preferences. Partly, this is because there is a general lack of understanding of the genetic basis of diet and diet preference [13].

Theory predicts that under lower resource availability there should be lower diet selectivity, less specialization and higher diet breadth [14]. Studies in guppies (P. reticulata, Peters) have shown that the diets of fish originating from populations differing in predation, density and resource availability differ in this predicted direction [15]. High-predation guppies, which live at lower densities and, subsequently, have higher per capita resources and less competition, have diets that are selective of high-quality food, namely invertebrates, that are richer in nitrogen. Conversely, low-predation guppies are less selective and feed more on lower-quality algae and detritus [15]. It is unknown, however, whether these differences have a genetic basis or are largely plastic.

In this study, we investigate the occurrence of diet preference evolution in a population of Trinidadian guppies that descend from fish that was experimentally translocated from a high-predation and high-resource environment to a low-predation environment where it achieved high densities. Specifically, we were interested in knowing whether (1) there had been heritable changes in food preference, and (2) this evolution occurred in the direction expected by theory: from higher to lower selectivity for high-quality food. We did so by comparing in laboratory assays the food preference of second generation common garden descendants of fish collected from the ancestral and derived populations.

2. Methods

In March 2008, as part of a larger study on eco-evolutionary feedbacks, 38 male and 38 female guppies from a high-predation site on the Guanapo River were introduced to the Lower La Laja, a guppy-free low-predation tributary [16,17]. The introduction site on the Lower La Laja is a 110 m reach bound by waterfalls that impede the movement of guppies upstream but not downstream. These fish then naturally moved downstream below the barrier waterfall to colonize further portions of the tributary, which did not previously have guppies. In March 2020, 12 years after the initial introduction, we collected 30 guppies from the Guanapo River high-predation ancestral population (GHP) as well as 30 from the lower regions of the Lower La Laja (LOL). These guppies were brought to the laboratory at Washington University in St. Louis and bred for two generations in common garden conditions. We kept all fish in flow-through tanks under the same conditions, with densities averaging 3 individuals per litre. Water temperature was set between 25°C and 28°C, conductivity between 750 and 850 µS, and pH between 7.5 and 8.5. Prior to our trials, fish were fed a combination of ad libitum flake food and quantified brine shrimp.

At a size ranging from 16.38 to 22.45 mm, we assigned 40 adult females from GHP and 40 from LOL to one of two food treatments: high quality (HQ) and low quality (LQ). We included food quality as a treatment because we expected habituation to food to have an effect in their subsequent food preference through learning. In order to make the high-quality food, we made a paste mix of 95% chicken liver and 5% fish oil by weight. To make the low-quality food we mixed 80% liver paste and 20% fish oil. Once mixed, we supplemented the resulting paste with baby food (Beech Nut) to ensure the presence of all essential nutrients. The final mix consisted of 1/3 baby food and 2/3 liver–oil mix by weight. The carbon to nitrogen (CN) ratio of the high- and low-quality food was measured in a CHNS analyser and determined to be an average of 8.7 and 17.5, respectively. These values are comparable with previously reported CN ratios of invertebrates (high-quality food) and algae (low-quality food) in Trinidadian streams [15,18]. Both food types were similar in appearance and consistency, and only differed in nutrient content. We fed each individual 10 μL of paste food twice a day and kept the animals under their different treatments for 12 weeks. This is on par with previous studies showing physiological and stoichiometric responses of guppies to diet quality [19,20]. In total, 20 individuals from each population were assigned to each food treatment.

We tested individual females for food preference at the end of the 12 weeks of treatment. For the trials, we placed a single female in a 22 L glass tank. After a period of 3.5 h of acclimatization, we presented 5 µL of high- and low-quality food to the fish on opposite sides of the tank so that they would be clearly separate when eating. The female was given 5 min to start to eat before the trial was considered invalid. We recorded behaviour for 5 min after the first bite. We recorded the latency to the first peck at the food, which food type was first pecked, and how long the guppy spent pecking at each food type. The latter was defined as the length of time the fish was actively pecking on the food and moving its mouth. On the day of the trial, individuals would eat in the morning (at approx. 9 am), but not in the afternoon. We carried out trials at approximately 4 pm. Guppies would be put in their experimental tanks at around 12:30 pm on the day of the trial, which gave a 3.5 h acclimation period before the trial. To avoid unwanted biases, we made all observations blindly, with the observer not knowing either the population or treatment the individual belonged to. Due to some mortality and failed trials where the female did not eat, the final sample sizes were: 14 GHP–HQ, 8 GHP–LQ, 7 LOL–HQ and 9 LOL–LQ.

We performed preliminary analyses to evaluate whether there were any differences in the size and motivation of individuals of different populations and food treatments, as well as the total time spent foraging. We did so using linear mixed models with standard length, latency and total time eating, respectively, as response variables, and population, food treatment and their interaction as explanatory factors. We included date of the trial as a random effect.

To evaluate treatment and population differences in food preference we analysed, using linear mixed models, the time spent feeding on the high-quality food as a function of the total time spent feeding, population, food treatment, initial food approached, and the two-way interactions between population, food treatment and initial food approached. The reason to include initial food approached is to account for the inertia of continuing eating the first food encountered. An effect of first food encountered would imply a low preference. We also included standard length and latency as covariates to control for individual differences in size and motivation. The rationale is that hungrier individuals will show shorter latencies and this may affect their food preference. Finally, we added date of trial as a random effect, to account for any date-related similarity in the behaviour of animals. We removed interaction terms when not significant. We performed all analyses using R package lme4.

3. Results

Total feeding time averaged 28.06 s out of the 5 min that a trial lasted. There were no differences in standard length, latency to eat, nor time feeding across treatments and populations (see electronic supplementary material).

With regard to food preference, there was a significant interaction between population and first food eaten in the trial, but no effect of food treatment (table 1). The ancestral Guanapo HP population displayed similar times eating high-quality food regardless of the first food encountered, while the derived Lower La Laja population showed an inertia to eat more low-quality food when that was the first food encountered (table 1; figure 1). In other words, when the first food encountered was of high quality, both populations spent similar amounts of time on high-quality food, but when the first food encountered was of low quality, Lower La Laja spent on average 21.7 s less feeding on high-quality food than Guanapo HP fish (table 1) population.

Table 1.

Linear mixed models for the time spent on high-quality food.

random effects	intercept s.d.	residual s.d.
date	5.627	12.159
fixed effects	estimate	s.e.	T22	p-value
(intercept)	−7.100	49.285	−0.144	0.887
total feeding time	0.856	0.105	8.118	<0.001
standard length	0.087	2.326	0.037	0.970
latency	0.002	0.035	0.068	0.946
first food (HQ)	−3.919	6.092	−0.643	0.527
food rearing (HQ)	−2.154	4.256	−0.506	0.618
population (LOL)	−15.805	7.283	−2.170	0.041
first food (HQ) × population (LOL)	21.704	9.772	2.221	0.037

Figure 1.

Estimated marginal means of the time guppies spent eating high-quality food in the behavioural trials by population and first food encountered. Error bars represent standard errors. Raw data are presented as gray dots. For ease of visualization, the y-axis is log transformed.

4. Discussion

Our results indicate that the ancestral high-predation guppies show consistent food preferences regardless of the food that they encounter first or the food they were reared with, spending more time on high-quality food than the derived low-predation population. By contrast, the feeding behaviour of the derived population is dependent on the first food approached, spending more time on the food first encountered regardless of quality, and therefore showing little evidence of a preference. More generally, our study suggests that feeding preferences in guppies can have a genetic basis and shows that they can rapidly evolve in response to environmental change. Given that diet is a key determinant of an organism's ecosystem function [21], if generalizable this result has important consequences for the study of eco-evolutionary feedbacks.

The direction of change shown in our study is consistent with observed diet patterns in the wild, such that guppies from high-predation populations typically have diets that are richer in nutrients [15] and include more invertebrates, which are richer in nutrients (and lower in CN ratios) than algae. Low-predation guppies, by contrast, have less selective diets that include more algae and detritus [15]. Our treatment diets simulated the stoichiometries of these two alternative food sources. Our study shows that after translocation from a high- to a low-predation environment, guppies of the Lower La Laja have evolved to be less selective of nutrient content. While not replicated, this change is consistent with the observed natural differences in the diet of natural high- and low-predation guppy populations [15].

Guppies inhabiting low-predation environments typically attain higher densities than high-predation populations [17], are strongly regulated [22] and have less and lower-quality resources available [23]. This means that competition for resources is likely to be stronger, and selectivity may be selected against. This pattern of lower selectivity at lower resource availability has been shown in a variety of organisms, including back bears Ursus americanus [24] and bluegill sunfish Lepomis macrohirus [25]. Similarly, a study in recently introduced populations of Podarcis sicula lizards into Adriatic islands [26] shows rapid increases in herbivory rates and associated morphological traits as densities increase. However, it is unknown whether those changes have a genetic basis. The strong environmental dependence of diet poses challenges for experimental studies of trophic evolution. One exception is the study showing heritable variation of omnivory in mullein bugs Campylomma verbasci [27]. Our study follows this, also showing that heritable change of diet preference, and therefore trophic position, can occur in less than 12 years or approximately 36 generations. Other traits such as male coloration [28], metabolic rate [29], or life history [30], have shown even faster rates of evolution in this population.

Our study leaves unclear the mechanism by which the evolved food preference operates. The fact that there was no effect of food treatment suggests that learning is not a necessary component of food preference. There was also not a significant effect of population on their first choice of food, suggesting that guppies may have had to try different food types to realize what they prefer when there is a preference at all. While in nature they are likely to use other cues, such as visual appearance, movement or position in the water column, it seems that visual cues are not necessary for diet choice evolution because our artificial food was identical in appearance. We hypothesize that our results could be driven by the evolution of taste, a mechanism proposed by Demi et al. [31] to be a response to trophic level and stoichiometric demands. This does not rule out that other modes of food recognition can evolve at the same time and reinforce preferences. Studies with more realistic food choices are needed to evaluate the ecological relevance of our results.

It is also important to acknowledge that our sample sizes are limited, and therefore may have precluded the detection of some meaningful effects, especially interactions. We did not find an effect of food treatment on preference. While we have interpreted this as a lack of learning or habituation, it could be due to a mere lack of power (although it is worth noting also the small effect size). We did not find any population differences in the latency or time spent feeding, yet it is likely that populations that differ in life history and food choice also differ in overall motivation and rates of feeding.

Guppies have become a model system for the study of eco-evolutionary dynamics. Their evolution has been shown to have profound effects on the ecosystem by altering their trophic positions in response to predation and competition levels [7,8]. At the same time, many traits in guppies have been shown to evolve rapidly in response to these conditions [32,33], including populations in the same tributary here studied [28,30]. However, for an eco-evolutionary feedback to occur, rapid evolution must occur in the traits that have ecological effects. We have here shown evidence for that missing piece of the puzzle: diet preference not only has important ecosystem consequences [7], but can evolve rapidly in response to ecological change. If this result generalizes to other introduced populations, it will have important consequences for our understanding of eco-evolutionary feedbacks in Trinidadian streams and other ecosystems.

Ethics

The collection of fish was approved by the Ministry of Agriculture, Land and Marine Resources, Republic of Trinidad and Tobago. All animals used were approved by the Washington University in St. Louis' Institutional Animal Care and Use Committee (protocol 19-1122).

Data accessibility

The data and RMarkdown code are available in the electronic supplementary material. The description of the data is included in the Rmarkdown code file, also available in pdf.

The data are provided in electronic supplementary material [34].

Authors' contributions

S.S.: conceptualization, data curation, investigation, methodology, project administration, writing—original draft, writing—review and editing; A.M.: conceptualization, investigation, methodology, project administration, supervision, writing—review and editing; J.R.A.: conceptualization, investigation, methodology, project administration, supervision, writing—review and editing; N.K.: conceptualization, investigation, methodology, writing—review and editing; A.S.: conceptualization, investigation, methodology, writing—review and editing; S.P.G.: conceptualization, formal analysis, funding acquisition, investigation, methodology, project administration, resources, supervision, writing---original draft, writing—review and editing; A.L.-S: conceptualization, data curation, formal analysis, funding acquisition, investigation, methodology, project administration, resources, supervision, visualization, writing—original draft, writing—review and editing.

All authors gave final approval for publication and agreed to be held accountable for the work performed therein.

Conflict of interest declaration

We declare we have no competing interests.

Funding

Funding was provided by start-up funds from Washington University in St. Louis to A.L-S and S.P.G.