Letter | Published:

Rapid evolution drives ecological dynamics in a predator–prey system

Nature volume 424, pages 303306 (17 July 2003) | Download Citation

Subjects

Abstract

Ecological and evolutionary dynamics can occur on similar timescales1,2,3,4,5,6,7. However, theoretical predictions of how rapid evolution can affect ecological dynamics8 are inconclusive and often depend on untested model assumptions8. Here we report that rapid prey evolution in response to oscillating predator density affects predator–prey (rotifer–algal) cycles in laboratory microcosms. Our experiments tested explicit predictions from a model for our system that allows prey evolution9. We verified the predicted existence of an evolutionary tradeoff between algal competitive ability and defence against consumption, and examined its effects on cycle dynamics by manipulating the evolutionary potential of the prey population. Single-clone algal cultures (lacking genetic variability) produced short cycle periods and typical quarter-period phase lags between prey and predator densities, whereas multi-clonal (genetically variable) algal cultures produced long cycles with prey and predator densities nearly out of phase, exactly as predicted. These results confirm that prey evolution can substantially alter predator–prey dynamics, and therefore that attempts to understand population oscillations in nature10,11 cannot neglect potential effects from ongoing rapid evolution.

Main

In our experiments, parthenogenetic planktonic rotifers, Brachionus calyciflorus, consumed unicellular, obligately asexual12 green algae, Chlorella vulgaris. These were cultured in continuous flow-through chemostat systems in a nitrogen-limited medium. A simple differential-equation model predicts the transitions between the qualitatively different population dynamics observed in this system (equilibria, population cycles and extinction) as a function of the nutrient concentration of the medium and the dilution (flow-through) rate of the chemostat13. However, cycle periods were far longer than predicted, and observed predator and prey cycles were almost exactly out of phase (predator maxima and prey minima were nearly simultaneous, and vice versa), which is impossible in a conventional predator–prey model13. These unexpected cycles had extended periods when algal biomass was high but rotifer densities remained low, followed by increased rotifer growth even though algal densities remained nearly constant.

To identify possible mechanisms that might account for these observations, we developed simple mathematical models for a series of alternative biologically plausible hypotheses9: changes in algal carbon:nitrogen ratio as a function of nutrient availability; rotifer self-limitation through toxin production or decreased egg viability when prey are scarce; and rapid prey evolution resulting from an evolutionary tradeoff between algal competitive ability and defence against rotifer predation. The model with rapid prey evolution reproduced the observed cycles much more accurately than our original model, whereas the non-evolutionary models failed to produce cycles similar to those observed9.

Here we present experimental tests of the prediction that the qualitative properties of rotifer–algal cycles in our system are the result of rapid prey evolution. To test for the evolutionary tradeoff postulated by the prey evolution model, we manipulated algal growth conditions. Algae cultivated under constant and intense rotifer grazing pressure became lower in food value (in the sense that rotifers that fed on these cells had a lower population growth rate) and were heritably smaller and competitively inferior relative to algae grown in the absence of rotifers but with a comparable mortality rate imposed by an elevated chemostat dilution rate (Table 1). These results show the existence of different genotypes in the algal population, and the presence of a tradeoff between algal food value and competitive ability among clones. The heritable response to rotifer predation also shows that the selected low-food-value algal genotypes are better able to survive rotifer grazing, although the mechanism for this remains to be determined.

Table 1: Evolutionary tradeoff for C. vulgaris

We added genetic structure, and therefore the possibility for algal evolution, to our original model13 by representing the algal population as an assortment of clones differing in food value and competitive ability (Fig. 1). Clonal selection is appropriate because C. vulgaris reproduces asexually12 and because we showed (above) the presence of different genotypes—clones—in our laboratory population (Table 1). The model incorporates the observed evolutionary tradeoff by assuming that lower-food-value clones have a higher half-saturation constant for nutrient uptake (Fig. 1a, c; see Methods for additional details).

Figure 1: Predicted effects of prey clonal diversity on population dynamics.
Figure 1

a, c, Modelled tradeoff curves between algal food value and competitive ability (represented by the half-saturation constant, Kc, for nutrient uptake by algae), with α′ >1 (a) and α′ <1 (c). b, d, Effects of clonal diversity on predator–prey cycle length for the two tradeoff curves between algal food value and competitive ability, with α′ >1 (b) and α′ <1 (d). Cycle periods in b and d are the result of 700 simulation runs each for one, two, three, five or seven clones present at the start of the run. Cycle periods are plotted against the number of clones remaining after running the model long enough for competitively inferior clones to be eliminated. ‘Degenerate’ cases (in which some of the initially present clones were eliminated) are plotted just to the right of non-degenerate cases in b and d. e, f, Cycles predicted by the model for the Brachionus predator (solid line) and Chlorella prey (dashed line) in a single-clone system (e) and a multiple-clone system (f). Model equations and further technical details are given in the Supplementary Information.

A fundamental prediction of the multiple-clone model is that prey evolution can substantially alter the population cycles (Fig. 1). Model simulations for systems composed of randomly generated sets of one, two, three, five or seven clones show that when multiple clones are present, long cycles are possible (Fig. 1b, d) irrespective of the tradeoff curve specified between food value and competitive ability (Fig. 1a, c). Cycle lengths for two or more coexisting clones comprised a range of periods. Short cycles can occur when the algal population consists of two very similar clones, but with more than two clones short cycles become highly unlikely. In contrast, a single-clone system always produces relatively short cycles, regardless of that clone's phenotype. Moreover, population cycles in a single-clone system exhibit the classic predator–prey phase relations in which the peaks in predator abundance follow the prey peaks by one-quarter of a cycle (Fig. 1e). However, if the prey population consists of two or more phenotypically divergent clones, and predation by rotifers in combination with competition for nutrients results in rapid changes in algal genotype frequencies, the resulting multi-clone population cycles show phase relations like those observed, with algae and rotifers almost exactly out of phase (Fig. 1f). These properties result from the increasing dominance of low-food-value clones as the algal population is grazed down. Therefore, when algal density increases again after the rotifer population has crashed, the rotifer population does not immediately respond. Rotifers cannot increase until the algae return to high density, at which point higher-food-value clones (which are better competitors) increase in abundance. Rotifer grazing then intensifies and the cycle begins anew.

To test these predictions we manipulated prey evolution by altering clonal diversity (that is, genetic variability) in the prey population. We initiated replicated chemostat trials either with a single clone as an evolutionarily ‘stagnant’ population or with multiple clones comprising an evolutionarily ‘active’ one, at dilution rates and nutrient concentrations that had previously been shown to produce population cycles13. We then compared cycle periods and phase relations between the treatments.

Our results are strikingly consistent with theoretical predictions. Population cycles of the single-clone evolutionarily stagnant treatment were much shorter and of smaller variance than those of the multiple-clone treatment (Figs 2 and 3). In addition, the single-clone treatments exhibited ‘classic’ predator–prey cycles with roughly a quarter-cycle delay between prey and predator maxima (Fig. 2; mean phase lag as a fraction of cycle period, 32%; range 26–36%). By contrast, in the multiple-clone evolutionarily active treatment, the predator and prey maxima were almost exactly out of phase (Fig. 2), as predicted by the clonal evolutionary model (mean phase lag, 56%; range 41–68%). The clonal evolutionary model is also able to match the experimentally determined bifurcation diagram of the system13 (transition points between stable and cyclic dynamics as the dilution rate is varied) more accurately than the non-evolutionary model (see Supplementary Information).

Figure 2: Experimental results showing the population cycles of rotifer–alga systems.
Figure 2

ad, Single-clone algal populations; ei, multiple-clone algal populations. Filled circles, B. calyciflorus (predator); open circles, C. vulgaris (prey). In ad there are short-period predator–prey cycles with the classical phase relations; in ei there are long cycles with predator and prey oscillations nearly out of phase with each other. Dilution rates, δ (d-1) a, 0.57; b, 0.65; c, 0.67; d, 0.68; e, 0.72; f, 0.64; g, 0.69; h, 0.95; i, 1.00. Data for four (fi) of the five multiple-clone trials were obtained from ref. 13. The fifth trial is a new experiment, conducted in parallel with the single-clone trials, to verify that the lapse of time did not result in any change of our rotifer or algal stock cultures that would change the qualitative dynamics of a system with multiple algal clones. Note that population cycles at relatively high dilution rates (h, i) were also long and with out-of-phase relations, as our mathematical model predicts.

Figure 3: Cycle periods for experimental populations of B. calyciflorus (filled circles) and C. vulgaris (open circles).
Figure 3

Lines connect period estimates from the same experimental trial, and the labelling of points corresponds to the panel labels in Fig. 2. The large difference between prey and predator period estimates by spectral analysis in experiment f resulted from the strongly non-sinusoidal shape of the cycles.

Predators can also evolve rapidly7. Although we have not yet investigated predator evolution, it is significant that prey evolution alone brought our model into close agreement with experiments. Two explanations are possible. One is that the selection pressure of predators on prey might be stronger than that of prey on predators (the ‘life–dinner’ dichotomy14,15). Alternatively, predator evolution simply might not have much influence on our system's dynamics. Although inducible defences by prey are often observed in nature16, none are evident in our system. Grazing-adapted cells are smaller but otherwise identical to non-adapted cells in morphology and chemical composition (C:N ratio; see Methods), and both types were predominantly unicellular. Inducible defences can also be ruled out as the cause of long-period cycles, because the cycle length and phase relations are both consequences of the fact that algal food value remains poor even after several algal generations without rotifer predation9.

Our results provide experimental support for model-based approaches to identifying the causes of natural population cycles11,17. An ideal experimental test removes or alters a candidate causal process (as we have done here) and compares observed with predicted effects on dynamics. However, in natural systems such tests are typically infeasible because of the large spatial and temporal scales at which systems would have to be manipulated18. Even when such experiments are possible, their interpretation can be problematic11. An alternative is to construct mechanistic models corresponding to each potential causal mechanism and to compare each model's consistency with available observational and experimental data11,17,19,20,21. Our results here confirm the conclusion previously derived by this approach: that among all candidate causal processes only rapid prey evolution could account for our system's dynamic patterns.

We have shown that ongoing rapid prey evolution can alter population dynamics (cycle period and phase relations) in a live predator–prey system. Population cycles can lead to fluctuating selection, so that rapid prey evolution occurs indefinitely and reciprocally influences population dynamics. The interactions governing ecological dynamics are then continually changing through rapid evolution. Our study indicates that ecologists must consider ongoing rapid evolution when exploring the mechanisms underlying the dynamics of populations and food webs. Evolutionary and ecological dynamics must be understood in concert.

Methods

Selection experiment

To assess variation in food value and competitive ability, C. vulgaris populations were exposed in two chemostats either to constant intense rotifer predation (dilution rate δ = 0.19 d-1) or to unselective mortality from elevated washout (δ = 1.24 d-1). Algal C:N ratios in the two treatments were similar (molar C:N = 13–14) and lower than in algae cultured under nutrient-deficient conditions (δ = 0.16 d-1, molar C:N = 20), indicating that both treatments experienced an equivalent high availability of nutrients. Algae were harvested after each chemostat had reached equilibrium and then used to determine the effects of previous selection under common conditions.

Algal food value was determined by bioassay: rotifer population growth rate when fed with algae at a density (7 × 106 cells ml-1) above the incipient limiting concentration22. Experimental procedures followed ref. 23. Algal competitive ability was estimated by measuring algal population growth rate under nutrient-deficient conditions in short-term batch cultures: 100 ml culture medium with 1 µM nitrate. Algae were inoculated at 500 cells ml-1, grown at 25 °C under constant 120 µE m-2 s-1 illumination (as in the chemostat experiments), and cell density and size distribution were monitored daily. Algal populations grew exponentially for the first 4–5 days; maximum growth rate was determined as the maximum derivative of a local polynomial curve fitted to the logarithm of algal density against time.

Simulation analysis

Model equations, parameter values and details are given in the Supplementary Information. Our model is based on that of ref. 13, a system of nonlinear differential equations for the state variables: limiting nutrient, algal abundance, breeding and total rotifer abundances. The two rotifer variables represent rotifer age structure, with rotifers gradually ceasing to breed with age; in other respects it is a standard chemostat model with Monod equations24 for algal nutrient uptake, and for predation by rotifers on algae, with fecundity proportional to consumption. Here, we replaced the single algal variable by a series of algal clones. Clones are characterized by their relative food value to rotifers, p, ranging between 1 (‘good’) and 0 (‘poor’), with lower p resulting in reduced predation risk. The defensive ‘low-food-value’ trait comes at the expense of reduced competitive ability (*). To model this tradeoff, lower-p clones are given a higher half-saturation constant for nutrient uptake and are therefore poorer competitors when nutrients are scarce (our previous model9 made different assumptions that are inconsistent with subsequent experimental results (Table 1, and T.Y., unpublished data) and used quantitative trait dynamics rather than clonal selection).

We initialized simulations with q = 1, 2, 3, 5 or 7 clones with random p values (uniformly distributed on [0.02, 1]) and competitive ability assigned according to one of two different tradeoff curves (Fig. 1a, c). Competitive ability (represented by the half-saturation constant Kc) is related to food value by Kc = K + α(1 - pα)1/α, where K is the minimum value applying when p = 1, α is a scale parameter describing the cost of decreased food value and α′ is a shape parameter determining the concavity of the tradeoff. Estimation of tradeoff curves is described in Supplementary Table 1. Clone sets were randomly drawn 700 times for each q. Clones comprising less than 1% of the final algal population were considered extinct. We then recorded the number of surviving clones and the cycle period (scored as period = 0 for coexistence at equilibrium).

Chemostat experiments

We established stock cultures of Chlorella vulgaris (UTEX no. 26) and Brachionus calyciflorus (taken originally from Milwaukee harbour, Wisconsin, and provided by M. Boraas). Although we do not know how much clonal (genotypic) variation for food value and competitive ability was present in our algal stock culture, populations derived from it in our selection experiment have been shown to possess distinct heritable phenotypes (T.Y., unpublished data), so we conclude that it consists of multiple clones. Isolated single algal clones were established by spot-plating cells from the stock culture on agar plates. We chose one of these arbitrarily for use in the experiments reported here.

The experimental chemostats followed ref. 13. To prevent algal contamination in the treatments using single algal clones, rotifers were taken from stock culture, rinsed thoroughly with sterilized medium, and held in sterilized medium for at least 6 h to ensure that all algal cells in their stomachs were digested. Rotifers were then thoroughly rinsed again with sterilized medium before being used to initiate chemostat trials. We set the dilution rate (0.57–1.00 d-1) and nutrient concentration (80 µM nitrate) to give population cycles (ref. 13 and our model). Organisms were sampled daily through ports near the bottom and top of each chemostat. Rotifers were counted under a dissecting microscope and algae were counted with a particle counter (CASY 1; Schärfe, Reutlingen, Germany). Organism abundance data are means of duplicate samples. We determined cycle periods by spectral analysis of organism counts after rotifer populations were established (more than one female ml -1) using spec.pgram in the R language25, and phase lags from the cross-correlation function between predator and prey counts (ccf in R).

References

  1. 1.

    et al. Rapid evolution revealed by dormant eggs. Nature 401, 446 (1999)

  2. 2.

    , , , & Rapid evolution of a geographic cline in size in an introduced fly. Science 287, 308–309 (1999)

  3. 3.

    , , , & Rapid evolution of reproductive isolation in the wild: Evidence from introduced salmon. Science 290, 516–518 (2000)

  4. 4.

    et al. Frontiers of ecology. BioScience 51, 15–24 (2001)

  5. 5.

    The Evolution Explosion: How Humans Cause Rapid Evolutionary Change (W. W. Norton, New York, 2001)

  6. 6.

    , & Density cycles and an offspring quantity and quality game driven by natural selection. Nature 406, 985–988 (2000)

  7. 7.

    & Linking genetic change to community evolution: Insights from studies of bacteria and bacteriophage. Ecol. Lett. 3, 362–377 (2000)

  8. 8.

    The evolution of predator–prey interactions: theory and evidence. Annu. Rev. Ecol. Syst. 31, 79–105 (2000)

  9. 9.

    , , & Predator–prey cycles in an aquatic microcosm: Testing hypotheses of mechanism. J. Anim. Ecol. 71, 802–815 (2002)

  10. 10.

    Berryman, A. (ed.) Population Cycles: The Case for Trophic Interactions (Oxford Univ. Press, 2002)

  11. 11.

    Complex Population Dynamics: A Theoretical/Empirical Synthesis (Princeton Univ. Press, 2003)

  12. 12.

    Green Algae: Structure, Reproduction and Evolution in Selected Genera (Sinauer Associates, Sunderland, Massachusetts, 1975)

  13. 13.

    , , & Crossing the Hopf bifurcation in a live predator–prey system. Science 290, 1358–1360 (2000)

  14. 14.

    Evolution and Escalation: An Ecological History of Life (Princeton Univ. Press, 1987)

  15. 15.

    The evolutionary interaction among species: selection, escalation, and coevolution. Annu. Rev. Ecol. Syst. 25, 219–236 (1994)

  16. 16.

    Tollrian, R. & Harvell, C. D. (eds) The Ecology and Evolution of Inducible Defenses (Princeton Univ. Press, 1999)

  17. 17.

    et al. Why do populations cycle? A synthesis of statistical and mechanistic modeling approaches. Ecology 80, 1789–1805 (1999)

  18. 18.

    , , & in Population Cycles: The Case for Trophic Interactions (ed. Berryman, A.) 155–176 (Oxford Univ. Press, 2002)

  19. 19.

    & The Ecological Detective: Confronting Models with Data (Princeton Univ. Press, 1997)

  20. 20.

    , , , & Large-amplitude cycles of Daphnia and its algal prey in enriched environments. Nature 402, 653–656 (1999)

  21. 21.

    et al. Dynamical effects of plant quality and parasitism on population cycles of larch budmoth. Ecology (in the press)

  22. 22.

    & Quantitative relations between phytoplankton and the population dynamics of the rotifer Brachionus calyciflorus Pallas. Results of laboratory experiments and field studies. Arch. Hydrobiol. 73, 273–309 (1974)

  23. 23.

    Algal nutrient limitation affects rotifer growth rate but not ingestion rate. Limnol. Oceanogr. 40, 1201–1208 (1995)

  24. 24.

    La technique de culture continue: theorie et applications. Ann. Inst. Pasteur Lille 79, 390–410 (1950)

  25. 25.

    & R: a language for data analysis and graphics. J. Comp. Graph. Stat. 5, 299–314 (1996)

Download references

Acknowledgements

We thank B. Kendall, K. Shertzer, J. Urabe and members of the EEB theoretical ecology ‘lunch-bunch’ for comments on the manuscript; A. Sasaki and C. Aquadro for discussions on clonal evolution; and M. Armsby, S. Hammer, M. Hung, C. Kearns, K. Keller and J. Meyer for assistance with the experiments. The study was supported by a grant from the Andrew W. Mellon Foundation to S.P.E. and N.G.H.

Author information

Author notes

    • Gregor F. Fussmann

    Present address: Institut für Biochemie und Biologie, Universität Potsdam, Maulbeerallee 2, D-14469 Potsdam, Germany.

Affiliations

  1. *Department of Ecology and Evolutionary Biology, Cornell University, Ithaca, NY 14853, USA

    • Takehito Yoshida
    • , Laura E. Jones
    • , Stephen P. Ellner
    • , Gregor F. Fussmann
    •  & Nelson G. Hairston Jr

Authors

  1. Search for Takehito Yoshida in:

  2. Search for Laura E. Jones in:

  3. Search for Stephen P. Ellner in:

  4. Search for Gregor F. Fussmann in:

  5. Search for Nelson G. Hairston in:

Competing interests

The authors declare that they have no competing financial interests.

Corresponding author

Correspondence to Nelson G. Hairston Jr.

Supplementary information

About this article

Publication history

Received

Accepted

Published

DOI

https://doi.org/10.1038/nature01767

Further reading

Comments

By submitting a comment you agree to abide by our Terms and Community Guidelines. If you find something abusive or that does not comply with our terms or guidelines please flag it as inappropriate.