ON THE MAINTENANCE OF GENETIC VARIATION BY DISRUPTIVE SELECTION AMONG HOSTS IN A PHYTOPHAGOUS MITE

The phytophagous mite Tetranychus urticae is broadly polyphagous, having been reported from hundreds of different host species in many different families (Jeppson et al., 1975). Some hosts, such as varieties of tomato and cucumber, typically cause high juvenile mortality of T. urticae (e.g., Gould, 1978a, 1979; Fry, 1988a, 1989), even though these hosts are often used by natural populations of the mites (Gilbert et al., 1966; Gould, 1978b; Fry, 1988a). In laboratory experiments, however, T. urticae populations have been shown to be able to rapidly adapt to tomato, cucumber, and other initially unfavorable hosts, sometimes experiencing more than two-fold gains in survival in under 10 generations (Gould, 1979; Fry, 1989). This raises the question of how genetic variation for fitness on these hosts is maintained. In other words, why should populations often be far below their fitness maxima in environments they normally encounter? One obvious way of answering this question is to invoke "trade-offs" in adaptation to different hosts (Gould, 1979; Rausher, 1983; Mitter and Futuyma, 1983; Futuyma and Peterson, 1985). If genotypes favored by selection on some hosts are selected against on others, this could keep populations from attaining their fitness maximum on any host. In fact, experiments by Gould (1979) and Fry (1990) support the trade-off hypothesis for T. urticae. In these experiments, laboratory populations that had been allowed to adapt to cucumber and tomato, respectively, lost a considerable proportion of that adaptation after being maintained on a third host, lima bean, for about 10 generations. Selection rather than genetic drift appeared to be the more likely explanation for this "reversion," because population sizes were on the order of 1,000 or more. While the reversion in the above experiments gave evidence that genotypes that were relatively well adapted to cucumber or tomato were selected against on bean, neither Gould nor Fry could detect statistically significant differences between the cucumber or tomato-adapted line and the control line in survival or fecundity on bean. In Fry's (1990) study, a confidence interval showed that fecundity on bean of the tomatoadapted line was probably not more than 10% lower than that of the control line. The fitness comparisons on bean in both studies were conducted under un-