Research
David Houle

For at least the last 50 years, evolutionary geneticists have held to the fundamental importance of studying genetic variation in populations. The logic is that, because genetic variation is essential for evolution to proceed, we can assume that the evolutionary process is limited by variation. If it is, then learning about the nature of genetic variation and the processes that maintain it will help us understand and predict adaptation. Until the last five years, I primarily followed this logic and devoted much of my research to tests of one particularly testable hypothesis for the maintenance of genetic variation, mutation-selection balance (see, e.g., Houle et al. 1994, 1996; Houle, 1998). More recently, I have suggested that many of the mechanisms that can maintain genetic variation will in fact maintain variation that is irrelevant to future adaptation (Houle et al. 1996; Houle, 2001). This suggestion implies that adaptation may be limited by mutation rather than variation. Surprisingly, tests of the relevance of standing variation to future evolution have not been carried out. The major project now under way in my lab is to examine the relationship between mutation, standing variation, and among-species variation.

If variation limits evolution, then variation within populations should correspond well with that among populations. Surprisingly, the number of attempts to test for such a correspondence is very small. I have recently obtained funding to test a series of conjectures about the dimensionality of variation and the predictability of evolution based on variation, using wing form in the family Drosophilidae (which includes the geneticist’s fruit fly) as a convenient model system. I have designed a semi-automated approach to measuring the entire vein structure that allows measurement of one wing about every 15 seconds with high repeatability. The system is described in (Houle et al. 2003), and the software is available from this web site. 

Many developmental mutants are known, and information about the genetics of wing development is rapidly accumulating. Most species of Drosophilids have the same homologous vein structures, and their arrangement is fairly conservative (see fig 1), although remarkably nearly every species can be discriminated on the basis of wing shape alone (Houle et al. 2003). This conservatism could be explained either by natural selection or by constraints on the nature of variation— that is by reduced dimensionality of the developmental system.

To estimate the dimensionality, we have measured the standing-genetic variation in Drosophila melanogaster and D. simulans. Our quantitative genetic experiments are unusually large in size with approximately 20,000 flies in 1,000 families measured in each species. This unusually rich data set shows that essentially all aspects of wing shape are genetically variable, although some are more variable than others. To be sure of this conclusion, we have developed and compared several statistical approaches to estimate the dimensionality of variation. These results strongly implicate natural selection as the force that causes wing form in the family Drosophilidae to be so conservative.

In addition, we have compared the genetic variation caused by spontaneous mutation within species and the variation among species to the standing genetic variation. The profile of genetic variation across these three levels, mutation, standing genetic variation, and among species variation, is remarkably similar. The same aspects of wing form that are most variable in the mutation data set are also the most variable within populations and among species. While there are interesting quantitative differences, overall these data suggest that the study of within species variation is an excellent indicator of the ability to evolve over long time scales.

In later parts of the project, I will study the effects of discrete wing mutations singly to see whether their effects fall along a small number of phenotypic dimensions. Mutant effects will be studied in pairs to test the effects of epistasis on dimensionality. Spontaneous and induced mutations will be studied within three different Drosophila species.

One of the challenges in this work is the lack of validated techniques for comparing the variation between different populations or at different levels. My recently departed postdoc Jason Mezey and I are devoting considerable effort to these questions. We recently published two investigations of the properties of one popular new approach (CPC analysis; Houle et al. 2002; Mezey and Houle, 1993) that demonstrate both the many shortcomings of the technique, and some potential advantages. Currently, we are investigating common space analysis, confirmatory factor analysis, and random skewers as alternatives. We plan to validate such measures of dimensionality using selection experiments in future years of the project.

Evolvability of the genetic system

Thomas Hansen and I have just obtained a three-year NSF grant to examine two potentially key determinants of the evolvability of traits under selection. The vast majority of work in evolutionary genetics is aimed at determining how evolution is capable of producing change over time. However, examination of the paleontological evidence suggests that this is only a small part of the story: more striking is the fact the phenotypes stay the same for very long periods of time (that is, show stasis), and only occasionally show major changes (Hansen and Houle, 2004). This new project will test two ideas that might help to explain stasis. The first idea is that the phenotype is so integrated that selecting on one trait necessarily drags so many other traits along with it that little overall progress can be made. The second idea is that the genetic system might actually resist change by damping the effects of mutations once the phenotype departs from its usual state.

To test these ideas, we are carrying out artificial selection experiments over a relatively long period. To test the first idea, we will look at the reduction (if any) in response to one selection pressure when seemingly unrelated traits are also subject to strong selection, versus when they are not. To test the second idea, we are generating lines with contrasting histories of selection - directional up and down, stabilizing, disruptive, fluctuating and of course controls. We now have completed 25 generations of selection on an index of wing shape. Results after 14 generations are shown in Houle et al. (2003). Once these selection histories are created, we will then substitute genes into each of these selected backgrounds to see if the magnitude of their effects is different. If a history of directional selection actually decreases the effect of new variation, then stasis might in part be explained by such effects.

Quantitative predictions about adaptation

An alternative approach to making  predictions about adaptation is to study natural selection.   In collaboration with Locke Rowe,  (Houle and Rowe, 2003) I have proposed that the study of natural selection in laboratory systems provides the most practical  opportunity to do this. Field studies have often identified strong natural selection, which can be combined with quantitative genetic estimates to predict evolution, but field tests of those predictions are difficult and have rarely been carried out. Experimental studies of adaptation in the laboratory are designed to create strong natural selection, and populations often show spectacular evolutionary responses. However, the actual form of natural selection is never estimated in such studies, so quantitative predictions are not possible. Laboratory studies of natural selection allow predictions in a context where those predictions can be tested.

As a test case, we have used manipulative experiments to estimate the form of selection in my laboratory fly population, IV, which has been maintained under the same conditions in the laboratory for more than 600 generations, enough time for it to approach evolutionary equilibrium. The IV population is transferred once every two weeks, a period a little longer than the generation time, so the IV population is analogous to a seasonal population. In contrast, wild D. melanogaster populations ancestral to the IV stock are multivoltine and have overlapping generations. We have confirmed that flies are under strong selection to lay eggs soon after transfer and to develop quickly, but a conflicting fecundity advantage accrues to large size. We have worked on a quantitative model to identify the optimal age at maturity on the basis of this information. This model shows that, for eggs laid early in a new bottle, the population mean is close to the optimum, but it also suggests an evolutionary constraint on plasticity in age at maturity. The nature of these constraints will be the subject of future experiments.

Why should animals care who they mate with?

In recent years, biologists studying sexual selection have been concerned with determining the evolutionary causes of mate choice. One of the most intuitive potential explanations is that animals choose their mates to increase the fitness of offspring—the so-called good-genes process. A traditional objection to this idea is that the variation in genetic quality is not high enough to support such a process. This objection has now been shown to be false, in part on the basis of my work showing that mean-standardized measures of variation are most appropriate for such comparisons (Houle, 1992). Some years ago, Locke Rowe and I (Rowe and Houle 1996) suggested that high variation in sexually selected traits is explained by a combination genetic variation in condition (overall quality of the individual) and the evolution of condition-dependence under sexual selection. Both are well supported by data. The necessary result is that the variance in sexually selected traits will come to reflect the genetic variance in condition, thus increasing the likelihood that sexually selected traits will serve as reliable indicators of mate quality. Thus all sexual selection may come to reflect good genes. This model has now stood up to several empirical tests and is being widely investigated.

More recently, I have extended this basic model to quantify the potential importance of the good-genes process. Some previous models had claimed that the good-genes process was not quantitatively sufficient to explain the extreme cases of male displays. Kondrashov and I (Houle and Kondrashov, 2002) built a new model, based on relative female preferences for male displays, that can explain arbitrarily large exaggerations of displays with very modest levels of female choice, even when choice is costly. It also shows that strong genetic benefits accrue to females and to the population as choice reduces the prevalence of deleterious alleles in the population.

These models suggest that the question of the effects of mate choice on population fitness have not received enough attention from empiricists. Drosophila melanogaster provides an ideal system in which to test for such effects, as mate choice can be short-circuited or enhanced by varying the numbers of potential mates. Furthermore, we can engineer populations that we know are segregating for deleterious genetic variation in a number of ways. So far we have run six such experiments. In two cases mutants were eliminated more quickly with sexual selection than without it, but in the other four cases sexual selection did not affect the rate of adaptation. This result suggests that good-genes sexual selection only affects a subset of all naturally selected alleles.

Seven years ago, I was asked to comment on a paper on the inheritance of small departures from bilateral symmetry, termed fluctuating asymmetry. I realized that the standard idea of how such asymmetries develop suggested that even identical individuals in the same environment would vary a great deal in degree of asymmetry (Houle, 1997). Subsequently, I have published several other papers that investigate the implications of this simple model for the study of fluctuating asymmetry (Houle, 2000; Fuller and Houle 2002, 2003). This simple consideration suggests that much recent work that claimed that asymmetry was a good predictor of organism quality was likely to be flawed. I was subsequently asked to review a book that championed this pro-asymmetry view. I found this book to be riddled with logical flaws, biased interpretations of the published literature, and shady use of the work of others. My critical review (Houle, 1998) had a large impact on this whole area of research.

Houle site index