Short-term evolution in the size and shape of pea aphids.
The disruption of the phenotype by recombination may be another situation in which evolutionary change can be witnessed. When the response to selection is attributable to dominance or epistasis, recombination can move the mean phenotype away from the direction of selection in a process called "genetic slippage" (Lynch and Deng 1994; note that this usage of genetic slippage differs from that of Dickerson 1955). Slippage is expected to be particularly pronounced in cyclically parthenogenetic species in which evolution by clonal selection on completely linked parthenogenetic genotypes can proceed for several generations between periods of recombination. In such species, slippage can cause pronounced maladaptive phenotypic change (Lynch and Deng 1994), providing the opportunity to witness yearly cycles of evolution without major environmental change.
Herbivorous insects spend their lives in close association with one or more host-plant species. Previous comparative studies of variation between species or populations have revealed that insect response to selection by a diversity of plant surfaces has resulted in morphological specialization, especially with respect to modes of grasping and locomotion (Southwood 1986; Kennedy 1986; Moran 1987), and modes of probing and penetrating surfaces for food or oviposition sites (Dixon 1987; Heie 1987, Carroll and Boyd 1992). In a comparative study of species in the aphid genus Uroleucon, morphological differentiation between species was shown to be correlated with host-plant morphology (Moran 1986). Moran found that Uroleucon species that use hairy plant species have longer rostra (beaks) and shorter hind tarsi than do those using smooth plant species, suggesting that the different host plants have acted as agents of natural selection. Our study was undertaken to complement the previous comparative studies by evaluating whether evolution of particular morphological traits could be detected within a single population of one species.
Within populations, the magnitude and direction of phenotypic evolution depends both on the extent to which characters are genetically (co)variable and on the specific action of phenotypic selection (Lande 1979; Lande and Arnold 1983). Designed experiments that use relationships between members of a sexual population or that replicate clonal lineages permit phenotypic variability within contemporary populations to be partitioned into its genetic and environmental components (e.g., Falconer 1989; Via 1984, 1991). The magnitude of natural selection on particular characters in these populations may then be estimated by regressing individual relative fitness on the phenotypic values of those traits (Lande and Arnold 1983).
If one observes evolutionary change in the phenotype through time, estimates of genetic (co)variability obtained before selection can permit hypotheses to be formulated about how natural selection might have acted to produce the observed change (Lande 1979; Schluter 1984). For example, only characters that are genetically variable are expected to evolve under selection, because genetic variability is required for evolution. In addition, correlated responses to selection can produce evolutionary change in heritable traits that are not themselves the target of natural selection (Lande 1979; Lande and Arnold 1983). Thus, understanding patterns of genetic correlations among characters is essential for interpreting the action of natural selection.
Here, we document genetically based morphological variability within pea-aphid populations and show that the mean morphological phenotype of the aphid population within a single field can evolve in time periods as short as a single summer. Using information on the genetic (co)variances of morphological traits, we hypothesize that clonal selection on overall body size has led to correlated short-term evolutionary changes in shape within this pea-aphid population. This hypothesis was tested by a separate analysis of natural selection on overall size under field conditions.
Because pea aphids have a single sexual generation each fall, the fitness gains from selection during the clonal phase may be partially lost as a result of recombination during the sexual phase (Lynch and Deng 1994). We compared the mean size at the end of one season with the size at the beginning of the next year and found evidence that genetic slippage may be occurring in this population, perhaps leading to persistent yearly cycles of selective advance and retreat in fitness.
MATERIALS AND METHODS
Study Organism and Field Collections
Acyrthosiphon pisum is a cyclical parthenogen. Ameiotic parthenogenesis (Blackman, 1979) begins upon egg hatch in early spring and continues through late fall, when declining photoperiods and temperatures induce the production of a single generation of sexuals (Lamb and Pointing 1972; Via 1992). When the asexual fundatrix generation hatches in the spring from the overwintering eggs, genetic variability is expected to be at a maximum due to recombination during the sexual phase (Lynch and Gabriel 1983). This variability is expected to be eroded through clonal selection (differential fitness of clones) as the population evolves to resemble the most-fit clone (Charlesworth 1980, pp. 65-69). If genetically based variability in morphology is correlated with fitness, a change in the mean morphological phenotype should thus be evident as the season progresses.
To measure genetic variability in morphometric traits and the potential for evolution, field collections were made at the beginning of two consecutive summer seasons. An "early" collection of 15 parthenogenetic clones was made in late May 1988 from a single alfalfa field in Tompkins County, New York. Unfortunately, at the time of the 1988 collection, about three asexual generations had elapsed; thus, some evolution of the mean phenotype may have been missed. To rectify this problem, a more intensive sampling of this same field took place in April 1989 when 28 fundatrices (the first clonal generation) were collected. Because asexual generations had not yet been produced, there was little exposure of the 1989 clones to selection except that which may have occurred during overwintering of the eggs or in the brief period before collection of the first spring generation. Because the improved time and size of the sample in 1989 permits the best test of the hypothesis that morphology evolves during the season, most of this paper will concern the results from that year. All collection locations within the field were separated by about 100 ft to increase the probability that unrelated individuals were collected, and only one individual was collected per location.
To determine whether the mean phenotype changed over the course of a season, "late" collections were made in the same field in both years using the same sampling methods as employed for the early collections. Both of these late collections were made about seven to eight generations (approximately three months) after the corresponding early collections [late 1988: August (15 clones), and late 1989: July (28 clones)].
Maintenance of Clones in Laboratory Culture
Laboratory-raised sublines of each clonal lineage were used for the morphometric analyses. To establish the laboratory cultures, each field-collected aphid was placed in a ventilated two-gallon plastic container containing three pots of alfalfa. Every 10-14 d, 10 newly produced nymphs (first to second instars) from each clonal line were added to a new bucket of plants. Each clonal lineage was replicated into two separate sublines at least three generations prior to collection for morphometric analysis in order to remove unwanted environmental variance from the variation between clones (Lynch 1985). All clones were maintained in the same constant temperature chamber at 17-18 [degrees] C on a cycle of 16 h light, 8 h dark. Sublines were maintained on different shelves in the chamber, and positions were rotated periodically.
Mounting and Measurement
Adult aphids were preserved in 70% ethanol prior to morphometric analysis, then cleared by heating to 80 [degrees] C in 30% lactic acid for 8-12 h. Two individuals were measured per subline. Before dissection, several characters (Table 1, [ILLUSTRATION FOR FIGURE 1 OMITTED]) were measured on each unmounted specimen using a Wild M-3 dissecting microscope attached to a video digitizing system in an IBM PC (digitizing board: Image Technology, software package: CODA, Pisces Microcomputer Consulting, Rochester, NY). After measuring the wet mounts, each aphid was dissected to remove the head, legs, and siphunculi. This permitted precise orientation of body parts in the horizontal plane. Dissected aphids were mounted in Hoyer's solution and were measured using the same video digitizing system attached to a Leitz Diaplan microscope. Measurement error, calculated from the variation of repeated measures taken on the same individuals, was 0.5% to 1.5%.
TABLE 1. Summary of morphological variables measured on Acyrthosiphon pisum. See also figure 1.
Description of measurement
Variables measured on specimens before mounting
Body length (from head to base of siphunculi) Body width (measured between siphunculi) Head width (from left to right margins of ocular tubercles) Coxa width (body width from left to right outer margins of fore coxae) Head length (from central margin between antennae to notch between fore coxae)
Variables measured on mounted specimens
Hind tibia length Hind femur length Siphunculus width (at base) Siphunculus length Apical rostrum length Stylet length Hind tarsal length (second segment)
Estimates of Genetic Variability in Size and Shape
Although investigators usually identify morphological traits or landmarks for measurement convenience, investigation of a set of linear measurements does not necessarily separate overall size from aspects of shape, which may be the actual targets of selection. The geometric mean of a set of linear measurments provides one useful estimate of overall size (Mosimann and James 1979; Darroch and Mosimann 1985). The log geometric mean of p original variables is calculated for each observation as [[Sigma]ln([X.sub.i])]/p, where [X.sub.i] is the value of the ith variable for a given observation. To place the geometric mean in context, note that on the log scale it is similar to the first principle component, which often used as a measure of size and is also a linear combination of the variables but with potentially unequal weightings. When the original variables are all positively correlated as they are here, all of the weights in the first principle component will be positive, making it very similar to the geometric mean.
Shape variables, which reflect the lengths of body parts relative to overall size, were calculated for each observation by subtracting the log geometric mean from the log of each of the original variables (Darroch and Mosimann 1985). We chose to focus on size (log geometric mean) and these shape variables for our study rather than on the original measurement variables because it seemed likely that a contrast between size and shape would provide greater insight into the possible patterns of natural selection than would examination of a set of highly positively correlated linear measurement variables.
To determine whether the clones sampled early in the summer varied genetically in size and/or shape, a multivariate analysis of variance (MANOVA, PROC GLM, SAS Institute 1985) was performed separately for each year. In these models, significant clonal genetic variability is expected only if the variation among clones is in excess of environmental variation among sublines nested within clones. The error term estimates variance among replicates of each subline. For all analyses, "clone" and "subline(clone)" were considered to be random effects. Because only observations on which all variables were measured could be included in the MANOVAs, the data set was slightly unbalanced (primarily due to loss of one observation in some sublines, though in a few sublines both observations were lost).
Although the MANOVAs test whether there is significant overall morphological variance among clones, we also wanted to obtain estimates of the proportion of total variability in each variable that was due to genetic causes (the heritability, [h.sup.2]). Clonal heritabilities, which reflect total genetic variance, were calculated for size and each of the shape variables as the clonal component of variance divided by the total phenotypic variance using variance component analysis in SAS (PROC NESTED, SAS Institute 1985).
Estimates of Genetic Correlations
Because evolutionary change is a function of both genetic variances and covariances, the genetic correlations estimated for the early season data may be useful for interpreting the pattern of morphological change observed over a season. Given our small sample size, the genetic correlations are presented primarily to provide a qualitative picture of the expected correlated responses to selection based on the genetic relationships between the characters. Much larger sample sizes would be required to provide estimates of the genetic correlations that are precise enough for quantitative prediction of evolutionary trajectories.
A nested analysis of variance and covariance (PROC NESTED in SAS) was used to estimate the variance component genetic correlations between overall size (log geometric mean) and the shape variables for the clones collected early in 1989. The clonal component correlations were estimated as
[r.sub.G] = [cov.sub.G](X, Y)/[[[var.sub.G](X)[var.sub.G](Y)].sup.1/2],
where [r.sub.G] is the clonal genetic correlation, [cov.sub.G](X,Y) is the clonal covariance of characters X and Y among clones, and [var.sub.G] is the clonal genetic variance for X or Y.
Clone mean correlations between size and the shape variables are also presented for comparison. Such product-moment correlations are useful because they can be readily assessed for a significant difference from zero; they tend to be quite similar in value to variance component correlations for characters with significant genetic variances (Via 1984, 1991).
Estimating the Extent of Evolutionary Change over a Season
A significant difference in morphology between clones collected early and late in the season and tested under common conditions would reveal evolutionary change in the average size and/or shape of the population. To determine whether there were differences between the mean vectors of morphological traits in collections made at the different times during the season, the clones from the early- and late-season collections in both 1988 and 1989 were compared in a MANOVA, with subline and clone nested within time of collection (early or late). In addition, univariate analyses were performed to determine which individual characters changed significantly during the season (sequential Bonferroni corrections were used in all cases, Rice 1989).
Natural Selection on Overall Size
To test the hypothesis that the observed evolutionary changes in size and shape could have resulted from natural selection on overall size, we measured phenotypic selection using a subset of the early 1989 clones (15 total observations) on which both fitness components and morphological traits were measured on individuals reared in the field. Individual relative fitnesses were calculated separately for age at first reproduction ([Alpha]) and fecundity during the first 12 d of adulthood (fec12). Following the methods of Lande and Arnold (1983), these were individually regressed on standardized overall size (the geometric mean for each individual standardized by subtracting the mean value and dividing by the standard deviation). To check the results of this univariate analysis of directional selection, we also performed a larger regression analysis in which the standardized values of the three shape variables that changed the most through the season were also included, but this analysis had very low power due to the small sample size. Finally, we checked for potential curvature in the fitness function and corresponding stabilizing or disruptive selection on the variance in size by introducing the quadratic term into regressions of the two fitness components on overall size (Lande and Arnold 1983).
Genetic Slippage and a Cost of Recombination
We compared the mean phenotype of the group of aphids collected late in 1988 with that of the first generation of 1989 (the early 1989 collection) in order to evaluate the possibility that the mean may have moved away from the direction of selection following recombination. Such a change, termed genetic slippage, provides evidence for a cost of recombination due to the presence of nonadditive genetic variance in the variation among clones (Lynch and Deng 1994).
Estimates of Genetic Variability in Size and Shape Early in the Season
The MANOVAs of size and 12 shape variables reveal highly significant morphological variability among clones collected from the same field early in the season in both 1989 (Table 2) and 1988 (P [less than] 0.0001, Shaw 1990). The relative contribution of size and shape to differences between clones was determined by taking the ratio of the summed eigenvalue for a model containing only shape variables to that from one that included size and shape. This ratio (52.7/63.7) revealed [TABULAR DATA FOR TABLE 2 OMITTED] that in 1989, 17% of the variance among clones early in the season was attributable to differences in overall size.
In the early 1989 collection there was considerable variation among clones for both overall size and many of the shape variables (Table 3), suggesting that various aspects of morphology have the potential to evolve through the season if morphometric variation is correlated with variation in fitness.
Genetic Correlations between Size and the Shape Variables
All of the shape variables that were significantly genetically variable were highly correlated with body size in the early 1989 collection (all shape variables are expressed as lengths relative to overall size): the relative siphunculus length, hind tibia length, and hind femur length were positively correlated with size, whereas relative head length, apical rostrum length, head width, and width at the coxae were negatively correlated with size (Table 4). In general, the clone mean correlations between size and shape were very similar to the variance component correlations, particularly for the characters that were significantly genetically variable.
Change in the Mean Morphological Phenotype of Clones across a Season
In 1989, the average size and shape differed between pea-aphid clones collected early and late in the season (the "time" [TABULAR DATA FOR TABLE 3 OMITTED] effect in Table 5 and [ILLUSTRATION FOR FIGURE 2 OMITTED]). Given that we controlled for phenotypic plasticity due to environmental effects such as different seasonal temperatures by measuring clonal replicates reared in a common environment, this seasonal change in size and shape can be interpreted as strong evidence for short-term evolutionary change. The ratio of the summed eigenvalue for the time effect from a model containing only shape (1.024) to that from a model containing both and size and shape (1.679) reveals that differences in size alone (defined as the geometric mean) explained 40% of the overall morphological change between early and late season clones in 1989 [ILLUSTRATION FOR FIGURE 2 OMITTED], despite the fact that only 17% of the genetic variance in the early collection was attributable to overall size.
In 1988, a trend toward larger size late in the season was also evident [ILLUSTRATION FOR FIGURE 2 OMITTED], but differences between times were not significant, either for multivariate morphology or for size alone. However, it is worth noting [ILLUSTRATION FOR FIGURE 2 OMITTED] both that the early 1988 collection (made in late May) was larger than the early 1989 collection (made in April), and that the late collection in 1988 (made in mid-August) appeared to be slightly larger than in the late collection in 1989 (made in early July). Thus, a consistent trend toward larger size as the season progresses was seen over both years. Because the rate of evolution under clonal selection is expected to be exponential (Charlesworth 1980), we may have missed much of the change in 1988 due to the fact that we did not collect until several clonal generations had passed. Figure 2 also shows a distinct absence of very small clones in the early 1988 sample compared with the early 1989 sample, which is consistent with the hypothesis that they had already been eliminated from the population by the time the early 1988 collection was made.
TABLE 4. Genetic correlations between overall size and different shape variables for early 1989 clones. Overall size was measured as the geometric mean of all other variables. Shape variables were measured as in the text. All variables were log transformed. Variables are listed in descending order of heritability. Bonferroni-corrected P-values for test that product-moment correlation is different from 0.
Character correlated Variance component Clone mean with size correlation correlation
Siphunculus length 0.866 0.732(**) Hind tibia length 0.825 0.712(**) Head length -0.881 -0.641(*) Hind femur length 0.628 0.683(**) Apical rostrum length -0.734 -0.796(**) Head width -0.833 -0.763(**) Body length -0.426 -0.034 Coxa width -1.016 -0.533(*) Hind tarsus length -0.162 -0.427 Siphunculus width 0.196 0.241 Stylet length -0.610 -0.612(**) Body width -0.495 -0.001
* P [less than] 0.05; ** P [less than] 0.01.
[TABULAR DATA FOR TABLE 5 OMITTED]
To determine which of the shape variables changed significantly during the 1989 season, univariate tests of the same model as in Table 5 were performed (Table 6). We found a mixed pattern of changes through the season in the shape variables. Increases were seen for hind femur length and siphunculus width relative to overall size, whereas relative head length, relative head width, and relative width of the body at the coxae decreased significantly through the season. Examination of the pattern of change in the entire set of shape variables in Table 6 suggests that, relative to their overall increased size, aphids collected at the end of the season tended to be longer and thinner and have longer legs and smaller heads than did aphids collected at the beginning of the season.
This is a rather complex pattern of shape change. Although it might be tempting to hypothesize that each of these changes is the direct result of natural selection, a simpler hypothesis is that the shape changes might just be correlated responses to selection on overall body size alone. When we compared the sign of the observed change in the mean of each variable with the sign of the correlated response expected from the genetic correlation early in the season between that variable and overall size (Table 6), we found that all of the changes were consistent in direction with those expected as correlated responses to selection for increased body size.
Natural Selection on Overall Size
For the observed changes in shape to be correlated responses to selection on body size there must be not only the appropriate pattern of genetic correlations, but also a relationship between overall size and fitness. To evaluate whether overall size might be under selection in this aphid population, we performed a regression analysis (e.g., Lande and Arnold 1983), using two fitness components and size and shape measures for a small subset of early 1989 clones reared under field conditions. We found significant standardized directional selection gradients on overall size with respect to both age at first reproduction ([Alpha]) and the fecundity in the first 12 d of adulthood ([[Beta].sub.alpha] = -0.12, P [less than] 0.0001; [[Beta].sub.fec12] = 0.26, P [less than] 0.0001, df = 14, [ILLUSTRATION FOR FIGURE 3 OMITTED]). These standardized selection gradients mean that an increase of body size of one phenotypic standard deviation would be expected to lead to a 12% decrease in age at first reproduction and a 26% increase in fecundity during the first 12 d of adulthood. In a species with broadly overlapping generations during the parthenogenetic phase, such changes could greatly enhance lifetime fitness. To put the change in overall size that we observed in 1989 (0.037) into perspective, it was just slightly larger than l phenotypic standard deviation (0.034) and thus could have been associated with quite a large fitness increase.
When we regressed the two fitness components on overall size plus the three shape variables that changed the most during the season (head length, hind femur length, and head width relative to body size), we found virtually identical values for the selection gradients on overall size as we had seen in the univariate model ([b.sub.alpha] = -0.10, [[Beta].sub.fec12] = 0.25). However, neither selection gradient was significant in the larger model due to the small sample size. Notably, the selection gradients for the shape variables in this analysis were three to 10 times smaller than the estimated selection on size, adding credence to our hypothesis that selection is potentially acting primarily on overall size.
The quadratic term in the regression of relative age at first reproduction on size was highly significant, suggesting a convex relationship between size and developmental time [ILLUSTRATION FOR FIGURE 3 OMITTED]. Observations of individuals with larger body sizes will be required to see whether age at first reproduction simply plateaus at larger sizes or actually increases, indicating stabilizing selection. The quadratic term in the regression of fecundity on size was not significant [ILLUSTRATION FOR FIGURE 3 OMITTED].
Between-Year Change in Overall Size: Genetic Slippage and a Cost of Recombination
On average, clones in the collections made late in the summer of 1988 were larger than were the clones in the early 1989 collection [ILLUSTRATION FOR FIGURE 2 OMITTED]. The late 1988 and early 1989 collections were separated by the yearly cycle of sexual reproduction, suggesting that the shift back to smaller body size at the beginning of the season may represent a cost of recombination. Unfortunately, the early 1988 clones are not useful for a test of slippage, because we have no data on the average size of aphids the previous fall.
We created shape variables by correcting each length variable for a measure of overall size (the geometric mean of all the variables). Thus, our measures of morphometric variation concern overall size and a collection of lengths of various body parts relative to overall size. In collections of pea aphids from a single field, we found considerable genetic variability both for overall size and for several of the shape variables (Table 3). The shape characters that were the most genetically variable showed a mixture of rather high positive and high negative genetic correlations with our measure of overall size (Table 4), leading to the expectation that the response to clonal selection on size and shape might be quite complex.
We found a significant change between the mean multivariate phenotype of aphid clones collected in the same field early and late in the summer of 1989. About 40% of this change was due to an increase in overall size, with the remainder attributable to changes in shape. We hypothesize that the change in overall size during the season reflects a response to a direct force of clonal selection on body size. This hypothesis is supported by our measures of selection on overall size [ILLUSTRATION FOR FIGURE 3 OMITTED] and by the observation that all of the shape changes were consistent in direction with what was predicted for correlated responses to selection on size alone (Table 6).
[TABULAR DATA FOR TABLE 6 OMITTED]
A genetic change in the average phenotype through time could result from forces other than natural selection. However, in view of our estimates of selection and various details of our experimental system and design, we believe that the major alternative explanations for the observed phenotypic change are not a problem in our study:
(1) In small populations, genetic drift, or random fluctuations in gene frequency due to sampling error, can alter the mean phenotype (Wright 1980; Endler 1986). Although we have no direct estimates of population size, several indirect lines of evidence suggest that pea aphid populations are not small. Samples from fields late in the season, near the time of mating, reveal a large number of different electrophoretic genotypes (S. Via unpubl. data). Moreover, after overwintering, fundatrices (each of which is potentially a unique genotype) were readily collected throughout the 15-acre study field in 1989, again suggesting that the genetic size of this aphid population was not very small. We therefore expect the selection that we estimated to be a more potent force than drift in altering the population mean.
(2) Biases in sampling could produce a significant difference in the morphological phenotypes of the early and late collections if the methods of collecting were inconsistent between the two time periods. Again, we believe this is unlikely to be the case, because great caution was taken to obtain samples across the entire field during both the early and late collections, and the same sampling methods were used both times.
(3) Although some migration occurred after the first sample was taken, migration is unlikely to have produced the pronounced phenotypic changes that we observed for the following reasons. Migrants can come only from one of two crop types in this region, alfalfa or clover, and a preliminary study of gene flow in this system suggests that migrants to alfalfa fields are nearly 10 times more likely to come from other alfalfa fields than they are to originate from a clover field (S. Via, unpubl. data). Because migrants from alfalfa are likely to have been subject to selection pressures that were similar to those in the collection field, they can be expected to be morphologically similar. Alternatively, the few migrants that may have come from clover fields can be expected to be much less fit than residents (Via 1989, 1991), and they are thus unlikely to represent a very large fraction of the population at the time of the second collection, which was taken several generations later. Finally, although it is possible that some migrants may have come from wild hosts, the acreage planted to alfalfa and clover in the study area is so great that it is unlikely that a very large percentage of the migrant pool comes from wild hosts. For these reasons, the selection that we estimated is likely to overwhelm any change in phenotype due to migration from other crops.
(4) Given our design, nutritional differences in plant quality early and late in the season cannot explain the observed changes in the mean phenotype. Although clones were collected at different times during the season, they were all established in individual laboratory cultures on greenhouse plants and were maintained under relatively uniform conditions. The clonal replicates that were used for morphological measurements were then sampled from the laboratory cultures at approximately the same time. Because measurements were not taken on the field-collected individuals, environmentally based variation (phenotypic plasticity) arising from differences in nutritional quality of field plants during the season cannot be an explanation for differences between early and late collections. In addition, the formation of two sublines from each clone ensured that within-clone environmental variation was taken into consideration when testing for the significance of higher effects, such as a difference between clones or collection times.
Thus, natural selection appears to be a reasonable mechanism for the change in the mean morphological phenotype that we observed in the 1989 collections. The nonsignificant trend toward increased body size in 1988 actually substantiates rather than weakens our hypothesis that the phenotypic change in 1989 was caused by selection. Because the early 1988 collection was made so much later than the early 1989 collection, we might have expected that the difference between the early and late groups would be in the same direction but smaller for the 1988 collections than for those made in 1989 if natural selection were responsible for the phenotypic change. This is what we saw. In view of the fact that clonal selection is expected to change the mean phenotype exponentially (Charlesworth 1980), it is not surprising that much of the change might have been missed in 1989 by waiting several generations to collect the early group. In addition, the mean overall size of individuals in the late 1988 collection (August) was slightly larger (and obtained later) than that in late 1989 (July, [ILLUSTRATION FOR FIGURE 2 OMITTED]), which is also consistent with the hypothesis that selection acted to increase body size over the summer months.
During the season, some of the shape variables increased in mean value, whereas others decreased. In general, relative to overall size, aphids after a period of selection were larger and thinner, with longer legs and a relatively shorter head and apical rostral segment. Because of a complex network of genetic correlations between size and shape (Table 4), as well as among the shape variables (Shaw 1990), it seems unlikely that direct forces of selection on each character are responsible for the observed changes: this would require postulating a rather complicated pattern of multivariate selection. Although we cannot completely eliminate the possibility that selection also acts directly on some components of shape, our data are fully consistent with the simple hypothesis that selection to increase overall size resulted in the complex pattern of shape changes through correlated responses. We observed significant selection on body size, and all of the observed phenotypic changes in the shape variables were in the direction predicted for correlated responses to selection on body size alone (Table 6). Unfortunately, a more quantitative prediction of the magnitude of change expected for each variable when genetic parameters are estimated with error is not a simple problem (McCulloch et al. 1996) and requires more precise estimates of genetic correlations and selection gradients than are available in this study.
In the selection analysis, we saw that larger overall size was associated with increased early fecundity. Positive associations between size and fecundity are also seen in Daphnia (Lynch and Spitze, 1994) and in a host of other insects (Robertson 1957; Parker 1970; Juliano 1985).
Roff (1981) and Wilkinson (1987) have suggested that correlations between body size and other life-history characters, such as developmental time, may produce selection against large body size. This pattern of opposing forces of selection mediated by correlations between body size and life history-traits led Roff (1981) to postulate that optimizing selection on body size might be widespread. We saw a strong positive association of size and developmental time in our selection analysis, suggesting the action of directional selection on size. The issue of stabilizing selection is less clear in our study. The significance of the quadratic term in our regression analysis of age at first reproduction on size [ILLUSTRATION FOR FIGURE 3 OMITTED] suggests that we might have seen an increase in development time at larger body sizes (causing a decline in fitness) had we possessed a larger sample. However, we cannot currently eliminate the alternative possibility that age at first reproduction simply remains at a constant low value at larger sizes.
Further regressions of individual fitness on the vector of size and shape variables (Lande and Arnold 1983) and analyses of selection at different phases of the life cycle (Schluter et al. 1991) would be invaluable for understanding why larger size may be favored in pea aphids during the summer months and for revealing the extent to which selection may also be acting on components of shape in the ways suggested by previous workers (e.g., Kennedy 1986, Moran 1986). In addition, by clarifying the morphological targets of direct selection, a larger phenotypic selection study would suggest whether any of the genetic correlations estimated here might act to constrain morphological evolution in pea aphids. Manipulative experiments (e.g., Wade and Kalisz 1990) could also be performed to try to distinguish among possible agents of selection, such as temperature, surface characteristics of the host plants, or presence of natural enemies.
It is notable that the mean overall size of the population was smaller at the beginning of 1989 than at the end of 1988 [ILLUSTRATION FOR FIGURE 2 OMITTED]. These two collections were separated by the sexual phase and overwintering of the eggs. The observed change in overall size away from the apparent direction of selection may be attributable to the bout of recombination that occurs during the sexual phase each fall, providing an example of a cost of recombination or "genetic slippage" (Lynch and Deng 1994). The expected magnitude of the genetic slippage after sex is directly proportional to the fraction of the clonal heritability (i.e., the total genetic variance) that is due to dominance or epistatic variance (Lynch and Deng 1994). In Daphnia, Lynch and Deng found slippage in the mean phenotype of several life-history traits on the order of 10% of a phenotypic standard deviation. However, the change toward smaller size that we observed at the start of the 1989 season was much larger than this (1.2 phenotypic standard deviations). Although the upper limit to slippage has not been formally estimated, a loss of almost the entire response to selection may be possible if most of the clonal heritability is due to nonadditive genetic variance (M. Lynch, pers. comm. 1994). Alternatively, we cannot currently eliminate the possibility that selection on overall size may have changed direction between the time of the late collection (July through August) and the end of the season (November), favoring smaller clones and driving the mean back toward the value observed in the spring. To distinguish genetic slippage from antagonistic selection as a cause of the decline in body size between years, experimental estimates of nonadditive genetic variance and further studies of temporal patterns of selection will be required.
Most previous examples of rapid evolution are associated with environmental changes in obvious selective agents such as the concentrations of heavy metals or the density of predators. However, yearly bouts of short-term evolution of the type documented here might also be quite prevalent, particularly in cyclical parthenogens that may be exposed to antagonistic forces of selection at different time periods during the clonal phase or in which the disruption of favorable genotypes caused by genetic slippage during recombination may regularly move populations away from a selectively advantageous phenotype.
We are grateful to C. E. McCulloch, A. R. McCune, J. Conner, and R. Lande for insightful discussions of this work. J. Conner, D. Hawthorne, and two anonymous reviewers provided useful comments on the manuscript. We also thank M. Lynch for sharing his manuscript on genetic slippage with us. Funding was provided by a Searle Scholar's Award (Chicago Community Trust), United States Department of Agriculture Competitive Grants 87-CRCR-1-2375 and 88-35713-3473, Hatch Project New York Cornell 139419, and National Science Foundation grant DEB-9207573.
Antonovics, J., A. D. Bradshaw, and R. G. Turner. 1971. Heavy metal tolerance in plants. Advances in Ecological Research 7:1-85.
Blackman, R. 1979. Stability and variation in aphid clonal lineages. Biological Journal of the Linnean Society 11:259-277.
Boag, P. T., and P. R. Grant. 1981. Intense natural selection in a population of Darwin's finches (Geospizinae) in the Galapagos. Science 214:82-85.
Charlesworth, B. 1980. Evolution in age-structured populations. Cambridge University Press, New York.
Carroll, S. P., and C. Boyd. 1992. Host race radiation in the soap-berry bug: Natural history with the history. Evolution 46:1052-1069.
Darroch, J. N., and J. E. Mosimann. 1985. Canonical and principle components of shape. Biometrika 72:241-252.
Dickerson, G. E. 1955. Genetic slippage in response to selection for multiple objectives. Cold Spring Harbor Symposium in Quantitative Biology 20:25-32.
Dixon, A. F. G. 1987. Aphid reproductive tactics. Pp. 3-18 in J. Holman, J. Pelikan, A. F. G. Dixon, and L. Weismann, eds. Population structure, genetics, and taxonomy of aphids and thysanoptera. SPB Academic Publishing, The Hague, Netherlands.
Endler, J. 1980. Natural selection on color patterns in Poecilia reticulata. Evolution 34:76-91.
Falconer, D. S. 1989. Introduction to quantitative genetics, 3d ed. Longman, New York.
Hairston, N. G., Jr., and T. A. Dillon. 1990. Fluctuating selection and response in a population of freshwater copepods. Evolution 44:1796-1805.
Heie, O. E. 1987. Morphological structures and adaptations. Pp. 393-400 in A. K. Minks and P. Harrewijn, eds. Aphids: Their biology, natural enemies and control. Vol. 2A Elsevier Press, New York.
Juliano, S. A. 1985. The effects of body size on mating and reproduction in Brachinus lateralis (Coleoptera: Carabidae). Ecological Entomology 10:271-280.
Kennedy, C. E. J. 1986. Attachment may be a basis for specialization in oak aphids. Ecological Entomology 11:291-300.
Lamb, R. J., and P. J. Pointing. 1972. Sexual morph determination in the aphid, Acyrthosiphon pisum. Journal of Insect Physiology 18:2029-2042.
Lande, R. 1979. Quantitative genetic analysis of multivariate evolution applied to brain:body size allometry. Evolution 33:402-416.
Lande, R., and S. J. Arnold. 1983. The measurement of selection on correlated characters. Evolution 37:1210-1226.
Lynch, M. 1985. Spontaneous mutations for life-history characters in an obligate parthenogen. Evolution 38:804-818.
Lynch, M., and H.-W. Deng. 1994. Genetic slippage in response to sex. American Naturalist 144:242-261.
Lynch, M., and W. Gabriel. 1983. Phenotypic evolution and parthenogenesis. American Naturalist. 122:745-764.
Lynch, M., and K. Spitze. 1994. Evolutionary genetics of Daphnia. Pp. 109-128 in L. Real, ed. Ecological genetics. Princeton University Press, Princeton, NJ.
McCulloch, C. E., M. D. Boudreau, and S. Via. 1996. Confidence regions for evolutionary trajectories. Biometrics. In press.
Moran, N. 1986. Morphological adaptation to host plants in Uroleucon (Homoptera: Aphididae). Evolution 40:1044-1050.
Moran, N. A. 1987. Evolutionary determinants of host specificity in Uroleucon. Pp. 29-38 in J. Holman, J. Pelikan, A. F. G. Dixon, and L. Weismann, eds. Population structure, genetics, and taxonomy of aphids and thysanoptera. SPB Academic Publishing, The Hague, Netherlands.
Mosimann, J. E., and F. C. James. 1979. New statistical methods for allometry with application to Florida red-winged blackbirds. Evolution 33:444-459.
Parker, G. A. 1970. Sperm competition and its evolutionary effect on copula duration in the fly Scatophaga stercoraria. Journal of Insect Physiology 16:13-1-1328.
Rice, W. R. 1989. Analyzing tables of statistical tests. Evolution 43:223-225.
Robertson, F. W. 1957. Studies on quantitative inheritance. XI. Genetic and environmental correlation between body size and egg production in Drosophila melanogaster. Journal of Genetics 55:428-443.
Roff, D. 1981. On being the right size. American Naturalist 118:405-422.
SAS Institute, Inc. 1985. SAS user's guide: Statistics. Version 5 ed. SAS Institute, Cary, NC.
Schluter, D. 1984. Morphological and phylogenetic relations among the Darwin's finches. Evolution 38:921-930.
Schluter, D., T. D. Price, and L. Rowe. 1991. Conflicting selection pressures and life history tradeoffs. Proceedings of the Royal Society B 246:11-17.
Seeley, R. H. 1986. Intense natural selection caused a rapid morphological transition in a living marine snail. Proceedings of the National Academy of Sciences 83:6897-6901.
Shaw, A. J. 1990. Genetic variation and natural selection for morphological characters in field populations of the pea aphid, Acrythosiphon pisum. Master's thesis. Cornell University, Ithaca, NY.
Southwood, R. 1986. Plant surfaces and insects - an overview. Pp. 1-22 in B. Juniper and R. Southwood, eds. Insects and the plant surface. Edward Arnold, London.
Stalker, H. D., and H. L. Carson. 1947. Morphological variation in natural populations of Drosophila robusta Sturtevant. Evolution 1:237-248.
Via, S. 1984. The quantitative genetics of polyphagy in an insect herbivore. I. Genotype-environment interaction in larval performance on different host plant species. Evolution 38:881-895.
-----. 1989. Field estimation of variation in host plant use between local populations of pea aphids from two crops. Ecological Entomology 14:357-364.
-----. 1991. The population structure of fitness in a spatial patchwork: Demography of pea aphid clones on two crops in a reciprocal transplant. Evolution 45:827-852.
-----. 1992. Inducing the sexual forms and hatching the eggs of pea aphids. Entomologia Experimentalis et Applicata. 65:119-127.
Wade, M. J., and S. Kalisz. 1990. The causes of natural selection. Evolution 44:1947-1955.
Wilkinson, G. S. 1987. Equilibrium analysis of sexual selection in Drosophila melanogaster. Evolution 41:11-21.
Wright, G. A. 1980. Genic and organismic selection. Evolution 34: 825-843.
|Printer friendly Cite/link Email Feedback|
|Author:||Via, Sara; Shaw, Alison J.|
|Date:||Feb 1, 1996|
|Previous Article:||A trade-off for host plant utilization in the black bean aphid, Aphis fabae.|
|Next Article:||Evolutionary loss of larval feeding: development, form and function in a facultatively feeding larva, Brisaster latifrons.|