Printer Friendly

Experimental evolution of resistance in Brassica rapa: correlated response of tolerance in lines selected for glucosinolate content.

Since Raven and Ehrlich (1964) advanced their theory of coevolution between plants and their insect herbivores, two major goals of plant-insect evolutionary ecology have been to determine, first, whether insects impose significant selection on plant populations and, second, if plant populations have the ability to respond to herbivore-imposed selection. Toward this aim numerous investigators have found that insects impose significant phenotypic selection on plant populations for defense (Berenbaum et al. 1986; Rausher and Simms 1989; Simms and Rausher 1989; Marquis 1990, 1992; Simms and Fritz 1990; Nunez-Farfan and Dirzo 1994; Sagers and Coley 1995; Pilson 1996). Further, significant genetic variation in specific defense traits as well as in overall defense have often been detected (Berenbaum et al. 1986; Fritz and Price 1988; Rausher and Simms 1989; Simms and Rausher 1989; Fritz 1990; Marquis 1990; Smith 1990; Berenbaum and Zangerl 1992; Sork et al. 1993; Stowe et al. 1994). The combination of these two results show that insect herbivores impose selection for defense and that plant populations have the potential to respond to this selection.

If selection for defense and additive genetic variation in defense were the only elements significantly affecting the evolution of defense, these studies would suggest that plant populations should evolve levels of defense that would allow them to completely escape herbivore damage. However, defense levels within plant populations are often variable and few individuals, and even fewer populations, escape damage completely (Berenbaum et al. 1986; Fritz and Price 1988; Rausher and Simms 1989; Simms and Rausher 1989; Fritz 1990; Marquis 1990; Smith 1990; Berenbaum and Zangerl 1992; Sork et al. 1993; Stowe et al. 1994). One possible factor that may prevent the evolution of levels of defense that allow individuals to escape herbivore damage is a trade-off between defense and some other important trait(s).

The cornerstone of both conceptual (Feeny 1976; Rhoades and Cates 1976; McKey 1979; Coley et al. 1985; Herms and Mattson 1992) and mathematical models (Simms and Rausher 1987; Adler and Karban 1994) concerning the evolution of plant defense against herbivores is the implicit or explicit assumption that the evolution of defense is constrained by trade-offs with other traits, that is, defense is costly. These costs would yield a stabilizing selection gradient on defense, resulting in the maintenance of an intermediate value of defense in plant populations where the difference between the benefit and cost of defense is maximized (Simms and Rausher 1987; Simms 1992). While these models predict the pattern of defense typically seen in nature, their predictions are only as robust as their assumptions. That is, if a cost of defense does not exist, these models yield erroneous predictions.

Costs of defense are typically envisaged as allocation costs (Simms and Rausher 1987; Simms 1992), manifested as lower fitness of more highly defend individuals growing in an herbivore-free environment (Simms and Rausher 1987; Simms 1992; Fineblum and Rausher 1995; Stowe 1997; Stowe and Marquis, unpubl. data). Allocation costs would result in a negative genetic correlation between fitness and the level of defense in an herbivore-free environment (Simms and Rausher 1987) due to the allocation of limited resources to defense, rather than to other fitness-enhancing traits (Simms 1992). Because allocation costs are the way costs of defense are typically conceived, this is the type of cost most often investigated (Simms 1992; Fineblum and Rausher 1995). However, though many researchers examining a diverse group of species have attempted to detect allocation costs of defense, they have met with varying results (Simms and Rausher 1987, 1989; Rausher and Simms 1989; Agren and Schemske 1993, 1994; Sagers and Coley 1995; Stowe 1997; Stowe and Marquis, unpubl. data; for review, see Simms 1992).

Whereas results from studies attempting to detect allocation costs of defense in terms of decreased fitness in the absence of herbivores have been equivocal, other types of costs have been virtually ignored (Simms [1992] and references therein; but see Berenbaum et al. 1986; Fineblum and Rausher 1995; Mauricio et al. 1997). Yet, costs of defense have also been hypothesized to occur when other ecological factors are considered and can be called ecological costs (Simms 1992). Such costs would result in negative genetic correlations between defense and other resistance and/or ecologically important traits. For example, Berenbaum et al. (1986) found that the production of various furanocoumarins in Pastinaca sativa were negatively genetically correlated. Thus, an evolutionary increase in production of any specific furanocoumarin would be constrained by selection for increased production of others. In addition, Sork et al. (1993) found that in Quercus rubra the evolution of the level of defense expressed in one environment may be constrained by the level of defense expressed in other environments. These two examples show that to accurately predict the response of plant populations to selection imposed by insect herbivores one must examine ecological as well as allocation costs of defense.

Although it is important to investigate both allocation and ecological costs, the distinction between the terms can be unclear because both types of costs may be the result of allocating limited resources to defense rather than some other function. For example, the cost of defense in P. sativa detected by Berenbaum et al. (1986) was due to the fact that resources allocated to the production of one furanocoumarin necessarily decreased those available for the production of others. Thus, this ecological cost is the result of the allocation patterns inherent to the plant's physiology. In contrast, Sork et al. (1993) describe what could be called an ecological cost of defense in Q. rubra that may or may not be due to allocation patterns. Thus, to alleviate any confusion as to the cause of any specific cost of defense, I propose two new terms to replace "allocation" and "ecological" that should prove more useful: "internal" and "external" costs of defense. Internal costs of defense are solely due to the internal resource-allocation patterns inherent in the physiology of the plant. As such they can be detected as decreased fitness of highly defended individuals in the absence of herbivores. Further, internal costs can be detected without taking any other ecological factor(s) into consideration. Thus, internal costs are those that have been termed allocation costs. In contrast, while external costs of defense may also be due to resource allocation patterns, to detect external costs some additional ecological factor, external to the plant, must be considered. In addition, external costs may not be incurred by the plant if that ecological factor is absent. For example, in Trifolium repens a cost of cyanogenesis in terms of a decrease in frost tolerance (Daday 1954a,b, 1965) would not be detected if the external ecological factor, ambient air temperature, were ignored and this cost would not be incurred in climates in which frost is rare or absent.

One type of external cost that has recently received attention is the relationship between the two components of resistance, defense and tolerance (Simms and Triplett 1994; Fineblum and Rausher 1995; Mauricio et al. 1997), or a tradeoff between the ability to avoid herbivore damage, defense, and the ability to maintain fitness despite herbivore damage, tolerance (Painter 1958). Although a trade-off between defense and tolerance has been hypothesized (van der Meijdan et al. 1988), the evolution of defense and/or tolerance are not necessarily mutually exclusive responses to selection imposed by herbivores for resistance (Smith 1990; Mauricio et al. 1997; but see Fineblum and Rausher 1995). Van der Meijdan et al. (1988) first hypothesized the existence of such a trade-off between defense and tolerance to explain the interspecific observation that species that suffer high levels of damage (low defense) had a high regrowth ability (high tolerance), while those species with high defense (incurring little damage) had low regrowth potential (low tolerance).

While this interspecific pattern is suggestive of a genetic trade-off between the components of resistance (defense and tolerance), concrete evidence for such a trade-off requires intraspecific studies. This is because interspecific correlations do not negate the possibility that different plant species may have undergone different selection histories. For example, those species that have high defense and low tolerance may have been selected for such traits. Conversely, when other species are considered, selection may have favored high tolerance and low defense. Thus, to determine if the macroevolutionary pattern of a negative correlation between defense and tolerance is determined by microevolutionary relationships between defense and tolerance, one must establish whether a genetically based trade-off exists between defense and tolerance within a species. Recently, Fineblum and Rausher (1995) demonstrated a negative genetic correlation between defense and tolerance to insects that cause apical meristem damage in Ipomoea purpurea. This result suggests the possibility that the evolution of resistance may be constrained by genetic trade-offs between defense and tolerance, but the generality of this phenomenon requires more research.

The purpose of this study was to investigate constraints on the evolution of resistance due to an external cost of defense, that is, a genetic trade-off between the chemical defense trait, glucosinolate production, and tolerance to herbivore damage. A rapid cycling variety of Brassica rapa (Williams and Hill 1986) was used as a model. To investigate the existence of such a trade-off, I examined variation in the response to artificial damage (tolerance) of populations of individuals that had been artificially selected for foliar glucosinolate content (Stowe 1997; Stowe and Marquis, unpubl. data) and differed in levels of realized defense (Stowe 1997; Stowe, in press). If a genetic trade-off between glucosinolate production and tolerance exists in this species, we would expect that those individuals selected for increased glucosinolate content would have a greater decrease in fitness in response to damage (express lower tolerance) than those selected for decreased glucosinolate content. Such an external cost of defense would result in a statistically significant selection line-by-damage treatment interaction as well as a correlated response of tolerance in the opposite direction of the direct response of foliar glucosinolate content. These two conditions would demonstrate a cost of defense in terms of tolerance in B. rapa.

Both variation or similarity in tolerance among selection lines may be due to variation or similarities in the response of specific fitness components to damage. For example, all selection lines may respond to damage by decreasing flower production, but not other fitness components (fruit or seed production). If high-defense lines exhibit a greater decrease in flower production than low-defense lines, high-defense lines would be less tolerant. Conversely, if high-defense lines respond to damage by solely decreasing the probability of a flower producing a fruit, whereas low-defense lines respond to damage only by decreasing flower production, then both lines could express similar levels of total fitness, tolerance. To determine which fitness components respond to damage and may contribute to the level of tolerance expressed, I examined the response to damage of: (1) number of flowers produced; (2) number of fruits produced per pollinated flower and; (3) number of seeds produced per fruit. Thus, this study answers two related questions: (1) Is there a genetic tradeoff between defense and tolerance? (2) If so, which components of fitness (flower production, fruit production, or seed production) contribute to the observed relationship between defense and tolerance? Artificial selection for foliar glucosinolate content did not result in a correlated response in seed weight (Stowe 1997; Stowe and Marquis, unpubl. data), so I did not consider this factor when examining the correlated response of tolerance.

METHODS

Study Species

Brassica rapa (syn. B. campestris: Brassicaceae; wild mustard) is an annual plant, native to Eurasia, that now grows worldwide in naturalized populations and as cultivars. Rapid cycling varieties (RCB) of B. rapa were developed at the University of Wisconsin by selecting for short generation time (approximately two months), lack of seed dormancy, and ability to grow under fluorescent lights (Williams and Hill 1986). As such, RCBs can be grown easily in the laboratory and a number of generations of selection can be carried out in a relatively short period of time. Glucosinolates, or mustard oil glycosides, are one of the chemical defenses of B. rapa (Louda and Mole 1991) and are ubiquitous throughout the plant family Brassicaceae (Rodman 1991). Glucosinolates have typically been classified as a qualitative defense (Feeny 1976), meaning that they can repel nonadapted insect herbivores at low concentrations. The seeds used for this study were obtained from the Crucifer Genetics Cooperative at the University of Wisconsin, Madison (B. rapa, CRGC #1-1 Aaa). Previous to this experiment, high-defense, control, and low-defense selection lines of B. rapa had been divergently selected for foliar glucosinolate content of the first true leaf 14 days after emergence, which resulted in significantly different levels of foliar glucosinolates ([Mathematical Expression Omitted], [Mathematical Expression Omitted] [Mathematical Expression Omitted]; Stowe 1997; Stowe and Marquis, unpubl. data). Further, high- and low-glucosinolate selection lines differed in the level of percentage leaf area damaged by the herbivores Pieris rapae and Trichoplusia ni in choice trials in the laboratory ([Mathematical Expression Omitted] damage, [Mathematical Expression Omitted] damage, and [Mathematical Expression Omitted] damage, [Mathematical Expression Omitted] damage for P. rapae and T. ni, respectively; Stowe 1997, in press). Thus, differences among selection lines in defense against both herbivore species may be attributed to differences in glucosinolate content, or another correlated defensive trait. However, trichome number, a physical defense trait (Agren and Schemske 1993, 1994) in this species, was uncorrelated with glucosinolate production (Stowe, unpubl. data).

Selection Methods

From a base population consisting of 300 RCB B. rapa individuals, two replicate populations were derived by randomly choosing 120 plants to begin each replicate population. From each replicate population, 20 random individuals were chosen to start the control-selection-line replicates. Following this, from the 100 remaining individuals within each replicate population, the 20 individuals with the highest and lowest foliar glucosinolate content were used to generate the high- and low-glucosinolate selection lines, respectively (for more details see Stowe 1997; Stowe and Marquis, unpubl. data). This was repeated within each replicate selection line for three generations of selection. Crosses among selected plants were random, with the constraint that matings between related individuals were avoided and control selection lines were inbred as much as the high and low selection lines to control for deviations due to inbreeding. Following the initial generation, each replicate line was maintained at approximately 100 individuals. To counteract the possible effects of natural selection opposing artificial selection for glucosinolate content, new generations were started from seeds chosen such that each parental plant (whether mother or father) within a replicate selection line was represented equally. Thus, only five seeds from each of the 20 selected plants (either as mother or father) were randomly chosen to start the next generation. Following the third generation of selection, 50 individuals from each replicate selection line were grown and randomly, reciprocally crossed within their own replicate selection line, thus yielding 25 separate full-sib families to be used in the following tolerance experiments.

Plants were grown in individual 5.5 x 5.5 x 5-cm square pots on shelves in the laboratory (Stowe 1997; Stowe and Marquis, unpubl. data). They were given 24 hours of fluorescent lighting and fertilized once, two to three days after emergence, with 10 ml of All Purpose Peter's 20-20-20 solution (3.41 g/l [H.sub.2]O). Plant positions were randomized twice weekly until all leaf samples were collected and weekly thereafter, to decrease position effects. During this experiment, temperatures that plants were grown at varied from 24 [degrees] C to 32 [degrees] C, but not in a systematic manner.

Tolerance Methods

To examine the tolerance of selection lines divergently selected for foliar glucosinolate content (defense), three artificial damage treatments were imposed on plants grown for this experiment: 0% (control), 20%, and 60% damage. These levels of damage were chosen following Stewart et al. (1990), who found a linear decrease in yield in Brassicaceae cultivars at these levels of defoliation. Plants for this experiment were grown under the same conditions as those in which selection was carried out (see above; Stowe 1997; Stowe and Marquis, unpubl. data) except artificial damage treatments were imposed. Three individuals from each of the resultant 25 families in each replicate selection line were randomly chosen to be used in this experiment. From these a single, randomly chosen individual from each family was placed in each of the three damage treatments. To increase the sample size, 15 additional individuals were randomly selected from each replicate selection line and five of these were randomly placed in each of the three damage treatments. This yielded a total of 540 individuals in the experiment (30 individuals per replicate selection line per damage treatment).

Fourteen days after emergence, damage treatments were established by measuring the area of each fully expanded leaf using transparent graph paper (grid = 0.04 [cm.sup.2]) and removing the appropriate amount of leaf area from each leaf with scissors. Following initial leaf area removal, subsequent leaf area was removed from newly expanded leaves weekly for five weeks. When flowers were initiated, individuals were randomly hand-pollinated within replicate selection line-damage treatment combination. Since pollinations were conducted within replicate selection line and damage treatments, any effect that damage treatment had on pollen production and/or viability (i.e., male fitness) is implicitly included in the estimate of tolerance. Each pollinated flower was marked with a jewelry tag, so that the total number of flowers pollinated on an individual could be determined whether or not the pollination resulted in the production of a fruit. Flower production was determined by counting the number of peduncles that had actually produced flowers each week. Thus, flower counts were of total flowers produced, rather than flowers produced since the previous count. Eight weeks after emergence, when most plants were senescing, fruits were collected and the total number of flowers pollinated and the number of fruits produced were recorded for each individual. Fruits were then air dried and the seeds contained in each fruit were counted.

Fruit set was calculated as the number of fruits an individual produced divided by the number of flowers pollinated. The mean number of seeds produced per fruit was calculated as the total number of seeds produced by an individual divided by the number of fruits produced. Defining the fitness components in this way allowed me to estimate total potential female fitness of an individual by multiplying total number of flowers produced, number of fruits produced per pollinated flower, and mean number of seeds produced per fruit. There were no significant differences in the number of flowers pollinated among damage treatments (mean [+ or -] SE: 0% damage, 10.62 [+ or -] 0.30; 20% damage, 11.19 [+ or -] 0.33; and 60% damage, 10.35 [+ or -] 0.31; df = 2,6, P = 0.1761) or selection lines (mean [+ or -] SE: low, 10.89 [+ or -] 0.30; control, 10.12 [+ or -] 0.30; and high, 10.41 [+ or -] 0.31; df = 2,3, P = 0.9664), nor was there a significant correlation between the number of flowers pollinated and estimated total potential female fitness (n = 540, r = 0.06, P = 0.1688).

Statistical Analysis

To evaluate the existence of a trade-off between defense and tolerance in B. rapa, I performed a single, mixed-model ANOVA using SAS PROC GLM (SAS Institute 1985). If the interaction between selection line and damage treatment was significant I examined the direction of this statistically significant interaction. The dependent variable for this analysis was relative fitness, that is, estimated total female fitness standardized to the control selection line of each replicate. Standardization was achieved by subtracting the mean of control selection replicates in a damage treatment from each individual in their respective high and low selection replicates within that same damage treatment. The main effects of selection line (low, control, or high glucosinolate content) and damage treatment (0%, 20%, or 60% damage) were considered fixed, whereas replicate nested within selection line and any interaction with replicate was considered random. All F-tests were based on Type Ill mean squares. Since total female fitness is a composite of the three multiplicative fitness components (total female fitness = total number of flowers x number of fruits/pollinated flower x mean number of seeds/fruit), an ANOVA using the estimate of total female fitness as the dependent variable is similar to performing a MANOVA. Further statistical analyses using separate mixed-model ANOVAs were performed using each of the three fitness components (total number of flowers, fruits/pollinated flower, and mean number of seeds/fruit) as the dependent variables to determine which multiplicative fitness component(s) contributed to the observed variation or similarity among selection lines in tolerance. Each model examining individual components of fitness was constructed as described above for estimated total potential female fitness. However, total flower number was log-transformed and mean number of fruits produced per pollinated flower was square-root arcsine-transformed for analysis.

To examine the differences among means of selection lines and damage treatments for estimated total fitness and each fitness component, Tukey's test was used. Tukey's Test was chosen because each experimental treatment had equal sample size. Further, this test provides the most powerful statistical tool for examining differences among means when comparisons are unplanned (Sokal and Rolf 1981). Tests were conducted using the TUKEY option (SAS Institute 1985). Replicate within selection line was used as the error sum of squares to test for differences among selection line means. In contrast, the interaction between replicate within selection line and damage treatment was used as the error sum of squares to examine differences among damage treatments and each selection line by damage treatment combination.

RESULTS

Damage treatment had a significant overall effect on estimated total potential female fitness (Table 1). As damage treatment increased, the estimated total potential female fitness of individuals decreased [ILLUSTRATION FOR FIGURE 1A OMITTED]. Selection line also had a significant effect on the overall estimated total potential [TABULAR DATA FOR TABLE 1 OMITTED] female fitness (Table 1). Estimated total potential female fitness, across all damage treatments, was greatest in low-defense selection lines followed by control and high-defense selection lines [ILLUSTRATION FOR FIGURE 2A OMITTED]. More importantly, there was a significant selection line-by-damage treatment interaction (Table 1). Further, this interaction was in the pattern demonstrating a cost of defense in terms of decreased tolerance in B. rapa [ILLUSTRATION FOR FIGURE 3 OMITTED].

Damage treatment had a significant overall effect on total number of flowers produced (Table 1), with the mean number of flowers produced across all lines decreasing as damage increased [ILLUSTRATION FOR FIGURE 1B OMITTED]. However, neither the overall effect of selection line on flower production (Table 1, [ILLUSTRATION FOR FIGURE 2B OMITTED]), nor the interaction between selection line and damage treatment (Table 1, [ILLUSTRATION FOR FIGURE 4A OMITTED]) were statistically significant.

Damage treatment had a significant overall effect on mean number of fruits produced per pollinated flower (Table 1) with those individuals in the 0% damage treatment producing the most fruits per pollinated flower, followed by the 20% and the 60% damage treatments [ILLUSTRATION FOR FIGURE 1C OMITTED]. Similarly, selection line had a significant overall effect on this fitness component (Table 1, [ILLUSTRATION FOR FIGURE 2C OMITTED]). In contrast to total flower production, mean number of fruits produced per pollinated flower resulted in a significant selection line-by-damage treatment interaction (Table 1). Further, the pattern of this interaction is in the direction expected if a trade-off between defense and tolerance in terms of fruit production exists. That is, low-defense selection lines decreased fruit production less than control and high-defense selection lines in response to damage [ILLUSTRATION FOR FIGURE 4B OMITTED].

Damage treatment had a significant overall effect on mean number of seeds produced per fruit (Table 1, [ILLUSTRATION FOR FIGURE 1D OMITTED]). Selection line also had a significant effect on this fitness component (Table 1) with low-defense selection lines producing more seeds per fruit than control selection lines, followed by high-defense selection lines [ILLUSTRATION FOR FIGURE 2D OMITTED]. There was also a significant interaction between selection line and damage treatment for mean number of seeds produced per fruit (Table 1), which supports the existence of a negative genetic correlation between defense and tolerance in terms of seed production [ILLUSTRATION FOR FIGURE 4C OMITTED].

DISCUSSION

Typically, tolerance is analyzed as the variation in the slope of the full/half-sib family regression of natural damage levels on fitness (Simms and Triplett 1994; Mauricio et al. 1997; D. Pilson, pers. comm.; but see Fineblum and Rausher 1995). However, using natural damage does not allow one to estimate fitness in the absence of damage unless some individuals remain free of damage (Simms and Triplett 1994). Thus, using natural damage levels prevents the estimation of the true value of tolerance. Further, to reliably examine the relationship between defense and tolerance, experimental measurements of both of these resistance components must be divorced to avoid the confounding effects of defense on tolerance and vice versa. For example, when natural damage levels are used to estimate both defense and tolerance, highly defended individuals will naturally suffer less damage, while poorly defended individuals will incur more damage. Yet, if highly defended individuals tolerate damage poorly and those with low defense are very tolerant, both may have the same resultant fitness. Thus, using natural damage obscures the resolution of the relationship between defense and tolerance. Employing artificial damage allows one to achieve an accurate estimate of tolerance by assuring that individuals in each damage treatment incur the same amount of damage regardless of their level of defense.

Further, if defense is measured as 1 - percentage leaf area damaged on a clonal or full/half-sib family, as is often done (Rausher and Simms 1989; Simms and Rausher 1989; Marquis 1990; Sork et al. 1993; Stowe et al. 1994), it is likely a composite of a number of traits. This measure of defense can obscure negative genetic correlations both among specific defense traits (e.g., Berenbaum et al. 1986) and between defense and tolerance (Simms and Triplett 1994), which is also likely a composite of a number of traits. For example, if a population contains two variable traits that confer defense against insect herbivores, yet one of these traits is positively correlated with tolerance while the other is negatively correlated with tolerance [ILLUSTRATION FOR FIGURE 5A OMITTED], then the detection of a negative correlation between defense (measured as 1 minus percentage damage) and tolerance would be hindered [ILLUSTRATION FOR FIGURE 5B OMITTED]. While this graphical model is only one possible scenario, in that it assumes that both defense traits act in an additive manner to determine the level of tolerance expressed even if they act epistatically to determine tolerance, the true relationship between specific defense and tolerance traits will still be obscured. Nonadditivity would yield a concave-up or concave-down relationship between damage and tolerance. Since the detection of a positive and/or negative relationship between damage and tolerance depends on linear models, if the relationship between damage and tolerance were curvilinear, results would depend on the extent of natural variation in percentage leaf area damage in the population. For example, if the relationship between damage and tolerance were concave down and the natural variation in herbivore damage only spanned between 1% and 10% the researcher would conclude that no trade-off between defense and tolerance exists. Conversely, if the researcher found natural variation between 60% and 75% damage, they would conclude that there was a cost of defense in terms of tolerance. Further, even if the natural variation in damage is large, if quadratic models are not examined, a relationship between defense (1 minus percentage damage) and tolerance may not be found. Thus, when attempting to ascertain whether a negative genetic correlation exists between defense and tolerance, it is important to specify the defense trait under consideration as well as to divorce measurements of defense and tolerance. To avoid such difficulties, this study investigated the genetic trade-off between the specific chemical defense trait, foliar glucosinolate content, and tolerance to artificial damage.

Results presented here demonstrate the existence of a genetic trade-off between glucosinolate production (defense) and the ability to maintain fitness despite the presence of varying levels of damage (tolerance) in B. rapa. Thus, the response of populations of B. rapa to selection imposed by insect herbivores for increased defense may be constrained by its negative genetic correlation with tolerance and vice versa. This result is similar to that found in I. purpurea (Fineblum and Rausher 1995) where apical meristem damage was considered, rather than foliar damage.

In I. purpurea, Fineblum and Rausher (1995) found a significant genetic trade-off between defense and tolerance only in terms of fruit (capsule) production in response to damage. In contrast, this study found that a genetic trade-off exists between defense and tolerance in terms of fruit production as well as seed production in B. rapa. Specifically in B. rapa, the external cost of defense in terms of tolerance is due to the greater ability of plants from lines selected for decreased foliar glucosinolate content (low defense) to maintain fruit and seed production despite the presence of damage, compared to plants from lines selected for increased foliar glucosinolate content (high defense). However, the ability of individuals of B. rapa to maintain flower production despite damage was not affected by selection line, a result similar to that found in I. purpurea (Fineblum and Rausher 1995).

Such results suggest that the microevolutionary response of resistance, and its two components (defense and tolerance), will be constrained by the external cost of defense in terms of tolerance to damage. Further, the existence of this microevolutionary trade-off between defense and tolerance in lines of B. rapa divergently selected for defense, in conjunction with the finding that defense and tolerance are negatively genetically correlated in 1. purpurea (Fineblum and Rausher 1995) lends support to the idea that the macroevolutionary pattern found by van der Meijdan et al. (1988) may be the result of microevolutionary processes.

Because this study was conducted exclusively in the laboratory, it is difficult to predict the degree to which the demonstrated negative genetic correlation between defense and tolerance would constrain the evolution of defense and/or tolerance in nature. For example, the effect of damage on attractiveness to pollinators and the contribution of attractiveness to fitness could not be determined (Conner and Rush 1996; Strauss 1997) because estimated total fitness was based on hand-pollinations. In turn, how pollinator attraction may alter the selection imposed on defense or tolerance remains unknown. Nonetheless, while flower number can affect attractiveness to pollinators (Conner and Rush 1996; Strauss 1997), tolerance in terms of flower production did not significantly vary among selection lines. Thus, even under conditions that include the interaction between B. rapa and pollinators, the fact that high-glucosinolate selection lines produce fewer fruits per pollinated flower and fewer seeds per fruit indicates that the observed trade-off between defense and tolerance would persist and constrain the evolution of both glucosinolate production and tolerance.

If costs of both defense and tolerance exist, then the evolution of either resistance component may be constrained. In this study, internal costs of glucosinolate production (defense) and tolerance can be determined by examining the fitness of individuals in the 0% damage treatment. Estimated total potential female fitness of lines selected for low defense was greater than high-defense selection lines in the 0% damage treatment [ILLUSTRATION FOR FIGURE 3 OMITTED], demonstrating an internal cost of defense. Similarly, an internal cost of tolerance would be represented by a negative relationship between tolerance and estimated total potential female fitness in the 0% damage treatment. In contrast to the presence of an internal cost of defense, no internal cost of tolerance was found. Thus, due to the positive correlation between internal costs of defense (fitness in the absence of damage) and external costs of defense (trade-off between defense and tolerance) the most tolerant selection lines (low glucosinolate content) were also the most fit in the absence of damage.

A naive interpretation of the lack of internal costs of tolerance in conjunction with the existence of a negative genetic correlation between defense and tolerance would be that the evolution of increased tolerance is more probable in B. rapa than increased defense. However, the observed pattern of resultant fitness among selection lines was more complex. Estimated total female fitness of low-defense selection lines in the 20% and 60% damage treatments were indistinguishable from the estimated total female fitness of control and high-defense selection lines in the 0% damage treatment [ILLUSTRATION FOR FIGURE 3 OMITTED]. Thus, if herbivore damage is negatively correlated with selection line (defense), poorly defended individuals would receive high damage levels, and the resultant fitness of both high- and low-defense individuals may be equal. This scenario suggests that even when a negative genetic correlation exists between defense and tolerance (Fineblum and Rausher 1995), the interaction between the internal and external costs of defense could result in the maintenance of variation in both defense and tolerance in natural populations of B. rapa with herbivores present. Alternatively, since highly defended individuals are less tolerant than poorly defended individuals, if high-defense individuals receive significant herbivore damage, then low-defense individuals may actually be more fit overall and be favored by natural selection, corresponding to the naive interpretation. Thus, to accurately predict whether this positive correlation between internal and external costs of defense in B. rapa would result in selection for decreased defense and increased tolerance, one would need to know both the frequency and amount of damage that individuals from each defense level receive.

The existence of a negative genetic correlation between defense and tolerance in natural populations of plants could be due to linkage disequilibrium between loci contributing to these components of resistance resulting from correlational selection on defense and tolerance when costs of both resistance components exist (Fineblum 1991). That is, the strength of selection imposed by herbivores for defense or tolerance depends on the levels of each resistance component expressed in the population. When a population receives little damage, that is, expresses high defense, selection by herbivores would impose stronger selection for increased defense. However, when a population receives significant damage, that is, expresses low defense, herbivores would impose stronger selection for increased tolerance. In natural populations, this correlational selection would cause linkage disequilibrium between those loci conferring defense and tolerance yielding a negative genetic correlation between the two traits (Fineblum 1991).

While correlational selection and linkage disequilibrium could be the cause of negative genetic correlations between defense and tolerance in some natural populations, this is not likely the cause of the negative genetic correlation observed in this study. In a laboratory population such as RCB B. rapa, there would be no correlational selection for defense and tolerance because this population would not undergo herbivore-imposed selection. Since a negative genetic correlation due to linkage disequilibrium would be transient unless there is something maintaining it, the linkage disequilibrium between defense and tolerance loci would dissipate in a laboratory population not undergoing correlational selection by herbivores (Falconer and Mackay 1996). However, the number of generations necessary for loci contributing to defense and tolerance in B. rapa to reach linkage equilibrium would depend on the number of loci involved and the recombination rate (Falconer and Mackay 1996). But, because the generation time of this RCB is approximately two months and they have been in the laboratory for over 10 years (Williams and Hill 1986), it is likely that linkage disequilibrium between defense and tolerance loci would have dissipated. Thus, if correlational selection were the only factor creating a negative genetic correlation between defense and tolerance, I should not have observed a correlated response of tolerance to artificial selection for defense. The fact that a correlated response was observed suggests that the origin of the negative genetic correlation between defense and tolerance is inherent to the physiology of the plant, rather than simply due to linkage disequilibrium created by correlational selection and costs of defense and tolerance. This external cost of defense in terms of tolerance may be due to the fact that allocating more resource to defense decreases those available to maintain fitness if an individual incurs damage. Such a physiological cause suggests that this result is a general phenomenon that would be applicable to natural populations of B. rapa and likely other species.

Models concerning the evolution of defense assume that a cost of defense exists, yet there is a significant lack of evidence for such costs. This study in combination with that of Berenbaum et al. (1986) and Fineblum and Rausher (1995) suggest that the detection of costs of defense may be hindered for three reasons. First, if costs of defense are manifested as a decrease in tolerance (Fineblum and Rausher 1995; this study), examining the correlation between defense and fitness in an herbivore-free environment would not detect this external cost of defense. Second, since numerous investigators have used damage as a bioassay of defense (Simms and Rausher 1987, 1989; Rausher and Simms 1989; Marquis 1990; Sork et al. 1993; Simms and Triplett 1994; Stowe et al. 1994) if there is both a negative genetic correlation between defense and tolerance and a cost of tolerance, detection of internal costs of defense may be hampered (Simms and Triplett 1994). Results presented here in conjunction with previous work on this system (Stowe 1997; Stowe and Marquis, unpubl. data) highlight the usefulness and strength of identifying and examining specific defense traits when attempting to detect costs. In addition, previous investigators have detected significant costs of defense when the trait(s) responsible for defense were known (Berenbaum et al. 1986; Han and Lincoln 1994; Sagers and Coley 1995; Stowe and Marquis, unpubl. data), but not when the trait(s) were unknown (Simms and Rausher 1987, 1989). Third, measuring defense as 1 minus percentage damage implicitly assumes that all herbivore species respond similarly to specific defense traits (but see Rausher and Simms 1989; Simms and Rausher 1989). Measuring defense in this way ignores possible variable responses of different herbivores to various traits and as such may obscure negative correlations between any specific defense trait and fitness. Thus, numerous previous attempts to detect costs of defense may have failed because they have declined to examine specific defense traits and/or have ignored trade-offs between defense and tolerance.

ACKNOWLEDGMENTS

I would like to thank B. Reuter and R. Reese for numerous hours of help in the lab, and C. Hochwender and R. Marquis for helpful discussions concerning this topic. I would also like to thank C. Hochwender, J. Le Corff, J. Lill, R. Marquis, K. Mothershead, J. Morisaki, B. Reuter, and S. Marchini for comments on this manuscript; and J. Cheverud, C. Kelly, R. Marquis, and V. Sork for their advice. This work was partially supported by the following grants: Sigma Xi; Garden Clubs of Missouri, Inc.; National Federation of State Garden Clubs; and National Science Foundation Doctoral Dissertation Improvement Grant DEB#9310992. This work was conducted as partial fulfillment for a Ph.D. at the University of Missouri-St. Louis.

LITERATURE CITED

ADLER, F. R., AND R. KARBAN. 1994. Defended fortresses or moving targets? Another model of inducible defenses inspired by military metaphors. Am. Nat. 144:813-832.

AGREN, J., AND D. W. SCHEMSKE. 1993. The cost of defense against herbivores: an experimental study of trichome production in Brassica rapa. Am. Nat. 141:338-350.

-----. 1994. Evolution of trichome number in a naturalized population of Brassica rapa. Am. Nat. 143:1-13.

BERENBAUM, M. R., AND A. R. ZANGERL. 1992. Quantification of chemical coevolution. Pp. 69-87 in R. S. Fritz and E. L. Simms, eds. Plant resistance to herbivores and pathogens: ecology, evolution, and genetics. Univ. of Chicago Press, Chicago.

BERENBAUM, M. R., A. R. ZANGERL, AND A. R. NITAO. 1986. Constraints on chemical coevolution: wild parsnips and the parsnip webworm. Evolution 40:1215-1228.

COLEY, P. D., J. P. BRYANT, AND F. S. CHAPIN III. 1985. Resource availability and plant antiherbivore defense. Science 230:895-899.

CONNER, J. K., AND S. RUSH. 1996. Effects of flower size and number on pollinator visitation to wild radish, Raphanus sativa. Oecologia 105:509-516.

DADAY, H. 1954a. Gene frequencies in wild populations of Trifolium repens L. I. Distribution by latitude. Heredity 8:61-78.

-----. 1954b. Gene frequencies in wild populations of Trifolium repens L. II. Distribution by altitude. Heredity 8:377-384.

-----. 1965. Gene frequencies in wild populations of Trifolium repens L. IV. Mechanisms of natural selection. Heredity 20:355-365.

EHRLICH, P., AND P. H. RAVEN. 1964. Butterflies and plants: a study in coevolution. Evolution 18:586-608.

FALCONER, D. S., AND T. F. C. MACKAY. 1996. Introduction to quantitative genetics. Longman Group Ltd., Essex, U.K.

FEENY, P. P. 1976. Plant apparency and chemical defenses. Recent Adv. Phytochem. 10:1-40.

FINEBLUM, W. L. 1991. Genetic constraints on the evolution of resistance to plant enemies. Ph.D. diss., Duke University, Durham, NC.

FINEBLUM, W. L., AND M.D. RAUSHER. 1995. Tradeoff between resistance and tolerance to herbivore damage in a morning glory. Nature 377:517-520.

FRITZ, R. S. 1990. Effects of genetic and environmental variation on resistance of willow to sawflies. Oecologia 82:325-332.

FRITZ, R. S., AND P. W. PRICE. 1988. Genetic variation among plants and insect community structure: willows and sawflies. Ecology 69:845-856.

HAN, K., AND D. E. LINCOLN. 1994. The evolution of carbon allocation to plant secondary metabolites: a genetic analysis of costs in Diplacus aurantiacus. Evolution 48:1550-1563.

HERMS, D. A., AND W. J. MATTSON. 1992. The dilemma of plants: to grow or defend. Q. Rev. Biol. 67:283-335.

LOUDA, S. M., AND S. MOLE. 1991. Glucosinolates: chemistry and ecology. Pp. 123-163 in G. A. Rosenthal and M. R. Berenbaum, eds. Herbivores, their interaction with secondary plant metabolites. Academic Press, San Diego, CA.

MARQUIS, R. J. 1990. Genotypic variation in leaf damage in Piper arieianum (Piperaceae) by a multi-species assemblage of herbivores. Evolution 44:104-120.

-----. 1992. The selective impact of herbivores. Pp. 301-325 in R. S. Fritz and E. L. Simms, eds. Plant resistance to herbivores and pathogens: ecology, evolution, and genetics. Univ. of Chicago Press, Chicago.

MAURICIO, R., M.D. RAUSHER, AND D. S. BURDICK. 1997. Variation in the defense strategies of plants: are resistance and tolerance mutually exclusive? Ecology 78:1301-1311.

McKEY, D. M. 1979. The distribution of secondary compounds within plants. Pp. 55-133 in G. A. Rosenthal and D. H. Janzen, eds. Herbivores, their interaction with secondary metabolites. Academic Press, New York.

NUNEZ-FARFAN, J. R., AND R. DIRZO. 1994. Evolutionary ecology of Datura stramonium L. in central Mexico: natural selection for resistance to herbivorous insects. Evolution 48:423-436.

PAINTER, R. H. 1958. Resistance of plants to insects. Annu. Rev. Entomol. 3:267-290.

PILSON, D. 1996. Two herbivores and constraints on selection for resistance in Brassica rapa. Evolution 50:1492-1500.

RAUSHER, M.D., AND E. L. SIMMS. 1989. The evolution of resistance to herbivory in Ipomoea purpurea. I. Attempts to detect selection. Evolution 43:563-572.

RHOADES, D. F., AND R. CATES. 1976. Toward a general theory of plant antiherbivore chemistry. Recent Adv. Phytochem. 10:168-213.

RODMAN, J. E. 1991. A taxonomic analysis of glucosinolate-producing plants. 2. Cladistics. Syst. Bot. 16:619-629.

SAGERS, S. L., AND P. D. COLEY. 1995. Benefits and costs of defense in a neotropical shrub. Ecology 76:1835-1843.

SAS INSTITUTE. 1985. SAS user's guide: statistics. Vers. 5. SAS Institute, Cary, NC.

SIMMS, E. L. 1992. Costs of plant resistance to herbivory. Pp. 392425 in R. S. Fritz and E. L. Simms, eds. Plant resistance to herbivores and pathogens: ecology, evolution, and genetics. Univ. of Chicago Press, Chicago.

SIMMS, E. L., AND R. S. FRITZ. 1990. The ecology and evolution of host-plant resistance to insects. Trends Ecol. Evol. 5:356-360.

SIMMS, E. L., AND M.D. RAUSHER. 1987. Costs and benefits of plants defense to herbivory. Am. Nat. 130:570-581.

-----. 1989. The evolution of resistance to herbivory in Ipomoea purpurea. II. Natural selection by insects and cost of resistance. Evolution 43:573-585.

SIMMS, E. L., AND J. TRIPLETT. 1994. Costs and benefits of plant responses to disease: resistance and tolerance. Evolution 48: 1973-1985.

SMITH, C. M. 1990. Plant resistance to insects: a fundamental approach. Wiley, New York.

SOKAL, R. R., AND F. J. ROHLF. 1981. Biometry. 2d ed. Freeman, New York.

SORK, V. L., K. A. STOWE, AND C. HOCHWENDER. 1993. Evidence for local adaptation in closely adjacent subpopulations of Northern Red Oak (Quercus rubra L.) expressed as resistance to herbivores. Am. Nat. 142:359-367.

STEWART, J. G., K. B. MCRAE, AND M. K. SEARS. 1990. Response of two cultivars of cauliflower to simulated insect defoliation. J. Econ. Entomol. 83:1499-1505.

STOWE, K. A. 1997. Experimental evolution of resistance in Brassica rapa: correlated responses of life history traits, tolerance and realized defense to selection for foliar glucosinolate content. Ph.D. diss., University of Missouri-St. Louis, St. Louis, MO.

-----. In press. Realized defense of lines of Brassica rapa artificially selected for foliar glucosinolate content. Environ. Entomol.

STOWE, K. A., V. L. SORK, AND A. W. FARRELL. 1994. Effect of water availability on the phenotypic expression of herbivore resistance in northern red oak seedlings (Quercus rubra L.). Oecologia 100:309-315.

STRAUSS, S. Y. 1997. Floral characters link herbivores, pollinators, and plant fitness. Ecology 78:1640-1645.

STRAUSS, S., J. K. CONNER, AND S. L. RUSH 1996. Foliar herbivory affects floral characters and plant attractiveness to pollinators: implications for male and female plant fitness. Am. Nat. 147: 1098-1107.

VAN DER MEIJDEN, E., M. WIJN, AND H. J. VERKAAR. 1988. Defence and regrowth: alternative native plant strategies in the struggle against herbivores. Oikos 51:355-363.

WILLIAMS, P. H., AND C. B. HILL. 1986. Rapid-cycling populations of Brassica. Science 232:1385-1389.
COPYRIGHT 1998 Society for the Study of Evolution
No portion of this article can be reproduced without the express written permission from the copyright holder.
Copyright 1998 Gale, Cengage Learning. All rights reserved.

Article Details
Printer friendly Cite/link Email Feedback
Author:Stowe, Kirk A.
Publication:Evolution
Date:Jun 1, 1998
Words:7549
Previous Article:Effects of different levels of inbreeding on progeny fitness in Plantago coronopus.
Next Article:Patterns of mating in wild sunflower hybrid zones.
Topics:

Terms of use | Privacy policy | Copyright © 2019 Farlex, Inc. | Feedback | For webmasters