Behavioral Ecology Vol. 13 No. 3: 301-311
© 2002 International Society for Behavioral Ecology
Predation and the evolution of prey behavior: an experiment with tree hole mosquitoes
Behavior, Ecology, Evolution, and Systematics Section, Department of Biological Sciences, Illinois State University, Normal, IL 61790-4120, USA
Address correspondence to S.A. Juliano. E-mail: sajulia{at}mail.bio.ilstu.edu . M.E. Gravel is now at the Ocean Sciences Centre, Memorial University, St. Johns, NF A1C 5S7, Canada.
Received 30 December 2000; revised 15 January 2001; accepted 24 June 2001.
 |
ABSTRACT |
|---|
 TOP ABSTRACT INTRODUCTIONMETHODSRESULTSDISCUSSIONREFERENCES |
|---|
We tested for facultative changes in behavior of an aquatic insect in response to cues from predation and for evolution of prey behavior in response to experimental predation regimes. Larvae of the tree hole mosquito Aedes triseriatus reduced filtering, browsing, and time below the surface in response to water that had held a feeding larva of the predator Toxorhynchites rutilus. We subjected experimental A. triseriatus populations to culling of 50% of the larval population, either by T. rutilus predation or by random removal. After two generations of laboratory culling, behavior of the two treatment groups diverged. Aedes triseriatus in control-culled lines retained their facultative shift from filtering to resting, but tended to lose the response of reduced browsing below the surface in water that had held a feeding predator. Predator-culled lines lost their facultative response of reduced filtering in water that had held a feeding predator and evolved toward more time resting and less time filtering in both water that had held a feeding predator and water that had held only A. triseriatus. Predator-culled lines retained their facultative response of reduced browsing below the surface in water that had held a feeding predator. Two field populations and their reciprocal hybrids responded similarly to cues from predation and did not differ in their evolutionary response to experimental culling. We conclude that consistent presence or absence of predation can select rapidly for divergence in prey behavior, including facultative behavioral responses to predators.
Key words: activity, Aedes triseriatus, behavioral responses, constitutive behavior, facultative antipredator behavior, feeding behavior, predation, predator avoidance behavior, Toxorhynchites rutilus.
|
INTRODUCTION |
|---|
|
TOPABSTRACTINTRODUCTIONMETHODSRESULTSDISCUSSIONREFERENCES |
|---|
Predation appears to cause strong selection on behavior of prey (reviewed by Lima and Dill, 1990
;
Sih, 1987). The result often
appears to be facultative predator avoidance behaviors (e.g., reduced feeding,
reduced activity, and refuge use) that are expressed when prey perceive the
predator (e.g., Ball and Baker,
1995,
1996;
Boersma et al., 1998;
Gilliam and Fraser, 1987;
Grostal and Dicke, 1999;
McPeek, 1990; Sih,
1980,
1986;
Stein and Magnuson, 1976;
Tikkanen et al., 1996;
Wisenden et al., 1999).
Facultative changes in behavior are often taken as evidence for evolution of
prey behavior in response to selection by predation
(Blumstein et al., 2000;
Gilliam and Fraser, 1987;
Main, 1987; Sih,
1980,
1987;
Stein and Magnuson, 1976;
Watkins, 1996) because
behavior patterns induced by the presence of a predator are often associated
with lower vulnerability to predation
(Grill and Juliano, 1996;
Lawler, 1989;
McPeek, 1990;
Peckarsky and Wilcox, 1989;
Skelly and Werner, 1990;
Woodward, 1983). Investigators
have rarely taken the next step of using controlled experiments to test the
prediction that populations subjected to different predation regimes diverge
in behavior. Such experimental tests would provide the strongest evidence that
behavior evolves in response to predation. In this study, we first documented
facultative changes in prey behavior in response to cues from predation and
then tested for evolution of prey behavior (including modification of
facultative responses to cues from predation) in response to controlled
predation regimes.
Our study organisms are larvae of the tree hole mosquitoes Aedes
triseriatus (Say) (prey) and Toxorhynchites rutilus (Coq.)
(predator). These subjects were chosen because (1) behavior patterns of
Aedes larvae are a major determinant of short-term probability of
predation by Toxorhynchites larvae. Toxorhynchites rutilus
is primarily an ambush predator (Steffan
and Evenhuis, 1981), though searching behavior by larvae sometimes
does occur (Linley and Darling,
1993). Prey appear to be detected primarily by mechanoreceptors
(Russo, 1986;
Steffan and Evenhuis, 1981).
More active individuals that move and feed more frequently, and individuals
below the water's surface, are more likely to be victims of predation, whereas
resting at the water's surface entails a relatively low risk of predation
(Grill and Juliano, 1996;
Juliano and Reminger, 1992;
Rubio et al., 1980;
Russo, 1986). (2) There is
significant interpopulation variation in behavior of A. triseriatus
(Juliano et al., 1993;
Juliano and Reminger, 1992),
suggesting local adaptation. (3) Late instar A. triseriatus alter
their behavior in the presence of T. rutilus larvae
(Juliano and Reminger, 1992);
however, these facultative behavioral responses by A. triseriatus to
this predator are not well understood. For example, these behavioral responses
are not detectable under all experimental conditions (e.g., not with caged
predators; Hechtel and Juliano,
1997), and they do not appear to vary consistently between
populations with different histories of exposure to this predator
(Hechtel and Juliano, 1997;
Juliano and Reminger,
1992).
We addressed two specific questions about the potential for predation by
T. rutilus to cause evolution of behavior of A. triseriatus.
First, does A. triseriatus adopt low-risk behaviors in response to
chemical and physical traces left by the act of predation (as opposed to
T. rutilus larvae themselves, as in past studies;
Juliano and Reminger, 1992)?
Based on previous studies with this predator—prey system
(Juliano and Reminger, 1992;
Juliano et al., 1993), we
predict that cues from T. rutilus predation should induce decreased
movement and feeding and increased time at the surface.
Second, does prey behavior change in response to controlled predation over several generations, and if so, how? Both constitutive and facultative behaviors could change in response to selection by predation. We quantified behavior under two conditions: with no cues to the presence of T. rutilus, and in water that has held a feeding T. rutilus, which may provide cues indicating the presence of this predator. By analyzing effects of selection by predation on behavior in these two environments, we sought to determine whether consistent selection by predation favors enhanced or modified facultative behavioral responses to the predator or favors a shift to constitutive patterns of low-risk behavior.
We addressed these questions in a laboratory experiment comparing experimental lines subjected to predation to control lines raised under identical conditions but without predation. Thus, we held constant other selective forces (e.g., density, competition, resource quantity and quality) and isolated the effect of predation alone on short-term evolution of prey behavior.
|
METHODS |
|---|
|
TOPABSTRACTINTRODUCTIONMETHODSRESULTSDISCUSSIONREFERENCES |
|---|
Source populations
We used two populations of A. triseriatus, both of which encounter T. rutilus in nature. One population was collected from tree holes at Parklands Merwin Reserve (and adjacent private lands) near Normal, Illinois, USA (88°50' W longitude, 40°39' N latitude), where T. rutilus is relatively rare (Juliano et al., 1993
). Toxorhynchites rutilus was collected in
only 4 of 10 years beginning in 1990, only in late summer and fall, and never
in spring, suggesting that they do not overwinter at this latitude
(Juliano, 1996). During the
two summers preceding this experiment, T. rutilus occurred in only 1
of 23 tree holes (4%) and 7 of 18 discarded tires (39%) sampled at this site
(Juliano and Gravel, unpublished data). The second population was collected
from tree holes at Horseshoe Lake State Park near Olive Branch, Illinois
(89°13' W longitude, 37°6' N latitude), where T.
rutilus appears to be more common
(Juliano et al., 1993). During
the two summers preceding this experiment, T. rutilus occurred in 9
of 36 tree holes (25%) and in 1 of 2 tires (50%) sampled at this site (Juliano
and Gravel, unpublished data). We collected A. triseriatus as larvae
and pupae, raised them to adulthood in environmental chambers, and allowed
them to breed in 0.6-m3 cages. We collected T. rutilus as
eggs or larvae in the field at Vero Beach, Florida, USA (82°23' W
longitude, 27°31' N latitude), the site where we conducted the
culling phase of the experiment. Toxorhynchites rutilus from this
site prey regularly on A. triseriatus in nature, and using this
population simplified the acquisition of sufficient numbers of predators and
does not affect our test of these hypotheses.
Field-collected A. triseriatus formed the parental (P) generation.
In addition to these populations, we created reciprocal hybrid populations,
which we deemed important for this design because of the possibility that
genetic variation in behavior within a single population may be low due to the
effect of selection on behavior (Hoffman,
2000; Mousseau and Roff,
1987). If there is interpopulation differentiation in behavior, as
suggested for the Normal and Olive Branch populations
(Juliano et al., 1993), and if
that differentiation has some genetic basis, hybrid populations may show
greater genetic variation and so may respond more rapidly to our laboratory
selection regimes. Thus, the hybrid populations may give us more power to
detect effects of selection by predation. The P generations for both pure
populations and for two reciprocal hybrid populations consisted of the
following numbers of adults: male Normal/female Normal (abbreviated NN) =
851/767; male Olive Branch/female Olive Branch (OO) = 524/602; male
Normal/female Olive Branch (NO) = 845/604; and male Olive Branch/female Normal
(ON) = 520/767. We reared the progeny of the P generation (F1) for
one generation in the laboratory in environmental chambers at 25°C with a
14:10 h day: night light cycle. We used their progeny (F2) to start
the culling experiment. Thus, we began the experiment with individuals only
two generations removed from field environments, minimizing the effects of
unintentional laboratory selection on behavior. We reared the F1
parents of the generation that initiated the experiment under identical
laboratory conditions, which should reduce any maternal effects of field
environments. Finally, any genetic variation in the hybrid populations that is
associated with recessive traits (and therefore masked in the F1
individuals) should be phenotypically expressed in the F2
individuals, increasing the amount of genetic variation upon which selection
may act. We cannot rule out the possibility that the two generations of
laboratory rearing prior to initiating the experiment altered the behavior of
larvae relative to that of field-collected larvae. However, because all
individuals were subjected to the same conditions during this period, this
preexperimental rearing cannot be responsible for any effects of our
treatments on behavior. Even if the behavior of larvae was affected by the
preexperimental rearing, our experiment still provides us with a test of
whether controlled predation regimes result in divergence of behavior.
Culling experiment
We chose to test for evolution of behavior in response to predation using a
laboratory culling experiment (Rose et al.,
1990,
1996). Culling experiments
differ from traditional artificial selection experiments, in which the
investigator measures phenotypic values for a number of individuals and then
chooses the parents for subsequent generations based on these values. In
contrast, in a culling experiment, replicate populations are exposed to some
potential agent of selection (in our case, sustained predation by a single
predator species) in circumstances that will yield a limited number of
survivors, and the survivors are used as parents for the next generation. The
traits of interest (in our case, prey behavior) in these selected populations
may then be compared to those in control populations without the agent of
selection (or with some alternative agent of selection). Culling experiments
can produce responses to selection quickly and, because they are less labor
intensive than selection experiments, may facilitate investigation of larger
experimental populations or greater replication (Rose et al.,
1990,
1996). Most important,
however, a culling experiment provides us with a means of testing the
effectiveness of a particular agent of selection (in this case predation) as a
cause of evolution. To our knowledge, this approach has not been used in the
study of evolution in response to predation but has been used to test for
evolution in response to several other factors
(Chippindale et al., 1994;
Goodnight and Craig, 1997;
Joshi and Mueller, 1988;
Mueller, 1988;
Rose, 1984;
Rose et al., 1990;
Zwaan et al., 1995).
From each of the four populations (i.e., NN, OO, NO, ON), we created replicate lines for predator and control culling regimes. For each replicate line, we placed 300 larvae in a 1-1 container with 800 ml of water. Initially, we established eight lines from each of the four populations. For the first generation of culling, when larvae were 8 days old (and culling was to begin), we equalized densities of all replicate lines by pooling within populations, resulting in a density at the start of culling of 250 larvae per line. This pooling is justifiable because at this point in the experiment, all larvae had been treated identically and had the same history. This pooling resulted in a reduction of replicate lines to four for OO and NO, six for ON, and eight for NN. For the second generation of culling, after each line had its own unique history, we could not equalize densities at the start of the culling phase; therefore, the number of larvae in each line when culling began on day 8 varied from 57 to 280. By the start of the second generation of culling, 6 of 22 lines had gone extinct due to combinations of poor production of adults and poor production of eggs. Though we lost lines from the experiment, by the second generation we still had eight control and eight predator lines; hence extinction of lines was not related to the culling treatments.
For each generation of culling, we subjected larvae in predator lines to predation by T. rutilus until the population of larvae in the line was reduced by approximately 50%. For different replicate lines, this required 2-6 days of exposure to the predator. We added the predator when the A. triseriatus larvae were 8 days old, at approximately the third instar. To keep effects of density consistent between culled and control lines, we culled larvae in control lines randomly, every other day, to densities necessary to produce the same proportion survivorship as the mean of the corresponding predator lines. For random culling of control larvae, we poured the water and larvae onto a screen marked with a numbered grid. We used a random number table to determine from which grid sections larvae were to be removed (regardless of instar). We chose grid sections until we had removed the required number of larvae. We also collected larvae in lines culled by the predator on a screen and counted them every other day to determine the appropriate reduction for the control lines. We fed larvae in each line 15 mg of bovine liver powder at the start of the experiment, and we also provided an additional 0.05 mg/larva every other day, beginning on day 8, after each count. Thus, all replicate lines received the same amount of liver powder per larva.
We reared survivors from each predator and control line to adulthood and
propagated adults in 15-1 bucket cages housed in a screened porch at Florida
Medical Entomology Laboratory, Vero Beach, Florida. Adults were subjected to
ambient temperature and photoperiod of subtropical Florida
(September—November). We provided adults in each line with sugar water
and weekly blood meals from restrained chickens. We collected and
synchronously hatched (Novak and Shroyer,
1978) eggs from each replicate line and subjected resulting larvae
to the next generation of culling. In each generation, we raised larvae from
the same hatch as those used for the culling experiment individually for
behavioral observations. We subjected lines to two generations of culling and
collected behavioral data for three generations: preculling (F2),
after one generation of culling (F3), and after two generations of
culling (F4). We refer to these as preculling, first generation,
and second generation, respectively.
Behavioral observations
Rearing of individuals
We placed newly hatched larvae individually into 18=ml vials with 11 ml of
water. Thus, in all cases, the tested individuals from predator lines had
never been exposed to T. rutilus or any stimuli emanating from T.
rutilus until the time they were tested. Therefore, any differences in
behavior between predator and control lines cannot be products of the
experience of the test larvae. We fed these larvae a liver powder suspension
(LPS) prepared by mixing 0.3 g of bovine liver powder with 1 l of water. The
suspension was stirred continuously on a magnetic stir plate and transferred
by Eppendorf variable-volume pipette to rearing vials. We provided each larva
with 0.50 ml of LPS on days 1 and 4, 1.00 ml of LPS on days 9 and 12, and 2 ml
of LPS on day 14, and every other day thereafter until pupation. We raised
these larvae in an environmental chamber at 23°C and a 14:10 h day:night
cycle.
Observation protocol
We videotaped behavior of one-day-old fourth instar larvae of the
preculling, first, and second generations. Before all observations, we held
larvae individually in a 50-ml cup with 30 ml of water with no food for 24 h
to standardize hunger before transfer to the test water. We recorded activity
and position of each larva every minute for 30 min in instantaneous scan
censuses (Juliano and Reminger,
1992; Martin and Bateson,
1986). We classified activity into four categories: (1)
browsing—the larva moved along the surfaces of the cup propelled by
movements of the mouth parts, which brushed the surfaces of the cup; (2)
resting—the larva was completely still and not feeding; (3)
filtering—the larva drifted through the water column, propelled by
movement of mouth parts, which created filtering currents; and (4)
thrashing—the larva propelled itself through the water by vigorous
lateral flexion of the body. We also classified position within the cup into
four categories: (1) surface—the larva's spiracular siphon was in
contact with the surface; (2) bottom—the larva was within 1 mm of the
bottom of the cup; (3) wall—the larva was within 1 mm of the sides of
the cup; and (4) middle—the larva was not in contact with the surface
and was more than 1 mm from the cup's surfaces. We classified larvae
simultaneously at the surface and wall as at the surface, and larvae
simultaneously at the bottom and wall as at the bottom. Among activities,
resting is the least likely to lead to predation, thrashing is the most likely
to result in predation, and feeding behaviors are intermediate
(Juliano and Reminger, 1992).
Among positions, the surface is the least likely to lead to predation, the
bottom is the most likely to lead to predation, and middle and wall are
intermediate (Grill and Juliano,
1996; Juliano and Reminger,
1992, Juliano et al.,
1993).
We tested for facultative behavioral changes in response to perception of the predator and examined whether such facultative changes are modified by selection due to predation by observing larvae under two experimental water treatments designed to provide cues to the presence of T. rutilus. We observed one group of larvae in water that had held a fourth instar T. rutilus larva feeding on A. triseriatus (predator + prey water). We observed the other group in water that had held only A. triseriatus (prey-only water). We prepared predator + prey water by placing one fourth instar T. rutilus in 50 ml of water with 10 third and fourth instar A. triseriatus larvae and replacing eaten or dead A. triseriatus larvae daily. We prepared prey-only water similarly, but with only 10 A. triseriatus larvae. We did not feed the A. triseriatus larvae during preparation of either water treatment. After 5-7 days of preparation, we removed all larvae from both treatments and recorded behavior patterns of test larvae as described above. We observed only one larva in each beaker of prepared water and then discarded the water.
Statistical analysis
Behavioral observations were converted to proportions of observations in
each activity or position category. To reduce the number of variables and to
obtain uncorrelated descriptors of behavior patterns, we summarized the
activity and position of the preculling generation using principal components
analysis (PCA; Hatcher and Stepanski,
1994). We estimated principal components (PCs) for activity and
position combined (SAS Institute Inc., 1990; PROC FACTOR), and retained
resulting principal components only if their eigen values exceeded 1.00
(Hatcher and Stepanski, 1994).
To follow changes in the same measure of phenotype over the subsequent
generations (first, second), we used the principal component loadings
generated during the preculling generation to determine PC scores in the first
and second generation. Because we used the same functions, we can compare
principal component scores across generations and determine how much these
summary variables change in response to experimental culling.
We then analyzed PC scores for all three generations using MANOVA (SAS
Institute Inc., 1990; PROC GLM). We interpreted MANOVA results via
standardized canonical coefficients
(Scheiner, 1993), which
indicate by their magnitudes which variables contributed to significant MANOVA
effects. When appropriate, we compared populations using multivariate
contrasts (Scheiner, 1993).
For each of the two postculling generations, we tested for effects of
population (Normal, hybrid, Olive Branch), culling regime (predator, control),
replicate line within population-culling regime, water (predator + prey, prey
only), and all interactions. We considered line within population-culling
regime, and all its interactions, to be random effects. For the preculling
generation, the analysis was similar, but without effects involving culling
regime and line within population-culling regime, because replicate lines were
only established after culling had begun.
For the first and second generations, we could not reject the null hypotheses of normally distributed residuals for PC scores. Residual plots indicated no pronounced heterogeneity of variances. For the preculling generation, residual plots indicated approximately homogeneous variances, but residuals were not normally distributed (p <.05, Shapiro-Wilk test). We found no transformation that improved this situation, and various transformations did not alter the overall conclusions concerning statistical significance. We therefore analyzed all PC scores without transformation.
For the preculling generation, we compared phenotypic variances for PCs. We first removed effects of water and population on the means by running ANOVA on each PC score and obtaining residuals. We then tested whether phenotypic variance of these residuals within hybrid populations (ON and NO pooled) was greater than phenotypic variance of these residuals within pure populations (NN and OO pooled) using F tests.
|
RESULTS |
|---|
|
TOPABSTRACTINTRODUCTIONMETHODSRESULTSDISCUSSIONREFERENCES |
|---|
Preculling generation
Across all populations and water treatments, resting was the most common activity, and surface the most common position (Table 1). There were significant positive correlations between position and activity categories, particularly for resting and surface, and for browsing and wall, bottom, and middle (Table 2). Within positions, there were significant positive correlations among bottom, wall, and middle (Table 2). Positive, but statistically nonsignificant, correlations were also present for browsing, filtering, and thrashing (Table 2). Resting and surface were significantly and negatively correlated with other activity and position variables, respectively (Table 2). Because such strong correlations complicate interpretations of any changes that may occur in activities and positions in response to the environment (i.e., prey only vs. predator + prey water) or to the culling regime (i.e., predator vs. control), our analysis focused on PCs.
|
|
Three PCs with eigen values >1 summarized >89% of the variation in activity and position (Table 3). The fourth eigen value was less than half of the third (Table 3). Rotated factor scores show that PC1 quantifies allocation of time between resting at the surface versus browsing at the wall and bottom (Table 4). Low scores on PC1 result from frequent resting at the surface (large negative coefficient for resting, surface; Table 4) and infrequent browsing of walls and bottom (large positive coefficients; Table 4). PC2 quantifies the frequency of thrashing in the middle (Table 4). High scores on PC2 result from frequent thrashing in the middle and infrequent other activities. PC3 quantifies the allocation of time to resting and filtering (Table 4). High scores on PC3 result from frequent filtering and infrequent resting, and low scores from the reverse (Table 4).
|
|
A significant water effect in MANOVA (Table 5) resulted primarily from larvae in predator + prey water having lower scores for PC1 and, to a lesser extent, PC3 (note large canonical coefficients for PC1 and PC3, relative to that for PC2; Table 5) than did larvae in prey-only water. Thus, predator + prey water induced reduced browsing below the surface and increased resting at the surface, and also some reduction in filtering (Figure 1A). In contrast to the water effect, the significant population effect resulted primarily from PC2 (note large canonical coefficient for PC2 relative to those for PC1 and PC3; Table 5), with hybrids (ON, NO) having lower PC2 scores than did pure populations (OO, NN). Thus, hybrids show a lower frequency of thrashing in the middle (Figure 1B). Pure populations did not differ, and reciprocal hybrids did not differ (Table 5). The nonsignificant interaction (Table 5) indicates that behavioral responses of these populations and the hybrids to cues from feeding T. rutilus are similar (Figure 1A).
|
|
Because resting and surface are both the safest
(Juliano and Reminger, 1992)
and most frequent behaviors, we also analyzed frequencies (arcsine square-root
transformed) of these two behaviors alone as a test of whether our conclusions
are robust relative to manipulations of the data. MANOVA on frequencies of
resting and surface showed effects similar to those derived from PCs, with a
highly significant effect of water (Pillai's trace = 0.507, df = 2, 118,
p =.0001) and significant effects of population (Pillai's trace =
0.106, df = 6, 238, p =.0418) and the interaction of the two
(Pillai's trace = 0.111, df = 6, 238, p =.0337). Proportion of time
at the surface (adjusted mean ± SE: prey only = 0.62 ± 0.02;
predator + prey = 0.94 ± 0.02) and time resting (adjusted mean ±
SE: prey only = 0.31 ± 0.03; predator + prey = 0.75 ± 0.03) both
increased significantly in predator + prey water. The difference between prey
only and predator + prey water was significant within all four populations,
though the magnitude of the difference was somewhat less for the NN population
(data not shown). Although analysis of only these two variables cannot
indicate the real nature of behavioral shifts in response to cues from the
predator (i.e., what decreases as resting and surface increase?), the
similarity of the significant effects supports our interpretation of PC1 and
indicates that cues from predation induce A. triseriatus to adopt a
safer pattern of behavior.
Variation of PCs within each population was similar (Figure 1). Our F tests on residuals indicated that phenotypic variance in PCs for behavior was not significantly greater for hybrid than for pure populations (F60,67 = 1.11, 0.27, and 1.16 for PC1, PC2, and PC3, respectively; p >> .10 in all cases).
Effects of culling by predation
Because there were no differences between the reciprocal hybrid populations
ON and NO (see above), we pooled the reciprocal hybrid populations into a
single hybrid population for analyses of the first and second generations.
After one generation of controlled culling in the laboratory, there were
neither significant differences among populations nor significant effects
involving culling for the three PCs (Table
6). The only significant MANOVA effect present after one
generation of culling was the effect of water treatment, which derived
primarily from PC1 (Table 6). For all populations combined, effects of water treatment on PC scores were
qualitatively similar to those observed in the preculling generation
(Figure 2). In predator + prey
water, larvae had lower mean scores on PC1 than did larvae in prey-only water
(Figure 2).
|
|
A simplified analysis of only frequencies of resting and surface after one generation of culling yielded the same conclusions as analysis of PCs, with only the effect of water significant (Pillai's trace = 0.859, df = 2,9, p =.0001). As in the preculling generation, frequencies of both resting (adjusted mean ± SE: prey only = 0.47 ± 0.02; predator + prey = 0.78 ± 0.02) and surface (adjusted mean ± SE: prey only = 0.74 ± 0.03; predator + prey = 0.95 ± 0.03) increased significantly in predator + prey water.
In contrast, after two generations of laboratory culling by the predator, there were significant effects of water and culling (Table 7). The culling-by-water interaction yielded p =.0521 (Table 7), which we interpret as a significant interaction, indicating that the behavioral response to water treatment differed between predator versus control culling regimes. Canonical coefficients indicated that PC1 and PC3 both contributed to the water and culling effects and that the culling-by-water interaction resulted primarily from PC3 (Table 7). In prey-only water, larvae in the predator-culling treatment browsed at the bottom and wall more, and filtered less, than did larvae in the control-culling treatment (Figure 2). In contrast, in predator + prey water, larvae in these groups differed little in frequencies of these activities and positions (Figure 2). The culling-by-water interaction and the trajectory of change in individual variables is most clearly seen when PC3 or PC1 scores are plotted individually versus by generation (Figure 3). For PC3, the variable most responsible for the significant culling-by-water interaction across the three generations of the experiment, there was an increase in allocation of time to resting at the expense of filtering in all groups (Figure 3A). For the control-culling regime, the difference in PC3 between water treatments remained similar across all three generations (Figure 3A). In contrast, for the predator-culling regime, the difference in PC3 between water treatments declined steadily across the three generations, so that by the second postculling generation, the values for PC3 in the two water treatments were virtually identical and low (Figure 3A). Culling by T. rutilus, then, appears to shift behavior of A. triseriatus to a pattern of reduced filtering even in water that contains no cues to the presence of the predator.
|
|
The behavioral response of second-generation larvae from lines culled by the predator remains evident in PC1, representing a shift from browsing below the surface to resting at the surface in predator + prey water (Figure 3B). In contrast, for PC1, second-generation larvae from control culled lines showed a reduced behavioral response to predator + prey water (Figure 3B). Effects of culling on PC1 (browsing below the surface vs. resting at the surface) are not as pronounced as those involving PC3 (filtering vs. resting; Table 7, Figure 3). However, when behavioral responses to predator + prey water for both PC1 and PC3 are considered together, second-generation predator-culled lines show uniformly low filtering (low PC1) and a strong reduction of browsing below the surface in predator + prey water (Figure 2). In contrast, second-generation control culled lines show a strong reduction of filtering, with little difference in browsing below the surface, in predator + prey water (Figure 2).
A simplified analysis of only frequencies of resting and surface after two generations of culling yielded similar conclusions to analysis of PCs, with only the effects of water (Pillai's trace = 0.704, df = 2,7, p =.0141) and culling-by-water interaction (Pillai's trace = 0.604, df = 2,7, p =.0390) significant. Larvae in the predator-culled treatment had similar frequencies of resting in prey-only and predator + prey water (adjusted mean ± SE: prey only = 0.36 ± 0.05; predator + prey = 0.44 ± 0.05), but frequencies of time at the surface (adjusted mean ± SE: prey only = 0.46 ± 0.03; predator + prey = 0.68 ± 0.03) that increased significantly in predator + prey water. In contrast, larvae in the control-culled treatment had greater frequencies of resting in predator + prey water (adjusted mean ± SE: prey only = 0.36 ± 0.04; predator + prey = 0.53 ± 0.04), but frequencies of time at the surface (adjusted mean ± SE: prey only = 0.69 ± 0.03; predator + prey = 0.71 ± 0.03) that did not differ between prey only and predator + prey water.
|
DISCUSSION |
|---|
|
TOPABSTRACTINTRODUCTIONMETHODSRESULTSDISCUSSIONREFERENCES |
|---|
Our results show that A. triseriatus adopts low-risk behaviors in response to chemical and physical traces of predation, and support the hypothesis that predation by T. rutilus can cause evolution of A. triseriatus behavior. The behavioral response of larvae to water that has held a feeding predator diverges in lines subjected to selection by predation versus control lines, with control-selected lines responding by reduced filtering and increased resting, but predator-selected lines showing constitutively low filtering. We reached these same conclusions regardless of whether we analyzed the PC scores derived from all behavioral variables or focused only on the safest and most common behaviors (resting, surface). Hence, our results are robust with respect to data manipulation, and we focus our discussion on the more complete analysis of PC scores. We now return to the questions posed concerning selection by predation in this system.
Behavioral change in response to the predator
Larvae of A. triseriatus dramatically increase the frequency of
low-risk behaviors in response to water that had held a feeding T.
rutilus. As predicted, movement in general and feeding (browsing,
filtering) in particular were significantly less common in predator + prey
water than in prey-only water (Figure
1A). In addition, as predicted, larvae spent more time at the
surface and less time at the bottom in predator + prey water compared to
prey-only water (Figure 1A).
These results suggest that A. triseriatus can perceive cues from
predation and modify behavior to reduce risk of predation. As previously
reported (Juliano and Reminger,
1992), there appears to be little differentiation among
populations in behavioral responses to T. rutilus. Both populations,
and their reciprocal hybrid populations, showed similar responses
(Figure 1A). The major
behavioral difference among populations in the preculling generation was that
the hybrid populations showed much lower values of PC2, indicating infrequent
thrashing in the middle, relative to the pure populations
(Figure 1B). Why hybrid
populations should be less active than the associated parental populations is
not obvious.
The observed changes in behavior of preculling generations of A. triseriatus when exposed to predator + prey water may be interpreted as adaptive predator avoidance behavior. However, an alternative explanation is that the change in behavior occurs in response to organic enrichment of the water, via predator feces and victim body parts, which increases microbial food for A. triseriatus and so modifies foraging behavior. This hypothesis suggests that the change in behavior in predator + prey water is not specifically related to predation. This alternative hypothesis is not supported by the apparent evolution of this facultative behavioral response under different regimes of predation, which suggests that the behavioral changes in response to predator + prey water are indeed adaptations specifically related to predation by T. rutilus.
Behavioral responses to cues from T. rutilus were similar for all
populations despite apparent differences in the regularity with which
individuals from Normal or Olive Branch would encounter T. rutilus
(Juliano et al., 1993;
Juliano, 1996; see Methods).
Rare occurrence of this predator at the Normal site is difficult to study
quantitatively, but it appears that in most containers for most of the active
season (April-October), developing A. triseriatus from Normal do not
encounter T. rutilus. Despite this, they show a clear and strong
behavioral response to cues from this predator. Though T. rutilus is
collected more frequently at the Olive Branch site, a substantial number of
containers at this site also lack this predator. Thus, at both sites, T.
rutilus predation varies in space and time, as it does at other sites in
North America (e.g., Lounibos et al.,
1997). Such variable occurrence of a predator is expected to favor
facultative, rather than constitutive, behavioral responses to predation
(Sih, 1987), as observed in
the populations of A. triseriatus we have studied.
Past tests for behavioral responses of A. triseriatus to T.
rutilus used different methods and obtained different results. Changes in
behavior in response to the current presence of T. rutilus
(Juliano and Reminger, 1992)
were similar to, but more subtle than, those observed in the preculling
generation in the present study. Juliano and Reminger
(1992) placed a T.
rutilus larva in the test water immediately before the trial, so that
there were cues from the predator itself but no cues from the act of predation
(e.g., victim body parts, predator feces). The likely source of quantitative
differences in results is the set of cues available from the predator. The
difference in results suggests that A. triseriatus larvae respond
more strongly to chemical or physical traces left by the act of predation and
less strongly to chemical or tactile cues to the current presence of the
predator itself. Such responses seem ideal for detecting an ambush predator
like T. rutilus, which may be difficult to detect if it is motionless
in a dark tree hole environment. Chemical and tactile traces of predation, in
contrast, are likely to be more reliably detectable and better indicators of
danger of predation (Sih,
1986).
Aedes triseriatus shows no behavioral response to caged, feeding
T. rutilus (Hechtel and Juliano,
1997). This lack of response to a caged predator, combined with
our results showing responses to water that has held a feeding predator,
suggests that solid residues from predation (e.g., prey body parts, predator
feces), which would have been retained inside cages used by Hechtel and
Juliano (1997), may be the
primary cues that induce behavioral changes in A. triseriatus.
Similar specificity of predator-derived cues that elicit behavioral responses
in prey is evident in diverse systems (e.g.,
Blumstein et al., 2000;
Grostal and Dicke, 1999; Sih,
1986,
1987;
Tikkanen et al., 1996;
Wisenden et al. 1999).
Which traits evolve in response to predation?
After only two generations of culling, individuals in the predator-culled
lines showed significantly reduced changes in behavior, as summarized by PC3,
in response to cues from predation by T. rutilus. Predator-culled
lines lost behavioral plasticity and showed increased baseline frequencies of
the safest behaviors, regardless of cues from T. rutilus predation
(Figures 2,
3A). In contrast,
control-called lines retained behavioral plasticity, as summarized by PC3, in
response to cues from predation (Figures
2,
3A). Thus, our data clearly
indicate that consistent selection by predation results in reduced facultative
behavioral responses and a shift in constitutive behavior patterns. Despite
the effect of culling by predation on the facultative switch from filtering to
resting, both predator-culled and control-culled lines appear to retain some
form of facultative response to predator-derived cues, but those responses are
not the same. Individuals in the preculling generation shifted from browsing
the walls and bottom to resting at the surface in the presence of
predator-derived cues (Figure
1), and this behavioral change remains strong in the
predator-culled lines after two generations (Figures
2,
3B). This shift from browsing
walls and bottom to resting at the surface tended to be reduced, however, in
the control-culled lines (Figures
2,
3B).
Perhaps the most interesting aspect of our results is that lines subjected
to different culling regimes clearly diverged in their behavioral responses to
this predator in only two generations of intense selection. This result
implies that there may be rapid differentiation of predator avoidance
behaviors if exposure to predators changes, as suggested for other
predator—prey systems (e.g.,
Blumstein et al., 2000).
Conditions in this laboratory experiment were, of course, different from those
encountered in the field by A. triseriatus, hence we are unwilling to
predict that field populations subjected to different levels of predation
would evolve the same behavior patterns as those in our experiment. However,
we do expect that behavior patterns, particularly facultative behavior
patterns, would diverge in some way in populations subjected to consistently
different predation regimes. We would have liked to continue this culling
experiment for additional generations to see if divergence of behavior
patterns continued, but losses of experimental lines rendered continuation of
this kind of controlled experiment with these species difficult, particularly
as this experimental protocol requires dramatic reduction in population size
every generation. However, even this relatively brief experiment showed that
divergence of behavior did occur. Further, the brief duration of the
experiment is realistic. For most of the range of these species, seasonal
environments dictate that predation by T. rutilus is unlikely to
continue at a high level for more than a few generations because of diapause
in A. triseriatus (Sims,
1985) and winter reductions of populations of T. rutilus
either due to freezing in the north or drought in the south
(Bradshaw and Holzapfel, 1985).
Spring generations of A. triseriatus may therefore escape predation
almost completely, and two generations of intense predation may be a
reasonable representation of what occurs over much of the range of these
species.
Although our results might suggest that there should be strong
interpopulation differentiation in behavioral responses to this predator, the
role of predation by T. rutilus in causing differentiation of A.
triseriatus behavior in nature is not yet clear. At sites where T.
rutilus occurs, container habitats form a temporal and spatial mosaic of
low versus high probability of predation (see Methods; see also Bradshaw and
Holzapfel, 1983,
1985;
Lounibos et al., 1997), so
that even where T. rutilus is common, the entire population of A.
triseriatus may not be subject to consistent directional selection. Such
variation, coupled with the rapid response of behavior to selection by
predation, may in fact enhance intrapopulation variation in behavioral
responses by causing small-scale intrapopulation differentiation. Subgroups
within the population (e.g., within one tree hole) may experience directional
selection on behavior, but because different subgroups are affected
differently, and because adults are mobile, evolution of behavior within the
entire population may show no directional trend.
Evolution of these behavioral differences in only a few generations of
laboratory husbandry raises questions about studying behavioral responses to
predation and other environmental cues of mosquitoes (and other organisms)
maintained in laboratory colonies for many generations. Laboratory-reared
lines may evolve modified facultative responses when removed from environments
with appropriate agents of selection. Facultative responses are particularly
problematic for laboratory study, as it seems likely that what selects for
facultative responses in nature is uncertainty (e.g., of predator presence;
Sih, 1987), and most
laboratory-rearing programs do not incorporate uncertainty.
Differences in behavior patterns between hybrids versus parental populations were evident in the preculling individuals (Table 1) but were apparently lost in both the culled and control lines in the first generation after culling (Table 2). This apparent loss of interpopulation behavioral differences should be interpreted with caution, however, as the design of the postculling experiment, with replicate lines, results in reduced denominator degrees of freedom for MANOVA statistics (from 357 to 18, sec Tables 2,3,4) for tests for population effects and thus reduced power for F tests to detect those population effects.
The lack of interaction of population and culling indicates that pure populations and the hybrids responded to culling regimes in similar ways. It thus appears that all populations harbored similar genetic variation for behavior (including facultative behavior) related to predation. This conclusion is consistent with the absence of differences in phenotypic variances for PC scores for pure versus hybrid populations in the preculling generation.
We have shown that populations that regularly encounter predation or the
absence of predation may diverge in their behavioral responses to predation
and that a constitutive pattern of increased low-risk behavior (increased
resting and reduced filtering in this case), regardless of the presence of
cues from the predator, may evolve in populations that regularly encounter a
predator. Our experiment isolates the effect of predation as a selective agent
acting on behavior, but behavior pattern is likely to affect feeding, growth,
and possibly competitive ability (Grill
and Juliano, 1996). Such competing selective forces make
prediction of evolution of behavior in nature more difficult, even if there is
consistent presence or absence of predation. Evolution of behavior in response
to predation regime may only be obvious when other factors are controlled and
selection by predation is applied uniformly and consistently, as was the case
in our laboratory experiment.
|
ACKNOWLEDGEMENTS |
|---|
We thank C.P. Grill, P.M. Léonard, M.A. Nannini, K. Nowers, and L.M. Turek for aid in the laboratory and field; L.P. Lounibos, R. Baker, and the faculty and staff of Florida Medical Entomology Laboratory for providing facilities; Parklands Foundation, D.E. Davis, D. Sears, Horseshoe Lake State Park, and Florida Medical Entomology Laboratory for permission to collect mosquitoes; and J.E. Armstrong, V.A. Borowicz, L.P. Lounibos, S.K. Sakaluk, S.M. Scheiner, and C.F. Thompson for useful discussion or comments on earlier versions of this paper. This research was funded by grants from National Science Foundation (BSR 90-06452), National Institutes of Health (AI-34139), and the Illinois State University Office of Research.
|
REFERENCES |
|---|
|
TOPABSTRACTINTRODUCTIONMETHODSRESULTSDISCUSSIONREFERENCES |
|---|
Ball SL, Baker RL, 1995. The nonlethal effects of predators and the influence of food availability on life history of adult Chironomus tentans (Diptera: Chironomidae). Freshwater Biol 34: 1-12.
Ball SL, Baker RL, 1996. Predator-induced life history changes: anti-predator behavior costs or facultative life history shifts? Ecology 77: 1116-1124.
Blumstein DT, Daniel JC, Griffin AS, Evans CS, 2000. Insular tammar wallabies (Macropus eugenii) respond to visual but not acoustic cues from predators. Behav Ecol 5: 528-535.
Boersma M, Spaak P, DeMeester L, 1998. Predator-mediated plasticity in morphology, life history, and behavior of Daphnia: the uncoupling of responses. Am Nat 152: 237-248.[Web of Science]
Bradshaw WE, Holzapfel CM, 1983. Predator-mediated, non-equilibrium coexistence of tree-hole mosquitoes in southeastern North America. Oecologia 57: 239-256.[Web of Science]
Bradshaw WE, Holzapfel CM, 1985. The distribution and abundance of treehole mosquitoes in eastern North America: perspectives from north Florida. In: Ecology of mosquitoes: proceedings of a workshop (Lounibos LP, Rey JR, Frank JH, eds). Vero Beach, Florida: Florida Medical Entomology Laboratory; 3-23.
Chippindale AK, Hoang DT, Service PM, Rose MR, 1994. The evolution of development in Drosophila melanogaster selected for postponed senescence. Evolution 48: 1880-1899.
Gilliam JF, Fraser DF, 1987. Habitat selection under predation hazard: test of a model with foraging minnows. Ecology 68: 1856-1862.[Web of Science]
Goodnight CJ, Craig DM, 1997. The effect of coexistence on competitive outcome in Tribolium castaneum and Tribolium confusum. Evolution 50: 1241-1250.
Grill CP, Juliano SA, 1996. Predicting species interactions based on behaviour: predation and competition in container dwelling mosquitoes. J Anim Ecol 65: 63-76.
Grostal P, Dicke M, 1999. Direct and indirect cues to
predation risk influence behavior and reproduction of prey: a case for acarine
interactions. Behav Ecol 10:
422-427.
Hatcher L, Stepanski EJ, 1994. A step-by-step approach to using the SAS system for univariate and multivariate analyses. Cary, North Carolina: SAS Institute.
Hechtel LJ, Juliano SA, 1997. Effects of a predator on prey metamorphosis: plastic responses by prey or selective mortality? Ecology 78: 838-851.
Hoffman AA, 2000. Laboratory and field heritabilities. Some lessons from Drosophila. In: Adaptive genetic variation in the wild (Mousseau TA, Sinervo B, Endler J, eds). New York: Oxford University Press; 200-219.
Joshi A, Mueller LD, 1988. Directional and stabilizing density-dependent natural selection for pupation height in Drosophila melanogaster. Evolution 47: 176-184.
Juliano SA, 1996. Geographic variation in Aedes triseriatus (Diptera: Culicidae): temperature-dependent effects of a predator on survival of larvae. Environ Entomol 25: 624-631.
Juliano SA, Hechtel LJ, Waters JR, 1993. Behavior and risk of predation in larval tree hole mosquitoes: effects of hunger and population history of predation. Oikos 68: 229-241.
Juliano SA, Reminger L, 1992. The relationship between vulnerability to predation and behavior of larval tree hole mosquitoes: geographic and ontogenetic differences. Oikos 63: 465-476.
Lawler SP, 1989. Behavioral responses to predators and predation risk in four species of larval anurans. Anim Behav 381: 1039-1047.
Lima SL, Dill LM, 1990. Behavioral decisions made under the risk of predation: a review and prospectus. Can J Zool 68: 619-640.[Web of Science]
Linley JR, Darling K, 1993. Search behaviour associated with egg cannibalism in Toxorhynchites amboinensis and Toxorhynchites rutilus rutilus (Diptera: Culicidae). J Med Entomol 30: 561-570.[Web of Science][Medline]
Lounibos LP, Escher RL, Nishimura N, Juliano SA, 1997. Long-term dynamics of a predator used for biological control and decoupling from mosquito prey in a subtropical treehole ecosystem. Oecologia 111: 189-200.
Main KL, 1987. Predator avoidance in seagrass meadows: prey behavior, microhabitat selection, and cryptic coloration. Ecology 68: 170-180.
Martin P, Bateson P, 1986. Measuring behaviour: an introductory guide. Cambridge: Cambridge University Press.
McPeek MA, 1990. Behavioral differences between Enallagma species (Odonata) influencing vulnerability to predators. Ecology 71: 1714-1726.
Mousseau TA, Roff DA, 1987. Natural selection and the heritability of fitness components. Heredity 59: 181-197.
Mueller LD, 1988. Density-dependent population growth and natural selection in food limited environments: the Drosophila model. Am Nat 132: 786-809.[Web of Science]
Novak RJ, Shroyer DA, 1978. Eggs of Aedes triseriatus and A. hendersoni: a method to stimulate optimal hatch. Mosq News 38: 515-521.
Peckarsky BL, Wilcox RS, 1989. Stonefly nymphs use hydrodynamic cues to discriminate between prey. Oecologia 79: 265-270.
Rose MR, 1984. Laboratory evolution of postponed senescence in Drosophila melanogaster. Evolution 38: 1004-1010.[Web of Science]
Rose MR, Graves JL, Hutchinson EW, 1990. The use of selection to probe patterns of pleiotropy in fitness characters. In: Insect life cycles: genetics, evolution, and co-ordination (Gilbert F, ed). London: Springer-Verlag; 29-42.
Rose MR, Nusbaum TJ, Chippindale AK, 1996. Laboratory evolution: the experimental wonderland and the Cheshire cat syndrome. In: Adaptation (Rose MR, Laudner GV, eds). New York: Academic Press; 221-241.
Rubio Y, Rodríguez D, Machado-Allison CE, León JA, 1980. Algunos apectos del comportamiento de Toxorhynchites theobaldi (Diptera: Culicidae). Acta Cient Venez 31: 345-351.
Russo R, 1986. Comparison of predatory behavior in five species of Toxorhynchites (Diptera: Culicidae). Ann Entomol Soc Am 79: 715-722.
SAS Institute, 1990. SAS/STAT users guide, version 6, 4th ed., vols. 1 and 2. Cary, North Carolina: SAS Institute.
Scheiner SM, 1993. MANOVA: Multiple response variables and multispecies interactions. In: Design and analysis of ecological experiments (Scheiner SM, Gurevitch J, eds). New York: Chapman and Hall; 94-112.
Sih A, 1980. Optimal behavior: can foragers balance
two conflicting demands? Science 210:
1041-1043.
Sih A, 1986. Antipredator responses and the perception of danger in mosquito larvae. Ecology 67: 434-441.
Sih A, 1987. Predator and prey lifestyles: an evolutionary and ecological overview. In: Predation: direct and indirect impacts on aquatic communities (Kerfoot WC, Sih A, eds). Hanover, New Hampshire: University Press of New England; 203-224.
Sims SR, 1985. Embryonic and larval diapause in Aedes triseriatus: phenotypic correlation and ecological consequences of the induction response. In: Ecology of mosquitoes: proceedings of a worksho. (Lounibos LP, Rey JR, Frank, JH, eds.) Vero Beach, Florida: Florida Medical Entomology Laboratory; 359-369.
Skelly DK, Werner EE, 1990. Behavior and life historical responses of larval American toads to an odonate predator. Ecology 71: 2313-2322.[Web of Science]
Sokal RR, Rohlf FJ, 1995. Biometry, 3rd ed. New York: W.H. Freeman.
Steffan WA, Evenhuis NL, 1981. Biology of Toxorhynchites. Annu Rev Ecol Syst 26: 159-181.
Stein RA, Magnuson JJ, 1976. Behavioral response of a crayfish to a fish predator. Ecology 57: 751-761.
Tikkanen P, Moutka T, Huhta A, 1996. Fishless-stream mayflies express behavioural flexibility in response to predatory fish. Anim Behav 51: 1391-1399.
Watkins TB, 1996. Predator-mediated selection on burst swimming-performance in tadpoles of the Pacific tree frog, Pseudacris redilla. Physiol Zool 69: 154-167.
Wisenden BD, Cline A, Sparkes TC, 1999. Survival benefit to anti-predator behavior in the amphipod Gammarus minus (Crustacea: Amphipoda) in response to injury-released chemical cues from conspecifics and heterospecifics. Ethology 105: 407-414.
Woodward BD, 1983. Predator-prey interactions and breeding pond use of temporary pond species in a desert anuran community. Ecology 64: 1549-1555.
Zwaan B, Bijlsma R, Hoekstra RF, 1995. Artificial selection for development time in Drosophila melanogaster in relation to the evolution of aging: direct and correlated responses. Evolution 49: 635-648.
CiteULike 
Connotea 
Del.icio.us What's this?
This article has been cited by other articles:
|
|
 |
|
  M. V. Orr, K. Hittel, and K. Lukowiak `Different strokes for different folks': geographically isolated strains of Lymnaea stagnalis only respond to sympatric predators and have different memory forming capabilities J. Exp. Biol., July 15, 2009; 212(14): 2237 - 2247. [Abstract] [Full Text] [PDF]
|
|
|
|
 |
|
 A. Ridley and N. Raihani Facultative response to a kleptoparasite by the cooperatively breeding pied babbler Behav. Ecol., March 1, 2007; 18(2): 324 - 330. [Abstract] [Full Text] [PDF]
|
|
|
|
 |
|
 R. C. Fuller, C. F. Baer, and J. Travis How and When Selection Experiments Might Actually be Useful Integr. Comp. Biol., June 1, 2005; 45(3): 391 - 404. [Abstract] [Full Text] [PDF]
|
|
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||