Dyer and Bowers: Iridoid glycosides as defenses

The importance of sequestered iridoid glycosides as a defense against an ant predator.

1,2,*Lee A. Dyer and 2M. Deane Bowers

1,*(to whom correspondence should be addressed)

Biology Department

Mesa State College

Grand Junction, CO 81502

PHONE: (970) 248-1124

FAX: (970) 248-1700

EMAIL: ldyer@mesa5.mesa.colorado.edu

2University of Colorado Museum

and Dept. of EPO Biology

Campus Box 334

University of Colorado

Boulder, CO USA 80309

Abstract

We reared larvae of Junonia coenia Hubner (Nymphalidae) on artificial diets with trace concentrations of iridoid glycosides and on leaf diets with higher concentrations of iridoid glycosides. We offered these caterpillars to predacious ants and observed the effects of the following on predation: diet (artificial vs. leaf), site (ant colonies in dry vs. wet areas), instar (early vs. late), and time (changes in predation over 5 days). Diet and site were consistently significant predictors of the ants' propensities to reject prey and the caterpillars' abilities to escape predation. Leaf-diet caterpillars escaped more frequently than artificial-diet caterpillars, and ants from dry sites were more likely to reject prey than ants from wet sites. The percentage of iridoid glycosides found in individual caterpillars was also a good predictor of the probability of rejection by predators and prey escape. Caterpillars with higher levels of iridoids were more likely to be rejected and to escape, suggesting that sequestered iridoid glycosides are a defense against predacious ants.

Key words

Predation, iridoid glycosides, Junonia coenia, Formica planipiles, Plantago lanceolata, specialist herbivore, larval defenses, unpalatability, sequestration

Introduction

Several recent studies have shown that some specialist insect herbivores are better protected than generalists against invertebrate predators (Bernays and Cornelius, 1989; Bernays, 1988; Stamp, 1992; Dyer and Floyd, 1993). However, it is still unknown what mechanisms are responsible for this difference. One possibility is that some specialist insects sequester plant secondary compounds that may be used as a defense (Duffey, 1980; Brower, 1984; Bowers, 1991; Dyer, 1995). Although some generalist insects can sequester plant secondary compounds (e.g. arctiids, Rothschild 1977; Rothschild et al. 1979), and not all specialists can sequester such compounds (Bowers, 1992), this ability may be critical in protecting many specialists against natural enemies.

The efficacy of such chemical defenses may be affected by several factors. Certain predators, for example, exhibit intraspecific phenological (Hagen et al., 1976) or spatial (Dyer and Floyd, 1993) variation in their responses to various defenses. Also, other prey characteristics, such as developmental rate or body size, could either augment or weaken chemical defenses. It is important to consider confounding factors such as these when testing the overall effectiveness of a chemical defense.

In this study we examined the importance of iridoid glycosides in the larval diet of the Buckeye, Junonia coenia Hubner (Nymphalidae), as a deterrent to the ant, Formica rufa planipiles Creighton (Formicidae). We also examined the effects on predation of prey developmental stage, previous predator exposure to prey, and predator habitat type.

The iridoids, aucubin and catalpol, are sequestered by Buckeye caterpillars (Bowers & Collinge, 1992), presumably for defense against predators (Stamp, 1992; De la Fuente et al. 1994). These caterpillars are specialists and regularly feed on narrow-leaved plantain (Plantago lanceolata L.: Plantaginaceae) (Bowers, 1984) and on members of three other families of plants that contain iridoid glycosides (Bowers, 1991). Iridoid glycosides are terpene-derived compounds that are synthesized in plants by the mevalonic acid pathway (Bowers, 1991). These compounds generally are noxious to herbivores (Bowers, 1991) and are used as defenses by herbivores against predators (Nishida and Fukami, 1989). Narrow-leaved plantain contains the iridoids aucubin and catalpol (Duff et al., 1965; Bobbitt and Segebarth, 1969).

Formica planipiles is sympatric with J. coenia in Northern California, and Dyer has observed F. rufa-group species preying on Buckeye caterpillars. Formica rufa-group species can be voracious predators on a variety of prey (Skinner and Whittaker, 1981; Cherix and Bourne, 1980); one colony, for example, can harvest over 21,000 arthropod larvae in a single day (H"lldobler and Wilson, 1990).

We asked the following questions about the interactions between F. rufa and Buckeye caterpillars that have sequestered variable levels of iridoid glycosides:

1) Are Buckeye caterpillars that have sequestered high concentrations of aucubin and catalpol better protected against ant predation than caterpillars that have sequestered low concentrations of aucubin and catalpol? 2) Do Buckeye caterpillars become more unpalatable to ants as they reach later developmental stages (because of increased levels of sequestered compounds)? 3) Do ants learn to avoid unpalatable Buckeye caterpillars over short periods of time? 4) Do colonies of ants with different resource availabilities show different responses to the Buckeye caterpillars?

Methods

Three field experiments were conducted in August, 1992 at the Sierra Nevada Aquatic Research Laboratory (SNARL) which is situated 37o36'N 118o51'W in the Great Basin Desert at the base of the eastern Sierra Nevada. We used two sites at SNARL; one site was on the upper 1/3 of a glacial moraine, about 0.25 km south of SNARL, the second site was along Convict Creek, which is at the base of the moraine and runs through SNARL. We used ants from 8 colonies of F. planipiles as predators; for each site, we used 4 colonies within 50 m of each other. Before choosing colonies, we examined them to subjectively estimate the mound size (above ground volume of nest materials) and foraging activity (number of ants leaving and returning to the nest) and tried to select colonies with approximately the same sizes and foraging rates.

Larvae from eggs laid by female Buckeyes were collected from Buck Meadows, California (about 80 km west of SNARL). The ants were an appropriate species of predator for these experiments because they co-occur with Buckeyes in many sites near SNARL (including the Buck Meadows site). However, we used the SNARL colonies because they were na‹ve to the prey (the host plants do not occur at SNARL), thus their reactions to caterpillars with variable amounts of iridoids would not be affected by previous encounters. Caterpillars in instars 2 and 3 were classified as "early-instar" caterpillars, while those in instars 4 and 5 were classified as "late-instar" caterpillars.

Larvae were reared on two different diets: a leaf diet consisting of fresh, new P. lanceolata leaves, which contain from 4 to 8% dry weight aucubin and catalpol (Bowers and Stamp, 1993), and an artificial diet containing only trace quantities (less than 1% dry weight) of P. lanceolata. The plantain leaves used in the artificial diet were of all ages, including ages with lower levels of iridoid glycosides (see Bowers and Stamp, 1993). Thus the iridoid glycoside concentration of the artificial diet was approximately 1/100th that of the leaves fed to larvae, and probably even lower. The P. lanceolata leaves were collected from Buck Meadows, California. The artificial diet was based on a recipe by Nijhout (unpublished data), and included the following for one 53 g batch (dry weight): 2.5 g Wesson salts, 6.3 g sucrose, 16.3 g wheat germ, 8.3 g casein, 5 g yeast, 1.7 g cellulose, 340 mg cholesterol, 300 mg methyl paraben, 2 g vanderzandt vitamin mix, 500 mg ascorbic acid, 500 mg aureomycin, 330 mg sorbic acid, 5 g dried and powdered P. lanceolata leaves, 4 g agar, 1 ml linseed oil, and 227 ml water.

An additional 50 caterpillars were reared on artificial or leaf diets (25 on each diet) to determine mean iridoid glycoside concentrations for caterpillars reared on these different diets. For these caterpillars, and for caterpillars used in Experiment 3, iridoid glycoside content was determined using gas chromatography of the tri-methyl silylated derivatives of the iridoids (Gardner and Stermitz, 1988; Bowers and Collinge, 1992). Caterpillars were starved for 24 hours, lyophilized, and ground to a fine powder for analysis.

EXPERIMENT 1

Individual Buckeye caterpillars were offered to 6 of the 8 F. planipiles colonies by placing them along main foraging trails and within 5 m of the colony entrance. This method was not utilized to mimic natural encounters between ants and caterpillars, but to maximize predator-prey interactions in order to test the efficacy of iridoids against the ants, which was the major goal of this study. However, the design was not entirely artificial. Foraging F. planipiles can encounter Buckeyes on the ground along and near the foraging trails, because caterpillars often move along the ground to access new host plants and often are found on lower leaves (personal observations).

The experimental design was a full factorial, repeated over 5 successive days, with site (3 creek colonies vs. 3 moraine colonies), diet (artificial vs. leaf), and instar (early vs. late) as main effects. Each cell in the experimental design contained 3 replicates which were spatially (different foraging trails) or temporally (different times of day) separated. Each colony therefore received 12 caterpillars per day: 3 early-instar artificial-diet caterpillars, 3 late-instar artificial-diet caterpillars, 3 early-instar leaf-diet caterpillars and 3 late-instar leaf-diet caterpillars.

The experiments were conducted between 0600 and 1930 hours, but were usually restricted to hours of greatest ant activity (0700-1000 and 1500-1800). The first colony visited each day was selected randomly, followed by a predetermined visitation sequence among the colonies. The caterpillars were offered to ant colonies by selecting them haphazardly from pre-packaged and marked containers; each container had the diet, colony, and instar marked. Caterpillars from different diet treatments were matched for similar size and mass. Two caterpillars (on two different foraging trails) were given to each colony per visit.

Each offering, lasting at least 5 minutes, began when an ant encountered a caterpillar. We recorded the following response variables: number of rejections, number of ants touching the caterpillar at the end of each minute, and whether the caterpillar escaped or was eaten. A rejection consisted of an individual ant approaching, touching, and permanently leaving a caterpillar. It was obvious when the ant was leaving the caterpillar for recruitment (rather than rejecting it), because it would walk in growing circles which always brought it back to the prey. In these cases, the individual would eventually return with other ants. If at least one ant was already touching a caterpillar, no further rejections (i.e. another ant encountering and leaving the caterpillar) were counted. Caterpillars "escaped" when they were rejected by at least 5 ants or when they were not eaten within the 5 minute time period. The rejection and escape variables are different because caterpillars that avoided being eaten for 5 minutes were not always touched and rejected by the ants. The number of caterpillars completely rejected was therefore not equal to the number of caterpillars that escaped because a caterpillar that was never touched (and therefore escaped predation) was counted as a missing value for the rejection response variable. This distinction between the two variables is important because some caterpillars were completely ignored by the ants and therefore could never be considered "rejected" (even though they may have escaped predation). If a caterpillar moved too far from the foraging trail before the 5 minutes was over, it was returned to the trail. By 5 minutes it usually was obvious if the ants were going to take the caterpillar back to the nest.

Rejections were pooled into three categories: no rejections, some rejections (less than 5), and completely rejected (5 or greater); this pooling was based on preliminary studies with the ants and previous work with other species of ants (Dyer and Floyd, 1993; Dyer, 1995, unpublished data). The rejection categories and the escape response variable were dependent variables in logit models (see Christensen, 1990), which is the most appropriate method for testing associations between categorical dependent and independent variables (Feinberg, 1970; Christensen, 1990). These models allowed us to test both the main effects and interactions of the variables diet, site, day, and instar on ant response.

To test the effects of the independent variables on recruitment, the number of ants visiting a given caterpillar every minute was the dependent variable in a repeated measures analysis of variance, with time as the repeated measure and diet, site, and instar as fixed effects. To test the effects of the independent variables on possible changes in recruitment due to learning over a longer period of time, the numbers of ants visiting every minute were totaled over the five minute period. This total was used as a dependent variable in a repeated measures analysis of variance, with day as a repeated measure and diet, site, and instar as fixed effects. All means for numbers of ants visiting are reported ñ 1 SE.

EXPERIMENT 2

Because it is possible that Buckeyes might be protected by automimicry (sensu Brower, 1984), we compared predation by na‹ve ants that were offered caterpillars reared on both diets with that of na‹ve ants that were only offered caterpillars reared on artificial diets. Caterpillars were offered to 2 of the 8 colonies (neither of which was used for Experiment 1) during a separate 5 day period. The methods used were the same as those in Experiment 1, with the exception of the diet treatment; the two colonies (one creek and one moraine colony) received only caterpillars reared on artificial diets. We used Chi-square contingency tables to compare predation on artificial-diet caterpillars from colonies fed caterpillars reared on both diets (data from the 6 colonies in Experiment 1) with that from colonies fed only artificial-diet caterpillars.

EXPERIMENT 3

To examine correlations between iridoid glycoside concentrations and ant responses, we conducted a third experiment in which the caterpillars were retained for iridoid glycoside analyses. This experiment was conducted after the experiments described above were completed and was identical in design to the 5-day experiments with the following exceptions: the experiment was conducted over a two day period, only the three colonies at the creek site were used, only fifth instar larvae were used, and the caterpillars were taken from ants before they were dragged into the nests. Caterpillars were allowed to recover for 24 hours after the experiment and were fed their original diets during this time; this recovery period permitted caterpillars to regain iridoids lost in secreted hemolymph or regurgitant. To clear the gut, the caterpillars then were starved for another 24 hours and lyophilized.

To test for correlations between levels of iridoids and ant response, the rejection categories and the escape response variable were dependent variables in logistic regressions (see Christensen, 1990) with percent aucubin and percent catalpol as predictors. The number of replicates was too small to examine relationships between iridoids and number of ants visiting the caterpillars.

Results

Caterpillars reared on leaf diets had significantly higher concentrations (percentage dry weight) of iridoid glycosides than caterpillars reared on artificial diets (leaf-diet aucubin mean = 4.53%, artificial-diet aucubin mean = 0.36%, unequal variances t[39.3] = -7.74, P < 0.001; leaf-diet catalpol mean = 3.42%, artificial-diet catalpol mean = 0.12%, unequal variances t[40] = -8.78, P < 0.001). Late instar caterpillars had significantly higher percentages of aucubin than early instars, but there were no significant differences between instar categories in the concentrations of catalpol (late-instar aucubin mean = 3.03%, early-instar aucubin mean = 1.31%, unequal variances t[23.5] = -2.41, P = 0.024; late-instar catalpol mean = 2.19%, early-instar catalpol mean = 1.08%, t[67] = -1.34, P = 0.184).

EXPERIMENT 1

A total of 360 caterpillars were offered to 3 creek colonies and 3 moraine colonies. Day was never a significant predictor of the dependent variables, so it was excluded from subsequent models to keep cell sizes large enough for meaningful comparisons (see Christensen, 1990). Diet (X2[1] = 25.97, P < 0.001) and site (X2[1] = 22.69, P < 0.001) were significant predictors of the escape variable, while instar was not (likelihood ratio for unsaturated logit model: X2[8] = 12.88, P = 0.116). Leaf-diet caterpillars were much more likely to escape than were artificial-diet caterpillars (Figure 1) and creek colonies were much more likely to eat caterpillars than were moraine colonies (Figure 2). None of the higher order interactions were significant.

Diet (X2[2] = 44.44, P < 0.001) and site (X2[2] = 10.27, P = 0.006) were also significant predictors of rejection, while instar was not (likelihood ratio for unsaturated logit model: X2[12] = 12.64, P = 0.396). Leaf-diet caterpillars were more likely to be rejected than artificial-diet caterpillars (Figure 1) and moraine colonies were more likely to reject caterpillars than were creek colonies (Figure 2). None of the higher order interactions were significant.

The number of ants visiting a caterpillar increased significantly over time (Wilks' Lambda F[4,343] = 32.39, P < 0.001). There were, however, significant time by diet (Wilks' Lambda F[4,343] = 10.98, P < 0.001) and time by site (Wilks' Lambda F[4,343] = 6.20, P < 0.001) interactions; the number of ants visiting a caterpillar increased significantly more over time for artificial-diet caterpillars than for leaf-diet caterpillars, and the number of ants visiting a caterpillar increased significantly more over time for creek colonies than for moraine colonies (Figure 3). For the between subjects effects (which are hypotheses tests with the dependent variable equal to the total number of ants over the 5 time categories), there was a significant three-way interaction between diet, site and instar (F[1,346] = 4.40, P = 0.037), indicating that total ant visits (for the five minute time period) depended on specific combinations of diet, colony and instar. Late-instar, leaf-diet caterpillars offered to moraine colonies had the fewest ants visiting (7.0 ñ 0.8), while late-instar, artificial-diet caterpillars offered to creek colonies had the most ants visiting (19.9 ñ 0.9).

The total number of ants visiting caterpillars (for the five minute time period) did not change between days (F[4,340] = 0.47, P = 0.760). There were significant diet (F[1,340] = 86.02, P < 0.001), site (F[1,340] = 18.47, P < 0.001), and instar (F[1,340] = 4.11, P = 0.044) effects; total number of ants visiting was greatest for artificial diet caterpillars (16.9 ñ 0.7), creek colonies (15.2 ñ 0.7), and late instars (14.5 ñ 0.6).

EXPERIMENT 2

A total of 60 caterpillars reared on artificial diets were offered to 1 creek colony and 1 moraine colony. The contingency table for rejections of artificial-diet caterpillars offered to: 1) colonies that received only artificial-diet caterpillars and 2) colonies that received both artificial-diet and leaf-diet caterpillars, revealed no significant association between colony type and level of rejection (X2[2] = 4.009, P = 0.135). There was, however, a significant relationship between colony type and level of escape (X2[1] = 6.187, P = 0.013). Artificial-diet caterpillars offered to colonies that received only artificial-diet caterpillars never escaped, while 9% of the artificial-diet caterpillars escaped at colonies that received caterpillars reared on both diets.

EXPERIMENT 3

We offered a total of 68 caterpillars to 3 ant colonies at the creek site. Percent aucubin was a significant predictor of escape (X2[1] = 9.99, P = 0.002) and rejection (X2[2] = 14.84, P < 0.001). Likewise, percent catalpol was a significant predictor of escape (X2[1] = 7.31, P = 0.007) and rejection (X2[2] = 14.66, P < 0.001), indicating that iridoid glycoside concentration of larvae influenced predator response to these larvae. Caterpillars that escaped and were completely rejected had much higher iridoid concentrations than caterpillars that were eaten and had no rejections (Figure 4).

Discussion

The results of the present study indicate that diet is an important component of the suite of defenses found in the specialist caterpillar, J. coenia. Larvae reared on leaves of P. lanceolata, which contain aucubin and catalpol (4 to 8% dry weight), are better protected against a voracious predator than larvae reared on an artificial diet with less than 0.4% dry weight iridoid glycosides. Sequestered iridoids seem to be the main component of the leaf-diet caterpillars' enhanced defenses. While it is true that iridoid concentrations are not the only differences between the two diets (i.e. one diet is plant material, the other is an agar-based gel), chemistry was the only obvious difference between the caterpillars reared on these different diets. For example, one might expect that the size and behavior of caterpillars would vary depending on their diet, but the caterpillars offered to the ants were matched for size and there were no apparent differences in their activity levels. Furthermore, the concentrations of iridoids found in caterpillars on both diets were good predictors of whether or not those caterpillars escaped predation or whether or not they were rejected by individual ants.

The fact that instar number was never a significant predictor of prey escape or predator rejection suggests that the low percentages of iridoid glycosides found in early instars are as effective as the higher percentages (especially for aucubin) found in later instars in deterring predation by ants. These chemicals, therefore, seem to act as "qualitative" defenses against predators--high concentrations are not necessary for effective deterrence. Furthermore, certain invertebrate predators usually prefer smaller and earlier instar caterpillars as prey, including those that are chemically defended (Dyer and Floyd, 1993; also see Reavey 1993), so the protection of early instars by low concentrations of iridoids is further evidence of the efficacy of this defense.

Our results provide evidence of automimicry (sensu Brower, 1984) in J. coenia larvae. When ants were offered both palatable (artificial diet) and unpalatable (leaf diet) caterpillars of the same species, some palatable caterpillars were able to escape, while no caterpillars escaped when ants were offered only palatable caterpillars. Palatable caterpillars therefore were afforded some protection by the presence of unpalatable conspecifics. This is a rather weak example of automimicry, since most (>90%) of the palatable caterpillars were eaten by ants that had also been offered unpalatable caterpillars; additionally, there was no significant association between colony type and the level of rejection for palatable caterpillars. The difference in results between escape and rejection response variables may indicate that ants were ignoring or avoiding all caterpillars (regardless of diet) because they have received unpalatable Buckeyes. Such avoidance would allow caterpillars to escape, as opposed to being examined and rejected by ants; as a result, the numbers of rejections for the palatable caterpillars offered to the same colonies would be lower (because they were simply ignored).

Despite the effectiveness of iridoids as a qualitative defense against F. planipiles, some leaf-diet caterpillars and caterpillars with high iridoids were still eaten by ants. In addition, F. planipiles showed no short-term ability to learn to avoid Buckeye caterpillars, despite the fact that other species of predaceous ants are often able to learn prey avoidance quickly (Dejean, 1988; Bowers and Larin, 1989). The weak example of automimicry discussed above illustrates the ants' abilities to retain some sort of information within a 12-hour period, but there were no day-to-day changes in the colonies' responses to unpalatable caterpillars over a period of five days. This inability to retain information has the potential effect of minimizing selective pressures exerted by these and similar predators on distasteful prey, since there is not a consistent advantage provided by unpalatability. Aposematism, for example, is not likely to evolve in organisms exposed mostly to predators lacking the ability to learn or remember. Nevertheless, the presence of this chemical defense does provide a significant selective advantage over caterpillars that have not sequestered iridoids.

Ants from certain colonies were less likely to reject Buckeyes. The creek colonies, which have high prey resources (mostly creek invertebrates; Dyer, personal observations), were more voracious: they recruited more, rejected fewer caterpillars, and the caterpillars were less likely to escape from these colonies. In contrast, the moraine colonies, which spend most of their time tending homopterans (Dyer, personal observations; McIver and Loomis, 1993) were much less interested in Buckeyes as prey. This type of intraspecific variation is important to consider when conducting assays with predators or when examining the selective pressures exerted by specific predators. If our study had included only moraine colonies, a result such as iridoid sequestering caterpillars being no better defended than non-sequestering caterpillars might only reflect the fact that these particular colonies were not interested in preying on Buckeyes. However, because there were no significant diet by colony interactions in the present study, the intraspecific variation in levels of predation by ants did not indicate that predators with fewer available prey were more likely than those with abundant prey to attack unpalatable prey; a pattern one might expect if prey abundance is an important selective pressure on prey choice.

In conclusion, we have demonstrated the defensive utility of a secondary compound sequestered by a specialist herbivore, and this contributes to a better understanding of how specialist herbivores protect themselves against invertebrate predators. We have also demonstrated intraspecific variation in prey preferences of an invertebrate predator, and such variation should be considered in the increasing numbers of studies on the natural enemies of insect herbivores.

Acknowledgments:

This research was supported by a University of Colorado Fellowship to LAD, a Dean's Small Grant Award to LAD, and NSF Grant BSR-8905895 to MDB. We thank Dan Dawson for the use of a laboratory at SNARL. Mark Camara collected many of the adult Buckeyes and much of the plantain used in these experiments. Jim McIver pointed out and identified the colonies of Formica rufa planipiles. An earlier version of this manuscript benefited from the critical comments of Mark Camara and four anonymous reviewers.

References:

Bernays, E.A. 1988. Host specificity in phytophagous insects: selection pressure from generalist predators. Entomol. Exp. Appl. 49:131-140.

Bernays, E.A. and Cornelius, M.L. 1989. Generalist caterpillar prey are more palatable than specialists for the generalist predator Iridomyrmex humilis. Oecologia 79:427-430.

Bobbitt, J.M. and Segebarth, D.P. 1969. Iridoid glycosides and similar substances, pp. 1-145 in: Taylor, W.I. and A.R. Battersby (eds.). Cyclopentanoid Terpene Derivatives. Academic Press, New York.

Bowers, M.D. 1992. The evolution of unpalatability and the cost of chemical defense in insects, pp. 216-244 in: Roitberg, B.D. and M.B. Isman (eds.). Insect Chemical Ecology: An Evolutionary Approach. Chapman & Hall, New York.

Bowers, M.D. 1991. Iridoid glycosides, pp. 297-326 in: Rosenthal, G.A. and M.R. Berenbaum (eds). Herbivores: Their Interactions with Secondary Plant Metabolites. Academic Press, New York.

Bowers, M.D. 1984. Iridoid glycosides and host-plant specificity in larvae of the Buckeye butterfly, Junonia coenia (Nymphalidae). J. Chem. Ecol. 10:1567-1577.

Bowers, M.D. and Collinge, S.K. 1992. Fate of iridoid glycosides in different life stages of the buckeye, Junonia coenia (Lepidoptera: Nymphalidae). J. Chem. Ecol. 18:317-331.

Bowers, M.D. and Larin, Z. 1989. Acquired chemical defense in the lycaenid butterfly, Eumaeus atala. J. Chem. Ecol. 15:133-146.

Bowers, M.D. and N.E. Stamp. 1993. Effects of plant age, genotype, and herbivory on Plantago performance and chemistry. Ecology 74:1778-1791.

Brower, L.P. 1984. Chemical defense in butterflies, pp. 109-134 in: Vane-Wright, R.I. and P.R. Ackery (eds.). The Biology of Butterflies: Symposium of the Royal Entomology Society Number 11. Academic Press, New York.

Cherix, D. and Bourne, J.D. 1980. A field-study on a super-colony of the Red Wood Ant Formica lugubris Zett. in relation to other predatory Arthropods (spiders, harvestmen and ants). Rev. Suisse Zool. 87:955-973.

Christensen, R. 1990. Log-linear models. Springer Verlag, New York.

Dejean, A. 1988. Prey capture by Camponotus maculatus (Formicidae - Formicinae). Biol. Behav. 13:97-115.

De la Fuente, M.A., Dyer, L.A. and Bowers, M.D. 1994. The iridoid glycoside, catalpol, as a deterrent to the predator Camponotus floridanus (Formicidae). Chemoecology 5:13-19.

Duff, R.B., Bacon, J.S.D., Mundie, C.M., Farmer, V.C., Russell, J.D. and Forrester, A.R. 1965. Catalpol and methylcatalpol: naturally occurring glycosides in Plantago and Buddleia species. Biochem. J. 96:1-5.

Duffey, S.S. 1980. Sequestration of plant natural products by insects. Ann. Rev. Entomol. 25:447-477.

Dyer, L.A. 1995. Tasty generalists and nasty specialists? Antipredator mechanisms in tropical lepidopteran larvae. Ecology 76:1483-1496.

Dyer, L.A. and Floyd, T. 1993. Determinants of predation on phytophagous insects: the importance of diet breadth. Oecologia 96:575-582.

Feinberg, S.E. 1970. The analysis of multidimensional contingency tables. Ecology 51:417-433.

Gardner, D.R. and Stermitz, F.R. 1988. Hostplant utilization and iridoid glycoside sequestration by Euphydryas anicia individuals and populations. J. Chem. Ecol. 15:2147-2168.

Hagen, K.S., Bombosch, S., and McMurtry, J.A. 1976. The biology and impact of predators, pp. 93-142 in: Huffaker, C.B. and P.S. Messenger (eds.). Theory and practice of biological control. Academic Press, New York.

H"lldobler, B. and Wilson, E.O. 1990. The Ants. P. 534. Belknap Press, Cambridge, MA.

McIver, J.D. and Loomis, C. 1993. A size-distance relation in Homoptera-tending thatch ants (Formica obscuripes, Formica planipilis). Insectes Sociaux 40:207-218.

Nishida, R. and Fukami, H. 1989. Host plant iridoid-based chemical defense of an aphid, Acyrthosiphon nipponicus, against ladybird beetles. J. Chem. Ecol. 15:1837-1845.

Reavey, D. 1993. Why body size matters to caterpillars, pp. 248-279 in: Stamp, N.E. and T.M. Casey (eds.). Caterpillars: Ecological and Evolutionary Constraints on Foraging. Chapman & Hall, New York.

Rothschild, M. 1977. Storage of cannabinoids by Arctia caja and Zonocerus elegans fed on chemically distinct strains of Cannabis sativa. Nature 266:650-651.

Rothschild, M., R.T. Aplin, P.A. Cockrum, J.A. Edgar, P. Fairweather, and R. Lees. 1979. Pyrrolizidine alkaloids in arctiid moths (Lep.) with a discussion on host plant relationships and the role of these secondary plant substances in the Arctiidae. Biol. J. Linn. Soc. 12:305-326.

Skinner, G.J. and Whittaker, J.B. 1981. An experimental investigation of inter-relationships between the wood-ant (Formica rufa) and some tree-canopy herbivores. J. Animal Ecol. 50:313-326.

Stamp, N.E. 1992. Relative susceptibility to predation of two species of caterpillar on plantain. Oecologia 92:124-131.

Figure Captions

Fig. 1. The association between larval diet and predation response variables. "Leaf" refers to caterpillars reared on Plantago lanceolata, while "artificial" refers to caterpillars reared on an artificial diet with trace quantities of P. lanceolata. The y-axes represent percentages of all larvae (n = 360) with specific diets that received the predator response indicated on the x-axis. For example, 91% of the artificial-diet caterpillars offered to the ants were eaten. The predator responses on the x-axis were categorical dependent variables in logit models.

Fig. 2. The association between predator habitat type and predation response variables. "Moraine" refers to ant colonies located on a glacial moraine, while "creek" refers to colonies located adjacent to a creek below the moraine. The y-axes represent percentages of all larvae (n = 360) offered to predators from specific habitats (n = 6 ant colonies) that received the predator response indicated on the x-axis. For example, 40% of the caterpillars offered to ant colonies on the glacial moraine were never rejected. The predator responses on the x-axis were categorical dependent variables in logit models.

Fig. 3. The effects of larval diet and predator habitat site on ant recruitment. The Y-axes represent mean number of ants visiting caterpillars reared on specific diets and mean number of ants from specific sites visiting caterpillars. The number of ants for each minute is stacked upon the previous minute. The standard error bars are for the numbers of ants totaled over the five minutes.

Fig. 4. The effects of iridoid glycoside concentrations (percent dry weight of aucubin and catalpol) on predation response variables. The Y-axes represent mean percent iridoids sequestered by caterpillars that received the predator response indicated on the x-axis. For example, the mean percent dry weight of aucubin sequestered by caterpillars that escaped was 0.085%. The standard error bars are for total iridoids. The predator responses on the x-axis were categorical dependent variables in logit models.