Host Effects on Herbivory and Pollination in a Hemiparasitic Plant

Ecology, 83(10), 2002, pp. 2700±2710 ᭧ 2002 by the Ecological Society of America

HOST EFFECTS ON HERBIVORY AND POLLINATION IN A HEMIPARASITIC PLANT

LYNN S. ADLER1 Center for Population Biology and Department of Entomology, University of California, Davis, California 95616 USA

Abstract. The indirect effects of hosts on interactions between parasites and other species are not well understood, and it may be dif®cult to predict the outcome of host species effects on parasite performance due to the complexity of potential direct and indirect effects. For example, parasitic plants obtain defensive compounds as well as nutrients from their hosts, and thus many attributes of parasitic plants are dependent on the quality of their host species. Here I measure the effect of a lupine host species (Lupinus argenteus) compared to other host species on herbivory, pollination, and female plant ®tness in the hemiparasite Indian paintbrush (Castilleja miniata) using a series of ®eld experiments. Association with lupine was determined in the ®eld by assaying Indian paintbrush leaves for lupine alkaloids. I found that Indian paintbrush plants parasitizing lupines experienced reduced herbivory from plume moth larvae, agromyzid ¯y larvae, and deer, relative to Indian paintbrush plants parasitizing other host species. However, there was no correlation between alkaloid content of in¯orescences and plume moth performance. Host species did not affect pollinator preference for Indian paintbrush in the ®eld. Indian paintbrush parasitizing lupines produced twice as many seeds overall as Indian paintbrush parasitizing other host species. Correlations suggest that this bene®t arises both from reduced herbivory and increased nitrogenous resources. The reduction of herbivory in Indian paintbrush plants parasitizing lupines indicates that host species can affect performance of hemiparasites via indirect pathways, and that the larger community of herbivores could alter the impact of a host on its plant parasites. Key words: Castilleja miniata; hemiparasitic plants; herbivory; host±parasite interactions; Indian paintbrush; indirect effects; lupine; Lupinus argenteus; plant chemical defense; pollination; quinolizidine alkaloids.

INTRODUCTION Graves 1998), different hosts can also affect their par- All organisms experience simultaneous, and at times asites by both direct and indirect pathways. Because con¯icting, selection pressures resulting from direct parasitic plants acquire both nutrients (Press and and indirect interactions with their biotic environment. Graves 1995) and defensive compounds (Arslanian et Indirect effects occur when the interaction between two al. 1990, Boros et al. 1991, Mead et al. 1992, Stermitz species is modi®ed by the presence of a third species and Pomeroy 1992, Stermitz et al. 1993) from their (Miller and Travis 1996), whereas direct effects involve hosts, many attributes of parasitic plants are dependent only two species in a pairwise interaction. Because on the qualities of their hosts. Thus, there is the po- indirect effects may occur in the same or opposite di- tential for tritrophic interactions between host plants, rection as direct effects, the combined outcome of mul- parasitic plants, and herbivores or mutualists of the tiple species interactions on a focal individual may be parasite, but the overall impact of a host on a parasitic dif®cult to predict from studying only pairwise inter- plant has rarely been addressed in the context of other, actions in the absence of a community context (Strauss interacting species. 1991, Schoener 1993, Wootton 1994). One of the most apparent ways that host species could affect herbivores of parasitic plants is by the Parasitic plants are diverse (Ͼ3000 species), are present in every major ecosystem (Kuijt 1969, Press and uptake of defensive compounds. Alkaloid uptake in one Graves 1995), and have strong impacts on community hemiparasite species, Castilleja indivisa, has been ex- structure and dynamics (Pennings and Callaway 1996, perimentally manipulated in the ®eld by growing plants with high-alkaloid and low-alkaloid lines of the host Marvier 1998b). While parasites often have strong neg- Lupinus albus (Adler 2000a). By using lines of the ative impacts on their hosts (e.g., Gibson and Watkin- same host species that varied in alkaloid production, son 1991, Graves 1995, Matthies 1995, Seel and Press the impact of alkaloid uptake was isolated from other 1996, Silva and Martinez Del Rio 1996, Davies and effects of parasitizing different host species. Castilleja indivisa grown with high-alkaloid hosts had reduced herbivory, increased pollinator visits, and increased 1 Current address: Department of Biology, Virginia Tech, lifetime seed set compared to C. indivisa parasitizing Blacksburg, Virginia 24061 USA. E-mail: [email protected] low-alkaloid hosts (Adler 2000a). Path analysis dem- 2700 October 2002 HERBIVORY AND POLLINATION IN CASTILLEJA 2701 onstrated that pollinators were not directly attracted to year period at the Rocky Mountain Biological Labo- alkaloids, but rather that alkaloids, by reducing bud ratory in Colorado, USA. Speci®cally, I addressed the herbivory, increased the number of open ¯owers on following questions: alkaloid-containing plants and thereby increased pol- 1) How is damage by the principal herbivores of linator attraction (Adler et al. 2001). Indian paintbrush affected by parasitizing a lupine, The previous study isolated the effect of one factor, compared to other host species? alkaloid content, on herbivores, pollinators, and life- 2) Is herbivore performance correlated with alkaloid time seed set of a parasitic plant. However, in the ®eld, content of Indian paintbrush? generalist parasitic plants are presented with a range 3) How is pollinator visitation in Indian paintbrush of host species that vary in many aspects including affected by parasitizing a lupine host, compared to oth- defensive chemistry. Lupine species are common hosts er host species? of several parasitic plant species (Stermitz and Harris 4) How is Indian paintbrush performance (biomass, 1987, Stermitz et al. 1989, Arslanian et al. 1990, seed production, seed mass, and nitrogen content) af- Schneider and Stermitz 1990, Boros et al. 1991, Baeu- fected by parasitizing a lupine host, compared to other mel et al. 1992, Stermitz and Pomeroy 1992, Jeschke host species? et al. 1994, Marvier 1995, 1998a, Adler and Wink Taken together, these studies quantify the net impact 2001), and they differ from nonleguminous hosts in of lupine hosts on a plant parasite and investigate the their ability to ®x nitrogen (via association with sym- importance of indirect effects of host species on a par- bionts) as well as in their alkaloid content. The impact asitic plant via changes in herbivory and pollination. of lupines or other alkaloid-containing host species on herbivores of a parasitic plant has only been assessed METHODS in a handful of studies to date, each with different Study system conclusions (Stermitz et al. 1989, Marko et al. 1995, Marvier 1995, 1998a). No study has yet considered the Castilleja miniata Douglas (Scrophulariaceae; In- effect of different host species on pollinators of a par- dian paintbrush), and Lupinus argenteus Pursh (Fa- asitic plant. baceae; lupine) are long-lived perennials that grow The indirect effects of hosts on interactions between abundantly at Emerald Lake (Gunnison County, Col- hemiparasites and other species are not well under- orado, USA; elevation 3000±3300 m), ϳ10 km from stood, and it may be dif®cult to predict the outcome the Rocky Mountain Biological Laboratory (RMBL). of host species effects on parasite performance in a The major habitat of C. miniata is subalpine meadows community context due to the complexity of potential surrounded by subalpine ®r forest; the dominant veg- direct and indirect effects. For example, the availability etation in these meadows is grasses and perennial forbs. of nitrogen from a host species might bene®t a parasite The growing season is July±August. Three sites were by providing greater resources for growth, defense, chosen based on accessibility and the abundance of and/or reproduction (Chapin 1980). However, in- Indian paintbrush (Ն75 individuals/site). Sites were creased nitrogen might also make hemiparasites more Ն200 m in diameter and in discrete meadows within palatable to their herbivores (Marvier 1995, Kyto et al. 500 m of each other, separated by wide stands of ®r 1996) and thus be detrimental to hemiparasite ®tness. trees. Sites 1 and 2 were adjacent (elevation 3336 m, Similarly, secondary compounds might bene®t a hem- above the high road on the east slope; site 1 at iparasite by increasing resistance to herbivory, which 39Њ00Ј42Љ N, 107Њ02Ј29Љ W; site 2 at 39Њ00Ј44Љ N, might also make plants more attractive to pollinators 107Њ02Ј32Љ W); site 1 was open, and site 2 was more (Adler 2000a, Adler and Wink 2001). However, these shaded. Site 3 was lower, sunny, and moister due to a compounds may also be detrimental if they are dam- small adjacent stream (3150 m, between the low and aging to the hemiparasite (McKey 1974, Chew and high roads on east slope; 39Њ00Ј41Љ N; 107Њ02Ј32Љ W). Rodman 1979, Fowden and Lea 1979) or deter mutu- Voucher specimens of C. miniata and L. argenteus from alists such as pollinators (Strauss et al. 1999). each site have been deposited at the RMBL Herbarium. Castilleja miniata, or Indian paintbrush, is a gen- The major potential host species in order of relative eralist hemiparasite that parasitizes many host species, abundance were: (site 1) various grasses (including including the native lupine Lupinus argenteus. Indian Bromus carinatus, Dactylis glomerata, Deschampsia paintbrush individuals obtain quinolizidine alkaloids cespitosa, Elymus trachycaulus, Festuca thurberi, Koe- when parasitizing several species of lupines, but not leria macrantha, Melica spectabilis, Phleum pratense, when parasitizing other host species (McCoy and Ster- Poa fendleriana, and Poa pratense), Fragaria virgi- mitz 1983, Stermitz et al. 1986, Stermitz and Harris niana (Rosaceae), Senecio integerrimus and S. cras- 1987, Adler and Wink 2001). To determine the effect sulus (Asteraceae), Thalictrum fendleri (Ranuncula- of different host species on Indian paintbrush perfor- ceae), Helianthella quinquenervis (Asteraceae), Ge- mance and on the interactions between Indian paint- ranium richardsonii (Geraniaceae), and Delphinium brush and its herbivores and pollinators, ®eld obser- barbeyi (Ranunculaceae); (sites 2 and 3): various grass- vations and experiments were conducted over a three- es (as for site 1), Heracleum lanatum (Apiaceae), Aq- 2702 LYNN S. ADLER Ecology, Vol. 83, No. 10 uilegia caerulea (Ranunculaceae; not in site 3), Poten- Herbivory tilla gracilis (Rosaceae), F. virginiana, Angelica grayi To determine the effect of host on herbivory, I sam- (Apiaceae), S. integerrimus and S. crassulus, T. fen- pled a total of 65 ¯owering Indian paintbrush individ- dleri, H. quinquenervis, and Mertensia ciliata (Bora- uals from all three sites in 1996 (20, 25, and 17 plants/ ginaceae). Although this list provides a general idea of site, respectively), and 60 individuals from sites 1 and the plant community, these species may not be para- 2 in 1997 (30 plants/site). Individuals parasitizing lu- sitized at a rate proportional to their abundance, since pines were identi®ed using Dragendorff reagent. The hemiparasites can preferentially parasitize some spe- two major herbivores, plume moth larvae and agro- cies over others in the ®eld (Gibson and Watkinson myzid ¯y larvae, both feed on in¯orescences and leave 1989, Nilsson and Svensson 1997). distinctive damage. Fly larvae remain in one fruit and Castilleja miniata individuals grow in discrete ro- tunnel through seeds until they emerge at pupation settes. In all experiments, the presence of alkaloids in ( personal observation), while plume moth larvae con- Indian paintbrush individuals was determined early in sume ¯owers, fruit, and seeds, leaving behind distinc- the ®eld season using Dragendorff reagent, which tive frass. I measured herbivory by assessing each ini- changes color in the presence of quinolizidine alkaloids tiated ¯ower on the longest in¯orescence. I recorded (Harborne 1984, Stermitz et al. 1989). A positive Dra- the number of seeds and type of damage (none, ¯y, or gendorff test indicated that Indian paintbrush were par- moth) for each fruit. I evaluated early bud damage us- asitizing L. argenteus, since this reagent does not detect ing the presence of bud scars and frass. One potential alkaloids from any other host species at these sites source of bias is that plume moth larvae may obscure (personal observation). Plants were sampled until an prior ¯y damage by consuming seeds with ¯y damage; even number was obtained of Indian paintbrush indi- there was a signi®cant negative correlation between the viduals testing positive and negative for alkaloid pres- proportion of fruits scored with damage by plume moth ence. If alkaloids were not detected, Indian paintbrush and ¯y larvae (␳ϭϪ0.34, P Ͻ 0.001, n ϭ 113). Thus, was considered to be parasitizing nonlupine host spe- I may have underestimated the importance of ¯y dam- cies. It is not possible to determine the hosts of Indian age. Deer browsed several plants at site 2 in 1997, and paintbrush that are not parasitizing lupines, since dig- the number of in¯orescences consumed per plant was ging out roots would be both laborious and destructive. recorded. Indian paintbrush parasitizing lupines and other host Herbivory by plume moths and ¯ies was measured species were interspersed at each ®eld site; there were as the proportion of fruit damaged per in¯orescence no distinct patches of Indian paintbrush with and with- and the number of seeds consumed per damaged fruit. out detectable alkaloids. I estimated the number of seeds consumed in each dam- Indian paintbrush experience little leaf herbivory, aged fruit using the mean number of seeds produced but have two common seed predators: larvae of the per undamaged fruit for each plant. Herbivory was an- plume moth Amblyptilia (Platyptila) pica Walsingham alyzed in a three-way ANOVA with host (lupine vs. (Pterophoridae) (Lange 1950, McCoy and Stermitz other hosts) and year as ®xed main effects, and site as 1983, Roby and Stermitz 1984, Stermitz et al. 1986) a random factor (Littell et al. 1991). However, only a and larvae of the ¯y Phytomyza subtenella (Agromy- subset of plants (50 of 122) had plume moth damage, zidae; S. Scheffer, L. S. Adler, and C. Gratton, unpub- and these were unbalanced across sites and years. lished data). Amblyptilia pica is a specialist on Cas- Therefore, only main effects were analyzed for this tilleja species, and Phytomyza subtenella is in a clade variable. Because deer browsed only one site in one that feeds largely on parasitic Scrophulariaceae (S. year, the effect of host on deer browsing was analyzed Scheffer, personal communication). Broad-tailed and with a separate one-way ANOVA. I also calculated the Rufous Hummingbirds (Selasphorus platycercus and S. correlation of plume moth and ¯y damage with per- rufus) are the predominant pollinators of Indian paint- centage alkaloid content of Indian paintbrush in¯ores- brush ( personal observation). Indian paintbrush and cences. In all analyses, data were transformed when lupines did not share any major pollinators or herbi- necessary to meet the assumptions of normality using vores during the course of this study ( personal obser- an arcsine square-root transformation for proportional vation). measures and a logarithmic transformation for numer- Because C. miniata plants are long-lived subalpine ical data. All ANOVA and MANOVA tests were per- perennials and could not be manipulated to grow suc- formed using the GLM procedure of SAS (Littell et al. cessfully with different hosts, I performed observations 1991), and correlations were calculated with the CORR and manipulations on established Indian paintbrush in procedure. these ®eld studies. Although I could not control for To determine the relationship between herbivore per- histories or microsite differences between individual formance and alkaloid content of Indian paintbrush, I plants, this study documents comparative differences chose 97 plants at site 1 in 1998. On 24 July, I collected in herbivory, pollination, and performance in Indian plume moth larvae from Indian paintbrush in a nearby paintbrush parasitizing lupines vs. other host species. meadow without lupines, randomly assigned one larvae October 2002 HERBIVORY AND POLLINATION IN CASTILLEJA 2703 to one undamaged in¯orescence per plant, and placed plants. The latter two variables were analyzed using larvae within ®ne mesh organdy bags (13 ϫ 18 cm). MANCOVA, and number of visits was analyzed sep- On 30 July, 12 and 15 Aug, I recorded the status of arately using ANCOVA because of the larger sample each larva (i.e., dead, missing, alive, pupated), and size. Independent variables were host (lupine vs. other collected and measured the masses of all pupae to the hosts), year, and the host ϫ year interaction (all ®xed nearest 0.1 mg. On 15 August I also collected between effects), with number of in¯orescences and plant height two and ®ve in¯orescences per plant for alkaloid anal- as covariates. ysis. In¯orescences were not collected during the experiment due to the large tissue samples needed for Plant growth and seed production alkaloid analysis and the possibility of inducing chang- The same plants used for the herbivory study were es that could affect herbivory (Karban and Baldwin also used to measure plant biomass and female ®tness. 1997). I stored pupae in 30-mL (one-ounce) plastic When fruits matured, I collected plants (including ma- portion cups (Sweetheart Cups, Chicago, Illinois, USA) jor tap root but not ®ne roots) and measured the number at room temperature until emergence. Of the 97 initial of seeds per undamaged fruit and fruit production for larvae, 73 pupated and 52 emerged as adults. the longest in¯orescence. There were 1±16 in¯ores- I quanti®ed alkaloids using gas chromatography. Tis- cences/plant (mean 4.6), 3±29 fruits/in¯orescence sue was air-dried for several weeks and then ground (mean 13.6), and 5±151 seeds/undamaged fruit (mean with a Krups type 203 coffee mill (Krups, Peoria, Il- 65.3). Due to the time constraints for counting this linois, USA). Alkaloids were extracted following the many seeds (an estimated 857 Ϯ 106 seeds/plant), I methods of Johnson et al. (1989). The methylene chlo- used one in¯orescence to estimate seed production per ride extract was injected into a HP 5890A gas chro- plant. Total seed production was estimated by multi- matograph (Hewlett Packard, Wilmington, Delaware, plying the number of seeds produced per centimeter of USA) with a DB-1 megapore capillary column (30 m, in¯orescence by the total length of in¯orescences. I 0.25 mm inner diameter, 0.25 ␮m ®lm thickness; J&W used the longest in¯orescence, rather than a randomly Scienti®c, Folsom, California, USA). Alkaloids were chosen in¯orescence, to provide the most data per in- eluted by temperature programming progressing ¯orescence, and to increase the chance of ®nding un- through 200±300ЊC over a 10-min period, followed by damaged fruits to estimate seed production. While us- ®ve minutes at 300ЊC. Alkaloids were identi®ed by ing the longest in¯orescence may have introduced a comparison to known standards (Johnson et al. 1989). bias into the results (e.g., if longer in¯orescences have I used linear regression to determine the relationship lower herbivory or larger seeds), the practice was con- between alkaloid concentration and plume moth pupal sistent across samples. In 1996, I measured the mass mass (wet). The relationship between larval status of a group of 20 seeds from one undamaged fruit per (dead or pupated), pupal eclosion, and alkaloid con- plant to estimate mean seed mass. centration at the end of the experiment was analyzed Plants are often limited by nitrogen acquisition with one-way ANOVA. (Chapin 1980), and one of the bene®ts for Indian paintbrush of parasitizing a lupine host may be the uptake Pollination of ®xed nitrogen. I measured nitrogen concentration in To determine the effect of host species on pollinator vegetative tissues of 30 plants sampled across sites in visitation, I observed pollinator visits to Indian paint- 1996. Nitrogen analysis is destructive due to the large brush for two seasons. I selected 39 and 62 ¯owering amount of tissue required; therefore I collected samples Indian paintbrush at site 3 in 1997 and 1998, respec- at the end of the ®eld season. Samples were dried at tively. Using Dragendorff reagent, I identi®ed Indian 50ЊC for one week and ground to pass through a 40- paintbrush parasitizing lupines, and I bagged plants mesh screen using a Wiley Mill (Thomas Scienti®c, with loose mesh to prevent removal of nectar when Swedesboro, New Jersey). Percent dry mass of nitrogen unobserved and to deter herbivory. Herbivores were was measured with a Carlo-Erba combustion gas an- also regularly removed by hand to minimize effects of alyzer (Pella 1990a, b) at the University of California herbivory on pollinator preference. I observed all Division of Agriculture and Natural Resources lab. I plants at this site simultaneously during the morning also measured alkaloid concentration (see Methods: when pollinators were most active. I recorded all hum- Herbivory), and subtracted the proportion of nitrogen mingbird visits, the number of ¯owers probed, and the in the predominant alkaloids (Adler and Wink 2001) time per visit. I observed ¯owering plants for a total from total nitrogen concentration to determine how of 30 h in 1997 and 20 h in 1998. In addition, I mea- much nitrogen could be ascribed to sources other than sured plant height and number of in¯orescences twice alkaloids. I calculated the correlation between nonal- during each ®eld season and averaged by plant within kaloid nitrogen concentration and Indian paintbrush year for use as covariates. biomass and seed production. Pollinator preference was measured as the number Biomass and seed number were analyzed using of visits to each plant, and the number of ¯owers probed MANOVA with the same model used for the herbivory per visit and time per ¯ower on the subset of visited study. This model was also analyzed with plume moth 2704 LYNN S. ADLER Ecology, Vol. 83, No. 10

TABLE 1. Effect of host species (lupine vs. other species), site, and year on herbivory by plume moth and agromyzid ¯y larvae in Castilleja miniata.

Plume moth Agromyzid ¯y Fruits Seeds Fruits Seeds damaged consumed damaged consumed

Effect df SS F SS F SS F SS F Host³ 1 0.48 15.93² 2.64 7.34** 0.002 0.08 1.56 1454.4*** Year§ 1 0.02 0.05 0.06 0.16 0.31 0.89 1.61 6.82 Site 2 0.26 1.04 1.34 1.86 0.16 0.99 0.44 0.39 Host ϫ Site 2 0.06 0.24 ´´´ ´´´ 0.06 0.41 0.002 0.00 Host ϫ Year࿣ 1 0.03 0.70 ´´´ ´´´ 0.38 1.98 3.16 2.76 Year ϫ Site 1 0.30 2.40 ´´´ ´´´ 0.35 4.44* 0.24 0.42 Host ϫ Year ϫ Site 1 0.04 0.33 ´´´ ´´´ 0.19 2.44 1.15 2.05 Error 14.28 ´´´ ´´´ ´´´ 9.62 ´´´ 46.34 ´´´ (103) (103) (92) Notes: Site was considered a random effect, and year and host species were ®xed effects. Measures of herbivory were the proportion of fruits damaged per in¯orescence and the number of seeds consumed per damaged fruit. The sample size for seeds consumed per fruit by plume moths was too small to calculate interaction terms, so only main effects were tested. Proportional data were arcsine square-root transformed, and numbers of seeds were log-transformed for analysis. Error df are in parentheses below error SS. * P Ͻ 0.05; ** P Ͻ 0.01; *** P Ͻ 0.001. ² P ϭ 0.0574. ³ Tested over the host ϫ site interaction term (except seeds consumed by plume moths). § Tested over the year ϫ site interaction term (except seeds consumed by plume moths). ࿣ Tested over the host ϫ year ϫ site interaction term (except seeds consumed by plume moths). and agromyzid ¯y damage (proportion of fruits dam- In the separate experiment examining plume moth aged) included as covariates to determine if the effect performance, alkaloid concentration did not correlate of host species on biomass or seeds was due to her- with plume moth pupal mass (linear regression; pupal bivory. I performed univariate tests when MANOVA mass: ␤ϭ10.23, P ϭ 0.1559, n ϭ 72), and there was analysis revealed signi®cant results of host as a main no effect of alkaloid concentration on eclosion from effect or interaction term (e.g, host ϫ site). Seed mass pupae to adults (F1,80 Ͻ 0.001, P Ͼ 0.95). and whole-plant nitrogen were only measured in one year and were analyzed with two-way ANOVA. Pollination Host species did not signi®cantly affect the number RESULTS of pollinator visits to Indian paintbrush in the ®eld

Herbivory (Table 2; MANCOVA: Wilks' ␭ϭ0.97, F2,64 ϭ 0.894, Indian paintbrush that were parasitizing lupines ex- P ϭ 0.4141). The number of observed visits to plants perienced decreased herbivory compared to Indian was greater for plants with more in¯orescences (Table paintbrush parasitizing other host species. Both plume 2) and also varied between years (Fig. 2; MANCOVA moth and ¯y larvae ate signi®cantly fewer seeds per host ϫ year interaction: Wilks' ␭ϭ0.89, F2,64 ϭ 4.00, damaged fruit (Table 1, Fig. 1), and plume moth larvae P ϭ 0.0232). However, when the data were analyzed damaged marginally fewer fruits per in¯orescence (P separately for each year, there were no signi®cant ef- ϭ 0.0574; Table 1, Fig. 1a). Deer also browsed three fects of host species on any measure of pollinator pref- times as many in¯orescences of Indian paintbrush par- erence (F Ͻ 3.0, P Ͼ 0.1). asitizing nonlupine host species, compared to Indian Plant growth and seed production paintbrush parasitizing lupines (mean in¯orescences browsed Ϯ 1 SE: lupine hosts, 0.71 Ϯ 0.30; other hosts Combining biomass and seed production in a MAN-

2.19 Ϯ 0.56; F1,15 ϭ 9.87, P ϭ 0.0078). OVA revealed a signi®cant effect of host species that Percent alkaloid content was negatively correlated varied across sites, and a signi®cant site effect (site, with the number of seeds consumed per fruit by plume Wilks' ␭ϭ0.78, F6, 198 ϭ 4.41, P Ͻ 0.001; host ϫ site, moths (␳ϭϪ0.33, P ϭ 0.026, n ϭ 47), but only mar- Wilks' ␭ϭ0.85, F6, 198 ϭ 2.70, P ϭ 0.015). However, ginally negatively correlated with the number of seeds there was no signi®cant main effect of host species on consumed per fruit by ¯y larvae (␳ϭϪ0.2, P ϭ 0.06, biomass and seed production in the MANOVA (Wilks' n ϭ 87). There was no signi®cant correlation between ␭ϭ0.04, F2,1 ϭ 12.78, P ϭ 0.2), although Indian alkaloid content and the proportion of fruits damaged paintbrush plants were on average 47% larger and pro- by either herbivore (P Ͼ 0.25 for both herbivores, n duced over twice as many seeds when parasitizing lu- ϭ 105). pines than when parasitizing other host species. Be- October 2002 HERBIVORY AND POLLINATION IN CASTILLEJA 2705

FIG. 1. Effect of host plant on herbivory in the hemiparasite Castilleja miniata by (a) plume moth larvae and (b) agromyzid ¯y larvae. Herbivory was measured as the proportion of fruits damaged per in¯orescence and the number of seeds consumed per damaged fruit. Data for each graph are averaged over all other main effects. Error bars represent Ϯ1 SE, and numbers in columns represent sample size. ²P Ͻ 0.06; **P Ͻ 0.01; ***P Ͻ 0.001; NS ϭ nonsigni®cant.

cause the host ϫ site interaction was signi®cant in the F1,99 ϭ 6.77, P ϭ 0.01, respectively). When these co- MANOVA, univariate analyses were examined; these variates were included, the effect of host species on showed signi®cant site and host ϫ site effects on bio- seed production was only marginally signi®cant (F1,1 mass (Table 3, Fig. 3), and a signi®cant main effect of ϭ 11.25, P ϭ 0.08), suggesting that the increased seed host species on seed production (Table 3, Fig. 4a). In- production of Indian paintbrush parasitizing lupine cluding plume moth and ¯y herbivory as covariates in hosts is due partially to a reduction in herbivory. the model did not substantially change results for bio- There was no effect of host species or site on indi- mass, but did in¯uence the analysis of seed production. vidual seed mass (F1,53 ϭ 0.22, P Ͼ 0.68; Fig. 4b), Both plume moth damage and ¯y damage signi®cantly suggesting that increased seed production is not at the affected seed production (F1,99 ϭ 10.88, P ϭ 0.001, and expense of seed quality. Indian paintbrush parasitizing

TABLE 2. Effect of host species (lupine vs. other species), year, number of in¯orescences, and plant height on pollinator preference in the ®eld.

No. ¯owers probed Time per Visits per visit ¯ower probe

Effect df SS F df SS F df SS F Host 1 0.48 0.19 1 12.19 1.04 1 0.14 0.51 Year 1 34.57 13.26*** 1 413.07 35.25**** 1 2.08 7.74** Host ϫ Year 1 0.11 0.04 1 65.05 5.55* 1 0.40 1.48 Number of in¯. 1 31.38 12.02*** 1 5.88 0.50 1 0.052 0.19 Height 1 2.05 0.79 1 22.33 1.91 1 0.065 0.24 Error 91 237.3 ´´´ 65 ´´´ 761.71 65 17.49 ´´´ Notes: ``Visits'' represents the number of times that pollinators approached and probed plants. Flowers probed per visit and time per ¯ower probe were analyzed only for plants that were visited by pollinators; i.e., zero values were not included. Number of visits was log(x ϩ 1)- transformed prior to analysis. * P Ͻ 0.05; ** P Ͻ 0.01; *** P Ͻ 0.001; **** P Ͻ 0.0001. 2706 LYNN S. ADLER Ecology, Vol. 83, No. 10

FIG. 2. Effect of host plant and year on pollinator pref- FIG. 3. Effect of host plant and site on biomass of Cas- erence in Castilleja miniata, measured as the number of ¯ow- tilleja miniata. Data are averaged over year, except that data ers probed per visit. Error bars represent 1 SE, and numbers Ϯ at site 3 were collected in 1996 only. Error bars represent Ϯ1 in bars represent sample size. SE, and numbers in bars represent sample size. lupines also had marginally higher nitrogen concentra- proximity of Indian paintbrush to alkaloid-containing tion, compared with Indian paintbrush parasitizing oth- lupines is another possible mechanism, this seems un- er host species (F1,24 ϭ 15.57, P ϭ 0.059; Fig. 4c). likely since both insect herbivores are specialists that Considering only nonalkaloid nitrogen concentration spend their larval development on one host plant (Ster- did not change the results; on average alkaloids ac- mitz et al. 1986, personal observation). This result is counted for only 0.03% of the percent dry mass of consistent with experiments in another Castilleja±Lu- nitrogen. There were strong positive correlations be- pinus system, in which experimentally manipulated al- tween nonalkaloid nitrogen and Indian paintbrush bio- kaloid uptake reduced ¯oral herbivory by the herbivore mass (␳ϭ0.6, P Ͻ 0.001, n ϭ 30) and seed production Endotheria hebesana (Lepidoptera: Tortricidae) (Adler (␳ϭ0.43, P ϭ 0.017, n ϭ 30). 2000a). Thus, the bene®ts for Indian paintbrush of par- DISCUSSION asitizing lupine include reduced herbivory and may therefore vary between populations that vary in her- Herbivory bivore pressure. Herbivory by two specialist insect herbivores and Lupine alkaloids are generally detrimental and act one generalist mammalian herbivore was reduced on to deter a wide range of organisms, including vertebrate Indian paintbrush parasitizing lupine hosts, compared and invertebrate herbivores and microbes (reviewed in to Indian paintbrush parasitizing other host species. Wink 1992, 1993). Although Indian paintbrush para- There was a negative correlation between herbivory by sitizing lupines had reduced plume moth herbivory, no both insects and alkaloid concentration of in¯ores- clear relationship emerged between alkaloid content of cences, suggesting that the mechanism of reduced her- in¯orescences and herbivore performance. Several hy- bivory is alkaloid uptake from lupine hosts. While potheses can be proposed to explain this lack of effect:

TABLE 3. Effect of host species (lupine vs. other species), sites, and year on biomass and estimated seed production in Castilleja miniata.

Biomass Seeds

Effect df SS F df SS F Host² 1 3.77 1.51 1 4.35 39.19* Year³ 1 7.78 38.94 1 67.63 4.27 Site 2 7.80 7.75*** 2 15.74 2.38 Host ϫ Site 2 4.99 4.96** 2 0.22 0.03 Host ϫ Year§ 1 0.53 3.55 1 0.43 0.49 Year ϫ Site 1 0.20 0.40 1 15.84 4.78* Host ϫ Year ϫ Site 1 0.15 0.30 1 0.89 0.27 Error 101 50.82 ´´´ 101 334.67 ´´´ Notes: Site was considered a random effect, and year and host plant were ®xed effects. Data were log transformed prior to analysis. * P Ͻ 0.05; ** P Ͻ 0.01; *** P Ͻ 0.001. ² Tested over the host ϫ site interaction term. ³ Tested over the year ϫ site interaction term. § Tested over the host ϫ year ϫ site interaction term. October 2002 HERBIVORY AND POLLINATION IN CASTILLEJA 2707

iparasite Castilleja indivisa indicate that alkaloids play an important role in resistance to herbivory (Adler 2000a), it is not possible to distinguish among these hypotheses without being able to manipulate alkaloids in C. miniata. The effect of host species with alkaloids, compared to hosts without alkaloids, on herbivory in parasitic plants has not been consistent in other studies. Both generalist and specialist herbivores (larvae of Tricho- plusia ni and Euphydryas anicia, respectively) performed more poorly in the laboratory on a diet of Cas- tilleja sulphurea containing alkaloids from the host Delphinium occidentale, compared with leaves from nonalkaloid C. sulphurea (Marko et al. 1995). Con- versely, in a greenhouse study, aphids survived and reproduced best on Castilleja wightii parasitizing the alkaloid-containing host Lupinus arboreus, compared to C. wightii parasitizing two nonleguminous, nonalkaloid host species (Marvier 1996). Increased total nitrogen in parasites on lupines may have been more important to aphids than the presence of alkaloids (Kyto et al. 1996). In a ®eld study, alkaloid uptake in the hemiparasite Pedicularis semibarbata from the host Lupinus fulcratus did not affect oviposition or larval performance by the specialist herbivore Euphydryas editha, compared to Pedicularis parasitizing other host species (Stermitz et al. 1989). These studies indicate that herbivore preference and performance on hemiparasites is not always predicted by the alkaloid content of host species. Pollination Pollinators did not discriminate between C. miniata parasitizing different host species in the ®eld. Thus, while lupines reduce herbivory on Indian paintbrush, there was no evidence for indirect effects of hosts on FIG. 4. Effect of host plant on (a) estimated total seed production, (b) mean seed mass, and (c) percentage total ni- parasites via pollinator preference. In a separate ex- trogen by dry mass in Castilleja miniata. Data for (a) are periment, hand pollination of Indian paintbrush ¯owers from 1996 and 1997 combined, and data for (b) and (c) are did not increase C. miniata seed set at one site in 1998 from 1996 only. Data for each graph are averaged over all (L. Adler, unpublished data), suggesting that seed pro- other main effects. Error bars represent Ϯ1 SE, and numbers duction is limited by resources rather than by pollen in bars represent sample size. ²P Ͻ 0.06; *P Ͻ 0.05; NS ϭ nonsigni®cant. transfer, at least in some years or sites. Therefore, pollinator preference may not affect hemiparasite female ®tness, even if pollinators did discriminate between (1) alkaloid content may be detrimental to larval per- Indian paintbrush with different hosts. formance, but correlated with other positive effects of Alkaloid-containing hosts have the potential to either lupine hosts, such as increased nitrogen, that obscure increase or decrease pollinator attraction to hemipar- this effect; (2) alkaloid uptake may reduce adult ovi- asites via different mechanisms. In some plant species, position preference, but have no effect on herbivore resistance to herbivores may cause decreased pollinator performance (Thompson and Pellmyr 1991); (3) the attraction due to secondary compounds in ¯oral struc- reduced herbivory of Indian paintbrush parasitizing lu- tures (Strauss et al. 1999, Adler 2000b). Neither Cas- pines may be due to some effect other than alkaloid tilleja miniata nor Castilleja indivisa contained alka- uptake; for example, increased nitrogen concentration loids in nectar when parasitizing alkaloid-containing may translate to less food required for herbivore de- hosts (Adler and Wink 2001). It is therefore not sur- velopment; and/or (4) variation in larval response to prising that lupine hosts did not deter pollinators of alkaloids or in the ability to accurately quantify alka- either Castilleja species. However, norditerpenoid al- loids may obscure any effect of alkaloids on larval kaloids from the host Delphinium occidentale have performance. Although studies of the congeneric hem- been found in the nectar of the hemiparasite Castilleja 2708 LYNN S. ADLER Ecology, Vol. 83, No. 10 sulphurea (Marko and Stermitz 1997). It would be in- speci®c variation in parasitic plant host preference has teresting to determine whether host association in¯u- only been examined in one study to my knowledge. ences pollinator preference in this closely related hem- Maternal sibships of the annual hemiparasite Triphy- iparasite. saria pusilla varied in their preference for a lupine host Alkaloid uptake from host plants could also increase (L. nanus) compared to a grass host (Bromus carinatus) pollinator attraction to hemiparasites by reducing her- (Adler 2000c). However, this variation appears to be bivory. Reduced ¯oral herbivory resulted in twice as determined more by maternal host than by genetic var- many pollinator visits to Castilleja indivisa parasitizing iation, suggesting that specialization is unlikely to high-alkaloid compared to low-alkaloid lines of a lu- evolve. In an unpredictable environment dominated by pine host (Adler et al. 2001). However, in the current annuals, maintaining a plastic response to hosts may study of C. miniata, herbivores were removed from be the most appropriate strategy; no studies have ex- Indian paintbrush in the pollination experiment to avoid amined whether perennial parasites have similar strat- confounding herbivore effects with pollinator effects. egies. Thus, while alkaloids in C. miniata were not directly Parasitic plants are found in every major ecosystem attractive to pollinators, there is still the possibility that (Kuijt 1969), and their presence can alter community alkaloid uptake could increase pollinator attraction in- composition and structure (Bennetts et al. 1996, Pen- directly via reduced plume moth herbivory in in¯ores- nings and Callaway 1996, Marvier 1998b). Several cences. studies, including the present one, have found that hemiparasites are most successful on leguminous hosts Plant growth and seed production (Gibson and Watkinson 1989, Seel and Press 1993, Indian paintbrush bene®ted from parasitizing lupine 1994, Matthies 1996, Press and Seel 1996, Marvier hosts, compared to other host species. Indian paint- 1998a). Hemiparasites can also alter the competitive brush parasitizing lupines produced more seeds across outcome between host species by having the most det- two years of sampling and had greater biomass, al- rimental impact on the species that is both the most though the extent of this effect varied between sites. bene®cial host and strongest competitor, typically a le- These bene®ts are most likely to arise from reduced gume (Gibson and Watkinson 1991, Matthies 1996). seed predation, which strongly affected seed produc- By reducing the competitive superiority of legumes that tion and was consistently reduced by parasitizing lu- can alter community dynamics and increase the inva- pines, and possibly by the marginal increase in nitrogen sion of non-native species via nitrogen ®xation (Maron concentration. However, the variation in biomass and Jefferies 1999), parasitic plants may play an im- across sites suggests that the interaction of these effects portant role in maintaining species diversity at the com- may be complex. While it is not possible without ex- munity level. The present study demonstrates that host± tensive manipulations to determine why such differ- parasite plant interactions have important consequenc- ences in plant biomass existed across nearby sites, var- es not just for host plants, but also for parasitic plants iation in abiotic resources is a possible explanation. and animals, such as herbivores interacting with those Indian paintbrush at site 2, a dry and shady site, had parasites. Thus, parasitic plant±host plant interactions the lowest overall biomass; while Indian paintbrush at may alter community composition via several mecha- site 1, which was dry but sunnier, had moderate bio- nisms. mass. The largest plants and the strongest host effects ACKNOWLEDGMENTS were at site 3, which in addition to being sunny was I thank Judah Dinnel, Susie Geer, and Amy Kreutzer for also moist and slightly lower in elevation. Thus, the their hard work in the ®eld, Barbara L. Bentley and Michael bene®ts of parasitizing lupines appear to be strongest Ziebell for invaluable assistance with alkaloid analysis, in- at the site with the most available resources for growth. cluding the use of B. L. Bentley's gas chromatograph (funded by USDA/National Resources Inventory [NRI] Grant Population- and community-level consequences 9401661), Nick Waser, Mary Price, and Bill Calder for hum- mingbird advice, and Kevin Taylor, David Inouye, Mary Price In this study, C. miniata had reduced herbivory and and Rick Williams for host plant identi®cations. The Rocky increased seed set when parasitizing lupine hosts com- Mountain Biological Laboratory provided facilities. Anurag pared to all other host species combined. Since many Agrawal, Keith Clay, Becky Irwin, Rick Karban, Kevin Rice, Jennifer Rudgers, Maureen Stanton, Sharon Strauss, Jennifer parasitic plants have greater ®tness on a subset of po- Thaler, Nora Underwood, Paige Warren, and four anonymous tential hosts, the question arises of whether generalist reviewers provided helpful comments on the manuscript. This parasites, like generalist herbivores, can evolve races research was funded by a Center for Population Biology Re- that specialize on bene®cial hosts. While some studies search Award, Jastro-Shields Research Award, Sigma Xi Grants-in-Aid of Research, National Science Foundation have documented host preference in generalist parasitic Graduate Research Fellowship, and National Science Foun- plants (Werth and Riopel 1979, Gibson and Watkinson dation Dissertation Improvement Grant DEB 98-00885. 1989, Kelly 1990, 1992, Nilsson and Svensson 1997, LITERATURE CITED Yoder 1997, Norton and De Lange 1999), for host pref- Adler, L. S. 2000a. Alkaloid uptake increases ®tness in a erence to evolve in response to selection there must be hemiparasitic plant via reduced herbivory and increased heritable variation in this trait (Falconer 1989). Intra- pollination. American Naturalist 156:92±99. October 2002 HERBIVORY AND POLLINATION IN CASTILLEJA 2709

Adler, L. S. 2000b. The ecological signi®cance of toxic nec- ences between nitrogen-®xing and nitrogen-limited plants. tar. Oikos 91:409±420. Journal of Chemical Ecology 15:2425±2434. Adler, L. S. 2000c. Herbivory and pollination in hemipara- Karban, R., and I. T. Baldwin. 1997. Induced responses to sitic plants: effects of host plant and host-obtained alka- herbivory. University of Chicago Press, Chicago, Illinois, loids. Dissertation. University of California±Davis, Cali- USA. fornia, USA. Kelly, C. K. 1990. Plant foraging: a marginal value model Adler, L. S., R. Karban, and S. Y. Strauss. 2001. Direct and and coiling response in Cuscuta subinclusa. Ecology 71: indirect effects of alkaloids on plant ®tness via herbivory 1916±1925. and pollination. Ecology 82:2032±2044. Kelly, C. K. 1992. Resource choice in Cuscuta europaea. Adler, L. S., and M. Wink. 2001. Transfer of alkaloids from Proceedings of the National Academy of Sciences 89: hosts to hemiparasites in two Castilleja±Lupinus associa- 12194±12197. tions: analysis of ¯oral and vegetative tissues. Biochemical Kuijt, J. 1969. The biology of parasitic ¯owering plants. Systematics and Ecology 29:551±561. University of California Press, Berkeley, California, USA. Arslanian, R. L., G. H. Harris, and F. R. Stermitz. 1990. New Kyto, M., P. Niemela, and S. Larsson. 1996. Insects on trees: quinolizidine alkaloids from Lupinus argenteus and its population and individual response to fertilization. Oikos hosted root parasite Castilleja sulphurea: stereochemistry 75:148±159. and conformation of some naturally occurring cyclic car- Lange, W. H. 1950. Biology and systematics of plume moths binolamides. Journal of Organic Chemistry 55:1204±1210. of the genus Platyptilia in California. Hilgardia 19:561± Baeumel, P., L. Witte, P. Proksch, and F. C. Czygan. 1992. 668. Uptake and metabolism of host plant alkaloids by parasit- Littell, R. C., R. J. Freund, and P. C. Spector. 1991. SAS izing Cuscuta species. Planta Medica 58:A671. system for linear models. Third edition. SAS Institute, Bennetts, R. E., G. C. White, F. G. Hawksworth, and S. E. Cary, North Carolina, USA. Severs. 1996. The in¯uence of dwarf mistletoe on bird Marko, M. D., M. F. Antolin, and F. R. Stermitz. 1995. The communities in Colorado ponderosa pine forests. Ecolog- effect of incorporated delphinium alkaloids by C. sulphurea ical Applications 6:899±909. on a generalist and a specialist insect herbivore. Bulletin Boros, C. A., D. R. Marshall, C. R. Caterino, and F. R. Ster- of the Ecological Society of America 76:168. mitz. 1991. Iridoid and phenylpropanoid glycosides from Marko, M. D., and F. R. Stermitz. 1997. Transfer of alkaloids Orthocarpus spp.: Alkaloid content as a consequence of from Delphinium to Castilleja via root parasitism. Nordi- parasitism on Lupinus. Journal of Natural Products (Lloy- terpenoid alkaloid analysis by electrospray mass spectrom- dia) 54:506±513. etry. Biochemical Systematics and Ecology 25:279±285. Chapin, F. S. I. 1980. The mineral nutrition of wild plants. Maron, J. L., and R. L. Jefferies. 1999. Bush lupine mortality, Annual Review of Ecology and Systematics 11:233±260. altered resource availability, an alternative vegetation Chew, F. S., and J. E. Rodman. 1979. Plant resources for states. Ecology 80:443±454. chemical defense. Pages 271±308 in G. A. Rosenthal and Marvier, M. A. 1995. Host use by parasitic plants: effects on D. H. Janzen, editors. Herbivores: their interactions with insect herbivore performance in a coastal plant community. secondary plant metabolites. Academic Press, New York, Bulletin of the Ecological Society of America 76:361±362. New York, USA. Marvier, M. A. 1996. Parasitic plant±host interactions: plant Davies, D. M., and J. D. Graves. 1998. Interactions between performance and indirect effects on parasite-feeding her- arbuscular mycorrhizal fungi and the hemiparasitic angio- bivores. Ecology 77:1398±1409. sperm Rhinanthus minor during co-infection of a host. New Marvier, M. 1998a. A mixed diet improves performance and Phytologist 139:555±563. herbivore resistance of a parasitic plant. Ecology 79:1272± Falconer, D. S. 1989. Introduction to quantitative genetics. 1280. Third edition. Longman Scienti®c and Technical, New Marvier, M. A. 1998b. Parasite impacts on host communities: York, New York, USA. Fowden, L., and P. J. Lea. 1979. Mechanism of plant avoid- plant parasitism in a California coastal prairie. Ecology 79: ance of autotoxicity by secondary metabolites, especially 2616±2623. by nonprotein amino acids. Pages 135±160 in G. A. Ro- Matthies, D. 1995. Host±parasite relations in the root hem- senthal and D. H. Janzen, editors. Herbivores: their inter- iparasite Melampyrum arvense. Flora 190:383±394. actions with secondary plant metabolites. Academic Press, Matthies, D. 1996. Interactions between the root hemiparasite New York, New York, USA. Melampyrum arvense and mixtures of host plants: hetero- Gibson, C. C., and A. R. Watkinson. 1989. The host range trophic bene®t and parasite-mediated competition. Oikos and selectivity of a parasitic plant: Rhinanthus minor L. 75:118±124. Oecologia 78:401±406. McCoy, J. W., and F. R. Stermitz. 1983. Alkaloids from Cas- Gibson, C. C., and A. R. Watkinson. 1991. Host selectivity tilleja miniata and Penstemon whippleanus, two host spe- and the mediation of competition by the root hemiparasite cies for the plume moth, Amblyptilia (Platyptilia) pica. Rhinanthus minor. Oecologia 86:81±87. Journal of Natural Products 46:902±907. Graves, J. D. 1995. Host-plant responses to parasitism. Pages McKey, D. 1974. Adaptive patterns in alkaloid physiology. 292 in M. C. Press and J. D. Graves, editors. Parasitic American Naturalist 108:305±320. plants. Chapman and Hall, London, UK. Mead, E. W., M. Looker, D. R. Gardner, and F. R. Stermitz. Harborne, J. B. 1984. Phytochemical methods: a guide to 1992. Pyrrolizidine alkaloids of Liatris punctata and its modern techniques of plant analysis. Second edition. Chap- root parasite, Castilleja integra. Phytochemistry 31:3255± man and Hall, London, UK. 3257. Jeschke, W. D., P. Baeumel, N. Rath, F. C. Czygan, and P. Miller, T. E., and J. Travis. 1996. The evolutionary role of Proksch. 1994. Modelling of the ¯ows and partitioning of indirect effects in communities. Ecology 77:1329±1335. carbon and nitrogen in the holoparasite Cuscuta re¯exa Nilsson, C. H., and B. M. Svensson. 1997. Host af®liation Roxb. and its host Lupinus albus L. II. Flows between host in two subarctic hemiparasitic plants: Bartsia alpina and and parasite and within the parasitized host. Journal of Pedicularis lapponica. Ecoscience 4:80±85. Experimental Botany 45:801±812. Norton, D. A., and P. J. De Lange. 1999. Host speci®city in Johnson, N. D., L. P. Rigney, and B. L. Bentley. 1989. Short- parasitic mistletoes (Loranthaceae) in New Zealand. Func- term induction of alkaloid production in lupines: differ- tional Ecology 13:552±559. 2710 LYNN S. ADLER Ecology, Vol. 83, No. 10

Pella, E. 1990a. Elemental organic analysis. Part 1: historical ibarbata (Scrophulariaceae) from Lupinus fulcratus (Le- developments. American Laboratory 22:116. guminosae) fail to in¯uence the specialist herbivore Eu- Pella, E. 1990b. Elemental organic analysis. Part 2: state of phydryas editha (Lepidoptera). Journal of Chemical Ecol- the art. American Laboratory 22:28. ogy 15:2521±2530. Pennings, S. C., and R. M. Callaway. 1996. Impact of a Stermitz, F. R., T. A. Foderaro, and Y. X. Li. 1993. Iridoid parasitic plant on the structure and dynamics of salt marsh glycoside uptake by Castilleja integra via root parasitism vegetation. Ecology 77:1410±1419. on Penstemon teucrioides. Phytochemistry 32:1151±1153. Press, M. C., and J. D. Graves. 1995. Parasitic plants. Chap- Stermitz, F. R., and G. H. Harris. 1987. Transfer of pyrrol- man and Hall, London, UK. izidine and quinolizidine alkaloids to Castilleja (Scrophu- Press, M. C., and W. E. Seel. 1996. Interactions between lariaceae) hemiparasites from composite and legume host hemiparasitic angiosperms and their hosts in the subarctic. plants. Journal of Chemical Ecology 13:1917±1925. Pages 151±158 in P. S. Karlsson and T. V. Callaghan, ed- Stermitz, F. R., G. H. Harris, and W. Jing. 1986. Iridoids and itors. Ecological Bulletins, no. 45. Plant ecology in the alkaloids from Castilleja host plants for Platyptilia pica. subarctic Swedish Lapland. Munksgaard, Copenhagen, Rhexifoline content of P. pica. Biochemical Systematics Denmark. and Ecology 14:499±506. Roby, M. R., and F. R. Stermitz. 1984. Pyrrolizidine and Stermitz, F. R., and M. Pomeroy. 1992. Iridoid glycosides pyridine monoterpene alkaloids from two Castilleja plant from Castilleja purpurea and C. indivisa, and quinolizidine hosts of the plume moth, Platyptilia pica. Journal of Natural alkaloid transfer from Lupinus texensis to C. indivisa via Products 47:846±853. root parasitism. Biochemical Systematics and Ecology 20: Schneider, M. J., and F. R. Stermitz. 1990. Uptake of host 473±475. plant alkaloids by root parasitic Pedicularis spp. Phyto- Strauss, S. Y. 1991. Indirect effects in community ecology: chemistry 29:1811±1814. their de®nition, study and importance. Trends in Ecological Schoener, T. W. 1993. On the relative importance of direct Evolution 6:206±210. versus indirect effects in ecological communities. Pages Strauss, S. Y., D. H. Siemens, M. B. Decher, and T. Mitchell- 365±411 in H. Kawanabe, J. E. Cohen, and K. Iwasaki, Olds. 1999. Ecological costs of plant resistance to herbi- editors. Mutualism and community organization: behav- vores in the currency of pollination. Evolution 53:1105± ioural, theoretical, and food-web approaches. Oxford Uni- 1113. versity Press, Oxford, UK. Thompson, J. N., and O. Pellmyr. 1991. Evolution of ovi- Seel, W. E., and M. C. Press. 1993. In¯uence of the host on position behavior and host preference in Lepidoptera. An- three sub-Arctic annual facultative root hemiparasites: I. nual Review of Entomology 36:65±90. Growth, mineral accumulation and above-ground dry-mat- Werth, C. R., and J. L. Riopel. 1979. A study of the host ter partitioning. New Phytologist 125:131±138. range of Aureolaria pedicularia (L.) Raf. (Scrophulari- Seel, W. E., and M. C. Press. 1994. In¯uence of the host on aceae). American Midland Naturalist 102:300±306. three sub-Arctic annual facultative root hemiparasites: II. Wink, M. 1992. The role of quinolizidine alkaloids in plant± Gas exchange characteristics and resource use-ef®ciency. insect interactions. Pages 133±169 in E. Bernays, editor. New Phytologist 127:37±44. Insect±plant interactions. CRC Press, Boca Raton, Florida, Seel, W. E., and M. C. Press. 1996. Effect of repeated par- USA. asitism by Rhinanthus minor on the growth and photosyn- Wink, M. 1993. Allelochemical properties or the raison d'etre thesis of a perennial grass, Poa alpina. New Phytologist of alkaloids. Pages 1±118 in G. A. Cordell, editor. The 134:495±502. alkaloids. Academic Press, San Diego, California, USA. Silva, A., and C. Martinez Del Rio. 1996. Effects of the Wootton, J. T. 1994. The nature and consequences of indirect mistletoe Tristerix aphyllus (Loranthaceae) on the repro- effects in ecological communities. Annual Review of Ecol- duction of its cactus host Echinopsis chilensis. Oikos 75: ogy and Systematics 25:443±466. 437±442. Yoder, J. I. 1997. A species-speci®c recognition system di- Stermitz, F. R., G. N. Belofsky, D. Ng, and M. C. Singer. rects haustorium development in the parasitic plant Tri- 1989. Quinolizidine alkaloids obtained by Pedicularis sem- physaria (Scrophulariaceae). Planta 202:407±413.