*Dr Peter Stiling, Department of Biology, University of South Florida, Tampa, Florida, 33620-5150, U.S.A. E-mail: email@example.com
Abstract. 1. Associational resistance theory suggests that the association of herbivore-susceptible plant species with herbivore-resistant plant species can reduce herbivore density on the susceptible plant species. Several casual mechanisms are possible but none has so far invoked natural enemies. Associational resistance mediated by natural enemies was tested for by examining densities of a gall fly, Asphondylia borrichiae (Diptera: Cecidomyiidae), and levels of parasitism on two closely related seaside plants, Borrichia frutescens and Iva frutescens, when alone and when co-occurring.
2. Both Borrichia and Iva grow alone or together on small offshore islands in Florida. Each host plant species has its own associated race of fly, but both races of fly are attacked by the same four species of parasitoids. Borrichia normally has a higher density of galls than Iva, and galls are larger on Borrichia than on Iva.
3. Gall size, gall abundance, parasitism levels, and parasitoid community composition were quantified on both Borrichia and Iva on islands where each species grew alone or together. Some islands were then manipulated by adding Borrichia to islands supporting only Iva, and by adding Iva to islands supporting only Borrichia. Subsequent gall densities and gall parasitism levels on the original native species were then examined.
4. On both natural and experimentally manipulated islands, gall densities on Iva were significantly lowered by the presence of Borrichia. This is because bigger parasitoid species that were common on Borrichia galls, which are bigger, spilled over and attacked the smaller Iva galls. Thus, parasitism rates on Iva were higher on islands where Borrichia co-occurred than on islands where Borrichia were absent. Most parasitoids from Iva were too small to successfully attack the large Borrichia galls and so gall density on Borrichia was unaffected by the presence of Iva.
Attack rates of plants by insect herbivores are affected by many factors, including host plant defences (Feeney, 1976; Fritz & Simms, 1992) and host plant nutrition (Mattson, 1980; Kyto et al., 1996); however most plant species exist in communities associated with other plant species. Community composition can also affect levels of herbivory on target species. The association of herbivore-susceptible plant species with herbivore-resistant plant species can reduce the herbivore density on the susceptible plant species. This is known as associational resistance (Tahvanainen & Root, 1972). Several mechanisms have been proposed to explain this phenomenon. Less palatable neighbours can repel herbivores from plant patches. This is known as the repellent plant hypothesis (Tahvanainen & Root, 1972; McNaughton, 1978; Pfister & Hay, 1988). Sometimes, one plant species may emit volatile chemical stimuli that interfere with the host-finding ability of phytophagous insects searching for other species (Stanton, 1983; Hamback et al., 2000). Finally, plants of higher palatability may attract herbivores away from other plant species. This is known as the attractant decoy hypothesis (Atsatt & O'Dowd, 1976). Trap crops are often used in this way to lure pests away from economically important crops. Here, a fourth mechanism is proposed to explain associational resistance, associational resistance mediated by the third trophic level–the natural enemies of the herbivores.
Changes in plant community composition are known to affect the abundance of natural enemies of herbivores. Root's (1973) natural enemies hypothesis proposed that natural enemies are often more abundant in diverse plant communities because most enemies are generalists and can better survive on the greater richness of herbivores there. As a result, generalist natural enemies often suppress herbivore populations more in polycultures than in monocultures (Russell, 1989). Much less, however, has been written about the effects of plant community composition on specialist enemies, particularly parasitoids. This is perhaps surprising given the importance of parasitoids in biological control (Hall et al., 1980; Stiling, 1990). Sheehan (1986) could find only three studies that examined parasitism levels of hosts in simple and diverse plant communities and all showed higher rates of parasitism in plant monocultures, presumably because of the ease of searching in monocultures and lack of disruption of chemical cues. Here, it is suggested that the addition of plant species to a community can actually increase the effectiveness of some specialist natural enemies, here parasitoids, on a given target herbivore as a result of them spilling over from herbivores on one host plant species to herbivores on another plant species. The result is the appearance of associational resistance because the herbivore population decreases on one host plant species when another plant species is added, a phenomenon defined here as associational resistance mediated by natural enemies.
The field sites
A natural intracoastal waterway runs between the mainland and the barrier islands of many U.S. coastal states. To facilitate boat traffic, the U.S. Army Corps of Engineers often deepen these natural channels by dredging. Along the Florida coast in the 1950s and 1960s, this dredge material was typically heaped together as small ‘spoil islands’ that were placed adjacent to the main channel at regular intervals of about 0.5 km. These islands served as the study sites.
While most spoil islands support a variety of vegetation, they are typically overgrown by the invasive exotics Brazilian pepper (Schinus terebinthifolia Raddi) and Australian pine (Casuarina equisetifolia L.) in their centres. Red mangrove (Rhizophora mangle L,), black mangrove (Avicennia germinans (L.) L), and white mangrove (Laguncularia racemosa (L.) Gaertn.f.), along with small patches of salt marsh cordgrass (Spartina alterniflora Loisel), grow near the waterline. Between the waterline and the island centres, two other native plants can be relatively common: sea oxeye daisy (Borrichia frutescens (L.) DC.), and marsh elder (Iva frutescens L.). Islands have one, both, or neither of these host species. Borrichia grows in rhizomatous clonal patches, to between 10 and 100 cm in height, while Iva is a salt marsh shrub which may grow up to 3 m tall.
The study system
The gall midge, Asphondylia borrichiae Rossi and Strong (Diptera: Cecidomyiidae), attacks the terminals of both Borrichia and Iva, with Borrichia suffering higher rates of galling (Rossi & Stiling, 1995). Host-plant choice experiments (Rossi et al., 1999) and recent electrophoretic studies (Dan Howard, New Mexico State University, unpubl. obs.) have shown that each host plant supports its own race of A. borrichiae, but both races share the same specialist parasitoids.
Asphondylia galls on both host plants are attacked by the same four species of hymenopteran parasitoids: Galeopsomyia haemon (Walker) (Eulophidae), Rileya cecidomyiae Ashmead (Eurytomidae), Tenuipetiolus teredon (Walker) (Eurytomidae), and Torymus umbilicatus (Gahan) (Torymidae) (Stiling & Rossi, 1994). Two species, R. cecidomyiae and T. teredon, are endoparasitic and two, G. haemon and T. umbilicatus, are ectoparasitic. Torymus umbilicatus and G. haemon are facultatively hyperparasitic. Most of these parasitoid species are of similar size, with similar length ovipositors, and they attack galls of similar size. The exception is T. umbilicatus, a large torymid with an ovipositor over twice as long as those of the other species (Stiling & Rossi, 1994). Torymus umbilicatus is more common in large galls since it can oviposit last and feed on either fly larvae or other parasitoid larvae (hyperparasitism). Bagging experiments confirmed that T. umbilicatus generally attacks galls after the other three species (Stiling & Rossi, 1994). Both parasitoids and gall flies are multivoltine in Florida with overlapping generations.
To determine the effects of plant community composition on Asphondylia abundance, densities of galls were counted on 14 islands with either one or two host plant species present. Jet Skis™ were used to travel between islands. Three islands supported Borrichia only, six Iva only, and five supported both Borrichia and Iva. All galls, which were easily visible with the naked eye, were counted on sets of 200 stems, or ramets, in three areas of a Borrichia patch or on three sets of 200 twigs or terminals on each of three Iva bushes, and expressed as an average count per 200 terminals. Counts were performed on random sets of 200 stems every other month, five times each during 1993, encompassing multiple generations of insects, to give a density of galls per 1000 stems.
At the same time as the gall densities were counted, the levels of parasitism of gall insects were also assayed. Flies leave characteristic puparia attached to the surface of the gall and create large emergence holes with ragged edges; parasitoids leave no puparia and create smaller exit holes with smooth edges. Rearing out gall insects, dissecting galls, and visual inspections of galls in the field give very similar estimates of parasitism rates (Stiling & Rossi, 1997). On each island, and each date, the galls were examined for exit holes and then scored for parasitism rates. Overall percent parasitism was obtained using data on exit holes from all five monthly samples. For each plant species, total gall densities per 1000 stems and overall percent parasitism data were analysed with t-tests with presence/absence of the other host plant as the treatment and island as the replicate.
While collecting the parasitism data, the size of mature galls (those with exit holes) was also determined, using calipers to measure galls at their widest point, and averaged the data to get a yearly average size of galls on Borrichia (n = 6 islands) and Iva (n = 8 islands). At the same time, galls were collected where they were abundant and patches of hosts were large, and the flies and parasitoids were reared to examine parasitoid community structure for Borrichia galls (n = 6 islands) and Iva galls (n = 6 islands). For Iva, the insects from galls were measured and reared out on three islands where Borrichia and Iva co-occurred and three islands where Iva occurred alone. The percentage of parasitism on Iva due to torymids was compared between islands with and without Borrichia using t-tests.
The year after the analysis of gall densities on naturally occurring plant communities on spoil islands, the host plant community on six islands was manipulated and then the insect community was censused as before. The choice of either adding host plant species to islands and monitoring gall densities and parasitism rates or removing host plant species was considered. It was decided not to remove plant species for two reasons. First, the integrity of the natural islands needed to be maintained and ripping out plant species would have dramatically changed the islands and possibly their abiotic conditions. Second, for many islands, removing many large Iva bushes would have been difficult; certainly all the Borrichia stems could not have been removed easily from an island. Thus, for these manipulations, species were added to the islands. The species were added not by planting them in the ground but by adding them in pots, as a monoculture. This accomplished two goals. First, at the end of the experiment all the pots could be removed, leaving the original plant community on the island unchanged. Second, adding plants in pots minimised the effects of the added plants on the soil properties of the island, and thus reduced the likelihood of effects via alteration of the abiotic conditions. Any changes in resultant gall densities would then likely be caused by the effects of the added plants on densities of gall midges and parasitoids, not from altered soil properties or plant chemistry. Finally, islands were manipulated where existing host patches were of moderate size, about 1500–3000 stems, and the potted plants could be added at equivalent densities. This was a frequently encountered patch size in the study system. Had the native patches been too large, then the effect of the added plants may have been swamped out by native vegetation.
Given the labour-intensive nature of these manipulations and the use of islands with only moderately sized host plant patches, it was not possible to replicate the treatments as much as it would have been liked. On three islands containing Iva only, Borrichia was added, and on three islands containing Borrichia, Iva was added. Borrichia and Iva were added by placing 50 potted plants on each island. Each pot contained 30 Borrichia stems or an Iva bush with an average of 30 stems, for a total of about 1500 stems. These plants were dug up from an unrelated site in February 1993, returned to the botanical garden at the University of South Florida and cultured in pots using a commercial potting soil. In September 1993 these potted plants, which supported no galls, were taken to experimental islands and the pots were placed in holes so that the potting soil was flush with the surrounding substrate. Censuses revealed that by February 1994 they had accrued galls at levels that fell well within the normal ranges of densities on natural islands. This experimental population was then available to influence fly and natural enemy populations on the naturally occurring Iva or Borrichia.
Gall abundances and parasitism levels were assessed on the naturally occurring Iva and Borrichia on manipulated islands every other month, for 10 months during 1994, the year following manipulations. Data were analysed by using t-tests to compare total gall densities and parasitism levels on the experimental islands in the year prior to the manipulation 1993 (control), to total densities in the year following the manipulation 1994 (treatment). To stop the gall densities on the experimental islands from decreasing, galls were not collected, which meant that there was no data on parasitoid communities on experimentally manipulated islands. In this experiment, treatment was confounded by year since gall densities in the pre-treatment year 1993 were being compared to gall densities in the post-treatment year 1994. As a partial control for temporal effects, the densities of galls on unmanipulated islands between 1993 and 1994 (four islands with Iva alone, and four with Iva and Borrichia together) were compared and the data were analysed using paired t-tests.
When data were summarised from all islands, gall densities were higher on Borrichia than on Iva (Fig. 1a; t17 = 2.443, P = 0.026). At the same time, parasitism levels were lower on Borrichia than on Iva (Fig. 1b; t17 = 5.727, P < 0.001). Gall size was significantly different between the two plant species (t12 = 5.627, P < 0.001) with galls on Borrichia (1.085 cm ± 0.081 SE) being almost double the size of those on Iva (0.577 cm ± 0.05 SE). Gall size on Iva was not significantly different between islands with and without Borrichia (t4 = 1.039, P = 0.358). The frequency of Torymus in the gall parasitoid community was 12.3% higher in galls on Borrichia than on Iva (t10 = 2.401, P = 0.037; Fig. 2a). All of these trends are consistent with results of earlier studies on this system (Stiling & Rossi, 1994; Rossi & Stiling, 1995; Rossi et al., 1999).
Gall densities on Iva were significantly reduced on islands where Borrichia was present (t9 = 2.306, P = 0.047; Fig. 1a). Gall densities on Borrichia were unchanged by the presence of Iva. Parasitism of Iva galls increased significantly (t9 = 3.468, P = 0.007) when Borrichia was present (Fig. 1b). At the same time, there was a 14% higher chance of Torymus in Iva galls when Borrichia galls were also present on the island (Fig. 2b), though this difference was not significant (t4 = 1.946, P = 0.124), probably because of small sample size. Parasitism rates on Borrichia galls were not significantly changed by the presence of Iva (t6 = 0.231, P = 0.825).
Gall densities on Iva were significantly reduced when Borrichia was added (t4 = 6.085, P = 0.004; Fig. 3a) and parasitism levels were significantly increased (t4 = 4.522, P = 0.011; Fig. 3b). Gall densities on Borrichia were not significantly changed when Iva was added to an island (t4 = 0.725, P = 0.539; Fig. 3a), nor were fly parasitism levels altered (t4 = 0.169, P = 0.874; Fig. 3b). These results are consistent with the data from natural islands. Gall density on unmanipulated islands did not change from 1993 to 1994 (Borrichia with Iva, t3 = 0.159, P = 0.884; Iva alone t3 = 2.098, P = 0.127; Iva with Borrichia t3 = 1.260, P = 0.297), indicating no significant year effects on gall density on islands.
Using both observations of natural communities and experimental manipulations, strong evidence was documented of an associational resistance mediated by natural enemies between Iva and Borrichia. In this case, the associational interactions are strongly asymmetrical. Asphondylia galls on Borrichia are much larger than those on the Iva. In Borrichia galls, the parasitoid with the longest ovipositor, Torymus umbilicatus, is the most common. When Borrichia is present on an island it provides a source of Torymus parasitoids, which are easily able to parasitise flies in the galls on the Iva species, because these galls are small and therefore vulnerable to attack. As a result, parasitism levels on Iva flies are significantly higher in the presence of Borrichia than they are on pure Iva patches. Because many of the parasitoids that attack the Iva flies are small, they cannot easily parasitise the fly larvae within the bigger Borrichia galls and hence parasitism on Borrichia is unchanged in the presence of Iva. These results are consistent with those reported earlier (Stiling & Rossi, 1994), with one exception. An earlier study reported that overall parasitism levels on Iva remained unchanged in the presence of Borrichia. The difference between the two studies stems from differences in reported gall sizes on Borrichia. In the earlier study, Borrichia galls present on the islands with Iva were small and contained relatively few torymids. The present study encompassed more islands, with bigger Borrichia galls, and these clearly influenced parasitism levels on co-occurring Iva. It is sobering to realise that increased replication over a broader range of conditions can change results.
This study is believed to be the first to document a case of association resistance mediated by parasitoids; however many cases of apparent competition between insects are caused by shared natural enemies (Holt & Lawton, 1994; Karban et al., 1994; Berdegue et al., 1996; Bonsall & Hassell, 1997; Redman & Scriber, 2000), illustrating how frequently associational resistance via shared natural enemies might be. Letourneau (1987) showed that the presence of maize in a tri-culture of maize–legume–squash caused a greater abundance of parasitoids and produced a higher rate of parasitism of a squash herbivore, despite lower initial levels of the herbivore in the tri-culture. Future work on determining the relative strength and frequency of associational resistance mediated by natural enemies could include the relative patch sizes of host plants (Root, 1973) (in these experiments they were about the same), relative distances between host patches (Karieva, 1985), host choice of natural enemies (Brown et al., 1995; Cronin & Abrahamson, 2001), and relative dispersal distances of both herbivores and their natural enemies (Coll & Bottrell, 1994).
Financial support was provided by National Science Foundation grants DEB 9309298 and DEB 00-89226. This study could not have been completed without field help from Alex Collazos, Christine Hilleary, Kate Frey, Ray Kraker, Leetha Menon, Beth Moses, and Terri Woods. Thanks to Alan Spencer at the University of South Florida Botanical Garden for helping to take care of the plants. The comments of David Andow, Earl McCoy, and Dick Root helped to improve earlier versions of the manuscript.