Trophic Interactions Between Two Herbivorous Insects, Galerucella calmariensis and Myzus lythri, Feeding on Purple Loosestrife, Lythrum salicaria, and Two Insect Predators, Harmonia axyridis and Chrysoperla carnea

The effects of two herbivorous insects, Galerucella calmariensis Duftschmid and Myzus lythri L. (Coleoptera: Chrysomelidae), feeding on purple loosestrife, Lythrum salicaria L. (Myrtiflorae: Lythraceae), were measured in the presence of two insect predators, Harmonia axyridis Pallas (Coleoptera: Coccinellidae) and Chrysoperla carnea (Stephens) (Neuroptera: Chrysopidae). A greenhouse cage experiment examined the direct effects of these predators on these herbivores, and indirect effects of predation on aboveground biomass, defoliation, number of leaves, and internode length. Eight treatment combinations with G. calmariensis, M. lythri, H. axyridis and C. carnea were applied to caged L. salicaria. The experiment ended when G. calmariensis adults were observed, 11 to 13 days after release of first instar G. calmariensis. G. calmariensis larvae alone removed significant amounts of leaf tissue and reduced the number of L. salicaria leaves. Predators did not reduce levels of defoliation by G. calmariensis. C. carnea had no effect on G. calmariensis survival, but H. axyridis reduced G. calmariensis survival in the presence of M. lythri. Both predators reduced the survival of M. lythri. This short duration greenhouse study did not demonstrate that predator-prey interactions altered herbivore effects on L. salicaria.


Introduction
Purple loostrife, Lythrum salicaria L. (Myrtiflorae: Lythraceae) is an invasive weed with high fecundity that displaces native vegetation (Thompson et al. 1987;Balough and Bookhout 1989;Blossey et al. 2001). In the early 1990s, two species of herbivorous insects, Galerucella calmariensis L. and Galerucella pusilla Duftschmid (Coleoptera: Chrysomelidae), were introduced in North America to reduce L. salicaria density (Hight et al. 1995). Since 1994, more than 1.4 million individual Galerucella spp. have been released in Iowa wetlands (J.J. Obrycki, unpublished data). In a field cage study, Cortilet (1998) demonstrated that the percentage defoliation and terminal bud damage of L. salicaria increased with increasing G. calmariensis larval density. After 47 days, 50 G. calmariensis larvae caused 25% defoliation and more than 20 terminal buds per stem were damaged (Cortilet 1998). In a second cage study, ten L. salicaria plants were enclosed with 45 G. pusilla eggs in individual cages for 35 days resulting in 14% defoliation (Wiebe 2001). Additionally, Katovich et al. (1999) released 50 Galerucella spp. adults and larvae on caged L. salicaria plants for two months, resulting in an average of 86% defoliation.
Myzus lythri (Schrank) (Homoptera: Aphididae), first recorded in the U.S. in the 1930s (Gillette and Palmer 1934), was observed feeding on L. salicaria in Indiana in 1992 (Voegtlin 1995). In a greenhouse study, significantly lower dry weight of roots and shoots were observed for plants infested with M. lythri, compared to plants without aphids (Voegtlin 1995).
Previous studies have shown negative effects on L. salicaria when either M. lythri (Voegtlin 1995) or G. calmariensis were the only herbivorous species (Cortilet 1998;Wiebe 2001;Katovich et al. 1999;Landis et al. 2003;Denoth and Myers 2005). However, to our knowledge no studies have examined interactions of G. calmariensis and M. lythri on L. salicaria. A previous discussion of the potential for biological control of L. salicaria predicted that combinations of insect herbivores would have a greater negative effect on L. salicaria than single herbivorous species (Malecki et al. 1993). Interspecific interactions of these herbivores on L. salicaria allow one to test whether an additive effect reduces L. salicaria biomass, or if interspecific competition between the herbivores inhibits reduction of L. salicaria.
In biological control, interactions are considered when multiple natural enemy species are introduced into a new habitat. Multiple agents are believed to increase cumulative stress on weeds (Myers 1985). Competing insects, located in the same areas of the plant, increase destruction of the plant, thus reducing plant growth (Harris 1981 (Matos and Obrycki 2006). M. lythri was highly suitable prey for H. axyridis and C. carnea (survival of 80% and 61%, respectively), whereas G. calmariensis was highly suitable prey for C. carnea (survival of 76%), but less suitable for H. axyridis (survival of 27%) (Matos and Obrycki 2006).
Predator-prey interactions have the potential to be detrimental to suppression of L. salicaria through trophic cascades. A trophic cascade occurs when top predators have an indirect influence on the abundance of plant species via their effect on the number of herbivores present (Schmitz et al. 2000). A field study in Sweden, demonstrated that defoliation of L. salicaria by G. calmariensis is higher when fewer insect predators are present . These results indicated that G. calmariensis abundance is affected by predation by lady beetles . In North American wetlands infested with L. salicaria, predation has been reported on Galerucella spp. eggs and larvae (Sebolt and Landis 2004;Wiebe and Obrycki 2004).
The objectives of this study were twofold; to determine if two herbivorous species (G. calmariensis and M. lythri) have an additive negative effect on selected plant measurements, and, second, whether the presence of insect predators that consume the herbivores reduce the effects of herbivory on L. salicaria, causing an indirect effect on the plant.

Experimental design
In early June 2004, L. salicaria seeds were planted in SunGro Sunshine LC1 Mix® and maintained in a greenhouse at 25 ± 5ºC 16:8 h (L:D). When L. salicaria plants reached a height of 60 ± 3 cm, each of 42 seedlings were transplanted into single 19-L pots and enclosed in mesh sleeves (120 cm tall x 70 cm wide) (No-See-Um netting, Balsom Hercules Group, Providence, RI) supported by tomato cages.
The experimental design was an incomplete block design with eight treatments and five replicates ( Table 1). The marsh habitat of purple loosestrife was simulated using potted plants in each of 10 wading pools (100 cm diameter, 30 cm height, containing 10 cm of water). Seven of the eight treatments were randomly selected and assigned to an individual potted plant in a wading pool, each of which served as a replication. The remaining treatments, which was the eighth pot (totaling five potted plants from the five blocks), were placed in a sixth pool, thereby creating an incomplete block. In the sixth pool, two additional potted plants were placed to create a pool environment similar to the other five pools with seven potted plants.
These treatments were selected because in a previous laboratory study, G. calmariensis and M. lythri were found to be suitable prey for the development and survival of H. axyridis and C. carnea (Matos and Obrycki 2006). In addition, due to logistical accommodation of the experimental units, we selected eight treatments out of all possible combinations. This selection was based on two preliminary studies and several prior experimental studies (Voegtlin 1995;Cortilet 1998;Wiebe 2001;Finke and Denno 2004). One plant in each block was randomly assigned to one of eight treatments (Table 1). Predator and herbivore densities were similar to densities observed in the field (Matos and Obrycki 2007 in press). The experiment concluded when G. calmariensis adults eclosed, because plant measurements were taken in response to G. calmariensis larval feeding.

Parameters measured
The number of herbivores and predators in each cage were counted at the end of the experiment. M. lythri infested plants for 25 days, G. calmariensis infested plants for 11-13 days, and predators were caged on plants for 9-11 days. Myzus lythri were released on the plants 14-16 days before predators were released and G. calmariensis first instars were released 2 days before the predators were released.
Five measurements of L. salicaria were taken: number of leaves, above ground biomass, internode length, leaf tissue removed, and total leaf area. Above ground dry biomass was determined by harvesting all live aboveground vegetation, which was dried in an oven for 7 days at 55ºC and then weighed. Skeletonizing of the leaves occurs as a result of G. calmariensis larvae feeding. Total leaf area and percentage defoliation (leaf tissue removed) were measured at the end of  (2002) who determined that a flatbed scanner is an accurate and precise tool for leaf area measurement. Two copies of the original images were made. To estimate leaf tissue removal, the area removed from a leaf in the copied image was filled with black pixels. The area was measured using the histogram option (selected from the Image menu in Adobe Photoshop®).

Data analysis
The herbivore-plant and predator-prey effects on number of leaves, internode lengths, leaf area measured, leaf tissue removal, and aboveground dry biomass of L. salicaria were analyzed independently with mixed-model analyses of variance in which a block (replication) was modeled as a random source of variation (SAS 2003

Herbivore -plant interactions
The greatest leaf tissue removal occurred in the G. calmariensis alone treatment (927.4 ± 305.3 cm 2 ; mean ± SE) ( Table 2). G. calmariensis removed a significantly larger amount of leaf tissue when compared to the control (P <0.0001; Table 3). Over 50% defoliation was measured in L. salicaria plants with G. calmariensis only. G. calmariensis also significantly reduced the number of leaves compared to the control  treatments (P < 0.0001; Table 3). M. lythri alone did not significantly affect internode length (P = 0.27; Table 3). Plants infested with M. lythri had more leaves at the end of the experiment compared to the M. lythri + G. calmariensis treatment (P = 0.0014; Table 3).

Interactions of two herbivore species
Over 29% of G. calmariensis and a mean of 93.9 M. lythri per 12 leaves survived in the M. lythri plus G. calmariensis treatment (Table 4). Total numbers of G. calmariensis and M. lythri surviving were similar to treatments in which each individual herbivore was alone. The results indicated that neither herbivore species influenced the survival of the other.

Herbivore -predator interactions
The herbivores plus predator treatments reduced leaf area compared to the control (P = 0.02; Table  3). The presence of C. carnea did not influence the number of leaves on plants exposed to G. calmariensis (P = 0.05). The presence of predators affected survival of herbivores but had no effect on plant measurements. Over 20% of G. calmariensis survived in the G. calmariensis plus C. carnea treatment (Table 4), which is similar to the G. calmariensis alone treatment. When both herbivores were in the same cage with C. carnea, G. calmariensis survival was unaffected (Table 4). G. calmariensis survival in the M. lythri with both G. calmariensis and H. axyridis treatment was significantly less (12.4%) than in the M. lythri plus G. calmariensis treatment (29.6%) ( Table 4). The lowest survival of M. lythri (2.2 per 12 leaves per plant) was observed in the M. lythri plus both G. calmariensis and H. axyridis treatment.

Plant-predator interactions
At least one of the three predators released in each cage survived, indicating that predation occurred within the cages. At the end of the experiment, the predators were either in their pupal stage or last instar. The presence of predators did not affect any plant measurement (Table 3).

Discussion
These findings have several implications for the understanding of multitrophic interactions associated with L. salicaria. First, the two herbivores (G. calmariensis and M. lythri) did not have an additive negative effect on L. salicaria. Second, two predatory species (C. carnea and H. axyridis) decreased M. lythri survival, and H. axyridis reduced G. calmariensis survival when M. lythri was present. Finally, predator presence did not indirectly benefit the plant, based upon the parameters measured.
Several studies have attempted to determine if multiple species of herbivores increase the success of weed biological control projects (Denoth et al. 2002). In the biological control program for the control of the invasive weed species, Lantana camara L., temporally separated natural enemies including (Teleonemia scrupulosa Stal (Hemiptera: Tingidae) during summer months and three species of Lepidoptera during winter months, contributed to the suppression of L. camara (Andres and Goeden, 1971). In some cases multiple herbivore species do not increase levels of biological control (Myers 1985;Hunt-Joshi and Blossey 2005a). In a 4-year field cage study, there was no increase in damage to L. salicaria when two spatially separated herbivores, G. calmariensis and a root feeding weevil, Hylobius transversovittatus Goeze (Coleoptera: Curculionidae), were present (Hunt-Joshi and Blossey 2004). By itself, G. calmariensis reduced L. salicaria height, reproductive ability, and aboveground biomass whereas H.
transversovittatus increased mortality of actively growing stems and thinned L. salicaria stands (Hunt-Joshi and Blossey 2004). However, in combined herbivore treatments, no increased suppression of L. salicaria was observed (Hunt-Joshi and Blossey 2004). In our study, an additive effect due to the herbivores G. calmariensis and M. lythri on L. salicaria was not observed because M. lythri alone did not reduce any of the plant characteristics assessed.
In the absence of predators, survival of one herbivore was not affected by the presence of the other herbivorous species. Previous studies examined coexistence of G. calmariensis and G. pusilla on L. salicaria (Blossey 1995) and interactions of the spatially separated root-feeding weevil, H. transversovittatus, and the herbivore, G. calmariensis on L. salicaria plants (Hunt-Joshi and Blossey 2005a). Although G. pusilla and G. calmariensis adults aggregate at the same sites and use the same host plant, these herbivorous species have similar competitive abilities and coexist (Blossey 1995 Although specialist predators and parasites are eliminated in the quarantine process before release (Harley and Forno 1992), naturally occurring generalist enemies in the release habitats have the potential to reduce establishment and success of the introduced herbivores (Goeden and Louda 1976). In our study species of generalist predators were chosen that occur in L. salicaria-infested wetlands (Sebolt and Landis 2004;Wiebe and Obrycki 2004). The results showed that survival of G. calmariensis in the treatment cages was not reduced by the presence of predators, except when H. axyridis and M. lythri were present. Previously, we showed that G. calmariensis was suitable for preimaginal development and survival of C. carnea and H. axyridis (Matos and Obrycki 2006). However, in the present study, C. carnea did not reduce G. calmariensis numbers compared to when G. calmariensis was alone. Neonate G. calmariensis feed within L. salicaria shoot tips presumably avoiding predation (Sebolt and Landis 2002). In the presence of the predator Coleomegilla maculata (De Geer) (Coleoptera: Coccinellidae), G. calmariensis neonate survival was higher in L. salicaria shoot tips (70%) than those neonates exposed on L. salicaria leaves (7.1%) (Sebolt and Landis 2002). A shift in behavior of herbivores or predators may alter the occurrence of a trophic cascade (Schmitz et al. 1997).
The suitability of prey for H. axyridis was higher for M. lythri compared to G. calmariensis larvae (Matos and Obrycki 2006). In the current study, it seems likely that M. lythri supplemented the low nutritional quality of G. calmariensis for H. axyridis development and survival. Similarly to G. calmariensis, M. lythri is found on leaves and stems and this may have created an opportunity for H. axyridis to attack G. calmariensis. Sebolt and Landis (2004) showed that the attack rates of H. axyridis on G. calmariensis first to third instars ranged from 60 to 100%. Our study showed higher G. calmariensis mortality when M. lythri was present and preyed upon by H. axyridis compared to the mortality when H. axyridis was absent. Because we did not include a treatment with only G. calmariensis plus H. axyridis, this treatment design did not determine whether the presence of M. lythri influenced predation of G. calmariensis.
M. lythri survival was significantly reduced in the presence of the predators. M. lythri did not reduce internode length; however, internode length was shorter in the M. lythri plus H. axyridis treatment than in the M. lythri alone treatment. Possibly M. lythri altered its behavior in the presence of H. axyridis and this behavior resulted in M. lythri moving to more protected sites where its feeding was detrimental to internode length.
C. carnea and H. axyridis did not indirectly benefit the growth of L. salicaria. Although H. axyridis caused G. calmariensis and M. lythri mortality and C. carnea caused M. lythri mortality, herbivory was not significantly reduced in this greenhouse cage study. During a 4-year field study, an opportunistic predator Plagiognathus politis (Hemiptera: Miridae) consumed large numbers of G. calmariensis eggs and young larvae (Hunt-Joshi et al. 2005b). However, after the second and third season, G. calmariensis populations increased to levels that caused significant defoliation (Hunt-Joshi et al. 2005b). Hunt-Joshi et al. (2005b) suggested that a more controlled experiment where predators and herbivore population levels were manipulated could result in a strong trophic cascade effect. Our study was a controlled short-term greenhouse experiment where herbivore and predator levels were manipulated, but no indirect positive effect of the two species of generalist predators on L. salicaria was found.