Acquisition of enemy-free-space has been suggested to reduce selective pressure against host range expansion in phytophagous insects. The gall midge, Asphondylia borrichiae Rossi and Strong (Diptera: Cecidomyiidae), which attacks the stem tips of its 3 host plants produces a spherical tumor-like growth (= gall). Juvenile stages (larvae and pupae) of A. borrichiae develop inside the gall; the midge spends approximately 95–98% of its life cycle embedded within the gall. During these juvenile stages, A. borrichiae are parasitized by 4 species of hymenopterans. Previous studies have found that one of the most common and the largest parasitoid, Torymus umbilicatus (Gahan), tends to dominate large galls owing to its significantly longer ovipositor, which enables it to penetrate the biggest galls and reach larvae and pupae that become unavailable to the other 3 parasitoids, which have much shorter ovipositors. Moreover, previous studies suggest that the gall midge is diverging both morphologically and genetically in sympatry. The current study is the first to provide morphological evidence that T. umbilicatus, which is a dominant member of the parasitoid guild that attacks A. borrichiae, may also be diverging in sympatry along with its host. Female T. umbilicatus from sea oxeye daisy (Borrichia frutescens [L.] DC) were significantly larger than those from alternative host plants of the gall midge, dune elder (Iva imbricata Walter) and marsh elder (I. frutescens L). Additionally, size of female T. umbilicatus collected from 2 geographically distant sites were significantly different and these differences were consistent with a latitudinal gradient in size between plant species. Although T. umbilicatus were larger from galls collected from B. frutescens compared to I. frutescens at both sites, gall diameter demonstrated a significant decline along a south-north latitudinal gradient. However, a significant interaction between plant species and site suggests that differences in T. umbilicatus size (and most likely their gall midge host) is caused either by phenotypic plasticity of the species at the 2 sites, or these insects (T umbilicatus and gall midges) tend to be smaller with increasing latitude. Moreover, galls on I, frutescens, owing to their smaller size and increased crowding, decline in size at a greater rate than those from B. frutescens which produces significantly larger and less crowded galls.
The gall midge, Asphondylia borrichiae (Diptera: Cecidomyiidae) Rossi and Strong, is a small ambrosia galler; females deposit eggs along with fungal conidia near the apical meristem, which results in a spherical fungal-lined gall, on its 3 host plant species (Rossi & Strong 1990; Stiling & Rossi 1997; Rossi & Stiling 1998; Rossi et al. 1999). The 3 host plants of the gall midge include sea oxeye daisy (Borrichia frutescens [L.] DC), marsh elder (Iva frutescens L.) and dune elder (I. imbricata Walter). All 3 plant species are closely related members of the aster family, occur in similar saline habitats (salt marshes, beaches, etc.) and they have highly sympatric distributions along the eastern and Gulf coasts of North America (USDAPlant Database Service at plants.usda. gov).
Asphondylia borrichiae galls are usually 4-chambered and each chamber contains a single larva; after eclosion larval development and pupation of the midge (and its parasitoids) occurs within the gall and adults emerge (depending upon the season) 4–8 weeks later (Rossi & Stiling 1995). Lifespan of A. borrichiae typically ranges from 6–12 weeks and immature midges overwinter in the galls (Rossi & Stiling 1995). Adult Asphondylia often live only 24–48 hours, which accounts for only 2–5% of the midge's lifecycle; most of their life (95–98%) is spent as juvenile stages embedded within the host plant's tissues (Gagné 1989; Rossi & Stiling 1995). Gall midges that develop on B. frutescens are significantly larger than those from either I. frutescens or I. imbricata because galls on Borrichia are larger and much less crowded than those from the other 2 plants (Rossi & Stiling 1995; 1998; Rossi et al. 1999; Stiling & Rossi 1994; 1998). Midges experience less competition in galls from B. frutescens compared to those on the 2 Iva species. Specifically, A. borrichiae that develop in sea B. frutescens are significantly larger than those from either I, imbricata or I. frutescens; the larger body size of midges from B. frutescens results in a 30–40% higher potential fitness (measured as the number of eggs at emergence) on this plant compared to the 2 Iva species (Rossi et al. 1999). Size- and/or host-assortative mating, coupled with different development rates of A. borrichiae on its host plants, may reduce gene flow between host-associated populations of the gall and promote divergence of the populations.
Four parasitoid wasps Torymus umbilicatus (Gahan), Galeopsomyia haemon (Walker), Rileya cecidomyiae Ashmead, and Tenuipetiolus teredon (Walker) parasitize juvenile stages of Asphondylia while they develop in the gall (Stiling et al. 1992; Stiling and Rossi 1994). Additionally, the largest and smallest species (T. umbilicatus and G. haemon respectively) are also hyperparasites that will attack the other members of the parasitoid guild as well as developing A. borrichiae (Stiling et al. 1992). Plant species and quality alter the relative abundance and attack rates of 4 species of parasitic wasps that parasitize the juvenile stages of the midge. For instance, the parasitoid guild that attacks the largest galls from B. frutescens and/or fertilized plants (which also produce larger galls) is dominated by the relatively large torymid wasp T. umbilicatus, while the parasitoid community of smaller galls and those on the 2 Iva species tend to be dominated by the smallest species of parasitoid (Galeopsomyia haemon). These 2 hyperparasites tend to dominate large and small galls respectively for different reasons. Specifically, T. umbilicatus owing to its longer ovipositor (which is more than twice as long as those from the next largest species) has the ability to oviposit last in a gall and attack juvenile stages of the midge or the other parasitoids when the galls are more mature and have a greater diameter. Conversely, the smallest species G. haemon is the only gregarious parasitoid in the gall community and its multiple larvae allow it to attack and overwhelm other members of the gall community including other members of the parasitoid guild and it often dominates the gall community in small galls (Stiling & Rossi 1994). Parasitism is a primary mortality factor for A. borrichiae and aggregate parasitism levels can reach 100% at some sites causing local extinction of the midges and it has been hypothesized that decreased size and fecundity of the midge on the 2 Iva spp. is partially balanced by the acquisition of enemyfree-space on these derived species (Stiling et al. 1992, 1994; Rossi & Stiling 1995; Moon & Stiling 2002; Rossi et al. 2006).
Although A. borrichiae is considered a single species field, laboratory and a recent genetic study provide evidence that A. borrichiae actually consists of distinct host-associated populations at the level of the plant genus (Rossi & Stiling 1998; Stiling & Rossi 1998; Rossi et al. 1999; Stokes et al. 2012). In particular, midges that develop in Borrichia are statistically larger and genetically distinct from the populations associated with the 2 species of Iva and much evidence suggests that the original host plant of the gall midge was sea oxeye daisy (B. frutescens) (Rossi & Stiling 1998; Stiling & Rossi 1994; 1998; Rossi et al. 1999). It has been hypothesized that the gall midge, probably through ovipositional mistakes, established populations on the 2 derived host plants (I. frutescens and I. imbricata) and midges from these 2 plants are significantly smaller (owing to increased larval crowding in their smaller galls) compared to midges that develop on B. frutescens. Asphondylia borrichiae exhibit consistent size and genetic differences that Stireman et al. (2006) refer to as host-associated differentiation (HAD) at the level of the plant genus. The apparent trade-off for decreased size and fecundity of A. borrichiae populations that utilize the 2 Iva spp. appears to be acquisition of enemy-free-space; midges on the derived host plants have lower parasitism rates especially during initial host plant range expansion (Stiling & Rossi 1994; Rossi & Stiling 1995; Rossi et al. 1999).
Plant-host-parasitoid systems are often tightly linked and changes in morphology, life history, genetics, etc. of the host(s) may produce a trophic cascade through the system. This study investigated whether the largest parasitoid, Torymus umbilicatus, exhibits HAD in size that is consistent with its gall midge host. Significant and consistent size differences suggest either phenotypic plasticity between populations of the parasitoid or a multi-trophic level cascading HAD in which both the host (A. borrichiae) and its primary parasitoid are diverging in sympatry at the level of the plant genus.
MATERIALS AND METHODS
Collection and Preservation of Parasitoids
To examine potential host-associated size divergence in T. umbilicatus, galls from each of the 3 host plants within their respective overlapping ranges in 2 distant geographic and noncontiguous locations in Florida at a distance of approximately 350 km were compared. Populations used in this study were from Little Talbot Island (Duval County) located north of Jacksonville and Honeymoon Island (Pinellas County) near Tampa. Approximately 100 galls were collected from each location and species except for I. imbricata from Talbot Island (only 4 galls were found on this plant at this site and none of them produced any insects); therefore, this species-site combination was excluded from the study. Late-stage galls were clipped from the host plants, returned to the lab and placed in large 30-mL plastic vials where they completed development. After emergence from the galls adult T. umbilicatus were collected from the vials and preserved in 70% ethanol. Vials were monitored for emerging insects for at least 2–4 weeks to allow ample emergence time for all juveniles developing within the galls.
Morphometric Analysis
After emergence, female T. umbilicatus were collected from the vials and placed under 5X magnification on a dissecting microscope (Leica, Model MZ95, Micro Optics of Florida, Inc., Plantation, Florida 33313) fitted with an ocular micrometer to measure wing and ovipositor length. Wing length (which is typically used as a measure of size and is highly positively correlated with fecundity in many insects) and ovipositor length were measured to the nearest 0.1 ocular micrometer units (omu) (1.0 omu = 0.025 mm). Wing length was measured from the wing base (tegula) to the wing tip. Ovipositor length was measured from apex to base and, because the ovipositor is partially concealed within the abdomen, the ovipositor was pulled completely free from the abdomen with forceps prior to measuring. Pearson's correlation was used to examine the general relationship between wing and ovipositor length; because no insects were obtained from I. imbricata at the northern site (Little Talbot Island) and because a site or latitudinal gradient may affect insect size, only T. umbilicatus collected from the 3 host plants from Honeymoon Island were used in the correlation. For this same reason, a oneway ANOVA to compare ovipositor length for torymids collected from the 3 host plants used only insects from the south Florida site (Honeymoon Island) since no insects successfully developed from I. imbricata at the northern site. Means were compared using Tukey's HSD post-hoc test.
The relationship between plant species and site on the ovipositor length of T. umbilicatus were assessed using a two-way ANOVA; as mentioned above, only B. frutescens and I. frutescens could be used in this analysis since no T. umbilicatus were collected from I. imbricata at the northern site. Data were log-transformed prior to analysis to meet homogeneity of variance assumption of ANOVA (but are presented non-transformed for clarity) (SPSS, Version 10, SPSS, Inc. Chicago, Illinois 60606).
RESULTS
Wing length and ovipositor length exhibited a significant positive correlation (r = 0.576; P < 0.001; n = 61) (Fig. 1). Variation in wing length explained approximately 33% of variation in ovipositor length. Torymus umbilicatus tended to be larger and have longer ovipositors if they developed in galls on B. frutescens compared to the 2 Iva spp., although small sample size (especially from I.imbricata; n = 8) reduced the power of the statistical tests and only differences in ovipositor length approached significance (F2,72 = 1.54; P = 0.221 and F272 = 2.416 ; P = 0.097 for wing length and ovipositor length respectively, Fig. 2). [Note: if the 2 Iva spp. are grouped together and compared to those from B. frutescens the difference in ovipositor length is significant; t = 2.137; P = 0.036)].
The two-way AN OVA revealed a significant effect of host plant (F1,114 = 23.142; P < 0.001), site (F1,114 = 25.526; P < 0.001) as well as a significant plant × site interaction on ovipositor length (F1,1 = 5.104; P = 0.026) (Fig. 3). Thus, while mean ovipositor length of T. umbilicatus is longer for parasitoids from B. frutescens compared to parasitoids from I. frutescens within a site, ovipositor lengths for both species exhibited significant differences that are consistent with a latitudinal gradient. Ovipositor lengths were much shorter at the northern site compared to the south, but the decrease in size affected parasitoids from I. frutescens (30% decrease) more than those from B. frutescens (19% decrease) resulting in a significant site × host plant interaction.
Fig. 1.
Correlation between wing length (mm) and ovipositor length (mm) of Torymus umbilicatus from populations of its 3 host plants (B. frutescens, I. frutescens and I. imbricata); r = 0.576; P < 0.001).
![f01_832.jpg](ContentImages/Journals/flen/96/3/024.096.0317/graphic/WebImages/f01_832.jpg)
Fig. 2.
Comparison of ovipositor length (mm) of Torymus umbilicatus from A. borrichiae galls from B. frutescens, I. frutescens and I. imbricata (Note: P = 0.097 for one-way ANOVA, but P = 0.036 for t-test in which the 2 Iva spp. were combined and compared to those from B. frutescens; see text for details). Values are mean + SE.
![f02_832.jpg](ContentImages/Journals/flen/96/3/024.096.0317/graphic/WebImages/f02_832.jpg)
DISCUSSION
The current study found consistent size variation between plant-associated populations of the torymid parasitoid, T. umbilicatus, which attacks the juvenile stages of the gall-inducing midge Asphondylia borrichiae. Torymus umbilicatus were consistently larger and had longer ovipositors from galls produced on B. frutescens compared to either species of Iva. These plant-associated size patterns between populations of the parasitoid are consistent with trends exhibited by the gall midge and, while some of these differences may be due to nutritional differences between the plants, much of the variability can be explained by gall size and larval crowding. Previous studies have shown that A. borrichiae from B. frutescens were significantly larger than those produced on the 2 Iva spp.; decreased size of the gall midge is a function of larval crowding and resource availability among the plant species (Clouse 1995; Rossi & Stiling 1995; Rossi et al. 1999). Although the average number of chambers (and hence gall midge larvae) is the same for all 3 plant species (4–5 chambers per gall), differences in gall size between the host plants results in significantly more crowded galls on the 2 Iva spp. compared to B. frutescens (Rossi & Stiling 1995; Rossi et al. 1999) Specifically, the smaller galls from I. imbricata and I. frutescens were 65% and 265% more crowded respectively compared to those from B. frutescens and increased larval crowding exhibited a highly negative correlation with gall midge size and potential fecundity (Rossi et al. 1999). However, the host × site study, which dropped I. imbricata from the analysis (since no T. umbilicatus from I. imbricata from the north site were collected), found that the effect of plant species was almost as strong as the effect of site.
Fig. 3.
Comparison of ovipositor length (mm) of Torymus umbilicatus from A. borrichiae galls collected from B. frutescens and I. frutescens from south Florida (Honeymoon Island) and north Florida (Little Talbot Island). Data were log-transformed prior to analysis, but are presented non-transformed; values are mean + SE.
![f03_832.jpg](ContentImages/Journals/flen/96/3/024.096.0317/graphic/WebImages/f03_832.jpg)
The current study also suggests that either phenotypic plasticity of insect populations account for differences in their size between sites or that they are influenced by a latitudinal gradient; populations of T. umbilicatus from northern populations were significantly smaller compared to southern ones. However, the latitudinal effect was greater for T. umbilicatus from I. frutescens than those from B. frutescens which resulted in a significant plant × site interaction. Decrease in gall size at the north Florida site (Little Talbot Island) compared to the south Florida site (Honeymoon Island) is consistent with cooler temperatures at the former; the 2 sites are approximately 350 km apart and fall in different temperature zones (USDA 2012). The greater decline in size of T. umbilicatus that develop in Asphondylia galls on Iva spp. suggests that galls from these host plants become more crowded (and at a faster rate) compared to the much larger galls produced on B. frutescens. The 3 plant species have similar ecological niches and occur largely in sympatry and it has been hypothesized that the ancestral host plant for the midge was B. frutescens (Rossi et al. 1999; although Clouse 1995 has suggested that I. frutescens may have been the original host). Populations of the midge most likely became established on the 2 Iva spp. through ovipositional mistakes. While female midges typically oviposit on the plant from which they emerged a field study by Rossi et al. (1999) demonstrated that a small amount of “leakage” occurs between host plants. Specifically, female midges from B. frutescens mistakenly oviposit on the Iva spp. and vice versa approximately 2–3% of the time.
Interestingly, the acquisition of “enemy-freespace” has been proposed as trade-off for reduced larval performance on non-natal host plants since parasitism is initially lower after establishment on a new plant (Mira & Bernays 2002; Murphy 2004; Grosman et al. 2005). However, once the parasitoids incorporate a novel plant into their search image parasitism rates will increase (reviewed by Ishii & Shimada 2010); and, if the development times of the gall midge are asynchronous (as is the case in this system), cascading HAD of populations of both the midge and its parasitoids may develop in sympatry. Allochronic isolation between host-associated populations of the midge has already been demonstrated in field studies (Rossi & Stiling 1995, 1998, 1999; Stiling & Rossi 1998) and a recent study by Stokes et al. (2012) found genetic differences between hostassociated populations of the midge at the level of the plant genus (i.e. divergence between B. frutescens and the 2 Iva spp. populations). While differences in development time between midge populations on B. frutescens and the 2 Iva spp. have already been established, galls may persist on the 3 plants throughout the year. However, B. frutescens is the primary host plant in the warmer months (spring-fall), while galls both Iva spp. are more abundant on during the colder months, (fallspring) (Rossi & Stiling 1995). Winters in Tampa are much warmer than winters in Jacksonville and may result in faster development and increased size of galls which produces less crowded galls and larger insects. Conversely, northern gall populations on I. frutescens would have a longer development period owing to longer cold spells compared to the southern populations. Warmer winters in Tampa would not only provide galls on southern I. frutescens longer growing periods, it would also give T. umbilicatus populations originating from B. frutescens in Tampa more time to incorporate a novel host into its search image.
Incorporation of new species into a phytophagous insect's host plant range and divergence in sympatry is increasingly accepted as a primary pathway of differentiation and speciation in phytophagous insects (Craig et al. 1997; Feder 1998; Via 2000; Abrahamson et al. 2001). However, the limited number of studies investigating cascading HAD of parasitoids, have produced mixed results even within closely related systems. For instance, Stireman et al. (2006) found that parasitoids of 2 goldenrod gall makers, which have diverged genetically along host plant lines, also exhibit cascading HAD. However the eurytomid, Eurytoma gigantea Walsh, which attacks the stem galling tephritid fly, Eurosta solidaginis Fitch, another galler of the same 2 goldenrod species has not diverged in parallel with its phytophagous host (Cronin & Abrahamson 2001). In this system, differences in seasonal gall distributions, insect size and development time probably account for reduced gene flow between host-associated populations of the midge and, possibly one of its parasitoids. Prolonged isolation and differences in host quality may result in host-plant associated variation in the size and search image of the midge's parasitoids. Previous bagging studies using galls in situ suggest that T. umbilicatus becomes the dominant member of the parasitoid guild when gall diameter exceeds 7-9 mm (most likely because smaller members of the parasitoid guild are unable to reach developing insects with their shorter ovipositors) (Weis & Abrahamson 1985; Stiling & Rossi 1994; Rossi et al. 2006). Lastly, it should be mentioned that I. imbricata, which primarily inhabits beach dunes (rather than marshes like B. frutescens and I. frutescens) was probably the last plant colonized by the midge (and its parasitoids). Only 8 T. umbilicatus were collected from the I. imbricata at the southern site and a total of only 4 galls (but no insects) were obtained from northern populations of I. imbricata. This indicates that midge and T umbilicatus populations in the north have yet to gain a substantial foothold on this relatively new third host plant. Future common garden and/or reciprocal transplant experiments in which parasitoids from each site and plant species are reared on their natal and alternative host plants, as well as molecular analysis of T. umbilicatus populations collected from galls on each plant species, should reveal whether differences in size between sites and plant species are caused by phenotypic plasticity or HAD of Asphondylia populations.
ACKNOWLEDGMENTS
We thank the Department of Biological Sciences and its Coastal Biology Program for support of this project. We also thank 3 anonymous reviewers who provided valuable comments that improved the manuscript.