a. Collections

We set out trap nests to survey cavity-nesting bees and wasps at approximately 200 locations, in four site types: urban parks, gardens, and green roofs, in the city of Toronto, Canada, over a three-year period (2011–2013). The trap nests were made of 30 cardboard tubes of three different inner widths (3.4 mm, 5.5 mm, 7.6 mm) (10 tubes per width) into which bees and wasps provision nests. Isodontia mexicana readily nests in trap nests [26,19,5] and exclusively used the largest of the three widths. All trap nests were opened and each individual wasp larva was labelled with a unique ID number, placed in an incubation tray (Corning 24-cell assay tray), and overwintered at 4°C. The individuals were moved to a sealed growth chamber in the April following their collection and artificial overwintering and were kept at 26°C and ~60% humidity until they emerged as adults. As a result of opening all nests and rearing each larva individually, any E. auripedis present would have come from the mother’s original provisioning and not from post-collection contamination.

The resulting adult I. mexicana wasps were pinned, labeled with the individual’s location and date of capture, and curated in the collection of the Biodiversity of Urban Green Spaces (BUGS) lab at the University of Toronto Scarborough. To complete this study, all pinned wasps were reexamined using an Olympus SZ61 microscope to quantify stylopization and body condition. Stylopized I. mexicana are identified by the characteristic, bulging extrusions on the abdomen of the wasp (Fig 2A). Each extrusion represented one E. auripedis male, and we had evidence of E. auripedis males emerging from only one wasp, in which two empty puparia and a partially emerged adult were observed (Fig 2B). No female E. auripedis were identified in the pinned specimens because none were visibly extruded and therefore undetectable. To preserve the wasp specimens, we did not dissect hosts to search for them. Of the 358 adults that emerged, we obtained sex, body size, and stylopization data for 321 individuals after excluding damaged specimens or those donated to synoptic collections. Of these, 40 stylopized wasps were identified and were parasitized by 69 individual E. auripedis males.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 2.

A. Stylopized Isodontia mexicana with six Eupathocera auripedis males embedded within the abdominal segments. Scale line represents 1mm. B. Another stylopized I. mexicana, showing one male that has eclosed but not yet exited its capsule, along with two empty capsules indicating previously emerged males. Photo credit: Steve Paiero.

https://doi.org/10.1371/journal.pone.0340623.g002

b. Morphometric measurements

Five morphological traits were measured for both stylopized and non-stylopized wasps. These morphological traits were head width (mm), head length (mm), body length (mm), wing length (mm), and abdomen width (mm) (S1 Table). These traits were individually measured using calipers (0–150 mm, 6-inch caliper, displayed two decimal places).

For stylopized individuals, the number of extrusions was used to count the number of parasitoids present per wasp. The location of each extrusion or empty puparia on the host body was recorded (Fig 2). Location of the individual was defined using three characters; left or right side of the abdomen; tergal or sternal side of the abdomen; and the abdomen segment number above the extrusion, counted from the front of the wasp to the rear.

c. Land cover classification

We used the land cover data calculated as in Xie et al. [27] from the 2008 Forest and Land Cover dataset (0.6 m raster pixel resolution of seven different cover classes) [28]. We calculated urban land cover as the sum of the proportion of buildings, roads, and other paved surfaces. Additionally, we calculated % open green space (meadows, gardens), and % closed green space (forests). All three were quantified at a 250m radius to test for relationships between land cover and I. mexicana infection rates. We could not calculate landscape variables for sites that were not completely contained within the raster dataset, hence two surveyed sites outside the City of Toronto were removed from analysis.

In total, I. mexicana were recorded from 52 locations in trap nests over all three years of the study, and morphometric data were obtained from individuals from 35 sites.

d. Analysis

All analyses were conducted using R statistical software (version 4.3.2; [29]) and collapsed over all three years of study (data used in this study are available in S1 File). First, a Pearson’s Chi-squared test with Yates’ continuity correction was used to compare the occurrence of strepsipterans between male (n = 17) and female I. mexicana (n = 23) after individuals of unknown sex were removed. Then, a Welch two sample t-test was used to compare strepsipteran abundance per individual between stylopized males and females. Additionally, generalized linear models (GLMs) were used to assess the effect of site type (predictor) on three response variables: I. mexicana abundance, number of stylopized wasps per site, and total number of strepsipterans.

Pairwise Pearson correlation coefficients were computed among the five morphometric variables and were highly correlated using the corr.test function in the psych package [30] (S2 Table). To account for multicollinearity, each variable was centered to mean zero and scaled to unit variance using z-score standardization and included in a principal component analysis (PCA) using the prcomp function in the stats package of base R. The first principal component (PC1) captured the dominant axis of body size (77.6% variance explained) with larger PC1 values representing a larger head, body, abdomen and wing length (S3 Table). To simplify analyses, we use PC1 as a single composite size variable. We fit two linear models. The first model tested the effects of sex, strepsipteran occurrence (presence/absence), and their interaction on PC1. The second model tested the same structure, but it replaced strepsipteran occurrence with strepsipteran abundance as the predictor. Site type was not included in these models.

During study, we observed that, Isodontia mexicana tended to nest in home gardens and parks disproportionately more than in community gardens and green roofs. We used a logistic regression to assess the probability of stylopization by site type to test this. Estimated marginal means were calculated for comparison using the emmeans package [31]. Probabilities of stylopization per site type were obtained by applying the inverse logit transformation to the log-odds estimated in the model.

To compare land cover with the abundance of I. mexicana, strepsipteran occurrence, and abundance, we first calculated Pearson correlation coefficients to assess collinearity among these variables (S4 Table). Percent tree cover and percent urban cover were strongly negatively correlated (r = −0.88, p < 0.001). Grass cover was moderately negatively correlated with tree cover (r = −0.39, p = 0.01) and weakly correlated with urban cover (r = −0.10, p = 0.49). These results served as justification to reduce dimensionality, we conducted a principal component analysis (PCA) on standardized values of the three landscape variables (S5 Table; S1 Fig). The first two components (PC1, PC2) explained 100% of the variance and were used as composite landscape predictors in subsequent models. PC1 (63.1%) represented a gradient from urban cover to treed area, with higher values indicating greater tree cover. PC2 (36.9%) represented a gradient of open green space (e.g., grass), with higher values indicating more open green area (S1 Fig).

We used GLMs to test whether I. mexicana abundance and stylopization were associated with the land use gradients. We first fit a Poisson GLM and identified overdispersion (S2 Fig) and so to account for this, we modelled I. mexicana abundance using a negative binomial GLM (glm.nb, MASS package; diagnostic plots: S3 Fig) with PC1, PC2, and site type as predictors and wasp abundance as the response variable.

To test whether site-level strepsipteran occurrence was associated with landscape composition, we fit a binomial GLM with PC1 and PC2 as predictors and strepsipteran occurrence (presence/absence) as the binary response. We also obtained odds ratios (OR) from the model coefficients to assess the odds of strepsipteran occurrence change with each unit increase in landscape variables. Finally, we modelled total strepsipteran abundance using the same procedure as for wasp abundance; by fitting a Poisson GLM first, identifying overdispersion (S4 Fig), and refitting with a negative binomial GLM (S5 Fig).

a. Stylopization is sex-specific in fitness costs

First, we found that stylopization was equally likely in both sexes, rejecting our first hypothesis that stylopization rates would be higher in female hosts because they are provisioned with more prey. As female and male I. mexicana offspring occur in the same natal nest, infection risk appears to be determined primarily by whether the provisioning female inadvertently introduces triungulin larvae into the nest. Evidence from our study suggests triungulin infect whichever sex is being provisioned at the time of interception.

We find partial support for our second hypothesis, which predicted body size reductions in stylopized hosts. Stylopized males were significantly smaller than uninfected males, whereas no size difference was detected among females. As body size is a key correlate of fitness in solitary wasps, including I. mexicana [20], this sex-specific pattern suggests sublethal effects, and that females may better buffer the energetic costs of stylopization. From the perspective of the strepsipteran, maintaining host females in good condition may also be adaptive. Larger females provision nests and may range farther, increasing opportunities for strepsipteran dispersal during flower visitation. In bees, larger body size is associated with greater foraging distance and efficiency [32], and similar relationships likely apply to solitary wasps. Thus, selection may favour stylopization strategies that minimize fitness costs to female hosts, allowing infected females to contribute to strepsipteran transmission dynamics, while smaller, shorter-lived males may experience greater body condition-related developmental constraints and associated behaviour from stylopization. Sex-specific behaviours related to increasing transmission have been documented among parasitized and non-parasitized insects [33]. In I. mexicana, this could serve to maintain dispersal via visits to high quality resources such as Asters and mints [23], creating hotspots for infection.

b. Density-dependent relationship

We found support for our third hypothesis that stylopization rates exhibit host density dependence. With higher I. mexicana abundance, there was significantly higher stylopization and marginally higher strepsipteran abundance. This aligns with classical host-parasite models and empirical studies in which parasite transmission scales positively with host density [34,35]. Greater host density likely increases encounter rates between ovipositing strepsipterans and susceptible I. mexicana hosts. As I. mexicana is widespread in North America, where it is native, and currently expanding throughout Europe [36], it would be interesting to know whether E. auripedis co-occurs in its non-native environment and whether the association is as widespread. Within hosts, density dependence is also evident: over half (55%) of infected adults harboured only a single strepsipteran. This pattern suggests that space and resource limitation within the host abdomen constrain the survival of multiple parasites [37], consistent with density-dependent resource competition [38]. With that said, we documented five and six strepsipterans on two individuals, respectively. This represents a level of superparasitism associated with substantial fitness costs and accelerated parasite development in other strepsipteran-host systems [39,40,41].

c. Land use gradients

We find support for our fourth and final hypothesis that stylopization would decline with increasing urbanization. Stylopization was negatively correlated with urbanization, but our principal coordinate analysis suggests this pattern is better explained by a positive association with tree cover, as indicated by the significant relationship between stylopization and PC1 (which reflects increasing tree cover and decreasing urban cover). It is important to note our results apply specifically to nesting sites, even though I. mexicana forage across many different urban habitat types. Tree cover appears to represent high quality environments for nesting in I. mexicana [27], however, foraging adults may fly to open green spaces such as pollinator gardens and meadows where tree crickets, nectar resources (asters, mints), and grass used to partition brood cells in nests, are all located. In sum, urbanization did not uniformly disrupt these host-parasite interactions; instead, stylopization was related more to habitat quality and host density.

These findings demonstrate the nuance and context-dependence of the enemy release hypothesis (e.g., [1]) in urban systems. While stylopization declined with urbanization, this pattern reflected habitat-mediated reductions in host wasp density rather than an intrinsic escape from natural enemies. In high quality urban green spaces, host-parasite interactions remained intact, consistent with evidence that urban-tolerant species can sustain parasites where suitable resources persist [42]. Rather than broadly suppressing stylopization, urbanization may result in heterogeneous enemy landscapes, with parasitism structured by resource availability and host density [43,42].