Within-host competition drives energy allocation trade-offs in an insect parasitoid

Host and parasitoid collection

Thirty-two M. sexta larvae in the fourth or fifth (final) instar were collected from the field near Portal, Arizona (~40 km radius) in August of 2017. Past work with Drino rhoeo showed that parasitization rate was high (~44%) in the field for M. sexta in fourth and fifth-instar stages (Wilson & Woods, 2015), so we collected a mixture of individuals that appeared healthy, and some that appeared to have been attacked by parasitoids (e.g., dried gut fluids on the skin, melanized spots where fly larvae burrowed inside the caterpillar (H.A. Woods and J.K. Wilson, 2014, unpublished data)). Larvae were raised together in a large plastic bin and fed cuttings from local Datura wrightii, their main host plant. Once larvae began to wander (Dominick & Truman, 1984), they were placed in individual plastic cups (13 cm × 12 cm × 14 cm) filled with soil. Larvae were allowed to burrow and begin pupation and were transported back to the University of Arizona in Tucson where they were kept in an experimental greenhouse for the duration of their development.

Head-capsule width as a measure of host quality

Here, we define host larval quality as the energy potential of a host to developing parasitoid larvae. Larval mass is not a good measure of larval quality, because growth, feeding and development are all affected by parasitism (Wilson & Woods, 2015), so we measured head-capsule width of hosts (all fifth-instar) as a proxy to estimate larval quality to developing flies. We examine the relationship between mass and head-capsule width in more detail in the Results section.

Fly emergence, weights and calorimetry

Seven of the 32 caterpillars (22%) collected were parasitized by Drino rhoeo and had successful fly emergence. Flies were allowed to emerge in the small plastic cups containing individual caterpillars described above. Once initial fly emergence was observed, we waited 48 h to allow all individuals to successfully emerge from the soil, and then placed the cups in a −20 °C freezer. Once flies were frozen, we transferred them to individually-labeled vials and scored sex using a dissecting scope to determine the presence of sexual patches on the ventral portion of the last abdominal segment on males. Files were sectioned into different body structures (head, abdomen, thorax, wings and legs) using a combination of scalpel, probe and forceps under a dissecting scope. Sectioned body structures were moved to small metal containers and placed in a drying oven at ~45 °C for 48 h before being placed back into vials and frozen again at −20 °C until further processing. Dried individual body sections were weighed on a microbalance (Mettler Toledo XS3DU), with legs and wings being weighed together. In total, we collected 104 individual flies from the seven hosts, though not all flies were used in all subsequent analyses because of some loss of body structures during the weighing or bombing process.

We used a Parr 6200 bomb calorimeter to determine energy content of heads, thoraces and abdomens. Because our calorimeter is designed to be used on larger tissue, we generated calibration curves for predicting caloric content based on tissue type and weight to be able to extrapolate to small weights (and respectively small energy content) that are below the threshold of detection for our calorimeter. Individual structures were mixed into different groups ranging from 1 to 10 structures from individual flies per bin (depending on the size of the tissue: more heads were needed to get to measurable weights than abdomens, because heads are smaller than abdomens) and separated by sex. We also varied the number of structures in binned groups to achieve enough variation in weight to generate accurate calibration curves. Binned tissue samples were weighed and placed in a crucible. We added 0.7 mL of mineral oil to samples as a heat spike (to increase the total energy content to a level readable by the machine). This method of measuring small samples is a standard procedure and the Parr calorimeter software automatically accounts for the mineral oil spike.

Zera & Harshman (2001) emphasize that to establish a physiological tradeoff among body functions, there is a need to establish that a specific resource (such as a specific lipid, protein or carbohydrate) is used by both functions which requires tracking the common resource in both functions. In this study we are interested in total amount of resources allocated among functions and not a specific resource. The use of energy as the common currency encompasses all resources allocated to a function and allows for identifying physiological allocation tradeoffs among functions. We note that this method cannot measure the cost of building a structure, just the energy content of the resources that are in the structure.

We created tissue-specific calibration curves to generate estimates of the calorimetric content of different tissues based on weight. We focused on the three main body segments (heads, thoraces and abdomens). To generate curves, we fit linear regression models for each tissue type that were forced through the origin. All models had good predictive power (p < 0.05 for all models), though the predictive power for estimating the energetic content of heads (R² = 0.58) was less than for thoraces (R² = 0.90) and abdomens (R² = 0.91) because of smaller sample sizes (it takes many more fly-heads to generate the measurable weights than it does thoraces or abdomens) and more variation in measurements at smaller weights. We did not have enough flies to generate separate calibration curves for males and females, so all bins consisted of flies from a single sex, and calibration curves are tissue-specific but not sex specific, which might result in additional variation in extrapolations if there were large differences between the weight-energy correlations of different sexes.

Head capsule and host tissue calorimetry

To determine the relationship between head-capsule width and the usable energetic content of hosts, we used late fifth-instar M. sexta caterpillars from our colony in Tucson, AZ, USA. We measured head-capsule width and then froze caterpillars in a −20 °C freezer. We then thawed and separated caterpillar structures into two groups: skin and head-capsule and internal tissues that would be available to developing parasitoids (this included hemolymph, tracheae, muscle, but not gut tissue or any remaining food). We dried samples for each caterpillar in a drying oven at ~45 °C for a minimum of 48 h. We crushed dried tissue samples and split samples that were too large for our calorimeter into three sub-samples before bombing. We used a bomb calorimeter (Parr 6200—methods described above) with a 600 µL mineral oil spike to measure the energetic content of each sub-sample before combining for further analysis.

Data analyses

All analyses were performed in R (Version 3.5.0 ‘Joy in Playing,’ www.r-project.org). We used linear mixed effects models (nlme) for modeling the relationship between adult fly size, host quality, cohort size and sex and AIC scores for model comparison and selection. Additional packages (effects, piecewiseSEM) were used to generate population level trend lines and marginal and conditional R² values. In spite of our limited sample size of hosts (n = 7), we included host as a random effect (with random intercepts) to help control for variance in conditions among hosts. Additionally, we used ordinary least squares (OLS) regression for examining allometric relationships between body weight and tissue weight and energetic content and AIC scores for model comparison and selection (Table S1). Though some researchers have advocated the use of reduced-major axis regression, recent work has shown that OLS regression is better suited in many cases, especially those similar to ours where there is comparatively little measurement error (Kilmer & Rodríguez, 2017). In comparisons of relative mass and energy allocation, polynomial models were fit where appropriate. All data and code are archived and available on Zenodo (DOI 10.5281/zenodo.3356991).

Cohort size, host quality and adult fly weight

On average, adult fly body size did not differ between males and females (males = 8.74 ± 3.58 mg, females = 8.51 ± 3.73 mg; F_{1, 90} = 0.091, p = 0.7641). On average, adult fly dry weight decreased with increasing cohort size (Fig. 1A), with an average fly from the largest cohort (38 flies) weighing 36% of an average fly from the smallest cohort (two flies). Additionally, adult fly dry weight increased with increasing host quality (head-capsule width; Fig. 1B). Overall, the best model that explained adult fly weight was one that included cohort size and host quality additively (Tables 1 and 2) with host included as a random effect (random intercepts). We also performed a multiple linear regression on a reduced number of data points (n = 7) that were average fly-weight values for each cohort, to confirm the biological pattern we show here and reaffirm that the linear-mixed effects model framework is accounting for any pseudo-replication of sampling multiple flies within a single host. This analysis showed that similar significant effects of cohort size (p = 0.05) and host head capsule width (p = 0.03) with good predictive power (p = 0.006, R² = 0.76). Head-capsule width is a frequently-used proxy for body size, and typically has strong positive correlations with body size (Smock, 1980; Potter, Davidowitz & Woods, 2011; D’Amico, Davidowitz & Nijhout, 2001)—it is also the best measurement of host-quality in this system because of the complex interactions between parasitism, host-feeding and body size (Wilson & Woods, 2015). We found some support that head-capsule width is positively associated with wandering-weight (F_{1, 5} = 6.961, p = 0.046, R² = 0.498) in M. sexta caterpillars in the field, despite a relatively small sample size and that caterpillars were parasitized by varying numbers of tachinid larvae. Additionally, in a series of measurements on lab-reared M. sexta, we found that pre-wander weight was positively correlated with the caloric content of hosts excluding skin and the gut (F_{1, 13} = 54.01, p < 0.0001, R² = 0.79). Together, these data suggest that head-capsule width functions as a good proxy for both body size and caloric content available to developing tachinids.

Size-relative allocation to different body structures

Overall, flies had relatively smaller heads as size increased, with no difference between males and females (t = 1.102, p = 0.273), though the best fit model for these data was a second-order polynomial, with a slight increase in investment in heads at large body sizes (F_{2, 89} = 51.8, p < 0.001, R² = 0.53; Fig. 2A; Table S1). Normalized wing weight (scaled to individual body size—we use this definition of normalized throughout) decreased with body size (F_{2, 89} = 10.74, p = 0.001, R² = 0.09), with no differences between the sexes (t = −0.768, p = 0.444; Fig. 2B). Normalized leg weight decreased linearly with body size (F_{2, 89} = 5.169, p = 0.008, R² = 0.50), and while there was a significant effect of sex (t = −2.617, p = 0.01), the effect size was small, with little difference between males and females, except only at small body sizes (Fig. 2C). Normalized thorax weight was best fit with a polynomial model that included sex as an additive effect (F_{3, 88} = 35.98, p < 0.001, R² = 0.54)—relative thorax weights were highest in medium-sized flies, and were higher overall in males (t = 7.909, p < 0.001), with the greatest difference between males and females occurring in medium-sized flies (Fig. 2D). Finally, normalized abdomen weight increased linearly with fly weight in both males and females (F_{2, 89} = 65.18, p < 0.001, R² = 0.56), though females showed increased relative investment across all body weights (t = −8.399, p < 0.001; Fig. 2E).

Energy tradeoffs and comparisons among body segments

Body segments differed in their average energy content with abdomens being the most energy dense at 4.36 ± 0.44 calories/mg, followed by thoraces at 4.05 ± 0.33 calories/mg, followed by heads at 1.96 ± 0.17 calories/mg. We compared the percentage of calories devoted to each body segment for male and female flies across body sizes and found strong potential trade-offs between thoraces and abdomens in both males and females, though the pattern is strongest for male flies of moderate size (Fig. 3A). We note here that while a negative correlation between two traits has often been used as evidence for trade-offs (Zera & Harshman, 2001; Stearns, 1989), the interactions among traits are complex, and indirect effects may impact negative correlations. We also compared the relative amount of energy devoted to one segment after controlling for the total energy content of the three main body segments. Both male and female flies showed a strong negative correlation between the relative amount of energy invested in thoraces and abdomens (F_{1, 99} = 1983, p < 0.001, R² = 0.952; Fig. 3B), with no difference between males and females (p = 0.209). There was no significant pattern of trade-offs between heads and thoraces or heads and abdomens for male or female flies (F_{1, 99} = 1.225, p = 0.271; F_{1, 99} = 1.187, p = 0.279).