Introduction
Food quality and quantity can influence a host's ability to resist and respond to infection through effects on host condition. Hosts in poor condition generally suffer greater effects of infection in terms of retarded development, reduced fitness and lower survival rates compared to hosts in relatively better condition (Brown et al., Reference Brown, Loosli and Schmid-Hempel2000). Many studies have been able to quantify increases in immune function in hosts in relatively better condition and this may drive much of the improved outcomes observed in these hosts. Hosts that have access to sufficient and high-quality nutrition should have more resources to invest in defense, as mounting an immune response is metabolically costly (Zuk and Stoehr, Reference Zuk and Stoehr2002). For example, nutritional deprivation has been shown to reduce the melanization response (Suwanchaichinda and Paskewitz, Reference Suwanchaichinda and Paskewitz1998; Lee et al., Reference Lee, Simpson and Wilson2008) and antimicrobial gene upregulation in insects (Brunner et al., Reference Brunner, Schmid-Hempel and Barribeau2014). Thus, well-fed hosts are generally able to invest more in immunity, fecundity and somatic maintenance in response to infection compared to their nutritionally deprived counterparts (Kopp and Medzhitov, Reference Kopp and Medzhitov2009).
Many studies that investigate how host food affects the cost or outcome of parasitism focus on the amount of food available to a host (e.g. Brown et al., Reference Brown, Loosli and Schmid-Hempel2000). While food availability is both crucial to host condition and highly variable in nature, food quality (e.g. micro and macro nutrients, plant secondary compounds) may be just as variable and important for generalist feeders. Generally, hosts feeding on a higher quality diet have greater measures of fitness proxies and infection outcomes. Protein, in particular, seems to increase host resistance to parasites and lead to more favorable outcomes for the host; for example protein supplementation increases body size and resistance to Batrachochytrium dendrobatidis infection in amphibians, and increases survival and resistance to nucleopolyhedrovirus in lepidopterans (Lee et al., Reference Lee, Cory, Wilson, Raubenheimer and Simpson2006; Venesky et al., Reference Venesky, Wilcoxen, Rensel, Rollins-Smith, Kerby and Parris2012). Additionally, the chemical composition and amount of secondary compounds produced by plants are known to alter insect herbivore-parasite interactions (reviewed by Cory and Hoover, Reference Cory and Hoover2006) and may also impact aquatic detritus-based communities (Davidson et al., Reference Davidson, Larsen and Palmer2012; Stoler et al., Reference Stoler, Berven and Raffel2016). Despite evidence that both the amount of available food and the nutritive quality of food being important for hosts, less is known about their relative importance to each other or potential interactive effects.
Host food can not only influence host condition and infection outcomes, but can potentially influence parasite fitness. There is considerable debate in the literature as to whether well-fed, good-condition hosts or ill-fed, poor-condition hosts will be better habitats for the parasites that exploit them. In some systems, poor quality hosts have parasites that produce more infectious stages or have faster growth rates (Ezenwa, Reference Ezenwa2004; Vale et al., Reference Vale, Choisy and Little2013), whereas the reverse is true in other systems (Pulkkinen and Ebert, Reference Pulkkinen and Ebert2004; Logan et al., Reference Logan, Ruiz-González and Brown2005; Dube et al., Reference Dube, Hund, Turbek and Safran2018). It is not hard to conceive why these discrepancies exist in the literature; parasites of well-fed hosts will have a host with plenty of resources to exploit but also a host capable of mounting a strong immune response (Bize et al., Reference Bize, Jeanneret, Klopfenstein and Roulin2008; Seppälä et al., Reference Seppälä, Liljeroos, Karvonen and Jokela2008). Thus, the specific mode of transmission, host immune strategy, host lineage, degree of environmental heterogeneity and evolutionary history of host and parasite likely will influence whether well-fed hosts produce more or less fit parasites (Lafferty and Holt, Reference Lafferty and Holt2003; Bedhomme et al., Reference Bedhomme, Agnew, Sidobre and Michalakis2004). Direct measures of parasite fitness, such as the basic reproductive number R0, are difficult to determine experimentally as they require measuring many aspects of a parasite's lifetime fitness (Hartemink et al., Reference Hartemink, Randolph, Davis and Heesterbeek2008). Thus, more easily determined proxy measures of fitness, such as parasite size, are often used and are correlated with fitness (Tseng and Myers, Reference Tseng and Myers2014) and many parasite taxa have strong size–fecundity relationships (Sasall and Morand, Reference Sasall and Morand1998; Holfeld, Reference Holfeld2000; Rowe et al., Reference Rowe, McMaster, Emery and Sangster2008; Costanzo et al., Reference Costanzo, Westby and Medley2018).
In recent decades, a growing body of work has demonstrated that both inter- and intra-specific competition for resources within hosts may have important consequences for parasite fitness correlates and transmission dynamics (Tompkins and Hudson, Reference Tompkins and Hudson1999; Wedekind et al., Reference Wedekind, Christen, Schärer and Treichel2000; de Roode et al., Reference de Roode, Read, Chan and Mackinnon2003). Resource competition in free-living organisms is well known to be context-dependent (Taniguchi and Nakano, Reference Taniguchi and Nakano2000; Murrell and Juliano, Reference Murrell and Juliano2008). Little research, however, has been conducted to understand under which conditions with-in host competition occurs. There may be strong seasonal patterns where density-dependent competition is only detected during certain times of the year (Irvine et al., Reference Irvine, Stien, Dallas, Halvorsen, Langvatn and Albon2001), may depend on host age and activation of the immune system (Patterson and Viney, Reference Patterson and Viney2002; Chylinski et al., Reference Chylinski, Boag, Stear and Cattadori2009), or may be altered by the supplementation of a nutrient required for parasite development (Wale et al., Reference Wale, Sim and Read2017). As host food quantity and quality impact host and parasite fitness, it is likely they also impact within-host dynamics such as competition.
Gregarine parasites (phylum Apicomplexa: Ascogregarina) are common obligate parasites of invertebrates. For gregarine species that infect mosquitoes, the infection begins when free-living oocysts are ingested by larvae in the aquatic habitat. The oocysts then migrate to the larval midgut where they release sporozoites that enter the epithelial cells of the gut and develop into trophozoites. New oocysts are formed in the malpighian tubules of adults and the infection cycle continues as new oocysts are returned to the larval habitat when infected adults or pupae die in or on the water, when adults defecate upon emergence, or when they return to lay eggs (more life-cyle detail available in Beier and Craig, Reference Beier and Craig1985; Lantova and Volf, Reference Lantova and Volf2014). Aedes triseriatus, the Eastern tree-hole mosquito, is parasitized by Ascogregarina barretti, and though there is little evidence that As. barretti infection causes direct mortality in Ae. triseriatus, infection can alter larval behavior (Soghigian et al., Reference Soghigian, Valsdottir and Livdahl2017), prolong female development time (Walker et al., Reference Walker, Poirier and Veldman1987) and decrease adult size (Walker et al., Reference Walker, Poirier and Veldman1987; Siegel et al., Reference Siegel, Novak and Maddox1992), reducing fecundity and population growth (Soghigian and Livdahl, Reference Soghigian and Livdahl2017).
In this paper, we tested the effects of host food quality and quantity on fitness proxies for both host and parasite in addition to within-host competition among parasites using the larval stage of the mosquito Ae. triseriatus and its protozoan parasite As. barretti. We predicted that if competition were occurring we would see a reduction in parasite size with increasing parasite counts. Using a model-selection approach, we were able to determine which host and parasite metrics influenced each other in terms of host size, parasite number, parasite size and extent of infection.
Materials and methods
Experimental design
Each experimental unit consisted of one second instar Ae. triseriatus larva placed in 12 mL of infusion in a 15 mL glass vial with 21.5 µL oocyst suspension containing approximately 7000 gregarine oocysts. There were 30 larvae per treatment at the start of the experiment (see Supplementary Table S1). Larvae were reared on infusions of one of four food treatments; (1) quality (animal or leaf detritus) and (2) quantity (high or low) at the following concentrations: animal-low (0.005 g per larva, n = 12), animal-high (0.01 g per larva, n = 21), leaf-low (0.084 g per larva, n = 19) and leaf-high (0.144 g per larva, n = 22). These two detritus types and quantities were chosen because they have been shown to have differential effects on mosquito growth and host quality (Yee and Juliano, Reference Yee and Juliano2006) and preliminary trials in our laboratory showed that they were sufficient to allow for development to the fourth instar with minimal death. Leaves and insect carcasses are common inputs in these aquatic container habitats where mosquito larvae live and become infected (Fish and Carpenter, Reference Fish and Carpenter1982; Beier and Craig, Reference Beier and Craig1985; Daugherty et al., Reference Daugherty, Alto and Juliano2000). The animal infusion consisted of camel crickets (Ceuthophilus spp.) collected at Tyson Research Center (Eureka, MO) dried at 50 °C for 48 h, and homogenized in a kitchen blender. The leaf infusion consisted of mixed species of senescent oak leaves (Quercus spp.) collected from the forest floor, dried for 1 week at 50 °C and stored at room temperature. Leaves were cut into one inch squares with galls and petioles removed. The four infusions were steeped for 8 days at 25 °C to allow the growth of microorganisms. Experimental mosquito larvae were reared in a growth chamber at 25 °C with a 14:10 day:light photoperiod.
Larval developmental stage (instar) was monitored daily by looking for the exuviae shed between each molt. Midguts were dissected from fourth instar larvae 9–10 days after the initial hatch and were examined for the presence of gregarine trophozoites (Munstermann and Wesson, Reference Munstermann and Wesson1990). The midguts were photographed at 100× power for further analysis. Larval heads were also removed and stored in 70% ethanol until they could be photographed at 100× magnification and were used as a measure of mosquito size (Alto et al., Reference Alto, Kesavaraju, Juliano and Philip Lounibos2009; Murrell and Juliano, Reference Murrell and Juliano2013). The images of both midguts and heads were given obscure names so that the researcher measuring them was blind to the treatment identity of the photos.
We measured four response variables for each mosquito larva: total parasite count, total parasite area, mean size of individual parasites and larval head width. Larval mosquito head width (mm) was measured from photos using imageJ (Schneider et al., Reference Schneider, Rasband and Eliceiri2012) and was measured at the outer edge of the eyes which is the furthest two points on the larval head from the dorsal perspective in Aedes mosquitoes. The larval head capsule size is fixed soon after eclosion and is a reliable correlate of larval size at the fourth instar (Daly, Reference Daly1985).
The body size (area, mm2) of individual parasites was measured from photos using imageJ (Schneider et al., Reference Schneider, Rasband and Eliceiri2012). Parasites were first classified into one of three size categories, (1) large enough to measure with imageJ's automated tracing tool (n = 1778, ⩾1.86 × 10−4 mm2), (2) too small to measure with the tracing tool (n = 2475, mean = 1.85 × 10−4, standard error = 9.41 × 10−6 mm2) and (3) discernably smaller than category two (n = 3381, mean = 5.89 × 10−5, standard error = 3.51 × 10−6 mm2). It was possible to measure parasites in the two former categories individually, but due to the time required to complete this, it was not considered feasible. In order to include the two smaller size classes in the analyses, we calculated the average parasite size for the two smaller size classes by individually measuring all of the smaller size-class parasites present in three randomly chosen mosquito larvae; the class two average was calculated among 39 individually measured parasites and the class three average was calculated among 42 individually measured parasites. Once an individual parasite was deemed too small to use the imageJ tracing tool, the parasite was recorded as falling into either size class two or three, determined by visual observation by one researcher for all parasites. Thus, the response variable ‘total parasite area’ in a mosquito larva is the sum of all of the individually measured category one parasites, plus the number of category two and three parasites multiplied by their average size. The response variable ‘mean parasite size’ within a larva is the average size of all the parasites from within individual a mosquito larva. This was calculated by averaging across all size classes by including the sizes of each individually measured category one parasites, in addition to sizes for each of class two and class three parasites where their individual sizes were assigned the average value calculated for each size category (two and three). In addition to calculating mean parasite size, we also tallied the number of parasites within each of the three categories to determine whether the quantity of parasite within each size class was more or less than expected.
Statistical analyses
We restricted our analysis to individual mosquito larvae for which we had collected data for all 4 response variables (total parasite count, total parasite area, mean size of individual parasites and larval head width; n = 74). Data were not captured for individual larvae if they died before reaching the forth instar, the midgut dissection was unsuccessful, the head was destroyed during dissection, or if they failed to reach the fourth instar (see Supplementary Table S1 for the distribution between treatment groups).
For the response variables total parasite count, total parasite infection area, mean parasite size and larval head width, we built generalized linear models to determine the effect of detritus quality and quantity on each variable. Additionally, we added our other response variables into the models as covariates (total parasite count, total parasite area and larval head size) where we had a priori hypotheses about how larval growth would affect parasite development and how parasitism would affect larval size (Table 1). For example, we predicted that evidence for within-host competition would manifest in a negative correlation between parasite count and individual parasite size. For each model, we chose the distribution that best fit the data and had minimal overdispersion as evaluated by the deviance for each model. For instance, for models fit using a Poisson or a negative binomial error distribution, the distribution that fit the data with lower deviance was selected for the entire model set for that response variable. We used the negative binomial distribution with a log link function for the total parasite count and linear models with normally distributed error for the remainder of the response variables. We created post-hoc qq plots and compared the distribution of model residuals to a randomly generated distribution (either binomial or normal) using chi-square diagnostics in R. We used information theoretic model selection to rank models for each response variable using Akaike's Information Criterion adjusted for a small sample size (AICc) (Burnham et al., Reference Burnham, Anderson and Huyvaert2010). Where appropriate, we conducted post-hoc Tukey's analyses to determine differences in pairwise comparisons for top models.
Table 1. Complete AIC table for the effects of detritus quality and quantity on parasite count, total area of parasites inside a host, mean parasite size, developmental time and larval head width
Models in bold are the best model determined by the lowest AIC score. T, detritus type (quality); A, detritus amount (quantity); HW, head width; PC, parasite count; PA, parasite area.
Some models included covariates that were also response variables where we had a priori hypotheses about how they might affect the data we collected.
As a complement to the generalized linear models assessing mean parasite area, we also evaluated frequencies of parasites within each of the three size classes by food type and quantity in a multi-way contingency table using hierarchical log-linear models in R with the base package (Gotelli and Ellison, Reference Gotelli and Ellison2004). We assembled a suite of nine hierarchical models including a fully saturated model and models with each of the variables removed, calculated G2 statistic for each and ranked individual models using AIC (Table 3, Burnham and Anderson, Reference Burnham and Anderson2002). To determine whether more or fewer parasites than expected within a given size class were present in larvae exposed to each food type and quantity, we assessed complete independence by comparing a saturated model (all possible interactions) and to one containing only additive effects of the three variables. To assess conditional independence among the variables, we compared pairs of models to one another within the hierarchical set; for instance, to determine conditional independence between food type and size class, we compared a model containing the all three possible interactions (food type × size class + food quantity × size class + food type × food quantity) to a similar model with the interaction between the variable of interest removed (e.g. food type × size class + food quantity × size class; a result of non-independence here would indicate the variables food type and food amount are not independent).
Collection of mosquitoes and parasites
Mosquito eggs were collected on seed germination paper taped above the water line in previously established water barrels in the forest at Tyson Research Center. Papers with eggs were dried and stored at 25° C with a 14:10 light:dark photoperiod until use (no longer than 2 months). Eggs were hatched by submersing egg papers in a solution of 0.35 g Difco™ Nutrient Broth (Becton, Dickinson and Company, Sparks, MD) per 1 L deionized water for 24 h. The hatched first instar larvae were rinsed and placed communally in a 7 g L−1 mixed-species oak leaf infusion with a small addition of a 1:1 mix of lactalbumen:liver-powder (MP Biomedicals, Solon, OH). After 72 h the larvae had molted to second instar, the first size at which they can be reliably identified to species (personal observation), and Ae. triseriatus larvae were identified and added individually to each of 120 experimental microcosms.
A gregarine oocyst suspension was prepared using modified methods of Beier and Craig (Reference Beier and Craig1985). To do this, we collected Ae. triseriatus larvae from extant water barrels at Tyson Research Center that had previous evidence of gregarine parasite occurrence (unpublished data), in addition to hatching mosquito larvae from field-collected eggs and rearing them in the lab with a previously prepared gregarine oocyst suspension. Pupae were allowed to eclose in a flask covered with netting with the adults left to die on the surface of the water. Contents of the Erlenmeyer flask were homogenized using a kitchen blender and the homogenate was sieved through progressively smaller filters (160 µm, 80 µm, 66 µm and 20 µm). The sieved suspension was transferred to 50 mL conical tubes and concentrated in an Eppendorf 5810R benchtop centrifuge for 7 min at 2500 rpm (978 ×g). The precipitate was collected, combined and the number of oocysts per 1 µL were quantified using a haemocytometer.
Results
The best model for total parasite count was the simplest model which included only the main effects of detritus quality and quantity (Table 1). Detritus quality significantly affected the number of parasites but the quantity was not significant (Table 2). Larvae reared with leaf detritus had more than twice the number of parasites than those reared with animal detritus (Fig. 1a). The best model for predicting total parasite area included the effects of detritus quality, quantity and the quality × quantity interaction, although the interaction was not significant (Tables 1 and 2). However, larvae reared with the higher quantity of leaf detritus had higher total parasite area in their guts compared to those reared with the lower amount (Fig. 1b). The best model for mean parasite size was the model with the effects of detritus quality, quantity, the quality × quantity interaction, larval head size, total parasite count and the parasite count × detritus quality interaction (Tables 1 and 2). Parasites in larvae reared on high detritus levels were larger than in larvae reared on low detritus levels (Table 2). Hierarchical log-linear models evaluating counts of parasites within size classes revealed significant associations between food type and quantity. A test of complete independence revealed food quality, quantity and size class were not independent (comparison of models 1 vs 8, Table 3; P < 0.001). Tests of conditional independence revealed food type and food quantity were not independent (comparison of models 7 vs 8, Table 3; P = 0.04), and food quantity and size class were not independent (comparison of models 6 vs 8, Table 3; P < 0.001). These significant associations are driven by a higher than expected frequency of size class one and a lower than expected frequency of size class three in the high-quantity leaf food source, in addition to a lower than expected frequency of size class one and higher than expected frequency of size class three in the low-quantity leaf food source (Table 4).
Fig. 1. (A) Means and standard errors for the effect of the two detritus types (animal and leaf) and quantities (low and high) on the total number of parasites in the host, (B) on the total area the parasites occupied in the host, and (C) host size (larval head width). Lower case letters indicate significant differences between bars (Tukey's HSD) and are not indicated for the total parasite count as there was not an appropriate post-hoc test with the negative binomial.
Table 2. Model output and parameter estimates for the best model, determined by the lowest AICc value, for each of the response variables
Table 3. Hierarchical models for multi-way contingency analyses of frequencies of each parasite size class within each treatment combination
T, detritus type (quality); A, detritus amount (quantity); HW, head width; PC, parasite count; PA, parasite area.
The G2 statistic results from a likelihood ratio test comparing observed counts of parasites within each size class with the distribution of expected counts. The lowest AIC value indicates the ‘best’ model fitting the data; the saturated model (model 9) by definition has the best fit/lowest AIC because it accounts for all possible variation.
Table 4. Multi-way contingency table for counts of parasites within each size class by food quality and quantity
Associated hierarchical models testing independence among categories and levels are provided in Table 2. Counts in bold represent observed frequencies that are either significantly higher (*) or lower than expected (**) based upon tests of complete and conditional independence between category levels.
The best model for larval head width included the effects of detritus quality, quantity and the quality × quantity interaction (Table 1). Larvae were largest when reared on the higher quality animal detritus, and larvae reared on leaf detritus were larger in the low quantity of food (Table 2 and Fig. 1c).
Discussion
In this study, we have demonstrated that both the quality and quantity of detritus in aquatic larval habitats affects the number of parasites in a host, the total parasite area within a host, the average size of individual parasites and the relationship between the number of parasites and the size of individual parasites. While detritus quality and quantity were both important, our results showed that each impacted host–parasite interactions in different ways.
We found that gregarine-infected mosquito larvae reared on a high quantity of animal detritus grew larger than those reared on plant detritus or a lower amount of animal detritus. Animal carcasses, primarily dead invertebrates, are a common and important detritus source in larval habitats (Bara et al., Reference Bara, Clark and Remold2014), and strong positive effects of animal detritus on larval mass and survivorship have been detected in other studies (Daugherty et al., Reference Daugherty, Alto and Juliano2000; Bara et al., Reference Bara, Clark and Remold2014; Yee and Juliano, Reference Yee and Juliano2006). In addition to animal detritus, plant detritus is a ubiquitous and important component of the food web in both natural and anthropogenic habitats of tree-hole mosquitos like Ae. triseriatus (Walker et al., Reference Walker, Poirier and Veldman1987; Walker and Merritt, Reference Walker and Merritt1988). Larvae feed on the microorganisms (e.g. bacteria, fungi and protists) that decompose detritus in the larval habitat (Fish and Carpenter, Reference Fish and Carpenter1982). Leaf detritus has been shown to be an inferior resource compared to animal detritus; larvae reared on leaf detritus have decreased mass, survivorship and fitness compared to larvae reared on animal detritus of the same concentration (Yee and Juliano, Reference Yee and Juliano2006). The leaching of secondary compounds such as tannins from high concentrations of leaf detritus can also be toxic to aquatic organisms (Mercer, Reference Mercer1993; Mercer and Anderson, Reference Mercer and Anderson1994; Earl et al., Reference Earl and Semlitsch2015). Tannins from leaf detritus have been shown to increase mortality, the duration of the larval period and decrease the adult size of Aedes albopictus mosquitoes (Sota, Reference Sota1993). Thus, larvae reared on higher amounts of animal detritus are generally well-fed hosts likely in better condition than their counterparts reared in either leaf detritus or on lower amounts of animal detritus. In this study, we did not include uninfected control larvae, but we did observe that larvae reared on the highest amount of animal detritus were the largest of the four treatment groups, and for insects, large size is indicative of higher fitness (Kingsolver and Huey, Reference Kingsolver and Huey2008; Costanzo et al., Reference Costanzo, Westby and Medley2018).
The quality of larval food source had a greater effect on the total parasite count than did the quantity of food. Larvae reared on animal detritus contained significantly fewer parasites than those reared on the lower quality leaf detritus. As all larvae in this experiment received the same infectious dose from the same population of parasites, the number of parasites present in each host is essentially the number of parasites that were able to infect and establish in the host's midgut. This is a measure of host susceptibility, and is not necessarily a measure of future relative parasite fitness. The high-quality, high-protein animal food diet led to hosts that were less susceptible to infection than hosts fed the lower quality leaf diet.
The total area the parasites occupied in the host was also substantially influenced by the quality of food available, but not by the quantity. Parasites in hosts reared on higher amounts of food had a larger total area than those reared on the lower amount of food. This implies that host susceptibility, in our study system, may be driven by the quality of food, but once a parasite is established in a host, the quantity of available food may be more important for parasite growth.
Both food quality and quantity influenced the average size of individual parasites. Parasites were smaller when in hosts reared on the greater amount of food for both types and were the largest when reared on the high amount of the low-quality food. Parasites were larger in larger hosts overall. Significant positive relationships between host and parasite size have been documented in many taxa (Taylor, Reference Taylor1988; Sasall and Morand, Reference Sasall and Morand1998; Holfeld, Reference Holfeld2000; Tsai et al., Reference Tsai, Li and Dai2001; Rowe et al., Reference Rowe, McMaster, Emery and Sangster2008) and many parasites exhibit strong size–fecundity relationships (Sasall and Morand, Reference Sasall and Morand1998; Holfeld, Reference Holfeld2000; Rowe et al., Reference Rowe, McMaster, Emery and Sangster2008). While we do not know if such a size–fecundity relationship exists for As. barretti, the gregarine parasite used in this study, it is reasonable to assume that larger individuals of this species have been able to extract more resources and enjoyed more rapid development than their smaller counter parts. We did not measure parasite fecundity in this experiment, but it is probable that larger individuals may also contribute more propagules to the subsequent generation.
We found evidence of within-host competition; there was a negative relationship between the total number of parasites in a host and the average size of parasites. This is consistent with results from a similar mosquito-gregarine system where parasite growth rates were reduced under high parasite doses (Soghigian and Livdahl, Reference Soghigian and Livdahl2017). Previous studies have shown that the addition of competing parasites of the same or different strains can reduce the size of cogeners with-in a host (Wedekind et al., Reference Wedekind, Christen, Schärer and Treichel2000; Fredensborg and Poulin, Reference Fredensborg and Poulin2005; Lagrue and Poulin, Reference Lagrue and Poulin2008). There was a significant interaction between food quality and total parasite count which implies that the relationship between parasite count and size is dependent on host food quality. Within-host competition is a widely accepted phenomenon (de Roode et al., Reference de Roode, Helinski, Anwar and Read2005; Mideo, Reference Mideo2009), but under which conditions it is expected to occur is much less studied. Additionally, it is not well understood if, or when, these density-dependent effects are driven by competition for limited resources or by top-down effects of the host immune system. Density-dependent effects in another study were more intense for malaria parasites in mosquito hosts that had supplemented nutrition (Wale et al., Reference Wale, Sim and Read2017). However, nematodes infecting rats do not show signs of density dependent regulation in immunocompromised hosts even at different infectious doses (Patterson and Viney, Reference Patterson and Viney2002) implying a role for the host's immune system.
While we did not measure any parameters related to host immune function, we speculate that the reduced size of parasites in the animal detritus treatment could be due to an increase in host immune function constraining parasite growth. One important component of the insect immune response against parasites, including gregarines (Comiskey et al., Reference Comiskey, Lowrie and Wesson1999), is encapsulation and melanization (Siva-Jothy et al., Reference Siva-Jothy, Moret and Rolff2005). This immune response was shown to decrease linearly as food levels declined in larvae of the mosquito Anopheles gambiae (Suwanchaichinda and Paskewitz, Reference Suwanchaichinda and Paskewitz1998). The total number of haemocytes, cells pivotal in many insect immunological pathways, in female Aedes aegypti decreased when larvae were reared under nutritional stress (Telang et al., Reference Telang, Qayum, Parker, Sacchetta and Byrnes2012). Additionally, high levels of tannins and other polyphenolic compounds leached from oak leaf detritus (used in this study) degrade the larval midgut epithelium, an important component of larval immune systems (David et al., Reference David, Rey, Pautou and Meyran2000). Tannins can also inhibit the growth of microorganisms that serve as the food source for larvae and sequester protein in the larval aquatic habitat (Hättenschwiler and Vitousek, Reference Hättenschwiler and Vitousek2000). Diets low in protein have been shown to reduce immunological activity in hymenopterans (Brunner et al., Reference Brunner, Schmid-Hempel and Barribeau2014) and amphibians (Venesky et al., Reference Venesky, Wilcoxen, Rensel, Rollins-Smith, Kerby and Parris2012). Alteration to the immune system, due to high tannin concentrations or reductions in available protein, in hosts fed higher amounts of leaf detritus may also explain the differences we observed in parasite area between the two food quantities.
In this experiment, we provide an example where hosts fed a high-quality diet are inferior hosts for their parasites. We also show that both food quality and quantity impact measures of parasite performance. Quality affected parasite number more than quantity, while quantity had a greater effect on parasite size. Food quality and quantity are variable in nature, both spatially and temporally. This variability will influence disease dynamics in addition to, as our data suggest, within-host dynamics. Future studies should expand the work presented here by testing many more combinations of food types and amounts to better understand the likely complicated interplay between host nutrition and host–parasite interactions.
Supplementary material
The supplementary material for this article can be found at https://doi.org/10.1017/S0031182019000994.
Acknowledgements
We thank Thomas Van Horn, Lexie Beckermann and Delilah Sayer for laboratory assistance and Susan Flowers for program support.
Financial support
This work was funded by Tyson Research Center.
Conflicts of interest
None.
Ethical standards
Not applicable.
