Experimental warming influences species abundances in a Drosophila host community through direct effects on species performance rather than altered competition and parasitism

Current global warming trends are expected to have direct effects on species through their sensitivity to temperature, as well as on their biotic interactions, with cascading indirect effects on species, communities, and entire ecosystems. To predict the community-level consequences of global change we need to understand the relative roles of both the direct and indirect effects of warming. We used a laboratory experiment to investigate how warming affects a tropical community of three species of Drosophila hosts interacting with two species of parasitoids over a single generation. Our experimental design allowed us to distinguish between the direct effects of temperature on host species performance, and indirect effects through altered biotic interactions (competition among hosts and parasitism by parasitoid wasps). Although experimental warming significantly decreased parasitism for all host-parasitoid pairs, the effects of parasitism and competition on host communities did not vary across temperatures. Instead, effects on host relative abundances were species-specific, with one host species dominating the community at warmer temperatures, independently of parasitism and competition treatments. Our results show that temperature shaped a Drosophila host community directly through differences in species’ thermal performance, and not via its influences on biotic interactions.

melanogaster. This host species is not present naturally at the field locations where hosts and parasitoids 120 originated, and was not used in the experiment, thus avoiding bias of host preferences. Single parasitoid 121 isofemale lines were used. 122

Experimental design 123
To disentangle the effects of warming temperatures on host species and their interactions, we 124 manipulated the presence of parasitoids and interspecific competition between host species in a fully 125 factorial design (Figure 1) at ambient and elevated temperatures. As the focus of the experiment was to 126 compare the direct and indirect effects of warming temperatures on host communities, competitive 127 interactions between parasitoids were not assessed nor manipulated, but potentially present in all treatments 128 with parasitoids. Parasitoid preferences were not quantified, but the two parasitoid species used were able 129 to parasitize all three hosts species during trials. Transparent plastic boxes (47cm x 30cm x 27.5cm) with 130 three ventilation holes (15 cm in diameter) covered with insect-proof nylon mesh served as the experimental 131 units. Each box contained three 90 mm high and 28 mm diameter glass vials containing 2.5 mL of 132 Drosophila food medium. Interactions were manipulated by establishing single-species (Figures 1a and 1c) 133 or mixed-species (Figures 1b and 1d) vials, and by including (Figures 1c and 1d) or excluding (Figures 1a  134 and 1b) parasitoids. A total of 60 three-day-old virgin adult hosts, with 1:1 sex ratio, were placed in each 135 vial for 48 hours to allow mating and oviposition (i.e., a total of 180 adults per box). In the multi-host 136 treatment, the 60 hosts were split evenly across the three species (i.e., 20 adults for each species). The 137 density of adult hosts was selected based on preliminary observations to achieve a high level of resource 138 competition (i.e., the density at which strong intraspecific competition was observed for all host species; 139 Supplementary Table S1) while keeping the number of adults for each of the three host species and the total 140 number of adult hosts consistent across treatments and species. The treatment allowed competition both at 141 the adult stage for oviposition space, and at the larval stage of their offspring for food resources [50,51], 142 but we did not aim to identify which was the primary source of competition. All results relate to the host 143 offspring (their abundances and frequencies). For treatments that included parasitoids (Figures 1c and 1d), ten parasitoids (3-7 days old, 1:1 sex ratio) from each species (n = 2, i.e. 20 parasitoids per box), 145 corresponding to 9 % of the total number of adult hosts, were placed in a box for 72 hours, creating high 146 but realistic parasitoid pressure (within the range of parasitism rate observed in this system in nature:  42% [47]). We aimed to study the impact and interaction of parasitism and host competition when both are 148 strong, but realistic, to detect any effect. Vials were removed from the boxes simultaneously with the 149 parasitoids and individually sealed. Each treatment was replicated once across four time-blocks, and each 150 treatment and replicate were therefore represented by three vials. The duration of the experiment 151 corresponded to a single generation of both the hosts and the parasitoids. Each set of treatments was 152 replicated once during a single day, and was repeated over four days (i.e., blocks). The experimental temperatures were chosen to simulate current mean yearly temperature at the two 162 study sites [47]: 23.2 ± 0.4°C (65.9 ± 2.8% humidity), and projected temperatures representing a plausible 163 future scenario under climate change: 26.7 ± 1.0°C (65.1 ± 2.8% humidity) [52]. The simulated difference 164 was therefore 3.5°C. Vials were placed at their corresponding temperature treatment from the first day the 165 adult hosts were introduced for mating and oviposition to the last emergence (up to 40 days). 166 To calculate parasitism rates for each host-parasitoid species pair, pupae from the three vials of 167 each box were randomly sampled 12 days after the initiation of the experiment. All sampled pupae were 168 transferred into one or two 96-well PCR plates (on average 169 ± 30 s.d. pupae sampled per box) and kept 169 at their corresponding temperature treatment until adult insects emerged (up to 40 days for the slowest-170 developing parasitoid species). Sampled pupae were identified to their corresponding host species, and the 171 outcome was recorded as either a host, a parasitoid, an empty pupal case, or an unhatched pupa. We assumed 172 that any pupae which were empty at the time of sampling resulted in adult hosts because this period was 173 too short for parasitoids to complete development and emerge. We calculated parasitism rates from the 174 pupae sampled in plates only. Parasitism rates were calculated as the proportion of each parasitoid species 175 that emerged from the total number of sampled pupae of each host species. Attack rates were not calculated 176 because the exact initial numbers of host larvae available for the parasitoids were unknown. 177 All hosts that emerged (from both vials and sampling plates) were used to quantify the following 178 aspects of host community structure: abundances of each host species, and their frequencies (i.e., the 179 fraction of all host individuals belonging to each host species). All hosts and parasitoids that emerged from 180 vials before and after subsampling for parasitism rates were collected, identified, and stored in 95% ethanol 181 until the second generation started emerging from the vials (i.e., hosts were no longer collected after four 182 days without any emergence). Effect of treatments on mean host body mass were also investigated, as an 183 increase in temperature generally produces smaller individuals, which could influence the outcome of 184 competition [27]. Individual dry body mass of hosts was measured with 1 μg accuracy using a Sartorius 185 Cubis ™ micro-balance. Only fully-eclosed and intact individuals were included in measurements. 186

Statistical analysis 187
All replicates with fewer than ten total emergences or pupae were removed from analyses of host 188 abundances, frequencies, and parasitism rates (Supplementary Table S2), as these outcomes were associated 189 with low success during the mating process and not with experimental treatments (results with the whole 190 dataset can be found in Supplementary Table S3 online). Data were analyzed with generalized linear models 191 (GLMs). After testing for overdispersion of the residuals, abundance data were modeled using a negative 192 binomial error distribution, host body mass using a gaussian error distribution, and frequencies of host species and parasitism rates using a quasibinomial error distribution. Parasitism (two levels), type of 194 competition (two levels), host species (three levels), parasitoid species (two levels), and temperature (two 195 levels) were included as categorical predictor variables within each model. Blocks were included in the 196 models as a fixed effect because of the small number of blocks. Each two-way interaction was tested and 197 kept in our models if judged to be significant on the basis of backward selection using ANOVA likelihood-198 ratio tests. Interaction between temperature and parasitism, temperature and competition, and parasitism 199 and competition were systematically kept in our models as the experiment was designed to test for the 200 significance of these interactions. The three-way interaction between temperature, parasitism, and 201 competition was tested for host abundances, host frequencies, and host body mass, but was not significant. 202 Significance of the effects was tested using type III analysis of deviance with F-tests. Factor levels were 203 compared using Tukey's HSD post hoc comparisons of all means, and the emmeans package. Model 204 assumptions were verified with the DHARMa package. All analyses were performed using R 3.5.2 [53] with 205 the packages stats, MASS[54], car [55], performance [56], DHARMa [57], and emmeans [58]. 206

207
In total, 7627 individuals (7063 hosts and 564 parasitoids) were reared across all treatments and 208 replicates (238.3 ± 13.3 s.d. on average per box). Across all treatments and replicates, 2717 pupae were 209 sampled in total for estimating parasitism rate, of which 2227 (82%) produced an adult host or parasitoid. 210 Mean host abundances, host body mass, and parasitism rates are presented for each treatment in 211 Supplementary Table S4 online. We focused on the effects of temperature, parasitism, competition and 212 their interactions on host abundances, host proportions, and host body mass (Table 1).  (intraspecific or  214 interspecific), host species (n = 3), parasitoid species (n = 2), interactions between terms, and block (n = 4) on host abundances, host frequencies, 215 host body mass, and parasitism rate. Degrees or freedom (Df) for each F-ratio are given for each factor and for the error. F values are presented with 216 the significance of the effect: (***) P < 0.001, (**) P < 0.01, (*) P < 0.05, (.) P < 0.1, (ns) P > 0. 05

Direct effect of warming on the host community
The effect of temperature on host relative abundances varied significantly across host species (Table 1 Figure 3b). However, temperature did not interact significantly with either parasitism or competition in affecting any of our measures of community structure (P > 0.05, Table 1), suggesting that the relative effects of parasitism and competition as structuring agents were not modified by experimental warming.

Discussion
Our experiment revealed that experimental warming directly affected host community structure through differences in thermal performance among species, and decreased parasitism rates. However, warming did not impact the effect of parasitism on host community structure over the timescale investigated.
Our results suggests that ongoing rises in global temperatures could directly alter host community structure through differences in thermal performance across species, as has been shown for communities of fish [59], plants [60], and insects [61]. Changes in host frequencies in warmer temperatures was primarily due to a dramatic increase in the relative abundance of a single host species, D. pseudoananassae, the species with the largest thermal performance breath [48]. This increase occurred across all combinations of parasitism and competition treatments, and without a change in Drosophila body mass, suggesting a direct effect of temperature on host fecundity due to the preferred temperature of the adults for egg-laying and/or offspring egg-to-adult viability related to their optimal temperature [62]. Surprisingly, the dominant species at warmer temperature, D. pseudoananassae, was not the one with the highest thermal performance optimum measured by MacLean et al. [48]. However, in our system, its distribution is limited to low elevation sites, and this species has a higher thermal tolerance than either of the other two species considered [48]. In nature, Drosophila species distributions are driven by differences in innate thermal tolerance limits, with low phenotypic plasticity for thermal tolerance limits in both widespread and tropical species [49].
This suggests that warming temperatures, in the context of global climate change, will have a strong effect on community composition through direct effect on fitness.
Our data also revealed a significant decrease in parasitism rates with warming. Reviews suggest that parasitism would decrease under global warming scenarios due to an increase in parasitoid mortality, and host-parasitoid spatial and temporal asynchrony [13,43]. However, presence of parasitoids significantly decreased abundances of the three host species independently of the temperature regime, suggesting that warming treatments did not decrease attack rate, but decreased parasitoid virulence. This experiment was performed over a single generation, so long-term consequences of decreased parasitism rates with elevated temperatures for host-parasitoid dynamic cannot be assessed, but a decrease in parasitism rates could lead to the release of hosts from top-down control. However, in the case of a simple linear tritrophic interaction, the results of Flores-Mejia et al. [63] suggest that parasitoid top-down control might be less sensitive to temperature than previously thought. Moreover, host immune responses might also be sensitive to temperature [64], increasing or decreasing host vulnerability to parasitoid attacks. Therefore, host immune function response to temperature should be considered alongside host thermal performance and tolerance to predict the effects of warming temperature on host communities [13].
Our results demonstrate that differences in thermal performance across host species may be a stronger determinant of how host communities respond to warming temperatures than shifts in the strength of biotic interactions. We used high, but realistic levels of competition and parasitism that would have allowed us to detect their effects on host species relative abundances if they were any. Aspects of our results contrast with those from a field transplant experiment on two species drawn from the same Australian Drosophila-parasitoid community [65]. Investigating fitness of D. birchii and D. bunnanda along an elevation gradient, the authors found an interacting effect between the abiotic environment and interspecific competition. However, the field experiment excluded parasitoids, and the elevational gradient studied is likely to include variations such as humidity as well as temperature, which might influence the outcome [66].
Our study serves as example of the mechanisms that can be expected to drive community responses to global warming, but general conclusions on the potential impact of warming temperature on hostparasitoid networks will require replication with different species compositions and different systems.
Especially, most host-parasitoid systems are tri-trophic (plants-arthropods-parasitoids), and climate warming is likely to impact host-parasitoid networks through bottom-up effects [67]. Few such experiments have been undertaken, despite the need to better disentangle direct and indirect effects of warming temperature on species communities. Ideally, future studies will also need to investigate the longer-term dynamics of such systems. Moreover, as temperatures continue to increase, species from diverse taxa are shifting their distribution worldwide to higher latitudes and elevations [68], changing their biotic environment with novel species interactions and different community assemblages [69]. Dispersal was not permitted in this study, but is likely to mediate some of the effects of warming temperature on species and their interactions [29,70].
Understanding the mechanisms driving community responses to warming scenarios is particularly important for tropical communities, which face more severe impacts of climate warming than temperate communities. Here, we demonstrate that warming had a direct effect on our focal tropical Drosophila host community through differences in thermal performance, without affecting the relative strength of parasitism and competition. The role of parasitoids as essential top-down control agents of insect populations was not reduced under experimental warming, but parasitism rate decreased, suggesting that an indirect effect of warming temperature on the structure of host community through the effect of parasitism could be observed after several generations.