Impact of heat stress on the fitness outcomes of symbiotic infection in aphids: a meta-analysis

Abstract

Beneficial microorganisms shape the evolutionary trajectories of their hosts, facilitating or constraining the colonization of new ecological niches. One convincing example entails the responses of insect–microbe associations to rising temperatures. Indeed, insect resilience to stressful high temperatures depends on the genetic identity of the obligate symbiont and the presence of heat-protective facultative symbionts. As extensively studied organisms, aphids and their endosymbiotic bacteria represent valuable models to address eco-evolutionary questions about the thermal ecology of insect–microbe partnerships, with broad relevance to various biological systems and insect models. This meta-analysis aims to quantify the context-dependent impacts of symbionts on host phenotype in benign or stressful heat conditions, across fitness traits, types of heat stress and symbiont species. We found that warming lowered the benefits (resistance to parasitoids) and costs (development, fecundity) of infection by facultative symbionts, which was overall mostly beneficial to the hosts under short-term heat stress (heat shock) rather than extended warming. Heat-tolerant genotypes of the obligate symbiont Buchnera aphidicola and some facultative symbionts (Rickettsia sp., Serratia symbiotica) improved or maintained aphid fitness under heat stress. We discuss the implications of these findings for the general understanding of the cost–benefit balance of insect–microbe associations across multiple traits and their eco-evolutionary dynamics faced with climate change.

1. Introduction

The ecology and evolution of most insect species are now understood through the lens of their mutualistic associations with diverse microorganisms (e.g. bacteria, fungi, viruses) [1,2]. These beneficial microbes shape the life histories and the evolutionary trajectories of their hosts, providing opportunities as well as imposing constraints for colonization of new ecological niches, depending on the nature of the association and the pace of environmental change [3]. On the one hand, acquisitions of obligate microbial partners involved in nutritional complementation are envisioned as key evolutionary innovations that enabled several insect taxa to specialize on unbalanced diets like plant sap and vertebrate blood [4,5]. However, these obligate symbionts underwent severe genome degradation (in terms of size, function and structural dynamic) as a product of their ancient coevolution history with their hosts. Such genetic features limit their potential to adapt to rapidly changing environmental conditions and the one of their hosts since they are required for insect successful growth and reproduction [6–8]. On the other hand, insects harbour facultative symbionts responsible for the expression of ecological traits such as plant exploitation and defence against parasitic enemies [9–11]. These symbiont-mediated traits might be adaptive under specific ecological contexts like those involving harsh biotic or abiotic stress, but infection by facultative symbionts is also known for inducing fitness costs under more benign environments. Hence, the overall impacts of facultative symbionts on the fitness and adaptive capacity of their hosts can be viewed as a cost–benefit balance tipping in a costly or beneficial state in relation with ecological contingency [12–15].

Temperature has been rapidly identified as a key environmental parameter determining the net fitness consequences of carrying a particular symbiont genotype (for obligate and facultative symbionts) or species (for facultative symbionts) (for reviews, see [16–19]). For obligate mutualisms, the insect resilience to stressful high temperatures can be curtailed by a single mutation affecting a gene encoding the production of heat-protective molecules by the obligate symbiont, as evidenced in aphids and their nutritional bacterial partner Buchnera aphidicola [20,21]. For facultative mutualisms, this resilience can be improved by the presence of non-essential symbionts bestowing physiological tolerance to heat [22,23]. Such symbiont-mediated changes in insect thermal ecology can involve the release of stress-lowering metabolites by the microbial player (e.g. Serratia symbiotica in aphids [24]). They may also rely on adjustments of host metabolic, physiological or behavioural profiles occurring with symbiotic infection. Convincing examples are given by Wolbachia-related modifications of dopamine metabolism and behavioural selection of microclimatic conditions in Drosophila melanogaster [25,26], or the induction of host genes involved in stress response triggered by the symbiont Rickettsia sp. in the whitefly Bemisia tabaci [22]. Conversely, temperature can mediate the expression of ecological traits displayed by symbiont-infected individuals, thereby affecting the adaptive value of hosting these facultative symbionts under some environmental contexts such as high parasitism or pathogenic pressure [27–29]. For example, temperature can alter the levels of protection granted by defensive symbionts towards parasitoids (e.g. Hamiltonella defensa in aphids and Spiroplasma sp. in drosophilids [27,28,30]) or viruses (e.g. Wolbachia sp. in drosophilids [31]). Similarly, symbiont-dependent detoxification of harmful compounds mediating host resistance to insecticides can be thermally sensitive (e.g. Wolbachia sp. in planthoppers [32]). Finally, thermal conditions determine the outcomes of interactions between insect hosts and facultative symbionts acting as reproductive manipulators, notably affecting the incidence of male killing and cytoplasmic incompatibility (e.g. Wolbachia sp. in drosophilids and mosquitoes; Cardinium hertiigi in parasitic wasps [33–35]).

Because they affect host performance as a function of temperature and the microbes themselves are sensitive to thermal variability, symbiotic associations can vary both temporally and spatially in prevalence, occurrence or strain type in many species such as honeybees, whiteflies, damselflies or aphids [15,36,37]. These different lines of evidence raise a concern about the short- and long-term maintenance of symbiotic associations enduring changing thermal conditions, such as increased mean temperature and frequency of heat events forecast under climate change [38]. Several putative evolutionary scenarios have been proposed to anticipate the fate of insect–microbe mutualisms in this climatic context, depending on the potential for symbiotic relationships to maintain sustainable insect performance and evolutionary capacity under heat stress [6,18,39]. As empirical evidence has accumulated over the last two decades, it seems now valuable to draw first general patterns about the thermal sensitivity of insect–microbe partnerships and their implications for understanding the ecological and evolutionary trajectories of these mutualistic associations under a changing climate.

Here, we propose a meta-analysis intended to quantify the impacts of microbial symbionts on the phenotype of their insect hosts under different thermal settings (control permissive versus stressful heat conditions). Our goal is to address and shed light on long-debated ecological and evolutionary questions about symbiont-mediated tolerance to heat in insects. We test for the broadening of the assumption that microbial players (obligate and facultative symbionts) should predominantly modulate the insect ability to withstand heat stress, or conversely, that temperature should act as an important driver of the expression of traits mediated by endosymbionts. Specifically, we aim at examining the generality and quantifying the intensity of the thermal modulation of symbiont-related alterations of insect phenotype among various biological and ecological contexts. These contexts include (i) life-history traits and fitness components (whether all insect traits are equally impacted by facultative symbionts and heat), (ii) the nature of the heat stress (whether protection granted by facultative symbionts is efficient under short- and long-term experimental heat stress), and (iii) symbiont species involved (whether obligate and facultative symbionts are heat protective). To fill these objectives, we focused our literature survey on aphid–bacterium associations as extensively studied models providing a unique dataset on insect–microbe thermal ecology, with a diversity of traits measured, temperature treatments applied and microbial species identified. Nevertheless, this analysis will allow one to infer general patterns of thermal sensitivity in insect–symbiont systems, with ecological and evolutionary scopes that go beyond the aphid study model.

2. Material and methods

(a) . Literature search

We collected data within a framework that allowed us to evaluate the interplay between the temperature treatment and the host symbiotic state on the set of aphid traits mentioned below. We used the following keywords in Google Scholar: (‘aphid*’ AND (symbiont* OR facultative OR secondary OR obligat*) AND ‘temperature*’ AND (trait* OR fitness OR development OR growth OR fecundity OR morpholog* OR longevity OR survival OR defens*)), bounded between 1991 (first experimental report of the impacts of temperature on an aphid-bacterium mutualism [40]) and 2020. Our main inclusion criteria were: (i) data with measures of at least one of the focus traits, reporting mean, variance and sample size; (ii) data with at least two temperature treatments, including a control (unheated conditions); and (iii) data with at least two aphid symbiotic states, including a control (either an individual hosting a heat-sensitive genotype of the obligate symbiont or an individual not infected by a given facultative symbiont).

Following our inclusion criteria, we retrieved data from a total of 18 relevant articles, published between 2000 and 2020 (list included as an electronic supplementary material file). Pairs of datapoints were extracted manually from (i) text and tables, (ii) raw data made available by the authors of the studies or (iii) figures by using WebPlotDigitizer [41]. One pair of datapoints corresponded, for a given temperature treatment (either control permissive or stressful heat conditions), to the mean of the trait value (with variation and sample size) for both control and infected individuals. This resulted in a total of 410 pairs of datapoints. Our dataset included a total of three aphid species developing on Fabaceae and belonging to two genera: Acyrthosiphon pisum (318 pairs of datapoints from 14 studies), Aphis fabae (76 pairs of datapoints from 3 studies) and Aphis craccivora (16 pairs of datapoints from 1 study).

For comparisons among fitness traits and types of heat stress (see the two following subsections), we excluded data related to the obligate symbiont (20 pairs of datapoints from 3 studies) and only retained those pertaining to the facultative symbionts (390 pairs of datapoints from 15 studies) to restrict the scope of the analyses to a biologically meaningful dataset. Indeed, obligate and facultative symbionts differ in degree of intimacy and coevolution history with their hosts, as well as regarding the source of variation in symbiotic state as defined in this meta-analysis (genetic variation versus absence/presence). They also differ in the underlying symbiont-related processes that allow a change in host physiology under heat stress. For these reasons, it was not relevant to directly compare or regroup the responses of obligate and facultative bacteria.

(b) . Data organization: fitness traits

We focused on aphid phenotypic traits classified into five categories: (i) defence against parasitic wasps (proportion of hosts surviving after exposure to parasitoids, parasitism rates, or emergence rates of the parasitoid) (151 pairs of datapoints from 7 studies), (ii) development (age to adulthood or duration of reproductive period) (54 pairs of datapoints from 5 studies), (iii) fecundity (total or daily number of offspring) (124 pairs of datapoints from 11 studies), (iv) morphology (body mass) (14 pairs of datapoints from 2 studies) and (v) survival (overall lifespan for long-term heat stress, number of days survived or survival rates after treatment for short-term heat stress) (47 pairs of datapoints from 4 studies). Contrary to other classical life-history traits, defence can be seen as an ‘ecological trait’, which implies that measures on the effect of temperature were done in the presence of the ecological conditions that the symbiont is purported to protect hosts from (i.e. presence of parasitoids).

(c) . Data organization: temperature treatments and types of heat stress

We established a distinction between control temperature treatment (encompassing permissive temperatures, 179 pairs of datapoints) and heat treatment (extending to stressful high temperatures, 211 pairs of datapoints). The distinction was based on what the authors of each study described in their methods and the thermal biology of studied aphid species, which usually endure a significant decline of performance at temperatures ≥ 25°C [19]. We then classified the heat treatments based on the duration of exposure to stressful high temperatures, either long term (i.e. for at least several days, extending over one or several insect developmental stages, with no return to favourable conditions before measurements of fitness traits) (275 pairs of datapoints from 7 studies) or short term (i.e. for a few hours with return to favourable conditions before measurements of fitness traits) (115 pairs of datapoints from 8 studies).

(d) . Data organization: symbiotic states and symbiont species

We included studies focusing on different haplotypes of the obligate bacterial endosymbiont B. aphidicola (20 pairs of datapoints from 3 studies) representing a form of ‘symbiotic plasticity’ for the insect (presence/absence of an unfavourable natural genetic mutation limiting symbiont transcriptional response to heat). To parallel the distinction made for facultative symbionts, aphids likely to be vulnerable to heat stress due to potentially limited symbiont protection (i.e. infected with the mutated heat-sensitive genotype of B. aphidicola) were considered as negative control, while those potentially benefiting from this protection (i.e. bearing the obligate symbiont genotype without the mutation) were labelled as treatment.

There was a total of five facultative symbiont species identified in our dataset: Fukatsuia symbiotica (45 pairs of datapoints from 6 studies), H. defensa (214 pairs of datapoints from 10 studies), Regiella insecticola (2 pairs of datapoints from 1 study), Rickettsia sp. (58 pairs of datapoints from 2 studies) and S. symbiotica (71 pairs of datapoints from 4 studies). In the case of single infection, host lines deprived of the facultative symbiont were considered as negative control and infected hosts were considered as treatment. In case of double infection (F. symbiotica/H. defensa or Rickettsia sp./S. symbiotica), individuals infected by only one of the two symbiont species were considered as negative control and compared with individuals co-infected with the two symbionts, thus allowing to specifically dissociate the effects of each microbe involved in the tripartite system. We could not consider the influence of genetic variation between strains of facultative symbionts because of an insufficient number of independent studies investigating these effects. In addition, unlike B. aphidicola, strains of a given facultative symbiont usually differed from one study to another.

(e) . Statistical analyses

The meta-analysis and meta-regression were conducted in R v. 4.0.1 and were performed using the metafor package [42] following the workflow proposed by Crystal-Ornelas [43]. We used standardized mean differences (SMDs, Hedges's g method) as effect sizes to compare symbiotic treatments (aphids infected with a heat tolerant genotype of obligate symbiont or a particular species of facultative symbiont) with control groups (infection with a heat-sensitive genotype of obligate symbiont or absence of facultative symbiont), under either permissive temperatures or heat stress [43].

To ensure that a negative effect size represents a net fitness cost for the host, we multiplied the effect sizes calculated for the developmental traits by −1, because slowly developing phytophagous insects might suffer from increased mortality risks due to an extended window of susceptibility to natural enemies (‘slow-growth-high-mortality hypothesis' [44]). We used the same data transformation for defence traits, when what was being measured was the benefit for the parasitoid at the expense of the aphid (e.g. parasitism rates and parasitoid emergence rates).

We ran multivariate linear mixed-effects models, using the identity of each independent study as a random effect to account for the nonindependence of observations coming from a same study (i.e. between-study random effect). We used the restricted maximum-likelihood (REML) estimator for the amount of heterogeneity to fit the random effect model [42,45,46]. We performed five distinct meta-regression analyses. The first analysis only included the temperature treatment (two levels: permissive temperatures or heat stress) as a moderator to determine if the overall fitness impacts of facultative symbionts on their hosts could be modified by the thermal context (n = 15 studies). All the other analyses integrated an additional moderator in interaction with the temperature treatment. In the second analysis, we integrated the fitness traits with five levels: defence, development, fecundity, morphology, or survival (n = 15 studies). In the third analysis, we integrated the type of heat stress involving two levels: long term or short term (n = 15 studies). In the fourth analysis, we integrated the symbiont species studied involving five levels: B. aphidicola, F. symbiotica, H. defensa, Rickettsia sp., or S. symbiotica (n = 18 studies). We removed the datapoints for R. insecticola from this analysis because of the limited number of observations available and the lack of replication (one single study involved). In a fifth analysis, we integrated the number of facultative symbionts hosted involving three levels: none, one or two (n = 18 studies) (electronic supplementary material, figure S1).

For each analysis, we tested the linear hypothesis that the estimated Hedges's g values for the effect of the symbiotic state differed from zero (package multcomp [47]). Each analysis was followed by a Tukey's contrasts post hoc test with Bonferroni adjustment method for p-value (package emmeans [48]) to compare values of effect size for permissive temperatures and heat stress, for each level of the moderator of interest. These post hoc analyses were done on a composite variable merging the levels of the two moderators included in each model, to avoid any issue regarding interactions when running the Tukey's test. The histogram of the frequency of effect sizes and the funnel plot analysis (electronic supplementary material, figure S2) showed relatively symmetrical data, although slightly skewed towards positive effect size values.

3. Results

(a) . Global effect of temperature on the fitness outcomes of symbiotic infection

When grouping all traits, types of heat stress and facultative symbiont species, there was a clear global thermal modulation of the fitness consequences of hosting a facultative symbiont. This interaction is indicated by a significant difference in effect sizes (quantifying the fitness impacts of symbiotic infection) between temperature treatments: warming overall increased the adaptive value (fitness gains) of symbiotic infection for the insect (z = 3.75, p < 0.001). This implies that hosting a facultative symbiont was more beneficial under heat stress than under permissive temperatures. Indeed, symbiotic infection had a neutral effect on aphid overall fitness under permissive temperatures (Hedges's g = −0.06, χ² = 0.11, p = 0.73), while it became significantly beneficial under heat stress (Hedges's g = 0.65, χ² = 11.7, p < 0.001) (figure 1).

Figure 1.

Global effect of temperature treatment on the fitness consequences of symbiotic infection (all traits, types of heat stress, and facultative symbiont species combined). Predicted overall Hedges's g values (standardized mean deviation, large circles) and 95% confidence intervals are shown, for control permissive temperatures (T−, blue) and heat stress (T+, red). A positive effect size value indicates that hosting a given facultative symbiont is favourable for the host, and a negative value indicates that it is costly. Confidence intervals that overlap zero indicate a non-significant effect of symbiotic infection on insect fitness. Note that only 60% of the datapoints are shown in this figure for aesthetic purposes. n = 15 studies, 390 datapoints. Asterisks highlight significant difference in effect sizes between temperature treatments (***p < 0.001). (Online version in colour.)

(b) . Fitness traits

The impacts of temperature on the fitness outcomes of symbiotic infection were trait-specific, with three different patterns of response. For defence against parasitoids, warming lowered the fitness benefits (protection) provided by some facultative symbionts (z = −9.85, p < 0.001). Still, in a parasitism context, a high fitness advantage remained of carrying the studied protective symbionts (H. defensa and F. symbiotica) under both permissive temperatures (Hedges's g = 3.20, χ² = 78.0, p < 0.001) and heat stress (Hedges's g = 2.13, χ² = 45.3, p < 0.001). This suggests that the symbiont-mediated protection towards parasitoids persisted under different temperature treatments. For development and fecundity, warming shifted the fitness consequences of symbiotic infection from a costly to a neutral outcome (development: z = 9.93, p < 0.001; fecundity: z = 16.3, p < 0.001). Under permissive temperatures, aphids infected by facultative symbionts incurred a prolonged development time (Hedges's g = −1.31, χ² = 15.9, p < 0.001) and a reduced fecundity (Hedges's g = −1.56, χ² = 22.7, p < 0.001) relative to uninfected individuals. These infection costs were no longer apparent under heat stress (development: Hedges's g = −0.24, χ² = 0.53, p = 0.47; fecundity: Hedges's g = −0.22, χ² = 0.48, p = 0.49). For morphology (body size) and survival, temperature did not significantly modulate the fitness consequences of symbiotic infection (morphology: z = 1.54, p = 1.00; survival: z = 3.10, p = 0.09). Aphids infected by facultative symbionts grew lighter (Hedges's g = −1.17, χ² = 9.41, p < 0.01 and Hedges's g = −0.85, χ² = 5.35, p < 0.05, for permissive and heat stress temperatures, respectively) and died faster (Hedges's g = −1.25, χ² = 14.3, p < 0.001 and Hedges's g = −0.90, χ² = 7.16, p < 0.01, for permissive and heat stress temperatures, respectively) compared with uninfected individuals (figure 2).

Figure 2.

Effect of temperature treatment on the fitness consequences of symbiotic infection for different aphid traits (all types of heat stress and facultative symbiont species combined). Predicted overall Hedges's g values (standardized mean deviation, large circles) and 95% confidence intervals are shown, for control permissive temperatures (blue) and heat stress (red). Note that only 75% of the datapoints are shown in this figure for aesthetic purposes. n = 7, 5, 11, 2 and 4 studies, for defence, development, fecundity, morphology and survival, respectively. Asterisks highlight significant difference in effect sizes between temperature treatments for a given trait, according to Tukey's post hoc comparisons (***p < 0.001, n.s. = not significant p > 0.05). (Online version in colour.)

(c) . Types of heat stress

The impacts of temperature on the fitness consequences of symbiotic infection varied according to the type of heat stress applied. Long-term heat treatment did not modulate these outcomes (z = 2.02, p = 0.26), while short-term heat treatment did (z = 8.14, p < 0.001). Under permissive temperatures, symbiotic infection had neutral effects on aphid overall fitness indicators in both long-term and short-term heat stress experimental procedures (long term: Hedges's g = 0.05, χ² = 0.04, p = 0.84; short term: Hedges's g = −0.24, χ² = 0.89, p = 0.34). Under heat stress, aphids infected by facultative symbionts benefited from a fitness advantage relative to their uninfected counterparts, irrespective of the duration of exposure to the heat treatment (long term: Hedges's g = 0.65, χ² = 6.42, p < 0.05; short term: Hedges's g = 0.58, χ² = 5.30, p < 0.05). However, the interaction mentioned above indicates that warming significantly increased the adaptive value of symbiotic infection only if applied through short-term heat stress (sporadic bout of heat shock) (figure 3).

Figure 3.

Effect of temperature treatment on the fitness consequences of symbiotic infection under different types of heat stress (all traits and facultative symbiont species combined). Predicted overall Hedges's g values (standardized mean deviation, large circles) and 95% confidence intervals are shown, for control permissive temperatures (blue) and heat stress (red). Note that only 63% of the datapoints are shown in this figure for aesthetic purposes. n = 7 and 8 studies, for long-term and short-term heat stress treatments, respectively. Asterisks highlight significant difference in effect sizes between temperature treatments for a given type of heat stress, according to Tukey's post hoc comparisons (***p < 0.001, n.s. = not significant p > 0.05). (Online version in colour.)

(d) . Symbiont species

The impacts of temperature on the fitness consequences of symbiotic infection were dependent on the microbial species involved. For the obligate symbiont, warming significantly enhanced the adaptive value of hosting the heat tolerant genotype of B. aphidicola (z = 3.73, p < 0.01). Insects infected with a heat tolerant genotype of B. aphidicola performed as well as those infected with a heat-sensitive genotype under permissive temperatures (Hedges's g = −0.31, χ² = 0.28, p = 0.59), and greatly better under heat stress (Hedges's g = 2.17, χ² = 12.7, p < 0.001).

For the facultative symbionts, warming significantly enhanced the adaptive value of hosting the bacteria Rickettsia sp. (z = 9.88, p < 0.001) and S. symbiotica (z = 15.7, p < 0.001). Harbouring Rickettsia sp. was slightly detrimental to aphids under permissive temperatures, since the CIs of effect size marginally overlap zero (Hedges's g = −0.76, χ² = 2.81, p = 0.09), while no fitness consequences related with this infection were seen under heat stress (Hedges's g = 0.17, χ² = 0.13, p = 0.72). Symbiotic association with S. symbiotica switched from a slightly detrimental to a slightly ameliorative impact on aphid fitness from permissive temperatures (Hedges's g = −0.69, χ² = 2.34, p = 0.13) to heat stress (Hedges's g = 0.65, χ² = 2.07, p = 0.15), respectively, although effect sizes and CIs only indicated trends. Warming did not affect the fitness outcomes of infection with F. symbiotica (z = 0.42, p = 1.00) and H. defensa (z = −3.19, p = 0.06). Hosting F. symbiotica remained neutral in terms of aphid fitness under both permissive temperatures (Hedges's g = −0.21, χ² = 0.50, p = 0.48) and heat stress (Hedges's g = −0.14, χ² = 0.22, p = 0.64). Individuals infected with H. defensa exhibited a net fitness advantage relative to their uninfected counterparts under both permissive temperatures (Hedges's g = 0.97, χ² = 12.0, p < 0.001) and heat stress (Hedges's g = 0.68, χ² = 6.10, p < 0.05) (figure 4).

Figure 4.

Effect of temperature treatment on the fitness consequences of symbiotic infection involving different symbiont species, engaged in either obligate or facultative relationship with their aphid hosts (all traits and types of heat stress combined). Predicted overall Hedges's g values (standardized mean deviation, large circles) and 95% confidence intervals are shown, for control permissive temperatures (blue) and heat stress (red). Obligate and facultative symbionts (separated here by a vertical solid line) differ in the processes involved in modulation of host physiology under heat stress. Note that only 74% of the datapoints are shown in this figure for aesthetic purposes. n = 3, 6, 10, 2 and 4 studies, for Buchnera aphidicola, Fukatsuia symbiotica, Hamiltonella defensa, Rickettsia sp. and Serratia symbiotica, respectively. Asterisks highlight significant difference in effect sizes between temperature treatments for a given symbiont species, according to Tukey's post hoc comparisons (***p < 0.001, **p < 0.01, n.s. = not significant p > 0.05). (Online version in colour.)

4. Discussion

This meta-analysis sought to quantify the fitness impacts (phenotypic effects) of symbiotic infection on aphid hosts under permissive temperatures and heat stress, with broader prospects about the generality of symbiont-mediated heat protection across biological and ecological contexts. We found that infection by facultative symbionts benefited the aphids in terms of overall fitness under heat stress or, conversely, that warming increased the general adaptive value of such infection for the insects. Warming lowered the benefits provided by some facultative symbionts (defence against parasitoids), but limited the costs of symbiotic infection on insect development and fecundity. Infection by facultative symbionts was mostly beneficial to the aphids in case of short-term exposure to high temperature spikes rather than prolonged warming. Heat tolerant genotypes of the obligate symbiont B. aphidicola on the one hand, and the presence of the facultative symbionts Rickettsia sp. and S. symbiotica on the other, specifically contributed to improve (or at least maintain) aphid fitness under heat stress. These findings highlight subtle trait- and environment-specific responses underlying complex ecological and evolutionary consequences for insect–microbe associations in a changing climate.

Several mechanistic explanations can be proposed to account for the trait-specific conclusions about the interactive effects of temperature and symbiotic infection on host fitness. First, variability could arise from energetic trade-offs shaping resource allocation between traits (e.g. defence and fecundity), particularly under physiologically constraining conditions such as infection by facultative symbionts and stress-inducing temperatures [14,49]. Second, life-history traits frequently differ in their thermal requirements and response to warming [50,51]. Heat-sensitive traits are thus likely to be the most responsive to temperature rise and its potential interaction with symbiotic infection. Third, trait expression can be modulated by different symbiotic processes in a temperature-dependent fashion. On the one hand, the direct involvement of protective facultative endosymbionts (especially toxin-producing H. defensa) in aphid defence against parasitoids should explain the considerable fitness gains provided by symbiotic infection and their decline under heat stress, since protective symbiosis is weakened at high temperatures [27–29]. However, we argue that the widespread idea that protection against parasitoids fails under heat stress could be largely overstated, as it remained overall beneficial to the insect even under higher temperature. On the other hand, some aphid vital functions (especially development and reproduction) are closely linked to titres of the obligate symbiont B. aphidicola [23,52,53]. Our study confirms the costly nature of symbiotic infection for several aphid fitness traits under permissive temperatures, thereby corroborating the results of a recent analysis [14]. However, under heat stress, symbiotic infection turned neutral for aphid development and fecundity, which may be explained by the protective role played by some facultative symbionts (F. symbiotica, S. symbiotica) in shielding populations of B. aphidicola from heat-related depletion [23,54]. An alternative explanation could be that heat stress triggers greater physiological costs relative to infection by facultative symbionts, and thereby masks potential fitness differences between infected and uninfected aphids.

Whatever their mechanistic basis, these trait-specific conclusions illustrate the concept of an environmentally contingent cost-benefit balance of symbiotic infection. They highlight the relevance of multi-trait approaches to understand how facultative symbionts can shape their host life histories under different thermal conditions. Although not all roles of symbionts are yet precisely known [10,55], there is great interest in looking beyond simple fitness traits to explore how symbiotic associations may affect ecological functions other than defence against parasitoids, in relation to temperature. For example, how temperature affects symbiont-mediated protection against pathogens, nutrient processing capabilities, population dynamics, interaction with the host plant, detoxification or body coloration are avenues to explore in different models (e.g. [56–58]).

The adaptive value of symbiotic infection was also contingent on the nature of the heat stress applied. Why symbiotic infection turned more advantageous under short-term heat stress can be explained by the underlying mechanisms of symbiont-mediated protection to heat. Specifically, the heat protection delivered by facultative symbionts is effective on the short term as it apparently consists of either maintaining populations of B. aphidicola or facilitating their recovery if aphids readily return to benign temperatures [23,24,54]. It might thus be speculated that long-term heat stress treatments (where insects are held in warm conditions for a prolonged duration without return to cooler environments) could hamper the efficiency of such protective mechanisms. This could occur if these mechanisms operate once temperatures shift from stressful to favourable conditions, or if they are triggered by threshold-like effects, as evidenced in the event of repeated exposure to heat shocks leading B. aphidicola abundances to decline below a value causing aphid sterility [59]. The results suggesting how impactful the temporal dynamic of temperatures is for aphid-symbiont associations should prompt further efforts to probe into the responses of these systems to different patterns of temperature fluctuations (e.g. diurnal and seasonal temperature cycles, (micro)climatic gradients) [16,18,60]. In a climate change context, this may also imply that heat waves rather than progressive warming could be the main driver of the evolution of the association.

When it comes to species-specific conclusions, the important contribution of different genotypes of B. aphidicola to aphid response to heat stress is not a surprising result, considering the high levels of mutual dependence between insect fitness and microbe abundance (as explained above). Moreover, studies compared B. aphidicola genotypes purposely chosen for differing a priori in their thermal sensitivity, because of a naturally recurring mutation that affects the symbiont transcriptional pathways involved in heat response [20,21,61]. Therefore, the prevalence and levels of heat protection mediated by the obligate symbiont could be overestimated relative to more ecologically relevant contexts [23,61]. Because B. aphidicola strains were selected based on their response to heat stress, there is a circular logic that does not allow to draw the same conclusions as for the facultative symbionts. Yet, the results we obtained are still interesting in the context of the evolution of mutualisms in various thermal conditions. The contribution of obligate endosymbionts to the evolvability of their carriers facing rising temperatures is thought to be rather minor across different phyla of sap-feeding insects, because of their genomic features limiting their evolutionary potential. They are thus frequently envisioned as the potential ‘Achilles' heel’ of the interaction in a rapidly changing environment [6,62,63]. However, we suggest that the genetic diversity in B. aphidicola strains shaping host heat tolerance could be an additional lever for natural selection, allowing adaptation to warming. Our results may therefore help temper the Achilles' heel statement, which may be excessive with respect to heat adaptation in such host-symbiont systems. In addition, the heat-resistant B. aphidicola strain associates poorly with facultative symbionts, probably because these non-essential partners are no longer beneficial for hosts in warmer conditions [23,24]. Hence, although natural variability in the obligate symbiont modulating insect heat tolerance can constitute a first evolutionary lever to cope with rapid temperature change, it may play on the persistence and evolution of facultative symbioses over the long term.

Outside of obligate symbiosis, four species of facultative symbionts have been reported to increase aphid heat tolerance: F. symbiotica, H. defensa, Rickettsia sp. and S. symbiotica [55]. From our quantitative study, it appears that only infection with Rickettsia sp. and S. symbiotica yielded different fitness impacts under permissive temperatures and heat stress (although CIs of effect sizes only indicated modest fitness consequences of symbiotic infection). This could alternatively indicate a direct symbiont involvement in heat protection, or a temperature-driven reduction of fitness costs incurred with infection. This does not necessarily imply that our study contradicts the literature, but rather that the net fitness gains of symbiotic infection under heat stress could not be apparent through the lens of a meta-analysis conducted at the species level. Plausible explanations could involve an insufficient set of independent observations from the literature, the influence of mechanisms acting at finer levels of biological entity (i.e. interactions between host and symbiont genotypes) or other environmental factors modulating heat protection (e.g. intensity of parasitism pressure exerted by parasitoids, host plant species) [64–66].

5. Conclusion

Understanding how temperature can modulate the fitness outcomes of symbiotic infection is crucial for predicting the eco-evolutionary dynamics of mutualisms facing temperature variations at different scales, especially in the era of climate change. It raises the issue of the condition-dependence of the evolutionary interest for an insect to host a facultative symbiont (cost–benefit balance of symbiotic infection in various thermal contexts). Speculatively, associations with some symbiotic bacteria may have evolved to withstand detrimental environmental conditions such as heat stress, so they are maintained even if fitness costs may occur under optimal temperatures. Our quantitative study confirms the net global increase in the adaptive value of infection by facultative symbionts under heat stress, thereby pointing towards a general role of these microorganisms in facilitating their host adaptation to a warming world. Relationships forged with microbial partners could thus become an additional evolutionary opportunity exploited by insects to persist under climate change, not only through dynamic of symbiont prevalence within host populations, but also through genetic innovations generated by rapid symbiont evolution [6,67].

An associated major question is whether conclusions drawn from aphids could be extrapolated to other insect–symbiont models. For instance, whiteflies resemble aphids from a symbiotic viewpoint in that they also harbour a heat-sensitive intracellular obligate symbiont and can benefit from infection with several facultative symbionts under heat stress [22,68]. Owing to these similarities, it could be postulated that heat stress could modulate this adaptive value, in a way similar to what we observed for aphids. By contrast, many bacterial symbionts exhibit a temporary extracellular lifestyle as they are acquired from environmental sources by diverse host taxa (e.g. beetles, heteropterans, hymenopterans) [69,70]. These models provide an outstanding field of research in thermal ecology of insect–symbiont mutualisms, and are particularly interesting since they may not display the same levels of genomic decay (at least from a functional perspective) and co-cladogenesis with the hosts than intracellular endosymbionts, but their impacts on host ecology are poorly documented [70].

To conclude, the trait-, environment- and species-specific conclusions we drew from aphid-bacterium mutualisms could imply that responses of insect–symbiont partnerships to temperature variations and climate change will probably depend on their life histories, thermal requirements and coevolution dynamics. The mechanisms underlying the thermal sensitivity of insect–symbiont associations are only beginning to be elucidated, but are undoubtedly a cornerstone of the structure and the functioning of many terrestrial ecosystems. We therefore recommend combining multi-trait cost–benefit approaches with realistic projections associated with different scenarios of climate change (e.g. [60,63]) to get a more integrative picture of symbiont-mediated insect responses to ongoing climate disturbance, especially in non-aphid-based systems.

Data accessibility

Our dataset is publicly available at: https://doi.org/10.5281/zenodo.5714555.

Authors' contributions

K.T.: conceptualization, data curation, formal analysis, funding acquisition, investigation, methodology, project administration, supervision, validation, visualization, writing—original draft, writing—review and editing; C.I.: conceptualization, data curation, formal analysis, funding acquisition, investigation, methodology, project administration, supervision, validation, writing—original draft, writing—review and editing.

All authors gave final approval for publication and agreed to be held accountable for the work performed therein.

Competing interests

We declare we have no competing interests.

Funding

K.T. was supported by the F.R.S.-FNRS. C.I. was a beneficiary of UCLouvain FSR Post-doctoral Fellowship.