Thermal niches of specialized gut symbionts: the case of social bees

Abstract

Responses to climate change are particularly complicated in species that engage in symbioses, as the niche of one partner may be modified by that of the other. We explored thermal traits in gut symbionts of honeybees and bumblebees, which are vulnerable to rising temperatures. In vitro assays of symbiont strains isolated from 16 host species revealed variation in thermal niches. Strains from bumblebees tended to be less heat-tolerant than those from honeybees, possibly due to bumblebees maintaining cooler nests or inhabiting cooler climates. Overall, however, bee symbionts grew at temperatures up to 44°C and withstood temperatures up to 52°C, at or above the upper thermal limits of their hosts. While heat-tolerant, most strains of the symbiont Snodgrassella grew relatively slowly below 35°C, perhaps because of adaptation to the elevated body temperatures that bees maintain through thermoregulation. In a gnotobiotic bumblebee experiment, Snodgrassella was unable to consistently colonize bees reared at 29°C under conditions that limit thermoregulation. Thus, host thermoregulatory behaviour appears important in creating a warm microenvironment for symbiont establishment. Bee–microbiome–temperature interactions could affect host health and pollination services, and inform research on the thermal biology of other specialized gut symbionts.

1. Introduction

Earth's climate is rapidly warming, and there is an urgent need to understand how organisms will respond [1]. One factor complicating such predictions is the role of interspecific interactions [2], and, in particular, symbiosis [3]. Many organisms closely associate with one or more distantly related partners that have highly distinct physiologies, as in the case of animal or plant hosts and their microbiomes. Hosts and microbes are likely to have different responses to temperature; yet if they are mutually dependent, the combined niche is restricted to that of the more sensitive partner. Furthermore, thermal niches can themselves evolve in response to symbiotic lifestyles. For example, obligate endosymbionts that undergo strong population bottlenecks during transmission may evolve unstable, easily denatured proteins as a consequence of mutation accumulation, leading to heat sensitivity [4,5]. Both of these factors may constrain the combined thermal niche of strongly symbiont-dependent organisms [6,7]. There is evidence for symbiont-imposed constraints on host thermotolerance in a variety of invertebrates such as aphids, stinkbugs, corals and sponges [8–11]. However, the wider prevalence of this phenomenon is unclear, and in general, we do not know how microbiomes will influence host responses to climate warming.

The eusocial corbiculate bees (hereafter ‘social bees') are a particularly important group in which to study symbiont thermal niches and their effects on hosts. This clade, comprising honeybees (Apis), bumblebees (Bombus) and stingless bees (Meliponini), is host to anciently associated, host-specialized and beneficial gut microbiomes [12,13]. Social bees are also key pollinators in both agricultural and natural ecosystems, but many are declining [14]. For bumblebees in particular, rising temperatures have been identified as a driver of range shifts and population declines in some species [15,16]. If gut symbionts are sensitive to heat stress, the microbiome could be one route through which climate change impacts bee health. Furthermore, because social bees as a whole exhibit extensive strain-level diversity in their microbiomes [12,17], strain variability in thermotolerance could partially underlie the corresponding variability among hosts, as was recently shown for endosymbionts of aphids [18].

Social bees present a uniquely complex thermal environment for their microbiome, making it challenging to predict their symbionts' thermal traits. They are not strictly poikilothermic; rather, they facultatively regulate the temperature of both their bodies (and individual body parts) as well as their shared nests [19–21]. These microenvironments are partially buffered from external fluctuations in temperature, but to a degree that is highly dynamic among individuals, over time and space, and across the bee phylogeny. Even within a single nest, the microbiome is distributed across individuals that vary in behaviours such as foraging or brood incubation, which involve changes in host body temperature [22–24]. Furthermore, social bee species regulate their nest temperatures to different set-points and exhibit different overwintering strategies [21,25,26]. For example, the microbiomes of temperate-zone bumblebees must overwinter within diapausing queens, while the microbiome of Apis mellifera is transmitted by a cluster of active, heat-generating workers [27,28].

As in most gut symbionts, symbionts of social bees experience a brief ex vivo phase during transmission, potentially imposing selection on thermal traits. The gut microbiome is transmitted via a faecal–oral route, usually between nest-mates within a hive [29,30], but horizontal transmission between bee species has also been inferred [12,28]. Although the symbionts cannot grow under ambient oxygen levels outside the bee gut [13], the ability to tolerate thermal stress while on flowers [31] or other external habitats could influence horizontal transmission rates and thus patterns of biogeography and host specificity. All of these factors add up to a complex selective landscape—even within a single host species—involving different castes, seasons and ex vivo phases.

Very little is currently known about the thermal biology of social bee microbiomes. Recent work on Bombus impatiens has shown that, once established in the gut, core symbionts are relatively robust to temperatures from 21 to 37°C [32]. However, honeybees and bumblebees emerge largely symbiont-free as adults and must acquire their symbionts from nest-mates [29,30,33]. It is not known if the colonization process, a crucial phase for both hosts and symbionts, is more temperature-sensitive than the maintenance of an established symbiosis. Furthermore, there has been no comparative work investigating how symbiont thermal traits have evolved across social bees. Social bee taxa maintain different nest temperatures and use different strategies to overwinter and to establish new colonies. They also occupy a climatically diverse range of environments, from arctic and alpine habitats to tropical forests [34]. This variability may impose divergent selection on symbiont thermal traits, with potential feedbacks on the thermal tolerance of the hosts themselves.

We used common garden experiments (in vitro) to characterize symbiont thermal niches with a culture collection of Snodgrassella and Gilliamella, two bacterial species that are ubiquitous across honeybees and bumblebees [12]. Symbiont strains from these two bee lineages belong to deeply divergent clades and appear restricted to their native host [12,35]. We measured thermal limits of growth from 12 to 48°C, tolerance of a brief heat exposure up to 52°C, and growth rates at 28°C versus 35°C. The 28°C is within the range of brood nest temperatures reported for some bumblebee species [36,37], while honeybee brood nest temperatures are typically approximately 33–36°C [25,26,38].

We hypothesized that for all three metrics, honeybee-associated strains would exhibit higher thermotolerance than bumblebee-associated strains, as a result of adaptation to a generally warmer host (nests and bodies) and external environment. We also examined whether the thermal environment impacts Snodgrassella establishment in the bumblebee B. impatiens. This experiment examined whether the colonization process is particularly vulnerable to thermal stress; our initial prediction was that an abnormally high nest temperature for B. impatiens (35°C) would impair symbiont acquisition. However, from a gut microbe's perspective, a ‘normal' temperature would depend more on the immediate surroundings (the gut) than the external environment (the bee nest). These are not necessarily equivalent, as bees can modulate their body temperature using thermoregulatory behaviours. Our findings suggest that bee symbionts may depend on host thermoregulation to provide a favourable microenvironment for rapid growth. Overall, this study provides a foundation for future work on the thermal ecology of bee gut microbiomes and raises new questions about the role of host thermoregulatory behaviour in mediating symbiosis.

2. Material and methods

(a). Culture collection

Strains of Snodgrassella alvi and Gilliamella spp. used in this study were collected as described in [35,39,40]. To our knowledge, no strains had undergone significant passaging in the laboratory since the original isolates were obtained, with the exception of S. alvi strain wkB2. Bumblebee host species were categorized as high-elevation or low-elevation species for electronic supplementary material, figures S1, S2 following range descriptions in [41]. Our culture collection represents isolates from 16 host species (electronic supplementary material, figure S1). Excepting A. mellifera, from which we tested three Snodgrassella strains, one Snodgrassella strain and/or one Gilliamella strain from each host species was included in the sample set.

(b). Thermal limits assay

From glycerol stocks, strains of Snodgrassella or Gilliamella were cultured on Columbia blood agar plates under 5% CO₂ and 35°C for 2 days. These were then restreaked, and the overnight cultures were resuspended in Insectagro medium (Corning) and adjusted to an OD₆₀₀ of 0.5. We further diluted these cell suspensions 1/20 and spotted 10 µl in triplicate onto the surface of fresh Columbia plates. Negative controls (10 µl Insectagro spotted onto plates) were included to ensure there was no background contamination. Given that the same OD can correspond to different densities of viable cells, we also quantified the corresponding colony-forming unit (CFU) count of the inoculum for each strain. Plates were incubated under 5% CO₂ at 12, 16, 20, 24, 28, 32, 36, 40, 44 or 48°C. After 48 h, we scored these plates for whether visible biomass was present or absent. In all but two of 285 strain-by-temperature combinations, the three technical replicates had the same result. In the two exceptions (Fer1.1 at 16°C and Occ4.2 at 28°C), growth was counted despite 1/3 replicates lacking visible biomass. If a strain could not grow at or below 28°C, it was not tested at colder temperatures. For example, strain wkB339 could not grow at 28°C so it was not further tested at 24°C or below.

We used logistic regression (implemented in R [42] as a generalized linear model with binomial error and logit link function) to test whether host genus (Apis versus Bombus) predicted the ability of symbionts to grow at 44°C, the upper thermal limit across strains in this assay. We chose this particular temperature because there was no strain-level variation in the ability to grow at the next lowest (40°C, all positive) or highest (48°C, all negative) tested temperatures (electronic supplementary material, figure S1). We took the same approach for the lower thermal limit, using growth data from the lowest temperature tested in our assay (12°C). To account for the potential influence of starting inoculum size on the probability of growth, we again used logistic regression, but with log-transformed CFU counts for each strain as a predictor.

(c). Heat stress assay

In this and the following assay, we focused on Snodgrassella. Cultures were initially prepared as above. We then resuspended overnight cultures in Insectagro, adjusted them to an OD₆₀₀ of 0.5 and further diluted 1/20 in 200 µl in PCR plates. Cells were subjected to a 1 h heat stress treatment using a thermocycler with a temperature gradient from 35.4°C to 51.6°C (three technical replicates per strain per temperature). They were then transferred to 96-well cell culture plates (Corning) in a microplate reader (Tecan) with 5% CO₂ and 35°C. Growth was monitored by OD₆₀₀ readings taken every 3 h for 66 h. We included blank wells (Insectagro only) as negative controls and subtracted their OD₆₀₀ values from those of the cultures.

(d). Thermal performance assay

Snodgrassella cultures were prepared as for the heat stress assay, except that they were transferred directly into 96-well cell culture plates in a microplate reader (without the thermocycler step) and incubated at either 28°C or 35°C for 72 h. We fit logistic curves to the data using the growthcurver package [43] in R and used a two-way ANOVA to test for effects of incubation temperature and host genus on the intrinsic growth rate (r). Linear regression was used to test whether starting inoculum size (log-transformed CFU counts) predicted growth rate.

(e). Colonization experiment

To obtain gnotobiotic bumblebees, we collected clumps of pupal cocoons from four separate commercial colonies of B. impatiens (Koppert USA). We then surface-sterilized the clumps in diluted bleach (0.2% NaOCl) for 90 s as described previously [44,45] to minimize contamination of the emerging workers. We maintained the sterilized cocoon clumps in sterile conditions in an incubator (Percival Scientific, model I36NL) at 35°C and monitored them daily for adult emergence. Newly emerging adults were transferred to sterile vials and randomly assigned to Snodgrassella or buffer-only treatments. The former were fed with 10 µl of filter-sterilized sugar syrup (50% v/v) containing approximately 10⁶ cells of S. alvi strain wkB12, following [35]. We prepared new Snodgrassella inocula daily from overnight cultures; these were not continuously propagated but rather independently obtained from the same frozen stock of wkB12. Negative-control bees were fed an identical solution but without cells. All bees were monitored to ensure that they consumed the entire 10 µl of inoculum.

We then transferred bees to sterilized 16 oz plastic containers in groups of 2–3 as microcolonies [46]. Bees within a microcolony were assigned to the same treatment and were obtained from the same source colony. Each container was provided with 10 ml of sterile 50% sugar syrup and 500 mg of sterile pollen dough (gamma-irradiated honeybee pollen mixed with sterile 50% sugar syrup). Microcolonies were reared in one of two incubators (Percival Scientific, model I36NL) set at either 29°C, representing a typical bumblebee rearing temperature [32,46–48], or at 35°C, a temperature typical of A. mellifera hives [26]. The pollen lump was replaced on the third day of rearing.

After 5 days, bees were briefly anaesthetized in ice and used for gut dissections. The gut (including hindgut and midgut) was removed from each bee, homogenized with a sterile plastic pestle and resuspended in 1 ml of Insectagro. This homogenate was serially diluted and plated on Columbia blood agar. We counted CFUs after 2 days of incubation at 35°C in 5% CO₂. In performing CFU counts of the focal Snodgrassella strain, we noticed the occasional presence (in 13% of bees overall) of an unidentified bacterium. This contaminant had distinct colony morphology and exhibited slow growth and haemolysis. There was no significant association between the inoculum or the rearing temperature with the presence of the contaminant (logistic regression; inoculum p = 0.19, temperature p = 0.066).

We used a Poisson generalized linear mixed model with the lme4 package [49], treating colony source as a random effect, to test whether rearing temperature was associated with Snodgrassella CFU counts. We also used logistic regression (as described above) to test whether bee survival was predicted by the inoculum treatment or by rearing temperature.

3. Results

We first used an in vitro assay to measure the temperature limits of growth of a collection of Snodgrassella and Gilliamella strains isolated from honeybees and bumblebees. While all strains could grow at 40°C (electronic supplementary material, figure S1) and none at 48°C, they varied in their ability to grow at 44°C (figure 1). Bumblebee strains were significantly less likely to be able to grow at this temperature than honeybee strains (logistic regression, p = 0.015).

Figure 1.

Ability of strains of two core bee gut symbionts, Gilliamella and Snodgrassella, to grow after 48 h incubation on solid media at 44°C. All strains could grow at 40°C while none could grow at 48°C. Honeybee (Apis) strains tend to be better able to grow at 44°C than bumblebee (Bombus) strains. Thermal limits broken down by host species of origin are shown in electronic supplementary material, figure S1. Growth data are based on three technical replicates per strain per temperature. (Online version in colour.)

There was also substantial strain-level variation in lower thermal limits, especially for Snodgrassella (electronic supplementary material, figure S1). However, probability of growth at the lowest temperature tested (12°C) was positively associated with starting inoculum size (logistic regression, p = 0.0047), which was not the case for growth at 44°C (logistic regression, p = 0.74). Hence, we regard the lower limit data as conservative estimates; many of these strains are likely to be able to grow at lower temperatures than indicated in electronic supplementary material, figure S1, especially with larger starting inocula or a longer assay duration.

In addition to characterizing the heat-sensitivity of symbiont growth under a constant temperature, we also sought to determine the ability of bee gut symbionts to tolerate short-term exposure to more extreme heat. Exposure to temperatures above 35°C for 1 h clearly delayed subsequent growth in vitro (figure 2). Overall, however, Snodgrassella appears to be quite heat-tolerant in that all six strains assayed could recover from exposure to 48.7°C. The exact limit was variable among strains. Although larger strain sample sizes are needed for a conclusive comparison, the three Apis­-associated strains tended to be slightly more heat-tolerant than the three Bombus-associated strains (figure 2). Strain robustness to short-term exposures was not simply a function of the initial inoculum size. For example, the only two strains that could recover following exposure to 51.6°C, wkB9 and wkB237, had the lowest inoculum sizes.

Figure 2.

Ability of the core bee gut symbiont Snodgrassella to recover from short-term exposure to high temperatures. Curves represent growth at 35°C and 5% CO₂ following a 1 h heat stress treatment applied using a gradient thermocycler, showing the mean OD₆₀₀ (optical density, a proxy for cell concentration) values of three replicates per strain per temperature. The host species from which each strain was isolated is indicated. Snodgrassella is generally robust to high temperatures, though tolerance varied among strains. For reference, host lethal limits are approximately 40–46°C for Bombus and approximately 50–52°C for Apis. (Online version in colour.)

Even when thermal limits do not vary among symbiont strains, the thermal optima could vary. To address this, we conducted Snodgrassella growth assays in liquid media at 28°C and 35°C, representing temperatures closer to typical Bombus or Apis nest temperatures, respectively. Inoculum size (number of CFUs at the start of the growth assay) did not predict growth rate at either 28°C or 35°C (linear regression, p = 0.69 and 0.82, respectively). Bumblebee-associated Snodgrassella typically had higher growth rates than honeybee-associated Snodgrassella (figure 3; two-way ANOVA, p = 0.0085). Although further tests using larger sample sizes—especially of Apis strains—are warranted, we found no significant interaction between host genus and temperature (two-way ANOVA, p = 0.24), as would be expected if symbiont growth were differentially adapted to the temperatures closest to those of their hosts' nests. Instead, growth rates of most strains were higher at 35°C (figure 3; two-way ANOVA, p = 0.0059). Among bumblebee strains, those from host species inhabiting higher elevations were not conspicuously better able to grow at the cooler temperature (28°C) than those from low-elevation hosts (electronic supplementary material, figure S2). In fact, the only strains that exhibited higher growth rates at 28°C than at 35°C were derived from A. mellifera (electronic supplementary material, figure S2). However, further tests including additional strains and bee species will be necessary to draw firm conclusions.

Figure 3.

In vitro growth rates of Snodgrassella strains at 28°C and 35°C. Dots connected by thin lines represent the maximum growth rates of each strain, coloured by whether the strains were isolated from honeybees (Apis) or bumblebees (Bombus). Thick lines connect the median growth rate for Apis or Bombus strains at each incubation temperature. All bumblebee Snodgrassella grow faster at 35°C, a temperature significantly higher than typical bumblebee nests. N = 5 Apis strains, N = 15 Bombus strains. (Online version in colour.)

We next examined whether symbiont thermotolerance might influence fitness in vivo using a colonization experiment with S. alvi strain wkB12 and gnotobiotic bumblebees. In vitro, this strain grows much more quickly at 35°C as compared with 28°C (figure 4b). No Snodgrassella colony-forming units (CFUs) were detected in any of the bees inoculated with sterile buffer (N = 15 reared at 29°C, N = 16 reared at 35°C). Among bees inoculated with a standardized dose of approximately 10⁶ Snodgrassella cells, gut colonization after 5 days was highly dependent on the thermal environment (figure 4a). Accounting for the different source colonies from which bees were obtained, CFU counts were significantly higher—by over 50-fold, comparing medians—in the 35°C rearing treatment (GLMM, p < 0.0001). Several bees reared at 29°C had no detectable Snodgrassella cells.

Figure 4.

(a) Effects of the thermal environment on colonization of Snodgrassella alvi wkB12 in gnotobiotic bumblebee workers (Bombus impatiens). Bees inoculated with S. alvi (approx. 10⁶ CFUs) were maintained for 5 days at 29°C, a typical temperature in B. impatiens nests (N = 22), or 35°C (N = 23). Colours indicate the four replicate colonies that were used for the experiment, and black bars indicate the median S. alvi titre in bee guts for each temperature treatment. Bees fed a sterile buffer and maintained under the same conditions had no detectable S. alvi colonization (not shown). CFUs = colony-forming units; ND = not detected. (b) Growth of S. alvi wkB12 in vitro when incubated at 28°C versus 35°C. Growth curves represent the mean OD₆₀₀ (optical density, a proxy for cell concentration) values of three replicates per temperature. (Online version in colour.)

We also examined whether bee survival was influenced by the experimental treatments. Overall, 92% of bees survived from adult emergence to dissection on day 5. The probability of survival was not affected by the inoculum (Snodgrassella versus buffer alone) or the rearing temperature (logistic regression; inoculum p = 0.49, temperature p = 0.33) (electronic supplementary material, figure S3). As only bees that survived for the entire 5-day rearing period were used for CFU counts, we were not able to examine whether gut colonization was predictive of survival.

4. Discussion

We first used in vitro assays to characterize bee symbiont thermal niches in a common environment and without potential host-mediated effects. The experiments were based on a collection of isolates of Snodgrassella and Gilliamella, which are ubiquitous, keystone members of the social bee gut microbiome [12,13]. Overall, we found that these symbionts are quite heat-tolerant relative to their hosts. All strains of both symbiont species can grow at a constant temperature of 40°C, and many can grow at 44°C (figure 1). Likewise, another core symbiont, Lactobacillus bombicola, has an optimal growth temperature around 40°C [50]. By contrast, honeybees and bumblebees do not normally allow their nests to reach temperatures above 35°C, which would harm brood development [25,26,36–38].

Bees can, however, maintain higher body temperatures for brief periods while foraging. For example, the abdomen (which contains the gut microbes [13]) of bumblebees foraging in full sunlight may reach close to 40°C [22]. The abdomen is also used to dissipate excess heat generated in the thorax [51]. We used short-term heat treatments of Snodgrassella to mimic this kind of temporary exposure and found that bumblebee-associated strains could recover from an hour-long exposure to at least 48.7°C, while honeybee-associated strains could recover from at least 50.4°C (figure 2). These limits are not greatly exceeded by the lethal limits (broadly defined) of their hosts, which are approximately 40–46°C for bumblebees [52,53] and approximately 50–52°C [54,55] for honeybees.

Bumblebees have experienced population declines partly linked to climate warming [15,16]; in this study, we asked whether heat-sensitive symbionts could constitute one underlying mechanism. The comparative robustness of Snodgrassella, Gilliamella and Lactobacillus bombicola [50] to high temperatures in vitro suggests that the gut microbiome does not constrain bee tolerance of heat. Rather, other factors rooted in host physiology, behaviour and ecology likely explain the observed impacts of climate change on bumblebee populations [56–58].

Snodgrassella and Gilliamella are heat-tolerant not only compared to their hosts, but also compared to many insect endosymbionts. For example, aphids, weevils and stinkbugs have obligate associations with highly heat-sensitive endosymbionts, which can be killed by temperatures as low as 30–35°C [8,9,59]. One trait these symbionts share is a strict maternal inheritance; this transmission mode enforces a clonal population structure that results in genome degeneration and, ultimately, impaired heat tolerance [4,5]. By contrast, social transmission of gut symbionts permits the strain mixing and recombination that is more typical for free-living bacteria. While mostly vertically transmitted between colonies, bee gut symbionts likely maintain larger population sizes and undergo recombination more frequently than endosymbionts. All bee gut symbiont species are culturable outside the host [13] and possess genomes that do not exhibit the hallmarks of degenerative evolution [35].

Like many gut microbes, bee gut symbionts experience a brief but potentially crucial ex vivo phase during transmission. Selection for persistence on nest substrates or on flowers—which may have elevated temperatures [31])—could influence symbiont heat tolerance. In vivo selective pressures related to the unique thermoregulatory behaviour of social bees may further explain the broad thermal range of bee symbionts. A bee-inhabiting microbial population will experience a wide range of body temperatures as its hosts forage in the environment, thermoregulate their nests and overwinter. Comparisons to non-bee-associated but related bacteria (e.g. Stenoxybacter from termites [60] or Orbus from butterflies [61]) would be useful to reconstruct how evolution in social bees specifically has shaped the thermal niches of the bee gut microbiome.

While heat-tolerant relative to their hosts and to other bacterial symbionts of insects, Snodgrassella and Gilliamella strains do vary in thermal traits. Specifically, honeybee strains are somewhat more heat-tolerant than bumblebee strains (figures 1 and 2), a pattern that matches the corresponding thermal traits of their hosts. For example, honeybees typically occupy tropical and subtropical environments (with the exception of introduced A. mellifera), maintain warmer nests, have higher upper thermal limits, and do not undergo diapause in winter [20,55]; bumblebees likely originated in montane environments and are considered a cold-adapted group [62]. Our findings are consistent with previously observed correlations between symbiont thermotolerance and the local thermal environment [18,63]. In the case of bees, even if divergent thermal niches of symbionts do not affect hosts, they could affect the potential for strains to successfully disperse between host colonies or even species, ultimately influencing their biogeography and degree of host specialization.

We did not find such host-symbiont matching for Snodgrassella thermal performance (i.e. relative growth rate at two temperatures). Bumblebee strains uniformly grew faster at 35°C, while most A. mellifera strains grew slightly faster at 28°C (figure 3, electronic supplementary material, figure S2). This pattern is the opposite of what would be expected if the ambient temperature in active colonies primarily determines the optimal growth temperature of symbionts, because honeybees generally maintain warmer nests. Overwintering biology may explain the A. mellifera-derived strains' comparatively higher growth rates at the cooler assay temperature. Unlike bumblebees, A. mellifera workers form active clusters that maintain above-ambient, but cool temperatures through the winter (average approximately 21°C; [26]).

What explains our observation that bumblebee-associated Snodgrassella, like Lactobacillus bombicola [50], grow faster at a temperature exceeding that of their nests (figure 3)? We suggest that the answer lies in the bee abdomen—the microenvironment inhabited by the symbionts, and a structure whose temperature is affected by host thermoregulatory behaviour. Specifically, both worker and queen bumblebees incubate larvae and pupae by lying on top of brood structures and elevating abdominal temperatures to approximately 35°C or above, exceeding ambient temperatures in the nest [64,65]. Given that bumblebees—especially foundress queens, the sole source of microbes for the colony—spend much of their time performing this behaviour [36,64], symbiont growth may be adapted to the locally heated conditions within the abdomen.

To further explore this possibility, we tested whether the warm-shifted growth preference of bumblebee-associated Snodgrassella (figure 3) might have fitness effects in vivo. We conducted an experiment on gnotobiotic B. impatiens, using conditions (microcolonies lacking brood) in which incubation behaviour is limited, and thus abdominal temperatures are expected to more closely match the ambient rearing temperatures (29°C and 35°C). Previous literature hinted that Snodgrassella colonization of B. impatiens might differ between these temperatures. In one study, inoculation of B. impatiens with Snodgrassella resulted in 100% colonization and consistently high titres [35], whereas another study of B. impatiens reported erratic colonization and frequently low titres [45]. These studies differed in the temperature at which bees were reared, with consistent colonization observed at 34°C and erratic colonization at approximately 26°C. In line with these experiments, we found that Snodgrassella colonization was variable and occasionally unsuccessful in bees reared at 29°C (figure 4a).

One possible explanation for this result has to do with the fact that the focal strain used, S. alvi wkB12, was not isolated from B. impatiens but rather B. bimaculatus. If bumblebees control colonization by other species' or even colonies’ symbiont strains, as has been suggested [45], robust growth at 35°C may reflect a weakening of putative strain-specific filtering mechanisms at this unnaturally high rearing temperature—e.g. through suppression of immune responses [66]. In this scenario, symbiont colonization might actually be detrimental for the host.

However, a recent study found that immune responses were not downregulated in bumblebees reared at 38°C [48]. Furthermore, we lack evidence that bumblebees can discriminate between their own symbiont strains and those from other Bombus, immunologically or otherwise. Experimental and phylogenetic evidence indicates that Snodgrassella strains can and do move between congeneric bee species [12,28], weakening host-species specificity of symbiont colonization. Furthermore, despite being isolated from B. bimaculatus, S. alvi wkB12 is closely related to strains from B. impatiens [17].

Another explanation for this result is based in the thermal niche of bumblebee-associated Snodgrassella strains. Snodgrassella alvi wkB12, like the other strains tested, grows more quickly at 35°C than 28°C in vitro (figure 4b). In bees reared in microcolonies at 29°C—where abdominal temperatures are expected to generally match ambient temperatures because of restricted thermoregulatory behaviour—ingested Snodgrassella cells may not be able to grow or adhere to the gut wall quickly enough to establish before being lost to defaecation. However, once established, they appear to be quite robust to a wide range of temperatures [32]. A subset of bees reared at 29°C did acquire high Snodgrassella titres that exceeded the number of cells in the inoculum, implying replication in the gut (figure 4a). One possibility is that these individuals had begun to incubate the provided pollen lump, a behaviour observed in microcolonies [67]. As a consequence, they may have maintained higher abdominal temperatures conducive to Snodgrassella colonization. Analogously, in a bumblebee nest, thermoregulatory behaviour may be important to gut microbiome acquisition.

Several caveats should be considered. First, 10 of the 76 experimental bees harboured a gut bacterium morphologically distinct from Snodgrassella, that was present in both Snodgrassella-inoculated and control bees, and in both rearing temperatures, and not significantly associated with either treatment. Hence, this bacterium was probably not responsible for the effect of temperature on Snodgrassella colonization. Second, incubator factors other than temperature cannot be entirely ruled out. However, these are unlikely, as the two incubators used are the same model and age and show no apparent differences. Third, our experiments using single isolates do not account for potential effects of microbe–microbe interactions on symbiont thermotolerance [68,69]. Such interactions—as well as the altered physiological state of cultured cells—may explain why monocolonizations of Snodgrassella into bees reared at 29°C or below have been erratic (figure 4a and [45]), while inoculations with faeces at these temperatures have not [30,44]. Finally, symbiont strains with different thermal niches might coexist within individual bees, which would complicate our findings. But while individual A. mellifera do harbour a diverse set of strains for each core gut bacterial species [70], in other social bees, within-individual strain diversity appears to be low or non-existent [17,70].

5. Conclusion

Our results imply that the gut microbiome does not constrain social bee responses to heat stress. However, we find that symbiont growth is somewhat sensitive to cold temperatures, which can occur during hibernation or periods of inclement weather, or due to stressors that disrupt host thermoregulation (e.g. pesticides [71]). Whether symbionts indeed constrain—or even improve [72]—bee responses to thermal challenges and other types of stressors remains an important priority for future research.

Our findings are also relevant beyond bees. We have identified a potential feedback between host behaviour and the microbiome: social bee thermoregulatory behaviours have provided elevated and buffered temperatures that may have shaped the evolution of their symbionts' thermal niches. Bees could be a useful model for mammals, including humans, which also create warm gut environments and harbour socially transmitted gut symbionts. There are potential parallels in how host thermoregulation influences, and even benefits from, the microbiome; for example, it may provide a useful means of speeding up or curtailing symbiont growth.

Ethics

Laboratory experiments were conducted on commercially raised bumblebees (Bombus impatiens), which are not subject to requirements on ethical approval.

Data accessibility

Data and R code used for analyses and visualizations are publicly available at: https://doi.org/10.6084/m9.figshare.12486395.

Authors' contributions

T.J.H. and N.A.M. designed the study. T.J.H. and E.L. conducted laboratory experiments. T.J.H. analysed data and made the figures. T.J.H. and N.A.M. drafted the manuscript with input from E.L. All authors gave final approval for publication and agree to be held accountable for the work performed therein.

Competing interests

The authors have no competing interests to declare.

Funding

This research was supported by a postdoctoral fellowship (2018-08156) from the USDA National Institute of Food and Agriculture to T.J.H., a University of Texas Undergraduate Research Fellowship to E.L. and National Institutes of Health award (R35GM131738) to N.A.M.