Cross-species gut microbiota transplantation predictably affects host heat tolerance

The host microbiota represent a complex biological community consisting of bacteria, fungi and viruses that has received increased focus over recent decades as its highly influential role in host physiology has become elucidated (Kohl and Carey, 2016; McFall-Ngai et al., 2013; Warne and Dallas, 2022). Resultantly, how the host microbiota responds to changes in environmental variables that are intimately tied to organismal biology, such as temperature, is of great interest. There is evidence that environmental temperature alters both the compositional and functional diversity of the gut microbiota in vertebrates and invertebrates (Hassenrück et al., 2021; Jaramillo and Castaneda, 2021; Kohl and Yahn, 2016; Li et al., 2020; Moghadam et al., 2018; Onyango et al., 2020; Wang et al., 2018; Woodhams et al., 2020). In their review, Sepulveda and Moeller (2020) found that warming temperatures are associated with an increased relative abundance of Proteobacteria in invertebrates and a reduction in Firmicutes in vertebrates, and these temperature-dependent changes in the microbiota community affect varying aspects of host fitness and performance (Fontaine et al., 2018; Horváthová et al., 2019; Kikuchi et al., 2016; Prado et al., 2010). Therefore, the risks of climate change, which are expected to worsen over the coming decades (IPCC, 2021), extend to the host microbiota and represent an important topic of study.

A fundamental aspect of an ectotherm's biology is their thermal tolerance, as it limits the spatial and temporal ranges they can occupy (Adolph and Porter, 1993; Hoffmann et al., 2013; Huey, 1982). Thermal tolerance of ectotherms is typically estimated via critical thermal (CT) limits that denote the endpoints of their thermal performance curve (TPC) where organismal performance reaches zero (Huey and Stevenson, 1979). Species with higher CT maxima (CT_max) are expected to be less susceptible to warmer environments, enabling them to persist in the projected climates (Gilbert and Miles, 2017; Huey et al., 2012; Roeder et al., 2021). Therefore, we suggest that a higher CT_max will be beneficial to many populations in the future. The factors that underlie organismal CT_max and, to a broader extent, heat death in ectotherms remain debated. Several proposed mechanisms include perturbations in the plasma membrane (Bowler, 2018), a decline in aerobic scope (Pörtner, 2002), loss of cardiovascular function (Somero, 2010) and disruption to mitochondrial activity (Chung and Schulte, 2020). Given the role of CT_max in promoting species persistence, there is foremost importance to exploring the different aspects of an individual's environment and biology that can elevate CT_max.

One potential, yet relatively understudied, factor of organismal biology that can influence host heat tolerance is the host microbiome. Across several invertebrate and vertebrate species, there is evidence from experimental manipulations linking the host microbiome to changes in CT_max (Baldassarre et al., 2022; Doering et al., 2021; Fontaine et al., 2022; Moeller et al., 2020). For example, an increasing prevalence of the Anaerotignum genus positively improved the CT_max of western fence lizards, Sceloporus occidentalis (Moeller et al., 2020). Additionally, Doering et al. (2021) showed that transplanting the microbiome of coral populations from thermally fluctuating waters to conspecifics from thermally stable environments enhanced their bleaching resistance, indicative of greater heat tolerance. There are several potential pathways through which the host microbiota can elevate host CT_max. For instance, mitochondrial enzyme activity was reduced in larval anurans with an experimentally depleted microbiota (Fontaine et al., 2022), which can decrease aerobic scope at high temperatures and limit CT_max (Pörtner, 2002). Furthermore, warm-acclimated microbiota communities exhibit higher expression of reactive oxygen scavengers (Fontaine and Kohl, 2023; Ziegler et al., 2017) that can limit the damage reactive oxygen species impart on mitochondrial mechanisms at high temperatures (Christen et al., 2018; Sokolova, 2023). Beyond enhanced CT_max, members of the host microbiota can promote enhanced host fitness in warmer temperatures as demonstrated in pea aphids, Acyrthosiphon pisum, hosting an essential bacterial symbiont (Buchnera aphidicola) (Montllor et al., 2002; Zhang et al., 2019). These examples suggest that the host microbiome confers benefits to the host in terms of heat tolerance that could facilitate survival in warmer environments.

Amphibians are imperiled by climate change as increases in temperature and unpredictable shifts in precipitation will restrict activity periods and place species at risk of experiencing heat stress (Campbell Grant et al., 2020; Greenberg and Palen, 2021; Hoffmann et al., 2021; Sunday et al., 2014). Based on the risks associated with warming temperatures, we sought to examine how early-life manipulation of the gut microbiota in a geographically widespread anuran influenced its heat tolerance. Specifically, we used a cross-species microbiota transplant to demonstrate that species-specific differences in thermal tolerance are mediated, at least partially, by the gut microbiota. While cross-species microbiota transplants have not been examined in terms of heat tolerance, Warne et al. (2019) showed that egg masses of wood frogs (Lithobates sylvaticus) inoculated with the gut microbiota of larval bullfrogs (Lithobates catesbeianus) displayed higher growth and developmental rates as well as enhanced disease resistance compared with those inoculated with a conventional wood frog microbiota.

To evaluate whether gut microbiota and inter-specific transplants influence thermal tolerance, we stripped the resident microbiota of field-collected wood egg masses using antibiotics and separated these ‘sterilized’ eggs into three microbiota treatments. Following the antibiotic wash, eggs were either inoculated with pond water from the collection site or with the gut microbiota of field-collected larval green frogs (Lithobates clamitans), or received no inoculation. As hatching represents a critical window for colonization of the gut microbiota (Warne et al., 2017), inoculating the eggs enables the microbiome transplants to become established in the developing larvae. As green frogs breed in warmer temperatures than wood frogs and their larvae are active over the summer months, whereas wood frogs metamorphose prior to the summer (Hulse et al., 2001), green frog larvae have been shown to exhibit a CT_max more than 1.5°C higher than that of wood frog larvae (Katzenberger et al., 2021). Resultantly, we predicted that inoculating wood frog eggs with the gut microbiota of larval green frogs would increase the heat tolerance of the recipient. We also evaluated whether a short-term acclimation period would alter the gut microbiota. Most studies measuring the effects of different acclimation temperatures on the gut microbiota exceed 10 days (e.g. Fontaine et al., 2022; Moeller et al., 2020; Zhu et al., 2021), and the impact of a short-term thermal acclimation period has not been thoroughly explored (but see Fontaine and Kohl, 2020). Therefore, we sought to identify how the gut microbiota changes in response to a brief acclimation period of 3 days and whether there was an interactive effect between acclimation temperature and microbiota treatment on wood frog CT_max.

Four freshly laid (<36 h old) wood frog, Lithobates sylvaticus (LeConte 1825), egg masses were collected from wetlands in Jackson Co., IL, USA, on 23 February 2022 (air temperature ∼0°C), under an Illinois Department of Natural Resources permit (HSCP 19-03). Additionally, 10 larval green frogs, Lithobates clamitans (Latreille 1801), were collected from Southern Illinois University Carbondale research ponds in Jackson Co., IL, USA on 23 February 2022. All experimental procedures were approved by the Southern Illinois University Institutional Animal Care and Use Committee (22–008).

On the day of collection, eggs were rinsed in antibiotics following a previously published protocol that disrupts and reduces the egg-associated microbiota, and affects subsequent larval gut microbiomes upon hatching (Warne et al., 2017, 2019). Briefly, the egg masses were separated into sterile 50 ml tubes (∼20 eggs per tube), and rinsed three times with 40 ml of autoclaved, aerated carbon-filtered water. The eggs were then treated by exposure to 500 µl penicillin–streptomycin (10,000 U ml⁻¹; Life Technologies #15140-122), 200 µl of kanamycin sulphate (25 µg ml⁻¹; Life Technologies #11815-032) and 50 µl of amphotericin B solution (250 µg ml⁻¹; Sigma-Aldrich #A2942) for 4 h on a nutator. The antibiotic-treated eggs were then triple rinsed with sterile water and placed into sterilized 6 l plastic containers containing 3 l of carbon-filtered sterile water and aerated with a HEPA inline filter disc (0.3 µm pore, Whatman Inc.) overnight. Additionally, we retained eggs from all four masses that were not exposed to any antibiotics.

The following day, the antibiotic-rinsed eggs were collected and separated into three different treatments: (1) wood frog (WF), (2) green frog (GF) and (3) no-inoculum (NI). The WF treatment eggs were placed in sterile 50 ml tubes and received 40 ml of pond water that was previously mixed with the unmanipulated eggs, with the intention of re-inoculating them with their natural microbiota. For the GF treatment, intestinal tracts were dissected from the larval green frogs (Gosner stage 27–30; Gosner, 1960), homogenized in 40 μl of autoclaved water, and added to sterile 50 ml tubes with the sterilized wood frog eggs. Both WF and GF treatments were mixed with their respective inoculum for 30 min on a nutator and were then returned to their respective containers (n=16 per treatment). For the NI treatment, eggs received no further manipulation beyond being placed in sterilized water and tubs (n=16), but were not maintained in sterile conditions, and thus should be considered to have disrupted rather than entirely sterilized microbiomes (Warne et al., 2017, 2019).

Eggs were maintained at room temperature and hatched within 4 days of the microbiota treatments. Larvae were allowed to feed on the inoculated egg jelly for 2 days to provide initial colonization of the gut microbiomes. Bubblers were removed 1 week after hatching. Initial feedings consisted of autoclaved algal flakes (Bug Bites Spirulina Flakes, Fluval Aquatics, Mansfield, MA, USA), after which they were fed autoclaved, crushed alfalfa pellets twice weekly. Water was changed weekly with aerated, carbon-filtered water. All treatment groups were exposed to the same diet and water.

Beginning on 30 March 2022, we randomly collected 40 larvae from each microbiota treatment over a 4 day period (n=10 per day). All larvae were staged, weighed and transferred to individual 750 ml plastic containers filled with 600 ml of aged (>24 h), aerated, carbon-filtered water. To reduce the effect of individual containers, only 1–3 larvae were collected from each container per day. Across all individuals, larval stages ranged from 27 to 36, and there was no difference among the treatments in terms of stage (F_2,107=0.92, P=0.40) or log-transformed mass (F_2,107=0.75, P=0.48), but collection day influenced larval stage (F_3,107=4.44, P=0.006) but not mass (F_3,107=1.96, P=0.13), and there was no interaction. Lastly, larvae (n=20 per microbiota treatment) were then split into two acclimation temperatures, low (15±0.2°C) and high (23±0.3°C), for a 3 day acclimation period during which individuals were fasted. Following an attempt to measure the CT minimum (CT_min) of larvae from the low acclimation temperature, the sample size was reduced to n=18.

After the acclimation period, all larvae were staged and weighed, and single larvae were placed into individual 125 ml flasks filled with 75 ml of aged, aerated, carbon-filtered water. Flasks were then submerged in a hot water bath (Isotemp 220, Fischer Scientific) and given 5 min to acclimate prior to beginning the assay. In each bath, there were eight flasks containing two individuals from each treatment. To record CT_max, temperature was increased ∼0.7°C min⁻¹ from a starting temperature (mean±1 s.e.m.) of 19.4±0.2°C. Beginning at 33°C, larvae were prodded with a spatula every 30 s until they failed to respond to the stimulus. At this point, a thermocouple probe (Physitemp BAT-12) was placed adjacent to the larvae and the water temperature was recorded, which represented larval CT_max. Flasks were then placed in a bath of room temperature water to facilitate larval recovery. While most larvae recovered within 3 min, 6 larvae died (WF n=4, NI n=1, GF n=1) from the heat shock and there was no treatment effect (χ²=3.31, d.f.=2, P=0.19), and two individuals (WF n=1, NI n=1) died prior to the assay; all were removed from further analyses.

To test for acclimation effects on the gut microbiome, four larvae were selected for microbial sequencing from each of the six groups (three microbiota treatments×two acclimation temperatures) after 3 days of temperature acclimation. We did not measure the CT_max of these 24 larvae to eliminate potential confounding effects of heat shock on the microbiota community. All larvae were euthanized via snap-freezing in −80°C ethanol and stored at −80°C until intestinal extraction.

The entire intestinal tract was removed and placed into an autoclaved 1.5 ml microcentrifuge tube. To reduce cross-contamination, the forceps were rinsed in 70% ethanol and then flame-sterilized prior to intestine removal. Additionally, the intestines were briefly rinsed with autoclaved water to remove any transient bacteria. To improve DNA extraction, the intestines were homogenized using sterilized forceps and then stored at −80°C. Microbial DNA was extracted using the GenCatch™ Plasmid DNA Mini-Prep Kit (Epoch Life Science) following the manufacturer’s instructions with minor modifications: the volume of proteinase K was increased to 25 μl and the initial incubation at 60°C was increased to 3 h to improve cellular digestion. Lab controls were used to identify any microbial DNA present in the extraction kit reagents. All extracted DNA samples were stored at −80°C prior to sequencing.

We used QIIME 2 v.2021.4 (Bolyen et al., 2019) to process a total of 7,573,956 raw sequences, resulting in 4,293,502 bacterial sequences after quality control for 87 samples. Note that only 24 of these 87 samples are for the current study, while the remaining samples are for a parallel study of larval wood frogs. We used QIIME 2 plugin cutadapt (Martin, 2011) to remove the primer sequences. Any sequences with ambiguous bases, with no primer, with greater than 0.1 error rate mismatch with primer or with any mismatches to the sample-specific 12 bp molecular identifiers (MIDs) were discarded. Following initial quality control, we used DADA2 (Callahan et al., 2016) to truncate the reads to a length where the 25th percentile of reads had a quality score below 15 (Forward 271 and Reverse 277). We used the pre-trained SILVA database (v.138) in QIIME 2 for taxonomic assignment of the bacteria. Sequences were annotated to amplicon sequence variants (ASVs), and any unknown or unclassified ASVs were removed from downstream analysis. We rarefied the dataset to 10,000 reads per sample (resulting in 870,000 high quality sequences variants in the 87 samples) to minimize bias resulting from differences in sequencing depth among samples before estimating diversity indices and downstream analyses (Gihring et al., 2012).

All statistical analyses were conducted using R version 4.3 (http://www.R-project.org/). Larval heat tolerance was assessed via a linear mixed model using the lmer package (Kuznetsova et al., 2017) with CT_max as the response variable and both microbiota treatment and acclimation temperature as fixed effects. We included larval mass, which was square root transformed to improve normality, and Gosner stage as covariates and the location of the 125 ml flask in the hot water bath as a random effect to account for potential differences in heating rates. In the results, CT_max is reported as mean±1 s.e.m. along with sample size.

Gut microbiota α- and β-diversity metrics were analyzed using the vegan package (https://CRAN.R-project.org/package=vegan). We compared three α-diversity metrics (ASV richness, Shannon index and Simpson index) using two-way ANOVA with microbiota treatment, acclimation temperature and their interaction as predictive variables, and Gosner stage as a covariate. We used the adonis2 function to perform PERMANOVA, with 999 permutations, using Bray–Curtis and Jaccard distance matrices based on microbiota treatment, acclimation temperature and their interaction along with Gosner stage as a covariate. Using the same distance indices, we compared intragroup heterogeneity via permutation tests with 999 permutations and the betadispr function.

We determined significant differences in bacterial phyla and families across the microbiota treatments and acclimation temperatures using the MaAsLin2 package (https://bioconductor.org/packages/Maaslin2/). We used an arcsine-square root transformation to improve normality of the relative abundance data. To reduce the risks of false positives, we corrected P-values using the Benjamini–Hochberg false discovery rate method.

We found that larval wood frog CT_max was affected by gut microbiota treatment and acclimation temperature, although the differences in heat tolerance were small (Fig. 1). In line with our prediction, microbiome treatment, controlling for both body mass and Gosner stage, influenced CT_max (Table 1; GLMM, χ²=15.10, P<0.001), with GF larvae having the highest CT_max on average. Within the low acclimation temperature treatment, GF larvae had the highest CT_max (38.3±0.2°C, n=13) while NI larvae (37.7±0.1°C, n=12) had the lowest (post hoc contrasts P=0.048). Acclimation temperature had a pronounced effect on CT_max­ in the expected direction, when controlling for body mass and Gosner stage, with those acclimated to warmer temperatures exhibiting higher heat tolerance (Table 1; GLMM, χ²=22.16, P<0.0001). A post hoc analysis found that only NI larvae showed a significant increase (P=0.021) in CT_max between the low and high acclimation temperature treatments (0.6°C), while the GF (0.3°C) and WF (0.4°C) treatments had smaller increases. There was no interaction between gut microbiota treatment and acclimation temperature (χ²=0.53, P=0.77) and the differences in CT_max were independent of body mass (χ²=0.16, P=0.69) and Gosner stage (χ²=0.61, P=0.44) (Table 1).

We analyzed a total of 1,221,946 sequences that were assigned to 145 unique ASVs across the 24 intestinal samples. Across the three gut microbiota treatments, ASV richness was lowest in the WF group and highest in the GF group, but there was no significant treatment or acclimation temperature effect (Fig. 2A). Using Shannon and Simpson indices, α-diversity was significantly different across microbiota treatments, acclimation temperature and larval stage. The WF larvae had significantly lower diversity metrics than both GF and NI larvae, and those acclimated to the high temperature had higher diversity, although the difference between the temperatures was minor (Fig. 2B,C; Fig. S1). Across both indices, diversity increased with larval stage, but the Shannon index model had a near-significant interaction of stage and treatment as NI larvae showed no change in diversity across different stages (Fig. 2B; Fig. S1).

The community composition and membership of the wood frog gut microbiota varied with respect to microbiota treatment, acclimation temperature and larval stage (Fig. 3; Fig. S2). Adonis PERMANOVA identified significant effects of microbiota treatment (Bray–Curtis and Jaccard distances, respectively: F_2,14=4.22, P<0.001; F_2,14=2.98, P<0.001), temperature (F_1,14=4.41, P<0.001; F_1,14=3.27, P<0.001) and Gosner stage (F_1,14=2.90, P=0.009; F_1,14=2.09, P=0.013). For both distance methods, there were no significant interactions. The PCoA plots of Bray–Curtis and Jaccard distances showed moderate overlap among the microbiota treatments and acclimation temperatures (Fig. 3). While the NI larvae at the low acclimation temperature showed the greatest variation, within-group dispersion was similar for both Bray–Curtis (F_5,18=1.15, P=0.36) and Jaccard (F_5,18=1.11, P=0.41) distance analyses. Among the most prominent phyla, variation along PC1 of both distance methods was negatively related to Bacteroidota abundance while Firmicutes and Proteobacteria abundance had positive effects. Actinobacteria accounted for the majority of variation along PC2, being negatively related in the Bray–Curtis distance PCoA and positively related in the Jaccard distance PCoA.

Across all wood frog larvae, gut bacterial communities were dominated by three phyla – Bacteroidota, Firmicutes and Proteobacteria – with relatively lower abundance of Actinobacteria, Desulfobacterota and Verrucomicrobiota (Fig. 4). Across both acclimation temperatures, the relative abundance of Bacteroidota was lower in NI larvae (mean±1 s.e.m.: 38.7±6.6%) than in GF (53.4±5.7%) and WF larvae (66.0±2.4%), but NI larvae had a higher relative abundance of Proteobacteria (38.7±6.1%) with respect to GF (22.1±3.8%) and WF larvae (16.8±2.8%). However, these differences were not statistically significant (MaAsLin2, across all treatments corrected P>0.09; Table S1). The relative abundance of these major phyla was unaffected by the acclimation temperature, indicating their resistance to the short-term acclimation period (MaAsLin2, corrected P>0.72). Across all treatments, Actinobacteria were enriched under the low acclimation temperature treatment (1.1±0.3%) relative to the high temperature treatment (0.45±0.2%), although the difference only approached statistical significance (MaAsLin2, corrected P=0.067).

When examining the relative abundance of bacterial families, only a single family differed between the acclimation temperatures, with Beijerinckiaceae being more abundant across all treatment groups following high temperature acclimation (MaAsLin2, corrected P<0.0001; Table S1). In contrast, there were 14 bacterial families that were significantly different among the three microbiota treatments (MaAsLin2, corrected P<0.05; Table S1).

Increased heat tolerance in organisms is predicted to be beneficial in the future through minimizing the mortality risk from acute heat extremes (Huey et al., 2012). In this study, we build upon current evidence that the gut microbiota influences host heat tolerance. By manipulation of larval wood frog gut microbiota, we observed that transferring the gut microbiota of a more heat-tolerant larval anuran (green frogs) to wood frogs resulted in elevated heat tolerance, indicating that the microbiota is partially tied to host thermal physiology. In these GF-treated larvae, several bacterial families were significantly enriched or depleted; for example, showing a higher relative abundance of Rikenellaceae (a short-chain fatty acid-producing bacteria) and a lower relative abundance of Aeromonadaceae (a potential pathogen) when compared with NI larvae. We found a significant, albeit small, effect of short-term acclimation on gut microbiota community composition and diversity, suggesting the gut microbiota is susceptible to changes following a brief change in environmental temperature. The significance of a cross-species microbiota transplant on increasing the recipient's CT_max underscores the role of the gut microbiota in host thermal tolerance and suggests this method could be useful in species conservation efforts.

Our study is the first to identify that cross-species gut microbiota transplantation influences the heat tolerance of the recipients in a predictive manner. In line with our prediction, inoculating wood frog eggs with the gut microbiota of larval green frogs, a species with greater heat tolerance than the former (Katzenberger et al., 2021), resulted in higher CT_max compared with the WF and NI treatments (Fig. 1). Previous studies in invertebrates have shown that, within species, microbiota transplants of individuals that were acclimated to warmer temperatures promoted enhanced heat tolerance in the recipient (Baldassarre et al., 2022; Doering et al., 2021; Moghadam et al., 2018). Such findings indicate that temperature-based restructuring of the gut microbiota selects for bacterial taxa and/or functions that improve survival under heat stress. For example, Fontaine and Kohl (2023) found that the gut microbiome of larval green frogs exposed to a 24 h period of heat stress significantly upregulated genes associated with carbohydrate metabolism, transcription and translation. Surprisingly, they did not find that the heat stress altered bacterial expression of heat shock proteins (HSPs) within the gut microbiota, which are commonly upregulated in response to thermally stressful environments to maintain protein structure integrity (Feder and Hofmann, 1999). This suggests that the gut microbiota improves host heat tolerance through increased metabolic function and DNA processes rather than bacterial HSP expression, but Fontaine and Kohl (2023) did not measure intestinal HSP expression, which have been shown to be modulated by the gut microbiota (Arnal and Lalles, 2016). We lack transcriptomic data to explore how the GF treatment promoted greater CT_max relative to the other treatments; therefore, our conclusions are limited to taxonomic differences we observed rather than altered metabolite and gene expression. Inclusion of metagenomic and transcriptomic analyses of the gut microbiota, along with expression of intestinal HSPs, in studies assessing heat tolerance would be the advisable next step in assessing the functional role the gut microbiota has in influencing host CT_max. Regardless, our findings are in line with other studies showing benefits of cross-species microbiota transplants in species conservation (Dallas and Warne, 2023) and could improve survival for threatened species under warming conditions.

There were several bacterial taxonomic groups that differed among the microbiota treatments. Most prominently, the Bacteroidota family Rikenellaceae, primarily Mucinivorans hirudinis, was significantly enriched in GF larvae. This species was initially described in leeches (Nelson et al., 2015), and produces the short-chain fatty acid acetate as a byproduct of metabolizing host mucin glycans (Bomar et al., 2011). Its prevalence in other taxa has not received much attention, but Mucinivorans species has been previously described in late-stage Asiatic toads (Bufo gargarizans) (Chai et al., 2018), suggesting this bacterial species is not limited to leeches. Other members of Rikenellaceae promote beneficial phenotypes such as enhanced heat tolerance in cows (Wang et al., 2022), improved health of late-stage giant spiny frogs (Paa spinosa) (Long et al., 2020) and lower obesity rates in humans (Pesoa et al., 2021). One potential mechanism through which M. hirudinis could promote heat tolerance is through upregulation of antioxidants as a byproduct of acetate production (Gonzalez-Bosch et al., 2021; Warne and Dallas, 2022). Heat stress is associated with an increased production of reactive oxygen species (ROS) in vertebrates that induce oxidative damage, which negatively affects cellular function and host survival, and can limit host CT_max (reviewed in Ritchie and Friesen, 2022). Acetate was shown to significantly reduce ROS production in human and mouse cells in vitro while promoting greater mitochondrial respiration rates (Hu et al., 2020; Huang et al., 2017). The beneficial role of acetate in promoting antioxidant concentrations has also been demonstrated in vivo using piglets (Pang et al., 2021), but direct links between M. hirudinis and antioxidant production or ROS reduction are not present. While the link between Rikenellaceae (M. hirudinis) and greater heat tolerance is not definitive, its production of acetate could represent a pathway by which GF larvae benefited from its increased relative abundance.

While the WF and NI larvae had similar relative abundance of bacterial families, the latter displayed a higher relative abundance of Aeromonadaceae, specifically those in the Aeromonas genus. Members of this family are Gram-negative bacteria that are potential pathogens of amphibians and fish (Janda and Abbott, 2010). The NI larvae received no bacterial inoculation, which means that upon hatching there would be a limited gut microbiota to become established. Resultantly, potentially pathogenic bacteria in the environment, including Aeromonas, and other transient species could readily colonize the gut as a consequence of a lack of bacterial competition (Buffie and Pamer, 2013). Aeromonas can reach relatively high abundance in the intestines of wild (Hird et al., 1983) and lab-reared anuran larvae (Chai et al., 2018). In a study on wild-collected Dybowski's frog (Rana dybowskii), Tong et al. (2020) found that the abundance of Aeromonas was enriched in diarrheic adults compared with healthy adults and they posited the presence of this genus led to the emergence of a pathogenic phenotype. While we did not observe direct evidence of such a phenotype, a reduction in the heat tolerance of NI larvae may be related to higher Aeromonas abundance. A potential mechanism for this outcome is through Aeromonas-induced systemic inflammation. As Aeromonas metabolites degrade the intestinal barrier (Dong et al., 2018; Feng et al., 2022), this could enable Gram-negative bacteria with their lipopolysaccharide (LPS) membrane to enter the bloodstream of the host. LPS promotes inflammation through increased production of pro-inflammatory cytokines and leukocyte activity (Boulange et al., 2016; Kamada et al., 2013; Wang et al., 2020). As an activated immune response can cause a decline in CT_max (Hector et al., 2020), the increased Aeromonadaceae abundance in NI larvae may have indicated increased immune costs. This may be indicative that the host could incur negative costs if Aeromonadaceae are enriched in the larval intestine, but further exploration of the topic is required to address this potential relationship.

In addition to increased Aeromonadaceae abundance, the work of Warne et al. (2019) and Fontaine et al. (2022) offers a potential explanation why NI larvae had the lowest CT_max. These studies demonstrated that larvae experiencing early-life bacterial depletion had lower mass-specific metabolic rates and reduced mitochondrial enzymatic activity, respectively. Fontaine et al. (2022) predicted that larvae with a diminished aerobic scope would have a constrained heat tolerance (Chung and Schulte, 2020; Pörtner et al., 2017), although the importance of such a link has been debated (e.g. Claunch et al., 2021; Ern et al., 2016; Semsar-Kazerouni and Verberk, 2018). Microbial derived short-chain fatty acids are also associated with increased mitochondrial activity (Schonfeld and Wojtczak, 2016), which may constrain the host's ability to maintain ATP production at high temperatures, resulting in a lower CT_max (Sokolova, 2023). Compared with GF larvae, NI larvae had a lower, although not statistically significant, relative abundance of Firmicutes and Bacteroidota, two phyla associated with short-chain fatty acid production (den Besten et al., 2013; Warne and Dallas, 2022). This could account for our observed CT_max pattern. Future studies should examine this link through direct supplementation of short-chain fatty acids and then measure the host's heat tolerance, or by measuring CT_max and short-chain fatty acid concentrations in the intestines. In addition, identifying whether changes in the gut microbiota composition affect a host's aerobic scope would determine whether changes in CT­_max are related to whole-body changes in metabolic rate.

While prolonged acclimation to different thermal regimes alters gut microbiota α- and β-diversity (Baldassarre et al., 2022; Fontaine et al., 2018; Moeller et al., 2020; Theys et al., 2023; Zhu et al., 2021), our short-term acclimation exposure had mixed effects on these metrics. While both bacterial membership and structure were dependent upon acclimation temperature (Fig. 3), only two metrics of α-diversity (Shannon and Simpson indices) displayed a small, temperature-dependent increase (Fig. 2B,C). A similar pattern was observed in larval Ischnura elegans damselflies exposed to a simulated 7 day heatwave (Theys et al., 2023). The temperature effect in the damselflies was primarily driven by increases in Proteobacteria, a pattern commonly observed in arthropods experiencing warmer temperatures (Sepulveda and Moeller, 2020). In a study on green frog larvae, Fontaine et al. (2022) showed a minor decline in α-diversity with increasing acclimation temperature, but bacterial composition did diverge, with the relative abundance of Proteobacteria negatively related to increasing temperature, although the effect was non-significant. This demonstrates the complex relationship between environmental temperature and the gut microbiota. In line with the predictions of Lozupone et al. (2012) and concept of disturbance ecology (Christian et al., 2015), we propose that exposure to different environmental temperatures represents a disturbance to the gut microbiota that causes a shift in bacterial diversity, composition and transcriptomics. While a rapid temperature shift can drastically alter gut microbiota transcriptomics of ectothermic vertebrates (Fontaine and Kohl, 2023; Zhou et al., 2022), community-wide changes are likely more pronounced following prolonged exposure to new temperatures, enabling a new community to become established (but see Fontaine and Kohl, 2020). Longitudinal studies measuring gut microbiota diversity metrics exposed to different acclimation temperatures (Baldassarre et al., 2022) would be beneficial in addressing this prediction.

We found that high temperature-acclimated wood frog larvae were depleted in the Actinobacteria phylum while the Proteobacteria family Beijerinckiaceae (Methylobacterium-Methylorubrum sp.) was enriched. The relative abundance of Actinobacteria in the gut microbiota, members of which are associated with antibiotic production (Bérdy, 2012), has been shown to be negatively impacted by exposure to warm temperatures in arthropods (Horváthová et al., 2019) and Mongolia racerunners (Eremias argus) (Zhang et al., 2022). However, the opposite was found in lab-reared northern leopard frog larvae (Lithobates pipiens) (Kohl and Yahn, 2016) and Chinese giant salamanders (Andrias davidianus) (Zhu et al., 2021). The pronounced increase in Beijerinckiaceae under the high temperature acclimation was surprising as members of this family have the capacity to grow across a broad thermal breadth in natural environments (Sharp et al., 2014). In previous studies on free-ranging Nile tilapia (Oreochromis niloticus) and northern leopard frog larvae, the relative abundance of Beijerinckiaceae (Methylobacterium spp.) was enriched at cool temperatures (Bereded et al., 2022; Kohl and Yahn, 2016), while pinfish (Lagodon rhomboides) had higher relative abundance at warmer temperatures (Givens, 2012). Many Beijerinckiaceae members are associated with nitrogen fixing (Marin and Arahal, 2014), which can provide their host with additional energy (Russell et al., 2009), although how this influences larval anuran performance or physiology requires further assessment. Our results show that complex relationships exist between environmental temperature and the relative abundance of Actinobacteria and Beijerinckiaceae that are likely host specific.

We acknowledge, however, that our speculations come with the caveat of small sample sizes for gut microbial sequencing (n=4 per microbiome treatment×acclimation temperature). These low sample sizes likely limit statistical power and our ability to properly discern subtle treatment differences in gut microbiota as a result of high within group variation. Follow-up research requires larger sample sizes to reduce variation and strengthen the conclusions of how cross-species microbiome transplants influence the gut microbiota of the recipients.

Another limitation of our study design is that we did not have an unmanipulated microbiota treatment that represents wildtype wood frog larvae. The WF treatment was meant to restore the wildtype microbiota following antibiotic exposure, but a subsequent study showed that unmanipulated wildtype larvae significantly diverged in both α- and β-diversity indices from WF larvae (J.W.D., A. Macedo, A.K., S.T.M.L. and R.W.W., unpublished observations). Therefore, the WF larvae did not accurately represent the wildtype condition. While GF larvae exhibited the highest CT_max among our microbiota treatments, their CT_max was comparable to that of less-developed wildtype larvae (Fig. S3). This indicates that transplanting the green frog microbiota improved heat tolerance of recipient larvae relative to the other antibiotic-treated larvae but not unmanipulated individuals. Notably, early-life antibiotic exposure in anurans leads to phenotypic costs including higher mortality risk, delayed metamorphosis and increased tail deformities (Peltzer et al., 2017; Warne et al., 2019; Zhu et al., 2022), suggesting that the green frog transplantation rescued antibiotic-treated larvae from the lower heat tolerance evident in the NI and WF larvae. A follow-up study incorporating a more successful reinoculation of the wildtype microbiome would provide greater support for linking cross-species gut microbiota transplants to warmer CT_max.

As the risk of extreme heat events becomes more prevalent under climate change, how the gut microbiota modulates host thermal physiology represents an intriguing research question. Our results represent an important step in linking this relationship by demonstrating that inoculating antibiotic-treated wood frog larvae, a heat-sensitive species, with the gut microbiota of green frog larvae, a more heat-tolerant species, promoted higher CT_max relative to that of other antibiotic-treated larvae. This suggests that interspecific differences in heat tolerance are partially driven by the gut microbiota, and that this difference can be transferred between species simply through microbial transplants. The green frog microbiota transplant resulted in large increases in the Bacteroidota family Rikenellaceae, a producer of the short-chain fatty acid acetate, which is linked to increased antioxidant activity (Gonzalez-Bosch et al., 2021). As heat shock increases the production of ROS that disrupt cellular and mitochondrial activities and so restrict heat tolerance, a greater abundance of antioxidants can likely help to maintain cell function at high temperatures. However, we acknowledge that our results are correlative but the inclusion of metagenomics to assess a mechanistic link between the gut microbiota and heat tolerance represents a future direction to explore. Microbial transplants represent a species-rescue technique against multiple threats, and this study indicates that cross-species transplants can rapidly induce greater heat tolerance, which promotes survival under warming temperatures.

We would like to thank Jacob Hutton for his assistance in collecting the larval green frogs used in the microbiome treatment, and both Jacob and Adrian Macedo for their assistance in the CT_max assays. We would also like to thank Alina Akhunova of the Kansas State Integrated Genomics Facility for help with 16S amplicon sequencing. We also wish to acknowledge Dr Frank Anderson and Dr Derek Fisher for their comments on previous versions of the manuscript. We also thank the two anonymous reviewers for their insightful comments to improve this manuscript.

This article has an associated ECR Spotlight interview with Jason Dallas.