Temperature and the microbial environment alter brain morphology in a larval amphibian

As the global climate continues to change, understanding how ecologically relevant shifts in temperature affect wildlife animal physiology remains pertinent. Perhaps no vertebrate group has been challenged by climate change more than ectotherms such as amphibians, as their biodiversity and global populations have been exponentially decreasing (Pounds et al., 2006). Temperature is particularly relevant for amphibians, as their body temperature largely tracks ambient temperature. Fluctuating ambient temperatures modulate numerous aspects of amphibian physiology, including metabolic rate, developmental rate, locomotion and digestive efficiency (Feder and Burggren, 1992; Fontaine et al., 2018; Gillooly et al., 2002; Huey and Stevenson, 1979). Interestingly, a literature review posited that rising temperatures stimulate neuronal activity in ectotherm brains up to a thermal threshold, beyond which it is detrimental to neuron development and nervous tissue (Beltrán et al., 2021). However, many of these temperature effects on neurodevelopment are likely species dependent and we currently lack a complete understanding of how ecologically relevant increases in temperature influence brain development across ectotherms (Beltrán et al., 2021).

Environmental temperature impacts the diversity and composition of host-associated microbial communities harbored by both aquatic and terrestrial amphibians, which can impact their physiology (Fontaine and Kohl, 2020; Fontaine et al., 2018; Kohl and Yahn, 2016). A particularly large microbial community composed of bacteria, archaea, fungi and protists resides in the gastrointestinal (GI) tract. Called the gut microbiota, this host-associated microbial community participates in a symbiotic relationship with the host and impacts host performance, health and fitness (Kohl and Carey, 2016). In species with external fertilization, like most amphibians and fish, the gut microbiota is likely colonized by microorganisms present in their aquatic habitat immediately upon hatching, just a few days after fertilization (Correa et al., 2020 preprint). Although the composition of the gut microbiota is malleable throughout development, its trajectory throughout ontogeny is heavily dependent on these initial microbial colonizers (Litvak and Bäumler, 2019; Martínez et al., 2018; Warne et al., 2019). Manipulation of the gut microbiota in early life through exposure to antibiotics or sterilized environments impacts host growth and immunity (Knutie et al., 2017; Warne et al., 2019). Often, these impairments can persist throughout life, emphasizing the importance of the gut microbial colonization process in vertebrate development (Cox et al., 2014).

Gut microbiomes specifically contribute to the development of the central and peripheral nervous systems through the microbiota–gut–brain (MGB) axis (Foster and McVey Neufeld, 2013; Perry et al., 2016). Gut microbes produce molecules such as neurotransmitters, short chain fatty acids, secondary bile acids and peptidoglycans that can stimulate peripheral sensory neurons of the vagus nerve, the enteric nervous system and the spinal cord (for a comprehensive review, see Cryan et al., 2019). Further, the gut microbiota and its products can also modulate non-neural routes of communication within the MGB axis such as the hypothalamic-pituitary–adrenal (HPA) axis, which coordinates the vertebrate stress response (Cryan et al., 2019).

The importance of the MGB axis is most revealed in studies using germ-free animals that lack a gut microbiota. Germ-free mice exhibit altered locomotory behavior and long-term synaptic gene expression in brain regions involved in motor control and anxiety (Heijtz et al., 2011). In addition, germ-free mice had increased amygdalar and hippocampal volume, which are brain regions implicated in fear and anxiety (Luczynski et al., 2016; Tovote et al., 2015). Of non-mammalian vertebrates, fish host–microbe interactions are relatively well studied because of their importance in aquaculture, and germ-free zebrafish are a common laboratory-based model. Zebrafish have ex utero development, which allows investigators to manipulate microbial exposure more easily throughout all stages of development (Davis et al., 2016; Lee et al., 2021). Zebrafish raised in germ-free conditions exhibited altered locomotory and anxiety-like behavior as well as altered stress responses compared with conventionally raised zebrafish, demonstrating a potential link between the microbiota and neurodevelopment (Davis et al., 2016). Additionally, adult zebrafish exposed to beneficial intestinal microbes (i.e. probiotics) exhibited reduced anxiety-like behavior which was accompanied by altered neurotransmitter signaling in the brain (Davis et al., 2016). These effects of intestinal microbes on anxiety and locomotory behavior in fish could potentially regulate feeding behavior and energy homeostasis, which would have significant ramifications on host fitness (Butt and Volkoff, 2019; Xia et al., 2022).

While the use of germ-free mammals and fish has advanced our understanding of host–microbe interactions, study of other vertebrate groups has lagged, especially in more ecological contexts (Colston and Jackson, 2016; Pascoe et al., 2017; Woodhams et al., 2020). For example, correlational evidence for the MGB axis in songbirds was only recently reported (Slevin et al., 2020). Likewise, in amphibians, a few studies have found limited evidence for a link between the gut microbiota and the brain. In captive-bred Chinese giant salamanders, exposure to sterile soil from their natural habitat altered gene expression related to energy metabolism in the brain compared with exposure to non-sterile soil (Zhu et al., 2022). In ranid frogs, exposure to an insecticide caused a shift in the composition of the gut microbiota and altered neurotransmitter production in the gut (Zhang et al., 2023).

To better understand the role of host microbiota in brain development in amphibians, we tested the hypothesis that the temperature and the microbial environment in which larval amphibians were raised shapes neurodevelopment, possibly through the effects on the MGB axis. To do this, we took advantage of a previous study where newly hatched green frog tadpoles (Lithobates clamitans) were raised in varying water temperatures; further, tadpoles were raised in natural pond water, or in pond water that was autoclaved to deplete the environmental microbial community available to colonize the gut (Fontaine et al., 2022). This environmental water sterilization resulted in tadpole gut bacterial communities that were reduced in diversity and altered in composition, but similar in overall bacterial density, when compared with that of tadpoles colonized with microbes from natural pond water (Fontaine et al., 2022). Using autoclaved water to manipulate the host-associated microbiota is a less invasive method to reduce the diversity and abundance of colonizing microbes compared with using antimicrobials, which have undesirable effects on host physiology (Fontaine et al., 2022; Knutie et al., 2017; Morgun et al., 2015; Patangia et al., 2022). Additionally, as noted above, tadpoles raised in autoclaved water still harbor a gut microbial community with indistinguishable bacterial density from that of tadpoles raised in pond water. Thus, our experimental groups compare tadpoles with varying microbial communities, rather than using the highly artificial state of germ-free systems. Specifically, we used the specimens from ‘Experiment 1’ of Fontaine et al. (2022) to examine the brains of tadpoles and test the predictions that (1) water temperature and the local microbial environment would alter relative brain size and morphology, and (2) the gut microbial diversity and composition would predict relative brain size and morphology.

Here, we present brief methodological details of animal husbandry and conditions. For detailed information on collection, rearing, treatment groups and microbiome sequencing, see Fontaine et al. (2022). All animal research was approved by the University of Pittsburgh IACUC (protocol #18062782) and animal collections were permitted by LA Department of Wildlife and Fisheries (Scientific Collecting Permit WDP-19-010).

In May 2019, a male and female adult green frog, Lithobates clamitans (Latreille 1801), were collected from a pond in Kisatchie National Forest (LA, USA) and transported to the University of Pittsburgh. In September 2019, the frogs were injected with hormones to induce spawning following methods outlined in Trudeau et al. (2010). Embryos were placed in a 16-quart (∼15 l) polypropylene tank containing autoclaved laboratory-treated water that was changed daily to ensure proper oxygenation. Tadpoles were fed weekly. Food consisted of three 0.5 g blocks of autoclaved rabbit chow suspended in autoclaved agar and supplemented with pet vitamins.

Free-swimming tadpoles (Gosner stage 25; Gosner, 1960) were raised in either 25% unmanipulated pond water or 25% autoclaved pond water at one of three temperatures, described below. The other 75% of water in each treatment consisted of autoclaved laboratory-treated water. Pond water was collected in June 2019 from the same pond from which the adult frogs were collected. Pond water was placed on ice and transported back to the University of Pittsburgh where it was filtered through a 500 µm sieve and stored at 4°C until use. Details on the microbial community of stored pond water versus pond water fresh from the pond are available in Fontaine et al. (2022). Tadpoles were raised in groups of 5 tadpoles in 900 ml of water in a 1 liter polypropylene bin. Eighteen bins were filled with the natural microbial environment and 18 were filled with the autoclaved microbial environment. Water treatments were re-administered during weekly water changes and were maintained for the duration of the experiment.

For the first 4 weeks of development, tadpoles were raised at 22°C by placing bins in water baths set to 22°C. There were three water baths, each with 12 bins, six each from the previously described microbial water treatments. After 4 weeks at 22°C, tadpoles were exposed to different acclimation temperatures. To do so, one of the three water baths was decreased to 14°C, and another water bath was increased to 28°C. The third water bath was maintained at 22°C. Tadpoles developed at these temperatures for an additional 3 weeks.

Each treatment combination (temperature treatment×microbial environment) consisted of six 1 liter polypropylene bins with five tadpoles per bin, for a total of 30 tadpoles per treatment combination. Final sample sizes were slightly lower than 30 tadpoles because of a small amount of mortality.

Seven weeks after the start of the experiment (November 2019), tadpoles were euthanized by immersion in buffered MS-222 (10 g l⁻¹) and then were weighed and staged (Gosner, 1960). Using sterilized tools, the entire gastrointestinal tract was removed within 10 min of euthanasia and immediately placed at −80°C. Carcasses were placed in 10% neutral buffered formalin.

In July–August 2020, brains were removed from carcasses using an Olympus SZ61 dissection scope with a camera attachment. Cranial nerves and the spinal cord were trimmed from the brain, brains were weighed, and the dorsal and ventral views of each brain were photographed three times for a total of six images per brain. In between taking each photograph, the brain was repositioned to produce three unique dorsal and ventral images. If damage to the brain occurred during dissection that could alter brain mass or brain shape measurements, the brain was excluded from analysis. The same investigator conducted all brain dissections and photography.

Tissues are known to shrink with time in fixative, so we ensured that time in fixative was the same among our treatment groups by completing all the weighing and imaging of the brains within 12 days. Furthermore, we processed the samples in blocks that consisted of representatives from each treatment group. Within each block, samples were randomized so that the investigator remained blind to treatments.

We used geometric morphometrics to evaluate brain shape (Adams et al., 2004). Four linear brain dimensions on each dorsal image and three linear brain dimensions on each ventral image were measured using ImageJ software (US National Institutes of Health, Bethesda, MD, USA) (Fig. 1). Each linear brain dimension was measured once from each of the three photographs, giving three measurements in total, which were averaged to give a single estimate for each brain dimension for each tadpole. The same investigator conducted all measurements.

We measured the length and width of the telencephalon, diencephalon and optic tectum, and we measured the width of the medulla. The telencephalon is involved in sensory processing, motor output, avoidance learning and social behavior (Northcutt, 1981). The diencephalon is implicated in homeostasis and endocrine function, and contains sensory neurons vital to the vertebrate stress response (Charmandari et al., 2005; Denver, 2009). The optic tectum processes visually guided motor behaviors in amphibians (Bestman et al., 2012). The medulla is implicated in respiration and other autonomic functions (Gdovin et al., 1999).

Total DNA was extracted from all gut samples using a QIAamp PowerFecal Pro DNA isolation kit (Qiagen) following kit directions. Extracted DNA was stored at −20°C until it was sent to the University of Illinois at Chicago's DNA Services Facility for library preparation, PCR and sequencing. The bacterial 16S rRNA gene was amplified and amplicons were sequenced on the Illumina Miseq platform. Raw sequence data were processed and analyzed using QIIME2 v2019.7 (Bolyen et al., 2019). Full microbiome sequencing methods, statistical analyses and results demonstrating relationships between temperature and microbial environment and bacterial community alpha diversity, beta diversity, bacterial taxa abundance and bacterial load are available in Fontaine et al. (2022).

Statistical analyses were performed in RStudio (v4.1.0) and using the lme4 package (Bates et al., 2015). All data analyzed with parametric statistics met assumptions of normal distribution and homogeneity of variance unless otherwise noted. Data were transformed to meet assumptions in some cases. Our criterion for statistical significance was P=0.05.

Because tadpole body mass varied as a result of temperature and microbial environment (Fontaine et al., 2022), we adjusted brain mass measurements for differences in body mass. We used an analysis of covariance (ANCOVA) with temperature and microbial environment as main effects, and body mass as a covariate. We confirmed that the slopes of the lines for brain mass were parallel across treatment groups by demonstrating a non-significant interaction between treatment and body mass. To calculate a measure for brain mass that was adjusted for body mass, the unstandardized brain mass residual value for each animal (from the ANCOVA) was added to the overall estimated marginal mean (EMM). To assess the effects of temperature, microbial environment and their interaction on our mass-adjusted brain mass measurements, we used generalized linear mixed models (GLMMs) including bin as a random effect.

Because linear brain dimensions covary with brain mass, we corrected brain dimensions for brain mass (McCoy et al., 2006). To do this, we used a multivariate analysis of covariance (MANCOVA) with temperature and microbial environment as main effects, and brain mass as a covariate. We confirmed that the slopes describing the relationship between brain mass and each brain dimension were parallel across each treatment group by demonstrating a non-significant interaction between treatments and brain mass. This was the case for all dimensions except optic tectum length and diencephalon length. Unstandardized residuals generated by the MANCOVA were added to the overall EMM to get a mass-adjusted value for each brain dimension for each tadpole.

Because mass-adjusted brain dimensions were highly correlated, we used a principal component analysis (PCA) with a varimax rotation to obtain uncorrelated principal components (PCs). Assumptions of the PCA were met, with Kaiser–Meyer–Olkin (KMO) >0.5 and Bartlett's tests ≤0.05. The PCA yielded three PCs with eigenvalues >1. The effects of temperature, microbial environment and their interaction on PCs were assessed using GLMMs, with bin as a random effect.

To further understand the effects of our treatments on the gut microbiota and brain development, we explored whether individual differences in the diversity of the bacterial gut microbiota predicted differences in brain mass and morphology. To do so, we used GLMMs that included bin as a random effect, with a measure of alpha diversity as predictor variables and brain mass or morphometric values as the response variable. We examined four alpha diversity metrics: the number of observed bacterial amplicon sequence variants (ASVs), Shannon diversity (Shannon, 2001), Faith's phylogenetic diversity (Faith, 1992) and Pielou's evenness (Pielou, 1966).

We were also interested in testing whether measurements of relative brain mass and shape were correlated with the relative abundance of specific gut bacterial taxa at the phylum and genus level. For this analysis, we used RStudio (v4.1.0) and the MaAsLin2 package (microbiome multivariable association with linear models; Mallick et al., 2021). P-values were corrected using the Benjamin–Hochberg false discovery rate (BH FDR) method.

Tadpoles raised at warmer temperatures had relatively larger brains (adjusted for body mass) compared with tadpoles raised at cooler temperatures (Fig. 2; χ²=22.8, P<0.001). For example, tadpoles raised at 28°C had relative brain masses that were roughly 20% larger than those raised at 14°C. There was no effect of microbial environment, and no interaction between temperature and microbial environment, on relative brain mass (Fig. 2).

The PCA of seven brain dimensions (adjusted for brain mass) yielded three PCs with eigenvalues greater than 1 (Table 1). PC1 loaded strongly with telencephalon width, telencephalon length, diencephalon width and medulla width. PC2 loaded strongly with optic tectum width and optic tectum length. PC3 loaded strongly with diencephalon length.

Relative brain dimensions described by PC1 were not affected by temperature or microbial environment, and there was no interaction between treatments (Table 1, Fig. 3A).

Relative brain dimensions described by PC2 were affected by temperature and microbial environment, but there was no interaction between treatments (Table 1). Specifically, tadpoles raised in warmer temperatures and autoclaved pond water had increased relative optic tectum width and length compared with tadpoles raised in cooler temperatures and natural pond water (Table 1, Fig. 3B). To better understand the change in PC2, we analyzed optic tectum width and length separately, and found that both were altered by temperature and microbial environment (Fig. S1).

There was no main effect of temperature or microbial environment on relative brain dimensions described by PC3, but there was a significant interaction between these variables (Table 1). Specifically, the effect of the microbial environment on diencephalon length depended on the temperature at which tadpoles were raised. At the intermediate temperature 22°C, the length of the diencephalon was shorter in tadpoles raised in autoclaved pond water versus natural water, but not at the extreme temperatures (14°C or 28°C) (Table 1, Fig. 3C).

Two measures of alpha diversity (observed ASVs and Faith's phylogenetic diversity) were predictors of the relative brain dimensions described by PC2 (optic tectum width and length) (Table 2, Figs 4 and 5; Table S1). Specifically, tadpoles that harbored more diverse gut microbial communities in terms of the number of different bacterial taxa present had narrower and shorter optic tecta. Additionally, Pielou's evenness was a predictor of relative brain dimensions described by PC3 (diencephalon length) (Table 2; Table S1). There was no relationship between alpha diversity metrics and relative brain mass and relative brain dimensions described by PC1 (Table 2; Table S1).

At the phylum taxonomic level, relative brain dimensions described by PC2 (optic tectum width and length) were associated with the relative abundance of 13 bacterial phyla (Table 3). Of the 13 phyla, tadpoles with a higher abundance of Firmicutes tended to have optic tecta with increased width and length, and this positive association is described in Table 3. In contrast, the relative abundance of the other 12 phyla had negative associations with optic tectum width and length (Table 3). Relative brain mass and brain dimensions described by PC1 and PC3 were not associated with the relative abundance of any bacterial phyla.

At the genus level, relative brain mass was associated with the relative abundance of five bacterial genera (Table S2). Additionally, relative brain dimensions described by PC2 (optic tectum width and length) were associated with the relative abundance of 41 bacterial genera (Table S2). Relative brain dimensions described by PC1 and PC3 were not associated with the relative abundance of any bacterial genera.

Here, we tested the effects of temperature and the microbial environment on neurodevelopment of green frog tadpoles, extending the results of a previous study (Fontaine et al., 2022). These tadpoles were raised in natural pond water or autoclaved pond water, as well as at three different temperatures. Tadpoles raised in warmer temperatures had a less diverse gut microbiota that varied in its composition compared with that of tadpoles raised in cooler temperatures (Fontaine et al., 2022). Similarly, tadpoles raised in autoclaved pond water had a less diverse and distinct gut microbiota compared with that of tadpoles that developed in natural pond water (Fontaine et al., 2022). We extend the results of Fontaine et al. (2022) by showing that (1) water temperature, but not the microbial environment, affected relative brain mass, (2) both the water temperature and the microbial environment affected relative optic tectum length and width, (3) there was an interactive effect between temperature and the microbial environment on relative diencephalon length, (4) metrics of alpha diversity predicted changes in relative brain shape, and (5) the relative abundance of several bacteria taxa, at the phylum and genus levels, was correlated with changes in relative brain mass and shape. Thus, water temperature and the microbial environment altered amphibian neurodevelopment, perhaps through changes to the gut microbiota and subsequently the MGB axis. Below, we discuss our results in more detail.

Brain mass was influenced by water temperature, such that tadpoles raised in warmer water had heavier brains, relative to body mass, than tadpoles raised in cooler water. As reported in Fontaine et al. (2022), the tadpoles that were raised in warmer temperatures also attained a larger body mass than those exposed to cooler temperatures. Thus, in addition to a larger overall body mass, the brain was proportionally larger with exposure to warmer temperatures.

It is likely that temperature-dependent alterations to host metabolism contributed to the larger brains seen in tadpoles raised at warmer temperatures. Increased sublethal temperatures enhance ectothermic growth and development (Goldstein et al., 2017; Marian and Pandian, 1985) as a result of increased metabolic rate (Feder and Burggren, 1992; Fontaine et al., 2022; Gillooly et al., 2001). Further, increased sublethal temperatures influence ectothermic digestive performance and energy assimilation, and increase appetite (Fontaine et al., 2018; McConnachie and Alexander, 2004). Interestingly, increasing temperatures have been shown to modulate tadpole feeding preferences such that they shift to a more herbivorous diet because of enhanced digestion of plant material at high temperatures compared with carnivorous diets that are high in lipids and proteins (Carreira et al., 2016). Although we did not measure food intake in our study and diet was consistent across all treatment groups, investigating tadpole feeding preferences and appetite in the context of climate change would be an interesting future direction to better understand our results.

Increased metabolic rate and energy uptake as a result of higher, non-lethal temperatures could result in tadpoles developing and maintaining larger brains, as nervous tissue is energetically expensive and has been shown to account for 2–10% of vertebrate metabolic output, despite nervous tissue representing a much smaller percentage of total body weight (Aiello and Wheeler, 1995; Mink et al., 1981). Whether this increase in relative brain size has cognitive benefits for the host is beyond the scope of this study, but increased size of the brain and larger body–brain ratios are associated with increased cognition in numerous animal taxa (Roth and Dicke, 2005).

In addition to increasing relative brain mass, we found that tadpoles raised in warmer water temperatures had relatively wider and longer optic tecta compared with tadpoles raised at cooler temperatures. It is important to note that these measurements of optic tectum size were corrected for relative brain mass, and changes in the size of this region were not driving the changes observed in overall brain size. A similar result to this was seen in minnows, such that minnows raised in warmer water temperatures had larger overall brains and larger relative medullas compared with minnows raised in cooler water temperatures (Závorka et al., 2020). As with relative brain mass, we suggest that changes in the optic tectum size could be related to temperature-induced increases in metabolic rate (Fontaine et al., 2022). We also found that tadpoles raised in autoclaved pond water had relatively wider and longer optic tecta compared with tadpoles raised in natural pond water. These changes could also be attributed to host metabolism, as other studies have found that newly hatched tadpoles that developed in sterilized water exhibited altered metabolism and growth rates, although this result may vary across different species (Warne et al., 2019).

Because of the energetic cost to maintain and develop nervous tissue, any changes in brain morphology are expected to be adaptive responses to the external environment (Aiello and Wheeler, 1995; Gonda et al., 2013). In particular, we expect that the changes in brain mass and shape induced by our treatments contribute to changes in behavior. For example, pesticide-induced changes in brain shape altered behavioral responses to novel visual stimuli in northern leopard frog tadpoles (Lithobates pipiens) (McClelland and Woodley, 2022). Because the optic tectum processes visually guided motor behaviors in amphibians (Bestman et al., 2012; González et al., 2020), the changes in the optic tectum found in our study could alter the ability of tadpoles to evade predators and capture prey, and affect their feeding behavior. For example, tadpoles raised in autoclaved pond water that had altered optic tectum width and length had slower escape responses when prodded by a blunt probe, although only after exposure to higher water temperatures (Fontaine et al., 2022). An intriguing possibility is that variation in the thermal and microbial environment in which tadpoles develop may contribute to natural behavioral variability of these aquatic animals through changes in the MGB axis. Future studies could incorporate behavioral assays to test whether variations in the thermal and/or microbial environment influence tadpole anti-predator and foraging behaviors.

In addition to the effect on optic tectum shape, our microbial treatment altered diencephalon length, but only at the intermediate temperature of 22°C. While the results are difficult to disentangle, alterations to diencephalon structure could alter the host stress response. Specifically, the vertebrate diencephalon contains neurons that produce corticotropin releasing hormones, which activate the HPA axis so that glucocorticoids are released into the circulation (Charmandari et al., 2005; Denver, 2009). Increased circulation of glucocorticoids has been associated with changes in brain morphology in amphibians. Specifically, exposure of northern leopard frog tadpoles to glucocorticoids in the water resulted in wider diencephalons compared with control (Cha et al., 2021). Some studies have shown that gut microbial diversity is associated with physiological markers of stress such as glucocorticoids, and this relationship appears to be present across vertebrates (Stothart et al., 2016). For example, germ-free zebrafish had a blunted stress response compared with conventionally raised zebrafish (Davis et al., 2016). Glucocorticoids were not measured here, but future studies could measure physiological markers of stress to further understand the relationship between the gut microbiota and the vertebrate stress response.

To further understand the effects of our treatments on the gut microbiota and brain development, we explored whether individual differences in the diversity of the gut microbiota predicted differences in brain mass and morphology. We found that relative optic tectum width and length were negatively associated with the number of observed ASVs and Faith's phylogenetic diversity. Specifically, a higher diversity of gut microbial communities in terms of the number of observed bacterial taxa tended to predict decreases in relative width and length of the tadpole optic tectum. Gut microbial diversity and composition have been associated with mammalian regional brain volume, as well as mammalian and avian cognitive ability (Carlson et al., 2018; Labus et al., 2017; Slevin et al., 2020), but the cellular causes driving these changes in brain morphology are unknown. Germ-free mice exhibit increased volume of brain regions such as the amygdala, hippocampus and forebrain compared with control mice, which is hypothesized to be due to neuronal remodeling that could alter their downstream stress responsivity and behavior (Cryan et al., 2019; Luczynski et al., 2016).

Alternatively, changes in brain morphology could be due to altered microglia morphology. The gut microbiota modulates the development of microglia, which are macrophages that serve as the main form of immune defense in the CNS to resolve threats to host health, such as neuroinflammation (Cryan et al., 2019; Erny et al., 2015). Germ-free mice exhibit increased proliferation of branched microglia with elongated processes, and their forebrains were 17% larger than those of control mice (Castillo-Ruiz et al., 2018). Future studies examining genetic, molecular and cellular differences in brains from tadpoles raised in different microbial environments are warranted.

To further understand the effects of our treatments on brain shape and the gut microbiota, we explored whether individual differences in brain shape were correlated with the relative abundance of specific bacterial taxa. While these results are correlational, they begin to give a picture of the wide-ranging interactions between hosts and the microbiome, especially in non-model systems. Making inferences as to the functional consequences of the microbiome (and specific microbial taxa within the microbiome) for host biology remains an important contribution and allows for downstream meta-analyses (Alberdi et al., 2021; Li et al., 2023).

Some of the bacterial phyla we found to be correlated with optic tectum shape have been previously linked to the nervous system. One example is the phylum Chlamydiae, which was more abundant in tadpoles that developed in natural pond water and was negatively correlated with optic tectum width and length (Fontaine et al., 2022). This phylum of bacteria has been shown to exploit host resources and elicit immune responses, and proliferation of this pathogenic bacteria has been seen in mammals experiencing neurodegenerative diseases such as Alzheimer's disease (Balin et al., 1998; Collingro et al., 2020; Gitsels et al., 2019). Similarly, we found that the phylum Cyanobacteria was more abundant in tadpoles that developed in natural pond water and was also negatively correlated with optic tectum width and length (Fontaine et al., 2022). This phylum contains aquatic microbes commonly found in freshwater systems that can contribute to harmful algal blooms and have been implicated in neurodegenerative diseases because of their ability to synthesize neurotoxins (Sini et al., 2021; Zehr, 2011). It is important to note that not all taxa within these phyla are pathogenic, and no symptoms of disease or infection were seen in experimental animals (Fontaine et al., 2022).

Alternatively, we found the phylum Firmicutes was more abundant in tadpoles that developed in autoclaved pond water and was positively correlated with optic tectum width and length (Fontaine et al., 2022). Further, we found that the genus Clostridium sensu stricto 5, a member of the Firmicutes phylum, was positively correlated with the width and length of the optic tectum. Our results are supported by a previous study that found positive correlations between the Firmicutes-associated Clostridium taxa and brain volume in regions involved in sensory integration in humans (Labus et al., 2017). While Clostridium has been implicated in the modulation of peripheral serotonin levels and aspects of neurophysiology (Yano et al., 2015), the exact mechanisms behind these correlations that are driving changes in brain structure in this study are unknown at this time. It is also possible that the changes in relative brain morphology are purely a response to other environmental or physiological factors beyond our treatments that could still impact the composition of the gut microbiota. Future studies could directly manipulate the abundance of specific microbial taxa and analyze changes in relative brain morphology to better understand functional roles of the gut microbiota on neurodevelopment.

We have shown that experimental manipulations of temperature and microbial environment alter tadpole relative brain mass and brain morphology. Further, we found these changes in neurodevelopment were significantly associated with metrics of gut microbial diversity and composition. Thus, our results provide some of the first evidence of the MGB axis in an amphibian model. Future experiments may include methods such as transcriptomics and/or isotropic fractionation to quantify neuronal and total cell counts in nervous tissue, which can help further investigation of the genetic, molecular and cellular causes driving these changes in brain development (Herculano-Houzel and Lent, 2005). Additionally, functional consequences of these changes in brain development may be further tested through behavioral assays, which would strengthen support for the MGB axis in amphibians.

There is value in broadening our understanding of the vertebrate microbiota in more natural developmental settings, even though within this experiment we are unable to determine which specific aspects of the microbiome (i.e. individual bacterial taxa) are driving our observed changes in neurodevelopmental endpoints. Ectotherms such as amphibians represent an ecologically relevant model to investigate host–microbe relationships. Global increases in environmental temperature appear to affect the gut microbial composition and diversity in addition to the other aspects of host physiology modulated by temperature (i.e. metabolic demands, hydration, foraging behavior, digestive performance and brain development; Beltrán et al., 2021; Feder and Burggren, 1992; Fontaine et al., 2018; Rohr and Palmer, 2013). Further, amphibians develop in freshwater systems that can be contaminated with environmental pollutants such as pesticides that alter host physiology as well as the community composition and diversity of colonizing microbes, and subsequently the host gut microbiota of inhabiting animals (Gao et al., 2017; Kohl et al., 2015; Kolpin et al., 2002; McClelland et al., 2018; Woodley et al., 2015; Zhang et al., 2020; Zhu et al., 2022). With these challenges in mind, further investigation of the host–microbe relationships in amphibians, and other wildlife animals facing similar challenges, can help inform wildlife conservation studies and expand the microbiome field beyond mammalian biomedical applications (Trevelline et al., 2019).

We thank M. Ohmer for field collection assistance; and K. Kohler, M. Maurer, M. Maier, S. Reilly, A. Haid, C. Duckworth and J. Adams for animal husbandry and DNA extraction assistance. We also thank Dr Peter Freeman for his counsel regarding statistical analyses. Additionally, we thank the DNA Services Facility at the University of Illinois at Chicago for sample processing.