Does brain size of Asiatic toads (Bufo gargarizans) trade-off with other energetically expensive organs along altitudinal gradients?

Study sites and sampling

A total of 434 individuals (233 males and 201 females) from 15 sites along three altitudinal transects were sampled from west of the Sichuan Basin in 2018 and 2019 (Supplementary Table S1; Supplementary Figure S1). All toads were breeding adults sampled during the breeding season near breeding ponds. Each transect followed one river valley from a low-altitude site to a high-altitude site. We selected river valleys with large altitudinal differences and steep ascendance. Four to six sites were sampled in each transect, and these sites were typically 300 m to 500 m apart by altitude. All sites from a transect were within a straight-line distance of 85 km to reduce confounding ecological background variables (Yao et al., 2021). The lowest altitude site was at an altitude of 785 m above sea level (site 1.1), while the highest altitude site was at an altitude of 3,239 m (site 2.6).

Toads were captured by hand and euthanized near the collecting site immediately after capture. For euthanasia, toads were submerged in a 0.25% MS-222 solution for 10–25 min (Mitchell et al., 2020). Death was confirmed when toads had their eyes closed and stopped moving. All animal processing procedures followed approved animal utilization protocols (Chengdu Institute of Biology, number 20171205).

Body size and body condition measurements

Snout-vent length (SVL) was used to represent body size and was measured from photos taken from a dorsal view using a ruler as a reference. We measured SVL three times using ImageJ version 1.53g (Broeke et al., 2015) and used the average value. Body mass was measured by an electronic balance to the nearest 0.1 g after euthanasia. The scaled mass index (SMI) was used to assess body condition (Peig & Green, 2009). SMI is a condition index based on body mass and morphometric measurements (e.g., body length); studies have demonstrated that it is comparable among different populations or species (Drakulić et al., 2016^; Peig & Green, 2009) and represents overall energy storage well in anuran amphibians (MacCracken & Stebbings, 2012).

SMI was calculated as

where M_i and L_i were body mass and SVL of individuals, respectively, while L₀ was an arbitrary value of L (e.g., here, the mean value within each population); b_SMA was the estimated slope of the standardized major axis (SMA) regression of ln-transformed values of body mass on body length (SVL) using lmodel2 package (Legendre, 2018).

Brain and visceral organ size measurements

Whole brain volume was used to represent brain size. We dissected the brains out and removed the cranial nerves, pineal organ, meninx, and pituitary glands (the pituitary infundibulum was preserved). Photographs of brains were taken from dorsal and lateral views while keeping the coronal plane and sagittal plane parallel to the camera sensor. Length (L), width (W), and height (H) of the brains were measured to the nearest 0.001 mm from the photographs using ImageJ. Each brain was measured three times by the same investigator (ZY), and the average value was used. The volume of the brain was calculated using an established ellipsoid model as V = (L × H × W) π/(6 × 1.43) (Jiang et al., 2015^; Liao et al., 2022). The parameters of this formula were empirically calibrated based on the application of the Cavalieri method to the brains of amphibians (Jiang et al., 2015). Brain size data from transects one and two were previously published (Yao et al., 2021), and data from transect three were acquired in this study.

In order to ensure comparability with previous studies investigating similar species (Zhao et al., 2019^; Zhong et al., 2017), dry mass was used to represent the size of visceral organs, including the heart, liver, alimentary tract, and kidney. All toad specimens were stored in neural buffered formalin (10%) over 6 months. Organs were first dissected out: the liver excluded the gall bladder; the alimentary tract included stomach, small and large intestine, and the contents in the tract were carefully removed; kidneys included the pair, and the heart included the overall heart (including conus arteriosus and sinus venosus). These organs were then washed with tap water overnight and dried in a baking oven (55 °C) until their weights stabilized. All dried organs were weighted to the nearest 0.0001 g by an electronic balance (STP2004). Organs that were damaged during the dissection process were excluded. In total, we obtained data for 367 brains, 298 hearts, 336 livers, 306 alimentary tracts, and 338 pairs of kidneys.

Data analysis

We first examined the relationships between SVL, SMI, and altitude to evaluate the assumed energetic constraint and its variation along the altitudinal gradients using general linear models (GLMs). In all models, SMI and SVL were the dependent variables, while sex and altitude were the independent variables. Sex was treated as a categorical variable, while altitude was considered as a continuous variable. We included sex in the model because this species has documented sexual size dimorphism (Fei et al., 2012). Both SVL and SMI were log-transformed to meet the assumption of normal distribution. All GLM analyses were conducted using R package stats.

We then examined relationships between the relative size of organs and altitude to test for potential impacts of altitude on organ size using GLMs. Log-transformed organ sizes, including brain volume and dry mass of heart, liver, alimentary tract, and kidneys, were dependent variables, while sex and altitude were independent variables, and log-transformed SVL was covariable. Model assumptions, including normality of residuals and homoscedasticity, were checked, and there was no significant violation of these assumptions.

Finally, we employed structural equation modeling (SEM) to test relationships between brain size and the sizes of visceral organs along altitudinal gradients. SEM allows us to summarize all four visceral organs into one unmeasured variable (latent variable) and test multiple causal relationships simultaneously. To prepare the data for model fitting, we conducted allometric correction for sizes of brain and visceral organs based on populations (= sampling sites) using R package GroupStruct version 0.1.0 (Chan & Grismer, 2022). All relative sizes, SVL, and SMI were further standardized to a mean of zero and a variance of one. Since the allometric correction process and the maximum likelihood method for SEM model evaluation did not accept missing data (Chan & Grismer, 2022^; Schumacker & Lomax, 2016), we removed individuals with missing data after confirming that the missing data were randomly distributed (p > 0.05, missing completely at random test with R package naniar; Tierney et al., 2024).

For SEM construction, we first defined a latent variable (“budget”), which represents the energy allocation budget to the four visceral organs, and constructed a measurement model for variable “budget” (Figure 1A). The sizes of the four visceral organs were indicator variables, and the variable “budget” was set as the predictor of these indicators. All connections in the model were linear regressions. We employed a confirmatory factor analysis (CFA) to validate the model. Multiple model fit indices, factor loadings, average variance extracted, composite reliability, and modification indices were estimated (Gana & Broc, 2019^; Rosseel, 2012^; Schumacker & Lomax, 2016). All indices suggested that the measurement model was a good fit and robust (Figure 1A; Supplementary Table S2; Supplementary Table S3). Secondly, a basic SEM model (SEM1) with brain size being the predictor variable and the latent variable “budget” being the effect was constructed based on the EBH hypothesis (Figure 1B). Three alternative models, in which SVL (SEM2), SMI (SEM3), and sex (SEM4) were covariables, respectively, were also proposed (Figure 1C–E). The categorical variable sex was transformed with dummy coding. All connections in these models were linear regressions. Lastly, to detect differences of relationships between brain size and the latent variable “budget” at different altitudes, we fitted data to a multiple-group SEM. All observations were assembled into three altitudinal groups: a high-altitude group (2,452–3,239 m), a middle-altitude group (1,223–2,136 m), and a low-altitude group (785–935 m). A principal component analysis of bioclimatic data of all sampling sites clustered all sites into three groups (data and analysis not shown), and the three-group design also provided reasonable sample sizes for each group. We applied a measurement invariance test to a multigroup CFA model for the latent variable “budget” (Figure 1A), and confirmed that the model coefficients were comparable among groups (“metric invariance”; Chen & Yung, 2023). We also used a chi-square difference test to assess difference between a free model, in which all path coefficients were allowed to vary across altitudinal groups, and a constrained model, in which the path coefficients were fixed to be equal (Chen & Yung, 2023).

Open multimedia modal

We employed Akaike information criterion (AIC), Bayesian information criterion (BIC), and a series of commonly used indices to test model fit and compare alternative models, including ratio of /df, comparative fit index, goodness of fit index, normed fit index, incremental fit index, root mean square error of approximation, and standardized root mean square residual (SRMR) (Gana & Broc, 2019^; Kang & Ahn, 2021^; Schumacker & Lomax, 2016^; Thakkar, 2020). Finally, model coefficients were estimated from the best-fit model. All model evaluations and estimates were conducted using R package Lavaan version 0.6-17 Rosseel, 2012).

All data processing and analyses were conducted in R version 4.1.2 (R Core Team, 2021). We completed GLMs plotting using R package ggplot2 (Wickham, 2016, p. 2) and R package ggeffects (Lüdecke, 2018). SEM path diagrams were constructed using R package semPlot version 1.1.6 (Epskamp, 2015) and semptools version 0.2.10.1 Cheung & Lai, 2023^; Epskamp, 2015).

SVL, SMI, and organs size change along altitudinal gradients

Both SVL and SMI significantly decreased with altitude (Figure 2A and B; Supplementary Table S4). GLMs indicated significant negative correlations between SVL and altitude (estimate ± SE = − 0.036 ± 0.008, p < 0.001) and between SMI and altitude (estimate ± SE = − 0.089 ± 0.021, p < 0.001). Toads from high-altitude sites had smaller SVL and SMI than those from low-altitude sites. For example, the mean SMI at the lowest (site 1.1) and highest (site 2.6) sites were 66.150 ± 1.112 (mean ± SE) and 47.391 ± 0.889, while the mean values of SVL were 88.755 ± 2.346 mm and 81.754 ± 1.945 mm, respectively. There was also significant sexual dimorphism in both SVL and SMI; males had smaller SVL and SMI than females (Supplementary Table S4).

Open multimedia modal

Brain volume also significantly decreased with altitude (Figure 2C; Supplementary Table S4). GLMs indicated a significant negative correlation between brain volume and altitude while controlling SVL (estimate ± SE = − 0.058 ± 0.010, p < 0.001). Toads from high-altitude sites had smaller relative whole brain volumes than those from low-altitude sites. For example, the mean brain volume at the lowest (site 1.1) and highest (site 2.6) altitudes were 77.239 ± 3.012 and 60.429 ± 2.741 mm³, respectively. Sexual dimorphism in brain size was also noticed; males had significantly smaller brains than females (Supplementary Table S4).

Sizes of the liver, alimentary tract, and kidneys revealed significant positive correlations with altitude while controlling SVL (p < 0.001; Figure 2E–G; Supplementary Table S4). The size of the heart also had a positive relationship with altitude, but the correlation was nonsignificant (p = 0.555; Figure 2D; Supplementary Table S4). Male toads often had smaller relative organ sizes than females, although the difference was significant only in the alimentary tract (p < 0.001; Supplementary Table S4).

Co-variation between brain size and visceral organs

Both AIC and BIC indicated that the basic SEM model (SEM1; Figure 1B) was the best-fit model, in which brain size had a causal connection to the latent variable “budget” without impacts from SVL, SMI, or sex. The model also had an acceptable fit measured by most indices (Supplementary Table S2). Based on this model, data were fitted to the multiple-group SEM, and model coefficients were extracted. In the high-altitude group, brain size had a negative path coefficient on the latent variable “budget” and was a significant predictor of the latter (prestandardized value, estimate ± SE = −0.40 ± 0.18, z = −2.22, p < 0.05; Supplementary Table S5), although it explained only a small proportion of variance in “budget” (R² = 0.11; Supplementary Table S5). This suggested a trade-off relationship between brain size and energy allocation to the visceral organs in high-altitudinal toads. On the other hand, path coefficients from the brain to the latent variable “budget” were positive but nonsignificant in both the middle-altitude group (prestandardized value, estimate ± SE = 0.06 ± 0.07, z = 0.81, p = 0.415; Supplementary Table S5) and the low-altitude group (prestandardized value, estimate ± SE = 0.17 ± 0.11, z = 1.60, p = 0.109; Supplementary Table S5). Furthermore, the chi-square difference test resulted in a significant difference between the free model and constrained model (p < 0.05), indicating a significant difference in the relationship between the brain and “budget” at different altitudes. Figure 3 presents standardized factor loading and path coefficient values of the model for better comparison.

Open multimedia modal

Factor loadings for all organs were statistically significant (p < 0.01; heart as the scaling parameter did not receive a p-value), and three organs, heart, liver, and kidney, consistently demonstrated high factor loadings, particularly in the middle- and high-altitudinal groups (Figure 3; Supplementary Table S5). In the high-altitude group, the heart had the highest standardized factor loading (0.85), suggesting a strong correlation between heart size and the latent variable “budget.” Any variation in the latent variable “budget” would like produce the largest relative size change in the heart, compared to other organs. In contrast, it was the liver that had the highest standardized factor loading in the low-altitude group, suggesting the liver had the strongest correlation with the latent variable “budget.” Additionally, the heart, liver, and kidney had high R² values (0.20–0.93; Figure 3), suggesting that the model explained substantial variance in these three organs but little in the alimentary tract (Figure 3; Supplementary Table S5). Together, these results suggested that brain size was a significant predictor of the latent variable “budget” in high-altitude toads, and the latent variable “budget” was a significant predictor of the sizes of heart, liver, kidney, and alimentary tract.

Trends of organ size change along altitudinal gradients

The brain of Asiatic toads displays a clear trend of decrease in relative size along altitudinal gradients (Figure 2C) and is consistent with the EBH. The brain is a metabolically expensive organ that uses disproportionally large amounts of energy. At high altitudes, resources are particularly limited, and their availability is highly variable; therefore, a small brain volume is likely adaptive under such energy limitations. We do not have direct evidence of the fitness consequence of brain size variation along altitudinal gradients in these toads. However, despite a decrease in whole brain size, Asiatic toads from high altitudes have relatively large telencephalons, which suggests a balance between the maintenance of cognitive functions and decreased energy expenditure achieved by reducing overall brain size (Yao et al., 2021). Their high-altitude habitats are generally homogeneous, barren, with few breeding ponds (our field observation). A large telencephalon, which is associated with spatial memory and behavioral flexibility in amphibians (Kaas, 2016), likely allows the toads to quickly locate breeding sites while a small total volume reduces energy expenditure. On the other hand, a large brain with an overall better cognitive ability is likely advantageous at low altitudes, where habitats are complex and resources are abundant. Previous works on brain size showed a similar altitudinal trend in rodents (Nengovhela et al., 2023).

Several visceral organs (i.e., liver, alimentary tract, kidneys) show a clear trend of increase in relative size in our study (Figure 2E–G); however, several recent works on B. gargarizans revealed different patterns. For example, Li et al. (2020) detected a negative correlation between visceral organ size and altitude; Zhao et al. (2019) found no significant correlation between altitude and the size of the heart, lungs, liver, or kidneys of B. andrewsi (often considered a synonymy of B. gargarizans; Fu, 2023). Studies of other amphibians also revealed an inconsistent pattern. Zhong et al. (2017) detected a positive correlation between heart size and altitude in the spot-legged treefrog (Polypedates megacephalus). However, in the plateau frog (Nanorana parkeri), heart size was smaller in high-altitudinal populations than those of low-altitudinal conspecifics (Hou et al., 2023). These inconsistent patterns suggest that the sizes of these organs are likely outcomes of interrelated and complex energetic trade-offs (more discussion below) as well as environment-dependent fitness consequences. For example, recent genetic data suggested that the heart likely plays a key role in high-elevation adaptation in Asiatic toads (Chen et al., 2022^; Yang W. et al., 2017). Zhong et al. (2017) suggested that large hearts likely have fitness benefits at high altitudes for amphibians, and Velotta et al. (2018) reached the same conclusion for mammals. A closer look at our data suggests that the heart size of Asiatic toads may have a “U”-shaped relationship with altitude (Supplementary Figure S2; Supplementary Table S6). At high altitudes, the fitness benefits of a large heart may outweigh the energetic costs, and at low altitudes, abundant resources may simply allow a large heart even with minimum fitness benefits. Furthermore, the inconsistent patterns may be the consequence of phenotypic plasticity. These organ sizes are highly plastic and are influenced by available resources during development and at the time of examination.

Energetic constraints and trade-offs between the brain and other expensive organs

Energetic constraint is a fundamental assumption of the EBH (Warren & Iglesias, 2012). The impacts of the constraints on trade-offs, however, depend on total energy acquisition and energy allocation; variation in either may change the trade-offs or render the constraints powerless to produce the expected trade-offs. Even though resources may be limited, organism’s acquisition strategy and ability are highly variable (Agrawal, 2020).

It is clear that Asiatic toads at high altitudes have low total energy acquisition and experience strong energetic constraints. Both SVL and SMI, two indicators of body growth and energy acquisition, consistently decrease with altitude (Figure 2A and B). In particular, SMI represents the overall energy storage well in anuran amphibians (MacCracken & Stebbings, 2012), and low SMI likely represents a decreased energy availability. Furthermore, a recent physiological examination of Asiatic toads from different altitudes detected no significant difference in their SMR (similar to BMR) when acclimated to the same temperature (i.e., no counter-gradient variation; Tan et al., 2021). Therefore, populations from high altitudes with low environmental temperatures likely have low body temperatures, low metabolic rates, and a short active season. All of these can result in low energy acquisition (e.g., low SMI). As a consequences of stronger energetic constraints imposed by low energy acquisition at high altitudes, the high-altitude toads evolve a trade-off between energy allocation to the brain and to other energetically expensive visceral organs (Figure 3). These patterns are consistent with the predictions of EBH. Furthermore, these visceral organs respond to energetic trade-offs differently, likely depending on their fitness consequence. At high altitudes, the heart appears to have the largest and most consistent response to variation of energy allocation to the visceral organs (standardized factor loading = 0.85, R² = 0.73; Figure 3). Any reduction in energy allocation to the brain can predict an increased energy allocation to the visceral organs, which in turn can consistently predict a large increase in heart size. The heart is known to play a crucial role in high-altitude adaptation in toads as well as other vertebrates (Chen et al., 2022^; Velotta et al., 2018^; Yang W. et al., 2017), and any size changes likely have significant fitness consequences.

No trade-offs between brain size and energy allocation to the visceral organs are detected in the middle- and low-altitude groups, which is most likely a consequence of high total energy acquisition at these locations. Particularly in the low-altitude group, a positive relationship is observed, although statistically nonsignificant (p = 0.109; Supplementary Table S5). This is the classic car-house trade-off paradox (Agrawal, 2020^; Houle, 1991). When variation in resource acquisition is greater than variation in resource allocation, a positive correlation often occurs between traits that compete for the resource. A similar lack of trade-off relationship was also observed in several other vertebrates (Mai et al., 2017^; Yang et al., 2018). Although EBH considers variation in total energy intake and predicts a positive correlation between brain size and energy acquisition (e.g., BMR), few have tested the prediction so far (e.g., Genoud et al., 2018). Clearly, additional research is needed in this area.

Implications for the EBH and future research

EBH is evidently applicable to poikilotherms, which represent the majority of animal diversity despite being developed in a mammalian context. Nevertheless, fundamental physiological differences of poikilotherms demand a refocus for testing the EBH in these animals. First, poikilotherms have high within-species variability of total energy acquisition, which is likely much higher than that of homeotherms (Liao et al., 2023^; Song et al., 2023). As evidenced by Asiatic toads, such variability has significant impacts on trade-off relationships. Therefore, to test the EBH in poikilotherms, their total energy intake should be emphasized as much as energy allocation. Second, poikilotherms have different energy allocation models than homeotherms, and energy demand and consumption of different expensive organs can vary drastically (Careau et al., 2014^; Song et al., 2023). Some poikilotherms are also known to have regional heterothermy (e.g., Cox et al., 2023). All these could lead to energy allocation change without organ size change. Third, environmental temperature likely plays a major role in determining brain size, not only through energy metabolism but also through development and function (fitness). For example, temperature is known to have a positive effect on neuronal development and activity (Beltrán et al., 2021), and lizards and fishes with lower developmental temperature show relatively poor cognitive performance (Amiel & Shine, 2012^; Brezden et al., 1975). Recently, Song et al. (2023) detected a positive relationship between brain size and body temperature across amphibians and reptiles that are active during daytime.

We strongly encourage more intra-specific comparisons evaluating the EBH. Most current supporting evidence for the EBH, i.e., trade-offs between brain and expensive functions, is from inter-specific comparisons (Heldstab et al., 2022). Our study is one of the few intra-specific studies supporting the EBH. Paradoxically, among species, trade-offs are more likely consequences of niche specialization and community interaction rather than limited resources (Agrawal, 2007^; Grabowski et al., 2023). Additionally, intra-specific comparison excludes confounding evolutionary history and allows us to focus on variables of interest (Axelrod et al., 2018). Furthermore, to move from correlation to causal relationships, a controlled experiment will be necessary (e.g., Kotrschal et al., 2013), which can be conducted only at the intra-specific level. It is also important to recognize that trait associations at different biological scales often involve different evolutionary processes (Agrawal, 2020).

A holistic look at the fitness consequences of size variation of the brain is essential to better understand brain size evolution. Recent research on the costs of a large brain has been very focused on energy constraints and resulting trade-offs (reviewed by Heldstab et al., 2022). This is sensible for encephalized species because energy constraints may play an over-sized role in determining the fitness consequence in these species. For the majority of animals with relatively small brains, however, the influences on fitness go beyond energy. This includes the examination of function-oriented brain region sizes (e.g., telencephalon in Asiatic toads) further than energy expenditure-oriented total brain volume. Similarly, the fitness consequences of size variation of other energetically expensive organs (e.g., heart in high-altitude toads) likely limit the degree of trade-offs that the brain may have. This has rarely been evaluated, and it requires an ecological context, such as an altitude gradient, in this study. Lastly, it should be recognized that community ecology, as opposed to energetic constraints, can also produce trade-offs (Agrawal, 2020). An ecological context is essential to understand the fitness consequences of those trade-offs.

Acknowledgments

We would like to thank L. Shu, Dr. Z. Liao, and Professor J. Huang for consultation with statistical analysis. We also thank B. Sun, Dr. Y. Cao, Dr. S. Tan, Y. Lin, Q. Yang, Y. Zhang, Y. Chen, and X. Zhu for field assistance. Drs. J. Cui and W. Yang kindly helped us with the project design. Drs. F. Laberge, O. Todorov, and two anonymous reviewers provided many valuable suggestions.