Skip to main content

Fast life-histories are associated with larger brain size in killifishes

Cite this dataset

Sowersby, Will et al. (2021). Fast life-histories are associated with larger brain size in killifishes [Dataset]. Dryad.


The high energetic demands associated with the vertebrate brain are proposed to result in a trade-off between the pace of life-history and relative brain size. However, because both life-history and brain size also have a strong relationship with body size, any associations between the pace of life-history and relative brain size may be confounded by coevolution with body size. Studies on systems where contrasts in the pace of life-history occur without concordant contrasts in body size could therefore add to our understanding of the potential coevolution between relative brain size and life-history. Using one such system - 21 species of killifish - we employed a common garden design across two ontogenetic stages to investigate the association between relative brain size and the pace of life-history. Contrary to predictions, we found that relative brain size was larger in adult fast-living killifishes, compared to slow-living species. Although we found no differences in relative brain size between juvenile killifishes. Our results suggest that fast- and slow-living killifishes do not exhibit the predicted trade-off between brain size and life-history. Instead, fast and slow-living killifishes could differ in the ontogenetic timing of somatic vs. neural growth or inhabit environments that differ considerably in cognitive demands.


During 2017-2018, we reared individuals from 21 species of killifish (Aplocheiloidei) from the egg stage to either non-sexually mature juveniles (in their linear growth phase, Sowersby et al. preprint) or sexually mature adults (adults: Nspp= 17, Nind = 234; juveniles: Nspp. = 18, Nind = 110; see Supplementary Information for a description of fish maintenance, and Table 1 and Supplementary Tables S1 to S7 for details on sample sizes, body size and age at sampling, and population origin). As environmental factors can induce plastic responses in brain size in fishes (Gonda et al. 2011), we kept all individuals in an environmentally standardized, common garden setting. The species included in the study were purposely selected to represent the five major evolutionary transitions between fast-living (annual) and slow-living (non-annual) life-history strategies (see Table 1; Table S1-S2), characterized by the presence or absence of eggs capable of entering embryonic diapause (as per Furness et al. 2015). Previous studies have confirmed that fast-living species exhibit faster life-history traits, compared to slow-living species (reproduction: Eckerstrm- Liedholm et al. 2017, development and growth: Sowersby et al. preprint), and that these traits are strongly correlated across species (the dominant eigenvector of a principal component analysis explain 75.4% of the total variation; Eckerstrm- Liedholm et al. 2019). After hatching, a subset of individuals were reared in 0.75-L plastic boxes (although some individuals were then moved to 13-L tanks in species that grew quickly), and were weighed and dissected as juveniles (at ~1 cm total body length) before they had reached sexual maturity (mean wet mass: 0.10 g; range: 0.015 to 0.34, no gonad tissue was observed during dissection). The remainder of the individuals were reared identically but were transferred to 13-L tanks 10 to 14 days after hatching . Each tank contained a mixed-sexed group of conspecifics of up to eight individuals, because some species produced more hatchlings, the number of groups and sample sizes are therefore not equal across species. Upon reaching sexually maturity these individuals were weighed and then dissected (i.e. after sexual maturity, but prior to showing signs of senescence). Sexual maturity was determined by the presence of species-specific male coloration. All individuals, juveniles and adults, were euthanized with a lethal dose of benzocaine solution, with experimental procedures approved by the Ethical Committee in Stockholm, Sweden (license N132/15). Individuals were euthanized, blotted with a paper towel, and their wet mass was recorded (precision: 1 mg; XS105, Mettler-Toledo GmbH, Giessen, Germany). Brains were fixed inside the head for 5-7 days in 4% phosphate-buffered formaldehyde, after which they were transferred to a phosphate- buffered saline solution and stored until dissection. Brains were dissected and photographed using a stereo microscope with a built-in 3 Mpixel digital camera (Leica EZ 4HD; Leica Application Suite EZ 3.4; Leica Microsystems GmbH, Wetzlar, Germany). Nerves were cut to ca. 0.3 mm length, the spinal cord was cut posterior to medulla oblongata, and fatty tissue, meninx, and blood vessels were removed to the extent possible. Brains were then photographed dorsally, ventrally, and laterally (left and right). Photographs were calibrated using a reference photo of a digital caliper (Absolute, Mitutoyo, Takatsu-ku Kawasaki, Japan) set at 6.00 mm, and measurements were taken using ImageJ 1.49 (Schneider et al. 2012). Length (L), width (W), and height (H) of olfactory bulbs, telencephalon, optic tectum, cerebellum, medulla oblongata and hypothalamus were recorded and used to calculate volumes (V) for each sub-region, using the idealized ellipsoid model (Huber et al. 1997): V = (L ∙ W ∙ H) / 6 For olfactory bulbs, telencephalon, optic tectum, and hypothalamus, the measurements of the two lobes were taken separately and thereafter added together; while the cerebellum and medulla oblongata were treated as single-lobed structures. The total brain volume was calculated by summing all sub-region volume estimates. Adult brains were blotted and weighed after photography (precision: 0.01 mg; MT5, Mettler-Toledo GmbH, Giessen, Germany). Correlation between brain mass and calculated total brain volume was high (r = 0.93, N = 234), but due to the adhesive nature of some brains, which led to slight damage and loss of tissue, we proceeded to analyze brain volume rather than mass. Juvenile brains were not weighed.


To control for phylogenetic non-independence, a phylogenetic effect was added to all analyses (Freckleton et al. 2002). For this purpose, we used a previously published time-calibrated phylogeny (Furness et al. 2015), with an additional four species inserted into said phylogeny (see Figure 1). These additional species were inserted into the phylogeny based on taxonomic information and other published phylogenies. Specifically: Ophthalmolebias constanciae was placed within the main Simpsonichthys clade (Pohl et al. 2015), Nothobranchius kadleci alongside its sister species N. furzeri (Dorn et al. 2015), Scriptaphyosemion cauveti and Rachovia aff. brevis (Monteria population) in their respective genera, and Millerichthys robustus was placed as a sister species of Rivulus cylindraceus (Gonzalez-Voyer et al. in prep). After these species were added, the tree was then pruned, leaving the overall structure of the tree unchanged. Therefore, the position of the additional species reflects the phylogenetic relationships that would be recovered from a tree with complete species sampling.

Statistical analyses

Relative brain-/sub-region size To test for differences in relative brain- and sub-region size between the two life- history strategies, we fit models with absolute brain volume (mm3; log10-transformed) as a response variable, and the explanatory variables: log10-transformed body mass (mean centered based on species means), life-history strategy (fast and slow), the interaction between body mass and life-history strategy, and sex (female and male). In addition, species identity and phylogenetic distance were added as random effects (Felsenstein 1985; Gelman 2007). Since the adult allometric slopes, between brain and body size, differed for different species (DIC = 25.7 for adults, DIC = 1.5 for juveniles, random slopes equal slopes), we chose to use random slope models. Adult and juvenile brain volumes were analyzed separately, but the models were identically applied, except for excluding sex as an explanatory factor in the juvenile analysis. In both adults and juveniles, we analyzed differences in intercepts and slopes of evolutionary static allometric slopes. As we only had one sampling point per individual, we did not obtain any estimates of ontogenetic slopes. In order to confirm that body size did not differ across the fast- and slow-living killifishes, we tested the effect of life-history strategy and sex (only for the adult model) on log10- transformed body mass, with species identity and phylogenetic relationship as random effects. Further, in the analysis above we used the classification of annual and non- annual as a proxy for fast and slow-living species. To validate our use of this proxy, we repeated the analysis (as described above), using the scores along the dominant eigenvector from a principal component analysis on growth rate, development time and reproductive rate (see Eckerstrm-Liedholm et al. 2019; Sowersby et al. preprint). These scores represent indirect measures of the pace-of-life and were available for all but two species. To analyze the relative volume of specific subregions of the brain (i.e. log10-transformed volumes of the olfactory bulbs, telencephalon, optic tectum, cerebellum, medulla oblongata and hypothalamus), we fitted one model per sub-region. Each model contained the following explanatory variables: log10-transformed brain volume (mean centered based on species means), life-history strategy (fast or slow), and their interaction. Species identity and the phylogeny were added as random effects. All models were analyzed using the MCMCglmm package Hadfield 2010 in R. Fixed effects were fitted with flat priors, while random effects were fitted with parameter-expanded locally non-informative priors (Murphy 2007). The parameter sampling was run for 2.01M iterations (burn-in: 10 000; thinning-interval: 2 000; posterior samples: 1 000). Autocorrelations between parameter estimates were within the interval of -0.1 and 0.1 for all analyses. For all analyses, the assumption of normally distributed residuals was assessed using visual examination. In the results, statistics from the Bayesian models are presented with parameter estimates () followed by their 95% credibility intervals (lower bound; upper bound), and Bayesian P-values (PMCMC).

Growth rate and brain size

We examined relative brain size under stages and conditions where energetic investment into life-history traits were realized; i.e. the subject fish were growing juveniles in isolation and reproducing (i.e. sexually mature) adults in mixed-sex groups. However, rearing density has previously been found to suppress growth rates in fish, both in the laboratory (Ribas et al. 2017) and in the wild (Lorenzen and Enberg 2002; Vrtlek et al. 2019), and if body size has a stronger plastic response than brain size, this could potentially affect relative brain size. As the fast- and slow-living species differ in terms of growth (Sowersby et al. preprint), it is plausible that plastic effects (suppression of growth) will be stronger in those species with faster growth rates (Auld et al. 2010). Hence, consistent differences in relative brain size among the fast-living and the slow-living groups could arise due to plastic effects of growth changes, rather than differential energetic budgets. To explore these options, we analyzed the relationship between brain size and growth rate (species means, cm ∙ day-1; Sowersby et al. preprint) using total brain volume as a response variable in a phylogenetic generalized least squares (PGLS) model with body mass and growth rate used as predictor variables, using a pruned version of the phylogenetic tree described above, with maximum likelihood estimation of (the strength of the phylogenetic signal; Freckleton et al. 2002). The model was fitted with the function pgls in the package caper (Orme 2018) in R.


Swedish Research Council, Award: 2013-5064

Swedish Research Council, Award: 2013-4834

Association for the Study of Animal Behaviour

Alice & Lars Siléns Fond

Alice & Lars Siléns Fond