Dataset of diverse evolutionary pathways shape cichlid egg size in Lake Tanganyika

Optimal egg size is a classic and important concept in life history theory. Here, we examined the ecological factors affecting egg size, using a comparative analysis of 83 cichlid species from Lake Tanganyika, which employ either a mouth-brooding or substrate-brooding. The two strategies differ substantially in regard to spatial limitations and food availability for offspring, potentially leading to divergent responses to the ecological factors influencing egg size. We observed a strong negative correlation between egg size and egg number in species that exhibit mouth-brooding, but not in substrate-brooding cichlids. Our results suggest that the importance of the relationship between clutch size and egg size varies substantially between these two types of parental care. Interestingly, in mouth-brooding species, we found that egg size increases when offspring exhibit external feeding behaviors (grazing behavior). We also demonstrate that substrate-brooding species with shorter periods of parental care typically produced larger eggs, compared to species with relatively longer periods of parental care. Overall, our results demonstrate that behavioral differences, including parental care type and duration, play an important but often overlooked role in the evolution of egg size across species.

Methods
Egg size, clutch size (i.e., the number of eggs released during a single spawning event), and adult female body mass were measured in 47 mouth-brooding (belonging to 11 tribes) and 40 substrate-brooding cichlid species (belonging to 2 tribes), all endemic to Lake Tanganyika, in East-Central Africa. The sexually mature female specimens were obtained mainly from local fishermen and fish markets (n = 62 species, n = 161 individuals) and from field sampling using a fine-mesh gill net, hand nets, and SCUBA (n = 19 species, n = 73 individuals). The captured fish were euthanized by overdosing with anesthetic (MS-222, Wako Pure Chemical Industries) and stored in a freezer until measurements were taken. Given the relatively small sample size of females carrying mature eggs collected during the field survey, we also used samples stored at Osaka Metropolitan University (n = 10 species, n = 22 individuals) and the Hokkaido University Museum (n = 7 species, n = 13 individuals). Standard length (in mm), body mass (i.e., gonad-free body weight to 0.001 g), and ovary mass (to 0.001 g) were measured. Mature eggs were distinguishable from immature eggs in the ovary based on coloration and size and were counted using a tally counter and preserved in 5% formalin. Immature eggs found in the same ovary were not included in the total and omitted from the analyses. In each clutch we randomly selected five eggs or used all eggs if the clutch size was smaller than five, to measure the mean egg size. Eggs were photographed using a digital camera equipped with a stereomicroscope (HD212; AMscope, USA), and the long and short diameter was measured on each egg to an accuracy of 0.01 mm using Image J ver. 1.52q software (Schneider et al. 2012). The mean value of the long and short diameter was used to define the mean egg size. In total, measurements were taken from 1326 eggs from 269 female individuals representing 87 species.

Literature survey

Information on parental care, i.e., brooding type - mouth or substrate-brooding, care type- maternal parental or biparental care, duration of parental care, and presence or absence of grazing behavior, was collected from the literature.

For 44 of 47 mouth-brooding species, we collated data on care type (biparental or maternal care) and grazing behavior (present or absent; Yanagisawa 1985, 1986; Kuwamura 1988b, 1997). We predicted that the acquisition of external food resources by offspring (i.e., grazing behavior) may allow for a reduction in egg yolk sac investment and consequently a reduction in egg size. Although we gained life history trait data from the field for Baileychromis centropomoides, Paracyprichromis nigripinnis, and Xenotilapia nigrolabiata, we were unable to determine care type or the presence/absence of grazing behavior due to limited ecological descriptions of these species in the literature.

For 39 of 40 substrate-brooding species, we collected information on care type (biparental or maternal care) and the duration of parental care (short or long). In this study, we defined the parental care period as days from the spawning event to dispersal and broadly classified the duration of parental care as either “shorter” or “longer”, based on the literature. Some cichlid species in Lake Tanganyika exhibit extremely short periods of parental care, often less than 20 days after hatching, while others have comparatively longer periods, ranging from 66–130 days . As far as we are aware, an intermediate period of parental care, between 20 and 65 days, has not been reported in any Lake Tanganyikan substrate-brooding species. Therefore, we defined species exhibiting parental care behaviors for ≤ 20 days as having “shorter” (n = 7 species) and > 20 days as having “longer” (n = 32 species) parentalcare duration. Information on parental care duration was lacking in one species Neolamprologus mustax again due to limited ecological descriptions.

Phylogenetic tree

To account for the effects of phylogeny, we included in our analyses a modified time calibrated phylogenetic tree of the cichlids that have radiated in Lake Tanganyika (Ronco et al. 2021). We pruned the Ronco et al. (2021) tree using the ‘droptip’ function in R ver 4.4.0 (R Core Team. 2024) package ape (Paradis et al. 2004) down to 85 species, excluding Oreochromis tanganicae and Tylochromis polylepis from our initial sample size of 87 species. We then obtained divergence ages of O. tanganicae and T. polylepis, which secondarily invaded Lake Tanganyika from a time estimated phylogenetic tree targeting representatives of all cichlid tribes (Ronco et al. 2021). These divergence ages were integrated into the previously pruned phylogenetic tree, resulting in a tree representing all 87 species used in our study. The ancestral states of egg size were estimated in phytools (Revell 2024) to visualize the posterior probabilities of each trait condition of mean egg size of species, along the branches of the cichlid phylogeny in Lake Tanganyika.

Phylogenetic signal life-history traits

We calculated phylogenetic signals for egg size, clutch size, and female body mass with the ‘phylosig’ function in the R package phytools. We note that measurement error can lead to an underestimation of the strength of phylogenetic signals and potentially obscure patterns of phylogenetic correlation among species in univariate cases. Therefore, for the 62 species for which repeated data were available, the strength of the phylogenetic signal was calculated, considering intraspecific variation. Each trait value was log10-transformed and the mean for all species and standard error for available species were calculated and used in the analysis.

Phylogenetic comparative analyses

We used Bayesian phylogenetic mixed model analyses (BPMM; R package “MCMCglmm”) to investigate the effects of brooding strategy (i.e. mouth- and substrate-brooding) and various parental care traits on the trade-off between clutch size and egg size. In each model, we fitted the pruned phylogenetic tree with 87 species as a random effect.

First, we used BPMM to investigate the effect of brooding strategy on mean egg size We fitted a model mean with egg size (log10-transformed) as the response variable and brooding type (mouth- or substrate brooding), mean clutch size (log10-transformed), mean female body mass (log10-transformed), and their interactions as fixed effects. The pruned phylogenetic tree was fitted as a random effect. The error distribution for the continuous variables followed a Gaussian distribution and were consequently standardized to facilitate the comparison of regression coefficients among the fixed effects. We used the uninformative priors for the random effect and the residual variance (inverse-Wishart distribution, nu = 1/2, V = 0.002) and non-informative priors for the fixed effects. The MCMC was performed with 205000 total iterations, a burn-in of 5000, and a thinning interval of 100 was applied. We evaluated model convergence based on the visual examination of trace plots. The model convergence was also checked by running the MCMC chain independently and using the Gelman-Rubin's diagnostic was less than 1.1 using the 'gelman.diag' function in the R package coda (Plummer et al. 2006). For all fixed effects, random effects and residuals, the effective sample size exceeded 1000. We also computed the variance inflation factor (VIF) to assess collinearity between log-transformed clutch size and body mass. The VIF for each fixed effect was lower than 10 (Clutch size = 8.26, female body mass = 4.96, brooding type = 1.75, clutch size: brooding type = 8.45, female body mass: brooding type = 4.67).

Next, since egg size variation was likely to account for different ecological conditions between mouth- and substrate brooders, we ran BPMMs separately for mouth-brooding and substrate-brooding species to investigate factors influencing variation in egg size within brooding strategies. For mouth-brooding species, we fitted a model with mean egg size as the response variable and with mean clutch size, mean female body mass, grazing behavior (presence or absence), and care type (biparental or maternal mouth-brooding) as explanatory variables, omitting any two-way interaction terms. For substrate-brooding species, we fitted a model with mean egg size as the response variable and with mean clutch size, mean female body mass, parental care duration (short or long), and care type (biparental or maternal brood guarding) as explanatory variables, again omitting any two-way interactions. The phylogenetic tree, isolated to either mouth-brooding or substrate-brooding species was incorporated as a random effect in each model. All other model components were conducted the same as described above.