Physiological predictors of reproductive performance in the European Starling (Sturnus vulgaris)

Abstract

BackgroundIt is widely assumed that variation in fitness components has a physiological basis that might underlie selection on trade-offs, but the mechanisms driving decreased survival and future fecundity remain elusive. Here, we assessed whether physiological variables are related to workload ability or immediate fitness consequences and if they mediate future survival or reproductive success. We used data on 13 physiological variables measured in 93 female European starlings (Sturnus vulgaris) at two breeding stages (incubation, chick-rearing), for first-and second-broods over two years (152 observations).ResultsThere was little co-variation among the physiological variables, either in incubating or chick-rearing birds, but some systematic physiological differences between the two stages. Chick-rearing birds had lower hematocrit and plasma creatine kinase but higher hemoglobin, triglyceride and uric acid levels. Only plasma corticosterone was repeatable between incubation and chick-rearing. We assessed relationships between incubation or chick-rearing physiology and measures of workload, current productivity, future fecundity or survival in a univariate manner, and found very few significant relationships. Thus, we next explored the utility of multivariate analysis (principal components analysis, Mahalanobis distance) to account for potentially complex physiological integration, but still found no clear associations.ConclusionsThis implies either that a) birds maintained physiological variables within a homeostatic range that did not affect their performance, b) there are relatively few links between physiology and performance, or, more likely, c) that the complexity of these relationships exceeds our ability to measure it. Variability in ecological context may complicate the relationship between physiology and behavior. We thus urge caution regarding the over-interpretation of isolated significant findings, based on single traits in single years, in the literature.Electronic supplementary materialThe online version of this article (10.1186/s12983-018-0288-3) contains supplementary material, which is available to authorized users.

Robust Physiological Metrics From Sparsely Sampled Networks

“…In particular, it can be sensitive to population composition, and structure can also vary depending on conditions or population ( Pigeon et al, 2013 ). For example, application of PCA to 13 circulating biomarkers in wild European starlings ( Sturnus vulgaris ) showed that the correlation structure—and thus the axes one might extract—varied across years and breeding season ( Fowler et al, 2018 ). This is to be expected: changes in food abundance, diet composition, or parasite burden should be expected to vary across seasons and years, and also to cause changes in which biomarkers correlate with which.…”

“…However, more generally, care should be used to apply PCA and related methods with appropriate cross-population validation to ensure that any findings are true reflections of biological organization rather than artifacts of sampling, population composition, or environmental heterogeneity ( Cohen et al, 2018b ). In the case of the Fowler et al (2018) study, no meaningful axes could be extracted due to the strong changes in correlation structure and the limited samples sizes in the relevant subgroups to robustly estimate structure. In contrast, integrated albunemia has been validated in multiple populations ( Cohen et al, 2015b ) and even species ( Wey et al, 2019 ).…”

“…To our knowledge, the first application was to measure physiological declines prior to death in fruit flies ( Drosophila melanogaster ) ( Shahrestani et al, 2012 ). It was subsequently and independently applied to human biomarker data for similar purposes, and has also been applied in wild animals, both with ( Milot et al, 2014 ) and without ( Fowler et al, 2018 ) success. The assumption in all these cases is that average state is close to optimal state, and that individuals far from optimal are likely to be more unhealthy.…”

“…The Cohen lab has unpublished analyses on gene expression data in which WGCNA failed to produce even minimally similar structures across different datasets; this may be due to the difficulty correcting for batch effects and how this impacts correlation structure, or to population differences in correlation structure ( Dufour et al, 2020 ). It is well-known that correlation structures do change, across age ( Barajas-Martínez et al, 2020 ), across environment ( Fowler et al, 2018 ), and across cell type ( Huttlin et al, 2020 ), among others. Such instability of the basic network structure could be problematic, but also could be the feature of interest ( Fowler et al, 2018 ; Barajas-Martínez et al, 2020 ).…”

“…It is well-known that correlation structures do change, across age ( Barajas-Martínez et al, 2020 ), across environment ( Fowler et al, 2018 ), and across cell type ( Huttlin et al, 2020 ), among others. Such instability of the basic network structure could be problematic, but also could be the feature of interest ( Fowler et al, 2018 ; Barajas-Martínez et al, 2020 ). It is also worth noting that such methods generally characterize the network structure of a population, but do not permit direct evaluation of an individual’s physiological state.…”