Some phenotypic dimensions are more developmentally variable than others. Such developmental variability (or bias) is common and uncontroversial. However, how and at what timescales these biases constrain or facilitate the emergence of standing genetic variation, plastic responses, as well as adaptation, remains contentious. To investigate the extent to which developmental variability shapes genetic variation, plasticity, and evolution, we first quantify developmental variability in the shape of the dung beetle foreleg — a functional trait critical for the excavation of breeding tunnels. We do so by testing how random developmental perturbations, manifesting themselves in fluctuating asymmetry, shape standing genetic variation within populations. Next, we investigate whether such developmental variability is aligned with thermal plasticity and recently evolved latitudinal variation. We find that, while developmental variability is a strong predictor of standing genetic (co)variance (i.e., the G-matrix), latitudinal population differentiation and thermal plasticity were unrelated to developmental variability. This suggests that, while developmental variability may shape standing genetic variation, it does not seem to constrain the evolution of putatively adaptive population differentiation and plastic responses. At least in this system, developmental biases do not seem to constrain morphological differentiation on ecological timescales.

Laboratory rearing and data acquisition

To investigate how developmental variability relates to genetic and environmental variation, we revisited a common garden study conducted by Rohner and Moczek 2021. In brief, this study sampled Onthophagus taurus beetles from one location in the native Mediterranean range (Italy) as well as four populations in the invasive range in the United States. Wild-caught females were brought into the laboratory, where they were then allowed to reproduce in plastic containers (27cm × 8cm × 8cm) filled with soil and defrosted cow dung. Reproductively active females excavate tunnels in the soil and construct so-called ‘brood balls’ underground. These brood balls were extracted from the soil using a sieve after five days. Because larval growth and adult morphology strongly depend on the size and quality of the brood ball, we removed eggs from their natural brood ball and placed them in standardized artificial brood balls (consisting of cow dung placed in 12-well plates). The brood of each female was then split evenly and kept at constant 19°C or 27°C. These temperatures were chosen as they approximate average soil temperatures during the breeding season at the northern and southern range edges in the exotic range (Florida vs. Michigan). After adult emergence, beetles were sacrificed and stored in 70% ethanol. This common garden rearing generated 679 individuals, including 321 females. We focused on females because this is the sex that primarily excavates tunnels in this species and correspondingly evolved an exaggerated tibia morphology.  

To increase the sample size for the estimation of developmental variability, we added an additional 113 laboratory-reared F1+ females that were originally collected in North Carolina (the same location as in the common garden experiment). This led to a sample size of n = 434 for O. taurus. In addition to this focal species, we also added an additional 83 female Digitonthophagus gazella. This species is morphologically and ecologically similar but shares its most recent common ancestor with the Onthophagus lineage about 40 million years ago. This species was added to provide an additional point of comparison for the statistical comparison of covariance matrices (see below).

We removed the left and right foretibiae of all O. taurus females and photographed them twice using a Pixelink camera (M20C-CYL) mounted on a Leica M205 stereoscope. For each photograph, the foretibia was placed on a piece of modeling clay and positioned in such a way under the stereoscope that all features that were used for landmarking were in focus. Each tibia was photographed twice. The tibia was removed from the clay and positioned anew between pictures to acquire two independent measurements of the same sample. We then placed eight two-dimensional landmarks on the pictures using TpsDig2. This led to a final dataset for O. taurus of 1,736 measurements of 16 variables (434 individuals in total × 2 sides (left/right) × 2 independent measurements). All landmark coordinates were then aligned simultaneously using Procrustes Analysis in the R package geomorph. Centroid size was extracted as a measure of structural size. Note that, while the original study, Rohner and Moczek 2021, also focused on tibia shape, the analysis here is based on a completely independent set of pictures and a slightly modified set of landmarks to enable comparisons across species.