Unravelling intravertebral integration, modularity and disparity in Felidae (Mammalia).

Marcela Randau,Anjali Goswami

doi:10.1111/ede.12218

Abstract

Morphological integration and modularity, which describe the relationships among morphological attributes and reflect genetic, developmental, and functional interactions, have been hypothesized to be major influences on trait responses to selection and thus morphological evolution. The mammalian presacral vertebral column shows little variation in vertebral count and therefore specialization for function occurs primarily through modification of vertebral shape. However, vertebral shape has been suggested to be under strong control from developmental canalization, although this has never been explicitly tested. Here, we assess hypotheses of developmental modules in the vertebrae of felids to determine whether developmental interactions are a primary influence on vertebral modularity. Additionally, we analyze the magnitudes of both intravertebral integration and disparity to evaluate if level of integration varies along the vertebral column and, if so, whether integration and disparity are associated. Our results confirm the hypothesis of vertebral developmental modularity, with most presacral vertebrae displaying two modules. Exceptions are concentrated in the boundaries among traditional and functional regions, suggesting that intravertebral modularity may reflect larger-scale modularity of the felid vertebral column. We further demonstrate that overall integration and disparity are highest in posterior vertebrae, thus providing an empirical example of integration potentially promoting greater morphological responses to selection.

Highlights

The dichotomy between maximum individual trait adaptation and cohesion between functioning parts is one that directly affects phenotypic response to selection (Klingenberg et al, 2003; Badyaev et al 2005; Hansen and Houle 2008; Porto et al 2009; Goswami and Polly 2010a; Goswami et al 2014)
The basis for understanding how organisms are organized was laid by the seminal work by Olson and Miller (1958) in which they described the fundamental concepts of phenotypic integration and modularity as can be ascertained through quantification of patterns of trait covariation
Both measures of disparity were calculated first per individual species per vertebra, and across taxa per vertebra, using the species mean shapes. Results from both RV and Covariance Ratio analysis (CR) analyses of modularity were consistent in all but one case, and strongly supported the twomodule model (P < 0.01) for all but six (C2, C7, T1, T8, L6, and L7) of the 19 analyzed vertebrae. They differed only with regards to T13, which was marginally significant for the tested modules with RV analysis, but significant when analysed with CR (P values 1⁄4 0.051 and 0.011, respectively; Table 2)

Summary

Introduction

The dichotomy between maximum individual trait adaptation and cohesion between functioning parts is one that directly affects phenotypic response to selection (Klingenberg et al, 2003; Badyaev et al 2005; Hansen and Houle 2008; Porto et al 2009; Goswami and Polly 2010a; Goswami et al 2014). Modules are a set of traits that show higher covariation among them than with other parts of the organism due to shared genetic or developmental origins or function, while integration is the overall pattern of intercorrelation (e.g., Hansen and Houle 2008; Klingenberg 2008; Goswami and Polly 2010b; Klingenberg and Marugan-Lobon 2013). Those two definitions are not contradictory and complex traits may present overall high integration and still be modular (Bookstein 2015), such as the mammalian skull (Goswami 2006a,b; Goswami and Polly 2010c). Trait integration has been shown to reflect shared developmental pathways in early ontogeny, postnatal function, and heterochronic shifts (Zelditch and Carmichael 1989; Goswami et al 2009; Zelditch et al 2009; Bennett and Goswami 2011; Goswami et al 2012, 2014), and to be susceptible to reorganization by extreme changes in selection (Drake and Klingenberg 2010)

Methods

Results

Discussion

Conclusion