Project Topics
Get Complete Project Material File(s) Now! »
The phylogenetic comparative method in palaeohistology
The link between bone histology and phylogeny has long been a matter of controversy among palaeohistologists. Originally thought to be of great interest in the field of systematics (Queckett, 1849a, 1855), osteohistological characters have progressively become mostly used, at least among tetrapods, to study biomechanical and other functional constraints, relegating the phylogenetic information they potentially contain to a ‘secondary’ signal.
Studies on the bone histology of vertebrates in a phylogenetic perspective have been in constant increase over the last decade (e.g. Padian et al., 2001; Padian, Horner & de Ricqlès, 2004; de Ricqlès, Castanet & Francillon-Vieillot, 2004; Cubo et al., 2005, 2008, 2012; Montes et al., 2007, 2010; Dumont et al., 2013; Houssaye et al., 2013; Houssaye, Tafforeau & Herrel, in press; Marín-Moratalla et al., in press), and the dichotomy between phylogenetic and functional signals for histological features has been described as misleading (since these explanatory factors are not mutually exclusive and any feature can significantly exhibit both of them). Yet, most incursions of palaeohistologists in the field of phylogeny have consisted in optimizing these descriptive characters on phylogenies taken from the literature without further investigation on a potentially informative signal. De Ricqlès et al. (2008: 73), while acknowledging the presence of a significant phylogenetic signal in the bone histology of archosaurs at the microanatomical level, considered features measured at the histological level as reflecting “mostly […] autapomorphic signals” (citing the results obtained by Cubo et al., 2005), and considered themselves “agnostic about whether the histological character-states that [they] tentatively used really depict apomorphic or plesiomorphic (or homoplastic) conditions at the nodes involved”. In a recently published book, Padian (2013: 4) identifies the “four signals” of fossil bone histology as being ontogeny, phylogeny, mechanics, and environment; the phylogenetic signal is described as being “persistent, but […] never the strongest signal”. Padian et al. (2013: 271) consider that “Separating phylogenetic signal of bone microstructure from ontogeny, biomechanics and environment will be extremely difficult, because these other signals are often directly related to phylogeny”.
The reason why qualitative, descriptive histological characters have mostly failed to reflect any significant phylogenetic signal and always been identified as being mostly nfluenced by autapomorphic, functional constraints, is that these characters are the result of the categorization of continuous features into discrete characters relative to bone matrix and bone tissue types, following a reference nomenclature established by Francillon-Vieillot et al. (1990) and de Ricqlès et al. (1991). These seminal works allowed a generation of palaeohistologists to provide accurate descriptions of the bone histology in major vertebrate taxa and to identify numerous characteristics of bone growth mecanisms, and brought evidence of extremely high levels of homoplasy ocurring at the histological level for discrete features. The discretisation of the variation present in vertebrate bone histology at all organization levels into unequivocal characters has been used to describe patterns of variation in entire bone sections. This may have resulted in ambiguous interpretations of some of these characters, especially because some of them have recently been considered dubious due to methodological biases for identifying some bone tissue types (Bromage et al., 2003; Lee, 2013; Stein & Prondvai, 2014). For this reason, the ‘traditional’, descriptive histological nomenclature can still be regarded as a very powerful tool to describe the comparative anatomy of bone, but it is unsuitable for a precise account of the variation of bone microstructure in a phylogenetic context, and its use may have prevented previous workers from identifying all signals accounting for this variation, including the phylogenetic one. It is worth noting that the bone histological features showing a high phylogenetic signal (and hence a potential interest in systematics) at the phylogenetic level considered in this study (palaeognaths) can show a homoplastic pattern of variation at more inclusive levels (birds, dinosaurs, archosaurs, diapsids, etc).
Phylogenetic Comparative Methods
Constructing paleobiological models of bone growth rate estimation. The paleobiological growth rate inference models were constructed using multiple linear regression tested for significance using permutations (Legendre et al., 1994) using the computer program Permute! version 3.4 alpha 9 (Casgrain, 2009). We obtained a standardized coefficient for each variable (a’), which was unstandardized into a raw coefficient (a) by using the expression: a1 = a’1 * (sY / sX1).
Evolution of Growth Patterns and Metabolic Rates in Archosauromorpha
The evolution of bone growth rates and metabolic rates in Archosauromorpha is a subject of major interest among paleontologists. Gross (1934) discovered fibrolamellar bone tissue (i.e. formed at high growth rate and compatible with a high metabolic rate) in Erythrosuchus – one of the most basal archosauriformes. This result, confirmed by de Ricqlès (1976), attracted the attention of many paleontologists towards non-archosaurian archosauromorphs as key taxa to understand the thermometabolism of the last common ancestor of archosaurs (e.g. de Ricqlès et al., 2008; Nesbitt et al., 2009; Botha-Brink & Smith, 2011; Werning et al., 2011). Among these contributions, that by Botha-Brink and Smith (2011) is particularly relevant because these authors performed for the first time bone growth rate estimations in nonarchosaurian archosauromorphs. Their estimates are the most conservative possible values because they were computed by measuring the amount of cortex deposited between two successive growth rings (assuming that a growth ring is deposited annually), divided by 380 days of a Triassic year (Botha-Brink and Smith, 2011). Considering that these taxa may have grown for only 6 months (i.e. spring and summer), the growth rate estimates would in fact double (J. Botha-Brink, pers. comm.).
Do Isotopic Analyses suggest that Crocodiles are Secondarily Ectothermic?
The analysis of the oxygen isotopic composition of phosphate in biogenic apatite (bone, teeth) has been used to estimate the ectothermic or endothermic status of extinct vertebrates (e.g. Barrick et al., 1996; Fricke & Rogers, 2000; Amiot et al., 2006). Considering that oxygen isotope fractionation between PO4 and body water is thermally dependant (Longinelli & Nuti, 1973), several studies have shown that for animals living in the same biota and having the same water strategies (obligate or non-obligate drinkers), the δ18OPO4 values are expected to be different between endotherms and ectotherms (Amiot et al., 2004, 2006). In terrestrial ecosystems, the water source depends on the isotopic composition of meteoric water, which is in turn controlled by latitude and air temperature (Dansgaard, 1964; Fricke & O’Neil, 1999). Model curves of present-day δ18O values of endothermic and ectothermic vertebrates have been established as a function of the latitude. According to these models, endothermic vertebrates are expected to have higher δ18O values than ectothermic ones above 50° latitude, but ectothermic vertebrates should be similar to endotherms, or display higher δ18O values at low latitudes (Amiot et al., 2004, 2006).
While results obtained in dinosaurs are compatible with endothermy (thus supporting the hypothesis of a widespread high metabolic rate in Cretaceous dinosaurs), the isotopic values obtained for Cretaceous crocodiles and turtles suggest that these animals were ectothermic (Amiot et al., 2006). This last result is compatible with our hypothesis of an ancestral endothermic state for the last common ancestor or archosaurs because, as quoted above, crocodylomorphs may have lost their endothermic condition when they became aquatic in the Jurassic. The other prediction derived from our hypothesis suggests that Triassic pseudosuchians and non-archosaurian archosauromorphs may have been characterized by δ18O values more similar to those of endothermic Triassic dinosaurs than to those of ectothermic Triassic lepidosaurs and turtles.
Comparison between methods used in this study and in Grady et al. (2014)
In the present study, we used osteohistological features and phylogenetic eigenvector maps (PEM) as independent variables in our predictive modeling of mass-specific RMR ; these variables were all measured for this study. A contrario, Grady et al. (2014) built their predictive model using growth rate as an independent variable; growth rate was compiled from measurements from previous studies, and expressed as maximum growth rate (Gmax), in g day-1. However, not all growth rates taken from literature were originally measured in g day- 1, which might generate some uncertainty and conversion errors. For example, Psittacosaurus mongoliensis is identified has having a Gmax of 5.82 kg year-1 (Erickson, Curry-Rogers & Yerby, 2001). In Grady et al. (2014) this number becomes 13.8 g day-1, which would imply a number of 422 days in an Early Cretaceous year, from which Psittacosaurus is dated; even when considering the variation of length of a day over time (Myhrvold, 2013), this estimation is incorrect. Similarly, Tenontosaurus tilletti has a Gmax of 27 kg year-1 (Lee & Werning, 2008), which would require an impossibly small number of days (139) in a Middle Cretaceous year to match the value of 194.5 g day-1 used by Grady et al. (2014). Some estimations of growth rates for extinct dinosaur species may thus be biased by these conversion rates.
The other major difference between this study and that of Grady et al. (2014) is the type of phylogenetic regression used for building predictive models. The construction of PEM for a given trait involve weighting the edges (i.e. branches) of the phylogenetic tree on the basis of the among-species phylogenetic covariance matrix of this trait, using a steepness parameter a – related to Pagel’s κ (Pagel, 1999) and to the Ornstein-Uhlenbeck selection strength parameter α (Butler & King, 2004) – to describe the relationship between changes in traits values and branch lengths in the tree (a = 0 under purely neutral evolution). This procedure is a significant improvement on the arbitrary assumption of purely neutral evolution (i.e. following a Brownian motion model) assumed by phylogenetic independent contrasts (PICs; Felsenstein, 1985) used by Grady et al. (2014), which imply a strict relationship between the variation of a given trait and branch length information for the corresponding phylogeny. Furthermore, although most of the species in their sample of nonavian dinosaurs have been the matter of precise dating studies in the literature (Cubo et al., 2012; Legendre et al., 2013), no branch length information was included by Grady et al. (2014) in their trees for both non-avian dinosaurs and crocodilians, which adds an important bias to the way their model takes into account phylogenetic information.
Table of contents :
Remerciements
Sommaire
Introduction
Partie I – Le signal phylogénétique dans la variation des caractères ostéohistologiques
I – 1. Phylogenetic signal in bone histology of amniotes revisited
Introduction
Material and methods
Results
Discussion
Acknowledgements
I – 2. Bone histology, phylogeny, and palaeognathous birds (Aves, Palaeognathae)
Introduction
Material and methods
Results
Discussion
Acknowledgements
Conclusions de la première partie
Partie II – L’évolution de la croissance osseuse et du thermométabolisme chez les archosauromorphes
II – 1. Evidence for high bone growth rate in Euparkeria obtained using a new paleohistological inference model for the humerus
Introduction
Material and methods
Results
Discussion
Acknowledgements
II – 2. Palaeohistological evidence for ancestral endothermy in archosaurs
Methods summary
Supplementary information
Acknowledgements
Author Contribution
Conclusions de la seconde partie
Conclusion générale et perspectives
Bibliographie
