Title: Patterns of covariation in the masticatory apparatus of hystricognathous rodents: implications for evolution and diversification. Running head: Patterns of covariation in the masticatory apparatus of hystricognaths Authors:

The mammalian masticatory apparatus is a highly plastic region of the skull. In this study, a quantification of shape variation, the separation of phylogeny from ecology in the genesis of shape brings new insights on the relationships between morphological changes in the cranium, mandible, and muscle architecture. Our study focuses on the Ctenohystrica, a clade that is remarkably diverse and exemplifies a rich evolutionary history in the Old and New World. Current and past rodent diversity brings out the limitations of the qualitative descriptive approach and highlights the need for using integrative quantitative methods. We present here the first descriptive comparison of the whole masticatory apparatus within the Ctenohystrica, by combining geometric morphometric approaches with a non-invasive method of dissection in 3D, iodine-enhanced microCT. We used these methods to explore the patterns of covariation between the cranium and the mandible, and the interspecific morphological variation of the skull with regard to several factors such as phylogeny, activity period, type of habitat, and diet. Our study revealed strong phylogenetic and ecological imprints on the morphological traits associated with masticatory mechanics. We showed that, despite a high diversification of lineages, the evolutionary history of Ctenohystrica comprises only a small number of morphotypes for the skull and mandible. The position of the eye was suggested as a key factor determining morphological evolution of the masticatory apparatus by limiting the number of possible pathways and promoting convergent evolution towards new habitats and diets between different clades. INTRODUCTION Rodents represent by far the largest mammalian order with more than 2200 species that occupy most of the ecosystems on the planet (Wilson and Reeder, 2005). But despite such diversification, all extinct and extant rodents share one of the most extreme specializations of the masticatory apparatus. Diprotodonty (i.e. the reduction of the upper and lower incisor series to a single pair) is a hallmark of the rodent masticatory apparatus and is accompanied by a reduction of the number of cheek teeth in association with the development of antero-posterior movements of the mandible for gnawing and chewing (Becht, 1953). Despite the apparent versatility of their masticatory apparatus, the order Rodentia has retained only a small number of different morphotypes for the skull and the mandible (Wood, 1965; Hautier et al., 2008, 2009; Cox and Jeffery, 2011). Different phylogenetic histories and selective pressures have moulded the characteristics of these morphotypes, while strong functional constraints affecting mastication have limited the number of possible evolutionary pathways and promoted convergent evolution. The Ctenohystrica (sensu Huchon et al. 2002: Ctenodactylidae+Diatomyidae and Hystricognathi; Fig. 1) exemplifies a rich evolutionary history in the Old and New World and is remarkable in showing multiple examples of parallel evolution. Both molecular (Huchon and Douzery, 2001; Huchon et al., 2007; Montgelard et al., 2008) and morphological analyses (Bugge, 1985; Luckett and Hartenberger, 1985; Woods and Hermanson, 1985; Marivaux et al., 2002) have long supported the monophyly of this major group of rodents. Like the great majority of living rodents, most of the members of Ctenohystrica are omnivorous or herbivorous (Landry, 1970); however, in contrast to this restricted variation in diet, they display a diverse array of ecological types. The South American Caviomorpha is arguably the most successful group of Ctenohystrica. Their fossil record attests for a rapid radiation, most of the modern caviomorph families appearing during the Paleogene (Lavocat, 1976). Because they diversified in complete isolation in South America during part of the Cenozoic period, they were able to fill niches usually occupied by other placental mammals (Elissamburu and Vizcaíno, 2004; Townsend and Croft, 2008). As a consequence, extant and extinct caviomorphs show a high anatomical and ecological diversity, ranging from the pseudo-ungulate maras (Dolichotis) to the fossorial tucotuco (Ctenomys). Interestingly, the differentiation in diet and habitat has occurred independently in two different monophyletic groups (the Cavioidea [Rowe and Honeycutt, 2002] and the Octodontoidea [Honeycutt et al., 2003]), but few studies have depicted the morphological characters of their masticatory apparatus as a whole. Such parallel evolution gives us a unique opportunity to estimate the role of phylogeny and evolutionary selective forces in driving the morphological evolution of the caviomorph skull. In the present study, we sought to determine whether ecological factors have influenced the evolution of the skull of Ctenohystrica. Hypotheses explaining the adaptive significance of these traits often relate to diet (e.g. Alvarez et al., 2011a, Croft et al., 2011). It is however reasonable to question whether animals subjected to intense predation pressure like rodents may evolve differently in different types of habitat or during different periods of the day. Phylogenetic constraints may have also played an important role in the morphological evolution of the skull precluding the occurrence of particular feeding modes in a given lineage (Claude et al., 2004). Broad cladewide studies combining analyses of cranial and mandibular variations are clearly lacking for rodents. Here, we use geometric morphometrics to explore the morphological variation of the skull of Ctenohystrica in relation to both phylogeny and ecology. Many works have been devoted to describing the morphological variation of the cranium or the mandible (e.g. Renaud and Michaux, 2003; Michaux et al., 2008; Hautier et al., 2009, 2011; Alvarez et al., 2011b), but the patterns of covariation between these two main elements of the masticatory complex have been largely unexplored. Thus, we strive to characterize how the morphological features of the cranium covary with the mandibular morphology, especially by characterizing the interplay between the position of the eye socket and the masticatory musculature. MATERIAL AND METHODS Sample composition The material studied came from the collection of the Museum national d’Histoire naturelle in Paris (MNHN, collection Vertébrés supérieurs Mammifères et Oiseaux), the Natural History Museum in London (BMNH), the Mahasarakham University Herbarium (MSUT), and of the Institut des Sciences de l’Evolution de Montpellier 2 (ISE-M). We analysed 177 mandibles and 196 skulls belonging to sciurognathous and hystricognathous rodents of both sexes, representing 41 genera and 16 families of Ctenohystrica (Fig. 1): Abrocomidae, Capromyidae, Cuniculidae, Caviidae, Chinchillidae, Ctenodactylidae, Ctenomyidae, Dasyproctidae, Diatomyidae, Dinomyidae, Echimyidae, Erethizontidae, Hystricidae, Octodontidae, Petromuridae and Thryonomyidae (see list in S1). The Ctenohystrica have the essential assets to fulfil the objectives set here: they are highly diversified, with a wide range of ecomorphological adaptations, and they include a wide range of mandibular morphologies (Hautier et al., 2011). Geometric morphometric methods – The mandibular and cranial forms were quantified with 23 and 73 anatomical landmarks respectively (Fig. 2). Digital data of all specimens were acquired using a Microscribe 3-D digitizer and using X-ray microcomputed tomography (μCT). Because the mandible of rodents is constituted by a unique dentary bone of relatively simple shape, most of the landmarks taken on the dentary were of type 2 (e.g. maxima of curvature – Fig. 2; Bookstein, 1991). All configurations (sets of landmarks) were superimposed using the Procrustes method of generalized least squares superimposition (GLS scaled, translated, and rotated configurations so that the intralandmark distances were minimized) following the method used by Rohlf (1999) and Bookstein (1991). Subsequently, mandibular and cranial forms of each specimen were represented by centroid size S, and by multidimensional shape vector v in linearized Procrustes shape space. Shape variability of the mandible was analysed by principal components analysis (PCA) of shape (Dryden and Mardia, 1998). Analysis and visualization of patterns of shape variation were performed with the interactive software package MORPHOTOOLS (Specht, 2007; Specht et al., 2007; Lebrun, 2008; Lebrun et al., 2010). Because it was impossible to remove the incisors from the CT scanned mandible, colors are mapped onto the mandibular incisors in all figures even if only two landmarks were actually taken on the incisors (mandible landmarks 1 and 2, Fig. 2). Thus, it is worth noting that the incisor structure is not analyzed by the set of landmarks used here, and no interpretation can be made on a putative link between incisor shape and ecology. A public version is currently being developed (contact [email protected] for further information). In order to take into account of the potentially confounding effects of size allometry on shape, size-corrected shapes were obtained as follows. Regressions of Procrustes coordinates against the logarithm of centroid size were computed for all families (except for mono-specific families), yielding family-specific allometric shape vectors (ASVf). The ASVf represent directions in shape space which characterize family-specific allometric patterns of shape variation. A common allometric shape vector (ASVc), obtained as the mean of all the ASVf, provided a direction in shape space that minimizes potential divergence in mandibular allometric patterns across families (see Lebrun et al., 2010 and Ponce de León and Zollikofer, 2006 for further details concerning this methodology). ASVc was then used to decompose the shape of each species-wise mean shape and of each family-wise mean shape into size-related (vs) and size independent (vi) components. Furthermore, covariation patterns between the crania and the mandibles were studied using 2-blocks partial least square analysis, as described by Bookstein et al. (2003), only adapted to allow for the use of 3D landmarks. For the N=164 specimens for which both cranial (k=73) and mandibular (l=23) landmarks had been digitized, cranial and mandibular landmark configurations were aligned separately using GLS, yielding a cranial matrix of N* 3k shape coordinates and a mandibular matrix of N *3l shape coordinates. The PLS analysis computed a series of pairs of unit vectors, the singular cranial and mandibular warps (Uc and Um), each being of length 3k and 3l, respectively. These pairs of singular warps maximize the covariance between the two sets of shape coordinates. Cranial and mandibular projection scores of the specimens on the singular warps were subsequently computed. Multivariate analysis of Variance (MANOVA) and Canonical Variate Analyses were performed on the principal component scores of each species-wise mandibular and cranial mean shapes (vi) in order to assess the effects of different factors on mandibular and cranial shape variation: clades (families), activity period (diurnal, nocturnal, and twilit), type of habitat, and diet (Nowak, 1999; Townsend and Croft, 2008). Following Townsend and Croft (2008), five categories of diets were considered: omnivorous, fruit-leaf, fruit-seed, grass, and roots. Four types of habitats were set apart: open areas, woody areas, burrowers, and ubiquists (Nowak, 1999). The terms “type of habitat” and “diet” refer to the usual habitat and principal diet and are given in S1. In order to quantify mandibular shape affinities at the family level, family-wise mean mandibular shapes were clustered using the UPGMA (unweighted pair-group method) on original shape data and shape data corrected for allometry. MANOVAs were performed with STATISTICA v6.0 (StatSoft Ltd., Milton Keynes, UK), Canonical Variate Analyses with MORPHOTOOLS. The UPGMA trees were computed using PHYLIP (Felsenstein, 1989). Imaging and reconstruction In order to reveal detail of both soft tissue and bony anatomy, formalin-fixed heads of Cavia porcellus and Proechimys cuvieri (representatives of the Cavoidea and Octodontoidea respectively) were imaged using the new technique of contrast-enhanced microCT (Jeffery et al., 2010). The specimens were supplied post-mortem by Biomedical Services, University of Liverpool, and François Catzeflis, Institut des Sciences de l’Evolution de Montpellier, respectively. The specimens were stained by immersion in an approximately 10% solution of iodine potassium iodide (I2KI) over a number of weeks. The stained specimens were then scanned with the Metris X-Tek custom 320kV bay system at the EPSRC funded Henry Moseley X-ray Imaging Facility, University of Manchester. Voxel resolutions were 0.08 mm (Cavia) and 0.04 mm (Proechimys). Threedimensional reconstructions of the skull, mandible, masticatory muscles and orbital contents (eye globe, extraocular muscles and lacrimal gland) were then created using the segmentation function of Amira 5.3.3 (Visage Imaging Inc., San Diego, CA, USA).