Evolutionary Patterns of Bone Histology and Bone Compactness in Xenarthran Mammal Long Bones

Bone microstructure reflects physiological characteristics and has been shown to contain phylogenetic and ecological signals. Although mammalian long bone histology is receiving increasing attention, systematic examination of the main clades has not yet been performed. Here we describe the long bone microstructure of Xenarthra based on thin sections representing twenty-two species. Additionally, patterns in bone compactness of humeri and femora are investigated. The primary bone tissue of xenarthran long bones is composed of a mixture of woven, parallel-fibered and lamellar bone. The vascular canals have a longitudinal, reticular or radial orientation and are mostly arranged in an irregular manner. Concentric rows of vascular canals and laminar organization of the tissue are only found in anteater bones. The long bones of adult specimens are marked by dense Haversian bone, a feature that has been noted for most groups of mammals. In the long bones of armadillos, secondary osteons have an oblique orientation within the three-dimensional bone tissue, thus resulting in their irregular shape when the bones are sectioned transversely. Secondary remodeling is generally more extensive in large taxa than in small taxa, and this could be caused by increased loading. Lines of arrested growth are assumed to be present in all specimens, but they are restricted to the outermost layer in bones of armadillos and are often masked by secondary remodeling in large taxa. Parameters of bone compactness show a pattern in the femur that separates Cingulata and Pilosa (Folivora and Vermilingua), with cingulates having a lower compactness than pilosans. In addition, cingulates show an allometric relationship between humeral and femoral bone compactness.


Introduction
The bone microstructure of the vertebrate skeleton has been studied since the nineteenth century, with Quekett [1,2], among others, laying the foundation for further research into comparative osteohistology. Later research included both fossil and extant samples, and provided a better overview of microscopic structures of bone among major groups of vertebrates (e.g. [3][4][5][6]). Cubo et al. [7] demonstrated that bone microstructure, like many other biological features, is the product of historical (phylogenetic), functional and structural factors. Bone microstructure does, in addition, also vary greatly over the course of the developmental history of an organism. Among functional factors, parameters of locomotion [8], metabolism [9], and lifestyle [10][11][12][13][14] have been analyzed. Lifestyle is mainly investigated via quantitative analyses of bone tissue by looking at properties such as cross-sectional geometry [15] or bone compactness [16]. Bone compactness parameters have been analyzed in a variety of vertebrates [10][11][12][13][17][18][19], and have shown to contain phylogenetic and ecological signals. These studies were, however, all conducted at a broad scale. Recently, Meier et al. [20] investigated bone compactness within a narrower clade, the moles (Talpidae), and found no significant differences between fossorial and non-fossorial taxa.
Bone histology of mammals in general has been investigated in the past, but in most cases only as a portion of a larger scale study of vertebrates (e.g. [2,5,21]). Studies focusing solely on the histology of extinct and extant mammals are still relatively rare, but important works published so far include those of Singh et al. [22], Klevezal [23], Sander and Andrássy [24] and Köhler et al. [25]. One of the most significant findings of these studies was the presence of lines of arrested growth (LAGs), a feature that was for a long time believed to be exclusive to ectotherms, the new reports thus raising questions about the ectothermic physiology that had once been hypothesized for dinosaurs [24][25][26][27]. But despite the increasing interest in mammalian bone microstructure and the promising results gained so far, little effort has been made to systematically investigate and compare mammalian microscopic bone features.
The Xenarthra represent one of the four major clades of placental mammals [28], and comprise extant armadillos, sloths and anteaters, and their extinct relatives. At present, 31 species of living xenarthrans are recognized, comprising six sloths, four anteaters, and 21 species of armadillos [29]. Most extant xenarthrans are found in Latin America, with the majority being endemic to South America. Dasypus novemcinctus, the nine-banded armadillo, has invaded and successfully colonized much of southern North America within the last two hundred years [30]. Xenarthra most likely evolved in South America, and the diversity of fossil taxa is very high, with more than 150 genera known [31] and with great morphological diversity.
Phylogenetic reconstructions estimated the origin of Xenarthra to be before or close to the Cretaceous-Paleogene boundary, about 80-65 mya [32][33][34][35][36]. On the basis of a combined phenomicmolecular analysis, O'Leary et al. [36] recently found high support for a split of Xenarthra from all other placental mammals (Epitheria) shortly after the Cretaceous-Paleogene boundary. Since then, three ecologically disparate and phylogenetically well supported groups have evolved within Xenarthra [37] (Figure 1). Cingulata includes modern armadillos and the extinct glyptodonts and pampatheres. Members of this group possess an armor that covers head, body and tail. In glyptodonts this armor is rigid, while it has movable bands in armadillos and pampatheres. Despite this morphological similarity to armadillos, the pampatheres are hypothesized to be more closely related to glyptodonts (e.g. [38,39]). The remaining two groups of xenarthrans, Vermilingua (anteaters) and Folivora (sloths), both lack armor and are widely considered to group together as Pilosa, based on morphological and molecular data [32,37,38,[40][41][42][43]. The Vermilingua feature elongate, tubular skulls, an absence of teeth, and a prehensile tongue used for the acquisition of prey. The Folivora are characterized by a hypselodont herbivorous dentition and a reduced dental formula that is exclusive to this group [43]. Both ground and tree sloths lack incisors and true canines, and their cheek teeth are restricted to the maxilla [44]. Folivorans are further well known for their very low metabolism, a feature that is actually common to all xenarthrans, and for their slow locomotion. These features have traditionally also been assumed for their fossil relatives [45]. More recent discoveries and analyses of their functional morphology have suggested more diverse lifestyles for some extinct folivorans, including aquatic habits [46].
Xenarthrans show a marked pattern of disparity in many biological parameters. Differences in body size between extinct and extant taxa are striking, with the weight of glyptodonts estimated at one to two tons and the weight of the giant ground sloth Megatherium americanum at almost six tons [47]. Recent studies have detected an unusually high degree of morphological variation within Xenarthra. Asher et al. [48] found substantial variation in vertebral counts, and Rager and Sánchez-Villagra [49] noted high variability in the timing of cranial suture closure. The latter is especially high among extant sloths, for which notable intraspecific variation has also been found in body size [50] and in the inner ear morphology [51] of species of Bradypus. On the microscopic level, the histology of osteoderms of cingulates and the mylodontids showed interspecific variation [52]. Since dermal ossifications provide information on soft tissue structures otherwise not Figure 1. Phylogenetic relationships between taxa included in this study. The phylogeny is based on [96] and [38], and placement of some fossil taxa follows [31]. doi:10.1371/journal.pone.0069275.g001 preserved in fossils, such studies on the histology of xenarthran osteoderms have also shed light on phylogenetic relationships within the group and on the functional morphology of xenarthrans [39,[52][53][54][55][56][57]. Within Pilosa, only species of mylodontid sloths are known to have possessed osteoderms (e.g. [58,59]).
This study represents the first fine-scale investigation into the bone histology and bone compactness of Xenarthra, with long bones representing all three major clades analyzed qualitatively and quantitatively. Our aim is to address the following questions: Do the histology and/or the compactness of xenarthran long bones show any patterns that could be related to the variation in body size or locomotory modes found in this clade? Do the histology and/or compactness of long bones of Cingulata and Pilosa differ substantially, e.g. due to the occurrence of armour in the former group? Does the long bone histology of xenarthrans comply with the general pattern of disparity for which the group is known?

Ethics Statement
All bones sampled for this study were loaned from the belowmentioned institutions after getting permission to access the collections.

Institutional Abbreviations
AMNH, American Museum of Natural History, New York, USA; AMU-CURS, Colección de Paleontología de Vertebrados

Material Sampled
A total of sixty-seven mostly adult long bones (humerus, femur, radius, ulna, and tibia) of fossil and extant xenarthrans, representing nineteen genera and twenty-two species, were sampled ( Table 1, Table S1). Fossil samples represent only a small subset of the large diversity of extinct xenarthrans [60][61][62][63][64][65], and they illustrate the spectrum of extinct diversity. Long bones are the most abundant remains of fossil xenarthrans, and they are well suited for future comparative work since the emphasis of studies on mammalian bone histology and tetrapod bone microanatomy has traditionally been on these bones (e.g. [10][11][12][13]17,19,22,24]).

Methods of Sectioning
Whenever possible, bones were sampled at the mid-diaphyseal level. Long bones undergo the least secondary remodeling at this level, yielding the most complete growth record [66,67]. Additionally, the mid-diaphyseal level has shown the maximum ecological signal [17]. With regard to the bone compactness analysis, this also ensured that variation in compactness profiles of different bones was not due to variation in connection with different sampling sites. In most cases, the mid-diaphyseal level corresponded to the area where the bone exhibited the smallest circumference. However, the sampling site had to be adjusted for humeri of anteaters and for all xenarthran ulnae ( Figure 2). Anteater humeri are characterized by a prominent deltoid tubercle located in the mid-diaphyseal region. Because bone drift and bone remodeling typically occur in such regions of bony outgrowth, sectioning through the deltoid tubercle would produce thin sections showing a distorted growth record of the humerus. Anteater humeri were thus sampled in the area with the least circumference between the deltoid tubercle and the lateral epicondyle. The sampling sites also had to be adjusted for ulnae, because in xenarthrans these bones vary greatly in shape and relative size, and the area with the smallest circumference did not always correspond to the mid-diaphyseal level. Therefore, to standardize sampling sites in ulnae, we took the distance between the coronoid process and the distalmost end of the bone, and then cut the bone halfway through this distance. Fossil bones considered too valuable or too large for producing regular cross sections were sampled by core drilling (see also [68]). For this purpose a 20 mm diamond-studded core drill bit (Metabo SBE 610) at 300 RPM was used.
For the production of thin sections, we followed standard procedures (e.g. [69,70]). Prior to sectioning, high-resolution photographs were taken, using a NikonH D26 digital reflex camera. Bones from the collection of the Zoologisches Museum Universität Zürich were additionally molded and cast. The molding agents used during this project were ElastosilH M 4503 and WackerH Catalyst T 46 (Ameba AG, CH-4056 Basel). The sampling sites of all extant and some fossil bones were mantled with either a two-component resin (TechnovitH 5071), superglue, or, in the case of some MNHSR-PV samples, paraffin wax to prevent additional damage during sawing and drilling. Core samples were embedded in transparent two-component epoxy resin (AralditeH 2020 XW 396/397, Huntsman A.M., B-Everberg). To produce thin sections, we cut the bones and the embedded core samples in half, using a rock saw (Diamant Boart) for fossil bones, a diamond blade saw for core samples (Hartig Export), and a band-saw (Inland DB-100 TM ) for bones of extant taxa. After letting the samples dry for at least 24 hours, cut surfaces were impregnated with transparent two-component epoxy resin (AralditeH 2020 XW 396, Huntsman A.M., B-Everberg), under vacuum conditions when possible. Most bones, however, were impregnated by applying a sufficient amount of resin to the cut surface and then placing the bone with the cut surface facing down on a piece of sheet protector attached to the tabletop. Once the resin had been allowed to harden completely, we polished the impregnated surfaces manually using a vertical or horizontal grinding disc (Crystal Master 8) and silicon carbide powder of variable grain sizes (220, 500, 800 grit, IEPCO AG, CH-Höri). One half of each core sample was additionally polished into a polished section sensu Sander [69] using 1200 grit silicon carbide powder and cerium(IV) oxide powder applied on leather. Polished samples intended for thin sections were then attached to roughened glass slides, using AralditeH. As a next step, we cut sections of 1 to 2 mm thickness using a diamond blade saw. These sections were reduced to a thickness of 100 to 150 mm with a sintered diamond grinding wheel (Wendt LM3). Then the sections were polished by hand to a final thickness of 60 to 80 mm. A cover slip was applied using UV glue (VitralitH 6127, Panacol AG, CH-Regensdorf). Finally, thin sections were examined by standard light microscopy (normal and cross-polarized) and polished sections under bright field illumination using a composite microscope (LeicaH DM2500 M). Photographs of histological features were taken with a LeicaH DFC 420 C digital camera mounted on the LeicaH DM2500 M composite microscope. Thin sections of all sampled bones have been returned to the respective repositories together with the long bones, with the exception of the sections of AMNH u1, AMNH u2 and AMNH u3.

Bone Compactness Profile Data
Bone compactness analysis was carried out for humeri and femora only, as these were the most abundant bones in the sample, especially for fossil taxa. Cross-sections were digitized using a flatbed scanner for big thin sections and a LeicaH DFC 420 C digital camera mounted on a LeicaH DM2500 M composite microscope for smaller thin sections. In most cases, the smaller thin sections had to be photographed partially, and images were then merged automatically in Adobe PhotoshopH CS5 (Adobe Systems Inc., San Jose, CA, USA). We transformed the photographs into binary images in Adobe PhotoshopH CS5 by marking bone in black and vascular spaces (medullary cavity, vascular canals, resorption spaces) in white. The binary images ( Figure 3, Figures S1.1-S1.5) were analyzed quantitatively with the software Bone Profiler Version 4.5.6 [16], which can extract several parameters of a bone compactness profile from crosssections. In brief, the program uses the bone center as the pivot point and estimates compactness in 60 radial sectors of 6u width, which are further divided into 51 concentric zones each. Compactness profiles were measured using the program's automatic mechanism for the detection of the bone center. However, in some cases it was necessary to adjust 'limit medulla' and 'section center' manually. Bone Profiler then plots the data using bone compactness (C) as a function of the distance from the bone center (d): From this function, four main parameters are extracted by the program: the relative distance from the section center to the point of most abrupt change in compactness, i.e., transition from medulla to cortex (P); the reciprocal of the slope at the inflexion point (S); the upper and the lower asymptote of the function (Max and Min respectively). The latter two correspond to the compactness in the outermost cortex (Max) and the compactness in the center of the medullary cavity (Min). Bone Profiler additionally extracts observed compactness (or global compactness, Cg), among other parameters.
Global compactness (Cg) is the parameter with the most relevance to this study, which aims at detecting general patterns in bone compactness. However, Canoville and Laurin [13] detected the most significant ecological signal in parameters S, P, Cc (compactness in the bone center), and Cp (compactness in the periphery of the section) in a study of amniote long bones. Additional plots were therefore produced for parameters S and P, but not for Cc and Cp, because all thin sections analyzed in this study had Cc values around 0 and Cp values around 1. S also showed an ecological signal in a fine-scale study of bone compactness in talpid humeri, in that values were significantly larger for fossorial than for non-fossorial forms [20].

Histological Features
The microstructure of long bones is described for two body mass categories of Cingulata, for two body mass categories of Folivora, and one category for Vermilingua. All descriptions follow the terminology of Francillon-Vieillot et al. [67]. In agreement with Werning [71], we refrain from using tissue-level terms like 'fibrolamellar bone'. Instead, we concentrate on the description of collagen orientation, vascular orientation, spatial arrangement of vascular canals, and osteocyte shape and arrangement. Body mass categories are based on data from Smith et al. [70] and White Tissue types external layer pf and la, av pf and la, almost av pf and la, av la, av majority of cortex dhb dhb, pf dhb, pf and la/wpl dhb, pf and la dhb, wpl innermost layer la, av pf & la, av la, av la, av la trabeculae pf and la pf and la pf and la la pf and la Body mass categories are based on data by Smith et al. [97] and White [98]. Abbreviations: av avascular; circ circular; dhb dense Haversian bone; elngtd elongated; irreg irregular; la lamellar; long longitudinal; pf parallel-fibered; rad radial; ret reticular; wpl woven, parallel and lamellar fiber arrangement. doi:10.1371/journal.pone.0069275.t002  [71]. A summary of histological features found in xenarthran long bones is provided in Table 2. In addition to the histological features described below, Sharpey's fibers indicating various muscle attachments are found in the majority of the bones. We do, however, refrain from describing and discussing them any further, as this would go beyond the scope of this paper. Cingulata: species .20 kg (all extinct). Long bones of large cingulates are characterized by a compact cortex and a dense network of thick trabeculae in the medullary region. In some samples, trabeculae had been previously destroyed by diagenetic processes, or were damaged during bone sectioning. The humerus of Pampatherium typum PIMUZ A/V 428 exhibits a high spatial organization of cancellous bone anteriorly, where three thin and long trabeculae are arranged parallel to each other. The cortices of the ulnae are thin in comparison to the cortices of other long bones included in this sample. Orientation of the vascular canals in the cortical tissue of cingulate long bones varies among the different skeletal elements. In the humeri, orientation of the vascular canals is variably longitudinal, radial, or reticular. In all other elements, vascular canals are mainly oriented longitudinally, with reticular canals occasionally occurring in the femur. The arrangement of vascular canals is, however, irregular in all elements. The cortices of the radii of Glyptodon clavipes, PIMUZ A/V 463 and PIMUZ A/ V 465, show expansion of the cancellous medullary bone into the posterior region of the cortex. Most of the cortex has been remodeled and dense Haversian bone is visible under crosspolarized light. Because the vascular canals are variously orientated in some areas of the bones, the remodeled tissue locally appears irregular (Figure 4a-b). Due to the preservational state, osteocyte lacunae are hardly visible. A thin layer of parallel-fibered and, in certain areas, lamellar bone tissue surrounds the exterior cortex. The medulla is lined by a thin strip of primary lamellar bone, and trabeculae show parallel-fibered and lamellar tissue organization.
Cingulata: species ,20 kg (all extant). The long bones of all sampled armadillos possess a thick, compact cortex. Only Dasypus hybridus PIMUZ A/V 4799 and Tolypeutes tricinctus PIMUZ A/V 4806 show thin cortices, and these two specimens were identified as juveniles on the basis of the level of epiphyseal fusion. The cortex shows three major layers of different bone tissues (Figure 5a-b). A layer of poorly vascularized parallel-fibered bone tissue lines the cortex at the periphery. This layer is of variable thickness within and between bones and sparsely vascularized. In most cases, a very thin layer of lamellar bone has been deposited external to this layer of parallel-fibered bone. This outer layer (OCL: outer circumferential layer or EFS: external fundamental system as discussed by Chinsamy [72]) is clearly delimited by a line representing a LAG or annulus. Within the OCL up to five closely spaced LAGs (or annuli) were found in adult specimens of armadillos (all except PIMUZ A/V 4799 and PIMUZ A/V 4806).
Arrangement of vascular canals in the parallel-fibered middle layer is random, and canals are partly oriented longitudinally, radially and reticularly. Only in certain areas of the humeri of   (Figure 6a-b). In some cases, remains of the primary tissue are still visible, suggesting that it had been of parallel-fibered and lamellar organization. In the case of MHNSR-PV 1002, MHNSR-PV 1006 and MHNSR-PV 1007 the primary matrix is composed of a mixture of woven, parallelfibered and lamellar tissue (Figure 6c-d). In areas of woven bone, osteocyte lacunae are round and randomly oriented, whereas in areas of parallel-fibered or lamellar bone osteocyte lacunae are elongated, flattened and arranged in a parallel manner. MHNSR-PV 1002 and MHNSR-PV 1006 also show strong laminar organization of the tissue in the exterior part of the cortex, with secondary osteons being elongated (Figure 6e-f). In this layer, two LAGs can be identified for MHNSR-PV 1006 (Figure 6e-f). The thin strips of bone lining the medulla and the trabeculae are composed of parallel-fibered and lamellar tissue. Trabeculae vary in thickness between samples but form a dense network in all elements sectioned.   laterally, the laminated organization of the tissue is lost and many obliquely or longitudinally sectioned secondary osteons can be found (Figure 7a-b). The density of the Haversian tissue with overlapping generations of secondary osteons (Figure 7c-d) indicates that the entire primary bone tissue has been remodeled. Lamellar bone lines the cavities in the cancellous area of the lateral cortex. AMU-CURS 128 and Megalocnus rodens AMNH u2 also show that the folivoran femur is locally cancellous in the lateral region.
Folivora: species ,20 kg (all extant). Long bones of smallsized folivoran specimens possess thick compact cortices and relatively loose networks of thick trabeculae. Trabeculae and the thin strip of secondary bone lining the medulla are composed of lamellar bone. The cortices are vascularized by mostly longitudinally oriented canals forming primary osteons, and many radially oriented canals in the external layer. Vascular canals are small, but a few large secondary cavities are found in the external layers of the cortices. These larger cavities could represent resorption cavities, since cement lines surrounding them are visible under cross-polarized light. All specimens show heavily remodeled cortical bone tissue, but they differ in the extent of remodeling (Figure 8a Vermilingua. The external cortices of anteater long bones are vascularized by circular, longitudinal and reticular canals of primary osteons, and these canals are arranged in concentric rows (Figure 9), thus resulting in a laminar organization. This laminar organization is most pronounced in Tamandua tetradactyla NMB 10420 in that vascular canals are larger than in the other two specimens of the same species (PIMUZ A/V 4807 and PIMUZ A/ V 4808) and the bright lines between laminae are clearly visible. This does, however, not imply that NMB 10420 is older than the other two specimens of T. tetradactyla. On the contrary, the cortex of PIMUZ A/V 4807 shows more laminae than NMB 10420 and the bone tissue of PIMUZ A/V 4808 has been secondarily remodeled, both indicating that NMB 10420 represents a younger specimen. This is in accordance with the less advanced state of epiphyseal fusion and the smaller size of NMB 10420.
In the humeri and femora of T. tetradactyla NMB 10420 and T. tetradactyla PIMUZ A/V 4807, arrangement of vascular canals is random on the lateral side and radial on the medial side of the bone. The internal part of the cortex is vascularized by randomly arranged longitudinal canals. Trabeculae in the cancellous medullary region are thick in NMB 10420, PIMUZ A/V 4807 and T. tetradactyla PIMUZ A/V 4808, but fine in Myrmecophaga tridactyla ZMZ 11119. In all bones, the trabeculae form a dense meshwork throughout the medulla.
The tissue in the cortices of anteater long bones is mainly composed of a mixture of woven, parallel-fibered and lamellar bone conforming to the arrangement of the vascular canals ( Figure 9). Up to five LAGs can be identified in the femur of ZMZ 11119, whereas the humerus only shows one distinct LAG. NMB 10420 shows no LAGs, but PIMUZ A/V 4807 and PIMUZ A/V 4808 show one LAG in the humerus and two LAGs in both humerus and femur, respectively. As already indicated above, secondary remodeling is not advanced in NMB 10420 and PIMUZ A/V 4807, and the few secondary osteons formed are mainly found in the medial and lateral regions of the bone. In PIMUZ A/V 4808, secondary remodeling is more advanced, with scattered secondary osteons occurring throughout the internal cortices. The internal layer of the cortices of ZMZ 11119 is heavily  (Figure 9), but patches of primary bone matrix are still visible between secondary osteons. Trabeculae are composed of parallel-fibered and lamellar bone tissue.

Bone Compactness Analysis
Bone global compactness of xenarthran humeri and femora. Bone global compactness (Cg) is highly variable across the whole group, within subgroups and also between individuals of the same species (Figure 10, Tables S2.1 and S2.2). The variation found in this study is further increased by the inclusion of the three juvenile specimens in this study. However, upon excluding these specimens, values still range between 0.600 (ZMZ 17223) and 0.879 (PIMUZ A/V 4808) for the humerus, and between 0.412 (ZMZ 11151) and 0.877 (PIMUZ A/V 438) for the femur. All adult specimens show a Cg value greater than 0.400, indicating that $40% of the cross-sectional area is composed of dense mineralized tissue. Cg values for anteater humeri are on average found to be slightly higher than those found for humeri of Cingulata and Folivora. However, considering the overall pattern of variation, this could be an artifact of the small sample size of Vermilingua. Furthermore, this pattern is no longer found when values are plotted according to species instead of individuals. In general, no distinct patterns in relation to phylogeny, body size or  The Cg values found for xenarthran femora do not suggest any striking differences between categories of body size or locomotion types either. However, the femora of sloths and anteaters show higher average values than those of cingulates. The same pattern is also found when plotting values according to species instead of individuals. Considering that sloths and anteaters form the monophyletic clade Pilosa, this may suggest a phylogenetic pattern for bone compactness in the femur.
A plot of Cg values of humeri and femora from individuals for which both bones were available is given in Figure 11. This group includes one fossil species (Lomaphorus ornatus). The distribution of the data points reflects the variation detected when plotting values for the humerus and the femur individually ( Figure 10). Overall, a positive correlation is found between the Cg of the humerus and the Cg of the femur, suggesting that these long bones do not represent independent modules. The data for extant sloths and for Myrmecophaga tridactyla falls on the identity line, suggesting that humeral Cg and femoral Cg in these species are equal. Data points for armadillos and Tamandua tetradactyla, on the other hand, are found below the identity line, suggesting an allometric relationship, with Cg in the humerus being higher than in the femur. Lomaphorus ornatus is the only species that plots above the identity line, showing a higher compactness in the femur than in the humerus.
P and S values of adult xenarthran humeri. Bone Profiler values for the relative distance from the section center to the inflexion point (P) and for the reciprocal of the slope at the inflexion point (S) [16] in xenarthran humeri do not show any distinct patterns related to phylogeny, body size or locomotion ( Figure 12; for exact values see Table S2.1). Variation of P values is moderate across Xenarthra, as well as within Cingulata, Pilosa and Vermilingua. Values for S, on the other hand, show extensive variation across Xenarthra, both within the three subgroups and within genera, signifying that the transition from the medulla to the compact cortex is gradual in some and abrupt in other taxa. No pattern related to phylogeny, body size or locomotion can be detected.
P and S values of adult xenarthran femora. Values for both P and S of the femur are variable within the three subgroups Cingulata, Folivora and Vermilingua, and within genera contained therein (Figure 13, for exact values see Table S2.2). However, P values are on average higher for femora of members belonging to Cingulata than for members of either of the two groups within the sister clade Pilosa. This signifies that the relative distance from the center of the bone to the transition point from cancellous medullary region to cortical compacta is greater in Cingulata than in Folivora and Vermilingua. In terms of crosssectional geometry, this indicates that the compact cortex of the cingulate femur occupies a smaller proportion of the crosssectional area than the cortex of a pilosan femur. This is consistent with the slightly lower average value of Cg found for Cingulata in comparison with the average values found for Pilosa. The S values show a reverse and more distinct pattern, with the average value for Pilosa being higher than the average value for Cingulata. The transition from a relatively cancellous medullary region to a relatively compact cortex is therefore more abrupt in cingulate femora than in folivoran or vermilinguan femora. In summary, these results suggest that a phylogenetic pattern is contained in P and S values for the femora.
Ontogenetic changes in bone compactness. Figure 14 shows a comparison of Cg values for adult and juvenile specimens of Dasypus hybridus, Tolypeutes tricinctus, and Tamandua tetradactyla. In armadillos, Cg values of juveniles are found to be smaller than Cg values of adults for both humerus and femur. In Dasypus hybridus, juvenile Cg for humeri amounts to 67.8% of the adult Cg value, while in Tolypeutes tricinctus it amounts to 61.9% of the adult Cg. For femora, juvenile Cg values represent 60.9% and 58.3% of the adult Cg value respectively. These results thus indicate that Cg in armadillo long bones increases extensively during development. In the anteater Tamandua tetradactyla juvenile and adult Cg values differ by a maximum of about 11% for both bones. The Cg value of the juvenile specimen is lower than both adult Cg values in the humerus and higher than both adult Cg values in the femur. This suggests that bone compactness in the humerus of Tamandua tetradactyla increases during development, whereas compactness of the femur decreases. However, the ontogenetic series for Dasypus hybridus, Tolypeutes tricinctus and Tamandua tetradactyla each include two growth stages only. A thorough investigation into ontogenetic changes in bone compactness of xenarthran long bones would be worthwhile in order to test whether the pattern found for T. tetradactyla is not just a product of chance but applies to vermilinguans in general, or even to all members of Pilosa.

Xenarthrans in the Context of Mammalian Bone Histology
Xenarthrans possess a long bone microstructure consistent with other placental mammals in general [2,5,21,22,24,25,[73][74][75][76][77]. Table 3 summarizes histological features described for long bones of four clades of placental mammals. Although the descriptions for some of these clades are based on limited sampling, this summary shows hypothetical patterns that can be further investigated. In terms of vascularization, xenarthran long bones show a pattern more similar to that of Euarchontoglires. Vascular canals are oriented longitudinally, reticularly or radially and they are arranged in an irregular manner. Only anteater long bones exhibit areas with a circular arrangement of vascular canals. According to the histological descriptions summarized in table 3 [5,22,24,25,76,77] members of Laurasiatheria and Afrotheria, on the other hand, often possess circumferential vascular canals and a laminar to plexiform vascular network. The primary bone tissue of xenarthran long bones comprises a mixture of woven, parallelfibered and lamellar bone and is thus more similar to that of Laurasiatheria and Afrotheria than to the primary bone tissue of Euarchontoglires. The latter can be fully lamellar or of the woven and parallel-fibered type. However, considering that the available data on the bone histology of Euarchontoglires is limited and that a mixed bone tissue (woven plus lamellar and/or parallel-fibered bone) was found in bones of stem eutherians [75], the bone tissue of Euarchontoglires is probably not so different from that of the remaining clades of placental mammals. Haversian bone has been found in all four clades of placental mammals. Xenarthra is, however, the only clade in which all investigated subgroups show this heavily remodeled bone tissue.
Information on the presence of LAGs within placental mammals is incomplete, but LAGs have been documented for at least some members of all four major clades (Table 3). Zonal bone tissue, indicating periodically interrupted deposition of bone during later ontogeny, has been documented for marine mammals (e.g. cetaceans and pinnipeds [78]), mammals of temperate regions [78], hibernators [75,78], polar bears [72], and ruminants from a variety of environments [25]. Additionally, Curtin et al. [77] investigated bone microstructure in fossil and extant neonatal elephantids and found hypercalcified lines of arrested osteogenesis in their femora and tibiae. The extinct Plio-Pleistocene bovid Myotragus [76] also showed zonal deposition of bone tissue, and Köhler and Moyà-Solà [76] argued that interrupted growth in Myotragus could have been an adaptive strategy in a resourcelimited environment. Sander and Andrássy [24] found LAGs in the large majority of their sample of large fossil mammals from Germany and thus proposed that LAGs might be a universal feature of large mammals. Indeed, flexible growth rates, indicated by the deposition of LAGs, are suggested to be plesiomorphic to the mammalian lineage and could have played a major role in the diversification pattern of mammals after the K/T extinction [75]. Hurum and Chinsamy [75] further suggested that eutherian mammals developed rapid, uninterrupted growth during the Paleocene, spurring an increase in body size, and that modern eutherians retained the plesiomorphic condition of interrupted growth (see also Chinsamy and Hurum [74]). Moreover, a recent study on seasonal growth in ruminants suggests that flexible growth is a universal feature of homeothermic endotherms [25].
The studied sample of long bones confirms that flexible growth rates, and thus LAGs, are common in Xenarthra. All extant specimens representing skeletally mature adults, i.e. those with fused epiphyses, show LAGs in the cortices of their long bones. The only exception to this is Choloepus didactylus ZMZ 17223, which is interpreted to be a specimen of advanced age since its bone microstructure is completely remodeled. In adult armadillo long bones the occurrence of LAGs is restricted to the outermost layer of the cortex, thus representing an OCL or EFS. In bones of anteaters and sloths, LAGs are not restricted to this outer layer but are also found within the middle part of the cortex, provided that the cortex is not completely remodeled. This pattern can be explained by taking the age at which these species reach sexual maturity into consideration. Most mammals reach sexual maturity shortly after achieving skeletal maturity, i.e. adult size [79], but slow circumferential periosteal growth of the bones in diameter, including the deposition of closely spaced LAGs, can continue beyond this point [80]. The extant folivoran taxa included in this study represent species that reach sexual maturity after two or more years [81,82]. Thus it is no surprise that the specimens assigned to these species show LAGs both in the middle part of the cortex and towards the periphery, as long as they are not older adults as is the case of Choloepus didactylus ZMZ 17223. The extant species of cingulates included in this study all reach sexual maturity within or shortly after their first year [83] and a clear line separates the inner remodeled part of the cortex from the thin outer parallel-fibered or lamellar layer in adult specimens. This line could mark the attainment of sexual maturity, after which bone tissue was only very slowly deposited and LAGs are closely spaced.
The presence of LAGs in MHNSR-PV 1006 raises the question of whether LAGs were formed in other fossil specimens but are simply masked by extensive secondary remodeling. The presence or absence of LAGs could provide information on the attainment of sexual maturity and the growth dynamics in extinct taxa. Since MHNSR-PV 1006 represents a pilosan taxon it can be assumed that LAGs are also common among extinct sloths and anteaters, and that these taxa probably reached sexual maturity after two or more years. Considering the size of glyptodonts and pampatheriids compared to their extant relatives, it would be of interest to find out whether they reached sexual maturity at about the same age or later than modern cingulates. Wolf [39,56] and Hill [52] found surface-parallel growth marks, in some cases possibly representing true LAGs, in the superficial layer of the cortex of most of the fossil osteoderms sampled. It is unclear, however, how growth of osteoderms relates to growth of long bones in these animals.

Secondary Remodeling and Size
The cortices of xenarthran long bones are well vascularized and often heavily remodeled. Enlow and Brown [5] noted that dense Haversian bone is a common feature found in mammalian bones, but is absent in some groups such as monotremes, marsupials, eulipotyphlans and rodents. The thin sections produced for this study show that secondary remodeling is found in the long bones of all xenarthrans. Armadillos exhibit a special type of Haversian bone in that certain areas show irregularly shaped secondary osteons. Among extant xenarthrans, sloths show the highest extent of secondary remodeling, and only fossil forms show a comparable extension of Haversian bone in the cortex.
The histology of cingulate and folivoran long bones included in this study differed between large (.20 kg) and small (,20 kg) taxa. It is noteworthy that for folivorans and cingulates included in this study, these body mass categories also correspond to the extinct/extant status of each taxon; i.e., large taxa represent extinct forms and small taxa represent extant forms. Secondary remodeling in bones of fossil cingulates and folivorans was much more extensive than in those of extant species, with the exception of the bones of Choloepus didactylus ZMZ 17223. Although the microstructure of fossil bones was greatly masked by diagenetic alteration of the tissue, it was still possible to detect dense Haversian bone that most likely constituted the entire cortices of these bones. The extent of secondary bone remodeling could be a feature in which extinct (and thus large) cingulates and folivorans differ from their extant (and thus small) relatives. A possible scenario is that extinct forms of Cingulata and Folivora lived longer than extant forms. A prolonged life span could also explain differences in size between extinct and extant xenarthrans, and increased size would have resulted in increased loading and consequently more remodeling in the larger (extinct) forms. On the other hand, the observed pattern could also be a result of chance, i.e. fossil specimens and Choloepus didactylus ZMZ 17223 represent more mature individuals than all remaining extant specimens. However, because we possess no data on the chronological age of fossil and extant specimens included in this study, other than that the epiphyses are fused in all but three specimens (PIMUZ A/V 4799, PIMUZ A/V 4806 and NMB 10420), this question remains unanswered. Bone histology in an ontogenetic series of fossil and extant xenarthran taxa could possibly reveal whether differences in the rate of secondary remodeling or the life span of extinct and extant taxa really exist.

Bone Compactness Parameters
The parameters Cg (global compactness), S (reciprocal of slope at inflexion point) and P (relative distance from center of bone to inflexion point) of the bone compactness profiles of xenarthran long bones show no ecological signal of the sort recently detected in humeri of a large sample of amniotes [13], a signal apparently lacking in talpid moles [20]. In fact, the compactness parameters of xenarthran humeri show neither a pattern related to locomotion or body size, nor a pattern related to phylogeny. However, the structural component sensu Cubo et al. [7] was not investigated in this study. A pattern possibly related to phylogeny is found for all three parameters of xenarthran femora. Phylogenetic signal has also been detected in these parameters for lissamphibian long bones [17] and in the radius of amniotes [10]. Cg and S values for femora of members of Pilosa are higher than values found for members of Cingulata. For P values, an inverse relationship is found with values being higher in Cingulata than in Pilosa. These results suggest a phylogenetic pattern in the bone compactness profile of the femur and also show that the parameters Cg, S and P are linked. The average S values for cingulate femora, for example, are lower than the S values found for Pilosa when Cg values of cingulate femora are lower than Cg values found for Pilosa. A negative relationship is found between values of P and Cg. This linkage of parameters indicates that not only the thickness of the compacta but also the transition (abrupt vs. gradual) from medulla to cortex have a direct influence on bone compactness.

Allometry in Bone Compactness
The pattern found in the relationship between humeral and femoral bone compactness in xenarthrans could possibly be related to locomotion and life habit. Figure 15 shows the traced history of the relationship between humeral and femoral bone compactness. The 1:1 relationship between bone compactness in the humerus and femur in modern sloths can be explained by the suspensory arboreality of these xenarthrans. In locomotory terms, the forelimb and the hind limb are equally important [84,85] and therefore one could assume that they bear the same amount of stress, which then induces a more or less equal compactness in the humerus and femur. Armadillos, on the other hand, are scratch diggers [86] and show many specializations for digging in their forelimbs. They possess large claws, tuberosities for strong muscular insertions and long lever arms for the line of action of the principal muscles involved in digging [87]. Indices of the armadillo forelimb, for example the index of fossorial ability (IFA), have shown to correlate well with digging, while indices of the hind limb are linked to body size [87,88]. The allometric relationship between humeral and femoral compactness found in this study, whereby humeral compactness is higher, is therefore most likely linked to digging and the associated morphological adaptations found in the humerus. Only members of Tolypeutini were noted to be non-diggers (e.g. [89]).
In the analysis of humeral and femoral compactness, Lomaphorus ornatus plotted as an outlier to all other cingulates ( Figure 11). Femoral compactness in this glyptodont was much higher than the compactness of the humerus. The humeri of glyptodonts have been shown to lack features that would indicate digging ability [90], and Quintana [91] noted that the general anatomy of these extinct xenarthrans prevented them from a fossorial habit. The allometric relationship between humeral and femoral bone compactness of the glyptodont L. ornatus found in this study fits with this scenario.
The position of the two anteaters Myrmecophaga tridactyla and Tamandua tetradactyla in Figure 11 question the proposed relationship of relative humeral and femoral bone compactness with locomotion and life habit in xenarthrans. M. tridactyla plotted along the same line as sloths, whereas the data point for T. tetradactyla lies on the same line as digging armadillos. Anteaters are generally considered hook-and-pull diggers [86] that apply their specialized digging method for obtaining food rather than constructing burrows [92]. M. tridactyla, the giant anteater, exhibits an almost wholly terrestrial habit, whereas T. tetradactyla moves freely between terrestrial and arboreal environments [93]. It would therefore fit better with the proposed locomotion-related pattern if M. tridactyla plotted along with digging armadillos and T. tetradactyla with modern sloths. Including the almost fully arboreal silky anteater (Cyclopes didactylus) in future studies could possibly shed more light on whether relative compactness of the humerus and of the femur in xenarthrans is indeed linked to locomotion. In this context it would also be important to include mylodontid ground sloths such as Scelidotherium and Glossotherium, as limb bone proportions and resistance to bending forces suggest that these taxa were capable of performing strenuous activities like digging [94,95]. Rare taxa such as the giant armadillo (Priodontes) or the extinct aquatic sloth Thalassocnus are difficult to obtain for histological analysis but might also reveal specialized patterns.