Bone cortical compactness in ‘tree sloths’ reflects convergent evolution

Abstract

Bone remodeling, one of the main processes that regulate bone microstructure, consists of bone resorption followed by the deposition of secondary bone at the same location. Remodeling intensity varies among taxa, but a characteristically compact cortex is ubiquitous in the long bones of mature terrestrial mammals. A previous analysis found that cortical bone in a few ‘tree sloth’ (Bradypus and Choloepus) specimens is heavily remodeled and characterized by numerous immature secondary osteons, suggesting that these animals were remodeling their bones at high rate until late in their ontogeny. This study aims at testing if this remodeling is generally present in ‘tree sloths’, using a quantitative analysis of the humeral cortical compactness (CC) among xenarthrans. The results of the investigation of humeral diaphyseal cross‐sections of 26 specimens belonging to 10 xenarthran species including specimens from both extinct and extant species indicate that in ‘tree sloths’ the CC is significantly lower than in the other sampled xenarthrans. No significant difference was found between the CC of the two genera of ‘tree sloths’. Our results are consistent with the hypothesis that the cortical bone of ‘tree sloths’ in general undergoes intense and balanced remodeling that is maintained until late (possibly throughout) in their ontogeny. In the light of xenarthran phylogeny, low CC represents another convergence between the long‐separated ‘tree sloth’ lineages. Although the exact structural and/or functional demands that are associated with this trait are hitherto unknown, several hypotheses are suggested here, including a relationship to their relatively low metabolism and to the mechanical demands imposed upon the bones by the suspensory posture and locomotion, which was independently acquired by the two genera of ‘tree sloths’.

Introduction

Bone remodeling consists in the resorption of bone tissue, i.e. the formation of resorption cavities (by osteoclasts), followed by deposition of secondary bone at the same location (by osteoblasts), forming secondary osteons (Francillon‐Vieillot et al. 1990; de Paula & Rosen, 2013). Remodeling involves a continuous replacement and a constant renewing of bone material. It is a process that can completely or partially affect primary, as well as secondary bone, and both the compact and spongy bone (Francillon‐Vieillot et al. 1990). In addition to bone modeling, i.e. primary formation, bone remodeling is the main mechanism that regulates bone microstructure and is therefore directly related to morphogenesis and growth at early life stages (Francillon‐Vieillot et al. 1990; Cambra‐Moo et al. 2014). Bone remodeling is furthermore necessary to maintain the structural integrity of bone and is involved in the homeostasis of calcium and phosphate (Francillon‐Vieillot et al. 1990; de Paula & Rosen, 2013). It is also involved in bone healing processes as a dramatically accelerated response in regions of bone injury (Shih & Norrdin, 1985). Bone remodeling is a lifelong process that influences bone structure until the oldest ages (Amprino & Godina, 1947). Intensity of remodeling and the resulting secondary osteons density varies among mammals, from no osteons at their long bone midshaft to a large number of them, which form in the latter case dense Haversian tissue (Kolb et al. 2015). Studies on human long bones indicate that young, growing individuals (before ca. 10 years old) are characterized by high levels of porosity at midshaft, resulting from a large number of erosion cavities throughout the cortex (Jowsey, 1960; Cambra‐Moo et al. 2014). In adult humans, Haversian tissue eventually occupies most of the humeral cortex at midshaft (Cambra‐Moo et al. 2014). But in deer, for instance, it is mostly the inner cortex that is heavily remodeled (Amson et al. 2015a). The cortex of long bones of adult (non‐osteoporotic) mammals usually feature a compact structure, comprising the so‐called compacta, or compact bone, with relatively few resorption cavities and immature secondary osteons (Amprino & Godina, 1947; Francillon‐Vieillot et al. 1990; Cambra‐Moo et al. 2014; Mitchell, 2016). Investigation of bone remodeling can yield ecological, functional, and phylogenetic information (Cubo et al. 2005; Mitchell et al. 2017), and it is thus of high interest to explore and compare this microanatomical feature of tetrapod long bones. The extent and pattern of distribution of Haversian tissue has hence been investigated in various clades, e.g. in relation to locomotion in primates (Schaffler & Burr, 1984) and birds (Ponton et al. 2007; Mitchell et al. 2017), and more generally in relation to lifestyle (Mitchell, 2016).

Placental mammals have been classified into four major clades: the afrotheres, the laurasiatheres, the euarchontoglires, and the xenarthrans (Asher et al. 2009). The xenarthrans are characterized by, among other features, the presence of extra joints in their spine (the eponymous xenarthrales), low body temperatures, and extremely low basal metabolic rates for their body size (McNab, 1982; Gilmore et al. 2000; Vendl et al. 2016). The xenarthran clade is divided into three orders that diverged ca. 60 Ma (Delsuc et al. 2001; Springer et al. 2003; dos Reis et al. 2012): Cingulata (armadillos and extinct close relatives), Vermilingua (anteaters), and Tardigrada (sloths) (Fig. 1). Tardigrada and Vermilingua form the clade Pilosa. The large and diverse group of extinct xenarthrans is characterized by a rich fossil record in which there are abundant remains of the long bones. In contrast, the extant diversity of the clade is impoverished and comprises only 21 species of armadillo, four anteaters, and six sloths (Möller‐Krull et al. 2007; Vizcaíno & Loughry, 2008). Whereas extant xenarthrans include only small to mid‐sized species, some species of extinct xenarthrans reached gigantic sizes. Xenarthrans include forms with subterranean, terrestrial, semi‐arboreal, arboreal, and aquatic lifestyles (Amson & Nyakatura, 2017). It is today well corroborated that the two genera of ‘tree sloths’ do not form a monophyletic group (hence the use of quotation marks herein), as one of the two genera, the two‐toed sloth Choloepus Illiger, 1811, is more closely related to the other sloths, the likewise non‐monophyletic ‘ground sloths’ (Fig. 1; Gaudin, 2004; Gibb et al. 2016). The ‘tree sloths’ are fully arboreal and are confined to Central and South America. The two constitutive genera, Choloepus and Bradypus Linnaeus, 1758 (three‐toed sloth), are morphologically distinct (Gaudin, 2004). Along with the ‘ground sloths’, Choloepus belongs to the Megalonychidae, and Bradypus, as the sister group to all other sloths, forms the Bradypodidae (Gaudin, 2004; Amson et al. 2016). The divergence of the two lineages dates back as far as 30 Ma according to a molecular clock (Gibb et al. 2016), and possibly 40 Ma based on the fossil record (Gaudin, 2004). Both genera of ‘tree sloths’ exhibit several differences in their biology, ecology, and morphology (Hayssen, 2009a,b, 2010, 2011), but have, however, convergently evolved many anatomical similarities as solutions to similar environmental challenges related to suspensory posture and locomotion (Nyakatura, 2012). These include the musculoskeletal specializations allowing a suspensory posture and locomotion (Mendel, 1985; Nyakatura, 2012), an aberrant cervical count (Buchholtz & Stepien, 2009), the retention of water and growth of algae in the hair (Aiello, 1985), and small and variable semicircular canals (Billet et al. 2012, 2013), with a weakly tilted lateral semicircular canal (Coutier et al. 2017).

Figure 1.

Time‐calibrated phylogeny of the eight xenarthran genera included in this study. Plot produced with library strap (Bell & Lloyd, 2015) of r version 3.4.0. LCA, last common ancestor of ‘tree sloths’.

The broad investigation into xenarthran long bone microstructure published by Straehl et al. (2013) indicates a general consistency between their long bone microstructure and that of other placental mammals. The pattern detected on the cortex of xenarthran long bones, however, includes a particularly remodeled tissue and high degree of vascularization. Those authors state that Xenarthra is indeed the only clade in which heavily remodeled bone tissue is consistently found in all genera. Nevertheless, it is the ‘tree sloths’ that show the most extensively remodeled bone (Straehl et al. 2013), followed by ‘ground sloths’ (which can even be taxonomically diagnostic; Amson et al. 2015b). Moreover, large and numerous resorption cavities or immature secondary osteons are described on the heavily remodeled cortical tissue of the long bones of skeletally mature ‘tree sloths’, therein represented by three specimens (and eight skeletal elements in total). In contrast, the cortices of skeletally mature Vermilingua and Cingulata were found mostly to lack large resorption cavities or immature secondary osteons (Straehl et al. 2013). A description of a ‘tree sloth’ femur and the radius of an armadillo by Amprino & Godina (1947) is consistent with these observations. Resorption cavities and immature secondary osteons were also reported in ‘ground sloths’ (Straehl et al. 2013) and, as in ‘tree sloths’, were interpreted as indicative that secondary bone remodeling was in progress at the death of the sampled individuals.

With the present analysis, we intend to test whether this pattern is generally present in ‘tree sloths’ (and absent in their close relatives). To increase sample size while avoiding the use of a destructive method, we use cortical compactness (CC, see Materials and methods below), which can be acquired with high resolution computed tomography (CT) scanning, to assess quantitatively the bone structure of xenarthrans. A low CC would correspond to the phenotype of the ‘tree sloths’ sampled by Straehl et al. (2013), i.e. strong remodeling that is in progress, whereas a high CC would be expected for the more usual compact bone, found in other xenarthrans, and mammals in general.

Materials and methods

Specimens

We studied the humerus to compare a homologous structure and to use one of the skeletal elements sampled by Straehl et al. (2013). A total of 26 specimens, corresponding to skeletally mature (as indicated by complete epiphyseal fusion), wild‐caught (for the extant species), and non‐pathological xenarthrans were included in this analysis (to avoid biases related to growth stage or injury; Shih & Norrdin, 1985). The set of specimens comprises ten ‘tree sloths’ and 16 ‘other xenarthrans’, the latter comprising three ‘ground sloths’, nine anteaters, and four armadillos (Fig. 1, Table 1). The species of ‘other xenarthrans’, used for comparison, were selected for their size to approach that of ‘tree sloths’: the ‘ground sloths’ Hapalops Ameghino, 1887 (Santa Cruz Formation, Early Miocene; Perkins et al. 2012), Parocnus brownii Matthew, 1931 (Las Breas de San Felipe tar pit, Cuba, Holocene; Jull et al. 2004) and Thalassocnus littoralis McDonald & Muizon, 2002 (Pisco Formation, Late Miocene; Amson et al. 2014), Myrmecophaga tridactyla Linnaeus, 1758, Tamandua tetradactyla Linnaeus, 1758, and Priodontes maximus (Kerr, 1792). Sex was unknown in most cases and is not provided here.

Table 1.

List of the 26 analyzed specimens and the corresponding measured cortical compactness (CC). For abbreviations of specimen numbers, refer to Institutional abbreviations in Materials and methods

Species	Specimen number	Data source	Cortical compactness, %
Bradypus sp.	ZMB‐MAM 33806	CT‐scan	98.911
Bradypus sp.	ZMB‐MAM 76147	CT‐scan	99.760
Bradypus variegatus	ZSM‐AM 1029	CT‐scan	99.198
Bradypus tridactylus	NMB 10488	Thin‐section	98.644
Bradypus torquatus	ZSM 1903‐9534	CT‐scan	98.978
Bradypus torquatus	ZMZ 11102	Thin‐section	96.857
Choloepus didactylus	ZMB‐MAM 35825	CT‐scan	98.642
Choloepus didactylus	ZMZ 17223	Thin‐section	97.757
Choloepus didactylus	MNHN 1999‐1062	CT‐scan	98.940
Choloepus didactylus	ZMB‐MAM 102636	CT‐scan	96.331
Mean for the ‘tree sloths’			98.401
Hapalops sp.	MNHN.F.SCZ 162	CT‐scan	99.583
Thalassocnus littoralis	MNHN.F.SAS 1605	CT‐scan	99.943
Parocnus brownii	AMNH u3	Thin‐section	99.035
Myrmecophaga tridactyla	MNHN 2005‐269	CT‐scan	99.994
Myrmecophaga tridactyla	ZMB‐MAM 77025	CT‐scan	99.987
Myrmecophaga tridactyla	SMF 43046	CT‐scan	99.842
Myrmecophaga tridactyla	ZMZ 11119	Thin‐section	99.807
Tamandua tetradactyla	ZLHUB‐1818‐20	CT‐scan	99.990
Tamandua tetradactyla	ZMB‐MAM 38396	CT‐scan	99.952
Tamandua tetradactyla	ZMB‐MAM 81448	CT‐scan	99.994
Tamandua tetradactyla	PIMUZ A/V 4807	Thin‐section	99.941
Tamandua tetradactyla	PIMUZ A/V 4808	Thin‐section	99.952
Priodontes maximus	ZMB‐MAM 6163	CT‐scan	99.992
Priodontes maximus	ZMB‐MAM 36422	CT‐scan	99.995
Priodontes maximus	SMF 42286	CT‐scan	99.967
Priodontes maximus	CeNaK‐S‐6038	CT‐scan	99.757
Mean for the ‘ground sloths’ and ‘non‐sloth xenarthrans’			99.858

Institutional abbreviations

AMNH, American Museum of Natural History, New York, NY, USA; CeNaK, Centrum für Naturkunde, Hamburg, Germany; MNHN, Muséum national d'Histoire naturelle, Paris, France; NMB, Naturhistorisches Museum, Basel, Switzerland; PIMUZ A/Z, Paläontologisches Institut und Museum der Universität Zurich, Switzerland; SMF, Smithsonian National Museum of Natural History, Washington, DC, USA; ZLHUB, Zoologische Lehrsammlung HU‐Berlin, Berlin, Germany; ZMB, Museum für Naturkunde, Berlin, Germany; ZMZ, Zoologisches Museum der Universität Zurich, Switzerland; ZSM, Zoologische Staatssammlung, Munich, Germany.

Data acquisition and processing

A first dataset of 19 specimens was acquired by CT‐scanning (see Table 1). The remaining seven humeri were analyzed based on images of conventional thin‐sections (see below). The CT‐scanners used were a Tomoscope Synergy Twin (Experimental Radiology Lab, Institute of General and Interventional Radiology, Jena University Hospital), a v|tome| × L 240 (GE Sensing & Inspection Technologies phoenix|x‐ray, AST‐RX plate‐forme, Muséum national d'Histoire naturelle, Paris, France), and a Nanotom m (phoenix|x‐ray, Zoologische Staatssammlung, Munich). The data from the CT‐scans corresponds to 16‐bit tif image stacks of the specimen's humeri with resolutions that range from 10.32 to 27.78 μm. The image stacks were visualized and processed with Fiji version 1.51 g, an image processing package distributed by imageJ (Schneider et al. 2012). Prior to the measurement of the bone CC, each image stack was processed with the same scheme, consisting of the standard orientation of the bone (centers of the metaphyses were aligned along the Z‐axis), the extraction of a specific cross‐section at the level of maximal cortical thickness, and the binarization of this cross‐section [‘optimise threshold’ function of the plugin boneJ (Doube et al. 2010) version 1.4.2; Figs 2 and 3]. In all specimens, the site of maximal cortical thickness was found to be just distal to the nutrient canal (see Fig. 2). In the case of Choloepus didactylus ZMB‐MAM 102636, a small part of the diaphysis was not scanned, so that the slice showing the thickest cortex could only be localized proximal to the nutrient canal (but qualitative observation revealed little difference in the CC along the shaft). If present, other bones and non‐bone material were deleted (i.e. given a gray value of 0). The fossil humeral cross‐section of Hapalops MNHN.F.SCZ 162 showed rests of dense infilling sediment within the medullary cavity that were removed in Adobe photoshop ^® CC 2015. Cracks in the cortex of this specimen were also digitally repaired to avoid biasing the following measurement.

Figure 2.

Coronal section (CT‐scan) of the humerus of C. didactylus ZMB‐MAM 35825. The sampled cross‐section was extracted from the level of greatest cortical thickness (dashed yellow line), which was just distal to the nutrient canal (indicated by blue arrowheads). Scale bar: 10 mm.

Figure 3.

Humeral cross‐sections (CT‐scans) of selected xenarthrans. (A) Bradypus torquatus ZSM 1903‐9534, (B) Priodontes maximus SMF 42286, and (C) Tamandua tetradactyla ZLHUB‐1818‐20. In (A), numerous cavities are present in the cortex (examples marked with blue arrowheads), whereas (B) and (C) lack apparent cortical porosity (at the resolution in use). cor, cortex; em, endosteal margin; mc, medullary cavity. Scale bars: 3 mm.

The first dataset deriving from CT‐scans was complemented with data from conventional humeral thin‐sections produced by Straehl et al. (2013). Images of these thin‐sections (see Table 1) have a higher resolution than the CT‐scans of the first dataset, hence showing a relatively greater number of small cavities, which mostly correspond to the central vascular canal of fully grown secondary osteons. Thus, the resolution of each thin‐section image was downgraded to the highest possible resolution that excludes the pixels representing the empty vascular canals (ca. 60% of that the CT‐scan of the specimen of the same species with the lowest resolution was enough for such purposes). These thin‐section images were then binarized. For the image of the conventional thin‐section of Bradypus torquatus ZMZ11102 deriving from the data of Straehl et al. (2013), it was necessary to fill in the medullary cavity, the cavities of the cortex, and the non‐bone material with white in Adobe photoshop ^® CC 2015.

Measurement of the cortical compactness (CC)

For the measurement of CC using Fiji, a region of interest (ROI) was manually selected on each humeral cross‐section [‘wand (tracing) tool’], including only the cortical bone, thus excluding from the selection the medullary cavity and surrounding trabeculae (see Fig. 4 and Supporting Information Fig. S1). At the level of the selected cross‐section, the diaphysis of ‘tree sloths’ lacks bony processes, in contrast to that of other xenarthrans (Fig. S1). At the level of these processes, the cortex is thin and an inner spongy area is present. Therefore, to compare the same structure across the whole dataset, the cortex just internal to bone processes was excluded from the ROI of the relevant specimens. In the latter cases, the selected ROI was defined as extending up to the level at which the cortex started to thin out and/or at which the inner spongy area appeared (Figs 4D–F and S1L,N–U,W–Z). Cracks in the cortex of the fossil humeral cross‐section of Parocnus brownii AMNH u3 were also excluded from the ROI (Fig. S1M). For the value of CC, the percentage of pixels corresponding to bone tissue within each ROI was acquired using the routine ‘area fraction’ of the ‘measure’ function.

Figure 4.

Binarized cross‐sections of the humerus of selected xenarthrans. The ‘tree sloths’ (A) Bradypus tridactylus NMB 10488 and (B) Choloepus didactylus ZMB‐MAM 102636 show numerous cavities (in white) within the cortical area. (C) Hapalops sp. MNHN.F.SCZ 162, (D) Myrmecophaga tridactyla ZMB‐MAM 77025, (E) Tamandua tetradactyla ZMB‐MAM 81448, and (F) Priodontes maximus CeNaK‐S‐6038 are examples of xenarthrans that show a small number of cavities within the cortical bone. The ROI where cortical compactness (CC) was measured is located within the area(s) delineated in red, which defines the cortex to the exclusion of those areas associated with a bony process [which alter cortical structure; (D–F)]. Scale bars: 3 mm.

Statistical analysis

r version 3.4.0 (R Development Core Team, 2008) was used to perform all statistical tests, for which the significance level was set to α = 0.05. We first checked for the presence of a phylogenetic signal within the CC measured for our dataset. We used the timetree of Gibb et al. (2016) and completed it with extinct species that we sampled according to published phylogenetic hypotheses and dating (Muizon et al. 2004; Perkins et al. 2012; Amson et al. 2016). Because the recovered Pagel's lambda (phylosig function of the phytools package; Revell, 2012) was not significantly different from 0, we proceeded with traditional statistics (as assuming any model of evolution other than White Noise would be spurious). Wilcoxon rank sum tests (wilcox.test) were used to test whether significant differences occurred in the humeral CC between ‘tree sloths’ and the ‘other xenarthrans’, and between the two genera of ‘tree sloths’. Furthermore, the Brown‐Forsythe Levene‐type test (levene.test) was carried out to test equality of variances between ‘tree sloths’ and ‘other xenarthrans’.

Results

CC of ‘tree sloths’ vs. ‘other xenarthrans’

Qualitative examination of the raw humeral CT‐scans indicated that high porosity marked the humeral cortex along the whole diaphysis in Choloepus and Bradypus, but not in other xenarthrans. Indeed, the cortical cross‐sections of the 10 sampled ‘tree sloths’ were characterized by numerous cavities (Figs 4A,B and S1A–J). The latter were distributed throughout the thickness of the cortex. In contrast, the other xenarthran species exhibited an extremely low number of cortical cavities (see Figs 4 and S1). Note, however, that the extinct sloths Hapalops MNHN.F.SCZ 162 and Parocnus brownii AMNH u3 also featured several cortical cavities of remarkable size. Furthermore, the few cortical cavities present on the cross‐sections of the ‘other xenarthrans’ were mostly located near the medullary cavity and in very few cases, e.g. Hapalops MNHN.F.SCZ 162, in the center or the periphery of the cortical bone.

Accordingly, the CC of ‘tree sloths’ was significantly lower than that of ‘other xenarthrans’ (Wilcoxon rank sum test, P < 10⁻⁴, Table 1, Fig. 5). Both categories exhibited a mean CC > 96%. The median CC for the ‘tree sloths’ was 98.778% and the minimum was observed in C. didactylus ZMB‐MAM 102636 with a value of 96.331% (Fig. 4B). Bradypus sp. ZMB‐MAM 76147 was the ‘tree sloth’ that exhibited the most compact cortex, with a CC value of 99.760% (Fig. S1B). Hence, a relatively large range of variation of CC existed within our sample of ‘tree sloths’ (Fig. 5). The conducted measurements for ‘other xenarthrans’ yielded values that ranged from 99.583% in Hapalops MNHN.F.SCZ 162, to 99.995% in P. maximus ZMB‐MAM 36422, except for a low value in one outlying specimen (99.035%), the ‘ground sloth’ P. brownii AMNH u3. The median of CC for the ‘other xenarthrans’ was 99.952%. This latter group exhibits a relatively small range of variation in CC values (Fig. 5A). The difference in variance with the CC in ‘tree sloths’ was significant (Brown‐Forsythe Levene‐type test, P = 0.0082). Direct comparison of the CC of T. littoralis MNHN.F.SAS 1605 (99.943%) with the CC of the ‘tree sloths’ showed that the value of this fossil specimen was higher than that of any ‘tree sloth’ specimen. The same applied for Hapalops MNHN.F.SCZ 162 and P. brownii AMNH u3, with the exception of the value of one Bradypus specimen (ZMB‐MAM 76147) and two Bradypus specimens (ZMB‐MAM 76147 and ZSM‐AM 1029), respectively.

Figure 5.

Box‐and‐whisker plots of humeral cortical compactness (CC) among xenarthrans. (A) all ‘tree sloth’ specimens (n = 10) show low CC and a wider range of variation than that of ‘other xenarthrans’ (n = 16). (B) Bradypus (n= 6; the outlier is Bradypus torquatus ZMZ 11102) and Choloepus (n = 4) exhibit similarly low levels of cortical compactness.

Because the recovered phylogenetic signal (Pagel's lambda) of CC across xenarthrans was not significantly different from 0, an ancestral state reconstruction for sloths assuming an evolutionary model was not performed. However, one can score the ‘tree sloths’ as having a ‘low CC’ and the rest of sampled taxa as having a ‘high CC’, and optimize this binary trait using parsimony (one should note, however, that two of the three sampled ‘ground sloths’ feature rather low values, albeit higher than the main distribution of ‘tree sloths’). Considering the armadillos and anteaters as outgroups (Barriel & Tassy, 1998), the most parsimonious scenario allows the state of the last common ancestor (LCA) of sloths to be reconstructed as ‘high CC’. This would also indicate that two transitions from ‘high CC’ to ‘low CC’ occurred, one for Bradypus and one for Choloepus (two evolutionary steps, the singlemost parsimonious scenario).

CC of Choloepus vs. Bradypus

Both Bradypus and Choloepus exhibited comparatively low CC values and there was no evident disparity in the distribution of the porosity between them, since cortical cavities were found scattered over the cortical area without a notable preferential zone of occurrence in any specimen of both genera. The levels of CC for Choloepus ranged from 96.331 to 98.940% with a median of 98.200%, and in Bradypus they ranged from 98.644 to 99.760% with a median of 98.945% (Fig. 5B). Bradypus torquatus ZMZ 11102, a specimen that had to be processed slightly differently (see Materials and methods above), was an outlier with a value of 96.857%. Statistical tests indicated that the humeral CC of Choloepus was not significantly different from that of Bradypus, regardless of whether the outlying specimen B. torquatus ZMZ 11102 was included (Wilcoxon Rank Sum test, P = 0.171) or not (P = 0.063).

Discussion

The investigation of diaphyseal cross‐sections derived from CT‐scans and thin‐sections reveals that specimens of the ‘tree sloths’ Choloepus and Bradypus exhibit a strikingly high number of large cavities in this area of the bone. The quantification of the area of these cavities yields surprisingly low cortical compactness values, which, in comparison with the values of other similar‐sized xenarthrans, reveals an additional morphological trait that distinguishes the very peculiar ‘tree sloths’ from other xenarthrans. Given these results, and our present understanding of xenarthran phylogeny, it is most parsimonious to recognize that the ancestral condition for all sloths is a high CC (as in other xenarthrans and, more generally, most terrestrial mammals; Fig. 1, LCA), and for the two genera of ‘tree sloths’ to have convergently acquired a low CC, which hence represents an addition to the growing list of their convergences (Nyakatura, 2012; Coutier et al. 2017). However, only three ‘ground sloths’ were included in the present analysis, and two of them exhibited rather low values of CC. A broader dataset is required to test whether there is a tendency in some ‘ground sloths’ to have a low CC.

Based on conventional thin‐sections, the highly remodeled bone tissue and presence of numerous cavities in the cortical bone of sloths were interpreted in a previous study as indicating that secondary bone remodeling was in progress (at the death of the sampled individuals) (Straehl et al. 2013). According to our observation of the images of these thin‐sections of ‘tree sloths’, we concur with their interpretation, adding that the cortical cavities often are the central canals of immature secondary osteons. This differs from a ‘pathological’ osteoporotic state for which intense resorption is observed, particularly in the inner region of the cortex (Jowsey, 1960; Francillon‐Vieillot et al. 1990). This also differs from the high porosity associated with the ‘non‐pathological’ osteoporotic phenotype, described for example in the humerus of extant cetaceans, which is due to intense bone resorption that takes place on the extensive spongiosa. This bone type fills the medullary cavity, their cortex being either reduced or completely absent (Buffrénil & Schoevaert, 1988). Therefore, based on the 10 skeletally mature, wild‐caught, and non‐pathological specimens sampled here, we infer that low humeral CC, a proxy in xenarthrans for bone remodeling in progress, is part of the phenotype that generally characterizes ‘tree sloths’. This entails, contrary to other xenarthrans (and probably most mammals), a remodeling in ‘tree sloths’ of their bones at a high rate in a balanced manner (without excess of resorption such as in pathological osteoporosis) until late in their ontogeny (possibly throughout life). Observed throughout the cortex of their humeral diaphysis, low CC indicate that the remodeling in ‘tree sloths’ is not restricted to a specific cortical area (see Figs 4A,B and S1A–J) or diaphyseal region of the humerus. Furthermore, qualitative observations of CT‐scans of other long bones (whole forelimb, femur, and radius), as well as elements of the axial postcranium (dorsal ribs and vertebrae) of skeletally mature ‘tree sloths’, suggest that low CC and, with it, long‐lasting intense bone remodeling, is in fact not restricted to their humerus, but is present throughout their postcranium.

The type of data analyzed here (CT‐scans and thin‐sections) does not allow us to ascertain the exact cause for the low CC, which we associate to ongoing bone remodeling in skeletally mature ‘tree sloths’. However, we here explore potential explaining factors that might be related to this trait in Bradypus and Choloepus. Even though the convergent acquisition of low CC in both ‘tree sloth’ genera might suggest a functional significance of this trait related to their common lifestyle, one should not exclude that it might be an example of a ‘spandrel’, i.e. a non‐functional trait that arises due to the evolution of another trait (Gould & Lewontin, 1979). Furthermore, the possibly explaining factors presented below are not necessarily mutually exclusive.

Is low CC related to a low whole‐body metabolic rate?

The existence of a relationship between bone remodeling and metabolic rate is well established (de Paula & Rosen, 2013). High metabolic rates correlate with high remodeling rates in early life stages and are associated with adaptations to more active lifestyles, in which bones are exposed to greater loading (Mitchell, 2016). Bone remodeling is a process that harnesses a significant amount of energy (de Paula & Rosen, 2013) and requires fast cell proliferation. Normally, low metabolic rates decelerate the formation rate of new bone cells and reduce the activity of the mechanisms involved in bone remodeling, e.g. signaling for repair and productivity by trigger hormones (Mitchell, 2016). Notable examples for the incidence of this phenomenon are hibernating mammals. Their temporarily decreased metabolic rates reduce the intensity of the bone remodeling process in order to minimize energy costs and, in addition, avoid bone loss (Mitchell, 2016). ‘Tree sloths’ are known to have a very low whole‐body (i.e. not tissue‐specific) metabolism for their size (34–56% of the basal metabolic rate of an average mammal of equivalent body mass; Vendl et al. 2016). It was therefore surprising to find in Choloepus and Bradypus that the bone tissue was heavily remodeled (Straehl et al. 2013). However, the low metabolism of Choloepus is associated with low thyroid activity (Lemaire et al. 1969; no data available to our knowledge for Bradypus). In mammals, the multifunctional thyroid hormones are essential for both the regulation of metabolism (Brent, 2012), and the development and maintenance of bones (Gorka et al. 2013). In humans, hyperthyroidism has been identified as one of possible causes for secondary osteoporosis. This hormonal dysfunction of the thyroid gland alters the normal remodeling cycles, increasing the ratio of bone resorption to bone formation (Gorka et al. 2013). It can therefore be suggested that the low thyroid activity of ‘tree sloths’ (at least Choloepus) may support the maintenance of a balanced bone remodeling process until late in the ontogeny (possibly even throughout life). Parathyroid activity might be associated with this phenomenon as well, but, to our knowledge, it was never monitored in sloths. Estrogen levels might also be linked to a particular bone remodeling pattern, but the available data did not report sloth estrogen levels to be different from those of other mammals. (Mühlbauer et al. 2006; Troll et al. 2015; M. Mühlbauer and D. Gilmore, pers. comm).

Another factor that is associated with the overall low metabolic rate of ‘tree sloths’ is that it may have freed some of their phenotypical traits from stabilizing selection by avoiding deleterious pleiotropic effects (Varela‐Lasheras et al. 2011; Böhmer et al. 2018). This provides one of the hypotheses explaining the number of seven cervical vertebrae that is conserved in all mammals (Arnold et al. 2017; the competing hypothesis being also linked to a low metabolic rate, Buchholtz, 2014) except ‘tree sloths’ and manatees. Pleiotropy occurs when mutations in a genetic locus affect more than one phenotypic trait. Such phenotypic variations often have negative effects on the fitness of an organism (Paaby & Rockman, 2013) and thus limit evolutionary change. According to this hypothesis, their overall low metabolic rate diminishes the incidence and gravity of cancer and other diseases associated with free radicals (Varela‐Lasheras et al. 2011). One could analogously suggest that the acquisition of a long‐lasting and intense bone remodeling process is prevented in most mammals by deleterious pleiotropic effects, but that this constraint was alleviated by the low whole‐body metabolism of ‘tree sloths’.

Other possible explaining factors

Functional demands of suspensory posture and locomotion

Since a fully arboreal lifestyle and a suspensory posture and locomotion are exclusive for Bradypus and Choloepus within the xenarthran clade, the question arises whether low CC is related to the mechanical environment entailed by their lifestyle (as for instance in the case of the highly plastic trabecular structure of their humerus, Amson et al. 2017). A strong functional signal was found in the mid‐diaphyseal structure of the humerus within a large sample of amniotes (Canoville & Laurin, 2010). More specifically, strain or exposure to great loading or torsion often results in the induction of fast bone remodeling (Lee, 1968; van Oers et al. 2008). However, low CC is not restricted to the humerus but is also observed (although only qualitatively) in other bones, e.g. ribs, which do not share the same mechanical demands. This would tend to support the rejection of this hypothesis. Nevertheless, Lieberman (1996) has suggested that a variation in the structure of single bones caused, for example, by specific mechanical demands or a change in the levels of physical activity of the organism, can become a systemic process and affect other bones as well (see also Amson et al. 2018).

A different speculation links the occurrence of lower CC not to the loads experienced by the bones during suspension but to the constant danger of falling. Maintaining a high degree of secondary bone remodeling throughout life could present an adaptation helping sloths to heal fall‐incurred bone injuries. However, to our knowledge, healed fractures are not frequently observed in ‘tree sloths’ specimens, which would have supported this hypothesis.

Mineral homeostasis

Bone remodeling is involved in mineral homeostasis through the storage and release of calcium and phosphate (Jowsey, 1960). However, no relevant physiological factors point to obvious differences between the normal calcium and phosphate homeostasis of ‘tree sloths’ and that of other xenarthrans. It is therefore not expected that the increased secondary bone remodeling is a response to mineral demands. Cases of special mineral need that can lead to increased mobilization of these are gestation and lactation in mammals (Mahan & Vallet, 1997), and development of temporary mineralized structures, such as antlers or eggshells in vertebrates (Francillon‐Vieillot et al. 1990; Mitchell, 2016). For the latter, the increase in bone remodeling is limited to the time of formation of the mentioned structures and is often linked to osteoporotic conditions or incomplete remodeling cycles that impede the formation of secondary osteons (Mitchell, 2016). This can therefore not be compared to the condition of ‘tree sloths’, whose bones appear to undergo constant and complete secondary remodeling cycles. Furthermore, an investigation on armadillos has revealed that female individuals in gestation or lactation states do not lose bone density, at least not from the carapace and femur (Actis et al. 2017). Assuming that the mineral homeostasis is similar in all xenarthrans, it can be expected that female reproduction under adequate diet conditions does not entail deflection of minerals from the bone tissue of xenarthrans in general. Hence, the very unlikely possibility of having sampled only pregnant or lactating ‘tree sloth’ females with inadequate diets and only non‐pregnant females or males of other xenarthran species can be safely rejected. It thus seems unlikely that low CC is a functional response of sloths to a specific physiological demand of mineral mobilization.

High variability in ‘tree sloths’

Finally, we address the fact that both skeletally mature Bradypus and Choloepus feature a relatively wide range of CC values in comparison with that of the ‘other xenarthrans’. Although the CC values found in ‘other xenarthrans’ all approach the maximal CC, which bounds unilaterally their distribution, we did recover a significantly narrower range of their variation. This could suggest that the rate of bone remodeling in ‘tree sloths’ is subject to high variability.

In ‘tree sloths’ several traits affected by high intraspecific variability have been described, including the morphology of the bony labyrinth (Billet et al. 2012), composition of the axial postcranium (Buchholtz & Stepien, 2009), and timing of cranial sutures closure (Rager et al. 2014). ‘Tree sloths’ show a tolerance towards intraspecific variation for many parts of their phenotype, which could be due to release of selection pressures (Billet et al. 2012). In the case of bone structure, high CC, which is normally conserved among terrestrial mammals, could have been released from selective pressure, yielding the more variable CC values observed in ‘tree sloths’.

Conclusions

The CC values in the humeri of skeletally mature ‘tree sloths’ are significantly lower than in the rest of the xenarthran species we sampled, which include three ‘ground sloths’ as well as anteaters and armadillos. This highly unique phenotype at the scale of mammalian bone microstructure is a trait that can be added to the list of the compelling convergences that characterize ‘tree sloths’. Thanks to osteo‐histological observations, we can associate the low CC observed in ‘tree sloths’ to a strong and balanced bone remodeling process that is maintained until late in ontogeny (possibly throughout life). Whether this carries direct functional significance or is an ‘evolutionary spandrel’ is so far unclear. But a relationship to the extremely low metabolism for their body size characterizing ‘tree sloths’ is viewed here as the most likely hypothesis.

Being non‐destructive, the method for the quantification of the humeral CC of different xenarthrans introduced in this study is a convenient procedure for generating information on bone microstructure. Being well suited for quantitative and comparative analyses, it could also be applied on elements other than long bones. This study constitutes a reliable groundwork for further and more detailed investigations, which are required to gain insight into the causes and implications of the particular microanatomical traits that skeletally mature ‘tree sloths’ exhibit. Future analyses, which should likely include an experimental approach, might reveal how ‘tree sloths’ seem to have dodged fundamental constraints that explain the high CC found in most other mammals.

Author contributions

I.M.R. performed data processing and analysis, statistical analysis, and drafted the manuscript. E.A. acquired raw data from museum specimens and designed the analysis. E.A. and J.A.N. conceived the study. All authors interpreted the data, and contributed to and approved the final version of the manuscript.

Supporting information

Fig. S1. Binarized cross‐sections of the humerus of the analyzed xenarthran specimens.