How does bone microanatomy and musculature covary? An investigation in the forelimb of two species of martens (Martes foina, Martes martes)

Abstract

The long bones and associated musculature play a prominent role in the support and movement of the body and are expected to reflect the associated mechanical demands. But in addition to the functional response to adaptive changes, the conjoined effects of phylogenetic, structural and developmental constraints also shape the animal's body. In order to minimise the effect of the aforementioned constraints and to reveal the biomechanical adaptations in the musculoskeletal system to locomotor mode, we here study the forelimb of two closely related martens: the arboreal pine marten (Martes martes) and the more terrestrial stone marten (Martes foina), focusing on their forelimb muscle anatomy and long bone microanatomy; and, especially, on their covariation. To do so, we quantified muscle data and bone microanatomical parameters and created 3D and 2D maps of the cortical thickness distribution for the three long bones of the forelimb. We then analysed the covariation of muscle and bone data, both qualitatively and quantitatively. Our results reveal that species‐specific muscular adaptations are not clearly reflected in the microanatomy of the bones. Yet, we observe a global thickening of the bone cortex in the radius and ulna of the more arboreal pine marten, as well a stronger flexor muscle inserting on its elbow. We attribute these differences to variation in their locomotor modes.

Analyses of our 2D maps revealed a shift of cortical thickness distribution pattern linked to ontogeny, rather than species‐specific patterns. We found that although intraspecific variation is not negligible, species distinction was possible when taking muscular and bone microanatomical data into consideration. Results of our covariation analyses suggest that the muscle‐bone correlation is linked to ontogeny rather than to muscular strength at zones of insertion. Indeed, if we find a correlation between cortical thickness distribution and the strength of some muscles in the humerus, that is not the case for the others and in the radius and ulna. Cortical thickness distribution appears rather linked to bone contact zones and ligament insertions in the radius and ulna, and to some extent in the humerus. We conclude that inference on muscle from bone microanatomy is possible only for certain muscles in the humerus.

Species‐specific muscular adaptations are not reflected in the bone microanatomy. Yet we observe a thickening of the bone cortex in the zeugopod of the more arboreal pine marten, as well as a muscular difference in the elbow of the two species, that we attribute to their different locomotor modes. We find a qualitative link between cortical thickness repartition and muscle strength in the humerus while it appears rather linked to bone contact zones and ligament insertions in the radius and ulna.

1. INTRODUCTION

The vertebrate skeleton ensures various functions, including movement, by being the passive structure on which the force‐producing muscles are attached. Like all biological structures, limb anatomy results from the conjoined effects of phylogenetic, structural and functional constraints (e.g., Cubo, 2004; Gould, 2002). Since long bones play a prominent role in the support and movement of the body, their external morphology is expected to reflect the biomechanical demands they face (Fabre et al., 2013a, 2015; Iwaniuk et al., 1999, 2000; Janis & Figueirido, 2014; Schmidt & Fischer, 2009). But their inner structure (bone microanatomy) also bears a strong functional signal (Ruff & Hayes, 1983; Turner, 1998; Ruimerman et al., 2005; Habib & Ruff, 2008; Nikander et al., 2009; Houssaye et al., 2018), and can thus reflect habitat (Laurin et al., 2011; Quemeneur et al., 2013; Nakajima et al., 2014), locomotor mode (Ryan & Ketcham, 2002, 2005; Carlson, 2005; Marchi, 2005; Carlson et al., 2006, 2008; Shaw & Stock, 2009, 2011; Bishop et al., 2018), body weight support requirements (Davies & Stock, 2014; Houssaye et al., 2018) and behaviour (Warden et al., 2007; Wilks et al., 2009). While bone microanatomical features are inherited from evolution, bone microanatomy also adapts to functional constraints during the lifetime of organisms (Reina et al., 2017; Sievänen, 2010; Warden et al., 2007). Wolff's law (1986) states that bones adapt, if they have time, to mechanical stresses and gravity (Ruff et al., 2006). Bones thus tend to be stiffer and stronger when subjected to high stresses, with an increase in cortical thickness and trabeculae orientated in the direction of the maximal strain (Barak et al., 2013; Ruimerman et al., 2004; Volpato, 2008; Wolff, 1986).

During an individual's lifetime, cortical bone can thicken in response to mechanical forces applied either through the direct insertion of the solicited muscles on the bone or through charge transfer by the trabeculae from the articulating surfaces (Henrikson et al., 1974; Hoyte & Enlow, 1966). In the case of muscle insertion, apophyses (i.e., bony tuberosities) may or may not form on the contact areas between the bone and muscle (Niinimäki et al., 2013; Sanchez et al., 2013).

The close association between muscles and bones is broadly intuitive because muscle forces and skeletal (cranial and postcranial) structure are linked through the process of bone modelling (i.e., bone formation during growth) and remodelling (i.e., bone resorption and deposition; Frost, 2001; Tatara et al., 2014). Thus, we could expect muscular strain to be reflected in the bone inner anatomy. However, the covariation between muscle and bone microanatomy has been scarcely studied, although it would enable to better understand bone functional adaptation. Muscles and bones are often studied separately, and most of the studies that do combine muscle and bone data have focused on the skull (e.g., Brassard et al., 2020; Fabre et al., 2014; Fabre et al., 2018; Sella‐Tunis et al., 2018; Toro‐Ibacache et al., 2016). Only a few studies have investigated the relationship between postcranial bones and muscles, and even fewer have done so quantitatively. Moreover, those few studies have focused exclusively on the external anatomy of the bones: Warburton et al. (2013) and Böhmer et al. (2018) both investigated the bone/muscle relationship in the forelimb of terrestrial mammals, finding significant sexual dimorphism (in the form of positive allometry in the musculature of male kangaroos) and species‐specific differences (stronger muscles in M. martes than in M. foina), respectively. In both cases, the differences were explored in muscular anatomy but not in muscle/bone covariation. Martin et al. (2019) found a significant correlation between muscle strength (approximated by muscle mass and cross‐sectional area) and bone external shape in digging marsupials.

Many bone microanatomical studies rely on two‐dimensional transverse slices of the diaphysis. However, cortical thickness and trabecular distribution are not uniform along the shaft, so that a single section provides limited information and thus can hardly be informative regarding muscular attachments. More recent studies use three‐dimensional (3D) visualisations of the whole bone instead, which allow investigating the relationship between muscle insertion and cortical bone: for example, Harbers et al. (2020) found that captivity induced an increase in cortical bone volume and muscle force in the humerus of suids. We here analyse the covariation of muscle and bone microanatomy in two phylogenetically closely‐related mustelids: the arboreal pine marten (Martes martes), and the more terrestrial stone marten (Martes foina). Although mustelids are an extremely diverse family, varying in size, geographic range and predation behaviour, the two studied species occur sympatrically and are very similar in overall appearance. They distinctly differ in habitat preference (forest versus urban environments) and locomotor mode (arboreal versus more terrestrial; Overskaug, 1994; Goszczyński et al., 2007; Wereszczuk & Zalewski, 2015). Differences in their shoulder and forelimb muscles have been highlighted and interpreted to reflect the greater climbing ability of the pine marten (Böhmer et al., 2018). This raises the question whether this difference can be observed in the shape of the bone, as well as at the bone microanatomical level.

The aim of this study is to investigate whether cortical thickness covaries with the strength of the muscles at zones of insertion by local thickening. In order to test this hypothesis, we first investigate how the differences in the locomotor mode of the two sympatric species are reflected (1) in the muscles attaching to the bone diaphysis and epiphyses; (2) in the microanatomy and, especially, cortical bone distribution; then we analyse (3) if there is a covariation between the distribution of cortical bone thickness and the insertion areas of the muscles in accordance with their relative strength and (4) the relative strength of this covariation at the intraspecific and interspecific levels.

2. MATERIALS AND METHODS

2.1. Sample

The study focused on M. foina (Erxleben, 1777), a medium‐sized terrestrial mustelid occurring in rural areas of Europe and Asia, with comparative specimens of its sister‐taxon M. martes (Linnaeus, 1758). Ten specimens of M. foina and three specimens of M. martes were analysed (Table 1). Carcasses were received from the Faculty of Veterinary Medicine of Ludwig‐Maximilians‐Universität in Munich (Germany), the INRAP Centre de Recherches Archéologiques de l'Oise in Compiègne (France), and the taxidermy laboratory of the Muséum National d'Histoire Naturelle in Paris (France).

TABLE 1.

Sample studied (A: adult, J: juvenile; MaxL: maximum length of the bone; NA: not available because one or both epiphyses were missing)

Specimen	Species	Age category	Humerus MaxL (cm)	Radius MaxL (cm)	Ulna MaxL (cm)
Mm1	Martes martes	A	6.77	5.13	6.31
Mm2	Martes martes	A	7.45	6.01	7.18
Mm3	Martes martes	A	7.04	5.32	6.49
Mf1	Martes foina	A	5.78	4.48	5.57
Mf2	Martes foina	A	6.44	4.79	5.81
Mf3	Martes foina	A	6.68	5.34	6.57
Mf4	Martes foina	J	NA	NA	NA
Mf5	Martes foina	A	6.51	5.31	6.57
Mf6	Martes foina	J	NA	NA	NA
Mf7	Martes foina	A	6.78	5.09	6.21
Mf8	Martes foina	J	NA	NA	NA
Mf9	Martes foina	A	7.03	5.38	6.72
Mf10	Martes foina	J	NA	NA	NA

Species discrimination was based on external attributes and dental characteristics: the throat patch in M. martes is yellowish, whereas it is white in M. foina; the crown morphology of the third maxillary premolar in occlusal view is concave in M. martes and convex in M. foina (Libois & Waechter, 1991; Llorente Rodriguez et al., 2011). Specimens did not display any pathology. All three M. martes specimens and six M. foina specimens were adult animals. Four M. foina specimens were juveniles (see Table 1). Age status was determined by the degree of fusion of the epiphyses of the long bones after dissection (completely fused in adults and unfused in juveniles; Nickel, 2003).

The present analyses on the M. foina specimens enabled us to study the link between bone and muscles at the intraspecific level, with specimens of various ontogenetic stages and body sizes. The comparison with M. martes allowed us to estimate species‐specific characteristics as well as the relationship between intra‐ and interspecific variation.

2.2. Muscle data

In total, we studied the left and right forelimbs of 13 specimens. Nine specimens (three M. martes and six M. foina) were previously dissected and muscle data were already reported in Böhmer et al. (2018). The remaining specimens (four juvenile M. foina specimens) were dissected for the present study. A number of 37 extrinsic and intrinsic muscles attach to at least one of the three long bones of the forelimb (humerus, radius and ulna; Table 2). Each muscle was identified and systematically dissected. The dissection protocol follows that described by Böhmer et al. (2018).

TABLE 2.

Muscles of the forelimb in martens (based on Martes martes)

Muscle	Acronym	Origin	Insertion	Main function	Functional group
M. supraspinatus	SSP	Supraspinous fossa and scapular spine	Greater tubercle of humerus	Shoulder joint extensor and humeral protractor
M. infraspinatus	ISP	Infraspinous fossa and scapular spine	Lateral on greater tubercle of humerus (Infraspinatus muscle facet)	Shoulder joint flexor and lateral humeral rotator
M. subscapularis	SUB	Subscapular fossa	Lesser tubercle of humerus	Scapular adductor
M. teres minor	TMN	Caudal border of scapula (near glenoid fossa)	Lateral on greater tubercle of humerus (distal to M. infraspinatus)	Shoulder joint flexor and lateral humeral rotator
M. triceps brachii caput accessorium	TBA	Distal caudomedial aspect of humeral diaphysis (along medial epicondylar crest)	Medial aspect of olecranon	Elbow joint extensor
M. extensor carpi ulnaris	ECU	Lateral epicondylar crest of humerus (distal to origin of M. extensor digitorum lateralis)	Base of metacarpal V	Elbow joint flexor and wrist joint extensor
M. extensor digitorum lateralis	EDL	Lateral epicondylar crest of humerus (distal to origin of M. extensor digitorum communis	Tendons into distal phalanges of digits IV‐V	Elbow joint flexor, wrist joint extensor and digital extensor
M. extensor digitorum communis	EDC	Lateral epicondylar crest of humerus (distal to origin of M.extensor carpi radialis)	Tendons into distal phalanges of digits II‐V	Elbow joint flexor, wrist joint extensor and digital extensor
M. extensor carpi radialis brevis	ECRB	Proximal lateral epicondylar crest of humerus (together with or distal to origin of M. extensor carpi radialis longus)	Base of metacarpal III	Elbow joint flexor and wrist joint extensor
M. palmaris longus	PL	Medial epicondyle of humerus	Tendons into distal phalanges of digits II‐V or palmar aponeurosis	Wrist joint flexor and digital flexor
M. flexor carpi ulnaris, caput humerale	FCUH	Medial epicondyle of humerus	Sesamoid proximal to metacarpal V (Pisiform)	Wrist joint flexor
M. flexor carpi radialis	FCR	Medial epicondyle of humerus	Base of metacarpal II and III	Wrist joint flexor	Fl
M. flexor digitorum profundus (4 heads)	FDP	Medial epicondyle of humerus and medial aspect of olecranon	Tendons into distal phalanges of digits II‐V	Wrist joint flexor and digital flexor	Fl
M. pectoantebrachialis	PAB	Ventrolateral surface of cranial portion of sternum (Manubrium)	Cranial proximal aspect of humeral diaphysis (superficial to insertion of M. pectoralis minor)	Humeral adductor	Add
M. pectoralis major	PMJ	Ventral surface of cranial portion of sternum (Manubrium) and body of sternum	Craniomedial middle of humeral diaphysis (superficial to insertion of M. pectoralis minor)	Humeral adductor and protractor	Add
M. pectoralis minor	PMN	Ventral surface of body of sternum	Cranial proximal aspect of humeral diaphysis (deep to insertion of M. pectoantebrachialis and M. pectoralis major)	Humeral adductor and retractor	Add
M. xiphihumeralis	XH	Ventrolateral surface of caudal portion of sternum (Xiphoid process)	Craniomedial middle of humeral diaphysis	Humeral adductor and retractor	Add
M. clavobrachialis	CB	Clavicle (if present) and raphe with M. clavotrapezius and M. cleidomastoideus	Cranial distal aspect of humeral diaphysis	Humeral protractor	Fl
M. teres major	TMJ	Caudal border of scapula (Teres major muscle facet)	Craniomedial on humeral diaphysis (Teres major tuberosity; near pectoral crest)	Shoulder joint flexor and humeral retractor	Fl
M. triceps brachii caput mediale, intermediate and long portion	TBM	Mediocaudal humeral diaphysis	Medial aspect of olecranon	Elbow joint extensor	Ext
M. articularis humeri	AH	Coracoid process of scapula	Medial proximal aspect of humeral diapyhsis	Shoulder joint stabiliser and humeral adductor	Add
M. triceps brachii caput laterale	TBLA	Proximal aspect of deltoid crest of humerus	Caudolateral aspect of olecranon (lateral to M. triceps brachii caput longum)	Elbow joint extensor	Ext
M. anconeus	ANC	Distal caudal aspect of humeral diaphysis (along lateral epicondylar crest)	Lateral aspect of olecranon (deep to M. triceps brachii caput laterale)	Elbow joint extensor and forearm pronator	Ext
M. acromiodeltoideus	DA	Acromion	Deltoid crest of humerus (superficial to M. spinodeltoideus)	Shoulder joint flexor and humeral abductor	Abd
M. spinodeltoideus	DS	Scapular spine (Superficial to M. infraspinatus)	Deltoid crest of humerus	Shoulder joint flexor and humeral abductor	Abd
M. extensor carpi radialis longus	ECRL	Proximal lateral epicondylar crest of humerus (distal to origin of M. brachioradialis ‐ if present)	Base of metacarpal II	Elbow joint flexor and wrist joint extensor	Fl
M. brachioradialis	BCR	Proximal lateral epicondylar crest of humerus (proximal to origin of M. extensor carpi radialis)	Distal medial aspect of radius	Elbow joint flexor and forearm supinator	Fl
M. brachialis	BCH	Proximal caudolateral humeral diaphysis	Tendon into bicipital tuberosity of radius or coronoid process of ulna (adjacent to insertion of M. biceps brachii)	Elbow joint flexor and forearm supinator	Fl
M. extensor pollicis	EP	Lateral ulnar and radial diaphysis (and interosseus membrane between radius and ulna)	Base of metacarpal I (and sesamoid proximal to metacarpal I)	Wrist joint extensor and extensor of digit I	Ext
M. pronator teres	PT	Medial epicondyle of humerus (dorsal to origin of M. flexor carpi radialis)	Medial aspect of radial diaphysis	Forearm pronator	Pro
M. pronator quadratus	PQ	Distal medioventral surface of ulna	Distal medioventral surface of radius	Forearm pronator	Pro
M. supinator	SUP	Lateral epicondyle of humerus	Dorsomedial aspect of radial diaphysis	Forearm supinator	Sup
M. epitrochlearis	EPI	Lateroventral on M. teres major and M. latissimus dorsi (Caudal border of scapula)	Caudal aspect of olecranon tuber	Humeral retractor and elbow joint extensor
M. triceps brachii caput longum	TBLO	Caudal border of scapula (near glenoid fossa; medial to M. teres minor)	Caudal aspect of olecranon tuber (deep to M. epitrochlearis)	Shoulder joint flexor and elbow joint extensor
M. extensor indicis	EI	Lateral middle of ulnar diaphysis	Tendons into distal phalanges of digits I and II	Extensor of digits I and II	Ext
M. flexor carpi ulnaris, caput ulnare	FCUU	Medial aspect of olecranon	Sesamoid proximal to metacarpal V (Pisiform)	Wrist joint flexor	Fl
M. biceps brachii	BB	Tendon from the supraglenoid tubercle of scapula (and from coracoid process of scapula if second head present)	Tendon into bicipital tuberosity of radius or coronoid process of ulna	Shoulder joint extensor and elbow joint flexor	Fl

Note: In grey, muscles inserted on the epiphyses.

Add, adductor; abd, abductor; fl, flexor; ext, extensor; pro, pronator; sup, supinator

The following architectural features were quantified by a single examiner (C.Bö) in order to avoid operator bias. First, the blotted dry muscles were weighed on a digital precision balance (Mettler; ±0.1 mg) and the muscle mass data were collected. Muscle belly length (the mean of the maximum and minimum lengths) was measured directly on the muscle using a standard ruler. Next, the fibre length (the mean of 15 randomly selected fibres) of each excised muscle was recorded. To do so, muscle fibres were separated by digesting the muscles in a 30% aqueous nitric acid solution for about 24 hours, after which they were transferred to a 50% aqueous glycerin solution (see Antón, 1999; Herrel et al., 2008). For each muscle, individual fibres were teased apart and scaled digital photographs were taken. The length of the 15 randomly selected fibres was measured using the software ImageJ v.1.48 (Schneider et al., 2012) and then mean fibre length was calculated. The documented parameters (muscle mass and fibre length) allowed us to determine the following variables: (a) Muscle volume (V) was calculated by dividing muscle mass (m) by a standard density (ρ) for mammalian muscles of 1.06 g/cm³ (Méndez & Keys, 1960): V [cm³] = m [g]/ρ [g/cm³] (1); (b) Anatomical cross‐sectional area (ACSA), which is a function of muscle volume and fibre length (l_f) (Powell et al., 1984; Sacks & Roy, 1982), was calculated using the following equation: ACSA [cm²] = (V [cm³]/(l_f [cm]) (2). Eventually, the ACSA of each muscle from the left and right forelimbs was used to calculate c) the mean ACSA of each muscle for each specimen.

Contrary to the physiological cross‐sectional area (PSCA; e.g., Kupczik et al., 2015; Rosin & Nyakatura, 2017; Böhmer et al., 2018), the ACSA does not take into account the pennation angle of muscle fibres. In muscles with high pennation angles, ACSA might be less accurate in predicting the force‐producing capability per muscle volume (Lieber & Friden, 2001). However, muscle output is related to the cosine of pennation and, thus, neglecting small angles causes only a small percentage of error in force estimates (Scott & Winter, 1991). Typically, the pennation angles appear to be rather small in carnivoran forelimb muscles (Böhmer et al., 2018; Moore et al., 2013; Williams et al., 2008). Furthermore, the two species of martens studied here (M. martes and M. foina) share similar pennation angles for the same muscles (Böhmer et al., 2018). Additionally, the surface pennation angle of a muscle may vary significantly from its deep pennation angle (Sopher et al., 2017) and, consequently, only micro‐dissection or micro‐computed tomography analyses may allow accurate analysis of the pennation of all fascicles that make up the muscle (e.g., Kupczik et al., 2015; Rosin & Nyakatura, 2017).

Although muscles are versatile organs that contribute to more than one functional role, the consideration of each muscle's main function facilitates interpretation. We assigned muscles to functional groups based on their topology and on the manipulation of dissected specimens (Table 2). Anatomical terminology primarily follows Böhmer et al. (2020).

2.3. Quantitative muscle analyses

Quantitative analyses were performed on two data sets: (1) A first data set comprising all M. foina specimens (intraspecific sample) and (2) a second data set comprising all adult specimens (interspecific sample).

To facilitate later interpretation, we calculated the mean ACSA for each long bone of each specimen by summing up the ACSA of all muscles attaching to the respective bone's diaphysis (since the subsequent covariation analyses were performed on diaphyseal microanatomical parameters) and dividing the obtained value by the number of muscles, which allows the muscles to be grouped into two categories: ‘strong’ muscles are muscles with an ACSA higher than the mean ACSA; ‘weak’ muscles are muscles with an ACSA smaller than the mean ACSA. We hypothesise that strong muscles apply more stress on the bone and thus may potentially induce bone thickening on their attachment sites. These strength categories were used for qualitative analyses only; quantitative analyses were conducted on the ACSA values of each muscle inserted on the bones using MANOVAs.

For size correction, the obtained ACSA data were logarithmically (log 10) transformed and then regressed against log 10‐transformed total ACSA (i.e., the sum of the ACSA of all muscles) for each specimen (Table S1). All subsequent statistical analyses were performed on (1) log‐transformed ACSA values (not size‐corrected) and (2) the resulting residuals of the regression (size‐corrected data), in order to differentiate between variations due to size and variations due to other factors.

Principal component analyses (PCA) were used to reduce the multidimensionality of the collected data and to visualise the distribution of the specimens in the morphospace delimited by the PCs. We established two data sets for each of the three bones of interest: (1) the first data set includes all muscles that attach to the bone under study and (2) the second data set includes only muscles that attach to the bone's diaphysis, as geometric morphometric analyses (GMMs) were performed on the diaphysis only (see below). Linear regressions of muscle ACSA values on muscle length were used to check for the presence of an allometric relationship in adult specimens (evolutionary allometry).

PCAs were performed and visualised using the FactoMineR package in R (Lê et al., 2008). Analyses of variance (MANOVA) were used to test for a difference of muscle ACSA between the two species of martens. All analyses were performed in R (R Core Team, 2020, version 4.0.2) using RStudio (RStudio Team, 2020, version 1.3.959–1).

2.4. X‐ray microtomography on bones

After dissection, the remaining skeleton was cleaned and the forelimb bones were collected from each specimen. The right humerus, radius and ulna of all M. martes and M. foina specimens were scanned using X‐ray microtomography (Easy Tom 40–150, RX Solutions) at the MRI‐ISEM (Montpellier Ressources Imagerie – Institut des Sciences de l'Évolution de Montpellier, UMR 5554, University of Montpellier), with reconstructions performed using X‐Act (RX Solutions). Voxel size varies between specimens depending on their size, from 238 μm to 413 μm. Bone tissues were segmented on the complete bones. The trabecular area, consisting of the medullary cavity and the trabeculae, was then separated from compact cortical bone manually following Houssaye et al. (2018) in order to calculate some microanatomical parameters (see below) and to generate bone cartographies in order to visualise the cortical thickness distribution. Image segmentation and visualisation were performed from the reconstructed image data using Avizo 9.4 (VSG, Burlington, MA, USA).

Bones were aligned along their longitudinal axis following Ruff (2002). Bone maximal length (MaxL) was obtained virtually by using the Landmark software (UC Davis, USA). The distance between proximal and distal extremities for each bone was measured in cranial view. In the humerus, this equates to the distance between the most proximal extremity of the humeral head and the most distal part of the medial epicondyle. In the radius, it is the distance between the head of the radius and the extremity of the styloid process, and in the ulna it is the distance between the most proximal part of the olecranon process and the extremity of the styloid process. The difference in adult bone length between the two species was tested by performing a t‐test.

Diaphyses needed to be isolated for subsequent quantitative analyses. Epiphyses were removed by choosing a homologous landmark on each bone based on Botton (2017), which enabled us to define an orthogonal cutting plane on each extremity of the bone (humerus: disto‐caudal tip of the humeral head, most proximal point of the caudal side of the supracondylar foramen; radius: most proximal point of the ulnar notch, maximum curvature of the depression on the cranial side of the radial tuberosity; ulna: proximal point of the M. brachialis insertion groove, most proximal point of the distal ulno‐radial articulation).

The conversion of the segmented scans into a binary image stack enabled measurements of microanatomical parameters (Table S2) using the BoneJ plugin (Double et al., 2010) of ImageJ (Wayne Rasband National Institutes of Health, USA): (1) 3D compactness of the complete bone (C), that is, the volume occupied by bone (cortex and spongious bone) divided by the whole volume; (2) Relative mean thickness of the cortical layer along the diaphysis (RmeanT), calculated as the absolute value of mean cortical thickness (AmeanT) divided by the radius of the bone's diaphysis if assimilated as a tube (R); and (3) Relative maximum thickness of the cortical layer along the diaphysis (RmaxT), calculated as the absolute value of maximum cortical thickness (AmaxT) divided by R. Finally, we used the ‘MaterialStatistics’ module in Avizo to calculate the trabecular ratio of the complete bone (%Trab), that is, the surface occupied by the trabecular bone over the total surface of osseous tissue (in 3D). Since the ossification was incomplete in juvenile specimens, some parameters (C, %Trab) could not be calculated and as such were excluded from analyses on the M. foina dataset.

All these parameters being ratios, no size‐correction was required. PCAs on the microanatomical parameters (performed using the ‘FactoMineR’ package in R) were used to visualise the distribution of the specimens in the morphospace delimited by the PCs. We performed linear regressions on the first two PCs using bone MaxL as a size estimate in order to check for a size effect within the PCAs.

2.5. Bone thickness mapping and geometric morphometrics

2.5.1. 3D mapping

3D‐mapping of the bone cortical thickness is an approach which provides both a graphical output and a set of numerical parameters, allowing for a holistic functional interpretation of the bone structure. Several methods have so far been used to measure and create 3D maps (see below).

In order to obtain 3D maps of the bone cortical thickness, cortical and trabecular bone need to be separated first. Some studies have focused on dissociating cortical bone from trabecular bone by automatic segmentation and analysing them separately (Gross et al., 2014; Lublinsky et al., 2007), using a thickness calculation algorithm (Hildebrand & Rüegsegger, 1997) in order to obtain 3D maps of the cortical thickness, while another (Tsegai et al., 2017) used an automatic threshold‐based segmentation of the bone using grey‐scale variation of the slices to obtain an outer surface and an inner surface, delimiting the cortex. These methods allow to map the entire bone and as such are also applicable to short and irregular bones.

Recently, the use of 2D maps of bone cortical thickness generated by unrolling 3D maps has allowed the quantitative comparison of cortical thickness maps using GMMs. These studies assimilated the diaphyses of long bones to cylinders (obtaining 3D cortical thickness maps of the diaphyses) before unrolling them to compare their topographies (Bondioli et al., 2010; Puymerail et al., 2012), thus obtaining two‐dimensional (2D) cortical thickness maps. However, this method is only applicable to long bone diaphyses, since the complex shape of epiphyses does not allow comparable unrolling. Similarly, it is not applicable to irregularly shaped bones.

Here, we use a combination of these methods, obtaining both 3D maps of the entire bones, allowing us to study cortical thickness variation in the whole bones including the epiphyses, and 2D maps of the diaphyses, to be analysed quantitatively in order to compare the cortical thickness distribution using GMMs. We first isolated an outer surface (corresponding to the outer surface of the bone) and an inner surface (corresponding to the inner limit of the compact cortex) for each bone. This is required for the removal of the cavities located in the compact cortex and the trabeculae in the medullary space. The removal was performed on Avizo. The ‘PointWrap’ function was added to manual segmentation in order to increase smoothness and avoid artificial discrepancies between sections. We then generated 3D bone cartographies using the ‘SurfaceDistance’ module in Avizo, i.e., calculating the thickness of cortical bone by measuring the distance between the outer and the inner surfaces of the cortex, and generating 3D cortical thickness maps of the entire bones using absolute values. These cartographies enabled the visualisation of variations in cortical thickness among each bone, to make comparisons (in absolute values) between the bones, with the objective to relate these results with the zones of insertion of muscles and ligaments on the bones.

2.5.2. 2D mapping

We used the ‘morphomap’ package in R (Profico et al., 2021) in order to quantitatively analyse bone cartographies by comparing their planar representations, using GMMs, once converted to 2D maps. 3D bone cartographies are required to create the 2D maps; they were obtained using the ‘morphomap3Dmap’ and ‘morphomapThickness’ functions. The rendering of the 3D maps made in Avizo (see above) enabled us a better visualisation of the cortical thickness distribution in the epiphyses, so that these ‘morphomap’ cartographies were only used for 2D mapping (that cannot be obtained based on the Avizo cartographies). The conversion of 3D cortical maps to 2D maps requires the 3D maps to be assimilated to tubes and thus the removal of the epiphyses. The epiphyses were removed by specifying in the ‘morphomapCore’ function the percentage of the bone's length where the mapping was to begin and to end. We chose to place this limit at 20% and 80% for the humerus and radius (following Profico et al., 2021) because it was consistent with the epiphyses' proportions, and 20% and 75% for the ulna, since the ulnar proximal epiphysis is proportionally longer. In order to obtain a planar representation of the topographic thickness variation, the 3D cortical thickness maps were virtually unzipped along a vertical line and unrolled into a plane using the ‘morphomap2Dmap’ function (Figure 1). During this operation, all maps were standardised to the same size, and the thickness value associated with each pixel (visually represented by its colour) was standardised between 0 and 1. All subsequent comparisons were thus made on relative (not absolute) values. Since the maps fully overlap and contain the same number of pixels, it is possible to perform GMM analyses by considering each pixel as a semi‐landmark and exporting them in a dataframe using the ‘morphomapDF’ function. GMM procedures were performed using the procSym function of the ‘Morpho’ package of R (Schlager, 2017), as well as the procD.lm and morphol.disparity functions of the ‘geomorph’ package (Adams & Otárola‐Castillo, 2013).

FIGURE 1.

2D cortical mapping of the ulna of a specimen of Martes martes (Mm2) with the corresponding 3D cortical maps in posterior, medial, anterior and lateral views (P: posterior, M: medial, A: anterior, L: lateral). Cortical thickness is represented by a gradient ranging from cold (low cortical thickness) to warm (high cortical thickness) colours. Refer to material and methods section for explanation on selected limits

2.6. Quantitative bone analyses

Analyses were performed on the two datasets described above, comprising all M. foina specimens and all adult specimens, respectively. We performed Procrustes ANOVAs on the 2D superimposed maps in order to check for differences in cortical thickness distribution between the two species of martens, as well as between the juvenile and adult M. foina specimens. Similarly, as for the quantitative muscle data, a PCA was performed to visualise the distribution of the specimens in the morphospace delimited by the PCs. Since all maps were standardised, this PCA was performed on the entirety of the sample, thus on adult and juvenile specimens and from the two species. We tested the effect of size within the PCAs using linear regression on the first two PCs with MaxL as a size estimate.

We used t‐tests to test for difference in mean thickness (AmeanT, RmeanT) and maximal thickness (AmaxT, RmaxT) in the M. foina and the adult datasets and for difference in compactness (C), and trabecular ratio (%Trab) in the adult dataset.

2.7. Covariation between muscle strength and inner bone anatomy

We investigated the correspondence between areas of high cortical thickness and the muscles' insertion areas, that is, the places where the mechanical forces of the muscles apply on the bone.

We first visually compared, on the entire bones, the areas of high cortical thickness (the highest values indicated in red on the 3D cortical bone maps) to the muscles' origin and insertion areas.

Then, using the 2D unrolled cortical maps, we performed two‐block partial least squares (2‐block PLS) analyses using the ‘two.b.pls’ function in the R package ‘geomorph’, to quantify the degree of covariation between muscle ACSA and superimposed bone thickness maps (following Harbers et al., 2020). PLS were performed on all muscles attached to the diaphysis regardless of their strength category or their functional group, as well as on the different functional groups of muscles (Table 2, Table S1).

A 2‐block PLS was also performed to quantify the degree of covariation between muscle ACSA and the microanatomical parameters for each bone.

3. RESULTS

3.1. Muscle analyses

3.1.1. Principal components analyses

The following results are for the size‐corrected muscle data; analyses on uncorrected muscle ACSA values yielded similar results (see Table 3 for details).

TABLE 3.

Results from the statistical analyses

		Bone microanatomy t‐test						Muscles (diaphysis only)MANOVA		Muscles (all)MANOVA		Cortical mappingProcrustes ANOVA
		%Trab	C	AmeanT	RmeanT	AmaxT	RmaxT	log(ACSA)	Residuals	log(ACSA)	Residuals	2D maps
Interspecific variation (n = 9)	Humerus	p = 0.60	p = 0.30	p = 0.08	p = 0.46	p = 0.09	p = 0.40	p = 0.85 r2 = 0.05	p = 0.91 r2 = 0.05	p = 0.03 r 2  = 0.25	p = 0.04 r 2  = 0.25	p = 0.09 r2 = 0.28
	Radius	p = 0.62	p = 0.25	p = 0.13	p = 0.004	p = 0.57	p = 0.008	p = 0.41 r2 = 0.1	p = 0.42 r2 = 0.12	p = 0.56 r2 = 0.1	p = 0.58 r2 = 0.10	p = 0.29 r2 = 0.13
	Ulna	p = 0.20	p = 0.19	p = 0.17	p = 0.003	p = 0.38	p = 0.007	p = 0.80 r2 = 0.07	p = 0.47 r2 = 0.10	p = 0.49 r2 = 0.1	p = 0.45 r2 = 0.10	p = 0.23 r2 = 0.16
Intraspecific variation (n = 10)	Humerus			p = 0.016	p = 0.94	p = 0.48	p = 0.72	p = 0.004 r 2  = 0.44	p = 0.99 r2 = 0.05	p = 0.003 r 2  = 0.47	p = 0.15 r2 = 0.15	p = 0.002 r 2  = 0.61
	Radius			p = 0.016	p = 0.48	p = 0.33	p = 0.36	p = 0.001 r 2  = 0.59	p = 0.81 r2 = 0.06	p = 0.001 r 2  = 0.57	p = 0.96 r2 = 0.04	p = 0.02 r 2  = 0.44
	Ulna			p = 0.03	p = 0.45	p = 0.19	p = 0.36	p = 0.002 r 2  = 0.52	p = 0.71 r2 = 0.07	p = 0.001 r 2  = 0.45	p = 0.93 r2 = 0.05	p = 0.08 r 2  = 0.26

Note: In grey: not available because of missing epiphyses in juvenile specimens

Abbreviations: %Trab, trabecular ratio; ACSA, anatomical cross‐sectional area; AmaxT, absolute maximum thicknes; AmeanT, absolute mean thickness; C, compactness; RmaxT, relative maximum thickness; RmeanT, relative mean thickness. Significant results are indicated in bold.

Results of the PCA on all muscles attaching to the humerus show that the two first PCs, which represent 37.3% and 23.1% of the variance, respectively, enable the differentiation between adult specimens of the two taxa (Figure 2a); the overlap on the first axis is due to a single specimen, Mf7. All muscles contribute to the separation (along PC1) similarly, though with differing intensities. The muscle that contributes the most to this separation is the M. flexor digitorum profundus (FDP), which is a major flexor of the wrist. There is no effect of size on the first two axes of the PCA (PC1: p = 0.42; PC2: p = 0.95). When taking only muscles inserting on the diaphysis into account, there is an almost complete overlap of the two taxa along the first axis (Figure 2b).

FIGURE 2.

Results of the Principal Components Analysis of size‐corrected anatomical cross‐sectional area (ACSA) considering (A) all muscles attached to the humerus and (B) muscles attached to the humeral diaphysis. The plots display the variation along the first two axes, along with the muscle contributions (arrows). For muscle abbreviations, see Table 2. Variables contributing the least are not shown in order to increase the visibility of the graph. Mm = Martes martes, Mf = Martes foina (adult specimens only)

PCAs on the ACSA of the muscles inserting on the radius and ulna show that the two taxa broadly overlap, whether we consider all muscles or only those that attach to the diaphysis (Figures S2).

Results of the PCAs on the M. foina dataset show that when all muscles attaching to the humerus are taken into account, adult and juvenile specimens are slightly differentiated along the first two axes (PC1 = 35.1%; PC2 = 18.5%), mainly under the influence of the FDP muscle. When taking only muscles inserting on the diaphysis into account, there is an almost complete overlap of adult and juvenile specimens.

Similarly as in the adults‐only dataset, we observe an almost complete overlap of adult and juvenile specimens of M. foina in the results of the PCAs on muscles inserted on the radius and ulna, whether we considered all muscles or only those that attach to the diaphysis (Figure S3).

3.1.2. MANOVAs

The linear regressions (ACSA~muscle length) detected no significant allometry within the entire muscle dataset (adult and juveniles) (p = 0.57, r² = 0.54).

MANOVAs on the adult only dataset indicated that when all muscles (n = 26) attaching to the humerus are taken into account, there is a significant difference between the two species (p = 0.04, r² = 0.25). No discrimination between the two species is observed when focusing only on muscles that attach to the humeral diaphysis (n = 15; MANOVA: p = 0.91, r² = 0.05). In order to investigate if that difference was also found when considering juvenile specimens, we used MANOVAs on the entire sample (adult and juveniles); we found no significant difference between the two species, whether considering all muscles attaching to the humerus (p = 0.27, r² = 0.1) or on the humeral diaphysis only (p = 0.96, r² = 0.03) (Figure S4).

MANOVAs on ACSA of all muscles show no significant difference between the two species neither for the radius (p = 0.58, r² = 0.10) nor for the ulna (p = 0.45, r² = 0.10); MANOVAs on ACSA of muscles inserting on the diaphysis yielded similar results (radius: p = 0.42, r² = 0.12; ulna: p = 0.47, r² = 0.10).

MANOVAs on uncorrected muscle data yielded similar results for all cases listed above (see Table 3).

MANOVA on the M. foina sample detected significant differences in muscle ACSA between adult and juvenile M. foina specimens when using uncorrected muscle data, in all three bones. These differences were not found when using size‐corrected muscle ACSA. Results were similar whether we considered all muscles or only muscles attached to the diaphyses (see Table 3).

3.2. Microanatomical analyses

3.2.1. Microanatomical parameters

PCAs on the microanatomical parameters show a distinction between adult specimens of the two taxa along the first axis for the three bones (Figure 3, Figure S5). PC1 represents over 65% of the variance in each case, and the four variables have a similar contribution along that axis. Although there is always a small overlap, the two species tend to be discriminated along the first axis, with RmeanT and RmaxT having a predominant contribution.

FIGURE 3.

Results of the Principal Components Analysis displaying the variation along the two first axes using microanatomical parameters of the ulna, along with their contributions (arrows). %Trab: trabecular ratio, C: Compactness, RmeanT: relative mean cortical thickness, RmaxT: relative maximal cortical thickness. Mm = Martes martes, Mf = Martes foina (adult specimens only)

In M. foina specimens, juvenile had a significantly lower absolute mean cortical thickness between than the adults, in all three bones (AmeanT: humerus: p = 0.016; radius: p = 0.016; ulna: p = 0.03). This difference was not detected when using relative values (RmeanT: humerus: p = 0.94; radius: p = 0.48; ulna: p = 0.45).

There was no significant difference in compactness nor in trabecular ratio between adults of the two species of martens (Table 3). Although the two taxa did not differ in mean and maximum cortical thickness for the humerus, difference was significant for the radius (RmeanT: p = 0.004; RmaxT: p = 0.008) and ulna (RmeanT: p = 0.003; RmaxT: p = 0.007), specimens of M. foina having a smaller RmeanT and RmaxT than those of M. martes.

3.2.2. 2D maps

PCAs on the 2D cortical thickness maps show a difference in distribution between the two taxa along the first axis for each bone, PC1 representing more than 80% of the variance for the humerus and radius (Figure 4, Figure S6). In each PCA, the specimens are distributed in the same order with a small overlap along the first axis: first the M. martes specimens, followed by the adult then juvenile M. foina specimens, with the exception of the Mm10 specimen in the PCA on ulna maps. Linear regression detected no effect of size in the first PC of the humerus PCA (p = 0.61, r² = 0.03), but a significant effect of size in its second PC (p = 0.02, r² = 0.35), this axis representing 3.33% of the variation. There was no effect of size along the first axes of the PCAs on the radius and ulna (radius: PC1: p = 0.09, PC2: p = 0.59; ulna: PC1: p = 0.09, PC2: p = 0.80).

FIGURE 4.

Results of the Principal Components Analysis displaying the variation along the two first axes using 2D cortical mappings of the humerus. Adults are visualised in the solid circles and juveniles in the open circles. Mm = Martes martes, Mf = Martes foina

Procrustes ANOVAs on the 2D maps detected no significant difference between the two taxa, but there was a significant difference between adult and juvenile M. foina specimens in the humerus (p = 0.002, r² = 0.61) and the radius (p = 0.02, r² = 0.44) but not the ulna (p = 0.014, r² = 0.64).

3.3. Covariation between muscle anatomy and inner bone structure

3.3.1. Qualitative comparisons based on 3D maps

Humerus

Among the muscles that attach to the humeral diaphysis (n = 15), seven muscles are generally considered ‘strong’ (i.e., ACSA larger than the mean ACSA of all humeral muscles). This includes all four pectoral muscles (PAB + PMJ and PMN + XH) (except PMN + XH for Mf7) and two triceps muscles (TBM and TBLA). In some specimens (Mm2, Mf3, Mf5, Mf6, Mf7, Mf8 and Mf10), one of the two deltoid muscles (DA) is considered ‘strong’ as well.

Some of the strong muscles are attached to areas of high cortical thickness: the zone of insertion of the TBM muscle (Figure 5) always appears thicker than the rest of the bone, either in its distal part only (Mf1, Mf2, Mf3, Mf7, Mf9 and Mm2) or in its entirety (Mf5, Mm1 and Mm3). The insertions of the superficial pectoral muscles (PAB + PMJ) are areas of high cortical thickness in most specimens (Mf1, Mf2, Mf3, Mf7, Mm2 and Mm3), but not all (Mf5, Mf9, Mm1). Similarly, the TBLA muscle is attached to areas of high cortical thickness in some specimens (Mf2, Mf3 and Mf5) but not in others. One of the strong muscles (PMN) is attached to areas of low cortical thickness. The DA is a particular case: it is strong in some specimens (Mf2, Mf9 and Mm3), in which the crest on which it inserts is thicker than the rest of the bone, and weak in others (Mf1, Mf3, Mf5, Mf7, Mm1 and Mm2), where it almost always inserts on thin cortical bone (except for Mf5).

FIGURE 5.

Representations of the origin (red) and insertion (blue) of muscles on the humerus, with the corresponding cortical mapping of the humerus of Martes martes (Mm2) in cranial (a, e), lateral (b, f), caudal (c, g) and medial (d, h) view. Cortical thickness is represented by a gradient ranging from cold (low cortical thickness) to warm (high cortical thickness) colours on the 3D mapping. For muscle abbreviations see Table 2

Some areas of high cortical thickness are insertion areas of weak muscles: The BCH muscle inserts on almost half of the diaphysis (Figure 5a–c). In all specimens, at least part of its insertion appears thick, whether it be more proximal (Mf5, Mm2) or distal (Mf2, Mf3, Mf7, Mm1 and Mm3). The ANC muscle is inserted on the caudal side of the lateral crest, which is always very thick as compared to the rest of the bone (Figure 5c,g). The CB muscle is also sometimes inserted on thick cortical bone since it is attached to the medial side of the cranial crest (Figure 5a), which is sometimes thicker than the rest of the bone (Mf1, Mf3 and Mf7). Additionally, there are two zones that are almost always thick but on which no muscle is inserted: the first one (except for Mf5) is the distal part of the medial side, which is the junction between the diaphysis and the medial epicondyle. This area bears several ligament insertions that contribute to the stability of the elbow joint capsule. The second one (in all specimens, although it is slightly less visible on Mf5) is the cranial extension of the TBM insertion, visible on the medial side.

There is almost no variation of the cortical thickness distribution in the proximal epiphysis: all adult specimens display the same pattern of thickening on both tuberosities (lesser and greater); although the proximal epiphysis bears several muscle attachments, these cortical thickenings do not correspond to any of those areas of attachment in particular. The proximal epiphysis is missing in two of the juvenile specimens (Mf8 and Mf10), but the pattern is similar in the other two. The cortex of the lesser tuberosity of the Mf4 specimen appears thicker than in the greater tuberosity.

There is more variation in the cortical thickness distribution in the distal epiphysis: all specimens display a very thick medial supracondylar ridge, on which both the PT and TBA muscles originate (Figure 5c,d), and a caudal part of the olecranon thicker than its cranial part. Both epicondyles are generally thicker than the rest of the epiphysis but there is no clear relationship with muscle insertions apart from that of the FCUH muscle on the medial side and sometimes EDL (Mf2 and Mf10) and/or ECU (Mf1, Mf2 and Mf10) on the lateral side. The most distal part of the epiphysis appears sometimes thicker than the rest and corresponds to the PL + FDP origin (Mf3, Mf4, Mf7, Mf9, Mf10 and Mm3).

Despite their absolute cortical thickness being distinctly thinner than in adults, juvenile specimens show the same pattern of cortical thickness distribution. The specimen with the thinnest cortex (Mf4) does not yet have sufficient variation in cortical thickness distribution along the shaft to clearly determine if there are cortical thickenings corresponding to muscle insertions: the lateral crest and the diaphysis are thicker than the epiphyses, but the posterior crest is still very thin. The other three juveniles show the same global pattern of cortical thickness distribution as adults, only thinner and less defined. The Mf10 specimen exhibits a thickening of the cortex in the very proximal part of the diaphysis, which corresponds to the TBM insertion.

We found no obvious differences in cortical thickness distribution between M. martes and M. foina.

This comparative analysis does not show a true correlation between muscle strength and cortical thickness distribution in the humerus.

Radius

Among the muscles that attach to the radial diaphysis (n = 6), only two are considered ‘strong’: the pronator teres muscle (PT) (except for Mf1, Mf2, Mf4 and Mf10) and the FDP muscle. Other muscles are ‘strong’ in two or three specimens only (the EP muscle in Mf3, Mf5 and Mf7; the FCR muscle in Mm3 and Mf5; and the BCR muscle in Mf1, Mf2 and Mf3), and as such are considered weak. The PT muscle inserts onto the proximo‐medial part of the radius diaphysis, which is an area of increased cortical thickness in all adult specimens (Figure 6a,d), although it seems to be slightly thinner in M. martes than in M. foina. The FDP muscle inserts into the lateral part of the diaphysis and partly on the proximal part of the distal crest (Figure 6d). Most of the time it is also attached to areas of increased cortical thickness, which corresponds to the zone of contact with the ulna and the interosseous membrane connecting the radius to the ulna. The four remaining muscles (EP, FCR, PQ and SUP), as well as the BCR muscle that inserts into the styloid process of the radius, are considered ‘weak’. The EP, PQ and BCR muscles insert into areas of ‘standard’ cortical thickness, while the FCR muscle is often attached to areas of increased cortical thickness. However, it is attached between the PT and FDP muscles (Figure 6d) so it is impossible to ascertain to which muscle insertion the cortical thickening is linked. Apart from the “cranial crest” leading to the radial styloid process that has a very thick cortex in all specimens (Figure 6f,g), there is no particular area of high cortical thickness that does not correspond to any muscle (with the exception of the proximal epiphysis of Mf10).

FIGURE 6.

Representations of the origin (red) and insertion (blue) of muscles on the radius (transparent ulna), with the corresponding cortical mapping of the radius of Martes martes (Mm2) in cranial (a, e), lateral (b, f), caudal (c, g) and medial (d, h) view. Cortical thickness is represented by a gradient ranging from cold (low cortical thickness) to warm (high cortical thickness) colours on the 3D mapping. For muscle abbreviations see Table 2

There are no muscle attachments to the proximal epiphysis of the radius, and one attachment (BCR muscle) on the distal one, on the styloid process of the radius (Figure 6a,d).

There is little variation in the proximal epiphysis. Almost all specimens display a slight thickening of the cortex along the articular circumference (Mf1, Mf2, Mf3, Mf6, Mf7, Mf8, Mm1, Mm2, Mm3) and/or the articular fovea, which is in contact with the medial epicondyle of the humerus (Mf1, Mf3, Mf5, Mf7, Mf8, Mm3), and around the radial tuberosity (except for Mf10), which bears a part of the PT muscle's insertion. Two specimens display a homogenous distribution (very thin) of the cortical thickness in the radial head: a juvenile (Mf4) and an adult (Mf9). Two of the juveniles (Mf6 and Mf8) display the same pattern as that of the adult, while Mf4 is too thin to see any cortical thickness variation. The last juvenile specimen, Mf10, is the only one to deviate from the general pattern, with an extremely thick lateral articular fovea and circumference.

The cortical distribution in the distal epiphysis of the radius is similar in all M. foina specimens, with a slightly thicker cortex in the carpal articular surface (CAS) and the styloid process. The three specimens of M. martes have a thinner cortex in the CAS than in the rest of the bone, and a thicker cortex in their styloid processes.

Juveniles exhibit a similar distribution pattern as adults. Mf4 is too thin to clearly distinguish cortical thickness variation but the cortical bone is thicker in the diaphysis than in the epiphyses, and the zone of contact with the ulna is very thin, surrounded by a thicker cortex, as found in adults. The distal epiphyses are missing in both Mf8 and Mf10. The lateral crest is not fully ossified in Mf6, and Mf10 displays a very thick cortex on the lateral part of the proximal epiphysis.

We found no obvious differences in cortical thickness distribution between M. martes and M. foina.

Ulna

Among the muscles that attach to the ulnar diaphysis (n = 7), two are strong muscles: the FDP muscle and the M. biceps brachii (BB except in Mf4 and Mf9). Another muscle (FCUU, the M. flexor carpi ulnaris) is strong in some specimens (Mm1, Mm2, Mf1, Mf2 and Mf3). The four remaining muscles (EI, EP, BCH and PQ) are considered ‘weak’.

Areas of high cortical thickness correspond to the contact zones between the radius and the ulna, the interosseous membrane connecting them on the cranial side (Figure 7e), and to the lateral crest on the caudal side on which the FDP, FCUU and EP muscles are inserted (Figure 7g). This pattern is similar in all adult specimens regardless of the FCUU muscle's status. All the strong muscles are thus inserted in areas of high cortical thickness. Similarly, juveniles (with the exception of Mf4 whose cortex is too thin to discern patterns) exhibit the same pattern, though they are naturally overall thinner. The distal crest corresponds to an area of increased cortical thickness in seven individuals (Mf1, Mf2, Mf3, Mf5, Mf9, Mm2 and Mm3).

FIGURE 7.

Representations of the origin (red) and insertion (blue) of muscles on the ulna (transparent radius), with the corresponding cortical mapping of the ulna of Martes martes (Mm2) in cranial (a, e), lateral (b, f), caudal (c, g) and medial (d, h) view. Cortical thickness is represented by a gradient ranging from cold (low cortical thickness) to warm (high cortical thickness) colours on the 3D mapping. For muscle abbreviations see Table 2

There are numerous muscle insertions on the ulnar head, and none on its distal epiphysis. The same pattern is found in almost all specimens: the cortex of the proximal epiphysis is thicker on the caudomedial side of the olecranon, which corresponds to the TBLO muscle insertion (Figure 7c,g), as well as on the trochlear notch and the medial coronoid process (Figure 7h).

There are two exceptions, both juveniles: Mf4, which shows a very thin cortex and a thickening of the TBLO insertion zone but not of the trochlear notch and coronoid process, and Mf10, which displays a completely different distribution of the cortical thickness. The olecranon of Mf10 is thicker on the craniomedial side, which corresponds to the TBLA insertion zone, and the area around the TBLO insertion zone is thicker than the insertion zone itself. The medial coronoid process is extremely thick.

The cortical thickness distribution of the distal epiphysis is similar in almost all specimens: both medial and lateral sides are very thin, with a thicker area above the styloid process. There are three exceptions (Mf4, Mf6 and Mm1), and two specimens with missing epiphyses (Mf8 and Mf10). There is no discernible cortical thickness variation in Mf4 and Mm1, and in Mf6 the styloid process appears to be thicker than the area above it.

We found no obvious difference in cortical thickness distribution between M. martes and M. foina.

3.3.2. Quantitative analyses of the covariation.

2b‐PLS on the superimposed maps and muscle data uncorrected for size detected no covariation between cortical thickness distribution and muscle ACSA (see Table 4). However, 2b‐PLS on the superimposed maps and size‐corrected muscle data indicate a correlation between the two parameters in all three bones. The results are similar whether we consider all muscles or functional groups separately, with the exception of the PQ muscle, for which no correlation is detected when using corrected muscle data (Table S3). An increase in muscle ACSA value is correlated to an increased contrast in thickness along the diaphyses (Figure 8a–c). The two extreme deformations of the cortical thickness maps appear to correspond to juvenile et adult patterns respectively, the adult pattern having the greater contrast in cortical thickness. And indeed, these correlations are found only when taking juvenile specimens into account, but not when considering adult specimens only (Table 4).

TABLE 4.

Results from the covariation analyses (Two‐block partial least squares analyses) using (a) the log(ACSA) values and (b) residuals from the linear regression

(a)	All specimens (n = 13)	M. foina (adults+juveniles)(n = 10)	Adult specimens (M. martes + M. foina) (n = 9)		Juvenile specimens (n = 4)	M. martes (n = 3)
	ACSA/2D maps	ACSA/2D maps	ACSA/MP	ACSA/2D maps	ACSA/2D maps	ACSA/2D maps
Humerus	p = 0.003 r‐PLS = 0.89	p = 0.02 r‐PLS = 0.85	p = 0.4 r‐PLS = 0.53	p = 0.32 r‐PLS = 0.68	p = 0.02 r‐PLS = 0.99	p = 0.91 r‐PLS = 0.94
Radius	p = 0.001 r‐PLS = 0.83	p = 0.02 r‐PLS = 0.76	p = 0.63 r‐PLS = 0.30	p = 0.76 r‐PLS = 0.54	p = 0.14 r‐PLS = 0.86	p = 0.08 r‐PLS = 0.99
Ulna	p = 0.003 r‐PLS = 0.85	p = 0.04 r‐PLS = 0.83	p = 0.29 r‐PLS = 0.48	p = 0.42 r‐PLS = 0.65	p = 0.02 r‐PLS = 0.96	p = 0.59 r‐PLS = 0.91

(b)	All specimens (n = 13)	M. foina (adults + juveniles)(n = 10)	Adult specimens (M. martes + M. foina) (n = 9)		Juvenile specimens (n = 4)	M. martes (n = 3)
	ACSA/2D maps	ACSA/2D maps	ACSA/MP	ACSA/2D maps	ACSA/2D maps	ACSA/2D maps
Humerus	p = 0.99 r‐PLS = 0.33	p = 0.86 r‐PLS = 0.61	p = 0.26 r‐PLS = 0.78	p = 0.80 r‐PLS = 0.62	p = 0.80 r‐PLS = 0.93	p = 0.91 r‐PLS = 0.86
Radius	p = 0.90 r‐PLS = 0.44	p = 0.82 r‐PLS = 0.56	p = 0.90 r‐PLS = 0.45	p = 0.50 r‐PLS = 0.69	p = 0.37 r‐PLS = 0.97	p = 0.26 r‐PLS = 0.99
Ulna	p = 0.72 r‐PLS = 0.48	p = 0.71 r‐PLS = 0.54	p = 0.74 r‐PLS = 0.49	p = 0.57 r‐PLS = 0.65	p = 0.37 r‐PLS = 0.83	p = 0.42 r‐PLS = 0.95

Abbreviations: ACSA, anatomical cross‐sectional area; MP, microanatomical parameters. Significant results are indicated in bold.

FIGURE 8.

Results of the two‐block partial least squares (2b‐PLS) regressions between the 2D cortical maps of the (a) humerus, (b) radius, (c) ulna and anatomical cross‐sectional area (ACSA) blocks. Adults are visualised by solid circles and juveniles by open circles. The black line represents the PLS regression line. Singular vectors for muscle ACSA blocks are shown using barplots, extreme cortical thickness patterns are shown using 2D mappings. Mm = Martes martes, Mf = Martes foina

In the humerus and radius, all strong muscles have a comparable impact on the covariation. In the ulna, the effect of the FCUU muscle is more important than that of the other two muscles.

In the humerus, an increase in muscle ACSA values is generally associated with a shift in higher cortical thickness towards the distal part of the caudal side, which corresponds to the insertion zone of three muscles (ANC, distal part of the TBM and BCH). In the radius, we observe almost an inversion of the cortical thickness distribution when the muscle ACSA values increase: high muscle ACSA values are associated with a thick cortex in the areas of contact with the ulna while low muscle ACSA values are associated with a thicker cortex surrounding these areas, with a very thin radial tuberosity. In the ulna, all muscle ACSA values are associated with similar patterns of cortical thickness distribution, but we observe a sharp increase in contrast between the different zones when muscle ACSA values increase, forming the cranial and caudal thickenings of the proximal part of the shaft.

4. DISCUSSION

4.1. Muscular variation at the inter‐ and intraspecific levels

Taverne et al. (2018) showed that the forelimb musculature evolved in a convergent manner in carnivoran arboreal species, resulting in more developed wrist flexors/rotators and elbow flexors. But if these authors studied the musculature of the forelimb as a whole, here we focused on the muscles attaching to the humerus, radius and ulna separately. Although there is a slight trend towards stronger wrist flexors in the more arboreal M. martes, the difference in quantitative muscular anatomy between M. martes and M. foina in our study is not significant. The only exception is found in the distal humeral epiphysis, on which a powerful flexor is attached (Figure 2a); this muscle allows the flexion of the wrist, its greater strength thus appears to be advantageous for arboreal locomotion. The main muscle contributing to this difference is the M. flexor digitorum profundus (FDP), a powerful elbow extensor and wrist flexor inserting on the medial epicondyle of the humeral distal epiphysis, that is also involved in pronation/supination movements. Studies have linked better radio‐ulnar rotation capacities to a more arboreal lifestyle in mustelids (Fabre et al., 2013b, 2015). The functional role of the FDP muscle suggests that this muscular difference between the two species is linked to the more arboreal lifestyle of the pine marten. This difference was less significant when adding the four M. foina juveniles (Figure S1). The humeral muscle strength of the M. foina juveniles appears to be more similar to that of the adult M. martes than to that of the adult M. foina. We lack data on the hindlimb, but it may be possible that a relatively strong forelimb in the young stone martens may compensate for their overall physiological immaturity. During ontogeny, relative muscle strength in the forelimb appears to decrease. In primates, a decrease in relative manual grasping force from juvenile to adult mouse lemurs has been shown and linked to a shift in the recruitment of both the fore‐ and hindlimbs (in juveniles) to a hindlimb‐dominated recruitment (in adults) (Boulinguez‐Ambroise et al., 2020).These results highlight the integrative nature of the animal's body plan. Indeed, if analysing one body part, such as a single long bone, is not sufficient to identify specific variations linked to differences in locomotor mode, because the forelimb works as a functional unit, these variations become noticeable by combining analyses on the three long bones of the forelimb, and their possible causes can be investigated.

There is a significant difference between the two taxa when focusing on all the muscles attaching to the humerus, but not when considering only muscles attaching to the humeral diaphysis. This suggests that the functional signal is stronger in muscles that insert or originate near the shoulder and elbow joint.

4.2. Bone microanatomy and cortical thickness distribution.

The external limb bone morphology of M. martes and M. foina is very similar (Fabre et al., 2013a, 2013b) despite their different locomotor modes. In light of almost unnoticeable external morphological differences, this makes them an ideal case study to investigate whether microanatomical changes do occur and whether they reflect different mechanical stress distribution.

Although the muscular differences between the two taxa lie predominantly in the elbow joint, no accompanying variation of cortical thickness distribution in the humeral distal epiphysis is observed between the two species. However, despite the absence of interspecific muscular differences in the zeugopod, we found a significantly thicker cortex (relative mean and maximal thicknesses) in the radius and ulna of M. martes and with no effect of size. Although they do not differ significantly, both the compactness and the trabecular ratio are higher in M. foina than in M. martes, suggesting a slight trend towards more compact bone in semi‐arboreal species.

Although M. martes and M. foina share similar feeding habits, M. foina's diet includes more plant food and insects than that of M. martes; it is especially the case when the two species are sympatric since they avoid competing with one another by assuming different ecological niches (Granata et al., 2021; Posłuszny et al., 2007). Additionally, pine martens are able to run along branches and jump from tree to tree in pursuit of prey, while stone marten tend to hunt on the ground, occasionally climbing trees to reach on bird nests but without pursuing prey in the treetops (Heptner & Sludskii, 2002; Grabham et al., 2019; Sidorovich et al., 2005). The differences we observed between the forelimbs of the two species could thus be related to their distinct locomotor modes (running along branches vs. running on the ground, jumping between trees vs. occasionally climbing one), but also by their different hunting behaviours (more grasping of prey for the pine marten, although Fabre et al., 2013b put both species in the same grasping category, namely “poorly developed grasping ability”). This suggests that the radius and ulna respond differently to an arboreal lifestyle and hunting habits than the humerus. Since there was no difference in maximal length nor absolute mean/maximal thickness, the difference in relative thickness might result in a slimmer medullar cavity in M. martes, without external shape modification.

The covariation of the radius and ulna is not surprising in itself since they are closely linked in their functional roles. The different response of the humerus is interesting since several studies found that the shape of the humerus covaried more with the radius than with the ulna in response to changes in body mass and muscle anatomy (Fabre et al., 2013a; Martin et al., 2019). Here, we found instead similar microanatomical variations in the radius and ulna, but not with the humerus.

In the three bones, we found no interspecific difference in the 2D diaphyseal maps. Both species exhibit the same distribution in cortical thickness, although it appears visually thicker (absolute values) in the arboreal pine marten. These results support the quantitative microanatomical analyses suggesting that the higher cortical thickness observed in the radius and ulna of M. martes was associated with a reduced medullar cavity.

If there was no difference between the two species, the 2D patterns of diaphyseal cortical topography were however significantly different between adult and juvenile specimens of M. foina. After each map's minimum and maximum values were standardised, we were able to compare the relative cortical thickness distribution of the whole sample without being hindered by lower absolute values in juvenile specimens. The extreme cortical topographies provided by the covariation analysis of the 2D diaphyseal cortical topography correspond to the shift from a juvenile to an adult pattern, with more contrast in relative thickness along the shaft: in the humerus, the thickest parts of the diaphysis shift from the proximal to the distal part of the shaft, which bears the insertions of two elbow extensors, the M. anconeus (ANC) and the M. triceps brachii (TBM) muscles. In the radius, we observe a sharp increase of relative cortical thickness in the zones of contact with the ulna where the cortex was previously very thin, while in the ulna these areas are already thicker than the rest of the shaft in juveniles, and their absolute thickness increases during growth without change in the relative thickness general pattern. This indicates that changes in cortical thickness distribution during growth are more pronounced in the humerus and radius than in the ulna. This shift is most likely related to muscular insertions that are not yet highly solicited in young individuals; a wider sample with more juveniles representing the various ontogenetic stages might allow for a better understanding of the link between muscle and bone microanatomy during the growth of the individuals.

4.3. Covariation between muscles and bone microanatomy.

The third objective of this study was to assess whether the ACSA of the muscles (as a proxy of muscle strength) is reflected in the bone microstructure considering that bones adapt their outer and inner structures to mechanical stresses.

Our 3D maps show mixed results concerning the relationship between cortical thickness and muscular attachments: we observed a clear link for some muscles, but not for all. It is coherent with the fact that muscle attachments can sometimes cover wide areas (e.g., the BCH muscle on the humerus). In those cases, the mechanical strain can be distributed unevenly along the bone surface, or be too spread out to have an impact on the bone cortical thickness. This result supports the observations of Cuff et al. (2020), who explored the relationship between bony attachment areas and both muscle mass and muscle insertions. Cuff et al. (2020) concluded that although “muscle scars” were rarely correlated with muscle mass, and as such could not be used for inferences in fossil taxa, those who were correlated were highly so. Further studies may be able to determinate for which muscles inferences are possible and to which extent.

We observed a covariation between muscle ACSA and cortical thickness distribution (2D maps) in all three bones, but only when juveniles were included, and only on data uncorrected for size. As our small sampling could be responsible for the non‐significance when testing for a muscle/bone covariation, the tests were repeated a number of times with a smaller sample (n = 5) of random specimens, and a covariation was observed only when juveniles were present. This suggests that the covariation is linked to changes during ontogeny; that is coherent with the limited variations in cortical thickness observed on 3D maps among adult specimens.

However, if the cortical thickness distribution does not seem to reflect muscle strength, it appears to reflect the attachment of ligaments. It is visible in the radius and ulna where the thickest areas correspond to the contact zones between the bones and to the membrane and ligament linking them, rather than to specific muscle insertions. These observations, while surprising, are consistent with recent results, like those of Harbers et al. (2020), who investigated the impact of captivity and domestication on limb bone cortical morphology in suids. In their study, Harbers et al. found no correlation between muscle ACSA and cortical thickness distribution in the humerus of adult wild boars but found that the cortical thickness distribution was correlated to the age and body mass of the specimens. These results are also consistent with those of Houssaye et al. (2021), who studied the microanatomy of the patella in perissodactyls and found a strong thickening of the cortex where the strong patellar ligament inserts, but observed no particular thickenings associated with muscle insertions. Similarly, on the calcaneum of suids, the ligaments insertion zones appear to be the main factors affecting the cortical thickness distribution (Cottereau, pers. com. 2021). All of these observations are congruent with the hypothesis of Zumwalt (2005), stating that muscular load does not affect the bony attachment areas unless they are pathological.

Muscle strength was not reflected at the microanatomical level. However, previous studies analysing 3D histology have found clearer correlations between bone histology and musculotendinous insertions: Sanchez et al. (2013) were able to determine the position of entheses in fossil vertebrates as well as the approximate orientation of the attached muscle. In addition, Cury et al. (2016) and Zhao et al. (2017) observed histological changes in tendon insertion zones and ligament insertion zones respectively, indicating that cortical bone holds significant information regarding muscular anatomy. Thus, if the histological level is impacted by muscle insertion, the microanatomical organisation appears rather poorly affected and a less efficient level of investigation to infer muscle structure based on skeletal elements.

4.4. Intraspecific versus interspecific variation.

We observed more separation between adult and juvenile specimens of M. foina than between adults of the two species, for all microanatomical parameters. This was due to a greater mean cortical thickness in the forelimb of the pine martens; in the M. foina sample, the thickness appeared to increase proportionally during growth.

Using muscle data, we also observed a greater distinction at the interspecific level than at the intraspecific level in the humerus. As stated above, this variation was almost entirely driven by the strength of one muscle, far greater in the arboreal pine marten than in the more terrestrial stone marten. The lack of differences between adult and juvenile specimens when using size‐corrected data suggests that the intraspecific variation is mainly driven by the increase of muscle force during growth.

Despite these results, the cortical maps revealed greater differences within the M. foina sample than compared to the M. martes sample. This can largely be attributed to the presence of juveniles within the former, in which the zones of high cortical thickness observed in adults are not yet fully developed. When taking only adult M. foina into consideration, the range of variation is similar to the one observed in the M. martes specimens. This is congruent with the fact that the only muscular difference was found in the humeral epiphysis, which is not visible on the 2D cortical maps. The greater cortical thickness detected in the pine marten's stylopod was not reflected in the maps either, since all values were standardised and showed relative variation in the cortical thickness pattern only.

Using muscular and microanatomical data, we found overall more differences between the two species than within the M. foina sample, indicating that we can distinguish between these two species, despite their strong similarity, using muscular and microanatomical data. However, the ontogenic variation is not negligible: muscle strength in juvenile specimens of M. foina is similar to that of the adult M. martes, thus masking the species‐specific muscular distinction in the humerus when taking all specimens into consideration.

The relatively limited number of specimens in this study, as well as the lack of juvenile specimens of M. martes, prevent us to conclude with certainty that the intraspecific variation, because of ontogenetic variation, is greater than the interspecific one. This would notably require a future study using a wider sample of both juvenile and adult specimens in order to better characterise the covariation of muscle and bone during ontogeny.

5. CONCLUSION

The present study of the forelimb muscular anatomy and bone inner structure in two sympatric species of Martes revealed a functional signal in the muscle anatomy: we observed a stronger flexor muscle in the elbow of M. martes than in the elbow of M. foina. While this difference was not directly reflected in the bones' microanatomy, we also observed a global thickening of the cortical bone in the radius and ulna, but not in the humerus, of M. martes, and conclude that the stylopod and zeugopod respond in a different manner to a more arboreal lifestyle. Our 2D maps revealed a clear difference between adult and juvenile specimens of M. foina, but not between the adults of the two species, and thus a shift of cortical thickness distribution pattern during ontogeny, rather than species‐specific patterns. However, using both muscle and bone microanatomical data, we were able to distinguish the two taxa, indicating that although it is not negligible, the intraspecific variation does not hinder species distinction, even between two very close species. Finally, our results show that cortical thickness varies only to some extent with muscular strength at zones of muscle attachment. While the correlation is clear for some muscles, it is not for others. It rather appears that cortical thickness responds prevalently to strains applied by contact between bones and ligament insertions. We thus conclude that inference of muscle information from cortical thickness distribution is possible but only for certain muscles in the humerus.

AUTHOR'S CONTRIBUTIONS

A.H., C.Ba. and C.Bö. designed the study. AH. did the bone data acquisition, C.Bö. did the muscular data acquisition. C.Ba. and M.A. conducted the segmentation. C.Ba. and C.Bö. conducted the analyses. C.Ba. and C.Bö. prepared the figures and drafted the manuscript. All authors contributed to the final manuscript, read it and approved it.

Supporting information

Figure S1

Figure S2

Figure S3

Figure S4

Figure S5

Figure S6

Figure S7A

Figure S7B

Figure S7C

Figure S8A

Figure S8B

Figure S8C

Table S1

Table S2

Table S3

Bader, C , Böhmer, C , Abou, M , Houssaye, A (2022) How does bone microanatomy and musculature covary? An investigation in the forelimb of two species of martens (Martes foina, Martes martes). Journal of Anatomy. 241:145–167. 10.1111/joa.13645

DATA AVAILABILITY STATEMENT

Data Availability Statement: 3D models of the bones used for this study will be uploaded on MorphoSource; R scripts used to perform analyses will be available on GitHub.