How moles destroy your lawn: the forelimb kinematics of eastern moles in loose and compact substrates

Moles (Family Talpidae) provide a classical example in functional morphology of extreme morphological specialization with their remarkable forelimb morphologies dedicated to burrowing (Kley and Kearney, 2007). Unlike those of most terrestrial mammals, the forelimbs of moles are displaced cranially and the palmar aspect of the manus faces laterally at rest, resulting in a secondarily derived, sprawling posture (Fig. 1A). The uniquely short and wide humerus provides large attachment sites for muscles such as teres major, which accounts for 75% of the forelimb muscle volume (Rose et al., 2013). The elongated olecranon process of the ulna provides substantial mechanical advantage for the elbow extensors and articulates with an expanded humeral trochlea, which is uncommon among mammals (asterisk in Fig. 1B). The flexor digitorum profundus muscle, which controls mammalian carpal and digital movements, is tendinous and thus facilitates the transmission of force from the humerus to the widened manus (Rose et al., 2013; Yalden, 1966). These specializations make moles one of the most accomplished tetrapod diggers: tunnels built by a single individual (body length <20 cm; Yates and Schmidly, 1978) can extend 100–1000 m (Arlton, 1936; Hickman, 1983). Moreover, moles may generate digging forces equivalent to 20–30 times their own body weight (Arlton, 1936; Gambaryan et al., 2002; Skoczeń, 1958).

Although morphological specializations of the mole forelimbs are celebrated in the morphological literature (Campbell, 1939; Gambaryan et al., 2002; Piras et al., 2012; Reed, 1951; Rose et al., 2013; Yalden, 1966), we know little about how the forelimb is used during digging. The shape of the shoulder joint (glenoid fossa and humeral head; white cross in Fig. 1B) and surrounding tendinous tissues might limit the mobility of the joint (Gambaryan et al., 2002). A similar limitation on excursion at the elbow is formed by the enlarged trochlear notch (asterisk in Fig. 1B), which is presumed to stabilize flexion and extension of the elbow as the ulnar notch articulates with the humeral trochlea. Despite these specializations, which appear to limit the range of motion of forelimb joints, basic field and laboratory observations have shown that moles vary their tunnel architecture depending upon the compactness of the soil (Arlton, 1936; Hisaw, 1923; Skoczeń, 1958; Lin et al., 2017). When in loose soil near the surface, tunnels form branching, raised mounds of soil on the surface of the ground. When in compact soil, moles excavate and transport soil and deposit it at the surface as mole hills (Fig. 2). These very different tunnel architectures raise the question: do moles always dig the same way or do forelimb kinematics change with the compactness of the substrate?

Here, we tested the hypothesis that moles burrow using different forelimb kinematics in loose and compact substrates. Based on previous studies of anatomy (Hildebrand, 1982, 1985; Reed, 1951; Rose et al., 2013; Yalden, 1966) and kinematics measured using cinefluorography (Gambaryan et al., 2002), we predicted that moles burrowing in loose substrates raise mounds of soil by performing compacting power-strokes mainly involving long-axis rotation of the humerus (i.e. pronation/supination). In compact soil, we predicted that moles shear soil by performing scratching power-strokes involving amplified elbow extension, similar to most scratching diggers (Hildebrand, 1985; Kley and Kearney, 2007; Milne et al., 2009; Olson et al., 2016; Vickaryous and Olson, 2007). We also predicted that in both types of substrates, moles displace soil rearward like other mammalian diggers.

Eastern moles, Scalopus aquaticus (Linnaeus 1758), are ideal for studies of forelimb kinematics as they are one of the most derived fossorial moles (Campbell, 1939; Sánchez-Villagra et al., 2006). They are also relatively common and can be kept in captivity. Three eastern moles (body mass 94.7±10.2 g) were captured in Hadley, MA, USA, and housed in captivity. All animals were transported to Brown University and anesthetized for surgery prior to X-ray videographic recordings (see below). Subjects were induced on 4% isoflurane in a chamber with 2 l min⁻¹ O₂ flow, transferred to a mask, and maintained on 1.5–2.5% isoflurane with 0.8 l min⁻¹ O₂ flow. All husbandry and experimental procedures were approved by the Institutional Animal Care and Use Committee of UMass Amherst (#2013-0023) and Brown University (#1409000093). For each sterile surgery, spherical tantalum markers (0.5–1.0 mm diameter) were implanted following the methodology detailed by Konow et al. (2015) (Fig. 3A). One marker was implanted in the proximal end of the right scapula, three to four in the humerus, and one in both the proximal and distal ends of the ulna. Three subcutaneous markers (0.5 mm) were implanted in the medial, distal and lateral side of the manus. These locations were deep to the false thumb (os falciforme), at the base of the third and fifth digits. Six subcutaneous markers (0.8 mm) were implanted along the dorsal and ventral midline to help define a reference body plane (Camp et al., 2015). After surgery, moles were allowed 3 days to recover and normal feeding and burrowing behavior were observed to resume within this time for all subjects, with no immediately discernible changes in patterns of movement and activity.

We allowed each subject to burrow in a 20 cm high×10 cm wide×50 cm long radio-translucent polycarbonate enclosure filled with couscous (Osem^®, original plain). Couscous was selected as the experimental substrate because of the even size and radio-opacity of the granules. Moles tend to occupy cohesive substrates that do not flow or collapse when they burrow through it. Therefore, we mixed couscous with water in a 2:1 volume ratio to increase the cohesion between couscous particles prior to each trial (i.e. cohesive couscous). We then created two substrate types for the experiments. The first mimicked natural loose substrates that moles burrow in (Lin et al., 2017) and was made by filling the enclosure with free-falling cohesive couscous to a depth of 10 cm (mean±s.e.m. compactness 0.37±0.03 kg cm⁻², n=10; measured with a Humboldt^® soil penetrometer). In soil with this compactness, which is typically located near the surface of the ground, moles raise soil up and compress it to the wall to open a tunnel (Lin et al., 2017). The second experimental substrate mimicked compact soil (Lin et al., 2017) and was made by filling the enclosure with free-falling cohesive couscous to a depth of 20 cm and then compressing it to 10 cm (mean±s.e.m. compactness 2.58±0.03 kg cm⁻², n=10). In soil of this compactness, which is typically found deeper underground, moles consistently excavate and transport substrate out of their tunnels (Lin et al., 2017). In both experiments, the enclosure was covered with a radio-translucent polycarbonate lid to ensure the substrate compactness remained constant and to avoid substrate being pushed out of the enclosure as a result of burrowing activity. The bottom of the enclosure was marked by radio-opaque markers that were used to reference the horizontal, ‘ground level’ plane for data analysis.

During the course of each burrowing trial, we recorded biplanar X-ray videos. For each lateral and dorsoventral view, X-ray images were generated by two X-ray machines (Radiological Imaging Systems) and captured by high-speed cameras (Phantom, V10, Vision Research) recording at 90 Hz with a shutter speed of 1/250 s. X-ray techniques were 80 kVp/160 mA for the dorsoventral view (left panel in Fig. 3A) and 80 kVp/200 mA for the lateral view (right panel in Fig. 3A). A calibration grid and cube were used at the beginning, middle and end of the experiments to calibrate the distortion of X-ray images using the software XMALab (v.1.3.3) (Knörlein et al., 2016). After completion of the trials, animals were euthanized by injection of sodium pentabarbitol (i.c.) for computed tomography (CT) scans. CT scans were used to build 3D mesh models of forelimbs and implanted markers, using Mimics (v.16.0, 64-bit) and Geomagic Studio (v.12).

We refer to the forelimb movement performed by moles in loose substrates as a ‘compacting stroke’ and that used in compact substrates as a ‘scratching stroke’. We chose three representative cycles of compacting strokes from each of two individuals and three representative cycles of scratching strokes from each of three individuals for analysis. The number of individuals and trials was constrained by the availability of animals, whether individuals used their marker-implanted right hand to move substrates, and the frequency with which complete stroke cycles were captured within the calibrated field of view. For each dataset, we used the marker-based XROMM workflow (Brainerd et al., 2010; Gatesy et al., 2010) to construct a model and calculate forelimb kinematics.

We calibrated distorted X-ray images and digitalized markers using XMALab 1.3.3. For bones implanted with at least three markers (humerus and manus), we combined marker coordinates from X-ray videos with marker coordinates from 3D bone models to calculate rigid-body transformations (Brainerd et al., 2010). For bones with fewer than three markers (scapula and ulna), we used Scientific Rotoscoping to align 3D mesh bone models to match their positions in both X-ray images (Gatesy et al., 2010).

We present two types of kinematic analysis in this study: (1) visualizations of the overall movement of the stroke within a world-space (Fig. 3B) and (2) descriptions of kinematics at the shoulder, elbow and carpal joints within a series of joint coordinate systems (JCSs; Fig. 3C). First, to visualize the overall movement of each burrowing stroke, we created a virtual marker (black sphere in Fig. 3B) in Maya^® 3D software and attached this marker digitally to the tip of the 3D model of the right third keratinous claw (digit III). Movements of the marker were thus linked with movements of the manus, as calculated from rigid-body transformations. As we were interested in how burrowing strokes move substrates in 3D space, we created a world coordinate system (WCS) with the y-axis set to align with the ground (horizontal) and the z-axis set to vertical. To account for the heading of the animal (body yawing), we aligned the x-axis of the WCS with the orientation of the sternum (from dorsal view; camera 1, left panel in Fig. 3A). Unlike body yawing, the effects of body pitching and rolling on stroke trajectories were not quantified in this analysis as we could not discriminate the pitch and roll of the sternum from both cameras or based on the subcutaneous markers we had. To understand how burrowing strokes move relative to their starting points, we set the center of the coordinate system (0,0,0) as the fixed location of the virtual point at the beginning of each stroke. We then tracked the movements of the virtual marker in this WCS over time. The displacements of the virtual claw marker along the x-, y-, z-axes were then analyzed and represented as the movements of the stroke as forward (x+)/rearward (x−), outward (y+)/inward (y−) from the midline and upward (z+)/downward (z−).

Second, each JCS described the movements of a distal bone relative to its proximal neighboring bone (Fig. 3C). For the shoulder and carpal joints, the coordinate system origin (0,0,0) was defined as the middle point of a line connecting the lateral and medial sides of the humeral and ulnar heads, assuming a hinge joint (Reed, 1951). For the elbow joint, a sphere was placed in the trochlear notch to approximate the center of rotation. After determination of the origin (0,0,0), we aligned orthogonal axes with the anatomical axes of the distal bone, by first determining the long axis of the bone [z-axis; supination (+)/pronation (−)], and then orienting the craniocaudal [y-axis; protraction (extension) (+)/retraction (flexion) (−)] and dorsoventral axis [x-axis; abduction (+)/adduction (−) for shoulder and carpal joints; medial (+)/lateral (−) rotation for elbow joint] (Pierce et al., 2012). For the shoulder, elbow and carpal joints, the craniocaudal axis was parallel to the humeral plane, perpendicular to the ulnar plane and parallel to the plane of the manus, respectively (green y-axis, Fig. 3C). After the origin and axes of each JCS were determined, we defined a reference angle for each JCS at each joint to describe joint movements relative to a neutral position. For the shoulder, when the humeral plane is perpendicular to the long axis of the scapula and its long axis is parallel to the glenoid fossa, the xyz angle of the humerus relative to the scapula is (0 deg, 90 deg, 90 deg), respectively. For the elbow, when the long axis of ulna is perpendicular to the humeral plane and the craniocaudal axis of the ulna is parallel to the trochlear fossa, the xyz angle of ulna relative to the humerus is (0 deg, 90 deg, 90 deg), respectively. For the carpal joint, when the plane of the manus is parallel to the plane of the ulnar head and its craniocaudal axis is parallel to the line connecting the two protruding ends of the ulnar head, the xyz angle of the manus relative to the ulna is (0 deg, 0 deg, 0 deg), respectively (Fig. 3C).

We used rigid-body transformations to define the movement of the humerus. For movements of the elongate scapula, we used one marker implanted on the proximal end and the glenoid fossa of the scapula as a bony marker to define the long-axis direction of the elongate scapula. This allowed us to analyze the humeral long-axis rotation (supination/pronation) and humeral protraction/retraction relative to the scapula at the glenohumeral (shoulder) joint. We do not present data on humeral rotation along the long axis of the scapula (i.e. humeral abduction/adduction) at the shoulder joint because we were unable to accurately define the long-axis rotation of the scapula itself based solely on the marker and landmark. For the ulna, we used markers implanted on its proximal and distal ends to define its long axis. This allowed us to analyze the ulnar protraction/retraction (i.e. elbow extension/flexion) and ulnar medial/lateral rotation relative to the humerus at the humeroulnar (elbow) joint. We do not present the results of ulnar long-axis rotation because it was minor and so could be overshadowed by error associated with manual rotoscoping. We used rigid-body transformations to define the movements of the manus and analyzed its rotations in three axes relative to the ulna at the ulnocarpal (carpal) joint.

We analyzed the displacements of the stroke and angular joint motions. Data points over one full stroke cycle were normalized to 101 points (corresponding to 0–100% of a stroke cycle) using a cubic spline interpolation (O'Neill et al., 2015). This allowed us to calculate the mean and standard error of the kinematic curves for each individual in each substrate. The beginning of forelimb retraction was defined as the moment at which the forelimbs were closest to each other in front of an animal's head. The beginning of forelimb protraction was defined as the moment at which the forelimbs of the animal were most distant from each other at the sides of body. We then calculated the durations of retraction and protraction within each cycle. We defined stroke velocity along the claw-tip trajectory in 3D space (cm s⁻¹), and stroke frequency as the inverse of stroke duration.

Biplanar X-ray kinematics measurements on the trajectory of the claw-tip revealed that Eastern moles moved their whole limbs in very different ways during compacting and scratching strokes (Fig. 4). During the retraction phase of a compacting stroke, the claw-tip moved primarily along the upward–downward (z-) axis in two individuals (Fig. 4C), indicating that the whole limb moved vertically from the bottom to the top (Fig. 4D,E). One individual also moved its claw outward from the body midline [along the inward–outward (y-) axis; mole 2 in Fig. 4B]. In contrast, during the retraction phase of a scratching stroke, the claw-tip moved primarily outward from the body midline (y-axis, Fig. 4B,D) and, to a lesser extent, rearward (forward–rearward x-axis, Fig. 4A,E). Compacting strokes were executed at lower velocity (mean±s.e.m. 13.83±1.16 cm s⁻¹) and frequency (1.91±0.14 Hz) compared with scratching strokes (21.21±1.22 cm s⁻¹, 3.39±0.17 Hz) (Mann–Whitney U-test, P=0.0004; P=0.002) (Table 1).

Kinematics analyses of joint movements revealed that substantial humeral long-axis pronation and retraction are the primary movements at the shoulder joint during both compacting and scratching strokes (Fig. 5A). There was a similar degree of humeral retraction at the shoulder between compacting and scratching strokes (Mann–Whitney test, P=1.00) but the range of pronation was greater during scratching than during compacting strokes (Mann–Whitney test, P=0.02) (Table 2). In addition, during scratching strokes, humeral pronation and retraction continued until the end of stroke retraction. In contrast, during compacting strokes, both rotations stopped earlier and the plane of the humerus was kept nearly perpendicular to the long axis of the scapula (Fig. 5A; the angle along the blue z-axis remains at ∼90 deg; this scapula–humerus position is illustrated in Fig. 3C in which the z-axis is at 90 deg) until the end of retraction.

In both compacting and scratching strokes, ulnar protraction (i.e. elbow extension, where the ulna rotated positively along the green y-axis in Figs 3C and 5B) started at the beginning of stroke retraction, followed by ulnar lateral rotation (i.e. ulna rotated negatively along the red x-axis in Figs 3C and 5B). However, ulnar movement with respect to the humerus at the elbow joint during the two strokes differed at the end of forelimb retraction. During scratching strokes, the ulna was retracted (i.e. elbow flexed) but continued to be rotated laterally until the end of stroke retraction. In contrast, during compacting strokes, the ulna was not retracted (i.e. the elbow was kept extended) by the performance of a minor secondary ulnar protraction (i.e. elbow extension) before the end of the stroke retraction phase (Fig. 5B).

Compacting and scratching strokes showed opposite patterns of movement at the carpal joint Fig. 5C). During scratching strokes, the manus was retracted and subsequently adducted and pronated until the end of stroke retraction. In contrast, adduction and pronation of the manus happened first during compacting strokes and were followed by retraction of the manus (carpal joint flexion). In both compacting and scratching strokes, the timing of manus retraction coincided with the point at which moles encountered challenges such as compressing soil to the wall or breaking compact soil. In contrast, the timing of manus adduction and pronation was synchronized with when moles moved loose or loosened substrates.

The results of our study support the hypothesis that eastern moles perform different joint movements while digging in loose and compact substrates. However, the nature of these differences was unexpected in three ways. First, there were clear differences in the movements of the whole limb between compacting and scratching strokes. These differences in limb trajectory were associated with different patterns of distal joint movement, whilst movements at the shoulder joint, where muscles presumably generate the greatest digging forces, were comparable in loose and compact substrates. In loose substrates, moles used compacting strokes to move the substrate upward. The power-stroke of this movement began with humeral pronation, humeral retraction and elbow extension, followed by a secondary elbow extension and flexion at the carpal joint at the end of stroke retraction. These secondary movements served to compress the substrate to open a tunnel (Movie 1). During scratching power-strokes, which were used to advance a tunnel in compact substrates, movements started with the same joint motions as compacting strokes, and were followed by ulnar lateral rotation, and manus adduction and pronation, which served to sweep substrates laterally before they were transported to the surface (Movie 2). Second, our results reveal that, contrary to historical speculations in the literature, amplified elbow extension is not the prime movement performed by moles in either loose or compact substrates. Third, we found that forelimb joint trajectories during burrowing strokes are unlike those of any other mammalian diggers; the power-stroke itself does not displace substrates directly rearward, but instead moves them upward or outward from the body midline.

Compared with other mammalian forelimb diggers, eastern moles exhibit unique stroke trajectories and mechanisms for soil displacement. Most fossorial mammals use scratch-digging or hook-and-pull digging primarily in a parasagittal plane to move substrate caudally during digging (Campbell, 1939; Dorgan et al., 2011; Hildebrand, 1985; Kley and Kearney, 2007; Moore et al., 2013; Reed, 1951). By contrast, evolution of the eastern mole forelimb has involved a cranially directed repositioning of the forelimbs, as well as a laterally facing manus to move substrates perpendicular to the body midline during forelimb retraction. It is this unique movement that allows moles to construct tunnels efficiently near the surface, and in relatively loose soil (Lin et al., 2017) where prey (e.g. earthworms) are particularly abundant (Edwards and Bohlen, 1996). When burrowing in this type of substrate, moles use a single compacting stroke to move soil upward and compress it into the side of the tunnel. At the same time, moles raise and rotate the front half of their body, supported by the other, non-digging forelimb, to further help direct soil upward (Arlton, 1936; Hisaw, 1923; Skoczeń, 1958). The combination of this efficient single stroke and body movements allows moles to move quickly through loose soil and construct tunnels when foraging in loose soil near the surface. However, when burrowing in deep, compact substrates that are more difficult to displace, moles advance their tunnel by scratching the soil and then transporting it out of the tunnels (Lin et al., 2017). To move soil, moles yaw their trunk to one side during a stroke in order to direct soil more directly caudally or use their reduced hindlimbs to assist in kicking soil caudally (Arlton, 1936; Hisaw, 1923; Skoczeń, 1958). We observed that moles also yaw their bodies when transporting soil backward to the surface. Although scratching strokes require these additional movements to remove soil, moles complete a scratching stroke much faster than a compacting stroke. High scratching stroke velocity and frequency may be compensating for time spent executing additional movements required to transport soil to the surface and explain the similar rate of tunnel construction in loose and compact soil (Lin et al., 2017).

The pronounced body rotations (yaw, pitch and roll), as well as the disassociation between movements of the cranial and caudal part of their body imposed limitations on our analyses. The pronounced body trunk rotations can affect the direction of the power-stroke and might explain the observed differences in stroke trajectories between individuals. For example, when moles burrow in loose substrates, rolling the trunk would lead to greater stroke excursion in the upward direction rather than outward from the body midline (mole 2, left panel in Fig. 4B). We were unable to use the array of subcutaneous markers we implanted to generate body reference planes. Therefore, we were unable to reconstruct the motion of the scapula as well as that of the entire limb, relative to the trunk. The development of alternative methods to create an anatomical reference plane would be a good subject for further study.

We discovered similarities in kinematics at the shoulder joint for compacting and scratching strokes, but substantial differences in elbow and carpal joint kinematics. These discoveries contribute to our understanding of how moles generate high forces during digging. Most of the force generated during digging has traditionally been attributed to the m. teres major, which acts at the shoulder joint and accounts for 75% of total forelimb muscle volume (Gambaryan et al., 2002; Rose et al., 2013). Contractions of m. teres major, together with mm. latissimus dorsi, subscapularis and pectoralis superficialis posticus, have previously been hypothesized to result in humeral pronation and retraction at the shoulder joint (Gambaryan et al., 2002; Hildebrand, 1982, 1985; Reed, 1951; Rose et al., 2013; Yalden, 1966). We observed these two elements of humeral movement at the initiation of both compacting and scratching strokes, thus hinting at the centrality of shoulder movements in both loose and compact substrates.

We also discovered that the long axis of the humerus remains nearly perpendicular to the long axis of the scapula as moles use compacting strokes to compress the substrate at the end of limb retraction. Prior studies have speculated that the humerus ends up parallel to the parasagittal plane at the end of retraction (e.g. Reed, 1951; Yalden, 1966). However, our findings combined with an earlier morphological study (Gambaryan et al., 2002) do not support those speculations and instead suggest that moles may be efficient burrowers because the perpendicular position of the humerus plane (relative to the long-axis of the scapula) provides the greatest possible moment arm for m. teres major to generate torque at the shoulder.

In contrast to the stereotyped movements at the shoulder, elbow movements were observed to differ between compacting and scratching strokes, particularly at the late stage of forelimb retraction. At the end of the retraction phase of the compacting stroke, the ulna resists retraction (i.e. the elbow resists flexion), which aids in compressing the substrates into the wall of the tunnel. This motion reinforces the tunnel wall whilst only involving a single stroke. By contrast, the ulna is continuously rotated laterally during scratching strokes to sweep soil in both lateral and caudal directions. This laterally rotated, sweeping motion may be reinforced by the expanded humeral trochlea and the enlarged humeral epicondyles, which prevent elbow dislocation (Gambaryan et al., 2002) during lateral rotation of the ulna. We speculate that this unusual ulnar rotation is achieved by using the olecranon fossa as a pivot that allows the ulna to glide around the expanded humeral trochlea.

Moles do not simply use amplified extension of the elbow for scratching or compacting strokes. This result was unexpected and might be explained by comparing the forelimb morphology of eastern moles with that of other scratching diggers (Fig. 1). A relatively long olecranon (the muscle moment arm) compared with the length of the radius (the output lever) has been suggested as characteristic of scratch diggers (Hildebrand, 1985; Kley and Kearney, 2007; Milne et al., 2009). This trait has been suggested to provide a large area of muscle insertion as well as increased mechanical advantage for the elbow extensors (e.g. m. triceps brachii, m. anconeus), leading to amplified elbow extension for soil shearing. In eastern moles, however, the ratio between olecranon length and radius length is much lower than in burrowing mammals like Spalax (mole rat) and Myospalax (zokor) (Gambaryan et al., 2002). Moreover, the vigorous elbow extensions seen in scratching diggers are less exaggerated in moles. These two facts suggest the possibility that the function of the extensor muscles in moles is to resist flexion at the elbow joint (Gambaryan et al., 2002; Rose et al., 2013). Anatomical measurements combined with our kinematic results confirm that amplified elbow extension is not the primary movement performed by moles during powerful digging. Instead, moles incorporate lateral rotation of the ulna (in compact substrates) and resist elbow flexion (in loose substrate), both of which would tend to avoid increases in the moment arm of the substrate reaction force at the elbow joint as the elbow extends (Gambaryan et al., 2002).

At the carpal joint, moles also perform different movements when moving loose versus compact substrates. When compressing or dislodging substrate, moles retract the manus (carpal joint flexion), but when moving loose or loosened substrates, the manus is adducted and pronated. Manus movements in eastern moles are primarily controlled by the m. flexor digitorum profundus, which is almost entirely tendinous (Rose et al., 2013). Given that this tendinous muscle originates at the medial epicondyle of the humerus, humeral pronation is proposed to place tension on the tendon to result in manus retraction (Reed, 1951; Yalden, 1966). Previous studies have speculated that carpal morphology limits carpal joint motion to retraction and protraction of the manus (Reed, 1951; Yalden, 1966). However, our results show that the manus adducts and pronates while removing relatively loose soil during the beginning of retraction in compacting strokes and at the end of retraction in scratching strokes, after the substrate has been sheared (Fig. 5C). Thus, we hypothesize that manus retraction assists the compression or dislodgement of soil by imposing a force perpendicular to the substrate surface, whereas manus adduction and pronation facilitate sweeping the substrate by covering more ground area in one stroke. As carpal joint motion is primarily controlled by the force transmitted from humeral pronation, via the tendinous m. flexor digitorum profundus, we speculate that changes in the carpal joint motions of eastern moles are largely passive responses (as opposed to being actuated by carpal joint muscles, which remain present, albeit not very well developed; Rose et al., 2013) to changes in substrate compactness. Passive engagement of the manus could be an energy-saving mechanism for eastern moles digging in both loose and compact substrates.

Researchers have long speculated about joint motion and burrowing mechanisms in fossorial tetrapods based on the morphological specializations of their forelimbs (Archer et al., 2011; Chen and Wilson, 2015; Hopkins and Davis, 2009; Kley and Kearney, 2007; Moore et al., 2013; Olson et al., 2016; Pritchard et al., 2016; Reed, 1951; Rose et al., 2013; Rupert et al., 2015; Stein, 2000; Woodman and Gaffney, 2014; Yalden, 1966). More recently, burrowing robots have been designed largely based on these speculations (Richardson et al., 2011; Scott and Richardson, 2005, 2006). Although frequently invoked, the structure–function relationship between morphological specializations and associated joint movements during mole burrowing has rarely been tested directly. Here, we present the first evidence that the expanded humeral trochlea and enlarged humeral epicondyles permit ulnar lateral rotation when moles move excavated soil laterally. In addition, a relatively low ratio of olecranon to radius length in moles, compared with that in other scratching diggers, is associated with moderate elbow extension that moves soil outward from the midline rather than rearward. These kinematics patterns contradict earlier speculations (Yalden, 1966) and may have affected the performance of burrowing robots that were built based on them (Richardson et al., 2011; Scott and Richardson, 2005, 2006). We suggest that pushing outward from the pectoral region might help moles (and mole-inspired robots) exert the highest possible forces (Hisaw, 1923). It may also provide the body with a better bilateral balance to prevent tumbling when encountering obstacles. Moles use other movements to displace soil backward, including body rotations. Our results imply that what makes moles powerful diggers is the combination of a powerful motor (the enlarged m. teres major driving stereotypical movements of shoulder) and an elastic spring (the tendinous m. flexor digitorum profundus) that transmits force distally along the burrowing appendage. This system provides a robust and potentially energy-saving solution that permits kinematic flexibility when burrowing appendages encounter complex substrates. Our findings likely have implications for the bio-inspired design of burrowing robots that can be used to navigate uncertain and dynamic terrain during urban search and rescue missions.

We thank Thomas Roberts for assistance with surgeries, Angela Horner for discussion of the experimental design, Elizabeth Brainerd, David Baier, Ariel Camp and Peter Falkingham for XROMM support, and Paul Spurlock, Joanne Huyler and Animal Care Services of UMass Amherst and Brown University for animal husbandry. We also thank Robert Kambic and two anonymous reviewers for their constructive suggestions to improve the final manuscript.