Keep calm and hang on: EMG activation in the forelimb musculature of three-toed sloths (Bradypus variegatus)

Tree sloths (Xenarthra, Pilosa) are one of only a few mammals for which suspensorial habits are almost obligatory (Hayssen, 2009, 2010). Sloths also have lower basal metabolism and mean body temperature compared with other xenarthran taxa, and placental mammals in general (Gardner, 2008). Specifically, three-toed sloths (Genus: Bradypus) are high-canopy folivores (Montgomery, 1983; Cliffe et al., 2015) and have the lowest field metabolic rate observed among non-hibernating mammals (Pauli et al., 2016). The brown-throated three-toed sloth (Bradypus variegatus) is the most accessible (non-endangered) species of this genus and it is a specialized folivore that subsists on a low food intake of exclusively leaves (Montgomery, 1983). Bradypus is diurnal but spends 15–18 hours a day sleeping (Sunquist and Montgomery, 1973; Janzen, 1983) and during this time, undergoes routine cycles of behavioral sleep (Barrat, 1965). The majority of the active, awake time in B. variegatus is filled by grooming or feeding while reclined in the recesses of tree branches; thereby, it does not engage in suspensory (below branch) behaviors as often as two-toed sloths (Urbani and Bosque, 2007). The collective sleep/wake patterns demonstrate that sloths are not particularly active mammals, supporting the hypothesis that their physiology, as well as their behavior, reflects a system extremely modified to conserve metabolic energy (Cliffe et al., 2015, 2018; Spainhower et al., 2018).

Sloths exhibit arboreal locomotor behaviors that include suspensory walking and vertical climbing, as well as suspensory postures (i.e. hanging) by some combination of limbs for extended periods of time (Hayssen, 2009, 2010). Despite these abilities that necessitate both muscular strength and fatigue resistance, the muscle mass of sloths is greatly reduced (Grand, 1978), which further necessitates modifications to their distal skeletal elements (Mendel, 1981b, 1985) and flexor muscle architectural properties (Olson et al., 2018), as well as the predominant expression of slow myosin heavy chain (MHC) fibers that atypically have anaerobic metabolic characteristics (Spainhower et al., 2018) to power contraction. Our previous work specifically reported that the extrinsic muscles of the forelimb in Bradypus have parallel fiber architecture and several have long moment arms, and with the exception of mainly the distal carpal/digital flexors and extensors, the majority of the intrinsic musculature is slow-contracting with relatively long, parallel or unipennate fibers (Olson et al., 2018). For example, the shoulder flexors m. pectoralis superficialis and m. deltoideus are ∼70% slow in their MHC fiber type distribution as are the elbow flexors/extensors, although large m. brachioradialis is the strongest elbow flexor and has a 95% expression of the slow MHC-1 isoform in B. variegatus. Conversely, the carpal/digital flexors have a larger percentage of fast MHC-2A fibers (Spainhower et al., 2018). In addition to generally strong, slow-contracting muscles, the distinctively slow and controlled movement patterns of sloths are considered to be an adaptive strategy for energy conservation (McDonald and De Iuliis, 2008; Cliffe et al., 2018), and this may be understood as a means of enhancing muscle force while limiting contractile velocity, thus lowering the ATP cost of contraction (Taylor et al., 1982; Heglund and Taylor, 1988; Roberts et al., 1997). However, in vivo patterns of muscle activation are needed to verify the muscle fiber type properties that we found previously in relation to locomotor performance in sloths.

The expectations for muscle function in a tensile limb system are formed by foundational fiber typing (Jouffroy and Medina, 2004) and telemetered electromyography (EMG) studies (Jouffroy and Stern, 1990; Jungers and Stern, 1980, 1981) in suspensorial primates. A near homogeneous expression of large, slow MHC-1 fibers is typical in a single belly (or head) of each major extensor functional group in primate limbs. These muscles show a modest level of EMG activation during arboreal postural behaviors, which correlates with an anti-gravity role in joint stabilization (Jouffroy et al., 1999). Sickles and Pinkstaff (1981a,b) notably observed a near uniform distribution of slow-contracting fiber types in both flexor and extensor muscles of the slow loris (Nycticebus spp.) corresponding to EMG activation patterns of selective flexors versus extensors during suspension (Jouffory and Stern, 1990), thus suggesting the possibility of shared body weight support between forelimb and hindlimb muscles. The absence of fast isoform MHC-2X and MHC-2B fibers in lorisids and sloths (Spainhower et al., 2018) further suggests specialization of their limb muscles for joint stabilization during their slow, deliberate suspensory habits, but this possibility has yet to be reported by companion EMG analyses from sloth limb muscles.

In this study, we aimed to evaluate EMG from the forelimb musculature of B. variegatus to determine the patterns and levels of muscle activation, in addition to the EMG spectral properties associated with suspensory habits versus those of metabolically costly vertical climbing (Hanna et al., 2008). Muscle activation patterns determined using EMG record electrical signals from action potentials as the motor units within muscles were recruited. Action potentials elicited from different motor units that comprise slow and fast fiber types leave characteristic signatures in the EMG frequency spectrum that can be detected using wavelet decomposition of EMG (Hodson-Tole and Wakeling, 2007; Blake and Wakeling, 2014). By use of wavelet analysis, the relationship between MHC fiber type expression and muscle activation in sloth limbs can be more critically assessed, and may offer insight into suspensory muscle function beyond that typical of numerous previous reports in primates.

We hypothesized that the neuromuscular properties of sloth muscles reflect a physiological system specialized to conserve metabolic energy via minimizing activation of flexor muscle mass for suspensory function. More specifically, it was expected that the slow fibers of the shoulder, elbow and digital flexors would be active for large portions of limb contact with the substrate. We also predicted that maximum EMG intensity would occur during vertical climbing (via greater recruitment of fast fibers), and would be intermediate during suspensory walking and minimal during suspensory posture. It is not known if the biomechanics of sloths resemble the forelimb-biased propulsion during vertical climbing in slow-moving lorisids (Hanna et al., 2017), but it is possible that sloth forelimbs have an even greater role in propulsion while moving vertically than they do when walking below-branch (Granatosky and Schmitt, 2017). The results of this study improve the understanding of muscle fiber recruitment patterns involved with suspensory habits and arboreal lifestyle and help elucidate the influence of the neuromuscular system on motor unit properties in sloths. The outcomes also help clarify mechanical design requirements for limbs that evolved to experience primarily tensile versus compressive loading.

Six three-toed sloths (Bradypus variegatus Schinz 1825; 3.9±0.6 kg) were used for this study at The Sloth Sanctuary in Penshurst-Limon, Costa Rica. No preference was given to the sex of the animals, although adult (or sub-adult) individuals were preferable to juveniles. Morphometric data for individuals are presented in Table S1. All experimental procedures complied with protocols approved by the Costa Rica Ministerio de Ambiente, Energia, y Telecomunicaciones (MINAE: R-008-2017 to M.B.) and adhered to the legal requirements of the USA.

Sloths were sedated (0.1 ml kg⁻¹ dexdomitor and 0.1 ml kg⁻¹ ketamine, injected into the m. gluteus medius) prior to electrode implantation. Supplemental doses of ketamine in increments of 0.1 ml were used as necessary to achieve the desired level of sedation. While under sedation, custom fine-wire, bipolar EMG electrodes (0.002 bifilar: California Fine Wire, Grover Beach, CA, USA) were implanted into selected muscles of the shoulder, brachium, and antebrachium. Specifically, electrodes (sterilized by hydrogen peroxide plasma) were implanted into either of two suites of four muscles in the left forelimb: suite 1 (proximal musculature): m. pectoralis superficialis (PS), m. deltoideus (DELT), m. biceps brachii (BB) and m. triceps brachii-long head (TBLO); suite 2 (distal musculature): m. brachioradialis (BCR), m. palmaris longus (PL), m. flexor digitorum profundus (FDP) and m. extensor digitorum communis (EDC). The electrodes for the PS were placed in the head approximating the m. pectoralis superficialis anterior (PSA) that originates from the manubrium/sternum and inserts on the pectoral ridge (line) of the humerus. The three heads of the DELT are inseparable in three-toed sloths and we placed the electrodes about the dorsolateral aspect of the shoulder joint approximating the merged bellies of the m. deltoideus pars scapularis/acromialis. See Olson et al. (2018) for a review of forelimb myology in B. variegatus.

Pairs of 4 electrodes were all implanted percutaneously into the center of each muscle belly within a suite by using gentle force with a 26 gauge needle to penetrate the skin, then a one-half twist, and removal of the needle that left the hooked electrode end anchored in the muscle belly. Target muscles were identified by palpation of the limb while sloths were sedated; prior to experimentation, muscle locations were verified by gross dissection of three cadaver sloth forelimbs. EMG wires were cabled, secured, and connected to a wireless EMG unit (BioRadio™, Great Lakes NeuroTechnologies, Cleveland, OH, USA) that was harnessed to the thorax of the sloth. Sedation was reversed with injection of Antisedan (0.1 ml kg⁻¹) into the m. gluteus medius.

Upon recovery from sedation, the limbs of sloths were placed in position to grasp a horizontal beam (4.57×0.22×1.52 m) constructed from bamboo. The beam was wrapped in flexible veterinary bandaging tape to provide extra grip. Sloths were required to both hang in a static suspensory posture and walk suspended below the beam. Several trials of suspensory hanging (SH) by forelimbs and hindlimbs were performed first in intervals of 1–3 min. Locomotor trials consisted of repeated short bouts of suspensory walking (SW), long enough to sample 2–3 footfalls of the instrumented forelimb (Fig. 1A). Sloths also performed short bouts of vertical climbing (VC) where individuals were allowed to freely climb the outside of a fenced enclosure (2.00×3.97×3.12 m) following walking trials (Fig. 1B). For each locomotor behavior, sloths were encouraged to move with the enticement of food (e.g. leaves or flowers) as needed, but were free to select their preferred speed.

All trials performed were videoed with a GoPro camera (HERO4: San Mateo, CA, USA) positioned 1.5 m from the animal; in the sagittal plane for the suspensory trials and in a coronal plane for the VC trials. This distance allowed nearly the entire length of the beam or height of the enclosure to be visualized. Video data (100 Hz) were synchronized with EMG (2000 Hz) with an LED light in the field of view of the camera. EMG was recorded via broadband streaming using BioCapture^™ software (Great Lakes NeuroTechnologies) for one suite of four muscles (proximal versus distal musculature) then the other on consecutive days for the same individual. At the completion of EMG recordings (∼2 h) the fidelity of each implant was verified before the fine wire electrodes were removed. The animals were then placed back in their enclosure and observed for 1–2 h to ensure full recovery.

Video data were first analyzed for frame numbers that corresponded to time (in s) in the EMG records marking the beginning and end of SH intervals, or to foot-on (grasp) and foot-off (release) events during locomotor trials. Footfall (i.e. each grasp) was defined as the first frame that we could determine when the claw of each digit was in full contact with the substrate and the limb showed evidence of weight-bearing support. EMG for SW and VC were analyzed as a series of strides consisting of consecutive pairs of limb contact and swing phases. Velocity (m s⁻¹) for SW and VC was calculated across a calibrated distance (15 cm) in the frame of view (Tracker software; https://physlets.org/tracker/): instantaneous velocity was determined at 1 s intervals throughout a stride and averaged across data points where velocity was similar. Except for velocity, temporal variables were calculated (LabScribe software: iWorx Systems Inc., Dover, NH, USA) for both the limb cycle (in s) and relative stride interval (% stride), where the sum of the contact and swing phases is 100% of the stride, and include duty factor, EMG onset, EMG offset and EMG burst duration. All temporal measurements were averaged for each muscle and individual.

Because both the magnitude of the EMG and the spectral properties that occurred in each EMG burst are the parameters of interest, this requires a time–frequency analysis that is not possible from more traditional approaches such as the root-mean-square (RMS) EMG (Jouffroy and Stern, 1990; Jouffroy et al., 1999). We used a wavelet technique that has been specifically adapted for EMG analysis (von Tscharner, 2000) and extensively tested across different species (e.g. Wakeling and Syme, 2002; Wakeling et al., 2002, 2012). The wavelet analysis determined an EMG intensity that was a close approximation of the power in the EMG signal. A priori, only bursts that were at least 2 times greater than baseline EMG voltage were considered for analysis. EMG intensities via wavelets were determined for a subset of 2–3 bursts per muscle for each individual across numerous strides (for locomotor trials) using custom software in Mathematica (v. 11.3; Wolfram Research, Champaign, IL, USA), with this intensity being localized across both time and frequency. Briefly, the normalized EMG was decomposed by a bank of 17 wavelets (range of central frequencies: 6.9–902 Hz), and EMG intensity spectra were generated for 0.5–1 s windows within each burst of activity (about 5 spectra per burst).

The EMG intensity was normalized to the mean of the peak EMG intensities across all individual trials and behaviors. The EMG intensity spectra for all muscles and functional behaviors across individuals were classified by principal components analysis (PCA) (Wakeling, 2004; Hodson-Tole and Wakeling, 2007): the EMG intensity was represented by the PC1 loading score for each window while the PC2 loading score represented activation frequency, and the shape of the EMG intensity spectra were given by the angle theta (θ) of the PC1–PC2 loading score eigenvector. A large θ characterized EMG intensity spectra with lower frequency components than spectra with small θ.

Values are reported as means±s.d. unless otherwise specified. MANOVA performed in SPSS (v. 20: IBM, Inc., Armonk, NY, USA) was used to test for mean differences in stride duration, stride frequency (inverse of stride duration), duty factor and velocity between SW and VC. Descriptive statistics for all temporal and EMG intensity measurements for each muscle were pooled and averaged across individuals to account for variance associated with the general patterns of activation timing and EMG maximum activation. Coefficients of variation (CV) were calculated for each EMG variable analyzed to justify use of mean data in additional multivariate testing. MANOVA was also employed to assess statistical significance in both EMG burst duration (% stride) and normalized EMG intensity, specifically how these parameters varied among flexor muscles (effect: muscle) and SH, SW and VC (effect: behavior). In cases where the assumption of homoscedasticity was not met (e.g. EMG burst durations) a Kruskal–Wallis test was performed, and then significant results from either this test or those from MANOVA were followed by a series of one-way ANOVAs and Dunnett’s T3 (burst duration) or Tukey's (burst activation) post hoc tests to determine all pair-wise significant outcomes. Significance for all statistical tests was accepted at P≤0.05.

The exact placement of EMG electrode implantation could not be verified in the present study. We did not have access to ultrasound or MRI to verify muscle identity or the location of the electrode wire once implanted in a muscle mid-belly. Moreover, implantation of fine-wire electrodes was approved as a recovery procedure where no animal could be sacrificed. However, despite the potential lack of exact placement of our electrode implants in some muscles, EMG activation broadly sampled from flexor/extensor musculature (per limb region) is adequate to achieve the objective and test our hypothesis. Also, we were not able to sample EMG from all muscles per behavior and individual owing to electrode failure. This is typical of in vivo studies and most often involved an individual scratching at the implantation site or implant failure following SW trials. We did not have the option (or time) to record SW and VC on separate days to minimize stress on the animals. Perhaps it would have been advantageous to alternate the order of SW and VC trials, although suspensory walking is stereotypical in sloths and we wanted to ensure adequate sampling of this locomotor behavior a priori. The combination of all EMG recordings for each individual, muscle and behavior are presented in Table S2.

Three other factors are offered as potential study limitations. First, all individuals were not capable of performing all behaviors equally well. This may be attributed to the time of day that sloths were required to perform locomotor trials and this time not coinciding with a typical active period for Bradypus. Additionally, measurements from some muscles could vary owing to fatigue and/or because muscle suites from the same individual were sampled on separate days. Second, it is also possible that sloths were not fully recovered from sedation at the outset of the EMG recordings. Sloths do not appear to metabolize drugs quickly, and despite administration of a reversal agent, individuals often performed better after multiple bouts of SW. For this reason, we often made 10–15 records of each behavior to ensure that we were sampling optimum performance, which included consecutive strides of near steady-state walking and climbing. It is noted that locomotor velocity was not controlled for in our experimental approach, although all trials analyzed show little variation in duty factor, indicating that preferred speeds were comparable across individuals. Moreover, sloths do not employ steady-state locomotion, favoring instead short bouts of intermittent movement. A third possibility could have been the large beam diameter. At 22 cm, the substrate was considered to be very large in previous studies of two-toed sloths (Granatosky et al., 2018). However, Bradypus may prefer or is more capable of grasping larger substrates (Mendel, 1985) and none of our individuals appeared to have difficulty placing their entire foot on the beam; this should therefore not have been a factor limiting SW performance.

Three-toed sloths performed SW using a lateral sequence diagonal couplet (LSDC) gait and VC using more diagonal sequences of footfalls (Fig. 2A,B). The LSDC gait observed during SW corresponded with the original formulations of Cartmill et al. (2002), as well as the predictions of Usherwood and Self Davies (2017) for slow gaits with large duty factors. Diagonality varied slightly among individuals, duty factors and behaviors, with VC correspondingly showing generally greater diagonality than SW in B. variegatus (Fig. 2C). Stride parameters for each individual are presented in Table S3. Stride frequencies were low and similar (P=0.62) between SW and VC, whereas stride durations were long but variable, and also did not differ significantly (P=0.95), resulting in slow velocities for both locomotor behaviors. Nevertheless, on average, three-toed sloths (N=6) walked nearly twice as fast (P=0.02) as they climbed (0.07 m s⁻¹ vs 0.04 m s⁻¹), but had the same (P=0.22) mean duty factors of 0.83±0.07 and 0.83±0.05 for SW and VC, respectively. Therefore, overall, these stride parameters between SW and VC were not statistically different (P=0.14).

A total of 117 trials, 224 strides and 856 EMG bursts were analyzed. Raw EMG traces are shown in Figs S1–S3. EMG activations during SH were long in duration (5–60 s), and each muscle generally showed intermittent activations over several minutes of recordings. For the locomotor behaviors, relative EMG durations were broadly similar between proximal and distal limb muscles, as well as flexor versus extensor muscle activations for both SW and VC. Means for all temporal variables during SW and VC are presented in Table 1.

All flexor muscles showed some pre-activation (i.e. activation in late swing phase into contact phase) and it was observed most frequently in BB, as well as in FDP (for VC), although no muscle averaged EMG onsets that occurred prior to grasp (negative %). In general, the timing of flexor muscle activity within stride cycles was similar between SW and VC (Fig. 3). Flexor muscles had earlier mean EMG onsets compared the extensors and all were activated within the initial 30% of the stride during SW, or except for DELT and PL, within the first 8% of the stride during VC (Table 1). The elbow flexor BCR activated earliest of all muscles, whereas TBLO showed the latest activation, typically at mid-contact phase during SW, and its activation was further delayed during VC (Fig. 3). Specifically, the PSA was activated latest among the flexors during SW, although it was active in early contact phase of VC. Due to multiple activations, the extensors averaged EMG onsets that approximated 50–60% of the stride, yet the initial activation for EDC was earlier, overlapping with EMG onsets of most flexors during SW and VC (Fig. 3). Consistent patterns were also observed for mean EMG offsets, but with additional variability among the proximal flexors between SW and VC (49–73% stride: Table 1). Flexor muscle activations ended earliest for BB and latest for FDP and DELT. Mean EMG offsets of EDC and TBLO were the latest of all muscles (Table 1). However, except for TBLO during VC, the initial activation of both extensors ended within the range of the flexor EMG offsets, while the second activation ended in late swing phase (Fig. 3).

Overall, relative EMG burst durations were similar (P=0.057–0.274) among the forelimb flexors for both SW and VC (P=0.546; MANOVA, effect: behavior). In particular, the distal flexors (BCR, PL, FDP) were activated for consistently long durations of the stride (∼50–60%), while the extensors TBLO and EDC had EMG burst durations that were on average 14–22% of the stride (Table 1), and these differences were significant (P<0.001; MANOVA, effect: muscle). All flexor muscles were mainly active during the contact phase of SW (>88% of trials analyzed) and VC (>83% of trials analyzed), with notably BB, BCR and FDP activated almost the entirety (97–100%) of limb contact during VC. The TBLO and EDC, which were most typically activated during both the contact and swing phases of the stride cycle, are the only muscles that showed biphasic activity, hence the generally shorter average EMG burst durations for both SW and VC (Fig. 3). The protracted EMG burst duration shown for EDC during the contact phase of VC resulted from fewer measurements for this muscle and behavior.

Mean EMG intensities across all muscles and functional behaviors are presented in Table 2. EMG intensity did not differ significantly among the main flexors PSA, BB, BCR and FDP across behaviors, although activity of FDP was consistently the lowest on average for each functional behavior (Fig. 4A). EMG intensity for the elbow flexors BB and BCR were the largest across all muscles and behaviors during SW with means of 0.71±0.19 and 0.72±0.16, respectively. The extensors TBLO and EDC showed the largest EMG intensity during VC (Table 2). In general, EMG intensity for all muscles was greatest during SW, moderate during VC and least during SH (Fig. 4B), and these differences were significant (P<0.001; MANOVA, effect: behavior). Compared with SH, the mean EMG intensities were nearly twice as large during SW (pairwise data: P<0.001) and 1.3-times greater during VC (pairwise data: P<0.001). Mean EMG intensity for all muscles during SW was also greater (pairwise data: P=0.022) than that for VC.

PCA for the EMG intensity spectra revealed that PC1 and PC2 weights explained 75.8±5.2% and 12.2±4.7% of the spectral signal, respectively. Combinations of these PCs allowed extreme low and high frequency components to be identified from the spectra for each muscle across all functional behaviors: the low frequency component had a mean of 147±55.7 Hz (range: 82–222 Hz) and the high frequency component had a mean of 350±69.7 Hz (range: 259–472 Hz) (Fig. 5). Notably, the low frequency components from PS, DELT and BCR (Fig. 5A,B,E) had low mean frequencies that averaged 88.9 Hz, 95.3 Hz and 81.8 Hz, respectively, while BB, TBLO and PL demonstrated intermediate low frequency components (Fig. 5C,D,F) and those for EDC and FDP had higher mean frequencies >200 Hz. The high frequency components from the EMG intensity spectra generally followed the same pattern for each muscle. On average, the FDP (383 Hz) and EDC (472 Hz) exhibited the two largest mean frequencies for their high frequency components (Fig. 5G,H).

The PC loading scores allowed the major properties of the EMG intensity spectra to be distinguished among the three functional behaviors and showed a lower angle theta (θ) for SH than that for either SW or VC, which were equally large (Fig. 6). SW had the highest intensity (shown by its PC1 loading score), while VC had the lowest frequency properties. Conversely, SH had lowest intensity and highest frequency properties of the three functional behaviors.

Interest in structure–function of sloths has increased over recent decades (Mendel, 1981a,b, 1985; Montgomery, 1983; Urbani and Bosque, 2007; Nyakatura and Fischer, 2010, 2011; Nyakatura et al., 2010; Nyakatura, 2012; Cliffe et al., 2015, 2018), although no previous study has evaluated sloth muscle function in vivo during suspensory locomotion and posture. Moreover, the majority of locomotor studies involving sloths (Nyakatura et al., 2010; Nyakatura and Andrada, 2013; Granatosky and Schmitt, 2017; Granatosky et al., 2018) have focused on only Linné's two-toed sloth (Choloepus didactylus) and were limited to small sample sizes, whereas the present study observed B. variegatus (N=6) for which less data are available.

Several of our major findings support the hypothesis that activation of sloth forelimb muscles reflects a neuromuscular system specialized to conserve metabolic energy. To summarize, first the shoulder, elbow and digital flexors of B. variegatus typically have long burst durations during the contact phase of locomotor behaviors, with notable early EMG onsets of the well-developed, strong elbow and digital flexors. Second, relatively low levels of muscle activation were detected during SH compared with significantly greater levels for all muscles when individuals performed SW and VC. Third, the forelimb musculature collectively had the signature of high frequencies in their EMG intensity spectra for SH, whereas the opposite pattern was similarly observed for SW and VC. Fourth, the shoulder (PSA and DELT) and elbow (BCR) flexors had characteristically low frequencies in their EMG intensity spectra across locomotor behaviors. These data are both surprising and informative, and overall, they help to validate understanding of neuromuscular recruitment patterns involved with suspensory habits. Importantly, our results refine our expectations for the mechanical properties of limb bones to endure tensile versus compressive loading, which we are currently testing in tree sloths.

The suspensory habits of sloths require their musculoskeletal system to resist mainly tensile loads. As predicted, the forelimb flexors are active during the majority of the contact phase of SW, whereas the extensors are activated in the contact and swing phases, and these in vivo patterns of muscle activation match mechanical predictions for sloths (Nyakatura, 2012; Nyakatura and Andrada, 2013) and previous EMG findings for Nycticebus (Jouffroy and Stern, 1990). The forelimb of B. variegatus notably contains strong flexors that are collectively capable of 2–5 times greater joint torque application than its extensor musculature (Olson et al., 2018). Hence, unlike pronograde arboreal locomotion, sloths rely on their forelimb flexors for below-branch stabilization (Miller, 1935; Mendel, 1985) and propulsion (Granatosky and Schmitt, 2017) during suspensory locomotion. However, during SW, the forelimb flexors of the slow loris appear to function mainly in support/stabilization, whereas activation of its hindlimb extensors late in the contact phase indicate a greater role in propulsion (Jouffroy and Stern, 1990). Comparable EMG data are not yet available for sloth hindlimbs.

One of the goals of this study was to interpret the functional roles of specific limb muscles during suspensory locomotion. Although, we acknowledge that formal studies of limb kinematics and kinetics in Bradypus are needed to confirm our interpretations of muscle function from EMG. From the video recordings, the forelimb is in a protracted position at the beginning of an SW stride, and as the contact phase progresses, the limb is retracted. Early activation of the biarticular BB may contribute to protraction of the forelimb during late swing and its sustained large activation intensity could be related to its ability to stabilize both the shoulder and elbow joints during contact. It is also possible that the lengthy (>50% stride) and large activations of BB (and BCR) provide some propulsion over the first half of limb contact. EMG onsets of the PSA and DELT occurred nearly 20–30% into the stride cycle also indicating a role in shoulder joint stabilization and controlled forward progress. Moreover, their intermediate activation intensities may be related to a limited degree of forelimb retraction in Bradypus, which agrees with kinematics observed for C. didactylus showing that its forelimb is most frequently in a neutral posture for SW (Granatosky et al., 2018). Unlike Choloepus, however, the elbow joint of B. variegatus is flexed throughout the majority of the stride cycle. In particular, the EMG activation patterns of the strong BCR most likely indicate the importance of its role as the main flexor of the elbow joint for body weight support.

The function of BCR in Bradypus during SW mimics the anti-gravity role of the m. brachialis in Nycticebus (Jouffroy and Stern, 1990). Both muscles are activated early in and maintain elbow flexion over the contact phase to control the distance between the body and support. Individuals in our study were observed to bring the body closer to the support early in the stride. Unlike BCR, the biarticular TBLO is activated nearly halfway into the stride, indicating that it does not play much of a role in elbow joint position control. We speculate that either the medial or lateral head of the m. triceps brachii performs this function in addition to active extension of the elbow joint during swing phase. However, the late biphasic activation pattern of TBLO matches that of the m. triceps brachii lateral head in Nycticebus (Jouffroy and Stern, 1990), suggesting the lack of early co-activation of major forelimb flexors and extensors during SW. The initial activation of TBLO occurs approximately when the ipsilateral hindlimb is being advanced and thus its activity could provide stabilization to the shoulder joint late in contact. Alternatively, activity of TBLO during its first burst likely provides some propulsion following mid-contact as indicated for the slow loris walking below-branch (Jouffroy and Stern, 1990), while activity of its second burst facilitates elbow joint extension in swing phase.

Digital flexion typically occurs just prior to complete purchase of the substrate by early activations of FDP during SW. Although carpal joint flexion and abduction (ulnar deviation) could not clearly be discerned from only sagittal video recordings, the early, elongated and moderately intense activations of both PL and FDP imply active maintenance of the carpus in this position. Kinematic evaluations of Choloepus show that it maintains the carpus almost exclusively in abduction during SW (Granatosky et al., 2018) and our individuals of Bradypus notably placed the majority of their forefeet on the beam to accommodate its large diameter, which requires both strong carpal flexion and abduction. Exact foot posture remains to be confirmed by comparable kinematic analyses in three-toed sloths. Nonetheless, active gripping of the substrate is prolonged until late into contact, as would be expected. The initial activation of EDC overlaps with activity of the distal flexors, suggesting that the digital extensors counterbalance flexion of the carpus and digits where PL and FDP activation intensity is largest. Similarly to TBLO, deactivation of EDC in contact and reactivation in swing phase (to extend the carpus/digits in preparation of a new grasp) of SW may demonstrate typical flexor versus extensor EMG activations in sloth limbs.

Beyond the similarity in stride parameters, the consistency in EMG patterns between SW and VC was notable, as was previously observed in climbing Nycticebus (Jouffroy and Stern, 1990). The forelimb is protracted at the beginning of a climbing stride and is retracted as the contact phase progresses, possibly by the action of PSA, which activates earlier (10% stride) compared with its delayed onset during SW. We speculate that m. latissimus dorsi could act as a major limb retractor during both SW and VC. Unfortunately, reliable percutaneous implantation of fine wire electrodes in the m. latissimus dorsi of Bradypus was not possible and the lack of complimentary data from this muscle could partially explain the relatively lower intensities of PSA and DELT during VC compared with those determined for SW. Nonetheless, as with walking, the elbow joint remains relatively flexed throughout the stride cycle except during extension in late contact and swing phase when the animal reaches for new purchase of the substrate. Thus, the hypothesized roles of BB, BCR and TBLO remain largely unchanged, even though TBLO likely provides greater propulsion by pushing throughout late contact during VC.

One difference between SW and VC, however, appears to be the position of the carpal joint. While early and prolonged activation of FDP is consistent with grasping the substrate, our video recordings of climbing in a frontal view do not allow for estimation of how much carpal flexion occurs as sloths ascended a fenced enclosure. Delayed EMG onset and low burst intensities of the PL suggest the carpal joint may remain relatively extended during VC on a fence, although our findings for PL during VC are tenuous, especially those for burst intensity. We recognize that allowing individuals to climb a natural substrate would have resulted in a forefoot position more similar to that previously observed in Bradypus (Mendel, 1985). Under these conditions, PL would be expected to exhibit an early EMG onset and prolonged activation as it does during SW. The EDC, however, shows the opposite pattern of longer, more intense activation in the contact phase of VC to maintain a relatively extended carpal joint. The TBLO and EDC are activated again to purportedly extend the elbow and carpus/digits, respectively, as the forelimb is protracted when sloths ascend the substrate. Moreover, both extensors were more intensely activated relative to the flexors during VC, indicating that beyond mid-contact, the extensors play a larger role in resisting gravity than do flexors when climbing (Jouffroy and Stern, 1990).

The forelimb flexors appear to play a role that minimizes overall muscle activity in postural suspension, consistent with our predictions. Perhaps most revealing was that the distal flexors consistently showed the lowest levels of activation among all flexor muscles across the three functional behaviors analyzed. This finding suggests that structure–function in sloth digital flexors may be analogous to a suspensory apparatus found in upright ungulates (Hildebrand, 1960; Butcher et al., 2007, 2010). For example, EMG recordings from the digital flexors and forelimb extensors of horses support this hypothesis by showing minimal activation during standing (Jansen et al., 1992; Harrison et al., 2012). The muscle activation patterns for the FDP in Bradypus are in accord with this previous observation and thus sloth distal flexor muscles may function to lower metabolic energy expenditure. By this mechanism, the FDP would produce enough force to counteract tensile loading of its tendons to sustain flexion of the digits/claws for grip on the substrate, all while requiring low-level muscle activity.

In contrast, EMG intensity of the forelimb muscles peaked during SW, which was not originally anticipated. Two aspects of sloth locomotor mechanics may explain this result: (1) greater levels of muscle activation during SW may be related to the application and maintenance of larger horizontal and medial substrate reaction forces by the forelimbs compared to those of the hindlimbs (Granatosky and Schmitt, 2017); and (2) body weight has been shown to be equally distributed between forelimb and hindlimb pairs during SW (Granatosky et al., 2018). Climbing should place relatively greater demands on the hindlimbs for body weight support, and while comparable kinetics data are not yet available for VC in sloths, EMG intensities were likely lower than those observed for SW because of shared roles in propulsion by forelimbs and hindlimbs to move up the substrate, as found in primates (Hanna et al., 2017). Studies of muscle activities in the hindlimb extensors similar to those for lorisids (Jouffroy and Stern, 1990) would be a good test of this hypothesis. Moreover, it is of particular interest to assess the functional roles of the hindlimb musculature because climbing is an energetically expensive behavior in mammals (Hanna et al., 2008) and EMG patterns from sloth hindlimbs also could be modulated, regardless of behavior, to offset metabolic cost. Such a proposed neuromuscular modification would match well with novel data (Spainhower et al., 2019) showing that sloth hindlimb muscles contract considerably more slowly than their forelimb musculature.

Large activations of the two elbow flexors contributed most to the significance in EMG intensity during suspensory walking. Even without an EMG recording of maximum voluntary contraction to normalize the EMG, relative mean ratios of EMG intensity greater than 0.70 for BB and BCR further emphasize the importance of elbow joint for suspensory function (Fujiwara et al., 2011). Sustained strong activations by the elbow flexors can mitigate tensile limb loading to relieve stresses acting at the shoulder joint during weight bearing, as previously suggested (Nyakatura et al., 2010; Granatosky et al., 2018). This could be chiefly important for Bradypus, which, unlike Choloepus (Miller, 1935), has a highly mobile pectoral girdle (Nyakatura and Fischer, 2010) because of the lack of muscular attachments to the cervical and cranial regions of the neck and head, respectively. Outside these differences, the intrinsic forelimb myology of B. variegatus is more conservative compared with that of Choloepus. In particular, two-toed sloths enhance elbow flexion for anti-gravity support via a muscle chain among m. biceps brachii long head, m. deltoideus and m. pectoralis superficialis posterior (Nyakatura and Fischer, 2011) that is somewhat similar to that in gibbon forelimbs (Jungers and Stern, 1980). In three-toed sloths, however, the strong BCR has the ability to apply substantial flexor torque owing to its large area and long moment arm at the elbow joint (Olson et al., 2018). Large activations of the elbow flexors in B. variegatus may explain the previous finding by Granatosky et al. (2018) that sloths exert appreciable and constant medial force on the substrate during SW.

In addition to the mechanics at the elbow joint, EMG activation patterns for the shoulder musculature offer new insights into locomotion in sloths. Suspensory locomotion in sloths does not involve pendulum-based energy exchanges (Nyakatura and Andrada, 2013), yet the mechanism by which they maintain a constant horizontal position of their center of mass (CoM) throughout a stride (Granatosky et al., 2018) is not fully understood. Muscle activations for B. variegatus help to reconcile this set of previous findings by the observed delayed activation of PSA, which may coincide with a hypothesized (Granatosky et al., 2018) co-activation of an ipsilateral hindlimb flexor to decelerate the CoM and prevent the substrate from oscillating. Although the two heads of m. pectoralis superficialis are slightly distinct in B. variegatus (Olson et al., 2018), we agree that the slow-contracting PSA (Spainhower et al., 2018) is most likely a main stabilizer of the shoulder joint during SW. Notably, the percentage of slow MHC-1 fibers in m. sartorius and m. adductor (Spainhower et al., 2019) is comparable to that of PS (both heads), and we further hypothesize that co-activation of these two muscles with PSA may be the mechanism by which sloths balance their body weight between the forelimbs and hindlimbs during SW. The complimentary pairing of opposing muscles with an ability to adduct the limbs at the shoulder and hip joints may also serve to make the body axis rigid and prevent swinging.

A suspensory system entirely driven by muscle activation requires modifications beyond passive tendon support (Mendel, 1981b) to minimize the metabolic cost of contraction. Two additional features of sloth limbs are possible strategies to economize locomotion. First, despite differences in behavioral and ecological preferences between Bradypus and Choloepus (Mendel, 1981a, 1985; Hayssen, 2010, 2011), members of both genera move at low speeds (Britton, 1941; Granatosky and Schmitt, 2017) and have large duty factors (Nyakatura et al., 2010). Their slow movements are emphasized by the LSDC gait of B. variegatus during SW. This pattern of footfall phasing may not only be more stable (Granatosky et al., 2019), but it is also expected to minimize mechanical work (Usherwood and Self Davies, 2017). It is interesting that diagonality and duty factors for VC parallel those observed for SW; thus velocity may be constrained in Bradypus as related to selection for stealth and cryptic movements, its main form of predatory avoidance (Carillo et al., 2009). Second, we suggest that sloth locomotor patterns are strongly related to selective recruitment of slow versus fast motor units and limb muscle fiber type characteristics (Spainhower et al., 2018), which in turn are correlated with extremely low basal metabolism (Pauli et al., 2016; Cliffe et al., 2018), therefore reflecting their critical need to conserve energy.

The activation parameters of flexor muscles inform the potential recruitment patterns of slow and fast motor units in sloth forelimbs consistent with our main objective. Wavelet analysis demonstrated that despite low EMG intensity, the EMG mean frequencies (from the EMG frequency spectra) of the forelimb musculature were overall the highest during SH. Higher mean EMG frequencies have previously been associated with faster muscle ‘fiber types’ when recorded using fine-wire EMG in animal studies (e.g. Anapol and Jungers, 1987; Hodson-Tole and Wakeling, 2007; Lee et al., 2013) and these results match the slow-to-fast gradient in MHC expression observed distally along the forelimbs of Bradypus. Specifically, high mean EMG frequencies of FDP and PL are consistent with their large distributions (54–65%) of fast MHC-2A fibers (Spainhower et al., 2018), and although fiber type percentage was not quantified for EDC, it is expected to consist of a majority of fast MHC-2A fibers based on the spectral properties of its EMG.

Wavelet results further imply that it is possible that motor units in sloth muscles are modified in both size and number, whereby smaller, faster-contracting MHC-2A fibers may be organized into smaller-sized motor units (fewer fibers) that are initially recruited when less muscle force is needed to balance the body weight. Lower magnitudes of muscle force should be needed to stabilize the shoulder, elbow and carpal/digital joints during SH, consequently placing greater emphasis on the strength/stiffness properties of the distal flexor tendons for postural support, which is supported by recent preliminary data (Mossor et al., 2019). Overall, the high mean EMG frequencies for the forelimb flexors during SH infer recruitment of small, fast MHC-2A fibers (i.e. activations of fast MU) from both the proximal and distal musculature, matching a major functional role of the limb flexors in joint stabilization and support.

Recruitment of potentially larger, slow motor units for enhanced force production is best reflected by the low mean EMG frequencies in the proximal flexors PSA and DELT, as well as the distal BCR. These muscles collectively showed the lowest mean EMG frequencies (82–95 Hz) in B. variegatus. While comparable wavelet data from other mammals capable of suspensory or clinging behaviors are not available, several studies of terrestrial locomotion (Wakeling et al., 2001; Hodson-Tole and Wakeling, 2007; Lee et al., 2013) in mammals with limb muscles that co-express MHC-1 and MHC-2A fibers only, report mean EMG frequencies for slow motor units of limb extensors (e.g. m. gastrocnemius) that generally range from 120 to 200 Hz. The lower mean EMG frequencies observed in PSA, DELT and BCR of Bradypus suggest a very low maximum contractile velocity (V_max) for their slow MHC-1 fibers that does not closely follow body size scaling relationships for a reduction in V_max with increasing body size (Toniolo et al., 2007).

The mean EMG frequencies for these three flexor muscles in sloths are also lower than those that have been estimated from the slowest motor units in fish and cats (Wakeling et al., 2002), rats (Wakeling and Syme, 2002) and goats (Lee et al., 2013); thus they may represent some of the slowest contractile properties of twitch muscle ever reported. Moreover, trading off contractile velocity for enhanced force output would match the most common pattern observed for sloth muscles: the lowest mean EMG frequencies, but significantly larger EMG intensity during SW (Fig. 4). Selectively recruiting potentially larger, slow motor units in the forelimb flexors appears to be necessary for propulsion during suspensory locomotion. Notably, this purported neuromuscular function involved sloth locomotor habits could provide advantages for enhanced force production despite a larger volume of active muscle recruited, but activation of large, slow fibers at relatively low EMG frequencies. The extremely slow-contracting BCR in B. variegatus (Spainhower et al., 2018) is perhaps the best example of a muscle that emphasizes force while minimizing shortening velocity (Olson et al., 2018) and these contractile features are generally reflective of sloth forelimbs.

Finally, the collective modulations in muscle activations between postural and locomotor behaviors are shown by the PCA loading scores (see Fig. 6), and these key findings help to clarify use of selective or preferential recruitment of motor units (Wakeling, 2004) in sloth muscles. The overall higher mean EMG frequencies observed for SH associated with recruitment of fast MHC-2A fibers, may be a fundamental example of how the neuromuscular system in sloths is ‘tuned’ to economize force production by recruiting the most appropriate motor units required for specific contractile tasks. By doing so, motor unit recruitment patterns in sloth forelimbs do not strictly follow the size principle (Henneman et al., 1974). However, this potentially unique feature in sloth limbs helps them reduce their overall metabolic energy expenditure, which is similar to that suggested for numerous other axial (Jayne and Lauder, 1994) and limb muscles (Smith et al., 1980; Wakeling, 2004) in diverse vertebrate taxa. Such a pattern of muscle recruitment may be critical to how B. variegatus offsets the cost of flexor activation for postural suspension, and the degree of neuromuscular modification that we propose may be even more specialized in Choloepus. We also hypothesize that suspensory primates have evolved similar neuromuscular control features in their forelimb muscles. For example, lorisids are ecologically and physiologically convergent with sloths. Last, it was essential to observe the similarity in muscle activation patterns between SW and VC. When moving either horizontally below or vertically up the substrate, sloth forelimb muscles appear to have largely similar performance requirements, which could be representative of energy savings by coupling the adaptive strategies of slow/intermittent movements with selective recruitment of large, slow MHC-1 fibers.

Despite the requirement of lengthy and large flexor muscle activations during suspensory locomotion, sloths appear to have neuromuscular modifications that can offset the cost of muscle contraction, including reduced EMG intensity for postural support and several specialized slow-contracting muscles with low mean EMG frequencies. In particular, B. variegatus may selectively recruit smaller, fast motor units at lower activation levels during suspensory hanging, but uses recruitment of larger, slow motor units during suspensory walking and vertical climbing when greater force production is needed. Three-toed sloths therefore demonstrate a division of labor in their forelimb musculature and this major finding provides the clearest support for the hypothesis that sloth forelimb muscles reflect a critical part of a physiological system specialized to conserve metabolic energy. We further clarified several muscular functional requirements for suspension and informed other limb mechanical features necessary to maintain even suspensory support of the body weight in sloths.

We sincerely thank Gerald Richardson (Dallas World Aquarium) for coordinating this research partnership. We also thank the staff at The Sloth Sanctuary of Costa Rica for care of the animals. Thanks to I. Pearce for help with electrode construction and B. Chadwell, C. Riwniak, and C. Lavin for assistance with data analysis. Special thanks to E. MacDonald, E. Harris, and N. Fomenky for light box construction, also T. Diggins for assistance with statistical analysis. Portions of the work were submitted as a Masters Thesis by M.A.G. The YSU College of Graduate Studies, College of STEM, and Office of Research, as well as the SFU Department of Biomedical Physiology and Kinesiology are gratefully acknowledged.