Longitudinal and transversal displacements between triceps surae muscles during locomotion of the rat

Animal movement is the result of the interaction and coordination of multiple muscles. Despite the importance of understanding such complex functional integration (Higham and Biewener, 2008), little is known about the consequences of differential fiber strain and level of activation occurring in vivo between individual muscles that cooperate within a synergistic muscle group. Regional variation in strain and heterogeneous patterns of activation have been observed between fascicles within muscles and between synergistic muscles during normal movements of the toad, cat, rat, guinea fowl and human (Ahn et al., 2003; English, 1984; Gillis and Biewener, 2001; Higham and Biewener, 2008; Lichtwark et al., 2007; Maas et al., 2009; Wakeling, 2009). Such variation is the result of morphological and architectural differences within and between muscles, such as cross-sectional area, fiber-type composition, pennation angle, activation history, level of fatigue, timing of activation and in-series tendon compliance (Higham and Biewener, 2011). These variations in strain may have effects on muscle force production through mechanical interactions between muscles mediated by in-series common elasticity (Sandercock, 2000; Tijs et al., 2014) and myofascial connections (Maas and Sandercock, 2010).

The triceps surae muscles of the rat have been widely used as an experimental model for investigating the mechanical properties of skeletal muscles (Bobbert et al., 1990; Devasahayam and Sandercock, 1992; Ettema and Huijing, 1994; Messner et al., 1999; Pickett et al., 2008). As a consequence, the anatomy, physiology and mechanical properties of these muscles are well characterized. Soleus (SO) and gastrocnemii (GAS; medial and lateral gastrocnemius, MG and LG, respectively) muscles share a similar anatomical location and function at the ankle joint, but differ substantially in pennation angle (Close, 1964; Heslinga and Huijing, 1990) and fiber type composition (Armstrong and Phelps, 1984). The SO is mainly composed of slow-twitch oxidative fibers (87%), while the LG and MG consist mainly of fast-twitch oxidative-glycolytic (28%) or fast-twitch glycolytic (65%) fibers (Armstrong and Phelps, 1984). The SO is much smaller than the GAS, with a physiological cross-sectional area of 8.8 and 63.3 mm² for SO and LG, respectively (Woittiez et al., 1985). In contrast to the SO, the GAS also span the knee joint. As a consequence, changes in knee joint angle will result in changes GAS length as well as changes in the relative positions of these muscles. Such anatomical and morphological organization can have consequences for muscle mechanics. When manipulating the relative position of muscle–tendon units (MTUs) within a physiological range in situ, mechanical interaction between the SO and LG was found (Bernabei et al., 2015). Previous studies have investigated the activation and length changes of the triceps surae muscles in vivo (Hodson-Tole and Wakeling, 2007, 2008), yet muscle relative displacements have not been measured.

Displacement between muscle bellies and/or tendons of neighboring muscles is associated with changes in the transmission of force via intermuscular or intertendinous connective tissues pathways. Changes in muscle relative position of the SO and LG are expected because of differential changes in MTU length and differences in contraction velocities of the muscle bellies. Such differences, in turn, can be the result of differential activation levels, asynchronous patterns of activation and differences in pennation angle, or a consequence of joint angle changes as the two neighboring muscles span different joints and have different moment arms. To date, a number of in situ studies have reported mechanical effects of changes in muscle relative position (reviewed in Huijing, 2009). More recently, ultrasound studies have reported displacements between SO and GAS in humans, but only for passive muscle conditions during isolated knee movements or selective electrical stimulation of targeted muscles at fixed joint angles (Bojsen-Moller et al., 2010; Finni et al., 2015; Oda et al., 2007; Tian et al., 2012). In addition, such relative displacements were measured in one dimension. Therefore, very little is known about the magnitudes of displacement between SO and GAS muscles during normal movements.

The aim of the present study was to investigate the coordinated action of ankle plantar-flexors during locomotion and to analyze how this adapts to varying locomotor demands. In particular, we assessed the patterns of activation and contractile behavior of triceps surae muscles, as well as the 3D relative displacement between the one-joint SO and the two-joint LG muscle bellies and distal tendons during level and upslope trotting. We expected that during upslope trotting, the increased mechanical demands and altered knee joint kinematics would increase the relative displacement between adjacent muscle bellies. We hypothesized that such increased relative displacement would be associated with a substantial increase of differential length changes of both muscles as well as deformation of the tissues connecting them, owing to differences in muscle activation patterns, effects of in-series compliance and intermuscular connections. These differences may affect the moment produced at the knee and ankle joints and, thus help elucidate how different muscles cooperate to achieve their functional role.

Seven male Wistar rats [Rattus norvegicus (Berkenhout 1769), body mass at time of measurements 277.6±15.7 g; 7 weeks old upon arrival] were tested in this study. Before implantation of sensors, each rat was trained to run on a Plexiglas-enclosed, custom-built, motor-driven treadmill (Westra et al., 1985) for 10 days. Food rewards were provided upon completion of the 2 min running task at increasing speed (up to 0.8 m s⁻¹) and slope (up to 20 deg). To promote running, a movable electrical grid placed at the back of the treadmill was used to deliver mild electrical shocks. Before touching the grid, the rats came into contact with a small object. After 2 days, the rats were trained to avoid the electrical shocks by responding to the object. During training and data collection, rats were housed in large cages (0.55×0.33×0.20 m) with access to food and water ad libitum. All surgical and experimental procedures were approved by the Committee on the Ethics of Animal Experimentation at the Vrije Universiteit Amsterdam (protocol number FBW 12-01) and in strict agreement with the guidelines and regulations concerning animal welfare and experimentation set forth by Dutch law. On completion of measurements, animals were euthanized with a pentobarbital overdose (Euthasol 20%) injected intracardially, followed by a double-sided pneumothorax.

All surgical procedures were performed under aseptic conditions. Rats were anesthetized using isofluorane gas (3% induction, 1.4–2% maintenance). A one-time subcutaneous injection of Buprenorphine (opiate, Temgesic^®, Schering-Plough, Maarssen, The Netherlands) at dosage of 0.02 mg kg⁻¹ was administered pre-operatively. The right hindlimb was surgically instrumented with sonomicrometry crystals and fine-wire electromyography (EMG) electrodes using procedures described previously (Maas et al., 2009). To measure muscle belly lengths and relative positions, two piezoelectric crystals (1 mm; Sonometrics, London, ON, Canada) were implanted near the proximal and distal aponeurosis of each muscle (see Fig. 1A, colored circles). To measure EMG activity, pairs of Teflon-insulated multi-stranded stainless steel fine-wire electrodes (25 µm un-coated diameter, 7SS-1T, Science Products GmbH, Hofheim, Germany) were chronically implanted in the mid-region of SO and LG muscle bellies. Muscle bellies were accessed by opening the skin and the crural fascia from the lateral side. The EMG electrodes were implanted approximately 2 mm deep and approximately 1 mm apart. A reference electrode was inserted underneath the skin, in the region above the gluteus maximus muscle. Given the distribution of fiber type in SO and LG muscle regions (Armstrong and Phelps, 1984) and muscle architecture (Prodanov et al., 2005), the EMG pick-up area consisted of slow oxidative fibers in the SO and a majority of fast glycolytic fibers in the LG. Electrode placement was verified by stimulation through the implanted wires. In addition, pictures of the implant were taken on the euthanized animal after data collection to verify consistent positioning and to exclude dislodging of EMG electrodes and sonomicrometry crystals.

Two other small incisions were made: (1) on the lateral aspect of the hindlimb over the biceps femoris fascia; and (2) in between the scapulae. A subcutaneous tunnel created between these two incisions was used to lead the wires from the site of sensor implantation to a transcutaneous connector mounted on the back of the animal, in the region of the shoulder blades. After the skin was sutured, the animal was allowed to recover for 2 weeks prior to data collection. Additional doses of Buprenorphine were given 1–2 days after the surgery if signs of pain were noticed.

Two weeks after sensor implantation, hindlimb kinematics, EMG activity and sonomicrometry crystal distances were recorded during level and upslope (+20 deg) trotting. Only steady-state stride cycles were selected for further analysis, excluding galloping gaits and forward–backward acceleration on the belt (Sabatier et al., 2011). Data from 10–15 s trials (30–50 stride cycles) were collected at a fixed treadmill speed of 0.6 m s⁻¹, to obtain at least 15 stride cycles recorded with the rat trotting at a constant speed (Table 1). Only stride cycles satisfying these requirements were retained for further analysis of EMG and sonomicrometry data. For each animal, we pooled all stride cycles from all trials at a given slope to generate a single representative mean time-series normalized to stride time for that animal and condition. EMG and sonomicrometry signals were sampled at 3120 Hz and digitized (Digital Sonomicrometer, Sonometrics, London, ON, Canada). Kinematic data and EMG–sonomicrometry signals from the implanted sensors were synchronized by an electronic trigger pulse to the sonomicrometry controller. Each measured variable was time-normalized to the stride-cycle time, and averaged first within and then across rats for the two different slopes.

Sagittal plane hindlimb kinematics were obtained from 2D videos during treadmill locomotion. The high-speed camera (A602f, Basler, Ahrensburg, Germany) was placed orthogonal to the treadmill, on the right side of the rat. Parallax errors were minimized by spatial transformation of the marker coordinates system with a calibration checkboard and by using a narrow (11 cm wide) Plexiglas-enclosed lane to limit medial–lateral movements on the treadmill. After shaving the right hindlimb, the following bony landmarks were marked using a permanent marker: the iliac crest, greater trochanter, lateral femoral epicondyle, lateral malleolus, fifth metatarsophalangeal (MTP) joint and the distal end of the fourth digit (Fig. 1A, white circles). To minimize skin movement artifacts (Bauman and Chang, 2010), the position of the knee joint axis was triangulated based on the position of the greater trochanter and the lateral malleolus, and using the thigh and shank lengths. Video data (400×180 pixels, 7.4 pixels mm⁻¹) were sampled at 200 frames s⁻¹, streamed to an IEEE-1394 port and recorded to the computer hard drive with custom software (LabView, National Instruments Inc., Austin, TX, USA).

A stride cycle was defined as the time between two toe-contacts of the same (right, implanted) hindlimb. For each stride cycle, the timing of two events was determined: (1) the onset of the stance phase, defined as the video frame in which the foot of the hindlimb touched the treadmill belt (paw-in), and (2) the onset of the swing phase, defined as the video frame in which the same foot left the ground completely (paw-off). The durations of stance and swing phases were calculated using the time points of these events. In addition, the stride length was calculated as the maximal absolute horizontal displacement of the toe marker during consecutive stance and swing phases. Angles of the hip, knee, ankle and MTP joints (Fig. 1A) were calculated based on the positional coordinates of the above-described anatomical landmarks. Maximum and minimum angles during the stance and swing phases were calculated for each joint.

MTU lengths of LG muscle were estimated using the ankle–knee geometric model of the rat described by Ettema (1997). Inputs for the model were the low-pass filtered (third-order, zero-lag Butterworth, cut-off 20 Hz) ankle–knee joint angles and the tibia lengths (see Table 1). In addition, displacements of the LG origin in response to changes in knee joint angle were calculated. The geometric model was adjusted to calculate MTU lengths of SO muscle using the SO origin on the tibia as measured after euthanasia. The range of MTU length changes was expressed as a percentage of the mean MTU length during the stride cycle.

EMG signals from SO and LG muscles (Fig. 1B) were amplified (100–2000×, common-mode rejection ratio >100 dB), filtered for aliasing (10–1000 Hz) and sampled (3120 Hz). Band-pass and notch digital filters were subsequently applied to remove movement artifacts (30–1000 Hz, third-order Butterworth) and power-line noise (50 Hz, third-order Butterworth). The envelope of the filtered EMG signal was computed as the magnitude of the discrete-time analytic signal calculated by the Hilbert transform, and subsequently time-normalized to 100 time bins per stride. For each rat, the amplitude of the EMG envelope for each time bin was normalized to the maximum EMG value across all time bins from all step cycles recorded during all conditions; this maximum was generally achieved in the upslope condition. The average EMG activity was then calculated as the mean of the normalized EMG envelope. The EMG data from one of the animals were excluded, because of insufficient signal-to-noise ratio. In addition to EMG amplitude, timing of EMG bursts was described by the absolute burst duration and the duration of the burst during the pre-stance phase. The onset–offset of muscle activity was determined by k-means cluster analysis applied on the band-pass filtered EMG signal (Den Otter et al., 2006), which corresponds to assigning values of the EMG amplitude to a user-defined number of clusters (n=3), with the lowest cluster reflecting inactivity.

The elapsed time for an ultrasound wave to be transmitted by one crystal and received by the other was used to assess the distance between two crystals, given the speed of sound at which the wave travels in vertebrate skeletal muscles (1580 m s⁻¹; Topp and O'Brien, 1999). Time series describing the muscle belly length changes of SO (l_SO) and LG (l_LG) during locomotion were low-pass filtered (second-order Butterworth, 50 Hz cut-off). Values of l_SO and l_LG were normalized to the reference length (l_REF), defined as the mean of the maximum and minimum muscle belly lengths during the swing phase of level running (Gabaldón et al., 2004). Muscle belly velocity of length change of SO (l′_SO) and LG (l′_LG) was computed using the method of finite differences. Positive velocity values indicated lengthening, while negative values indicated shortening. The relative contraction velocity between SO and LG was calculated at each time bin as the difference l′_SO–l′_LG.

where T_prox and T_dist represent the reference length of the proximal and distal tendons, respectively. These were measured in previous experiments in our lab (C. Tijs and H.M., unpublished data) on animals with matching characteristics (Wistar male rats, mass±s.d.: 321.8±34.2, n=24). Distal tendon length changes were expressed relative to T_dist.

the ratio between the major and minor diagonals was used as a comprehensive descriptor of 2D intermuscular deformation. This ratio was normalized to the mean value over stride.

Paired-sample t-tests were performed to assess effects of slope condition on joint angles, MTU length changes, stride characteristics and amplitude of SO and LG activity. For the following groups of similar outcome variables, P-values were Bonferroni corrected (P×n) for multiple comparisons (n): stance-swing duration and stride length (n=3), hip peak extension/flexion in stance and swing (n=4), MTP peak extension/flexion in stance and swing (n=4), and Q_RATIO peak and range (n=2). Two-way repeated-measures ANOVAs were performed to test for effects on timing of peak extension/flexion (independent factors ‘slope’ and ‘joint’). Two-way repeated-measures ANOVAs with independent factors ‘slope’ and ‘muscle’ were formed to test for effects on amplitude of SO and LG activity, on timing of muscle activation, tendon lengthening, muscle belly length and velocity. If significant interaction was found, post hoc tests with Bonferroni correction were performed. Statistical tests were performed using SPSS (V. 21, IBM, Armonk, NY, USA). In all statistical analysis, results were considered to be significant at P<0.05. Variability of descriptors is expressed as standard deviation (s.d.).

The time devoted to stance and swing was 0.16±0.01 and 0.12±0.01 s (57.1% and 42.8% of the stride time), respectively, and stride length was 0.18±0.01 m. Changing the slope did not result in different spatio-temporal characteristics of locomotion (stance, swing and stride length, all P=1.000). Details on all statistical analyses are reported in Table 2.

During level trotting, ankle and knee joint patterns were characterized by a yield phase in early stance followed by joint extension in late stance. The timing of peak extension and flexion of these two joints differed. Peak knee extension in stance occurred 8.6±3.6 ms earlier than peak ankle extension (P=0.005). During swing, knee peak flexion was observed 22.4±6.1 ms earlier than ankle peak flexion (P<0.001; Fig. 2A,B). The hip joint angle extended from 60 to 104 deg during stance and flexed during swing. The MTP joint angle decreased towards a flexion peak at 116±5 deg during stance, then it started increasing before paw-off up to a maximum angle of 194±7 deg during swing.

During upslope trotting, hindlimb joint angles were very similar to those during level trotting (Fig. 2). The timing of maximum knee flexion during swing was different than that of the ankle joint (20.0 ms earlier, P<0.001), but no differences in timing were found for peak extension (P=0.788). Compared with the level trotting condition, the hip joint peak flexion increased during stance phase (P=0.020; Fig. 2).

These data indicate that only small changes in kinematics were used to accommodate the slope difference. The different timing of maximum and minimum angles of knee and ankle joints will cause changes in relative position between the SO and LG MTUs (see below).

During level trotting, the onset of activation for both SO and LG occurred in late swing (Fig. 3A,B). Then, the activity of both plantar-flexors increased up to a peak during stance, which occurred earlier (P=0.005) for SO than for LG (8.8±3.7 and 47.6±33.0 ms after paw-in, respectively). Compared with LG, the SO showed a much steeper increase and decrease in EMG activity (Fig. 3A). The onset of SO activity was earlier than that of LG (Fig. 4C): the duration of pre-stance activation was longer in the SO than in the LG (P=0.013, 0.05±0.01 and 0.03±0.01 s, or 16.7±4.5% and 10.7±4.1% relative to stride duration, respectively).

Compared with level trotting, the EMG amplitude (Fig. 4A) increased substantially during upslope trotting in the LG (by 53.6%, P=0.041), but not in the SO (P=0.657). Total burst duration did not differ between level and upslope conditions (Fig. 4B). No difference in the duration of the pre-stance activation between the SO and LG was found for the upslope condition (P=0.570).

Overall, these patterns indicate co-activation of the SO and LG, with SO onset preceding LG onset. Although they were largely active simultaneously, the pattern of activity was rather different between these muscles, suggesting differential contractile conditions.

During level trotting, the LG showed only small length changes in stance phase, while SO muscle belly lengthened and shortened substantially (up to 15.8% of l_REF; P=0.012; Figs 3C,D, 4D,E,G). During swing, SO muscle lengthened by 33% of l_REF, while this was 21% for the LG (significantly shorter than SO: P=0.011). In addition, pre-stance shortening of the SO was considerably larger than that of the LG (30% and 13% of l_REF, respectively; P<0.001; Fig. 4F).

Peak lengthening and shortening velocities of the SO and LG were different (86.8±22.5 versus 39.6±8.0 mm s⁻¹ and −76.1±22.5 versus −53.8±19.2 mm s⁻¹, respectively, P<0.001 and P=0.025), and the time point at which these occurred was slightly shifted between the two muscles (Fig. 5A,B). This yielded a pattern of relative velocities between SO and LG muscle lengths with a peak of differential velocity in swing at 64% of the stride cycle, when SO starts shortening earlier than LG, resulting in a positive relative velocity of 55.9 mm s⁻¹ for level trotting (Fig. 5C). An effect of slope was found only for muscle lengthening during stance (P=0.010; Fig. 4E). All other measures of muscle lengths and velocities of SO and LG during contraction and relaxation did not significantly change between level and upslope locomotion (Fig. 4D–H). Therefore, for most parameters, only the data for level locomotion were reported. These results indicate that changes in muscle relative positions and relative velocities are substantial at specific instances of the stride cycle.

With combined knee and ankle angular displacement, the estimated MTU length varied between 29.3 and 35.7 mm for the SO and between 37.6 and 44.9 mm for the LG (corresponding to 19.8% and 17.6% of the average MTU length across the stride cycle for the SO and LG, respectively; Fig. 6A). The LG MTU length changes caused by knee joint angular displacement only, i.e. proximal lengthening, were between 40.7 and 42.1 mm during level trotting, which corresponds to 3.4% of the average LG MTU length. This did not differ between level and upslope conditions (data not shown), as expected from kinematics data. These results show that changes in LG MTU length were mainly determined by ankle joint angular displacement.

Estimates of distal tendon lengths of the SO and LG showed that most of the LG MTU length changes were the result of changes in tendon length, up to 3.1 mm during stance (7.5% of mean LG MTU length over the stride), but this was not the case for the SO, the tendon of which lengthened by a maximum of 0.9 mm during stance (2.5% of mean SO MTU length; Fig. 6C). This is consistent with the rather isometric behavior of the muscle belly observed during stance for the LG (Fig. 3D), resulting in differential tendon lengthening, especially just before paw-off at 52% of stride cycle (1.9 mm or 4.7% of mean LG MTU) and in swing at 74% of stride (2.0 mm or 4.8% of LG MTU length; Fig. 6C). SO and LG tendon length changes and differential lengthening were not altered by the slope (2.0 and 1.6 mm in stance and swing, respectively; data not shown). Given that the distal SO and LG tendons merge into a common tendon, these results suggest unequal strain of Achilles tendon fascicles.

The maximum distance between sonomicrometry crystals (0.57 mm relative to the mean) implanted in the proximal region (Fig. 7A) of SO and LG muscle bellies (Q_prox) was found at 52% of the stride, i.e. at the end of stance phase, while a minimum (−0.74 mm relative to the mean) occurred at 78% of the stride, i.e. during swing. In contrast, the distance between sonomicrometry crystals implanted in the distal regions of SO and LG muscles (Q_dist) did not show a consistent pattern that could be related to phases in the stride cycle. This distance varied by a maximum of 0.37 mm (Fig. 7B). The major (Q_D1, Fig. 7C) and minor (Q_D2, Fig. 7D) diagonals of the quadrangle defined by the four crystals are indicative of the quadrangle shape and were used to assess the shear deformation between SO and LG muscle bellies. Q_D1 was nearly constant during stance, but increased during swing to a maximum of 1.24 mm (relative to mean) observed at 85% of stride. Also, for Q_D2, the maximum of 1.50 mm (relative to mean) was found at 80% of the stride. Q_RATIO reached its maximum (17.3%) at paw-in, and decreased to its minimum (−21.8%) at 80% of stride cycle (swing phase). Trotting upslope did not result in significant changes in the deformation pattern (Fig. 7E, lower panel: mean±s.d.). The similarity between patterns of Q_D1 and Q_D2 and those of LG and SO muscle belly lengths (see Fig. 3C,D) suggests that the deformation of the intermuscular volume was determined by SO and LG contractile behavior. These results indicate that both longitudinal and transversal deformations of intermuscular tissues would mainly be found in the proximal intermuscular region, while the distal muscle belly ends did shear relative to each other without much lengthening of connective tissues between the muscles at this location, as the distal distance was nearly constant with substantial changes in Q_RATIO.

Adjacent muscles with different functional roles can exhibit substantial differences in their pattern of activation and contractile behavior, especially in a multi-joint muscle system such as the ankle plantar-flexor group. Mechanical interaction between these muscles via common tendons and connective tissues at their shared interface may affect force and strain distribution among the individual actuators. Understanding how these muscles are controlled and how they eventually deliver force at the joints requires detailed knowledge about the coordinated action of muscle groups. This is the first study providing a measure of the deformation of the intermuscular volume between two adjacent muscles during locomotion. We confirmed that the different functional roles of the SO and LG requires these muscles to undergo differential length changes, despite the fact that they merge in a common tendon distally. We observed displacements between the SO and LG muscles during locomotion, which may have implications for intermuscular force transmission. In contrast to our expectations, the extent of muscle relative displacement was not substantially affected by the imposed increase in the slope of the treadmill. Differential changes in muscle belly lengths and relative velocities of SO and LG were likely the result of non-simultaneous muscle activation. The relative displacements and activation levels of the muscles studied were those occurring in the freely moving animal, thus providing information about the extent of mechanical and functional interaction between muscles in an intact environment and with physiological muscle recruitment. Differential action of synergistic muscles has been observed in several species with a similar approach (Biewener et al., 1998; Daley, 2003; Ellerby and Marsh, 2010; Higham et al., 2008; Maas et al., 2009).

We found substantial differences in length changes and muscle activation between the SO and LG muscles of the rat. By active lengthening during the stance phase, the SO muscle belly likely contributed to negative work, absorbing the body inertial load after paw-in. In contrast, the almost isometric behavior of the LG suggests that this muscle stored elastic energy in its tendon (Griffiths, 1991). These data indicate a potential difference in the timing and the amount of energy absorbed by the two muscle bellies during active lengthening. Although frequently co-activated, the timing of EMG bursts between these muscles was rather asynchronous, with the SO onset of activation occurring earlier during pre-stance than that of LG. In addition, only the EMG amplitude of the LG increased substantially during upslope trotting (by 54%) whereas no changes were observed in the SO. Heterogeneity of muscle lengths and velocity between synergists during contraction may be explained by such differences in the neural drive, but also by architectural properties, such as pennation angle and fiber type composition (Armstrong and Phelps, 1984). Previous studies have shown that muscle activation and muscle fascicle deformation differ between ankle plantar-flexors (Cronin et al., 2013; Maas et al., 2009). Also, the force produced by ankle plantar-flexors and their contractile behavior has been reported to be quite diverse (Daley, 2003; Prilutsky et al., 1996). In humans, for instance, asynchronous shortening of the SO and LG during the limb extension phase of jumping allows the LG to store the power produced by the knee extensors in the series elastic element at the ankle, acting in a strut-like manner (Bobbert et al., 1986; Farris et al., 2016). It is conceivable that architectural diversity, together with a fine-tuned neural drive, allows for adaptation to varying constraints, e.g. different locomotor speeds and motor tasks (Gillis and Biewener, 2001; Gregor et al., 2006; Hodson-Tole and Wakeling, 2008; Hutchison et al., 1989), as also suggested by the different responses in EMG amplitude to changes in slope between the LG and SO reported here.

Although different activation and contractile behavior for muscles arranged in parallel and merging in a common tendon may seem paradoxical, such diversity may constitute a key feature for flexibility, subserving the rat's behavioral repertoire. The rat triceps surae allows both damping and energy recoil: while the SO absorbed energy by deformation, the LG responded more isometrically, thereby storing energy in its tendon during gait. In addition, only the activation of the LG muscle was modified in response to environmental changes (slope). This suggests that by selectively activating one muscle in the synergistic muscle group, its specific mechanical response can be exploited. In more specialized animals, the anatomy appears more tuned to just one of these functions. For instance, plantaris and LG muscles fuse for much of their length in the possum (Trichosurus vulpecula), forming one functional unit, while the SO is absent (Hore and Porter, 1972). In the camel (Camelus dromedarius), plantaris muscle fibers are vestigial, replaced by a long tendon (Alexander et al., 1982). Likewise, in the horse (Equus caballus), the digital flexor muscles have extremely short muscle fibers with very long tendons spanning several joints (Wilson et al., 2001). In other species, energy storage efficiency is maximized by shortening or removing muscle fibers (for a comparison, see Biewener and Roberts, 2000). These observations on specific mammals suggest that evolutionary changes of the triceps surae architecture would have involved separation or fusion of a previously present and functionally unsuitable set of muscles.

Our data indicate that displacements and velocities between neighboring muscle bellies can be substantial in the freely moving animal. We found the displacement between the SO and LG to be maximal in the proximal region (Q_prox), consistent with the location of their origins on the skeleton. In fact, the distance between the SO and LG is expected to increase proximally with knee extension, as the SO and LG are set apart because of their independent origin and because the LG spans the knee joint and the SO does not. Moreover, substantial changes of the quadrangle shape at paw-in and during swing (+17.3% and −21.7% relative to the mean diagonals' ratio) indicate that intermuscular displacement does not occur only in one dimension; rather, it is a combination of longitudinal and transversal displacements. As a consequence of different shape changes of neighboring muscles owing to variation in pennation angles (Azizi and Roberts, 2014) and relative muscle belly length changes in response to muscle excitation, deformation of the intermuscular volume must occur. Knee–ankle joint coordination and subsequent changes of orientation between one- and two-joint muscles' lines of pull also cause deformations of the intermuscular volume. It has been shown that changes in relative position of neighboring muscles affect epimuscular myofascial force transmission (Huijing and Baan, 2003; Maas et al., 2003). Several structures are present in the intermuscular volume between ankle plantar-flexors and, with substantial relative displacements, likely change their tensional state, possibly modifying force transmission between the SO and LG. Connective tissues are found across the whole surface between SO and LG muscle bellies, with the neurovascular tract running centrally between SO and LG muscle bellies (Fig. 8A; see Bernabei et al., 2016b for a description). Bundles of connective tissues surrounding blood vessels and nerves reaching the posterior crural muscles can be divided into two main force transmission pathways. The first pathway runs from the proximal-ventral side of LG to the dorsal side of SO. The second pathway contains branches of the popliteal artery and tibial nerve running from the popliteal fossa to the proximal-dorsal surface of LG and MG muscle bellies. In situ, mechanical interaction between SO and LG of the rat has been reported (Bernabei et al., 2015) for a range of MTU length changes (6 mm), which were consistent with the ones estimated in this study (7.2 and 6.4 mm for LG and SO, respectively; Fig. 6A). However, in the study by Bernabei et al. (2015), both muscles were maximally activated, while the extent of mechanical interaction has been shown to decrease with submaximal activation and passive muscle conditions (Tijs et al., 2015a, 2016a). Moreover, the maximal MTU relative displacement imposed in the study by Bernabei et al. (2015) was larger (3.0 mm) than that found in vivo in the present study (1.4 mm). It should also be noted that in previous experiments only one-dimensional (longitudinal) relative displacements between MTUs were imposed. However, we found here that an increase in longitudinal relative position was accompanied by a decrease in orthogonal intermuscular distance (Fig. 7E), thereby reducing the strain of the intermuscular connective tissues. This could partially explain the limited or no mechanical interaction between muscles when intact muscle compartments and physiological joint excursions in rats and cats hindlimb muscles were tested (Sandercock and Maas, 2009; Tijs et al., 2015b). According to previous studies relating the amount of intermuscularly transmitted force with the active and passive state of muscles (Bernabei et al., 2016a; Tijs et al., 2016b), the largest effects of intermuscular force transmission are expected during the pre-stance and stance phases, when both the LG and SO are active. When the relative position changes between passive muscles, compliant intramuscular structures will likely fully absorb the resulting shear strain and, hence, the strain on intermuscular connections will be minimal. Therefore, the largest displacement and differential velocity of muscle belly observed during mid-swing likely involves negligible amounts of intermuscularly transmitted forces.

Our observed relative displacements and the previously reported minimal mechanical interaction between submaximally activated muscles suggest that during level and upslope trotting, the extent of epimuscular force transmission between ankle plantar-flexors is not significant in healthy rats. This may change in pathological conditions of the musculoskeletal system, such as those caused by muscle injury and surgical interventions, for which enhanced mechanical interactions between muscles have been reported (Bernabei et al., 2016a; Maas and Huijing, 2012).

We estimated the difference in length changes between SO and LG distal tendons during locomotion to be maximally 4.8% of the mean LG MTU length. In contrast, we showed that the distance between the distal regions of SO and LG muscle bellies (Q_dist) did not undergo substantial changes over the stride cycle. This could be explained by a stiff intertendinous connection between SO and LG tendons, which causes transversal displacement during longitudinal displacements (see Fig. 8C–E). In the rat, the Achilles tendon can be divided into a proximal portion with tendon bundles separated from the SO, LG and MG, and a distal portion in which these bundles are fused together and no longer separable without dissection (Fig. 8B). The total length of the tendon is approximately 10 to 15 mm in an adult rat (male Sprague-Dawley, 400–425 g; Krapft et al., 2012). A recent study reported that the length of the merged distal portion (∼45% of the total Achilles tendon length) can change from 6.6 to 6.9 mm depending on knee extension and SO+GAS activation (Tijs et al., 2015b; male Wistar rat 315–340 g). Only a few studies have assessed the mechanical response of the SO and LG common tendon in physiological conditions (Finni et al., 1998; Franz et al., 2015; Slane and Thelen, 2014). Interestingly, the consistent finding that the Achilles tendon fascicles in series with the SO or LG undergo differential displacements was interpreted rather differently (Arndt et al., 1999; Bojsen-Møller and Magnusson, 2015). On the one hand, this relative sliding may indicate that individual muscles transmit their forces independently; on the other hand, relative sliding may result in intertendinous force transmission. With regard to the latter, it has recently been shown that the matrix between tendon fascicles of equine tendons has elastic properties (Thorpe et al., 2012, 2015). This implies that forces can be transmitted between different tendon fascicles when substantial sliding occurs. Connective tissue linkages between tendons could serve as a stress-release mechanism protecting tendon fibers from damage by redistributing forces to neighboring fascicles. Moreover, tendon bundles from the different muscles of the triceps surae could cooperate for elastic recoil as a system of elements arranged in a parallel fashion. An example of optimization of this mechanism for elastic recoil is given by the complex system of accessory ligaments in the horse lower limb (Wilson et al., 2001). Non-linear summation of joint moments of the triceps surae muscles found in the rat indicates that SO and LG tendons are not mechanically independent (Tijs et al., 2014) and, hence, stress distribution seems likely. However, the function and implications of the complex arrangement of the triceps surae MTU are still not fully understood (Bojsen-Møller and Magnusson, 2015).

Coordinated activation and contractile behavior of two ankle plantar-flexors were directly measured by implanted sensors, while MTU and tendon length changes were estimated. As a consequence, several issues need to be considered when interpreting the results presented here.

It should be noted that the MG is part of the triceps surae muscle group and it shares an aponeurosis with the LG distally. However, in this study we focused on SO–LG interaction exclusively for several reasons. First, while there is evidence of direct connective tissues linkages between SO and LG muscle bellies (Fig. 8A), the MG and SO do not share an interface. Second, previous studies have quantified the force transmitted via inter- and extramuscular connective tissues between SO and LG as a function of relative displacement (Bernabei et al., 2015, 2016b), while this information is not available for the LG–MG or SO–MG interaction. Third, we reduced the number of implanted sensors to minimize effects on the muscle and connective tissues in response to implantation surgery. As a consequence, the present data do not allow us to draw conclusions regarding the interactions with MG.

2D data obtained from X-ray motion tracking and bone-derived markers (Bauman and Chang, 2010) yielded similar patterns of hindlimb joint angular displacements on the sagittal plane. However, a biplanar motion tracking system is preferable to describe joint flexion–extension angles in 3D space if out-of-plane motion is substantial. Comparison between 2D and 3D motion data of joint angular displacement revealed that large differences are found in the frontal and horizontal planes (Joãoa et al., 2010), while hip–knee–ankle flexion–extension during gait can be well approximated by motion on the sagittal plane.

The use of 2D instead of 3D motion data limits the accuracy of the kinematic descriptors and the estimates of MTU lengths presented here. Based on previously reported 3D kinematics of the rat hindlimb (Thota et al., 2005) collected during walking at 0.3 m s⁻¹ (nb. not trotting; much slower speed than in our experiment), we estimated a maximum projection error of 17.9 and 17.5 deg for knee and ankle joint angle changes, respectively. This discrepancy results in a maximum underestimation of 5.4% and 3.4% of the average SO and LG MTU length, respectively.

Sonomicrometry crystals were implanted to measure relative displacement between muscle bellies and differential muscle belly strain. Owing to the fact that LG is a pennated muscle, muscle belly length changes do not fully reflect the length changes of muscle fibers.

The assessment of distal tendon length changes is affected by out of plane (sagittal) movements of SO and LG muscle bellies. Because of projection errors, this would yield an overestimation of distal tendon lengths, especially during the swing phase and for LG muscle, because the sonomicrometry crystal distance vector may not be in the line of pull of the LG muscle belly and tendon. Also, 2D/3D motion data discrepancy (see above) may influence this estimate, as modeled MTU lengths were used. Therefore, the estimated tendon length changes during swing should be interpreted with caution. More research is required to obtain accurate predictions of tendon forces from estimated tendon lengths.

This study provides evidence of differential muscle activation and contractile behavior between rat plantar-flexors. We observed substantial longitudinal and transversal displacements between muscle bellies and subsequent deformation of the intermuscular volume. Such conditions may affect transmission of force between neighboring muscles via epimuscular myofascial pathways and/or intertendinous pathways, but knowledge about the stiffness of these pathways is required to predict the exact magnitude of the resultant mechanical interaction. Studying not only the contractile behavior of individual muscles, but also the multi-dimensional deformations between them, provides a more comprehensive view of the mechanical interactions within a synergistic muscle group and may be needed to fully understand neural control of movement.

The authors thank Milanka Visser, Guus Baan and Wendy Noort for their assistance with data collection. We thank Frans den Boer for the implantable connectors, as well as Bert Coolen for technical assistance with kinematics recordings. We further thank Maike Visschedijk and Julie Teeuwen for the motion tracking analysis.