Left–right coordination from simple to extreme conditions during split‐belt locomotion in the chronic spinal adult cat

Abstract

Key points

• Coordination between the left and right sides is essential for dynamic stability during locomotion.

• The immature or neonatal mammalian spinal cord can adjust to differences in speed between the left and right sides during split‐belt locomotion by taking more steps on the fast side.

• We show that the adult mammalian spinal cord can also adjust its output so that the fast side can take more steps.

• During split‐belt locomotion, only certain parts of the cycle are modified to adjust left–right coordination, primarily those associated with swing onset.

• When the fast limb takes more steps than the slow limb, strong left–right interactions persist.

• Therefore, the adult mammalian spinal cord has a remarkable adaptive capacity for left–right coordination, from simple to extreme conditions.

Abstract

Although left–right coordination is essential for locomotion, its control is poorly understood, particularly in adult mammals. To investigate the spinal control of left–right coordination, a spinal transection was performed in six adult cats that were then trained to recover hindlimb locomotion. Spinal cats performed tied‐belt locomotion from 0.1 to 1.0 m s⁻¹ and split‐belt locomotion with low to high (1:1.25–10) slow/fast speed ratios. With the left hindlimb stepping at 0.1 m s⁻¹ and the right hindlimb stepping from 0.2 to 1.0 m s⁻¹, 1:1, 1:2, 1:3, 1:4 and 1:5 left–right step relationships could appear. The appearance of 1:2+ relationships was not linearly dependent on the difference in speed between the slow and fast belts. The last step taken by the fast hindlimb displayed longer cycle, stance and swing durations and increased extensor activity, as the slow limb transitioned to swing. During split‐belt locomotion with 1:1, 1:2 and 1:3 relationships, the timing of stance onset of the fast limb relative to the slow limb and placement of both limbs at contact were invariant with increasing slow/fast speed ratios. In contrast, the timing of stance onset of the slow limb relative to the fast limb and the placement of both limbs at swing onset were modulated with slow/fast speed ratios. Thus, left–right coordination is adjusted by modifying specific parts of the cycle. Results highlight the remarkable adaptive capacity of the adult mammalian spinal cord, providing insight into spinal mechanisms and sensory signals regulating left–right coordination.

Key points

• Coordination between the left and right sides is essential for dynamic stability during locomotion.

• The immature or neonatal mammalian spinal cord can adjust to differences in speed between the left and right sides during split‐belt locomotion by taking more steps on the fast side.

• We show that the adult mammalian spinal cord can also adjust its output so that the fast side can take more steps.

• During split‐belt locomotion, only certain parts of the cycle are modified to adjust left–right coordination, primarily those associated with swing onset.

• When the fast limb takes more steps than the slow limb, strong left–right interactions persist.

• Therefore, the adult mammalian spinal cord has a remarkable adaptive capacity for left–right coordination, from simple to extreme conditions.

Abbreviations

CPG

central pattern generator

LG

lateral gastrocnemius

LH

left hindlimb

MG

medial gastrocnemius

RH

right hindlimb

Srt

sartorius

VL

vastus lateralis

Introduction

Proper coordination of the left and right sides is essential for locomotion, ensuring dynamic stability when modulating speed or when adjusting to task demands and environmental constraints. It is well established that the core control centres for locomotion are within the spinal cord, the so‐called central pattern generators (CPGs), and that each limb is controlled by its own CPG (reviewed by Kiehn, 2006, 2011; Frigon, 2012). There is compelling evidence that neuronal connections between the left and right sides (Butt & Kiehn, 2003; Lanuza et al. 2004; Zhong et al. 2006; Quinlan & Kiehn, 2007; Talpalar et al. 2013) and between cervical and lumbar levels (Juvin et al. 2005, 2012) coordinate the activities of spinal CPGs. The strength of neuronal connections between limbs is constantly regulated during locomotion by supraspinal signals, peripheral sensory feedback and intrinsic changes in spinal excitability (Rossignol et al. 2006). Despite progress made in understanding neuronal connectivity within the mammalian spinal cord, primarily via in vitro neonatal rodent preparations, the spinal control and adaptive capacity of left–right coordination during real locomotor behaviours, where biomechanical factors must be accounted for, remain poorly understood, particularly in adult systems.

The presence of locomotor CPGs within the lumbar spinal cord allows for the recovery of hindlimb locomotion after spinal transection in several adult mammalian species, including mice (Leblond et al. 2003), rats (Slawinska et al. 2012; Alluin et al. 2015) and cats (Lovely et al. 1986; Barbeau & Rossignol, 1987). These chronic spinal animal models are powerful tools to investigate the control of left–right coordination during real locomotor movements by spinal mechanisms interacting with sensory feedback from the periphery. In a landmark study, Forssberg et al. (1980) showed that kittens, spinal‐transected (spinalized) soon after birth, could adjust hindlimb locomotion when the left (LH) and right (RH) hindlimbs stepped at different speeds on a split‐belt treadmill. Split‐belt locomotion simulates many aspects of stepping along a curved trajectory where the inner limb(s) steps slower than the outer one(s) in cats and humans (Forssberg et al. 1980; Reisman et al. 2005; D'Angelo et al. 2014; Frigon et al. 2015). Studies in spinal kittens (Forssberg et al. 1980) and human infants (Thelen et al. 1987; Yang et al. 2004, 2005) have shown that when the speed ratio between LH and RH increases sufficiently, the fast limb can take more steps than the slow limb. In contrast, human adults maintain a 1:1 relationship in the number of steps taken between the left and right legs during split‐belt locomotion, even with high left–right speed ratios (Dietz et al. 1994; Reisman et al. 2005). Whether the adult mammalian spinal cord is capable of adjusting its output to produce different rhythms on the left and right sides while maintaining dynamic stability remains unclear.

We recently showed that chronic spinal adult cats could adjust their hindlimb pattern during split‐belt locomotion over a narrow range of left–right speed ratios (1.25–2.5) with the slow limb stepping at 0.4 m s⁻¹ and the fast limb stepping at speeds of up to 1.0 m s⁻¹ (Frigon et al. 2013). With the slow limb stepping at 0.4 m s⁻¹, a 1:1 relationship in left–right step number was maintained up to a speed of 1.0 m s⁻¹ (maximal speed tested) of the fast limb. The first goal of the present study was to assess the full adaptive potential for left–right coordination in the chronic spinal adult cat by evaluating hindlimb locomotion during split‐belt locomotion with low to high (1.25–10) left–right (slow–fast) speed ratios. The secondary aim was to characterize the kinematic and electromyographic adjustments that take place to produce stable locomotion to gain insight into the potential mechanisms regulating left–right coordination.

Methods

Ethical approval

All procedures were approved by the Animal Care Committee of the Université de Sherbrooke and were in accordance with policies and directives of the Canadian Council on Animal Care (Protocol number 362‐14). Six adult cats (two males, four females) weighing between 3.0 and 4.5 kg and all over 1 year of age were used. Before and after experiments, cats were housed and fed in a dedicated room within the animal care facility of the Faculty of Medicine and Health Sciences at the Université de Sherbrooke. As part of our effort to maximize the scientific output of each animal, cats were used in other studies to provide answers to other scientific questions (Frigon et al. 2013, 2014, 2015; D'Angelo et al. 2014; Dambreville et al. 2015; Hurteau et al. 2015). The experiments comply with the ARRIVE guidelines (Kilkenny et al. 2010) and principles of animal research established by the Journal of Physiology (Grundy, 2015).

Surgical procedures

The implantation and spinal transection (spinalization) surgeries were performed on separate days under aseptic conditions in an operating room with sterilized equipment. Prior to surgery, the cat was sedated with an intramuscular (i.m.) injection of butorphanol (0.4 mg kg⁻¹), acepromazine (0.1 mg kg⁻¹) and glycopyrrolate (0.01 mg kg⁻¹). Induction was done with ketamine/diazepam (0.11 ml kg⁻¹ in a 1:1 ratio, i.m.). The fur overlying the back, stomach and hindlimbs was shaved. The cat was then anaesthetized with isoflurane (1.5–3%) using a mask for a minimum of 5 min and then intubated with a flexible endotracheal tube. Isoflurane concentration was confirmed and adjusted throughout the surgery by monitoring cardiac and respiratory rates, by applying pressure to the paw to detect limb withdrawal and by assessing muscle tone. A rectal thermometer was used to monitor body temperature and keep it between 35 and 37°C using a water‐filled heating pad placed under the animal and an infrared lamp positioned ∼50 cm above the cat. In two cats, the implantation surgery was performed before spinalization while in the other four cats, it was made after cats had recovered hindlimb locomotion following spinalization. During surgery, an antibiotic (cefovecin, 0.1 ml kg⁻¹) was injected subcutaneously and a transdermal fentanyl patch (25 mg h⁻¹) was taped to the back of the animal 2–3 cm rostral to the base of the tail. During surgery and approximately 7 h later, another analgesic (buprenorphine 0.01 mg kg⁻¹) was administered subcutaneously. After surgery, cats were placed in an incubator and closely monitored until they regained consciousness. At the conclusion of the experiments, cats received a lethal dose of pentobarbital through the left or right cephalic vein.

Spinal transection

The spinal cord was completely transected at low thoracic levels. A small laminectomy was performed between the junction of the 12th and 13th vertebrae. After exposing the spinal cord, lidocaine (Xylocaine, 2%) was applied topically and injected within the spinal cord. The spinal cord was then transected with surgical scissors. Haemostatic material (Spongostan) was then inserted within the gap and muscles and skin were sewn back to close the opening in anatomical layers. Following spinalization and for the remainder of the study, the bladder was manually expressed 1–2 times each day. The hindlimbs were frequently cleaned by placing the lower half of the body in a warm soapy bath.

Implantation

Cats were implanted with electrodes to chronically record muscle activity (EMG). Pairs of Teflon insulated multistrain fine wires (AS633; Cooner wire, Chatsworth, CA, USA) were directed subcutaneously from one or two head‐mounted 34‐pin connectors (Omnetics Connector Corporation, Minneapolis, MN, USA) and sewn into the belly of selected hindlimb muscles for bipolar recordings. Electrode placement was verified by electrically stimulating each muscle through the appropriate head connector channel.

Locomotor training

One week after spinalization, cats were trained five times a week (20–30 min per session) to recover involuntary hindlimb locomotion. Early after spinalization, training consisted of two experimenters moving the hindlimbs over the moving treadmill belt to simulate locomotion with similar joint kinematics and paw contacts while the forelimbs were positioned on a fixed platform located ∼1 cm above the belt. Weight support and equilibrium were provided by an experimenter gently holding the base of the tail. A Plexiglas separator was placed between the hindlimbs to prevent them from impeding each other. After a few days, the skin of the perineal region was stimulated to evoke stepping movements. Experimenter‐assisted weight support was gradually reduced as extensor tonus recovered. Recording sessions started once the animals attained a stable locomotor pattern with full weight bearing and consistent plantar foot placement. One experimenter continued to provide lateral equilibrium by holding the tail.

Experimental paradigms

All experiments were performed on an animal treadmill with two independently controlled running surfaces 120 cm long and 30 cm wide (Bertec Corporation, Columbus, OH, USA). Cats performed three locomotor paradigms: (1) tied‐belt locomotion from 0.1 to 1.0 m s⁻¹ in 0.1 m s⁻¹ increments, (2) split‐belt locomotion with LH (slow side) stepping at 0.4 m s⁻¹ and RH (fast side) stepping from 0.5 to 1.0 m s⁻¹ in 0.1 m s⁻¹ increments and (3) split‐belt locomotion with LH (slow side) stepping at 0.1 m s⁻¹ and RH (fast side) stepping from 0.2 to 1.0 m s⁻¹ in 0.1 m s⁻¹ increments. Split‐belt locomotion with the slow side stepping at 0.4 m s⁻¹ was chosen because pilot studies determined that chronic spinal adult cats maintained a 1:1 relationship in cycle duration between LH and RH up to speeds of 1.0 m s⁻¹ on the fast side. With the slow hindlimb stepping at speeds below 0.4 m s⁻¹, a dissociation of the left–right rhythms was observed. We chose split‐belt locomotion with the slow side stepping at 0.1 m s⁻¹ to challenge left–right coordination up to a 10‐fold ratio in speed between the slow LH and fast RH. Experiments were also performed with RH as the slow side and the results were essentially a mirror image of those obtained with LH as the slow side. As such, only results with LH as the slow side are shown. In all paradigms, the forelimbs remained on a stationary platform with a Plexiglas separator placed between the hindlimbs.

Data acquisition and analysis

Videos of the left and right sides were captured with two cameras (Basler AcA640‐100g) at 60 frames per second with a spatial resolution of 640 by 480 pixels. A custom‐made Labview program acquired images and synchronized the cameras with the EMG. Videos were analysed off‐line at 60 frames per second using custom‐made software. Contact of the paw and its most caudal displacement were determined for both hindlimbs by visual inspection. Paw contact was defined as the first frame where the paw made visible contact with the treadmill surface while the most caudal displacement of the limb was the frame with the most caudal displacement of the toe. Cycle, stance and swing durations were measured for each hindlimb. Cycle duration was measured from successive paw contacts while stance duration corresponded to the interval of time from paw contact to the most caudal displacement of the limb. Swing duration was measured as cycle duration minus stance duration. Durations from 10–15 cycles for each limb were averaged for an episode. The phasing between contacts was calculated as the interval of time between LH and RH contacts divided by LH cycle duration. When RH performed more steps than LH during split‐belt locomotion with LH stepping at 0.1 m s⁻¹, the phase interval between contacts of each RH step was measured.

EMG was pre‐amplified (×10, custom‐made system), bandpass filtered (30–1000 Hz) and amplified (×100–5000) using a 16‐channel amplifier (AM Systems Model 3500, Sequim, WA, USA). EMG data were digitized (2000 Hz) with a National Instruments card (NI 6032E) and acquired with custom‐made acquisition software and stored on computer. The current EMG data set was obtained from bilateral recordings in the anterior sartorius (Srt, hip flexor/knee extensor), the vastus lateralis (VL, knee extensor), the medial gastrocnemius (MG, ankle plantarflexor/knee flexor) and/or the lateral gastrocnemius (LG, ankle plantarflexor/knee flexor). The extensor muscle (VL, MG or LG) with the best signal‐to‐noise ratio was retained for analysis. Due to the loss of some EMG signals over time, we did not have a reliable left Srt in one cat and a right extensor in another cat. Burst onsets and offsets were determined by visual inspection by the same experimenter (A.F.) from the raw EMG waveforms using a custom‐made program. Burst duration was measured from onset to offset while mean EMG amplitude was measured by integrating the full‐wave rectified EMG burst from onset to offset and dividing it by its burst duration.

Statistical analyses

Statistical tests were performed with IBM SPSS Statistics 19.0 (IBM Corporation, Armonk, NY, USA). A two‐factor (side × speed) repeated measures ANOVA was performed on cycle, phase and burst durations during tied‐belt locomotion. A one‐factor (RH speed) repeated measures ANOVA was performed on cycle, phase and burst durations during split‐belt locomotion with LH at 0.1 and 0.4 m s⁻¹. A one‐factor (speed or LH–RH speed ratios) repeated measures ANOVA was also performed on stance phase onsets and offsets during tied‐belt locomotion and split‐belt locomotion with LH at 0.1 and 0.4 m s⁻¹. Paired t‐tests were used to assess adjustments in RH cycle and phase durations between RH steps with 1:2 and 1:3 relationships in the number of steps taken by LH and RH during split‐belt locomotion with LH at 0.1 m s⁻¹. Group data in the graphs are the mean ± the standard deviation. Significance level was set at P < 0.05.

Results

The main objective of the present study was to characterize adjustments in the hindlimb pattern during split‐belt locomotion with slow–fast speed ratios ranging from 1:1.25 to 10. As a basis for comparison, the results first describe the adjustments observed during tied‐belt locomotion followed by those occurring during split‐belt locomotion with the slow limb stepping at 0.4 m s⁻¹ and then at 0.1 m s⁻¹. All cats recovered full weight‐bearing hindlimb locomotion with treadmill training 6–10 weeks after spinalization and could perform the three experimental paradigms.

Tied‐belt locomotion

Kinematic and EMG recordings were made in chronic spinal adult cats during tied‐belt locomotion from 0.1 to 1.0 m s⁻¹. Although kinematic or EMG adjustments to speed have been well documented in intact cats during locomotion (Frigon et al. 2013, 2014; Goslow et al. 1973; Halbertsma, 1983), it has rarely been the main focus in spinal animals (but see Frigon et al. 2013; Dambreville et al. 2015). Here, we show adjustments to treadmill speed in chronic spinal cats over a 10‐fold range. It should be noted that adjustments to speed during tied‐belt locomotion are similar to single‐belt locomotion (Frigon et al. 2014).

Figure 1 shows the hindlimb pattern during tied‐belt locomotion at three speeds (0.1, 0.4 and 1.0 m s⁻¹) in one chronic spinal adult cat. These speeds were chosen for representation because 0.1 and 0.4 m s⁻¹ are the speeds of the slow belt during split‐belt locomotion while 1.0 m s⁻¹ is the fastest treadmill speed tested. The left panels show the EMGs of five hindlimb muscles along with the stance phases while the right panels show diagrams of the hindlimbs at contact and swing onset for the middle LH cycle shown in the left panel. As expected, stance phase and extensor burst durations decreased bilaterally with increasing treadmill speed while swing phase and Srt burst durations remained relatively invariant (compare Fig. 1 A, B and C, noting the difference in timescale). The time scales are different to clearly show the adjustments that occur within a cycle in the different tied‐belt and split‐belt conditions. Stance durations were approximately of equal duration on the left and right sides, particularly at higher speeds, demonstrating left–right symmetry during tied‐belt locomotion with increasing speed. The limb diagrams show that the distance of the paw at contact relative to the hip remained approximately the same with increasing speed while it was progressively more caudal at swing onset.

Figure 1. Hindlimb muscle activity and phase durations during tied‐belt locomotion in one spinal cat.

The pattern is shown during tied‐belt locomotion at (A) 0.1 m s⁻¹, (B) 0.4 m s⁻¹ and (C) 1.0 m s⁻¹. The black horizontal bars at the bottom of each panel on the left show left (LSTA) and right (RSTA) stance phase durations. The limb diagrams on the right show the positions of the hindlimbs at phase transitions: left (LHC) and right (RHC) hindlimb contacts and at left (LHSO) and right (RHSO) swing onsets. The dotted vertical line from the hip marker is shown for reference. Data shown are from cat BL. L, left; LG, lateral gastrocnemius; R, right; Srt, sartorius; STA, stance; VL, vastus lateralis.

It has been well established that cycle duration is reduced with increasing speed and that this is mostly due to a reduction in the duration of the stance/extension phase while the swing/flexion phase remains relatively invariant (reviewed by Gossard et al. 2011). Figure 2 shows cycle, phase and burst durations in LH and RH during tied‐belt locomotion from 0.1 to 1.0 m s⁻¹ in 0.1 m s⁻¹ increments for the group. As can be observed, cycle (P < 0.001) and stance (P < 0.001) durations were significantly reduced with increasing speed while swing duration remained unchanged (P > 0.05) (Fig. 2 A, B). In a similar vein, extensor burst duration decreased significantly (P < 0.001) with increasing speed while the duration of the hip flexor (Srt) burst was unaffected (P > 0.05) (Fig. 2 C, D). There was no significant effect of side (LH versus RH) on cycle, stance and swing durations or on extensor and Srt burst durations (P > 0.05). Thus, similar to intact animals (Halbertsma, 1983; Frigon et al. 2013, 2014), a change in cycle duration in spinal cats is accompanied by a change in the stance or extension phase over a wide range of treadmill speeds.

Figure 2. Cycle, phase and burst durations during tied‐belt locomotion across cats.

Cycle and phase durations are shown for (A) the left hindlimb (LH) and (B) the right hindlimb (RH) for the group (n = 6 cats). Burst durations are shown for (C) an LH extensor (LH ext, n = 6) and an LH flexor (LH flex, n = 5) and for (D) an RH extensor (RH ext, n = 5) and an RH flexor (RH flex, n = 6). The extensor burst was obtained from the lateral or medial gastrocnemii or vastus lateralis while the flexor burst was obtained from the anterior sartorius. Note that due to the loss of some EMG signals, LH Srt and an RH ext are missing in one cat each. At every speed, 10–15 cycles were averaged for each cat. Individual cat averages were then averaged for the group. Each data point is the mean ± SD.

During treadmill or overground locomotion from slow to moderately fast speeds in intact cats, homologous limbs (forelimbs or hindlimbs) are almost perfectly out‐of‐phase, meaning that contact of one limb occurs at 50% of the cycle of the contralateral homologous limb (English & Lennard, 1982; Frigon et al. 2014). In our chronic spinal cats, the phasing between the left and right hindlimbs was also maintained around 0.5 during tied‐belt locomotion from 0.1 to 1.0 m s⁻¹, showing that contact of the right hindlimb occurred at ∼50% of the cycle of the left hindlimb (Fig. 3 A). With increasing speed, stance offset (or swing onset) occurred significantly earlier (P < 0.001) for both hindlimbs, thus reducing periods of double support (Fig. 3 A). In a recent study in chronic spinal cats, we showed that the horizontal distance of the paw at contact relative to the hip remained the same with an increase in speed from 0.1 to 0.4 m s⁻¹ during tied‐belt locomotion, while the distance at swing onset progressively increased (Dambreville et al. 2015). Here, we show similar results up to a speed of 1.0 m s⁻¹. With increasing speed, the distance of the paw at stance onset remained unchanged (P > 0.05) while the distance at swing onset was progressively more caudal for the group (P < 0.001) (Fig. 3 B). Thus, the timing of stance onsets and paw placements are invariant to a change in speed while the timing of swing onset and the distance of the paw at swing onset are modified.

Figure 3. The bilateral phasing of stance phases and the distance of the paw at stance and swing onsets during tied‐belt locomotion across cats.

A, stance onsets and offsets for the left (LH) and right (RH) hindlimbs were normalized to LH cycle duration. The dotted vertical line represents the end of the normalized LH cycle and is shown for reference. B, the horizontal distance between the paw and the hip marker was measured at stance (values at the right of the bars) and swing (values at the left of the bars) onsets. The vertical dotted line at a value of 0 represents the point directly beneath the hip marker. Each bar in A and B shows the mean ± SD.

Split‐belt locomotion with the slow limb stepping at 0.4 m s⁻¹

Split‐belt locomotion inevitably generates an asymmetry between the left and right sides as one limb is stepping faster than the other, thus requiring bilateral adjustments in the locomotor pattern. This section describes temporal and spatial adjustments in the hindlimb pattern during split‐belt locomotion with LH (the slow limb) stepping at 0.4 m s⁻¹ and RH (the fast limb) stepping at speeds of 0.5–1.0 m s⁻¹. A speed of 0.4 m s⁻¹ was chosen for the slow limb because, at all slow–fast speed ratios tested, a 1:1 relationship in the number of steps taken by LH and RH was maintained. It is important to note that the adjustments to split‐belt locomotion were observed immediately upon exposure to the differences in speed between LH and RH. In other words, there were no step‐by‐step adjustments in the first few steps required to reach a steady‐state. Figure 4 shows data from the same cat as in Fig. 1 and is organized in the same way. In contrast to tied‐belt locomotion, clear left–right asymmetries appeared in the pattern during split‐belt locomotion. The duration of the stance phase and extensor burst in LH (the slow limb) was longer than in RH (the fast limb) and this asymmetry became greater with an increase in the L‐R speed ratio (compare Fig. 4 A and B). In contrast, the duration of the swing phase and of the flexor (Srt) burst was longer in RH than in LH. The limb diagrams show that the distance of the paw at contact relative to the hip remained approximately the same for both LH and RH while the distance of the paw at swing onset was more caudal for RH compared to LH.

Figure 4. Hindlimb muscle activity and phase durations during split‐belt locomotion with the slow limb stepping at 0.4 m s⁻¹ in one spinal cat.

The pattern is shown during split‐belt locomotion with the left hindlimb stepping at 0.4 m s⁻¹ and the right hindlimb stepping at (A) 0.5 m s⁻¹ and (B) 1.0 m s⁻¹. The black horizontal bars at the bottom of each panel show left (LSTA) and right (RSTA) stance phase durations. The limb diagrams on the right show the positions of the hindlimbs at phase transitions: left (LHC) and right (RHC) hindlimb contacts and at left (LHSO) and right (RHSO) swing onsets. The dotted vertical line from the hip marker is shown for reference. Data shown are from cat BL. L, left; LG, lateral gastrocnemius; R, right; Srt, sartorius; STA, stance; VL, vastus lateralis.

To equalize cycle duration bilaterally during split‐belt locomotion, stance and swing phase durations are differentially modified in the slow and fast limbs (Kulagin & Shik, 1970; Forssberg et al. 1980; Dietz et al. 1994; Frigon et al. 2013; D'Angelo et al. 2014). We recently showed adjustments in cycle and phase durations in the slow and fast limbs during quadrupedal split‐belt locomotion in intact adult cats with the slow side stepping at 0.4 m s⁻¹ and the fast side stepping at speeds of 0.5–1.0 m s⁻¹ (D'Angelo et al. 2014). Here, we show adjustments in cycle and phase durations in chronic spinal adult cats over the same slow–fast speed ratios. Interestingly, LH and RH cycle durations were not significantly affected (P > 0.05) by increasing RH speed, despite an increase in mean treadmill speed (Fig. 5 A, B). The lack of change in cycle duration with an increase in left–right speed ratio was due to the fact that LH stance duration did not change significantly (P > 0.05), despite a small but significant decrease (P < 0.01) in LH swing duration (Fig. 5 A). For RH, the fast limb, stance duration was shorter than on the left side and despite a significant decrease (P < 0.001) with increasing RH speed, it was insufficient to significantly affect cycle duration (Fig. 5 B). Moreover, the decrease in RH stance duration was counterbalanced by a significant increase (P < 0.05) in RH swing duration, which occupied a progressively greater proportion of the cycle with increasing RH speed. Extensor and flexor burst durations for the slow and fast limbs followed changes in kinematic phase durations. LH extensor burst duration was not affected (P > 0.05) by increasing RH speed while Srt burst duration showed a small but significant decrease (P < 0.05) (Fig. 5 C). For RH, extensor burst duration decreased significantly (P < 0.01) with increasing RH speed and although Srt burst duration was not significantly affected (P > 0.05), there was a trend showing an increase (Fig. 5 D).

Figure 5. Cycle, phase and burst durations during split‐belt locomotion with the slow limb stepping at 0.4 m s⁻¹ across cats.

In the panels shown above, the left hindlimb (LH) was stepping at 0.4 m s⁻¹ while the right hindlimb (RH) stepped from 0.5 to 1.0 m s⁻¹ in 0.1 m s⁻¹ increments. Cycle and phase durations are shown for (A) LH and (B) RH for the group (n = 6 cats). Burst durations are shown for (C) an LH extensor (LH ext, n = 6) and an LH flexor (LH flex, n = 5) and for (D) an RH extensor (RH ext, n = 5) and an RH flexor (RH flex, n = 6). The extensor burst was obtained from the lateral or medial gastrocnemii or vastus lateralis while the flexor burst was obtained from the anterior sartorius. Note that due to the loss of some EMG signals, LH Srt and an RH ext are missing in one cat each. Also, one cat could not perform split‐belt locomotion with RH stepping at 0.9 and 1.0 m s⁻¹. At every RH speed, 10–15 cycles were averaged for each cat. Individual cat averages were then averaged for the group. Each data point is the mean ± SD.

As shown above, the hindlimbs maintained a strict out‐of‐phase alternation during tied‐belt locomotion with increasing speed. Moreover, the placement of the paw at contact was invariant while it was progressively more caudal at swing onset with increasing speed. Despite the fact that split‐belt locomotion prolonged and shortened stance duration in the slow and fast hindlimbs, respectively, a strict out‐of‐phase alternation between limbs was maintained. During split‐belt locomotion with LH at 0.4 m s⁻¹, stance onset for RH was around 0.5 (50% of the cycle), although there was a small but significant increase (P < 0.01) with increasing RH speed (from 0.48 to 0.54 at RH speeds of 0.5 and 1.0 m s⁻¹, respectively), indicating that RH stance onset occurred slightly later within the LH cycle with increasing RH speed (Fig. 6 A). Due to a longer LH stance duration, and a significant increase (P < 0.01) in LH stance offset with increasing RH speed, there was a long period (∼20% of the cycle) of double support from RH contact to LH swing onset. On the other hand, the progressively earlier (P < 0.01) offset of the RH stance phase with increasing RH speed reduced the period of double support from RH stance offset to LH stance onset, even producing a period with no support, or an aerial period, at an RH speed of 1.0 m s⁻¹ (see also Fig. 4 B). Thus, split‐belt locomotion with a 1:1 relationship produced differential changes in the relative phasing between limbs at the stance‐to‐swing and swing‐to‐stance transitions for the slow and fast limbs. As shown above, increasing speed during tied‐belt locomotion only affected the placement of the paw at swing onset and not at contact. During split‐belt locomotion with the slow limb stepping at 0.4 m s⁻¹, the distances of the paw at both stance and swing onsets were not significantly affected by RH speed (P > 0.05) (Fig. 6 B). However, although there was no significant difference in the distance at paw contact between limbs (P > 0.05), there was a significant difference at swing onset (P < 0.01), with the fast limb transitioning to swing at a more caudal position compared to the slow limb at every slow–fast speed ratio. Thus, during split‐belt locomotion with a 1:1 relationship, only certain temporal and spatial aspects are modified to adjust left–right coordination.

Figure 6. The bilateral phasing of stance phases and the distance of the paw at stance and swing onsets during split‐belt locomotion with the slow limb stepping at 0.4 m s⁻¹ across spinal cats.

A, stance onsets and offsets for the left (LH) and right (RH) hindlimbs were normalized to LH cycle duration. The dotted vertical line represents the end of the normalized LH cycle and is shown for reference. B, the horizontal distance between the paw and the hip marker was measured at stance (values at the right of the bars) and swing (values at the left of the bars) onsets. The vertical dotted line at a value of 0 represents the point directly beneath the hip marker. Each bar in A and B shows the mean ± SD.

Split‐belt locomotion with the slow limb stepping at 0.1 m s⁻¹

Previous studies in spinal kittens (Forssberg et al. 1980) and human infants (Yang et al. 2005) have shown that the fast limb can take two or more steps for every step of the slow limb during split‐belt locomotion when the slow–fast speed ratio is large enough. Here, we show that this can also occur in chronic spinal adult cats. This section describes temporal and spatial adjustments in the hindlimb pattern during split‐belt locomotion with LH (the slow limb) stepping at 0.1 m s⁻¹ and RH (the fast limb) stepping at speeds of 0.2–1.0 m s⁻¹.

Figure 7 shows data from the same cat as in Figs 1 and 4 and is organized in the same way. In Fig. 7 A and B, 1:2 and 1:3 relationships in the number of steps taken by LH and RH are shown with RH speeds of 0.4 and 0.7 m s⁻¹, respectively. The stance phase of the slow limb, LH, occupied the majority (> 80%) of its cycle, allowing the fast limb, RH, to take two or three steps. As can be observed, each time RH made contact, LH extensor burst activity was reduced and increased again when RH transitioned to swing. In other words, as RH extensor burst increased and the limb supported some of the weight, LH extensor burst activity decreased. The last RH cycle, or the RH cycle when LH transitioned to swing, was always the longest RH cycle, with greater extensor burst activity than previous RH cycles. Figure 7 C shows a closer view of the middle LH cycle of Fig. 7 B to better illustrate that RH extensor burst activity started before RH made contact and that this pre‐contact activity was greater in the last RH cycle. Thus, the central drive to RH extensors is already preparing for the stance‐to‐swing transition of LH. In contrast, the multiple ‘sub‐bursts’ shown during the long LH extensor burst decreased progressively as LH was preparing to eventually transition to swing.

Figure 7. Hindlimb muscle activity and phase durations during split‐belt locomotion with the slow limb stepping at 0.1 m s⁻¹ in one spinal cat.

The pattern is shown during split‐belt locomotion with the left hindlimb stepping at 0.1 m s⁻¹ and the right hindlimb stepping at (A) 0.4 m s⁻¹ and (B) 0.7 m s⁻¹. The black horizontal bars at the bottom of each panel show left (LSTA) and right (RSTA) stance phase durations. C, an expanded view of the middle left hindlimb cycle shown in B. The limb diagrams on the right show the positions of the hindlimbs at phase transitions: left (LHC) and right (RHC) hindlimb contacts and at left (LHSO) and right (RHSO) swing onsets. The dotted vertical line from the hip marker is shown for reference. Data shown are from cat BL. L, left; LG, lateral gastrocnemius; R, right; Srt, sartorius; STA, stance; VL, vastus lateralis.

When LH, the slow limb, stepped at a speed of 0.1 m s⁻¹ with RH, the fast limb, stepping at speeds of 0.2–1.0 m s⁻¹, 1:1, 1:2, 1:3, 1:4 and 1:5 relationships in the number of steps taken by LH and RH could be observed, indicating that the fast limb could take two, three, four or five steps for every step of the slow limb. An unequal number of steps between LH and RH was observed in five cats while one cat maintained a 1:1 relationship over the entire range of slow–fast speed ratios. The appearance of two or more RH steps for every LH step was not linearly related with slow–fast speed ratio (Fig. 8). Relationships of 1:1 and 1:2 predominated at the lowest RH speeds of 0.2 and 0.3 m s⁻¹. At RH speeds of 0.4 and 0.5 m s⁻¹, a 1:2 relationship dominated. At RH speeds of 0.6–1.0 m s⁻¹, 1:1, 1:2 and 1:3 relationships were found in approximately equal proportion with the appearance of 1:4 relationships. Two 1:5 relationships were observed in two different cats with RH stepping at 0.9 or 1.0 m s⁻¹ (data not shown). Our findings are similar to those found in human infants where 1:2 and 1:3 relationships were common while 1:4 and 1:5 relationships were occasionally observed (Yang et al. 2005).

Figure 8. Percentage of cycles with 1:1, 1:2, 1:3 and 1:4 relationships in the number of steps taken by the slow and fast limbs during split‐belt locomotion with the slow limb stepping at 0.1 m s⁻¹ across spinal cats.

In the graph, the left hindlimb (LH, slow limb) stepped at 0.1 m s⁻¹ while the right hindlimb (RH, fast limb) stepped from 0.2 to 1.0 m s⁻¹ in 0.1 m s⁻¹ increments. The number of LH cycles at each RH speed varied from 63 to 81 cycles with a mean of 76.2 ± 6.5 cycles.

To determine if these relationships were stable, we counted the number of consecutive LH cycles with the same type of relationship. A stable relationship was defined as one that showed three or more consecutive cycles of the same type. Table 1 shows the number of stable 1:1, 1:2, 1:3 and 1:4 relationships as a function of RH treadmill speed with LH stepping at 0.1 m s⁻¹. As can be observed, stable 1:1, 1:3 and 1:4 relationships were rare while stable 1:2 relationships were more frequent.

Table 1.

Number of stable 1:1, 1:2, 1:3 and 1:4 relationships

RH speed (m s–1)	Stable 1:1	Stable 1:2	Stable 1:3	Stable 1:4
0.2	2 in 2 cats	4 in 4 cats	–	–
0.3	2 in 2 cats	9 in 5 cats	–	–
0.4	1 in 1 cat	9 in 5 cats	–	–
0.5	1 in 1 cat	7 in 5 cats	1 in 1 cat	–
0.6	1 in 1 cat	5 in 3 cats	3 in 2 cats	–
0.7	1 in 1 cat	2 in 2 cats	5 in 3 cats	–
0.8	1 in 1 cat	2 in 2 cats	4 in 3 cats	1 in 1 cat
0.9	2 in 2 cats	3 in 2 cats	4 in 3 cats	–
1.0	1 in 1 cat	7 in 4 cats	3 in 2 cats	–

The appearance of additional steps taken by the fast limb for every step of the slow limb indicates a difference in cycle duration between the slow and fast sides. To maintain stable locomotion, this requires temporal adjustments in both the slow and the fast limbs. To determine how the cycles were reorganized bilaterally, we measured cycle and phase durations in LH and RH with 1:1, 1:2 and 1:3 relationships, respectively (Fig. 9). The numbers at the top of the upper panels indicate the number of cats that were included in the average at each RH speed. Due to the fact that not all cats displayed 1:1, 1:2 and 1:3 relationships at all RH speeds, a one‐factor (RH speed) repeated measures ANOVA was performed for cycle and phase durations at six speeds for 1:2 (0.2, 0.4, 0.5, 0.7, 0.9 and 1.0 m s⁻¹) and 1:3 (0.5, 0.6, 0.7, 0.8, 0.9 and 1.0 m s⁻¹) relationships where the same four cats could be paired for statistical analysis. A sufficient number of cats could not be paired with a 1:1 relationship so no statistical tests were performed. Despite these limitations, some interesting observations emerged. For instance, a consistent finding was the prolongation of the cycle and stance durations in the slow limb with 1:2 and 1:3 relationships to enable the fast limb to take additional steps. Indeed, LH cycle and stance durations were shortest at all RH speeds with a 1:1 relationship and increased with 1:2 and 1:3 relationships while swing duration changed only slightly (compare Fig. 9 A, B and C). Despite the appearance of 1:2 and 1:3 relationships, cycle and phase durations of the slow limb were reduced by increasing the speed of the limb on the faster belt, indicating that the fast limb still influenced the slow limb. LH cycle, stance and swing durations were significantly reduced with increasing RH speed with 1:2 (Fig. 9 B, P < 0.001) and 1:3 (Fig. 9 C, P < 0.05) relationships. For the fast limb, cycle and phase durations were also modulated with speed with one notable difference compared to the slow limb. In general, RH cycle and stance durations tended to decrease with increasing RH speed in the single step of a 1:1 relationship (Fig. 9 D), the first (Fig. 9 E) and second (Fig. 9 G) steps of a 1:2 relationship, and in the first (Fig. 9 F), second (Fig. 9 H) and third (Fig. 9 I) steps of a 1:3 relationship. On the other hand, swing duration tended to increase. Reductions in cycle and stance durations with increasing RH speed were significant for the first (Fig. 9 E, P < 0.01) and second (Fig. 9 G, P < 0.001) steps of a 1:2 relationship and the second step of a 1:3 relationship (Fig. 9 H, P < 0.05). Although stance duration was significantly reduced with increasing RH speed in the third step of a 1:3 relationship (Fig. 9 I, P < 0.001), cycle duration was unaffected due to one cat showing a prolonged swing phase at an RH speed of 0.8 m s⁻¹. Therefore, to adjust to large slow–fast speed ratios and the appearance of extra steps for the fast limb, cycle and stance durations of the slow limb were increased while swing duration remained relatively unchanged.

Figure 9. Cycle and phase durations during split‐belt locomotion with the slow limb stepping at 0.1 m s⁻¹ across cats.

The left hindlimb (LH) stepped at 0.1 m s⁻¹ while the right hindlimb (RH) stepped from 0.2 to 1.0 m s⁻¹ in 0.1 m s⁻¹ increments. Cycle and phase durations are shown for 1:1 (left panels), 1:2 (middle panels) and 1:3 (right panels) relationships in the number of steps taken by LH (slow) and RH (fast). Cycle and phase durations are shown in A–C for LH, in D–F for the first RH step, in G and H for the second RH step and in I for the third RH step. The numbers at the top of the upper panels indicate the number of cats that were included in the average at each RH speed. Each data point is the mean of available cats ± SD.

To better characterize adjustments in the fast limb when it performed two or three steps for every step of the slow limb, RH cycle, stance and swing durations of the second or third steps were expressed as a percentage of RH cycle duration of the first step at four (0.2, 0.4, 0.6 and 1.0 m s⁻¹) and three (0.6, 0.8 and 1.0 m s⁻¹) RH speeds with 1:2 and 1:3 relationships, respectively. In 1:2 relationships, RH cycle (Fig. 10 A), stance (Fig. 10 B) and swing (Fig. 10 C) durations of the second step were significantly longer than the first. In 1:3 relationships, RH cycle (Fig. 10 D), stance (Fig. 10 E) and swing (Fig. 10 F) durations of the third step were significantly longer than in the first and second steps. Thus, the last cycle in 1:2 or 1:3 relationships is longer due to an increase in the duration of both the stance and the swing phases.

Figure 10. Relative durations of each step of the fast limb in 1:2 and 1:3 relationships during split‐belt locomotion with the slow limb stepping at 0.1 m s⁻¹ across spinal cats.

Right hindlimb (RH) cycle (left panels), stance (middle panels) and swing (right panels) durations expressed as a percentage of the first RH cycle duration for (A–C) 1:2 and (D–F) 1:3 relationships in the number of steps taken by LH (slow) and RH (fast) at four (0.2, 0.4, 0.6 and 1.0 m s⁻¹) and three (0.6, 0.8 and 1.0 m s⁻¹) RH speeds, respectively. These RH speeds were chosen because four cats could be paired for statistical analysis. ^*** P < 0.001, ^** P < 0.01 and ^* P < 0.05 (paired t‐tests).

Although we did not have a sufficient sample size to conduct statistical tests on EMG bursts (n = 2–3 cats per RH speed), RH extensor burst duration followed changes in RH stance durations with longer bursts in the last RH cycle of 1:2 (Fig. 11 A) and 1:3 (Fig. 11 B) relationships. Additionally, the mean amplitude of the RH extensor burst was greater in the last RH cycle of 1:2 (Fig. 11 C) and 1:3 (Fig. 11 D) relationships. This greater extensor activity in the last step of the fast limb is necessary to allow the slow limb to transition to its swing phase.

Figure 11. Extensor burst durations and amplitudes for each step of the fast limb in 1:2 and 1:3 relationships during split‐belt locomotion with the slow limb stepping at 0.1 m s⁻¹ across spinal cats.

Burst durations of the RH extensor (RH ext) expressed as a percentage of the first RH cycle duration are shown for (A) 1:2 and (B) 1:3 relationships in the number of steps taken by LH (slow) and RH (fast) at four (0.2, 0.4, 0.6 and 1.0 m s⁻¹) and three (0.6, 0.8 and 1.0 m s⁻¹) RH speeds, respectively. Mean amplitudes of RH ext expressed as a percentage of the mean RH ext amplitude of the 1st step are shown for (C) 1:2 and (D) 1:3 relationships. Statistical tests were not performed on burst durations and mean amplitudes due to an insufficient number of muscles per RH speed. However, the trends are consistent across RH speeds.

The temporal phasing between hindlimbs becomes increasingly important to maintain dynamic stability when the fast limb starts to take more steps than the slow limb. To determine how the two hindlimbs remained temporally coordinated when the fast limb performed two or three steps for every step of the slow limb, LH and RH stance onsets and offsets were measured and normalized to LH cycle duration with 1:1, 1:2 and 1:3 relationships. As stated earlier, due to the fact that not all cats displayed 1:1, 1:2 and 1:3 relationships at all RH speeds with LH stepping at 0.1 m s⁻¹, to determine the effect of slow–fast speed ratios on stance onsets and offsets, a one‐factor (RH speed) repeated measures ANOVA was only performed at six RH speeds for 1:2 (0.2, 0.4, 0.5, 0.7, 0.9 and 1.0 m s⁻¹) and 1:3 (0.5, 0.6, 0.7, 0.8, 0.9 and 1.0 m s⁻¹) relationships. Although no statistical tests were performed for a 1:1 relationship (Fig. 12 A), LH stance always terminated around 0.8 or 80% of LH cycle. RH stance began well before LH stance offset, providing a long period of double support. On the other hand, RH stance offset occurred around LH contact and this period of double support from stance offset of the fast limb to stance onset of the slow limb was reduced with increasing RH speed. With 1:2 and 1:3 relationships, the pattern of RH stance offsets was similar but with one and two interposed RH steps, respectively (Fig. 12 B, C). Moreover, LH stance was proportionally longer, ending at 86–91% of the LH cycle for 1:2 and 1:3 relationships. The small variability between animals in stance onsets and offsets, as indicated by the error bars, suggests that coordination between the slow and fast limbs is tightly coupled when one limb takes more steps than the other.

Figure 12. The bilateral phasing of stance phases during split‐belt locomotion with the slow limb stepping at 0.1 m s⁻¹ across spinal cats.

Stance onsets and offsets for the left (LH) and right (RH) hindlimbs were normalized to LH cycle duration and are shown for (A) 1:1, (B) 1:2 and (C) 1:3 relationships in the number of steps taken by the slow and fast limbs. The dotted vertical line represents the end of the normalized LH cycle and is shown for reference. Each bar in A–C shows the mean ± SD.

The distance of the paw at stance onset for both LH, the slow limb, and RH, the fast limb, was not significantly affected by RH speed (P > 0.05) and did not differ between limbs (P > 0.05) (Fig. 13). In contrast, there was a clear difference for the distance of the paw at swing onset between the slow and fast limbs and between the first and last steps of the fast limb. LH, the slow limb, was more rostral at swing onset than RH, the fast limb, with the asymmetry between the slow and fast limbs increasing with RH speed. For RH, the position of the paw at swing onset was more caudal for the first step compared to the next steps of 1:2 and 1:3 relationships. This is because the first RH swing onset (see 1^st RHSO in Fig. 7) is the RH cycle where the slow limb is making its initial contact and starting its cycle. In temporal terms, the first RH swing onset is within the last RH cycle (determined from successive RH contacts) of 1:2 and 1: 3 relationships, which is characterized by a longer stance phase, allowing the fast limb to extend more caudally than successive steps. Thus, during split‐belt locomotion with a 1:2 and 1:3 relationships, only certain temporal and spatial aspects are modified to adjust left–right coordination, primarily those associated with swing onset.

Figure 13. The distance of the paw at stance and swing onsets during split‐belt locomotion with the slow limb stepping at 0.1 m s⁻¹ across spinal cats.

The horizontal distance between the paw and the hip marker was measured at stance (values at the right of the bars) and swing (values at the left of the bars) onsets and are shown for (A) 1:1, (B) 1:2 and (C) 1:3 relationships in the number of steps taken by the slow and fast limbs. The vertical dotted line at a value of 0 represents the point directly beneath the hip marker. Each bar in A–C shows the mean ± SD.

Discussion

The main objective of this study was to characterize changes in the hindlimb pattern of spinal adult cats during split‐belt locomotion with low to high ratios in the difference in speed between the slow and fast hindlimbs to gain insight into the control of left–right coordination by the adult mammalian spinal cord. The results show that chronic spinal adult cats can adjust their hindlimb pattern in locomotor conditions where left–right coordination is challenged, from simple to extreme conditions, producing different rhythms for the slow and fast sides while maintaining stable locomotion. The following sections discuss how these adjustments take place and the potential mechanisms involved.

Reorganization of the hindlimb pattern during split‐belt locomotion

Adjusting the pattern bilaterally is required when stepping on a curved trajectory and split‐belt locomotion can reproduce many aspects of this phenomenon. In the present study, when a 1:1 coupling between the slow and fast limbs was maintained during split‐belt locomotion in chronic spinal adult cats, the cycle and its phases were reorganized in a way similar to spinal kittens (Forssberg et al. 1980) and human infants (Yang et al. 2005), indicating that the adult mammalian spinal cord has the required mechanisms to mediate these adjustments. Specifically, stance duration increased and decreased for the slow and fast limbs while the swing phase showed an inverse pattern of change. However, there were differences in the adjustments that took place when compared to intact adult cats with a speed of 0.4 m s⁻¹ for the slow side. In our spinal cats, cycle duration and stance duration of the slow limb were not significantly affected by increasing mean treadmill speed (the average speed of the slow and fast belts) (Fig. 5). In contrast, cycle durations and stance durations of the slow limb were significantly decreased (P < 0.001) by increasing mean treadmill speed during split‐belt locomotion at the same slow–fast speed ratios in six intact adult cats (D'Angelo et al. 2014). There are several explanations for the differences in spinal and intact cats. In spinal cats, the swing phase occupies a greater proportion of the cycle at all slow–fast speed ratios, particularly in the fast limb (compare our Fig. 5 with fig. 3 of D'Angelo et al. 2014). As such, changes in stance duration have relatively less effect on cycle duration because the swing phase occupies a greater proportion of the cycle. Although this could be due to differences in supraspinal influences between the spinal and intact preparations, another explanation relates to the nature of hindlimb‐only locomotion, as the swing phase occupies a greater proportion of the cycle at a given speed during hindlimb‐only locomotion compared to quadrupedal locomotion in intact cats (Zelenin et al. 2011).

A consistent finding during 1:2 and 1:3 relationships was that the last step of the fast limb was longer than the preceding step(s), with longer stance and swing durations (Fig. 9). The duration of extensor bursts of the fast limb was also prolonged in the last step and showed greater activity (Fig. 10). These results are similar to those found in human infants (see fig. 5 of Yang et al. 2005), suggesting that adjustments found in human infants during split‐belt locomotion are primarily mediated at a spinal level. The longer last step of the fast limb and increased extensor activity probably facilitate the transition from stance to swing for the slow limb. Increased extensor activity in the fast limb starts well before onset of the swing phase or flexor burst of the slow limb (see Fig. 7), indicating that it is mediated centrally. Therefore, the adult mammalian spinal cord can precisely time events on the left and right sides to maintain stable locomotion, even when one hindlimb takes more steps than the other.

Our results also show that some parts of the cycle are more modifiable than others. For instance, during split‐belt locomotion with 1:1, 1:2 and 1:3 relationships, the timing of stance onset of the fast limb relative to the slow limb and placement of both limbs at contact were invariant with increasing slow–fast speed ratios (Figs 12 and 13). In contrast, the timing of stance onset of the slow limb relative to the fast limb and the placement of both limbs at swing onset were modulated with slow–fast speed ratios. At swing onset, the paw of the slow limb was more rostral (horizontal distance from the hip) than that of the fast limb. As discussed below, this affects the relative contribution of different sensory signals in the control of phase transitions.

An important question is why the fast limb decides to take more steps than the slow limb. This appears related to the degree of hip extension, the position of the slow limb in relation to the fast limb and the level of loading for each limb (see Fig. 7). At the first stance‐to‐swing transition of the fast limb, the slow limb is either preparing to accept weight support or has just started its stance phase. As the slow limb is moving more slowly and the swing phase of the fast limb is relatively rapid, the fast limb makes contact when the slow limb has just started its stance phase. At this point, two choices are available to the spinal locomotor network: (1) either the slow limb transitions to swing or (2) the slow hindlimb remains in stance and lets the fast limb perform its stance phase followed by a swing phase and thus take an extra step(s). If the slow limb transitions to swing and the fast limb is too rostral, the animal would fall backwards. This important decision is undoubtedly informed by sensory signals from both limbs, as discussed in the next section. The fact that not all animals adjusted similarly to slow–fast speed ratios (e.g. presence and frequency of 1:1, 1:2, 1:3, 1:4, 1:5 relationships) and that one animal maintained a 1:1 relationships up to a 10‐fold left–right speed ratio indicates individualized strategies that could reflect intrinsic inter‐animal differences in neuronal communication between the left and right CPGs, in the strength of reflex pathways to spinal neuronal networks or in the relative position of the limbs at stance onsets and offsets. Inter‐animal variability is a complex phenomenon that requires further investigation (Frigon, 2011).

Interdependence of left and right spinal pattern generators

It was proposed that left and right pattern generators, defined as the CPG working with sensory feedback, could operate independently of one another, to a certain degree, because the left and right limbs could take a different number of steps or step in different directions (Yang et al. 2005; Choi & Bastian, 2007). However, are extra steps taken by the fast limb or simultaneous opposite directions indicative of independence of the left and right pattern generators? Several lines of evidence indicate that the left and right spinal pattern generators do not, from simple to extreme circumstances that challenge left–right coordination, operate independently. First, the phasing of stance onsets and offsets of the fast limb relative to the slow limb was extremely consistent, even with 1:2 and 1:3 relationships (Fig. 12). Such consistent phasing was also shown in human infants during split‐belt locomotion (Yang et al. 2005) and in spinalized turtles during fictive scratching (Stein & McCullough, 1998). Second, the flexion phases of the left and right limbs never overlap during 1:1, 1:2 or 1:3 relationships [see Srt bursts in Figs 4 and 7 (see also Forssberg et al. 1980; Yang et al. 2005)], indicating that flexor‐generating centres on the left and right sides remain strongly reciprocally coupled. Although swing phases can overlap, such as in Fig. 4 B, the flexion phase of the slow limb was finished and had started its extension (E1 phase) to prepare for contact. Previous studies also showed that even with the legs stepping in opposite direction, the flexion phases never overlap (Yang et al. 2005; Choi & Bastian, 2007). Third, the cycle and phase durations of the fast limb during 1:2 or 1:3 relationships were not equal. The last cycle and its phases were always longer to accommodate the stance‐to‐swing transition of the slow limb. Fourth, one study showed that an isolated spinal pattern generator (lateral hemisection with a longitudinal myelotomy caudal to the hemisection) could produce locomotor movements without signals from the other limbs (Kato, 1989). Although this is convincing evidence that a locomotor pattern generator controlling one limb can function independently, the ability of the isolated hindlimb to stand and step required approximately 3 weeks to recover, indicating that the neural control of the hindlimbs is normally intimately intertwined.

Scratching and paw shake behaviours are evidence for autonomous functioning of a single pattern generator. However, when both the left and the right sides are considered, the neuronal composition of the spinal networks generating locomotion, scratch and paw shake are undoubtedly different, even though they share common elements (Frigon, 2012).The neuronal pathways that project contralaterally to coordinate the left and right sides must be depressed during unilateral rhythmic motor activities. Therefore, although the left and right spinal locomotor CPGs are probably anatomically and functionally distinct, in simple to extreme conditions that challenge left–right coordination they remain strongly coupled, as shown in the present study.

Sensory control of phase transitions and left–right interactions between spinal locomotor networks

Phase transitions are key events in the locomotor cycle that require precise coordination between the left and right sides to maintain stable locomotion. For instance, during walking or trotting, as one hindlimb transitions from stance‐to‐swing, the contralateral hindlimb must be in a position to provide weight support. Sensory feedback is important in regulating the timing of phase transitions. In particular, stretch‐sensitive inputs from hip flexor and hip extensor muscles and load‐sensitive inputs from extensor muscles and cutaneous mechanoreceptors of the plantar surface have been shown to influence phase transitions in cats and humans (reviewed by Duysens et al. 2000; Rossignol et al. 2006; Pearson, 2008). In the present study, the degree of stretch on hip flexors and hip extensors was estimated by the horizontal distance of the paw relative to the hip at swing and stance onsets, respectively, while loading was estimated by the timing of the stance phase and the level of EMG in extensor muscles. The level of feedback related to loading, hip extension and hip flexion varies throughout the step cycle.

According to experimental and modelling studies, the transition from stance to swing is regulated by a reduction in loading‐dependent feedback from extensor muscles and cutaneous afferents from the paw pad, as well as an increase in stretch‐dependent feedback from hip flexor muscles as the hip is extended (Grillner & Rossignol, 1978; Pearson, 2008; Markin et al. 2010). On the other hand, the transition from swing to stance appears to be mainly controlled by stretch‐dependent feedback from hip extensor muscles (McVea et al. 2005). Figure 14 shows a modified version of the model proposed by Pearson (2008) to explain the effects of sensory feedback related to loading (group Ib inputs from extensor muscles and cutaneous inputs from paw pad) and hip extension (group Ia and II inputs from hip flexors) on the regulation of the stance to swing transition. Here, it is expanded to include the role of stretch‐sensitive inputs related to hip flexion (group Ia and II inputs from hip extensors) and the interactions with the contralateral limb. In the model, extensor and flexor half‐centres on each side of the spinal cord mutually inhibit each other for extensor–flexor alternation while flexor half‐centres on the two sides of the spinal cord inhibit each other for left–right alternation.

Figure 14. Schematic diagram illustrating the sensory control of phase transitions and interactions between the left and right spinal pattern generators.

Central pattern generators on both sides of the spinal cord are represented as extensor (E) and flexor (F) half‐centres that mutually inhibit each other while left–right interactions are assumed by mutual inhibition between flexor half‐centres. Extensor half‐centres project to extensor motoneurons (Ext) that produce the left (LSTA) or right (RSTA) stance phases while flexor half‐centres project to flexor motoneurons (Fle) that produce the left (LSW) and right (RSW) swing phases. The effects of loading feedback (group Ib and cutaneous afferents from plantar surface), hip extension (group Ia and II inputs from hip flexors) and hip flexion (group Ia and II inputs from hip extensors) are shown. Projections with filled circles or with a straight line represent inhibitory and excitatory influences, respectively.

During tied‐belt locomotion, where the hindlimbs are approximately perfectly out‐of‐phase, the stance‐to‐swing transition for the right hindlimb occurs when the left hindlimb is in early stance and vice versa (Fig. 1). With increasing speed, the swing‐to‐stance transition occurs increasingly earlier in the stance phase of the contralateral limb. In our model, at the stance‐to‐swing transition for the right hindlimb, the right hip is extended and group Ia and II inputs from right hip flexors provide excitation to the right flexor half‐centre. At the same time, loading on the right hindlimb is reduced, decreasing loading feedback to the right extensor half‐centre and disinhibiting the right flexor half‐centre. Excitation of the right flexor half‐centre activates flexor motoneurons, initiating swing of the right hindlimb, and inhibits the left flexor half‐centre. Concurrently, the left flexor half‐centre is inhibited by loading‐related feedback, as the left hindlimb is starting its support phase, and by stretch‐sensitive inputs from hip extensors, as the hip is flexed.

During split‐belt locomotion with a 1:1 relationship, the stance‐to‐swing transitions for the slow (left) and fast (right) hindlimbs are not mirror images of one another (Fig. 4). The stance‐to‐swing transition for the right hindlimb occurs during early stance of the left hindlimb and at the highest slow–fast speed ratio it can even precede contact of the left hindlimb (i.e. before loading feedback can contribute) (see Fig. 4 B). At this transition, the right hindlimb is extended past the hip marker and stretch‐sensitive inputs from hip flexors are probably the main regulator of the stance‐to‐swing transition. On the other hand, the stance‐to‐swing transition for the left hindlimb occurs relatively later during the stance phase of the right hindlimb. As the right hindlimb is moving faster it can extend further back before the left hindlimb transitions from stance to swing. Although the left hindlimb also makes its stance‐to‐swing transition extended past the hip marker, the level of hip extension at this transition is less than that of the right hindlimb. In our model, stretch‐sensitive feedback from left hip extension and right loading feedback facilitate the stance‐to‐swing transition of the left hindlimb by exciting and disinhibiting the left flexor half‐centre, respectively.

During split‐belt locomotion with 1:2 and 1:3 relationships, interactions between the left and right sides become increasingly important to maintain stable locomotion. Figure 15 illustrates the series of events with a 1:2 relationship during split‐belt locomotion depicted in Fig. 7 A. For the left hindlimb, at the swing‐to‐stance transition, the hip is flexed and stretch‐related inputs from hip extensors activate the extensor half‐centre to prepare the limb for contact. At the same time, the right hindlimb is extended and hip flexor inputs activate the ipsilateral flexor half‐centre. The left hindlimb (the slow limb) remains in stance for the next three events, providing force feedback to the ipsilateral extensor half‐centre and preventing the limb from transitioning to swing. During this period, the right hindlimb makes contact, transitions to swing and then makes another contact, guided by stretch‐sensitive inputs from hip extensors, hip flexors and hip extensors, respectively. The left hindlimb transitions to swing as it becomes sufficiently extended and as force feedback increases in the right hindlimb, disinhibiting the flexor half‐centre of the left hindlimb. The left hindlimb then makes contact and the cycle begins anew.

Figure 15. Sensory control of phase transitions and left–right interactions during split‐belt locomotion with a 1:2 relationship in the number of steps taken by the slow and fast limbs.

The model described in Fig. 14 is used to explain the middle cycle shown in Fig. 7 A with the left (LH) and right (RH) hindlimbs stepping at 0.1 and 0.4 m s⁻¹, respectively. The limb diagrams on the left show the positions of the hindlimbs at the phase transitions: left (LHC) or right (RHC) contact and at left (LHSO) or right (RHSO) swing onset.

Concluding remarks

The present study showed that pattern generators within the adult mammalian spinal cord can adjust hindlimb locomotion when left–right coordination is challenged from simple to extreme conditions. Although extra steps can be taken by the fast limb, the left and right pattern generators remain tightly coupled. As stated by Forssberg et al. (1980), this complex level of control within the spinal cord means that the brain does not need to concern itself with the trajectory of the hindlimbs. Sensory inputs from the moving hindlimbs working with spinal mechanisms are sufficient to produce complex adjustments in the pattern.

Additional information

Competing interests

The authors declare no conflicts of interest.

Author contributions

All experiment were performed in the Frigon laboratory. A.F. was involved in the conception and design of the work. A.F., E.D., Y.T., M‐F.H and C.D. were involved the acquisition, analysis or interpretation of data. A.F. drafter the work while all authors revised it critically for intellectual content. All the authors approved the final version of the manuscript and agree to be accountable for all aspects of the work in ensuring that questions related to the accuracy or integrity of any part of the work are appropriately investigated and resolved. All authors qualify for authorship and all those who qualify for authorship are listed.

Funding

The present research was funded by grants from the Natural Sciences and Engineering Research Council of Canada (NSERC) and the Beverly Petterson Bishop Award from the American Physiological Society to A.F. M.‐F.H. was supported by a master's scholarship from the Fonds de recherche du Québec ‐ Santé (FRQS) and Y.T. was supported by a doctoral scholarship from the FQRNT.