A secondary motor area contributing to interlimb coordination during visually guided locomotion in the cat

Abstract

We investigated the contribution of cytoarchitectonic cortical area 4δc, in the caudal bank of the cruciate sulcus of the cat, to the control of visually guided locomotion. To do so, we recorded the activity of 114 neurons in 4δc while cats walked on a treadmill and stepped over an obstacle that advanced toward them. A total of 84/114 (74%) cells were task-related and 68/84 (81%) of these cells showed significant modulation of their discharge frequency when the contralateral limbs were the first to step over the obstacle. These latter cells included a substantial proportion (27/68 40%) that discharged between the passage of the contralateral forelimb and the contralateral hindlimb over the obstacle, suggesting a contribution of this area to interlimb coordination. We further compared the discharge in area 4δc with the activity patterns of cells in the rostral division of the same cytoarchitectonic area (4δr), which has been suggested to be a separate functional region. Despite some differences in the patterns of activity in the 2 subdivisions, we suggest that activity in each is compatible with a contribution to interlimb coordination and that they should be considered as a single functional area that contributes to both forelimb–forelimb and forelimb–hindlimb coordination.

Introduction

Locomotion over uneven terrain requires the capacity to visually analyze the irregularities in the ground and to plan how to step over, onto, or around obstacles. As we have detailed elsewhere (Marigold et al. 2011; Drew and Marigold 2015; Nakajima et al. 2019), this simple, everyday activity requires a complex processing of visual information to ensure that progression is appropriately modified to allow successful negotiation of an impediment. In particular, after detection of an obstacle, one needs to determine its dimensions and the relative distance of that obstacle as one approaches it. This information is used to adjust gait so that one forelimb is placed (planted) firmly in front of the obstacle while the other is brought over it. In the case of a quadruped, there is the additional task of ensuring that the activity in the hindlimbs is also modified appropriately to step over the obstacle. In this latter situation, modifications to hindlimb activity are made in the absence of visual information, which is lost as soon as the obstacle begins to pass under the body. Temporal changes in hindlimb activity have therefore to be predicted and retained based both on the prior visual information and on the modification in timing, paw placement, and limb trajectory of the preceding forelimb activity. Appropriate changes in hindlimb timing and trajectory lead to the seamless passage of the hindlimb over the obstacle; the absence of any such adaptations leads to stumbling.

The planning and execution of visually guided gait modifications undoubtedly depend on a distributed network of cortical areas and subcortical structures (Drew and Marigold 2015). In the case of forelimb–hindlimb coordination, spinal and brainstem pathways are sufficient to ensure one-to-one coupling of the forelimb and hindlimb during unobstructed locomotion at different speeds (Grillner 1981; Rossignol 1996; Frigon 2017), although damage to the corticospinal tract impairs coupling (Bem et al. 1995; Jiang and Drew 1996).

A cortical contribution to the control of the hindlimb during gait modifications is supported by the finding that cells in the hindlimb representation of the primary motor cortex (area 4γ) show large changes in their discharge patterns related to the execution of the step over an obstacle (Widajewicz et al. 1994). However, in both the forelimb and the hindlimb representations of the motor cortex, the majority of the cells are tightly related to the execution of the step over the obstacle. Few cells discharge between the passage of the fore- and hindlimb as one might expect for cells that are coordinating the activity of the hindlimbs on the basis of the movements in the forelimb. The question then arises as to the source of the signal responsible for such a coordination.

A partial reply to this question comes from our previous recordings in area 5b of the posterior parietal cortex (PPC; Andujar et al. 2010; Lajoie et al. 2010). There, we detailed the properties of a population of cells in the medial part of the PPC that changed their discharge activity between the passage of the forelimbs and the hindlimbs over the obstacle, as predicted above. Moreover, in experiments in which forward progression was halted when an obstacle was located under the body, cells in this same area continued to discharge until progression resumed (Lajoie et al. 2010). Such cells have been proposed to hold information about upcoming movements of the hindlimb in working memory (see also McVea et al. 2009; Wong et al. 2017). However, such cells are unlikely to be directly responsible for producing changes in hindlimb activity as the discharge of a large proportion of these cells was independent of which limb was the first to step over the obstacle. This means that they could not directly specify motor activity in a given limb. Instead, such cells may provide information on the localization of the obstacle as it passes under the body (Pearson and Gramlich 2010; Drew and Marigold 2015), leaving open the question of how and where this limb-independent signal is transformed into a limb-dependent signal that can be used to adapt the motor activity in the hindlimb with that of the forelimb.

One possibility is that such a transformation may be performed by the cat secondary motor areas in the same way that primate secondary motor areas contribute to the control of visually guided reaching (see Cisek and Kalaska 2010; Rizzolatti et al. 2014). In the cat, these secondary motor areas are found mostly within the cruciate sulcus and extend also onto the anterior sigmoid gyrus (see Fig. 1). They include 4 cytoarchitectonically identified subdivisions of area 6 (6aα, 6aβ, 6aγ, and 6iffu) and 3 subdivisions of area 4 (4δ, 4sfu, and 4fu) (Hassler and Muhs-Clement 1964; Avendaño et al. 1992; Ghosh 1997b). More recently, on the basis of differential anatomical projections, Ghosh (1997a, 1997c) subdivided area 4δ into a caudal region (4δc) and a rostral (4δr) region, divided by the fundus of the cruciate sulcus.

Fig. 1.

Histological localization of the recordings. A–C) Reconstructions of selected penetrations in area 4δc in cats P1 (A), P2 (B), and P5 (C) corresponding to cells illustrated in Fig. 2 (correspondence with Fig. 2 in top left of each rectangle). Dashed line indicates layer V of the cortex and the filled symbol indicates the location of the recording. D) Surface view of the cat’s brain; the horizontal line indicates the approximate location of the section shown in C. E–G) Flattened representation of the pericruciate cortex, illustrating the location of all cells (circles) recorded in areas 4δc and 4δr in each cat. Larger symbols indicate cells that showed changes in activity either during unobstructed locomotion and/or during steps over the obstacle. Cells with identical coordinates have been jittered by 0.2 mm. Dashed lines indicate divisions between cytoarchitectonic regions; solid lines indicate the fundus of the cruciate sulcus (Cru), the lip of the anterior and posterior sigmoid gyrus (ASG and PSG, respectively), the fundus of the ansate sulcus (Ans), and the fundus of the coronal sulcus (Cor). Horizontal lines indicate the location of the penetrations illustrated in A–C; the location of the cell illustrated in Fig. 2E is also illustrated in Fig. 1F. Other abbreviations: Lat, lateral sulcus; N, number of modified and/or modulated cells as a function of the number of total cells recorded in areas 4δc and 4δr; SS, suprasylvian sulcus.

In a recent microstimulation study of the functional connectivity of these secondary motor areas (Fortier-Lebel et al. 2021), we showed that intracortical microstimulation in area 4δc during locomotion produced phase-dependent responses in both contralateral forelimb and hindlimb muscles, albeit at a higher threshold for activation than in the primary motor cortex. In addition, some sites produced simultaneous movements of both the forelimb and hindlimb, predominantly on the contralateral side of the body. Moreover, in a previous study examining the properties of cells in area 6iffu and area 4δr (Nakajima et al. 2019), we found discharge characteristics that were compatible with a contribution of these areas in transferring global properties of the obstacle into a limb-specific signal that contributes to the step over it by the contralateral forelimb. We postulate here that area 4δc may perform a similar function in transferring the limb-independent discharge of the cells in the medial portions of the PPC into a limb-dependent signal that serves to coordinate the hindlimb movement with that of the forelimb. If this hypothesis was correct, one would expect to observe cells that change their discharge activity between the passage of the forelimb and hindlimb over an obstacle and that show predominantly limb-dependent activity.

We also had 2 secondary aims in these experiments. One of these was to determine if areas 4δc and 4δr should be considered as a single functional area, with similar discharge characteristics, or whether they should be considered as 2 different functional areas, as suggested by Ghosh (see above). Arguments that they should be considered as a single functional area come from their definition as a single cytoarchitectonic region and the fact that intracortical microstimulation (ICMS) produces similar effects, at similar thresholds in both areas (Ghosh 1997c; Fortier-Lebel et al. 2021). Arguments against come from the suggestion of Ghosh (1997a) that the 2 areas have different patterns of connectivity with surrounding areas.

The other aim was to verify the suggestion, originally made by Ghosh on the basis of his stimulation experiments (Ghosh 1997c) and supported by our own results (Fortier-Lebel et al. 2021) that area 4δc (and perhaps, by extension, area 4δr) has analogies with the primate supplementary motor area (SMA, or area F3). Testing such a hypothesis requires comparison not only of the properties of cells in area 4δc but also of the corticocortical projections to this region. We therefore complemented our electrophysiological study of area 4δc with an anatomical one in which we injected retrograde tracers into area 4δc and quantified the density of the projections from the surrounding frontoparietal and cingulate cortical areas. These anatomical experiments also allowed us to determine if area 4δc receives input from the medial regions of the PPC as one would expect on the basis of our primary hypothesis.

Our results support our suggestion for a contribution of area 4δc to the coordination of activity in the forelimb and hindlimb during locomotion and also provide support for considering areas 4δc and 4δr as a single, secondary motor area with a common function in interlimb coordination. We further discuss the evidence supporting similarities in function and connectivity of feline area 4δ with primate SMA.

Methods

Task and training

Experiments were performed on the same 2 adult cats (P1 and P2, weight 4–5 kg) as used in a previous study (Nakajima et al. 2019) as well as on 2 additional cats (P4 and P5, weight also 4–5 kg). All 4 cats were used in a previous study examining the effects of ICMS on muscle activity during locomotion (Fortier-Lebel et al. 2021). All cats were initially trained to walk on a treadmill at 0.45 m/s and subsequently to step over obstacles attached to the moving belt (Drew 1988, 1993; Nakajima et al. 2019). In all experiments, 2 obstacles were attached to the treadmill belt, placed equidistantly 3 m apart, allowing the cat to take 12–14 steps in between each obstacle. The treadmill was designed so that the cats were able to see the advancing obstacle 10–12 steps before they stepped over it. The size of the obstacles for the 4 cats varied from 5 to 10 cm high, in all cases allowing the cats to step over the obstacles without interruption of their cadence.

Surgery

Surgical details are provided in a previous report (Nakajima et al. 2019). In brief, cats were appropriately premedicated and then anesthetized with a mixture of 2%–3% isoflurane and oxygen. Temperature, respiration rate, heart rate, and blood oxygenation were monitored throughout the surgery and maintained at steady levels. Analgesics (buprenorphine 5 μm/kg) were administered at the beginning and the end of the surgery. Drying of the cornea was prevented by the application of petroleum jelly and xylocaine was applied to the atraumatic ear bars before placing the cat in a stereotaxic apparatus. A craniotomy was made to provide access to the pericruciate cortex and a stainless steel baseplate with internal dimensions of 10 × 15 mm was attached to the cranium with dental acrylic and stainless steel screws (Drew 1993). The chamber was implanted over the right pericruciate cortex in all 4 cats. Pairs of Teflon-insulated, braided stainless steel wires were implanted into the bellies of selected forelimb and hindlimb muscles (Drew et al. 1986). Wires were led subcutaneously to a connector on the cranium. Arrays of microwires were implanted into either the cerebral peduncle (stereotaxic coordinates A4, L3–L5, Berman 1968) and/or the pyramidal tract (P7, L1) on the side of the recording chamber by using a harpoon assembly (Palmer 1978; Drew 1993). Analgesics were given as required and as advised by the institutional veterinary surgeon, and the animal was left to recover for 1–2 weeks. All procedures followed the recommendations of the Canadian Council of Animal Care and of the local ethics committee.

Protocol

Recordings were made 2–4 times a week. In each session, a single glass-insulated tungsten microelectrode (impedance 0.5–1.5 MΩ) was inserted into the brain at a given location. In the experiments targeting area 4δc, penetrations initially traversed the posterior sigmoid gyrus of the pericruciate cortex before entering the caudal bank of the cruciate sulcus (see Fig. 1A–C). Penetrations into area 4δr passed through the posterior sigmoid gyrus and the caudal bank of the cruciate sulcus before entering the ventral bank. Thus, we generally traversed 2 or 3 different bands of cells during the experiments.

As the electrode was advanced, stimulation of the pyramidal tract or cerebral peduncle electrodes was used to identify layer V in each band of cells that we traversed, based on the presence of antidromically activated action potentials (as determined by the presence of a fixed latency and the collision test, Lipski 1981). In each band of cells that the electrode traversed, we initially recorded cell activity as the cats stepped over the obstacles attached to the moving belt. Recording was maintained until we recorded 10 steps over the obstacle with each limb leading or until isolation of the recorded unit was lost. We then applied ICMS with the cat at rest, and in most sites, also during locomotion (Fortier-Lebel et al. 2021). Following this, the electrode was advanced to the next band of cells. Small lesions (20–50 μA, DC cathodal current) were made just above or below layer V in selected penetrations to aid in histological reconstruction.

Electromyographic (EMG) activity during locomotion was bandpass filtered (100–450 Hz) and digitized at 1 kHz simultaneously with the action potentials from the cells, which were digitized at 50 or 100 kHz to maintain the full waveform. Video recordings (60 frames/s) were made during all experiments and synchronized with the EMG and cell recordings by using a digital time code.

Analysis

Sections of data with stable locomotion and cell recordings were selected for analysis based on the video recordings and examination of the digitized action potentials. Cells with stable locomotion were analyzed providing that the recording contained at least 4 steps over an obstacle with each limb leading. Action potentials from single cells were discriminated from background activity by using commercially available software (Plexon Offline Sorter). Cells were isolated based on principal component analysis of their waveforms with subsequent verification by one of the experimenters (TN). Generally, only 1 cell was isolated from each recording although occasionally 2 or 3 cells could be discriminated.

Task-related cells during the step over the obstacle

We used a custom program to identify the onset and the offset of the period of EMG activity in selected forelimb and hindlimb flexor muscles during locomotion. Step cycles were defined as from the onset of activity of the contralateral (co) or ipsilateral (i) forelimb flexor (cleidobrachialis [ClB] or brachialis [Br]), until the onset of the next period of activity in the same muscle. Based on the synchronized video recordings, we then identified those step cycles in which the cat stepped over the obstacle and categorized them according to whether the leg contralateral or ipsilateral to the recording site was the first to step over the obstacle (respectively, contralateral, and ipsilateral lead conditions). We also identified the step cycle that occurred 4 step cycles (8 steps) before the step over the obstacle and used this cycle as a control to determine changes in activity on the approach to, and/or during the step over, the obstacle. Activity during this step cycle was uninfluenced by even the most precocious changes in activity observed in our prior recordings in either the PPC (Andujar et al. 2010; Marigold and Drew 2011; Marigold and Drew 2017) or in area 6iffu (Nakajima et al. 2019), both areas that included many cells that modified their discharge activity in advance of the step over the obstacle.

For the purposes of display and analysis, cell activity was transformed into its instantaneous frequency (1,000/interspike interval) and filtered at 25 Hz (fourth-order low pass Butterworth). Cell activity for each step cycle was then aligned on the onset of the coClB/Br or the iClB/Br, according to which limb was the first to step over the obstacle. We then used the routine interpft from Matlab to allocate the averaged frequency for each step cycle into 512 equal bins (binwidth ~2 ms). For the purposes of the current analysis, we performed these procedures for the 2 step cycles preceding the onset of the step over the obstacle, for the step cycle in which the cat stepped over the obstacle, and for the following step cycle. In the illustrations of cell activity (see Fig. 2), the averaged activity in the control step cycle (see above) is repeated 4 times, once for each of the step cycles that we illustrate.

Fig. 2.

Examples of task-related cells in area 4δc. A–C) Three examples of limb-dependent cells, modified only (A, C) or primarily (B) during the contralateral lead condition (red traces). E, F) Two limb-dependent cells that each discharged similarly during the contralateral and ipsilateral lead conditions. For each of these cells, in A–C and E, F, we illustrate peri-event histograms (PEHs), and raster displays of cell activity synchronized to the onset of activity in the contralateral cleidobrachialis or brachialis (coClB/Br) in both the contralateral and ipsilateral lead conditions: see EMGs in D and Gi. G) An example of a limb-independent cell. In Gi, the activity is triggered on the coClB in both conditions, as for A–F, while in Gii, the data are triggered on the onset of activity in the ClB of the lead limb. All raster displays are rank-ordered according to the duration of the flexor burst. Red traces indicate the contralateral lead condition; green traces represent the ipsilateral lead condition; blue traces indicate activity during unobstructed (control) locomotion (see key in part Gii). Shaded areas around the blue traces on the PEHs indicate the interval of confidence (P = 0.01) of the standard error of the mean of the control activity. Red horizontal line below the PEH indicates the period of significant task-related modifications in discharge activity during contralateral lead; green line indicates significant activity during the ipsilateral lead. D) Averaged EMG activity recorded simultaneously with the cell illustrated in B. Vertical black lines in A–G delimit step cycles (−2 to +2). Numbers in parentheses below the x-axis and adjacent to the coClB and iClB muscle bursts indicate steps with respect to the step over the obstacle. coLead, step over the obstacle when the contralateral limb leads; iLead, activity when the ipsilateral limb leads; N, number of step cycles included for each condition and the control; Srt, anterior head of sartorius; St, semitendinosus. The color key in part Gii pertains to the entire figure.

Each step cycle was then divided into 2 steps; for the contralateral lead condition, for example this was from the onset of the coClB/Br to onset of the ipsilateral ClB/Br (iClB/Br), the first step, and then from the iClB/Br to the onset of the next coClB/Br, the second step. For the ipsilateral lead condition, the first step was from the onset of the iClB/Br to the onset of the coClB/Br, and so on. Each step was divided into 10 equal windows, each of ~50 ms, and the instantaneous frequencies from each bin in the window were averaged. We then used a moving window average, consisting of 4 windows displaced by 1 window at a time to perform a 2-sample, 2-tailed t-test between the average activity during the step over the obstacle and the control activity. The moving window average began 6 steps before the step over the obstacle and continued for 3.8 steps following the onset of that step. We considered changes to be significantly task-related if more than 4 consecutive windows were significant (P < 0.05). Periods of significant activity that were separated by 5 or fewer nonsignificant bins were considered to be continuous (see Nakajima et al. 2019).

To determine if these significant changes were also meaningful, we applied a second criterion to determine if periods of discharge identified as being significant by the t-test deviated from the interval of confidence (P < 0.01) of the standard error of the mean for at least 10% of the step cycle (~100 ms). Periods of activity that fulfilled both criteria were classified as task-related.

Step-advanced versus step-related

As in our previous manuscripts (Andujar et al. 2010; Marigold and Drew 2017; Nakajima et al. 2019), we distinguish between step-advanced (SA) and step-related (SR) cells. SA cells show significant task-related changes in activity that begin more than 0.2 step cycles (0.4 steps) before the onset of the activity in the lead limb and that continue until at least 0.2 step cycles before the step over the obstacle. SR changes include those that occur within 0.2 step cycles before the onset of activity of coClB or that occur following the onset of activity in the coClB/Br. We also classify as SR activity those changes that occur earlier in the sequence, leading up to the step over the obstacle, but that end more than 0.2 step cycles before the onset of the coClB/Br. Most of these latter types of discharge would be expected to be involved in modifying activity in the steps preceding the step over the obstacle (see, e.g. Fig. 6).

Fig. 6.

Population plots of activity in different subregions of area 4. We compare the phase of the significantly modified periods of discharge activity for the populations of cell recorded in areas 4δc (A), the hindlimb representation of the primary motor cortex, 4γHL (B), the forelimb representation of the primary motor cortex, 4γFL (C), and 4δr (D). For each area, we illustrate rank-ordered phase plots of the cell discharge in the contralateral (part i) and the ipsilateral (part ii) lead conditions. Red and green lines in (i) and (ii) indicate cusums of the onset of cell discharge. We also illustrate the summed activity of the discharge during the contralateral (red) and ipsilateral (green) lead conditions (part iii), together with cusums of this summed activity (see Methods). Data in all plots are aligned to the onset of activity in the coClB/Br (phase = 0). In cells with more than one burst of activity, the period of discharge showing the largest change in activity was used for the alignment. Black horizontal bars in parts (i) and (ii) indicate increases of activity; gray horizontal bars indicate decreases of activity. Binwidth in part iii = 0.1 step, ~50 ms. N, number of cells in each plot that show modified activity in the indicated condition.

Limb-dependent versus limb-independent

Cells that show significant modification of discharge activity at the same phase of the step cycle (with respect to the onset of the coClB/Br) in both the contralateral and ipsilateral lead conditions are defined as limb-dependent. In the primary motor cortex, this cell discharge is almost invariably related to the contralateral limb. Cells that show task-related activity only in one condition (contralateral or ipsilateral lead condition) are, by definition, also limb-dependent. In contrast, the discharge of limb-independent cells overlaps when the activity is triggered on the onset of activity in the ClB/Br of the leading limb. As a result, there is a substantial phase shift, centered around 0.5 step cycles (1 step) when a cell is triggered on the coClB/Br in both conditions. In this manuscript, we consider cell discharge to be limb-independent when there is a phase shift ≥0.4 step cycles in the onset of cell activity when triggered on the coClB. We use 0.4 rather than 0.5 to allow for small variations in the time of onset of activity.

Population plots

Population plots (as in Fig. 6) were constructed by rank ordering the cells according to the phase of the onset of the principal period of modified activity with respect to the onset of activity in the coClB/Br. In cells with only one period of significantly modified activity, this was straightforward. In cells with multiple periods of significantly modified activity, we used the period of activity that showed the largest change. We also created histograms indicating the sum activity of the population of cells during the gait modifications. For these histograms, we calculated the total number of cells that showed activity during each of the 10 bins comprising each step (see above). These totals were then divided by the total number of recorded cells to provide a percentage response across the studied time period before, during, and after the step over the obstacle. We used only the principal period of activity for this calculation. For both the phase plots and the summed activity, we also calculated the cumulative sum (cusum) of the change in activity, normalized to 100%. For the phase plots, the calculation was based on the phase of onset of the change in activity, while for the summed activity, we used the activity in each bin.

Cell modulation during unobstructed locomotion

To determine whether a cell was modulated during the control step cycle, we used the Rayleigh test of directionality (Batschelet 1981; Drew and Doucet 1991). Cells that showed significant directional tuning (P < 0.05) were considered to be modulated during locomotion.

Histology

At the end of all experimental sessions, we made lesions (50–100 μA DC cathodal current) in selected locations adjacent to the explored region. The cat was then anaesthetized (ketamine induction followed by i.v. pentobarbital sodium, Somnotol, 30 mg/kg) and perfused “per cardium” with a formaldehyde solution. The brain was removed, blocked, cryoprotected, and sectioned in the parasagittal plane (40 μm sections). Sections were stained with cresyl violet. Recording sites were identified based on the marking lesions made during and after the experiments and on the depth of layer V (as determined with respect to a reference point during each penetration) with respect to the geometry of the cortex. Allocation of recording sites to different cytoarchitectonic subdivisions was based on previous descriptions of these regions (Hassler and Muhs-Clement 1964; Avendano et al. 1988; Ghosh 1997c) as used in a previous publication (Fortier-Lebel et al. 2021).

The location of each recording site was quantified on the basis of its laterality and linear distance from the fundus of the cruciate sulcus. The position was then mapped onto a flattened representation of the cortex centered on the fundus of the cruciate sulcus (rostrocaudal position = 0; see Fig. 1E–G) as in our previous experiments (Jiang and Drew 1996; Andujar and Drew 2007; Nakajima et al. 2019; Fortier-Lebel et al. 2021). Negative numbers to the left of zero indicate locations in the rostral bank of the cruciate sulcus; positive values indicate sites caudal to the fundus.

Anatomical experiments

To determine the cortical regions that project to our areas of interest, we injected the retrograde tracers Texas Red and Alexa Fluor 488 into area 4δc in 3 cats. These experiments were performed on cats that were prepared for surgery in the same manner as for the chronic experiments (see above). A craniotomy was made to provide access to the caudal bank of the cruciate sulcus and a glass-insulated tungsten microelectrode was inserted into area 4δc at mediolateral and rostrocaudal coordinates that were calculated on the basis of our unit recordings. The electrode was advanced as described above and we identified layer V in the caudal bank of the cruciate sulcus on the basis of the depth of the recording and by the presence of large action potentials. The electrode was then withdrawn, and a Hamilton syringe (32 gauge, sharp-tip needle) was inserted at the same coordinates and advanced to the measured depth of layer V. One or 2 injections of 0.1–0.3 μL of tracer (0.1 μL/min) were made in and above layer V (Table 5), resulting in labeling throughout all 6 layers. Two adjacent sites in area 4δc were targeted in each cat. The needle was left in situ for 5 min prior to and following the injection of each aliquot.

Table 5.

Labeled cells in different parts of the cerebral cortex following injections into 4δc.

Cat	Tracer	Volume	Total cells	Area5(med)	S1	4γ	4δc	4sfu	Area 6	Cing
Prem7	AF	0.70	3085	8%	29%	35%	16%	7%	2%	3%
Prem9	TR	0.85	8449	14%	22%	25%	5%	5%	7%	8%
Prem10	TR	0.45	8406	12%	22%	34%	8%	4%	6%	7%

Total number of cells labeled following each injection together with the percentage of cells labeled in areas in which the average percentage of cell (from the 3 cats) was ≥5% of the total. Area 5(med) indicates those regions of areas 5a and 5b that are medial to the lateral sulcus; area 6 includes all 4 subdivisions of this cytoarchitectonic area. Volume indicates the total volume of tracer injected in the 2 sites used for each injection (in microliters). AF, Alexa-Fluor; TR, Texas Red.

To examine the overlap between regions of area 4δc that projected to the forelimb and hindlimb regions of the primary motor cortex, in one cat, we made large injections of Texas Red into the forelimb representation of area 4γ and large injections of Alexa Fluor 488 into the hindlimb representation. We pretreated the cat as above and then continued anesthesia with a constant i.v. perfusion of ketamine under veterinary guidance. Heart rate, temperature, and respiration rate were monitored as above. The anesthesia level was maintained so as to eliminate the withdrawal reflex and the corneal reflex while permitting responses to ICMS, albeit at relatively high stimulus strengths. We stimulated at various regions within the anterior and posterior sigmoid gyri, as well as in the caudomedial regions of the cruciate sulcus, and identified sites in which we could evoke EMG responses specific to either the contralateral forelimb or hindlimb. Tracer injections were then made in sites with pure forelimb or hindlimb representation, as described in the preceding paragraph.

After ~14 days, each cat was anesthetized (ketamine induction followed by i.v. pentobarbital sodium, Somnotol, 30 mg/kg) and perfused “per cardium” with 4% paraformaldehyde in phosphate buffer (pH 7.4) (Andujar and Drew 2007). The brain was removed and sectioned in the sagittal plane (40 μm sections). Every third section was mounted for fluorescence imaging and the adjacent section was stained with cresyl violet to identify cytoarchitectonic boundaries. Labeled cells were identified at a magnification of 100× using a fluorescence microscope with appropriate filters and their location marked using the Neurolucida software (MBF Bioscience). We also digitized the parasagittal path of layer V in each section. Using custom software, we then straightened the layer V path into a line and, after collapsing the labeled cortical cells onto this line, calculated the number in each 200 μm segment. The coordinates of labeled cells were calculated as described above for the recording sites and plotted on the flattened maps of the cortex aligned on the fundus of the cruciate sulcus.

To provide quantification of the density of the labeling in each area, we used 2 types of analysis. For each of these, we first calculated the mean number of labeled cells in each bin, together with the standard deviation (SD). In one analysis, we then divided each bin into one of 5 equal groups based on their density, with the first group containing all bins in the 1%–20% range of the mean + SD and the fifth group containing all cells in the 80%–100% of the mean + SD. Bins that contained more cells than the mean + SD were allocated to the highest density group. These groups were then plotted on the flattened maps using a color code (see Fig. 9). For the second analysis, we followed a similar process, except that no grouping occurred. The density of the labeling in each bin was first plotted as a Z-axis value on the flattened cortical map. The resulting surface plot was then divided into 3,600 nodes (60 × 60) and smoothed using SYSTAT’s NEXPO function. It was then partitioned into 20 contours representing concentric isodensity lines of increasing value (contours, each of 5%, from 5% to 100%). Again, the maximum contour was set to the mean + SD; this had the effect of reducing the influence of bins with very large numbers of labeled cells.

Fig. 9.

Quantitative analysis of the distribution of labeled cells following tracer injections in area 4δc. A, B) Number of labeled cells in different regions of the frontal cortex plotted on flattened representations of the cortex in 2 cats, Prem10 (A, same cat as in Fig. 8) and Prem9. Each circle represents a 200 μm bin and the color of that circle indicates the relative proportion of cells labeled in the bin (see color key). For example, a red symbol indicates all bins containing between 80% and 100% of the maximum value (mean + SD; see Methods). C, D) Contour plots in which the maximum contour is again set to the mean + SD and the dataset is divided into 20 levels. E) Contour plot of the distribution of cat Prem7, with an injection in area 4δc. F) Contour plot of the cells that were labeled following an injection into area 6iffu, also in cat Prem7 (taken from Nakajima et al. 2019). Filled black areas and surrounding white region in all parts of the figure indicate the location of the injection site and the surrounding exclusion site (same convention in Fig. 10). In the larger of the 2 boxes in each part of the figure, cells are aligned with respect to the fundus of the cruciate sulcus with negative values indicating cells rostral to the fundus of the cruciate sulcus and positive values cells that are caudal to the fundus. In the smaller box, cells are aligned on the fundus of the splenial sulcus with cells in the cingulate cortex that are caudal and ventral (CV) to the fundus being indicated by negative values and cells that are caudal and dorsal (CD) being indicated by positive values.

To allow the location of cells recorded from different cats to be plotted on a single representation of the cortex, we used a morphing routine previously developed by us based on an algorithm by Archibald (2020) (see also Dea et al. 2016). The method for the morphing is fully described in a previous publication (Fortier-Lebel et al. 2021). In brief, we placed fiduciary marks on multiple landmarks, including sulci and cytoarchitectonic boundaries, on each cat used in the study and used these marks to align the different landmarks to a single reference cat. The locations of the recording sites were proportionately reassigned to the reference cat in the same manner.

Note on colors

In Figs 2,5,6,9, and10, in which green and red are juxtaposed, we use the Okabe–Ito palette (Okabe and Ito 2002) for color-blind readers. However, for simplicity’s sake, we refer to these as red (vermillion), green (bluish green), and blue (sky blue) in the legend and text.

Fig. 5.

Cell activity in the ipsilateral lead condition. A, B) Two examples of limb-independent cells. Ai and Bi illustrate the cell discharge activity of 2 cells, each triggered on the onset of activity in the coClB muscle during both the contralateral (red traces) and ipsilateral (green traces) lead conditions. Aii, Bii: Activity in the same 2 cells is now triggered on the activity of the lead ClB (i.e. the period of activity in the first limb to step over the obstacle). These 2 cells are indicated by “5A,B” in Fig. 3. Arrows in Fig. 5Bii indicate the period of activity of the coSrt in the co- and iLead conditions. C, D: Two cells that showed significant modification of activity only in the ipsilateral lead condition. Ci, Di: The activity of the cells is illustrated triggered on the activity of the coClB in both coLead and iLead conditions. The numbers beside the periods of EMG activity indicate the sequence of activation of each muscle as each limb in turn steps over the obstacle. Numbers are color-coded in the same way as for the traces. Only the sequence of activity during the iLead condition is illustrated in Di. Cii) The activity of the cell illustrated in Ci is now synchronized to the onset of activity in the iSrt muscle as the ipsilateral hindlimb passes over the obstacle. Dii) The activity of the cell in Di is synchronized to the onset of activity in the coSrt as the contralateral hindlimb passes over the obstacle in the iLead condition. B1 and B2 indicate the same 2 bursts of activity in Di and Dii, displaced because of the difference in the burst used to trigger the display. The EMG burst #4 corresponds to that in Di (green traces). Bars under the histograms indicate the periods of significant modified activity. Figure otherwise arranged as in Fig. 2.

Fig. 10.

Distribution of retrogradely labeled cells following injections into areas 4γFL and 4γHL. A) Photograph of cortical surface showing sites in which injections of Texas Red (TR, red squares) and Alexa Fluor 488 (AF488, green triangles) were made. B) Averaged EMG responses to intracortical microstimulation at the 2 sites indicated. C–E) Three sagittal sections through the pericruciate cortex at different lateralities indicating the location of retrogradely labeled cells in the dorsal and ventral banks of the cruciate sulcus and in surrounding areas. Red symbols indicate Texas Red labeled cells from the injections in the forelimb representation and green symbols indicate cells labeled by the Alexa Fluor injections in the hindlimb representation. F, G) Distribution of cells labeled from the injections into the forelimb (F) and hindlimb (G) representations of the motor cortex. Color of symbols indicates the percentage of cells in each 200 μm bin (see key in Fig. 9A). H) The areas with the largest percentage of labeled cells (≥80% per bin) are displayed. For this part of the figure, red indicates cells labeled from the forelimb representation and green indicates cells labeled from the hindlimb representation. St, semitendinosus; Srt, anterior head of sartorius; TA, tibialis anterior; TriL, lateral head of triceps brachii.

Results

Database and localization

The present report concentrates on the discharge characteristics of 114 cells recorded from area 4δc located within the caudal bank of the cruciate sulcus (Fig. 1, Table 1). Additionally, for the purposes of comparison, we also include 55 cells recorded in area 4δr, data for some of which (cat P2) were included in a previous publication (Nakajima et al. 2019).

Table 1.

Cell database.

Area	Cat P1	Cat P2	Cat P4	Cat P5	Total
4δc	18	69	-	27	114
4δr	-	33	12	10	55

The table indicates the number of cells recorded in the 2 subdivisions of area 4δ of the cat pericruciate cortex for the 4 cats used in this study.

In area 4δc, 47/114 (41%) of the recorded cells were antidromically activated from either stimulation of the electrodes in the pyramidal tract or the cerebral peduncle and were therefore confirmed as being recorded in layer V. A similar proportion was found in area 4δr in which 24/55 (44%) of the recorded cells were antidromically activated. All other cells included in the database from both areas 4δc and 4δr were identified as being from layer V by their location in close proximity (generally <200 μm) to antidromically identified cells.

Figure 1A–C shows example trajectories into area 4δc in cats P1, P2, and P5 while Fig. 1E–G illustrates the localization of all cells recorded in area 4δc. Cells were recorded from throughout the extent of area 4δc in all 3 cats. We also illustrate the location of the cells recorded in area 4δr in cats P2 and P5.

Baseline activity of cells in area 4δc during unobstructed locomotion

We quantitatively analyzed the discharge patterns of all 114 cells recorded in area 4δc. Of these, 67/114 (59%) were modulated during unobstructed locomotion, based on the presence of a statistically (P < 0.05) nonuniform discharge pattern as determined by the Rayleigh test of directionality. Examples of cells with strong modulation can be seen in Fig. 2A and B (blue traces). Conversely, this means that a large proportion of the analyzed cells showed no significant modulation (see Fig. 2C, F, G). A similar proportion of cells were modulated (24/47, 51%) when considering only pyramidal tract neurons, as we did in earlier papers (Drew 1993; Widajewicz et al. 1994; Yakovenko and Drew 2015).

Discharge characteristics of cells in area 4δc during the gait modifications

A total of 84/114 (74%) analyzed cells recorded in region 4δc, including cells that were both modulated and unmodulated during unobstructed locomotion, were categorized as being task-related because they showed a significant modification of their discharge activity during the approach to and/or during the step over the obstacle. Of these, 68/84 (81%) cells were modified during the contralateral lead condition and 59/84 (70%) during the ipsilateral lead condition (Table 2).

Table 2.

Significantly modified cells in area 4δc (N = 84).

	Task-related	Limb-dependent	Limb-independent	SR		SA		Modulated
Modified during coLead	68	55	13	58		10		46
Modified during iLead	59	46	13	40		19		36
				Co	I	Co	I
Modified only in coLead	25	25		22		3		20
Modified only in iLead	16	16			14		2	10
Modified in coLead and iLead	43	30	13	36	26	7	17	26
Subtotal	84			58	40	10	19

The table indicates the number of significantly modified (task-related) cells together with the number that were identified as limb-dependent or limb-independent and as SR or SA. Co and I indicate the number of cells that were SR during contralateral or ipsilateral lead (coLead and iLead) and the number that were modified during both. Similarly, for SA. For example, 58 cells were SR during coLead, 22 of which were active only in coLead and 36 which were active in coLead and iLead. Modulated indicates the number of task-related cells that were also rhythmically modulated during the control cycles.

Activity of cells in the contralateral lead condition

In the contralateral lead condition, cells discharged at different times during the gait modification, with some modifying their activity just before or during the passage of the forelimb over the obstacle (red traces in Fig. 2A), others during that of the hindlimb (Fig. 2B), and still others in the period between the passage of the forelimb and the hindlimb (Fig. 2C, E, F). In some cases, cells (25/84) modified their activity only when the contralateral limb was the first to step over the obstacle, as for the examples in Fig. 2A and C. However, a large proportion, 43/84 cells, showed changes in discharge activity during both the contralateral and ipsilateral lead conditions, as in Fig. 2E–G. Finally, 16/84 cells discharged only in the ipsilateral lead condition (Table 2, see later). These latter cells were heterogeneous in nature and included cells that discharged with respect to the passage of the contralateral limbs over the obstacles as well as others that discharged during the passage of the ipsilateral limbs. In the text that immediately follows, we place the emphasis on the activity of cells active in the contralateral lead condition and address cells active during ipsilateral lead in a later section.

Cells discharging only in one condition (contralateral or ipsilateral lead), are by definition limb-dependent. Cells with modified activity during both the contralateral and ipsilateral lead conditions, however, can be defined as either limb-dependent or limb-independent based on whether cell discharge maintains a more constant relationship to a given limb during both contralateral and ipsilateral lead or whether the discharge maintains a more constant relationship to the lead limb regardless of condition (see Methods). The major period of task-related activity of the cells illustrated in Fig. 2B, E, F, for example, is significantly modified during both the contralateral and ipsilateral lead conditions and maintains a relatively constant relationship with respect to the onset of the activity in the coClB as shown by the superimposition of the red and green traces; such cells are classified as limb-dependent. The discharge of the cell illustrated in Fig. 2Gi, in contrast, shows a substantial phase shift in activity with respect to the onset of the coClB in the contralateral and ipsilateral lead conditions. If activity is aligned instead on the onset of activity of the ClB in the lead limb, the traces now superimpose (Fig. 2Gii). Such a cell is considered to discharge in a limb-independent manner.

To objectively differentiate between these 2 categories, we defined limb-independent cells as having an onset of discharge that differed by more than 40% of the step cycle in the 2 conditions (see Methods). Using this criterion, 13/43 cells in area 4δc discharging in both conditions were defined as limb-independent (Table 2). These limb-independent cells are illustrated by the red symbols in the plots of Fig. 3 that illustrate the relative onset (Fig. 3A) and offset (Fig. 3B) of the discharge activity in the contralateral and ipsilateral lead conditions. Most of these limb-independent cells either discharged with respect to the passage of the forelimb over the obstacle, as in Fig. 2G, or to the period between the passage of the forelimb and the hindlimb (see below). In most of these limb-independent cells, the offset also occurred relatively earlier in the ipsilateral lead condition (with respect to the onset of the coClB, Fig. 3B). In contrast, the onset and offset of most of the limb-dependent cells all lay close to the line of equivalence (blue symbols).

Fig. 3.

Relative timing of the onset and offset of cell activity in area 4δc. A) The phase of the onset of cell activity in the ipsilateral lead condition is plotted as a function of the phase of onset in the contralateral lead condition for those cells that showed task-related activity in both conditions. Blue symbols indicate limb-dependent cells; red symbols indicate limb-independent cells. Solid diagonal line indicates the line of equivalence; dotted lines delimit cells in which the onset of the modification in activity occurred 0.4 step cycles earlier or later in the ipsilateral than the contralateral lead condition. B) Similar plot for the phase of offset of activity. In all cases, phases are calculated with respect to the onset of the period of activity in the coClB/Br. “2G” indicates the cell illustrated in Fig. 2G. “5A,B” indicates the cells illustrated in Fig. 5A and B. Values in parentheses indicate steps.

Spatial organization

Given that we found cells discharging during the passage of the forelimbs and the hindlimbs over the obstacle, as well as between the passage of the 2 limbs, we investigated whether there was any topographical organization, based on discharge patterns, within area 4δc. Spatial separation of different cell groups would suggest that there might be a sequential processing of information within the area. Intermingling of cell types would be indicative of a more distributed organization. To examine this issue, we objectively divided the population into 3 major groups based on the time of the onset and the offset of the period of task-related activity in the contralateral lead condition (Fig. 4A). These groups corresponded to cells showing modified activity before or during the step over the obstacle by the forelimb (e.g. Fig. 2A), to those with modified activity occurring between the forelimbs and the hindlimbs (Fig. 2C, E, F), and to those showing changes during the step over the obstacle by the hindlimb (Fig. 2B) (see legend of Fig. 4A for details). The group with modified activity during the passage of the forelimb over the obstacle was further subdivided into those cells in which the onset began within 0.2 step cycles of the step over the obstacle (SR) and those in which the onset began earlier than that (SA): 5 (5/6) of the latter were limb-dependent cells. The division of the population into the 3 principal groups resulted, by definition, in temporally segregated populations of forelimb- and hindlimb-related cells and a larger population of cells that temporally overlapped both the forelimb and hindlimb populations. Each population contained both cells with increased activity (red bars) and those with decreased activity (blue bars). The exemplar cells illustrated in Fig. 2A–G are divided between the 3 main categories (a–g in Fig. 4A).

Fig. 4.

Spatial representation of the population of cells recorded in area 4δc. A) The population of cells is divided into 3 main groups (forelimb, FL; hindlimb, HL; and forelimb–hindlimb, FL–HL) according to the phase of the modified cell activity with respect to the onset of the coClB/Br in the contralateral lead condition. We also show the activity of a small number of cells (N = 3) that discharged only in advance of the step over the obstacle. Cells active during the passage of the forelimb were defined as those in which the onset of activity began before the end of the burst of activity in the coClB (<0.8 steps, ≡ 0.4 step cycles) and ended <1.6 steps after the onset of activity in the coClB, approximating to the onset of activity in the coSrt. These FL cells were then divided into 2 subgroups depending on whether cell discharge began during, or just before (less than 0.2 step cycles) the step over the obstacle (SR cells), or earlier than that (SA cells). Cells active during the passage of the hindlimb were defined as those whose onset was >1.5 steps after the onset of activity in the coClB. Cells discharging between the passage of the forelimb and the hindlimb were defined as those whose discharge began <1.5 steps after the onset of activity in the coClB and in which the end of the period of discharge was >1.6. Only the principal burst of activity is plotted. (One cell was unclassified and is not plotted, hence N = 67 and not 68). Cells illustrated in Fig. 2A–G are indicated by the respective letters (a-g) and arrows. B) Localization of the 3 main groups of cells morphed onto the flattened map of the cortex taken from cat P1 (see full method in Fortier-Lebel et al. 2021). N = 64 as the 3 cells that did not fall into one of the main groups are not plotted. Cell locations were jittered by a maximum of 0.2 mm, as in Fig. 1.

We then morphed the location from which these cells were recorded onto the flattened map of the pericruciate cortex taken from cat P1 (as in a previous publication, Fortier-Lebel et al. 2021). The plot in Fig. 4B shows a propensity for cells discharging during the passage of the hindlimb (red symbols) to be located more medially in the cortex (8/10, 80%, medial to 3.0 mm), while cells related to the passage of the forelimbs over the obstacle were more frequent in the more lateral regions of area 4δc (18/27, 67%, lateral to 3.0 mm). Cells discharging in the period between the passage of the forelimbs and the hindlimbs were more evenly distributed (16/27, 59% lateral to 3.0 mm). Overall, however, cells discharging during each of the 3 periods were found intermingled throughout the mediolateral extent of area 4δc. We equally observed no spatial distinction between SR and SA, between limb-dependent and limb-independent cells, or between modulated and unmodulated cells (not illustrated). The analysis therefore provides some support for topographical organization and the possibility of sequential processing within the area but without the clearer demarcations observed in the primary motor cortex.

Receptive fields were located for 19/21 of the cells which were classified as SR forelimb cells (Table 3). Of these, 14/19 had a receptive field that included the contralateral forelimb, with the majority of these being located on the more distal forelimb, including the pads. For the hindlimb cells, receptive fields were determined for 9/10 cells and 8/9 of these had a receptive field located on the contralateral hindlimb. In 6/8 of these cells, the input was from the entire hindlimb with additional input from the forelimb in 2 of them. For both the forelimb and hindlimb cells, the receptive fields were generally larger than those found in area 4γ (Armstrong and Drew 1984; Drew 1993; Widajewicz et al. 1994).

Table 3.

Receptive fields.

		Forelimb			Hindlimb			FL + HL	Head	None
	Total	Dist	Prox	All	Dist	Prox	All
FL(SR)	21	8	5	1	1			1	2	2
HL	10				2		4	2		1
FL–HL	27	4	1		2		11(3)	2		3

Receptive fields for those cells that we tested, organized according to the classification of their discharge patterns (first column) as illustrated in Fig. 4. Receptive fields were divided into those that were distal (dist: distal to elbow or knee), proximal (prox), or which included the entire limb. Value in parenthesis for the FL–HL cells indicates that for 3 cells the receptive field also included the trunk. FL + HL, receptive field included forelimb and hindlimb. None, no receptive field was located. Receptive fields were not tested for 1 HL cell and for 1 FL–HL cell.

Cells discharging in the period between the passage of the forelimb and hindlimb over the obstacle had more varied receptive fields, but again these receptive fields were generally large. Of the 26/27 cells for which we tested a receptive field, 18/26 had a receptive field that included the hindlimb (Table 3). Of these cells, the receptive field of the majority (16/18) included the entire contralateral hindlimb. An additional 5 cells had a receptive field that was restricted to the forelimb, while for 3 cells we were unable to determine a receptive field.

Cell discharge properties were equally mostly compatible with the effects of ICMS at the location from which cells were recorded. Hindlimb cells, for example, were recorded from 8 sites and in 6/8 sites, ICMS evoked hindlimb movements as the threshold response, mostly at the knee. Forelimb cells were recorded from 17 sites in area 4δc and microstimulation was applied in 16 of these. Movements of the forelimb were evoked from 9 sites (6 distal responses and 3 proximal) while hindlimb responses were produced from only 1 site. No responses were evoked from the other 6 sites. The 27 forelimb–hindlimb cells were recorded from 21 sites and the movement of the hindlimbs was evoked from 12/21 of these sites (2 of these also produced forelimb movements at the same threshold as the hindlimb movement). All 12 sites included movements of the ankle and/or knee. A further 4 sites produced movement only of the forelimb and 5 sites were without effect.

Activity of cells during the ipsilateral lead condition

As indicated in Fig. 3 (blue symbols), the majority of cells maintained the same temporal relationship of the period of modified activity to the onset of activity in the coClB during both the contralateral and ipsilateral lead conditions, indicative of a contribution to the control of the contralateral limb regardless of the lead limb. However, the limb-independent cells (red symbols in Fig. 3) clearly showed differences in the pattern of modified activity during the contralateral and ipsilateral lead conditions, indicative of differences in their contribution to the control of motor activity in a given limb in these 2 conditions. In most of these cells, activity was displaced by a phase difference of ~0.5 step cycles as in Fig. 2G, indicative of a contribution to the lead limb, regardless of which limb was the first to step over the obstacle. While the cell illustrated in Fig. 2G discharged during the passage of the forelimb, the cell illustrated in Fig. 5A shows a modification of discharge activity between the passage of the forelimb and the hindlimb over the obstacle. This period of modified activity maintains a constant relationship to the lead limb in the 2 conditions (Fig. 5Aii). Altogether, 8/13 limb-independent cells showed a constant relationship either to the passage of the lead forelimb over the obstacle or to the time between the passage of the lead forelimb and the hindlimb.

In a few cells, there was evidence of more complex activity as illustrated in Fig. 5B. The onset of the task-related period of activity in this cell equally showed a phase difference of ~0.5 with respect to the onset of the period of activity in the coClB between the contralateral and ipsilateral lead conditions (Fig. 5Bi). The end of this period of modified activity, however, occurred at the same time, approximating to the time that the contralateral hindlimb stepped over the obstacle. When triggered on the lead limb in the 2 conditions (Fig. 5Bii), the onset occurs at the same time as the onset of activity in the ClB of the lead limb, but the offset differs, corresponding to its fixed relationship to the onset of activity in the contralateral hindlimb as indicated by the red and green arrows. Four other cells showed qualitatively similar complex periods of modified activity.

This emphasizes that while most cells in area 4δc show changes in activity related to the execution of the step over the obstacle, other cells in this area may be implicated in more complex aspects of the control of gait modifications.

We also recorded a small population of limb-dependent cells in which activity was modified only during the ipsilateral lead condition (Table 2). These cells included a proportion (6/16) in which the discharge activity covaried with the period of activity of the ipsilateral hindlimb. Figure 5C, for example, illustrates such a cell that discharged only in the ipsilateral lead condition (green trace). The timing of this period of discharge coincided with the period of activity in the ipsilateral sartorius (iSrt) which is indicated by EMG burst #3 (in green) in Fig. 5Ci. The fixed temporal relationship between cell activity and the end of the period of activity in the iSrt is illustrated in Fig. 5Cii.

Some cells active only in the ipsilateral lead condition were better related to changes in muscle activity in the contralateral limbs. Figure 5D, for example, illustrates a cell with a receptive field on the contralateral hindlimb. During unobstructed locomotion, the discharge activity in the cell was modulated in phase with the activity of the contralateral sartorius (coSrt), as indicated by the blue traces in Fig. 5Dii. It is likely, therefore, that this cell would be related to changes in activity in the contralateral hindlimb. When the activity of this cell was triggered on the period of activity in the coClB, as in our standard analyses (Fig. 5Di), however, there was no change in activity as the contralateral hindlimb stepped over the obstacle in either the contralateral or ipsilateral lead conditions, but there was a decrease in activity in the preceding step cycle in the ipsilateral lead condition. This decrease in activity is indicated by the arrow and label, B1. When the activity of the cell was synchronized to the onset of activity in the coSrt (Fig. 5Dii), it can be better seen that this decrease in activity (burst B1) occurs in the step cycle preceding the step over the obstacle by the contralateral hindlimb. This pattern of discharge is analogous to that which we previously recorded in some cells in area 4δr (Nakajima et al. 2019) that discharged to the plant of the contralateral forelimb before the passage of the lead, ipsilateral forelimb over the obstacle (see later and Section 4). A small increase in activity (B2) can also be observed during the step over the obstacle by the contralateral hindlimb, not visible in Fig. 5Di because of the effect of smearing. Notably, no cells discharged during the period of activity of the iClB in the ipsilateral lead condition, although this may be because of the relatively small size of the database.

Population activity

Despite the presence of cells discharging at discrete periods during the passage of the contralateral limbs over the obstacle in the contralateral lead condition (Fig. 2), the onset of the periods of modified activity in the overall population of cells forms a continuum, as illustrated in Fig. 6Ai. As also indicated in Fig. 4, few of the cells exhibited SA changes of activity in the contralateral lead condition and only 3 cells modified their discharge earlier than 1 step before the step over the obstacle by the forelimb. This is in marked contrast to area 6iffu which had a large number of cells (74/90, 82%) that showed SA changes of activity in which 45/90 (50%) cells showed such changes of activity more than 2 steps before the step over the obstacle (Nakajima et al. 2019).

The periods of modified activity of the 59 cells that changed their activity when the ipsilateral limb was the first to step over the obstacle showed a similar continuum of activity (Fig. 6Aii). However, in this case, many cells showed a relatively earlier onset of the period of modified activity with respect to the onset of the burst of activity in the coClB than in the contralateral lead condition. Indeed, during the ipsilateral lead condition, changes of activity in 19/59 cells (32%, Table 2) were SA. This population included 10/13 limb-independent cells (see Figs 2G and5A and B).

As illustrated in Fig. 6Aiii, the summed activity from the principal periods of modified discharge (see Methods) during the contralateral lead condition (red trace) covered the period from 1 step before the step over the obstacle to 4 steps after; this is also reflected in the cusum. The peak phase of the activity occurred 1.2 steps after the step over the obstacle. During the ipsilateral lead condition (green lines), the summed activity was phase-advanced with respect to activity during the contralateral lead condition, but the phase of peak activity was similar.

Comparison of population discharge activity from different subregions of area 4

We compared the pattern of modified activity observed in area 4δc with that of 3 populations of cells from other subregions of area 4, namely: 1, cells recorded from the hindlimb region of the primary motor cortex, M1 (4γHL), which included the total population of cells recorded in the publication by Widajewicz et al. (1994); 2, data from the forelimb region of the primary motor cortex (4γFL), which were taken from Yakovenko and Drew (2015) (see also Nakajima et al. 2019); and 3, a population of cells from adjacent area 4δr, which includes the 33 cells detailed in Nakajima et al. (2019) together with 22 additional cells recorded from cats P4 and P5 (Table 1).

As illustrated in Fig. 6 and quantified in Table 4, the distribution of the periods of modified activity in these 3 populations showed both similarities and differences with those described for the population from area 4δc. In the contralateral lead condition, the peak of activity of the summed discharge of the periods of modified activity in the cells in area 4γHL (Fig. 6Bi and iii) occurred 1.8 steps after the onset of the coClB, corresponding to the passage of the hindlimb over the obstacle. Using the same criteria for the classification of cells as used for Fig. 4, 41% of cells changed their activity during the step over the obstacle by the hindlimb, while only 16% of cells discharged during the passage of the forelimb (Table 4). The inverse pattern of activity was observed in 4γFL in which a total of 70% cells modified their discharge activity before and during the passage of the forelimb over the obstacle and only 7% during the passage of the hindlimb. The peak of activity was also relatively early, occurring 0.2 steps after the onset of activity in the coClB/Br. The pattern of activity in area 4δr during the contralateral lead condition was similar to that in area 4γFL, but with a total of 41% of cells (FL-SA and Pre-FL) with an onset of discharge earlier than −0.2 step cycles compared to 17% in area 4γFL. A small proportion of cells in area 4δr changed their activity between the forelimb and hindlimb. Peak activity of the population was also slightly earlier than in area 4γFL, occurring at a phase of −0.1 steps, i.e. just before the onset of activity in the coClB/Br (Fig. 6Diii).

Table 4.

Comparison of discharge patterns in subdivisions of area 4 during coLead.

	4δc (N = 114)	4γFL (N = 60)	4γHL (N = 58)	4δr (N = 55)
Modulated	46 (40%)	55 (92%)	47 (81%)	30 (55%)
Modified	84 (74%)	58 (97%)	57 (98%)	50 (91%)
N: coLead	68 (60%)	55 (92%)	51 (88%)	42 (76%)
HL	15%	7%	41%	10%
FL–HL	40%	11%	41%	12%
FL(all)	40%	70%	16%	62%
Pre-FL	4%	6%	2%	17%
FL(SR)	31%	58%	16%	38%
FL(SA)	9%	11%	0%	24%

Characteristics of the cell discharge in different subdivisions of area 4, including the number and percentage of modulated and modified (task-related) cells together with the number in which discharge was modified during coLead. For the latter, we indicate the percentage of cells discharging at different phases of the step cycle during the coLead condition, as illustrated in Fig. 4A for the population recorded in area 4δc. The cells discharging during the period of activity of the forelimb are divided into those that were SR and those that were SA, as in Fig. 4A. Pre-FL indicates cells in which discharge activity ended < −0.2 step cycle before coClB/Br onset.

In the ipsilateral lead condition, as for area 4δc, periods of modified cell activity in areas 4γHL, 4γFL, and 4δr were generally phase-advanced with respect to those observed in the contralateral lead condition, as can also be observed in the histograms of the summed activity. Moreover, in the ipsilateral lead condition, there was a tendency in the 4γFL population, and particularly in the 4δr population, for the change in discharge activity to begin before the step over the obstacle. However, the characteristics of the changes in discharge in these 2 populations during ipsilateral lead were different. Many of the cells in area 4δr (15/48) started to discharge more than 2 steps before the step over the obstacle and, in most cells, this was the only change in activity. As we have discussed previously (Nakajima et al. 2019), we interpret this activity as being related to the need to control the placement of the contralateral limb in front of the obstacle prior to the step over the obstacle by the ipsilateral limb. In area 4γFL, only 5/54 cells discharged more than 2 steps before the obstacle, and none discharged as early as those in area 4δr. Moreover, in area 4γFL, we also observed multiple smaller bursts of activity prior to the step over the obstacle that may serve to adjust motor activity leading up to the step over the obstacle.

These results show that the pattern of cell discharge activity varies between the 4 subdivisions of area 4 that we studied and that the pattern of activity in area 4δc is intermediate between that observed in the forelimb and hindlimb representations of the primary motor cortex.

A more direct comparison of these 4 areas is shown in Fig. 7, in which the cusums describing the periods of task-related activity of these 4 cortical areas during the contralateral lead condition are superimposed. Both the distributions of the onset of the period of activity of the main burst (Fig. 7A), and of the distributions of the summed activity (Fig. 7B), show the same order of activation. Cells in 4δr are activated initially, followed by cells in 4γFL and then cells in 4δc, with the cells in 4γHL being activated last. Activation of cells in area 4δc is thus intermediate between 4γFL and 4γHL, commensurate with the preponderance of cells discharging between the passage of the forelimb and of the hindlimb. The Kolmogorov–Smirnov test shows that in both Fig. 7A and Fig. 7B, the distribution of the cells in area 4δr is significantly different from that of each of the other 3 regions (P < 0.01 for all pairs), but that there are no significant differences between the other 3 distributions. Restricting the analysis to only the period between the 5% and 95% activation levels for the onset of activity in area 4δc (between −1 step to +2 steps) shows an additional significant difference between the 4γFL and the 4γHL population (P = 0.01) but no differences between these populations and the 4δc population.

Fig. 7.

Distribution of population activity during the contralateral lead condition. A) Superposed cusums from the 4 subregions of area 4 illustrated in Fig. 6 (part i), calculated from the time of onset of the main period of discharge activity. B) Similar plot for the cusums representing the summed activity as illustrated in part iii of Fig. 6.

Cortical projections to area 4δc

We made injections of retrograde tracers limited to area 4δc in 3 cats. Two of these injections were made by using Texas Red (cats Prem9 and Prem10) and the other with Alexa Fluor 488 (cat Prem 7: Table 5). All injections were in the more medial regions of the area.

Data from one example cat (Prem10) are illustrated in Fig. 8. In this cat, a total of 0.45 μL of Texas Red was injected into 2 sites in the medial aspect of area 4δc (Figs 8E and F and9A); there was no spread into either adjacent areas 4γ or 4sfu. Labeling was particularly strong in the more medial regions of the cerebral cortex. Figure 8A and B shows that the major concentration of cells in these medial regions was observed in the dorsal bank of the cruciate sulcus encompassing areas 4fu, 4sfu, 4δc, and 4γHL. The labeling continued onto the posterior sigmoid gyrus where strong labeling was observed in most areas of the primary somatosensory cortex (S1). Strong labeling was also observed within the ansate sulcus, corresponding to areas 5a and 5b of the PPC. There was also weak labeling around the splenial sulcus, corresponding to the cingulate cortex. In the 2 more lateral sections (Fig. 8C and D), the labeling in the caudal bank of the cruciate sulcus remained strong and dense labeling was also found within the rostral bank, corresponding to areas 4δr and 4γFL. There was very little labeling in the ansate sulcus lateral to the lateral sulcus and no labeling in area 7. There was equally very little labeling in any of the subdivisions of area 6 (6aα, 6aβ, 6aγ, or 6iffu) or in area 4δr. Percentage values for the extent of the labeling in different areas can be found in Table 5.

Fig. 8.

Examples of cell labeling observed after injections of Texas Red into area 4δc. A–D) Tracings of parasagittal sections showing cells retrogradely labeled in the frontal cortex at 4 different lateralities, ranging from 2.52 to 6.84 mm from the midline, after injection of 0.45 μL of dextran amine-bound Texas Red into area 4δc in cat Prem10. Scale in (A) pertains also to (B–D). E) One of the 2 injection sites in the medial region of area 4δc, located in the dorsal bank of the cruciate sulcus as indicated in (A). F) Dorsal view of the cat cortex showing the approximate laterality of the 4 sections illustrated in (A)–(D): Red circles indicate the approximate sites of entry for the 2 injections into area 4δc. dlPFC, dorsolateral prefrontal cortex; dmPFC, dorsomedial prefrontal cortex; Lat, lateral sulcus; Prs, presylvian sulcus; SS, suprasylvian sulcus.

The extent of the labeling in this example is shown quantitatively in Fig. 9A and C. In Fig. 9A, the percentage of labeled cells in each 200 μm bin is shown by the color of the filled symbols (see Methods as well as legend and key), while in Fig. 9C the same data are represented as contour plots of the density of the labeling (see Methods). In both illustrations, it can be seen that the highest densities of retrogradely labeled cells are to be found in area 4δc and in the medial regions of adjacent area 4γFL, as well as in S1 and in areas 5a and 5b. This medial region of the PPC corresponds to the region from which cells related to the hindlimb and to forelimb–hindlimb coordination are most prevalent (Lajoie et al. 2010). Labeling in the cingulate cortex was relatively weak (right side of Fig. 9A and C; see also Fig. 8). A very similar pattern of labeling was observed in the other 2 cases that we analyzed as illustrated in Fig. 9B, D, and E. For comparison, the results of an injection in area 6iffu in the same cat as illustrated in Fig. 9E are shown in Fig. 9F (see Nakajima et al. 2019). The 2 injections differ particularly in the regions of cortex within and adjacent to the cruciate sulcus that are labeled, the labeling in area 7, and the density of the labeling in the cingulate cortex.

Corticocortical projections from area 4γ

In a previous study (Andujar and Drew 2007), we made injections of retrograde tracers in multiple regions of the forelimb and hindlimb representations of the motor cortex (area 4γ) and examined the projections from the PPC to the injected regions. Data on the projections from subregion 4δc to these same regions were available but were not quantitatively analyzed at the resolution necessary to determine the nature of the connections with these regions. However, the illustrations from that manuscript clearly show retrogradely labeled cell in regions corresponding to area 4δc following injections into both 4γFL and 4γHL (Figs 6,9, and10 in Andujar and Drew 2007). For the current study, we examined this projection in more detail in 1 case in which we made large injections of dextran amines into electrophysiologically defined regions of both 4γFL and 4γHL. The results from this cat are compatible with those illustrated qualitatively in our previous publication.

ICMS (11 pulses, each of 0.2 ms at a frequency of 330 Hz) was used to identify the forelimb and hindlimb representations of area 4γ, as illustrated in Fig. 10A and B. Because of the level of the anesthesia, we needed to use currents of ~100 μA to evoke clearly reproducible responses at each site. In penetrations that entered the cortex in the most medial regions of the posterior sigmoid gyrus and that were targeted to the hindlimb representation of M1 within the dorsal bank of the cruciate sulcus, ICMS evoked responses in all 3 hindlimb flexors that we recorded (see legend) but in none of the forelimb flexor muscles. Conversely, stimulation in the anterior sigmoid gyrus evoked responses in all 3 forelimb flexor muscles that we recorded and in none of the hindlimb flexor muscles. After identifying the forelimb and hindlimb regions, we made injections of 1.9 μL of Texas Red distributed among 4 sites in the forelimb representation and 2.0 μL of Alexa Fluor 488 distributed among 4 sites in the hindlimb representation. These large injections were designed to provide an overview of the spatial organization of the projections from 4δc to the forelimb and hindlimb regions of the primary motor cortex.

The 3 tracings illustrated in Fig. 10C–E are taken from different lateralities of the pericruciate cortex (see dorsal view in the inset of Fig. 10E) and show the presence of multiple labeled cells within and surrounding the cruciate sulcus. In medial regions (Fig. 10C), there was labeling from the hindlimb representation (green symbols) particularly within area 4γ. This gave way to predominant labeling from the forelimb representation (red symbols) within area 4δc. There was then mixed labeling from both forelimb and hindlimb representations within area 4δc. In addition, there was labeling in area 5 in the ansate sulcus, predominantly from the hindlimb representation, and labeling from the forelimb representation in areas 6iffu, 6aα, and 6aγ. This pattern of labeling was maintained at intermediate lateralities of the cortex (Fig. 10D) with, however, a more intermingled representation of the forelimb and hindlimb within area 4δc. More laterally (Fig. 10E), cells labeled from the 4γHL injection were still observed in 4γ in the caudal bank of the cruciate sulcus, while in area 4δc most cells were labeled from the 4γFL injection. Cells in area 4δr were also labeled from the forelimb representation. Labeling in the posterior bank of the ansate sulcus (area 5) was also mostly from the injection into 4γFL at this laterality.

A summary of the regions of cortex labeled by the 2 injections is illustrated in Fig. 10F and G. The injection into the forelimb representation of area 4 (Fig. 10F) produced widespread labeling throughout the cruciate sulcus and the surrounding cortical regions. The bins with major concentrations of labeled cells (red symbols, 80%–100% percentiles, see key in Fig. 9A) were located in areas 4γ (mostly lateral and rostral), in area 4δ, and in lateral regions of area 5. However, weaker labeling was also observed throughout areas 6iffu, 6aα, and 6aγ. Also note the lack of labeling in the caudomedial aspect of the pericruciate sulcus, corresponding to the hindlimb representation of area 4γ. The injection into the hindlimb representation (Fig. 10G) caused the densest labeling in medial regions of area 4δc, in the adjacent regions of area 4γ, and in the medial regions of area 5. There was also heavy labeling at the lateral margin of the cruciate sulcus corresponding to that part of area 4γ in which corticospinal cells projecting to the lumbar regions of the spinal cord were found (see Fig. 13, Fortier-Lebel et al. 2021). Figure 10H synthesizes these data by superimposing the bins with heaviest labeling (≥80%, see key in Fig. 9A) from the forelimb (red symbols) and hindlimb (green symbols) representations. This shows that although cells in area 4δc which were labeled from the forelimb and hindlimb representations were intermingled, the major concentration of cells projecting to 4γHL was medial to that projecting to area 4γFL. This is compatible with the tendency for cells discharging during steps over the obstacle with the forelimb to be located medially to those discharging during steps over the obstacle with the hindlimbs (Fig. 4B).

Discussion

The results from this study provide the first information on the functional characteristics of cells recorded in area 4δc, lying in the caudal bank of the feline cruciate sulcus. We show that a large proportion of cells in this area modified their discharge activity during the step over the obstacle, including a substantial proportion of cells that modified their discharge between the passage of the contralateral forelimb and hindlimb over the obstacle. We suggest that these latter cells contribute to the coordination of the forelimb and hindlimb during voluntary gait modifications. We further suggest that the rostral and caudal regions of area 4δ, which have been sometimes treated as distinct regions, should be considered as a single functional area with an important role in regulating interlimb coordination.

Contribution of area 4δc to the regulation of interlimb coordination

We posited in Section 1 that cells involved in directly coordinating forelimb and hindlimb activity during gait modifications should show changes in activity between the passage of the forelimbs and the hindlimbs that are largely limb-dependent. Many of the cells recorded in area 4δc showed both of these characteristics, being activated primarily during steps over the obstacle, and then only when the contralateral limb led. This pattern of activation is consistent with a function in assuring that the modification of the activity in the hindlimb is coordinated with that of the forelimb and adapted to take into account the passage of the obstacle under the body.

At the same time, many of these same cells showed no modulation of their discharge activity during unobstructed locomotion and changed their discharge activity only during the steps over the obstacle. This suggests that such cells in area 4δc are primarily involved in forelimb–hindlimb coordination only when there is a need for modification of the baseline pattern, as in this task when cats step over an obstacle. Moreover, although a majority of cells discharged between the passage of the forelimb and hindlimb, there was also a substantial percentage of cells that discharged either during the passage of the forelimb over the obstacle or only during the passage of the hindlimb. As such, area 4δc should not be viewed as an area that contributes only to interlimb coordination but rather as an area which regulates interlimb coordination as one of its functions.

In addition, despite the preponderance of limb-dependent signals, a smaller population of cells in area 4δc showed a limb-independent modification of activity, showing activity related to the passage of the lead limb over the obstacle. This suggests that while this area is involved primarily in determining the motor pattern in the contralateral limbs (as in the primary motor cortex), it might make an additional contribution to more global aspects of the gait modifications, such as limb selection or limb-independent representations of the position of the obstacle with respect to the body (as in area 5b, Lajoie et al. 2010). Possibly, the limb-independent cells might reflect an intermediate contribution to the transformation of a global signal to a limb-dependent one.

When considering the function of area 4δc, it should be emphasized that there are few, if any, anatomical connections between the forelimb and hindlimb representations of the primary motor cortex in either the cat or the primate (Fig. 10H, Ghosh 1997a; Hatanaka et al. 2001, see also Capaday et al. 2009). Area 4δc therefore provides a potentially important anatomical substrate for coordinating activity between the forelimb and the hindlimb. It receives projections from area 4γFL that provide information on the state of the forelimb, and it projects to area 4γHL which regulates muscle activity as the hindlimb is brought over the obstacle (Widajewicz et al. 1994). In addition, it also receives a large input from the medial regions of the PPC, providing information on the passage of the obstacle under the body (see below). We suggest that area 4δc integrates these sources of information to ensure that activity in the hindlimb is appropriately timed and scaled to that of the forelimb. In this view, the motor command that modifies the hindlimb activity is a function of the corticocortical projections from area 4δc to area 4γHL.

However, we cannot completely rule out other possibilities. For example, it is possible that area 4δc might exert an effect directly on spinal circuits via its corticospinal projections, although these projections are relatively sparse (Ghosh 1997c). Equally, it might act through the pontomedullary reticular formation, which is also involved in interlimb coordination (Drew et al. 1986; Drew 1991; Bem et al. 1995; Brustein and Rossignol 1998; Prentice and Drew 2001), although again the projections to the brainstem from the caudal bank of the cruciate sulcus are sparse (Rho et al. 1997), making this possibility unlikely.

A unified role for interlimb coordination in areas 4δc and 4δr

Ghosh (1997a) divided cytoarchitectonic area 4δ into 2 parts, largely on the differences in strength and pattern of their respective anatomical connections with other cortical areas, including area 4γ. However, there are several reasons to suggest that they should be considered as a single functional area. In addition to the fact that the cytoarchitecture is the same in the 2 subdivisions, microstimulation at rest and during locomotion produces similar motor activation, although with a stronger bias toward the contralateral forelimb in area 4δr (Fortier-Lebel et al. 2021). In addition, although the relative strength of the corticocortical connections to areas 4δc and 4δr shows differences, qualitatively they are very similar (Ghosh 1997a).

On the other hand, the separation of areas 4δc and 4δr is, at first sight, supported by some of the differences in cell characteristics in the 2 subdivisions. For example, while cells related to the passage of the forelimbs over the obstacles are found in both 4δc and 4δr, other cell types are concentrated in one area or the other. Thus, in area 4δc, a large proportion of cells discharged between the passage of the forelimbs and the hindlimbs over the obstacle, while such cells were less frequent in area 4δr. In contrast, in area 4δr, we have detailed the presence of a population of cells that discharge when the contralateral forelimb is placed on the support surface in advance of the step over the obstacle by the ipsilateral forelimb (Nakajima et al. 2019) while such cells were rare in area 4δc.

We suggest that a unified view of the apparent differences in cell activity in these 2 areas can be reconciled if one considers that the underlying function in each subdivision is one of coordination of limb activity in preparation for the upcoming step over the obstacle. Thus, the cells recorded in area 4δr can be viewed as coordinating the activity in the contralateral forelimb to ensure the appropriate step over the obstacle by the ipsilateral limb. The cells in area 4δc discharging between the passage of the forelimb and the hindlimb are more clearly interpreted as coordinating contralateral hindlimb activity on the basis of the passage of the contralateral forelimb. In addition, cells in both 4δc and 4δr participate in regulating the passage of the contralateral forelimb and hindlimb over the obstacle. Moreover, the data in Fig. 5D raise the possibility that some cells in area 4δc might control placement of the hindlimb in the step before the step over the obstacle in the same way as cells in area 4δr for the forelimb. As such, cells in areas 4δc and 4δr might be considered as having more general features in common than not. Area 4δr and the lateral regions of 4δc are more related to activity in the forelimb and medial 4δc is more related to activity in the hindlimb. In this view, area 4δ as a whole can be considered to have a role in coordinating interlimb activity in situations in which the normal pattern of activity during unobstructed locomotion has to be adapted to maneuver around or over obstacles on the basis of visual information.

Area 4δ as part of the cortical network controlling locomotion

Figure 11 summarizes the major cortical projections to and from areas 4δc and 4δr and places these areas into context with other structures that have been demonstrated to participate in the control of visually guided locomotion. In the figure, we represent regions that are more related to control of the hindlimb in red and those more related to the forelimb in blue. The shading indicates that in several regions there is not a clear demarcation between forelimb and hindlimb representations but rather a graded change from one to the other. Major inputs to area 4δc originate from surrounding regions of the motor cortex, from S1 and from area 5. In the case of S1, the inputs are from areas mostly related to hindlimb function, and in the case of area 5, they are from areas that we have previously shown to be involved in hindlimb and in forelimb–hindlimb coordination (Lajoie et al. 2010). This input from the medial regions of area 5 might be particularly important in transferring information to area 4δc on the predicted location of the obstacle as it passes underneath the body and may be a major source of input driving the cell activity in area 4δc.

Fig. 11.

Principal corticocortical projections to area 4δ. We illustrate the principal connections from selected frontoparietal areas and from cingulate cortex to area 4δc. Only those projections quantified in Table 5 are presented. For comparison, we also present the principal inputs to area 6iffu as determined in a previous publication (Nakajima et al. 2019). In each block, the color indicates hindlimb (red) and forelimb (blue) representations in each of the indicated cortical areas in which somatotopy has been established. Note that for areas 5b, 4δc, and 4δr, we use gradations of blue and red to indicate that somatotopy is not as well developed as in M1 and S1. The direction and the thickness of the arrows gives an approximation of the relative strength of different connections and are based on our own observations (Table 5) as well as those of Ghosh (1997a).

Moreover, the more lateral regions of area 4δc receive similar input from the more lateral regions of area 5 (Ghosh 1997a), where cells discharge before and during steps over the obstacles (Lajoie et al. 2010; Drew and Marigold 2015; Marigold and Drew 2017). As such, both regions of area 5 could contribute to the production and/or regulation of the signals observed in cells in area 4δc. Possibly, the inputs from the more lateral regions of area 5 to the more lateral regions of area 4δc, in which forelimb-related cells are more concentrated, may make a similar contribution to determining the initiation of the step over the obstacle by the forelimb (Andujar et al. 2010; Drew and Marigold 2015; Marigold and Drew 2017).

Moreover, the presence of limb-independent cells in area 4δc suggests that planning is not a serial process passing from area 5b to area 4δc, but rather one in which some activities occur in parallel. Possibly, the transformation from a global signal to a muscle-related one arises through a combination of gating through strongly limb-dependent rhythmical inputs from S1 and surrounding M1, together with transmission of the parietal signals via either corticocortical or subcortical structures. Although the mechanisms remain unclear, the results do suggest that the predominantly limb-independent cell discharge in area 5b is transformed into a predominantly limb-dependent signal in area 4δc and that this latter discharge contributes particularly to the control of interlimb coordination via its connections to the primary motor cortex.

We cannot rule out the possibility that information to area 4δc is transmitted from regions of area 6, including area 6iffu, in which we have also detailed cells discharging in advance of and during the step over the obstacle (Nakajima et al. 2019). However, projections from these areas are relatively weak (Table 5; Ghosh 1997a), suggesting that area 6iffu is not the major determinant of the pattern of activity observed in area 4δc.

Comparison to the primate

Most studies on the cortical control of locomotion, and particularly visually guided locomotion have been performed on cats (refs in Section 1, see also Beloozerova and Sirota 1993, 2003) and, to a lesser extent, rodents (DiGiovanna et al. 2016; Miri et al. 2017; Laflamme et al. 2019; Omlor et al. 2019). Data from single unit recording studies in nonhuman primates are limited and come mostly from recordings in primary motor cortex during unobstructed locomotion directed at studying the utility of such signals for controlling neuroprosthetic devices (e.g. Fitzsimmons et al. 2009; Capogrosso et al. 2016; Xing et al. 2019). Few studies have examined activity in secondary motor areas in primates during locomotion (Foster et al. 2014). The body of work on cats therefore potentially provides an important database when considering neuroprosthetic controllers aimed at higher level processes necessitating modification of gait rather than simply producing a rhythmical output. However, in transferring data from one species to another, and especially primates, the question arises as to how similar the control mechanisms are in cats and primates and whether one can make inter-species analogies between different cortical areas. While there are undoubtedly species-specific differences, the available literature from nonhuman primates (see above) and from human studies (see, e.g. Petersen et al. 2001; Petersen et al. 2012) suggest that the functional characteristics of the motor cortical control of locomotion hold across species. Similarly, electroencephalographic recordings from walking humans (Wagner et al. 2014, 2016) and imaging studies during imagined locomotion (Malouin et al. 2003; Wang et al. 2009) suggest that the overall cortical network involved in locomotion control (Drew and Marigold 2015) is also similar across species. As such, comparing the results from experiments in cats with those in primates is essential for cross-species transfer of information that is currently impossible to obtain in humans and unavailable in the primate literature.

With respect to the current study, Ghosh 1997a, 1997c, on the basis of his microstimulation and anatomical studies, proposed that area 4δ might be the feline equivalent of the primate SMA. In particular, the presence of a corticospinal projection (Ghosh 1997c), the presence of a substantial projection from area 4δc to area 4γ, and the presence of a representation of the forelimb and hindlimb in this area are all similar to the properties of primate SMA (Mitz and Wise 1987; Dum and Strick 1991, 1992; Luppino et al. 1991, 1993, 1994; He et al. 1995; Hamadjida et al. 2016).

A major question then is the extent to which the discharge properties of cells in area 4δ are similar to those described for cells recorded from the SMA of primates. We have addressed this issue for area 4δr in a previous publication (Nakajima et al. 2019), arguing that the general properties of cells in this subdivision of area 4δ are compatible with the properties of cells recorded in primate SMA. Our present results suggest that this is equally true for the cells recorded in area 4δc. Thus, at the simplest level, cells in both area 4δ and SMA show changes in discharge activity that occur just before and during a voluntary movement (Tanji and Kurata 1979; Matsuzaka and Tanji 1996; Tanji and Mushiake 1996). Similar changes have been described for the hindlimb, although no cells discharging to both forelimb and hindlimb movements were observed (Tanji and Kurata 1982). Moreover, we also have a small population of cells that discharge in relationship to activity in the ipsilateral limb during the ipsilateral lead condition (see Tanji et al. 1988), although we only found cells related to the ipsilateral hindlimb, and not to the ipsilateral forelimb, perhaps a result of insufficient sampling.

With respect to the cells involved in forelimb–hindlimb coordination in area 4δc, it is difficult to claim similarities in discharge properties between cat area 4δ and primate SMA as no one has recorded single-cell activity during coordinated movements of the forelimbs and hindlimbs in this area in the primate. However, a study using functional magnetic resonance imaging in subjects making coordinated movements of the wrist and ankle did report increased activity in the SMA (Debaere et al. 2001). In addition, anatomical (He et al. 1995; Hatanaka et al. 2001) and microstimulation (Luppino et al. 1991) studies show that there is a small area of overlap in forelimb and hindlimb representations in primate SMA, where coordination between the forelimb and hindlimb might be favored. However, it should be emphasized that cells that are proposed to be involved in forelimb–hindlimb coordination exert a motor effect only on the hindlimbs. As such, it is quite possible that cells with a function in forelimb–hindlimb coordination might be found in areas in which microstimulation reveals representation of only the hindlimb, so that overlap of forelimb and hindlimb output areas is not necessarily indicative of, or necessary for, a contribution to forelimb–hindlimb coordination.

One characteristic that separates SMA from M1 is the presence of cells that signal more complex aspects of a task including differential responses to an instruction, context dependence, and the sequence in which movements are performed (Shima and Tanji 1998; Tanji 2001; Nakajima et al. 2013; Russo et al. 2020). In fact, pharmacological inactivation and ablation of SMA have been shown to cause deficiency in performing motor sequences (Shima and Tanji 1998) and bimanual coordination (Brinkman 1984) in primates. Although we could not manipulate our task in the same manner as in many primate experiments, we consider that the cells that we recorded in area 4δ that discharge with respect to the placement of the contralateral limb in anticipation of the step over the obstacle by the ipsilateral limb are such an example of a more complex discharge pattern. Cells in primary motor cortex overwhelmingly discharge maximally during the step over the obstacle by the contralateral limb. The cells in area 4δr discharge in the step before, and we have suggested that this is not to increase the level of activity in the limb but rather to ensure that it is appropriately placed to allow the gait modification by the ipsilateral limb. We would argue that this is a cell property that is qualitatively similar to that observed in the SMA. Indeed, in some cells in area 4δr, their only change in discharge occurred in this specific condition (Nakajima et al. 2019), supporting the view that activity in some cells is highly context-dependent. Similarly, some of the cells in area 4δc showing limb-independent modifications of activity might also be considered as having complex patterns of activity, in that whether the discharge covaries with activity in either the contralateral or ipsilateral limb depends on which limb is the first to step over the obstacle.

Although one may also suggest analogies to studies that have examined the contribution of SMA to bilateral movements (Kermadi et al. 1998; Kazennikov et al. 1999; Donchin et al. 2002) of the limbs or to comparisons of discharge depending in which limb movements were instructed (Hoshi and Tanji 2004), the nature of the tasks is different. In the bimanual tasks in primates, the 2 limbs are generally controlled to produce a single movement to which both limbs contribute. In studies comparing discharge activity in movements of the contralateral or ipsilateral limbs, only one limb moves at a time. However, in locomotion, even if the activity between the limbs is coordinated, the movement of each limb has to be sequential, and in this condition, activity is biased to the contralateral limb.

In addition to cell properties, there are also some concordances in the anatomical connections of the 2 areas. For example, as shown for the SMA (Luppino et al. 1993; Fang et al. 2005; Hamadjida et al. 2016), area 4δ receives input from area 5 of the PPC and from M1 (see also Ghosh 1997a). However, while SMA in the primate also receives input from the cingulate cortex, the dorsal and ventral premotor areas (PMd and PMv) (Luppino et al. 1990; Luppino et al. 1993; Wang et al. 2001), in the cat, the inputs to area 4δ from these regions are relatively weak (Table 5). In addition, we identified a strong input from S1 that does not appear to be found in SMA in the primate, although a small projection was observed after injections into areas 3a and 3b in the marmoset (Krubitzer and Kaas 1990; Huffman and Krubitzer 2001). Although we cannot discount the possibility that the labeling of cells in S1 in our study may result from contamination from the injection needle that passed through S1, we saw no evidence of a halo of diffuse tracer in S1 nor any indication of labeled axons leaving this region. Moreover, injections into area 6iffu (Nakajima et al. 2019), which equally traversed S1, did not lead to labeled cells in this region.

Combined with previous studies, the cellular activity properties during locomotion and the corticocortical connectivity shown in the present study strengthen the view that feline area 4δ shows several analogies with the primate SMA. Further approaches, including single-unit recordings in primate SMA during locomotion and local inactivation of area 4δ in cats, are needed to substantiate this view.

Conclusion

We present arguments in this manuscript to suggest that area 4δc, together with its cytoarchitectonic continuation onto the rostral bank of the cruciate sulcus, area 4δr, should be considered as a single functional region, involved in coordinating limb activity during visually guided gait modifications. We have suggested 2 specific situations in which area 4δ may make an essential contribution. The first of these is in adjusting the position of the contralateral forelimb, the plant limb, in front of the obstacle so as to permit the passage of the ipsilateral forelimb in the ipsilateral lead condition. The second is in assuring that the hindlimb gait modification is activated at the appropriate time following the passage of the forelimb over the obstacle in the contralateral lead condition. In most cells, this activity is lateralized to control the execution of movements of the contralateral limb, in contrast to activity in the PPC (Andujar et al. 2010; Drew and Marigold 2015; Marigold and Drew 2017) and area 6iffu (Nakajima et al. 2019), which provide more global, nonlateralized, information.

Although we are using the cat as a model for studying secondary motor areas, requirements for interlimb coordination are as important in primates as they are for cats. Many primates, such as macaques, adopt quadrupedal gaits (see, e.g. Mori et al. 1996) and face the same challenges during overground locomotion as cats. Moreover, as we have previously suggested (Lajoie et al. 2010), cortical areas, such as the PPC and area 4δ that are involved in interlimb coordination during locomotion, may well have evolved in primates to equally control the coordination of activity in the forelimb and hindlimb based on the need to control the appropriate sequence and timing of activity required for visually guided arboreal locomotion.

Acknowledgments

We would like to acknowledge the assistance of T. Ariel, M. Bourdeau, N. De Sylva, P. Drapeau, F. Lebel, and J. Soucy for technical assistance in the performance and analysis of these experiments. We thank Dr Stephane Menard for his excellent veterinary assistance and acknowledge the participation of Nabiha Yahiaoui in some of these experiments. We thank Drs Elaine Chapman and Numa Dancause for helpful comments on this manuscript.

Funding

This work was supported by an operating grant (PJT-156281) from the Canadian Institutes of Health Research (TD) and JSPS KAKENHI Fostering Joint International Research (A) Grant Number JP19KK0414 (TN). NF-L received a studentship from the Fonds de Recherche Santé, Quebec (FRQS).

Conflict of interest statement. None declared.