Abstract
Recovery of motor function after stroke is accompanied by reorganization of movement representations in spared cortical motor regions. It is widely assumed that map reorganization parallels recovery, suggesting a causal relationship. We examined this assumption by measuring changes in motor representations in eight male and six female squirrel monkeys in the first few weeks after injury, a time when motor recovery is most rapid. Maps of movement representations were derived using intracortical microstimulation techniques in primary motor cortex (M1), ventral premotor cortex (PMv), and dorsal premotor cortex (PMd) in 14 adult squirrel monkeys before and after a focal infarct in the M1 distal forelimb area. Maps were derived at baseline and at either 2 (n = 7) or 3 weeks (n = 7) postinfarct. In PMv the forelimb maps remained unchanged at 2 weeks but contracted significantly (−42.4%) at 3 weeks. In PMd the forelimb maps expanded significantly (+110.6%) at 2 weeks but contracted significantly (−57.4%) at 3 weeks. Motor deficits were equivalent at both time points. These results highlight two features of plasticity after M1 lesions. First, significant contraction of distal forelimb motor maps in both PMv and PMd is evident by 3 weeks. Second, an unpredictable nonlinear pattern of reorganization occurs in the distal forelimb representation in PMd, first expanding at 2 weeks, and then contracting at 3 weeks postinjury. Together with previous results demonstrating reliable map expansions in PMv several weeks to months after M1 injury, the subacute time period may represent a critical window for the timing of therapeutic interventions.
SIGNIFICANCE STATEMENT The relationship between motor recovery and motor map reorganization after cortical injury has rarely been examined in acute/subacute periods. In nonhuman primates, premotor maps were examined at 2 and 3 weeks after injury to primary motor cortex. Although maps are known to expand late after injury, the present study demonstrates early map expansion at 2 weeks (dorsal premotor cortex) followed by contraction at 3 weeks (dorsal and ventral premotor cortex). This nonlinear map reorganization during a time of gradual behavioral recovery suggests that the relationship between map plasticity and motor recovery is much more complex than previously thought. It also suggests that rehabilitative motor training may have its most potent effects during this early dynamic phase of map reorganization.
Introduction
Stroke remains one of the major causes of death and disability worldwide (Gorelick, 2019). Although advances have been made in acute stroke therapy, interventions generally are limited to a few hours following stroke (Hollist et al., 2021). Delays in treatment often result in chronic motor impairments requiring intensive physical rehabilitation and restorative therapies to improve function. Animal models remain useful for addressing underlying mechanisms of recovery and offer therapeutic targets to develop or refine current rehabilitative therapies (Bundy and Nudo, 2019). Principally, a large effort has been made to understand the capacity of the brain to reorganize after an ischemic injury and how therapies can modulate plasticity processes to facilitate recovery.
Intracortical microstimulation (ICMS) techniques (Asanuma and Sakata, 1967) have been used frequently to derive highly detailed maps (<0.5 mm resolution) of spared motor representations after injury, demonstrating cortical plasticity in intact animals after acquisition of new motor skills (Nudo et al., 1996a) or in brain-injured animals following natural motor recovery or recovery aided by rehabilitative training, pharmacological treatment, or neuromodulatory treatment (Nudo et al., 1996b; Nudo and Milliken, 1996; Frost et al., 2003; Conner et al., 2005; Gharbawie et al., 2005; Piecharka et al., 2005; Dancause et al., 2005, 2006b; Porter et al., 2012; Wyss et al., 2013; Barbay et al., 2015; Nishibe et al., 2015; Okabe et al., 2016; Plautz et al., 2016). In most animal studies, postinjury ICMS motor maps are derived after motor recovery has become stable, typically several weeks to months after injury. The results of such studies are consistent across laboratories and in both rodent and nonhuman primate models, demonstrating varying degrees of expansion in spared premotor areas ipsilateral to the injury (Bundy and Nudo, 2019). The potential causal role of spared motor map expansion has been strengthened by demonstrations that secondary injury to the reorganized cortex can reinstate motor deficits after recovery (Murata et al., 2015; Okabe et al., 2016; S. Y. Kim et al., 2018; Y. K. Kim et al., 2018).
Relatively few studies have examined changes in spared motor maps in the first few weeks after primary motor cortex (M1) injury, that is, the period of rapid functional improvement. Those that have been done suggest that map expansion is delayed relative to improvements in behavioral function. For example, 2.5 weeks following an ischemic infarct in M1 of rats, a time when behavioral performance had already reached a plateau, the forelimb representation in the premotor cortex was decreased to about half its size relative to that of healthy rats (Nishibe et al., 2015). Map contractions occurred regardless of whether rats received rehabilitative training. Subsequent maps derived ∼5 weeks postinjury were substantially expanded compared with early maps and within the range of intact rats but only in rats that received rehabilitative training. Thus, map expansions in secondary motor areas clearly appear to be dependent on rehabilitative training, but their role in functional recovery is obscured because of the delayed reorganization.
The purpose of the present study in a large cohort of nonhuman primates was to determine whether (1) M1 lesions result in early contraction of spared premotor maps (as they do in rats), and (2) whether premotor map reorganization is temporally related to functional recovery.
Materials and Methods
Fourteen adult squirrel monkeys (genus Saimiri) were used in this study. Animals were assigned randomly to one of two experimental groups (2-Week Group, 3-Week Group; n = 7 per group) that differed with respect to the timing of the postinfarct cortical mapping procedure (see below). Eight adult monkeys were male (weight range, 811–1111 g; mean, 942 g) and six were female (597–700 g; mean, 640 g). Both experimental groups contained males and females (2-Week Group, 5 males, 2 females; 3-Week Group, 3 males, 4 females). Monkeys were experimentally naive and free of obvious physical and neurologic problems at the time of study initiation. All procedures were conducted in accordance with federal guidelines for the care and use of laboratory animals and with University of Kansas Medical Center Insttutional Animal Care and Use Committee approval.
Experimental design
Hand preference on a pellet retrieval task was determined according to the procedures of Plautz et al. (2000). Then monkeys underwent 2 weeks of behavioral training on a pellet retrieval task (10 training sessions/week, ∼400–500 trials/d). The following week, monkeys underwent a surgical procedure to expose regions of the motor cortex in the hemisphere opposite the preferred hand. ICMS techniques were used to derive functional maps of movement representations within primary motor cortex (M1) and its subdivisions dorsal premotor cortex (PMd) and ventral premotor cortex (PMv). In the same procedure, an ischemic infarct was created within M1 that included ∼100% of the distal forelimb (DF) representation (i.e., digit, wrist, and forearm movements) and a portion of the adjacent proximal forelimb representation (i.e., elbow and shoulder movements). Monkeys were recovered from anesthesia and returned to their home cages. Limited behavioral testing on the pellet retrieval task was performed during the postinfarct period (two test sessions/week, ≤50 trials/test) to track behavioral deficits and recovery. At either 2 weeks (n = 7) or 3 weeks (n = 7) postinfarct, monkeys underwent a second surgical procedure to rederive motor maps in M1, PMd, and PMv. Motor representations from baseline (preinfarct) maps were compared with maps derived at 2 and 3 weeks postinfarct.
Surgical procedures
Surgical procedures were conducted according to the methods of Plautz et al. (2000, 2016). Briefly, monkeys were anesthetized with ketamine (20 mg/kg, i.m.), the trachea intubated, and the saphenous vein catheterized. Atropine (0.1 ml/kg, i.m.), dexamethasone (0.5 mg/kg, i.v.) and penicillin (15 000 units/kg, s.c.) were administered. Surgical anesthesia was produced using inhaled halothane (2–3% induction, 1–2% maintenance) in a mixture of 75% nitrous oxide and 25% oxygen delivered at 1 L/min via a nonrebreathing gas scavenging circuit, and monkeys were secured in a stereotaxic frame. Under aseptic conditions, a craniotomy was performed over an ∼1.5 × 1.5 cm region of the frontal lobe, and the dura was excised to expose the primary and premotor cortex. Warm (37°C) mannitol (25% solution, 8 ml/kg, i.v., before craniotomy and 3–4 ml/kg, i.v., supplementally as needed) was administered to control cortical edema. A plastic cylinder was attached to the skull around the cranial opening with dental acrylic and filled with warm (37°C) silicone oil to prevent desiccation of the cortex. After completion of the surgical opening, gas anesthesia was withdrawn and ketamine (∼20 mg/kg/h, i.v.) supplemented with diazepam (∼0.01 mg/kg/h, i.v.) were administered as needed to maintain a stable anesthetic state for ICMS mapping procedures (see below). Throughout the surgical and mapping procedures, vital signs (heart rate, oxygen saturation, respiration rate, expired carbon dioxide, and core body temperature) were monitored and maintained within normal ranges, and intravenous fluids (3% dextrose in lactated Ringer's solution, 10 ml/kg/h) were continuously delivered. At the conclusion of the ICMS mapping procedure, gas anesthesia was reinstated, the plastic cylinder removed, and the cortical infarct induced (see below). Then, the excised dura was replaced with a thin (0.007 inch thick) silicone sheet cut/shaped such that its edges extended beneath the cut edges of the dura, thus preventing formation of tissue and vascular anastomoses. The craniotomy was closed with dental acrylic applied over either the original bone flap or over a thin piece of gelfoam shaped to the cranial opening. The scalp incision was sutured and treated with Marcaine (local anesthetic) and a topical antibiotic, and penicillin and dexamethasone were administered again. Monkeys were removed from the stereotaxic frame, gas anesthesia was withdrawn, and monkeys were monitored in a temperature-controlled and humidified incubator until alert and then were returned to their home cages. Total duration of anesthesia was typically 12–16 h, with a 1–3 h duration for anesthetic recovery.
ICMS procedures
ICMS procedures were conducted as in Barbay et al. (2015) to derive detailed maps of movement representations in the distal and proximal forelimb areas of M1, PMd, and PMv. A digital image of the cortical surface was acquired, imported into a graphics program, and overlaid with a 250 µm grid so that grid intersections could be used to guide microelectrode penetrations. Under ketamine anesthesia, a microelectrode (tapered and beveled glass micropipette, 10–25 µm o.d. tip, filled with 3.5 m NaCl) was introduced perpendicularly into the cortex to a depth of ∼1750 µm (cortical layer V) based on grid intersection points, adjusted for presence of surface vasculature. At each site, the ICMS stimulus was applied (13 monophasic pulses, 200 µs pulse width, 3.3 ms interpulse interval; pulse train repeated once/s) and the movement(s) evoked at threshold current levels (restricted to ≤30 µA) were determined. Evoked movements (joint and direction of movement) were described with conventional terminology. Movements included digits (one or more fingers and/or thumb), wrist, forearm, elbow, shoulder, face (jaw, tongue, vibrissa, nose, pinna, etc.), trunk (chest, back), hindlimb (hip, knee, ankle, toes), and tail. Evoked movements and current thresholds were defined by one investigator and confirmed by at least one additional investigator. Occasionally, two distinct movement types could be detected at the same stimulation site. When these movements had thresholds within 2 µA of each other, they were considered to co-occur and were classified as dual-response sites. Dual-responses were included in areal measurements of digit and wrist/forearm for initial analysis. A secondary areal analysis was then conducted for exclusive dual site areas. Site interpenetration distances were 500 µm throughout M1 and PMd and were 350 µm in PMv. Mapping continued until all DF representations (digit, wrist, forearm) were fully surrounded by proximal representations (i.e., any movement not including a DF response) and/or nonresponsive sites. Typically, surrounding borders were extended 500–1000 µm from the closest DF response to ensure that all DF sites were identified; ∼250–350 ICMS sites were performed during each mapping session (Fig. 1). In each of the second (postinfarct) mapping procedures, evoked movements were defined by investigators who were blind to the intermap interval (i.e., whether the second map was derived at week 2 or week 3 postinfarct).
Cortical infarct procedure
At the conclusion of the first surgical/mapping procedure and based on ICMS mapping data, an intended infarct region was defined that included 94–100% of the M1 DF representation, extending ∼500 µm beyond the edge of the DF map and thus also included a portion of the proximal forelimb representation (Fig. 1). The ischemic infarct was produced by bipolar electrocoagulation of the pial vasculature within the intended target region (Plautz et al. 2016). Small arterioles and veins were occluded. Only large-diameter vessels that passed-through the target region to serve other cortical areas were spared, whereas all branches from these vessels that served the target region were occluded up to their most proximal branch point from the larger vessel (Fig. 2). This infarct technique allows for precisely controlled borders and takes advantage of the radial penetration of pial vessels to produce a focal region of cortical damage extending through the depths of the gray matter while sparing underlying white matter. Because of constraints on occludable vasculature in relation to the location of DF sites, it was not always possible to target 100% of the DF representation in each monkey (Table 1). The intended infarct was defined conservatively on the caudal side of M1 to minimize potential inclusion of somatosensory area 3a (i.e., <250 µm from caudal edge of DF representation). A digital image of the cortex was acquired after infarct creation to compare the actual versus intended extent of the infarct.
Verification of infarct
The infarct location and effectiveness were evaluated in three ways. First, as noted above, an immediate postinfarct image was acquired and compared with the intended target region. Following the infarct procedure, the ischemic region is characterized visibly by the absence of intact vasculature and blanching (loss of color) within the cortical tissue. Second, during the second mapping experiment, ICMS sites were performed within the infarcted region, with the absence of evocable movements taken as verification of infarction. Third, behavioral testing following the infarct procedure was used to assess the presence and severity of motor deficits (see below, Behavioral assessment). Because these monkeys were part of a long-term recovery experiment with chronic rehabilitative interventions (data not shown), histologic evaluation was not performed as it would have been affected by months of tissue necrosis and scavenging. The digital images of the infarct and the neurophysiological assessments of viable tissue were considered to be reliable indicators of infarct extent.
Behavioral assessment
Motor performance, and thus the assessment of motor deficits after ischemic infarct, was based on dwell time (in seconds), that is, the time to remove single food pellets from wells of an automated training/testing apparatus (Plautz et al., 2003, 2016). A performance index of dwell time was derived for each monkey to control for individual differences in preinfarct performance, that is, (postinfarct dwell time) ÷ (preinfarct dwell time). An index of 1 indicates average baseline performance; an index above 1 indicates an impairment in performance or an increase in dwell time. For example, an index of 2 would be double the time it would normally take to remove a food pellet from the Klüver board.
Analysis of ICMS-derived motor maps
The x–y coordinate positions of all ICMS sites were determined using National Institutes of Health (NIH) Image (version 1.63), and color-coded maps were generated by an in-house program via a space-filling algorithm (Nudo et al., 1992; Frost et al., 2003; Barbay et al., 2015; Plautz et al., 2016; Frost et al., 2022). The area of each colored region, representing a specific movement type (or category), was measured using NIH Image. Area data were generated from each cortical area for the four movement categories of total DF, digit and wrist/forearm; dual responses (combined digit, wrist, and proximal arm) are also examined separately. The designation of ICMS DF sites as being in M1, PMv, or PMd was based on several considerations including (1) topographic location relative to the central sulcus and arcuate dimple, (2) published descriptions of DF map locations in motor cortex, and (3) that in the postinfarct map, the known physiology and corresponding vascular patterns in the preinfarct map were used to distinguish peri-infarct M1 sites from PMd sites. ICMS current thresholds were also examined for differences between regions, representational categories, and between preinfarct and postinfarct maps (Fig. 1).
Statistical analysis
Motor performance on the reach and grasp task was obtained for 10 of the 14 monkeys (n = 5/group). Performance data were not available for two of the monkeys in the 2-Week Group and two of the monkeys in the 3-Week Group because of issues with the automated behavioral apparatus. A linear trend analysis was performed over the first 3 weeks of recovery to assess consistency of improvement. Box and whisker plots showed a wide distribution of areal measurements within some movement categories; thus, we used the more conservative nonparametric tests for statistical comparisons. Statistical testing of mapping data were performed with the Mann–Whitney U test for between-group comparisons and Wilcoxon matched-pairs signed-rank test for within-group comparisons. Preinfarct area data were compared between groups to determine whether maps in the two groups were similar before infarct. Size of the infarct produced was also compared between groups (total infarct area, total DF area affected by infarct). Within-group changes between preinfarct and postinfarct maps for individual movement categories were compared on independent pairs of data. All statistical tests were conducted using Prism (GraphPad) software, with two-tailed hypotheses and alpha at 0.05.
Results
Lesion verification
In the entire cohort of 14 monkeys, an average of 90.4% of mapping sites contained within the intended infarct border were found to be nonresponsive during the second ICMS procedure, with 7.8% of sites (∼5 sites per map) producing proximal responses and 1.8% of sites (∼1 per map) producing DF responses. Sites that were responsive were universally found at the edge of the intended infarct region.
Behavioral effect of M1 infarct
Before the M1 infarct there were no differences in dwell time (seconds) between the two groups for removing food pellets from the automated training/testing apparatus (2-Week Group, n = 5; median, 0.76 s; range, 0.53–0.95 s; 3-Week Group, n = 5; median, 0.80 s; range, 0.5–0.95 s; Mann–Whitney U = 10, p = 0.69). After the infarct, monkeys initially displayed flaccid paresis of the hand contralateral to the infarct, which resolved within the first week and before the first assessment. Postinfarct data are represented as an index score as follows: (postinfarct dwell time)/(preinfarct dwell time). Figure 3 shows natural (spontaneous) recovery over time for all monkeys (n = 10). A linear regression analysis showed a significant linear trend toward baseline performance over the first 3 weeks postinfarct, F(1,28) = 5.57, p = 0.026. A Wilcoxon paired comparison of performance between postinfarct week 1 (median, 3.10 s; range, 2.04−5.97 s) and postinfarct week 3 (median, 2.21 s; range, 1.29–3.79 s) showed a significant improvement from the first to third week after the infarct (W = −41, p = 0.04).
Preinfarct movement representations in M1, PMv, and PMd
General topography
The general topography of movement representations in M1, PMv, and PMd were similar to previous descriptions in primate motor cortex (Gould et al., 1986; Nudo et al., 1992; Plautz et al., 2000; Park et al., 2001; Frost et al., 2003; Dancause et al., 2005; 2006a; Card and Gharbawie, 2020). The DF region in M1 was surrounded by, and occasionally interleaved with, more proximal motor representations, primarily including elbow and shoulder regions, although trunk/torso sites (medially) and face sites (rostrolaterally) could be found in the proximity of the M1 DF representation (Fig. 1). Caudally, M1 was bounded by a nonresponsive region previously attributed to the border of somatosensory area 3a (Dancause et al., 2005). The DF region in PMv was similarly surrounded by proximal movements, with elbow and shoulder sites predominating on the medial, lateral, and rostral margins, and face sites on the caudal and medial margins. Nonresponsive sites were also common laterally and rostrally. The DF region in PMd was substantially smaller than in the M1 and PMv regions and in fact could be best described as being embedded within a larger proximal forelimb representation. Torso/trunk and/or face responses were sometimes found at the lateral border between the PMd and PMv forelimb representations. PMd was typically bounded rostrally by nonresponsive sites and medially by torso/trunk and hindlimb. The PMv and PMd DF representations were closely aligned along a similar medial/lateral line (or AP bregma coordinate). The face representation was generally a reliable indicator of the transition from M1 to PMv. In contrast, there was no clear physiological demarcation between M1 and PMd; instead, the border was characterized by a continuous region of proximal arm sites. Based on the maps derived in this study, the most caudal PMd DF sites were typically at least 1 mm rostral from the edge of the M1 DF region. These borders are consistent with those shown in a previous study (Dancause et al., 2008). This separation in baseline maps was used as a basis for designating DF sites as M1 or PMd in postinfarct maps; the border was assumed to evenly bisect the intervening proximal arm region in the baseline maps, and thus vascular landmarks could be used to redefine the border in the postinfarct maps (which often featured vascular remodeling related to the infarct cavitation).
Areal measurement
The baseline (preinfarct) areal size of the total DF representations (mean ± SD) for all 14 monkeys was 13.78 ± 2.65 mm2 (median, 12.75 mm2; range, 11.0–18.95 mm2) in M1, 3.34 ± 0.80 mm2 (median, 3.19 mm2; range, 2.24–5.15 mm2) in PMv, and 1.27 ± 0.93 mm2 (median, 1.18 mm2; range, 0.22–3.72 mm2) in PMd. The size of the M1 and PMv DF areas were consistent with previous ICMS mapping reports in squirrel monkeys (Nudo et al., 1992, 1996a; Plautz et al., 2000; Frost et al., 2003; Plautz et al., 2003, 2016). The composition of the DF representations was a mix of digit, wrist/forearm, and dual-response areas. For M1, this mix was 35%, 51%, and 14%, respectively. This is also consistent with previous reports (Nudo et al., 1992, 1996a; Plautz et al., 2000). For PMv, the ratios were 43%, 43%, and 14%, respectively. For PMd, the ratios were 29%, 33%, and 38%, respectively.
Stimulation thresholds
ICMS current thresholds (mean ± SD) for evoking DF movements for 14 baseline maps in 14 monkeys were 14.67 ± 2.90 µA (median, 13.65 µA; range, 10.7–22.8 µA) in M1, 18.72 ± 2.54 µA (median, 18.8 µA; range, 15.0–23.7 µA) in PMv, and 18.10 ± 3.87 µA (median, 19.00 µA; range, 11.7–24.0 µA) in PMd. Thresholds in M1 were consistent with previous reports (Nudo et al., 1992, 1996a; Plautz et al., 2000; Frost et al., 2003; Dancause et al., 2008) as were thresholds for PMv and PMd (Dancause et al., 2008).
There were no significant differences between groups at baseline for total DF area or for component representations (digit and wrist/forearm) in any of the three regions (M1, PMv, PMd) in the preinfarct map (nine comparisons, all p > 0.05; complete data in Table 2). Baseline ICMS thresholds (total DF) also were not significantly different between the 2-Week and 3-Week Groups for the three motor regions (M1, Mann–Whitney U = 16, p = 0.30; PMv, Mann Whitney U = 16, p = 0.32; PMd, Mann–Whitney U = 8.5, p = 0.94). In short, baseline maps from both groups were statistically equivalent.
Infarct dimensions and effect on M1 representations
As reported above in Materials and Methods, the infarct was intended to target the entire M1 DF representation, including some adjacent proximal arm areas, while also considering the associated vascular pattern with respect to how occlusion of a particular vessel might or might not have an impact on regions that were intended to be spared. For example, large arteries that passed through the DF map but also supplied substantial regions of the surrounding cortex were spared from complete occlusion; instead, only the branches that terminated within the DF map were occluded. The result of these several considerations was to produce infarcts that were larger in areal extent than the size of the DF map and yet occasionally were still unable to target the entire extent of the map.
The infarct size was equivalent for both groups (2-Week Group, 21.62 ± 1.98 mm2; median, 20.23 mm2; range, 15.23–28.36 mm2; 3-Week Group, 23.0 ± 1.26 mm2; median, 22.24 mm2, range 17.62–27.31 mm2; Mann–Whitney U = 21, p = 0.71). The size of the M1 DF representation was similarly equivalent (2-Week Group, 13.47 ± 0.90 mm2; median, 12.83 mm2; range, 11.0–16.75 mm2; 3-Week Group, 14.10 ± 1.15 mm2; median, 12.66 mm2; range, 11.41–18.95 mm2; Mann–Whitney U = 21, p = 0.71). Accounting for intentionally spared M1 DF, the actual DF infarct was equivalent between groups (2-Week Group, 13.22 ± 0.91 mm2; median, 12.58 mm2; range, 10.96–16.75 mm2; 3-Week Group, 13.78 ± 1.12 mm2, median, 12.66, range, 10.66−18.20 mm2; Mann–Whitney U = 21, p = 0.71). In sum, to the extent possible, the infarcts were equivalent in actual size and functional target.
Effect of postinfarct duration on movement representations in PMv and PMd
Changes in premotor DF representation
Neither PMv nor PMd were directly targeted by the infarct. However, both areas underwent dynamic changes in response to the injury. At 2 weeks postinfarct, the DF representation expanded by an average of 7.4% in PMv and by 110.6% in PMd. In contrast, at 3 weeks postinfarct, the DF representation was reduced by an average of 42.4% in PMv and by 57.4% in PMd. Further, the balance between different movement types was altered. Notably, for PMv, digit area decreased by an average of 34.1% at 2 weeks and by 73.2% at 3 weeks. However, both wrist/forearm and dual response areas increased by an average of 39.0% and 67.9%, respectively, at 2 weeks but decreased by an average of 26.3% and 11.7%, respectively, at 3 weeks. For PMd, the pattern was similar, with digit area decreasing by an average of 20.7% and 82.6% at 2 and 3 weeks, respectively, whereas wrist/forearm and dual-response areas increased by an average of 175.3% and 142.8%, respectively, at 2 weeks but decreased by an average of 59.6% and 10%, respectively, at 3 weeks.
PMv statistical analysis
Paired nonparametric comparisons were used to analyze differences in preinfarct and postinfarct map areas for each group independently. Results are presented in Table 3 and Figure 4. Statistically, these analyses show that in PMv there was no significant difference in the total DF, digit, or wrist/forearm movement representation area 2 weeks after the infarct, but there was a significant reduction in total DF area and digit area 3 weeks postinfarct.
PMd statistical analysis
Paired nonparametric comparisons were used to analyze differences in preinfarct and postinfarct map areas for each group independently. These results are presented in Table 4 and Figure 5. Statistically, these analyses show that in PMd there was a significant increase in total DF and wrist/forearm movement representation areas 2 weeks after an M1 infarct, but there was no significant change in digit movement area. Three weeks after an M1 infarct there was a reduction in total DF area and digit area, but there was no significant change in wrist/forearm movement area compared with preinfarct wrist/forearm area.
Changes in ICMS thresholds
ICMS thresholds for matched pairs of sites were compared between preinfarct and postinfarct maps for both groups. Map regions with fewer than three DF sites (in either map) were excluded from analysis. This rule excluded 11 M1 map pairs (because of loss of sites in postinfarct maps) and 5 PMd map pairs (1 preinfarct, 2 postinfarct, and 2 preinfarct and postinfarct). All PMv map pairs met the criteria. Only total DF was analyzed as there were too many map pairs excluded when the data were subdivided into digit and wrist/forearm response categories.
PMv, and PMd analysis
Wilcoxon matched pairs signed-rank test was used to analyze differences in preinfarct and postinfarct thresholds (µA) for each group independently. Results are presented in Table 5. Thresholds remained relatively stable in both the 2-Week and 3-Week Groups in PMv and PMd after the infarct despite a small but statistically insignificant decrease in PMv in postinfarct week 3 and PMd in postinfarct week 2.
Discussion
To date, this study presents the most convincing evidence that early improvements in motor function after stroke precede expansion and stabilization of spared cortical map topography. A nonlinear relationship between map expansion and recovery in primates was first reported in a study of supplementary motor area (SMA) in the medial premotor cortex of squirrel monkeys using ICMS procedures (Eisner-Janowicz et al., 2008). Digit and wrist/forearm representations in SMA were significantly reduced 3 weeks after an ischemic injury that included M1, PMv, and PMd. Although these monkeys were chronically impaired, the reduction in SMA occurred at a time when improvements in motor performance on a reach task had already plateaued. In the present study, confining injury to the DF representation of M1, functional improvement of reach and grasp was more pronounced during the early stages of recovery. Both studies showed a reduction in cortical map size 3 weeks after a cortical infarct. Delayed change in map size is consistent with findings from an fMRI study by Wang et al. (2010) assessing early cortical reorganization in people following a subcortical stroke. In their study Wang et al. (2010) reported that destabilization of the motor network was not immediate, being undetected until 10–14 d after stroke. These results put into question the assumption that expansions in spared motor maps represent a reliable biomarker of adaptive mechanisms related to recovery. Although spared premotor DF map expansion is commonly observed in the chronic postinjury stage when motor performance has plateaued, the present results and others demonstrate a temporal mismatch between motor map changes and motor recovery in the subacute stage, when natural behavioral improvements are most rapid. As early recovery of motor skills after M1 DF injury is accompanied by nonlinear changes in spared premotor DF representations, including significant reductions at 3 weeks postinjury, a reinterpretation of the functional significance of motor output maps in recovery seems to be demanded.
In the healthy brain, cortical motor maps remain relatively stable over time; plastic enough to accommodate acquisition of new motor skills while being resistant to transient environmental demands. The stability of motor map organization has been demonstrated by Plautz et al. (2000) with squirrel monkeys and by Kleim et al. (1998, 2004) with rats after repetitive performance on an unskilled task to receive a food reward. Repetitive training on behavioral tasks requiring motor skill acquisition results in enduring shifts in motor map organization. Newly developed skills can alter topography in a task-specific redistribution of movement representations within motor maps and sometimes result in the expansion of maps (e.g., areas defined as digit and wrist movements may expand into adjacent proximal arm areas). Nudo et al. (1996a) demonstrated this in squirrel monkeys, showing that movement representation maps in primary motor cortex could be alternated between wrist or digit representation depending on the tasks required (e.g., removing food pellets from a small well with digits or using wrist/forearm for rotating a bolt to receive a reward). Milliken et al. (2013) also showed an expansion of wrist/forearm related to a decrease in digit representation in motor cortex when use of digits was restricted for up to 35 weeks. Maps returned to normal topography of each monkey in 50–130 d once the limb restriction was removed. Kleim et al. (1998) demonstrated that repetition of a skilled reach and retrieval task expanded DF wrist and digit areas in the rat primary motor cortex, but repetition of a nonskilled bar press task did not have the same effect. These learning-dependent physiological changes in the forelimb motor map of the rat were shown to be associated with synaptogenesis (Kleim et al., 2002). In a subsequent study, Kleim et al. (2004) investigated when cortical changes developed during motor skill acquisition. Their results showed that map reorganization and synaptogenesis were not seen as the performance of the rat improved but occurred after skill acquisition. The authors concluded that cortical motor representations may be important for refined motor skill development but not during the early stages of learning the task. The initial process of developing motor skills seems to depend on rapid changes in synaptic efficacy as demonstrated by Rioult-Pedotti et al. (2000). These rapid changes can transiently alter activation of new task-related muscle synergies (Donoghue et al. (1992). With repeated practice these new muscle synergies can become consolidated and maintained within motor maps as mediated by local homeostatic processes that stabilize synaptic connectivity (Monfils et al., 2005).
Injury to the motor cortex can lead to a disruption of homeostatic processes leading to rapid alterations in synaptic efficacy and instability of muscle synergies and consequently motor maps during the acute/subacute stages of recovery. An initial breakdown of homeostatic processes is related to loss of neural connectivity (e.g., imbalance of GABAergic inhibition and glutamatergic excitability) but eventually rebalances during recovery (Kleim et al., 2004; Monfils et al., 2005; Dancause and Nudo, 2011; Grefkes and Fink, 2014). This process is sometimes referred to as synaptic scaling (Pozo and Goda, 2010; Stampanoni Bassi et al., 2019; Susman et al., 2019; Moulin et al., 2020; van Vugt et al., 2020) and seems to regulate synaptic strength relative to resting-state network activity (Davis, 2013; George et al., 2018; Moulin et al., 2020). In the present study, we assume deafferentation of premotor cortex after an infarct in M1 DF disrupts synaptic scaling, which leads to random shifts in ipsilateral or interhemispheric modulation of neural activity causing disinhibition and hyperexcitability in perilesional and remote cortical areas (Witte, 1988; Redecker et al., 2002). During the acute/subacute stages of map destabilization and hyperexcitability, various movement strategies may be attempted to compensate for injury-related motor impairment (Cirstea and Levin, 2000). Newly formed muscle synergies may gradually be consolidated while motor maps begin to stabilize as new motor skills develop (Monfils et al., 2005). Nonlinear changes observed in premotor maps early after an M1 DF injury as seen in the present study may be in part because of attempts at different behavioral strategies to compensate for motor impairment while homeostatic processes are unstable, creating a hyperplastic neural environment (e.g., increase in PMd wrist/forearm representation in the 2 week maps and reduction in the 3 week maps). Transient, compensatory movements may have more of an impact on map topography during this period, resulting in rapid changes in movement representations.
During the chronic period of recovery from an M1 DF injury, motor maps seem to become more stable as recovery of motor skills plateaus. In nonhuman primates, ICMS maps derived in PMv and PMd several weeks after an MI DF infarct show expansion of digit and wrist/forearm representations (Frost et al., 2003; Murata et al., 2015; Plautz et al., 2016; Frost et al., 2022). The extent of map reorganization in PMv observed during this late recovery period has been shown to be linearly related to the extent of injury to M1 (Frost et al., 2003; Dancause et al., 2006b). Because interregional connectivity between distal forearm areas in M1 and premotor cortex are functionally organized (Dancauseet al., 2006a; Dea et al., 2016), changes in premotor maps will depend on the subregions of the DF affected by injury. Ischemic injuries that encompass all subregions in the M1 DF area (a total M1 DF infarcts) have been shown to result in reorganization of connectivity patterns between cortical areas formally connected to M1 before the injury (Dancause et al., 2005). Novel anatomic projections between PMv and the caudal aspect of the primary somatosensory hand area (S1) were seen 3 months after a total M1 DF injury; this may be evidence of compensatory sensorimotor network reintegration because of loss of M1 connectivity. That is, novel injury-induced axonal sprouting may be contributing to restabilization of an injured motor network in the process of modifying functional contributions of premotor cortices for motor skill acquisition and performance such as reach and grasp.
Functional modifications of premotor cortices after an M1 infarct may be driven by attempts to regain volitional use of the hand and distal forearm. These compensatory behaviors may explain the differences in map changes seen in the current study between PMv and PMd within the first 2–3 weeks after the M1 DF infarct, a period of postinjury enhanced or reactive plasticity. Furthermore, the differential effects of various compensatory strategies on PMv and PMd may involve their unique functional characteristics (Hoshi and Tanji, 2007; Kantak et al., 2012; Takahashi et al., 2017). PMv functions from mostly an extrinsic frame of reference mediating reaching and grasping behavior in space relative to external sensory information and controls the DF almost exclusively through M1 (Shimazu et al., 2004; Hao et al., 2014). Early loss of connectivity with M1 may lead to an immediate inability for PMv to initiate use of digit and wrist/forearm for target location and grasping; initial disuse of digits and grasping may result in progressive loss of DF representation, especially digit representation (Milliken et al., 2013). In contrast, PMd is more functionally related to M1 as it is involved in executing behavior from an intrinsic frame of reference that is strongly related to the kinematics of reaching and grasping (Hoshi and Tanji, 2007). Also, PMd is not as reliant on M1 output for mediating purposeful movements as PMv because PMd sends corticospinal projections to upper and lower cervical segments of the spinal cord (He et al., 1993; Dum and Strick, 2005). This gives PMd more direct influence on upper limb control and wrist/forearm motor control via projections to propriospinal neurons that terminate at cervical level C7. More recent studies have shown that there is not a clear separation between PMv and PMd for reach and grasping behaviors as has been commonly held (Liu and Rouiller, 1999; Raos et al., 2003; Hao et al., 2014; Takahashi et al., 2017). Given that PMd can mediate reach and grasping behaviors independently of PMv after an M1 DF injury (Liu and Rouiller, 1999) and is involved in associative learning and action selection (Kantak et al., 2012), the early improvements in task performance seen in the current study over the first 3 weeks after the M1 DF infarct may have been mediated through PMd.
Considering the functional differences between PMd and PMv, the nonlinear PMd motor map expansion and later contraction not seen in PMv maps may be because of variability in PMd action selection after the M1 injury. Repetitive rehabilitative training could reduce variability in action selection and prevent maladaptive strategies that can interfere with early recovery resulting in chronic impairments. We believe that the synaptic hyperexcitability seen early after injury creates an optimal opportunity for implementing behavioral therapy and thus directing adaptive behavioral strategies (Allred and Jones, 2008; Jones, 2017). Early rehabilitative training has been shown to be effective in rodents (Kleim et al., 2003; Conner et al., 2005; Tennant et al., 2015; Okabe et al., 2016) monkeys (Nudo et al., 1996b; Friel et al., 2000; Plautz et al., 2003) and humans (Liepert et al., 1998; Liepert et al., 2001; Wittenberg et al., 2003). So implementing therapeutic interventions to control adaptive or successful compensatory movement strategies early after injury may avoid later development of maladaptive muscle synergies from consolidating within premotor maps.
Do the present results suggest that this period of early map instability is the optimal time for therapeutic rehabilitation? During the acute/subacute period, when maps are unstable and compensatory movement strategies are used, rehabilitative training could have its most potent effect on shaping the subsequent organization of motor maps in chronic stages. Also, as neuroanatomical sprouting of corticocortical fibers is occurring during these early stages, rehabilitative training may shape connectivity of corticocortical circuits. During chronic stages, neurophysiological and neuroanatomical plasticity has largely stabilized, and rehabilitative training may be less impactful on motor map organization and functional recovery. Although several studies have suggested that early rehabilitation is more effective, the molecular and cellular events that contribute to stabilization of motor maps and better functional recovery remain unclear (Paolucci et al., 2000; Salter et al., 2006; Wieloch and Nikolich, 2006).
Footnotes
This work was supported by National Institues of Health Grants U54 NS048126, NIH R01 NS30853, and P30 HD002528 and Northstar Neuroscience. We thank Bob Cross, Caleb Dunham, Erica Hoover, Diane Larson, and Phuong Nguyen for technical assistance and Scott Bury, Pei-chun Fang, David McNeal, Michael Taylor, and Ed Urban for assistance with collection of mapping data.
The authors declare no competing financial interests.
- Correspondence should be sent to Randolph J. Nudo at rnudo{at}kumc.edu or Erik Plautz at erik.plautz{at}utsouthwestern.edu