Behavioral recovery after a spinal deafferentation injury in monkeys does not correlate with extent of corticospinal sprouting

Abstract

A long held view in the spinal cord injury field is that corticospinal terminal sprouting is needed for new connections to form, that then mediate behavioral recovery. This makes sense, but tells us little about the relationship between corticospinal sprouting extent and recovery potential. The inference has been that more extensive axonal sprouting predicts greater recovery, though there is little evidence to support this. Here we addressed this by comparing behavioral data from monkeys that had received one of two established deafferentation spinal injury models in monkeys (Darian-Smith et al., 2014, Fisher et al., 2019, 2020). Both injuries cut similar afferent pools supplying the thumb, index and middle fingers of one hand but each resulted in a very different corticospinal tract (CST) sprouting response. Following a cervical dorsal root lesion, the somatosensory CST retracted significantly, while the motor CST stayed largely intact. In contrast, when a dorsal column lesion was combined with the DRL, somatosensory and motor CSTs sprouted dramatically within the cervical cord. How these two responses relate to the behavioral outcome was not clear. Here we analyzed the behavioral outcome for the two lesions, and provide a clear example that sprouting extent does not track with behavioral recovery.

1.1. Introduction

Recovery of function following spinal cord injury is notoriously difficult to predict, and experimental models have long used axonal sprouting, and most commonly corticospinal tract (CST) outgrowth, as an anatomical biomarker of functional recovery [1–7]. Here we compared behavioral recovery across two spinal injury models in the macaque monkey, that cut primary afferent inputs exclusively from the thumb, index and middle finger of one hand [8–11], to determine the validity of this relationship. Lesions involved either the dorsal rootlets alone (DRL), or in combination with a lesion in the dorsal column (i.e. cuneate fasciculus, DRL/DCL). Both lesions cut a similar (though not identical) afferent population of fibers, and produced a sensory deficit in the opposing digits (i.e. the thumb, index, and middle fingers) of the affected hand. Despite the similarities of the two lesions, each is known to induce a dramatically different sprouting response in the spinal cord from the primary somatosensory (S1) and primary motor (M1) CSTs during the first 6 post-injury months. These data have been published [2, 8–12] (see summary data in Figures 1D–E) and set the stage for the current investigation. When the dorsal roots alone are transected, the S1 CST on the lesioned side retracts to 60% of its original terminal territory within the dorsal horn of the cervical cord (Figure 1D, green vs orange distributions), and the M1 CST remains robust and largely unchanged from its normal range (Figure 1E, green vs orange). In contrast, following a DRL/DCL, both the S1 and M1 CST terminals sprout dramatically (and bilaterally) beyond their normal range within the cervical and thoracic cord (Figure 1D–E, compare blue with green territories) [2, 8, 9, 11, 12]. The injury models used were easily replicated, and ideal for testing the relationship between terminal sprouting and functional recovery extent [1, 13]. Injuries that transect the CSTs directly (e.g. hemisections, or contusions) are clinically relevant, but also more difficult to precisely reproduce, which makes it challenging to interpret the impact of terminal sprouting on behavioral recovery. Given our findings, and the different CST responses following the two lesions, we hypothesized that there is no clear link between the CST terminal sprouting extent and the post-injury behavioral recovery, at least for the sensory lesions examined.

Figure 1.

A–C Summary of the experimental design of the study, with the behavioral manipulandum/apparatus used to assess reach-grasp-retrieval (A), the location of, and pathways impacted by the two lesion models (B), and the placement of tracers within the sensorimotor cortical regions during a craniotomy (C), which labeled corticospinal projections to the cervical and thoracic cord. These anatomical findings have been published [2, 8, 9, 11, 12] and they provide the context for this report and behavioral analysis. (D–E) Summarizes these published data for (D) S1 CST and (E) M1 CST terminal territories. Data are shown for injections placed into the region representing the thumb, index, and middle fingers (D1–D3) of S1 and M1 bilaterally, 4–6 months after a unilateral primary afferent lesion was made (indicated vertically through D and E). The S1 CST is strictly a contralateral projection, even following a spinal lesion. M1 CSTs are known to have a small ipsilateral projection (~2%), and crossover component, but these did not alter the terminal domains. Following a DRL alone, there was a 40% retraction of S1 CST input to the ipsilateral cord, and the CST projection to the contralateral cord was not affected. Thus, green shows both the ‘normal’ spread of input to contralateral cord, as well as the terminal territory within the contralateral cord following a DRL. Blue shows the large expansion of the terminal territory on both sides following a DRL/DCL. Note that there was extensive bilateral CST sprouting in DRL/DCL animals that was not seen following a DRL alone. This was especially dramatic for the S1 CST (top panel). (F) provides specific details for the 7 monkeys used.

1.2. Materials and Methods

1.2.1. Subjects

Subjects were 7 young adult (3–4 years old with an average weight of 3.9 ± 0.5kg) colony bred (Charles River) male monkeys (Macaca cynomolgus). Monkeys were housed individually at the Stanford Research Animal Facility, in four unit cages (64×60×77cm, depth × width × height per each unit) in a room with other monkeys. They were kept to a 12 hour light/dark cycle and the room was maintained at 72–74°F. ARRIVE guidelines were followed, with the exception that only male monkeys were used. This was due to availability and a lack of gender differences with respect to hand function, sensorimotor pathways and recovery following SCI.

Animal procedures were carried out in accordance with National Institutes of Health guidelines and the Stanford University Institutional Animal Care and Use Committee.

Only one monkey used in this paper (M1106) has had anatomical data published (Darian-Smith et al., 2014), and none of the behavioral data in this study has been reported elsewhere. Anatomical data from M1804 and M1805 forms part of an ongoing investigation (Fisher et al., unpublished).

1.2.2. Behavioral assessment

All monkeys were matched for species, age, and sex, and all had identical DRLs or combined DRL/DCLs that only affected the thumb, index and middle fingers (D1–D3) in one hand (see Figure 1). All lesions were unilateral, and made on the side of the dominant hand. Hand preference was determined over 2–3 weeks prior to the laminectomy. Monkeys were handed fruit or other small food items, and the preferred hand for reaching and grasping the food was scored over dozens of reaches during this time period. All monkeys had identical behavioral training routines. As such, the monkeys in each lesion group were otherwise indistinguishable.

The reach-grasp-retrieval task involved a natural movement, so that monkeys learned quickly and performed at a consistent speed within 4–6 weeks. Monkeys were trained to sit in a plexiglass box and reach through a window (located on either the right or left depending on the hand being tested), to retrieve a candy pellet held in one of 4 clamps (see Figures 1, 3 and 4). Each clamp held an identical candy pellet at one of 4 different forces (0.5, 1.0, 1.5 or 2.0 Newtons), and clamps (which were visually indistinguishable to the monkey), were presented pseudo-randomly. Training sessions were filmed using a high shutter speed (30 fps at 1/500s) digital camcorder (Canon XF200) and data analyzed offline using Edius Pro 9 (Grass Valley) software. ‘Contact time’ was the time interval between first contact with the target pellet and its successful displacement from the clamp.

Figure 3.

Representative manual retrieval stratagems used during behavioral assessment. Three different frame sequences from monkey M902 (DRL), showing the predominant stratagem used prelesion (column 1) and postlesion (columns 2 and 3). Contact times for each sequence are given at the bottom of each column. The times for each frame are indicated, with red stars illustrating the initial digit contact with the object and the point at which contact times begin. Prior to the lesion (column 1), performance was smooth and efficient. The target was contacted simultaneously with the opposing distal pads of the thumb and index finger, and the object was retrieved from the clamp. In column 2 (2 weeks postlesion), performance was visibly impaired, and retrieval took much longer than it had prior to the lesion. Retrieval stratagem was also different, as initial contact with the object no longer involved simultaneous opposition of the thumb and index fingers. Instead, only the distal pad of the index finger made initial contact, which then scooped the object until it contacted the distal pad of the thumb. Several attempts were needed to dislodge the object from the clamp, after which the object was supported by the dorsum of the distal segment of the thumb and the palmar pads of the curled index finger. Placement of both digits were abnormal. There was also a profound extension of the wrist, which was not seen prelesion. In column 3 (10 weeks postlesion for M902), a similar scooping motion was used, though the overall proficiency of the performance had increased and the contact times decreased. Although a compensatory stratagem was adopted to retrieve the object, this new strategy produced the same efficiency seen prior to the lesion. Alternate stratagems used post-lesion are shown to the right.

Figure 4.

Shows manual retrieval stratagems used during behavioral assessment in a representative monkey that received a DRL/DCL (M1803). No behavioral differences were observed between monkeys receiving a DRL (Figure 3) and those receiving a DRL/DCL. Conventions as in Figure 3. Performance was smooth and efficient pre-lesion, and the target was contacted simultaneously with the opposing distal pads of the thumb and index finger. One week postlesion, performance was clearly impaired, contact times were much longer, and the retrieval stratagem had changed so that only the distal pad of the index finger made initial contact. Several attempts were needed to dislodge the object from the clamp, after which the object was supported by the dorsum of the distal segment of the thumb and the palmar pads of the curled index finger. Placement of opposing digits was abnormal, and there was a profound extension of the wrist. A similar scooping motion was used at 8 weeks, though overall performance was more efficient and contact times had decreased. Alternate stratagems used post-lesion are shown (right) for different monkeys.

Data were obtained from a minimum of 20 trials per week, pooled from 3–5 training sessions of 20 minutes duration, and conducted at the same time each day. Monkeys had chow and water available ad libitum, but fruit was provided after training and candy pellets were only available during training.

1.2.3. Surgical Procedures

All monkeys underwent two aseptic surgical procedures, including a laminectomy to make the lesion, and a craniotomy, during which anterograde tracers were injected into reorganized primary sensory and motor cortex. The latter labeled CST terminals within the spinal cord, and the anatomical data have been described and published in detail elsewhere [8, 9, 11, 12]. The laminectomy was made after the initial behavioral assessment, and the craniotomy was made after the behavioral assessment was completed, and 6~7 weeks before terminating the experiment.

Surgeries

All surgical procedures involved initial sedation with ketamine hydrochloride (10mg/kg), and animals were maintained under gaseous anesthesia (isofluorane, 1–2% / O₂). Atropine sulfate (0.05mg/kg), buprenorphine (0.015mg/kg) and the antibiotic cefazolin (20mg/kg) were also given, and Normasol-R was infused intravenously throughout surgery to maintain fluid balance. Dexamethasone (0.25mg/kg) was also given before craniotomy procedures to minimize brain edema.

Physiological signs (i.e. blood pressure, heart rate, pulse oximetry, capnography, and core temperature) were continuously monitored throughout surgery to ensure a proper depth of anesthesia. A post-operative analgesic (buprenorphine, 0.015mg/kg) was given after surgery and monkeys were typically awake and alert within one hour. Oral meloxicam (0.1mg/kg) was administered for 3–5 days post-surgery, and Buprenorphine (0.05mg/kg, oral route) was also given as indicated in the days following surgery.

1.2.3.1. Making the lesions

The laminectomy involved exposing the C5-T1 region of the cord, cutting the dura mater along the midline and reflecting it to expose dorsal rootlets for these segments. Unitary recordings were made from dorsal root fascicles to create a microdermatome map of cutaneous receptive fields on the hand, and rootlets with axons supplying the thumb, index and middle fingers were then cut to create the DRL. In monkeys receiving a DRL/DCL, an additional cuneate fasciculus lesion was made level with the most rostral cut dorsal rootlet, using a micro blade, as described in earlier publications [8, 9, 11, 12].

1.2.3.2. Making tracer injections into cortex

In order to place tracer injections into the region of reorganization, unitary recordings were made to map hand representation in the primary somatosensory cortex (Areas 3b/1), as described in earlier work [8, 9, 11, 12]. Once the mediolateral boundaries of the thumb, index and middle fingers (D1–D3) had been determined electrophysiologically, anterograde tracer injections (i.e. either biotin dextran amine or Lucifer yellow dextran; 6–8 injections, 0.3μl, each) were made in a strip in the Area 3b/1 D1–D3 region. Primary motor cortex (D1–D3), was also injected, using a different tracer (again either BDA or LYD). Injections were made either bilaterally or unilaterally (contralateral to the side of the lesion) in monkeys used in this and earlier studies. We know from our published [8, 9, 11, 12] and unpublished work (Fisher and Darian-Smith, unpublished), that the S1 CST is always a contralateral projection, while the M1 CST has a known small ipsilateral and callosal component. In normal monkeys [14], and following a DRL (Fisher and Darian-Smith, unpublished), this comprises ~2 % of the total ‘hand’ projection from the motor cortex. Following a DRL/DCL, the ipsilateral component (contralateral to the side of the lesion) sprouts to expand its terminal territory, but it never extends for more than ~20% of its normal range, and never beyond the terminal domain observed ipsilateral to the lesion.

1.3. Results

All monkeys in the current study underwent the same experimental sequence shown in Figure 1, and published elsewhere [8, 9, 11, 12]. Briefly, all were trained and assessed in a reach-grasp-retrieval task requiring sensory feedback, and once performance had stabilized, a laminectomy was performed to make the lesion. Electrophysiological recordings were made in dorsal rootlets in C5-T1 to identify and target only the rootlets with detectable cutaneous RFs on the thumb, index and middle fingers (D1–D3). This is necessary for lesion replication, due to substantial inter-animal variability of inputs to the cervical segments [15]. In monkeys receiving a DRL/DCL, the DCL was always placed at the rostral border of the DRL, and only the cuneate fasciculus was cut, as described elsewhere [8, 9, 11]. Dorsal column lesions were later identified in cervical spinal cord tissue sections and their extent is shown in Figure 2. Once lesioned, behavioral data were collected (3–5 days per week) for a minimum of three additional months. A craniotomy was then made contralateral to the lesion, to record and locate the reorganized D1–D3 region of the primary somatosensory cortex, so that tracers could be injected to label corticospinal projections to the cord [8, 9, 11]. Monkeys were euthanized 6–7 weeks after the craniotomy.

Figure 2.

Photomicrographs of the dorsal column lesions in each of the four monkeys that received a combined DRL/DCL. In all cases, the dorsal column lesion involved the cuneate fasciculus component and was located within segment C5 adjacent to the most rostral of the cut dorsal rootlets. Scale bars = 1mm.

Behavioral data were obtained for both the lesioned and non-lesioned hands throughout the assessment period (Figures 3–4). This showed inter-hand, and inter-animal differences in performance times, as well as stable and consistent performance metrics in the non-lesioned hand in all monkeys throughout the study.

The reach-grasp-retrieval task has been described elsewhere in detail [16]. Briefly, monkeys reached to retrieve a pellet from one of four clamps (on a rotating turret) at a set distance in front of them. Each clamp held an identical treat at a different resistive force (0.5, 1.0, 1.5 and 2.0 Newtons), which made trials dependent on sensory feedback from the opposing digits, for effective retrieval. Monkeys could not see the clamp after the initial reach.

Baseline performance was taken as the data obtained during the last two weeks of pre-lesion training, since monkeys typically followed a learning curve during the early training period before their performance stabilized. Two parameters were used to assess the deficit and recovery of function in the hand: (1) contact time (CT), or the time from initial contact with the pellet to its removal from the clamp, and (2) the digit stratagem used to grasp and successfully retrieve the pellet. Representative examples of the behavioral consequences of each lesion are shown in Figures 3 and 4. All monkeys showed an initial deficit, and a subsequent recovery of digit function during the early post-lesion weeks. Monkeys performed the retrieval task pre-lesion using a precision grip involving the opposition of the distal pads of digits 1 and 2. After either lesion, however, performance was clumsy. Monkeys could pre-shape the affected hand, and their reach trajectory was not affected, but they were unable to accurately locate the target using the opposing distal digit pads once they reached the clamp, or to apply the correct force needed to effectively retrieve the target. This meant that contact times were initially significantly longer than baseline (Figure 5). Over 5–8 weeks, however, all animals developed novel and effective grip strategies (Figure 5), which allowed them to perform the task at pre-lesion speed. This was functional compensation, rather than a complete restoration of the original strategy. All DRL and DRL/DCL animals (targeting D1–D3), without exception, have followed a similar deficit and post-injury recovery of hand/digit function.

Figure 5.

(A–B) Contact time data showing the raw data for lesioned and non-lesioned hands in monkeys from each lesion group, and (C) a summary of data from the lesioned hand in all monkeys. Recovery of function is shown by a decrease in contact time following the lesion. The graph shows the line of best fit for the two lesion types, figured for the average force (Newtons). The shaded area is the +/− SE of each line. Importantly, the lines would have to be separated by approximately 3 standard errors to be significantly different. The decrease in contact time is highly significant, but the difference between the two curves is not significant. The analysis was performed in log-log space to model these curves plotted in linear space on the figure.

1.3.1. Statistical Analysis

For the statistical analysis, we used data from all 7 monkeys and compared CTs for the two lesion types (DRL versus DRL/DCL), over pre-lesion and post-lesion weeks, for the different clamp forces (0.5, 1.0, 1.5, and 2.0 Newtons). Only data from the side of the lesion was assessed statistically, as contact times, as well as stratagems were unaffected in the contralateral hand throughout the assessment period (Figure 5).

Contact time was log-normal distributed (as is typical for latencies), so CTs were logged and averaged for analysis. To test whether lesion type (DRL vs DRL/DCL) affected recovery post lesion, we performed a repeated measures REML mixed model in JMP 14 Pro. Monkeys were nested within lesion type, and treated as a random effect. Force and week were treated as continuous variables. Week was logged to model curves of decreasing return. Interactions between lesion type, and force and lesion type and week were tested to determine whether the effect of force or week differed between the lesions. Appropriate error terms for repeated measures mixed models were included.

We found that the change in contact times (i.e. slope) did not differ significantly between the two lesion types (F_1,5.00=1.322; P=0.3022; Figure 5). CT increased significantly with force (F_1,157 = 33.05; P<0.0001), but this effect did not differ between lesion types (F_1,157=0.0557; P=0.8138). CTs also decreased significantly with week post lesion (F_1,157.1 = 182.7; P<0.0001), and this effect did not differ between lesion types (F_1,157.1=0.2327; P=0.6302; Figure 5).

1.4. Discussion

Given that the two lesions differed so dramatically with respect to S1 and M1 corticospinal tract responses, but not with respect to the monkey’s behavioral deficit or recovery of hand/digit function, our findings support our hypothesis, that there is no clear link between the CST terminal sprouting extent and the post-injury behavioral recovery. We provide the first clear demonstration that more sprouting does not predict a better behavioral outcome, which means that CST sprouting is not in itself a good biomarker of behavioral recovery.

There are a number of considerations worth noting in the present study. First, the lesion models used compared a dorsal root (peripheral to the spinal cord) with a combined peripheral/central injury. These were purely sensory lesions and are not typical of clinical spinal injuries, where there is usually a mix of sensorimotor and central and peripheral elements. However, we do not see this as a confound. The approach taken provided a unique opportunity to demonstrate the lack of a relationship between terminal sprouting and behavioral recovery, and this would have been extremely difficult to demonstrate or assess in a more typical clinical model. Importantly, the principle is likely to apply well beyond the spinal cord, to sensorimotor injuries occurring at all levels of the neuraxis.

Another consideration is that the descending CST fibers were not themselves directly cut by either of the deafferentation lesions examined. When the CST itself is transected, as can happen following a contusion or hemisection injury [3, 17, 18], the cut axons do not regrow spontaneously, though spared CST axon terminals (including presumptive collateral branches), can sprout above and below the injury. As a result, a considerable body of research has focused on inducing CST terminal growth (e.g., with the introduction of stem cells, neurotrophic factors, CSPGs, various permissive cytokines, etc.), with some impressive and important results [3, 4, 19, 20]. However, these studies do not directly demonstrate that additional sprouting (beyond a certain level) is necessary for enabling a correspondingly greater functional recovery. The findings of the present study suggest that our current understanding of the relationship is incomplete at best, and may even differ according to injury type. Either way, it requires closer examination and clarity.

Key questions remain. Do the newly formed axons form functional synapses? This is difficult to demonstrate directly, especially following a spinal injury, but electrophysiological and neuroanatomical studies are starting to address this, and several groups have shown that at least within the subchronic timeframe, new synapses do form within the reorganized circuitry [4, 12, 21–23]. Does the extent of the CST terminal sprouting change over time? Little is currently known about what happens to terminal sprouts over an extended time period, so it will be important to determine if an initial exuberance in the early post-injury months, gives rise to a pruning of these ‘sprouts’ as the pathways stabilize and enter a chronic state. This may mimic events normally associated with developmental maturation, though little detail is yet understood about these dynamics. Finally, the CST projection to the spinal cord is likely to have an important role in recovery, as it provides cortical modulation of the spared reorganizing primary afferents (S1 CST), in addition to the intermediate and ventral horn output circuitry, as both adapt over time. However, this is clearly only part of the affected pathway. A number of studies show that axonal sprouting is widespread [2, 24–26], and that neuronal circuitry is affected distant to the lesion site, though not a lot is yet known about level specific changes.

1.5. Conclusions

Importantly, our findings do not suggest that axon terminal growth is not important (or even critical) in the formation of new connections and behavioral recovery. Clearly some outgrowth is needed for the formation of new synapses, and the establishment of effective pathways. However, this may require only very localized sprouting from a greatly reduced number of axons. It is already known that significant recovery is possible when the injury leaves even a tiny proportion of pathways intact. The present study is a case in point, where the DRL and DRL/DCLs permanently removed >95% of the original primary afferent input to the cord from the affected digits [2]. Other examples have also been reported that support this redundancy and plasticity [27–29]. The relationship between CST terminal sprouting extent and behavior is obviously complex and nuanced, involving all levels of the neuraxis [24], the mix of peripheral to central pathways, and factors such as the post-lesion time point at which responses are assessed. The inflammatory response [30], and molecular and cellular changes [13] are also clearly important in determining outgrowth post-injury [11, 20, 31, 32]. However, excess axonal growth cannot be used to infer a better behavioral outcome, and future studies are needed to establish more reliable predictors of behavioral recovery, both in the experimental and clinical settings.

Abbreviations

CST

corticospinal tract

CT

contact time

DRL

dorsal root lesion

DCL

dorsal column lesion

M1

primary motor cortex (Brodmann area 4 or new motor cortex)

S1

primary somatosensory cortex (Brodmann areas 3b/1)

D1-D5

digits 1 to 5 (D1= thumb, D2= index, etc)