Motor Learning Changes the Axon Initial Segment of the Spinal Motoneuron

Abstract

We are studying the mechanisms of H-reflex operant conditioning, a simple form of learning. Modeling studies in the literature and our previous data suggested that changes in the axon initial segment (AIS) might contribute. To explore this, we used blinded quantitative histological and immunohistochemical methods to study in adult rats the impact of H-reflex conditioning on the AIS of the spinal motoneuron that produces the reflex.

Successful, but not unsuccessful H-reflex up-conditioning was associated with greater AIS length and distance from soma; greater length correlated with greater H-reflex increase. Modeling studies in the literature suggest that these increases may increase motoneuron excitability, supporting the hypothesis that they may contribute to H-reflex increase. Up-conditioning did not affect AIS ankyrin G (AnkG) immunoreactivity (IR), p-p38 protein kinase IR, or GABAergic terminals.

Successful, but not unsuccessful H-reflex down-conditioning was associated with more GABAergic terminals on the AIS, weaker AnkG-IR, and stronger p-p38-IR. More GABAergic terminals and weaker AnkG-IR correlated with greater H-reflex decrease. These changes might potentially contribute to the positive shift in motoneuron firing threshold underlying H-reflex decrease; they are consistent with modeling suggesting that sodium channel change may be responsible. H-reflex down-conditioning did not affect AIS dimensions.

This evidence that AIS plasticity is associated with and might contribute to H-reflex conditioning adds to evidence that motor learning involves both spinal and brain plasticity, and both neuronal and synaptic plasticity. AIS properties of spinal motoneurons are likely to reflect the combined influence of all the motor skills that share these motoneurons.

Graphical Abstract

Summary of results and possible mechanisms of H-reflex up-conditioning and down-conditioning.

The cartoon shows the motoneuron soma, axon, and axon initial segment (AIS), GABAergic terminal input on soma and AIS, and the final H-reflex (HR) from Naive Control (NC) rats (black), Successful Up-Conditioned (US) rats (red, HR up-arrow), and Successful Down-Conditioned (DS) rats (blue, HR down-arrow). The relative H-reflex heights indicate the actual magnitude of the H-reflex size change produced by up- or down-conditioning. AIS length and location in US and DS rats relative to NC rats are indicated qualitatively. The concentrations of somatic p-p38, and GABAergic input (indicated by immunoreactivity (IR)), the concentrations of AIS AnkG, p-p38, and GAD-67, and the number of GABAergic terminals, all relative to NC rats, are indicated qualitatively by font size. In US rats, the AIS is longer and is located farther from the soma, while the AIS concentration of AnkG, the somatic and AIS concentrations of p-p38, and the number of AIS GABAergic terminals do not differ from those of NC rats. Computational modeling available in the literature suggests that the increases in AIS length and location in US rats are likely to increase motoneuron excitability; this is consistent with the hypothesis that they may contribute to the increase in H-reflex size (see text). In DS rats, AIS length and location do not differ from those of NC rats. At the same time, in DS rats, AIS AnkG immunoreactivity (IR) is weaker, AIS p-p38-IR and AIS and somatic GAD-67-IR are stronger, and AIS GABAergic terminal number is greater. These results extend previous data. Together with computational modeling, the results support the hypothesis that these AIS changes may possibly contribute to the operantly conditioned H-reflex decrease. Furthermore, if the AnkG decrease and p-p38 increase are also present at the axonal nodes of Ranvier, they could conceivably contribute to the decreased axonal conduction velocity that is associated with successful down-conditioning in both rats and monkeys (indicated in the figure by the longer H-reflex latency) (Carp, Wolpaw 1994; Halter et al. 1995; Carp et al. 2001). See text for further explication and discussion of these findings and their potential implications concerning the etiology of the H-reflex changes produced by operant conditioning.

INTRODUCTION

We are using operant conditioning of the H-reflex to explore the mechanisms of motor learning. The H-reflex is the electrical analog of the simplest vertebrate behavior, the spinal stretch reflex (e.g., the knee-jerk reflex). Like the spinal stretch reflex, the H-reflex is mediated primarily by a two-neuron, monosynaptic pathway comprising the primary afferent, its synapse on the spinal motoneuron, and the motoneuron itself (Magladery et al., 1951; Matthews, 1972; Baldissera et al., 1981; Henneman, Mendell 1981; Brown 1984; Pierrot-Deseilligny, Burke 2012). Because this spinal pathway is influenced by descending input from the brain, the H-reflex can be operantly conditioned. In response to an operant conditioning protocol, monkeys, humans, rats, and mice can gradually increase (up-condition) or decrease (down-condition) the H-reflex (Wolpaw 1987; Chen, Wolpaw 1995; Carp et al. 2006; Thompson et al. 2009; Pierrot-Deseilligny, Burke 2012). The larger or smaller H-reflex that results is a simple motor skill (i.e., an adaptive behavior acquired through practice (Compact OED 1993; Shmuelof, Krakauer 2011)). This simple skill is associated with plasticity at multiple sites in the brain and the spinal cord (reviewed in Wolpaw, Chen 2009; Thompson, Wolpaw 2014; Wolpaw 2018; Wolpaw, Kamesar 2022). Figure 1a shows the H-reflex operant conditioning protocol in rats and its impact on H-reflex size. The human, monkey, and mouse protocols are similar and produce similar results.

Figure 1.

a: The rat H-reflex operant conditioning protocol and its results. The mouse, monkey, and human protocols are similar and produce similar results. Left: Soleus EMG is monitored 24 h/day in a rat with implanted EMG electrodes and a tibial nerve cuff. The wires pass subcutaneously to a head-mounted connector and through a flexible cable and a commutator to EMG amplifier and stimulator. The rat can move freely about the cage. The diagram shows the main pathway of the H-reflex: the Ia afferent fiber and the spinal motoneuron. (Other large afferents and oligosynaptic paths may also contribute to the H-reflex.) Whenever the absolute (i.e., rectified) value of soleus EMG activity stays in a defined range for a randomly varying 2.3–2.7 s period, a nerve cuff stimulus elicits a threshold M wave (a direct muscle response) and an H-reflex. The trace illustrates one trial. A rat averages 2000–6000 trials/day. Center: The rat is first exposed to the control (baseline) mode, in which no reward occurs and the H-reflex is simply measured to define its initial size. For Days 0–50, it is exposed to the HRup or HRdown mode, in which a food-pellet reward occurs when the H-reflex is above (HRup) or below (HRdown) a criterion. Background EMG activity and M wave are kept stable throughout. Successful conditioning, defined as a change of at least 20% in the correct direction, occurs in most (~80%) rats; the others almost invariably remain within 20% of their initial value. The graphs show average (±SE) daily H-reflex sizes for 59 successful HRup rats (red) and 81 successful HRdown rats (blue). In both, mode-appropriate H-reflex change develops steadily over the 50 days. (If the mode switches from HRup to HRdown mode (or vice versa), the change reverses in the same gradual fashion.) Right: Average post-stimulus EMG activity (absolute value) for a day from an HRup rat (top) and an HRdown rat (bottom) in control mode (continuous line) and after conditioning (dashed line). The H-reflex is larger after up-conditioning and smaller after down-conditioning. Background EMG activity (shown here by EMG at time zero) and M waves are not changed. From Wolpaw (1997, 2010).

b: Experimental Schedule. The initial H-reflex size is defined as the average H-reflex size for the last 10 days of the control period (first heavy line on the x-axis). The final H-reflex size is defined as the average H-reflex size for the last 10 days of the conditioning period (Days 41–50, second heavy line).

c: Labeling of the soleus motoneuron and its axon initial segment (AIS). CTB-Alexa Fluor 647-labeled soleus motoneuron in rat spinal cord section (scale bar 20 μm). The arrowhead indicates the AIS, which shows ankyrin-G immunoreactivity (AnkG-IR, green). The inset shows the motoneuron location (arrow) in right ventral horn (mirror image) (scale bar 200 μm).

d-e: Determination of AIS location (distance from soma), length, and area with the Fiji-Image J program. d: A line is drawn from the soma (the zero point) to the beginning of the AIS (as defined in f below). The length of this line segment (d1) defines AIS location (AIS distance from the soma). The line then continues along the midline of the AIS to its end. The length of this line segment (d2) defines AIS length. (Scale bar 20 μm.)

e: The AIS starting and ending points (left and right arrows, respectively) are defined from the smoothed plot of AnkG-IR versus the path from the soma to the beginning of the AIS and then through the length of the AIS and beyond. (See text for full measurement details.)

f: Effect of H-reflex up-conditioning (HRup) or down-conditioning (HRdown) on H-reflex size. Average H-reflex size for each rat in each of the four experimental rat groups (Up-Successful (US; n=11), Up-Failed (UF; n=4), Down-Successful (DS; n=13), Down-Failed (DF; n=7) for the final 10 days of conditioning in percent of the rat’s initial H-reflex size (average size for the last 10 control days). For each rat, success was defined as a change of at least 20% in the correct direction (Wolpaw et al., 1993; Chen & Wolpaw, 1995). The bars show mean ± SD. For a rat group, a significant change in the correct direction from initial H-reflex size was indicated by a p-value <0.0125 (one-tailed paired t-test with Bonferroni correction). Exact p values are shown for the two rat groups (US and DS) that satisfied this criterion.

H-reflex down-conditioning is largely attributable to a positive shift in the firing threshold of the soleus motoneurons, probably due to a change in sodium channels (Carp et al. 1994; Halter et al. 1995; Carp et al. 2001). Successful down-conditioning is accompanied by an increase in GABAergic inhibitory terminals on the soleus motoneuron (Feng-Chen, Wolpaw 1996; Wang et al 2006; 2009). These findings prompted the present study of the impact of H-reflex conditioning on the motoneuron axon initial segment (AIS), the site of action potential generation. Guided by the findings and by recent AIS studies and related computational modeling available in the literature (Bréchet et al 2008; Grubb, Burrone 2010; Kuba 2012; Barry et al 2014; Evans et al 2015; Wefelmeyer et al 2015; Gulledge & Bravo 2016; Hines et al 2018; Nathanson et al 2019; Goethals & Brette 2020; Rotterman et al. 2021), we focused on: AIS length and location (i.e., distance from the soma); AIS ankyrin G (AnkG) immunoreactivity (IR) and phosphorylated p38 mitogen-activated protein kinase (p-p38) IR; and GABAergic terminals on the AIS. AnkG is important for Na channel function (Barry et al 2014; Bréchet et al 2008; Hedstrom et al 20008; Xu & Shrager 2005); p-p38 promotes internalization of sodium channel Na_v1.6 (Wittmack et al 2005, Gasser et al 2010); and axonal GABAergic terminal innervation affects neuronal excitability (Wefelmeyer et al 2015). We studied five rat groups: unconditioned naive control (NC) rats; H-reflex up-conditioned (HRup) rats in which the H-reflex increased; HRup rats in which the H-reflex failed to increase; H-reflex down-conditioned (HRdown) rats in which the H-reflex decreased; and HRdown rats in which the H-reflex failed to decrease. The results show that AIS plasticity is associated with successful H-reflex conditioning, and they suggest that this plasticity might have a role in this simple form of motor learning.

METHODS

Subjects were 49 young adult Sprague-Dawley rats (male, 14–15 weeks old). Thirty-five underwent electrode implantation followed by either H-reflex up-conditioning (15 HRup rats) or H-reflex down-conditioning (20 HRdown rats). The other 14 rats were not implanted and provided a naïve control group (NC rats). All procedures satisfied the Guide for the Care and Use of Laboratory Animals of the Institute of Laboratory Animal Resources, Commission on Life Sciences, National Research Council (National Academy Press, Washington, DC, 2011) and were reviewed and approved by the Institutional Animal Care and Use Committees of the Wadsworth Center (Protocol D16–00115 (A3183–01)) and the Stratton VA Medical Center (Protocol #1537733). The H-reflex conditioning paradigm, and the methods for motoneuron retrograde labeling, animal perfusion, and histological preparation have been fully described (Wolpaw & Herchenroder, 1990, Chen & Wolpaw, 1995, Wang et al 2006, 2009). They are summarized here. The procedures for histological and immunocytochemical processing and analysis are described in detail. Figure 1b shows the experimental schedule.

Electrode implantation, H-reflex elicitation, and operant conditioning

Under general anesthesia (ketamine HCl, 80 mg ⁄ kg and xylazine, 10 mg ⁄ kg, both i.p.), chronic stimulating and recording electrodes were implanted in the right hindlimb. To elicit the soleus H-reflex, a nerve-stimulating cuff was placed around the right posterior tibial nerve just proximal to the triceps surae branches. To record soleus EMG activity, a pair of fine-wire electrodes were inserted in the right soleus muscle. The Teflon-coated wires from the nerve cuff and muscle passed subcutaneously to a connector plug mounted on the skull.

Data collection started at least 20 days after implantation. During data collection, each rat lived in a standard rat cage with a flexible cable attached to the head plug. The cable, which allowed the animal to move freely about the cage, carried the wires from the electrodes to a commutator above the cage that connected to an EMG amplifier (gain 1000, bandwidth 100–1000 Hz) and a nerve-cuff stimulation unit. The rat had free access to water and food, except that during H-reflex conditioning it received food mainly by performing the task described below. Animal well-being was carefully checked several times each day, body weight was measured weekly. Laboratory lighting was reduced from 2100 to 0600 h each day.

Stimulus delivery and data collection were controlled by a computer that monitored soleus EMG activity (sampled at 5,000 Hz) continuously for the entire period of data collection. The soleus H-reflex was elicited as follows: Whenever the absolute value (equivalent to the full-wave rectified value) of background (i.e., ongoing) EMG activity in the soleus muscle remained within a pre-defined range for a randomly varying 2.3–2.7 s period, the computer initiated a trial. The EMG range was based on the rat’s typical soleus EMG level as it moved about the cage; it was typically 1–2% of the maximum possible EMG activity as assessed by the maximum M wave (i.e., the direct muscle response (Pierrot-Deseilligny & Burke 2012)). In each trial, the computer stored the most recent 50 ms of soleus EMG activity (i.e., the background EMG interval), delivered a monophasic stimulus pulse (typically 0.5 ms in duration) through the cuff on the posterior tibial nerve, and stored soleus EMG activity for the next 100 ms. Pulse amplitude was initially set just above M-wave threshold; it was subsequently automatically adjusted by the computer after each trial to maintain M-wave size unchanged throughout the entire period of data collection. Thus, the background EMG (reflecting soleus motoneuron tone at the time of H-reflex elicitation) and the M-wave (reflecting the effective strength of the nerve-cuff stimulus) remained stable throughout data collection.

Under the control mode, the computer simply digitized and stored the absolute value of EMG activity from the soleus muscle for 100 ms following nerve stimulation. Under the soleus H-reflex conditioning mode, it gave a food-pellet reward 200 ms after stimulation if EMG amplitude in the H-reflex interval (typically 6–10 ms after stimulation) was above (H-reflex up-conditioning mode) or below (H-reflex down-conditioning mode) a criterion value. The criterion was set and adjusted daily as needed so that the rat received an adequate amount of food (~1000 pellets/d for a 500-g rat). H-reflex size was calculated as average EMG amplitude in the H-reflex interval minus average background EMG amplitude and was expressed in units of average background EMG amplitude.

Each rat was first exposed to the control mode for 20 days to determine the control H-reflex size, and then exposed to the up-conditioning (HRup rats) or down-conditioning (HRdown rats) mode for 50 days. The last 10 control-mode days and the last 10 up- or down-conditioning days (i.e., days 41–50 of conditioning) provided the data used to determine the impact of H-reflex conditioning on soleus H-reflex size. Average final (days 41–50) H-reflex size was calculated as percent of the control H-reflex size. Successful H-reflex conditioning was defined as a change of at least 20% in the correct direction (i.e., to ≥120% for HRup rats, to ≤ 80% for HRdown rats) (Wolpaw et al., 1993; Chen & Wolpaw, 1995). Thus, the rats fell into five groups: naive control (NC) rats; successful HRup (US) rats; failed HRup (UF) rats; successful HRdown (DS) rats; and failed HRdown rats (DF) rats.

Soleus motoneuron labeling and spinal cord section

At the end of the 50-day period of HRup or HRdown conditioning, each rat was anesthetized with isoflurane (0.5–1.5% in O₂ via inhalation) and injected in the right soleus muscle with 50 μl 0.1% cholera toxin subunit B-conjugated Alexa Fluor (CTB-Alexa Fluor 647, Life Technology Inc, Eugene, OR) to retrogradely label the soleus motoneurons (Wang et al. 2006). The 14 naive control (NC) rats (i.e., rats not implanted or exposed to H-reflex conditioning) were similarly injected. Three days later, the rats were euthanized with an overdose of sodium pentobarbital and perfused through the aorta with 0.05 M phosphate-buffer containing 137 mM NaCl (PBS, pH 7.3) followed by 4% paraformaldehyde (PFA) in 0.1 M phosphate buffer (Wang et al 2012). The 4% PFA fixative minimized protein degradation during storage at −80°C.

The lumbosacral spinal cord was removed and postfixed for 2 hr, washed with PBS (pH 7.4), and infiltrated with 30% sucrose for 24 hr. The spinal cord containing the soleus motoneuron pool was blocked, embedded in OCT compound (Tissue-Tek), frozen on dry ice, and transversely cut into 25- and 16-μm sections with a cryostat. Sections were mounted onto precoated glass slides (Superfrost; Fisher) and stored at −80°C prior to immunocytochemistry processing.

Every individual processing session included tissue from NC rats as well as tissue from experimental group rats; the NC rats provided an internal control for each session. Furthermore, those processing the tissue were blinded as to which sections came from which rat groups.

Fluorescent immunocytochemistry

Sections for AIS labeling were immersed in sodium citrate buffer (10 mM Tri-sodium citrate, 0.05% Tween 20, pH 6.0) and underwent heat-induced antigen retrieval at 95–100°C for 25 min to break protein cross-linking in the PFA-fixed tissue. After they cooled down to room temperature, they underwent immunofluorescent-cytochemistry staining. The sections were washed three times (7–8 min each) with PBS (pH 7.4), immersed in 0.5% sodium borohydride for 30 min to quench PFA auto-fluorescence, rinsed again with PBS containing 0.1% Triton X-100 (PBST, pH 7.4), blocked with 5% normal donkey serum (in PBST) for 3 hr, and placed in a humid chamber to incubate with primary antibodies diluted in PBST containing 1% BSA at 4°C for 24 hrs. Two primary antibodies were applied simultaneously for labeling different biomarkers on the AIS and the motoneuron. The antibodies (abs) included: mouse anti-ankyrin G monoclonal ab (AnkG, Santa Cruz Biotech, 1:100); rabbit anti-AnkG (SYSY 1:1000); rabbit anti-phosph-Thr¹⁸⁰/Tyr¹⁸²p38 MAPK (p-p38) (PhosphoSolutions 1:100); mouse anti-GAD₆₇ ab (EMD Millipore 1:1000); rabbit anti-GAD₆₇ ab (abcam 1:500); and goat anti-CTB ab (Calbiochem 1:4000).

Sections labeled for AnkG and p-p38 were: (1) washed in HEPES-buffered phosphate-free medium solution (HEPES); (2) blocked with normal donkey serum diluted with HEPES; (3) incubated in HEPES containing p-p38 primary antibody; (4) washed again in HEPES; (5) incubated with one or more secondary antibodies (donkey anti-mouse conjugated with Alexa Fluor 488 (1:2000), donkey anti-rabbit conjugated with Alexa Fluor 546 (1:2000), and/or donkey anti-goat conjugated with Alexa Fluor 647 (1:4000)) for 2 hr to visualize the specific immunoreactivities (IR). Finally, the rinsed sections were mounted with ProLong Gold anti-fade reagent (Life Technology INC) and allowed to cure at 4°C before being photographed.

Image stack acquisition

All epifluorescent microphotographs were taken with a CCD camera (model C8484–03G02, Hamamatsu Photonics) using an Olympus BX61 motorized epifluorescence microscope (Zeiss) with a 40x objective (Olympus UPlanSApo) and a fixed illumination time. The computer-controlled motorized microscope automatically acquired image stacks with a multi-dimensional acquisition program (Metamorph for Olympus) at the appropriate wavelengths (for excitation channels) and z-axis (depth) settings. Stacks were collected to locate the best optical sections for analysis.

Soleus motoneurons were recognized by their CTB labeling in soma and/or primary dendrites and by their Lamina-IX location; to minimize the risk of counting a motoneuron twice, only soleus motoneurons with visible nuclei were assessed (Fig. 1c) (Wang et al. 2006). The motoneuron axon began at the axon hillock or, rarely, on a proximal dendrite. The axon initial segment (AIS) was identified as the part of the axon with AnkG-IR labeling (arrowhead in Fig. 1c). It was considered to be complete if it had a gradually tapering shape, a well-defined distal end with CTB-Alexa 647 fluorescence visible beyond the end, and was at least 15 μm long (see below). With the Metamorph program, each motoneuron or AIS was focused and exposed to two or three different excitation wavelength channels matching the specific fluorescent secondary antibodies; this yielded an image stack comprising 13–15 parallel image layers for each channel of secondary antibodies. Retrograde CTB labeling was usually visualized by Alexa Fluor 647 channel (blue), p-p38 MAPK or GAD₆₇ labeling by Alexa Fluor 546 (red), and AnkG-labeling by Alexa 488 (green).

The sections were cut in the tranverse (i.e., coronal) plane relative to the spinal cord because the axons of spinal motoneurons pass from the motoneuron to a nearby ventral root; thus, their course is predominantly in the transverse plane. This was reflected in our finding that 80–90% of soleus motoneuron AISs were visible in transverse sections, while only 10–12% were visible in sagittal sections. Most motoneuron AISs did fall entirely within a single tranverse section; and the 2-D AIS length and location (distance from soma) measurements provided by our epifluorescence methodology (determined as described below) were expected to be close to the 3-D length and location measurements of the AIS. This expectation was confirmed by the subsequent confocal analysis described below. Thus, we are confident that our epifluorescence measurements provided an accurate assessment of actual AIS lengths and locations.

Prior to analysis, all images were coded by a person who was not otherwise involved in this study. This ensured that all subsequent measurements were obtained in a blinded fashion (i.e., the rater did not know from which rat group a given image came).

Soleus Motoneuron Assessment

Each CTB-labeled soleus motoneuron and its AIS were traced with the Image J program and the following measurements were obtained: soma size and somatic immunoreactivity (IR) for p-p38; and AIS location and size and AnkG-IR or p-p38-IR (illustrated in Fig. 1c,d; details below). After the program subtracted the background IR, the immunoreactivity measurements were converted into percentage of the average value for naive control (NC) rats.

AIS Assessment

AISs that emerged from the soma (or rarely (4%) from a primary dendrite) of CTB-labeled soleus motoneurons were assessed with the Image J program (ver. 1.54f) as follows (Fig. 1d). A line was drawn from the axon hillock to the proximal end of the AIS and along the AIS midline until just past its distal end. The Image J plug-in plot profile program plotted AnkG-IR luminance vs. pixel location along this midline path. These AnkG-IR luminance values were expressed in percent of the highest value. The data were smoothed with a polynominal regression function to yield the final plot of AnkG-IR vs. pixel location along the path (Fig. 1e). The difference between the minimum value from the axon hillock to the AIS and the maximum value for the AIS was determined; and the proximal end of the AIS was defined as the point on the plot where AnkG-IR reached 15% of this difference above the minimum value (left arrow in Fig. 1e). The distal end of the AIS was defined as the point where AnkG-IR fell back to 5% of this difference (right arrow in Fig. 1e).

We defined AIS distance from the soma (i.e., location) as the length (“d1” in Fig.1d,e) of the axon segment from the axon hillock to the proximal end of the AIS (defined above). For the few AISs emerging from a proximal dendrite, AIS location was defined as the length of the axon segment from the dendrite to the proximal end of the AIS. AIS length was defined as the total length of the path from the proximal end of the AIS to its distal end (“d2” in Fig. 1e, f). All AISs ≥15 μm long (i.e., 87.5% of the visible AIS) are included in the analysis presented here. Most of the excluded <15-μm AISs (12.5%) were believed to have been cut short by sectioning or to appear short due to bending that made parts of them perpendicular to the section. As previously noted, all analyses were blinded as to which sections came from which rat groups.

GABAergic terminal assessment

GAD₆₇-IR for soma and for AIS were measured for each soleus motoneuron. For the soma, GAD₆₇-IR was measured along the soma perimeter; for the AIS, which is only 4–6 um in diameter, it was measured for the entire AIS (i.e., the entire AnkG-positive area). The GAD₆₇-positive terminals on each AIS were counted.

Confocal assessment of AIS length and distance from soma

Subsequent to completion of data collection and epifluorescence analysis, we performed a confocal analysis of AIS length and distance from soma (i.e., location). Previously analyzed sections could not be reanalyzed with the confocal microscope because the epifluorescence had faded due to the earlier analysis and the passage of time. Instead, we used remaining tissue to perform confocal analysis of the AISs of 155 soleus motoneurons in AnkG-labeled 25-μm sections from 6 NC, 6 US, and 6 DS rats. (From the lesser numbers of UF and DF rats, we did not have enough remaining tissue to perform a meaningful confocal analysis.)

Confocal images were collected with an Andor spinning disk confocal (Dragonfly) microscope with a 40X/.95 oil immersion lens. A z-stack (0.2-μm depth) of each AnkG-positive MN was obtained using Nyquist sampling rates. The number of slices in each image stack averaged 34(±8SD); the number occupied by an AIS averaged 14±6 (i.e., average total depth 2.8 μm). Only 10 AISs (~6.5%) occupied >35 slices (i.e., >7 μm total depth); these 10 were distributed across NC, US, and DS rats.

To measure AIS length and location, each confocal image stack was imported in high-resolution auto-scaled hyper-stack format using Fiji-Image J with the neuroanatomy plugin. The AIS tracing channel was selected with the SNT (simple neurite tracer) (part of the Image J neuroanatomy plugin) (Arshadi et al 2021). Each AIS was traced as a single 3-D path using the A search algorithm with autotracing enabled. AIS length and location were calculated from the path using the active window of SNTv4.2.1. In accord with our standard practice, this confocal analysis was blinded as to which sections came from which rat groups.

Data analysis

After data acquisition, all results from the renamed images were decoded, collated, and sorted by rat and group. Statistical comparisons were performed among the five groups by nested ANOVA, with rats nested in groups. If this test showed an overall difference among the groups with p<0.05, the difference between each of the experimental groups (US,UF,DS,DF) and the naïve control (NC) group was evaluated by the LSMeans contrast test. Because we were making these four comparisons, we used p<0.0125 as the criterion for significance (i.e., Bonferroni correction).

RESULTS

Animal well-being and postmortem examination

All rats remained healthy and active throughout data collection. Average body weight increased from 397(±63SD) g at the beginning of the study to 553±72 g at the time of perfusion. Right and left soleus muscle weights (measured as percentage of body weight) were 0.042(±0.005SD)% and 0.041(± 0.005)%, respectively. They did not differ significantly from each other (p=0.30 by paired t test), nor from the soleus muscle weights of normal rats (Chen and Wolpaw, 1995, 1997, 2002; Chen et al., 1996, 1999, 2001, 2002, 2005, 2006). Postmortem examination confirmed that the EMG electrodes and the nerve cuffs were located where they had been implanted. The nerve cuffs were covered by connective tissue and the tibial nerves appeared well preserved inside the cuff.

H-reflex conditioning

As Figure 1f and Table I summarize, conditioning had its expected effects on the right soleus H-reflex in the up-conditioned (HRup) and down-conditioned (HRdown) rats. By the standard definition of successful conditioning as a change of at least 20% in the rewarded direction (Chen and Wolpaw, 1995, 1997, 2001; Chen et al., 2006), the conditioned rats fell into four groups: Up-Successful (US; n=11); Up-Failed (UF; n=4); Down-Successful (DS; n=13); and Down-Failed (DF; n=7). Final H-reflex size averaged 154(±32SD)% of its initial value in US rats and 62(±17%) in DS rats (both p=0.008 vs. initial value by one-sided paired t-test). Naive control (NC) rats (n=14) did not undergo electrode implantation, H-reflex data collection, or conditioning. They were simply intact animals comparable in age and body weight to the rats of the experimental groups.

Table 1.

Average (±SD) values from epifluorescence analysis for the five rat groups for: final H-reflex size; motoneuron area and diameter; AIS dimensions; AnkG, p-p38, and GAD67 immunoreactivities (IR); and AIS GABAergic terminal number.

#Rats in Group (n)	NC (13)	US (11)	UF (4)	DS (13)	DF (7)
Final H-Reflex (% initial)	N/A	154±9*	101±14	62±4*	118±5
Motoneuron (n)	233	178	75	138	93
Area (μm2)	1246±284	1247±306	1163±185	1201±272	1175±306
Diameter (μm)	50.7±6.7	51.2±7.6	49.5±6.9	50.6±7.8	50.9±9.6
AIS Dimensions (μm)
Distance from soma	7.1±3.9	9.3±5.1**	8.7±5.1	7.3±3.4	8.8±4.9*
Confocal analysis 1	6.6±2.8	7.9±3.2		6.5±2.4
Length	25.9±3.3	27.7±3.4*	24.9±2.9	25.9±3.3	25.6±3.6
Confocal analysis 1	25.9±3.0	28.4±3.4 **		25.8±3.1
Area (μm2)	61.3±13.6	65.3±13.8	53.3±12.9*	59.7±12.7	58.6±12.9
Immunoreactivity (IR) (% of NC) and Terminal (Term) #
Rat Group (n)	NC (14)	US (11)	UF (4)	DS (13)	DF (7)
AnkG-IR
AIS	100±26	99±35	100±16	83±28*	96±33
p-p38-IR 2
AIS	100±16	99±25	97±12	113±14*	107±18
Soma	100±12	100±18	101±9	109±13	109±16
GAD67-IR 3
AIS	100±35	115±46	96±39	138±55**	90±37
Soma membrane	100±36	114±43	117±39	129±56*	102±33
AIS GABA Term #	3.4±1.5	3.4±1.5	2.8±1.5	5.0±2.0**	3.1±1.8

^*:p<0.0125 (Bonferroni correction) and

^**:p≤0.001 vs NC (exact p values are in the Figures). In italics are average (±SD) values from confocal analysis of AIS dimensions¹.

¹After data collection and epifluorescence analysis were completed, AIS dimensions were assessed by confocal analysis in the tissue remaining from a subset of NC, US, and DS rats.

#rats/#motoneurons/#AIS for NC, US, and DS groups were 6/46/46, 6/56/56, and 6/53/53, respectively.

**: p≤0.001 vs NC (exact p value is in Figure 2d).

²p-p38-IR in AIS was assessed in a subset of rats. #rats/#motoneurons/#AIS for NC, US, UF, DS, & DF groups were 11/175/131, 7/81/75, 4/43/42, 13/160/81, and 7/74/73, respectively.

³GAD67-IR and GABAergic terminal were assessed in a subset of rats. #rats/#motoneurons/#AIS for NC, US, UF, DS, & DF groups were 13/129/109, 9/60/60, 3/18/18, 8/69/61, and 5/39/29, respectively.

Motoneuron and AIS morphology

Somatic diameter and area and AIS location, length, and area were assessed in 717 Alexa Fluor-identified (i.e., CTB-labeled) soleus motoneurons from the five groups of rats. Table 1 summarizes the results. Motoneuron somatic diameter and area did not differ across the groups.

AIS morphology was similar to that reported by Xu and Shrager (2005) and Rotterman et al. (2021). The AIS emerged from the soma in 96% of the motoneurons, and from a primary dendrite in the remaining 4%. AIS location (distance from the soma (or primary dendrite)) varied from 1–30 μm. Most (85%) AIS were 22–32 μm long. Those emerging from the soma were significantly longer than those emerging from a dendrite (26.16 (±3.40SD) and 24.4 (±3.41) μm, respectively) (p=0.003 by t-test). Motoneuron soma area correlated with AIS length (r=+0.23; p<0.001) but not with AIS location (r= −0.04, p=0.31). Thus, the AIS was usually longer in larger motoneurons.

AIS length, location, and area

Table 1 shows for the five rat groups average AIS length, location (i.e., distance from soma (or primary dendrite)), and area. Figure 2a shows the values for the individual rats of the five rat groups (NC, US, UF, DS, DF). Significant differences from NC rats are noted in Table 1; and the exact p values are shown in Figure 2a. AIS length and location were significantly greater in US rats; AIS distance was significantly greater in DF rats; and AIS area was significantly smaller in UF rats. As noted above, the five groups did not differ significantly in motoneuron somatic diameter or area.

Figure 2. Successful H-reflex up-conditioning is associated with greater AIS length and location.

a: AIS location (i.e., distance from soma), length, and area from epifluorescence analysis for: Naïve Control (NC) rats; Successful Up-Conditioned (US) rats; Failed Up-Conditioned (UF) rats; Successful Down-Conditioned (DS) rats; and Failed Down-Conditioned (DF) rats. The box includes the second and third quartiles; its middle line is the median value. The top and bottom bars indicate the extremes. The dotted line is the mean value for the NC rats. P values are shown for groups that are significantly different from the NC rat group with p<0.0125 (t-test with Bonferroni correction). See Methods for full description of the statistical analysis. AIS location and length are both significantly increased in US rats. AIS location is significantly increased in DF rats; and AIS area is significantly decreased in UF rats. DS rats do not differ from NC rats in AIS, length, or area. The moderate decrease in AIS length in UF rats does not reach significance (p=0.029).

b: Final H-reflex size is significantly correlated with AIS length in up-conditioned (US,UF) rats, but not in down-conditioned (DS,DF) rats. Each filled symbol represents a successful (US (red) or DS (blue)) rat; each open symbol represents an unsuccessful (UF (red) or DF (blue)) rat. (As noted in the text, success is indicated by change of at least 20% in the correct direction.)

c1–3: Representative photomicrographs of soleus motoneuron soma (blue) and AIS (green) from an NC rat, a US rat, and a DS rat. The AIS is longer and farther from the soma in the US rat. Note also the weak AnkG-IR in the DS rat (see Fig. 3). (Scale bar 50 μm.)

d: AIS location (i.e., distance from soma) and length from confocal analysis of: Naïve Control (NC) rats; Successful Up-Conditioned (US) rats; and Successful Down-Conditioned (DS) rats. AIS length is significantly increased in US rats. AIS location also appears to be increased in US rats, although its p=0.041 vs. NC rats does not satisfy the Bonferroni-corrected significance value of p<0.025 (i.e., for two tests: US and DS rats, each vs. NC rats).

Figure 2b shows that final H-reflex size correlated positively with AIS length in up-conditioned (US, UF) rats; it did not correlate with AIS location. Thus, larger final H-reflex size was associated with longer AIS length. In down-conditioned (DS, DF) rats, final H-reflex size did not correlate with AIS length or location. Figure 2c shows representative AISs. AIS length and location (i.e. distance from the soma) are greater in the US rat.

Furthermore, as Table 1 and Figure 2d show, the later confocal analysis of AIS location and length in tissue remaining from NC, US, and DS rats supported the epifluorescence analysis. In US rats, AIS length was significantly increased vs. NC rats (p=0.001); and AIS location was also increased, though its p=0.041 did not reach the Bonferroni-corrected p<0.025 (i.e., for two tests: US and DS rats, each vs. NC rats). DS rats did not differ from NC rats in length or location.

The changes in AIS location in DF rats and AIS area in UF rats add to indications in our previous studies that conditioning failure does not necessarily mean that nothing has happened. The changes found in DF and UF rats may represent compensatory plasticity associated with preservation of other behaviors that use the same motoneurons (see Discussion).

AnkG immunoreactivity (IR)

Table 1 shows for the five rat groups average AIS AnkG-IR in percent of the naïve control (NC) group values. Figure 3a shows the values for the individual rats of the five rat groups. Significant differences from NC rats are noted in Table 1, and exact p values are shown in Figure 3a. In DS rats, AnkG-IR was significantly weaker in the AIS. In contrast, US, UF, and DF values were not significantly different from NC values.

Figure 3. Successful H-reflex down-conditioning is associated with weak AnkG-IR.

a: AIS AnkG-IR (in % of mean values for naive control (NC) rats) for: Successful Up-Conditioned (US) rats; Failed Up-Conditioned (UF) rats; Successful Down-Conditioned (DS) rats; and Failed Down-Conditioned (DF) rats. The box includes the second and third quartiles; its middle line is the median value. The top and bottom bars indicate the extremes. The dotted line is the mean value for the NC rats. The p value is shown for the group that is significantly different from the NC rat group with p<0.0125 (t-test with Bonferroni correction). See Methods for full description of the statistical analysis. AIS AnkG-IR is significantly weaker in DS rats. AIS AnkG-IR does not differ from that of NC rats in DF, US, and UF rats.

b: Final H-reflex size is positively correlated with AnkG-IR in down-conditioned (DS,DF) rats. In contrast, final H-reflex size is negatively correlated with AnkG-IR in up-conditioned (US,UF) rats. Filled symbols represent successful (US (red) and DS (blue)) rats; open symbols represent unsuccessful (UF (red) and DF (blue)) rats. As noted in the text, success is indicated by change of at least 20% in the correct direction.

c: Representative photomicrographs of AIS labeled for AnkG-IR (green) from an NC rat, a US rat, and a DS rat. AnkG-IR (shown in the optical section with the largest and best-focused AIS image) is similarly intense in the AISs of the NC rat and the US rat. In contrast, AnkG-IR is weaker and shows less clustering in the AIS of the DS rat. (Scale bar 20 μm.) The decreased background in the DS rat image is likely to reflect similarly weak AnkG-IR in the AISs of the many surrounding soleus motoneurons.

Figure 3b shows that final H-reflex size in down-conditioned (DS, DF) rats correlated positively with AIS AnkG-IR. Thus, H-reflex decrease correlated with weaker AIS AnkG-IR. Figure 3c shows representative AIS examples from control, US, and DS rats labeled for AnkG (green). The weak AnkG-IR in the DS rat is evident. The associated weak background is likely to reflect similarly weak AnkG-IR in the AISs (and possibly also at the nodes of Ranvier) of the many surrounding soleus motoneurons.

Another aspect of Figure 3b is also of interest. As Table 1 indicates, neither successful nor failed up-conditioning was associated with significant change in AIS AnkG-IR; in neither US nor UF rats does this measure differ from that of NC rats. However, as Figure 3b shows, final H-reflex size in up-conditioned (US, UF) rats is negatively correlated with AIS AnkG-IR. If the weak AnkG-IR in DS rats does contribute to their smaller H-reflexes, its negative correlation with final H-reflex size in up-conditioned rats implies that the large H-reflex increase in US rats would be even larger if this negative correlation were absent. The potential import of this finding is considered in the last section of the Discussion.

Phosphorylated Mitogen-activated Protein Kinase (p-p38) immunoreactivity (IR)

Phosphorylated p38 mitogen-activated protein kinase (p-p38MAPK) promotes internalization of sodium channel Na_v1.6 (Wittmack et al 2005, Gasser et al 2010). Channel Na_v1.6 is prominent in the AIS and is thought to have a lower firing threshold than other sodium channels (Royeck et al. 2008; Akin et al. 2015). Thus, a decrease in Na_v1.6 concentration or availability (due to internalization) might be expected to reduce AIS excitability. As Table 1 summarizes and Figure 4 illustrates, p-p38-IR was significantly stronger in the AISs of DS rats. In US, UF, and DF rats, p-p38-IR in AIS and soma did not differ from NC rats.

Figure 4. Successful H-reflex down-conditioning is associated with stronger AIS p-P38-IR.

a: p-p38-IR in AIS and soma (in % of mean values for naive control (NC) rats) for: Successful Up-Conditioned (US) rats; Failed Up-Conditioned (UF) rats; Successful Down-Conditioned (DS) rats; and Failed Down-Conditioned (DF) rats. The box includes the second and third quartiles; its middle line is the median value. The top and bottom bars indicate the extremes. The dotted line is the mean value for the NC rats. P values are shown for groups that are significantly different from the NC rat group with p<0.0125 (t-test with Bonferroni correction). See Methods for full description of the statistical analysis. p-p38-IR is significantly stronger in the AIS of DS rats. p-p38-IR in US, UF, and DF rats does not differ from that in NC rats.

b: Representative photomicrographs of soleus motoneuron soma and AIS from an NC rat, a US rat, and a DS rat triple-labeled for CTB (blue), AnkG (green), and p-p38 (red). The enlarged insets show their AISs labeled for p-p38. In the NC and US rats, AIS p-p38-IR is weak and confined to the distal AIS. In contrast, in the DS rat, p-p38-IR is strong in the soma and throughout the AIS.

GAD₆₇ immunoreactivity and AIS terminal contacts

As Table 1 summarizes and Figure 5 illustrates, down-conditioning has its expected effects (Feng-Chen, Wolpaw 1996; Wang et al 2006; 2009) on soma GAD₆₇-IR: GAD₆₇-IR was significantly greater in DS rats than in NC rats. AIS GAD₆₇-IR was also significantly greater in DS rats than in NC rats; and the number of GABAergic terminals on the AIS was greater in DS rats than in NC rats. US, UF, and DF rats did not differ from NC rats in any of these measures. Finally, as Figure 5b shows, AIS GAD₆₇-IR and GABAergic terminal number correlate significantly with H-reflex size. Thus, increased AIS GAD₆₇-IR or AIS GABAergic terminal number was associated with decreased H-reflex size.

Figure 5. Successful H-reflex down-conditioning is associated with stronger AIS and somatic GAD₆₇-IR and more AIS GABAergic terminals.

a: AIS and Soma GAD₆₇-IR (in % of mean values for naive control (NC) rats) and AIS GABAergic terminal numbers for: Successful Up-Conditioned (US) rats; Failed Up-Conditioned (UF) rats; Successful Down-Conditioned (DS) rats; and Failed Down-Conditioned (DF) rats. The box includes the second and third quartiles; its middle line is the median value. The top and bottom bars indicate the extremes. The dotted line is the mean value for the NC rats. P values are shown for groups that are significantly different from the NC rat group with p<0.0125 (t-test with Bonferroni correction). See Methods for full description of the statistical analysis. AIS and somatic GAD₆₇-IR and the number of AIS GABAergic terminals are significantly greater in DS rats. DF, US, and UF rats do not differ from NC rats in these measures.

b: Final H-reflex size is significantly correlated with AIS GAD₆₇-IR and with GABAergic terminal number in down-conditioned (DS,DF) rats, but not in up-conditioned (US,UF) rats. Stronger AIS GAD₆₇-IR and more AIS GABAergic terminals correlate with smaller H-reflexes. Filled symbols represent successful (US (red) and DS (blue)) rats; open symbols represent unsuccessful (UF (red) and DF (blue)) rats. As noted in the text, success is indicated by change of at least 20% in the correct direction.

c: Representative photomicrographs of soleus motoneuron somata (insets) and their AISs from an NC rat, a US rat, and a DS rat, with triple-immunofluorescent labeling for CTB (blue somata in insets), AnkG (green AISs), and GAD₆₇ (red terminals). AIS GABAergic terminals (white arrows) are similar in number in NC and US rats and more numerous in the DS rat. (Scale bars: 5 μm for main images; 50 μm for insets.)

DISCUSSION

Neuronal action potentials usually begin in the axon initial segment (AIS); thus, AIS plasticity can affect neuronal excitability (Araki, Otani 1955; Coombs et al. 1957; Buffington, Rasband 2011; Kole, Stuart 2012). The AIS mechanisms that initiate action potentials display activity-dependent plasticity during development, in pathological situations, and in response to laboratory manipulations (Yoshimura, Rasband 2014; Petersen et al. 2017; Leterrier 2018; Huang, Rasband 2018 for review). AIS plasticity may also contribute to homeostatic plasticity (Jamann et al. 2021). Whether AIS plasticity contributes to the acquisition of motor skills during adult life is not known. To begin to address this question, the present study asked whether simple motor learning – operant conditioning of a spinal reflex – in adult rats is associated with changes in the AIS of the spinal motoneuron that produces this reflex, and, if so, whether this AIS plasticity might contribute to the change in reflex size.

Operant conditioning of the H-reflex, the electrical analog of the spinal stretch reflex (e.g., the knee-jerk reflex), is a well-characterized model for studying motor learning and the plasticity in spinal cord and brain that underlies it (Wolpaw, Tennissen 2001; Wolpaw 2006; Wolpaw, Chen 2009; Pierrot-Deseilligny, Burke 2012; Pearson, Gordon 2013). The spinal plasticity includes changes in spinal motoneuron properties and in several synaptic populations on the motoneuron (reviewed in Wolpaw, Chen 2009; Thompson, Wolpaw 2014; Wolpaw 2018; Wolpaw, Kamesar 2022). Primate and rodent data show that successful down-conditioning of the H-reflex is accompanied by a positive shift in motoneuron firing threshold that can largely account for the smaller H-reflex. In both primates and rats, this shift is accompanied by a drop in motoneuron axonal conduction velocity that may reflect a similar shift in threshold at each node of Ranvier (Carp, Wolpaw 1994; Halter et al. 1995; Carp et al. 2001). Unsuccessful down-conditioning is not associated with these changes in motoneuron firing threshold and axonal conduction velocity. Our computational modeling indicates that the most plausible mechanism of these changes is a 2- to 3-mV positive shift in the activation voltage of sodium channels in the motoneuron membrane (Halter et al. 1995). H-reflex up-conditioning is not a mirror image of down conditioning (Carp, Wolpaw 1995); its mechanisms may include other changes in the motoneuron, in oligosynaptic pathways that contribute to the H-reflex, and/or in other spinal interneurons (e.g., Feng-Chen, Wolpaw 1996; Wolpaw, Chen 2001).

This study of the motoneuron AIS, the usual site of action potential generation, was motivated by the strong evidence that a positive shift in motoneuron firing threshold caused the smaller H-reflex produced by down-conditioning; it was also guided by computational modeling by others (Gulledge, Bravo 2016) and by the AIS changes that occur in development, with disease, and with laboratory manipulations (Yoshimura, Rasband 2014; Petersen et al. 2017; Leterrier 2018; Huang, Rasband 2018 for review).

The study was blinded and quantitative, and the main results are clear. The Abstract Figure illustrates them. Successful H-reflex up-conditioning is associated with changes in AIS length and location (i.e., distance from the soma); in contrast, successful down-conditioning is associated with changes in AIS proteins (AnkG and p-p38) that are closely associated with AIS sodium channel function, and in GABAergic input to the AIS. For both up-conditioning and down-conditioning, the magnitudes of several of these AIS changes correlate with the magnitude of H-reflex change. Furthermore, unsuccessful conditioning is not associated with these AIS changes. These results are, to our knowledge, the first evidence that AIS plasticity accompanies, and might possibly contribute to normal learning.

This discussion addresses the implications of these results for H-reflex conditioning specifically and for motor learning in general. It focuses on: (1) AIS plasticity associated with H-reflex up-conditioning; (2) AIS plasticity associated with down-conditioning; (3) how the reward contingency might guide the AIS plasticity; and (4) the impact of this AIS plasticity on the many other behaviors that use the same spinal motoneurons.

AIS plasticity with up-conditioning of the H-reflex

Successful up-conditioning (US rats) was associated with significantly greater AIS length and AIS location (i.e., distance from the motoneuron soma). Furthermore, in US rats, final H-reflex size correlated significantly with AIS length. Neither unsuccessful up-conditioning (UF rats) nor successful down-conditioning (DS rats) was associated with such changes. Table 1 and Figure 2 summarize and illustrate these results.

Changes in AIS length and location can affect AIS function; they can thereby affect neuronal excitability (Baalman et al. 2013; Yoshimura, Rasband 2014; Yamada, Kuba 2016; Jamann et al 2021). Their correlations with excitability vary across neuronal types and laboratory models (e.g., Grubb, Burrone 2010; Chand et al. 2015; Wefelmeyer et al. 2015; Jensen et al. 2020; Rotterman et al. 2021). Gulledge and Bravo (2016) and Goethals and Brette (2020) used computational modeling to assess the impact of AIS length and location on neuronal excitability. For large neurons with extensive dendritic trees, such as spinal motoneurons, their modeling predicts that increases in AIS length and/or location (i.e., distance from the soma) will increase excitability. Thus, these studies suggest that the increases in AIS length and location found in the present study might contribute to the larger H-reflex in US rats. The significant positive correlation (Fig. 2b) between final H-reflex size in up-conditioned rats and AIS length also supports this possibility. In addition, Jensen et al. (2020) found that increases in AIS length and location were associated with increased excitability. However, Rotterman et al. (2021) found that greater AIS location (i.e., distance from soma) was associated with decreased excitability. While both studied rat gastrocnemius motoneurons, the mean AIS location was nearly twice that in Jensen et al. (2020) or in the present study (Table 1)). Given the many factors that affect AIS excitability, the greater mean location in Rotterman et al. (2021) than in Jensen et al. (2020) or in the present study may have a role in the differing relationship reported.

Changes in AIS length or location may result from increases or decreases in AnkG and/or AnkB that shift the AIS intra-axonal boundaries (Galiano et al. 2012)). In our study, the AIS length and location increases in US rats were not associated with AnkG-IR change; AnkB-IR was not assessed.

In summary, the significant positive correlation between greater AIS length and H-reflex increase, the associated increase in AIS location, the absence of these changes in rats in which H-reflex up-conditioning was unsuccessful, and computational modeling studies in the literature suggest that this AIS plasticity could possibly contribute to the operantly conditioned increase in the H-reflex. At the same time, the impact on action potential generation of interactions among AIS, soma, and somatic synaptic inputs, and the variations in these interactions across neuronal types remain to be clearly defined (e.g., Balbi et al. 2015; Kole & Brett 2018; Goethals et al. 2021). Better understanding of these factors should clarify the potential relationship of the AIS plasticity described here to the associated behavioral change (i.e., a larger or smaller soleus H-reflex).

AIS plasticity with down-conditioning of the H-reflex

The results for down-conditioning were very different from those for up-conditioning. Successful down-conditioning (i.e., DS rats) was not associated with change in AIS length or location. However, unlike successful up-conditioning, successful down-conditioning was associated with decrease in the AIS scaffolding protein ankyrin G (AnkG), increase in AIS phosphorylated p38 mitogen-activated protein kinase (p-p38MAPK) (p-p38), and increase in GABAergic synaptic input to AIS. As Table 1 summarizes and Figures 3–5 illustrate, in DS rats AIS AnkG-IR was decreased, AIS p-p38-IR was increased, and AIS GAD₆₇-IR and GABAergic terminal number were increased. Unsuccessful down-conditioning (i.e., DF rats) was not associated with these changes.

Voltage-gated sodium (Na_v) channels in the AIS have a central role in action potential initiation. The cytoskeletal protein AnkG enables appropriate clustering of AIS Na_v channels, and is important for their function (Jenkins, Bennett 2001; Gasser et al. 2012; Huang, Rasband 2018). Thus, the decrease in AIS AnkG in DS rats (indicated by the weak AIS AnkG-IR) may potentially contribute to the smaller H-reflex. The significant positive correlation of final H-reflex size with AnkG-IR is also consistent with the hypothesis that the AnkG decrease contributes to the smaller H-reflex. A similar decrease might possibly underlie the positive shift in motoneuron firing threshold and the smaller H-reflex found in monkeys in which down-conditioning is successful (Carp, Wolpaw 1994).

Several protein kinases have post-translational effects on AIS Na_v channels (Bréchet et al. 2008; Hedstrom et al. 2008; Wittmack et al. 2005; Fotia et al. 2004; Gasser et al. 2010; Ogino et al. 2015; Galiano et al. 2012). These include phosphorylated p38 mitogen-activated protein kinase (p-p38MAPK), which promotes ubiquitination and internalization of the Na_v1.6 sodium channels that are likely to be most responsible for setting the motoneuron firing threshold (Royeck et al. 2008; Akin et al. 2015). By increasing Na_v1.6 sodium channel internalization, the p-p38MAPK increase in DS rats (indicated by increased AIS p-p38-IR) might produce the change in sodium channels that our modeling (Halter et al. 1995) suggested was the most likely origin of the positive shift in firing threshold that could largely account for the H-reflex decrease produced by successful down-conditioning.

GABAergic inhibitory input regulates neuronal excitability in many settings (Wefelmeyer et al, 2015; Hines et al, 2018; Nathanson et al, 2019). Previous studies in rats and monkeys found that successful down-conditioning of the H-reflex was associated with significant increase in inhibitory terminals (monkeys) and in GABAergic terminals specifically (rats) on the spinal motoneurons that produce the reflex; furthermore, unsuccessful down-conditioning was not associated with such increases (Feng-Chen, Wolpaw 1996; Wang et al 2006, 2009). The present study extends this finding to the motoneuron AIS. Successful down-conditioning was associated with significant increases in AIS GAD67-IR and number of GABAergic terminals on the AIS. Unsuccessful down conditioning was not associated with these increases. These results are consistent with studies showing that GABAergic input to AIS reduces neuronal excitability in dentate granule cells, in cortical pyramidal cells, and in hippocampus (Wang et al 2014; Gojas et al 2011; Wefelmeyer et al 2015; Hines et al 2019). The present results, together with these earlier studies, are consistent with the hypothesis that increased GABAergic inhibition of AIS could contribute to the acquisition of a new motor skill (i.e., operantly conditioned H-reflex decrease). This contribution would appear to be in addition to the contribution made by the decrease in AIS AnkG and increase in p-p38. For example, the increase in AIS GABAergic terminals, which seems to be most prominent on the proximal AIS (e.g., Fig. 5c (DS image), might impede generation and propagation of the action potential (Wefelmeyer et al 2015; Duflocq et al 2008, 2011). Finally, to the best of our knowledge, GABAergic terminals on the AIS of spinal motoneurons have not previously been reported.

As Table 1 indicates, successful down-conditioning was also associated with significant somatic GAD67-IR increase. If this somatic change extends to the nodes of Ranvier in the motoneuron axon, it may possibly explain the decreased motoneuron axonal conduction velocity that is associated with successful, but not unsuccessful down-conditioning in both rats and monkeys (Carp, Wolpaw 1994, 1995; Carp et al. 2001). This hypothesis is consistent with our computational modeling suggesting that the positive shift in motoneuron firing threshold found in successfully down-conditioned monkeys and the accompanying decrease in conduction velocity are most readily explained by change in sodium channel function (Halter et al. 1995).

In summary, the results are consistent with the hypothesis that AIS plasticity may contribute to operantly conditioned decrease in the H-reflex, and they suggest the responsible mechanisms. They do this by demonstrating that: (1) successful down-conditioning is associated with significant decrease in AIS AnkG-IR and significant increases in AIS p-p38MAPK-IR, GAD67-IR, and number of AIS GABAergic terminals; (2) the changes in AIS AnkG-IR, GAD67-IR, and GABAergic terminal number correlate significantly with the H-reflex decrease; (3) unsuccessful down-conditioning is not associated with these changes. Furthermore, the hypothesis that these changes contribute to operantly conditioned H-reflex decrease is consistent with our computational modeling (Halter et al. 1995) and with the contributions of AIS plasticity and increases in AIS GABAergic terminals to changes in neuronal excitability that occur in development, in disease, and in other laboratory models (Bréchet et al 2008; Kuba 2012; Barry et al 2014; Hines et al 2018; Nathanson et al 2019).

The reward contingency and AIS plasticity

The up-conditioning mode rewards H-reflexes larger than a criterion; the down-conditioning mode rewards those that are smaller. It is presumably these reward contingencies that produce the AIS changes found in US rats and the very different AIS changes found in DS rats. As the previous sections indicate, this mode-specific AIS plasticity may possibly contribute to the change in reflex size. The reward contingency has its effect in the brain. The corticospinal tract (CST) is the only pathway from brain to spinal cord that is essential for successful conditioning (Chen, Wolpaw 1997, 2002). While the rat CST is not thought to contact lumbosacral spinal motoneurons directly, it does contact spinal interneurons (Yang, Lemon 2003). Successful conditioning is associated with mode-specific changes in GABAergic and other synaptic inputs on the motoneuron that may come from these interneurons (Feng-Chen, Wolpaw 1996; Wang et al. 2006,2009,2012). These inputs are likely to convey the essential CST influence through synapses on the motoneuron soma and dendrites and/or near or on the AIS (Wefelmeyer et al. 2015; Christie, De Blas 2003; Mitchell et al 2016, Ueno et al. 2018). Further studies are needed to establish the CST to spinal interneuron pathway responsible for conveying the essential descending input to the motoneuron, and to define the intra-motoneuronal mechanisms through which this input changes the AIS. As discussed above, increased GABAergic inhibitory input to AIS might contribute to H-reflex down-conditioning.

Behavioral impact of the AIS plasticity associated with H-reflex conditioning

The spinal cord is, in Sherrington’s terminology, the final common pathway for behavior. All muscle-based behaviors pass through it (or through its analogous brainstem nuclei) to reach the muscles. In a similar fashion, the spinal motoneuron is the final common pathway for the spinal cord, and the motoneuron AIS is the final common pathway for the motoneuron. AIS properties determine whether and when the many synaptic inputs to the motoneuron combine to produce an action potential and thereby excite the muscle fibers of the motor unit. In short, excitation of the motoneuron AIS produces behavior.

In the past, the spinal cord was assumed to be hardwired; once developed, its neuronal and synaptic components were thought to remain stable through life. It thereby provided a reliable final common pathway for the many behaviors that use it. However, the advances of the past 50 years have shown that the spinal cord, like the rest of the CNS, is plastic through life (for review: Mendell 1984; Wolpaw, Tennissen 2001; Wolpaw 2010; Bertrand, Cazalets 2013). This new knowledge raises a new question. Given that the spinal cord changes as new motor skills are acquired, and as growth, aging, and other life events occur, how are motor skills maintained through life? Why are they not continually disrupted by the ongoing plasticity of the spinal cord?

The heksor and negotiated equilibrium concepts were introduced to begin to address this question (Wolpaw 2010 & 2018). These concepts are formally explicated and supported in Wolpaw and Kamesar (2022). A heksor (from the ancient Greek hexis) is a network of neurons and synapses that produces a skill. Heksors often extend from cortex to spinal cord. Thus, they often overlap each other; they share neurons and synapses. Each heksor changes itself continually to ensure that its skill remains satisfactory despite plasticity caused by other heksors, by creation of new heksors (i.e., acquisition of new skills), and by growth, aging, and other life events. Through their concurrent changes, heksors negotiate the properties of the neurons and synapses that they all use. They keep the CNS, including the spinal cord, in a state of negotiated equilibrium that serves them all.

AIS plasticity is part of this ongoing negotiation; physiological or pathological changes in AIS function may evoke compensatory plasticity that counteracts their functional effects, responses that could be labeled homeostatic plasticity (Huang, Rasband 2018; Jamann et al. 2021; Wolpaw, Kamesar 2022). The present study shows that AIS plasticity occurs with learning, specifically with the creation of the new heksor responsible for a larger or smaller H-reflex. As Table 1 summarizes and the Abstract Figure illustrates, the most prominent aspects of this AIS plasticity might be part of the new heksor; that is, they might contribute to the new skill.

The results may provide some evidence for compensatory AIS plasticity as well. Up-conditioned rats (US or UF) did not differ from naïve control (NC) rats in AnkG-IR (Table 1). However, H-reflex size in up-conditioned rats was negatively correlated with AnkG-IR (Fig. 3b). As the H-reflex got larger, AnkG-IR got weaker. This correlation, which might be expected to reduce the H-reflex increase associated with up-conditioning, may reflect compensatory plasticity in old heksors. Without this compensation, the H-reflex might be even larger. This correlation may represent one aspect of the negotiation that ensures that the creation of a new heksor does not prevent old heksors from maintaining their skills.

Conclusion

Successful H-reflex up-conditioning and successful down-conditioning are associated with very different effects on the motoneuron AIS. Successful up-conditioning is associated with greater AIS length and distance from the soma. AIS length increase correlates with H-reflex increase. Unsuccessful up-conditioning is not associated with these AIS changes. Successful down-conditioning is associated with weaker AIS AnkG-IR, stronger AIS p-p38MAPK-IR and GAD67-IR, stronger somatic GAD67-IR, and more AIS GABAergic terminals. The AIS AnkG-IR, GAD67-IR, and GABAergic terminal changes correlate with H-reflex decrease. Unsuccessful down-conditioning is not associated with these changes. Related data and computational modeling suggest that these AIS changes could potentially contribute to the operantly conditioned changes in H-reflex size. In sum, AIS plasticity is associated with and may possibly contribute to this simple motor learning. The results imply that the properties of the motoneuron AIS reflect the combined influence of the many heksors that use the motoneuron to produce their skills.

Key Points:

• Neuronal action potentials normally begin in the axon initial segment (AIS). AIS plasticity affects neuronal excitability in development and disease. Whether it does so in learning is unknown.

• Operant conditioning of a spinal reflex, a simple learning model, changes the rat spinal motoneuron AIS. Successful, but not unsuccessful, H-reflex up-conditioning is associated with greater AIS length and distance from soma. Successful, but not unsuccessful, down-conditioning is associated with more AIS GABAergic terminals, less ankyrin G, and more p-p38 protein kinase.

• The associations between AIS plasticity and successful H-reflex conditioning are consistent with those between AIS plasticity and functional changes in development and disease, and with those predicted by modeling studies in the literature.

• Motor learning changes neurons and synapses in spinal cord and brain. Because spinal motoneurons are the final common pathway for behavior, their AIS properties probably reflect the combined impact of all the behaviors that use these motoneurons.

Biographies

Yu Wang is an NIH-funded principal investigator in the National Center for Adaptive Neurotechnologies at the Stratton VA Medical Center in Albany, New York. She received her Ph.D. from Nanjing University of China in 2002 and postdoctoral training at the Wadsworth Center of the New York State Department of Health. Her research is elucidating the neuronal and synaptic substrates of H-reflex conditioning (a simple form of motor learning). She is also exploring the neuroanatomical and genomic bases of the long-term effects of cortical stimulation on spinal cord function, with a focus on potential therapeutic applications.

Yi Chen is an NIH-funded research scientist in the National Center for Adaptive Neurotechnologies at the Stratton VA Medical Center in Albany, New York. He received his Ph.D. from Ohio State University and postdoctoral training at the Wadsworth Center of the New York State Department of Health. His graduate studies demonstrated for the first time that spinal reflex conditioning could improve motor function in rats with incomplete spinal cord injuries. His research is defining the CNS plasticity underlying spinal reflex conditioning and is exploring the application of reflex conditioning and other forms of stimulation to restore motor skills that have been impaired by CNS injury or disease.

Lu Chen was, prior to her recent retirement, a Research Scientist of the NIH-funded National Center for Adaptive Neurotechnologies at the Stratton VA Medical Center in Albany, New York. Over 30 years, she made key contributions to development and validation of the novel surgical techniques and other unique methodologies needed for long-term 24/7 interactive collection of physiological data from freely moving rats, and she participated in the design, implementation, analyses, presentation, and publication of numerous studies that used these methods.

Bruce Herron is a research scientist at the Wadsworth Center in the New York State Department of Health and an Assistant Professor within the School of Public Health at the University at Albany, New York. He received his Ph.D. from the University at Albany in 1999 and postdoctoral training in the Division of Genetics at Harvard Medical School. His research in mammalian genomics focuses on factors that contribute to neurological development and disease. In addition to his current interest in discovering the molecular mechanisms of motor learning, he has mapped genetic factors that influence complex traits including epilepsy and angiogenesis.

Xiang Yang Chen was, prior to his recent retirement, a Research Scientist of the NIH-funded National Center for Adaptive Neurotechnologies at the Stratton VA Medical Center and an associate Professor of Biomedical Sciences at the State University of New York in Albany. He is a neurophysiologist who engaged in basic neuroscience research for over 30 years. He led development and use of spinal reflex operant conditioning in rats to explore the plasticity underlying learning and also to enhance functional recovery after CNS injury. His group’s research was funded for many years by NIH and private foundations and their discoveries are now featured in basic textbooks.

Jonathan R. Wolpaw is Director of the NIH-funded National Center for Adaptive Neurotechnologies at the Stratton VA Medical Center and a Professor of Biomedical Sciences at the State University of New York in Albany. He is a neurologist who has been engaged in basic and clinical neuroscience research for 50 years. He and his colleagues developed and are using operant conditioning of spinal reflexes to explore the plasticity underlying learning and to enhance functional recovery for people with spinal cord injuries and other disorders. His research program has been funded for many years by NIH, other federal agencies and private foundations, and has received numerous national and international awards.

Data Availability Statement

All relevant processed data are available in the paper and/or in the references cited. Those wishing to see the entire raw data set may obtain it by contacting the authors (wolpaw@neurotechcenter.org).