Rats were sedated with ketamine (80 mg/kg, IP) during stimulation and recording for all experiments. Animals were placed on a table in a prone position with normal body temperature maintained with a heating blanket. The forepaw was placed in a posture that extended the wrist without inducing background electromyographic (EMG) activity. The sedation level was checked by monitoring the breathing rate, spontaneous vibrissae whisking, and hindlimb withdrawal to toe pinch. Supplemental doses of ketamine (25 mg/kg) were administered as needed to maintain the required sedation during the experiment. After final testing, rats were euthanized by IP administration of a Ketamine/Xylazine overdose.

Two 1.5 cm × 2 cm gel patch electrodes (StimTent Com.) were used to deliver tsDCS. First, the hair over the dorsal neck and chest of the animal was shaved and further removed with Nair. Second, electrically conductive adhesive was sprayed over the contact surface of electrodes, to optimize and stabilize electrical conduction between the skin and electrodes, and then applied to fixed locations in each animal (dorsally, over C4-T1 and the other was placed over the chest; Figure 1A) according to the results of prior modeling experiments (Song et al., 2015). The polarity of tsDCS stimulation was referenced to the dorsal electrode (cathode). tsDCS was generated with an analog isolated stimulator (model 2200, A-M Systems), that was controlled by an analog output channel of an experimental control and data acquisition system (CED, Inc., Cambridge, United Kingdom). Twenty seconds of c-tsDCS (-3 mA) or a-tsDCS (+3 mA) was tested with a 3s ramp for both up and down phases (Figures 1A,B). To avoid the effect of cathodal stimulation on the anodal response, and vice versa, we typically recorded unit activity either in response to one or the other polarity during a given day. The tsDCS intensity of ±3 mA was chosen based on our previous observation and a modeling study indicating consistent effects on MEP enhancement with c-tsDCS (Song et al., 2015). This intensity (current density: 1 mA/cm²; total charge density: 26 mC/cm²) could induce consistent effects, and is below the threshold value for tissue damage (Yuen et al., 1981; McCreery et al., 1990; Liebetanz et al., 2009). This current produced reddening of the skin, especially at the edges of the electrode, but no damage.

Pairs of PFA-insulated stainless steel microwire electrodes (0.002” diameter, A-M Systems) were deinsulated at the tip (about 1 mm) and were inserted into the flexor carpi radialis (FCR) and/or extensor carpi radialis (ECR) muscles. EMGs from the recorded muscles were filtered (300–5000 Hz) and amplified (×1,000), and then digitized at a sampling frequency of 10 kHz with an acquisition system (CED, Inc). Raw EMG records (Figure 1C) were analyzed offline for single motor units. Spikes of single motor units were sorted according to principle component analysis (PCA) or template matching from the recorded EMG signals (Figures 1C,D) and further manually cluster-cut with a customized script written with Spike 2 (Figure 1A, inset). The firing rate of each sorted single motor unit was smoothed with a 1s sliding window, and the ensemble response was constructed from the smoothed firing of all the recorded motor units during c-tsDCS and a-tsDCS. The majority of the recorded motor units were from the ECR muscle, with additional recordings from FCR (c-tsDCS: ECR, 16; FCR, 4) and 22 (a-tsDCS ECR, 17; FCR, 5). No differences were found between ECR and FCR responses; thus, we combined data from the two muscles.

We classified two response components of motor units (Figure 1E). The (1) direct response corresponds to unit firing during the ramp and plateau phases of tsDCS and (2) the persistent response corresponded to firing after the tsDCS had returned to baseline. Motor units were characterized according to the following five metrics: (1) the percentage of units that showed a direct response during tsDCS; (2) the firing rate during tsDCS; (3) the percentage of units that showed persistent firing after tsDCS; (4) persistent firing duration, which is the duration of elevated firing (over baseline) after cessation of tsDCS; and (5) persistent response gain, which is the ratio of the area of the persistent unit response over area of the direct response.

Since PICs are primarily mediated by an increase in inward calcium current (Li and Bennett, 2003), we used the L-type calcium channel blocker Nimodipine (125 mg, Sigma Inc., Burlington, MA, United States, USP grade) to determine if the persistent effect of tsDCS on motor unit firing is mediated by a calcium PIC. Nimodipine is a non-selective L-type Ca²⁺ channel blocker (Carlson et al., 2020). Nimodipine was dissolved in vehicle (ethanol, DMSO, polyethylene glycol and saline in the following proportions: 1:1:8:10), and then sterilized using a syringe filter (30 μm), and stored in a sterile dark bottle at room temperature. It was injected into the tail vein over a 3–5 min period (5 mg/kg). Induction time for the drug is approximately 30 min after IP injection (Marcantoni et al., 2020). We choose to use IV administration to obtain a faster response. We examined open field behavior after Nimodipine administration and did not observe any behavioral changes at this dose (data not shown).

Experimental observations were compared with a computer simulation using a two-compartment neuron model. With this model, PICs were previously found to be changed after spinal cord injury (Kurian et al., 2011). This model also has been used to study normal spinal motor neuron behavior (Booth et al., 1997; Kim, 2017). The motor neuron was modeled with two simplified compartments: dendritic and axosomatic. Both compartments are described by active and passive conductances following the Hodgkin-Huxley formalism. The somatic compartment contains ionic conductances that generate action potentials. The dendritic compartment contains conductances responsible for plateau potentials for generating PICs. Voltage-dependent Na and Ca²⁺ channels were modeled in the dendrite only. The model included the following dendritic conductances: Na_p (Sodium, persistent); K_Ca (Potassium, Ca²⁺ dependent); Ca_p (Calcium, persistent); L (maximal leakage conductance). The following somatic conductances were modeled: Na (maximal Sodium conductance); Ca_N (Calcium, N-like); L (maximal leakage conductance); K_Ca (Potassium, Ca²⁺-dependent); K_dr (Potassium, delayed rectified). All channel dynamics and transition rates, along with conductances for the model, were set according to published values (Kurian et al., 2011). tsDCS was modeled as an external electrical field, as previously described (Lafon et al., 2017). During direct current stimulation, the neuron will polarize along the direction of the electric field. This results in opposite changes in membrane potential at the dendritic and somatic sties. The numerical solutions of the model were computed using ode15s function of Matlab (Mathworks, Inc., Nattic, MA, United States).

The differences between two conditions (pre vs. post Nimodipine or a-tsDCS vs. c-tsDCS) within each group were assessed by parametric tests (t-test, MATLAB). We performed a linear regression. The non-parametric Kolmogorov–Smirnov test was used to assess differences in the distributions of two groups (K-S test, MATLAB). The significance level was set at 0.05. All data analyses were performed using MATLAB (The Math Works). In accordance with the Journal’s Data Availability Declaration, for quantitative presentation of data, all values are shown.

We first distinguished the effect of tsDCS during the period of stimulation (direct response) from unit firing that was changed after tsDCS ended (persistent response). A representative ECR EMG recording is shown (Figures 2A,B) and two units activated in response to c-tsDCS were isolated (Figure 2C). Motor unit activity is transformed into a continuous frequency plot (Figure 2D). C-tsDCS produces a robust increase in firing during stimulation for both units. Remarkably, the firing of one of the motor units (blue) is maintained for 210 seconds after stimulation ends. Whereas a-tsDCS (Figure 2E) also produced a direct effect, there was little or no maintenance of persistent unit firing (Figures 2F–H). Note, the delay in firing from stimulus onset of the red unit is due to the current ramp and the response threshold for the particular unit. In addition to showing modulatory effects on spontaneous single motor unit activity by tsDCS, we demonstrate that c-tsDCS produces a persistent response similar to the increase in motoneuronal firing observed during activation of a PIC. We also noticed for each of these units that c-tsDCS induced a stronger effect than a-tsDCS during testing (data not shown).

We recorded from a total of 20 motor units (ECR, 16; FCR, 4) during c-tsDCS and 22 (ECR, 17; FCR, 5) during a-tsDCS in five rats across multiple sessions with at least 7 days between any two sessions in the same animal. No differences were found between ECR and FCR responses; thus, we combined date from the two muscles. The percentage of units that had a direct response was slightly higher in c-tsDCS (100%) than a-tsDCS (86%). The cumulative distribution function (Figure 3A) plots the mean firing rate during the stimulation period. There was no significant difference between c-tsDCS and a-tsDCS distributions (K&S test), nor was there a difference in mean firing rate (Figure 3A, inset; t-test, p > 0.05).

Similar to the direct response, most units showed some persistent response with either polarity (90% in c-tsDCS vs. 77% in a-tsDCS). However, the cumulative distribution functions of persistent firing duration were significantly different for the two polarities (Figure 3B; K-S test, p < 0.05) and show a wide range of response durations. The mean duration of the persistent increase in motor unit firing was approximately six-times longer after c-tsDCS than a-tsDCS (Figure 3B, inset; t-test, p < 0.05). We computed a measure of the gain of the persistent response for each unit (Figure 3B; persistent response area divided by direct response area) and found this was also approximately 6 times greater for cathodal than anodal stimulation (t-test, p < 0.05). Our findings suggest that c-tsDCS activates a PIC to prolong the elevated motor unit activity beyond the stimulation period.

To better understand the underlying mechanism of the persistent increase in motor unit firing during tsDCS, and to distinguish the mechanisms underlying the direct from the persistent responses, we administered the L-type Ca²⁺ channel blocker Nimodipine (5 mg/kg, IV; tested 15 min post-injection). We determined if Ca²⁺ channel blockade changed the properties of the direct and persistent responses. For the two units shown, the direct responses to c-tsDCS and a-tsDCS were similar (Figures 4A,B), whereas the persistent response was only produced by c-tsDCS. Nimodipine completely blocked the persistent response in the motor unit, with a smaller effect on the direct response (Figure 4A). Although there was no persistent response produced by a-tsDCS in the unit shown (Figure 4B) there was a small reduction in firing during the direct response. Interestingly, there was a small increase in the direct response duration but insufficient to produce a persistent effect after stimulation ceased. Ensemble responses (Figures 4C,D; 20 motor units in c-tsDCS and 22 in a-tsDCS) show complete elimination of the persistent response after c-tsDCS with approximately a 50% decrease in the peak of the direct response after Nimodipine (Figure 4C). In contrast, Nimodipine had a minimal effect on the direct response to a-tsDCS and the small persistent response that was produced, was eliminated (Figure 4D).

To better reveal how variable the response to Ca²⁺ channel blockade is, we plot the cumulative distribution function for the change in the direct response firing rate (Figure 5A; pre-Nimodipine minus post-Nimodipine). The plot shows a consistent reduction with c-tsDCS (rightward shift) and a mixture of symmetrical changes around zero for a-tsDCS. The effect of blockade on the two polarities was significant (K-S test, p < 0.05; Figure 5A). The percentage of motor units showing a direct response was significantly decreased after Nimodipine for c-tsDCS (−50%). than for a-tsDCS (−11%). The mean reduction in firing rate of the direct response after blockade was also significantly different between c-tsDCS and a-tsDCS (t-test, p < 0.05; Figure 5A, inset). The effect of calcium channel blockade on the direct response was linearly correlated with the mean firing rate of the direct response (Figure 5B). The larger the direct response the stronger the blockade effect for both c-tsDCS (slope: −0.70, R-sq: 0.59; p < 0.05) and a-tsDCS (slope: −0.62, R-sq 0.20; p < 0.05).

After Nimodipine administration, the duration of the persistent response showed a consistent reduction for the c-tsDCS and, not surprisingly, a negligible change for a-tsDCS (Figure 5C). Most of the persistent responses of motor units were blocked by Nimodipine for both c-tsDCS (−67%) and a-tsDCS (−58%). Similar to the direct response, the blockade effect was linearly correlated with the strength of persistent response: the larger the persistent response the stronger the blockade effect for both c-tsDCS and a-tsDCS (Figure 5D). The slope of this relationship is negative one (R:1, p < 0.05), thus affirming the strong L-type calcium channel dependence. Not surprisingly, the blockade effect was stronger for c-tsDCS than for a-tsDCS; for both persistent response duration and persistent response gain (Figure 5C, insets). Our findings reveal a robust effect of Nimodipine on abrogating the persistent response produced by c-tsDCS suggesting that the persistent response is mediated by a Ca²⁺ PIC.

We used a computer simulation of a two-compartment neuron model (Booth et al., 1997; Kurian et al., 2011; Kim, 2017) to inform our finding that the persistent motor unit response is mediated by a Ca²⁺ PIC. Figure 6A shows a schematic representation of the neuron model as well as its coupling to the external field through the extracellular potential difference V_E. Voltage-dependent sodium channels and calcium channels were modeled in the dendrite only (Heckman et al., 2003; Murray et al., 2010; Marcantoni et al., 2020). The motor neuron time constant and conductance parameters used in our simulation were chosen from a published paper (Kurian et al., 2011; Mekhael et al., 2019). The strength of the external applied direct current stimulation (DCS) was modeled according to our published data (Song et al., 2015), and the effect of calcium channel blockade produced by Nimodipine was modeled by a partial Ca²⁺ conductance block. We used a 10% reduction (from 0.35 to 0.315); however, progressive Ca²⁺ conductance reduction did not systematically and linearly reduce the direct response.

The model replicated the key features of the experimental findings. Parameters set for the model account for the presence of spontaneous activity (black line before and after DCS). Modeling cathodal DCS produced both direct and persistent responses (Figure 6B), similar to our experimental findings using c-tsDCS (Figures 2D, 4C, pre-Nimodipine). Partial Ca²⁺ channel blockade reduced the direct response and eliminated the persistent response, which is similar to our experimental findings (Figures 2D, 4C). By contrast, modeling anodal DCS produced only a direct response and this was unaffected by partial Ca²⁺ channel blockade (Figure 6C). Additionally, the direct response modeled by anodal DCS is smaller than during cathodal DCS. Results of our simulation suggest that both c-tsDCS and a-tsDCS activate motor units during tsDCS and this direct response is a balance between activation of the dendrites and the soma, whereas the persistent response is only present during c-tsDCS and is modulated by dendritic calcium channels.

It is not understood if the excitatory actions of c-tsDCS reflect membrane depolarization of spinal cord neurons or if it indirectly affects spinal neurons by depolarizing the axons of peripheral somatic sensory afferents and other intraspinal axons and terminals (Bolzoni and Jankowska, 2015; Formento et al., 2018), or supraspinal processing through ascending synaptic effects (Ahmed, 2013b; Bocci et al., 2014, 2015). Some studies suggest that a-tsDCS may affect axonal conduction, while c-tsDCS modulates interneuronal spinal networks (Bocci et al., 2014). Locally-applied (intraspinal) DC fields, in the microampere range, can enhance the excitability of intraspinal afferent fiber terminals (Bolzoni and Jankowska, 2015) and epidural direct current stimulation produces similar effects (Jankowska et al., 2017). Whereas there are multiple targets that could contribute to persistent firing, the well-known role for motoneuronal PICs is likely to be a dominant factor. We found that both cathodal and anodal stimulation produced a direct facilitatory effect (albeit a weaker anodal than cathodal response), but only cathodal stimulation produced a persistent response. Perhaps the absence of a direct cathodal facilitation with intracellular motoneuron recordings (Baczyk et al., 2019, 2020a) reflects the deeply anesthetized preparation. It is known that pentobarbital anesthesia does not support PICs (Button et al., 2006); there may be other molecular targets of tsDCS that are affected by deep anesthesia. The discrepancy may also be due to misalignment of the polarization field and dendrite orientation or sampling the activity of large motoneurons (Ahmed, 2016). Note that the enhanced motor unit firing during a-tsDCS is not significantly reduced by nimodipine, suggesting a source for activation other than the L-type Ca²⁺ channel. It is likely that there are multiple neural targets, especially if one considers both the direct and persistent effects. Further experiments and more sophisticated modeling [i.e., multi-compartment (Elbasiouny and Mushahwar, 2007)]; are needed to dissect the specificity of tsDCS actions. However, based on pharmacological blockade we propose that the L-type Ca²⁺ is a key mediator of the PIC.

The effects of tsDCS, in both human and animal models, has been studied predominantly at the level of changes in motor cortex-evoked motor output [e.g., MEPs; also spinal LFPs; Ahmed (2011); Knikou et al. (2015), Song et al. (2015), and Song and Martin (2017)]. Changes in motoneuron excitability may reflect membrane polarization in response to the external applied electrical field, but also can be modulated through different ion channels (Booth et al., 1997; Rahman et al., 2013). Thus, changes in spontaneous activity of single motor units can provide insight into the molecular mechanism of tsDCS on motoneurons.

Similar to the facilitatory effect on motor cortex MEPs (Song et al., 2015), we also find facilitation of the spontaneous single motor unit response during c-tsDCS. However, in contrast to prior findings that a-tsDCS reduced motor cortex-evoked MEPs, the direct response during a-tsDCS increased, albeit less than for cathodal stimulation. This suggests differential actions of a-tsDCS on synaptically-evoked MEPs (Bolzoni and Jankowska, 2015) and on spontaneous single motor unit activity during the direct response (Bolzoni and Jankowska, 2015). Here, we show that tsDCS modulation of spontaneous activity is driven by intrinsic excitability mechanisms; in particular, the L-type calcium channel.

Polarization of hippocampal neurons in brain slices affects both spontaneous firing rate and synaptic efficacy (Lafon et al., 2017). We also found that tsDCS could modulate motoneuron PIC gain, with the c-tsDCS effect being substantially greater than that of a-tsDCS. The two-compartment model showed that the PIC arises predominantly from dendritic Ca²⁺ influx (Kurian et al., 2011). Although c-tsDCS could hyperpolarize the somatic membrane potential (i.e., DC-induced neuronal polarization), the strong inward current from the dendrite would be expected to depolarize the soma, causing a plateau potential. Our findings point to the principal mechanism for this augmentation as activation of the voltage-dependent calcium channel within the dendritic compartment to produce a PIC-like motoneuronal response.

We hoped to model the question of whether tsDCS field polarization, through actions on dendritic Ca²⁺ channels and PICs, can lead to persistent motor unit firing. Whereas tsDCS could act on spinal neurons through synaptic and network actions, persistent firing and PICs are largely an intrinsic membrane property and the action of motoneurons. This suggests that a two-compartment motoneuron model would be sufficient. The model accounted for the smaller anodal direct effect and the larger cathodal direct and persistent effects we observed experimentally. Further, it also accounted for L-type Ca channel blockade and the loss of the persistent effect, but not the direct effect. The two-compartment model did not capture the very long duration persistent responses after c-tsDCS. Although we were able to model the reduction in the c-tsDCS direct response and elimination of the persistent response to both polarities with a 10% reduction in Ca²⁺ conductance, systematically smaller reductions did not yield progressive response reductions. These findings stress non-linear dynamics underlying tsDCS neuromodulation (Elbasiouny and Mushahwar, 2007). What might contribute to these differences between experimental and modeling effects? The orientation of motoneuron dendritic arbors is heterogeneous. During c-tsDCS, some motoneurons would be expected to experience full cathodal stimulation, whereas others would simultaneously experience anodal stimulation because of differences in their somato-dendritic axes. The particular activity change of each motor unit during tsDCS could be the effect of a combination of c-tsDCS or a-tsDCS depending on its soma-dendritic orientation and susceptibility to polarization by the applied tsDCS. Although most dendrites of motoneurons are located dorsal to the soma (Balaskas et al., 2019), with a net orientation parallel to the applied tsDCS field, the soma-dendritic orientation relative to the external field varies (see Figure 6). A multicompartment motoneuron model showed that the non-linear properties of the voltage-gated Ca²⁺ channel could lead to suppression of PICs in both depolarized or hyperpolarized dendritic regions (Elbasiouny and Mushahwar, 2007). However, motoneuron dendrites were modeled with a radially symmetrical morphology around the soma, which may have led to different predictions and did not accord with our experimental results. Considering the single dendritic compartment in the model we developed, it suggests that two dominant factors leading to persistent firing after c-tsDCS are polarization of the dorsally- (or dorsomedially-) directed dendritic arbor and dendritic localization of the Ca channels.

Phasic activation of spinal motor circuits holds much promise to improve motor function in humans after spinal cord injury (SCI) (Angeli et al., 2014, 2018; Pena Pino et al., 2020), and the effects of a similar phasic stimulation showed a frequency and segmental-level dependence (Vogelstein et al., 2006). Phasic spinal stimulation is thought to activate CPGs, possibly through activation of large-diameter proprioceptive inputs (Formento et al., 2018). Moreover, non-invasive phasic stimulation methods are being developed to target the human spinal cord after injury (Inanici et al., 2021). The noninvasive application of tsDCS using surface electrodes is appropriate for behaving animals and humans. By developing an understanding of the molecular mechanisms of tsDCS, we can better inform therapeutic strategies of this method for promoting function after spinal injury. The specific effect of augmenting PIC-like responses with c-tsDCS is a novel molecular target. It must be stressed that PIC-like responses are produced during naturally-occurring motor actions in rats and humans (Kiehn and Eken, 1989; Gorassini et al., 2002a,b). PICs are regarded to be an important component of normal motor control and, together with a wide-range of channel types in motoneurons, offer extraordinary flexibility for muscle force control (Heckmann et al., 2005). Enhancing the PIC-like response of motor units induced by c-tsDCS would be well-suited to potentiate motor strength after SCI.

After complete sacral transection, the 5-HT2_C receptor caudal to injury can become constitutively active, resulting in unregulated PICs that contribute to hyperreflexia and spasms (Murray et al., 2010). This is thought to reflect the loss of descending monoaminergic regulation of motoneuronal excitability. Since hyperreflexia and spasticity are considered to reflect enhances spinal excitability, would c-tsDCS exacerbate these conditions in the injured spinal cord? Or through its targeted use to counter the loss of descending excitatory control signals, might it interrupt the circle of loss of excitability leading to maladaptive hyperreflexia that occurs after SCI, and especially after perinatal corticospinal system injury (Cavarsan et al., 2019; Steele et al., 2020)? The mechanism for PIC-like motor unit firing induced by c-tsDCS, L-type Ca²⁺ channel activation, has recently been targeted pharmacologically with Nimodipine to ameliorate spasticity in mice after complete sacral SCI (Marcantoni et al., 2020). This adds to the well-known amelioration of spasticity after rehabilitation (Cote et al., 2014; Beverungen et al., 2020). C-tsDCS neuromodulation enhances PICs, and presumably force capacity, which is necessary to improve motor capacity post-injury. By contrast, Nimodipine reduces PICs and presumably force capacity, thereby exacerbating weakness but ameliorating hyperreflexia. Intriguingly, these two interventions might be recruited in different combinations and different times after motor system injury—as spinal excitability changes evolve and hyperreflexia and spasms develop—to modify PIC production bidirectionally for different rehabilitation goals.