Facilitation of sensory transmission to motoneurons during cortical or sensory‐evoked primary afferent depolarization (PAD) in humans

Sensory and corticospinal tract (CST) pathways activate spinal GABAergic interneurons that have axoaxonic connections onto proprioceptive (Ia) afferents that cause long‐lasting depolarizations (termed primary afferent depolarization, PAD). In rodents, sensory‐evoked PAD is produced by GABAA receptors at nodes of Ranvier in Ia afferents, rather than at presynaptic terminals, and facilitates spike propagation to motoneurons by preventing branch‐point failures, rather than causing presynaptic inhibition. We examined in 40 human participants whether putative activation of Ia‐PAD by sensory or CST pathways can also facilitate Ia afferent activation of motoneurons via the H‐reflex. H‐reflexes in several leg muscles were facilitated by prior conditioning from low‐threshold proprioceptive, cutaneous or CST pathways, with a similar long‐lasting time course (∼200 ms) to phasic PAD measured in rodent Ia afferents. Long trains of cutaneous or proprioceptive afferent conditioning produced longer‐lasting facilitation of the H‐reflex for up to 2 min, consistent with tonic PAD in rodent Ia afferents mediated by nodal α5‐GABAA receptors for similar stimulation trains. Facilitation of H‐reflexes by this conditioning was likely not mediated by direct facilitation of the motoneurons because isolated stimulation of sensory or CST pathways did not alone facilitate the tonic firing rate of motor units. Furthermore, cutaneous conditioning increased the firing probability of single motor units (motoneurons) during the H‐reflex without increasing their firing rate at this time, indicating that the underlying excitatory postsynaptic potential was more probable, but not larger. These results are consistent with sensory and CST pathways activating nodal GABAA receptors that reduce intermittent failure of action potentials propagating into Ia afferent branches.

in turn synapse onto specialized GABAergic interneurons (termed GABA axo neurons; GAD2+) with axoaxonic contacts onto other afferents, forming the classic trisynaptic pathway (Alvarez, 1998;Jankowska, McCrea, Rudomin et al., 1981;Lalonde & Bui, 2021;Lin et al., 2022;Zimmerman et al., 2019). The activation of GABA A receptors on these sensory axons produces a local depolarization of the afferent due to an outward flow of chloride ions (Gallagher et al., 1978, also reviewed in Rudomin & Schmidt, 1999;Willis, 1999). Although paradoxically excitatory, this GABA A receptor-mediated depolarization (referred to as primary afferent depolarization, PAD) was previously thought to shunt or inactivate action potentials invading the afferent terminal, thereby inhibiting neurotransmitter release to produce presynaptic inhibition (Willis, 2006). This inhibitory role was postulated because the PAD evoked by a flexor afferent followed a similar time course to the suppression of Ia-mediated excitatory postsynaptic potentials (EPSP) in an extensor motoneuron when conditioned by the same flexor afferent (Eccles et al., 1961;Frank & Fuortes, 1957;Willis, 1999). Because there appeared to be no direct effects of the conditioning flexor nerve stimulation on the extensor motoneuron, a presynaptic inhibitory mechanism of PAD on afferent transmission was assumed and this idea has prevailed over the past 60 years (Willis, 2006;Zimmerman et al., 2019). However, recent studies reveal that GABA A receptors are generally not found on Ia afferent terminals (Alvarez et al., 1996;Hari et al., 2022;Lucas-Osma et al., 2018) and do not appreciably depolarize the Ia afferent terminals during PAD Lucas-Osma et al., 2018), conflicting with the concept of GABA A -mediated presynaptic inhibition. Instead, GABA B receptors are found at afferent terminals  and likely mediate presynaptic inhibition of sensory transmission to motoneurons (Curtis & Lacey, 1998;Fink, 2013). In contrast, GABA A receptors are found mostly at or within 100 μm of the nodes of Ranvier (nodes) in the many large, myelinated branches of Ia afferents throughout the dorsal and ventral spinal cord . Accordingly, a novel role for GABA in facilitating, rather than inhibiting, afferent conduction has been proposed where the activation of GABA A receptors in or near the afferent nodes produces 0 Krista Metz received her PhD in Neuroscience from the University of Alberta, Edmonton, in 2022. She received her Bachelor's degree in Kinesiology in 2016 from the University of Regina, Saskatchewan. Her PhD work focused on sensorimotor circuits in the human spinal cord and how changes to these circuits after injury to the central nervous system could be involved in sensorimotor impairment and spasticity. Currently, Krista works as a neurophysiologist for the Saskatchewan Health Authority at the Royal University Hospital in Saskatoon, Saskatchewan, and continues research related to sensorimotor systems in the human spinal cord.
a local depolarization that facilitates nodal sodium channels to secure action potential propagation and decreases downstream branch-point failure (termed nodal facilitation). Interestingly, axon depolarization from any source, including PAD, can have a bipolar action: moderate subthreshold depolarizations near rest (−70 mV) promote axon conduction, while large suprathreshold depolarizations of the axon above −50 mV cause spike failure (Lüscher et al., 1994), but the latter rarely occurs because PAD is not large enough to cause a depolarization block of the Na + spike (see Extended data in Fig. 3A-C in Hari et al., 2022). In rodents, these GABA A receptors on the dorsal Ia afferents are readily activated by cutaneous or pain afferent pathways and produce nodal facilitation, rather than presynaptic inhibition, of the Ia-mediated monosynaptic reflex Lucas-Osma et al., 2018). Our preliminary data from the soleus muscle suggest that this may also occur in humans following cutaneous afferent conditioning to evoke PAD . In this paper, we expand these studies to provide further evidence consistent with nodal facilitation in human Ia afferents by examining additional muscle groups with various forms of conditioning stimuli to induce PAD. The conditioning included stimulation of proprioceptive and cutaneous afferents and descending corticospinal tract (CST) pathways, the latter which directly synapse onto GABA axo neurons (Ueno et al., 2018) and thus should produce a PAD as previously shown for afferents in the dorsal horn following trains of motor cortex stimulation (Andersen et al., 1964;Carpenter et al., 1963;Liang et al., 2022;Seki et al., 2003). As well, we analysed single motor unit recordings to estimate the reduction in the probability of intermittent failures of motoneuron EPSPs with conditioning, to further support the idea that long-lasting conditioning-evoked PAD decreased the intermittent failures in sensory axon transmission to motoneurons. Previous studies in humans have shown that cutaneous and CST stimulation facilitates proprioceptive sensory transmission to motoneurons, as assessed by H-reflexes (Aimonetti et al., 2000;Berardelli et al., 1987;Iles, 1996;Iles & Roberts, 1987;Meunier & Pierrot-Deseilligny, 1998;Nakashima et al., 1990). At the time, the explanation for this facilitation was that cutaneous and CST activity somehow inhibits the tonic, spontaneous activation of K. Metz and others J Physiol 601.10 GABA axo neurons and their associated PAD that were thought to cause presynaptic inhibition of the H-reflex. However, numerous animal studies have shown that cutaneous and CST activity does not usually by itself reduce spontaneously active PAD, as assessed directly (Lund et al., 1965;Lundberg, 1964) or with the Wall excitability test method (Rudomin et al., 1983; although see Mendell & Wall, 1964), and often increases PAD (Andersen et al., 1964;Hari et al., 2022;Liang et al., 2022;Lucas-Osma et al., 2018;Lund et al., 1965;Seki et al., 2003), making this explanation unlikely. The brief phasic PAD and associated GABA axo neuron activity induced by proprioceptive stimulation is reduced by prior cutaneous or supraspinal activity (Lund et al., 1965;Lundberg, 1964;Rudomin et al., 1983), but this is not relevant when the GABA axo neurons are inactive at rest, as appears to be the case, based on the lack of reduction in ongoing spontaneous PAD with cutaneous stimulation alone. Likewise, the previous human studies cited above have shown that cutaneous and CST-driven pathways reduce the phasic inhibition of the H-reflex produced by a conditioning proprioceptive stimulation, but it is unclear how this complex disinhibition is relevant to the resting state where these GABA axo neurons are likely silent. Thus, both the animal and human work suggest that GABA axo neurons evoke a phasic PAD following proprioceptive stimulation that can somehow be reduced by cutaneous and CST stimulation. However, these neurons appear to be silent without conditioning, leaving uncertain the mechanism for how cutaneous or CST stimulation directly facilitates the H-reflex. Consistent with observations that the GABA axo neurons that evoke PAD from proprioceptive and cutaneous afferents are two spatially separated populations (Jankowska, McCrea, Rudomin et al., 1981;Rudomin & Schmidt, 1999;Zimmerman et al., 2019), with the latter more dorsal, it seems likely that the PAD evoked by these two pathways can act independently on sensory transmission to motoneurons.
Considering that cutaneous conditioning stimulation generally increases, rather than decreases, GABA axo activity and PAD in proprioceptive afferents Lucas-Osma et al., 2018), we propose a simpler mechanism to explain how cutaneous and CST pathways facilitate sensory transmission to motoneurons as reflected in the H-reflex. That is, we suggest that cutaneous or CST stimulation directly or indirectly activates dorsally located GABA axo neurons that release GABA onto proprioceptive afferents to increase their conduction as they traverse through the dorsal horn toward motoneurons. As detailed above, nodal GABA receptors on proprioceptive afferents facilitate conduction by depolarizing nodes to prevent action potential failure that is otherwise common at branch points without GABA. The GABA axo neurons activated by cutaneous or CST inputs are different from those activated by proprioceptive inputs, and so these neurons might well inhibit one another during combined activation of these two sensory systems, as conceived by Rudomin's group (Fig. 11A;Rudomin et al., 1986), but overall they both depolarize proprioceptive sensory axons and increase axon conduction. This simpler mechanism would not only explain the older data that show cutaneous and CST conditioning causes a reduced inhibition of the H-reflex (disinhibition) but also accounts for the direct facilitation of the Ia afferents mediating the H-reflex by these conditioning pathways (Meunier & Pierrot-Deseilligny, 1998;Nakashima et al., 1990;Roby-Brami & Bussel, 1990). Here we provide further evidence to support this view.
There are two types of PAD that have been measured in proprioceptive and cutaneous afferents. The first is a short-duration (phasic) depolarization lasting for 100-200 ms that is evoked from a brief stimulation train (1-3 pulses at 200 Hz) of low-threshold proprioceptive, cutaneous or pain afferents (Eccles, Kostyuk et al., 1962;Lucas-Osma et al., 2018;Rudomin & Schmidt, 1999;Willis, 2006). This phasic PAD is mediated by synaptic GABA A receptors with α1, α2 and γ 2 subunits that are located adjacent to sodium channels in afferent nodes . The second type of PAD is a longer-duration (tonic) depolarization that lasts for tens of seconds (Delgado-Lezama et al., 2013;Lomeli et al., 1998;Wall, 1998) and is activated by longer trains of stimulation, most effectively from a fast stimulation train (0.5 s at 200 Hz) or from a relatively slower but longer frequency train (20 s at 0.2 to 2 Hz) (Lucas-Osma et al., 2018). This tonic PAD is mediated by the activation of extra-synaptic GABA A receptors with α5 subunits (α5 GABA A ) that are also located near nodal Na + channels and specifically reduced by the α5 GABA A receptor blocker L655708 with inverse agonist properties Lucas-Osma et al., 2018).
Importantly, the characteristic time course of phasic PAD (100-200 ms), and the reduction in branch-point failure it produces in the Ia afferent, is reflected in a similar time course of facilitation of the motoneuron EPSP following a conditioning stimulation that evokes PAD . Thus, in the present study, we first examined in human participants whether the facilitation of Ia-mediated H-reflexes from such conditioning had a similar time course to the phasic PAD measured in rodent Ia afferents. To do this we measured the facilitation of the resting H-reflex (as a surrogate measure of Ia afferent transmission to the motoneuron) produced by a brief, peripheral sensory or CST conditioning stimulation at interstimulus intervals (ISIs) between 0 and 200 ms. Second, to determine whether we could induce long-lasting increases in H-reflexes consistent with tonic PAD in the Ia afferents, we used long trains of cutaneous stimulation (0.5-10 s). Tonic PAD from repeated cutaneous stimulation is thought to result from GABA spillover in the dorsal horn and subsequent long-lasting activation of extra-synaptic α5 GABA A receptors on the Ia afferent (Lucas-Osma et al., 2018). Since Lucas-Osma et al., 2018 found that faster cutaneous stimulation trains were more effective in inducing tonic PAD, we examined whether fast, compared with slower, trains produced more long-lasting H-reflex facilitation.
In all experiments, special attention was given to ensure there was no direct postsynaptic facilitation on the motoneuron produced by the sensory or CST conditioning stimulation to support the idea that any facilitation of H-reflexes was likely presynaptic in origin, somewhere on the sensory axon. To rule out direct effects on the motoneuron pool, we examined whether the conditioning stimulation itself modulated the firing rate of tonically firing motor units in the test muscle, because the generation of action potentials in a motoneuron is very sensitive to small changes in synaptic inputs, even at distal dendrites (Frank, 1959), making firing rate a sensitive measure of postsynaptic effects on the motoneuron (Powers & Binder, 2001).
Lastly, we measured changes in the firing rate and discharge probability of single motor units (motoneurons) activated during the H-reflex to estimate the underlying Ia-EPSP by using the peristimulus frequencygram (PSF) method (Türker & Powers, 2005). This allowed us to examine whether spontaneous failures in Ia-EPSP activation, consistent with failures in Ia axon branch-point conduction, decreased with cutaneous conditioning known to evoke long-lasting tonic PAD in Ia afferents (Lucas-Osma et al., 2018). Remarkably, with the PSF method we found that increases in H-reflex amplitude from cutaneous conditioning were not associated with an increase in the underlying estimated unitary Ia-EPSP size, ruling out changes in presynaptic neurotransmitter release and/or postsynaptic effects, where both should produce graded changes in the size of the EPSP. Instead, the all-or-none probability of motor unit firing during the estimated Ia-EPSP window increased, suggesting that the underlying EPSP was more probable, but not larger. This is consistent with results from the rat where cutaneous-evoked tonic PAD increased the probability of Ia afferent conduction through branch points and the resultant unitary EPSPs . Taken together, these human experiments show that sensory and CST pathways facilitate H-reflexes with a time course similar to phasic and tonic PAD in the rodent, the latter which overcomes intermittent branch-point failure in the Ia afferent by nodal facilitation. Parts of the data from the long-lasting, cutaneous H-reflex facilitation experiments were adapted from .

Ethical approval
Experiments were approved from the Human Research Ethics Board at the University of Alberta (Pro 00 078057), were performed with informed consent, and adhered to the Declaration of Helsinki. Our sample comprised 40 participants (22 male) with no known neurological injury or disease, ranging in age from 18 to 57 years (27.8 (10.0), mean (standard deviation)).

Experimental setup
Participants were seated in a reclined, supine position on a padded table. The right leg was bent slightly to access the popliteal fossa and padded supports were added to facilitate complete relaxation of all leg muscles during H-reflex testing because CST activation could potentially activate spinal GABA axo circuits and interfere with the activation of PAD by the conditioning inputs (Ueno et al., 2018). For the transcranial magnetic stimulation (TMS) experiments, participants sat in a padded chair with the right leg slightly extended to 100°at both the knee and ankle joint and the foot was strapped to a supporting platform. The upper leg was also supported with straps and padding as above. The head was supported by a headrest to allow minimal movement during TMS. During H-reflex recordings, participants were asked to rest completely with no talking, hand or arm movements.

Surface EMG recordings
To measure M-wave and H-reflexes, a pair of Ag-AgCl electrodes (Kendall, Chicopee, MA, USA, 3.2 cm by 2.2 cm) was used to record surface EMG from the soleus, tibialis anterior (TA), abductor hallucis (AbHal), biceps femoris, medial gastrocnemius and vastus lateralis (quad) muscles with a ground electrode placed just below the knee. The EMG signals were amplified by 200-1000 and band-pass filtered from 10 to 1000 Hz (Octopus, Bortec Technologies; Calgary, AB, Canada) and digitized at 5000 Hz using Axoscope 10 hardware and software (Digidata 1400 Series, Axon Instruments, Union City, CA). To examine whether the conditioning inputs had any direct effects on the motoneuron pool, the surface electrodes were also used to record single motor unit activity in the soleus and AbHal muscles by placing them on the border of the muscle as per Matthews (1996).
Motor unit activity from the soleus muscle was also recorded at higher levels of contraction using a high-density surface EMG (HDsEMG) electrode (OT Bioelettronica, Torino, Italy, semi-disposable adhesive matrix, 64 electrodes, 5 × 13, 8 mm inter-electrode J Physiol 601.10 distance) with differential and ground electrodes wrapped above the ankle. Signals were amplified (150 times), filtered (10-900 Hz) and digitized (16 bit at 5120 Hz) using the Quattrocento Bioelectrical signal amplifier and OTBioLab+ v.1.2.3.0 software (OT Bioelettronica, Torino, Italy). The EMG signal was decomposed into single motor units using a convolutive blind source separation algorithm implemented in MatLab R2020b with additional quality assessment and accuracy improvement of the automatically decomposed motor unit action potential spike (pulse) trains as described previously (Afsharipour et al., 2020;Martinez-Valdes et al., 2017;Negro et al., 2016). Single motor units that were measured from standard surface and intramuscular EMG were discriminated visually.

Nerve stimulation to evoke homonymous and heteronymous reflexes
The tibial nerve (TN) was stimulated using a constant current stimulator (1 ms rectangular pulse width, Digitimer DS7A, Hertfordshire, UK) to evoke a homonymous H-reflex in the soleus and AbHal muscles. After searching for the TN with a surface probe, an Ag-AgCl cathode electrode (Kendall, Chicopee, MA, USA, 2.2 cm by 2.2 cm) was placed in the popliteal fossa, with the anode electrode (Axelgaard, Fallbrook, CA, USA, 5 cm by 10 cm) placed on the patella. If an AbHal H-reflex was not readily evoked from TN stimulation behind the knee, the posterior TN was stimulated below the medial malleolus. The stimulation intensity for the homonymous soleus and AbHal H-reflexes was set to evoke a test (unconditioned) H-reflex below half maximum on the ascending phase of the H-reflex recruitment curve (∼30% of the maximum H-reflex) to reduce the possibility of evoking polysynaptic reflexes so that we could better attribute increases in H-reflexes to an increase in Ia transmission to the motoneurons . A small heteronymous reflex (260.00 (65.192) μV peak-to-peak, n = 5 participants) was also evoked in the biceps femoris, which needed a weak contraction (∼5% maximum voluntary contraction, MVC) to be evoked, by stimulating the TN to the medial gastrocnemius (MG) in the popliteal fossa (1 ms pulse width, ∼3.3 × M-wave threshold measured in MG muscle). The heteronymous biceps femoris reflex had an onset time that was consistent with a monosynaptic response because it had a mean latency of 29.98 (2.28) ms, which was on average 4.51 (0.613) ms earlier than the more distal homonymous H-reflex evoked in the MG muscle, the latter with a mean latency of 34.38 (1.89) ms. Assuming the conduction velocity of the motor axons in the TN is 60 m/s (or 6 cm/ms) and the distance between the biceps femoris to the MG EMG electrodes was ∼24 cm, the biceps femoris reflex should, in theory, arrive 4 ms earlier than the H-reflex measured in the more distal MG muscle (24 cm/6 cm/ms = 4 ms), which is close to the measured latency difference of ∼4.5 ms between the two reflexes. Reflexes were evoked every 5 s to minimize post-activation depression of the Ia afferents (Hultborn et al., 1996). At least 20 test reflexes were evoked before conditioning to establish a steady baseline since the Ia afferent activation itself could also activate spinal GABA axo networks and facilitate the reflexes (i.e. self-priming, Hari et al., 2022). All reflexes were recorded at rest except for the heteronymous biceps reflex and during the firing probability experiments described below.

Cutaneous and proprioceptive conditioning stimulation.
To condition the H-reflex by mainly cutaneous afferents, the medial (cutaneous) branch of the superficial fibular nerve (SFN, also known as the superficial peroneal nerve) was stimulated on the dorsal surface of the ankle using a bipolar arrangement (Ag-AgCl electrodes, Kendall, Chicopee, MA, USA, 2.2 cm by 2.2 cm). A short train (three pulses, 200 Hz for 10 ms) of SFN stimulation was applied at intensities corresponding to the perception threshold but below the radiating threshold (3.0-7.4 mA) to avoid direct activation of motoneurons as assessed below. A train of pulses was used for the cutaneous nerve stimulation because it evokes larger phasic PAD in Ia afferents compared with single-pulse stimulation . Approximately 20 baseline soleus H-reflexes were elicited to ensure the H-reflex was stable, followed by seven conditioned H-reflexes at one of the ISIs (0,30,60,80,100,150 or 200 ms). Following this, seven unconditioned H-reflexes were evoked to re-establish baseline and another run of seven conditioned H-reflexes was applied at another randomly chosen ISI. This was repeated until all ISIs were applied and the order of ISIs chosen was randomly applied across different participants. To condition the H-reflex by mainly low-threshold proprioceptive afferents, the common peroneal nerve (CPN, also known as common fibular nerve, but CPN used here for historical purposes) was stimulated in a bipolar arrangement just below the head of the fibula (three pulses at 200 Hz) at 1.0× motor threshold measured in the TA muscle, being careful to elicit a pure dorsiflexion response at higher intensities.
CST conditioning stimulation. The CST to the soleus or AbHal motoneuron pool was activated by applying TMS to the contralateral motor cortex using a custom-made figure-of-eight batwing coil (P/N 15,857; 90 mm diameter; Nielsen & Petersen, 1994) that was connected to a Magstim 200 stimulator (Magstim, Dyfed, UK). The coil was typically positioned 2 cm lateral to vertex to target the lower leg muscles. The active motor threshold (AMT) was determined by the lowest-intensity, single TMS pulse that produced a discernible and reproducible motor-evoked potential in the tested muscle while the participant held a small voluntary contraction of ∼5% MVC. The TMS intensity was set to 0.9× AMT in the resting muscle to avoid direct activation of the motoneurons. H-reflexes were conditioned by a single-pulse TMS at the same ISIs as for the sensory conditioning experiment above, in addition to the 250 and 300 ms ISIs given the longer duration of facilitation in some participants.
Data analysis. For both the sensory and CST conditioning experiments, the unrectified, peak-to-peak amplitude of the seven test H-reflexes immediately preceding the seven conditioned H-reflexes for a given ISI were averaged together because test H-reflexes could grow over time (self-priming, Hari et al., 2022). The effect of the conditioning stimulation on the test H-reflex was measured using the formula: percentage change H-reflex = ([(conditioned H -test H)/test H] × 100%). Data were also analysed by averaging the amplitude of all test H-reflexes in a trial and this provided similar results throughout so only calculations of percentage change using the immediately preceding test H-reflexes are reported here. The mean percentage change H-reflex for each ISI was averaged across participants. Because the profile of H-reflex facilitation could be variable between participants, the maximum or peak percentage change H-reflex, irrespective of ISI, was also averaged across participants.
Postsynaptic effects of conditioning stimuli. To determine whether there were any direct effects on the motoneurons from the conditioning stimulation at the time the H-reflexes were evoked, we measured whether the sensory or CST stimulation applied alone produced any changes in the tonic firing rate of single motor units or changes in the amplitude of the rectified surface EMG. Single motor units were activated in the soleus or AbHal muscle while the participant held a small voluntary contraction at around 5% of maximum. Both auditory and visual feedback were used to keep the firing rates of the units steady while the conditioning cutaneous or CST stimulation was applied every 3-5 s. The firing rate profiles from many stimulation trials (∼150) were superimposed and time-locked to the onset of the conditioning stimulation to produce a PSF as done previously (Norton et al., 2008;Türker & Powers, 2005). A mean PSF for each participant was produced by averaging the frequency values into 20 ms bins, ensuring there were ∼15 frequency values on average per bin.
The mean rate in each 20 ms bin was compared with the mean rate measured in a 100 ms window before the conditioning stimulation and expressed as: percentage change PSF = ([(mean bin rate -mean pre-stimulus rate)/mean pre-stimulus rate] × 100%). A similar binning and percentage change calculation was performed for the rectified surface EMG using a 100 ms window before the conditioning stimulation to measure the pre-stimulation background EMG. The mean PSF or EMG profile for each participant was then averaged across participants in each experiment.

Motoneuron and Ia afferent excitability during
conditioned H-reflex. We wanted to estimate both the excitability of the motoneuron (as reflected in the PSF and rectified EMG), and the predicted amplitude of PAD in the Ia afferent, at the time the conditioned Ia afferents activated the motoneurons at the spinal cord to ensure any increase in H-reflex size was not due to an increase in motoneuron excitability from the conditioning stimulation but would be explained better by an increase in transmission of Ia inputs to the motoneuron (Fig. 1). First to align motoneuron excitability, the conditioned H-reflex values, as a function of the various ISIs (see bottom graph, Fig. 1), were shifted to the right of the onset of the conditioning (SFN) stimulation in the PSF (light blue trace) and EMG profiles (not shown) by an amount equivalent to the average latency of the H-reflex (∼30 ms, solid black arrows in Fig. 1). The rightward shift accounted for the time it takes the Ia afferent volley to reach the spinal cord plus the time it takes the motoneuron response to reach the muscle where the H-reflex is measured (grey arrows). Thus, the PSF (or EMG) values occurring near the shifted conditioned H-reflexes can be used as an estimate of the motoneuron excitability at the time the conditioned H-reflexes were activated at the spinal cord. Second, plotting the estimated time course of PAD measured in a rodent Ia afferent after the conditioning stimulation helps to predict when the Ia afferents mediating the H-reflexes should be facilitated by the conditioning PAD. Because the conditioned-evoked PAD in the Ia afferent is activated at the cord with a similar delay as the Ia-EPSP mediating the H-reflex (top two traces in Fig. 1, see Hari et al., 2022), a 0 ms ISI corresponds to just before the conditioning afferent stimulation can influence the H-reflex. In contrast, at the 80 ms ISI while PAD is predicted to be activated in the Ia afferent, the H-reflex can be facilitated ( Fig. 1; compare pink and blue dots representing H-reflexes in inset at 0 and 80 ms ISI, respectively). Moreover, if the mean firing rate of the motor units (PSF, light blue line) near the 80 ms ISI is at or below the pre-conditioned level (below red line at red double arrow), this would indicate that the facilitation of the conditioned H-reflex at this ISI was not J Physiol 601.10

Figure 1. Time course of predicted phasic PAD and its relationship to H-reflex facilitation
Top three traces: Trace 1, theoretical example of phasic primary afferent depolarization (PAD) recorded in Ia afferent (top dark blue trace, profile taken from a rat Ia afferent recording). Trace 2, EPSP recorded in spinal motoneuron and H-reflex recorded in muscle (pink traces) from simultaneous SPN and TN stimulation to produce a conditioned H-reflex as measured for the 0 ms ISI (pink dot, bottom trace). Time of stimulation marked by rectangular pulses on left. PAD profile taken from intracellular recording in rat afferent (sacral S3) in response to stimulation of an adjacent dorsal root (S4: 1.1 × threshold, 0.1 ms pulse, Hari et al., 2022). Trace 3, EPSP (upper) and H-reflex (lower dark blue traces) in response to TN stimulation applied 80 ms after the SPN stimulation (ISI of 80 ms, dashed arrow). Bottom three traces: Trace 4, example mean PSF (light blue trace) to represent profile of motoneuron excitability in response to conditioning SPN stimulation alone. Red trace represents mean firing rate before SPN stimulation was applied. Double red arrow marks the excitability of the motoneuron (membrane potential estimated from the PSF) during activation of the H-reflex. Trace 5, violet trace is the PAD profile shifted to the right by the conduction time from the spinal cord to the muscle to visually line up the excitability of the Ia afferents at the time the motoneuron is being activated during the H-reflex and how this affects the resulting amplitude of the H-reflex recorded at the muscle. Trace 6, percentage change in H-reflex amplitude ((conditioned H -test H)/test H × 100%) at the various ISIs, highlighting the 0 and 80 ms ISIs. H-reflex values shifted to the right of the onset of the conditioning stimulation by an amount equal to the H-reflex latency (solid black arrows) plus the ISI interval (e.g. dashed line for 80 ms ISI). [Colour figure can be viewed at wileyonlinelibrary.com] mediated by a depolarization of the motoneurons from the conditioning stimulation.

Cutaneous and proprioceptive facilitation of H-reflexes during tonic PAD
(a) Slow (0.2 Hz) vs. moderate (2 Hz) cutaneous frequency trains: a slow (0.2 Hz) and moderate (2 Hz) frequency train of SFN stimulation was applied as per animal studies (Lucas-Osma et al., 2018). A stimulation intensity at two times perception threshold (10.6 (1.6) mA, n = 16 participants) was used to induce (theoretically) GABA spillover. Following a baseline of 20 test H-reflexes (delivered every 5 s), the slow or moderate SFN stimulation train was applied along with the test H-reflexes for 10 s, with each SFN stimulation occurring 500 ms before any H-reflex. Test H-reflexes were evoked for another 120 s to examine after-effects from the SFN stimulation trains. This stimulation protocol was repeated three times for both the 0.2 and 2 Hz frequencies in each participant, waiting at least 1 min at rest between each trial. (b) Fast (200 Hz) cutaneous frequency train: a faster (200 Hz) but shorter (500 ms) train of SFN stimulation was also used to condition the H-reflex as per Hari et al. (2022), using a protocol similar to the slower 10 s trains. Here, the SFN train was applied 700 ms before the test H-reflex and following this, H-reflexes continued to be evoked for another 90 to 120 s. A very low intensity of stimulation below radiating threshold (3.0-4.5 mA, n = 15 participants) was used to ensure that the high-frequency stimulation was not painful. (c) Slow (0.2 Hz) proprioceptive frequency train: because repetitive activation of Ia afferents also produces a build-up in tonic PAD and associated monosynaptic reflexes in rodents , we examined whether there was a similar build-up of test H-reflexes from repetitive stimulation of the TN afferents alone. In rodents, Ia afferent collaterals activate PAD networks that connect back to the same afferent to produce self-facilitation, which is revealed during low-intensity, repetitive stimulation of Ia afferents (1.1× afferent threshold, 0.

Firing probability of single motor units during the H-reflex with cutaneous conditioning
The firing probability of single motor units during the H-reflex window (approximately 30 to 45 ms post-TN stimulation) was measured with and without SFN conditioning in 13 participants. In 12 participants, single motor units were identified from HDsEMG and in one participant with intramuscular EMG to verify the HDsEMG recordings. The size of the unconditioned test H-reflex was set to just above the motor threshold (average amplitude of 252.7 (141.4) mV peak-to-peak (range 33.7-545.7 mV) with an average stimulation intensity of 5.2 (3.9)% of Mmax (0.66%-15.16%) during a small plantarflexion so that single motor units at the time of the H-reflex could be distinguished from the compound H-reflex potential (Nielsen et al., 2019;Yavuz et al., 2015). Importantly, the test H-reflex stimulation was set low enough that the probability of the motor unit participating in the reflex was low (∼30%), allowing us to examine intermittent failures in the EPSPs underlying the motor unit responses, as detailed below. In a single trial run, test H-reflexes were evoked every 3−5 s for the first 100 s followed by SFN-conditioned H-reflexes for the next 100 s using a 200 Hz, 50 ms pulse train (3.0 to 4.5 mA, below radiating threshold) applied 500 ms before each H-reflex. Approximately 40−50 usable test and conditioned firing rate profiles (PSFs) were produced for a single trial run where the motor units had a steady discharge rate 400 ms before and 600 ms after the SFN stimulation. Trial runs were repeated 3−6 times to obtain a sufficient number of frequency profiles (∼200) to J Physiol 601.10 construct a PSF with an average of ∼30 frequency points in each 20 ms bin (Norton et al., 2008).

Motor unit analysis.
For each test or conditioned PSF, the probability that a motor unit discharged during the entire ∼15 ms H-reflex window, or during the first 0.5 ms at the start of the rise in the PSF, was measured using the formula: (number of discharges during the H-reflex window/total number of sweeps) × 100%. The mean background firing rate 100 ms before the TN stimulation with and without conditioning was also measured. For both the test and conditioned trials, the average firing probability during the H-reflex window and the mean background rate were measured for each participant and then averaged across participants. The mean firing rate during the H-reflex window was also measured as an estimate of EPSP size (Norton et al., 2008), and also expressed as a percentage change between the test and conditioned H-reflex trials ((conditioned rate-test rate/test rate) × 100). As previously observed , the probability of the motor unit firing during the H-reflex time window can increase while the Ia-EPSP (PSF) remains unchanged with conditioning stimulation. This is likely explained by conditioning reducing intermittent failures in Ia afferent conduction and associated failures in unitary EPSPs on the motoneuron , as considered in the following analysis. That is, if the net Ia-EPSP in a motoneuron activated during the H-reflex is made up of one or more large unitary EPSPs that spontaneously fail in an all-or-nothing manner (e.g. due to axon conduction failure), then the motoneuron will only reach threshold and fire a spike when some of these unitary EPSPs are active during the H-reflex window. This spike will result in an increase in firing rate of the motor unit (PSF) above the ongoing baseline firing rate (i.e. pre-Ia-EPSP) that is mostly independent of the probability of these EPSPs occurring. To understand this, first consider the simplest case where there is only one unitary EPSP (Ia-EPSP) dominating the composite EPSP response of a motoneuron (motor unit) during the H-reflex window, which is the mostly likely case since we used near threshold test TN stimuli to recruit motor units with low firing probability. Here, the magnitude of the increased firing rate of this single motor unit (PSF) during this unitary EPSP is unchanged when the probability of the unitary EPSP occurring changes since the increase in rate only occurs when the unitary Ia-EPSP occurs in the H-reflex window. Thus, the constructed PSF from many trials is unchanged when the probability of the unitary EPSP increases with cutaneous conditioning, though more trials have a motor unit spike. Likewise, when there are only a few unitary EPSPs dominating the H-reflex, then the increased firing rate (PSF) when these unitary EPSPs are active does not depend on the probability of the unitary Ia-EPSP events if these events do not occur together in the H-reflex window, which is the most likely situation since these unitary EPSPs occur infrequently near threshold. Occasionally, these events occur together though, and this does lead to a small increase in the net EPSP (and associated PSF), as detailed below, though this has a negligible impact on the overall average net EPSP from many stimulation trials when we focus on low probability unitary EPSPs near threshold, as we have done when constructing the PSF. For example, consider two unitary EPSPs of similar amplitude A and probability of 10% each (P = 0.1, with a 20% chance of either EPSP occurring, similar to in Results). Out of every 100 H-reflex trials these two unitary EPSPs will contribute 20 EPSPs over 19 trials (2•100•p -100•p2), since in one of the trials they will occur together (100•p2 = 100•0.01 = 1); so the overall average net EPSP that drives the PSF has an amplitude of (20/19) where A = 1.053 A, or more generally the PSF is proportional to (2p/(2p -p2) A = (2/(2-p))A. If the probability of each unitary EPSP is increased by 5% (from 10% to 15%) with conditioning, which would increase the probability of the motor unit firing by 10% (5% + 5%; triggered from either unitary EPSP; as we see in Fig. 8B), then the amplitude of the average net EPSP that drives the PSF will increase to A (2/(2-0.15)) = 1.08 A, which is only a 2.7% increase. This would increase the PSF by 2.7% while the motor unit firing probability increases by 10%. Thus, in general when cutaneous conditioning increases the probability of these near threshold unitary EPSPs, then the motor unit likewise increases its probability of discharging during the Ia-EPSP by this amount, but the average firing rate of the motor unit during the Ia-EPSP/H-reflex (PSF) is relatively unchanged if the size of the all-or-nothing unitary EPSPs do not change much (indicating a presynaptic effect), as has been observed in rats .
Finally, we consider the factors that lead to the motor unit firing during the H-reflex window, and which of these matter to the analysis considered above. First, the motor unit may simply fire by chance unrelated to the weak TN stimulation during the 15 ms Ia-EPSP/H-reflex window, but this probability is very low, since the units are only firing on average at 7 Hz with an interspike interval of about 140 ms (∼10.7% chance; see Results) and a 0.035% random probability during the period of monosynaptic activation of the motoneuron in the first 0.5 ms of the Ia-EPSP. Thus, when the unitary Ia-EPSPs do not occur we can also expect the motor unit to usually not fire in the H-reflex window, and this random firing will not change with conditioning (since background firing is unchanged; Fig. 8D). Note though that even when an EPSP occurs, the motor unit only ever fires once during the H-reflex window, due to these long interspike intervals and the short duration of the Ia-EPSP. Second, any unitary EPSPs that are smaller in comparison with the dominant Ia-EPSP, but always occur (never fail), will make the motor unit fire its spike during the H-reflex window only when the unit is very near the end of its interspike interval, but again the probability of this event is very low, since the interspike interval is long. Furthermore, even when these events occur they are infrequent relative to the dominant unitary Ia-EPSPs and so contribute few points to the overall average PSF. Third, the dominant unitary Ia-EPSPs that do fail intermittently (discussed in the analysis above) have a much higher probability of making a motor unit fire during the H-reflex window since they cause increases in firing rates of up to 30 Hz (see Results), which corresponds to the motor unit firing within 27 ms of the prior spike (1/(30 Hz + 7 Hz), for a 7 Hz baseline firing rate), meaning that the motor unit can fire a spike during the H-reflex window at any time in most of its whole 140 ms interspike interval (from 27 to 140 ms, or ∼80% of the time). Thus, the large dominant unitary EPSPs cause by far the most points (motor units spikes) in the cumulative PSF during the H-reflex window, determine most of the final average firing rate response, and cause most of the changes in motor unit probability with conditioning.

Statistical analysis
Statistical analysis was performed in Sigma Plot 11. The percentage change of various measures (conditioned J Physiol 601.10 H-reflex, PSF and rectified EMG) across the different ISIs were compared with a 0% change using either a one-way ANOVA for repeated measures for normally distributed data (determined by the equal variance test) or by a Friedman repeated measures analysis of variance on ranks for data that were not normally distributed. Post hoc Tukey's tests for the ANOVA and Friedman's test were used to determine which ISIs were significantly different from a 0% change. A two-way ANOVA for repeated measures was used to compare the percentage change in H-reflexes from the 2 Hz and 0.2 Hz conditioning cutaneous stimulation trains, with frequency and time as factors. Post hoc Tukey's tests were used to determine which time bins were significantly different between the two stimulation frequencies. A Mann-Whitney U test was used to compare group values that were not normally distributed and Student's t tests were used for normally distributed data. Data are presented in figures and in the text as means (standard deviation). Significance was set as P < 0.05 and n refers to the number of participants tested in each experiment.

Cutaneous and proprioceptive facilitation of H-reflexes during phasic PAD
To explore whether H-reflex facilitation occurred with a similar time course to phasic PAD , we started by examining whether the soleus H-reflex was facilitated by a cutaneous conditioning stimulation (SFN: 4.1 (1.1) mA at perception threshold) at ISIs between 0 and 200 ms (see schematic of experimental protocol, top of Fig. 2). H-reflexes were initially measured at rest so as not to introduce possible additional activation of PAD and/or facilitation of spinal neurons by voluntarily activated descending pathways during testing. Consistent with the time course of phasic PAD measured in rodents, the soleus H-reflex measured at rest was facilitated when the brief SFN train was applied between 60 and 150 ms earlier, as shown for two participants in Fig. 2A, C and B, D, respectively. The change in the conditioned H-reflex with respect to the test H-reflex is plotted as a function of the ISI (Fig. 2C and D) to evaluate how the facilitation changes with the expected time course of phasic PAD (up to 200 ms; see below). For this analysis, the H-reflex and conditioning SFN volley are expected to have a similar latency in reaching the spinal cord (see Methods, Fig. 1) and thus, an ISI of 0 ms corresponds to a time just before the expected onset of PAD where appreciable H-reflex facilitation is not expected. Longer ISIs correspond to times during the activation of PAD where Ia afferent and reflex facilitation is expected, as highlighted for the H-reflexes at the 60 and 80 ms ISIs in the two participants (see insets in Fig. 2A and B with corresponding percentage change values shown between the grey dashed lines in Fig. 2C and D).
The facilitation of the soleus H-reflex occurred even though the conditioning SFN stimulation itself (when applied alone during a weak contraction) did not facilitate the soleus motoneuron pool as reflected in the mean motor unit firing profile (PSF, light blue line in Fig. 2E and F), remaining close to or slightly below the mean firing rate before the SFN stimulation was applied (red line), and thus ruling out postsynaptic facilitation of the H-reflex. Likewise, the SFN stimulation did not produce an increase in the mean rectified EMG, representing the activity of a larger number of motor units and ruling out the addition of newly recruited units from the conditioning input ( Fig. 2G and H). For comparing the PSF or EMG after the SFN stimulation ( Fig. 2E-H) to the conditioned H-reflex changes ( Fig. 2C and D), the H-reflex-ISI plots were shifted to the right by the latency of the H-reflex, to determine how the SFN stimulation affected the excitability of the soleus motoneurons at the time the H-reflex was activated (if at all; see Methods, Fig. 1). As highlighted for the 60 and 80 ms ISIs at a time comparable to when the conditioned H-reflex was maximally facilitated (between grey dashed lines in Fig. 2), both the PSF and EMG remained close to or slightly below the pre-stimulus values indicating an unfacilitated (or even a weakly inhibited) soleus motoneuron pool when the conditioned H-reflexes were evoked and facilitated.
Overall, from the 16 participants, the mean profile of soleus H-reflex facilitation from the conditioning SFN stimulation resembled the profile of afferent depolarization (PAD) evoked by a cutaneous afferent stimulation , lasting for ∼150 ms but with a later peak at 80 ms (Fig. 3 Ai) and with a significant facilitation of the reflex at the 80 and 100 ms ISIs (see legend for statistics). In contrast, there was no significant change in the mean PSF or EMG for all ISIs tested, again indicating an absence or potentially a very weak inhibition of the soleus motoneurons from the conditioning SFN stimulation. The maximal facilitation of the soleus H-reflex across the different condition-test ISIs in each participant was 42.0 (27.4)% (P < 0.001, Fig. 3B left bar), whereas the SFN-conditioning stimulation alone did not increase the PSF (without H-reflex testing) at these times of H-reflex testing, with the average rate instead reduced slightly compared with the firing rate before the SFN stimulation (−3.4 (4.8)%, P = 0.020, Fig. 3B middle bar). Although there was no significant change in the overall rectified EMG (Fig. 3B right bar), in three participants the rectified EMG was increased compared with the mean pre-stimulus level (at 80 or 100 ms after the SFN stimulation) but this was likely in response to motor unit synchronisation following a preceding EMG suppression, highlighting the fact that caution should be used when using EMG to estimate soleus H-reflex measured at rest was 0.86 (0.76) mV (or 11.4 (9.2)% of Mmax). There was an effect of the conditioning-test ISI on the H-reflex (F(15,7) = 4.3, P < 0.001, one-way ANOVA), being significantly greater than a 0% change at the 80 (P < 0.001, Tukey's test) and 100 ms (P < 0.008, Tukey's test) but no effect of ISI for both the mean PSF (Chi-square, DF 7, P = 0.520) and EMG (F(13,7) = 1.1, P = 0.351), one-way ANOVA). The PSF profiles from the conditioning stimulation alone comprised, on average, 2.33 (2.50) units (range 1−9), 176.47 (105.40) sweeps (range 54−420), a mean pre-stimulation firing rate of 6.95 (1.23) Hz (range 4.79-9.10 Hz) with a firing rate coefficient of variation of 17.0 (5.0)% (range 11.0-31.0%) and 18.02 (14.36) frequency values in each 20 ms bin (range 6.60-54.60). Values were averaged for each participant and then averaged across the group. B, left bar: maximum (peak) percentage change of the conditioned SOL H-reflex for each participant (white circles, n = 16 participants), mean represented by black circle and median by horizontal line, 25th and 75th percentiles by the box bounds, and the 95th and 5th percentiles by whiskers (median greater than a 0% change, P < 0.001, Student's t test). Middle bar: percentage change PSF at the time bin where the peak conditioned H-reflex would have been activated at the spinal cord (n = 16 units, one unit measured in each participant with five units recorded using intramuscular or surface EMG, 11 units decomposed from HDsEMG, median smaller than a 0% change, P = 0.018). Right bar: percentage change of rectified surface EMG as in PSF (median not different from 0% change, P = 0.507). C, the maximum percentage change in the homonymous soleus H-reflex (H) from conditioning of CPN (n = 5 participants: median greater than 0% change, P = 0.008) and corresponding percentage change in EMG (median not greater than 0% change, P = 0.556). Data for conditioning of soleus H-reflex by SFN are replotted from B for comparison. Length of vertical bar represents the mean and whiskers the SD. D, same as in C but for maximum percentage change from CPN conditioning in the heteronymous biceps femoris (Biceps Hetero) reflex (median greater than 0% change, P = 0.008) and associated EMG (median not greater than 0% change, P = 0.690) and for SFN-conditioning of biceps femoris reflex (median greater than a 0% change, P = 0.029) and associated percentage change EMG (median not greater than a 0% change, P = 1.000) (n = 5 participants for CPN, n = 4 participants for SFN), reflexes recorded during a mild contraction. Mann-Whitney's U test used for all pairwise comparisons (B-D) to a 0% change with * P < 0.05, * * P < 0.01, * * * P < 0.001. [Colour figure can be viewed at wileyonlinelibrary.com] J Physiol 601.10 motoneuron membrane potential (Türker & Powers, 2005). In summary, the facilitation of the H-reflex by the SFN stimulation was likely not produced by a direct facilitation of the soleus motoneuron pool at the time of Ia activation based on the PSF and EMG. Rather, facilitation likely occurred at a pre-motoneuron level, as further demonstrated in the motor unit firing probability experiment detailed below.
Conditioning stimulation of proprioceptive afferents also produced facilitation of both homonymous H-reflexes and heteronymous (presumably monosynaptic) reflexes. A low intensity of stimulation to the CPN (also called common fibular nerve, 1.0 × MT), to recruit mainly proprioceptive afferents from the TA muscle (antagonist to the soleus), produced a maximum facilitation of the homonymous extensor soleus H-reflexes of 59.1% (30.9) in five participants (P = 0.008), with no corresponding increase in EMG activity (P > 0.556, Fig. 3C). Like the cutaneous SFN stimulation (replotted in Fig. 3C), the maximum facilitation of the soleus H-reflex from the antagonist CPN conditioning occurred between the 60 and 80 ms ISIs (average 68.0 (11.0) ms, not shown). The conditioning CPN stimulation also facilitated a heteronymous reflex in the biceps femoris muscle (Biceps Hetero) activated by stimulating the MG nerve during a weak biceps contraction. This biceps femoris reflex was evoked at a predicted monosynaptic latency (see Nerve stimulation in Methods), with a maximum facilitation of 27.4 (11.2)% (P < 0.001, n = 5 participants) that occurred between the 60 and 80 ms ISIs (64.0 (8.9) ms, Fig. 3D). Likewise, the cutaneous SFN stimulation facilitated the heteronymous biceps femoris reflex by 35.4 (7.4)% at an average ISI of 60.0 (14.1) ms (P < 0.001, Fig. 3D, n = 4 participants), with both low-intensity conditioning stimuli producing no significant change in the surface EMG, reflecting the excitability of the motoneuron pool at the time the H-reflex was evoked (Fig. 3D, see legend for statistics). Thus, the facilitation of the H-reflex by sensory conditioning not only follows the time course of phasic PAD, but also spatially mimics the widespread distribution of PAD in Ia afferents across many spinal segments from both flexor and extensor muscles (Rudomin & Schmidt, 1999;Willis, 2006).

CST facilitation of the H-reflex during phasic PAD
Similar to the action of sensory conditioning, the H-reflex measured at rest was also facilitated by prior conditioning from the CST activated by TMS to the contralateral motor cortex, which should produce PAD (Carpenter et al., 1963;Liang et al., 2022). This occurred for both the soleus and AbHal H-reflexes (representative AbHal H-reflex data shown for two participants in Fig. 4A and B, inset). A very weak TMS intensity (0.9 × AMT applied at rest, single pulse) was used to condition the resting H-reflex to avoid direct facilitation of the motoneuron pool. The profile of AbHal H-reflex facilitation from TMS conditioning at the different ISIs was similar to the cutaneous conditioning, lasting for ∼150 ms and peaking near 80 and 60 ms in the two participants ( Fig. 4C and D). When applied alone, the conditioning TMS did not produce an increase in the mean firing rate of the tonically active motor units (PSF in Fig. 4E and F) or in the rectified surface EMG ( Fig. 4G and H), again suggesting that the facilitation of the H-reflex was not due to postsynaptic facilitation in the motoneurons.
In the group average (n = 9 participants), the AbHal H-reflex was facilitated across the various ISIs with TMS conditioning (Fig. 5 Ai, see statistics in legend), similar to the sensory afferent conditioning of the H-reflex, being significant at the 80 ms ISI (51.1 (41.8)%, P = 0.033, Tukey's test), and similar in duration to the expected CST-evoked phasic PAD (Carpenter et al., 1963). In contrast, the PSF or background EMG profiles averaged across the group were not significantly altered by the conditioning TMS when applied alone (Fig. 5 Aii and Aiii), with minor fluctuations unrelated to the profile of H-reflex facilitation, and as shown for the EMG, within the variability occurring before the conditioning stimulation. This indicated a lack of effect from the TMS on the AbHal motoneurons. The peak facilitation of the AbHal H-reflex across the various ISIs in each participant was 71.2 (33.7)% (P < 0.001, Fig. 5B, Mann-Whitney's U test), with no change in the firing rate of the AbHal motor units or rectified EMG at time points when the maximal
Cutaneous and proprioceptive facilitation of H-reflexes to produce tonic PAD Slow (0.2 Hz) vs. moderate (2 Hz) speed of cutaneous afferent trains. We next examined whether a more intense and longer duration of cutaneous SFN stimulation could produce a longer-lasting facilitation of the H-reflex consistent with the tonic PAD in Ia afferents produced by such stimuli Lucas-Osma et al., 2018). The SFN intensity was increased to 2 × radiating threshold (10.6 (1.7) mA), potentially to produce GABA spillover, and applied for 10 s at either 2 Hz or 0.2 Hz to determine whether there was a frequency-dependent effect, as previously shown for tonic PAD mediated by α5 GABA A receptors (Lucas-Osma et al., 2018). During both 10 s stimulation trains (2 or 0.2 Hz), the soleus H-reflex was facilitated (pink circles at 0 s time bin, Fig. 6A) compared with the unconditioned test H-reflexes (<0 s time bins), with direct motoneuron depolarization occurring during the period of SFN stimulation (not shown). The H-reflexes following the 0.2 Hz or 2 Hz conditioning SFN stimulation trains (>0 s time bins) increased across all time points (see statistics in legend) with post hoc analysis revealing that H-reflexes were significantly increased (above 0%) at all time points for the 2 Hz train only (all P < 0.05). The facilitation of the H-reflex was greater following the 2 Hz stimulation train compared with the 0.2 Hz train at several of the condition-test ISIs (P < 0.05, marked by asterisks in Fig. 6A).

Fast (200 Hz) cutaneous afferent train.
Given the larger sustained facilitation of H-reflexes by the faster conditioning stimulation train of 2 Hz, we also examined a much higher frequency conditioning stimulation train of the SFN at 200 Hz, but of shorter duration (500 ms, 700 ms ISI). This was very effective in facilitating the H-reflex, consistent with the long-lasting tonic PAD and facilitation of monosynaptic reflexes evoked by this stimulation in animals, lasting for upwards of a minute after the train was terminated Lucas-Osma et al., 2018). Because high-frequency stimuli can be painful, we used an intensity of SFN stimulation that was below the radiating threshold (3.6 (0.3) mA). The 500 ms, 200 Hz SFN stimulation produced a large facilitation of the H-reflex immediately after the stimulus train (pink circle at 0 s time bin, Fig. 6B) compared with the unconditioned H-reflexes (black circles, at time bins <0 s). This large H-reflex facilitation was associated with a ∼500 ms increase in the PSF of the soleus motor units following the SFN train as measured in three participants (not shown), with an average increase in the PSF of 516.67 (175.59) ms and thus may have been partly produced by direct facilitation of the soleus motoneurons. However, after the PSF returned to baseline (e.g. by 700 ms in Fig. 6C, note different time scale in ms), the H-reflex continued to be facilitated for at least 95 s after the high-frequency train ended. The H-reflexes were greater than a 0% change at many time points following the conditioning SFN train (at time bins >0 s, marked by * , P < 0.05, Tukey's test) and larger than any of the H-reflexes evoked before the SFN train (at time bins <0 s).

Repeated slow (0.2 Hz) proprioceptive stimulation
induces self-facilitation. The small gradual increase in H-reflexes from the 0.2 Hz SFN stimulation could have been produced by a small tonic PAD from cutaneous afferent pathways with a build-up in extra-synaptic GABA Lucas-Osma et al., 2018). Similarly, a tonic PAD may also have been produced by the repeated activation of the soleus Ia afferents themselves. During repetitive stimulation of Ia afferents in the rodent, tonic PAD can be produced in other Ia afferents via the classic trisynaptic pathway, and also in the stimulated Ia afferent itself, the latter termed self-facilitation . Thus, we predicted that repeated activation of Ia afferents by TN stimulation alone may also produce long-lasting facilitation of the soleus H-reflex, consistent with self-facilitation from tonic PAD (see schematic in Fig. 7). To examine this, we measured whether there was any build-up of test H-reflexes during baseline measures before any conditioning stimulation was applied. The repetition rate of every 5 s of TN stimulation (0.2 Hz) likely produces a small amount of post-activation depression (Hultborn et al., 1996), but the self-facilitation of the Ia afferents from tonic PAD may override this. In the 19 participants tested, H-reflexes evoked from the 1st to the 14th TN stimulation (set 1, Fig. 2) gradually increased in amplitude over a period of 1 min compared with the test H-reflexes that reached a steady state after the first run of conditioning stimuli (set 2, see statistics in legend). Post hoc, the 6th to 14th H-reflexes were each larger than the 1st H-reflex, suggesting a self-facilitation of the Ia afferents arising from the repeated TN stimulation.

Cutaneous facilitation of single motor unit discharge probability
To provide more direct evidence that a conditioning cutaneous input facilitates the soleus H-reflex by increasing the probability of Ia afferents transmitting spikes to motoneurons, we examined whether the firing probability of single soleus motor units within the H-reflex (Ia-EPSP) window was increased without producing an increase in the amplitude of the motoneuron EPSP estimated by the PSF method. The PSF requires many motor unit firing trials to be averaged, and thus we used cutaneous stimulation trains that produced long-duration increases in afferent depolarization (tonic PAD), where we repeatedly tested the H-reflex before and then after conditioning. The intensity of the TN stimulation used to evoke the H-reflex was set near threshold and specifically adjusted to produce a low probability of the motor unit firing during the H-reflex window. On average the probability was ∼30%, where the random chance of the unit firing within the H-reflex window is only 10% (detailed in Motor unit analysis in Methods). This simplifies the analysis by making the likelihood of intermittent failure of a single afferent branch dominate the Ia-EPSP for that motoneuron, especially since intermittent failure in afferent branch points and associated failure in unitary EPSPs are more readily with H-reflexes after the 2 Hz stimulation greater than a 0% change at all time points (not shown, P < 0.05, Tukey's test). The increase in H-reflex was larger for the 2 Hz stimulation compared with the 0.2 Hz stimulation (F(1,17) = 2.2, P = 0.005, two-way repeated measures ANOVA) with * indicating post hoc analysis where 2 Hz values >0.2 Hz values (P < 0.05, Tukey's test). B, same format as A but in response to a fast cDPN train (200 Hz for 500 ms, n = 15 participants). Pink circle represents conditioned H-reflex where the start of the SFN stimulation train preceded the H-reflex by 700 ms. There was an effect of time on the H-reflex (Chi-square = 100.6, DF = 14 (P < 0.001)) with * indicating significant difference from a 0% change (P < 0.05, Tukey's test). C, PSF of soleus motor unit in response to fast SFN stimulation used in B (note different time scale). Overlay of 100 stimulation trials from one participant. Blue line is binned average of frequency points (20 ms bin width). Red line is pre-stimulus rate. B and C adapted from Hari et al. 2022. [Colour figure can be viewed at wileyonlinelibrary.com] J Physiol 601.10 seen at low stimulation intensities . In this way the motor unit failure can be more readily interpreted as all-or-nothing unitary EPSP failures, as detailed in the Methods.
Participants held a weak plantarflexion to recruit a motor unit with a steady background firing rate, upon which we could evaluate changes in firing rate with TN stimulation (PSF, Fig. 8A) in addition to measuring the compound H-reflex. As shown by representative data from a single participant, the firing rate in the PSF increased at the start of the H-reflex window (left vertical line, Fig. 8A) compared with the background rate, consistent with the underlying Ia-EPSP mediating the H-reflex discharging the motor unit sooner than would be expected from the steady, voluntary drive (PSF in light blue), as previously detailed (Türker & Powers, 2005). A brief duration (50 ms), low-intensity SFN stimulation train (4.0 (0.55) mA, 200 Hz) was applied 500 ms before each TN stimulation to avoid any postsynaptic effects on the motoneuron at the time of Ia-EPSP activation. By ∼200 ms after the conditioning SFN train, any depolarization of the soleus motoneuron subsided as reflected in the PSF returning to baseline before the H-reflex was evoked (light blue trace near red line, Fig. 8 Aii). In this participant, the probability of motor unit discharge during the 15 ms H-reflex window increased from 58% during the test-alone trials (Fig. 8 Ai) to 71% during the SFN-conditioning trials (Fig. 8 Aii). The increased probability of motor unit discharge occurred even though the PSF, representing the profile of the Ia-evoked EPSP (∼10 ms in duration in rats, Hari et al., 2022), was not altered by the cutaneous conditioning stimulation (note overlay of blue test PSF and pink conditioned PSF, inset of Fig. 8 Aii).
Across the group (n = 13), the firing probability of the motor units within the test H-reflex window was 32.9 (14.2)% and increased to 46.3 (12.8)% during the cutaneous conditioning trials (Fig. 8B, see statistics in legend). Importantly, the average PSF within the H-reflex window (13.6 (4.6) ms in duration; reflecting the Ia-EPSP size) did not change in response to the cutaneous conditioning (2.59 (14.95)% change, Fig. 8C right bar), suggesting again that an increase in firing probability of the soleus motoneurons activated by the Ia afferents occurred without an increase in the amplitude of the EPSP. This is consistent with an increased probability of unitary Ia-EPSPs arising from an increased probability of afferent transmission with conditioning. Note that motor units may simply fire by chance (i.e. unrelated to the Ia-EPSP) within the ∼15 ms H-reflex window with a probability of about 10% (see Motor unit analysis in Methods). Thus, the unitary Ia-EPSP that drives the motor unit firing in the H-reflex window likely had a probability that was 10% less than the overall firing probability of the motor unit. This indicates that, on average, the unitary Ia-EPSPs had a low probability of occurring prior to conditioning (∼22.9% probability), that increased by ∼30% with conditioning (to ∼36.3% probability), consistent with the amount of H-reflex increase with conditioning (Fig. 8C). Moreover, the firing probability during the first 0.5 ms at the start of the PSF rise (e.g. near 39 ms in Fig. 8A), which likely corresponds to the period  Fig. 2, set 2 average = 100% as marked by red horizontal line). The test H-reflexes increased over the 14 tibial nerve stimulations (F(18,13) = 3.93, P < 0.001, one-way ANOVA) with H-reflexes 6 to 14 all larger than the first H-reflex (all P < 0.001, Tukey's test). Right, theoretical schematic of soleus Ia afferent collateral activating PAD circuit that synapses back onto its own branch-point node to produce tonic PAD and self-facilitation. [Colour figure can be viewed at wileyonlinelibrary.com] Figure 8. Probability of single motor unit discharge A, a representative PSF of soleus motor units recorded from a single participant using HDsEMG, time-locked to the tibial nerve (TN) stimulation (at 0 ms, 270 μV peak-to-peak test reflex amplitude (5.87% Mmax, 29.8% Hmax near H-threshold)) without (i, test) and with (ii, cond) SFN conditioning (10 pulses, 200 Hz, 500 ms ISI, 4.0 mA below radiating threshold). Activity of 11 units are superimposed over 302 sweeps in i (test H-reflexes) and 311 sweeps in ii (conditioned H-reflexes). To reduce variability in baseline firing rates across units, the mean pre-stimulus firing rate was subtracted from each unit (S-PSF). The mean S-PSF (blue line) is plotted over the mean pre-stimulus rate (red line). The firing probability of the units were measured within the H-reflex window between the grey vertical lines. Inset: estimated EPSP from the test (blue) and conditioned (pink) S-PSF measured in 2 ms bins. B, probability of motor unit (MU) discharge during the H-reflex window before (test) and after (cond) SFN conditioning for each participant (white circles, n = 13 participants, P = 0.014, Mann-Whitney's U test), mean represented by the black circle and median by the horizontal line, 25th and 75th percentiles by the box bounds, and the 95th and 5th percentiles by whiskers. Unit activity was measured with HDsEMG in 12 PSFs and with intramuscular EMG in 1 PSF. C, percentage change of the conditioned soleus H-reflex (left bar) and percentage change of the PSF within the H-reflex window (right bar), with the H-reflex greater than a 0% change (P = 0.001) but no change in the PSF (P = 0.74), both using Mann-Whitney's U test. Inset: example test (blue) and conditioned (pink) H-reflex obtained during PSF recordings; bars = 20 ms, 50 μV. D, average firing rate of the soleus motor units measured 100 ms before TN nerve stimulation from the test (left bar) and conditioned (right bar) stimulation trials, with no difference between the two (P = 0.710, Student's t test). * P < 0.05, * * P < 0.01). The test soleus PSF profiles comprised, on average, 2.56 (1.36) units (range 1.0-6.38), 225.38 (210.73) sweeps (range 55−879), a mean pre-stimulation rate of 7.09 (1.04) Hz (range 5.86-9.20 Hz) with a firing rate coefficient of variation of 20.0 (6.0)% (range 9.0%-32.0%) and 30.48 (26.21) frequency values in each 20 ms bin (range 7.40-106.60). The conditioned soleus PSF profiles comprised, on average, 2.56 (1.36) units (range 1.0-6.38), 223.38 (218.71) sweeps (range 55−899), a mean pre-stimulation rate of 6.95 (1.12) Hz (range 5.03-9.17) with a coefficient of variation of 20.0 (5.0)% (range 9.0%-28.0%) and 26.72 (21.11) frequency values in each 20 ms bin (range 7.40-89.00).
[Colour figure can be viewed at wileyonlinelibrary.com] J Physiol 601.10 of monosynaptic Ia activation of the motoneuron, also increased from a median of 1.51% in the test trials (range 0.70%-2.02%) to 2.70% (1.10%-4.70%) in the conditioned trials (P = 0.024, Mann-Whitney's rank U test), where in comparison the random probability of firing in a 0.5 ms window is around 0.35% (see Motor unit analysis in Methods). In contrast, the firing rates remained similar at 7.39 Hz (1.19 Hz) and 7.12 Hz (1.06 Hz, P = 0.729, Student's t test), consistent with an unchanged monosynaptic Ia-EPSP. This comparison was performed in nine of the 13 participants having enough motor unit spikes in the small 0.5 ms window (see also Aimonetti et al. (2000) for similar findings).
In eight of the 13 participants, the mean PSF during the H-reflex window was unchanged or even decreased in response to the cutaneous conditioning (Fig. 8C right), ruling out changes in motoneuron facilitation (postsynaptic) accounting for the increased motor unit firing probability. The overall soleus H-reflex measured from the surface EMG also increased with conditioning (23.56 (23.6)%, Fig. 8C left bar, see example test and conditioned H-reflexes in inset), as expected from Fig. 6, consistent with the increased probability of the EPSP(s) dominating over any changes in EPSP size and leading to a net increase in the H-reflex. Across the group, the mean background firing rate of the motor unit PSF 100 ms before the TN stimulation with cutaneous conditioning (7.19 (2.1) Hz cond, right bar Fig. 8D) was not different from that without conditioning (7.37 (2.2) Hz test, left bar) in all participants, further supporting the conclusion that changes in firing probability of the units were not mediated by postsynaptic facilitation of the soleus motoneurons. Overall, the increased probability of the motor unit firing during the H-reflex, while the Ia-EPSP (PSF) and background motoneuron activity remained unchanged with conditioning stimulation, is consistent with an increased probability of Ia afferent transmission to motoneurons, without changes in transmitter release, as detailed in rodents ; see also Motor unit analysis in Methods).

Discussion
Our results are consistent with cortical or sensory conditioning stimulation in humans facilitating proprioceptive axon transmission to motoneurons over long periods, as reflected by H-reflex changes, when only the long-lasting actions of GABA axo neurons that innervate these axons are likely to account for this increased transmission. This is specifically consistent with recent observations in rodents that long-lasting activation of GABA A receptors at the nodes of Ranvier in large myelinated proprioceptive axons increases reflex transmission by depolarizing sodium channels closer to threshold, especially at nodes near branch points where spike failure is probable . In rodents, cats and primates, axonal GABA A receptor activation produces characteristically long-lasting depolarizations of Ia axons (PAD; Rudomin & Schmidt, 1999;Willis, 2006) and accordingly, facilitates sensory transmission to motoneurons for a long time Wall, 1958), making the long-lasting facilitation of Ia-mediated reflexes in humans by PAD a likely possibility. Facilitation of sensory transmission by cortical or sensory conditioning in humans is not likely mediated by postsynaptic actions on the motoneuron since the facilitation lasts far too long (minutes with repeated conditioning), outlasts any short latency postsynaptic action of the conditioning, and occurs in the absence of changes in ongoing motor unit firing produced during voluntary contractions. Also, this facilitation is not likely mediated by an inhibition of the GABA axo neurons that innervate sensory axons and cause presynaptic inhibition (via GABA B receptors), because the unitary EPSPs that we estimated from the motor unit PSFs are not increased with conditioning stimulation, even though the probability of the EPSP occurring increases, as assessed by an increased likelihood of motor units firing during the H-reflex. The latter is also broadly consistent with the findings in animals that cutaneous or cortical stimulation alone never decreases activity in the GABA axo neurons that generate PAD (see Introduction), and instead, PAD is often increased by such stimulation and leads to increased reflex transmission by nodal facilitation Liang et al., 2022), consistent with our observations in humans.
Taken together, the facilitation of the H-reflex observed in this human study is consistent with a PAD-mediated facilitation of Ia afferent transmission to motoneurons given that it has similar characteristics to the facilitation of Ia afferents by PAD studied in animals where: (1) the time course of both phasic and tonic PAD in animals (Eccles, Kostyuk et al., 1962;Lomeli et al., 1998) matches the time course of facilitation of the H-reflex; (2) the activation of PAD by a broad range of inputs in animals (CST, cutaneous and proprioceptive; Rudomin & Schmidt, 1999) matches the broad range of inputs facilitating the H-reflex; (3) the frequency-dependence of tonic PAD (0.2 Hz < 2 Hz; Lucas-Osma et al., 2018) matches that of H-reflex facilitation; (4) the self-facilitation of Ia afferents by PAD  matches that of H-reflex self-facilitation; and (5) the increased probability of unitary EPSPs and reduction of axon failure seen during PAD  matches the increased motor unit firing probability during the H-reflex in humans (as detailed above). However, given the indirect measurement of Ia conduction from the H-reflex and single motor unit probability measures, we also provide alternative considerations of this data.

Short-duration facilitation of H-reflexes by sensory and CST pathways
The profile of H-reflex facilitation may reflect the time course of PAD evoked in the contributing Ia afferents. The brief conditioning stimulation of sensory or CST pathways facilitated the H-reflexes for about 100−200 ms and peaked near the 60 or 80 ms ISI. A similar profile of monosynaptic reflex facilitation was produced in rodents following either a brief cutaneous stimulation or from direct light activation of the GABA axo interneurons . In both cases, we propose that the time course of monosynaptic reflex facilitation follows the time course of the evoked phasic PAD in the Ia afferent. An illustration of the general profile of H-reflex facilitation in relation to the estimated PAD evoked in the Ia afferents is provided in Fig. 1. For instance, at the 0 ms ISI, the phasic PAD from the conditioning stimulation likely was not yet activated in the Ia afferents when the afferent was activated by the TN stimulation for the H-reflex. This likely produces a motoneuron EPSP and H-reflex that is uninfluenced by PAD. However, when the TN stimulation followed the conditioning stimulation by 60 to 80 ms, the activation of the TN Ia afferents likely occurs during the presence of the PAD, allowing the TN stimulation to activate more Ia afferent branches and produce more or larger EPSPs and a larger H-reflex (H-reflex at 60 to 80 ms ISI > at 0 ms ISI). A presynaptic mechanism of H-reflex facilitation is likely because when the conditioning stimulation was applied alone, the motor unit PSF did not increase above the mean pre-stimulus rate, indicating that the conditioning stimulation itself did not depolarize the motoneuron to facilitate the H-reflex. It is important to note that the effect of the conditioning stimulation was only tested on the motoneurons recruited during the weak voluntary contractions and may have had a different effect on higher threshold motoneurons also activated during the H-reflex. For example, the cutaneous inputs may have had an excitatory effect on the higher threshold motoneurons, although this effect would not likely have lasted for as long (<60 ms; Nielsen & Kagamihara, 1993) as the long-lasting facilitation consistent with PAD (100-200 ms).
Although the mean profile of H-reflex facilitation closely follows the estimated profile of phasic PAD from the 60 to 80 ms ISI onwards, the facilitation of the H-reflex at the earlier ISIs is smaller than expected based on the PAD profile. This may be due to direct effects on the motoneuron from the conditioning stimulation that mask the facilitation of Ia transmission by PAD. For example, any excitatory or inhibitory activation of the motoneuron may have prevented full H-reflex facilitation at these earlier ISIs due to postsynaptic shunting or direct inhibition of the motoneuron as shown in rodents . Small decreases in the PSF during these earlier ISIs provide some evidence that the motoneuron may have been weakly inhibited by the conditioning sensory and CST stimulation. Likewise, if there was a balanced excitation and inhibition onto the motoneuron to produce an unmodulated PSF, this likely would have increased cell conductance to reduce the Ia-EPSP. In the rodent, when the direct effects on the motoneuron from the conditioning stimulation are removed with voltage clamp, the full effect of the PAD facilitation on monosynaptic reflexes is unmasked. Thus, if anything, subtle postsynaptic effects on the motoneuron from the conditioning stimulation tend to decrease the Ia-EPSP, strengthening the conclusion that H-reflex facilitation occurred from an increase in Ia input to the motoneuron, consistent with a facilitation of Ia afferent conduction given the similarities in the profile and time course of H-reflex facilitation to PAD.
The net facilitating action of GABA on afferent conduction is likely due to the relatively greater expression of GABA A receptors on the dorsally located nodes of the myelinated segments of Ia afferents, compared with the sparser receptor expression found on the unmyelinated terminals of these Ia afferents Lucas-Osma et al., 2018). The few GABA A receptors at the terminals could, in principle, provide a graded shunting of current to produce presynaptic inhibition of Ia inputs onto the motoneuron (shown in invertebrates; Cattaert & El Manira, 1999;Clarac & Cattaert, 1996). However, mathematical models demonstrate that this shunting is not sufficient to reduce the size of the action potential invading the terminal Walmsley et al., 1995). Moreover, PAD measured at the terminal is small (Lucas-Osma et al., 2018), likely owing to the small number of terminal GABA A receptors (Alvarez et al., 1996;Betley et al., 2009;Fink et al., 2014) compared with dorsal parts of the afferent (Lucas-Osma et al., 2018) and the large electrotonic attenuation of current from the last node to the terminal . However, GABA axo neurons may also activate GABA B receptors on the afferent terminal and GABA A receptors on the motoneuron to reduce the size of the monosynaptic reflex (Curtis & Lacey, 1998;Fink, 2013;Hari et al., 2022;Hughes et al., 2005;Pierce & Mendell, 1993). Thus, the net increase in monosynaptic reflexes from the activation of GABA axo interneurons is likely mediated by the activation of GABA A receptors on the dorsal regions of the Ia afferent that have a stronger facilitatory effect on Ia afferent conduction compared with the inhibitory effect of GABA B receptors activated on afferent terminals and the GABA A receptors on the motoneurons. This balance may favour the facilitation of reflexes when the conditioning stimuli are moderate or small and instead favour a suppression of H-reflexes when higher-intensity conditioning stimuli are applied (Pierrot-Deseilligny et al., 1973), which may have stronger effects on GABA B receptor-mediated presynaptic inhibition and/or direct motoneuron inhibition J Physiol 601.10 Metz et al., 2023). It remains to be determined whether the same or separate GABA axo neurons innervate nodes (GABA A ) and terminals (GABA B ) of the Ia afferent. For example, perhaps a separate group of GABAergic neurons innervate the ventral terminals (and GABA B receptors) and are driven by homonymous nerve stimulation, causing the strong depression of the H-reflex with repeated stimulation (i.e. rate-dependent or post-activation depression) as examined in Metz et al. (2023), and another more dorsal group of GABAergic neurons mediate nodal facilitation (via GABA A receptors) that are driven by more diverse afferent (e.g. cutaneous) and descending inputs (Lalonde & Bui, 2021).

Long-lasting facilitation of Ia afferents by cutaneous and proprioceptive inputs
The minutes-long facilitation of the H-reflex from the conditioning stimulation trains is consistent with a long duration of PAD evoked in Ia afferents from multiple, especially high-frequency, sensory inputs produced by the activation of extra-synaptic α5 GABA A receptors on the Ia afferent nodes, potentially from GABA spillover produced by the repeated activation of GABA axo interneurons (Lucas-Osma et al., 2018). We show that trains of cutaneous stimulation (0.2, 2 and 200 Hz) facilitate the soleus H-reflex for up to 2 min, similar to the duration of long-lasting (tonic) PAD recorded in rodent Ia afferents in response to identical stimulation trains applied to a dorsal root Lucas-Osma et al., 2018). It is unlikely that the motoneuron is continually facilitated for 2 min by these high-frequency stimulation trains given that the membrane potential of the motoneuron in the rat, and the motor unit firing rates in the human (PSF), return to pre-stimulation baseline by less than 1 s after the stimulation train. The trains of low frequency (0.2 Hz) cutaneous stimulation produce a smaller sustained facilitation of the H-reflex, comparable to the low-amplitude tonic PAD produced from the same stimulation train in the rodent (Lucas-Osma et al., 2018), and thus, this small facilitation is likely due to a small tonic PAD. It is likewise possible that the gradual increase in H-reflexes over the 2 min recording is produced by tonic PAD evoked by the repeated (0.2 Hz) activation of the soleus Ia afferents themselves when evoking the H-reflex (repetitive TN stimulation). The gradual increase in test H-reflexes with TN stimulation alone (Fig. 7) supports this hypothesis of self-facilitation where collaterals of the Ia afferent activate a PAD network that synapses back onto its own branch-point nodes. Self-facilitation of the Ia afferents is most readily revealed when we use low intensities of TN stimulation that produced an H-reflex of ∼30% of maximum, giving headroom for recruiting new afferent branches with α5 GABA A receptor-mediated tonic PAD . Although the time course and frequency dependency of the H-reflex facilitation by the cutaneous conditioning was similar to tonic PAD in animals, antagonism of the α5 GABA A receptor in humans, like with the neutral antagonist S44819 used in clinical stroke trials (Chabriat et al., 2020;Darmani et al., 2016), is needed to definitively prove that tonic PAD has a role in increasing Ia afferent conduction.

Probability of motor unit firing
To strengthen the conclusion that the facilitation of the H-reflex by the sensory or CST conditioning is mediated by increasing Ia transmission to the motoneurons, we found it useful to measure single motor unit (motoneuron) activity before and during the H-reflex (Ia-EPSP) window. This allows us to: (1) examine whether the conditioning stimulation alone changes the motoneuron depolarization by examining baseline motor unit firing rates just before the H-reflex is evoked (postsynaptic actions), (2) examine whether the conditioning stimulation changes the EPSP size in a graded manner by measuring the PSF during the Ia-EPSP (H-reflex) window since graded changes in EPSP would be mediated by changes in either presynaptic inhibition or postsynaptic facilitation and (3) examine the probability of the evoked Ia-EPSP (all-or-nothing failure) reflected in whether the motor unit participated in the H-reflex or not. Overall, we found that the conditioning-evoked PAD was not associated with an increase in baseline motor unit firing rate or the size of the estimated Ia-EPSP, consistent with a lack of postsynaptic facilitation or a decrease in presynaptic inhibition that would both otherwise grade the EPSP size. If anything, conditioning tended to slightly slow motor unit firing or hyperpolarize motoneurons, as in rats , which would decrease the probability of the motor unit contribution to the H-reflex. Moreover, it is unlikely that cutaneous conditioning increased the gain (output/input) of the motoneurons (Nielsen et al., 2019) since there was no change in the firing rate response of the motoneuron (output) in response to the Ia-or voluntarily evoked EPSPs (input). It should be noted that Nielsen et al. (2019) reported that cutaneous conditioning increased the slope of the relationship between the firing of all the motoneurons (monosynaptic reflex) and the Ia-EPSP size in a single motoneuron in the MG, but not soleus, which they referred to as the recruitment gain (reflex/EPSP). However, we point out that this is unlikely to represent the actual gain of the single motoneurons under consideration (firing/EPSP size) because although Ia afferents project to the majority of motoneurons in the homonymous muscle (Mendell & Henneman, 1971), not all motoneurons in the reflex are activated by each Ia afferent and an associated common EPSP, due to branch-point failure in the afferents Henneman et al., 1984). Thus, the recruitment gain analysis of Neilson confounds changes in afferent (presynaptic branch-point conduction) and motoneuron (postsynaptic) properties and does not represent the motoneuron gain.
In contrast, we found that the conditioning increased the probability of the motor unit participating in the H-reflex without increasing its firing rate, in agreement with findings from rats that PAD prevents branch-point failure and reduces the probability of intermittent, all-or-nothing unitary EPSP failures. Low-amplitude TN stimulation intensities were used with the intention of evoking monosynaptic soleus H-reflexes only and not polysynaptic reflexes and to more readily follow the all-or-nothing activation of single motor units that reflect the all-or-nothing activation of single unitary EPSPs, as detailed in the Methods (Motor unit analysis). Thus, the PSF within the 15 ms H-reflex window likely reflected motor unit discharge during the monosynaptic EPSP and its increased probability by increases in single Ia afferent conduction branches. However, excitatory interneurons in the fast, polysynaptic H-reflex pathway (Burke et al., 1984;Jankowska, McCrea, Mackel et al., 1981) may also be facilitated by conditioning and lead to an increase in the H-reflex that is unrelated to afferent facilitation. Likewise, conditioning may also reduce the activation of inhibitory (e.g. Ib) interneurons activated by the fast TN pathway at latencies 1 ms or longer. However, similar to Aimonetti et al. (2000), we demonstrated that cutaneous conditioning increases the probability of motor unit discharge within the first 0.5 ms of the H-reflex, when polysynaptic components of the H-reflex are unlikely to contribute. This does not by itself rule out conditioning increasing interneuron activity (or reducing inhibitory interneuron activity) in the later polysynaptic portions of the H-reflex, but we argue that such changes in interneuronal activity are unlikely since it would also likely lead to an increase in postsynaptic motoneuron activity, which we do not observe.
Outside of sensory-evoked PAD preventing branch-point failure, there are a few other possible explanations for the increased probability of motor unit discharge contributing to the facilitated H-reflex during conditioning without changing the Ia-EPSP which we must consider. First, the conditioning input might somehow increase the probability of quantal transmitter release at the Ia afferent terminal, thereby increasing the probability of the EPSP. This could occur by a yet undescribed non-GABAergic innervation of the Ia afferent terminal that may facilitate intracellular calcium and neurotransmitter release. However, GABAergic effects on the afferent terminal seem unlikely to increase firing probability because imaging in rodents indicate that Ia afferent terminals mainly only express GABA B receptors  which decrease, rather than increase, transmitter release via its inhibitory Gi protein-coupled pathways (Curtis & Lacey, 1998). Even if a few GABA A receptors were on the Ia terminals, these have little practical depolarizing action as shown from direct terminal recordings (Lucas-Osma et al., 2018) and, if anything, likely inhibits transmitter release, although terminal GABA A receptor activation can facilitate transmission of neurons in the brainstem (Calix of Held) and cerebellum (Purkinje cells) (Trigo et al., 2008;Zorrilla de San Martin et al., 2017). Second, activation of GABA A receptors on the axon hillock near sodium channels also facilitates action potential generation in cortical cells (Szabadics et al., 2006), but whether local GABA action occurs in the dorsal root ganglion of the Ia afferent, like that shown for nociceptors (Du et al., 2017), is an open question. In any case, the action of GABA on the sodium channels of the axon hillock is likely similar to that at the node, so these two mechanisms are complementary. Third, conditioning may somehow decrease GABA B receptor-mediated presynaptic inhibition of Ia afferents; however, this is not likely as this would potentially increase the EPSP, unlike what we observed. Fourth, conditioning may increase the excitatory postsynaptic activation of the motoneurons to facilitate the H-reflex. However, we can rule this out from the finding that the motoneuron did not increase its firing rate with conditioning. Finally, a portion of the H-reflex may be mediated by fast polysynaptic pathways, though as mentioned above, this is unlikely to contribute to our results.
Thus, we conclude that the most likely explanation for the increased H-reflex and motor unit discharge probability during conditioning is a PAD-mediated facilitation of branch-point conduction in the afferents mediating the H-reflex. It is interesting that in these probability experiments (and also for the heteronymous biceps femoris reflex), facilitation of the H-reflex by cutaneous conditioning still occurred even though there may have been additional activation of PAD circuits from descending inputs during the weak contraction. Future experiments will examine the effect of increasing levels of voluntary activation on the long-lasting facilitation of the H-reflex by conditioning sensory inputs. In addition, further work is needed to support the role of Ia afferent conduction, including showing that when the motor unit fires during the H-reflex window, there is a uniformly bigger H-reflex than when the motor unit fails to fire during the reflex, consistent with recruitment of unitary EPSPs that result from an afferent branch that is recruited into action. dorsal columns to the brain ; the latter allowing GABA to modulate sensation directly on the sensory axon.
The regulation of afferent conduction by GABA may also be affected by brain or spinal cord injury given the known changes to spinal GABAergic networks following these insults (Faist et al., 1994;Kapitza et al., 2012;Khalki et al., 2018;Lalonde & Bui, 2021;Mende et al., 2016;Tillakaratne et al., 2000). Perhaps some of the problems with movement control, neuropathic pain and development of spasticity from injury may be produced by alterations in GABAergic control of nodal facilitation in afferents, a topic we are currently exploring.