Central sensitization of dorsal root potentials and dorsal root reflexes: An in vitro study in the mouse spinal cord.

Axo‐axonic contacts onto central terminals of primary afferents modulate sensory inputs to the spinal cord. These contacts produce primary afferent depolarization (PAD), which serves as a mechanism for presynaptic inhibition, and also produce dorsal root reflexes (DRRs), which may regulate the excitability of peripheral terminals and second order neurons. We aimed to identify changes in these responses as a consequence of peripheral inflammation.


| INTRODUCTION
Inside the spinal cord, central terminals of primary afferents receive axo-axonic contacts that regulate their excitability, modulating neurotransmitters' release and the entry of somatosensory information from the periphery (Rudomin & Schmidt, 1999). Circuits responsible for presynaptic contacts are formed by interneurons located in the spinal cord that are activated by segmental and descending inputs (Jimenez et al., 1987;Rudomin, 1990). Axo-axonic contacts are produced by GABAergic neurons, and identified subpopulations are implicated in inhibiting the information coming into the cord through specific afferents (Barber et al., 1978;Boyle et al., 2019;Fink et al., 2014). Opening of GABA-A receptors produces depolarization by chloride ions' efflux, due to the particular chloride gradient present in primary afferent terminals (Price et al., 2009). Transmitter release from afferents can be reduced by inactivation of voltage-dependent sodium channels, which may alter action potential amplitude and shape at the terminal, and/or inactivation of calcium channels, reducing calcium entry into the presynaptic terminal (French et al., 2006;Rudomin & Schmidt, 1999). In addition, action potential amplitude can also be reduced by an increase in membrane conductance or shunting caused by GABA-A receptors' opening (French et al., 2006). Additional neurotransmitters and modulators may shape presynaptic control of peripheral inputs. Among others, the implication of NMDA and GABA-B receptors, as well as serotonin, noradrenaline and opiates has been described (Willis, 1999;Zimmerman et al., 2019).
Presynaptic activity onto central terminals can also elicit action potentials that may travel along the afferents in antidromic direction towards the periphery (dorsal root reflexes-DRRs-). DRRs have been reported in cutaneous and muscle afferents during fictive locomotion (Dubuc et al., 1985;Gossard et al., 1999), as well as in nociceptive fibres after peripheral inflammation (Lin et al., 2000;Sluka et al., 1995). DRRs in stretch-sensitive muscle afferents interfere with the mechanisms of spike generation in peripheral receptors, altering the discharge of action potentials (Gossard et al., 1999). In nociceptive afferents, DRRs may invade peripheral terminals releasing proinflammatory mediators and leading to neurogenic inflammation and nociceptor sensitization (Willis, 1999). In addition, it has been postulated that DRRs elicited by presynaptic contacts onto primary afferent must also travel in orthodromic direction. If such activity affected nociceptive primary afferents, this would lead to the activation of nociceptive pathways (Cervero & Laird, 1996;.
Here, we aimed to study changes in the spinal circuits implicated in the generation of PAD and DRRs due to a sensitization process. For this purpose, we used a model of peripheral inflammation and in vitro electrophysiological recordings of spontaneous PAD and DRRs reflecting the intrinsic activity of these circuits. Spontaneous activity in motor neurons was simultaneously recorded for comparison on the effects of inflammation in the motor output of the cord and for studying sensory-motor communication.
The influence of the main excitatory and inhibitory neurotransmitters and the descending adrenergic pathway on the spinal circuits responsible for spontaneous activity was also tested by pharmacological assays. This work provides insights into the functioning of the spinal circuits mediating PAD and DRRs and their modifications after peripheral inflammation.

| MATERIALS AND METHODS
Experiments were performed on a total of 97 CD1 mice of either sex, aged 6-11 days and weighing between 4.0 and 8.4 g. Animals were maintained with their litter mates under light-dark cycles of 12 h, at a temperature of 21 ± 2ºC and humidity of 55 ± 15%. All procedures were carried out following European Union and Spanish Government regulations on animal handling, were approved by the local ethics committee and the regional government of Madrid (Ref. PROEX 018/16 and PROEX 51.0/20) and comply with ARRIVE guidelines.

| Induction of inflammation
Experiments were performed using spinal cord preparations obtained from naïve and paw-inflamed mice. Peripheral inflammation was induced by intraplantar injections of carrageenan (3% in saline, 30 μl) in both hind paws. Paw diameter and mechanical withdrawal threshold were measured before and 20 h after carrageenan injection. The withdrawal threshold was defined as the increased the spontaneous activity of dorsal roots, which may be secondary to an enhanced output of spinal generators. This can be considered as a novel sign of central sensitization. minimum mechanical force applied with von Frey filaments that elicited a withdrawal response in at least 3 of 5 trials.

| In vitro spinal cord preparations
For this work, longitudinal slices and hemisected cord preparations were used. Each single preparation was used once and considered as an individual observation, no repeated recordings were obtained from the same preparation and only one drug was applied on each preparation. Extraction and conservation procedures have been described before (Rivera-Arconada et al., 2004;Roza et al., 2016). Briefly, mice were anaesthetized with intraperitoneal urethane (2 mg/kg) and the spinal cords were extracted following a dorsal laminectomy. The whole spinal cord was placed on a plate filled with cold artificial cerebrospinal fluid (ACSF at 4ºC) to remove the outer meninges.
For hemisection, cords were gently separated all along the midline isolating left and right sides. Slices containing the superficial laminae of the cord and attached dorsal roots were prepared in a vibratome (400 µm thick).

| Electrophysiological recordings
For hemisected cords, spontaneous activity from primary afferents and motor neurons was recorded by introducing the L4 dorsal and ventral roots into tight fitting glass suction electrodes. When using slices only the L4 dorsal root was recorded. Signals were obtained using a Multiclamp 700A amplifier (Molecular Devices), sampled at 6 kHz and stored for offline analysis using Spike2 software (CED). Recordings from suction electrodes were amplified (×500) and filtered (10 KHz) to obtain DC signals that were digitized and used to study slow depolarizations. Spontaneous depolarizations recorded from dorsal roots (spontaneous dorsal root potentials, S-DRP) are generated by synchronous depolarizations in unidentified primary afferents. Spontaneous depolarizations recorded from ventral roots (spontaneous ventral root potentials, S-VRP) are generated by depolarizations of populations of motor neurons.
This original DC signal was digitally band-pass filtered between 200 and 1200 kHz to obtain an AC signal that allowed recording of fast spikes (Rivera-Arconada et al., 2004). Spikes recorded from dorsal roots correspond to spontaneous dorsal root reflexes (S-DRRs) generated by synchronous firing of primary afferents, which travel in antidromic direction from the spinal cord. Spikes recorded from the ventral root (spontaneous ventral root reflexes, S-VRRs) reflect the synchronous firing of populations of motor neurons. All measurements were taken from 10 min periods of continuous recordings presenting unperturbed spontaneous activity.

| Histological analysis of spinal cord slices
Slices were immersed in fixative and kept in cooled (4ºC) paraformaldehyde 4% overnight. Then the slice was washed in phosphate buffer, cryoprotected in sucrose, included in gelatine and frozen. Transverse sections of 50 µm were then obtained in a cryostat (Leica CM1950). Sections were stained with 0.1% toluidine blue to visualize the main histological landmarks of the dorsal horn.

| Drugs and chemicals
ACSF components, carrageenan lambda, UK 14,304 and the synaptic blockers picrotoxin, gabazine and strychnine were purchased from Sigma Aldrich. NBQX and dAP5 were acquired from Alomone Labs. The ACSF was prepared daily. Drugs were dissolved in milliQ water or DMSO as concentrated stocks (1-50 mM) and conserved frozen at −20ºC. Compounds were diluted in ACSF just before use and applied to the entire preparation during periods of 30-60 min to allow for a complete equilibration within the tissue.

| Measurements
Depolarizations in DC signals were detected using a home-made algorithm that identifies upward deflections and the time at maximum amplitude. This process was manually supervised to avoid inclusion of artefacts. The maximum amplitude from baseline was then measured using spike2 software. For analysis, we only included S-DRP that exceeded 10 µV or S-VRP larger than 20 µV (to avoid the influence of basal potential oscillations present in motor neurons). Time at maximum amplitude and amplitude for each individual depolarization was annotated.
A threshold criterion was used to count S-DRRs and S-VRRs in AC signals. The threshold was set at four times the root mean square value (RMS) of baseline noise. The total number of spikes counted as well as the time point for threshold crossing was annotated for each spike.

| Data analysis
Relations between S-DRPs and S-VRPs were studied by cross-correlation analysis using Spike2 software. Time at maximum amplitude of S-DRPs was chosen as trigger for correlograms. When a S-VRP occurred in a time window between −80 ms (IQR −88 and −73 ms) and +18 ms (IQR 10 and 25 ms) from the trigger, the two events were classed as time-locked.
The time course for the rising and decay phases of S-DRP and S-VRPs was measured. The whole population of events of the same class (time-locked or independent) was averaged for each individual experiment. Rising phase was measured as the time to change from 10% to 90% of the maximum amplitude of the response. Decay phase was characterized by the time constants obtained by fitting the data to a one or two phase exponential function. Curve fitting was made using GraphPad Prism 7.0 software (GraphPad Software), that calculate the parameters that best adjust the mathematical equation to the data.
It was common to observe S-DRRs immediately preceding S-DRPs. S-DRRs appearing in the 200 ms before the maximum amplitude of each depolarization were classified as associated activity.
Statistical analysis was performed using GraphPad Prism 7.0 software. All data sets were analysed using nonparametric tests and all data are represented as median and interquartile range (IQR, 25th and 75th quartiles). Wilcoxon matched-pairs signed rank test for paired and Mann-Whitney test for unpaired data comparisons were used.

| Paw inflammation enhances spontaneous activity
Recordings from dorsal and ventral roots obtained from spinal cord preparations in vitro showed abundant spontaneous activity in the form of depolarizations and spike firing (Figure 1a,b). In 19 hemisected spinal cords studied, depolarizations in the dorsal root had a median amplitude of 46 µV (IQR 40-60 µV) and a median frequency of 25.9 S-DRP/min (IQR 19.9-35.7 S-DRP/min). Spike firing in the dorsal root occurred both in association and absence of depolarizations, and showed a frequency of 45.5 S-DRR/min (IQR 31.6-63.1 S-DRR/min). In the ventral root, spontaneous activity consisted of depolarizations that showed a median amplitude of 95 µV (IQR 68-114 µV) and a frequency of 50.7 S-VRP/min (IQR 26.2-96.9 S-VRP/min). Spike firing at a frequency of 26.0 S-VRR/min (IQR 20.7-37.1 S-VRR/min) was also observed. Equivalent recordings were obtained from spinal cords extracted from mice subjected to an experimental inflammation of the hindpaws 20 h prior to cord extraction. The intraplantar injection of carrageenan into the hindpaws produced a significant increase of paw diameter from 2.2 mm (IQR 2.1-2.2) to 2.7 mm (IQR 2.5-2.9; n = 33, p < 0.001). Mechanical nociceptive threshold assayed with Von Frey filaments was reduced from a median force of 1 g (IQR 0.8-1.2) before to 0.04 g (IQR 0.04-0.07; n = 33, p < 0.001) after inflammatory treatment, indicating the development of sensitization. Spinal cords extracted from paw-inflamed mice showed an altered spontaneous activity in vitro. Both the amplitude and the frequency of S-DRP, as well as the frequency of S-DRR, increased compared to cords obtained from naïve mice (Figure 1c,d). In the ventral root the median amplitude, but not the frequency, of S-VRP was increased, the frequency of S-VRR did not change after paw inflammation (Figure 1e,f).

| Spontaneous activity in spinal cord slices from naïve and paw-inflamed mice
Dorsal root recordings were also obtained from spinal cord slices containing the dorsal aspect of the spinal cord and the attached dorsal roots. Histological analysis of the preparations used showed that slices included lamina I through to V (see Figure 2). In slices from naïve mice (n = 10), values for amplitude (48.5 µV, IQR 39.2-56.5 µV) and frequency of S-DRP (27.7 S-DRP/min, IQR 20.9-34.6 S-DRP/min), as well as firing of S-DRR (36.4 spikes/min, IQR 27.6-57.9 spikes/min) were not statistically different to those obtained from hemisected cord preparations. These results suggest that circuits responsible of dorsal root activity should be contained within the dorsal horn.
In 10 spinal cord slices obtained from mice that had undergone paw inflammation, the activity recorded from the dorsal root was significantly increased compared to slices from naïve mice, indicating that the increase in activity seen after sensitization involves neurons contained within the dorsal horn. The median amplitude of S-DRP (69.5 µV, IQR 60-117.5; p < 0.001) and the firing of S-DRR (77.1 S-DRR/min, IQR 57.7-263.6; p < 0.01) augmented, although the frequency of S-DRP did not change significantly (32.4 S-DRP/min, IQR 25.4-44.3).

| Classification and characteristics of spontaneous activity
A more detailed analysis of the recordings allowed identifying that some spontaneous depolarizations in dorsal and ventral roots were time-locked, whereas others were independent (Figure 3a-c). In order to study the properties of the different events, we used correlograms (see methods, Figure 3d) to classify them as dorsal root time-locked, ventral root-time locked, dorsal root independent and ventral root independent.
To analyse the temporal relationship between timelocked dorsal and ventral depolarizations, they were averaged and superimposed together with the accumulated number of S-DRRs. This analysis was performed for each experiment (see example on Figure 3e) and the results show that spiking in the dorsal root precedes depolarizations.
Comparing time-locked versus independent S-DRPs, the former had larger amplitude and frequency ( Figure  3f,g), but both had similar time course (Table 1). S-DRRs were often associated to time-locked depolarizations (35.3%, IQR 27.3-51.8%) and less commonly to independent ones (5.4% of S-DRR, IQR 2.9-8.5%; p < 0.001). There were still a remaining 56.1% S-DRRs (IQR 40.1-66.5%) that occur in the absence of any recorded depolarization. Time-locked S-DRP had a stronger relation to S-DRRs F I G U R E 1 Spontaneous activity recorded simultaneously from dorsal and ventral roots in a hemisected spinal cord preparation. Electrophysiological recordings from dorsal (a) and ventral roots (b) showed the spontaneous activity in primary afferent terminals and motor neurons, respectively. Spontaneous activity consisted in slow depolarizations (DC signal) and spike firing (AC signal). Graphs show a quantification of the spontaneous activity in cords from naïve (blue) and paw-inflamed mice (red). In the dorsal root the amplitude and frequency of S-DRP (c) as well as the frequency of S-DRR (d) was increased after inflammation. Graphs in e and f show the equivalent quantification for S-VRP (e) and S-VRR (f). Each data point represents mean values for an individual spinal cord preparation. Interquartile range for each data set is shown superimposed to individual data points. Asterisks indicate statistically significant differences between naïve and paw-inflamed mice (*p < 0.05; **p < 0.01 Mann-Whitney test). than independent ones (52.0% of depolarizations with firing, IQR 31.8-56.7% vs. 29.4%, IQR 14.4-37.1%, p < 0.001) and a larger number of associated spikes (2.06 S-DRR/S-DRP (IQR 1.45-3.17) vs. 1.25 S-DRR/S-DRP, IQR 1.00-1.52; p < 0.001).
In the ventral root, comparison between time-locked and independent S-VRPs showed that the first had a larger amplitude but both had similar frequency (Figure 3h,i). These two types of events had also different kinetics (Table  1) with time-locked depolarizations showing longer rising and decay times.
In general terms, cords from paw-inflamed mice showed an enhanced spontaneous activity that reflected in significant changes in several parameters, as summarized in Figure 3. In the dorsal root, all S-DRPs had larger amplitude than their respective controls whereas frequency increased only in independent S-DRPs ( Figure  3f,g). Rising and decay phases of S-DRP did not change because of the inflammatory insult. Similarly, S-DRRs showed larger frequency in the treated group, showing an increase in spikes per depolarization, which affected to time-locked (3.22 S-DRR/S-DRP, IQR 2.09-4.98; p < 0.05) and independent events (1.45 S-DRR/S-DRP, IQR 1.21-1.85; p < 0.05). However, the percentage of S-DRRs that occur associated or in the absence of any depolarization did not change.
In the ventral root, both time-locked and independent depolarizations increased in amplitude compared to cords from naïve mice, with no change in frequency or kinetics (Figure 3h,i).
These results indicate that different types of spontaneous depolarizations can be recorded from both roots and may represent the output of different spinal circuits. Central sensitization may differentially affect these circuits.

| Pharmacological characterization of spontaneous activity
Pharmacological assays were carried out to define the implication of the main neurotransmitter´ systems in generating spontaneous activity in the spinal cord. Depolarization in dorsal and ventral roots were abolished after the blockade of non-NMDA receptors with 5 µM NBQX (n = 8),

F I G U R E 3
Characteristics of time-locked and independent depolarizations in both dorsal and ventral roots in cords from naïve and paw-inflamed mice. Original recordings exemplify the different classes of depolarizations studied. Some depolarizations were recorded simultaneously in both roots (a) and were classified as time-locked depolarizations. Note the S-DRRs associated with the depolarizations in AC signal. In both dorsal (b) and ventral root recordings (c) was possible to observe independent depolarizations. Graph in (d) shows a cross correlogram between the time at maximum deflection for S-DRP and the occurrence of S-VRPs from a representative experiment. A clear peak indicative of the existence of correlated activity was observed at ≈ −25 ms. Bin size for cross correlogram is 5 ms. Graph (e) includes the averaged shape of time-locked depolarizations in dorsal and ventral roots as obtained from a representative experiment. In addition, S-DRR counts accumulated for the total period of analysis are included (bin size 5 ms). All events are aligned to the maximum depolarization in S-DRP. S-DRRs were concentrated forming a peak that precedes the rising phase of depolarizations. Graphs f-i show a quantification of mean amplitude (f, h) and frequency (g, i) of depolarizations in dorsal and ventral roots as recorded from hemisected spinal cord preparations obtained from naïve (n = 19) and paw-inflamed mice (n = 23). Values include data for the depolarizations after the classification in timelocked (diamonds) and independent events ("Indep.", triangles). Asterisks stand for statistically significant differences between naïve and paw-inflamed mice obtained using Mann-Whitney test (*p < 0.05; **p < 0.01; ***p < 0.001).
indicating their synaptic origin (Figure 4a). Bursts of S-DRR and S-VRR that occurred in association to depolarizations were absent after NBQX, but some background activity was still present.
Blockade of NMDA receptors with 50 µM dAP5 showed depressant effects on spontaneous activity recorded from both roots (n = 10, Figure 4b). In the dorsal root the frequency of all S-DRPs and the amplitude of time-locked S-DRPs were significantly reduced ( Table 2). The frequency of S-DRR was significantly reduced as well. In the ventral root the amplitude and frequency of S-VRP was depressed. The frequency of S-VRRs were also reduced by dAP5. These results suggest that spontaneous activity depends on neuronal circuits connected by excitatory glutamatergic synapses acting on both NMDA and non-NMDA receptors.
In addition, dorsal root activity was also dependent on the activation of GABA-A receptors (Figure 4c). Application of 3 µM gabazine (n = 8) almost abolished S-DRP, with only few low amplitude depolarizations (0.1-0.7 S-DRP/min) remained in 5 of 8 preparations (see Figure 4c). In addition, gabazine reduced the frequency of S-DRR, eliminating spiking associated to S-DRP.
In contrast, the activity in the ventral root was greatly increased. After gabazine, time-locked depolarizations were virtually absent. Independent S-VRP did not change in amplitude but increased in frequency from 24.4 S-VRP/ min in control (IQR 15.6-73.1 S-VRP/min) to 81.1 S-VRP/ min after GABA-A blockade (IQR 56.1-175.5 S-VRP/min; p < 0.01). The frequency of S-VRR was also increased from 31.2 S-VRR/min (IQR 17.3-66.0 S-VRR/min) to 148.4 S-VRR/min (IQR 54.8-553.1 S-VRR/min; p < 0.01) by gabazine. The same pattern of effects was observed with 20 µM picrotoxin (n = 5). These experiments indicate that GABAergic activity onto primary afferents is dependent on GABA-A receptors, and that ventral root activity originates from spontaneous circuits controlled by inhibitory neurons.
Blocking glycine receptors seems to have less influence on rhythmicity, but potentiates the amplitude of the spontaneous activity, principally in the ventral root.

| Modulation of spontaneous activity by the α2-adrenergic agonist UK 14,304
Finally, we wanted to explore the influence of the descending adrenergic system in the spontaneous activity observed in dorsal and ventral roots. Adrenergic modulation T A B L E 1 Rising and decay kinetics of time-locked and independent depolarizations recorded in dorsal and ventral roots

S-VRR
Frequency (1/min) 16.9 (11.6-29.2) 10.8 (6.5-14.7) ** Note: Median (IQR) frequency and amplitude of events are included for the events classified as dorsal and ventral time-locked and independent (Indep.). Median (IQR) frequency of S-DRR and S-VRR is also included. Asterisks stand for statistically significant differences obtained using Wilcoxon matched-pairs signed rank test between values in control conditions and after dAP5 application.
is determinant for spinal cord processing of nociceptive information and the activation of α2-adrenergic receptors at the spinal level has analgesic effects. In naïve animals, the application of the α2-adrenergic agonist UK 14,304 produced a depression of spontaneous activity in dorsal and ventral roots (see Table 3). The median frequency and amplitude of time-locked S-DRP was significantly reduced. In contrast, independent S-DRP were not affected by UK 14,304. S-DRR frequency was significantly depressed. In the ventral root, frequency of both types of depolarizations was strongly reduced without changing their median amplitude. No effects were observed on the frequency of S-VRR.
In spinal cord preparations obtained from pawinflamed mice, UK 14,304 also reduced the frequency of time-locked S-DRP and S-DRR (Table 3). However, effects of UK 14,304 on the amplitude of S-DRP showed some differences between naïve and paw-inflamed mice. Contrary with the actions observed in naïve, UK 14,304 failed to depress the amplitude of time-locked S-DRP. In addition, the amplitude of independent S-DRP was significantly increased. In the ventral root, UK 14,304 showed the same effects observed in naïve mice. These results indicate that adrenergic transmission may regulate spontaneous activity in the spinal cord, in addition the modulation of S-DRP may change during central sensitization.

| Changes in spontaneous activity after peripheral inflammation
In this study, we show that spontaneous activity in the dorsal and ventral roots is modified by a peripheral inflammatory insult, which highlights a novel component of central sensitization. This effect is preserved in slices containing only the dorsal horn, which indicates that it is produced in sensory areas of the cord. The inflammatory insult produced signs of sensitization in the behaving animal, as was reported on previous occasions (Hedo et al., 1999;Rivera-Arconada & Lopez-Garcia, 2010). Moreover, the amplitude and frequency of DRPs, the experimental correlate of PAD, was significantly augmented on the isolated spinal cord under inflammatory conditions. Several mechanisms for the enhanced PAD during F I G U R E 4 Pharmacology of spinal circuits responsible for spontaneous activity. Figures show simultaneous recordings from dorsal and ventral roots obtained from naïve mice before (Control) and after drug application. The non-NMDA receptor blocker NBQX at 5 µM depressed spontaneous activity abolishing depolarizations (a). Blockade of NMDA receptors with dAP5 (50 µM) also reduced spontaneous activity (b). Application of the GABA-A receptor antagonist gabazine (GZ, 3 µM) produced a profound depressant effect on dorsal root activity, but showed excitatory actions in the ventral root (c). Arrow point at one of the few small depolarizations that can be detected in the dorsal root during gabazine application. Recordings in (d) show the excitatory effects of strychnine (Strych, 1 µM), an antagonist of glycine receptors, on the activity recorded from both roots.
inflammation have been proposed. Previous studies have shown increased GABAergic currents in afferent neurons , increased activity of NKCC1 transporters (Price et al., 2009), reduced potassium currents in DRG neurons and depolarization-induced activation of low voltage-activated Ca 2+ currents . In addition, the frequency of DRRs was significantly higher after inflammation. This observation is consistent with previous reports showing backfiring on primary afferent neurons, including nociceptive ones, in live adult rats, cats and monkeys after inflammation (Lin et al., 2000;Sluka et al., 1995). We observed that motor neurons enhanced subthreshold activity, reflecting an increased synaptic input that may alter their excitability. In this context, increased responses to afferent stimulation have been reported under in vivo and in vitro conditions (Hedo et al., 1999;Woolf, 1983). Given their characteristics, the in vitro preparations constitute an excellent model for the study of the mechanisms of sensitization at a cellular and circuit level.
In isolation from peripheral receptors and superior centres, it would be expected that spontaneously active neurons in the spinal cord would constitute the source of excitation to the circuits mediating PAD. Several groups have reported the presence of spontaneous activity in spinal cord neurons in vitro and in vivo (Li & Baccei, 2011;Lucas-Romero et al., 2018;Luz et al., 2014;Roza et al., 2016;Sandkuhler & Eblen-Zajjur, 1994), and the local circuits involved in the control of afferent excitability include neurons with spontaneous activity (Chavez et al., 2012). Furthermore, under pathological conditions, the synchrony between spontaneously active neurons may increase (Roza et al., 2016), which may promote the activation of pathways generating PAD (Chavez et al., 2012) and subsequently contribute to the augmented output on the dorsal root.
In our study, the observed spiking in primary afferents was considerable in comparison with reports from in vivo experiments. This may have been due to factors such as the animal's age and the recording temperature (Bagust et al., 1989;Brooks et al., 1955;Sluka et al., 1995). During postnatal development, the proportion of neurons with intrinsic firing may be reduced (Li & Baccei, 2011), and there is a maturation of the descending control system (Schwaller T A B L E 3 Summary of the effects of 0.1 µM UK 14,304 on spontaneous activity recorded from dorsal and ventral roots in cords from seven naïve and seven paw-inflamed mice  et al., 2016), which likely reduces the occurrence of spontaneous activity. Additionally, cooling the cord may enhance the excitatory actions of spinal neurons, promoting the appearance of dorsal root activity (Brooks et al., 1955).

| Spinal circuits mediating spontaneous activity of dorsal and ventral roots
Our results suggest the existence of different generators for the spontaneous activity recorded. The simultaneous recordings from dorsal and ventral roots showed events that were clearly time-locked, suggesting a common generator mechanism. In the dorsal root, these were large amplitude depolarizations, which are often associated with spike firing. Our analysis suggests that afferent fibres having functional connectivity to motor neurons, generate depolarization that is recorded from the ventral root (Bos et al., 2011;Duchen, 1986). In contrast, we observed independent S-DRPs with lower amplitude and frequency than time-locked events. Further considering the differential effects of strychnine on both types of S-DRPs, it is possible that different generators promote both types of events in the dorsal root. Moreover, it has been reported that PAD in large-calibre fibres of cutaneous origin is subjected to phase-dependent modulation that is related to the locomotor cycle. This happens through pathways and/ or mechanisms that are different to those responsible for PAD when originated by peripheral inputs (Gossard et al., 1990). In addition, it has also been noted that PAD onto identified cutaneous mechanoreceptors of different sensory submodalities may be elicited by separated systems (Schmidt, 1971). The existence of independent circuits with different characteristics governing the excitability of primary afferents may allow for the modulation of different sets of afferents within the context of diverse events. Further to time-locked S-VRP, ventral roots also showed activity that was independent of dorsal events with characteristic amplitude and time-course. These events require a different generator. As such, collecting on the previous considerations, we propose the existence of several local generator circuits within the cord, which may serve to modulate excitability in different sensory afferents, motor neurons and sensory-motor transmissions. Our results also indicate that generators may be differentially affected by central sensitization since it was only the independent S-DRP that increased in frequency after paw inflammation.
In addition, this variety of spontaneous events in the dorsal root could be generated on different types of afferents. Under the present experimental conditions, it is not possible to establish the nature of the afferents implicated.
However, since the spontaneous activity is preserved in a slice containing only the dorsal horn, it is likely that terminals of fine afferents and mechanoreceptors of cutaneous origin may be major contributors (Bernardi et al., 1995;Rudomin et al., 2013). Furthermore, neurons that form presynaptic contacts with proprioceptive afferents have been described in deeper laminae, including the intermediate zone of the spinal cord and motor nuclei (Eccles et al., 1963;Fink et al., 2014;Jankowska et al., 1981). The excitability in different types of cutaneous afferents may be regulated by related but independent circuits, as has been previously suggested . Since independent S-DRP increased in frequency after inflammation, it is tempting to speculate whether this type of depolarization may involve nociceptive afferents/circuits. However, this is a possibility that will have to be addressed in future studies. Since recording from individual terminals of fine afferents within the cord is technically challenging, other alternatives, such as calcium imaging in specific populations of primary afferents, may be successful (Chen et al., 2014).

| Pharmacological characterization of the recorded activity
The main mechanism for the depolarization in the afferents is produced by GABAergic interneurons that form axo-axonic contacts with afferent terminals (Barber et al.,1978;Fink et al., 2014). The release of GABA produces PAD, which reduces the efficacy of transmitter release. However, if the threshold is surpassed, then action potentials are fired at the terminal, which propagates in orthodromic and antidromic directions (Willis, 1999). In our study, a blockade of GABA-A receptors abolished the activity in the afferents confirming this mechanism. The increase in ventral root activity is likely due to disinhibition after GABA-A blockade.
The blockade of glutamatergic transmission also suppresses activity in primary afferents. This suggests that local circuit neurons with autogenous spiking provide excitatory input to the GABAergic interneurons contacting primary afferents. The spinal circuits controlling the excitability of afferents not only integrate inputs from neighbouring fibres but also descending commands from the brain (Jimenez et al., 1987). Dorsal root responses elicited by the electrical stimulation of a nearby dorsal root have shown similar pharmacological sensitivity to both the GABAergic and glutamatergic blockade (Rivera-Arconada & Lopez-Garcia, 2006). Here, we also show that spontaneous activity was depressed by the α2-adrenergic receptor agonist UK14,304, indicating its modulation by descending pathways.