GABAA and Glycine Receptor-Mediated Inhibitory Synaptic Transmission onto Adult Rat Lamina IIi PKCγ-Interneurons: Pharmacological but Not Anatomical Specialization

Mechanical allodynia (pain to normally innocuous tactile stimuli) is a widespread symptom of inflammatory and neuropathic pain. Spinal or medullary dorsal horn (SDH or MDH) circuits mediating tactile sensation and pain need to interact in order to evoke mechanical allodynia. PKCγ-expressing (PKCγ+) interneurons and inhibitory controls within SDH/MDH inner lamina II (IIi) are pivotal in connecting touch and pain circuits. However, the relative contribution of GABA and glycine to PKCγ+ interneuron inhibition remains unknown. We characterized inhibitory inputs onto PKCγ+ interneurons by combining electrophysiology to record spontaneous and miniature IPSCs (sIPSCs, mIPSCs) and immunohistochemical detection of GABAARα2 and GlyRα1 subunits in adult rat MDH. While GlyR-only- and GABAAR-only-mediated mIPSCs/sIPSCs are predominantly recorded from PKCγ+ interneurons, immunohistochemistry reveals that ~80% of their inhibitory synapses possess both GABAARα2 and GlyRα1. Moreover, nearly all inhibitory boutons at gephyrin-expressing synapses on these cells contain glutamate decarboxylase and are therefore GABAergic, with around half possessing the neuronal glycine transporter (GlyT2) and therefore being glycinergic. Thus, while GABA and glycine are presumably co-released and GABAARs and GlyRs are present at most inhibitory synapses on PKCγ+ interneurons, these interneurons exhibit almost exclusively GABAAR-only and GlyR-only quantal postsynaptic inhibitory currents, suggesting a pharmacological specialization of their inhibitory synapses.


Introduction
Mechanical allodynia (pain sensation in response to a normally innocuous mechanical stimulus) is one of the most prevalent symptoms of both inflammatory (initiated by tissue damage/inflammation) and neuropathic (following nervous system lesion) chronic pain syndromes [1,2]. How can light touch evoke pain? It is now well established that mechanical allodynia is associated with the activation, within the spinal dorsal horn (SDH) or medullary dorsal horn (MDH), of dorsally directed polysynaptic circuits. These allow tactile inputs (Aβ-, Aδ-, and C-fiber low-threshold mechanoreceptors, LTMRs), which terminate within or below inner lamina II (II i ) [3,4], to gain access to the pain circuitry in the superficial SDH/MDH [5][6][7][8]. Mechanical allodynia thus results from a miscoding, with cells that normally respond to and transmit only noxious stimuli being activated by tactile inputs.
It is assumed that LTMR inputs onto these PKCγ + interneurons activate convergent feedforward inhibitory circuits that prevent such LTMR inputs from activating PKCγ + interneurons and, in turn, from being transferred to pain circuitry. Allodynia is thought to stem from the loss of feedforward inhibition onto these PKCγ + interneurons [5,6,14]. This conclusion is consistent with recent circuit-based evidence. Using paired recordings, Lu et al. [20] showed that PKCγ + interneurons are normally under feedforward glycinergic disynaptic inhibition from lamina III neurons, thus preventing Aβ LTMRs from activating PKCγ + interneurons. PKCγ + interneurons are also inhibited by parvalbumin-expressing inhibitory interneurons [15], which receive strong monosynaptic input from several classes of myelinated LTMRs [25]. Notably, these inhibitory connections onto lamina II i PKCγ + interneurons appear to be lost under neuropathic conditions [15,20].
Both glycine (GlyR) and GABA A receptors (GABA A R) are present on lamina II i PKCγ + interneurons [5,22]. However, functional evidence suggests a pharmacological specialization of inhibitory inputs onto the lamina II i PKCγ + interneurons. Thus, Aβ LTMR inputs onto these PKCγ + interneurons were shown to be under exclusively feedforward glycinergic inhibition [20,26]. Moreover, mechanical allodynia is specifically dynamic under glycinergic disinhibition but static under GABA A ergic disinhibition [5,6,14]. This issue has already been addressed for lamina II neurons with whole-cell patch-clamp recordings in ex vivo SDH/MDH slices and analysis of action potential-independent miniature inhibitory postsynaptic currents (mIPSCs). Such studies revealed that, whereas GABA and glycine are co-released from interneurons [27][28][29] and GABA A R and GlyR are colocalized on postsynaptic neurons [30], quantal postsynaptic currents are either GlyR-or GABA A R-mediated, but never both, in adult neurons [27,28,[31][32][33][34][35][36][37]. Importantly, in all these studies, the identities of the recorded neuron were never examined. Therefore, whether such a pharmacological specialization of inhibitory synapses exists in the lamina II i PKCγ + interneurons and whether it is related to any anatomical specialization are still unknown.
To address these issues, we carried out (i) an electrophysiological analysis (whole-cell patch-clamp recordings) of spontaneous IPSCs (sIPSCs) and mIPSCs recorded from PKCγ + interneurons in the adult rat MDH and (ii) an immunocytochemical analysis of GlyRs and GABA A Rs, present at inhibitory synapses on these cells. We found that three distinct populations of quantal inhibitory synaptic currents can be recorded from MDH lamina II i PKCγ + interneurons according to decay kinetics: fast monoexponential, slow monoexponential, and biexponential mIPSCs. Notably, all PKCγ + interneurons exhibit these three types of mIPSCs in equal proportions. Pharmacological analysis indicates that fast monoexponential mIPSCs are GlyR-mediated, and slow monoexponential ones are GABA A R-mediated. Therefore, most, if not all, mIPSCs onto MDH lamina II i PKCγ + interneurons are either GlyR-or GABA A R-mediated, not both. On the other hand, our immunocytochemical study provides evidence in favor of colocalization of GABA A R and GlyR on MDH lamina II i PKCγ + interneurons. Altogether, this allows for a functional specialization of the feedforward inhibition of the different LTMR inputs onto PKCγ + interneurons (see the Figure 1 in the work of [9]) but raises the question as to how anatomical data can account for such specialization.

Animals
Adult male Sprague-Dawley rats (21-35 days old, 50-100 g, n = 13; 42-49 days old, 280-290 g, n = 3) were obtained from Charles River (L'Arbresle, France) and Harlan (Loughborough, UK) Laboratories, respectively, and housed 3-4 per cage under standard laboratory conditions (22 ± 1 • C, 12 h light/dark cycle, lights on at 07:00 a.m., food and water ad libitum). All efforts were made to minimize the number of animals used. All animal experiments were performed in accordance with the ethical guidelines of the International Association for the Study of Pain (IASP) [38], the Directive 2010/63/UE of the European Parliament, the Council on the Protection of Animals Used for Scientific Purpose, and the UK. Animals (Scientific Procedures) Act 1986. Protocols applied in this study were approved by the local animal experimentation committees: CEMEAA "Comité d'Ethique en Matière d'Expérimentation Animale Auvergne" (n • CE 28-12) and the Ethical Review Process Applications Panel of the University of Glasgow.

Patch-Clamp Recordings
Slices were transferred into a recording chamber (volume ≈ 1 mL) and held down with a C-shape platinum wire supporting transverse nylon fibers. The chamber is mounted on an upright microscope fitted with fluorescence optics (AxioExaminer, Carl Zeiss, Hamburg, Germany) and linked to a digital camera QImaging Exi Aqua (Ostrava, Czech Republic). Slices were continuously perfused at 3.0 mL/min with aCSF solution maintained at room temperature (≈25 • C). The substantia gelatinosa of MDH has a distinct translucent appearance that can be easily distinguished under the microscope using the 10× objective lens. Lamina II can be divided into outer and inner (II i ) laminae II of equal size, and neurons were recorded within lamina II i under visual control using a 63× water-immersion objective lens with combined infrared and differential interference contrast (DIC).
Spontaneous and miniature IPSCs (sIPSCs and mIPSCs, respectively) were recorded using the patch-clamp technique in whole-cell configuration and voltage-clamp mode. Acquisitions were performed using Clampex 10 software (Molecular Devices, Sunnyvale, CA, USA) connected to a Multiclamp 700B amplifier (Molecular Devices, Sunnyvale, CA, USA) via a Digidata 1440A digitizer (Molecular Devices, Sunnyvale, CA, USA), or Patch Master v2x65 software (HEKA Elektronik, Lambrecht, Germany) connected to an EPC10 amplifier (HEKA Elektronik, Lambrecht, Germany). Voltage-clamp data were low pass filtered at 3 kHz and digitized at 10 kHz. Series resistance was monitored throughout the experiments and was not compensated. Data were discarded if the series resistance varied more than ±20 MΩ.

Immunofluorescence Detection of PKCγ/Neurobiotin Interneurons
At the end of the recordings, epifluorescence was used to ensure that the recorded cells were filled with neurobiotin. Slices were then transferred into 4% paraformaldehyde in 0.1 M phosphate-buffered solution (pH 7.4) and stored overnight at 4 • C. Double labeling PKCγ/neurobiotin was performed for post-hoc analysis. Slices were first incubated with Avidin DCS-rhodamine (1:200, Vector Laboratories, Burlingame, CA, USA) and then placed into primary antibody solution containing polyclonal guinea pig anti-PKCγ (1:4000; Frontier Institute Co., Ltd, Hokkaido, Japan; RRID: AB-2571826). After subsequent washes, slices were incubated with Alexa Fluor ® 488-conjugated goat anti-guinea pig secondary antibody (1:200, Jackson Immunoresearch Laboratories, Inc., West Grove, PA, USA; RRID: AB-2337438). Slices were then transferred onto gelatinized slides and dehydrated before being coverslipped with DPX mountant for histology. Double-immunolabeling was examined with a Zeiss LSM 510 confocal laser scanning microscope (Carl Zeiss, Hamburg, Germany) by using 488 and 532 nm excitation laser light in the original thick slices. In order to suppress emission cross-talk, the microscope was configured to perform all scanning in sequential mode. Z-series (z-step of 0.38 µm) were scanned at x40 magnification with an oil-immersion lens. Neurite reconstruction and morphological classification of neurons were performed as previously described in the work of [39].

Data Analysis
IPSCs were analyzed offline using Clampfit 10.0 software (Axon Instrument, Molecular Devices, Sunnyvale, CA, USA) and Electrophysiology Data Recorder/Whole-Cell Analysis Program (WinEDR/WinWCP; Dr. J. Dempster). Events were detected and analyzed over a 10 min-long data segment using an automated low-amplitude (−10 pA, duration 3 ms) threshold detection algorithm. Each event was visually examined, and any noise that spuriously met trigger specifications was rejected. Several characteristics of the accepted events were analyzed: peak amplitude, instantaneous frequency, inter-event interval, and rise and decay tau (τ). A least-squares minimization algorithm was used to determine the decay τ of IPSCs. The decay phase of individual IPSCs was fitted (90%-10% of the peak amplitude) by either a monoexponential (y(t) = Ae−t/τ) or biexponential (y(t) = Afast (−t/τ fast) + Aslow(−t/τ slow)) function, where t is time, A amplitude, and τ the decay τ. An improvement of the fit by two exponentials compared to one resulted in a significant reduction in the standard deviation of the residuals, as confirmed by the use of the F-test. Traces containing multiple events were discarded, and only events that had stable baselines before the rise and after the end of the decay were kept for analysis. For IPSC averaging, the initial rising phases of the successive events were artificially aligned using software.
We chose the guinea pig anti-GABAAα2 receptor antibody for two reasons: firstly, because it tolerated the pepsin treatment, which was required for antigen retrieval, and secondly, because it was raised in a different species from the anti-GlyRα1 antibody. Although early in situ hybridization studies failed to detect mRNA for the GABA A α2 subunit in the spinal dorsal horn [41], in recent transcriptomic studies [42,43], the mRNA (Gabra2) has been found in many neurons in the dorsal horn of the mouse, including excitatory interneurons that express PKCγ. In addition, in situ hybridization studies using the very sensitive RNAscope method have revealed that many excitatory neurons in the superficial dorsal horn of the rat spinal cord contain Gabra2 mRNA [44,45].

GABAergic and Glycinergic Boutons on PKCγ Interneurons
Transverse sections (70 µm thick) through the MDH from the same 3 male Sprague-Dawley rats were immersed in 50% ethanol for 30 min to facilitate antibody penetration [46]. Immunofluorescent staining for the glycine transporter GlyT2, glutamic acid decarboxylase GAD (both isoforms), gephyrin, and PKCγ was carried out by incubating sections for 2 days at 4 • C in the following mixture of primary antibodies: goat anti-PKCγ (a gift from M Watanabe; 1:1000), rabbit anti-GlyT2 (a gift from F Zafra; 1:1000), mouse monoclonal anti-GAD65 (GAD6, Developmental Studies Hybridoma Bank, University of Iowa; 1:100), and mouse monoclonal anti-GAD67 (Merck, Watford, UK, catalog number mAb5406; 1:5000). These were revealed by overnight incubation in species-specific secondary antibodies raised in donkey and conjugated to Rhodamine Red, Alexa488, and Alexa647 (Jackson Immunoresearch), respectively. To avoid binding of the mouse secondary antibody to primary antibodies used in the subsequent step, a Fab' fragment of donkey anti-mouse IgG (conjugated to Alexa 647) was used. The sections were then incubated overnight in mouse monoclonal antibody against gephyrin (clone 7a), conjugated to the fluorescent dye Oyster550 (catalog number 147 011C3, Synaptic Systems; 1:500), rinsed, and mounted in anti-fade medium and stored at −20 • C.

Confocal Microscopy and Analysis
Sections were scanned with a Zeiss LSM710 confocal microscope with argon multi-line, 405 nm diode, 561 nm solid-state, and 633 nm HeNe lasers. For analysis of GABA A and glycine receptor expression, confocal image stacks consisting of 15-27 optical sections at 0.5 µm z-separation were acquired from the MDH of the 3 rats (6-9 scans per animal) by scanning through a 63× oil-immersion lens (numerical aperture 1.4). Sections were analyzed with Neurolucida for Confocal software (MBF Bioscience, Williston, VT, USA). Twenty-five PKCγ-immunoreactive (IR) cells were identified and selected in scans from each animal before the distribution of receptor staining was viewed. Staining for the GABA A α2 subunit was then viewed, and the locations of all immunoreactive puncta on the cell bodies of the selected PKCγ cells were noted. Staining for GlyRα1 was then viewed, and the presence or absence of this subunit at each GABA A α2 punctum was recorded. In addition, a search was performed for GlyRα1-positive puncta that lacked GABA A α2-immunoreactivity. A separate analysis was performed to determine whether the relationship between the two receptor subunits was similar in the dendrites of PKCγ neurons. For this analysis, we initially viewed the PKCγ and GABA A α2 channels and selected~100 GABA A α2-IR puncta that were located on PKCγ dendrites in sections from each of the 3 rats. We then viewed the GlyRα1 channel and noted the presence or absence of this type of immunoreactivity on each of the selected GABA A α2 puncta. We subsequently performed the reverse analysis by viewing PKCγ and GlyRα1 channels and selecting~100 GlyRα1-immunoreactive puncta that were on PKCγ dendrites from each rat. The GABA A α2 channel was then viewed, and the presence or absence of GABA A α2 immunostaining was noted for each GlyRα1 punctum.
To analyze the relationship between GAD, GlyT2, gephyrin, and PKCγ, the MDH of the 3 rats was scanned (between 1 and 6 sections were scanned through the x63 lens, z-step 0.3 µm). Gephyrin-IR puncta on the cell bodies or dendrites of PKCγ cells were initially identified (73-86 puncta per animal, mean 77.7). The channels corresponding to GADs and GlyT2 were then viewed, and the presence or absence of staining for each of these was noted for all of the selected gephyrin puncta.

Characterization of Antibodies
The guinea pig and goat antibodies against PKCγ were raised against amino acids 684-697 of the mouse protein and recognized a single protein band of 75 kDa on immunoblots (manufacturer's specification). The rabbit anti-PKCγ has been shown to stain identical structures to the guinea pig antibody [47]. The GABA A α2 antibody, raised against amino acids 1-9, recognizes a single band of 52 kDa on the Western blot of mouse cerebral cortex [48]. The GlyRα1 antibody was raised against a recombinant protein corresponding to amino acids 1 to 457 from rat glycine receptor α1 and is specific for the N-terminal 10 residues of this subunit [49]. The gephyrin antibody detects the brain-specific 93 kDa splice variant, and specificity has been verified in knock-out tissue (manufacturer's specification). The GlyT2 antibody was raised against amino acids 1-193 and detected a 90-110 kDa band in immunoblots [50]. Both monoclonal GAD antibodies are highly selective for the corresponding isoform [51] (manufacturer's specification).

Statistical Analysis
All statistical analyses were performed with the software GraphPad Prism (v7.1, La Jolla, CA, USA). Continuously distributed data were displayed either by showing all data points or by using box-and-whisker plots with all elements (mean, median, interquartile interval, minimum, maximum). Categorical variables were described by relative frequencies. The minimal sample size required to detect a standardized effect size of at least 20% with a power analysis >70% was calculated and respected as much as possible. The normality distribution of data was determined using Kolmogorov-Smirnov tests. The comparison between groups was made using one-way or two-way analysis of variance (ANOVA) followed by Tukey's post-hoc test for multiple comparisons. F-values were expressed with their associated degrees of freedom (df). Factors were designed as follows: IPSC type for comparisons between fast monoexponential, slow monoexponential, and biexponential mIPSC or sIPSC, Cell type for comparisons between PKCγ + and PKCγ, and Interaction for effect between two factors on the dependent variable. Grubbs' test was used to identify outlier data. The level of significance was set at p < 0.05. Statistical details for each quantitative experiment were illustrated in Tables 1 and 2  Interaction F(2,54) = 2.6 0.08 , and the amplitude (F) of the fast monoexponentially decaying, the slowly monoexponentially decaying, and the slowly biexponentially decaying mIPSCs; the τ values of the latter were obtained by forcing monoexponential fits. Data are presented as box-and-whisker plots depicting median, interquartile interval, minimum, and maximum. The "+" represents the medians, and the boxes present the quartiles. Data were averaged from 5 lamina II i PKCγ + interneurons. The comparison between groups was made using one-way ANOVA followed by Tukey's post-hoc test. * p < 0.05.

Electrophysiological Analysis
We recorded spontaneous IPSCs, in the presence (i.e., mIPSCs) or absence of TTX (i.e., sIPSCs), from lamina II i neurons under both CNQX (10 µM) and APV (40 µM) in the bath. To be included in this analysis, each recorded neuron had to be successfully filled with neurobiotin, be PKCγ phenotyped (Figure 1(A1)), and show a sufficiently high frequency of sIPSCs or mIPSCs to permit analysis. A total of 10 PKCγ + interneurons and 10 PKCγ − interneurons fulfilled such criteria.

Three Kinetically Distinct Populations of mIPSCs onto MDH Lamina II i PKCγ + Interneurons
It is now well established that GlyR-and GABA A R-mediated evoked IPSCs or mIPSCs recorded from lamina I-II neurons in the MDH [32] as well as SDH [27,32,33,35,37,52] can be distinguished on the basis of their decay kinetics, the decay τ of GlyR-mediated IPSCs being on average a factor of 2.4 [35], 3.6 [28], or 5.0 [27] faster than that of GABA A Rmediated IPSCs. Therefore, to assess the nature of inhibitory synaptic inputs onto MDH lamina II i PKCγ + interneurons, we first examined the decay kinetics of mIPSCs recorded from 5 PKCγ + interneurons. Heterogeneous kinetics of mIPSCs were identified ( Figure 1B). The decay time courses of most mIPSCs were best fitted by monoexponential functions. Close inspection of the distribution of the decay τ of these monoexponential mIPSCs revealed a heterogeneous population of decays, which were best fitted by the sum of two Gaussians ( Figure 1C). One population of monoexponential mIPSCs exhibited a fast decay phase (peak of decay τ: 8.9 ± 0.9 ms; Figure 1D). The other population of monoexponential mIPSCs had a slower peak of decay τ (31.7 ± 1.3 ms; Figure 1D). Notably, the decay τ of the fast monoexponential mIPSCs was, on average, a factor of 3.6 faster than that of the slow monoexponential mIPSCs. In addition, there was another population of mIPSCs with a slow decay phase, which required fitting with two exponentials. Biexponential mIPSCs yielded components that had a fast (τ 1 = 7.9 ± 3.3 ms) and a slow (τ 2 = 116.7 ± 40.2 ms) component (data not shown). The decay τ 1 was thus a factor of 14.8 faster than the decay τ 2 . These data show that, according to decay kinetics, three types of mIPSCs can be recorded from MDH lamina II i PKCγ + interneurons: fast monoexponential, slow monoexponential, and biexponential mIPSCs. The same decay kinetics were discriminated when sIPSCs were analyzed (Table 3). Data show mean ± SEM of decay τ (ms), rise τ (ms), instantaneous frequency (Hz), and amplitude (pA) of fast monoexponential, slow monoexponential, and biexponential miniature (mIPSCs) and spontaneous inhibitory postsynaptic currents (sIPSC) from (n) lamina II i PKCγ + or PKCγ − interneurons. Tau (τ) values were obtained by forcing monoexponential fits. The comparison between groups was made using two-way ANOVA followed by Tukey's post-hoc test. a p < 0.001 Fast monoexp compared to Slow monoexpo PKCγ + , b p < 0.05 Fast monoexp compared to Biexpo PKCγ + , and c p < 0.05 Slow monoexpo compared to Biexpo PKCγ + .~Not analyzed.
A comparison of the other functional properties of the fast monoexponential, slow monoexponential, and biexponential mIPSCs indicated that the three types of mIPSCs were not significantly different in rise time ( Figure 1E) or in amplitude ( Figure 1F). Finally, the three types of mIPSCs exhibited the same instantaneous frequencies (Table 3).

Co-Occurrence of Fast Monoexponential, Slow Monoexponential, and Biexponential Spontaneous IPSCs in All MDH Lamina II i PKCγ + Interneurons
We then asked whether all PKCγ + interneurons exhibit the three types of IPSCs: fast monoexponential, slow monoexponential, and biexponential IPSCs. Since the same three kinetically defined types of sIPSCs (n = 5) or mIPSCs (n = 5) could be recorded from MDH lamina II i PKCγ + interneurons (Table 3), we examined the proportions of each of them. On average, 34.5% ± 5.4% (fast monoexponential), 28.9% ± 4.2% (slow monoexponential), and 36.7% ± 3.3% (biexponential) of the total number of spontaneous IPSCs were displayed by MDH lamina II i PKCγ + interneurons ( Figure 2). These results thus indicate that all MDH lamina II i PKCγ + interneurons are bombarded by the three kinetically defined types of spontaneous IPSCs in rather equal proportions.

Co-Occurrence of Fast Monoexponential, Slow Monoexponential, and Biexponential IPSCs in MDH Lamina II i PKCγ + and PKCγ − Interneurons
MDH lamina II i PKCγ + and PKCγ − neurons are different according to membrane properties and excitatory synaptic inputs [26,39]. To test whether synaptic inhibitions onto the two cell phenotypes are also different, we compared the sIPSCs or mIPSCs recorded from 10 PKCγ + and 10 PKCγ − interneurons. The total proportions of fast monoexponential, slow monoexponential, and biexponential sIPSCs or mIPSCs were similar in both neuronal populations ( Figure 2). Nevertheless, there were differences in the total proportions of IPSCs when compared within PKCγ − interneurons (Figure 2). On the other hand, the PKCγ + (n = 5) and PKCγ − interneurons (n = 5) displayed the same kinetic properties, except for the instantaneous frequency of mIPSC that was significantly lower in PKCγ + interneurons compared to PKCγ − interneurons (Table 3). Altogether, these results thus suggest that the synaptic inhibitions onto lamina II i PKCγ + and PKCγ − interneurons are similar.

GABA A R-and GlyR-Mediated sIPSCs in MDH Lamina II i PKCγ + and PKCγ − Interneurons
We finally examined whether these sIPSCs were mediated by activation of GABA A Rs or GlyRs by using the selective GABA A R antagonist, bicuculline (10 µM), and GlyR antagonist strychnine (0.5 µM). Since sIPSCs and mIPSCs recorded from MDH lamina II i PKCγ + and PKCγ − interneurons displayed the same kinetic properties, the effects of bath-applied bicuculline and strychnine were assessed on sIPSCs recorded from four lamina II i interneurons (two PKCγ + and 2 PKCγ − interneurons) and result with PKCγ − interneurons were compared to those from PKCγ + interneurons (Figure 3). Bicuculline blocked a large population of sIPSCs (76.3% ± 1.3% for PKCγ + and 75.6% ± 1.3% for PKCγ − ), and the remaining bicuculline-insensitive sIPSCs were all abolished by the GlyR antagonist, strychnine, indicating that the latter sIPSCs were mediated via activation of GlyRs ( Figure 3A). Notably, all lamina II i neurons displayed bicuculline-sensitive as well as bicuculline-insensitive, strychnine-sensitive sIPSCs, implying that they all received both GABA A R-and GlyRmediated inhibition. The decay phase of individual GlyR-mediated sIPSCs could be fit by a monoexponential function (τ = 8.9 ± 0.9 ms). Therefore, these results suggest that GlyR-mediated sIPSCs represent a population of fast monoexponential sIPSCs, while slow monoexponential sIPSCs (τ > 20 ms) are mediated by GABA A R in both PKCγ + and PKCγ − lamina II i interneurons within the rat MDH. This is in agreement with previous findings obtained in MDH [32] and SDH lamina I-II neurons [27,32,33,35,37,52], where GlyR-and GABA A Rmediated IPSCs were shown to exhibit rapid and slow decay time courses, respectively.
We conclude that at least the vast majority of quantal events (about 70%) onto lamina II i PKCγ + as well as PKCγ − interneurons within the rat MDH are mediated by GlyR or GABA A R only, but not by both. Moreover, all lamina II i interneurons receive GlyR as well as GABA A R-only-mediated inputs.

Inhibitory Receptors and Synapses onto PKCγ + Interneurons
To assess the presence of GABA A R and/or GlyR onto lamina IIi PKCγ + interneurons, we labeled MDH sections with antibodies against the α2 subunit of GABA A R (GABA A Rα2), the α1 subunit of GlyR (GlyRα1), and PKCγ.
Triple-immunofluorescence showed that the soma of PKCγ + interneurons (25 PKCγ + interneurons/rat in three rats) received a high density of inhibitory synapses: the number of receptor-IR puncta that were identified on the somata of the PKCγ + cells ranged from 3 to 28 with an average of 11. In most PKCγ + interneurons (49 of 75 cells, 65%; Figure 4A,B), all puncta were labeled for both GABA A Rα2 and GlyRα1. Nevertheless, a small number of PKCγ + interneurons had only single-labeled puncta: either GlyRα1-only (5 of 75 cells, 7%) or GABA A Rα2-only (2 of 75 cells, 3%) puncta ( Figure 4B). Each of the remaining 19 cells displayed double-labeled puncta together with either GlyRα1-only (13 of 75 cells, 17%) or GABA A Rα2-only (6 of 75 cells; 8%) puncta ( Figure 4B). Pooling data from all 75 cells analyzed, we identified 843 receptor-IR puncta, of which 660 (78.3%) were positive for both GlyRα1 and GABA A Rα2, 164 (19.5%) were positive for only GlyRα1, and 19 (2.3%) were positive for only GABA A Rα2. Considering each subunit, we identified 824 GlyRα1 puncta, of which 660 (80%) were also GABA A Rα2-IR, and 679 GABA A Rα2 puncta, of which 660 (97%) were also GlyRα1-IR. Of note, we never observed singlelabeled GlyRα1 and GABA A Rα2 puncta on the same PKCγ interneuron. This suggests that most PKCγ + interneurons (68 of 75 cells; 91%) receive mixed GABA A R-GlyR inhibitory synapses. Some PKCγ + interneurons also receive GABA A R-only or GlyR-only synapses, alone or together with mixed GABA A R-GlyR synapses. Our analysis of PKCγ dendrites yielded quantitatively similar results. We identified a mean of 106.7 (range 102-115) GABA A Rα2 puncta on PKCγ dendrites, of which 96.9% (92.8-100%) were also GlyRα1-IR, and 104 (100-111) GlyRα1 puncta, of which 76% (57.3-96.1%) were also GABA A Rα2-IR. We also looked for appositions of GABAergic and glycinergic boutons onto lamina IIi PKCγ + interneurons. Analysis of MDH transverse sections that had been reacted with antibodies against GlyT2, GADs, gephyrin, and PKCγ revealed that both the somata and dendrites of PKCγ + interneurons had numerous gephyrin puncta, which represent the sites of inhibitory synapses ( Figure 4C) [47]. We analyzed between 73 and 86 (mean 77.7) gephyrin puncta that were on cell bodies or dendrites of PKCγ + cells in sections from the three rats. We found that 42.2% of these (range 37.8-47.9%) were adjacent to boutons that were positive for both GAD and GlyT2, while 49.5% (range 46.5-54.1%) were in contact with GAD-positive boutons that lacked GlyT2, and 6.3% (range 4.1-9.3%) were contacted by a bouton that was GAD-negative and GlyT2-positive. In the remaining 2.1% (range 0-3.5%) of cases, we did not observe structures with either GAD-IR or GlyT2-IR. This suggests that most inhibitory boutons that synapse on the PKCγ + cells release GABA, with around half also being glycinergic.

Discussion
We performed electrophysiology and immunochemistry to characterize synaptic inhibition onto lamina II i PKCγ + interneurons in the adult rat MDH. According to decay time courses, three distinct populations of sIPSCs or mIPSCs can be recorded from these lamina II i PKCγ + interneurons: fast monoexponential, slow monoexponential, and biexponential sIPSCs or mIPSCs. All three types of sIPSCs or mIPSCs are present in all recorded PKCγ + with rather similar frequencies. Moreover, the same quantal events are recorded from neighboring lamina II i PKCγ − interneurons, suggesting that lamina II i PKCγ + and PKCγ − interneurons display similar types of inhibitory inputs. The sensitivity of fast and slow quantal events to GABA A R and GlyR antagonists suggests that monoexponential sIPSCs or mIPSCs are mediated by GlyR and GABA A R, respectively. Altogether, our electrophysiological results suggest that inhibitory inputs onto lamina II i PKCγ + or PKCγ − interneurons in the adult rat MDH are predominantly (at least 70%) mediated by either GlyR or GABA A R but not both. On the other hand, immunohistochemistry reveals that the vast majority (~80%) of inhibitory synapses onto PKCγ + interneurons contain both GABA A Rs and GlyRs. We also show that the great majority (~90%) of inhibitory boutons presynaptic to the PKCγ + cells possess GAD and can therefore presumably release GABA, while just under half of these also express GlyT2 and are therefore likely to co-release glycine.
Two types of biexponential-evoked IPSCs and mIPSCs were previously recorded from lamina II neurons: pure GABA A R-mediated within the MDH [32] and mixed GABA A R/GlyR IPSCs (that is, combining two components, a fast decaying, GlyR-mediated and a slowly decaying, GABA A R-mediated) in the SDH [35,36]. The biexponential sIPSCs or mIPSCs that we recorded from MDH lamina II i PKCγ + interneurons at post-natal day (P) 21-35 are unlikely to be mixed GABA A R/GlyR IPSCs since these very IPSCs, in SDH lamina II neurons, are only present during early development and become virtually undetectable by approximately P23 [35,37]. They are rather GABA A R-mediated. First, similar biexponential GABA A R-mediated mIPSCs were already recorded in this very area in juvenile to adult rats [32]. Second, compared with slow monoexponential, GABA A R-mediated mIPSCs, biexponential mIPSCs exhibit larger amplitudes and longer decay τ of the slow component (here and in the work of [32]). Finally, bath application of bicuculline strongly reduced sIPSCs frequency (more than 2/3 reduction), consistent with the observation that GABA A R-mediated mIPSCs occur at a higher frequency than GlyR-mediated mIPSCs in MDH lamina II neurons [32]. The proportion of GlyR-only sIPSCs recorded from PKCγ + interneurons in the presence of bicuculline (~24%) is in the same range as that of kinetically isolated fast monoexponential IPSCs (~34%). Indeed, if our biexponential sIPSCs were mixed GABA A R/GlyR IPSCs, the proportion of GlyR-only sIPSCs in the presence of bicuculline should have been double; that is, the sum of the proportions of kinetically isolated fast monoexponential IPSCs and biexponential IPSCs. Therefore, our electrophysiological data strongly suggest that all quantal inhibition onto lamina II i PKCγ + interneurons in adult rat MDH is mediated by either GlyR or GABA A R but not by both types of receptors concurrently.
This conclusion holds true for lamina II i PKCγ − neurons as the same quantal events were recorded from PKCγ + and PKCγ − interneurons. Therefore, whereas these two classes of lamina II i interneurons exhibit differences in their membrane properties and excitatory inputs [26,39], they receive similar inhibitory inputs, at least in the adult rat MDH.
Altogether, our results are consistent with the conclusion that lamina II neurons in the adult rat MDH [32] as well as SDH [27,31,32,[34][35][36] exhibit quantal events mediated by GlyR and by GABA A R, but not by both types of receptors concurrently. However, whereas GABA A R-only IPSCs in the SDH are exclusively monoexponentially decaying, those in the MDH can be mono-as well as biexponentially decaying (present results; [32]). It is interesting to note that the decay τ of the fast component of biexponential mIPSCs is within the same range as that of GABA A R-mediated mIPSCs recorded from pyramidal neurons in the rat somatosensory cortex [53]. Such fast GABA A R-mediated mIPSCs might be related to the expression of the GABA A Rα1 subunit [54]. Thus, with respect to GABA A R-mediated inhibition, MDH would be intermediate between the SDH and higher brain areas.

PKCγ + Interneurons Co-Express GABA A R and GlyR at the Same Inhibitory Synapses
The present results establish that the subunits belonging to GABA A R and GlyR are colocalized at the great majority (~80%) of inhibitory synapses on PKCγ + cells and that nearly all (91%) of these cells have synapses at which the receptors are coexpressed. Nevertheless, some PKCγ + interneurons (9%) also displayed synapses with only GABA A Rs or GlyRs, which may reflect the existence of different subpopulations of PKCγ + interneurons [26,39].
However, we have to consider certain technical issues. We have previously demonstrated with confocal microscopy that virtually all GlyRα1-IR puncta in the rat spinal dorsal horn are also immunoreactive for gephyrin [30]. We have also shown that gephyrin puncta seen at the light microscope level are invariably associated with inhibitory axonal boutons identified by expression of the vesicular GABA transporter (which also transports glycine) [47] and that they correspond to the postsynaptic specialization of inhibitory synapses when viewed with electron microscopy [55]. It is, therefore, very likely that all of the GlyRα1 puncta that we observed in the membranes of PKCγ cells represent inhibitory synapses, and as we show here, most of these (80%) were also GABA A Rα2-IR. The GABA A Rα2 is also expressed by primary afferent axons, and so some of the immunostaining for this receptor might have been associated with primary afferent terminals. However, this is highly unlikely to account for any of the GABA A Rα2 puncta that we identified on PKCγ neurons because the vast majority of these puncta (98%) also contained the GlyRα1 subunit, which is not expressed by primary afferents [56][57][58].
Negative findings with immunohistochemistry (i.e., lack of immunoreactivity for one or other receptor subunit at individual puncta) could result from low levels of the receptor (rather than its absence) at a synapse. Indeed, "single-labeled" puncta were often quite weakly stained for either GABA A R or GlyR, so they may have had low levels of the other receptor that were below the detection threshold. Another possibility is that a synapse may only express other subunits for the receptor. We assessed the expression of GABA A Rα2 and GlyRα1 on lamina II i PKCγ + interneurons because these subunits are known to be abundant in SDH [59][60][61]. For the few cells that completely lacked GlyRα1, it is unlikely that they have functional glycine receptors (although they could presumably express GlyRα3) [62]. However, for cells that lack the GABA A Rα2 subunit, it is quite possible that they express another GABA A α subunit. It is thought that individual GABA A receptors may contain two different α subunits [44], and it is also possible that individual neurons may possess receptors that contain different subunits (e.g., α2 or α3). Thus, we might have underestimated the proportion of cells and synapses that express both GABA A R and GlyR.
Previous immunohistochemical studies showed that 30-50% of superficial SDH neurons are GABAergic, and about half of these are also enriched with glycine [63][64][65]. This is consistent with our finding that most inhibitory boutons presynaptic to PKCγ + interneurons stained for GAD, while around 40% of them co-expressed GlyT2.
Detection of synaptic receptors required pepsin treatment, which presumably exposes epitopes that are normally masked by fixation [40,66]. This treatment also destroys many other antigens [40], and to avoid this problem, we used tyramide signal amplification to reveal PKCγ. However, it is difficult to combine antigen unmasking and receptor immunostaining with the detection of more than other antigens, and so we were unable to directly compare the expression of presynaptic markers (GADs, GlyT2) with that of postsynaptic receptors on the PKCγ + cells.

Can Anatomical Data Account for the Functional Specialization of Inhibitory Synapses onto MDH Lamina II i PKCγ + Interneurons?
We can make certain predictions concerning the arrangement of neurotransmitters and receptors based on our anatomical finding that~80% of inhibitory synapses had both GABA A Rα2 and GlyRα1 subunits and that >90% of presynaptic boutons were GAD + , whereas only around half were GlyT2 + . Thus, many inhibitory synapses onto PKCγ + interneurons may have both types of receptors but only involve GABA release. The few synapses that only express GABA A Rs should also operate as purely GABAergic. Similarly, inhibitory synapses that have only GlyRs should function as purely glycinergic.
However, GlyR-only inhibitory synapses, as revealed by immunohistochemistry, are unlikely to account for the 34% of GlyR-only-mediated sIPSCs or mIPSCs, even though such synapses would display high release probabilities. Moreover, our anatomical data suggest that, in numerous inhibitory synapses onto PKCγ + interneurons, GABA and glycine are co-released, and GlyR and GABA A R are both present on the postsynaptic membrane. One would therefore expect mixed GABA A R/GlyR-mediated mIPSCs to be recorded. However, our electrophysiological data show that, in normal conditions, lamina II i PKCγ + interneurons in adult rat MDH are exclusively bombarded by GABA A R-only-and GlyR-only-mediated IPSCs, similar to lamina II neurons in young rat SDH [27,31,32,[34][35][36]. Recently, while recording ex vivo from lamina II i PKCγ + interneurons in 6-8-week-old mice, Wang et al. [26] observed not only pure Aβ fiber-evoked GlyR-mediated IPSCs in most neurons (7 out of 12) but also mixed Aβ fiber-evoked GABA A R-GlyR-mediated IPSCs in the other five. However, in this report, dorsal roots were electrically stimulated, and therefore, many Aβ fibers were activated simultaneously: the few mixed GABA A R-GlyR-mediated IP-SCs might thus result from summated Aβ fiber-evoked pure GABA A R-and GlyR-mediated IPSCs. Altogether, this suggests that some synapses can function as purely GABAergic or glycinergic even though presynaptic buttons co-release GABA/glycine and postsynaptic membranes contain both GABA A and glycine receptors. One possibility is that GABA A R and GlyR are differentially located within inhibitory synapses. Indeed, bath application of a benzodiazepine [27] or inflammatory conditions [37] leads to the unmasking of a longduration GABA A R-mediated component and the appearance of mixed GABA A R/GlyR mIPSCs in adult SDH lamina II neurons. This might stem from a different localization of GABA A R and GlyR in inhibitory synapses onto SDH neurons, with GABA A Rs being extrasynaptically distributed [27]. However, our anatomical data in adult rat MDH are not consistent with this hypothesis, as the two types of receptors were co-extensive at presumed inhibitory synapses. A more likely explanation is that although there is co-expression of the GABA A Rα2 and GlyRα1 subunits at these synapses, one or another type of receptor is not functional, perhaps related to the differential expression of other subunits.
The present finding of GABA A R-only-and GlyR-only-mediated sIPSCs or mIPSCs recorded from PKCγ + interneurons suggests separate roles for these two inhibitory systems. There are two clinical forms of mechanical allodynia: dynamic and static. Whereas both types are impacted by modulating PKCγ activity [5,6,14,16], glycinergic disinhibition produces specifically dynamic mechanical allodynia [5] and GABA A ergic disinhibition, a static one [6,22]. How can mechanical hypersensitivity depend on the type of disinhibition? LTMRs transmitting the dynamic (Aβ-LTMRs) and static (Aδ-LTMRs) mechanical inputs to PKCγ + interneurons might be gated by specifically glycinergic and GABA A ergic feedforward inhibitory microcircuits, respectively (see Figure 1 in the work of [9]). There is already evidence for feedforward glycinergic circuits preventing the activation of PKCγ + interneurons and superficial nociceptive circuits by Aβ inputs, which is reduced following nerve injury [20,26]. It is, therefore, possible that differential modulation of glycinergic and GABA A ergic feedforward inhibitions onto PKCγ + interneurons might be key determinants of the manifestation of dynamic and static mechanical allodynias, respectively.

Conclusions
Although our anatomical data suggest that GABA A R and GlyR are colocalized at most inhibitory synapses on these cells and that GABA and glycine are often co-released, PKCγ + interneurons exclusively displayed GABA A R-only-and GlyR-only-mediated spontaneous IPSCs. Behavioral evidence suggests that this specialization of inhibitory inputs onto PKCγ + interneurons is functionally relevant.
Data represent F-values for one or two-way ANOVA with corresponding degrees of freedom (DFn, DFd) and P-values. Factors for ANOVA were designed as follows: Cell type, for comparisons between PKCγ+ and PKCγ-; IPSC type for comparisons between fast monoexponential, slow monoexponential, and biexponential mIPSC; and Interaction, for effect between two factors on the dependent variable.