Pain Inhibits GRPR Neurons via GABAergic Signaling in the Spinal Cord

It has been known that algogens and cooling could inhibit itch sensation; however, the underlying molecular and neural mechanisms remain poorly understood. Here, we show that the spinal neurons expressing gastrin releasing peptide receptor (GRPR) primarily comprise excitatory interneurons that receive direct and indirect inputs from C and Aδ fibers and form contacts with projection neurons expressing the neurokinin 1 receptor (NK1R). Importantly, we show that noxious or cooling agents inhibit the activity of GRPR neurons via GABAergic signaling. By contrast, capsaicin, which evokes a mix of itch and pain sensations, enhances both excitatory and inhibitory spontaneous synaptic transmission onto GRPR neurons. These data strengthen the role of GRPR neurons as a key circuit for itch transmission and illustrate a spinal mechanism whereby pain inhibits itch by suppressing the function of GRPR neurons.


Characterization of GRPR neuron membrane properties.
To characterize the properties of GRPR neurons, electrophysiological recordings were obtained from a total of 230 GRPR-eGFP neurons in the spinal cord slices from P16-P25 mice. Action potential firing patterns were determined from a sample of 39 GRPR neurons, by recording, in current clamp, the responses to injections of depolarizing current. Most neurons (56.4%) exhibited a delayed firing pattern, when current steps were applied from a membrane potential of about −80 mV ( Fig. 2A-D). This pattern is characterized by a delay in the generation of the first action potential, that is larger than the average interspike interval (Fig. 2B). Other subpopulations of GRPR neurons showed a tonic (23.1%) or a phasic (15.4%) firing pattern ( Fig. 2A-D). The tonic pattern is characterized by an action potential discharge that persists during the whole current step and often decreases in frequency. The delay of the first action potential is comparable to the average interspike interval (Fig. 2B). Neurons showing the phasic pattern fire only at the beginning of the current step, with a variable number of action potentials. Only two neurons exhibited a single spike pattern. Similar results have been recently reported 31 , showing a prevalence of the delayed firing pattern in GRPR neurons recorded at their resting potential (about −73 mV).
A recent study showed that GRPR neurons predominantly exhibit a tonic firing pattern at their resting potential 20 . Accordingly, almost half of the delayed firing neurons at −80 mV exhibited a tonic pattern when maintained in current clamp at −60/−65 mV ( Supplementary Fig. 2). Voltage clamp recordings performed from delayed firing neurons showed the presence of a transient, voltage-dependent A current, that was activated by holding the cells at negative potentials (−100 mV) and applying depolarizing voltage steps (Fig. 2C). Activation of the A current is likely responsible for the delayed firing pattern observed in most GRPR neurons at −80 mV, as shown in previous studies 32, 33 .
To confirm the expression and functionality of GRPR on GRPR neurons, we tested the response of 22 neurons to GRP. Application of 1 µM of GRP induced a slow inward current in 73% (16/22) of GRPR neurons, held at −50 mV in voltage clamp. A second application of GRP, performed on a subpopulation of 7 responsive neurons, elicited in 3 cells an inward current of smaller amplitude (11.5 ± 6.8 pA versus 22.6 ± 10.1 pA at the first GRP application). The suppression of an inward rectifier K + current and/or the activation of a non-selective cation conductance could contribute to the GRP-generated current ( Supplementary Fig. 3) 34 .

GRPR neurons receive direct inputs from primary afferents.
Previous studies showed that abundant GRP fibers are present in the dorsal horn 9,35,36 and Immuno-EM studies confirmed that GRP fibers form contacts with dendrites of GRPR neurons 37 . To characterize the primary afferent fibers synapsing onto GRPR neurons, we stimulated the dorsal root attached to the slice and recorded evoked excitatory postsynaptic currents (EPSCs) from GRPR neurons (Fig. 3). Stimulus intensities were determined in separate sets of experiments, by stimulating one end of the dorsal root and recording compound action potentials from the other end (n = 8, Fig. 3A). Intensities of stimulation ranged between 10 and 25 μA for Aβ fibers, 25 and 100 μA for Aδ fibers, and between 200 and 500 μA for C fibers. These values are consistent with those reported by previous studies performed in mice of comparable age 38 .
Of 54 GRPR neurons, the majority of cells received synaptic input from both Aδ and C fibers (n = 36, 66.8%), while a small proportion exhibited EPSCs mediated by Aδ or C fibers only (n = 11 and 5, respectively; Fig. 3B,C,E). EPSCs mediated by Aδ were mostly polysynaptic (n = 38, 81%), while almost half of the C fiber mediated responses were monosynaptic (n = 18, 43.1%). Aβ fiber mediated EPSCs (of polysynaptic nature), evoked at intensities lower than 25 µA, were observed in only 2 of the 54 neurons tested, indicating that GRPR neurons mainly receive primary afferent inputs recruited by high intensity stimulation. Other 4 cells, showing polysynaptic EPSCs evoked at 25 µA, were classified as receiving Aδ inputs, although at this stimulus intensity some additional Aβ fibers could have been also recruited (Fig. 3A). Being polysynaptic connections, it was not possible to apply the high frequency stimulation protocol to distinguish the 2 types of fibers. EPSCs evoked on GRPR neurons by dorsal root stimulation were completely blocked by co-application of the AMPA receptor antagonist NBQX and the NMDA receptor antagonist D-APV, showing that they are mediated by glutamate receptors (n = 10, Fig. 3D). Stimulation of Aδ and C fibers was also able to evoke inhibitory postsynaptic currents (IPSCs) on GRPR neurons held at −10 mV (Fig. 4A). Similar to the EPSCs, most neurons exhibited evoked IPSCs mediated by both Aδ and C fibers (15 out of 25 neurons tested, Fig. 4B). Neurons showing IPSCs mediated by Aδ fibers include also 3 cells where IPSCs were evoked at 25 µA, an intensity also compatible with Aβ stimulation (see www.nature.com/scientificreports www.nature.com/scientificreports/ above). Application of bicuculline produced more than a 90% block of the evoked IPSCs, showing that they were mainly mediated by GABA A receptors (n = 10, Fig. 4C).

Inhibition of GRPR neurons by pain and cooling.
The observation that evoked IPSCs on GRPR neurons are mediated by primary afferents stimulated at high intensity suggests that these neurons are inhibited by nociceptive inputs. To test this hypothesis, we examined the effect of capsaicin, allyl isothiocyanate (AITC, a key component of mustard oil) and menthol on spontaneous IPSCs (sIPSCs) recorded from GRPR neurons. Upon application of menthol (500 µM), we observed a significant increase of sIPSC frequency in 26.3% of GRPR neurons, with an average 3.3-fold increase (Fig. 5A,B,I). AITC (500 µM) significantly increased the sIPSC frequency in 21.6% of neurons, with an average 5.3-fold increase (Fig. 5C,D,I). An increase of sEPSC frequency was observed in one neuron out of 13 in menthol and in one neuron out of 14 in AITC (Fig. 5J). Remarkably, application of capsaicin (1 µM) affected both sEPSCs and sIPSCs of GRPR neurons ( Fig. 5E-J), with a potent effect (23.5-fold increase) on sEPSC frequency in the large majority of cells tested (91.6%). By contrast, sIPSC frequency was increased 4.6-fold in 34.3% of the recorded neurons. These results confirm that pain and cooling can activate inhibitory spinal interneurons, producing an increase of the inhibitory tone in GRPR neurons.

Discussion
In this study, using classic neuroanatomical tracing and immuno-EM approach we demonstrate that GRPR neurons are excitatory interneurons that make contacts with NK1R PBN-and STT-projecting neurons, supporting earlier studies 13,14 . Electrophysiological studies show that the majority of GRPR neurons exhibit a delayed firing pattern at hyperpolarized potentials, typical of dorsal horn excitatory interneurons 33,39 , which is consistent with the study by Aresh et al. 20 , showing that the majority of GRPR neurons express Vglut2 mRNA.
Importantly, we have characterized, for the first time, the primary afferent fibers mediating excitatory and inhibitory synaptic transmission onto GRPR neurons. GRPR neurons receive both direct and indirect high threshold C/Aδ excitatory inputs from primary afferents, in line with earlier studies in primates and humans showing that primary afferent pruriceptors are high threshold C/A fibers [40][41][42][43] . However, it is surprising that we failed to find significant Aβ inputs onto GRPR neurons using transverse spinal cord slices, a preparation that has www.nature.com/scientificreports www.nature.com/scientificreports/ been shown to be suitable for studying mono-and polysynaptic responses mediated by Aβ fibers 44,45 . However, it is possible that synaptic circuits activating superficial GRPR neurons (sometimes selected because of the stronger fluorescent signal) were not preserved in our preparation, thereby contributing to an underestimation of the amount of Aβ inputs received by GRPR neurons in the present study. The fact that EPSCs evoked by the dorsal root stimulation were blocked by NBQX and D-APV demonstrates that fast glutamatergic transmission constitutes an integral mechanism for relaying itch information from primary afferents to GRPR neurons. While a high concentration of GRP can directly evoke spikes on GRPR neurons 20 , at a lower dose that is of more physiological relevance, GRP could only depolarize cells, suggesting that GRP in vivo may modulate glutamatergic transmission under normal physiological condition 16,46 . GRP may modulate an indirect glutamatergic transmission from NMBR neurons to GRPR neurons 46 , rather than primary afferent-dependent glutamatergic input, which is not required for nonhistaminergic itch transmission 47,48 . It is worth noting that spinal neurons expressing Grp mRNA has been suggested to be important in itch transmission through activation of natriuretic peptide receptor A (NPRA) by B-type natriuretic peptide released from sensory neurons 49 . Using intersectional spinal restricted ablation of Grp neurons, we found that spinal Grp neurons are dispensable for itch transmission 12 . Equally important is the finding that conditional knockout of Grp in sensory neurons revealed the deficit in nonhistaminergic but not histamine itch 12 , consistent with earlier studies 9,10,46 .
Another novel finding is that inhibitory synaptic transmission onto GRPR neurons is mediated by high threshold Aδ/C fibers. The observation that over 90% of evoked IPSCs were blocked by bicuculline indicates that fast GABAergic transmission is necessary and sufficient to mediate inhibition of GRPR neurons. These results are consistent with recent studies showing anti-itch effects of GABA, a major neurotransmitter to inhibit laminae I-II excitatory interneurons located by GABA 50 , in various itch conditions 51,52 . One important question arising from the present study is which subset of inhibitory neurons release GABA to inhibit GRPR neurons. One candidate is SST2A inhibitory neurons whose inhibition by exogenous somatostatin (SST) may result in disinhibition of itch transmission 26,53,54 . It is possible that SST2A interneurons could be activated by painful stimuli, since mustard oil, capsaicin and menthol increase glutamatergic transmission on these neurons 26 .
By directly recording the responsiveness of GRPR neurons, we show that the same counterstimuli increase spontaneous inhibitory currents on many GRPR neurons. Thus, our data represent the first evidence that GABAergic interneurons, innervated by sensory neurons responsive to counterstimuli, are synaptically connected to GRPR neurons. On the other hand, since SST2A neurons are also required for pain inhibition and i.t. SST-elicited scratching reflects both itch and pain, the question remains as to whether there is dedicated neuronal pathway for itch inhibition. In addition, future studies are necessary to determine which GABA subtype receptors are expressed in GRPR neurons 55 .
Peripherally administered mustard oil (AITC) can induce both pain and itch in a dose-dependent manner 56 . Application of AITC on lamina II neurons in slices potentiates both spontaneous and miniature EPSCs 57 . In our experiments, AITC seems to act predominantly on nociceptive afferents, increasing inhibition on GRPR neurons. Excitatory responses recorded from GRPR neurons were not significantly affected by AITC, suggesting a lack of effect of mustard oil on central terminals of pruriceptive fibers. Topical application of capsaicin can induce a mix of pain and itch sensations 58,59 . Interestingly, we found that capsaicin exhibited dual effect on GRPR neurons: while inhibiting some, it also activates GRPR neurons. These results suggest that capsaicin induces pain by activating non-MrgprA3 TRPV1 nociceptors 60 , which in turns inhibits GRPR neurons, while concurrently inducing itch by activating TRPV1 in pruriceptors 61 and/or MrgprA3 neurons, which mediates itch via innervating GRPR neurons 60 . Coupled with previous studies, GRPR neurons receive at least three different kinds of inputs from primary afferents: a direct input from C/Aδ pruriceptors, an indirect glutamatergic input from NMBR neurons 30 , and an indirect inhibitory input from GABAergic neurons, in part mediated by Vglut2-expressing primary afferents 47,48 . Therefore, there are several modes of action of GRPR neurons, whose activation or inhibition could underlie the responses of itch behavior mediated by various counter-stimuli. Immunohistochemistry. The procedures were described previously 37,62 . Deeply anesthetized mice (ketamine, 90 mg/kg and Xylazine, 10 mg/kg) were perfused transcardially with 0.01 M PBS (PH 7.4) and paraformaldehyde (PFA) (4% in PBS). Spinal cord and brain were removed and post-fixed in 4% PFA for 2-4 h. The tissues were then cryoprotected in 20% sucrose overnight at 4 °C. Free-floating frozen sections were incubated with 2% donkey serum and 0.3% Triton X-100 for 1 h at room temperature followed by incubation with primary antibodies overnight at 4 °C. The sections were then washed and incubated with secondary antibodies for 2 h at room temperature. The following primary antibodies were used: chicken anti-GFP (1:500, Aves Labs, GFP-1020), guinea-pig anti-NK1R (1:500, AB15810, EMD Millipore), rabbit anti-FG (1:5000, AB153, Millipore). The following secondary antibodies were used: Alexa-Fluor 488 conjugated donkey anti-chicken (1:1000, Jackson ImmunoResearch, 703-545-155), Cy3-conjugated donkey anti-rabbit (1:1000, Jackson ImmunoResearch, 711-165-152) and Cy5-conjugated donkey anti-guinea pig 1(:1000, Jackson ImmunoResearch, 703-175-148). Fluorescent Images were taken using a Nikon C2+ confocal microscope system (Nikon Instruments, Inc.). www.nature.com/scientificreports www.nature.com/scientificreports/ Immuno-electron microscopy. To observe the connections between GRPR+ neurons and FG retrograde labelled PBN projection neurons in the spinal dorsal horn, immuno-electron microscopic studies were performed as previously described 37 . Briefly, for GRPR/FG double staining, cross sections of lumbar spinal cord of adult GRPR-eGFP mice were double immune-labeled by chicken anti-GFP antibody (1:500; Aves Labs) and rabbit anti-FG (1:5000, AB153, Millipore) using immunogold-silver method and immunoperoxidase method, respectively. Further, 50-nm-thick ultrathin sections were cut and examined with a JEM-1400 electron microscope (JEM, Tokyo, Japan). The digital micrographs were captured by VELETA (Olympus,Tokyo, Japan).
Excitatory or inhibitory currents (EPSCs or IPSCs) were evoked by stimulating the dorsal root attached to each slice using a suction electrode. Stimulus duration was 0.1 ms, stimulus intensities were determined by performing extracellular recordings of compound action potentials from the dorsal root (see Results and Fig. 5A). Monosynaptic vs polysynaptic EPSCs were identified following the procedure described by Torsney and  Active and passive membrane properties were determined by applying current steps (10 or 20 pA of amplitude) in current clamp, starting from a membrane potential of −60/−65 mV. Membrane resistance was calculated from responses to hyperpolarizing pulses, rheobase was defined as the current step amplitude able to elicit the lowest number of action potentials 64 . Action potential threshold was determined for the first action potential evoked at rheobase, at the point where the increase of membrane potential exceeds 20 mV/ms. Drugs were bath-applied for 1 min. All drugs were obtained from Sigma-Aldrich (Saint Louis, USA), except for GRP, that was purchased from Genscript (Piscataway, USA), and tetrodotoxin (TTX) from Tocris (Bristol, UK). Data were analyzed off-line using pClamp10 or MiniAnalysis (Synaptosoft, Decatur, USA). Graphs were obtained using Sigmaplot 11 (Systat software, San Jose, USA).

Experimental design and statistical analysis.
Electrophysiology experiments were performed on spinal cord slices from GRPR eGFP mice of either sex. A total of 92 mice have been used. Typically, 2-3 viable slices, with well-preserved synaptic connections, were obtained from the lumbar segment of the spinal cord (L3-L5). One to three neurons were recorded from each slice (only one cell if drugs were applied). Only neurons showing stable recording conditions (constant series resistance and membrane potential) were included in this study. Sample sizes were established according to similar studies of firing pattern and synaptic input characterization 33,38 . www.nature.com/scientificreports www.nature.com/scientificreports/ Statistical tests are indicated in figure legends when performed. Values are reported as the mean ± standard error of the mean (SEM). Statistical analyses were performed using Prism 7 (v7.0c, GraphPad, San Diego, CA) or Sigmaplot 11. Normality and equal variance tests were performed for all statistical analyses. P < 0.05 was considered statistically significant.