Spinal neuropeptide Y Y1 receptor-expressing neurons are a pharmacotherapeutic target for the alleviation of neuropathic pain

Significance

Neuropathic pain is a debilitating chronic pain condition that is poorly responsive to currently available analgesic drugs. A widespread clinical need remains for novel, safe, and efficacious nonopioid therapeutics for afflicted patients. Using a multidisciplinary approach, we provide evidence that hyperexcitability of a subclass of glutamatergic spinal cord dorsal horn interneurons, those that express the inhibitory G-protein-coupled (G_i/o) neuropeptide Y Y1 receptor (Y1-INs), drives neuropathic pain. Conversely, we demonstrate that inhibition of Y1-INs attenuates the behavioral signs of neuropathic pain. Our findings promote spinally directed Y1 receptor agonists as a promising future therapeutic strategy for the treatment of neuropathic pain.

Abstract

Peripheral nerve injury sensitizes a complex network of spinal cord dorsal horn (DH) neurons to produce allodynia and neuropathic pain. The identification of a druggable target within this network has remained elusive, but a promising candidate is the neuropeptide Y (NPY) Y1 receptor-expressing interneuron (Y1-IN) population. We report that spared nerve injury (SNI) enhanced the excitability of Y1-INs and elicited allodynia (mechanical and cold hypersensitivity) and affective pain. Similarly, chemogenetic or optogenetic activation of Y1-INs in uninjured mice elicited behavioral signs of spontaneous, allodynic, and affective pain. SNI-induced allodynia was reduced by chemogenetic inhibition of Y1-INs, or intrathecal administration of a Y1-selective agonist. Conditional deletion of Npy1r in DH neurons, but not peripheral afferent neurons prevented the anti-hyperalgesic effects of the intrathecal Y1 agonist. We conclude that spinal Y1-INs are necessary and sufficient for the behavioral symptoms of neuropathic pain and represent a promising target for future pharmacotherapeutic development of Y1 agonists.

Nociception serves as a danger signal to prevent tissue damage and promote survival (1). However, peripheral nerve damage can lead to pathological allodynia (normally innocuous sensory input is amplified and conveyed as painful), debilitating spontaneous pain, and affective comorbidities such as anxiety and depression (2–4). These features of neuropathic pain are poorly responsive to analgesic drugs (5–8), necessitating the need for new pharmacological targets. Such targets are likely to be found in the spinal cord dorsal horn (DH) (9–12) as peripheral nerve injury increases the intrinsic excitability of DH neurons (13, 14) and disinhibits excitatory (glutamatergic) DH neurons (15–17) to produce allodynia (18–22). Key populations of excitatory neurons that mediate allodynia include those that express protein kinase C gamma (PKCγ) (18–20, 23–27), somatostatin (Sst) (28, 29), cholecystokinin (CCK) (20, 30), neurokinin-1 receptor (NK1R) (31), and transient vesicular glutamate transporter 3 during development (tVGLUT3) (32, 33); however, these neural subpopulations do not represent druggable pharmaceutical targets. By contrast, pharmacological agonism at excitatory interneurons that express a neurotransmitter receptor coupled to inhibitory G proteins (G_i/o), such as the neuropeptide Y Y1 receptor, should reduce pronociceptive signaling. Indeed, application of NPY Y1-selective agonists to spinal cord slices reduces the excitability of NPY Y1 receptor-expressing interneurons (Y1-INs) and decreases pronociceptive signaling (34–36). In this study, we use a multipronged approach including [³⁵S]GTPγS binding in spinal cord slice, in vivo spinal cord pharmacology, optogenetics, chemogenetics, conditional genetic knockout mice, and ex vivo slice electrophysiology in the spared nerve injury (SNI) model of neuropathic pain. We demonstrate that spinal Y1-INs facilitate allodynia and mediate the anti-hyperalgesic effects of intrathecally-administered NPY. These results promote Y1-INs as a promising pharmacotherapeutic target for the treatment of neuropathic pain with Y1-selective agonists.

Results

Nerve Injury Increases the Efficiency of Coupling between NPY Y1 Receptors and G Proteins.

A multitude of G protein-coupled receptors (GPCRs) have been targeted for the development of new analgesic drugs (37, 38). However, peripheral injury can alter both the efficacy and potency of GPCR-agonist interactions (39–43). To determine whether nerve injury changes NPY-Y1-G protein signaling, we assessed guanosine-5′-O-(3-[³⁵S]thio)triphosphate ([³⁵S]GTPγS) binding in lumbar spinal cord slices 14 d after SNI, a common animal model of neuropathic pain (Fig. 1A), or sham surgery. Nerve injury is associated with two phases of behavioral hypersensitivity: a development phase (1 to 7 d post-SNI) and a maintenance/establishment phase (8+ d post-SNI) (44). We performed our studies at the well-established 14 d or 21 d post-SNI time points (15, 30, 45) to study Y1 as a potential therapeutic target for established neuropathic pain. As illustrated in Fig. 1 B–D and SI Appendix, Fig. 1, the NPY Y1 receptor-selective agonist, [Leu³¹, Pro³⁴]-NPY, increased [³⁵S]GTPγS binding with a maximum physiological effect (E_max) and effective concentration (EC₅₀) of 37.90 ± 1.32% and 2.80 ± 0.31 nM in the left DH and 37.40 ± 1.33% and 2.87 ± 0.32 nM in the right DH of sham-14 d mice. In SNI-14 d mice, the EC₅₀ of [Leu³¹, Pro³⁴]-NPY was reduced in the left (ipsilateral to injury, 1.84 ± 0.23 nM) but not right (contralateral, 2.56 ± 0.57 nM) DH compared to sham surgery, respectively. SNI did not change [Leu³¹, Pro³⁴]-NPY-stimulated E_max in the ipsilateral (40.16 ± 1.91%) or contralateral (41.28 ± 2.81%) dorsal horns compared to sham surgery (Fig. 1 C and D). These EC₅₀ data indicate that Y1 not only retains its G protein activation capacity in the setting of nerve injury, but also that nerve injury increases the efficiency of coupling between NPY Y1 receptors and G proteins, possibly to increase the analgesic potential of NPY-Y1 selective agonists at the DH.

Fig. 1.

Intrathecal administration of a Y1 agonist alleviates behavioral and immunohistochemical markers of SNI-induced neuropathic nociception. (A) Schematic representation of the SNI model of neuropathic pain. (B) Representative pseudocolor images of [³⁵S]GTPγS binding quantitative autoradiography in mouse spinal cord sections. Binding assays were performed in the absence (basal) or presence of 45 µM [Leu³¹, Pro³⁴]-NPY in sections obtained from mice given sham surgery (nerve exposure but no ligation) or SNI 14 d before the experiment. (C) SNI did not change the E_max of [Leu³¹, Pro³⁴]-NPY-stimulated [³⁵S]GTPγS binding (n = 5 to 10 animals/group). Student’s unpaired two-tailed t test. (D) SNI reduces the EC₅₀ of [Leu³¹, Pro³⁴]-NPY-stimulated [³⁵S]GTPγS binding in the ipsilateral (Left) DH (n = 5 to 10 animals/group). Student’s unpaired two-tailed t test. (E) Experimental timeline for SNI, intrathecal pharmacology, and evoked/reflexive mechanical (von Frey) and cold (acetone droplet withdrawal) behavioral testing. (F and G) Intrathecally administered [Leu³¹, Pro³⁴]-NPY (0.1 µg, 1.0 µg, or 10.0 µg) dose-dependently reduces SNI-induced mechanical and cold allodynia (n = 8 to 13 mice/group). Two-way repeated measure (RM) ANOVA. Holm-Sidak post hoc test. (H and I) The Y1 (BIBO 3304, 10.0 µg) but not Y2 receptor (BIIE 0246, 5.0 µg) antagonist abolishes Y1 agonist-induced ([Leu³¹, Pro³⁴]-NPY, 10.0 µg) mechanical and cold anti-allodynia (n = 4 to 11 mice/group). Two-way RM ANOVA. Holm-Sidak post hoc test. (J and K) Intrathecally administered [Leu³¹, Pro³⁴]-NPY (10.0 µg) reduced SNI-induced mechanical and cold hypersensitivities in an analgesic manner at early (3 wk post-SNI) but anti-hyperalgesic fashion at later (13 wk post-SNI) time points of neuropathic pain (n = 6 to 8 mice/group). Two-way RM ANOVA. Holm-Sidak post hoc test. (L) Experimental timeline for SNI, intrathecal pharmacology, light brush of the lateral hindpaw, and immunohistochemical staining for pERK immunoreactivity in the lumbar DH. (M–P) Representative images of ipsilateral DH light-touched evoked pERK immunoreactivity after intrathecal administration of agents. (Q) Intrathecally administered [Leu³¹, Pro³⁴]-NPY (10.0 µg) reduces light-touched evoked pERK immunoreactivity in the ipsilateral DH and this effect is abolished with coadministration of the Y1 antagonist, BIBO 3304 (10.0 µg) (n = 4 to 6 mice/group). Two-way ANOVA. Holm-Sidak post hoc test. Data are shown as means ± SEM. *P < 0.05, **P < 0.01, ***P < 0.001, ****P < 0.0001. Dots represent data points from individual animals.

A Y1 Selective Agonist Acts at DH Interneurons, Rather Than the Central Terminals of Primary Afferent Neurons, to Inhibit Mechanical and Cold Allodynia.

Intrathecal administration of NPY Y1 selective agonists attenuate DH neuron activation and neuropathic allodynia (46–48). However, NPY Y1 receptors are expressed on both excitatory interneurons in the DH and on the central terminals of primary afferent neurons arising from the dorsal root ganglion (DRG) (41, 49–51). To resolve the specific site of antihyperalgesic action by intrathecal Y1 agonists, we first extended previous results in the rat (47, 48), to our mouse SNI model of peripheral nerve injury. Fourteen days after SNI, rodents exhibit long-lasting mechanical and cold hypersensitivity in the sural innervation territory of the afflicted hindpaw (lateral surface) (47, 52–54). At this time point, intrathecal administration of [Leu³¹, Pro³⁴]-NPY (Fig. 1E) dose-dependently reduced both SNI-induced mechanical and cold hypersensitivity in male and female mice (Fig. 1 F and G). These antihyperalgesic effects were abolished with coadministration of the Y1 receptor antagonist, BIBO 3304, but not the Y2 receptor antagonist, BIIE 0246 (Fig. 1 H and I), suggesting on-target agonist binding at the Y1 receptor. Additionally, intrathecal administration of [Leu³¹, Pro³⁴]-NPY increased mechanical thresholds at both early (3 wk post-SNI) and later time points (13 wk post-SNI) after the induction of nerve injury (Fig. 1 J and K). [Leu³¹, Pro³⁴]-NPY increased thresholds in the sham group as well at the 3 wk time point, suggestive of an analgesic effect, but this effect was gone by 13 wk. (Fig. 1 J and K). Light brush of the injured hindpaw of SNI mice induced phosphorylated extracellular signal-regulated kinase (pERK) in ipsilateral but not contralateral superficial DH neurons, a proxy for neuronal activation (55). [Leu³¹, Pro³⁴]-NPY reduced light touch-evoked pERK expression in superficial DH neurons and this effect was abolished with coadministration of BIBO 3304 (Fig. 1 L–Q). These results indicate that Y1 agonism at the spinal cord potently inhibits behavioral and immunohistochemical markers of neuropathic pain in mice.

Next, we crossed Npy1r^loxP/loxP mice (56) with either Pirt^Cre (57) or Lbx1^Cre mice (58) to selectively knockout Npy1r in the DRG, or DH, respectively (Fig. 2 A and B). [Leu³¹, Pro³⁴]-NPY reduced SNI-induced mechanical and cold hypersensitivity in both control (Npy1r^loxP/loxP) and DRG conditional knockout mice (Npy1r^loxP/loxP;Pirt^Cre), but not in DH conditional knockout mice (Npy1r^loxP/loxP;Lbx1^Cre) (Fig. 2 C and D). These results indicate that intrathecal NPY Y1 agonists act at spinal cord interneurons rather than the peripheral terminals of DRG neurons to inhibit behavioral signs of neuropathic pain.

Fig. 2.

Y1 selective agonist acts at spinal cord interneurons rather than the central terminals of primary afferent neurons to inhibit SNI-induced neuropathic pain. (A) Fluorescence in situ hybridization of sections through DRG (Top row) and DH of the spinal cord (Bottom row) demonstrate that Npy1r^loxP/loxP mice express Npy1r in DRG and spinal cord, Npy1r^loxP/loxP;Pirt^Cre mice lack expression of Npy1r in the DRG, and Npy1r^loxP/loxP;Lbx1^Cre mice lack expression of Npy1r in the DH of the spinal cord. (B) Quantification demonstrating complete ablation of Npy1r-expressing cells in L3-L4 DRGs of Npy1r^loxP/loxP;Pirt^Cre, and L4 superficial DH of Npy1r^loxP/loxP;Lbx1^Cre mice as compared to Npy1r^loxP/loxP (control) expression. (C) [Leu³¹, Pro³⁴]-NPY (10.0 µg) abolishes SNI-induced mechanical allodynia in Npy1r^loxP/loxP and Npy1r^loxP/loxP;Pirt^Cre mice but not in Npy1r^loxP/loxP;Lbx1^Cre mice (n = 9 mice/group). Three-way RM ANOVA. Holm-Sidak post hoc test. (D) Leu³¹, Pro³⁴]-NPY (10.0 µg) abolishes SNI-induced cold allodynia in Npy1r^loxP/loxP and Npy1r^loxP/loxP;Pirt^Cre mice but not in Npy1r^loxP/loxP;Lbx1^Cre mice (n = 9 mice/group). Three-way repeated measure (RM) ANOVA. Holm-Sidak post hoc test. Data are shown as means ± SEM. *P < 0.05, **P < 0.01, ****P < 0.0001. Dots represent data points from individual animals.

Y1-INs Are Necessary for the Sensory and Affective Components of SNI-Induced Neuropathic Pain.

Our conditional genetic knockout of peripheral and central Npy1r data indicate that Y1 receptor agonists act at spinal cord interneurons (Y1-INs) to inhibit SNI-induced neuropathic mechanical and cold hypersensitivities. To directly test the hypothesis that Y1-INs are necessary for the full manifestation of SNI-induced hyperalgesia, we used intraspinal inhibitory chemogenetics to inhibit the adult spinal Y1-IN population. We injected a Cre-dependent inhibitory DREADD (AAV8-hSyn-DIO-hM4D_Gi) into the left lumbar (targeting L3–L4) DH of Npy1r^Cre mice (Fig. 3A). As described previously (15, 32), the AAV8-hSyn serotype selectively transfected neurons in the spinal cord but not DRG (SI Appendix, Fig. 2 A and B). As expected, AAV transfection selectively occurred in Npy1r-expressing cells (SI Appendix, Fig. 2C). Immunohistochemical verification of AAV expression was detected ipsilateral to viral injection (left but not right DH) and dorsal to the PKCγ band that delineates lamina IIi, indicating that its expression was restricted to the superficial DH (laminae I-IIo) (Fig. 3B). When administered after SNI (but not before), clozapine N-oxide (CNO 3 mg/kg, intraperitonially [i.p.]) abolished both mechanical and cold hypersensitivity (Fig. 3 C–G) and reduced light-touch evoked pERK in the superficial DH (Fig. 3 H and I). We selected the 3 mg/kg dose of CNO as this dose is sufficient to alter DREADD-dependent neuronal activity within 15 min of administration, low enough to minimize off target effects (59) and is a commonly used dose for spinal cord chemogenetics (19, 60). In addition to stimulus-evoked features of pain such as allodynia, nerve injury also elicits stimulus-independent ongoing affective pain which can be assessed with conditioned place preference (CPP) paradigms (61). Following SNI and conditioning, mice injected with inhibitory DREADD (hM4D_Gi) but not control virus (mCherry) spent more time in the CNO-paired chamber (Fig. 3 J and K). These findings indicate that chemogenetic inhibition of spinal Y1-INs reduces stimulus-dependent and -independent components of neuropathic pain.

Fig. 3.

Chemogenetic inhibition of spinal Y1-INs reduces behavioral and immunohistochemical markers of SNI-induced nociceptive and affective pain. (A) Strategy for selectively targeting Y1-INs in Npy1r^Cre mice with intraspinal injections of the Cre-dependent virus AAV8-hSyn-DIO-hM4D_Gi-mCherry. (B) Representative immunohistochemistry image of the spatial distribution of AAV-hM4D_Gi-mCherry neurons (red) that is largely restricted to the ipsilateral (Left) superficial DH (laminae I-IIo) and dorsal to the PKCγ band (green). DAPI (blue). (C) Experimental timeline of chemogenetic reflexive behavioral testing at both pre- and post-SNI time points. (D–G) Chemogenetic inhibition of Y1-INs with CNO (3 mg/kg) does not alter mechanical or cold withdrawal thresholds before SNI but dramatically reduces both SNI-induced mechanical and cold allodynia (n = 7 mice/group). Two-way ANOVA. Holm-Sidak post hoc tests. (H) Representative images of light-touched evoked pERK in the spinal cord DH following intraperitoneal administration of CNO (3 mg/kg). (Scale bars, 100 µm). (I) Chemogenetic inhibition of Y1-INs with CNO (3 mg/kg) reduces light-touched evoked pERK in SNI mice (n = 7 mice/group). Student’s unpaired two-tailed t test. (J) Protocol for conditioned place preference. (K) Chemogenetic inhibition of Y1-INs induces CPP (increased time in the CNO-paired chamber) in SNI mice (n = 10 to 11 mice/group). Student’s unpaired two-tailed t test. Data are shown as means ± SEM. *P < 0.05, ***P < 0.001, ****P < 0.0001. Dots represent data points from individual animals.

Nerve Injury Depolarizes the Resting Membrane Potential and Increases the Excitability of Delayed Firing Y1-INs.

Neurochemical, neurophysiological, and morphological characterization of Y1-INs.

Spinal cord DH neurons are classified by neurochemical gene/protein expression, neurophysiological firing pattern, and/or cellular morphology (62, 63). Using these criteria, we characterized Y1-INs in lumbar (L3–L4) sections obtained from BAC transgenic mice that express enhanced green fluorescent protein (eGFP) under the control of the Npy1r promoter (Npy1r^eGFP). First, we investigated the neurochemical identity of Y1-INs with fluorescence in situ hybridization. Consistent with our previous results in the rat (49, 53), we found that almost all Npy1r^eGFP neurons coexpress neurochemical markers of excitatory (Lmx1b) but not inhibitory (Pax2) neurons (64–67); we show that this was not altered by nerve injury (SI Appendix, Fig. 3). Second, we characterized superficial DH Npy1r^eGFP neurons using ex vivo slice electrophysiology recordings to evaluate the firing patterns evoked by steady-state current injection as described previously (36) (Fig. 4A). We identified four main firing patterns in Npy1r^eGFP neurons: delayed long latency firing (DLLF) neurons, delayed short latency firing (DSLF) neurons, initial burst firing (IBF) neurons, and phasic firing (PF) neurons (Fig. 4 B and C). Third, with respect to each firing pattern, we used neurobiotin labeling and confocal imaging to characterize the cellular morphology of Y1-INs as described by Grudt and Perl (68). As illustrated in Fig. 4 D and E, most Y1-INs, and particularly the DSLF and IBF firing types, exhibited a morphology similar to the central cells described by Grudt and Perl. In summary, most Y1-INs are excitatory neurons that exhibit a DSLF firing pattern and a central cell morphology.

Fig. 4.

Electrophysiological and morphological characterization of Y1-INs. (A) Schematic illustration of parasagittal slices from the L3/L4 segment of adult mice that were used for electrophysiological recordings in Npy1r^eGFP neurons. (B) Representative examples and (C) incidence of firing patterns in Npy1r^eGFP neurons (DLLF, DSLF, IBF, or PF) (n = 100 neurons from 36 mice). (D) Representative examples of biotin filled Npy1r^eGFP neuron morphologies. (Scale bars, 50 µm). (E) Quantification of morphology incidence within each electrophysiological firing pattern (DLLF [n = 15 cells]; DSLF [n = 13 cells]; IBF [n = 9 cells]; and PF [n = 11 cells]). (F) Nerve injury did not change the incidence of firing patterns in Npy1r^eGFP neurons (n = 100 sham neurons and n = 100 SNI neurons). χ² Test. (G) Nerve injury produces a depolarizing shift in the resting membrane potential in DSLF Npy1r^eGFP neurons (DLLF [n = 10 to 13]; DSLF [n = 28 to 33]; IBF [n = 4 to 7]; and PF [n = 4 to 7]). χ² Test. (H) Nerve injury did not change the membrane capacitance of any Npy1r^eGFP neuron firing types (DLLF (n = 14 to 19 neurons), DSLF (n = 41 to 45 neurons); IBF (n = 4 to 15 neurons), and PF (n = 4 to 15 neurons). χ² Test. (I–L) Nerve injury did not change current-voltage (I-V) plots for any Npy1r^eGFP neuron firing types (DLLF [n = 7 to 11 neurons]; DSLF [n = 15 to 23 neurons], IBF [n = 4 to 12 neurons]; and PF [n = 8 to 11 neurons]). Two-way ANOVA. Data are shown as means ± SEM. *P < 0.05.

Effect of nerve injury on firing type and passive membrane properties.

The response properties of DH neurons are shaped by their passive biophysical membrane properties (69). Interestingly, peripheral nerve injury modestly affects or does not alter the passive membrane properties of randomly sampled interneurons in the rat or GABAergic interneurons in the mouse DHs (18, 22, 70–74). For this reason, we hypothesized that SNI would not alter the passive membrane properties of Y1-INs. To test this, we conducted ex vivo slice electrophysiology recordings in both naïve and SNI Npy1r^eGFP mice. As hypothesized, SNI did not change the percentages of DSLF, DLLF, IBF or PF Y1-IN firing patterns (Fig. 4F). To our surprise, SNI produced a depolarizing shift of the resting membrane potential for specifically DSLF neurons (Fig. 4G), though did not alter membrane capacitance (Fig. 4H) nor the current–voltage (I–V) relations (Fig. 4 I–L) in any Npy1r^eGFP firing populations. In summary, a more positive resting membrane potential in DSLF neurons indicates that SNI increases intrinsic excitability, a key feature of central sensitization (13).

Effect of nerve injury on active membrane properties.

Peripheral nerve injury can enhance the membrane excitability of DH neurons (13). To test the hypothesis that SNI increases the membrane excitability of Y1-INs, we employed depolarizing current ramps to elicit action potential firing (70, 71). We focused our analysis on the DSLF subtype for two reasons: 1) DSLF neurons were the only population found to have a depolarizing shift in resting membrane potential after SNI (Fig. 4G) and 2) the DSLF population is the most abundant population of Y1-INs when segregated by firing type (Fig. 4F). Neurons were held at −60 mV and current ramps were applied at 67 pA/s (Fig. 5A). SNI did not change the number of action potentials in DSLF neurons but decreased the latency to first action potential (Fig. 5 B and C). We then examined the effect of SNI on membrane excitability in DSLF Npy1r^eGFP neurons using stepped depolarization from hyperpolarized conditions. All DSLF Y1-INs were held at −80 to −85 mV for 500 msec before application of current. Steady-state 1-s current steps from 0 to 80 pA were used to elicit action potentials. Representative firing patterns in DSLF Npy1r^eGFP neurons from SNI mice are shown in Fig. 5D. SNI increased the number of action potentials per step during the 1-s pulse in DSLF Y1-INs (Fig. 5E), decreased the current threshold needed to elicit first action potentials (Fig. 5F), increased the number of neurons exhibiting rebound spiking (action potential spike at 0 pA from conditioning hyperpolarized state is used to measure rebound spiking). SI Appendix, Fig. 4 illustrates that there are no differences in the afterhyperpolarization between naïve and SNI mice; further studies are needed to clarify the cause of rebound depolarization) (Fig. 5G), and increased average firing frequency (but not spike amplitude) at current steps to 40 pA (Fig. 5 H and I).We also found that SNI robustly increased the membrane excitability of DLLF neurons (SI Appendix, Fig. 5). We conclude that SNI increases the membrane excitability of delayed firing Y1-INs.

Fig. 5.

SNI increases the excitability of DSLF Y1-INs in the superficial DH. (A) Representative example of a DSLF Npy1r^eGFP neuron firing pattern upon injection of ramped current at 67 pA/s. (B) SNI did not alter the number of action potential spikes but (C) decreased the latency to first spike elicited by ramp current in DLSF Npy1r^eGFP neurons (n = 11 to 13 neurons). Student’s unpaired two-tailed t test. (D) Representative example of a DSLF Npy1r^eGFP neuron held at hyperpolarized condition (−80 mV) for 500 msec before application of steady-state 1-s current steps from 0 to 40 pA. (E) SNI increased the number of action potential spikes at current injection steps, (F) lowered the Rheobase, (G) increased the percentage of neurons exhibiting rebound spiking and increased (H) firing frequency but (I) not amplitude in DLSF Npy1r^eGFP neurons (n = 10 to 33 neurons). Student’s unpaired two-tailed t test. (J and K) Cumulative distribution plots (Left) and group data (Right) show SNI increases (J) the average amplitude and (K) frequency of sEPSCs in DSLF Npy1r^eGFP neurons (n = 16 neurons/group). Kolmogorov-Smirnov test and Student’s unpaired two-tailed t test. Data are shown as means ± SEM. *P < 0.05, **P < 0.01, ***P < 0.001, ****P < 0.0001. Dots represent data points or averages from individual cells.

Effect of nerve injury on synaptic excitability.

Peripheral nerve injury dramatically enhances the spontaneous excitatory synaptic activity of excitatory DH neurons (70, 71, 75). To assess changes in excitatory synaptic transmission in Y1-INs, we recorded spontaneous excitatory postsynaptic currents (sEPSCs) from both naïve and SNI Npy1r^eGFP mice, with the assumption that changes in amplitude and frequency will be attributed to postsynaptic and presynaptic adaptations (70, 71), respectively. We found that SNI increased both amplitude and frequency (decrease in interevent interval, IEI) of sEPSCs (Fig. 5 J and K and SI Appendix, Figs. 6 and 7), suggesting that both pre- and postsynaptic adaptations contribute to an increased excitatory drive in DSLF, IBF, and PF Y1-INs.

Spinal Y1-INs Are Sufficient for the Behavioral Manifestations of Pain.

Our intrathecal pharmacology (Figs. 1 and 2) and inhibitory chemogenetic data (Fig. 3) indicate that spinal Y1-INs are necessary for the full manifestation of the behavioral symptoms of neuropathic pain. To test the hypothesis that activation of Y1-INs is sufficient to induce pain, we chemogenetically activated Y1-INs and measured both evoked and nonevoked pain-like behaviors in uninjured mice. We injected a Cre-dependent excitatory DREADD (AAV8-hSyn-DIO-hM3D_Gq) into the left lumbar DH of Npy1r^Cre mice (Fig. 6A), and we measured both CNO-elicited spontaneous and evoked pain-like behaviors (76) (Fig. 6B). CNO (3 mg/kg, i.p.), but not saline, induced robust spontaneous nocifensive (lifting, flapping, shaking, licking, guarding) but not itch-like (biting or scratching) behaviors directed to the left ipsilateral (virus-injected side), but not contralateral, hind limb in mice injected with the excitatory DREADD (hM3D_Gq) but not control (mCherry) virus (Fig. 6 C and D and Movie S1). CNO also elicited both mechanical and thermal hypersensitivities (Fig. 6 E–H). Next, we tested the hypothesis that chemogenetic activation of Y1-INs is sufficient to produce avoidance using a conditioned place aversion paradigm (CPA) (Fig. 6I). We found that chemogenetic Y1-IN activation produced a robust CPA (Fig. 6 J and K).

Fig. 6.

Activation of spinal Y1-INs in uninjured mice is sufficient to induce behavioral nocifensive signs of neuropathic pain. (A) Strategy for selectively targeting Y1-INs in Npy1r^Cre mice with intraspinal injections of the Cre-dependent virus AAV8-hSyn-DIO-hM3D_Gq-mCitrine. (B) Experimental timeline for chemogenetic activation of Y1-INs with CNO (3 mg/kg) and the videotaping of spontaneous hind left-limb directed nocifensive behaviors. (C) Representative images of mice engaged in ipsilateral hindpaw-specific nocifensive behaviors. (D) CNO (3 mg/kg) but not saline induces spontaneous ipsilateral-directed (hind left paw) nocifensive behaviors (lifting, licking, flapping, shaking) in hM3D_(Gq)- but not mCherry-injected Npy1r^Cre mice. (n = 3 to 5 mice/group). Three-way repeated measure (RM) ANOVA. Holm-Sidak post hoc test. (E) Experimental timeline of chemogenetic activation of Y1-INs with CNO (3 mg/kg) and behavioral testing. (F–H) Chemogenetic activation of Y1-INs with CNO (3 mg/kg) induces (F) mechanical (vF), (G) cold (acetone droplet withdrawal), and (H) heat (52.5 C hotplate) hypersensitivity (n = 8 mice/group). (F and G) Two-way ANOVA. Holm-Sidak post hoc tests. (H) Student’s unpaired two-tailed t test. (I) Experimental protocol for conditioned place aversion. (J and K) Activation of Y1-INs induces conditioned place aversion (n = 10 to 11 mice/group). (J) Student’s paired two-tailed t test. (K) Two-way ANOVA. Holm-Sidak post hoc test. (L) Experimental timeline of chemogenetic activation of Y1-INs with CNO (3 mg/kg) and then immediate placement of mice into an open field testing environment. (M–P) Chemogenetic activation of Y1-INs with CNO (3 mg/kg) produces robust discomfort represented by (M) continuous movement, (N) lack of rest, and (O) increased grooming-like (stereotypic episode activity count) behaviors. (P) Heat maps depicting averaged activity for all mice for the entire 60-min duration of testing (n = 6 mice/group). (M–O) Student’s unpaired two-tailed t test. (Q) Strategy for selectively targeting Y1-INs in Npy1r^Cre mice with intraspinal injections of the Cre-dependent virus AAV8-hSyn-Flex-Chronos-GFP and implantation of a wireless optogenetic 473 nm spinal LED implant. Optogenetic activation of Y1-INs induces a frequency dependent increase in (R) mechanical and (S) cold hypersensitivity (n = 8 mice/group). Two-way RM ANOVA. Holm-Sidak post hoc tests. Data are shown as means ± SEM. *P < 0.05, **P < 0.01, ***P < 0.001, ****P < 0.0001. Dots represent data points from individual animals.

We also evaluated the effect of chemogenetic activation of Y1-INs on nonevoked nocifensive behaviors. Immediately following CNO administration (3 mg/kg, i.p.), mice injected with control virus (mCherry) or excitatory DREADD (hM3D_Gq) were placed into a novel open field apparatus for 60 min (Fig. 6K). Chemogenetic activation of Y1-INs produced a dramatic increase in ambulation (pain-like agitation), inability to rest, and robust grooming-like behaviors (stereotypic episode activity count) (Fig. 6 L–O and SI Appendix, Fig. 8 A–C). These results are consistent with increased locomotion/agitation and grooming seen following injection of formalin into the hindpaw (77–79). We believe that the chemogenetic Y1-IN activation-induced increase in movement is an index of nociception rather than anxiety. First, Y1-IN activation reduced thigmotaxis (mice spent more time in the center and less time in the margin of the open field) (SI Appendix, Fig. 8 D and E) (80, 81). Second, pretreatment with the first-line neuropathic pain analgesic, gabapentin (30 mg/kg, i.p (82, 83).), reduced the nocifensive Y1-IN activation-induced increases in movement and grooming-like behaviors (SI Appendix, Fig. 8 F–I). Thus, Y1-IN activation produces spontaneous nocifensive-like (not anxiety-like) behaviors that can be abolished with the pain analgesic gabapentin.

Chemogenetic activation provides a powerful tool for excitation of Y1-INs with high spatial resolution, whereas it lacks the frequency-dependent and temporal activation that is afforded by excitatory optogenetic stimulation. To perform in vivo spinal optogenetic stimulation of Y1-INs and behavioral testing in awake and freely moving mice, we took advantage of newly developed wireless, lightweight, flexible, and implantable optoelectronic devices (84, 85). These spinal implants do not alter baseline somatosensation or motor coordination (SI Appendix, Fig. 9). We injected a Cre-dependent excitatory channelrhodopsin with fast kinetics and enhanced light sensitivity (AAV8-hSyn-Flex-Chronos) (86) into the left lumbar DH of Npy1r^Cre mice before implanting a 473-nm, spinal optoelectronic device overlying the dura mater (Fig. 6P). Delivery of light to the surface of the dorsal spinal cord in mice injected with the excitatory opsin (Chronos), but not control virus (mCherry), produced a robust and frequency-dependent (0, 5, 10, 20 Hz) increase in mechanical and cold hypersensitivity in the ipsilateral hindpaw (Fig. 6 Q and R). These data indicate that activation of Y1-INs is sufficient to produce pain-like behaviors.

Discussion

We propose that Npy1r-expressing neurons in the DH are fundamental to both the initiation and maintenance of neuropathic pain.

Intrathecal NPY Y1 Agonists Act at Spinal Cord Interneurons Rather Than the Peripheral Terminals of DRG Neurons to Inhibit Behavioral Signs of Neuropathic Pain.

Several decades of research have established that intrathecal NPY reduces the behavioral and spinal molecular signs of hyperalgesia in rodent models of chronic pain by targeting Y1 receptors (48, 49, 51, 87–89). However, the exact target for Y1-mediated anti-nociception has remained elusive due to the dual expression of Y1 at both the central terminals of primary afferents and on spinal interneurons, both of which can be engaged following intrathecal injection. To resolve the Y1 site of action, we utilized the Cre-lox recombination system to conditionally knockout Npy1r in either the DRG or the DH (Fig. 2). The resulting data demonstrate that intrathecal [Leu³¹, Pro³⁴]-NPY acts exclusively at Y1 on DH neurons to inhibit allodynia. In future studies this same methodology can be used to resolve the Y1 agonist target for analgesia in other pain models. In particular, models of peripheral inflammation should be explored as Npy1r in the DRG is abundantly expressed within peptidergic C fibers (41, 49, 51, 90), thus, peripheral Y1 may have more analgesic efficacy in these models.

We demonstrate that the Y1 receptor can be targeted with agonists delivered at early (3 wk post-SNI) and later time points (13 wk post-SNI) to alleviate behavioral signs of allodynia, consistent with previous chronic pain studies in the rat reporting that spinal Y1 could be repeatedly engaged to alleviate chronic constriction injury-induced neuropathic allodynia (48). In contrast to opioids which fail to alleviate neuropathic pain (likely due to down-regulation of receptors in neuropathic pain states (42, 91)), our results suggest that spinally administered Y1 agonists remain highly efficacious in chronic neuropathic pain states long-term, a critical result for future clinical translation.

Intrathecal injection of the Y1 agonist at 3 wk postsurgery increased mechanical withdrawal thresholds not only in SNI mice but also in Sham controls. Injection of the Y1 agonist at a much later time point (13 wk postsurgery) continued to increase mechanical and cold withdrawal thresholds in the SNI group, but this effect was lost in the sham group. These results are consistent with literature indicating that Y1 agonists behave as analgesic drugs for heat pain (92), but exhibit largely antihyperalgesic actions for mechanical pain (49).

SNI Enhances the Potency of Intrathecal [Leu³¹, Pro³⁴]-NPY.

GPCRs play fundamental roles in regulating biophysical functions and are pharmacological targets for many classes of drugs. [³⁵S]GTPγS binding assays measure the level of G protein activation following the occupation of a GPCR by an agonist to provide pharmacological values of potency, efficacy, and affinity (93). We used [Leu³¹, Pro³⁴]-NPY-stimulated [³⁵S]GTPγS binding to analyze the functional coupling of Y1 receptors to G proteins in the DH after SNI. SNI did not change the E_max of NPY-stimulated G protein activation through Y1 receptors. This demonstrates that the magnitude of spinal Y1-G protein coupling is not diminished or desensitized following nerve injury. We found that SNI increased the potency (decreased the EC₅₀ ipsilateral to SNI) of [Leu³¹, Pro³⁴]-NPY as compared to sham surgery (Fig. 1), thus revealing an enhanced affinity of Y1 coupling to G protein activation. This increase in NPY potency after SNI indicates that lower concentrations of NPY can activate or maintain Y1-driven antinociceptive signaling during neuropathic pain. The molecular mechanisms underlying this phenomenon are not well-defined but could reflect the regulation of Y1/G_i/o coupling dynamics via altered accessory protein partnering, modified stoichiometry of Y1 receptors to available G protein pools, receptor phosphorylation states, movement between membrane lipid domains, or even oligomerization of receptor proteins (94). We propose that the increased activation of G proteins will amplify the effect of [Leu³¹, Pro³⁴]-NPY in the setting of peripheral nerve injury, profoundly increasing the anti-hyperalgesic efficacy of [Leu³¹, Pro³⁴]-NPY. In summary, our findings promote spinal Y1 as an extremely promising G_i/o therapeutic target for intervention with intrathecal Y1 selective agonists to treat neuropathic pain.

SNI Increases the Excitability of Y1-INs.

The balance between excitation and inhibition in the DH determines the setpoint of somatosensory processing (9). A shift in this balance toward an enduring increase in excitability can be denoted as central sensitization (13). Our neurophysiological recordings of passive membrane, active membrane, and synaptic properties demonstrate that SNI increases excitability in Y1-INs. First, we note that SNI increased the resting membrane potential of DSLF Y1-INs (Fig. 4), suggesting that Y1-INs are more likely to fire an action potential following synaptic input. Second, we note that SNI increased the membrane excitability of DSLF and DLLF Y1-INs (Fig. 5 and SI Appendix, Fig. 5). These results are consistent with peripheral nerve injury-induced hyperexcitability/central sensitization/long term potentiation of DH neurons (13, 95–97). Third, we note that SNI robustly increased both the frequency and amplitude of sEPSCs in DSLF, IBF, and PF Y1-INs (Fig. 5 and SI Appendix, Figs. 6 and 7). The SNI-induced increase in sEPSC frequency is likely a result of an increased frequency of presynaptic action potentials (98). The SNI-induced increase in amplitude of sEPSCs may be a result of increased postsynaptic effectiveness of glutamate, perhaps due to central sensitization and the up-regulation of ionotropic glutamate receptors (13). These results are consistent with increased synaptic excitability (indicated by both increases in frequency and amplitude of sEPSCs/mEPSCs) of excitatory DH neurons following peripheral nerve injury (including SNI, chronic constriction injury, and sciatic nerve axotomy) or application of brain derived neurotrophic factor (BDNF) (70, 71, 75, 99). Interestingly, gabapentin, the first-line analgesic prescribed for neuropathic pain, acts by reducing both the frequency and amplitude of sEPSCs to spinal cord transient firing neurons (100). Although not tested, perhaps one of the mechanisms by which gabapentin is effective in neuropathic pain patients is via reducing the excitability of Y1-INs. This hypothesis is moderately supported by pretreatment of gabapentin preventing Y1-IN chemogenetic activation-induced nocifensive behaviors (SI Appendix, Fig. 8). Together, our results indicate that nerve injury robustly enhances the excitability of Y1-INs. One limitation of our recordings is that we utilized a nonsurgical (anesthesia) control and therefore we cannot exclude the possibility that some excitability changes associated with the nerve injury surgery may be due to skin or muscle damage. Nevertheless, we propose that the hyperexcitability of Y1-INs produces an abnormal amplification of innocuous inputs that drives the manifestation/facilitation of allodynia (2, 101) (Fig. 7A).

Fig. 7.

The proposed contribution of Y1-INs to the ascending peripheral nerve injury-induced mechanical allodynia circuit in the DH. (A) Graphical abstract summarizing key data from this article implicating Npy1r-expressing interneuron excitability in the manifestation of pain. (B) Innocuous mechanical inputs activate Aβ/Aδ myelinated afferents (red) that project into deeper laminae of the DH and synapse onto interneurons marked by the expression of CCK (purple) and PKCγ (yellow). Normally, feedforward inhibition prevents the activation of these interneurons and as a result light touch is perceived as nonpainful. However, in the context of neuropathic pain, feedforward inhibition onto PKCγ interneurons is lost and innocuous light touch inputs activate a theorized dorsally directed microcircuit to allow innocuous light touch sensory information to be perceived as painful. In this theorized circuit, activated PKCγ interneurons excite transient central cells (speculated as Y1-INs), that in turn synapse onto vertical cells, which then activate ascending PNs that travel via the spinothalamic and spinoparabrachial tracts to be processed via higher order pain centers such as the lateral parabrachial nucleus. Inhibitory NPY interneurons (light gray) may gate some of these nociceptive inputs at the Y1-IN and normally prevent these neurons from being activated and driving pain-like behaviors. Exogenous NPY or Y1 agonist binding to the G_i-coupled NPY Y1 receptor results in inhibition of Y1-INs and thus the abolishment of peripheral nerve injury-induced mechanical allodynia.

Investigation of Adult DH Npy1r-Expressing Neurons.

In this study we took advantage of NPY Y1 receptor-selective pharmacological agents, conditional genetic knockouts, intraspinal administration of Cre-dependent AAV constructs into the spinal cords of adult mice, and electrophysiological recordings in a reliable Npy1r^eGFP mouse line (36) to thoroughly characterize spinal interneurons that genuinely express the NPY Y1 receptor or Npy1r mRNA in adulthood. One previous study characterized Npy1r^Cre-lineage neurons, however, Npy1r is transiently expressed during DH development and these genetic crosses labeled most of the excitatory neurons in the DH, many of which that were Npy1r-lacking (102). Our approach circumvented this developmental issue. Further, through the use of cell type–specific inhibition tools, we have built upon prior results lesioning the spinal Y1-INs with NPY-saporin conjugates in rats (53, 103, 104), however, we avoided the negative pitfalls of lesion studies that include toxicity and circuit rearrangements. Importantly, our viral AAV transfected adult Npy1r-expressing cells (SI Appendix, Fig. 2C) and we did not find effects of CNO administration or blue light stimulation in mCherry controls (Figs. 3 D–K and 6 F–S), or effects of saline administration in hM3D_Gq-transfected mice (Fig. 6D), indicating that we manipulated Npy1r-expressing adult Y1-INs with high fidelity. Our findings reveal the critical importance of DH Y1-INs to the manifestation and maintenance of neuropathic pain-like behaviors.

Cell Type–Specific Interrogation of In Vivo Y1-INs Finds a Key Role in the Manifestation of Pain.

With modern genetic technologies, researchers can permanently or transiently activate, inhibit, or ablate spinal excitatory interneuron populations in vivo to test predications about the identity of pain-transmitting cells (105). Using this approach, we found that cell type–specific activation of excitatory Y1-INs elicited spontaneous nocifensive behaviors, mechanical and thermal hypersensitivities, and conditioned place aversion. Conversely, chemogenetic inhibition of Y1-INs reduced pain-like behavior. These functional results are markedly similar to SST-IN inhibition and activation: ablation of SST-INs prevented the induction of both neuropathic and inflammatory allodynia (28), and chemogenetic or optogenetic activation of SST-INs induced spontaneous nocifensive behaviors, mechanical and thermal hypersensitivities, and conditioned place aversion (29). These similar results between our functional studies and SST-IN studies are likely due to the extensive overlap of Y1-INs with SST-INs (49, 53, 106). Thus, in this study we are likely chemogenetically/optogenetically activating a subset of the SST population that is sufficient for nocifensive behaviors and fundamental to the manifestation of peripheral nerve injury-induced mechanical pain. Additionally, our results are consistent with NPY-saporin lesion studies which find a primary role for Y1-INs in the development and maintenance of both neuropathic and inflammatory pain (53, 103).

We propose that Y1-INs are normally under strong inhibition from NPY-expressing inhibitory interneurons. Indeed, Y1-INs receive direct and functional synaptic contacts from inhibitory NPY-expressing interneurons (102). We believe that chemogenetic inhibition of Y1-INs does not alter baseline responsiveness to mechanical or cold stimuli for this reason (Fig. 3)- inhibition of an already silenced interneuron population has no net effect. However, following peripheral injury, Y1-INs lose their inhibition, become hyperexcitable, and drive pain; consequently, their chemogenetic inhibition potently reduces pain-like behavior. Recently it was shown that chemogenetically activating inhibitory NPY-expressing neurons potently reduced both neuropathic behavioral and spinal molecular signs of hyperalgesia (107). We believe that the chemogenetic activation of NPY-inhibitory interneurons results in indirect Y1-IN inhibition (gating) to prevent allodynia. In summary, Y1-INs function like a pain rheostat: noxious stimuli increase the activity of Y1-INs to promote pain, whereas their inhibition reduces pain-like behaviors (Fig. 7A).

Proposed Model of Y1-INs within an Ascending DH Microcircuit That Develops after Nerve Injury.

Our study implicates Y1-INs as a key neuron population in the manifestation of peripheral nerve injury-induced mechanical allodynia. Currently, mechanical allodynia is hypothesized to occur via a polysynaptic DH microcircuit that allows Aβ-fibers to transmit innocuous mechanical input as painful (18). Within this model, CCK and PKCγ interneurons have been identified/labeled as key neuron populations (18, 20). However, as this model was first discovered with random patch-clamp recordings in unlabeled DH cells, two other key interneuron populations within the model, excitatory transient central cells [neurons in lamina II outer with a central morphology that discharge action potentials transiently during a depolarizing step (68)], and vertical cells [neurons in lamina II outer with a large vertically oriented dendritic arborization (68)], both remain unclassified by neurochemical gene/protein expression. We find that Y1-INs mainly exhibit a central morphology and exhibit a DSLF firing type that closely mirrors the transient firing type (Fig. 4) (36, 68). Further, Y1-INs densely overlap with the Grp-expressing population of spinal cord interneurons (49, 108) which to date is the only identified class of transient central cells (109). These data suggest that Y1-INs may be the transient central cell population in the peripheral nerve injury-induced mechanical allodynia model (Fig. 7B).

Materials and Methods

Details are provided in SI Appendix. In brief, male and female mice were used for all studies. A model of neuropathic pain, SNI, was developed by ligating and cutting the tibial and common peroneal branches of the sciatic nerve. [Leu³¹, Pro³⁴]-NPY-evoked [³⁵S]GTPγS binding was performed in lumbar sections from both SNI and sham mice. Behavioral testing was largely performed by applying calibrated von Frey filaments or a 10 μL droplet of acetone to the lateral hindpaw of mice. Intrathecal administration of agonists/antagonists or chemogenetic/optogenetic AAV vectors injected into the spinal cord DH were used to investigate the role of Y1 and Y1-INs, respectively. Whole-cell slice neurophysiology recordings from GFP-labeled Y1-INs in lumbar sections of both sham and nerve-injured mice was used to investigate neural excitability. AAV- or GFP-labeled neurons were further investigated using immunohistochemistry and RNAscope fluorescence in-situ hybridization. Statistical analyses were performed using GraphPad Prism or SigmaPlot. All procedures were approved by the Institutional Animal Care and Use Committees of the University of Pittsburgh and University of Kentucky.

Data, Materials, and Software Availability

All study data are included in the article and/or supporting information.