Medullary kappa-opioid receptor neurons inhibit pain and itch through a descending circuit

Abstract

In perilous and stressful situations, the ability to suppress pain can be critical for survival. The rostral ventromedial medulla contains neurons that robustly inhibit nocioception at the level of the spinal cord through a top-down modulatory pathway. Although much is known about the role of the rostral ventromedial medulla in the inhibition of pain, the precise ability to directly manipulate pain-inhibitory neurons in the rostral ventromedial medulla has never been achieved. We now expose a cellular circuit that inhibits nocioception and itch in mice.

Through a combination of molecular, tracing and behavioural approaches, we found that rostral ventromedial medulla neurons containing the kappa-opioid receptor inhibit itch and nocioception. With chemogenetic inhibition, we uncovered that these neurons are required for stress-induced analgesia. Using intersectional chemogenetic and pharmacological approaches, we determined that rostral ventromedial medulla kappa-opioid receptor neurons inhibit nocioception and itch through a descending circuit. Lastly, we identified a dynorphinergic pathway arising from the periaqueductal grey that modulates nociception within the rostral ventromedial medulla. These discoveries highlight a distinct population of rostral ventromedial medulla neurons capable of broadly and robustly inhibiting itch and nocioception.

Through a combination of molecular, tracing and behavioural approaches, Nguyen et al. show that neurons in the rostral ventromedial medulla containing the kappa-opioid receptor robustly and bidirectionally modulate itch and pain behaviours.

See Fields (https://doi.org/10.1093/brain/awac212) for a scientific commentary on this article.

See Fields (https://doi.org/10.1093/brain/awac212) for a scientific commentary on this article.

Introduction

The brain exerts powerful control over nocioceptive processing at the level of the spinal cord. This top-down control occurs through a mechanism known as the descending modulation of pain.^1–3 Critically, the rostral ventromedial medulla (RVM) represents one final common node within the brain that directly modulates the activity of spinal neurons involved in pain transmission.⁴ Placebo and stress have also been shown to alter the perception of pain through the endogenous opioidergic pain modulatory system, centred upon the RVM.^5–8 Elucidating how neurons within the RVM can modulate pain presents unique opportunities to harness the brain’s natural pain-inhibitory system to potentially treat chronic pain disorders, thereby avoiding unwanted and even harmful side effects of existing pharmacological approaches.

Neurons in the RVM have been identified through extensive electrophysiological and pharmacological studies.^4,9 For example, single unit recordings have provided significant insight into three types of neurons in the RVM ON, OFF and NEUTRAL cells based on their firing responses to noxious stimulation and sensitivity to pharmacological agents.^4,10,11 ON cells are active during noxious stimulation, are inhibited by morphine and have been proposed to facilitate nocioception, whereas OFF cells are inactive or exhibit silent ongoing activity during noxious stimulation, are excited by morphine and are thought to inhibit nocioception.^4,12–14 Although these studies have established a robust framework by which to classify RVM neurons, many were conducted in anaesthetized rodents and the molecular identities of ON and OFF cells remain unclear.

The role of inhibitory RVM neurons is also controversial. One study uncovered that RVM^Vgat (GABAergic) neurons facilitate mechanical, but not heat, nocioception.¹⁵ However, a different study found that RVM^Gad2 neurons (which comprise a subpopulation of GABAergic neurons) were involved in the inhibition of heat nocioception.¹⁶ These discordant findings may reflect the fact that GABAergic RVM neurons comprise numerous cell types, and different Cre alleles likely target somewhat different subsets. Both ON and OFF cells are thought to be GABAergic, and glutamic acid decarboxylase (GAD)+ immunoreactivity has even been identified in some NEUTRAL cells.¹⁷ In addition to these apparent discrepancies, the role of RVM neurons, including the Gad2/Vgat populations, in the modulation of itch has not been tested.

Previous work has highlighted the existence of OFF cells in the RVM that robustly inhibit nocioception,^4,12–14 yet it has been difficult to identify distinct cell types. One candidate population of OFF cells in the RVM includes neurons expressing the kappa-opioid receptor (KOR).^18–20 Microinjections of kappa agonists into the RVM have been shown to elicit mechanical hypersensitivity following injury and attenuate morphine-induced analgesia.^18–20 Because KOR agonists traditionally reduce neuronal excitability through Giα signalling, these studies, therefore, suggest that RVM^KOR neurons participate in the inhibition of nocioception. However, this idea remains controversial; microinjections of kappa agonists into the RVM have also resulted in elevations of withdrawal thresholds, which are proposed to occur through the presynaptic inhibition of excitatory inputs onto ON cells,^21,22 thereby resulting in an anti-nocioceptive response. One explanation for the observed discrepancies may be due to the site at which kappa agonists exert their action, which could include RVM^KOR cell bodies as well as KOR-expressing presynaptic terminals within the RVM. Thus, one advantage of using genetic tools is the ability to manipulate neurons directly, which may shed light on the precise role of RVM^KOR neurons.

Here, we identified a population of spinally projecting, GABAergic RVM^KOR neurons that inhibit nocioceptive thresholds, itch and pain behaviours following acute and chronic injury. We also determined that RVM^KOR neurons tonically inhibit nocioception that was unmasked following pharmacological and chemogenetic inhibition. Furthermore, we found that RVM^KOR neurons were required for the behavioural expression of stress-induced analgesia. Lastly, we identified that dynorphin signalling from the periaqueductal grey (PAG) to the RVM bidirectionally modulates nociception. These findings emphasize the role of the endogenous pain modulatory pathway, centred on RVM^KOR neurons, that robustly inhibits nocioception and itch.

Materials and methods

Mice

All animals were of the C57Bl/6 background. Cre mouse lines used are available at Jackson Laboratory: Pdyn^Cre (#027958), KOR^Cre (#035045), Oprk1^fl/fl (#030076). The studies were performed in both male and female mice 8–10 weeks of age. Even numbers of male and female mice were used for all experiments and no clear sex differences were observed so data were pooled. Mice were given free access to food and water and housed under standard laboratory conditions. The use of animals was approved by the Institutional Animal Care and Use Committee of the University of Pittsburgh.

Pharmacologic agents

Clozapine-N-oxide (Tocris) was dissolved in phosphate-buffered saline (PBS) and administered intraperitoneally (5 mg/kg), intrathecally (50 μg/kg in 10 μl) or locally via cannula injection (1 mmol in 300 nl). Salvinorin B (SalB; Tocris) was dissolved in dimethyl sulfoxide (DMSO) and administered subcutaneously (10 mg/kg). U69,593 (Sigma; 40 ng in 0.25 μl) was dissolved in 45% (2-hydroxypropyl)-β-cyclodextrin (HBC; Sigma). Morphine sulfate (Henry Schein; 1 μg in 0.25 μl) was diluted in sterile saline. Nor-binaltorphimine dihydrochloride (norBNI; Sigma, 100 ng in 250 nl) was dissolved in sterile saline. U69,593, morphine and norBNI were microinjected into the RVM. Capsaicin (0.1%) in 10% EtOH in PBS was injected (10 μl) into the plantar hindpaw. Chloroquine diphosphate salt (Sigma) was dissolved in physiological saline (100 μg in 10 μl) and administered intradermally.

Intradermal and intrathecal injections

For intradermal injection of chloroquine, hair was removed at least 24 h before the experiment. Chloroquine (100 μg in 10 μl) was administered into the nape of the neck or calf, which could subsequently be visualized by the formation of a small bubble under the skin. For intrathecal (IT) injections, hair was clipped from the back of each mouse at least 24 h before the experiment. All IT injections were delivered in a total volume of 10 μl using a 30-gauge needle attached to a luer-tip 25 μl Hamilton syringe. The needle was inserted into the tissue at a 45° angle and through the fifth intervertebral space (L5–L6). Solution was injected at a rate of 1 μl/s. The needle was held in position for 10 s and removed slowly to avoid any outflow of the solution. Only mice that exhibited a reflexive flick of the tail following puncture of the dura were included in our behavioural analysis. These procedures were performed in awake mice.

Intraspinal injections

Mice were anaesthetized with 100 mg/kg ketamine and 10 mg/kg xylazine. An incision was made at the spinal cord level corresponding to the L4–L6 dermatome. The IT space was exposed and two injections of approximately 300 nl of virus were infused 300 μm below the surface of the spinal cord at 5 nl/s via glass pipette through the IT space corresponding to L4–L6 of the spinal cord. The same approach was used to target the cervical spinal cord. The glass pipette was left in place for an additional 5 min before withdrawal. The incision was closed with 6-0 vicryl suture. Ketofen was delivered intraperitoneally (10 mg/kg), and mice were allowed to recover over a heat pad.

Stereotaxic injections and cannula implantation

Animals were anaesthetized with 2% isoflurane and placed in a stereotaxic head frame. A drill bit (MA Ford, #87) was used to create a burr hole and a custom-made metal needle (33 gauge) loaded with virus was subsequently inserted through the hole to the injection site. Virus was infused at a rate of 100 nl/min using a Hamilton syringe with a microsyringe pump (World Precision Instruments). Mice received 250–500 nl of virus. The injection needle was left in place for an additional 15 min and then slowly withdrawn over another 15 min. Injections and cannula implantations were performed at the following coordinates: RVM: anterior-posterior −5.80 mm, medial-lateral 0.00 mm, dorsal-ventral −6.00 and PAG: anterior-posterior −4.70 mm, medial-lateral ± 0.74 mm, dorsal-ventral −2.75. For implantation of guide cannulas (P1 Technologies), implants were slowly lowered 0.300–0.500 mm above the site of injection and secured to the skull with a thin layer of Vetbond (3M) and dental cement. The incision was closed using Vetbond and animals were given ketofen (IP, 10 mg/kg) and allowed to recover over a heat pad. Mice were given 4 weeks to recover prior to experimentation.

Spared nerve injury

For spared nerve injury surgery, the sural and superficial peroneal branches of the sciatic nerve were ligated (size 6.0 suture thread) and transected, leaving the tibial nerve intact. The muscle tissue overlying the injury was placed back together gently and the skin sutured. Animals were tested 3 weeks post-injury.

Capsaicin-induced injury

Animals received 10 μl of intraplantar capsaicin (0.1% w/v in 10% ethanol diluted in saline) and the total time spent licking the injured paw was quantified over 20 min.

Behaviour

All assays were performed in the Pittsburgh Phenotyping Core and scored by an experimenter blind to treatment.

Observation of scratching behaviour

Scratching behaviour was observed using a previously reported method.²³ On the testing day, the mice were individually placed in the observation cage (12 × 9 × 14 cm) to permit acclimation for 30 min. Scratching behaviour was videotaped for 30 min after administration of chloroquine. The temporal and total numbers of scratch bouts by the hind paws at various body sites during this period were counted. For the experiments targeting the lumbar spinal cord, the amount of time spent biting following intradermal chloroquine was summated.

Hargreaves testing

Animals were acclimated on a glass plate held at 30°C (Model 390 Series 8, IITC Life Science Inc.). A radiant heat source was applied to the hindpaw and latency to paw withdrawal was recorded.²⁴ Two trials were conducted on each paw, with at least 5 min between testing the opposite paw and at least 10 min between testing the same paw. To avoid tissue damage, a cut-off latency of 20 s was set. Values from both paws were averaged to determine withdrawal latency.

Stress-induced analgesia (forced swim test)

Mice were placed in a water bath maintained at 30°C and forced to swim for 3 min. After the swim, mice were returned to their home cages and tested 30 min later for stress-induced analgesia to hotplate and capsaicin (described further in their respective sections).

Hotplate

Twenty minutes after the forced swim test, mice were placed on a 55°C hotplate. The latency to first nocifensive response (lick or jump) and total number of licking bouts (defined as the number of times an animal lifted its paw to its mouth and removed the paw from the mouth) over a 60 s period were measured. Values were averaged across two trials for each mouse.

von Frey testing

Mechanical sensitivity was measured using the Chaplan up–down method of the von Frey test.²⁵ Calibrated von Frey filaments (North Coast Medical Inc.) were applied to the plantar surface of the hindpaw. Lifting, shaking and licking were scored as positive responses to von Frey stimulation. Average responses were obtained from each hindpaw, with 3 min between trials on opposite paws and 5 min between trials on the same paw.

Cold plantar assay

Paw withdrawal to cold sensitivity was measured as previously described.^26,27 Briefly, animals were acclimated to a 1/4′′ glass plate and crushed dry ice was packed into a modified 3 ml syringe and used as a probe. The dry ice probe was applied to the glass beneath the plantar hindpaw and the latency to withdrawal was recorded. Two trials were conducted on each hindpaw, with 3 min between trials on opposite paws and 5 min between trials on the same paw. A cut-off latency of 20 s was used to prevent tissue damage. Withdrawal latencies for each paw were determined by averaging values across two trials per paw.

Tail-flick assay

Tails were immersed 3 cm into a water bath at 48°C, and the latency to tail-flick was measured three times with a 5 min interval between trials.

Rotarod

Coordination and balance were determined using an accelerating rotarod. A session involving 120 s on a non-accelerating rotarod was prior to experimentation. Five consecutive trials were performed while the rotarod accelerated from 4 to 40 rpm in 30 s increments. The latency to fall off (s) was recorded for each trial.

Open field activity

Spontaneous activity in the open field was conducted over 30 min in an automated Versamax Legacy open field apparatus for mice (Omnitech Electronics Inc.). Distance travelled was measured by infrared photobeams located around the perimeter of the arena interfaced to a computer running Fusion v.6 for Versamax software (Omnitech Electronics Inc.), which monitored the location and activity of the mouse during testing. Activity plots were generated using the Fusion Locomotor Activity Plotter analyses module (Omnitech Electronics Inc.). Mice were placed into the open field 30 min after receiving chemogenetic activators [clozapine N oxide (CNO) or SalB].

Acoustic startle

Prepulse inhibition of acoustic startle was measured by placing mice into a soundproof chamber (Kinder Scientific Startle Monitor). After 5 min of white noise acclimation, mice were exposed to randomized trials of 500 ms exposures to sound pressure levels of to 65 or 115 dB with a 500 ms intertrial interval. Trials were repeated between seven and eight times for each mouse in a randomized order. Startle was measured as the maximum force in Newtons and the average response across the trial repetitions for each mouse was used for data analysis.

RNAscope in situ hybridization

Multiplex fluorescent in situ hybridization (FISH) was performed according to the manufacturer’s instructions (Advanced Cell Diagnostics #320850). Briefly, 14-µm thick fresh-frozen sections containing the RVM were fixed in 4% paraformaldehyde, dehydrated, treated with protease IV (#322336) for 15 min and hybridized with gene-specific probes to mouse. Probes were used to detect tdTomato-C2 (#317041-C2), mCherry-C2 (#431201), Mm-Oprm1-C1 (#315841), Mm-Oprk1-C1 (#316111), Mm-Oprk1-C2 (316111-C2), Mm-Penk-C1 (#318761), Mm-Nos1-C2 (#437651-C2), Mm-Tac1-C1 (#517971), Mm-Tph2-C2 (#318691-C2), Mm-Fos-C3 (#498401-C3), Mm-Pdyn-C2 (#31877-C2), Mm-Slc32a1-C3 (#319191-C3) and Mm-Slc17a6-C3 (#319171-C3). DAPI (#320858) was used to visualize nuclei. Three-plex positive (#320881) and negative (#320871) control probes were tested. Three to four z-stacked sections were quantified and averaged for each mouse (technical replicates were averaged) and 3–4 mice (biological replicates) were used per experiment.

Immunohistochemistry

Mice were anaesthetized with an intraperitoneal (IP) injection of urethane, transcardially perfused and post-fixed at least 4 h in 4% paraformaldehyde. RVM, 40 and 25 μm thick, and spinal cord sections were collected on a cryostat and slide-mounted for immunohistochemistry. Sections were blocked at room temperature for 2 h in 5% donkey serum and 0.2% Triton X-100 in PBS. Primary antisera were incubated for 14 h overnight at 4°C: rabbit anti-red fluorescent protein (RFP; Rockland #600-401-379; 1:1K), chicken anti-green fluorescent protein (GFP; Aves Laboratory #GFP-1020; 1:1K), rabbit anti-tyrosine hydroxylase (Millipore #AB152; 1:1000), rabbit anti-tryptophan hydroxylase (Abcam #AB228588; 1:1000) or mouse anti-NeuN (Millipore #MAB377; 1:500). Sections were subsequently washed three times for 20 min in wash buffer (0.2% Triton X-100 in PBS) and incubated in secondary antibodies: donkey anti-rabbit (IgG) Alexa Fluor 555 secondary antibody (ThermoFisher #A-31572; 1:500), donkey anti-chicken (IgG) Alexa Fluor 488 secondary antibody (Jackson ImmunoResearch #703-035-155; 1:500), donkey anti-mouse (IgG) Alexa Fluor 647 (ThermoFisher #A-31571; 1:500) at room temperature for 2 h. Sections were then incubated in Hoechst (ThermoFisher, 1:10K) for 1 min and washed three times for 15 min in wash buffer, mounted and cover-slipped.

Fos experiments

Morphine sulphate (1 μg in 0.25 μl) was microinjected into the RVM. Twenty minutes thereafter, brain samples were collected and processed for FISH. Spinal cords were harvested 20 min and 90 min following CNO administration for FISH and immunohistochemistry, respectively.

Image acquisition and quantification

Full-tissue thickness sections were imaged using either an Olympus BX53 fluorescent microscope with UPlanSApo 4×, 10× or 20× objectives or a Nikon A1R confocal microscope with 20× or 60× objectives. All images were quantified and analysed using ImageJ. To quantify images in RNAscope in situ hybridization experiments, confocal images of tissue samples (3–4 sections per mouse over 3–4 mice) were imaged and only cells whose nuclei were clearly visible by DAPI staining and exhibited fluorescent signal were counted. To quantify spinal fluorescence, we measured the mean intensity of fluorescence within the spinal cord (both dorsal and ventral horns) using ImageJ.

Tissue clearing by iDisco and light sheet imaging

The iDisco+ protocol was followed as previously described.²⁸ Samples were immunolabelled with rabbit anti-RFP (Rockland #600-401-379) at 1:200 during a 7-day incubation at 37°C. Secondary antibody AlexaFluor 647-highly cross-adsorbed donkey anti-rabbit (Thermo Fisher, Cat #A-31573) was used at 1:400. iDisco-processed samples were imaged using light sheet microscopy (Miltenyi Biotec Ultramicroscope II with the 1× and 4× objectives). All images were acquired with illumination from a single side (three light sheets) with dynamic focusing (10 positions) enabled. The ImSpector software ‘Blend’ algorithm was used for dynamic focus processing. Images were stitched using Arivis Vision4D software.

RVM slice preparation and electrophysiology

Acute brain slices were generated from KOR^Cre mice in which spinally projecting RVM^KOR neurons had been labelled by two unilateral intraspinal injections of 300 nl each of AAVr-FLEX-tdTomato into the lumbar cord (as described above under the section ‘Intraspinal injections’). Under ketamine/xylazine anaesthesia, mice were perfused transcardially with ice-cold N-methyl-d-glucamine (NMDG) artificial CSF and horizontal sections of the brainstem, 300 µm thick, were cut on a Leica VT1200s vibratome. Slices were allowed to recover for 7 min in NMDG artificial CSF at 33°C and then transferred to normal artificial CSF to recover for 1 h at room temperature before recording. Whole cell recordings were obtained from fluorescent cells with borosilicate glass pipettes containing (in mM): 70 K-gluconate, 60 KCl, 10 HEPES, 1 MgCl₂, 0.5 EGTA, 2.5 Mg-ATP, 0.2 GTP–Tris, with pH 7.3 and 285 mOsm. Miniature inhibitory postsynaptic currents were isolated with bath application of (in µM): 0.5 tetrodotoxin, 20 cyanquixaline (CNQX) and 50 2-amino-5-phosphonovalerate. The µ-opioid receptor agonist DAMGO (Abcam) was bath applied at 10 nM. Data were acquired using MultiClamp 700B amplifier and Axon Digidata 1550B and stored on a computer running Clampex software (Molecular Devices, San Jose CA). Spontaneous inhibitory postsynaptic currents (sIPSC) were isolated with MiniAnalysis (Synaptosoft) and analysed in Prism (GraphPad, San Diego CA).

Spinal cord slice preparation and electrophysiology

Four weeks after surgery, spinal cord slices were prepared using methods described previously.²⁹ In brief, mice were deeply anaesthetized with a ketamine/xylazine cocktail (90 mg/kg and 10 mg/kg, respectively) and decapitated. The spinal cord was dissected in ice-cold NMDG artificial CSF containing (in mM): 93 NMDG, 2.5 KCl, 1.2 NaH₂PO₄, 30 NaHCO₃, 20 HEPES, 25 glucose, 5 sodium ascorbate, 2 thiourea, 3 sodium pyruvate, 10 MgSO₄.7H₂O, 0.5 CaCl₂.2H₂O, pH 7.3–7.4 with HCl. Parasagittal slices (200 µM) from lumbar spinal cord were made using a vibrating microtome (Leica VT1200s). Slices were transferred to oxygenated, room temperature artificial CSF containing (in mM): 124 NaCl, 2.5 KCl, 1.2 NaH₂PO₄, 24 NaHCO₃, 5 HEPES, 12.5 glucose, 2 MgSO₄.7H₂O, 2 CaCl₂.2H₂O, pH 7.3–7.4, and stored in an immersion incubation chamber for at least 1 h before recording.

Neurons were visualized using a 40× water immersion objective and infrared-differential interference contrast optics. Recordings were then made from randomly selected neurons within or dorsal to the substantia gelatinosa, easily visualized under brightfield illumination by its translucent appearance. Patch pipettes (6–9 MΩ) were filled with a CsCl-based internal solution containing (in mM): 130 CsCl, 1 MgCl₂.6H₂O, 10 HEPES, 2 Mg₂ATP, 0.3 Na₃GTP, 10 EGTA, pH 7.3 with CsOH. Data were captured using a Multiclamp 200B amplifier (Molecular Devices), digitized with an ITC-18 (Instrutech) and stored using Axograph X software (Molecular Devices).

Photostimulation of KOR^Cre fibres originating from the RVM was achieved using a high-intensity LED light source evoked by the data acquisition program (Axograph) delivered through the microscope’s optical path. Photostimulation (488 nm) of increasing duration was delivered at a rate of 0.5 Hz (1 ms, 5 ms, 50 ms, 500 ms and 1 s). Responses of dorsal horn neurons to photostimulation were recorded in voltage clamp with a holding potential of −70 mV. CNQX (10 µM, Sigma-Aldrich) was included in all recordings to ensure capture of inhibitory events only. Each light stimulation protocol was repeated 10 times per cell with a 7 s gap between each sweep. In a subset of recordings, bicuculline (10 µM, Sigma-Aldrich) was added to the bath solution. Data were collected from a total of nine animals, which on average yielded three slices per animal and two cells per slice.

All data were analysed offline using Axograph software. Connections between KOR^Cre fibres originating from the RVM and spinal cord neurons were counted if optically evoked currents were present in at least 50% of trials. Optically evoked inhibitory postsynaptic currents for each stimulation duration were averaged and maximum amplitude and latency from stimulus was measured. Optically evoked inhibitory postsynaptic current latency was also measured from individual trials to determine jitter (defined as 2 × SD) for each stimulus duration.

Statistical analysis

For behavioural, FISH and neurochemical experiments, statistical analyses were performed using GraphPad Prism 8. Values are presented as mean ± SEM. Statistical significance was assessed using tests indicated in the applicable figure legends. Significance was indicated by P < 0.05. Sample sizes were based on pilot data and are similar to those typically used in the field.

Data availability

Further information and requests for resources and reagents should be directed to and will be fulfilled by the corresponding authors and Lead Contact, Dr Sarah E. Ross (saross@pitt.edu).

Results

RVM^KOR neurons inhibit nocioceptive thresholds, itch and acute and chronic pain

To visualize and manipulate RVM^KOR neurons, we used the KOR^Cre mice generated by our group³⁰ and targeted the RVM in these mice through stereotaxic injections of Cre-dependent adeno-associated viruses (AAVs) encoding hM3Dq-mCherry Designer Receptors Exclusively Activated by Designer Drugs (DREADDs) for activation or mCherry alone as a control. We validated our injections using FISH, which revealed that mCherry-expressing RVM^KOR neurons comprised 88.6 ± 3.6% of Oprk1 neurons in the RVM and that these neurons were 100% specific to the Oprk1 population (Fig. 1A).

Figure 1.

Activation of RVM^KOR neurons is analgesic. (A) Validation of the viral approach (AAV-hSyn-DIO-hM3Dq-mCherry) to target RVM^KOR neurons using FISH. Data are mean + SEM with dots representing individual mice (n = 4). Scale bar = 50 μm. (B) FISH characterization of Oprk1-expressing neurons in the RVM relative to other markers of interest. Data are mean ± SEM with dots representing individual mice (n = 3–4 mice per group). (C) Effect of chemogenetic activation on locomotor activity in an open field test. Data are mean ± SEM with dots representing individual mice (n = 7 controls, n = 8 hM3Dq mice). NS indicates the results of unpaired t-test (ns P > 0.05). (D–H) Effect of chemogenetic activation of RVM^KOR neurons with IP CNO (5 mg/kg) on naive responses to (D) cold plantar assay, (E) Hargreaves assay, (F) chloroquine-evoked itch, (G) von Frey assay and (H) tail-flick assay. Data are mean ± SEM with dots representing individual mice (n = 7 controls, n = 8 hM3Dq). Asterisks indicate the results of two-way repeated-measures ANOVA with Bonferroni’s correction (**P < 0.01, ***P < 0.001, ****P < 0.0001). (I) Effect of activation of RVM^KOR cells on thermal and mechanical responses with intraplantar capsaicin-induced injury. Data are mean ± SEM with dots representing individual mice (n = 14 controls, n = 15 hM3Dq). Asterisks indicate the results of two-way repeated-measures ANOVA with Bonferroni’s correction (ns P > 0.05, ****P < 0.0001). (J) Effect of chemogenetic activation of RVM^KOR on mechanical hypersensitivity with spared nerve injury. Data are mean ± SEM (n = 14 controls, n = 15 hM3Dq). Asterisks indicate the results of two-way repeated-measures ANOVA with Bonferroni’s correction (ns P > 0.05, ****P < 0.0001). BL = baseline.

We then used FISH to characterize RVM^KOR neurons against several known markers of RVM populations that have previously been implicated in descending modulation. These markers included: serotonergic neurons (5HT; marked by Tph2), Gad65 (Gad2), neuronal nitric oxide synthase (Nos1), substance P (Tac1) and MOR (Oprm1).^{15,16,18–20,31–38} RVM neuronal nitric oxide synthase neurons are thought to mediate anti-nocioception through local RVM circuits and projections to cortical and subcortical areas.^34,35 Serotonergic neurons have classically been presumed to be NEUTRAL cells,³⁹ although optogenetic manipulations of neurons containing Tph have revealed their role in facilitating mechanical and thermal hypersensitivity.³¹Tac1 and Oprm1 neurons in the RVM are thought to facilitate nocioception^18,19,36,40 and the role of GABAergic neurons (Gad2 and Slc32a1/Vgat) remains controversial.^15,16 We found that the KOR (Oprk1) neurons in the RVM were diverse with respect to neurotransmitter, with 75–80% of Oprk1 neurons coexpressing Slc32a1 (Vgat) and Gad2, 25% coexpressing Slc17a6 (Vglut2), 46% coexpressing Tph2 (serotonin) and 30% overlapping with Penk (Fig. 1B). Moreover, approximately 30% of Oprk1 neurons coexpressed Oprm1 (MOR), consistent with prior reports.^17,41 In contrast, there was no overlap of Oprk1 with either Nos1 or Tac1 (Fig. 1B and Supplementary Fig. 1A). Thus, our FISH analysis indicated that although a majority of Oprk1 neurons in the RVM were inhibitory, they comprised a molecularly heterogeneous population.

Next, we tested the effect of chemogenetically activating RVM^KOR neurons. Because we observed diffuse innervation of the spinal cord (including both dorsal and ventral horns) as well as supraspinal areas (Fig. 1A, Supplementary Fig. 1B), we first tested whether mice would exhibit overt motor deficits in the open field test (Fig. 1C). We found that chemogenetic activation of RVM^KOR neurons did not affect performance in an open field assay. However, in somatosensory assays, chemogenetic activation of RVM^KOR neurons produced elevated thresholds to cold, heat and mechanical testing as well as decreased itch behaviour (Fig. 1D–G). Because RVM neurons have been traditionally classified based on their firing patterns in relation to the tail-flick response,^4,10,11 we performed this assay as well (Fig. 1H). Activation of RVM^KOR neurons likewise caused elevated tail-flick latencies, consistent with the notion that they could be OFF cells. To test the role of RVM^KOR neurons in ongoing pain, we analysed hypersensitivity following intraplantar capsaicin (acute pain) and following nerve injury (chronic pain). Following intraplantar capsaicin, mice developed thermal and mechanical hypersensitivity and this effect was reversed with activation of RVM^KOR neurons (Fig. 1I, Supplementary Fig. 2A and B). Similarly, following spared nerve injury, both control and hM3Dq groups exhibited pronounced mechanical allodynia, but this hypersensitivity was abrogated following chemogenetic activation of RVM^KOR neurons (Fig. 1J). Together, these data suggest that RVM^KOR neurons are sufficient for the inhibition of nocioception, itch and the reversal of pain induced by acute (capsaicin) and chronic (spared nerve) injury.

Microinjection of a kappa agonist into the RVM has previously been shown to block the analgesia driven by mu agonists within other sites of the CNS^18,22 and can precipitate mechanical allodynia following spinal nerve ligation.⁴² Thus, we sought to determine whether microinjection of a kappa agonist, U69,593, could affect naïve responses to nocioceptive stimuli (Fig. 2A). When microinjected into the RVM, U69,593 had a pro-nocioceptive effect and reduced the latencies to respond to cold and thermal stimuli (Fig. 2B and C) as well as mechanical thresholds (Fig. 2D). Consistent with previous studies,¹⁸ microinjection of U69,593 did not affect latencies in response to tail-flick (Fig. 2E). These data support that, at least for some stimuli, kappa-sensitive neurons in the RVM tonically suppress responses to nocioception that may be unmasked with their temporary pharmacological inhibition.

Figure 2.

Pharmacological and chemogenetic inhibition of RVM^KOR neurons facilitates nocioception. (A) Model to test whether the microinjection of a kappa agonist inhibits RVM^KOR neurons to facilitate nocioception. (B–E) Effects of microinjection of U69,593 (40 ng) into the RVM on sensitivities to (B) cold plantar assay, (C) Hargreaves assay, (D) von Frey assay and (E) tail-flick assay. Data are mean ± SEM with dots representing individual mice (n = 8 vehicle, n = 7 U69,593 treated animals). Asterisks indicate the results of unpaired t-test (ns P > 0.05, *P < 0.05, **P < 0.01). (F) Model of how a chemogenetic strategy using hM4Di (pink) or KORD (green) DREADDs mimics the effects of pharmacological inhibition of RVM^KOR neurons. Representative images of the RVM and spinal cord following infection are shown. Scale bar = 50 μm. (G–K) Effect of chemogenetic inhibition of RVM^KOR neurons with CNO (IP 5 mg/kg; pink, left) or Sal B (SC 10 mg/kg; green, right) in mice receiving hM4Di or KORD, respectively, on (G) cold plantar assay, (H) Hargreaves, (I) chloroquine-evoked itch, (J) von Frey assay and (K) tail-flick assay. Data are mean ± SEM with dots representing individual mice (n = 10 hM4Di control, n = 11 hM4Di, n = 7 KORD control, n = 8 KORD mice). Asterisks indicate the results of two-way repeated-measures ANOVA with Bonferroni’s correction (ns P > 0.05, *P < 0.05, **P < 0.01, ***P < 0.001, ****P < 0.0001). BL = baseline.

Kappa agonists delivered to the RVM could affect either RVM neurons or presynaptic terminals from neurons that project into the RVM. Moreover, although kappa agonists traditionally reduce neuronal excitability through Giα signalling, they have also been shown to activate cellular kinases that facilitate neuronal activity.^43,44 Thus, to specifically address the effect of inhibiting RVM neurons that express KOR, we sought to chemogenetically inhibit RVM^KOR neurons using the hM4Di DREADD and kappa-opioid receptor DREADD (KORD) (Fig. 2F). We found that inhibition of RVM^KOR neurons with either hM4Di or KORD enhanced sensitivities to cold, thermal and mechanical stimuli but did not enhance chloroquine-evoked itch (Fig. 2G–J). In the tail-flick assay, only inhibition with hM4Di, but not KORD, reduced tail-flick latencies (Fig. 2K). Silencing of RVM^KOR neurons with hM4Di facilitated itch (Supplementary Fig. 3A) but did not exacerbate hyperalgesia following acute injury with intraplantar capsaicin (Supplementary Fig. 3B–E). Chemogenetic inhibition with either hM4Di or KORD had no effect on locomotor activity in an open field or rotarod assay, but KORD inhibition resulted in elevated responses to the acoustic startle test (Supplementary Fig. 3F–H). In agreement with our pharmacological inhibition of RVM^KOR neurons (Fig. 2B–D), these experiments suggest that RVM^KOR neurons are required for the descending inhibition of nocioception, and their chemogenetic inhibition exposes a pro-nocioceptive state.

RVM^KOR neurons are required for stress-induced analgesia

In acute settings, stress inhibits pain⁴⁵ and stress-induced analgesia has been blocked by the RVM microinjection of a KOR agonist.^6,46 To test the contributions of RVM^KOR neurons to stress-induced analgesia, we used the forced swim model of stress-induced analgesia.^47,48 We looked at stress-induced analgesia by using the hotplate assay and by examining nociofensive responses to intraplantar capsaicin (Fig. 3A). In control mice, forced swim decreased licking behaviour and response latency on a hotplate, consistent with stress-induced analgesia. Upon activation of RVM^KOR neurons with CNO alone (no stress), hotplate responses were decreased, similar to stress-induced analgesia, and stress had no further effect (Fig. 3B and C). Next, we examined the effect of inhibiting RVM^KOR neurons on stress-induced analgesia. Following inhibition with hM4Di, mice no longer exhibited stress-induced analgesia to intraplantar capsaicin or hotplate testing (Fig. 3D–F). Together, these results suggest that RVM^KOR neurons are necessary and sufficient for stress-induced analgesia.

Figure 3.

RVM^KOR neurons are required for stress-induced analgesia. (A) Model to test stress-induced analgesia. Mice were forced to swim for 3 min and later tested either on a hotplate (total licking bouts and latency to lick or jump) or their licking responses to intraplantar capsaicin were quantified. (B and C) Effect of chemogenetic activation of RVM^KOR neurons (IP CNO 5 mg/kg) on responses to stress-induced analgesia as measured by (B) hotplate licking bouts and (C) hotplate licking latencies. Data are mean ± SEM with dots representing individual mice (n = 14 control, n = 15 hM3Dq mice). Asterisks indicate the results of two-way repeated-measures ANOVA with Bonferroni’s correction (ns P > 0.05, ***P < 0.001, ****P < 0.0001). (D–F) Effect of chemogenetic inhibition of RVM^KOR neurons (IP CNO 5 mg/kg) on stress-induced analgesia as measured by (D) nociofensive responses to capsaicin, (E) hotplate licking bouts and (F) hotplate licking latencies. Data are mean ± SEM with dots representing individual mice (n = 11 control, n = 12 hM4Di mice). Asterisks indicate the results of two-way repeated-measures ANOVA with Bonferroni’s correction (ns P > 0.05, **P < 0.01, ***P < 0.001, ****P < 0.0001). BL = baseline.

Targeting of spinally projecting RVM^KOR neurons

Given the robust innervation of the spinal cord by RVM^KOR neurons (Fig. 1A) and the role of the RVM in the descending modulation of pain, we sought to characterize the organization of RVM^KOR neurons that project to the spinal cord. Previous studies have reported variations in descending modulation between spinally innervated lumbar segments and uncovered the differential modulation of thermal sensitivity in the tail and foot.⁴⁹ Furthermore, RVM inactivation has been shown to completely block allodynia in facial regions but not plantar hindpaw in an inflammatory pain model.⁵⁰ Therefore, we examined the anatomical innervation of RVM^KOR neurons at different dermatomal segments. We used multicolour tracers with retrograde properties [AAVr-Ef1a-DIO-eYFP, AAVr-hSyn-DIO-mCherry and cholera toxin subunit B (CTB)-647] and injected these tracers at different segmental levels to examine whether RVM^KOR neurons are somatotopically organized (Fig. 4A). We found that RVM^KOR neurons were frequently infected by both eYFP and mCherry, suggesting that the same RVM^KOR neurons innervate both the cervical and lumbar spinal segments. This observation suggests that these neurons are not segmentally organized. From these experiments, we also determined that RVM^KOR neurons comprised 44.9 ± 1.5% of the total CTB-647-labelled spinally projecting RVM neurons, which underscores that descending RVM^KOR neurons represent a significant fraction of total descending neurons from the RVM (Fig. 4A).

Figure 4.

Activation of descending RVM^KOR neurons is analgesic. (A) Labelling of spinally projecting RVM^KOR neurons using retrograde AAV and CTB introduced to the lumbar and cervical spinal cords. Data are mean ± SEM with dots representing individual mice (n = 3–5). Scale bar = 50 μm. (B) FISH characterization of spinally projecting RVM^KOR neurons. Data are mean ± SEM with dots representing individual mice (n = 4). Scale bar = 50 μm. (C) Representative image of iDisco clearing of brainstem–spinal cord specimen following intraspinal injection of AAVr-FLEX-tdtomato. Scale bar = 1 mm (D) Approach to selectively activate spinally projecting RVM^KOR neurons using two complementary approaches. One involves intraspinal AAVretro-hM3Dq and RVM CNO (1 mmol in 300 nl). The other involves injection of AAV2-hM3Dq into the RVM with the activation of descending input with IT CNO (50 μg/kg). (E–H) Effect of chemogenetic activation of spinally projecting RVM^KOR neurons using RVM CNO or IT CNO on responses to (E) cold plantar assay, (F) Hargreaves assay, (G) chloroquine-evoked itch and (H) von Frey assay. Data are mean ± SEM with dots representing individual mice (AAVretro: n = 13 controls, n = 13 hM3Dq; AAV2: n = 7 controls, n = 8 hM3Dq). Asterisks indicate the results of two-way repeated-measures ANOVA with Bonferroni’s correction (ns P > 0.05, *P < 0.05, **P < 0.01, ***P < 0.001, ****P < 0.0001). (I) Effect of chemogenetic activation of spinally projecting RVM^KOR neurons with RVM CNO on thermal and mechanical responses with intraplantar capsaicin-induced injury. Data are mean ± SEM with dots representing individual mice (n = 13 controls, n = 13 hM3Dq). Asterisks indicate the results of two-way repeated-measures ANOVA with Bonferroni’s correction (****P < 0.0001). (J) Effect of chemogenetic activation of spinally projecting RVM^KOR neurons with RVM CNO on mechanical hypersensitivity with SNI. Data are mean ± SEM (n = 7 controls, n = 8 hM3Dq). Asterisks indicate the results of two-way repeated-measures ANOVA with Bonferroni’s correction (ns P > 0.05, ***P < 0.001). (K) Electrophysiological recordings of lamina II spinal neurons to examine descending inputs from RVM^KOR neurons following the injection of AAV2-Ef1a-DIO-ChR2-eYFP in the RVM of KOR^Cre mice. Representative traces are shown. Eleven of 48 cells in lamina II received input from RVM^KOR terminals following optogenetic activation (5 ms duration) and these responses are abolished in the presence of bicuculline (10 µM). All recordings were made in the presence of CNQX (10 µM). BL = baseline.

Next, we molecularly characterized these spinally projecting RVM^KOR neurons and found that they were virtually all GABAergic (based on their 100% overlap with Slc32a1) and that they exhibited limited overlap with other markers of RVM neurons (Fig. 4B). This finding is in striking contrast with RVM^KOR neurons as a heterogeneous group (Fig. 1B), suggesting that spinally projecting RVM^KOR neurons comprise a molecularly distinct population. Using this approach, we were also able to capture 88.1 ± 3.1% of Oprk1 neurons in the RVM with 100% specificity (Fig. 4B). Furthermore, we performed whole-tissue clearing of intact brainstem–spinal cord specimens in which descending RVM^KOR neurons were labelled with intraspinal AAVr-FLEX-tdtomato to visualize this pathway (Fig. 4C, Supplementary Video 1). Using this technique, we were able to label the cell bodies of RVM^KOR neurons that project to the spinal cord. These anatomical and molecular approaches, therefore, allowed us to selectively target, characterize and visualize descending RVM^KOR neurons.

Descending RVM^KOR neurons inhibit itch and nociception through a spinal circuit

Although our experiments suggested that RVM^KOR neurons inhibit nocioception and itch, the specific KOR neurons that were involved remained unclear. Our analysis had revealed that RVM^KOR neurons as a whole include GABAergic, glutamatergic and serotonergic subtypes; however, those that project to the spinal cord were exclusively GABAergic, and we hypothesized that it was these descending neurons that mediate the inhibition of nocioception and itch. To test this idea, we used two approaches to manipulate spinally projecting RVM^KOR neurons (Fig. 4D). In one set of experiments, we injected AAVr-hSyn-DIO-hM3Dq-mCherry into the lumbar spinal cord of KOR^Cre mice and implanted guide cannulas over the RVM, allowing us to infuse CNO directly into the RVM. In a complementary set of experiments, we tested animals receiving AAV2-hSyn-DIO-hM3Dq-mCherry into the RVM and administered IT CNO, which allowed us to spatially restrict our chemogenetic activation to RVM^KOR spinal projections. We confirmed that chemogenetic activation, with either approach, did not result in unusual motor or arousal behaviours on an open field, rotarod and acoustic startle test (Supplementary Fig. 2C–G). With both manipulations, activation of RVM^KOR cells inhibited responses to cold, thermal plantar testing and chloroquine-evoked itch (Fig. 4E–G, Supplementary Fig. 2H–K), but only activation of descending fibres with IT CNO reduced mechanical responses to von Frey testing (Fig. 4H). Lastly, activation of these descending neurons inhibited capsaicin-induced thermal and mechanical hypersensitivities (Fig. 4I, Supplementary Fig. 2L and M) as well as mechanical allodynia following chronic injury with SNI (Fig. 4J). These findings highlight the key role of spinally projecting RVM^KOR neurons in the inhibition of nocioception, itch and acute and chronic pain.

To determine the nature of RVM^KOR inputs to the spinal cord, we conducted recordings in spinal cord slices of KOR^Cre animals that had received AAV2-hSyn-DIO-ChR2-eYFP in the RVM (Fig. 4K, Supplementary Fig. 4A–D). We patched random cells in lamina II and found that nearly a quarter of these neurons received inhibitory inputs from RVM^KOR neurons upon optogenetic activation (Fig. 4K). We further validated that the nature of this input was inhibitory by blockade with bicuculline (Fig. 4K). Together, these findings suggest that the anti-nocioceptive effects of activating RVM^KOR neurons may occur through direct inhibition of spinal neurons.

We then applied our intersectional viral and pharmacological approach to determine the effects of inhibiting descending RVM^KOR neurons. We selectively removed the KOR from descending neurons by injecting AAVr-hSyn-eGFP-Cre into the lumbar spinal cord of KOR^fl/fl mice (Fig. 5A). Consistent with our previous findings that microinjection of U69,593 into the RVM is pro-nociceptive in wild-type mice (Fig. 2B–D), administration of U69,593 also generally unveiled a pro-nocioceptive state upon cold and mechanical testing, but did not result in thermal hypersensitivity in KOR^fl/fl mice (Fig. 5B–E). As expected from our previous experiments following microinjection of a kappa agonist in wild-type mice (Fig. 2E), local injection of U69,593 did not affect responses to the tail-flick test (Fig. 5E). However, following spinal delivery of AAVr-hSyn-eGFP-Cre in KOR^fl/fl mice, we found that kappa agonism no longer elicited cold and mechanical hypersensitivity (Fig. 5B and D). This highlights that expression of KOR on descending RVM^KOR neurons, and not non-spinally projecting RVM^KOR neurons or presynaptic KOR-containing terminals within the RVM, is required for the pro-nocioceptive effects following kappa agonism. Therefore, kappa agonists likely exert their pro-nocioceptive effects on KOR-containing RVM neurons that project to the spinal cord.

Figure 5.

Inhibition of descending RVM^KOR neurons facilitates nocioception. (A) Approach to selectively remove the KOR from descending RVM neurons. KOR^fl/fl animals received AAVr-hSyn-eGFP-Cre into the lumbar spinal cord. Validation of the deletion of KOR within the RVM was determined by FISH. Data are mean ± SEM with dots representing individual mice (n = 3 mice). Scale bar = 50 μm. (B–E) Effects of microinjection of U69,593 (40 ng) into the RVM before and after Cre-mediated deletion of KOR within the RVM. Sensitivities to (B) cold plantar assay, (C) Hargreaves assay, (D) von Frey assay and (E) tail-flick assay were tested. Data are mean ± SEM with dots representing individual mice (n = 9 vehicle, n = 7 U69,593-treated animals). Asterisks indicate the results of two-way ANOVA with Bonferroni’s correction (ns P > 0.05, *P < 0.05, **P < 0.01, ***P < 0.001). (F) Strategy to selectively inhibit descending RVM^KOR neurons using AAVr-hSyn-DIO-hM4Di-mCherry into the lumbar spinal cord followed by implantation of a guide cannula over the RVM to permit the local infusion of CNO. Representative targeting of the spinal cord and labelling of cell bodies within the RVM are shown. Scale bar = 50 μm. (G–J) Effect of chemogenetic inhibition of RVM^KOR neurons with local CNO (1 mmol in 300 nl) on (G) cold plantar assay, (H) Hargreaves, (I) chloroquine-evoked itch and (J) von Frey assay. Data are mean ± SEM with dots representing individual mice (n = 9 AAVretro control, n = 10 AAVretro hM4Di mice). Asterisks indicate the results of unpaired t-test (ns P > 0.05, *P < 0.05).

In a complementary set of experiments, we selectively inhibited descending RVM^KOR neurons by viral delivery of AAVretro-hSyn-DIO-hM4Di-mCherry into the lumbar spinal cord of KOR^Cre mice and local CNO into the RVM (Fig. 5F). The chemogenetic inhibition of descending RVM^KOR neurons increased sensitivity to cold and Hargreaves testing but did not affect itch or mechanical withdrawal (Fig. 5G–J). Furthermore, selective inhibition of spinally projecting RVM^KOR neurons attenuated stress-induced analgesia to capsaicin and hotplate testing (Supplementary Fig. 3I–K). Together, our combinatorial pharmacological and viral approaches to selectively inhibit descending RVM^KOR neurons support the idea that these neurons function to inhibit nocioception.

Dynorphin signalling between the PAG and RVM bidirectionally modulates nociception

Next, we investigated potential inputs to RVM^KOR neurons. As dynorphin is the endogenous peptide for the KOR, we injected Cre-dependent retrograde viral tracers into the RVM in Dyn^Cre mice to identify areas that could provide dynorphinergic inputs to the RVM (Fig. 6A). Robust labelling was identified in many areas, including the parabrachial nucleus (PBN), dorsal reticular nucleus (Drt), and the agranular insular area (AIp); however, the most robust labelling was in the ventrolateral PAG (vlPAG; Fig. 6A). Previous studies have highlighted the RVM as a necessary component within the descending modulatory circuit activated by stimulation of the PAG. For example, lesioning or inactivation of the RVM (with local anaesthetics) blocks the anti-nocioceptive action of PAG stimulation.^51–53 Therefore, we focused on RVM-projecting PAG^Dyn neurons as possible inputs to RVM^KOR cells. We characterized RVM-projecting PAG^Dyn neurons using CTB and AAVr-DIO-tdt in Dyn^Cre mice (Fig. 6B) and found that PAG^Dyn neurons represent 10.4 ± 4.0% of PAG neurons that project to the RVM. Using a complementary approach involving retrograde viral labelling and FISH, we found that this RVM-projecting population represented 32.3 ± 2.1% of PAG^Dyn neurons. Additionally, glutamatergic (Slc17a6-expressing) neurons comprised 99.3 ± 0.7% of all RVM-projecting PAG^Dyn neurons (Fig. 6C). Lastly, PAG^Dyn neurons represented a population that was molecularly distinct from PAG^Tac1 and PAG^Th neurons that have recently been described in the modulation of itch and nocioceptive behaviours, respectively^54,55 (Supplementary Fig. 5A–C). Together, these findings identify a unique population of glutamatergic PAG^Dyn neurons that project to the RVM.

Figure 6.

PAG^Dyn neurons provide input to the RVM and bidirectionally modulate pain. (A) Experimental strategy to trace dynorphinergic inputs to the RVM. Robust labelling was observed in the PAG of Dyn^Cre mice receiving AAVr-Ef1a-DIO-eYFP in the RVM. vlPAG: ventrolateral periaqueductal grey, PBN: parabrachial nucleus, AIp: agranular insular area, Drt: dorsal reticular nucleus. Scale bar = 50 μm. (B) Approach to characterize dynorphin neurons projecting to the RVM. In one experiment, Dyn^Cre mice received Cre-dependent AAVr-DIO-tdt in the RVM and RVM-projecting neurons were further characterized using FISH. In a separate experiment, Dyn^Cre; tdtomato mice received CTB and back-labelled CTB neurons were quantified. Scale bar = 50 μm. (C) Summary of RVM-projecting PAG^Dyn neurons characterization. Data are mean ± SEM with dots representing individual mice (n = 3). (D) Experimental strategy and effect of bidirectional modulation of PAG^Dyn neurons using chemogenetic activation (purple) and inhibition (green) of with IP CNO (5 mg/kg). Validation of viral targeting using FISH is shown. Data are mean ± SEM with dots representing individual mice (n = 3 mice). Scale bar = 50 μm. (E–I) Effect of chemogenetic activation with hM3Dq (purple, left) and inhibition with hM4Di (green, right) on (E) cold plantar assay, (F) Hargreaves assay, (G) chloroquine-evoked itch, (H) von Frey assay and (I) tail-flick assay. Data are mean ± SEM with dots representing individual mice (hM3Dq: n = 8 control, n = 6 hM3Dq; hM4Di: n = 7 control, n = 7 hM4Di mice). Asterisks indicate the results of two-way repeated-measures ANOVA with Bonferroni’s correction (ns P > 0.05, *P < 0.05, **P < 0.01, ***P < 0.001, ****P < 0.0001). (J) Experimental approach to selectively activate PAG^Dyn projections to the RVM by delivery of RVM CNO (1 mmol in 300 nl). Effect of chemogenetic activation on responses to Hargreaves, cold plantar and von Frey testing. Data are mean ± SEM with dots representing individual mice (n = 7 control, n = 9 hM3Dq mice). Asterisks indicate the results of unpaired t-test (ns P > 0.05, *P < 0.05). (K) Experimental approach to block dynorphin input from chemogenetically activated PAG^Dyn neurons by delivery of RVM norBNI (100 ng in 250 nl) and its effects on Hargreaves, cold plantar and von Frey testing. Data are mean ± SEM with dots representing individual mice (n = 7 control, n = 9 hM3Dq mice). Asterisks indicate the results of two-way repeated-measures ANOVA with Bonferroni’s correction (ns P > 0.05, *P < 0.05, **P < 0.01, ***P < 0.001, ****P < 0.0001). BL = baseline.

Recent work has revealed that glutamatergic and GABAergic PAG neurons divergently modulate nocioception and itch.^56,57 We bidirectionally modulated the dynorphin population of PAG neurons using Cre-dependent hM3Dq and hM4di in Dyn^Cre mice to activate and inhibit these neurons, respectively (Fig. 6D). Chemogenetic activation and inhibition with IP CNO bidirectionally modulated thresholds to cold plantar and Hargreaves testing (Fig. 6E and F) and scratching behaviours in response to chloroquine (Fig. 6G). Although activation of PAG^Dyn neurons reduced mechanical thresholds, inhibition did not elevate thresholds above baseline (Fig. 6H). Neither activation nor inhibition of PAG^Dyn neurons affected latencies to tail-flick (Fig. 6I). Overall, we found that activation of PAG^Dyn neurons generally facilitated nocioception and itch, whereas inhibition of these neurons reduced sensitivities to cold, thermal and itch responses.

We also confirmed that PAG^Dyn neurons facilitated nociception through their input to the RVM by implanting cannulas in the RVM for the local delivery of CNO (Fig. 6J). This allowed us to selectively activate RVM-projecting PAG^Dyn neurons. We found that the selective chemogenetic activation of this pathway generally recapitulated our findings following IP CNO (Fig. 6F and H) and facilitated sensitivity to thermal and mechanical testing, but not to cold testing (Fig. 6J).

To test more directly whether PAG^Dyn neurons could exert their effects through kappa signalling in the RVM, we used a pharmacological approach to block dynorphin input to the RVM from PAG^Dyn neurons (Fig. 6K). We found that IP CNO reduced latencies to Hargreaves testing and mechanical thresholds in the von Frey assay. However, microinjection of norBNI, a kappa antagonist, prior to the administration of IP CNO blocked the development of thermal and mechanical hypersensitivity (Fig. 6K). Thus, these data suggest that PAG^Dyn neurons facilitate nociception through kappa signalling in the RVM.

Mu agonists disinhibit RVM^KOR neurons

Previous models have suggested that morphine inhibits nocioception, at least in part, through the indirect activation of OFF cells in the RVM, potentially through a mechanism of disinhibition. In particular, OFF cells are proposed to be directly inhibited by kappa agonists, but disinhibited by mu agonists.^18–20 We set out to test whether the morphine-active cells are Oprk1-expressing neurons. As expected, microinjection of morphine into the RVM increased response latencies to thermal stimulation in the Hargreaves assay (Fig. 7A). This anti-nocioception was accompanied by the induction of Fos in 40.2 ± 9.9% of Oprk1-expressing neurons, which represented 40.9 ± 8.3% of Fos-expressing cells (Fig. 7B). When we recorded from spinally projecting RVM^KOR neurons by injecting AAVr-FLEX-tdtomato into the spinal cord of KOR^Cre mice, we found that DAMGO substantially reduced the frequency of miniature inhibitory postsynaptic currents in a subset of these cells (Fig. 7C, Supplementary Fig. 4E and F). Importantly, no RVM^KOR neurons showed a direct effect of DAMGO compared to baseline (Supplementary Fig. 4F). Thus, our findings support the idea that morphine enhances the activity of RVM^KOR neurons to produce anti-nocioception (Fig. 7D).

Figure 7.

RVM^KOR neurons are disinhibited by mu agonists. (A) Model to test if microinjection of morphine disinhibits RVM^KOR neurons to produce analgesia. Effect of microinjection of morphine (1 μg in 250 nl) into the RVM on latencies in the Hargreaves assay. Data are mean ± SEM with dots representing individual mice (n = 7 saline, n = 8 morphine-treated animals). Asterisks indicate the results of unpaired t-test (****P < 0.0001). (B) Fos expression is enriched in Oprk1 neurons following microinjection of morphine into the RVM. Data are mean ± SEM with dots representing individual mice (n = 4 saline, n = 4 morphine-treated animals). Scale bar = 50 μm. Asterisks indicate the results of unpaired t-test (*P < 0.05). (C) Recordings of labelled spinally projecting RVM^KOR neurons revealed a reduction in inhibitory synaptic transmission in the presence of DAMGO (10 and 100 nM). Of 11 cells, six showed a reduction in mIPSC frequency, but only four of those were significant at P < 0.05 according to a Kruskal–Wallis test. (D) Model for the role of RVM^KOR neurons in the descending modulation of nocioception. PAG^Dyn neurons provide inputs to the RVM. RVM^KOR neurons target the spinal cord and bidirectionally modulate nocioception. BL = baseline; Mor = morphine; Sal = saline.

Discussion

Our study identifies the role of RVM^KOR neurons in the inhibition of pain and itch. Previous work has also implicated RVM^KOR neurons as the neurons that are disinhibited to produce analgesia in response to mu agonists.^18–20 However, there was controversy as to whether RVM neurons responsive to kappa agonists could be pro- or anti-nocioceptive.^21,22 One limitation of previous studies is that pharmacological manipulation cannot distinguish between direct effects on RVM neurons^18–20 or presynaptic terminals.^21,22 Our neurochemical characterization of RVM^KOR neurons also reveals that they comprise a heterogeneous population that likely has diverse effects on the descending modulation of pain (that could include both pro-nocioception and anti-nocioception). However, our ability to specifically capture and manipulate spinally projecting RVM^KOR neurons identifies them as a molecularly distinct subgroup of all RVM^KOR neurons. We found that the targeting of a small number of RVM^KOR neurons that project to the spinal cord was sufficient to inhibit itch, produce anti-nocioception and modulate stress-induced analgesia.

We found that all RVM^KOR neurons that project to the spinal cord are GABAergic. However, the KOR-expressing subset only represents 40% of all spinally projecting GABAergic neurons in the RVM, suggesting that spinally projecting GABAergic neurons likely represent numerous cell types. Consistent with this idea, various groups that have modulated GABAergic RVM neurons had different findings. One study, using very similar approaches to those used in ours, uncovered the role of RVM^Vgat neurons in the facilitation of mechanical, but not heat, nocioception¹⁵; however, another study found RVM^Gad2 neurons (which capture a subpopulation of Vgat neurons) to be involved in the inhibition of heat nocioception.¹⁶ Our manipulation of inhibitory RVM^KOR neurons is consistent with the latter findings, where we found that activation of inhibitory cells was anti-nocioceptive. Thus, it is likely that GABAergic neurons in the RVM span several functional subgroups and that differences in behavioural findings may reflect the relative subpopulations that are captured. Future efforts involving intersectional genetic approaches will be required to identify and directly manipulate unique RVM subpopulations in freely behaving mice.

Our viral targeting of RVM^KOR neurons revealed that they did not exclusively innervate the dorsal horn. We observed projections from mCherry-labelled RVM^KOR neurons diffusely throughout the spinal cord, including the ventral horn. However, we tested our animals using a variety of assays including motor (open field test), coordination (decreased rotarod latency) and acoustic startle tests and did not find any deficits. Nevertheless, the functional role of RVM^KOR projections to spinal structures beyond the dorsal horn remains unknown.

Another outstanding question involves the identities of the neurons in the spinal cord targeted by descending RVM^KOR neurons. Our results suggest that neurons in the spinal cord receive inhibitory inputs from descending RVM^KOR neurons, but this downstream circuit was not explored in our study. RVM^KOR neurons may exert their influence by directly inhibiting spinal neurons involved in itch and pain, such as those containing NK1R.^58,59 In our recordings from LII neurons, we found that a quarter of spinal neurons received inputs from RVM^KOR neurons. We speculate that the spinal neurons targeted by RVM^KOR neurons may comprise a subpopulation of excitatory neurons.⁶⁰ Alternatively, RVM^KOR neurons may also inhibit primary afferents to suppress pain and itch by presynaptic inhibition,¹⁶ but this was not examined in our study.

Several groups have examined the contribution of the RVM to the state-dependence of somatosensation, including nocioception.^11,61–65 Our findings suggest that RVM^KOR neurons are engaged during states of acute stress and that they mediate stress-induced analgesia. Such a protective mechanism may explain, for example, how the survival of an organism depends on its ability to appropriately respond to nocioceptive stimuli in emergency situations. The precise circuitry for stress-induced analgesia remains to be further elucidated.

Dynorphin signalling within the brain has been shown to be involved in many behaviours, including stress, addiction and analgesia.^66,67 We found that the selective activation and inhibition of dynorphin neurons in the PAG facilitates and inhibits nocioception, respectively. Using pharmacological approaches, we determined that these PAG neurons influence nocioception through their release of dynorphin within the RVM. This finding, together with the observation that we and others have made that RVM neurons are sensitive to kappa agonism and antagonism, suggests that the descending modulation of pain is influenced by dynorphin signalling between the PAG and RVM. Identification of this dynorphinergic pathway may provide opportunities to take advantage of the body’s endogenous opioid system to manage pain. Harnessing the endogenous pain-modulatory system may reduce the need for pharmacological treatments, such as opioids, with fewer side effects and improved outcomes for the treatment of pain disorders.

Abbreviations

AAV =

adeno-associated virus

CNO =

clozapine N oxide

FISH

fluorescent in situ hybridization

IP

intraperitoneal

IT =

intrathecal

KOR

kappa-opioid receptor

PAG

periaqueductal grey

RVM

rostral ventromedial medulla.

Funding

Research reported in this publication was supported by the Virginia Kaufman Endowment Fund, NIH/NIAMS grant AR063772, NIH/NINDS grant NS096705 (S.E.R.), NRSA F31 grant F31NS113371 and NIGM/NIH T32GM008208 (E.N.) and ZIA-HD008966 (NICHD) to (C.E.L.P.).

Competing interests

The authors report no competing interests.

Supplementary material

Supplementary material is available at Brain online.