Novel inhibitory brainstem neurons with selective projections to spinal lamina I reduce both pain and itch

Abstract Sensory information is transmitted from peripheral nerves, through the spinal cord, and up to the brain (“bottom up” pathway). Some of this information may be modulated by “top‐down” projections from the brain to the spinal cord. Discovering endogenous mechanisms for reducing pain and itch holds enormous potential for developing new treatments. However, neurons mediating the top‐down inhibition of pain are not well understood, nor has any such pathway been identified for itch sensation. Here, we identify a novel population of GABAergic neurons in the ventral brainstem, distinguished by prodynorphin expression, which we named LJA5. LJA5 neurons provide the only known inhibitory projection specifically to lamina I of the spinal cord, which contains sensory neurons that transmit pain and itch information up to the brain. Chemogenetically activating LJA5 neurons in male mice reduces capsaicin‐induced pain and histamine‐induced itch. Identifying this new pathway opens new treatment opportunities for chronic, refractory pain, and pruritis.


| INTRODUCTION
Chronic pain and itch afflict millions of patients, and current treatments are largely ineffective (Mansfield et al., 2016;Yosipovitch & Bernhard, 2013). We do not fully understand the neural control of pain and itch, and are thus limited in developing therapies targeting the underlying circuitry in the brain or spinal cord. Current models show that lamina I of the spinal cord contains the neurons that transmit pain and itch information from the skin up to the brain (Andrew & Craig, 2001;Craig, 2002;Dong & Dong, 2018).
Most of what we know about the neural control of pain and itch involves this bottom-up pathway. Many studies have elucidated mechanisms by which the peripheral nervous system first detects pain and itch, then relays this information from skin to spinal cord (Caterina et al., 2000;Duan et al., 2014;LaMotte et al., 2014;Liu et al., 2009;Sun et al., 2009). However, it is not well understood how the brain modulates pain and itch sensations in a top-down fashion.
The simplest and most direct way for the brain to suppress pain and itch would be a direct, inhibitory projection to lamina I sensory neurons, but previously identified projections do not specifically target lamina I. The idea that our brains can suppress unpleasant sensations began with a report in 1969 that electrically stimulating the brainstem in rats could produce a surgical level of anesthesia, without anesthetic drugs (Reynolds, 1969). A roadblock to developing this observation is that the stimulated region of the brainstem does not send any inhibitory projections to the spinal cord. Prior investigators have focused on brainstem regions that project axons nonspecifically targeting many laminae of the spinal cord (Basbaum et al., 1978;Fields et al., 1991;Millan, 2002). In addition to top-down inhibition of pain, pharmacological studies suggest the possibility of top-down itch control, but this has no known neural substrate (Gotoh et al., 2011;Zhao et al., 2014). Therefore, the specific neural pathway the brain uses to modulate pain remains incompletely understood, and whether the brain can inhibit itch in a top-down fashion is unknown.
We discovered a novel group of inhibitory neurons in the ventrolateral brainstem that sends dense axonal projections specifically to lamina I of the spinal cord (Agostinelli et al., 2020). Given that lamina I neurons transmit pain, temperature, and itch (Craig, 2002;Dong & Dong, 2018), we hypothesized that these neurons inhibit those sensory modalities. To test this hypothesis, we used a genetically targeted, excitatory designer receptor exclusively activated by designer drug (DREADD) to test the ability of these neurons to reduce pain, itch, and temperature sensation.

| MATERIALS AND METHODS
Note: All experiments and analysis were completed before March 2020.

| Animals
To generate prodynorphin reporter mice, we crossed Pdyn-IRES-Cre mice with R26-lsl-L10-green fluorescent protein (GFP) mice, two lines of mice that were produced and characterized in the Lowell lab and were published previously (Geerling et al., 2016;Krashes et al., 2014).
Pdyn-IRES-Cre mice are available from The Jackson Laboratory (stock No. 027958). Pdyn-IRES-Cre mice selectively express Cre recombinase in cells that have expressed prodynorphin, and L10-GFP mice have a loxP-flanked STOP cassette which prevents the transcription of downstream GFP. We crossed these mice with Pdyn-IRES-Cre mice because in the presence of Cre recombinase, the STOP cassette is deleted, resulting in offspring that express GFP selectively in Creexpressing dynorphin cells, which we abbreviate as Pdyn-GFP.
We housed all mice on a 12:12 light:dark cycle with lights on at 6:00 a.m. at 22 C ambient temperature with ad libitum access to food and water. All protocols and care of the mice followed National Institute of Health guidelines and were approved by the University of Iowa Institutional Animal Care and Use Committee.
For stereotaxic delivery of these AAVs, we first anesthetized mice with ketamine/xylazine (100/10 mg/kg. i.p.) and unilaterally microinjected 90 nl of AAV-DIO-synaptophysin-mCherry into the right side of the hindbrain in the LJA5 region (coordinates from bregma: AP -4.83 mm, RL 1.35 mm, DV -5.50 mm from the top of the skull). We injected AAV-DIO-hM3Dq-mCherry bilaterally, using the same coordinates.
Four weeks after stereotaxic injection, we perfused the mice with AAV-DIO-synaptophysin-mCherry, and performed behavioral experiments on the mice with AAV-DIO-hM3Dq-mCherry before perfusion.
We later sectioned the brains and spinal cords at 30 μm into a 1:4 series on a freezing microtome.
We stained a series of spinal cord sections from AAV-DIO-synaptophysin-mCherry injected brains with dorsal root ganglia markers.
After immunolabeling, we mounted and dried sections on Superfrost Plus slides. For sections with fluorescence immunolabeling, we coverslipped with fade-retardant mounting medium containing DAPI (Vectashield; Vector Labs). For DAB-labeled tissue, we dehydrated sections in graded ethanols and cleared in xylenes before coverslipping with a toluene-based mounting media (Cytoseal; Thermo Scientific).
We acquired whole-slide images using a slide-scanning microscope (Olympus VS120) and reviewed slides in OlyVIA software (Olympus).

| Validation of neurochemical specificity
Pdyn-GFP mice were described previously (Geerling et al., 2016;Krashes et al., 2014). To confirm that in the LJA5 region specifically, GFP expression was limited to prodynorphin-expressing neurons, we performed in situ hybridization with RNAscope using a probe directed against mouse Pdyn (318771) and the Advanced Cell Diagnostics RNAscope multiplex V1 fluorescent assay.

| Analysis of injection sites
To measure the specificity of neuronal transduction, we immunostained sections for GFP (to enhance the native GFP of reporter neurons) and for mCherry, then counted the number of GFP+ cells that expressed mCherry across all sections containing LJA5 neurons (spanning typically six axial sections from a 1:4 series through the rostral hindbrain). Three mice injected with AAV-DIO-hM3Dq-mCherry had missed injection sites (only 10 or fewer mCherry+ cells) and were thus excluded from the study.

| Bouton counting of anterograde tracing
To estimate the relative density of LJA5 projections to efferent targets, we counted LJA5 boutons in each region. We used AAV-DIOsynaptophysin-mCherry cases because the mCherry (fused to a synaptically localized protein) robustly labels axon terminals and en passant synaptic boutons. We imaged whole brain and spinal cord sections with a slide-scanning microscope (Olympus VS120), using a ×40 objective and extended focal imaging (EFI). EFI takes a z-series of images through the tissue and combines them into one in-focus image. We loaded images of selected brain regions into Adobe Illustrator, and using the symbols tool, quantified the number of boutons in a given brain region. We quantified boutons in two sections containing the PAG terminal field, three sections through the PB, and five spinal cord sections (cervical, upper thoracic, lower thoracic, lumbar, and sacral levels). We added the bouton counts from sample sections through each region and multiplied by the total number of sections that contain the axon-terminal field in that region. Then, to estimate the relative proportion of boutons, we divided this number by the estimated total number of boutons across all three sample regions (PAG, PB, and spinal cord).

| DREADD behavior experiments
Behavior experiments were performed in male mice. Three weeks after stereotaxic microinjection of AAV-DIO-hM3Dq-mCherry, mice (14 weeks of age) were habituated to testing chambers, handling, and i.p. injections. One week later, we began behavior testing. Testing consisted of an i.p. injection of saline or clozapine-N-oxide (CNO) (1.0 mg/kg, 0.1 ml/10 g diluted in 0.9% saline) 30 min prior to testing. Tests were conducted in the light phase at approximately the same circadian time each day (ZT4 10:00 am). Each mouse was tested in a given experiment twice, and received both saline and CNO in a randomized order.
All behavior scoring was completed by research assistants blinded to the experimental conditions.

| Pain
Our testing chamber consisted of a clear glass floor with a camera below, and plexiglass dividers to separate the mice. Mice were habituated to the chamber for 30 min on three separate days. On test days, mice received either saline or CNO and were placed in the chamber for 30 min of habituation. We then injected 1 μg capsaicin in a volume of 10 μl into the plantar surface mice of the hindpaw (0.1% capsaicin dissolved in saline with 5% ethanol and 5% tween 80, randomized injections into left vs right hindpaw).
The mice were returned to the Hargreaves apparatus and recorded from below for 15 min. The amount of time spent licking and flinching the paw was scored. After 8 days, mice were given saline or CNO and then tested again with a capsaicin injection in the opposite hindpaw.

| Mechanical hypersensitivity testing
One hour after the capsaicin injection (90 min after the i.p. injection of saline or CNO), we placed mice on a wire mesh floor for 15 min of habituation. Then we performed a common test for mechanical hypersensitivity, the von Frey test (Chaplan et al., 1994). We applied graded mechanical pressure using calibrated nylon monofilaments (von Frey filaments) from below the wire floor to the hindpaws of the mice. This application of low-pressure von Frey filaments is innocuous, but if the animal experiences discomfort it will withdraw its paw (with a brisk flexion movement). We applied a series of increasingly stiff von Frey filaments (each filament applied eight times) to the plantar surface of the hindpaw, from below, until reaching a threshold pressure, defined as greater than or equal to 50% hindpaw withdraws for a given filament.

| Itch
At least 2 days before testing, we shaved a patch of hair off the nape of each mouse. On testing days, mice received either saline or CNO 30 min before testing, and were placed in the recording chamber for these 30 min for habituation. Intradermal (i.d.) injections of histamine (500 μg in 50 μl saline) were made into the nape of the neck, producing a 3 mm intradermal bleb. One trial each was excluded from two mice, then repeated, due to a subcutaneous injection (no intradermal bleb

| Ambulation
To measure mouse activity after CNO and saline i.p. injections, we used computer analysis of 10-min temperature preference video recordings when both Plates A and B set to the same preferred temperature (30 C). We used Ethovision software (Noldus) to calculate the total distance traveled and time spent moving.

| Statistical analysis
Data are presented as mean ± SEM. n represents the number of mice analyzed. Most statistical comparisons were conducted by two-tailed, paired Student's t test. We used two-way analysis of variance with Holm-Sidak's multiple comparisons test to compare histamine-induced responses. Power analysis was used to determine sample size using pilot histamine-induced scratching data and capsaicin-induced pain behavior; with an alpha of .05 and power of 0.8, the projected sample sizes were n = 8 and n = 9, respectively.

| LJA5 inhibitory brainstem neurons selectively project to lamina I of the entire spinal cord
We discovered prodynorphin-expressing neurons in the rostral, ventrolateral hindbrain (Figure 1(a)), which were not reported previously (Fallon & Leslie, 1986;Merchenthaler et al., 1997). We named this population LJA5 to describe their location in the lateral pons, juxta A5 (noradrenergic cell group). LJA5 neurons lie dorsal to the superior olivary complex. They are also immediately rostral to A5 and the exiting facial nerve root (Figure 1(a)). LJA5 neurons are separate from the catecholaminergic group A5 as they do not express TH (Figure 1 (b)). We discovered these neurons in Cre-reporter mice (Pdyn-GFP) made by crossing a prodynorphin-Cre mouse with an L10-GFP reporter mouse (see Methods) so that all neurons that have expressed prodynorphin will constitutively express GFP. We confirmed that LJA5 GFP+ neurons  (2020)). Conversely, none of the LJA5 GFP+ neurons express Slc17a6 (Vglut2, vesicular transporter for glutamate) (Figure 1(c)).
To investigate the efferent projections of these inhibitory neurons, we injected a Cre-dependent anterograde tracer, AAV-DIOsynaptophysin-mCherry, into the LJA5 region of Pdyn-GFP mice (Figure 2(a)). Four weeks later, we analyzed the brain and spinal cord for mCherry-labeled axons and presynaptic boutons, and found dense terminal fields throughout lamina I of the entire spinal cord and spinal trigeminal nucleus (although sparse labeling is scattered across deeper spinal lamina) (Figure 2(b)).
To confirm that fibers were limited to lamina I, we labeled markers of peripheral sensory axons (CGRP and IB4) that termi- (k) CNO activation of LJA5 neurons had no effect on latency of paw withdraw in the Hargreaves radiant-heat test (n = 7). (l) CNO activation of LJA5 neurons had no effect on behavioral temperature preference (n = 7). Y-axis is the percent of time spent on Plate A when Plate B is 30 C. (m,n) CNO activation of LJA5 had no effect on ambulation (n = 7)

| Activation of LJA5 neurons inhibits pain and itch behaviors
Lamina I neurons transmit pain, temperature, and itch (Craig, 2002;Dong & Dong, 2018)  were transduced with AAV-DIO-hM3Dq-mCherry in each mouse. To F I G U R E 4 Inhibition of pain and itch transmission by LJA5 neurons. Histamine and capsaicin in the skin are detected by receptors on DRG nerve fibers that transmit pain and itch information to the spinal cord and up to the brain. LJA5 Pdyn-expressing GABAergic neurons in the brainstem send inhibitory descending projections to sensory relay neurons in lamina I of the dorsal horn, lateral parabrachial nucleus (LPB), and periaqueductal gray (PAG). Activating this inhibitory "top-down" projection suppresses pain and itch signaling confirm that CNO activates hM3Dq-expressing neurons, we counted the number of transduced cells (mCherry+) expressing an activity marker, Fos, after injecting saline or CNO. Ninety minutes after injection, 1.9% ± 0.3 (saline) or 85.3% ± 1.4 (CNO) of LJA5 neurons with mCherry also expressed Fos (Figure 3(b)).
We then tested the effects of activating LJA5 neurons on pain, itch, and temperature sensation. The mice were used as their own controls, with i.p. injection of saline or CNO on different days, 30 min before each test. Activating LJA5 neurons with CNO markedly suppressed capsaicin-induced pain behavior and mechanical hypersensitivity (Figure 3(c,e,f)). CNO also reduced histamine-induced scratching (Figure 3(g,h)). To ensure that CNO itself is not analgesic or anti-pruritic, we made i.p. injections of saline or CNO in Pdyn-GFP mice without AAV-DIO-hM3Dq-mCherry, and we saw no differences in pain or itch behavior (Figure 3(d,i,j)). Finally, in contrast to the large reductions in pain and itch-associated behaviors, DREADD activation of LJA5 neurons did not alter temperature preference, reflexive thermal pain, or ambulation ( Figure 3(k-n)). These results indicate that LJA5 neurons specifically suppress pain and itch without altering thermosensory or motor functions.

| DISCUSSION
The idea that our brains can suppress unpleasant sensations began with a report in 1969 that electrically stimulating the brainstem in rats could produce a surgical level of anesthesia, without anesthetic drugs (Reynolds, 1969). A roadblock to developing this observation is that the stimulated region of the brainstem does not send any inhibitory projections to the spinal cord. Additionally, no lamina I-selective projection was identified until now. Here, we identify a group of prodynorphin-expressing GABAergic brainstem neurons with specific inhibitory projections to lamina I of the spinal cord. We have previously investigated the inputs and outputs of this group (Agostinelli et al., 2020), and here we further explore the efferent spinal projections. We further show that chemogenetic activation of these neurons reduces pain and itch, while sparing temperature preference or reflexive thermal responses. Inhibition by LJA5 neurons updates current models of pain and itch circuitry (Figure 4).

| Characterizing LJA5 neurons
Previously, the only known neuronal population in this region of the brainstem was the noradrenergic cell group, A5. A5 neurons also project to the spinal cord, but their projection to deep autonomic laminae is distinct from the lamina I-targeted projections of prodynorphinexpressing LJA5 neurons (Bruinstroop et al., 2012;Loewy et al., 1979). Also, LJA5 neurons are distributed rostral to A5 neurons, with the tail end of LJA5 neurons tapering medial to the facial nerve root exit, where rostral A5 neurons are located. Additionally, LJA5 neurons lack the catecholamine-synthesizing enzyme TH, which is expressed by A5 neurons (Figure 1(b)). Therefore, to describe their location in relation to A5, we named this population LJA5 as it sits in the lateral pons, juxta A5. We have additionally shown that Pdyn+ neurons in a similar location in human brainstems (Agostinelli et al., 2020).
We discovered these prodynorphin-expressing neurons in a Pdyn-GFP reporter mouse and showed that these GFP-expressing neurons express Pdyn mRNA (Figure 1(c)).
Previous studies in the rat brain did not comment on these neurons, although neurons containing Pdyn mRNA or dynorphin immunoreactivity can be seen above the superior olive (where we see LJA5 neurons in mice) in their figures (Fallon & Leslie, 1986;Merchenthaler et al., 1997).
Additionally, LJA5 neurons are both dynorphinergic and GABAergic (Figure 1(c), (Agostinelli et al., 2020). Whether the antipruritic and analgesic effects of activating LJA5 neurons are mediated by dynorphin, GABA, or another neurotransmitter remains unknown. Dynorphin is an endogenous, inhibitory opioid peptide located in many regions of the forebrain, brainstem, and spinal cord (Merchenthaler et al., 1997). Four identified opioid receptors respond to at least four different endogenous peptide ligands (Waldhoer et al., 2004). Dynorphin peptides derive from the precursor prodynorphin, and bind to the kappa opioid receptor, which is expressed by neurons in many regions, including lamina I of the spinal cord (Harris et al., 2004;Snyder et al., 2018;Waldhoer et al., 2004). Opioid agonists that bind to the mu opioid receptor (including morphine) are most commonly used for analgesic treatment, but these agonists have unpleasant and serious side effects including pruritis, constipation, and respiratory suppression (Chaney, 1995). Worse, due to their development of tolerance, mu opioid agonists are highly addictive (Waldhoer et al., 2004), with drugs like heroin and fentanyl contributing to an increasing epidemic of drug overdose deaths (Rudd et al., 2016). More recently, kappa opioid agonists have been shown to be antipruritic (Inan & Cowan, 2004;Kardon et al., 2014;Kumagai et al., 2012;Wikstrom et al., 2005) and analgesic (Ansonoff et al., 2006;Kivell & Prisinzano, 2010) in mice and humans, and may be a safer alternative to mu opioid agonists due to their low addictive potential and lack of respiratory depression.
In mice, intrathecal administration of kappa-opioid agonists reduces itch (Kardon et al., 2014). Initially, this finding led to the suggestion that dynorphin released by interneurons in the spinal cord inhibits itch (Kardon et al., 2014). Our discovery that dynorphinergic neurons in the brainstem project to lamina I of the spinal cord, and that activating these neurons inhibits pain and itch, suggests LJA5 as a key source of dynorphin for kappa-opioid inhibition of pain and itch sensory transmission. Our finding that LJA5 neurons can reduce both itch and pain holds immense therapeutic potential. They represent the first known source of dynorphinergic projections from the brain to the spinal cord for itch and pain modulation, and the first known GABAergic projection that selectively targets lamina I. LJA5 neurons therefore are a promising target for understanding endogenous modulation of noxious sensory circuits and treating patients with chronic pruritis or pain.

| Spinal projections
Perhaps the most remarkable feature of this group of neurons is that they are the first and only known group of inhibitory neurons that provide direct axonal projections selectively targeting lamina I of the spinal cord, whose neurons transmit pain, itch, and thermal sensory information from the skin. Outside lamina I, the only substantial LJA5 terminals we find in the spinal cord target the adjoining LSN, a loose collection of neurons in the dorsolateral funiculus. LSN neurons receive nociceptive input via primary afferents, similar to lamina I, but from subcutaneous muscle and visceral tissues (Neuhuber, 1982;Neuhuber et al., 1986;Sikandar et al., 2017).  (Basbaum et al., 1978;Fields et al., 1991;Millan, 2002). Additionally, the locus coeruleus sends noradrenergic projections to all laminae of the full spinal cord (Bruinstroop et al., 2012;Clark & Proudfit, 1991). These neurons are thought to mediate anti-nociceptive effects that can be modulated pharmacologically with anti-depressant medications (serotonin and norepinephrine reuptake inhibitors) sometimes used clinically to treat chronic neuropathic pain (Kremer et al., 2016). However, these drugs only partially reduce pain symptoms and are not effective in treating all types of pain (Gilron, 2016), suggesting that there are additional antinociceptive mechanisms.
Outside the brainstem, a small number of hypothalamic neurons also project axons to the spinal cord, without specifically targeting lamina I. A subset of neurons in the paraventricular hypothalamic nucleus sends axons to the IML and multiple laminae including I, II, and X (Saper et al., 1976;Swanson & McKellar, 1979). Orexin neurons of the lateral hypothalamus project to the spinal trigeminal nucleus (Peyron et al., 1998)  It should also be noted that the pain has been discussed as eliciting multidimensional responses, with emphasis on the reflexive verses affective/coping responses to pain (Fields, 1999;Han et al., 2015;Huang et al., 2019 Traditionally, sensory information was thought to be transmitted from the periphery to the brain via the spinothalamic pathway. However, multiple studies have shown that lamina I neurons of the spinal cord directly project to the PB (Cechetto et al., 1985;Craig, 1995;Panneton & Burton, 1985), comprising the spinoparabrachial pathway.
Additional studies confirmed connectivity between nociceptivespecific lamina I neurons and PB neurons (Bester et al., 2000;Light et al., 1993). Furthermore, PB neurons respond to pain (Campos et al., 2018), and PB inhibition is both antinociceptive (Han et al., 2015;Rodriguez et al., 2017) and antipruritic (Campos et al., 2018;Mu et al., 2017). Since LJA5 neurons send inhibitory projections to the PB, future studies will help clarify the role of LJA5 in modulating sensory transmission in the PB.
Studies on the descending control of pain have focused on the PAG projection to the RVM, which projects nonspecifically across the spinal cord (Fields et al., 1991;Millan, 2002). Additionally, there are direct projections from lamina I of the spinal cord to the PAG (Craig, 1995), implicating the PAG in both nociceptive reception and descending modulation. The PAG has mixed effects on pain and itch sensation; PAG stimulation induces analgesia (Reynolds, 1969;Yeung et al., 1977) and scratching behavior (Gao et al., 2019), and inhibiting PAG neurons reduces scratching in mice (Gao et al., 2019). Given this heterogeneity among PAG neurons in regards to nociception, further studies are required to clarify the functional relationship between LJA5 and PAG neurons.
LJA5 inhibitory axonal projections to the spinal cord are the first and only known top-down pathway by which the brain may directly and selectively inhibit pain and itch. This discovery represents a major development in our basic understanding of how the brain not only receives, but modifies noxious sensory information. Given that the LJA5 region receives input from many stress-related forebrain structures (Agostinelli et al., 2020), we suspect that descending stress signals activate LJA5 neurons, which would then in turn inhibit pain or itch. This phenomenon is known as stress-induced analgesia, and it might be evolutionarily beneficial animal can focus on behavioral responses during an acutely noxious situation without much distraction from pain or itch (Amit & Galina, 1988). Inhibition by LJA5 neurons updates current models of pain and itch circuitry and opens the possibility of novel analgesic and anti-pruritic therapeutics for patients with chronic pain and pruritis.