TMEM184B promotes pruriceptive neuron specification to allow itch sensitivity

Nociceptive and pruriceptive neurons in the dorsal root ganglia (DRG) convey sensations of pain and itch to the spinal cord, respectively. One subtype of mature DRG neurons, marked by Somatostatin (Sst) expression, is responsible for sensing mediators of acute itch and atopic dermatitis, including the cytokine IL-31. How itch-sensitive (pruriceptive) neurons are specified is unclear. Here we show that Tmem184b, a gene with roles in axon degeneration and nerve terminal maintenance, is required for the expression of a large cohort of itch receptors, including those for IL-31, Leukotriene C4, and Histamine. Mice lacking Tmem184b fail to respond to IL-31, but maintain normal responses to pain and mechanical force, indicating a specific behavioral defect in pruriception. Lineage-tracing studies using Sst-driven Cre recombinase show a loss of pruriceptive neurons in Tmem184b-mutant mice, suggesting a defect in neuron subtype specification. We identify an early failure of proper Wnt-dependent transcriptional signatures and signaling components in Tmem184b mutant mice that can explain the improper DRG neuronal subtype specification. Lentiviral re-expression of Tmem184b in mutant embryonic neurons restores Wnt signatures, whereas re-expression of Tmem184b in adult DRG fails to restore itch responses. Together, these data demonstrate that Tmem184b promotes adult somatosensation through developmental Wnt signaling and specification of pruriceptive neurons. Our data illuminate a new key regulatory step in the processes controlling the establishment of diversity in the somatosensory system.


Introduction
Neurons of the somatosensory system are responsible for detecting temperature, touch, pain, and itch. Somatosensory neurons also relay the first warning signals when peripheral tissues are damaged. For example, following a skin wound, local inflammatory processes occur near the terminals of somatosensory neurons, causing the release of cytokines and neuroactive peptides that can either directly depolarize neurons or result in their sensitization to other agonists (Dong and Dong, 2018;Solinski et al., 2019;Wilson et al., 2013). A common consequence of increased skin inflammation is the triggering of itch.
Acute and chronic itch are significant morbidities in humans. Itch-causing pathological conditions have a variety of causes, including skin disorders (atopic dermatitis), drug-induced reactions, systemic disorders such as liver disease, and other neurological disorders (Dong and Dong, 2018). The sensation of itch is distinct from pain and is carried by a subset of itch-specific somatosensory neurons called pruriceptive neurons with cell bodies in the dorsal root ganglia (DRG). Skin itch is triggered by release of pro-inflammatory compounds such as cytokines , thymic stromal lymphopoietin (TSLP), and histamine from mast cells and epithelial cells, which act upon membrane receptors and channels present in epidermal nerve endings (Oetjen et al., 2017;Solinski et al., 2019;Wilson et al., 2013). In addition to chemical itch, mechanical stimuli can evoke similar itch sensations but rely on a distinct spinal circuit (Acton et al., 2019;Bourane et al., 2015;Pagani et al., 2019;Pan et al., 2019).
To convey these sensations, somatosensory neurons display diversity in both protein expression and nerve terminal morphology. Somatosensory neurons have been classically categorized using soma size, electrical properties (including myelination state), and expression of a few markers that include lectins (IB4), neuroactive peptides, and ion channels. Recently, multiple groups have described the diversity of somatosensory neurons from the adult DRG using single-cell RNA sequencing, a powerful approach that has allowed an improved itch, but not pain, when Tmem184b is absent, showing that Tmem184b is required for proper pruriception. Tmem184b loss causes a reduction in the total number of NP3/C2 neurons in adults. We show that Tmem184b is required in embryonic neurons for expression of critical drivers of sensory neurogenesis, suggesting an early role for Tmem184b in sensory development. Rescue experiments in both adults and embryos support a primary role for Tmem184b in pruriceptive neuron specification through control of Wnt signaling. Our data identify a critical role for Tmem184b in the establishment of the neurons responsible for primary pruriception.

Tmem184b is Required for Expression of Pruriceptor-Specific Markers
In studying the nervous system phenotypes of loss of the transmembrane protein Tmem184b, which include axon protection as well as dysmorphic nerve terminals (Bhattacharya et al., 2016), we sought to identify expression changes that could contribute to these phenotypes. We performed RNA sequencing on isolated adult DRGs of 6-month-old Tmem184b global knockout mice and compared sequences to age-matched wild-type mice. We observed 316 genes with significantly downregulated (FDR ≤ 0.05) transcript expression in Tmem184bmutant ganglia, and 90 genes that had significantly upregulated transcripts (Fig. 1A). Strikingly, the downregulated genes contained a large fraction of markers previously identified in singlecell RNAseq studies as being unique to a population of pruriceptive neurons (called NP3 or C2 in previous studies) ( Chiu et al., 2014;Li et al., 2016b;Usoskin et al., 2015). The key markers of NP3/C2 neurons are dramatically downregulated, with reductions in Il31ra, Nppb, and Sst being among the largest changes seen in the dataset (Fig. 1B). However, other types of neurons also showed decreased expression of their unique markers, including NP2/C4 (identified by Mrgpra3) and NP1/C5-6 (identified by Mrgprd) (Fig. 1C). Taken together, this data identifies Tmem184b as a major regulator of somatosensory gene expression, particularly in pruriceptive populations.
To predict the functional significance of these expression changes, we performed pathway analysis using Panther. Downregulated transcripts were enriched in pathways such as GPCR signaling, serotonin receptor signaling, and synaptic transmission (Fig. 1D). When human phenotype ontology was examined, pruritus was the top and most significantly changed feature (p < 0.003). These data show that Tmem184b is required for appropriate transcription of signaling pathways critical to normal sensory function.

Tmem184b is required for itch, but not pain, responses
To test the behavioral consequences of the expression changes observed in Tmem184b-mutant mice, we challenged them with agonists that promote itch or pain. The cytokine IL-31, implicated in atopic dermatitis, activates the IL31RA/OSMR heterodimeric receptor on NP3/C2 DRG neurons, whereas the anti-malarial agent chloroquine (CQ) activates MRGPRA3 on NP2/C4 DRG neurons (Cevikbas et al., 2014;Dillon et al., 2004;Liu et al., 2009).
Tmem184b-mutant mice are resistant to scratching evoked by IL-31 injection ( Fig. 2A). In addition, mutant mice have a reduced response to chloroquine, although the reduction does not reach statistical significance (Fig. 2B, p = 0.07). To examine the effects on other nociceptive populations, we performed mechanical threshold and thermal pain testing. Tmem184b-mutant mice responded normally to these stimuli . Taken together, this data shows that Tmem184b is selectively required for behavioral responses to pruriceptive agonists.

mutant mice
To determine whether functional shifts in adult somatosensory neuron populations had occurred, we performed calcium imaging on dissociated adult DRG neurons. We saw significant reductions in the percentage of neurons that responded to the broad nociceptive receptor agonists, capsaicin (a TRPV1 agonist) and allyl isothiocyanate (AITC, a TRPA1 agonist) ( Fig.   2F-G). The percent of neurons responding to two agonists specific for the NP3/C2 population, CYM5442 (an S1PR1 agonist) and LY344862 (an HTR1F agonist), was reduced ( Fig. 2I-J).
These results suggest that cell loss contributes to our pruriceptive phenotypes. Paradoxically, mutant mice showed a small increase in the percentage of neurons responding to both IL-31 and chloroquine ( Fig. 2K-L). Because in wild type mice, S1pr1, Htr1f, and Il31ra are all unique markers of NP3/C2 neurons, we interpret these dynamic changes as evidence that receptor gene expression has been at least partially uncoupled with cell fate in

Tmem184b regulates key developmental pathways in somatosensory neurons
To evaluate whether the somatosensory neurons corresponding to the NP3/C2 population had truly been reduced, we sought to use a genetic fate-mapping strategy to label this population as soon as it arises. NP3/C2 DRG neurons are uniquely marked by Somatostatin expression, therefore we used Sst Cre combined with Cre-dependent tdTomato to identify this group of neurons. We found that the percent of adult neurons labeled with tdTomato is reduced but not completely eliminated in Tmem184b-mutant mice ( Fig. 3A-B). This result supports our calcium imaging data and indicates that reduction of this subpopulation of neurons contributes to the reduction of gene expression seen in RNAseq experiments. When taken together with the calcium imaging data, it suggests that reductions in both neuron numbers and response amplitude contribute to the loss of IL-31 induced pruriception.
We sought to identify the mechanism by which Tmem184b promotes NP3/C2 neuron specification by evaluating gene expression changes during the formation and differentiation of sensory ganglia. In mice, unique DRG subpopulations arise from a uniform group of neural crest precursors over a period of about three weeks in late embryonic and early postnatal life (from E10-P10) (Lallemend and Ernfors, 2012). Tmem184b is expressed in DRG neurons as early as E11.5 according to a recent single-cell RNAseq study (Sharma et al., 2020); we independently confirmed its expression at E13 in somatosensory ganglia (Fig. 3C). Embryos homozygous for the T184b GT allele show greater than 75% loss of Tmem184b transcripts in DRG at this timepoint. To identify the developmental pathways dependent on Tmem184b, we isolated RNA from embryonic day 13 (E13) dorsal root ganglia and compared transcriptional signatures between mutant and wild type ganglia using RNAseq. We identified 1635 genes for which transcripts significantly changed in the absence of Tmem184b at this timepoint (Adjusted P < 0.01) (Fig. 3D). Using pathway analysis, we identified multiple key developmental processes, including axonogenesis and neuronal development and differentiation, that are substantially affected by the loss of Tmem184b at E13 ( Fig. 3E-F). These results underscore that Tmem184b controls key aspects of the establishment of the somatosensory system. To analyze the trajectories of somatosensory neuron populations as they are adopting their adult fates, we combined our E13 analysis with a parallel transcriptomic evaluation of gene expression in developing ganglia at postnatal day 0 and postnatal day 10. At the latest time point, expression of NP3/C2 DRG markers has been previously observed (Lallemend and Ernfors, 2012). Somatostatin expression in wild-type began at P0 and was maintained at high levels through P10, when NP3/C2 neurons express the majority of their unique markers (Fig.   3J). In Tmem184b-mutant mice, Sst failed to turn on at P0 but had a burst of expression at P10.
This burst of Sst expression in mutants at P10 likely causes tdTomato expression, as these neurons would be permanently marked following the burst even though they express little Sst as adults (Sst transcript levels are 27% of wild-type in adult DRGs, Fig. 1A). In comparison, Il31ra expression is suppressed throughout these developmental timepoints in Tmem184b mutants ( Fig. 3K). This data argues for a significant loss of NP3/C2 neurons in Tmem184b-mutant mice.
Many other genes necessary for somatosensory development or mature function are also dysregulated (Supp Fig. 1). Taken together, our data supports a model in which Tmem184b, acting early in neuronal development, controls a Wnt-dependent developmental program leading to the proper specification of pruriceptive neurons.

Tmem184b controls expression of Wnt signaling components in embryonic DRG neurons
To more directly test if Tmem184b activity in DRG neurons controls critical developmental gene expression networks, we performed embryonic rescue of Tmem184b in vitro. We cultured E13 embryonic DRG neurons from wild type or mutant mice for 14 days (DIV 14, roughly equivalent timewise to postnatal day 8) (Fig. 4A). In a subset of neurons from each mutant embryo, we re-expressed Tmem184b using lentiviral transduction on the day of dissociation. At day 14, we collected RNA and performed RNA sequencing for each condition.
We identified differentially expressed genes in DIV 14 mutant neurons relative to wild type (Fig.   4B), and in mutant neurons with and without restoration of Tmem184b (Fig. 4C). Within these genes, we identified those that both decreased in the absence of Tmem184b (blue, Fig. 4B) and were increased in mutants upon re-expression of Tmem184b (gold, Fig. 4C). Our analysis resulted in the identification of 304 genes matching this pattern. Gene ontology analysis revealed biological processes significantly over-represented by these genes, representing processes positively regulated by Tmem184b activity (Fig. 4D). The most over-represented process among genes positively regulated by Tmem184b is the planar cell polarity (PCP) pathway (23.5-fold enriched), a non-canonical Wnt signaling pathway. Canonical Wnt signaling was also enriched in this group of genes (7.93-fold). Tmem184b regulated expression of components at multiple levels of the Wnt pathway, including receptors (Fzd1, Fzd10) as well as transcription factors (Tcf7l2, Taz) (Fig. 4E). Many more genes that positively influence Wnt signaling (green) were identified as compared to those that inhibit Wnt signaling (red). Of note, because these cultures were significantly enriched in neurons over glia (due to the presence of the anti-mitotic factor FdU for the duration of culture), it suggests that the expression changes to Wnt components are likely occurring within the DRG neurons themselves. Many of the processes controlled by Tmem184b also influence critical neuronal functions including calcium regulation (7.86-fold enrichment) and axon guidance (4.18-fold enrichment). This data implicates Tmem184b in controlling key developmental steps in the establishment of the somatosensory system. Our data support a model in which Tmem184b controls the proper expression of Wnt signaling components necessary for the cascade of developmental events leading to pruriceptive neuron differentiation.

In vivo re-expression of Tmem184b in adult mutant ganglia does not increase IL-31 responses
Given our results showing reductions in the NP3/C2 population and effects on Wnt signaling, we sought to evaluate whether Tmem184b contributes primarily to early neuronal specification or if it can influence pruriceptive function in adulthood. If Tmem184b's role is primarily developmental, then re-expression in adults should not be able to rescue IL-31 responsiveness. We constructed an adeno-associated (AAV9) virus encoding Tmem184b under the control of the neuronal-specific Synapsin promoter, and injected this virus, or control AAV9 expressing only mCherry, into the cervical intrathecal space (Fig. 5A). Histological analysis showed that virus was able to spread to DRG from C1-C10 (Fig. 5B) and also into some of the central axonal projections of these DRGs into the dorsal horn (Fig. 5C). Virus expression was also seen in the dorsal gray matter (Fig. 5C). In these mice, we evaluated scratching responses to subsequent IL-31 injection. We noted a slight, though statistically insignificant, rise in IL-31 induced scratching in all AAV-infected mutant mice, indicating that AAV injection and expression may influence the baseline levels of IL-31 responsiveness (Fig. 5D). Surprisingly, we found that in wild-type mice, overexpression of Tmem184b attenuates scratching to IL-31 injection.
However, we were unable to alter mutant responsiveness with adult re-expression of Tmem184b. The failure to rescue in adults is consistent with a primary effect of Tmem184b activity during development of pruriceptive neurons.

Discussion
Our data support a significant role for Tmem184b in the control of pruriception. Using a genetic labeling strategy, we found that NP3/C2 neuron numbers are reduced in the absence of Tmem184b, though they are not completely eliminated. Using calcium imaging, we found a reduction of nociceptive neurons generally and skewed numbers and/or sensitivity of remaining pruriceptive populations. In addition, at E13, we see reductions in transcripts encoding critical developmental factors including Wnt3a, Neurog1, and Neurog2. These three factors are essential for the establishment and differentiation of neurons within sensory ganglia (Lallemend and Ernfors, 2012). Based on this evidence, we believe that Tmem184b acts early in sensory development to promote the proper specification of nociceptive and pruriceptive neurons. It is possible that Tmem184b plays an additional role in adult nociceptor maintenance. In wild type mice, over-expression of Tmem184b causes reduction in IL-31-induced scratching (Fig. 5D). One possible explanation is that Tmem184b expression at high levels may be promoting apoptosis. In support of this model, we see decreased expression of apoptotic pathway components when Tmem184b is absent and restoration of these components upon Tmem184b re-expression in cultured DRGs (data not shown). Another possibility is that Tmem184b over-expression may promote degeneration of pruriceptive axons. Future experiments deleting Tmem184b solely in adults would be necessary to clarify roles for Tmem184b in long-term maintenance of pruriceptors.

Tmem184b-mutant
In summary, we show that Tmem184b activity critically affects the development of pruriceptive neurons in mouse DRG and that this effect is likely due to its ability to induce components of Wnt signaling during neurogenesis. Our data illuminates a new key regulatory step in the processes controlling the establishment of diversity in the somatosensory system.

Animal Models
All animal treatment was approved by the Institutional Animal Care and Use Committee at the

RNA sequencing
Total RNA was isolated from adult DRG from 6-month old mutant and wild-type mice (4 per genotype, mixed male and female groups) using the RNaqueous Micro kit (Ambion). All DRGs were pooled for each sample to obtain enough RNA for analysis. Following total RNA extraction, samples were ethanol precipitated to increase purity. Library preparation and sequencing was performed at the Washington University Genome Technology Access Center (GTAC). Data were analyzed using Salmon and DeSeq2 (on our servers or with Galaxy, www.usegalaxy.org). Volcano plots were generated in RStudio; heatmaps of genes for which adjusted P values were less than 0.05 were created using R or with Cluster 3.0 and Java 1994;Hargreaves et al., 1988) For pain and mechanical threshold testing, the experimenter was blinded to genotype, and males and females were separated during analysis.

AAV Injection
Custom AAV9 was built and titered by Vector Biolabs. Viruses were intrathecally injected (10ul, 2.0 x 10^6 virus particles/mL) at 8 to 9 weeks of age. Each mouse received intradermal (i.d.) saline injections 3 weeks after AAV delivery, and one week later received i.d. IL-31 injections.
One Tmem184b mutant mouse with ~10x average scratching behavior to both saline and IL-31 was treated as an outlier and removed from our analysis.

DRG Immunohistochemistry and Image Analysis
Isolated ganglia or spinal cords were fixed with 4% paraformaldehyde for 1 hour, immersed in 30% sucrose overnight (4°C), and embedded in OCT cryo-compound (Tissue-Tek® O.C.T.) using isopentane cooled with dry ice. Spinal cords were dissected into segments containing two spinal segments (e.g. cervical 1 and 2). DRG cryosections (14 µm thickness) or spinal cord sections (20 µm thickness) were-mounted onto charged microscope slides. Following washes in PBST and 1 hour of blocking in 5% goat serum in PBST, sections were incubated overnight at 4°C with rabbit NeuN (1:250, Proteintech), or for 1 hour with Alexa Fluor® 488 Mouse anti-β-Tubulin, Class III (BD Pharmingen). Secondary antibody for NeuN was goat anti-rabbit Alexa Fluor 633 (Thermo Fisher). Sections were mounted in Vectashield (Vector Biolabs). Spinal cords were imaged with a ZEISS AxioZoom V16 Fluorescent Microscope. DRGs were imaged using a ZEISS Axio Observer Z1 or LSM880 inverted confocal microscope. For DRG counting, only neurons with a visible NeuN-positive nucleus in the section were counted. For each genotype, n = 5 mice. For each mouse, ≥800 neurons were counted.

Neuronal Culture
For adult neurons, ganglia from all spinal levels were pooled in DMEM on ice, followed by digestion with Liberase TM (Roche) and 0.05% Trypsin (Gibco). Neurons were dissociated with a P1000 pipette tip, spotted in dense culture spots (20uL) on poly-D-lysine (Sigma) and laminin coated chambered coverglass (Nunc) or 100mm glass coverslips, and grown overnight in Dulbecco's Modified Eagle Medium (Gibco) with 10% fetal bovine serum (Atlas Biologicals) and Penicillin/Streptomycin (Gibco). For embryonic DRG culture, ganglia were dissociated with Trypsin, triturated with a P1000 pipette tip, and plated in 5-10 µl spots in a 24-well dish previously coated with poly-D-lysine and laminin. Media contained B27 (Gibco), 5-fluoro-deoxyuridine (FDU) and nerve growth factor (NGF) (Invitrogen). Half of the media volume was exchanged every 5 days until cells were collected for analysis.

Calcium Imaging
Adult DRG neurons were imaged either with Fluo-4 dye on a Zeiss Observer Z1 microscope, or with Fura-2 dye on an Olympus BXW microscope under a 10X immersion objective lens with a filter wheel and Hamamatsu camera, each with a frame rate of one image/3 sec. Data shown used ratiometric, perfusion based Fura-2 imaging with the exception of IL-31 and chloroquine, which used Fluo-4 due to quantity constraints (IL-31) or autofluorescence in UV wavelengths (chloroquine). Fluorescence videos were acquired via Zeiss or HCImage Software, processed and analyzed using MATLAB, and RStudio for quantification of fluorescent responses. We used a custom-written R script to identify neurons via responses to high potassium in each experiment. Responses to individual agonists were determined manually for all neurons. Each agonist was evaluated using cultures from at least 3 mice, and most agonist-genotype combinations have >1000 neurons analyzed (minimum 659, maximum 2333). Coverslips with extensive motion or other artifacts were excluded from analysis. Perfusion artifacts (single frame spikes) were identified and removed prior to analysis.

Data availability
The data sets generated and analyzed in the current study will be deposited at NCBI GEO upon publication and linked to this manuscript.