key: cord-0260536-tf3547nm
authors: Avraham, Oshri; Chamessian, Alexander; Feng, Rui; Yang, Lite; Halevi, Alexandra E.; Moore, Amy M.; Gereau, Robert W.; Cavalli, Valeria
title: Profiling the molecular signature of Satellite Glial Cells at the single cell level reveals high similarities between rodent and human
date: 2022-02-08
journal: bioRxiv
DOI: 10.1101/2021.04.17.440274
sha: 881caad26313cfd267f75fe4a5e762a0a13bbf25
doc_id: 260536
cord_uid: tf3547nm

Peripheral sensory neurons located in dorsal root ganglia relay sensory information from the peripheral tissue to the brain. Satellite glial cells (SGC) are unique glial cells that form an envelope completely surrounding each sensory neuron soma. This organization allows for close bi-directional communication between the neuron and it surrounding glial coat. Morphological and molecular changes in SGC have been observed in multiple pathological conditions such as inflammation, chemotherapy-induced neuropathy, viral infection and nerve injuries. There is evidence that changes in SGC contribute to chronic pain by augmenting neuronal activity in various rodent pain models. SGC also play a critical role in axon regeneration. Whether findings made in rodent model systems are relevant to human physiology have not been investigated. Here we present a detailed characterization of the transcriptional profile of SGC in mouse, rat and human at the single cell level. Our findings suggest that key features of SGC in rodent models are conserved in human. Our study provides the potential to leverage rodent SGC properties and identify potential targets in humans for the treatment of nerve injuries and alleviation of painful conditions.

Peripheral sensory neurons located in dorsal root ganglia (DRG) relay sensory information from the peripheral tissue to the brain. Satellite glial cells (SGC) are unique glial cells in that they form an envelope that completely surround each sensory neuron [25; 57-59] . This organization allows for close bidirectional communication between SGC and their enwrapped soma. SGC actively participate in the information processing of sensory signals [28] . Morphological and molecular changes are elicited in SGC by pathological conditions such as inflammation, chemotherapy-induced neuropathic pain as well as nerve injuries [1; 9; 23; 24; 31; 80] . These studies also point to the contribution of SGC to abnormal pain conditions under injurious conditions [19; 25; 28] .

Our recent analysis of SGC at the single cell level revealed that SGC share functional and molecular features with astrocytes [1; 2] . Despite great morphological differences, SGC and astrocytes share many signaling mechanisms, including potassium buffering via the inwardly rectifying potassium channel Kir4.1 and intercellular signaling via gap junctions [26] . Both cell types also undergo major changes under pathological conditions, which can have neuroprotective function but can also contribute to disease and chronic pain [26] .

Most of the available information on SGC function has been obtained in rodents, which limits the potential for clinical translation. Only a small number of studies have investigated the molecular characteristics of human SGC [21] . Some evidence points to a role of SGC in human pathological conditions. In patients with Friedreich Ataxia, an autosomal recessive neurodegenerative disease, SGC proliferate, form gap junctions and abnormal multiple layers around the neurons [33; 34] , likely leading to alterations in the bidirectional communication between SGC and neurons. SGC also play a role in viral infection such as herpes simplex virus or varicella zoster virus [16; 83] . In HIV-1 infection of macaques, the virus that causes AIDS, in which peripheral neuropathy and pain are common, an upregulation of GFAP in SGC was observed [44] . Hemagglutinating encephalomyelitis virus belongs to the family of coronavirus and 5 Accreditation of Laboratory Animal Care (AALAC) and conforms to the PHS guidelines for Animal Care. Accreditation -7/18/97, USDA Accreditation: Registration # 43-R-008. [8] [9] [10] [11] [12] week old female C57Bl/6 mice and adult male Lewis rats were used for single cell RNAseq studies.

L4 and L5 DRG's from mice (2 biological independent samples, n=3 mice for each sample) and rat (2 biological independent samples, n=2 rats for each sample) were collected into cold Hoechst dye was added to distinguish live cells from debris and cells were FACS sorted using MoFlo HTS with Cyclone (Beckman Coulter, Indianapolis, IN). Sorted cells were washed in HBSS + Hepes+0.1%BSA solution and manually counted using hemocytometer. Solution was adjusted to a concentration of 500cell/microliter and loaded on the 10X Chromium system. The mouse data set used in this study is the naïve data set used in our previous study [2] .

Single-cell RNASeq libraries were prepared using GemCode Single-Cell 3′ Gel Bead and Library Kit (10x Genomics). A digital expression matrix was obtained using 10X's CellRanger pipeline (Build version 3.1.0) (Washington University Genome Technology Access Center).

Quantification and statistical analysis were done with Partek Flow package (Build version 9.0.20.0417). Filtering criteria: Low quality cells and potential doublets were filtered out from analysis using the following parameters: total reads per cell: 600-15000, expressed genes per cell: 500-4000, mitochondrial reads <10%. A noise reduction was applied to remove low expressing genes < = 1 count. Counts were normalized and presented in logarithmic scale in CPM (count per million) approach. We applied variance stabilizing transformation to count data using a regularized Negative Binomial regression model (Seurat::SCTransform) followed by a removal of unwanted variation caused by known nuisance and/or batch factors (Scale expression).

Principal component analysis (PCA) using Louvain clustering algorithm was then undertaken followed by an unbiased clustering (Graph-based clustering) algorithm implemented in Partek.

Clustering was performed using Compute biomarkers algorithm, which compute the genes that are expressed highly when comparing each cluster. Seurat3 integration was used to obtain cell type markers that are conserved across samples and clusters were assigned to a cell population by at least three established marker genes. Clusters are presented in t-SNE (t-distributed stochastic neighbor embedding) plot, using a dimensional reduction algorithm that shows groups of similar cells as clusters on a scatter plot. Differential gene expression analysis performed using three different models: Compute biomarkers following regularized Negative Binomial regression, non-parametric ANOVA and the Partek algorithm GSA that integrate multiple statistical models.

Gene lists from each statistical model were intersected to remove potentially false positive genes.

The intersected listed were then applied for all downstream analyses. A gene was considered differentially expressed (DE) if it has a false discovery rate (FDR) step-up (p value adjusted). p ≤ 0.05 and a Log2fold-change ≥ ± 2. The DE genes were subsequently analyzed for enrichment of GO terms and the KEGG pathways using Partek flow pathway analysis. Partek was also used to generate figures for t-SNE and scatter plot representing gene expression.

For snRNAseq and TEM, human dorsal root ganglia (hDRG) were obtained from Anabios, 

To make the tissue suitable for nuclei isolation, the entire DRG was processed into smaller pieces by cryopulverization using the CryoPrep (Covaris; CP02). Nuclei were isolated according to Martelotto with some modifications [45] . We elected to apportion the cryopulverized tissue and process the portions in parallel using two different buffers in order to evaluate potential effects on nuclei representation. One homogenization buffer was EZ Nuclei Lysis Buffer (Sigma; NUC101-1KT) with 0.5% RNasin Plus (Promega; N2615), 0.5% SUPERase-In (ThermoFisher; AM2696) 8 and 1mM DTT [22] . The other homogenization buffer was CST buffer (NaCl2 146 mM, Tris HCl pH 7,5 10 mM, CaCl2 1mM, MgCl2 21 mM, 0.49% CHAPS (Millipore-Sigma), 0.01% BSA, 0.5% SUPERasin-in, 0.5% RNasin Plus), as described in Slyper et al. [68] . Using the EZ buffer, the samples were homogenized on ice using 18 strokes of Pestle A followed by 18 strokes of Pestle B. The homogenate was filtered through a 50 . m filter (Sysmex; 04-004-2327) into a 2 mL microcentrifuge tube (Eppendorf; 022431048). An additional 0.5 mL of homogenization buffer was used to wash the Dounce homogenizer and filter. The sample was then placed on ice while the remaining samples were processed. The sample was centrifuged at 500g at 4ºC for 5 mins to obtain a crude pellet containing spinal nuclei. The supernatant was removed and discarded, being careful to not disturb the pellet. The pellet was resuspended in 1.5 mL of Homogenization Buffer and allowed to sit on ice for 5 mins. at 500g, 4ºC for 5 mins. This wash step was repeated twice more for a total of 3 washes. The final pellet was resuspended in 0.5 mL of NRB containing 6 M 4′,6-diamidino-2-phenylindole (DAPI, ThermoFisher; D1306). The suspension was filtered through a 20 m filter (Sysmex; 04-004-2325) into a polypropylene tube and kept on ice. Using the CST buffer, the samples were homogenized on ice using 18 strokes of Pestle A followed by 18 strokes of Pestle B in 1 mL of CST buffer. The homogenate was filtered through a 50 m filter into a 15 mL conical. An additional 1 mL was used to wash the filter and then 3 mL of CST was added, bringing the total volume to 5 mL. The suspension was spun down at 500g for 5 mins at 4ºC. The supernatant was removed and the pellet was resuspended in 0.5 mL of CST containing 6 M of DAPI. The suspension was filtered through a 20 m filter into a polypropylene tube and kept on ice. Fluorescence Activated Nuclear Sorting (FANS) was performed to purify nuclei from debris on a FACSAria II (BD). Gates were set to isolate DAPI+ singlet nuclei based on forward scatter and side scatter as well as fluorescence intensity. The instrument was set to 45 pounds per square inch (psi) of pressure and a 85 nozzle was used, with sterile PBS sheath fluid.

Nuclei were sorted into a 1. 

To define the similarities and differences between SGC across different species, we performed snRNA seq of L4,L5 human DRG and scRNA-seq of L4,L5 mouse and rat DRG using the Chromium Single Cell Gene Expression Solution (10x Genomics) (Fig. 1A) . We chose to perform scRNAseq in rodents because we previously showed that this method efficiently captures Table 1 ). The cell clusters obtained from the mouse data set matched our previous dataset [1] ( Supplementary Fig. 1D ).

We previously showed that although the actual percentage of neuronal cells in the mouse DRG is about 12%, the number of neurons detected in our scRNAseq analysis was only about 1% [1; 2] , which might be a result of the dissociation protocol that is biased towards non-neuronal cells. Another possibility is neuronal damage during the tissue dissociation process or the fact that sensory neurons are relatively large cells and are less amenable for single cell studies. In the rat samples, only 0.5% of cells were neurons and no neuronal cells were detected in the human sample ( Fig. 1A) . Nevertheless, our protocol achieved recovery of SGC from all species with 42% in human, 55% in mouse and 74% in rat (Fig. 1A) , allowing us to compare the molecular profile of SGC across species.

We recently described that Fabp7 (Fatty acid binding protein 7) is a specific marker gene for SGC and that the FABP7 protein is highly enriched in mouse SGC compared to other cells in the DRG [1] . t-SNE plots overlaid for Fabp7 demonstrated that Fabp7 is also enriched in human and rat SGC (Fig. 1C ). To validate Fabp7 expression at the protein level, we performed immunostaining of DRG sections from mouse and human, which revealed specific FABP7

labelling of SGC surrounding sensory neurons in both species ( Fig. 2A,B) .

In adult animals, SGC tightly enwrap the soma of each sensory neuron [56; 57; 59] . The gap between SGC and the neuronal surface is only about 20 nm, which is similar to that of the synaptic cleft. This close association between the two cell types is essential for efficient mutual neuron-SGC interactions [25] . The detailed morphology of the human neuron-SGC unit has only been examined at the light microscopy level in human [21] . To further compare the SGC organization surrounding sensory neurons across species, we performed Transmission Electron Microscopy (TEM) of human and mouse DRG sections (Fig. 2C) , which demonstrated the tight contact between SGCs and neurons in both mouse and human and the increased human sensory neuron soma size, which can be up to 5 times larger than mouse sensory neurons ( Fig 2C) [21] .

Quantification of the number of SGC surrounding sensory neurons revealed that human sensory neurons are surrounded by significantly more SGC compared to mouse neurons (Fig. 2D ),

consistent with the observations that the number of SGC surrounding sensory neurons increases with increasing soma size in mammals [36; 56; 59] .

We next examined the expression of known SGC marker genes in rodent and human.

Cells in the clusters identified as SGC were pooled together. We identified 7,880 SGC in human, 3,460 SGC in mouse and 8,428 SGC in rat ( Fig. 3A-C) .

demonstrated that in all species, Fabp7 is expressed at high levels in a majority of SGC (mouse 92%, rat 88% and human 96% ( Fig.3A -D). Cadherin19 (Cdh19) has been described as a unique SGC marker in rat Schwann cell precursors [71] and in adult rat SGC [20] . We found that Cdh19 was expressed in most human SGC (98%), whereas only half of SGC in rodents expressed this gene (56% in mouse and 48% in rat ( Fig. 3A-D) . Glutamine synthetase (GS/Glul) has been suggested as an SGC specific marker in rat [46] and mouse DRG [30; 31] . Our previous finding indicated a nonspecific expression of Glul in almost all cells in the DRG at the transcript level [1] .

Our current analysis suggests differences in Glul expression between species with more than half of SGC expressing Glul in rodents (~60%), and only around 10% in human ( Fig. 3A-D) .

One of the characteristics of SGC, which is similar to astrocytes, is the ability to control the microenvironment via expression of transporters and ion channels [26] . The potassium channel Kir4.1/Kcnj10 is a known marker of SGC. Kir4.1/Kcnj10 is expressed in rat SGC and influence the level of neuronal excitability, which has been associated with neuropathic pain conditions [79] . Our analysis demonstrates that Kcnj10 is expressed in almost half of mouse SGC (42%), whereas it is expressed in fewer SGC in rat (6%) and human (17%) (Fig.3A-D) .

Interestingly, we found that the potassium channel Kir3.1/Kcnj3 is widely expressed in human (82%) but not in rodents SGC (Fig. 3D ). The diversity in potassium channel expression in SGC might suggest a special role for this channel in defining the physiological characteristics of SGC across species. Another potassium channel that has been shown to be expressed specifically in rat SGC is SK3/Kcnn3 [79] . Our analysis suggests that only a small subset of rat SGC (12%) and

human SGC (6%) express Kcnn3 (Fig.3D) , whereas mouse SGC do not express Kcnn3 (Fig.3D ).

Another main property of SGC that is shared with astrocytes is functional coupling by gap junctions, with SGC surrounding the same neuron connected by gap junctions [25; 29] . Gap junction protein alpha 1 (CX43/Gja1) is the most abundant connexin (Cx) and was shown to modulate pain responses [69] . We found that Gja1 is expressed in a majority of human SCG (70%) but less prevalent in mouse (30%) and rat (40%) (Fig. 3D ). Other known gap junctions proteins expressed in SGC include Cx32, followed by Cx30.2, Cx37, Cx26, Cx30, Cx45 and Cx36 [25] .

Although these gap junction genes were reported to be expressed in SGC, we found that less than 5% of SGC expressed them across all species, except for Cx30.2/Gjd3 which was expressed in 35% of mouse SGC and Cx45/Gjc1 that is expressed in 30% of human SGC (Fig. 3D ).

Membrane channels related to gap junctions are Pannexins (Panx), which do not form cell-to-cell channels but are highly permeable to ATP [28] . Pannexin1(Panx1) was reported to be expressed in sensory ganglia where it is increased in pain models [69] and there is evidence that Panx1 mediated ATP release is implicated in nociception [25] . Our analysis demonstrated moderate expression of Panx1 in human SGC (30%) with lower expression in mouse and rat ( Fig.3D ). These observations indicate variability in Gap junction and pannexin gene expression between rodent and human, which may suggest functional differences in SGC communication and function in nociception. SGC also express glial fibrillary acidic protein (GFAP) and similarly to astrocytes, GFAP expression is increased under pathological conditions, which can have a protective function [82] . In mouse, Gfap is one of the top upregulated genes in SGC upon nerve injury [1; 2; 10; 17], but it is not expressed in all SGC [2; 47] . In uninjured SGC, the distribution of Gfap was relatively low, with ~20% expression in rat, 1.5% in mouse and undetectable levels in human DRG (Fig. 3D ).

While SGC do not typically myelinate neuronal soma, except in the spiral ganglion [62] , some myelin associated genes such as Mpz, Mbp, and Plp1 are highly expressed in SGC [1] .

Proteolipid protein (Plp1) is the major myelin protein in the central and peripheral nervous system.

Plp1 is expressed in all rodents SGC and in more than half of human SGC (Fig.3D ). Another gene that is expressed in all rodent and human SGC is apolipoprotein E (ApoE) (Fig.3D ). APOE is a multifunctional protein, mainly involved in lipid synthesis and transport. High levels of APOE production occurs in the brain, where it is primarily synthesized by astrocytes [18] . We recently found that one of the genes enriched in mouse SGC is fatty acid synthase (Fasn) [1] , which controls the committed step in endogenous fatty acid synthesis [14] . Examination of Fasn transcript expression revealed high distribution in rat SGC (70%) and lower in mouse (40%) and human (33%) SGC (Fig. 3D) .

To validate the expression of unique SGC marker genes across species at the protein level, immunostaining for selected markers was performed in human and mouse DRG sections.

Immunostaining for FASN revealed its specific expression in SGC surrounding neurons (stained for TUJ1) in both human and mouse DRG tissue (Fig. 4A) . These results suggest that the expression of genes related to lipid metabolism and transport in SGC, such as Fasn are conserved between rodents and human. Gfap was detected only at low levels in our scRNAseq analysis and immunostaining with GFAP further demonstrated expression only in very few SGC in both mouse and human (Fig. 4B ). The SGC marker Glul was highly expressed in mouse SGC and less in human at the RNA level (Fig. 3D) . Staining for GLUL also revealed higher expression around most mouse SGC, with lower detection in human SGC (Fig.4C ). Another striking difference between human and mouse SGC is the expression of the potassium channel Kcnj3

(Kir3.1), which is highly expressed in human SGC but not in rodent (Fig 3D) . Immunostaining for Kcnj3 (Kir3.1) further confirmed expression in human SGC, with almost undetectable expression in mouse SGC (Fig. 4D) . Quantification of the percent of neurons associated with SGC expressing FASN, GFAP, GLUL or KCNJ3 (Fig. 4E ) confirms the important similarities and differences in ion channels and gap junction genes between human and mouse SGC that might impact their function in controlling neuronal activity and nociceptive thresholds.

To further examine the biological properties of SGC across species, we calculated the top differentially expressed genes in SGC in human (2,070 genes), mouse (1,622 genes) and rat (993 genes) (fold-change >1.5, significant differences across groups, and p < 0.05 compared to average gene expression in all other populations in the DRG in the same species). This analysis might be influenced by the fact that the representation of cell population differs in each data set ( Fig 1A) . Nonetheless this analysis allowed us to compare the three gene sets, which revealed 193 genes shared between SGC in human and rodents (Fig. 5A Fig. 2A) , consistent with our previous findings that 10% of top enriched genes in mouse SGC were shared with brain astrocytes [1] . We next examined if human SGC also share unique expressed genes with human mature and fetal astrocytes [84] . We found that human SGC shared 152 genes with human mature astrocytes, 107 genes with human fetal astrocytes and 260 genes were shared with both mature and fetal astrocytes (Supplementary Fig. 2A , Supplementary Table 2 ). Despite the diversity in morphology and signaling mechanisms between astrocytes and SGC, these results support that important parallels between these two cell types are conserved between mice and human.

We next compared human SGC to Schwann cells, the other main types of glial cells in the peripheral nervous system [81] . Schwann cells are divided to two types; myelinating (mySC) and

non-myelinating (nmSC). We identified both types of Schwann cells in our human scRNAseq data set and compared their similarity to SGC. We found that human SGC share 453 genes with nmSC, Table 4 ). Our recent work also identified that mouse SGC do not represent a uniform cell population and at least four subtypes exist [2] . One of the sublcuster (mSGC3) we defined showed the highest similarity to astrocytes and another subcluster (mSGC 4) resembled the most to mySC [2] . We found that human SGC expressed a unique gene set that showed the highest similarity to the mSGC3 ( Supplementary Fig. 2D, Supplementary Table 5 ). These results further support the high similarity at the gene expression level between SGC and astrocytes and extend this similarity to human tissue.

We next analyzed the enriched biological pathways using KEGG 2021 (Kyoto Encyclopedia of Genes and Genomes) and Gene Ontology (GO). We found that human and rodents SGC show enriched molecular functions (GO MF) related to enzymatic activity and ion channel and transport activity (Fig. 5B-D) . This further supports the important role SGC may play in human DRG to control neuronal excitability and pain thresholds. Both rodents and human SGC shared some cellular components (GO CC), including involvement in plasma membrane and synaptic activity (Fig. 5B-D) . The biological process enriched pathways (GO BP) in human and rodents were particularly similar, with enrichment for mainly metabolic pathways of lipids and cholesterol, as well as processes related to nervous system development ( Fig. 5B-D) . We previously revealed that fatty acid synthesis and PPARα signaling pathway were enriched in mouse SGC and those pathways were also upregulated after peripheral nerve injury in SGC [1], but not after central axon injury [2] . We demonstrated that PPARα activity downstream of fatty Table 2 ) [84] . Aquaporin-4 is a water channel predominantly found in astrocytes in the central nervous system and is believed to play a critical role in the formation and maintenance of the blood-brain barrier and in water secretion from the brain [51] , further highlighting that the similarity between SGC and astrocytes is conserved in human.

Examination of enriched molecular pathways (KEGG) of the enriched channels and receptors gene sets in SGC for each species revealed that the top pathways were related to neuroactive ligand-receptor interaction, GnRH secretion, insulin secretion and endocannabinoid signaling pathway (Fig. 6C ). Mouse and human SGC were also enriched for glutamatergic and dopaminergic synapse (Fig. 6C) . The similarity in ion channel and receptor composition between human and rodent SGC suggests conserved physiological roles of SGC across species. However, the differences in subset of channels and receptors may inform future study design for therapeutic development targeting SGC in pain conditions and other peripheral neuropathies.

SGC have been implicated in pain conditions related to viral infection such as herpesvirus, varicella zoster virus and also swine hemagglutinating encephalomyelitis virus, which is related to the coronavirus family [25] . Current models suggest that SGC surrounding virally infected neurons may restrict the virus spread [25] . A recent study suggested that sensory neurons could be potential targets for the infection of SARS-CoV-2, with SARS-CoV-2 gaining access to the nervous system through entry into nociceptor nerve endings in the skin and luminal organs [67] .

COVID-19, the disease caused by the SARS-CoV-2 can trigger many unexplained neurological effects including chronic pain. Price and colleagues found that a subset of human DRG neurons express the SARS-CoV-2 receptor angiotensin-converting enzyme 2 (ACE2) at the RNA and protein level [67] . DRG neurons also express SARS-CoV-2 coronavirus-associated factors and receptors (SCARFs), which were shown to be expressed in DRG at the lumbar and thoracic level as assessed by bulk RNA sequencing of human DRG tissue [67] . Having the resolution of single cell in human DRG, we assessed the expression of Ace2 and SCARF genes specifically in SGC.

While the Ace2 receptor was lowly expressed in SGC in all three species, the assembly/trafficking factors Rab10 and Rab1a, Rab14, RhoA and the restriction factors Ifitim3 and Ifitim2 were highly expressed (Fig. 7A,B , Table 3 ). Further examination of these abundant genes in human SGC revealed high similarities in expression across individual donors (Fig. 7B) . Interferon-induced transmembrane proteins (IFITMs) restrict infections by many viruses, but a subset of IFITMs can enhance infections by specific coronaviruses. Recently, it has been showed that human IFITM3

with mutations in its endocytic motif enhances SARS-CoV-2 Spike-mediated cell-to-cell fusion and thus raised the concept that polymorphisms in IFITM3 can positively or negatively influence COVID-19 severity [66] . Weather IFITM3 expression in SGC enhance or limit viral infection in sensory ganglia remains to be determined. Together, these results suggest that SARS-CoV-2 may gain access to the nervous system through entry into sensory neurons at free nerve endings in organs and that SGC may attempt to restrict the local diffusion of the virus [25; 40] .

The biology of SGC has remained poorly characterized under normal or pathological condition. Most of the current knowledge on SGC function stems from studies in rodents. Here we present a direct comparison of transcriptional profile of SGC in mouse, rat and human at the single cell level. Our findings suggest that key features of SGC in rodent models, such as similarities with astrocytes and enrichment for lipid metabolism and PPAR signaling, are conserved in human. However, notable differences exist in ion channels and receptors expression, which may suggest differences in SGC-neuron communication and functions in pain conditions and other peripheral neuropathies. Our study provides the potential to leverage rodent and human SGC properties and unravel novel mechanisms and potential targets for treating nerve injuries and other pathological conditions. Furthermore, depending on the specific target that is under study for therapeutic development, these parallels and differences should be considered in study design.

Our previous study demonstrated that SGC contribute to the nerve repair process in mice [1] and that the FDA approved PPARα agonist fenofibrate, which is used in dyslipidemia treatment, can increase axon regeneration after the dorsal root injury, a model of poor sensory axon growth [2] . Fenofibrate was surprisingly shown in clinical trials to have neuroprotective effects in diabetic retinopathy [5; 48] and traumatic brain injury [8] . Fenofibrate was also shown to exert analgesic and neuroprotective effects in rodent models of chronic neuropathic pain and inflammation as well as in some human studies [15; 53; 55] . The neuroprotective role of fenofibrate was also recently observed in a paclitaxel chemotherapy-induced peripheral neuropathy [7] . It is possible that this neuroprotection is mediated by an effect of PPAR activation in SGC. Together, these studies further support a central role for SGC in multiple pathological conditions affecting peripheral nerves. The observation that PPAR signaling is similarly enriched in human and rodents SGC opens the potential pharmacological repurposing of fenofibrate and that manipulation of SGC could lead to avenues to promote functional recovery after mechanical or chemical nervous system injuries.

Our study also highlights that the functional similarity of SGC and astrocytes is conserved across species. Both cell types undergo major changes under pathological conditions, which can have a protective function, but can also contribute to disease, and chronic pain [26] . One of the main functional similarities is buffering of extracellular potassium. In the CNS, glial buffering of extracellular potassium is carried out by astrocytes and consists of potassium uptake by inwardly rectifying potassium (Kir) channels [35] . Kir channels are key regulators of glial functions, which in turn determine neuronal excitability and axonal conduction [6] . Functionally, Kir channels can be divided into different subtypes based on their biophysical properties. Kir4.1 is an ATPdependent potassium channel, whereas Kir3.1 is a G protein-activated potassium channel [6] .

Most astrocytes express Kir4.1 but rat astrocytes and guinea-pig Muller glia have been shown to affects astrocyte ability to regulate neuronal activity [11] . The diversity in potassium channel expression in SGC might suggest that differences in signaling mechanisms related to ATP or G protein coupled receptors define the physiological characteristics of SGC across species.

We also observed an enrichment for other types of channels in human SGC, including sodium and water channels, chloride channels and TRP channels. Transient receptor potential (TRP) proteins consist of a superfamily of cation channels that have been involved in diverse physiological processes in the brain as well as in the pathogenesis of neurological disease. TRP channels are widely expressed in the brain, including neurons and glial cells. Channels of TRP family have been shown to be involved in sensation and modulation of pain in peripheral ganglia but their expression in SGC have not been demonstrated. TRP channels contribute to the transition of inflammation and immune responses from a defensive early response to a chronic and pathological conditions. The expression of purinergic and glutamate receptor was also highly conserved between human and mice, suggesting that the ATP and glutamate dependent communication between neuron and glia in response to neuronal activity and pathological conditions is largely conserved. Whether different channels are expressed in SGC surrounding different types of sensory neurons remains to be determined. In mice we detected at least 4 SGC subtype with no evidence that one of the subtypes is dedicate to one class of sensory neurons [2] . Spatial transcriptomics approaches or in situ hybridization to molecularly characterize transcriptomes of DRG and their adjacent SGC [74] might shed light on the molecular properties of individual neuron-SGC units in human.

Another main feature common to astrocytes and SGC and conserved across species is the enrichment for genes related to lipid metabolism and expression of ApoE. In astrocyte, lipid metabolism is critical for synapse development and function in vivo [3; 76] . ApoE is predominantly secreted by astrocytes in the brain and functions as a major transporter of lipoproteins between cells. Of the three ApoE alleles, the ApoE4 allele is associated with an increased risk for

Alzheimer's disease (AD) [18; 42] . ApoE likely regulates AD risk in large part via effects on amyloid pathology [38] . However, several studies revealed a role for ApoE in lipid delivery for axon growth [39; 49; 77; 78] . We previously found that ApoE expression is increased in SGC via activation of the PPAR signaling after nerve injury [1] . The pathophysiological changes in AD are believed to arise in part from defects in neuronal communication in the central nervous system [32; 65] . However, decline in different sensory modalities are suggested to be a primary first-tier pathology [12] . In cultured sensory neurons, exogenously applied ApoE4 directly inhibits neurite outgrowth, whereas ApoE3 stimulates neurite outgrowth [50] . These studies suggest that the 

The authors declare no conflict of interest

The raw Fastq files and the processed filtered count matrix for scRNA sequencing were deposited at the NCBI GEO database under the accession number GSE158892 (mouse), GSE169301 (rat and human). Data analysis and processing was performed using commercial code from Partek Table 1 Top 10 differentially expressed genes in each cell population compared to all other cell types in the DRG (fold change threshold >1.5) Table 2 Top normalized counts expression of ion channels and receptors genes in human SGC Table 3 SCARF genes expression in human SGC mouse SGC (40) Human SGC (57) Rat SGC (18) 

Satellite glial cells promote regenerative growth in sensory neurons

Profiling sensory neuron microenvironment after peripheral and central axon injury reveals key pathways for neural repair

Cholesterol--making or breaking the synapse

Loss of gait control assessed by cognitive-motor dual-tasks: pros and cons in detecting people at risk of developing Alzheimer's and Parkinson's diseases

Effect of fenofibrate on retinal neurodegeneration in an experimental model of type 2 diabetes

Inwardly rectifying potassium channels (Kir) in central nervous system glia: a special role for Kir4.1 in glial functions

Imad Damaj M. Targeting Peroxisome Proliferator-Activated Receptor-alpha (PPAR-alpha) to Reduce Paclitaxel-Induced Peripheral Neuropathy

Neurological recovery-promoting, anti-inflammatory, and anti-oxidative effects afforded by fenofibrate, a PPAR alpha agonist, in traumatic brain injury

The effects of axotomy on neurons and satellite glial cells in mouse trigeminal ganglion

Intraganglionic interactions between satellite cells and adult sensory neurons

Astroglial Kir4.1 in the lateral habenula drives neuronal bursts in depression

Dysfunctional Sensory Modalities, Locus Coeruleus, and Basal Forebrain: Early Determinants that Promote Neuropathogenesis of Cognitive and Memory Decline and Alzheimer's Disease

Human sensory neurons: Membrane properties and sensitization by inflammatory mediators

Structural and functional roles of ether lipids

Preferential PPAR-alpha activation reduces neuroinflammation, and blocks neurodegeneration in vivo

Human herpes simplex virus infections: epidemiology, pathogenesis, symptomatology, diagnosis, and management

Glial cell and macrophage reactions in rat spinal ganglion after peripheral nerve lesions: an immunocytochemical and morphometric study

The Role of APOE4 in Disrupting the Homeostatic Functions of Astrocytes and Microglia in Aging and Alzheimer's Disease

Satellite Glial Cells in Pain Research: A Targeted Viewpoint of Potential and Future Directions

Satellite glial cells represent a population of developmentally arrested Schwann cells

Massively parallel single-nucleus RNA-seq with DroNc-seq

Satellite glial cells in sensory ganglia: from form to function

Glial cell plasticity in sensory ganglia induced by nerve damage

Emerging importance of satellite glia in nervous system function and dysfunction

Satellite Glial Cells and Astrocytes, a Comparative Review

Hearing and dementia

Communication between neuronal somata and satellite glial cells in sensory ganglia

Dye coupling among satellite glial cells in mammalian dorsal root ganglia

Isolation of satellite glial cells for high-quality RNA purification

Changes in the transcriptional fingerprint of satellite glial cells following peripheral nerve injury

APP processing and synaptic function

Friedreich Ataxia: Hypoplasia of Spinal Cord and Dorsal Root Ganglia

Dorsal root ganglia in Friedreich ataxia: satellite cell proliferation and inflammation

Developmental and activity-dependent changes in K+ currents in satellite glial cells in mouse superior cervical ganglion

Ratios between number of neuroglial cells and number and volume of nerve cells in the spinal ganglia of two species of reptiles and three species of mammals

Relationship between postural instability and subcortical volume loss in Alzheimer's disease

Glial contributions to neurodegeneration in tauopathies

An apolipoprotein E-mimetic stimulates axonal regeneration and remyelination after peripheral nerve injury

Coronavirus infection of rat dorsal root ganglia: ultrastructural characterization of viral replication, transfer, and the early response of satellite cells

Association of Hearing Loss With Dementia

Alzheimer Disease: An Update on Pathobiology and Treatment Strategies

Association of Age-Related Hearing Loss With Cognitive Function, Cognitive Impairment, and Dementia: A Systematic Review and Meta-analysis

SIV-Induced Immune Activation and Metabolic Alterations in the Dorsal Root Ganglia During Acute Infection

Frankenstein' protocol for nuclei isolation from fresh and frozen tissue for snRNAseq V

Glutamine-, glutamine synthetase-, glutamate dehydrogenase-and pyruvate carboxylase-immunoreactivities in the rat dorsal root ganglion and peripheral nerve

Discrepancy in the Usage of GFAP as a Marker of Satellite Glial Cell Reactivity

In search for novel strategies towards neuroprotection and neuroregeneration: is PPARalpha a promising therapeutic target?

Neurite outgrowth stimulation by n-3 and n-6 PUFAs of phospholipids in apoE-containing lipoproteins secreted from glial cells

Differential effects of apolipoproteins E3 and E4 on neuronal growth in vitro

Specialized membrane domains for water transport in glial cells: high-resolution immunogold cytochemistry of aquaporin-4 in rat brain

Electrophysiological and transcriptomic correlates of neuropathic pain in human dorsal root ganglion neurons

Antinociceptive and antiedematogenic activities of fenofibrate, an agonist of PPAR alpha, and pioglitazone, an agonist of PPAR gamma

Mislocalization of Kir channels in malignant glia

Fenofibrate lowers atypical sphingolipids in plasma of dyslipidemic patients: A novel approach for treating diabetic neuropathy?

Number and Structure of Perisomatic Satellite Cells of Spinal Ganglia under Normal Conditions or during Axon Regeneration and Neuronal Hypertrophy

The satellite cells of the sensory ganglia

Perikaryal surface specializations of neurons in sensory ganglia

The structure of the perineuronal sheath of satellite glial cells (SGCs) in sensory ganglia

Diversity of Kir channel subunit mRNA expressed by retinal glial cells of the guinea-pig

Hearing loss and Alzheimer's disease: A Review

Qualitative and quantitative observations of spiral ganglion development in the rat

Human vs. Mouse Nociceptors -Similarities and Differences

The Effect of Hearing Aid Use on Cognition in Older Adults: Can We Delay Decline or Even Improve Cognitive Function?

Synapses and Alzheimer's disease

Opposing activities of IFITM proteins in SARS-CoV-2 infection

ACE2 and SCARF expression in human dorsal root ganglion nociceptors: implications for SARS-CoV-2 virus neurological effects

A single-cell and single-nucleus RNA-Seq toolbox for fresh and frozen human tumors

Gap junction mediated signaling between satellite glia and neurons in trigeminal ganglia

Auditory system dysfunction in Alzheimer disease and its prodromal states: A review

Identification of a novel type II classical cadherin: rat cadherin19 is expressed in the cranial ganglia and Schwann cell precursors during development

Peripheral inflammation suppresses inward rectifying potassium currents of satellite glial cells in the trigeminal ganglia

Inwardly rectifying potassium channel Kir4.1 is responsible for the native inward potassium conductance of satellite glial cells in sensory ganglia

Spatial transcriptomics reveals unique molecular fingerprints of human nociceptors

Hearing loss as a risk factor for dementia: A systematic review

Astrocyte lipid metabolism is critical for synapse development and function in vivo

The synthesis and transport of lipids for axonal growth and nerve regeneration

Evidence that the major membrane lipids, except cholesterol, are made in axons of cultured rat sympathetic neurons

Silencing the Kir4.1 potassium channel subunit in satellite glial cells of the rat trigeminal ganglion results in pain-like behavior in the absence of nerve injury

The contribution of satellite glial cells to chemotherapy-induced neuropathic pain

Redefining the heterogeneity of peripheral nerve cells in health and autoimmunity

Satellite cells surrounding axotomised rat dorsal root ganglion cells increase expression of a GFAP-like protein

Varicella-zoster virus infection of human dorsal root ganglia in vivo

Purification and Characterization of Progenitor and Mature Human Astrocytes Reveals Transcriptional and Functional Differences with Mouse