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

Sensory neurons with cell bodies in dorsal root ganglia (DRG) represent a useful model to study axon regeneration. Whereas regeneration and functional recovery occurs after peripheral nerve injury, spinal cord injury or dorsal root injury is not followed by regenerative outcomes. This results in part from a failure of central injury to elicit a pro-regenerative response in sensory neurons. However, regeneration of sensory axons in peripheral nerves is not entirely cell autonomous. Whether the different regenerative capacities after peripheral or central injury result in part from a lack of response of macrophages, satellite glial cells (SGC) or other non-neuronal cells in the DRG microenvironment remains largely unknown. To answer this question, we performed a single cell transcriptional profiling of DRG in response to peripheral (sciatic nerve crush) and central injuries (dorsal root crush and spinal cord injury). Each cell type responded differently to peripheral and central injuries. Activation of the PPAR signaling pathway in SGC, which promotes axon regeneration after nerve injury, did not occur after central injuries. Treatment with the FDA-approved PPARα agonist fenofibrate, increased axon regeneration after dorsal root injury. This study provides a map of the distinct DRG microenvironment responses to peripheral and central injuries at the single cell level and highlights that manipulating non-neuronal cells could lead to avenues to promote functional recovery after CNS injuries.


INTRODUCTION
Peripheral sensory neurons activate a pro-regenerative program after nerve injury to enable axon regeneration and functional recovery. In contrast, axons fail to regenerate after central nervous system injury, leading to permanent disabilities. Sensory neurons with cell bodies in dorsal root ganglia (DRG) represent one of the most useful models to study axon regeneration.
Sensory neurons send a single axon which bifurcates within the ganglion; one axon proceeds centrally along the dorsal root into the spinal cord and the other proceeds along peripheral nerves.
Whereas regeneration and functional recovery can occur after peripheral nerve injury, dorsal root injury or spinal cord injury is not followed by regenerative outcomes (Attwell et al., 2018;He and Jin, 2016;Mahar and Cavalli, 2018;Tran et al., 2018). This results in part from a failure of central injury to elicit a pro-regenerative response in sensory neurons (Attwell et al., 2018;Fagoe et al., 2014;Mahar and Cavalli, 2018;Tran et al., 2018).
The dorsal root injury is a useful model to understand how to promote axon growth into the central nervous system (Smith et al., 2012). Dorsal root disruption can occur in brachial plexus injuries, leading to paralysis of the affected arm (Smith et al., 2012). Regeneration following dorsal root crush can occur along the growth-supportive environment of Schwann cells, but stops as the axons reach the transition between the peripheral nervous system and the central nervous system, termed the dorsal root entry zone, where a variety of inhibitory factors block further growth (Smith et al., 2012). However dorsal root axonal growth occurs only at half the rate of peripheral axons (Oblinger and Lasek, 1984;Wujek and Lasek, 1983). The histological difference between dorsal roots and peripheral nerve is not sufficient to alter the rate of axonal regeneration (Wujek and Lasek, 1983). Rather, the availability of trophic factors and other target derived influences via the peripheral axon were suggested to prevent the upregulation of pro-regenerative genes such as Jun (Broude et al., 1997) or Gap43 (Schreyer and Skene, 1993). Interestingly, dorsal root injury causes up-regulation of the pro-regenerative gene Atf3 (Huang et al., 2006), but only in large diameter neurons, whereas Atf3 and Jun are upregulated in a majority of neurons after 4 peripheral nerve injury (Chandran et al., 2016;Renthal et al., 2020;Seijffers et al., 2007;Tsujino et al., 2000). Spinal cord injury also leads to activation of Atf3 in large diameter neurons, but this is not sufficient to promote regenerative growth (Ewan et al., 2021). Another possibility explaining the slow growth capacity of axons in the injured dorsal root is the contribution of non-neuronal cells.
Regeneration of axons in peripheral nerves is not cell autonomous. At the site of injury in the nerve, Schwann cells (Jessen and Mirsky, 2016) and macrophages (Zigmond and Echevarria, 2019) contribute to promote axon regeneration. In the DRG, macrophages are involved in eliciting a pro-regenerative response after peripheral but not central injury (Kwon et al., 2013;Niemi et al., 2016;Niemi et al., 2013;Zigmond and Echevarria, 2019), with anti-inflammatory macrophages believed to be more involved in the regenerative process than pro-inflammatory macrophages (Zigmond and Echevarria, 2019). We recently revealed that satellite glial cells (SGC), which completely surround sensory neuron soma, also contribute to promote axon regeneration (Avraham et al., 2020). PPARα signaling downstream of fatty acid synthase in SGC promote axon regeneration in peripheral nerves, in part via regulating the expression of pro-regenerative genes in neurons, such as Atf3 (Avraham et al., 2020). Whether the different regenerative capacities after peripheral or central injury result, at least in part from a lack or an altered response of macrophages, SGC or other non-neuronal cells in the DRG microenvironment remains largely unknown.
To answer this question, we performed a comprehensive single cell transcriptional profiling of DRG cells after peripheral injury (sciatic nerve crush) and central injuries (dorsal root crush and spinal cord injury). Sciatic nerve crush injures approximately half the axons projecting into the peripheral nerves (Laedermann et al., 2014;Rigaud et al., 2008) and dorsal root crush injures all axons projecting into the dorsal root. Dorsal column lesion of the spinal cord damages the ascending axon branches of most large diameter neurons and leaves the descending axon branches in the spinal cord intact (Attwell et al., 2018;Niu et al., 2013;Zheng et al., 2019). We 6 2020). Contralateral uninjured DRG were used as control and referred thereafter to naïve. The sciatic nerve is composed of axons projecting from sensory neurons residing in multiple DRG, and SNC results in ~50% of lumbar DRG neurons being axotomized (Laedermann et al., 2014;Renthal et al., 2020;Rigaud et al., 2008). SNC is followed by activation of a pro-regenerative program that allows functional recovery (He and Jin, 2016;Mahar and Cavalli, 2018). SCI injures the ascending axons of a subset of large diameter sensory neurons, leaving the descending axon branches in the spinal cord intact (Attwell et al., 2018;Niu et al., 2013;Zheng et al., 2019), and is not followed by regenerative outcomes. DRC damages all centrally projecting sensory axons in the PNS, without causing an impassable glial scar, and is followed by a slower regenerative growth compared to SNC that stops as axons reach the scar-free dorsal root entry zone (Oblinger and Lasek, 1984;Smith et al., 2012;Wujek and Lasek, 1983), providing an additional model to unravel the mechanisms promoting axon regeneration. The percent of DRG neurons lesioned under the three injury paradigms and the distance of the injury to the DRG may impact the injury responses of the microenvironment. However, all three injury paradigms are widely used models to study the mechanisms promoting axon regeneration.
Our scRNAseq protocol achieves efficient recovery of non-neuronal cells compared to other protocols that use single nuclear RNAseq to analyze neuronal responses to injury (Avraham et al., 2020;Renthal et al., 2020). While scRNAseq captures transcriptional responses, changes in RNA stability may also contribute to the differential profile and the depth of sequencing obtained in scRNAseq analyses might not allow to capture low level transcripts. The number of total sequenced cells from all conditions was 25,154 from 2 biological replicates for naïve, SCI and Schwann cells, macrophages and SGC (FDR≤0.05, FC≥2). Heat map of differential gene expression in the indicated cell types revealed that the magnitude of gene expression changes was the largest after SNC, but also occurred after DRC and SCI ( Figure 1I and Figure 1-Source Data 1), as previously suggested (Palmisano et al., 2019;Stam et al., 2007).
To further investigate how the neuronal microenvironment is affected by the different injuries, we performed cell-cell interaction analysis based on ligand-receptor expression in the different cell types for every injury condition using CellPhoneDB repository (Figure 1-Source Data 2). This analysis revealed that the cell-cell interaction network changed significantly after SNC compared to naïve, and that these changes are distinct from those elicit by DRC. SCI had limited influence on the cellular network interaction compared to naïve (Figure 1-figure supplement 1E).
This analysis further highlights the importance of the microenvironment response and the potential extrinsic influence on axon regeneration.

Alterations in blood-nerve-barrier markers in response to central and peripheral injuries
Blood-tissue barriers play an essential role in the maintenance and homeostasis of the tissue environment. Integrity of the peripheral nervous system is maintained by the blood-nervebarrier (BNB), which shares many structural features with the blood brain barrier (Richner et al., 2018). An essential component of the BNB cellular architecture is tight junctions (TJ) in the endoneurial vascular endothelium or the perineurium that surrounds the nerve fascicle.
Endothelial cells comprise the inner lining of vessels, while pericytes encompass blood microvessels such as blood capillaries (Sims, 2000). Sensory ganglia are highly vascularized ( Figure 2A) (Jimenez-Andrade et al., 2008), with blood vessels in sensory ganglia being more permeable than their counterpart in the brain (Kiernan, 1996;Reinhold and Rittner, 2017) or the nerve (Hirakawa et al., 2004;Jimenez-Andrade et al., 2008). Unlike in the brain, pericytes do not fully cover the blood vessel in peripheral nerve (Stierli et al., 2018). We observed a similar situation in the DRG, with the presence of blood vessel not fully covered by pericytes ( Figure 2B).
We examined changes in gene expression that occurred in endothelial cells and pericytes following peripheral and central injuries (FDR≤0.05, FC≥2) (Figure 2-Source Data 1), as the magnitude of gene expression changes was the largest in these cells after SNC ( Figure 1I). t-SNE plots of endothelial cells and pericytes demonstrated different clustering of cells after SNC or DRC, while similar clustering in naïve and after SCI were observed ( Figure 2C,D). Increased BNB permeability in the nerve is linked to changes in the expression of TJ genes, in particular a reduced expression of ZO-1 (Tjp1) in endoneurial cells (Richner et al., 2018). We thus examined the expression of tight junction (TJ) as well as adherens junction (AJ) genes. Heat map of TJ and AJ genes indicated that the response of barrier components was affected by SNC differently than DRC, with numerous junction genes being differentially expressed following SNC and DRC compared to naive and SCI condition ( Figure 2E,F). Changes in Tjp1 and Tjp2 expression suggest that the BNB may be more permeable after SNC and DRC compared to naive and SCI. KEGG pathway analysis of DE genes in endothelial cells and pericytes further suggest that the BNB may be differentially altered after SNC and DRC ( Figure 2G,H). The enrichment of the cell cycle pathway after SNC and DRC suggests that endothelial cell division may regulate blood vessel angiogenesis (Zeng et al., 2007).
After nerve injury, dedifferentiation of Schwann cells into repair Schwann cells at the site of injury as well as resident macrophages in the nerve elicits breakdown of the BNB (Mellick and Cavanagh, 1968;Napoli et al., 2012;Richner et al., 2018). Although Schwann cells in the DRG are far away from the injury site in axons, we found that they undergo transcriptional changes that Notably, Ngf, which is known to promote myelination by Schwann cells in peripheral nerves (Chan et al., 2004) is downregulated after all injuries. VEGF is known to increase BNB permeability (Lim et al., 2014) and Vegfb is differentially regulated after peripheral and central injuries ( Figure 2J), suggesting that Schwann cells may influence BNB permeability in the DRG. Shh is strongly upregulated after DRC. Shh signaling in Schwann cells in the DRG after SNC and DRC may have neuroprotective functions (Hashimoto et al., 2008) and may facilitate axon regeneration . KEGG analysis also revealed that the hedgehog signaling pathway and axon guidance is upregulated specifically after DRC ( Figure 2K). The hippo signaling pathway, which plays multiple cellular functions, such as proliferation, apoptosis, regeneration and organ size control (Yu and Guan, 2013;Zhao et al., 2011), is downregulated specifically after SCI and DRC.
Key transcription factors in the Hippo pathway, Yap and Bmp5 were downregulated after DRC and SCI, and upregulated after SNC (FDR≤0.05, FC≥2) (Figure 2-Source Data 1). These results suggest that Schwann cell in the DRG respond differently to peripheral and central injuries, with central injury potentially limiting their plasticity.

Macrophages proliferate in response to peripheral but not central injuries
After nerve injury, breakdown of the BNB allows the influx of inflammatory cells at the site of injury in the nerve to promote repair (Mellick and Cavanagh, 1968;Napoli et al., 2012). In addition to their role at the site of injury in the nerve, macrophages regulate axon regeneration and pain responses acting at the level of the ganglia (Kwon et al., 2013;Niemi et al., 2013;Yu et al., 2020). Both resident and infiltrating macrophages were found in the DRG (Zigmond and Echevarria, 2019). To understand if macrophages in the DRG include the two major macrophages subsets found in the nerve, snMac1 that reside in the endoneurium or snMac2 that reside in the connective tissue surrounding nerve fascicles (Ydens et al., 2020), we analyzed the percent of DRG macrophages expressing marker genes for these two subtypes. This analysis revealed that most DRG macrophages express the snMa1c genes (Cbrr2, Mgl2) ( Figure 3A, light blue), whereas few DRG macrophages express the snMac2 genes (Retnnla, Clecl10a, Folr2) ( Figure 3A, blue). The DRG macrophages, similarly to nerve macrophages (Wang et al., 2020;Ydens et al., 2020), also express CNS associated microglia genes such as Teme119, P2ry12 and Trem2 ( Figure 3A, pink) and common microglia/macrophages markers Ccl12, ,Gpr34, Gpr183, Hexb, Mef2c, St3gal6 and Tagap ( Figure 3A, green). The common macrophages markers Cd68, Emr1 and Aif1 were expressed in >80% of cells in the macrophage cluster ( Figure   3A, orange and Figure 3 -Source Data 1). These results suggest that DRG macrophages share similar properties to snMac1 residing in the nerve endoneurium and with CNS microglia.
We next examined the injury responses of DRG macrophages. The number of macrophages increased after SNC compared to naïve and also increased to a lesser extent after DRC and SCI ( Figure 1G). Macrophages displayed a similar gene expression profile in naive and SCI condition, but SNC and DRC elicited large changes in genes expression ( Figure 3B). KEGG pathway analysis of DEG (FDR≤0.05, FC≥2) (Figure 3-Source Data 1) revealed up regulation of cell cycle and DNA replication after SNC, while DRC and SCI macrophages mainly showed up regulation of metabolic pathways such as steroid biosynthesis and glycolysis/gluconeogenesis pathways ( Figure 3C). Interestingly, macrophages from all injury conditions down regulated genes related to antigen processing and presentation as well as genes involved in phagosome activity ( Figure 3C). We further validated the down regulation of genes involved in antigen processing and presentation associated with class II major histocompatibility complex (MHC II) Cd74, H2-Aa and Ctss in qPCR experiments ( Figure 3D and figure 3-Source Data1). Several studies suggested a predominant anti-inflammatory macrophage phenotype in the DRG following sciatic nerve injury (Komori et al., 2011;Kwon et al., 2013;Lindborg et al., 2018;Niemi et al., 2013). Heat map of cytokines and other macrophages polarization markers suggest that macrophages responses to injury in the DRG are complex and may not easily follow the classical proinflammatory and anti-inflammatory polarization scheme ( Figure 3E and figure 3-Source Data1).
Furthermore, expression of the anti-inflammatory marker gene Arg1 was only detected after DRC and the pro-inflammatory marker Nos2 was not detected in any conditions ( Figure 3E). Among the significant down regulated genes after SNC, we found the cytokines Ccl2, Il1b and Tnf, which we validated for their downregulation also by qPCR experiments (figure 3D and figure 3-Source Data1). Hematopoietic cell lineage pathway, which is involved in the formation of macrophages from myeloid cells, was down regulated specifically after DRC and SCI, but not after SNC ( Figure   3C). We next explored cell division in the macrophage cluster. The cell cycle and DNA replication pathways were upregulated only after SNC ( Figure 3C). Heatmap analysis of the proliferation markers Mki67, Cdk1 and Top2a further demonstrated higher expression of proliferation genes after SNC compared to naive macrophages and following central injuries ( Figure 3E). qPCR analysis of DRG cells after SNC showed increase in Mki67 expression compared to naïve ( Figure   4D). t-SNE plots overlaid with the proliferation marker genes Mki67, Cdk1 and Top2a revealed expression mainly in one macrophage subtypes following SNC (~60%) and smaller clusters of proliferating cells after DRC (~20%), SCI (~14%) and in naive condition (~6%) ( Figure 3F).
Validation of the scRNAseq data by immunostaining of DRG sections with the macrophage specific marker CD68 and the proliferation marker MKI67 further revealed a higher number of . These results suggest that macrophages represent a large proportion of proliferating cells in the DRG after nerve injury. This is consistent with the recent observation that macrophage expansion after nerve injury in the DRG involves proliferation (Yu et al., 2020).
However, whether the proliferating macrophages originate from resident macrophages or from the infiltration of monocytes-derived macrophages remains to be determined. These results highlight that central and peripheral nerve injury differently affect gene expression in macrophages and that a better understanding of these responses may highlight their role in pain and nerve regeneration.

A subset of macrophages expressing glial markers is increased by injury
The macrophage cluster was classified in our scRNAseq analysis by differential expression of macrophage specific markers such as Cd68 and Aif1 ( Figure 4B). This agrees with other studies reporting that SGC can express immune markers (Donegan et al., 2013;Huang et al., 2021;Jasmin et al., 2010;Mapps et al., 2021;van Velzen et al., 2009;van Weperen et al., 2021). Violin plots for Cd68 and Fabp7 13 expression across all cell types in the DRG further demonstrate that Cd68 and Fabp7 are highly expressed in cluster 13, which we named the 'Imoonglia' cluster ( Figure  Macrophages can also be located in close proximity to SGC ( Figure 4F bottom), suggesting that their localization around sensory neurons can resemble SGC (Avraham et al., 2020;Hanani, 2005). Examination of the extent of the Imoonglia population in the DRG revealed that this is a rare population, representing ~1% of all DRG cells ( Figure 4G). Interestingly, the representation of Imoonglia in the DRG increased to ~ 2.5% after both peripheral and central injuries ( Figure   4G). t-SNE plot of Imoonglia cells revealed similar clustering across all conditions, suggesting similar gene expression that is not affected by injury (Figure 4-Figure supplement 1C). We then pooled the Imoonglia cluster from all conditions and calculated the expression difference of each gene between that cluster and the average in the rest of the clusters (>1.5-fold change p-value<0.05). We then compared the genes uniquely expressed in Imoonglia (1,097 genes) to macrophages (2,469 genes) and SGC (2,331 genes) (Figure 4-Source Data 1). This analysis revealed a higher similarity of Imoonglia cluster to macrophages (863 shared genes) than SGC

SGC represent a diverse cell population
We next determined if SGC represent a diverse glial population in the DRG in naïve conditions. We previously described that Fabp7 is a specific marker gene for SGC in DRG and that the FABP7 protein is expressed in all SGC in the DRG ( Figure 5A) (Avraham et al., 2020).
The DRG encompasses different types of sensory neurons such as nociceptors, mechanoreceptors and proprioceptors (Renthal et al., 2020;Usoskin et al., 2015), with each type controlling a different sensory function. To determine if SGC also exist as different subtypes in DRG, we examined the expression of other known SGC markers, Cadh19, Kcnj10 and Glul (GS) in addition to Fabp7, by pooling naïve SGC from our scRNAseq analysis. To determine if a given subtype is associated with a specific neuronal subtype, we examined expression of cluster 3 specific gene Aldh1l1 using the Aldh1l1::Rpl10a-Egfp reporter mouse (Doyle et al., 2008). Although Scn7a appeared to be the most cluster-specific gene marker, it is also expressed in neurons, impairing the precise examination of its cellular localization. Aldh1l1 is typically used to label all astrocytes in the CNS and we found that Aldh1l1 drove expression of Rpl10a-Egfp in a subset of SGC, consistent with the single cell data ( Figure   5G). Rpl10a-Egfp expression was detected in SGC surrounding both TRKA positive nociceptor neurons and TRKA negative neurons ( Figure 5G), suggesting that cluster 3 SGC are not specifically associated to a given neuronal subtype. Our results are also consistent with the recent finding that Aldh1l1::Rpl10a-Egfp mice express Egfp in a subset of SGC (Rabah et al., 2020). We next compared each SGC subcluster with astrocytes (Zhang et al., 2014), myelinating Schwann cells and non-myelinating Schwann cells (Wolbert et al., 2020). Cluster 3, which expresses the astrocyte marker Aldh1l1 shares the most genes with astrocytes such as Glul, Kcnj10 and Slc1a3  Figure 5J). Whether the SGC subtypes represent functionally distinct populations remains to be determined.

A distinct SGC cluster appears in response to peripheral nerve injury
We previously revealed the contribution of SGC to axon regeneration (Avraham et al., 2020). To determine if the different regenerative capacities after peripheral or central injury result in part from different responses in SGC, we examined the SGC responses to SCI and DRC compared to SNC. Separate clusters emerged in SGC after SNC and DRC injuries but were similar in naïve and SCI conditions ( Figure 6A). We next determined if the 4 sub clusters identified in naïve conditions ( Figure 5D Figure 6C). After SNC, cluster 1 (blue) decreased, whereas cluster 6 (light blue) emerged and accounted for 40% of all SGC. In contrast, SGC after SCI showed a decrease in cluster 4 (green), with cluster 7 appearing specifically after SCI (pink). Dot plot analysis further supports sub cluster changes in SGC following different injuries, revealing the percentage of cells in each cluster together with the level of expression of cluster specific genes ( Figure 6D). Analysis of DEG for every subcluster in each injury condition revealed that the majority of gene expression changes occurred following SNC in all subclusters, with the highest changes in subcluster 4 (FDR≤ 0.05, fold-change ≥2) ( Figure 6E, Figure 6-Source Data 1). GFAP is a known marker of injured SGC Woodham et al., 1989;Xie et al., 2009) and GFAP expression was observed in cluster 6 after SNC, but also in SGC after DRC in cluster 2, 3 and 4 ( Figure  6F,G). Immunostaining in DRG sections confirmed that ~25% of neurons in both SNC and DRC conditions were surrounded by GFAP expressing SGC, with no changes after SCI ( Figure 6H,I and Figure 6-Source Data 1). These results suggest that GFAP expression is a marker for SGC injury but does not entirely relate to the different axon regenerative capabilities in peripheral nerve and dorsal root. We next performed GO and KEGG pathway analysis of cluster 6 marker genes, which revealed enrichment for pathways involved in axon regeneration, calcium signaling pathway and mineral absorption ( Figure 6J and Figure 6-Source Data 1). To further characterize the unique cluster 6 marker genes induced by SNC, we performed a transcription factor binding site analysis, which revealed enrichment for Rest, Trp53, Rad21, Ctcf and Zeb1 ( Figure 6K). We next used STRING to determine the functional protein interaction of these transcription factors and found that the transcription repressor CTCF was highly associated with the RAD21 and p53, less with ZEB1 and not at all with REST ( Figure 6L). Zeb1 is known to control epithelial to mesenchymal transition (EMT) leading to a more plastic state , while Rest is involved in the signaling pathways regulating pluripotency (Singh et al., 2008), suggesting that cluster 6 adopts a more plastic state after SNC that might play a role in nerve regeneration.

Activation of PPARα with fenofibrate increases axon regeneration after dorsal root crush
We recently revealed that PPARα signaling downstream of fatty acid synthase in SGC promotes axon regeneration after peripheral neve injury (Avraham et al., 2020) We previously showed that the specific PPARα agonist fenofibrate, an FDA approved drug used to treat dyslipidemia with minimal activity towards PPARγ (Kim et al., 2016;Rosenson, 2008), upregulated PPARα target genes in the DRG and rescued the impaired axon growth in mice lacking fatty acid synthase in SGC (Avraham et al., 2020). Since SGC do not activate PPARα after DRC ( Figure 7A), a model in which axonal growth occurs at about half the rate of peripheral axons (Oblinger and Lasek, 1984;Wujek and Lasek, 1983), we tested if fenofibrate treatment improved axon regeneration after DRC. Mice were fed with fenofibrate or control diet for 2 weeks as described (Avraham et al., 2020) and then underwent DRC injury. We measured the extent of axon regeneration past the injury site three days later by labeling dorsal root sections with SCG10, a marker for regenerating axons (Shin et al., 2014) ( Figure 7E). The crush site was determined according to highest SCG10 intensity along the nerve. First, we measured the length of the 10 longest axons, which reflect the extent of axon elongation, regardless of the number of axons that regenerate ( Figure 7F). Second, we measured a regeneration index by normalizing the average SCG10 intensity at distances away from the crush site to the SCG10 intensity at the crush site ( Figure 7G). This measure takes into accounts both the length and the number of regenerating axons past the crush site. Both measurement methods revealed improved regeneration in mice treated with fenofibrate compared to control diet ( Figure 7F,G and Figure 7-Source Data 1).
These results indicate that the lack of PPARα signaling in SGC after central injury contributes to the decreased regenerative ability. Whether PPARα activation could also promote axon entry passed the dorsal root entry zone and improve axon regeneration of dorsal column axons after SCI remains to be determined. This study provides a map of the distinct DRG microenvironment responses to peripheral and central injuries at the single cell level and highlights that manipulating SGC in the DRG could lead to avenues to promote functional recovery after CNS injuries.

DISCUSSION
Our unbiased single cell approach fills a critical gap in knowledge for the field and enables in depth characterization of the molecular profile of cells comprising the neuronal microenvironment in the DRG following acute peripheral and central injuries. Our analysis demonstrates major, yet distinct molecular changes in non-neuronal cells in response to peripheral nerve injury and dorsal root injury, with more limited responses after SCI. It is possible that the percent of DRG neurons lesioned under any injury paradigms may impact the injury response of the microenvironment.
Nonetheless, all three injury paradigms are widely used models to study the mechanisms promoting axon regeneration. Our study highlights that manipulating non-neuronal cells could lead to avenues to promote functional recovery after CNS injuries or disease.
In endothelial cells, pericytes, Schwann cells, macrophages and SGC, gene expression changes were the largest after peripheral injury, but also occurred after dorsal root crush and spinal cord injury. How non-neuronal cells in the DRG sense a distant axon injury remains poorly understood. The mechanisms underlying SGC responses to nerve injury were proposed to depend on early spontaneous activity in injured neurons as well as retrograde signaling and direct bidirectional communication Xie et al., 2009). DLK was shown to be important for retrograde injury signaling in sensory neurons (Shin et al., 2019), but whether DLK also regulate SGC response to injury remains to be determined. Macrophages have been proposed to respond to chemokines expressed by injured neurons such as CCL2 for their recruitment to the DRG after nerve injury (Niemi et al., 2016;Niemi et al., 2013) and DLK was shown to regulate the expression of cytokines such as CCL2 in neurons (Hu et al., 2019). Injured sensory neurons also express colony-stimulating factor 1 (CSF1) in a DLK dependent manner (Guan et al., 2016;Wlaschin et al., 2018), which could contribute to recruit CSF1R expressing macrophages to the DRG. Whether DLK activity in neurons regulate communication to immune cells directly or via SGC will require further investigations. Our observation that SGC express CCL5 after nerve injury suggests that SGC may contribute to recruit macrophages to the DRG.
The breakdown of the BNB in response to nerve damage can lead to neuronal dysfunction and contribute to the development of neuropathy (Richner et al., 2018). While the impact of physical nerve damage or disease state such as diabetic neuropathy on the BNB are being studied (Richner et al., 2018), whether and how nerve injury affect the BNB in the sensory ganglion is not known. BNB leakage can give blood derived molecules direct access to sensory neurons and promote infiltration of inflammatory cells to engage inflammatory responses. Our results suggest that endothelial cells and pericytes respond differently to peripheral and central injuries, with alteration in expression of tight junction related genes, potentially underlying changes in BNB permeability. Similar to the situation in the nerve, Schwann cells and potentially macrophages may secrete factors such as VEGF and cytokines, altering endothelial cell function.
Sympathetic innervation of blood vessels in the DRG following nerve injury, which depends in part on IL-6 signaling, may also underlie the changes observed in endothelial cells (Ramer and Bisby, 1998;Ramer et al., 1998b).
In the injured nerve, Schwann cells guide regenerating axons to support distal innervation Jessen and Mirsky, 2016). To accomplish these regenerative functions, Schwann cells are reprogrammed to a repair state, which relies in part on the transition from an epithelial fate to a more plastic mesenchymal fate (Arthur-Farraj et al., 2017;Clements et al., 2017). Our studies highlight that away from the injury site in the DRG, Schwann cells respond to injury in part by regulating the hippo pathway, a central pathway in cellular growth and plasticity (Yu and Guan, 2013;Zhao et al., 2011). The transcription factor Yap, which is regulated by the hippo pathway, is upregulated after nerve injury but downregulated after both central injuries. The Nf2-Yap signaling was shown to play important roles in controlling the expansion of DRG progenitors and glia during DRG development (Serinagaoglu et al., 2015). These results suggest that Schwann cells in the DRG respond differently to distant peripheral and central injuries, and that central injury may limit their plasticity. Prior studies demonstrated that glial overexpression of NGF enhances neuropathic pain and adrenergic sprouting into DRG following chronic sciatic constriction in mice (Ramer et al., 1998a). Since Ngf is downregulated in Schwann cell after all injuries, Schwann cells in the DRG may regulate pain and adrenergic sprouting after injury.
Macrophages are known to regulate regenerative responses. In the nerve, macrophages function primarily to assist Schwann cells for debris clearance. Nerve resident macrophages in naive conditions share features with CNS microglia (Wang et al., 2020;Ydens et al., 2020). In the injured nerve, resident macrophages represent only a small subset that secretes chemoattractants to recruit circulating monocytes-derived macrophages (Ydens et al., 2020).
These recruited monocytes-derived macrophages express Arg1 and represent the main macrophage population guiding nerve repair (Ydens et al., 2020). In the naïve DRG, we found that resident macrophages share similar properties with endoneurial macrophages in the nerve.
However, Arg1 expression is only observed after dorsal root crush and not after peripheral nerve injury or SCI. The increase in cell cycle marker after nerve injury suggest that the increase in macrophage numbers in the DRG results largely from macrophage proliferation. The increase in IL-6, which has been associated with nerve regeneration (Cafferty et al., 2004;Cao et al., 2006), is associated only with nerve injury and not central injuries. Immune cells in the DRG may thus not follow a strict classification and complex subtypes exist in naive conditions that are differentially regulated by peripheral and central injury.
Our data also unravels the existence of a small proportion of cells that share expression of macrophage and glial genes, which we named Imoonglia. These cells increase in number after all injuries. Previous studies suggested that after injury some macrophages penetrate into the space between SGC and neurons (Hu and McLachlan, 2002). Several other studies interpreted  (Mapps et al., 2021;van Weperen et al., 2021). Whether immune cells move to a peri-neuronal position and become Imoonglia or whether SGCs transition from a progenitor state to acquire an immune signature in stressful condition such as after nerve injury remains to be tested. Our single cell profiling, flow cytometry and immunofluorescence data suggest that Imoonglia represent a specific cell type that is increased after injury and share the spatial arrangement of SGC surrounding sensory neurons.
This may be similar to astrocytes in the CNS, which in addition to their homeostatic functions can undergo an inflammatory transcriptional transition following inflammation after acute insults like stroke (Zamanian et al., 2012), spinal cord injury (Karimi-Abdolrezaee and Billakanti, 2012) and systemic inflammation (Hasel et al., 2021). It is tempting to speculate that Imoonglia may function in immune surveillance in the DRG and specifically in the case of viral infection. Indeed, SGC were suggested to restrict the local diffusion of viruses such as herpes simplex virus, varicella zoster virus, HIV-1 and haemagglutinating encephalomyelitis virus, which belongs to the coronavirus family (Hanani and Spray, 2020).
SGC have been previously characterized as a neural crest derived uniform glial population that plays a major role in pain (Hanani and Spray, 2020). Our single cell analysis reveals that SGC do not represent a uniform population but that several subtypes exist. In addition to the Imoonglia discussed above, we identified four SGC clusters in naïve conditions and up to seven clusters after injury. The enrichment of different biological pathways in each cluster suggest that they each have a specialized function. Cluster 3 shares the most similarities with astrocytes and is enriched for PPARα signaling and fatty acid biosynthesis. Cluster 1 and 2 represent the most unique SGC subtypes. Interestingly, cluster 1 is enriched for calcium signaling pathway, which in SGC is known to play a role in pain (Hanani and Spray, 2020). It will be important to determine in future studies whether the most unique SGC subtypes (cluster 1 and 2) surround specific neuronal subtypes, or if a mosaic organization exists, with different SGC surrounding the same neuron. The appearance of a specific cluster (cluster 6) after nerve injury suggest that this cluster plays an essential role in nerve injury responses. Cluster 6 is also enriched for the transcription factor REST, which in astrocyte regulates gliosecretion (Prada et al., 2011). Transcription factor binding site analysis of genes in cluster 6 revealed enrichment for the EMT gene Zeb1. EMT is often linked to increased plasticity and stem cell activation during tissue regeneration, suggesting that cluster 6 is related to plasticity of SGC.
In neurons, axonal PPARγ contributes to the pro-regenerative response after axon injury (Lezana et al., 2016) and our recent study suggest that FASN may generate ligands for PPARγ in neurons (Ewan et al., 2021). We recently demonstrated that in SGC, FASN is required for the activation of PPARα and that PPARα signaling promote axon regeneration in adult peripheral nerves (Avraham et al., 2020). Here we showed that this PPARα signaling pathway is not activated in SGC after injury to centrally projecting axons (SCI and DRC). Further, we demonstrate that the FDA approved PPARα agonist fenofibrate increased axon regeneration in the dorsal root, a model of poor sensory axon regeneration. In our previous study, we showed that removing the enzyme FASN, which is upstream of PPARα activation, specifically in SGC, decreases axon growth in the sciatic nerve (Avraham et al., 2020). Altogether, these findings suggest that the lack of PPARα activation after dorsal root crush contributes to the low regeneration rates of axons in the dorsal root and provide insights into the translational potential of fenofibrate. Indeed, fenofibrate is used clinically to treat lipid disorders, and has unexpectedly been shown in clinical trials to have neuroprotective effects in diabetic retinopathy (Bogdanov et al., 2015;Moreno and Ceru, 2015) and in traumatic brain injury (Chen et al., 2007). In mice, fenofibrate was shown to modestly increase tissue sparing following spinal contusion injury (Almad et al., 2011). The neuroprotective role of fenofibrate was also recently observed in a paclitaxel chemotherapy-induced peripheral neuropathy (Caillaud et al., 2021). The enrichment of biological pathways related to PPARα signaling are largely conserved between rodent and human SGC . Together, these findings support the notion that PPARα activation is a promising therapeutics for neurologic disease and CNS injury (Mandrekar-Colucci et al., 2013). The transcriptional profiling of SGC in response to peripheral and central injury highlights that manipulating non-neuronal cells could lead to avenues to treat CNS injuries.  Accreditation -7/18/97, USDA Accreditation: Registration # 43-R-008.

Animals and Procedures
During surgery, 8-12-week-old female C57Bl/6 mice were anesthetized using 2% inhaled isoflurane. Sciatic nerve injuries were performed as previously described (Avraham et al., 2020;Cho et al., 2015;Cho et al., 2013). Briefly, the sciatic nerve was exposed with a small skin incision (~1cm) and crushed for 5 seconds using #55 forceps. The wound was closed using wound clips and injured L4 and L5 dorsal root ganglia were dissected at the indicated time post-surgery.
Contralateral DRG served as uninjured control. For spinal cord injury (SCI), a small midline skin incision (~1cm) was made over the thoracic vertebrae at T9−T10, the paraspinal muscles freed, and the vertebral column stabilized with metal clamps placed under the T9/10 transverse processes. Dorsal laminectomy at T9/10 was performed with laminectomy forceps, the dura removed with fine forceps, and the dorsal column transversely cut using fine iridectomy scissors.
Dorsal root injury was performed similarly as SCI, except that procedures were performed at the L2-L3 vertebral level, and fine forceps used to crush the right L4 and L5 dorsal roots for 5 seconds.
L4 and L5 roots are in close proximity anatomically hence both roots were crushed simultaneously where the distance from the crush site to L4 DRG is 4-5mm and 7-8mm to L5 DRG. During dorsal root crush, the roots are forcefully squeezed causing the disruption of nerve fibers without interrupting the endoneurial tube. 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. An unbiased clustering (graph-based clustering) was done and presented as t-SNE (tdistributed 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 an ANOVA model; a gene is considered differentially-expressed (DE) if it has an FDR≤ 0.05 and a fold-change≥2. The data was subsequently analyzed for enrichment of GO terms and the KEGG pathways using Partek flow pathway analysis.
A differential trajectory map of single cells was performed using 'Monocle2' with standard setting. The algorithm orders a set of individual cells along a path / trajectory / lineage and assign a pseudotime value to each cell that represents where the cell is along that path. This method facilitates the discovery of genes that identify certain subtypes of cells, or that mark intermediate 32 states during a biological process as well as bifurcate between two alternative cellular fates.
Partek was also used to generate figures for t-SNE, scatter plot and trajectory analysis representing gene expression.
Cell-cell interaction analysis was performed based on CellPhoneDB repository (v2.1.6) ,which was developed for human, the mouse genes were converted to human genes first.
Statistical iterations were set at 1000 and gene expressed by less than 10% of cells in the cluster were removed. Network visualization was performed with Cytoscape (v3.8.2) using the identified significant interactions between the clusters.

Flow cytometry
Ganglia were enzymatically and mechanically dissociated to a single cell suspension as All primer sequences were obtained from PrimerBank (Harvard) and product size validated using agarose gel electrophoresis.

Tissue Preparation and Immunohistochemistry
After isolation of either dorsal root or DRG, tissue was fixed using 4% paraformaldehyde for 1 hour at room temperature. Tissue was then washed in PBS and cryoprotected using 30% sucrose solution at 4C overnight. Next, the tissue was embedded in O.C.T., frozen, and mounted for cryosectioning. All frozen sections were cut to a width of 10µm for subsequent staining. Catalog#DL-1178-1) was used per mouse by injection into the tail vein. Mice were sacrificed after 20mins of injection. Samples were collected following the procedure described above. 20um thickness of DRG cryosections were used for immunofluorescence staining and image was captured under LSM880 confocal microscope.

Data collection and analyses.
Data collection and analyses were performed blind to the conditions of the experiments.
Single cell RNAseq analysis was performed in an unbiased manner using established algorithms.

Quantification and statistical analysis
Quantifications were performed by a blinded experimenter to injury type and treatment.
Fiji (ImageJ) analysis software was used for immunohistochemistry images quantifications. Nikon Elements analysis software was used to trace regenerating axons in the dorsal root sections.
Statistics was performed using GraphPad (Prism8) for t-test and one/two-way ANOVA followed by Bonferroni's multiple comparisons test. Error bars indicate the standard error of the mean (SEM). Heatmaps were calculated as fold change of normalized counts compared to naïve or as z-scores. The formula for calculating z-score used z = (x-μ)/σ, where x is the expression of the gene in all cells for each condition, μ is the sample mean of the gene in all conditions, and σ is the sample standard deviation in all conditions. Sample was calculated as the average expression of cells in each condition.