Temporal Analysis of Gene Expression in the Murine Schwann Cell Lineage and the Acutely Injured Postnatal Nerve

Schwann cells (SCs) arise from neural crest cells (NCCs) that first give rise to SC precursors (SCPs), followed by immature SCs, pro-myelinating SCs, and finally, non-myelinating or myelinating SCs. After nerve injury, mature SCs ‘de-differentiate’, downregulating their myelination program while transiently re-activating early glial lineage genes. To better understand molecular parallels between developing and de-differentiated SCs, we characterized the expression profiles of a panel of 12 transcription factors from the onset of NCC migration through postnatal stages, as well as after acute nerve injury. Using Sox10 as a pan-glial marker in co-expression studies, the earliest transcription factors expressed in E9.0 Sox10+ NCCs were Sox9, Pax3, AP2α and Nfatc4. E10.5 Sox10+ NCCs coalescing in the dorsal root ganglia differed slightly, expressing Sox9, Pax3, AP2α and Etv5. E12.5 SCPs continued to express Sox10, Sox9, AP2α and Pax3, as well as initiating Sox2 and Egr1 expression. E14.5 immature SCs were similar to SCPs, except that they lost Pax3 expression. By E18.5, AP2α, Sox2 and Egr1 expression was turned off in the nerve, while Jun, Oct6 and Yy1 expression was initiated in pro-myelinating Sox9+/Sox10+ SCs. Early postnatal and adult SCs continued to express Sox9, Jun, Oct6 and Yy1 and initiated Nfatc4 and Egr2 expression. Notably, at all stages, expression of each marker was observed only in a subset of Sox10+ SCs, highlighting the heterogeneity of the SC pool. Following acute nerve injury, Egr1, Jun, Oct6, and Sox2 expression was upregulated, Egr2 expression was downregulated, while Sox9, Yy1, and Nfatc4 expression was maintained at similar frequencies. Notably, de-differentiated SCs in the injured nerve did not display a transcription factor profile corresponding to a specific stage in the SC lineage. Taken together, we demonstrate that uninjured and injured SCs are heterogeneous and distinct from one another, and de-differentiation recapitulates transcriptional aspects of several different embryonic stages.

Introduction processing techniques and reagents. Additionally, many of these comparisons are made at the level of mRNA, or in-vitro, making it difficult to accurately appreciate the in vivo de-differentiated phenotype, and how closely it recapitulates developmental SC programs. Here, we have described the transcriptional profile of developing and de-differentiated SC in vivo in a comparable and relevant manner. We conducted an extensive spatio-temporal analysis through five key stages of embryonic mouse development (E9.0, E10.5, E12.5, E14.5, E18.5), postnatal stages P7 and P65, as well as within the P65 nerve following acute injury, when SCs are actively acquiring a reparative state.
Through these studies, we identified distinct expression profiles for NCCs, NCC precursors, SCPs, iSCs, pro-myelinating SCs and mature myelinating/non-myelinating SCs, and demonstrated an underlying heterogeneity of the SC pool. Our findings also demonstrated that the de-differentiated SC is a unique SC subtype, distinct from any one developmental stage in the SC lineage. Given that this 'repair' SC subtype is the driving force behind efficient regeneration in uncompromised nerve injuries, methods to recapitulate this phenotype could be further investigated as a therapeutic avenue to treat chronic nerve injury and demyelinating disease.

Materials and Methods Animals
CD1, C57/BL6, and Sox2eGFP mice [46] mice were purchased from Charles River Laboratories (Senneville, QC) and Jackson Laboratory (ME, United States) and maintained in a 12 hr light cycle. Embryos were staged using the morning of the vaginal plug as embryonic day (E) 0.5. Pregnant females were housed individually after mating, and euthanized using cervical dislocation for embryo collection. Adult mice used for peripheral nerve harvesting were group housed before injury, and then singly housed with enrichment post-injury. These animals were euthanized using an overdose (0.1mL) of Sodium Pentobarbital (54.7mg/mL, Ceva Sante Animale). Animal procedures were approved by the University of Calgary Animal Care Committee in compliance with the Guidelines of the Canadian Council of Animal Care.

Embryo processing
Whole embryos were collected for stages E9.0 and E10.5, while only bodies were collected for stages E12.5, E14.5 and E18.5. For post-natal studies, sciatic nerves were harvested from the limbs of P7/P65 pups. The embryos and nerves were fixed in 4% paraformaldehyde (PFA)/1X diethyl-pyrocarbonate (DEPC) treated phosphate-buffered saline (PBS) for~4-20 hours at 4°C. The embryos and nerves were rinsed in DEPC-PBS and transferred to 20% sucrose/1X DEPC-PBS, and were kept overnight at 4°C. The embryos and nerves were then embedded using O.C.T™ (Tissue-Tek1, Sakura Finetek U.S.A. Inc., Torrance, CA) and stored at -80°C. For the injury study, five days after injury (P65) mice were sacrificed by overdose of sodium pentrobarbital (i.p.; CEVA, Sante Animale). Nerves were removed and fixed in 4% PFA for two hours and subsequentally placed in 30% sucrose overnight. The next day, nerves were embedded in O.C.T.™ compound, frozen on dry ice and stored at -80°C before cutting cryosections on a Leica cryostat (Richmond Hill, ON).

Surgery
For crush injury, P60 mice were anesthetized using isofluorane (5% induction and 2% maintenance) and then given a preoperative subcutaneous injection of 0.1 mL (0.03 mg/mL) buprenorphine. Hindlimbs were shaved and then cleaned twice with 70% EtOH followed by 10% providine iodine. On only the right hindlimb, the sciatic nerve was crushed at mid-thigh using #10 forcep for one minute. Muscle and skin were sutured back together (7-0 Prolene and 7-0 Silk, black braided; Ethicon Inc.) and buprenorphine was administered once a day for 4 days following surgery.

RNA in situ hybridization
RNA in situ hybridization was performed as previously described [47].

Microscopy and image processing
For the embryonic and P7 sections, images were captured with a QImaging RETIGA 2000R or QImaging RETIGA EX digital camera and a Leica DMRXA2 optical microscope using OpenLab5 software (Improvision; Waltham MA). P65 nerve sections (injured and un-injured) were imaged using an inverted epifluorescent microscope (40x objective with oil, z-stack, Axio Observer Research Microscope; Zeiss Observer.Z1). Negative controls were included to distinguish non-specific secondary antibody binding. The captured images were processed using Adobe Photoshop software. Quantification in the embryonic studies was restricted to the Sox10 + NCCs at E9.0, and Sox10 + cells populating the dorsal and ventral root and DRG, and the exiting spinal nerve at the remaining stages. Quantification in the injury studies was conducted using a minimum of three images taken distal to the crush site from three different tissue sections for each animal (n = 3 mice per group). Double and single-positive cells (protein of interest co-localizing with Sox10 + SCs and Sox10 + SCs) were manually counted using Adobe Photoshop software. An unpaired student's t-test was performed using Prism software (Graph-Pad) to determine whether there was a difference in protein expression of intact or actuely injured adult nerves (significance p<0.05).

Expression of Schwann cell lineage markers in neural crest cells
We first assessed the expression of SC markers in delaminating trunk NCCs at E9.0 ( Fig 1A). To reliably label NCCs, and at later stages, to mark peripheral cells fated for the glial lineage, we used Sox10 as a co-label in all marker studies. Sox10 is continually expressed in NCCs, SCs and satellite glia throughout development [26,28], and is required for the differentiation of all peripheral glia [28,49,50]. At E9.0, Sox10 was expressed in trunk NCCs delaminating from the neural tube, including those following both dorsolateral and ventral migratory routes ( These observations were consistent with previous reports documenting the activation of a Nfat transcriptional reporter [45] and the expression of Sox9 [26], AP2α [39], and Pax3 [51] in migrating NCCs. Of these markers, Sox9 induces a NCC phenotype [25], and its expression biases migrating NCCs towards glial and melanocyte lineage selection [26], whereas essential roles have only been documented at later developmental stages for the remaining transcription factors in the SC lineage: AP2α maintains a SCP fate, impeding the transition to an iSC [39], Pax3 regulates SC proliferation [52] and Nfatc4 acts synergistically with Sox10 to initiate Egr2 expression in SCs [45].
At E9.0, Etv5 was also expressed, but only in a very small number of Sox10 + NCCs (S1A-S1C Fig), which may be why previous reports have suggested that Etv5 is not expressed in the E9.0 neural crest [42]. Jun expression was also detected in a subset of NCCs, but instead of labeling cells in the ventral migratory pathway, it was primarily expressed in NCCs following a dorsolateral migratory route, which are pre-destined to a melanocyte fate (S1D-S1F Fig). In contrast, we did not detect the expression of Oct6 (S1G-S1I Thus, Sox9, AP2α, Pax3, and Nfatc4 are widely co-expressed with Sox10 in E9.0 NCCs following a ventral migratory route, albeit not in all SCs for AP2α and Nfatc4, whereas Etv5 is only expressed in a small subset of NCCs, and Jun instead marks dorsolaterally migrating NCCs.

Expression of Schwann cell lineage markers in migratory NCC precursors
By E10.5, NCCs have coalesced to form the DRG, which at this stage, are comprised of sensory neurons and migratory NCC precursors that are destined to become SCPs and satellite glial cells (Fig 1B). Emanating from the DRG are the dorsal and ventral roots, which coalesce to form the mixed sensory/motor spinal nerve. NCCs also give rise to a subset of multipotent cells called boundary cap cells that are located at the dorsal root entry zone and motor exit points. SCs populating the dorsal and ventral roots find their origin in these boundary cap cells, as do a few satellite glial cells [15].  At E10.5, Sox10 was expressed in boundary cap cells in the dorsal and ventral roots of the DRG, in presumptive SCPs, in satellite glia in the periphery of the DRG, and in migratory NCCs in the spinal nerve (Fig 3A-3C). Sox10 expression was for the most part excluded from the central DRG, where NeuN + neuronal cells are located (S2D and S2E Fig). Similar to their co-expression profiles in E9.0 NCCs, Sox10 was largely co-expressed with Sox9 (100±0% Sox9 -+ Sox10 + /Sox10 + cells; Fig 3A-3C' and 3P) and AP2α (100±0% AP2α + Sox10 + /Sox10 + cells; Fig   3D-3F' and 3P). Importantly, not all Sox9 + cells were Sox10 + , as Sox9 was expressed in a larger subset of non-glial cells (as seen in other stages as well), confirming that the antibodies are not recognizing epitopes shared between Sox family members. We further confirmed the specificity of the antibodies by demonstrating that Sox9 and Sox10 have distinct staining patterns in the CNS (S3A- S3F Fig). Pax3 also continued to be co-expressed with Sox10, however at reduced levels (50.1±7.4% Pax3 + Sox10 + /Sox10 + cells; Fig 3G-3I' and 3P), and a much smaller number of Sox10 + NCCs coalescing in the E10.5 DRG and in the ventral root expressed Nfatc4 (3.7 ±0.8% Nfatc4 + Sox10 + /Sox10 + cells; Fig 3J-3L' and 3P). In addition, Etv5 expression was initiated at E10.5 in Sox10 + NCC precursors in the DRG (83.9±6.3% Etv5 + Sox10 + /Sox10 + cells; Fig  3M-3O' and 3P), consistent with previous reports [42,53].
In vitro studies suggested that a block of Etv5 function in NCCs affects neuronal and not glial fate specification [54]. Consistent with these findings, Etv5 (S2J-S2L Fig) as well as AP2α (S2G- S2I Fig) were also expressed in the neuronal-rich central part of the DRG, where they were co-labeled with NeuN, a pan-neuronal marker. In contrast, Sox9 (Fig 3A-3C) was exclusively co-expressed with Sox10 in the DRG periphery, where presumptive peripheral glia are located. Sox9 was also expressed with Sox10 in the ventral root, developing spinal nerve, and in mesenchymal tissue between the DRG and somites (Fig 3A-3C). In contrast, Oct6 (S4A-S4E In summary, at E10.5, Sox9, AP2α, Pax3, Nfatc4 and Etv5 are co-expressed with Sox10 in presumptive NCC-derived glial cells in the dorsal and ventral roots, DRG and developing spinal nerve, and a subset of these cells have progressed to a SCP fate based on the co-expression of glial markers (data not shown).

Expression of Schwann cell lineage markers in Schwann cell precursors
By E12.5, the vast majority of NCC precursors destined for a glial lineage have differentiated into either SCPs or satellite glia. Morphologically, SCPs are distinguished from migrating NCCs as they associate directly with growing axon bundles, but they lack the basal lamina secreted by iSCs ( Fig 1C). SCPs are located proximal to the growing nerve tip, and they participate in compacting the nerves while also guiding axons to their targets [10]. Satellite glia can be partially distinguished from SCPs based on their location; satellite glia are in the DRG but are excluded from the nerves, whereas SCPs are found in both locations. However, because of the lack of specific markers, satellite glial cells are not easily distinguished from SCPs within the DRG, although they do have a more flattened nuclear morphology [55,56]. For simplicity, we use the SCP nomenclature for precursor cells for both satellite glia and SCs.
At E12.5, Sox10 expression was slightly more widespread in the DRG compared to E10.5, marking both peripheral and central DRG cells (Fig 4A and 4B). The extension of Sox10 expression into the central DRG did not include sensory neurons, as Sox10 was not coexpressed with NeuN at this stage (S5D- S5F Fig). Instead, Sox10 was expressed exclusively in presumptive peripheral glia, as previously suggested [26,28]. Sox10 was also expressed in the dorsal (Fig 4A and 4C) and ventral (Fig 4A and 4D) roots, where migrating SCPs are located.  In co-expression studies, Sox10 continued to be highly co-expressed with Sox9 (99.8±0.2% Sox9 + Sox10 + /Sox10 + cells; Fig 4A- ). Most Sox10 + Etv5 + cells lined the periphery of the DRG and were likely satellite glial cells based on their flattened nuclei. Etv5 expression was also detected in a few SCPs in the dorsal and ventral root. In addition, Etv5 was co-expressed with NeuN + in DRG sensory neurons (S5M- S5O Fig). This data is consistent with previous reports indicating that Etv5 transcripts are detected in satellite glial cells and DRG sensory neurons [42,53], although this previous study did not detect Etv5 transcripts in the sciatic nerve [54]. In contrast to Etv5, Sox9 (S5A- In summary, E12.5 SCPs undergo a temporal shift in their expression profile, retaining the expression of Sox10, Sox9, AP2α and Pax3, as observed in E10.5 NCCs, while losing the expression of Nfatc4 and Etv5, and gaining the expression of Sox2 and Egr1. The initiation of Sox2 ( Fig 4Q-4T') and Egr1 (Fig 4U-4X') expression in E12.5 peripheral glia is consistent with previous reports documenting the expression of Sox2 in SCPs and iSCs [23] and Egr1 in SCPs [35]. However, Egr1 is also co-expressed with NeuN in the DRG (S5S-S5U Fig), indicating that it also labels sensory neurons. Of the genes that are newly expressed at this stage, Sox2 expression is upregulated in a subset of cells that develop into the PNS [23], and within the SC lineage, Sox2 regulates the differentiation of SCPs into myelinating SCs versus melanocytes [24], while Egr1 is considered a non-myelinating SC marker [35].

Expression of Schwann cell lineage markers in immature Schwann cells
As development proceeds, SCPs can either give rise to iSCs, or alternatively, endoneurial fibroblasts and melanocytes [18,19]. iSCs appear from E14.5 and persist until just before birth [10] ( Fig 1D). iSCs cluster around several axons and deposit a basal lamina that surrounds both the iSCs and the axonal bundle [20]. iSCs then penetrate axonal bundles, positioning larger diameter axons in the periphery for radial sorting. A characteristic feature of iSCs is that they secrete autocrine survival factors so that they are no longer entirely dependent on axon-derived Neure-gulin1, present on the surface of axons [10,57].
By E14.5, Sox10 was expressed in scattered cells throughout the DRG, including in the center and periphery, where iSCs and satellite glia are located (Fig 5A and 5B). Sox10 was not coexpressed with NeuN, confirming that it is exclusively labeling glial precursors at E14.5 (data not shown). In addition, Sox10 was expressed in the dorsal (Fig 5A and 5C) and ventral roots (Fig 5A and 5D) and in the exiting spinal nerve (data not shown). In co-expression studies at E14.5, Sox10 was still highly co-expressed with Sox9 (97.1±1.8% Sox9 + Sox10 + /Sox10 + cells;  Thus, the major difference between E12.5 SCPs and E14.5 iSCs is the loss of Pax3 expression. Previous studies had detected Pax3 transcripts in SCPs as well as in iSCs, but indicated that Pax3 transcript levels decline in late iSCs undergoing radial sorting, and protein levels were not assessed [43]. In summary, E14.5 iSCs are characterized by the expression of Sox10, Sox9, AP2α, Sox2 and Egr1, as well as glial lineage markers (data not shown), and they differ from E12.5 SCPs in that they no longer express Pax3.

Expression of Schwann cell lineage markers in pro-myelinating Schwann cells
As iSCs develop, they extend cytoplasmic processes that penetrate axonal bundles, helping to distinguish large and small diameter axons. Larger axons are rearranged to the periphery of the bundle, with iSCs associating in a 1:1 proportional manner with these large diameter axons, resulting in radial sorting [20]. iSCs that associate with large diameter axons are a transient population termed pro-myelinating SCs; these are the SCs that will progress towards the myelinating stage (Fig 1E). In addition, a subset of late iSCs persists at E18.5 in association with multiple smaller axons, which are destined to become non-myelinating SCs. Thus, pro-myelinating SCs represent a transient phase in the SC lineage, first appearing just prior to birth, and expanding greatly on the first postnatal day [16].
At E18.5, Sox10 was expressed throughout the DRG, including in the center and periphery (Fig 6A and 6B). In the DRG center, Sox10 + late iSCs and pro-myelinating SCs amalgamated around the growing spinal nerve (Fig 6G and 6H). In dual labeling studies, the only transcription factors expressed at E14.5 that continued to be co-expressed with Sox10 at E18.5 were Sox9 (98.6±1.4% Sox9 + Sox10 + /Sox10 + cells; Fig  Thus, three new transcription factors are expressed in E18.5 late iSCs and pro-myelinating SCs: Jun, Oct6 and Yy1. Jun has previously been shown to be expressed in late immature SCs and downregulated with the onset of myelination [41]. Oct6 + SCs were restricted for the most part to the boundary cap or the ventral root and exiting spinal nerve (Fig 6G-6H), most likely representing pro-myelinating SCs. Indeed, Oct6 is well studied for its role as a cell autonomous regulator of SC development [37] and a pro-myelinating SC marker [38], and is also essential for bringing about the pro-myelinating to myelinating SC transition [38]. Finally, Yy1 is important for attaining the myelination phenotype, such that conditional knockdown of Yy1 in SCs results in hypomyelinated nerves with poor expression of the myelin genes MPZ and Pmp22 [36].
At In summary, E18.5 late iSCs and pro-myelinating SCs in the DRG and nerve roots are characterized by the expression of Sox10, Sox9, Jun, Oct6, Yy1 and AP2α (in the DRG only), and the loss of expression of Sox2 and Egr1.
While Egr2 promotes the terminal differentiation of SCs to a myelinating phenotype, Egr1 and Pax3 are considered non-myelinating SC markers [35,44]. Sox9 is also expressed later in neonatal myelinating and non-myelinating SCs [58], and can cooperatively bind the P0 promoter, a mature SC marker [30], suggesting that Sox9 may also function in postnatal SCs. Nfatc4 acts synergistically with Sox10 to activate the expression of Egr2 in the embryonic nerve [45,59], which suggests a later role for Nfatc4 as Egr2 is required for the terminal differentiation of SCs to a myelinating phenotype [35]. Similarly, Oct6 [38] and Yy1 [36] are required for the myelination of peripheral nerves. In contrast, Jun expression is downregulated by Egr2 upon the onset of myelination, and is thus considered a marker of non-myelinating SCs [41].
Thus, the P7 nerve is primarily populated by Sox10 + myelinating SCs that co-express Sox9, Nfatc4, Jun, Oct6, Yy1 and Egr2, differing from E18.5 pro-myelinating SCs by the initiation of Nfatc4 and Egr2 protein expression. Non-myelinating SCs may also be present based on the expression of Jun, but they fail to express Egr1 (at least in the nucleus) and Pax3 at this stage.
In summary, early postnatal SCs are characterized by the continued expression of Sox10, Sox9, Jun, Oct6 and Yy1, the acquisition of Nfatc4 and Egr2 protein expression, and the loss of AP2α expression. In contrast, late postnatal (i.e., adult) SCs initiate Sox2 and Egr1 expression at low levels, and begin to downregulate Jun and Oct6 expression.

Nerve injury triggers adult Schwann cells to recapitulate a unique pattern of embryonic glial lineage transcription factors
Peripheral nerve injury has been suggested to induce SC de-differentiation and an iSC phenotype [11,17]. However, while there are clearly global changes in SC gene expression post-injury [17,40,62], whether a specific embryonic SC stage is recapitulated, and which set of gliogenic transcription factors are deregulated, remains poorly understood. We thus assessed our panel of transcription factors in "repair" SCs by performing a crush injury on the P60 sciatic nerve, and assessing gene expression at P65, 5 days post injury (dpi).
We failed to detect Pax3 protein in Sox10 + SCs 5 dpi (S11J-S11L Fig), even though Pax3 transcripts have been isolated from SCs following acute injury [43]. Similarly, Etv5, which marks satellite glial cells and is downregulated in maturing SCs [42], was not expressed in Sox10 + SCs (S11P-S11R Fig). Finally, AP2α, which is expressed in SCPs and involved in negatively regulating SC maturation [39], was also absent in the distal segment at 5 dpi (S11D-F). In summary, with the exception of Egr2, SCs retain the expression of the core transcriptional program of a SC identity post-injury, including Sox9, Nfatc4 and Yy1. In addition, a 'repair' SC phenotype is characterized by an increase in expression of the SC lineage markers Sox2, Jun and Oct6, but other embryonic SC markers (AP2α, Pax3 and Etv5) are not induced, suggesting that SCs do not de-differentiate to a particular embryonic state.

Discussion
Generation of SCs from NCCs is a progressive process characterized by at least five transient embryonic stages of development. Here we have defined these developmental stages by examining the expression patterns of 12 transcription factors (Sox2, Sox9, Sox10, AP2α, Pax3, Nfatc4, Etv5, Jun, Yy1, Egr1, Egr2, Oct6) (Fig 11). While other studies have performed more global analyses of gene expression in SC lineages between E9.5 and P0 [67], similarly delineating the dynamic changes that occur over these time points, our focus on protein expression levels of a core set of transcription factors provides a framework for scientists to reliably follow temporal identity transitions in this lineage. Based on this temporal profile, we can begin to delineate the potential diverse functional contributions of these genes during SC development and their recapitulation following peripheral nerve injury.
At E9.0, the earliest NCC stage is characterized by the expression of Sox10, Sox9, AP2α, Pax3 and Nfatc4, a gene expression profile that is maintained at E10.5. However, a distinct feature of NCC precursors coalescing in the DRG is the initiation of Etv5 expression and concomitant loss of Nfatc4 expression. Next, E12.5 SCPs in the DRG and dorsal and ventral roots retain the expression of Sox10, Sox9, AP2α, and Pax3 while also initiating the expression of Sox2 and Egr1, but losing the expression of Etv5. This is followed at E14.5 by the loss of Pax3 expression, while Sox10, Sox9, AP2α, Sox2 and Egr1 continue to be expressed in iSCs. A more dramatic change in gene expression is observed in E18.5 late immature/pro-myelinating SCs, which express Sox10, Sox9, Jun, Oct6 and Yy1, but lose the expression of AP2α, Sox2 and Egr1 in the nerve. At P7 and P65, we observed a very similar expression profile as seen at E18.5 (i.e., Sox10, Sox9, Jun, Oct6, Yy1) with the added expression of Nfatc4 and Egr2 at P7 and P65, and Egr1 and Sox2 expression in the P65 nerve.
Following peripheral nerve injury, SCs assume a transient 'de-differentiated' phenotype that is critical for supporting axonal regeneration and nerve repair. Hence, we characterized expression of the transcription factor panel following injury in the adult P65 nerve, comparing it against the embryonic and postnatal profile. We reveal that denervated SCs upregulate the expression of only a subset of early glial-lineage transcription factors. Indeed, expression of genes associated with both SCPs (Sox2, Egr1) and pro-myelinating/late immature SCs (Oct6, Jun) are elevated following denervation of adult SCs, while genes involved in the myelination program are actively lost (Egr2). Importantly, the absence of expression of several embryonic genes in denervated SCs (AP2α, Etv5, Pax3) may provide a strategy to enhance or prolong the repair phenotype to improve recovery of function following PNS injury or neuropathy.

Sustained expression of Sox9 and Sox10 across the Schwann cell lineage
An intricate network of transcription factors controls the timely development of peripheral glia, including several of the transcription factors examined in this study. One of the core regulators of SC development is Sox10, which we used to mark peripheral glial cells in our co-labeling experiments. Indeed, we (this study) and others [26,28] found that Sox10 is continually expressed in Schwann and satellite glial cells throughout development (Figs 1-9) and into adulthood. Prior studies have revealed that Sox10 is required for the specification and terminal differentiation of iSCs, and to maintain a peripheral glial phenotype [28,49]. Mechanistically, Sox10 directly activates Egr2 expression, acting synergistically with Oct6 and Nfatc4 [30][31][32][33][34] to induce the expression of peripheral myelin genes such as myelin basic protein (MBP), myelin protein zero (MPZ), myelin associated glycoproteins and connexin-32 (Cx32) [32,34,45]. Consequently, deletion of Sox10 results in loss of Egr2 expression, as well as myelin sheath degeneration and axonal death resulting in declined nerve conduction [29]. However, conditional Sox10 ablation studies also revealed that Sox10 is essential for survival of early migrating trunk NCCs, but not the survival of adult SCs, indicating that it is a critical player early in the SC lineage [29,68,69] and later to maintain functional myelination.
Interestingly, we also found that Sox9 follows a similar temporal pattern with sustained, overlapping expression with Sox10. Sox9 induces a NCC phenotype [25], and its expression biases migrating NCCs towards glial and melanocyte lineage selection [26]. While previous studies have observed Sox9 expression in the peripheral nerve at E14.5 [58], earlier stages of Fig 11. Summary of temporal expression profiles of key transcription factors in the SC lineage. Sox9 and Sox10 are expressed throughout SC genesis, beginning at E9.0 in migrating NCCs and persisting until P65 in myelinating and non-myelinating SCs in the sciatic nerve. AP2α, Pax3 and Etv5 are also expressed in NCCs, persisting until E12.5 in SCPs with Nfatc4 expression being restricted to NCCs. Egr1 and Sox2 expression is initiated in SCPs at E12.5, persisting until E14.5 in iSCs. At E18.5, AP2α is expressed in the DRG but is undetectable in the nerves while the expression of Jun, Oct6, and Yy1 is initiated and persists till P65. Rare Sox2 + Sox10 + positive cells are also detected in the P65 nerve. Egr2 expression is not detected until P7 in myelinating SCs in the sciatic nerve. Post-injury, high upregulation in expression of Sox2, Oct6, and Jun is observed, while distinct nuclear Egr1 expression is also detected. The straight lines represent continued expression of the markers through the different stages, while the dotted lines represent declining or low expression. The asterisk indicates where expression is restricted to the DRG and is undetectable in the nerves at E18.5. Markers expressed in the nerve after an acute injury have been denoted with ' p '. The green cells represent the developing axon, while the beige cells represent the NCCs at E9.0-E10.5 and the developing SCs at E12.5 to post-natal stages. doi:10.1371/journal.pone.0153256.g011 Sox9 expression have not been documented. Interestingly, recent work has suggested that isolated human SCs show negligible Sox9 expression, but Sox9 is over-expressed in neurofibromatosis 1 tumor derived SCs [70], hinting to a role for Sox9 in promoting SC proliferation. It is possible that sustained Sox9 activation in Schwann cells may enable re-establishment of the immature SC phenotype and re-entry into the cell cycle, and should be addressed in future studies.

Transcriptional regulators expressed at early stages in the Schwann cell lineage
In addition to Sox9 and Sox10, four other transcriptional regulators in our panel were expressed at the NCC stage; AP2α, Pax3, and Etv5, with Etv5 appearing one day later than the others. One novel observation was that Nfatc4, which is a calcium-responsive transcription factor, exhibits transient early expression in NCCs but is rapidly lost by E10.5, before re-initiating expression in maturing SCs at P7. A role for Nfatc4 in early NCCs has not previously been reported, however at later stages, Nfatc4 has been reported to bind a myelin specific enhancer in Egr2, cooperatively with Sox10, to activate Egr2 and other myelin genes during the pro-myelination to myelination transition [45]. AP2α expression persists until 18.5, however, AP2α + Sox10 + cells at this stage are confined to the DRG, and AP2α expression is not seen in the Sox10 + SCs lining the nerve. This could be suggestive of a role for AP2α in satellite glial cells in the DRG. Also, AP2α is co-expressed with majority of Sox10 + cells throughout development, except at E18.5 when AP2α expression becomes restricted to the DRG. Interestingly, overexpression of AP2α in vitro blocks the transition of SCPs to iSCs [39], even though we found that this transcription factor is expressed in iSCs and pro-myelinating SCs. The in vivo requirement for AP2α in the SC lineage has not yet been determined.
Pax3 is co-expressed with most Sox10 + glial lineage cells at early embryonic stages, and may play a role in regulating the proliferation of these early glial cells. Indeed, Pax3 induces proliferation in SCPs, and Pax3 transcript levels decline at the onset of differentiation [43]. Finally, we also found that Etv5 expression is limited to E10.5 NCC precursors in the DRG and in satellite glia cells at later stages. The function of Etv5 in the SC lineage has not been elucidated, although misexpression of dominant negative Etv5 in NCCs affects neuronal and not glial specification [54]. The absence of Etv5 at all later time points of the SC lineage (including after injury) suggest that it does not play a role in SC differentiation or myelination.

Transcriptional regulators expressed at late stages in the Schwann cell lineage
Several of the transcription factors in our panel were not expressed in NCCs, but were expressed in definitive cells in the SC lineage, including Sox2, Egr1, Jun, Oct6, Yy1 and Egr2. In contrast to Sox9 and Sox10, Sox2 is expressed in a more limited window of SC development, appearing in E12.5 SCPs and E14.5 iSCs. A decline in number of Sox2 + Sox10 + cells is observed at E14.5. This data is consistent with previous studies demonstrating that Sox2 expression declines upon neuronal commitment, and continues at low levels in SCPs and iSCs [23]. Interestingly, persistent Sox2 expression suppresses myelin-associated genes such as Egr2 and MPZ whilst maintaining cells in an undifferentiated state [22]. Notably, cross-repressive interactions between Sox2 and Mitf/Egr2 regulate the differentiation of SCPs into either myelinating SCs or melanocytes [24]. Sox2 is thus considered a negative regulator of myelination.
The zinc finger transcription factors Egr1 (Early growth response 1) and Egr2 have nearly identical DNA binding domains but opposite effects on myelination; Egr2 (Krox-20) promotes the differentiation of SCs to a myelinating phenotype while Egr1 is a non-myelinating SC marker [16,35]. Hence, we (this study) and others [35] found that Egr1 is expressed in SCPs but is downregulated as the cells mature. At postnatal stages, Egr1 expression is also re-initiated in non-myelinating SCs [35], where a modest cytoplasmic expression pattern is observed, and is sustained in a subset of SCs within the adult (P65) sciatic nerve. Conversely, Egr2 transcripts are detected in the dorsal and ventral roots from E10.5 onwards but are absent from the SCs in the DRG and peripheral nerves throughout embryogenesis [15]. Strikingly, we did not detect Egr2 protein in the dorsal and ventral roots, suggesting that it may not be translated until postnatal stages. Indeed, we (this study) and others [35] observed Egr2 protein in SCs lining the postnatal peripheral nerve, which is expected considering its requirement for myelination. e observed Jun expression in late immature/pro-myelinating SCs, P7 and in rare Jun + Sox10 + SCs in the adult nerve (P65). Downregulation of Jun expression is mediated by Egr2 just prior to onset of myelination [40]. Overexpression of Jun has been associated with a decline in myelination and de-differentiation of SCs, along with a reduction in Egr2 and MPZ levels [40]. Similarly, we detected Oct6 expression in a subset of Sox10 + pro-myelinating SCs at E18.5, in mature SCs at P7 and in rare cells at P65. Oct6 acts in synergy with Sox10 to induce Egr2 expression, which in turn promotes the expression of several myelin proteins [34,37]. Oct6 deficient mice show a transient arrest at the pro-myelinating stage, which is overcome by P10, with a late onset of Egr2 expression and myelin formation [37]. Oct6 not only promotes myelination by promoting terminal differentiation from pro-myelinating to myelinating SC via induction of Egr2, but also prevents premature myelination by repression of MBP and MPZ. A progressive reduction in Oct6 levels allows MBP and MPZ to be activated, thereby initiating a temporally controlled myelination program. Notably, constitutive overexpression of Oct6 results in a persistent hypomyelination phenotype in mice and gradual axonal loss [71], suggesting that varied levels of Oct6 allow for diverse functional contributions across the SC lineage.
Yy1, expressed at E18.5 and postnatally at P7/P65, is important for attaining the myelination phenotype, such that conditional knockdown of Yy1 in SCs results in hypomyelinated nerves with deficient expression of MPZ and Pmp22 [36]. The Egr2 promoter and Myelinating Schwann cell Element (MSE) has multiple binding sites for Yy1. Activation of MEK pathway occurs in response to the axonal signaling molecule, Neuregulin1. This in turn results in serine phosphorylation of Yy1. The phosphorylated Yy1 is recruited to binding sites in the Egr2 promoter and MSE, activating Egr2 expression [36] and underscoring its important role in SC myelination.

Injury activates SC lineage genes that recapitulate features of both Schwann cell precursors and pro-myelinating Schwann cells
Following a nerve crush injury we observed continued expression of Sox10 in nerve SCs and co-expression with Sox9, Nfatc4 and Yy1. Previous work showed that Sox9 is expressed in isolated SCs from P3 nerves [58] but its presence in adult SCs in vivo or following injury has not been reported. Constitutive expression of Sox9 in adult (uninjured) SCs and following injury in vivo, suggesting that Sox9 may play a continued role in the maintenance of the SC fate or in sustaining SC competence to re-acquire a de-differentiated state, particularly since it has a demonstrated role in both induction and maintenance of self-renewal capacity in subependymal neural stem cells [72] and various epithelial stem cell types [73,74]. Indeed, SC de-differentiation encompasses the hallmark features of a stem cell, exhibiting both the capacity for self-renewal, and the ability to generate mature cell types. Future conditional knockout studies will need to be done to determine the role of Sox9 in adult SCs and its potential contribution to the de-differentiation process.
Acutely injured SCs exhibit robust activation of the myelin-inhibitory gene Sox2, as has been previously reported [22], and elevated levels of Egr1, both of which are unique to SCPs and iSCs and absent in late immature/pro-myelinating SCs (Fig 10). However, several other transcription factors that are upregulated in denervated SCs are markers of late immature/promyelinating SCs, including Jun, and Oct6 (Fig 11). SC de-differentiation was also associated with a concomitant loss of mature myelinating genes such as Egr2.
Egr1 is a transcriptional activator that is normally active during cell cycle-re-entry [75]. Although the frequency of Egr1 + Sox10 + SCs did not change, Egr1 protein exhibited a marked increase in intensity and nuclear translocation following injury, suggesting that denervation causes a change in Egr1 function [35]. Egr1 may thus be an important modulator of SC plasticity. Egr1 and Egr2 appear to play opposing roles in modulating the acquisition of non-myelinating versus myelinating phenotypes [35]. Sustained Egr1 expression in a subset of SCs at all postnatal stages, and its known role in regulating cell cycle entry, suggests that this factor might also enable SC proliferation after injury and/or activation of other genes that are necessary for de-differentiation, partly through its known cooperation with Egr3 [76]. Future studies using conditional knockout approaches will need to be done in order to determine the ultimate role for Egr1 and Egr3 in acquisition of the de-differentiated SC state.
Interestingly, the frequency of pro-myelinating genes Yy1 and Nfatc4, did not change following injury further indicating the retention of late immature/pro-myelinating SC traits. Despite their putative pro-myelination function, many Yy1 + and Nfatc4 + cells that co-localized with Sox10, were also mitotically active, suggesting that both of these transcription factors are permissive of the proliferative, de-differentiated state.
The most notable change after injury was observed with respect to Jun and Oct6 expression. A robust increase in Jun expression has been reported in denervated SCs [40,62]. Indeed, Jun is a critical regulator of de-differentiation such that loss of Jun results in an inability to downregulate myelination genes and severely impairs regeneration. The POU domain transcription factor Oct6 initiates the transition from pro-myelinating to myelinating SC [77]. Interestingly, a shift from cytoplasmic to nuclear localization in Oct6 expression has been reported in axonal regeneration in peripheral neuropathic conditions [78]. It is noteworthy that Oct6 expression is highly upregulated at a time when myelin is being degraded and SCs are repressing their myelin program. Oct6 could be playing multiple roles in governing SC function, a possibility that requires further exploration. Since neither Jun nor Oct6 are expressed in SCPs but only by late immature/pro-myelinating SCs at E18.5, it suggests that at the level of protein expression, the de-differentiated SC state cannot be equated to a single embryonic SC stage, but rather is unique and includes features from multiple developmental stages.
Several key transcription factors including AP2α, Etv5, and Pax3 that are associated with early stages of development were not expressed within denervated SCs. We observed only extremely rare Pax3 + cells in the adult nerve (data not shown). A recent report suggests that Pax3 labels approximately 1% of cells in the adult nerve and marks non-myelinating SCs [44]. This suggested that possibly only a subset of Sox10 + non-myelinating SCs express Pax3. Notably, we did not observe Pax3 expression following injury either, despite seeing robust expression in NCCs that were immunostained in parallel as a positive control. This is in contrast to a previous study [43] that reported Pax3 transcripts were detected in the denervated distal stump at seven days post transection injury. This discrepancy may be due to several factors: 1) the temporal expression profile of Pax3 may be delayed, such that it does not peak until after the 5 day time-point we examined, 2) severity of nerve injury (transection versus crush) may be an important determinant of the SC transcriptional response within denervated SCs, 3) young mice (3 weeks of age) may elicit a different response compared to the adult animals (P65) used in our study, and 4) Pax3 transcripts may not be translated. Future studies using a conditional Pax3 gene deletion could determine its role in establishing the SC repair phenotype after injury.
Our spatio-temporal expression study provides a comprehensive glimpse into the expression profiles of the various transcriptional regulators involved in SC development and in SC injury response. To summarize, the injured peripheral nerve contains a highly dynamic and heterogeneous population of glia that undergoes phenotypic reversion to a de-differentiated state by recapitulating a subset of early glial-associated transcription factors. Taken together, we provide evidence that "repair" SCs retain their core SC transcriptional program, while initiating the expression of a subset of embryonic genes that represent several embryonic SC stages. Since SC function is diminished in the adult aging PNS it may be necessary to artificially activate additional genes, particularly those that fail to be re-activated or are diminished following prolonged denervation, in order to maximize nerve regeneration.