How scars shape the neural landscape: Key molecular mediators of TGF-β1’s anti-neuritogenic effects

Following injury to the peripheral and central nervous systems, tissue levels of transforming growth factor (TGF)-β1 often increase, which is key for wound healing and scarring. However, active wound regions and scars appear to inhibit process outgrowth by regenerating neurons. We recently showed that corneal wound myofibroblasts block corneal nerve regeneration in vivo, and sensory neurite outgrowth in vitro in a manner that relies critically on TGF-β1. In turn, delayed, abnormal re-innervation contributes to long-term sensory dysfunctions of the ocular surface. Here, we exposed morphologically and biochemically-differentiated sensory neurons from the ND7/23 cell line to TGF-β1 to identify the intracellular signals regulating these anti-neuritogenic effects, contrasting them with those of Semaphorin(Sema)3A, a known inhibitor of neurite outgrowth. Neuronal morphology was quantified using phase-contrast imaging. Western blotting and specific inhibitors were then used to identify key molecular mediators. Differentiated ND7/23 cells expressed neuron-specific markers, including those involved in neurite extension and polarization. TGF-β1 increased phosphorylation of collapsin response mediator protein-2 (CRMP2), a molecule that is key for neurite extension. We now show that both glycogen synthase kinase (GSK)-3β and Smad3 modulate phosphorylation of CRMP2 after treatment with TGF-β1. GSK-3β appeared to exert a particularly strong effect, which could be explained by its ability to phosphorylate not only CRMP2, but also Smad3. In conclusion, TGF-β1’s inhibition of neurite outgrowth in sensory neurons appears to be regulated through a highly-conserved signaling pathway, which involves the GSK-3β/CRMP-2 loop via both canonical and non-canonical mechanisms. It is hoped that by defining the signaling pathways that control neurite outgrowth in wound environments, it will become possible to identify optimal molecular targets to promote re-innervation following injury.


Introduction
Nerve injuries pose many challenges to patients, ranging from mild discomfort to life-long impairment due to pain, loss of sensitivity, motor function and/or autonomic control [1]. While adult peripheral nerves can regenerate, reinnervation does not always recapitulate the original anatomic distributions, nor do all patients with injured nerves regain appropriate functions [2,3]. Successful structural and functional regeneration depends on many factors, including the intrinsic growth capacity of the injured cells, the presence of a local, permissive environment and appropriate axonal guidance cues [4][5][6]. Wound environments are not generally permissive to nerve regeneration. One factor that contributes to this phenomenon is transforming growth factor (TGF)-β1, which is normally present at low levels in intact peripheral and central nervous systems, but increases massively both proximal and distal to injury sites [7,8]. TGF-β1 controls master switches for key events in extracellular matrix formation and wound healing [9,10], via both canonical, Small Mothers Against Decapentaplegic (Smad) signaling and noncanonical pathways [11]. The increased expression of TGF-β1 after injury plays positive roles by regulating the immune response, modulating neuronal phenotype and regulating expression of growth factors important for neuronal repair [12][13][14]. However, TGF-β1 also stimulates the formation of fibrotic scars at the lesion site; this can be problematic for nerve regeneration, as scar-forming cells can inhibit regrowth, plasticity and recovery of damaged neurons [15][16][17][18]. Scar formation is particularly detrimental in the cornea, where it decreases transparency and the ability to see. The cornea is also interesting in the context of TGF-β1-regulated scarring, because it is the most densely innervated peripheral tissue in the human body [19]. Corneal nerves arise from the trigeminal ganglion, and as schematically represented in Fig 1, are distributed through the anterior half of the stroma and the epithelium. They are predominantly mechano-sensory and nociceptive, and serve to protect the eye from outside elements [20][21][22]. Although corneal nerves are part of the peripheral nervous system and able to regenerate, wounding of the cornea often results in abnormal reinnervation, with surprisingly serious and long-lasting effects [23][24][25][26][27][28][29][30]. We recently showed that corneal myofibroblasts, which differentiate largely from stromal keratocytes in and around the wound site [31], inhibit the regrowth of corneal nerves into the wound area [16,18] -Fig 1. Key to the present experiments, this effect was reproduced in vitro, when corneal myofibroblasts were co-cultured with neurons derived from the ND7/23 cell line, they inhibited neurite outgrowth [16]. ND7/23 cells are created by fusing N18tg2 mouse neuroblastoma cells and neonatal rat dorsal root ganglion cells [32], and they are often used as an immortalized proxy for peripheral [including corneal] sensory neurons because they are readily differentiated by addition of nerve growth factor (NGF) [33][34][35]. Differentiated ND7/23 neurons extend neurites and express molecular markers that identify them as A-fiber mechano-sensors and C-fibers [36]-two major classes of sensory nerves found in the cornea and other peripheral tissues. Importantly, TGF-β1 was shown to be both necessary and sufficient for the anti-neuritogenic effect of myofibroblasts on differentiated ND7/23 cells [16], allowing us to now use this growth factor as a surrogate [in place of myofibroblasts] to define molecular pathways that inhibit neurite outgrowth in pure neuronal cultures (i.e., cellautonomously).
Neurites are projections from the filopodia or lamellipodia of a neuron, which can subsequently become polarized into an axon or dendrite [37,38]. During their elaboration, neurites undergo stereotypical changes [39] that rely on the response of the intracellular cytoskeleton to a variety of extracellular cues, including secreted factors [38]. Microtubules, cylindrical polymers composed of α-and β-tubulin heterodimers, are the primary cytoskeletal elements forming the core of cylindrical neurites [40]. Collapsin-response mediator protein-2 (CRMP2) functions as a carrier of free tubulin heterodimers [41], which it delivers to the assembly-plus ends of nucleating sites in growing microtubules, consequently promoting acetylation and neurite extension. We previously reported that TGF-β1-induced reduction in neurite outgrowth by ND7/23 cells is associated with increased phosphorylation of CRMP2 [16]. One class of molecules previously shown to phosphorylate CRMP2 in developing central neurons and a number of commercially-available neuronal cell lines is glycogen synthase kinase (GSK)-3β [42,43]. GSK-3s are Ser/Thr protein kinases whose GSK-3α and GSK-3β isoforms are ubiquitously expressed [44] and play important roles in fundamental cellular processes such as differentiation, proliferation, motility and inflammation [45]. The present set of experiments asked if the inhibitory effects of TGF-β1 on neurite outgrowth occur through activation (i.e., de-phosphorylation) of GSK-3β and whether this is the key event causing increased CRMP2 phosphorylation, and the associated decreased neurite elongation (Fig 1). However, because TGF-β1 is well-known to exert many of its effects through canonical Smad signaling [11], we also investigated the role of this pathway in CRMP2 phosphorylation and GSK-3β activation (summarized schematically in Fig 1). . Cells were multiplied until passage 10 (freshly thawed cells were considered passage 1). They were passaged by trypsinization (Corning Inc.) every 3-5 days, when approximately 70-90% confluent. Cells were seeded at a density of 5x10 4 /cm 2 . After attachment, they were washed twice with 1x Dulbecco's phosphate-buffered saline (dPBS; Thermo Fisher Scientific, Waltham, MA) and with serum free medium (SFM) and then incubated with SFM for 1 day to inhibit mitosis before adding nerve growth factor (NGF) [46,47]. Such SFM-primed cells were used in all experiments below, unless otherwise specified.

Phenotypic characterization of ND7/23 cells after TGF-β1 treatment
To investigate the effects of TGF-β1 on neuritogenesis, we examined the morphology of ND7/ 23 cells after 1 day in culture, and measured their relative expression of pGSK-3β/tGSK-3α/β and pCRMP2/tCRMP2 using western blots. As a control, we contrasted the strength of TGF-β1' effects with those of Semaphorin 3A (Sema3A), a well-known inhibitor of neurite extension [49][50][51]. While earlier studies [49][50][51] found 4ng/ml Sema3A sufficient to produce 50% collapse of DRG growth cones, 10ng/ml of recombinant Sema3A was necessary to exert a comparable effect in SFM-primed ND7/23 cells and was thus used in all experiments.
Quantitative analysis of neuronal polarity and neurite length. SFM-primed ND7/23 cells (8x10 4 cells/6cm primaria culture dish) were pretreated with 1, 10 ng/ml TGF-β1 (240-B-010, R&D Systems) or 10ng/ml Sema3A (1250-S3-025, R&D Systems). After 1hr, either 0.5ng/ ml or 50ng/ml rNGF (R&D Systems) was added for 1 day. Phase-contrast images were captured using an inverted Olympus IX73 microscope (Center Valley, PA) under 20X magnification. A 19-hole imaging template was used for analysis, covering~8.6% of the dish area and 100 cells/dish. The development of polarity by cultured cells can be divided into 5 stages [39,40], but only stages 1 to 3 were evident after 1 day in our culture system (see examples in Fig  3A-3C). In Stage 1, short neurites emerge from the cell body, extending as protrusions and lamellipodia (lamellipodial veil). In Stage 2, some lamellipodia are replaced by short (less than 10μm) neurites, which appear as flattened processes with extensive, protrusive activity at their tips. In Stage 3, one of these neurites elongates at a faster rate, developing into an axon-like process (at least 10μm longer than the other neurites) [52], causing the neuron to become polarized. The CellSens imaging software (Olympus CellSens Standard ver. 1.12, Olympus) was used to manually tag and count the number of cells in Stages 1, 2 and 3 in each of the 19 photographs taken for each culture plate. For each of the neurons in Stages 2 and 3, we then traced and measured the length of neurites extruded using the CellSens imaging software. From these tracing, we counted the number of neurites >40μm in length (two-fold longer than the average diameter of a cell soma) per image. These analyses allowed us to evaluate the neuronal differentiation attained following exposure to different factors, as well as the proclivity for significant (>40μm long) neurite extension.

Statistical analysis
When three or more intervention groups were compared, inter-group differences were assayed with one or two-way ANOVAs, followed by Tukey's post-hoc tests, as appropriate. When only two groups were compared, two-tailed paired or unpaired Student's t-tests were performed. A probability of error of P<0.05 was considered statistically significant.

ND7/23 cells become neuron-like after sequential treatment with SFM and rNGF
When cultured with 10% FBS/DMEM, ND7/23 cells proliferated rapidly, with a strong tendency to form clumps that were loosely attached to the culture dish and occasionally became free-floating (Fig 2A). When FBS was removed from the culture medium to block proliferation, the majority of the attached, clumped cells spread into a monolayer. After 1day of addition of rNGF, the cells began to display clear neurite outgrowth, with the majority belonging to the Stage 1 phenotype. Over ensuing days, an increasing proportion of ND7/23 cells transitioned to Stage 2 or Stage 3 morphologies (Fig 2B), similar to DRG neurons in culture [55]. After 7 days, ND7/23 cell cultures developed aggregations of rounded, phase-bright cell bodies and an extensive network of dendrites that spread over the culture dishes' surface between these aggregations (Fig 2C).
We then asked if the morphological changes observed in ND7/23 cells were accompanied by the appearance of neuron-specific markers and molecules involved in neurite extension and polarization. Compared to undifferentiated ND7/23 cells (grown in 10% FBS, lane 1, Fig  2D), culture in SFM for 1 day caused an upregulation of GAP-43, Ac-Tub and CGRP (lane 2, Fig 2D). Even more dramatic changes were seen in these molecules after rNGF was added (lanes 3 -6, Fig 2D). The relative expression of the NGF receptor TrkA was only seen to increase after addition of rNGF (lanes 3 -6, Fig 2D). NF-L and t-CRMP2 increased more slowly, peaking at 7 days in culture (lane 6, Fig 2D). Although substance P was not expressed by proliferating cells (lane 1, Fig 2D) and did not appear during the first 4 days of culture with rNGF ( lanes 3-4, Fig 2D), it was induced in detectable amounts after 5 and 7 days in culture (lanes 5 -6, Fig 2D). In contrast, the expression of t-GSK-3β and β-actin (the loading control) remained relatively unchanged throughout (lanes 1-6, Fig 2D).  6). The relative expression of TrkA, GAP43, acetylated α-tubulin (Ac-Tub), NF-L, CGRP, substance P, and t-CRMP2 all increased when cells were exposed to rNGF, albeit at different rates. t-GSK3 α/β expression appeared relatively stable. β-actin was used as a loading control. https://doi.org/10.1371/journal.pone.0234950.g002

Effect of TGF-β1 on intracellular signals associated with neurite outgrowth
Consistent with the notion that GSK-3β activity is a key regulator of NGF's neuritogenic effects [56] (S2 Fig), 100ng/ml rNGF dramatically increased levels of p-GSK-3β (the inactive form) while leaving levels of t-GSK-3 relatively unaffected (Fig 4A and 4B-lane2). As a result, there was a significant increase in the ratio of p-GSK-3β/t-GSK-3 after treatment with 50-100ng/ml rNGF (two-tailed Student's t-test: t 2 = 4.48, p = 0.046). Of note, the rNGF-induced increase in p-GSK-3β/t-GSK-3 ratio could be blocked effectively by the addition of 0.6μg/ml anti-NGF antibody (S2 Fig). rNGF with or without TGF-β1 or Sema3A and probed with antibodies against phosphorylated GSK-3β (at Ser9) or total GSK-3α/β. Levels of phosphorylated (i.e. inactive) GSK-3β (lane 2) are increased after treatment with 100ng/ml rNGF compared to 0.5 ng/ml rNGF (lane1). Both 10ng/ml TGF-β1 and 10ng/ml Sema3A blocked this increased phosphorylation of GSK-3β. (B) Plot of the ratio of p-GSK-3β/t-GSK-3α/β normalized to levels obtained following stimulation with 100ng/ml rNGF and under the different conditions shown in A. Data are means ± standard deviations. (C) Representative Western blot of SFM-primed ND7/23 cells treated with 100ng/ml rNGF alone (lane 1), or with TGF-β1 or Sema3A with and without lithium chloride (LiCl), a blocker of GSK-3β activity (see Fig 5). Both TGF-β1 and Sema3A cause an increase in levels of phosphorylated (p)-CRMP2 compared to total (t)-CRMP2 levels (compare lane1 with lanes 2,3). Pre-treatment with LiCl (lanes 4 and 5), which maximally phosphorylates (i.e. inactivates) GSK-3β prior to the addition of TGF-β1 or Sema3A completely blocks the ability of these two molecules to generate p-CRMP2. (D) Plot of the ratio of p-/t-cRMP2 measured from blots such as those shown in C and normalized to lane 2 (10ng/ml TGF-β1 condition). Values are means ± standard deviations.
Our prior works showed that TGF-β1 stimulation regulates phosphorylation of CRMP2 in ND7/23 cells [16]. Here, we replicated this result, showing that addition of 10ng/ml TGF-β1 or 10ng/ml Sema3A to SFM-primed ND7/23 cells cultured with 100ng/ml rNGF were equally effective at increasing levels of p-CRMP2, doing so by about 1.5-fold (Fig 4C and 4D). In fact, when the ratio of p-CRMP2/t-CRMP2 was normalized to levels attained following stimulation with 100ng/ml rNGF, there was no significant difference between the effects of 10ng/ml of TGF-β1 and 10ng/ml Sema3A (Fig 4D, two-tailed Student's t-test: t 4 = 0.04, p = 0.970).
Since GSK-3β is a candidate kinase thought to regulate CRMP2 phosphorylation and thus, its activity [42], the prediction is that when TGF-β1 and Sema3A decrease p-GSK-3β in ND7/ 23 cells, this should be associated with a corresponding increase in levels of p-CRMP2. Indeed, when SFM-primed ND7/23 cells were treated with 100ng/ml rNGF, the ratio of p-GSK-3β/t-GSK-3β increased (lane 2, Fig 4A, and 4B), levels of p-CRMP2 were relatively low (lane 1, Fig  4C and 4D). In other words, CRMP2 was in its active (non-phosphorylated) form, associated with neurite extension (see Introduction , Figs 1 and 6). When either 10 ng/ml TGF-β1 or 10ng/ml Sema3A were added, the relative levels of p-GSK-3β decreased (Fig 4A and 4B) and those of p-CRMP2 increased (Fig 3C and 3D). Because levels of total GSK-3 and CRMP2 remained relatively unchanged, we concluded that an inverse relationship exists between levels of p-GSK-3β and p-CRMP2 in differentiated ND7/23 cells.
We then critically tested this hypothesis using lithium chloride (LiCl), a natural GSK-3α/β inhibitor, which is thought to work by activating an upstream protein kinase (AKT) and blocking protein phosphatases-actions that ultimately saturate the phosphorylation of GSK-3α/β on Ser 21/9 [53]. We verified that pre-treatment of ND7/23 cells with 10mM LiCl blocked the ability of both TGF-β1 and Sema3A to convert p-GSK-3β to GSK-3β; the end result was a maintenance of high levels of p-GSK-3β/t-GSK-3 in the cells (lanes 4 and 5, Fig 4C). Associated with this effect, LiCl completely blocked the ability of both TGF-β1 and Sema3A to increase p-CRMP2 relative to t-CRMP2 (compare lane 2-3 with lanes 4-5, Fig 4C and 4D). A one-way ANOVA for p-/t-CRMP2 ratios across all treatment groups in this experiment showed a significant main effect of treatment (F (4,8) = 8.3, p = 0.006), with post-hoc Tukey HSD tests confirming significant differences between TGF-β1, Sema3A and TGF-β1+LiCl, Sema3A+LiCl conditions (p<0.05 in all cases). Therefore, it appears that levels of p-GSK-3β and p-CRMP2 are inversely and causally linked in ND7/23 cells, and that manipulations which either increase (e.g. LiCl) or decrease (e.g. TGF-β1, Sema3A) p-GSK-3β levels impact p-CRMP2.

Discussion
In the present study, we developed an optimized protocol to generate morphologically and biochemically-differentiated sensory neurons from the ND7/23 cell line. We then used these differentiated neurons to detail the signaling pathways mediating the inhibitory effects of TGF-β1 on neurite outgrowth. Specifically, we uncovered that TGF-β1 works via both canonical and non-canonical signals, with the latter appearing to exert the strongest influence on CRMP2 phosphorylation, and consequently, neurite elongation.

ND7/23 cells become neuron-like after sequential treatment with SFM and rNGF
Because our experiments were performed in ND7/23 cells, an immortalized cell line often used as an proxy for peripheral sensory neurons [16,[33][34][35][36], it was essential to first attain reliable differentiation of these cells. Neuroblastoma cells express few NGF receptors and can thus have low sensitivity to this growth factor [57,58]. We overcame this hurdle by first "starving" ND7/23 cells with SFM for 1 day in order to increase responsiveness to NGF [55]. As a result, cells began to reliably express high levels of TrkA upon subsequent exposure to NGF, sustaining those levels across multiple (up to 7) days in culture. These cells also extended neurites, becoming multipolar (Stage 1), then pseudo-unipolar (Stages 2-3). Neurite extension was associated with increased acetylated α-tubulin (necessary for stable microtubule assembly), as well as NF-L. In addition, differentiated cells expressed neuronal markers such as GAP43 (associated with growing neurites), CGRP and substance P, which characterize major sub-populations of sensory nerves in the cornea [20]. Thus, in contrast with other neuroblastoma cells, which can extend neurites, but do not become polarized [59], we successfully developed culture conditions that allowed us to differentiate ND7/23 cells into mature, polarized, sensory, peripheral neurons.
Treatment of SFM-primed cells with 50ng/ml rNGF for 1 day decreased the proportion of Stage 1 cells and increased the proportion of Stage 3 cells by a corresponding amount; the proportion of Stage 2 cells remained unaffected. This suggests that NGF caused a proportion of Stage 1 cells to reach Stage 2 and almost immediately transition into Stage 3, showing the greatest level of polarization with the extension of an axon-like process. Treatment of these cells with either TGF-β1 or Sema3A, even in the presence of 50ng/ml rNGF, prevented neurite elongation, and thus cell polarization. As such, there was no decrease in the proportion of Stage 1 cells, nor increase in Stage 3 cells. While we recently showed that ND7/23 cells co-cultured with corneal fibroblasts are sensitive to the effects of TGF-β1 [16], the cells in our prior publication were not SFM-primed, nor as carefully differentiated and staged as the cells in the present study. As such, we can now confidently state that TGF-β1 exert a strong, anti-neuritogenic effect on sensory neurons differentiated from the ND7/23 cell line.

Molecular substrates of TGF-β1 's anti-neuritogenic effects in differentiated, sensory neurons
We began our investigations by ascertaining that SFM-primed, differentiated ND7/23 cells strongly express both TGF-βRl and TGF-βRll in their soma and neurites. This confirms a simple molecular initiator for TGF-β1 signaling in these neurons, as previously reported for primary sensory neurons [60], and in contrast with some neuroblastoma cells, whose low responsiveness to TGF-β1 was due to low levels of its receptor(s) [61]. Our earlier observation that the inhibitory effect of TGF-β1 on neurite outgrowth is critically mediated by its receptor [16] is consistent with studies in embryonic hippocampal and human iPSC-derived neurons  [62], in cerebellar granule neurons [63], in experimental diabetic neuropathy and in embryonic DRG cells [64]. However, TGF-β1 was reported to increase neurite length in differentiated retinal ganglion cells [65], suggesting that the effect of TGF-β1 on neuronal morphogenesis may be influenced by cell type, developmental state, health or disease, experimental manipulation, as well as by the local microenvironment. Nonetheless, most of the evidence in both central and peripheral nervous systems suggests that TGF-β1 has largely antineuritogenic effects, and this is also the case in differentiated neurons from the ND7/23 cell line. In fact, in these cells, present results show that TGF-β1 exerted the same level of inhibition on neurite outgrowth as Sema3A [49][50][51], at the same effective concentration (10ng/ml). So how does TGF-β1 exert its anti-neuritogenic effect in cultured, sensory neurons?
We previously showed that in ND7/23 cells, TGF-β1 increases phosphorylation of CRMP2 [16], a molecule that delivers free tubulin heterodimers to the assembly-plus ends of growing microtubules, thus promoting acetylation and neurite extension [41]. Here, we asked how TGF-β1 increased p-CRMP2, taking our cues from primary DRG cells, where Sema3A exerts is anti-neuritogenic effect by increasing GSK-3β activity, causing phosphorylation of CRMP2 [42,66]. Conversely, neurotrophin-induced axonal outgrowth in hippocampal neurons occurs via decreased GSK-3β activity, which reduces phosphorylation of CRMP2 [42]. As such, it appears that in many neuronal types, CRMP2 is a physiological substrate of GSK-3β [67,68] and that GSK-3β regulates neurite outgrowth through phosphorylation of CRMP2. In ND7/23 cells, treatment with either TGF-β1 or Sema3A prevented NGF-induced increases in p-GSK-3β levels to a similar extent. Because this caused a simultaneous increase in p-CRMP2, it suggests that modulating GSK-3β activity was key to regulating neurite outgrowth of differentiated ND7/23 cells (Fig 6). Our experiments are the first to show that TGF-β1 and Sema3A have comparable anti-neuritogenic effects-via GSK-3β/pCRMP2-in differentiated ND7/23 cells. Additionally, since the effects of TGF-β1 and Sema3A were similar in magnitude and operated via the same intracellular pathways in ND7/23 cells as in other neuronal cell types [42,69,70], we posit that neuritogenesis may be regulated through a very conserved signaling pathway that relies on a GSK-3β/CRMP-2 loop (Fig 6).
Finally, the actions of TGF-β1 in both neurons and other tissues of the body are known to be mediated by canonical Smad2/3 signaling [11]. Although not known to be regulated by TGF-β1, Smad1 signaling, which is key for regeneration of peripheral sensory neurons, is critically regulated by GSK-3β [71]. Thus, we also asked if TGF-β1 stimulation recruited Smads (specifically Smad3) to mediate its anti-neuritogenic effects in ND7/23 cells, and how this was related to its regulation of GSK-3β activity. SIS3 did indeed suppress TGF-β1-induced phosphorylation of Smad3 in these neurons, just as it does in fibroblasts [54]. However, here, this action prevented TGF-β1 from increasing levels of p-CRMP2, without affecting changes in p-GSK-3β. In contrast, LiCl, which blocks GSK-3β activity, decreased Smad3 phosphorylation at 2 different sites (impacted by TGF-β1 [72] and NGF, respectively). That GSK-3β can phosphorylate Smad3 at Ser-204 was previously reported [72] but its ability to also influence Smad3 phosphorylation at Ser-423/5 is newly described here. Thus, it appears that GSK-3β not only phosphorylates CRMP2 directly, it additionally phosphorylates (and activates) Smad3 (Fig 6). This explains why LiCl, TGF-β1 and other ways of modulating GSK-3β activity have such a strong effect on p-CRMP2 and neurite outgrowth. However, as Smad3 is not a kinase, how its phosphorylation causes increased levels of p-CRMP2 in differentiated ND7/23 cells, and whether it is sufficient and necessary, remain to be determined.
In conclusion, the present study used an optimized protocol to generate morphologically and biochemically-differentiated sensory neurons from the ND7/23 line. In these cells, we quantified the anti-neuritogenic effects of TGF-β1, contrasting them with those of Sema3A. We identified a key role of GSK-3β and p-CRMP2 in mediating the effects of both molecules.
The prevalence of a GSK-3β/CRMP2 loop for regulating neurite extension across a wide range of neuronal types and species, suggests this to be a highly-conserved signaling pathway. Ultimately, it also means that areas of central and peripheral nervous system damage, which are characterized by dramatically-increased levels of TGF-β1 relative to the healthy environment, provide a direct, highly-effective inhibitory stimulus to prevent neurons from rapidly innervating (or re-innervating) damaged regions. In the cornea and other tissues where scars form, the region of fibrosis is denuded of sensory nerve endings, and thus, of sensation. This may be necessary initially, to avoid the largely noxious stimulation that may result from exposure to molecules characteristic of the wound environment. However, one could also envisage a time when it would be desirable to restore normal innervation, a situation that also applies to the central nervous system where axons may need to regrow through glial scars. Thus, understanding key signaling pathways that control neurite outgrowth in wound environments, and the seemingly ubiquitous nature of these pathways across neuronal types, should allow us to better target interventions so as to promote re-innervation in a wide range of injury conditions. One option of course, is to pharmacologically block TGF-β1's binding to its receptor using for example, SB431542 [16]. While this works well to restore neurite outgrowth in vitro, the danger in vivo is that blocking all of TGF-β1's effects may inhibit beneficial functions of this growth factor in tissue repair. One alternative is to block CRMP2 phosphorylation by modulating the activity of GSK-3β in neurons. LiCl is well known as a neuro-active drug, and as a non-selective blocker of GSK-3β activity [53]. In the last decade, several selective, blood-brain-barrier penetrant GSK3 inhibitors have been described (reviewed in [73]Among them is SAR502250 (a.k.a. UDA-680), which was recently shown to exhibit neuroprotective activity in rodent model of Alzheimer's disease [74] Topical application of such pharmacologics to the cornea have not yet been attempted. Nor do we know if topical applications to the ocular surface can sufficiently target corneal nerve terminals, of whether more sophisticated, cell specific approaches (such as transfection with genetically-encoded materials-e.g. [75,76] are needed to ultimately manipulate GSK-3β activity in damaged, corneal nerves.