EGFR Activation Mediates Inhibition of Axon Regeneration by Myelin and Chondroitin Sulfate Proteoglycans

See allHide authors and affiliations

Science  07 Oct 2005:
Vol. 310, Issue 5745, pp. 106-110
DOI: 10.1126/science.1115462


Inhibitory molecules associated with myelin and the glial scar limit axon regeneration in the adult central nervous system (CNS), but the underlying signaling mechanisms of regeneration inhibition are not fully understood. Here, we show that suppressing the kinase function of the epidermal growth factor receptor (EGFR) blocks the activities of both myelin inhibitors and chondroitin sulfate proteoglycans in inhibiting neurite outgrowth. In addition, regeneration inhibitors trigger the phosphorylation of EGFR in a calcium-dependent manner. Local administration of EGFR inhibitors promotes significant regeneration of injured optic nerve fibers, pointing to a promising therapeutic avenue for enhancing axon regeneration after CNS injury.

Failure of successful axon regeneration in the CNS is attributed not only to the intrinsic regenerative incompetence of mature neurons but also to the environment encountered by injured axons (17). The inhibitory activity is principally associated with components of CNS myelin and molecules in the glial scar at the lesion site (15). Recent studies have suggested that three myelin proteins—myelin-associated glycoprotein (MAG), Nogo-A, and oligodendrocyte myelin glycoprotein (OMgp)—collectively account for the majority of the inhibitory activity in CNS myelin (35). The inhibitory activity of MAG, OMgp, and the extracellular domain of Nogo-A (Nogo-66) may be mediated by common receptor complexes that consist of the ligand-binding Nogo-66 receptor (NgR) and its signaling coreceptors p75/TROY and Lingo-1 (814). However, the intracellular mechanisms transducing these signals to the cytoskeleton remain unclear. For instance, it is known that MAG and perhaps other myelin inhibitors are able to induce an elevation of intracellular calcium levels (1517), but it is not known how intracellular calcium signaling leads to inhibition of axon regeneration.

In a systematic approach to identify signaling events required for the inhibitory activity of CNS myelin, we performed a small molecule screen to search for compounds with the ability to neutralize neurite outgrowth inhibitory activity associated with CNS myelin. We screened approximately 400 well-characterized small molecules in a neurite outgrowth assay using cerebellar granule cells (CGNs) on an immobilized myelin substrate. The majority of compounds tested did not have a noticeable effect on neurite outgrowth, and a small number of them were toxic (table S1). Most prominently, several EGFR kinase inhibitors showed a remarkable ability to counter the effects of myelin inhibition, suggesting the involvement of EGFR kinase activity in the inhibitory effects of myelin inhibitors.

To confirm our observations, we tested two well-characterized EGFR inhibitors, a competitive inhibitor AG1478, and an irreversible inhibitor PD168393 (18) in neurite outgrowth assays. Both inhibitors, but not a control compound AG1288, effectively promoted neurite outgrowth from both CGNs (Fig. 1, A to C) and dorsal root ganglion (DRG) neurons (fig. S1) when grown on an immobilized substrate of either whole myelin or individual myelin inhibitors. In contrast, none of the treatments affected neurite outgrowth on a control poly-d-lysine (PDL) substrate (Fig. 1D). Similarly, EGFR kinase inhibitors were able to block neurite outgrowth inhibition by myelin in retinal explant cultures grown in a collagen matrix laden with myelin (Fig. 1, E and F). Although retinal explants are mixtures of retinal ganglion cells (RGCs) and other cells, dissociated cultures contained primarily CGNs and DRG neurons. Thus, it is likely that EGFR inhibitors act directly on neurons to block their inhibitory responses to myelin inhibitors.

Fig. 1.

EGFR kinase activity is required for myelin inhibition. (A to D) Quantitation of the effect of EGFR inhibitors on neurite outgrowth from p7-9 CGNs on myelin (A), Nogo-66 (B), MAG (C), or PDL (D). Data represent mean neurite lengths derived from at least three individual experiments, representing a total of about 500 neurons per condition. AG1478 and PD168393 treatments significantly enhanced neurite outgrowth on inhibitors [(A) to (C)] but not on PDL (D). The concentrations of each drug are as follows: AG1288: 1 and 10 μM; AG1478: 10 and 100 nM; PD168393: 10 and 100 nM. Statistical analyses were performed by analysis of variance (ANOVA) [myelin (A): F = 64.8, df = 7, P < 0.0001; Nogo-66 (B): F = 20.5, df = 7, P < 0.001; MAG (C): F = 21.91, df =7, P < 0.0001; PDL (D): F = 2.102, df = 6, P = 0.1767), followed by Dunnett's posttest to compare outgrowth on each inhibitor with or without drug treatments. Both AG1478 and PD168393 treatments on inhibitors, but not AG1288, were significant when compared with no treatment on inhibitors alone (P < 0.001, except for 10 nM PD168393 on MAG, where P < 0.05) [(A) to (C)]. None of the three drug treatments showed a significant change in outgrowth on PDL alone (D). Error bars show mean + SEM. (E and F) Effects of PD168393 on retinal neurite outgrowth inhibition by myelin. Retinal explants from P5-6 mice were cultured within a collagen matrix with and without myelin (10 μg/ml) in the presence or absence of PD168393 (10 nM) for 72 hours. Explants were fixed and stained with antibodies to tubulin (E). Scale bar, 200 μm. The mean total neurite lengths (+SEM, N = 10) are shown in (F). PD168393 significantly promoted outgrowth of explants when compared to myelin alone (*, Student's t test, P < 0.0001). (G) Wild-type (wtEGFR) and kinase-deficient EGFR (kdEGFR) HSV-infected CGNs were stimulated with 1-ng/ml EGF for 5 min and the lysates were immunoblotted with an antibody against pTyr1173 EGFR (pEGFR) and reprobed with antibodies to EGFR. (H) DRGs were infected with HSV viruses and plated on control and myelin substrates. Average neurite lengths from three different experiments (at least 100 neurons from each experiment) were obtained as above. kdEGFR significantly promoted DRG outgrowth on myelin when compared with wtEGFR (*, Student's t test, P < 0.01). Error bars show means ± SEM.

To complement our pharmacological results, we made recombinant herpes simplex viruses (HSVs) to transduce the expression of a mutant form of human EGFR in neurons (19, 20). The kinase deficient EGFR-K721A (kdEGFR) carries a point mutation in the kinase adenosine 5′-triphosphate–binding site (19), and when overexpressed, can inhibit the activity of endogenous EGFR in a dominant-negative manner (Fig. 1G) (20). As expected, DRGs infected with the mutant but not with wild-type EGFR exhibited extensive neurite outgrowth on myelin (Fig. 1H). Taken together, these results suggest that EGFR kinase activity is required for myelin-dependent neurite outgrowth inhibition.

We next examined the expression of EGFR in the adult nervous system. Using in situ hybridization, we found that EGFR is expressed in most parts of the mature nervous system, including the cerebral cortex, cerebellum, most DRG neurons, and RGCs (fig. S2, A to C). Furthermore, immunostaining with antibodies to EGFR demonstrated the presence of EGFR protein in both cell bodies and neurites of cultured DRG neurons (fig. S2D) and retinal explants (fig. S2E).

To examine whether myelin inhibitors directly influence the activity of EGFR in myelin-responsive neurons, we treated serum-starved CGNs with recombinant soluble myelin inhibitors and assessed the phosphorylation of the endogenous EGFR receptor. By blotting neuronal lysates with antibodies directed against phosphorylated EGFR, we found that both Nogo-66 and OMgp triggered rapid EGFR phosphorylation (Fig. 2, A to B). However, neither control alkaline phosphatase (AP) protein nor the chemorepellant Semaphorin3A induced detectable levels of EGFR phosphorylation (Fig. 2B). Additional evidence for EGFR activation was obtained by observing that Nogo-66–dependent extracellular signal–regulated kinase 1/2 mitogen-activated protein kinase (ERK1/2 MAPK) activation occurred in an EGFR kinase-dependent manner (Fig. 2C).

Fig. 2.

EGFR activation by myelin inhibitors. (A) Serum-starved CGNs were stimulated with 5 nM AP–Nogo-66 for durations indicated. Lysates were blotted with an antibody to pEGFR, stripped, and reblotted with an antibody to EGFR, as in Fig. 1. (B) CGNs were stimulated with AP (5 nM), Semaphorin3A (Sem, 100 ng/ml), AP–Nogo-66 (N66, 5 nM), AP-OMgp (OM, 5 nM), and EGF (1 ng/ml) for 4 min, and the lysates were probed as above. (C) Nogo-66 activates ERK1/2 MAPKs in an EGFR-dependent manner. CGNs were preincubated (30 min) with AG1478 or PD16933 (both at 100 nM) and stimulated with Nogo-66 for 10 min. Lysates were blotted with antibodies against phospho-ERK1/2 and ERK1/2. (D) NgR-dependent EGFR activation. CGNs were infected with lentiviruses expressing full-length (FL NgR) or dominant-negative NgR (DN NgR) (10, 12) and stimulated with AP–Nogo-66 as in (B). The lysates were blotted with an antibody to pEGFR and an antibody to NgR. (E and F) CGNs were preincubated with AG1478 (100 nM), EGTA (3 mM), BAPTA-AM (5 μM), Go6976 (100 nM), TAPI-0 (1 μM), PTX (100 ng/ml), PP2 (10 nM), and GM6001 (100 nM) and stimulated with AP–Nogo-66 (5 nM) (E) or EGF (1 ng/ml) (F) for 5 min.

How do myelin inhibitors activate EGFR in cultured neurons? Previously, we developed a truncated form of NgR that can bind to ligands but not its signaling coreceptors (911). When overexpressed, the truncated receptor can compete with the endogenous NgR for ligand binding, thus blocking the signaling pathways induced by myelin inhibitors (911). We found that expression of the truncated but not the full-length NgR efficiently blocked the EGFR phosphorylation triggered by Nogo-66 (Fig. 2D), suggesting that EGFR activation is NgR-complex dependent. Next, we performed cell surface binding and coimmunoprecipitation experiments but failed to detect either direct binding of EGFR to inhibitor ligands or a physical association with receptor components NgR or p75 (fig. S3). These results suggested that EGFR is not a likely receptor for myelin-derived inhibitors or part of the NgR receptor complex. It is known that in addition to activation by its cognate ligands, EGFR phosphorylation can also result from “trans-activation” by other signaling pathways (2123). Failure to detect EGFR in the NgR receptor complex suggests trans-activation of EGFR by means of signaling pathways downstream of the active NgR receptor. In support of this, we found that the extent of EGFR phosphorylation triggered by optimal concentrations of myelin inhibitors was comparable to that resulting from low concentrations (1 to 2 ng/ml) of epidermal growth factor (EGF) (Fig. 2B), reminiscent of what has been reported previously for EGFR trans-activation (22).

Several signaling molecules have been implicated in EGFR trans-activation, including calcium, protein kinase C (PKC), nonreceptor tyrosine kinases (Src and Pyk2), G protein–coupled receptors, and metalloproteases that generate EGF-like ligands (2123). We examined whether any of these mechanisms are involved in NgR-dependent EGFR trans-activation by inhibiting specific signaling pathways pharmacologically. As shown in Fig. 2E, only the calcium chelators EGTA and 1,2-bis(2-aminophenoxy)ethane-N,N,N′,N′-tetraacetic acid tetra(acetoxymethyl)ester (BAPTA-AM) substantially decreased EGFR phosphorylation resulting from Nogo-66 treatment. In contrast, a PKC inhibitor Go6976, metalloprotease inhibitors TAPI and GM6001, an Src inhibitor PP2, and pertussis toxin had no effect on Nogo-66–elicited EGFR phosphorylation. In further support of the idea of EGFR trans-activation, neither EGTA nor BAPTA-AM had any effect on EGF-induced EGFR phosphorylation (Fig. 2F). Moreover, treatment of cultured neurons with EGF or heparin-bound EGF triggered robust EGFR phosphorylation but did not inhibit neurite outgrowth (24), suggesting that EGFR activation may be a required, but not sufficient, signaling step in the response to myelin inhibitors.

In addition to myelin inhibitors, chondroitin sulfate proteoglycans (CSPGs) in the glial scar represent another major hurdle for regenerating axons. Consistent with previous observations that CSPGs induce an elevation in calcium levels in responding neurons (25), we found that EGFR inhibitors could neutralize the neurite outgrowth inhibitory activity of chicken brain-derived CSPG preparations in which the major components include neurocan, phosphacan, versican, and aggrecan (Fig. 3, A and B). Furthermore, soluble CSPGs added to serum-starved CGNs also elicited EGFR phosphorylation in a calcium-dependent manner (Fig. 3C). In contrast, neither growth cone collapse nor repulsive responses induced by Semaphorin3A were affected by EGFR inhibitors (Fig. 3, D to G). Because Semaphorin3A has previously been suggested to act independently of intracellular calcium (26), these results support the idea that EGFR might be a calcium-specific signaling molecule in axon guidance pathways.

Fig. 3.

Requirement of EGFR activity for CSPGs and Nogo-66 inhibition but not Semaphorin3A repulsion. (A) Retinal explants were grown in collagen gels with and without CSPGs (200 ng/ml, Chemicon) and PD168393 (100 nM) for 3 days, fixed and stained with antibodies to tubulin. Scale bar, 200 μm. (B) Average total neurite lengths of retinal explants in the conditions described in (A). PD168393 significantly increased neurite length of explants when compared with CSPGs alone (*, Student's t test P = 0.0015). Error bars show means + SEM. (C) CSPGs activate EGFR in a Ca2+-dependent manner. Serum-starved CGNs were stimulated with CSPGs for 5 min with and without 3 mM EGTA and the lysates were blotted. (D and E) Growth cone collapse after 30-min addition of AP–Nogo-66 (∼5 nM) and Semaphorin3A (Sema3A) (100 ng/ml) to embryonic day 13 (E13) chick DRG cultures infected with or without HSVs expressing wtEGFR or kdEGFR. Representative growth cone morphology is shown in (D) and the average percentage of collapsed growth cones + SEM from quadruplicate experiments in (E). kdEGFR significantly decreased growth cone collapse in response to Nogo-66 when compared with wtEGFR (ANOVA, F = 298, df = 8, P < 0.001, using Dunnett's posttest). (F and G) EGFR inhibitors did not affect the repulsive activity of Semaphorin3A. E14 rat DRGs were embedded in collagen adjacent to 293 cells expressing Sema3A. AG1288 (10 μM) and PD168393 (PD, 100 nM) were added at the same time, and cells were grown for 48 hours. Neurites were visualized by anti-tubulin staining. Representative images are shown in (F) and the quantitation from quadruplicate experiments in (G). Repulsive effects were quantified by comparing the ratio of proximal to distal neurite length (P/D) with respect to the 293 cell aggregates. No significant differences between AG1288 or PD168393-treated groups and their controls were detected by Student's t test. Error bars in (G) show means + SEM.

We next examined whether EGFR inhibitors introduced at a CNS lesion site could promote axon regeneration in an optic nerve crush model (2729). Immediately after injury in adult mice, gelfoam soaked in a solution containing PD168393 or vehicle [dimethyl sulfoxide (DMSO)] was placed around the crush site of the nerve and replaced after three days. The extent of axonal regrowth was assessed 2 weeks after injury by immunohistochemistry, which used antibodies to GAP-43 to detect regenerating axons. Although little regeneration was detected in DMSO-treated control mice (Fig. 4, A, C, and E), PD168393 treatment resulted in substantial axonal regrowth with a ninefold increase in the number of regenerating axons 0.25 mm beyond the injury site, as compared with control mice (Fig. 4, B, D, and E). To test the possibility that the observed axon regrowth after PD168393 treatment was a consequence of improved cell survival, we stained retinal sections with the antibody to the tubulin Tuj1, which stains RGCs in the retina, and counted surviving RGCs. No detectable effect of PD168393 on RGC survival was found (Fig. 4, F and G). The extent of axon regeneration observed after PD168393 treatment was comparable to that induced by a Rho inhibitor C3, but less than the combined effect of C3 and lens injury, a procedure that can enhance the regenerative capacity of RGC axons (28). We estimated that about 0.5% of RGC axons regenerated beyond the lesion site after PD168393 treatment. This low percentage of axon regeneration may be explained by previous observations that a small portion of postnatal RGCs retains rapid axonal growth ability (6). Independent experiments suggested that locally administrated AG1478 also induced significant regeneration of injured optic nerve fibers (fig. S3). In comparison with PD168393, the relatively weaker effects of AG1478 might reflect the reversible EGFR inhibition mechanism of this compound. These findings suggest that local blockade of EGFR activity could alleviate inhibitory influences and promote the regeneration of lesioned CNS fibers in adult mice.

Fig. 4.

PD168393 promotes optic nerve regeneration. (A to D) Representative images of optic nerves stained with antibodies to GAP43 from control [(A) and (C)] or PD168393-treated [(B) and (D)] mice. The injury site was identified visually and with lectin staining (marked by C). The images (C) and (D) are magnified views of the postcrush area. (C) The control nerve where few GAP-43 fibers are evident. (D) Numerous regenerating fibers, including some that have turned (filled triangle) as well as those that are straight (open triangle). Scale bar, 100 μm in (A) and (B) and 50 μm in (C) and (D). (E) Quantitation of regenerating fibers in control and PD168393-treated mice. Fibers were counted at 250-μm intervals from the crush site from three nonconsecutive sections and the number of fibers at a given distance was calculated as (27). There is a significant difference between DMSO and PD168393 treatment groups by ANOVA (F = 63, df = 1, P < 0.001, n = 6 for each group), with Bonferroni posttests at each distance indicating a significant difference between DMSO and PD168393 treatment at 250 μm (P < 0.001) and 500 μm (P < 0.01). Error bars show means ± SEM. (F and G) Anti-tubulin stained RGCs in control and PD168393-treated retinas. Note large axon bundle (filled triangle) and retinal ganglion cell (open triangle) in (F). Scale bar, 10 μm. (G) Quantitation of surviving RGCs in control and PD168393-treated mice revealed no difference in survival between the two groups (Student's t test, P = 0.349). The results are plotted against the results from noncrushed controls. Error bars show means + SEM.

Taken together, our results suggest that EGFR activation might be a critical signaling event downstream of the intracellular calcium influx produced by myelin inhibitors and CSPGs. Interestingly, we found that Erlotinib (Tarceva), an EGFR inhibitor approved for the treatment of cancer (30), could also block neurite outgrowth inhibition by myelin inhibitors (fig. S5). Thus, such compounds might prove useful for promoting axon regeneration after brain and spinal cord injury.

Supporting Online Material

Materials and Methods

Figs. S1 to S5

Table S1

References and Notes

View Abstract

Navigate This Article