Chondroitin sulfate proteoglycans (CSPGs) and myelin-based inhibitors are the most studied inhibitory molecules in the adult central nervous system. Unlike myelin-based inhibitors, few studies have reported ways to overcome the inhibitory effect of CSPGs. Here, by using regenerating adult dorsal root ganglion (DRG) neurons, we show that chondroitin sulfate proteoglycans inhibit axon assembly by a different mechanism from myelin-based inhibitors. Furthermore, we show that neither Rho inhibition nor cAMP elevation rescues extracellular factor-induced axon assembly inhibited by CSPGs. Instead, our data suggest that CSPGs block axon assembly by interfering with integrin signaling. Surprisingly, we find that nerve growth factor (NGF) promotes robust axon growth of regenerating DRG neurons over CSPGs. We have found that, unlike naive neurons that require simultaneous activation of neurotrophin and integrin pathways for axon assembly, either neurotrophin or integrin signaling alone is sufficient to induce axon assembly of regenerating neurons. Thus, our results suggest that the ability of NGF to overcome CSPG inhibition in regenerating neurons is probably due to the ability of regenerating neurons to assemble axons using an integrin-independent pathway. Finally, our data show that the GSK-3β-APC pathway, previously shown to mediate developing axon growth, is also necessary for axon regeneration.
Failure of axon regeneration in the adult central nervous system (CNS) is due to both a muted response of CNS neurons to axon injury and a CNS environment that is hostile to axon growth. Myelin-based proteins, such as Nogo, MAG and OMgp, together with chondroitin sulfate proteoglycans (CSPGs) in the glial scar, are two major categories of CNS inhibitory molecules demonstrated to impede axon regeneration after injury (for a review, see David and Lacroix, 2003). Data from several studies demonstrate that Nogo, MAG and OMpg block axon growth by binding to their common receptor complex that includes NgR, p75/TROY and Lingo (Liu, B. P. et al., 2002; Mandemakers and Barres, 2005; Mi et al., 2004; Park et al., 2005). Recent studies indicate that the inhibitory effects of these myelin-based molecules are mediated by Rho activation downstream of PKC (Sivasankaran et al., 2004), p75-TROY (Mandemakers and Barres, 2005; Park et al., 2005; Yamashita et al., 2002), and probably EGFR activation induced by calcium influx (Koprivica et al., 2005). As a result, inhibition of these signaling mediators is able to antagonize the effects of myelin-based inhibitors. By contrast, the mechanisms by which CSPGs block axon growth are less clear because no CSPG receptor has been identified. Early studies suggested that CSPGs may impede axon growth through interfering with extracellular matrix (ECM)-integrin interactions (McKeon et al., 1995; Smith-Thomas et al., 1994). Indeed, endogenous upregulation as well as overexpression of integrin has been shown to promote axon growth on CSPG (Condic, 2001; Condic et al., 1999). However, recent studies suggest that CSPG may inhibit axon growth by similar mechanisms to those of Nogo, MAG and OMgp through Rho activation (Monnier et al., 2003; Sivasankaran et al., 2004).
Axon growth requires both gene transcription for the synthesis of raw materials and coordinated assembly of cytoskeletal elements in the extending axon. CNS inhibitory molecules are thought to act primarily at the growth cone to shut down axon assembly. Axon assembly is a highly regulated process strongly influenced by extracellular factors during both development and successful peripheral nervous system (PNS) regeneration (Kuruvilla et al., 2004; Patel et al., 2000; Werner et al., 2000). Two well-studied groups of extracellular factors that mediate axon assembly are neuronal growth factors, such as neurotrophins, and adhesion proteins, which include ECM proteins and cell adhesion molecules (CAMs). Recent in vitro studies of highly purified neurons in defined media indicate that either neurotrophins or ECM proteins alone induce only limited axon growth from either PNS or CNS neurons during development (Goldberg et al., 2002; Lentz et al., 1999; Liu, R. Y. et al., 2002). However, neurotrophins and ECM together induce robust axon growth (Goldberg et al., 2002; Liu, R. Y. et al., 2002), suggesting that coordinated activation of neurotrophin and ECM-integrin signaling is necessary for efficient and long-distance axon extension. CNS inhibitory molecules could block axon assembly by interfering with either signaling cascades that drive axon assembly induced by extracellular factors, or the intracellular machinery that assembles axons, or both. Most previous studies have focused on how CNS inhibitors affect the intracellular axon assembly machinery in the absence of extracellular axon-promoting factors. Less attention has been paid to the interaction between CNS inhibitors and axon growth-promoting signaling activated by extracellular factors.
In this study, we investigated the effects of Nogo and aggrecan on axon assembly of adult dorsal root ganglion (DRG) neurons promoted by either nerve growth factor (NGF) treatment or a pre-conditioning lesion (PCL) in the presence of ECM laminin. We have found that Nogo has no effect on either NGF or PCL-induced axon assembly when DRG neurons are cultured on laminin. By contrast, aggrecan significantly blocks both NGF and PCL-induced axon assembly in the presence of laminin. Surprisingly, we found that the addition of NGF overcomes the inhibitory effect of aggrecan and induces robust axon assembly of PCL neurons over CSPGs. We also discovered that, unlike naive neurons that require simultaneous activation of both growth factor and integrin pathways for axon assembly, either growth factor or ECM-integrin signaling alone is sufficient to induce axon assembly of PCL neurons. Furthermore, NGF mediates axon assembly of PCL neurons by a distinct ECM-integrin independent pathway, suggesting that NGF can overcome the inhibitory effect of aggrecan by bypassing the requirement of ECM-integrin signaling for axon growth in PCL neurons. Finally, we show that the previously identified GSK-3β–APC pathway is a convergent point for axon assembly of PCL neurons induced by both NGF and ECM-integrin. Taken together, our data provide new insight into the mechanisms of CNS inhibitory molecules, and thus suggest a new direction to achieve axon regeneration over a hostile CNS environment.
Different effects of Nogo and aggrecan on axon assembly from adult DRG neurons
Most studies of CNS inhibitory molecules have examined their effects on developing neurons in the absence of extracellular signals that promote axon assembly (e.g. on poly-D-lysine in the absence of neurotrophic factors). Since extracellular axon-promoting factors are required for efficient axon growth, it is of considerable interest to study the effects of inhibitory molecules on adult, especially regenerating neurons, in the presence of defined signals promoting axon growth (e.g. NGF, ECM, etc.). We first tested how Nogo affects axon assembly from adult naive DRG neurons cultured on laminin in the presence of NGF. Our results showed that Nogo had no inhibitory effect on NGF-induced axon assembly from adult naive DRG neurons plated on laminin (Fig. 1A). Additionally, regenerating adult DRG neurons primed with conditioning lesions and plated on laminin were also able to overcome the inhibitory effect of Nogo (Fig. 1A). These results are consistent with earlier studies showing that signaling events activated by laminin can abolish the inhibitory effect of CNS myelins (David et al., 1995), and a more recent study that laminin overrides the inhibitory effect of MAG (Laforest et al., 2005).
In significant contrast, aggrecan, a component of CSPGs, significantly abolished NGF-induced axon assembly of naive neurons cultured in the presence of laminin (Fig. 1B,C). Immunostaining (supplementary material Fig. S1) showed that NGF-induced ERK phosphorylation was not affected by aggrecan, suggesting that ECM-integrin signaling is probably the major target of CSPGs (McKeon et al., 1995). Further, PCL neurons plated on a mixture of laminin and aggrecan also exhibited minimal axon growth (Fig. 1D). To determine whether other CSPGs had similar growth inhibitory effects we plated PCL neurons on a commercially available mixture of CSPGs. Similar inhibitory effects were observed when a mixture of CSPGs was used instead of aggrecan (data not shown). These data are in agreement with previous results that regenerating neurons triggered by cell dissociation (axotomy) cannot overcome the inhibitory effect of CSPGs of the glial scar (Davies et al., 1997). Taken together, our results show that in the presence of signals promoting axon growth, myelin-based inhibitors and CSPGs act through distinctly different pathways to interfere with local signaling events that regulate axon assembly.
NGF but not Rho inactivation stimulates robust axon assembly from PCL neurons on CSPG
It is unclear how CSPGs block axon assembly, because no CSPG receptor has been identified. Recent studies suggest that CSPGs may inhibit basal axon assembly in the absence of extracellular signals by activation of Rho, EGFR or PKC, similar to the mechanism of Nogo, MAG and OMgp (Koprivica et al., 2005; Monnier et al., 2003; Sivasankaran et al., 2004). In fact, treatment of neurons with pharmacological inhibitors of Rho, PKC or EGFR completely rescued the basal axon assembly inhibited by CSPGs.
To examine whether similar mechanisms mediate the inhibitory effect of aggrecan on axon assembly of adult DRG neurons in the presence of axon-growth-promoting factors, we treated adult naive neurons with Rho kinase inhibitor Y27632. We found that the Rho kinase inhibitor did not overcome the inhibitory effect of aggrecan (Fig. 2A). A similar result was observed when a dominant-negative RhoA construct was expressed in naive neurons (data not shown). Similarly, addition of Y27632 to PCL neurons plated on aggrecan did not rescue axon assembly (Fig. 2B). These results indicate that axon assembly of adult neurons in the presence of extracellular factors can be inhibited by CSPGs and that Rho inactivation alone is not sufficient to reverse this axon growth inhibition. Neurons treated with PKC or EGFR inhibitors also did not show rescued axon assembly under similar experimental settings (data not shown). Moreover, previous studies have suggested that elevation of cAMP levels may underlie peripheral injury-induced intrinsic changes of DRG neurons that allow axon assembly on myelin-based inhibitory molecules (Neumann et al., 2002; Qiu et al., 2002). We therefore tested whether naive adult neurons in the presence of elevated cAMP could overcome aggrecan and grow axons. Our results showed that treating cells with db-cAMP did not promote axon assembly from naive adult neurons cultured on aggrecan (Fig. 2A). Taken together, these data further suggest that CSPGs inhibit extracellular-signal-induced axon assembly by different mechanisms to that of myelin-based inhibitors.
Surprisingly, when we added NGF to PCL neurons cultured on aggrecan, it significantly restored axon growth (Fig. 3A-E), even though NGF was unable to initiate axon assembly in naive neurons cultured on aggrecan (see Fig. 1B). The average axon length and the percentage of neurons with axons are similar to those of control neurons on laminin (Fig. 3D,E). Addition of other neurotrophic factors, NT3 and GDNF, together with NGF further increased the percentage of neurons that extend axons over aggrecan (data not shown). To determine whether these effects of NGF are through activation of the TrkA receptor followed by PI3K activation, neurons were treated with a TrkA receptor antagonist K252a. Axon growth from neurons treated with NGF and K252a cultured on aggrecan was significantly decreased. The phosphoinositide 3-kinase (PI3K) inhibitor, LY294002, had an even greater effect preventing NGF-induced growth on aggrecan (Fig. 3D,E). These results clearly demonstrated that adult regenerating neurons are able to overcome glial scar inhibitors when a neurotrophic pathway is activated. Axon growth of PCL neurons on laminin has been shown to be independent of transcription (Smith and Skene, 1997). The NGF-induced axon growth of PCL neurons over aggrecan was also transcription independent, as the transcription inhibitor DRB showed no inhibitory effect (Fig. 3E). Finally, a β1 integrin function-blocking antibody showed no effect on NGF-induced axon growth of PCL neurons (Fig. 3E), indicating that NGF induces axon growth over aggrecan by an integrin-independent pathway.
PCL allows adult DRG neurons to extend axons in response to either ECMs or NGF alone
In order to determine how a conditioning lesion allows neurotrophic factors to promote axon assembly of PCL neurons but not naive neurons over CSPGs, we compared how extracellular signals mediate axon assembly from naive or PCL adult DRG neurons.
We took advantage of the fact that there is a 24-hour window in which to study axon assembly from naive neurons. Longer culture times mimic axotomy effects and switch naive neurons into a regeneration mode similar to PCL neurons (Smith and Skene, 1997). Our results demonstrate that similar to the case for embryonic DRG neurons (Liu, R. Y. et al., 2002), NGF and laminin are both required as extracellular signals to mediate efficient axon assembly from naive adult DRG neurons (supplementary material Fig. S2). NGF or laminin alone were unable to induce significant axon outgrowth in naive neurons cultured for 24 hours. Importantly, experiments with a transcription inhibitor verify that there is a requirement for NGF in axon growth that is independent of its effects on gene transcription.
We then examined how axon assembly of PCL neurons was regulated. We found that no significant axon growth occurred in 24 hours when PCL neurons were cultured in the absence of any extracellular factors (on poly-D-lysine in serum-free medium). However, in significant contrast to naive neurons, when PCL neurons were plated in serum-free medium on laminin, long-distance axon growth occurred from roughly 45% of neurons (Fig. 4A-C). Presumably the other 55% of the DRG neurons that do not extend axons during the 24-hour period were either not axotomized by transection of the sciatic nerve, or extend axons so slowly that significant growth does not occur over this time frame. Interestingly, other ECMs, such as fibronection and tenascin C, were also able to induce axon growth (data not shown). However, L1 CAM, a non-integrin-mediated adhesion protein could not support significant axon growth from PCL neurons (supplementary material Fig. S3), suggesting a specific role of ECMs in mediating regenerative axon growth from adult DRG neurons. When we examined PCL-induced axon assembly in the presence of the β1 integrin function-blocking antibody, it significantly impaired axon assembly induced by laminin, whereas the control IgG protein had little effect (Fig. 4D), indicating that ECMs induce axon growth from PCL neurons through the activation of integrin signaling. To rule out the possibility that neurotrophins secreted by the neurons themselves mediate the axon assembly, the Trk inhibitor K252a was added to the medium. Results showed that K252a had no effect on regenerative axon growth (Fig. 4C), even though the same concentration of K252a significantly blocked NGF-mediated axon assembly. Together, these results indicate that axon growth of PCL neurons on laminin is mediated by the activation of integrin signaling alone.
To determine whether other growth factor signaling pathways can mediate axon assembly from PCL neurons, we screened various polypeptide growth factors in the absence of integrin activation. When NGF was added to the medium, we saw a significant increase in the percentage of PCL neurons that were able to extend long axons on poly-D-lysine (Fig. 4A). The percentage of neurons with axons includes the neurons that respond to both axotomy and NGF. In addition to NGF, NT3 and GDNF were also capable of promoting axon assembly from PCL neurons (data not shown), whereas the addition of IGF or EGF that also act through receptor tyrosine kinases did not support axon growth from these neurons (Fig. 4E). Additionally, the growth-promoting effect of NGF was TrkA dependent because K252a significantly attenuated NGF-induced axon assembly (Fig. 4E). When axon length was measured, we found that NGF could induce axon assembly from the PCL neurons to the same extent as that induced by laminin alone (Fig. 4F). Both integrin activation and NGF stimulation have been shown to inactivate RhoA (Arthur and Burridge, 2001; Yamashita et al., 1999), however when regenerating neurons were treated with a Rho kinase inhibitor Y27632 alone, little axon assembly was induced when compared with NGF or laminin treatments (Fig. 4F). This result suggests that inactivation of RhoA by itself is not sufficient to mediate the axon-promoting effects of integrin or NGF stimulation.
Since ECM-integrin is a major target of CSPGs, together our data suggest that the ability of NGF to promote axon assembly of PCL neurons over CSPGs is due to the fact that PCL allows neurons to extend axons in an ECM-integrin independent manner in response to NGF.
NGF and ECM-integrin signaling mediate axon assembly of PCL neurons by divergent signaling pathways
We next determined whether growth factor or integrin activation require the same downstream signaling mediators to promote axon assembly of PCL neurons. We first examined a series of pharmacological inhibitors to determine their effects on axon assembly of PCL neurons. We found that NGF-mediated axon assembly of PCL neurons requires classical PI3K and PKC activities (Fig. 5A). In significant contrast, ECM-integrin-mediated axon assembly from PCL neurons is independent of PI3K, the major signaling mediator of neurotrophins, but requires atypical PKC and Src activity (Fig. 5B). To confirm the fact that integrin-mediated axon assembly from PCL neurons does not depend on PI3K, we overexpressed a dominant-negative form of PI3K, previously shown to block NGF-induced axon assembly from embryonic neurons (Markus et al., 2002b), in PCL neurons. Results showed that dominant-negative PI3K had no effect on axon assembly from PCL neurons induced by laminin (Fig. 5C). We also tested whether Raf, another major signaling mediator of NGF, is required for ECM-integrin induced axon assembly of PCL neurons by expressing a dominant-negative form of C-Raf that is sufficient to block NGF-induced axon assembly (Markus et al., 2002b). We found that Raf activation was not required for ECM-integrin mediated axon assembly from PCL neurons (Fig. 5D). Finally, the requirement of Src for integrin-mediated axon growth was confirmed with a dominant-negative Src construct (Fig. 5E). These results indicate that neurotrophins and ECMs use distinct signaling mechanisms to mediate axon assembly of PCL neurons. Consistent with these functional data, western blot results showed that PI3K and ERK pathways were not activated in PCL neurons upon laminin stimulation (Fig. 6G). Together, these results suggest that, in PCL neurons, ECM-integrin mediates axon assembly by different pathways to NGF. As a result, CSPGs can specifically block ECM-integrin-induced axon assembly while having no effect on NGF-mediated axon assembly of PCL neurons.
Previously we showed that integrin-linked kinase (ILK) mediates NGF-induced axon assembly from embryonic DRG neurons downstream of PI3K (Zhou et al., 2004). We therefore examined whether ILK activity was also required for regenerative axon growth from adult PCL neurons. Application of a specific ILK kinase inhibitor KP74728 significantly blocked both NGF and laminin-induced axon assembly from PCL neurons (Fig. 5A,B), indicating that ILK is a converging signaling mediator in these neurons. Consistent with an important role of ILK, the protein level of ILK was significantly increased in DRGs after peripheral axotomy (data not shown).
We have shown previously that NGF can mediate axon assembly by the regulation of localized inactivation of GSK-3β and the interaction of APC with microtubules downstream of ILK and PI3K. Similar localization of inactivated GSK-3β and APC was observed at the distal tips of regenerating axons (Zhou et al., 2004), suggesting that the GSK-3β–APC pathway may also play an important role in regulating regenerative axon assembly in PCL neurons. To investigate whether inactivation of GSK-3β and the subsequent APC-microtubule interactions are necessary for axon assembly of PCL neurons plated on laminin, we expressed constructs that interfere with this pathway to determine how they affect laminin-induced axon assembly of PCL neurons. Results showed that expression of a control EGFP construct had little effect on axon assembly (Fig. 6A). By contrast, overexpression of a mutant GSK-3β, GSK-3β(S9A), which cannot be phosphorylated at Ser9, significantly blocked regenerative axon growth induced by laminin (Fig. 6B,E). This result indicates that inactivation of GSK-3β by phosphorylation is necessary for laminin to induce axon assembly from the PCL neurons. Consistent with this idea, plating PCL neurons on laminin increased GSK-3β phosphorylation, whereas laminin alone was not sufficient to induce GSK-3β phosphorylation in naive and embryonic DRG neurons in the absence of NGF (Fig. 6G). Finally, we examined whether the interaction between APC and microtubules is also required for integrin-mediated axon assembly. To interrupt this interaction, we overexpressed a mutant EB1 (C-EB1) that can sequester endogenous APC (Zhou et al., 2004). We found that C-EB1 almost completely abolished integrin-induced axon assembly from regenerating neurons (Fig. 6C,F), whereas C-EB1 lacking the APC binding domain had little effect (Fig. 6D-F). Together, these results indicate that the GSK-3β–APC pathway is a converging point for both ECM-integrin- and NGF-induced regenerative axon growth of PCL neurons.
In this study, we examined how two CNS inhibitory molecules, Nogo and aggrecan, affect axon assembly of naive and regenerating adult DRG neurons in the presence of extracellular factors that promote axon assembly. Under these experimental settings, we find that Nogo and aggrecan affect axon assembly of adult DRG neurons quite differently. Our data suggest that aggrecan interferes with an ECM-integrin-mediated extracellular signal that promotes axon assembly (Fig. 7). Thus, CSPGs inhibit axon assembly mediated by ECM-integrin signaling, which is necessary for axon assembly of developing and naive adult DRG neurons. However, a pre-conditioning lesion induces intrinsic changes in neurons so that PCL neurons are able to support ECM-integrin-independent axon assembly. As a result, neurotrophins can bypass the ECM-integrin signaling and promote axon assembly of PCL neurons in the presence of CSPGs (Fig. 7). Finally, we show that both ECM-integrin and NGF promote axon assembly of PCL neurons by a conserved GSK-3β–APC pathway.
CSPGs inhibit axon regeneration by a different mechanism to myelin-based inhibitors in the presence of axon-growth-promoting factors
Axon extension is controlled by intracellular machinery that assembles cytoskeletal elements and membrane components into new axons. An intracellular signaling cascade activated by extracellular factors is necessary to power the machinery and drive axon assembly, even though the basal activities of intracellular signaling molecules (e.g. kinases, second messengers, etc.) can mediate limited axon extension in the absence of extracellular factors. Recent studies suggest that CSPGs inhibit basal axon growth in the absence of extracellular signals by a similar mechanism to that of Nogo, MAG and OMgp (Monnier et al., 2003; Sivasankaran et al., 2004; Koprivica et al., 2005). In this study, we examined the effects of Nogo and aggrecan on axon assembly of adult DRG neurons in the presence of defined extracellular axon growth factors, including neurotrophins and ECMs. Our results clearly demonstrate that Nogo and aggrecan have different effects on axon assembly of adult DRG neurons.
First, when adult DRG neurons were cultured on laminin, Nogo had no inhibitory effect on axon assembly induced by either NGF or PCL. Since acute NGF treatment of DRG neurons was unable to overcome the inhibitory effect of MAG when neurons were cultured on poly-D-lysine (Cai et al., 1999), our result suggests that laminin acts to antagonize the effect of Nogo. Indeed, a previous observation has shown that laminin is able to overcome the inhibitory effect of CNS myelins (David et al., 1995). Moreover, a recent study has also demonstrated that laminin is able to override the inhibitory effect of MAG by regulation of Rac activity (Laforest et al., 2005).
By contrast, aggrecan significantly blocked axon assembly under the same experimental conditions. Furthermore, in NGF-induced axon growth, aggrecan did not affect ERK activation, the major signaling pathway downstream of NGF. This result indicates that NGF signaling is not a major target of aggrecan and implies that ECM-integrin signaling is probably the major target of aggrecan (Carulli et al., 2005; Smith-Thomas et al., 1994). This idea is further supported by the result that aggrecan blocked laminin-induced axon assembly of PCL neurons, as laminin-integrin activation is the only extracellular signal that mediates axon assembly under such conditions. The inhibitory effect of aggrecan shown here is unlikely to be due to its ability to directly interfere with the basic intracellular axon assembly machinery through Rho activation (see Discussion below). Together, the different functional effects of Nogo and aggrecan on axon assembly shown in this study suggest that myelin-based inhibitors and CSPGs inhibit axon assembly by different mechanisms. This difference between myelin-based inhibitors and CSPGs only becomes apparent when axon assembly is induced by defined extracellular axon growth factors, especially the ECMs. The effect of myelin-based inhibitors can be overcome by ECM-integrin activation, whereas CSPGs act to block ECM-integrin signaling (Fig. 7).
Second, both inactivation of Rho and elevation of cAMP have been used to overcome the inhibitory effect of myelin-based inhibitors (Fournier et al., 2003; Neumann et al., 2002; Qiu et al., 2002). We showed that these two treatments did not significantly rescue axon assembly of adult neurons plated on aggrecan, further indicating that CSPGs block axon assembly via different mechanisms from myelin-based inhibitors. Our interpretation of these conflicting results is that the Rho-mediated inhibitory effect mainly serves as a `brake' that inhibits the intracellular axon assembly machinery (Fig. 7). Therefore, Rho inactivation alone is not sufficient to drive axon assembly in the absence of axon growth-promoting signals (e.g. NGF, laminin). We hypothesize that the effect of Rho inactivation in promoting axon assembly on CSPG demonstrated in previous studies from Monnier et al. (Monnier et al., 2003) and Sivasankaran et al. (Sivasankaran et al., 2004) are due to the fact that neurons used in those studies exhibit limited degree of basal axon growth that is ECM-integrin independent. Taken together, we believe our data show that although CSPGs may impede basic axon assembly via Rho activation, they inhibit axon growth mainly through blocking integrin signaling (Fig. 7).
Finally, it is worth mentioning that neurons primed with PCL have been shown to overcome the inhibitory effect of myelin-based inhibitors (Neumann et al., 2002; Qiu et al., 2002). However, in our study, laminin-induced axon assembly of PCL neurons was still inhibited by aggrecan, further indicating that CSPGs act differently from myelin-based inhibitors to block axon assembly. This result is also consistent with in vivo observations that transplanted adult DRG neurons (mimicking PCL) can support robust axon growth over spinal cord myelin but stop at the boundary of CSPGs (Davies et al., 1997).
Either integrin or neurotrophin signaling can induce axon assembly in PCL neurons
It is well established that peripheral axotomy induces intrinsic changes in adult DRG neurons that lead to a robust axon regenerative response (for a review, see Snider et al., 2002). Although numerous studies have addressed changes in gene expression induced by axotomy (Costigan et al., 2002; Tanabe et al., 2003), little attention has been paid to axotomy-induced changes in the signaling events that mediate axon assembly. In this study, we show that in striking contrast to the situation in developing and naive neurons, which require simultaneous activation of integrin and growth factor signaling for axon growth, integrin activation by ECM alone is sufficient to induce robust axon assembly of PCL neurons. Since integrin signaling by itself is not sufficient to induce efficient axon assembly in the absence of neurotrophic peptides in developing neurons (for a review, see Goldberg, 2003), our study suggests that a different integrin-mediated pathway may specifically control regenerative axon assembly.
More importantly, we have demonstrated that robust axon assembly from PCL neurons is induced when NGF signaling is activated in the absence of ECM-induced integrin activation. Thus, when PCL neurons are plated on a neutral substrate, addition of NGF triggers axon growth similar in extent to that which occurs when PCL neurons are plated on laminin. This ECM-integrin-independent axon assembly is also unique to regenerating neurons. As a result of these changes, regenerating axons are presumably more adaptive to the local surface molecules, which may be markedly different from those encountered in the developing nervous system. How pre-conditioning lesion induces such changes is unclear. Interestingly, a similar growth phenomenon has been observed with cancer cells. Normal cells require both growth factor signaling and attachment to the ECM (integrin activation) for growth. By contrast, cancer cells are able to grow in suspension independently of integrin activation by ECMs (for a review, see Giancotti and Ruoslahti, 1999). The idea of parallels between signaling in regeneration and malignant transformation is supported by recent gene-profiling studies, in which a group of cancer-related genes was identified to be upregulated after a peripheral nerve injury (Cameron et al., 2003).
Furthermore, our studies also demonstrate that, in PCL neurons, ECM-integrin and NGF mediate axon assembly by different signaling pathways. ECM-integrin-induced axon assembly of PCL neurons does not depend on PI3K and Raf-ERK activities, which are major mediators of NGF signaling (for a review, see Markus et al., 2002a). However, by using a constitutively active GSK-3β construct and overexpressing a mutant APC binding protein, we demonstrate that the GSK-3β–APC pathway, previously shown to mediate NGF-induced axon assembly, is also crucial for integrin-mediated regenerative axon assembly of PCL neurons. This result indicates that both ECM-integrin and NGF need to converge onto the GSK-3β–APC pathway to mediate axon assembly of PCL neurons. Indeed, we show that laminin alone is sufficient to induce GSK-3β phosphorylation of PCL neurons, further supporting the idea that ECM-integrin and NGF mediate axon assembly independently of each other in PCL neurons. Given the importance of GSK-3β–APC pathway in mediating axon assembly of regenerating neurons, it will be interesting in the future to investigate how it is regulated downstream of integrin or neurotrophins in PCL neurons and whether it is the target of different CNS inhibitors. However, these results do not necessarily mean that the ILK–GSK-3β pathway is the only pathway that mediates axon growth downstream of both NGF and laminin-integrin signaling.
Neurotrophins mediate `regenerative' axon assembly on CSPGs by bypassing ECM-integrin signaling
Although much effort has been devoted to finding ways to promote axon regeneration over inhibitory CNS molecules, few studies have successfully induced robust axon assembly over CSPGs, the major components of the glial scar. The main reason is due to limited knowledge of mechanisms that underlie the inhibitory effect of CSPGs. Although recent studies suggest that CSPGs may use the same molecular mechanisms as myelin-based inhibitors, in this study treatments shown to overcome the myelin-based inhibitors failed to rescue axon growth induced by extracellular axon-promoting factors. By contrast, we show here, for the first time, that neurotrophins are able to induce long axon growth of PCL neurons in the presence of CSPGs. The rescue effect of NGF was significant when either axon length or percentage of neurons with axons was measured. Since our results, together with previous observations, suggest that CSPGs can inhibit axon assembly by interfering with ECM-integrin signaling, one way to overcome its inhibitory effect is to induce axon assembly by pathways independent of ECM-integrin signaling. Indeed, we find that in PCL neurons, either ECM-integrin or NGF can induce axon assembly independently of each other by distinct signaling pathways. As a result, neurotrophins are able to bypass the ECM-integrin signaling to induce axon assembly over CSPGs. Therefore, approaches that directly activate integrin signaling may be a potential way to promote CNS axon regeneration over glial scar.
Another way to overcome the inhibitory effect of CSPGs may be to increase the expression of integrins and thus their responsiveness to ECMs (Condic, 2001; Condic et al., 1999). Interestingly, several integrins are upregulated after sciatic nerve injury in DRG neurons (Wallquist et al., 2004). However, our data that PCL neurons were able to grow long axons in the absence of any ECMs (i.e. on poly-D-lysine) suggest that increases in responsiveness to ECMs may not explain the conditioning lesion effect. Instead, our results suggest that the activation of signaling mediators localized downstream of integrin at the convergent point with growth factor signaling, such as Src, ILK and FAK-Pyk2 (Ivankovic-Dikic et al., 2000), might underlie the ability of NGF to promote axon assembly of PCL neurons on poly-D-lysine or CSPGs.
We show here that GSK-3β inhibition is necessary for both NGF and ECM-integrin signaling in PCL neurons to mediate axon assembly, suggesting GSK-3β as a key regulator of axon regeneration. In addition, inhibition of GSK-3β has also been implicated in CNS regeneration by inducing multiple axons from hippocampal neurons (Jiang et al., 2005). However, global application of GSK-3β inhibitors on adult DRG neurons was unable to promote axon assembly over CSPGs (unpublished data), indicating that GSK-3β inactivation alone is not sufficient to overcome the inhibitory effect of CSPGs, or a localized inactivation of GSK-3β is required (Zhou et al., 2004). Indeed, NGF-induced activation and inactivation of GSK-3β simultaneously is necessary to ensure optimal axon growth by regulation of different microtubule-binding proteins (for a review, see Zhou and Snider, 2005).
In conclusion, our data indicate that CSPGs and myelin-based inhibitors block axon regeneration by different mechanisms in the presence of extracellular axon-growth-promoting factors. In addition to acting through the same pathway as Nogo by Rho activation, CSPGs may also interfere with extracellular axon growth promoting signaling activated by ECM-integrin. After a peripheral conditioning lesion, adult DRG neurons change their intrinsic properties, thus allowing either NGF or ECMs to induce axon assembly by distinct pathways. As a result, NGF is able to overcome the inhibitory effect of CSPGs by bypassing the ECM-integrin signaling in PCL neurons. Our results suggest a novel approach to overcome CSPG inhibitory influences, namely, the direct activation of signaling pathways downstream of integrin.
Materials and Methods
LY294001, Wortmannin, K252a, Y27632, BIS, Go6976 and PP1 were from Calbiochem (San Diego, CA). NGF was obtained from Harlan Bioproducts (Indianapolis, IN). The specific ILK inhibitor KP-074728 is an analog of the previously reported KP-392 (Tan et al., 2004). Aggrecan, IGF, EGF and db-cAMP were from Sigma (St Louis, MO). Transcription inhibitor DRB was from MP Biomedicals (Irvine, CA). Mouse anti-neurofilament antibody SMI-31 was from Sternberger/Meyer Immunochemicals (Jarrettsville, MD). Anti-β1 integrin (CD29, clone Ha2/5) was from BD Pharmingen (San Diego, CA). Phospho-GSK-3β, phospho-ERK1/2 and phospho-Akt (Ser473) were all from Cell Signaling Technology (Beverly, MA). L1 cell adhesion molecule is from R&D Systems (Minneapolis, MN). All fluorescence-tagged secondary antibodies were from Molecular Probes (Eugene, OR). Nogo66 peptide was a generous gift from Stephen Strittmatter (Yale University, New Haven, CT).
Dominant-negative PI3K, dominant-negative C-Raf, GSK-3β(S9A), C-EB1 and C-EB1ΔAPC constructs were generated as described previously (Markus et al., 2002b; Zhou et al., 2004). Dominant-negative Src was a generous gift from David Shalloway (Cornell University, Ithaca, NY).
Cell culture and immunocytochemistry
L4, L5 and L6 DRGs were dissected from 8- to 12-week-old adult naive or pre-conditioning lesioned (Liu and Snider, 2001) CF-1 mice and digested with collagenase (1 mg/ml) for 90 minutes followed by trypsin-EDTA (0.05%) for 15 minutes at 37°C. The DRGs were then washed three times with plating medium (MEM with L-glutamine and 1× penicillin/streptomycin) plus 5% fetal calf serum and dissociated with a 1 ml pipette tip in plating medium. Glass coverslips were coated with poly-D-lysine (100 μg/ml), laminin (10 μg/ml) with or without Aggrecan (52 μg/ml; Sigma) for 60 minutes before plating. DRGs were plated and cultured at 37°C in serum-free MEM containing N2 supplement. Embryonic DRG culture for western blot study was as described in (Zhou et al., 2004).
After 20-24 hours of culture, neurons were fixed in 4% paraformaldehyde for 20 minutes, washed in PBS and blocked in blocking solution (2% BSA, 0.1% Triton X-100 and 0.1% sodium azide in PBS) for 60 minutes. Coverslips were incubated with primary monoclonal anti-neurofilament antibody SMI-31 (1:200) for 1 hour, washed and incubated with goat anti-mouse Alexa Flour 594-conjugated secondary for 1 hour, washed in distilled H2O and attached to slides with Mowiol antifade mounting media.
Dissociated adult PCL neurons or embryonic neurons were plated on culture dishes coated with poly-D-lysine or laminin in the absence of NGF. After culture for 3-4 hours, neurons were lysed with lysis buffer and prepared for western blot assay. For embryonic neurons, cells cultured on laminin were also treated with NGF for 30 minutes before cell lysis.
Gene transfection and pharmacology
DRG neurons were transfected with various DNA constructs using the electroporation technique from Amaxa (Cologne, Germany). The transfection procedure was according to the Amaxa protocols for mouse neurons. Briefly, dissociated neurons were spun down to remove the supernatant completely and resuspended in 100 μl specified Amaxa electroporation buffer with 10-20 μg plasmid DNA. Suspended cells were then transferred to a 2.0 mm cuvette and eletroporated with an Amaxa Nucleofector™ apparatus. After electroporation, cells were immediately transferred to the desired volume of culture medium and plated onto coated coverslips. After neurons fully attached to the substrates (2-4 hours), the medium was changed to remove the remnant transfection buffer.
All growth factors and pharmacological reagents were added directly to the neuronal culture medium at indicated concentration at the time of plating. For anti-integrin antibody treatment, cells were incubated with anti-beta1 integrin (20 μg/ml) at 4°C for 30 minutes before plating. Cells were then fixed for analysis following overnight culture. For each plasmid and pharmacological treatment, at least three independent experiments were conducted.
Image analysis and statistics
Images were taken with Spot imaging software and a CCD camera (Diagnostic Instruments) attached to a Nikon Eclipse microscope. A 4× objective (0.45 NA) was used to record neurons with axons. 10-12 images were acquired for each coverslip. All image analysis was done with IPLab software. To quantify the percentages of neurons that grow axons, all cells of each experimental condition were recorded by the camera. The number of neurons with axons longer than two cell bodies was then counted. To measure axon length, the first 50-60 neurons with axons in each condition were selected for measurement. The longest axon of each neuron was traced manually and the length was then calculated.
All data were reported as mean ± s.e.m., and an unpaired Student's t-test was used to determine the significance of the data between groups. For multiple group comparison, a one-way ANOVA was used followed by a t-test post-hoc analysis.
We would like to thank Sam Snider for help in the data analysis, and You-Jun Chen for help with western blot experiments. This study was supported by a Spinal Cord Research Foundation fellowship (to F.Q.Z.) and NIH grants NS031768 and NS050968 (to W.D.S.).
Supplementary material available online at http://jcs.biologists.org/cgi/content/full/119/13/2787/DC1
- Accepted April 11, 2006.
- © The Company of Biologists Limited 2006