The glial cell-line derived neurotrophic factor (GDNF) family of trophic factors, including GDNF, neurturin (NTN), persephin (PSP), and artemin (ART) are members of the transforming growth factor-β family and are secreted molecules that play a fundamental role during inductive events of organogenesis such as kidney formation, as well as in the cell survival and differentiation of the developing nervous system. The GDNF family receptors are multicomponent complexes formed by the tyrosine kinase receptor RET and one of the GPI-linked GDNF family receptors (GFRα1-α4) that binds the ligand with high affinity and determines ligand specificity. RET is unable to bind GDNF on its own but can be activated when GDNF is in complex with GFRα1 (Jing et al., 1996; Tang et al., 1998). The two GDNF receptor components do not necessarily have to act in a cis configuration in a cell autonomous way. Consistently, GFRαs are much more widely expressed than Ret, and their expression are not always colocalized with Ret.
In addition to regulating neuronal survival, GDNF has been shown to stimulate neurite growth in cultured enteric and motor neurons (Zurn et al., 1996; Schafer and Mestres, 1999). During motor neuron regeneration, GDNF and GFRα1 are expressed in Schwann cells particularly in the distal nerve stump. This finding agrees with alternative mechanism of Ret signaling that is consistent with the findings showing that GDNF elicits a variety of effects at different developmental stages and on several different cell types.
Based on the observation that GDNF supports the survival of postnatal but not embryonic dorsal root ganglion (DRG) neurons in vitro, despite that its receptors are expressed at both ages, we postulated that GDNF might have other functions in the embryonic ganglia (Baudet et al., 2000). In a recent microarray experiment, GDNF treatment of DRG neurons was shown to lead to a regulation of several classes of genes, including genes related to neurite branching and axonal growth (Linnarsson et al., 2001). However, also GFRαs might be important in nerve growth. Both GDNF and GFRα1 are markedly up-regulated after sciatic nerve lesion in Schwann cells along the path of nerve regeneration (Naveilhan et al., 1997; Trupp et al., 1997) and GFRα1 can be released by the action of membrane proteases or phospholipases. Thus, based on these findings, two non–cell autonomous modes of so-called action in trans of GFRαs have been proposed. GFRα1 present on the membrane of Schwann cell can interact with growing nerves (Fundin et al., 1999), and alternatively, GFRα1 can be released from the Schwann cell and interact with the nerve in a soluble form (Naveilhan et al., 1997; Trupp et al., 1997, Paratcha et al., 2001). Localization studies of GFRα1 on terminal Schwann cells have indicated that it is localized in an apical–basal concentration gradient in these cells (Fundin et al., 1999). Such concentration gradients generate the positional information that is the basis for directional nerve growth (Knöll and Drescher, 2002). Consistently, uniformly presented GDNF with GFRα1 provided in a gradient either in a soluble form or bound to microbeads promotes directional growth of chick sympathetic and nodose ganglion neurons (Ledda et al., 2002). Naturally occurring release of GFRα1 from neuronal as well as Schwann cells has also been shown to potentiate GDNF signaling and nerve outgrowth in culture (Paratcha et al., 2001).
All previous studies on the roles of the broadly expressed GFRα1 on nerve growth have been limited to its roles as a GDNF co-receptor. Ligand-independent roles of GFRα have not been possible to address, because the neurons under study die in the absence of GDNF. In this study, we have used Bax−/− neurons that survive independent of trophic support in culture. This strategy has allowed us to characterize the effects of GDNF and soluble or bound form of GFRα1 (sGFRα1 and bGFRα1, respectively) on neurite growth, independent of survival. Presence of GDNF and GFRα1 receptors led to marked and different effects on neurite growth. Interestingly, GFRα1 elicited ligand-independent effects on neurite growth. Further investigation on the intracellular activity induced by the soluble form of GFRα1 was performed. Addition of ET-18-OCH3, a PLCγ inhibitor, specifically blocked the effect of soluble GFRα1 in a dose-dependent manner, whereas PD98059 (inhibitor of mitogen-activated protein kinase, MAPK), LY294002 (inhibitor of phosphoinositide 3-kinase, PI-3K), or PP2 (inhibits all known Src family kinases) had no effect on soluble GFRα1-induced neurite growth response.
GDNF-Induced Modulation of Neurite Growth
To explore the role of GDNF, independent of neuronal survival, on primary sensory neurons in vitro, we dissected DRG neurons from embryonic day (E) 12 Bax−/− mice. The neurons were dissociated and cultured on laminin-coated culture dishes under serum-free conditions. Neurites were scored in unit length (1–12, each unit being 85 μm), and the length was plotted in a frequency histogram. Addition of GDNF to these cultures caused a minor but significant shift toward longer neurites compared with control cultures without any trophic factors (Fig. 1).
Opposite Effects of Soluble and Bound GFRα1 Neutralized by GDNF
Previous results support an important role of GFRα1 in trans for neuronal growth and guidance. We, therefore, explored the biological significance of stimulation in trans by GFRα1 by studying the effect of soluble or bound GFRα1 on neurite length in the presence or absence of GDNF. The effect of different concentrations of sGFRα1 on neurite growth was investigated by the addition of sGFRα1 at a concentration of 50, 150, 300, or 600 ng/ml. sGFRα1 was found to cause a statistically significant reduction on neurite growth in all conditions compared with control cultures with no addition of trophic factor (not shown). Neurons grown only in the presence of soluble GFRα1 (sGFRα1) displayed a significant shift toward shorter neurites. Of interest, this effect of sGFRα1 was neutralized by the addition of GDNF (Fig. 2). Hereafter, sGFRα1 was used at a concentration of 300 ng/ml. To evaluate a putative effect of endogenously produced GDNF, we added neutralizing antibodies against GDNF, which did not alter the effect of sGFRα1 on neurite lengths.
DRG neurons were next cultured with GFRα1 immobilized on the surface of the culture dish. Bound GFRα1 in the absence of GDNF led to extension of neurites leading to overall increased length. This effect of GFRα1 was partially neutralized by the addition of GDNF (Fig. 3). Furthermore, in control cultures, the distribution of neurite length was not affected with soluble or immobilized Baculo-Fc (not shown). To investigate whether the different effects of sGFRα1 and bGFRα1 are due to a difference in concentration and/or level of dimerization, we used enhancer antibodies for aggregating tagged soluble GFRα1 before its addition to the culture. This method did not cause any statistical difference from the effect of sGFRα1 when added alone (not shown). Sole enhancer antibody did not have any effect on neurite growth compared with control conditions with no factor added to the cultures, neither did control cultures with soluble Baculo-Fc or enhancer antibody and Baculo-Fc.
We next examined if the effects of sGFRα1 and bGFRα1 are limited to neurons cultured on laminin that interacts with α1β1 integrin by culturing the neurons on fibronectin. Neurons cultured on fibronectin (binding α5β1 integrin) showed an overall smaller neurite growth compared with those grown on laminin, consistent with that neural crest and sensory neurons use primarily α1β1 integrin for migration and neurite outgrowth. sGFRα1 and bGFRα1 elicited similar responses on fibronectin as on laminin. Neurons grown only in the presence of sGFRα1 displayed a significant shift toward shorter neurites and those on bGFRα1 significantly longer neurites. Furthermore, these effects were reversed by GDNF (data not shown).
Soluble GFRα1 and PLCγ Activity
Next, we examined the intracellular activity induced by the soluble form of GFRα1 by using selected inhibitors with their respective activities in the known RET and GFRα1 signaling pathways indicated in Figure 4.
Neither MAPK inhibitor PD98059 (10 μM), PI-3 kinase inhibitor LY294002 (50 μM), nor PP2 (0.33 μM), which inhibits all known Src family kinases, had any effect on the sGFRα1-induced neurite growth response. However, addition of ET-18-OCH3 (1 μg/ml), a known PLCγ inhibitor, blocked specifically the effect of soluble GFRα1. This effect was further investigated by adding various concentrations of ET-18-OCH3 to the cultures (50, 10, 1, 0.1, 0.01 μg/ml). A total of 50 μg/ml was found to be toxic to the cells. Although 10 and 1 μg/ml was found to inhibit the effect of sGFRα1 in a dose-dependent manner, 0.1 and 0.01 μg/ml did not block the effect of sGFRα1 (Fig. 5).
Several studies have shown that GDNF and GFRαs are important for neurite growth (Naveilhan et al., 1997; Trupp et al., 1997; Paratcha et al., 2001; Ledda et al., 2002) and both GDNF and GFRαs are expressed during the period of nerve growth and target innervation in the developing sensory nervous system (Fundin et al., 1999). However, a major difficulty in studying the role of GDNF and GFRα1 in axon growth at early developmental stages has been that GDNF does not promote DRG neuron survival, and its putative role in neurite outgrowth at these stages, therefore, has not been established. Furthermore, in studies of chick neuronal populations at developmental periods where GDNF supports neuronal survival, it has not been possible to examine axon outgrowth in vivo or in vitro independent of the survival promoting effect of GDNF. The establishment of null mutant mice for the death-promoting member Bax, revealed that it is required for neonatal sympathetic neurons and facial motor neurons to undergo programmed cell death in the setting of trophic factor or target deprivation (Deckwerth et al., 1996). Cultured sympathetic and DRG neurons that are prevented from dying by Bax deficiency have been shown to grow neurites in the absence of NGF (Deckwerth et al., 1996; Lentz et al., 1999), although to a lesser extent than in the presence of NGF. The use of Bax-deficient mice permits the separation of the survival requirements for neurotrophic factors from their effects on guidance, target innervation, and differentiation. With the aim to investigate the role of GDNF and its receptor GFRα1 (in soluble or bound form) on neurite outgrowth in developing sensory neurons, we cultured dissociated Bax−/− DRG neurons in the presence and absence of GDNF and/or GFRα1. Our results show that GDNF, as well as GFRα1, modulates axonal growth of embryonic sensory neurons that does not depend on GDNF for survival. However, the DRG is heterogeneous and contains several subpopulations of neurons (Snider and Wright, 1996), and we cannot exclude that the response to GDNF and GFRα1 occurs in selective subpopulations in our assay.
GDNF-Independent GFRα1-Induced Neurite Growth Responses in Sensory Neurons
Our findings clearly show a shift toward shorter neurite lengths in the presence of sGFRα1, which is reversed by the addition of GDNF. GDNF alone caused a significant but minor shift toward longer neurites, and the most part of the length distribution (i.e., neurites within the length intervals 2–7, see Fig. 1) is not different within 95% statistical probability limits compared with control conditions without any addition of trophic factors. When GFRα1 is immobilized to the culture dish, an opposite response is triggered, i.e., the neurites are longer. The addition of GDNF shortens the neurites. Previous reports on GFRα1 have been limited to studying its effects on neurite outgrowth in the presence of GDNF, because the neurons otherwise die from trophic factor deprivation. In agreement with the conclusion that GFRα1 also elicits neurite growth activity in the absence of GDNF, neutralizing antibodies against GDNF did not alter the effect of sGFRα1. Furthermore, anti-GDNF antibodies added alone to the cultures did not have any effect compared with control cultures without any addition of factor. These results show that GFRα1 may signal without GDNF and that this signaling controls neurite outgrowth but not neuronal survival.
GFRα1 Presented From the Target of Innervation Guides Neurite Growth
GPI-anchored proteins are largely localized to detergent-insoluble sphingolipid and cholesterol-rich lipid microdomains that exist as phase-separated so-called “lipid rafts” in the plasma membrane, where downstream signaling components (e.g., Src family kinases, see Fig. 4) are enriched (Simmons and Ikonen, 1997; Brown and London, 1998). Mobilization of RET to lipid rafts leads to more efficient intracellular signaling. Tansey and coworkers showed that only GPI-anchored but not soluble or transmembrane-anchored GFRα1 could recruit RET to lipid rafts (Tansey et al., 2000). However, another report finds that RET is recruited to lipid rafts by cis as well by trans mechanisms (Paratcha et al., 2001). The recruitment in cis is mediated by the GPI-anchored GFRα1 receptor in the cytoplasmic membrane, whereas upon activation in trans, it is mediated by intracellular proteins that normally reside in these rafts. RET can be activated in trans by exogenous GFRα1 when present either as a soluble or immobilized molecule in vitro. After sciatic nerve lesion, there is a marked up-regulation of mRNA expression of both gdnf and gfrα1 in Schwann cells. The increased expression occurs in a gradient with the highest levels of expression found proximal to the site of transection in the distal nerve (Naveilhan et al., 1997; Trupp et al., 1997). These findings suggested that GFRα1 might be involved in nerve regeneration. Supporting evidence for this notion was found in a recent study where cultured neuronal cells, Schwann cells, and injured sciatic nerve were shown to release biologically active soluble GFRα1 to the extracellular space, which indicates a naturally occurring release of this molecule in vivo and a biological relevance of trans activation by GFRα1 (Paratcha et al., 2001). Both soluble and immobilized GFRα1 potentiated neuronal differentiation and survival in response to GDNF independently of GPI-linked GFRα1 (Paratcha et al., 2001). Exogenous GFRα1 in the presence of GDNF, provides positional information and guides the directional growth of chick nodose and sympathetic nerves in culture (Ledda et al., 2002). The ability of soluble GFRα1 to activate the receptor might explain the different tissue distribution of the transcripts in vivo: gfrαs are more widely expressed than ret in the nervous system (Trupp et al., 1996; Yu et al., 1998; Fundin et al., 1999; Baudet et al., 2000; Mikaels et al., 2000), and there is a rapid and dynamic regulation of GFRα1 expression in terminal Schwann cells and targets of innervation that coincide with nerve invasion and formation of specific sets of sensory endings (Fundin et al., 1999). Our results confirm and extend these findings and show that GFRα1 presented as an artificial target in either soluble or bound form markedly affects neurite outgrowth independent of GDNF. Thus, there is biochemical and functional evidence that GFRα1 presented from the target could both in soluble and membrane bound form influence neurite growth. In the present study, we also investigated the GDNF-independent signaling pathways activated by soluble GFRα1 in cultured DRG neurons by the addition of known specific inhibitors to PI-3 kinase (LY294002), MAPK (PD98059), Src family kinases (PP2), and PLCγ (ET-18-OCH3), respectively. We found that presence of ET-18-OCH3 specifically blocked the effect of sGFRα1-induced neurite growth response in a dose-dependent manner. It has been shown in vitro that GFRα1, to some degree, can interact with RET in the absence of GDNF (Eketjäll et al., 1999). It is possible that GFRα1 plays an active role in axonal growth and guidance by binding to Ret, but we cannot exclude an interaction with partners different from Ret. For instance, the nerve growth-regulating transmembrane glycoprotein L1 was shown recently to interact with neuropilin-1 and to be required for sema3A to elicit a response on DRG axons (Castellani et al., 2000). Analogous to our results, the repulsion of sema3A can be converted to attraction by presenting L1 in a soluble form instead of bound. Candidates for GFRα1 cross-talk with other receptors are the integrins that transduces the axonal growth signal promoted by both fibronectin and laminin.
GPI-Anchoring May Allow Multiple Effects of GFRαs
GFRα1 receptors might have different effects, depending on whether they are either in soluble or bound form (in trans) or GPI-anchored to the membrane of the responsive cell (in cis). Indeed, several other GPI-anchored molecules have been postulated to have dual effects, depending on whether they are in a soluble form or anchored to membrane (for review, see Faivre-Sarrailh and Rougon, 1997). Schwann cell-bound GFRα1 might activate RET located only at the closely associated axons and, therefore, act as a local source for axonal guidance/growth to their correct target and finally to participate in the formation of nerve endings. In our culture experiment, we investigated the role of this type of in trans activation by the binding of GFRα1 to the culture dish, which leads to a shift toward longer neurites. GFRα1 might also be cleaved off and released from the surface of the terminal Schwann cells. This type of activation was investigated by adding soluble GFRα1 in the medium, which leads to a shift toward shorter neurites. We investigated whether the different outcome of bound and soluble GFRα1 was due to a difference in concentration and/or level of dimerization of the receptor, by using antibodies for aggregating tagged soluble GFRα1. We found no difference in neurite growth response in the presence of sGFRα1 when added alone to the cultures or when sGFRα1 had been preincubated with enhancer antibodies. Our findings that GFRα1 elicits opposite effects when present in different trans configurations, even in the absence of GDNF, further broadens the view of the physiological role of these GPI-anchored molecules.
Mice heterozygous for Bax (Knudson et al., 1995) were mated and their embryos were collected at E12. The plug date was considered as E0. The genotypes were determined by a polymerase chain reaction–based technique by using DNA extracted from embryonic tissues (White et al., 1998).
Separate dissociated cultures of DRG neurons were set from individual E12 embryos in litters resulting from crosses of Bax+/−/Bax+/− mice. DRGs were collected in PBS/glucose. To dissociate the neurons, the DRGs were incubated at 37°C with trypsin/DNAse 0.05% (Gibco/BRL/ Sigma). After removal of the trypsin solution, the ganglia were washed once with DMEM/10% heat-inactivated horse serum and twice with defined medium (see below). The ganglia were then gently triturated to give a single-cell suspension. Non-neuronal cells were removed by preplating. The neurons were plated in duplicates or triplicates at a low density on 24-well plates (Nunc) and precoated with poly-D-lysine (PDL, 1 mg/ml, 30 min, Sigma) and laminin (20 μg/ml, overnight, Gibco/BRL) in a defined medium consisting of Ham's F14 supplemented with 2 mM glutamine, 0.35% bovine serum albumin, 60 ng/ml progesterone, 16 μg/ml putrescine, 400 ng/ml L-tyroxine, 38 ng/ml sodium selinite, 340 ng/ml triiodo-thyronine, 60 μg/ml penicillin, and 100 μg/ml streptomycin.
GDNF (Promega) was added at 150 ng/ml and GFRα1-Fc (R&D systems) at 300 ng/ml (Paratcha et al., 2001), if not stated otherwise. Goat anti-rhGDNF neutralizing antibody (R&D systems) was added at 1 μg/ml before addition of soluble GFRα1-Fc. For cluster experiments, anti-human IgG antibody (H+L, Jackson Lab.) was preincubated (300 ng/ml) together with GFRα1-Fc (300 ng/ml) 30 min at 37°C before addition to the cultures. Baculo-Fc (300 ng/ml) was preincubated with anti-human IgG (300 ng/ml) in additional control cultures.
Protein tyrosine kinase inhibitors (all from Calbiochem, CA) were used as follows: MAPK inhibitor PD98059 (10 μM); PI-3 kinase inhibitor LY294002 (50 μM), PP2 (1 μM and 0.33 μM), inhibits all known Src family kinases; PP3 (negative control for PP2, used at the same concentrations as PP2). A concentration of 1 μM PP2 and PP3 was found to be toxic (not shown). ET-18-OCH3 (PLCγ inhibitor) was used at 50, 10, 1, 0.1, 0.01 μg/ml.
The primary cultures were maintained for 48 hr at 37°C in a humidified incubator under 5% CO2. For studying the effect of immobilized GFRα1, the plates were coated with GFRα1-Fc (10 μg/ml) for 2 hr at 37°C, followed by the coating of PDL and laminin (as above). Immobilized or soluble Baculo-Fc (10 μg/ml and 300 ng/ml, respectively) was used on additional control cultures.
Neurite Outgrowth Assay
For the neurite outgrowth assay, neurons were fixed after 48 hr with 4% paraformaldehyde in phosphate-buffered saline (PBS). The cultures were washed in PBS and incubated in blocking buffer consisting of 1% bovine serum albumin, 0.3% Triton X-100, and 0.02% NaN3 in PBS (PBT) for 20 min and stained with anti-neurofilament (200 kDa) polyclonal antibody (Affiniti) at 1:500 in PBT using the ABC Elite kit (Vectastain), according to the manufacturer's instructions.
The length of the longest neurite of the bipolar neurons was measured with a grid and plotted in a frequency histogram by unit length (1–12, each unit = 85 μm). The relative frequencies fk/n were calculated for each group. Here, fk is the total number of neurons with neurite lengths between k μm and k + 1 μm, and n is the total number of neurons analyzed in one experiment. Each set of neurite lengths fk is semi-closed in the sense that neurites with length k + 1 belongs to fk whereas neurites with neurite length k does not belong to fk (fk ∈ ]k,k = 1]). For the confidence intervals of the difference between two groups (i.e., all cultures with addition factor, compared with control cultures without any factor), we proceeded as follows: estimates of the standard error of the mean of the difference are given by (Mansfield, 1996)
where Δk is the difference in group k for the control and the noncontrol group. fk,c is the relative frequency of neurons in group k for the control, and nc is the total number of neurons analyzed in the control experiment. fk,A is the relative frequency of neurons in group k for the noncontrol, and nA is the total number of neurons analyzed in the noncontrol experiment. Confidence intervals around each estimated relative frequency indicating 95% probability values is obtained by multiplying the standard error of the mean by λ0.025 = 1.96 (Mansfield, 1996). Confidence intervals indicating 95% probability values for the difference in relative frequency between each group of neurite length between the control and the noncontrol is given by Ik = [Δk + λ0.025 × d (Δk), Δk − λ0.025 × d (Δk)], where k = neurite lengths between k μm and k + 1 μm. If zero is not included in Ik, the control and the noncontrol is significantly different at the level of 95% (*) for the group of neurite length indicated by the index k.
The data are compiled from three separate culture experiments from at least 3 to 20 (n = 3–20) embryos of at least three separate matings. At least 150–300 neurons have been analyzed for all conditions. The neurons analyzed from different mice were assumed to be equally statistically distributed.
We thank Lars Edman for statistical advice and Lotta Johansson for secretarial assistance.