Growth cone transduction: G_o and GAP-43

The function of the central nervous system is absolutely dependent on the highly complex synaptic connections established during brain development. The ability of the extending neurite to identify and grow along appropriate pathways, and then to stop at synaptic targets, is mediated to a great extent by the axonal growth cone (reviewed by Lockerbie, 1987). The growth cone responds to extracellular signals such as soluble factors, matrix components and cell-bound adhesion molecules. Among the important soluble factors are neurotransmitters (Haydon et al. 1984; Lankford et al. 1988) and growth factors (Yankner and Shooter, 1982), both of which bind to specific transmembrane receptors. The matrix component with the most dramatic effect on growth cone motility is laminin in its various isoforms (Letourneau, 1975; Rogers et al. 1983; Carbonetto et al. 1983; Tomaselli et al. 1986). Its action is mediated primarily through the integrins (Turner et al. 1989; Bozyczko and Howitz, 1986; Horwitz et al. 1985; Hynes, 1987). Other matrix components which affect neurite outgrowth are fibronectin, tenascin, and type IV collagen (Grumet et al. 1985; Chiquet-Ehrismann et al. 1986; Rogers et al. 1983; Carbonetto et al. 1983). Cellbound adhesion molecules, such as NCAM, N-cadherin, LI, TAG-1 and Fll, also are important in controlling growth cone guidance (reviewed by Jessell, 1988). They function primarily via homophilic binding to like molecules on adjacent cells. It is clear that the growth cone has the capacity to transduce a wide variety of extracellular signals into altered motility and directed growth.

In addition, there is evidence that neurons exhibit various intrinsic growth potentials (Banker and Cowen, 1979), which may reflect the regulated intracellular expression of certain genes (Skene and Willard, 1981; Benowitz and Lewis, 1983; Strittmatter and Fishman, 1991). GAP-43 is one such protein, with expression that is closely correlated with periods of active growth cone function. The levels of GAP-43 are dramatically increased within regenerating nerves after axonal injury (Benowitz and Lewis, 1983; Skene and Willard, 1981), and are higher during brain development than in the adult (Kams et al. 1987; Jacobsen et al. 1986). The expression pattern of GAP-43 within the developing mouse brain closely reflects areas of active neurite outgrowth (Biffo et al. 1990). When cultured neurons are exposed to synaptic targets, the level of GAP-43 synthesis and axonal transport falls to 20 % of control levels (Baizer and Fishman, 1987). The correspon-dence of GAP-43 expression with active neurite outgrowth suggests that it may have a role in determining or modifying neurite extension.

GAP-43 is highly concentrated at the cytosolic face of the plasma membrane, especially in the growth cone (Meiri et al. 1986; Skene et al. 1986; Van Hooff et al. 1989). There is some data that GAP-43 directly regulates cell morphology. The expression of GAP-43 in non-neuronal cells enhances their propensity to form filopodia (Zuber et al. 1989a). This suggests that GAP-43 can interact with some generalized cell machinery to alter cell shape. In coordination with other growth cone proteins, it may modulate growth cone motility. Overexpression of GAP-43 in NGF-treated PC-12 cells does increase the extent of neurite outgrowth modestly (Yankner et al. 1990), although PC-12 strains with low levels of GAP-43 can extend neurites normally (Baetge and Hammang, 1991).

What molecular interactions might underlie the function of GAP-43? The cDNA and the gene for GAP-43 have been cloned (Karns et al. 1987; Basi et al. 1987; Cimier et al. 1987; Grabczyk et al. 1990). The deduced sequence encodes a 226-amino acid protein with many charged residues and very little hydrophobicity. GAP-43 has been shown to bind to calmodulin in vitro (Alexander et al. 1988; Chapman et al. 1991). This interaction is strongest in the absence of calcium and at low ionic strength. The calmodulin binding site has been localized to amino acid residues 39-56 (Fig. 1). The region contains a serine residue at position 41, which is subject to phosphorylation by protein kinase C (Coggins and Zwiers, 1989; Chapman et al. 1991). Phosphorylation of GAP-43 at this site inhibits calmodulin binding. The level of phosphorylation has also been correlated with the development of long term potentiation in vivo (Lovinger et al. 1985). Long term potentiation is a model for learning and memory, which may require synaptic remodelling. How calmodulin binding by GAP-43 might relate to growth cone motility is not well defined, but it has been suggested that GAP-43 serves as a ‘calmodulin sink’, regulating the level of free calmodulin (Alexander et al. 1988). There are a number of studies showing the importance of calcium levels in growth cone regulation (reviewed by Kater and Mills, 1991).

Several other actions of GAP-43 have been described. Antibody to GAP-43 can alter the release of neurotransmitter from isolated, permeabilized synaptosomes (Dekker et al. 1989), and the phosphorylated protein may inhibit phosphatidylinositol 4-phosphate (PIP) kinase activity (Van Dongen et al. 1985).

We have studied how GAP-43 binds to the growth cone membrane, since this is a prerequisite for specific action at this site. GAP-43 behaves as an integral membrane protein, resisting extraction by high ionic strength, despite its lack of a domain containing hydrophobic amino acid residues (Skene and Virag, 1989; Karns et al. 1987). When expressed in non-neuronal cells, GAP-43 binds to the membrane (Zuber et al. 1989a). However, point mutations of GAP-43 at position 3 or 4 (the only two cysteines in the molecule) prevent membrane association in non-neuronal cells, demonstrating that the amino terminus is necessary for membrane association (Zuber et al. 19896, Fig. 1). The amino terminus is, in fact, sufficient for membrane binding. We showed this by expressing fusion proteins between the amino terminus of GAP-43 and chloramphenicol acetyl transferase (CAT). The first 10 amino acids of GAP-43 are sufficient to localize the normally cytosolic CAT to the membrane, even in nonneuronal cells. This fusion protein behaves as an integral membrane protein, requiring detergent for extraction from the membrane and resisting solubilization with 1 M NaCl (Fig. 2).

The amino terminus can undergo palmitoylation at positions 3 and 4 (Skene and Virag, 1989), and this may make the protein hydrophobic enough to partition into the membrane fraction. Alternatively, if the palmitates are added after membrane association, they might stabilize an interaction initiated by some other mechanism. The GAP-CAT fusion proteins also contain information for distribution to certain sites within the membrane, as demonstrated by CAT immunohistochemistry after transfection into PC-12 cells (Zuber et al. 19896). The chimeric proteins are localized to the same regions as native GAP-43, including the growth cone. This suggests that this region of GAP-43 may interact with other proteins having restricted distributions in the plasma membrane. Such growth cone ‘receptors’ for intracellular GAP-43 have not yet been identified. Recent work by Storm and colleagues has shown that although the GAP-43 amino terminus can also localize a fl-gal fusion protein to the membrane, growth cone targeting does not occur for this protein in neurons (Liu et al. 1991). This may reflect a steric problem in targeting the larger β-gal chimeric protein, or be specific to the small minority of neurons successfully transfected in these experiments, or it may be that other portions of GAP-43 are required for transport through the entire axonal length of nerve cells. The ability of the amino terminus of GAP-43 to target at least 50 % of theβ-gal to the membrane in COS cells confirms this domain is largely responsible for membrane binding (Liu et al. 1991).

In order to provide a better understanding of the molecular events which underlie the growth cone’s ability to respond to this wide variety of extracellular and intracellular signals, we sought to identify some of the major proteins in the growth cone membrane. Fortunately, a subcellular fractionation method for enriching in growth cone particles had been developed by Pfenninger et al. (1983). Analysis of growth cone membrane fractions reveals a remarkably simple protein composition (Simkowitz et al. 1989). Apart from actin and tubulin, there are two other major proteins detectable by SDS-PAGE. We employed a combination of electrophoresis, immunoblotting and partial protein sequence to identify these two proteins as the a and b subunits of the GTP-binding protein, G_o (Strittmatter et al. 1990). Immunoblots of various fractions from this subcellular fractionation demonstrated enrichment of a₀ in the growth cone membrane (Fig. 3). We also confirmed the growth cone localization by immunohistology, in both PC-12 cells and chick sympathetic neurons (Fig. 4).

What might be the function of G_o in the growth cone? G_o is a member of the large family of heterotrimeric GTP-binding proteins which link transmembrane receptors to intracellular second messenger systems (reviewed by Gilman, 1987). For example, G_o in the growth cone might respond to the neurotransmitter receptors for serotonin and dopamine, which are known to regulate potently growth cone motilty in certain neurons (Haydon et al. 1984; Lankford et al. 1988). Furthermore, the ability of pertussis toxin to abolish the effect of antibodies to LI and NCAM on intracellular second messengers suggests that a pertussis toxin-sensitive G protein, possibly G_o, may be important in the action of the cell adhesion molecules (Schuch et al. 1989).

G_o and GAP-43 have indistinguishable distributions at the light microscopic level in NGF-treated PC-12 cells. Both are highly concentrated in the tips of many neuritic processes (Fig. 5), whereas other processes appear to lack both (not shown). This high degree of concordance between G_o and GAP-43 immunoreactivity led us to explore the possibility that GAP-43 might interact with G_o.

The activation of G_o can be monitored by its guanine nucleotide-binding properties (Gilman 1987). In the basal state, G_o contains tightly bound GDP. When activated by receptor-ligand complexes, the rate of GDP release is increased, and GTP then binds. The GTP-bound α subunit activates appropriate second messenger systems. GTP hydrolysis terminates activation. GAP-43 alters the guanine nucleotide-binding characteristics of G_o, by increasing its level of GTPγS binding (Fig. 6, Strittmatter et al. 1990). The increase in GTPγS binding cannot be explained by changes in G_o thermal stability, as GAP-43 has no effect on this parameter (not shown). This effect is saturable, with an EC₅₀ of about 300 nM GAP-43 (Fig. 6). The concentrations of G_o and GAP-43 in vivo probably exceed this level.

GAP-43 is the first potential intracellular regulator of G proteins. Therefore, we compared the mechanism of G_o activation by GAP-43 and receptors (Table 1). GAP-43 is clearly a guanine nucleotide release protein (GNRP), like receptors. GAP-43 nearly doubles the rate of release of bound GDP. Since GDP release is the rate limiting step in GTP binding, GAP-43 also doubles the initial rate of GTPγS binding to G_o and the steady state GTPase activity of G_o. Also like receptors, the intrinsic k_cat of G_o for the hydrolysis of bound GTP is not altered by GAP-43. Since, in all of these regards, the GAP-43 effect resembles that of the G protein-linked receptors (Gilman 1987), it is clear that GAP-43 is a GNRP, and that it does not have activity as a GTPase-activating protein (GAP).

In several other respects, however, GAP-43 action on G_o can be distinguished from that of receptors (Table 1). Pertussis toxin treatment of G_o prevents its regulation by receptors, but not by GAP-43. GAP-43 can double the rate of GTPγS binding and GTPase of isolated recombinant α_o, while receptors require β γ in addition to α subunits for productive interactions. GAP-43 is equally effective at stimulating detergent soluble- and vesicle-incorporated G_o, whereas receptors stimulate G proteins only when the two are incorporated into phospholipid vesicles. Thus, by these mechanistic studies GAP-43 is a GNRP, but is clearly distinguishable from those previously studied.

We have begun to define the domains of GAP-43 that contribute to its stimulation of G_o. The amino terminus of GAP-43 is known to be critical for association of GAP-43 with the plasma membrane and for growth cone localization. Residues 39-56 include a calmodulin-binding site, and there is a phosphorylation site for protein kinase C at serine 41 (Fig. 1, Alexander et al. 1988; Chapman et al. 1991; Coggins and Zwiers, 1989). A synthetic peptide corresponding to the calmodulin binding site has no effect on G_o activity (Strittmatter et al. 1990), and 40 "M calmodulin does not alter the GAP-43/G₀ interaction (not shown). This suggests that the calmodulin-binding domain of GAP-43 is not important for G protein interaction.

In contrast, a peptide composed of the first 25 amino acids of GAP-43 does stimulate GTPγS binding to G_o (Strittmatter et al. 1990). Although the amino terminal peptide requires a twenty to hundred-fold higher concentration than does intact GAP-43 for G_o activation, the effect on G_o is two-fold and saturable, as it is for the protein. These findings demonstrate the importance of this domain in G protein interaction.

We compared this domain of GAP-43 with the portions of G protein-linked receptors necessary for coupling (Strittmatter et al. 1990). Despite the fact that the two proteins have disparate overall structures, a small region of homology exists between GAP-43 and receptors. This stretch, at the GAP-43 amino terminus, contains the sequence nonpolar-nonpolar-cys-cys-x-basic-basic, with the cysteines subject to palmitoylation (Skene and Virag, 1989, O’Dowd et al. 1989).

The evidence that G_o is a major component of the growth cone membrane suggests a critical role for this signaltransducing protein in coordinating growth cone function. Fig. 7 is a model incorporating notions from several investigators as to how the growth cone might be regulated, one that places heavy emphasis upon G_o, GAP-43 and the calcium ion. Although the signals are not known with certainty, growth cone G_o might respond to the neurotransmitter receptors and possibly to some celladhesion molecules. The potential role of G_o in transducing a number of other extracellular signals has not been investigated. There is evidence in other systems that G_o can link receptors to several effector systems: Ca²⁺ channels (Holz et al. 1986; Heschler et al. 1987), phospholipase C (Moriarty et al. 1990), K⁺ channels (Von Dongen et al. 1988) and phospholipase A₂ (Bloch et al. 1989). Since the first three of these effectors can alter intracellular Ca²⁺ levels, and since the level of Ca²⁺ is a major regulator of growth cone activity (Kater and Mills, 1991), many G₀-linked signals may regulate growth cone dynamics via the Ca²⁺ ion. Alternatively, G_o activation of phospholipase C may lead to diacylglycerol-mediated protein kinase C phosphorylation of various substrates in the growth cone. Protein kinase C activation has been implicated in laminin stimulation of neurite outgrowth (Bixby, 1989). Whether the final mediator in these pathways is the cytoskeleton (Yamada and Wessells, 1973; Smith, 1988; Bamburg et al. 1986), membrane cycling (Lockerbie et al. 1991; Cheng and Reese, 1987) or some other mechanism remains to be seen.

G_o responds to both transmembrane receptors (Gilman, 1987) and the intracellular protein GAP-43 (Strittmatter et al. 1990). GAP-43 has GNRP activity as do receptors, and the two proteins share a small region of sequence homology. However, the mechanism of action of GAP-43 can be distinguished from that of receptors in several ways. This defines GAP-43 to be in a new class of GNRP. The differences between GAP-43 and receptors suggest that GAP-43 interacts with different or additional region of α_o. GAP-43 also appears to affect a different domain than do β γ subunits because there is no competition or synergism between GAP-43 and β γ. This raises the possibility that GAP-43 might interact directly with either the guanine nucleotide-binding domain or the effector domain. If GAP-43 acted at the effector domain, its primary role might be as downstream responder to activated G_o rather than a positive upstream regulator of G₀-like receptors. However, since it enhances GDP release, and therefore interacts with the GDP-bound form, it is more likely an upstream positive regulator of G_o.

The possibility that G_o might respond to both trans-membrane receptors and GAP-43 places it at a central position in the growth cone control apparatus (Fig. 7). Extracellular signals could bind to transmembrane receptors and then be transduced into second messanger changes by G_o. In addition, G_o might respond to intracellular gene programs for growth, as signaled by increased GAP-43 synthesis. Thus both extracellular and intracellular signals to the growth cone would converge on, and be integrated by, G_o. Whether such G_o activation enhances or retards neurite extension is under study.