Nerve Growth Factor in Neuronal Development and Maintenance

The discovery of the flow of nerve growth factor (NGF) from the target tissue of an appropriate neurone to the neuronal cell body provides the framework for an understanding of how the action of N G F leads to the selective survival of a fraction of a group of nerve cells at a critical time in development. The influence of the target on this process of naturally occurring neuronal cell death has been known for a long time, stemming from the findings that neuronal cell death is enhanced by removal of target tissue but reduced by provision of additional target tissue (reviewed in Oppenheim, 1981). That this influence is mediated in targets of sympathetic and some sensory neurones by a trophic substance became apparent in the experiments which led to the discovery of NGF (Levi-Montalcini & Hamburger, 1951, 1953). A particularly striking demonstration of the role of NGF in controlling neuronal survival came from the experiments of Levi-Montalcini & Booker (1960a), who found that injections of NGF antiserum in neonatal mice produced massive neuronal death in sympathetic ganglia. Different methods of delivering anti-NGF antibodies extended these observations to dorsal root ganglion sensory neurones (Gorin & Johnson, 1979, 1980; Johnson, Gorin, Brandeis & Pearson, 1980), while all the antibody experiments emphasized that the antibody-induced neuronal cell death occurred at a specific time in the development of the particular neurones. Furthermore, considerable evidence exists to show that neuronal cell death is a result of depriving the animals of endogenous NGF rather than of a complement-mediated cytotoxicity (Goedert et al. 1980). The in vitro experiments of Campenot (1977) using multi-compartment chambers also demonstrated that NGF interacting with the nerve terminals of sympathetic neurones is sufficient to support neuronal survival. All these data suggest that sensory and sympathetic neurones compete for limiting quantities of NGF when their axons reach their peripheral targets and that the supply of NGF determines the extent of neuronal cell death. A consequence of this hypothesis is that neuronal cell death should be reduced by providing NGF over and above that derived from the target tissue, and this has been found to be true. Again, some of the earlier work showed that NGF administration to neonatal animals increased the size of sympathetic and sensory ganglia as a result of both increased neuronal cell numbers and size (Levi-Montalcini & Angeletti, 1968; Levi-Montalcini & Booker, 1960b; Hendry, 1976; Hendry & Campbell, 1976), processes which correlated well with the timing of neuronal cell death (Cowan, 1975; Carr & Simpson, 1978). More recently, the timing and extent of neuronal cell death of VL and DM sensory neurones has been determined and exogenous NGF has been shown to rescue a significant fraction of VL and essentially all DM neurones (Hamburger, Brunso-Bechtold & Yip, 1981). Clearly NGF acts as an endogenous trophic factor for perhaps all sensory neurones, although it is present in their targets in amounts insufficient to sustain all the neurones during development. It is reasonable to assume that the same conclusion applies to sympathetic neurones.

Besides this key role in development, NGF continues to play an important role in the maintenance of the differentiated state of mature sympathetic and sensory neurones. This is demonstrated by the reduction in protein content, norepinephrine content and tyrosine hydroxylase activity in sympathetic ganglia after administration of anti-NGF antibodies to adult rodents (Goedert, Otten & Thoenen, 1978; Otten, Goedert, Schwab & Thibault, 1979; Ebendal, Olsen, Seiger & Hedlund, 1980) as well as the reduction of substance P levels in the sensory neurones of these animals (Otten, Goedert, Mayer & Lembeck, 1980; Ross et al. 1981; Mayer, Lembeck, Goedert & Otten, 1982; Schwartz, Pearson Sc Johnson, 1982; Kessler, Bell Sc Black, 1983). Correspondingly, increases in neuronal cell volume and tyrosine hydroxylase activity in the superior cervical ganglion are observed when NGF is injected into the targets (eye or submaxillary gland) of these sympathetic neurones (Paravicini, Stoeckel & Thoenen, 1975; Hendry, 1977). NGF also partially obviates the decrease in substance P and extralysosomal acid phosphatase seen in dorsal root ganglia after peripheral axotomy (Fitzgerald, Wall, Goedert & Emson, 1985). Since NGF is found in adult rat sciatic nerve in culture (Richardson & Ebendal, 1982) and is produced by non-neuronal cells at the site of and distal to an injury of the same nerve in vivo (Heumann, Korsching, Bandtlow & Thoenen, 1987), it is possible that NGF acts to promote sympathetic and sensory axon regrowth in vivo as it does in vitro (Levi-Montalcini, Meyer & Hamburger, 1954).

Underlying the actions of NGF is its ability to convey information from the peripheral targets to the neuronal cell body by retrograde flow (Fig. 1). The phenomenon was first demonstrated by injecting [¹²⁵I]NGF unilaterally into the anterior chamber of the eye of a mouse and observing accumulation of radioactivity, subsequently shown to be in the form of intact, biologically active [¹²⁵I]NGF, in the superior cervical ganglion on the side of the injection (Dumas, Schwab & Thoenen, 1979; Johnson, Andres & Bradshaw, 1978). Transection of postganglionic nerves abolishes the preferential labelling, showing that transport occurs in axons. This has also been confirmed by following retrograde transport of [^l25I] NGF in sympathetic neurites in culture in compartmentalized chambers (Claude, Hawrot, Dunis & Campenot, 1982). More recently, the retrograde flow of endogenous NGF in the rat sciatic nerve has been demonstrated by its accumulation on the distal, but not the proximal, side of a crush (Korsching & Thoenen, 1983b). This experiment and measurement of both NGF and NGF mRNA in a typical sympathetic target (iris) provide the most convincing proof that endogenous NGF is produced in the target, sequestered by nerve terminals and carried retrogradely to the neuronal cell bodies (Korsching & Thoenen, 1983a; Heumann, Korsching, Scott & Thoenen, 1984; Shelton & Reichardt, 1984). The actual flow of NGF occurs in membrane-bound vesicles along microtubules (Schnapp, Vale, Sheetz & Reese, 1985), at a rate similar to the fast axonal flow of other molecules (Grafstein & Forman, 1980). Interruption of the retrograde flow of NGF leads to degeneration of developing - and chromatolysis and loss of differentiated functions in mature — sympathetic and sensory neurones. This may be achieved by severing the axon, disintegrating the microtubules with chemical agents such as vinblastine or colchicine (Purves, 1976; Chen, Chen, Calissano & Levi-Montalcini, 1977; Johnson, 1978) or by destroying the nerve terminals with 6-hydroxydopamine (Levi-Montalcini et al. 1975). The effects of such treatments are similar to the changes seen after administration of anti-NGF antibodies has been used to deplete supplies of NGF. The retrograde transport of NGF is independent of neuronal activity but depends instead on a specific interaction with NGF receptors at the nerve terminal (Dumas et al. 1979; Schwab, Heumann & Thoenen, 1982) (Fig. 2). The NGF is then internalized by receptor-mediated endocytosis, a process characterized in some detail in the NGF-responsive pheochromocytoma cell line, PC12 (Layer & Shooter, 1983). After transport of the vesicles (endosomes) along microtubules to the cell body, the endosomes are acidified by insertion of proton pumps into their membranes, causing the NGF to dissociate from the NGF receptor at the low pH (Vale & Shooter, 1984). The NGF, now in the fluid phase of the endosome, is translocated finally to lysosomal structures and degraded (Schwab, 1977; Schwab & Thoenen, 1977; Claude et al. 1982). The fate of the NGF receptor is unknown, although if it behaves in an analogous manner to other receptors which efficiently internalize ligands, it will be recycled to the nerve terminal for further use.

As NGF proceeds from target tissue to neuronal cell body along the retrograde pathway it generates intracellular signals which mediate its various functions at the molecular level. However, the exact location at which this signal or signals are produced is not yet known. The possibilities are (1) at the receptor, (2) after internalization in the nerve terminal, (3) during retrograde flow or (4) in the cell body, or various combinations of these. Certain possibilities have been eliminated. The degradation of NGF in the lysosomes appears not to play a role in nerve growth factor action since inhibition of lysosomal function by chloroquine has no effect on NGF-induced neurite outgrowth (Shooter, Yankner, Landreth & Sutter, 1981) or on choline acetyltransferase activity in PC12 cells (Heumann, Schwab, Merkl & Thoenen, 1984). Similarly anti-NGF antibodies introduced into the cytoplasm of PC12 cells have no effect on the ability of these cells to respond to NGF, indicating that immune-precipitable fragments of NGF produced by lysosomal degradation are not involved (Heumann, Schwab & Thoenen, 1981; Seeley, Keith, Shelanski & Greene, 1983). Furthermore, since neurite outgrowth from PC12 cells is not induced by introducing NGF into the cells cytoplasm, any signals generated by NGF with flow the cell must come from NGF within vesicles. NGF enters these vesicles by interacting with the specific NGF receptors on the nerve terminal. Considerable effort has gone into characterizing these interactions and the receptors themselves since it is at this level that intracellular signals may first be generated.

All the NGF-responsive cells display two classes of NGF receptors, one of high affinity and low capacity and the other of low affinity and high capacity (Vale & Shooter, 1984) (Table 1). Both types of receptor are observed when NGF binding experiments are carried out at low, as well as at higher, temperatures, indicating the cell surface localization of both the receptors (Sutter, Riopelle, Harris-Warrick & Shooter, 1979a). The high-affinity receptors show a relatively low rate of dissociation, and are referred to as slow receptors or SNGFRs, while the dissociation rate of the low-affinity receptors is approx. 100-fold faster, leading to their identification as fast receptors or FNGFRs. Pharmacological dose-response curves indicate that the SNGFRs mediate long-term NGF responses such as neurite outgrowth (Sutter, Riopelle, Harris-Warrick & Shooter, 1979b). High-affinity SNGFR occupancy also correlates with survival and neurite outgrowth of sensory neurones (Zimmermann, Sutter, Samuelson & Shooter, 1978). In contrast, the low-affinity FNGFRs mediate at least some of the rapid responses to NGF, such as stimulation of amino acid uptake (McGuire & Greene, 1979; Kedes, Gunning & Shooter, 1982). This receptor heterogeneity can be explained by one or more models. For example, the two receptor types could be independent receptor molecules derived from different genes or by differential splicing of mRNA or protein precursors. Alternatively, the SNGFRs could be generated from the FNGFRs by association with another protein, by crosslinking with the divalent ligand NGF or by a change in conformation. These last three hypotheses imply that NGF binds first to the low-affinity FNGFRs thereby converting them to high-affinity SNGFRs. Current evidence favours the idea that the FNGFR is the common NGF binding subunit of the two receptor types with a second protein associating with the FNGFR to create the SNGFR.

The key pieces of evidence for this conclusion are the relative molecular masses (M_r) of the two receptor species (determined by crosslinking experiments), the structure of the FNGFR (determined by gene transfer and cloning) and the demonstration of possible receptor conversion by agents which cluster the receptors.

Several experiments suggest that rapidly (FNGFR) and slowly (SNGFR) dissociating receptors are interconvertible. Addition of the plant lectin wheat germ agglutinin (WGA) alters the ratio of the two receptor types on PC12 cells without affecting total binding (Vale & Shooter, 1982; Grob & Bothwell, 1983). Specifically, WGA converts a major proportion of the FNGFRs which are trypsin-sensitive to trypsin-resistant, slowly dissociating receptors. Whether these receptors are exactly the same as the naturally occurring trypsin-resistant SNGFRs is unclear. However, the WGA-induced change in the dissociation characteristics of the receptor also correlates with the conversion of the receptor to a form insoluble in Triton X-100: another characteristic of the SNGFRs (Vale & Shooter, 1982). It seems likely that WGA acts by crossbridging the FNGFR to one or more proteins in the cell membrane which themselves are anchored to the cytoskeleton (Vale, Ignatius & Shooter, 1985). That receptor clustering may be involved in the conversion is also shown by the finding that anti-NGF antibodies, but not monovalent Fc fragments, convert occupied FNGFRs to a slowly dissociating, trypsin-resistant form insoluble in Triton X-100 (Vale & Shooter, 1983). Since NGF is potentially divalent with respect to receptor binding, receptor clustering may occur simply through the action of NGF. However, the differences in the relative molecular masses of the two receptors and the finding that PC12 plasma membranes display only FNGFRs (Block & Bothwell, 1983) suggest that another component in the intact cell, possibly a cytoskeletal-linked protein, may be responsible for receptor conversion.

Following earlier work (Massague et al. 1981), which showed that rabbit superior cervical ganglion membranes had two differently sized NGF receptor species, Hosang & Shooter (1985) used the same hetero-bifunctional, photoactivated crosslinking agent to demonstrate that on PC12 cells [^l25I]NGF also crosslinked into two major complexes with relative molecular masses of 158 000 and 100000, respectively. Assuming that only one chain of NGF is covalently crosslinked in these complexes, it follows that the actual values for the receptors are approx. 85 000 and 145 000, respectively. Moreover, on the basis of its low rate of dissociation and resistance to trypsin (binding of [¹²⁵I]NGF is unaffected but the receptor loses a peptide with an M_r of 10000), the larger receptor was identified as the SNGFR. Similarly, since the species with the lower M_r dissociated rapidly and was degraded by trypsin it has the characteristics of the FNGFR. Treatment of both receptors with neuraminidase removes terminal sialic acid residues and lowers the apparent relative molecular masses of both receptors without significantly affecting their capacity to bind NGF (Vale, Hosang & Shooter, 1985).

The relative molecular mass of the purified low-affinity human NGF receptor from the A875 melanoma cell is 85 000 (Puma et al. 1983) and a similar figure was obtained for the biosynthetically labelled receptor in these cells (Ross et al. 1984). Taken together with the results obtained from the cloning of the rat low-affinity FNGFR (see later), it is clear that the low-affinity NGF receptor from both species is a single, glycosylated peptide chain with a relative molecular mass of approx. 85 000.

The crosslinking technique can be used to follow the internalization and intracellular fate of the [^l25I]NGF-receptor complexes in PC12 cells. When the latter are incubated for relatively long periods with [¹²⁵I]NGF prior to crosslinking, a progressive decrease in the labelling of the 4/_r 158000, but not the M_r 100000, complex is observed, suggesting internalization only of the former (Hosang & Shooter, 1987). In keeping with this idea is the further finding that the lower relative molecular mass receptor remains trypsin-sensitive during the prolonged incubation, indicating a cell surface localization, while the larger receptor is protected from the characteristic tryptic release of the small peptide, indicating that it is internalized. It is of considerable interest that the receptor which is internalized in PC12 cells, the SNGFR, is also the one implicated in mediating NGF-induced neurite outgrowth. The internalized, crosslinked SNGFR remains in the endosomes of the PC12 cell where, apparently, the inability of the low pH environment to bring about the usual dissociation of the (crosslinked) [^l25I]NGF from its receptor (Vale & Shooter, 1984) inhibits both transfer of the [^l25I]NGF to the lysosomes and the potential recycling of the receptor.

The identification of the initial intracellular signal generated by the binding of NGF to its receptors requires information on the molecular structures of the receptors. Since no information was available for the amino acid sequences of either receptor, the technique of gene transfer was used (i.e. the transfer of the gene for the rat NGF receptor to a cell line from another species, mouse). The success of this approach depends on the availability of a monoclonal antibody reacting with one or other of the rat, but not the mouse, NGF receptors. One such antibody exists that enhances NGF binding on rat PC12 cells by specifically increasing the affinity of the low-affinity FNGFR (Chandler, Parsons, Hosang & Shooter, 1984). This antibody was used in conjunction with a second fluorescein-conjugated goat anti-mouse antibody to detect, and isolate in a cell sorter, cells expressing the rat FNGFR. Transfection of mouse LTK⁻ cells with 100 to 200-kb pieces of PC12 genomic DNA and a plasmid containing the chicken thymidine kinase gene was followed by selection of cells in a medium which allows only for the survival of cells integrating and expressing the thymidine mouse genes (Radeke et al. 1987). These cells also integrate and express one of the PC 12 genomic fragments and it is from this population that the few cells expressing the rat FNGFR gene were selected with the anti-receptor antibody as described above. In one of the transfectants obtained in this manner the phenotype was unstable, resulting in a continuous amplification of the expression of the FNGFR during growth in the selective medium after cell sorting. The phenotype finally stabilized after 10 rounds of division, giving a transfected cell line (PCNA10) expressing approx. 2·5×10⁶ FNGFRs per cell. All the receptors expressed were rapidly dissociating with an equilibrium dissociation constant the same as that of the low-affinity NGF receptors. Their relative molecular mass was 83 000, again consistent with the size of the FNGFR. It is of interest that the expression of the rat FNGFR gene in the mouse L cell gives rise only to the FNGFR and not to SNGFRs.

The rat FNGFR cDNA was recovered by subtractive hybridization of the transfectant cell single-stranded cDNA with polyA⁺ RNA from the recipient L cell (Fig. 3), and the use of this FNGFR-enriched cDNA pool to screen a double stranded transfectant cell cDNA library. Of the 19 clones which hybridized very strongly with this probe, six cross-hybridized and the longest insert was approx. 3·4kb. The latter hybridized with a common message from all cell types expressing NGF receptors but not from cells lacking the receptors and, when transfected in an appropriate expression vector into mouse L cells, gave rise to the expression of FNGFR. The nucleotide sequence of this FNGFR cDNA showed the start of an open reading frame close to the 5′-end and continuing for 1275 bases, resulting in a receptor precursor containing 425 amino acids (Fig. 4). The FNGFR receptor itself, with 396 residues, has a relative molecular mass of 42478, indicating that it is heavily glycosylated to bring its size up to that of the mature membrane-bound receptor (M_r83 000) indicated by electrophoretic analysis. It has one membrane-spanning segment separating an extracellular domain containing four cysteine-rich repeating units, characteristic of protein- or peptide-binding receptors (Fig. 5), from an intracellular domain rich in serine and threonine, but lacking an ATP-binding site characteristic of an endogenous kinase. The FNGFR shows no significant homologies with any of the other known receptors (or indeed any other protein), in spite of the extracellular repeating units, and by itself offers no clue to the receptor-mediated transduction signal. The structure of the human FNGFR on melanoma cells is, in contrast, highly homologous to that of the rat FNGFR (Johnson et al. 1986).

The fact that the FNGFR cDNA hybridizes to a single mRNA species in cells which express both types of NGF receptors suggests that the two receptors either have no relationship to one another, except for the property of binding NGF, or that they share the FNGFR protein as a common subunit. It follows, if the latter explanation is correct, that the high-affinity SNGFR will contain a second cellular protein, with a relative molecular mass of approx. 60000, which is the key to the signal transduction mechanism. As anticipated from the demonstration that only the high-affinity SNGFR is internalized after NGF binding to PC12 cells, the low-affinity FNGFR expressed in mouse L cells does not internalize NGF, suggesting that the ability to internalize is also dependent on the interaction of the FNGFR with another protein. One of the possible candidates for this second protein is the product of the proto-oncogene c-src, a 60 000.M_r tyrosine-specific kinase. One of the advantages of the gene transfer technique is that it can be used to identify the gene coding for this putative second protein in the SNGFR by its ability to convert some or all of the FNGFR in the mouse L cell transfectant to the high-affinity SNGFR. These and similar studies using, for example, the antisense mRNA for the FNGFR to suppress FNGFR expression in PC12 cells, should finally permit the relationship of the two receptor types to be firmly established.

This work was supported by grants from NIH (NS04270, NS07638), NIMH (MH 17047), the American Cancer Society and by the Isabelle M. Niemela Trust.