Control of microtubule assembly in growth cones

The growth cone’s remarkable ability to find its way through the complex terrain of the developing nervous system and successfully locate its synaptic partner has been known since the time of its discovery by Ramón y Cajal, just over one hundred years ago (Ramón y Cajal, 1890). Growth cone navigation depends on a system of plasma membrane receptors for extrinsic guidance cues which are coupled to a cytoskeleton capable of motility (Dodd and Jessell, 1988). In addition to pathfinding activities and synaptogenesis, growth cones are also involved in growth of the neurite (Gordon-Weeks, 1989, 1991). This involves the assembly of the cytoskeleton of the neurite and the addition of new surface and internal membrane systems and other elements of the cytoplasm.

In a growing neurite, as in most mature axons and dendrites, microtubules are a dominant feature and are often observed in bundles running parallel to the longitudinal axis of the neurite. Two functions of microtubules within neurites have been identified: a specific one, the substrate for fast axonal (and dendritic) transport, and a more general one, the maintenance of the shape and structural integrity of the neurite. At the distal end of a growing neurite, where it enlarges into the growth cone, the individual microtubules of the bundle diverge, like the ribs of a fan, and extend into the central domain (C-domain) of the growth cone (Fig. 1). Microtubules only rarely extend beyond the C-domain and into the peripheral region (P-domain) of the growth cone, where the cytoskeleton is mainly composed of microfilaments, and have never been found in the filopodia (Gordon-Weeks, 1988). As they diverge from each other on entering the C-domain, microtubules may have a winding course at their distal ends, whereas in the neurite they are straight and presumably bundled by microtubule-associated proteins (MAPs) (Cheng and Reese, 1985). Furthermore, there is a marked preference for smooth endoplasmic reticulum and vesicles to be associated with the straight regions but not the winding regions (Cheng and Reese,1985).

The microtubules that enter the growth cone from the neurite shaft are all oriented with their so-called ‘plus’ ends, the end at which tubulin subunit addition takes place preferentially, located distally (Baas et al. 1987) (Fig. 1). We now know from recent work that microtubule elongation takes place in the growth cone by the addition of tubulin onto the distal ends of the neurite microtubules that enter the C-domain of the growth cone. It has been known for some time that neurite advance is critically dependent on microtubule assembly, since neurite extension in cultured neurones can be inhibited or reversed by the addition of agents that interfere with microtubule assembly/disassembly (Yamada et al. 1970; Daniels, 1972). More recent experiments have shown that it is the assembly of microtubules in the growth cone, rather than the cell body or neurite, that is important for neurite extension. The evidence for this derives from a number of observations. First, as discussed above, only the ‘plus’ ends of the neurite microtubules extend into the growth cone. Second, the direct application onto the growth cone by micropipette of agents that depolymerise or stabilise microtubules prevents neurite growth (Bamburg et al.1986). Furthermore, the growth cone is the most sensitive region of the neurone to these agents, by orders of magnitude (Bamburg et al. 1986). Third, there is a large, assembly competent, pool of tubulin in the growth cone that is the immediate source of tubulin for microtubule elongation (Letourneau and Ressler, 1984; Gordon-Weeks, 1987). This was first shown by observing the effects of the agent taxol on growth cones in culture (Letourneau and Ressler, 1984). Taxol lowers the critical concentration point for tubulin assembly within cells and thus forces the soluble pool of tubulin to form microtubules. The effect of taxol treatment on growth cones is to assemble the soluble pool of tubulin onto the ‘plus’ ends of the microtubules that enter the C-domain from the neurite shaft (Letourneau and Ressler, 1984; Gordon-Weeks, 1987; Gordon-Weeks et al. 1989). At high concentrations of taxol (low,MM), when this effect goes to completion, microtubule loops appear in the C-domain of the growth cone because of the large volume of the soluble tubulin pool. This effect has also been seen in growth cones isolated from developing brain as a subcellular fraction (Gordon-Weeks, 1987; see review by Lockerbie, 1990). Normally, microtubules are absent from these growth cones, for reasons that are not clear, but they appear on incubation in taxol because of the presence of the large pool of soluble tubulin (Gordon-Weeks, 1987).

The soluble pool of tubulin in the growth cone can be visualised in cultured neurones using antibodies to the carboxy (C)-terminal of n-tubulin such as YL 1/2 (Gordon-Weeks et al. 1989; Mansfield and Gordon-Weeks, 1991). YL 1/2 labels both the microtubules and the soluble tubulin which, unlike the microtubules, is distributed throughout the P- and C-domains, including the filopodia. As

described above, taxol treatment accelerates the assembly of the soluble pool of tubulin onto the ends of the microtubules in the growth cone forming microtubule loops in the C-domain. This process exhausts the pool of soluble tubulin and consequently results in a loss of the staining of the P-domain and filopodia with tubulin antibodies (Gordon-Weeks et al. 1989). If neuronal cultures are fixed with fixatives containing detergent, the soluble pool of tubulin is removed leaving behind the cytoskeleton including the microtubules and actin filament bundles in the filopodia (P. R. Gordon-Weeks, unpublished observations). Under these circumstances, only the microtubules are seen in growth cones after immunofluorescence staining with tubulin antibodies. Occasionally, in these cytoskeletal preparations, individual microtubules can be seen to extend across the P-domain and run alongside the actin filament bundles of filopodia (Fig. 2). These observations suggest that there is a functional link between growing microtubules and filopodia in growth cones (Fig. 1).

These studies provide a view of microtubule assembly in extending neurites in which microtubules are formed and elongated in the growth cone and deposited in the neurite as the growth cone advances. The immediate source of the tubulin to support microtubule formation is the large pool of soluble tubulin in the growth cone. Microtubules are clearly necessary in the neurite both to provide structural support and for the substrate of fast axonal (and dendritic) transport. Microtubules in the neurite shaft are stationary (Okabe and Hirokawa, 1990; Lim et al. 1990), and less dynamic than in the growth cone (Lim et al. 1990). Unlike the neurite shaft, however, the growth cone is a highly motile structure and it seems that when the numbers of microtubules are artifactually increased, by the action of an agent such as taxol, motility is compromised (Letourneau and Ressler, 1984; Mansfield and Gordon-Weeks, 1991). This observation and the fact that the large pool of tubulin in the growth cone is assembly competent and yet remains unassembled (Letourneau and Ressler, 1984; Gordon-Weeks, 1987; Gordon-Weeks et al. 1989), despite the presence of the ‘plus’ ends of microtubules (Baas et al. 1987), implies that microtubule assembly in the growth cone is precisely controlled (Gordon-Weeks, 1991).

What factors might be involved in influencing the assembly of microtubules in growth cones and the crosslinking of microtubules into bundles in the neurite shaft? To approach these questions we have been investigating the types of post-translational modifications of tubulin in growth cones and the presence and distribution of MAPs.

Several post-translational modifications of tubulin have been identified including phosphorylation, acetylation and a reversible removal of the C-terminal tyrosine of n-tubulin by specific enzymes (Serrano and Avila, 1990). Although the function of these modifications has not been established, it is possible that they may affect microtubule assembly either directly or through modifying the binding of MAPs-which may themselves alter microtubule assembly.

With a few minor exceptions, most of the genes for α-tubulin code for a protein with a carboxy-terminal tyrosine (Serrano and Avila, 1990). This tyrosine can be selectively removed by a specific tubulin tyrosine carboxypeptidase (Argarana et al. 1978) or added by a specific tubulin tyrosine ligase (Barra et al. 1973; Raybin and Flavin, 1977). These post-translational modifications are unique to tubulin. Most of the a-tubulin in growth cones is carboxy-terminal tyrosinated, as indicated by biochemical experiments with isolated growth cones (Gordon-Weeks and Lang, 1988; Gordon-Weeks et al. 1989). Furthermore, immunofluorescence studies of cultured neurones using antibodies specific for either tyrosinated or de-tyrosinated α-tubulin support this finding (Lim et al. 1989; Robson and Burgoyne, 1989; Mansfield and Gordon-Weeks, 1990, 1991). For instance, in dorsal root ganglion cells in culture, the majority of the axonal growth cones (these cells do not have dendrites) stain for antibodies specific for tyrosinated n-tubulin but not for de-tyrosinated n-tubulin (Robson and Burgoyne, 1989). This is also the case with PC12 growth cones (Lim et al. 1989), and the axonal and dendritic growth cones of cerebral cortical neurones in culture (Mansfield and Gordon-Weeks, 1990, 1991).

The function of these post-translational modifications of tubulin are not known but it has been found that n-tubulin becomes de-tyrosinated and acetylated sometime after assembly in non-neuronal cells (Schulze et al. 1987). A similar event occurs in neurones and neuroblastoma cells extending neurites in culture (Lim et al. 1989; Robson and Burgoyne, 1989; Mansfield and Gordon-Weeks, 1990, 1991). When the growth cone has moved on and the microtubules have become incorporated into the neurite cytoskeleton the "-tubulin is de-tyrosinated at its carboxyterminal and acetylated (Lim et al. 1989; Robson and Burgoyne, 1989; Mansfield and Gordon-Weeks, 1990, 1991). These post-translational modifications of tubulin correlate with an increase in the stability of the microtubules to depolymerization by cold shock and microtubule depolymerizing agents such as nocodazole, but they are not causal to microtubule stability (Schulze et al. 1987). The events that stabilise microtubules to depolymerisation are not known. Once they have become incorporated into the neurite cytoskeleton, although immobile, microtubules can still exchange tubulin subunits at their ‘plus’ ends (Okabe and Hirokawa, 1988) and here also the added a-tubulin is initially tyrosinated at its carboxy-terminal and unacetylated (Baas and Black, 1990).

It is likely that microtubules in growth cones are dynamically unstable, that is to say they are alternately growing slowly and shrinking back rapidly (Mitchison and Kirschner, 1988). Although dynamic instability has been directly observed in some cell types (Schulze and Kirschner, 1988), it has not been looked for systematically in growth cones. However, in a recent study of Aplysia neurones observed in culture with video enhanced microscopy, microtubules showing dynamic instability were observed in growth cones, although the authors were not able to rule out forward sliding of the microtubules (Forscher and Smith, 1988). The proposal that growth cone microtubules are undergoing dynamic instability is also consistent with the observation that these microtubules are turning-over more rapidly than in other regions of the growing neurone (Lim et al. 1989), and that these microtubules are largely composed of tyrosinated a--tubulin (see above) which correlates with microtubule instability. Recently we have found that the distal regions of microtubules in the growth cones of neurones growing in culture, but not the microtubules in the neurite shaft, depolymerise in microtubule assembly buffers containing detergents, a finding which also suggests that these microtubules are relatively labile (P. R. Gordon-Weeks, unpublished observations).

Tyrosinated tv-tubulin is no less able to polymerise than de-tyrosinated a-tubulin. However, if the carboxy-terminal tyrosine is phosphorylated then assembly is markedly impaired (Wandosell et al. 1987). Even if assembly occurs, the binding of MAPs may be altered by phosphorylation and this in turn may lead to less stable microtubules (see below). Experiments with isolated growth cones have shown that tubulin can be phosphorylated on tyrosine residues in the growth cone but the location of the tyrosine within the molecule is not known (Cheng and Sahyoun, 1988; Lockerbie et al. 1989). The tyrosine kinase pp60^c-src is present in an active form in growth cones (Maness et al. 1988) and phosphorylates tubulin in them, although probably not at the C-terminal tyrosine (Matten et al. 1990).

MAPs have been shown to affect profoundly microtubule dynamics both in vitro and in vivo and they are candidates, therefore, for controlling microtubule assembly in growth cones. There is considerable circumstantial evidence that MAP IB may play an important role in neurite outgrowth. MAP IB, also known as MAP 5 (Riederer et al. 1986), MAP 1.2 (Aletta et al. 1988) and MAP lx (Calvert and Anderton, 1985, is present in growth cones as judged by immunocytochemical and biochemical data (Mansfield et al. 1991). Furthermore, when PC 12 cells are induced to form neurites in culture by the action of nerve growth factor, MAP IB is rapidly up-regulated, suggesting that it is required for neurite outgrowth (Greene et al. 1983; Drubin et al. 1985; Brugg and Matus, 1988). MAP IB is post-translationally phosphorylated by a casein kinase Il-like activity (Diaz-Nido et al. 1988) and there is strong developmental downregulation of the phosphorylated form (Viereck and Matus, 1990; Fischer and Romano-Clarke, 1990). Interestingly, it is the phosphorylated form of MAP IB that is most strongly induced during neurite outgrowth (Aletta et al. 1988; Diaz-Nido et al. 1988), and recently we have found that MAP IB is phosphorylated in both isolated growth cones and the growth cones of rat cerebral cortical neurones in culture (Mansfield et al. 1991). In some axons in culture, the phosphorylated form of MAP IB is distributed in a striking gradient which is highest distally and lowest near the cell body whereas the nonphosphorylated form is distributed throughout the neurone (Mansfield et al. 1991). The phosphorylated form of MAP IB has also been found in high levels in growing axons in vivo (Viereck and Matus, 1990; Sato-Yoshitake et al. 1989). In Alzheimer’s disease, in which it is thought there is a massive attempt at axonal and dendritic regeneration, it is interesting to note that there is an accumulation of hyperphosphorylated MAP IB (Hasegawa et al. 1990).

Collectively, these results suggest an important role for MAP IB, particularly the phosphorylated form, in neurite outgrowth (Matus, 1988). MAP IB probably cross-links microtubules (Sato-Yoshitake et al. 1989) and therefore one possible role for MAP IB in growth cones is to bundle the newly-formed microtubules in the C-domain as it transforms into the neurite (Fig. 1) (Gordon-Weeks and Mansfield, 1991; Mansfield et al. 1991).

Naturally, the experiments described here, while providing answers to some questions, engender more. For instance, how is the soluble tubulin in the growth cone maintained and delivered to the growth cone from the cell body where, at least in axons, it is synthesized? Perhaps it comes down in the so-called ‘slow component a particles’ described by Weisenberg et al. (1987) which may correspond to the motile varicosities seen in some cultured neurites (Koenig et al. 1985). Another basic question is how is microtubule assembly coupled to growth cone motility (Gordon-Weeks, 1989; Mitchison and Kirschner, 1988)? The evidence that these events are linked includes the observation that the formation of microtubule loops following taxol treatment is associated with the collapse of the growth cone (Letourneau and Ressler, 1984; Gordon-Weeks et al. 1989; Mansfield and Gordon-Weeks, 1990, 1991) and that individual growing microtubules in growth cones may be coupled to filopodial actin bundles (Figs 1 and 2).

The work in my laboratory is supported by the MRC, Wellcome Trust, British Council and the Royal Society.