Tau protein and the establishment of an axonal morphology

Primary neurons grown in dissociated culture are able to recapitulate some aspects of their in vivo morphology. One morphological feature of some dissociated neurons is their ability to elaborate two distinct populations of neurites which have features of axons and dendrites. The development of neuronal polarity was first characterized in detail in cultured rat hippocampal neurons (Bartlett and Banker, 1984a,6). These investigators demonstrated that the plated neurons were able to generate correct axonal and dendritic morphologies in the absence of environmental cues. This segregation occurs in sparsely plated cultures without cell-cell contact and upon a patternless background of polylysine. The distinctive properties of these ‘axons’ and ‘dendrites’, including their synaptic structure, receptors, ion channels, ribosomal content, presence of Golgi elements, and cytoskeletal compositions are retained in culture. Briefly, ribosomes and Golgi elements are present in the somatodendritic compartment, and absent from the axon. Neurofilaments, particularly phosphorylated neurofilaments, are more abundant and closely packed in the axon than in the dendrite. Microtubule-associated protein 2 (MAP2), a somatodendritic marker (Matus et al. 1981; Miller et al. 1982), only appears in the somatodendritic compartment of cultured neurons after they mature (Kosik and Finch, 1987; Caceres et al. 1986). In contrast MAP2 is compartmentalized to the somatodendritic portion of the cell from the time it is first expressed in brain tissue (Crandall et al. 1986). The orientation of the microtubules differs in the axons and dendrites of cultured neurons. In the axon, microtubules have their plus ends aligned distally (Heidemann et al. 1981; Burton and Paige, 1981; Filliatreau and DiGiamber-dino, 1981) while a mixed microtubule orientation prevails in the mid-portion of the dendrite (Baas et al. 1988; Burton, 1988).

Tau protein is a microtubule-associated protein that appears on gels as a group of isoforms with relative molecular masses from 45 to 62xlO³M_r (Cleveland et al. 1977). Tau was the first MAP shown to promote microtubule assembly in vitro (Weingarten et al. 1975). Tau is a single gene (Neve et al. 1986) from which diversity arises by alternative splicing (Himmler, 1989) and multiple phosphorylation states. The human tau gene has been mapped to the long arm of chromosome 17 (Neve et al. 1986) and located on a genetic linkage map of the chromosome (Haines et al. 1990). The exon structure of some of the brain isoforms has been characterized and shown to consist of six variants that arise from the inclusion or exclusion of three alternatively spliced exons (Goedert et al. 1989). Tau isoforms with higher masses have been described from certain neural tissues and neural cell lines, however their primary structures have not yet been established.

When tau cDNAs are hybridized to brain tissue sections the signal is located predominantly in neurons (Kosik et al. 1989a). Whether the very sparse silver grains over nonneuronal elements represent tau mRNAs is unclear. Using the technique of Northern blots of total RNA, tau is not detected in non-neural tissues (Neve et al. 1986). Within neurons, the tau signal, as detected by in situ hybridization, is primarily within the cell soma and the most proximal portion of the dendrite. This distribution contrasts with published reports of the localization of MAP2 mRNAs, which extend well into the dendrites (Brucken-stein et al. 1990; Kleiman et al. 1990; Gamer et al. 1988), but is similar to the distribution of the axonally transported protein GAP43, and tubulin. While at the protein level the distribution of MAP2 approximates to its site of synthesis, the tau protein is localized to the axon (Binder et al. 1985; Peng et al. 1986; Kowall and Kosik, 1987; Migheli et al. 1988; Brion et al. 1988). Perhaps positioning of ribosomes within the dendritic compart-ment may help in localizing a protein to the dendrite. Axons, on the other hand, lack ribosomes, and therefore face a distinct problem. The targeting of tau to the axon must involve its selective association with those microtubules destined for the axon.

The study of the localization of the tau protein in dissociated neuronal cultures has given rise to some conflicting observations. One study cited above demonstrated by western blots axonal enrichment of tau protein in sympathetic neurons grown in culture (Peng et al.1986). In dissociated rat hippocampal cultures tau was reported not to segregate to neurons (Dotti et al. 1987); however in dissociated rat cerebro-cortical neurons tau did segregate to an axon-like structure (Kosik and Finch,1987). In the later case, tau segregation did not occur until after the development of morphological polarity. That is, neurites were clearly identifiable as axons or dendrites before tau (or MAP2) localized correctly. There was a period of time during which antibodies to tau and MAP2 co-localized throughout the neuron, despite the formation of distinct axon-like and dendrite-like neurites.

The immunocytochemical findings described above make the attribution of a polarity function to tau problematic. However, the use of tau antisense to create a pseudo-genetic null mutation in cerebellar macroneurons suggested some role for tau in the cascade of events that leads to a polarized neuron (Caceres and Kosik, 1990). In these experiments a DNA oligonucleotide synthesized in the inverse orientation to a portion of the tau nucleotide sequence was added to the culture media one hour after plating. This sequence did not occur in the database other than in tau. Twenty four hours after plating the antisense treated cultures failed to react with tau antibodies by immunocytochemistry, dot immuno-binding, and by western blots, whereas the control sense treated cultures all reacted. It was also possible to demonstrate by polymerase chain reaction (PCR) that the tau mRNA was not detectable in the antisense treated cultures only (K. S. Kosik and R. L. Neve, unpublished observations). Similar results were obtained with a second tau antisense oligonucleotide that did not overlap with the first and was complementary to a region located entirely within the 5’ untranslated region of tau. Under the antisense conditions minor neurite outgrowth still occurred; however the neurons failed to elongate a single neurite and form an axon-like structure. Cells with minor neurites did label with antibodies to other proteins including MAP2 and tubulin. If the cells were not fixed at 24 h, but permitted to remain in the antisense-containing media beyond 24 h, they did develop polarity over the ensuing 24-48 h. However, when tau antisense was added repeatedly every 24 h, i.e. at time zero, at time 24 h and at time 48 h, and then fixed at 72 h, the cells remained locked at a stage with only minor neurites (Caceres et al. 1991). While it is unlikely that tau initiates polarity, we concluded that tau is necessary for the ultimate generation of polarity.

A second group of experiments were performed to assess the contribution of tau to the maintenance of polarity (Caceres et al. 1991). In these experiments long nontapering neurites were referred to as axon-like, and a shorter population of neurites which tapered were referred to as dendrite-like. Under normal conditions it is possible to categorize nearly all the neurites according to this schema with a very high level of agreement between observers. Tau antisense was added to the cerebellar macroneurons 72 h after plating. By 72 h in culture the cerebellar neurons have acquired polarity. One process is clearly distinct from the others and appears to be axonal; the remaining minor neurites are beginning dendritic differentiation at this time. After 24 h in tau antisense the cells were fixed and reacted with tau antibodies. Tau-immunoreactivity was considerably diminished in the antisense treated cultures. What minimal reaction remained had a granular appearance in contrast to the more homogeneous staining of control cultures. When the morphology of the neurons was compared under the sense and antisense conditions, it was not possible to identify an axon-like neurite in most of the antisense treated cells. However, during the 24 h in tau antisense the minor neurites continued to grow and differentiate into dendrites (Fig. 1). In fact the complexity of the dendrite-like network was greater than in the controls. Administration of tau antisense oligonucleotides at 72 h thus resulted in the loss of axon-like neurites, while dendrite-like neurites continued to grow.

The continued administration of tau antisense from the time of plating resulted in the persistence of minor neurites and their failure to differentiate into axon-like or dendrite-like neurites. However, when tau antisense was added after the onset of axonal differentiation, the minor neurites were able to undergo dendritic differentiation. We concluded that the establishment of an incipient axon provides a signal that permits dendritic differentiation.

At any time within the first 72 h after plating, it is possible to abolish or nearly abolish tau-immunoreactivity after 24 h in antisense. After 72 h in culture, a 24 h period of tau antisense is no longer sufficient to diminish tau-immunoreactivity. While there are many explanations for this observation, we have hypothesized that the half-life of tau increases as function of neuronal maturity. The increased half-life of the protein would make antisense experiments performed over a 24 h period less effective. One determinant of the tau half-life is its binding to microtubules. Microinjection experiments have shown that free tau has a short half-life and microtubule-bound tau has a longer half-life (Okabe and Hirokawa, 1989). Since tau is present in all minor neurites before the onset of polarity, its role in selectively elongating a single neurite may involve the enhanced binding of tau to microtubules in the elongating neurite and/or an inhibitory state in those minor neurites not undergoing elongation. Our preliminary data suggests that in rat cerebro-cortical cultures there is only a negligible amount of tau bound to microtubules before the onset of polarity, and as development progresses the amount of tau in the microtubule-bound pool increases.

At least two categories of modifications can modulate the binding of tau to microtubules. One of these modifications is a splicing event in the microtubule-binding domain of tau. The carboxy terminus of tau has either three or four imperfectly repeated sequences of 31-32 amino acids in tandem. The transition from a three repeat form of tau to a four repeat isoform is developmentally regulated and in rat brain occurs at approximately post-natal day 8 (Kosik et al. 19896). The insertion of the repeated sequence occurs as a cassette exon, which according to the bovine tau gene structure is referred to as exon 10 (Himmler, 1989). Four repeat isoforms, in contrast to three repeat isoforms, show enhanced binding to microtubules and when transfected into certain mammalian cell lines are able to bundle microtubules (Lee and Rook-Greenhalgh, 1990). However, in neuronal culture polarity is evident well before a shift to the adult tau isoforms can be detected. It is therefore unlikely that splicing of exon 10 regulates polarity.

There is also evidence that the phosphorylation state of tau can affect its binding to microtubules. Alkaline phosphatase treatment of tau resulted in a more rapid and more extensive polymerization of microtubules (Lindwall and Cole, 1984). Ca²⁺/calmodulin-dependent protein kinase (CaM kinase) phosphorylation of tau inhibits its ability to promote assembly of microtubules, whereas the catalytic subunit of type-II cyclic AMP-dependent protein kinase had no effect on the microtubule assembly properties of tau (Yamamoto et al. 1985). Although these authors noted phosphorylation on both seryl and threonyl residues, Steiner et al. (1990) have described a single site of CaM kinase phosphorylation of tau in the carboxy terminus beyond the repeated sequences. In essence the question of which phosphorylation sites in tau modulate microtubule binding and which are the responsible kinases is very much open.

Neurite outgrowth immediately after plating permits observation of the transition from a symmetric array of actively elongating and retracting neurites to the establishment of a single stable neurite as the axon. Before polarity is established, the initial minor neurites are highly dynamic structures that do not achieve net elongation (Dotti and Banker, 1987). Viewed at shorter time intervals, though, these structures can achieve significant lengthening before they ultimately retract or become established neurites. Neurites must coalesce from the veil-like lamellopodial edge that precedes their formation into a short elongated structure with formed microtubules that can undergo rapid depolymerization during retraction events. While extrapolation from the behavior of individual microtubule systems to the larger regulation of cell shape is speculative, the predictions of dynamic instability, as observed in vivo (Horio and Hotani, 1986; Sammak et al. 1987; Sammak and Borisy, 1988), provide a parallel to minor neurite behavior. In brief, dynamically unstable microtubules are short, and completely and rapidly depolymerize. These catastrophic events increase the monomer or oligomer pool and thus slowly lengthen the few long microtubules, resulting in a length redistribution of the population. Polymer dynamics are probably capable of generating the push and pull forces required for the movement of supramolecular structures at their ends. Deformation of liposome membranes as a result of microtubule polymerization has been demonstrated (Hotani and Miyamoto, 1990) and a measure of the force generated by microtubule depolymerization has recently been published (Coue et al. 1991).

There is some evidence that binding of MAPs to microtubules reduces dynamic instability (Keates and Hallett, 1988; Bre and Karsenti, 1990). Since the binding of MAPs to microtubules is reduced by MAP phosphorylation (Lindwall and Cole, 1984; Jameson et al. 1980), and MAP phosphorylation can reverse MAP-induced stabilization of microtubules (Job et al. 1985), the likely consequence is an increase in dynamic instability. Along these lines it is reasonable to suggest for heuristic purposes that phosphorylation of certain MAPs, particularly tau, maintains minor neurites as minor neurites.

Associated with neurite outgrowth is a progressive stabilization of the microtubules (Black and Greene, 1982; Lim et al. 1989). Axons contain stable microtubules; however, even in regions of high microtubule stability, such as the axonal shaft of a mature axon, there remains a significant pool of non-polymerized tubulin. For instance in the squid giant axon, 22-36% of the tubulin is monomeric (Morris and Lasek, 1984). Therefore the maintenance of a significant monomeric or oligomeric pool of tubulin requires that any MAPs free in the cytoplasm must be inhibited from promoting polymerization. That MAPs do spend a certain portion of their time free of microtubules is suggested by the off times in fluorescent labeling experiments (Olmsted et al. 1989). An explanation for this observation may be that free MAPs are readily and rapidly phosphorylated. Under these circumstances nearly all of the free MAP pool would be maintained in the phosphorylated state and the degree of MAP dephosphorylation would regulate binding and subsequent microtubule stabilization.

Some support for a role of selective microtubule stabilization in the generation of morphology comes from myogenic differentiation in which a stabilized microtubule array is specifically detected in elongating portions of the cell and precedes the accumulation of muscle myosins (Gunderson et al. 1989). The establishment of polarity in MDCK II cells involves increased microtubule stabilization (Bre et al. 1990). In cerebellar macroneurons selective stabilization of a microtubule population within the developing axon during the establishment of polarity has been suggested by experiments with antibodies to acetylated tubulin (Ferreira et al. 1989).

As a working model, we propose that an intrinsic mediator of polarity could be a phosphatase, which stochastically acts upon the tau molecules within each neurite. The neurite, with a certain threshold of dephosphorylated tau, is able to attain maximal microtubule stability and consequently elongate. Once begun the phenomenon may occur cooperatively such that the elongating neurite continues to enhance its growth in comparison to the others. Once a population of stable microtubules is present in a neurite, further extension of the microtubule system could arise by nucleation from more proximal microtubules. A model of this sort involving stochastic events and cooperative phenomena (although not necessarily a phosphatase) circumvents the conceptual problem in the existing models of neuronal polarity which attribute polarity to the earliest immuno-detectable protein present in the nascent axon. Proteins with which the establishment of polarity has been correlated are GAP43 (Goslin et al. 1988), synapsin and synaptophysin (Fletcher et al. 1989). Still unexplained in such a mechanism is how these molecules are selectively segregated to a single neurite. On the other hand, unexplained by a stochastic model is the observation that daughter cells develop neurites as mirror images (Mattson et al. 1989), which could suggest some intrinsic predetermination of the direction of outgrowth. A second problem with the model proposed here is the prediction that over time the other minor neurites should also elongate. Some feedback inactivation from the elongating neurite, perhaps of the putative phosphatase, is required to prevent elongation of the other neurites.

We are indebted to Sonja Potrebic for her expertise with the Eutectics Neuron Tracing System. The studies were supported by NIH grants AG06601 and AG06172.