MUCH MORE ON MICROTUBULE SELF-ASSEMBLY

Enzymologists typically consider catalytic rate enhancement as the cardinal feature of biological catalysis, and enormous intellectual energy has been gainfully invested in the root causes of such accelerative mechanisms. Nonetheless, many regulatory signals are transduced by enzymatic activities that are far more impressive in terms of their information-transferring capacity than their respective turnover constants. One may justifiably argue that the key cytoskeletal components, tubulin and actin, are low-kcat nucleoside-5'-triphosphatases that act as highly regulated switches governing cell shape and motility. There is correspondingly an even larger family of GTPases commonly referred to as GTP- (or G-) regulatory systems. Indeed, the cardinal feature of G-protein-mediated regulation of receptors stems from GTPase-linked modulation of inhibitor or activator potency (Gilman, 1987; Stryer, 1991). Each G-protein has two states, each characterized by the binding of GTP or GDP, with their interconversion accomplished directly by GTPase or indirectly by nucleotide exchange. In much the same way, tubulin exists in two principal states: the assembly-competent Tb·GTP complex and the assembly incompetent Tb·GDP complex.

Signal-processing arises from the conformationally restricted properties of Tb·GTP and Tb·GDP in unpolymerized and polymerized forms, and although the microtubule literature is burgeoning, the scope of this discussion is limited to this issue. Likewise, the intent is not to develop quantitative kinetic models; rather, the goal here is to deal with the scaffolding of logic that serves as a basis for studying Òlow-kcatÓ enzyme activitiess of this sort. Readers might also wish to consult several other reviews for extended discussions of tubulin structure and polymerization kinetics (Purich and Kristofferson, 1984; Carlier, 1989), enzyme-catalyzed interconversion of microtubule cytoskeletal proteins (Terry and Purich, 1982), and control of tubulin gene expression (Cleveland, 1988; Cleveland and Sullivan, 1984). The author apologizes to those investigators whose studies of tubulin-nucleotide interactions are not mentioned in this discussion.

II. Guanine Nucleotide Interactions with Tubulin--- Twenty years ago, Weisenberg (l968ab) first reported that tubulin could undergo self-assembly in the presence of GTP. Earlier studies by Weisenberg et al. (1968ab) and Berry and Shelanski (1972) showed that tubulin is a heterodimer containing a single exchangeable guanine nucleotide site and another nonexchangeable guanine nucleotide site. MacNeal and Purich (1978a) first demonstrated that only the exchangeable site is engaged in GTP-dependent assembly. GDP can bind in place of GTP at the exchangeable nucleotide site, and Brylawski and Caplow (1983) demonstrated that nucleotide exchange occurs with a rate constant of about 0.14 s-1. Photo-affinity cross-linking experiments with 8-azido GTP indicate that the exchangeable site is located on the ß-subunit. Strong evidence against any coupling of the tubulin N-site to assembly was provided by Speigelman et al. (1977) who isolated tubulin from Chinese hamster ovary cells that had been pulse-labeled with [35S]methionine and [32P]orthophosphate to label protein and nucleotide during the biosynthesis of new tubulin molecules. These studies showed that the half-life of N site[32P]GTP label and the half-life of tubulin were 33 and 45 hours, respectively. This excluded the exchange of phosphoryls of the N-siteGTP with the E-site nucleotide during assembly. The heterodimer arrangement of E- and N-sites also rationalizes the observed head-to-tail arrangement of tubulin subunits in each of the thirteen protofilaments that run longitudinally in microtubules (Purich and Kristofferson, 1984). This intrinsic polarity results in biased tubulin addition to the two microtubule ends (Rosenbaum and Child, 1976; Witman, 1975), and they are designated (+) for the faster growing end and (-) for the more slowly growing end. Microtubule polarity is also evident in terms of dynein binding (Haimo et al., 1979), and the main globular head of dynein tilts at an angle of 55° toward the (+)-end. The minus ends are tethered to centrosomes in cells, and the plus end of each microtubule terminates at or near the cell margin (Bray, 1992). Moreover, the ATP-motors dynein and kinesin recognize the intrinsic structural polarity of the heterodimers within the microtubule lattice, and they are multi-subunit complexes that contain "cargo"-recognition sites to bind and transport specific cellular components. Depending on the motor type, they move unidirectionally toward or away from the centrosome (Bray, 1992; Lloyd et al., 1986).

III. Bioenergetics of the Tubulin·GTP/GDP Cycle--- GTP hydrolysis appears to be tightly coupled to microtubule self-assembly (MacNeal and Purich, 1978a). Our original time courses for self-assembly and E-site GTP hydrolysis, using cycle-purified microtubule protein preparations containing both tubulin and microtubule-associated proteins (MAPs). MAPs can greatly stimulate tubulin polymerization by increasing the efficiency of nucleation, which was reflected in the study by Sloboda et al. (1976) as a decrease in the average polymer length. MacNeal and Purich (1978a) found that the presence of MAPs leads to stimulated microtubule assembly rates and GTPase activity, and the availability of large quantities of bacterially expressed MAP-2 tubule-binding domain (Coffey et al., 1993) should now permit reinvestigation of the rates of assembly and GTP hydrolysis at MAP levels corresponding to their high cellular content. These observations, along with changes in the critical concentration in the presence of GDP, led Karr et al. (1979) to propose their boundary stabilization model for microtubule growth and stabilization. Here, Tb·GTP participates in tight-binding interactions that result in tubule stabilization. In this model, Tb·GDP (itself arising from assembly-induced hydrolysis) occurs within the interior microtubule lattice. Each internal Tb·GDP is stabilized by its surrounding Tb·GDP neighbors or by Tb·GTP at the microtubule ends which are referred to as boundaries. Loss of the single terminal layer (or boundary) would raise the critical concentration as a consequence of the much weaker binding of Tb·GDP molecules at the tubule ends. Such a model is also fully consistent with GDP inhibition of assembly as well as with the ability of enzymatic conversion of GTP to GDP resulting in microtubule disassembly. Delayed GTP hydrolysis can also be represented as the second model wherein Tb·GTP molecules persist randomly throughout the microtubule lattice for a period following polymerization. To date, there has been no evidence supporting this mechanistic possibility, and the isotope exchange studies of Angelastro and Purich (1990) would suggest that this is an unlikely mechanism. Indeed, the microtubule lattice may exclude tubulin-bound GTP except at the terminal boundary. A third model involves formation of a multi-layer "cap" comprised of many Tb·GTP molecules, and this was suggested on the basis of experiments with the microtubule-stabilizing drug taxol (Carlier and Pantaloni, 1981; Hill and Carlier, 1983). Assembly under such "forcing conditions" may have no relevance to microtubule stabilization in vivo, especially when one recognizes that Tb·GTP, Tb·GDP, and even E-site nucleotide-free Tb can polymerize readily in the presence of taxol.

The kinetic coupling of GTP hydrolysis and microtubule self-assembly (MacNeal and Purich, 1978a) has been verified by several investigators who subsequently found no evidence for a deep "cap" of Tb·GTP molecules, even at very high tubulin concentrations favoring rapid polymerization. Indeed, even when working over an extraordinarily large range of tubulin concentration to increase the celerity of assembly, O'Brien and Erickson (1989) did not observe any net accumulation of GTP beyond that expected on the basis of the boundary stabilization model of Karr et al. (1979). This was also confirmed through independent studies by Jordan and Wilson (1990). Finally, although one might seek to circumvent these mechanistic constraints by proposing a role for Tb·GDP·Pi in place of Tb·GTP complex, the former cannot have significantly similar thermodynamic stability as that of Tb·GTP. Were they isoenergetic, or even nearly so, the stable oxygen-18 exchange studies would have readily indicated this through the loss of oxygen-18 atoms during such reversals. Based on the use of aluminum tetrafluoride anion (AlF4-) as a phosphate analogue, Carlier et al. (1988b, 1989) suggested that orthophosphate may bind within the gamma-phosphoryl pocket of the exchangeable nucleotide site in a manner that confers at least some of the stabilization displayed by the bound-GTP conformation. In particular, they proposed that microtubule-bound tubulin-GDP-Pi is a stable intermediate in the GTPase reaction, suggesting then that the dynamic instability of microtubules could be governed by loss of such "caps" or by Pi release into the medium. Formation of such a stable intermediate creates a paradox; if release of Pi increases the instability of microtubule ends, this would require a correspondingly less favorable dissociation of Pi.

The heterodimeric structure of tubulin results in different configurations at each microtubule end. Thus microtubules have protofilaments running parallel and tubulin protomers regularly arranged in a head-to-tail manner. Biased addition is a consequence of the nature of microtubule tubulin interactions, where the transition states for protomer-polymer reactions need not be identical. Thus, the corresponding activation energies for the reactions at the two ends will in general be different, and the corresponding bimolecular rate constants will then be different. Template-directed assembly with axonemes serving as microtubule-organizing centers is known to occur at the distal or (+)-ends (Olmsted et al., 1974; Dentler et al., 1974). This is also true for basal bodies (Snell et al., 1974) and centrioles (McGill and Brinkley, 1975; Telzer and Rosenbaum, 1979). This polarity also accounts for the maintenance of different critical concentrations, each representing the respective equilibrium constants for tubulin addition/release reactions at each end. Such differential affinity arises from the fractional retention of free energy of GTP hydrolysis, and the less stable end (i.e., the (-)-end) does not release as much free energy of hydrolysis during each step in microtubule elongation.

IV. Microtubule Assembly/Disassembly Turnover--- Oosawa (1975) was among the first to consider how supramolecular protein assemblages can arise via a condensation equilibrium model involving nucleation, elongation, monomer/polymer equilibration, and polymer length redistribution from an initially kinetically controlled condition to a final, equilibrium controlled situation. These processes, as they apply to microtubule assembly, have been considered in an earlier review (Purich and Kristofferson, 1984). An additional feature of such processes, one that was not developed in the Oosawa treatment, is the involvement of nucleotide hydrolysis in priming the system for eventual turnover, and this feature is the primary subject of consideration here.

A. Treadmilling--- In the original formulation for actin dynamics (Wegner, 1976), the concept of cytoskeletal polymer "treadmilling" was shown to arise from ATP hydrolysis-induced differences in the critical subunit concentrations in equilibrium with each filament end. Such a process implies that the macroscopic critical concentration [i.e., K° equals the algebraic sum of the off-rate constants for both ends divided by the algebraic sum of the on-rate constants for both ends] lies between the microscopic critical concentrations for the more and less stable filament ends; at steady state of assembly/disassembly, this will consequentially lead to loss of subunits from the less stable filament end and accumulation of subunits at the more stable end. In the context of microtubule dynamics,

Margolis and Wilson (1978) first proposed a treadmilling model operating via exclusive addition and release of tubulin dimers from opposing assembly and disassembly ends of microtubules. Because rapid dilution of microtubule protein to below the critical concentration resulted in prompt depolymerization with a rate constant of around 0.1 min-1 for complete loss of polymer mass, Karr and Purich (1979) noted that this disassembly rate exceeded the steady-state treadmilling rate by 1000-times. They therefore proposed a minimal model wherein the possibility of reversibility at the so-called primary disassembly-end was also indicated by use of a broken arrow. Later analysis indicated that all four rate constants must be considered to account for the assembly/disassembly kinetics of tubulin dimer interactions with microtubule ends (Caplow, 1992), resulting in the following scheme for subunit addition and loss from both ends. These distinctions only apply to untethered microtubules, and they have only limited relevance to intracellular microtubules which are bound to microtubule-organizing centers (MTOCs). An amended phase-dynamics model that represents a hybrid of the treadmilling and dynamic instability models has also appeared (Farrell et al., 1987). To date, however, despite the use of a variety of techniques and cell types, there is really no compelling evidence for treadmilling in intracellular microtubule dynamics. This does not mean that microtubule treadmilling is only an interesting in vitro property, and the precedent for treadmilling as a feature of the actin cytoskeleton seems well founded. It is, however, difficult to reconcile how tethered microtubules can treadmill in the absence of populations of polymers with two free ends of differential stability.

B. Dynamic Instability---The biological implications of models for GTPase involvement became especially evident when Mitchison and Kirschner (1984) presented an explanation for microtubule steady-state dynamics in terms of GTP hydrolysis and the stochastics of losing the stabilizing cap (or boundary) of tubulin molecules containing unhydrolyzed GTP at their E-sites. In the "dynamic instability" model, length changes in microtubules at steady-state are thought to arise from the overall balance of two phases: the first involving slow growth of the majority of microtubule polymers; and the second arising from the rapid disassembly of a smaller fraction of polymers. They proposed that microtubules at steady-state contain Tb·GTP protomers at the polymer ends, forming a cap of stably bound protomers. In contrast, the microtubule interior lattice contains largely Tb·GDP protomers that are lost by endwise depolymerization whenever microtubules lose Tb·GTP protomers stabilizing their ends. Taken with the mounting evidence (vide supra) that there is no cap, the dynamic instability model most probably arises from the presence or loss of the stabilizing boundary formed by a few terminally bound Tb·GTP protomers. Tubules grow and remain stable as a consequence of Tb·GTP bound at the growth sites. They undergo stochastic disassembly in the improbable case that all growth points lose their Tb·GTP molecules. Under the influence of as yet unidentified intracellular signal(s), terminal Tb·GTP molecules are probably released from growth points. Then, disassembly promptly takes place in what can be regarded as a catastrophe. Depending on the circumstance, disassembling tubules may completely depolymerize or may recover such that regrowth begins the cycle anew. Kristofferson et al. (1986) used biotinylated tubulin and antibody methods to analyze the time evolution of microtubule length redistribution which is a measure of steady-state microtubule dynamics. They were able to confirm the basic tenets of the dynamic instability model, and they clearly demonstrated that some tubules exhibit catastrophic depolymerization. These investigators found that microtubules showed no evidence of treadmilling, first proposed for tubules by Margolis and Wilson (1978). Video microscopy has permitted direct observation of microtubule assembly/disassembly dynamics in vitro. Horio and Hotani (1986) first used dark-field optics to observe the growth and shrinkage phases, but so-called Allen video enhanced contrast microscopy has become most convenient.

As noted earlier, the concept of stabilizing "caps" has been advanced to explain microtubule stability, and Carlier et al. (1988b,1989) suggested that microtubule-bound tubulin-GDP-Pi is a stable intermediate and that the dynamic instability of microtubules could be governed by loss of such "caps" or by Pi release into the medium. Nonetheless, two other groups have concluded that such is not the case for either reassembled brain tubules (Caplow et al., 1989) or avian erythrocyte marginal band microtubules (Trinczek et al., 1993). In both studies, the presence of elevated orthophosphate concentrations were without any significant effect on the observed dynamics of subunit addition during elongation, subunit release during the rapid shortening phase, or the frequency of transitions from shortening to regrowth phases.

V. Intracellular Microtubule Dynamics--- Evaluating microtubule assembly/disassembly dynamics directly in living cells has proven to be a challenging task that requires special insights about cellular behavior as well as technical ingenuity. In practice, no single method has proved to provide both temporal and spatial resolution required to characterize all aspects of processes (which include accurate estimates of kinetic constants and both the size and intracellular location of stable and dynamic microtubule pools). Direct microscopic observation of microtubules, for example, offers an attractive means for analyzing net rates of tubulin gain or loss (Horio and Hotani, 1986; Walker et al., 1988; Cassimeris et al., 1988; Saxton et al., 1984), but the resolving power of light and fluorescence microscopy cannot distinguish single tubules from bundled microtubules or even a pair of microtubules that run closely parallel to each other. Thus, even as image reconstruction techniques are improved electronically, the physics of light refraction still will limit the technique to cells or cell regions containing only a few microtubules. In the case of microinjection, proteins can be introduced into unfertilized oocytes (Tanaka and Kirschner, 1991; Reinsch et al., 1991; Sabry et al., 1991), thereby allowing the investigator to observe details of cytoskeletal assembly and disassembly after the subsequently fertilized egg proceeds through embryogenesis. Such an approach works best with large oocytes, as in the case of Xenopus, so that the embryo's development does not lead to significant dilution of the originally microinjected reporter protein. Recent studies in Xenopus exemplify how microinjection of rhodamine-labeled or bis-"caged" fluorescein-labeled tubulin during the first cleavage division can be used to examine microtubule movement. At the appropriate stage of embryonic development, the "caged" fluorophore can be photoactivated by brief exposure to a focused light beam, and intracellular tubule dynamics can be recorded with an intensified silicon-intensified target video camera. In experiments of this sort one must contend with problems of photo-induced oxidation of the fluorescent tag itself as well as damage to light-sensitive subcellular components. Because dioxygen is essential for photo-oxidation, efforts to eliminate molecular oxygen by redox scavengers or by physically excluding this gas helps to minimize photo-damage. Nonetheless, one must be somewhat wary of such practices, because oxygen per se is an essential substrate for maintaining the [ATP]/[ADP] and [GTP]/[GDP] concentration ratios via oxidative phosphorylation. The sensitivity of the assembly reactions of actin and tubulin, respectively, to these ratios cannot be overemphasized. Likewise, the [ATP]/[ADP] ratio is important in modulating the action of microtubule-based motors [i.e., dynein and kinesin], and various regulatory protein kinases. While the metabolic labeling technique can provide estimates of rates and extents of tubulin microtubule exchange, results obtained with this approach must be interpreted using other data to gain insight about the cellular locations of stable and dynamic tubules. For example, there is mounting evidence for the occurrence of stable microtubules located in neurite outgrowths of PC12 cells grown in the presence of nerve growth factor (Okabe and Hirokawa, 1990), and PC12 cells grown in the presence of NGF doubled their tubulin content (Drubin et al., 1985; Schulze and Kirschner, 1987). Based on these observations, results of metabolic labeling experiments (Angelastro and Purich, submitted) indicate that PC12 cells grown in the presence of NGF contain two microtubule pools: the first characterized by rapid assembly/disassembly and demonstrating only a three-to-four min lag period relative to the rise in radiospecific activity of the cellular guanine nucleotides; and a second less abundant, non-dynamic microtubule pool (i.e., corresponding to about 40% of the microtubules). Before leaving these brief descriptions of methods for examining intracellular dynamics, it is appropriate to acknowledge that only isolated cells have been studied thus far. Many obstacles confound the extension of these methods to the investigation of whole tissues or even tissue slices. From the table given below, one can readily appreciate that substantial turnover (i.e., amounting to 50% or more subunits replaced by disassembly and subsequent reassembly) occurs within thirty minutes. These observations indicate that the microtubule cytoskeleton is highly dynamic. In the case of PC12 cells, the microtubule-bound GDP pool undergoes disassembly and equilibrates with the cellular guanine nucleotide pool in less than three minutes. The microtubule bound GDP pool was found to be about 20 µM, suggesting that microtubule dynamics in PC12 cells expends as much as 7-10 µM GTP each minute. For comparison, brain cells typically produce in excess of 30 mM nucleoside-5'-triphosphate (predominantly as ATP) over such a period; we can thus estimate that far less than one percent of cellular nucleoside-5'-triphosphate is consumed in maintaining microtubules as a dynamically responsive organelle. The GTP pool must undergo rather rapid turnover, suggesting a clear role for nucleoside-5'-diphosphate kinase in the transfer of phosphoryls from the ATP pool to the GTP pool (Terry and Purich, 1982).

VI. Options for Controlling Microtubule Dynamics--- The abrupt and highly committed nature of cytoskeletal rearrangements during the cell cycle is accomplished, at least in part, by complete microtubule cytoskeleton disassembly. The interphase microtubule cytoskeleton, for example, must be fully dismantled to allow mitosis to commence. Thus, upon transition from a growing to a shrinking tubule in the dynamic instability model, totality of depolymerization occurs, and this may be a very desirable outcome. During microtubule disassembly, Tb·GDP molecules rapidly issue from microtubule ends and probably crowd the vicinity of each unwinding microtubule end, and any entry and binding of Tb·GTP molecules would probably be most unlikely. Moreover, restoration of a Tb·GTP boundary layer and restabilization of a microtubule end would be especially improbable if Tb·GTP molecules must coexist at several growth sites simultaneously before stabilization can be restored.

When viewed as a fully loaded pistol with a "hair trigger", there must be multiple hierarchies of regulatory interactions, including those encoded in the bound-GDP and bound-GTP states of unpolymerized tubulin and microtubules, that endow tight control on the triggering device per se. Modulation of microtubule dynamics may be elicited by microtubule-associated proteins and enzymes, metabolic signals, as well as other low-molecular-weight factors. Sammak et al. (1987) first suggested a tempered instability model in which MAPs may play a key role. Indeed, the heat-stable fibrous MAPs (particularly MAP-2 and tau) do appear to confer microtubule stability in terms of critical concentration behavior and length redistribution kinetic properties. Protein kinase-mediated MAP phosphorylation releases this constraint of the latter in vitro (Raffaelli et al., 1992). Even more intriguing is the possibility of enhanced instability that may be achieved through the action of a Xenopus oocyte protein found to sever microtubules after mitotic activation (Vale, 1991).

Microtubules have their (+)-ends near the cell margin and distal to the centrosomes which appear to bind and stabilize the (-)-ends; loss of subunits only from the (+) ends may be insufficient to permit rapid microtubule disassembly during transitions in the cell cycle. If both ends were free to disassembly, tubulin dimer release could proceed with rate constants of 200-500s-1, corresponding to length changes of about 6-12 µm/min. Severing long tubules into several shorter fragments could thus increase both the number and kind of disassembling ends, thereby allowing for rates upwards of 50 µm/min. Vale's experiments on the severing of taxol-stabilized tubules are consistent with extensive, multiple fragmentation. In terms of control mechanisms, one should also not discount the opportunity for tubule disassembly in response to local surges of calcium ion, and any factor causing a drop in the cellular [GTP]/[GDP] poise could likewise tip the balance to favor disassembly. The latter may be especially relevant to transformed cells which may be influenced by oncogene-associated GTPases or altered nucleoside 5'-diphosphate kinase; their effects on cellular [GTP]/[GDP] and tubule dynamics remain to be elucidated. Enzymatic modification of tubulin [e.g., acetylation (Maruta et al., 1986), tyrosination (Arce et al., 1975; Webster et al., 1987), glutamylation (Paturle-Lafanechere et al., 1991), and even ADPribosylation (Scaife et al., 1992)] also may impact on the stability and dynamics of microtubules. Such modifications may change the on- and off-rate constants for tubulin dimer addition and release, and they may also modulate the susceptibility of microtubules to the action of severing proteins.

Efforts aimed at determining intracellular microtubule dynamics led to the discovery of 2' deoxyGTP in nascent tubulin in response to nerve growth factor (NGF) treatment of PC12 and embryonic chick dorsal root ganglion neurons (Angelastro and Purich, 1993). Identity was confirmed by comparison of HPLC elution characteristics with genuine dGTP, ultraviolet absorption spectra, and failure to react with sodium metaperiodate. This unexpected finding of dGTP incorporation into the nonexchangeable site of tubulin may reflect NGF-induced changes in guanine nucleotide metabolism and coincident induction of tubulin synthesis. In this regard, the high abundance of tubulin in neuronal cells may afford a mechanism for sequestering significant stores of dGTP, such that NGF evokes a pause in DNA synthesis and cell proliferation and likewise stimulates neurite outgrowth and neuronal morphogenesis. Although the physiologic consequences of N-site dGTP remains to be elucidated, the finding clearly indicates that nucleotide interactions with microtubules may provide cells with a robust range of regulatory options, beyond those considered in other sections of this review.

VII. Concluding Remarks--- Issues described in this report allow one to recognize that the dynamic properties of microtubules are inextricably linked to the bioenergetics of assembly-induced GTP hydrolysis. Various regulatory opportunities can arise whenever structural perturbations in the tubulin molecule are communicated to the GTPase active site in a manner that affects the activation energy for this reaction or alters the stability of the initial GTP-bound or final GDP-bound states. Accordingly, voids in our current understanding of the biochemistry of the microtubule cytoskeleton limit a fuller consideration of the GTPase reaction. Among the most salient lacunae are the following. First, we lack X-ray crystallographic data at any level of resolution on the tubulin dimer, and this places the tubulin field at a distinct disadvantage, relative to actin, in reaching definitive mechanistic conclusions about the nucleotide binding site. Second, the detailed lattice structure of microtubules and the geometry of tubule ends are not settled, and recent reports (Mandelkow et al., 1991; Song and Mandelkow, 1993) call into question earlier conclusions about singlet tubule fiber structure, and these new findings suggest that a regular helical arrangement does not obtain either within the lattice or at the tubule ends. Furthermore, knowledge of the number of growth sites on each microtubule is requisite for estimating the magnitude rate constants in endwise polymerization and depolymerization experiments (Kristofferson, et al., 1980; Karr et al., 1980). Third, despite the identification of specific tubulin isoforms, there is still no satisfactory accounting of the role(s) of tubulin structural diversification in cellular architecture or physiology. Fourth, we are likewise at a loss to explain how post translational modification alters tubulin's participation in various cytoskeletal processes. And fifth, any allosteric cross-talk between tubulin subunits is still uncharacterized, and one must consider the possibility that the N-site plays more than a cofactor role in the proper folding and assembly of the tubulin dimer. Ideally, such information will provide a clearer picture of the mechanistic relatedness of systems involving GTP and ATP hydrolysis to modulate ligand binding affinity. One excellent example of such a system is RecA, and Story and Steitz, 1992) have adduced a potentially general mechanism by which differential affinity of a protein for a ligand is coupled to nucleoside-5'-triphosphate hydrolysis.