Effect of ATP on the Kinetics of Microtubule Assembly*

We investigated the role of ATP in the assembly of microtubules. Tubulin, prepared by chromatography on DEAE-cellulose, was nearly devoid of nucleoside diphosphokinase activity. ATP induced assembly in such preparations for a single assembly/disassembly cycle; then further assembly could not be induced by ATP unless the system was supplemented with additional GTP. This suggests that the E-site must contain GTP for polymerization and ATP interacts at a different site on tubulin. Although tubulin can be assembled into microtubules in 1.0 m~ GTP, the inclusion of 0.2 m~ ATP along with the GTP increases the rate and extent of assembly. The enhancement increased with increasing ATP concen- trations. The inclusion of the critical concentration from 1.5 to 0.9 mg/ml. Analysis of assembly rate versus protein concentration suggested that ATP also affects nuclea- tion.

The role of ATP is not clearly defined. ATP can promote microtubule assembly (12, 13) through the action of an NDP kinase' which can phosphorylate tubulin-associated GDP at the expense of the y-phosphate of ATP (4, 14, 15). The NDP kinase activity is not inherent to the tubulin and can be removed from microtubule protein by ion-exchange chromatography (4,15-17). Tubulin has been reported to be incapable of binding ATP (3, 14) although there is more recent evidence to the contrary (18).
There is evidence that ATP functions in a regulatory capacity in microtubule assembly. ATP was shown to alter the assembly characteristics and stability of microtubules ( 19-21). Increased rates of microtubule assembly and disassembly at steady state occurred in the presence of physiological concentration of ATP (20). We reported that ATP induces the formation of aggregates of tubulin rings (18) which are thought to function as nucleation centers in microtubule polymerization.
There are several models for microtubule assembly which incorporate ringlike aggregates of tubulin as nucleation sites (22-24). Kinetic studies (25, 26) are consistent with a condensation-polymerization mechanism where there is an initiation phase (lag) followed by an elongation phase. Electron microscopy reveals that rings, initially present, disappear early in the polymerization process (23). Rings could uncoil to become protofiaments which aggregate laterally to form a microtubule. Bryan, however, claims that rings are not obligatory intermediates in microtubule formation (26) although they might have appeared transiently during the assembly pathway (27) and, therefore, overlooked in his analysis.
We previously showed that, at elevated M P concentrations, ATP can induce the formation of tubulin rings (18). We now report that ATP can promote microtubule formation in the absence of NDP kinase activity and that it can significantly alter the assembly kinetics. We propose that ATP effects nucleation and provides further support for the existence of an ATP binding site on tubulin independent of the GTP binding sites.

MATERIALS AND METHODS
Tubulin was prepared from fresh bovine brains using DEAE-cel-Idose (Whatman DE52) chromatography according to Weisenberg et al. (6). This was performed either by the batch procedure as described or on a DEAE column (5.0 X 4.0 cm) using a step elution of 1 column volume of 0.25 M NaCl in PM buffer (10 m~ phosphate, 5 mM Mgz+, pH 7.0) and then eluting the tubulin with 1 column volume of 0.5 M NaCl in PM buffer. The latter procedure yielded a product with more consistently low NDP kinase activity. Tubulin was stored at -20 "C in PM buffer containing 1.0 M sucrose. All buffers used in the purification also contained 0.1 mM GTP. For assembly experiments, tubulin was dialyzed overnight against approximately 400 volumes of assembly buffer (10 m~ 2-(N-morpholino)propanesulfonic acid, 1.0 mM ethylene glycol bis(,C?-aminoethylether)-N,N,N',N'-tetraacetic acid, The abbreviations used are: NDP k i n a s e , nucleoside diphosphokinase; PM, phosphate-magnesium; HPLC, high pressure liquid chromatography; AMP-PNP, adenylyl imidodiphosphate.
3.4 M glycerol, pH 6.4) containing the appropriate concentration of Mg". For assembly, the indicated nucleotides were added and then the sample was warmed to 37 "C in a thermostatically controlled cuvette chamber of a Cary 15 recording spectrophotometer. The increase in turbidity was monitored at 350 nm.
For purposes of comparison, tubulin was prepared by repeated cycles of assembly and disassembly according to the method of Shelanski et al. (28). The procedure was performed both with and without (29) glycerol. The products were stored at -80 "C.
Sodium dodecyl sulfate-polyacrylamide gel electrophoresis was performed with vertical slab gels containing 7.5% acrylamide following the procedure of Laemmli (30). NDP kinase activity was determined by direct measurement of GTP produced from GDP and ATP with the use of high pressure liquid chromatography. PM buffer was made 0.5 m M in ATP and GDP. Protein was added and the mixture was incubated for 30 min at 37 "C. (The assay was shown to be linear with time and protein concentration over the range used.) The reaction was stopped and protein removal was facilitated by the addition of 100% trichloroacetic acid to a final concentration of 5% followed by centrifugation. Nucleotides were separated and quantified on a reversed phase column (Altex, Ultrasphere-ODS, 25 cm) equilibrated with 5 m~ tetrabutylammonium phosphate (Altex), 30 mM KH2P0,, 4% CHXN. Nucleoside di-and triphosphates are well separated with a linear gradient of 20-358 CHICN. NDP kinase activity was also determined by measuring ADP production in a coupled enzyme assay (31).
All nucleotides were obtained from Sigma (>98% pure). Protein concentration was determined by the method of Lowry et al. (32). Electron microscopy was performed in a Philips EM301. Samples were futed with 2 4 % glutaraldehyde, placed onto carbon-coated grids, and negatively stained with 1.5% uranyl acetate.

RESULTS
Microtubule protein prepared by cycles of assembly and disassembly contains approximately 15-25% non-tubulin proteins (33,34). When tubulin is purified by chromatography on DEAE-cellulose, most of the microtubule accessory proteins are removed and the final product is greater than 98% tubulin.   preparations. Tubulin, prepared by: I , chromatography on DEAEcellulose, 2, cycle procedure with glycerol, and 3, cycle procedure without glycerol, was run on an sodium dodecyl sulfate-polyacrylamide gel (7.5%) to display the impurities. HMWs, high molecular weight proteins.
A protein that is found in many tubulin preparations is an NDP kinase. This activity can be removed from tubulin by chromatography on ion-exchange resins (4, 15, 17). That the NDP kinase was depleted in our preparation is illustrated in Table I. NDP kinase activity was determined by direct analysis with HPLC and by the coupled enzyme assay for tubulin fractions prepared by DEAE chromatography and by the cycle procedure. It is clear that the NDP kinase activity was substantially reduced in our preparations compared to once and twice cycled tubulin. It is interesting to note that when assayed by HPLC, the activity in cycled tubulin was 20% or less of the value determined by the coupled enzyme assay. Whereas the latter assay will respond to various kinases and phosphatases, the HPLC assay is specific for the formation of G T P from ATP and GDP. The discrepancy may indicate that several kinases and/or phosphatases accompany tubulin during the purification.
The level of NDP kinase remaining in our preparation was not sufficient for ATP-supported microtubule assembly. Fig.  2 illustrates what happened when tubulin was warmed in the presence of 1 mM ATP as the only added nucleotide. There was the usual turbidity increase resulting from microtubule formation as verified by electron microscopy. It is important to note that no assembly occurred without added ATP. When the sample was cooled to disassemble the microtubules and then rewarmed to 37 "C, no further assembly took place. Competency to form microtubules was returned when this sample was supplemented with GTP. We interpret this as follows. Following dialysis, the tubulin contained a small amount of bound GTP. ATP was able to induce microtubule assembly with the subsequent hydrolysis of GTP. Once the G T P was depleted, no further assembly could occur until additional G T P was added. In tubulin preparations containing sufficient NDP kinase (ie. cycle-purified tubulin), ATP alone TABLE I NDP kinase activity NDP kinase activity was measured in tubulin prepared by DEAE chromatography and by cycles of assembly/disa.ssembly with glycerol in the assembly buffer. Specific activities were determined by both HPLC and a coupled enzymatic assay. Microtubule assembly was followed by the increase in turbidity at 350 nm. After microtubule formation had reached steady state, the sample was cooled to 4 "C to depolymerize the microtubules. This was warmed and cooled two more times with GTP added to 1.0 m~ prior to the second cycle. can lead to several cycles of assembly tmough the regeneration of GTP from GDP.
Two additional points were revealed by this experiment. First, the addition of ATP was capable of inducing assembly in a system where this would not otherwise have occurred. Second, it is apparent that in order for ATP to support assembly, the E-site must contain GTP which then becomes hydrolyzed. Therefore, ATP cannot be interacting with tubulin at this site.
The level of Mg2+ in the assembly buffer affected the tubulin polymer induced by ATP. In a previous report (18), we described aggregates of tubulin rings which were formed upon warming tubulin in the presence of 1.0 ~l l~ ATP and 5.0 mM MgC12. When the Mg2+ concentration was lowered to approximately 2.5 m under the same conditions, extensive microtubule formation occurred with few, if any, aggregated rings, as revealed by electron microscopy. This is also supported by the previous experiment; whereas the formation of aggregated rings at 5.0 mM M 2 + was repetitive, the formation of microtubules at 2.5 m Mg2+ was not, that is, not until additional GTP was added.
The addition of small amounts of ATP to tubulin in the presence of GTP increased the rate and extent of microtubule assembly. Fig. 3 shows the polymerization profile of two tubulin samples that contained 1.0 m GTP and which were identical in every other respect except that one had ATP added to 0.2 m. The ATP was added as the magnesium salt so the total free M F level should have remained constant. The sample with ATP assembled with a shorter lag, a faster elongation rate, and a higher plateau value. By electron microscopy, the two preparations were indistinguishable, showing mostly microtubules whether ATP was added or not. When the experiment was repeated keeping the total nucleotide concentration constant (Le. 1.0 l~l~ GTP in one sample and 0.8 mM GTP plus 0.2 mM ATP in the other), the same effect was seen, indicating that the increased level of nucleotide was not responsible for the enhancement of microtubule formation. A detectable enhancement of assembly was seen at a GTP-to-ATP ratio of 10 to 1 implying that the site of ATP interaction is specific for ATP over GTP.
An increase in the concentration of ATP resulted in a proportional increase in the level of assembly. This is illustrated in Fig. 4 in which the plateau level of turbidity is plotted as a function of ATP addition. To demonstrate that this effect is not a result of added Mg2+ or the increased level of nucleoside triphosphate, the same experiment was performed in which comparable amounts of M e or GTP were added. The addition of these factors had little or no effect (Fig. 4), confirming that ATP was responsible for the enhancement of polymerization. The presence of ATP resulted in a reduction of the critical concentration for microtubule assembly. Two tubulin samples were prepared in assembly buffer (with 1.0 mM GTP) and ATP was added to one to a final concentration of 0.2 mM. These samples were polymerized by warming to 37 "C until they reached m.aximum turbidity, and then they were depolymerized by cooling to 4 "C. Dilution and repolymerization were carried out for several cycles. Fig. 5 presents the final change in turbidity as a function of the tubulin concentration. Extrapolation to zero turbidity change yields the critical concentration (C,) (36). Including ATP in the assembly buffer decreased the C, from 1.5 to 0.9 mg/ml. In Fig. 6, the rate of assembly is plotted as a function of the tubulin concentration. The initial rates were determined from the linear portion of the polymerization curves which represent the elongation phase of microtubule assembly. In this experiment, each point represents a different tubulin sample. Two curves are presented corresponding to assembly in 1.0 m GTP or 1.0 mM GTP plus 0.2 m~ ATP. The data obtained from assembly in GTP alone fall on a straight line while those with added ATP are distinctly nonlinear. The significance of this will be discussed later.
To test whether ATP hydrolysis was required for the enhancement of assembly, a comparison was made of the assembly kinetics for samples containing 1.0 rrm GTP or 1.0 mM GTP plus 0.2 m~ AMP-PNP, a nonhydrolyzable ATP analog. The polymerization curves were essentially the same, or perhaps in the sample containing the AMP-PNP slightly slower    such that spontaneous assembly could not occur at 37 "C, or if it did, it would do so at a very slow rate. To such preparations of tubulin, aliquots of preformed aggregated rings, prepared as described (la), were added at the times indicated in Fig. 7. The increase in turbidity indicated that the formation of microtubules resulted from this addition. Furthermore, the rate of assembly was dependent upon the concentration of rings (inset). A control experiment demonstrated that the small amount of ATP which was added with the ring fraction was not responsible for the initiation of assembly.

DISCUSSION
ATP as the only added nucleotide was capable of inducing microtubule assembly through a mechanism that does not involve transphosphorylation. The experiment described in Fig. 2 demonstrated that there was not sufficient NDP kinase present to support assembly and that ATP appears to interact at a site independent of the exchangeable GTP site.
That ATP does indeed stimulate the assembly rate is borne out by comparison of assembly kinetics in the presence and absence of a small amount of ATP. With the GTP level at 1.0 m~ (well in excess of tubulin), and ATP at 0.2 m, the rate and extent of polymerization were dramatically increased over those of an identical sample of tubulin with GTP but without ATP. Electron microscopy proved that there was extensive microtubule formation in both cases under the conditions used. A critical factor here was the M e concentration of 2.5 mM. We previously described (18) a similar enhancement of tubulin assembly in 1.0 m~ ATP compared to 1.0 m GTP, when both samples were at 5.0 n" Mg2+. At this level of Mg2+, the polymers formed with ATP were extensively aggregated tubulin rings whereas polymerization with GTP produced only microtubules. It is clear that the Mg2' concentration is an important factor in the microtubule assembly process. ATP may be acting to favor the formation of tubulin rings which are forced to aggregate by high M$+ levels, and are thereby prevented from proceeding onto the further stages in the assembly of microtubules. If rings are indeed structures for the nucleation of microtubule formation, an influence of ATP on ring formation would be consistent with the shorter lag time and increased assembly rate which were observed when a small amount of ATP was added.
Microtubule formation has been described as a nucleated In the absence of ATP, microtubule polymerization can take place. With bound ATP and at a Mg2+ concentration of 5 m~, aggregated rings of tubulin are formed, which, condensation mechanism (25, 35) involving a critical concentration below which net polymerization cannot occur. It is supposed that nuclei form first, and then tubulin dimers add onto the nuclei in an elongation reaction. The theoretical treatment of such a mechanism was developed by Oosawa and Higashi (36) to describe the polymerization of actin. In this treatment, the critical concentration (C,) is seen to be the inverse of the equilibrium association constant (K,) of the polymer. Thus C, = l / K p k-l/kl (1) where k-, and k , are rate constants for depolymerization and polymerization. Numerous factors including microtubule accessory proteins, have been reported to influence the critical concentration (24, 37, 38), and we show here that small amounts of ATP reduce C,. The latter observation implies that ATP increases k l , the rate of addition of tubulin dimers onto the ends of microtubules, or it decreases k-,, the rate of release of dimers, or both. As illustrated in Fig. 6, there is evidence that ATP affects nucleation as well as the critical concentration. Gaskin et al. (25) and others, using preparations of microtubule proteins that contained accessory proteins, reported that microtubule assembly exhibited mixed kinetics of an order only slightly higher than first. For several reasons (25,38), it was concluded that the kinetics predominantly reflects a rate-limiting step in the nucleation process. As shown in Fig. 6, in the absence of accessory proteins and ATP, the dependence of assembly rate on tubulin concentration was strictly first order (above the C,) . Apparently, this strictly first order dependence represents some step in nucleation that is not only predominant in the kinetics, but absolutely rate-limiting. This rate limitation, which presumably involves the addition of a tubulin dimer to some intermediate structure, was at least partially relieved by ATP. This was demonstrated by the fact that assembly rates in the presence of ATP were higher and the kinetics was greater than first order (nonlinear curue in Fig. 6), just as it is when accessory proteins are present. Therefore, ATP enhances nucleation, in addition to reducing the critical concentration. We demonstrated that ATP interacts directly with tubulin and is responsible for changes in the assembly characteristics. Although the effects of ATP were clear, the mechanism by which ATP influences polymerization remained to be studied.
The first question to be addressed was whether ATP hydrolysis is involved. Consistent with the requirement for hydrolysis was the finding that the nonhydrolyzable ATP analog, AMP-PNP, had no effect on the assembly kinetics. Also, this analog would not induce the formation of aggregated rings as did ATP. While these results would be necessary to demonstrate ATP hydrolysis, they are not sufficient to prove the case. Therefore, we looked for the release of inorganic phosphate using [y3'P]ATP. No hydrolysis was detected over the time necessary for microtubule formation to reach steady state.' The fact that AMP-PNP was unable to substitute for ATP can be reconciled to the lack of hydrolysis if ATP is bound with a precise and crucial molecular fit and the difference in structure of the analog prevents such binding.
For some time, tubulin rings have been postulated to be nucleation centers for microtubule formation. Indeed, aggregated tubulin rings induced with ATP proved capable of initiating assembly in a solution of tubulin which otherwise would not have polymerized. A crucial point to be re-emphasized is that while the Mg2' concentration was 5.0 mM during the formation of the ring aggregates, it was diluted to 2.5 m~ when the aliquot of aggregated rings was added to the tubulin solution. Microtubule formation resulted from this addition. When the Mg'+ concentration was maintained at 5.0 mM in parallel experiments, assembly was aborted by the aggregation of the rings. Fig. 8 presents a scheme for the assembly of tubulin into microtubules incorporating the interaction of ATP at a third nucleotide binding site. We also illustrate some aspects of the role of magnesium in tubulin polymer formation. Since little information is available on the role of the N-site, GTP was simply placed at this site in every case. When GDP is on the E-site, no polymer formation can occur. With GTP at the E- Mg2+, warming to 37 "C results in the formation of aggregated tubulin rings (18). Tubulin is apparently locked in this polymeric form and prevented from entering the next stage in the formation of microtubules. When the magnesium concentration is lowered, these aggregates break down and the individual rings become nucleation centers for microtubule assembly. With the same nucleotide configuration but a t lower M P concentration (2 m), individual rings are formed and without the hindrance of aggregation they initiate microtubule formation with subsequent E-site GTP hydrolysis. The last line of Fig. 8 indicates that there is not an absolute requirement for ATP, at least not under the conditions used here. In the absence of ATP, microtubule formation proceeds through the formation of rings as intermediates with E-site GTP hydrolysis, but with ATP bound to the tubulin, ring formation and microtubule assembly are speeded and the incorporation of tubulin into microtubules is more extensive.