Proc. Nat!. Acad. Sci. USAVol. 78, No. 3, pp. 1586-1590, March 1981Biochemistry

Role of GTP hydrolysis in microtubule treadmilling and assembly(tubulin/polymerization/podophyllotoxin)

ROBERT L. MARGOLISThe Fred Hutchinson Cancer Research Center, 1124 Columbia Street, Seattle, Washington 98104

Communicated by Katherine Esau, December 10, 1980

ABSTRACT GTP hydrolysis accompanies addition of tubulinto microtubules. I find that hydrolysis is a requirement for theopposite-end assembly/disassembly of microtubules and conse-quent subunit treadmilling from one end to the other of the poly-mer. Neither GDP nor guanosine 5'-[fi,y-imido]triphosphate al-lows or participates in the treadmilling reaction. Therefore, thereis a requirement for hydrolysis in the addition of subunits to thefavored assembly end of the microtubule. Podophyllotoxin, anassembly inhibitory drug, "caps" the microtubule assembly end,preventing subunit loss from that site to equilibrium. Continuedhydrolysis of GTP is required to maintain the podophyllotoxin cap.A corollary of this finding is that GTP hydrolysis is required forcap formation. Microtubules assembled in GTP enter a metastablestate when all remaining GTP is hydrolyzed. This state is char-acterized by its ability to maintain indefinitely a subunit/polymerdistribution ratio that is arbitrary and that can be altered at willby brief chilling or by addition of small amounts of GTP. Thismetastable state is labile to podophyllotoxin. Use of podophyllo-toxin allows measurement of the microtubule treadmilling rate;use of podophyllotoxin in the absence of GTP allows measurementof the overall rate of dimer dissociation from the microtubule.Measurement of these rates has permitted determination of theefficiency with which adding dimers incorporate into the micro-tubule treadmill and are not lost to assembly end equilibrium. Theefficiency varies with GTP concentration for unknown reasons,being high at 0.1 mM GTP and low at higher GTP concentrations.

GTP binds rapidly and reversably to one exchangeable bindingsite (E-site) per tubulin subunit (1). During microtubule assem-bly, the E-site GTP undergoes a hydrolysis step as the subunitadds (2-5); and the GDP that is formed remains tightly bound,nearly one per dimer, in the polymer (3, 4, 6).

Because nucleotides that are incapable of 13-y hydrolysis alsosustain microtubule assembly [guanosine 5'-[,B, y-imido]tri-phosphate (p[NH]ppG), guanosine 5'-[,8,y-methylene]tri-phosphate (p[CH2]ppG) (7-10) and, reportedly, GDP (10)],the necessity for a hydrolysis step during assembly has beenunclear.

Microtubules undergo a constant opposite-end assembly/disassembly mechanism in vitro (6). At an apparent equilibriumstate for the polymer, subunits add at one end [the net assemblyor (A) end] at a rate that is greater than their rate of loss at thisend. As a consequence, the constant shuttling of subunits fromone end to the other of a microtubule has been designated a"treadmilling" mechanism (11). A similar treadmilling mecha-nism occurs in actin polymers (12, 13). It has been suggestedthat ATP hydrolysis sustains the (A) end steady-state or tread-milling behavior in actin polymers (12).

I report here that microtubules do not treadmill in the ab-sence of GTP hydrolysis, although net assembly will occur inthe presence of p[NH]ppG and a metastable state can be sus-tained in the presence of GDP.

Furthermore, by examining the state of microtubules in thepresence of GDP only, I find that hydrolysis creates stability.

When microtubules are assembled with GTP and the GTP insolution is subsequently hydrolyzed, the microtubules maintaina metastable state characterized by a stable polymer-subunitdistribution ratio that can nonetheless be shifted about at will.The GDP metastable state is labile to podophyllotoxin (PLN).Conversely, a PLN "cap" which prevents subunit addition orloss at the microtubule (A) end (6) is sustained by continued GTPhydrolysis.PLN creates an (A) end cap, and complete hydrolysis of GTP

removes the cap and allows free dissociation of dimers at bothmicrotubule ends. I have used these observations to measurethe efficiency with which dimers are incorporated into the mi-crotubule in a treadmilling manner and are not lost to equilib-rium at the assembly end. I find that treadmilling efficiency canbe quite high under proper conditions in vitro. This findingincreases the probability that microtubule treadmilling hasphysiological meaning in vivo.

MATERIALS AND METHODSPreparation of Microtubule Protein. Microtubule protein

was prepared from bovine brain by three cycles of temperature-dependent polymerization and depolymerization, essentially bythe method of Borisy et al. (14) as modified by Asnes and Wilson(15). Purification was carried out in 20 mM sodium phosphate/100 mM sodium glutamate/1.0 mM MgSO1.0 mM ethyleneglycol bis[,B-aminoethyl ether]-N,N,N',N'-tetraacetic acid(EGTA), 2.5 mM GTP, pH 6.75 (PG buffer). All chemicals, ex-cept when indicated otherwise, were from Sigma. After thethird cycle of polymerization, the preparation was cleaned bycentrifugation of microtubules through 50% (wt/vol) sucrosein PG buffer (200,000 X g, 2.0 hr, Beckman 50 Ti rotor, 30°C).Final pellets were stored in liquid nitrogen until used. Thepellets used in these experiments contained 75% tubulin and25% microtubule-associated proteins (MAPs) (16).

Assembly Assays. Microtubule pellets were solubilized in100 mM 2'-[N-morpholino]ethanesulfonic acid/1.0 mMMgSOJ1.0 mM EGTA, pH 6.75 (MEM buffer) containing 0.1mM GTP, and the product was centrifuged at 30,000 X g for10 min. Microtubule assembly at 30°C was followed by mea-surement of turbidity development at 350 nm in a Cary record-ing spectrophotometer (Varian model 219) with a constant-tem-perature cuvette chamber. Protein concentrations, determinedby Lowry assay (17) with a bovine serum albumin standard,were usually 2.0 mg/ml.

For experiments in which the effect of GDP on microtubuleassembly was characterized, 10 mM fructose 6-phosphate wasincluded in the buffer prior to assembly. Phosphofructokinase

Abbreviations: E-site, exchangable GTP binding site on tubulin; (A)end, net assembly end of microtubule; (D) end, net disassembly end;p[NH]ppG, guanosine 5'-[3, y-imido]triphosphate; p[CH2]ppG, guan-osine 5'-[P,y-methylene]triphosphate; MAPs, microtubule associatedproteins; PLN, podophyllotoxin; EGTA, ethylene glycol bis(f3-amino-ethyl ether)-N,NN',N',-tetraacetic acid.

1586

The publication costs of this article were defrayed in part by page chargepayment. This article must therefore be hereby marked "advertise-ment" in accordance with 18 U. S. C. §1734 solely to indicate this fact.

Proc. Natl. Acad. Sci. USA 78 (1981) 1587

(rabbit muscle, Sigma, type III) was added, from the supplier'sstock solution, to a final concentration of 0.3 unit/ml at theappropriate times in these experiments. PLN (Aldrich) wasadded when indicated from a X 100 stock solution in dimethylsulfoxide. Final PLN concentration was 50 AuM, sufficient toblock in vitro assembly of microtubules totally.

Experiments with [3H]GTP (ICN; 17.1 Ci/mmol; 1 Ci = 3.7x 10 becquerels) as a marker for microtubule kinetics atsteady-state were performed as reported (6).

RESULTSHydrolysis Is Necessary for Treadmilling. Hydrolysis of

GTP accompanies subunit addition to microtubules. However,assembly of microtubules is not strictly dependent on GTP hy-drolysis, and nonhydrolyzable analogs of GTP, GMPPNP and(GMPPCP) will support the formation of polymers apparentlyindistinguishable in structure from those formed by GTP.A steady-state behavior of microtubules which involves the

net addition of subunits at one end and the net removal of sub-units at the other end of the polymer when it is in a state ofoverall apparent equilibrium has been reported (6). This unusu-al behavior of microtubules constitutes an opposite end assem-bly/disassembly mechanism. Accompanying this process is aconstant unidirectional shuttling of subunits from the assemblyend to the disassembly end, or treadmilling. A similar processhas been observed in actin polymers. Wegner (12) offered thepossibility that the treadmilling behavior of F-actin was drivenby the continuous ATP hydrolysis that accompanies assembly.To determine whether the treadmilling behavior of micro-

tubules is dependent on continuous GTP hydrolysis, I measuredthe ability of dimers to treadmill if they add to a microtubulewith either GDP or p[NH]ppG, and therefore without accom-panying hydrolysis. To make this measurement I took advantageof the fact that the tubulin subunit readily exchanges GTP atan E-site (1) but that, once in the polymer, the guanine nu-cleotide (now GDP) is nonexchangeable (3, 4, 6). Subunit entryand exit from microtubules at steady-state (by determining therate of uptake and loss of labeled guanine nucleotides) and,therefore, the rate of uptake and loss of the subunits to whichthey are bound have been measured (6).A typical result for microtubules assembled in GTP then

pulsed with [3H]GTP at steady state is shown in Fig. 1. Theincrement of label in the polymer is linear with time. On thebasis of other supporting evidence (6), this increment has beendetermined to represent a continuous treadmilling flow of sub-units into the microtubule from the assembly end. Howeverwhen microtubules are assembled in GTP and, then at steady-state, are pulsed in [3H]p[NH]ppG, there is no incorporationof this label into the polymer with time. Subunits that add withp[NH]ppG appear not to be participating in the treadmillingprocess.

It can be argued that the affinity of the subunit forp[NH]ppGis so weak relative to its affinity for GTP that p[NH]ppG sub-units add too infrequently to the polymer to be detected astreadmilling. However, when microtubules are assembled inp[NH]ppG and then pulsed with [3H]p[NH]ppG once an as-sembly plateau (as determined by turbidity measurement) hasbeen attained, [3H]p[NH]ppG does not add to microtubules ina linear treadmilling manner under these circumstances either(Fig. 1).To determine whether GDP could support the treadmilling

reaction, microtubules were assembled to steady state with 0.1mM GTP and, once the maximal extent of assembly had beenreached, all remaining GTP was hydrolyzed to GDP by using10 mM fructose 6-phosphate and rabbit muscle phosphofruc-tokinase. [This enzyme utilizes GTP with almost the same fa-

Time, min

FIG. 1. Effect of GTP hydrolysis on treadmilling. Bovine brain re-cycled tubulin was assembled to apparentequilibrium in 0.1 mM GTPor in 1.4 mM p[NHIppG in MEM buffer. At apparent equilibrium, theability of subunits to treadmill into-the polymer was assayed by ex-posure to a pulse of 3H-labeled nucleotide for different lengths of time.Finally, label incorporation and protein content were determined formicrotubules that were centrifuged through sucrose cushions at thetermination of the pulse period. For the case of microtubules in GDPand pulsed with [3H]GDP, microtubules were first assembled with GTP.and residual triphosphate was removed enzymatically by addition ofphosphofructokinase (0.3 ,ul/ml) to a microtubule solution that hadbeen assembled with 10 mM fructose 6-phosphate present. The pulseperiod began once a new apparent equilibrium had been established.Tubulin isolated by centrifugation of microtubules through 50% su-crose prior to the experiment assembled well in p[NH~ppG althoughat increased nucleotide concentration (1.4 mM p[NH]ppG. Final spe-cific activity of all label added was 0.125 mCi/mmol. Nonspecific labelbackgrounds have been subtracted for all experiments. -.*, Micro-tubules in GTP, pulsed with [3HIGTP; A---A, microtubules in GTP,pulsed with [3H]p[NH]ppG; o----o, microtubules in GDP, pulsed with[3H]GDP; o---o, microtubules in p[NH]ppG, pulsed with[3Hlp[NH]ppG.

cility as its usual substrate, ATP (18).] Complete hydrolysis ofGTP was accompanied by a partial disassembly, after which astable plateau, as observed by spectrophotometer assay, wasreached (see Fig. 5). Once this GDP plateau had been reached[3H]GDP was pulsed into microtubule protein aliquots for in-creasing periods of time, and the protein was then treated thesame as in the [3H]GTP pulsing experiments. There was notreadmilling incorporation of GDP into microtubules (Fig. 1).This result, again, is consistent with hydrolysis as a requirementfor treadmilling.

Nature of the PLN Cap at the Microtubule Assembly End.It has been reported (6) that, when added to microtubules atsteady state, PLN prevents all further assembly reaction andproduces a rate of microtubule subunit loss that is equal to thenet disassembly end [(D) end] loss of subunits at steady stateas measured in [3H]GTP pulse-chase experiments. Eitherthere is little equilibrium loss of subunits at the (A) end or PLNuniquely stabilizes the (A) end, producing a cap that preventsall further subunit addition and loss at this site (6).The possibility that PLN produces a cap is supported by the

following experiment. Microtubules were assembled to steadystate in 0.1 mM GTP and then pulsed with [3H]GTP for 1 hr.This pulse labels only 7-10% of the microtubule, with all thelabel adding at the (A) end (6). In a chase, therefore, one is ob-serving only what happens to (A) end subunits. After the pulse,50 ,uM PLN was added, along with 2.5 mM GTP for chase, andthe fate of the label was followed during increasing times inPLN. After an initial loss of some subunits at the (A) end, a sta-ble plateau was attained (Fig. 2). It would seem that a cap does

Biochemistry: Margolis

Proc. NatL Acad. Sci. USA 78 (1981)

a)

a)

U.

100

80

60

40

20[

0 15 30 45Time, min

60 75 90

FIG. 2. Stability of an (A) end pulse to PLN. Recycled microtubuleprotein was assembled to steady state with 0.1 mM GTP and 10 mMfructose .6-phosphate in MEM buffer. At steady state the system waspulsed with [3HIGTP (0.25 mCi/mmol) for 1 hr, in either 10 mM acetylphosphate (GTP chase) or 2 mM acetyl phosphate (GDP chase) andacetate kinase (0.1 unit/ml). The pulse was terminated by.addition of50 uM PLN, and samples were chased for increasing periods of timein PLN. Fifteen minutes after addition of PLN, the samples that wereGDPchased received phosphofructokinase (0.3 unit/ml). Samples wereassayed at 15-min intervals after enzyme addition. o---o, .GTP chase;.* -, GDP chase.

indeed form at an unknown time within approximately 15 min;before that time, one can observe an equilibrium loss of sub-units from the (A) end in the absence of further assembly.

Furthermore, the cap that forms is maintained by continuingGTP hydrolysis, or at least by the continued presence of thetriphosphate. When microtubules were treated as above butexposed to phosphofructokinase and fructose 6-phosphate at 15min after exposure to PLN, the GTP was totally hydrolyzed andthe label was not retained in a stable plateau but rather contin-ued to be lost in a relatively linear manner (Fig. 2).

Measurement of the Efficiency of the Treadmilling Reac-tion. The ability of PLN to form an (A) end cap and of completehydrolysis of GTP to destroy the cap can be exploited to mea-sure the relative number of subunits that add at the (A) end andare not lost to equilibrium but continue to treadmill through themicrotubule. An expression that describes the efficiency of (A)end subunit addition, in terms of the subunit on and off rateconstants at the (A) and (D) ends of a polymer that is tread-milling, has been derived by Wegner (12):

S = [(kil * c,) - k'l]/(k'l + k22) [1]

in which k1l is the rate constant for (A) end subunit addition,cC is the critical subunit concentration for assembly, k21 is therate constant for (A) end subunit loss, and k' is the rate constantfor (D) end subunit loss. S is the fraction of a population of sub-units that binds at the (A) end and is not lost to equilibrium butproceeds unidirectionally through a polymer. When S = 1.0,treadmilling is completely efficient.

In order to determine S experimentally, one need not de-termine each of the individual rate constants. The numeratorin Eq. 1 expresses the net rate of subunit addition at the (A) end,which is identical to the rate of treadmilling. In turn, for thespecial case of beef brain microtubules, the rate of treadmillingis equal to the rate of PLN-induced disassembly (6, 19). Thus,the numerator can be quickly determined by measurement ofthe rate of mnicrotubule disassembly in the presence of 50 uMPLN. Two physical phenomena must account for the ability ofPLN to yield the treadmilling rate. First, beef brain microtu-bules appear to have a negligible rate of subunit addition at the(D) end at steady-state [which is not the case with sea urchin

flagellar outer doublet tubulin reassembly (20)], making the rateof loss at the (D) end equal to the treadmilling rate. Second,use of PLN allows one to measure the (D) end loss rate inde-pendently of the (A) end reactions because the (A) end is cappedby PLN (Fig. 2). The denominator in Eq. 1' expresses the sumof the rate constants for subunit loss at both the (A) and (D) endsof the microtubule. In order to obtain this value, one may takeadvantage of the fact that the PLN (A)- end cap is destroyed bythe complete hydrolysis of GTP in phosphofructokinase (Fig.2). In the presence of PLN and GDP, one may use turbidityto measure the sum of the rate constants for the loss of subunitsat the (A) and (D) ends. Under these conditions the rate con-stants for subunit addition to the microtubule are negligiblebecause the measured disassembly rate remains apparentlyfirst-order until depolymerization is complete (not shown).

The S ratio may thus be quickly determined, under variousconditions, by assembling microtubules to steady state in abuffer containing 10 mM fructose 6-phosphate. At steady state,50 tLM PLN is added and the treadmilling rate is determined;then phosphofructokinase is added and the sum of (A) and (D)end disassembly rates is determined. Fig. 3 shows a typicalexperimental result and values of S are given in the legend.The S ratio is quite sensitive to the concentration of GTP in

vitro (Fig. 4). When microtubules are assembled in 0.1 mMGTP, S is quite high (0.8), indicating highly efficient tread-milling of subunits. At higher GTP concentrations, S becomesprogressively lower.

There are two points to stress. First S is highly variable asconditions of assembly are altered in vitro, as are, presumably,the rate constants which S reflects. Second, treadmilling canbe shown to be a highly efficient process, given the proper invitro circumstances.

In GDP Microtubules Exist in a Metastable State. In a mi-crotubule solution, once GTP is completely hydrolyzed, it isapparent that two events may happen: treadmilling of subunitsceases (Fig. 1), and any preexisting (A} end PLN cap is lost (Fig.2). I examined the fate of microtubules in GDP further and

0.25

0.20 I fA B 2

0.05m_AAA_

0 10 20 30 40 50 60Time, min

70 80 90

FIG. 3. Effect of guanine nucleotide concentration on the micro-tubule disassembly rate. Microtubules were assembled at 30°C fromrecycled microtubule protein in MEM buffer containing 10 mM fruc-tose 6-phosphate and either 0.3 mM (curves 1 and 3) or 1.0mM (curves2 and 4) GTP. At 30 min, once assembly reached plateau, 50 JuM PLNwas added (arrow A) to all samples. At 45 min, PFK was added to sam-ples 3 and 4 to hydrolyze all remaining GTP (arrow B). The rates ofloss for all samples were determined by- tangents to the slope at times>70 min. The PLN rate of loss (curves 1 and 2) was 11%/hr. The 0.3mM GDP/PLN rate of loss (curve 3) was 26%/hr. The 1.0 mM GDP/PLN rate of loss (curve 4) was 57%/hr. From these rates, S values of0.42 (0.3mM GTP) and 0.19 (1.0mMGTP) were obtained. , 1.0mMGTP during assembly; ---, 0.3 mM GTP during assembly.

1588 Biochemistry: Margolis

Proc. Nati Acad. Sci. USA 78 (1981) 1589

0.8

0.6

s0.S0.4

0.21

0 1 1U51/(GTP, mMI)

FIG. 4. Change in S with GTP concentration. The data representthe collected results of a number of experiments conducted to deter-mine the change in S as GTP levels are changed. All data are from asingle brain preparation.

found that a curious new state referred to as a metastable state,is created.

Microtubules were assembled in 0.1 mM GTP in MEMbuffer containing 10 mM fructose 6-phosphate. Curve 1 in Fig.5A shows a typical GTP assembly curve. Curve 2 shows theresult of adding phosphofructokinase (0.3 unit/ml) to the as-

A0.2

0.1

o0.15-

Mv~~~~~~~~2

0.0

0 24 48 72 96

0.2.-Is; t

00.15 X =

°0.1./\0.05

0 24 48 72 96Time, min

FIG. 5. The microtubule metastable state in GDP. Recycled mi-crotubule protein was assembled in MEM buffer containing 0.1 mMGTP and 10.0 mM fructose 6-phosphate. Assembly was measured bychange in turbidity at 350 nm. (A) Two samples were assembled. Onesample (curve 2) received phosphofructokinase (0.3 unit/ml) duringassembly; then it was chilled to 000 for 1 min and rapidly rewarmed;and finally it was given an additional 0.1 mM GTP. (B) Three sampleswere assembled. One sample (curve 3) received phosphofructokinasewhen approximately half assembled. Another sample (curve 2) re-ceived phosphofructokinase after steady state was reached; after itsnew metastable state was attained, it received 50 MM PLN.

sembling microtubules. Assembly ceased almost immediatelyas. all available GTP was quickly hydrolyzed to GDP and a newapparent equilibrium was reached. If this state represents atrue microtubule-subunit equilibrium, then one should be ableto perturb.this particular polymer-subunit distribution and ob-serve a return to the original state. However, any disturbanceof this state created a unique new "equilibrium. " The-polymer-subunit distribution could be shifted about at will, and each newstate obtained was quite stable-e.g., by chilling the micro-tubules in GDP for 1 min at 00C and then rewarming the so-lution, or by adding 0.1 mM GTP to this solution (new assemblywas induced until the 'GTP was again depleted).

This result was completely reproducible regardless of theorder of chilling or GTP addition or level of nucleotide present.In this set of conditions, there is no true equilibrium in GDPbut rather a stable polymer state that nonetheless can be shiftedto higher or lower stable states by simple manipulation. It mustbe emphasized that these microtubules do not assemble inGDP, nor are they capable of further assembly in GDP fromany preexisting assembly state. This result is consistent with theresults of Weisenberg et al. (3) suggesting a microtubule non-equilibrium in GDP.

Fig. 5B shows the result of addition of phosphofructokinaseat different stages in assembly. Again, two different metastablestates were attained, depending only on the extent of polymerformation prior to the addition of phosphofructokinase. Al-though the state of microtubules in GDP is not an apparent trueequilibrium, the state is labile to PLN which can be expectedto bind to free subunits and interfere with their ability to in-teract with the microtubule ends. Thus, the effect of PLN incombination with GDP is the same regardless of their order ofaddition. GDP disrupts the PLN cap, and PLN disrupts theGDP metastable state.

DISCUSSION

Although hydrolysis is not an absolute requirement for micro-tubule assembly, it normally accompanies assembly. I have pre-sented evidence that hydrolysis enhances subunit addition tothe favored assembly end of the microtubule and that it is re-quired for the microtubule's treadmilling behavior. Others havesuggested previously that one role of hydrolysis may be to drivethe assembly end steady-state that produces treadmilling in F-actin (12) and in microtubules (21). The result here clearly ful-fills the prediction of Wegner's hypothesis on the role of hy-drolysis in actin (and microtubule) assembly (12).GDP does not support microtubule assembly under the con-

ditions used here. Microtubules that have assembled in GTP,however, are remarkably stable in GDP. This stable state is notan equilibrium because, when disturbed, the polymer-subunitdistribution does not return to its former state. I have charac-terized this state as a metastable state. How this state is main-tained is not understood. There are some considerations thatmight help in eventually establishing a mechanism. Loss of sub-units from the polymer only occurs when the dimers occupy thevery end of the microtubule (6, 22, 23). Continual subunit lossfrom the (D) end normally accompanies the GTP steady-state(6), yet this loss is blocked in GDP. Recently, Sandoval andWeber (24) have reported that guanosine [a,3]methylene tri-phosphate blocks treadmilling in GTP assembled microtubules.The specific blockage appears to occur at the (D) end. Perhapsmy result, showing blockage of (D) end loss in GDP, is mecha-nistically similar to their finding. Finally, by a mechanism thatis not clear, PLN is able to terminate the GDP-tubulin meta-stable state and permit the disassembly of microtubules in aGDP environment.

Biochemistry: Margolis

0 -1"

1.. 0

a

1.0.

Proc. Natl. Acad. Sci. USA 78 (1981)

PLN produces a cap on the microtubule (A) end-that preventsfurther dimer addition or loss at this site. This capping behavioris the apparent cause of substoichiometric poisoning of micro-tubule assembly by PLN (6). There are two points to make asa result of the work presented here. First, the cap forms onlymany minutes after PLN addition, and GTP hydrolysis is nec-essary to form the cap. Before this time, (A) end subunits arelost to the solution (the PLN concentrations used here were inmolar excess to tubulin). Second, continuing presence of GTPis required to maintain a PLN cap once formed. It is possiblethat subunits add with GTP hydrolysis, either adjacent to oratop PLN-bearing dimers at the (A) end, and stabilize the PLN-bound dimers in the microtubule. However, the PLN has al-tered either the bond stability or the structure of the polymerthat forms beyond this point, or both, so that further assemblyis not possible. This result may explain why tubulin and micro-tubule GTP hydrolysis rates are insensitive to the presence ofassembly-blocking drugs (25).

It would appear, then, that the native mechanism of (A) endmicrotubule assembly augments the potency of microtubulepoisons in the living cell, making them effective at very lowconcentrations.

Recently, some research groups have confirmed that micro-tubule treadmilling occurs, but with low efficiency, and havecalled into question whether it represents a viable cellularmechanism (21, 26). By using the techniques presented in thisarticle, I was able to measure the relative efficiency with whichsubunits that add to the microtubule (A) end contribute to tread-milling. Under certain circumstances (low GTP concentration),treadmilling efficiencies were quite high. Under the conditionsthat Karr and'Purich (26) and Bergen and Borisy (21) used (highGTP concentration), I find a low S ratio that is close to what theyreported.

There is an initial rapid (A) end loss of tubulin from PLN-poisoned microtubules, both in GTP and in GDP. This rateexceeds the rate of PLN/GDP disassembly observed at latertimes. Initial measurements (unpublished data) indicate thatthere is a correspondingly higher rate of [3HI]GTP incorporationat the (A) end during the first 10 min of pulse at steady state.I believe that the initial rapid loss is reciprocal to the initial rapidsubunit gain, representing a transition zone of unstable dimer-dimer bonds close to the microtubule (A) end. If this is true,it may be possible that this unstable transition zone must be lostbefore a PLN cap is formed on more stable dimers. Such a pos-sibility could explain the lag before a PLN cap is formed.

In summary, I find that GTP hydrolysis serves two functionsin microtubule assembly: (i) permitting a dimer-dimer bondstrengthening; and (ii) permitting subunit treadmilling to occur.I find that PLN (and perhaps other microtubule inhibitors) actsin two ways: (i) as a substoichiometric (A) end cap maintainedby continuing GTP hydrolysis; and (ii) as an inhibitor (possibly

stoichiometric) of the GDP-dependent "equilibrium." Also, Ifind that there is a GDP metastable state of microtubules. PLNappears to abrogate the GDP metastable state by making GDPsubunit addition unfavored. Last, I find that the efficiency ofmicrotubule treadmilling can be made to vary greatly in vitroby simple manipulations of GTP concentration and that tread-milling efficiency can be quite high. The fact that GTP can beexpended economically to produce treadmilling argues that thetreadmilling mechanism may be an important reason far theevolutionary design that causes GTP hydrolysis to accompanymicrotubule assembly normally.

I thank Charles T. Rauch for excellent technical assistance. This workwas supported by start-up funds from the Fred Hutchinson CancerResearch Center and by U.S. Public Health Service Grant GM 28189.

1. Weisenberg, R. C., Borisy, G. G. & Taylor, E. W. (1968) Bio-chemistry 7, 4466-4479.

2. Kobayashi, T. (1975)J. Biochem. (Tokyo) 77, 1193-1197.3. Weisenberg, R. C., Deery, W. J. & Dickinson, P. J. (1976) Bio-

chemistry 15, 4248-4254.4. Arai, T. & Kaziro, Y. (1977) J. Biochem. (Tokyo) 82, 1063-1071.5. MacNeal, R. K. & Purich, D. L. (1978) FEBS Lett. 96, 83-86.6. Margolis, R. L. & Wilson, L. (1978) Cell 13, 1-8.7. Arai, T. & Kaziro, Y. (1976) Biochem. Biophys. Res. Commun. 69,

369-376.8. Penningroth, S. M. & Kirschner, M. W. (1977)J. Mol. Biol. 115,

643-673.9. Weisenberg, R. C. & Deery, W. J. (1976) Nature (London) 263,

792.10. Karr, T. L., Podrasky, A. E. & Purich, D. L. (1979) Proc. Natl

Acad. Sci. USA 76, 5475-5479.11. Wilson, L. & Margolis, R. L. (1978) in Cell Reproduction, ICN-

UCLA Symposia on Molecular and Cellular Biology, eds. Dirk-sen, E. R. Prescott, D. M. & Fox, C. R. (Academic, New York),Vol. 12, pp. 241-258.

12. Wegner, A. (1976)1. Mol. Biol. 108, 139-150.13. Brenner, S. L. & Korn, E. D. (1979)J. Biol Chem. 254, 9982-9985.14. Borisy, G. G., Marcum, J. M., Olmsted, J. B>, Murphy, D. B.

& Johnson, K. A. (1975) Ann. N.Y. Acad. Sci. 253, 107-132.15. Asnes, C. R. & Wilson, L. (1979) AnaL Biochem. 98, 64-73.16. Sloboda, R. D., Rudolph, S. A., Rosenbaum, J. L. & Greengard,

P. (1975) Proc. Natl. Acad. Sci. USA 72, 177-181.17. Lowry, 0. G., Rosebrough, N. J., Farr, A. L. & Randall, R. J.

(1951)J. Biol. Chem. 193, 265-275.18. Barman, T. E. (1969) Enzyme Handbook (Springer, New York).19. Margolis, R. L. & Wilson, L. (1979) Cell 18, 673-679.20. Farrell, K. W., Kassis, J. A. & Wilson, L. (1979) Biochemistry 12,

2642-2647.21. Bergen, L. G. & Borisy, G. G. (1980)1. Cell Biol 84, 141-150.22. Johnson, K. A., & Borisy, G. G. (1975) in Molecules and Cell

Movement, eds. Inoue, S. & Stephens, R. E. (Raven, New York),pp. 119-141.

23. Bryan, J. (1976) J. Cell Biol 71, 749-767.24. Sandoval, I. V. & Weber, K. (1980)J. Biol. Chem. 255, 6966-6974.25. Maccioni, R. & Seeds, N. W. (1977) Proc. Natl Acad. Sci. USA 74,

462-466.26. Karr, T. L. & Purich, D. L. (1979)j. Biol Chem. 254,10885-10888.

1590 Biochemistry: Margolis