Tubulin Exchanges Divalent Cations at Both Guanine Nucleotide-binding Sites*

The tubulin heterodimer binds a molecule of GTP at the nonexchangeable nucleotide-binding site (N-site) and either GDP or GTP at the exchangeable nucleo- tide-binding site (E-site). Mg2* is known to be tightly linked to the binding of GTP at the E-site (Correia, J. J., Baty, L. T., and Williams, R. C., Jr. (1987) J. Biol. Chem. 262, 17278-17284). Measurements of the ex- change of Mn2+ for bound M 8 + (as monitored by atomic absorption and EPR) demonstrate that tubulin which has GDP at the E-site possesses one high affinity metal-binding site and that tubulin which has GTP at the E- site possesses two such sites. The apparent association constants are 0.7-1.1 X lo6 M" for M 8 + and -4.1-4.9 X lo7 M" for Mn2+. Divalent cations do bind to GDP at the E-site, but with much lower affinity (2.0-2.3 x 10' M" for M S + and 3.9-6.6 X 10' M-' for Mn2+). These data suggest that divalent cations are involved in GTP binding to both the N- and E-sites of tubulin. The N-site metal exchanges slowly (kaPp = 0.020 min"), suggesting a mechanism involving protein "breathing" or heterodimer dissociation. The

with known nucleotide-binding proteins (2). This difference in M$+ linkage may be due to the presence of two different conformations of tubulin allosterically induced by the binding of GDP or GTP at the E-site (2, 3). Other divalent cations are known to exchange for bound M$+ on tubulin and to promote tubulin polymerization. Mn2+ (4, 5), Coz+ (6, 7), and Zn2+ (6, 8,9) are believed to compete for the same high affinity site, although Co2+ and Zn2+ induce aberrant assembly of tubulin into sheets of protofilaments. The binding of Mn2+ to tubulin has been described as occurring at two classes of noninteracting sites: one high affinity site (Kl = 6.3 X lo6 to 4.0 X lo6 M-') and eight weak sites (Kz = 2.2-2.6 X lo3 M-') (4, 5 ) . The fractional number of high affinity sites per heterodimer varies with the nucleotide content of tubulin, suggesting that bound nucleotide is part of the site. GTPCr also binds to the E-site and promotes microtubule assembly (6, 10). Recently, a tight metal-binding site has been shown to be within 6-8 A of the y-phosphate of E-site GTP (11). These results strongly suggest that guanine nucleotides bind to the E-site of tubulin as a metal complex.
Here we have reinvestigated the exchange of Mn2+ for bound M P on tubulin and show that the number of high affinity ( K > lo6 M-') metal sites is correlated with the total GTP content of tubulin and not with the GDP content. We have found that the E-site exchanges the two cations rapidly, binding Mn2+ strongly when it is occupied by GTP and weakly when occupied by GDP. A second high affinity site, most likely the GTP-occupied N-site, binds divalent cations strongly, but exchanges them very slowly (tH = 35 min). Thus, the binding of divalent cations to tubulin is best described as occurring at four possible classes of noninteracting sites. It involves high affinity coordination with GTP at the E-site and with GTP at the N-site, as well as low affinity coordination with GDP at the E-site and with other weak binding sites most likely associated with the acidic carboxyl-terminal region of both subunits. These results have important consequences for the detailed mechanism of divalent cation release from microtubules after GTP hydrolysis.

MATERIALS AND METHODS
Reagents-Pipes, EGTA, dithioerythritol, GDP (Type I), and GTP (Type 11-S) were from Sigma. MgSO, and MnC12 were ACS reagentgrade from Fisher. Magnesium and manganese atomic absorption standard solutions were from Sigma.
Preparation of Tubulin-Microtubule protein was purified from bovine brain by three assembly/disassembly cycles according to the method of Shelanski et al. (12), with modifications described by Williams and Lee (13). ATP (to 2.5 mM) was added to the supernatant of the first high speed centrifugation (14) to increase the yield. Tubulin was separated from microtubule-associated proteins by chromatography on Whatman P-11 phosphocellulose in 0.1 M Pipes, 2 mM EGTA, 2 mM dithioerythritol, 1 mM MgSO,, pH 6.9, plus 0.1 mM GDP (1, 15); MgS04 (to 1 mM) was added to fractions of the eluate that contained tubulin (16). Each preparation was checked for purity 10682 Exchange of Divalent Cations by Tubulin by sodium dodecyl sulfate gel electrophoresis on heavily overloaded gels, frozen dropwise in liquid nitrogen, stored at -75 "C, and used within 3 months. The GDP content of this tubulin is typically 46-49% of the total bound nucleotide, and this protein is designated TBGDP, i.e. tubulin with GDP bound at the E-site. To make tubulin with predominant GTP at the E-site (TBGTP), a solution of tubulin purified by phosphocellulose chromatography (PC-tubulin) was brought to 1 mM GTP and 1 mM MgS04, incubated for 5 min, and column-centrifuged (17) into 0.1 M Pipes, 1 mM MgS04, pH 6.9, and the process was repeated. The tubulin was exposed to 1 mM GTP and MgSO4 a third time and column-centrifuged into 0.1 M Pipes, pH 6.9.
All steps were performed at 0-4 "C. The bound nucleotide content of this material is typically 5-10% GDP, or 0.1-0.2 molecules of GDP/ E-site. Prior to an experiment, TBGDP or TBGTP in Pipes buffer was diluted to the desired experimental concentration, and other buffer components (dithioerythritol, Mn2+, M e , glycerol, etc.) were added.
Determination of Concentrations-The concentrations of M$+ and Mn2+ in stocks and buffers and tightly bound to tubulin were determined by atomic absorption on a Perkin-Elmer Model 603. The concentration of tubulin was determined from its absorbance: €278.5 = 1.23 ml mg" cm" (18). The molecular weight of the dimer was taken to be 100,000 (19-21). The number of moles of guanine nucleotide bound to 1 mol of tubulin dimer (denoted GXP/TB) was determined as described by Croom et al. (22). Tubulin was freed of unbound nucleotide (and weakly bound metal) by column centrifugation (17); the protein was precipitated with perchloric acid at 0 "C; and the amount of guanine nucleotide was determined from the absorbance of the supernatant: €256 = 1.24 X lo' M" cm", applicable near pH 1.
Assessment of Purity of GTP and GDP-Guanine nucleotides were separated on a SynChropak AX-300 anion-exchange column (Syn-Chrom, Inc.) with a 1 M NaCl, 0.35 M NaH2P04 solvent system, pH 5.0. This solvent resolves GMP, GDP, and GTP clearly. Samples of GTP were found to contain 95-96% GTP and 4-5% GDP. The sample of GDP contained less than 1% GTP and 3-4% GMP.
To assay the relative amount of GDP or GTP bound to tubulin after column centrifugation, an aliquot of the perchloric acid extract used to determine the nucleotide concentration was mixed with 4 M KCH3COO (12%, v/v) to precipitate the perchloric acid and centrifuged. The resulting solution of GTP and GDP was then analyzed by HPLC to determine the fractional amount of each nucleotide.
Circular Dichroism-Far-UV CD spectra were measured on a Jasco Model 500 at 2 "C in 0.01 M Pipes, 0.1 mM dithioerythritol, pH 6.9, at a protein concentration of 5 p~ in a 0.1-cm jacketed cell. Molar residue ellipticities were calculated assuming a mean residue weight of 110.
Electron Paramagnetic Resonance-EPR spectra were collected on a Varian E-112 X-band spectrometer interfaced to a PDP 11/73 microcomputer, which drove the spectrometer magnetic field and collected signals digitally. A Varian E-257 temperature control unit regulated the temperature of the Varian E-238 TMllo mode high volume aqueous cavity. The spectra for Fig. 3 were recorded at 2 "C in 50-pl borosilicate capillaries at 9.432 GHz, 60-milliwatt microwave power, and 5-G modulation amplitude (peak to peak) with a filter time constant of 0.032 s. The field was scanned over 1000 G in 1024 X-band EPR spectrum of Mn2+ bound to tubulin is severely broad-increments, and each point was averaged over 2" readings. Since the ened by virtue of its coordination environment (Ref. 23; data not shown), it is possible to determine the concentration of free MnZ+ directly from the amplitude of its characteristic six-line spectrum. Each scan was blank-corrected prior to determining the free Mn2+ concentration by comparison with the spectrum of a known concentration of free Mn2+ (23). Kinetic measurements (Fig. 26) were performed in a TE-flat cell at 9.42 GHz and 30-milliwatt microwave power. The appropriate Mn2+ concentration (-100 p~) was mixed with the protein (-40 VM), and the cell was filled and allowed a few minutes to cool in the cavity. A single peak-to-peak data region was collected (usually the third or the fifth pair) every 5 min until the signal no longer changed.
Measurement of Binding of Mn2+ by GTP and GDP-The association constants for the binding of Mn2+ to GDP and GTP under the conditions of the experiments were measured by EPR (2 "C) and by the resin-binding method (0 "C) of Walaas (24), as modified by Jenkins (25). Bio-Rad AG MP-1 resin was used in place of the Dowex I-X2 recommended by Jenkins. The apparent equilibrium constants were corrected for contaminating GMP in the GDP samples and the binding of Mn2+ by contaminating GDP in the GTP samples. The following results were obtained K G D~ = 6.5 X lo3 M" (EPR) -6.0 X lo3 M-' (resin) and K G T~ = 5.6 X lo4 M" (EPR) -5.2 X lo' M" (resin).

RESULTS
The time course of the exchange of Mn2+ for M e bound to tubulin was investigated to test for complete and rapid binding. PC-tubulin (in the absence of excess nucleotide) with predominantly GTP at the E-site (TBGTP) and tubulin with Time ( Thus, it appears that high affinity metal sites on tubulin are associated with bound GTP on both the a-subunit (N-site) and @-subunit (E-site) of tubulin. The rapid exchange seen for TBGTP samples occurs at the E-site, and the slow exchange (identical for both TBGTP and TBGDP) occurs at the N-site.
Kinetic analysis of the slow exchange processes was carried out assuming pseudo first-order kinetics. The apparent dissociation rate for M r is 0.0204 min-' (Fig. 24. Monitoring the free Mnz+ EPR signal after mixing with tubulin (Fig. 2b) gives a time to equilibrium for Mnz+/TB that is equal to the time determined from the comparable data in Fig. 1: hap* = 0.039 min-'. Thus, the Mnz+/TB values appear to reach a plateau before the M P / T B values stop decreasing. However, when corrected for protein denaturation ( k = 8.55 x min"), the apparent Mn2+ binding rate (0.0212 min-l) is nearly identical to the apparent M$+ dissociation rate. This result is consistent with a one-to-one exchange of Mnz+ for The experiment summarized in Fig. 1 involves a rapid gel filtration step. The show similar results in a nonseparation experiment, PC-tubulin was mixed with various concentrations of Mnz+ at 0 "C and allowed to react until the exchange was complete (monitored by EPR peak heights), and then an EPR spectrum was collected. The results are shown as a Scatchard plot in Fig. 3. The GXP/TB values for all of these samples are nearly identical. The fraction of bound GDP, however, is different in each of the four experiments shown. As the occupancy of the E-site by GDP increases, the curve shifts to the left, corresponding to the presence of fewer high affinity metal sites. These results are consistent with the presence of one high affinity site in TBGDP (GTP at the Nsite) and two such sites in TBGTP (GTP at the Eand Nsites). Analysis of the scatchard data in Fig. 3  An alternate method for removing free and weakly bound metals from tubulin has been to use phosphocellulose (7,9) or Chelex 100 (4). However, since GTP binds weakly to the E-site unless it is complexed with M P or metal (I) and since E-site Mg2+ is in rapid equilibrium, E-site GTP will also be partially or completely removed by cation-exchange resins. In the presence of excess GTP, contaminating GDP will fill the vacated sites; in the absence of excess nucleotide, the vacated sites will remain empty (22). Rapid column centrifugation avoids these problems. will be presented elsewhere?) The values of the Mn2+ affinity for the E-site that is inferred from those experiments are 4.1-4.9 x lo7 "' for MnZ+ binding to TBGTP and 3.9-6.6 X lo3 M" for Mnz+ binding to TBGDP. As anticipated, these affinities are stronger than the corresponding M P affinitie~.~ They are also consistent with the data in Fig. 3. That the slow exchange and binding of Mn2+ to tubulin do not involve a significant conformational change in the protein J. J. Correia and R. C. Williams, Jr., manuscript in preparation.
' The Me-tubulin association constants were originally determined by fitting competitive binding data (1) to a scheme that did not include the N-site or seven weak sites. Refitting those data to the complete scheme at the same total M8+ concentrations gives the following values: K G T~ = 2.2-5.0 X 10' M-', K G T~M~ = 8.5-12.2 X lo7 M-' , and KGDPM~ = 4.8-5.5 X lo7 "I, and thus, the affinity of M e for GTP at both the E-and N-sites is 0.7-1.1 X lo6 M" and for GDP at the E-site is 2.0-2.3 X lo3 M-'. The chi-square values for these new fits are worse than the original fits (1) by a factor of 1.31-1.11. If the total M e concentrations are allowed to vary, the chi-square values can be reduced to the original values, but then the best fitted value of K G T~ = 2.00-1.90 M-'; and by the same criteria as before, we conclude that a reasonable upper limit for KGTP is 1.4 X lo' M-' (1).
Clearly, KGTP is poorly determined because it is at least 240-8700fold smaller than K G T~M~ Competitive binding experiments with MnZ+ will help set a definite range for KGTP. The qualitative conclusions of this study are not affected by these considerations. or Mn2+ and with 100 p~ GDP or GTP and incubated at 0 "C for at least 45 min before spectra were recorded between 260 and 200 nm. Each spectrum was scanned four times to obtain an average spectrum. The experiment was repeated on two successive days, but in reverse order to correct for time variations. The two experimental spectra for each sample were then averaged to give the final spectra shown. The maximum deviation at 217 nm represents +Z.O% S.D. and is consid- is apparent in Fig. 4, where the CD spectra of tubulin with Mn2+ or M e and with GDP or GTP are compared. All four spectra are nearly identical. Similar experiments with 16 mM M e , 100 p~ GDP or GTP, and 5 p~ tubulin also showed no significant change in the CD spectrum (data not shown). This result is in contrast to the differences observed by Howard and Timasheff (26). Buffers or temperature may account for the discrepancy. The protein is assembly-competent in the presence of Mn2+. In a 2 M glycerol assembly system, Mn2+tubulin and Me-tubulin react identically to warm and cold stimulus, giving rise to identical rates of assembly and disassembly as well as overall changes in turbidity per milligram polymerized (Fig. 5). The relative affinity of GTP for the tubulin E-site versus GTPMg (or GTPMn) (1) clearly implies that the concentration of GTPMg (or GTPMn) is the critical species in inducing microtubule assembly. The only polymers visible by electron microscopy at the plateaus were microtubules (data not shown). Buttlaire et al. (4) had reported different rates of assembly (in 10% dimethyl sulfoxide and 1 mM EGTA), but similar extents of assembly, with M g + and Mn2+. The changes in turbidity they observed (>0.43 A r 3~/ mg polymerized) are larger than those shown here (0.23 AT~SO/ mg polymerized) and may represent the presence of some nonmicrotubular forms (27). When tubulin is cycled in the presence of M P , Mn2+, or Mg+ and Mn2+ (Table I), the final assembly-competent protein contains amounts of bound metal, ( M e + Mn2+)/bound GTP, that are slightly larger than those shown in Fig. 1, but still consistent with high affinity metal-binding sites being at or near GTP-binding sites on both subunits. Metal-free GTP at the N-site may exist transiently prior to denaturation, and it may be removed by cycling. Similar levels of Mn2+ binding to tubulin after cycling were also observed by Buttlaire et al. (4).
The slow exchange of Mn2+ for M e at the N-site is consistent with a mechanism involving protein "breathing"' The term breathing is meant to imply the slow diffusion and release of M e through movements of a dynamic protein backbone.
The rate of Mn2+ binding (exchange) apparently involves the diffusion of M$+ through the protein or along the a/@interface, the dissociation rate of M e binding from the protein surface, the association rate of MnZ+ binding to the protein surface, and the diffusion of Mn2+ to the N-site. The M$+ dissociation rate is extremely slow (0.020 rnin") and approximately equal to the Mn2+ association rate, suggesting that the exchange process is kinetically limited by a slow isomerization involving the protein, possibly a local unfolding of the structure (28).  Rather than attempting to monitor metal exchange, we followed GDP binding by HPLC. Tubulin (initially 46% GDP) was incubated with 500 p~ GDP at 0 "C. At various times, aliquots were column-centrifuged and analyzed (see "Materials and Methods") for bound GDP and bound GTP. If dimer dissociation allows metal exchange, then it should also allow GDP to exchange for GTP at the N-site. The fraction of bound GDP did not change during more than 6 h (*2.46%). When 10 mM EDTA was included, the same result was observed, but the total binding of nucleotide (GXP/ TB) dropped at a rate four times faster than in the absence of EDTA. This observation suggests that removing metal from the N-site rapidly leads to denaturation of both subunits. If dimer dissociation were occurring, then either GDP cannot bind to the a-subunit, i.e. the N-site, or the subunit denatures faster than GDP binding can occur. Sternlicht et al. (2) concluded that a-tubulin lacks the guanine-specific binding loop (loop 111) found in P-tubulin and suggests that "the GTPbinding domain in a-tubulin is conformationally different from that in @-tubulin." This difference may in part explain the inability of GDP to replace GTP on the a-subunit, even though both the a-and @-tubulin GTP-binding domains do appear to bind GTP in an Mp-dependent manner. Additional support for an exchange mechanism involving a slow isomerization or unfolding of the protein comes from a lack of concentration dependence in the measured apparent Mg2+ dissociation rate and Mn2+ association rates. Over a protein concentration of 25.20-68.46 p~ and an Mn2+ concentration of 50-1000 p~, the apparent rates, when corrected for denaturation, were nearly identical. This observation supports the one-to-one exchange of MnZ+ for M p by a process that is kinetically limited by a protein isomerization step.

DISCUSSION
The results described above show that Mn2+ binds strongly to the GTP-occupied E-site and weakly to the GDP-occupied E-site of tubulin. In addition, Mn2+ can bind to the GTPoccupied N-site of tubulin by slowly exchanging for the MgZ+ usually bound there. The binding of M F has previously been measured by competitive binding studies (l), and its affinity for the E-site of TBGDP was found to be much weaker than its affinity for the E-site of TBGTP. On the basis of the affinities of Mn2+ for GDP and GTP and in the absence of steric constraints, it would be expected that Mn2+ will bind with the same relative affinities to the GDP-and GTPoccupied E-site. The binding data presented in Fig. 3 are consistent with this assertion. Competitive binding data for Mn2+, GTP, and GDP as well as nonlinear least-squares fitting and a detailed error analysis of the confidence intervals for the appropriate association constants will be presented elsewhere? Here we can summarize that the association constants for binding to GTP at the E-and N-sites are 0.7-1.1 X lo6 M" for M e and 4.1-4.9 X lo7 I"' for Mn2+. For the binding to GDP at the E-site, the affinities are 2.0-2.3 X lo3 M-' for M$+ and 3.9-6.6 X lo3 M" for Mn2+.
The evidence for binding and exchange of divalent cations at the N-site rests on two findings. First, as shown in Fig. l b and Table I, there appears to be a one-to-one correspondence between bound GTP and bound cation. Jemiolo and Grisham (5) established that the apparent total number of high affinity sites varies with the amount of nucleotide bound to tubulin, but included no measurement of amounts of GTP or GDP.
Rather, they assumed that tight binding occurred to either GTP or GDP and only when the nucleotide was bound to the E-site. Our work shows that these assumptions were too simple. Second, as shown in Fig. 3, when the E-site is occupied by GDP, tubulin still has a single high affinity metal-binding site. Because of the homology of the a-and @-subunits of tubulin, one would expect similar GTP-binding sites, i.e. sites that require M e coordination of comparable affinities. Both subunits contain an aspartic acid in position 205 that is in a sequence that has been identified as an M e coordination site by homology with other nucleotide-binding proteins (2).
EDTA does not completely remove bound M F unless the protein is first denatured (4). Our results explain this observation as being due to the slow release of M$+ from the Nsite. The finding that there is both a rapidly exchanging and a slowly exchanging site strongly supports the hypothesis that the N-site metal, unlike the N-site nucleotide, can be exchanged. This phenomenon has not been widely noted before, probably because the process of removing free metal in preparing the sample causes the dissociation of E-site GTP and the subsequent binding of GDP there (1). Himes et al. (7) have observed one tightly bound M e and one tightly bound Co2+ per tubulin. Definitive location of the divalent cation awaits application of more sensitive techniques, e.g. 35-GHz EPR (29) or NMR (11).
The CD (Fig. 4) and polymerization (Fig. 5 ) data clearly demonstrate no large conformational or functional changes in the protein when Mn2+ is exchanged for MgZ+ and binds to the guanine nucleotide sites. The absence of a major change in secondary structure between TBGDP and TBGTP is not necessarily inconsistent with the model hypothesized by Sternlicht et al. (2). The number of tightly bound metals per bound GTP observed after a polymerization cycle (-1.0; see Table I) also suggests that this one-to-one correspondence is typical of native, functional tubulin. These results explain the M~+ composition of TBGDP after gel filtration (1) or phosphocellulose chromatography (16) and are, in fact, consistent with the weak binding of M$+ to E-site GDP.
It is not clear whether other metals will bind to the E-site in the same nucleotide-dependent manner. Zn2+, Co2+, Cr2+, and A13+ (30) are all believed to bind to the tubulin E-site and to stimulate assembly. It has also been suggested that weak binding sites are required to promote the formation of both microtubules and sheets, with sheets being favored at higher metal concentrations. The data are also consistent with the binding of these metals to the N-site since slow exchange (4, 5,7) or nonexchangeable characteristics have been described.
Our data are consistent with the idea that N-site metal is directly important in protein stability, and thus only indirectly in microtubule formation. Within the microtubule, metal at both sites may be slowly exchangeable. Carlier and Pantaloni (31) have reported the slow release of Pi from actin filaments subsequent to ATP hydrolysis. The mechanism may be similar to that reported here. A13+ apparently promotes microtubule assembly but not GTP hydrolysis (30,32). The extra charge on A13+ might possibly prevent the hydrolysis step by stabilizing the TBGTP conformation over an intermediate enzymatic conformation. In summary, high affinity divalent cation ( M e and Mn2+) binding to tubulin occurs at GTP-occupied guanine nucleotide-binding domains. The strength of binding to GDP at the E-site is comparable to the strength of binding to GDP in solution and suggests little or no coordination of the metal with the protein. Metal binding to the TBGTP E-site is consistent with a coordination with the protein and possibly involves a conformational change in the protein. (Kinetically, it may actually involve GTPMg binding to tubulin.) The slow exchange of metal at the N-site may provide a new means of differentially labeling tubulin and represents a new kind of evidence of homology between the a-and @-subunits: the Me-dependent binding of GTP. This heterogeneity in high affinity metal-binding sites on tubulin may complicate the interpretation of more quantitative structural studies of the metal-nucleotide coordination environment by, for example, 35-GHz EPR. Alternatively, if the sites are identical, it may increase the sensitivity of such experiments.