The Majority of the Saccharomyces cerevisiae Septin Complexes Do Not Exchange Guanine Nucleotides*

We show here that affinity-purified Saccharomyces cerevisiae septin complexes contain stoichiometric amounts of guanine nucleotides, specifically GTP and GDP. Using a 15N-dilution assay read-out by liquid chromatography-tandem mass spectrometry, we determined that the majority of the bound guanine nucleotides do not turn over in vivo during one cell cycle period. In vitro, the isolated S. cerevisiae septin complexes have similar GTP binding and hydrolytic properties to the Drosophila septin complexes (Field, C. M., al-Awar, O., Rosenblatt, J., Wong, M. L., Alberts, B., and Mitchison, T. J. (1996) J. Cell Biol. 133, 605-616). In particular, the GTP turnover of septins is very slow when compared with the GTP turnover for Ras-like GTPases. We conclude that bound GTP and GDP play a structural, rather then regulatory, role for the majority of septins in proliferating cells as GTP does for α-tubulin.

Septins are a family of proteins initially identified in Saccharomyces cerevisiae and later found in most eukaryotic organisms (1)(2)(3)(4). They are important for cytokinesis (2,4) and membrane trafficking (5)(6)(7). The S. cerevisiae septins play additional roles in bud morphogenesis (1,8), chitin deposition (9), and spindle positioning (10). The molecular function of septins in all of the above processes is not well understood. Electron microscopy (11) and immunofluorescence (12)(13)(14) experiments in wild-type and temperature-sensitive septin mutants suggest that the S. cerevisiae septins assemble into higher-order structures in vivo. One hypothesis is that septins form molecular scaffolds at specific cortical locations and recruit diverse effector proteins (15,16). Recruitment to the septin scaffold might then catalyze interactions between certain effectors by proximity effects.
Multiple septin proteins (polypeptides) are expressed in each of the studied organisms. When isolated, the septin polypeptides are found in tight complexes with defined stoichiometries (17)(18)(19). Five vegetative septins were identified in S. cerevisiae: Cdc3, Cdc10, Cdc11, Cdc12, and Shs1 (Sep7) (1,20). When a peptide antibody against Cdc3 was used for affinity purification, the four Cdc septins were isolated as a defined complex (19). Although the precise stoichiometry of each protein in the complex is not known, an approximate stoichiometry was cal-culated using gel filtration, sucrose gradients, and quantitation of Coomassie R-250-stained bands (19). These experiments suggest that Cdc3, Cdc10, Cdc11, and Cdc12 are the major constituents of the septin complex, and the approximate stoichiometry of these proteins is 2:2:1:2, respectively (19).
Septins have a characteristic domain structure: an N terminus with variable length and sequence, a conserved central region, and with a few exceptions, a C terminus predicted to form coiled-coils (15). The conserved central region contains Ras-like GTP-binding motifs (4). Indeed, Field et al. (17) showed that Drosophila septin complexes copurify with tightly bound GTP and GDP and have GTPase activity in vitro. However, GTP binding/exchange of the Drosophila septin complex was very slow in vitro when compared with the activities of generic Ras-like small GTPases (21). In contrast, individual recombinant septins (Xenopus (22), mammalian (4,23)) and reconstituted partial septin complexes (mammalian (23)) show considerably faster GTP binding/exchange kinetics.
The role of septins' GTP binding and hydrolysis is not well understood, and it constitutes the subject of this report. GTP might turn over rapidly and regulate the function of septins, as in the cases of ␤-tubulin (24,25), FtsZ (26), and small GTPases. Alternatively, GTP might incorporate once during folding or complex assembly and play no further role in the function of septins, analogous to GTP bound to ␣-tubulin (27,28). In vitro experiments implicate GTP binding in filament formation by an individual Xenopus septin (22) and possibly in the interaction between different septin polypeptides (23). However, the in vivo significance of these observations has not been tested. The cited papers echo the prevailing assumption in the field that GTP binding plays a regulatory role. Here we use in vivo exchange and biochemical assays to test whether GTP plays a regulatory or a structural role. Our data favor the latter model, which should trigger a significant change in the direction of the field.

Reagents
Anti-Cdc3 antibody was raised against a synthetic peptide corresponding to the C terminus of the Cdc3 protein ([C]NHSPVPTKKKGFLR), as described previously (19). Anti-Shs1 antibody was a gift from Prof. Douglas Kellogg (University of California at Santa Cruz) (30). ( 15 NH 4 ) 2 SO 4 98ϩ atom % 15 N was purchased from Aldrich Chemical Co.

Septin Complex Purification
S. cerevisiae septin complexes were purified using a variation of a purification method published previously (19). Briefly, YEF473 frozen yeast was pulverized with a liquid N 2 -chilled mortar and pestle. The resulting yeast powder was thawed shortly at room temperature. All of the following steps were performed at 4°C. Extract buffer (50 mM Tris-HCl, pH 7.8, 100 mM KCl, 2 mM MgCl 2 , 0.5 mM EGTA, 0.8% Triton X-100, 1 mM dithiothreitol, 1 mM phenylmethylsulfonyl fluoride, 1 mM leupeptin, 1 mM chymostatin, and 1 mM pepstatin) was added to the thawed powder in a ratio of 1:1.3 (mass/ml). The mixture was gently rotated for 1 h and then spun for 12 min at 500 ϫ g. The supernatant was supplemented with 1 mM phenylmethylsulfonyl fluoride, 1 mM leupeptin, 1 mM chymostatin, and 1 mM pepstatin and incubated for ϳ210 min with (anti-Cdc3)-protein A beads (Affi-Prep Protein A Support, Bio-Rad; previously incubated with anti-Cdc3 antibody for 2 h at 4°C). The beads were pelleted, and the supernatant depletion of Cdc3, as assessed by Western blotting analysis, was Ͼ80%. The beads were washed four times with eight volumes of wash buffer (50 mM Tris-HCl, pH 7.8, 250 mM KCl, 2 mM MgCl 2 , 0.5 mM EGTA, and 0.2% Triton X-100) over a period of ϳ90 min. The spins during the washes were performed at 350 ϫ g, 350 ϫ g, 500 ϫ g, and 3000 ϫ g for 5 min each. The washed beads were incubated over night in 1.6 volumes of elution buffer (20 mM Tris-HCl, pH 7.8, 1 M KCl, 2 mM MgCl 2 , 0.5 mM EGTA, 4% sucrose, and 0.8 mg/ml Cdc3 peptide). The beads were pelleted at 500 ϫ g, and the supernatant containing the septin complexes was used for further experiments.

Determination of Septin Complex Concentrations
The purified septin complexes together with bovine serum albumin (BSA) standards were run on a 10% SDS-PAGE and stained with Coomassie G-250 (31). The amount of Cdc12 in the sample was determined based on the Coomassie G-250 staining intensity of the BSA standards using Quantity One software. The precise stoichiometry of the septin complex is not known. Two extreme stoichiometries were assumed for determining the molarity of the GTP-binding sites (Table  I): one was 2 Cdc3:2 Cdc10:1 Cdc11:2 Cdc12; and the other was 1 Shs1:2 Cdc3:2 Cdc10:2 Cdc11:2 Cdc12 (boldface used for clarity). The molar concentration of the septin complex is the same in both cases using our analysis method, but the molarity of the GTP-binding sites is different by a factor of 9/7 ϭ 1.28. The molar concentration of the septin complexes in our preparations ranged from ϳ100 to 300 nM.

Nucleotide Content Analysis
The nucleotide content of the septin proteins was determined using a variation of a procedure published previously (17,32,33). The following manipulations were performed at 4°C. Poly(ethylene glycol) (PEG) 1 8000 was added to the septin complex preparation at a final concentra-tion of 10%. The mixture was incubated for 2 h and then spun for 25 min at 390,880 ϫ g using a TLA 100 rotor and a Beckman Coulter Optima TL Ultracentrifuge. Most of the septin proteins were pelleted in this way. The pellet was resuspended in 20 l of 8 M urea, heated at 100°C for 1 min, and diluted with 20 l of water. The resulting 40 l were spun through a 30-kDa cut-off filter (Microcon, Millipore) at 10,000 ϫ g. Water (40 l) was then added to the top of the filter, and the spin was repeated. The total filtrate was applied to a Pharmacia Biotech SMART system MonoQ PC 1.6/5 column. The nucleotide bound to the column was eluted using a gradient of NH 4 HCO 3 (from 100 to 500 mM over 3 ml). The supernatant of the PEG 8000-pelleted septins was supplemented with 8 M urea, heated 1 min at 100°C, diluted to 2 ml in water, and then applied to the MonoQ column. We found that the supernatant contained only trace amounts of nucleotides. Alternatively, the septin complex was dialyzed for 2 h at 4°C against 20 mM Tris-HCl, pH 7.8, 75 mM KCl, 2 mM MgCl 2 , 0.5 mM EGTA, 1 mM dithiothreitol, and 4% sucrose, and the resulting septin filaments (most of the septin protein) (19) were pelleted at 390,880 ϫ g for 15 min. The septin pellet was resuspended in 8 M urea, and the nucleotides were isolated as above. For nucleotide quantitation, we added known amounts of ATP to the septin pellet prior to urea denaturation, injected the total isolated nucleotides on the SMART system, and monitored the absorbance at 254 nm. The ATP signal was corrected for the difference in extinction coefficient between ATP and guanine nucleotides and was used to calculate the absolute amounts of GTP and GDP.

Filter-binding Assay
Septin complexes in 1 M KCl (as eluted from the antibody beads) were dialyzed for 2 h at 4°C against 20 mM Tris-HCl, pH 7.8, 75 mM KCl, 2 mM MgCl 2 , 0.5 mM EGTA, 1 mM dithiothreitol, and 4% sucrose. The dialyzed septins were supplemented with [␣-32 P]GTP at a final concentration of 2 M (specific activity (SA), 150 -300 Ci/mmol). The samples were incubated at 30°C. Aliquots were withdrawn at specific times, diluted in ice-cold wash buffer, and applied to 1.3 cm ϫ 1.3 cm nitrocellulose filters (Scheicher & Schuell, 0.45 m) connected to a vacuum system. Two to three aliquots were used for every time point. The filters were then washed with 5 ml of wash buffer (20 mM Tris-HCl, pH 7.8, 100 mM KCl, 0.5 mM EGTA, and 2 mM MgCl 2 ) and dissolved in 250 l of acetone. Scintillation mixture (ScintiSafe Plus 50%, Fisher Scientific, 5.6 ml) was added to each of them. The mixture was briefly vortexed and counted for 2 min.

GTP Hydrolysis Assay
This experiment was performed in parallel to the filter-binding assay. Two extra samples were taken at each time point. These samples were filtered through nitrocellulose and washed as above. Urea solution (8 M, 50 l) was added immediately to each filter to denature the bound protein. Samples of the nucleotides released (25 l) were spotted onto a TLC plate (Baker-flex, cellulose PEI-F) together with GTP and GDP standards (34). The plates were washed ascendantly in water to separate urea from the nucleotides. The plate was air dried and then developed in 1 M LiCl until the front reached 11 cm from the sample application line. The TLC plate was dried and exposed to a Molecular The results of six independent experiments are shown. The yeast extract was obtained either by bead beating or by the mortar and pestle method, as indicated. The septin complexes were precipitated either by incubation with PEG 8000 for 2 h at 4°C followed by 25 min centrifugation at 390,880 ϫ g, or by 2-h dialysis to 75 mM KCl at 4°C followed by 15 min centrifugation at 390,880 ϫ g. Dialysis promotes the association of septin complexes into long filaments that can be pelleted using the above conditions. The bound nucleotide was quantified using co-injection with ATP standards into the SMART system (see "Experimental Procedures"). The septin protein retrieved from the 30-kDa cut-off filter (see "Nucleotide Content Analysis") was quantified by running it on a 10% polyacrylamide gel together with BSA standards of known concentrations. Quantity One software was used for analysis. The protein concentration is probably underestimated by our method because the limitations of the technique (using BSA as a standard) and to the multiple inherent manipulations (spinning through the 30-kDa cut-off filter, transferring to tubes) before the final measurement. The nucleotide concentration is internally controlled using ATP and therefore much more accurate. Thus, the molar ratio of total nucleotide to protein is overestimated. The first number in the Total protein column and the first number in the Nucleotide/protein ratio columns were determined using the 2:2:1:2 stoichiometry, whereas the second one of each set was determined using the 1:2:2:2:2 stoichiometry (see "Experimental Procedures"). Dynamics phosphorimager screen. The screen was analyzed using a Bio-Rad Molecular Imager and Quantity One software.

UV Cross-linking
Samples of the dialyzed septin complexes were incubated with [␣-32 P]GTP (2 M GTP SA, 300 Ci/mmol) and 300 M ATP at 30°C with or without competing GTP at 300 M. Aliquots (25 l) were taken at each time point and used as follows. Three samples of 5 l each were counted using the filter-binding assay. The remaining 10 l were mixed with additional ATP and 2-mercaptoethanol at a final concentration of 300 M ATP and 1% 2-mercaptoethanol. The resulting mixture was incubated on ice for 1 min and then cross-linked on ice for five periods of 3 min each (with 1-min pause between each cross-linking period) using a 254-nm Stratagene UV Stratalinker 2400. The sample was set at 7 cm from the UV lamps, and the power of the instrument was ϳ1640 W/cm 2 . The cross-linked sample was quenched with 3.7 l of 4ϫ sample buffer and heated for 30 s at 100°C. 200 l of 20 mM Tris-HCl, pH 7.8, 100 mM KCl, 0.5 mM EGTA, and 2 mM MgCl 2 solution was added to each sample. The mixture was filtered through a 30-kDa cut-off Microcon filter until 98% of the liquid went through (ϳ15 min at 8,000 ϫ g). The top of the filter was resuspended in 30 l of 1ϫ sample buffer, heated for 1 min at 100°C, and subjected to 10% SDS-PAGE. The gel was stained using Colloidal Coomassie G-250 (31), dried, and exposed to the phosphorimager screen.

In Vivo GTP Exchange Assay
Yeast Growth, Collection, and Processing-An FY4 yeast strain was grown in liquid minimal medium containing 2.5 g/l ( 15 NH 4 ) 2 SO 4 , 1.7 g/l yeast nitrogen base without amino acids and (NH 4 ) 2 SO 4 , and 20 g/l glucose. The culture was grown for ϳ24 h to a final A 600 of 7. Four flasks containing 600 ml of 15 N minimal medium were inoculated with 45 ml of the overnight culture each. The cultures were grown 5 h at 30°C until A 600 was 1.2. The cultures were filtered through a 90-mm Durapore membrane filter (0.22 m GV, Millipore), and the filters were transferred to new flasks containing 650 ml of chase medium (5 g/l ( 14 NH 4 ) 2 SO 4 , 1.7 g/l yeast nitrogen base without amino acids and (NH 4 ) 2 SO 4 , 0.06 g/l guanosine, and 20 g/l glucose) each. The doubling time was ϳ3 h under these conditions. The cultures were grown for 0, 1, 2, and 3 h in the chase medium, respectively. From each culture, 600 ml were spun at 2,600 ϫ g for 25 min. The pelleted yeast was frozen in liquid N 2 . The remaining 45 ml were filtered through a 47-mm Millipore membrane (type GS, 0.22 M). The membrane was immersed in 1 ml of ice-cold 1 M formic acid saturated in 1-butanol to lyse the cells, precipitate the proteins, and release the intracellular pool of small molecules (35). The mixture was kept on ice for 20 min. The liquid and the precipitate scraped from the membrane were spun at 20,800 ϫ g for 15 min. The supernatant was frozen in liquid N 2 and lyophilized. The cellular GTP and GDP were isolated from the other small molecules by dissolving the lyophilized sample in water, running it over a SMART system MonoQ column, and pooling together the fractions containing GTP and GDP (see "Nucleotide Content Analysis"). The pooled fractions were lyophilized and analyzed by mass spectrometry.
Septin Polypeptides and Septin-bound Nucleotide Preparation-Purified septin complex was obtained from the frozen yeast samples essentially as described above for the YEF473 yeast. However, the yeast was broken by bead beating at 4°C using 0.5-mm glass beads (BioSpec Products). The Mini Bead Beater (BioSpec Products) was set to "homogenize." Three pulses of 15 s each were applied to the yeast-and-beads slurry. Between each of the pulses, the tubes were cooled on ice for 5 min. The nucleotides bound to the purified samples were isolated by precipitating the septins with PEG 8000, denaturing them with 8 M urea, and filtering the released nucleotide through a 30-kDa cut-off filter (see "Nucleotide Content Analysis"). The filtrate containing the bound nucleotides was further purified on a capillary immobilized metal affinity (IMAC) solid-phase extraction (SPE) device (as described below) and analyzed by mass spectrometry. Sample buffer (1ϫ, 30 l) was added to the top of the filter, and the filter was heated at 100°C for 1 min. The sample buffer solution containing the septin polypeptides was removed and subjected to 10% SDS-PAGE. The resulting gel was stained with Coomassie R-250. The septin protein bands were cut out of the gel and analyzed by mass spectrometry.
IMAC Purification and Precursor-Ion Nanoelectrospray of GDP and GTP-IMAC SPE devices were prepared as described previously using gel-loader pipette tips for a total bed volume of ϳ1 l (36). The filtrate containing the bound nucleotide (ϳ40 l) was diluted 1:1 with 0.3 M acetic acid and slowly loaded onto a washed and equilibrated IMAC SPE device (load rate, Ͻ3 l/min). The SPE device was then washed with 40 l of 0.1 M acetic acid, followed by 40 l of a 3:1 mixture of 0.1 M acetic acid and acetonitrile. Immediately before analysis, the SPE bed was washed with 3 l of a 3:1 mixture of acetonitrile and water, followed by elution with 1 l of 0.1 M ammonia/50% methanol/water directly into a medium tip-length nanoelectrospray emitter (Proxeon Biosystems) and negative-ion nanoelectrospray. The lyophilized cellular GTP ϩ GDP sample was reconstituted in 75 l of purified water, of which 1 l was diluted in 3 l of 0.1 M ammonia/50% methanol/water and loaded into a nanoelectrospray emitter and negative-ion nanoelectrospray. Precursor-ion scanning was performed on a ThermoFinnigan TSQ Quantum triple quadrupole mass spectrometer at optimum spray voltages between Ϫ575 and Ϫ800 V. The nozzle-skimmer (declustering) offset was 15 V. The negative PO 3 Ϫ product ion at 79 m/z (Q3 resolution, 0.8 full width at half maxim (FWHM)) was used to scan for precursors containing phosphate esters (Q2 argon pressure, 1.5 mtorr; collision energy, 40 V). For GDP, Q1 resolution was 0.6 FWHM, and the scan was performed from 420 to 470 m/z; for GTP Q1 resolution of 0.2 FWHM, a scan was performed from 250 to 275 m/z. Optimum spectral signal-tonoise and peak resolution was obtained using long scan times (30 s) for 10 scans.
Nanoscale Microcapillary Peptide Liquid Chromatography-Selected Ion Monitoring-Sections of gel corresponding to Cdc3, Cdc10, Cdc11, and Cdc12 migration at each time point were excised individually and trypsinized as described previously (37). The final peptide extracts were dried by vacuum centrifugation and stored at Ϫ 20°C until immediately before analysis. For analysis, peptide samples were reconstituted with 20 l of 6% acetic acid/5% methanol/0.05% trifluoroacetic acid solution, of which 5 l was transferred to deactivated glass limited volume inserts (Agilent) and crimp-sealed in 2-ml autosampler vials. Each peptide sample (4 l) was analyzed by nanoscale microcapillary liquid chromatography-selected ion monitoring on in-house fabricated, 100 m ϫ 12 cm fused silica reverse-phase microcapillary columns (200 Å pore, 5 m particle C 18 -AQ, Michrom Bioresources). A FAMOS capillary autosampler (LC Packings), an Agilent 1100 liquid chromatograph, and a ThermoFinnigan TSQ Quantum (ThermoElectron) triple quadrupole mass spectrometer were used.
The peptide selected ion monitoring windows (in m/z) for each protein were first determined via sequencing analysis of each individual native protein tryptic digest by liquid chromatography-tandem mass spectrometry using a ThermoFinnigan LCQ DecaXP ion trap mass spectrometer. Peptide ions for selected ion monitoring were chosen so that they exhibited a favorable combination of high signal-to-noise and lack of interfering ions in the m/z space corresponding to any combination of 15 N incorporation, based on the peptide sequence. In addition, the sequences of selected peptides were confirmed by comparison of the corresponding tandem mass spectra against the non-redundant yeast protein database using the computer algorithm SEQUEST (38). The peptide sequences, corresponding selected ion monitoring ranges, and relevant TSQ Quantum instrument settings are as follows. For all (MϩH ϩ ) ϩ ions, the selected ion monitoring scan time was 400 millisec- Peptide samples were resolved across a 20-min gradient of 10% acetonitrile/ 0.1% formic acid/0.005% heptafluorobutyric acid to 35% acetonitrile/ 0.1% formic acid/0.005% heptafluorobutyric acid.

RESULTS
We purified S. cerevisiae septin complexes by immunoprecipitation with an anti-Cdc3 peptide antibody (Fig. 1A) (19). The isolated septin complexes contained the previously described septin proteins Cdc3, Cdc10, Cdc11, and Cdc12 (19). Shs1 (Sep7) was also present in our preparations. Shs1 migrated on SDS-PAGE as multiple species that partially overlapped with Cdc3 (Fig. 1A). Coomassie G-250 staining (Fig. 5) suggested that each of the Shs1 species is sub-stoichiometric to the other septin proteins. However, quantitation of the Shs1 species staining intensities suggested that their sums are present in a stoichiometric amount (Table II, Column II).
We released the small molecules associated with the septin complexes by urea denaturation and identified and quantified the bound nucleotides by anion exchange chromatography/UV absorbance ( Fig. 1B; Table I). The septin complexes contain bound guanine nucleotides, specifically GTP and GDP. The ratio of bound GDP to bound GTP is ϳ2.2:1 ( Fig. 1B; Table I), similar to the Drosophila septin complexes (17). Comparing nucleotide to protein, the S. cerevisiae septin complexes contain approximately one mol of guanine nucleotide per mol of septin polypeptide. Because the precise stoichiometry of the septin complex is not known, we estimated the ratio of bound nucleotide to the septin protein using two possible stoichiometries for the septin complex (see "Experimental Procedures"). The results are shown in Table I.
To understand whether these guanine nucleotides play an active, regulatory role in the biology of septins, we measured their turnover rates in vivo and compared them to the turnover rates of the septin proteins. We reasoned that if bound GTP and GDP play an active, regulatory role in the biology of septins, they would turn over at least once during one cell cycle period. If, in contrast, they play a static, structural role, their turnover rate would be similar to the turnover rate of the septin proteins. This approach is analogous to the experiments that showed that GDP bound to ␤-tubulin turns over fast, consistent with a regulatory role, whereas GTP bound to ␣-tu-bulin turns over at the same rate as the protein, consistent with a structural role (28,39).
To determine turnover rates of nucleotide and protein in the same experiment, we used a 15 N-dilution assay. Specifically, we labeled the nucleotide and protein pools with 15 N by growing yeast in medium containing ( 15 NH 4 ) 2 SO 4 as the sole nitrogen source. The label was then chased with medium containing exclusively 14 N nitrogen sources. Subsequently, samples were analyzed by mass spectrometry, and the ratio of 15 N to ( 14 N ϩ 15 N) was measured in three samples: total cellular GTP ϩ GDP (to determine how fast the label was diluted in the pool), septin-bound GTP ϩ GDP, and septin polypeptides (Fig. 2B). The turnover rate of the cellular GTP ϩ GDP pool was not measured precisely. However, we found that the 15 N GDP (M-H ϩ ) Ϫ ion (that originates from both GTP and GDP) was almost completely turned over during the first hour of the chase (ϳ1/3 of the cell cycle under these growth conditions) ( Fig. 2A, first row, and Fig. 2B, solid circles). This is sufficiently fast to allow turnover analysis for septin-bound GTP and GDP. The turnover rate for the soluble amino acids pool was not measured, but we assume that it is at least as fast as the total guanine nucleotide pool. Note that because we did not make this measurement, we cannot exactly compare the turnover rates of protein and bound nucleotides.
We determined that there was no significant in vivo exchange of the septin-bound GTP and GDP. Because yeast actively grew during our chase period (approximately one cell cycle), the 15 N-labeled amino acids and nucleotides were diluted with the newly synthesized 14 N material. If a particular protein and/or nucleotide pool did not turn over, the 15 N to ( 14 N ϩ 15 N) ratio would be 0.5 by the time the cells doubled once. This is approximately what we observe for both septinbound GTP ϩ GDP (data for GDP (M-H ϩ ) Ϫ shown in Fig. 2A,  second row, and Fig. 2B, solid squares; similar data not shown for GTP (M-2H ϩ ) 2Ϫ ), and the septin proteins (data for the SWDPIIK Cdc3 peptide shown in Fig. 2A, third row; data for peptides from other septin proteins shown in Fig. 2B, open squares and solid squares). Thus, septin-bound GTP and GDP do not turnover during one cell cycle period. Because the yeast was not synchronized, each cell would have visited every cell cycle state at the end of the chase period (approximately one cell cycle). Thus, we can rule out the possibility that a certain cell cycle transition could cause rapid GTP turnover in the bulk population of septins. This scenario would have resulted in an almost complete turnover of the labeled GTP by the end of the chase period. Because this is a bulk measurement, we would, however, miss fast GTP turnover in a small subpopulation of septin complexes, or in a single, special GTP-binding site within the septin complex (see "Discussion").  Fig. 5 was scanned, and the protein bands were quantified using Quantity One software. The intensity of each band after background subtraction was normalized by the molecular weight of that particular protein, and percentages for each of the normalized intensities from the total normalized intensities were calculated (Column II). The intensity of the time-dependent UV cross-links was calculated as follows. Intensities of each band of the phophorimager scan were calculated for both the 0-h and the 5-h time points using Quantity One software. The 0-h intensities were subtracted from their respective 5-h intensities. A percentage was calculated for the adjusted intensity from the sum of the total adjusted intensities (Column III). The percentage of time-dependent cross-links normalized to the molar amount of protein is shown in Column IV. For Shs1, the protein quantitation data represent the sum of the Shs1 bands that do not overlap with Cdc3, whereas the UV cross-linking quantitation data represent the slower migrating form that significantly cross-links GTP. To confirm the lack of GTP exchange in the bulk septin population and to test for the presence of faster exchanging guanine nucleotide sites, we examined GTP exchange and GTP hydrolysis in vitro. We isolated S. cerevisiae septin complexes by immunoprecipitation and peptide elution, incubated them with [␣-32 P]GTP, and filtered them through nitrocellulose at FIG. 2. In vivo turnover analysis of cellular GTP ؉ GDP, septin-bound GTP ؉ GDP, and septin polypeptides. A prototrophic yeast strain was grown in minimal medium containing ( 15 NH 4 ) 2 SO 4 as the only nitrogen source. The 15 N labeling of the cellular GTP ϩ GDP and septin pools was ϳ98% as determined by mass spectrometry. The label was then chased by changing the yeast to medium containing ( 14 NH 4 ) 2 SO 4 . The yeast was collected at 0, 1, 2, and 3 h after the start of the chase. The total cellular GTP ϩ GDP, the septin proteins, and the septin-bound GTP ϩ GDP were purified from the collected yeast. Each of the isolated fractions was analyzed by mass spectrometry. A, sample spectra for each of the analyzed pools are shown. Each row represents a particular ion (as indicated), and each column represents a particular time point of the chase (as indicated). Of the protein pool, a Cdc3 peptide, SWDPIIK (MϩH ϩ ) ϩ 858.47 m/z, is shown. specific time points. Because the nitrocellulose filters retain the septin complexes, bound radioactive nucleotides can be measured. We found that the isolated septin complex binds GTP in vitro (Fig. 3A, solid diamonds). The GTP-binding activity of the S. cerevisiae septins is competed by excess cold GTP (Fig. 3A, solid triangles) but not by excess cold ATP (Fig.  3A, solid squares). The maximum binding of ϳ0.5 mol of guanine nucleotide per mol of septin complex was achieved after 8 h of incubation at 30°C (Fig. 4). This binding represents ϳ7% of the binding sites of the complexes given that one mol of the septin complex contains 7-9 mol of GTP-binding sites (depending on the stoichiometry of the complex, see "Experimental Procedures"), and assuming that all binding sites are equally susceptible to this reaction. Note that given the sub-stoichiometric nature of the GTP-binding reaction and our results described in Fig. 4, we cannot discriminate between binding attributable to exchange of the already bound GTP and GDP and binding attributable to the presence of empty septin GTPbinding sites (see "Discussion").
To test whether the S. cerevisiae septins hydrolyze the GTP they bind, we analyzed the radioactive nucleotides retained on the nitrocellulose filter by the septin complexes using TLC for 0.5-8-h incubation periods (Fig. 3B). TLC plate phosphorimager scans show that t1 ⁄2 of bound GTP being converted to GDP is Ͻ30 min (Fig. 3B). This half-life is similar to the half-life of GTP bound to Ras-like small GTPases in the absence of GTPase-activating proteins (21). In contrast to Ras-like GTPases (21), the hydrolytic activity of the septins is limited by their slow GTP-binding or exchange activity (Figs. 3A and 4).
To determine whether the GTP binding we observed in vitro (Fig. 3A) was caused by the exchange of the already bound GTP and GDP that co-purified with the complex, we analyzed the association and dissociation kinetics of the GTP-binding reaction. We incubated septin complexes with [␣-32 P]GTP, and at time ϭ 4.5 h, we added excess cold GTP to an aliquot of the reaction. We quantified the bound nucleotides at earlier and later time points using the filter-binding assay (Fig. 4). We found that the GTP association and dissociation reactions have different kinetics parameters. The t1 ⁄2 of the association reaction (ϳ3.5 h) is shorter than the t1 ⁄2 of the dissociation reaction (ϳ12.5 h). This may indicate that different rate-limiting steps govern these reactions (see "Discussion").
To identify which polypeptides are responsible for the observed in vitro GTP-binding activity, we incubated our septin preparation with [␣-32 P]GTP for different amounts of time (0 and 5 h) and then UV cross-linked the septin complex to the bound radioactive nucleotide (27). The identity of the crosslinked proteins was determined by SDS-PAGE and phosphorimager analysis (Fig. 5). The septin incubation reaction contained 300 M ATP throughout the incubation period, and additional 300 M ATP and 1% 2-mercaptoethanol were added just before the UV irradiation to suppress nonspecific crosslinking. Although all septin polypeptides showed GTP crosslinks competed by excess cold GTP (300 M) (Fig. 5), two of the septins, Cdc10 and Cdc12, were responsible for the majority of time-dependent cross-links (Fig. 5) (Table II, Column III). DISCUSSION We developed a novel 15 N-dilution assay for analysis of nucleotide turnover and determined that the majority of the yeast septin complexes do not turn over GTP in vivo during one cell cycle period (Fig. 2). In vitro, the GTP binding/exchange of septins is sub-stoichiometric and slow when compared with the The results are given as the average of the three aliquots counted at each time point. The upper limit of the vertical line corresponding to each symbol represents the highest value, and the lower limit represents the lowest value. f, samples from the incubation reaction with no additional GTP; OE, samples from the incubation reaction with an additional 1.5 mM GTP. The association data (f) were fitted to the onephase exponential association equation Y ϭ Y max (1 Ϫ exp(ϪkX)). The dissociation data (OE) were fitted to the one-phase exponential decay equation Y ϭ Span(exp(ϪkX)) ϩ Plateau. The R 2 values for the resulting curves are ϳ0.95. The data analysis was performed using the GraphPad Prism software. The observed rate constant for the association reaction k ob is ϳ0.2 h Ϫ1 , and the dissociation rate constant k off is 0.055 h Ϫ1 , as determined from the equations of the fitted curves. The association rate constant duration of the cell cycle (Fig. 3A). Therefore, the majority of the septin complexes do not turn over GTP. These data strongly support a structural role for septin-bound GTP and GDP, analogous to the role of GTP for ␣-tubulin. In the case of ␣-tubulin, GTP is non-exchangeable (27,28) and thought to be important for ␣-tubulin folding and tubulin dimer assembly. Similarly, septin-bound GTP and GDP might be important for individual septin protein folding and/or septin complex assembly.
Although our in vivo experiments suggest that GTP exchange does not play a major role in the biology of septins during exponential growth, we nevertheless characterized exchange in vitro. GTP exchange could be important for a subset of septins in vivo, and even if it is not, analysis of exchange biochemistry may reveal clues as to the organization of the septin complex. We found very slow binding of exogenous GTP to all the septins, slightly faster to Cdc10, Cdc12, and Shs1 (Figs. 3A and 5; Table II, Column IV). Once exchanged, the rate of dissociation of the bound GTP was slower than its association rate (Fig. 4), which is not consistent with the dissociation rate-limiting the exchange reaction. We can see two possibilities for this discrepancy. One is that the GTP binding we observed in vitro is attributable to GTP binding to a small subpopulation of empty binding sites rather than to exchange of the bound nucleotide. The other is that GTP binding is attributable to exchange, but the dissociation rate we meas-ured is slower than the real dissociation rate. This would be the case if, for example, the septin complexes aggregate over time, because of partial denaturation, and physically prevent the release of the radioactive bound nucleotide.
The properties of the bound guanine nucleotides of the S. cerevisiae septins are remarkably similar to those of the Drosophila septin complex (17) and of the reconstituted mammalian septin complex (18). In particular, the nucleotide:septin protein ratio (1:1) and the ratio of bound GDP:bound GTP (2.2:1) are highly conserved between these organisms, despite different numbers of septin polypeptides in the complex. This evolutionary conservation suggests that GTP/GDP-related biochemistry is critical for the function of septins. We propose that GTP binding is important for stabilizing the septin polypeptides. GTP hydrolysis by certain septin polypeptides (Cdc10, Cdc12, and Shs1) during complex assembly may control the architecture of the septin complex (for example, by forcing certain polypeptides to be neighbors, or by locking the conformation of the septin complex). Indeed, a defined architecture of the septin complex might be required for septins to accomplish their function as molecular scaffolds (15,16) so as to bring together diverse effector molecules in a defined spatial pattern.
Previous experiments showed that mutagenesis of Cdc11conserved amino acids eliminated GTP binding in vitro but failed to reveal a phenotype in vivo at normal temperatures (40). This result agrees with our data and suggests that GTP binding does not play an active, regulatory role for the function of Cdc11. Moreover, this result points toward other mechanisms additional to GTP binding that might stabilize the septin polypeptides.
The caveats of our conclusions pertain to the limitations of the techniques used. Specifically, for the in vivo experiment, we may not have purified a special subpopulation of septin complexes that turn over GTP. Our isolation procedure-immunoprecipitation-defines and thus potentially limits the septin population we analyzed. Also, we may have missed exchange in a single GTP-binding site within the septin complex because of the limited sensitivity of the 15 N-dilution experiment. Although the in vitro filter-binding assay has the potential to identify a single, fast-turnover GTP-binding site, it would miss it if additional cofactors (guanine nucleotide exchange factors, for example) were required for stoichiometric binding or exchange.
Of special interest is Shs1 (Sep7). This protein is present in our preparation as multiple bands that partially overlap with Cdc3. Whereas each individual Shs1 band is sub-stoichiometric to the other septin proteins, the sum of the Shs1 bands represents a stoichiometric amount (Table II, Column II). As was shown by Mortensen et al. (30), Shs1 is a phosphoprotein, and the phosphorylated form migrates slower on SDS-PAGE. Interestingly, one of the slowly migrating forms of Shs1 exchanges and cross-links GTP faster than other polypeptides (Table II, Column IV). It may be that Shs1 can turn over GTP more rapidly than the other septins and that this potentially rapid GTP turnover is regulated by phosphorylation.
An intriguing possibility not ruled out by our experiments is that individual septin proteins, not part of the septin complex, function in vivo. Those might turn over GTP rapidly and behave like conventional GTPases, as suggested by the behavior of individual recombinant septin proteins in vitro (22,23). Given that we used an antibody against one of the septins, Cdc3, to isolate the septin complex, we may have not isolated the other potentially free septin proteins. During our purification procedure, the depletion of Cdc3 from the clarified extract (see "Experimental Procedures") was Ͼ80%. Additionally, we did not detect free Cdc3 protein in our preparation, as evi-  3 and 4), 25-l aliquots were withdrawn. Three samples of 5 l each were processed via the filter-binding assay. The rest of 10 l was mixed with additional ATP and 2-mercaptoethanol at a final concentration of 300 M and 1%, respectively. The mixture was UV cross-linked and subjected to 10% SDS-PAGE. The left panel represents the scan of the Coomassie G-250-stained gel, and the right panel represents the phosphorimager scan. Lanes 1 and 3, 2 M GTP ϩ 300 M ATP; Lanes 2 and 4, 300 M GTP ϩ 300 M ATP. The values determined by filter-binding were 8,500 cpm/5 l for the Lane 1 sample and 150,000 cpm/5 l for the Lane 3 sample. The highest cross-linked band is Shs1, based on the analysis presented in Fig. 1A. Given the results of Mortensen et al. (30), this band should correspond to the phosphorylated form of Shs1. The cross-linked band labeled as Cdc3 might be attributable entirely or in part to Shs1. We assumed it to be attributable to Cdc3 in calculations for Table II. denced by gel filtration and sucrose gradients. We obtained similar results using an anti-Cdc12 antibody. 2 Thus, our data provide no support for a model in which free Cdc3 and Cdc12 represent a major fraction of the functional septin pool.
Although we do not exclude the possibility of a regulatory role for a small subset of septin-bound GTP and GDP, we provide strong evidence that the majority of bound guanine nucleotides play a structural role. Our results support a model of septins being scaffold proteins rather than signaling GTPases (15). The septin complex contains multiple polypeptides, GTP, and GDP in a tight complex. Such a defined architecture may catalyze interactions between different effectors by bringing them in proximity to each other. Additionally, through its polyvalency, the septin complex may help coordinate mechanistically independent processes (spindle positioning, chitin deposition, cytokinesis) in the context of cell division. We think that structural parameters, such as the architecture of the septin complex and its higher order organization in vivo, are key for understanding how septins regulate the processes in which they are involved.