The Mechanism of GTP Hydrolysis by Ras Probed by Fourier Transform Infrared Spectroscopy*

Time-resolved Fourier transform infrared spectroscopy (FTIR) in combination with photo-induced release of 18 O-labeled caged nucleotide has been employed to address mechanistic issues of GTP hydrolysis by Ras protein. Infrared spectroscopy of Ras complexes with nitrophenylethyl (NPE)-[ a - 18 O 2 ]GTP, NPE-[ b - 18 O 4 ]GTP, or NPE-[ g - 18 O 3 ]GTP upon photolysis or during hydrolysis afforded a substantially improved mode assignment of phosphoryl group absorptions. Photolysis spectra of hydroxyphenylacyl-GTP and hydroxyphenylacyl-GDP bound to Ras and several mutants, Ras(Gly 12 )-Mn 2 1 , Ras(Pro 12 ), Ras(Ala 12 ), and Ras(Val 12 ), were obtained and yielded valuable information about structures of GTP or GDP bound to Ras mutants. IR spectra revealed stronger binding of GDP b - PO 32 2 moiety by Ras mutants with higher activity, suggesting that the transition state is largely GDP-like. Analysis of the photolysis and hydrolysis FTIR spectra of the [ b -nonbridge- 18 O 2 , ab bridge- 18 O]GTP isotopomer allowed us to probe for positional isotope exchange. Such a reaction might signal the existence of metaphosphate as a discrete intermediate, a key species for a dissociative mechanism. No positional isotope exchange was observed. Overall, our results support a concerted mechanism, but the transition state seems to have a considerable amount of dissociative character. This work demonstrates that time-re-solved FTIR is highly suitable for monitoring positional isotope the literature 25). A TR-FTIR 1 study on NPE-GTP-Ras has revealed strong binding between the Ras and b -PO 2 2 group of Ras-bound GTP, which led to the suggestion that the transition state is dissociative (24). A Raman study provided an independ-ent vibrational assignment and semiempirical vibrational analysis to simulate the band shifts of g -PO 32 2 group. On the basis of the simulation result, the formation of a weak bond between the attacking water molecule and the g -phosphorus atom was proposed (25). Here we report our results on GTP hydrolysis by Ras using time-resolved FTIR spectroscopy. Four nucleotides, [ a -nonbridge- 18 O 2 ]GTP ([ a - 18 O 2 ]GTP), [ b -non-bridge 18 O 2 , ab -bridge- 18 O, bg -bridge- 18 O]GTP ([ b - 18 O 4 ]GTP), [ g -nonbridge- 18 O 3 ]GTP ([ g - 18 O 3 ]GTP), and [ b -nonbridge- 18 O 2 , ab -bridge- 18 O]GTP ([ b - 18 O 3 ]GTP) were synthesized. Each was caged by the NPE group and prepared as caged nucleotide-Ras complex. IR spectroscopy following the photolysis of NPE-[ a - 18 O 2 ]GTP, NPE-[ b - 18 O 4 ]GTP, or NPE-[ g - 18 O 3 ]GTP bound to Ras allowed a complete assignment of phosphoryl group absorptions. Hydroxyphenylacyl (HPA)-GTP and HPA-GDP were also synthesized and employed in the IR study involving Ras revealed that no change occurred to the sample. All experiments were performed at 21 °C. Time-resolved FTIR experiments were conducted on a Bruker model IFS88 spectrometer equipped with a HgCdTe detector (model KMPV 11-1-J2; Kolmar Technologies). An infrared filter (model W07100-11X; Optical Coating Laboratory) was employed to cut off IR radiation above 2250 cm 2 1 . The lower frequency limit was the cut-off of the BaF 2 plate around 900 cm 2 1 . The spectra were recorded in the rapid scan mode with a mirror velocity of 160 kHz and a spectral resolution of 4 cm 2 1 (41). This resulted in a single interferogram scan time of 125 ms (double sided, forward/backward). Before each experiment, a series of spectra were collected over a half hour period to check the base-line stability.

for phosphoryl transfer reactions, which ATP and GTP hydrolysis are examples of, are usually described by a structure somewhere between dissociative and associative extremes (6). In the fully dissociative mechanism, reactions proceed via a metaphosphate intermediate, which is subsequently attacked by the nucleophile. In contrast, a fully associative mechanism involves a trigonal bipyramidal intermediate, followed by the departure of the leaving group. Between these two limits exist concerted pathways featuring trigonal bipyramidal transition states with no distinct intermediate. For GTP hydrolysis by Ras, several distinct models have been developed to explain the biochemical and mutagenesis data (4,5,7,8). While some groups have been looking for a general base that activates the water molecule in the active site to render it a better nucleophile (8 -10), others emphasized the contribution from transition state stabilization by Ras (7,11). An underlying feature common to these models is the assumption that the transition state of the GTP hydrolysis is associative in nature. It is based on the observation that the ␥-phosphoryl group is surrounded by positive charges capable of neutralizing the negative charges accumulating on the ␥-phosphoryl oxygens in an associative transition state (7,(12)(13)(14). Evidence supporting an associative mechanism includes the following: (i) an observed linear free energy relationship between the rates of GTP hydrolysis by Ras mutants and the pK a values of the ␥-PO 3 2Ϫ group of GTP bound to these mutants (5); (ii) the observation that the catalytically important residues interact extensively with the transition state analog GDP-AlF 4 Ϫ in the crystal structures of Ras-GDP-AlF 4 Ϫ -GTPase-activating protein (15), G i␣1 -GDP-AlF 4 Ϫ (16), and transducin ␣-GDP-AlF 4 Ϫ (17). This assumption has been strongly challenged by Maegley et al. (4), who argued that the transition state is more likely dissociative in nature based on the following: (i) a wealth of physical organic data that implicates a dissociative, metaphosphate-like transition state in solution reactions of phosphate monoesters, acyl phosphates, phosphorylated amines (6), and phosphoanhydride (18); and (ii) the fact that the localization of positively charged side chains and metal ions in an enzymatic active site may not change the dissociative transition state of a solution reaction to a more associative one in an enzymatic reaction, as shown for the Escherichia coli alkaline phosphatase (19) and other phosphatases (20 -23).
The resolution of this question clearly requires further studies on this systems and probably even on other kinases, ATPases, GTPases, and phosphatases. Vibrational spectroscopy can provide detailed, mechanistically important information about the structures of substrate(s), cofactor(s), even transition state analog bound to the enzymes. Time-resolved vibrational spectroscopy has been established as a valuable tool to study kinetics of enzymatic reactions. In the case of Ras, two studies employing vibrational spectroscopy have appeared in the literature (24,25). A TR-FTIR 1 study on NPE-GTP-Ras has revealed strong binding between the Ras and ␤-PO 2 Ϫ group of Ras-bound GTP, which led to the suggestion that the transition state is dissociative (24). A Raman study provided an independent vibrational assignment and semiempirical vibrational analysis to simulate the band shifts of ␥-PO 3 2Ϫ group. On the basis of the simulation result, the formation of a weak bond between the attacking water molecule and the ␥-phosphorus atom was proposed (25). Here we report our results on GTP hydrolysis by Ras using time-resolved FTIR spectroscopy. Four nucleotides, [␣-nonbridge- 18 18 O 3 ]GTP bound to Ras allowed a complete assignment of phosphoryl group absorptions. Hydroxyphenylacyl (HPA)-GTP and HPA-GDP were also synthesized and employed in the IR study involving Ras mutants. The photolysis chemistry of NPE-GTP and HPA-GTP is shown in Scheme 1. The band shifts of the vibrational modes of GTP or GDP bound to Ras mutants compared with that bound to wild type Ras were discussed in the context of the mechanism of GTP hydrolysis. We also employed TR-FTIR to probe the positional isotope exchange (PIX) in the process of [␤- 18 (33).
Carbamate kinase was previously suggested to be useful in the synthesis of labeled GTP (34). We were able to use this enzyme to make [␤- 18  ]phosphate (400 l, 500 mM, tetrabutylammonium form), sodium acetate (160 l, 1 M, pH 4.9) and potassium cyanate (400 l of 1 M) were mixed and incubated at 37°C for 15 min. An additional 400 l of KCNO was added, and the mixture was incubated for an additional 15 min. The solution was added to a mixture of GDP (80 mg in 3.2 ml of water), Tris-HCl (800 l, 1 M, pH 8.0), MgSO 4 (160 l, 1 M), and carbamate kinase (50 units). A white precipitate of an unknown nature appeared upon the mixing of the two solutions. The reaction nevertheless proceeded when we kept the solution on a rocking plate for 5 h at 37°C. Afterward, the sample was purified according to the procedure described above for [␤- 18 O 4 The labeled nucleotides were examined by electrospray mass spectrometry. The degree of 18  The coupling between the GTP and the nitrosoacetophenone cage was carried out according to a procedure described previously (36).
HPA-GTP was synthesized by coupling HPA-phosphate with GDP. The synthesis of HPA-phosphate was performed according to a published procedure (37). Activation of GDP by carbonyldiimidizole allowed the coupling between GDP and HPA-P i to generate HPA-GTP. This coupling was carried out according to a procedure described by Martin Webb (29). The reaction mixture was fractionated by either a DEAE or a Q column. Although HPA-GTP was the sole expected product, the main product was HPA-GDP (80%) with HPA-GTP as only a minor product (20%). The identities of HPA-GDP and HPA-GTP were firmly established by comparing the photolysis IR difference spectra with those of NPE-GDP and NPE-GTP (see "Results") and by mass spectrometry. This is a very surprising finding, since a very similar procedure resulted in HPA-ATP as the only product (37,38). No further attempt was made to understand this observation.
Preparation of Caged GTP-Ras and Caged GDP-Ras Complex-The purification of wild type Ha-Ras was described previously (39). The Ras mutants were purified similarly from E. coli strains harboring the clones of respective mutants. The tightly bound GDP has to be removed from the Ras protein before the complex with the caged-GTP can be formed. Degradation of GDP to guanosine by alkaline phosphatase affords efficient replacement by caged GTP. This exchange process was performed according to a published procedure with some modifications (40). In a typical preparation, 2.5 ml of Ras (3 mg/ml) in stock buffer (50 mM Hepes, 200 mM NaCl, 10 mM MgCl 2 , 50% glycerol, pH 7.5) was applied on a Sephadex PD-10 column, which was preequilibrated with exchange buffer (30 mM Tris, 200 mM ammonium sulfate, pH 8.0). The protein was then eluted by 3.5 ml of exchange buffer. The eluted protein solution was concentrated by Centricon-10 followed by readding the exchange buffer to further deplete the Mg 2ϩ . A 3-fold equivalent of caged GTP or caged GDP and 70 l of alkaline phosphatase on agarose with storage buffer (Roche Molecular Biochemicals) were added to the protein solution. The mixture was placed on a rotating plate to ensure an even distribution of alkaline phosphatase in solution during the process. After the reaction was complete, the mixture was filtered to SCHEME 1. Photolysis of HPA-GTP and NPE-GTP. SCHEME 2. Positional isotope exchange.
remove the alkaline phosphatase on agarose and a small amount of denatured Ras. The protein solution was subject to repeated concentration-dilution-concentration cycles (3 ϫ 500 l of H 2 O) with Centricon-10 to remove the buffer, guanosine, and excess caged GTP or caged GDP.
To improve the yield of the caged nucleotide-Ras complex, a small amount of caged nucleotide can be added to the water used in the washing process. The final caged GTP-Ras complex in water was divided into 1-mg portions, and each was lyophilized. Photolysis and Rapid Scan FTIR Procedure-The infrared sample cell consists of two BaF 2 plates separated by a 12-m mylar spacer. To load the Ras into the IR cell, 1 mg of lyophilized protein was first dissolved in 5 l of reaction buffer (200 mM Tris-HCl, 40 mM MgCl 2 , pH 8.0). The sample was briefly centrifuged to remove any precipitate, and the protein drop was pipetted to the center of a BaF 2 plate, with spacer in place. The second plate was then carefully mounted on top in order to spread the protein drop across the area confined by the spacer. The plates were sealed by vacuum grease to prevent the sample from drying. For the photolysis of NPE-caged compound, 355-nm laser pulses generated by a Nd:YAG laser (model Quanta Ray DCR-2; Spectra Physics) at 10 Hz were employed. The HPA cage has no absorption at 350 nm, so a dye laser and a wavelength extender (model PDL-1; Spectra Physics) were combined to produce 308-nm light. The light beam covered the entire sample area of 1-cm diameter. For most of the experiments, photolysis was performed with power 1 mJ/pulse for a period of 30 s. For some experiments involving NPE cages, higher pulse energy (6 mJ) and correspondingly shorter irradiation time (5 s) was used. The photolysis did not induce photochemistry to the aromatic groups of the Ras. When the Ras-GDP complex was subjected to the same photolysis method, infrared spectroscopy revealed that no change occurred to the sample. All experiments were performed at 21°C.
Time-resolved FTIR experiments were conducted on a Bruker model IFS88 spectrometer equipped with a HgCdTe detector (model KMPV 11-1-J2; Kolmar Technologies). An infrared filter (model W07100-11X; Optical Coating Laboratory) was employed to cut off IR radiation above 2250 cm Ϫ1 . The lower frequency limit was the cut-off of the BaF 2 plate around 900 cm Ϫ1 . The spectra were recorded in the rapid scan mode with a mirror velocity of 160 kHz and a spectral resolution of 4 cm Ϫ1 (41). This resulted in a single interferogram scan time of 125 ms (double sided, forward/backward). Before each experiment, a series of spectra were collected over a half hour period to check the base-line stability. Upon photolysis, data collection by the spectrometer was triggered by a TTL pulse generated manually through an external pulse generator. During the first 5 min after photolysis, 240 of the 125-ms scans were recorded and averaged, furnishing a 30-s time resolution spectrum. At later times, averaging was extended over one to several min as the kinetics of the observed process occurred on a time scale of hours.
Calculation of Difference Spectra-The spectra presented in this paper are difference spectra as defined by Ϫlog (I 2 /I 1 ), where I 2 and I 1 refer to two single beam spectra of the same sample at different times. Hence, spectra manifest exclusively changes of the sample while suppressing the static absorptions, which are 2-3 orders of magnitude larger.

Photolysis and Hydrolysis Spectra of NPE-GTP-Ras
Complexes-Laser flash photolysis of caged GTP-Ras complexes combined with FTIR monitoring furnishes two types of difference spectra that reveal the absorptions of the transient Rasbound GTP. The first is the photolysis difference spectrum, in which the positive bands originate from GTP-Ras. The other, the hydrolysis spectrum, shows the GTP modes as negative bands. The absorbance scale of the spectra has been normalized such that the extent of photolysis is about the same for all samples, which is indicated by the same amount of depletion of caged GTP at 1344 cm Ϫ1 (symmetric stretch of NO 2 group). Fig. 2 depicts the GTP-Ras spectra of the same isotopic modifications in the form of hydrolysis difference spectra. Each trace was obtained by calculating the ratio of a 5-min average, taken 50 min after photolysis, against a 2-min average at 30 min after photolysis. The hydrolysis spectra for [␤-18 O 3 ]GTP is not duplicated here, since it is shown in Fig. 6. The hydrolysis spectra for unlabeled GTP, [␤- 18 O 4 ]GTP, and [␥-18 O 3 ]GTP were normalized by the amplitude difference of peaks at 1263/1236 cm Ϫ1 . Because of the absence of these peaks in the spectra for [␣-18 O 2 ]GTP, a small protein absorption at 1324 cm Ϫ1 was used for its normalization. Fig. 3 shows the double difference photolysis spectra.
Photolysis Spectra of HPA-GTP and HPA-GDP Complexed with Ras Mutants-A number of Ras mutants have either impaired or elevated GTPase activity compared with wild type Ras. IR spectra can reveal structural differences between the GTP or GDP bound to wild type Ras(Gly 12 ) and Ras mutants. A correlation between these structural differences and reaction rates could potentially shed light on the mechanism of GTP hydrolysis. For this reason, we have performed TR-FTIR experiments on Ras(Gly 12 )-Mn 2ϩ , Ras(Pro 12 ), wild type Ras-(Gly 12 ), Ras(Ala 12 ), and Ras(Val 12 ) with HPA-GTP or HPA-GDP as caged substrate. Ras(Gly 12 )-Mn 2ϩ refers to wild type Ras with Mn 2ϩ , instead of Mg 2ϩ , as metal cofactor. In experiments with HPA-GTP and Ras complexes, curve fitting of IR absorption change as a function of time reveals first order kinetic constants of 0.018, 0.015, 0.011, 0.0027, and 0.0012 min Ϫ1 for Ras and Ras mutants listed in the above sequence, respectively. The photolysis spectra are shown in Fig. 4. Fig. 5 shows photolysis spectra of HPA-GDP bound to Ras or its mutants. Precision in locating the absorption becomes very important in the discussion of bandshifts, since the observed shifts are often small if at all discernible. Although the spectra were collected at 4-cm Ϫ1 resolution, the precision of peak position is much higher. For example, locating peaks in six separate photolysis spectra collected on HPA-GTP-Ras(Gly 12 ) of the same batch gave S.D. values of 0.17 and 0.12 cm Ϫ1 for bands of 1215.5 cm Ϫ1 and 1143.4 cm Ϫ1 , respectively. Larger standard deviations, about 0.5 cm Ϫ1 , were observed in spectra for samples of different batches, suggesting that the subtle conformational differences between these samples contribute significantly to the variations. The S.D. values reported in Fig. 4 were obtained on 17 spectra collected on HPA-GTP-Ras(Gly 12 ) samples originating from eight batches. Those reported in Fig. 5 for HPA-GDP-Ras(Gly 12 ) were obtained from 10 spectra, all of different batches.  Ras complex for the first 20 min after photolysis. The single beam spectrum recorded during the first 30 s after photolysis was used as reference. This presentation offers a chance to observe transient IR peaks of the isotopically scrambled GTP, if it is generated in the early phase of the reaction. An alternative way of detecting scrambled GTP is by the hydrolysis difference spectra, presented in Fig. 7. Here, spectral changes of [␤- 18 18 O 3 ]GTP) allowed us to make a complete assignment of phosphoryl group stretching absorptions of GTP and GDP bound to Ras. This is essential for studying the structures of GTP or GDP bound to Ras and inferring the mechanisms of GTP hydrolysis. It is also a prerequisite for interpreting PIX spectra. Assignment of GTP absorptions was derived from both photolysis spectra (Fig. 1) and hydrolysis spectra (Fig. 2) of the GTP-Ras complex. GDP assignment was based on the hydrolysis spectra (Fig. 2) and photolysis spectra of HPA-GDP-Ras (Fig. 5c). The results for the assignment are summarized in Tables I and II for GTP and GDP, respectively.
The negative peak at 1263 cm Ϫ1 in the hydrolysis spectrum of GTP-Ras (Fig. 2a) remains unchanged in the hydrolysis spectra of both ␤-labeled and ␥-labeled GTP-Ras (Fig. 2, c and  d) but shifts down to 1216 cm Ϫ1 upon labeling of the ␣-phosphoryl group (Fig. 2b). Therefore, this peak is assigned to a (␣-PO 2 Ϫ ). The corresponding positive peak is readily discerned in the photolysis spectra of ␤or ␥-labeled samples (Fig.  1, c-e), but clearly shifts down in the case of ␣-labeled GTP (Fig. 1b). Its presence in the spectra for unlabeled GTP (Fig. 1a) is obscured by its overlap with three negative peaks due to the three asymmetric PO 2 stretch modes in the caged GTP-Ras complex. In a recent FTIR study of caged GTP-Ras photolysis and hydrolysis by Cepus et al. (24) using singly 18 O-labeled GTP at the ␣, ␤, or ␥ position, the 1263 cm Ϫ1 band was also observed in the hydrolysis spectra, but no spectral assignment was given. The small residual peaks at 1263 and 1242 cm Ϫ1 in the hydrolysis spectra of [␣-18 O 2 ]GTP (Fig. 2b) (24). This is supported by the red shift of this band to 1183 cm Ϫ1 upon labeling of the ␤-phosphoryl group (Fig. 1, c  and e).
The ␣and ␤-PO 2 Ϫ symmetric stretching vibrations s (␣-PO 2 Ϫ ) and s (␤-PO 2 Ϫ ) and the ␥-PO 3 2Ϫ vibration a (␥-PO 3 2Ϫ ) are expected around 1100 cm Ϫ1 (42,43). Four peaks (1158, 1142, 1124, and 1090 cm Ϫ1 ) are observed in this region (Fig. 1a). Bands at 1124 and 1090 cm Ϫ1 shift to the red upon labeling at either ␣or ␤-positions ( Fig. 1, b, c, and e) and therefore are assigned to coupled symmetric stretching modes of ␣and ␤-phosphoryl groups. This is consistent with a recent Raman spectroscopic study on GTP-Ras, in which the 1124-cm Ϫ1 absorption is attributed to the in-phase combination and the 1090-cm Ϫ1 band is attributed to the out-of-phase combination (25). Not all of the positions of down-shifted peaks are clear in the corresponding spectra ( Fig. 1, b, c, and e). The values reported in Table I were obtained in double difference photolysis spectra, which were calculated by subtracting the photolysis spectra of unlabeled NPE GTP-Ras from those of 18 O samples (Fig. 3). This pair of peaks due to symmetric stretching mode also shift down to 1106 and 1089 cm Ϫ1 in the spectra for ␥-labeled samples (Fig. 1d), indicating some coupling of the symmetric ␣and ␤-PO 2 Ϫ stretching modes with s (␥-PO 3 2Ϫ ). The other two peaks in this region, 1158 and 1142 cm Ϫ1 , remain the same in the case of ␣or ␤-labeled sample (Fig. 1, b, c, and e) but shift to 1138 and 1118 cm Ϫ1 , respectively, in the case of ␥-labeled GTP (Fig. 1d). Given the intensities and isotopic shifts of these peaks, the only interpretation we can conceive of is assignment to the two a (␥-PO 3 2Ϫ ) modes. It should be noted that double difference spectra must be interpreted with caution. First, a double difference spectrum includes spectral features arising from four sample states. Consequently, there is more overlap among the infrared absorptions, and each positive or negative peak could originate from two sample states. Second, 18 O labeling at one position can induce a relatively large change in the amplitude of another absorption through vibrational coupling, even when the two vibrational modes seem decoupled based on frequencies. For example, although the frequencies of a (␣-PO 2 Ϫ ) and a (␤-PO 2 Ϫ ) are decoupled from the above analysis, 18 O labeling at the ␣-position seems to increase the amplitude of a (␤-PO 2 Ϫ ) at 1216 cm Ϫ1 (Fig. 3a). The absorption undergoing a change in amplitude can be revealed in the double difference spectra and misinterpreted as a shifted absorption. Based on these considerations, double difference spectra were used in this paper only to locate peaks labeled in Fig. 3 whose true positions were not clear in difference spectra. s (␥-PO 3 2Ϫ ) is expected in the region from 900 to 1000 cm Ϫ1 (42,43) and was assigned to 917 cm Ϫ1 according to the Raman study (25). A positive peak at this position would be overwhelmed by the strong negative peak around 913 cm Ϫ1 in Fig.  1a. The peak 951 cm Ϫ1 in Fig. 1a shifts to 939 cm Ϫ1 in [␤-18 O 3 ]GTP spectra (Fig. 1e) and shifts further down to 918 cm Ϫ1 in the spectra for [␤- 18 O 4 ]GTP (Fig. 1c). It shifts only a few wave numbers down upon labeling at the ␥-position (Fig.  1d). These large shifts upon ␤-labeling suggest a very localized POP backbone stretch, possibly coupled with symmetric ␤-PO 2 Ϫ . Peaks at 897 and 880 cm Ϫ1 show a similar pattern of shift upon 18 O labeling at the ␤or ␥-positions and are assigned to POP backbone stretching modes as well.
The GDP-Ras complex is expected to have bands due to a (␣-PO 2 Ϫ ), s (␣-PO 2 Ϫ ), a (␤-PO 3 2Ϫ ), s (␤-PO 3 2Ϫ ), and (POP). The only positive peak at 1236 cm Ϫ1 in the asymmetric PO 2 Ϫ stretch region of the hydrolysis spectra (Fig. 2a) is assigned to a (␣-PO 2 Ϫ ), since it shifts to 1196 cm Ϫ1 for ␣-labeled GTP (Fig. 2b) but remains at the same position in the case of the ␤and ␥-labeled species (Fig. 2, c and d). This contrasts with a previous assignment of this absorption to a ␤-PO 3 2Ϫ stretching mode (24). In the HPA GDP-Ras photolysis spectrum (Fig. 5c), three peaks are observed at 1137, 1129 (shoulder), and 1100 cm Ϫ1 in the region from 1150 to 1100 cm Ϫ1 , where the bands due to s (␣-PO 2 Ϫ ) and a (␤-PO 3 2Ϫ ) are expected. The band at 1100 cm Ϫ1 is readily visible in the hydrolysis spectra (Fig. 2a), while the peaks at 1137 and 1129 cm Ϫ1 are strongly overlapped by the negative GTP bands at 1142 and 1124 cm Ϫ1 . However, the 1137-cm Ϫ1 band is exposed in the hydrolysis spectra of the ␥-labeled sample (Fig. 2d). In the hydrolysis spectra of [␤-18 O 4 ]GTP (Fig. 2c), the 1100-cm Ϫ1 band shifts to 1072 cm Ϫ1 , but a new peak appears at 1105 cm Ϫ1 . The most likely counterpart of this band in the unlabeled GDP-Ras complex is 1137 cm Ϫ1 as seen in the HPA GDP-Ras photolysis spectrum (Fig.  5c). The parallel shift of 1137 and 1100 cm Ϫ1 to 1105 and 1072 cm Ϫ1 for [␤- 18 O 4 ]GTP was also observed in the difference spectra between the end state (after hydrolysis is over) and initial state (before photolysis), in which the overlapping of negative bands happens to be absent (data not shown). Additional support for this assignment comes from the photolysis spectra of    the HPA GDP-Ras in the presence of Mn 2ϩ (Fig. 5a). Upon metal ion exchange, the pair of peaks at 1137 and 1129 cm Ϫ1 shift to 1133 and 1121 cm Ϫ1 . The relative intensities of these peaks suggest that the 1137-cm Ϫ1 band shifts down to 1121 cm Ϫ1 and that the 1129 cm Ϫ1 band shifts up to 1133 cm Ϫ1 . Crystallographic and EPR studies have established that the ␤-PO 3 2Ϫ group, instead of the ␣-PO 2 Ϫ group, interacts directly with the metal ion in the GDP binding site (44 -46). The large shift of 1137 cm Ϫ1 (16 cm Ϫ1 ) and small shift by 1129 (4 cm Ϫ1 ) justify the assignment of 1137 cm Ϫ1 to a (␤-PO 3 2Ϫ ) and 1129 cm Ϫ1 to s (␣-PO 2 Ϫ ). As in the case of GTP-Ras, a splitting of the absorption of the terminal PO 3 2Ϫ group asymmetric stretch is observed. The released free phosphate (mostly HPO 4 2Ϫ at pH 8.0) gives rise to a peak at 1079 cm Ϫ1 in the hydrolysis spectra (Fig. 2a) (47). The peak at 913 cm Ϫ1 in Fig. 5c is the only positive peak in that region. It is assigned to s (␤-PO 3 2Ϫ ), since it shifts down to 879 cm Ϫ1 in the hydrolysis spectra of [␤ 18 O 4 ]GTP (Fig. 2c), which generates a ␤-P 18 O 3 2Ϫ group upon hydrolysis.
Compared with two previous studies, the vibrational assignments reported here are more complete (24,25). With the exception of s (␥-PO 3 2Ϫ ), all of the major stretching modes have been accounted for, and all of the major peaks sensitive to 18 O labeling have been assigned. One intriguing new insight is the observed splitting of the absorption of the terminal PO 3 2Ϫ group of Ras-bound GTP or GDP. In principle, this can result from the existence of two major conformational states in equilibrium or a removal of the degeneracy. For the Ras-bound GDP, the latter is likely to be the case, since no multiple conformation has been implicated by the crystallographic or NMR studies on GDP-Ras complex structure. In the case of GTP-Ras complex, it is possible that coexistence of alternative conformational states is causing the splitting. Two alternative conformations have been proposed for GTP-Ras complex (48,49). Loop 2 of the Ras is disordered in the crystal structures (13,14,46) but has to adopt a particular conformation for catalysis (7,9). Further IR work on Ras with mutation in the PO 3 2Ϫ binding region and ab initio calculations might resolve this issue.
In the Raman study (25), the authors proposed that the water molecule in the active site forms a weak bond with the ␥-phosphorus atom of bound GTP. The central evidence is that as the water-phosphorus distance decreases in the simplified active site model, the experimental band shift of s (␥-PO 3 2Ϫ ) was predicted by empirical formulas that relate geometry to vibrational frequencies of the PO 3 2Ϫ group. However, the same active site model and empirical equations also predicted a 1-cm Ϫ1 red shift of asymmetric stretch a (␥-PO 3 2Ϫ ) (Table III of  Ref. 25). This does not agree with our experimental data. Our photolysis spectrum of Ras-bound [␥-18 O 3 ]GTP (Fig. 1d) allows the assignment of the 1158-and 1142-cm Ϫ1 bands to a (␥-PO 3 2Ϫ ). Compared with 1121 cm Ϫ1 in solution (43), this is an upshift of at least 21 cm Ϫ1 . A more sophisticated calculation, possibly including GTP-Ras interactions, is probably needed to explain the removal of degenaracy and band shifts.
IR Spectra of GTP Bound to Ras and Ras Mutants-The previous FTIR study has identified a critical role of ␤-PO 2 Ϫ in the GTP hydrolysis by Ras (24). The frequency of a (␤-PO 2 Ϫ ) of Ras-bound GTP is unusually low as compared with free GTP, revealing strong hydrogen bonding and electrostatic interactions between the ␤-PO 2 Ϫ group of GTP and the active site. This strong binding may preferentially stabilize a dissociative transition state over ground state, since the interactions get stronger in a dissociative transition state as negative charge builds up on the ␤-PO 2 Ϫ . The same interactions are not thought to catalyze a reaction with an associative transition, since no charge builds up on ␤-PO 2 Ϫ group. Therefore, a downshift of a (␤-PO 2 Ϫ ) upon Ras binding suggests a dissociative transition state for GTP hydrolysis. If this notion is indeed correct, it is conceivable that mutants with weaker binding to the ␤-PO 2 Ϫ group of GTP would have diminished hydrolysis activity, and the opposite trend is predicted for the mutants with an enhanced hydrolysis rate. To further explore this possibility, we performed experiments on complexes of HPA GTP with Ras-(Gly 12 )-Mn 2ϩ , Ras(Pro 12 ), Ras(Gly 12 ), Ras(Ala 12 ), or Ras-(Val 12 ), with the photolysis spectra shown in Fig. 4. The enzymes are listed according to GTPase activity. It is interesting to see that a (␤-PO 2 Ϫ ) shifts down 10 cm Ϫ1 upon the substitution of Mg 2ϩ by Mn 2ϩ , indicating stronger interaction of ␤-PO 2 Ϫ in the case of the Ras-Mn 2ϩ complex, which is accompanied by an enhanced hydrolysis efficiency. However, no systematic upshift of a (␤-PO 2 Ϫ ) can be observed in the case of Ras(Pro 12 ), Ras(Gly 12 ), Ras(Ala 12 ), Ras(Val 12 ) (i.e. in the direction of decreased hydrolysis efficiency). In fact, a (␤-PO 2 Ϫ ) of Ras(Val 12 ), the least effective enzyme of the five, also shifts down to 1211 cm Ϫ1 . We conclude that the binding strength of Ras or its mutants to GTP simply does not correspond to the same binding strength in the transition state. This point will be discussed below in conjunction with the photolysis spectra of HPA GDP bound to Ras mutants.
Another interesting absorption to study is a (␥-PO 3 2Ϫ ). Early studies have found that the terminal ␥-PO 3 2Ϫ group of GTP bound to different Ras mutants have different pK a values, which eventually led to the proposal of a substrate-assisted mechanism for GTP hydrolysis by Ras (10). It is expected that the Ras-bound GTP with the higher pK a has more electron density in the ␥-phosphoryl group, which leads to higher frequency of vibrational modes. This possibility was examined on the peak at 1143 cm Ϫ1 in our photolysis spectra of mutant complexes (Fig. 4). The peak at 1158 cm Ϫ1 was not used, since it is on the shoulder of 1142 cm Ϫ1 and cannot be located accurately. With a S.D. value of 0.45 cm Ϫ1 for this band, the relative shifts of this absorption for Ras(Pro 12 ), Ras(Gly 12 ), and Ras(Ala 12 ) are not significant. However, the 1.34 cm Ϫ1 downshift of Ras(Val 12 ) relative to Ras(Gly 12 )-Mn 2ϩ is 3 times the S.D. value and should be considered significant. The direction of the shift is also consistent with the fact the pK a of Ras-(Val 12 )-bound GTP is 0.8 unit lower than that of GTP in Ras-(Gly 12 )-Mn 2ϩ complex (10). Since there is only one data point that can be interpreted with confidence, we conclude that our infrared data is consistent with, but does not prove, the observed correlation between pK a values of mutant-bound GTP and catalytic efficiency by Ras mutants, which was established by 31 P NMR and biochemical methods.
IR Spectra of GDP Bound to Ras and Ras Mutants-It is interesting to see the degeneracy of a (␤-PO 3 2Ϫ ) removed, giving rise to two peaks at 1137 and 1101 cm Ϫ1 . We studied how these two peaks shift in the spectra of Ras mutants. The S.D. values are 0.03 and 0.37 cm Ϫ1 for 1137 and 1101 cm Ϫ1 , respectively, as determined by 10 spectra of independently prepared HPA GDP-Ras. The peak at 1101 cm Ϫ1 does not have a consistent shift pattern. However, the peak at 1137 cm Ϫ1 shifts consistently to higher frequencies for mutants with reduced GTPase activity. This reveals that the mutants of lower hydrolysis rate also have weaker binding to the ␤-PO 3 2Ϫ group of GDP. It is noted that the precision of the peak at 1237 cm Ϫ1 is the highest among the peaks under discussion. In fact, the largest variation in this peak for any Ras mutant is 0.11 cm Ϫ1 . With this small margin of error, it can be stated with confidence that the shifts shown by the peak at 1137 cm Ϫ1 are statistically significant. Since an enzyme catalyzes a reaction by preferentially binding to the transition state of the chemical reaction, this shift pattern reveals a positive correlation between the binding to transition state and the binding to ␤-PO 3 2Ϫ of GDP by Ras mutants. This suggests that the transition state is GDP-like (i.e. has significant dissociative character).
One question that needs to be addressed is whether the small band shifts are consistent with the magnitude of the changes in the hydrolysis rate. Although the substitution of Mg 2ϩ by Mn 2ϩ leads to a 15-cm Ϫ1 downshift, the largest shift for mutants is only 1.3 cm Ϫ1 for Ras(Val 12 ), whose GTPase activity is about 10 times lower than Ras(Gly 12 ). We are not aware of another study on the correlation between vibrational frequency and reactivity of the phosphoryl groups. However, a 2-cm Ϫ1 shift of the carbonyl stretch is enough to induce a 2-fold change in the rate of deacylation in the case of acyl-serine proteases (50). Considering the vast difference in the chemical nature and vibrational modes of these two functional groups, we argue that a larger sensitivity of reactivity to change in vibrational frequency for phosphoryl groups as observed here is indeed plausible.
It remains to be explained why there is no correlation between catalytic efficiency and binding strength of the GTP ␤-PO 2 Ϫ moiety by Ras mutants. Crystallographic studies have shown that metal ion, the side chain of lysine 16, and the main chain of loop 1 are involved in interaction with ␤-PO 2 Ϫ of GTP (13,14). We propose that a local conformational change has to occur to accommodate the transition state. As a consequence of this, an initially tighter binding in the ground state may not be favored in the transition state. The Ras conformation in the transition state is probably close to that in Ras-GDP complex, which accounts for the observed positive correlation between hydrolysis efficiency and binding of GDP ␤-PO 3 2Ϫ moiety. This idea is consistent with crystallographic and biochemical studies on this issue. Crystallographic studies suggested that amino acids other than glycine at position 12 would interfere with the transition state complex (7). By substituting glycine 12 with a number of unusual amino acids, Schultz and coworkers (51) found that only a replacement by amino acids that can adopt unusual backbone conformations allowed Ras to retain GTPase activity similar to wild type protein. These data suggest that the steric hindrance of alanine and valine side chains and their inability to relieve unfavorable side chaintransition state interaction at position 12 cause low catalytic efficiency for the respective mutants, although they bind to ␤-PO 2 Ϫ of GTP equally well or even better, as in the case of Ras(Val 12 ). In the absence of this steric hindrance, Ras-Mn 2ϩ , which binds to ␤-PO 2 Ϫ of GTP better than wild type Ras, is also more catalytically efficient.
Positional Isotope Exchange?-While the study on Ras mutants points to a largely dissociative transition state, it does not tell us whether a discreet metaphosphate intermediate exists in the reaction pathway. This information can be potentially provided by studying PIX reaction. The knowledge of peak assignments of the GTP phosphoryl groups is a prerequisite for finding out whether or not isotopic scrambling takes place. In the case of [␤- 18 O 3 ]GTP, the ␤-P 18 O 2 Ϫ group would be converted to ␤-P 18 O 16 O Ϫ for 2 ⁄3 of the population if isotopic scrambling were to occur. Two features in the asymmetric stretch region of phosphoryl groups make this mode ideal for spectroscopic detection of the exchange reaction. First, the analysis presented under "Band Assignment of Phosphoryl Group-associated IR Absorption" indicates that the asymmetric stretch frequencies of ␣and ␤-PO 2 Ϫ groups are not coupled. They depend only on their local isotopic composition and are insensitive to the labeling at other phosphoryl groups and bridge positions. Second, the asymmetric stretching modes of ␣and ␤-phosphoryl groups give rise to fairly sharp and well separated peaks. The absorptions of a (␤-P 16 18 O-labeled at the nonbridge ␤-position at 1205 cm Ϫ1 . By contrast, spectral changes in the PO 2 Ϫ symmetric stretch region as a consequence of scrambling would be much less interpretable because of the coupling of the symmetric stretches of the ␣and ␤-phosphoryl groups.
The kinetic behavior of any isotopic scrambling would depend on the relative ease of the PIX reaction compared with the hydrolysis reaction. If PIX proceeds much faster than hydrolysis, a rapid equilibrium between as-synthesized [␤- 18 18 O 3 ]GTP in the hydrolysis spectra. We did not notice any sign of growth at 1200 cm Ϫ1 even as long as 100 min after start of the hydrolysis, as can be seen from the 5-min time windows plotted in Fig. 7. We conclude that the infrared spectroscopic study does not reveal any positional isotope exchange during GTP hydrolysis by Ras.
We can conceive of two reasons that would account for the lack of PIX reaction even if metaphosphate were an intermediate. One is that the rotation of the ␤-PO 3 2Ϫ group of GDP could be suppressed during the lifetime of the metaphosphate. The other possibility is that the reaction rate of metaphosphate with water is faster than the rate of back reaction. The first factor seems to depend on the ratio of the lifetime of metaphosphate to the time required to reach torsional equilibrium. In this case, a suppression of the rotation is very likely, considering the strong binding between Ras and the ␤-PO 3 2Ϫ group of GDP. However, this scenario involving a system in equilibrium probably underestimates the chance for a rotation to occur. The relaxation of GDP from a conformation in the transition state to one in the Ras-GDP complex, and the reverse process in the case of recombination, are invariably coupled to rotational movement of the PO 3 2Ϫ group of GDP. Crystal structures of Ras-GDP, Ras-GPPNP, and Ras-GDP-AlF 4 Ϫ seem to support this argument (11,14,46). When the guanosine part of the nucleotides in the three complexes were aligned, large conformational differences were observed between the three ␤-phosphoryl and even the three ␣-phosphoryl groups. Although it is dangerous to take these structures as the real structures on the reaction pathway, this finding does illustrate how the relaxation of both GDP and Ras in response to the dissociation/ recombination process can activate the internal rotations of ␤-PO 3 2Ϫ of GDP. Regarding the second possibility, it is unlikely that metaphosphate has similar reactivity toward GDP and H 2 O, because GDP is a much better nucleophile than water. Also, while GDP is always aligned for nucleophilic attack, the water molecule may need to be aligned by Ras residues (7,8). Hence, metaphosphate is expected to undergo a large number of dissociation/recombination events with GDP before it reacts with water, and this large number of forward/backward steps should bring the PIX reaction to a detectable level. Based on these considerations, we conclude that the absence of positional isotope exchange supports a concerted mechanism while disfavoring a mechanism with metaphosphate as a discrete reaction intermediate. This result reinforces a previous finding from a stereochemical study that hydrolysis occurs with inversion of configuration at the ␥-phosphorus (52). Both observations are best described by a concerted mechanism with a single transition state.
However, the absence of a positional isotope exchange reaction does not rule out the existence of the metaphosphate intermediate on an ultrafast time scale (less than nanoseconds). As the intermediate becomes increasingly unstable, the selectivity for nucleophile disappears, and the backward commitment is no longer dominant, a PIX reaction is not expected.
It is important to add that our work rules out the existence of a reversible reaction between the GDP and the phosphate at the enzymatic site. If phosphate were produced and recombined with GDP, the back reaction would certainly be slow on the time scale of internal rotation of the ␤-phosphoryl group of GDP in light of the inherent stability of phosphate. Hence, a PIX reaction is expected to occur in this case as well. Isotopic scrambling of ␤␥-bridge oxygen to ␤-nonbridge positions catalyzed by myosin is a precedent for such a reaction (27).
FTIR technique is advantageous over the 31 P NMR or mass spectrometry methods for PIX studies in the following ways. (i) TR-FTIR offers time-resolved capability, by contrast to 13 P NMR or mass spectrometry, which usually require that molecules involved in PIX reactions be purified or even derivatized for analysis. (ii) Only a modest amount of sample is needed. The sample quantity used in one FTIR experiment (typically a few l of 10 mM GTP-Ras complex) is comparable with the quantity required by mass spectrometry but 100-fold lower than that required by 31 P NMR. Of course, the infrared method is readily expandable to other types of enzymatic systems featuring different functional groups. A limitation of the method is the rather high enzyme concentration needed (several mM) in order to achieve useful signal-to-noise levels, which may prevent application to some systems.
Conclusions-IR spectra revealed stronger binding of the GDP ␤-PO 3 2Ϫ moiety by Ras mutants with higher activity, suggesting that the transition state is largely GDP-like. Analysis of the photolysis and hydrolysis FTIR spectra of the [␤-nonbridge-18 O 2 , ␣␤-bridge- 18 O]GTP isotopomer allowed us to probe for positional isotope exchange. No positional isotope exchange was observed during GTP hydrolysis by Ras. Overall, our results support a concerted mechanism, but the transition state seems to have a considerable amount of dissociative character.