NMR Investigation of Tyr105 Mutants in TEM-1 β-Lactamase

The existence of coupled residue motions on various time scales in enzymes is now well accepted, and their detailed characterization has become an essential element in understanding the role of dynamics in catalysis. To this day, a handful of enzyme systems has been shown to rely on essential residue motions for catalysis, but the generality of such phenomena remains to be elucidated. Using NMR spectroscopy, we investigated the electronic and dynamic effects of several mutations at position 105 in TEM-1 β-lactamase, an enzyme responsible for antibiotic resistance. Even in absence of substrate, our results show that the number and magnitude of short and long range effects on 1H-15N chemical shifts are correlated with the catalytic efficiencies of the various Y105X mutants investigated. In addition, 15N relaxation experiments on mutant Y105D show that several active-site residues of TEM-1 display significantly altered motions on both picosecond-nanosecond and microsecond-millisecond time scales despite many being far away from the site of mutation. The altered motions among various active-site residues in mutant Y105D may account for the observed decrease in catalytic efficiency, therefore suggesting that short and long range residue motions could play an important catalytic role in TEM-1 β-lactamase. These results support previous observations suggesting that internal motions play a role in promoting protein function.

regarded as relatively static entities, increasing evidence now suggests that enzymes behave as dynamic machines and that motions on various time scales play important roles in promoting enzyme catalysis (4). To this day, only a small number of enzymes have been shown to rely on essential proximal and/or distal coupled residue motions for catalysis, among which dihydrofolate reductase (5)(6)(7)(8), cyclophilin A (9, 10), liver alcohol dehydrogenase (11)(12)(13)(14), triose-phosphate isomerase (15)(16)(17)(18)(19), and ribonuclease A (20 -22) remain some of the best characterized systems (for recent reviews see Refs. 23 and 24). Analogous behavior among other enzymes remains to be elucidated, although the confirmation of such phenomena in structurally and functionally unrelated protein families and folds suggests that this may be a widespread process (24). A general view of such correlation between structure, function, and dynamics in enzymes would greatly improve our current understanding of these powerful catalysts. In this study, we provide experimental evidence supporting the importance of active-site residue motions in the enzyme TEM-1 ␤-lactamase through the characterization of various Y105X mutants by NMR spectroscopy.
We have previously investigated the role of the active-site residue Tyr 105 in TEM-1 ␤-lactamase using saturation mutagenesis, enzyme kinetics, and in silico molecular dynamics studies (25). Our results show that this residue is mainly involved in substrate discrimination and stabilization at the active site of TEM-1. Aromatic residues at position 105 were shown to play an important role in substrate stabilization by preventing steric hindrance with substrate molecules through the formation of a rigid, stabilizing wall that restricts the activesite cavity size, and therefore substrate movement. Most nonaromatic residue replacements at position 105 were found to possess too many degrees of freedom for appropriate substrate stabilization, thus explaining the strong aromatic bias observed in other class A ␤-lactamases at this active-site position. Interestingly, Y105G, Y105N, and Y105A were moderately active in hydrolysis of penicillin substrates, despite the fact that they are not aromatic. Increasing side chain length (and flexibility) generally resulted in important decreases in activity. Charged side chains (except the aromatic His) were also poorly compatible with reactivity. These kinetic observations correlated with the extent of Y105X side chain motion upon molecular modeling in the presence of a penicillin substrate; only the aromatic and the small residues provided a stable, well organized binding environment over the short time scale tested (picosecond). To further explore dynamics of these TEM-1 ␤-lactamase active-site mutants, we turned to nuclear magnetic resonance.
NMR relaxation experiments are powerful techniques that can provide valuable information on the dynamic effects of mutations in the active-site cavity of enzymes during catalysis (10). In addition to providing information on fast dynamics (picosecond-nanosecond (ps-ns)) of protein backbones (26), the time scale of NMR dynamics ranges up to the catalytically relevant microsecond-millisecond (s-ms) (27). The recent and complete 1 H, 15 N, 13 C backbone resonance assignments of TEM-1 (E28G) ␤-lactamase by NMR (28) as well as 15 N relaxation backbone dynamics studies on the same enzyme (29) now pave the way to the motional characterization of important active-site residues with respect to their effect on catalysis in this enzyme family. To improve the interpretation of our previous molecular modeling and kinetics observations at the molecular level and to verify the proposed motion of residue 105 in TEM-1, this study describes the backbone resonance assignments of TEM-1 mutants Y105D, Y105G, Y105N, and Y105W as well as 15 N relaxation and backbone dynamics of wild-type TEM-1 and mutant Y105D. To our surprise, the localized dynamic effects we had originally observed by molecular modeling in the area of position 105 extend to a far broader environment, affecting catalytically relevant motional time scales of important catalytic residues in the active-site cavity of TEM-1. The dynamic investigation of the Y105D mutant and the effects of the Y105X mutations on the surrounding environment provide evidence for the importance of active-site residue motions in TEM-1 ␤-lactamase as well as their possible role in substrate stabilization and catalysis. In addition to long range effects observed for residues distal to the active site as well as evidence offered by previous molecular dynamics investigation performed on a TEM-1 mutant (30), these experimental observations suggest that class A ␤-lactamases may rely on long range residue motions in substrate recognition as well as for catalysis.

EXPERIMENTAL PROCEDURES
Reagents-Unless otherwise indicated, all chemicals were purchased from Sigma. Bistris propane 4 was purchased from GE Healthcare, and nitrocefin was purchased from Oxoid (Nepean, Ontario, Canada). Restriction and DNA-modifying enzymes were purchased from MBI Fermentas and New England Biolabs. 15 (31), in which the wild-type bla TEM-1 gene was fused to the leader sequence of ompA was a generous gift from Marvin D. Makinen (University of Chicago). It was maintained using 30 g/ml kanamycin and was used for extracellular protein expression under the control of the T7 promoter in E. coli BL21(DE3). The construction of plasmids pQE32Chl-TEM(Y105X) containing the Y105X mutations of TEM-1 was described elsewhere (25).
After overnight dialysis at 4°C against a 10 mM Bistris propane buffer (pH 6.6), purification of WT and TEM(Y105X) mutants was performed on an ÄKTAexplorer chromatography system from GE Healthcare as reported previously (29). In all cases, purity was estimated to be higher than 95% by SDS-PAGE, and liquid chromatography/mass spectrometry/electrospray ionization and yields were typically ϳ50 mg/liter of pure protein for all TEM-1(Y105X) mutants.
NMR Samples-For the acquisition of NMR spectra, WT and mutants Y105D, Y105G, Y105N, and Y105W were lyophilized after extensive dialysis against H 2 O. The enzymes were subsequently dissolved to a concentration of 0.8 mM in a 90% H 2 O, 10% 2 H 2 O solution containing 4 mM imidazole and 0.1 mM 2,2dimethylsilapentane-5-sulfonic acid for internal pH and chemical shift referencing, respectively. All experiments were performed at pH 6.6.
NMR Spectroscopy-All NMR experiments were performed at 30°C on a Varian INOVA 600 spectrometer operating at a proton frequency of 599.739 MHz equipped with a z axis gradient and a triple resonance cryoprobe. Two-dimensional 1 H-15 N HSQC, three-dimensional HNCO, three-dimensional HN(CO)CA, and three-dimensional CBCA(CO)NH spectra (Biopack, Varian Inc., Palo Alto, CA) together with assignments obtained for TEM-1 E28G (28) were used to determine sequencespecific assignments for the polypeptide backbone of WT and mutant Y105W. Other mutants (Y105D, Y105G, and Y105N) were assigned by comparison using a combination of two-dimensional 1 H-15 N HSQC and three-dimensional HNCO spectra. 15 N relaxation experiments were performed on WT and on the Y105D mutant using 15 N-13 C double-labeled samples for all experiments. 15 N-T 1 experiments were performed using sensitivity-enhanced inversion-recovery pulse sequence with pulsed field gradients developed by Kay and co-workers (32). 15 N-T 2 experiments were performed using the BioPack pulse sequence from Varian, Inc. (Palo Alto, CA) (33). An RF field strength of 6.579 kHz was used for the 15 15 N steady state NOEs were obtained by acquiring spectra with and without 1 H saturation applied before the start of the experiments using a pulse sequence obtained from Kay and co-workers (32). A saturation time of 4 s was used for { 1 H}-15 N NOE experiments. To eliminate the potential effect of sample or field homogeneity degradation over time on measured exponential decays, relaxation delays were acquired in an interleaved manner. For example, the acquisition order for the 15 N-T 2 relaxation experiments was 10, 50, 90, 130, 30, 70, 110, 150, and 190 ms (34, 35).
Data Analysis-All NMR data were processed with NMRPipe/NMRDraw (36) and analyzed with NMRView (37). For relaxation experiments, 1 H-15 N spectra were processed using either a 90°or a 60°shifted sine apodization function in F2 ( 1 H) and either a 90°or a 60°shifted sine-squared function in F1 ( 15 N). The 90°processing was used for the great majority of residues, and the 60°processing allowed the separation of a few peaks exhibiting slight overlapping. Linear prediction was performed in F1 to extend the time domain by a factor of 1.5, and both dimensions were zero-filled to the next power of 2. For each 15 N-T 1 and 15 N-T 2 experiment, the spectrum with the shortest relaxation delay (highest intensities) was peak-picked with NMR-View, and each ellipse was manually adjusted to fit the peak. The same procedure was used for { 1 H}-15 N NOE spectra.
The 15 N R 1 and R 2 relaxation rates were determined by fitting T 1 and T 2 curves to a two-parameter exponential decay of the form shown in Equation 1, where V(t) is the volume after a delay time t; V 0 is the volume at time t ϭ 0; and R is either R 1 ϭ 1/T 1 or R 2 ϭ 1/T 2 . Fitting was accomplished using the program CURVEFIT (AG Palmer, Columbia University). R 1 and R 2 uncertainties were calculated using Jackknife simulations (63), and for each data set the minimum error used for further calculation was set to the mean error. { 1 H}-15 N NOE values were obtained from the ratio of the volumes of experiments recorded with and without proton saturation. The uncertainties on the { 1 H}-15 N NOE values were obtained using the method described by Nicholson et al. (38).
Model-free-The internal motion parameters were optimized for the relaxation data according to the model-free formalism pioneered by Lipari and Szabo (39,40) and extended by Clore et al. (41,42) using the program ModelFree 4.16 (AG Palmer, Columbia University) and the statistical approach of Mandel et al. (43). An axially symmetric diffusion model was used in our analysis. Initial estimates of the global tumbling parameters were obtained using the program QUADRIC (AG Palmer, Columbia University). Residues with { 1 H}-15 N NOE Ͻ0.65 were not considered, and neither were residues with high R 2 (R 2 Ն ͗R 2 ͘ ϩ R2 ), unless their corresponding R 1 values were low (R 1 Յ ͗R 1 ͘ Ϫ R1 ) (44). The value used for 15 N chemical shift anisotropy was Ϫ172 ppm and the N-H bond length was set to 1.02 Å. For each simulation, 500 randomly distributed data sets were generated, and discrimination between models was performed using F-statistics analysis.
The five models used to describe the spin-relaxation data were as follows: model 1, S 2 ; model 2; S 2 , e ; model 3, S 2 , R ex ; model 4, S 2 , e , R ex ; and model 5, S 2 , e , S 2 f , where S 2 is the order parameter used to characterize the amplitude of the internal motions on the ps-ns time scale. S 2 is a measure of the degree of spatial restriction of the 1 H-15 N bond vector and has values ranging from 0, indicating unrestricted motions, to 1, for completely restricted motions; S 2 f is the order parameter for fast motions; e is the effective correlation time for internal motions; and R ex is an exchange term to account for contributions to R 2 from s-ms time scale motions.  (47). Minimized PDB files were created using the InsightII package, version 2000.1 (Accelrys, San Diego, CA) according to the following protocol. For position A, the 1.8-Å crystallographic structure of the E. coli TEM-1 ␤-lactamase (PDB coordinates 1BTL) was used as starting coordinates. The crystallographic water molecules and the active-site SO 4 molecule were deleted, and hydrogen atoms were added at the normal ionization state of the amino acids at pH 7.0. The structure was energy-minimized by applying 1000 steps of steepest descent followed by a conjugate gradient minimization until convergence of 0.001 kcal mol Ϫ1 Å Ϫ1 . For position B, backbone atoms of 1BTL and 1BT5 were superimposed, and the Tyr 105 side chain of 1BTL was repositioned according to Tyr 105 in 1BT5 by applying 1 and 2 angle torsions. The structure was then minimized using the same protocol as for position A. The contribution of Tyr 105 to ring current shifts was estimated with all aromatic residues mutated to Ala, except for Tyr 105 . SHIFTS was also used to predict the effect of the mutation on chemical shifts of other residues of the protein. Observed chemical shift changes were considered meaningful when they were significantly greater than the predicted ones.

Estimation of the Tyr 105 Ring Current Effects and Prediction of Chemical Shift Changes Induced by the Mutation-Ring
Sequence Numbering-Because of technical requirements in NMR analysis, sequence numbering used for TEM-1 and mutants Y105X is different from the classical nomenclature proposed by Ambler et al. (48). Although sequence numbering starts at 26 to give the active-site serine residue the number 70, residues 239 and 253 are not skipped, and the numbering is sequential from 26 to 288. To avoid any confusion in catalytic and structural interpretation, both nomenclatures are appropriately labeled.

RESULTS
15 N-13 C-Labeled proteins corresponding to mutants Y105D, Y105G, Y105N, and Y105W of TEM-1 ␤-lactamase were expressed and purified to homogeneity. These mutants were chosen based on the following structural and functional considerations: Trp because of its aromatic similarity with the native Tyr, the native-like activity of mutant Y105W toward penicillins and cephalosporins, and also because of its frequent occurrence in other class A ␤-lactamases; Asn because its side chain is much smaller than the native, aromatic Tyr although still being a highly active mutant also occasionally represented in other class A ␤-lactamases; Gly because, like Ala, it exhibits discrimination with respect to penicillins (high catalytic efficiency) and cephalosporins (low catalytic efficiency); and Asp as a representative of a low activity mutant despite its structural similarity with the highly active Asn mutant (25).
Chemical Shift Differences-Although our previous NMR studies of TEM-1 were carried on the E28G mutant (28,29), this study was performed without this mutation, and assignments of WT TEM-1 were used (Biological Magnetic Resonance Data Bank accession number 6357). A combination of two-dimensional 1 H-15 N HSQC, three-dimensional HNCO, three-dimensional HN(CO)CA, and three-dimensional CBCA(CO)NH spectra (Biopack, Varian Inc., Palo Alto, CA) were used to sequentially assign nearly all the backbone 1 H, 15 N, and 13 CЈ atoms (BMRB numbers for Y105W, Y105G, Y105N, and Y105D are 7236, 7237, 7238, and 7239, respectively). More than 99% of backbone 1 H, 15 N, and 13 CЈ assignments were obtained for non-proline residues of each enzyme tested. For WT and mutants Y105D, Y105G, and Y105N, the missing assignments are 1 H/ 15 N-Ser 70 and 1 H/ 15 N-Ala 237 , whereas for mutant Y105W, the missing assignments are 1 H/ 15 N-Ser 70 , 1 H/ 15 N-Asn 132 , and 1 H/ 15 N-Ala 237 . As observed previously for the E28G assignments (28), the missing chemical shifts are attributed to peaks overlapping or missing resonances caused by line broadening. Regardless of the functional differences resulting from the Y105X mutations, the twodimensional and three-dimensional NMR spectra of all mutants are quite similar to those of the WT enzyme (supplemental Fig. S1). However, depending on the mutation at position 105, important chemical shift differences were observed at specific residues relative to WT. Fig. 1 presents the structural mapping of backbone amide chemical shift differences (⌬␦ HN ) observed between WT and mutants Y105X, whereas Fig. 2 presents the magnitude of these ⌬␦ HN displayed on the primary sequence of the enzyme. For all mutants, the most important effects observed on ⌬␦ HN occur in three major areas of the enzyme corresponding to residues 100 -115, 120 -140, and 213-218 (Fig. 2). To a lesser extent, all mutants also display higher-than-background effects on ⌬␦ HN in regions encompassing residues 68 -80, 163-170, and 235-246. Finally, with the exception of residue 215, ⌬␦ HN is generally smaller for mutant Y105W than for the three other Y105X mutants investigated.
Residues in contact with substrate molecules or directly implicated in catalysis in TEM-1 ␤-lactamase are all located within active-site walls encompassing residues Met 69 -Lys 73 , . Because these residues are generally located in the immediate vicinity of the substrate molecule and, in some cases, are very close to the mutated residue, some ⌬␦ HN may be the result of a direct short range interaction with residues in close proximity to the Tyr 105 mutation. For instance, the shortest distance between Tyr 105 and Asn 132 (O 105 -N 132 ) is only 3.0 Å in the crystal structure of the free enzyme (PDB coordinates 1BTL), therefore providing an explanation for the magnitude of ⌬␦ HN observed at position Asn 132 in all Y105X mutants. In contrast, although they form the sub-  strate cavity in TEM-1, some of these active-site walls are constituted by residues separated by large distances (e.g. the shortest distance between Tyr 105 and Glu 166 , O 105 -O⑀1 166 ϭ 7.6 Å). Interestingly, the active-site walls are generally more affected by the Y105X mutation than any other portion in the enzyme, and most segments of each wall contain at least one residue displaying higher-than-average ⌬␦ HN (Fig. 2). Moreover, whereas being generally concentrated near the active-site cavity, significant effects on ⌬␦ HN are nevertheless observed throughout the enzyme for all mutants, sometimes more than 20 Å from the site of mutation (Fig. 1). This result illustrates that the effect of mutations at position 105 is not restricted to a local environment. In fact, although 57 residues of mutant Y105W display backbone ⌬␦ HN greater than 11 Hz when compared with WT (22% of the total enzyme), this number jumps to 78 residues (30%) in Y105G, 91 residues (35%) in Y105N, and 104 residues (40%) in Y105D (Table 1), thus clearly extending beyond the immediate environment of position 105. In fact, the Y105X mutation is too far away from several residues displaying significant ⌬␦ HN to result in any direct contribution to chemical shift changes (e.g. backbone 15 N 105 -15 N x distances ϭ 12.7 Å for Gly 238 , 12.9 Å for Leu 169 , and 19.0 Å for Leu 76 ), therefore suggesting the existence of coupled long range effects caused by the mutation at position 105.
The -system of aromatic residues such as tyrosine can generate a local magnetic field known as ring current, which may significantly affect the chemical shift of surrounding nuclei. To verify whether short and long range chemical shift differences observed in the Y105X mutants could simply be attributed to the disappearance of the aromatic hydroxyphenyl side chain of Tyr 105 , we used the program SHIFTS (45) to predict ring current effects originating from Tyr 105 . Ring current shifts were calculated for both previously observed conformations of the Tyr 105 side chain in crystal structures of TEM-1 (46,47). Significant ring current shifts for Tyr 105 were predicted in the immediate vicinity of position 105 (90 Hz predicted for backbone 1 H N of residue 106, 72 Hz for 108, and less than 36 Hz for residues 109, 110, and 130 -132), but no other significant effect (Ͼ15 Hz) was predicted for either of the two side chain conformations of residue 105 (results not shown). This indicates that significant long range chemical shift differences observed in all Y105X mutants are not attributed to direct electronic perturbation caused by the elimination of the hydroxyphenyl side chain of Tyr 105 , and therefore must rely on a combination of concerted effects that have consequences throughout the enzyme.
It is also interesting to note that the number and magnitude of the effects on ⌬␦ HN generally correlate with the previously reported catalytic efficiencies of the same mutants for the classical substrate benzylpenicillin (Table 1) (25), suggesting that electronic perturbations and/or dynamic effects observed by NMR may adequately reflect catalytic effects caused by this mutation in TEM-1. Thus, mutant Y105W displays fewer affected residues (22%), consistent with its high catalytic efficiency, whereas 40% of the residues of the weakly active Y105D mutant show significant 1 H-15 N chemical shift perturbation. This mutant also displays chemical shift perturbation for a greater number of catalytic residues than the other Y105X mutants, consistent with its low catalytic efficiency (Y105Dk cat ϭ 255 s Ϫ1 , Y105D-K m ϭ 369 M, Y105D-k cat /K m ϭ 6.9 ϫ 10 5 M Ϫ1 s Ϫ1 versus WT-k cat ϭ 1240 s Ϫ1 , WT-K m ϭ 43 M, WT-k cat /K m ϭ 2.9 ϫ 10 7 M Ϫ1 s Ϫ1 ) ( Table 2) (25). In addition, these ⌬␦ HN are generally of greater magnitude as catalytic efficiency decreases. 15 N Backbone Relaxation Dynamics-Previous experimental observations made by x-ray crystallography on TEM-1 have shown that Tyr 105 can adopt two alternate conformations in the presence of substrates or inhibitors (47). Although the Tyr 105 hydroxyphenyl side chain points toward Val 216 in the free enzyme, a 1 angle rotation of more than 110°has been   (46,49) and confirmed by a lower-than-average order parameter of this amide in solution as evaluated by NMR (29). However, our previous dynamic modeling studies of this residue showed a low propensity of Tyr 105 for conformational change on the ps time scale relative to more flexible Y105X replacements, suggesting that positioning and restricted dynamic motions of the Tyr 105 side chain could be a determinant of recognition for substrate stabilization in TEM-1 ␤-lactamase (25). TEM-1 Y105D displayed the most important effects on ⌬␦ HN relative to WT as well as a reduction of 2 orders of magnitude in its catalytic efficiency for benzylpenicillin (25). We therefore conducted 15 N backbone relaxation dynamics studies to evaluate the importance of local and global dynamic effects caused by the Y105D mutation. To obtain dynamic information on time scales of ps-ns and s-ms, both longitudinal (R 1 ) and transverse (R 2 ) 15 N relaxation rates as well as { 1 H}-15 N NOE values were measured and are presented in Fig. 4 for both WT and mutant Y105D (raw data presented in supplemental material). We observed a significant variation of relaxation data and global tumbling times as a function of protein concentration. For example, the global correlation time for TEM-1 varied from 12.8 ns at 0.4 mM to 13.8 ns at 0.8 mM. This increase in correlation time is most likely due to an increase in viscosity at high protein concentration. It was therefore crucial for our analysis that both proteins be at exactly the same concentration. From the 1 H-15 N HSQC spectra recorded for the relaxation experiments, it was possible to obtain reliable data for 206 and 230 out of 250 potentially observable amides for WT and mutant Y105D, respectively. Table 3 presents average values for relaxation data and parameters obtained for TEM-1 and mutant Y105D.
The general comparison of R 1 , R 2 , and { 1 H}-15 N NOE values for TEM-1 and mutant Y105D shows that both enzymes behave in a very similar manner, displaying comparable values throughout the sequence (Table 3 and Fig. 4, A-C). The general  constant pattern observed in the relaxation data from one extremity of the enzyme to the other is uncommon relative to the pattern generally observed in other proteins, where an important decrease in both N-and C-terminal regions as well as in unstructured regions is frequently observed. This feature reflects the high rigidity of both enzymes in solution, a property that we have also previously observed in TEM-1 (E28G) (29). However, there are significant local differences between the relaxation data for both enzymes, especially concentrated in regions showing important chemical shift differences (e.g. residues 70 -80, 124 -135, and most importantly 211-221) (Fig. 4, A-C). Fig. 5 shows the Y105D/WT ratios for all relaxation parameters, highlighting the residues displaying the most important differences between both enzymes. Changes in R 1 and { 1 H}- 15 N NOE values reflect differences in the ps-ns dynamics of proteins, whereas changes in R 2 values may also reflect changes in s-ms motions. Interestingly, 88 residues are significantly affected in either R 1 or { 1 H}-15 N NOE relaxation parameters, suggesting significant dynamic differences on the ps-ns time scale for these residues (supplemental Table S3). Among these, 14 belong to the active-site walls of TEM-1 and may be implicated in substrate stabilization (Table 4). Because it has been proposed that motions on the ps-ns time scale may influence the thermodynamics of binding as well as the kinetics of enzyme-catalyzed reactions (50 -52), disruption of ps-ns motions among these active-site residues may reduce substrate stabilization and/or catalysis in mutant Y105D. Similarly, among the 30 residues significantly affected in R 2 (supplemental Table S3), 11 belong to the active-site walls (Table 4), suggesting that differences in s-ms motions of these residues between WT and mutant Y105D may also affect substrate stabilization and/or catalysis. Residues 211-221 correspond to the region where the R 2 values are the most affected by the Y105D mutations, both in terms of magnitude and number of residues

TABLE 4 Active-site wall and invariant residues displaying significant relaxation parameter variation between wild-type TEM-1 and mutant Y105D
Variations in R 1 , R 2 , and NOE were considered significant if the Y105D/WT ratio was larger than the average ratio Ϯ1. Variations in S 2 were considered significant if the difference was larger than the sum of the errors ϩ 0.01. Variations in R ex were considered significant if the difference was larger than the sum of the errors ϩ 1.0 s Ϫ1 . a Residues defining active-site walls are in boldface type and residues intolerant to any amino acid substitution in TEM-1 (55) are underlined. Sequential numbering is used. b As the 15 N-HSQC correlation of Lys 215 was significantly weaker for the WT than for Y105D, it was impossible to obtain relaxation data for the WT. We therefore conclude that the R 2 value was significantly higher for the WT, reflecting significantly higher R ex .

Relaxation parameter Residues
affected. In addition, because s-ms dynamics are directly related to the time scale of catalysis, these modified motions may also affect turnover in mutant Y105D. Residue Lys 234 is also a particularly good candidate for this, as it has been shown to be an essential member of the catalytically important hydrogen-bonding sub-network of class A ␤-lactamases through the formation of a hydrogen bond with Ser 130 (an equally important member of the SDN loop implicated in catalysis) (53).
Model-free Analysis-To correlate our relaxation data with the internal dynamics of the protein, further dynamic analyses were conducted using the model-free formalism pioneered by Lipari and Szabo (39,40). Such analyses allow for the direct investigation of local and global dynamic effects observed in the protein of interest, namely through the extraction of the order parameter (S 2 ), the conformational exchange parameter (R ex ), and the overall correlation time of the molecule ( m ) ( Tables 3  and 4 and Fig. 4, D and E).
Average values of model-free parameters for WT and mutant Y105D confirm that both enzymes behave very similarly with respect to their global dynamic properties (Table 3). Following the model selection, there was only a slight divergence between both proteins; for WT and mutant Y105D, respectively, 82 and 81% of the residues fitted well for model 1, 7 and 9% for model 2, and 10 and 8% for model 3. No residue was fitted to model 4 nor model 5 in either protein. Both WT and Y105D display a small prolate axial anisotropy with D ʈ /D Ќ values of 1.16 and 1.18 and similar global correlation times ( m ) of 13.8 and 13.7 ns, respectively ( Table 3). Considering that the two enzymes differ only by a single mutation, this result was expected. In addition, the average order parameter values (S 2 ) obtained for TEM-1 and mutant Y105D are exceptionally high (Ͼ0.9 for both enzymes), confirming previous observations reported for TEM-1 (E28G) (29). Because the order parameter measures the amplitude of ps-ns motions and varies from 0 for unrestricted internal motions to 1 for completely restricted motions (43), values Ͼ0.9 are indicative of highly ordered proteins in solution.
Six residues display distinct behavior patterns in their order parameters (S 2 ) (supplemental Table S3), two of which are located in the active-site walls (Ser 106 and Glu 239 ) ( Table 4). These residues display a significant decrease in their order parameters indicating an increase in ps-ns motion amplitude in the mutant compared with WT. Furthermore, five residues display distinct s-ms motions when compared with WT (supplemental Table S3), four of which belong to active-site walls (Lys 215 , Val 216 , Gly 218 , and Lys 234 ) ( Table 4). The model-free analysis for these residues required an R ex term, which is related to local conformational exchange and refers to motions observed on the s-ms time scale (43). As the 15 N-HSQC correlation of residue Lys 215 was significantly weaker for the WT than for Y105D, it was impossible to obtain relaxation data for this residue in the WT. We therefore conclude that the R 2 value was significantly higher for the WT, reflecting significantly higher R ex for Lys 215 in the WT. The important decrease of R 2 and R ex for residues 215, 216, 218, and 234 in mutant Y105D (Fig. 4E) suggests a slowdown in motions of these residues on the catalytically relevant s-ms time scale. These residues delineate two active-site walls of TEM-1, and Lys 215 , Val 216 , and Gly 218 are located in the most dynamically affected region observed between WT and mutant Y105D (Fig. 4). It is interesting that for this region (211-221), only two residues exhibit small but significant chemical shift differences (Lys 215 and Val 216 ), although changes in motional parameters occur for a greater number of residues and are more important, as reflected in the fact that this region displays the most significant changes in R 2 and R ex values. Assuming exchange in a two-state model, this could suggest that for these residues the rate of exchange between the two states is significantly affected by the mutation, but the population of each state is roughly the same in the WT and in mutant Y105D. Therefore, only the rate of exchange would be significantly affected by the mutation.
Despite the fact that the model-free approach is not the most comprehensive evaluation of s-ms time scale motions, dynamics on this time scale can be inferred from R 2 . Crude R 2 values do not necessarily provide an adequate portrayal of s-ms dynamics for an enzyme, but differences in R 2 for WT and mutant Y105D suggest s-ms motion differences resulting from this mutation. In addition to the important R ex differences noted for positions 215, 216, 218, and 234, 13 additional activesite wall or invariant residues may have different s-ms motions between WT and Y105D, based on significant variation in R 2 ( Table 4). Motions of Val 216 and Lys 234 on a catalytically relevant time scale could affect catalysis by perturbing the hydrogen bonding network observed in WT (Fig. 6), therefore partly explaining the differences in k cat observed previously with mutant Y105D (25).

DISCUSSION
The investigation of the correlations between enzyme dynamics and function is required to gain a detailed understanding of the mechanisms underlying the catalytic activity of these important molecules. Implication of ps-ns and s-ms motions in enzyme activities and catalytic rates is now well accepted, and NMR spectroscopy is a valuable tool to probe these time windows (21). In an attempt to explain the differences in the catalytic efficiency of several Y105X mutants of TEM-1, we previously conducted a short 200-ps molecular dynamics simulation that allowed for the partial explanation of differences in affinity through the formation of a stabilizing wall created by residues exhibiting few degrees of freedom at position 105, therefore restricting substrate motion in the active site (25). However, this molecular dynamics simulation model was relatively limited in that it only allowed for the investigation of small motions explored on a short time scale and exclusively concentrated in the local environment of position 105. To better characterize these motions, we investigated the role of the conserved active-site residue Tyr 105 by comparing its structural and dynamic features with respect to the Y105W, Y105G, Y105N, and Y105D mutants using NMR spectroscopy. The chemical shift differences observed between TEM-1 and these various mutants allowed us to focus our dynamic characterization on the Y105D mutant.
Overall, although the Y105D mutation considerably affects the electronic and dynamic environment of several residues throughout the enzyme, the backbone dynamics of residue 105 are not significantly affected relative to WT, suggesting that local dynamics at position 105 are not the sole element contributing to the differences observed in catalysis. Indeed, despite the important steric and ionic alterations offered by the Y105D replacement, R 1 , R 2 , and { 1 H}-15 N NOE values of mutant Y105D remain similar to WT. On the other hand, our present analyses revealed significant short and long range changes in motion throughout the enzyme, providing further clarification of the effect of this mutation in substrate stabilization and catalysis. We show significant alterations in the ps-ns and s-ms dynamics of important residues, often either in or near the active-site walls of TEM-1, as a result of this mutation at position 105. For instance, Ser 106 and Glu 239 show significant increase in ps-ns motions with respect to WT. Such a decrease of the order parameter for residues located in the active site could result in a higher conformational entropy cost associated with substrate binding, hence contributing to the catalytically impaired active-site observed in the Y105D mutant. In addition, Arg 164 is a highly conserved residue in class A ␤-lactamases that is considered to be important in anchoring the base of the ⍀-loop through a salt bridge with Asp 179 (54). Although more than 17 Å away from the mutation at position 105, a significant change in the R 1 is observed. This change in ps-ns dynamics through a possible network of active-site motions could affect the stability of the ⍀-loop and thus reduce appropriate substrate stabilization and catalysis.
Moreover, our results show a significant decrease in the s-ms motions for residues Glu 212 , Lys 215 , Val 216 , Gly 218 , and Lys 234 . It is interesting to note that among these residues, Val 216 and Lys 234 were shown to be intolerant to any amino acid substitutions with respect to benzylpenicillin hydrolysis in TEM-1 (55). These results suggest that this activity requirement may partly be governed by s-ms dynamics in the vicinity of these residues. Lys 215 , Val 216 , and Lys 234 are positioned in the immediate vicinity of the substrate molecule, and their ground-state dynamic behavior may have a direct impact on substrate recognition and stabilization, therefore partly explaining the decreased affinity observed for mutant Y105D (25). In fact, although Lys 234 is implicated in the initial recognition of the substrate molecule as an electrostatic anchor for the carboxylate moiety of substrates (56), its implication in proton shuttling during catalysis is still the subject of debate (57). On the other hand, with help from the guanidinium group of Arg 244 , the backbone carbonyl group of Val 216 has been shown to anchor a conserved water molecule that interacts with the C3 carboxylate group of the substrate for appropriate stabilization (58) (Fig. 6). Thus, this residue has been suggested to influence substrate binding and catalysis in both TEM-1 (55) and PSE-4 (59). Because the s-ms motions surrounding Glu 212 , Lys 215 , Val 216 , and Lys 234 are decreased in mutant Y105D, these changes on the catalytically relevant time scale of their local environment may reduce substrate stabilization in the active site of Y105D, consistent with the large increase in K m previously observed with this mutant (25). In addition to affecting the correct stabilization of the substrate through a conserved water molecule by Val 216 (Fig. 6), the attenuation of s-ms motions in the Y105D mutant could reflect a reduction of the possible conformations that the enzyme can adopt on this time scale. These "productive" conformations could be responsible for the appropriate positioning of the substrate in the active site, and their loss upon mutation could account for the decline of the enzymatic activity. It should be noted that the present study confirmed the presence as well as the importance of previously observed s-ms motions in the active site and in the ⍀-loop of TEM-1 (29). Although it was observed that s-ms motions were present in the vicinity of the active site (29), it was impossible to determine whether these motions were catalytically relevant. The changes in s-ms time scale motions we observe upon a mutation that affects kinetic properties of TEM-1 indicates a correlation between motion time scales and the kinetic properties, consistent with causality, although causality has not been demonstrated. To date, the numerous x-ray studies of TEM-1 have provided no indication of catalytically related motions. As a result of the current NMR studies, we propose that the combination of these subtle but significant effects within the activesite cavity are directly related to the 2 orders of magnitude reduction in catalytic efficiency observed for the Y105D mutant of TEM-1 (25). It is important to note that the available data do not allow us to differentiate whether the s-ms motions observed for the amides of Glu 212 , Lys 215 , Val 216 , Gly 218 , and Lys 234 are resulting from their own motions or from motions of other surrounding residues. However, it is now clear that these s-ms motions are correlated with the fine-tuning of catalytic properties of TEM-1, and possibly of other ␤-lactamases.
Long Range Dynamic Effects-Although important residues displaying modified ps-ns and s-ms dynamics in mutant Y105D are elements of the active-site walls and therefore generally considered in close proximity to the substrate molecule, it is important to keep in mind that motions characterized in this study are exclusively focused on the backbone relaxation of 15 N atoms. To that extent, it is important to estimate relevant distances separating residues of interest and therefore to clarify what is considered short range (Ͻ5 Å) or long range (Ͼ5 Å) interactions. For instance, the shortest distance between Tyr 105 and Val 216 (OH 105 -C␥1 216 ϭ 4.2 Å) may be considered short range, whereas the distance between both their 15 N atoms would be considered long range (14.7 Å). This important difference in atom distances is observed for several residues located in the active-site walls of TEM-1. Nonetheless, most of the calculated distances among dynamically affected residues should be considered long range because they do not permit any direct contact with residue 105. For instance, Lys 234 displays a shortest residue distance of 7.8 Å with Tyr 105 (C⑀2 105 -N 234 ) and a 15 N 105 -15 N 234 distance of 18.7 Å (both long range).
This observation raises two important points regarding dynamic results characterized in this study. The first point is the fact that, except for dynamically affected residues located in the immediate vicinity of the mutation (e.g. Ser 106 , Asn 132 , and to a certain extent Val 216 ), both shortest inter-residue or NMR observable (backbone 15 N atoms) distances between position 105 and any other dynamically affected residue are too important to account for any direct interaction (e.g. Lys 234 , Leu 76 , Gly 218 , etc.). This observation raises the second important point: the long range dynamic effects observed as a consequence of the Y105D mutation is consistent with the existence of a network of motions among residues of TEM-1 ␤-lactamase. This hypothesis is strengthened by a previous molecular dynamics study performed on the inhibitor-resistant M69L mutant of TEM-1, suggesting that only differences in dynamics of this mutant account for the resistance to clavulanate (30). This type of dynamic network explaining long range dynamic effects has been characterized previously in detail for other enzymes and has often been shown to play a crucial role in catalysis (reviewed in Refs. 23 and 24). It has also been observed in allosteric (60) and other noncatalytic proteins (61) with long range motions characterized as contiguous (displaying a traceable pathway) or disperse (noncontiguous with an untraceable pathway). The evidence presented here not only suggests dynamic "cross-talk" between residues constituting opposite walls of the active-site cavity of TEM-1 (often separated by large distances) but also suggests that residues distal to the active-site cavity may disrupt catalysis in TEM-1 through their altered motional behavior. For instance, it is interesting to note that several residues displaying significant ps-ns and/or s-ms motional differences have been shown to be intolerant to any amino acid substitution in TEM-1 (Table 4), among which several are conserved in all class A ␤-lactamases (55). Although some residues of the active-site walls are expected to be intolerant to any mutation because of their direct importance in catalysis (e.g. Lys 234 and Val 216 ), others are distal to the activesite cavity and display significant dynamic alterations between TEM-1 and mutant Y105D. For example, two dynamically affected residues (Tyr 46 and Leu 76 , both Ͼ20 Å from the mutation) are completely buried in TEM-1, and their motional disruption through a possible long range network of coupled motions may affect the appropriate packing of the hydrophobic core of the enzyme and/or catalysis.
The fact that our NMR studies on WT and mutants Y105X were performed in the absence of any substrate or inhibitor is of considerable interest because it supports previous investiga-tions showing that catalytically relevant motions are often observed in the free enzyme (10,20). The conservation of motions in absence of substrate may be an essential component of enzyme evolution and may contribute to explain the exceptionally high acceleration rates observed in enzyme catalysis. In addition, as previously pointed out (23), this finding may have interesting implications for the understanding of the secondary and tertiary elements conserved in several protein folds. Assuming that the conserved fold observed for all class A ␤-lactamases is partly governed by an evolutionary constraint preserving elements that define dynamic motions essential to their catalytic function, the characterization of such elements would greatly advance our understanding of class A ␤-lactamase catalysis as well as their importance in antibiotic resistance. A common structural organization of functionally relevant regions that undergo similar concerted movements was recently revealed in the protease enzymatic superfamily (62). Whether the presence of such a network of motions is relevant to the class A ␤-lactamase fold and similar catalytic mechanisms is beyond the scope of this study. Nevertheless, considering the long range dynamic effects characterized here, it is reasonable to assume that dynamic differences between TEM-1 and various mutants are involved in the observed differences in activity.