Triose-phosphate Isomerase (TIM) of the Psychrophilic BacteriumVibrio marinus

The purification and characterization of triose-phosphate isomerase from the psychrophilic bacteriumVibrio marinus (vTIM) is described. Crystal structures of the vTIM-sulfate complex and the vTIM-2-phosphoglycolate complex (at a 2.7-Å resolution) are also presented. The optimal growth temperature of Vibrio marinus is 15 °C. Stability studies show that vTIM is an unstable protein with a half-life of only 10 min at 25 °C. The vTIM sequence is most closely related to the sequence ofEscherichia coli TIM (eTIM) (66% identity), and several unique structural features described for eTIM are also seen in vTIM, but eTIM is considerably more stable. The T d values of vTIM and eTIM, determined by calorimetric studies, are 41 and 54 °C, respectively. Amino acid sequence comparison reveals that vTIM has an alanine in loop 8 (at position 238), whereas all other TIM sequences known to date have a serine. The vTIM mutant A238S was produced and characterized. Compared with wild type, the catalytic efficiency of the A238S mutant is somewhat reduced, and its stability is considerably increased.

physical, enzymological, and computational studies. Its catalytic properties and mechanism have been studied in detail (1,2). It has been established that TIM is a very efficient catalyst, since the reaction rates are diffusion-controlled. Neither cofactors nor metal ions are required in this reaction, and there is no evidence of allostery or cooperativity among the subunits.
Structurally TIMs do form a well characterized family. To date, x-ray structures of TIM from seven different sources have been solved (wild type or/and in complex with substrate analogues): chicken (3), yeast (4), Trypanosoma brucei (5), Escherichia coli (6), human (7), Bacillus stearothermophilus (8), and Plasmodium falciparum (9). In addition, the structures have been determined of Leishmania mexicana TIM 2 and of Thermotoga maritima TIM. 3 Also, 45 amino acid TIM sequences from a wide variety of organisms have been determined and are available in the data bases.
Each TIM monomer has a globular shape with two protruding loops. The globular part is formed by an 8-fold repeat of a ␤-strand, loop, ␣-helix, loop motif. The ␤␣ units fold up in a regular way, such that the ␤-strands (numbered ␤1 to ␤8) form an eight-stranded ␤-barrel surrounded by the eight ␣-helices (numbered ␣1 to ␣8) on the outside. This folding motif is also referred to as the TIM barrel motif. In TIM, the two protruding loops are after ␤-strand three and after ␤-strand six.
The TIM barrel structural motif is found for a large number of other enzymes with widely different functions (10) that have little or no sequence homology. As such this stable framework with tolerant sequence variations has already been used as an interesting scaffold to design loops (11,12) and even artificial proteins (13,14). The active sites of all these TIM barrel enzymes are located at the C-terminal end of the ␤-barrel, with catalytic residues contributed by the ␤-strands and the loops connecting the ␤-strands to subsequent ␣-helices numbered loop 1 to loop 8 in correspondence with the numbering of the preceding ␤-strands. The loops preceding the ␤-strands are located on the other side of the barrel and are believed to be important for the stability of the TIM barrel (15).
In the TIM dimer, the dimer interface is formed mainly by loops 1, 2, 3, and 4. The protruding loop 3 of one subunit docks into a deep pocket between loop 1 and loop 4 of the other subunit (16). The catalytic residues, a lysine, a histidine, and a glutamate are located in loop 1, loop 4, and loop 6, respectively. Binding of substrate or substrate analogues is accompanied by * This work was supported by Services Fédéraux des Affaires Scientifiques, Techniques, et Culturelles Grant PAI P3-044, European Union Grants BIOT CT90-0182 and BIO4-CT96-0670, and Belgian National Science Foundation Grant FRFC 2.4545.96. The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked "advertisement" in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.
The nucleotide sequence (s)  changes in conformation of loops 5, 6, and 7. The most remarkable change occurs in loop 6, "the flexible loop," the tip of which undergoes a C ␣ movement of some 7 Å in response to ligand binding (17,18). As such, the active site is closed off from solvent to prevent an unwanted phosphate elimination reaction that may otherwise occur (1). This state is called "the closed state" in contrast to the unliganded "open state." In the liganded form the oxygen atoms of the phosphate interact with main chain atoms of loops 6, 7, and 8. Loop 8 is called the phosphate binding loop, because it comprises a 3/10 helix (the phosphate binding helix) whose N terminus points directly to the phosphate moiety, allowing for hydrogen bonds between main chain NH groups and phosphate oxygen atoms. These interactions keep the phosphate moiety rigidly in place, whereas in the triose part of the ligand molecule, the catalytic conversion takes place.
The glutamate at the beginning of loop 6 is implicated in the transfer of a proton between the two substrate C atoms. The side chain of this glutamate can occupy two different positions, depending on the contents of the active site. In ligand-free TIM, it is found in a "swung out" conformation, whereas in liganded TIM, it occupies a "swung in" conformation, ideal for carrying out catalysis (19).
In this paper, we focus on the TIM enzyme from the psychrophilic bacterium Vibrio marinus. We will refer to the sequence numbering of V. marinus TIM (vTIM); according to this numbering, the catalytic residues are Lys-11, His-92, and Glu-169 (see Fig. 2). V. marinus is a member of the Vibrionaceae family, able to grow at 4°C and having an optimal growth temperature of about 15°C (20). We cloned the gene encoding TIM. Recombinant vTIM was overproduced in Escherichia coli and purified. The kinetic properties as well as the stability of the enzyme were determined.
Two structures of this enzyme, one complexed with sulfate and one with the substrate analogue 2-phosphoglycolate (2PG) (Fig. 1), were solved by x-ray crystallography at a resolution of 2.7 Å. Because the sequence of vTIM is most closely related to that of E. coli TIM (eTIM) (66% identity) (see Fig. 2), the latter enzyme will be used as a reference for comparison throughout this paper.
Furthermore, sequence comparison of the vTIM sequence with all other known TIM sequences reveals that vTIM has a unique residue in the phosphate binding helix, where the otherwise completely conserved Ser-238 is replaced by an alanine. The implications of this unique sequence feature are discussed by comparing the stability and the kinetic parameters from the native vTIM with its A238S mutant. It turns out that this point mutation decreases the physiological efficiency at 10°C and it increases the thermal stability, as evidenced by calorimetric studies.
Cloning of the vTIM Gene-A fragment of the vTIM gene was synthesized by polymerase chain reaction using primers corresponding to conserved regions of the molecule (22). This fragment was cloned, 32 P-labeled, and used to screen a V. marinus genomic library constructed by partially digesting the chromosomal DNA with XbaI and ligating the fragments to XbaI-cut and dephosphorylated pUC13. A 2,100-bp XbaI fragment of V. marinus DNA cloned in pUC13 gave a strong hybridization signal with the specific 500-bp polymerase chain reaction fragment used as a probe. DNA sequencing revealed that this 2,100-bp fragment contains the entire TIM gene.
Construction of the Expression Vector-The vTIM gene cloned in pUC13 was digested with Sau3AI and PstI. This fragment (about 780 bp) contained the entire TIM gene except for its first 20 codons. These first 20 codons were synthesized chemically with an NdeI restriction site adjacent to the initiation ATG codon and a Sau3AI protruding end at the opposite site. The resulting oligonucleotide and its complementary strand were annealed, phosphorylated, and ligated to the Sau3AI-PstI 780-bp TIM gene fragment as well as to the pT7CTBR plasmid cut with NdeI and PstI. The ligation products were used to transform E. coli HB101 cells. The correct recombinant plasmid, called pT7-Vib, was used to direct expression of vTIM.
Production and Purification of vTIM and Its A238S Mutant-Expression of the TIM gene was performed using the T7 system (23). E. coli BL21 (DE3) cells carrying the pT7-Vib recombinant vector were grown for 16 h at 37°C in L-broth medium containing 100 mg/liter ampicillin. Induction with isopropyl-␤-D-thiogalactopyranoside was unnecessary for the production of vTIM and its mutant. The cells were harvested by centrifugation (15 min, 4,500 ϫ g) and resuspended in 20 mM Tris-HCl (pH 8.0), 1 mM EDTA. After lysis in a high pressure cell disruption system (Inceltech), the cell debris was removed by centrifugation (45 min, 10,000 ϫ g), and the supernatant containing the soluble TIM was recovered. TIM was partially purified by stepwise ammonium sulfate precipitation. Monitoring was performed by SDS-polyacrylamide gel electrophoresis (15% polyacrylamide). TIM precipitated between 60 and 90% saturation. The TIM-containing precipitate was collected by centrifugation (15 min, 20,000 ϫ g, 4°C) and dissolved in 20 mM Tris-HCl (pH 8.0), 1 mM EDTA. The sample was dialyzed against 20 mM triethanolamine-HCl (pH 7.6) overnight at 4°C. The sample was then concentrated by ultrafiltration with Centricon-10 (Amicon Corp.) and applied onto a Mono Q ion-exchange column (HR 10/10, Pharmacia Biotech Inc.) equilibrated in 20 mM triethanolamine-HCl (pH 7.6). TIM was eluted with a gradient of 0 -500 mM NaCl, and the fractions containing TIM were pooled and concentrated with Amicon YM-10 membranes (Amicon). The protein concentration was determined by the method of Bradford (24) using the Bio-Rad protein assay and bovine serum albumin as the standard. An average of 80 mg of TIM was obtained/liter of bacterial culture.
Production and Purification of eTIM-eTIM was produced and purified as described previously (6).
Activity Assays-TIM activity was determined by a coupled enzyme assay based on the decrease of absorbance at 340 nm due to oxidation of NADH, according to Misset and Opperdoes (25). The assay was monitored with a thermostated Uvikon 930 spectrophotometer (Kontron). The reaction mixture (1 ml) contained 0.24 mM NADH (Boehringer Mannheim, Gmbh) and 20 g/ml glycerol-3-phosphate dehydrogenase (Boehringer Mannheim, Gmbh) in 100 mM triethanolamine-HCl (pH 7.6). To initiate the reaction, 2.5 ng of vTIM or 4 ng of its A238S mutant were added to each sample. The enzymatic assay was performed by adding varied concentrations of the substrate D-glyceralde-FIG. 1. The reaction catalyzed by triose-phosphate isomerase, and the covalent structure of the substrate analogue 2PG is shown.
Stability Assays-For the stability assays, samples containing 100 mM triethanolamine-HCl (pH 7.6) and TIM were incubated at various temperatures (5-25°C) in a thermocycler (Programmable TRIO-thermoblock TB-1, Biometra BAU) for 10 -60 min. The remaining activity was measured as a function of incubation time. From these measurements an inactivation rate constant, k inact (sec Ϫ1 ), was calculated according to Rentier-Delrue et al. (22).
Differential Scanning Calorimetry-Thermal unfolding of TIM was studied using a microcal scanning calorimeter differential scanning calorimetric unit (MicroCal, Inc.). Protein samples of 1.3 mg/ml in triethanolamine-HCl buffer (pH 7.6) were used. Samples were heated from 15 to 60°C, and the scan rate was 1°C/min.
Crystallization and Data Collection-Crystals were grown at 4°C using the hanging drop vapor diffusion technique. The protein solutions used in the crystallization setups contained 10 mg/ml protein in 10 mM triethanolamine-HCl, 25 mM NaCl, 1 mM dithiothreitol, 1 mM EDTA, 1 mM NaN 3 . For the vTIM(sulfate) crystallization, 100 mM ammonium sulfate was added, and for the vTIM(2PG) crystallization, 20 mM 2PG was added. The crystallization conditions were found using the "Fast Screen" crystallization protocol (28) and are summarized in Table I.
Crystals of a size up to 0.3 ϫ 0.3 ϫ 0.5 mm 3 grew within a week, and one of each was used for data collection. Data were collected at 25°C on a BigMar image plate detector mounted on an Enraf Nonius rotating anode FR571. Data collection was set up to provide a 2.7-Å data set of the vTIM(2PG) crystal and a 2.65-Å data set of the vTIM(sulfate) crystal. The data were processed with DENZO (29). The crystal lattice for both growth conditions is primitive monoclinic. The ␤-angle for both data sets is about 91°and the a and c cell dimensions are very similar (89.7 and 89.5 Å), agreeing closely with tetragonal symmetry. Indeed at low resolution the data could be merged in space group P4, but for the higher resolution shells this results in an R-merge above 40%. The systematic absences of the 0k0 reflections indicated space group P2 1 . This symmetry and the cell volume (1.11 ϫ 10 6 Å 3 ) suggest that the asymmetric unit contained four TIM dimers, giving a V m of 2.5 Å 3 Da Ϫ1 . Data collection statistics for both data sets are shown in Table I.
Molecular Replacement-The molecular replacement calculations were done with the vTIM(sulfate) data set. To find the orientation of the TIM dimers in the unit cell, molecular replacement was performed with the program AMORE (30) using the crystal structure of the eTIM dimer (1TRE; Ref. 6) as the search model. The flexible loop (residues 169 -180) known to be variable in conformation was deleted in both monomers. The rotation function showed eight unique solutions, as expected for four dimers per asymmetric unit. A translation function was calculated for each of the eight rotation peaks. The best translation function solution was used to find the position of the second dimer. This proce-  (6) is the r.m.s. difference between observed 1 2 value and the nearest preferred cluster values, as observed in a data base of well refined structures. b As defined by PROCHECK (34).
dure was repeated until the positions of all four dimers were found. The R-factor for data between 15 and 3.5 Å for this complete solution was 49.1%, which dropped to 45.7% after rigid body refinement. The correctness of the solution was confirmed by the good packing of the four dimers and the quality of the fit of the omitted loop in the electron density map. Structure Refinement-First the structure of the vTIM(sulfate) complex was refined. The sequence of the molecular replacement model was mutated from that of eTIM to that of vTIM in both monomers. The model was optimized to improve its fit in a 2Fo-Fc SIGMA A weighted map (CCP4; Ref. 31) using O (32) running on an Evans and Sutherland workstation. For refinement, a subset of 5% of the data (the test set) was used for R-free calculations. Refinement was pursued with a mixture of simulated annealing x-ray refinement and conventional positional and thermal factor refinement using the X-PLOR package (33) for all refinement calculations. The positions and the thermal B-factors of the atoms of both monomers in the dimer were refined independently. The atom positions of the three other dimers were generated by strict noncrystallographic symmetry operators, recalculated at several stages of the refinement by rigid body refinement.
Electron density maps indicated an open conformation for loop 6, which was manually rebuilt into the density. In addition, a sulfate ion was found in the active site of the first subunit. Its coordinates were included in the model. In the final stages of the refinement, water molecules were added at sites, displaying a peak larger than 3 S.D. above the mean in an Fo-Fc map and having a potential hydrogen bonding partner. In total, 44 water molecules were added per dimer. The final R-factor was 20.3% and R-free was 22.0% to a resolution of 2.65 Å (Table I).
Subsequently this structure was used for the rigid body refinement against the vTIM(2PG) data set. Electron density maps showed clear density for the closed loop 6 and the 2PG bound in the active site of both subunits. A similar protocol as for the native structure refinement was used, loop 6 was rebuilt, the inhibitor 2PG was built into the active site of both subunits, and 77 waters per dimer were added. It resulted in a final R-factor of 19.2% (R-free is 21.5%) for this vTIM(2PG) structure. The refinement statistics are given in Table I.
Structure and Sequence Analysis-The quality of the structures was analyzed using the programs PROCHECK (34) and WHAT IF (35). For the purpose of analysis, TIM structures were superimposed on the basis of the 129 C ␣ positions forming the eight ␣-helices and eight ␤-strands using the lsq option in O. Cavities were calculated with the program MSP (36), intermolecular contacts were calculated with WHAT IF, and molecular graphs were drawn with the programs MOLSCRIPT (37) (Fig. 6) and ICM 4 (38) (Figs. 7 and 8). The consensus sequence shown in Fig. 2 was obtained from a sequence alignment of 45 sequences currently available in the data bases using PILEUP (39) as well as a superposition analysis of known structures.

Kinetic and Stability Properties of vTIM
In the present work we have cloned the gene encoding TIM of V. marinus. The corresponding protein was expressed in E. coli. The thermostability of this recombinant TIM was assessed by measuring the inactivation rate constant (k inact ) at various temperatures (Fig. 3). At 25°C the value recorded was 1 ϫ 10 Ϫ3 sec Ϫ1 . The calculated half-life (t 1/2 ) was 10 min at this temperature, showing the intrinsic psychrophilic character of this TIM. Because of its psychrophilic properties, we choose to measure its kinetic parameters at 10°C; at this temperature t 1/2 is 58 min. The kinetic properties are summarized in Table II. It is interesting to point out that the k cat (vTIM, 10°C) is very similar to the k cat (eTIM, 25°C), as shown in Table II. An important difference between vTIM and eTIM is the K i (2PG), which for vTIM is approximately 10 times higher, showing that the affinity of the vTIM active site for 2PG is much weaker. It is not clear from the structure how to explain this difference. For example an analysis of the presence of the charged side chains within a sphere of radius 8 Å from the phosphate position showed no difference between vTIM and eTIM, except for the presence of an asparagine at position 216 in eTIM, where in vTIM a lysine is found (the NZ atom is at a distance of 7.1 Å from the phosphate). The Asn-216 in eTIM is responsible for the N-capping of helix 7, whereas in vTIM Pro-217 has the same role.

Characterization of the A238S Mutant
To assess the role of the unique Ala-238 in loop 8 of vTIM, we decided to study the point mutant A238S. The mutant was produced and purified in the same manner as the native TIM. Its catalytic efficiency and its stability were analyzed (Table  II). Activity measurements at different temperatures revealed that the mutant has a higher thermal stability than the corresponding wild type; for example the half-lifes of wild type and A238S at 25°C are 10 and 27 min, respectively (Table II). This was confirmed by microcalorimetric studies (Fig. 4), showing that thermal unfolding of the mutant occurs at higher temperatures. In both cases, denaturation was found to be irreversible under the conditions tested. The values of the temperature of half-denaturation T d , calculated for a protein concentration of 1.3 mg/ml, were 40.8°C for the wild type enzyme and 45.8°C for the mutant.
The kinetic parameters of the A238S mutant were determined at 10°C and are compared with those of the wild type vTIM and eTIM (Table II). The mutant has a k cat value that is somewhat lower than that observed for the wild type, whereas its K m is somewhat higher, implying that the mutant is 3.8 times less efficient at 10°C. The K i (2PG) for wild type and mutant is rather similar. These data demonstrate that an alanine at position 238 contributes to a lower thermal stability (more psychrophilic) but allows for a higher catalytic efficiency at low temperatures.

X-ray Structure
The vTIM(sulfate) Structure-The model of vTIM consists of residues 2-250 for subunit A and subunit B. The N-terminal methionine was not included in the model because the electron density map was not clear in this region in both subunits. Nevertheless, the highest peak in the final (Fo-Fc) ␣ c map is at the N termini of both subunits, in agreement with the notion that the methionine is present but disordered.
The stereochemical quality of the refined structure was analyzed with the program PROCHECK (34). All stereochemical parameters for the main chain and side chain atoms are better than those for proteins refined to a similar resolution. The Ramachandran plot shows all the residues in the allowed regions of conformational space with one exception. The catalytic residue Lys-11 has an unusual conformation in both subunits ( Ϸ 58°; Ϸ 140°), although it has low B values and is well determined in the electron density. The lysine residue has to adopt this unusual conformation to position its side chain at a very precise location in the active-site pocket, a general feature seen in all TIM structures.
In Fig. 5 the real space density correlation coefficient for the main chain atoms is displayed as a function of residue number for both monomers. The lowest values are found in loop 6, which is poorly defined in the electron density map. Loop 6 has the highest B-factors as also shown in Fig. 5. The B-factors of the second subunit are consistently higher than the B-factors of the first subunit. This is due to the different packing interactions; for each of the dimers in the asymmetric unit it is observed that the first subunit has more crystallographic contacts when compared with the second subunit. For the first subunit approximately 45 intermolecular contacts are observed, and for the second subunit only 23 (when using a cutoff of 3.5 Å) are observed.
Both subunits are very similar with a C ␣ r.m.s. deviation of 0.29 Å for the framework C ␣ atoms and of 0.45 Å for all main chain C ␣ atoms. Loop 6 is in the open conformation, and Glu-169 adopts the swung-out conformation (19), contacting Ser-98 of loop 4. A sulfate ion occupies the active-site pocket of subunit A but not of subunit B. The average thermal B-factor of the atoms of this ion is high (70 Å 2 ), suggesting a low occupancy of this site. and stability data of vTIM, its A238S variant, and eTIM The kinetic parameters of vTIM and its A238S variant were determined at 10°C, whereas those of eTIM were measured at 25°C. t 1/2 values were calculated with data as shown in Fig. 3; T d , the temperature of half-denaturation, was determined from Fig. 4 a NA, not applicable. These data were not measured because thermal inactivation of eTIM is undetectable at these temperatures (22). The vTIM(2PG) Structure-The structure of vTIM(2PG) is very similar to the structure of vTIM(sulfate), with a C ␣ r.m.s. deviation of 0.18 Å for the framework C ␣ atoms and of 0.95 Å for all main chain C ␣ atoms of the first monomer. The most important differences are the presence of 2PG in both active sites of the dimer and, consequently, the closed conformation of loop 6 in both subunits (Fig. 6). In the vTIM(2PG) structure, loop 6 is well defined by the electron density map, and the loop 6 atoms do have lower B-factors when compared with the vTIM(sulfate) structure (Fig. 5). The catalytic glutamate, Glu-169, at the beginning of loop 6 is now seen in the swung-in conformation, contacting the 2PG molecule, as already seen in other TIM complexes. The 2PG phosphate position does not exactly superimpose on the sulfate of the vTIM(sulfate) structure. The distance between the equivalent S and P atoms is 1.25 Å. As shown in Fig. 6, the phosphate moiety is bound deeper into the active-site pocket. This fact has also been seen in similar complexes of trypanosomal TIM (tTIM) (19).
Comparison of the vTIM with the eTIM Structure-The sequences of vTIM and eTIM have almost the same length (256 for vTIM and 255 for eTIM) with a quite high sequence homology (66% sequence identity), suggesting a close evolutionary relationship. There is only one insertion of two residues at the C-terminal end of helix ␣2 (Fig. 1), resulting in a longer helix ␣2 in vTIM (Fig. 6). This difference is located at the N-terminal end of the ␤-barrel, and concerns residues not involved in catalysis and dimer formation. The overall fold of both vTIM structures and the unliganded eTIM structure has been compared in Fig. 6. The rms difference after superposition of the 129 framework C ␣ atoms of the vTIM(sulfate) and the eTIM structures is 0.59 Å for these framework atoms. A detailed structural analysis of the eTIM structure has shown some unique structural features (6), the majority of which are also present in vTIM. For example in vTIM and eTIM a methionine is located at position 7 in strand ␤1, where in other known TIM structures a glycine or an alanine is located. This methionine points into the interior of the barrel. The larger space required is compensated for by smaller residues in the environment. Another unique common feature of vTIM and eTIM is found in the region between loop 1 and loop 8, as shown in Fig. 7, in which the structures of vTIM, eTIM, and tTIM are superimposed. A buried water molecule is observed in this region in vTIM and eTIM, hydrogen-bonded to NE1(Trp-10) and O(Ala-238). The presence of this water molecule does correlate with two sequence changes in this region; in particular it concerns the mutation of an alanine (in tTIM) to a glycine (in vTIM and eTIM) at position 8 and of a phenylalanine (in tTIM) to a leucine (in vTIM and eTIM) at position 26. As a result of these two changes to a smaller side chain, a cavity is formed that is filled up by a water molecule. A small rotation of the Trp side chain is also observed. Consequently, the direct hydrogen bond of NE1(Trp10) and O(238) between loop 1 and loop 8 as seen in tTIM is replaced by the above-mentioned water-mediated hydrogen bond (Fig. 7). Another common feature of vTIM and eTIM is an intersubunit salt bridge between residue 79 and His-104 that is not observed in the other structures, where an intersubunit salt bridge exists between Glu-79 and Arg-100. The latter contact is absent in eTIM due to a reorientation of the side chains from the residues involved. In vTIM an aspartate is found at the position of Glu-79 in eTIM, and the previously mentioned salt bridge between residue 79 of one subunit and residue 100 of the other subunit is also present, as well as the Asp-79 -His-104 salt bridge.
There are three unique residues in vTIM when compared with the other nine known structures: Asp-79, Ala-182, and Ala-238. Only the latter is a unique residue when comparing the 45 known sequences of TIM. Position 79 is at the tip of loop 3; in most sequences (41 times) this is a glutamate involved in an intersubunit contact as discussed above; sometimes an aspartate (three times) or a threonine (one time) is observed. Position 182 is at the N terminus of helix 6; in most other sequences a proline (34 times) is seen at this N-cap position, but in some sequences a glycine (1 time), an alanine (6 times), a threonine (2 times), a valine (1 time), or an asparagine (1 time) is observed. As previously mentioned, at position 238 an alanine is found in vTIM, where in all the other sequences a serine is observed. The importance of Ala-238 is discussed in the next section.
The A238S Mutation in the Phosphate Binding Helix-Position 238 is at the beginning of the conserved phosphate binding 3/10 helix in loop 8. The structural and stability effects of an alanine to serine mutation has been studied in at least two other well characterized proteins: lysozyme and barnase (40,41). In all studied examples, the structural changes of this mutation are minimal. In none of the cases does the mutation significantly enhance the stability; moreover in almost all cases this mutation is destabilizing. The effects depend on the structural context of the mutated alanine; for fully exposed alanines, the mutation into a serine does only have minimal effects on the stability, but for fully buried alanines, the new variants are much less stable (41). The alanine to serine mutation in vTIM causes a significant increase in stability. This can be understood from the structural context of Ala 238 in vTIM. The C ␤ atom of Ala-238 is completely buried inside the protein interior, as is the serine side chain in the other TIM structures. Loop 8 of vTIM does superimpose very well on loop 8 of the other TIMs. Only small rearrangements are seen (Fig. 8). As a result no cavity can be detected in this region of the vTIM structure. Apparently the loop 8 helix is very rigid, as seen from the B-factor plot in Fig. 5. Important for the structure comparison is the observation that in the other TIMs, for example eTIM (Fig. 8), the buried O ␥ atom of the conserved serine makes two hydrogen bonds to main chain oxygen atoms. As can be seen in Fig. 8, each of these two buried main chain oxygens is making another hydrogen bond with a neighboring main chain peptide NH. These hydrogen bonds are also present in wild type vTIM. Assuming no major structural rearrangements, the alanine to serine mutation in vTIM will generate two additional buried hydrogen bonds, which rationalizes the extra stability of this variant (Table II). The assumption of only minor structural rearrangements is entirely consistent with the observation that the kinetic properties of vTIM and A238S are rather similar. The increase in stability, as expressed in a T d increase of approximately 5°C is quite significant. For example it is larger than the most stabilizing point mutation observed in more than 190 lysozyme variants (only 48 stabilizing, from which only 4 with an increase of T d higher than 1.5°C and with a maximum of 2.8°C) (42). This is partly due to the fact that TIM is a dimer; therefore the more favorable hydrogen bond interactions occur twice. Furthermore it is also due to the very good hydrogen bonding complementarity of the environment for a serine at this site (Fig. 8).
Concluding Remarks-In this paper we reported the purification and the kinetic and stability properties of triose-phosphate isomerase from the psychrophilic bacterium V. marinus and its A238S mutant as well as two x-ray structures of the native enzyme, one complexed with sulfate and one complexed with the substrate analogue 2PG. The vTIM structure is very similar to the structure of eTIM, in agreement with the high sequence identity. The comparison of the structures of vTIM and eTIM does not provide an explanation for the 10-fold lower affinity of 2PG for vTIM, when compared with eTIM. One unique sequence feature of vTIM has been found, which is Ala-238, whereas in all other TIM sequences a serine is observed. The importance of the alanine at position 238 has been investigated further by studying the properties of the A238S mutant. A238S is considerable more stable than wild type, and it catalyzes the reaction less efficiently; therefore, there is good evidence that the alanine at position 238 is a successful adaptation to the psychrophilic character of V. marinus.