Analysis of an invariant cofactor-protein interaction in thiamin diphosphate-dependent enzymes by site-directed mutagenesis. Glutamic acid 418 in transketolase is essential for catalysis.

A homologous expression system and a purification protocol for pure, highly active recombinant yeast transketolase have been developed. The invariant transketolase residue Glu418, which forms a hydrogen bond to the N-1' nitrogen atom of the pyrimidine ring of the cofactor thiamin diphosphate has been replaced by glutamine and alanine. Crystallographic analyses of the mutants show that these amino acid substitutions do not induce structural changes beyond the site of mutation. In both cases, the cofactor binds in a manner identical to the wild-type enzyme. Significant differences in the CD spectra of the mutant transketolases compared with the spectrum of wild-type enzyme indicate differences in the electron distribution of the aminopyrimidine ring of the cofactor. The E418Q mutant shows 2% and the E418A mutant shows about 0.1% of the catalytic activity of wild-type enzyme. The affinities of the mutant enzymes for thiamin diphosphate are comparable with wild-type transketolase. The hydrogen bond between the coenzyme and the side chain of Glu418 is thus not required for coenzyme binding but essential for catalytic activity. The results demonstrate the functional importance of this interaction and support the molecular model for cofactor deprotonation, the first step in enzymatic thiamin catalysis.

Fa: 46-18-536971. ThDP-dependent transketolase catalyzes the interconversion of monosaccharides by transferring glycoaldehyde groups from ketoses to aldoses (for a review see Kochetov (1982)) . The crystallographic analyses of transketolase in complex with the coenzyme ThDP or the coenzyme analogue thiamin thiazolon diphosphate, respectively, have provided detailed information on the mode of coenzyme binding and the interactions of ThDP with the enzyme (Lindqvist et ai., 1992, Nilsson et al., 1993, Nikkola et al., 1994.
In the crystal structure of transketolase, there is no enzymic base in a position suitable for proton abstraction at the C-2 atom of the thiazolium ring of ThDP. Instead, a molecular mechanism for proton abstraction has been proposed that suggests a cofactor-assisted deprotonation (Lindqvist et dl., 1992;. Essential to this model is the observed hydrogen bond between the side chain of the invariant residue G1u*Is and the N-1' nitrogen atom of ThDP ( Fig. 11. This interaction has also been found in other ThDP-dependent enzymes (Muller and Schulz, 1993;Dyda et al., 19931, and it has been proposed that the mechanism for generating the C-2 carbanion, the first step in enzymatic thiamin catalysis, is common to all ThDP-dependent enzymes (Muller et at., 1993, SEhneider and.
Chemical modification of ThDP through replacement of the N-1' nitrogen by a carbon atom has provided evidence that the N-1' nitrogen atom of the pyrimidine ring is indeed required for catalytic activity (Schellenberger, 1967;Golbik et al., 1991). Subsequently it was shown that the binding mode of this cofactor analogue to transketolase is identical to the binding of ThDP (Ktinig et al., 19941, and structural differences in cofactor binding can therefore be ruled out as being responsible for the loss of catalytic function. Replacement of the N-3' nitrogen atom of the pyrimidine ring results in a cofactor analogue that is catalytically active.
In this paper, we describe an efficient expression system for yeast transketolase with high yields of pure and highly active recombinant enzyme, We also report on the results of a mutagenesis study to probe the interaction between ThDP and the invariant residue G I U~'~ in transketolase and discuss general implications of these results for the enzymatic reaction mechanisms of thiamin catalysis. EXPERIMENTa PROCEDURES ~u t e r~u~s~o m m o n l y used chemicals and reagents were of the highest purity readily available. Glyceraldehyde-3-phospha~ dehydrogenase was prepared fmm rabbit muscle as described by Eladi and Szorengi ( 1956 Yeast and Bacterial Strains and Plasmids-The yeast-Escherichia coli shuttle vector pTKL1, used for both site-directed mutagenesis and expression of transketolase, was derived from the genomic library made in the plasmid pHR81 (Nehlin et al., 1989;Sundstrom et al., 1993). pTKLl consists of a 2664-bp fragment inserted into the BamHI site of pHR81. This fragment contains the 2040-bp-long T a l , flanked by an upstream sequence of 472 bp and a sequence of 152 bp downstream. E. coli strain BMH 71-18 mutS was used in the mutagenesis step. All other manipulations of pTKLl in E. coli were done using the strain JM109 (Yanisch-Perron et aE., 19851. The yeast strain H402, with a disrupted TKLl, is a tklZ::HIS3 derivative of W303-1A Thomas and Rothstein, 1989). H402 was transformed with pTKL1 using the method described by Soni et at. (19931, selecting for URA3 in pTKL1. Site-directed Mutagenesis-Unless otherwise indicated, standard molecular biology procedures (Maniatis et al., 1982;Sambrook et al., 1989) were used. Single-and double-stranded plasmid DNA was prepared with the Wizard kits (Promega). Mutagenesis was performed using the reagents in the Altered Sites kit (Promega) using singlestranded template ipTKL1) prepared with the fl helper phage R408 (Russel et al., 1986). Selection was based on the elimination of the unique Sal1 site in pTKLl according to the method described by Deng and Nickoloff (1992). Templates for DNA sequencing were obtained by preparing plasmid DNA (Caldwell and Becker, 1993) from colonies of yeast harboring the mutated genes and amplifying the plasmids in E. coli JMlO9. The transketolase mutants were sequenced throughout their entire coding region.
Expression of ~n s~e~o l a s e -H 4 O~p T~l cells were cultured in a modification of the synthetic medium described by Sherman et al., (1986). Glucose was replaced by galactose as carbon source, and ammonium sulfate i5 g/liter) was added separately. Leucine was omitted to obtain a high copy number of the plasmid (Erhart and Hollenberg, 1983). The cells were grown at 30 "C, 150 rpm. A 100-ml preculture was grown for 2 days. This culture was used to inoculate a 5-liter culture that was grown for approximately 36 h and then harvested.
Protein Purification-All steps were carried out at 4 "C. The cells were collected by centrifugation at 2500 x g for 4 min. The pellet (approximately 60 g of packed cells) was resuspended in a small volume of 100 m M Tris (pH 7.6), containing 100 m~ lithium chloride, 100 m M EDTA, 1 m M phenylmethylsulfonyl fluoride, 1 m M benzamidine bydrochloride, 3.2 RM ieupeptin, 4.4 p~ pepstatin A, 0.2% (w/v) sodium dode-cy1 sulfate, and 1% (w/v) dimethyl sulfoxide. The cells were disrupted in a cell mill (Biospec Products) for 30 see 6 times with 3 min of cooling between each round. The homogenate was clarified by centrifugation at 10,000 x .g for 20 min, and the supernatant was filtered through glass wool. Streptomycin sulfate was added to a final concentration of 1%. The mixture was stirred gently for 30 min, and the precipitate was removed by centrifugation at 10,000 x g for 20 min. Ammonium sulfate was added to 70% saturation, and the precipitate was collected by centrifugation at 10,000 x g for 20 min. The pellet was resuspended in 40 ml of 20 m M Tris buffer (unless stated otherwise, the pH of the buffers used was 8.21, containing 300 m M NaCl. Undissolved protein was removed by centrifugation at 25,000 x g for 30 min. The supernatant was loaded on a Sephadex G-25 column (2.6 x 40 cm) equilibrated with the same buffer. Desalted protein was concentrated in an Amicon ultrafiltration cell to approximately 20 ml. Concentrated protein was loaded on a Sephacryl S-300 HR column (5.0 x 100 cm), equilibrated with 20 m M Tris containing 300 m M NaCl and eluted with the same buffer. Fractions containing transketolase activity were concentrated in an Amicon ultrafiltration cell to 10 ml. The concentrated protein was loaded on a Sephadex G-50 coiumn (1.5 x 25 cm) equilibrated with 20 m M Tris. Desalted protein was divided into aliquots of approximately I00 mg of protein and loaded on a Mono-& HR 10/10 column equilibrated with 20 m M Tris. Transketolase was eluted at a flow rate of 1 mVmin with a 0.0-75 m M linear gradient of NaCl in 20 m M Tris buffer. Transketolase was detected in two peaks. Pure, homogeneous enzyme was found in the first peak at about 22-33 m M NaCl. The second peak of transketolase appeared immediately after the first hut was not homogeneous and was therefore discarded. Transketolase was precipitated by addition of solid ammonium sulfate to 70% saturation and collected by centrifugation at 25,000 x g for 15 min. The pellet was dissolved in 25 m M glycyl-glycine (pH 7.6) to a final concentration of transketolase of approximately 20 mg/ml. The enzyme can be stared as ammonium sulfate precipitate at 4 "C without loss of enzymic activity. For long term storage, samples of transketolase were frozen in liquid nitrogen and stored at -80 "C.
Gel Electrophoresis and Western Blots-Polyacrylamide gel electrophoresis was run under denaturating conditions in the presence of sodium dodecyl sulfate in 10% (w/v) slab gels in the Trisiglycine system (Laemmli, 1970). Gels were stained with Coomassie Brilliant Blue R.
PhastTransfer (Pharmacia) and ProtoBlot I1 AP System (Promega) were used for Western blotting on PVDF-PLUS membranes according to the manufacturer's recommendations.
Antiserum against Zkansketolase-Polyclonal antiserum against transketolase was obtained from rabbits immunized and boosted with wild-type transketolase. A partial purification of the antiserum was performed using a DEAE-Sepharose CL-GB ion exchanger. A sample of polyclonat antibodies against transketolase was also provided by Dr. N. Tikhamirova tMoscow).
Activity Measurements and Protein Determination-The specific activity of transketolase was measured spectrophotometrically in two different ways (de la Haba et ai., 1955;Racker, 1961;Kochetov, 1982). In both assays, 1 unit is defined as the formation of 1 pmol of glyceraldehyde 3-phosphate per minute. In one assay, the reaction is followed by the rate of NAD+ reduction in a coupled system with glyceraldehyde-3phosphate dehydrogenase (Racker, 1961). In the second assay, activity is measured by the rate of NADH oxidation in a coupled system with triose-phosphate isomerase and glycerol-3-phosphate dehydrogenase (de la Haba et al., 1955). The first method has been used for all of the kinetic experiments, and the second assay was used for activity measurements in the crude extract and during the purification procedure.
The dependence of transketolase activity on pH has been determined in the pH interval 6.6-8.6 using MOPS buffer in the pH interval 6.6-7.6 and glycyl-glycine buffer in the pH interval 7.6-8.6. The rate of the reaction was followed spectrophotometrically (Racker, 1961) and by formation of sedoheptulose 7-phosphate. The concentration of sedoheptulose 7-phosphate was determined colorimetrically (Villafranea et al., 1971;Osawa et al., 1972). Sedoheptulose anhydride was used as a standard. Protein concentrations were determined in the crude extract by the method of Lowry (Lowry et al., 1951) using the Bio-Rad DC protein assay with bovine serum albumin as a standard. Otherwise, protein Concentrations were determined using the extinction coefficient E&cm = 14.5 (Heinrich et al., 1972a).
Determination of Kinetic Parameters-For the recombinant enzyme and the mutant E418Q, kinetic parameters were determined from measurements of initial rates at different ThDP or substrate concentrations with D-xylulose 5-phosphate as donor and D-ribose 5-phosphate as acceptor substrate. Before the start of the reaction by the addition of substrates, transketolase was incubated with ThDP and metal ions for 15 min at 25 "C. All measurements were performed at 25 "C in triplicate. The data were analyzed by nonlinear regression analysis resulting in values of K , and V, , and standard errors with the program ULTRAFIT (Biosoft).
Binding of ThDP-In the case of the E418A mutant, the binding constant of ThDP was estimated by equilibrium dialysis. The concentration of free ThDP was determined spectroscopically based on = 7.500 M" cm" (Schellenberger et al., 1965).
Circular Dichroism-Near-UV CD spectra were measured with a Jasco 5-41 A spectropolarimeter using a path length of 1.0 cm at 25 "C.
Crystallography-Mutant transketolase was crystallized by vapor diffusion under conditions similar to those used for hoiotransketolase (Schneider et al., 1989). 7.5 111 of a 12 mgiml protein solution were mixed with the same amount of mother liquid, 13-17% (w/w) of polyethylene I, The R-factor for the E418Q mutant is calculated on all data in the resolution interval 6.5-2.3 A and for the E418A mutant in the interval 6.5-2.9 A.
glycol 6000 in 50 mM glycyl-glycine buffer, containing 5 mM CaCI, and 5 mM ThDP a t pH 7.9, and the droplets were left to equilibrate with 1 ml of the mother liquor. The crystals obtained were isomorphous to those of holotransketolase, space group P2,2,2, with cell dimensions a = 76.5 A, b = 113.3 A, and c = 160.9 A.
The x-ray data set for the E418Q mutant was collected with a R-AXIS I1 imaging plate mounted on a Rigaku rotating anode. The data were processed with the MSC software (Sato et al., 1992). The data set for the E418A mutant was collected with a Xentronics multiwire area detector mounted on a Rigaku rotating anode. Frames were evaluated with the BUDDHA program (Blum et al., 19871, and the data were scaled with the CCP4 program suite (SERC, 1979). Initial phase information for the calculation of electron density maps was derived from the model of holotransketolase refined to 2.0-A resolution, which has a crystallographic R-factor of 15.8% (Nikkola et al., 1994). For crystallographic refinement, the program package XPLOR (Briinger et al., 1989) was used. Inspection of the electron density maps, model building, and structural comparisons were carried out with the program 0 (Jones et al., 1991). Details of data collection statistics and crystallographic refinement are given in Table I. Structure factors and atomic coordinates for the E418Q and E418Amutants have been deposited with the Protein Data Bank, Brookhaven.

Gene Disruption and Background Activity in the Expression
Strain-The yeast strain H402, which has a disrupted T a l gene, cannot grow on the culture medium used in this work . As a control of the T a l gene disruption experiments, a 2467-bp Sac1 fragment, containing the promoter region and the complete transketolase gene (with the exception of the last 60 bp) was removed from pTKL1. The plasmid was religated, yielding pTKA. After transformation of the Tal-disrupted yeast strain H402 with the plasmid pTKA, it is possible to analyze strain H402 under identical expression conditions as when producing recombinant enzyme. In cell extracts of H402 and H402xpTKA, no trace of transketolase activity was found. SDS-PAGE gels of the crude extract revealed that transketolase is absent (Fig. 2). Western blots of crude extract verified the absence of a protein able to cross-react with polyclonal antibodies raised against wild-type transketolase (data not shown).
Expression and Purification of Recombinant Dansketolase-Yeast strain H402xpTKL1 expresses recombinant transketolase a t high levels, about 15% of total cell protein. A summary of the purification procedure starting from 50 g of wet yeast cells is shown in Table 11. The specific activity of purified transketolase was about 37 unitslmg at 25 "C. The purification protocol results in about 60-80 mg of pure transketolase from 5 liters of cell culture. SDS-PAGE electrophoresis (Fig. 2) shows that purified recombinant transketolase was homogenous and electrophoretically indistinguishable from wild-type enzyme.
Site-directed Mutagenesis and Properties of Mutant Enzyme-The invariant residue Glu418 was replaced by gluta- mine and alanine, respectively. As judged from polyacrylamide gels and total yield of purified protein, the levels of mutant transketolase expressed in H402xpTKL1 cells are similar to the level of wild-type transketolase. Both mutants, E418A and E418Q could be readily purified according to the protocol for the wild-type recombinant enzyme with similar final yields of pure protein (Fig. 2).
The E418Q mutant retained only 2% of the wild-type catalytic activity (Table III), but the apparent K, for ThDP (0.86 PM) is similar to the K,,, for wild-type enzyme (0.47 p~) .
The K, value for the donor substrate xylulose 5-phosphate is about twice the value for the wild-type enzyme, whereas the K, value for the acceptor substrate ribose 5-phosphate is, within the error limits of the experiment, identical to that of wild-type transketolase (Table 111). The E418Q mutant enzyme has the same pH optimum (pH 7.6-8.2) as wild-type transketolase.
Substitution of glutamic acid 418 with alanine has a more profound effect on catalytic activity. This mutant shows a 1000fold decrease in activity (Table 111). The affinity of the mutant for the cofactor ThDP is, however, essentially unchanged as can be seen from the similarities in binding constants (Table 111). The residual catalytic activity remains constant in the pH range from 7.6 to 8.6, and there is no decrease in activity at high pH, as observed in wild type and the E418Q mutant. Because of the very low activity of this mutant, K, values for substrates could not be determined.
Circular Dichroism Measurements-Binding of ThDP to transketolase can be monitored by characteristic changes in the near-W region of the CD spectrum. Addition of ThDP to the enzyme leads to an increase of the maximum at 280 nm and the appearance of a negative broad band with a minimum at 320 nm (Kochetov and Usmanov, 1970). The CD spectra of the recombinant enzyme and the E418Aand E418Q mutants in the absence of ThDP are very similar to the spectrum of apotransketolase. Upon addition of ThDP, recombinant transketolase (Fig. 3) shows the same spectral change as nonrecombinant enzyme. The CD spectra of the mutants in the presence of ThDP are qualitatively similar to each other but are very different from the spectrum of wild-type transketolase. In the mutant spectra, the maxima a t 280 nm and 320 nm are shifted toward shorter wavelengths at 265 and 295 nm, respectively (Fig. 3). HR a One unit is defined as 1.0 pmol of glyceraldehyde 3-phosphate produced per min in a coupled assay with ribose 5-phosphate and xylulose 5-phosphate as substrates.
a ICm values were determined for xylulose 5-phosphate as donor and Mean value of five independent measurements. ND, not determined because of low catalytic activity. tion also shows that no significant local structural changes are found in the two mutants when compared with the structure of wild-type transketolase (Fig. 4). Similarly, the positions of the Ca2' ion and ThDP are, within the error limits of the electron density maps, identical. Fig. 5 shows a view of the electron density maps at the site of mutation. In the E418Q mutant, the observed electron density at position 418 is identi5al to the one found in wild-type transketolase. Indeed, at 2.3-Aresolution, a crystallographic distinction between glutamate and glutamine side chains is not possible. In the refined model of the E418Q mutant, the glutamine side chain forms a hydrogen bond to the N-1' atom of the pyrimidine ring. Other interactions of the cofactor with the enzyme are unchanged upon amino acid substitution.
In the I F, I -IF, I electron density map for the E418A mutant, calculated with phases from the wild-type enzyme, negative electron density is found at the position of the glutamate side chain, indicating that this residue has been replaced by an alanine. In the electron density maps calculated from the refined model of the E418A mutant, strong electron density is found in the proximity of the N-1' nitrogen atom of the pyrimidine ring of ThDP (Fig. 5b). Most likely, this electron density represents a solvent molecule tightly bound in the cavity created by the removal of the glutamic acid side chain. This solvent molecule is close to the 0 -y oxygen atom of the side chain of Thr441 and to the main chain oxygen atom ~f A l a~*~, and it is very likely that it is held firmly in place through hydrogen bonds to these atoms and the N-1' nitrogen of ThDP.

DISCUSSION
Transketolase catalyzes two reactions of the nonreductive pentose phosphate pathway and is found in all organisms. To avoid misinterpretations of mutagenesis studies caused by copurification of endogenous enzyme, we have produced the yeast strain H402 where the TKLl gene has been disrupted ( S~d s t r o m et ai., 1993). The present study provides evidence by activity measurements and gel electrophoresis that cell extracts from this TKLI-disrupted yeast strain indeed do not contain any transketolase. Using similar gene disruption experiments with the TKLJ gene, Zimmermann and co-workers (Schaff-Gerstenschlager et al., 1993) also reported a yeast strain without endogenous transketolase.
Transketolase from S a c c~a~o~~c e s cereuisiae has been purified to homogeneity earlier, however, with low yields (de la Haba et al., 1955;Saitou et al., 1974). The expression system described here produces considerable amounts of recombinant enzyme from modest amounts of yeast cells. The recombinant enzyme is highly active, with a specific activity between 35 and 40 unitdmg. This value is higher than for most transketolase preparations described earlier, which had specific activities ranging from 12 to 20 unitdmg. Tikhomirova et al. (1991) described a purification protocol using affinity chromatography resulting in a transke~lase with specific activities between 12 and 60 units/mg dependent on the yeast source, with activities typically around 30 unitslmg.
High resolution crystallographic studies of recombinant enzyme (Konig et al., 1994) did not detect any structural differ-ences between wild-type and recombinant transketolase. The binding constant of the cofactor ThDP to recombinant enzyme is, within the error limit of the experiment, identical to the value found using wild-type enzyme (Heinrich et ai., 197213). CD spectra of recombinant apo-and holotransketolase are identical to the corresponding spectra obtained from wild-type enzyme. We conclude from these data that recombinant yeast transketolase is structurally and functionally indistinguishable from the endogenous enzyme.
Substitution of the invariant residue Glu418 with a glutamine or an alanine residue has little effect on the affinity for substrate and, surprisingly, ThDP (Table 111). From the three-dimensional structure, it is obvious that the side chain of GluQ1' is shielded from the substrate by the cofactor ThDP and cannot form any direct contacts to the substrate. As anticipated from the three-dimensional structure, mutations at position 418 do indeed have little effect on the affinities of the enzyme for the substrates. The side chain of Glu418 forms, however, a hydrogen bond to the N-1' nitrogen atom of the cofactor, an interaction found in all ThDP-dependent enzymes (Muller et aE., 1993). Nevertheless, removal of this hydrogen bond by either chemical modification of the cofactor (Golbik et al., 1991;Konig et al., 1994) or site-directed mutagenesis of the enzyme has little effect on the binding constants of ThDP or its analogues and does not contribute to coenzyme binding in a significant manner.
Removal of this hydrogen bond results, however, in a drastic drop in catalytic activity. Structural effects caused by the amino acid substitution can be excluded as revealed by the crystallographic analysis of the mutants. The results of the mutagenesis study therefore must reflect functional deficiencies in the mutant enzymes, and we conclude that the interaction of the side chain of E418 with the cofactor is not required for cofactor binding but for catalytic activity.
The appearance of the strong absorbance band in the CD spectrum upon binding of the coenzyme has been interpreted as being caused by the formation of a charge-transfer complex between ThDP and an aromatic residue at the coenzyme binding site of the enzyme (Kochetov and Usmanov, 1970). The crystal structure analysis of holotransketolase (Lindqvist et al., 1992;Nikkola et al., 1994) shows that the pyrimidine ring of the cofactor binds in a hydrophobic pocket and interacts tightly with side chains of conserved aromatic residues, in particular Phe445 and vU8, which could give rise to the formation of a charge-transfer complex. If this interpretation of the near-UV CD spectra is correct, then they will provide a sensitive probe to analyze the microenvironment of the pyrimidine ring, since the charge-transfer band will depend strongly on the electron distnbut~on in the pyrimidine ring.
The replacement of Glu418 by alanine or glutamine shifts the absorption maximum in the near-W region of the CD spectrum by 25 nm, indicating a change in the distribution of the n-electron system of the pyrimidine ring. The spectral changes are largest in the E418A mutant, where the hydrogen bond to the N-1' nitrogen atom has been removed completely, and less in the E418Q mutant, where this hydrogen bond, albeit weaker, is still present.
Based on the crystal structure of transketolase, we have suggested a scheme for cofactor-assisted deprotonation of the C-2 atom of the thiazolium ring, the first step of enzymatic thiamin catalysis (Fig, 6), and the chemical arguments behind this model have been discussed at length elsewhere (Lindqvist et al., 1992;. In this model, the base abstracting the proton from the C-2 carbon atom is the 4-imino group of the pyrimidine ring. The hydrogen bond between the N-l' nitrogen atom and the side chain of Glu4'8 is The refined protein model of the mutant is superimposed. The contour level of the electron density is at the standard deviation of the electron density map. Hydrogen bonds are indicated by dashed lines. The position of the putative solvent molecule close to the N-1' atom of the coenzyme in the E418A mutant is indicated by a sphere. essential for the conversion of the 4-amino group to the 4-imino group during the catalytic cycle. It is known from experiments with model compounds that pro~nation of the N-1' nitrogen in pyrimidine derivatives will shift the p K , of the 4-amino group (Jordan et al., 1982). Protonation of the N-1' atom will also perturb the distribution of ?r-electrons in the ring by promoting resonance form I1 (see Fig. 6 ) . This change in electron distribution will certainly influence the charge-transfer complex of the cofactor with the aromatic side chain&) in the hydrophobic pocket at the coenzyme binding site and might be the cause for the observed changes in the CD spectrum of the mutants. In this context, it is interesting to note that mutations in transketolase at positions not directly interfering with the interactions of the aminopyrimidine ring and the protein have CD spectra very similar to wild-type enzyme.' The 2% residual activity observed in the E418Q mutant fits rather well into this model since, as revealed by the crystal structure analysis, this side chain still forms the hydrogen bond to the N-1' nitrogen atom. However, for chemical reasons this hydrogen bond is expected to be weaker than the corre-C. Wlkner, S. Backstrom, T. Kostikowa, L. Meshalkina, Y. Lindqvist, and G. Schneider, unpublish~d results. sponding interaction in the wild-type enzyme, possibly resulting in a smailer pKa shift for the 4-amino group and, therefore, lower activity. The low activity of the E418A mutant is then readily explained by the complete absence of this interaction in this mutant. One might have anticipated that the activity in the E418A mutant caused by the complete absence of this interaction should be even lower than 0.1% of wild-type activity. The crystal structure of this mutant offers a possible explanation. Despite the moderate resolution of the electron density map, clear indications of a water molecule close to the N-1' atom were found. It is conceivable that this solvent molecule can act to a certain extent as a hydrogen bond donor and thus be responsible for the residual activity of this mutant.
In conclusion, the functional significance of the invariant cofactor-enzyme interaction between a glutamate side chain and the N-1' nitrogen atom of the cofactor has been analyzed by chemical replacement of the nitrogen atom by a carbon atom (Golbik et al., 1991;Konig et al., 1994) and site-directed mutagenesis reported here. The results from both studies show that this interaction is of functional importance and support the general model for cofactor-assisted deprotonation as the first step in enzymatic thiamin catalysis.