Probing the catalytic mechanism of GDP-4-keto-6-deoxy--mannose epimerase/reductase by kinetic and crystallographic characterization of site-specific mutants1

doi:10.1006/jmbi.2000.4106 available online at http://www.idealibrary.com on J. Mol. Biol. (2000) 303, 77±91 Probing the Catalytic Mechanism of GDP-4-keto-6deoxy-D-mannose Epimerase/Reductase by Kinetic and Crystallographic Characterization of Site-specific Mutants Camillo Rosano1, Angela Bisso2, Gaetano Izzo1, Michela Tonetti2 Laura Sturla2, Antonio De Flora2 and Martino Bolognesi1* 1 Department of Physics-INFM and Advanced Biotechnology Center-IST, University of Genova Largo Rosanna Benzi 10 I-16132 Genova, Italy 2 Department of Experimental Medicine, Section of Biochemistry, University of Genova, Viale Benedetto XV 1 I-16132 Genova, Italy GDP-4-keto-6-deoxy-D-mannose epimerase/reductase is a bifunctional enzyme responsible for the last step in the biosynthesis of GDP-L-fucose, the substrate of fucosyl transferases. Several cell-surface antigens, including the leukocyte Lewis system and cell-surface antigens in pathogenic bacteria, depend on the availability of GDP-L-fucose for their expression. Therefore, the enzyme is a potential target for therapy in pathological states depending on selectin-mediated cell-to-cell interactions. Previous crystallographic investigations have shown that GDP-4-keto-6-deoxy-Dmannose epimerase/reductase belongs to the short-chain dehydrogenase/reductase protein homology family. The enzyme active-site region is at the interface of an N-terminal NADPH-binding domain and a C-terminal domain, held to bind the substrate. The design, expression and functional characterization of seven site-speci®c mutant forms of GDP-4-keto-6-deoxy-D-mannose epimerase/reductase are reported here. In parallel, the crystal structures of the native holoenzyme and of three mutants (Ser107Ala, Tyr136Glu and Lys140Arg) have been investigated Ê resolution, based on synchrotron data and re®ned at 1.45-1.60 A (R-factors range between 12.6 % and 13.9 %). The re®ned protein models show that besides the active-site residues Ser107, Tyr136 and Lys140, whose mutations impair the overall enzymatic activity and may affect the coenzyme binding mode, side-chains capable of proton exchange, located around the expected substrate (GDP-4-keto-6-deoxy-D-mannose) binding pocket, are selectively required during the epimerization and reduction steps. Among these, Cys109 and His179 may play a primary role in proton exchange between the enzyme and the epimerization catalytic intermediates. Finally, the additional role of mutated active-site residues involved in substrate recognition and in enzyme stability has been analyzed. # 2000 Academic Press *Corresponding author Keywords: GDP-L-fucose; enzyme structure; short-chain dehydrogenase; NADP‡; epimerization Introduction C.R. and A.B. contributed equally to this work. Abbreviations used: GMD, GDP-D-mannose 4,6dehydratase; GMER, GDP-4-keto-6-deoxy-D-mannose epimerase/reductase; RED, reductase, epimerase, dehydrogenase superfamily; SDR, short-chain dehydrogenase, reductase family; TLC, thin-layer chromatography; TFA, tri¯uoroacetic acid; UGE, UDP-galactose 4-epimerase. E-mail address of the corresponding author: bolognes@®sica.unige.it 0022-2836/00/010077±15 $35.00/0 Substantial evidence has accumulated in the past few years on the key role that glycoconjugates play in cell-to-cell interaction and adhesion processes, both in prokaryotes and in higher organisms (Feizi & Childs, 1987; Feizi, 1990; Zhang et al., 1997). In particular, some glycoconjugates containing the 6-deoxy-hexose L-fucose, the Lewis system antigens, have been shown to be the selectin ligands, involved in leukocyte and tumor cell adhesion to # 2000 Academic Press 78 the endothelium, as well as in development (Feizi & Childs, 1987; Brandley et al., 1990; Varki, 1994; Lowe, 1997). Lack of constituent L-fucose from these glycoconjugates, as observed in the human genetic syndrome LAD II (leukocytes adhesion de®ciency type II), leads to impaired binding to selectins, and eventually to severe symptoms, such as immunode®ciency and psychomotor retardation (Etzioni, 1992; Phillips et al., 1995). Oligosaccharide structures containing L-fucose are involved in microorganism and pathogen interactions with host tissues, as observed for Helicobacter pylori, where the expression of antigens related to the Lewis system contributes to molecular mimicry with the host and to the development of an autoimmune response (Appelmelk et al., 1997). In vivo, L-fucose is made available for insertion into cell-surface antigens as GDP-L-fucose, the substrate of fucosyl transferases. The biosynthesis of this nucleotide-sugar molecule occurs via a threestep metabolic pathway, common to both bacterial and animal cells, starting from GDP-D-mannose (Ginsburg, 1960, 1961; Tonetti et al., 1998a). The ®rst reaction in the pathway is the elimination of a water molecule from GDP-D-mannose, catalyzed by GDP-D-mannose 4,6-dehydratase (Sturla et al., 1997; Sullivan et al., 1998; Somoza et al, 2000), leading to the formation of GDP-4-keto-6-deoxy-Dmannose (see Figure 1). This intermediate undergoes subsequent epimerization reactions at the C-3 and C-5 hexose ring centers, leading to a change from the D to the L-con®guration, followed by an NADPH-dependent reduction on C-4. Both epimerase and reductase reactions are catalyzed by a single enzyme, GDP-4-keto-6-deoxy-D-mannose epimerase/reductase (GMER) (Tonetti et al., 1996), which was recognized as the long-known protein FX (Morelli & De Flora, 1977). Recent studies have shown that GMER can support the epimerization of the substrate also in the absence of NADPH from the incubation mixture, indicating independence of the epimerization and of the reduction reactions (Menon et al., 1999). Inspection of amino acid sequences indicates that GMER belongs to the continuously expanding family of short-chain dehydrogenase/reductases GDP-L-fucose Biosynthesis (SDRs), a protein homology family hosting enzymes involved in different biological functions such as alcohol, hydroxyprostaglandin, hydroxysteroid dehydrogenase, dihydropteridin reductase, carbonyl reductase and UDP-galactose epimerase (UGE) (Persson et al., 1991; Varughese et al., 1994; JoÈrnvall et al., 1995; Gosh et al., 1995; Tanaka et al., 1996a,b; Ensor & Tai, 1996; Thoden et al., 1997; Benach et al., 1998), which display conservation of a residue triad (Ser, Tyr and Lys) at their catalytic centers. In more general evolutionary terms, GMER, together with UGE and homologous SDR enzymes, is a component of the reductase-epimerase-dehydrogenase (RED) protein homology superfamily (Labesse et al., 1994). GMER from Escherichia coli, obtained as recombinant protein, has been crystallized in our laboratory, both as the apoenzyme and in a complex with NADP‡, and its X-ray structure has been Ê resolution (Tonetti et al., 1998b; elucidated at 2.1 A Rizzi et al., 1998). Similar results were obtained independently by Somers et al. (1998). Both studies have shown that wild-type GMER is associated in a dimeric structure in the crystalline state, closely matching the quaternary structure observed in homologous SDRs (Gosh et al., 1995; Benach et al., 1998). Each GMER subunit (321 amino acid residues; see Figure 2) is composed of two domains, that can be de®ned as mostly N-terminal (ca. 190 residues, adopting a Rossmann fold topology; Branden & Tooze, 1991; Rizzi et al., 1998), and mostly C-terminal (ca 100 residues). An extended NADP‡ coenzyme molecule is properly located at the Rossmann fold topological switch-point, next to the domain interface. Moreover, an evident cleft Ê 3), present in the C-terminal domain (>200 A region, hosting 25 ordered water molecules in the inhibitor-free holoenzyme, has been proposed as the substrate-binding site, based on homology modeling studies (Rizzi et al., 1998; Somers et al., 1998; Somoza et al., 2000). The crystallographic analyses have shown that the GMER active-site region is located at the domain interface, close to the nicotinamide reducing end and to the evolutionarily conserved residues Ser107, Tyr136 and Lys140 (see Figure 2). Figure 1. The GDP-L-fucose biosynthetic pathway and the products obtained after NaBH4 reduction of the intermediate compounds. GMD, GDP-D-mannose 4,6 dehydratase; GMER, GDP4-keto-6-deoxy-D-mannose epimerase/reductase; R-, GDP. 79 GDP-L-fucose Biosynthesis Results Epimerase and reductase activities in GMER mutants Figure 2. Ribbon representation of the GMER subunit structure, displaying the N and the C-terminal domains (lower and upper in the Figure, respectively) and the active-site cleft (central). The bound NADP‡ coenzyme (in yellow) and the active-site residues Ser107, Tyr136 and Lys140 (purple) are displayed. This Figure was drawn with DINO (Philippsen, 2000 http://www.bioz.unibas.ch/  xray/dino). Crystallization of GMER ternary complexes, containing the coenzyme and substrate-like inhibitors, has so far proved elusive (Rizzi et al., 1998; Somers et al., 1998). In the absence of direct crystallographic evidence, a study of the enzyme's catalytic mechanism must therefore rely on rational modeling of substrate binding, on the design of sitespeci®c mutants and on the analysis of their functional properties. In this context, we have undertaken a systematic investigation on GMER active-site mutants, focusing on the residues held responsible for the catalytic activity, or its assistance, and on those likely involved in substrate recognition. Here, we report kinetic and functional data for seven different GMER mutants, together with the crystal structures of three of these, and of the wild-type holo-enzyme, all determined Ê and 1.60 A Ê resolution. between 1.45 A Wild-type and mutant GMER forms were overexpressed in E. coli and puri®ed to homogeneity as GST-fusion proteins. The GST-Tag was then removed by proteolytic cleavage. High levels of expression were obtained for both wild-type and mutant proteins. One liter of bacterial culture yielded approximately 30 mg of pure wild-type GMER, while the yield for mutants ranged between 10 and 20 mg per liter of culture. The puri®ed proteins were found to be homogeneous by SDS-PAGE, migrating as single bands, with an apparent molecular mass of approximately 3536 kDa. Purity and stability of the substrate GDP4-keto-6-deoxy-D-mannose were determined by HPLC and electrospray-mass spectroscopic analysis, showing the absence of residual GDP-D-mannose and that the compound was stable for at least one month when stored at ÿ80  C. Total enzymatic activities (epimerase plus reductase) were analyzed for wild-type and mutant enzymes by monitoring absorbance at 340 nm to determine initial rates of NADPH oxidation, or by analysis of GDP-L-fucose production by HPLC. The kinetic constants for wild-type GMER enzymatic reaction and for the seven mutant enzymes are listed in Table 1. The catalytic activity was severely affected in all mutants, with the exception of Arg187Ala, whose kcat was lower by approximately twofold as compared to the wild-type enzyme. The Tyr136Glu mutant proved to be completely inactive; Lys140Arg exhibited a 20-fold reduction in kcat, while the other mutants featured an activity below 0.1 % of that of wildtype protein. A slight Km increase for both NADPH and GDP-4-keto-6-deoxy-D-mannose (between two- and threefold) was measured for Lys140Arg GMER, while a sixfold increase in Km for the substrate was observed for the Arg187Ala mutant (see Table 1). Conversely, it was not possible to measure Km values towards the substrate for the Lys140Ser and His179Asn mutants, since their rate versus substrate concentration plots exhibited a roughly biphasic rather than hyperbolic pro®le; this anomaly suggests the presence of different conformational forms of the mutant enzymes featuring different af®nities for GDP-4keto-6-deoxy-D-mannose. In order to follow the effects of the individual mutations on GMER epimerase activity, formation of GDP-4-keto-6-deoxy-L-galactose was monitored by means of TLC. Upon chemical reduction and hydrolytic removal of the dinucleotide moiety, GDP-4-keto-6-deoxy-D-mannose is expected to produce D-rhamnose and 6-deoxy-D-talose (see Figure 1), while the epimerization intermediate GDP-4-keto-6-deoxy-L-galactose yields L-fucose and 6-deoxy-L-glucose. When wild-type GMER was incubated in the absence of NADPH, after 80 GDP-L-fucose Biosynthesis Table 1. Kinetic parameters of wild-type and mutant GMER forms NADPH GMER form Wild-type Ser107Ala Cys109Ala Tyr136Glu Lys140 Ser Lys140Arg His179Asn Arg187Ala Vmax (mmol/hour per mg) kcat (s ) Km (mM) kcat/Km (sÿ1 mMÿ1) 738.9  16.3 0.808  0.019 0.137  0.007 ND 0.494  0.025 36.7  2.0 0.703  0.080 320  9.2 7.08  0.16 0.00774  0.00018 0.00131  0.00007 0.00473  0.00024 0.351  0.019 0.00673  0.00077 3.06  0.09 14.1  1.3 10.0  0.9 7.7  1.5 ND 18.5  3.3 42.8  4.6 11.1  2.3 9.1  1.1 502  58 0.774  0.088 0.170  0.042 0.256  0.059 8.20  1.32 0.606  0.194 336  50 ÿ1 GDP-4-keto-6-deoxy-mannose kcat/Km Km (mM) (sÿ1 mMÿ1) 38.6  4.4 15.3  0.3 26.2  1.6 ND 85.3  11.4 250  26 183  25 0.506  0.022 0.05  0.006 4.11  0.77 12.2  1.6 Data were obtained spectrophotometrically by measuring NADPH oxidation as the change in absorbance at 340 nm, at 25  C. ND, not detectable. NaBH4 reduction, equal amounts of both 6-deoxyglucose and fucose were formed, together with rhamnose and 6-deoxytalose (see Figure 3), indicating the partial formation of GDP-4-keto-6-deoxy-Lgalactose via the epimerase reaction, as observed previously (Menon et al., 1999). On the other hand, when NADPH was added to the reaction mixture, fucose was the only product observed, as expected. TLC analyses showed that the GMER mutants Lys140Ser, Cys109Ala, Tyr136Glu and His179Asn are not endowed with epimerase activity under the incubation mixture conditions (see Figure 3). A very low epimerase activity was detectable when enzyme concentrations higher than 200 mg/ml were employed. When the Lys140Arg mutant, which in the experimental conditions used displays also a residual reductase activity, was incubated in the presence of NADPH, the formation of both 6-deoxy-glucose and fucose was observed. However, in this case, the different yields in fucose and 6-deoxy-glucose, which are expected to be formed in equal amounts after chemical reduction of the Figure 3. TLC analysis of the products obtained after chemical reduction of the intermediate compounds GDP-4keto-6-deoxy-D-mannose and GDP-4-keto-6-deoxy-L-galactose. GMER 14C-labeled substrate (250 mM) was incubated at 37  C with 5 mg/ml wild-type and mutant enzymes, either in the presence or absence of 1 mM NADPH. At different time-points, aliquots were withdrawn and the enzymes were heat-inactivated. Samples were subjected to chemical reduction and hydrolysis as described in the text. The Figure represents samples after ®ve minutes of incubation. Standard sugars were co-chromatographed with samples and detected by a colorimetric method. Radioactive compounds were detected and quanti®ed by autoradiography using the Packard Cyclone Phosphor Storage system. (Data for GMER Cys109Ala mutant not shown). 81 GDP-L-fucose Biosynthesis 4-keto group, indicate that fucose derives partly from the enzyme reductase activity and partly from chemical reduction of GDP-4-keto-6-deoxy-Lgalactose, which accumulates during the course of the reaction. The Ser107Ala mutant featured a signi®cant epimerase activity, revealed by formation of both 6-deoxy-glucose and fucose, even when no residual reductase activity could be observed under the conditions used for incubation, where 5 mg/ml of protein was used (see Figure 3). It is interesting that epimerization was observed mainly when NADPH was added to the incubation mixture, suggesting that even if NADPH is not used for the reduction of the 4-keto group, it has an indirect effect on the ef®ciency of the epimerization reaction. When the samples were not treated with NaBH4, no formation of 6-deoxy-hexoses could be detected, con®rming the lack of reductase activity for the Ser107Ala mutant, under the adopted experimental conditions. Lastly, the Arg187Ala mutant displayed an epimerase activity comparable to that of the wild-type protein (data not shown). To address the role of the coenzyme in the epimerization reaction, we analyzed the GDP-4-keto6-deoxy-L-galactose formation rates for the wildtype enzyme and the Ser107Ala mutant, in the presence of both NADPH or NADP‡. As shown in Figure 4(a) and (b), the presence of the coenzyme affected the initial epimerization rates signi®cantly in both wild-type and mutant GMER. In particular, the lack of reductase activity for the Ser107Ala mutant allowed to demonstrate that the effect of NADPH on the epimerization reaction is more substantial than that of NADP‡. For wild-type GMER, the initial rates of GDP-4-keto-6-deoxy-L-galactose formation were 18.0 mmol/hour per mg and 127.5 mmol/hour per mg, for the apoenzyme or for GMER in the presence of NADP‡, respectively. For the Ser107Ala mutant, the rates were 7.2, 67.4 and 132 mmol/hour per mg for the apoenzyme, for GMER supplemented with NADP‡ and with NADPH, respectively. The equilibrium of the epimerization reaction between GDP-4-keto-6-deoxyD-mannose and GDP-4-keto-6-deoxy-L-galactose, determined for both the apoenzyme and for the enzyme supplemented with NADP‡, was approximately 50:50 for both wild-type and Ser107Ala mutant GMER. Wild-type GMER maintained its enzymatic activity for several weeks, when stored in phosphate-buffered saline, at 4  C, at concentrations higher than 10 mg/ml. In a similar way the Lys140Arg, Ser107Ala and Arg187Ala mutants displayed limited loss of their enzymatic activity, upon storage at 4  C for up to three weeks. On the contrary, the Lys140Ser and His179Asn mutants were highly unstable and became almost completely inactive in less than one week. Stability of the Tyr146Glu and Cys109Ala mutants could not be determined because their initial activity was too low. Figure 4. (a) TLC analysis of samples obtained after incubation of wild-type and Ser107Ala GMER for 15 minutes at 37  C with 250 mM 14C-labeled GDP-4-keto-6deoxy-D-mannose, either alone or in the presence of 1 mM NADP‡ or NADPH. Experimental conditions were as described for Figure 3. (b) Effects of NADP‡ and NADPH on the rate of the epimerization reaction for wild-type and the Ser107Ala mutant. GDP-4-keto-6deoxy-L-galactose production in the incubation mixture was derived by fucose and 6-deoxy-glucose formation determined by TLC analysis, as described above. Crystal structures of wild-type GMER and of its mutants at atomic resolution Table 2 lists the X-ray data collection (in the Ê resolution range) and the ®nal crystal1.45-1.60 A lographic re®nement statistics (R-factors vary between 12.6 % and 13.9 %) for the three holo- 82 GDP-L-fucose Biosynthesis Table 2. X-ray data collection and re®nement statistics Completeness (%) Ê) l (A Unique reflections Redundancy Rmerge (%) I/s(I) overall I/s(I) outer shell Resolution range used in Ê) refinement (A Total number of non-hydrogen protein atoms Number of water molecules Number of atoms in non-water solvent peaks Number of coenzyme atoms R-factor R-free rmsd from ideal geometry Ê) Bond lengths (A Ê) Bond angles (A Ê) Planes (1-4) (A Ê 2) Averaged B-factors (A Main-chain Side-chain Water molecules Coenzyme Ê) Cruickshank DPI (A Wild-type GMER Ser107Ala Tyr136Glu Lys140Arg 99.1 0.855 80,994 4.5 5.7 14.8 Ê) 5.5 (1.48-1.45 A 10.0-1.45 99.1 0.844 60,724 9.0 4.0 20.5 Ê) 2.9 (1.63-1.60 A 10-1.60 99.3 0.844 60,793 6.2 4.3 9.7 Ê) 2.1 (1.63-1.60 A 10-1.60 98.8 0.902 73,654 8.4 16.2 8.64 Ê) 2.45 (1.53-1.50 A 12-1.50 2502 2497 2497 2497 379 31 325 36 352 34 425 30 48 0.127 0.167 48 0.138 0.182 48 0.139 0.180 48 0.126 0.162 0.015 0.032 0.032 0.016 0.035 0.037 0.017 0.038 0.031 0.014 0.029 0.028 23.4 28.5 52.7 41.9 0.047 25.5 30.9 48.4 45.7 0.068 26.5 30.7 51.7 43.1 0.069 21.9 26.4 45.0 38.2 0.052 GMER mutants analyzed, including information on the reference GMER wild-type structure, re®ned Ê resolution (Rosano et al., 2000). Inspection at 1.45 A of the re®ned models shows that all residues (for each protein structure) lie within the allowed Ramachandran plot regions, with close to ideal values for the stereochemical parameters, as analyzed by PROCHECK (Engh & Huber, 1991; Wallace et al., 1995). Continuous and clearly interpretable electron density is available for all mutant proteins in the Lys3-Gln316 protein region. Residue Ala2 could be modeled in the electron density of the wild-type native protein only, while residue 118 is always found as cis-proline. The NADP‡ coenzyme electron density is well de®ned in all the structures analyzed, with the exception of the nicotinamide ring, whose carboxamide substituent can be properly located only in the Tyr136Glu and Lys140Arg mutants. Moreover, the Lys140Arg mutant structure shows a substantial conformational readjustment (a shift of approximately Ê ) in the ribose-nicotinamide end of the coen10 A zyme. Clear electron density is present for more than 320 water molecules, for Tris molecules and for sulfate ions in each of the re®ned structures (see Table 2). In all re®ned structures, electron density accounting for a small unidenti®ed molecule is present next to the protein regions Glu67-Ala71 and Ser176-Ser 178. The estimated atomic coordinate error (Cruickshank DPI; Murshudov et al., Ê for all the re®ned structures. 1997) is <0.1 A Inspection of the re®ned model and electron density shows that the Ser107Ala mutation has minor effects on the structure of its protein sur- roundings. Signi®cant side-chain conformational changes are observed at Cys109, Ile110 and Met162 (Met162 is spatially close to Ile110); in particular, the modi®ed side-chain conformation of Cys109 may be responsible for the modest solvent-accessiÊ 2). The backbone conforbility of Ala107 (11 A mation in the 108-111 segment is partly affected. Minor side-chain readjustments are present also at Asn133 and His179, while the active site-solvent structure is mostly coincident with that observed in the wild-type holoenzyme. Although the Ser107Ala mutation affects one of the three recognized GMER catalytic residues, neither Tyr136 nor Lys140 side-chain conformation is perturbed. Similarly, the NADP‡ coenzyme molecule displays a conformation that is coincident with that of the wild-type enzyme, although a precise orientation of the nicotinamide ring cannot be assigned due to poor electron density for this part of the coenzyme. The Tyr136Glu mutant structure shows clearly de®ned electron density for the mutated residue, fully compatible with a Glu side-chain whose Ca and Cb atoms are essentially coincident with those of Tyr136 in the wild-type GMER structure (see Figure 5(a)). On the other hand, the glutamate carboxylate group is shifted towards the solvent space and held in its positions by hydrogen bonds to Ê ), and potentially to Asn72 Asn133 ND2 (3.30 A Ê ) and Cys109 SG (3.63 A Ê ). Moreover, OD1 (3.53 A Glu136 OE2 is hydrogen bonded through bridging water molecules to the carbonyl groups of Gly67 (through Wat36) and Lys65 (through Wat230). The Tyr136Glu substitution affects signi®cantly the side-chain conformation of residues Val66, Cys109, GDP-L-fucose Biosynthesis Lys140 and His179, all surrounding the nicotinamide end of the bound coenzyme. The protein backbone structure is shifted in the Phe104-Leu110 Ê with respect segment, with a deviation of ca 1 A to the wild-type enzyme, at residue Cys109. 83 Clear conformational readjustment is observed at C-2 and C-3 of the N-ribose (N- identi®es the nicotinamide-end of NADP‡), which moves to ®ll a cavity left by the Tyr136Glu substitution. As a result, Lys140 NZ can hydrogen bond to the Figure 5. (a) Stereo view of the Tyr136Glu mutant active-site region including the mutated residue and part of the neighboring amino acid residues involved in the GMER catalytic mechanism. For reference, the NADP‡ molecule (blue) fully de®ned in the mutant electron density is displayed overlaid upon its own structure in the wild-type enzyme (yellow). Helix a-E, in the mutant structure, is displayed as a blue ribbon. (b) Stereo view of the Lys140Arg mutant active-site region in an orientation comparable to that of (a), with the same colour conventions. The space®lling nucleotide-sugar molecule highlights the expected substrate-binding mode, as inferred from homology modeling based on the structure of the UGE:UDP-glucose:NADH complex (Thoden et al., 1997; Rizzi et al., 1998; Somers et al., 1998). This Figure was drawn with MOLSCRIPT (Kraulis, 1991). 84 Ê ) and O-30 atoms (3.03 A Ê ), N-ribose O-20 (2.84 A while Glu136, which is solvent-accessible only for Ê 2, is hydrogen bonded to the N-ribose O-20 14 A atom via a bridging water molecule (Wat129). The nicotinamide ring is well de®ned in the electron density, and displays an intramolecular hydrogen bond between the carboxamide N atom and the bridging b-phosphate group. Moreover, the nicotinamide carboxamide O atom is hydrogen bonded Ê ). The to the Leu166 peptidic N atom (3.25 A NADP‡ binding region is characterized by substantial redistribution of the ordered water molecules as compared to wild-type GMER structure. The third mutant structure analyzed, bearing the substitution of the catalytic residue Lys140 with an arginyl side-chain, shows substantial perturbation of the active-site region, despite the conservation of a positively charged residue at position 140 (see Figure 5(b)). In fact, the presence of the bulky guanidino group of Arg140, which would potentially collide with the N-ribose ring, forces the ribosenicotinamide segment of NADP‡ to rotate by about 120  around the ribose-phosphate bonds, in the direction of the solvent. The mutated Arg140 residue, which matches the side-chain conformation of Lys140 in the wild-type enzyme, is Ê ) and, hydrogen bonded to Tyr136 OH (3.01 A through four active-site water molecules, to Leu61, Ala63, Glu87 and to a phosphate group of the coenzyme. Side-chain readjustments can be found at residues Leu105, Cys109, Met162 and His179, the latter being hydrogen bonded to the nicotinic Ê ). As observed for the carboxamide O atom (3.01 A other mutants, the protein main chain displays a signi®cant deviation from the wild-type protein structure in the Phe104-Ile110 segment (about Ê at the Ca atom of Gly106). Moreover, the 1.9 A ordered water molecules surrounding the coenzyme are structured quite differently as compared to the other GMER structures examined, as a result of the internal cavity left by the shift of the NADP‡ ribose-nicotinamide segment. As listed in Table 2, one Tris molecule and three sulfate ions were located on the molecular surface of the re®ned structures. In particular, the Tris molecule is located next to residues Arg21, His170 and Trp311, in an inter-domain location. Next, one of the sulfate anions, common to all the re®ned structures, is electrostatically linked to His11, Arg12 (to which it is also hydrogen bonded) and to Arg20. The constant presence of the anion may prevent direct interaction of Arg12 with the NADP‡ ribose Ê away from the Arg12 phosphate, which is kept 7 A guanidino group. A distinct case concerns a small molecular compound, of unknown origin, constantly observed between the GMER N and Cterminal domains in the region between residues Gly67-Ala71 and Ser176-Ser178. In all the re®ned structures, the electron density accounting for such a ligand displays a constant shape, reminiscent of an ethyl phosphate molecule. One end of the liganded compound, displaying tetrahedral atomic structure, provides hydrogen bonding to residue GDP-L-fucose Biosynthesis Lys262 of a symmetry-related GMER molecule. In the Lys140Arg mutant, which shows a modi®ed NADP‡ conformation in this region, the putative ethyl phosphate molecule can still be clearly recognized, but rotated by about 180  with respect to the other re®ned structures, in interaction with the nicotinamide carboxamide N atom. Related to such modi®ed binding mode, shifting the location of the putative phosphate group, Lys262 of the symmetry-related GMER molecule adopts two alternative conformations. Discussion The wealth of structural and mutational data available on members of the SDR homology family (Gosh et al., 1994, 1995; Thoden et al., 1996a,b,c, 1997; Breton et al., 1996; Tanaka et al., 1996a,b; Benach et al., 1998), has indicated that the GMER catalytic mechanism is based on the concerted action of residues Ser107, Tyr136 and Lys140. These residues are spatially close, and fall next to the NADP‡ N-ribose, to the nicotinamide ring and to the expected 4-ketopyranose substrate binding site (Rizzi et al., 1998; Somers et al., 1998). In wildtype GMER, Tyr136 OH atom is hydrogen bonded Ê ), Lys140 to the 30 OH atom of the N-ribose (2.76 A Ê) NZ is hydrogen bonded to both 20 OH (2.93 A 0 Ê and 3 OH (2.96 A) atoms, but the two catalytic residues are not mutually hydrogen bonded (the Ê ). More136 OH to 140 NZ distance being 4.28 A over, the Tyr136 OH and Lys140 NZ side-chain atoms cannot hydrogen bond to Ser107, the third Ê catalytic residue, whose OG atom is about 4.38 A Ê away, respectively (see Figures 2 and and 6.97 A 6). A very similar structural organization is conserved in the active site of homologous enzymes active on related substrates, including GMD, the enzyme preceding GMER in the GDP-L-fucose biosynthetic pathway, in UGE, but also in more distantly related SDRs active on substrates other than nucleotide-sugars (Varughese et al., 1994; Tanaka et al., 1996a,b; Gosh et al., 1995; Benach et al., 1998; Somoza et al., 2000; Thoden et al., 1996a,b). The epimerization reaction catalyzed by GMER occurs at the C-3 and C-5 centers of the 4-ketopyranose substrate (see Figure 1). From a mechanistic viewpoint, the active-site structure suggests that, due to the role played by the Lys140 positive charge (lowering the pKa of Tyr136 to ca 6.1), Tyr136 OH may act as a general acid/base with respect to the substrate 4-keto center, during catalysis (Jornvall et al., 1995; Liu et al., 1997). Tyr136 is held to donate/accept a proton to/from the substrate C-4 oxygen atom, promoting the transition between keto/enolic forms at this center (and therefore at the adjacent C-3 or C-5), required for the two epimerization reactions. In this key action, Tyr136 is likely assisted by the hydrogen bonding capabilities of neighboring residues Ser107, Ser108 and Cys109. Moreover, Tyr136 can assist the reduction step of the epimerized intermediate, GDP-L-fucose Biosynthesis GDP-4-keto-6-deoxy-L-galactose, by donating a proton to the 4-keto group, concerted with the stereospeci®c hydride transfer from NADPH to the pyranose C-4 center (Burke & Frey, 1993; Menon et al., 1999). In this mechanistic context, the catalytic center Ser107Ala mutation, re¯ected by minor structural perturbation of the wild-type GMER structure, is found to decrease the kcat value for the overall enzymatic reaction by a factor of 1000, with very moderate decrease in Km values for NADPH and for the substrate. Thus, whereas Ser107 appears to be a key residue for catalysis, it plays a secondary role in coenzyme and substrate binding (see Table 1). Hydrogen bonding between Ser107 and the coenzyme nicotinamide ring is structurally possible in wild-type GMER (see Figure 5(a)), but cannot be ®rmly established due to conformational disorder displayed by the nicotinamide ring in the crystal structure. In the homologous enzyme UGE, Ser124, the Ser107 structurally equivalent residue, has been shown to be involved in hydrogen bonding to the nicotinamide carboxamide group, in an NAD(H) redox-dependent manner (Thoden et al., 1997). Mutation of UGE Ser124 to Thr maintains an active enzyme, whereas virtually inactive forms are obtained for the Ser124Ala or Ser124Val residue substitutions, which impair proton exchange capabilities (Liu et al., 1997). In a similar way the Thr133Val mutation, in the structurally equivalent site of GMD, reduces the kcat value by 3000-fold, leaving the Km values for NADP‡ and for the substrate almost unaltered (Somoza et al., 2000). All these data are consistent with two complementary roles for Ser107 in the GMER catalytic mechanism. On one hand, Ser107 may assist proton exchange between Tyr136 and the substrate, likely supported by Ser108 or Cys109. On the other, Ser107 may play a structural role, stabilizing the nicotinamide orientation required for the NADPH to substrate stereospeci®c hydride transfer, in the reduction step. In this respect, a Ser107 to nicotinamide hydrogen bond may be favored in a lower polarity environment, when both NADPH and the substrate are bound. In dihydropteridine reductase, the presence of Ala135 (Varughese et al., 1994), instead of Ser at this site, is compensated by a trapped water molecule, which may provide the required proton exchange capabilities. However, no trapped water molecule is present in the GMER Ser107Ala mutant (holoenzyme) at a structural location compatible with the role played by the Ser107 OG atom. The Ser107Ala mutant is nevertheless endowed with epimerase activity (both in the presence of oxidized or reduced coenzyme; see Figures 3 and 4). Such an observation would suggest that stabilization of the enediol intermediates required to achieve epimerization at the C-3 and C-5 substrate centers does not require a protein OH group at this site, in keeping with the minor effects that the Ser107Ala mutation has on substrate recognition. In this respect, it should be considered that the 85 nearby residues Ser108 and Cys109, both spatially close to the proposed 4-ketopyranose binding site, are also valid candidates for proton shuttling between the enzyme and the substrate, and may provide stabilization of the substrate or reaction intermediate(s) during the catalytic cycle. Residues Ser107, Ser108 and Cys109 are at the base of a round protein surface pocket, whose outer rim is essentially de®ned by residues Asn72, Lys113, Glu130, Asn133, His179 and Ly283, having the nicotinamide moiety at the ¯oor (see Figures 5 and 6). Modeling indicates that the substrate 4-ketopyranose ring may be located in the pocket, with C-3, C-4 and C-5 centers facing Ser107, Ser108, Cys109 and His179 residues (Rizzi et al., 1998). Although Ser108 and Cys109 are not strongly conserved in related SDR enzyme sequences, they are present in human erythocyte enzyme FX, which also displays a two-center epimerase activity, and align with Thr115 and Cys116 in the murine transplantation antigen P35B (Tonetti et al., 1996; Szikora et al., 1990). Moreover, residue Glu135 of GMD, which is structurally equivalent to Cys109 of GMER, has been proposed to act as a base during the GDPmannose 4,6-dehydratase reaction (Somoza et al., 2000). Impairment of the catalytic activity in the Cys109Ala GMER mutant, which however maintains virtually unaltered Km values for NADPH and for the substrate, is in keeping with the proton exchange role for Cys109, proposed above. Moreover, the lack of epimerase activity shown by the Cys109Ala mutant, as opposed to conservation of the epimerase activity displayed by the Ser107Ala mutant GMER, supports such a proton shuttle role for Cys109 as opposed to a nicotinamide-orienting role (during NADPH reduction of the epimerized intermediate GDP-4-keto-6-deoxy-L-galactose) for Ser107. In particular, we note that Cys109 may display a lowered pKa value, resulting from a strong Cys109 SG - - - Ser107 OG hydrogen bond, constantly present in the pertinent GMER wildtype or mutant structures analyzed here (see Figures 5 and 6). Related to the discussion on residues assisting catalysis, and in agreement with the known lability of the C-3 proton during epimerization, a complementary general acid/base role may be played by residue His179, which, in modeling studies, falls next to the C-2 and C-3 centers of the 4-ketopyranose substrate ring (see Figure 6; Chang et al., 1988; Rizzi et al., 1998; Somers et al., 1998; Somoza et al., 2000). Such a role is supported by the present analysis, which shows a 1000-fold loss of total activity for the His179Asn mutant, with very modest decrease in Km for NADPH. Moreover, mutation of His179 impairs the GMER epimerase activity, suggesting that a multi-residue proton donor/acceptor array is required in the strict neighborhood of the pyranose ring to promote the epimerization reaction and/or stabilize the different keto-enolic intermediate forms. In fact, mutations of residue Lys140 (into Arg or Ser), 86 GDP-L-fucose Biosynthesis Figure 6. Stereo view of the proposed binding mode for a nucleotide-sugar molecule (shown as a space-®lling model) relative to active-site residues discussed in the text. The C-2 and C-3 centers can be recognized as those closest to His179 side-chain; the C-4 center falls next to the nicotinamide carboxamido group. which is structurally farther away and unlikely to interact directly with the substrate pyranose ring, do not affect the epimerase activity signi®cantly (see Figure 6). The Tyr136Glu mutation has evident structural effects, which are localized within the GMER active site despite the unusual residue substitution and the buried location of the affected site (see Figure 5(a)). Accommodation of a negatively charged Glu residue at site 136 is in keeping with the proposed ionization state of Tyr136, and with its electrostatic coupling to Lys140 in the wild-type enzyme. Therefore, the structural perturbations observed in the Tyr136Glu mutant can be mainly ascribed to the steric and hydrogen bonding effects exerted by the mutated residue on its surroundings. In this respect, it should be noted that, unlike residue Tyr136, Glu136 is connected to the N-ribose only through a bridging water molecule, loosing an SDR conserved protein-coenzyme interaction. This fact is re¯ected by different sugar puckering modes of the N-ribose, which is present as C-20 -endo, C-30 -exo in the mutant (versus C-20 -exo, C-30 -endo in wild-type GMER). Nevertheless, the catalytic residue Lys140 maintains the same pair of hydrogen bonds (to N-ribose 20 OH and 30 OH groups) observed in the wild-type GMER (see Figure 5(a)). The total loss of activity in Tyr136Glu GMER can be related to the following considerations (see Figures 5 and 6). The structural environment in the 4-ketopyranose ring binding site is perturbed in such a way that the general acid/base residue at site 136 cannot be properly positioned for proton transfer to the C-4 oxygen atom during the different steps of the catalytic cycle. Moreover, Glu136 is expected to display a pKa value lower by at least two pH units as compared to that of Tyr136, possibly too acid to provide equally ef®cient proton donor and acceptor capabilities during catalysis. Several different mutations of the catalytic center Tyr residue, the only truly invariant residue in the whole SDR family, failed to provide signi®cantly active enzymes in UGE, Drosophila alcohol dehydrogenase and in dihydropteridine reductase, stressing the requirement for a properly positioned general acid/base group within hydrogen bonding distance from the substrate reactive center (Chen et al., 1993; Cols et al., 1993; Varughese et al., 1994; Liu et al., 1997). The two mutations engineered at site 140 (Lys140Arg and Lys140Ser) affect quite differently the GMER catalytic activity (with 20-fold and 1500fold drops, respectively). Inspection of the crystal structure shows that in the Lys140Arg mutant these effects may be related to the increased size of residue Arg140, which perturbs the active-site structure, particularly at the NADP‡ ribose-nicotinamide end, resulting in a direct hydrogen bonding interaction between Arg140 and Tyr136 (see Figure 5(b)). Perturbation of the coenzyme binding mode is re¯ected by the Km value for NADPH in the Lys140Arg mutant, as opposed to that of the Lys140Ser mutant (Table 1), where the smaller 87 GDP-L-fucose Biosynthesis Ser140 residue is unlikely to alter the coenzyme binding mode. On the other hand, the Arg140 positive charge supports maintenance of the electrostatic coupling with Tyr136, and should keep the Tyr136 pKa value within the useful range for catalysis as a general acid/base residue. In fact, inspection of the kcat values given in Table 1 shows that the residual activity of the Lys140Arg mutant is signi®cant and 75-fold higher than in the Lys140Ser mutant, despite some loss of substrate af®nity. Lastly, residue Arg187 is located at the upper part of the proposed substrate-binding cleft (see Figures 2 and 6), next to the guanine-ring site suggested by modeling. Although some conformational readjustment in this region may be required to improve the enzyme-substrate ®t, mutation of residue Arg187 to Ala increases the substrate Km by approximately sixfold, suggesting the net contribution of this residue to substrate recognition, presumably through hydrogen bonding to guanine hetero-atoms (Rizzi et al., 1998). On the other hand, in accordance with the substantial distance of Arg187 from the coenzyme binding site Ê ), no effect on the Km value for NADPH is (>19 A observed. Sequence alignment of GMER and UGE shows that the Arg187 equivalent residue in UGE is Val, a residue variation that may be related to the different substrate speci®cities displayed by the two homologous enzymes (whose sites recognize guanine and uridine bases, respectively), but also to other residue substitutions in this area. On the other hand, residue Arg209, which is located in the central part of the proposed GMER substrate cleft, is conserved in UGE, where it plays an electrostatic and hydrogen bonding role in recognition of the substrate pyrophosphate bridge. The functional data here presented allow us to make some distinction between the epimerase and reductase activities in GMER. The structural data indicate that the two enzymatic reactions may be based on a common subset of the active-site residues, centered around Tyr136. Maintenance of residual epimerase activity in wild-type GMER (even in the absence of NADP‡) is in keeping with the essentially conserved active-site structure and location of residues Cys109, Tyr136, His179 and Lys140, observed in the crystal structure of the apoenzyme (Rizzi et al., 1998). On the other hand, a fully structured active site with ideal dielectric environment, linked to the presence of bound NADPH (and substrate), is required for the achievement of the full epimerase activity, as compared to apo-GMER. In fact, our data show that additional epimerase ef®ciency, relative to the apoenzyme, can be gained also in the presence of NADP‡, i.e. in a non-reducing but properly structured holo-GMER active center. Moreover, the epimerase reaction ef®ciency in the Ser107Ala mutant is dependent on the presence of either NADP‡ or NADPH (see Figure 4(b)). Such observations may be related to different active-site structuring, electrostatic charges, conformations or mobilities achieved by the oxidized versus reduced nicotin- amide ring, which would support increasingly productive binding modes for the substrate (or catalytic intermediates) pyranose ring relative to residues Cys109, Tyr136 and His179. Remarkably, no NADP‡ effect on the epimerization reaction catalyzed by GDP-fucose synthetase (an alternative name for GMER) was reported by Menon et al. in a recent characterization of the wild-type enzyme (Menon et al., 1999). As a whole, the mutational studies presented here show that the active tertiary and quaternary structures of GMER can support a wide range of amino acid substitutions, while preserving rather strictly the enzyme's wild-type fold, a prerequisite for site-directed mutagenesis characterization of the catalytic mechanism. In this respect, we note that decreased stability has been observed for only two GMER mutants, which affect the active site electrostatics. In fact, unlike the Tyr136Glu mutant, which maintains the Tyr136 negative charge, both the meta-stable Lys140Ser and His179Asn mutants remove a positive charge from the active site, with possible effects on the local enzyme structure, as re¯ected by failure to grow crystals of the latter mutants under the common physicochemical conditions employed in this study. In view of the key role played in the de novo biosynthesis of GDP-L-fucose and, as consequence, of fucosylated glycoconjugates, GMER is a potential target for the therapeutical treatment of pathological conditions arising from abnormal selectinmediated cell-to-cell interaction processes. These include acute and chronic in¯ammation, graft rejection and metastatic states. The elucidation of GMER catalytic mechanism, together with identi®cation of the residues involved in substrate recognition and enzymatic activity, is the basis for the rational development of inhibitory molecules to be used in the development of leads. Materials and Methods Cloning of bacterial GMER Wild-type GMER was overexpressed in E. coli as described by Tonetti et al. (1998a). Gene-speci®c primers (TibMolBiol, Genova, Italy), containing EcoRI and XhoI restriction sites in sense and antisense primers, respectively, were used to amplify the GMER gene, using puri®ed E. coli K12 genomic DNA as template. The PCR fragment was puri®ed by agarose gel, digested with EcoRI and XhoI, and ligated in pGEX-6-P1 vector (Amersham-Pharmacia Biotech, Milan, Italy). The construct obtained was used to transform E. coli strain JM109 and to express the recombinant GST-fusion protein as described by Sturla et al. (1997). Site-directed mutagenesis Site-directed mutagenesis was performed by the unique site elimination technique (Deng & Nikoloff, 1992), using the U.S.E. Mutagenesis Kit from AmershamPharmacia Biotech and following the manufacturer's instructions. Mutagenesis was performed directly on pGEX-6P-1 vector containing the full-length coding 88 sequence of GMER (Tonetti et al., 1998a), annealed with the selection primer, which substitutes the unique restriction site in the plasmid DNA, and with the target mutagenic primer, which introduces the desired mutations into the GMER DNA. The selection primer was pGEX U.S.E. Selection Primer (PstI to SacII) and was obtained from Amersham-Pharmacia Biotech. Target mutagenic primers, obtained from TibMolBiol, were FPLC-puri®ed and were 30 to 32 bases in length. Plasmids containing the mutated restriction site (SacII instead of PstI) were analyzed by DNA sequencing to verify the presence of the desired mutations in the GMER coding region. DNA sequencing was performed by M-Medical (Florence, Italy). The vectors containing the correct mutation were used to trasform E. coli JM109 strain; the colonies that produced the greatest amount of protein were used to express the mutant enzymes as inducible GST-fusion proteins. Protein expression was performed as described (Sturla et al., 1997), except that the bacterial cells were grown at 22  C, both before and after induction with 0.1 mM IPTG. Wild-type and mutant GMERs were puri®ed to homogeneity by af®nity chromatography as described by Tonetti et al. (1998a), and the GST-tag was removed using Pre-Scission Protease (AmershamPharmacia). Proteins were concentrated using YM10 ultra®ltration membrane (Amicon, Millipore, Milan, Italy) to a ®nal concentration above 10 mg/ml. The recombinant proteins were analyzed by SDS-PAGE (Laemmli, 1970), followed by Coomassie blue staining to check their purity. Synthesis of GDP-4-keto-6-deoxy-D-mannose Unless otherwise speci®ed, all reagents were obtained from Sigma (St. Louis, MO, USA). Synthesis of GDP-4keto-6-deoxy-D-mannose was performed from GDP-Dmannose using recombinant human GDP-D-mannose 4,6 dehydratase (GMD) (Bisso et al., 1999). Reaction mixtures containing 50 mM Tris-HCl (pH 7.0), 150 mM NaCl, 1 mM DTT, 1 mM EDTA, 2 mg/ml bovine serum albumin, 150 mg/ml GMD, and 1 mM GDP-D-mannose, or GDP-[U-14C]-D-mannose (Amersham-Pharmacia Biotech, speci®c activity 10.6 GBq/mmol) were incubated for 45 minutes at 37  C. Samples were heat-denatured at 100  C for one minute and centrifuged to remove precipitated proteins. Reaction products were analyzed by HPLC using a C18 (Bondapack column (Waters, Milford, MA; 3.9 mm  300 mm, 10 mm particle size), as described by Tonetti et al. (1996). Electrospray mass spectrometry was performed by direct ¯ow injection of the sample through a Valco valve into the atmospheric pressure ionization electrospray ion source of the mass spectrometer (5989A single quadrupole Hewlett-Packard Engine). The spectra were performed in the negative ion mode in a range including the expected molecular mass. The mixture used as eluent was water/methanol/tri¯uoroacetic acid (TFA), 49.5:49.5:1 (by vol.), the drying gas was nitrogen and the capillary exit voltage was set at ÿ200 V. GDP-4keto-6-deoxy-D-mannose was stored at ÿ80  C in small aliquots. Enzymatic assays Spectrophotometric assays of the reductase activity were performed using a Beckman DU 640 spectrophotometer, equipped with a cell-holder maintained at 25  C, by monitoring NADPH disappearance as a change in absorbance at 340 nm. Reaction mixtures contained GDP-L-fucose Biosynthesis 50 mM Tris-HCl (pH 7.0), 150 mM NaCl and 2.5 mM MgCl2, with different concentrations of NADPH (ranging from 2.5 to 250 mM) and of GDP-4-keto-6-deoxy-D-mannose (ranging from 10 to 750 mM), in a total volume of 350 ml. Reactions were initiated by addition of GMER and initial rates were recorded. The kinetic parameters for wild-type and mutant enzymes were determined by using at least six substrate concentrations, for both NADPH and GDP-4-keto-6-deoxy-D-mannose. Km and Vmax values were determined using a non-linear regression method. To con®rm the speci®city of the reaction for both wild-type and mutant enzymes, the identity of the product GDP-L-fucose was con®rmed by HPLC and TLC analyses. To evaluate the epimerase activity, 14 C-labeled GDP-4-keto-6-deoxy-D-mannose (see above) at saturating concentration (250 mM), was incubated at 37  C with wild-type or mutant GMER, in the presence of 2.5 mM MgCl2, either with or without 1 mM NADP(H). Incubation mixtures were then reacted with 1 mg/ml NaBH4, for one hour at room temperature, and acid-hydrolyzed by addition of 5 % (v/v) TFA (incubation for 15 minutes at 100  C). Samples were vacuumdried, suspended in 1 ml of water and desalted using 0.5 g of Amberlite MB-150 (Sigma) for each milliliter of incubation. Samples were vacuum-dried again and TLC analysis was performed on Silica-gel 60 TLC plates (Merck, Milan, Italy), pre-treated as described (Tonetti et al. 1996). TLC plates were developed four times using an acetonitrile/water (95:5, v/v) mixture. Standard unlabeled sugars were chromatographed together with samples and detected with a diphenylamine/aniline/ phosphoric acid reagent (Chaplin, 1994). The radiolabeled compounds were detected and quanti®ed by autoradiography using the Cyclone System (Packard, Milan, Italy). Crystallographic analyses Crystals of native and mutant GMER forms were grown as described by Tonetti et al. (1998b), with slight adjustments according to the mutant species considered. In particular, protein solutions ranging from 11 to 20 mg/ml were equilibrated against 1.5 M lithium sulfate, 0.1 M Mes or Tris buffers (pH 6.5-7.8) at 21  C, yielding trigonal bipyramidal crystals (ca 0.25 mm per edge) within one week. All crystals grown for GMER mutant forms proved isomorphous with the wild-type protein crystals (trigonal space group P3221, Ê , c ˆ 74.9 A Ê , g ˆ 120  , one GMER mola ˆ b ˆ 103.0 A ecule per asymmetric unit). High-resolution X-ray diffraction data for the wild-type protein were collected using synchrotron radiation (ELETTRA, beam line Ê , 100 K). X-ray diffracXRD1, Trieste, Italy; l ˆ 0.855 A tion data for the GMER Tyr136Glu and Ser107Ala mutants were collected at the EMBL/DESY synchrotron source (beam line BW7A, Hamburg, Germany; Ê , 100 K). The Lys140Arg mutant data were l ˆ 0.844 A collected at ESRF (beam line ID14-1, Grenoble, France; Ê , 100 K). A summary of the X-ray data collecl ˆ 0.902 A tion statistics is provided in Table 2. Holo-GMER (wildtype and mutant) crystals were prepared from the apoenzyme crystals by soaking in NADP‡-containing solutions. For the purpose of cryoprotection, the mother liquor solution was supplemented with 20 % (v/v) glycerol. Diffracted intensities were integrated and scaled, in all cases, using the HKL program suite (Otwinowski et al., 1997). Crystal structure analyses were carried out starting Ê wild-type GMER with phases calculated from the 2.1 A GDP-L-fucose Biosynthesis structure previously determined (PDB code 1bws; Rizzi et al., 1998), omitting the NADP‡ molecule, solvent and the relevant mutated side-chain. Crystallographic re®nement of the GMER native and mutant structures was performed using the program REFMAC (CCP4 suite, Murshudov et al., 1997). Rigid-body re®nement of the 1bws model (determined at 300 K) was preliminarily Ê resolution range. Subperformed in the 30.0-3.5 A sequently, alternate cycles of positional re®nement and model inspection/rebuilding were performed, based on the O program suite (Jones et al., 1991). When re®nement of the protein component was at convergence, the mutated side-chains and the NADP‡ molecule were modeled in the respective electron density peaks, followed by additional individual B-factor and atomic coordinate re®nement cycles, with water and solvent molecule location, until convergence was achieved at the maximum resolution. Eventually, individual anisotropic B-factors were re®ned for all NADP‡ complexes. Final statistics for all the re®ned structures are reported in Ê resolution Table 2. A preliminary account on the 1.45 A structure of wild-type GMER has been reported (Rosano et al., 2000). Protein Data Bank accession codes Atomic coordinates and structure factors for the Ê resolution wild-type holo-GMER structure and 1.45 A for the Ser107Ala, Tyr136Glu and Lys140Arg mutants have been deposited with the RCSB Protein Data Bank, with accession codes 1E6U, 1E7Q, 1E7R, 1E7S, respectively Berman et al. (2000). Acknowledgments This work was supported by grants from the Italian Ministry for University, Scienti®c and Technological Research (``Structural bases and functional consequences of cell-surface recognition processes'' and PRIN 98 ``Molecular mechanism of intercellular communication'') and CNR (Target Project ``Biotechnology'', to M.B. and M.T.). We are grateful to Professor Menico Rizzi (Universita' del Piemonte Orientale) for helpful comments and to Dr Gianluca Damonte (DIMES, University of Genova) for mass spectroscopy analysis. Finally, support from the synchrotron radiation facilities (ELETTRA-Trieste, ESRFGrenoble and EMBL/DESY-Hamburg) is acknowledged. References Appelmelk, B. J., Negrini, R., Moran, A. P. & Kuipers, E. J. (1997). Molecular mimicry between Helicobacter pylori and the host. Trends Microbiol. 5, 70-73. Benach, J., Atrian, S., Gonzalez-Duarte, R. & Ladestein, R. (1998). The re®ned crystal structure of Drosophilia Ê resollebanonensis alcohol dehydrogenase at 1.9 A ution. J. Mol. Biol. 282, 383-399. Berman, H. M., Westbrook, J., Feng, Z., Gilliland, G., Bhat, T. N. & Weisseg, M. (2000). The Protein Data Bank. Nucl. Acids Res. 28, 235-242. Bisso, A., Sturla, L., Zanardi, D., De Flora, A. & Tonetti, M. (1999). Structural and enzymatic characterization of human recombinant GDP-D-mannose-4,6-dehydratase. FEBS Letters, 456, 370-374. 89 Branden, C. & Tooze, J. (1991). Alpha/Beta structures. In Introduction to Protein Structure, 2nd edit., pp. 47-66, Garland Publishing Inc., New York. Brandley, B. K., Swiedler, S. J. & Robbins, P. W. (1990). Carbohydrate ligands of the LEC cell adhesion molecules. Cell, 63, 861-863. Breton, R., Housset, D., Mazza, C. & Fontecilla-Camps, J. C. (1996). The structure of a complex of human 17b-hydroxysteroid dehydrogenase with estradiol and NADP‡ identi®es two principal targets for the design of inhibitors. Structure, 4, 905-915. Burke, J. R. & Frey, P. A. (1993). The importance of binding energy in catalysis of hydride transfer by UDP-galactose 4-epimerase: a 13C and 15N NMR and kinetic study. Biochemistry, 32, 13220-13230. Chang, S., Duerr, B. & Serif, G. (1988). An epimerasereductase in L-fucose synthesis. J. Biol. Chem. 263, 1693-1697. Chaplin, M. F. (1994). Monosaccharides. In Carbohydrate Analysis: A Pratical Approach, pp. 1-42, IRL Press, Oxford. Chen, Z., Jiang, J. C., Lin, Z.-G., Lee, W. R., Naker, M. E. & Chang, S. H. (1993). Site-speci®c mutagenesis of Drosophila alcohol dehydrogenase: evidence for involvement of tyrosine-152 and lysine-156 in catalysis. Biochemistry, 32, 3342-3346. Cols, N., Marfany, G., Atrian, S. & Gonzalez-Duarte, R. (1993). Effect of site-directed mutagenesis on conserved positions of Drosophila alcohol dehydrogenase. FEBS Letters, 319, 90-94. Deng, W. P. & Nickoloff, J. A. (1992). Site directed mutagenesis of virtually any plasmid by eliminating a unique site. Anal. Biochem. 200, 81-88. Engh, R. & Huber, R. (1991). Accurate bond and angle parameters for X-ray structure re®nement. Acta Crystallog. sect. A, 47, 392-400. Ensor, C. M. & Tai, H. (1996). Site-directed mutagenesis of the conserved serine 138 of human placental NAD‡-dependent 15-hydroxyprostaglandin dehydrogenase to an alanine results in an inactive enzyme. Biochem. Biophys. Research Communications, 220, 330-333. Etzioni, A., Frydman, M., Pollack, S., Avidor, I., Phillips, M. L. & Paulson, J. C. (1992). Brief report: recurrent severe infections caused by a novel leukocyte adhesion de®ciency. New Engl. J. Med. 327, 17891792. Feizi, T. (1990). The major blood group ABO(H) determining genes are isolated. Trends Biochem. Sci. 15, 330-331. Feizi, T. & Childs, R. A. (1987). Carbohydrates as antigenic determinants of glycoproteins. Biochem. J. 245, 1-11. Ginsburg, V. (1960). Formation of guanosine diphosphate L-fucose from guanosine diphosphate D-mannose. J. Biol. Chem. 235, 2196-2201. Ginsburg, V. (1961). Studies on the biosynthesis of guanosine diphosphate L-fucose. J. Biol. Chem. 236, 2389-2393. Gosh, D., Wawrzak, Z., Weeks, C. M., Duax, W. L. & Erman, M. (1994). The re®ned three-dimensional structure of 3a,20b-hydroxysteroid dehydrogenase and possible roles of the residues conserved in short-chain dehydrogenases. Structure, 2, 629-640. Gosh, D., Pletnev, V. Z., Zhu, D.-W., Wawrzak, Z., Duax, W. L., Pangborn, W., Labrie, F. & Lin, S. X. (1995). Structure of estrogenic 17b-hydrosysteroid dehydrogenase at 2.20 resolution. Structure, 3, 503513. 90 Jones, T. A., Zou, J. Y., Cowan, S. W. & Kjeldgaard, M. (1991). Improved methods for building protein models in electron density maps and the location of errors in these models. Acta Crystallog. sect. A, 47, 110-119. JoÈrnvall, H., Persson, B., Krook, M., Atrian, S., Gonzalez-Duarte, R., Jeffery, J. & Ghosh, D. (1995). Short-chain dehydrogenase/reductase (SDR). Biochemistry, 34, 6003-6013. Kraulis, P. J. (1991). MOLSCRIPT: a program to produce both detailed and schematic plots of protein structures. J. Appl. Crystallog. 24, 946-950. Labesse, G., Vidal-Cros, A., Chomilier, J., Gaudry, M. & Mornon, J. P. (1994). Structural comparisons lead to the de®nition of a new superfamily of NAD(P)(H)accepting oxidoreductase: the single-domain reductase/epimerase/dehydrogenase (the ``RED'' family). Biochem. J. 304, 95-99. Laemmli, U. K. (1970). Cleavage of structural proteins during the assembly of the head of bacteriophage T4. Nature, 227, 680-685. Liu, Y., Thoden, J. B., Kim, J., Berger, E., Gulik, A. M., Ruzicka, F. J., Holden, H. M. & Frey, P. A. (1997). Mechanistic roles of tyrosine 149 and serine 124 in UDP-galactose 4-epimerase from Escherichia coli. Biochemistry, 36, 10675-10684. Lowe, J. B. (1997). Selectin ligands, leukocyte traf®cking, and fucosyltransferase genes. Kidney Int. 51, 14181426. Menon, S., Stahl, M., Kumar, R., Xu, G. Y. & Sullivan, F. X. (1999). Stereochemical course and steady state mechanism of the reaction catalyzed by the GDPfucose synthetase from Escherichia coli. J. Biol. Chem. 274, 26743-26750. Morelli, A. & De Flora, A. (1977). Isolation and partial characterization of a NADP and NADPH-binding protein from human erythrocytes. Arch. Biochem. Biophys. 179, 698-705. Murshudov, G. N., Vagin, A. A. & Dodson, E. J. (1997). Re®nement of macromolecular structures by the maximum-likelihood method. Acta Crystallog. sect. D, 53, 240-257. Otwinowski, Z. & Minor, W. (1997). Processing of X-ray diffraction data collected in oscillation mode. Methods Enzymol. A276, 307-326. Persson, B., Rook, K. & JoÈrnvall, H. (1991). Characteristic of short-chain alcohol dehydrogenase and related enzymes. Eur. J. Biochem. 200, 537-543. Phillips, M. L., Schwartz, B. R., Etzioni, A., Bayer, R., Ochs, H. D., Paulson, J. C. & Harlan, J. M. (1995). Neutrophil adhesion in leukocyte adhesion de®ciency syndrome type 2. J. Clin. Invest. 96, 28982906. Rizzi, M., Tonetti, M., Vigevani, P., Sturla, L., Bisso, A., De Flora, A., Bordo, D. & Bolognesi, M. (1998). GDP-4-keto-6-deoxy-D-mannose epimerase/reductase from Escherichia coli, a key enzyme in the biosynthesis of GDP-L-fucose, displays the structural characteristics of the RED protein homology superfamily. Structure, 6, 1453-1465. Rosano, C., Izzo, G., Sturla, L., Bisso, A., Tonetti, M. & Bolognesi, M. (2000). The high resolution structure of GDP-4-keto-6-deoxy-D-mannose epimerase/ reductase. Croatica Chem. Acta. In the press. Somers, W. S., Stahl, M. L. & Sullivan, F. X. (1998). GDP-fucose synthetase from Escherichia coli: structure of a unique member of the short-chain dehydrogenase/reductase family that catalyzes two GDP-L-fucose Biosynthesis distinct reactions at the same active site. Structure, 6, 1601-1612. Somoza, J. R., Menon, S., Schmidt, H., Joseph-McCarthy, D., Dessen, A., Stahl, M. L., Somers, W. S. & Sullivan, F. X. (2000). Structural and kinetic analysis of Escherichia coli GDP-mannose 4,6-dehydratase provides insights into the enzyme's catalytic mechanism and regulation by GDP-fucose. Structure, 8, 123-135. Sturla, L., Bisso, A., Zanardi, D., Benatti, U., De Flora, A. & Tonetti, M. (1997). Expression, puri®cation and characterization of GDP-D-mannose 4,6-dehydratase from Escherichia coli. FEBS Letters, 412, 126130. Sullivan, F. X., Kumar, R., Kriz, R., Stahl, M., Xu, G., Rouse, J., Chang, X., Boodhoo, A., Potvin, B. & Cummings, D. (1998). Molecular cloning of human GDP-mannose 4,6-dehydratase and reconstitution of GDP-fucose biosynthesis in vitro. J. Biol. Chem. 273, 8193-8202. Szikora, J. P., Van Pel, A., Brichard, V., Andre, M., Van Baren, N., Henry, P., De Plaen, E. & Boon, T. (1990). Structure of the gene of tum- transplantation antigen P35B: presence of a point mutation in the antigenic allele. EMBO J. 9, 1041-1050. Tanaka, N., Nonaka, T., Tanabe, T., Yoshimoto, T., Tsuru, D. & Mitsui, Y. (1996a). Crystal structures of the binary and ternary complexes of 7a-hydroxysteroid dehydrogenase from Escherichia coli. Biochemistry, 35, 7715-7730. Tanaka, N., Nonaka, T., Nakanishi, M., Deyashiki, Y., Hara, A. & Mitsui, Y. (1996b). Crystal structure of the ternary complex of mouse lung carbonyl Ê resolution: the structural origin reductase at 1.8 A of coenzyme speci®city in the short-chain dehydrogenase/reductase family. Structure, 4, 33-45. Thoden, J. B., Frey, P. A. & Holden, H. M. (1996a). Crystal structure of the oxidized and reduced forms of UDP-galactose 4-epimerase isolated from Escherichia coli. Biochemistry, 35, 2557-2566. Thoden, J. B., Frey, P. A. & Holden, H. M. (1996b). Molecular structure of the NADH/UDP-glucose abortive complex of UDP-galactose 4-epimerase from Escherichia coli: implications for the catalytic mechanism. Biochemistry, 35, 5137-5144. Thoden, J. B., Frey, P. A. & Holden, H. M. (1996c). High resolution X-ray structure of UDP-galactose 4-epimerase complexed with UDP-phenol. Protein Sci. 5, 2149-2161. Thoden, J. B., Hegeman, A. D., Wesenberg, G., Chapeau, M. C., Frey, P. A. & Holden, H. M. (1997). Structural analysis of UDP-sugar binding to UDP-galactose 4-epimerase from Escherichia coli. Biochemistry, 36, 6294-6304. Tonetti, M., Sturla, L., Bisso, A., Benatti, U. & De Flora, A. (1996). Synthesis of GDP-L-fucose by the human FX protein. J. Biol. Chem. 271, 27274-27279. Tonetti, M., Sturla, L., Bisso, A., Zanardi, D., Benatti, U. & De Flora, A. (1998a). The metabolism of 6-deoxyhexoses in bacterial and animal cells. Biochimie, 80, 923-931. Tonetti, M., Rizzi, M., Vigevani, P., Sturla, L., Bisso, A., De Flora, A. & Bolognesi, M. (1998b). Preliminary crystallographic investigations of recombinant GDP4-keto-6-deoxy-D-mannose epimerase/reductase from E. coli. Acta Crystallog. sect. D, 54, 684-686. Varki, A. (1994). Selectin ligands. Proc. Natl Acad. Sci. USA, 91, 7390-7397. GDP-L-fucose Biosynthesis Varughese, K. I., Xuong, N. H., Kiefer, P. M., Matthews, D. A. & Whiteley, J. M. (1994). Structural and mechanistic characteristics of dihydropteridine reductase: a member of the Tyr-(Xaa)3-Lys-containing family of redictases and dehydrogenases. Proc. Natl Acad. Sci. USA, 91, 5582-5586. Wallace, A. C., Laskowski, R. A. & Thornton, J. M. (1995). LIGPLOT: a program to generate schematic 91 diagrams of protein-ligand interaction. Protein Eng. 8, 127-134. Zhang, L., Radziejewska-Lebrecht, J., KrajeweskaPietrasik, D., Toivanen, P. & Skurnik, M. (1997). Molecular and chemical characterization of the lipopolysaccharide O-antigen and its role in the virulence of Yersinia enterocolitica serotype O:8. Mol. Microbiol. 23, 63-76. Edited by K. Nagai (Received 15 May 2000; received in revised form 7 August 2000; accepted 15 August 2000)