Structure-Guided Engineering of Xylitol Dehydrogenase Cosubstrate Speciﬁcity

Home , Mannitol dehydrogenase

Andreas H. Ehrensberger,1,2,3 Robert A. Elling,1,2 is catalyzed by xylose reductase (XR), an NADPH-pre- and David K. Wilson1,2,* ferring enzyme which reduces xylose to form xylitol (Fig- 1 Section of Molecular and Cellular Biology ure 1). Xylitol dehydrogenase (XDH) reoxidizes xylitol University of California, Davis to form xylulose, using NAD+ exclusively as a cosub- Davis, California 95616 strate. Finally, xylulose kinase phosphorylates xylulose at the 5 position using ATP. The xylose-to-xylulose conversion can also be catalyzed in one step by xylose Summary isomerase, a prokaryotic enzyme. Attempts to express various forms of this enzyme in yeast have been largely Xylitol dehydrogenase (XDH) is one of several en- unsuccessful, with one recent exception (Kuyper et al., zymes responsible for assimilating xylose into eukary- 2005). otic metabolism and is useful for fermentation of xy- One possible solution to the cosubstrate recycling lose contained in agricultural byproducts to produce problem can be obtained with an NADH-dependent ethanol. For efficient xylose utilization at high flux XR. Progress in this area has been made by searching rates, cosubstrates should be recycled between the for enzymes with a rare natural preference for this co- NAD+-specific XDH and the NADPH-preferring xylose substrate. The XR from Candida parapsilosis possesses reductase, another enzyme in the pathway. To under- an approximately 100-fold preference for NADH over stand and alter the cosubstrate specificity of XDH, NADPH based on the ratio of catalytic efficiencies, we determined the crystal structure of the Glucono- kcat/Km (Lee et al., 2003). Alternatively, the abundance bacter oxydans holoenzyme to 1.9 A˚ resolution. The of biochemical and structural knowledge of the XR structure reveals that NAD+ specificity is largely con- from Candida tenuis has been exploited to design mu- ferred by Asp38, which interacts with the hydroxyls tants with altered cosubstrate specificity. Currently, of the adenosine ribose. Met39 stacked under the pu- the wild-type enzyme, which has a 33-fold preference rine ring and was also located near the 20 hydroxyl. for NADPH, has been altered to exhibit a 5-fold prefer- Based on the location of these residues and on se- ence for NADH (Petschacher et al., 2005). quence alignments with related enzymes of various As an alternative approach to alleviate the cosub- cosubstrate specificities, we constructed a double strate recycling problem, the cosubstrate specificity of mutant (D38S/M39R) that was able to exclusively use XDH can be changed to NADP+. Xylitol dehydrogenase, NADP+, with no loss of activity. also known as D-xylulose reductase, from Glucono- bacter oxydans (E.C. 1.1.1.9) is among the best-studied XDHs and has been shown to exclusively use NAD+.No Introduction structural data are, however, available for this or any other XDH. Based on its sequence, G. oxydans XDH The metabolic integration of xylose, the second most has been classified as a member of the superfamily of abundant sugar in nature after glucose, by yeast could short chain dehydrogenases/reductases (SDR, also play a key role in the bioconversion of lignocellulose- known as the short chain oxidoreductase/SCOR family), containing agricultural waste products to ethanol via which currently encompasses approximately 3000 pro- fermentation. A majority of postharvest residues are cur- teins. As a group, SDRs can accommodate a wide range rently disposed of through burning, a practice that pro- of substrate sizes and polarities and are found in all spe- duces greenhouse gases and leads to air pollution and cies. Enzymes with XDH activity also exist in the diver- associated human health problems. These waste prod- gent medium chain dehydrogenase/reductase (MDR) ucts could serve as an energy source if converted to eth- superfamily and cosubstrate engineering experiments anol, which has the potential to drive hydrogen fuel cells converting the cosubstrate specificity have been carried (Deluga et al., 2004). Although many five-carbon sugars out on a form from Pichia stipitis (Watanabe et al., 2004). are utilized by Saccharomyces cerevisiae, an organism In addition to its ability to oxidize xylitol, G. oxydans XDH that has been optimized for ethanol production, xylose has previously been shown to use sorbitol and mannitol is not. Efficient fermentation of xylose in budding yeast as relatively good substrates (Sugiyama et al., 2003). is not possible because the organism lacks the enzymes SDRs are characterized by a set of recurring se- necessary to integrate this sugar into its metabolism. quence motifs (Oppermann et al., 2003) and by a cata- Previous attempts have been made to express the en- lytic Ser-Tyr-Lys-Asn tetrad. Most are oxidoreductases zymes necessary for conversion of xylose into xylu- that utilize NAD+ or NADP+, and of the over 40 structures lose-5-phosphate, which can enter metabolism via the determined so far, all contain a single Rossmann fold pentose phosphate pathway (most recently reviewed domain. In many cases, SDRs assemble into dimers or by Jeffries and Jin, 2004). The first of these reactions tetramers, although monomers have been found to exist as well (Ghosh et al., 2001; Oppermann et al., 2003). As a result of these structural studies, the atomic interac- *Correspondence: [email protected] tions determining cosubstrate specificity are under- 2 Lab address: http://alanine.ucdavis.edu/wdave/wilsonlab/index. html stood and it has been possible to make relatively accu- 3 Present address: Biophysics Program, Stanford University, Stan- rate predictions regarding this selectivity (Duax et al., ford, California 94305. 2003; Persson et al., 2003). Structure 568

Figure 1. A Schematic of the Eukaryotic Xy- lose Assimilation Pathway

The crystal structure of NAD+-bound XDH has been NAD+ Binding determined to a resolution of 1.9 A˚ with an R factor of The axis of the NAD+ runs roughly parallel to the b sheet 16.7%. The high-resolution structure reveals details and perpendicular to the six Rossmann fold helices. The about catalysis and substrate and cosubstrate specific- N-terminal region of the primary sequence is responsi- ity, and allowed us to design a double mutant specific ble for binding the adenine portion of the cofactor and for NADP+ utilization. By introducing two mutations establishing selectivity for NAD+ or NADP+. The nicotin- into XDH, we have developed an enzyme that may allow amide ring which defines the active site of the enzyme is S. cerevisiae cells to metabolize xylose without the confined between helices aFG1 and aFG2 and the C-ter- problem of cosubstrate depletion. minal side of the b sheet. The NAD+ is anchored to the enzyme with the nicotinamide ring in the syn conformation, which is stabilized in this conformation by an intra- Results and Discussion molecular hydrogen bond between the nicotinamide amide and the pyrophosphate (Figure 3). Both ribose hy- Overall Structure droxyls engage in hydrogen bonds with Lys161, and the 20 hydroxyl forms a hydrogen bond with the catalytic G. oxydans XDH crystallized in space group P21 with two tetramers in the asymmetric unit. Phasing was carried residue; Tyr157 has been shown to form part of the pro- out via molecular replacement using the 38% identical ton conduction system to and from the active site (Filling Corynebacterium aquaticum levodione reductase tetra- et al., 2002). mer (Sogabe et al., 2003) as the search model. Each tet- The adenine ribose is positioned in anti conformation, ramer is assembled via noncrystallographic 222 symme- with the adenine ring stacking adjacent to Met39. Both try and contains one clearly observable NAD+ molecule ribose hydroxyls form hydrogen bonds with Asp38, the + in each of its monomers (Figure 2A). The tetramer re- key polar determinant of NAD specificity. The adenine sembles a flat torus and has an approximate diameter primary amine forms a hydrogen bond with Asp64. 0 of 80 A˚ . Of the 262 residues constituting each of the in- Both riboses on the molecule adopt the C2 -endo con- tact monomers (designated A–H), residues 3–262 were formation. The pyrophosphate moiety interacts primar- fit into the electron density. There is very little structural ily with water molecules and contacts the protein only divergence between the monomers, with the rms devia- through Asn18 and the backbone amide nitrogen of tion between them ranging from 0.18 to 0.28 A˚ . All refer- Ile19. ences to a particular feature of the XDH structure will in this paper refer to monomer E, as it subjectively ex- Magnesium and Metal Binding to XDH hibited the best-defined electron density. Five small spherical regions of high electron density

Interactions between monomers within the tetramer (5 to 6 s in the 2Fo 2 Fc map) were found in the asymmet- are stabilized on two sides by hydrogen bonds between ric unit. Because the crystallization buffer contained adjacent 7-stranded b sheets, forming two 14-strand 100 mM MgCl2, and the spheres exhibited octahedral ge- wide parallel sheets. The other two sides of the tetramer ometry typical for divalent cations, five magnesium ions are stabilized by four-helix bundle interactions between were assigned to these densities. All five ions are located two helices on one monomer and two helices on the far from the active sites and substrate binding pockets neighboring monomer (Figure 2B). and are therefore likely to have no direct impact on catal- Each monomer contains a single Rossmann fold do- ysis. The shortest distance from the C4 of a nicotinamide main consisting of a large 7-stranded b sheet flanked to a magnesium is 16 A˚ . Two Mg2+ ions lie directly on by three short helices on one side (aB, aC, and aG, using a noncrystallographic 2-fold axis (the horizontal axis in conventional SDR nomenclature) and three longer heli- Figure 2A) between two monomers in similar environ- ces (aD, aE, and aF) on the other side of the sheet ments, while the fifth ion lies outside both tetramers. (Figure 2B). The two remaining short helices, aFG1 and Each ion is coordinated by six water molecules, which aFG2, are suspended outside and above the Rossmann bind mostly to backbone carbonyl oxygens. The average fold in the perspective shown in Figure 2B. distance of the magnesiums to their ligands is 2.17 A˚ The core of XDH, encompassing everything but the (max = 2.29 A˚ , min = 2.07 A˚ , standard deviation = peripheral helices aFG1 and aFG2, shows strong resem- 0.06 A˚ ). As determined by others (Sugiyama et al., blance to most other SDRs. A flexible hinge connects 2003), XDH is inhibited neither by Mg2+ nor by the Mg2+- helices aFG1 and aFG2 to the core, and its orientation sequestering agent EDTA. Magnesium has been found shows some variation between SDRs. Helix aFG1 is in- to be integral in dimeric interfaces of other SDRs volved in binding both substrate and cosubstrate. such as dTDP-6-deoxy-L-lyxo-4-hexulose reductase Xylitol Dehydrogenase Crystal Structure 569

Figure 2. The Structure of XDH (A) Stereoview of the XDH tetramer with bound NAD+. Each monomer is shown in a different color. The two magnesium ions are shown in ma- genta. Figures 2A, 4, and 5B were generated using the programs MOLSCRIPT (Kraulis, 1991) and Raster3D (Merritt and Bacon, 1997). (B) XDH monomer. Locations of residues directly involved in catalysis are colored blue and residues responsible for establishing cosubstrate speciﬁcity are shown in red. Secondary structures involved in tetramerization are colored in gray. Figures 2B and 5A were created using the program PyMOL (DeLano, 2002).

(Blankenfeldt et al., 2002), suggesting that it is important is oxidized by the enzyme to a carbonyl, hydrogen for stabilization and/or formation of the oligomer. bonds with the catalytic base Tyr157. The C2 carbon, which transfers its hydride to NAD+ during the course Substrate Binding Pocket of the reaction, has a distance of 2.6 A˚ to the nicotin- To understand how the substrate is bound by the active amide C4, suitable for hydride transfer. In this position, site of the enzyme, a xylitol molecule was modeled into three more hydroxyls on the substrate form hydrogen the putative substrate binding pocket (Figure 4). The bonds with the enzyme: the hydroxyls at positions 2 sterically and catalytically most feasible position resem- and 4 with Ser144, and the C5 hydroxyl to the side chain bles the arrangement of estrogen in the ternary complex amide oxygen of Gln199. Ser144 is highly conserved of 17b-hydroxysteroid dehydrogenase (Mazza et al., among SDRs and, in agreement with our model, is gen- 1998). In this position, the C2 hydroxyl of xylitol, which erally believed to be essential for stabilization of the Structure 570

tions of xylitol inhibited the enzyme, kinetic parameters were determined by ﬁtting data to Equation 1 (see Ex- perimental Procedures), which accounts for substrate inhibition. Others have noted that sorbitol is a much bet- ter substrate than xylitol, which suggests that, in vivo, XDH may therefore function as a sorbitol dehydrogenase rather than a xylitol dehydrogenase.

Catalytic Site The widely accepted general catalytic model for SDRs depends primarily on a highly conserved and essential tyrosine which is believed to act as the general acid/ base (Filling et al., 2002). In the reducing direction, a proton is channeled from the solvent to the tyrosine via a se- ries of interactions involving the nicotinamide ribose 20 hydroxyl, a conserved lysine, and an ordered water molecule which interacts with bulk solvent. In XDH, Tyr157 and Lys161 are observed in structurally conserved con- formations, in agreement with this general mechanism (Figure 3). A conserved asparagine at position 116 is also found. The main chain carbonyl of this residue has been shown to be responsible for anchoring the water molecule involved in proton channeling, as is the case in 3b/17b-hydroxysteroid dehydrogenase (Filling et al., 2002). Finally, Ser144 is also conserved and poised to stabilize a negative charge on the catalytic intermediate. Based on the essential and highly con-

+ served roles of Tyr157, Lys161, Ser144, and Asn116, Figure 3. A Schematic of Polar Interactions Involved in NAD Bind- ing to XDH the former catalytic Tyr-Lys-Ser triad has recently For interactions only seen in some of the molecules within the asym- been renamed a Tyr-Lys-Ser-Asn tetrad (Filling et al., metric unit, the name of the residue is enclosed by parentheses. 2002; Oppermann et al., 2003). catalytic intermediate (Filling et al., 2002; Oppermann Comparison with Other Short Chain et al., 2003). In addition to these interactions between Dehydrogenases/Reductases xylitol and XDH, the model predicts that the C1 hydroxyl In addition to the catalytic tetrad, three sequences are would form a hydrogen bond with the NAD+ amide oxy- highly conserved among SDRs: an N-terminal Thr-Gly- gen. The C5 hydroxyl occupies a site currently filled by X3-Gly-X-Gly segment, a central Asn-Asn-Ala-Gly se- a water molecule, which is held into place by other water quence, and a C-terminal Pro-Gly (Duax et al., 2003; Op- molecules in an open pocket of the enzyme. Overall, this permann et al., 2003). All of these sequences are present set of interactions ensures that only C2 and not C1 or C3 in XDH. The first is present in a slightly modified form as of xylitol is oxidized. Other possible orientations for the Thr13-Gly14-Ala15-Gly16-Gly17-Asn18-Ile19-Gly20, the xylitol lead to collisions between molecules or lack fea- second as Asn90-Asn91-Ala92-Gly93, and the last as sible hydrogen bonds to the protein. The proposed Pro187-Gly188. The first two regions are believed to be placement of the substrate agrees with the finding that responsible for b sheet stabilization and coenzyme bind- the six-carbon polyol sorbitol also serves as an excellent ing (Oppermann et al., 2003), roles that stand in agree- substrate for XDH (Sugiyama et al., 2003). The first four ment with this structure. Residues Gly17, Ile19, and carbons are stereochemically identical in xylitol and sor- Asn91 on these segments form direct hydrogen bonds bitol, and a cavity on the enzyme extending beyond C5 to the NAD+ (Figure 3). The Pro-Gly sequence has allows the larger sorbitol to fit into the active site as well. been associated with determining the preference of We tested and quantitated the ability of XDH to use the enzyme for binding either the alcohol or the ke- xylitol as a substrate (Table 1). Because high concentra- tone/aldehyde. In XDH, Pro187 sits adjacent and parallel

Figure 4. Xylitol Modeling into the Active Site XDH carbons are shown in gray, NAD+ carbons in black, and xylitol carbons in green. Hydrogen bonds are drawn as dashed yellow lines, and the path of hydride transfer between xylitol C2 and NAD+ C4 is marked with green spheres. Xylitol Dehydrogenase Crystal Structure 571

Table 1. Kinetic Parameters for Wild-Type and Double Mutants Pro187 and C5 in five of eight subunits (Figure 3). Or- dered water molecules were responsible for disrupting Substrate Wild-Type Double Mutant interactions in the last two cases. NAD+ Km (mM) 348 NMA Conversion of Cosubstrate Specificity 21 kcat (s ) 27.2 NMA To examine which residues on XDH should be mutated k /K (s21mM21) 0.08 NMA cat m to reverse the specificity from NAD+ to NADP+,we NADP+ made a structural alignment of the XDH structure with Km (mM) NMA 206 21 kcat (s ) NMA 20.5 11 other SDRs with known cosubstrate specificity (Table 21 21 kcat/Km (s mM ) NMA 0.10 2; Figure 5B). By looking for the residues known to con- Xylitol fer cosubstrate specificity in SDRs (Persson et al., 2003), Km (mM) 13.7 100 we concluded that several of these structural elements k (s21) 24.6 17.9 cat establish the specificity of XDH for NAD+ as well. First, k /K (s21mM21) 1.79 0.18 cat m XDH contains the residue Asp38, completely conserved NMA, no measurable activity. among NAD+-specific SDRs. Each of the carboxylate oxygens in Asp38 forms a hydrogen bond with one of the two adenine ribose hydroxyls, preventing the inser- to the NAD+ nicotinamide and its carbonyl appears to tion of the NADP+ phosphate group. The 20 and 30 hy- engage in hydrogen bonding in some cases (Duax droxyls of ribose are commonly recognized by a similar et al., 2003). interaction with carboxylate-containing residues in A conserved threonine, usually found around position many other unrelated enzymes, including NADH-depen- 188 and believed to be essential for proper coenzyme dent enzymes that do not adopt the Rossmann fold (Ka- binding in other SDRs (Zhou and Tai, 1999), is absent vanagh et al., 2003). in XDH. This residue usually interacts with the nicotin- In addition to the NAD+-dependent enzymes, five amide amide and was believed to be necessary for the NADP+-specific enzymes were also examined (Table 2) proper orientation of the ring during catalysis. In XDH, (Andersson et al., 1996; Horer et al., 2001; Liu et al., a water molecule is found in place of this threonine. 1997; Nakajima et al., 1998; Somers et al., 1998). Most It has been suggested that hydrogen bonding notably, XDH lacks the two basic residues commonly between the nicotinamide ring carbons and protein found at positions 17 and 39 in the NADP+-specific en- hydrogen bond acceptors may play a role in catalysis zymes. Each of these five enzymes contains at least by fine-tuning its environment (Duax et al., 2003). They one arginine or lysine at those sites. Four of these examined 11 different SDR structures, and interactions NADP+-specific SDRs contained an arginine at position were consistently found between Ser144 (XDH number- 39. Both residues are located in the binding site for the ing) and the ring C5 and often between the carbonyl of phosphate and help stabilize its negative charge. Spe- Pro187 and the ring C5. Various other interactions cifically, residue 39 often forms a direct salt bridge to were also observed between protein hydrogen bond ac- the critical phosphate (Figure 5). ceptors and either the C4 or C5 atoms. When a similar Based on these observations, we designed a double analysis of all eight molecules in the XDH asymmetric (D38S, M39R) and a triple mutant (D38S, M39R, G17K). unit was done, the only interaction that was consistently The double mutant expressed with a yield similar to present was between the carbonyl of Ala143 and the nic- that of the wild-type enzyme. Severe solubility and sta- otinamide C6 (average distance = 3.33 A˚ ; standard devi- bility problems, which prevented a detailed kinetic and ation = 0.05 A˚ ). Other interactions with plausible geom- structural analysis, were encountered with the triple mu- etry were seen between the side chain of Ser144 and tant. Assays demonstrated that wild-type XDH reduced C5 in three of eight subunits and the carbonyl of NAD+ with a catalytic efficiency of 0.08 s21mM21

Table 2. Identity of Cosubstrate Speciﬁcity Deﬁning Residues in Several Short Chain Dehydrogenases/Reductases

Speciﬁcity Residues Enzyme PDB ID Cofactor(s) Asp38 Lys/Arg 17 Lys/Arg 39 Reference Xylitol dehydrogenase 1ZEM NADH Asp Gly Met Levodione reductase 1IY8 NADH Asp Ser Leu (Sogabe et al., 2003) HSDH 2HSD NADH Asp Arg Leu (Ghosh et al., 1994) SBP 1QRR NADH Asp Gly Asn (Mulichak et al., 1999) Tropinone reductase 1AE1 NADPH Ser Lys Arg (Nakajima et al., 1998) Carbonyl reductase 1CYD NADPH Thr Lys Arg (Liu et al., 1997) Mannitol dehydrogenase 1H5Q NADPH Tyra Arg Arg (Horer et al., 2001) THNR 1YBV NADPH Tyra Arg Ala (Andersson et al., 1996) GDP-fucose synthase 1BSV NADPH b Gly Arg (Somers et al., 1998) 17b-hydroxysteroid DH 1FDW NAD(P)H Ile Ser Gly (Mazza et al., 1998) AMHE 1EQ2 NAD(P)H Asp Gly Asn (Deacon et al., 2000) HSDH, 3a20b-hydroxysteroid dehydrogenase; SBP, sulfolipid biosynthetic protein; THNR, 1,3,8-trihydroxynaphthalene reductase; AMHE, ADP- mannoheptose epimerase. a Tyrosine rotated away from NADP+, causing altered hydrogen bonding (Duax et al., 2003). b Different active site architecture. Structure 572

Figure 5. Cosubstrate Binding to XDH and Other SDRs (A) A stereoview of NAD+ and catalytic resi-

dues with density from an Fo 2 Fc omit map contoured at 3s. The catalytic base Tyr157, and Lys161, which form part of the catalytic tetrad, are also shown. (B) Overlay of the XDH NAD+ binding site with the tropinone reductase (TR) NADP+ binding site in stereo. Based on the identity of its three speciﬁcity residues, TR is a prototypical NADP+-dependent SDR. XDH and XDH- bound NAD+ are colored white, and TR and TR-bound NADP+ are colored black. Protein and adenine ribose atoms are colored red (oxygen), blue (nitrogen), and yellow (sulfur). The two XDH residues responsible for confer- ring NAD+ speciﬁcity to the enzyme (Met39 and Asp38) and the corresponding residues in TR (Arg53 and Ser52) are labeled. The rele- vant hydrogen bonds between XDH Asp38 and the NAD+ ribose and the salt bridge between TR Arg53 and the NADP+ 20 phosphate are depicted in yellow.

(kcat/Km) and showed no measurable activity with straightforward and accurate. NAD(H) utilization de- NADP+, in agreement with prior qualitative results (Su- pends on the presence of a carboxylate-containing res- giyama et al., 2003). On the other hand, the D38S/ idue (usually an aspartate) at the position equivalent to M39R double mutant was entirely inactive with NAD+ Asp38 in XDH. NADP(H) specificity is provided by the re- and exhibited a catalytic efficiency of 0.10 s21mM21 for placement of this carboxylate (usually by a serine or the reduction of NADP+ using saturating concentrations threonine) and the presence of a basic residue at the ad- of xylitol, comparable to that found in the wild-type with jacent position (39 in XDH). An additional basic residue NAD+ (Table 1; Figure 6). Both the catalytic rate constant is sometimes present as well in the NADP(H)-dependent and the Km were lowered by approximately a third. With enzymes at a position equivalent to residue 17 in XDH + aKm of 206 mM, mutant XDH binds NADP more tightly (Table 2). + than the wild-type, whose Km for NAD is 348 mM. The Complete reversal of cosubstrate specificity by muta- Km for xylitol is much more sensitive to the two mutations in the active site can be more problematic. In the tions, and the double mutant binds xylitol with a Km case of XDH, the double mutant D38S/M39R completely almost 7-fold higher than the wild-type. converted the specificity, while the triple mutant G17K/ D38S/M39R was unstable for reasons that are unclear. Conclusions Others have found that introducing similar mutations In general, the prediction of cosubstrate specificity from produces enzymes with altered cosubstrate preference sequence data in SDRs has been shown to be relatively but often without a change in exclusive specificity. This Xylitol Dehydrogenase Crystal Structure 573

Protein Expression and Puriﬁcation E. coli BL21* cells were transformed with the plasmid coding for XDH and were grown in 6 liters of Luria broth medium supplemented with

100 mg/ml ampicillin. At an OD600 of 0.6, cells were cooled to 15ºC for 1 hr and then induced with 750 mM isopropylthiogalactoside for 16 hr. After centrifugation, the pellet was resuspended in column buffer (0.5 M NaCl, 0.1 mM EDTA, 20 mM Tris [pH 8.0]) supplemented with 0.1% Triton X-100, then ruptured using microfluidization and sonication. The cell extract was loaded on 15 ml chitin beads (New England Biolabs) to purify XDH through the chitin binding/intein domain fused to the N terminus of XDH which was incorporated from the pTYB1 plasmid. The loaded column was washed overnight in column buffer with 0.1% Triton X-100 followed by a wash with column buffer alone to remove detergent. XDH was eluted by quickly washing the column with column buffer containing 0.3% (v/v) b-mer- captoethanol and incubating overnight to cleave the protein from the chitin binding/intein domain. The protein was concentrated in a 10,000 MW cutoff filter and the buffer was changed to 5 mM HEPES (pH 7.0). In a secondary purification step, the protein was loaded on an HQ anion exchange column, and washed and eluted in a 0–450 mM NaCl gradient at pH 7.5. The major peak containing XDH activity eluted at 20 mS. Purified XDH was concentrated to 13 mg/ml and the buffer was changed to 5 mM HEPES (pH 7.0). The final yield was 3.1 mg pure XDH per liter culture.

Figure 6. Activities of Wild-Type and Mutant XDH Dynamic Light Scattering Velocity plotted as a function of cofactor concentration. Wild-type To determine the XDH oligomerization state in solution, dynamic XDH was assayed using NAD+ (:) while the double mutant was light scattering measurements were taken. The experiment was per- assayed with NADP+ ( ). The data were ﬁt to Equation 1. formed on a Protein Solutions Dynapro-99 using 1.6 mg/ml XDH in 10 mM HEPES (pH 7.5).

Kinetic Assays has, for example, been observed when attempting to All kinetic assays were performed in 100 mM Tris (pH 7.0). Enzyme change the specificity of the NADP(H)-specific mouse was used at concentrations ranging from 28 to 280 nM. When assay- lung carbonyl reductase by mutating a threonine to an ing activity with varied xylitol or sorbitol concentrations, [NAD(P)+] aspartate (Nakanishi et al., 1997). Complete conversion was kept at saturating concentrations of 2 to 10 mM. Xylitol was of NAD(H) to NADP(H) dependence of an SDR domain kept at saturating concentrations of 1.5 M (wild-type XDH) and 2 M + + was successfully accomplished with the double muta- (double mutant) for determining the Km of NAD and NADP . Saturat- ing concentrations were always at least ten times the K of the cor- tion D36A/K37R in the case of the bifunctional human m responding substrate or cosubstrate. Activity was determined by type 1 3b-hydroxysteroid dehydrogenase/isomerase measuring the increase in absorbance at 340 nm corresponding to (Thomas et al., 2003). Other examples have been less the reduction of NAD(P)+ to NAD(P)H using a Shimadzu UV-160U successful, although the nature of the mutations may spectrophotometer. The noncatalytic reduction rate was negligible. not have been optimal in view of the recent structural The resulting values were fit to the Michaelis-Menten equation using data. The aspartate in the NAD(H)-preferring rat dihy- the program DeltaGraph (SPSS, Inc.) (Figure 6). When wild-type XDH was assayed for sorbitol or xylitol activity, we observed inhibition at dropteridine reductase was replaced with an isoleucine high substrate concentrations. This effect was accounted for by fit- to produce an enzyme which has improved NADP(H) uti- ting to the following equation: lization (Grimshaw et al., 1992). The homologous aspar- V ½S tate in Drosophila alcohol dehydrogenase was changed v = max : (1) ½S2 to an asparagine, making the enzyme dually NADH/ KmS + ½S + NADPH dependent (Chen et al., 1991). Finally, the role KiS of a carboxylate in establishing specificity for the cosubstrate is not exclusive to SDRs and has been recognized Preparation of Mutant Xylitol Dehydrogenases in other families of oxidoreductases, including the me- The XDH double and triple mutants were prepared using a previously described protocol (Hemsley et al., 1989). For the D38S/M39R dou- dium chain dehydrogenases/reductases and the aldo- ble mutant, we used the wild-type plasmid as the template and am- keto reductases (Kavanagh et al., 2003; Watanabe plified it using the primers 50-GGGCACGGCCATCGCCCTTCTGAG et al., 2004). CCGGAACCGCGA-30 and 50-CCAGCGCCTCGCGGTTCCGGCTCA GAAGGGCGATGG-30, where mismatched bases are underlined. This resulted in a plasmid coding for the D38S/M39R double mutant Experimental Procedures XDH. The triple mutant (G17K/D38S/M39R) was generated from the double mutant plasmid as a template using the primers 50-CAC Cloning CGGCGCGGGTAAGAATATCGGTCTTGC-30 and 50-GCAAGACCGA The gene for xylitol dehydrogenase (GenBank accession number TATTCTTACCCGCGCCGGTG-30. The plasmids were inserted into AB091690) was amplified from lyophilized ATCC 3448 Glucono- BL21* E. coli cells and protein was expressed and purified using bacter oxydans cells by the polymerase chain reaction using the the same protocol as for wild-type XDH. The yield of double mutant proofreading Vent DNA polymerase (New England Biolabs). The se- was 3.5 mg per liter culture. quences of the primers were 50-GCTAGTGCATATGTCGAAGAA GTTTAACGGTAAAGTCTG-30 and 50-GGTGGTTGCTCTTCCGCAAC Crystallization and Data Collection CGCCAGCAATCGGCAGGTTC-30. The resulting DNA fragment was XDH crystals were grown at room temperature using the hanging inserted into the NdeI and SapI restriction sites on the pTYB1 plas- drop vapor diffusion method with a ratio of 1 ml protein (13 mg/ml) mid (New England Biolabs). The construct was then sequenced to to 1 ml mother liquor. NAD+ was added to the drop at a final concen- confirm the fidelity of the PCR reaction. tration of 5 mM. Initial crystallization conditions were obtained using Structure 574

lowship. The data collection facilities at the Stanford Synchrotron Table 3. Data Collection and Structure Reﬁnement Statistics Radiation Laboratory are funded by the U.S. Department of Energy for NAD+-Bound XDH and by the National Institutes of Health. Data Collection Statistics Wavelength (A˚ ) 0.953695 Received: June 22, 2005 Resolution (A˚ ) (highest shell) 100–1.9 (1.97–1.90) Revised: October 18, 2005 Reﬂections (observed/unique) 447,789/137,448 Accepted: November 4, 2005 Completeness (%) 92.1 (84.0) Published online: March 14, 2006 I/s(I) 12.6 (6.2)

Rmerge 0.064 (0.191) References Refinement Statistics Andersson, A., Jordan, D., Schneider, G., and Lindqvist, Y. (1996). Space group P21 Unit cell parameters (A˚ , º) a = 87.959, b = 64.717, Crystal structure of the ternary complex of 1,3,8-trihydroxynaphtha- c = 168.239, b = 93.058 lene reductase from Magnaporthe grisea with NADPH and an active- Monomers per A.U. 8 site inhibitor. Structure 4, 1161–1170. Matthews coefficient 2.19 Blankenfeldt, W., Kerr, I.D., Giraud, M.F., McMiken, H.J., Leonard, Resolution (A˚ ) 30–1.9 G., Whitfield, C., Messner, P., Graninger, M., and Naismith, J.H.

Rcryst (%) 16.7 (2002). Variation on a theme of SDR. dTDP-6-deoxy-L-lyxo-4-hexu- 2+ Rfree (%) 21.8 lose reductase (RmlD) shows a new Mg -dependent dimerization Average temperature factors (A˚ 2) mode. Structure 10, 773–786. Tetramer 1 (monomers A–D) 17.7 Bru¨ nger, A.T., Adams, P.D., Clore, G.M., DeLano, W.L., Gros, P., Tetramer 2 (monomers E–H) 16.4 Grosse-Kunstleve, R.W., Jiang, J.S., Kuszewski, J., Nilges, M., + NAD chains A–D 19.9 Pannu, N.S., et al. (1998). Crystallography & NMR system: a new + NAD chains E–H 19.1 software suite for macromolecular structure determination. Acta 2+ Mg ions 13.9 Crystallogr. D Biol. Crystallogr. 54, 905–921. Rmsd bond lengths (A˚ ) 0.019 Chen, Z., Lee, W.R., and Chang, S.H. (1991). Role of aspartic acid 38 Rmsd bond angles (º) 2.01 in the cofactor specificity of Drosophila alcohol dehydrogenase. Eur. Protein atoms 15,512 J. Biochem. 202, 263–267. NAD+ atoms 352 Mg2+ ions 5 Deacon, A.M., Ni, Y.S., Coleman, W.G., Jr., and Ealick, S.E. (2000). Water molecules 1,648 The crystal structure of ADP-L-glycero-D-mannoheptose 6-epimer- Ramachandran plot ase: catalysis with a twist. Structure 8, 453–462. % in most favorable regions 91.5 DeLano, W.L. (2002). The PyMOL Molecular Graphics System (San % in allowed regions 8.5 Carlos, CA: DeLano Scientific). Parentheses indicate values for the highest resolution shell. Deluga, G.A., Salge, J.R., Schmidt, L.D., and Verykios, X.E. (2004). Renewable hydrogen from ethanol by autothermal reforming. Sci- ence 303, 993–997. Duax, W.L., Pletnev, V., Addlagatta, A., Bruenn, J., and Weeks, C.M. the Hampton crystallization screen. Conditions were optimized to (2003). Rational proteomics I. Fingerprint identification and cofactor 20% PEG-4000, 100 mM MgCl2, 70 mM Tris (pH 7.0) and yielded crystals that grew in 1 day. For cryocooling, the crystals were har- specificity in the short-chain oxidoreductase (SCOR) enzyme family. vested in well solution with 10% ethylene glycol and 5 mM NAD+ Proteins 53, 931–943. added. Diffraction data were collected on beamline 9-1 at the Stan- Filling, C., Berndt, K.D., Benach, J., Knapp, S., Prozorovski, T., Nor- ford Synchrotron Radiation Laboratory and processed using dling, E., Ladenstein, R., Jornvall, H., and Oppermann, U. (2002). DENZO and SCALEPACK (Otwinowski and Minor, 1997)(Table 3). Critical residues for structure and catalysis in short-chain dehydro- Attempts were also made to obtain structures of XDH-NAD(H) genases/reductases. J. Biol. Chem. 277, 25677–25684. 2+ bound to xylitol or to the known inhibitor Ni (Sugiyama et al., Ghosh, D., Wawrzak, Z., Weeks, C.M., Duax, W.L., and Erman, M. 2003), but failed to yield interpretable electron densities. (1994). The refined three-dimensional structure of 3 a,20 b-hydroxysteroid dehydrogenase and possible roles of the residues con- Structure Refinement and Model Building served in short-chain dehydrogenases. Structure 2, 629–640. Initial phases were obtained by molecular replacement using the program EPMR (Kissinger et al., 1999) and the levodione reductase Ghosh, D., Sawicki, M., Pletnev, V., Erman, M., Ohno, S., Nakajin, S., tetramer (Protein Data Bank ID code 1IY8)(Sogabe et al., 2003), and Duax, W.L. (2001). Porcine carbonyl reductase. Structural basis which shares 38% sequence identity with XDH, as the search model. for a functional monomer in short chain dehydrogenases/reduc- This yielded a structure containing two tetramers per asymmetric tases. J. Biol. Chem. 276, 18457–18463. unit and with a correlation coefficient of 44.9% and an R factor of Grimshaw, C.E., Matthews, D.A., Varughese, K.I., Skinner, M., 54.7%. The sequence was adjusted to the XDH sequence, and the Xuong, N.H., Bray, T., Hoch, J., and Whiteley, J.M. (1992). Character- structure further refined using CNS (Bru¨ nger et al., 1998) and fit ization and nucleotide binding properties of a mutant dihydropteri- with O (Jones et al., 1991). Noncrystallographic symmetry restraints dine reductase containing an aspartate 37-isoleucine replacement. were applied for the initial rounds of refinement and removed for fi- J. Biol. Chem. 267, 15334–15339. nal stages. Five percent of the reflections were excluded from refine- Hemsley, A., Arnheim, N., Toney, M.D., Cortopassi, G., and Galas, ment and constituted the Rfree set. In the latter stages of refinement, D.J. (1989). A simple method for site-directed mutagenesis using weighting placed upon geometric restraints was optimized to pro- the polymerase chain reaction. Nucleic Acids Res. 17, 6545–6551. duce the lowest R value. The final geometry was assessed using free Horer, S., Stoop, J., Mooibroek, H., Baumann, U., and Sassoon, J. PROCHECK (Laskowski et al., 1993). (2001). The crystallographic structure of the mannitol 2-dehydrogenase NADP+ binary complex from Agaricus bisporus. J. Biol. Chem. Acknowledgments 276, 27555–27561.

We would like to thank Drs. Kathryn L. Kavanagh and Udo Opper- Jeffries, T., and Jin, Y. (2004). Metabolic engineering for improved mann for helpful discussions. This work was supported by NIH grant fermentation of pentoses by yeasts. Appl. Microbiol. Biotechnol. GM66135 to D.K.W. A.H.E. was funded by a President’s Undergrad- 63, 495–509. uate Fellowship and is currently supported by a Munger, Pollack, Jones, T.A., Zou, J.-Y., Cowan, S.W., and Kjeldgaard. (1991). Im- Reynolds, Robinson, Smith, and Yoedicke Stanford Graduate Fel- proved methods for building protein models in electron-density Xylitol Dehydrogenase Crystal Structure 575

maps and the location of errors in these models. Acta Crystallogr. A dans and improved production of xylitol from D-arabitol. Biosci. 47, 110–119. Biotechnol. Biochem. 67, 584–591. Kavanagh, K.L., Klimacek, M., Nidetzky, B., and Wilson, D.K. (2003). Thomas, J.L., Duax, W.L., Addlagatta, A., Brandt, S., Fuller, R.R., and The structure of xylose reductase bound to NAD+ and the basis for Norris, W. (2003). Structure/function relationships responsible for single and dual cosubstrate specificity in family 2 aldo-keto reduc- coenzyme specificity and the isomerase activity of human type 1 tases. Biochem. J. 373, 319–326. 3b-hydroxysteroid dehydrogenase/isomerase. J. Biol. Chem. 278, Kissinger, C.R., Gehlhaar, D.K., and Fogel, D.B. (1999). Rapid auto- 35483–35490. mated molecular replacement by evolutionary search. Acta Crystal- Watanabe, S., Kodaki, T., and Makino, K. (2004). Complete reversal logr. D Biol. Crystallogr. 55, 484–491. of coenzyme specificity of xylitol dehydrogenase and increase of Kraulis, P.J. (1991). MOLSCRIPT: a program to produce detailed and thermostability by the introduction of structural zinc. J. Biol. schematic plots of protein structures. J. Appl. Crystallogr. 24, 946– Chem. 280, 10340–10349. 950. Zhou, H., and Tai, H.H. (1999). Threonine 188 is critical for interaction + + Kuyper, M., Hartog, M.M., Toirkens, M.J., Almering, M.J., Winkler, with NAD in human NAD -dependent 15-hydroxyprostaglandin de- A.A., van Dijken, J.P., and Pronk, J.T. (2005). Metabolic engineering hydrogenase. Biochem. Biophys. Res. Commun. 257, 414–417. of a xylose-isomerase-expressing Saccharomyces cerevisiae strain for rapid anaerobic xylose fermentation. FEMS Yeast Res. 5, 399– 409. Accession Numbers Laskowski, R.A., MacArthur, M.W., Moss, D.S., and Thornton, J.M. Coordinates have been deposited in the Protein Data Bank under ID (1993). PROCHECK: a program to check the stereochemical quality code 1ZEM. of protein structures. J. Appl. Crystallogr. 26, 283–291. Lee, J., Koo, B., and Kim, S. (2003). Cloning and characterization of the xyl1 gene, encoding an NADH-preferring xylose reductase from Candida parapsilosis, and its functional expression in Candida tropi- calis. Appl. Environ. Microbiol. 69, 6179–6188. Liu, Y., Thoden, J.B., Kim, J., Berger, E., Gulick, A.M., Ruzicka, F.J., Holden, H.M., and Frey, P.A. (1997). Mechanistic roles of tyrosine 149 and serine 124 in UDP-galactose 4-epimerase from Escherichia coli. Biochemistry 36, 10675–10684. Mazza, C., Breton, R., Housset, D., and Fontecilla-Camps, J.C. (1998). Unusual charge stabilization of NADP+ in 17b-hydroxysteroid dehydrogenase. J. Biol. Chem. 273, 8145–8152. Merritt, E.A., and Bacon, D.J. (1997). Raster3D: photorealistic molecular graphics. Methods Enzymol. 277, 505–524. Mulichak, A.M., Theisen, M.J., Essigmann, B., Benning, C., and Ga- ravito, R.M. (1999). Crystal structure of SQD1, an enzyme involved in the biosynthesis of the plant sulfolipid headgroup donor UDP-sulfo- quinovose. Proc. Natl. Acad. Sci. USA 96, 13097–13102. Nakajima, K., Yamashita, A., Akama, H., Nakatsu, T., Kato, H., Hashi- moto, T., Oda, J., and Yamada, Y. (1998). Crystal structures of two tropinone reductases: different reaction stereospecificities in the same protein fold. Proc. Natl. Acad. Sci. USA 95, 4876–4881. Nakanishi, M., Matsuura, K., Kaibe, H., Tanaka, N., Nonaka, T., Mit- sui, Y., and Hara, A. (1997). Switch of coenzyme specificity of mouse lung carbonyl reductase by substitution of threonine 38 with aspartic acid. J. Biol. Chem. 272, 2218–2222. Oppermann, U., Filling, C., Hult, M., Shafqat, N., Wu, X., Lindh, M., Shafqat, J., Nordling, E., Kallberg, Y., Persson, B., and Jornvall, H. (2003). Short-chain dehydrogenases/reductases (SDR): the 2002 update. Chem. Biol. Interact. 143–144, 247–253. Otwinowski, Z., and Minor, W. (1997). Processing of X-ray diffraction data collected in oscillation mode. Methods Enzymol. 276, 307–326. Persson, B., Kallberg, Y., Oppermann, U., and Jornvall, H. (2003). Coenzyme-based functional assignments of short-chain dehydrogenases/reductases (SDRs). Chem. Biol. Interact. 143–144, 271– 278. Petschacher, B., Leitgeb, S., Kavanagh, K., Wilson, D., and Nidetzky, B. (2005). The coenzyme specificity of Candida tenuis xylose reductase (AKR2B5) explored by site-directed mutagenesis and X-ray crystallography. Biochem. J. 385, 75–83. Sogabe, S., Yoshizumi, A., Fukami, T.A., Shiratori, Y., Shimizu, S., Takagi, H., Nakamori, S., and Wada, M. (2003). The crystal structure and stereospecificity of levodione reductase from Corynebacterium aquaticum M-13. J. Biol. Chem. 278, 19387–19395. Somers, W.S., Stahl, M.L., and 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 distinct reactions at the same active site. Structure 6, 1601– 1612. Sugiyama, M., Suzuki, S., Tonouchi, N., and Yokozeki, K. (2003). Cloning of the xylitol dehydrogenase gene from Gluconobacter oxy-