Fine tuning of coenzyme specificity in family 2 aldo-keto reductases revealed by crystal structures of the Lys-274 fi Arg mutant of

FEBS 29220 FEBS Letters 579 (2005) 763 767 Fine tuning of coenzyme specificity in family 2 aldo-keto reductases revealed by crystal structures of the Lys-274 fi Arg mutant of Candida tenuis xylose reductase (AKR2B5) bound to NAD + and NADP + Stefan Leitgeb a,b, Barbara Petschacher a, David K. Wilson b, Bernd Nidetzky a, * a Institute of Biotechnology and Biochemical Engineering, Graz University of Technology, Petersgasse 12/I, A-8010 Graz, Austria b Section of Molecular and Cellular Biology, University of California, Davis, CA 95616, USA Received 5 November 2004; revised 10 December 2004; accepted 15 December 2004 Available online 11 January 2005 Edited by Hans Eklund Abstract Aldo-keto reductases of family 2 employ single site replacement Lys fi Arg to switch their cosubstrate preference from NADPH to NADH. X-ray crystal structures of Lys- 274 fi Arg mutant of Candida tenuis xylose reductase (AKR2B5) bound to NAD + and NADP + were determined at a resolution of 2.4 and 2.3 Å, respectively. Due to steric conflicts in the NADP + -bound form, the arginine side chain must rotate away from the position of the original lysine side chain, thereby disrupting a network of direct and water-mediated interactions between Glu-227, Lys-274 and the cofactor 2 0 -phosphate and 3 0 -hydroxy groups. Because anchoring contacts of its Glu-227 are lost, the coenzyme-enfolding loop that becomes ordered upon binding of NAD(P) + in the wild-type remains partly disordered in the NADP + -bound mutant. The results delineate a catalytic reaction profile for the mutant in comparison to wild-type. Ó 2005 Federation of European Biochemical Societies. Published by Elsevier B.V. All rights reserved. Keywords: Dual coenzyme specificity; Selectivity; Loop ordering; Xylose fermentation 1. Introduction There is a fundamental distinction between NADH and NADPH in metabolic pathways. NADH is the reductant in catabolic reactions, whereas NADPH generally provides reducing power for biosynthesis. Therefore, most nicotinamide cofactor-dependent enzymes are specific for either NADH or NADPH. Xylose reductase (XR) from the yeast Candida tenuis is a dual specific NADPH- and NADHdependent enzyme that converts D-xylose into xylitol [1]. The xylitol is oxidized to yield D-xylulose which in turn is phosphorylated and enters the pentose phosphate pathway (PPP) as D-xylulose 5-phosphate. The PPP is a main source of NADPH and supplies intermediates for energy production and biosynthesis [2]. The co-substrate selectivity of the XR probably reflects the capacity of the PPP to regenerate * Corresponding author. Fax: +43 316 873 8434. E-mail address: bernd.nidetzky@tugraz.at (B. Nidetzky). Abbreviations: AKR, aldo-keto reductase; XR, xylose reductase; CpXR; Candida parapsilosis XR; CtXR, Candida tenuis XR; PPP, pentose phosphate pathway; r.m.s., root mean square the NADPH consumed by XR and the extent to which carbon from xylose partitions between catabolism and anabolism in C. tenuis physiology. Based on sequence homology, C. tenuis XR (CtXR) has been classified into family 2 of the aldo-keto reductase (AKR) superfamily of proteins and enzymes and is systematically identified as AKR2B5 [3]. Family 2 is unusual among the 14 AKR families currently recognized because it includes dual specific NADPH/NADH-dependent reductases as well as NADPH-specific enzymes. Crystal structures of wild-type CtXR bound to NADP + and NAD + show that a conformational change of the side chain of Glu-227, to make strong interactions with the 2 0 -OH and 3 0 -OH groups when the cofactor 2 0 -phosphate group is lacking, is most critical for utilization of NADH [4,5]. Glu-227 of CtXR is widely conserved among the dual specific AKRs of family 2. The side chain of Lys-274 forms a strong hydrogen bond with the 2 0 -phosphate group and participates in a network of water-mediated contacts with Glu-227 and the cofactor 3 0 -hydroxyl group. In the NAD + -bound conformation, the side chain of Lys-274 rotates away slightly and is re-directed to hydrogen bond with solvent molecules. Recent results of mutational studies [6] corroborate the structure-derived notion that 20-fold tighter binding of NADPH than NADH results mainly from the interactions of Lys-274, which are specific to the NADP complex. Lys-274 is conserved in all members of family 2 except for XR from Candida parapsilosis which shows a Lys fi Arg replacement and is the only natural XR known to prefer NADH [7]. A K274R mutant of CtXR utilizes NADH and NADPH with the same specificity, whereas the wild-type prefers NADPH about 33-fold [6]. To provide a molecular basis of how coenzyme selectivity in family 2 AKRs is adjusted by the substitution of Lys-274 with Arg, we determined the X-ray crystal structures of the K274R mutant of CtXR bound to NADP + and NAD +. 2. Materials and methods 2.1. Protein preparation and crystallization All chemicals used were reported previously [6]. The purified K274R mutant [6] was crystallized by using the hanging drop vapor diffusion method at 25 C and a drop consisting of 1 ll well solution and 1 ll protein solution (15 mg/ml) with 2.5 mm cofactor added. Crystals of the NAD + -bound holoenzyme were obtained using a well solution of 2.0 M ammonium sulfate and 100 mm sodium citrate, ph 6.2, in the presence of NADP + the well solution contained additional 160 mm 0014-5793/$30.00 Ó 2005 Federation of European Biochemical Societies. Published by Elsevier B.V. All rights reserved. doi:10.1016/j.febslet.2004.12.063

764 S. Leitgeb et al. / FEBS Letters 579 (2005) 763 767 sodium acetate. Crystals of the NAD + - and NADP + -bound enzymes took 2 and 4 months to grow, respectively, and had final dimensions of approximately 0.5 0.1 0.1 mm. 2.2. Data collection, structure determination and refinement The crystals were flash-cooled in a 100 K nitrogen stream in a buffer containing 80% (v/v) of the well solution, 5 mm coenzyme and 20% (v/ v) glycerol as cryo-protectant. Data were collected at the Stanford Synchrotron Radiation Laboratory beamline 9-1 on a MAR Research 345 detector and were integrated and reduced using the programs DENZO and SCALEPACK [8]. The data collection statistics along with the respective space group and lattice constants are presented in Table 1. Since the mutant enzyme unit cells were isomorphous with the wildtype, rigid body refinement using the program CNS [9] commenced with the respective native structures stripped of NADP + or NAD + and water molecules as starting models (PDB-codes 1K8C and 1MI3). Before any refinement was performed, 5% of the observed reflections were flagged for calculation of the free R-factor. The structures were refined with CNS [9] and manually refitted into the electron density with the program O [10]. Many iterations of this cycle resulted in the final structures. Waters were automatically picked by CNS [9] and manually checked for electron density and appropriate hydrogen bonding. The statistics associated with the final round of refinement are shown in Table 1. Table 1 Statistics for the crystal structures of CtXR K274R mutant bound to NAD + and NADP + K274R-NAD + K274R-NADP + Data collection Space group C2 C2 Unit cell dimensions a (Å) 182.83 180.01 b (Å) 128.67 127.99 c (Å) 80.10 80.07 b ( ) 90.11 90.22 Overall temperature factors (Å 2 ) Monomer A 31.4 27.0 Monomer B 25.7 25.3 Monomer C 29.9 25.1 Monomer D 21.6 21.6 Ramachandran plot [20] Most favoured regions 90.3% 90.8% Additional allowed regions 9.5% 9.1% Generously allowed regions 0.2% Disallowed regions 0.1% Resolution range (Å) 30 2.4 30 2.3 R merge (overall/high 0.080/0.257 0.094/0.347 resolution shell a ) Completeness (overall/high 96.2/95.4 99.7/98.8 resolution shell a ) (%) Mosaicity ( ) 0.96 0.62 I/r(I) (overall/high resolution shell a ) 10.61/3.2 13.4/4.5 Refinement R cryst 0.206 0.192 R free 0.256 0.229 r.m.s. deviation from 0.007 0.006 ideal bond length (Å) r.m.s. deviation from ideal bond angle ( ) 1.3 1.3 a The high resolution shell is 2.49 2.40 Å for the NAD + -bound and 2.38 2.30 Å for the NADP + -bound structure. 2.3. Steady-state kinetic measurements Initial rates of NADP + - and NAD + -dependent oxidations of xylitol catalyzed by K274R were recorded spectrophotometrically (340 nm) at 25 C in 50 mm potassium phosphate buffer, ph 7.0. Xylitol was varied in seven concentrations between 48 and 2400 mm, and NAD + or NADP + was present at several constant concentrations between 1 and 300 lm. The concentration of K274R was approximately 3 lm in all assays. NADP + -dependent wild-type rates were obtained under identical conditions using similar reactant concentrations. Data fitting was performed as described elsewhere [6]. 3. Results and discussion 3.1. Kinetic consequences of replacing Lys-274 by arginine Kinetic parameters of K274R-catalyzed oxidation of xylitol by NAD + and NADP + are summarized in Table 2 along with previously determined kinetic parameters of K274R for the reduction of xylose by NADH and NADPH [6]. Table 2 also shows the relevant wild-type parameters. For a comparison of the wild-type and the mutant, the ratio of K i values which is a measure of the selectivity of binding is most informative. In the wild-type, apparent binding of NADPH and NADP + was about 19- and 39-fold tighter than binding of NADH and NAD +, respectively. Ratios of K i values for the K274R mutant were 1.33 (=K inadh /K inadph ) and 7.3 (=K inad+ / K inadp+ ). Therefore, replacement of Lys-274 by an arginine caused a change in NADP versus NAD binding selectivity for reduced and oxidized coenzymes by factors of 14 (=19/ 1.33) and 5.3 (=39/7.3), respectively. Comparison of catalytic efficiencies for reactions with NADP(H) and NAD(H), indicated in Table 2, shows that the wild-type preferred NADPH and NADP + in the directions of xylose reduction (33-fold) and xylitol oxidation (23-fold). The mutant displayed a slight preference for reactions with NADH (1.5-fold) and NAD + (2-fold). With the exception of k cat for NADPH-dependent reduction of xylose (discussed later), turnover numbers of K274R and wild-type were identical within experimental error. In the absence of a catalytic reaction profile for xylose reduction by CpXR, comparison of the literature data [7] and our results must be qualitative. It indicates however that the Lys- 274 fi Arg replacement induced kinetic properties in the mutant that most clearly distinguish the NADH-preferring CpXR from CtXR wild-type. In CpXR and K274R, apparent binding affinity for NADH seems to be gained partly at the expense of affinity for NADPH; and the K m for xylose in the NADPHlinked reaction is strongly increased, compared to CtXR. 3.2. Overall structures of K274R mutant The K274R mutant bound to NADP + or NAD + folds into the (b/a) 8 barrel previously described for wild-type CtXR [4,5]. There were four K274R molecules, two enzyme dimers, within the respective crystallographical asymmetric units. Numbering starts with 1 for the initiator methionine. Results are summarized in Table 1. 3.3. Structure of K274R mutant bound to NAD + The NAD + -bound structure of the mutant showed well-defined electron density around residues 4 322 predicted from the protein and the NAD + coenzyme molecules. A total of 681 ordered waters were observed in the final structure of K274R but none was placed at a position possibly important for the enzyme function. The root mean square (r.m.s.) devia-

S. Leitgeb et al. / FEBS Letters 579 (2005) 763 767 765 Table 2 Kinetic characterization of K274R mutant Oxidation Reduction NAD + NADP + NADH NADPH K ma (lm) 8 a (27) 4 a (9) 21 (38) 5 a (3) K ia (lm) 44 a (195) 6 (5 b ) 4 c (19) 3 d (1 d ) k cat (s 1 ) 0.93 (1.07) 0.91 (0.79) 8 (11) 26 (13) k cat /(K mb K ia ) e (s 1 mm 2 ) 0.113 (0.026) 0.062 (0.59) 37 (4) 24 (135) The standard deviations (S.D.) were <15%, unless indicated: a <25%; b <30%; c <60%; and d <40%. Values in parentheses are for the wild-type [1,6], except for the NADP + -dependent oxidation of xylitol (this work). k cat is the turnover number; K ia is the apparent coenzyme dissociation constant; and K ma and K mb are the Michaelis constants for coenzyme and substrate, respectively; e catalytic efficiency. tions between the Ca values of the models ranged from 0.23 to 0.30 Å, indicating that the four K274R monomers A D in the asymmetric unit were virtually identical. Since the overall temperature factor for molecule D was the lowest, measurements and conclusions in regard to specific atomic interactions were drawn from it. All of the non-hydrogen atoms in the NAD + were unambiguously observed in the final unbiased electron density map (Fig. 1A) with refined B factors for individual coenzyme atoms that ranged from 13 to 25 Å 2. Binding of the NAD + coenzyme to K274R occurred in a manner indistinguishable to that seen for the wild-type. Like Lys-274 of wild-type CtXR, Arg-274 does not engage in hydrogen bonding with NAD + (Fig. 2A). Its guanidinium group is P4.6 Å away from the adenosine 3 0 -hydroxyl. Mutation of Lys-274 did not cause significant structural changes in and around (60.25 Å) the active site. The significant kinetic consequences in the mutant regarding reactions with NADH and NAD + (see Table 2) must therefore reflect the ability of CtXR to accommodate the site-specific replacement through multiple conformational changes, which are too small to be resolved in a structural comparison of the wild-type and the mutant. 3.4. Structure of K274R mutant bound to NADP + Superpositioning of the models showed that the largest deviations occurred at the binding site of the adenosine moiety of the coenzyme. In particular, monomers A and C were different from monomers B and D. The final unbiased high-resolution electron density maps revealed that NAD + instead of NADP + was present in both monomers B and D. We have no way of Fig. 1. Unbiased F o F c electron densities contoured at 3.0 r of the cofactor adenosine moiety and Arg-274 in the NAD + -bound (A) and NADP + - bound mutant enzyme (B). The maps were derived from the refined models with the atoms of NAD(P) + and the side chain of Arg-274 removed. Figures were generated with MOLSCRIPT [16], BOBSCRIPT [17] and RASTER3D [18].

766 S. Leitgeb et al. / FEBS Letters 579 (2005) 763 767 Fig. 2. Stereoviews of overlap of NAD + -bound structures (A) and NADP + -bound structures (B) of wild-type (yellow) and K274R (blue), illustrating the adenosine binding pocket. Hydrogen bonds are indicated as dashed lines and are shown for the mutant structures. The distances are given in Å, with wild-type values shown in parentheses. The overlays were made with the program EdPDB [19]. knowing the true source of NAD + in the experiment but conversion of NADP + may have occurred during the long time needed for crystallization. By contrast, the 2 0 -phosphate group of NADP + was unambiguously observed in structures of monomers A and C. The r.m.s. deviation of Ca values between these two monomers in the asymmetric unit was 0.21 Å. Of the two, monomer C had the lower overall temperature factor and therefore was chosen for further analysis. The final structure of CtXR K274R bound to NADP + contained 760 ordered water molecules. Important differences in interactions are noted when the adenosine moiety and its 2 0 -phosphate group is examined and compared to what is observed in structures of holo forms of the wild-type and the NAD + -bound structure of K274R. As in the wild-type apo enzyme [4], loop 7 was partly disordered in the K274R mutant bound to NADP +. In the wildtype, loop 7 becomes ordered upon binding of NADP + and is held in place by multiple interactions of Glu-227 and hydrogen bonds between Ser-224 and the juxtaposed Lys-25, which is contributed by a short turn connecting b-strand 1 and a- helix 1. The side chain of Glu-227 is part of a network of water-mediated contacts that is centered around the side chain of Lys-274 and also involves the 3 0 -hydroxy group of NADP +. The side chain of Lys-274 engages in a charged hydrogen bond (2.68 Å) with an oxygen of the 2 0 -phosphate group. All the interactions just discussed were disrupted in NADP + -bound K274R, clearly because of steric conflicts for the bulky arginine side chain. The conformation available to Arg-274 positions its guanidinium group 3.06 Å away from the 2 0 -phosphate (Fig. 2B). The distance between the hydroxyl of Ser-275 and the phosphate oxygen is longer in the NADP + -complex of K274R (2.91 Å) than in the same complex of the wild-type (2.72 Å). The backbone amide of Asn-276 is moved away slightly from its position in the wild-type, thereby increasing its original distance to the 2 0 -phosphate (2.71 Å) to a value of 2.85 Å. The distance between the phosphate oxygen and the side chain of Arg-280 is however shortened significantly in the mutant complex (2.66 Å) in comparison to the wild-type complex (2.75 Å). Summarizing, these observations suggest that the binding selectivity NADP + versus NAD + will be decreased in the mutant in comparison to the wild-type, in good agreement with the experimental ratios K inad+ /K inadp+ of 39 and 7.3 for wild-type and K274R, respectively. Interestingly, K inadp+ for K274R is not strongly increased, compared to the wild-type value. 3.5. Conformational changes involved in coenzyme binding and release Previous structures of wild-type CtXR [4,5] have shown that loop 7, which closes over the bound NAD(P) + in the holoenzyme, must undergo a conformational change to enable release of the coenzyme. The rearrangement of the loop from the closed to the open position has been correlated with a kinetically observed isomerization of the NAD + -bound enzyme, which is the main rate-limiting step of NADH-dependent reduction of xylose by wild-type CtXR [11]. Aside from the contacts made by Glu-227, the interactions between the side chains of Lys-25 and Ser-224 stabilize the closed conformation. In the NAD + -bound structure of the wild-type, the hydroxy group and backbone carbonyl oxygen of Ser-224 are

S. Leitgeb et al. / FEBS Letters 579 (2005) 763 767 767 2.56 and 2.78 Å, respectively, away from the e-amino group of Lys-25. The corresponding distances in K274R bound to NAD + are similar and 2.91 and 2.66 Å, respectively. An overlay of the structures revealed the side chains of Lys-25 and Ser-224 to be almost superimposable in the two enzymes. These results suggest that the NAD + release rates should be similar for reactions catalyzed by K274R and the wild-type, in accordance with experimental turnover numbers for NADH-dependent reduction of xylose. The disorder seen in loop 7 in the NADP + -bound structure of K274R implies that the interactions between Lys-25 and Ser-224 are lost in this complex. The inability of K274R to anchor the coenzyme-enfolding loop in the closed position when NADP + is bound should lead to a faster net rate of NADP + release in the mutant, compared to wild-type. This explains the unusually high k cat for NADPH-dependent reduction of xylose by K274R. Also consistent with the proposed scenario, the turnover numbers for xylitol oxidation by wild-type and the mutant are identical, irrespective of whether NAD + or NADP + is used as a coenzyme. Because hydride transfer is the rate-limiting step of xylitol conversion by the wild-type [11], one does not expect that the k cat changes in that direction as a result of the mutation, considering that replacement Lys- 274 fi Arg affects mainly the steps leading to and from the binary complexes. Summarizing, this work provides a molecular basis of how the naturally occurring replacement Lys-274 fi Arg alters the coenzyme selectivity of family 2 AKRs towards improved acceptance of NADH, relative to NADPH. Lys-274 is positionally conserved across 5 of the current 14 AKR families, and several AKR structures outside family 2 confirm that the function of Lys-274 in hydrogen bonding with the co-substrate 2 0 -phosphate group is also conserved [12 15]. Therefore, this underscores the general relevance of our findings. The coordinates for the K274R mutant bound to NAD + and NADP + have been deposited into the Protein Data Bank under the ID codes 1YE4 and 1YE6, respectively. Acknowledgments: Financial support is acknowledged from the Austrian Science Funds (FWF project 15208 to B.N.), the National Institutes of Health (GM66135 to D.K.W.), and the Keck Foundation (D.K.W.). S.L. acknowledges support from a travel grant from Graz University of Technology. The data collection facilities at Stanford Synchrotron Radiation Laboratory are funded by the U.S. Department of Energy and the National Institutes of Health. We thank Dr. Ulrike Wagner (Department of Chemistry, University of Graz) for support with the preparation of the figures. References [1] Neuhauser, W., Haltrich, D., Kulbe, K.D. and Nidetzky, B. (1997) NAD(P)H-dependent aldose reductase from the xyloseassimilating yeast Candida tenuis. Isolation, characterization and biochemical properties of the enzyme. Biochem. J. 326, 683 692. [2] Bruinenberg, P.M., de Bot, P.H.M., van Dijken, J.P. and Scheffers, W.A. (1983) The role of redox balances in the anaerobic fermentation of xylose by yeasts. Eur. J. Appl. Microbiol. Biotechnol. 18, 287 292. [3] Jez, J.M. and Penning, T.M. (2001) The aldo-keto reductase (AKR) superfamily: an update. Chem. Biol. Interact. 130 132, 499 525. [4] Kavanagh, K.L., Klimacek, M., Nidetzky, B. and Wilson, D.K. (2002) The structure of apo and holo forms of xylose reductase, a dimeric aldo-keto reductase from Candida tenuis. Biochemistry 41, 8785 8795. [5] Kavanagh, K.L., Klimacek, M., Nidetzky, B. and Wilson, D.K. (2003) Structure of xylose reductase bound to NAD + and the basis for single and dual co-substrate specificity in family 2 aldoketo reductases. Biochem. J. 373, 319 326. [6] Petschacher, B., Leitgeb, S., Kavanagh, K.L., Wilson, D.K. 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. [7] Lee, J.K., Koo, B.S. and Kim, S.Y. (2003) Cloning and characterization of the xyl1 gene, encoding an NADH-preferring xylose reductase from Candida parapsilosis, and its functional expression in Candida tropicalis. Appl. Environ. Microbiol. 69, 6179 6188. [8] Otwinowski, Z. and Minor, W. (1997) Processing of X-ray diffraction data collected in oscillation model. Methods Enzymol. 276, 307 326. [9] Brunger, A.T., et al. (1998) Crystallography & NMR system: a new software suite for macromolecular structure determination. Acta Crystallogr., Part D 54, 905 921. [10] Jones, T.A., Zou, J.Y., Cowan, S.W. and Kjeldgaard, M. (1991) Improved methods for building protein models in electron density maps and the location of errors in these models. Acta Crystallogr., Part A 47, 110 119. [11] Nidetzky, B., Klimacek, M. and Mayr, P. (2001) Transientstate and steady-state kinetic studies of the mechanism of NADH-dependent aldehyde reduction catalyzed by xylose reductase from the yeast Candida tenuis. Biochemistry 40, 10371 10381. [12] Ye, Q., Hyndman, D., Green, N.C., Li, L., Jia, Z. and Flynn, T.G. (2001) The crystal structure of an aldehyde reductase Y50F mutant NADP complex and its implications for substrate binding. Chem. Biol. Interact. 130 132, 651 658. [13] Wilson, D.K., Bohren, K.M., Gabbay, K.H. and Quiocho, F.A. (1992) An unlikely sugar substrate site in the 1.65 Å structure of the human aldose reductase holoenzyme implicated in diabetic complications. Science 257, 81 84. [14] Penning, T.M., Bennett, M.J., Smith-Hoog, S., Schlegel, B.P., Jez, J.M. and Lewis, M. (1997) Structure and function of 3 alphahydroxysteroid dehydrogenase. Steroids 62, 101 111. [15] Khurana, S., Powers, D.B., Anderson, S. and Blaber, M. (1998) Crystal structure of 2,5-diketo-D-gluconic acid reductase A complexed with NADPH at 2.1-Å resolution. Proc. Natl. Acad. Sci. USA 95, 6768 6773. [16] Kraulis, P.J. (1991) MOLSCRIPT: a program to produce both detailed and schematic plots of protein structures. J. Appl. Crystallogr. 24, 946 950. [17] Esnouf, R.M. (1997) An extensively modified version of Mol- Script that includes greatly enhanced coloring capabilities. J. Mol. Graph. Model. 15, 132 134, 112 113. [18] Merritt, E.A. and Bacon, D.J. (1997) Raster3D: photorealistic molecular graphics. Methods Enzymol. 277, 505 524. [19] Zhang, X.-J. and Matthews, B.W. (1995) EDPDB: a multifunctional tool for protein structure analysis. J. Appl. Crystallogr. 28, 624 630. [20] Laskowski, R.A., MacArthur, M.W., Moss, D.S. and Thornton, J.M. (1993) PROCHECK: a program to check the stereochemical quality of protein structures. J. Appl. Crystallogr. 26, 283 291.