- Email: [email protected]
The crystal structure of the GCY1 protein from S. cerevisiae suggests a divergent aldo-keto reductase catalytic mechanism
Chemico-Biological Interactions 130 – 132 (2001) 527 – 536
www.elsevier.com/locate/chembiont
The crystal structure of the GCY1 protein from S. cere6isiae suggests a divergent aldo-keto reductase catalytic mechanism Eugene Hur, David K. Wilson * Section of Molecular and Cellular Biology, Uni6ersity of California, One Shields A6enue, Da6is, CA 95616, USA
Abstract The crystal structure of the GCY1 gene product from Saccharomyces cere6isiae has been determined to 2.5 A, and is being refined. The model includes two protein molecules, one apo and one holo, per asymmetric unit. Examination of the model reveals that the active site surface is somewhat flat when compared with the other aldo-keto reductase structures, possibly accommodating larger substrates. The Km for NADPH (28.5 mM) is higher than that seen for other family members. This can be explained structurally by the lack of the ‘safety belt’ of residues seen in other aldo-keto reductases with higher affinity for NADPH. Catalysis also differs from the other aldo-keto reductases. The tyrosine that acts as an acid in the reduction reaction is flipped out of the catalytic pocket. This implies that the protein must either undergo a conformational change before catalysis can take place or that there is an alternate acid moiety. © 2001 Elsevier Science Ireland Ltd. All rights reserved. Keywords: Aldo-keto reductase; Crystal structure; Gcy1p
1. Introduction Aldo-keto reductases (AKRs) are a large group of mostly NAD(P)H-dependent enzymes currently consisting of 112 recognized proteins in 13 families that are believed to be a result of divergent evolution [1]. Members of families 1, 5 and 6 have been crystallographically studied [2–9]. These structures indicate that the * Corresponding author. Fax: + 1-530-7523085. E-mail address: [email protected] (D.K. Wilson). 0009-2797/01/$ - see front matter © 2001 Elsevier Science Ireland Ltd. All rights reserved. PII: S 0 0 0 9 - 2 7 9 7 ( 0 0 ) 0 0 2 9 6 - 9
528
E. Hur, D.K. Wilson / Chemico-Biological Interactions 130–132 (2001) 527–536
proteins fold in a (b/a)8 barrel, bind cosubstrate in a similar manner and appear to share a common catalytic mechanism. In order to determine the effects of sequence divergence on the structure and function of AKRs, we initiated a kinetic and crystallographic study of Gcy1p (AKR3A1) from Saccharomyces cere6isiae, a member of family 3. The current structure also provides an understanding of the protein’s unusual catalytic mechanism and the nature of its substrate and cosubstrate binding sites. Prior studies of GCY1 indicated that it is a transcribed gene encoding a 1300 base pair mRNA encoding a 312 amino acid protein bearing 59% sequence homology with frog r-crystallin [10]. It was later determined that expression was upregulated by galactose and repressed by glucose; however, mutants showed no obvious phenotype under normal conditions [11,12]. Expression is also enhanced by the DNA-alkylating agent methyl methanesulfonate [13]. Assays using enzyme contained in crude cell extracts produced from bacteria and yeast expressing recombinant Gcy1p suggested that the protein did have NADPH-dependent aldoketo reductase activity using a number of typical AKR carbonyl substrates [11]. The chemical mechanism involved in the AKR-catalyzed reduction of a carbonyl to its corresponding alcohol involves the transfer of a hydride from the NAD(P)H co-substrate directly to the carbonyl carbon and the subsequent or concerted abstraction of a proton from an acidic moiety, presumably on the protein. Crystal structures of the AKR holoenzymes, aldose reductase [3], aldehyde reductase [5], 3a-hydroxysteroid dehydrogenase [4], FR-1 [6], 2,5-diketo-gluconic acid reductase [7], the potassium channel b subunit [8] and CHO reductase [9] are all consistent with the role of a conserved tyrosine as the general acid. Direct mutational evidence in aldose reductase also appears to confirm this. Changing the tyrosine to phenylalanine in aldose reductase abolishes all catalytic activity. Moreover, mutating residues implicated in perturbing the pKa of this tyrosine, a necessity for optimum catalytic activity, have lesser effects [14].
2. Methods The intron-less GCY1 gene was initially polymerase chain reaction-amplified from S. cere6isiae genomic DNA using the primers 5%-CCCATATGCCTGCTACTTTACATGAT and 5%-TCCCCCGGGCTTGAATACTTCGAAAGGAGACCA, and either Taq or VENT polymerase (New England Biolabs). The resulting insert was placed into the pCR-BLUNT vector (Invitrogen) and transferred to the Escherichia coli expression vector pTYB2 (New England Biolabs) using NdeI and SmaI restriction sites. This vector fuses an intein domain and a chitin-binding domain to the target protein for affinity purification and cleavage of the target protein from the column. The resulting construct was used to transform the expression strain BL21 (DE3). The initial clone was created from an insert that contained a double mutation (Pro268 Leu, Ser281Phe), which was a result of a Taq polymerase error. After sequencing indicated mutations were present, the insert was generated again using VENT, a proofreading polymerase, and inserted
E. Hur, D.K. Wilson / Chemico-Biological Interactions 130–132 (2001) 527–536
529
into the expression vector using similar methods. Cells were grown in the presence of 100 mg/ml ampicillin to an optical density at 600 nm of 0.6 and induced with 750 mM IPTG for a period of 16 h. Cells were harvested, resuspended in 20 mM Tris, 0.5 M NaCl, 0.5 mM ethylenediamine tetraacetic acid, 0.1% (v/v) Triton X-100 (pH 7.5) and lysed using a microfluidizer (Microfluidics Inc.). The supernatant was applied to a 20 ml chitin column at a rate of 1 ml/min and then washed with large volumes of lysis buffer until baseline was reached. Detergent was then removed by a further wash with the identical buffer minus Triton X-100. Intein-mediated cleavage was initiated by adding 45 mM 2-mercaptoethanol to the detergent-free buffer, quickly passing approximately one column volume over the column and incubating overnight. Eluted protein was pooled, concentrated and desalted. This was applied to a POROS CM cation exchange column on a Perkin Elmer Biocad Sprint system at pH 7.5. A 0 – 600 mM NaCl gradient was run and Gcy1p was found as the only major elution peak. These fractions were pooled, concentrated to 20 mg/ml with a tenfold excess of NADP+ and the buffer changed to 10 mM HEPES, 24 mM NaCl, 60 mM 2-mercaptoethanol (pH 7.4) for use in crystallization and kinetic experiments. Kinetic experiments were carried out by measuring oxidation of NADPH to NADP+ at 340 nm on either a Shimadzu UV160U or Hewlett Packard 8453 spectrophotometer. Km and kcat values were determined by fitting data to the hyperbolic form of the Michaelis–Menten equation using KALEIDAGRAPH version 3.09. Crystallization was carried out according to a previously described protocol [15]. Briefly, hanging drop vapor diffusion experiments were set up using a 2 ml drop containing 10 mg/ml Gcy1p, 14% (w/v) PEG 8000, 0.1 M ammonium sulfate, 50 mM sodium citrate (pH 5.5) over a well containing 28% (w/v) PEG 8000, 0.2 M ammonium acetate, 0.1 M sodium citrate (pH 5.5). After reaching full size, the crystals were placed in Paratone-N oil, solvent was removed and the mounted crystal was cooled in a 100 K cold stream for data collection. Frames were collected, and unit cell parameters and space group were determined on a Rigaku R-AXIS IV using the DENZO package of programs [16]. Data collection statistics are shown in Table 1. The structure was determined using the molecular replacement method as implemented in AMORE [17]. The structure of human aldose reductase holoenzyme [3] with cosubstrate removed was used as a search object. One clear rotation peak was found (11.9s). Using this rotation peak, two translation peaks were found. Initial rigid body refinement reduced the Rcryst from 53.1 to 49.3% for 20–2.5 A, resolution data. Individual amino acids were changed from the aldose reductase sequence to Gcy1p and loops were inserted and removed as appropriate to yield the correct Gcy1p sequence. Additionally, one NADP+ molecule was inserted into one of the subunits. Subsequent iterations of refinement using CNS and manual refitting have reduced the Rcryst and Rfree to 26.7 and 32.7%, respectively. Current refinement statistics are shown in Table 1.
E. Hur, D.K. Wilson / Chemico-Biological Interactions 130–132 (2001) 527–536
530
Table 1 Crystallographic parameters and results of refinementa Space group Unit cell Resolution Vm (assuming two molecules/asymmetric unit) Reflections/observations Completeness I/s(I) Rmerge
P21 a = 50.94 A, , b =65.64 A, , c= 86.23 A, ; b =92.64° 100–2.49 A, (2.59–2.49 A, ) 2.06 A, 3 Da−1 19 639/59 321 97.1% (80.7%) 22.4 (6.0) 5.4% (16.8%)
Atoms in model Protein NADP+ Waters
4512 48 0
Rcryst Rfree
26.7% 32.7%
a
Values in parentheses are for the highest resolution shell.
3. Results and discussion The majority of the NADPH-binding residues and all of the catalytic residues were conserved when compared with other aldo-keto reductases, which suggested that the protein was likely to function as an enzyme. Experiments using crude cellular extracts from yeast and bacteria overexpressing Gcy1p also suggested that it was an enzymatically active protein [11]. Using a recombinant, inadvertent double mutant protein (Pro268 Leu, Ser281 Phe), kinetic constants were determined for several typical AKR substrates (Table 2). After sequencing the expression construct, the protein was found to in fact be a mutant and experiments were repeated later using wild-type enzyme. These demonstrated that the mutations did not significantly affect the catalytic properties of the enzyme. Results from the kinetic experiments showed that Gcy1p is a NADPH-specific enzyme. No activity was measurable when NADH was used as a cosubstrate. Furthermore, the enzyme exhibited AKR activity versus several substrates, albeit with lower catalytic effiTable 2 Kinetic data for various Gcy1p substrates Substrate
Km
kcat (s−1)
D,L-Glyceraldehyde
11.3 mM 20.0 mM 144 mM 28.5 mM None
3.83 1.50 2.3 4.57 None
Glucuronatea Proprionaldehydea NADPH NADH, galactose a
Values obtained from Pro268Leu, Ser281Phe double mutant.
E. Hur, D.K. Wilson / Chemico-Biological Interactions 130–132 (2001) 527–536
Fig. 1. (Continued)
531
532
E. Hur, D.K. Wilson / Chemico-Biological Interactions 130–132 (2001) 527–536
ciency when compared with other AKRs (Table 2). Interestingly, despite the fact that it is transcriptionally induced by galactose, no activity was detected when galactose was used as a substrate. The current model of Gcy1p derived from X-ray diffraction consists of two molecules of protein, one in the apo form and one in the hapo form (Fig. 1). The holo molecule is much better ordered than the epo form. Currently, 297 residues are included in the holo model and 277 residues are in the apo model. The majority of missing chains is in loop regions, which have also been difficult to locate in other aldo-keto reductases. Looking at the Ca traces at a gross level, both molecules show little deviation from the AKR fold. One slight difference is seen at the amino terminus of the protein. Gcy1p has seven extra residues relative to human aldose reductase. Structurally, these residues are present as a meander that packs next to the a8 helix and the strand connecting a8 with the carboxyl terminal helix H2. These residues precede the b hairpin that initiates the other aldo-keto reductases. Since they reside in the amino terminal region of the barrel, these residues are not likely to affect catalysis. Another feature that differs from other AKRs is the added length in loop 4 (between b4 and a4). This loop in human aldose reductase is approximately 26 residues compared with the 34 seen in Gcy1p. The functional significance of this extension is not clear. Another significant difference between Gcy1p and aldose reductase is seen in loop 7, which is reduced in length from 12 residues in aldose reductase to four residues in Gcy1p (Fig. 2). The net effect gained from the variation in loop lengths is a substrate-binding site without the pronounced invagination observed in other AKR structures. This active site could conceivably accommodate a much larger substrate than other AKRs. The open nature of the active site also offers, in part, an explanation as to why Gcy1p is not able to reduce smaller substrates as efficiently as other AKRs. As with other AKRs, the NADPH cosubstrate binds in an extended conformation across the carboxyl terminal end of the b barrel, protruding between b/a repeats 7 and 8. Assays with NADH indicated that the enzyme has an absolute requirement for NADPH. This specificity is conferred by salt links and main chain interactions with the 2% phosphate, analogous to those seen in aldose reductase. In the case of Gcy1p, these are provided by the guanidinium group of Arg-270 and hydrogen bonding with the polar main chain amide nitrogens in residues 266 and 267. One structural feature believed to be key in the high affinity of aldose reductase and other AKRs for NAD(P)H is the ‘safety belt’ loop of residues seen to fold over Fig. 1. An alpha carbon trace of the two Gcy1p molecules in the asymmetric unit. The holo molecule is in the upper left and the apo molecule is in the lower right. Both are in similar orientations, looking down the barrel. The chains are colored by the temperature factors of the underlying Ca atoms from blue (B15 A, 2) to red (\ 50 A, 2). Figures of molecular models were produced using the programs MOLSCRIPT [20] and RASTER3D [21]. Fig. 2. Overlap of aldo-keto reductase holo structures. Models are colored according to the following scheme: Gcy1p, White; 3a-hydroxysteroid dehydrogenase, green; aldose reductase, blue; aldehyde reductase, red; FR-1, orange; 2,5-diketo-gluconic acid reductase, purple.
E. Hur, D.K. Wilson / Chemico-Biological Interactions 130–132 (2001) 527–536
Fig. 3. (Continued)
533
534
E. Hur, D.K. Wilson / Chemico-Biological Interactions 130–132 (2001) 527–536
the cosubstrate after binding. This loop is held closed by the salt bridge between Asp-216 on one side and Lys-21 and Lys-262 on the other. A conformational change in this loop is required for the release of the consumed cosubstrate and is believed to be the rate-limiting step in the reaction [18]. Gcy1p completely lacks this loop (Fig. 3), a feature also seen in 2,5-diketo-D-gluconic acid reductase A [7]. It is likely that the reduced affinity for NADPH when compared with aldose reductase (28.5 versus 2 mM) is due to the absence of this loop [19]. Examination of the active site reveals substantial differences when compared with the aldose reductase active site and those of other AKRs. Although the main chain trace is similar, two key amino acid side chains have altered positions, requiring a conformational change to take place in order for catalysis to take place. The first is Trp-28, analogous to Trp-20 in human aldose reductase [3]. The side chain is positioned over the nicotinamide ribose in aldose reductase. In Gcy1p, the side chain stacks on the nicotinamide ring, preventing access to the ring by even the smallest substrates. The second structural perturbation involves Tyr-56, the residue that functions as the general acid in other AKRs (Tyr-48 in human aldose reductase). The pKa of the phenolic hydrogen is depressed in other AKRs by its proximity to a lysine (Lys-77 in human aldose reductase), which in turn is engaged in a salt bridge to an aspartate (Asp-43 in human aldose reductase). Gcy1p preserves the positions of the lysine and aspartate but the conformation of the loop consisting of residues 52–59 and containing the catalytic tyrosine is changed (Fig. 4). When compared with the human aldose reductase structure, the tyrosine Ca position is displaced by more than 4 A, . A rigid movement of this loop would bring the tyrosine into position to hydrogen bond with Lys-81 and allow catalysis to occur. Whether the conformational changes involving Trp-28 leaving the active site and the tyrosine entering are concerted is not yet clear. It is necessary for the tryptophan to move, however, to provide room for the tyrosine.
4. Conclusions Gcy1p, a yeast protein with homology to AKRs, including catalytic and cosubstrate-binding residues, has been shown to possess NADPH-specific reductase activity. The structure of Gcy1p, currently being refined at 2.5 A, , has revealed similarities and differences with other AKR structures. As expected, the overall Fig. 3. Residues 21, 216 and 262 form a bidentate interaction (yellow bonds), closing a loop or ‘safety belt’ of residues over the top of the NADPH cosubstrate (blue CPK model) in aldose reductase (shown in blue). No such loop exists in Gcy1p (white), possibly accounting for its lower affinity for NADPH. Fig. 4. A conformational rearrangement of active site residues is seen in Gcy1p (white) when compared with the human aldose reductase structure (blue). Trp-28 in Gcy1p (Trp-20 in aldose reductase) is rotated so that it stacks on top of the nicotinamide where the carbonyl containing substrate should bind. Tyr-56 (Tyr-48 in aldose reductase), which normally functions as a general acid, is shifted out of the active site. Activating residues such as Lys-81 and Asp-51 are in conserved positions.
E. Hur, D.K. Wilson / Chemico-Biological Interactions 130–132 (2001) 527–536
535
(b/a)8 structure shows little deviation from those of other AKRs. The majority of NADPH binding interactions is also well conserved. A major exception is the complete absence of the belt of residues that fold over the NADPH, creating a higher affinity binding site. Differences are also seen in the substrate-binding site. This region is more open and flatter than similar areas in other AKR structures, suggesting that larger substrate may be able to fit into it. The substrate-binding pocket immediately above the nicotinamide ring is blocked by a tryptophan, which must undergo a conformational change so that the cosubstrate may be accessible to the substrate. Finally, differences are seen in the position of the tyrosine general acid, which also requires a conformation change so that the protein can exist in a catalytically competent configuration. It is not yet clear why the enzyme has evolved this unconventional catalytic mechanism and what implications it may have on substrate binding and other enzymatic properties.
Acknowledgements Many thanks are due to Prof. Mark Petrash and coworkers for collaborating and sharing results before publication. E.H. was supported in part by an undergraduate fellowship from the California Foundation for Biomedical Research. This research was supported by a grant from the NIH to D.W. and a grant from the W.M. Keck Foundation.
References [1] J.M. Jez, T.G. Flynn, T.M. Penning, A nomenclature system for the aldo-keto reductase superfamily, in: H. Weiner, R. Lindahl, D. Crabb, T.G. Flynn (Eds.), Enzymology and Molecular Biology of Carbonyl Metabolism 6, Plenum Press, New York, 1996, pp. 579 – 589. [2] J.M. Rondeau, F. Tete-Favier, A. Podjarny, J.M. Reymann, P. Barth, J.F. Biellmann, D. Moras, Novel NADPH-binding domain revealed by the crystal structure of aldose reductase, Nature 355 (1992) 469–472. [3] D.K. Wilson, K.M. Bohren, K.H. Gabbay, F.A. Quiocho, An unlikely sugar substrate site in the 1.65 A, structure of the human aldose reductase holoenzyme implicated in diabetic complications, Science 257 (1992) 81–84. [4] S.S. Hoog, J.E. Pawlowski, P.M. Alzari, T.M. Penning, M. Lewis, Three-dimensional structure of rat liver 3a-hydroxysteroid/dihydrodiol dehydrogenase: a member of the aldo-keto reductase superfamily, Proc. Natl. Acad. Sci. USA 91 (1994) 2517 – 2521. [5] O. El-Kabbani, K. Judge, S.L. Ginell, D.A.A. Myles, L.J. DeLucas, T.G. Flynn, Structure of porcine aldehyde reductase holoenzyme, Nat. Struct. Biol. 2 (1995) 687 – 692. [6] D.K. Wilson, T. Nakano, J.M. Petrash, F.A. Quiocho, 1.7 A, structure of FR-1, a fibroblast growth factor-induced member of the aldo-keto reductase family, complexed with coenzyme and inhibitor, Biochemistry 34 (1995) 14323–14330. [7] S. Khurana, D.B. Powers, S. Anderson, M. Blaber, Crystal structure of 2,5-diketo-D-gluconic acid reductase A complexed with NADPH at 2.1 A, resolution, Proc. Natl. Acad. Sci. USA 95 (1998) 6768–6773. [8] J.M. Gulbis, S. Mann, R. MacKinnon, Structure of a voltage-dependent K+ channel b subunit, Cell 97 (1999) 943–952.
536
E. Hur, D.K. Wilson / Chemico-Biological Interactions 130–132 (2001) 527–536
[9] Q. Ye, D. Hyndman, X. Li, T.G. Flynn, Z. Jia, Crystal structure of CHO reductase, a member of the aldo-keto reductase superfamily, Proteins 38 (2000) 41 – 48. [10] U. Oechsner, V. Magdolen, W. Bandlow, A nuclear yeast gene (GCY) encodes a polypeptide with high homology to a vertebrate eye lens protein, FEBS Lett. 238 (1988) 123 – 128. [11] V. Magdolen, U. Oechsner, P. Trommler, W. Bandlow, Transcriptional control by galactose of a yeast gene encoding a protein homologous to mammalian aldo/keto reductases, Gene 90 (1990) 105–114. [12] M. Angermayr, W. Bandlow, The type of basal promoter determines the regulated or constitutive mode of transcription in the common control region of the yeast gene pair GCY1/RIO1, J. Biol. Chem. 272 (1997) 31630–31635. [13] S.A. Jelinsky, L.D. Samson, Global response of Saccharomyces cerevisiae to an alkylating agent, Proc. Natl. Acad. Sci. USA 96 (1999) 1486 – 1491. [14] I. Tarle, D.W. Borhani, D.K. Wilson, F.A. Quiocho, J.M. Petrash, Probing the active site of human aldose reductase: site directed mutagenesis of Asp-43, Tyr-48, Lys-77 and His-110, J. Biol. Chem. 268 (1993) 25687–25693. [15] E. Hur, D.K. Wilson, Crystallization and aldo-keto reductase activity of Gcy1p from Saccharomyces cere6isiae, Acta Crystallogr. D56 (2000) 763 – 765. [16] Z. Otwinowski, W. Minor, Processing of X-ray diffraction data collected in oscillation mode, Methods Enzymol. 276 (1997) 307–326. [17] J. Navaza, AMoRe: an automated package for molecular replacement, Acta Crystallogr. A50 (1994) 157–163. [18] C.E. Grimshaw, M. Shahbaz, C.G. Putney, Mechanistic basis for nonlinear kinetics of aldehyde reduction catalyzed by aldose reductase, Biochemistry 29 (1990) 9947 – 9955. [19] K.M. Bohren, J.L. Page, R. Shankar, S.P. Henry, K.H. Gabbay, Expression of human aldose and aldehyde reductases. Site-directed mutagenesis of a critical lysine 262, J. Biol. Chem. 266 (1991) 24031–24037. [20] P.J. Kraulis, MOLSCRIPT: a program to produce both detailed and schematic plots of protein structures, J. Appl. Crystallogr. 24 (1991) 946 – 950. [21] E.A. Merritt, M.E.P. Murphy, RASTER3D Version 2.0 — a program for photorealistic molecular graphics, Acta Crystallogr. D50 (1994) 869 – 873.
.
www.elsevier.com/locate/chembiont
The crystal structure of the GCY1 protein from S. cere6isiae suggests a divergent aldo-keto reductase catalytic mechanism Eugene Hur, David K. Wilson * Section of Molecular and Cellular Biology, Uni6ersity of California, One Shields A6enue, Da6is, CA 95616, USA
Abstract The crystal structure of the GCY1 gene product from Saccharomyces cere6isiae has been determined to 2.5 A, and is being refined. The model includes two protein molecules, one apo and one holo, per asymmetric unit. Examination of the model reveals that the active site surface is somewhat flat when compared with the other aldo-keto reductase structures, possibly accommodating larger substrates. The Km for NADPH (28.5 mM) is higher than that seen for other family members. This can be explained structurally by the lack of the ‘safety belt’ of residues seen in other aldo-keto reductases with higher affinity for NADPH. Catalysis also differs from the other aldo-keto reductases. The tyrosine that acts as an acid in the reduction reaction is flipped out of the catalytic pocket. This implies that the protein must either undergo a conformational change before catalysis can take place or that there is an alternate acid moiety. © 2001 Elsevier Science Ireland Ltd. All rights reserved. Keywords: Aldo-keto reductase; Crystal structure; Gcy1p
1. Introduction Aldo-keto reductases (AKRs) are a large group of mostly NAD(P)H-dependent enzymes currently consisting of 112 recognized proteins in 13 families that are believed to be a result of divergent evolution [1]. Members of families 1, 5 and 6 have been crystallographically studied [2–9]. These structures indicate that the * Corresponding author. Fax: + 1-530-7523085. E-mail address: [email protected] (D.K. Wilson). 0009-2797/01/$ - see front matter © 2001 Elsevier Science Ireland Ltd. All rights reserved. PII: S 0 0 0 9 - 2 7 9 7 ( 0 0 ) 0 0 2 9 6 - 9
528
E. Hur, D.K. Wilson / Chemico-Biological Interactions 130–132 (2001) 527–536
proteins fold in a (b/a)8 barrel, bind cosubstrate in a similar manner and appear to share a common catalytic mechanism. In order to determine the effects of sequence divergence on the structure and function of AKRs, we initiated a kinetic and crystallographic study of Gcy1p (AKR3A1) from Saccharomyces cere6isiae, a member of family 3. The current structure also provides an understanding of the protein’s unusual catalytic mechanism and the nature of its substrate and cosubstrate binding sites. Prior studies of GCY1 indicated that it is a transcribed gene encoding a 1300 base pair mRNA encoding a 312 amino acid protein bearing 59% sequence homology with frog r-crystallin [10]. It was later determined that expression was upregulated by galactose and repressed by glucose; however, mutants showed no obvious phenotype under normal conditions [11,12]. Expression is also enhanced by the DNA-alkylating agent methyl methanesulfonate [13]. Assays using enzyme contained in crude cell extracts produced from bacteria and yeast expressing recombinant Gcy1p suggested that the protein did have NADPH-dependent aldoketo reductase activity using a number of typical AKR carbonyl substrates [11]. The chemical mechanism involved in the AKR-catalyzed reduction of a carbonyl to its corresponding alcohol involves the transfer of a hydride from the NAD(P)H co-substrate directly to the carbonyl carbon and the subsequent or concerted abstraction of a proton from an acidic moiety, presumably on the protein. Crystal structures of the AKR holoenzymes, aldose reductase [3], aldehyde reductase [5], 3a-hydroxysteroid dehydrogenase [4], FR-1 [6], 2,5-diketo-gluconic acid reductase [7], the potassium channel b subunit [8] and CHO reductase [9] are all consistent with the role of a conserved tyrosine as the general acid. Direct mutational evidence in aldose reductase also appears to confirm this. Changing the tyrosine to phenylalanine in aldose reductase abolishes all catalytic activity. Moreover, mutating residues implicated in perturbing the pKa of this tyrosine, a necessity for optimum catalytic activity, have lesser effects [14].
2. Methods The intron-less GCY1 gene was initially polymerase chain reaction-amplified from S. cere6isiae genomic DNA using the primers 5%-CCCATATGCCTGCTACTTTACATGAT and 5%-TCCCCCGGGCTTGAATACTTCGAAAGGAGACCA, and either Taq or VENT polymerase (New England Biolabs). The resulting insert was placed into the pCR-BLUNT vector (Invitrogen) and transferred to the Escherichia coli expression vector pTYB2 (New England Biolabs) using NdeI and SmaI restriction sites. This vector fuses an intein domain and a chitin-binding domain to the target protein for affinity purification and cleavage of the target protein from the column. The resulting construct was used to transform the expression strain BL21 (DE3). The initial clone was created from an insert that contained a double mutation (Pro268 Leu, Ser281Phe), which was a result of a Taq polymerase error. After sequencing indicated mutations were present, the insert was generated again using VENT, a proofreading polymerase, and inserted
E. Hur, D.K. Wilson / Chemico-Biological Interactions 130–132 (2001) 527–536
529
into the expression vector using similar methods. Cells were grown in the presence of 100 mg/ml ampicillin to an optical density at 600 nm of 0.6 and induced with 750 mM IPTG for a period of 16 h. Cells were harvested, resuspended in 20 mM Tris, 0.5 M NaCl, 0.5 mM ethylenediamine tetraacetic acid, 0.1% (v/v) Triton X-100 (pH 7.5) and lysed using a microfluidizer (Microfluidics Inc.). The supernatant was applied to a 20 ml chitin column at a rate of 1 ml/min and then washed with large volumes of lysis buffer until baseline was reached. Detergent was then removed by a further wash with the identical buffer minus Triton X-100. Intein-mediated cleavage was initiated by adding 45 mM 2-mercaptoethanol to the detergent-free buffer, quickly passing approximately one column volume over the column and incubating overnight. Eluted protein was pooled, concentrated and desalted. This was applied to a POROS CM cation exchange column on a Perkin Elmer Biocad Sprint system at pH 7.5. A 0 – 600 mM NaCl gradient was run and Gcy1p was found as the only major elution peak. These fractions were pooled, concentrated to 20 mg/ml with a tenfold excess of NADP+ and the buffer changed to 10 mM HEPES, 24 mM NaCl, 60 mM 2-mercaptoethanol (pH 7.4) for use in crystallization and kinetic experiments. Kinetic experiments were carried out by measuring oxidation of NADPH to NADP+ at 340 nm on either a Shimadzu UV160U or Hewlett Packard 8453 spectrophotometer. Km and kcat values were determined by fitting data to the hyperbolic form of the Michaelis–Menten equation using KALEIDAGRAPH version 3.09. Crystallization was carried out according to a previously described protocol [15]. Briefly, hanging drop vapor diffusion experiments were set up using a 2 ml drop containing 10 mg/ml Gcy1p, 14% (w/v) PEG 8000, 0.1 M ammonium sulfate, 50 mM sodium citrate (pH 5.5) over a well containing 28% (w/v) PEG 8000, 0.2 M ammonium acetate, 0.1 M sodium citrate (pH 5.5). After reaching full size, the crystals were placed in Paratone-N oil, solvent was removed and the mounted crystal was cooled in a 100 K cold stream for data collection. Frames were collected, and unit cell parameters and space group were determined on a Rigaku R-AXIS IV using the DENZO package of programs [16]. Data collection statistics are shown in Table 1. The structure was determined using the molecular replacement method as implemented in AMORE [17]. The structure of human aldose reductase holoenzyme [3] with cosubstrate removed was used as a search object. One clear rotation peak was found (11.9s). Using this rotation peak, two translation peaks were found. Initial rigid body refinement reduced the Rcryst from 53.1 to 49.3% for 20–2.5 A, resolution data. Individual amino acids were changed from the aldose reductase sequence to Gcy1p and loops were inserted and removed as appropriate to yield the correct Gcy1p sequence. Additionally, one NADP+ molecule was inserted into one of the subunits. Subsequent iterations of refinement using CNS and manual refitting have reduced the Rcryst and Rfree to 26.7 and 32.7%, respectively. Current refinement statistics are shown in Table 1.
E. Hur, D.K. Wilson / Chemico-Biological Interactions 130–132 (2001) 527–536
530
Table 1 Crystallographic parameters and results of refinementa Space group Unit cell Resolution Vm (assuming two molecules/asymmetric unit) Reflections/observations Completeness I/s(I) Rmerge
P21 a = 50.94 A, , b =65.64 A, , c= 86.23 A, ; b =92.64° 100–2.49 A, (2.59–2.49 A, ) 2.06 A, 3 Da−1 19 639/59 321 97.1% (80.7%) 22.4 (6.0) 5.4% (16.8%)
Atoms in model Protein NADP+ Waters
4512 48 0
Rcryst Rfree
26.7% 32.7%
a
Values in parentheses are for the highest resolution shell.
3. Results and discussion The majority of the NADPH-binding residues and all of the catalytic residues were conserved when compared with other aldo-keto reductases, which suggested that the protein was likely to function as an enzyme. Experiments using crude cellular extracts from yeast and bacteria overexpressing Gcy1p also suggested that it was an enzymatically active protein [11]. Using a recombinant, inadvertent double mutant protein (Pro268 Leu, Ser281 Phe), kinetic constants were determined for several typical AKR substrates (Table 2). After sequencing the expression construct, the protein was found to in fact be a mutant and experiments were repeated later using wild-type enzyme. These demonstrated that the mutations did not significantly affect the catalytic properties of the enzyme. Results from the kinetic experiments showed that Gcy1p is a NADPH-specific enzyme. No activity was measurable when NADH was used as a cosubstrate. Furthermore, the enzyme exhibited AKR activity versus several substrates, albeit with lower catalytic effiTable 2 Kinetic data for various Gcy1p substrates Substrate
Km
kcat (s−1)
D,L-Glyceraldehyde
11.3 mM 20.0 mM 144 mM 28.5 mM None
3.83 1.50 2.3 4.57 None
Glucuronatea Proprionaldehydea NADPH NADH, galactose a
Values obtained from Pro268Leu, Ser281Phe double mutant.
E. Hur, D.K. Wilson / Chemico-Biological Interactions 130–132 (2001) 527–536
Fig. 1. (Continued)
531
532
E. Hur, D.K. Wilson / Chemico-Biological Interactions 130–132 (2001) 527–536
ciency when compared with other AKRs (Table 2). Interestingly, despite the fact that it is transcriptionally induced by galactose, no activity was detected when galactose was used as a substrate. The current model of Gcy1p derived from X-ray diffraction consists of two molecules of protein, one in the apo form and one in the hapo form (Fig. 1). The holo molecule is much better ordered than the epo form. Currently, 297 residues are included in the holo model and 277 residues are in the apo model. The majority of missing chains is in loop regions, which have also been difficult to locate in other aldo-keto reductases. Looking at the Ca traces at a gross level, both molecules show little deviation from the AKR fold. One slight difference is seen at the amino terminus of the protein. Gcy1p has seven extra residues relative to human aldose reductase. Structurally, these residues are present as a meander that packs next to the a8 helix and the strand connecting a8 with the carboxyl terminal helix H2. These residues precede the b hairpin that initiates the other aldo-keto reductases. Since they reside in the amino terminal region of the barrel, these residues are not likely to affect catalysis. Another feature that differs from other AKRs is the added length in loop 4 (between b4 and a4). This loop in human aldose reductase is approximately 26 residues compared with the 34 seen in Gcy1p. The functional significance of this extension is not clear. Another significant difference between Gcy1p and aldose reductase is seen in loop 7, which is reduced in length from 12 residues in aldose reductase to four residues in Gcy1p (Fig. 2). The net effect gained from the variation in loop lengths is a substrate-binding site without the pronounced invagination observed in other AKR structures. This active site could conceivably accommodate a much larger substrate than other AKRs. The open nature of the active site also offers, in part, an explanation as to why Gcy1p is not able to reduce smaller substrates as efficiently as other AKRs. As with other AKRs, the NADPH cosubstrate binds in an extended conformation across the carboxyl terminal end of the b barrel, protruding between b/a repeats 7 and 8. Assays with NADH indicated that the enzyme has an absolute requirement for NADPH. This specificity is conferred by salt links and main chain interactions with the 2% phosphate, analogous to those seen in aldose reductase. In the case of Gcy1p, these are provided by the guanidinium group of Arg-270 and hydrogen bonding with the polar main chain amide nitrogens in residues 266 and 267. One structural feature believed to be key in the high affinity of aldose reductase and other AKRs for NAD(P)H is the ‘safety belt’ loop of residues seen to fold over Fig. 1. An alpha carbon trace of the two Gcy1p molecules in the asymmetric unit. The holo molecule is in the upper left and the apo molecule is in the lower right. Both are in similar orientations, looking down the barrel. The chains are colored by the temperature factors of the underlying Ca atoms from blue (B15 A, 2) to red (\ 50 A, 2). Figures of molecular models were produced using the programs MOLSCRIPT [20] and RASTER3D [21]. Fig. 2. Overlap of aldo-keto reductase holo structures. Models are colored according to the following scheme: Gcy1p, White; 3a-hydroxysteroid dehydrogenase, green; aldose reductase, blue; aldehyde reductase, red; FR-1, orange; 2,5-diketo-gluconic acid reductase, purple.
E. Hur, D.K. Wilson / Chemico-Biological Interactions 130–132 (2001) 527–536
Fig. 3. (Continued)
533
534
E. Hur, D.K. Wilson / Chemico-Biological Interactions 130–132 (2001) 527–536
the cosubstrate after binding. This loop is held closed by the salt bridge between Asp-216 on one side and Lys-21 and Lys-262 on the other. A conformational change in this loop is required for the release of the consumed cosubstrate and is believed to be the rate-limiting step in the reaction [18]. Gcy1p completely lacks this loop (Fig. 3), a feature also seen in 2,5-diketo-D-gluconic acid reductase A [7]. It is likely that the reduced affinity for NADPH when compared with aldose reductase (28.5 versus 2 mM) is due to the absence of this loop [19]. Examination of the active site reveals substantial differences when compared with the aldose reductase active site and those of other AKRs. Although the main chain trace is similar, two key amino acid side chains have altered positions, requiring a conformational change to take place in order for catalysis to take place. The first is Trp-28, analogous to Trp-20 in human aldose reductase [3]. The side chain is positioned over the nicotinamide ribose in aldose reductase. In Gcy1p, the side chain stacks on the nicotinamide ring, preventing access to the ring by even the smallest substrates. The second structural perturbation involves Tyr-56, the residue that functions as the general acid in other AKRs (Tyr-48 in human aldose reductase). The pKa of the phenolic hydrogen is depressed in other AKRs by its proximity to a lysine (Lys-77 in human aldose reductase), which in turn is engaged in a salt bridge to an aspartate (Asp-43 in human aldose reductase). Gcy1p preserves the positions of the lysine and aspartate but the conformation of the loop consisting of residues 52–59 and containing the catalytic tyrosine is changed (Fig. 4). When compared with the human aldose reductase structure, the tyrosine Ca position is displaced by more than 4 A, . A rigid movement of this loop would bring the tyrosine into position to hydrogen bond with Lys-81 and allow catalysis to occur. Whether the conformational changes involving Trp-28 leaving the active site and the tyrosine entering are concerted is not yet clear. It is necessary for the tryptophan to move, however, to provide room for the tyrosine.
4. Conclusions Gcy1p, a yeast protein with homology to AKRs, including catalytic and cosubstrate-binding residues, has been shown to possess NADPH-specific reductase activity. The structure of Gcy1p, currently being refined at 2.5 A, , has revealed similarities and differences with other AKR structures. As expected, the overall Fig. 3. Residues 21, 216 and 262 form a bidentate interaction (yellow bonds), closing a loop or ‘safety belt’ of residues over the top of the NADPH cosubstrate (blue CPK model) in aldose reductase (shown in blue). No such loop exists in Gcy1p (white), possibly accounting for its lower affinity for NADPH. Fig. 4. A conformational rearrangement of active site residues is seen in Gcy1p (white) when compared with the human aldose reductase structure (blue). Trp-28 in Gcy1p (Trp-20 in aldose reductase) is rotated so that it stacks on top of the nicotinamide where the carbonyl containing substrate should bind. Tyr-56 (Tyr-48 in aldose reductase), which normally functions as a general acid, is shifted out of the active site. Activating residues such as Lys-81 and Asp-51 are in conserved positions.
E. Hur, D.K. Wilson / Chemico-Biological Interactions 130–132 (2001) 527–536
535
(b/a)8 structure shows little deviation from those of other AKRs. The majority of NADPH binding interactions is also well conserved. A major exception is the complete absence of the belt of residues that fold over the NADPH, creating a higher affinity binding site. Differences are also seen in the substrate-binding site. This region is more open and flatter than similar areas in other AKR structures, suggesting that larger substrate may be able to fit into it. The substrate-binding pocket immediately above the nicotinamide ring is blocked by a tryptophan, which must undergo a conformational change so that the cosubstrate may be accessible to the substrate. Finally, differences are seen in the position of the tyrosine general acid, which also requires a conformation change so that the protein can exist in a catalytically competent configuration. It is not yet clear why the enzyme has evolved this unconventional catalytic mechanism and what implications it may have on substrate binding and other enzymatic properties.
Acknowledgements Many thanks are due to Prof. Mark Petrash and coworkers for collaborating and sharing results before publication. E.H. was supported in part by an undergraduate fellowship from the California Foundation for Biomedical Research. This research was supported by a grant from the NIH to D.W. and a grant from the W.M. Keck Foundation.
References [1] J.M. Jez, T.G. Flynn, T.M. Penning, A nomenclature system for the aldo-keto reductase superfamily, in: H. Weiner, R. Lindahl, D. Crabb, T.G. Flynn (Eds.), Enzymology and Molecular Biology of Carbonyl Metabolism 6, Plenum Press, New York, 1996, pp. 579 – 589. [2] J.M. Rondeau, F. Tete-Favier, A. Podjarny, J.M. Reymann, P. Barth, J.F. Biellmann, D. Moras, Novel NADPH-binding domain revealed by the crystal structure of aldose reductase, Nature 355 (1992) 469–472. [3] D.K. Wilson, K.M. Bohren, K.H. Gabbay, F.A. Quiocho, An unlikely sugar substrate site in the 1.65 A, structure of the human aldose reductase holoenzyme implicated in diabetic complications, Science 257 (1992) 81–84. [4] S.S. Hoog, J.E. Pawlowski, P.M. Alzari, T.M. Penning, M. Lewis, Three-dimensional structure of rat liver 3a-hydroxysteroid/dihydrodiol dehydrogenase: a member of the aldo-keto reductase superfamily, Proc. Natl. Acad. Sci. USA 91 (1994) 2517 – 2521. [5] O. El-Kabbani, K. Judge, S.L. Ginell, D.A.A. Myles, L.J. DeLucas, T.G. Flynn, Structure of porcine aldehyde reductase holoenzyme, Nat. Struct. Biol. 2 (1995) 687 – 692. [6] D.K. Wilson, T. Nakano, J.M. Petrash, F.A. Quiocho, 1.7 A, structure of FR-1, a fibroblast growth factor-induced member of the aldo-keto reductase family, complexed with coenzyme and inhibitor, Biochemistry 34 (1995) 14323–14330. [7] S. Khurana, D.B. Powers, S. Anderson, M. Blaber, Crystal structure of 2,5-diketo-D-gluconic acid reductase A complexed with NADPH at 2.1 A, resolution, Proc. Natl. Acad. Sci. USA 95 (1998) 6768–6773. [8] J.M. Gulbis, S. Mann, R. MacKinnon, Structure of a voltage-dependent K+ channel b subunit, Cell 97 (1999) 943–952.
536
E. Hur, D.K. Wilson / Chemico-Biological Interactions 130–132 (2001) 527–536
[9] Q. Ye, D. Hyndman, X. Li, T.G. Flynn, Z. Jia, Crystal structure of CHO reductase, a member of the aldo-keto reductase superfamily, Proteins 38 (2000) 41 – 48. [10] U. Oechsner, V. Magdolen, W. Bandlow, A nuclear yeast gene (GCY) encodes a polypeptide with high homology to a vertebrate eye lens protein, FEBS Lett. 238 (1988) 123 – 128. [11] V. Magdolen, U. Oechsner, P. Trommler, W. Bandlow, Transcriptional control by galactose of a yeast gene encoding a protein homologous to mammalian aldo/keto reductases, Gene 90 (1990) 105–114. [12] M. Angermayr, W. Bandlow, The type of basal promoter determines the regulated or constitutive mode of transcription in the common control region of the yeast gene pair GCY1/RIO1, J. Biol. Chem. 272 (1997) 31630–31635. [13] S.A. Jelinsky, L.D. Samson, Global response of Saccharomyces cerevisiae to an alkylating agent, Proc. Natl. Acad. Sci. USA 96 (1999) 1486 – 1491. [14] I. Tarle, D.W. Borhani, D.K. Wilson, F.A. Quiocho, J.M. Petrash, Probing the active site of human aldose reductase: site directed mutagenesis of Asp-43, Tyr-48, Lys-77 and His-110, J. Biol. Chem. 268 (1993) 25687–25693. [15] E. Hur, D.K. Wilson, Crystallization and aldo-keto reductase activity of Gcy1p from Saccharomyces cere6isiae, Acta Crystallogr. D56 (2000) 763 – 765. [16] Z. Otwinowski, W. Minor, Processing of X-ray diffraction data collected in oscillation mode, Methods Enzymol. 276 (1997) 307–326. [17] J. Navaza, AMoRe: an automated package for molecular replacement, Acta Crystallogr. A50 (1994) 157–163. [18] C.E. Grimshaw, M. Shahbaz, C.G. Putney, Mechanistic basis for nonlinear kinetics of aldehyde reduction catalyzed by aldose reductase, Biochemistry 29 (1990) 9947 – 9955. [19] K.M. Bohren, J.L. Page, R. Shankar, S.P. Henry, K.H. Gabbay, Expression of human aldose and aldehyde reductases. Site-directed mutagenesis of a critical lysine 262, J. Biol. Chem. 266 (1991) 24031–24037. [20] P.J. Kraulis, MOLSCRIPT: a program to produce both detailed and schematic plots of protein structures, J. Appl. Crystallogr. 24 (1991) 946 – 950. [21] E.A. Merritt, M.E.P. Murphy, RASTER3D Version 2.0 — a program for photorealistic molecular graphics, Acta Crystallogr. D50 (1994) 869 – 873.
.