Proc. Natl. Acad. Sci. USA Vol. 92, pp. 11666-11670, December 1995 Biochemistry

A highly active decarboxylating dehydrogenase with rationally inverted coenzyme specificity RIDONG CHEN, ANN GREER, AND ANTONY M. DEAN Department of Biological Chemistry, The Chicago Medical School, 3333 Green Bay Road, North Chicago, IL 60064 Communicated by Daniel E. Koshland, Jr., University of California, Berkeley, CA, September 1, 1995

ABSTRACT The of Escherichia phosphate of bound NADP (Fig. 1B; E. coli IDH numbering). coli, which lacks the Rossmann fold common to other dehy- These residues are conserved in prokaryotic NADP-de- drogenases, displays a 7000-fold preference for NADP over pendent IDHs and replaced with a variety of residues in the NAD (calculated as the ratio of kcat/Km). Guided by x-ray NAD-dependent dehydrogenases (Table 1). In T. thermophilus crystal structures and molecular modeling, site-directed mu- IMDH, there is no site equivalent to position 395, while tagenesis has been used to introduce six substitutions in the replacements Ser-292', Ile-345, and Pro-391 eliminate all adenosine binding pocket that systematically shift coenzyme favorable interactions with the 2'-phosphate (Fig. 1B). Spec- preference toward NAD. The engineered displays an ificity in IMDH is conferred by the conserved Asp-344, which 850-fold preference for NAD over NADP, which exceeds the forms a double hydrogen bond with the 2'- and 3'-hydroxyls of 140-fold preference displayed by a homologous NAD- the adenosine ribose of NAD, shifting its position and perhaps dependent enzyme. Of the six mutations introduced, only one changing the ribose pucker from C3'-endo to C2'-endo. Not is identical in all related NAD-dependent enzyme sequences- only are these movements incompatible with the strong 2'- strict adherence to homology as a criterion for replacing these phosphate interactions seen in IDH but also the negative amino acids impairs function. Two additional mutations at charge on Asp-344 may repel NADP. Indeed, the dramatic remote sites improve performance further, resulting in a final drop in the specificity of E. coli IDH toward isocitrate upon mutant enzyme with kinetic characteristics and coenzyme phosphorylation of an Ser is caused by electrostatic preference comparable to naturally occurring homologous repulsion of the y-carboxylate of its carboxylic acid NAD-dependent . (10, 11). Herein, guided by a knowledge of the determinants of Descriptions of the determinants of specificity based on pro- coenzyme specificity and molecular modeling, we engineer a tein structures represent plausible hypotheses that beg exper- highly active NAD-specific enzyme in the nucleotide binding imental verification. A thorough understanding of the deter- domain of the NADP-dependent IDH of E. coli (5). Only one minants of specificity is demonstrated whenever the prefer- of the six substitutions introduced around the binding pocket ence for two substrates is inverted by rational means. are identical in all NAD-dependent IMDHs. Dehydrogenases discriminate among nicotinamide coenzymes through interactions established between the protein and the MATERIALS AND METHODS 2'-phosphate of NADP and the 2'- and 3'-hydroxyls of NAD. Engineering dihydrolipoamide and malate dehydrogenases Site-Directed Mutagenesis. Plasmid pTK513, which carries demonstrates that changing the preference of an NAD- the kcd gene inserted into pEMBL18- (12), was used to dependent enzyme can be achieved by introducing positively generate uridine-labeled template in E. coli CJ236 with helper charged residues to neutralize the negatively charged 2'- phage R401. Oligonucleotide primers containing the necessary phosphate of NADP (1, 2). Yet, as earlier attempts to invert mismatches were synthesized on a Biosearch model 8700 DNA the preference of glutathione reductase and glutamate dehy- synthesizer and were used to introduce mutations into the kcd drogenase illustrate, engineering the preference of an NADP- gene by the Kunkel method (6) with a kit from Bio-Rad. dependent enzyme toward NAD is more troublesome (3, 24). Putative mutants were screened by kinetic analysis and then Perhaps, the strict reliance on homology as a criterion for confirmed by dideoxynucleotide sequencing (13). replacing amino acids is insufficient to optimize directional Cell Growth and Enzyme Purification. After transforma- interactions, such as hydrogen bonds to the 2'- and 3'- tion of the mutated plasmids into E. coli SL4 (AIDH), cultures hydroxyls of NAD. were grown to full density overnight in 100 ml of Luria broth The decarboxylating dehydrogenases, of which Escherichia at 37°C in the presence of ampicillin (60 gg/ml). Purification coli isocitrate dehydrogenase (IDH) and Thermus thermophilus of the enzymes, by the procedure of Garnak and Reeves (14) isopropylmalate dehydrogenase (IMDH) are members, form as modified by Dean and Koshland (15), involves ammonium an ancient family of dehydrogenases sharing 25% amino acid sulfate precipitation, DEAE anion chromatography, and af- sequence identity and a common catalytic mechanism (4, 5). finity chromatography using Affi-Gel Blue. All preparations They also share a common protein fold (Fig. 1A), one that is are 95% free of contaminating enzyme, as judged by Coomas- topologically distinct from other dehydrogenases of known sie blue staining after SDS/PAGE electrophoresis. structure and that lacks the afc3ap binding motif characteristic Kinetic Analyses. The kinetics of IDHs were determined in of the nucleotide binding Rossmann fold. Instead, the aden- KAC buffer (25 mM Mops/100 mM NaCl/5 mM MgCl2/1 mM osine moiety of coenzyme binds in a pocket constructed from dithiothreitol, pH 7.5) at 21°C in the presence of 1 mM two loops and an a-helix in IDH, although the latter is DL-isocitrate (10). Data were collected on a Hewlett-Packard substituted by a (3-turn in IMDH (Fig. 1A) (4, 5). model 8452A single-beam diode-array spectrophotometer. Specificity in E. coli IDH is conferred by interactions among The rates of reaction were determined by monitoring the Arg-395, Tyr-345, Tyr-391, and Arg-292' with the 2'- production of NAD(P)H at 340 nm in a 1-cm light path by using a molar extinction coefficient of 6200 M-1cm-1. Protein The publication costs of this article were defrayed in part by page charge payment. This article must therefore be hereby marked "advertisement" in Abbreviations: IDH, isocitrate dehydrogenase; IMDH, isopropyl- accordance with 18 U.S.C. §1734 solely to indicate this fact. . 11 tt Downloaded by guest on September 25, 2021 Biochemistry: Chen et al. Proc. Natl. Acad. Sci. USA 92 (1995) 11667

FIG. 1. (A) Ribbon trace of a monomer of E. coli IDH (,B-sheets are green, a-helices are purple, and loops are white), showing the positions of the six Cys residues (numbered left to right are Cys-405, -332, -127, -301, and -201 with Cys-194 in the lower loop: all marked by yellow side chains). The large cleft between the two domains contains the active site marked by isocitrate and NADP (carbon is white, nitrogen is blue, is red, and phosphorous is light yellow). (B) Superposition of the binding pockets of the NADP-dependent E. coli IDH (5, 6) with two waters (red spheres) and the NAD-dependent T. thermophilus IMDH (ref. 4; yellow). Side chains of Ile-37, Val-41, Ile-320, His-339, Ala-342, and Val-351, the aliphatic portion of the side chains of Asn-352 and Asp-392, and the main-chain residues Gly-321 and Asn-352 form the binding pocket (E. coli IDH numbering). All residues are identical in T thernophilus IMDH except for conservative substitutions replacing Ile-320 with Leu and Val-351 with Ala. N2 and N6, common to the adenosine 2',3'-bisphosphate moiety of NADP, form hydrogen bonds to the main-chain amide and carbonyl of Asn-352. A dipole-quadrupole interaction between the adenine N6 and the His ring is evident in IMDH, but the low pH conditions necessary for crystallization of IDH may have disrupted this interaction. Specificity in IDH is conferred by interactions among residues Arg-292' (on the second domain of the second subunit), Arg-395, Tyr-345, and Tyr-391 with the 2'-phosphate of bound NADP. Specificity in IMDH is conferred by Asp-344, which forms a double hydrogen bond with the 2'- and 3'-hydroxyls of the adenosine ribose of NAD and may also repel the 2'-phosphate of NADP. concentrations were determined at 280 nm by using a molar tate, the performance with NADP was greatly reduced (Table extinction coefficient of 66,330 M-1-cm-1 (15). Nonlinear least 2). Although preference no longer favored NADP, the per- squares Gauss-Newton regressions were used to determine the formance with NAD suggests that no new interactions were fit of the data to the Michaelis-Menten model. established with this coenzyme. Molecular Modeling. Molecular modeling was conducted on Val-351 of IDH is either conserved or replaced by Ala in the a Silicon Graphics 4D120/GTX using QUANTA/CHARMM soft- NAD-dependent enzymes (Table 1). Modeling suggested that ware program and visualized using the Crystal Eyes stereo the reduced bulk of Ala might allow the adenosine to shift, viewing system. X-ray crystallographic structures of the binary bringing the 2'- and 3'-hydroxyls of the attached ribose closer complexes of IDH with NADP and IMDH with NAD were to Asp-344. Site-directed mutagenesis was used to generate the superimposed by least squares minimization of main-chain triple mutant with striking results. The 14-fold increase in atoms surrounding the nucleotide binding pockets. Amino acid performance with NAD was consonant with the formation of subsitutions were modeled assuming that the polypeptide a hydrogen bond. Preference now favored NAD by a factor of backbone remained unchanged, and side chains were adjusted 4 (Table 2). by rotating torsional bonds to establish favorable interactions Tyr-391 is replaced with Pro in T. thermophilus IMDH (Fig. with NAD. 1B). This substitution removes a hydrogen bond to the 2'- phosphate and alters the local secondary structure from RESULTS a-helix to 3-turn. Pro, common to many IMDHs, was not introduced at site 391 to avoid disrupting the a-helix of IDH. Substitutions in the Coenzyme Binding Pocket. Site- Nor was Phe chosen because it might become buried in the directed mutagenesis was used to replace Lys-344 with Asp and hydrophobic pocket, thereby hindering the approach of the Tyr-345 with Ile. Both Asp-344 and Ile-345 are identical in all adenosine of NAD toward Asp-344. Sequence alignments known prokaryotic NAD-dependent decarboxylating dehydro- suggested that either Gly or Arg might be introduced at this genases (Table 1). As expected from a loss of hydrogen site. Further along the a-helix of IDH is Arg-395, which also bonding and the introduction of a potentially repulsive aspar- hydrogen bonds to the 2'-phosphate, but which has no equiv- Downloaded by guest on September 25, 2021 11668 Biochemistry: Chen et aL Proc. Natl. Acad. Sci. USA 92 (1995) Table 1. Alignments of the primary sequences of the decarboxylating dehydrogenases around the adenosine binding pocket Enzyme Sequence

** * * ** * * * NADP-dependent IDH 283 294 316 325 334 344 351 357 338 395 E. coli DAFLQQILLRPAEY QVGGIGIAPGAN LFEATHGTAPKYAGQDKVNPGSIILS TVTYDFERLM T. thermophilus DNAAHQLVKRPEQF LIGGLGFAPSAN IFEAVHGSAPKYAGKNVINPTAVLLS VLTGDVVGYD Vibrio sp. DAMLQQVLLRPAEY QVGGIGIAPGAN VFZATHGTAPKYAGKNKVNPGSVILS TVTYDFERLM NAD-dependent IMDH T. thermophilus DAMAMHLVRSPARF LPGSLGLLPSAS VFEPVHGSAPDIAGKGIANPTAAILS TPPPDLGGSA T. aquaticus DAMAMHLVKNPARF LPGSLGLLPSAS VFZPVHGSAPDIAGKGIANPTAAILS TPPPDLGGSA E. coli DNATMQLIKDPSQF ITGSMGMLPSAS LYZPAGGSAPDIAGKNIANPIAQILS IRTGDLARGA S. typhimurium DNATMQLIKDPSQF ITGSMGMLPSAS LYEPAGGSAPDIAGKNIANPIAQILS VRTGDLARGA B. subtilis DNAAMQLIYAPNQF LTGSLGMLPSAS LFZPVHGSAPDIAGKGMANPFAAILS KRTRDLARSE L. interrogans DNAAMQLIVNPKQF ITGSIGMLPSAS LYEPSGGSAPDIAGKGVANPIAQVLS KRTRDIEVGS A. tumefaciens DAGGMQLVRKPKQF LTGSLGMLPSAS MYEPVHGSAPDIAGKSIANPIAMIAS IRTADIMADG Y lipolytica DSAAMILIKQPSKM IPGSLGLLPSAS LYEPCHGSAPDL.GKQKVNPIATILS ITTADIGGSS S. cerevisiae DSAAMILVKNPTHL IPGSLGLLPSAS LYEPCHGSAPDL.PKNKVNPIATILS IRTGDLGGSN C. utilis DSAAMILIKYPTQL IPGSLGLLPSAS LYEPCHGSAPDL.PANKVNPIATILS IRTGDLKGTN S. pombe DSAAMLLVKSPRTL IPGSLGLLPSAS LVEPIHGSAPDIAGKGIVNPVGTILS LYTRDLGGEA NAD-dependent IDH S. cerevisiae DNSVLKVVTNPSAY SAGSLGLTPSAN IFZAVHGSAPDIAGQKDANPTALLLS NRTGDLAGTA NAD-dependent TDH P. putida DILCARFVLQPERF CAGTIGIAPSAN LFEPVHGSAPDIFGKNIANPIAMIWS SVTPDMGGTL IDH numbering is used throughout, asterisks denote sites subjected to mutagenesis, and boldface type denotes rigidly conserved amino acids. IDH was from E. coli, T. thermophilus, and Saccharomyces cerevisiae (7), and Vibrio sp. (8); IMDH was from T-. thermophilus, Thermus aquaticus, E. coli, Bacillus subtilis, Leptospira interrogans, Agrobacterium tumefaciens, Yarrowia lipolytica, Saccharomyces cerevisiae, Candida utilis, Schizosac- charomyces pombe, and Salmonella typhimurium (7). TDH, tartrate dehydrogenase (9). P. putida, Pseudomonas putida.

alent in the (3-turn of IMDH. If sequence alignments suggest Arg-395 in the a-helix with Val, Thr, Leu, Met, or Ile might that Arg occupies this "site" in several IMDHs, the secondary generate steric effects with adjacent residues. Hence, Ser, a structure of the (3-turn in IMDH shows that the side chain will polar residue suitable for replacements at the surface of a point away from the nucleotide . Sequence align- protein, was chosen to replace Arg-395. The introduction of ments suggested the Arg might be replaced with Gly or Ala. Y391K and R395S to generate K344D/Y345I/V351A/ The introduction of Gly residues at both sites improved Y391K/R395S caused a dramatic decrease in performance preference by destroying activity with NADP (no was with NADP, as expected from the loss of two hydrogen bonds detectable). However, the performance with NAD was re- to the 2'-phosphate and a modest increase in performance with duced below that of the wild-type enzyme (K344D/Y345I/ NAD (Table 2). The preference for NAD over NADP was thus V351A/Y391G/R395G had a Km of 1400 pLM, a kcat of 0.284 100-fold. sec-1, and a performance kcat/Km of 0.0002 ,uM-1lsec-1). A A shift in the loop formed by residues 316-322 alters the second mutant (K344D/Y345I/V351A/Y391R/R395A) pro- orientation of Leu-320 in IMDH, nudging the adenosine duced comparable results. Instead, Tyr-391 was replaced by moiety of NAD toward Arg-344. This shift, generated by hydrophilic Lys that, being shorter than the Arg found in many Pro-317, is stabilized by a hydrogen bond between Asp-392 and IMDHs, was less likely to interact the 2'-phosphate of NADP. the hydroxyl of Ser-319 that precisely replaces by a bound Although at the surface, modeling indicated that replacing water with a similar function in IDH. All attempts to engineer Table 2. Kinetic parameters of wild-type and mutant enzymes toward NADP and NAD NADP NAD Performance, Performance Preference, Km, kcat, kcat/Km, Km, kcat, kcat/Km, NADP performance Enzyme ,uM sec- 1 M-1 . sec-I1 uM sec-1 .&M-1-sec-1 /NAD performance E. coli IDH at 21°C abcdefgh KYVYRRCC (wild type) 17 80.5 4.7 4700.0 3.22 0.00069 6900 DI ------7,300 6.3 0.00086 3300.0 2.59 0.00078 1.1 DIA----- 6,400 18.0 0.0028 850.0 9.39 0.011 0.25 DIAKS--- 11,300 2.0 0.00018 290.0 6.00 0.021 0.009 DIAKSD-- 32,200 0.39 0.000012 924.0 9.74 0.011 0.001 DIAKS-Y- 2,800 3.34 0.0012 108.0 11.4 0.106 0.011 DIAKS-YI 5,800 4.70 0.00081 99.0 16.20 0.164 0.005 Saccharomyces cerevisiae Wild-type IDH at 24°C (16, 17) 210 40 0.190 Wild-type IMDH at 30°C (18) 140 14.48 0.103 Salmonella typhimurium Wild-type IMDH at 24°C (19) 100 31.5 0.315 T. thermophilus Wild-type IMDH at 65°C (7) 12,300 29.9 0.00243 40 13.6 0.34 0.007 All apparent standard errors are <15% of the estimates. For E. coli IDH, residues at the following positions are shown: a, 344; b, 345; c, 351; d, 391; e, 395; f, 292; g, 332; h, 201. Dashes indicate wild-type residues. Downloaded by guest on September 25, 2021 Biochemistry: Chen et al. Proc. Natl. Acad. Sci. USA 92 (1995) 11669 the loop in IDH (e.g., Q316L, V317P, G319S, and/or I320L) malate dehydrogenase merely generates an enzyme with poor reduced performance toward both coenzymes. Engineering substrate specificity (21). Similarly, introducing positively the entire loop produced similar results (the performance of charged side chains into the Rossmann fold of the E. coli K344D/Y3451/V351A/Y391K/R395S/Q316L/V317P/ NAD-dependent dihydrolipoamide dehydrogenase generates G319S/1320L was 0.00096 uM-lsec-1 with NAD and 0.00001 a highly specific NADP-dependent enzyme (1), while substi- AM-1 sec-1 with NADP). All substitutions in the loop were tutions in the Rossmann fold of the NADP-dependent E. coli omitted from the next round of engineering. glutathione reductase fail to generate an enzyme with a strict Arg-292' forms a hydrogen bond from the small domain of preference for NAD, even at optimal pH (3). Indeed, engi- the second subunit to the 2'-phosphate of NADP and is neering IDH performance is frequently impaired when ho- replaced by Ser in T. thermophilus IMDH, and a wide variety mology is the sole criterion governing the introduction of of amino acids in other NAD-dependent enzymes (Table 1). residues. For example, the substitution of Gly at sites 391 and The introduction of Asp improved the preference for NAD 395 reduces performance with NAD by a factor of 55, while over NADP to 850-fold, exceeding the 140-fold preference engineering the loop between Gln-316 and Gly-321 reduces displayed by T. thermophilus IMDH (Table 2). However, this performance with both coenzymes by factors >10. substitution also increased the Km of NAD by a factor of 3 and In this three substitutions based on ho- was, therefore, omitted from the next round of engineering. study approaches, Substitutions Outside the Coenzyme Binding Pocket. As mology, alternative substitutions within binding sites, and judged by improved performance with NAD by a factor of 30, substitutions outside binding sites, were combined to invert the K344D/Y345I/V351A/Y391K/R395S was the most success- coenzyme preference of E. coli IDH while retaining perfor- ful mutant so far generated. Nevertheless, it was far less active mance. than natural homologous NAD-dependent decarboxylating Substitutions Based on Homology. The first two mutations, dehydrogenases (Table 2). We decided to investigate the K344D and Y345I, were based on homology. Sequence align- possibility that amino acid substitutions outside the nucleotide ments indicated that Asp-334 is rigidly conserved in all related binding pocket might promote binding and catalysis. Six Cys NAD-dependent enzymes, and structural analysis indicated residues were targeted because they are limited in number and distributed haphazardly with respect to the coenzyme binding 1.2 pocket: Cys-405 is immediately adjacent, Cys-332 lies on a a. 1 F. (3-sheet that traverses the pocket, Cys-127 supports Arg-159, 0.-. which hydrogen bonds to the a-carboxylate of isocitrate, s 0.8 Cys-301 lies deep in the hydrophobic core of the second domain, Cys-201 lies adjacent to a loop that forms part of the O 3r03: 0.6 active site ip the dimer, and Cys-194 lies at the dimer interface in the extended loop (Fig. 1A). Amino acid substitutions were 0.4 introduced by using degenerate oligonucleotides, one encod- ing bulkier residues (Phe, Tyr, and His) and the other encoding o-IL 0 0.21 longer residues (Ile, Lys, and Met). In accommodating these bulky residues, subtle conformational shifts might be trans- 0 mitted into the adenosine binding pocket and the active site, fortuitously improving performance. Molecular modeling suggested that substitutions at Cys-127 < _ 200 would disturb isocitrate binding by displacing Arg-159, those at z a the deeply buried Cys-301 would disrupt the hydrophobic core, 3-= 150 while those at Cys-194 would interfere with subunit interac- 310 tions. Indeed, substitutions at these three sites greatly reduced 00Io performance of the K344D/Y345I/V351A/Y391K/R395S mutant by at least a factor of 100 (data not shown). Although Cys-405 is adjacent to the adenosine binding pocket, all 50 substitutions were predicted to have negligible structural ef- fects because minor torsion changes allow side chains to be exposed to solvent. As expected, none of the substitutions introduced influenced enzyme performance, which in all cases D-- 1 UUU remained comparable to the parent K344D/Y345I/V351A/ z Y391K/R395S mutant (Table 2 and data not shown). 100

The partially buried Cys-332 in the (3-sheet that traverses the < 1 0 adenosine binding pocket is also adjacent to active site resi- z dues. Replacements with Phe, and especially Tyr, increased 0 ; kcat by 2-fold and decreased Km by 3-fold (Table 2). Cys-201 lies adjacent to the super secondary structure containing an active 0.1 site Lys-230'. Replacement with Ile caused a modest improve- CL 0.01 1 ment in kcat (Table 2). All bulkier residues impaired perfor- i 0.001 mance by factors >100 (data not shown). 00Q H X C) >4 >d > cnuuv

DISCUSSION ', H H H H H studies that as a for Previous reveal using homology strict guide FIG. 2. Systematic shift in coenzyme preference generated in determining the introduction of amino acid substitutions suc- engineered mutants of IDH. Differences in the height of the histogram ceeds as often as it fails. For example, the replacement of a columns represent the degree to which performance (kcat/Km) and single Glu with Arg converts Bacillus stearothermophilus lac- preference [calculated as the ratio (kcat/Km)NADp/(kcat/Km)NAD] is tate dehydrogenase into a highly specific malate dehydroge- shifted by each set of mutations (denoted by the single-letter amino nase (20), while the replacement of Arg with Gln in E. coli acid code). Downloaded by guest on September 25, 2021 11670 Biochemistry: Chen et al. Proc. Natl. Acad. Sci. USA 92 (1995) that it hydrogen bonds to the 2'- and 3'-hydroxyls of the IMDHs from various sources (Table 2). The fact that the adenosine ribose of NAD while potentially repelling the Michaelis constant of the mutant enzyme is higher toward 2'-phosphate of NADP. Structural analysis also revealed that NAD than that ofthe wild-type enzyme toward NADP appears the substitution Y3451, a conserved residue in many related typical of natural IDHs: the Michaelis constants of NAD- NAD-dependent enzymes, not only removes a hydrogen bond dependent enzymes range from 150 ,uM to 800 j,M, whereas to the 2'-phosphate of NADP but might eliminate any steric those of the NADP-dependent enzymes range from 2 ,M to hindrance preventing the formation of the hydrogen bonds 20 ,uM (22). Perhaps the strong electrostatic interactions with between the ribose hydroxyls and Asp-334. Nevertheless, the the 2'-phosphate of NADP, evident in the crystal structure of resulting double mutant showed no improved performance E. coli IDH with NADP (Fig. 1B), are responsible for this with NAD (Fig. 2). difference. The reason why the maximum rate of catalysis Substitutions Based on Design. Structural analysis sug- displayed by wild-type IDH utilizing NADP is 5-fold higher gested that the bulk of Val-351 might prevent the adenosine than with the engineered mutant (Table 2) is unclear. Con- shifting, thereby preventing Asp-344 from forming hydrogen formational changes induced by NADP binding in IDH are bonds to 2'- and 3'-hydroxyls of the adenosine ribose of NAD. restricted to side-chain movements in the immediate vicinity of Replacement with Ala, found in some but not all NAD- the coenzyme binding pocket (5). The structure of a pseudo- dependent enzymes, produced the desired effect: the triple Michaelis ternary complex with NADP and isocitrate has been mutant had a 14-fold increase in performance with, and a determined (23), but again, no obvious changes in the active 4-fold preference for, NAD (Fig. 2). The next two substitutions site can be ascribed to changes induced in the coenzyme introduced were Y391K and R395S. The criteria for introduc- binding pocket, which lies some 14 A away. Hence, our final ing these substitutions were removal of hydrogen bonds to the mutant is probably as efficient an NAD-dependent IDH as is 2'-phosphate of NADP, maintenance of local secondary struc- currently possible to design. ture, avoidance of steric effects hindering the approach of the adenosine ribose toward Asp-344, and hydrophilicity, these We thank Eric Walters, Jim Hurley, and Bob Kemp for their being surface residues. The 2-fold increase in performance and thoughtful suggestions. This work was supported by Public Health 30-fold increase in preference with NAD (Fig. 2) was achieved Service Grant GM-48735 from the National Institutes of Health. solely by rational design since these sites have no direct counterparts in IMDH, where the a-helix is replaced by a 1. Bocanegra, J. A., Scrutton, N. S. & Perham, R. N. (1993) Bio- ,-turn. chemistry 32, 2737-2740. Substitutions Outside the Coenzyme Binding Pocket. Ad- 2. Nishiyama, M., Birktoff, J. J. & Beppu, T. (1993) J. Bio. Chem. ditional mutations in the binding pocket impaired perfor- 268, 4656-4660. mance with NAD. the failure of several other 3. Scrutton, N. S., Berry, A. & Perham, R. N. (1990) Nature (Lon- Furthermore, don) 343, 38-43. studies to invert preference based on homology or the use of 4. Hurley, J. H. & Dean, A. M. (1994) Structure 2, 1007-1016. alternative' substitutions within binding sites, suggested the 5. Hurley, J. H., Dean, A. M., Koshland, D. E., Jr., & Stroud, R. M. possibility that substitutions outside the binding site might (1991) Biochemistry 30, 8671-8678. improve performance with NAD. 6. Kunkel, T. A., Roberts, J. D. & Zakour, R. A. (1987) Methods The choice of replacing the six Cys residues (Fig. 1A) was Enzymol. 154, 367-382. arbitrary, save their limited number and haphazard distribu- 7. Miyazaki, K. & Oshima, T. (1994) Protein Eng. 7, 401-403. tion with respect to the coenzyme binding site. The degenerate 8. Ishii, A. (1993) J. Bacteriol. 175, 6873-6880. oligonucleotides encoded larger amino acids to force changes 9. Tipton, P. A. & Beecher, B. S. (1994) Arch. Biochem. Biophys. in conformation. A net 8-fold increase in performance was 313, 15-21. at two 10. Dean, A. M. & Koshland, D. E., Jr. (1990) Science 249, 1044- generated by substitutions of the six sites. The mech- 1046. anisms by which this improvement was generated awaits 11. Hurley, J. H., Dean, A. M., Sohl, J. L., Koshland, D. E., Jr., & further structural analysis. However, the implication is that the Stroud, R. M. (1990) Science 249, 1012-1016. combined effects of many substitutions outside binding sites 12. Thorsness, P. E. & Koshland, D. E., Jr. (1987) J. Biol. Chem. 262, and catalytic enters may be considerable in aggregate. These 10422-10425. observations may help explain both the general difficulty 13. Sanger, F., Nicklen, S. & Coulsen, R. (1977) Proc. Natl. Acad. Sci. encountered in engineering enzyme function and the reason USA 74, 5463-5467. why enzymes are so large. 14. Garnak, M. & Reeves, H. C. (1979) J. Biol. Chem. 254, 7915- Kinetic Properties of the Final Mutant. Seven amino acid 7920. substitutions introduced into wild-type E. coli IDH cause a 15. Dean, A. M. & Koshland, D. E., Jr. (1993) Biochemistry 32, shift in preference from NADP to NAD by a factor >106: 9302-9309. 16. Barnes, L. D., Kuehn, G. D. & Atkinson, D. E. (1971) Biochem- performance with NADP was reduced 6000-fold, and perfor- istry 10, 3939-3945. mance with NAD increased 240-fold (Fig. 2). Of the seven 17. Cupp, J. R. & McAlister-Henn, L. (1993) Biochemistry 32, 9323- successful substitutions introduced into IDH, only two were 9328. based on strict homology with the NAD-dependent enzymes 18. Hsu, Y.-P. & Kolhaw, G. B. (1980)J. Biol. Chem. 255,7255-7260. (Table 1). Our final mutant displays a preference for NAD 19. Parsons, S. J. & Burns, R. 0. (1970) Methods Enzymol. 17A, over NADP of 200-fold; has an NAD Michaelis constant of 100 793-799. ,tM, representing an improvement by a factor of 50, and a kcat 20. Wilks, H. M., Hart, K. W., Feeney, R., Dunn, C. R., Muirhead, of 16.2 sec-1, which represents an increase by a factor of 5 H., Chia, W. N., Barstow, D. A., Atkinson, T., Clarke, A. R. & (Table 2). The results suggest that a hydrogen bond between Holbrook, J. J. (1988) Science 242, 1541-1544. the adenosine ribose of NAD and Asp-344, as seen in the x-ray 21. Nicholls, D. J., Miller, J., Scawen, M. D., Clarke, A. R., Hol- brook, J. J., Atkinson, T. & Goward, C. R. (1992) Biochem. structure of the IMDH binary complex, may have been Biophys. Res. Commun. 189, 1057-1062. successfully established. Note that the Michaelis constant of 22. Chen, R. & Gadal, P. (1990) Plant Physiol. Biochem. 28, 411-418. isocitrate remains unchanged at 10 ,uM, suggesting that these 23. Stoddard, B. L., Dean, A. M. & Koshland, D. E., Jr. (1993) mutations have no effect on substrate binding. Biochemistry 32, 9310-9316. The kinetic characteristics of the final engineered enzyme 24. Haeffner-Gormley, L., Chen, Z., Zalkin, H. & Colman, R. (1992) compare favorably with natural NAD-dependent IDHs and Biochemistry 31, 7807-7814. Downloaded by guest on September 25, 2021