PEDS Advance Access published December 23, 2014 Protein Engineering, Design & Selection, 2014, 1–7 doi: 10.1093/protein/gzu053 Original Article
Original Article
Alteration of substrate specificity of alanine dehydrogenase Puja Fernandes1,†, Hannah Aldeborgh2,†, Lauren Carlucci1, Lauren Walsh1, Jordan Wasserman1, Edward Zhou1, Scott T. Lefurgy1, and Emily C. Mundorff1,* Downloaded from http://peds.oxfordjournals.org/ at Selcuk University on January 19, 2015
1
Chemistry Department, Hofstra University, Hempstead, NY 11549, USA, and 2Chemistry Department, Vassar College, Poughkeepsie, NY 12604, USA *To whom correspondence should be addressed. E-mail: [email protected] † These authors contributed equally to this work. Edited by Dan Tawfik Received 21 July 2014; Revised 30 October 2014; Accepted 19 November 2014
Abstract The L-alanine dehydrogenase (AlaDH) has a natural history that suggests it would not be a promising candidate for expansion of substrate specificity by protein engineering: it is the only amino acid dehydrogenase in its fold family, it has no sequence or structural similarity to any known amino acid dehydrogenase, and it has a strong preference for L-alanine over all other substrates. By contrast, engineering of the amino acid dehydrogenase superfamily members has produced catalysts with expanded substrate specificity; yet, this enzyme family already contains members that accept a broad range of substrates. To test whether the natural history of an enzyme is a predictor of its innate evolvability, directed evolution was carried out on AlaDH. A single mutation identified through molecular modeling, F94S, introduced into the AlaDH from Mycobacterium tuberculosis (MtAlaDH) completely alters its substrate specificity pattern, enabling activity toward a range of larger amino acids. Saturation mutagenesis libraries in this mutant background additionally identified a double mutant (F94S/Y117L) showing improved activity toward hydrophobic amino acids. The catalytic efficiencies achieved in AlaDH are comparable with those that resulted from similar efforts in the amino acid dehydrogenase superfamily and demonstrate the evolvability of MtAlaDH specificity toward other amino acid substrates. Key words: alanine dehydrogenase, amino acid dehydrogenase, evolvability, saturation mutagenesis, superfamily
Introduction It has become clear that some enzymes are more evolvable than others (Aharoni et al., 2005; Bloom et al., 2006; O’Loughlin et al., 2006). The determination of this evolvability trait has been the target of recent work (Dellus-Gur et al., 2013), but it is still unclear how to predict the capacity of a specific enzyme to evolve new activities or substrate specificities. One assumption is that the evolutionary history of an enzyme is a predictor of its evolutionary potential (O’Loughlin et al., 2006; Romero and Arnold, 2009). This principle is based on the idea that an enzyme derived from a fold family that has evolved
multiple new functions in nature would be more amenable to the development of new activities in the laboratory. However, this assumption ignores the multitude of constraints that may limit the evolutionary potential of an enzyme family in natural evolution, such as regulatory strategies, expression levels, metabolic flux and functional non-redundance. Such constraints disappear once the enzyme is in the artificial environment of a laboratory (Arnold et al., 2001). Also, given that we do not have a complete evolutionary tree, our understanding of an enzyme’s natural history may be misleading. As a counterexample to the natural evolvability argument, we present the
© The Author 2014. Published by Oxford University Press. All rights reserved. For Permissions, please e-mail: [email protected]
1
P. Fernandes et al.
2
presented (Hutter and Singh, 1999), the MtAlaDH would be considered a highly specific enzyme from a highly conserved family that appears to be unrelated to other AADHs. If the apparent evolutionary history is a guide to evolvability, one would reasonably expect to generate novel amino acid substrate activity more easily within an enzyme in the AADH superfamily than in MtAlaDH. We report here a single mutation of AlaDH (F94S) that switches its specificity from vastly preferring alanine to being generally promiscuous toward hydrophobic amino acids (L-norleucine, L-leucine, L-methionine, L-norvaline, L-2-amino-4-pentenoic acid, L-lysine, L-ornithine, L-serine, L-homoserine, L-homophenylalanine and L-allysine ethylene acetal). Saturation mutagenesis libraries in this mutant background additionally identified a double mutant (F94S/Y117L) showing improved activity toward L-norleucine, L-leucine, L-norvaline, L-methionine and L-homophenylalanine, and further depression of activity toward the native substrate L-alanine. The results obtained are comparable with those found within the amino acid superfamily and demonstrate the evolvability of MtAlaDH specificity to other amino acid substrates.
Materials and methods Materials All chemicals were purchased from Sigma-Aldrich (St. Louis, MO) at reagent grade or higher.
Molecular modeling The crystal structure of MtAlaDH with NAD+ and pyruvate bound ( pdbid: 2VHX) was used as the basis for all modeling studies using the software package, Discovery Studio 3.0 (Accelrys, San Diego, CA, USA). Positions to test were identified visually as those residues with side chains pointing toward the side chain of the substrate, Phe94 and Leu130. Mutations were made using the protein mutate function and models were minimized to an RMS gradient of less than 0.02 kcal/(mol × Å) with the generalized Born with molecular volume solvent model. All models were screened by manually modeling 4-methyl-2-oxopentanoic acid (the product of oxidative deamination of L-leucine) in a reactive conformation to identify those mutants that would best accommodate this substrate.
Generation of saturation mutagenesis libraries Site-directed saturation mutagenesis libraries were made using the QuikChange Lightning Site-Directed Mutagenesis kit (Agilent, Santa Clara, CA, USA) and the enclosed protocol was followed with the exception of a 1-hour DpnI digestion. The polymerase chain reaction products were transformed into XL-10 gold cells and selected for antibiotic resistance on LB agar with 30 μg/ml kanamycin. Colonies that appeared in 24–36 h were collected into 3 ml of LB broth and the plasmids were extracted using a spin miniprep kit from Qiagen (Valencia, CA). The miniprep plasmid DNA was used to transform BL21 competent cells and transformants were selected on LB-kanamycin agar plates. The colonies were picked by toothpicks, 176 per library, transferred to two 96-well plates, and grown overnight shaking at 200 rpm at 30°C with four wells each of positive (F94S) and negative (empty vector) controls. 180 μl of LB broth was inoculated with 10 μl of overnight growth and grown for 3 h at 30°C shaking at 200 rpm. The plates were induced with 1 mM β-D-isopropylthiogalactoside (IPTG), shaken at 200 rpm for ∼18 h at 30°C, and then centrifuged at 4000 rpm for 10 min. Lysis of the cell pellets commenced through resuspension in 120 μl per well in Tris–HCl (50 mM, pH7.2), NaCl
Downloaded from http://peds.oxfordjournals.org/ at Selcuk University on January 19, 2015
alteration of substrate specificity of L-alanine dehydrogenase (AlaDH), an enzyme from a family that looks particularly unpromising if one were to judge it based purely on its natural history. The members of the amino acid dehydrogenase (AADH) superfamily include the glutamate dehydrogenase (GluDH), leucine dehydrogenase (LeuDH), valine dehydrogenase and phenylalanine dehydrogenase (PheDH) families, but not AlaDH. The member families of the AADH superfamily have a long and rich history of being targeted in protein engineering to alter substrate specificity due to its members’ usefulness in diagnostic kits and in the biocatalysis of chiral amines and unnatural amino acids (Brunhuber and Blanchard, 1994). There is a clear evolutionary relationship among the AADH families, which demonstrate homology on the basis of sequence and structural similarities (Britton et al., 1993; Baker et al., 1997). Researchers have exploited the conserved features of these enzymes to alter their substrate specificities using rational design (Wang et al., 2001; Seah et al., 2003). Guided by homology and the hypothesis that certain positions in the enzyme exert discrete control over substrate specificity, researchers were able to successfully transform both a GluDH (Wang et al., 2001) and a PheDH (Kataoka and Tanizawa, 2003) into a L-leucine/L-norleucine dehydrogenase using three or fewer mutations. By contrast, AlaDH enzymes share no apparent evolutionary relationship with the AADH superfamily (Baker et al., 1998). There is no sequence similarity between AlaDH and any member of the AADH superfamily beyond both having the NAD+/NADH-binding Rossmann fold. Mechanistically, the AlaDH and PheDH differ in that they donate the hydride to opposite sides of the NAD+ moiety, indicating convergent evolution. AlaDH is classified in the same superfamily as NAD(P) transhydrogenase (AlaDH_PNT superfamily) due to the sequence and structural similarities between the two enzyme families (Jackson et al., 1999). The two families bind the adenosine moiety of NAD+ similarly, but bind the nicotinamide in different regions of the structure, such that the catalytic residues are not in structurally conserved positions. The only other family with significant structural similarity to AlaDH is saccharopine dehydrogenase (NAD+, L-lysine forming) (SacDH) (Burk et al., 2007). Representative structures in the PDB showed 14% sequence identity and 2.67 Å RMSD (PDBID:4LMP for MtAlaDH and PDBID:2Q99 for SacDH from Saccharomyces cerevisiae). The two enzymes share a common partial mechanism and three conserved catalytic residues, but perform overall different reactions (Burk et al., 2007). SacDH catalyzes the oxidative deamination of D-saccharopine to generate L-lysine and α-ketoglutarate; SacDH releases a primary amine, whereas the oxidative deamination of L-alanine by AlaDH releases ammonia. AlaDH has been studied from a variety of bacteria, generally maintaining ∼50% sequence identity among family members (Baker et al., 1998). In Bacillus subtilis it is known to be essential for the utilization of alanine as a sole carbon and nitrogen source and is required for normal sporulation (Siranosian et al., 1993). Increased levels of AlaDH have also been associated with both the maintenance of the NAD+ pool as well as generation of precursors for peptidoglycan biosynthesis (Hutter and Dick, 1998; Hutter and Singh, 1999; Betts et al., 2002; Feng et al., 2002). Why this enzyme family has remained highly conserved is unclear, but there appears to be some constraint on its evolution. It is also interesting that there is not a specific alanine dehydrogenase within the AADH superfamily. It is possible, of course, that an alanine dehydrogenase exists within the AADH superfamily but has yet to be identified, or that there are other, unknown AADHs related to MtAlaDH. However, based on the limited number of sequences available, and the enzymological data previously
Alteration of substrate specificity of alanine dehydrogenase (50 mM), EDTA (5 mM), lysozyme (1 mg/ml) and polymyxin B sulfate (500 μg/ml) and shaking at 250 rpm at room temperature for two hours. The plates were centrifuged at 4000 rpm for 10 min and the supernatant was used for screening. All variants in retest plates were screened in quadruplicate.
Purification of wild-type and single mutant enzymes
Library screening The libraries were screened by following the formation of NADH at 340 nm in a Spectromax 384 Plus UV/Vis plate reader. To each well in a UV clear flat bottom 96-well plate was added 43.3 μl of sodium carbonate buffer (150 mM, pH 10.2), 6.7 μl NAD+ (37.5 mM), 50 μl cell lysate and 100 μl of amino acid substrate. The substrates were dissolved in sodium carbonate buffer to concentrations of 100 mM for L-leucine, L-norleucine and L-alanine and 250 mM for L-norvaline. The reactions were monitored for 30 minutes. The initial library plates were screened against L-norleucine and L-norvaline. Retest plates were compiled based on the initial data and these variants were tested in quadruplicate against all four substrates.
Enzyme assays The kinetic properties of wild type and selected variant dehydrogenases were measured using the appearance of NADH (ε339 = 6220 M−1 cm−1) in a continuous spectroscopic assay monitored at 339 nm in a Spectromax384 Plus UV/vis plate reader. For the specific activity determinations, 0.4–1.5 nmol of enzyme were added to a reaction mixture containing the amino acid (100 mM; except 3.5 mM + L-homophenylalanine), NAD (5 mM), EDTA (3.25 mM) and sodium carbonate (100 mM, pH 10.2). The reaction was monitored for 5 minutes and the initial rate of reaction was recorded. Background activity was determined using reactions lacking enzyme; significant backgrounds (seen only for L-lysine and L-cysteine) were subtracted from the initial rates prior to data analysis. For determination of
Michaelis–Menten parameters, the amino acid substrate concentration was varied using seven values that span a 10-fold range. The reaction mixture also contained the enzyme, NAD+ (15 mM, ∼15•Km), sodium carbonate (100 mM), EDTA (3.25 mM) and pH 10.2. Reactions were monitored for 5 minutes and the initial rate of reaction was recorded. Averaged data from at least three trials were fit to the Michaelis–Menten equation (v = kcat [E][S]/(Km + [S]) (Cleland, 1979) to estimate kcat and Km using Kaleidagraph (Synergy Software, Reading, PA, USA) Since none of the enzyme–substrate pairs reached saturation, except for wild-type enzyme and alanine or serine, only the catalytic efficiency kcat/Km is reported. The highest tested concentration varied with the solubility of the amino acid in pH 10.2 buffer, from L-homophenylalanine (5.85 mM) to glycine (325 mM).
Results Superfamily analysis The two most comprehensive methods for enzyme superfamily classification are CATH and SCOP. The CATH classification (Class, Architecture, Topology, Homology) is based on a combination of manual procedures aided by structural comparison and hidden Markov model-based methods (Cuff et al., 2011). Members of CATH superfamilies are separated into functional families (FunFams). AlaDH is classified in the 3.40.50.1770 CATH superfamily in a FunFam along with NAD(P)(+) transhydrogenase, and saccharopine dehydrogenase FunFams. The Glu/Leu/Val/Phe dehydrogenase FunFams are classified in the CATH Superfamily 3.40.192.10 along with methylenetetrahydrofolate dehydrogenase/cyclohydrogenase and shikimate dehydrogenase which, unlike the rest of this superfamily, converts an alcohol to a carbonyl group (Supplementary Fig. S1A). SCOP, the Structural Classification of Proteins, classifies protein structures into superfamilies purely on structural similarities (Andreeva et al., 2007). There are four classifications in SCOP: superfamily, family, protein and species. AlaDH is classified in the AlaDH/ PNT family along with NAD(P) transhydrogenase and Saccharopine dehydrogenase. This family is classified in the formate–glycerate superfamily along with a variety of dehydrogenases that convert carbonyls to alcohols as well as the S-adenosylhomocysteine hydrolase family. The Glu/Val/Leu/Phe dehydrogenases are classified in the AADH family which is part of the AADH-like superfamily. Other members of this superfamily include methylenetetrahydromethanopterin dehydrogenase and tetrahydrofolate dehydrogenase, malic enzyme (a decarboxylase) and shikimate-5-dehydrogenase (which reduces a carbonyl to an alcohol) (Supplementary Fig. S1B). From this analysis, it is clear that regardless of how AlaDH is classified, it is the only member of its superfamily that carries out amino acid dehydrogenation.
Molecular modeling Molecular modeling using the crystal structure of MtAlaDH bound with pyruvate and NAD+ ( pdbid: 2VHX chain F) indicated that the sidechains of F94 and L130 would prevent larger substrates from binding through steric clashes. Models of all smaller residues were generated in these two positions with the exception of glycine (to avoid increased flexibility). In both positions alanine and serine appeared the most promising. The serine side chain modeled in both positions is rotated to H-bond away from the substrate binding pocket. The mutations of F94 to alanine and serine appeared to be
Downloaded from http://peds.oxfordjournals.org/ at Selcuk University on January 19, 2015
The genes for the wild-type MtAlaDH, F94A, F94S, L130A and L130S were purchased from DNA2.0 in the PJ401express vector with a C-terminal His6 tag. The mutants were transformed into chemically competent BL21 Escherichia coli cells (Sigma, St. Louis, MO, USA). To express the enzymes, 5 ml of an overnight growth was used to inoculate 1 l of LB broth with kanamycin (30 μg/ml). The cultures were incubated at 37°C while shaking at 250 rpm. Expression of the AlaDH gene or its variants was induced by the addition of 1 mM IPTG at a culture OD600 of 0.6–0.9 and cells were harvested via centrifugation at 5000 × g three hours later. The harvested bacteria were lysed by sonication and the lysate was clarified by centrifugation at 10 000 × g. All enzymes were purified by batch purification with Ni-NTA (Sambrook and Russell, 2006). After binding for one hour at 4°C, the protein-bound resin was poured into an empty column and was washed successively with 10 column volumes each of 20 and 100 mM imidazole; the protein was eluted with 250 mM imidazole. Purity of the collected enzyme-containing fractions was determined by sodium dodecyl sulphate-polyacrylamide gel electrophoresis (SDS-PAGE) (10% acrylamide). The enzyme concentration was determined by ultraviolet (UV) spectroscopy (ε280 = 25 900 M−1 cm−1). Pure enzymes were stored in 50 mM Tris–HCl, pH 7.2, 50 mM NaCl and 5 mM EDTA, pH 7.2 at 4°C and were used within three days of purification. For long-term storage, glycerol (25%) was added as a cryoprotectant for storage at −80°C, and thawed samples were extensively dialyzed against storage buffer prior to assay.
3
P. Fernandes et al.
4
Fig. 1. Modeling the AlaDH substrate binding pocket. (A) The wild-type AlaDH structure, 2VHX chain F, with NAD+ and pyruvate bound. (B) Molecular modeling of 4-methyl-2-oxopentanoic acid (the product of oxidative deamination of L-leucine) in the wild-type AlaDH. A steric clash is evident with F94. (C) Mutation F94S relieves this clash. All modeling was done with Discovery Studio 3.0 (Accelrys, 2012).
Wild-type
Ala,
L-alanine;
Ser,
L-serine;
F94A
Cys,
L-cysteine;
Val,
L-valine;
hSer,
F94S or F94S/Y117L
L-homoserine;
Ile,
L-isoleucine;
Leu,
L-leucine;
Phe,
L-phenylalanine;
Orn,
L-ornithine;
Lys,
L-lysine; AEA, L-allysine ethylene acetal; nVal, L-norvaline; APE, L-2-amino-4-pentenoic acid; nLeu, L-norleucine; Met, L-methionine; hPhe, L-homophenylalanine.
especially promising as larger substrates were able to be manually modeled in a reactive conformation without steric clashes (Fig. 1).
Substrate activity screen of wild-type and mutant AlaDH To test the hypothesis that replacement of F94 and L130 would admit larger substrates in AlaDH, the wild-type MtAlaDH and the mutants F94S, F94A, L130A and L130S were constructed by whole-gene synthesis, expressed in E. coli, and tested for in vitro activity. Expression of either L130 mutant resulted in protein in the insoluble fraction after lysis (tested at 20, 30 and 37°C expression temperatures), which suggested that this position is important for protein folding. Both F94S and F94A expressed solubly and were purified to >95% purity based on SDS-PAGE gel analysis. A specific activity assay was carried out to identify substrates whose reactivities increase upon mutation of the AlaDH active site. The assay revealed that the wild-type AlaDH is more promiscuous than previously
reported (Table I and Fig. 2; Supplementary Table SI). Its specific activity against non-native substrates (at 100 mM) L-cysteine, L-norvaline, L-valine, L-2-amino-4-pentenoic acid, L-serine and L-homoserine ranged from 0.2% (L-cysteine) to 2% (L-homoserine) of the activity with L-alanine. This group includes small, polar, and nonpolar amino acids. Although the branched amino acid L-valine showed a very small degree of activity, no activity was seen against isoleucine, highlighting the steric limitations of the wild-type binding pocket. The single mutant enzyme variants F94A and F94S demonstrated activity toward L-norleucine, L-leucine, L-methionine, L-homophenylalanine (screened at 3.5 mM) and L-allysine ethylene acetal, whereas none of these activities were seen with wild-type enzyme (Table I and Fig. 2). Promiscuous activity displayed by the wild-type AlaDH toward L-cysteine, L-valine, L-serine and L-homoserine was diminished by the mutations, whereas activity toward L-norvaline and L-2-amino-4-pentenoic acid increased in at least one single
Downloaded from http://peds.oxfordjournals.org/ at Selcuk University on January 19, 2015
Table I. Preferred enzyme for amino acid substrates in specific activity screen
Alteration of substrate specificity of alanine dehydrogenase
5
mutant variant. Activity toward the native substrate, L-alanine, was decreased by three orders of magnitude by these mutations, yet the residual activity against L-alanine was comparable to the novel activities against L-leucine and L-methionine found in these variants. These results suggested that the mutations broadened substrate specificity. Notably, glycine did not appear to be a substrate of the wild-type enzyme (nor of any variants described here) under the conditions of the specific activity assay; MtAlaDH has been shown to form glycine from glyoxylate and ammonium, but not to carry out the reverse reaction (Giffin et al., 2012).
Saturation mutagenesis To improve the catalytic efficiency toward the new substrates identified by our specific activity screen of the single mutants, saturation mutagenesis libraries were made using F94S as a backbone. All positions comprising the amino acid binding pocket, but without a defined catalytic role, were chosen for saturation mutagenesis, including Y117, L130, M133 and L313. The libraries were screened (>99% coverage) initially against L-norleucine and L-norvaline; hits were retested in quadruplicate with L-norleucine, L-norvaline, L-alanine and L-leucine. Variants in Y117X library showed improvements towards L-norleucine, L-norvaline and L-leucine with Y117L being the most active. The F94S/Y117L variant showed higher activity toward these substrates in the specific activity screen (Fig. 2), as well as L-homophenylalanine and L-allysine ethylene acetal (Table I). Modeling of the Y117L mutant suggests that the leucine side chain may create beneficial hydrophobic interactions with the larger amino acid substrates (Table I). It is possible that the original F94S mutation created too large of a binding area for the substrate and Y117L mutant was able to create more positive interactions. The L130X library produced a handful of variants whose activities were similar to the F94S positive control, suggesting that this position does not tolerate
substitution well, confirming the single mutant results. Variants from libraries M133X and L313X retained partial activity, but none were more active than the backbone. The original mutation position, F94, was also subjected to saturation mutagenesis. This library identified F94A and F94S as the only variants active against L-norleucine, and also confirmed the wild-type activity against L-alanine and L-norvaline.
Initial-rate kinetics To determine the mechanistic basis for the observed activity differences among the variants with different amino acids, initial-rate kinetic measurements were carried out (Tables II and III; Supplementary Fig. S2). The concentration of NAD+ was held at a fixed, saturating concentration (15 mM, ∼15•Km) and the amino acid concentration was varied over a 10-fold range, limited by either its solubility in carbonate buffer or an arbitrary maximum of 325 mM. Under these conditions, only the wild-type dehydrogenase could be saturated with amino acid (L-alanine and L-serine) (Table II); the kinetic constants for L-alanine agreed well with previously determined values (Giffin et al., 2012). All other enzyme–substrate combinations yielded velocity vs. substrate curves that did not saturate; therefore, only the catalytic efficiency (kcat/Km) is reported. Catalytic efficiency values can be used to compare substrate specificity among enzyme variants. For the purposes of this study, catalytic efficiency values were determined only for substrates whose activity increased upon mutation. The trend in catalytic efficiency from F94S to F94S/Y117L generally mirrors that of the specific activity, because both the initial-rate and activity screen studies were carried out at subsaturating concentrations of amino acid. For most substrates showing improvement (L-norvaline, L-norleucine, L-leucine, L-homophenylalanine), the additional Y117L mutation approximately doubles both the specific activity and catalytic efficiency. L-Methionine showed a slight decrease in
Downloaded from http://peds.oxfordjournals.org/ at Selcuk University on January 19, 2015
Fig. 2. Specific activity of AlaDH variants with amino acid substrates. Initial reaction rates of oxidative deamination were measured at 339 nm and normalized to enzyme concentration. Conditions: 100 mM L-amino acid (3.25 mM for L-homophenylalanine), 5 mM NAD+, 2–7 µM enzyme (monomer), 3.25 mM EDTA, 100 mM sodium carbonate, pH 10.2. No activity was observed for any of the enzymes with glycine, L-histidine, L-glutamate, L-glutamine or L-2-aminoadipic acid.
P. Fernandes et al.
6
Table II. Kinetics of wild-type AlaDH −1
Amino acid
Km (mM)
kcat (s )
L-Alanine
5.3 ± 0.8 – 24 ± 2 – –
7.5 ± 0.4 – 0.24 ± 0.01 – –
L-Homoserine L-Serine L-Valine L-2-Amino-4-pentenoic
kcat/Km (M−1 s−1) 1400 ± 200 15 ± 2 10 ± 1 4.6 ± 0.5 1.5 ± 0.1
acid L-Norvaline
–
–
0.40 ± 0.07
Table III. Catalytic efficiency of AlaDH variants Amino acid
L-2-Amino-4-pentenoic
F94S (M−1 s−1)
F94S/Y117L (M−1 s−1)
1400 ± 200 1.5 ± 0.1
2.2 ± 0.1 4±1
0.35 ± 0.05 5±1
0.40 ± 0.07
Original Article
Alteration of substrate specificity of alanine dehydrogenase Puja Fernandes1,†, Hannah Aldeborgh2,†, Lauren Carlucci1, Lauren Walsh1, Jordan Wasserman1, Edward Zhou1, Scott T. Lefurgy1, and Emily C. Mundorff1,* Downloaded from http://peds.oxfordjournals.org/ at Selcuk University on January 19, 2015
1
Chemistry Department, Hofstra University, Hempstead, NY 11549, USA, and 2Chemistry Department, Vassar College, Poughkeepsie, NY 12604, USA *To whom correspondence should be addressed. E-mail: [email protected] † These authors contributed equally to this work. Edited by Dan Tawfik Received 21 July 2014; Revised 30 October 2014; Accepted 19 November 2014
Abstract The L-alanine dehydrogenase (AlaDH) has a natural history that suggests it would not be a promising candidate for expansion of substrate specificity by protein engineering: it is the only amino acid dehydrogenase in its fold family, it has no sequence or structural similarity to any known amino acid dehydrogenase, and it has a strong preference for L-alanine over all other substrates. By contrast, engineering of the amino acid dehydrogenase superfamily members has produced catalysts with expanded substrate specificity; yet, this enzyme family already contains members that accept a broad range of substrates. To test whether the natural history of an enzyme is a predictor of its innate evolvability, directed evolution was carried out on AlaDH. A single mutation identified through molecular modeling, F94S, introduced into the AlaDH from Mycobacterium tuberculosis (MtAlaDH) completely alters its substrate specificity pattern, enabling activity toward a range of larger amino acids. Saturation mutagenesis libraries in this mutant background additionally identified a double mutant (F94S/Y117L) showing improved activity toward hydrophobic amino acids. The catalytic efficiencies achieved in AlaDH are comparable with those that resulted from similar efforts in the amino acid dehydrogenase superfamily and demonstrate the evolvability of MtAlaDH specificity toward other amino acid substrates. Key words: alanine dehydrogenase, amino acid dehydrogenase, evolvability, saturation mutagenesis, superfamily
Introduction It has become clear that some enzymes are more evolvable than others (Aharoni et al., 2005; Bloom et al., 2006; O’Loughlin et al., 2006). The determination of this evolvability trait has been the target of recent work (Dellus-Gur et al., 2013), but it is still unclear how to predict the capacity of a specific enzyme to evolve new activities or substrate specificities. One assumption is that the evolutionary history of an enzyme is a predictor of its evolutionary potential (O’Loughlin et al., 2006; Romero and Arnold, 2009). This principle is based on the idea that an enzyme derived from a fold family that has evolved
multiple new functions in nature would be more amenable to the development of new activities in the laboratory. However, this assumption ignores the multitude of constraints that may limit the evolutionary potential of an enzyme family in natural evolution, such as regulatory strategies, expression levels, metabolic flux and functional non-redundance. Such constraints disappear once the enzyme is in the artificial environment of a laboratory (Arnold et al., 2001). Also, given that we do not have a complete evolutionary tree, our understanding of an enzyme’s natural history may be misleading. As a counterexample to the natural evolvability argument, we present the
© The Author 2014. Published by Oxford University Press. All rights reserved. For Permissions, please e-mail: [email protected]
1
P. Fernandes et al.
2
presented (Hutter and Singh, 1999), the MtAlaDH would be considered a highly specific enzyme from a highly conserved family that appears to be unrelated to other AADHs. If the apparent evolutionary history is a guide to evolvability, one would reasonably expect to generate novel amino acid substrate activity more easily within an enzyme in the AADH superfamily than in MtAlaDH. We report here a single mutation of AlaDH (F94S) that switches its specificity from vastly preferring alanine to being generally promiscuous toward hydrophobic amino acids (L-norleucine, L-leucine, L-methionine, L-norvaline, L-2-amino-4-pentenoic acid, L-lysine, L-ornithine, L-serine, L-homoserine, L-homophenylalanine and L-allysine ethylene acetal). Saturation mutagenesis libraries in this mutant background additionally identified a double mutant (F94S/Y117L) showing improved activity toward L-norleucine, L-leucine, L-norvaline, L-methionine and L-homophenylalanine, and further depression of activity toward the native substrate L-alanine. The results obtained are comparable with those found within the amino acid superfamily and demonstrate the evolvability of MtAlaDH specificity to other amino acid substrates.
Materials and methods Materials All chemicals were purchased from Sigma-Aldrich (St. Louis, MO) at reagent grade or higher.
Molecular modeling The crystal structure of MtAlaDH with NAD+ and pyruvate bound ( pdbid: 2VHX) was used as the basis for all modeling studies using the software package, Discovery Studio 3.0 (Accelrys, San Diego, CA, USA). Positions to test were identified visually as those residues with side chains pointing toward the side chain of the substrate, Phe94 and Leu130. Mutations were made using the protein mutate function and models were minimized to an RMS gradient of less than 0.02 kcal/(mol × Å) with the generalized Born with molecular volume solvent model. All models were screened by manually modeling 4-methyl-2-oxopentanoic acid (the product of oxidative deamination of L-leucine) in a reactive conformation to identify those mutants that would best accommodate this substrate.
Generation of saturation mutagenesis libraries Site-directed saturation mutagenesis libraries were made using the QuikChange Lightning Site-Directed Mutagenesis kit (Agilent, Santa Clara, CA, USA) and the enclosed protocol was followed with the exception of a 1-hour DpnI digestion. The polymerase chain reaction products were transformed into XL-10 gold cells and selected for antibiotic resistance on LB agar with 30 μg/ml kanamycin. Colonies that appeared in 24–36 h were collected into 3 ml of LB broth and the plasmids were extracted using a spin miniprep kit from Qiagen (Valencia, CA). The miniprep plasmid DNA was used to transform BL21 competent cells and transformants were selected on LB-kanamycin agar plates. The colonies were picked by toothpicks, 176 per library, transferred to two 96-well plates, and grown overnight shaking at 200 rpm at 30°C with four wells each of positive (F94S) and negative (empty vector) controls. 180 μl of LB broth was inoculated with 10 μl of overnight growth and grown for 3 h at 30°C shaking at 200 rpm. The plates were induced with 1 mM β-D-isopropylthiogalactoside (IPTG), shaken at 200 rpm for ∼18 h at 30°C, and then centrifuged at 4000 rpm for 10 min. Lysis of the cell pellets commenced through resuspension in 120 μl per well in Tris–HCl (50 mM, pH7.2), NaCl
Downloaded from http://peds.oxfordjournals.org/ at Selcuk University on January 19, 2015
alteration of substrate specificity of L-alanine dehydrogenase (AlaDH), an enzyme from a family that looks particularly unpromising if one were to judge it based purely on its natural history. The members of the amino acid dehydrogenase (AADH) superfamily include the glutamate dehydrogenase (GluDH), leucine dehydrogenase (LeuDH), valine dehydrogenase and phenylalanine dehydrogenase (PheDH) families, but not AlaDH. The member families of the AADH superfamily have a long and rich history of being targeted in protein engineering to alter substrate specificity due to its members’ usefulness in diagnostic kits and in the biocatalysis of chiral amines and unnatural amino acids (Brunhuber and Blanchard, 1994). There is a clear evolutionary relationship among the AADH families, which demonstrate homology on the basis of sequence and structural similarities (Britton et al., 1993; Baker et al., 1997). Researchers have exploited the conserved features of these enzymes to alter their substrate specificities using rational design (Wang et al., 2001; Seah et al., 2003). Guided by homology and the hypothesis that certain positions in the enzyme exert discrete control over substrate specificity, researchers were able to successfully transform both a GluDH (Wang et al., 2001) and a PheDH (Kataoka and Tanizawa, 2003) into a L-leucine/L-norleucine dehydrogenase using three or fewer mutations. By contrast, AlaDH enzymes share no apparent evolutionary relationship with the AADH superfamily (Baker et al., 1998). There is no sequence similarity between AlaDH and any member of the AADH superfamily beyond both having the NAD+/NADH-binding Rossmann fold. Mechanistically, the AlaDH and PheDH differ in that they donate the hydride to opposite sides of the NAD+ moiety, indicating convergent evolution. AlaDH is classified in the same superfamily as NAD(P) transhydrogenase (AlaDH_PNT superfamily) due to the sequence and structural similarities between the two enzyme families (Jackson et al., 1999). The two families bind the adenosine moiety of NAD+ similarly, but bind the nicotinamide in different regions of the structure, such that the catalytic residues are not in structurally conserved positions. The only other family with significant structural similarity to AlaDH is saccharopine dehydrogenase (NAD+, L-lysine forming) (SacDH) (Burk et al., 2007). Representative structures in the PDB showed 14% sequence identity and 2.67 Å RMSD (PDBID:4LMP for MtAlaDH and PDBID:2Q99 for SacDH from Saccharomyces cerevisiae). The two enzymes share a common partial mechanism and three conserved catalytic residues, but perform overall different reactions (Burk et al., 2007). SacDH catalyzes the oxidative deamination of D-saccharopine to generate L-lysine and α-ketoglutarate; SacDH releases a primary amine, whereas the oxidative deamination of L-alanine by AlaDH releases ammonia. AlaDH has been studied from a variety of bacteria, generally maintaining ∼50% sequence identity among family members (Baker et al., 1998). In Bacillus subtilis it is known to be essential for the utilization of alanine as a sole carbon and nitrogen source and is required for normal sporulation (Siranosian et al., 1993). Increased levels of AlaDH have also been associated with both the maintenance of the NAD+ pool as well as generation of precursors for peptidoglycan biosynthesis (Hutter and Dick, 1998; Hutter and Singh, 1999; Betts et al., 2002; Feng et al., 2002). Why this enzyme family has remained highly conserved is unclear, but there appears to be some constraint on its evolution. It is also interesting that there is not a specific alanine dehydrogenase within the AADH superfamily. It is possible, of course, that an alanine dehydrogenase exists within the AADH superfamily but has yet to be identified, or that there are other, unknown AADHs related to MtAlaDH. However, based on the limited number of sequences available, and the enzymological data previously
Alteration of substrate specificity of alanine dehydrogenase (50 mM), EDTA (5 mM), lysozyme (1 mg/ml) and polymyxin B sulfate (500 μg/ml) and shaking at 250 rpm at room temperature for two hours. The plates were centrifuged at 4000 rpm for 10 min and the supernatant was used for screening. All variants in retest plates were screened in quadruplicate.
Purification of wild-type and single mutant enzymes
Library screening The libraries were screened by following the formation of NADH at 340 nm in a Spectromax 384 Plus UV/Vis plate reader. To each well in a UV clear flat bottom 96-well plate was added 43.3 μl of sodium carbonate buffer (150 mM, pH 10.2), 6.7 μl NAD+ (37.5 mM), 50 μl cell lysate and 100 μl of amino acid substrate. The substrates were dissolved in sodium carbonate buffer to concentrations of 100 mM for L-leucine, L-norleucine and L-alanine and 250 mM for L-norvaline. The reactions were monitored for 30 minutes. The initial library plates were screened against L-norleucine and L-norvaline. Retest plates were compiled based on the initial data and these variants were tested in quadruplicate against all four substrates.
Enzyme assays The kinetic properties of wild type and selected variant dehydrogenases were measured using the appearance of NADH (ε339 = 6220 M−1 cm−1) in a continuous spectroscopic assay monitored at 339 nm in a Spectromax384 Plus UV/vis plate reader. For the specific activity determinations, 0.4–1.5 nmol of enzyme were added to a reaction mixture containing the amino acid (100 mM; except 3.5 mM + L-homophenylalanine), NAD (5 mM), EDTA (3.25 mM) and sodium carbonate (100 mM, pH 10.2). The reaction was monitored for 5 minutes and the initial rate of reaction was recorded. Background activity was determined using reactions lacking enzyme; significant backgrounds (seen only for L-lysine and L-cysteine) were subtracted from the initial rates prior to data analysis. For determination of
Michaelis–Menten parameters, the amino acid substrate concentration was varied using seven values that span a 10-fold range. The reaction mixture also contained the enzyme, NAD+ (15 mM, ∼15•Km), sodium carbonate (100 mM), EDTA (3.25 mM) and pH 10.2. Reactions were monitored for 5 minutes and the initial rate of reaction was recorded. Averaged data from at least three trials were fit to the Michaelis–Menten equation (v = kcat [E][S]/(Km + [S]) (Cleland, 1979) to estimate kcat and Km using Kaleidagraph (Synergy Software, Reading, PA, USA) Since none of the enzyme–substrate pairs reached saturation, except for wild-type enzyme and alanine or serine, only the catalytic efficiency kcat/Km is reported. The highest tested concentration varied with the solubility of the amino acid in pH 10.2 buffer, from L-homophenylalanine (5.85 mM) to glycine (325 mM).
Results Superfamily analysis The two most comprehensive methods for enzyme superfamily classification are CATH and SCOP. The CATH classification (Class, Architecture, Topology, Homology) is based on a combination of manual procedures aided by structural comparison and hidden Markov model-based methods (Cuff et al., 2011). Members of CATH superfamilies are separated into functional families (FunFams). AlaDH is classified in the 3.40.50.1770 CATH superfamily in a FunFam along with NAD(P)(+) transhydrogenase, and saccharopine dehydrogenase FunFams. The Glu/Leu/Val/Phe dehydrogenase FunFams are classified in the CATH Superfamily 3.40.192.10 along with methylenetetrahydrofolate dehydrogenase/cyclohydrogenase and shikimate dehydrogenase which, unlike the rest of this superfamily, converts an alcohol to a carbonyl group (Supplementary Fig. S1A). SCOP, the Structural Classification of Proteins, classifies protein structures into superfamilies purely on structural similarities (Andreeva et al., 2007). There are four classifications in SCOP: superfamily, family, protein and species. AlaDH is classified in the AlaDH/ PNT family along with NAD(P) transhydrogenase and Saccharopine dehydrogenase. This family is classified in the formate–glycerate superfamily along with a variety of dehydrogenases that convert carbonyls to alcohols as well as the S-adenosylhomocysteine hydrolase family. The Glu/Val/Leu/Phe dehydrogenases are classified in the AADH family which is part of the AADH-like superfamily. Other members of this superfamily include methylenetetrahydromethanopterin dehydrogenase and tetrahydrofolate dehydrogenase, malic enzyme (a decarboxylase) and shikimate-5-dehydrogenase (which reduces a carbonyl to an alcohol) (Supplementary Fig. S1B). From this analysis, it is clear that regardless of how AlaDH is classified, it is the only member of its superfamily that carries out amino acid dehydrogenation.
Molecular modeling Molecular modeling using the crystal structure of MtAlaDH bound with pyruvate and NAD+ ( pdbid: 2VHX chain F) indicated that the sidechains of F94 and L130 would prevent larger substrates from binding through steric clashes. Models of all smaller residues were generated in these two positions with the exception of glycine (to avoid increased flexibility). In both positions alanine and serine appeared the most promising. The serine side chain modeled in both positions is rotated to H-bond away from the substrate binding pocket. The mutations of F94 to alanine and serine appeared to be
Downloaded from http://peds.oxfordjournals.org/ at Selcuk University on January 19, 2015
The genes for the wild-type MtAlaDH, F94A, F94S, L130A and L130S were purchased from DNA2.0 in the PJ401express vector with a C-terminal His6 tag. The mutants were transformed into chemically competent BL21 Escherichia coli cells (Sigma, St. Louis, MO, USA). To express the enzymes, 5 ml of an overnight growth was used to inoculate 1 l of LB broth with kanamycin (30 μg/ml). The cultures were incubated at 37°C while shaking at 250 rpm. Expression of the AlaDH gene or its variants was induced by the addition of 1 mM IPTG at a culture OD600 of 0.6–0.9 and cells were harvested via centrifugation at 5000 × g three hours later. The harvested bacteria were lysed by sonication and the lysate was clarified by centrifugation at 10 000 × g. All enzymes were purified by batch purification with Ni-NTA (Sambrook and Russell, 2006). After binding for one hour at 4°C, the protein-bound resin was poured into an empty column and was washed successively with 10 column volumes each of 20 and 100 mM imidazole; the protein was eluted with 250 mM imidazole. Purity of the collected enzyme-containing fractions was determined by sodium dodecyl sulphate-polyacrylamide gel electrophoresis (SDS-PAGE) (10% acrylamide). The enzyme concentration was determined by ultraviolet (UV) spectroscopy (ε280 = 25 900 M−1 cm−1). Pure enzymes were stored in 50 mM Tris–HCl, pH 7.2, 50 mM NaCl and 5 mM EDTA, pH 7.2 at 4°C and were used within three days of purification. For long-term storage, glycerol (25%) was added as a cryoprotectant for storage at −80°C, and thawed samples were extensively dialyzed against storage buffer prior to assay.
3
P. Fernandes et al.
4
Fig. 1. Modeling the AlaDH substrate binding pocket. (A) The wild-type AlaDH structure, 2VHX chain F, with NAD+ and pyruvate bound. (B) Molecular modeling of 4-methyl-2-oxopentanoic acid (the product of oxidative deamination of L-leucine) in the wild-type AlaDH. A steric clash is evident with F94. (C) Mutation F94S relieves this clash. All modeling was done with Discovery Studio 3.0 (Accelrys, 2012).
Wild-type
Ala,
L-alanine;
Ser,
L-serine;
F94A
Cys,
L-cysteine;
Val,
L-valine;
hSer,
F94S or F94S/Y117L
L-homoserine;
Ile,
L-isoleucine;
Leu,
L-leucine;
Phe,
L-phenylalanine;
Orn,
L-ornithine;
Lys,
L-lysine; AEA, L-allysine ethylene acetal; nVal, L-norvaline; APE, L-2-amino-4-pentenoic acid; nLeu, L-norleucine; Met, L-methionine; hPhe, L-homophenylalanine.
especially promising as larger substrates were able to be manually modeled in a reactive conformation without steric clashes (Fig. 1).
Substrate activity screen of wild-type and mutant AlaDH To test the hypothesis that replacement of F94 and L130 would admit larger substrates in AlaDH, the wild-type MtAlaDH and the mutants F94S, F94A, L130A and L130S were constructed by whole-gene synthesis, expressed in E. coli, and tested for in vitro activity. Expression of either L130 mutant resulted in protein in the insoluble fraction after lysis (tested at 20, 30 and 37°C expression temperatures), which suggested that this position is important for protein folding. Both F94S and F94A expressed solubly and were purified to >95% purity based on SDS-PAGE gel analysis. A specific activity assay was carried out to identify substrates whose reactivities increase upon mutation of the AlaDH active site. The assay revealed that the wild-type AlaDH is more promiscuous than previously
reported (Table I and Fig. 2; Supplementary Table SI). Its specific activity against non-native substrates (at 100 mM) L-cysteine, L-norvaline, L-valine, L-2-amino-4-pentenoic acid, L-serine and L-homoserine ranged from 0.2% (L-cysteine) to 2% (L-homoserine) of the activity with L-alanine. This group includes small, polar, and nonpolar amino acids. Although the branched amino acid L-valine showed a very small degree of activity, no activity was seen against isoleucine, highlighting the steric limitations of the wild-type binding pocket. The single mutant enzyme variants F94A and F94S demonstrated activity toward L-norleucine, L-leucine, L-methionine, L-homophenylalanine (screened at 3.5 mM) and L-allysine ethylene acetal, whereas none of these activities were seen with wild-type enzyme (Table I and Fig. 2). Promiscuous activity displayed by the wild-type AlaDH toward L-cysteine, L-valine, L-serine and L-homoserine was diminished by the mutations, whereas activity toward L-norvaline and L-2-amino-4-pentenoic acid increased in at least one single
Downloaded from http://peds.oxfordjournals.org/ at Selcuk University on January 19, 2015
Table I. Preferred enzyme for amino acid substrates in specific activity screen
Alteration of substrate specificity of alanine dehydrogenase
5
mutant variant. Activity toward the native substrate, L-alanine, was decreased by three orders of magnitude by these mutations, yet the residual activity against L-alanine was comparable to the novel activities against L-leucine and L-methionine found in these variants. These results suggested that the mutations broadened substrate specificity. Notably, glycine did not appear to be a substrate of the wild-type enzyme (nor of any variants described here) under the conditions of the specific activity assay; MtAlaDH has been shown to form glycine from glyoxylate and ammonium, but not to carry out the reverse reaction (Giffin et al., 2012).
Saturation mutagenesis To improve the catalytic efficiency toward the new substrates identified by our specific activity screen of the single mutants, saturation mutagenesis libraries were made using F94S as a backbone. All positions comprising the amino acid binding pocket, but without a defined catalytic role, were chosen for saturation mutagenesis, including Y117, L130, M133 and L313. The libraries were screened (>99% coverage) initially against L-norleucine and L-norvaline; hits were retested in quadruplicate with L-norleucine, L-norvaline, L-alanine and L-leucine. Variants in Y117X library showed improvements towards L-norleucine, L-norvaline and L-leucine with Y117L being the most active. The F94S/Y117L variant showed higher activity toward these substrates in the specific activity screen (Fig. 2), as well as L-homophenylalanine and L-allysine ethylene acetal (Table I). Modeling of the Y117L mutant suggests that the leucine side chain may create beneficial hydrophobic interactions with the larger amino acid substrates (Table I). It is possible that the original F94S mutation created too large of a binding area for the substrate and Y117L mutant was able to create more positive interactions. The L130X library produced a handful of variants whose activities were similar to the F94S positive control, suggesting that this position does not tolerate
substitution well, confirming the single mutant results. Variants from libraries M133X and L313X retained partial activity, but none were more active than the backbone. The original mutation position, F94, was also subjected to saturation mutagenesis. This library identified F94A and F94S as the only variants active against L-norleucine, and also confirmed the wild-type activity against L-alanine and L-norvaline.
Initial-rate kinetics To determine the mechanistic basis for the observed activity differences among the variants with different amino acids, initial-rate kinetic measurements were carried out (Tables II and III; Supplementary Fig. S2). The concentration of NAD+ was held at a fixed, saturating concentration (15 mM, ∼15•Km) and the amino acid concentration was varied over a 10-fold range, limited by either its solubility in carbonate buffer or an arbitrary maximum of 325 mM. Under these conditions, only the wild-type dehydrogenase could be saturated with amino acid (L-alanine and L-serine) (Table II); the kinetic constants for L-alanine agreed well with previously determined values (Giffin et al., 2012). All other enzyme–substrate combinations yielded velocity vs. substrate curves that did not saturate; therefore, only the catalytic efficiency (kcat/Km) is reported. Catalytic efficiency values can be used to compare substrate specificity among enzyme variants. For the purposes of this study, catalytic efficiency values were determined only for substrates whose activity increased upon mutation. The trend in catalytic efficiency from F94S to F94S/Y117L generally mirrors that of the specific activity, because both the initial-rate and activity screen studies were carried out at subsaturating concentrations of amino acid. For most substrates showing improvement (L-norvaline, L-norleucine, L-leucine, L-homophenylalanine), the additional Y117L mutation approximately doubles both the specific activity and catalytic efficiency. L-Methionine showed a slight decrease in
Downloaded from http://peds.oxfordjournals.org/ at Selcuk University on January 19, 2015
Fig. 2. Specific activity of AlaDH variants with amino acid substrates. Initial reaction rates of oxidative deamination were measured at 339 nm and normalized to enzyme concentration. Conditions: 100 mM L-amino acid (3.25 mM for L-homophenylalanine), 5 mM NAD+, 2–7 µM enzyme (monomer), 3.25 mM EDTA, 100 mM sodium carbonate, pH 10.2. No activity was observed for any of the enzymes with glycine, L-histidine, L-glutamate, L-glutamine or L-2-aminoadipic acid.
P. Fernandes et al.
6
Table II. Kinetics of wild-type AlaDH −1
Amino acid
Km (mM)
kcat (s )
L-Alanine
5.3 ± 0.8 – 24 ± 2 – –
7.5 ± 0.4 – 0.24 ± 0.01 – –
L-Homoserine L-Serine L-Valine L-2-Amino-4-pentenoic
kcat/Km (M−1 s−1) 1400 ± 200 15 ± 2 10 ± 1 4.6 ± 0.5 1.5 ± 0.1
acid L-Norvaline
–
–
0.40 ± 0.07
Table III. Catalytic efficiency of AlaDH variants Amino acid
L-2-Amino-4-pentenoic
F94S (M−1 s−1)
F94S/Y117L (M−1 s−1)
1400 ± 200 1.5 ± 0.1
2.2 ± 0.1 4±1
0.35 ± 0.05 5±1
0.40 ± 0.07
