Mapping the Substrate Binding Site of Phenylacetone Monooxygenase from Thermobifida fusca by Mutational Analysis^,

Abstract

Baeyer-Villiger monooxygenases catalyze oxidations that are of interest for biocatalytic applications. Among these enzymes, phenylacetone monooxygenase (PAMO) from Thermobifida fusca is the only protein showing remarkable stability. While related enzymes often present a broad substrate scope, PAMO accepts only a limited number of substrates. Due to the absence of a substrate in the elucidated crystal structure of PAMO, the substrate binding site of this protein has not yet been defined. In this study, a structural model of cyclopentanone monooxygenase, which acts on a broad range of compounds, has been prepared and compared with the structure of PAMO. This revealed 15 amino acid positions in the active site of PAMO that may account for its relatively narrow substrate specificity. We designed and analyzed 30 single and multiple mutants in order to verify the role of these positions. Extensive substrate screening revealed several mutants that displayed increased activity and altered regio- or enantioselectivity in Baeyer-Villiger reactions and sulfoxidations. Further substrate profiling resulted in the identification of mutants with improved catalytic properties toward synthetically attractive compounds. Moreover, the thermostability of the mutants was not compromised in comparison to that of the wild-type enzyme. Our data demonstrate that the positions identified within the active site of PAMO, namely, V54, I67, Q152, and A435, contribute to the substrate specificity of this enzyme. These findings will aid in more dedicated and effective redesign of PAMO and related monooxygenases toward an expanded substrate scope.

INTRODUCTION

Enzymes have been gaining increasing attention as efficient and selective catalysts to be used in synthetic chemistry. Baeyer-Villiger monooxygenases (BVMOs) comprise a group of enzymes that are particularly interesting for synthetic applications. These biocatalysts employ molecular oxygen as a mild oxidant to oxidize carbonylic compounds. Apart from catalyzing Baeyer-Villiger reactions, BVMOs are capable of oxidizing a range of heteroatoms (e.g., sulfur, nitrogen, and boron). Furthermore, they often perform these reactions with high chemo-, regio-, and enantioselectivity (5, 15, 30). The use of oxygen, which is a cheap and clean oxidant, and their diversity of catalyzed reactions make BVMOs attractive candidates for biocatalytic processes.

The identification of phenylacetone monooxygenase (PAMO) in the moderately thermophilic bacterium Thermobifida fusca has brought about a breakthrough in the research on BVMOs (11). PAMO is a thermostable enzyme that can easily be expressed in Escherichia coli and purified. Besides, its crystal structure has been solved as the first structure of a BVMO (17). While PAMO shows excellent stability, even in the presence of organic solvents (6, 28), its substrate specificity is rather restricted. The enzyme accepts mainly small aromatic ketones and sulfides (7, 26), whereas the oxidations of bulkier ketones occur with lower activity and selectivity (27). However, the unique robustness of PAMO and the availability of its three-dimensional structure could facilitate the redesign of this enzyme in order to alter the substrate scope. It would be valuable to obtain PAMO variants displaying the same stability but presenting a relaxed substrate specificity and, thus, accepting alicyclic or bulky substrates.

Members of different enzyme classes, including BVMOs, have been successfully modified by directed evolution. Cyclohexanone monooxygenase (CHMO) from Acinetobacter sp. NCIB 9871 (3) was subjected to random mutagenesis, and mutants with improved enantioselectivity were identified (20, 21). In a similar fashion, a BVMO from Pseudomonas fluorescens DSM 50106 was evolved toward increased enantioselectivity (16). The elucidation of the crystal structure of PAMO has laid the foundation for more rational engineering studies. A homology model of cyclopentanone monooxygenase (CPMO) from Comamonas sp. strain NCIMB 9872 (13) based on the structure of PAMO supported a redesign study of CPMO by which the enantioselectivity of the enzyme could be improved (4). Recently, the Reetz group described a few cases in which residues in the active site of PAMO were targeted by semirandom mutagenesis (22, 23, 33). They obtained a number of mutants that act on substrates not accepted by the wild-type PAMO. As an example of rational engineering of PAMO, shortening an active-site loop by the deletion of residues S441 and A442 resulted in a variant accepting bulkier substrates (2). Moreover, a model of CPMO built on the basis of the PAMO structure inspired the preparation of the mutant M446G, which turned out to have a different substrate scope than the wild-type enzyme (31).

It should be noted that the molecular basis of the narrow substrate specificity of PAMO is not clear. The crystal structure of PAMO lacks any substrate, product, or inhibitor from which the binding site for substrates can be inferred. Taking into account the complex catalytic mechanism of BVMOs, probably involving domain movements (19, 29), it is hard to speculate on how substrates enter and bind in the active site. Some clues about the substrate binding site originated from previous mutagenesis studies. Residues belonging to the active-site loop spanning the region 440 to 446 have been repeatedly found to play a role in the substrate specificity and the enantioselectivity of PAMO (2, 22, 23, 31). On the other hand, examination of the crystal structure reveals that this region probably forms only part of the binding site. Nevertheless, detailed knowledge of the residues interacting with the substrate in the active site is necessary to efficiently alter properties such as the substrate range or the enantioselectivity of PAMO by enzyme engineering.

To address the above-described problem, we strove for a detailed comparison of PAMO and CPMO structures. The latter enzyme can catalyze Baeyer-Villiger oxidations of numerous aliphatic and aromatic substrates, including small alicyclic ketones, which are not accepted by PAMO (13). We prepared a homology model of CPMO based on the crystal structure of PAMO. A comprehensive inspection of the active sites of these two enzymes let us identify 15 residues that differed between PAMO and CPMO (Fig. 1). We hypothesized that a subset of these residues is involved in the substrate recognition. In a systematic site-directed mutagenesis study, we have mapped the active site of PAMO with respect to substrate specificity. By exchanging amino acids at the selected positions for their counterparts from CPMO, several mutants with significantly altered substrate scope and enantioselectivity have been identified. These findings indicate that the respective residues interact with substrates during catalysis. Better understanding of the substrate specificity determinants in PAMO opens up new routes for more effective redesign of this enzyme.

Fig. 1.

Active-site residues of PAMO targeted in the mutagenesis study. FAD is shown in black. For clarity, only the side chains of amino acid residues are presented. Proposed mutations are indicated. The schematic was prepared using PyMol software and the structure of PAMO (PDB ID 1W4X_A).

MATERIALS AND METHODS

Enzymes and reagents.

Oligonucleotide primers for mutagenesis were purchased from Sigma. PfuTurbo DNA polymerase was from Stratagene. DpnI was from New England BioLabs. E. coli TOP10 cells were obtained from Invitrogen. Yeast extract and Bacto tryptone for preparation of culture media were purchased from Becton, Dickinson & Company. Substrates 1 to 8, 10, and 12 to 14 (see Fig. 3), as well as ethyl benzoate (product 1a), δ-valerolactone (product 3a), and other chemicals, were obtained from Acros Organics, Sigma-Aldrich, Merck, Julich Chiral Solutions GmbH, Clontech, and Roche Diagnostics GmbH. Sulfide 9 was prepared as described previously (25), while ketone 11 was synthesized by treating 1-(4-chlorophenyl)propan-2-one with methyl iodide and NaOH in a biphasic water/CH₂Cl₂ system. Racemic sulfoxides 5a to 8a were prepared by treatment of the starting sulfides with H₂O₂ in methanol at room temperature (yields higher than 80%). Compound 11a was obtained by acetylation of 1-(4-chlorophenyl)ethanol with acetic anhydride and pyridine. Ketoester 12a was achieved according to the literature (7). His-tagged phosphite dehydrogenase used for regeneration of NADPH was overexpressed in E. coli TOP10 from a pBAD plasmid containing a codon-optimized ptxD gene and purified on a Ni²⁺-Sepharose column (GE Healthcare).

Fig. 3.

Substrates used in the whole-cell screening (A) and in conversions with the isolated enzymes (B).

Homology modeling and structural analysis.

The sequence of CPMO retrieved from GenBank (gi∣62286566) was submitted to the Protein Structure Prediction MetaServer (http://meta.bioinfo.pl) (12), which joins various homology modeling and threading algorithms. The results obtained were ranked according to the 3D-Jury method as described previously (12). The mappings obtained were submitted to Modeller (9).

Construction of PAMO mutants.

Single mutants were prepared by QuikChange site-directed mutagenesis using as a template a modified pBAD plasmid containing the pamO gene fused to a C-terminal His tag (11). The construction of the mutants with mutations M446G, Q152F/L153G, and Q152F/L153G/M446G has been reported before (31). The mutagenic primers used in this study are listed in Table S1 in the supplemental material. Prior to transformation, a DpnI-treated reaction mixture was purified using a NucleoSpin extract II kit (Macherey-Nagel). The introduced mutations were verified by sequencing (GATC Biotech, Konstanz, Germany). Mutations of two or three neighboring residues were typically introduced in one QuikChange reaction. The Q152F mutant was obtained by curing the mutation of L153G in the Q152F/L153G mutant. Multiple mutants were constructed by introducing mutations in consecutive rounds of mutagenesis. In each round, a plasmid with the verified sequence was used as a template. The triple mutant with mutations Q152F/L153G/M446G was a starting point for the construction of the 8-fold mutant (V54I/C65V/I67T/Q93W/Q152F/L153G/I339S/M446G) and the 7-fold mutant (Q152F/L153G/S441A/A442G/S444C/M446G/L447P). The 12-fold mutant was obtained by recombination of the 8-fold mutant with the 7-fold mutant using PvuI sites. The 10- and 11-fold mutants were achieved by reversing the Q152F and/or L153G mutations in the 12-fold mutant.

Screening of mutants.

The activities of the mutants on ketones 1 to 4 were tested in reactions with whole cells. For these reactions, E. coli strains transformed with plasmids containing the wild-type or mutant pamO genes were grown in CellStar 24-well plates (Greiner Bio-One). First, 1 ml of Luria-Bertani medium supplemented with 50 μg ml⁻¹ ampicillin was dispensed into each well of a plate and inoculated with a glycerol stock of the respective PAMO mutant. The wild-type PAMO, the inactive mutant R337A, and a well with no cells were included on every plate. Next, the plate was covered with an AeraSeal film (Excel Scientific, Inc.) to reduce evaporation of the medium and incubated overnight at 37°C with shaking. On the next day, 20 μl of a preculture from each well was used to inoculate 2 ml of Luria-Bertani medium supplemented with 50 μg ml⁻¹ ampicillin and 0.02% (wt/vol) l-arabinose for the induction of protein expression in a 24-well plate. The plate was covered with an AeraSeal film and incubated overnight at 37°C. On the third day, the 2-ml cultures were transferred into 10-ml Pyrex glass tubes, and a substrate (as 1 M stock in dimethyl sulfoxide [DMSO]) was added to a final concentration of 10 mM. After incubation at 37°C for 24 h, the reaction mixtures were extracted with 1 ml of tert-butyl methyl ether supplemented with 0.1% (vol/vol) mesitylene as an internal standard, dried over anhydrous magnesium sulfate, and analyzed by gas chromatography (GC).

Oxidations of thioanisole (substrate 5) were tested using cell extracts. Precultures were prepared in 24-well plates as described above. On the next day, 40 μl of a preculture from each well was used to inoculate 4 ml of LB medium supplemented with 50 μg ml⁻¹ ampicillin and 0.02% (wt/vol) l-arabinose in a test tube. The tubes were incubated overnight at 37°C, after which the cells were harvested by centrifugation, resuspended in 1 ml of 50 mM Tris-HCl, pH 7.5, and disrupted by sonication. Each reaction mixture was comprised of 450 μl of a cell extract, 2.5 mM thioanisole, 1% (vol/vol) DMSO, 100 μM NADPH, the coenzyme regeneration system (20 mM phosphite and 2 μM phosphite dehydrogenase), and 50 mM Tris-HCl, pH 7.5, up to 2 ml. In a control reaction mixture, the cell extract was replaced with 450 μl Tris buffer. The reaction mixtures were incubated and extracted as described above and analyzed by GC.

Since PAMO shows higher activity at slightly basic pH, wild-type PAMO and the selected mutants (A442G, S441A, I67T, L443F, M446G, and I67T/L338P/A435Y/A442G/L443F/S444C) were tested for conversion of substrate 4 at pH 9.0. For this, cell extracts were prepared and reaction mixtures were set up as described above, except that 50 mM Tris-HCl, pH 9.0, was used.

Protein expression and purification.

Wild-type and mutant PAMO proteins were overexpressed and purified by affinity chromatography as described previously (8). The extinction coefficients of flavin adenine dinucleotide (FAD) bound in PAMO mutants were determined as described previously (11). UV and visible-light absorption spectra were collected on a Perkin-Elmer Lambda Bio40 spectrophotometer.

Conversions with isolated enzymes.

Isolated enzymes were used to analyze conversions of compounds 5 to 14. The reaction mixtures were typically comprised of 2.5 to 10 mM substrate, 4 μM enzyme, 200 μM NADPH, 20 mM sodium phosphite, 3 μM phosphite dehydrogenase, and 50 mM Tris-HCl. The reactions were carried out at pH 9.0, except for substrates 11 and 12, for which pH 7.5 was used. The oxidations were usually performed at 30°C and 200 rpm. For substrates 9 and 13, reactions were also tested at 37°C. The conversion levels and enantiomeric purity of the products were assessed by GC or high-performance liquid chromatography (HPLC).

GC and HPLC analyses.

A Shimadzu GC 2014 chromatograph equipped with an AT5 column (30 m by 0.25 mm by 0.25 μm; Grace) and a Hewlett Packard HP 6890 GC system with an HP1 column (30 m by 0.32 mm by 0.25 μm; Hewlett Packard) were employed to identify product formation catalyzed by the mutants. Enantioselective oxidations of compounds 2 and 11 were analyzed on a Hewlett Packard HP 6890 chromatograph equipped with a chiral column, the Chiralsil Dex CB (25 m by 0.32 mm by 0.25 μm; Varian), while a Chiraldex G-TA column (30 m by 0.25 mm by 0.25 μm; Alltech) was used to separate enantiomeric products of oxidations of compounds 5 and 12. A Chiralcel OD column (25 cm by 0.46 cm; Daicel) was employed for the separation of compounds 6a to 8a. The products of conversion of ketone 2 were assigned by comparing them with the products of reactions catalyzed by CHMO (1), CPMO (18), and 4-hydroxyacetophenone monooxygenase (14). An oxidation product of ketone 4 was identified by comparing it with the product of a reaction catalyzed by the M446G mutant of PAMO (24). The absolute configurations of sulfoxides 5a to 8a and esters 11a to 12a were assigned by comparison of the GC or HPLC retention times with the published data (7, 27). The temperature programs used for separation and retention times of substrates and products in GC and HPLC are in Tables S2 and S3, respectively, in the supplemental material.

Determination of melting temperatures.

The ThermoFAD method (10) was employed to measure the melting temperatures of the wild-type PAMO and selected mutants. Fifty-microliter samples containing pure proteins diluted with 50 mM Tris-HCl, pH 7.5, to a final concentration of 16 μM were placed in a 96-well real-time (RT)-PCR plate (Bio-Rad Laboratories). Measurements were conducted using a MyiQ real-time PCR detection system (Bio-Rad Laboratories), which allows excitation in a range of 475 to 495 nm. The RT-PCR apparatus was equipped with a fluorescence emission filter of 515 to 545 nm. Unfolding was conducted in a range from 20°C to 90°C. Fluorescence was recorded after every 0.5°C temperature increase, followed by a 10-s stabilization step. Each measurement was performed in duplicate. As this method does not determine the equilibrium constant for the folded and unfolded enzyme, the derived melting temperatures are indicated as apparent melting temperature, T′_m. The T′_m values were derived from the peaks of the first derivatives of fluorescence intensity.

RESULTS

Comparative analysis of the CPMO model and the PAMO crystal structure.

On the basis of the PAMO structure (PDB ID 1W4X_A), a homology model of CPMO (41% of sequence identity) was calculated with Modeller. According to the 3D-Jury results, the protein was modeled on the basis of an alignment prepared by the FUGUE server (32). The model of CPMO was superimposed onto the PAMO structure. The isoalloxazine moiety of the FAD cofactor was used as a reference for the reaction center. Residues in the closest proximity (8 Å) to the C4a of the FAD cofactor were annotated and screened for differences. The assumed 8-Å radius covered the region surrounding the active site and identified 49 amino acids, of which 30 were facing the surface of the cavity, and only 12 of those were variable when comparing PAMO and CPMO. Thus, they could represent determinants of substrate specificity (marked on the alignment in Fig. 2; also see Fig. S1 in the supplemental material). By comparison with CPMO, the suggested mutations were V54I, C65V, I67T, Q93W, Q152F, L153G, I339S, S441A, A442G, S444C, M446G, and L447P. During the course of this study, the structure of PAMO (19a) in complex with both FAD and NADPH has been elucidated, which enabled further refinement of the residues flanking the active site. As a result, the following additional mutants were proposed: L338P, A435Y, and L443F (Fig. 2; also see Fig. S1 in the supplemental material).

Fig. 2.

Sequence alignment of PAMO and CPMO. Mutations proposed after the initial analysis are indicated with an asterisk (*), while mutations proposed after the analysis of the new structure of PAMO are indicated with a pound sign (#).

Substrate profiling using whole cells or cell extracts.

Based on the structural analysis described above, we prepared 15 single and 15 multiple mutants (containing 2 to 12 substitutions) of PAMO in which selected residues were replaced by amino acids occurring at the respective positions in CPMO (Fig. 2). These mutants were initially screened in conversions catalyzed by whole cells or cell extracts for activity against four ketones and one sulfide (Fig. 3 A). The results are summarized in Table 1. Only one of the single mutants, the L153G mutant, was found to be inactive with all of the tested substrates. Inactivation was also observed for the 12-fold mutant which contains this substitution. One single mutant, the A435Y mutant, was active only with bicyclo[3.2.0]hept-2-en-6-one (substrate 2). All of the other single mutants oxidized substrates 1, 2, and 5. None of the mutants tested showed activity in the oxidation of cyclopentanone (substrate 3). The only mutant able to convert 1-indanone (substrate 4) was the M446G mutant, yielding the unexpected lactone 1-isochromanone (compound 4b) as the only product (Fig. 4). This unique reactivity of the M446G mutant has already been reported (24).

Table 1.

Oxidations of substrates 1 and 2 catalyzed by the PAMO mutants in whole cells and oxidations of substrate 5 catalyzed by cell extracts^a

Description of enzyme	RA (%)b of substrate 1	Substrate 2c			Substrate 5c
		RA (%)b	2a:2bd	ee (%) (2a/2b)e	RA (%)b	ee (%)
Wild type	100 (12%)	100 (9%)	73:27	94/82	100 (28%)	26 (R)
V54I	65	54	58:42	90/76	38	34 (R)
C65V	94	54	74:26	92/84	74	22 (R)
I67T	16	33	54:46	84/28	140	74 (S)
Q93W	65	97	70:30	91/78	89	27 (R)
Q152F	35	17	30:70	66/(-)84	12	ND
L153G	≤3	≤3	ND	ND	≤3	ND
L338P	59	34	74:26	88/80	82	13 (R)
I339S	81	74	71:29	93/86	62	42 (R)
A435Y	≤3	25	29:71	46/(-) ≥90	≤3	ND
S441A	73	62	50:50	88/2	110	22 (R)
A442G	75	96	30:70	89/(-)14	270	68 (S)
L443F	65	75	59:41	92/8	150	28 (R)
S444C	66	110	80:20	97/78	66	38 (R)
M446G	49	27	65:35	84/21	250	93 (R)
L447P	85	77	74:26	94/92	44	14 (R)
I67T/L338P	≤3	8	60:40	59/(-)48	130	≤3
C65V/I67T	47	37	52:48	81/17	81	65 (S)
C65V/I67T/Q93W	54	41	57:43	85/46	12	ND
S441A/A442G	56	100	24:76	76/(-)61	170	30 (S)
Q152F/A442G	37	54	23:77	75/(-)88	35	57 (R)
Q152F/S441A/A442G	33	58	22:78	78/(-)87	40	50 (R)
C65V/I67T/Q152F/S441A/A442G	31	48	23:77	61/(-)25	45	63 (S)
Q93W/A442G/S444C/M446G/L447P	38	27	52:48	85/(-)86	13	ND
S441A/A442G/S444C/M446G/L447P (5-fold mutant)	41	32	49:51	87/(-)89	62	≥80 (R)
Q93W/S441A/A442G/S444C/M446G/L447P	40	45	47:53	81/(-)91	14	ND
I67T/L338P/A435Y/A442G	≤3	≤3	ND	ND	47	5 (R)
I67T/L338P/A435Y/A442G/L443F/S444C	≤3	5	ND	ND	23	≤3
V54I/C65V/I67T/Q93W/I339S/S441A/A442G/ S444C/M446G/L447P (10-fold mutant)	5	26	40:60	85/(-)96	12	ND
V54I/C65V/I67T/Q93W/Q152F/I339S/S441A/ A442G/S444C/M446G/L447P (11-fold mutant)	≤3	≤3	ND	ND	≤3	ND
V54I/C65V/I67T/Q93W/Q152F/L153G/I339S/S441A/A442G/ S444C/M446G/L447P (12-fold mutant)	≤3	≤3	ND	ND	≤3	ND

^aSubstrate 1 is phenylacetone, substrate 2 is bicyclo[3.2.0]hept-2-en-6-one, and substrate 5 is thioanisole.

^bRA, relative activity. For a fair comparison, all incubations were for 24 h and relative activities are expressed as the rate of conversion normalized to that obtained with wild-type PAMO. Absolute conversion values for the wild-type PAMO are given in parentheses.

^cND, not determined.

^dRatio of compound 2a to 2b.

^eIn most cases, the (1S,5R)-enantiomer is the main product; (-) indicates that (1R,5S)-lactone is the main product.

Fig. 4.

(A) The Baeyer-Villiger oxidation of bicyclo[3.2.0]hept-2-en-6-one (racemic compound 2) leads to the formation of the “normal” (2a) and “abnormal” (2b) lactones. (B) Similarly, 1-indanone (4) can be oxidized to the “normal” (4a) and “abnormal” (4b) products.

Interestingly, the L338P mutant, in which a drastic substitution was introduced next to the catalytically essential R337, presented activity comparable to that of the wild-type enzyme. However, when this mutation was combined with I67T or more amino acid substitutions, we observed a significant drop in the activity in the Baeyer-Villiger reactions. For the I67T/L338P, I67T/L338P/A435Y/A442G, and I67T/L338P/A435Y/A442G/L443F/S444C mutants, no oxidation or only traces of esters could be observed, while the formation of sulfoxide 5a, albeit of poor enantiomeric purity, was detected.

Remarkably, for a number of mutants, the enzymatic regio- and enantioselectivities were altered. The Baeyer-Villiger oxidation of the racemic compound 2 can result in the formation of four products: two enantiomers of the “normal” (product 2a) and two enantiomers of the “abnormal” (product 2b) lactones (Fig. 4). The wild-type PAMO produces both the “normal” and “abnormal” lactones in a ratio of about 3:1, with high enantiomeric excesses (ee) in favor of the (1S,5R)-enantiomers. The I67T and S441A mutants yielded the two regioisomers in almost equal amounts. This trend, although less pronounced, was also observed for the V54I and L443F mutants. Furthermore, the Q152F, A435Y, and A442G mutants generated compound 2b as the main product. When the enantioselectivity in the oxidation of compound 2 was analyzed, it turned out that all of the mutants produced mainly the (1S,5R)-enantiomer of compound 2a. Interestingly, some mutations led to a decrease in the enantioselectivity of product 2b (I67T, S441A, and L443F), while three mutants (Q152F, A435Y, and A442G) showed preference toward the opposite (1R,5S)-enantiomer.

In some situations, combining mutations resulted in synergistic effects. When the S441A mutant, which displayed the low enantioselectivity for product 2b, was merged with the A442G variant, showing little preference for the (1R,5S)-enantiomer, the resulting S441A/A442G mutant catalyzed the production of the (1R,5S)-enantiomer of product 2b with high optical purity. This effect was even more striking for the double mutant I67T/L338P: combining two mutants selective for the (1S,5R)-enantiomer of product 2b created an enzyme producing the opposite enantiomer. Furthermore, the fact that several multiple mutants (Q93W/A442G/S444C/M446G/ L447P, Q93W/S441A/A442G/S444C/M446G/L447P, and S441A/A442G/S444C/M446G/L447P) feature very similar enantioselectivity for ketone 2 illustrates that the effects of mutations depend very much on the context. The A442G mutation, which was the only substitution increasing the selectivity for the (1R,5S)-enantiomer of product 2b that was present in all three mutants, did not appear to be able to account by itself for such a high ee value for the (1R,5S)-enantiomer of product 2b of the multiple mutants. The S441A mutation, which by itself is less selective for the (1R,5S)-enantiomer of product 2b, was missing in the Q93W/A442G/S444C/M446G/L447P mutant. This indicates that one or more of the other mutations (S444C, M446G, or L447P), which did not have much effect in the single mutants, caused this significant change in the enantioselectivity in the multiple mutants. This observation is in line with the predicted role of the mutated residues: together, they are involved in shaping the substrate binding cavity.

Sulfoxidation of thioanisole (substrate 5) by the wild-type PAMO leads to the formation of the (R)-methyl phenyl sulfoxide with a moderate ee and conversion level. Several mutants afforded significantly increased activity for this sulfide (I67T, A442G, L443F, M446G, I67T/L338P, and S441/A442G). For two of the mutants tested, the opposite enantiomeric preference was observed: I67T and A442G formed the (S)-sulfoxide with high optical purities. Interesting effects occurred when the multiple mutants were tested with this sulfide. When the substitution Q152F was introduced into the A442G or S441A/A442G mutant, the resulting enzymes showed a change in the enantiomeric preference, forming the (R)-sulfoxide. Moreover, the addition of the mutation I67T (along with C65V, which, in contrast, did not seem to have an effect) to the triple mutant with mutations Q152F/S441A/A442G resulted in a significant change in the enantioselectivity, from 50% ee for the (R)-enantiomer to 63% ee for the (S)-sulfoxide.

The 11-fold mutant, from which the deleterious substitution L153G had been removed, still did not perform any oxidation. Upon curing the Q152F mutation, which in the single mutant caused a significant decrease in the activity, small amounts of Baeyer-Villiger oxidation and sulfoxidation products were observed. The regio- and enantioselectivities of this 10-fold mutant in the oxidation of ketone 2 were similar to those of the 5- and 6-fold mutants.

Substrate profiling using isolated mutant enzymes.

Eight of the mutant enzymes featuring altered biocatalytic properties (V54I, I67T, Q152F, A435Y, A442G, M446G, Q152F/A442G, and S441A/A442G/S444C/M446G/L447P) were chosen for purification and further characterization. As isolated enzymes, they were tested in conversions of a number of sulfides (substrates 6 to 10) and ketones (substrates 11 to 14), which are not recognized or are poorly recognized by the wild-type PAMO. In addition, the mutants were applied in oxidations of prochiral (substrates 5, 6, 12) or racemic (substrate 11) compounds in order to further assess their enantioselective behavior. Tables 2 and 3 provide experimental data on these reactions.

Table 2.

Conversions of sulfides 5 to 8 by the isolated PAMO mutants^a

Description of enzyme	Substrate 5			Substrate 6			Substrate 7			Substrate 8
	cb (%)	ee (%)	Config	cb (%)	ee (%)	Config	cc (%)	ee (%)	Config	cb (%)	ee (%)	Config
Wild type	70	34	R	97	97	S	5	ND	ND	18	37	S
Q152F	18	54	R	70	90	S	≤3	ND	ND	36	40	S
M446G	78	90	R	82	60	S	80	97	R	60	78	S
V54I	70	50	R	95	96	S	13	46	R	51	45	S
A442G	42	10	S	45	42	S	17	34	R	36	28	R
5-Fold mutantd	71	50	R	96	82	S	75	92	R	25	56	S
Q152F/A442G	25	75	R	92	93	S	7	ND	ND	41	40	S
I67T	40	25	R	97	95	S	28	93	R	25	19	S
A435Y	20	72	R	97	97	S	≤3	ND	ND	13	35	S

^aAll reactions were performed at 30°C. For all compounds, a small amount of sulfone was observed (≤12%). c, conversion; Config, configuration; ND, not determined.

^bReactions were stopped after 24 h.

^cReactions were stopped after 48 h.

^dThe 5-fold mutant has mutations S441A/A442G/S444C/M446G/L447P.

Table 3.

Conversion of ketones 11 and 12 by the isolated PAMO mutants^a

Description of enzyme	Substrate 11			Substrate 12
	t (h)	c (%)	E	cb (%)	ee (%)	Config
Wild type	0.5	36	20	61	82	R
Q152F	1	43	44	≤3	ND	ND
M446G	1	11	19	20	81	R
V54I	1	10	11	8	ND	ND
A442G	0.5	34	17	8	ND	ND
5-fold mutantc	2	35	28	7	ND	ND
Q152F/A442G	8	56	12	≤3	ND	ND
I67T	1	9	18	43	82	R
A435Y	1	26	54	≤3	ND	ND

^aAll conversions were performed at 30°C. c, conversion; Config, configuration; ND, not determined.

^bReactions were stopped after 2 h.

^cThe 5-fold mutant has mutations S441A/A442G/S444C/M446G/L447P.

In the case of thioanisole, the results from conversions by cell extracts and by isolated enzymes could be compared. In general, the findings from the initial screening (Table 1) were in agreement with the results of the isolated enzyme conversions (Table 2). One outlier was found, the I67T mutant, which produced the (S)-sulfoxide in the initial cell extract-based screening, while it turned out to be selective for the (R)-enantiomer when the isolated enzyme was used. The reason for this change in enantiomeric preference remains unclear. Previous studies have shown that the medium conditions, e.g., pH or additives/solvents, can influence the enantioselectivity of PAMO, which may suggest that intracellular components may influence the enantioselectivity of this mutant (6, 34). In addition, 68% ee was observed for the A442G mutant when using the cell extract, which was much higher than the 10% ee seen when using the isolated enzyme. Although overoxidation of sulfides catalyzed by PAMO, namely, the formation of sulfones, can take place, the detected levels of sulfones (5 to 10%) indicate that this second reaction could not affect the ee significantly. This was further supported by the results of the oxidation of racemic sulfoxide 5a, which was performed for the tested mutants: the enantiomeric ratio values (E) obtained were low, ranging from 2 to 16 (data not shown). Finally, as analyzed in the time-resolved conversion of thioanisole by the A442G mutant, the enantiomeric excess of product 5a did not change during the progress of the reaction (data not shown).

Conversions of the benzyl methyl sulfide (compound 6) by the mutants were performed with usually good yields and high optical purities, leading to (S)-benzyl methyl sulfoxide. The two exceptions were the A442G and M446G mutants, for which only moderate ee values were obtained. The change in the enantioselectivity of the A442G mutant appears to be consistent with the inverted enantioselectivity of this specific mutant toward thioanisole. The M446G mutant was already characterized to some extent, and the results of conversion of sulfide 6, as well as substrate 5, obtained in the current investigation are similar to the published data (31).

For testing whether mutants were more effective with relatively bulky substrates, compounds 7 and 8 were tested. While these sulfides are poorly accepted by the wild-type PAMO, they were found to be converted more efficiently by several mutants. The variants M446G and S441A/A442G/S444C/M446G/L447P were able to catalyze the sulfoxidation of benzyl phenyl sulfide (substrate 7) with good yields and high ee values. The I67T mutant also performed this reaction in a highly selective manner, but the yield was only moderate. Furthermore, a few mutants oxidized sulfide 8 with activities and enantioselectivity that were higher than those of the wild-type enzyme. It is worth mentioning that the M446G mutant produced 60% of the sulfoxide with an enantiomeric excess of 78% in favor of the (S)-enantiomer, whereas the A442G mutant yielded the (R)-sulfoxide with 28% ee. While a number of mutants showed increased activities with compounds 7 and 8, none of them were able to convert more challenging substrates, such as sulfides 9 and 10 or ketones 13 and 14, to a significant extent (<2%).

The wild-type PAMO was previously reported to oxidize 3-phenylpentan-2,4-dione (substrate 12) with a good yield and optical purity (7). Two of the mutants tested (M446G and I67T) catalyzed the oxidation with similar enantioselectivity but lower conversion rates than the wild-type enzyme. In the kinetic resolution of the racemic benzylketone (compound 11), all of the biocatalysts preferentially oxidized the (S)-enantiomer. The mutants A435Y and Q152F displayed increased E values (54 and 44, respectively) compared to that of the wild-type PAMO (E = 20), while for the rest of the mutants, the enantioselectivity achieved was similar to that of the wild-type enzyme.

Expression level and thermostability of PAMO mutants.

The variants of PAMO with a high mutation load created during this research did not show compromised expression levels. In fact, the expression levels of all of the mutants constructed have been assessed by SDS-PAGE, and no differences between the wild-type PAMO and the mutants were observed. The multiple mutants, including the 12-fold mutant that was found to be inactive in Baeyer-Villiger reaction or sulfoxidation, were still able to incorporate FAD, as indicated by the A₂₈₀/A₄₄₁ ratio (Table 4). This demonstrates their ability to fold properly, although in the case of the M446G and S441A/A442G/S444C/M446G/L447P mutants and the 12-fold variant, the enzymes were not fully occupied with the cofactor. The thermostability of the purified wild-type PAMO and the mutants was evaluated by the determination of apparent melting temperatures by the ThermoFAD method (10). As can be seen from the data in Table 4, the single mutants did not suffer from significant decreases in their apparent melting temperatures when compared to the relatively high T′_m of the wild-type PAMO (60°C). The only exception to this trend was the A435Y mutant, for which a T′_m of 55.5°C was observed. In this variant, an alanine residue was replaced by a bulky amino acid, which could affect the packing of residues inside the protein and, thus, its stability. The 5-fold S441A/A442G/S444C/M446G/L447P mutant displayed a T′_m of 57°C. This loss of three degrees in the T′_m could be attributed to the additive effects of the substitutions A442G and M446G.

Table 4.

Absorbance ratios and apparent melting temperatures of the wild-type PAMO and the selected mutants

Description of enzyme	A280/A441	T′m (°C)
Wild type	20	60
Q152F	23	60
A442G	22	59
V54I	25	58
I67T	27	60
M446G	∼50	58.5
A435Y	21	55.5
Q152F/A442G	26	59
5-Fold mutanta	∼65	57
10-Fold mutantb	22	NDc
12-Fold mutantd	∼65	ND

^aThe 5-fold mutant has mutations S441A/A442G/S444C/M446G/L447P.

^bThe 10-fold mutant has mutations V54I/C65V/I67T/Q93W/I339S/S441A/A442G/S444C/M446G/L447P.

^cND, not determined. The denaturation curve showed a gradual unfolding from 30°C to 58°C with no clear transition.

^dThe 12-fold mutant has mutations V54I/C65V/I67T/Q93W/Q152F/L153G/I339S/S441A/A442G/S444C/M446G/L447P.

DISCUSSION

An extensive site-directed mutagenesis study was performed on PAMO. Fifteen residues in this enzyme were selected and mutated to amino acids present in CPMO at the corresponding positions. The targeted positions are not conserved among different BVMOs (see Fig. S1 in the supplemental material), which supports the concept that they may account for the observed differences in the substrate scope. Our data show that many of the introduced mutations influence the regio- and enantioselectivities of PAMO.

Oxidation of ketone 2 is often used as a probe for the selectivity of BVMOs. PAMO is able to produce both regioisomers, similar to CHMO and unlike CPMO, which generates mainly the “normal” lactone (product 2a) (1, 18). The “normal” lactone is produced as the (1S,5R)-enantiomer by PAMO, similar to other BVMOs. PAMO also generates the (1S,5R)-enantiomer of compound 2b, in contrast to CPMO and CHMO, which provide the (1R,5S)-enantiomer of compound 2b (1, 18). The Q152F, A435Y, and A442G mutants have manifested inverted enantioselectivity of product 2b, which makes them similar to CPMO and CHMO with respect to the enantioselective conversion of this substrate. However, their regioisomeric preference for the production of compound 2b rather than compound 2a is a unique property among the BVMOs. Three other mutants (I67T, S441A, and L443F) generated similar amounts of both lactones, and the enantiomeric purity of compound 2b was lost. The drop in the enantioselectivity, even though it is an undesired property, could be interpreted as a step toward the inversion of the enantioselectivity. Therefore, it is reasonable to conclude that mutations at these positions can be used to tune the enantioselectivity of PAMO. Lastly, a change of the regioselectivity was observed for the V54I mutant, which produced similar amounts of compounds 2a and 2b while retaining the enantiomeric preference. In summary, our data confirm the hypothesis that positions V54, I67, Q152, and A435, in addition to previously analyzed S441, A442, L443, and M446 (22, 31), contribute to the substrate specificity and the enantioselectivity of PAMO.

It is worth noting that among the eight mutants tested as isolated proteins, all except I67T and A442G converted thioanisole to the (R)-sulfoxide with an ee higher than that of the wild-type PAMO. Remarkably, the enantiomeric preference of the A442G mutant in oxidations of compounds 5, 6, and 8 was different than that of all the other mutants tested. The I67T and M446G mutants oxidized sulfide 7 with excellent ee values, while mutations Q152F and A435Y resulted in increased E values in the kinetic resolution of ketone 11. These data further support the idea that positions V54, I67, Q152, A435, A442, and M446 can be useful targets for tuning the enantioselectivity of PAMO in Baeyer-Villiger reactions or sulfoxidations.

As illustrated in Fig. 5, the above-mentioned amino acids cluster around the R337 residue, which has been found to be essential for catalysis (29). R337 in turn is predicted to interact with NADP⁺, which is bound during the oxygenation step. S441, A442, L443, and M446 belong to a loop whose importance for the substrate specificity of PAMO is well known (2, 22, 31). A435 can also be considered a part of this extended loop, and this residue points toward the putative binding site. While located close to the 440-to-446 loop as well, Q152 stays in the proximity of the ribose moiety of FAD, which indicates that it might interact with the cofactor and a mutation of Q152 may affect the position of FAD. Similarly, a mutation at position V54 could influence the orientation of the flavin. I67 is located on the opposite site of the isoalloxazine ring from the other residues discussed but also in the direct neighborhood of the flavin. Upon mutation to a polar residue, i.e., a threonine, this residue could form a hydrogen bond with the cofactor and thereby subtly change the architecture of the active site. In summary, the residues found to influence the substrate specificity and the selectivity of PAMO can directly or indirectly, via interactions with the FAD cofactor, influence the positioning of a substrate in the active site. Nevertheless, it must be emphasized that conclusions on structural effects of the mutations should be treated with caution, since it has been postulated that BVMOs adopt multiple conformations and undergo domain rearrangement during the catalytic cycle (17, 19, 29).

Fig. 5.

A schematic representation of the active site of PAMO. Residues identified as hot spots for substrate specificity are displayed as sticks. Residue R337 that is essential for catalysis is also shown. FAD is presented as black sticks. The schematic was prepared using PyMol software and the structure of PAMO (PDB ID 1W4X_A).

The mutation of L153 to a glycine fully inactivates PAMO in Baeyer-Villiger and sulfoxidation reactions. Even though it was already known that the double mutant Q152F/L153G did not show any Baeyer-Villiger activity (31), it was not clear at that time which of the two mutations inactivates PAMO. Our results demonstrate that the Q152F mutant could catalyze oxidations, while the substitution L153G was detrimental to PAMO either alone or in any combination. In this mutant, a leucine, which is a relatively large amino acid, was exchanged for a small one. Such a mutation could affect the orientation of FAD, which must be tightly controlled in order to react with NADPH, oxygen, and a substrate. Strikingly, the corresponding glycine in CPMO could be mutated into much bulkier residues, for example, a phenylalanine, a tyrosine, or a leucine, without impairing the catalytic properties and could even improve the enantioselectivity of the enzyme (4). Perhaps a more conservative replacement of L153, such as L153I, could still yield active enzyme, which may also display altered biocatalytic properties.

Combining mutations into multiple mutants has in some cases resulted in unexpected properties. Most strikingly, the substitution L338P, which had virtually no effect in the single mutant, led to almost a complete loss of activity when combined with other mutations. Furthermore, the highly mutated variants of PAMO (10 to 12 mutations) showed no or little Baeyer-Villiger activity. This could be explained by the fact that some of the introduced mutations had negative effects on the enzyme's catalytic performance, and they resulted in severe decreases in its activity upon accumulation in one mutant. Meanwhile, it should be pointed out that the multiple mutants were still viable with respect to expression and the binding of the flavin cofactor. This demonstrates again that PAMO is ideally suited for enzyme redesign studies because it can tolerate a high load of mutations.

Due to the chosen approach, which entails a limited capacity for analysis, only certain specific substitutions were tested. In some cases, quite dramatic changes were introduced (e.g., A435Y and L447P), while the mutations in other variants caused only slight changes (e.g., S444C and V54I). It could be argued that in order to get a complete view, one should perform a site saturation mutagenesis on all of the active-site residues and assess the effects of all possible amino acid exchanges. However, it would entail considerably more effort, particularly if more substrates were to be tested. Since we wanted to test as many as 15 positions with at least five substrates, the present approach appeared to be more appropriate. Certainly, our results have proven the concept of testing single substitutions to be valuable. We assume that if a mutation to a significantly different amino acid does not have any effect on the enzyme's activity or selectivity, this position is not involved in interactions with a substrate. This appears to be the case for the C65V, Q93W, I339S, and L447P mutants. On the other hand, if subtle mutations bring about a clear effect on the enantio- or regioselectivity of the enzyme, as was the case for the V54I mutant, it emphasizes the importance of the position for substrate recognition and positioning in the active site.

A number of positions investigated in this paper, Q152, L153, L338, I339, and M446, had formerly been subjected to partial randomization in several PAMO mutant libraries, but no hits were found in the screening for activity with 2-phenylcyclohexanone (23). While one of these positions, namely, I339, seems not to be involved in substrate recognition in the present study, our data indicate that replacing the other residues can affect the biocatalytic properties of PAMO. Apparently, the mutations of Q152, L338, and M446 tested in the current project, which were also included in the libraries prepared in the research of Reetz and Wu, are not sufficient to introduce activity against this specific compound. Moreover, the double Q93N/P94D mutant afforded a PAMO variant that converted 2-ethylcyclohexanone, as well as other 2- and 3-substituted cyclohexanones, with high enantioselectivity (33). In our investigation, the mutation of Q93W did not influence the substrate specificity or the enantioselectivity of PAMO. This is in line with the findings of Wu et al., who reported that no activity on the above-mentioned substrates could be detected when single mutants or single-position libraries were tested. It suggests that the effects of mutations at position 93 can only be observed when an accompanying mutation, e.g., P94D, is introduced. In fact, the properties of the Q93N/P94D mutant have been explained by a conformational change relatively far from the active site that translates into altered accessibility of the active site. This agrees with our conclusion that Q93 is not directly involved in substrate recognition.

This thorough site-directed mutagenesis study has allowed us to identify four positions in PAMO, V54, I67, Q152, and A435, which in addition to the previously known hot spots, such as S441, A442, L443, and M446, contribute to substrate specificity. With these data, the substrate binding pocket of PAMO is becoming more defined, which will help to efficiently engineer this enzyme to accept new substrates. As is clear from our results, even single, arbitrarily designed substitutions of indicated amino acids can result in mutants with increased activity and enantioselectivity. This supports the idea that screening libraries of PAMO randomized at identified positions could bring greatly improved variants. Finally, the ability of PAMO to accommodate numerous mutations highlights the exceptional robustness of this protein and underlines its suitability as a starting point for future directed-evolution experiments.