Effect of Mutation and Substrate Binding on the Stability of Cytochrome P450_BM3 Variants

Abstract

Cytochrome P450_BM3 is a heme-containing enzyme from Bacillus megaterium that exhibits high monooxygenase activity and has a self-sufficient electron transfer system in the full-length enzyme. Its potential synthetic applications drive protein engineering efforts to produce variants capable of oxidizing nonnative substrates such as pharmaceuticals and aromatic pollutants. However, promiscuous P450_BM3 mutants often exhibit lower stability, thereby hindering their industrial application. This study demonstrated that the heme domain R47L/F87V/L188Q/E267V/F81I pentuple mutant (PM) is destabilized because of the disruption of hydrophobic contacts and salt bridge interactions. This was directly observed from crystal structures of PM in the presence and absence of ligands (palmitic acid and metyrapone). The instability of the tertiary structure and heme environment of substrate-free PM was confirmed by pulse proteolysis and circular dichroism, respectively. Binding of the inhibitor, metyrapone, significantly stabilized PM, but the presence of the native substrate, palmitic acid, had no effect. On the basis of high-temperature molecular dynamics simulations, the lid domain, β-sheet 1, and Cys ligand loop (a β-bulge segment connected to the heme) are the most labile regions and, thus, potential sites for stabilizing mutations. Possible approaches to stabilization include improvement of hydrophobic packing interactions in the lid domain and introduction of new salt bridges into β-sheet 1 and the heme region. An understanding of the molecular factors behind the loss of stability of P450_BM3 variants therefore expedites site-directed mutagenesis studies aimed at developing thermostability.

Graphical Abstract

P450_BM3 (CYP102A1) belongs to the P450 cytochrome superfamily of heme b-dependent monooxygenases and has been extensively studied in recent years because of its potential as a biocatalyst in the production of fine chemicals and environmental remediation. It has the highest known monooxygenase activity among the P450 enzymes,¹ which is attributed to efficient electron transfer due to fusion of the reductase and heme domains in the single polypeptide chain.² P450_BM3 catalyzes the hydroxylation and/or epoxidation of fatty acids, fatty amides, and alcohols. Random mutagenesis and directed evolution have produced more promiscuous variants capable of catalyzing the oxidation of nonnative substrates, particularly drugs normally metabolized by human P450s. For example, P450_BM3 containing the D251G and Q307H mutations is active toward nonsteroidal anti-inflammatory drugs,³ that containing F87V and A82F toward proton pump inhibitors,⁴ and that containing R47L, F87V, and L188Q [triple mutant (TM)] toward druglike molecules such as dextromethorphan and 3,4-methylenedioxymethylamphetamine.⁵

There has been interest in using TM P450_BM3 as a platform from which to develop additional variants that exhibit activity toward other nonnative substrates.^6-9 A prior computational study of TM, as well as the R47L/F87V/L188Q/E267V quadruple mutant (QM) and the R47L/F87V/L188Q/E267V/F81I pentuple mutant (PM), linked the enhanced catalytic activity of the mutants to the more closed conformation of the substrate channel and possibly to electrostatic effects resulting from the bending of heme propionate A toward the active site.¹⁰ Mutations that are beneficial for nonnative substrate catalysis, however, often cause loss of protein stability, limiting the application of P450_BM3 mutants as industrial biocatalysts. It also precludes their use as a starting point in engineering more promiscuous and active variants, as they are less tolerant of destabilizing mutations that might be required for nonnative substrate binding.^11-13

Protein stability pertains to (a) thermodynamic stability, the resistance to unfolding defined by the free energy difference between the folded and unfolded states (ΔG_stab) and melting temperature (T_m, the temperature at which 50% of the protein is unfolded), and (b) kinetic stability, the resistance to irreversible inactivation defined by the half-life of the enzyme (t_1/2) at a specific temperature.^14,15 These two definitions of stability involve different processes but are usually related when the protein follows the classical two-step N ↔ U → D process, where the native structure (N) first undergoes reversible unfolding (U), leading to irreversible denaturation (D) and, eventually, to permanent inactivation due to aggregation, misfolding, covalent changes, cofactor loss, or oxidation of sulfur-containing residues.^15-18 Though kinetic stability may be a more important parameter in cost-effective industrial enzyme utilization, thermodynamic stability is inextricably linked to kinetic stability through the unfolding process. As increasing resistance to unfolding, via any mechanism, will likely also enhance kinetic stability, the study presented here focuses on the thermodynamic stability of P450_BM3 variants given the relative accessibility of thermodynamic stability as an engineering target.

Effective means of overcoming the destabilization of TM (and other variants derived from it) remain largely unknown because the mechanisms behind the phenomenon are not well understood. In the case of full-length wild-type (WT) P450_BM3, the relatively low stability results from the low T_m of the reductase domain (47.5 °C) compared to that of the heme domain (64.9 °C and a shoulder at 58.7 °C).¹⁹ Thus, one approach to improving stability has been to replace the reductase domain with a more thermostable analogue, for example, that of CYP102A3.²⁰ However, chimeras such as this generally have lower coupling efficiencies and turnover rates.²¹ Alternatively, Arnold et al. developed an efficient peroxide-driven variant of the P450_BM3 heme domain, 21B3, which requires neither the reductase domain nor NADPH or O₂.²² Prior to that, the F87A and F87G mutants were observed to perform oxidation via the peroxide shunt pathway.^23,24 The 21B3 variant has since been used as a starting point to create more thermostable variants through directed evolution,^11,25,26 though this approach suffers from the required time-intensive variant screening process.

As a rational approach to selection of mutation sites for the stabilization of P450_BM3 variants, the molecular-level contributions to the stability of the heme domain were determined using biochemical, structural, and computational methods. As the heme domain can be used as a catalytic center using the peroxide shunt, the studies were focused on this one domain. The stabilities of WT, TM, QM, PM, and the single mutants E267V and F81I, in the presence and absence of small active site ligands, were experimentally measured by chemical denaturation using pulse proteolysis and circular dichroism (CD).²⁷ Structural data for WT and PM, as well as PM in complex with the native substrate, palmitic acid, and the inhibitor, metyrapone, were obtained to directly examine the effects of mutation and binding. High-temperature molecular dynamics (MD) simulations were performed to model the unfolding process and describe the interactions underlying the biochemical results. Labile regions in PM, which can be targeted for mutation to improve stability, were identified on the basis of the experimental and computational findings.

EXPERIMENTAL PROCEDURES

Cloning and Site-Directed Mutagenesis of P450_BM3.

The heme domain (Thr 1–Thr 463) of P450_BM3 containing a C-terminal six-His tag was cloned into the pCWori vector. All point mutations were incorporated into the heme domain using the QuikChange (Stratagene) site-directed mutagenesis procedure. The DpnI-digested polymerase chain reaction mixture was transformed into ultracompetent cells (Stratagene) and screened for ampicillin resistance. Colonies were grown in Luria broth with 100 μg/mL ampicillin, followed by plasmid isolation. All mutations were confirmed by sequence analysis (Eurofins Scientific).

Expression and Purification.

The P450_BM3 heme domain proteins were expressed in Escherichia coli BL21(DE3) cells. Cells were grown in 1 L of Terrific Broth at 37 °C while being shaken at 150 rpm until an OD₆₀₀ of 0.7–0.8 was reached. Protein expression was then induced via addition of isopropyl β-d-1-thiogalactopyranoside at a final concentration of 0.5 mM. The temperature was then reduced to 30 °C, and expression was continued for 20 h before the cells were harvested by centrifugation at 4000g for 15 min at 4 °C. The supernatant was decanted, and the cell pellet was stored at −80 °C. For purification, the cell pellets were resuspended in buffer A [50 mM NaH₂PO₄, 300 mM NaCl, 10 mM imidazole, and 0.1 mM ethylenediaminetetraacetic acid (EDTA) (pH 8.0)] with the addition of 0.1 mM phenylmethanesulfonyl fluoride. Lysis was then conducted by sonication on ice for 15 min using a microtip with an output control of 3 and a duty cycle of 50% (Branson Sonifier 250). Cell debris was cleared by centrifugation at 20000g for 1 h at 4 °C. The supernatant was decanted and passed through a 0.45 μm polytetrafluoroethylene filter before being loaded onto a His-Trap column (GE Healthcare) equilibrated with 5 column volumes of buffer B [50 mM NaH₂PO₄, 300 mM NaCl, and 20 mM imidazole (pH 8.0)]. The protein was eluted over a gradient where the imidazole concentration was linearly increased from 20 to 300 mM. The fractions containing heme-bound protein were identified by measuring absorption at 420 nm by UV–vis spectroscopy (Figure S1) and concentrated to approximately 2 mL using Amicon Ultracel-30K Millipore centrifugal units at 4500g and 4 °C. The protein was then loaded onto a Hi-Prep 26/60 Sephacryl S200 HR column equilibrated in 20 mM 2-amino-2-hydroxymethylpropane-1,3-diol (Tris) and 150 mM NaCl (pH 8.0). Protein fractions containing an A₄₂₀/A₂₈₀ absorption ratio of >1.4 for WT, >1.3 for TM, and >1.2 for F81I, QM, and PM were collected and concentrated to 10–20 mg/mL using an Ultracel-30K centrifugal filter unit. For E267V, the protein was isolated in the high-spin state, so fractions with an A₃₉₄/A₂₈₀ ratio of >1.3 were collected and concentrated. The protein concentration and quality were determined by CO binding (Figure S2).²⁸ For storage, glycerol was added to the protein to a final concentration of 20%, and aliquots were frozen in a dry ice/ethanol bath and kept at −80 °C.

Pulse Proteolysis.

The pulse proteolysis procedure was adapted from the work of Park and Marqusee.²⁹ The protein was prepared by thawing on ice followed by buffer exchange into 10 mM Tris (pH 7.4) using a Micro Bio-Spin 6 chromatography column (Bio-Rad). The protein was then diluted to 1.7 mg/mL in pulse buffer [20 mM Tris, 10 mM CaCl₂, and 20 mM NaCl (pH 7.4)] with urea added to achieve concentrations ranging from 0 to 6.8 M. The samples were then incubated for 2 h at room temperature. For experiments involving ligands, the midpoint concentration required to saturate binding (K_d) was first determined by measuring the change in heme absorption as a function of ligand concentration. The protein was then equilibrated in pulse buffer containing the ligand at 10 times the K_d for 10 min prior to the addition of urea.

After incubation with urea, thermolysin was added to the solution to a final concentration of 0.6 mg/mL, gently vortexed to mix, and allowed to react for 1 min at room temperature. The reaction was immediately quenched upon addition of EDTA to give a final concentration of 37 mM and the mixture placed on ice. Laemlli sample buffer was added, and the samples were kept at 95 °C for 1 min, followed by centrifugation for 30 s at 17000g. Samples (7 μg of P450_BM3) were resolved on a 4 to 12% Tris-glycine gel, followed by Coomassie Blue staining. Gels were then imaged and quantified using a ChemiDoc MP instrument equipped with Image Quant (Bio-Rad), where the intensities of the P450_BM3 bands were quantified and plotted against urea concentration. The concentration midpoint (C_m) was determined by normalizing the data to the fraction folded. C_m values were averaged from three replicate experiments.

Circular Dichroism.

The heme domain was equilibrated with urea at a concentration of 1.7 mg/mL at room temperature as described above, and spectra were recorded using a Jasco J-815 CD spectrometer. Because of spectral interference, far-UV spectra were recorded from 190 to 250 nm in 50 mM KH₂PO₄ buffer (pH 7.6), while near–UV–visible spectra were recorded between 300 and 450 nm with the sample in 20 mM Tris, 10 mM CaCl₂, and 20 mM NaCl (pH 7.6). Ligands were added as described for pulse proteolysis. The change in ellipticity at 222 and 415 nm was determined, and the data were plotted as a function of urea concentration.

Complex Formation for Crystallization.

Separate samples of both purified WT and PM P450_BM3 heme domain proteins were diluted in 20 mM Tris-HCl and 150 mM NaCl (pH 8.0) to a final concentration of 20 mg/mL with either 3 mM metyrapone (Sigma) or 5 mM sodium palmitate (Sigma). Free protein and small molecule complexes were flash-frozen in liquid nitrogen and stored in 50 μL aliquots at −80 °C.

Crystallization.

P450_BM3 heme domain protein crystals and complex cocrystals were grown at room temperature by the hanging-drop vapor-diffusion method. One microliter of protein or protein:small molecule complex (~20 mg/mL) was mixed with 1 μL of the reservoir solution containing 0.10–0.25 M NiCl₂ and 5–10% polyethylene glycol monomethyl ether (PEG 2000 MME) and then suspended in a sealed compartment over 1 mL of the reservoir solution. Ruby red hexagonal disc crystals grew to a final size of up to 0.3 mm × 0.3 mm × 0.2 mm in 3–4 days.

Diffraction Data Collection and Processing.

P450_BM3 heme crystals and ligand-bound cocrystals were harvested with nylon loops and immersed in the reservoir solution supplemented with 20 or 30% ethylene glycol before being flash-cooled directly in liquid nitrogen. Diffraction data were collected at 100 K with 0.1° oscillations in continuous (shutterless) mode on a Pilatus-6MF pixel array detector at Advanced Photon Source (Argonne National Laboratory, Argonne, IL) NE-CAT beamline 24ID-C. Because of one large unit cell dimension (c = 705–720 Å) in the crystals, diffraction data were collected using a mounted mini-kappa goniometer at a crystal-to-detector distance of 900 mm. Raw diffraction intensities were indexed, integrated, and scaled in X-ray Detector Software as part of the RAPD package.³⁰ Data collection statistics are listed in Table 1.

Table 1.

Data Collection, Structure Solution, and Model Refinement Statistics of Forms of the P450_BM3 Heme Domain

	wild-type	pentuple mutant
	no substrate	no substrate	metyrapone	palmitic acid
Data Collection
X-ray source	APS 24ID-C	APS 24ID-C	APS 24ID-C	APS 24ID-C
wavelength (Å)	0.9792	0.9792	0.9792	0.9792
space group unit cell (Å)	P6522	P6522	P6522	P6522
a	55.5	56.0	55.9	55.9
b	55.5	56.0	55.9	55.9
c	717.7	711.2	711.0	707.6
no. of molecules per asymmetric unit	1	1	1	1
resolution rangea	179.44–2.76 (2.91–2.76)	177.81–2.77 (2.92–2.77)	177.76–2.77 (2.92–2.77)	176.89–2.84 (2.99–2.84)
Rsymm	0.077 (0.435)	0.070 (0.323)	0.065 (0.373)	0.091 (0.431)
no. of observations	135949	136501	134033	127981
no. of unique reflections	18145	18183	18059	16866
completeness (%)	97.0 (80.2)	97.2 (80.6)	96.2 (74.7)	98.5 (92.3)
redundancy	7.5 (3.5)	7.5 (3.4)	7.4 (3.7)	7.6 (5.0)
average intensity (I/σ)	16.2 (2.3)	17.7 (3.1)	18.0 (2.4)	14.2 (2.8)
Structure Solution by Molecular Replacement
probe	3NPL	3NPL
resolution range (Å)	10.0–4.0	10.0–4.0
Z-scoreb	11.1	12.8
LLGb	3113.11	3083.44
Refinement
resolution range (Å)	59.83–2.77 (2.92–2.77)	59.27–2.77 (2.92–2.77)	48.42–2.77 (2.91–2.77)	117.93–2.84 (3.02–2.84)
no. of unique reflections	17928 (1853)	17997 (1228)	17883 (1732)	16706 (2322)
no. of protein atoms	3699	3688	3688	3688
no. of ligand atoms	67	67	64	69
no. of water atoms	57	61	83	56
Rwork	0.2466 (0.3347)	0.2008 (0.2923)	0.2176 (0.3219)	0.2207 (0.3075)
Rfreec geometry	0.3017 (0.4245)	0.2572 (0.4153)	0.2627 (0.4448)	0.2741 (0.3784)
rmsd for bond lengths (Å)	0.002	0.004	0.003	0.003
rmsd for bond angles (deg)	0.689	0.947	0.739	0.785
mean B (Å2)
protein	51.96	49.11	59.21	49.98
ligands	46.73	46.97	49.54	45.47
waters	47.97	39.33	51.77	43.22
Clashscore	5.61	4.55	5.09	4.94
Ramachandran plotd
favored	93.9	94.3	95.5	92.6
disallowed	0.2	0.2	0.2	0
MolProbity scoree	1.71	1.62	1.64	1.72
PDB entry	4ZFA	4ZF6	4ZF8	4ZFB

^aData in parentheses are for the highest-resolution shell.

^bPHENIX.³²

^cCalculated against a cross-validation set of 5% of the data selected at random prior to refinement.

^dMolProbity.³⁵

^eCombines Clashscore, rotamer, and Ramachandran evaluations into a single score, normalized to the same scale as X-ray resolution.

X-ray Structure Solution and Refinement.

The X-ray crystal structures of the P450_BM3 heme domains were determined by molecular replacement using PHASER within the PHENIX crystallography suite.^31,32 The coordinates that were used for molecular replacement were from the P450_BM3 heme domain in space group P4₁2₁2 [Protein Data Bank (PDB) entry 3NPL, no associated publication]. The structure was modified to remove all nonbonded atoms other than those in the heme prosthetic group, which were set to an occupancy of zero, and the sites of the five mutated amino acids were replaced with alanines. The raw data scaled as P622, and systematic absences along the 001 face suggested a 6-fold screw axis along c. When rotation and translation functions were performed on the probe against data from 10 to 4 Å, a clear solution arose for space group P6₅22. The solution contained one complex in the asymmetric unit. Model building was conducted in COOT using 2F_o – F_c electron density maps and maximum likelihood refinement with REFMAC5.^33,34 Stereochemical analysis and final adjustments to the model were directed by MolProbity.³⁵ WT and PM models were then used to determine complex cocrystal structures by removing nonbonded atoms, reassigning B factors to 15.00 Ų, and performing rigid body refinement against all data to a limit of 3.4 Å. The resulting models were then refined against all data, and the electron density of the bound small molecules was identified unambiguously from F_o – F_c difference maps. Statistics for the refined crystallographic models are listed in Table 1.

MD Simulations.

A total of 11 systems were simulated: (1) substrate-free WT, R47L, F81I, F87V, L188Q, E267V, and PM and (2) WT and PM in complex with palmitic acid (modeled as palmitate), and (3) WT and PM in complex with metyrapone. System preparation, minimization, heating, and equilibration are described in detail in the Supporting Information. The enzyme was modeled using the AMBER ff14SB force field³⁶ and the solvent using TIP3P.³⁷ Force field parameters for the heme–Cys moiety and partial charges for the high-spin, pentacoordinate form (characteristic of the substrate-bound enzyme) were taken from a previous study.³⁸ Partial charges for the low-spin, hexacoordinate (water being the sixth ligand) form (characteristic of the substrate-free enzyme) were calculated using the method described therein. The suitability of the heme–Cys parameters was verified by monitoring the rmsd and planarity of the heme (Figure S3). Parameters for unbound metyrapone and palmitate were derived using the antechamber module³⁹ at the HF/6-31G* level to be consistent with the GAFF force field.⁴⁰ The bond between the pyridine N and heme Fe in the metyrapone complexes was restrained using a harmonic force constant of 10 kcal mol⁻¹ Å⁻¹ and an equilibrium bond length of 2.6 Å.

Using AMBER 14,⁴¹ production MD simulations in the NVT ensemble at 300 K were run for 100 ns using the same simulation parameters that were used during equilibration. For the high-temperature simulations, 550 K was chosen for the observation of significant structural changes in the protein on a feasible simulation time scale (50 ns). A similar temperature (500 K) was used to study substrate-bound mesophilic (CYP101 and CYP176A) and thermophilic (CYP119, CYP231A2, and CYP175A1) P450s.⁴² Previous MD studies, employing a temperature as high as 600 K, have demonstrated that the unfolding mechanism of an enzyme remains essentially the same regardless of the simulation temperature.^43-45 Three independent trials, each with a new set of velocities, were performed for the 550 K simulations. Trajectories were analyzed using the cpptraj module of AmberTools 14.^41,46 Native contacts and salt bridge networks were determined from the 300 K simulations. A native contact is defined as contact between Cα atoms that (a) is at least three residues away in sequence, (b) is within a distance cutoff of 6.5 Å, and (c) has an occupancy of >67%. A salt bridge is defined as interaction between the O atom of Asp/Glu and the protonated N atom of Arg, Lys, or His within a distance cutoff of 4.0 Å.

RESULTS

Mutation-Induced Substrate Promiscuity Reduces P450_BM3 Stability.

It has previously been shown that the introduction of five mutations (R47L, F81I, F87V, L188Q, and E267V) into the heme domain of P450_BM3 allows the enzyme to shift from selectively oxidizing fatty acids to modifying larger, more complicated druglike molecules.^6,9,47,48 To gain insight into the effect of these mutations on the global stability of the protein, the heme domain of WT and the promiscuous mutant, PM, were subjected to chemical denaturation studies (Figure 1). The urea concentration midpoint for stability (C_m) was measured for WT and PM (Table 2), with a C_m of 5.4 M for substrate-free WT. The PM P450_BM3 protein was much less stable, with a midpoint of 2.7 M.

Figure 1.

Mutations in P450_BM3 that result in substrate promiscuity destabilize the protein. (A) Wild-type (WT) P450_BM3 equilibrated in urea followed by pulse proteolysis was resolved by SDS–PAGE, where the upper band corresponds to P450_BM3 and the lower band to thermolysin. The amount of intact P450_BM3 was quantified for each urea concentration (in molar units) to determine the midpoint for stability. (B) SDS–PAGE of the pentuple mutant, PM, after pulse proteolysis at varying urea concentrations. (C) Normalized data from panels A and B, where the fraction of intact P450_BM3 is plotted vs urea concentration. The filled squares correspond to data for PM and the filled circles to data for WT.

Table 2.

Urea Concentrations at Half-Denaturation (C_m, molar) of Substrate-Free and Substrate-Bound P450_BM3 Variants^a

	no substrate	metyrapone	palmitic acid
WT	5.4 ± 0.5	3.9 ± 0.2	4.9 ± 0.4
F81I	4.7 ± 0.1	4.8 ± 0.1	4.8 ± 0.1
E267V	4.4 ± 0.1	3.6 ± 0.2	4.5 ± 0.1
TM	3.2 ± 0.1	3.4 ± 0.2	3.3 ± 0.1
QM	3.2 ± 0.2	4.0 ± 0.2	3.3 ± 0.1
PM	2.7 ± 0.2	3.8 ± 0.2	2.7 ± 0.2

^aResults are from three independent experiments.

Because pulse proteolysis primarily uncovers changes in tertiary structure, the effect of the mutations on the stability of protein secondary structure was measured in the presence of urea by CD (Figure 2A). As anticipated, the WT and PM denaturation C_m values were higher than those observed by pulse proteolysis, with values of 5.9 M for WT and 4.1 M for PM. This is consistent with the expectation that loss of secondary structure will require higher concentrations of denaturant than for disruption of the tertiary structure. The stability of the heme cofactor environment was also measured by following the change in the Soret region at 415 nm (Figure 2B). The stability of the heme in PM was reduced compared to that of WT, with a C_m of 3.1 M versus a C_m of 5.4 M for WT. In both cases, the stability of the heme environment was lower than that of the overall protein secondary structure. Pulse proteolysis of the single-point variants, E267V and F81I, and multiple mutants, TM and QM, indicated that all mutations had a destabilizing effect on the substrate-free enzyme (Table 2).

Figure 2.

Mutations induce the destabilization of secondary and tertiary structure. (A) Changes in secondary structure measured by circular dichroism as a function of urea concentration. (B) Stability of the heme monitored in the Soret region. In all plots, data for WT are shown as filled circles and those for PM as filled squares.

Inhibitor and Substrate Binding Modulates Stability.

The effects of inhibitor binding on the stability of PM and WT were measured using metyrapone, which inhibits P450 enzymes through direct coordination of the pyridine nitrogen with the heme Fe; this prevents the enzyme from oxidizing its substrate. Surprisingly, the stability of the WT enzyme decreased when it was bound to metyrapone, from 5.4 to 3.9 M (Figure 3A). The destabilizing effect of metyrapone on WT was also observed by CD, with the C_m reduced from 5.9 to 5.1 M, and the heme environment reflected a disruption, with a reduction in C_m from 5.4 to 4.7 M (Figure 3C,D). In marked contrast, the stability of metyrapone-bound PM increased from a C_m of 2.7 M to a C_m of 3.8 M (Figure 3B). No significant modulations of the stability of the secondary structure were observed by CD [4.1 M vs 4.2 M (Figure 3C)], but the stability of the heme environment increased, with a shift from 3.1 to 4.1 M (Figure 3D). Metyrapone-bound E267V was less stable, while the F81I, TM, and QM complexes were more stable than their respective substrate-free forms, with the greatest effect seen for the least stable variants (Table 2).

Figure 3.

Metyrapone inhibitor shifts the stability of WT and PM in opposite directions. (A) WT stability as measured by pulse proteolysis, where filled circles correspond to data for the enzyme without the inhibitor and empty circles to data for the enzyme with the inhibitor. (B) PM stability measured by pulse proteolysis, where filled squares correspond to data for the enzyme without the inhibitor and empty squares to data for the enzyme with the inhibitor. (C) Stability of the secondary structure and (D) heme environment in the presence of metyrapone measured by circular dichroism. For all data in panels C and D, data for WT are shown as empty circles and data for PM as empty squares.

The WT P450_BM3 enzyme is selective for the hydroxylation of fatty acid substrates, such as palmitic acid. Binding requires the end of the fatty acid hydrocarbon chain to be in the proximity of the heme for hydroxylation to occur. Substrate binding was expected to increase the stability of WT based on previous studies.⁴⁹ Instead, only a minor change was observed, with a decrease to 4.9 M determined by pulse proteolysis and a slight decrease detected by CD, to 5.7 M. For PM, palmitic acid had no significant impact on the stability of the secondary structure, with the C_m decreasing from 4.1 to 4.0 M. This was also seen with pulse proteolysis, where the C_m was 2.7 M in the presence or absence of palmitic acid. The stabilities of E267V, F81I, TM, and QM were also unaffected by palmitic acid binding.

The X-ray Crystal Structure of the P450_BM3 Heme Domain Is in a Closed Conformation.

To directly observe the effects of mutation and ligand binding on the P450_BM3 structure, X-ray crystallography was conducted on WT and PM P450_BM3 proteins. The C-terminal His-tagged proteins used in this study failed to crystallize under conditions reported in previous structural studies.^50,51 However, crystals of a unique hexagonal crystal form grew under conditions of nickel and low-molecular weight polyethylene glycol (PEG). Refinement of the WT X-ray crystal structure was performed against diffraction data at a limit of 2.76 Å resolution and resulted in a model with excellent stereochemistry (4ZFA). The structure revealed that, in this hexagonal crystal form, P450_BM3 exhibits the closed conformation that is typically observed in substrate-bound complex models (Figure 4A). Movement between the open and closed conformations involves en bloc rotation of a segment composed roughly of amino acids 168–267 that encompass helices F–I, as well as the loops that join them (Figure 4B). Helices F and G, along with helix B′, form the lid domain of the substrate access channel, which is lined by the F/G loop on one side and the 3₁₀ helix (residues 16–20) and β-sheet 1 on the other (Figure 4C,D).

Figure 4.

Crystal structures of substrate-free (open conformation) and substrate-bound (closed conformation) WT P450_BM3. The protein is shown in cartoon representation, and the heme and residues of interest are shown as sticks. (A) The closed conformation, represented by 4ZFA, is colored gray, while the fully open conformation is represented by 1BU7,⁵⁰ molecule B (green). (B) Helices F, G, and I undergo significant displacement upon substrate binding. A PEG molecule found in the substrate channel of 4ZFA is shown in ball-and-stick form. E267 forms a salt bridge with K440 in both conformations and a hydrogen bond with T438 in the closed conformation. F81 forms contacts with L181, F205, I209, and I263, which rearrange upon substrate binding. (C and D) The region nomenclature and mutated residues in PM P450_BM3 are illustrated to aid in discussion. R47 is located in β-sheet 1, F81 in helix B′, F87 between helices B′ and C, L188 in helix F, and E267 in helix I.

Crystal packing is mediated by a nickel ion that is coordinated by amino acid residues D338 and E348 and the N-terminal threonine at the interface between crystallographic neighbors along the 6-fold screw axis. However, besides slight changes to the side chain geometries, this interaction does little to perturb the local structure relative to either the open or closed conformation P450_BM3 crystallographic models (1BU7,⁵⁰ molecule B, or 1FAG,⁵¹ respectively). Similarly, coordination of a nickel ion by the side chains of H138 and H426 appears to do little more than change these surface-exposed residues to alternate rotamers. H285 rotates to contact a surface-exposed nickel ion. Even less change is observed due to nickel ion binding at H236. Therefore, the involvement of nickel in mediating crystal packing does not appear to be responsible for the closed conformation observed in the substrate-free WT model.

Analysis of the heme does not reveal any differences from previously reported P450_BM3 crystal structures. What is evident within the proximity of the enzyme’s active site, however, is a strong peak of electron density that occupies the space previously observed to house palmitoleic acid in the complex X-ray cocrystal structure (1FAG). The ligand density refines best as a PEG fragment that is five ethylene glycol units in length (Figure 5A,B). As in the previously reported P450_BM3:palmitoleic acid complex model, the bound ligand displaces R47, F87, and L188, as well as I263 and L437. Therefore, it is apparent that binding of the linear PEG molecule at a distance of >8 Å from the heme iron within the base of the substrate access channel casts the WT P450_BM3 crystal structure in its closed conformation. The structure also contains an ethylene glycol molecule bound at the active site where a water molecule is typically found (Figure 5C,D).

Figure 5.

Omit maps for bound ligands in WT and PM P450_BM3 X-ray crystal structures. (A) 2F_o – F_c difference electron density map for the crystallographic model of WT (4ZFA) contoured at 2.0σ with a stick model showing the heme and PEG. (B) F_o – F_c omit map contoured at 7.0σ and calculated for a modified final WT model lacking only the PEG, shown covering the same portion of the final refined model. (C) Portion of the final refined 2F_o – F_c difference electron density map (blue) for the WT model (4ZFA) contoured at 2.0σ with the corresponding portion of the final refined model containing the heme and bound ethylene glycol, shown as sticks (yellow carbon atoms). (D) F_o – F_c omit map (green/red) contoured at 9.0σ and calculated for a modified final WT model lacking only the ethylene glycol molecule shown covering the same portion of the final refined model. (E) 2F_o – F_c difference electron density map for the crystallographic model of PM bound to palmitic acid (4ZFB) contoured at 2.0σ with a stick model showing the heme and palmitic acid. (F) F_o – F_c omit map contoured at 7.0σ and calculated for a modified final PM:palmitic acid complex model lacking only the palmitic acid, shown covering the same portion of the final refined model. (G) 2F_o – F_c difference electron density map for the crystallographic model of PM bound to metyrapone (4ZF8) contoured at 2.0σ with a stick model showing the heme bound to metyrapone. (H) F_o – F_c omit map contoured at 9.0σ and calculated for a modified final PM:metyrapone complex model lacking only the metyrapone, shown covering the same portion of the final refined model.

The X-ray Crystal Structure of PM P450_BM3 Is Similar to That of WT.

Crystals of the PM P450_BM3 formed in the same space group as the WT protein, and X-ray diffraction data of nearly identical quality were obtained for both crystals. Independent solution and refinement of the PM P450_BM3 X-ray crystal structure revealed that it also adopts the closed conformation (4ZF6), with a PEG molecule positioned within the normally solvent-filled base of the substrate access channel.

Though the replacement of five amino acid side chains in PM does not significantly influence the fold or conformation of the P450_BM3 protein, there are a few notable differences (Figure 6). Replacement of R47 with leucine creates a cavity into which the side chain of Q73 moves, positioning it for interaction with Q188, which occupies the same position in WT near the end of the bound PEG. The E267V mutation eliminated the interaction with the T438 and K440 side chains observed in the WT closed conformation model.

Figure 6.

Crystal structures of WT (4ZFA, gray) and PM (4ZF6, green) P450_BM3. The protein is shown in cartoon representation, and the heme and mutated residues (47, 81, 87, 188, and 267) are shown as sticks. The PEG molecule in the channel is not shown for the sake of clarity. E267 forms interactions with T438 and K440, which are lost upon mutation to valine. Q73 moves into the cavity left by the R47L mutation and could potentially interact with Q188.

Mutation of F87 to valine creates space above the heme porphyrin ring system adjacent to the O₂ binding site. The effect of this mutation on the environment at the O₂ binding site is evidenced by the severe change in orientation of the proximally bound ethylene glycol molecule relative to its position in the WT structure. Moreover, replacement with valine obviates the need for movement of the longer F87 side chain upon substrate binding. That movement, together with similar rearrangement of the I263 side chain, creates a hydrophobic cavity into which the hydrocarbon tail of long chain fatty acids can anchor themselves. Therefore, the F87V mutation expands the active site, which would allow closer approach of substrates to the heme.

Comparison of the crystal structures of the PM:PEG complex (4ZF6) and the substrate-free R47L/Y51F/F87V/E267V/I401I mutant (4RNS⁵²) (Figure S4A) suggests that the predominant form of PM in the absence of a ligand in the channel would be the closed conformation. Among the common substitutions of the two mutants, only L47 differed significantly in side chain position. Additionally, the neighboring Q73 side chain points outward unlike in PM (Figure S4B).

X-ray Cocrystal Structures of PM P450_BM3 with Substrate or Inhibitor.

Complexes of PM P450_BM3 with the ligands palmitic acid (substrate) and metyrapone (inhibitor) were crystallized in the same crystal form as the WT and PM proteins described previously, and X-ray cocrystal structures were refined to 2.84 Å (4ZFB) and 2.77 Å (4ZF8), respectively. The PM:palmitic acid complex model is similar to the WT P450_BM3 structure in complex with palmitoleic acid. In both, the carboxylate group of the substrate forms a hydrogen bond with the hydroxyl group of Y51. However, the PM P450_BM3 also involves the side chain of Q188 in anchoring the substrate. At the opposite end, the F87V mutation leaves a hydrophobic pocket that is amenable to accommodating the hydrocarbon tail without requiring movement of the phenylalanine side chain. In general, the PM P450_BM3 appears to be better suited to stably enfold its substrate, palmitic acid, in its binding site (Figure 5E,F).

The PM:metyrapone complex crystal structure reveals a clear density for the inhibitor ligand at the active site and no evidence of elongated electron density in the vicinity of the fatty acid/PEG binding site (Figure 5G,H). The pyridine nitrogen of metyrapone coordinates with the heme iron, as observed in complexes of P450_cam and P450 3A4.^53,54 The Fe–N distance in the P450_BM3 PM:metyrapone complex is 2.6 Å. The average error in coordinate positions throughout the model as estimated by the maximum likelihood target function is 0.4 Å. The F87V mutation allows for metyrapone binding because the F87 side chain in either the substrate-bound or unbound conformation of WT would collide with metyrapone as it appears in the refined PM:metyrapone crystallographic model. This provides direct evidence of how mutating residues near the heme active site can significantly alter both the activity and the specificity of the P450_BM3 enzyme toward substrates.

In addition to these differences at the site of the heme, the lack of a fatty acid or PEG molecule within the enzyme causes the leucine side chain that substitutes for R47 in the PM enzyme to move into a solvent-exposed position. It is unclear what stabilizes this conformation. However, it appears that the absence of a fatty acid or its analogue allows for conformational flexibility within this region. None of the remaining mutated residues, F81I, L188Q, or E267V, appear to change relative to their positions in the PM and PM:palmitic acid complex crystallographic models.

Molecular Dynamics Simulations.

Simulations at 300 K were performed primarily to determine the native contacts in each system. Several interesting structural changes were also observed during the simulations. In substrate-free WT, E267V, and L188Q (modeled from 1BU7, molecule B), helix A and β-sheets 1-1 (residues 38–44) and 1-2 (residues 47–53) moved closer to the core of the protein, resulting in a partially closed substrate channel (Figure S5A and Table S1) similar to 1BU7 (molecule A). Helices D–H and β-sheet 4 were also shifted in position in E267V because of the elimination of the salt bridge with K440. Structural changes in R47L, F81I, and F87V were relatively less significant and mainly occurred in helices F and G (Table S1). On the other hand, the lid domain of substrate-free PM opened from the initial closed conformation (4ZF6) during the 300 K simulation (Figure S5B). For the substrate-bound enzymes, larger backbone rmsds, with respect to the crystal structure, were observed for PM:palmitic acid (1.5 Å) and PM:metyrapone (1.1 Å) complexes. In addition to the movement of helix F to the N-terminal end of helix I, the substrates also changed in position in the PM active site. Palmitic acid moved closer to the heme, with its carboxylate group within hydrogen bonding distance of S72 (Figure S5C). Metyrapone rotated slightly, bringing the uncoordinated pyridine ring in close contact with V87, T260, and I263 (Figure S5D).

The unfolding of substrate-free and substrate-bound WT and PM was simulated using high-temperature (550 K) MD to rationalize the biochemical results. Simulations of the individual mutations R47L, F81I, F87V, L188Q and E267V were also conducted to isolate the effect of each mutation on stability. As a technique, the thermally induced unfolding of these proteins provides mechanistic insights into regions of relative instability, which are identified by monitoring the native contacts of each residue. Native contact plots, averaged over three simulations, for helices F and G [residues 171–226 (Figure S6)] and the Cys ligand loop [residues 393–400 (Figure S7)] of substrate-free WT and PM are provided in the Supporting Information as examples. Approximately 50% of the contacts were lost at the end of the 50 ns run for all simulations of substrate-free and substrate-bound enzymes.

The different regions of P450_BM3 are illustrated in panels C and D of Figure 4 to aid in the discussion of the results. Loss of contacts in substrate-free WT began at 3 ns in β-sheet 3, particularly those at the C-terminal end of the enzyme. A few contacts at helix B′ were broken at 4 ns, followed by those of helix G residues near the F/G loop at 6 ns (Figure S6A). In all simulations, the F/G loop and helix B′ continued to unfold, along with β-sheet 1 and the N-terminal end, exposing the substrate channel (Figure 7A and Figure S8A,B).

Figure 7.

Snapshots at 25 ns from the 550 K simulations of substrate-free (A) WT and (B) PM viewed through the channel entrance. The protein is shown in cartoon representation, with helices colored purple, β-sheets green, and loops blue. The heme, bound water molecule, Cys ligand, and F393 at the other end of the loop are shown as sticks. The hydrogen bond distance between the Cys amide H and F393 carbonyl O (d_O–H) is represented by a dashed line. The Cys ligand loop is still intact in WT (d_O–H = 1.75 Å) but has begun to unfold in PM (d_O–H = 3.98 Å).

The total number of native contacts in substrate-free PM is lower (659) than in WT (680), which could explain the loss of PM stability. The increase in backbone rmsd within the first 10 ns was slightly faster than in WT (Figure S9A,B). In addition to helix G, contacts were also lost in helices C and F (Figure S6B), the Cys ligand loop [a β-bulge segment below the heme (Figure S7B)], and β-sheet 1 within the first 5 ns. This was followed by helix K′ and β-sheet 2, which are connected to the latter, and helix D. The unfolding process in all simulations is similar to that of WT, but with helices D and E unfolding concurrently (Figure 7B and Figure S8C,D). Among the single mutants, early loss of contacts in the β-sheets was also observed with L188Q, helix D with L188Q and E267V, and the Cys ligand loop with F81I, L188Q, and E267V. Surprisingly, F87V and R47L did not exhibit a significant decrease in the number of native contacts in the first 10 ns, although the increase in backbone rmsd was similar to those of the other mutants (Figure S9C-G).

The unfolding of metyrapone-bound WT also began at β-sheet 3, albeit earlier in the simulation (1 ns). Unlike the substrate-free enzyme, the Cys ligand loop started to unfold after 4 ns, followed by helices C, F, and G and β-sheet 4 (5–6 ns). The latter is inserted into the channel and forms close contacts with the substrate. Shortly thereafter, the other side of the channel (β-sheets 1 and 2 and helix A) unfolded as well (7 ns). In comparison, fewer regions unfolded in the PM:metyrapone complex within the first 10 ns. Helix B′ lost contacts first (2 ns) followed closely by helices F and G (3–4 ns) and then, β-sheet 1 (6 ns). As with the substrate-free and metyrapone-bound WT, helix G (3 ns) and β-sheet 3 (5 ns) in the WT:palmitic acid complex were involved in the early stages of unfolding. For the PM:palmitic acid complex, loss of contacts began at β-sheet 1 (3 ns), followed by β-sheet 3 (4 ns), helices G and H (5 ns), the Cys ligand loop (5 ns), β-sheet 4 (7 ns), and helix J′ (8 ns). Generally, the substrate-bound variants displayed the same unfolding process as the substrate-free ones (Figures S10 and S11). Each side of the substrate access channel continued to unfold, with the destruction of helices F and G extending to β-sheet 5 and, in some cases, even to part of helix I. The unfolding of the other portion of the lid domain (helix B′) also extended to helix C, which consequently allowed the solvent to enter the active site. In contrast, a few strands of β-sheet 1 on the other side of the channel remained intact at the end of the simulation. For all substrate-bound enzymes, the hierarchy of unfolding in each trial simulation is essentially similar. However, the rate at which each side unfolds usually differs after ~10 ns (Figure S12); hence, the extent of damage to the channel at the end of the 50 ns simulation was not the same.

DISCUSSION

The effects of mutations in the substrate access channel (R47L, F81I, F87V, L188Q and E267V) and binding of the ligands palmitic acid and metyrapone on protein conformation and stability were analyzed using experimental and computational methods. Pulse proteolysis, which probes the global stability of a protein, indicated that PM is less stable than WT P450_BM3. The crystal structures of WT and PM provide a possible explanation of how the mutations destabilize the enzyme. At the entrance of the channel, mutation of residue 188 in helix F from the nonpolar leucine to the polar asparagine can impact its interaction with Q73 in helix B′. Previous studies have shown that variants with substitution at or near this residue (L188P, F162I/K187E, F162I/K187E/M237I, and F162I/K187E/L188P/M237I) are less stable.^55,56 Within the channel, the mutation of E267 in helix I eliminated the salt bridge with K440 in β-sheet 4, which would account for the 1 M decrease in the C_m of E267V. However, its effect on global stability appears to be less significant when combined with R47L, F87V, and L188Q mutations, as the similar C_m of TM and QM indicates. F81, located in the lid domain (helix B′), is within contact distance of L181 in helix F, F205 and I209 in helix G, and I263 in helix I. Its mutation to the smaller isoleucine would therefore affect hydrophobic packing in this region, although pulse proteolysis of F81I suggests that this has a relatively minor effect on stability compared to the other mutations.

The chemical denaturation stabilities could be correlated to high-temperature simulations demonstrating the effect of mutations on the unfolding process. A linear correlation between C_m and T_m determined from heat inactivation curves of CO binding difference spectra was observed for P450_BM3.¹¹ An MD study of chymotrypsin inhibitor 2 also indicated that the overall unfolding mechanisms of chemical denaturation with urea and thermal denaturation are similar, in that key residues in the hydrophobic core are exposed first.⁵⁷ Monitoring the decrease in the number of native contacts per residue over time revealed that the most labile regions of substrate-free WT P450_BM3 are helices B′ and G in the lid domain and β-sheet 3 at the C-terminal end of the protein. Hydrophobic contacts in the lid domain, aside from those observed in the crystal structure, include F205 (helix G)–A180 (helix F) and V211/M212 (helix G)–F173 (helix F) contacts.

Additional regions were shown to unfold earlier in PM, notably β-sheet 1 and the Cys ligand loop. Early unfolding of the Cys ligand loop is consistent with CD data indicating that the heme environment of substrate-free PM is less stable than that of WT. L188Q similarly lost contacts early in both regions, suggesting that this mutation has a significant contribution to destabilization. Simulations of the single-point mutants at 300 K showed that the L188Q mutation caused the most significant structural change, based on backbone rmsd at the N-terminal end up to β-sheet 1 of the substrate-free enzyme (Table S1), which is presumably destabilizing and might have affected the Cys ligand loop. On the basis of analysis of time-correlated atomic motions, these two regions were assigned to the same protein domain, which implies that they move cooperatively and cohesively during structural transitions of the protein.⁵⁸

As for E267, simulations confirmed the importance of its salt bridge with K440 (Table S2), as it was maintained even at high temperature in WT. The concurrent unfolding of helices D and E with sections of the substrate channel in PM may also be attributed to the E267V mutation, which disrupted hydrophobic contacts such as L150 (helix E)–H266 (helix I) and I122 (helix D)–L148 (helix E) contacts. R47L did not exhibit significant unfolding during the early stages of high-temperature simulations, despite the elimination of a salt bridge with another β-sheet 1 residue, E352 (Table S2). This may be explained by the fact that leucine is a better β-sheet former.⁵⁹ F87V was also relatively stable during the simulations, which is inconsistent with differential scanning calorimetry data showing that its melting temperature is lower by ~4 °C than that of WT.⁴⁹ The discrepancy may be attributed to the difficulty in accurately predicting slight differences in stability in the case of conservative (in terms of hydrophobicity) substitutions such as F81I and F87V using MD simulations, which is exacerbated by the lack of available crystallographic evidence; initial coordinates for single-point mutant and WT simulations were all derived from 1BU7 (molecule B), which might not be adequately representative of the F87V structure.

Substrate binding generally stabilizes an enzyme.^27,60 However, pulse proteolysis and CD data demonstrated that this was not the case with the WT:metyrapone complex. High-temperature simulations showed that additional regions, particularly the Cys ligand loop, unfolded early, which is consistent with the observed decrease in C_m for the heme environment. In contrast, the stability of the heme region of PM was enhanced by metyrapone, although it is not certain whether this can be attributed to covalent bonding with the heme iron because simulations of P450_cam in complex with 4-phenylimidazole (Fe–N bond length of 2.21 Å in the crystal structure⁵³) showed instability of the Cys ligand loop.⁴² However, a new salt bridge, E380–K312, formed during the simulation (Table S2), and its proximity to the heme might contribute to the stabilization of PM. On the other hand, the stability of WT and PM was not significantly affected by the presence of palmitic acid, as indicated by pulse proteolysis. Although the labile regions in the palmitic acid-bound complex of WT and PM are different from those of the corresponding substrate-free enzyme, the unfolding process was observed to be similar. The Cys ligand loop unfolded early in both forms of PM but was relatively stable in both forms of WT.

Increasing the stability of PM is advantageous if it is to be used as a starting point to develop more promiscuous variants. The lid domain, β-sheet 1, and Cys ligand loop would be the logical targets for mutation based on the hypothesis that delaying the unfolding of the most labile regions would increase the global stability of the protein. On the basis of analysis of physical factors that differentiate (hyper)thermophilic proteins from mesophilic ones,^61-63 several approaches for protein stabilization have been attempted such as introducing salt bridges,⁶⁴ increasing structural rigidity,⁶⁵ and improving hydrophobic core packing.⁶⁶

The latter method is promising for P450_BM3 on the basis of a previous study of the 21B3 variant of this enzyme. Five of the eight mutations introduced through directed evolution, resulting in an increase of 15 °C in T₅₀, happened to conserve hydrophobicity. L52I (β-sheet 1) and A184V (helix F) are buried residues, while L324I (helix K), V340M (β-sheet 2), and I366V (helix K″) are located on the surface. The authors hypothesized that these stabilizing mutations counteracted structural perturbations caused by previously introduced activity-enhancing mutations given the proximity of some residues in the two sets of mutations.²⁵ Interestingly, most of the stabilizing mutations are located at or near the labile regions identified in this study. Packing interactions in PM P450_BM3 can be enhanced by introducing substitutions at the hydrophobic patches in the (a) lid domain, consisting of I81, A180, L181, F173, F205, I209, V211, M212, and I263, and (b) heme region, including W367, F379, and F390. This cluster of aromatic residues presumably stabilizes the Cys ligand loop, as is the case in the thermophilic CYP119.⁴²

Salt bridges contribute to the stability of thermophilic P450s at elevated temperatures through hydration effects.⁶⁷ For instance, CYP175A1 (T_m = 88 °C), whose closest homologue is P450_BM3 (26% sequence identity), has eight salt bridge networks (i.e., one that involves more than two charged residues).⁶⁸ MD simulations of P450_BM3 variants showed that there are only three salt bridge networks in substrate-free WT (R323/E320/R378/D370, K3/E344/R56/E38, and K94/E247/K98) and four in PM (R375/D370/R378, K3/E344/R56/E38, K94/E247/K98/D250, and E292/R296/E293) (Table S2). At the N-terminal end of P450_BM3, T339 (β-sheet 2) can be mutated to Asp to form an extended network with R66 (β-sheet 1) and E60 (helix B). New salt bridges in the other labile regions can also be introduced, for example, the D388 (mutated from His)–K391 salt bridge in the loop region near the heme and the K445 (mutated from Val)–E140 salt bridge in β-sheet 3.

Pulse proteolysis, CD spectroscopy, X-ray crystallography, and MD simulations elucidated how mutations that impart affinity for nonnative compounds such as metyrapone destabilize P450_BM3. The method-dependent variation in stability measurements indicates that use of both global and active site characterizing methods is necessary to effectively evaluate the effects of mutation on P450_BM3 stability. Destabilization generally arose from disruption of important salt bridges and hydrophobic contacts and unfolding of the Cys ligand loop connected to the heme. The identification of the conserved Cys ligand loop as a key contributor to instability suggests that some findings may be generalizable to other CYP102A subfamily fatty acid hydroxylases.⁶⁹ A commonly held view among the protein design community is that increasing the promiscuity of an enzyme comes at the expense of protein stability. However, Arnold et al. demonstrated that this is not necessarily the case, as more thermostable variants have been produced through further directed evolution of an existing mutant without compromising enzyme activity.^11,25,26 The characterization of the unfolding process of the P450_BM3 variants presented herein provides fundamental knowledge that could be used to rationally design stability upon an enhanced specificity platform. Regions involved in the early stages of unfolding are potential targets for mutation to develop thermostable variants of the promiscuous PM P450_BM3.

ABBREVIATIONS

WT

wild-type

TM

R47L/F87V/L188Q

QM

R47L/F87V/L188Q/E267V

PM

R47L/F87V/L188Q/E267V/F81I

CD

circular dichroism

MD

molecular dynamics

Tris

2-amino-2-hydroxymethylpropane-1,3-diol

EDTA

ethylenediaminetetraacetic acid

PEG

polyethylene glycol

SDS–PAGE

sodium dodecyl sulfate–polyacrylamide gel electrophoresis

rmsd

root-mean-square deviation