Engineering Streptomyces coelicolor Carbonyl Reductase for Efficient Atorvastatin Precursor Synthesis

ABSTRACT

Streptomyces coelicolor CR1 (ScCR1) has been shown to be a promising biocatalyst for the synthesis of an atorvastatin precursor, ethyl-(S)-4-chloro-3-hydroxybutyrate [(S)-CHBE]. However, limitations of ScCR1 observed for practical application include low activity and poor stability. In this work, protein engineering was employed to improve the catalytic efficiency and stability of ScCR1. First, the crystal structure of ScCR1 complexed with NADH and cosubstrate 2-propanol was solved, and the specific activity of ScCR1 was increased from 38.8 U/mg to 168 U/mg (ScCR1_I158V/P168S) by structure-guided engineering. Second, directed evolution was performed to improve the stability using ScCR1_I158V/P168S as a template, affording a triple mutant, ScCR1_{A60T/I158V/P168S}, whose thermostability (T₅₀¹⁵, defined as the temperature at which 50% of initial enzyme activity is lost following a heat treatment for 15 min) and substrate tolerance (C₅₀¹⁵, defined as the concentration at which 50% of initial enzyme activity is lost following incubation for 15 min) were 6.2°C and 4.7-fold higher than those of the wild-type enzyme. Interestingly, the specific activity of the triple mutant was further increased to 260 U/mg. Protein modeling and docking analysis shed light on the origin of the improved activity and stability. In the asymmetric reduction of ethyl-4-chloro-3-oxobutyrate (COBE) on a 300-ml scale, 100 g/liter COBE could be completely converted by only 2 g/liter of lyophilized ScCR1_{A60T/I158V/P168S} within 9 h, affording an excellent enantiomeric excess (ee) of >99% and a space-time yield of 255 g liter⁻¹ day⁻¹. These results suggest high efficiency of the protein engineering strategy and good potential of the resulting variant for efficient synthesis of the atorvastatin precursor.

IMPORTANCE Application of the carbonyl reductase ScCR1 in asymmetrically synthesizing (S)-CHBE, a key precursor for the blockbuster drug Lipitor, from COBE has been hindered by its low catalytic activity and poor thermostability and substrate tolerance. In this work, protein engineering was employed to improve the catalytic efficiency and stability of ScCR1. The catalytic efficiency, thermostability, and substrate tolerance of ScCR1 were significantly improved by structure-guided engineering and directed evolution. The engineered ScCR1 may serve as a promising biocatalyst for the biosynthesis of (S)-CHBE, and the protein engineering strategy adopted in this work would serve as a useful approach for future engineering of other reductases toward potential application in organic synthesis.

INTRODUCTION

Chiral secondary alcohols are important intermediates and precursors for producing pharmaceuticals, flavors, and fragrances (1). For example, ethyl-(S)-4-chloro-3-hydroxylbutyrate [(S)-CHBE] is a key intermediate for hydroxymethylglutaryl-coenzyme A (HMG-CoA) reductase inhibitors, the precursor of the cholesterol-lowering drug Lipitor (2, 3). In the past decade, great effort has been made toward the practical synthesis of (S)-CHBE. Compared with chemical synthesis, biocatalytic asymmetric synthesis of (S)-CHBE from ethyl-4-chloro-3-oxobutyrate (COBE) has attracted more and more attention because it provides eco-friendly reaction conditions, satisfactory selectivity, 100% theoretical yield, and fewer by-products, with simpler and easier operations (3–6). Therefore, several biocatalytic approaches for the synthesis of (S)-CHBE from COBE employing different carbonyl reductases have been developed (3, 7–11). However, these methods usually require expensive NADH or NADPH as a cofactor, which represents a great challenge for practical application. Due to the high cost of the cofactor (NADH or NADPH), efficient and cost-effective in situ cofactor regeneration is very important for the economic viability of industrial-scale biotransformation (12, 13). Two systems have already been developed for cofactor regeneration. One is the enzyme-coupled system, which employs formate dehydrogenase or glucose dehydrogenase to recycle nicotinamide coenzyme (14); the other is the substrate-coupled system, which recycles NAD⁺ or NADP⁺ with 2-propanol as a cosubstrate due to its low cost and the feasibility of pushing the reaction toward completion by removing the coproduct, acetone, under reduced pressure (15).

In our previous work (16), a novel NADH-dependent reductase, Streptomyces coelicolor CR1 (ScCR1), was discovered by genome data mining, which could efficiently convert COBE to (S)-CHBE with high enantiomeric excess (ee) (>99%) using 2-propanol as a cosubstrate (Fig. 1). By process optimization and scale-up, the specific production could reach levels as high as 36.8 g product/g biocatalyst using lyophilized recombinant Escherichia coli cells expressing ScCR1 as biocatalyst, the highest value reported so far (17). However, the specific activity of ScCR1 toward COBE was relatively low (38.8 U/mg protein); therefore, high biocatalyst loading (e.g., 20 g/liter) was required, and significant emulsion was formed during downstream processing, which would in turn increase the cost of the bioprocess (16). Meanwhile, ScCR1 shows a thermostability (T₅₀¹⁵, where T₅₀¹⁵ is defined as the temperature at which 50% of initial enzyme activity is lost following a heat treatment for 15 min) of 47.4°C and a substrate tolerance (C₅₀¹⁵, where C₅₀¹⁵ is defined as the concentration at which 50% of initial enzyme activity is lost following incubation for 15 min) of 34 mM. The poor thermostability and substrate tolerance of ScCR1 represent another hurdle for its practical application.

FIG 1.

Bioreduction of COBE for the synthesis of an atorvastatin precursor by the novel reductase ScCR1.

Protein engineering has been shown to be a powerful tool for engineering the catalytic properties (e.g., activity, thermostability, and stereoselectivity, etc.) of biocatalysts (18–23), especially in industrial processes (2, 24). However, little is known about the engineering of the carbonyl reductase involved in asymmetric reduction of COBE for the synthesis of (S)-CHBE. Here, we report on the progress of the engineered ScCR1 with increased catalytic activity, thermostability, and substrate tolerance.

RESULTS

Structure-guided engineering to improve the activity toward COBE.

To increase the positive hits for identifying ScCR1 variants with improved catalytic activity and to keep the screening effort at a minimum, the structure of ScCR1 in complex with NADH and the cosubstrate 2-propanol was solved and refined to 2.3 Å (Table 1; see Fig. S2 in the supplemental material). The substrate COBE was then docked into the active site of ScCR1, and amino acids (Ala156, Ile158, Leu159, Gly160, Phe164, Ser167, Pro168, and Val171) surrounding the substrate within 8 Å, with the exception of catalytic residues (Ser157, Tyr170, and Lys174) and conserved amino acids in classical short-chain dehydrogenase (see Fig. S3 in the supplemental material), were subjected to saturation mutagenesis. Two mutants, the I158V and P168S mutants, with 2.4-fold- and 3-fold-higher specific activities than the wild-type enzyme, respectively, were identified. These two mutations were then combined, aiming to further increase the catalytic activity, and as expected, the specific activity of the resultant double mutant (I158V/P168S) reached as high as 168 U/mg, which was about 4.3-fold that of the wild-type enzyme (Table 2). It should be noted that substitution of Phe164 completely deleted the enzyme activity, indicating that this residue is irreplaceable. Although active mutants could be found from mutations at the other 5 positions, the activities of the most active mutants with mutations at positions Ala156, Gly160, Ser167, and Val174 were all less than 60% of that of the wild-type enzyme, and the activity of the most active mutant, mutated at position Leu159, was even worse, at about 30% of that of the wild-type enzyme. The expression levels of these mutants were very close to that of the wild-type enzyme (see Fig. S4 in the supplemental material), indicating that introduction of the two mutations into ScCR1 did not significantly affect its protein expression level.

TABLE 1.

Data collection and refinement statistics for ScCR1 in complex with NADH/isopropanol

Parameter	Value(s) for ScCR1/NADH/isopropanola
PDB ID	5H5X
Data collection statistics
Wavelength (Å)	1.5418
Space group	P31
Unit cell	a = b = 187.79 Å; c = 80.86 Å; α = β = 90.0°; γ = 120.0°
Resolution (Å)	32.5–2.3 (2.38–2.3)
No. of measured reflections	251,561
No. of unique reflections	118,981 (11,886)
Completeness (%)	84.0 (83.7)
Rmergeb (%)	0.14 (0.63)
Redundancy	2.1 (2.2)
Avg I/σ	4.0 (1.3)
Refinement statistics
No. of reflections	118,935
Rwork/Rfreec	0.179/0.243
RMSDd
Bonds (Å)	0.019
Angles (°)	1.42
Ramachandran plot (%)
Favored	94.8
Allowed	4.5
Outliers (%)	0.7
Avg B factor (Å2)	35

^aNumbers in parentheses refer to data for highest-resolution shells.

^bR_merge = ∑_hkl∑_i|I_i − 〈I〉|/∑_hkl∑_i|〈I〉|, where I_i is the intensity for the ith measurement of an equivalent reflection with indices h, k, and l.

^cR_free was calculated with 5% of reflections set aside randomly throughout the refinement.

^dRMSD, root mean square deviation.

TABLE 2.

Specific activities and thermostabilities of ScCR1 and its variants^a

Enzyme	Sp actb (U/mg protein)	T5015 (°C)
ScCR1	38.8 ± 0.8	47.4 ± 0.2
ScCR1I158V	93.2 ± 0.5	44.8 ± 0.5
ScCR1P168S	116 ± 5	49.1 ± 0.6
ScCR1I158V/P168S	168 ± 2	49.2 ± 0.1
ScCR1A60T	40.2 ± 0.6	50.5 ± 0.3
ScCR1A60T/I158V/P168S	260 ± 3	53.6 ± 0.2

^aValues are means and standard deviations.

^bSpecific activity of the purified enzyme toward COBE.

Directed evolution for thermostability enhancement.

After successfully increasing the catalytic activity of ScCR1 to a satisfactory level, we then focused on its thermostability, since poor stability of ScCR1 is another bottleneck for practical application. At first, the thermostabilities (T₅₀¹⁵) of the previously identified mutants with higher specific activities than the wild-type enzyme were measured. Interestingly, besides catalytic activity, variant ScCR1_I158V/P168S also showed slightly better thermostability (1.8°C increase in T₅₀¹⁵) than the wild-type enzyme (Table 2). To further improve the thermostability, error-prone PCR was performed using ScCR1_I158V/P168S as the template. After screening about 5,000 colonies, six variants with obviously better thermostability than the wild-type enzyme and the double mutant and with comparable catalytic activity were selected for further confirmation in shaking flasks (see Fig. S5 in the supplemental material). Finally, a triple mutant, ScCR1_{A60T/I158V/P168S}, showing 6.2°C and 4.4°C increases in T₅₀¹⁵ compared to those of the wild-type and the double mutant, respectively, was obtained. Interestingly, activity assay showed that the specific activity of ScCR1_{A60T/I158V/P168S} was further increased to 260 U/mg, which is 6.7-fold higher than that of the wild-type enzyme (Table 2; see Table S1 in the supplemental material). The protein folding prediction software FoldX, which calculates free energies of folding from van der Waals interactions, solvation energy of polar and apolar groups, hydrogen bonds, water bridges, electrostatic contributions, and entropy changes of the side chain and main chain, was also employed to help design more stable variants (25). Among the 4,997 potential point mutations calculated by FoldX, 3 potential candidates (ScCR1_{A60T/I158V/P168S/S135L}, ScCR1_{A60T/I158V/P168S/E139M}, and ScCR1_{A60T/I158V/P168S/A184F}) were selected for experimental validation, since they were predicted to stabilize the enzyme with a ΔΔG of <−4 kJ/mol (see Fig. S6 in the supplemental material). Unfortunately, only ScCR1_{A60T/I158V/P168S/E139M} showed slightly better thermostability than the starting triple mutant ScCR1_{A60T/I158V/P168S} (a 1.3°C increase in T₅₀¹⁵) but with 20% loss of activity (Table S1). Therefore, this variant was not considered in our following experiments.

Kinetic constants.

The kinetic constants of the purified ScCR1 and its variants were investigated and calculated based on Lineweaver-Burk double-reciprocal plots (Table 3). The steady-state kinetic constants for COBE indicated that the catalytic efficiencies of the variants were obviously higher than those of the wild-type enzyme, with the exception of ScCR1_A60T. Although the catalytic efficiency of ScCR1_A60T was comparable to that of the wild-type enzyme, it was very important for thermostability, since it could increase the T₅₀¹⁵ by 3°C. For the variant ScCR1_I158V/P168S, the catalytic efficiency (k_cat/K_m) toward COBE was about 5.6-fold that of the wild-type enzyme, while for the variant ScCR1_{A60T/I158V/P168S}, although the K_m was increased by 1.7-fold, a 12-fold increase in k_cat contributed to a 6.7-fold increase in k_cat/K_m. In the case of ScCR1_{A60T/I158V/P168S} toward NADH, a 5.3-fold decrease in K_m and a 7.0-fold increase in k_cat resulted in a 37-fold higher catalytic efficiency (k_cat/K_m) than that of the wild-type enzyme. The kinetic constants for 2-propanol illustrated that the catalytic efficiencies of all the variants were similar to those of the wild-type enzyme, with the exception of ScCR1_A60T. In the case of ScCR1_A60T, a 3.5-fold decrease in K_m and a 2.3-fold increase in k_cat resulted in an 8.0-fold increase in k_cat/K_m. It is noteworthy that the catalytic efficiency of ScCR1_A60T toward NAD⁺ was also higher (2.8-fold) than those of the wild-type enzyme and other variants.

TABLE 3.

Apparent kinetic constants of ScCR1 and its variants

Enzyme	Parameter	Valuea for:
		COBE	NADH	Isopropanol	NAD+
ScCR1	Km (mM)	0.59 ± 0.01	0.048 ± 0.002	217 ± 13	0.24 ± 0.06
	kcat (s−1)	20.5 ± 1.5	35.6 ± 1.4	39.2 ± 4.1	34.5 ± 4.3
	kcat/Km (s−1 mM−1)	34.7	742	0.181	144
ScCR1I158V	Km (mM)	0.44 ± 0.02	0.021 ± 0.001	187 ± 3	0.12 ± 0.01
	kcat (s−1)	39.3 ± 1.0	47.5 ± 0.8	40.5 ± 1.2	33.7 ± 2.1
	kcat/Km (s−1 mM−1)	89.3	2.26 × 103	0.217	281
ScCR1P168S	Km (mM)	0.54 ± 0.03	0.0031 ± 0.0002	82.7 ± 3.2	0.25 ± 0.01
	kcat (s−1)	54.4 ± 1.8	48.0 ± 0.9	35.6 ± 1.1	34.2 ± 1.4
	kcat/Km (s−1 mM−1)	101	1.55 × 104	0.430	137
ScCR1I158V/P168S	Km (mM)	0.92 ± 0.07	0.012 ± 0.003	183 ± 20	0.17 ± 0.04
	kcat (s−1)	180 ± 13	178 ± 10	55.5 ± 8.3	55 ± 5.2
	kcat/Km (s−1 mM−1)	196	1.48 × 104	0.303	324
ScCR1A60T/I158V/P168S	Km (mM)	1.0 ± 0.1	0.0090 ± 0.0002	202 ± 24	0.16 ± 0.04
	kcat (s−1)	241 ± 20	248 ± 11	62.4 ± 2.2	54.2 ± 3.2
	kcat/Km (s−1 mM−1)	241	2.76 × 104	0.309	339
ScCR1A60T	Km (mM)	0.58 ± 0.01	0.042 ± 0.009	61.8 ± 1.8	0.27 ± 0.02
	kcat (s−1)	19.6 ± 0.3	22 ± 1.5	89.8 ± 4.1	108 ± 3
	kcat/Km (s−1 mM−1)	33.8	524	1.45	400

^aValues are means and standard deviations.

Temperature optima and thermostabilities of the variants.

The temperature dependence of ScCR1 and its variants was determined by measuring the enzyme activity at 20 to 60°C, as shown in Fig. 2. The optimum reaction temperature for ScCR1_I158V was found to be 35°C, which was 5°C lower than that of the wild-type enzyme. However, both ScCR1_I158V/P168S and ScCR1_{A60T/I158V/P168S} showed an optimum reaction temperature of 55°C, which was 15°C and 20°C higher than those of the wild-type enzyme and the variant ScCR1_I158V, respectively. T₅₀¹⁵, which allows the intuitive comparison of enzymes with varied thermostabilities, was investigated for ScCR1 and its variants (Table 2). ScCR1_I158V displayed a T₅₀¹⁵ of 44.8°C, which was 2.6°C lower than that of the wild-type enzyme, while the T₅₀¹⁵ of ScCR1_{A60T/I158V/P168S} was determined to be 53.6°C, which was 6.2°C and 8.8°C higher than those of the wild-type enzyme and ScCR1_I158V, respectively. These results are in good agreement with the optimum reaction temperatures. To gain more insight concerning the improved thermostability, the melting temperatures (T_m) of ScCR1 and ScCR1_{A60T/I158V/P168S} were also investigated. In accordance with the results for T₅₀¹⁵, the T_m of ScCR1_{A60T/I158V/P168S} was 7.4°C higher than that of the wild-type enzyme. Additionally, the thermostability of the variant ScCR1_{A60T/I158V/P168S} was also studied at temperatures of 30, 40, and 50°C to further confirm its increased thermostability (see Fig. S7 in the supplemental material), giving half-lives (t_1/2) of 315, 131, and 4.33 h at 30, 40, and 50°C, respectively (Table 4). All these results demonstrated that the thermostability of ScCR1 was successfully improved by protein engineering.

FIG 2.

Effects of temperature on the activities of ScCR1 (●) and its variants ScCR1_I158V (▲), ScCR1_I158V/P168S (◆), and ScCR1_{A60T/I158V/P168S} (■).

TABLE 4.

Half-lives of ScCR1 and ScCR1_{A60T/I158V/P168S} at 30, 40, and 50°C

T (oC)	ScCR1A60T/I158V/P168S		ScCR1a
	kD (h−1)	t1/2 (h)	kD (h−1)	t1/2 (h)
30	0.0022	315	0.0041	169
40	0.0053	131	0.0086	80.6
50	0.16	4.33	0.55	1.30

^aData reported by Wang et al. (16).

Substrate tolerance of the variants against COBE.

Like T₅₀¹⁵, C₅₀¹⁵ was measured to investigate the substrate tolerance of ScCR1 and its variants against COBE, the results are shown in Fig. S8 in the supplemental material and in Table 5. The C₅₀¹⁵ of ScCR1_I158V was as low as 6.01 mM, which was about 6-fold lower than that of the wild-type enzyme, while the substrate tolerances of ScCR1_I158V/P168S and ScCR1_{A60T/I158V/P168S} against COBE were improved significantly, giving C₅₀¹⁵ values of 122 mM and 162 mM, respectively, which were approximately 4-fold and 5-fold higher than that of the wild-type enzyme. These results were consistent with those of thermostability investigation.

TABLE 5.

Substrate tolerance of ScCR1 and its variants against COBE

Enzyme	C5015 (mM)a
ScCR1	34.0 ± 0.2
ScCR1I158V	6.01 ± 0.08
ScCR1I158V/P168S	122 ± 4
ScCR1A60T/I158V/P168S	162 ± 1

^aValues are means and standard deviations.

Asymmetric reduction of COBE.

To evaluate whether the higher catalytic activity and better stability of the variant are superior to those of the wild-type enzyme for synthetic application, asymmetric reduction of COBE was performed comparatively using both ScCR1 and the best variant, ScCR1_{A60T/I158V/P168S}. Two parallel reactions were run at the same time, one with an equal weight of biocatalyst (Fig. 3A) and the other with an equal total activity of biocatalyst (Fig. 3B). As shown in Fig. 3A, when an equal weight of biocatalyst was employed, the reaction rate of ScCR1_{A60T/I158V/P168S} was obviously higher than that of the wild-type enzyme, achieving 99.9% conversion within 2 h, while the conversion of the wild-type enzyme was only 81%, indicating that the higher catalytic activity of the variant did significantly accelerate the reaction. As shown in Fig. 3B, even if the biocatalyst loading of the variant was only 25% of that of the wild-type enzyme (based on equal total enzyme activity), the reaction rate mediated by the variant was comparable to that for the wild-type enzyme. Most importantly, the formation of an emulsion could be avoided during downstream processing. Furthermore, the introduction of mutations into ScCR1 did not change the stereoselectivity of the enzyme, and the product (S)-CHBE was obtained with an ee of >99.9%.

FIG 3.

(A) Time course of COBE bioreduction catalyzed by lyophilized ScCR1 (○) or ScCR1_{A60T/I158V/P168S} (●) with the same amount of enzyme. (B) Time course of COBE bioreduction catalyzed by lyophilized ScCR1 (□) or ScCR1_{A60T/I158V/P168S} (■) with equal total enzyme activity.

The preparative synthesis of (S)-CHBE was also performed on a 300-ml scale using ScCR1_{A60T/I158V/P168S} as a biocatalyst to evaluate its feasibility for practical application. As shown in Fig. 4, 30 g (100 g/liter) COBE could be completely transformed within 9 h by only 2 g/liter biocatalyst loading, affording 28.5 g (S)-CHBE with an excellent ee value of >99.9% (see Fig. S9 in the supplemental material) and a space-time yield of 255 g liter⁻¹ day⁻¹, which was 6-fold higher than the value reported by Pan et al. (17).

FIG 4.

Biosynthesis of (S)-CHBE with the variant ScCR1_{A60T/I158V/P168S} on a preparative scale.

DISCUSSION

A novel carbonyl reductase (ScCR1) from Streptomyces coelicolor has been reported for synthesis of (S)-CHBE from COBE using 2-propanol as a cosubstrate (16, 17); however, the low activity and poor stability of ScCR1 inspired us to improve its catalytic activity by protein engineering. To increase the positive hits of ScCR1 variants with improved catalytic activity while maintaining the screening effort at a minimum, we focused on the amino acids surrounding the active pocket because they were considered to have great effects on the activity of an enzyme (26–28). Since the oxidized nicotinamide ring was considered to be contacting 2-propanol, great effort was made to solve the crystal structures of ScCR1 complexed with NAD⁺ and 2-propanol and also of ScCR1 complexed with COBE, NAD⁺, and 2-propanol; however, we failed to obtain qualified crystal structures. As an alternative, the crystal structure of ScCR1 complexed with NADH and 2-propanol was successfully solved and refined to 2.3 Å (Table 1; see Fig. S2 in the supplemental material), and we believed that the structure was reliable enough to guide our protein engineering, since the binding modes of NADH and NAD⁺ should be very similar because the major parts of these two molecules are identical. This was also confirmed by the high number of positive hits in our screening results (2 positive mutants from a theoretical 160 variants based on 8 positions).

Since COBE was selected as the target substrate for the screening process, the k_cat/K_m values of ScCR1 and its variants toward COBE were significantly higher than those for 2-propanol, but this was mainly due to the huge K_m value (two or three orders of magnitude higher) of ScCR1 toward 2-propanol compared with COBE, since their k_cat values were comparable (Table 3). In the asymmetric reduction of COBE, the substrate concentrations (COBE and 2-propanol) were 10 to 100 times higher than their K_m values, although the reaction was performed in a toluene-water biphasic reaction system, in which large fraction of the hydrophobic substrate was retained in the organic phase. However, the concentration of COBE employed in the bioconversion was 600 mM, and the partition coefficient of COBE in the toluene-water biphasic system was 21.3 (16); therefore, the actual concentration of COBE in the aqueous phase was still 27 mM, more than 20-fold higher than the K_m values of ScCR1 and its variants toward COBE, while 2-propanol is water soluble in nature, and its concentration in the aqueous phase would be at least 10-fold higher than the K_m value. Under these conditions, the reaction rate determinant was k_cat. Considering the comparable k_cat values of ScCR1 and some of its variants (the I158V, P168S, and A60T variants) toward COBE and 2-propanol, the NADH regeneration rate should be high enough for the bioreduction of COBE. For the other two mutants (the I158V/P168S and A60T/I158V/P168S mutants), although significantly higher reaction rates for the bioreduction of COBE compared with those of the wild-type enzyme were observed (Fig. 3A), the possibility that the rate-limiting step might be the regeneration of NADH could not be excluded, since their k_cat values toward 2-propanol were lower than those for COBE.

Homology modeling and docking analysis based on the crystal structure of ScCR1 was then performed to shed light on the origin of improved activity and stability. The location of each mutation in the structure of ScCR1 is shown in Fig. 5. Both Ile158 and Pro168 are located at the entrance of the active site, while Ala60 is located on the surface of the enzyme. In the case of I158V, the smaller side chain of valine rather than that of isoleucine would reduce the steric hindrance of the active site entrance, thereby allowing better accommodation and transportation of substrate/product into and out of the active site (Fig. 6A and B). This is consistent with the fact that the specific activity and catalytic constant (k_cat) of the I158V variant are both higher than those of the wild type (Tables 2 and 3). Pro168 connects a loop and an α-helix, and replacement of the more rigid proline with serine increases the flexibility of the loop, which is beneficial for the transportation of substrate/product into and out of the active site, thereby resulting in better catalytic activity of the P168S mutant (Fig. 6C and D). This observation is also in good accordance with the activity and kinetic constant studies (Tables 2 and 3). Besides better catalytic activity, P168S also increases the T₅₀¹⁵ by 1.7°C (Table 2), which might be due to additional hydrogen bonds formed between serine and other amino acids nearby or water molecules, since Pro168 also is located in proximity to the surface of the enzyme (Fig. 5). For mutation A60T, threonine forms two additional hydrogen bonds with Glu57 and Arg64, which would strengthen the rigidity of the protein structure, thereby improving the stability of the enzyme (Fig. 6E and F). This is also consistent with the 3.1°C enhancement of the T₅₀¹⁵ of the A60T mutant compared to the wild-type enzyme (Table 2). The protein engineering progress for ScCR1 is shown in Fig. 7.

FIG 5.

Locations of the activity- and thermostability-related amino acid residues identified in this study.

FIG 6.

Molecular docking of the substrate COBE (orange sticks) into ScCR1 and its variants. The catalytic triad and NADH are shown as green and yellow sticks, respectively. (A and B) Local environments near the isoleucine (A) to valine (B) substitutions in ScCR1; (C and D) detailed view of local environments near the proline (C) to serine (D) substitutions in ScCR1; (E and F) hydrogen bonds formed by the alanine (E) to threonine (F) substitutions in ScCR1.

FIG 7.

Protein engineering route for ScCR1.

In practical application, ScCR1_{A60T/I158V/P168S} significantly outperformed the wild-type enzyme, with a much higher reaction rate or much less biocatalyst loading (Fig. 3A and B). It should be noted that less biocatalyst loading of the variant ScCR1_{A60T/I158V/P168S} (25% of the wild-type) would reduce the cost of the biocatalyst up to 75%, which is very attractive for industrial application since a great challenge for industrial applications using biocatalysts is the relatively high cost of enzymes. Furthermore, in the preparative scale synthesis of (S)-CHBE from COBE by the engineered variant ScCR1_{A60T/I158V/P168S}, the process parameters of the biocatalytic reaction are comparable or even superior to those required for industrial application (Table 6) (29). Although the current process parameters (e.g., product titer) might still be lower than those reported by Ma et al. (2) and Kizaki et al. (30), it is noteworthy that the cofactor (NAD⁺) used in this study is less expensive than those in previous processes, in which more expensive NADP⁺ was used. Another advantage in our process is that only a single biocatalyst (ScCR1) is enough for the bioreduction of COBE and regeneration of NADH through oxidation of 2-propanol, which is very simple for process industrialization, while in the case of previous studies, another enzyme (glucose dehydrogenase [GDH]) was required to regenerate the cofactor NADPH. Careful pH control was also required due to excess gluconic acid production from glucose oxidation, and the resultant gluconate salt in the reaction medium would complicate the downstream process and also represents a difficult environmental issue. As a result, our process might be a promising alternative for the production of (S)-CHBE.

TABLE 6.

Process parameters of variant ScCR1_{A60T/I158V/P168S} in comparison with the industrial requirements

Parameter	Required value	Value from this work
Substrate-to-enzyme ratio (kg/kg)	≥50	50
[Cofactor] (g/liter)	<0.5	0.1 (0.1 mM)
[Substrate] (g/liter)	≥100	100
Conversion (%)	>95 (≤24 h)	>99 (≤9 h)
(Optical) purity of product (% ee)	>99.5	>99.9

In summary, the activity, thermostability, and substrate tolerance of ScCR1 were significantly improved by protein engineering through a structure-guided approach and directed evolution. Docking analysis based on the crystal structure of ScCR1 shed light on the origin of improved activity and stability. The resultant variant shows great potential for practical synthesis of the atorvastatin precursor (S)-CHBE.

MATERIALS AND METHODS

Chemicals and materials.

Ethyl-4-chloro-3-oxo-butanoate was kindly provided by Nantong Chengxin Amino Acid Co., Ltd. (Jiangsu, China). All other chemicals were obtained commercially and used without further purification. PrimerSTAR HS and rTaq polymerase, restriction enzymes (DpnI, NdeI, and HindIII), and T4 DNA ligase were purchased from TaKaRa Biotechnology Co., Ltd. (Dalian, China). The expression vector pET28a was obtained from Shanghai Bioleaf Biotech Co., Ltd. (Shanghai, China), and the host strain E. coli BL21(DE3) was purchased from Tiangen (Beijing, China). Primers were synthesized by Generay Biotech Co., Ltd. (Shanghai, China).

Library construction.

Saturation mutagenesis was applied to generate a mutant library of ScCR1 on the amino acids (Ala156, Ile158, Leu159, Gly160, Phe164, Ser167, Pro168, and Val171) around the substrate within 8 Å, with the exception of catalytic residues (Ser157, Tyr170, and Lys174) and conserved amino acids in classical short-chain dehydrogenase. Saturation mutagenesis was performed using PrimerSTAR HS. The PCR mixture contained 80 ng of plasmid pET28a-ScCR1, 1.25 U PrimerSTAR HS, and a 0.2 μM concentration of each of the two degenerate primers (Table 7) in a total volume of 50 μl. The reaction mixture was preincubated at 94°C for 2 min, preheated at 98°C for 10 s, annealed at 55°C for 30 s, and elongated at 68°C for 5.5 min. The PCR product was digested with 20 U DpnI at 37°C for 1 h. Plasmids containing the mutated gene were transformed into E. coli BL21(DE3) host cells and then plated on an LB agar plate with 50 μg/ml kanamycin.

TABLE 7.

Primers used for saturation mutagenesis and site-directed mutagenesis

Primer	Oligonucleotide sequence (5′→3′)a
A156-F	TCGATCGTGAACGTCNNKTCCATCCTCGGC
A156-R	GCCGAGGATGGAMNNGACGTTCACGATCGA
I158-F	ACGTCGCCTCCNNKCTCGGCTCGGTCGGCT
I158-R	AGCCGACCGAGCCGAGMNNGGAGGCGACGT
L159-F	AACGTCGCCTCCATNNKAGGCTCGGTCGGCTTC
L159-R	GAAGCCGACCGAGCCTMNNATGGAGGCGACGTT
G160-F	TCGCCTCCATCCTCNNKTCGGTCGGCTTCG
G160-R	CGAAGCCGACCGAMNNGAGGATGGAGGCGA
S167-F	CTTCGCCGGCNNKCCCGCCTACGT
S167-R	ACGTAGGCGGGMNNGCCGGCGAAG
P168-F	TTCGCCGGCTCCNNKGCCTACGTCGCCGCCA
P168-R	TGGCGGCGACGTAGGCMNNGGAGCCGGCGAA
V171-F	CCCCGCCTACNNKGCCGCCAAGCACGGCG
V171-R	CGCCGTGCTTGGCGGCMNNGTAGGCGGGG
S135L-F	GGCGTCTTCTACTTAATGCGCTACGAACTG
S135L-R	CAGTTCGTAGCGCATTAAGTAGAAGACGCC
E139 M-F	ACTCGATGCGCTACATGCTGCCCGCCATCGA
E139 M-R	TCGATGGCGGGCAGCATGTAGCGCATCGAGT
A184F-F	GCTGACGAAGGCGTTCGCCGCCGAGTACG
A184F-R	CGTACTCGGCGGCGAACGCCTTCGTCAGC
A22V-F	TTCGCCGGCCGTACCGTACTCGTCACCGGTGC
A22V-R	GCACCGGTGACGAGTACGGTACGGCCGGCGAA
A60T-F	AGGGCGCCGAGAAGGCCACAGCCGAGCTGCGGGCCGGT
A60T-R	ACCGGCCCGCAGCTCGGCTGTGGCCTTCTCGGCGCCCT
S167W-F	TTCGCCGGCTGGTCCGCCTACGT
S167W-R	ACGTAGGCGGACCAGCCGGCGAA
P168S-F	TTCGCCGGCTCCTCAGCCTACGTCGCCGCCA
P168S-R	TGGCGGCGACGTAGGCTGAGGAGCCGGCGAA

^aUnderlining indicates the codon used for mutagenesis.

Error-prone PCR was carried out using the plasmid pET-28a-ScCR1_I158V/P168S as the template and 5′-GGAATTCCATATGACTGTCGAAACCGCCACC-3′ and 5′-CGCGGATCCCTAGACAGAACAGTAACCACCT-3′ as forward and reverse primers, respectively. For error-prone PCR, 100 μM Mn²⁺ was selected for the desired mutagenesis rate (1 to 3 mutation sites per gene based on sequence analysis of 100 samples). The PCR product was digested with NdeI and HindIII and ligated into the corresponding sites of pET-28a. The recombinant plasmid was then transformed into competent E. coli BL21(DE3) cells.

Site-directed mutagenesis was performed similarly, with primers listed in Table 7. The mutation was confirmed by DNA sequencing by Sunny Biotechnology Co., Ltd. (Shanghai, China).

Library screening.

Library colonies were separately inoculated into 200 μl LB medium containing 50 μg/ml kanamycin in 96-deep-well plates and cultured at 37°C for 10 h before inoculating 600 μl LB medium in new 96-deep-well plates. The plates were incubated at 37°C for 3 h, and protein expression was induced with 0.1 mM IPTG (isopropyl-β-d-thiogalactopyranoside) at 16°C for another 24 h. The cells were harvested by centrifugation (3,420 × g, 10 min) and lysed by freeze-thawing using 200 μl lysis buffer containing 0.75 mg/ml lysozyme and 0.01 mg/ml DNase I at 37°C for 1.5 h. The supernatant was obtained by centrifugation at 3,420 × g for 20 min at 4°C. High-throughput screening was performed by recording the decrease in the absorbance of NADH at 340 nm for a defined period (3 min 40 s) at 30°C.

To screen the variants with enhanced activity toward COBE, the supernatant was subjected to activity assay as follows. Fifty microliters of supernatant with proper dilution was added to a microtiter plate, and a 150-μl mixture containing 50 mM phosphate buffer (pH 6.5), 0.2 mM NADH, and 2 mM COBE was then added to the reaction system. With a microtiter spectrophotometer (PowerWave XS2; BioTek Instruments Co., Ltd.), the activity of each variant was determined. Variants showing higher catalytic activity than the wild type were chosen for rescreening in 96-deep-well plates, and then the top variants were selected for further confirmation in shake flasks.

For thermostability screening, 100 μl supernatant with proper dilution was transferred to another 96-well plate, heated at 37°C for 30 min, and then cooled down on ice before activity assay. The following reaction system was used: 10 μl supernatant after heat treatment and a 190-μl mixture containing 50 mM phosphate buffer (pH 6.5), 0.2 mM NADH, and 2 mM COBE were then added to a microtiter plate for activity assay. Similarly, the residual activity of every variant was measured by the high-throughput screening method. Variants with improved thermostability were screened and selected for rescreening in a 96-deep-well plate and confirmed in shake flasks.

Protein expression and purification.

The cultivation of recombinant E. coli cells expressing ScCR1 and its variants was performed as described previously (16). Cells were harvested by centrifugation (6,000 × g) at 4°C for 10 min, resuspended in buffer A (20 mM sodium phosphate, 500 mM NaCl, 10 mM imidazole, pH 8.0), and disrupted with an ultrasonic oscillator (JY92-II; Scientz Biotech. Co., Ltd.). The cell debris was then removed by centrifugation (30,000 × g) at 4°C for 1 h. The supernatant was loaded onto a His trap Ni-nitrilotriacetic acid (Ni-NTA) fast-flow (FF) column (1 ml) preequilibrated with buffer A and eluted with buffer A with an increasing gradient of imidazole from 20 mM to 500 mM at a flow rate of 1 ml/min. The fractions containing target protein as verified by SDS-PAGE were combined and dialyzed, and the purified enzyme was then concentrated to around 20 mg · ml⁻¹. Next, typically 10 mg concentrated protein was applied to a Superdex 200 size exclusion column equilibrated with 25 mM Tris-HCl (pH 7.5) containing 150 mM NaCl and 1 mM dithiothreitol (DTT). The eluted fractions of ScCR1 were pooled, concentrated to 20 to 25 mg · ml⁻¹, flash frozen in aliquots of 110 μl in liquid nitrogen, and stored at −80°C. Before crystallization experiments, the sample was centrifuged for 15 min at 15,000 × g.

Crystallization.

Crystals were grown at 4°C using the sitting-drop vapor diffusion method with a reservoir solution containing 20% polyethylene glycol (PEG) 8000, 0.1 M cacodylate sodium (pH 6.5), and 0.2 M magnesium acetate [Mg(OAc)₂]. The protein sample (1 μl) and precipitant (1 μl) were mixed, and crystals from these drops appeared within 1 day and developed to full size within 3 days. Optimal crystals were obtained after initial rounds of optimization using the same batch aliquots. Dimethyl sulfoxide (DMSO) (0.2 μl, 30%) was chosen as an additive after screening with an additive kit (Hampton Research). Complexed crystals of ScCR1 were obtained under the same conditions with addition of 5 mM NADH in protein solution and then soaked in a reservoir solution with 5% of 2-propanol before freezing with liquid nitrogen.

Data collection and processing.

X-ray diffraction data sets with resolution of 2.30 Å were collected using CuKα generated by a Rigaku Micromax007 HF with an R-AXIS IV++ detector system for the complexed crystals of ScCR1. The diffraction data were indexed and processed with HKL2000 (31) and the CCP4 program suite (32). The complexed crystals belonged to space group P3₁ with cell parameters a = 188 Å, b = 188 Å, and c = 81 Å.

Structure determination and refinements.

The initial phase of the ScCR1 structure was determined by molecular replacement with the program Balbes in the CCP4 suite and using the crystal structure of levodione reductase (PDB code 1IY8) as the search model. The model was built in Coot (33), and model refinement was performed in Phenix (34).

Enzyme assay.

The activities of ScCR1 and its variants toward COBE were measured spectrophotometrically at 30°C by recording the decrease in the absorbance of NADH at 340 nm within 2 min. The reaction mixture (1 ml) contained 50 mM phosphate buffer (pH 6.5), 2.0 mM COBE, 0.1 mM NADH, and an appropriate amount of enzyme. One unit of enzyme activity was defined as the amount of enzyme catalyzing the oxidation of 1 μmol NADH per min under the standard assay conditions.

Kinetic constants.

The kinetic constants of ScCR1 and its variants were investigated by measuring the activity under different substrate (COBE) concentrations ranging from 0.15 mM to 4 mM at a fixed NADH concentration (0.2 mM). The apparent K_m value for NADH was measured by assaying the activity with NADH concentrations varied from 0.02 mM to 4 mM in the presence of 5 mM COBE. Similarly, to determine the kinetic constants for 2-propanol, the concentrations of 2-propanol were varied from 0.05 M to 1.5 M at a fixed NAD⁺ concentration (5 mM), and the apparent K_m value for NAD⁺ was measured with varied NAD⁺ concentrations (0.2 to 4 mM) and a fixed 2-propanol concentration (1.5 M). The K_m and V_max values were calculated from Lineweaver-Burk plots.

Temperature optima and thermostability.

The optimum temperatures for ScCR1 and its variants were determined under the standard assay conditions at various temperatures from 20°C to 70°C with 5°C increments. T₅₀¹⁵, which is an indicator of thermal stability, was defined as the temperature at which 50% of initial enzyme activity is lost following a heat treatment for 15 min (22). To determine the T₅₀¹⁵ of ScCR1 and its variants, 100 μl (1 mg/ml) purified enzyme was distributed into 8 PCR tubes, and subjected to incubation at a gradient of temperatures (40 to 60°C) for 15 min with a thermocycler (Bio-Rad), and then cooled down on ice before activity assay. The residual activities of ScCR1 and its variants were assayed under the standard assay conditions. The melting temperature (T_m) was measured with circular dichroism (CD) spectroscopy as described previously (35). The enzyme unfolding curves were measured from 200 nm to 280 nm against temperatures ranging from 20°C to 90°C with a gradient of 1°C min⁻¹ (36).

Thermostability was determined by assaying the residual activity of enzyme after incubation at the desired temperatures (30, 40, and 50°C separately) for the required period. All the enzyme assays were performed in triplicate.

Substrate tolerance of the variants against COBE.

Like T₅₀¹⁵, C₅₀¹⁵ was introduced as an indicator of substrate tolerance and defined as the concentration at which 50% of initial enzyme activity is lost following incubation for 15 min. To determine the C₅₀¹⁵ of ScCR1 and its variants, 100 μl (1 mg/ml) purified enzyme was distributed into 8 PCR tubes, and subjected to incubation at a gradient of substrate concentrations (10 to 200 mM) for 15 min with a vortex mixer, and then the residual activity was measured. The residual activities of ScCR1 and its variants were assayed under the standard assay conditions after COBE was diluted to 2 mM.

Asymmetric reduction of COBE by ScCR1 and ScCR1_{A60T/I158V/P168S}.

The reaction using a mixture (10 ml) containing 5 ml sodium phosphate buffer (50 mM, pH 6.5), 5 ml toluene, 1.0 g (6 mmol) COBE, 1.0 μmol NAD⁺, 20 μmol MgCl₂, 2-propanol (9 mmol, 1.5 equivalent to COBE), and 40 mg lyophilized cell extract was performed at 30°C with magnetic stirring. In addition, the reaction was also carried out with equal activity (1.6 kU) of ScCR1 and ScCR1_{A60T/I158V/P168S}. Samples were taken periodically to determine the conversion.

For bioreduction of COBE on a preparative scale, the reaction using a mixture (300 ml) containing 150 ml sodium phosphate buffer (50 mM, pH 6.5), 150 ml toluene, COBE (30 g, 0.18 mol), 2-propanol (20 ml, 0.26 mol), 0.03 mmol NAD⁺, 0.6 mmol MgCl₂, and 0.6 g lyophilized ScCR1_{A60T/I158V/P168S} cell extract was performed in a bioreactor (Reactor-Ready, 500 ml) with two bladed turbine impellers driven by a speed-variable motor at 30°C and 250 rpm. The temperature of the reactor contents was maintained by a thermostat water bath. After 9 h, the reaction mixture was extracted twice with ethyl acetate (150 ml). The organic layer was dried over anhydrous sodium sulfate and then evaporated with a rotary evaporator. The conversion and enantiomeric purity for CHBE during the reaction process were determined by chiral gas chromatography (GC). GC analysis was performed as described previously (16). The isolated product was validated by ¹H nuclear magnetic resonance (NMR) spectra (see Fig. S1 in the supplemental material).

Docking and computational methods.

Molecular docking was performed to analyze the interactions of ScCR1 or its variants with the ligand COBE (37). The FoldX algorithm (version 3.0 beta5.1, available at http://foldx.crg.es/) was employed to estimate protein stability by introducing the other 19 amino acids to each residue of ScCR1 and calculating the free energy of each mutant (25). The energy of the structure of ScCR1 was first minimized with the “RepairPDB” program, and then the “BuildModel” program was applied to evaluate the effect of every substitution. The relative change in folding free energy (ΔΔG) based on point mutation was predicted using the FoldX software. The predicted ΔΔG equals the ΔG for the protein carrying the point mutation minus the ΔG for the wild-type protein (38). The potentially stabilizing mutations were selected based on the predicted ΔΔG and verified experimentally.

Accession number(s).

The coordinates of the ScCR1/NADH/2-propanol complexed structure have been deposited in the Protein Data Bank with the accession code 5H5X.