Coenzyme A-free activity, crystal structure, and rational engineering of a promiscuous β-ketoacyl thiolase from Ralstonia eutropha

Abstract

Thiolases catalyze the formation of carbon-carbon bonds in diverse biosynthetic pathways. The promiscuous β-ketoacyl thiolase B of Ralstonia eutropha (ReBktB) has been utilized in the in vivo conversion of Coenzyme A (CoA)-linked precursors such as acetyl-CoA and glycolyl-CoA into β-hydroxy acids, including the pharmaceutically-important 3,4-dihydroxybutyric acid. Such thiolases could serve as powerful carbon-carbon bond-forming biocatalysts in vitro if handles less costly than CoA were employable. Here, thiolase activity is demonstrated toward substrates linked to the readily-available CoA mimic, N-acetylcysteamine (NAC). ReBktB was observed to catalyze the retro-Claisen condensation of several β-ketoacyl-S-NAC substrates, with a preference for 3-oxopentanoyl-S-NAC over 3-oxobutanoyl-, 3-oxohexanoyl-, and 3-oxoheptanoyl-S-NAC. A 2.0 Å-resolution crystal structure, in which the asymmetric unit consists of four ReBktB tetramers, provides insight into acyl group specificity and how it may be engineered. By replacing an active site methionine with an alanine, a mutant possessing significant activity towards α-methyl substituted, NAC-linked substrates was engineered. The ability of ReBktB and its engineered mutants to utilize NAC-linked substrates will facilitate the in vitro biocatalytic synthesis of diketide chiral building blocks from feedstock molecules such as acetate and propionate.

Graphical Abstract

1. Introduction

Biosynthetic thiolases catalyze carbon-carbon bond formation in many pathways. The first step in the production of isoprenoids, ketone bodies, and polyhydroxyalkanoate (PHA) polymers is the thiolase-mediated condensation of two molecules of acetyl-CoA into acetoacetyl-CoA [1,2]. Such carbon-carbon bond-forming reactions are of biocatalytic interest since they could enable the generation of complex organic compounds from feedstock chemicals.

The bacterium Ralstonia eutropha utilizes β-ketoacyl thiolases such as RePhbA (also known as RePhaA) and ReBktB in the biosynthesis of PHA polymers that store both carbon and energy (Figure 1a) [3-8]. While RePhbA is specific for acetyl-CoA and helps generate 3-hydroxybutyryl monomers in the biosynthesis of poly-(R)-3-hydroxybutanoate, ReBktB can condense acetyl-CoA with propionyl-CoA or butyryl-CoA to generate 3-hydroxypentanoyl or 3-hydroxyvaleroyl monomers in the biosynthesis of PHA copolymers. Recently, the promiscuous substrate specificity of ReBktB was harnessed by incorporating it within a pathway engineered into E. coli to yield chiral building blocks such as the pharmaceutically-important 3,4-dihydroxybutyric acid (Figure 1b). In these studies, ReBktB was observed to condense acetyl-CoA with acetyl-, propionyl-, butyryl-, isobutyryl-, and glycolyl-CoA [9,10]. Fascinatingly, the chiral building block 2,3-dihydroxybutyric acid was also generated by the engineered pathway, indicating that ReBktB produced the α-substituted intermediate 2-hydroxy-3-oxobutryl-CoA from acetyl-CoA and glycolyl-CoA.

Figure 1.

Natural and engineered pathways utilizing biosynthetic thiolases. (a) Polyhydroxyalkanoate (PHA) synthesis. After the biosynthetic thiolase PhbA or BktB condenses two acyl-CoA's into a β-ketoacyl-CoA, PhbB can generate a R-β-hydroxy butyryl monomer for PhbC-mediated addition to a growing PHA polymer. (b) In vivo pathway engineered to yield β-hydroxy acids. After condensation and reduction by BktB and PhbB, respectively, the thioesterase TesB releases β-hydroxy acid chiral building blocks. (c) Proposed in vitro pathway. Modular polyketide synthase ketoreductases (PKS KRs) coupled with an NADPH-regeneration system could both drive the production of α-substituted, β-ketoacyl-SNAC's by BktB as well as set the stereochemistry of α- and β-substituents of generated α-substituted, β-hydroxyacyl chiral building blocks [31].

A biosynthetic thiolase fuses a priming acyl-CoA and an extending acyl-CoA into a β-ketoacyl-CoA through a Claisen-like condensation that proceeds through a two-step ping-pong mechanism [1,11,12]. Under standard conditions, the reverse reaction is thermodynamically favored; however, in biological contexts flux is usually in the forward direction owing to the removal of product by downstream enzymes in the biosynthetic pathway. Thus, for thiolases to be employed in the in vitro synthesis of chiral building blocks, they need to be coupled with energetically-favorable reactions (Figure 1c). Another obstacle in employing thiolases in vitro is the expense of the CoA handle. Binary complexes of CoA-bound biosynthetic thiolases, such as the thiolase from Zoogloea ramigera (referred to here as ZrPhbA, PDB 1DLV), show interactions with the CoA adenosine and pantetheinyl moieties that may accelerate catalysis [13]. To our knowledge, the importance of these interactions for thiolase-mediated catalysis has not been determined.

Here, we demonstrate the activity of ReBktB toward β-ketoacyl substrates linked to the readily available N-acetylcysteamine (NAC) handle that lacks the phosphoadenosine moiety and most of the pantetheinyl arm of CoA. In the degradative direction, ReBktB is shown to prefer 3-oxopentanoyl-S-NAC over 3-oxobutanoyl-S-NAC and is also active towards 3-oxohexanoyl-S-NAC and 3-oxoheptanoyl-SNAC. A 2.01-Å resolution crystal structure of ReBktB helps to provide a structural basis for the broad substrate specificity of the thiolase. Through rational engineering, a mutant was generated with significant activity towards α-methyl substituted substrates.

2. Results

2.1. In vitro thiolysis of β-ketoacyl-S-NAC substrates

ReBkB (1 μM) was incubated with 5 mM NAC and 1 mM β-ketoacyl-S-NAC substrate: 3-oxobutanoyl-S-NAC (1), 3-oxopentanoyl-S-NAC (2), 3-oxohexanoyl-S-NAC (3), 3-oxoheptanoyl-S-NAC (4), or (2RS)-2-methyl-3-oxopentanoyl-S-NAC (5) (Figure 2). After 1 h and 24 h at 22 °C, reactions were quenched with acid and analyzed by HPLC. The α-branched 5 was found to be an unsuitable substrate for the enzyme even after a 24-h incubation (Figure S1). However, after 1 h reactions containing 1, 2, 3, or 4 yielded acetyl-S-NAC (6) as well as another molecule of 6, propionyl-S-NAC (7), butyryl-S-NAC (8), or valeryl-S-NAC (9), respectively (products showed the same retention times and absorbance spectra as synthetic standards, and their identities were corroborated by LC/MS: m/z = 162.2, 176.2, 190.2, and 204.4 for 6-9); after 24 h, the β-ketoacyl-S-NAC substrates had completely converted to acyl-S-NAC products. Peaks for product 6 were integrated and their areas were compared to a standard curve to determine the concentration of 6 formed (Figure S2). ReBktB exhibited preference for the β-ketoacyl-SNAC thioesters in the order 2 > 1 > 3 ~ 4 (0.873 ± 0.031 mM, 0.238 ± 0.007 mM, 0.089 ± 0.006 mM, and 0.082 ± 0.005 mM 6 formed, respectively). This differs from the previously determined specificity of ReBktB toward β-ketoacyl-CoA thioesters (3-oxopentanoyl-CoA > 3-oxohexanoyl-CoA > 3-oxobutanoyl-CoA) [5]. Increasing the concentration of ReBktB to 10 μM resulted in the complete thiolysis of the 7-carbon NAC thioester 4 (the least favorable substrate tested) into 6 and 9 within 1 h (the chromatogram resembles that of the 24 h reaction with 1 μM ReBktB; Figure S1). To our knowledge, utilization of β-ketoacyl-S-NAC substrates by a β-ketoacyl thiolase has not been reported.

Figure 2.

In vitro thiolytic cleavage of β-ketoacyl-S-NAC thioesters by ReBktB. (a) Reactions assayed with ReBktB. R-groups of β-ketoacyl-S-NAC substrates and acyl-S-NAC products are indicated. ReBktB did not catalyze the thiolysis of compound 5. (b) Chromatograms (235 nm) of reactions containing 1 (left panel) and 2, 3, or 4 (right panel; overlay) (1 μM ReBktB was incubated with 1 mM β-ketoacyl-S-NAC and 5 mM NAC for 1 h at 22 °C, then acid-quenched and analyzed by C₁₈ reverse-phase HPLC). A shallower gradient was employed for reactions containing 1 to obtain better resolution. (c) ReBktB preference: 2 > 1 > 3 ~ 4. The concentration of product 6 was determined by integrating the area of peak 6, subtracting that of corresponding –ReBktB controls, and accounting for dilution factor (0.873 ± 0.031 mM for 2, 0.238 ± 0.007 mM for 1, 0.089 ± 0.006 mM for 3, and 0.082 ± 0.005 mM for 4). For reactions containing 1, peak areas for product 6 were divided in half. Error bars indicate standard deviation of triplicate measurements.

2.2. ReBktB crystallography

ReBktB crystallized in space group P1 with 16 monomers per asymmetric unit (Figure 3a and Table 1). Molecular replacement was performed with Phaser by searching for eight copies of Clostridium difficile thiolase dimer (PDB 4DD5) [14]. Twinning was suspected after residual factors stalled during refinement despite the high quality of the electron density map [15]. The dataset was processed in a primitive triclinic unit cell and submitted to phenix.xtriage of the PHENIX software bundle [16], which estimated the twin fraction α > 0.41 (where α = 0.5 for perfectly twinned and α = 0 for untwinned) and identified three pseudo-merohedral twin operators: (1) h, -k, -l; (2) -h, k, -l; and (3) -h, -k, l (Table S1). The software also performed an L Test [17], calculating 〈|L|〉 = 0.421 and 〈L²〉 = 0.248 for acentric reflections, which deviate significantly from expected values for normal, untwinned data (Figure S3). Twin refinement in REFMAC5 [18] of the CCP4 software suite [19] drastically improved the residual factors after a single cycle (Table S2). In the final model, R_work = 17.6% and R_free = 20.8%. The only region between residues 2 and 393 that shows weak electron density is loop residues 207-210. The presented crystal structure is higher quality (resolution: 2.01 vs. 2.29 Å, B-factors: 42.2 vs. 64.6 Ų, R_work/R_free: 17.6%/20.8% vs. 25.3%/31.6%, completeness: 92.9% vs. 85.9%) than a recently-reported crystal structure of ReBktB [20].

Figure 3.

Crystal structure of ReBktB. (a) The asymmetric unit of the ReBktB crystal contains four tetramers (distinguished by color). The P1 unit cell is displayed as a wire box. (b) Stereoview of superposed ReBktB (green) and CoA-bound ZrPhbA (cyan, PDB 1DLV). C_α r.m.s.d. = 0.64 Å over 339 residues.

Table 1.

Crystallographic data collection and refinement statistics for ReBktB

Data collection
Space group	P1
Resolution (Å)	50-2.00 (2.03-2.00)
Unit cell parameters
a, b, c (Å)	72.05, 105.99, 201.14
α, β, γ (°)	89.97, 89.98, 89.94
Rmerge (%)	5.6 (44.4)
Average I/σ(I)	10.4 (2.0)
No. of observed reflections	644,834
No. of unique reflections	369,679 (18,535)
Redundancy	1.7 (1.7)
Refinement
Resolution (Å)	41.8-2.01 (2.06-2.01)
Completeness (%)	92.9 (78.7)
Rwork/Rfree (%)	17.6/20.8 (23.2/28.3)
No. of atoms
Overall	46,064
Protein	45,388
Water	676
Average B-factors (Å2)
Overall	42.2
Protein	39.1
Water	29.3
Root-mean-square deviation
Bond lengths (Å)	0.005
Bond angles (°)	0.68
Ramachandran plot (%)
Favored	96.8
Allowed	2.2
Outliers	1.0
PDB ID	4W61

Values for the highest-resolution shell are given in parentheses.

The tertiary fold of ReBktB is highly similar to that of ZrPhbA, a biosynthetic thiolase from Z. ramigera (PDBs 1DLV and 1DM3, 51% sequence identity) [13,21], with C_α r.m.s.d. = 0.64 Å over 339 residues (Figures 3b and S4). The enzyme adopts the five-layered αβαβα fold of the thiolase superfamily in which two β-sheets are sandwiched between three clusters of α-helices [22] (Figures 4 and S5-S7). Like other biosynthetic thiolases, ReBktB is a tetramer in which residues 120-143 from each monomer contribute to an eight-stranded, antiparallel β-barrel. Soaking of crystals with CoA and acetyl-CoA was attempted; however, samples rapidly deteriorated in their presence. Crystal structures were superposed using PyMol to compare thiolase active sites [23] (Figure 5).

Figure 4.

Active site elements of ReBktB. (a) Condensation and thiolysis reactions (clockwise and counterclockwise, respectively). Select atoms are shown from substrates, products, C90, H350, C380, and G382. Binding pockets for CoA, R₁, and R₂ are drawn as red, green, and blue semicircles, respectively. (b) Stereoview of superposed active sites of ReBktB (light and dark green chains) and acetyl-CoA-bound acetyl-ZrPhbA (cyan and blue chains; PDB 1DM3). ReBktB possesses unique features in helices α3 (kinked at Y66) and α5 (narrower) that expand the R₁ binding pocket relative to ZrPhbA. The loop following α5 has been hidden for clarity (indicated with black circles). (c) Alignment of thiolase sequences corresponding to ReBktB helices α3 and α5 (residues 62-75 and 144-149, ReBktB numbering). PhbA-type biosynthetic thiolases, but not ReBktB, share strong conservation in these regions, while degradative thiolases possess different architectures in these regions. Biosynthetic thiolases: R. eutropha BktB (NCBI YP_725948); R. eutropha PhbA (P14611); Z. ramigera PhbA (P07097); Pseudomonas nitroreducens (WP_017518929); Lamprocystis purpurea (WP_020506577); Thiothrix nivea (WP_002708073); C. difficile (Q18AR0); Mus musculus cytosolic (NP_033364); H. sapiens mitochondrial (P24752); and S. cerevisiae cytosolic (XXBYAC). Degradative thiolases: S. typhimurium (P0A2H7); Yersinia pestis (A4TR28); Achromobacter xylosoxidans (WP_020928315); Leptonema illini (WP_002774022); A. thaliana peroxisomal (Q56WD9); Glycine max (NP_001237076); Macaca mulatta peroxisomal (AFH32148); S. cerevisiae peroxisomal (P27796); Trypanosoma brucei (Q57XD5); and Aspergillus niger (EHA20372). Sequences were aligned with Clustal Omega [21].

Figure 5.

Superposition of thiolase active sites validates differing substrate specificities. (a) Superposed stereoview of ZrPhbA (cyan carbons, with acetyl-CoA and enzyme-bound acetyl carbons in white; PDB 1DM3), RePhbA (orange carbons; PDB 4O9A), and ReBktB (green carbons). (b) Superposed stereoview of ZrPhbA and the sixteen ReBktB monomers from the unit cell. L89 and M158 adopt more conformations than neighboring residues. (c) Model of acetyl-CoA-bound, valeryl-ReBktB intermediate. The valeryl group appears to fit within a hydrophobic pocket with proper geometry and no serious steric clashes.

2.3. Rational engineering

The M290A mutant catalyzed the retro-Claisen condensation of (2RS)-2-methyl-3-oxopentanoyl-S-NAC (5), albeit complete conversion in 24 h required 25 μM ReBktB (Figure 6). Only trace activity was noted for wild-type and the M290L mutants in the same conditions. The M290A and M290L mutants showed greatly decreased activity towards 3-oxopentanoyl-S-NAC (2) compared to wild-type ReBktB (Figure S8).

Figure 6.

The rationally-engineered M290A mutant is active toward an α-substituted substrate. (a) Wild-type ReBktB, (b) the M190A mutant, and (c) the M190L mutant incubated with (2RS)-2-methyl-3-oxopentanoyl-S-NAC (5) (25 μM enzyme; 0, 1, and 24 h timepoints). Significant thiolysis to propionyl-SNAC (7) is observed for the M190A mutant.

3. Discussion

ReBktB has shown biocatalytic potential in vivo, aiding in the production of valuable chiral building blocks [10]. In order to realize the biocatalytic potential of ReBktB in vitro, alter its substrate specificity, and enhance its catalytic properties, interactions between potential substrates and regions of ReBktB need to be more fully characterized - especially the CoA-binding region, the acyl group binding pocket for the priming acyl-CoA, and the acyl group binding pocket for the extending acyl-CoA (Figure 4a).

That ReBktB can catalyze retro-Claisen reactions on NAC-linked substrates reveals that the residues contacting the adenosine portion of CoA are not essential for catalysis, even though they likely enhance catalytic efficiency (Figure 5b). By analogy with the CoA-bound structures of ZrPhbA (PDBs 1DLV and 1QFL) [12,13], when an acyl-CoA is bound by ReBktB its adenine ring is sandwiched between R222 and L232 and its ribose 2′-OH forms a charged hydrogen bond with the R222 guanidinium. Since the CoA pantetheinyl arm of the priming and extending acyl-CoA substrates are differentially positioned during catalysis, the CoA tunnel accommodates at least two conformations of the pantetheinyl moiety. That the CoA binding tunnel does not tightly bind the pantetheinyl moiety is also indicated by crystal structures of CoA-bound thiolases (PDBs 1DLV, 1DM3, 1QFL, 1M10, and 4O9C) [12,13,24,25]. Thus, the catalytic efficiency of ReBktB towards NAC-linked substrates may suffer most in comparison with CoA-linked substrates due to the relative loss of interactions with the CoA adenosine moiety.

By analogy with ZrPhbA, when the priming acyl-CoA is positioned for the transthioesterification reaction the thioester carbonyl forms hydrogen bonds with C90 NH and G382 NH (Figure 4b). If an acyl group bound to ReBktB points in the same direction as the acetyl group in the crystal structure of acetylated ZrPhbA (PDB 1DM3) [13], it would be located in a hydrophobic pocket created, in part, by Y66′ (the prime indicates a residue from the adjacent monomer) and α5. The equivalent pocket in ZrPhbA, partially constructed by Q64′ and α5, is smaller. The side chain of the conserved loop residue Q64′ in ZrPhbA is closer to the catalytic residues compared to the side chain of helix-embedded Y66′ of ReBktB (Figure 4c). Both residues form hydrogen bonds with the C-terminal end of α5; however, because α5 is narrower in ReBktB (being one residue shorter) it creates more space for the acyl group of the priming acyl-CoA.

A recent crystal structure reveals the active site architecture of RePhbA (PDB 4O9C) [25] to be nearly identical to that of ZrPhbA (they share 63% sequence identity, while ReBktB and RePhbA only share 52% sequence identity) [21], with the conserved glutamine and α5 in equivalent positions (Figure 5a). The small pocket they create may explain why both of these biosynthetic thiolases are specific for acetyl-CoA as a priming acyl-CoA. Degradative thiolases, which often perform thiolysis on long β-ketoacyl substrates, do not contain the glutamine or α5 (e.g., PDBs 1AFW, 2WU9, and 3GOA, from the Saccharomyces cerevisiae peroxisome, the Arabidopsis thaliana peroxisome, and Salmonella typhimurium) [26,27] (Figure 4c). The region where the glutamine and α5 would be located in degradative thiolases is hypothesized to bind fatty acyl chains. In crystal structures of degradative thiolases from S. cerevisiae and A. thaliana (PDBs 1AFW and 2WU9) [26,27], 2-methyl-2,4-pentanediol and ethylene glycol are observed in this region, and in the crystal structure of a Mycobacterium tuberculosis degradative thiolase a steroid substrate mimic can be observed (PDB 4UBT) [28].

We sought to better understand interactions between known ReBktB substrates and the ReBktB active site. The presented 2.01-Å resolution crystal structure provides a unique opportunity to visualize the conformations available to active site residues since the asymmetric unit possesses 16 crystallographically-distinct monomers. The side chains of L89 and M158 show more conformational freedom compared to the side chains of the catalytic residues C90, H350, and C380, which are observed in the same orientation within each monomer (Figure 5b). Since ReBktB can accept acetyl-, propionyl-, glycolyl-, butyryl-, isobutyrl-, and valeryl-CoA as priming substrates, the orientations that these acyl groups adopt within the active site were investigated. After placing the first two carbons of a valeryl group as observed for the acetyl group bound to ZrPhbA and rotating the C_α-C_β torsion angle ~40° as in several acyl-enzyme complexes of the thiolase superfamily (PDBs 1EK4, 1M4T, 1TQY, 4NA2, 4NA3) [23, 29-31], the next three carbons were modeled such that no clashes were made either with side chains or a bound acetyl-CoA (Figure 5c). As modeled, the valeryl group makes hydrophobic contact with the side chains of V57, L89, A148, M158, and I352. Shorter acyl chains may bind similarly; the glycolyl hydroxyl group would be in position to form a hydrogen bond with the A148 carbonyl.

The superposition of biosynthetic thiolases with ReBktB shows that the active site region where the extending acetyl-CoA is bound during carbon-carbon bond-formation is equivalent in all three enzymes (Figure 5a). A methionine (M290 in ReBktB) is positioned to make hydrophobic contact with the acetyl methyl group, as observed in the crystal structure of the acetylated ZrPhbA in complex with acetyl-CoA (PDB 1DM3) [13]. That M290 was observed in the same conformation in each of the 16 crystallographically-distinct ReBktB monomers may indicate the importance of this residue in naturally selecting against extending acyl-CoA's larger than acetyl-CoA (Figure 4b). Interestingly, ReBktB does not exclude glycolyl-CoA as an extending acyl-CoA but does exclude propionyl-CoA [10].

In an attempt to engineer ReBktB to generate α-methyl substituted products, M290A and M290L mutants were expressed, purified, and assayed (Figure 6). At a concentration of 25 μM, M290A catalyzes the complete thiolysis of (2RS)-2-methyl-3-oxopentanoyl-S-NAC (5) in 24 h. At the same concentration, wild-type ReBktB and the M290L mutant only show trace activity. Both the M290A and M290L mutants show greatly diminished activity toward 3-oxopentanoyl-S-NAC (2) compared to the wild-type enzyme (Figure S8). These results indicate that further rational engineering to tune the substrate specificity of ReBktB is possible.

4. Conclusion

The presented structural and functional analysis will be useful in engineering desired biocatalytic properties into ReBktB through rational or directed evolution approaches. Mutating V57, L89, A148, M158, and I352 to smaller residues may increase the range of priming acyl thioesters accepted by ReBktB and allow the incorporation of synthetically-useful functional groups. As suggested by the M290A mutant reported here, substitution of M290 by smaller residues may enable the condensation of extending acyl thioester substrates other than acetyl and glycolyl thioesters to generate α-substituted, β-ketoacyl thioesters. Such products could be reduced in a stereocontrolled manner (e.g., by modular polyketide synthase ketoreductases) to yield libraries of two-stereocenter chiral building blocks (Figure 1c) [32]. That ReBktB is active towards NAC-linked substrates enables its usage in practical in vitro biocatalytic schemes, which would be attractive since such reactions contain fewer components and are more controllable than those in vivo. Engineering greater extending acyl thioester promiscuity within ReBktB could result in mixtures of reaction products; however, valuable small molecule products generated from such in vitro biocatalytic schemes could be chromatographically resolved or selectively utilized by downstream enzymes. In the current advent of biosynthetic biocatalysis, the studies reported here will aid in engineering and utilizing ReBktB to generate valuable molecules from abundant precursors.

5. Materials and methods

5.1. Cloning, expression, and purification of ReBktB

The ReBktB gene was amplified from the pCDF/pct/tesB plasmid [10] with primers 5′-ATGCTTGCAcatatgACGCGTGAAGTGGTAGTGGTA-3′ and 5′-GTACGAACGgaattcTCAGATACGCTCGAAGATGGCGGC-3′ (restriction sites in lower case; stop codon underlined), digested with NdeI and EcoRI, and ligated into pET28b (Novagen). E. coli BL21(DE3) cells carrying this plasmid were grown in LB media with 50 mg/L kanamycin at 37 °C. Upon reaching OD₆₀₀ = 0.4, cultures were cooled to 15 °C and induced with 0.5 mM IPTG. After 17 h, cells were pelleted, resuspended in lysis buffer (500 mM NaCl, 30 mM HEPES pH 7.5, 10% (v/v) glycerol) and sonicated. Cell lysate was centrifuged at 30,000 × g for 1 h, and the supernatant was poured over a column of Ni-NTA beads (Expedeon) equilibrated with lysis buffer. Beads were washed with 10 column volumes of 15 mM imidazole in lysis buffer and protein was eluted with 1.5 column volumes of 150 mM imidazole in lysis buffer. Eluate was further purified on a Superdex 200 gel filtration column (GE Healthcare Life Sciences) equilibrated with 150 mM NaCl, 10 mM HEPES pH 7.5, 10% (v/v) glycerol. Column fractions were exchanged in 25 mM NaCl, 10 mM HEPES pH 7.5, 10% (v/v) glycerol and brought to 37 mg/mL in a centrifugal concentrator. Aliquots were flash frozen in liquid nitrogen and stored at −80 °C until further use. The isolated yield of ReBktB was ~10 mg/L culture.

Mutations of ReBktB were generated by PCR amplification of the construct described above using primers 5′-GTGGACCCGAAGGCCGCCGGCATCGGCCCGGTG-3′ and 5′-CACCGGGCCGATGCCCGGGGCCTTCGGGTCCAC-3′ for M290A, and primers 5′-CGTGGACCCGAAGGCCCTGGGCATCGGCC-3′ and 5′-GGCCTTCGGGTCCACGCCGGCAT-3′ for M290L. The resulting products were digested with DpnI to remove the template plasmid and subsequently transformed into E. coli BL21(DE3) cells. Expression and purification of the ReBktB mutants followed that of the wild-type enzyme.

5.2. Syntheses of NAC and β-ketoacyl-S-NAC substrates

Syntheses of NAC, 1, 2, 3, and 5 were carried out as described [32]. Syntheses of 4 and 6 were carried out as described [33].

5.3. In vitro thiolytic cleavage of β-ketoacyl-S-NAC substrates

Enzyme aliquots were rapidly thawed and placed on ice. ReBktB (1 μM final) was added to 100 mM NaCl, 100 mM HEPES pH 7.5, 10% (v/v) glycerol, 5 mM NAC, and 1 mM 1, 2, 3, 4, or 5 in a total volume of 1 mL in triplicate. Negative controls lacking either ReBktB or NAC were also prepared. After incubating for 1 or 24 h at 22 °C, quenching was effected by mixing 190 μL reaction solution with 10 μL 7 M HCl. Negative controls lacking both ReBktB and HCl were also prepared to ensure that quenching did not promote acid-catalyzed hydrolysis of thioester substrates. Samples were centrifuged and their supernatants were analyzed at 235 nm by reverse-phase HPLC on a Varian Microsorb-MV 300-5 C18 250 × 4.6 mm column (Agilent Technologies; 50-μL injection loop) and Waters 2998 PDA detector. Mobile phases consisted of 0.1% TFA in water (solution A) and 0.1% TFA in methanol (solution B). Reactions containing 1 were injected (20 μL) on a 30-min linear gradient of 0-20% solution B with a flow rate of 1 mL/min. Reactions containing 2, 3, 4, or 5 were injected (20 μL) on a 30-min linear gradient of 5-50% solution B with a flow rate of 1 mL/min. Peaks for product 6, monitored at 235 nm, eluted at 17.3 min for reactions containing 1 or 10.2 min for reactions containing 2, 3, or 4 (Figures 2 and S1). Peaks for product 6 were integrated in Breeze 2 software (Waters) and their areas were compared to those of a standard curve of 0.1-1 mM 6 to determine concentrations (Figure S2). For reactions containing 5, peaks for product 7, which eluted at 15.3 min, were monitored at 235 nm and integrated. Background from reaction solutions lacking enzyme was subtracted from peak areas, and calculated concentrations were adjusted for dilution factor. For reactions containing 1, peak areas for product 6 were divided by 2 for direct comparison with other reactions. Product fractions were collected and further analyzed by positive-ESI LC/MS (Agilent Technologies 1200 Series HPLC with a Gemini C18 5 micron 2.1 × 50 mm column coupled to an Agilent Technologies 6130 Series Quadrupole MS). Mobile phases consisted of 0.1% formic acid in water (solution C) and 0.1% formic acid in acetonitrile (solution D). Samples were run on a 12-min linear gradient of 5-95% solution D with a flow rate of 1 mL/min.

To compare the activities of ReBktB, the M290A mutant, and the M290L mutant toward 5, each enzyme was added at a concentration of 25 μM to 100 mM NaCl, 100 mM HEPES pH 7.5, 10% (v/v) glycerol, 5 mM NAC, and 1 mM 5 in a total volume of 0.5 mL. To compare the activities of ReBktB, the M290A mutant, and the M290L mutant toward 2, each enzyme was added at a concentration of 1 μM to 100 mM NaCl, 100 mM HEPES pH 7.5, 10% (v/v) glycerol, 5 mM NAC, and 1 mM 2 in a total volume of 1 mL. In both assays samples were analyzed as above at 0, 1, and 24 h.

5.4. Crystallization, data collection, processing, phasing, and refinement

Enzyme aliquots were rapidly thawed and placed on ice. Crystals were grown by sitting-drop vapor diffusion by adding 1 μL protein solution to 0.5 μL 22% (w/v) PEG 4000, 100 mM HEPES pH 7.0, 12% (v/v) glycerol at 22 °C. Crystals were harvested from the mother liquor and immediately flash frozen in liquid nitrogen without addition of cryoprotectants; they dissolved unless removed from the mother liquor within 48 h of drop preparation.

A 2.0-Å dataset was collected from a single crystal at 100 K at Advanced Light Source beamline 5.0.3 (λ = 0.9765 Å) and indexed, integrated, and scaled in HKL2000 [34] (Table 1). Intensities were converted to amplitudes, and 5% of scaled reflections were reserved for calculating R_free in TRUNCATE [35] of the CCP4 software suite [19]. Phases were solved via molecular replacement by searching for eight copies of a Clostridium difficile thiolase (PDB 4DD5) dimer in space group P1 in Phaser [14] of CCP4. To convert side chains to those of the R. eutropha enzyme, a SCWRL model [36] obtained from the Fold & Function Assignment System (FFAS) server [37] was superposed onto each monomer of the Phaser solution. Pseudo-merohedral twinning and associated operators were identified (Table S2) in phenix.xtriage of the PHENIX software bundle [16]. Iterative refinement was performed in Coot [38] and REFMAC5 [18] of CCP4 (with intensity-based twin refinement, distance and angle restraint weights of 4.0, and automatically generated local NCS restraints). Throughout refinement, the MolProbity server was consulted for structure validation [39]. Structure factor amplitudes and atomic coordinates for ReBktB were deposited in the Protein Data Bank with PDB ID 4W61.

Highlights.

The thiolase ReBktB shows activity on N-acetylcysteamine bound substrates.

ReBktB shows a preference for the five-carbon β-ketoacyl substrate.

The 2.0-Å resolution crystal structure of ReBktB was solved with 16 monomers/ASU.

The structure helped to analyze handle and acyl group promiscuity.

A mutant active towards α-methyl substituted substrates was rationally engineered.