A ketoreductase domain in the PksJ protein of the bacillaene assembly line carries out both α- and β-ketone reduction during chain growth

Abstract

The polyketide signaling metabolites bacillaene and dihydrobacillaene are biosynthesized in Bacillus subtilis on an enzymatic assembly line with both nonribosomal peptide synthetase (NRPS) and polyketide synthase (PKS) modules acting along with catalytic domains servicing the assembly line in trans. These signaling metabolites possess the unusual starter unit α-hydroxyisocaproate (α-HIC). We show here that it arises from initial activation of α-ketoisocaproate (α-KIC) by the first adenylation domain of PksJ (a hybrid PKS/NRPS) and installation on the pantetheinyl arm of the adjacent thiolation (T) domain. The α-KIC unit is elongated to α-KIC-Gly by the second NRPS module in PksJ as demonstrated by mass spectrometric analysis. The third module of PksJ uses PKS logic and contains an embedded ketoreductase (KR) domain along with two adjacent T domains. We show that this KR domain reduces canonical 3-ketobutyryl chains but also the α-keto group of α-KIC-containing intermediates on the PksJ T-domain doublet. This KR activity accounts for the α-HIC moiety found in the dihydrobacillaene/bacillaene pair and represents an example of an assembly-line dual-function α- and β-KR acting on disparate positions of a growing chain intermediate.

Polyketide (PK) and nonribosomal peptide (NRP) natural products span a wide cross-section of bioactive molecules, including human therapeutics and signaling agents (1, 2). The hybrid PK/NRPs dihydrobacillaene and bacillaene are biosignaling agents from Bacillus subtilis where each can be described as a diamide of ω-amino polyenoic acids that is N-capped with an α-hydroxyisocaproate (α-HIC) moiety (3). The metabolite initially arising from their shared assembly line is dihydrobacillaene, which is later converted to bacillaene by the cytochrome P450 PksS, which oxidizes the saturated C14′–C15′ bond of dihydrobacillaene to an olefin (3–5). Previous work has shown the dihydrobacillaene assembly line to be a rich source of unusual biochemistry, where hybrid PK synthase (PKS)/NRP synthetase assembly-line proteins are used in conjunction with enzyme machinery borrowed from both isoprenoid and polyunsaturated fatty acid biosynthesis (6). Little precedent exists for α-hydroxyacyl units in PK/NRPs, and for this reason we chose to probe the origins of the unusual α-hydroxyacyl N-cap of dihydrobacillaene.

The dihydrobacillaene biosynthetic machinery has been partially characterized and is split among five large megasynthases (PksJLMNR) and several accessory proteins (Fig. 1) (3, 7). Like other modular megasynthases for PK/NRP hybrids, catalytic domains are arranged into modules with each module containing thiolation (T) domains that act as covalent way stations for PK/NRP chain-elongation intermediates. The T domains are posttranslationally modified with phosphopantetheinyl (Ppant) units to afford the mobile thiol arm upon which chain growth occurs. For dihydrobacillaene, a trans-acyltransferase (AT) loads T domains with malonyl units. Ketosynthase (KS) domains catalyze Claisen-like decarboxylative condensations of assembly-line-tethered malonyl units to yield β-ketothioesters. Further processing of β-ketothioesters occurs, in part, by reductive loop domains such as β-ketoreductases (β-KRs) and dehydratases (DHs) found immediately adjacent to the acylated T domains. The three NRPS modules found in the dihydrobacillaene assembly line include an adenylation (A) domain for adenylating and loading of amino acid units and a condensation (C) domain that catalyzes peptide bond formation.

Fig. 1.

Biosynthesis of dihydrobacillaene and bacillaene. (A) Gene cluster responsible for dihydrobacillaene/bacillaene biosynthesis. (B) Domain organization of PksJ with predicted intermediates shown. The module three and module four intermediates were observed by Moldenhauer et al. (5); the module one and module two intermediates are predicted based on the structure of dihydrobacillaene. The cytochrome P450, PksS, oxidizes the C14′-C15′ bond of dihydrobacillaene after biosynthesis to generate bacillaene. DH, DH domain; D, acyltransferase docking domain.

At the beginning of the first dihydrobacillaene assembly line enzyme, PksJ, are two NRPS modules, an initiating (A₁-T₁) and an elongating (C₁-A₂-T₂) module, which are followed by two PKS modules. Working in concert, the two NRPS modules would appear to condense the α-HIC N-cap with glycine while tethered to the dihydrobacillaene assembly line. Previous work on dihydrobacillaene biosynthesis revealed that by the time the growing molecule reaches the third module of PksJ it already possesses the α-hydroxyl, demonstrating either (i) that α-HIC is directly loaded onto the bacillaene assembly line, or (ii) that an unknown α-HIC precursor is converted to the hydroxy acid while it is tethered to a T domain in one of the first three modules (Fig. 1B). Unlike β-hydroxyacyl units, α-hydroxyacyl units are relatively rare in PKs and nonribosomal peptides. Among the few characterized examples is the fungal secondary metabolite enniatin, in which the α-hydroxyacid is directly adenylated for incorporation into the natural product (8). Alternatively, in the cereulide depsipeptide assembly-line machinery, α-ketoacids are activated and reduced with chiral specificity by an A domain-embedded KR to the corresponding α-hydroxyacids while tethered to a T domain on the assembly line (9). No such module organization (A-KR-T) is found in the dihydrobacillaene assembly line.

During the course of exploring the origins of the α-HIC unit in dihydrobacillaene/bacillaene, we identified a unique dual-function KR domain that acts as a canonical β-KR and a noncanonical α-KR to convert the remote chain-initiating α-ketoamide to an α-hydroxyamide. We also identified the stereochemical course of both of these reactions.

Results

The First Two Modules of PksJ Load and Transfer α-KIC.

Our initial experiments set out to identify the substrate of the first A domain (A₁) in PksJ. The 80-kDa PksJ initiation module (A₁-T₁) was heterologously expressed in Escherichia coli, and test substrates included isocaproate derivatives differing in the oxidation state of the α-carbon: α-ketoisocaproate (α-KIC), (R)- and (S)-HIC, isocaproic acid (IC), and (S)-leucine (α-aminoisocaproic acid). Based on substrate activation measured indirectly through ATP-PP_i exchange α-KIC was revealed as the preferred substrate of the PksJ A₁ domain (Fig. 2A) with a K_m of 403 nM. In the depsipeptide α-ketoacid-activating A domains, a neutral residue replaces the aspartate that interacts with the positively charged substrate amine in α-amino acid-activating A domains (9). This mutation is also observed in the α-KIC-activating PksJ A₁ domain, wherein a valine (V240) replaces the conserved aspartate, further supporting the preference for α-KIC as a substrate.

Fig. 2.

Origin of the bacillaene α-hydroxyacid. (A) Adenylation catalyzed by PksJ(A₁-T₁), as assayed by ATP/pyrophosphate exchange. (B) The α-KIC-Gly dipeptide is formed exclusively at T₂ when holo-PksJ(A₁-T₁-C₁-A₂-T₂) is incubated with α-KIC, IC, α-HIC, and Gly in the presence of ATP, as detected at the intact peptide (Upper) and Ppant ejection ion (Lower) levels by FTMS. No Ppant ejection products are observed other than from the α-KIC-Gly dipeptide. N/D, ion not detected.

Next, we attempted to observe covalent loading of α-KIC onto phosphopantetheinylated (holo) PksJ(A₁-T₁). Mixing holo-PksJ(A₁-T₁) with α-KIC and ATP and subsequent analysis by Fourier transform mass spectrometry (FTMS) did not reveal any stable loading of the α-KIC in our hands, perhaps because the presumed α-ketothioester generated on PksJ(A₁-T₁) is hydrolytically unstable (data not shown). To confirm that α-KIC is in fact activated and loaded onto PksJ(A₁-T₁), we examined loading on a 197-kDa construct corresponding to the first two modules of PksJ [PksJ(A₁-T₁-C₁-A₂-T₂)]. We speculated that the condensation domain would catalyze transfer of the T₁-loaded α-KIC to the glycine tethered to T₂, generating the more stable α-ketoamide at a rate sufficiently high relative to α-ketothioester hydrolysis so as to allow observation of the α-KIC-Gly dipeptide by MS. The PksJ(A₁-T₁-C₁-A₂-T₂) construct was expressed and purified from E. coli and phosphopantetheinylated by the Ppant transferase, Sfp (10). When we incubated bis-holo-PksJ(A₁-T₁-C₁-A₂-T₂) with Gly and α-KIC in the presence of adenylation substrates (±)-HIC and IC, we observed exclusive formation of the T₂-tethered α-KIC-Gly dipeptide by FTMS analysis and the Ppant ejection assay (11), confirming loading α-KIC onto the assembly line (Fig. 2B).

PksJ-KR Reduces T Domain-Tethered β-Ketothioesters.

Based on the above results, α-KIC is loaded onto the PksJ assembly line as a precursor to the dihydrobacillaene α-HIC N-cap and is transported to the second NRPS module as an α-ketoacyl unit. In contrast to α-keto acid depsipeptide modules it lacks an in cis fused α-KR (9), implying that the KR functions are supplied in trans or by a KR-containing module elsewhere in the assembly line. Surveying the unassigned ORFs in the bacillaene gene cluster failed to identify any candidate dehydrogenase enzymes.

The above observations combined with work by Moldenhauer et al. (5) suggest reduction occurs while an α-KIC-containing intermediate is tethered to PksJ module 3 (KS-DH-KR-T₃-T₄). The jump from module two to three is a NRPS to PKS switchover, with PKS module three possessing features that demand attention for understanding the α-ketoreduction scaffold. First, module three has a doublet of T domains (T₃-T₄), a noncanonical domain architecture perhaps indicative of noncanonical domain function. Second, the tandem T domains are in juxtaposition with a KR domain. Based on inspection of the dihydrobacillaene structure this first PksJ KR domain (PksJ-KR) can be inferred to act as a β-KR to generate a cryptic C15′ hydroxyl that is subsequently dehydrated by the first PksJ DH domain and saturated by an in trans-enoyl reductase to yield the C14′-C15′ alkane (unpublished work). However, the position of the tandem T domains in module three led to the hypothesis that PksJ-KR could work twice on two different keto groups: the β-keto group and the α-keto group within the starter unit, converting α-KIC to α-HIC.

To test this hypothesis, the KR-T₃-T₄ region of PksJ was heterologously overexpressed (12). Tests of the KR domain catalytic action as a canonical β-KR used the model β-ketoacyl substrate bis-acetoacetyl (Acac)-S-PksJ(KR-T₃-T₄) generated by incubation of apo-PksJ(KR-T₃-T₄) with Acac-CoA (CoA) and Sfp (10). Although we were able to distinguish substrates linked to each of the two T domains by protease digestion, we also generated mutants of PksJ(KR-T₃-T₄) in which the phosphopantetheinylated serine residue in each T domain was replaced by alanine. In both the WT and mutant constructs, we observed rapid formation of active-site peptide and Ppant ejection ions with a mass shift of +2.0166 Da in an NAD(P)H-dependent fashion, consistent with ketoreduction and formation of β-hydroxybutyryl-S-PksJ(KR-T₃-T₄) (Fig. 3). Both NADH and NADPH served as hydride donors with approximately equal efficacy, and we were unable to detect any kinetic differentiation between the two T domains as β-reduction scaffolds. These results demonstrate clearly that the PksJ-KR domain is active as a canonical β-KR.

Fig. 3.

Analysis of PksJ ketoreduction by FTMS. (A) The KR domain in PksJ(KR-T₃-T₄) can perform β-ketoreduction. (Upper) Active-site peptides and Ppant ejection ions observed for Acac-S-T₃ and Acac-S -T₄ in the absence of NAD(P)H. (Lower) Upon the addition of NAD(P)H, a mass increase of 2.0166 Da is observed in both the peptides and Ppant ejection ions. (B) The KR domain of PksJ(KR-T₃-T₄) can perform α-ketoreduction. (Upper) Active-site peptides and Ppant ejection ions observed for α-KIC-GABA-S-T₃ and α-KIC-GABA-S-T₄ in the absence of NAD(P)H. (Lower) Upon the addition of NAD(P)H, a mass increase of 2.0154 Da is observed in the peptides and Ppant ejection ions. Lines denote the most abundant peak in the peptide isotopic distribution (solid: absence of NAD(P)H; dashed: presence of NAD(P)H). (Insets) Structures of theoretical Ppant ejection ions.

PksJ-KR Reduces α-KIC to α-HIC.

We next tested the ability of Pks-KR to reduce the remote α-KIC amide. We tested α-ketoisocaproyl-γ-aminobutyrate (α-KIC-GABA) as the PksJ α-KR substrate. According to canonical PKS logic, the reductive loop domain sequence in PksJ module three (DH-KR) would yield an α-KIC-γ-aminocrotonyl KR substrate. However, we have shown that this substrate is reduced to α-KIC-GABA in trans while it is tethered to this module (unpublished work), and Moldenhauer et al. (5) observed in vivo that the α-ketoreduced product is indeed the α-HIC-GABA-S-T thioester.

To access α-KIC-GABA-S-PksJ(KR-T₃-T₄), α-KIC-GABA-CoA was synthesized and used as a donor in the Sfp-catalyzed phosphopantetheinylation reaction see supporting information (SI) Text. The acyl-holo WT and mutant PksJ(KR-T₃-T₄) constructs were incubated with NAD(P)H, and a 2.0154-Da mass shift (calculated from the observed Ppant ejection ions) indicated that an α-ketoreduction had occurred (Fig. 3B). As observed in β-reduction, NADH and NADPH were interchangeable as hydride donors, and reduction occurred on both T domain scaffolds with approximately equal efficiency. Of note, the apparent rate of α-reduction was ≈10-fold less than the apparent rate of β-reduction (data not shown).

Reduction Chirality.

Finally, we sought to determine the stereochemistry of the reductions catalyzed by the dual-function PksJ-KR domain. Again, using acetoacetyl-S-T thioesters as model β-ketoacyl substrates in the PksJ(KR-T₃-T₄) construct, we repeated the β-ketoreduction assay, followed by TycF-catalyzed cleavage of the resulting β-hydroxybutyrate from the T-domain scaffold (13) and derivatization with an amine fluorophore. The stereochemistry of the cryptic C15′ β-hydroxyl was identified by chiral HPLC to be S (Fig. 4A).**

Fig. 4.

Determination of ketoreduction stereochemistry by chiral HPLC. (A) Acac-S-PksJ(KR-T₃-T₄) was subjected to β-ketoreduction, and the resulting 3-hydroxybutyryl (3HB) unit was cleaved from PksJ(KR-T₃-T₄) by action the thioesterase TycF. The 3HB was amidated in situ with AMC, and the 3HB-AMC amide was subjected to chiral chromatography to identify the stereochemical course of reduction. (B) α-KIC-GABA-S-PksJ(KR-T₃-T₄) was subjected to α-reduction and treated as in A. (C) The observed stereochemical courses of β- and α-ketoreduction can be rationalized by a model in which the substrates are held in similar extended conformations and the hydride is delivered from the same face (top as drawn).

We performed a similar set of experiments on the authentic α-KIC-GABA-S-T thioester α-ketoreduction substrate, revealing that PksJ-catalyzed α-ketoreduction produces (S)-α-HIC-GABA (Fig. 4B). Based on these observations and the apparent lack of an epimerase that could act on the C2″ center after reduction, we propose that the previously unknown chirality of the dihydrobacillaene C2″ is S.

Discussion

The B. subtilis secondary metabolites bacillaene and dihydrobacillaene both begin with a relatively rare α-hydroxyacyl N-capping (3). Two precedents for incorporation of α-hydroxyacid residues into NRP scaffolds are relevant. In the enniatin synthetase the hydroxyacid (α-hydroxyisovaleric acid) is selected by the A domain and then loaded as the S-pantetheinyl thioester on the adjacent T domain (8). In the case of the potassium ionophores valinomycin and cereulide the modules that insert and elongate α-hydroxyacids have been shown by us (9) to select and directly install the α-ketoacids. α-KR domains embedded within the A domains then reduce the tethered ketoacid to the α-hydroxyacyl moiety before chain elongation. Such α-ketoacid-activating A domains typically lack the active-site aspartate that normally ion pairs with the α-H₃N⁺ of α-amino acids and is a key selection determinant for amino acid substrates (14, 15).

The initial adenylation domain of PksJ indeed activates and installs α-KIC on the holo form of the T₁ domain, suggesting an “on assembly-line” reduction step occurs. There is no α-KR domain embedded within NRPS module one or NRPS module two of PksJ, nor is there a discernible separate KR protein encoded within the bacillaene gene cluster. We validated that the α-KIC moiety is transferred unreduced to module two by showing that α-KIC-Gly-S-T₂ forms in a purified PksJ(A₁-T₁-C₁-A₂-T₂) construct by mass spectral interrogation of the acylated holo-peptide derived proteolytically from T₂.

Moldenhauer et al. (5) created a mutant dihydrobacillaene-producing strain that allowed detection of chain-elongation intermediates, including α-HIC-GABA-S-pantetheinyl thioesters tethered to the third module. Importantly, this showed that the chain-elongation intermediate that transits module three of PksJ has an α-HIC rather than α-KIC N-cap, establishing that α-KIC reduction occurs on one of the first three modules in PksJ. In this work, we identify the intermediates on modules one and two to be α-KIC and α-KIC-Gly (5). Module three of PksJ has the domains KS-DH-KR-T₃-T₄. The doublet of active T domains, T₃-T₄, in module three is atypical for modular assembly lines and served as a clue that unusual biosynthetic reactions occur on module three.

To assay the KR domain of module three of PksJ we expressed the three-domain fragment PksJ(KR-T₃-T₄) in apo form. This methodology also allowed us to load PksJ(KR-T₃-T₄) with acetoacetyl units from Acac-CoA and generate a model 3-ketoacyl-S-pantetheinyl-T₃ and-T₄ form of the protein. Using FTMS we were able to detect the acyl-S-pantetheinyl-peptides from both T₃ and T₄.and show PksJ-KR behaves as a canonical β-KR.

To assess whether the PksJ-KR domain also had α-KR activity we prepared the chain-elongation intermediate we anticipated is presented to the third module of PksJ in vivo. As in the assay of β-KR function, there was an NAD(P)H-dependent increase of 2 mass units on the peptide fragment from T₃ and T₄ validating that the only keto group on the α-KIC-GABA chain, the α-ketoamide moiety, was reduced to the (S)-α-hydroxyamide (Fig. 5). Therefore PksJ-KR can reduce both the ketone in the canonical β-ketoacyl-S-pantetheinyl arm of an adjacent T domain and the ketone in the α-ketoisocaproyl amide moiety of the growing dihydrobacillaene chain.

Fig. 5.

Role of dual-function PksJ-KR in dihydrobacillaene biosynthesis. The dual-function KR processes a bis-ketone thioester, reducing the α-KIC amide to an α-HIC amide and the β-ketone to a β-hydroxyl. Separate machinery catalyzes the dehydration of the β-hydroxythioester and subsequent enoyl reduction. The reductions are depicted as occurring on T₃, but they occur on both T₃ and T₄ in vitro. The relative ordering of α- and β-reduction is unknown.

In the absence of more detailed structural characterization of PksJ-KR, it is difficult to assess the molecular determinants of α- and β-ketoreduction stereoselectivity, but it is possible to draw several inferences. PksJ-catalyzed generation of the (S)-hydroxylated products when acting in both the β-KR and α-KR modes can be rationalized by a model in which the β-KR and α-KR substrates are recognized by the PksJ-KR active site in a similar fashion and reducing hydride equivalents are delivered from NAD(P)H to the same face of the substrate (Fig. 5). How both the β- and α-ketoreduction substrates change register to be processed by a single set of active-site residues is not yet known.

A suite of β-KR active-site residues has been identified to determine the stereochemical course of ketoreduction with “B-type” (d-hydroxy-specific) β-KRs possessing a conserved Leu–Asp–Asp motif and catalyzing β-ketoreductions with the same stereochemical sense as the PksJ-KR, and “A-type” (l-hydroxy-specific) β-KRs possessing a conserved tryptophan residue and yielding products epimeric to the PksJ-KR β-hydroxyacyl product (12, 16–18). The PksJ-KR has a completely different set of residues lining the active site, suggesting a different recognition “code.”

Because few growing chains with both α- and β-keto groups have been examined on PKS assembly-line modules, it is not yet known how many dual-functioning α/β KR domains exist. Blastp searches of the dual-function PksJ-KR revealed numerous orthologs predominantly in trans-AT clusters, including those responsible for mupirocin, bryostatin, and rhizoxin biosynthesis (19, 20), with 44%, 44%, and 43% sequence identity, respectively. Notably, many of these possible dual-function α/β KRs are predicted to act on C₂-extended α-ketoacyl-Gly-S-T thioesters, perhaps reflecting a preference for conformationally flexible substrates (21). For example, the first KR domain in Ta1 from the myxovirescin biosynthetic cluster is 38% identical to PksJ-KR. Like dihydrobacillaene, which begins with an α-hydroxyisocaproyl-glycine unit, the initial stages of myxovirescin are predicted to yield an α-hydroxyvaleryl-glycine moiety (22). In the case of myxovirescin, the α-hydroxy group serves as the nucleophile for a cyclization reaction to afford the myxovirescin macrocycle. A complex assembly of the α-hydroxyvalerate moiety has been proposed which includes use of two PK synthases (TaI and TaL) (22). An alternative based on this work is that α-ketovaleryl-glycine is generated and tethered to the first T domain in Ta1 by an as-yet unknown mechanism, which, like in dihydrobacillaene biosynthesis, is extended by the first PKS module (KS-KR-T) followed by α- and β-ketoreduction.

To date we have not carried out kinetic studies on the timing of β- and α-ketoreduction timing. Such experiments would require proteolysis and MS analysis of the acyl-pantetheinyl peptides from T₃ and T₄, which exist only in stoichiometric amounts relative to the cis-acting KR catalyst. We do not yet know whether the tandem T-domain arrangement of T₃ and T₄ acts, among other things, as a pause site before chain transfer to the next, fourth module of PksJ. If so, extra dwell time might allow the KR of module three to catalyze β-ketone reduction on a typical tethered acyl substrate and then carry out reduction of the nearby α-keto moiety of KIC. Although the apparent rate of α-reduction using substrates tested here is ≈10-fold less than the apparent rate of β-reduction, the current data do not allow a conclusion regarding the sequence of reduction. Among the complicating factors is that along with α- and β-reduction, dehydration and enoyl reduction also occur while the substrate is tethered to the tandem T domains in PksJ; although β-reduction must precede dehydration and enoyl reduction, α-reduction could occur at any point in the sequence. Attempts to generate the β-keto and olefinic α-reduction substrates corresponding to earlier intermediates were abandoned because of issues of synthetic accessibility and instability.

The above experiments also did not provide insight into a possible preference for one or the other T domain as an α- or β-reduction scaffold. Both α- and β-reductions were observed to occur on each T domain, regardless of whether the domains were independently assayed by protease digestion followed by mass spectral analysis or by independent mutation of the phosphopantetheinylated serine residue in each domain. Similar T-domain doublets and triplets have been observed in several systems, including sterigmatocystin, albicidin, curacin, and other secondary metabolites, and in polyunsaturated fatty acid biosynthetic clusters, where up to nine consecutive T domains have been observed (23–27).

Conclusion

The unusual dual-function KR found in PksJ represents a violation of canonical PK biosynthetic logic, whereby a given module acts to tailor a monomer incorporated by an upstream module. By exploiting this noncanonical strategy, an assembly line can access α-hydroxyacyl-containing structures even when α-hydroxyacids may be lacking in the primary metabolic pool. Continuing work should be directed at identifying additional examples of this class of α/β-KRs to help ascertain their catalytic scope and biosynthetic prevalence.

Materials and Methods

General Methods, Cloning and Expression, Substrate Synthesis, Enzyme Essays, and FTMS.

See SI Text.

Ketoreduction Chirality Determination.

For additional details see SI Text.

Determination of α-Reduction Stereoselectivity.

Cloning of pksJ fragments was preformed according to standard procedures (28), and recombinant forms of PksJ were expressed and purified (see SI Text). A preparation of α-KIC-GABA-S-CoA (see SI Text) was transferred to apo-PksJ(KR-T₃-T₄) using Sfp in used Hepes buffer containing MgCl₂, 20 mM α-KIC-GABA-CoA, and Sfp, and incubated at 23°C for 60 min to generate α-KIC-GABA-S-PksJ(KR-T₃-T₄) after which NADPH was added and further incubated at 23°C. After 60 min, the solution was filtered, concentrated, brought up in a Mops buffer (pH 7.5), and to which TycF was added. After 1-h incubation at 23°C reactions were filtered and combined with an equal volume of 20% pyridine in acetonitrile, 20 mM amine fluorophore (AMC) in acetonitrile, and 1 M 1-ethyl-3-[3-dimethylaminopropyl]carbodiimide (EDC) in water, and further incubated at 23°C. The reaction was purified by HPLC (C₁₈ stationary phase) on a gradient of 20–100% acetonitrile in 0.1% aqueous trifluoroacetic acid over 20 min. The α-HIC-GABA-AMC was identified by coelution with authentic standard, collected, and lyophilized. The pure enzymatically generated α-HIC-GABA-AMC was injected on an AD-H chiral HPLC column (Chiraltech) and subjected to isocratic elution using 30% ethanol in hexanes. The stereochemistry of the enzymatic product was identified by coinjection with authentic (S)-α-HIC-GABA-AMC and racemic α-HIC-GABA-AMC.

Determination of β-Reduction Stereoselectivity.

Acac-loaded PksJ(KR-T₃-T₄) was generated by incubation (30°C) of apo-PksJ(KR-T₃-T₄) with Sfp and Acac-CoA (see SI Text) in Hepes (pH 7.8) supplemented with MgCl₂. After a 1-h incubation NADPH was added and the reduction reaction was transferred to 23°C. After 60 min, the protein solution was filtered and concentrated by centrifugation. The Acac-S-PksJ(KR-T₃-T₄) concentrated sample was brought up in a Mops buffer (pH 7.5) containing TycF, and the mixture was incubated at 23°C. After 1 h, the reaction was once again filtered and the filtrate was combined with an equal volume of 20% pyridine in acetonitrile, 20 mM AMC in acetonitrile, and 1 M EDC in water and further incubated at 23°C for 1 h. The derivatization reaction was purified by HPLC (C₁₈ stationary phase) on a gradient of 20–100% acetonitrile in 0.1% aqueous trifluoroacetic acid over 20 min. The 3HB-AMC was identified by coelution with authentic standard, collected, and lyophilized (see SI Text). Like the enzymatically generated α-HIC-GABA-AMC, the 3HB-AMC product was injected on an AD-H chiral HPLC column (Chiraltech) and subjected to isocratic elution using 30% ethanol in hexanes. The stereochemistry of the enzymatic product was identified by coinjection with authentic 3HB-AMC stereoisomers.

Active-site peptide mapping of PksJ(A₁-T₁-C₁-A₂-T₂) can be seen in Fig. S1; dipeptide formation on PksJ(A₁-T₁-C₁-A₂-T₂) can be seen in Fig. S2; active-site peptide identification of PksJ(KR-T₃-T₄) can be seen in Fig. S3; and qualitative time course of α- and β-ketoreduction catalyzed by PksJ KR using PksJ(KR-T₃-T₄) can be seen in Fig. S4. Theoretical and experimental masses of detected peptides are provided in Table S1.