Introduction

Polyketide synthases (PKS) are large multidomain enzyme complexes that synthesize a variety of bioactive natural products.1, 2, 3 One family of PKS systems, most vividly illustrated by the 6-deoxyerythronolide B synthase (DEBS; Figure 1a), has an assembly line architecture with individual catalytic domains organized into successive modules.4, 5 Each catalytic domain performs a unique reaction in the overall catalytic cycle of the PKS. A typical module of an assembly line PKS consists of a ketosynthase (KS), acyltransferase (AT) and acyl carrier protein (ACP), which together yield a β-ketoacyl-ACP intermediate. This intermediate is then modified by other tailoring domains, such as the ketoreductase (KR), dehydratase (DH) and enoyl reductase (ER) domains, to introduce functional group diversity along the carbon chain backbone of the polyketide.

Figure 1
figure 1

(a) DEBS schematic representation. (b) Reactions carried out by KR domains of DEBS catalyzed by (i) KR1; (ii) KR2, KR5 and KR6; (iii) KR3; and (iv) KR4. Following the condensation reaction, the substrate for each KR is a (2R)-2-methyl-3-ketoacyl-ACP which is then processed to yield a (i) (2S, 3R)-2-methyl-3-hydroxy; (ii) (2R, 3S)-2-methyl-3-hydroxy; (iii) (2S)-2-methyl-3-keto; or (iv) (2R, 3R)-2-methyl-3-hydroxy, acyl product. A full color version of this figure is available at The Journal of Antibiotics journal online.

Full size image

The KR domains of assembly line PKSs are the most prevalent tailoring domains, and are responsible for establishing the stereo-configuration of both the α- and β-carbon atoms of their (2R)-2-methyl-3-ketoacyl-ACP substrates (Figure 1b). In addition to catalyzing NADPH-dependent ketone reduction, some KR domains also harbor epimerase activity, allowing them to catalyze epimerization of the methyl groups of their acyl-ACP substrates. Earlier studies have primarily focused on elucidating the stereocontrol principles of individual KR domains,6, 7, 8 including their ability to catalyze α-methyl epimerization.9 Here we have utilized a previously developed assay system to interrogate the contribution of protein–protein interactions to the overall substrate specificity of KR domains.10

Results

The catalytic activities of stand-alone DEBS KR1 and KR6 were initially quantified by oxidation of 2-methyl-3-hydroxyacyl-CoA substrates that correspond to their respective natively produced ACP-bound (2S, 3R)- and (2R, 3S) enantiomers. As seen in Figure 2a, both enzymes oxidize their substrates with comparable catalytic efficiency; the kcat/KM of DEBS KR1 for NDK-CoA was 4 mm−1 h−1, whereas the corresponding value of KR6 for EDK-CoA was 0.9 mm−1 h−1. Neither enzyme had a particularly high affinity for its CoA substrate; only KR1 showed evidence of saturation under our assay conditions, with an estimated KM of 0.2 mm.

Figure 2
figure 2

Activity of DEBS KR1 (squares) and KR6 (diamonds) with (a) their preferred acyl-CoA substrates and (b) the corresponding acyl-ACP substrates ACP1 and ACP6, respectively. The preferred substrates for KR1 were NDK ((2S, 3R)-2-methyl-3-hydroxypentanoyl-) thioesters, whereas the preferred substrates for KR6 were EDK ((2R, 3S)-2-methyl-3-hydroxypentanoyl-) thioesters. All reactions were performed in triplicate with error bars showing 1 s.d. The lower limit of quantification of turnover in this assay is ~0.01 h−1.

Full size image

To evaluate the specificity of the same KR domains for their acyl-ACP substrates, we first needed to attach the acyl substrate onto the pantetheine arms of these small helical proteins. To do so, a phosphopantetheinyl transferase, Sfp, with broad substrate specificity for preformed acyl-CoA substrates was used.11 In the presence of NDK-CoA or EDK-CoA, Sfp was used to prepare acyl-ACP derivatives of DEBS ACP1, ACP2, ACP3, ACP5 and ACP6, which were then individually incubated with DEBS KR1, KR2, KR5 or KR6 in the presence of saturating amounts of NADP+. As seen in the cases of KR1 and KR6 (Figure 2b), acyl-ACP substrates are substantially better (kcat/KM 10–100-fold higher) than their CoA thioester counterparts, suggesting that protein–protein recognition between the KR and the ACP contributes significantly to catalysis of this critical reaction in polyketide biosynthesis.

To assess the ability of representative KR domains to discriminate between alternative ACP carriers, the kcat/KM values of each of the above four DEBS KR domains was measured against a panel of acyl-ACP substrates harboring the identical acyl group. As summarized in Table 1, KR1 shows ~10–50-fold kinetic preference for its cognate (2S, 3R) attached to its native ACP1 over the other four ACPs tested. In contrast, the specificities of KR2 and KR5 for their cognate (2R, 3S) ACP were more modest, showing a smaller 4–5-fold advantage in measured kcat/Km. Unexpectedly, KR6 was relatively non-discriminatory between its cognate (2R, 3S) ACP substrate and either ACP1, ACP2 or ACP3, but showed an approximately 10-fold preference for substrate attached to ACP5. It should be noted that the absolute turnover rates of KR5 and especially KR2 are low, and therefore present lower confidence limits to the reported specificity analysis of these two domains.

Table 1 Specificity constants (kcat/KM; mm−1 h−1) for KR catalyzed oxidations of their preferred acyl-ACP substrates (as before, KR1 preferred NDK thioesters whereas KR 6 preferred EDK thioesters) loaded onto a panel of ACPs
Full size table

Given the enhanced reactivity of acyl-ACP substrates relative to acyl-CoA substrates, we attempted to determine if a given KR could oxidize the enantiomer of its natural substrate. To do so, KR1 was presented with 43 μM EDK-ACP1. With NDK-ACP1 at a comparable concentration, KR1 displayed turnover of ~20 h−1; however, with the enantiomeric counterpart, no reduction of NADP+ above background levels could be detected. Similar results were obtained with KR 6 and NDK-ACP6. Thus, KRs appear to have at least a 1000-fold specificity for acyl chains with the natural α- and β-carbon stereochemistry as compared with enantiomeric forms.

Discussion

The importance of specific protein–protein interactions in the catalytic cycles of assembly line PKSs has been well established.12 For example, the KS domains of DEBS have significant preference (~10–1000-fold) for their cognate ACP domains when catalyzing polyketide chain elongation13 as well as intermodular chain translocation.14 Similarly, AT domains show preference for their cognate ACP domains in the context of extender unit transacylation.15 Our data presented here suggests that the specificity of KR domains for cognate ACP domains is relatively low. Moreover, given the unexpected preference of KR6 for a neighboring ACP domain (ACP5), there appears to be little evolutionary pressure on KR-ACP recognition.

Computational analysis of the recognition of ACP1 by KR1 of DEBS identified possible interactions between the two domains, including Arg 1857 (KR): Asp 45 (ACP), Phe 1856 (KR): Leu 47 (ACP) and four other residues on the KR that were predicted to make contact with the ACP.16 Both the Asp and Leu residues of ACP1 are conserved across the five ACPs tested in our study, suggesting that conserved interactions contribute to the bulk of KR-ACP recognition in assembly line PKSs. Our findings have encouraging implications for the engineering of PKSs through KR domain substitution, as they suggest that protein–protein interactions are unlikely to be important considerations in the design of a chimeric assembly line.

Materials and methods

Instrumentation

UV–vis spectrophotometric assays were performed on a BioTek Synergy HT Microplate Reader (BioTek Instruments, Inc., Winooski, VT, USA). Liquid chromatography-electronspray ionization-tandem mass spectrometry analysis was carried out using a Waters single quadrupole mass spectrometer (Waters Corporation, Milford, MA, USA) and Acquity H Class UPLC with photodiode array detector (Waters Corporation). Liquid chromatography-electronspray ionization-tandem mass spectrometry samples were run on a 10 min gradient, eluting with increasing acetonitrile containing 0.1% formic acid over a C8 column (Zorbax 300SB-C8, Agilent Technologies, Inc., Santa Clara, CA, USA).

Plasmids

Plasmids expressing the Sfp phosphopantetheinyl transferase, KRs and ACPs were constructed using standard molecular cloning procedures. Detailed descriptions of these plasmids can be found in Supplementary Tables 1 and 2.

Protein purification

Expression plasmids were introduced into Escherichia coli BL21(DE3) cells via electroporation, then plated onto Luria-Bertani media plates with the appropriate antibiotic. Colonies were selected and grown in 5 ml seed cultures for 10–15 h at 37 °C, then used to inoculate 1 l autoclaved Luria-Bertani with the appropriate antibiotic. Cells were grown to an A600 of 0.6–0.8 then induced with 250 μl of 1 M IPTG, cooled to 18 °C, and incubated at this temperature for 12–15 h.

Cells were centrifuged at 6370 g, then resuspended in 30 ml lysis buffer (50 mm sodium phosphate, 450 mm sodium chloride, 10 mm imidazole, 20% glycerol, adjusted to pH 7.6) per liter of original cell culture. The resulting cell slurry was lysed on ice, using a Branson Sonifier 450 (Branson Ultrasonics, Danbury, CT, USA). The lysate was centrifuged at 29 000 g for 1 h, and the supernatant was mixed with Ni-NTA agarose resin (2.5 ml resin per liter of culture) and incubated with gentle agitation for at least 1 h. After incubation, the loaded resin was applied to a Kimble-Kontes Flex column (Thermo Fisher Scientific, Waltham, MA, USA) and washed with 20 column volumes of lysis buffer, 10 column volumes of wash buffer (50 mm sodium phosphate, 300 mm sodium chloride, 25 mm imidazole, 10% glycerol, adjusted to pH 7.6), and eluted with 2–6 column volumes of elution buffer (150 mm sodium phosphate, 40 mm sodium chloride, 500 mm imidazole, 10% glycerol, adjusted to pH 7.6). For ACPs, the eluate was further processed with GE PD-10 desalting columns and Buffer A (50 mm sodium phosphate, 10% glycerol, adjusted to pH 7.6) in order to improve the efficacy of downstream purification steps. The eluate was further purified using anion exchange chromatography over a GE HiTrap Q column (GE Healthcare Bio-Sciences, Pittsburgh, PA, USA) with Buffer A (as above) and a linear gradient of Buffer B (50 mm sodium phosphate, 500 mm sodium chloride, 10% glycerol, adjusted to pH 7.6). FPLC fractions were tested using SDS–polyacrylamide gel electrophoresis, and fractions containing the protein of interest were pooled, concentrated, and buffer exchanged into Buffer A using Amicon Ultra-15 centrifugal filter devices (EMD Millipore, Billerica, MA, USA). Purified proteins were aliquoted and flash frozen in liquid nitrogen, and stored at −80 °C. Protein concentrations were determined using the Pierce BCA Protein Assay Kit (Thermo Fisher Scientific) with BSA as a reference.

Synthesis of (2S, 3R)-2-methyl-3-hydroxypentanoyl-CoA (NDK-CoA) and (2S, 3R)-2-methyl-3-hydroxypentanoyl-CoA (EDK-CoA)

NDK-CoA and EDK-CoA were synthesized, as described previously.17, 18, 19 Ellman’s reagent was used to precisely quantify the concentration of each stock solution. Specifically, 20 μl of 2 m NaOH was added to a 40 μl aliquot of each purified acyl-CoA and incubated at RT for 15 min to hydrolyze the thioester. The reaction was quenched with 20 μl 2 m HCl. The amount of acyl-CoA was determined by measuring the thiol content using Ellman’s reagent, and comparison to a standard curve generated from different concentrations of L-cysteine.

Synthesis of NDK-ACP and EDK-ACP substrates

A typical acyl-ACP substrate was synthesized by co-incubating the corresponding apo-ACP (100 μm) and acyl-CoA (100–600 μm) with 1 μm Sfp phosphopantetheinyl transferase in a buffer containing 10 mm magnesium chloride and 100 mm sodium phosphate, pH 7.2.20 After reaction at RT for ~2 h, the mixtures were applied serially to five Zeba spin desalting columns with a 7 kDa molecular weight cutoff to remove unreacted acyl-CoA (using 100 mm sodium phosphate, pH 7.2, as equilibration buffer). Thereafter, a portion of the sample was concentrated using Amicon Ultra-0.5 ml centrifugal filter devices.

Ketoreductase activity assays

Each batch of a purified KR was tested for activity as well as consistency with previous preparations of the same enzyme by measuring its ability to reduce the model substrate, trans-decalone.21 Assays were performed in 100 μl volumes with 2 mm trans-decalone, 1 mm NADPH and 5 μm KR in 100 mm sodium phosphate buffer, pH 7.2. After initiation with NADPH, reaction curves were monitored at 340 nm corresponding to oxidation of the cofactor NADPH to NADP+ (Supplementary Figure 1).

KR activity against acyl-CoA substrates was quantified in 100 mm sodium phosphate, pH 7.2, with ~25 μg ml−1 BSA, 1 μM KR and 2 mm NADP+. After initiation with NADP+, the entire reaction volume was immediately transferred into a 96-well plate (Corning 3993), and readings were taken using a 360/460 nm filter to monitor the production of NADPH. Reaction progress curves in relative fluorescence units were converted to μM NADPH by comparison to a linear standard curve of NADPH in 100 mm sodium phosphate, pH 7.2.

KR activity against acyl-ACP substrates was quantified analogously in 50 μl reaction volumes containing 100 mm sodium phosphate, pH 7.2, with 0.5–1 μm KR and ~25 μg ml−1 BSA to stabilize protein components. Reactions were initiated by adding the protein mixture directly to 2.5 μl of 40 mm NADP+, resulting in a final concentration of 2 mm NADP+. Negative controls contained the cofactor NADP+ and included the acyl-ACP without KR, the KR only, and the KR with desalted ACP that had been mixed with acyl-CoA in the absence of Sfp (to confirm retention of acyl-CoA by the desalting column). Assays were performed in a 96-well plate (Corning 3993) using a 360/460 nm filter to monitor the production of NADPH. Reaction progress curves in relative fluorescence units were converted to μm NADPH by comparison to a linear standard curve of NADPH in 100 mm sodium phosphate, pH 7.2. A portion of the desalted acyl-ACP was set aside. Part of this aliquot was flash frozen and later processed with the BCA assay to determine the amount of acyl-ACP recovered after the desalting procedure. In addition, the chemical composition of each acyl-ACP was verified by liquid chromatography-electronspray ionization-tandem mass spectrometry. Most ACPs had minor post-translational modifications such as acetylation, gluconoylation and phosphogluconoylation.22 When modified with the acyl-CoA and Sfp, the apo-ACP peaks all increased by the mass corresponding to the added acyl-phosphopantetheine group. In all cases, only small amounts (<10%) of unloaded apo-ACP remained (Supplementary Figures 2 and 3).