SUMMARY
Modular type I polyketide synthases (PKSs) are versatile biosynthetic systems that initiate, successively elongate and modify acyl chains. Intermediate transfer between modules is mediated via docking domains, which are attractive targets for PKS pathway engineering to produce novel small molecules. We identified a Class 2 docking domain in cyanobacterial PKSs and determined crystal structures for two docking domain pairs, revealing a novel docking strategy for promoting intermediate transfer. The selectivity of Class 2 docking interactions, demonstrated in binding and biochemical assays, could be altered by mutagenesis. We determined the ideal fusion location for exchanging Class 1 and Class 2 docking domains and demonstrated effective polyketide chain transfer in heterologous modules. Thus, Class 2 docking domains are new tools for rational bioengineering of a broad range of PKSs containing either Class 1 or 2 docking domains.
INTRODUCTION
Polyketide natural products provide the chemical backbone for a large percentage of pharmaceuticals now in clinical use to treat human and animal diseases (Newman and Cragg, 2007). Exploring new chemical space by manipulation of microbial biosynthetic pathways for these complex and chemically diverse molecules is an area of active investigation with broad applications for synthetic biology (Floss, 2006; Kittendorf and Sherman, 2006; Menzella and Reeves, 2007; Walsh, 2002).
Polyketides are synthesized in stepwise fashion from acyl-CoAs by polyketide synthases (PKS) (Fischbach and Walsh, 2006). The type I modular PKSs may be the most amenable for rational engineering due to the diversity of their products and their modular organization (Donadio et al., 1991; Sherman, 2005). Minimally, each module contains three domains for two carbon extension of the polyketide intermediate: an acyl carrier protein (ACP) domain to carry pathway intermediates or extender units through a phosphopantetheine-mediated thioester bond, an acyltransferase (AT) to load extender units on the ACP from acyl-CoAs, and a ketosynthase (KS) to catalyze carbon-carbon bond formation between the chain elongation intermediate from the previous module and the extender on the ACP within the module. Modules may also contain various combinations of ketoreductase (KR), dehydratase (DH), and enoylreductase (ER) catalytic domains that successively process β-ketones to β-hydroxy groups, double bonds, and single bonds, respectively. Moreover, a C-terminal thioesterase (TE) in the final module removes the mature intermediate from the ACP to provide a linear acyl carboxylic acid or cyclic macrolactone product.
Biosynthetic pathway and product fidelity are critically dependent on the correct transfer of chain elongation intermediates from one PKS module to the next. This is straightforward in the case of bimodule proteins, as an upstream ACP is fused directly to the adjacent downstream KS domain. When successive modules operate from independent proteins, non-covalent association of C- and N-terminal docking domains promote protein-protein interaction of the upstream ACP and downstream KS (Gokhale et al., 1999) (Figure 1A). Docking domains, ACPdd at the ACP C-terminus of the upstream module and ddKS at the KS N-terminus of the downstream module, are essential to ensure correct transfer of polyketide chain elongation intermediates (Gokhale et al., 1999; Kittendorf et al., 2007; Kumar et al., 2003; Tsuji et al., 2001; Weissman, 2006a, b; Wu et al., 2002; Wu et al., 2001), and thus are essential structural elements for engineering these pathways to generate novel small molecules by rearrangement or recombination of PKS modules (Menzella et al., 2007; Menzella et al., 2005; Reeves et al., 2004; Yan et al., 2009). Although early studies demonstrated that cognate docking domains can facilitate intermediate transfer between modules that do not naturally associate (Menzella et al., 2007; Menzella et al., 2005; Reeves et al., 2004; Wu et al., 2002; Yan et al., 2009), none of the systems explored docking domain structure and function across broad phylogenetic groups.
Figure 1. Curacin docking domains.
(A)PKS portion of the curacin A (Cur) pathway from Moorea producens (Chang et al., 2004). Matched docking domain pairs used for the binding experiments are indicated with the same color. Crystallized docking domain pairs are indicated by red boxes.
(B) Sequence alignment of Cur ACPdds (top, orange) and Cur ddKSs (bottom, cyan) used for the crystallization experiments extracted from a sequence alignment of 23 cyano- and myxobacterial docking domain pairs (Figure S1). Matched docking domain pairs are indicated by the same color. Residues are colored by type (hydrophobic-yellow, polar-green, acidic-red, basic-blue, and P or G-purple) and conservation (darker colors indicate higher conservation). Secondary structure elements are annotated above the sequence alignments (rectangles are helices). Purple stars indicate amino acids involved in the extended interface of the Class 2 dock, red circles indicate those involved selectivity-promoting electrostatic interaction, and the a and d positions of the coiled coil heptad repeats are labeled. See also Figure S1.
Previously we demonstrated the specificity of protein-protein interactions in binding studies of all pairs of docking domains from two actinobacterial PKS pathways, including 6-deoxyerythronolide B synthase (DEBS) and pikromycin synthase (Pik) (Buchholz et al., 2009). In these two systems binding occurs only for cognate docking domain pairs, demonstrating that the relatively weak interaction (Kd ~20 μM) has the requisite specificity to maintain biosynthetic fidelity in the pathway. High-resolution structure analysis (Broadhurst et al., 2003; Buchholz et al., 2009) of docking domains from the DEBS and Pik pathways revealed their dimeric form, consistent with the oligomeric state of full-length PKS modules (Aparicio et al., 1994; Staunton et al., 1996). We refer to ACPdd and ddKS from actinobacterial PKS modules as “Class 1” docking domains. The ACPdd consists of two dimerization helices that form a four-helix-bundle dimer, followed by a C-terminal docking helix, which binds to the coiled-coil formed by the dimerization of the single ddKS helix. The docking interface consists of small (~550Å), complementary hydrophobic surfaces surrounded by electrostatic interactions that promote specificity. The ACPdd docking helices and the ddKS coiled-coil helices are approximately parallel, with the connection to the upstream ACP directed away from the downstream KS. Long linkers (30–50 amino acids) between the ACP and ACPdd promote movement of the carrier protein to both the downstream KS and catalytic domains in the upstream module (JRW et al. submitted 2013).
The cyanobacterial PKSs such as those in the curacin A pathway (Cur) (Figure 1A) of Moorea producens (Chang et al., 2004), appear to employ a different strategy for mediating intermodule protein-protein interactions, as their ACPdd and ddKS amino acid sequences cannot be classified with the corresponding actinobacterial Class 1 docking domains (Thattai et al., 2007). The Cur pathway is particularly well suited to investigation of cyanobacterial docking domain interactions as it contains seven mono-module PKS polypeptides and six cognate docking domain pairs (Chang et al., 2004). Here we identify a novel “Class 2” docking domain with significant potential as a new tool for engineering PKS pathways, present the structural and biochemical characterization of cyanobacterial docking domains from Cur, and demonstrate selectivity determinants for cognate docks. Class 2 docking domains are able to direct the upstream ACP towards the downstream KS, and thus may be especially effective at transfer of polyketide intermediates between modules. We also determined the ideal fusion location for replacement of Class 1 docking domains with their Class 2 counterparts and demonstrate effective association and transfer of chain elongation intermediates between modules from the Pik actinobacterial modular PKS.
RESULTS
Analysis of cyanobacterial docking domain sequences
We aligned amino acid sequences of cyanobacterial, myxobacterial and actinobacterial PKS docking domains as a first step to obtain insights into structural similarities and differences among the relevant phylogenetic bacterial groups (Figure S1). The cyanobacterial and myxobacterial docking domains, mostly from mixed PKS-NRPS pathways, clustered in a group, here designated Class 2 (Figure 1B and S1), that suggests a unique structural paradigm compared to the one previously established for Class 1 actinobacterial docking domains (Broadhurst et al., 2003; Buchholz et al., 2009). The Class 2 ACPdds are ~40 amino acids shorter than the Class 1 ACPdds, and are too short to include the dimerization region. The Class 1 and Class 2 ddKSs are of similar length, but the N-terminal region of the Class 2 ddKSs is too polar to form a coiled-coil, in contrast to the complete coiled-coil of Class 1 ddKSs (Broadhurst et al., 2003; Buchholz et al., 2009).
To understand the Class 2 docking interaction in greater detail, we investigated the affinities of four docking domains from the curacin pathway and solved crystal structures for two. The Cur ACPdd was found to lack a dimerization region (Broadhurst et al., 2003), as all five ACP-dd constructs from our binding experiments were monomeric. To facilitate crystallization, cognate ACP and KS docking domain pairs were fused, in anticipation that the protein-protein binding interaction may be weak, as for the Class 1 docking domains (Buchholz et al., 2009). Structures of Class 1 docking domains had an ACPdd fused directly to a ddKS, limiting information about the natural chain termini (Broadhurst et al., 2003; Buchholz et al., 2009). Thus, we fused the ACPdd C-terminal region (~30 residues predicted to be helical, Figure 1B) to the N-terminus of the cognate ddKS via an eight amino acid (Gly3Ser)2 flexible linker (ACPdd-G3SG3S-ddKS) to allow the natural docking interaction. As expected for a coiled-coil ddKS, the fusion proteins were dimeric in solution. Crystals and high-resolution structures (Table 1) were obtained for the CurG-CurH ACPddG-G3SG3S- (GH dock) and the CurK-CurL ACPddK-G3SG3S- (KL dock) (Figure 2). Additionally, we determined the crystal structure of the CurL KS-AT di-domain including the N-terminal ddKS (dd-KS-AT).
Table 1.
Crystallographic data
GH dock SeMet | GH dock Native | KL dock L68M SeMet | KL dock Native | CurL dd-KS-AT | |
---|---|---|---|---|---|
Data Collection | |||||
Space group | P3121 | P3121 | C2 | C2 | P212121 |
Cell dimensions | |||||
a,b,c (Å) | 36.7, 36.7,186.9 | 36.8, 36.8,187.2 | 90.9, 59.5,52.9 | 90.9, 59.5, 52.9 | 69.3, 150.8, 236.0 |
α, β, γ (°) | 90, 90, 120 | 90, 90, 120 | 90, 124.9, 90 | 90, 124.9, 90 | 90, 90, 90 |
X-ray source | APS 23ID-D | APS 23ID-D | APS 23ID-B | APS 23 ID-D | APS 23ID-D |
Wavelength (Å) | 0.979 | 1.033 | 0.979 | 1.033 | 1.033 |
dmin (Å) | 2.13 (2.21–2.13) a | 1.68 (1.74–1.68) a | 2.5 (2.54–2.5) a | 1.5 (1.55–1.5) a | 2.8 (2.9–2.8) a |
Rsymm (%) | 8.7 (47) | 6.1 (72) | 8.3 (18) | 4.5 (65) | 11 (69) |
Ave I/σI | 37.3 (5.3) | 28.4 (3.4) | 20.6 (8.2) | 28.1 (2.5) | 25.0 (3.5) |
Completeness (%) | 98.6 (97.0) | 97.3 (96.1) | 87.1 (59.7) | 96.9 (98.7) | 99.8 (99.6) |
Average redundancy | 19.6 (12.6) | 10.7 (11) | 6.6 (4.8) | 3.7 (3.7) | 6.5 (6.3) |
Refinementb | |||||
Data range (Å) | 50–1.68 | 46.5–1.5 | 50–2.8 | ||
No. reflections | 16,316 | 33,881 | 61,718 | ||
Rwork/Rfree | 0.19/0.25 | 0.18/0.20 | 0.19/0.23 | ||
Polypeptide chains | 2 | 3 | 2 | ||
Atoms(#) | |||||
Protein | 1,274 | 1,201 | 12,810 | ||
Water | 122 | 160 | 282 | ||
Ion | 15 | - | 2 | ||
RMS deviations | |||||
Bond length (Å) | 0.0089 | 0.0096 | 0.01 | ||
Bond angle (°) | 1.23 | 1.19 | 1.22 | ||
Avg B-factor (Å2) | |||||
Protein | 15.2 | 22.2 | 60.5 | ||
Water | 36.4 | 50.6 | 47.6 | ||
Ligand | 58 | - | 47.4 | ||
Ramachandran plot | |||||
Allowed (%) | 100 | 100 | 99.59 | ||
Outliers (%) | 0 | 0 | 0.41 |
Values in parentheses pertain to the outermost shell of data.
The final structures are deposited in the PDB with accession codes 4MYY for GH dock Native, 4MYZ for KL dock Native, and 4MZ0 for CurL dd-KS-AT.
See also Table S2.
Figure 2. Class 2 docking domain structures.
(A) GH dock. In the upper stereo diagram, monomers are shown with different shades of orange (ACPdd) or cyan (ddKS). The GlySer linker (green) connects the upstream ACPdd C-terminus and the downstream ddKS N-terminus. Connections to the ACP and KS domains are shown with thick lines. Close-up stereoviews are shown of the Class 2-specific ACPddG α1 interface with αA and αB (purple box) and the common ACPddG α2 interface with α B coiled-coil (red box). Hydrophobic contacts in the docking interface are shown as sticks with orange C for the ACPdd, cyan C for the ddKS, red O and blue N.
(B) KL dock. Coloring and labeling are as in A. Zoomed-in stereoviews of the boxed regions of the upper stereo diagram are highlighted below. Electrostatic interactions that may promote selectivity are labeled, and hydrogen bonds are indicated with red dashes (detail in Figure S2E, F). See also Figure S2.
Structure of Class 2 docking domains
The crystal structures of the GH dock, KL dock, and CurL dd-KS-AT revealed a novel interaction strategy for the Class 2 docks (Figure 2). Each docking domain consists of two α-helices connected by a sharp bend (αA and αB in the ddKS, α1 and α2 in the ACPdd). In the ddKS, helix αB from the two monomers forms a parallel coiled-coil dimer, as predicted, in which hydrophobic residues at the a and d positions of a heptad repeat form the dimer interface. However, in a striking departure from the Class 1 docking domain structures (Broadhurst et al., 2003; Buchholz et al., 2009; Tang et al., 2006), the full ddKS is not a coiled-coil, and the more polar helix αA extends away from the dimer interface. The ddKS in one subunit of the CurL dd-KS-AT structure forms a single extended helix including both αA and αB regions, but the partner subunit cannot form a coiled-coil dimer with the αA region because the αA helix has no hydrophobic surface to support coiled-coil formation (Figure S1B, S2A, and S2B). The break in the coiled-coil occurs at equivalent positions in the GH and KL docks and in CurL dd-KS-AT, where a short αA-αB connecting loop (ddKS loop) creates a sharp bend in the polypeptide chain (Figure 2, 3A). Like the ddKS, the ACPdd also has a sharp bend between α1 and α2 (ACPdd loop) (Figure 2, 3B).
Figure 3. Superposition of Class 2 ddKS and ACPdd.
(A) Superposition of Class 2 ddKS. The CurL ddKS from KL dock structure (cyan) was superimposed with the CurL ddKS (yellow) from the CurL dd-KS-AT structure (RMSD = 1.3 Å) and with the CurG ddKS (magenta) from the GH dock structure (RMSD = 1.0 Å). Superposition of Fo-Fc simulated annealing (SA) density contoured at 3σ for the ddKS loops from each of the three structures are shown below the superposition.
(B) Superposition of the Class 2 ACPdd. The ACPdd from GH dock structure (yellow) and KL dock (orange) were superimposed (RMSD = 0.564 Å). Fo-Fc SA density contoured at 3σ for the ACPdd loops from the GH and KL dock structures are shown below the superposition.
All helices in ACPdd and ddKS contribute to the Class 2 dimeric docking interaction (Figure 2). Two ACPdd α2 helices bind in a parallel orientation to the ddKS αB coiled-coil dimer, similar to the sole interface of Class 1 docking domains. Additionally the ACPdd α1 helix (not present in Class 1 docking domains) contacts both αA and αB within a ddKS monomer in predominantly hydrophobic interactions of generally conserved amino acids. The sharp bend in both the ACPdd loop and the ddKS loop enables α1 interaction with both ddKS helices, and orients the link to the upstream ACP domain towards the downstream KS domain (Figure 3, 4). This is a marked difference compared to the Class 1 docking domains where the links to ACP and KS are on opposite sides of the docking interaction. Moreover, by directing the ACP towards the KS, Class 2 docking domains may be especially efficient in promoting transfer of polyketide chain elongation intermediates to the downstream module (Figure 4).
Figure 4. Comparison of Class 2 (top) and Class 1 (bottom) docking domains.
The Class 2 GH dock (top left) and KL dock (top right) (orange ACPdd and cyan ddKS) were modeled onto the CurL dd-KS-AT (blue KS, green AT, red KS-AT linker domain, KS active site cysteine in yellow spheres). The Class 1 dimerization helices (orange) and ACPdd (orange)/ddKS (cyan) complex from the DEBS module 2–3 interface (Broadhurst et al., 2003) was modeled onto the KS-AT di-domain from DEBS module 5 (Tang et al., 2006). In the Class 2 GH dock (top left), the ACP is directed towards the downstream KS while the KS and ACP are much further apart in the Class 1 dock. Flexibility in the Class 2 dock, as seen in the KL dock (top right) allows the upstream ACP to interact with catalytic domains of the upstream module.
Key features of the Class 2 docking domains are essential for the formation and stabilization of the additional docking interaction between ACPdd helix α1 and both ddKS helices and are conserved in the GH and KL dock structures and also among other Class 2 docking domain sequences (Figure 1B, 3, S1). First, the Class 2 ddKS sequences have polar N-termini that are incompatible with coiled-coil formation (Figure S1B, S2B). Only the 20 amino acids proximal to the KS catalytic domain have the requisite hydrophobicity (three heptad repeats) to form a coiled-coil dimer (Figure 1B, S1B, S2B). The lack of a coiled coil in the αA region allows formation of the ddKSloop and helix αA that are necessary for the extended interface with ACPdd helix α1in Class 2 docking domains. In contrast, Class 1 docks form a longer coiled coil (4 heptad repeats) and lack the ddKSloop and helix αA (Figure S1D) (Broadhurst et al., 2003; Buchholz et al., 2009). Second, the ddKS loop has identical structures in the GH and KL docks and the CurL dd-KS-AT subunit A (Figure 3A). Each ddKS loop contains a “helix cap” hydrogen bond (N-terminus of helix αB with the side chain of CurH ddKS Ser 22 and CurL ddKS Ser14) and a large hydrophobic residue (CurH ddKS Leu21, CurL ddKS Leu13) that contributes to the additional interface of Class 2 docks (Figure 1B, 2A, 2B, 3A, S2B). Like for the ddKS loop, the ACPdd loops have identical structures in the GH and KL docks, a conserved ACPdd loop Ser forms a “helix cap” hydrogen bond with the N-terminus of the following helix α2 (CurG ACPdd Ser 1566 CurK ACPdd Ser 2213), and a large hydrophobic residue (Leu2212 of CurK ACPdd) contributes the extended hydrophobic interface of the Class 2 dock (Figure 1B, 3B, S1A).
The structures also provide evidence for motion of the ACP-proximal end of the docking domains (helices α1 and αA). In the GH dock, helicesα1 and αA have slightly different positions on opposite halves of the dimer (Figure S2C). By contrast, these helices are disordered in the KL dock, and in the CurL dd-KS-AT structure αA-αB forms a continuous helix in one monomer and αA is disordered in the other. This flexibility may facilitate ACP interaction with both the downstream KS and the catalytic domains of its own module.
The Class 2 docking domains bury significantly more surface area than their Class 1 counterparts (1834 Å2 for the GH PKS dock vs. 1100 Å2 for DEBS 2–3 dock) (Broadhurst et al., 2003). The core of the Class 2 docking domain interface is comprised of hydrophobic interactions of surfaces with high shape complementarity. The hydrophobic character is well conserved among docking domain sequences, but variability of the hydrophobic residues among docking domain pairs may promote specificity (Figure 1B). For example, in the GH dock α2 Ile1574 interacts with αB Ala28 whereas in the KL dock contact is between α2 Val2221 and αB Leu16. As a result, the size and shape complementarity may be lost in a non-cognate complex of CurG ACPdd and CurL ddKS. Non-conserved electrostatic interactions also may be important for specificity, for example Glu2224 of α2 interacts with Lys24 and Arg27 of the coiled-coil αB in the KL dock (Figure 1B, 2B, S1). In Class 2 docks, the residues at these positions are generally complementary. The equivalent residues in the GH dock are Ile1577, Lys32, and Glu35, respectively. However, due to a different position of the CurG ACPdd α2 helix relative to the coiled-coil, the Ile1577 side chain is in the hydrophobic core, away from the charged side chains of the coiled coil (Figure 2A and S2D). CurG ACPdd may not be able to bind to CurL ddKS due to an unfavorable interaction of Ile1577 with Lys24 and Arg27.
Affinity of cyanobacterial docking domains
Binding affinities of the Class 2 docking domains were determined with fluorescence polarization (FP). Five full-length ACP domains with their C-terminal docking domains (ACP-dd) from the CurG, H, I, K, and L PKS monomodules containing an engineered C-terminal cysteine were conjugated to a BODIPY fluorophore. Binding affinities for all combinations of ACP-dd and dd-KS-AT (modules CurH, I, L, and M) were determined.
The binding affinities (Kd) of matched docking domain pairs ranged from 4.5–20.5 μM (Table 2 and Figure S3A). Biolayer interferometry was subsequently employed to confirm the binding affinity of CurG ACP-dd/CurH dd-KS-AT, resulting in a Kd of 16.5 μM, in excellent agreement with the 15.7 μM Kd determined by FP (Figure S3B). The dd-KS-AT proteins varied in the extent of dimer formation, which correlated with the variation in docking affinity (Table 2). For example, CurI dd-KS-AT (100 kDa) was predominantly monomeric (apparent molecular weight 138 kDa by gel filtration) and had an affinity for its cognate CurH ACP-dd of 20.5 μM (Figure S3C). When a natural CurI dimerization element (Zheng et al., 2013) was included at the C-terminus of the CurI dd-KS-AT (CurI dd-KS-AT dimer, 114 kDa), the protein was predominantly dimeric (apparent molecular weight 174 kDa) and had two-fold greater affinity (9.4 μM) for CurH ACP-dd (Table 2 and Figure S3C).
Table 2.
Binding Data
dd-KS-AT | |||||||
---|---|---|---|---|---|---|---|
CurH | CurI | CurI dimer | CurL | CurM | CurH no αA | ||
ACP-dd | CurG | 15.7 ± 5.4 μM | NB | ND | NB | NB | 117 ± 26 μM |
CurH | NB | 20.5 ± 5.3 μM | 9.4 ± 2.3 μM | NB | NB | ND | |
CurI | NB | NB | ND | NB | NB | ND | |
CurK | 23.7 ± 7.7 μM | NB | ND | 4.5 ± 2.1 μM | 52 ± 9.4 μM | ND | |
CurK E2224R | 9.2 ± 0.6 μM | ND | ND | NB | NB | ND | |
CurL | NB | NB | ND | NB | 5.4 ± 1.2 μM | ND |
Interestingly, Class 2 and Class 1 docking domains have similar affinities, despite the larger interaction surface in Class 2 docks (Broadhurst et al., 2003; Buchholz et al., 2009). This raises the question whether the additional sequence elements of the Class 2 docking domains (helix α1 and the ACPdd loop in the ACPdd and helix αA and the ddKS loop in the ddKS) are critical to the protein-protein interaction. To answer this question, we engineered a CurH dd-KS-AT that lacked helix αA and the ddKS loop (ΔαA-dd-KS-AT). The ΔαA-dd-KS-AT protein had 7-fold weaker affinity for the cognate CurG ACP-dd (117 μM) compared to the wild type CurH dd-KS-AT (16 μM), indicating the importance of the extended interface for Class 2 docking domains.
Selectivity of Class 2 docking interactions
The Class 2 docking interactions were generally specific, as no binding was detected between most of the mismatched pairs (Table 2). Interestingly, CurK ACP-dd was promiscuous in our assay and had detectable affinity for two non-cognate partners, CurH dd-KS-AT(24 μM) and CurM dd-KS-AT(52 μM) (Table 2). A similar analysis using Class 1 docking domains did not reveal non-specific interactions between non-cognate pairs (Buchholz et al., 2009).
As discussed above, the electrostatic interaction between Glu2224 of CurK ACPdd and Lys24 and Arg27 of CurL ddKS may be important for docking selectivity, as residues at these positions are generally complementary in Class 2 docks (Figure 1B, 2B, S1, red circles). We tested this hypothesis with a Glu-to-Arg substitution in the CurK ACP-dd, which exhibits promiscuous binding with non-cognate dd-KS-ATs. Consistent with the hypothesis of electrostatic selectivity, no binding was detected between CurK ACP-dd E2224R and the cognate CurL dd-KS-AT. The Arg substitution also abolished binding to the non-cognate CurM dd-KS-AT. Surprisingly, CurK ACP-dd E2224R exhibited 2-fold greater affinity for the non-cognate CurH dd-KS-AT(10 μM) than did the wild type CurK ACP-dd (24 μM). Therefore, the single Glu to Arg substitution altered the selectivity of CurK ACP-dd from the cognate CurL dd-KS-AT to the non-cognate CurH dd-KS-AT, demonstrating the importance of the electrostatic interaction for the selectivity of Class 2 docking domains.
Effectiveness of Class 2 docking interactions
The unique docking configuration suggested that Class 2 docking domains might promote highly efficient intermodule ACP→KS polyketide transfer and thus may be effective tools for bioengineering efforts. Thus, we determined if Class 2 docking domains could promote the transfer of natural polyketide substrates between engineered PKS monomodules containing endogenous Class 1 docking domains. Furthermore, employing the same assay, we probed the ideal fusion location within a module for exchanging Class 1 docking domains with their Class 2 counterparts. Protein chimeras were engineered by substituting Cur docking domains onto the final two modules (PikAIII and PikAIV) of the pikromycin pathway (Xue et al., 1998). In reactions that require a functional docking domain interaction, PikAIII and PikAIV produce the 14-membered macrolactone by two successive rounds of methylmalonyl (MeMal)-CoA extension with the natural Pik pentaketide substrate followed by keto group processing and cyclization of the heptaketide to narbonolide (nbl) (Figure 5A). Through a domain skipping mechanism that also requires a functional docking domain interaction, this system can catalyze a single pentaketide elongation by MeMal-CoA, with processing of the hexaketide followed by off-loading and ring closure to form the 12-membered macrolactone 10-deoxymethynolide (10-dml) (Kittendorf et al., 2007) (Figure 5A).
Figure 5. PikAIII/PikAIV assay of docking effectiveness.
(A) Schematic of the two-module throughput reaction. The 10-dml and nbl macrocycles are formed by the PikAIV TE following processing of the pentaketide substrate by PikAIII and PikAIII/PikAIV, respectively.
(B) Assay results. The levels of 10-dml and nbl produced by each combination of the PikAIII/PikAIV chimeras are shown as percents of the levels with wild type (WT) PikAIII/PikAIV. The levels of starting material (thiophenol-pentaketide) consumed in each reaction were determined based on peak areas normalized to a control, which lacked enzyme. Domains are colored according to their source: PikAIII red, PikAIV magenta, CurG/CurH orange, CurK/CurL green. Catalytic and carrier domains are labeled circles, dimerization helices are boxes and ACPdd are rectangles, ddKS are crossed rectangles. ND-not detected.
See also Figure S4 and Table S1 and S2.
PikAIII docking domain chimeras were engineered by exchanging CurG or CurK ACPdd in place of the PikAIII ACPdd5 (PikAIII-ACPddG, PikAIII-ACPddK). We also created a chimera with the CurG ACPdd fused to the PikAIII ACPdd5 dimerization helices (PikAIIID-ACPddG). PikAIV chimeras were engineered exchanging dd KS from CurH or CurL in place of . An additional chimera was made with a CurH ddKS that lacked helix αA and the ddKS loop ( ). All Cur PikAIII and PikAIV chimera combinations were tested for consumption of pentaketide substrate, as well as for production of 10-dml and nbl, and directly compared to the catalytic activity and formation of products by wild type PikAIII/PikAIV with its natural Class 1 docking domains.
Macrolactone products 10-dml and nbl were both produced by the engineered Cur PikAIII-ACPdd/ddKS-PikAIV combinations for which the Cur docking domains exhibited detectable binding affinity (Table 2, Figure 5B, and Figure S4A). The chimeric docking domain interaction that included PikAIII dimerization helices ( ) was most efficient at transfer of polyketide chain elongation intermediates, consuming 86% of the substrate and producing nbl at 70%, and 10-dml at 188% of the levels of the wild type PikAIII/PikAIV monomodule pair. The high efficiency of is likely due to the predominantly dimeric state of PikAIIID-ACPddG, as chimeras lacking the dimerization helices (PikAIII-ACPddG and PikAIII-ACPddK) were monomeric (Figure S4B).
Yield of macrolactone products was correlated with binding affinity of the docking domains (Table 2, Figure 5B). As expected, engineered PikAIII and PikAIV bearing the mismatched CurK ACPdd and CurH ddKS docking domains were less effective than either cognate pair. The PikAIV chimera lacking ddKS helix αA and the ddKS loop ( ) was less efficient than with the PikAIIID-ACPddG and PikAIII-ACPddG chimeras. Finally, yielded no detectable product, either 10-dml or nbl, while produced higher levels of 10-dml and nbl than did . These results are consistent with the binding data and underscore the importance of Glu2224 and equivalent residues for promoting the association and selectivity of Class 2 docking interactions (Figure S4C).
Interestingly, the Cur PikAIII/PikAIV chimeras shifted the product ratio, producing higher levels of 10-dml and lower levels of nbl (3:1 nbl to 10-dml) compared with wild type PikAIII/PikAIV (7:1 nbl to 10-dml) (Table S1). The shift towards 10-dml suggests that Class 2 docking domains more effectively deliver the hexaketide directly to PikAIV TE, which is likely due to the docking mechanism of Class 2 docking domains that appears to direct the upstream ACP to the downstream module (Figure 4).
DISCUSSION
In this work we have identified and characterized a novel class of PKS docking domains that predominates in cyanobacterial and myxobacterial mixed PKS/NRPS pathways. These efforts have established three structures and provided high-resolution details on the protein-protein interactions and docking mechanism for the Class 2 docking domains that are distinct from the previously characterized Class 1 (actinobacterial) counterparts (Broadhurst et al., 2003; Buchholz et al., 2009). The Class 2 docking domain consists of two ACPdd helices (α1 and α2) and two ddKS helices (αA and αB). The interaction of the ACPdd α2 and the ddKS αB coiled-coil is similar to the interface that is characteristic of the Class 1 docking domains. The Class 2 docking interactions include an additional interaction between ACPdd α1 and ddKS αA and αB, which requires key structural features that are conserved in Class 2 docks but do not exist in Class 1 docks, that extend the docking interface, and that direct the upstream ACP toward the downstream KS. The additional interactions provided by the ddKS αA and the ddKS loop are critical to both binding and intermediate transfer, as an αA-truncated ddKS formed a lower affinity docking interaction and was less effective in intermediate transfer (Table 2, Figure 5B). Loops between the helices within each partner, ACPdd loop and ddKS loop, which are conserved in Class 2 docking domains but do not exist in Class 1 docking domains, provide flexibility.
The shape complementary of hydrophobic surfaces likely provides specificity for Class 2 docking interactions. Complementary peripheral charges also play a role in maintaining selectivity, for example CurK Glu2224 in ACPdd α2 and CurL Lys24 and Arg27 in ddKS αB (Figure 2B), as these residues are generally complementary among Class 2 docks. Mutagenesis of Glu2224 to Arg in CurK ACPdd not only eliminated binding to the cognate partner CurL ddKS and the non-cognate partner CurM ddKS, but also enhanced binding to and intermediate transfer via the non-cognate partner CurH ddKS. Based on the KL dock crystal structure, CurK ACPdd Glu2224 interacts primarily with Arg27 of CurL ddKS (Figure 2B). The analogous residue to Arg27 is Glu35 in CurH ddKS and Arg25 in CurM ddKS. As a result, CurK ACPdd E2224R may have introduced unfavorable Arg/Arg contacts in the interactions with CurL ddKS and CurM ddKS, and a favorable Arg/Glu contact in the interaction with CurH ddKS, thus explaining the observed selectivity. These results underscore the importance of electrostatic selectivity in Class 2 docking domains and may provide a viable method for altering docking selectivity through single residue substitutions, which could be used to engineer PKS pathways. Furthermore, the electrostatic interaction may explain the promiscuity of CurK ACPdd, which binds the non-cognate CurH and CurM ddKSs. The complementary charged residues may offset the effects of non-complementary hydrophobic surfaces to enable the observed non-specific, promiscuous interaction of the wild type docks (Figure S2E, F).
Class 2 docking domains do not contribute to PKS module dimerization, unlike the Class 1 docks where the ACPdd includes two dimerization helices (Broadhurst et al., 2003). Nevertheless the highest affinity and most effective Class 2 docking interactions require that both the upstream and downstream modules are dimeric. This is supported by the observation that the predominantly dimeric variant of CurI dd-KS-AT had twofold greater affinity for CurH ACP-dd than did the monomeric variant (Table 2), and the dimeric CurG ACPdd PikAIII chimera (PikAIIID-ACPddG) was more effective than the monomeric analog (PikAIII-ACPddG) in transfer of the hexaketide chain elongation intermediate to CurH ddKS PikAIV chimeras ( and ) (Figure 5B). All PKS modules we identified with Class 2 docking domains have domains that promote module dimerization in addition to ddKS and KS, for example a dehydratase (Akey et al., 2010; Keatinge-Clay, 2008) or a dimerization element (Zheng et al., 2013) in modules lacking a dehydratase. Our binding measurements employed exclusively monomeric ACP-dd proteins, and we expect greater affinity for dimeric docks in the context of dimeric modules.
The sequence analysis, structures and biochemical data presented here demonstrate that Class 2 docks have a distinct docking mechanism, in which the upstream ACP can be directed towards both the downstream module and the upstream module, and thus may be effective tools for PKS bioengineering efforts (Figure 4). Furthermore, when fused to the dimerization helices, Class 2 docking domains ( ) promoted intermediate transfer between PikAIII and PikAIV with total efficiency comparable to the native Class 1 docking domains (Figure 5B). However, the Class 2 docking domain interaction yielded nearly twofold more 10-dml compared to the wild type Class 1 docking domains (Table S1). This shift in product suggests that the Class 2 docking domains deliver the PikAIII hexaketide product more rapidly to the PikAIV module TE than do the Class 1 counterparts, consistent with the unique Class 2 docking mechanism.
SIGNIFICANCE
The structural and biochemical analysis presented here identified a Class 2 docking domain from cyanobacterial and myxobacterial PKS pathways that is distinct from the well characterized Class 1 docking domain of actinobacterial PKS pathways. Crystal structures of two bound docking domains pairs and of a dd-KS-AT from the Cur pathway demonstrate a novel docking strategy for Class 2 docking domains in which the upstream ACP is directed towards the downstream KS. The novel docking strategy was supported by binding and biochemical experiments. A key electrostatic interaction at the docking interface is a determinant of docking domain selectivity. Mutagenesis of the interacting residues can change the selectivity of docking interactions and thus may be a tool for engineering PKS pathways. Furthermore, Class 2 docking domains can effectively mediate intermediate transfer between modules with Class 1 docking domains. Therefore, the Class 2 docking domains are new tools with the potential to broaden the scope of modules that can be used for pathway engineering efforts to produce novel small molecules. In addition, the results presented here provide a firm foundation to guide the use of Class 2 docking domains in these engineering efforts.
EXPERIMENTAL PROCEDURES
Design of expression constructs
All PCR primers are listed in supplementary Table S2. Cur docking-domain fusions, GH dock and KL dock, were constructed with overlap PCR. Fragments encoding ACPdd and ddKS were amplified from cosmid pLM9A (Chang et al., 2004) and then used as templates in a second PCR to amplify the fusion construct. Fusion constructs were inserted into pMCSG7 (Donnelly et al., 2006) with LIC. For selenomethionyl KL dock, Leu68 was substituted with methionine by site-directed mutagenesis (KL dock L68M).
Full-length ACP domains with their C-terminal docking domains (ACP-dd) from modules CurG – CurL were amplified from cosmid pLM9A. The reverse primer appended a C-terminal cysteine for site-specific conjugation to a BODIPY fluorophore. The natural cysteine in CurL ACP-dd was substituted with serine by site-directed mutagenesis. CurG, H, I, J and L ACP-dd were inserted into pMCSG7 and CurK ACP-dd was inserted into pMocr with LIC (DelProposto et al., 2009). Glu2224 of CurK ACP-dd was substituted with an Arg by site directed mutagenesis. In addition CurG ACP-dd was inserted into pMCSG16 (Scholle et al., 2004), which contains an N-terminal avitag, for octet red (forte bio) binding experiments (Avitag-CurG ACP-dd).
KS-AT di-domains with N-terminal docking domains (dd-KS-AT) from modules CurH – CurL, CurH dd-KS-AT lacking helix αA and the ddKS loop ((ΔαA-dd-KS-AT), and CurI dd-KS-AT with the post-AT dimerization element (CurI dd-KS-AT dimer) were amplified from pLM9A. CurM dd-KS-AT was amplified from cosmid pLM14 (Chang et al., 2004). Each dd-KS-AT construct was ligated into the pSUMO vector (Weeks et al., 2007) to produce a protein with a native N-terminus.
PikAIII-ACPddG, PikAIIID-ACPddG, PikAIII-ACPddK were made with overlap PCR. For the PikAIII-ACPddG and PikAIII-ACPddL constructs, a PikAIII fragment without ACP or dimerization helices was amplified from pPikAIII (Beck et al., 2003), and fragments encoding CurG ACPdd and CurK ACPdd were amplified from pLM9A. For the PikAIIID-ACPddG chimera, a PikAIII fragment that included the ACPdd dimerization helices was amplified from pPikAIII. The fragments were used as templates in a second PCR to amplify the fusions. The fusions were digested with FseI (natural enzyme site) and HindIII and inserted into pPikAIII digested with the same enzymes. Glu2224 of PikAIII-ACPddK was substituted with an Arg by site directed mutagenesis.
The three PikAIV chimeras, were constructed following a published protocol (Menzella et al., 2005). An MfeI site was engineered at the final codon for the ddKS in pPikAIV (Beck et al., 2003) by site directed mutagenesis (pPikAIVMfeI). Fragments encoding CurH ddKS, CurH ddKS lacking helix αA and the ddKS loop, and CurL ddKS were amplified from pLM9A with 3′ NdeI and 5′ MfeI sites. The ddKS -encoding fragments were digested with NdeI and MfeI and ligated into pPikAIVMfeI digested with the same enzymes. All constructs were confirmed by sequencing.
Expression and purification of expression constructs
GH dock, KL dock (WT and L68M), and all Cur ACP-dd plasmids, except Avitag-CurG ACP-dd, were expressed in E. coli BL21(DE3) cells. Avitag-CurG ACP-dd was expressed in BL21 (DE3) cells expressing BirA (biotin ligase) for in vivo biotinylation of the Avitag. All Cur dd-KS-AT expression plasmids were expressed in E. coli BL21 AI (Invitrogen) containing the pRARE2-CDF plasmid. To construct the pRARE-CDF plasmid, the secondary plasmid pRare2 (Novagen) and the plasmid pCDF-1b (Novagen) were digested with DrdI and XbaI. The larger fragment from pRare2 carrying the tRNA genes and the smaller fragment from pCDF-1b, which carries the origin of replication and the antibiotic resistance marker gene, were gel purified and ligated to create pRARE2-CDF. WT PikAIII and PikAIV constructs and chimeric PikAIII and PikAIV constructs were expressed in E. coli BAP1 cells (Pfeifer et al., 2001).
Transformed bacteria were grown in 0.5 L of TB media with the appropriate antibiotic at 37°C to an OD600=1. Cells were cooled to 20°C for 1hr, induced with 200μM IPTG, and allowed to express for approximately 18 hr. For Avitag-CurG ACP-dd, the cells were cultured to an OD600=1 at 37°C, cooled 1hr to 20°C, induced with 0.2mM IPTG and 50μM biotin in 10mM Bicine pH 8.3, and allowed to express for approximately 18 hr
To produce selenomethionyl GH dock and KL dock L68M, cells were cultured 18 hr at 37°C in TB, centrifuged, re-suspended in 1L minimal media (AthenaES) containing 100 μg/mL D,L-selenomethionine to give an A600 = 0.4, cultured at 37°C to A600 = 0.6, incubated 1 hr at 20°C, induced with 200μM IPTG, and allowed to express 18 hr at 20°C.
Cell pellets were re-suspended in 300mM NaCl, 10% glycerol with either 50mM Tris pH 7.5 (Tris buffer A; Cur docking domain fusion) or 50mM HEPES pH 7.4 (HEPES buffer A; all Cur ACP-dd, Cur dd-KS-AT, WT PikAIII and PikAIV, and Cur PikAIII and PikAIV) containing 0.1mg/mL lysozyme, 0.05mg/mL DNase, and 2mM MgCl2. Cells were lysed by sonication and cleared by centrifugation. The supernatant was loaded onto a 5-mL His trap column (GE Healthcare). Proteins were eluted using a gradient of 15–400 mM imidazole in buffer A over 10 column volumes.
His6-tags were removed by incubation with protease (TEV protease, 2mM dithiothreitol (DTT) for docking domain fusions and all ACP-dd except Avitag-CurG ACP-dd; SUMO hydrolase, 1mM DTT for Cur dd-KS-AT), overnight dialysis in buffer A, and separation from tagged proteins on a His trap column. Proteins were further purified by gel filtration with buffer A (HiLoad 16/60 Superdex S75 for Cur docking domain fusions and Cur ACP-dd; HiPrep 16/60 Sephacryl S300 HR for Cur dd-KS-AT, WT PikAIII and PikAIV, and Cur PikAIII and PikAIV). ACP-dd from CurJ and dd-KS-AT from CurJ and CurK were insoluble.
Protein crystallization
GH dock (native and SeMet), KL dock (native and SeMet L68M), and CurL dd-KS-AT were crystallized by sitting-drop vapor diffusion. GH dock (15 mg/mL) crystallized in 24hr from 3M (NH4)2SO4, 8% glycerol (native and SeMet). KL dock (15 mg/mL in 20mM Tris pH 7.5, 100mM NaCl) crystallized in 1 week in 3.4M (NH4)2SO4, 0.1M Bis-Tris propane pH 6.5 (native) and 4–5 weeks from 3.5M (NH4)2SO4, Bis-Tris propane pH8 (SeMet L68M). CurL dd-KS-AT (10 mg/mL) crystallized in 3 days from 32% PEG 2000, 12% glycerol, 200mM calcium acetate, 100mM Bis-Tris propane pH 6.5. Crystals were grown at 4°C, harvested in loops, cryo-protected in the corresponding well solution with 15% glycerol, and flash cooled in liquid nitrogen.
Data collection and structure determination
Data were collected at the Advanced Photon Source (APS), GM/CA beamline 23ID-D at Argonne National Laboratory (Argonne, IL). All data were processed using HKL2000 (Otwinowski, 1997) (Table 1). The GH dock and KL dock structures were determined by single-wavelength anomalous diffraction using SOLVE (Terwilliger, 2003) and RESOLVE (Wang et al., 2004) to locate selenium atoms, determine initial phases, perform density modification, and build 95% complete initial models (FOM=0.34 for GH dock, FOM=0.30 for KL dock L68M). The CurL dd-KS-AT structure was solved by molecular replacement using Phenix (Adams et al., 2010) with the KS-AT di-domain from DEBS module 3 as the search model (Tang et al., 2007).
Refinement was performed with native data sets (Table 1) with REFMAC5 (Murshudov et al., 1997) of the CCP4 (Collaborative Computational Project, 1994) suite for GH and KL docks and Buster (Bricogne G., 2010) for CurL dd-KS-AT. Coot (Emsley and Cowtan, 2004) was used for model building. Waters were added using the routines in REFMAC5 and Buster and were edited as needed after visual inspection of hydrogen bonding geometries. In the GH dock structure, amino acids 1–30 are CurG 1554–1583, 31–38 are the (Gly3Ser)2 linker, and 39–82 are CurH 1–44. The GH dock structure is complete except for residues 34–36 of (Gly3Ser)2 linker in chain A and the C-terminal residue of chain B. Three residues of the His6-tag not removed by TEV protease are ordered at the N-terminus. In the KL dock structure, amino acids 6–30 are CurK 2208–2232, and 47–73 are CurL 9–35. Residues 1–5, 31–49, and 74–77 of chain A, residues 1–5, 31–46, and 74–77 of chain B, and residues 1–5, 32–46, and 72–77 of chain C are missing in the KL dock structure. The CurL dd-KS-AT structure is complete except for residues 1–2, 465–488, 617–621, and 633–644 of chain A, and residues 1–11, 465–488, 610–647, 702–717, 740–741, 778, 786–804, and 822–830 of chain B. Structures were validated with MolProbity (Davis et al., 2004), sequence alignments were done with MUSCLE (Edgar, 2004), and molecular figures were prepared with PyMOL (http://www.pymol.org/).
FP binding assays
A BODIPY FL C1-IA, N-(4,4-difluoro-5,7-dimethyl-4-bora-3a,4a-diaza-s-indacene-3-yl)methyl) iodoacetamide (Invitrogen) fluorophore was conjugated to the C-terminal cysteine of each Cur ACP-dd construct according to manufacturer’s instructions. Labeled proteins were separated from unreacted fluorophore with a PD-10 column (GE) equilibrated with 10mM HEPES pH 7.4, 150mM NaCl (buffer B) and overnight dialysis in buffer B.
FP binding assays were performed in 384-well black opaque Corning plates in 50μL reaction volume. 10nM of BODIPY-labeled ACP-dd was incubated with varying concentrations of dd-KS-AT (200nM-200μM) in buffer B at room temperature for 30min. Fluorescence polarization measurements were taken with a SpectraMax M5 (Molecular Devices) using 485nm excitation, 538nm emission, and 530nm cutoff filter. Affinities were determined by fit of the data to the equation Y=Bmax*X/(Kd + X) (synergy software).
Octet red binding assays
Avitag-CurG ACP-dd was loaded onto streptavidin biosensors (Forte Bio) for 1,500 seconds. The loaded biosensor tips were quenched with 1μg/μl biocytin for 60 seconds and put into binding buffer (10mM HEPES pH 7.4 and 150mM NaCl) for 600 seconds. Then the loaded tips were put into CurH dd-KS-AT (concentrations ranged from 1–70μM) in binding buffer for 300 seconds for an association step and then put into binding buffer for 300 seconds for a dissociation step. All experiments were performed at 25°C. The increase in biolayer interferometry (BLI) was plotted for each concentration and the data were fit with the equation Y=Bmax*X/(Kd + X) to determine binding affinity.
PikAIII/PikAIV assays
For PikAIII/PikAIV assays a 200μL mixture (1μM PikAIII, 1μM PikAIV, 0.5mM NADP+, 0.5 U/uL glucose-6-phosphate dehydrogenase, 5mM glucose-6-phosphate) was pre-incubated 10 min at room temperature in buffer C (400mM sodium phosphate pH 7.2, 5mM NaCl, 20% glycerol). The reaction was initiated by addition of 20mM methylmalonyl-SNAC, 1mM thiophenol-pentaketide (Hansen et al., 2013), 8mM 2-vinylpyridine, incubated 2hr at room temperature, and quenched by addition of a threefold excess of methanol. The mixture was vortexed, incubated 15min at −20°C, and centrifuged 20min at 14,000 RPM and 4°C. Reactants and products were analyzed by reverse-phase HPLC using a Luna C18(2) (5μm, 250 × 4.6mm) column (Phenomenex) with flow rate 1.5mL/min and a protocol as follows: 5% solvent B (acetonitrile with 0.1% formic acid) for 1min, 5–100% solvent B for 10min, 100% solvent B for 4 min, and 5% solvent B for 2.5min. Solvent A was water with 0.1% formic acid. The elution times of 10-dml and nbl were confirmed with authentic standards. Products were quantitated by the peak areas of 10-dml and nbl and for each combination of chimeric PikAIII and PikAIV, normalized to the values for WT PikAIII and PikAIV. The levels of thiophenol-pentaketide consumed by each reaction were determined based on peak areas normalized to a control, which lacked enzyme.
Supplementary Material
HIGHLIGHTS.
Cyanobacterial PKS docking domains are a new class (Class 2) of docking domains
Novel docking mechanism of Class 2 docking domains directs the ACP towards the KS
Mutation of electrostatic residues alters selectivity of Class 2 docking domains
Class 2 docking domains can effectively replace Class 1 docking domains
Acknowledgments
We thank the staff at the GM/CA beamlines (supported by the NIH National Institute of General Medical Sciences (GM, Y1-GM-1104) and National Cancer Institute (CA, Y1-CO-1020)), Advanced Photon Source (supported by the United States Department of Energy), Argonne National Laboratory. This work was supported by Chemical Biology interface training grant (J.R.W.), Rackham Merit and American Foundation for Pharmaceutical Education pre-doctoral fellowships (D.A.H.), NIH grants GM076477 (D.H.S. and J.L.S.) and DK042303 (J.L.S.), the Hans W. Vahlteich Professorship (to D.H.S.).
Footnotes
Publisher's Disclaimer: This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to our customers we are providing this early version of the manuscript. The manuscript will undergo copyediting, typesetting, and review of the resulting proof before it is published in its final citable form. Please note that during the production process errors may be discovered which could affect the content, and all legal disclaimers that apply to the journal pertain.
References
- Adams PD, Afonine PV, Bunkoczi G, Chen VB, Davis IW, Echols N, Headd JJ, Hung LW, Kapral GJ, Grosse-Kunstleve RW, et al. PHENIX: a comprehensive Python-based system for macromolecular structure solution. Acta Crystallogr D Biol Crystallogr. 2010;66:213–221. doi: 10.1107/S0907444909052925. [DOI] [PMC free article] [PubMed] [Google Scholar]
- Akey DL, Razelun JR, Tehranisa J, Sherman DH, Gerwick WH, Smith JL. Crystal structures of dehydratase domains from the curacin polyketide biosynthetic pathway. Structure. 2010;18:94–105. doi: 10.1016/j.str.2009.10.018. [DOI] [PMC free article] [PubMed] [Google Scholar]
- Aparicio JF, Caffrey P, Marsden AF, Staunton J, Leadlay PF. Limited proteolysis and active-site studies of the first multienzyme component of the erythromycin-producing polyketide synthase. J Biol Chem. 1994;269:8524–8528. [PubMed] [Google Scholar]
- Beck BJ, Aldrich CC, Fecik RA, Reynolds KA, Sherman DH. Iterative chain elongation by a pikromycin monomodular polyketide synthase. J Am Chem Soc. 2003;125:4682–4683. doi: 10.1021/ja029974c. [DOI] [PubMed] [Google Scholar]
- Bricogne G, BE, Brandl M, Flensburg C, Keller P, Paciorek W, Roversi P, Sharff A, Smart OS, Vonrhein C, Womack TO. BUSTER. Global Phasing; Cambridge, UK: 2010. Version 2.9. [Google Scholar]
- Broadhurst RW, Nietlispach D, Wheatcroft MP, Leadlay PF, Weissman KJ. The structure of docking domains in modular polyketide synthases. Chem Biol. 2003;10:723–731. doi: 10.1016/s1074-5521(03)00156-x. [DOI] [PubMed] [Google Scholar]
- Buchholz TJ, Geders TW, Bartley FE, 3rd, Reynolds KA, Smith JL, Sherman DH. Structural basis for binding specificity between subclasses of modular polyketide synthase docking domains. ACS Chem Biol. 2009;4:41–52. doi: 10.1021/cb8002607. [DOI] [PMC free article] [PubMed] [Google Scholar]
- Chang Z, Sitachitta N, Rossi JV, Roberts MA, Flatt PM, Jia J, Sherman DH, Gerwick WH. Biosynthetic pathway and gene cluster analysis of curacin A, an antitubulin natural product from the tropical marine cyanobacterium Lyngbya majuscula. J Nat Prod. 2004;67:1356–1367. doi: 10.1021/np0499261. [DOI] [PubMed] [Google Scholar]
- Collaborative Computational Project N. The CCP4 suite: programs for protein crystallography. Acta Crystallogr D Biol Crystallogr. 1994;50:760–763. doi: 10.1107/S0907444994003112. [DOI] [PubMed] [Google Scholar]
- Davis IW, Murray LW, Richardson JS, Richardson DC. MOLPROBITY: structure validation and all-atom contact analysis for nucleic acids and their complexes. Nucleic Acids Res. 2004;32:W615–619. doi: 10.1093/nar/gkh398. [DOI] [PMC free article] [PubMed] [Google Scholar]
- DelProposto J, Majmudar CY, Smith JL, Brown WC. Mocr: a novel fusion tag for enhancing solubility that is compatible with structural biology applications. Protein Expr Purif. 2009;63:40–49. doi: 10.1016/j.pep.2008.08.011. [DOI] [PMC free article] [PubMed] [Google Scholar]
- Donadio S, Staver MJ, McAlpine JB, Swanson SJ, Katz L. Modular organization of genes required for complex polyketide biosynthesis. Science. 1991;252:675–679. doi: 10.1126/science.2024119. [DOI] [PubMed] [Google Scholar]
- Donnelly MI, Zhou M, Millard CS, Clancy S, Stols L, Eschenfeldt WH, Collart FR, Joachimiak A. An expression vector tailored for large-scale, high-throughput purification of recombinant proteins. Protein Expr Purif. 2006;47:446–454. doi: 10.1016/j.pep.2005.12.011. [DOI] [PMC free article] [PubMed] [Google Scholar]
- Edgar RC. MUSCLE: multiple sequence alignment with high accuracy and high throughput. Nucleic Acids Res. 2004;32:1792–1797. doi: 10.1093/nar/gkh340. [DOI] [PMC free article] [PubMed] [Google Scholar]
- Emsley P, Cowtan K. Coot: model-building tools for molecular graphics. Acta Crystallogr D Biol Crystallogr. 2004;60:2126–2132. doi: 10.1107/S0907444904019158. [DOI] [PubMed] [Google Scholar]
- Fischbach MA, Walsh CT. Assembly-line enzymology for polyketide and nonribosomal Peptide antibiotics: logic, machinery, and mechanisms. Chem Rev. 2006;106:3468–3496. doi: 10.1021/cr0503097. [DOI] [PubMed] [Google Scholar]
- Floss HG. Combinatorial biosynthesis--potential and problems. J Biotechnol. 2006;124:242–257. doi: 10.1016/j.jbiotec.2005.12.001. [DOI] [PMC free article] [PubMed] [Google Scholar]
- Gokhale RS, Tsuji SY, Cane DE, Khosla C. Dissecting and exploiting intermodular communication in polyketide synthases. Science. 1999;284:482–485. doi: 10.1126/science.284.5413.482. [DOI] [PubMed] [Google Scholar]
- Hansen DA, Rath CM, Eisman EB, Narayan AR, Kittendorf JD, Mortison JD, Yoon YJ, Sherman DH. Biocatalytic Synthesis of Pikromycin, Methymycin, Neomethymycin, Novamethymycin, and Ketomethymycin. J Am Chem Soc. 2013 doi: 10.1021/ja404134f. [DOI] [PMC free article] [PubMed] [Google Scholar]
- Keatinge-Clay A. Crystal structure of the erythromycin polyketide synthase dehydratase. J Mol Biol. 2008;384:941–953. doi: 10.1016/j.jmb.2008.09.084. [DOI] [PMC free article] [PubMed] [Google Scholar]
- Kittendorf JD, Beck BJ, Buchholz TJ, Seufert W, Sherman DH. Interrogating the molecular basis for multiple macrolactone ring formation by the pikromycin polyketide synthase. Chem Biol. 2007;14:944–954. doi: 10.1016/j.chembiol.2007.07.013. [DOI] [PMC free article] [PubMed] [Google Scholar]
- Kittendorf JD, Sherman DH. Developing tools for engineering hybrid polyketide synthetic pathways. Curr Opin Biotechnol. 2006;17:597–605. doi: 10.1016/j.copbio.2006.09.005. [DOI] [PubMed] [Google Scholar]
- Kumar P, Li Q, Cane DE, Khosla C. Intermodular communication in modular polyketide synthases: structural and mutational analysis of linker mediated protein-protein recognition. J Am Chem Soc. 2003;125:4097–4102. doi: 10.1021/ja0297537. [DOI] [PubMed] [Google Scholar]
- Menzella HG, Carney JR, Santi DV. Rational design and assembly of synthetic trimodular polyketide synthases. Chem Biol. 2007;14:143–151. doi: 10.1016/j.chembiol.2006.12.002. [DOI] [PubMed] [Google Scholar]
- Menzella HG, Reeves CD. Combinatorial biosynthesis for drug development. Curr Opin Microbiol. 2007;10:238–245. doi: 10.1016/j.mib.2007.05.005. [DOI] [PubMed] [Google Scholar]
- Menzella HG, Reid R, Carney JR, Chandran SS, Reisinger SJ, Patel KG, Hopwood DA, Santi DV. Combinatorial polyketide biosynthesis by de novo design and rearrangement of modular polyketide synthase genes. Nat Biotechnol. 2005;23:1171–1176. doi: 10.1038/nbt1128. [DOI] [PubMed] [Google Scholar]
- Murshudov GN, Vagin AA, Dodson EJ. Refinement of macromolecular structures by the maximum-likelihood method. Acta Crystallogr D Biol Crystallogr. 1997;53:240–255. doi: 10.1107/S0907444996012255. [DOI] [PubMed] [Google Scholar]
- Newman DJ, Cragg GM. Natural products as sources of new drugs over the last 25 years. J Nat Prod. 2007;70:461–477. doi: 10.1021/np068054v. [DOI] [PubMed] [Google Scholar]
- Otwinowski Z, Minor V. Processing of X-ray Diffraction Data Collected in Oscillation Mode. Methods Enzymol. 1997;276:307–326. doi: 10.1016/S0076-6879(97)76066-X. [DOI] [PubMed] [Google Scholar]
- Pfeifer BA, Admiraal SJ, Gramajo H, Cane DE, Khosla C. Biosynthesis of complex polyketides in a metabolically engineered strain of E. coli. Science. 2001;291:1790–1792. doi: 10.1126/science.1058092. [DOI] [PubMed] [Google Scholar]
- Reeves CD, Ward SL, Revill WP, Suzuki H, Marcus M, Petrakovsky OV, Marquez S, Fu H, Dong SD, Katz L. Production of hybrid 16-membered macrolides by expressing combinations of polyketide synthase genes in engineered Streptomyces fradiae hosts. Chem Biol. 2004;11:1465–1472. doi: 10.1016/j.chembiol.2004.08.019. [DOI] [PubMed] [Google Scholar]
- Scholle MD, Collart FR, Kay BK. In vivo biotinylated proteins as targets for phage-display selection experiments. Protein Expr Purif. 2004;37:243–252. doi: 10.1016/j.pep.2004.05.012. [DOI] [PubMed] [Google Scholar]
- Sherman DH. The Lego-ization of polyketide biosynthesis. Nat Biotechnol. 2005;23:1083–1084. doi: 10.1038/nbt0905-1083. [DOI] [PubMed] [Google Scholar]
- Staunton J, Caffrey P, Aparicio JF, Roberts GA, Bethell SS, Leadlay PF. Evidence for a double-helical structure for modular polyketide synthases. Nat Struct Biol. 1996;3:188–192. doi: 10.1038/nsb0296-188. [DOI] [PubMed] [Google Scholar]
- Tang Y, Chen AY, Kim CY, Cane DE, Khosla C. Structural and mechanistic analysis of protein interactions in module 3 of the 6-deoxyerythronolide B synthase. Chem Biol. 2007;14:931–943. doi: 10.1016/j.chembiol.2007.07.012. [DOI] [PMC free article] [PubMed] [Google Scholar]
- Tang Y, Kim CY, Mathews II, Cane DE, Khosla C. The 2.7-Å crystal structure of a 194-kDa homodimeric fragment of the 6-deoxyerythronolide B synthase. Proc Natl Acad Sci U S A. 2006;103:11124–11129. doi: 10.1073/pnas.0601924103. [DOI] [PMC free article] [PubMed] [Google Scholar]
- Terwilliger TC. Automated main-chain model building by template matching and iterative fragment extension. Acta Crystallogr D Biol Crystallogr. 2003;59:38–44. doi: 10.1107/S0907444902018036. [DOI] [PMC free article] [PubMed] [Google Scholar]
- Thattai M, Burak Y, Shraiman BI. The origins of specificity in polyketide synthase protein interactions. PLoS Comput Biol. 2007;3:1827–1835. doi: 10.1371/journal.pcbi.0030186. [DOI] [PMC free article] [PubMed] [Google Scholar]
- Tsuji SY, Cane DE, Khosla C. Selective protein-protein interactions direct channeling of intermediates between polyketide synthase modules. Biochemistry. 2001;40:2326–2331. doi: 10.1021/bi002463n. [DOI] [PubMed] [Google Scholar]
- Walsh CT. Combinatorial biosynthesis of antibiotics: challenges and opportunities. Chembiochem. 2002;3:125–134. doi: 10.1002/1439-7633(20020301)3:2/3<124::AID-CBIC124>3.0.CO;2-J. [DOI] [PubMed] [Google Scholar]
- Wang JW, Chen JR, Gu YX, Zheng CD, Jiang F, Fan HF, Terwilliger TC, Hao Q. SAD phasing by combination of direct methods with the SOLVE/RESOLVE procedure. Acta Crystallogr D Biol Crystallogr. 2004;60:1244–1253. doi: 10.1107/S0907444904010674. [DOI] [PubMed] [Google Scholar]
- Weeks SD, Drinker M, Loll PJ. Ligation independent cloning vectors for expression of SUMO fusions. Protein Expr Purif. 2007;53:40–50. doi: 10.1016/j.pep.2006.12.006. [DOI] [PMC free article] [PubMed] [Google Scholar]
- Weissman KJ. Single amino acid substitutions alter the efficiency of docking in modular polyketide biosynthesis. Chembiochem. 2006a;7:1334–1342. doi: 10.1002/cbic.200600185. [DOI] [PubMed] [Google Scholar]
- Weissman KJ. The structural basis for docking in modular polyketide biosynthesis. Chembiochem. 2006b;7:485–494. doi: 10.1002/cbic.200500435. [DOI] [PubMed] [Google Scholar]
- Wu N, Cane DE, Khosla C. Quantitative analysis of the relative contributions of donor acyl carrier proteins, acceptor ketosynthases, and linker regions to intermodular transfer of intermediates in hybrid polyketide synthases. Biochemistry. 2002;41:5056–5066. doi: 10.1021/bi012086u. [DOI] [PubMed] [Google Scholar]
- Wu N, Tsuji SY, Cane DE, Khosla C. Assessing the balance between protein-protein interactions and enzyme-substrate interactions in the channeling of intermediates between polyketide synthase modules. J Am Chem Soc. 2001;123:6465–6474. doi: 10.1021/ja010219t. [DOI] [PubMed] [Google Scholar]
- Xue Y, Zhao L, Liu HW, Sherman DH. A gene cluster for macrolide antibiotic biosynthesis in Streptomyces venezuelae: architecture of metabolic diversity. Proc Natl Acad Sci U S A. 1998;95:12111–12116. doi: 10.1073/pnas.95.21.12111. [DOI] [PMC free article] [PubMed] [Google Scholar]
- Yan J, Gupta S, Sherman DH, Reynolds KA. Functional dissection of a multimodular polypeptide of the pikromycin polyketide synthase into monomodules by using a matched pair of heterologous docking domains. Chembiochem. 2009;10:1537–1543. doi: 10.1002/cbic.200900098. [DOI] [PMC free article] [PubMed] [Google Scholar]
- Zheng J, Fage CD, Demeler B, Hoffman DW, Keatinge-Clay AT. The missing linker: A dimerization motif located within polyketide synthase modules. ACS Chem Biol. 2013 doi: 10.1021/cb400047s. [DOI] [PMC free article] [PubMed] [Google Scholar]
Associated Data
This section collects any data citations, data availability statements, or supplementary materials included in this article.