Crystal structure of the putative cyclase IdmH from the indanomycin nonribosomal peptide synthase/polyketide synthase
aAstbury Centre for Structural Molecular Biology and School of Molecular and Cellular Biology, University of Leeds, Leeds LS2 9JT, England, and bSchool of Biochemistry, University of Bristol, University Walk, Bristol BS8 1TD, England
*Correspondence e-mail: firstname.lastname@example.org, email@example.com
Indanomycin is biosynthesized by a hybrid nonribosomal peptide synthase/polyketide synthase (NRPS/PKS) followed by a number of `tailoring' steps to form the two ring systems that are present in the mature product. It had previously been hypothesized that the indane ring of indanomycin was formed by the action of IdmH using a Diels–Alder reaction. Here, the crystal structure of a selenomethionine-labelled truncated form of IdmH (IdmH-Δ99–107) was solved using single-wavelength anomalous dispersion (SAD) phasing. This truncated variant allows consistent and easy crystallization, but importantly the structure was used as a search model in molecular replacement, allowing the full-length IdmH structure to be determined to 2.7 Å resolution. IdmH is a homodimer, with the individual protomers consisting of an α+β barrel. Each protomer contains a deep hydrophobic pocket which is proposed to constitute the active site of the enzyme. To investigate the reaction catalysed by IdmH, 88% of the backbone NMR resonances were assigned, and using chemical shift perturbation of [15N]-labelled IdmH it was demonstrated that indanomycin binds in the active-site pocket. Finally, combined quantum mechanical/molecular mechanical (QM/MM) modelling of the IdmH reaction shows that the active site of the enzyme provides an appropriate environment to promote indane-ring formation, supporting the assignment of IdmH as the key Diels–Alderase catalysing the final step in the biosynthesis of indanomycin through a similar mechanism to other recently characterized Diels–Alderases involved in polyketide-tailoring reactions. An animated Interactive 3D Complement (I3DC) is available in Proteopedia at https://proteopedia.org/w/Journal:IUCrJ:S2052252519012399.
Polyketides and nonribosomal peptides are two major classes of natural products which give rise to nearly one third of the current pharmacopoeia (Newman & Cragg, 2012). They exhibit high structural diversity and are proven to be excellent therapeutics, but there is an increasing interest in diversifying current natural product libraries to produce analogues with improved or novel biological activity (Cummings et al., 2014). Natural products classed as polyketides, nonribosomal peptides and hybrids of both are often biosynthesized by giant, complex enzymes known as polyketide synthases (PKS) and nonribosomal peptide synthetases (NRPS). Both systems consist of a range of domains which incorporate a number of starter and extender units to build a linear structure. In the case of polyketides, the linear enzyme products can then be further tailored by various enzymes, including cyclases, oxidases, reductases and methylases, to yield a diverse array of bioactive compounds (Olano et al., 2010).
Indanomycin (1) is an antibiotic from the pyrroloketoindane family which is known to act as an effective ionophoric agent against Gram-positive bacteria (Dutton et al., 1995). Indanomycin is synthesized in Streptomyces antibioticus NRRL 8167 by a hybrid nonribosomal peptide synthetase/polyketide synthase (NRPS/PKS; Li et al., 2009). The NRPS portion of the pathway is proposed to generate a pyrrole moiety from L-proline, which is then extended by the sequential addition of malonyl-CoA, methylmalonyl-CoA and ethylmalonyl-CoA building blocks by ten predicted PKS modules to yield the linear nonribosomal peptide–polyketide hybrid natural product 2 (Fig. 1). At least two cyclization reactions are then needed to generate the tetrahydropyran and tetrahydroindane rings of indanomycin (1; Fig. 1). The former could be installed while the polyketide is still tethered to the PKS by the cyclase domain, Cyc11, in the terminal, 11th PKS module. Indane-ring formation has been postulated to be catalysed by a separate indane cyclase, IdmH (Li et al., 2009; Rommel et al., 2011), using a Diels–Alder [4+2] intramolecular cyclization (Fig. 1). The feasibility of indane-ring formation in indanomycin proceeding via a Diels–Alder reaction was first shown in its complete chemical synthesis (Edwards et al., 1984), but to date it has not been possible to characterize the activity of IdmH in molecular detail.
The proposed biosynthetic cycloaddition would require the appropriate diene and dienophile within the intermediate. The indanomycin PKS harbours catalytic domains which would produce the suitable diene intermediate for the cycloaddition to be feasible. However, the required dienophile would not be generated by straightforward use of the catalytic domains present in the indanomycin PKS/NRPS, since the second PKS module is predicted to install a hydroxyl group at C19 (Fig. 1, compound 2; Rommel et al., 2011). To form the appropriate dienophile, dehydration of the alcohol at C19 is therefore needed, either while the linear polyketide is tethered to the synthase or after it has been released. One suggested possibility is that a dehydratase (DH) domain, for example DH3, from a neighbouring module might catalyse the required dehydration in a manner similar to that used in epothilone biosynthesis (Tang et al., 2004; Li et al., 2009).
Irrespective of the activity required to dehydrate C19, IdmH has been proposed to catalyse the Diels–Alder reaction which would lead to the final indane-ring formation in indanomycin. In support of this idea, an idmH deletion mutant of S. antibioticus exhibited both a significant reduction in indanomycin yield and the production of a previously unobserved linear tetraene alternative product (Rommel et al., 2011). Notably, indanomycin-production levels were restored when the idmH gene was introduced to S. antibioticus in trans (Rommel et al., 2011). IdmH is thus thought to be the key cyclase responsible for the formation of the indane ring and hence mature indanomycin. IdmH is a protein consisting of 145 amino acids (Supplementary Fig. S1) exhibiting sequence similarity to a number of natural product-modifying enzymes. These include the epoxide hydroxylase MonBI from S. cinnamonensis (50% identity over 14% of the sequence; Minami et al., 2014) and the polyketide cyclase SnoaL from S. nogalater (27% identity over 91% of the sequence; Sultana et al., 2004) (Supplementary Fig. S2). Both homologues exhibit an α+β barrel fold and are involved in ring closure and aromatic ring hydroxylations in the monensin and nogalamycin biosynthetic pathways, respectively. Beyond polyketide-modifying enzymes, other α+β barrel proteins include Δ5-3-ketosteroid isomerase and scylatone dehydratase (Ha et al., 2001; Lundqvist et al., 1994), showing that nature has made broad use of this fold in many biosynthetic pathways. Given the diversity of reactions catalysed by enzymes displaying this fold, it is challenging to predict the function of IdmH based on sequence similarity alone.
As a step towards elucidating the underlying enzymatic mechanism of indanomycin maturation, we set out to determine the three-dimensional structure of IdmH. Here, we present crystal structures of IdmH and a deletion variant, IdmH-Δ99–107, at 2.7 and 2 Å resolution, respectively. Structural analysis shows that IdmH has a similar fold to that of the putative hydroxylases SnoAL2 and AcIR (Beinker et al., 2006) and the polyketide cyclases SnoaL (Sultana et al., 2004) and AknH (Kallio et al., 2006). Apparent differences in the putative active site, however, suggest that this enzyme catalyses a distinct reaction from that of its closest structural homologues. We therefore utilized NMR spectroscopy and in silico modelling [including quantum mechanical/molecular mechanical (QM/MM) reaction simulations] to probe product binding and the likely reaction catalysed by IdmH, with our analyses supporting the notion that this enzyme is indeed a Diels–Alderase.
The gene sequence encoding the full-length idmH gene (ACN69984.1) was amplified by PCR using the genomic DNA from S. antibioticus (strain NRRL 8167) as a template, and the forward and reverse primers 5′-ctg gtg ccg cgc ggc agc cat ATG GCT CAT CAG CCT TCG-3′ and 5′-tcc acc agt cat gct agc caT CAC AGG GAC GCC TTC AC-3′, respectively. The amplified sequence contained overhang sequences (in lower case letters) which facilitated ligation to an NdeI-linearized pET-28(a)+ vector using the NEBuilder HiFi DNA-assembly kit (NEB). The generated plasmid also encoded an N-terminal hexahistidine tag on the protein and a thrombin cleavage site. The gene insert was confirmed by sequencing with T7 and T7term universal primers before transformation into Escherichia coli strain BL21(DE3) cells for overexpression. Site-directed mutagenesis to generate the deletion variant IdmH-Δ99–107 was performed using wild-type pET-28_IdmH as a template and the forward and reverse primers 5′-CGC GAC CGC GAG GGG TGG-3′ and 5′-GGA GTG CGT TCC CCG TGC-3′, respectively. The site-directed mutagenesis PCR product was treated with KLD Enzyme Mix (NEB) before transformation into E. coli strain 5-α competent cells.
Recombinant IdmH was expressed in E. coli strain BL21(DE3) cells harbouring the pET-28_IdmH plasmid. The cells were grown in 2TY medium containing 50 µg ml−1 kanamycin at 37°C to an OD600 of 0.4 before lowering the temperature to 15°C. Expression was then induced with 0.4 mM isopropyl β-D-1-thiogalactopyranoside (IPTG) at an OD600 of 1.0 and the cells were allowed to grow overnight. The cells were harvested by centrifugation at 6000g and subsequently resuspended in buffer A (50 mM Tris–HCl pH 7.4, 0.5 M NaCl) containing 20 mM imidazole, 0.1 mg ml−1 lysozyme, 0.05 mg ml−1 DNase and 2 mM MgCl2. After disrupting the cells with a high-pressure press (Constant Systems), the cell debris was removed by centrifugation at 50 000g. The supernatant was then loaded onto a 5 ml HisTrap HP column (GE Healthcare, USA) which had been pre-equilibrated with buffer A. Following sample loading, the column was washed with ten volumes of buffer A before the target protein was eluted from the column with buffer A supplemented with 200 mM imidazole. The eluted recombinant protein was desalted into buffer B (10 mM Tris–HCl pH 7.2, 50 mM NaCl) using a PD-10 desalting column (GE Healthcare, USA) and the hexahistidine tag was then removed by incubating the sample with 1 U thrombin (GE Healthcare, USA) per milligram of recombinant IdmH overnight at 4°C. Thrombin was removed by loading the protein mixture onto a 5 ml HiTrap Q column (GE Healthcare, USA) equilibrated with buffer B. The column was washed with five column volumes of buffer B before a linear gradient of increasing NaCl concentration was applied to elute the bound proteins. 2.5 ml fractions were collected across the gradient, which were analysed by SDS–PAGE. Fractions containing the protein were pooled, concentrated and applied onto a size-exclusion chromatography column (HiLoad 26/600 Superdex 75 pg, GE Healthcare, USA) pre-equilibrated with buffer B. After a void volume of 40 ml, 1 ml fractions were collected. These were once more analysed using SDS–PAGE and the fractions containing the most highly pure IdmH were combined to form the final sample. This was concentrated to 20 mg ml−1 (as determined from the A280 with an extinction coefficient of 17 990 M−1 cm−1) using a centrifugal filter concentration device (Vivaspin 20; 10 000 molecular-weight cutoff). Finally, samples were flash-frozen in liquid nitrogen and stored at −80°C until use. The IdmH-Δ99–107 variant was expressed and purified in exactly the same way.
Production of selenomethionine-labelled IdmH-Δ99–107 was achieved by cultivation of the methionine-auxotrophic E. coli strain B834 (DE3) cells in M9 minimal medium supplemented with BME vitamins (Sigma–Aldrich) and 50 µg ml−1 L-(+)-selenomethionine (Anatrace). All expression and purification steps were performed as for the native protein. Incorporation of selenomethionine was confirmed by LC/MS analysis.
Initial crystallization conditions for full-length IdmH were identified with The JCSG Core Suites I–IV (Qiagen) using Formulatrix NT8 robotic mixing of protein at 20 mg ml−1 with crystallization solution at 1:1, 2:1 and 1:2 volumetric ratios in separate drops. IdmH crystal hit conditions were subsequently optimized with respect to pH and precipitant concentration using hanging-drop vapour diffusion, giving the crystals used for data collection in 0.2 M calcium acetate, 0.1 M MES pH 5.5, 18.6%(w/v) polyethylene glycol 8000. The crystals were cryoprotected by soaking them in mother-liquor solution containing 20%(v/v) glycerol for about 30 s before flash-cooling them by plunging them into liquid nitrogen. X-ray diffraction data for full-length IdmH were collected on beamline I04 at Diamond Light Source.
The structure of IdmH could not be determined by molecular replacement using these data and the structure was therefore solved using selenomethionine-labelled IdmH-Δ99–107. Crystals were obtained in 3.6–4 M sodium formate, 10–15%(v/v) glycerol using the same approach as described for the wild-type enzyme. The crystals were flash-cooled in liquid nitrogen for data collection, this time in the absence of cryoprotectant. X-ray diffraction data for selenomethionine-labelled IdmH-Δ99–107 were recorded on MASSIF-1 at the European Synchrotron Radiation Facility (Bowler et al., 2015, 2016; Svensson et al., 2015, 2018; Nurizzo et al., 2016), while native IdmH-Δ99–107 data were recorded on beamline I03 at Diamond Light Source.
All diffraction data were indexed and integrated using XDS (Kabsch, 2010) before subsequent scaling in AIMLESS and data processing using the CCP4i2 graphical user interface to CCP4 (Potterton et al., 2018). The structure of IdmH-Δ99–107 was determined to 2 Å resolution by single-wavelength anomalous dispersion (SAD) phasing using the SHELX pipeline followed by density modification using Parrot and automated model building in Buccaneer. Iterative rounds of manual model building and refinement were then performed in Coot and REFMAC5, respectively (Cowtan, 2006, 2010, 2012; Murshudov et al., 2011; Sheldrick, 2010). The structure of wild-type IdmH was solved by molecular replacement using a single monomer of IdmH-Δ99–107 as a search model in Phaser (McCoy, 2007; McCoy et al., 2007). The final model was once more generated following iterative rounds of manual model building in Coot and refinement using REFMAC5 imposing local noncrystallographic symmetry (NCS) restraints throughout (Emsley & Cowtan, 2010; Murshudov et al., 2011). The model was validated using MolProbity (Chen et al., 2010). All data-collection and structure-refinement statistics are listed in Table 1.
950 MHz Bruker Ascend Aeon and 750 MHz Oxford NMR spectrometers equipped with cryogenically cooled triple-resonance probes were used for all NMR experiments. Spectra were recorded at 25°C in NMR buffer [20 mM Tris pH 7.3, 50 mM NaCl, 1 mM DTT, 0.02%(w/v) NaN3, 5%(v/v) D2O]. 2D NMR experiments were carried out using Wilmad NMR tubes (Precision, limit 600 MHz frequency), while Shigemi NMR tubes were used for triple-resonance experiments.
Triple-labelled protein was obtained by growing E. coli BL21(DE3) cells containing the wild-type IdmH expression plasmid in M9 medium prepared in D2O and supplemented with [13C]-D-glucose and 15NH4Cl (Cambridge Isotope Laboratories). Labelled protein was purified as described for the wild type, except that no thrombin-cleavage step was carried out. A set of 3D BEST-TROSY backbone resonance assignment spectra [HNCA, HN(CO)CA, HNCO, HN(CA)CB and HN(CO)CACB] were recorded at 750 MHz, while a 3D BEST-TROSY HN(CA)CO spectrum was recorded at 950 MHz, using pulse sequences (Solyom et al., 2013) obtained from the Institut de Biologie Structurale (IBS), Grenoble, France for [15N,13C,2H]-labelled IdmH at a concentration of 0.8 mM. Non-uniform sampling was used to speed up the acquisition time and obtain high-resolution spectra.
The raw spectra were pre-processed with NMRPipe and NMRDraw (Delaglio et al., 1995). Non-uniform sampled data were reconstructed using the MddNMR suite (Jaravine & Orekhov, 2006; Jaravine et al., 2008). Backbone nuclei resonance assignment was performed using CCPNmr Analysis (Vranken et al., 2005) and AutoAssign (Zimmerman et al., 1997).
2.4.2. Chemical shift perturbation studies
15N-labelled protein was prepared by growing E. coli strain BL21(DE3) containing the wild-type idmH expression plasmid in M9 minimal medium using 15NH4Cl as the sole nitrogen source. 1H–15N HSQC-TROSY spectra were collected at an IdmH concentration of 0.2 mM with varying concentrations of indanomycin (0–200 µM) at 25°C at a field strength of 750 MHz. The raw data were pre-processed using TopSpin (Bruker). The NMR spectra were analysed using CCPNmr Analysis (Vranken et al., 2005). The minimal chemical shift perturbation was calculated from 1H and 15N chemical shift differences using 0.154 as a scaling factor for 15N shift changes and calculating the closest chemical shift distance between peaks in the spectra with and without indanomycin (Williamson, 2013).
Molecular docking of indanomycin (1) and an analogue truncated at C8 was performed using AutoDock Vina 1.1.2 (Trott & Olson, 2010). AutoDockTools 1.5.4 was used to merge the nonpolar hydrogens and define all formally single bonds as rotatable bonds for docking. Subsequently, docking was performed with AutoDock Vina using a grid of 25 × 20 × 25 Å centred on the active-site cavity (in either chain A or chain B of the crystal structure). Three top-scoring poses in different orientations were selected by filtering out highly similar poses and poses that did not place the indane ring inside the cavity. Subsequently, the protein–product poses (with the complete indanomycin) were prepared for molecular-dynamics simulation with AmberTools17 (Case et al., 2017) using the Enlighten protocols (see https://github.com/marcvanderkamp/enlighten). For one of the poses, a different hydrogen-bonding network was obtained by `flipping' the Asn117 side chain. Simulations were thus performed on four different initial protein–product complexes. The PREP protocol was used to prepare the protein–ligand system for simulation, which involved (i) adding hydrogens [all residues in their standard protonation states in line with prediction by PROPKA 3.1 (Olsson et al., 2011; Søndergaard et al., 2011); His44 singly protonated on Hδ1 with all other His singly protonated on H∊2 as predicted by the AmberTools program reduce], (ii) parameterization of indanomycin using AM1-BCC partial charges and GAFF parameters assigned by Antechamber (Wang et al., 2004, 2006) and (iii) solvation of the active site with TIP3P water molecules (in a 20 Å sphere centred on the N atom of indanomycin, in addition to crystallographically determined water molecules). After a brief structural optimization (using the STRUCT protocol, which involves simulated annealing and energy minimization), a molecular-dynamics simulation of 200 or 300 ps was performed for each pose (using the DYNAM protocol at a constant temperature of 298 K).
Snapshots from molecular-dynamics simulations between 100 and 300 ps were used to start approximate QM/MM umbrella sampling simulations of the reverse Diels–Alder reaction (at least 10 ps apart) using the same protocol as used previously (Byrne et al., 2016) using DFTB2 (Elstner et al., 1998) for indanomycin (QM region) and the Amber force field ff14SB (Maier et al., 2015) for the enzyme environment. In short, the reaction coordinate used to follow the Diels–Alder reaction was the distance between the centre of mass of the dienophile carbons (C19 and C20) and the centre of mass of the diene carbons to which they bond (C15 and C12). By using this reaction-coordinate definition, no bias is applied as to which carbon–carbon bond is formed or broken first (or indeed whether or not the bonds are formed/broken synchronously). Umbrella sampling was performed from reaction coordinate value 1.2 to value 3.8 Å (in steps of 0.1 Å, with additional windows at 2.15, 2.25 and 2.35 Å). The same conditions were used as in the DYNAM protocol, apart from using a 1 fs timestep (instead of 2 fs). Snapshots were saved every 0.5 ps. A restraint of 100 kcal mol−1 Å−1 and 2 ps of simulation was used for each umbrella sampling window. Reaction-coordinate values were recorded every 1 fs and used for input to the Weighted-Histogram Analysis Method (WHAM; Grossfield, 2013), resulting in the potential of mean force (free energy) along the reaction coordinate (bin width of 0.05 Å). For each docking pose, 11 QM/MM reaction simulations were performed, each starting from a different snapshot. The QM/MM protocols are designed to limit computational resources and are thus approximate. Our aim is not to analyse the reaction in detail, but to distinguish between different possible reactive binding poses. For example, the use of a single, combined reaction coordinate (obtaining a 1D-PMF) is less accurate than a 2D-PMF with sampling along two reaction coordinates representing the carbon–carbon bonds formed [see, for example, Świderek & Moliner (2016), where this is discussed in detail]. With the combined reaction coordinate used here, reasonable sampling is obtained along the minimum free-energy region indicated in a 2D-PMF (see Supplementary Fig. S3). Although longer sampling (for example 20 ps per window after a 2 ps equilibration) affects the free-energy barriers for simulations, we have tested that this does not significantly change the differences in the barrier between different poses (similar to as reported by Hirvonen et al., 2019). To obtain representative snapshots for the approximate transition state, the centroid of the highest populated cluster was used, as obtained with hierarchical agglomerative clustering on the r.m.s.d. of the indanomycin atoms after alignment on the protein backbone on the structures sampled in the windows with reaction-coordinate restraints at 2.2 and 2.3 Å.
The N-terminally hexahistidine-tagged wild-type IdmH was successfully expressed in E. coli strain BL21(DE3) cells and purified to homogeneity, and the hexahistidine tag was removed by thrombin cleavage. Size-exclusion chromatography and LC/MS analysis (Supplementary Fig. S4) suggested a dimeric oligomeric state for this protein. Initial attempts at crystallization yielded small needle-like crystals, but optimization of these initial conditions produced larger rod-shaped crystals. X-ray diffraction data were collected to 2.7 Å resolution, but attempts to determine the structure by molecular replacement were unsuccessful, presumably owing to the lack of a suitable search model and the large number of molecules in the asymmetric unit as estimated using MATTPROB (Kantardjieff & Rupp, 2003).
To improve crystallization, a number of variants were produced to reduce the surface entropy (Goldschmidt et al., 2007) or to truncate predicted surface loops (Roy et al., 2010; Supplementary Fig. S1 and Table S1). One variant, containing a deletion of residues 99–107, resulted in large, easily reproducible crystals in the cubic space group F23 [Supplementary Fig. S5(b)] from new crystallization conditions. This protein was also produced labelled with selenomethionine and crystallized under identical conditions, allowing the crystal structure of IdmH-Δ99–107 to be determined to 2 Å resolution using selenium SAD phasing. The asymmetric unit comprises one copy of the IdmH-Δ99–107 dimer with the entire polypeptide chain successfully modelled into density, except for the N- and C-terminal regions. Both the selenomethionine-labelled protein structure and the isomorphous native structure of IdmH-Δ99–107 were modelled independently (Table 1), giving almost identical structures that superpose with an r.m.s.d. of 0.12 Å over 257 amino acids. The remainder of the structural description will focus on the unlabelled native protein as this should best represent the natural form of the enzyme. With this structure in hand, the crystal structure of wild-type IdmH was determined by molecular replacement using a single subunit of IdmH-Δ99–107 as a search model (Fig. 2). Electron-density maps were well defined for all ten protomers in the asymmetric unit, allowing all chains to be modelled, with the exception of some residues at the extreme N- and C-termini (residues in the ranges 1–6 and 143–148). The final model was refined at 2.7 Å resolution with an Rwork of 0.21 and an Rfree of 0.24. Both models were checked with MolProbity, revealing a high-quality model (MolProbity scores of 1.00 and 0.91 for wild-type IdmH and IdmH-Δ99–107, respectively) with no Ramachandran outliers.
The ten molecules in the asymmetric unit for the wild-type protein form five almost identical dimers. The following discussion of the IdmH structure will therefore focus on the dimer formed by chains A and B unless otherwise stated, as this represents the most well ordered dimer based on an analysis of the overall B factors of the chains. The structure of IdmH reveals a homodimer, with individual protomers consisting of a core α+β barrel. Each subunit is composed of a five-stranded mixed β-sheet with four strands (β2–β5) continuous in sequence and an additional β-strand (β1), considerably shorter in length, provided later in the polypeptide chain (Fig. 2). The β-sheet of each subunit is curved and forms the central part of a distorted α+β barrel. The barrel is completed by three α-helices (α1–α3) on the periphery of the dimer and a fourth short helix (α4) completing the fold. The dimer assembly is driven by hydrophobic interactions between the β-sheets from the two subunits, generating the dimer shown in Fig. 2.
At the centre of each protomer, a large hydrophobic pocket is found which is proposed to form the IdmH active site (see below). Two loops (residues 38–46 and 99–108) are located near to the entrance of this hydrophobic pocket and limit the access to the proposed active site from the bulk solvent. Interestingly, the loop, which comprises residues 99–108 and is partially truncated in the IdmH-Δ99–107 variant, appears to occupy two distinct conformations within each of the five dimers in the asymmetric unit. Indeed, in each dimer each protomer has this loop occupying either a more open or a more closed conformation above the hydrophobic cavity. Whilst we cannot rule this out as an artefact of the crystallization, this suggests that this loop is likely to be highly dynamic in solution and may also hint at a potential change in conformation between the two protomers, suggesting the potential for allostery in this enzyme (Supplementary Fig. S6). Furthermore, studies of a number of Diels–Alderases have shown that a loop covering the active site becomes more ordered upon substrate binding and this loop could also play such a role here (Byrne et al., 2016; Zheng et al., 2016).
A deep hydrophobic pocket penetrating towards the core of the barrel can be observed in each IdmH protomer. It is likely that this constitutes the active site of the enzyme (Fig. 3 and Supplementary Fig. S7). Each pocket is lined by the side chains of residues Tyr16, Phe19, Leu39, Trp59, Val62, Trp63, Met115, His129, Ser133 and Asn135. We propose that the preponderance of hydrophobic residues might act to steer the hydrophobic substrate (4 in Fig. 1) into an appropriate conformation for facile ring cyclization by acting as a natural product template, as observed by others in structurally related cyclases (Barajas et al., 2017; Li et al., 2010; Liu et al., 2015; Carey & Sundberg, 2007).
BLASTp (Supplementary Fig. S2) and DALI searches against the PDB revealed that the closest sequence and structural homologue to IdmH is the putative polyketide cyclase from Chromobacterium violaceum (PDB entry 4lgq; 30% sequence identity; Joint Center for Structural Genomics, unpublished work). Many of the closest structural and sequence matches, however, stem from structural genomics efforts and thus lack accompanying biochemical data and publications supporting their functional annotation. The putative hydrolases AcIR (PDB entry 2gey) and SnoaL2 (PDB entry 2gex) from S. galilaeus and S. nogalater, respectively (Beinker et al., 2006) as well as the polyketide cyclase SnoaL (PDB entry 1sjw; Sultana et al., 2004) did appear as close structural matches to the structure of IdmH even though they share at best 28% sequence identity. We therefore focused our structural comparisons against these biochemically characterized enzymes.
We superimposed the structure of IdmH on its sequence homologues using PyMOL (Schrödinger) and found that the polyketide cyclase SnoaL from the nogalamycin biosynthetic pathway (PDB entry 1sjw; Sultana et al., 2004) exhibited the highest structural similarity to IdmH as measured by the root-mean-square deviation of the fit (r.m.s.d. = 1.56 Å over 549 atoms) [Fig. 4(a)]. Crystallographic analysis and gel-filtration chromatography experiments both suggest that SnoaL forms a tetramer in solution (Sultana et al., 2004), in contrast to IdmH which is homodimeric. SnoaL catalyses an intramolecular aldol condensation using the conserved residue Asp121 as an acid/base catalyst during the reaction. Other essential residues for the activity of SnoaL include Asn33, Gln105 and two histidines, His107 and His119. In contrast, the active site of IdmH does not contain residues that can perform acid/base chemistry at equivalent positions to the catalytic Asp121 and His107 of SnoaL, but rather contains Thr131 and Asn117 instead [Fig. 4(b)]. Of note, Thr131 is implicated as important in the suggested catalytic mechanism of IdmH discussed later. Both active sites include many aromatic residues, and some of them are conserved between the two proteins (for example Phe19 in IdmH and Phe15 in SnoaL and His129 in IdmH and His119 in SnoaL). There are also two tryptophan residues (Trp63 in IdmH and Trp54 in SnoaL) that appear to be conserved between these proteins; however, they appear to be in entirely different orientations. These different orientations could reflect genuine differences between these two structures or they may arise from the fact that the SnoaL structure contains a bound substrate. Furthermore, there are additional tryptophan (Trp59) and tyrosine (Tyr16) residues in IdmH which seem to be unique to this protein (SnoaL contains Tyr31 and Met11 at these locations) [Fig. 4(b)]. Taken together, these data suggest that IdmH is unlikely to catalyze an aldol condensation as found for SnoaL and is likely to interact with its substrate in a different manner.
We had no access to the potential IdmH substrate 4, so we chose to investigate the binding of the product, indanomycin (1), to the enzyme by 2D nuclear magnetic resonance (NMR) spectroscopy. Although this did not allow us to directly test the hypothesis that IdmH catalyses indane-ring formation, we postulated that the enzyme would bind the product of the reaction, providing evidence to support its role in indanomycin biosynthesis.
IdmH was labelled with 13C, 15N and 2H and, following the collection and processing of a number of triple-resonance spectra (Yamazaki et al., 1994; Supplementary Table S2), the NMR resonances of backbone nuclei were assigned for 88% of residues. To investigate indanomycin binding, [15N]-IdmH was mixed with a stoichiometric amount of ligand (indanomycin) and a 1H–15N HSQC-TROSY spectrum was acquired (Fig. 5). Significant chemical shift perturbations (CSPs) for a number of backbone amides compared with the native protein were seen on addition of indanomycin, suggesting a binding event.
To investigate the binding further and to localize the indanomycin-binding site, titration spectra with substoichiometric ligand:protein ratios were recorded. Significant shifts in both the 1H and 15N spectra were observed for a number of the assigned residues (see Supplementary Fig. S8 for examples of specific residue chemical shift perturbations during the titration). Since shifts can be influenced by through-bond, through-space and allosteric interactions, it is difficult to determine whether the CSPs in IdmH are a result of direct binding or conformational change. However, to localize the protein–ligand interactions, the geometrical distance moved by each peak was calculated for all assigned residues [Fig. 6(a)] and chemical shifts larger than the standard deviation of all shifts were mapped onto the IdmH diagram using yellow, orange and red colours [Fig. 6(b)].
In the presence of indanomycin (the bound spectrum), some resonances moved a substantial distance from their resonance counterparts in the native protein (the free spectrum) making it impossible to match the two accurately. For this reason, a `minimum chemical shift procedure' was employed (Farmer et al., 1996; Lüttgen et al., 2002; Williamson, 2013). This method links each assigned free resonance to the signal in the bound spectrum that has moved the least from the position in the free spectrum. Therefore, each resonance is assigned a CSP, and while the true change might be larger than that assigned, it will never be smaller.
The results of chemical shift mapping suggest that the most significant CSPs are localized at the top of the hydrophobic cavity of IdmH (α-helices 3–4) and on the β-sheet inside the pocket (β-strands 2–4) [Figs. 2 and 6(c)], consistent with the predicted location of the IdmH active site. A closer look at the active-site residues [Fig. 6(d)] revealed that the majority of the resonances exhibited a moderate (Val62 and Met115) to large (Trp59, Trp63 and Tyr16) shift. Two resonances, Ser133 and Ile37, exhibited a shift below the level of significance (<0.72 p.p.m.) and three remaining resonances, Thr131, Phe19 and His129, do not have their backbone resonances assigned. These results thus suggest that indanomycin can bind to IdmH in the proposed active site, and that binding, and potentially catalysis, could induce conformational changes which spread to other areas of the enzyme, as highlighted by the more distal CSPs.
In a bid to obtain further details of indanomycin binding to IdmH, we performed both co-crystallization and crystal-soaking experiments to try and determine a structure of the product complex. Co-crystallization with the ligand appeared to inhibit crystal formation, likely owing to the high concentration of DMSO required to keep the ligand in solution, whilst crystal soaking did not result in significant new electron density that could be confidently attributed to the ligand. In the light of these results, we therefore elected to explore computational methods as a means of further analyzing the biochemical function of IdmH.
Having determined that indanomycin binds in the active site of IdmH, we wanted to further explore the likely chemical reaction catalysed by IdmH. Indanomycin was placed in the wild-type IdmH active sites (chain A and B) using in silico molecular docking. Three significantly different binding poses were found (two in chain A and one in chain B). For the binding pose in chain B, the conformer for Asn117 was adjusted to provide a slight difference in the hydrogen-bonding environment. This led to the following four options: pose A, the indane ring bound in the chain A pocket but without any specific hydrogen bonds being made; pose B, the indane ring in the chain A pocket with a hydrogen bond formed between Ser133 and the substrate carbonyl group (C21=O); pose C, the indane ring in the chain B pocket with a hydrogen bond formed between Ser133 and the substrate carbonyl group (C21=O) but in a different orientation to pose B (Fig. 8); pose D, the indane ring in the chain B pocket with a hydrogen bond formed between Thr131 and the substrate carbonyl group in which a new hydrogen-bond network is formed in the protein between Asn117 and Tyr16. Poses C and D are shown in Fig. 7 (after brief structural optimization), with the main difference in binding pose arising from the side-chain orientation of Asn117. Subsequently, brief structural optimization and 100–300 ps molecular-dynamics simulation was performed for each of the four poses.
11 QM/MM reaction simulations (starting from independent snapshots of the molecular-dynamics simulations) were run for all four possible enzyme–substrate complexes using the same approximate protocol as described in Byrne et al. (2016). Representative (approximate) transition-state structures of these four complexes are shown in Fig. 8, together with the free-energy profile obtained from the 11 simulations together. Based on the energy barriers, reactions proceeding from starting poses C and D are the most likely (Fig. 7). The low barriers compared with poses A and B can be rationalized based on the electron-withdrawing nature of the hydrogen bond formed between Ser133 or Thr131 and the substrate carbonyl at C21 (depending on the pose). Between the two, pose D seems to be the most likely, based on a good fit with the cavity and some possible catalytic stabilization by the positioning of Trp59. The reaction barrier for option A is highest, which could be expected owing to the lack of a (good) hydrogen-bonding interaction in this pose; during the QM/MM simulations a hydrogen bond to Trp59 can be formed (see Fig. 8), but this interaction is more transient and thus does not offer the same benefit as the hydrogen bonds formed in poses C and D. MD and QM/MM simulation of option B lead to a poorer fit with the active-site cavity, suggesting that this is an unlikely starting point for catalysis by IdmH (Fig. 8).
The chemical shift mapping results (Fig. 6) further support the observations from the approximate QM/MM simulations. Pose D involves a hydrogen-bonding network between Thr131, Asn117 and Tyr16 as well as catalytic stabilization from Trp59. According to the NMR titration experiments, all of these residues (except for Thr131, which could not be assigned) are involved in protein–ligand interactions as measured by significant CSPs (0.22 p.p.m. for Tyr16, 0.12 p.p.m. for Asn117 and 0.20 p.p.m. for Trp59). There are also van der Waals interactions between the substrate and the other residues with significant CSPs. It is further worth noting that Ser133, which is involved in a hydrogen bond to the substrate in pose C, displayed a CSP (0.012 p.p.m.) below the level of significance (0.07 p.p.m.), making pose C less likely than pose D. The reaction simulations indicate that an asynchronous concerted mechanism is followed, with the formation of the C15–C19 bond occurring first (and representing the reaction barrier; see Fig. 8); this is similar to the situation in the intramolecular Diels–Alder reaction in AbyU and indeed the uncatalysed reaction (Byrne et al., 2016).
In summary, chemical shift perturbation mapping, in silico docking and approximate QM/MM simulations reveal the structure of the likely reactive IdmH–substrate complex and show that the active-site shape, along with a critical hydrogen bond, is likely to promote the correct conformation of the substrate 4 (bringing together the diene and dienophile) and provides an appropriate environment for the indane-ring cyclization reaction to occur.
The Diels–Alder reaction is of major synthetic value in organic chemistry for the preparation of substituted six-membered rings with the creation of up to four new stereocenters (Stocking & Williams, 2003). The potential use of proteins to catalyse Diels–Alder reactions stereoselectively under mild conditions therefore remains of high interest. Until recently, no natural enzymes that specifically catalyse Diels–Alder reactions (i.e. Diels–Alderases) were known (Jamieson et al., 2019), and some progress was made with obtaining catalytic antibodies and de novo designed enzymes that catalyse a bimolecular Diels–Alder reaction (Gouverneur et al., 1993; Siegel et al., 2010). During the late 1990s and early 2000s, three putative Diels–Alderases involved in natural product synthesis were reported: macrophomate synthase (Watanabe et al., 2000), lovastatin nonaketide synthase (LovB; Auclair et al., 2000) and solanapyrone synthase (Sol5; Katayama et al., 1998; Kasahara et al., 2010). LovB and Sol5 have activities other than catalysing a Diels–Alder reaction, which makes proving any catalysis of the cycloaddition reaction difficult. The bimolecular reaction catalysed by macrophomate synthase is now thought to follow a stepwise mechanism: Michael addition followed by an aldol reaction (Guimarães et al., 2005). In the 2010s, the first enzyme specifically catalysing only a Diels–Alder reaction was reported: SpnF, which is involved in the biosynthesis of spinosyn A (Kim et al., 2011). Subsequently, numerous standalone Diels–Alderases have been identified, especially from bacterial biosynthesis pathways, including the cofactor-free PyrI4 (Zheng et al., 2016) from the pyrroindomycin biosynthetic pathway and its homologues AbyU (Byrne et al., 2016) and VstJ (Hashimoto et al., 2015) from other spirotetronate-containing natural products.
Here, we propose that IdmH from the biosynthesis of indanomycin is another natural standalone Diels–Alderase. Similar to the structurally characterized natural Diels–Alderases AbyU (Byrne et al., 2016) and PyrI4 (Zheng et al., 2016), the enzyme acts in the (final) tailoring steps of natural product biosynthesis, does not require a cofactor and the reaction catalysed is intermolecular. The IdmH structure is a homodimeric `barrel' that contains a `capping loop'. With the capping loop closed, a cavity is formed that fits the substrate in a conformation in which the diene and dienophile moieties are positioned in line with a Diels–Alder reaction. Interestingly, the structure and sequence of IdmH are substantially different from those of AbyU and PyrI4 (as well as their homologues from other spirotetronate and spirotetramate biosynthesis pathways; Supplementary Fig. S9). Instead of a β-barrel, IdmH is an α+β barrel. There is no significant sequence similarity between IdmH on the one hand and AbyU and PyrI4 on the other, nor is there an obvious mapping of the four antiparallel β-strands to the four antiparallel β-strands of the AbyU and PyrI4 β-barrels. Despite the absence of a clear evolutionary relationship between the IdmH α+β barrel and the β-barrels of AbyU and PyrI4, the overall shape is similar, with the substrate cavity near the capping loop in each case (Supplementary Fig. S9; Hashimoto & Kuzuyama, 2016). This raises the question of whether the emergence of these intramolecular Diels–Alderases with similar structures provide an example of convergent evolution. Further, it indicates that not only a β-barrel scaffold, but also a α+β barrel scaffold, could be used for the design of engineered or de novo designed Diels–Alderases, thereby potentially allowing a broader range of substrates.
The NMR assignment data for IdmH have been deposited in the Biological Magnetic Resonance Data Bank (BioMagResBank) with accession number 27838.
We thank Diamond Light Source for access to beamlines I03 and I04 (MX15378) and the European Synchrotron Radiation Facility for access to MASSIF-1 (MX1828) that contributed to the results presented here. NMR spectroscopy was carried out at the Astbury Biostructure Laboratory using 950 MHz Bruker Ascend Aeon and 750 MHz Oxford NMR magnets (University of Leeds ABSL award and Wellcome Trust 104920/Z/14/Z). We thank Dr Anastasia Zhuravleva (University of Leeds) and Dr Gary Thompson (University of Kent) for their help in acquiring and processing 3D NMR spectra, and Dr Ryan Seipke and Dan Van (University of Leeds) for helpful discussions during this work. The simulation work was conducted using the computational facilities of the Advanced Computing Research Centre of the University of Bristol.
ID was supported by the Wellcome Trust (105214/Z/14/Z). MWvdK and GRH were supported by BBSRC David Phillips Fellowships (BB/M026280/1) to Marc W. van der Kamp and (BB/N019970/1) to Glyn R. Hemsworth.
Auclair, K., Sutherland, A., Kennedy, J., Witter, D. J., Van den Heever, J. P., Hutchinson, C. R. & Vederas, J. C. (2000). J. Am. Chem. Soc. 122, 11519–11520. Web of Science CrossRef CAS Google Scholar
Barajas, J. F., Shakya, G., Moreno, G., Rivera, H., Jackson, D. R., Topper, C. L., Vagstad, A. L., La Clair, J. J., Townsend, C. A., Burkart, M. D. & Tsai, S.-C. (2017). Proc. Natl Acad. Sci. USA, 114, E4142–E4148. CrossRef CAS PubMed Google Scholar
Beinker, P., Lohkamp, B., Peltonen, T., Niemi, J., Mäntsälä, P. & Schneider, G. (2006). J. Mol. Biol. 359, 728–740. Web of Science CrossRef PubMed CAS Google Scholar
Bowler, M. W., Nurizzo, D., Barrett, R., Beteva, A., Bodin, M., Caserotto, H., Delagenière, S., Dobias, F., Flot, D., Giraud, T., Guichard, N., Guijarro, M., Lentini, M., Leonard, G. A., McSweeney, S., Oskarsson, M., Schmidt, W., Snigirev, A., von Stetten, D., Surr, J., Svensson, O., Theveneau, P. & Mueller-Dieckmann, C. (2015). J. Synchrotron Rad. 22, 1540–1547. Web of Science CrossRef IUCr Journals Google Scholar
Bowler, M. W., Svensson, O. & Nurizzo, D. (2016). Crystallogr. Rev. 22, 233–249. Web of Science CrossRef CAS Google Scholar
Byrne, M. J., Lees, N. R., Han, L.-C., van der Kamp, M. W., Mulholland, A. J., Stach, J. E. M., Willis, C. L. & Race, P. R. (2016). J. Am. Chem. Soc. 138, 6095–6098. CrossRef CAS PubMed Google Scholar
Carey, F. A. & Sundberg, R. J. (2007). Advanced Organic Chemistry. Part B: Reactions and Synthesis, 5th ed. New York: Springer. Google Scholar
Case, D. A., Cerutti, D. S., Cheatham, T. E., Darden, T. A., Duke, R. E., Giese, T. J., Gohlke, H., Götz, A. W., Greene, D., Homeyer, N., Izadi, S., Kovalenko, A., Lee, T. S., Le Grand, S., Li, P., Lin, C., Liu, J., Luchko, T., Luo, R., Mermelstein, D., Merz, K. M., Monard, G., Nguyen, H., Omelyan, I., Onufriev, A., Pan, F., Qi, R., Roe, D. R., Roitberg, A. E., Sagui, C., Simmerling, C. L., Botello-Smith, W. M., Swails, J. M., Walker, R. C., Wang, J., Wolf, R. M., Wu, X., Xiao, L., York, D. M. & Kollman, P. A. (2017). AMBER 2017. University of California, San Francisco. Google Scholar
Chen, V. B., Arendall, W. B., Headd, J. J., Keedy, D. A., Immormino, R. M., Kapral, G. J., Murray, L. W., Richardson, J. S. & Richardson, D. C. (2010). Acta Cryst. D66, 12–21. Web of Science CrossRef CAS IUCr Journals Google Scholar
Cowtan, K. (2006). Acta Cryst. D62, 1002–1011. Web of Science CrossRef CAS IUCr Journals Google Scholar
Cowtan, K. (2010). Acta Cryst. D66, 470–478. Web of Science CrossRef CAS IUCr Journals Google Scholar
Cowtan, K. (2012). Acta Cryst. D68, 328–335. Web of Science CrossRef CAS IUCr Journals Google Scholar
Cummings, M., Breitling, R. & Takano, E. (2014). FEMS Microbiol. Lett. 351, 116–125. CrossRef CAS PubMed Google Scholar
Delaglio, F., Grzesiek, S., Vuister, G. W., Zhu, G., Pfeifer, J. & Bax, A. (1995). J. Biomol. NMR, 6, 277–293. CrossRef CAS PubMed Web of Science Google Scholar
Dutton, C. J., Banks, B. J. & Cooper, C. B. (1995). Nat. Prod. Rep. 12, 165–181. CrossRef CAS PubMed Google Scholar
Edwards, M. P., Ley, S. V., Lister, S. G., Palmer, B. D. & Williams, D. J. (1984). J. Org. Chem. 49, 3503–3516. CSD CrossRef CAS Google Scholar
Eisenberg, D., Schwarz, E., Komaromy, M. & Wall, R. (1984). J. Mol. Biol. 179, 125–142. CrossRef CAS PubMed Web of Science Google Scholar
Elstner, M., Porezag, D., Jungnickel, G., Elsner, J., Haugk, M., Frauenheim, T., Suhai, S. & Seifert, G. (1998). Phys. Rev. B, 58, 7260–7268. Web of Science CrossRef CAS Google Scholar
Emsley, P., Lohkamp, B., Scott, W. G. & Cowtan, K. (2010). Acta Cryst. D66, 486–501. Web of Science CrossRef CAS IUCr Journals Google Scholar
Farmer, B. T., Constantine, K. L., Goldfarb, V., Friedrichs, M. S., Wittekind, M., Yanchunas, J., Robertson, J. G. & Mueller, L. (1996). Nat. Struct. Biol. 3, 995–997. CrossRef CAS PubMed Google Scholar
Goldschmidt, L., Cooper, D. R., Derewenda, Z. S. & Eisenberg, D. (2007). Protein Sci. 16, 1569–1576. Web of Science CrossRef PubMed CAS Google Scholar
Gouverneur, V. E., Houk, K. N., de Pascual-Teresa, B., Beno, B., Janda, K. D. & Lerner, R. A. (1993). Science, 262, 204–208. CrossRef CAS PubMed Web of Science Google Scholar
Grossfield, A. (2013). WHAM: The Weighted Histogram Analysis Method v.184.108.40.206. https://membrane.urmc.rochester.edu/content/wham. Google Scholar
Guimarães, C. R., Udier-Blagović, M. & Jorgensen, W. L. (2005). J. Am. Chem. Soc. 127, 3577–3588. PubMed Google Scholar
Ha, N.-C., Choi, G., Choi, K. Y. & Oh, B.-H. (2001). Curr. Opin. Struct. Biol. 11, 674–678. CrossRef PubMed CAS Google Scholar
Hashimoto, T., Hashimoto, J., Teruya, K., Hirano, T., Shin-ya, K., Ikeda, H., Liu, H., Nishiyama, M. & Kuzuyama, T. (2015). J. Am. Chem. Soc. 137, 572–575. CrossRef CAS PubMed Google Scholar
Hashimoto, T. & Kuzuyama, T. (2016). Curr. Opin. Chem. Biol. 35, 117–123. CrossRef CAS PubMed Google Scholar
Hirvonen, V. H. A., Hammond, K., Chudyk, E. I., Limb, M. A. L., Spencer, J., Mulholland, A. & van der Kamp, M. W. (2019). J. Chem. Inf. Model. 59, 3365–3369. CrossRef CAS PubMed Google Scholar
Jamieson, C. S., Ohashi, M., Liu, F., Tang, Y. & Houk, K. N. (2019). Nat. Prod. Rep. 36, 698–713. CrossRef CAS PubMed Google Scholar
Jaravine, V. A. & Orekhov, V. Y. (2006). J. Am. Chem. Soc. 128, 13421–13426. CrossRef PubMed CAS Google Scholar
Jaravine, V. A., Zhuravleva, A. V., Permi, P., Ibraghimov, I. & Orekhov, V. Y. (2008). J. Am. Chem. Soc. 130, 3927–3936. CrossRef PubMed CAS Google Scholar
Jurrus, E., Engel, D., Star, K., Monson, K., Brandi, J., Felberg, L., Brookes, D., Wilson, L., Chen, J., Liles, K., Chun, M., Li, P., Gohara, D., Dolinsky, T., Konecny, R., Koes, D., Nielsen, J., Head-Gordon, T., Geng, W., Krasny, R., Wei, G., Holst, M., McCammon, J. & Baker, N. (2018). Protein Sci. 27, 112–128. CrossRef CAS PubMed Google Scholar
Kabsch, W. (2010). Acta Cryst. D66, 133–144. Web of Science CrossRef CAS IUCr Journals Google Scholar
Kallio, P., Sultana, A., Niemi, J., Mäntsälä, P. & Schneider, G. (2006). J. Mol. Biol. 357, 210–220. Web of Science CrossRef PubMed CAS Google Scholar
Kantardjieff, K. A. & Rupp, B. (2003). Protein Sci. 12, 1865–1871. Web of Science CrossRef PubMed CAS Google Scholar
Kasahara, K., Miyamoto, T., Fujimoto, T., Oguri, H., Tokiwano, T., Oikawa, H., Ebizuka, Y. & Fujii, I. (2010). Chembiochem, 11, 1245–1252. CAS PubMed Google Scholar
Katayama, K., Kobayashi, T., Oikawa, H., Honma, M. & Ichihara, A. (1998). Biochim. Biophys. Acta, 1384, 387–395. Web of Science CrossRef CAS PubMed Google Scholar
Kim, H. J., Ruszczycky, M. W., Choi, S.-H., Liu, Y.-N. & Liu, H. (2011). Nature (London), 473, 109–112. CrossRef CAS PubMed Google Scholar
Li, C., Roege, K. E. & Kelly, W. L. (2009). Chembiochem, 10, 1064–1072. CrossRef PubMed CAS Google Scholar
Li, Y., Xu, W. & Tang, Y. (2010). J. Biol. Chem. 285, 22764–22773. CrossRef CAS PubMed Google Scholar
Liu, L., Zhang, Z., Shao, C.-L., Wang, J.-L., Bai, H. & Wang, C.-Y. (2015). Sci. Rep. 5, 10463. CrossRef PubMed Google Scholar
Lundqvist, T., Rice, J., Hodge, C. N., Basarab, G. S., Pierce, J. & Lindqvist, Y. (1994). Structure, 2, 937–944. CrossRef CAS PubMed Web of Science Google Scholar
Lüttgen, H., Robelek, R., Mühlberger, R., Diercks, T., Schuster, S. C., Köhler, P., Kessler, H., Bacher, A. & Richter, G. (2002). J. Mol. Biol. 316, 875–885. PubMed Google Scholar
Maier, J. A., Martinez, C., Kasavajhala, K., Wickstrom, L., Hauser, K. E. & Simmerling, C. (2015). J. Chem. Theory Comput. 11, 3696–3713. Web of Science CrossRef CAS PubMed Google Scholar
McCoy, A. J. (2007). Acta Cryst. D63, 32–41. Web of Science CrossRef CAS IUCr Journals Google Scholar
McCoy, A. J., Grosse-Kunstleve, R. W., Adams, P. D., Winn, M. D., Storoni, L. C. & Read, R. J. (2007). J. Appl. Cryst. 40, 658–674. Web of Science CrossRef CAS IUCr Journals Google Scholar
Minami, A., Ose, T., Sato, K., Oikawa, A., Kuroki, K., Maenaka, K., Oguri, H. & Oikawa, H. (2014). ACS Chem. Biol. 9, 562–569. CrossRef CAS PubMed Google Scholar
Murshudov, G. N., Skubák, P., Lebedev, A. A., Pannu, N. S., Steiner, R. A., Nicholls, R. A., Winn, M. D., Long, F. & Vagin, A. A. (2011). Acta Cryst. D67, 355–367. Web of Science CrossRef CAS IUCr Journals Google Scholar
Newman, D. J. & Cragg, G. M. (2012). J. Nat. Prod. 75, 311–335. CrossRef CAS PubMed Google Scholar
Nurizzo, D., Bowler, M. W., Caserotto, H., Dobias, F., Giraud, T., Surr, J., Guichard, N., Papp, G., Guijarro, M., Mueller-Dieckmann, C., Flot, D., McSweeney, S., Cipriani, F., Theveneau, P. & Leonard, G. A. (2016). Acta Cryst. D72, 966–975. Web of Science CrossRef IUCr Journals Google Scholar
Olano, C., Méndez, C. & Salas, J. A. (2010). Nat. Prod. Rep. 27, 571–616. CrossRef CAS PubMed Google Scholar
Olsson, M. H., Søndergaard, C. R., Rostkowski, M. & Jensen, J. H. (2011). J. Chem. Theory Comput. 7, 525–537. Web of Science CrossRef CAS PubMed Google Scholar
Potterton, L., Agirre, J., Ballard, C., Cowtan, K., Dodson, E., Evans, P. R., Jenkins, H. T., Keegan, R., Krissinel, E., Stevenson, K., Lebedev, A., McNicholas, S. J., Nicholls, R. A., Noble, M., Pannu, N. S., Roth, C., Sheldrick, G., Skubak, P., Turkenburg, J., Uski, V., von Delft, F., Waterman, D., Wilson, K., Winn, M. & Wojdyr, M. (2018). Acta Cryst. D74, 68–84. Web of Science CrossRef IUCr Journals Google Scholar
Rommel, K. R., Li, C. & Kelly, W. L. (2011). Org. Lett. 13, 2536–2539. CrossRef CAS PubMed Google Scholar
Roy, A., Kucukural, A. & Zhang, Y. (2010). Nat. Protoc. 5, 725–738. CrossRef CAS PubMed Google Scholar
Sheldrick, G. M. (2010). Acta Cryst. D66, 479–485. Web of Science CrossRef CAS IUCr Journals Google Scholar
Siegel, J. B., Zanghellini, A., Lovick, H. M., Kiss, G., Lambert, A. R., St Clair, J. L., Gallaher, J. L., Hilvert, D., Gelb, M. H., Stoddard, B. L., Houk, K. N., Michael, F. E. & Baker, D. (2010). Science, 329, 309–313. CrossRef CAS PubMed Google Scholar
Solyom, Z., Schwarten, M., Geist, L., Konrat, R., Willbold, D. & Brutscher, B. (2013). J. Biomol. NMR, 55, 311–321. CrossRef CAS PubMed Google Scholar
Søndergaard, C. R., Olsson, M. H., Rostkowski, M. & Jensen, J. H. (2011). J. Chem. Theory Comput. 7, 2284–2295. PubMed Google Scholar
Stocking, E. M. & Williams, R. M. (2003). Angew. Chem. Int. Ed. 42, 3078–3115. CrossRef CAS Google Scholar
Sultana, A., Kallio, P., Jansson, A., Wang, J.-S., Niemi, J., Mäntsälä, P. & Schneider, G. (2004). EMBO J. 23, 1911–1921. Web of Science CrossRef PubMed CAS Google Scholar
Svensson, O., Gilski, M., Nurizzo, D. & Bowler, M. W. (2018). Acta Cryst. D74, 433–440. Web of Science CrossRef IUCr Journals Google Scholar
Svensson, O., Malbet-Monaco, S., Popov, A., Nurizzo, D. & Bowler, M. W. (2015). Acta Cryst. D71, 1757–1767. Web of Science CrossRef IUCr Journals Google Scholar
Świderek, K. & Moliner, V. (2016). J. Phys. Chem. B, 120, 2053–2070. PubMed Google Scholar
Tang, L., Ward, S., Chung, L., Carney, J. R., Li, Y., Reid, R. & Katz, L. (2004). J. Am. Chem. Soc. 126, 46–47. CrossRef PubMed CAS Google Scholar
Trott, O. & Olson, A. J. (2010). J. Comput. Chem. 31, 455–461. Web of Science PubMed CAS Google Scholar
Vranken, W. F., Boucher, W., Stevens, T. J., Fogh, R. H., Pajon, A., Llinas, P., Ulrich, E. L., Markley, J. L., Ionides, J. & Laue, E. D. (2005). Proteins, 59, 687–696. Web of Science CrossRef PubMed CAS Google Scholar
Wang, J., Wang, W., Kollman, P. A. & Case, D. A. (2006). J. Mol. Graph. Model. 25, 247–260. Web of Science CrossRef PubMed Google Scholar
Wang, J., Wolf, R. M., Caldwell, J. W., Kollman, P. A. & Case, D. A. (2004). J. Comput. Chem. 25, 1157–1174. Web of Science CrossRef PubMed CAS Google Scholar
Watanabe, K., Mie, T., Ichihara, A., Oikawa, H. & Honma, M. (2000). J. Biol. Chem. 275, 38393–38401. CrossRef PubMed CAS Google Scholar
Williamson, M. P. (2013). Prog. Nucl. Magn. Reson. Spectrosc. 73, 1–16. Web of Science CrossRef CAS PubMed Google Scholar
Yamazaki, T., Lee, W., Arrowsmith, C. H., Muhandiram, D. R. & Kay, L. E. (1994). J. Am. Chem. Soc. 116, 11655–11666. CrossRef CAS Web of Science Google Scholar
Zheng, Q., Guo, Y., Yang, L., Zhao, Z., Wu, Z., Zhang, H., Liu, J., Cheng, X., Wu, J., Yang, H., Jiang, H., Pan, L. & Liu, W. (2016). Cell Chem. Biol. 23, 352–360. CrossRef CAS PubMed Google Scholar
Zimmerman, D. E., Kulikowski, C. A., Huang, Y., Feng, W., Tashiro, M., Shimotakahara, S., Chien, C.-Y., Powers, R. & Montelione, G. T. (1997). J. Mol. Biol. 269, 592–610. CrossRef CAS PubMed Google Scholar
This is an open-access article distributed under the terms of the Creative Commons Attribution (CC-BY) Licence, which permits unrestricted use, distribution, and reproduction in any medium, provided the original authors and source are cited.