CHEMBIOCHEM FULL PAPERS DOI: 10.1002/cbic.201402206
Predicted Incorporation of Non-native Substrates by a Polyketide Synthase Yields Bioactive Natural Product Derivatives Kenny Bravo-Rodriguez,[a] Ahmed F. Ismail-Ali,[b] Stephan Klopries,[b] Susanna Kushnir,[b] Shehab Ismail,[c, d] Eyad K. Fansa,[c] Alfred Wittinghofer,[c] Frank Schulz,*[b] and Elsa SanchezGarcia*[a] The polyether ionophore monensin is biosynthesized by a polyketide synthase that delivers a mixture of monensins A and B by the incorporation of ethyl- or methyl-malonyl-CoA at its fifth module. Here we present the first computational model of the fifth acyltransferase domain (AT5mon) of this polyketide synthase, thus affording an investigation of the basis of the relaxed specificity in AT5mon, insights into the activation for the nucleophilic attack on the substrate, and prediction of the incorporation of synthetic malonic acid building blocks by this
enzyme. Our predictions are supported by experimental studies, including the isolation of a predicted derivative of the monensin precursor premonensin. The incorporation of nonnative building blocks was found to alter the ratio of premonensins A and B. The bioactivity of the natural product derivatives was investigated and revealed binding to prenyl-binding protein. We thus show the potential of engineered biosynthetic polyketides as a source of ligands for biological macromolecules.
Introduction Polyketide synthases (PKSs) are natural multienzyme complexes of high relevance for medicine as they synthesize compounds used as immunosuppressants, antibiotics, and anticancer drugs, among others. PKSs catalyze a cascade of Claisen condensations between enzyme-bound acyl thioesters and malonic acid thioesters, which serve as extender units for the growing polyketide chain. In bacterial type I PKSs, acyltransferase (AT) domains serve as gatekeepers to recruit the malonicacid-based building blocks required for chain extension. The mechanism of this process is largely not understood, thus rendering the task of engineering a PKS for the incorporation of non-native building blocks a formidable challenge.[1–6] In this context, elucidation of the molecular mechanisms underlying [a] K. Bravo-Rodriguez,+ Dr. E. Sanchez-Garcia Department of Theoretical Chemistry Max-Planck-Institut fr Kohlenforschung Kaiser-Wilhelm-Platz 1, 45470 Mlheim an der Ruhr (Germany) E-mail: [email protected] [b] A. F. Ismail-Ali,+ S. Klopries, Dr. S. Kushnir, Prof. Dr. F. Schulz Fakultt fr Chemie und Biochemie, Ruhr-Universitt Bochum Universittsstrasse 150, 44780 Bochum (Germany) E-mail: [email protected] [c] Dr. S. Ismail, E. K. Fansa, Prof. Dr. A. Wittinghofer Max-Planck-Institut fr Molekulare Physiologie Otto-Hahn-Strasse 11, 44227 Dortmund (Germany) [d] Dr. S. Ismail Present address: CR-UK Beatson Institute, Garscube Estate Switchback Road,Glasgow G61 1BD (UK) [+] These authors contributed equally to this work. Supporting information for this article is available on the WWW under http://dx.doi.org/10.1002/cbic.201402206.
2014 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim
substrate incorporation into PKS is a task of high importance, but this is currently hindered by the lack of structural information on the megadalton-sized PKSs. The AT domain in module five of the monensin PKS (AT5mon) is known to accept both ethyl-malonyl- and methyl-malonylCoA (EtMCoA and MMCoA) as substrates, thereby giving rise to monensins A and B.[7, 8] Previous mutagenesis experiments have shown the flexibility of this PKS in processing non-native substrates with different redox patterns in the nascent polyketide chain, thus implying opportunities for further biosynthetic derivatization of monensin.[9] Furthermore, our earlier investigations on artificially induced substrate promiscuity in the sixth AT domain of the erythromycin assembly line (AT6DEBS) showed that a non-natural building block can be introduced as a source of a propargylated side chain in erythromycin A by targeted mutagenesis of the AT6DEBS domain.[5] However, it is important to broaden and deepen the understanding of substrate recognition by giant and complex PKSs, in particular of the AT domains, as these are crucial for the introduction of new functional groups and stereocenters.
Results and Discussion Molecular modeling of acyltransferase domain 5 of monensin PKS We built a homology model of the AT5mon domain in order to investigate its substrate recognition mechanism (see Computational details and the Supporting Information). Based on the amino acid sequence of AT5mon, we used the X-ray structures ChemBioChem 0000, 00, 1 – 8
&1&
These are not the final page numbers! ÞÞ
CHEMBIOCHEM FULL PAPERS of AT3DEBS, AT5DEBS, and malonyl CoA/acyl carrier protein transacylase (MAT) as templates (Section I.1 in the Supporting Information). After equilibration, we performed molecular dynamics (MD) simulations of solvated AT5mon without ligand. These simulations indicated that the overall protein structure is conserved. However, in the absence of substrate, the region formed by residues 267–339 was most flexible, as it tended to move away from the rest of the protein. This movement makes the active site of AT5mon more open and exposed to the ligand than in AT6DEBS. MD simulations of AT5mon models with the natural substrates (MMCoA or EtMCoA) in different orientations showed, in all cases, a structured active site able to accommodate the ligands. We found that the main interaction between MMCoA and the active site is the salt bridge formed between the carboxylate group of MMCoA and R257, which is located at the bottom of the active site. The thioester and amide groups of MMCoA further contribute to keeping the ligand in a suitable position within the active site (Section I.2 in the Supporting Information). The simultaneous interactions of the thioester moiety and the closest amide group of MMCoA with the backbone of Q149 help the substrate to adopt a conformation in which S232 is close to the thioester, thus enabling the nucleophilic attack (Figure 1 A). Our work also provides insights into the activation of S232. In all simulations, S232 interacted with H231 (Section I.2 and Table S1 in the Supporting Information) or with the carboxylate group of the ligand, but never with H334. This suggests that either H231 or the carboxylate group (but not H334) acts as the base to promote activation of S232 in AT5mon. By incorporating EtMCoA, AT5mon exhibits a relaxed specificity compared to AT6DEBS, which is highly selective for MMCoA. A comparison of AT5mon with our AT6DEBS model[5] provides interesting insights into possible factors contributing to substrate promiscuity in AT5mon. In AT6DEBS, the YASH sequence strongly interacts with the substrate and together with V295 prevents the incorporation of substrates larger than MMCoA in the active site. We established that the V295A mutation lowers the specificity of AT6DEBS.[5] In AT5mon, the sequence equivalent to
www.chembiochem.org
Figure 1. A) Residue S232 interacts with H231 but not with H334. B) A comparison of AT5mon with AT6DEBS (italics) indicates that the VAGH sequence and S329 might play a role in the substrate promiscuity of the monensin PKS.
YASH is VAGH, and S329 is the residue equivalent to V295 (Figure 1 B). Our simulations indicated that these differences result in a more open active site that is thus able to accept not only MMCoA but also EtMCoA as a substrate. EtMCoA establishes an interactions network with the active site similarly to MMCoA. The most favorable orientation of this substrate is with the ethyl group pointing towards V331 (Figure 2 A). Unlike MMCoA, EtMCoA was found in a location distant from S232 if the ethyl moiety was not in the correct orientation (Section I.2 in the Supporting Information). In all cases, the in-
Figure 2. Binding at the active site of AT5mon by A) EtMCoA (native substrate), B) MCoA, and C) propargyl-MSNAC. The substrates are shown as thicker stick models. Unlike EtMCoA, MCoA lacks many stabilizing interactions with the active site. Propargyl-MSNAC is accommodated in the active site in an orientation similar to that of EtMCoA.
2014 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim
ChemBioChem 0000, 00, 1 – 8
&2&
These are not the final page numbers! ÞÞ
CHEMBIOCHEM FULL PAPERS teraction with R257 was conserved, and thus EtMCoA remained inside the active site. Given the experimental complexity of PKS manipulation, the ability to computationally predict the incorporation of nonnative substrates is important, not only as a guide to experiments but also to understand the fundamental substrate selectivity of PKS. The relaxed specificity of AT5mon relative to AT6DEBS and other AT domains does not automatically establish AT5mon as a very promiscuous enzyme, as MMCoA and EtMCoA differ by only a methylene group. In this context, we wanted to establish whether AT5mon is able to incorporate synthetic building blocks with larger and more rigid moieties than a methyl or ethyl group. Thus, we computationally investigated the incorporation of two ligands of different sizes relative to MMCoA: the smaller natural malonyl-CoA (MCoA) and the larger synthetic propargyl-malonyl-SNAC (propargyl-MSNAC; Figure 2 B and C). Simulations of AT5mon with bound MCoA suggested significantly weakened interaction with the active site: average values of the distance between R257 and the carboxylate moiety of MCoA were large, the substrate was far from S232, and no secondary stabilizing interactions between the ligand and the active site were found (Figure 2 B; Section I.2, and Table S1 in the Supporting Information). This is consistent with experimental observations: in premonensin and monensin biosynthesis, the abundantly available natural MCoA does not lead to the corresponding desmethyl derivatives of the natural products in significant quantities.[9] Propargyl-MSNAC, on the other hand, was significantly better accepted by the active site and exhibited a behavior similar to that of the natural substrate, EtMCoA. However, unfavorable orientations of this ligand led to disruption of several interactions important for substrate binding and activation (with residues such as Q149, Q233, and R257; Section I.2 in the Supporting Information). Our results thus indicate that the insertion of propargyl-MSNAC into the active site depends on the orientation of its triple-bond moiety, which should point toward V331 similarly to the ethyl group in EtMCoA (Figure 2 A and C). In addition to homology modeling and MD simulations we performed quantum mechanics/molecular mechanics (BP86D2/SVP//CHARMM) calculations of the AT5mon–propargylMSNAC system; the results were in very good agreement with the MD results (Supporting Information, Section I.2 and Table S5). The relative free energy differences associated with the transformations MMCoA!EtMCoA and MMCoA!propargyl-MCoA were also calculated. To directly compare the effect of the substituents, we calculated propargyl-MCoA instead of propargyl-MSNAC, which was used in the experiments. As both MMCoA and EtMCoA are natural substrates of AT5mon, DDG should have values close to 0. As expected, the calculated value for the MMCoA!EtMCoA transformation was 0.12 0.02 kcal mol1. The small DDG value for MMCoA!propargylMCoA (0.53 0.04 kcal mol1) means that, thermodynamically, the active site of AT5mon should be able to accommodate the bulky propargyl group as it does with the natural substrates.
2014 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim
www.chembiochem.org Experimental investigation of substrate promiscuity of AT5mon Malonic acid derivatives, each thioester-activated with N-acetylcysteamine (SNAC), were synthesized either from commercially available derivatives of malonic acid diesters or by reductive alkylation of Meldrum’s acid (Scheme 1, Figure 3 A; and Section IV in the Supporting Information).[10]
Scheme 1. Synthesis of compounds 4–8; a 3.0 equiv LiOH·H2O, H2O, 18 h, RT; b) 1.01 equiv isoprenylacetate, 0.06 equiv H2SO4, neat, 18 h, RT; c) 1.0 equiv borane dimethylamine complex, 3.0 equiv propyl, butyl, or hexyl aldehyde, 1 h, RT; d) tBuOH, 6 h, 90–100 8C; e) 1.1 equiv CDI, 0.3 equiv DMAP, 1.2 equiv SNAC, THF, 18 h, RT; f) 2.5 equiv TiCl4, DCM, 6 h, RT. CDI: N,N’-Carbonyldiimidazole, DMAP: 4-dimethylaminopyridine, SNAC: N-acetylcysteamine.
Cultures of Streptomyces cinnamonensis A495 were supplemented with each derivative (10–30 mm). This strain produces premonensin, an artificial shunt-product of the monensin biosynthetic pathway.[8] Analysis of the fermentation extracts by LC-ESI-MS and HRMS indeed revealed significant PKS promiscuity (Section II.3 in the Supporting Information). Four of the five synthetic building blocks were incorporated in significant amounts. 2-Hexylmalonyl-SNAC was found to be toxic for the bacterium and to abolish the fermentation of premonensin. LC-ESI-MS analysis showed that each new premonensin derivative was as a third fermentation product, in addition to premonensins A and B. This indicates specific incorporation of the building blocks by the AT5mon domain. As a general trend, the ChemBioChem 0000, 00, 1 – 8
&3&
These are not the final page numbers! ÞÞ
CHEMBIOCHEM FULL PAPERS
www.chembiochem.org
Figure 3. A) Malonic acid derivatives used in this study (those investigated by molecular modelling are in bold letters). From the synthetic building blocks, analogues 4–7 were experimentally incorporated into premonensin. Compound 5 a was isolated from the fermentation broth and analyzed by NMR. The corresponding propargylated ER20-derivative, however, was not found by LC-ESI-MS. B) LC-ESI-MS total ion chromatogram of a crude fermentation extract of S. cinnamonensis A495 supplemented with 5. C) High-resolution MS results from the same fermentation extract; 5 a is identified by its H + , NH4 + , and Na + adducts. D) A segment of the 1H NMR analysis of premonensin A (3 a, lower trace) in comparison to 5 a (upper trace).
2014 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim
ChemBioChem 0000, 00, 1 – 8
&4&
These are not the final page numbers! ÞÞ
CHEMBIOCHEM FULL PAPERS experiments revealed higher relative incorporation efficiency at lower concentration of the respective malonic acid derivative (Section II.2 in the Supporting Information). Higher concentrations were increasingly toxic to the bacterium. Addition of propyl-malonyl-SNAC nearly abolished the production of premonensin and resulted in only trace amounts of the natural product derivative. Interestingly, the relative amounts of the naturally produced products premonensins A and B appeared to be inverted in the LC-ESI-MS analysis upon addition of the artificial building blocks. For propargyl-MSNAC, the feeding experiment was scaled up to 1.8 L, as its preparative incorporation might give rise to a synthetically useful orthogonal functional group in the polyketide (Section II.4 in the Supporting Information). After purification, the identity of propargyl-premonensin was confirmed by 1H- and 13C NMR (Figure 3 D and Section II.5 in the Supporting Information), thus proving specific incorporation of the building block by AT5mon. The new product was isolated in a yield of 0.55 mg L1 (5.4 mg L1 premonensin A and 1.8 mg L1 premonensin B from the same fermentation). This confirmed the inversion of the incorporation efficiencies of ethyl- and methyl-malonyl-CoA upon addition of the artificial malonic acid derivative (control fermentations yielded 8.2 mg L1 premonensin A and 15.2 mg L1 premonensin B). In further investigation of the substrate flexibility of the PKS, all artificial building blocks except butylmalonyl-SNAC were supplied to cultures of the mutant strain S. cinnamonensis A495-ER20.[9] This strain gives rise to a redox derivative of premonensin (Figure 3 A) and was the result of previous PKS-engineering experiments to yield polyketide redox derivatives.[9] Analysis of these fermentations revealed interesting differences from the wild-type PKS and indicates “crosstalk” between different segments of the multienzyme complex by unknown mechanisms. As expected, the overall amounts of the polyketide derivatives were significantly lower than for the wildtype PKS. Interestingly, however, propargyl-MSNAC did not give rise to the expected premonensin derivative but instead largely abolished productivity. Previously, a small amount of a biosynthetic byproduct corresponding to the incorporation of MCoA by AT5mon was isolated.[9] Propyl- and allyl-malonylSNAC on the other hand led to significant incorporation levels relative to wild type. These experiments indicate a subtle but effective crosstalk mechanism between module 2 and module 5. Apparently, the differences in the surrounding PKS machinery can induce alterations of the substrate specificity of individual domains or modules. This could be the result of dedicated proofreading mechanisms or simply substrate-induced alteration of catalytic activity.[9, 11] Bioactivity of premonensin and its derivative Polyketides have specific chemical structures to interact with biological macromolecules.[12] Here we found that the non-natural biosynthetic shunt-product premonensin binds tightly to the human phosphopdiesterase 6 delta subunit (PDEd). This protein plays a pivotal role in K-Ras trafficking in human cells, and its inhibition by suitable binders was shown to have cyto 2014 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim
www.chembiochem.org toxic effects on K-Ras-dependent human pancreatic tumor cells.[13] Premonensin B was found to bind to PDEd in a competitive fluorescence polarization assay with farnesylated Rheb peptide, which has been used as a mimic of the K-Ras C-terminal peptide (KD = (214 10) nm, see Figure 4 and Section III in
Figure 4. A 1 mm solution of fluorescently labeled Rheb peptide in the presence of 1 mm PDEd was titrated against increasing concentrations of premonensin B (2 a) and propargyl-premonensin. Premonensin B showed an IC50 of 0.81 mm whereas propargyl-premonensin (5 a) showed an IC50 of 2.00 mm. ~: negative control, &: premonensin B, and *: propargyl-premonensin.
the Supporting Information). Premonensin B was able to displace the natural binder from the complex with PDEd. The propargylated derivative showed a 2.5-fold decreased affinity towards the protein (Figure 4 and Section III in the Supporting Information). This result shows that modified biosynthesis can produce suitable starting compounds for natural-product based structure-activity studies. In this context, the newly attached propargyl moiety will, after further improvement of this precursordirected biosynthesis experiment, offer opportunities for straightforward chemical modification in a chemistry–biology– chemistry reaction sequence.[14]
Conclusion Here we present the first computational model of the AT5mon domain of the monensin PKS; this domain plays an essential role in building-block selection and incorporation. This model allowed us to predict the incorporation of non-native substrates in the wild-type system and to investigate the basis of substrate recognition by this enzyme. Our predictions were corroborated and extended by systematic screening for the incorporation of non-natural malonic acid derivatives as building blocks for this PKS. Important findings are that the inclusion of artificial building blocks seemed to reverse the relative amounts of the shunt products premonensin A and B. Moreover, premonensin B and propargyl-premonensin (obtained after theoretically predicting the incorporation of propargylMSNAC by wild-type AT5mon) bound to PDEd. The incorporation of the new propargyl moiety into the structure of premonensin opens perspectives for the discovery of further bioactive derivatives through targeted biosynthetic derivatization and chemical conversions. ChemBioChem 0000, 00, 1 – 8
&5&
These are not the final page numbers! ÞÞ
CHEMBIOCHEM FULL PAPERS Experimental Section Computational details: Based on its amino acid sequence, the homology model of the AT5 domain of the monensin PKS (AT5mon) was generated from the X-ray structures of DEBS domains AT3DEBS (PDB ID: 2QO3)[15] and AT5DEBS (PDB ID: 2HG4)[16] and of malonylCoA:acyl carrier protein transacylase (MAT, to be published) as templates. The program Modeller (v. 9.10) was used.[17] The program NAMD2.9[18] was used for all MD simulations with the CHARMM22 force field[19] for the protein and the TIP3P model[20] for water. Parameters for the substrates were generated by Swissparam[21] and previously tested by us.[5] The PME method was used for the treatment of the electrostatic interactions.[22] QM/MM (BP86-D2/SVP//CHARMM22)[23–25] optimizations were also used to study the interaction between propargyl-MSNAC and the active site. Snapshots taken at the beginning and end of the simulations were optimized and compared for an additional assessment of the evolution of key contacts between the substrate and residues in the active site. The QM regions were defined as the entire substrate. All residues within 30 of the substrate were allowed to move freely during the optimizations. The HDLC optimizer was used.[26] QM calculations were handled with Turbomole5.10;[27] DL_POLY was used for the MM calculations.[28] All QM/MM optimizations were performed with Chemshell3.5.[29] Relative free energy differences (DDG) were calculated to assess the preference of the active site for different lateral chains of the substrates. The transformations of MMCoA in EtMCoA, and propargyl-MCoA were computed by using Free Energy Perturbation theory.[30, 31] The free energy calculations were analyzed with the parseFEP plugin of VMD.[32] The Bennet acceptance ratio was used to calculate the error in DDG.[33] Experimental details: S. cinnamonensis A495 (BHDCIBIBII)[8] and A495-ER20[9] was grown in tryptic soy broth (Fluka Analytical) at 30 8C. SM16 medium (MOPS (20.9 g L1), l-proline (10 g L1), glucose (20 g L1), NaCl (0.5 g L1), K2HPO4 (2.10 g L1), EDTA (0.25 g L1), MgSO4·7 H2O (0.49 g L1), CaCl2·2 H2O (0.029 g L1)), supplemented with the respective malonyl-SNAC derivative (10, 20, or 30 mm) was inoculated with pre-culture (5 %). SM16 cultures were grown for five days at 30 8C. On days 4 and 5, SM16 cultures were supplemented with XAD16 resin (20 g L1, Sigma–Aldrich). S. cinnamonensis A495 and A495-ER20 were also cultured without derivative as a control. Analysis of fermentation products: For characterization by HPLCESI-MS, cell paste and XAD16 resin obtained by centrifugation of the fermentation cultures were extracted with ethyl acetate. The solvent was evaporated, and the residue was dissolved in methanol (0.5 mL). Biological profiling of premonensin and its derivatives: Escherichia coli Rosetta 2 (Novagen) was used for expression of PDEd as described.[34] . Cells were grown in TB medium supplied with ampicillin (100 mg/ml) and chloramphenicol (170 mg/ml) at 25 8C. Protein expression was induced by addition of IPTG (100 mm), and cells were grown at 18 8C overnight; the resulting protein was purified by nickel affinity chromatography followed by size exclusion chromatography.[34] Fluorescence polarization: Fluorescence polarization measurements were carried out at 20 8C in Tris·HCl (30 mm, pH 7.5) containing NaCl (150 mm) and DTT (3 mm). Data were recorded by using a Fluoromax-4 spectrophotometer (HORIBA Jobin Yvon, Munich, Germany). For fluorescein-labeled RheB peptide, excitation and 2014 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim
www.chembiochem.org emission wavelengths of 495 nm and 520 nm, respectively, were used. Data analysis was performed with Grafit 5.0 (Erithacus Software, East Grinstead, UK).[34] Stock solutions of premonensin B (10 mm, 2 a) and propargyl-premonensin (10 mm, 5 a) were prepared in Milli-Q water/DMSO (20:80, v/v) supplemented with CAVASOL W7 HP Pharma (16.8 mg mL1; Applichem) and methanol (5 mL to pre-dissolve the compound before dilution into DMSO/H2O). Equivalent concentrations of Milli-Q water and DMSO supplemented with methanol and CAVASOL W7 HP Pharma (16.8 mg mL1) served as negative control. More details of the experimental and computational procedures are in the Supporting Information.
Acknowledgements E.S.-G. acknowledges a Liebig-stipend from the Fonds der Chemischen Industrieand the support of the SFB1093 funded by the Deutsche Forschungsgemeinschaft. A F.I.-A. is a fellow of the International Max Planck Research School of Chemical Biology. This work was supported by the Cluster of Excellence RESOLV (EXC 1069) funded by the Deutsche Forschungsgemeinschaft. Keywords: computational chemistry · free energy calculations · molecular dynamics · monensin · polyketide biosynthesis · QM/MM
[1] Z. Xu, L. Ding, C. Hertweck, Angew. Chem. Int. Ed. 2011, 50, 4667 – 4670; Angew. Chem. 2011, 123, 4763 – 4766. [2] M. C. Walker, B. W. Thuronyi, L. K. Charkoudian, B. Lowry, C. Khosla, M. C. Y. Chang, Science 2013, 341, 1089 – 1094. [3] B. J. Dunn, C. Khosla, J. R. Soc. Interface 2013, 10, 20130297. [4] B. J. Dunn, D. E. Cane, C. Khosla, Biochemistry 2013, 52, 1839 – 1841. [5] U. Sundermann, K. Bravo-Rodriguez, S. Klopries, S. Kushnir, H. Gomez, E. Sanchez-Garcia, F. Schulz, ACS Chem. Biol. 2013, 8, 443 – 450. [6] I. Koryakina, J. McArthur, S. Randall, M. M. Draelos, E. M. Musiol, D. C. Muddiman, T. Weber, G. J. Williams, ACS Chem. Biol. 2013, 8, 200 – 208. [7] M. Oliynyk, C. B. W. Stark, A. Bhatt, M. A. Jones, Z. A. Hughes-Thomas, C. Wilkinson, Z. Oliynyk, Y. Demydchuk, J. Staunton, P. F. Leadlay, Mol. Microbiol. 2003, 49, 1179 – 1190. [8] A. Bhatt, C. B. W. Stark, B. M. Harvey, A. R. Gallimore, Y. A. Demydchuk, J. B. Spencer, J. Staunton, P. F. Leadlay, Angew. Chem. Int. Ed. 2005, 44, 7075 – 7078; Angew. Chem. 2005, 119, 7237 – 7240. [9] S. Kushnir, U. Sundermann, S. Yahiaoui, A. Brockmeyer, P. Janning, F. Schulz, Angew. Chem. Int. Ed. 2012, 51, 10664 – 10669; Angew. Chem. 2012, 124, 10820 – 10825. [10] S. Klopries, U. Sundermann, F. Schulz, Beilstein J. Org. Chem. 2013, 9, 664 – 674. [11] Y. Sugimoto, L. Ding, K. Ishida, C. Hertweck, Angew. Chem. Int. Ed. 2014, 53, 1560 – 1564; Angew. Chem. 2014, 126, 1586 – 1590. [12] R. Breinbauer, I. R. Vetter, H. Waldmann, Angew. Chem. Int. Ed. 2002, 41, 2878 – 2890; Angew. Chem. 2002, 114, 3002 – 3015. [13] G. Zimmermann, B. Papke, S. Ismail, N. Vartak, A. Chandra, M. Hoffmann, S. A. Hahn, G. Triola, A. Wittinghofer, P. I. H. Bastiaens, H. Waldmann, Nature 2013, 497, 638 – 642. [14] A. Kirschning, F. Hahn, Angew. Chem. Int. Ed. 2012, 51, 4012 – 4022; Angew. Chem. 2012, 124, 4086 – 4096. [15] Y. Tang, A. Y. Chen, C.-Y. Kim, D. E. Cane, C. Khosla, Chem. Biol. 2007, 14, 931 – 943. [16] Y. Tang, C.-Y. Kim, I. I. Mathews, D. E. Cane, C. Khosla, Proc. Natl. Acad. Sci. USA 2006, 103, 11124 – 11129. [17] M. A. Mart-Renom, A. C. Stuart, A. Fiser, R. Snchez, F. Melo, A. Sˇali, Annu. Rev. Biophys. Biomol. Struct. 2000, 29, 291 – 325.
ChemBioChem 0000, 00, 1 – 8
&6&
These are not the final page numbers! ÞÞ
CHEMBIOCHEM FULL PAPERS [18] J. C. Phillips, R. Braun, W. Wang, J. Gumbart, E. Tajkhorshid, E. Villa, C. Chipot, R. D. Skeel, L. Kal, K. Schulten, J. Comput. Chem. 2005, 26, 1781 – 1802. [19] A. D. Mackerell Jr., M. Feig, C. L. Brooks III, J. Comput. Chem. 2004, 25, 1400 – 1415. [20] W. L. Jorgensen, J. Chandrasekhar, J. D. Madura, R. W. Impey, M. L. Klein, J. Chem. Phys. 1983, 79, 926 – 935. [21] V. Zoete, M. A. Cuendet, A. Grosdidier, O. Michielin, J. Comput. Chem. 2011, 32, 2359 – 2368. [22] U. Essmann, L. Perera, M. L. Berkowitz, T. Darden, H. Lee, L. G. Pedersen, J. Chem. Phys. 1995, 103, 8577 – 8593. [23] A. D. Becke, J. Chem. Phys. 1992, 97, 9173 – 9177. [24] S. Grimme, J. Comput. Chem. 2006, 27, 1787 – 1799. [25] A. Schfer, H. Horn, R. Ahlrichs, J. Chem. Phys. 1992, 97, 2571 – 2577. [26] S. R. Billeter, A. J. Turner, W. Thiel, Phys. Chem. Chem. Phys. 2000, 2, 2177 – 2186.
2014 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim
www.chembiochem.org [27] R. Ahlrichs, M. Br, M. Hser, H. Horn, C. Kçlmel, Chem. Phys. Lett. 1989, 162, 165 – 169. [28] W. Smith, T. R. Forester, J. Mol. Graph. 1996, 14, 136 – 141. [29] ChemShell, a Computational Chemistry Shell, http://www.chemshell.org. [30] P. Kollman, Chem. Rev. 1993, 93, 2395 – 2417. [31] R. W. Zwanzig, J. Chem. Phys. 1954, 22, 1420 – 1426. [32] P. Liu, F. Dehez, W. Cai, C. Chipot, J. Chem. Theory Comput. 2012, 8, 2606 – 2616. [33] C. H. Bennett, J. Comput. Phys. 1976, 22, 245 – 268. [34] D. Wtzlich, I. Vetter, K. Gotthardt, M. Miertzschke, Y.-X. Chen, A. Wittinghofer, S. Ismail, EMBO Rep. 2013, 14, 465 – 472.
Received: April 30, 2014 Published online on && &&, 0000
ChemBioChem 0000, 00, 1 – 8
&7&
These are not the final page numbers! ÞÞ
FULL PAPERS K. Bravo-Rodriguez, A. F. Ismail-Ali, S. Klopries, S. Kushnir, S. Ismail, E. K. Fansa, A. Wittinghofer, F. Schulz,* E. Sanchez-Garcia* && – && Predicted Incorporation of Non-native Substrates by a Polyketide Synthase Yields Bioactive Natural Product Derivatives
2014 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim
PKS analysis: Which substrate is incorporated, how does it happen and what is produced? Computational modeling of the fifth acyltransferase domain of a polyketide synthase predicts that the enzyme incorporates non-native building blocks into the polyether ionophore monensin to give natural product derivatives with biological activity.
ChemBioChem 0000, 00, 1 – 8
&8&
These are not the final page numbers! ÞÞ
Predicted Incorporation of Non-native Substrates by a Polyketide Synthase Yields Bioactive Natural Product Derivatives Kenny Bravo-Rodriguez,[a] Ahmed F. Ismail-Ali,[b] Stephan Klopries,[b] Susanna Kushnir,[b] Shehab Ismail,[c, d] Eyad K. Fansa,[c] Alfred Wittinghofer,[c] Frank Schulz,*[b] and Elsa SanchezGarcia*[a] The polyether ionophore monensin is biosynthesized by a polyketide synthase that delivers a mixture of monensins A and B by the incorporation of ethyl- or methyl-malonyl-CoA at its fifth module. Here we present the first computational model of the fifth acyltransferase domain (AT5mon) of this polyketide synthase, thus affording an investigation of the basis of the relaxed specificity in AT5mon, insights into the activation for the nucleophilic attack on the substrate, and prediction of the incorporation of synthetic malonic acid building blocks by this
enzyme. Our predictions are supported by experimental studies, including the isolation of a predicted derivative of the monensin precursor premonensin. The incorporation of nonnative building blocks was found to alter the ratio of premonensins A and B. The bioactivity of the natural product derivatives was investigated and revealed binding to prenyl-binding protein. We thus show the potential of engineered biosynthetic polyketides as a source of ligands for biological macromolecules.
Introduction Polyketide synthases (PKSs) are natural multienzyme complexes of high relevance for medicine as they synthesize compounds used as immunosuppressants, antibiotics, and anticancer drugs, among others. PKSs catalyze a cascade of Claisen condensations between enzyme-bound acyl thioesters and malonic acid thioesters, which serve as extender units for the growing polyketide chain. In bacterial type I PKSs, acyltransferase (AT) domains serve as gatekeepers to recruit the malonicacid-based building blocks required for chain extension. The mechanism of this process is largely not understood, thus rendering the task of engineering a PKS for the incorporation of non-native building blocks a formidable challenge.[1–6] In this context, elucidation of the molecular mechanisms underlying [a] K. Bravo-Rodriguez,+ Dr. E. Sanchez-Garcia Department of Theoretical Chemistry Max-Planck-Institut fr Kohlenforschung Kaiser-Wilhelm-Platz 1, 45470 Mlheim an der Ruhr (Germany) E-mail: [email protected] [b] A. F. Ismail-Ali,+ S. Klopries, Dr. S. Kushnir, Prof. Dr. F. Schulz Fakultt fr Chemie und Biochemie, Ruhr-Universitt Bochum Universittsstrasse 150, 44780 Bochum (Germany) E-mail: [email protected] [c] Dr. S. Ismail, E. K. Fansa, Prof. Dr. A. Wittinghofer Max-Planck-Institut fr Molekulare Physiologie Otto-Hahn-Strasse 11, 44227 Dortmund (Germany) [d] Dr. S. Ismail Present address: CR-UK Beatson Institute, Garscube Estate Switchback Road,Glasgow G61 1BD (UK) [+] These authors contributed equally to this work. Supporting information for this article is available on the WWW under http://dx.doi.org/10.1002/cbic.201402206.
2014 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim
substrate incorporation into PKS is a task of high importance, but this is currently hindered by the lack of structural information on the megadalton-sized PKSs. The AT domain in module five of the monensin PKS (AT5mon) is known to accept both ethyl-malonyl- and methyl-malonylCoA (EtMCoA and MMCoA) as substrates, thereby giving rise to monensins A and B.[7, 8] Previous mutagenesis experiments have shown the flexibility of this PKS in processing non-native substrates with different redox patterns in the nascent polyketide chain, thus implying opportunities for further biosynthetic derivatization of monensin.[9] Furthermore, our earlier investigations on artificially induced substrate promiscuity in the sixth AT domain of the erythromycin assembly line (AT6DEBS) showed that a non-natural building block can be introduced as a source of a propargylated side chain in erythromycin A by targeted mutagenesis of the AT6DEBS domain.[5] However, it is important to broaden and deepen the understanding of substrate recognition by giant and complex PKSs, in particular of the AT domains, as these are crucial for the introduction of new functional groups and stereocenters.
Results and Discussion Molecular modeling of acyltransferase domain 5 of monensin PKS We built a homology model of the AT5mon domain in order to investigate its substrate recognition mechanism (see Computational details and the Supporting Information). Based on the amino acid sequence of AT5mon, we used the X-ray structures ChemBioChem 0000, 00, 1 – 8
&1&
These are not the final page numbers! ÞÞ
CHEMBIOCHEM FULL PAPERS of AT3DEBS, AT5DEBS, and malonyl CoA/acyl carrier protein transacylase (MAT) as templates (Section I.1 in the Supporting Information). After equilibration, we performed molecular dynamics (MD) simulations of solvated AT5mon without ligand. These simulations indicated that the overall protein structure is conserved. However, in the absence of substrate, the region formed by residues 267–339 was most flexible, as it tended to move away from the rest of the protein. This movement makes the active site of AT5mon more open and exposed to the ligand than in AT6DEBS. MD simulations of AT5mon models with the natural substrates (MMCoA or EtMCoA) in different orientations showed, in all cases, a structured active site able to accommodate the ligands. We found that the main interaction between MMCoA and the active site is the salt bridge formed between the carboxylate group of MMCoA and R257, which is located at the bottom of the active site. The thioester and amide groups of MMCoA further contribute to keeping the ligand in a suitable position within the active site (Section I.2 in the Supporting Information). The simultaneous interactions of the thioester moiety and the closest amide group of MMCoA with the backbone of Q149 help the substrate to adopt a conformation in which S232 is close to the thioester, thus enabling the nucleophilic attack (Figure 1 A). Our work also provides insights into the activation of S232. In all simulations, S232 interacted with H231 (Section I.2 and Table S1 in the Supporting Information) or with the carboxylate group of the ligand, but never with H334. This suggests that either H231 or the carboxylate group (but not H334) acts as the base to promote activation of S232 in AT5mon. By incorporating EtMCoA, AT5mon exhibits a relaxed specificity compared to AT6DEBS, which is highly selective for MMCoA. A comparison of AT5mon with our AT6DEBS model[5] provides interesting insights into possible factors contributing to substrate promiscuity in AT5mon. In AT6DEBS, the YASH sequence strongly interacts with the substrate and together with V295 prevents the incorporation of substrates larger than MMCoA in the active site. We established that the V295A mutation lowers the specificity of AT6DEBS.[5] In AT5mon, the sequence equivalent to
www.chembiochem.org
Figure 1. A) Residue S232 interacts with H231 but not with H334. B) A comparison of AT5mon with AT6DEBS (italics) indicates that the VAGH sequence and S329 might play a role in the substrate promiscuity of the monensin PKS.
YASH is VAGH, and S329 is the residue equivalent to V295 (Figure 1 B). Our simulations indicated that these differences result in a more open active site that is thus able to accept not only MMCoA but also EtMCoA as a substrate. EtMCoA establishes an interactions network with the active site similarly to MMCoA. The most favorable orientation of this substrate is with the ethyl group pointing towards V331 (Figure 2 A). Unlike MMCoA, EtMCoA was found in a location distant from S232 if the ethyl moiety was not in the correct orientation (Section I.2 in the Supporting Information). In all cases, the in-
Figure 2. Binding at the active site of AT5mon by A) EtMCoA (native substrate), B) MCoA, and C) propargyl-MSNAC. The substrates are shown as thicker stick models. Unlike EtMCoA, MCoA lacks many stabilizing interactions with the active site. Propargyl-MSNAC is accommodated in the active site in an orientation similar to that of EtMCoA.
2014 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim
ChemBioChem 0000, 00, 1 – 8
&2&
These are not the final page numbers! ÞÞ
CHEMBIOCHEM FULL PAPERS teraction with R257 was conserved, and thus EtMCoA remained inside the active site. Given the experimental complexity of PKS manipulation, the ability to computationally predict the incorporation of nonnative substrates is important, not only as a guide to experiments but also to understand the fundamental substrate selectivity of PKS. The relaxed specificity of AT5mon relative to AT6DEBS and other AT domains does not automatically establish AT5mon as a very promiscuous enzyme, as MMCoA and EtMCoA differ by only a methylene group. In this context, we wanted to establish whether AT5mon is able to incorporate synthetic building blocks with larger and more rigid moieties than a methyl or ethyl group. Thus, we computationally investigated the incorporation of two ligands of different sizes relative to MMCoA: the smaller natural malonyl-CoA (MCoA) and the larger synthetic propargyl-malonyl-SNAC (propargyl-MSNAC; Figure 2 B and C). Simulations of AT5mon with bound MCoA suggested significantly weakened interaction with the active site: average values of the distance between R257 and the carboxylate moiety of MCoA were large, the substrate was far from S232, and no secondary stabilizing interactions between the ligand and the active site were found (Figure 2 B; Section I.2, and Table S1 in the Supporting Information). This is consistent with experimental observations: in premonensin and monensin biosynthesis, the abundantly available natural MCoA does not lead to the corresponding desmethyl derivatives of the natural products in significant quantities.[9] Propargyl-MSNAC, on the other hand, was significantly better accepted by the active site and exhibited a behavior similar to that of the natural substrate, EtMCoA. However, unfavorable orientations of this ligand led to disruption of several interactions important for substrate binding and activation (with residues such as Q149, Q233, and R257; Section I.2 in the Supporting Information). Our results thus indicate that the insertion of propargyl-MSNAC into the active site depends on the orientation of its triple-bond moiety, which should point toward V331 similarly to the ethyl group in EtMCoA (Figure 2 A and C). In addition to homology modeling and MD simulations we performed quantum mechanics/molecular mechanics (BP86D2/SVP//CHARMM) calculations of the AT5mon–propargylMSNAC system; the results were in very good agreement with the MD results (Supporting Information, Section I.2 and Table S5). The relative free energy differences associated with the transformations MMCoA!EtMCoA and MMCoA!propargyl-MCoA were also calculated. To directly compare the effect of the substituents, we calculated propargyl-MCoA instead of propargyl-MSNAC, which was used in the experiments. As both MMCoA and EtMCoA are natural substrates of AT5mon, DDG should have values close to 0. As expected, the calculated value for the MMCoA!EtMCoA transformation was 0.12 0.02 kcal mol1. The small DDG value for MMCoA!propargylMCoA (0.53 0.04 kcal mol1) means that, thermodynamically, the active site of AT5mon should be able to accommodate the bulky propargyl group as it does with the natural substrates.
2014 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim
www.chembiochem.org Experimental investigation of substrate promiscuity of AT5mon Malonic acid derivatives, each thioester-activated with N-acetylcysteamine (SNAC), were synthesized either from commercially available derivatives of malonic acid diesters or by reductive alkylation of Meldrum’s acid (Scheme 1, Figure 3 A; and Section IV in the Supporting Information).[10]
Scheme 1. Synthesis of compounds 4–8; a 3.0 equiv LiOH·H2O, H2O, 18 h, RT; b) 1.01 equiv isoprenylacetate, 0.06 equiv H2SO4, neat, 18 h, RT; c) 1.0 equiv borane dimethylamine complex, 3.0 equiv propyl, butyl, or hexyl aldehyde, 1 h, RT; d) tBuOH, 6 h, 90–100 8C; e) 1.1 equiv CDI, 0.3 equiv DMAP, 1.2 equiv SNAC, THF, 18 h, RT; f) 2.5 equiv TiCl4, DCM, 6 h, RT. CDI: N,N’-Carbonyldiimidazole, DMAP: 4-dimethylaminopyridine, SNAC: N-acetylcysteamine.
Cultures of Streptomyces cinnamonensis A495 were supplemented with each derivative (10–30 mm). This strain produces premonensin, an artificial shunt-product of the monensin biosynthetic pathway.[8] Analysis of the fermentation extracts by LC-ESI-MS and HRMS indeed revealed significant PKS promiscuity (Section II.3 in the Supporting Information). Four of the five synthetic building blocks were incorporated in significant amounts. 2-Hexylmalonyl-SNAC was found to be toxic for the bacterium and to abolish the fermentation of premonensin. LC-ESI-MS analysis showed that each new premonensin derivative was as a third fermentation product, in addition to premonensins A and B. This indicates specific incorporation of the building blocks by the AT5mon domain. As a general trend, the ChemBioChem 0000, 00, 1 – 8
&3&
These are not the final page numbers! ÞÞ
CHEMBIOCHEM FULL PAPERS
www.chembiochem.org
Figure 3. A) Malonic acid derivatives used in this study (those investigated by molecular modelling are in bold letters). From the synthetic building blocks, analogues 4–7 were experimentally incorporated into premonensin. Compound 5 a was isolated from the fermentation broth and analyzed by NMR. The corresponding propargylated ER20-derivative, however, was not found by LC-ESI-MS. B) LC-ESI-MS total ion chromatogram of a crude fermentation extract of S. cinnamonensis A495 supplemented with 5. C) High-resolution MS results from the same fermentation extract; 5 a is identified by its H + , NH4 + , and Na + adducts. D) A segment of the 1H NMR analysis of premonensin A (3 a, lower trace) in comparison to 5 a (upper trace).
2014 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim
ChemBioChem 0000, 00, 1 – 8
&4&
These are not the final page numbers! ÞÞ
CHEMBIOCHEM FULL PAPERS experiments revealed higher relative incorporation efficiency at lower concentration of the respective malonic acid derivative (Section II.2 in the Supporting Information). Higher concentrations were increasingly toxic to the bacterium. Addition of propyl-malonyl-SNAC nearly abolished the production of premonensin and resulted in only trace amounts of the natural product derivative. Interestingly, the relative amounts of the naturally produced products premonensins A and B appeared to be inverted in the LC-ESI-MS analysis upon addition of the artificial building blocks. For propargyl-MSNAC, the feeding experiment was scaled up to 1.8 L, as its preparative incorporation might give rise to a synthetically useful orthogonal functional group in the polyketide (Section II.4 in the Supporting Information). After purification, the identity of propargyl-premonensin was confirmed by 1H- and 13C NMR (Figure 3 D and Section II.5 in the Supporting Information), thus proving specific incorporation of the building block by AT5mon. The new product was isolated in a yield of 0.55 mg L1 (5.4 mg L1 premonensin A and 1.8 mg L1 premonensin B from the same fermentation). This confirmed the inversion of the incorporation efficiencies of ethyl- and methyl-malonyl-CoA upon addition of the artificial malonic acid derivative (control fermentations yielded 8.2 mg L1 premonensin A and 15.2 mg L1 premonensin B). In further investigation of the substrate flexibility of the PKS, all artificial building blocks except butylmalonyl-SNAC were supplied to cultures of the mutant strain S. cinnamonensis A495-ER20.[9] This strain gives rise to a redox derivative of premonensin (Figure 3 A) and was the result of previous PKS-engineering experiments to yield polyketide redox derivatives.[9] Analysis of these fermentations revealed interesting differences from the wild-type PKS and indicates “crosstalk” between different segments of the multienzyme complex by unknown mechanisms. As expected, the overall amounts of the polyketide derivatives were significantly lower than for the wildtype PKS. Interestingly, however, propargyl-MSNAC did not give rise to the expected premonensin derivative but instead largely abolished productivity. Previously, a small amount of a biosynthetic byproduct corresponding to the incorporation of MCoA by AT5mon was isolated.[9] Propyl- and allyl-malonylSNAC on the other hand led to significant incorporation levels relative to wild type. These experiments indicate a subtle but effective crosstalk mechanism between module 2 and module 5. Apparently, the differences in the surrounding PKS machinery can induce alterations of the substrate specificity of individual domains or modules. This could be the result of dedicated proofreading mechanisms or simply substrate-induced alteration of catalytic activity.[9, 11] Bioactivity of premonensin and its derivative Polyketides have specific chemical structures to interact with biological macromolecules.[12] Here we found that the non-natural biosynthetic shunt-product premonensin binds tightly to the human phosphopdiesterase 6 delta subunit (PDEd). This protein plays a pivotal role in K-Ras trafficking in human cells, and its inhibition by suitable binders was shown to have cyto 2014 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim
www.chembiochem.org toxic effects on K-Ras-dependent human pancreatic tumor cells.[13] Premonensin B was found to bind to PDEd in a competitive fluorescence polarization assay with farnesylated Rheb peptide, which has been used as a mimic of the K-Ras C-terminal peptide (KD = (214 10) nm, see Figure 4 and Section III in
Figure 4. A 1 mm solution of fluorescently labeled Rheb peptide in the presence of 1 mm PDEd was titrated against increasing concentrations of premonensin B (2 a) and propargyl-premonensin. Premonensin B showed an IC50 of 0.81 mm whereas propargyl-premonensin (5 a) showed an IC50 of 2.00 mm. ~: negative control, &: premonensin B, and *: propargyl-premonensin.
the Supporting Information). Premonensin B was able to displace the natural binder from the complex with PDEd. The propargylated derivative showed a 2.5-fold decreased affinity towards the protein (Figure 4 and Section III in the Supporting Information). This result shows that modified biosynthesis can produce suitable starting compounds for natural-product based structure-activity studies. In this context, the newly attached propargyl moiety will, after further improvement of this precursordirected biosynthesis experiment, offer opportunities for straightforward chemical modification in a chemistry–biology– chemistry reaction sequence.[14]
Conclusion Here we present the first computational model of the AT5mon domain of the monensin PKS; this domain plays an essential role in building-block selection and incorporation. This model allowed us to predict the incorporation of non-native substrates in the wild-type system and to investigate the basis of substrate recognition by this enzyme. Our predictions were corroborated and extended by systematic screening for the incorporation of non-natural malonic acid derivatives as building blocks for this PKS. Important findings are that the inclusion of artificial building blocks seemed to reverse the relative amounts of the shunt products premonensin A and B. Moreover, premonensin B and propargyl-premonensin (obtained after theoretically predicting the incorporation of propargylMSNAC by wild-type AT5mon) bound to PDEd. The incorporation of the new propargyl moiety into the structure of premonensin opens perspectives for the discovery of further bioactive derivatives through targeted biosynthetic derivatization and chemical conversions. ChemBioChem 0000, 00, 1 – 8
&5&
These are not the final page numbers! ÞÞ
CHEMBIOCHEM FULL PAPERS Experimental Section Computational details: Based on its amino acid sequence, the homology model of the AT5 domain of the monensin PKS (AT5mon) was generated from the X-ray structures of DEBS domains AT3DEBS (PDB ID: 2QO3)[15] and AT5DEBS (PDB ID: 2HG4)[16] and of malonylCoA:acyl carrier protein transacylase (MAT, to be published) as templates. The program Modeller (v. 9.10) was used.[17] The program NAMD2.9[18] was used for all MD simulations with the CHARMM22 force field[19] for the protein and the TIP3P model[20] for water. Parameters for the substrates were generated by Swissparam[21] and previously tested by us.[5] The PME method was used for the treatment of the electrostatic interactions.[22] QM/MM (BP86-D2/SVP//CHARMM22)[23–25] optimizations were also used to study the interaction between propargyl-MSNAC and the active site. Snapshots taken at the beginning and end of the simulations were optimized and compared for an additional assessment of the evolution of key contacts between the substrate and residues in the active site. The QM regions were defined as the entire substrate. All residues within 30 of the substrate were allowed to move freely during the optimizations. The HDLC optimizer was used.[26] QM calculations were handled with Turbomole5.10;[27] DL_POLY was used for the MM calculations.[28] All QM/MM optimizations were performed with Chemshell3.5.[29] Relative free energy differences (DDG) were calculated to assess the preference of the active site for different lateral chains of the substrates. The transformations of MMCoA in EtMCoA, and propargyl-MCoA were computed by using Free Energy Perturbation theory.[30, 31] The free energy calculations were analyzed with the parseFEP plugin of VMD.[32] The Bennet acceptance ratio was used to calculate the error in DDG.[33] Experimental details: S. cinnamonensis A495 (BHDCIBIBII)[8] and A495-ER20[9] was grown in tryptic soy broth (Fluka Analytical) at 30 8C. SM16 medium (MOPS (20.9 g L1), l-proline (10 g L1), glucose (20 g L1), NaCl (0.5 g L1), K2HPO4 (2.10 g L1), EDTA (0.25 g L1), MgSO4·7 H2O (0.49 g L1), CaCl2·2 H2O (0.029 g L1)), supplemented with the respective malonyl-SNAC derivative (10, 20, or 30 mm) was inoculated with pre-culture (5 %). SM16 cultures were grown for five days at 30 8C. On days 4 and 5, SM16 cultures were supplemented with XAD16 resin (20 g L1, Sigma–Aldrich). S. cinnamonensis A495 and A495-ER20 were also cultured without derivative as a control. Analysis of fermentation products: For characterization by HPLCESI-MS, cell paste and XAD16 resin obtained by centrifugation of the fermentation cultures were extracted with ethyl acetate. The solvent was evaporated, and the residue was dissolved in methanol (0.5 mL). Biological profiling of premonensin and its derivatives: Escherichia coli Rosetta 2 (Novagen) was used for expression of PDEd as described.[34] . Cells were grown in TB medium supplied with ampicillin (100 mg/ml) and chloramphenicol (170 mg/ml) at 25 8C. Protein expression was induced by addition of IPTG (100 mm), and cells were grown at 18 8C overnight; the resulting protein was purified by nickel affinity chromatography followed by size exclusion chromatography.[34] Fluorescence polarization: Fluorescence polarization measurements were carried out at 20 8C in Tris·HCl (30 mm, pH 7.5) containing NaCl (150 mm) and DTT (3 mm). Data were recorded by using a Fluoromax-4 spectrophotometer (HORIBA Jobin Yvon, Munich, Germany). For fluorescein-labeled RheB peptide, excitation and 2014 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim
www.chembiochem.org emission wavelengths of 495 nm and 520 nm, respectively, were used. Data analysis was performed with Grafit 5.0 (Erithacus Software, East Grinstead, UK).[34] Stock solutions of premonensin B (10 mm, 2 a) and propargyl-premonensin (10 mm, 5 a) were prepared in Milli-Q water/DMSO (20:80, v/v) supplemented with CAVASOL W7 HP Pharma (16.8 mg mL1; Applichem) and methanol (5 mL to pre-dissolve the compound before dilution into DMSO/H2O). Equivalent concentrations of Milli-Q water and DMSO supplemented with methanol and CAVASOL W7 HP Pharma (16.8 mg mL1) served as negative control. More details of the experimental and computational procedures are in the Supporting Information.
Acknowledgements E.S.-G. acknowledges a Liebig-stipend from the Fonds der Chemischen Industrieand the support of the SFB1093 funded by the Deutsche Forschungsgemeinschaft. A F.I.-A. is a fellow of the International Max Planck Research School of Chemical Biology. This work was supported by the Cluster of Excellence RESOLV (EXC 1069) funded by the Deutsche Forschungsgemeinschaft. Keywords: computational chemistry · free energy calculations · molecular dynamics · monensin · polyketide biosynthesis · QM/MM
[1] Z. Xu, L. Ding, C. Hertweck, Angew. Chem. Int. Ed. 2011, 50, 4667 – 4670; Angew. Chem. 2011, 123, 4763 – 4766. [2] M. C. Walker, B. W. Thuronyi, L. K. Charkoudian, B. Lowry, C. Khosla, M. C. Y. Chang, Science 2013, 341, 1089 – 1094. [3] B. J. Dunn, C. Khosla, J. R. Soc. Interface 2013, 10, 20130297. [4] B. J. Dunn, D. E. Cane, C. Khosla, Biochemistry 2013, 52, 1839 – 1841. [5] U. Sundermann, K. Bravo-Rodriguez, S. Klopries, S. Kushnir, H. Gomez, E. Sanchez-Garcia, F. Schulz, ACS Chem. Biol. 2013, 8, 443 – 450. [6] I. Koryakina, J. McArthur, S. Randall, M. M. Draelos, E. M. Musiol, D. C. Muddiman, T. Weber, G. J. Williams, ACS Chem. Biol. 2013, 8, 200 – 208. [7] M. Oliynyk, C. B. W. Stark, A. Bhatt, M. A. Jones, Z. A. Hughes-Thomas, C. Wilkinson, Z. Oliynyk, Y. Demydchuk, J. Staunton, P. F. Leadlay, Mol. Microbiol. 2003, 49, 1179 – 1190. [8] A. Bhatt, C. B. W. Stark, B. M. Harvey, A. R. Gallimore, Y. A. Demydchuk, J. B. Spencer, J. Staunton, P. F. Leadlay, Angew. Chem. Int. Ed. 2005, 44, 7075 – 7078; Angew. Chem. 2005, 119, 7237 – 7240. [9] S. Kushnir, U. Sundermann, S. Yahiaoui, A. Brockmeyer, P. Janning, F. Schulz, Angew. Chem. Int. Ed. 2012, 51, 10664 – 10669; Angew. Chem. 2012, 124, 10820 – 10825. [10] S. Klopries, U. Sundermann, F. Schulz, Beilstein J. Org. Chem. 2013, 9, 664 – 674. [11] Y. Sugimoto, L. Ding, K. Ishida, C. Hertweck, Angew. Chem. Int. Ed. 2014, 53, 1560 – 1564; Angew. Chem. 2014, 126, 1586 – 1590. [12] R. Breinbauer, I. R. Vetter, H. Waldmann, Angew. Chem. Int. Ed. 2002, 41, 2878 – 2890; Angew. Chem. 2002, 114, 3002 – 3015. [13] G. Zimmermann, B. Papke, S. Ismail, N. Vartak, A. Chandra, M. Hoffmann, S. A. Hahn, G. Triola, A. Wittinghofer, P. I. H. Bastiaens, H. Waldmann, Nature 2013, 497, 638 – 642. [14] A. Kirschning, F. Hahn, Angew. Chem. Int. Ed. 2012, 51, 4012 – 4022; Angew. Chem. 2012, 124, 4086 – 4096. [15] Y. Tang, A. Y. Chen, C.-Y. Kim, D. E. Cane, C. Khosla, Chem. Biol. 2007, 14, 931 – 943. [16] Y. Tang, C.-Y. Kim, I. I. Mathews, D. E. Cane, C. Khosla, Proc. Natl. Acad. Sci. USA 2006, 103, 11124 – 11129. [17] M. A. Mart-Renom, A. C. Stuart, A. Fiser, R. Snchez, F. Melo, A. Sˇali, Annu. Rev. Biophys. Biomol. Struct. 2000, 29, 291 – 325.
ChemBioChem 0000, 00, 1 – 8
&6&
These are not the final page numbers! ÞÞ
CHEMBIOCHEM FULL PAPERS [18] J. C. Phillips, R. Braun, W. Wang, J. Gumbart, E. Tajkhorshid, E. Villa, C. Chipot, R. D. Skeel, L. Kal, K. Schulten, J. Comput. Chem. 2005, 26, 1781 – 1802. [19] A. D. Mackerell Jr., M. Feig, C. L. Brooks III, J. Comput. Chem. 2004, 25, 1400 – 1415. [20] W. L. Jorgensen, J. Chandrasekhar, J. D. Madura, R. W. Impey, M. L. Klein, J. Chem. Phys. 1983, 79, 926 – 935. [21] V. Zoete, M. A. Cuendet, A. Grosdidier, O. Michielin, J. Comput. Chem. 2011, 32, 2359 – 2368. [22] U. Essmann, L. Perera, M. L. Berkowitz, T. Darden, H. Lee, L. G. Pedersen, J. Chem. Phys. 1995, 103, 8577 – 8593. [23] A. D. Becke, J. Chem. Phys. 1992, 97, 9173 – 9177. [24] S. Grimme, J. Comput. Chem. 2006, 27, 1787 – 1799. [25] A. Schfer, H. Horn, R. Ahlrichs, J. Chem. Phys. 1992, 97, 2571 – 2577. [26] S. R. Billeter, A. J. Turner, W. Thiel, Phys. Chem. Chem. Phys. 2000, 2, 2177 – 2186.
2014 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim
www.chembiochem.org [27] R. Ahlrichs, M. Br, M. Hser, H. Horn, C. Kçlmel, Chem. Phys. Lett. 1989, 162, 165 – 169. [28] W. Smith, T. R. Forester, J. Mol. Graph. 1996, 14, 136 – 141. [29] ChemShell, a Computational Chemistry Shell, http://www.chemshell.org. [30] P. Kollman, Chem. Rev. 1993, 93, 2395 – 2417. [31] R. W. Zwanzig, J. Chem. Phys. 1954, 22, 1420 – 1426. [32] P. Liu, F. Dehez, W. Cai, C. Chipot, J. Chem. Theory Comput. 2012, 8, 2606 – 2616. [33] C. H. Bennett, J. Comput. Phys. 1976, 22, 245 – 268. [34] D. Wtzlich, I. Vetter, K. Gotthardt, M. Miertzschke, Y.-X. Chen, A. Wittinghofer, S. Ismail, EMBO Rep. 2013, 14, 465 – 472.
Received: April 30, 2014 Published online on && &&, 0000
ChemBioChem 0000, 00, 1 – 8
&7&
These are not the final page numbers! ÞÞ
FULL PAPERS K. Bravo-Rodriguez, A. F. Ismail-Ali, S. Klopries, S. Kushnir, S. Ismail, E. K. Fansa, A. Wittinghofer, F. Schulz,* E. Sanchez-Garcia* && – && Predicted Incorporation of Non-native Substrates by a Polyketide Synthase Yields Bioactive Natural Product Derivatives
2014 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim
PKS analysis: Which substrate is incorporated, how does it happen and what is produced? Computational modeling of the fifth acyltransferase domain of a polyketide synthase predicts that the enzyme incorporates non-native building blocks into the polyether ionophore monensin to give natural product derivatives with biological activity.
ChemBioChem 0000, 00, 1 – 8
&8&
These are not the final page numbers! ÞÞ
