Divergent Mechanistic Routes for the Formation of gem-Dimethyl Groups in the Biosynthesis of Complex Polyketides**

Sean Poust, Ryan M. Phelan, Kai Deng, Leonard Katz, Christopher J. Petzold,* and Jay D. Keasling*

Abstract: The gem-dimethyl groups in polyketide-derived natural products add steric bulk and, accordingly, lend increased stability to medicinal compounds, however, our ability to rationally incorporate this functional group in modified natural products is limited. In order to characterize the mechanism of gem-dimethyl group formation, with a goal toward engineering of novel compounds containing this moiety, the gem-dimethyl group producing polyketide synthase (PKS) modules of yersiniabactin and epothilone were charac- terized using mass spectrometry. The work demonstrated, contrary to the canonical understanding of reaction order in PKSs, that methylation can precede condensation in gem- dimethyl group producing PKS modules. Experiments showed that both PKSs are able to use dimethylmalonyl acyl carrier protein (ACP) as an extender unit. Interestingly, for epothilone module 8, use of dimethylmalonyl-ACP appeared to be the sole route to form a gem-dimethylated product, while the yersinia- bactin PKS could methylate before or after ketosynthase condensation.
Approximately 10% of all approved drugs contain a geminal dimethyl group. The introduction of this group into com-
pounds can decrease their rates of chemical and metabolic degradation, thus improving the efficacy of drugs.[1] Con- sequently, there is great potential utility for the regiospecific incorporation of gem-dimethyl groups in engineered poly-
ketide products. Regiospecific methylation of polyketide products using organic semisynthesis is challenging. To enable biobased approaches to form new gem-dimethyl- containing pharmaceutical agents, and to clarify discrepancies with the current mechanistic understanding, we sought to determine the reaction mechanism of PKS-based C-methyl- ation.[2]
Type I polyketide synthases catalyze Claisen condensa- tions and tailoring reactions in an assembly line fashion to elaborate a remarkably diverse collection of secondary metabolites, many of which have medicinal and industrial applications. Polyketide chains are extended by sequential homologation reactions between ketosynthase (KS)-bound thioesters and a-carboxy building blocks (traditionally malonyl- or methylmalonyl-acyl carrier protein (ACP)) to form b-keto-polyketides. Tailoring reactions, such as the reduction of the b-keto group on the growing polyketide chain, and the subsequent sulfonation or O-methylation of the corresponding b-hydroxy group, customarily follow con- densation. Incorporation of methyl or gem-dimethyl groups by methyltransferase (MT) containing PKS modules is thought to follow this biosynthetic paradigm.[2,3]
The currently accepted view of PKS-based reactions is that KS-mediated condensation precedes all subsequent events that take place within the module, including mono- or dimethylation of the b-ketoacyl-ACP when an MT domain is present (Scheme 1B, Route 1).[2] However, another route is

[*] S. Poust, Prof. Dr. J. D. Keasling
Department of Chemical and Biomolecular Engineering University of California—Berkeley
Berkeley, CA 94270 (USA)
E-mail: [email protected]
Dr. R. M. Phelan, Dr. K. Deng, Dr. C. J. Petzold, Prof. Dr. J. D. Keasling
Joint BioEnergy Institute, Lawrence Berkeley National Lab 5885 Hollis Street, Emeryville, CA 94608 (USA)
Dr. L. Katz, Prof. Dr. J. D. Keasling
Synthetic Biology Engineering Research Center 5885 Hollis Street, Emeryville, CA 94608 (USA)
[**] The authors would like to thank Isu Yoon for assistance with protein purification and Satoshi Yuzawa for helpful discussions. This work was supported by the Joint BioEnergy Institute, which is funded by the Office of Science, Office of Biological and Environmental Research of the U.S. Department of Energy (Contract No. DE-AC02- 05CH11231), by the National Science Foundation (award No. EEC- 0540879 to the Synthetic Biology Research Center), by the Depart- ment of Energy, ARPA-E Electrofuels Program (Contract No. DE- 0000206-1577), and by the National Science Foundation Graduate Research Fellowship Program (Grant No. DGE 1106400, to SP). Supporting information for this article is available on the WWW under http://dx.doi.org/10.1002/anie.201410124.
possible in which methylation precedes the condensation (Scheme 1B, Route 2). A pKa-based argument supports Route 1, as the pKa of a b-ketoacyl-ACP intermediate of Route 1 is about two units lower than the analogous malonyl- ACP of Route 2 and therefore more easily deprotonated.[4] In further support of the proposed Route 1, attack of the enolate on the upstream acyl-KS should occur more readily with the less sterically hindered acetyl enolate, than the more sterically hindered isobutyryl enolate of Route 2. On the other hand, if methylation precedes condensation as in Route 2, the observed production of isobutyryl-ACP in the yersiniabactin PKS (see below) is rationalized by a similar argument that malonyl-ACP is significantly easier to deprotonate than acetyl-ACP (approximate ten unit difference in pKa) and decarboxylation of dimethylmalonyl-CoA would be facili- tated by the presence of electron-donating methyl groups.
There is precedent for the involvement of dimethyl- malonyl moieties in biochemical reactions. When provided in vitro to carboxymethylproline synthase (CarB), a member of the crotonase superfamily, a dimethylmalonyl moiety has
been shown to be a source of a nucleophilic enolate for CtiC bond-forming reactions, despite its steric bulk.[5] Dimethyl-
Angew. Chem. Int. Ed. 2015, 54, 1 – 5 ti 2015 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim 1















Scheme 1. A) Representative gem-dimethyl group containing com- pounds. B) Potential mechanistic routes for the methylation and condensation in gem-dimethyl group producing polyketide synthase domains. Route 1: KS-catalyzed condensation precedes methylation. Route 2: Methylation precedes KS condensation. Acetyl-ACP and isobutyryl-ACP represent potential side products of unproductive decarboxylation (shown in gray). AT = acyltransferase; KS = ketosyn- thase; MT = methyltransferase. R represents the acyl chain transferred from the previous module.

malonyl groups are also naturally added to sugar residues in the biosynthesis of cervimycin,[6] and have been proposed as a source of isobutyryl groups in polyketide synthesis, but the latter has not been conclusively demonstrated.[7]
Examples of polyketides containing gem-dimethyl groups include the potential anticancer agents epothilone, pederin, and bryostatin, and the siderophore yersiniabactin (Sche- me 1A).[8] Currently, the biochemical characterization of gem-dimethyl group producing PKSs is limited to the yersiniabactin PKS. Using an in vitro system, Miller and co- workers[9] showed that the yersiniabactin PKS used malonyl- CoA and two S-adenosyl methionine (SAM) molecules to produce its gem-dimethyl group containing extended product, 3-(hydroxyphenylthiazolinylthiazolinyl) [HPTT]-b-keto-2,2- dimethyl-ACP. They also established that the yersiniabactin AT domain has a 500-fold kinetic preference for malonyl- CoA over methylmalonyl-CoA. Additionally, Mazur and co- workers[10] observed that the condensation reaction was dependent on SAM and that isobutyryl-ACP was a major side product of the reaction; approximately 75% of the ACP domains of the PKS were occupied by an isobutyryl moiety following the reaction. The observed dependence of the condensation reaction on SAM[10] could be explained by the MT domain generating dimethylmalonyl-ACP, which partic-

ipates in the decarboxylative homologation reaction. Epothi- lone PKS module 8 (EpoM8) forms a similar gem-dimethyl group based on the final structure of epothilone, but this module has not been previously studied biochemically. The AT domain of EpoM8 is bioinformatically predicted to prefer methylmalonyl-CoA over malonyl-CoA.[11] This prediction was verified biochemically in a competition assay, though a limited amount of malonyl-ACP was formed when EpoM8 was incubated with malonyl-CoA in the absence of methyl- malonyl-CoA (Figure SI-2.1 in the Supporting Information). Thus, the MT domain in the module is predicted to naturally carry out only a single C methylation. Despite these initial studies, the exact mechanism and biosynthetic timing that PKSs employ to produce gem-dimethyl groups remains unclear.
In order to test if Route 2 is used, we performed in vitro chain elongation reactions with dimethylmalonyl-ACP, gen- erated through directed acylation of the apo-ACP domain using two gem-dimethyl group producing PKS modules: yersiniabactin and EpoM8. In addition, we sought to under- stand if the KS domains of these PKSs demonstrated a preference for the methylation state of the ACP-bound substrate, as reported by Mazur and co-workers.[10] Determi- nation of these parameters would provide a clearer picture of which path(s) are operative in the production of this unusual group. Accordingly, we characterized these gem-dimethyl group producing PKS clusters using a well-established tandem mass spectrometry method[12] that can detect inter- mediates covalently attached to ACP domains in a single- turnover in vitro assay.
The yersiniabactin and EpoM8 PKSs were expressed in Escherichia coli strains BLR and BAP1,[13] respectively. These proteins were purified using Ni-NTA (nickel nitrilotriacetate) chromatography. The purified yersiniabactin PKS was approximately 70% holo form, while the EpoM8 was only the holo form. As both malonyl-CoA and SAM are present in the expression strains, we analyzed the acylation state of each PKS immediately after purification to determine if acylation or methylation occurred during expression. Examination of the freshly purified yersiniabactin PKS and EpoM8 ACPs showed isobutyryl-ACP (a decarboxylation product of dime- thylmalonyl-ACP) on EpoM8 (Figure 1A) and both dime- thylmalonyl-ACP and isobutyryl-ACP appended to the yersiniabactin PKS (Figure 1B). This observation shows that at least the first half of Route 2 is viable for the yersiniabactin PKS. Interestingly, these results suggest that EpoM8 MT is able to methylate malonyl-ACP loaded during in vivo expres- sion in E. coli, even though the preferred substrate for the EpoM8 AT is methylmalonyl-ACP (which is not natively present in E. coli). Furthermore, in vitro studies of the yersiniabactin PKS showed that acetyl-ACP was not methy- lated in the presence of SAM (Figure SI-2.2), contrary to previous proposals, which suggested the methylation of acetyl-ACP accounted for the production of isobutyryl- ACP.[10] These facts strongly suggest that isobutyryl-ACP is derived from dimethylmalonyl-ACP and implies that Route 2 contributes in the overall formation of a gem-dimethyl group.
To unambiguously determine if the dimethylmalonyl moiety loaded on the yersiniabactin and EpoM8 PKSs can
2 www.angewandte.org ti 2015 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim Angew. Chem. Int. Ed. 2015, 54, 1 – 5

These are not the final page numbers!















Figure 1. Analysis of the acylation state of epothilone module 8 PKS (EpoM8 PKS) and yersiniabactin PKS after Ni-NTA purification. Reten- tion times are shown on top of peaks. A) EpoM8 isobutyryl-ACP chromatogram. B) Yersiniabactin PKS dimethylmalonyl-ACP chromato- gram and isobutyryl-ACP chromatogram. All XIC m/z transitions monitored in this study are summarized in the Supporting Informa- tion. XIC = extracted ion chromatogram; cps = counts per second.











Figure 2. Dimethylmalonyl-ACP is extended by the EpoM8 and yersi- niabactin PKS. A) Experiment overview: dimethylmalonyl-ACP was formed using Sfp, dimethylmalonyl-CoA, and apo-PKS. Then, sub- strates to form acyl-KS were added to generate extended products (experimental details in the Supporting Information). Extracted ion chromatogram (XIC) peak areas for dimethylmalonyl-ACP extension experiments for B) EpoM8 and C) yersiniabactin PKS. For the yersinia- bactin case, R represents the acyl chain transferred from the previous NRPS module: a HPTT moiety. All XIC m/z transitions monitored in this study are summarized in the Supporting Information.

To determine if Route 1 is also operative for each PKS, we

be utilized by the KS domain in an in vitro extension reaction, we expressed apo-PKS (expressed in E. coli strain BLR, this resulted in approximately 100% apo form of the EpoM8). We then synthesized authentic dimethylmalonyl-CoA[5,6] and loaded the ACP domains with the dimethylmalonyl moiety using the phosphopantetheinyl transferase Sfp.[14] Loading of dimethylmalonyl-ACP was verified by LC-MS/MS (Fig- ure SI-2.3). The acyl group to be extended was provided to the appropriate KS domain by either a diketide-S-N-acetyl- cysteamine substrate,[15] the acyl group of which mimics the structure of the native intermediate in the case of EpoM8, or by reconstitution of the upstream nonribosomal peptide synthetase (NRPS) enzymes in the case of the yersiniabactin PKS. As shown in Figure 2, both EpoM8 and yersiniabactin extended dimethylmalonyl-ACP in the absence of SAM to produce the correct fully extended and methylated product with retention times matching those observed during fully reconstituted in vitro reactions (Figure SI-2.4). Additionally, the side product isobutyryl-ACP was produced from dime- thylmalonyl-ACP by both enzymes (Figure SI-2.5). A large amount of isobutyryl-ACP was produced by both enzymes relative to the extended methylated product, suggesting that Sfp loading experiments may not fully capture the precisely orchestrated catalytic cycle of these enzymes. The formation of acyl-KS also seems to have activated the KS for decarbox- ylation of dimethylmalonyl-ACP to isobutyryl-ACP (Fig- ure SI-2.5), potentially indicating a conformational shift in the enzyme upon KS acylation.
attempted to form the b-keto unmethylated (yersiniabactin PKS) or monomethylated (EpoM8 PKS) products by incu- bating malonyl-CoA with acyl-KS, while omitting SAM and subsequently adding SAM (Figure 3A). We then examined products before and after SAM addition using LC-MS/MS. Contrary to the results of Mazur et al.[10] the yersiniabactin KS domain was found to use malonyl-ACP as a substrate for condensation in the absence of SAM, and the MT domain methylated the resulting b-keto acyl-ACP moiety after SAM was added (Figure 3B). Small amounts of the unmethylated and singly methylated yersiniabactin-derived PKS product were also observed during reactions with the native malonyl- CoA substrate in the presence of SAM (Figure SI-2.6). EpoM8 was unable to extend methylmalonyl-ACP or malonyl-ACP in the absence of SAM, but it was able to form the dimethyated final product for both extender units in the presence of SAM (Figure 4 and SI-2.7). This suggests that EpoM8 methylates methylmalonyl-ACP and then uses dime- thylmalonyl-ACP as the substrate in the condensation reaction. Additionally, EpoM8 did not form the singly methylated intermediate (i.e. an intermediate resulting solely from the extension of methylmalonyl-ACP with no action by the MT) when full reconstitution reactions were performed in the presence of methylmalonyl-CoA and SAM. Together, these findings demonstrate that in vitro, the yersi- niabactin PKS is able to employ both pathways in Scheme 1, while EpoM8 is restricted to using solely Route 2.
We have clearly demonstrated that the yersiniabactin and epothilone PKSs are capable of forming dimethylmalonyl-
Angew. Chem. Int. Ed. 2015, 54, 1 – 5 ti 2015 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim www.angewandte.org 3
electron microscopy.[16] Future studies will employ cryo electron microscopy to examine the spatial relationship between MT with the ACP during the catalytic cycle to determine the partition between Routes 1 and 2 for the yersiniabactin PKS. Based on our results, we would suggest that PKS engineering strategies to incorporate this group using combinatorial biosynthesis should swap whole modules, instead of swapping individual MT domains into modules unaccustomed to a particular order of reactions. Ultimately, an improved understanding of this remarkable reaction will allow easier integration of this monomer unit in novel, useful PKS-based compounds.
Figure 3. Characterization of Route 1 (KS-catalyzed condensation pre-

cedes methylation) for the yersiniabactin PKS. A) Experiment overview: in part a, SAM was omitted, while malonyl-CoA and acyl-KS were included to allow the formationof unmethylated extended-ACP. In
part b, SAM was added (experimental details in the Supporting Information). B) Unmethylated extended-ACP extracted ion chromato- gram (XIC) peak areas (gray) and methylated extended-ACP peak areas (black). R is a HPTT moiety. All XIC m/z values monitored in this
study are summarized in the Supporting Information.







Figure 4. Characterization of Route 1 (KS catalyzed condensation pre- cedes methylation) for the epothilone PKS. A) Experiment overview: using methylmalonyl-CoA as an extender unit, SAM was omitted from the overall reaction (no product formed), or SAM was included, allowing the complete reaction to take place. B) Unmethylated
extended-ACP extracted ion chromatogram (XIC) peak areas (gray) and methylated extended-ACP peak areas (black). All XIC m/z values monitored in this study are summarized in the Supporting Informa- tion.

ACP and isobutyryl-ACP, its decarboxylated derivative, and, contrary to the traditional understanding of methylation in type I PKSs, that both enzymes can use dimethylmalonyl- ACP as an extender unit in acyl chain elongation. For EpoM8, formation of dimethylmalonyl-ACP through Route 2 appeared to be the only means to form a gem-dimethyl group (at least in vitro), while the yersiniabactin PKS could methylate before or after condensation. The increased decarboxylation of dimethylmalonyl-ACP we observed upon acylation of the KS suggests that conformational changes in the enzyme are important during the catalytic cycle of these gem-dimethyl group producing PKSs, as recently demonstrated for pikromycin module 5 using cryo
Received: October 15, 2014 Revised: November 18, 2014
Published online: && &&, &&&&

Keywords: biosynthesis · dimethylmalonyl-ACP · methylation ·
polyketides · transferases
[1]J. A. Burkhard, G. Wuitschik, M. Rogers-Evans, K. Mtiller, E. M. Carreira, Angew. Chem. Int. Ed. 2010, 49, 9052 – 9067; Angew. Chem. 2010, 122, 9236 – 9251.
[2]A. T. Keatinge-Clay, Nat. Prod. Rep. 2012, 29, 1050 – 1073.
[3]a) M. A. Fischbach, C. T. Walsh, Chem. Rev. 2006, 106, 3468 – 3496; b) C. Hertweck, Angew. Chem. Int. Ed. 2009, 48, 4688 – 4716; Angew. Chem. 2009, 121, 4782 – 4811.
[4]D. A. Evans, “pKa’s of CH bonds in Hydrocarbons and Carbonyl Compounds ”, can be found under http://evans.rc.fas.harvard. edu/pdf/evans_pKa_table.pdf, 2005.
[5]E. T. Batchelar, R. B. Hamed, C. Ducho, T. D. W. Claridge, M. J. Edelmann, B. Kessler, C. J. Schofield, Angew. Chem. Int. Ed. 2008, 47, 9322 – 9325; Angew. Chem. 2008, 120, 9462 – 9465.
[6]T. Bretschneider, G. Zocher, M. Unger, K. Scherlach, T. Stehle, C. Hertweck, Nat. Chem. Biol. 2012, 8, 154 – 161.
[7]J. Young, D. C. Stevens, R. Carmichael, J. Tan, S. Rachid, C. N. Boddy, R. Mtiller, R. E. Taylor, J. Nat. Prod. 2013, 76, 2269 – 2276.
[8]S. Anand, M. V. R. Prasad, G. Yadav, N. Kumar, J. Shehara, M. Z. Ansari, D. Mohanty, Nucleic Acids Res. 2010, 38, W487 – W496.
[9]D. A. Miller, L. Luo, N. Hillson, T. A. Keating, C. T. Walsh, Chem. Biol. 2002, 9, 333 – 344.
[10]M. T. Mazur, C. T. Walsh, N. L. Kelleher, Biochemistry 2003, 42, 13393 – 13400.
[11]B. Julien, S. Shah, R. Ziermann, R. Goldman, L. Katz, C. Khosla, Gene 2000, 249, 153 – 160.
[12]P. C. Dorrestein, S. B. Bumpus, C. T. Calderone, S. Garneau- Tsodikova, Z. D. Aron, P. D. Straight, R. Kolter, C. T. Walsh, N. L. Kelleher, Biochemistry 2006, 45, 12756 – 12766.
[13]B. A. Pfeifer, S. J. Admiraal, H. Gramajo, D. E. Cane, C. Khosla, Science 2001, 291, 1790 – 1792.
[14]L. E. N. Quadri, P. H. Weinreb, M. Lei, M. M. Nakano, P. Zuber, C. T. Walsh, Biochemistry 1998, 37, 1585 – 1595.
[15]K. K. Sharma, C. N. Boddy, Bioorg. Med. Chem. Lett. 2007, 17, 3034 – 3037.
[16]J. R. Whicher, S. Dutta, D. A. Hansen, W. A. Hale, J. A. Chemler, A. M. Dosey, A. R. H. Narayan, K. Hakansson, D. H. Sherman, J. L. Smith, G. Skiniotis, Nature 2014, 510, 560 – 564.



4 www.angewandte.org ti 2015 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim Angew. Chem. Int. Ed. 2015, 54, 1 – 5

These are not the final page numbers!

S. Poust, R. M. Phelan, K. Deng, L. Katz, C. J. Petzold,*
Order of events : In order to elucidate the mechanism of gem-dimethyl group for- mation in polyketides, the gem-dimethyl group producing polyketide synthase

J. D. Keasling* &&& — &&& (PKS) modules of yersiniabactin and
epothilone were characterized using

Divergent Mechanistic Routes for the Formation of gem-Dimethyl Groups in the Biosynthesis of Complex Polyketides
mass spectrometry. The study demon- strated, contrary to the canonical under- standing of reaction order in PKSs, that methylation can precede condensation in PKS modules that produce gem-dimethyl groups.



























Angew. Chem. Int. Ed. 2015, 54, 1 – 5 ti 2015 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim www.angewandte.org 5