Specificity and Versatility in Erythromycin Biosynthesis Rembert Pieper Camilla Kao and Chaitan Khosla" Department of Chemical Engineering Stanford University Stanford CA 94305-5025 USA Guanglin Luo and David E. Cane* Department of Chemistry Brown University Box H Providence RI 02972 USA 1 Introduction More than 40 years ago Woodward and Gerzon suggested that macrolide antibiotics such as erythromycin A (1) might be formed from simple propionate building blocks.'** Since that time not only has this prediction proved to be remarkably perceptive but a wealth of information has been gained concerning the biosynthesis of this medicinally important class of natural products. Incorporation experiments with [ I4C]- [ 13C]- [ 1801-,and [*H]-labelled sub- strates and intermediate analogues have confirmed the propionate origin of erythromycin and related metabolites and established that formation of the parent erythromycin macrolide 6-deoxyery- thronolide B (6-dEB 2) occurs by a processive mechanism in which the oxidation level and stereochemistry of the growing polyketide chain are adjusted immediately after each step of polyketide chain el~ngation."~ Intriguingly although more than 100 individual macrolides made up of various combinations of acetate propionate and butyrate subunits have been identified all of these metabolites can be described by a general stereochemical model illustrated in Figure 1 first proposed by Celmer." Indeed the existence of such striking regularities among a large number of metabolites produced by a wide range of Actinomycete species first suggested the possible modularity of the biosynthetic enzymes responsible for the formation of these polyketide natural products.' The central challenge has therefore been to unravel the mystery that shrouds the molecular genetic and biochemical basis for the intricate programming of the biosynthesis of complex polyketides. 2 Isolation and Characterization of the Genes for a Modular Polyketide Synthase The first direct experimental evidence for the modular hypothesis came from the work of the groups of Peter Leadlay at the University Rembert Pieper was born in 1963 in Osnabruck Germany and graduated from Philipps Universitat Marburg Germany with a degree in pharmacy in 1989. This was followed by a PhD from the Technische Universitat Berlin in I993 on the isolation and char- acterization of enniatin synthetase postdoctoral work (1994-1 996) at Stanford University on the enzymology of polyketide synthases and a current position as research associate at the National institutes of Health Bethesda. Camilla Kao was born in 1970 in Midland MI and graduated in chemical engineering j-om Rice University Houston TX in May 1992. This was followed by graduate school at Stanford University in September 1992 also in chemical engineering and a current position as Lieberman Fellow at Stanford University. Chaitan Khosla (born 1964) is Assistant Professor of Chemical Chaitan Khosla (left) Engineering of Chemistry (by David Cane (right) courtesy) and of Biochemistry 0 It '"'0" ErythromycinA (1) R Celmer Macrolide Model Figure 1 The broad spectrum antibiotic erythromycin A 1 and Celrner's macrolide stereochemical model. of Cambridge and Leonard Katz at the Abbott Laborat~ries.*.~ Working independently these two groups of investigators demon- strated that the structural genes responsible for the formation of 6-dEB consist of three contiguous open reading frames of 10kb each encoding three large (ca. 3000 amino acid) multidomain proteins designated deoxyerythronolide B synthase (DEBS) 1 2 and 3 by Leadlay (Figure 2). Detailed sequence comparisons revealed that each of these proteins consists of eight to ten domains with consid-erable sequence similarity to enzymes responsible for each of the individual steps of fatty acid biosynthesis. Moreover these domains (by courtesy) at Stanford University. He received his PhD in I990 at Caltech. After completing his postdoctoral studies in the lahorn- tory of Professor Sir David Hopwood at the John lnnes Centre in the UK he joined Stanford in 1992. His current research interests focus at the interface of biological chemistry and hiomolecular engineering. Guanglin Luo (horn 1967) graduated from Fudan University in Shanghai with a BSc in 1987. AJter receiving his MSc in organic chemistry in 1990 from the Shanghai Institute of Organic Chemistry Academia Sinica where he obtained the First Grade Award as one of the top four students he entered the PhD pro- gramme in chemistry at Brown University. Upon receiving his PhD in 1996 he was awarded the Potter Prize for the outstanding thesis in chemistry. He is currently a postdoctoral researcher in the laho- ratory of Professor E. J. Corey at Harvard University. David E. Cane (born 1944) is Vernon K. Krieble Professor of Chemistry and Professor of Biochemistry at Brown University. After receiving his PhD in 1971 from Harvard University under Prof. E. J. Corey he carried out postdoctoral research with Professor Duilio Arigoni at the ETH in Zurich. Hejoined the,fucult-y at Brown in 1973. His current research interests are centred on the biosynthesis of natural products with special emphasis on the mechanistic enzymology of isoprenoid und polyketide biosynthesis . 297 CHEMICAL SOCIETY REVIEWS 1996 7f DET ,~ ,D X DEBf module 1 module 3 module 5 module2 module4 module6 7 AT ACP KSAT KR ACP KS AT KR ACP KS AT ACP KS AT DH ER KR A KS AT KR ACP KSAT KR ACP TE I.-is HO I HO HOu HO HO 0 2 Figure 2 Model for the modular organization of 6 deoxyerythronolide B synthase (DEBS) and the biosynthesis of 6-deoxyerythronolide B (2) by DEBS 1 + 2 + 3 Each DEBS subunit carries two complete modules and each of the six modules accounts for one cycle of polyketide chain extension and 0-ketore duction as appropriate The active sites are designated as follows acyltransferase (AT) P-ketoacyl ACP transferase (KS). acyl carrier protein (ACP). p-ketoreductase (KR) dehydratase (DH) enoylreductase (ER) and thioesterase (TE) are arranged such that each protein contains two functional units or modules each of which carries all the requisite catalytic activities for one of six cycles of polyketide chain elongation and reductive modification of the resultant P-ketoacyl thioester The significance of this discovery cannot be overstated Not only did the availability of the structural genes for 6-DEB synthase provide an invaluable tool for much of the subsequent experimental work in the polyke- tide synthase area but the model that emerged from these studies reshaped the thinking about the programming of complex polyke- tide synthases and has provided the conceptual framework for the design of many of the most important experiments carried out over the last several years According to the now widely accepted model (Figure Z) the acyl- transferase (AT) domain at the N-terminus of DEBSl initiates the polyketide chain-building process by transferring the propionyl- CoA primer unit via the pantetheinyl residue of the first acyl carrier protein (ACP) domain to the active site cysteine of the ketosyn- thase of module 1 (KSl) The acyltransferase in module 1 (ATl) loads methylmalonyl-CoA onto the thiol terminus of the ACP domain of module 1 KSI then catalyses the first polyketide chain elongation reaction by decarboxylative acylation of the methyl- malonyl residue by the propionyl starter unit resulting in the for- mation of a 2-methyl-3-ketopentanoyl-ACPthioester The latter intermediate is then reduced by the ketoreductase of module 1 (KR l) giving rise to enzyme-bound (2S,3R)-2-methyl-3-hydroxy-pentanoyl-ACP At this point module 1 has finished its task and the diketide product is transferred to the core cysteine of KS2 where- upon it undergoes another round of condensation and reduction resulting in the formation of the corresponding triketide This process is repeated several times with each module being respon- sible for a separate round of polyketide chain elongation and reduc- tion as appropriate of the resulting P-ketoacyl thioester Finally the thioesterase (TE) at the C-terminus of DEBS3 is thought to catalyse release of the finished polyketide chain by lactonization of the product generated by module 6 Following the characterization of the DEBS genes Leadlay and coworkers succeeded in purifying the corresponding three DEBS proteins from the natural erythromycin producer Sac-charopolyspora erythraea The three proteins were as predicted from the DNA sequence of unusually large size -DEBSl (M 370000) DEBS2 (M 380000) and DEBS3 (M 330000) Partial proteolysis studies further established that propionyl-CoA specifi- cally acylates the N-terminal domain of DEBS 1 consistent with the proposed role of this region in loading the propionate starter on the polyketide synthase I1 In a very important set of experiments the Cambridge group also established that (2S)-methylmalonyl-CoA is the exclusive substrate for polyketide chain elongation based on the stereospecific acylation both of intact DEBS proteins and of selected partial proteolytic fragments I2 Unfortunately neither the native protein preparations isolated from Sac erythraea nor recombinant derivatives expressed in Escherichia coli which lacked the requisite pantetheinyl moieties were able to catalyse polyketide chain elon- gation In fact until recently there had been no reports of successful cell-free synthesis of macrolide-type polyketides in spite of more than 30 years of intense efforts by numerous research groups Very recently the Cambridge group have reported the results of experiments which shed light on the subunit organization of the DEBS multienzyme system Gel filtration and ultracentrifugation experiments both indicate that the individual DEBS proteins (as well as entire modules derived therefrom) are associated as homo- dimers These conclusions were reinforced by crosslinking experi- ments in which purified module 5 obtained by partial elastase digestion of DEBS3 was crosslinked with 1,3-dibromopropanone a reagent previously used to crosslink the sulfhydryl residues of the 4'-phosphopantetheine of the ACP domain and the active site cys- teine of the ketosynthase in yeast and animal fatty acid synthases In a control experiment an elastase fragment representing module 6 but lacking its ACP did not undergo dimerization upon addition of 1,3-dibromopropanone These results suggest that the ACP of one module interacts with the KS from its identical partner within each homodimeric unit 3 Genetic Manipulation of DEBS The availability of the cloned DEBS structural genes opened the door to genetic modification of the DEBS proteins themselves In a SPECIFICITY AND VERSATILITY IN ERYTHOMYCIN BIOSY NTHESIS-R PIEPER ET AL pioneering experiment Katz and his coworkers generated a Sac erythraea mutant carrying a large in-frame deletion in the ketore- ductase domain of DEBS module S (KRS) and demonstrated that this mutant produced erythromycin analogues derived from 3 with a keto group at the predicted site C-5' (Figure 3) The latter exper- iment provided not only direct experimental verification of the modular hypothesis suggested by the DEBS gene sequences but established that the downstream domains in module 6 were capable of processing modified polyketide chain-elongation intermediates thereby opening up the exciting possibility of rationally engineer- ing the production of novel polyketide metabolites The Katz group also effected a similar reprogramming of polyketide synthesis by mutation of the presumed NADPH binding motif of the enoyl reductase domain of module 4 (ER4) The resulting mutant strain produced macrolides derived from 4 with the predicted A6 7-anhy-droerythronolide skeleton l4 0 3 4 Figure 3 Novel analogues of 6-dEB produced by DEBS mutants carrying a deletion in KR5 (compound 3) or a mutation in ER4 (compound 4) In 1994 using a specially engineered host-vector system for the expression of recombinant polyketide synthases we succeeded in expressing the complete set of DEBS structural genes in an actino- mycete host species Streptomyces coeficolor,which normally pro- duces neither erythromycin nor any other macrolides Is The resultant strain designated S coeficolor CH999/pCK7 produced substantial quantities (> 40 mg dm ?) of 6-deoxyerythronolide B 2 accompanied by a novel cometabolite 8,8a-deoxyoleandolide 5 (> 10 mg dm ') (Figure 4a) Analysis of the protein constituents of S coeficolor CH999/pCK7 by sodium dodecyl sulfate -polyacryl-amide gel electrophoresis (SDS-PAGE) revealed the presence of three characteristically large proteins DEBS 1 2 and 3 The pro- duction of 6-dEB demonstrates that DEBS I 2 and 3 carry all the necessary biosynthetic activities to support generation of the full- length polyketide and cyclization to the aglycone 2 Furthermore it is evident that ancillary activities required for the essential phos- phopantetheinylation of the ACP domains are present in the host strain and that the recombinant DEBS is fully functional in the het- erologous host That 6-dEB is being formed by the normal biosyn- thetic pathway was confirmed by the incorporation of [ 1-13C~pr~pionate,giving rise to the expected labelling pattern in the 13CNMR spectrum of the resultant sample of 6-dEB In an anal- ogous experiment the starter unit of 8,8a-deoxyoleandolide 5 was labelled by 1 1,2-I3C2jacetate thereby confirming that DEBS can tolerate an acetate starter in place of its normal propionyl-CoA sub- strate and re-emphasizing a potential catalytic flexibility first sug- gested by the molecular genetic experiments of Katz The utilization of acetyl-CoA as a starter is presumably due to the lower intracellular concentration of propionyl-CoA in S coelicofor as compared to the native erythromycin producer Sue er>thraea Together these results have raised intriguing possibilities for the use of PKS systems for the controlled formation of 'unnatural' natural products by rational control of the modular composition of the PKS as well as the structures and relative amounts of the avail- able substrates In further experiments with S coelicolor CH999/pCK7 feeding of (2S,3R)-I 2,3- I 3C I-2-methyl-3-hydroxypentanoyl-N-acetylcys-teamine (NAC) thioester 6 to the engineered organism resulted in the formation of 6-dEB 2 labelled with I 3C at C- 12 and C- 13 as evi- denced by the appearance of the predicted set of enhanced and coupled doublets consistent with the intact incorporation of the diketide chain elongation intermediate'" (Scheme 1) The level of 0 0 - SNAC CH999/pCK7 "'OH I ' 6 O.i\/ 1 "'OH 2 Scheme 1 Incorporation of the intact chain elongation intermediate 6 into 6 dEB 2 by S corlicolor CH999/pCK7 enrichment (1 5-20 atom%) was especially noteworthy. being nearly 100 times more efficient than the levels usually observed for the incorporation of NAC thioesters into microbial polyketide metabolites Whatever the levels of incorporation however it is evident that the DEBS protein can recognize the relevant structural and stereochemical features of the exogenously administered NAC thioester and load the intermediate analogue on to the appropriate ketosynthase domain presumably KS2 from where it will be processed in the normal manner These results as well as the intact incorporation of advanced polyketide chain elongation intermedi- ates into a wide variety of other polyketides strongly suggest that superimposed on the purely organizational level of control over the programming of polyketide biosynthesis intrinsic to the sequential organization of the modular DEBS proteins there is an additional level of substrate molecular recognition exercised by the various catalytic domains In order to explore further the function of the modular DEBS proteins we next constructed a plasmid pCK9 carrying only the genes for the first open reading frame of the DEBS clusteri7 (Figure 4b) When this plasmid was used to transform S c odic olor CH999 protein extracts of the resultant recombinant strain were found to contain a protein shown to be identical with DEBS 1 by a variety of methods Moreover S coelicolor CH999/pCK9 pro- duced 1-3 mg dm ? of a triketide lactone (2R,X,4SSR)-2,4- dimethyl-3,5-dihydroxy-vi-heptanoicacid CZ-lactone 7 the structure of which was unambiguously established by direct spec- troscopic and chromatographic comparison with an authentic sample of 7 prepared by total synthesis Feeding of [I -I 3C]propi-onate resulted in the formation of 7enriched as expected at C-1 C-3 and C-5 These results established for the first time that the DEBSl protein is fully competent to support the first two cycles of polyketide biosynthesis involved in the formation of 6-dEB and does not require association with either DEBS2 or DEBS3 for activity Interestingly 7,which is generated by lactonization of the corresponding acyclic triketide intermediate attached to the ACP of module 2 has previously been reported by Katz to be an abortive chain-elongation product generated by the DEBS con- struct carrying the deletion in KR5 It is apparent that the triketide lactone can be released from DEBSl alone without a requirement for the thioesterase (TE) domain In order to analyse the substrate specificity and function of the thioesterase we next constructed yet another mutant CH999IpCK12 which expresses a PKS DEBS 1 + TE in which the thioesterase. originally at the C-terminus of DEBS3 has now been fused to the C-terminus of DEBSl (Figure 4c) In fact the latter strain not only produced substantially enhanced amounts of the j),triketide lactone 7 (> 20 mg dm but up to 10 mg dm of a cometabolite originating from incorporation of an acetate starter (2R,3S,4S,5R)-2,4-dimethyl-3,5-dihydroxy-vz-hexanoicacid 6-lactone 8 As with the formation of 8,8a-deoxyoIeandolide 5 from an acetate starter the production of 8 most likely reflects some combination of the limited availability of propionyl-CoA in the S coelicolnr host strain as well as the greater abundance of the alter- native substrate acetyl-CoA Meanwhile the Cambridge group had reported analogous results CHEMICAL SOCIETY REVIEWS 1996 OH AT ALP KSAT KR ALP KS AT KR ACP Ks AT ALP K9 AT DH ER KR ACP KS AT KR ACP KS AT KR ALP y 2 5 (40mg dm-3) (10 mg dm-3) ‘1,. b) AT ACP KS AT KR ACP KS AT KR A cox,I 7 (3 mg drns) c) AT ACP KS AT KR ACP KS AT KR ACP I ,a\\I! :.:111)) HOd) 1ATACP KS AT KRACP KS AT KR A+I KSAT ACP KSAT DH ERKR A+F~ ”‘OH 9-(20 mg dmA3) Figure 4 Polyketides produced by engineered S corlrcolor CH999 strains (a) CH999IpCK7 (b) CH999/pCK9 (c) CH999/pCK 12 (d) CH999/pCK 15 on the formation of 7 2o Using genetic recombination techniques in Sac erythrueu they fused the thioesterase domain from DEBS3 to the C-terminus of DEBSl Through a second recombination they also deleted the structural genes for DEBS2 and DEBS3 Each of the resultant mutants produced the expected triketide lactone 7 while neither produced the macrolactone 6-dEB indicating that not only was the thioesterase capable of cyclizing the normal triketide product of DEBSl but that this process completely suppresses transfer of the acyclic triketide intermediate to the downstream enzymes By contrast a mutant carrying an inactive TE fused to the C-terminus of DEBS 1 produced significantly lower levels of trike- tide lactone (<0 I mg dm 3 These results not only confirmed that DEBS 1 is fully competent to support the formation of triketide but that the thioesterase can play an active role in catalysing the release of this substrate In a further series of experiments the Cambridge group also expressed a DEBSl + TE construct in S coelicolor 21 Consistent with our own results with DEBSl + TE in this organism both the triketide lactone 7 and the C,-lactone 8 were produced but the natural tripropionate triketide 7 was found to be the minor con- stituent The reasons for these differences in product ratios is unclear given the more than 30-fold preference of DEBS 1 + TE for a propionyl-CoA over an acetyl-CoA primer (see below) but most likely reflects differences in timing and levels of protein expression as well as the varying sizes of intracellular pools of the two primers and the methylmalonyl extender under the growth conditions used for S coelicolor by the two research groups Our group has constructed yet another deletion mutant of DEBS expressed by S coelicolor CH999/pCK 15,containing the first four DEBS modules present as DEBSl and DEBS2 plus a chimeric fifth module in which the ACP6 TE didomain region has now been fused just downstream of KR5I9 (Figure 4d) Assuming that KS5 and ACP6 could productively interact to catalyse a fifth condensa- tion it was expected that the mutant strain would be able to support five complete rounds of polyketide chain elongation and that the thioesterase/cyclase would catalyse the lactonization of the resul- tant polyketrde Indeed S coelicolor CH999/pCK 15 produced 20 mg dm of a completely new macrolactone (8/?,9S)-8,9-dihydro- 8-methyl-9-hydroxy- 10-deoxymethynolide 9 The structure of 9 which is a novel analogue of 10-deoxymethynolide the aglycone of the macrolide antibiotic methymycin was unambiguously con- firmed by extensive 2-D NMR and mass spectrometric analysis supplemented by specific labelling by a variety of 1 i3C]enriched propionate precursors The latter results established that the TE activity which naturally supports the formation of a 14-membered ring product catalysed exclusive formation of the 12-membered ring lactone by esterification of the acyl thioester with the C-11 hydroxy in preference to the ordinarily kinetically favoured gener- ation of a &lactone by esterificatron with the C-5 hydroxy These results further confirmed the structural and functional independence of individual modules of modular PKSs They also demonstrated the feasibility of constructing hybrid modules viu genetic engineer- ing 4 Cell-free Formation of Polyketides Unlike the related peptide synthetases for which well-developed enzymological methods have been available for the past several decades,22 detailed mechanistic studies on PKSs have been seri- ously hampered by a lack of fully active cell-free systems Indeed until quite recently 6-methylsalicylic acid synthase and its closely related homologue orsellinic acid synthase were the only known examples of microbial PKSs the activities of which (Including chain-elongation activity) had been reconstituted in vitro 23-25 The situation has changed dramatically over the past few years with several reports demonstrating enzymatic synthesis of pol yketrdes using the PKS responsible for the pol yketide component of cyclosporin?6 the tetracenomycin synthase?’ truncated forms of SPECIFICITY AND VERSATILITY IN ERYTHOMYCIN BIOSY NTHESIS-R PIEPER ET AL modular PKSS,~~ 29 and the complete DEBS assembly 28 In at least the last three cases the development of fully active cell-free systems has benefited from the availability of high-level expression systems derived via genetic engineering As described above DEBS 1,2 and 3 are large multifunctional proteins carrying a total of at least 28 distinct active sites The DEBS proteins were present in a cell-free protein preparation from S coelicolor CH999/pCK7 at a level of ca 3-5% total cellular protein 3o In the presence of 1-50 mmol dm-3 sodium phosphate buffer the multi-enzyme assembly was found to catalyse the for- mation of 6-dEB 2 as well as the abortive chain elongation product 7 upon addition of propionyl-CoA (2RS)-methylmaIonyl CoA and NADPHz8 (Scheme 2a) A high phosphate concentration in the protein preparation and reaction buffers was found to be very important for the observed enzymatic activity presumably by enhancing the assembly of the multi-enzyme complex via hydrophobic interactions z8 3o The DEBS-catalysed formation of 6-dEB as well as the formation of the triketide lactone 7 described below is inhibited by both cerulenin and N-ethylmaleimide each a well-known inhibitor of the condensation reactions of fatty acid biosynthesis 28 The apparent k, parameters for the formation of 6-dEB and the triketide lactone 7by DEBS 1,2 and 3 are 0 5 and 0 23 rnin I respectively pointing to the relative inefficiency of chain transfer from DEBSl to DEBS2 in vitro The complete DEBS system has been substantially purified (to > 50% homogeneity) without significant losses in specific activity The individual recom- binant proteins behave as expected as homodimers upon gel filtra- tion Analogous to the above studies with the complete DEBS system similar (and more extensive) investigations have also been carried out on the formation of the triketide lactone 7 by the truncated DEBS1 + TE protein (Figure 2b) Despite the relative simplicity of the system it harbours key molecular recognition features of the overall DEBS system the first two modules generate methyl- branched carbon centres as well as secondary alcohols with both D and L stereochemistry Furthermore this mini-PKS has been shown to be highly active as judged by a k, value (3 4 min I) that is com- parable to the estimated rate constant in VIVO~~and by the fact that products can be synthesized on scales that facilitate structural analysis via NMR spectroscopyzR The apparent K, for methyl- malonyl-CoA in DEBSl + TE catalysed synthesis of the tripropi- onate lactone 7 is 24 pmol dm 30 In contrast the K for propionyl-CoA is not easily measured since the enzyme can readily decarboxylate methylmalonyl-CoA (or methylmalonyl-ACP) to 0 MeMal-CoA NADPH * a) FSCoA DEBSl+2+3 2 b) 8SCoA MeMal-CoA NADPH DEBSI+TE t '4 I+ a0 ' 7 c' SNAC;:I"'0H MeMal-CoA NADPH DEBSI+TE rn 't JY',a0 30 1 generate a propionate primer which is turned over into 7 without any effect on the apparent k, 3o In the presence of (2RS)-methyl- malonyl-CoA and NADPH DEBSl + TE could also convert the exogenously added diketide chain elongation intermediate (2S,3R)-2-methyl-3-hydroxypentanoyl-NACthioester 6 to the triketide lactone 7,as verified by both I4C- and 13C-labelling exper- imentsz8 (Scheme 2c) completely consistent with the previously described experiments with intact cells which had demonstrated the incorporation of 6 into 6-deoxyerythronolide B 2 DEBSl + TE has a remarkably broad specificity towards alter- native primer units In addition to propionyl-CoA both acetyl- and butyryl-CoA can serve as surrogate chain initiators giving rise to the corresponding triketide lactones 8 and 10 respectivelyz931 (Scheme 3) Consistent with these observations DEBSl + TE can be acylated by radiolabelled acetyl- propionyl- and butyryl-CoA with comparable efficiency 31 Preincubation of DEBS 1 + TE with iodoacetamide fails to inhibit acylation by propionyI-CoA,28 while partial proteolysis of the labelled protein indicates exclusive acyla tion of the N-terminal portion of the protein containing the first AT domain Notwithstanding this tolerance for different starter units however DEBS 1 + TE exhibits a 32-fold and an eight-fold kinetic preference towards propionyl primers over acetyl and butyryl primers respectively suggesting the existence of one or more active sites with discriminating molecular recognition features The broad substrate specificity of DEBS towards unnatural sub-strates 1s also illustrated by two other types of experiments First DEBSl + TE can also recognize the unnatural diketide 11 pro-cessing it to the corresponding C triketide lactone (Scheme 3c) Second if NADPH is excluded from the reaction mixture the enzyme can convert propionyl-CoA and methylmalonyl-CoA to the pyran-2-one 12 presumably formed by lactonization of the unre- duced diketoacylthioester product3' (Scheme 4a) Alternatively incubation of DEBSl + TE with methylmalonyl-CoA and the NAC-diketide 6 in the absence of NADPH leads to formation of the ketolactone 1332(Scheme 4b) emphasizing the ability of the DEBS protein to mediate the formation of polyketides in a variety of oxi- dation states 5 Programming and Reprogramming of Modular Polyketide Synthases The example of the erythromycin PKS presents an elegant evolu- tionary solution to the problem of programming a complex sequence of biosynthetic reactions based on the repetitive utilization of a small repertoire of biochemical transformations Two levels of catalytic control are evident in this multi-enzyme assembly First the modular OH /)#,0 MeMal-CoA NADPH a) ACOA + \\" a0 DEBS1+TE 8 y MeMal-CoA NADPH b) DEBS 1+TE NADPH I DEBSl+TE 11 6 6 Scheme 3 Broad substrate specificity of the DEBS enzyme Processing of Scheme 2 Enzyme catalysed formation of (a) 6 dEB catalysed by DEBS 1 anomalous primer substrates (a) acetyl CoA (b) butyryl CoA and (c)11 + 2 + 3 (b) and (c)triketide lactone 7by DEBS1 + TE by DEBS1 + TE CHEMICAL SOCIETY REVIEWS 1996 MeMal-CoA b FXOA DEBS 1-tTE 12 MeMal-CoA L ""IX'""' ,\\" &6'"OH DEBS1+TE ' 13 Scheme 4 Catalysis of polyketide chain elongation by DEBS 1 + TE In the absence of NADPH using a) propionyl-CoA and b) 6 as substrates structure of the proteins provides organizational control at the level of dictating the sequence of reactions to be employed in the overall catalytic cycle Secondly the molecular recognition features of some or all individual domains introduce an additional level of selectivity into the multi-step transformation Importantly neither of these two control mechanisms results in absolute specificity as illustrated vividly by several examples reviewed here Thus the intrinsic toler- ance within modular PKSs towards reprogramming presents an exciting opportunity for the rational design of novel 'unnatural' natural products and for the combinatorial generation of molecular diversity within this medicinally important family of molecules Several diverse and complementary strategies for genetic and/or chemical reprogramming can be envisioned As discussed above a few examples suggest the feasibility of genetically knocking out individual active sites without impairing the remainder of the cat- alytic cycle In an extreme case it may even be possible to delete entire modules as illustrated by experiments in which the terminal thioesterase is fused to various upstream modules Beyond inacti- vation it will be interesting to explore the extent to which individ- ual modules (or domains therein) can be substituted by heterologous modules/domains with unnatural molecular recogni- tion features This has been successfully demonstrated in the cases of the structurally smaller aromatic PKSs as well as the modular peptide synthetases Finally the potential for genetically engineer- ing new catalytic functions into existing PKS pathways also remains to be evaluated In addition to genetic reprogramming the availability of fully active cell-free systems in conjunction with facile mutagenesis tools opens up new possibilities for chemically reprogramming modular PKSs to produce polyketides that might otherwise be inac- cessible viu itz vzvo engineered biosynthesis For example as reviewed above modular PKSs can turn over a variety of unnatural substrates into polyketide products They can also function under non-biological conditions such as in the absence of reducing equiv- alents Given the structural complexity of most natural or engi- neered products of modular PKSs 'one-pot' enzymatic synthetic methodologies could be an attractive complement to established chemical synthesis efforts aimed at elucidating the structure-activ- ity relationships of lead molecules with potential human therapeu- tic veterinary and agrochemical utility Acknowledgments. This work was supported by a grant from the National Institutes of Health (GM22172) to D E C and in part by a grant from the National Institutes of Health (CA 66736-Ol) a National Science Foundation Young Investigator Award and a David and Lucile Packard Fellowship for Science and Engineering Oto C K 6 References 1 R B Woodward Angew Chem 1957,69,50 2 K Gerzon E H Flynn M V Sigal P F Wiley R Monahan and U C Quarck,J Am Chem Soc 1956,78,6398 3 D E Cane H Hasler P B Taylor and T -C Liang Terrahedron 1983 39.3449 4 D E Cane TC Liang P B Taylor C Chang and C C Yang J Am Chem Soc 1986,108,4957 5 D E Cane and C -C Yang,J Am Chem Soc . 1987,109,1255 6 W D Celmer J Am Chem SOC ,1965,87 1801 7 D E Cane. W D Celmer and J W Westley. J Am Chem Sue. 1983 105,3594 8 J Cortes S F Haydwk G A Roberts D J Bevitt and P F Leadlay Nature 1990,348,176 9 S Donadio M J Staver J B McAlpine S J Swanson and L Katz Science 1991,252.675 I0 P Caffrey D J Bevitt J Staunton and P F Leadlay FEBS Lett 1992 304,225 I 1 J F Aparicio P Caffrey A F A Marsden J Staunton and P F Leadlay J Bid Chem 1994,269,8524 12 A F A Marsden P Caffrey,J F Aparicio M S Loughran J Staunton and P F Leadlay Scfence. 1994,263,378 13 J Staunton P Caffrey J F Aparicio G A Roberts S S Bethel1 and P F Leadlay Nuture Struct Biol 1996,3 188 14 S Donadio,J B McAlpine,P J Sheldon,M JacksonandL Katz Proc Nutl Acad SCI USA 1993,90,7119 15 C M Kao L Katz and C Khosla Science 1994,265,509 16 D E Cane,G Luo,C Khosla,C M KaoandL Katz,J Antibiot ,1995 48,647 17 C M Kao G Luo L Katz D E Cane and C Khosla J Am Chem Soc ,1994,116 I1612 18 S Donadio. M J Staver J B McAlpine S J Swanson and L Katz Gene 1992,115,97 19 C M Kao G Luo L Katz D E Cane and C Khosla J Am Chem Soc 1995,117,9105 20 J Cortes,K E H Wiesmann,G A R0berts.M J B Brown,J Staunton and P F Leadlay Science 1995,268 1487 21 M J B Brown J Cortes,A L Cutter P F Leadlay and J Staunton,./ Chem SOC Chem Commun ,1995,1517 22 H Kleinkauf and H von Dohren Eur J Biochem 1990,192,l 23 P Dimroth H Walter and F Lynen Eur J Biochem ,1970,13,98 24 J B Spencer and P M Jordan Biochemi 'fry 1992.31,9 107 25 E R Wo,I Fujii,Y Ebizuka,U Sankawa,A Kawaguchi,S Huang,J M Beale M Shibuya U Mwek and H G Floss. J Am Chem Soc 1989,111,5498 26 M Offenzeller Z Su G Santer H Moser R Traber K Memmert and E Schneider Scherzer J Biol Chem 1993,268,26127 27 B Shen and C R Hutchinson Science 1993,262,1535 28 R Pieper G Luo. D E Cane and C Khosla Nature 1995,378.263 29 K E H Wiesmann. J Cortes M J B Brown A L Cutter. J Staunton and P F Leadlay Chem Biol ,1995,2,583 30 R Pieper S Ebert Khosla D E Cane and C Khosla Biochemistry 1996.35.2054 31 R Pieper G Luo D E Cane and C Khosla J Am Chem Soc ,1995 117. 11373 32 G Luo R Pieper C Khosla and D E Cane Bioorg Med Chem ,1996 4.995