Sie sind auf Seite 1von 66

Published by the Royal Society of Chemistry http://pubs.rsc.

org/ej/NP/2001/A909079G/

Natural Product Reports


DOI: 10.1039/a909079g

Polyketide biosynthesis: a millennium review


Review Article
James Staunton and Kira J. Weissman
Department of Chemistry, University of Cambridge, Lensfield Road, Cambridge, CB2 1EW E-mail:
js24@cam.ac.uk or kjw21@cus.cam.ac.uk

Received (in Cambridge, UK) 13th March 2001

Published on the Web 4th June 2001

Contents
1. Introduction
2. Development of the biosynthetic hypothesis
2.1 The polyketide field is revived
3. The application of isotopic labelling to studying biosynthesis
4. Biomimetic syntheses of aromatic polyketides
5. Enzymology of fatty acid and polyketide biosynthesis
5.1 Fatty acid biosynthesis
6. The genetic approach to biosynthetic studies
6.1 Genetics of type I or modular polyketide synthases
6.1.1 The erythromycin polyketide synthase
6.1.1.1 Organisation and function of the erythromycin PKS
6.2 Other type I modular PKSs
7. Type I iterative polyketide synthases
7.1 6-Methylsalicylic acid synthase
7.2 Lovastatin polyketide synthases
8. Structural studies on the type I modular polyketide synthases
8.1 Experiments to probe the structure of the erythromycin PKS
8.2 Proposed structural models for the erythromycin PKS
8.3 Docking of modular proteins
9. Genetic engineering of type I modular polyketide synthases
9.1 Polyketides as targets for combinatorial biochemistry
9.2 Expression of polyketide synthases
9.3 Genetic engineering strategies
9.3.1 Domain mutagenesis
9.3.2 Linking of modules
9.3.3 Combining protein subunits
9.3.4 Mutational biosynthesis (mutasynthesis)
9.3.5 Post-PKS processing
10. Genetic engineering of type I iterative polyketide synthases

1 von 66 20.11.2006 11:04


Published by the Royal Society of Chemistry http://pubs.rsc.org/ej/NP/2001/A909079G/

11. Mechanistic studies in vitro on type I modular polyketide synthases


11.1 Studies in vitro with the DEBS PKS
11.2 Model systems for studying erythromycin biosynthesis: DEBS 1-TE and DEBS 1+TE
11.3 Substrate specificity of DEBS 1
11.4 The molecular basis of stereocontrol
11.5 Studies in vitro of the erythromycin thioesterase
11.5.1 Substrate specificity of the TE
12. Genetic studies on the type II PKS systems
13. Studies in vitro of type II PKSs
14. Summary of the advancements in the understanding of type II polyketide biosynthesis
15. Biosynthesis of plant polyketides
15.1 Structural studies of the chalcone synthases
15.2 Bacterial analogues of CHSs
16. Conclusions
Jim Staunton graduated from Liverpool University in 1959 with a PhD based on elucidation of the
structures of the fungal metabolites, sclerotiorin and rotiorin. After a postdoctoral fellowship in Carl
Djerassi s group at Stanford, California, he returned to the UK to take up a lectureship post at Liverpool
University. In 1969, following a move to a similar post in Cambridge, he began his research into
polyketide biosynthesis. Although he was primarily interested in the elucidation of biosynthetic pathways,
he has made important contributions to biomimetic synthesis and enantiospecific synthesis in this field.
He also developed novel technologies for tracing stable isotopes which had a strong influence on the
development of biosynthetic studies. In recent years he has collaborated with Professor Peter Leadlay in
running a multidisciplinary group of chemists and biologists dedicated to biosynthetic studies of complex
polyketides.

Kira Weissman was born in Stanford, CA in 1973. She studied chemistry at Stanford University, CA,
obtaining a BS in 1995. She then moved to the University of Cambridge, UK where she obtained both her
MPhil (1996) and PhD (1999) in chemistry under the guidance of Professor James Staunton. Her doctoral
research concerned various aspects of erythromycin biosynthesis, particularly the mechanism of
stereocontrol. Kira is continuing her research on aliphatic polyketides as a Junior Research Fellow at
Newnham College, working in the lab of Professor Peter Leadlay.

2 von 66 20.11.2006 11:04


Published by the Royal Society of Chemistry http://pubs.rsc.org/ej/NP/2001/A909079G/

1 Introduction

The polyketide natural products (Fig. 1) are a remarkable class of compounds. In addition to exhibiting a
staggering range of functional and structural diversity, they boast a wealth of medicinally important
activities, including antibiotic, anticancer, antifungal, antiparasitic and immunosuppressive properties.
Even before the full extent of their utility was known, scientists became interested in how these
complicated molecules are assembled. This review highlights the developments of the past 100 years in
the study of polyketides, but to begin, we must delve into the 19th century.

3 von 66 20.11.2006 11:04


Published by the Royal Society of Chemistry http://pubs.rsc.org/ej/NP/2001/A909079G/

Fig. 1 Examples of polyketide secondary metabolites. The type of polyketide and its primary biological activity are
indicated.

In 1893 at London University, James Collie, then a chemist aged just 25, serendipitously discovered a
remarkable set of synthetic reactions. 1 While attempting to prove the structure of dehydroacetic acid 1 by
degradation chemistry (Fig. 2), he boiled it with barium hydroxide and then worked up the reaction with
acid. To his surprise, he discovered that an aromatic compound, orcinol 2, was one of the products.

4 von 66 20.11.2006 11:04


Published by the Royal Society of Chemistry http://pubs.rsc.org/ej/NP/2001/A909079G/

Fig. 2 Collie s synthesis of orcinol 2 from


dehydroacetic acid 1.

Establishing the structure of orcinol was no mean achievement in those early pioneering days. Collie s
analytical data included the melting point and, remarkably, the sweet taste of the product (so much for
health and safety!). 1 Even a simple structure like orcinol was a challenging problem for organic analysis,
but Collie was convinced, and correctly, by his deductive analysis. In an even more amazing leap of
intuition, he came up with an explanation for the mechanism of orcinol formation in terms of a key
polyketone intermediate 3. 1 This intermediate could be formed from the -pyrone starting material by
addition of water and ring opening. It could then spit out water to form a ring as indicated in the
mechanistic scheme (Fig. 2).
Since Collie s pioneering work, the polyketide field has advanced enormously. Not only have
researchers elucidated in full the structures of hundreds of these complicated molecules, but they have also
begun to clone, isolate and characterise the enzymes responsible for their biosynthesis. It is now known,
for example, that in the bacteria and fungi that produce polyketides, Collie chemistry takes place on
polyketo thioesters attached to enzymes called acyl carrier proteins (ACPs). The structure of the ACP
involved in actinorhodin biosynthesis has been established by Tom Simpson and co-workers from the
University of Bristol, UK 2,3 and is given in Fig. 3. Despite these very significant achievements, the
subject of Collie s early speculations the precise mechanism of aromatic ring formation in vivo remains
unknown.

5 von 66 20.11.2006 11:04


Published by the Royal Society of Chemistry http://pubs.rsc.org/ej/NP/2001/A909079G/

Fig. 3 Structure of the actinorhodin ACP based on 1H


NMR.

2 Development of the biosynthetic hypothesis

When Collie originally proposed his ideas on the biosynthesis of polyphenols in 1907 4 organic chemists
had only the haziest idea of reaction pathways. Reactions were rationalised in terms of lasso mechanisms
in which a condensation would be portrayed with departing atoms ringed by dotted lines (Fig. 2). Collie s
realisation that the triketone might be an intermediate in orcinol biosynthesis was inspirational. Even more
daring was his proposal that such intermediates might be generated inside the living cells that produced
such compounds naturally.
Sadly, other influential organic chemists of the day lacked his vision and his hypotheses were deemed to
be not just premature, but presumptuous. Perhaps he was overly ambitious in his conjectures, because he
went on, erroneously, to suggest that the polyketones might be generated from suitable sugars known to be
present in polyketide-producing organisms. Similarly far-fetched hypotheses are broached today by
chemists who speculate on the origins of life. Contemporary chemists, however, are much more receptive
to such theories and so the best of them will probably be investigated. Not so with Collie s ideas they
were buried by the indifference of his contemporaries, and the polyketide field went dead for half a
century.

2.1 The polyketide field is revived

Just how great a loss this was can be appreciated by looking at the history of subsequent theorising on the
biosynthetic pathway to the alkaloids. Robert Robinson proposed his scheme for tropinone 4 biosynthesis
(Fig. 4) 5 in 1917, just ten years after Collie stopped publishing. Perhaps Robinson was more convincing
or charismatic than Collie because his ideas took hold. He went on to dazzle the organic chemical
community for another thirty years, culminating in his book published in 1955. 6 Robinson s hypotheses
focused on alkaloids and terpenes, but, intriguingly, he did give passing mention to the idea that
polyphenols might be produced from polyketones. 6 Was this his original idea, or was he aware of Collie s
earlier speculations?

Fig. 4 Robinson s stunningly simple biomimetic synthesis


of tropinone 4.

The main impetus to the resurgence of interest in polyketides, however, came from Arthur Birch in the
1950s. Birch, like most other leading natural product chemists of the era, had spent a major part of his
early training in Robinson s laboratory at Oxford. His contributions were decisive, for two reasons. First,
he recognised that polyketones could be generated from acetate units by repeated condensation reactions,
and second, he was willing to test his theory by administering an isotopically-labelled version of his
proposed precursor to a suitable polyketide-producing organism. 7
Birch s chosen subject was the aromatic polyketide 6-methylsalicylic acid (6-MSA) 5 (Fig. 5) produced
by the fungus Penicillium patulum. 6-MSA is an intermediate in the biosynthesis of the toxin patulin 6 and
has some importance as an antibiotic in its own right. 8 The mechanistic basis of his proposal is shown in
Fig. 5. Four acetate units are linked head-to-tail to generate a triketo acid 7, and then one of the resulting
keto groups is reduced to a hydroxy group. Generation of a carbanion at the -keto residue would then

6 von 66 20.11.2006 11:04


Published by the Royal Society of Chemistry http://pubs.rsc.org/ej/NP/2001/A909079G/

allow an aldol condensation to form a six-membered carbocyle (the order of these steps is only
hypothetical). Finally, a sequence of plausible dehydration and enolisation reactions would lead to the
aromatic natural product.

Fig. 5 Birch s proposed sequence of reactions in the


biosynthesis of 6-methylsalicylic acid 5.

Birch tested these ideas by feeding acetate labelled with 14C at C-1 (Fig. 6). 7 According to his
hypothesis, four sites in 6-MSA should incorporate the radioactivity as indicated. To confirm his idea, it
was essential to determine the pattern of labelling in the molecule. This analysis entailed laborious
experiments in which samples of the acid were subjected to controlled chemical degradation to produce
fragments that could be reliably correlated with specific sites in the natural product. The three pieces
shown were isolated and subjected to radioactivity measurement to determine their relative molar activity.
The results were consistent with Birch s predicted pattern of incorporation with, importantly, uniform
isotopic enrichment at the expected sites. Rumour has it that Birch experienced a hostile reception to his
ideas from influential members of the British chemical community (shades of Collie!). As a native
Australian working in Britain, he must have derived considerable satisfaction from the publication of these
ground-breaking results in The Australian Journal of Chemistry. 7

7 von 66 20.11.2006 11:04


Published by the Royal Society of Chemistry http://pubs.rsc.org/ej/NP/2001/A909079G/

Fig. 6 Birch s demonstration that 6-methysalicylic acid 5 is


derived from four acetate units.

Birch s paper inaugurated a period of frenetic activity in the natural product community. Feeding of
radiolabelled acetate to many organisms known to produce secondary metabolites yielded structures
exhibiting a measles pattern of isotopic spots. These experiments established not only that many
polyphenolic aromatic molecules were biosynthesised from acetate units according to what came to be
known as the CollieBirch polyketide hypothesis, but also that many non-aromatic compounds were also
formed by further transformations of such products. As an illustration, the pathway from 6-MSA 5 to
patulin 6 is shown in Fig. 7. 8

Fig. 7 Biosynthetic route to the toxin patulin 6 from 6-methylsalicylic acid 5.

3 The application of isotopic labelling to studying biosynthesis

The growth of this field of research was well timed because it coincided closely with the development of
nuclear magnetic resonance (NMR) and mass spectroscopy (MS) as powerful new techniques for the
structural elucidation of complex molecules. Up until 1960, the main tool available to the natural product
chemist for use in structural studies was chemical degradation to produce recognisable fragments. These
pieces were then assembled on paper, often with great leaps of imagination, to produce an idea for the
overall structure. Standard, well-tried methods of degradation had been developed over several decades,
but all of this technology and accumulated knowledge were made redundant almost overnight by the
advent of NMR. A structural study that might take several person-years of intensive work using the
decomposition methodology, might be solved in only a matter of weeks or even days by the new
spectroscopic technologies. Ever resourceful, the natural products community recycled their hard-won
chemical skills to the elucidation of radioactive isotopic labelling patterns.
The reprieve was only temporary, however, because stable isotopic labels that were open to direct

8 von 66 20.11.2006 11:04


Published by the Royal Society of Chemistry http://pubs.rsc.org/ej/NP/2001/A909079G/

detection by NMR spectroscopy, were becoming commercially available. Of particular significance was
the isotope 13C, which like 1H, has a suitable nuclear spin for NMR observation. Inevitably, biosynthetic
investigators turned to this isotope. Labelled precursors were administered in the usual way and then the
natural products isolated. It was then a simple matter to determine the sites of isotopic enrichment by
measurement of the 13C NMR spectrum. In a successful experiment, sites where the isotope was
incorporated would give rise to signals with intensity enhanced over the natural abundance level (1%).
Ideally, for reliable evidence, it is customary to look for at least a doubling of the signal size. Fortunately,
the majority of polyketides are produced in micro-organisms that often readily take up labelled substrates.
NMR spectroscopy proved an especial boon in the polyketide field where molecules are built up from
multiple units of acetate. A simple 1H-decoupled 13C NMR spectrum containing a singlet signal from each
carbon could give the complete labelling pattern and so obviate months of laborious effort in degradation
studies. Care was required to obtain quantitative evidence concerning the uniformity of labelling. It was
also necessary to have a rigorous assignment of the 13C NMR spectrum, which sometimes took a
considerable amount of effort.
Even if only singly-labelled acetate had been available, 13C would have displaced its radioactive cousin
from its dominating position in investigations of polyketide biosynthesis. However, the extra power of
using acetate doubly labelled with 13C made an even greater impact. As long as the enrichment is
sufficiently high (>90%), the dominant signals in a doubly labelled acetate unit are a pair of 13C13C
coupled doublets. This pattern persists in the spectrum of the natural product as long as the relevant
carboncarbon bond remains intact. As has already been mentioned, in many biosynthetic pathways an
initially formed aromatic product undergoes a profound structural reorganisation with CC bond cleavages
and rearrangements to give a complex new skeleton. Use of doubly-labelled acetate gave biosynthetic
chemists a powerful tool to probe and to prove such structural changes.
A spectacular example of late stage reorganisation is in the biosynthesis of aflatoxin B1 8 (Fig. 8). 9,10
The predicted pattern of intact and cleaved two-carbon (C2) residues is indicated in the conventional way
by heavy lines and single blobs, respectively. In ideal circumstances, each pair of coupled 13C nuclei will
have a unique coupling constant which helps to confirm the assignment of the 13C NMR spectrum. Each
coupled atom gives a signal in which a 13C doublet is symmetrically distributed around the natural
abundance singlet; 13C nuclei that have lost their original partner signal appear as an enriched singlet. An
essential factor for interpreting this simple pattern of signals is that the labelled precursor is diluted to a
sufficient extent by unlabelled acetate (usually as a natural consequence of the incorporation process);
otherwise, more than one doubly-labelled acetate is incorporated into a single molecule and the resulting
spectrum may then be further complicated by inter-unit 13C13C couplings.

9 von 66 20.11.2006 11:04


Published by the Royal Society of Chemistry http://pubs.rsc.org/ej/NP/2001/A909079G/

Fig. 8 Summary of steps in the late-stage reorganisation in the biosynthesis of the polyketide aflatoxin B1 8. The
predicted pattern of intact and cleaved two-carbon (C2) residues is shown in the conventional way by heavy lines an
single blobs, respectively.

Additional insights into the mechanisms of biosynthetic processes can be obtained by studying the fate
of hydrogen atoms in intermediates. Various experiments have been developed involving either direct
observation of 2H or 3H by NMR spectroscopy or indirect detection through the effect of the hydrogen
isotope on an adjacent or nearby 13C label in the original precursor. An example of the use of just one of
these technologies, the -shift, will suffice to illustrate the extra information which can come from tracing
the path of hydrogen atoms through the biosynthesis. In this technology, the hydrogens to be followed are
replaced by deuterium and the carbons to which they are attached are labelled with 13C. The 13C NMR
spectrum then reveals the number of attached deuteriums in individual molecules. In molecules with one
deuterium, the 13C signal is shifted upfield by about 0.3 ppm and it appears as a 1 1 1 triplet (J typically
20 Hz) due to 2H13C coupling; for each additional deuterium there is a further upfield shift and a
corresponding increase in the multiplicity of the signal (Fig. 9). 11,12 With a suitable spectrometer,
therefore, it is possible to determine the extent of deuterium labelling in considerable detail. Any change
in the deuterium content at the corresponding 13C-labelled site in the natural product will be reflected in
the pattern of 13C signals.

10 von 66 20.11.2006 11:04


Published by the Royal Society of Chemistry http://pubs.rsc.org/ej/NP/2001/A909079G/

Fig. 9 Hypothetical 13C NMR spectrum illustrating the


-effect. Each additional deuterium label at a carbon centre
shifts the carbon signal upfield by 0.3 ppm and also
increases its multiplicity.

One important use of this technology was in the study of starter units. Acetate simultaneously labelled in
the methyl group with three deuteriums and a 13C, was incorporated into terrein 9 (Fig. 10). 13 From the
13
C spectrum of the metabolite it was clear that a small proportion of the molecules carried three
deuteriums in the starter methyl group. This result proved that acetate could serve as a starter acid.
However, the majority of molecules had only one or two labels, showing that metabolic processes can
remove hydrogen from the labelled carbon at some intermediate stage.

Fig. 10 Incorporation of deuterium-labelled acetate


into terrein 9. The presence of three deuteriums in
the starter group of some molecules demonstrated
that acetate serves as a starter acid in terrein
biosynthesis.

This technology was also successfully applied to understanding the biosynthesis of erythromycin A 10.
The putative precursor 11 (Fig. 11) was labelled simultaneously with 2H and 13C labels as indicated.
Incorporation of the precursor into erythromycin resulted in the metabolite retaining both the 2H and 13C
labels. 14 This result pointed to a biosynthetic route to the reduced aliphatic polyketide in which each
successive addition of a building block is immediately followed by reductive modification of the
newly-formed keto group (so-called processive biosynthesis), as opposed to construction of a polyketone
chain followed by reduction. This observation is highly significant, as will become clear later.

11 von 66 20.11.2006 11:04


Published by the Royal Society of Chemistry http://pubs.rsc.org/ej/NP/2001/A909079G/

Fig. 11 Incorporation of diketide intermediate 11 into


erythromycin A 10. The fact that analogue 11 was
incorporated intact into erythromycin A provided
important evidence for a processive mode of
biosynthesis.

Other isotope technologies were developed to monitor the fate of oxygen atoms from simple precursor
molecules to the complex products. There are a number of excellent reviews on the subject (see, e.g., ref.
11 ) which makes it unnecessary to give details here. Suffice it to say that, as a result of many detailed and
elegant investigations, chemists established a large body of knowledge upon which they have built a
detailed picture of the types of biosynthetic processes employed in polyketide pathways.

4 Biomimetic syntheses of aromatic polyketides

The new biosynthetic insights that emerged from precursor studies inspired a flurry of activity in synthetic
organic chemistry towards the generation of -polyketones. The aim was to develop new synthetic routes
to polyketide aromatic systems (benzenes, naphthalenes, anthracenes, etc.). A second goal, of greater
relevance to the subject of this review, was to carry out mechanistic studies on the chemical factors that
govern the inherent preference for one cyclisation mode over another. This is, of course, germane to the
role of enzymes in controlling ring formation in vivo.
One notably successful approach developed by Tom Harris in the 1970s was the condensation of
poly-anions of polyketones with acylating agents. 1518 A striking example is the condensation of acetate
with the acetoacetate anion 12 (Fig. 12). The resulting polyketone had precisely the type of structure
predicted some 65 years earlier by Collie. Several important conclusions emerged from the Harris
investigations. Most importantly, it was clear that the folding of a polyketone chain formed in vivo must be
under enzymatic control, so that unwanted cyclisation modes are suppressed. In particular, it is necessary
to prevent premature cyclisation of the chain in the early stages of elongation. 19

Fig. 12 Generation of a linear triketone by reaction of


the dianion of acetylacetone 12 with an acylating
agent, ethyl acetate.

Several other groups used pyrone chemistry to generate transient intermediates with polyketone chains.
2022
The researchers were motivated by the idea that enzymatic control in vivo might be achieved by
selection of the folding point of the chain so that the first carbocyclic ring is formed at the appropriate site.
The preferred mode for subsequent cyclisations would then be progressive formation of extra
six-membered rings until a linear polycyclic system is generated. A striking example of this principle in
the biosynthesis of the fungal metabolites rubrofusarin 13 and alternariol 14 is shown in Fig. 13. 23

12 von 66 20.11.2006 11:04


Published by the Royal Society of Chemistry http://pubs.rsc.org/ej/NP/2001/A909079G/

Fig. 13 Folding of a heptaketide chain to give the phenolic


natural products rubrofusarin 13 and alternariol 14. Control
of cyclisation in both cases is achieved by formation of the
first carbocyclic ring at the appropriate position in the
chain.

5 Enzymology of fatty acid and polyketide biosynthesis

Biomimetic syntheses could do no more than provide indirect insights into the reactions occurring in vivo.
Therefore, there was a powerful incentive to investigate the biosynthetic enzymes directly. Before
introducing the enzymes of polyketide biosynthesis (the so-called polyketide synthases (PKSs)), it is
appropriate to digress briefly into the closely related field of fatty acid biosynthesis. The two pathways
have strong homologies, both in the nature of the chemistry used in chain extension from a common pool
of simple precursors and in the character of the enzymes that are used for chain assembly, and so insights
into fatty acid assembly are relevant to polyketide biosynthesis. In fact, research in the fatty acid field has
almost always predated equivalent studies in the polyketide field. Even today, the precedents of fatty acid
biosynthesis guide polyketide researchers.

5.1 Fatty acid biosynthesis

Fatty acids are assembled from C2 units by repeated head-to-tail linkage, until a chain of the required
length is reached. The chemical steps of chain extension are shown in Fig. 14. A starter acyl unit (usually
acetyl) is condensed with a malonyl unit that undergoes concerted decarboxylation to furnish the electrons
for the new carboncarbon double bond. The resulting -keto ester is successively reduced to a hydroxy,
dehydrated and finally reduced again to give a saturated chain longer than the original by two methylene
units.

13 von 66 20.11.2006 11:04


Published by the Royal Society of Chemistry http://pubs.rsc.org/ej/NP/2001/A909079G/

Fig. 14 The fatty acid biosynthetic cycle. MAT: malonyl-acetyl transferase; ACP: acyl carrier protein; KS: ketosynt
KR: ketoreductase; DH: dehydratase; ER: enoyl reductase.

The acyl participants in these chemical reactions are bound as thioesters to the set of enzymes used for
chain extension. The starter acyl is attached to a cysteine thiol of the first enzyme, a keto synthase (KS),
which catalyses condensation. The chain extender malonate unit is bound to a thiol residue of a protein
designated as an acyl carrier protein (ACP). The thiol in this case is not part of the primary protein chain,
but is the terminus of a phosphopantetheine residue added to the protein in a post-translational
modification. The phosphopantetheine functions as a long flexible arm that carries the growing chain and
delivers it to the various enzymes responsible for chain extension. Condensation results in an extended
chain in the form of a -keto ester bound to the ACP. In subsequent steps, the keto ester is reduced by a
keto reductase (KR), dehydrated by a dehydratase (DH) and finally reduced again by an enoyl reductase
(ER). This sequence of reactions completes the first round of chain extension. The cycle is then repeated
after transfer of the saturated chain from the ACP to the KS. Successive cycles lead eventually to a chain
of the required length (usually 14, 16 or 18 carbons). At that stage, the chain is passed to a thioesterase
(TE) enzyme from which it is released, usually as a free acid or an acyl ester. All fatty acid synthases
(FASs) have the same set of components (KS, ACP, KR, DH, ER and TE). There is also a seventh partner,
a malonyl-acetyl transferase (MAT). The role of this protein is to transfer the building blocks acetate (to
start) and malonate (to extend) from the respective coenzyme A pools onto the active thiol resides of the
appropriate domains.
The structural organisation of the many activities of the FASs depends on the type of organism. At one
extreme, in bacteria, the fatty acid synthase consists of a set of discrete proteins that can be isolated
separately. In mammals, on the other hand, all of the enzyme activities are covalently linked to form a
large multifunctional protein. Various intermediate stages of organisation are found in other organisms.
The fully dissociable proteins are designated type II; conversely, the multifunctional variety are called type
I. It is likely, however, that the dissociable type II systems, like the type I systems, are organised into a
highly structured multifunctional array in their active state. Investigators ignore this organisational aspect
at their peril, as we shall see.
The enormous progress in studies of fatty acid synthases made after the 1970s was not matched in the
polyketide field. Only one PKS was isolated and characterised in sufficient purity for definitive studies.

14 von 66 20.11.2006 11:04


Published by the Royal Society of Chemistry http://pubs.rsc.org/ej/NP/2001/A909079G/

This was 6-methylsalicylic acid synthase (6-MSAS) isolated from a fungal source. It proved to be a type I
synthase and a tetramer (MW of each monomer, 190 kDa). 2426 A fuller account of this fascinating
synthase will be given later. Despite extensive efforts, attempts to find other synthases, especially those for
macrolides like erythromycin, drew a blank. Two decades passed before the polyketide synthase
responsible for erythromycin biosynthesis was found. In the next section, we will describe the genetic
advances that underpined this breakthrough.

6 The genetic approach to biosynthetic studies

Just as the introduction of isotopes opened up the elucidation of biosynthetic pathways in the 1950s, the
development of genetic techniques in the 1980s opened the way to the discovery of the enzymes. The
principal pioneer in this area was Sir David Hopwood who adopted the polyketide actinorhodin 15 (Fig.
15(a)) as his test system on the basis of its characteristic blue colour. Cessation of production in any
mutant bacteria by inactivation of specific genes was immediately obvious by lack of the blue colour
combined with loss of antibiotic activity. Through a series of carefully designed experiments over a
number of years, the genes coding for the enzymes of actinorhodin biosynthesis were identified. 27 The
genes were sequenced (a formidable task in the early days of this research) and so the primary sequence of
the various proteins was established. The next stage was to assign functions to the enzymes by comparison
with the sequence of known proteins. A number of components showed convincing homology with the
various enzymes of fatty acid synthases (KS, ACP, KR) and so were assigned to the role of polyketide
chain extension. The organisation of these genes on the genome is shown in Fig. 15(b). The genes are
discrete and so, clearly, the actinorhodin PKS is a type II dissociable system just like the FAS of bacteria.

Fig. 15 (a) The polyketide antibiotic actinorhodin 15. (b) Organisation of the gene cluster for
the type II PKS responsible for actinorhodin biosynthesis.

We will return to further studies of the type II genes later, but first we will consider the macrolide
systems because these turned out to be type I synthases and were much more amenable to systematic
studies.

6.1 Genetics of type I or modular polyketide synthases

In the past five years, there has been an explosion of interest in sequencing and manipulating the genes for
type I polyketide synthases, as well as understanding the structure and function of the extraordinary
proteins that they encode. 28,29 As a consequence, it is impossible to present a comprehensive review of all
of the advancements in the field. We will therefore attempt only to be representative, relying most often on
the erythromycin PKS as an example.

6.1.1 The erythromycin polyketide synthase. Inspired by the discovery of the genes encoding for the
biosynthesis of actinorhodin, 27 the search for the elusive erythromycin genes in the bacterium
Saccharopolyspora erythraea began in earnest in the late 1980s. Within a few years, two labs had

15 von 66 20.11.2006 11:04


Published by the Royal Society of Chemistry http://pubs.rsc.org/ej/NP/2001/A909079G/

succeeded in finding the genes using different but complementary approaches. Peter Leadlay s group at
Cambridge University worked on the assumption that the antibiotic biosynthetic genes in S. erythraea and
the related Streptomyces bacteria are clustered around the gene that confers self-resistance. 30,31 By
walking along the chromosome away from the erythromycin resistance gene, ermE, and searching for
sequences with similarity to FAS and other PKS enzymes, they discovered the ery genes. 31 Leonard Katz
and co-workers at Abbott Laboratories, on the other hand, knew of a region in the genome of S. erythraea
near ermE in which mutations disabled the synthesis of 6-deoxyerythronolide B (6-dEB) 16, the putative
product of the polyketide synthase. 32,33 Sequencing within this area also revealed the PKS genes. Partial
sequence information was published by both groups in 1990, 31,34 while more rigorous analyses of the
gene cluster followed shortly after. 35,36 A map showing the regions of the genome is given in Fig. 16. On
both sides of ermE are genes coding for non-PKS enzymes involved in the late or tailoring stages of the
pathway in which 6-dEB is converted into the active antibiotic erythromycin A 10. Further away from
ermE are the genes for the PKS itself.

Fig. 16 Map of the regions of the Saccharopolyspora erythraea genome containing the genes associated with erythr
biosynthesis.

The structural genes responsible for the biosynthesis of the first macrolide intermediate 6-dEB 16 are
three enormous open reading frames (ORFs), eryAI,eryAII and eryAIII ( 10 kbp each, Fig. 16), coding for
three gigantic ( 350 kDa each) multienzyme polypeptides, 6-deoxyerythronolide B synthase (DEBS) 1, 2
and 3 respectively. The size of the ORFs places the erythromycin PKS in a select group of enzymes that
synthesise secondary metabolites using giant, multifunctional polypeptides. Even more remarkably, the
DEBS proteins are dwarfed by those involved in the biosynthesis of, for example, rapamycin, 37,38
rifamycin 39 and nystatin. 40 The catalytic activities or domains within each of the giant proteins were
assigned functions by comparison to the domains in fatty acid synthases. 31,3436 The homologies between
the FAS and putative PKS activities gave the investigators confidence that they had discovered the genes
for the PKS proteins.
In the linear representation of the primary sequence of the proteins shown in Fig. 17, coded blocks
indicate sections thought to be associated with specific domains. Each domain is presumed to form a
localised globular structure with a catalytic role determined by conserved residues at its specific active site
(46 amino acids). The linker regions between the domains are believed to play a vital structural role in
maintaining the correct topology of the activities for co-operation in the biosynthetic process. These
linkers, up to 100 residues long, are frequently rich in alanine, proline and charged residues, which,
depending on the particular amino acid composition, may allow them to serve either as rigid connectors or
to provide flexibility to allow segmental motion of the multienzymes. 35,41 There are also intriguing, much
longer sections of primary sequence preceding the KR domains, which have not yet been associated with
any particular catalytic function.

16 von 66 20.11.2006 11:04


Published by the Royal Society of Chemistry http://pubs.rsc.org/ej/NP/2001/A909079G/

Fig. 17 Predicted domain organisation of the DEBS proteins. The ketosynthase (KS),
acyltransferase (AT), dehydratase (DH), enoyl reductase (ER), ketoreductase (KR), acyl
carrier protein (ACP) and thioesterase (TE) domains are represented by coded boxes
whose lengths are proportional to the sizes of the domains. KR indicates the inactive
ketoreductase domain. The ruler shows the residue number within the primary structure
of the constituent proteins. The linker regions are also given in proportion.

6.1.1.1 Organisation and function of the erythromycin PKS. To clarify the synthetic implications of the
domain organisation, a cartoon version of the PKS is shown in Fig. 18, in which circles depict domains
and the linker regions are omitted. From this picture it is clear that each of the DEBS proteins contains
two functional units or modules, hence the name modular PKS. Each module contains the three domains
required to catalyse one cycle of chain extension (ketosynthase (KS), acyltransferase (AT) and acyl carrier
protein (ACP)) as well as a variable set of domains (ketoreductase (KR), dehydratase (DH) and enoyl
reductase (ER)) associated with keto group modification. 35,36 In this representation, the set of domains
starting with a KS and finishing with an ACP is taken to be a modular unit. DEBS 1 is fronted by a
loading didomain (AT and ACP) which accepts the starter unit propionate from propionyl-CoA, while
DEBS 3 terminates with a thioesterase (TE) activity which is thought to catalyse the off-loading and
cyclisation of the fully-formed heptaketide intermediate 17 to give 6-dEB 16. Throughout the entire
biosynthetic sequence, the polyketide chains remain bound to the PKS.

17 von 66 20.11.2006 11:04


Published by the Royal Society of Chemistry http://pubs.rsc.org/ej/NP/2001/A909079G/

Fig. 18 Domain organisation of the erythromycin polyketide synthase. Putative domains are represented as circles a
the structural residues are ignored. Each module incorporates the essential KS, AT and ACP domains, while all but
include optional reductive activities (KR, DH, ER). The one-to-one correspondence between domains and biosynthe
transformations explains how programming is achieved in this modular PKS.

Closer inspection of the modules (Fig. 18) allows some predictions to be made concerning the structure
of the chain extension unit formed in the growing chain. The three essential domains KS, AT and ACP,
co-operate to catalyse carboncarbon bond formation by Claisen condensation, which results in a -keto
ester intermediate. The variable set of domains positioned between the AT and ACP (depicted as a loop
above the line of essential domains) then carry out reductive modification of the keto group before the
next round of chain extension. Module 3 has no reductive domains and so the keto group should survive
intact. In contrast, modules 1, 2, 5 and 6 each incorporate a KR domain, which is expected to catalyse
reduction of the keto group to a hydroxy group. In module 4, there is a full set of reductive activities (DH
+ ER + KR) and so complete reduction to produce a methylene is anticipated. Therefore, if the modules
operate in the order in which they are positioned, the polyketide product would have the following
sequence of functional groups on alternate carbons, starting with the first chain extension unit (alkyl end
of the chain): hydroxy, hydroxy, keto, methylene, hydroxy, hydroxy and finally an acyl group.
Examination of the uncyclised form of the heptaketide 17 shows that it has precisely this substitution
pattern.
The correspondence between the domain composition of consecutive modules and the structure of each

18 von 66 20.11.2006 11:04


Published by the Royal Society of Chemistry http://pubs.rsc.org/ej/NP/2001/A909079G/

newly added three-carbon unit in successive intermediates was persuasive, but it did not prove that the
modules function in the linear order indicated by the sequence. The proteins could fold or associate in
many different ways to bring sets of non-contiguous domains together to form functional modules. It was
therefore a high priority to establish that modules are indeed made up of adjacent domains, and that the
individual modules and domains are used in the order suggested by the genes. The first strong supporting
evidence came from Leonard Katz and co-workers who carried out pioneering gene-disruption
experiments. 36,42 In 1991, the researchers disabled the KR of module 5 by removing a large portion of the
gene (Fig. 19(a)). 36 The resulting mutant strain produced two erythromycin analogues in which a keto
group had survived at the predicted position in place of the normal hydroxy function. Subsequently, they
inactivated the ER in module 4 by introducing mutations in its putative NADPH-binding site. 42 The
resulting mutant produced small amounts of an erythromycin analogue incorporating a double bond at the
expected position (Fig. 19(b)). From these experiments, it was concluded that the deactivated KR operates
in the fifth cycle of chain extension and that the disrupted ER functions in the fourth cycle, as predicted by
the arrangement of modules and domains shown in Fig. 18. The co-linearity of protein structure and
function has been confirmed by further experiments (see, e.g., ref. 43 ) and is now accepted as a general
property of these megasynthases.

19 von 66 20.11.2006 11:04


Published by the Royal Society of Chemistry http://pubs.rsc.org/ej/NP/2001/A909079G/

Fig. 19 Experimental evidence for the modular organisation of DEBS. (a) Inactivation of KR5 of DEBS resulted in
production of erythromycin analogues with keto groups at the C-5 position. (b) Inactivation of ER4 resulted in an
analogue of erythromycin with a double bond at the expected site.

This analysis has revealed one way in which nature has solved the problem of how to generate complex
programming in a biosynthetic pathway: the order of operations in modular PKSs is written in the primary
structure of the proteins. Thus, the choices for the number and type of building blocks for chain extension

20 von 66 20.11.2006 11:04


Published by the Royal Society of Chemistry http://pubs.rsc.org/ej/NP/2001/A909079G/

are encoded in the gene sequence itself, and translated in the proteins as a series of active sites
appropriately arrayed along the multienzyme. In this simple scheme, each enzyme only has one role. In
effect, therefore, the multienzyme DEBS proteins appear to function in succession as a molecular
assembly line.

6.2 Other type I modular PKSs

In the years since the discovery of the eryA genes, many other genes from bacteria encoding for macrolide
polyketide synthases have been sequenced, in whole or in part, including those for rapamycin 18, 37,38
FK506 19, 44,45 spiramycin 20, 46 oleandomycin 21, 47,48 the avermectins (e.g., avermectin B1b 22), 49,50
niddamycin 23, 51 methymycin 24/pikromycin 25, 52 pimaricin, 53 nystatin 40,54 and tylosin 26 55 (Fig. 20).
Screening has, in many cases, been carried out with probes derived from the erythromycin PKS (see, e.g.,
refs. 37, 51, 52). PKS clusters encoding for as yet unknown metabolites have also been uncovered in a
number of bacterial chromosomes, while others have been located through the inadvertent cloning of
genes that were not originally targeted. 56,57 Without exception, each PKS has the modular organisation of
DEBS with each module catalysing a single condensation and reduction cycle.

Fig. 20 Examples of macrolide polyketides for which the genes have been sequenced in whole or in part.

One of the largest sets of genes sequenced in its entirety is rapAC, which encodes for the rapamycin
synthase (RAPS) responsible for assembling the polyketide core of rapamycin 18 (Fig. 20). Rapamycin

21 von 66 20.11.2006 11:04


Published by the Royal Society of Chemistry http://pubs.rsc.org/ej/NP/2001/A909079G/

and the structurally related compounds FK506 19 and FK520 belong to the family of polyketide polyenes
and demonstrate antitumor, antifungal and immunosuppressant activities. It is the latter
immunosuppressant activity that has generated considerable interest due to its applications to organ
transplant surgery. From a biosynthetic standpoint the molecules are very interesting because they
incorporate both polyketide and polypeptide structure. The RAPS cluster consists of three extraordinarily
large ORFs encoding the multienzymes RAPS 1 ( 900 kDa, 4 modules), RAPS 2 (1.07 MDa, 6 modules)
and RAPS 3 (660 kDa, 4 modules) (Fig. 21). In total, the three proteins contain 70 catalytic functions,
making this one of the most complex multienzyme systems identified in nature. 37,38 In contrast to the
erythromycin PKS, the domain structure of RAPS may not correspond in every detail to the pattern
expected from the sequence of domains. In modules 3 and 6, for example, there appear to be potentially
active KR and DH domains which are not required; module 3 also contains a nominally active but
functionally redundant ER domain (Fig. 21). It is possible that the active sites of these extra domains have
been disabled in a way that is not apparent from the primary sequence, and that the now superfluous
protein residues remain to be completely edited out by the random processes of evolution. There is also a
chance that these domains are active, and that extra post-PKS oxidations are required to reintroduce the
oxygen functionality at the relevant sites in the final structure. This discrepancy between the sequence of
domains and the required set of transformations is not unique to this synthase.

Fig. 21 Domain organisation of the rapamycin polyketide synthase (RAPS). As with the erythromycin PKS there is
the order of biosynthetic steps. The activities shown in grey are presumed to be inactive because their predicted site
structure of rapamycin 18 (Fig. 20). Cyclisation of the fully-processed chain mediated by the pipecolate-incorporati

22 von 66 20.11.2006 11:04


Published by the Royal Society of Chemistry http://pubs.rsc.org/ej/NP/2001/A909079G/

The RAPS PKS differs from DEBS in other interesting ways. The first gene, rapA, encodes a set of
loading functions consisting of domains resembling a coenzyme A (CoA)-ligase, an ER and an ACP,
which are thought to activate and reduce a dihydrocyclohexane starter unit and then transfer the resulting
starter unit to the KS domain of the first module. 37,58 Rapamycin also has an unusual mechanism of chain
termination and cyclisation (Fig. 22): within the PKS genes and translationally coupled to rapC is a gene
rapP 59 with sequence similarity to genes involved in nonribosomal polypeptide biosynthesis. 60 The
corresponding protein (called pipecolate-incorporating enzyme (PIE)) is believed to catalyse the formation
of ester and amide bonds to pipecolic acid. It has therefore been proposed that the polyketide chain of
rapamycin is transferred from a thioester linkage on RAPS 3 directly to the amino group of enzyme-bound
pipecolate 27, which in turn is attacked by the terminal hydroxy of the chain to form the macrolide (Fig.
22). 37 PIE therefore provides the pivotal link between polypeptide and polyketide biosynthesis.

Fig. 22 Formation of the rapamycin macrocycle. The


fully-processed chain is first coupled to the nitrogen of
pipecolate 27 and then the carboxyl group of the amino acid
is esterified with the C-34 hydroxy group at the remote end
of the chain, to give the macrolide (the order of steps could
be reversed). The pipecolate-incorporating enzyme (PIE) is
responsible for both transformations, and as such is the
pivotal link between polyketide and polypeptide
biosynthesis.

7 Type I iterative polyketide synthases

Like DEBS, the polyketide synthases responsible for biosynthesis of the fungal metabolites 6-MSA 5 and
lovastatin 28 (also called mevinolin or monacolin K) have a type I or covalent architecture. However,
whereas DEBS contains a distinct module for each round of chain extension, the single modules of
6-MSAS and the lovastatin nonaketide synthase (LNKS) are used iteratively, that is, at least some of the
constituent enzymes act in every successive cycle. The molecular basis for this complex programming has

23 von 66 20.11.2006 11:04


Published by the Royal Society of Chemistry http://pubs.rsc.org/ej/NP/2001/A909079G/

generated considerable interest. 61,62

7.1 6-Methylsalicylic acid synthase

6-Methylsalicylic acid 5 is assembled from four ketide units (one molecule of acetate and three of
malonate) by the sequence of events discussed earlier (see Section 2.1). The PKS responsible for its
biosynthesis, 6-MSAS, contains the following domains (in order): KS, MAT, DH, KR and ACP. 8 By
some as yet undetermined mechanism, these activities act repeatedly to catalyse three rounds of chain
extension, carrying out different levels of reductive processing at each stage (Fig. 23). For example, the
first condensation is followed directly by reaction with a second equivalent of malonate extender unit,
while the second condensation is followed instead by reduction and dehydration of the newly-formed keto
group. After the third cycle, the chain undergoes cyclisation, dehydration and enolisation to give the
aromatic 6-MSA. The absence of a thioesterase domain suggests that release of the chain from the PKS
does not occur by hydrolysis and an alternative mechanism involving a ketene intermediate has been
proposed, although not experimentally verified. 8

24 von 66 20.11.2006 11:04


Published by the Royal Society of Chemistry http://pubs.rsc.org/ej/NP/2001/A909079G/

Fig. 23 The biosynthetic pathway for the fungal polyketide 6-methylsalicylic acid (6-MSA) 5.

7.2 Lovastatin polyketide synthases

25 von 66 20.11.2006 11:04


Published by the Royal Society of Chemistry http://pubs.rsc.org/ej/NP/2001/A909079G/

Another example of an iterative type I PKS is that responsible for assembling part of the non-aromatic
polyketide lovastatin 28 (Fig. 24). Lovastatin is a polyketide metabolite produced by the filamentous
fungus Aspergillus terreus. It inhibits the rate-limiting step in the biosynthesis of cholesterol, the
conversion of (3S )-hydroxymethylglutaryl-CoA (HMG-CoA) into mevalonate by the enzyme HMG-CoA
reductase. Lovastatin therefore exhibits strong cholesterol lowering activity and is in clinical use for
treatment of hypercholesterolemia. 62,63

26 von 66 20.11.2006 11:04


Published by the Royal Society of Chemistry http://pubs.rsc.org/ej/NP/2001/A909079G/

Fig. 24 Hypothetical pathway for biosynthesis of the fungal metabolite lovastatin 28. The nonaketide
and diketide portions of lovastatin are generated by two independent pathways, catalysed by the
lovastatin nonaketide synthase (LNKS) and the lovastatin diketide synthase (LDKS), respectively. Both
enzymes build their products from an acetate starter unit and malonate extender units, and both add a
methyl group from S-adenosyl methionine. The intermediates shown bound to the PKSs are
hypothetical, but consistent with experimental results. Cyclisation has been shown to proceed via a

27 von 66 20.11.2006 11:04


Published by the Royal Society of Chemistry http://pubs.rsc.org/ej/NP/2001/A909079G/

DielsAlder condensation.

Lovastatin is composed of two polyketide chains joined through an ester linkage (Fig. 24). One chain is
a nonaketide that undergoes cyclisation to form an octahydronaphthalene ring system, and the other is a
diketide, (2R)-2-methylbutyrate. A hypothetical pathway for lovastatin biosynthesis is presented in Fig.
24. 62 Several experiments have demonstrated that the two portions of the molecule are assembled by
separate large multifunctional PKSs, called the lovastatin nonaketide synthase (LNKS) and the lovastatin
diketide synthase (LDKS). 61,62 Both PKSs generate their polyketide product from an acetate starter unit
and malonate extender units 64,65 and both add a methyl group from S-adenosyl methionine. The PKSs
interact and are modulated by other enzymes of the pathway, which somehow enable the LNKS to
discriminate between carbon chain intermediates at different stages of assembly, and may cause the LDKS
to behave noniteratively, like a bacterial modular PKS. 61,62
The LNKS is unimodular and contains six active sites KS, AT, DH, ER (thought to be non-functional),
KR and ACP in the same order as in the animal FAS and the modular PKSs. 61,62 A significant
difference is the presence of a methyl transferase (MT), which has been shown to add a methyl group from
S-adenosyl methionine during chain extension, and substitution of the terminal TE domain often found in
FASs and PKSs with a putative peptide synthetase elongation (PSED) domain. The PSED is similar to the
corresponding domains of peptide synthetases, but as there is no known role for amino acids or other
nitrogen-containing compounds in lovastatin biosynthesis, the function of this domain is unknown. It is
believed that the LNKS, in co-operation with a discrete type II protein called LovC (a putative ER), acts
iteratively to synthesise dihydromonacolin L 29 (Fig. 24), the first enzyme-free intermediate in lovastatin
biosynthesis. The formation of the octahydronaphthalene ring system of 29 has been shown to involve an
intramolecular DielsAlder condensation. 6468 As 29 incorporates two hydroxy groups, one double bond
and a methyl group at C-6, the LNKS is presumably able to use different sets of activities during each
chain extension cycle.
The second PKS, the LDKS, is also unimodular and contains the six active sites typical of PKSs and an
additional MT domain. Unlike the LNKS, however, it does not contain a PSED domain and its ER activity
is presumed to be functional. 62 All evidence suggests that the LDKS is a noniterative type I PKS which
catalyses a single condensation, followed by methylation, ketoreduction, dehydration and enoyl reduction
to yield (2R)-2-methylbutyrate 30. This diketide remains bound to the LDKS until it is attached to
monacolin J 31 by a specific transesterase enzyme (Fig. 24).
As is apparent from these descriptions, the iterative pathways of 6-MSA and lovastatin biosynthesis
require that the PKS activities have the ability to differentiate between substrates that vary both in their
state of reduction and chain length. The basis for this programming remains an intriguing mystery, because
it is not obvious how a single set of activities can make the choice of oxidation level at each
chain-extending condensation. Very clever experiments will be required to tease out the molecular
acrobatics behind these reactions.

8 Structural studies on the type I modular polyketide synthases

X-ray crystallography has dramatically advanced our understanding of protein structure, and it goes
without saying that such information is required to fully comprehend the behaviour of a chemical system.
69
However, the majority of enzymes for which a structure has been determined act as discrete catalytic
units, and are of manageable dimensions (on the scale of 1050 kDa). The complexity and sheer size of
the polyketide synthases, therefore, presents a significant challenge to the crystallographer. Although a
number of structures of individual (type II) domains and plant PKSs are in the literature, 2,7074
high-resolution information about type I systems from X-ray crystallography or NMR spectroscopy is
lacking. 69
The modular function of the PKSs raises fascinating organisational questions. While the proteins that
need to co-operate are tethered in groups, there must also be some three-dimensional arrangement that
allows the interacting partners to be within reach of each other. The key player in the biosynthetic pathway
is the acyl carrier protein, which participates in reactions with all of the other components of each module.
A reacting species is attached as a thioester on the end of the phosphopantetheine chain, and it is believed

28 von 66 20.11.2006 11:04


Published by the Royal Society of Chemistry http://pubs.rsc.org/ej/NP/2001/A909079G/

that the chain acts as a flexible arm that can reach approximately 20 . The volume of space though which
the arm can pass will be extended by whatever degree of relative movement is allowed to the ACP itself. It
seems likely that to avoid cross talk between modules, other than the essential transfer of the product of
each chain-extension cycle to the downstream KS, the domains within each module must occupy a fixed
array. Obviously this geometrical and spatial arrangement must be respected in experiments to alter the
functionality of modules. There is therefore a strong imperative to define the three-dimensional shape of
polyketide systems.
The following features of the synthases must be accounted for by any model of their structure: (i)
variation in the number of cycles and therefore modules required to synthesise individual polyketides; (ii)
substantial disparities in the size of different modules arising from the presence or absence of reductive
activities; and (iii) variation in the number of modules in each multienzyme subunit. In lieu of
crystallographic information, data obtained from a variety of structural studies on DEBS 7577 have been
used to generate a model for the structure of a modular PKS that fulfils all of these requirements. 77
As mentioned earlier, the domains of the modular PKSs and the type I animal FASs not only carry out
similar operations, but share strong sequence homologies. 31 In fact, the domain composition of a typical
FAS is equivalent to a single PKS chain extension module. Therefore, it was initially speculated that the
overall PKS structure might be similar to that of the FAS. Until recently, the accepted model of the FAS
was a planar dimer composed of two identical polypeptide chains aligned side-by-side and head-to-tail
(Fig. 25). 78,79 In this model, the FAS supports two catalytic centres for fatty acid biosynthesis, one at
either end of the dimer. Each centre is composed of the KS, MAT and DH of one subunit and the ER, KR,
ACP and TE of the other. Evidence from recent elegant complementation experiments with the rat FAS,
however, has repeatedly challenged this view, as domains were shown to co-operate within the same
subunit, 80,81 a mode of operation not accommodated by this model. An alternative structure for the FAS 41
accounts for all of the available data, but it has yet to be verified experimentally.

Fig. 25 Schematic model of the dimeric animal fatty


acid synthase (FAS). The two polypeptides are
aligned head-to-tail, creating two catalytic centres
as shown (dashed line). The overall shape of the
FAS was suggested by early experiments using
small-angle neutron scattering and electron
microscopy.

29 von 66 20.11.2006 11:04


Published by the Royal Society of Chemistry http://pubs.rsc.org/ej/NP/2001/A909079G/

8.1 Experiments to probe the structure of the erythromycin PKS

To determine the structure of the erythromycin PKS, DEBS 1, 2 and 3 were purified 8284 and subjected to
limited proteolysis, chemical cross-linking and analytical ultracentrifugation. 7577 For each proteolytic
enzyme and each of the DEBS proteins, the most rapid cleavages occurred between the two chain
extension modules housed in each DEBS multienzyme (Fig. 26). Other initial cuts released the N-terminal
loading didomain from DEBS 1 and the C-terminal ACP-TE didomain from DEBS 3. 76,77 All of the
modular fragments had molecular masses in agreement with those predicted from their amino acid
sequences, and all except the loading didomain behaved as homodimers by gel filtration.

Fig. 26 Pattern of fragments generated by controlled proteolysis of DEBS 1 and DEBS 3. Signifies a
homodimeric protein.

More extensive proteolysis was required to cleave in the interdomain regions of each module, suggesting
that the domains must be folded into tight protease-resistant structures separated by relatively unstructured
linker regions. 7577 The KS and AT domains were released together and were determined to be
homodimeric by gel filtration (i.e. KS2AT2), as was the thioesterase (Fig. 26). On this basis it was
reasoned that each KS-AT pair is probably in contact with the equivalent domains in the opposite chain in
the intact DEBS protein, and that these four activities form the core of the modular structure. In contrast,
the activities involved in the optional reductive modifications the KR domains of modules 1, 3, 5 and 6
and the single ER domain from module 4 were all characterised as monomers. 77 Treatment of the DEBS
proteins with the chemical cross-linking agent 1,3-dibromopropanone followed by limited proteolysis gave
a different pattern of fragments, consistent with the formation of covalent bonds between the two protein
chains. 77 By analogy with the results from the cross-linking of animal FAS activities, 85 it was concluded
that the links were formed between the free thiols of the KS of one chain and the ACP in the
complementary chain, and therefore that these activities co-operate in catalysis. 41

8.2 Proposed structural models for the erythromycin PKS

30 von 66 20.11.2006 11:04


Published by the Royal Society of Chemistry http://pubs.rsc.org/ej/NP/2001/A909079G/

The data from the structural studies, particularly the finding that the KS-AT pair and the TE are
homodimeric while the reductive activities are monomeric, inspired a three-dimensional model for the
erythromycin PKS. 41,77 In this so-called Cambridge model proposed by Staunton, Leadlay and
co-workers, the identical PKS subunits are associated head-to-head and tail-to-tail, and twisted together to
from a helix (Fig. 27). At the core of the helix is a tetrahedron formed by the KS and AT domains of each
module, an arrangement consistent with their homodimeric behaviour; likewise, the TE domains are in
intimate contact at the end of the DEBS 3 dimer (Fig. 27). In contrast, the optional reductive activities
form loops which protrude out from the central core while remaining within range of the
phosphopantetheine arm of the ACP on the same chain. In this arrangement there is no contact between
reductive domains of the complementary chains, which agrees with the monomeric nature of these
activities when released by limited proteolysis.

Fig. 27 The double helical Cambridge model for the


erythromycin PKS, showing the tetrahedral relationship
between the co-operating KS and ACP domains in a given
module. The TE domains are also in intimate contact at the
base of the dimer, while the reductive activities form loops
out from the core of the structure.

In invoking the double helical model, the Cambridge group arbitrarily chose a left-handed
(counter-clockwise) helical twist (looking down the module from the KS) and it is equally probable that
the twist could instead run clockwise (Fig. 28). Within each module, the angle of coil brings the ACP of
one chain below the KS of the opposite subunit, which would facilitate the interaction of these domains
across the dimer interface (Fig. 27). Considering what happens when one module is stacked on top of
another, it is sensible that the axis of the first module serves as the axis for the second. There are two
possible arrangements for sequential modules, however, (Fig. 29): in the first, both helical modules have
the same direction of twist, and the bimodular pair forms a rope-like structure. In the second, the chirality
of the helix alternates from one module to the next. This latter organisation seems most unlikely because
there is strong homology in primary protein sequence between successive modules in a typical PKS, 37
which suggests that multiple modular structures evolved by successive gene duplication. 86 If so, the
chirality would not be expected to change from one module to the next.

31 von 66 20.11.2006 11:04


Published by the Royal Society of Chemistry http://pubs.rsc.org/ej/NP/2001/A909079G/

Fig. 29 The two possible arrangements of successive


modules in a type I PKS. (a) In this model, both helical
modules have the same direction of twist, and the bimodular
pair forms a rope-like structure. (b) In this arrangement, the
chirality of the helix alternates from one module to the next.

The Cambridge model satisfactorily accounts for all of the observed features of modular PKSs. Because
each helix has the same handedness, modular units can be added or subtracted without disrupting the
overall topology of the PKS. This structure also allows for variation in the number of modules in each
protein subunit and therefore in the overall number of modules in a PKS. Differences in the size of
modules due to the presence or absence of reductive domains are accommodated by the fact that these
activities loop out from the central core, and therefore do not interfere with the central structure of the
helix.
There have also been speculations from the US-based groups of David Cane and Chaitan Khosla
concerning the architecture of the DEBS proteins. 8789 The first, a linear head-to-tail model, was proposed
in 1996, to account for the results of complementation studies. 87 In this model, by analogy to the
conventional planar model for the animal FAS, 79,90,91 the DEBS modules are arranged in an antiparallel,
side-by-side configuration (Fig. 30). The resulting flat modules are then stacked one above the other and
connected by linker regions running vertically at the outside of the stack. This model satisfactorily
accommodates the data from complementation experiments 87,88 but it fails to account for the
homodimeric behaviour of key proteolysis fragments. 7577 In the head-to-tail model, the KS and AT
domains are placed far apart and so therefore could not possibly be in contact. Likewise, the TE domains
are greatly separated at the terminus of module 6, so again homodimer formation is not allowed. In
contrast, the helical model brings these activities into close proximity. Most tellingly, the results of recent
complementation studies on the animal FAS 80,81 have largely discredited the head-to-tail planar model for
the FAS on which this DEBS structure was modelled.

32 von 66 20.11.2006 11:04


Published by the Royal Society of Chemistry http://pubs.rsc.org/ej/NP/2001/A909079G/

Fig. 30 Alternative model for the structure of the DEBS PKS based on the conventional linear head-to-tail
model for the animal FAS. Each module forms a head-to-tail homodimer (as shown in the side views). Two
equivalent catalytic centres are present at opposite ends of the PKS complex, with each subunit contributing
half of the KS and ACP domains to each centre. The weakness of this model is that it fails to account for the
homodimeric behaviour of proteolysis fragments, including the KS-AT pairs and the TE.

More recently, Cane and Khosla have outlined an alternative model that, like the Cambridge structure,
has a head-to-head and tail-to-tail arrangement of the two co-operating chains. 89 This topology is
essentially equivalent to the earlier Cambridge proposal, and therefore needs no further discussion.

8.3 Docking of modular proteins

An intriguing feature of the multi-subunit organisation of the modular PKSs is the mechanism that
controls the order of the proteins in the synthases so that correct transfer of the growing chain from one
multienzyme to the next is achieved. Two experiments have recently begun to shed light on the structural
features required for productive communication between co-operating subunits. 92,93 A recombinant PKS
was constructed which contained two proteins derived from DEBS 1 and DEBS 2. 92 The first
multienzyme consisted of the loading module of DEBS and DEBS module 1 AT, KS and KR fused to the
ACP of module 2 (Fig. 31). The second protein contained DEBS module 3 fused to the terminal
thioesterase. A strain containing these multienzymes produced the expected product, demonstrating that
the ACP domain of module 2 can mediate the successful transfer of a polyketide chain from module 1 to
module 3 of DEBS. In another report, it was elegantly shown that, in several cases, the hand over could be
effective even if only the regions after the ACP domain and before the KS domain of the next subunit (the
linker regions) are maintained. 93

33 von 66 20.11.2006 11:04


Published by the Royal Society of Chemistry http://pubs.rsc.org/ej/NP/2001/A909079G/

Fig. 31 Docking of modular proteins. A recombinant PKS


constructed from portions of DEBS 1 and DEBS 2
demonstrated that communication between the two
proteins can be mediated by the ACP domain of module 2.

These results provide confirmation of the predictions embodied in the structural model for the modular
PKSs, 77 in which the strongest contacts in a PKS are between two copies of each module, interacting
across the dimer interface, and the interaction between successive modules is limited to adjacent ACP and
KS domains. 92

9 Genetic engineering of type I modular polyketide synthases

9.1 Polyketides as targets for combinatorial biochemistry

By analogy to combinatorial chemistry in which highly diverse libraries of compounds are assembled by
linking simple chemical building blocks, (for recent reviews, see, e.g., ref. 94 ), the emerging field of
combinatorial biochemistry is aimed at generating similarly rich libraries of novel molecules by mixing
and matching pieces of known natural products. The strategy relies on manipulating the genes that encode
the enzymes responsible for biosynthesising these metabolites. Using standard methods of genetic
engineering, enzymes are deleted or added, or their substrate specificity is tuned in a particular manner.
Depending on the possible ways in which this can be done, the number of molecules potentially accessible
by this route could be enormous. Recent advances in the development of oligonucleotide (reviewed in
refs. 95 and 96) and peptide (reviewed in ref. 97) libraries have highlighted some significant advantages of
using natural products to create molecular diversity: very large libraries can easily be made, stored,
amplified and screened, and candidate molecules obtained from such libraries can be serially modified and
reselected. 29
The structure of the modular PKSs offers an obvious starting point for manipulation, because there is
apparently a one-to-one correspondence between the complement of enzymes and the number of catalysed
steps. Therefore, when targeted changes are made in the genes, a clear prediction can be made for the
precise point at which chain assembly will be affected, assuming that all of the constituent domains act
independently. Possible alterations include deletion, substitution or addition of activities. Although not all
of the modified assembly lines will be viable, this approach offers a facile way of generating molecular
diversity. The idea of using combinatorial biochemistry to accelerate the pace of drug discovery and
development in the polyketide family is now a major area of study. 28,29
Many recent reviews have covered in detail the progress in genetic engineering in this field 28,29,89,98102
and so the strategies and results will only be summarised here.

9.2 Expression of polyketide synthases

The crucial technology for the expression of engineered PKSs is the availability of a host strain whose
gene set for biosynthesis of a particular polyketide (typically the one whose structure is being altered) has

34 von 66 20.11.2006 11:04


Published by the Royal Society of Chemistry http://pubs.rsc.org/ej/NP/2001/A909079G/

been eliminated. The absence of the cluster, including PKS and tailoring enzymes, allows individual
components of a PKS to be introduced and their function to be investigated in a clean background. The
first such strain to be created was a Streptomyces coelicolor derivative called CH999 from which the
majority of genes from the cluster encoding actinorhodin (15) biosynthesis had been deleted. 103
The desired PKS genes were then introduced into CH999 using a series of plasmids designed to
self-replicate in S. coelicolor, 104 and containing a thiostrepton resistance gene to enable selection. The
vectors also carry an origin of replication for E. coli as well as the ampicillin resistance gene for selection
in this bacterium. This combination of genes allows for rapid engineering to be carried out in E. coli
before transfer of the finished construct to S. coelicolor CH999. An additional component of the plasmids
is the actII-ORF4 gene, the natural pathway-specific activator of actinorhodin biosynthesis; 105 its protein
product serves to activate transcription from the PactI and PactIII promoters which are also cloned into the
plasmid. The PKS genes are introduced downstream of one or both of these promoters, so that the
appropriate growth signals drive expression of the synthase proteins. For example, the plasmid pRM5
(Fig. 32(a) 103 ) contains the act KS, chain length factor (CLF), ACP, aromatase (ARO), cyclase (CYC)
and KR genes cloned downstream of PactI and PactIII.

35 von 66 20.11.2006 11:04


Published by the Royal Society of Chemistry http://pubs.rsc.org/ej/NP/2001/A909079G/

Fig. 32 Vectors for expression of polyketide synthase


genes in Streptomyces. (a) pRM5 contains the activator
gene actII-ORF4 for regulated expression in a strain of S.
coelicolor (CH999) from which the act genes have largely
been deleted. The activator drives expression from genes
located downstream of the bidirectional promoter pair PactI
/PactIII. (b) pCJR24 also contains the actII-ORF4 activator
and PactI /PactIII promoter pair, but is designed for
integration into a strain of S. erythraea (JC2) from which
almost all of the eryA region has been removed.

CH999 was used initially to explore the roles of various genes involved in the biosynthesis of
actinorhodin itself, 103,106109 and later to express heterologous genes such as those involved in assembling
tetracenomycin 103,106,107,110113 (these experiments will be discussed in greater depth later in this review).
Subsequently, the strain was exploited for the expression of modular and fungal PKSs including those
responsible for erythromycin A 10, 114 epithilone 115 and 6-MSA 5 116 biosynthesis. Recently, an
engineered Streptomyces lividans strain (K4-114) which is also deficient in actinorhodin production has
been emerging as an alternative host for expression of heterologous PKSs. 117 As with CH999, the
plasmids used with K4-114 have relied on the actII-ORF4/actI activatorpromoter couple to stimulate
protein expression. DEBS 118,119 and portions of the pikromycin and oleandomycin PKSs 120 have been
expressed in this host. In both S. coelicolor and S. lividans it appears that the additional activities required
to activate the ACP domains by transfer of a phosphopantetheine arm are present. 99
A significant drawback to using unnatural hosts such as S. coelicolor or S. lividans for expression of
modular PKSs is that they lack the appropriate post-PKS tailoring enzymes, and consequently only the
polyketide macrolactone or its intermediates in the synthesis can be produced. 89,121 To achieve post-PKS
modification of erythromycin analogues, it was therefore desirable to express altered versions of the DEBS
PKS in its natural host Saccharopolyspora erythraea. Unlike S. coelicolor, however, S. erythraea cannot
transcribe genes that are introduced on autonomous plasmids, so any new PKS genes must be integrated
into its genome. By analogy to CH999, a strain of S. erythraea called JC2 was genetically engineered to
remove all of the PKS genes except for the terminal region of eryAIII encoding the chain-terminating
thioesterase. 122 Leaving this DNA intact allows for genes also bearing the TE to be delivered into the
chromosome by homologous recombination. A new integrative vector, pCJR24 (Fig. 32(b)), was then
developed for use with JC2 and other actinomycetes. 122 Like the plasmids employed with CH999 and
K4-114 it contains the actII-ORF4 activator gene and the actI promoter to turn on protein synthesis. To
demonstrate the utility of this system, the entire gene set for erythromycin biosynthesis was introduced
into S. erythraea. The resulting strain produced copious quantities of erythromycins and precursor
metabolites when compared to the unmodified organism. 122
A new and very promising alternative approach is to use E. coli itself for expression of PKS proteins; 93
E. coli does not biosynthesise any polyketides, and so the strain background is naturally clean . This
strategy has been enabled by the finding that co-expression in E. coli of a heterologous
phosphopantetheinyl (Ppant) transferase gene (e.g., sfp Ppant transferase from Bacillus subtilis 123 ) can
compensate for the deficiency of the host strain to carry out this crucial post-translational modification of
the PKS. 93 This strategy was first demonstrated by Simpson and co-workers, 206 who used a co-expression
system to ensure high levels of PKS holo-ACP expression in E. coli. To date, single module portions of
DEBS have been expressed successfully in E. coli. 93,124 The use of E. coli should increase as strains are
further engineered for the efficient production of polyketide metabolites from the introduced proteins. 125

9.3 Genetic engineering strategies

Genetic engineers have taken four basic approaches to modifying modular polyketide biosynthesis: (i)
altering of individual domains within modules; (ii) joining of intact modules; (iii) combining both natural
and engineered multimodular subunits; and (iv) feeding of synthetic precursors (termed mutasynthesis ).

36 von 66 20.11.2006 11:04


Published by the Royal Society of Chemistry http://pubs.rsc.org/ej/NP/2001/A909079G/

Additional diversity has been introduced by modifying the enzymes involved in post-PKS processing.

9.3.1 Domain mutagenesis. Genetic engineering of domains (by deletion, inactivation or swapping) has
been the most widely practised technique. To date, changes have been made to the nature of the starter
unit by exchange of loading domains, 46,126129 to the level and stereochemistry of keto reduction by
inactivating or swapping KR domains or by introducing additional reductive activities, 130133 to the
degree of chain branching by substituting AT domains of alternative specificity, 56,134137 and to chain
length by relocating the thioesterase domain to an upstream location in a PKS. 133,138140 Single mutations
have also been combined sequentially to produce multiple changes to a PKS. 141 These later experiments
have been reported to yield a small combinatorial library containing over 50 novel polyketides. Although
these studies demonstrate the feasibility of redirecting polyketide biosynthesis, the methods used are far
from trivial and are not yet amenable to large library production. Another significant limitation is the
generally low yields of product analogues increasing the number of mutations in a single PKS typically
results in a significant decrease in yield of polyketide product. 118,141

9.3.2 Linking of modules. An alternative strategy is to take an entire module as the combinatorial unit.
92,93
For this approach to succeed, however, the modular unit must be correctly identified. While the
structure of PKSs suggests that modules begin with a KS domain and end in an ACP, recent evidence from
two different experiments has challenged this view. 92,93 The first experiment demonstrated that a DEBS
module terminating in an ACP could not communicate effectively with a RAPS module introduced
downstream of it (Fig. 33). 92 Instead, the DEBS ACP required the presence of its partner KS in the next
module in order for the polyketide chain to progress between modules. This result has been refined by
experiments designed to investigate the role of linker domains in polyketide biosynthesis. 93 In creating
bimodular PKSs in which DEBS module 1 was fused to DEBS modules 3 and 6, the investigators found
that preserving the natural linker region between modules 1 and 2 gave a functional multienzyme (Fig.
34). Previous experiments in which the joins between the modules were made at arbitrary points in the
linkers failed to yield any polyketides, although the proteins were stable, could be purified and were at
least partially functional. 93 This result was extended to create a hybrid PKS containing DEBS module 1
and module 5 from the rifamycin PKS. These experiments narrowed the modular unit to include not the
entire KS domain of the next module, but only the linker region located in front of it. It is interesting that
while the first module requires its downstream linker the second module exhibits a higher degree of
tolerance to the protein structure which precedes it.

37 von 66 20.11.2006 11:04


Published by the Royal Society of Chemistry http://pubs.rsc.org/ej/NP/2001/A909079G/

Fig. 33 Engineered heterologous modular fusions. (a) An attempt to establish communication between module 1 of
and module 12 of the rapamycin PKS (RAPS) failed to result in the expected polyketide product. (b) An analogous
to that in (a), but with the natural ACP1-KS2 interaction preserved at the intermodular junction. This protein biosyn
the anticipated lactone. (c) The result in (b) was extended to give a trimodular PKS. In this case, KS2 was able to m
chain transfer between module 1 of DEBS and module 11 of RAPS to give the predicted tetraketide. Domains in gre
derived from the RAPS PKS.

Fig. 34 The role of linker domains in the docking of modular proteins. Constructs (a) and (b) are bimodular proteins
engineered from the erythromycin PKS, where successful chain transfer was observed between module 1 and an unn
downstream module (modules 3 and 6, respectively), as long as the intermodular linker that is naturally present be
modules 1 and 2 (shown in red) is maintained. In construct (c), module 1 from the erythromycin PKS was able to
communicate with module 5 from the rifamycin synthase, through module 1 s natural intermodular linker. The cons
shown in (d) demonstrates successful interpolypeptide chain transfer between module 5 of the rifamycin PKS and m
3 from DEBS. In addition to maintaining the module 1-module 2 linker from DEBS (shown in red), the interpolype
linker between module 2 and module 3 from the erythromycin PKS (shown in blue) was also preserved.

9.3.3 Combining protein subunits. A promising new approach is to engineer novel combinations of
whole protein subunits. 118,120 In the first example of this strategy, 118 each of the three DEBS genes were
cloned into three compatible Streptomyces vectors containing mutually selectable resistance markers; the
genes were introduced in both their native forms and containing a single mutation (domain swap or point
mutation). Introducing the plasmids sequentially into S. lividans generated 64 different triple
transformants, 46 of which produced detectable levels of polyketides. This system was also exploited to
create a small library of 12-membered lactones.
More recently, this multiplasmid approach has been used to form hybrid PKSs incorporating both native
and mutant subunits from the pikromycin, erythromycin and oleandomycin polyketide synthases. 120
Because the polyketide chains synthesised by these PKSs resemble each other it was reasoned that the
modules of one PKS would recognise and process the structurally related intermediates from another PKS.
In this case, the genes encoding subunits 1 and 2 (modules 14) of the pikromycin PKS (PikPKS) and the
gene encoding DEBS 3 were cloned into two compatible Streptomyces expression vectors. A strain of S.
lividans co-transformed with the two plasmids produced the expected hybrid polyketide (Fig. 35).
Co-expression of the same pik genes with the gene for the third subunit from the oleandomycin PKS was
also successful (Fig. 35). The PikPKS subunits 1 and 2 were also combined with genetically altered
versions of DEBS 3 containing a deleted or substituted domain (Fig. 35). Each of the combinations

38 von 66 20.11.2006 11:04


Published by the Royal Society of Chemistry http://pubs.rsc.org/ej/NP/2001/A909079G/

produced at least one polyketide product.

Fig. 35 Hybrid PKSs produced by combining whole protein subunits from the pikromycin (red), DEBS (yellow) and
using altered versions of DEBS 3 containing domains from RAPS (shown in green) or a domain deletion (indicated

Taken together, these studies demonstrate that a range of engineering strategies, from domain and
module substitutions to module fusion and subunit complementations, can be used for biosynthesising
new polyketides.

9.3.4 Mutational biosynthesis (mutasynthesis). A promising alternative towards increasing polyketide


diversity is mutational biosynthesis, or mutasynthesis, which couples genetic engineering with organic
synthesis. In this technique, a mutant PKS is constructed in which the first chain extension cycle is
disabled, typically by inactivating the KS domain in the first module so that the synthase cannot initiate
biosynthesis through a starter unit. Chain assembly is then kick-started by supplying a synthetic
non-natural diketide or triketide intermediate which includes novel functionality. If the added compound
is recognised by the PKS, the resulting polyketide will be altered at the appropriate position(s). 142147
This strategy was used initially to produce an analogue of avermectin with a non-natural starter unit 145

39 von 66 20.11.2006 11:04


Published by the Royal Society of Chemistry http://pubs.rsc.org/ej/NP/2001/A909079G/

and has been exploited more recently to produce similarly modified versions of 6-dEB and erythromycin
(Fig. 36). 146,147

Fig. 36 Mutasynthesis using a DEBS mutant in which KS1 had been selectively inactivated (designated DEBS (KS
Incorporation in vivo of synthetic analogues of chain-extension intermediates resulted in novel derivatives of
6-deoxyerythronolide B 16, which in some cases were further modified towards erythromycin analogues by post-PK
enzymes.

9.3.5 Post-PKS processing. After release from the PKSs, the carbon framework of polyketide
metabolites is typically enhanced with additional functionality; it is this secondary modification that yield
biologically active compounds. For example, the skeleton can be oxidised to introduce hydroxy or
carbonyl groups, methylated at oxygen, nitrogen or carbon centres, or decorated with deoxysugar
molecules. This post-PKS processing is another promising source of diversity in polyketide biosynthesis,
as there is enormous scope for mixing and matching of the tailoring enzymes to produce altered structures

40 von 66 20.11.2006 11:04


Published by the Royal Society of Chemistry http://pubs.rsc.org/ej/NP/2001/A909079G/

(see, e.g., refs. 148158 ). Additional efforts have been directed towards creating new sugars by modifying
the genes responsible for their biosynthesis, 158,159 and it should also be possible to tune or broaden the
specificity of the sugar transfer enzymes (glycosyl transferases) by DNA shuffling. 102 These approaches
are likely to be a major new area of natural products research, complementing efforts to alter polyketide
backbones.

10 Genetic engineering of type I iterative polyketide synthases

In contrast to the modular systems, it is not obvious how to best reconfigure the biosynthesis of reduced
fungal metabolites like lovastatin 28 (Fig. 24). Because the individual activities in fungal type I PKSs are
used iteratively, inactivation or swapping of a domain to cause a particular change could have
unpredictable and uncontrollable consequences for the rest of the pathway. It remains to be seen if the
methods of combinatorial biosynthesis will be successfully applied to these fascinating multienzymes.

11 Mechanistic studies in vitro on type I modular polyketide synthases

The rational design of large-scale combinatorial libraries remains a distant goal as many aspects of
polyketide biosynthesis remain poorly understood. Some important issues include the balance between
context and inherent specificity in substrate recognition by individual domains, the tolerance of activities
to changes in the structure and stereochemistry of intermediates, the precise mechanism and timing of
stereocontrol, and the relative importance of proteinprotein and proteinsubstrate interactions in guiding
the movement of biosynthetic intermediates between active sites.
The most direct way to gain insight into these sorts of questions is to carry out experiments with PKS
proteins outside of the bacterial cells (in vitro). The strength of such cell-free systems is that the properties
of the multienzymes can be probed under controlled conditions (e.g., temperature and buffer composition)
in the absence of metabolic constraints. Additionally, the in vitro environment provides a clean
background from which to isolate the often very small quantities of polyketide products.
The interpretation of such experiments, however, hinges on the use of purified proteins rather than crude
cell extracts, as other enzymes may have unpredictable effects on the reaction of interest. Additionally,
rigorous chemical analysis of all of the products should be carried out by high-resolution mass
spectrometry and NMR, rather than by chromatographic comparison with standards using thin-layer
chromatography. In summarising efforts in this area, we have therefore decided to emphasise the studies
that were carried out with pure multienzymes and where products were subjected to the highest standards
of analysis.

11.1 Studies in vitro with the DEBS PKS

Synthesis of 6-dEB 16 by DEBS in a cell-free extract was reported in 1995, 160 but has not been
subsequently reproduced. However, as the DEBS multienzymes constituted only 35% of total cellular
protein, no mechanistic conclusions can be drawn from this experiment. In early studies with purified
DEBS proteins, it was discovered that specific domains, such as the ATACP pair, were active in partial
reactions. 75,76 Further studies with intact DEBS proteins showed that all of the AT domains catalysed the
hydrolysis of the extender unit methylmalonyl-CoA from the AT active site in the absence of biosynthesis.
75
The finding that this reaction was selective for only the (2S )-isomer of methylmalonyl-CoA provided
important but circumstantial evidence for chain extension involving exclusively this isomer (more about
the significance of this observation later).
Although studies in vitro with the whole DEBS proteins provided some insights into erythromycin
biosynthesis, they were generally hampered by low quantities of protein as well as poor turnover to
product. 160 The low yield of product can most likely be attributed to the inefficiency of both the docking
of the three DEBS proteins in vitro and the transfer of the growing chain between the multienzymes.

11.2 Model systems for studying erythromycin biosynthesis: DEBS 1-TE and DEBS 1+TE

41 von 66 20.11.2006 11:04


Published by the Royal Society of Chemistry http://pubs.rsc.org/ej/NP/2001/A909079G/

The enormous size of the three-protein DEBS PKS makes it a very cumbersome system for detailed
mechanistic study. To facilitate such experiments, the PKS has been shortened to yield a significantly
smaller model system called DEBS 1-TE (Fig. 37). 138 This truncation was accomplished by moving the
thioesterase (TE) domain from the terminus of DEBS 3 to the end of DEBS 1 to cause premature release
of the chain at the triketide stage. A mutant strain of S. erythraea containing this bimodular mini-PKS
assembled two six-membered lactones, one with a propionate starter unit 32 and the other with acetate 33
(Fig. 37). To prove that the TE was playing an active role in forming these products, another PKS was
constructed in which a non-functional copy of the TE was placed at the end of DEBS 1. 138 This mutant
produced the two lactones in much lower amounts, demonstrating that the TE participates in chain release.
In contemporary experiments, biosynthesis of the lactones was also demonstrated with a similar construct
called DEBS 1+TE. 139 Both DEBS 1-TE and DEBS 1+TE have been purified to homogeneity and
characterised kinetically. 161163

Fig. 37 Construction of the bimodular PKS, DEBS 1-TE. Repositioning of the thioesterase domain
from the C-terminus of module 6 in DEBS 3 to the end of module 2 in DEBS 1, led to interruption of
the chain extension process and premature release of triketide intermediates as lactones 32 and 33.

The two mini-PKSs offer several obvious advantages over the whole DEBS system for studies in vitro.
Most importantly, at approximately 400 kDa apiece, they are significantly smaller than their parent
proteins (combined molecular weight of DEBS 13, approximately 1 MDa). And unlike the three DEBS
proteins that must dock together to form 6-dEB, purified DEBS 1-TE or DEBS 1+TE can operate
independently to make lactones. Finally, the lactone products include much of the functionality and all of
the stereochemistry present in the larger macrolide 6-dEB, but are significantly easier to analyse by
GC-MS, NMR and other techniques. The model systems can therefore serve as convenient substitutes for
the far more elaborate DEBS system.

42 von 66 20.11.2006 11:04


Published by the Royal Society of Chemistry http://pubs.rsc.org/ej/NP/2001/A909079G/

11.3 Substrate specificity of DEBS 1

DEBS 1-TE was first obtained in a high (95%) state of purity by the Cambridge groups of Leadlay and
Staunton in 1995. 164 When incubated with the appropriate substrates for synthesising lactone 32
propionyl-CoA, (2RS)-methylmalonyl-CoA and NADPH the protein functioned as expected; this
experiment was the first unequivocal demonstration of polyketide chain extension on a modular PKS in
vitro. As the protein was substantially pure it was apparent that DEBS 1-TE is self-sufficient for lactone
biosynthesis and does not require assistance from auxiliary activities. Another important discovery was
that the loading domain is more tolerant to alternative substrates than is evident in vivo; in addition to
acetyl-CoA and propionyl-CoA, both n-butyryl-CoA and isobutyryl-CoA were accepted as starter units.
This broad substrate specificity has recently been confirmed through studies in vitro on purified loading
didomain. 165

11.4 The molecular basis of stereocontrol

In 1965, Walter Celmer noted that, despite the variation within a single structure, there are strong
position-specific homologies among diverse polyketides (Fig. 38). 166,167 The fact that hundreds of
polyketides fit into common structural and stereochemical patterns 166169 suggests that the synthases
responsible for their assembly evolved from the same precursor and therefore that they act by common
mechanisms. It seems likely, therefore, that determining the mode of stereocontrol for one system should
provide insight into hundreds of others.

Fig. 38 Celmer model for the structural and stereochemical


homologies among the macrolide polyketides.

The key to understanding the control of stereochemistry is to identify the individual domains involved in
setting up the methyl and hydroxy configurations. The enzyme candidates are the AT domains that select
the chirality of the chain-extender unit, the KS domains that catalyse the condensation reactions and the
KR activities that are responsible for keto reduction. It was hoped that the cloning of PKS clusters such as
that for erythromycin would reveal the molecular basis for the control of stereochemistry in these systems.
However, while the genes readily predict the structures of the units added to the growing chains, there are

43 von 66 20.11.2006 11:04


Published by the Royal Society of Chemistry http://pubs.rsc.org/ej/NP/2001/A909079G/

no features of the sequences to aid in predicting the configurations of the many chiral centres in these
molecules.
The first significant insight into this question came instead from in vitro studies with purified DEBS
1-TE. The experiments revealed that the lactone is assembled exclusively from the (2S )-isomer of
methylmalonyl-CoA, 164 a finding which excluded the AT domains from playing a role. A subsequent
series of in vitro assays also with DEBS 1-TE and incorporating deuterium labelled substrates showed that
the condensation reactions in modules 1 and 2 occur with inversion of configuration, 170 as has been found
in fatty acid biosynthesis. 171 The experiments also went some way to revealing the complex interplay
between the KS and KR domains in achieving stereocontrol. It was discovered that in the second module
of DEBS, the KS establishes the methyl stereochemistry and the KR the hydroxy. In contrast, in module 1,
the KR likely determines the configuration at both stereocentres (Fig. 39). The suggestion that KR1 can
discriminate between its substrates was bolstered by further in vitro studies with DEBS 1-TE. 172 Although
some mechanistic details remain unexplained, taken together, these experiments revealed that there are
two distinct modes of co-operation between the KS and KR domains of each module. This complexity
needs to be considered carefully in future attempts to design synthases with altered stereochemistry of
chain extension.

Fig. 39 Stereochemistry of the condensation step in erythromycin biosynthesis. Experiments


with bimodular DEBS 1-TE in vitro have demonstrated that condensation of (2S
)-methylmalonyl-CoA proceeds with decarboxylative inversion. In module 2, this
condensation is followed directly by KR-catalysed reduction and, as a result, the KS sets the
methyl stereochemistry and the KR the hydroxy. The module 1 condensation, however, also
incorporates an obligatory epimerisation at the methyl centre. As the KR must wait until it
sees its preferred substrate for reduction, it controls the stereochemistry of the methyl and
hydroxy groups.

11.5 Studies in vitro of the erythromycin thioesterase

In addition to moving the TE to the end of DEBS 1, 138,139 the domain has been repositioned to cause
premature chain release at the diketide, 140 tetraketide 173 and hexaketide 139 stages (Fig. 40). Modified
versions of these products have also been created by additional engineering of these truncated PKSs.
131134
The viability of this approach for making novel polyketides depends critically on the ability of the
TE to release each new substrate. Knowing which structural features promote or prevent product release

44 von 66 20.11.2006 11:04


Published by the Royal Society of Chemistry http://pubs.rsc.org/ej/NP/2001/A909079G/

by the TE will therefore be vital for future genetic experiments.

Fig. 40 Chromosomal repositioning of the thioesterase domain to locations within the


erythromycin PKS. The TE has been successfully relocated to the end of modules 2, 3 and 5,
and in each case has catalysed the formation of the anticipated lactone.

To facilitate studies, the TE has been expressed in E. coli either with its companion ACP (Fig. 41) 174,175
or as a monodomain. 176

45 von 66 20.11.2006 11:04


Published by the Royal Society of Chemistry http://pubs.rsc.org/ej/NP/2001/A909079G/

Fig. 41 Cloning and overexpression in E. coli of the TE


along with its partner ACP.

11.5.1 Substrate specificity of the TE. The ACP-TE didomain has been used to investigate the
behaviour of the TE when challenged with a variety of synthetic substrates (Fig. 42). 174,177,178 The
qualitative results of these studies have demonstrated that the TE is broadly tolerant to a range of
functionality and chain length, although it does exhibit a preference for medium sized chains ( 10
carbons); 174 compounds with terminal hydroxy functionality are better substrates than simpler compounds
of the same length. 174 It was also shown that, in addition to chain release via macrolactone ring formation,
the TE can off-load substrates as free acid and as esters. 174,177,178 Finally, the presence of a methyl group
at C-2 and, less importantly, a hydroxy functionality at C-3 (in any stereochemical configuration) was
found to suppress the hydrolytic release mechanism. 177,178 These findings will be important in attempts to
achieve ring closure of novel polyketides. Disappointingly, the thioesterase failed to lactonise any of the
synthetic substrates, even those incorporating the normal pattern of structure at C-2 and C-3. 178,179
Apparently, ring formation requires some extra as yet determined structural features in the polyketide
chain.

46 von 66 20.11.2006 11:04


Published by the Royal Society of Chemistry http://pubs.rsc.org/ej/NP/2001/A909079G/

47 von 66 20.11.2006 11:04


Published by the Royal Society of Chemistry http://pubs.rsc.org/ej/NP/2001/A909079G/

Fig. 42 Synthetic compounds used to investigate the structural and stereochemical specificity of the erythromycin
thioesterase. The natural substrate of the TE is shown at the top. (X = p-nitrophenol and/or N-acetylcysteamine).

Experiments with the TE monodomain and a more limited range of substrates (Fig. 43), though differing
in quantitative aspects, produced similar conclusions. 176 The studies also demonstrated that the covalent
link between the TE domain and DEBS 1 in DEBS 1-TE is crucial for the TE s role in release of the
lactone.

Fig. 43 Synthetic compounds used to investigate the structural and stereochemical specificity of the
erythromycin thioesterase as a monodomain.

12 Genetic studies on the type II PKS systems

It will be apparent from the account of the erythromycin PKS that significant advances have been made
over the last ten years in studies of the structure, function and genetic modification of this type I system. It
is now interesting to review in a comparative way the results of investigations into the most heavily
studied type II system, the actinorhodin PKS.
The characteristic products of this type of PKS are phenolic aromatic compounds. The biosynthetic
reactions are, therefore, involved with the condensation of building blocks to produce polyketone chains
and the subsequent cyclisation and aromatisation processes to produce aromatic compounds. In short, we
are now concerned with the biological processes that were the subject of Collie s speculations at the start
of the twentieth century. As we mentioned earlier, it was Hopwood and his colleagues who pioneered the
genetic approach to understanding the biosynthesis of type II polyketides. The background to these
developments has been comprehensively covered in two fascinating reviews by Hopwood. 86,180 In the
present review, therefore, we will concentrate on the chemistry and enzymology rather than the genetics.
The spark that ignited interest in the genetics of the type II polyketides was Hopwood s sequencing of
the genes associated with the biosynthesis of the antibiotic actinorhodin 15. 27 The key genes associated
with polyketide chain assembly, cyclisation and aromatisation are clustered on the genome as indicated in
Fig. 44. A second seminal contribution of the Hopwood group, mentioned in Section 9.2, was the
engineering of a mutant (CH999) of the producer organism (S. coelicolor) in which these key genes and
others associated with the later stages of the pathway were removed, and also a complementary plasmid,
pRM5 (Fig. 32(a)), which contained the necessary genetic apparatus to reintroduce the missing genes in
different combinations. This technology made possible an elegant series of plug-and-play experiments in
which the behaviour of different combinations of genes could be investigated. Hopwood recognised that
the products of such synthases are chemical artefacts produced from the proposed biosynthetic

48 von 66 20.11.2006 11:04


Published by the Royal Society of Chemistry http://pubs.rsc.org/ej/NP/2001/A909079G/

intermediates, probably after release from the enzyme.

Fig. 44 (a) Organisation of the act genes of Streptomyces coelicolor. (b) The act minimal PKS, consisting of the ket
the acyl carrier protein (ACP, actI/ORF3), is believed to assemble an unreduced polyketone chain from one unit of
minimal PKS is also thought to catalyse the formation of the first carbocyclic ring. The enzyme products of the rem
actIV (cyclase, CYC) are proposed to carry out the reactions as shown. The metabolites DMAC 35 and aloesaponar
genes.

In an early experiment, cultures of CH999/pRM5 were found to make aloesaponarin II 34, as well as its
expected undecarboxylated precursor, 3,8-dihydroxy-1-methylanthraquinone-2-carboxylic acid (DMAC)
35 (Fig. 44). 103 This result confirmed conclusions from earlier studies 181 and encouraged the subsequent
deletion experiments, because it was evident that pRM5 contained all of the genes needed to encode a
correctly programmed PKS without the complication of tailoring enzymes. 180 Progressive deletion of
genes led to a series of mutants which generated increasingly primitive products (in terms of degree of
cyclisation and aromatisation) 106109 until finally the minimal PKS was reached, consisting of a KS, an
ACP and a mystery component which had no obvious active site or predictable catalytic role. 106 The
product of the act minimal PKS was SEK4 36, with an unreduced C16 chain and just the first carbocyclic

49 von 66 20.11.2006 11:04


Published by the Royal Society of Chemistry http://pubs.rsc.org/ej/NP/2001/A909079G/

ring correctly formed by aldol condensation between C-7 and C-12; the rest of the structure arose by the
most probable uncatalysed chemistry, in this case to give a hemiketal at the methyl end of the chain and a
pyrone at the carboxyl end. 110
The early advances in the sequencing of the actinorhodin cluster were rapidly followed by similar
sequencing successes with other type II clusters (see, e.g., refs. 182193 ). The core genes associated with
formation of the polyketide skeletons showed remarkable homology, both in terms of the component
activities and the way they were organised on the respective genomes. A representative set of four such
gene sets is shown in Fig. 45. 180 The first three (act, tcm and otc) have a standard set of clustered enzymes
for the minimal PKS; in the fourth (dps/dau), the gene for the ACP is located at a site remote from the
other two. The evolutionary significance of these patterns of genes has been discussed by Hopwood. 180

Fig. 45 Organisation of gene clusters for type II PKSs (act, actinorhodin; tcm, tetracenomycin; otc,
oxytetracycline; dps/dau, daunorubicin).

The emergence of these additional sets of genes allowed Hopwood, now in collaboration with Khosla, to
carry out experiments in which genes from different PKSs were mixed and matched. 103,106113,194,195
These studies led to the production of an extensive set of primitive polyketide products reminiscent of
SEK4 36, but with different chain lengths and modes of cyclisations. In rationalising their results,
Hopwood noted that switching ACP components did not affect product structure, but that swapping the
mystery component of the minimal PKS could alter chain length. They therefore assembled a set of design
rules 194 which sought to predict the synthetic outcome of specific combinations of genes from different
sources (for a summary of the rules, see ref. 180 ). An essential element of the design rules was the
proposal that control of chain length rested with the mystery component, and so this was designated the
chain length factor (CLF ) . 113
Meanwhile, Dick Hutchinson and co-workers, on the basis of similar experiments with the
tetracenomycin PKS, came to the view that the mystery component of the minimal PKS is not the sole
arbiter of chain length control. 196 They concluded, instead, that both the KS and the CLF co-operate to
determine chain length and proposed a different nomenclature for the minimal PKS: KS for the KS, KS

50 von 66 20.11.2006 11:04


Published by the Royal Society of Chemistry http://pubs.rsc.org/ej/NP/2001/A909079G/

for the mystery component, and ACP; 197 these designations are now accepted by most practitioners in the
field. An additional set of experiments demonstrated that the tetracenomycin minimal PKS itself is
inefficient at biosynthesising a chain of the appropriate length. Hutchinson and co-workers prepared a
cell-free extract of a mutant that possessed a set of genes corresponding to the minimal PKS of the
tetracenomycin cluster. 198 When supplied with standard polyketide precursors, acetyl-CoA and
malonyl-CoA, this extract made a mixture of products from which SEK15 37 and Tcm F2 38 were
characterised (Fig. 46); the latter compound has the correct chain length and folding to produce the
tetracenomycin system while SEK15 has the appropriate length but incorrect folding. Intriguingly, there
were also a number of uncharacterised products. Addition of another component of the tetracenomycin
PKS (a cyclase called TcmN) radically altered the composition of the product mixture, so that TcmF2 was
the major metabolite. Hutchinson concluded that the minimal PKS does not have the ability to efficiently
control the chain length and manner of folding of the polyketone chain to produce a single product, and
that the optimal TcmPKS is instead a complex that also includes the TcmN protein. The wider aspects of
anthracycline biosynthesis are covered in four excellent reviews. 199202

Fig. 46 Role of TcmN in the biosynthesis of tetracenomycin F2 38. In the absence of TcmN, the TcmJ, K, L and M
enzymes together made SEK15 37 and a mixture of uncharacterised products, along with the natural intermediate
tetracenomycin F2. When TcmN was added, Tcm F2 became the major product.

A recent paper by the group of Iain Hunter describes results which show that the products of a minimal
PKS can not only have different folding patterns, but also different chain lengths. 203 In this work, the gene
coding for the putative aromatase/cyclase of the oxytetracycline cluster was disrupted. The resulting
mutant (which would be expected to have a complete normal minimal PKS set) generated a mixture of
products from which the range of compounds shown in Fig. 47 were isolated. One is a tetraketide, one a
heptaketide and the other two are octaketides with different folding patterns. All four compounds are
probably chemical artefacts derived from spontaneous cyclisations of incompletely formed polyketone
chains produced by the minimal PKS. No nonaketide compounds corresponding to the completely formed
polyketide intermediate were reported, but small amounts may have escaped detection.

51 von 66 20.11.2006 11:04


Published by the Royal Society of Chemistry http://pubs.rsc.org/ej/NP/2001/A909079G/

Fig. 47 Disruption of an aromatase/cyclase from the


oxytetracycline gene cluster of Streptomyces rimosus led to
the production of four polyketides (shown in relationship to
their putative acyclic precursors). One is a tetraketide, one a
heptaketide and the other two are octaketides with different
folding patterns.

13 Studies in vitro of type II PKSs

Two other recent discoveries that relate to the mechanism of chain extension by the minimal PKS, rather
than its possible role in folding or cyclisation, also merit highlighting. The Simpson group at Bristol have
carried out a focused programme on the structure and function of the ACP components of a series of
minimal PKSs, including the actinorhodin ACP. 2,3,204206 The full solution structure of the act ACP, 2
assigned by high field NMR, shows remarkable similarity to the simple fatty acid ACP derived from E.
coli, 207210 except that the putative chain binding pocket in the polyketide ACP is less lipophilic in

52 von 66 20.11.2006 11:04


Published by the Royal Society of Chemistry http://pubs.rsc.org/ej/NP/2001/A909079G/

character.
The availability of relatively large quantities of reliably pure ACP has allowed the Bristol group to
discover a remarkable and exceptional property of the type II polyketide ACPs: the ability to accept a
malonate and other acyl groups from their respective CoA thioesters to catalyse acyl transfer reactions
apparently without external assistance. 204,211 Recently, Kevin Reynolds has confirmed the result and
provided further evidence that a type II ACP is not just an ACP, but is also endowed with a powerful AT
activity. 212 The absence of an acyltransferase activity in the type II polyketide clusters has been a
long-standing and intriguing puzzle. It has been suggested that the AT component of the fatty acid
synthase in S. coelicolor may also serve the PKS. 213,214 With the Simpson discovery, it is no longer
necessary to invoke this device, but, nonetheless, the true mechanism of malonylation in vivo remains to
be established.
The second important discovery came about as a consequence of an inspired piece of deductive
reasoning by Leadlay concerning the role of KS . He noted, as others had before, that this component
showed remarkable sequence homology to its partner, KS , with the important exception that its active
site cysteine was missing. However, he also noticed that the residue that replaced the cysteine was not
random, but was always a glutamine. He therefore suggested that this replacement might cause KS to be
a decarboxylase toward malonate rather than acting as a conventional KS partner in chain extension, and
would have the required properties of a chain initiation factor that generates acetyl-ACP in situ on the
minimal PKS. 215 In collaboration with Simpson s group at Bristol, this idea was put to the test.
A mutant minimal PKS was created by substituting the KS active site cysteine for alanine, a change
that disabled polyketide biosynthesis. The resulting KS /KS complex was rigorously purified, then
tested for its ability to decarboxylate malonyl-ACP. Analysis by electrospray MS showed that the
malonyl-ACP was indeed readily decarboxylated to acetyl-ACP (Fig. 48(a)). A mutant in which both the
KS /KS active site residues were changed to alanine had no decarboxylase activity. Together, these
results unequivocally implicated the KS and its active site glutamine in the provision of starter units for
aromatic polyketide assembly. 215 These results were bolstered by contemporary experiments carried out
by Stuart Smith and co-workers in the context of the fatty acid synthase. These researchers demonstrated
that a condensing KS could be transformed into a decarboxylase simply by replacing its active site
cysteine with a glutamine. 216 So KS , originally wrongly assigned the role of chain length factor, now
moves to centre stage once more as a chain initiation factor. It has also been established that in certain
type I modular PKSs a similar decarboxylative component (KSQ) forms part of the loading module. 215

53 von 66 20.11.2006 11:04


Published by the Royal Society of Chemistry http://pubs.rsc.org/ej/NP/2001/A909079G/

Fig. 48 Sequence of events in the biosynthesis of aromatic polyketides. (a) Chain extension is initiated by
decarboxylation of malonyl-ACP to acetyl-ACP by the KS domain. (b) The acetyl group is then passed from
the ACP to the active site cysteine of the KS . The ACP is loaded with another unit of malonate and then
KS-catalysed condensation takes place resulting in acetoacetyl-ACP. The ketoester is then transferred back to
the KS , and another cycle of chain extension can begin.

14 Summary of the advancements in the understanding of type II polyketide


biosynthesis

We will now try to draw together these various advances to present the current understanding of the field.
In some respects, our task is complicated by a lack of consensus between the various groups working in
the area. It would be a pity, however, not to give an overview, first of chain formation and secondly of
cyclisation and aromatisation.
After decarboxylation of malonyl-ACP, the acetyl group is then transferred to the KS active site to
provide the starter unit for the first condensation step (Fig. 48(b)). The ACP is malonylated again and as
the condensation reaction of KS is available the reaction proceeds to give acetoacetyl-ACP. This
ketoester is then transferred from the ACP back to KS and the second cycle of chain extension can
commence. For this scheme to occur, some mechanism of control is necessary to ensure that the correct
balance is maintained between the chain initiation and chain extension chemistry. How this control is
achieved remains to be established.
The mechanism responsible for governing chain length is still shrouded in mystery despite the heroic
efforts of Hopwood, Khosla, Hutchinson, Hunter and others to investigate the role of the minimal PKS and
the various putative aromatases and cyclases found associated with most type II PKS clusters. One
significant challenge of investigating this question lies in the difficulty of the chemical analysis. All too
often a complex mixture of PKS products with very similar properties is produced by mix-and-match
experiments. Additionally, in an in vivo study, the polyketide metabolites have to be isolated from a
complex mixture of materials used in the fermentation media. Under these circumstances, there is a high
risk that important products of the biosynthetic system will not be detected. We will therefore confine our
comments to a few general observations in which we draw parallels with the earlier efforts of synthetic
chemists to carry out biomimetic syntheses.

54 von 66 20.11.2006 11:04


Published by the Royal Society of Chemistry http://pubs.rsc.org/ej/NP/2001/A909079G/

To produce a specific polyketide skeleton leading to an aromatic polyketide, two characteristics have to
be determined: length of the chain and site of its folding back on itself to produce the first aromatic ring.
In contrast to earlier speculations that each component would have a single function, all of the genetic
experiments suggest that many activities have more than one role and that in fact these roles depend on the
other components present in the PKS. It is likely, therefore, that a complex association of many proteins is
needed for the effective and controlled operation of a type II PKS and that removal of one component
could have unpredictable indirect effects on the behaviour of the remaining activities.
Whatever the mechanism, it is reasonable to speculate concerning the general principle of control of
chain length and cyclisation. Using actinorhodin as an example, the issues are timing of cyclisation to
form the first ring in the octaketide intermediate (Fig. 44), and also the timing of the reductive step at C-9.
Two possible schemes are shown in Fig. 49. In the first pathway (Fig. 49(a)), the whole series of seven
chain extensions is accomplished prior to cyclisation and also possibly before the reduction. The
disadvantage of this mechanism is that the long reactive polyketone chain must be protected during the
chain extension steps. An alternative pathway shows cyclisation taking place at the hexaketide stage (Fig.
49(b)). In this scheme, all that is required to select the desired cyclisation mode is to protect the
intermediate acetoacetate residue from forming a carbanion, so that the first opportunity to cyclise occurs
at the hexaketide stage. The formation of the first ring removes the need to control the regiochemistry of
subsequent cyclisations, as was pointed out earlier. 20 After cyclisation, the chain is then extended by two
additional ketide units to produce a molecule of the correct length; at this point, we rejoin the Hopwood
scheme (Fig. 44).

55 von 66 20.11.2006 11:04


Published by the Royal Society of Chemistry http://pubs.rsc.org/ej/NP/2001/A909079G/

Fig. 49 Possible schemes for the timing of cyclisation in actinorhodin biosynthesis. (a) The whole series of
chain extensions is carried out before cyclisation and possibly before reduction. In this scheme, a long
reactive -keto chain must be protected from spontaneous cyclisation throughout chain extension. (b) In
this alternative mechanism, cyclisation takes place at the hexaketide stage. Ring formation at this point in
chain extension only requires that the initially generated acetoacetate is protected. Additionally, creation of
the first ring removes the need to control the regiochemistry of subsequent cyclisations.

The important difference between the two proposals is that in the first case there needs to be some
mechanism to protect a long polyketone chain containing eight keto carbonyl groups. In the alternative
scheme, only the first two groups need protecting, but, more significantly, control of chain length is
broken into two separate operations the length to the first cyclisation and the number of subsequent
chain extensions to achieve the finished product. In this scheme, then, there could be two separate
components controlling different aspects of chain length, an idea that would go some way to explaining
the confusing picture that has emerged from the mix-and-match experiments. The possible candidates for
the role of ketone chain protection are the active site of the KS , the arginine-rich core of the ACP as has
been suggested by Simpson et al., 2 or even one of the active sites involved in the control of cyclisation.
The other factor that must play a role in determining chain length is the mechanism of release of the
fully extended molecule. Puzzlingly, there is no dedicated TE associated with type II PKS systems such as

56 von 66 20.11.2006 11:04


Published by the Royal Society of Chemistry http://pubs.rsc.org/ej/NP/2001/A909079G/

the act cluster. One possible mode of release for these systems is through a ketene mechanism related to
that proposed by Peter Shoolingin-Jordan. 8 The timing of release would then have to be determined by
some agent in the PKS complex. As an aside, it is interesting to note that a survey of the chemical artefacts
produced by the many experiments with minimal or depleted PKSs reveals an overwhelming
predominance of -pyrones following the latest ring. Formation of an -pyrone from a 3,5-diketo ester
would be a very favourable chemical mechanism for chain release. It may be, therefore, that the pattern of
chain length observed in in vitro experiments is a reflection of this anomalous mechanism of chain release
rather than the intrinsic chain length preference of the PKS system under study. This may be a further
source of confusion in attempts to elucidate the design rules for chain length.

15 Biosynthesis of plant polyketides

Before concluding this review, it is appropriate to briefly discuss a third family of polyketide synthesising
enzymes, the chalcone/stilbene synthase-type family of enzymes found in higher plants. 217 Chalcone
synthases (CHSs), the most well-known representatives of this family, provide the starting materials for a
diverse set of metabolites (phenylpropanoids) with roles in pigmentation, protection against UV
photodamage, defence against pathogens (phytoalexins) and interactions with micro-organisms. 218 Of
greater interest to human health, the phenylpropanoids also exhibit cancer chemopreventive, 219
antimitotic, 220 estrogenic, 221 antimalarial, 222 antioxidant 219 and antiasthmatic 223 activities. It is these
latter activities that have sparked interest in reengineering the pathways towards the production of novel
compounds. 217,224
Synthesis of chalcone by CHS involves the sequential condensation of a starter unit p-coumaroyl-CoA
with three extender units (most typically malonyl-CoA) (Fig. 50). 224 The resulting tetraketide intermediate
is ultimately cyclised by a Claisen condensation into a hydroxylated aromatic ring system. While this
mechanism superficially resembles that biosynthesis by type I and II systems, there are a number of
important differences. Most significantly, while PKSs rely on the phosphopantetheine arms of ACPs to
shuttle substrates and intermediates among active sites, the CHSs instead uses CoA thioesters directly.
Also, in contrast to the modular (type I) or multicomponent (type II) PKSs, the CHSs carry out a complete
series of decarboxylation, condensation, cyclisation and aromatisation reactions with a single active site.
224

57 von 66 20.11.2006 11:04


Published by the Royal Society of Chemistry http://pubs.rsc.org/ej/NP/2001/A909079G/

Fig. 50 Reactions of chalcone synthase (CHS). 217 The


starter unit is coloured in blue, and the numbers indicate the
three condensation reactions.

Although nearly 150 CHS-related genes have been sequenced, significant insight into how a single
active site controls the chemistry of multiple decarboxylation and condensation reactions only emerged
with the determination of the crystal structure of CHS2 from alfalfa. 224

15.1 Structural studies of the chalcone synthases

In 1999, Joseph Noel and co-workers determined the crystal structures of CHS2 alone and complexed with
substrate and product analogues. 224 These structures were the first ever of any PKS component! The
researchers found that CHS2 is a homodimer (MW 84 kDa) whose overall fold resembles that of
-ketoacyl synthase (KAS) II of E. coli (involved in fatty acid biosynthesis) 74 as well as the thiolase from
Saccharomycescerevisiae. 225,226 The structure also suggested that four residues, Cys164, Phe215, His303,
and Asn336 which are conserved among all known CHS-related enzymes, are the primary participants in
catalysis. The roles of these residues have been pinpointed by structural and functional characterisation of
16 CHSs point mutants the Cys serves as an active-site nucleophile, the His and Asn are responsible for
malonyl-CoA decarboxylation, and the Phe may help to orient substrates during chain elongation (Fig. 51).
227
The His has also been shown to stabilise the active site nucleophile through formation of a
thiolate-imidazolium ion pair (Fig. 51). 228 The crystal structure further revealed three interconnected
cavities that intersect with the four catalytic residues and form the active site architecture of the CHS: a
CoA-binding tunnel, a coumaroyl-binding pocket and a cyclisation pocket. 224 This division of the active
site into discrete pockets provides a structural basis for the ability of the CHSs to orchestrate the multiple
reactions of chalcone synthesis.

Fig. 51 Proposed reaction mechanism for the CHS, showing the loading, decarboxylation and elongation
steps. 228 R is the coumaroyl moiety in the first reaction cycle, coumaroyl-acetyl in the second, and
coumaroyl-diacetyl in the third. In the loading step, the proposed cysteinehistidine ionic interaction is
indicated.

Noel and co-workers have also determined the crystal structure of a related enzyme, 2-pyrone synthase
(2-PS), which forms a triketide methylpyrone from acetyl-CoA and two units of malonyl-CoA. 229 They
found that while 2-PS and CHS share a common fold, the same catalytic residues and a similar CoA
binding site, the active site cavity of 2-PS is smaller than that of CHS. Armed with this information, the
researchers were able to functionally convert the CHS into a 2-PS by reducing the size of its
initiation/elongation cavity. This experiment demonstrated that cavity volume influences both the choice
of starter molecule and the final length of the polyketide. 229
Although the sequences of CHS-related enzymes have no obvious similarity to the type I and type II
PKSs, 217 the homology of active site residues 228 as well as shared structural features 227 suggest that
observations about CHSs may well have relevance to the productive reengineering of bacterial PKSs.

15.2 Bacterial analogues of CHSs

58 von 66 20.11.2006 11:04


Published by the Royal Society of Chemistry http://pubs.rsc.org/ej/NP/2001/A909079G/

One of the more surprising findings to emerge from this field was that the chalcone-related synthases are
not after all unique to plants. 230 Sueharu Horinouchi and co-workers demonstrated that the rppA gene of
Streptomyces griseus encodes a homodimeric, small molecular weight protein (RppA) (MW 90 kDa)
which assembles and cyclises a pentaketide product. Site-directed mutagenesis showed that RppA has a
catalytically essential cysteine, as expected for a CHS-type enzyme. They therefore concluded that RppA
is a CHS-related synthase. A computer-aided search further revealed that RppA enzymes are involved in
the biosynthesis of a wide range of secondary metabolites, not only in the Streptomyces, but in many other
bacteria.

16 Conclusions

Having surveyed the developments in this field over the last century, it is clear that enormous progress has
been made, but that much more remains to be discovered. It seems fitting therefore to speculate on the
future of polyketide research.
One area that should see enormous interest is the determination of PKS structure. Such information is of
obvious academic importance, but is also vital for the rational re-configuring of these synthases for the
production of novel compounds. In the modular type I field, various analytical techniques have yielded
some insight into the overall topology of these gigantic multienzymes. However, these experiments have
yet to explain how the component domains co-operate through space, as well as more subtle behaviours
such as substrate specificity and stereochemical control. It will therefore be necessary to not only
determine the three-dimensional structure of the individual enzyme activities with and without bound
substrates and/or inhibitors, but to further explore the quaternary architecture which maintains the correct
spatial arrangement of the interacting domains. The iterative type I PKSs represent an even greater
challenge as we must explain how a given set of domains can operate repeatedly yet carry out different
chain extension processes in different cycles. The type II dissociable systems, on the other hand, pose the
opposite problem. While NMR and X-ray crystallography have already yielded high-resolution structures
of individual domains, the nature of the quaternary association that must take place in these systems is
unknown. It is likely that this lack of knowledge has contributed to the failure of mix-and-match
experiments to yield predictable products. The architecture of the type II iterative PKSs must be
determined before their biosynthesis can be efficiently re-engineered.
Issues such as the catalytic mechanism of individual domains as well as the molecular determinants of
substrate and stereochemical specificity will continue to be probed by coupling standard mutagenesis
techniques with studies of purified enzymes in vitro. Efforts are also likely to focus on understanding the
means by which the individual activities, arrayed in their various synthases, communicate with each other
as they guide the growing chain between them. These types of experiments should provide crucial
information on exactly which subunits or groups of domains can be swapped without compromising the
workings of the integrated complexes.
On a more applied level, researchers in industry will be seeking to maximise the efficiency of analogue
generation by developing more sophisticated genetic tools, engineering improved producing strains and by
optimising fermentation procedures. 125 These efforts should help to bring the goal of large library
production within reach. Additionally, as bacterial and fungal hosts may not be able to tolerate novel
metabolites, work should also continue on the development of alternative hosts for polyketide expression.
Finally the fields of combinatorial chemistry and biochemistry may be united by using engineered
polyketides as scaffolds for further chemical elaboration.
Over the past 50 years, the polyketide natural products and their semi-synthetic derivatives have played
an enormous role in human and veterinary medicine. The total market for these compounds exceeds $10
billion per annum. 231 As organisms become resistant to these drugs and many other classes of molecules,
the race to find new drug candidates becomes increasingly urgent. It is likely that over the next half
century, these new unnatural natural products will come increasingly from the genetic engineering of
existing producing organisms rather than through the discovery of novel natural natural products as yet
unidentified in the biosphere. The polyketide field will, therefore, continue to attract interest in the
commercial sphere as well as in academia. We can guarantee that the hiatus that afflicted this field in the
first half of the twentieth century will not be repeated in the twenty-first.

59 von 66 20.11.2006 11:04


Published by the Royal Society of Chemistry http://pubs.rsc.org/ej/NP/2001/A909079G/

References

1 J. N. Collie and W. S. Myers, J. Chem. Soc., 1893, 63, 122.


2 M. P. Crump, J. Crosby, C. E. Dempsey, J. A. Parkinson, M. Murray, D. A. Hopwood and T. J.
Simpson, Biochemistry, 1997, 36, 6000.
3 M. P. Crump, J. Crosby, C. E. Dempsey, M. Murray, D. A. Hopwood and T. J. Simpson, FEBS Lett.,
1996, 391, 302.
4 J. N. Collie, J. Chem. Soc., 1907, 91, 1806.
5 R. Robinson, J. Chem. Soc., 1917, 111, 762.
6 R. Robinson , The Structural Relations of Natural Products, Clarendon Press, Oxford, 1955.
7 A. J. Birch, P. A. Massy-Westropp and C. J. Moye, Aus. J. Chem., 1955, 8, 539.
8 P. M. Shoolingin-Jordan and I. D. G. Campuzano , in Comprehensive Natural Products, ed. U.
Sankawa, Elsevier, Oxford, 1999, pp. 345365.
9 R. E. T. Minto and C. A. Townsend, Chem. Rev., 1997, 97, 2537.
10 C. A. Townsend and R. E. Minto , in Comprehensive Natural Products, ed. U. Sankawa, Elsevier,
Oxford, 1999, pp. 443471.
11 T. J. Simpson, Biosynthesis, 1998, 195, 1.
12 T. J. Simpson, Chem. Soc. Rev., 1987, 16, 123.
13 R. A. Hill, R. H. Carter and J. Staunton, J. Chem. Soc., Perkin Trans. 1, 1981, 2570.
14 D. E. Cane, P. C. Prabhakaran, W. Tan and W. R. Ott, Tetrahedron Lett., 1991, 32, 5457.
15 C. R. Hauser and T. M. Harris, J. Am. Chem. Soc., 1958, 80, 6360.
16 T. M. Harris and G. P. Murphy, J. Am. Chem. Soc., 1971, 93, 6708.
17 T. M. Harris and T. P. Murray, J. Am. Chem. Soc., 1972, 94, 8253.
18 T. M. Harris , C. M. Harris and K. B. Hindley , in Progress in the Chemistry of Organic Natural
Products, ed. W. Herz, H. Grisebach and G. W. Kirby, Springer Verlag, Vienna, 1973.
19 T. M. Harris and P. J. Wittek, J. Am. Chem. Soc., 1975, 97, 3270.
20 F. J. Leeper and J. Staunton, J. Chem. Soc., Perk. Trans. 1, 1984, 1053.
21 T. F. Money, F. W. Comer, G. R. B. Webster, I. G. Wright and A. I. Scott, Tetrahedron, 1967, 23,
3435.
22 L. J. Crombie and A. W. G. James, J. Chem. Soc., Chem. Commun., 1966, 357.
23 C. Abell, B. D. Bush and J. Staunton, J. Chem. Soc., Chem. Commun., 1986, 15.
24 P. M. Shoolingin-Jordan and J. B. Spencer, Biochem. Soc. Trans., 1993, 21, 222.
25 J. B. Spencer and P. M. Shoolingin-Jordan, Biochem. J., 1992, 288, 839.
26 C. J. Child, J. B. Spencer, P. Bhogal and P. M. Shoolingin-Jordan, Biochemistry, 1996, 35, 12 267.
27 F. H. Malpartida and D. A. Hopwood, Nature, 1984, 309, 462.
28 J. Staunton, Curr. Opin. Chem. Biol., 1998, 2, 339.
29 P. F. Leadlay, Curr. Opin. Chem. Biol., 1997, 1, 162.
30 F. H. Malpartida and D. A. Hopwood, Biotechnology, 1992, 24, 342.
31 J. Corts, S. F. Haydock, G. A. Roberts, D. J. Bevitt and P. F. Leadlay, Nature, 1990, 348, 176.
32 J. M. Weber and R. Losick, Gene, 1988, 68, 173.
33 R. Stanzak, P. Matsushima, R. H. Baltz and R. N. Rao, Biol. Technology, 1986, 4, 229.
34 J. S. Tuan, J. M. Weber, M. J. Staver, J. O. Leung, S. Donadio and L. Katz, Gene, 1990, 90, 21.
35 D. J. Bevitt, J. Corts, S. F. Haydock and P. F. Leadlay, Eur. J. Biochem., 1992, 204, 39.
36 S. Donadio, M. J. Staver, J. B. McAlpine, S. J. Swanson and L. Katz, Science, 1991, 252, 675.

60 von 66 20.11.2006 11:04


Published by the Royal Society of Chemistry http://pubs.rsc.org/ej/NP/2001/A909079G/

37 T. Schwecke, J. F. Aparicio, I. Molnr, A. Knig, L. E. Khaw, S. F. Haydock, M. Oliynyk, P.


Caffrey, J. Corts, J. B. Lester, G. A. Bhm, J. Staunton and P. F. Leadlay, Proc. Natl. Acad. Sci.
U.S.A., 1995, 92, 7839.
38 J. F. Aparicio, I. Molnr, T. Schwecke, A. Knig, S. F. Haydock, L. E. Khaw, J. Staunton and P. F.
Leadlay, Gene, 1996, 169, 9.
39 T. Schupp, C. Toupet, N. Engel and S. Goff, FEMS Microbiol. Lett., 1998, 159, 201.
40 T. Brautaset, O. N. Sekurova, H. Sletta, T. E. Ellingsen, A. R. Strom, S. Valla and S. B. Zotchev,
Chem. Biol., 2000, 7, 395.
41 J. Staunton and B. Wilkinson, Chem. Rev., 1997, 97, 2611.
42 S. Donadio, J. B. McAlpine, P. J. Sheldon, M. Jackson and L. Katz, Proc. Natl. Acad. Sci. U.S.A.,
1993, 90, 7119.
43 T. W. Yu, Y. Shen, Y. Doi-Katayama, L. Tang, C. Park, B. S. Moore, C. R. Hutchinson and H. G.
Floss, Proc. Natl. Acad. Sci. U.S.A., 1999, 96, 9051.
44 H. Motamedi, S.-J. Cai, A. Shafiee and K. Elliston, Eur. J. Biochem., 1997, 244, 74.
45 H. Motamedi and A. Shafiee, Eur. J. Biochem, 1998, 256, 528.
46 S. Kuhstoss, M. Huber, J. R. Turner, J. W. Paschal and N. Rao, Gene, 1996, 183, 231.
47 D. G. Swan, A. M. Rodriguez, C. Vilches, C. Mendez and J. A. Salas, Mol. Gen. Genet., 1994, 242,
358.
48 B. Shen, R. G. Summers, H. Gramajo, M. J. Bibb and C. R. Hutchinson, J. Bacteriol., 1992, 174,
3818.
49 D. J. MacNeil, J. L. Occi, K. M. Gewain, T. MacNeil, P. H. Gibbons, C. L. Ruby and S. J. Danis,
Ann. N.Y. Acad. Sci., 1994, 721, 123.
50 H. Ikeda, T. Nonomiya, M. Usami, T. Ohta and S. Omura, Proc. Natl. Acad. Sci. U.S.A., 1999, 96,
9509.
51 S. J. Kakavas, L. Katz and D. Stassi, J. Bacteriol., 1997, 179, 7515.
52 Y. Xue, L. Zhao, H.-W. Liu and D. H. Sherman, Proc. Natl. Acad. Sci. U.S.A., 1998, 95, 12 111.
53 J. F. Aparicio, A. J. Colina, E. Ceballos and J. F. Martn, J. Biol. Chem., 1999, 274, 10 133.
54 S. Zotchev, K. Haugan, O. Sekurova, H. Sletta, T. E. Ellingsen and S. Valla, Microbiology, 2000,
146, 611.
55 A. R. Gandecha, S. L. Large and E. Cundliffe, Gene, 1997, 184, 197.
56 X. Ruan, A. Pereda, D. L. Stassi, D. Zeidner, R. G. Summers, M. Jackson, A. Shivakumar, S.
Kakavas, M. J. Staver, S. Donadio and L. Katz, J. Bacteriol., 1997, 179, 6416.
57 D. V. Santi, M. A. Siani, B. Julien, D. Kupfer and B. Roe, Gene, 2000, 247, 97.
58 P. A. S. Lowden, G. A. Bhm, J. Staunton and P. F. Leadlay, Angew. Chem., Int. Ed. Engl., 1996,
35, 2249.
59 A. Knig, T. Schwecke, I. Molnr, G. A. Bhm, P. A. S. Lowden, J. Staunton and P. F. Leadlay,
Eur. J. Biochem., 1997, 247, 526.
60 H. V. Kleinkauf and H. Von Doren, Eur. J. Biochem, 1996, 236, 335.
61 L. Hendrickson, C. R. Davis, C. Roach, D. K. Nguyen, T. Aldrich, P. McAda and C. D. Reeves,
Chem. Biol., 1999, 6, 429.
62 J. Kennedy, K. Auclair, S. G. Kendrew, C. Park, J. C. Vederas and C. R. Hutchinson, Science, 1999,
284, 1368.
63 I. Fujii , in Comprehensive Natural Products Chemistry, ed. U. Sankawa, Elsevier, Oxford, 1999,
pp. 409441.
64 R. N. Moore, G. Bigam, J. K. Chan, A. M. Hogg, T. T. Nakashima and J. C. Vederas, J. Am. Chem.
Soc., 1985, 107, 3694.

61 von 66 20.11.2006 11:04


Published by the Royal Society of Chemistry http://pubs.rsc.org/ej/NP/2001/A909079G/

65 K. Wagschal, Y. Yoshizawa, D. J. Witter, Y. Q. Liu and J. C. Vederas, J. Chem. Soc., Perkin Trans.
1, 1996, 2357.
66 Y. Yoshizawa, D. J. Witter, Y. Q. Liu and J. C. Vederas, J. Am. Chem. Soc., 1994, 116, 2693.
67 D. J. Witter and J. C. Vederas, J. Org. Chem., 1996, 61, 2613.
68 K. Auclair, A. Sutherland, J. Kennedy, D. J. Witter, J. P. Van den Heever, C. R. Hutchinson and J.
C. Vederas, J. Am. Chem. Soc., 2000, 122, 11 519.
69 C. A. Townsend, Chem. Biol., 1997, 4, 721.
70 J. M. Jaz, J. L. Farrar, M. Bowman, R. A. Dixon and J. P. Noel, FASEB J., 1999, 13, A1392.
71 M. He, M. Varoglu and D. H. Sherman, J. Bacteriol., 2000, 182, 2619.
72 J. L. Ferrer, J. M. Jez, M. E. Bowman, R. A. Dixon and J. P. Noel, Nat. Struct. Biol., 1999, 6, 775.
73 J. M. Jez, J. L. Ferrer, M. E. Bowman, R. A. Dixon and J. P. Noel, Biochemistry, 2000, 39, 890.
74 W. Huang, J. Jia, P. Edwards, K. Dehesh, G. Schneider and Y. Lindqvist, EMBO J., 1998, 17, 1183.
75 A. F. A. Marsden, P. Caffrey, J. F. Aparicio, M. S. Loughran, J. Staunton and P. F. Leadlay, Science,
1994, 263, 378.
76 J. F. Aparicio, P. Caffrey, A. F. A. Marsden, J. Staunton and P. F. Leadlay, J. Biol. Chem., 1994,
269, 8524.
77 J. Staunton, P. Caffrey, J. F. Aparicio, G. A. Roberts, S. S. Bethell and P. F. Leadlay, Nat. Struct.
Biol., 1996, 3, 188.
78 J. K. Stoops and S. J. Wakil, J. Biol. Chem., 1981, 256, 5128.
79 J. K. Stoops, S. J. Wakil, E. C. Uberbacher and G. J. Bunick, J. Biol. Chem., 1987, 262, 10 246.
80 A. K. Joshi, A. Witkowski and S. Smith, Biochemistry, 1997, 36, 2316.
81 A. K. Joshi, A. Witkowski and S. Smith, Biochemistry, 1998, 37, 2515.
82 P. Caffrey, D. J. Bevitt, J. Staunton and P. F. Leadlay, FEBS Lett., 1992, 304, 225.
83 G. A. Roberts, J. Staunton and P. F. Leadlay, Biochem. Soc. Trans., 1992, 21.
84 G. A. Roberts, J. Staunton and P. F. Leadlay, FEBS Lett., 1993, 214, 306.
85 S. J. Wakil, Biochemistry, 1989, 28, 4523.
86 D. A. Hopwood and D. H. Sherman, Annu. Rev. Genet., 1990, 24, 67.
87 C. M. Kao, R. Pieper, D. E. Cane and C. Khosla, Biochemistry, 1996, 35, 12 363.
88 R. S. Gokhale, J. Lau, D. E. Cane and C. Khosla, Biochemistry, 1998, 37, 2524.
89 D. E. Cane, C. T. Walsh and C. Khosla, Science, 1998, 282, 63.
90 S. Smith, FASEB J., 1994, 8, 1248.
91 J. K. Stoops and S. J. Wakil, Fed. Proc., 1987, 46, 2216.
92 A. Ranganathan, M. Timoney, M. Bycroft, J. Corts, I. P. Thomas, B. Wilkinson, L. Kellenberger,
U. Hanefeld, I. S. Galloway, J. Staunton and P. F. Leadlay, Chem. Biol., 1999, 6, 731.
93 R. S. Gokhale, S. Y. Tsuji, D. E. Cane and C. Khosla, Science, 1999, 284, 482.
94 Curr. Opin. Chem. Biol., 2000, 4, 243.
95 R. R. Breaker, Curr. Opin. Chem. Biol., 1997, 1, 26.
96 T. Pan, Curr. Opin. Chem. Biol., 1997, 1, 17.
97 V. J. Hruby, J. M. Ahn and S. Liao, Curr. Opin. Chem. Biol., 1997, 1, 114.
98 Chem. Rev., 1997, 97, 2463.
99 C. Khosla, Chem. Rev., 1997, 97, 2577.
100 R. Lal, R. Kumari, H. Kaur, R. Khanna, N. Dhingra and D. Tuteja, Trends Biotechnol., 2000, 18,
264.
101 C. W. Carreras and D. V. Santi, Curr. Opin. Biotechnol., 1998, 9, 403.
102 C. R. Hutchinson, Curr. Opin. Microbiol., 1998, 1, 319.

62 von 66 20.11.2006 11:04


Published by the Royal Society of Chemistry http://pubs.rsc.org/ej/NP/2001/A909079G/

103 R. McDaniel, S. Ebert-Khosla, D. A. Hopwood and C. Khosla, Science, 1993, 262, 1546.
104 D. J. Lydiate, F. Malpartida and D. A. Hopwood, Gene, 1985, 35, 223.
105 M. A. Fernandez-Moreno, E. Martinez, J. L. Caballero, K. Ichinose, D. A. Hopwood and F.
Malpartida, J. Biol. Chem., 1994, 269, 24 854.
106 R. McDaniel, S. Ebert-Khosla, H. Fu, D. A. Hopwood and C. Khosla, Proc. Natl. Acad. Sci. U.S.A.,
1994, 91, 11542.
107 P. J. Kramer, R. J. X. Zawada, R. McDaniel, C. R. Hutchinson, D. A. Hopwood and C. Khosla, J.
Am. Chem. Soc., 1997, 119, 635.
108 R. McDaniel, S. Ebert-Khosla, D. A. Hopwood and C. Khosla, J. Am. Chem. Soc., 1994, 116, 10
855.
109 H. Fu, D. A. Hopwood and C. Khosla, Chem. Biol., 1994, 1, 205.
110 H. Fu, S. Ebert-Khosla, D. A. Hopwood and C. Khosla, J. Am. Chem. Soc., 1994, 116, 4166.
111 H. Fu, R. McDaniel, D. A. Hopwood and C. Khosla, Biochemistry, 1994, 33, 9321.
112 R. McDaniel, C. R. Hutchinson and C. Khosla, J. Am. Chem. Soc., 1995, 117, 6805.
113 R. McDaniel, S. Ebert-Khosla, D. A. Hopwood and C. Khosla, J. Am. Chem. Soc., 1993, 115, 11
671.
114 C. M. Kao, L. Katz and C. Khosla, Science, 1994, 265, 509.
115 L. Tang, S. Shah, L. Chung, J. Carney, L. Katz, C. Khosla and B. Julien, Science, 2000, 287, 640.
116 D. J. Bedford, E. Schweizer, D. A. Hopwood and C. Khosla, J. Bacteriol., 1995, 177, 4544.
117 R. B. Ziermann and M. C. Betlach, Biotechniques, 1999, 26, 106.
118 Q. Xue, G. Ashley, C. R. Hutchinson and D. V. Santi, Proc. Natl. Acad. Sci. U.S.A., 1999, 96, 11
740.
119 R. B. Ziermann and M. C. Betlach, J. Ind. Microbiol. Biotechnol., 2000, 24, 46.
120 L. Tang, J. Fu and R. McDaniel, Chem. Biol., 2000, 7, 77.
121 L. Katz, Chem. Rev., 1997, 97, 2557.
122 C. J. Rowe, J. Corts, S. Gaisser, J. Staunton and P. F. Leadlay, Gene, 1998, 216, 215.
123 L. E. Quadri, P. H. Weinreb, M. Lei, M. M. Nakano, P. Zuber and C. T. Walsh, Biochemistry, 1998,
37, 1585.
124 N. Wu, F. Kudo, D. E. Cane and C. Khosla, J. Am. Chem. Soc., 2000, 122, 4847.
125 C. Khosla, J. Org. Chem., 2000, 65, 8127.
126 A. F. A. Marsden, B. Wilkinson, J. Corts, N. J. Dunster, J. Staunton and P. F. Leadlay, Science,
1998, 279, 199.
127 M. S. Pacey, J. P. Dirlam, R. W. Geldart, P. F. Leadlay, H. A. I. McArthur, E. L. McCormick, R. A.
Monday, T. N. O Connell, J. Staunton and T. J. Winchester, J. Antibiot., 1998, 51, 1029.
128 M. S. Brown, J. P. Dirlam, H. A. I. McArthur, E. L. McCormick, B. K. Morse, P. A. Murphy, T. N.
O Connell, M. Pacey, D. M. Rescek, J. Ruddock and R. G. Wax, J. Antibiot., 1999, 52, 742.
129 I. C. Parsons, J. R. Everett, M. S. Pacey, J. C. Ruddock, A. G. Swanson and C. M. Thompson, J.
Antibiot., 1999, 52, 190.
130 C. M. Kao, M. McPherson, R. N. McDaniel, H. Fu, D. E. Cane and C. Khosla, J. Am. Chem. Soc.,
1998, 120, 2478.
131 C. M. Kao, M. McPherson, R. N. McDaniel, H. Fu, D. E. Cane and C. Khosla, J. Am. Chem. Soc.,
1997, 119, 11 339.
132 R. McDaniel, C. M. Kao, H. Fu, P. Hevezi, C. Gustafsson, M. Betlach, G. Ashley, D. E. Cane and C.
Khosla, J. Am. Chem. Soc., 1997, 119, 4309.
133 R. McDaniel, C. M. Kao, S. J. Hwang and C. Khosla, Chem. Biol., 1997, 4, 667.
134 M. Oliynyk, M. J. B. Brown, J. Corts, J. Staunton and P. F. Leadlay, Chem. Biol., 1996, 3, 833.

63 von 66 20.11.2006 11:04


Published by the Royal Society of Chemistry http://pubs.rsc.org/ej/NP/2001/A909079G/

135 J. Lau, H. Fu, D. E. Cane and C. Khosla, Biochemistry, 1999, 38, 1643.
136 L. Liu, A. Thamchaipenet, H. Fu, M. Betlach and G. Ashley, J. Am. Chem. Soc., 1997, 119, 10 553.
137 D. L. Stassi, S. J. Kakavas, K. A. Reynolds, G. Gunawardana, S. Swanson, D. Zeidner, M. Jackson,
H. Liu, A. Buko and L. Katz, Proc. Natl. Acad. Sci. U.S.A., 1998, 95, 7305.
138 J. Corts, K. E. H. Wiesmann, G. A. Roberts, M. J. B. Brown, J. Staunton and P. F. Leadlay,
Science, 1995, 268, 1487.
139 C. M. Kao, G. Luo, L. Katz, D. E. Cane and C. Khosla, J. Am. Chem. Soc., 1995, 117, 9105.
140 I. Bhm, I. E. Holzbaur, U. Hanefeld, J. Corts, J. Staunton and P. F. Leadlay, Chem. Biol., 1998, 5,
407.
141 R. McDaniel, A. Thamchaipenet, C. Gustafsson, H. Fu, M. Betlach, M. Betlach and G. Ashley,
Proc. Natl. Acad. Sci. U.S.A., 1999, 96, 1846.
142 R. Pieper, G. Luo, D. E. Cane and C. Khosla, J. Am. Chem. Soc., 1995, 117, 11 373.
143 J.-A. Chuck, M. McPherson, H. Huang, J. R. Jacobsen, C. Khosla and D. E. Cane, Chem. Biol.,
1997, 4, 757.
144 K. J. Weissman, M. Bycroft, A. L. Cutter, U. Hanefeld, E. J. Frost, M. C. Timoney, R. Harris, S.
Handa, M. Roddis, J. Staunton and P. F. Leadlay, Chem. Biol., 1998, 5, 743.
145 C. J. Dutton, A. M. Hooper, P. F. Leadlay and J. Staunton, Tetrahedron Lett., 1994, 35, 327.
146 J. R. Jacobsen, A. T. Keatinge-Clay, D. E. Cane and C. Khosla, Bioorg. Med. Chem., 1998, 6, 1171.
147 D. Hunziker, N. Wu, D. E. Cane and C. Khosla, Tetrahedron Lett., 1999, 40, 635.
148 S. Gaisser and P. F. Leadlay, Nat. Biotechnol., 1998, 16, 19.
149 S. Gaisser, G. A. Bohm, M. Doumith, M. C. Raynal, N. Dhillon, J. Corts and P. F. Leadlay, Mol.
Gen. Genet., 1998, 258, 78.
150 S. Gaisser, G. A. Bohm, J. Corts and P. F. Leadlay, Mol. Gen. Genet., 1997, 256, 239.
151 S. Gaisser, J. Reather, G. Wirtz, L. Kellenberger, J. Staunton and P. F. Leadlay, Mol. Microbiol.,
2000, 36, 391.
152 R. J. Summers, S. Donadio, M. J. Staver, E. Wendt-Pienkowski, C. R. Hutchinson and L. Katz,
Microbiol. UK, 1997, 143, 3251.
153 K. Salah-Bey, M. Doumith, J. M. Michel, S. Haydock, J. Corts, P. F. Leadlay and M. C. Raynal,
Mol. Gen. Genet., 1998, 257, 542.
154 L. Zhao, D. H. Sherman and H.-W. Liu, J. Am. Chem. Soc., 1998, 120, 10 256.
155 L. Zhao, D. H. Sherman and H.-W. Liu, J. Am. Chem. Soc., 1998, 120, 9374.
156 J. M. Weber, J. O. Leung, S. J. Swanson, K. B. Idler and J. B. McAlpine, Science, 1991, 252, 114.
157 D. Stassi, S. Donadio, M. J. Staver and L. Katz, J. Bacteriol., 1993, 175, 182.
158 L. Zhao, J. Ahlert, Y. Xue, J. S. Thorson, D. H. Sherman and H.-W. Liu, J. Am. Chem. Soc., 1999,
121, 9881.
159 K. Madduri, J. Kennedy, G. Rivola, A. Inventi-Solari, S. Filippini, G. Zanuso, A. L. Colombo, K. M.
Gewain, J. L. Occi, D. J. MacNeil and C. R. Hutchinson, Nat. Biotechnol., 1998, 16, 69.
160 R. Pieper, G. Luo, D. E. Cane and C. Khosla, Nature, 1995, 378, 263.
161 R. Pieper, S. Ebert-Khosla, D. E. Cane and C. Khosla, Biochemistry, 1996, 35, 2054.
162 R. Pieper, R. S. Gokhale, G. Luo, D. E. Cane and C. Khosla, Biochemistry, 1997, 36, 1846.
163 M. Bycroft, K. J. Weissman, J. Staunton and P. F. Leadlay, Eur. J. Biochem., 2000, 267, 520.
164 K. E. H. Wiesmann, J. Corts, M. J. B. Brown, A. L. Cutter, J. Staunton and P. F. Leadlay, Chem.
Biol., 1995, 2, 583.
165 J. Lau, D. E. Cane and C. Khosla, Biochemistry, 2000, 39, 10 514.
166 W. D. Celmer, J. Am. Chem. Soc., 1965, 87, 1799.
167 W. D. Celmer, J. Am. Chem. Soc., 1965, 87, 1801.

64 von 66 20.11.2006 11:04


Published by the Royal Society of Chemistry http://pubs.rsc.org/ej/NP/2001/A909079G/

168 D. E. Cane, W. D. Celmer and J. W. Westley, J. Am. Chem. Soc., 1983, 105, 3594.
169 D. O Hagan, Tetrahedron, 1988, 44, 1691.
170 K. J. Weissman, M. Timoney, M. Bycroft, P. Grice, U. Hanefeld, J. Staunton and P. F. Leadlay,
Biochemistry, 1997, 36, 13 849.
171 B. Sedgewick, J. W. Cornforth, S. J. French, R. T. Gray, E. Kelstrup and P. Willadsen, Eur. J.
Biochem., 1977, 75, 481.
172 I. E. Holzbaur, R. C. Harris, M. Bycroft, J. Corts, C. Bisang, J. Staunton, B. A. M. Rudd and P. F.
Leadlay, Chem. Biol., 1999, 6, 189.
173 C. M. Kao, G. Luo, L. Katz, D. E. Cane and C. Khosla, J. Am. Chem. Soc., 1996, 118, 9184.
174 R. Aggarwal, P. Caffrey, P. F. Leadlay, C. J. Smith and J. Staunton, J. Chem. Soc., Chem. Commun.,
1995, 1519.
175 P. Caffrey, B. Green, L. C. Packman, B. J. Rawlings, J. Staunton and P. F. Leadlay, Eur. J.
Biochem., 1991, 195, 823.
176 R. S. Gokhale, D. Hunziker, D. E. Cane and C. Khosla, Chem. Biol., 1999, 6, 117.
177 K. J. Weissman, C. J. Smith, U. Hanefeld, R. Aggarwal, M. Bycroft, J. Staunton and P. F. Leadlay,
Angew. Chem., Int. Ed., 1998, 37, 1437.
178 M. L. Heathcote , PhD Thesis, University of Cambridge, Cambridge, UK, 2000.
179 C. J. Smith , PhD Thesis, University of Cambridge, Cambridge, UK, 1995.
180 D. A. Hopwood, Chem. Rev., 1997, 97, 2465.
181 P. L. Bartel, C. B. Zhu, J. S. Lampel, D. C. Dosch, N. C. Connors, W. R. Strohl, J. M. Beale and H.
G. Floss, J. Bacteriol., 1990, 172, 4816.
182 M. J. Bibb, D. H. Sherman, S. Omura and D. A. Hopwood, Gene, 1994, 142, 31.
183 T. W. Yu, M. J. Bibb, W. P. Revill and D. A. Hopwood, J. Bacteriol., 1994, 176, 2627.
184 K. Ichinose, D. J. Bedford, D. Tornus, A. Bechthold, M. J. Bibb, W. P. Revill, H. G. Floss and D. A.
Hopwood, Chem. Biol., 1998, 5, 647.
185 A. Bechthold, J. K. Sohng, T. M. Smith, X. Chu and H. G. Floss, Mol. Gen. Genet., 1995, 248, 610.
186 S. L. Otten, X. Liu, J. Ferguson and C. R. Hutchinson, J. Bacteriol., 1995, 177, 6688.
187 A. Grimm, K. Madduri, A. Ali and C. R. Hutchinson, Gene, 1994, 151, 1.
188 N. Tsukamoto, I. Fujii, Y. Ebizuka and U. Sankawa, J. Antibiot., 1992, 45, 1286.
189 K. Ylihonko, J. Tuikkanen, S. Jussila, L. Cong and P. Mantsala, Mol. Gen. Genet., 1996, 251, 113.
190 L. Han, K. Yang, E. Ramalingam, R. H. Mosher and L. C. Vining, Microbiology, 1994, 140, 3379.
191 E. S. Kim, M. J. Bibb, M. J. Butler, D. A. Hopwood and D. H. Sherman, Gene, 1994, 141, 141.
192 H. C. Gramajo, J. White, C. R. Hutchinson and M. J. Bibb, J. Bacteriol., 1991, 173, 6475.
193 F. Lombo, G. Blanco, E. Fernandez, C. Mendez and J. A. Salas, Gene, 1996, 172, 87.
194 R. McDaniel, S. Ebert-Khosla, D. A. Hopwood and C. Khosla, Nature, 1995, 375, 549.
195 M. A. Alvarez, H. Fu, C. Khosla, D. A. Hopwood and J. E. Bailey, Nat. Biotechnol., 1996, 14, 335.
196 B. Shen, R. G. Summers, E. Wendt-Pienkowski and C. R. Hutchinson, J. Am. Chem. Soc., 1995,
117, 6811.
197 G. Geurer, M. Gerlitz, E. Wendt-Pienkowski, L. C. Vining, J. Rohr and C. R. Hutchinson, Chem.
Biol., 1997, 4, 433.
198 B. Shen and C. R. Hutchinson, Proc. Natl. Acad. Sci. U.S.A., 1996, 93, 6600.
199 C. R. Hutchinson, Chem. Rev., 1997, 97, 2525.
200 I. Fujii and Y. Ebizuka, Chem. Rev., 1997, 97, 2511.
201 B. J. Rawlings, Nat. Prod. Rep., 1999, 16, 425.
202 B. Shen, Top. Curr. Chem., 2000, 209, 1.

65 von 66 20.11.2006 11:04


Published by the Royal Society of Chemistry http://pubs.rsc.org/ej/NP/2001/A909079G/

203 H. Petkovic, A. Thamchaipenet, L. H. Zhou, D. Hranueli, P. Raspor, P. G. Waterman and I. S.


Hunter, J. Biol. Chem., 1999, 274, 32 829.
204 A.-L. Matharu, R. J. Cox, J. Crosby, K. J. Byrom and T. J. Simpson, Chem. Biol., 1998, 5, 699.
205 J. Crosby, K. J. Byrom, T. S. Hitchman, R. J. Cox, M. P. Crump, I. S. C. Findlow, M. J. Bibb and T.
J. Simpson, FEBS Lett., 1998, 433, 132.
206 R. J. Cox, T. S. Hitchman, K. J. Byrom, I. S. C. Findlow, J. A. Tanner, J. Crosby and T. J. Simpson,
FEBS Lett., 1997, 405, 267.
207 T. A. Holak, M. Nilges, J. H. Prestegard, A. M. Gronenborn and G. M. Clore, Eur. J. Biochem.,
1988, 175, 9.
208 M. Andrec, R. B. Hill and J. H. Prestegard, Protein Sci., 1995, 4, 983.
209 Y. P. Kim and J. H. Prestagard, Biochemistry, 1989, 28, 8792.
210 Y. P. Kim and J. H. Prestagard, Proteins, 1990, 8, 377.
211 T. S. Hitchman, J. Crosby, K. J. Byrom, R. J. Cox and T. J. Simpson, Chem. Biol., 1998, 5, 35.
212 P. Zhou, G. Florova and K. A. Reynolds, Chem. Biol., 1999, 6, 577.
213 W. P. Revill, M. J. Bibb and D. A. Hopwood, J. Bacteriol., 1995, 177, 3946.
214 R. G. Summers, A. Ali, B. Shen, W. A. Wessel and C. R. Hutchinson, Biochemistry, 1995, 34, 9389.
215 C. Bisang, P. F. Long, J. Corts, J. Westcott, J. Crosby, A.-L. Matharu, R. J. Cox, T. J. Simpson, J.
Staunton and P. F. Leadlay, Nature, 1999, 401, 502.
216 A. Witkowski, A. K. Joshi, Y. Lindqvist and S. Smith, Biochemistry, 1999, 38, 11 643.
217 J. Schrder, Nat. Struct. Biol., 1999, 6, 714.
218 J. Schrder , in Comprehensive Natural Products, ed. U. Sankawa, Elsevier, Oxford, 1999,
749771.
219 M. S. Jang, E. N. Cai, G. O. Udeani, K. V. Slowing, C. F. Thomas, C. W. W. Beecher, H. H. S.
Fong, N. R. Fransworth, A. D. Kinghorn, R. G. Mehta, R. C. Moon and J. M. Pezzuto, Science,
1997, 275, 218.
220 M. L. Edwards, D. M. Stemerick and P. S. Sunkara, J. Med. Chem., 1990, 33, 1948.
221 B. D. Gehm, J. M. McAndrews, P.-Y. Chien and J. L. Jameson, Proc. Natl. Acad. Sci. U.S.A., 1997,
94, 14 138.
222 R. S. Li, G. L. Kenyon, F. E. Cohen, X. W. Chen, B. Q. Gong, J. N. Dominquez, E. Davidson, G.
Kurzban, R. E. Miller, E. O. Nuzum, P. J. Rosenthal and J. H. McKerrow, J. Med. Chem., 1995, 38,
5031.
223 M. E. Zwaagstra, H. Timmerman, M. Tamura, T. Tohma, Y. Wada, K. Onogi and M. Q. Zhang, J.
Med. Chem., 1997, 40, 1075.
224 J.-L., Ferrer, J. M. Jez, M. E. Bowman, R. A. Dixon and J. P. Noel, Nat. Struct. Biol., 1999, 6, 775.
225 M. Mathieu, J. P. Zeelen, R. A. Pauptit, R. Erdmann, W. H. Kunau and R. K. Wierenga, Structure,
1994, 2, 797.
226 M. Mathieu, Y. Modis, J. P. Zeelen, C. K. Engel, R. A. Abagyan, A. Ahlberg, B. Rasmussen, V. S.
Lamzin, W. H. Kunau and R. K. Wierenga, J. Mol. Biol., 1997, 273, 714.
227 J. M. Jez, J.-L. Ferrer, M. E. Bowman, R. A. Dixon and J. P. Noel, Biochemistry, 2000, 39, 890.
228 J. M. Jez and J. P. Noel, J. Biol. Chem., 2000, 275, 39 640.
229 J. M. Jez, M. B. Austin, J.-L. Ferrer, M. E. Bowman, J. Schrder and J. P. Noel, Chem. Biol., 2000,
7, 919.
230 N. Funa, Y. Ohnishi, I. Fujii, M. Shibuya, Y. Ebizuka and S. Horinouchi, Nature, 1999, 400, 897.
231 A. D. Buss , personal communication, 1997.

This journal is The Royal Society of Chemistry 2001

66 von 66 20.11.2006 11:04

Das könnte Ihnen auch gefallen