Type II Polyketide Synthases: Gaining a Deeper Insight into Enzymatic Teamwork

Article (PDF Available)inNatural Product Reports 24(1):162-90 · March 2007with676 Reads
DOI: 10.1039/b507395m · Source: PubMed
This review covers advances in understanding of the biosynthesis of polyketides produced by type II PKS systems at the genetic, biochemical and structural levels.
REVIEW www.rsc.org/npr | Natural Product Reports
Type II polyketide synthases: gaining a deeper insight into enzymatic
Christian Hertweck,
Andriy Luzhetskyy,
Yuri Rebets
and Andreas Bechthold*
Received (in Cambridge, UK) 15th August 2006
First published as an Advance Article on the web 22nd November 2006
DOI: 10.1039/b507395m
Covering: 2002–2005
This review covers advances in understanding of the biosynthesis of polyketides produced by type II
PKS systems at the genetic, biochemical and structural levels.
1 Introduction
2 Biological activities of type II PKS-derived natural
3 Enzymes involved in the formation of aromatic
3.1 Production of aromatic polyketides by teamwork
between enzymes
3.2 Polyketide chain assembly by the minimal PKS
3.3 Chain length
3.4 PKS priming
3.5 Ketoreductases
3.6 Cyclases
3.7 Aromatases
4 Polyketide-tailoring reactions
4.1 Methyltransferases
4.2 Oxygenases
4.3 Glycosyltransferases
5 Regulation of aromatic polyketide biosynthesis
5.1 Genes that affect secondary metabolism and
morphological differentiation
5.2 Genes that affect secondary metabolism and growth
5.3 General signals affecting secondary metabolism
5.4 SARPs, pathway-specific regulators affecting
secondary metabolism
5.5 Neg ative regulators of secondary metabolism
5.6 Post-translational modifications of biosynthetic
6 Acknowledgements
7 References
1 Introduction
Natural products formed by type II polyketide synthases (PKSs)
comprise an important and structurally diverse group of bacterial
secondary metabolites.
Many of these compounds or their
semisynthetic derivatives have emerged as clinically useful drugs
or are promising drug candidates. The soil-borne and marine
Leibniz Institute for Natural Product Research and Infection Biology, HKI,
Beutenbergstr. 11a, 07745, Jena, Germany
at Freiburg, Institut f
ur Pharmazeutische Wissenschaften, Phar-
mazeutische Biologie und Biotechnologie, Stefan-Meier-Straße 19, 79104,
Freiburg, Germany. E-mail: andreas.bechthold@pharmazie.uni-freiburg.de
Gram-positive actinomycetes are a particularly rich source of
such bioactive polyphenols, and represent the only known group
of organisms that employ type II PKS systems for polyketide
biosynthesis. With the advent of molecular tools and recombinant
methods applicable to actinomycetes,
it has become feasible
to investigate bacterial aromatic polyketide biosynthesis at the
genetic and biochemical levels,
which has finally provided a
basis for engineering novel natural product derivatives.
enzymology of type II PKS systems has been the subject of
excellent general reviews by Rawlings,
and Reeves,
as well as specialized reviews on PKS starter units by Moore
and Hertweck,
polyketide tailoring reactions by Rohr and co-
and pathway engineering approaches by Mendez and
Nonetheless, the rapid developments in both cloning and
analyzing type II PKS gene clusters, and biochemical as well
as structure-biological investigations,
call for an update on the
recent developments in type II PKS research. After a short
introduction into the mode of action of the most valuable aromatic
polyketides, this review summarizes the recent literature (2002–
2005) pertaining to the type II PKS genes and enzymes that
catalyze and regulate the biosynthesis of aromatic polyketides.
2 Biological activities of type II PKS-derived natural
On the basis of the polyphenolic ring system and their biosynthetic
pathways, metabolites produced by type II PKSs are classi-
fied as anthracyclines, angucyclines, aureolic acids, tetracyclines,
tetracenomycins, pradimicin-type polyphenols, and benzoisochro-
manequinones. This brief overview introduces bioactive aromatic
polyketides that have been the subject of recent biosynthetic
studies at the molecular level.
Anthracyclines rank among the most effective anticancer drugs
ever developed.
The first anthracyclines were isolated in
the 1950s and were named rhodomycin and cinerubicin.
few years later daunomycin and doxorubicin were discovered.
The aglycone of anthracyclines consists of a tetracyclic ring system
with quinone–hydroquinone groups in rings C and B, a methoxy
substituent at C-4 in ring D and a short side chain at C-9, with a
carbonyl at C-13. A sugar, daunosamine, is attached at C-7 of ring
A (Fig. 1). The difference between doxorubicin and daunorubicin
is that the side chain of doxorubicin terminates with a primary
alcohol and daunorubicin terminates with a methyl. This minor
structural difference, however, has an important impact on the
162 | Nat. Prod. Rep., 2007, 24, 162–190 This journal is
The Royal Society of Chemistry 2007
Downloaded on 06 February 2013
Published on 22 November 2006 on http://pubs.rsc.org | doi:10.1039/B507395M
View Article Online
/ Journal Homepage
/ Table of Contents for this issue
Christian Hertweck (born 1969) studied chemistry, and obtained his Ph.D. on the subject of natural product synthesis under the supervision
of Professor Boland at the Kekul
e-Institute for Organic Chemistry and Biochemistry, Bonn, and at the Max-Planck-Institute for Chemical
Ecology, Jena. In 1999, he became a Humboldt (Feodor Lynen) postdoctoral fellow of Professors Floss and Moore at the University of
Washington, Seattle. From 2000 to 2005 he was head of a Junior Research Group at the Hans-Kn
oll-Institute, Jena. In 2006 he was appointed
as Full Professor for Natural Product Chemistry at the Friedrich-Schiller-Univers ity and Head of the Department of Biomolecular Chemistry
at the Leibniz-Institute for Natural Product Research and Infection Biology, Jena. His current research interests involve the elucidation and
manipulation of microbial biosynthetic pathways.
Andriy Luzhetskyy was born in Busovisko, Ukraine, in 1977. He studied genetics and biotechnology at Ivan Franko National University,
Lviv, and in 2003 obtained his Ph.D. on the subject of the genetics of streptomycetes under the supervision of Professor Fedorenko, Lviv. In
2001 and 2002 he was an INTAS fellow of Professor Bechthold at the Albert-Ludwigs Universit
at, Freiburg. In 2004 he started his work
as a postdoctoral researcher in the laboratory of Professor Bechthold. His major interests are the elucidation of biosynthetic pathways of
antibiotics, the generation of new hybrid natural products through combinatorial biosynthesis, and the development of new genetic tools for
the manipulation of actinomycetes.
Yuriy Rebets was born in Zhovkva, Ukraine, in 1978. He studied biology at Ivan Franko National University, Lviv. In 2001 he joined a
graduate programme under the supervision of Dr Fedorenko, and in 2006 he obtained his Ph.D. on the subject of molecular genetics. In 2003
and 2004 he was a research fellow of Professor Nakamura at Niigata University of Pharmacy and Applied Life Sciences, Nigata, Japan.
From 2002–2005 he was a DAAD fellow of Professor Bechthold at the Albert-Ludwigs Universit
at, Freiburg. In 2006 he started his work as a
postdoctoral researcher in the laboratory of Professor Nakam ura at the Niigata University of Pharmacy and Applied Life Sciences, Nigata,
Japan. His current research is devoted to control of antibiotic production at the level of pathway-specific regulatory genes, and investigation
of ion transporters in bacteria.
Andreas Bechthold (born 1962) studied pharmacy, and obtained his Ph.D. on the subject of natural product biosynthesis under the
supervision of Professors Heide and Leistner at Rheinische Friedrich-Wilhelms-Universit
at, Bonn. In 1992, he became a DFG (Deutsche
Forschungsgemeinschaft) postdoctoral fellow of Professor Floss at the University of Washington, Seattle. In 1993, he became a JSPS
(Japanese Society for Pharmaceutical Sciences) postdoctoral fellow of Professor Tabata at Kyoto University. From 1994 to 1999 he
was head of a Junior Research Group at Eberhard-Karls-Universit
at, T
ubingen. In 2000 he was appointed as Associate Professor for
Pharmaceutical Biology at the Christian-Albrechts-Universit
at, Kiel. In 2001 he was appointed as Full Professor for Pharmaceutical Biology
and Biotechnology at the Albert-Ludwigs-Universit
at, Freiburg. His current research interests involve the elucidation and manipulation of
microbial biosynthetic pathways.
Christian Hertweck Andriy Luzhetskyy Yuri Rebets Andreas Bechthold
spectrum of activity. Doxorubicin is an essential component
in treatments for breast cancer, childhood solid tumours, soft
tissue sarcomas, and aggressive lymphomas, while daunorubicin
shows activity in acute lymphoblastic or myeloblastic leukaemias.
Unfortunately, the use of these drugs may cause serious problems,
such as cardiomyopathy and congestive heart failure. In addition,
drug-resistant tumour cells survive treatment with both anticancer
Attempts to generate or find novel anthracyclines with
improved activity or reduced toxicity has resulted only in a few
valuable compounds. Among these, epirubicin and idarubicin
Epirubicin differs from
doxorubicin in the stereochemistry at position C-4 of the sugar.
It shows different pharmacokinetic parameters from doxorubicin,
resulting in an enhanced total body clearance. Idarubicin does
not contain the 4-methoxy group that is present in ring D of
daunorubicin, and shows a broader spectrum of activity. Piraru-
bicin, aclacinomycin A (aclarubicin) and nogalamycin (Fig. 1) are
further anthracyclines that have received clinical approval.
It was indicated in early reports that pirarubicin is less
cardiotoxic than doxorubicin, but disadvantages concerning the
activity and toxicity of this drug were described later.
The activity
of anthracyclines seems to be based on different mechanisms
of action. Intercalation into DNA, generation of free radicals,
alkylation, DNA cross-linking, interference with helicase activity,
inhibition of topoisomerase II and induction of apoptosis have
been considered as being the most important mechanisms.
This journal is
The Royal Society of Chemistry 2007 Nat. Prod. Rep., 2007, 24, 162–190 | 163
Downloaded on 06 February 2013
Published on 22 November 2006 on http://pubs.rsc.org | doi:10.1039/B507395M
View Article Online
Fig. 1 Structures of selected anthracyclines.
The formation of an anthracycline–DNA complex relies on
defined structural determinants. The planar ring system is im-
portant for intercalation into DNA. Crystallographic studies of a
complex of daunomycin and duplex DNA also showed the amino
sugar daunosamine to be located in the minor groove of the DNA
duplex, with the protonated daunosamine potentially contributing
to the non-covalent binding of daunomycin. The amino and
hydroxyl groups of the sugar face out of the minor groove,
possibly contributing to the topoisomerase-inhibiting activity of
the drug.
Interestingly, DNA binding has also been shown for
the sugar components of aclacinomycin A and nogalamycin.
The aureolic acids represent another group of potent anticancer
drugs. Members of this family are chromomycin and mithramycin
(Fig. 2), olivomycin, chromocyclomycin, UCH9 and durhamycin
Mithramycin (which has been used clinically in several cancer
therapies and Paget’s disease), as well as the chromomycins,
contain a tricyclic chromophore with two aliphatic side chains at
C-3 and C-7. Two sugar side chains are attached to each aglycone
at positions C-2 and C-6. Both compounds are DNA-dependent
RNA polymerase inhibitors forming symmetrical dimers with
DNA; the sugars of the trisaccharide side chains are essential
for stabilizing the dimers. Interestingly, the acetoxy groups in two
of the sugars of chromomycin contribute distinctively to DNA
complex formation by providing an additional H-bond with the
2-amino groups of G bases and thus adding more specificity to the
DNA binding.
One of the most important classes of antibiotics formed by
type II PKSs are the tetracyclines. The first generation of tetracy-
clines, introduced in the 1950–1960s, includes the broad-spectrum
antibiotics tetracycline, oxytetracycline, chlorotetracycline and
demeclocycline (Fig. 3). The second generation of tetracyclines
includes doxycycline and minocycline (Fig. 3).
Tetracyclines are
amphoteric due to a dimethylamino group and an acidic phenolic
residue. Tetracyclines inhibit protein synthesis by interfering with
the binding of aminoacylated tRNA to the A-site of the 30S
subunit. In 2001 six different tetracycline-binding sites on the
30S subunit were identified.
Binding site I is located near to
the A-site, the docking place of the aminoacylated tRNA. The
five other binding sites are found at various locations in the 30S
subunit, which explains the sometimes contradictory biochemical
and functional data for tetracycline binding to the 30S subunits.
The intensive use of tetracyclines has caused bacterial resistance
against tetracyclines, resulting in severe drawbacks in their use.
Attempts to develop tetracycline analogues that overcome these
resistance mechanisms led to the discovery of the glycylcyclines.
The most potent glycylcycline is the 9-tert-butylglycylamido
derivative of minocycline, also known as tigecycline (Fig. 3). The
glycylcyclines exhibit antibacterial activities comparable to the
first-generation tetracyclines, but have a higher potency against
tetracycline-resistant pathogens with efflux and ribosomal protec-
tion mechanisms of resistance.
Despite the structural similarities
between the linear tetracyclic ring systems of the tetracyclines,
anthracyclines, and aureolic acid precursors, they differ in their
cyclization patterns and thus belong to separate classes of
aromatic polyketides. The tetracenomycins (tetracenomycin and
elloramycin) (Fig. 3), which have been isolated from Streptomyces
glaucescens and Streptomyces olivaceus, also form a separate group
of tetracyclic antibiotics, which exhibit moderate antibacterial
Landomycin A and urdamycin A belong to the angucycline type
of antibiotics (Fig. 4). The biological activities of the angucycline
group metabolites are not restricted to any particular type of
action. To date, five main areas of biological activities have
been described: cytostatic activities, enzyme inhibition activities,
inhibition of platelet aggregation, antibacterial and antiviral
One of the most important compounds in the angucy-
cline group of antibiotics is landomycin A. It possesses strong
antitumour activities, in particular against prostate cancer cell
lines. It inhibits thymidine uptake in murine smooth muscle cells
and cell cycle progression.
Its unusual activities depend on the
extended oligosaccharide chain, since landomycins with shorter
sugar chains, such as landomycin E, containing three sugars, have
much weaker antitumour activity.
Benzo[a]naphthacenes, such as the pradimicins and pluramycins
(Fig. 4), also feature an angular polyphenolic ring system, and may
be regarded as ‘extended angucyclines’. The pradimicins exhibit
broad-spectrum antifungal and antiviral activity.
It has been
164 | Nat. Prod. Rep., 2007, 24, 162–190 This journal is
The Royal Society of Chemistry 2007
Downloaded on 06 February 2013
Published on 22 November 2006 on http://pubs.rsc.org | doi:10.1039/B507395M
View Article Online
Fig. 2 Structures of aureolic acid antibiotics.
Fig. 3 Structures of selected tetracyclines, the glycylcycline tigecycline
and the tetracenomycins.
shown that the D-alanine side chain, as well as the sugar moiety,
are essential for the antifungal activity.
The structurally related
benastatins inhibit glutathione-S-transferase (GST) activities.
This finding is of particular interest because some isoenzymes of
GST play a crucial role in the detoxification of tumour cell lines
that have acquired antitumour drug resistance.
The pradimicin-related pluramycins altromycin B and
hedamycin (Fig. 4) are strong alkylating agents. They contain a
4H-anthra[1,2-b]pyrene structural motif. Both compounds exhibit
antitumour and antimicrobial activities. The mechanism of action
is by direct interaction with DNA through intercalation between
the bases of the nucleotide, and alkylation of a guanine residue
at 5
sites. NMR studies revealed that the dimethylated
amino-substituted sugar vancosamine, a component of both
compounds, interacts directly with pyrimidine units of the DNA.
Gilvocarcins belong to the benzo[d]naphtho[1,2-b]pyran-6-one
C-glycoside antibiotics.
The most widely used compound is
gilvocarcin V
(Fig. 4), which shows excellent antitumour activity
and remarkably low toxicity.
The jadomycins are a unique family
of angucycline-derived antibiotics because of their pentacyclic
8H-benz[b]oxazolo[3,2-f ]phenanthridine backbone. Jadomycin B
(Fig. 4) is the principal product of Streptomyces venezuelae
ISP5230 when this strain is fermented under stress, such as heat
shock, ethanol treatment, or phage infection.
The jadomycins
show biological activity against Gram-positive and Gram-negative
bacteria, and against yeast.
Interestingly, both gilvocarcin V and
jadomycin B are angucycline derivatives (see below).
The benzoisochromanequinone class of antibiotics
the antibiotics actinorhodin from Streptomyces coelicolor,med-
ermycin from Streptomyces sp. AM7161, frenolicin from Strep-
tomyces roseofulvus, griseusin A from Streptomyces griseus,and
the antineoplastic agent granaticin from Streptomyces olivaceus
(Fig. 5). The blue pigment actinorhodin is the best-studied type II
It should be highlighted once again
that the pioneering experiments by Hopwood and co-workers
on combining actinorhodin and medermycin biosynthesis genes
led to the production of the first type II PKS hybrid metabolite,
Various polycyclic polyketide metabolites do not fit into the
above-mentioned general classes. For example, the R1128 complex
This journal is
The Royal Society of Chemistry 2007 Nat. Prod. Rep., 2007, 24, 162–190 | 165
Downloaded on 06 February 2013
Published on 22 November 2006 on http://pubs.rsc.org | doi:10.1039/B507395M
View Article Online
Fig. 4 Structures of landomycin A, urdamycin A, altromycin, hedamycin, pradimicin, pluramycin, benastatin, gilvocarcin V and jadomycin B.
comprises several anthraquinones, which have been identified as
novel non-steroidal estrogen-receptor antagonists.
The potent antitumour polyketide glycosides chartreusin
(Fig. 6) and elsamicin feature a rare p entacyclic bislactone
aglycone, named chartarin, which results from an unusual rear-
rangement sequence.
Pharmacological studies revealed that
chartreusin and its derivatives exert their antitumour activities
through binding to DNA,
radical-mediated single strand breaks,
and inhibition of topoisomerase II.
Resistomycin (Fig. 6) is an
unusual naphthanthrene derivative isolated from Streptomyces
that exhibits a variety of pharmacologically
relevant properties, e.g. inhibition of HIV- 1 protease,
as well as
RNA and DNA polymerase, and activity against Gram-positive
bacteria and mycobacteria.
More recently, it has also been
166 | Nat. Prod. Rep., 2007, 24, 162–190 This journal is
The Royal Society of Chemistry 2007
Downloaded on 06 February 2013
Published on 22 November 2006 on http://pubs.rsc.org | doi:10.1039/B507395M
View Article Online
Fig. 5 Structures of selected benzoisochromanequinones.
implicated as a modulator of apoptosis, and may thus serve as
a valuable tool in cell biology.
The structurally fascinating b-rubromycin
and griseorhodin
as well as fredericamycin
(Fig. 6) are potent antitumoural
agents. According to their biosynthetic models, these unusual
polyphenols share early biosynthetic s teps with pradimicin-type
metabolites and then undergo intriguing oxidative rearrangement
The most unusual metabolite formed by type II PKS is the
bacteriostatic agent enterocin
(Fig. 6), which
has been isolated from terrestrial and marine Streptomyces species.
Interestingly, the only aromatic part of the heavily rearranged
molecule is the benzoyl-derived starter unit.
3 Enzymes involved in the formation of aromatic
3.1 Production of aromatic polyketides by teamwork between
In analogy to type II bacterial and plant fatty acid synthases
(FASs), type II PKSs are comprised of several individual
Sequence information from numerous type II PKS
gene clusters revealed a hallmark for the biosynthesis of aromatic
polyketides: in all cases, a minimal set of iteratively used enzymes,
each expressed from a distinct gene, is involved. In general, this so-
called ‘minimal PKS’ consists of two ketosynthase units (KS
) and an acyl carrier protein (ACP), which serves as an anchor
for the growing polyketide chain.
With a few exceptions, genes
encoding these three proteins are grouped together, and show a
typical KS
/ACP architecture. Additional P KS subunits,
including ketoreductases, cyclases and aromatases define the
folding pattern of the nascent poly-b-keto intermediate. Finally,
Fig. 6 Structures of chartreusin, resistomycin, b-rubromycin, grise-
orhodin A, fredericamycin A and enterocin.
the polyphenols are tailored by oxygenases, glycosyl and methyl
transferases. Typically, genes encoding these enzymes are clustered
together. Since 2000, over 14 publications on the identification,
cloning and sequencing of type II PKS gene clusters have appeared
in the literature, highlighting the great interest in these systems
(Table 1). In addition, with the advent of partial and whole-genome
sequencing programs, as-yet uncharacterized and/or silent type II
PKS gene clusters have been detected in actinomycetes.
3.2 Polyketide chain assembly by the minimal PKS
The minimal PKS catalyses the iterative decarboxylative conden-
sation of malonyl-CoA extender units with an acyl starter unit
(Fig. 7). Both subunits, KS
and KS
, show high sequence
similarities. However, in contrast to KS
, the KS
is lacking an active site cysteine, which is crucial for polyketide
This journal is
The Royal Society of Chemistry 2007 Nat. Prod. Rep., 2007, 24, 162–190 | 167
Downloaded on 06 February 2013
Published on 22 November 2006 on http://pubs.rsc.org | doi:10.1039/B507395M
View Article Online
Fig. 7 Basic mechanisms of aromatic polyketide biosynthesis.
Table 1 Type II PKS gene clusters analyzed since 2000.
Metabolite Abbrev. Strain/source Year Ref.
Enterocin Enc Streptomyces maritimus 2000 80
R1128 Zhu Streptomyces sp. R1128 2000 81
Nogalamyin Sno Streptomyces nogalater 2001 82
Rubromycin Rub Streptomyces collinus 2001 67
Griseorhodin Grh Streptomyces sp. JP95 2002 64
Medermycin Med Streptomyces sp. AM-7161 2003 83
Gilvocarcin Gil Streptomyces griseoflavus 2003 84
Hedamycin Hed Streptomyces griseoruber 2004 85
Chromomycin Cmm Streptomyces griseus 2004 30
Resistomycin Rem Streptomyces resistomycificus 2004 86
Cervimycin Cer Streptomyces tendae 2004 87
Chartreusin Cha Streptomyces chartreusis 2005 56
Oviedomycin Ovi Streptomyces antibioticus 2005 88
Fredericamycin Fdm Streptomyces griseus 2005 66
Cluster sequence only published in GenBank.
Only partial sequence
assembly. Before direct experimental evidence existed, it had been
generally accepted that KS
and KS
form a heterodimer in
analogy to the KS homodimers (FabF) from bacterial FAS. While
the ketosynthase (KS
) subunit obviously catalyzes Claisen-type
C–C bond formations from activated acyl and malonyl building
blocks, the role of the KS
subunit had not been obvious and
became the subject of intense research.
In fact, despite the
mutated active site, the KS
is not ‘catalytically silent’ (Fig. 8).
Leadlay and co-workers showed that in the KS
unit, the active
site cysteine was mutated into a glutamine (Q) residue, which
is, at least in the actinorhodin (act) and tetracenomycin (tcm)
PKSs, involved in loading malonyl-CoA and generating acetyl
KS from decarboxylation of malonyl-ACP.
Interestingly, an
analogous condensation-deficient enzyme (KS
) often functions
in modular PKSs (modular PKSs, named type I PKSs, are large
multifunctional proteins that build macrocyclic polyketides) to
provide an acetate starter unit by decarboxylation of malonyl-
CoA. Another very important function of the KS
subunit is that
it is the primary determinant of carbon chain length, and it has
thus also been termed ‘chain length factor’ (CLF, see below).
addition to chain assembly, the min PKS can partially control the
regiochemistry of the first cyclization. Nonetheless, products from
aberrant cyclizations can be observed, e.g. in the act pathway.
Interestingly, there is no report on an aromatic PKS utilizing an
alternative elongation unit than malonyl-CoA, which is in marked
contrast to modular polyketide synthases.
There has been
an ongoing debate regarding the transfer of the malonyl exten-
der units in type II PKS and the involvement of a potential fourth
Fig. 8 Model for a minimal PKS complex.
component of a minimal PKS, a malonyl-CoA:ACP transferase
(MCAT). Since genes encoding MCATs are absent in most type
II PKS gene clusters, it was proposed that an endogenous MCAT
is recruited from fatty acid biosynthesis.
An alternative model
was proposed by Simpson and co-workers, who demonstrated by
in vitro studies that the ACP is capable of self-malonylation, albeit
only at elevated concentrations of malonyl-CoA. Recently, in vitro
self-malonylation of a chemically synthesized ACP was observed,
clearly ruling out any contamination of purified ACP with traces
of MCAT.
In vivo, however, at physiologically relevant concen-
trations, MCAT-dependent malonyl-ACP formation is still a likely
In this context the catalysis, specificity, and
ACP docking site of Streptomyces coelicolor malonyl-CoA:ACP
transacylase has been investigated.
Polyketide assembly by type II PKSs is notoriously difficult
to study, since it is carried out by a multienzyme complex and
not by a single enzyme. Furthermore, the proposed poly-b-keto
intermediates are highly unstable and prone to spontaneous
cyclization. Consequently, the biosynthetic machinery has long
been regarded as a ‘black box’. The recent structural study by
Stroud, Khosla and co-workers, in which the authors were able
to ‘catch a type II PKS in action’, was a breakthrough.
provided direct proof for a KS heterodimer in showing that KS
and CLF (or KS
) have evolved highly complementary contacts. In
addition, they deduced from the structure that a protein cleft keeps
the nascent polyketide chain extended. In this fashion the reactive
ketide groups are separated and probably contact the KS-CLF
mainly as enols.
This study complements well the previously
available structures of type II PKS components; primarily an X-
ray structure of a priming KS (KAS III homologue) from the
R1128 pathway,
and NMR solution structures of ACPs from
the actinorhodin,
and frenolicin
168 | Nat. Prod. Rep., 2007, 24, 162–190 This journal is
The Royal Society of Chemistry 2007
Downloaded on 06 February 2013
Published on 22 November 2006 on http://pubs.rsc.org | doi:10.1039/B507395M
View Article Online
3.3 Chain length
The chain length of polyketides synthesized by type II PKS is
usually either 16 (octaketides, such as actinorhodin), 20 (decake-
tides, such as tetracenomycin), or 24 (dodecaketides, such as
pradimicin). In a few cases, for example in the biosynthesis of
griseorhodin (a tridecaketide), benastatin (a tetradecaketide) and
fredericamycin (a pentadecaketide), even longer chains are formed
(Fig. 9).
In vivo experiments with various PKS hybrids suggested that
the KS
controls the number of Claisen condensations during
polyketide assembly, and was thus termed chain length factor
Furthermore, hybrid KS and CLF genes designed from
sequences of the act (octaketide) and tcm (decaketide) genes
reinforced the importance of CLF in defining chain length.
How is chain length determined? In principle, the size of
the polyketide could be controlled by counting the number of
condensation reactions, as has been suggested for 6-methylsalicylic
acid (6-MSAS)
or by measuring the length of the chain, as
in fatty acid
or plant polyketide biosynthesis.
Cox and co-
workers have addressed this issue by in vitro directed polyketide
biosynthesis using a set of acylated C17S actinorhodin ACP
surrogates and a mutated KS
system, which is not capable
of synthesizing the endogenous acetyl ACP starter. It was found
that the number of elongations is dependent of the length of
the starter unit, and consequently, the act min PKS governs
chain length by measurement. Analogous observations with the
cervimycin (cer)PKS
strongly support the model of a protein
cavity, which controls the size of the growing chain (C. Hertweck,
unpublished observations).
Results from homology modelling based on crystal structures
of the KS homodimer from E. coli FA S
suggested that regions
defining chain length are located at the interface of the dimer.
Khosla and co-workers found that residues 109, 112 and 116 in
the act CLF serve as gatekeepers in the polyketide tunnel, and
that reducing the size of these residues lengthens the channel,
allowing two more elongation cycles.
Stroud, Khosla and co-
workers extended their model of this ‘gating mechanism’ in the
context of KS/CLF structural studies, including Trp194, Leu143
and Phe140.
They proposed that novel chain lengths could be
engineered by opening or closing the gates.
Although the KS
(CLF) represents the primary determinant of
polyketide chain length, it may not be the only crucial regulatory
element. Other type II PKS components, in particular cyclases, can
also influence the number of chain elongations, as demonstrated
by Shen et al. in a remarkable study on whiE spore pigment
(see below). Almost simultaneously, Hunter and
co-workers reported that disruption of an aromatase/cyclase from
the oxytetracycline (otc) gene cluster of Streptomyces rimosus
results in the production of novel polyketides with shorter chain
Consequently, it seems that the chain length is at least to a
limited degree determined not only by the CLF, but also by the
entire PKS complex. Interestingly, in the context of engineering
hybrid systems, a fundamental incompatibility between antibiotic
and spore pigment biosynthesis has been observed.
Also it
should be noted that to date no factor mediating the release of
the fully elongated polyketide chain has been identified.
3.4 PKS priming
In most cases aromatic polyketide synthases are formally primed
with acetate. At least in actinorhodin and tetracenomycin biosyn-
thesis, it has been implicated that the PKS is primed by decar-
boxylation of a malonyl unit to yield an acetyl-S-KS intermediate
(Fig. 10). Leadlay and co-workers demonstrated that the KS
(CLF) has decarboxylase activity towards malonyl-CoA, in close
analogy to the KS
of modular PKS (see above).
Stroud et al.
argued against this model on the basis of the solved act KS/CLF
protein structure and suggested that CLF is a function of the KS
and thus this issue is still a matter of debate. It is also
relevant to point out that in the case of decarboxylation there is
Fig. 9 Examples of varying polyketide chain lengths.
This journal is
The Royal Society of Chemistry 2007 Nat. Prod. Rep., 2007, 24, 162–190 | 169
Downloaded on 06 February 2013
Published on 22 November 2006 on http://pubs.rsc.org | doi:10.1039/B507395M
View Article Online
Fig. 10 Mechanism of type II PKS priming. a) Priming involving decarboxylation. b) Priming involving a KSIII/ACP module. c) Priming involving the
direct loading of an acyl-CoA starter.
the requirement for transfer of the resulting acetate unit to the
active site of the KS domain.
Various PKS employ alternative primers, such as propionate,
(iso)butyrate, malonamate and benzoate. The biosynthesis and
attachment of such alternative type II PKS primers has been ex-
tensively reviewed by Moore and Hertweck.
In short, in addition
to decarboxylation of malonates, two alternative pathways have
been proposed for non-acetate PKS primers.
One alternative involves an additional KSIII (and ACP) for the
biosynthesis and attachment of short-chain fatty acids (Fig. 10).
The latter scenario was first described in the context of an-
thracycline biosynthesis. It should be noted that in these cases,
decarboxylative chain initiation by malonyl-ACP should even be
suppressed in favour of unusual starter units.
Sequence analyses of the daunorubicin (dnr) and doxorubicin
(dox) gene clusters by the Hutchinson
and Strohl
revealed the presence of two additional PKS components, which
were implicated in starter unit selection and chain initiation. These
additional activities are represented by a FabH-like KS
and an
AT, encoded by dpsC and dpsD, respectively. As in bacterial fatty
acid biosynthesis,
it has been suggested that DpsC functions
as a KSIII, which catalyzes the first condensation of malonyl-
CoA with propionyl-CoA to form a b-ketoester. Subsequently,
the diketide would be loaded onto the min PKS by means of the
AT DpsD.
Independent in vivo studies by the Hutchinson and
Strohl groups on the function of DpsC and DpsD showed that in
the absence of the priming KSIII component DpsC, the dps PKS
exhibits a relaxed starter unit specificity for acetyl-CoA instead
of propionyl-CoA. However, propionate-derived polyketides may
also be formed by the dps PKS in the absence of heterologously
produced DpsC and DpsD.
More recent in vitro studies
by Hutchinson and co-workers unequivocally demonstrated that
DpsC is indeed responsible for the choice of the correct starter unit
for daunorubicin biosynthesis.
Other examples involving related PKS priming mechanisms
include the biosynthesis of frenolicin, R1128, and hedamycin.
In all cases, genes encoding KSIII components were identified
in the respective gene clusters. Analysis of the frenolicin (frn)
biosynthetic genes by the Hopwood group not only revealed an
ORF encoding a FabH orthologue (FrnI), but also an additional
ACP (FrnJ), which m ay serve as an intermediate anchor for the
Most likely the ACP-bound b-keto ester is processed to
the fully saturated butyryl moiety before priming the PKS. Since
no b-keto-processing enzymes were found in the sequenced frn
gene locus, it was proposed that these activities were recruited
from fatty acid metabolism (Fig. 11).
A similar scenario was found in the pathway of the R1128
complex, which incorporates butyryl, valeryl or 4-methylvaleryl
starters. In addition to the KSIII gene (zhuH), Khosla and co-
workers identified further malonyl AT (ZhuC) and priming ACP
(ZhuG) components encoded by the R1128 (zhu) biosynthetic gene
cluster. In analogy to anthracycline biosynthesis, it is assumed
that ZhuH functions as a priming KS that is solely dedicated
to the first malonyl-CoA condensation cycle.
In vitro studies
on the substrate specificity of ZhuH by Meadows and Khosla
strongly supported this model.
In addition, it was found that the
additional ACP (ZhuG) is indispensable for the incorporation of
non-acetate starter units.
It should be highlighted that the zhu
ketosynthases in the initiation (ZhuH) and elongation modules
(ZhuA) of aromatic polyketide synthases have orthogonal acyl
carrier protein specificity.
The crystal structure of the priming
KSIII (ZhuH) has been disclosed by the Stroud and Khosla
170 | Nat. Prod. Rep., 2007, 24, 162–190 This journal is
The Royal Society of Chemistry 2007
Downloaded on 06 February 2013
Published on 22 November 2006 on http://pubs.rsc.org | doi:10.1039/B507395M
View Article Online
Fig. 11 PKS priming involving KSIII components.
Furthermore, Khosla et al. succeeded in engineering
modified aromatic polyketides by recombination of the R1128
‘loading module’ consisting of ZhuC, ZhuH, and ZhuG, with
other min PKS components.
In the biosynthesis of hedamycin, which belongs to the plu-
ramycin group of antibiotics, an iterative type I PKS is utilized
for the formation of a short-chain starter unit (Fig. 12). Thorson
and co-workers reported that an additional ketosynthase (KSIII),
encoded by hedS, presumably plays a role in starter unit specificity
in utilizing the unusual hexenoate starter unit in favour of the
regular malonyl decarboxylation mechanism.
Here, the KSIII
component represents a unique link between a type I (HedT)
Fig. 12 KSIII as shuttle between type I and type II PKS components.
This journal is
The Royal Society of Chemistry 2007 Nat. Prod. Rep., 2007, 24, 162–190 | 171
Downloaded on 06 February 2013
Published on 22 November 2006 on http://pubs.rsc.org | doi:10.1039/B507395M
View Article Online
and a type II PKS (HedCDE). In addition, a gene encoding a
putative acyl transferase (HedF) has been located in the hed gene
cluster, which might be involved in malonyl-CoA:ACP transfer or
in loading the hexadienoate primer onto the PKS.
In conclusion, there is increasing evidence that KSIII compo-
nents serve as gatekeepers for the min PKS. The presence of a
KSIII and an additional ACP is always a good indication for
non-acetate priming.
While at least empirically the choice of
starters can be rationalized by the presence of these components,
the exact PKS priming mechanism remains obscure.
The second alternative priming pathway is the direct loading by
an acyl-CoA ligase and an AT, which are responsible for activation
and transfer of the starters onto the PKS, respectively (Fig. 10).
This mechanism has been proposed for malonamate and benzoate
starters in the tetracycline and enterocin pathways, respectively
(Fig. 13). It would be rational to assume that these carboxylic acids,
which clearly cannot be loaded by any of the other two priming
modes, are activated as CoA thioesters by an acyl-CoA ligase
(LIG) and then transferred onto the KS by an acyl transferase.
Analysis of the enterocin biosynthesis gene cluster
by Moore
and co-workers immediately lent support to this model, since gene
candidates for AT (encL) and LIG (encN) were identified. The
involvement of the ligase EncL was demonstrated by in vivo and
in vitro studies.
Substrate flexibility of both EncL and the
enc PKS allowed for the activation and attachment of benzoyl-
CoA surrogates. In a study on precursor-directed biosynthesis
of enterocin and wailupemycin derivatives, it was shown that
a number of non-natural aromatic and alicyclic primers were
accepted by the enc PKS and fully processed.
however, EncL does not seem to be required for chain initiation,
and also in this case the exact mechanism of KS priming remains
to be established.
To date, the biosynthesis of the putative malonamyl-CoA starter
in tetracycline and cervimycin biosynthesis has not been fully
elucidated. However, it is likely that it involves an asparagine
synthase-like amidase.
3.5 Ketoreductases
Ketoreductases (KRs) catalyze the stereospecific hydrogen trans-
fer from NAD(P)H onto a keto group, resulting in the formation
of a secondary alcohol. In aromatic polyketide biosynthesis, this
reaction can either influence the orientation of the poly-b-keto
chain for a favoured aldol condensation, which delineates the first
ring closure, or define the configuration of a persisting carbinol
moiety. In most cases, KRs are an integral part of bacterial
aromatic polyketide pathways. In such systems, KRs can be the
first enzymes to affect the nascent polyketide chain prior to
cyclization. Since the ketoreduction results in a transition from sp
to sp
hybridization, inducing a defined bend of the poly-b-keto
intermediate, the carbon chain may take up a favoured orientation
for an aldol reaction between the carbons involved in the first
ring formation. Consequently, ketoreduction has a pronounced
role in dictating the structure of the final product, and thus at
least indirectly contributes to the polyketide cyclization process
(Fig. 14). On the other hand, incomplete KR-deficient sets of
type II PKS enzymes may result in incorrectly folded polyketides.
When the act KR was omitted from its endogenous type II PKS,
the correctly folded act octaketide SEK4, as well as the incorrectly
folded shunt product SEK4b, was formed.
Another very important observation of such mixed first ring
cyclization events was reported by Kunnari et al. with randomly
generated S. peucetius and S. galilaeus KR mutants.
tivation and complementation experiments using mithramycin,
nogalamycin and aclacinomycin biosynthesis genes further sup-
port the involvement of ketoreductase in regiospecific cyclization
(Fig. 15). It is remarkable that all type II PKS-
derived polyketides that undergo ketoreduction, including ben-
zoisochromanones, tetracyclines, and anthracyclines, are similarly
reduced at the ninth carbon from the carboxy terminus of the
assembled polyketide. This strict regiochemistry is independent
from the polyketide chain length (octaketide to decaketide) and
the nature of the starter unit.
Similarly, it is assumed that a KR may also be involved
in preforming the ‘bent’ part of the polycyclic ring system of
pradimicin-type polyketides (Fig. 16). In this regard it should be
noted that KR may occur in cyclase–reductase didomain enzymes,
as in griseorhodin biosynthesis.
Structural investigations of KR proteins could shed light on the
biochemical basis of this important step in polyketide biosynthesis.
In two independent studies the crystal structure of the
actinorhodin polyketide reductase (ActIII) has been investigated.
Hadfield et al. proposed a mechanism for ACP and polyketide
whereas Korman et al. focused on cofactor binding
Fig. 13 PKS priming with benzoate and malonamate.
172 | Nat. Prod. Rep., 2007, 24, 162–190 This journal is
The Royal Society of Chemistry 2007
Downloaded on 06 February 2013
Published on 22 November 2006 on http://pubs.rsc.org | doi:10.1039/B507395M
View Article Online
Fig. 14 Impact of the KR on cyclization pattern.
Fig. 15 Regioselectve ketoreduction.
Fig. 16 Ketoreduction in pradimicin-type polyketide biosynthesis.
and substrate specificity.
Apart from the act KR, which
was the first cloned KR controlling C-9 ketoreduction, genes
encoding closely related enzymes were found in various other
gene clusters involved in benzoisochromanequinone biosynthesis,
e.g . medermycin,
and enterocin.
In many cases
it has been possible to interchange ketoreductases from these
different systems, and KRs have been shown to function with
hybrid PKSs.
Iterative type II polyketide synthases, cyclases
and ketoreductases exhibit context-dependent behaviour in the
biosynthesis of linear and angular decapolyketides.
This journal is
The Royal Society of Chemistry 2007 Nat. Prod. Rep., 2007, 24, 162–190 | 173
Downloaded on 06 February 2013
Published on 22 November 2006 on http://pubs.rsc.org | doi:10.1039/B507395M
View Article Online
various mix-and-match experiments, it was hypothesiz ed that
the polyketide chain is completely assembled before reduction
of the keto group takes place.
However, as a more recent
study revealed, the model for aromatic polyketide biosynthesis
is likely to be oversimplified. While most minimal PKSs are
able to operate in the absence of a KR and assemble polyketide
chains, the enterocin PKS is surprisingly unable to produce
polyketides without its endogenous KR. This suggests that if
chain growth terminates at this stage, truncated pentaketides
are either not released from the PKS complex, or are simply
catabolized. Point mutations in the active site (S to A, and Y to
F) of EncD implied that the KR is required to serve as a catalytic
and not just a structural role.
Consequently, ketoreduction
in enterocin biosynthesis occurs during the chain elongation
and not after, as proposed for most other type II PKS systems.
Furthermore, it is remarkable that although co-expression of the
act minimal PKS genes with encD results in the formation of the
same shunt products, DMAP and aloesaponarin,
no polyketide production was observed even when the encD null
mutant was complemented with the functionally equivalent act
KR. The observed differences between the enc and act KRs
suggest that there may be two pathways for ketoreduction in type
II PKS systems, either during polyketide assembly (EncD) or
post-polyketide assembly (ActIII). The likely mode of catalysis
by EncD involves its association with the KS and reduction of
the KS-bound pentaketide intermediate. Two routes involving
either cyclized (Staunton model) or open-chain intermediates are
conceivable (Fig. 17).
Fig. 17 Model of ketoreduction in the enterocin pathway.
In a few aromatic polyketide biosynthetic pathways the action of
an additional KR can result in a chiral carbon centre, which does
not undergo elimination and is still present in the final product. A
prominent example is the antitumoural polyphenol doxorubicin
from S. peucetius, which has chiral secondary hydroxy groups at C-
17 and on the side chain. The C-17 keto group is stereospecifically
reduced by aklaviketone reductase (DnrH), a result which has been
experimentally proven by expression and feeding experiments
(Fig. 18). Genes encoding related enzymes were found in the
aclacinomycin/aklavinone (aknU),
nogalamycin (snoaF)
and chartreusin (chaL)
biosynthesis gene clusters. Inactivation
and complementation experiments showed that DnrU is required
for reduction of the side chain keto group of anthracyclic
Fig. 18 C-17 ketoreduction in anthracycline biosynthesis.
The benzoisochromanequinone frameworks of actinorhodin (or
medermycin) and granaticin feature chiral carbon centres with
opposite configuration. The key to the S-orR-configuration of
the pyran rings are C-3 KRs, which catalyze hydride transfer from
the pro-S or pro-R site, respectively. Ichinose et al. demonstrated
that ActVI-ORFI converts a bicyclic intermediate to a chiral
alcohol, which is subsequently converted to a key intermediate in
BIQ biosynthesis, (S)-DNPA.
An orthologue of ActVI-ORFI is
encoded in the medermycin (med) biosynthesis gene cluster (med-
(Fig. 19). A corresponding KS from the granaticin pathway
exhibits the opposite stereospecificity, yielding the enantiomeric
The C-3 KRs are capable of converting even synthetic
surrogates to their corresponding secondary alcohols with very
high enantioselectivity.
It has been shown previously that
inactivation of act-ORF1 results in the formation of DMAC and
Surprisingly, co-expression of the act min
PKS, CYC and aromatase genes with gra KR, as well as
complementation of the act KR with the gra C-3 KR (Gra-ORF6)
revealed that the KRs act context-dependently.
3.6 Cyclases
In the current model for aromatic polyketide biosynthesis, highly
reactive poly-b-keto intermediates are formed, which can undergo
spontaneous aldol reactions without enzyme stabilization or
directed cyclizations (Fig. 20). The occurrence of spontaneous
cyclizations resulting from incomplete sets of type II PKS
systems was most impressively demonstrated by Shen et al. in
a study on spore pigment biosynthesis. The metabolic profile of a
174 | Nat. Prod. Rep., 2007, 24, 162–190 This journal is
The Royal Society of Chemistry 2007
Downloaded on 06 February 2013
Published on 22 November 2006 on http://pubs.rsc.org | doi:10.1039/B507395M
View Article Online
Fig. 19 Enantioselective C-3 ketoreduction in BIQ biosynthesis.
Fig. 20 Different folding patterns of polyketides resulting from sponta-
neous aldol reactions.
recombinant Streptomyces strain expressing the S. coelicolor
whiE min PKS revealed the presence of a multitude of polyketides
with varying chain lengths and cyclization patterns.
results not only highlight the fascinating metabolic diversity that
results from type II PKSs, but also that these systems need to work
in a highly concerted action in order to yield a defined polyketide
So-called cyclases (CYCs), which function in a ‘chaperone-like’
manner, help in directing nascent polyketide intermediates into
particular reaction channels. In the presence of these enzymes,
spontaneous aldol chemistry is efficiently suppressed.
Since biochemical investigations of cyclases and their unstable
substrates are particularly difficult, to date very little is known
about the exact mode of action of these intriguing cofactor-free en-
zymes. With a few exceptions, their function could only be deduced
by inactivation and recombination experiments, and elucidation of
the (shunt) metabolites produced. Also, it could be demonstrated
that (at least in e.g. anthracycline biosynthesis) cyclase-catalyzed
C–C bond formations take place prior to dehydration mediated by
aromatases (ARO, see below). Analyses of deduced gene products
from various type II PKS gene clusters revealed that cyclases
may occur as ‘didomain’ enzymes with internally duplicated
motifs, such as ActVII in actinorhodin biosynthesis,
and that
expression of the N- or C-termini alone resulted in loss of function.
Alternatively, CYCs may also exist as ‘monodomain’ enzymes, e.g.
In various cases, these CYC domains are fused with other
functionalities, for example as CYC/MT, or CYC/KR. Despite
the large number of cloned CYC genes, the only biochemically
characterized cyclases are the Tcm F2 cyclase
and TcmN
the tetracenomycin pathway, and the ester cyclases SnoaL and
AknH involved in nogalamycin and aclacinomycin biosynthesis,
In solution, these cyclases exist as a homotrimer
(Tcm F2 CYC), a homodimer (TcmN), and a monomer (SnoaL).
Interestingly, the different groups of cyclases do not share unique
common motifs (except for HxGTHxDxPxH, which is likely to
form part of the active site of cyclases), and exhibit no phylogenetic
There is an ongoing discussion on whether or not KS
first ring cyclization. It has been reported that the first ring cycliza-
tion largely depends on the type of KS and, more importantly,
on regioselective ketoreduction. However, it was implicated that
folding of the polyketide chain is not dictated by the minimal
polyketide synthase in the biosynthesis of mithramycin and
anthracyclines, but rather by ketoreduction and cyclases.
in tetracenomycin biosynthesis, the cyclase TcmN seems to play a
crucial role in first ring formation, directing the initial cyclization
between C-9 and C-14. Meurer et al. concluded that KS, CYC
and KR components exhibit context-dependent behaviour in the
biosynthesis of linear and angular decapolyketides.
Early inactivation and recombination experiments, which have
been thoroughly reviewed by Rawlings
and Shen,
gave the first
clue on the functions of downstream cyclases in tetracenomycin
biosynthesis, such as TcmN, TcmI, and TcmJ, which were identified
as second, third and fourth ring cyclases, respectively
(Fig. 21).
The cyclization events in angucycline biosynthesis have also
been the subject of many studies. Recently, functional analy-
ses of jadI revealed the role of the deduced gene product in
angucyclinone cyclization.
Hautala et al. studied second and
third ring cyclization in anthracycline biosynthesis, and identified
dpsY and snoaM as cyclase genes from S. peucetius and S.
noglalater, respectively. From analysis of the metabolites produced
by mutants they concluded that the third ring closure is most
likely mediated by the same enzyme that catalyzes the second
ring closure.
It is remarkable that in many cases cyclase genes
are not interchangeable. While the incompatibility of enzymes
This journal is
The Royal Society of Chemistry 2007 Nat. Prod. Rep., 2007, 24, 162–190 | 175
Downloaded on 06 February 2013
Published on 22 November 2006 on http://pubs.rsc.org | doi:10.1039/B507395M
View Article Online
Fig. 21 Cyclization events in tetracenomycin biosynthesis.
involved in the biosynthesis of non-reduced (tcm) and partially
reduced (jad, dps) polyketides can be rationalized from the
deviating folding of the polyketide chain, in other cases the
failure of hybrid systems lacks an explanation. A recent example
was reported by Wohlert et al., who found that heterologously
produced cyclases DpsY and JadI are incompatible in the pres-
ence of both the tetracenomycin (tcm) and daunorubicin (dps)
Conversely, Mets
a et al. succeeded in engineering
anthracycline biosynthesis (nogalamycin, aclacinomycin) toward
angucyclines by co-expressing the angucycline-specific cyclase
gene pgaF from a silent gene cluster.
In this study, PgaF, which
is similar to angucycline cyclases LanF, UrdF and JadI from
the landomycin,
and jadomycin
pathways, respectively, was identified as the third and fourth ring
cyclase in angular systems (Fig. 22).
In anthracycline biosynthesis, cyclases such as CmmY
SnoaM (nogalamycin),
and AknW
are required for the complete cyclization of the
second (and most probably the third) ring of linear polyketides.
The fourth ring formation is catalyzed by a group of ester cyclases
(DnrD, DauD, RdmA, AcmA, SnoaL). The enzyme-catalyzed
aldol reaction usually results in the formation of a secondary
alcohol with 9R configuration, except for nogalamycin and
steffimycin, which are both 9S-configured. The cyclase SnoaL,
which dictates the 9S configuration of anthracyclines from S.
nogalater, has been characterized by Torkkell et al.
(Fig. 23). In
addition, Schneider, M
a and co-workers were able to solve
the protein crystal structure of SnoaL, providing the first insight
into a novel mechanism for enzymatic aldol condensation.
Individual amino acids were spotted that differ between the
Fig. 22 Directing anthracycline biosynthesis towards angucyclines.
Fig. 23 Stereochemistry of the fourth ring ester cyclases.
9R and 9S specific enzymes, and the recent structural analysis
of AknH helped to shed more light on the molecular basis of
enantiospecificity of the ester cyclases.
It is remarkable that only very few basic cyclization patterns
are realized in aromatic polyketide biosynthesis. Virtually all
aromatic polyketides from bacteria are either linear (e.g. biq,
anthracycline, aureolic acids, tetracenomycins and tetracyclines)
or angular (e.g . angucyclines and pradimicin-type polyketides).
A common structural feature is that in these systems rings
are annellated only on one or two sides. A clear exception
is the antibacterial polyketide resistomycin from Streptomyces
The naphthanthrene framework of this rare
metabolite contains multiply annellated rings, and represents
the only ‘discoid’ polyketide ring system found so far. Jakobi
176 | Nat. Prod. Rep., 2007, 24, 162–190 This journal is
The Royal Society of Chemistry 2007
Downloaded on 06 February 2013
Published on 22 November 2006 on http://pubs.rsc.org | doi:10.1039/B507395M
View Article Online
Fig. 24 U-shape vs. S-shape folding in angular, linear and discoid polyketide pathways.
and Hertweck cloned and sequenced the gene locus encoding
resistomycin (rem) biosynthesis, and identified several cyclase gene
candidates, remI, remF ,andremL.
Mutational analyses and
recombination experiments, together with elucidation of the shunt
metabolites, produced gave an unexpected result. The resistomycin
pathway shares early biosynthetic steps with tetracenomycin, but
diverges from all other paths by folding the polyketide chain into
an S-shape as opposed to the regular U-shape (Fig. 24).
3.7 Aromatases
Apart from cyclases, the cyclization process is often supported
by so-called aromatases. These enzymes are known to dehydrate
cyclic alcohols to yield aromatic ring systems. In anthracycline
biosynthesis, aromatases AknE1, RdmK, and SnoaE
responsible for the aromatization of the A-rings. Recently Yang,
Rohr and co-workers identified JadH as an aromatase involved
in the biosynthesis of jadomacin A and B. When jadH was
deleted in the genome of Streptomyces venezuelae ISP5230, a
novel angucycline metabolite, 2,3-dehydro-UWM6, was formed.
2,3-Dehydro-UWM6 and UWM6, obtained by expression of
the jadomycin PKS gene cluster in S. lividans, could be trans-
formed into jadomycin B by a ketosynthase (jadA)mutantofS.
venezuelae. These angucycline intermediates were also converted
to the non-glycosylated jadomycin B derivative jadomycin A by
transformants of the heterologous host S. lividans expressing
the jadFGH oxygenases in vivo. After overexpression JadH was
isolated for enzyme assays. UW6 was converted to rabelomycin
and 2,3-dehydro-UW6 to dehydrorabelomycin, indicating that
JadH possesses dehydrase and oxygenase activity
(Fig. 25).
A further example of an aromatase is LanV, which is involved in
landomycin biosynthesis. Expression of lanV in the rabelomycin-
producing urdM mutant of Streptomyces fradiae resulted in
Fig. 25 Ring aromatization in jadomycin biosynthesis.
the formation of 9-C-D-olivosyltetrangulol, urdamycin B and
urdamycinone B. This result indicated that LanV is a bifunctional
enzyme: it catalyzes the ketoreduction step at position C-6 of
the polyketide core followed by a dehydration reaction, and thus
aromatization of ring A of the angucycline (Fig. 26).
This journal is
The Royal Society of Chemistry 2007 Nat. Prod. Rep., 2007, 24, 162–190 | 177
Downloaded on 06 February 2013
Published on 22 November 2006 on http://pubs.rsc.org | doi:10.1039/B507395M
View Article Online
Fig. 26 Function of LanV as ketoreductase and aromatase involved in
landomycin biosynthesis.
4 Polyketide-tailoring reactions
4.1 Methyltransferases
Most methyltransferases use the ubiquitous cofactor S-adenosyl-
L-methionine (SAM) and transfer its activated methyl group to
nitrogen, carbon or oxygen in a variety of target molecules. To
date over twenty deduced amino acid sequences have been reported
that might function as methyltransferases in Streptomyces. Among
these are DnrK from S. peuceticus,DauKfromS. sp. C5,
CmmM1 from S. griseus, OxyF from S. rimosus, GilMT from
S. griseoflavus, AlpH from S. ambofaciens and TcmN and TcmO
from S. glaucescens, all involved in the biosynthesis of aromatic
The degree of amino acid conservation among
these enzymes is low, although they often share a chain-fold
consisting of a central b-sheet with surrounding a-helices. In
addition, a SAM binding site can be found in all of these
enzymes. The function of some of these genes has been identified
by gene disruption experiments and structure elucidation of
the metabolites produced by the mutants. Recently the crystal
structure of a ternary complex of DnrK with bound SAM
and 4-methoxy-e-rhodomycin T has been solved.
DnrK, a
carminomycin 4-O-methyltransferase, methylates the 4-hydroxyl
group of daunorubicin at a very late step in the biosynthesis.
DnrK is a homodimer, and the subunit displays the typical fold
of small molecule O-methyltransferases. The structure provides
insights into the recognition of the anthracycline substrate and
also suggests conformational changes as part of the catalytic
cycle of the enzyme. The position and orientation of the bound
ligands are consistent with an S
2 mechanism of methyl transfer.
Mutagenesis experiments on a putative catalytic base confirm
that DnrK most likely acts as an entropic enzyme, in that rate
enhancement is mainly due to orientational and proximity effects.
DnrK exhibits a rather relaxed substrate specificity, as it accepts
carminomycin, 13-deoxycarminomycin and e-rhodomycin T as
The crystal structure of RdmB, an enzyme involved
in rhodomycin biosynthesis in Streptomyces purpurascens, has also
been elucidated. RdmB strongly resembles DnrK (52% identical
amino acids) and shows the typical methyltransferase fold and a
SAM binding site. However, most surprisingly RdmB does not act
as a methyltransferase but as a regiospecific hydroxylase.
4.2 Oxygenases
Oxidative tailoring reactions, which are catalyzed by oxygenases,
are the key to the vast structural diversity of polyketides and to
their biological activity.
In general, these enzymes promote
the incorporation of oxygen into a substrate. Oxygenases are
divided into monooxygenases and dioxygenases, depending on the
number of oxygen atoms that are inserted into the substrates. Since
monooxygenases reduce the second oxygen atom of dioxygen to
water, they have also been termed ‘mixed function oxygenases’.
Several types of oxygenases are known that play a role in
post-PKS II modification: cytochrome P-450 monooxygenases
(CYP450), flavin-dependent mono- and dioxygenases, and an-
throne oxygenases. These enzymes catalyze different reactions,
e.g . hydroxylation, epoxidation, quinone formation, and oxidative
rearrangement of the Baeyer–Villiger or the Favorskii type. Oxy-
genases differ significantly in their mode of action, requirement of
cofactors and degree of flexibility toward the substrate.
Anthrone-type oxygenases. The oxidation of naphthacenone-
and anthrone-type precursors of aromatic polyketides to the
corresponding quinone derivatives is a common and important
tailoring step in the biosynthesis of aromatic polyketides. An-
throne oxygenases, classified as internal monooxygenases, are
responsible for these reactions. Interestingly, these enzymes lack
the requirement for any cofactor.
Anthrone oxygenases use
their substrates as a reducing equivalent for the reduction of
one oxygen atom (from dioxygen) to water.
monooxygenases involved in aromatic polyketide biosynthesis
were first studied by Shen and Hutchinson in 1993 in the
context of tetracenomycin biosynthesis in S. glaucescens.The
monooxygenase TcmH was found to catalyze the conversion of the
naphthacenone tetracenomycin F1 at position C-5 into the 5,12-
naphthacenequinone tetracenomycin D3 (Fig. 27).
TcmH has a
molecular weight of 12.5 kDa and forms a homotrimer in solution,
as revealed by gel filtration. Inhibitor studies suggested that
sulfhydryl groups and histidine residues are essential for catalysis.
The second quinone-forming monooxygenase studied was ActVA-
orf6 from S. coelicolor A3(2), which shows 39% identity to
It was proposed that ActVA-orf6 catalyzes the formation
of dihydrokalafungin (Fig. 27), an intermediate in actinorhodin
biosynthesis. Since the natural substrate of ActVA-orf6 was not
available, its catalytic activity was studied using tetracenomycin
F1, the substrate of TcmH. Besides this tetracyclic compound,
the protein also oxidizes tricyclic anthrones, suggesting a relaxed
substrate specificity.
The ActVA-orf6 protein also does not re-
quire exogenous cofactors or metal ions, and there is no indication
of any bound cofactors. In contrast to TcmH, which presumably
possesses catalytically relevant sulfhydryl groups,
does not contain any cysteine residues. Site-directed mutagenesis
experiments implicate that residue His-52 is important for the
catalytic activity of ActVA-orf6, as well as of TcmH.
internal monooxygenase is AknX, which has been purified and
178 | Nat. Prod. Rep., 2007, 24, 162–190 This journal is
The Royal Society of Chemistry 2007
Downloaded on 06 February 2013
Published on 22 November 2006 on http://pubs.rsc.org | doi:10.1039/B507395M
View Article Online
Fig. 27 Reactions catalyzed by the anthrone-type oxygenases TcmH,
ActVA-orf6, and AknX.
characterized by Ebizuka and co-workers more recently.
enzyme is involved in the oxidation of aklanonic acid anthrone
to aklanonic acid ( Fig. 27). In accordance with the findings
regarding ActVA-orf6, Cys residues of AknX seem not to be
important for catalytic activity of the enzyme, as a C52S mutant
of AknX retained its oxygenase activity. Shen and Hutchinson
proposed that the reaction mechanism of TcmH follows a radical
process, which includes the generation of a superoxide anion
However, the anthrone oxygenase activity of AknX
was not affected by superoxide dismutase.
Moreover, mutants
of AknX His-49, which corresponds to His-52 of ActVA6 and
TcmH, retained 40–45% enzyme activity. This implies that His-49
at least in the AknX enzyme reaction is not an essential residue.
Kinetic analyses of AknX mutants suggested that the Trp-67
residue plays a key role in the AknX oxygenation reaction.
to the unique features of quinone-forming monooxygenases, it was
assumed that they might share a novel protein fold. In 2003, X-
ray analyses of ActVA-orf6 crystals revealed a ferredoxin-like fold
with a novel dimeric assembly, but also indicating that the widely
represented ferredoxin fold may sustain another functionality.
The structure of ActVA-Orf6 suggests that four residues may be
important for binding and/or catalysis: Tyr51, Asn62 and Trp66,
which belong to b-sheets, and Tyr72 as part of an a-helix.
In addition to aknX, tcmH, and actVA6, several genes, such as
elmH of Streptomyces olivaceusTu2353,
dnrG of S. peucetius,
dauG of Streptomyces sp. strain C5,
snoaB of S. nogalater,
mOIII of S. argillaceus,
and frnU of S. roseofulvus,
have been
reported to be possibly involved in similar oxygenation reactions.
The putative aklanonic acid anthrone oxygenases AknX, DnrG,
DauG, and SnoaB are closely related to each other and have a
distal relationship with the other oxygenases like TcmH, ActVA6,
MtmOIII and FrnU. Their amino acid sequence alignment shows
that only four amino acid residues, Pro-44, Gly-45, Ala-65 and
Trp-67, are fully conserved in these proteins.
Flavin-dependent oxygenases. Flavoenzymes are very versatile
catalysts and are involved not only in the activation of oxygen
for oxidation and hydroxylation reactions, but also in dehy-
drogenation reactions, in light emission, and in one- and two-
electron transfer reactions.
Flavin-dependent oxygenases
often show high homologies within two regions of their amino
acid sequences. The first region, which is usually located close
to the N-terminus of the protein, has been implicated in the
binding of the adenosine moiety of the FAD cofactor, whereas the
second motif is associated with the binding of the flavin portion of
Flavoprotein-catalyzed oxidative reactions associated with
post-PKS modifications include hydroxylations, epoxidations,
Baeyer–Villiger and Favorskii rearrangements.
Villiger rearrangements are one of the most widespread FAD-
dependent oxygenase reactions in the biosynthesis of type II PKS-
derived metabolites.
In general, flavin-containing Baeyer–
Villiger monooxygenases (BVMO) employ NADPH and molec-
ular oxygen to catalyze the insertion of an oxygen atom into a
carbon–carbon bond of a carbonylic substrate.
The catalytic
mechanism of BVMO, which has been rewieved in detail,
involves the formation of two crucial intermediates: flavin peroxide
generated by the reaction of the reduced flavin with molecular
oxygen, and a ‘Criegee’ intermediate resulting from the attack of
the flavin peroxide onto the substrate. One of the best examples
indicating the importance of BVMO as post-PKS II enzymes
has been outlined in the biosynthetic pathway of the clinically
used antitumour antibiotic mithramycin (Fig. 28) produced by S.
argillaceus. The process that is responsible for conferring biolog-
ical activity to the molecule has been identified by Prado et al.
as the Baeyer–Villiger oxygenation of the tetracyclic intermediate
premithramycin B, which is followed by hydrolytic ring-opening
Fig. 28 Baeyer–Villiger oxygenation of premithramycin B leads to the
active form of mithramycin.
This journal is
The Royal Society of Chemistry 2007 Nat. Prod. Rep., 2007, 24, 162–190 | 179
Downloaded on 06 February 2013
Published on 22 November 2006 on http://pubs.rsc.org | doi:10.1039/B507395M
View Article Online
of the lactone intermediate and the formation of a tricyclic
mithramycin precursor.
Recently, Rohr and co-workers have
purified the protein MtmOIV, which is responsible for the oxidative
C–C bond cleavage of premithramycin B that leads to the active
molecule mithramycin after decarboxylation. The authors isolated
the lactone intermediate, premithramycin B-lactone, proving the
Baeyer–Villiger mechanism in the formation of MtmOIV.
The involvement of the three another oxygenases (mtmOI,
mtmOII, mtmOIII) in mithramycin biosynthesis was extensively
reviewed by Rohr and co-workers in 2002.
BVMOs have been
investigated, in particular in the biosynthesis of angucyclines, such
as urdamycin, jadomycin, and gilvocarcin. A unique hydroxylation
mechanism was described for the biosynthesis of urdamycin
A. The bifunctional oxygenase/reductase UrdM introduces the
angular 12b-hydroxy group into urdamycin molecules by a
Baeyer–Villiger mechanism.
The accumulation of rabelomycin
and urdamycin L in S. fradiae DurdM can be explained by the
reaction scheme shown in Fig. 29.
Fig. 29 12b-Hydroxylation of urdamycin A catalyzed by UrdM involves
a Baeyer–Villiger monooxygenation reaction. Enz = enzyme.
Recently, Rohr and co-workers have demonstrated that BVMOs
are involved in ring-opening reactions during the biosynthesis of
jadomycin and gilvocarcin, something which is essential for the
biological activity of these polyketides.
One of the key steps in gilvocarcin and jadomycin biosynthesis
is the oxidative rearrangement sequence by which their unique
chromophores are generated. GilOI and GilOIV, both responsible
for the oxidative C–C bond cleavage, are the key enzymes in gilvo-
carcin biosynthesis. Gene inactivation of gilOI and gilOIV and
structural elucidation of the metabolites produced by the mutants
helped deduction of the ring-opening mechanisms (Fig. 30). It was
shown that GilOI and GilOIV cooperate in a sequential action to
promote the C–C bond cleavage, with GilOIV acting prior to
It has been shown that JadF, JadH, and JadG are involved
in extremely complex oxygenation/dehydration processes during
jadomycin A biosynthesis. The inactivation and heterologous ex-
pression experiments show that JadF, JadH and JadG are arranged
in an oligomeric protein complex.
Strikingly, non-functionality
of one of these enzymes causes the loss of the activity of the other
two enzymes. In addition to oxygenase activity, JadF and JadH
also function as 2,3- and 4a,12b-dehydratases, respectively, and are
responsible for the aromatization of ring A.
Apparently, these
additional enzyme functions are not dependent on correct protein
Fig. 30 Model for the stepwise oxidative C–C bond cleavage in the
biosyntheses of the gilvocarcins and the jadomycins.
180 | Nat. Prod. Rep., 2007, 24, 162–190 This journal is
The Royal Society of Chemistry 2007
Downloaded on 06 February 2013
Published on 22 November 2006 on http://pubs.rsc.org | doi:10.1039/B507395M
View Article Online
complex formation. A general scheme of oxidative rearrangements
during the biosyntheses of the jadomycins and the gilvocarcins is
The FAD-dependent monooxygenase ChaZ, which seems to
work as a Baeyer–Villigerase, plays a key role in the biosynthesis
of the potent antitumour agent and topoisomerase inhibitor
chartreusin. The unusual structure of the bislactone aglycone
and puzzling results of
C labelling experiments hinted towards
an intriguing biosynthetic scheme. Hertweck and co-workers
demonstrated that inactivation of chaZ results in the accumulation
of the anthracyclic intermediate or shunt product resomycin
which revealed that chartreusin biosynthesis involves the
unprecedented rearrangement of an anthracyclic ring system.
According to the
C-acetate labelling pattern, a quinone carbon
bond is cleaved and a new C–C bond is formed between the
carbonyl and the unsubstituted carbon of the C ring. A hydroxyl
group could then be introduced either during the ring cleavage or
by hydroxylation of the angucyclic intermediate by the putative
hydroxylase ChaX. After formation of the polycyclic framework,
the quinone oxygen, probably in the enol form, could attack the
methyl ester carbonyl with substitution of methanol, forming
the first lactone ring. Finally, the putative dioxygenase ChaP,
which belongs to the vicinal oxygen chelate (VOC) superfamily of
enzymes, could disrupt the dione moiety by attack of a dioxygen
species and loss of carbon by rearrangement of pradimicin-type
polyketides, yielding the spiro metabolite griseorhodin oxide. Sub-
sequent lactone formation would ultimately furnish the chartarin
aglycone (Fig. 31).
Baeyer–Villigerases have also been implicated in the biosyn-
theses of the rare spiroketal metabolites rubromycin and grise-
orhodin. Cloning and analysis of the griseorhodin biosynthesis
gene cluster from a tunicate-derived Streptomyces strain re-
vealed the presence of a plethora of genes encoding putative
On the basis of deduced gene functions, Li
and Piel established a fascinating albeit hypothetical biosyn-
thetic scheme. In this scheme, the formation of the spiroketal
pharmacophore would require the cleavage of three C–C bonds
to extrude a C
unit, followed by the rearrangement of a
proposed pradimicin-type intermediate. Li and Piel pointed out
the striking similarity between GrhO6 and MtmOIV, which
suggests that this enzyme acts as a Baeyer–Villigerase. The model
involving a tridecaketide intermediate is strongly supported by
independent studies of Minas, Bailey and co-workers, who found
that heterologous expression of a fragment of the rubromycin
biosynthetic gene cluster from S. collinus in S. coelicolor resulted
in the formation of the hexacyclic shunt product collinone.
Interestingly, fredericamycin, which features a unique spiro carbon
centre, seems to share early biosynthetic steps with griseorhodin
and rubromycin. Most recently, Hutchinson, Shen and co-workers
unveiled the molecular basis for fredericamycin biosynthesis by
sequence analysis of the fredericamycin biosynthesis gene cluster
from S. griseus.
Fig. 31 Model of chartreusin biosynthesis involving the oxidative rearrangement of an anthracyclic precursor.
This journal is
The Royal Society of Chemistry 2007 Nat. Prod. Rep., 2007, 24, 162–190 | 181
Downloaded on 06 February 2013
Published on 22 November 2006 on http://pubs.rsc.org | doi:10.1039/B507395M
View Article Online
Favorskiiase. In contrast to Baeyer–Villiger rearrangements,
the occurrence of Favorskii-type oxidative rearrangements of a-
ketones to the corresponding acids or esters is scarce.
To d ate
only one ‘favorskiiase’, EncM, which is involved in enterocin
biosynthesis, has been characterized.
The presumed substrate
of EncM is a linear C-9-reduced octaketide that is oxidized
at C-12 to form the 11,12,13-trione intermediate (Fig. 32).
Moore and co-workers demonstrated that EncM facilitates two
aldol condensations (between C-6 and C-11, and C-7 and C-
14), generating chiral centres at each position. Usually, cyclases
catalyze such aldol condensation reactions during aromatic
polyketide assembly. However, the enterocin gene cluster is an
exception among all other known type II PKS gene sets, as it
lacks typical cyclase- and aromatase-encoding genes.
also participates in two heterocycle-forming reactions during the
formation of desmethyl-5-deoxyenterocin. It would be tempting
to employ this unique enzyme in combinatorial biosynthesis
approaches to produce novel derailed polyketide metabolites.
Unfortunately, the co-expression of encM with actinorhodin PKS
genes did not lead to the production of any new compounds,
suggesting that EncM may be specific for its endogenous type
II PKS or for benzoyl-primed polyketide precursors.
The final
hydroxylation step from 5-deoxyenterocin to enterocin is catalyzed
by EncR, a cytochrome P-450 monooxygenase (see below) as-
sisted by the translationally coupled ferredoxin-like encQ gene
Cytochrome-dependent P450 monooxygenases. Surprisingly, to
date few CYP450 monooxygenases have been described in the
context of bacterial aromatic polyketide biosynthesis. In addition
to EncR, the monooxygenase responsible for the transformation of
deoxyenterocin into enterocin (see above), DoxA from S. peucetius
(and S. sp. strain C5) has been studied. The cytochrome P450
oxygenase participates in the biosynthesis of the clinically impor-
tant antitumour antibiotics daunorubicin (DNR) and doxorubicin
DoxA exhibits a broad substrate specificity, catalyzing
the anthracycline hydroxylation on both C-13 and C-14. Priestley
and co-workers demonstrated that DoxA catalyzes the conver-
sion of the anthracycline analogue desacetyladriamycin into the
novel anthracycline metabolite 10-hydroxydesacetyladriamycin
(Fig. 33). In these experiments DoxA showed an absolute require-
ment for molecular oxygen and NADPH.
Fig. 32 Proposed biosynthetic pathway of enterocin, indicating the role of E ncM in post-PKS modifications.
Fig. 33 Enzymatic reactions catalyzed by DoxA.
182 | Nat. Prod. Rep., 2007, 24, 162–190 This journal is
The Royal Society of Chemistry 2007
Downloaded on 06 February 2013
Published on 22 November 2006 on http://pubs.rsc.org | doi:10.1039/B507395M
View Article Online
4.3 Glycosyltransferases
Many bacterial aromatic polyketides contain deoxysugars as
pivotal structural elements. The formation and attachment of these
sugars is catalyzed by several enzymes encoded by genes that are
also located in the biosynthetic gene clusters. The formation of
nucleotide-activated sugars has been reviewed several times
and is therefore not part of this review