Angewandte
Chemie
We individually expressed in E. coli each of AM-2699,
AM-6565, AH-6564, AH-7510, CoL-2700, CoL-7500, AT-
2704, and AT-7501, as well as the N-terminal ACP domains
from the polyene PKS multienzyme CleAI (STRVN_2730)
and the macrocyclic polyol PKS multienzyme AzlA1
(STRVN_7492). All were obtained as purified recombinant
proteins (Figure S2 in the Supporting Information). For
assays of arginine monooxygenase activity, l-arginine (3)
was incubated with purified AM in phosphate buffer and the
reaction was monitored by using HPLC–MS. Arginine
monooxygenase, first described in Streptomyces griseus,[15] is
one of very few flavin-dependent enzymes capable of directly
decarboxylating an a-amino acid in the presence of oxygen to
yield an amide. With both AM-2699 and AM-6565, rapid
conversion to 4 was observed, with no evidence either of
further hydrolysis to 4-guanidinobutyric acid (5) or of
a competing oxidative deamination to form 4-guanidinyl-2-
oxo-butanoate (Figure 2a). In contrast, there was no observ-
able reaction when arginine was replaced by either lysine,
glutamine, asparagine, tryptophan, ornithine, or histidine.
When fully [13C,15N]-labeled l-arginine was used as substrate,
the labeled amide product 4 was obtained in essentially
quantitative yield, and its mass was shifted higher by nine
mass units, as expected (Figure S3). The structure of labeled
product 4 was further confirmed by 13C NMR analysis (Fig-
ure S4). The labeled 3 and 4 were then fed to the DSM 4137
strain as potential precursors of the guanidinyl-containing
secondary metabolites (Figures S5 and S6). Both precursors
were incorporated intact, into both azalomycin and desulfo-
clethramycin (clethramycin, the final product of the pathway,
is a minor coproduct under these conditions). The efficiency
of incorporation from 3 and 4 was 10% and 72%, respec-
tively, into azalomycin F3a, and 17.5% and 93%, respectively,
into desulfoclethramycin. These results provided the first
direct evidence of the importance of the AM-initiated
pathway for starter unit biosynthesis in S. violaceusniger.
Amide hydrolase activity was also monitored using
HPLC–MS, facilitated by the use of labeled 4 as substrate.
Purified recombinant AH-6564 and AH-7510 both efficiently
catalyzed this reaction (Figure 2b and Figure S7). Likewise,
both of the predicted ligases, CoL-2700 and CoL-7500,
efficiently catalyzed the reaction of 4-guanidinobutyric acid
(5) with ATP and CoASH in the presence of magnesium ions
to give the thioester 6 (Figure 2c and Figure S8). Neither 3-
guanidinopropionic acid nor 4-aminobutyric acid were sub-
strates for these ligase enzymes. These results taken together
suggest that there is a single intracellular pool of 6 in DSM
4137 providing starter units for both the linear polyene and
the macrocyclic polyketide.
Scheme 1. Proposed pathway in S. violaceusniger DSM 4137 for produc-
tion of 4-guanidinobutyryl-CoA from arginine, and transfer of the
guanidinobutanoyl group to the N-terminal acyl carrier protein (ACP)
domain of the azalomycin (azl) and clethramycin (cle) modular
polyketide synthases.
The mechanism by which a guanidine-containing starter
unit is incorporated into antifungal polyketides has been
a matter of conjecture, despite early attempts to incorporate
label from plausible precursors.[10a] We considered the mech-
anistic precedent offered by known examples in which
a nonribosomal peptide synthetase (NRPS) module inte-
grated into a hybrid PKS–NRPS multienzyme serves to
introduce a specific amino acid unit, either to initiate[11] or to
terminate[12] the polyketide chain. However, a better model is
provided by recently characterized PKSs[13] in which the N-
terminal loading module of the PKS consists of a single ACP,
which is aminoacylated by a separate, specific acyltransferase.
In particular, the sequencing of two different gene clusters for
nitrogen-containing linear polyenes, ECO-02301 from Strep-
tomyces aizunensis[10b] and ECO-0501 from Amycolatopsis
orientalis[14] has revealed a lone ACP domain in the loading
module, as well as candidate genes for a starter unit pathway
clustered together with the polyketide synthase genes. These
genes were predicted to encode an arginine monooxygenase
(AM), a 4-guanidinobutanoate:CoA ligase (CoL) and a 4-
guanidinobutyryl-CoA:ACP acyltransferase respectively. No
4-guanidinobutyramide hydrolase (AH) candidate was
reported from either cluster (Scheme 1).
These putative precursor genes were used to screen the
complete genome of S. violaceusniger DSM 4137 for homol-
ogous sequences (Figure 1b and the Supporting Information).
The azl cluster located at 9.2 Mbp on the linear genome
houses eight PKS multienzymes (open reading frames (Orfs)
STRVN_7492 to STRVN_7499), adjacent to a ligase (CoL;
SRTVN_7500) and acyltransferase (AT; STRVN_7501), with
a candidate amide hydrolase (AH) nearby (STRVN_7510)
but no amino acid monooxygenase. The cle cluster at 3.5 Mbp
on the genome contains nine PKS multienzymes (Orfs
We next determined whether the putative cle 4-guanidi-
nobutanoyl-CoA:ACP acyltransferase (AT-2704) and its azl
counterpart (AT-7501) catalyze the reaction predicted for
them, and whether they show specificity for their own ACP
substrate. Modeling of the AT sequences (Figure S9) showed
that both are likely to adopt the fold of well-studied malonyl-
CoA:ACP acyltransferases.[16] The recombinant ACPs were
converted to the holo form by co-expression with 4’-phos-
phopantetheinyl transferase, and incubated with 4-guanidi-
nobutyryl-CoA and either of the two ATs. HPLC–MS
STRVN_2722-2730), clustered near
a
monooxygenase
(STRVN_2699), a ligase (STRVN_2700), and an acyltransfer-
ase (STRVN_2704) but no amide hydrolase. The DSM4137
genome also contains, at around 8.0 Mbp, adjacent genes
encoding AH (STRVN_6564) and AM (STRVN_6565; Fig-
ure 1b).
Angew. Chem. Int. Ed. 2013, 52, 13096 –13099
ꢀ 2013 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim
Hong, Hui
