Divergent evolution of an atypical S-adenosyl-L-methionine-dependent monooxygenase involved in anthracycline biosynthesis

Article abstract of DOI:10.1073/pnas.1501765112

Bacterial secondary metabolic pathways are responsible for the biosynthesis of thousands of bioactive natural products. Many enzymes residing in these pathways have evolved to catalyze unusual chemical transformations, which is facilitated by an evolutionary pressure promoting chemical diversity. Such divergent enzyme evolution has been observed in S-adenosyl-L-methionine (SAM)-dependent methyltransferases involved in the biosynthesis of anthracycline anticancer antibiotics; whereas DnrK from the daunorubicin pathway is a canonical 4-O-methyltransferase, the closely related RdmB (52% sequence identity) from the rhodomycin pathways is an atypical 10-hydroxylase that requires SAM, a thiol reducing agent, and molecular oxygen for activity. Here, we have used extensive chimeragenesis to gain insight into the functional differentiation of RdmB and show that insertion of a single serine residue to DnrK is sufficient for introduction of the monooxygenation activity. The crystal structure of DnrK-Ser in complex with aclacinomycin T and S-adenosyl-L-homocysteine refined to 1.9-? resolution revealed that the inserted serine S297 resides in an a-helical segment adjacent to the substrate, but in a manner where the side chain points away from the active site. Further experimental work indicated that the shift in activity is mediated by rotation of a preceding phenylalanine F296 toward the active site, which blocks a channel to the surface of the protein that is present in native DnrK. The channel is also closed in RdmB and may be important for monooxygenation in a solvent-free environment. Finally, we postulate that the hydroxylation ability of RdmB originates from a previously undetected 10-decarboxylation activity of DnrK.

Full text of DOI:10.1073/pnas.1501765112

Divergent evolution of an atypical S-adenosyl-L-
methionine–dependent monooxygenase involved
in anthracycline biosynthesis
Thadée Grocholski, Pedro Dinis, Laila Niiranen, Jarmo Niemi, and Mikko Metsä-Ketelä1
Department of Biochemistry, University of Turku, FIN-20014 Turku, Finland
Edited by Frank M. Raushel, Texas A&M University, College Station, TX, and accepted by the Editorial Board July 7, 2015 (received for review January 27, 2015)
Bacterial secondary metabolic pathways are responsible for the
biosynthesis of thousands of bioactive natural products. Many
enzymes residing in these pathways have evolved to catalyze
unusual chemical transformations, which is facilitated by an evolu-
tionary pressure promoting chemical diversity. Such divergent
enzyme evolution has been observed in S-adenosyl-L-methionine
(SAM)-dependent methyltransferases involved in the biosynthesis
of anthracycline anticancer antibiotics; whereas DnrK from the dau-
norubicin pathway is a canonical 4-O-methyltransferase, the closely
related RdmB (52% sequence identity) from the rhodomycin path-
ways is an atypical 10-hydroxylase that requires SAM, a thiol reduc-
ing agent, and molecular oxygen for activity. Here, we have used
extensive chimeragenesis to gain insight into the functional differ-
entiation of RdmB and show that insertion of a single serine residue
to DnrK is sufficient for introduction of the monooxygenation activity.
The crystal structure of DnrK-Ser in complex with aclacinomycin
T and S-adenosyl-L-homocysteine refined to 1.9-Å resolution revealed
that the inserted serine S297 resides in an α-helical segment adjacent
to the substrate, but in a manner where the side chain points away
from the active site. Further experimental work indicated that the
shift in activity is mediated by rotation of a preceding phenylalanine
F296 toward the active site, which blocks a channel to the surface of
the protein that is present in native DnrK. The channel is also closed in
RdmB and may be important for monooxygenation in a solvent-free
environment. Finally, we postulate that the hydroxylation ability of
RdmB originates from a previously undetected 10-decarboxylation
activity of DnrK.
metabolites by pathway and protein engineering. In particular,
the discovery of thousands of unknown biosynthetic pathways by
next-generation sequencing and the emergence of synthetic bi-
ology holds great promise for increasing the chemical space of
natural products (10). However, identification of the function of
gene products residing in these pathways by bioinformatic
analysis alone is complicated by the atypical chemistry the cor-
responding enzymes may catalyze. Unexpected chemical trans-
formations appear to be especially common in the biosynthesis of
aromatic polyketides, where biochemical studies have frequently
revealed unusual reaction mechanisms (2, 11), which includes
instances where the reactivity of the substrates themselves are
used in catalysis (12–14). In addition, homologous enzymes have
been noted to catalyze diverse chemical transformations (15).
Recently, the use of protein engineering for altering the reaction
specificities of biosynthetic enzymes has emerged as another
developing field for modification of natural products (16, 17).
This unusually rapid differentiation of enzyme function may
be related to the evolution of secondary metabolism, where traits
enabling a swift increase in the diversity of natural products are
beneficial (18). To promote chemical diversity rather than for-
mation of a single product, specialized metabolism enzymes ap-
pear to have evolved as generalists, which is manifested as
promiscuity and slow reaction rates typical to these proteins (19).
One intriguing example of such divergent evolution is the enzyme
pair DnrK and RdmB from the daunorubicin and rhodomycin
Significance
enzyme evolution protein engineering Streptomyces polyketide
|
|
|
|
natural products
Natural products produced by Streptomyces are widely used in
the treatment of various medical conditions. Over the years,
thousands of metabolites with complex chemical structures
have been isolated from cultures of these soil bacteria. An
evolutionary pressure that promotes chemical diversity ap-
pears to be critical for generation of this rich source of bi-
ologically active compounds. This is reflected in the bio-
synthetic enzymes, where functions of similar proteins may
greatly differ. Here, we have clarified the molecular basis of
how a classical methyltransferase has evolved into an unusual
hydroxylase on the biosynthetic pathways of two anthracy-
cline anticancer agents. Detailed understanding of enzymes
involved in antibiotic biosynthesis will facilitate future protein
engineering efforts for generation of improved bioactive na-
tural products.
nthracyclines are a medically important class of natural
products, which belong to the type II family of aromatic
A
polyketides (1–3). These compounds, produced by Streptomyces
bacteria, are among the most effective anticancer drugs available
and are currently included in 500 clinical trials worldwide (4).
Doxorubicin is widely used as a first-choice chemotherapeutic
agent for many tumors, including various carcinomas and sar-
comas (5), whereas aclacinomycin A has been used in the
treatment of hematological malignancies (6). The biological
activity of doxorubicin and aclacinomycin is mediated through
topoisomerase II (7), while additional mechanisms of these
drugs include the recently discovered histone eviction activity
(4). Despite the success of anthracyclines in cancer chemo-
therapy, improved compounds are still urgently needed, be-
cause the clinical efficacy of these metabolites has been limited
by severe side effects (8, 9).
Author contributions: T.G., J.N., and M.M.-K. designed research; T.G., P.D., L.N., and M.M.-K.
performed research; T.G., P.D., L.N., J.N., and M.M.-K. analyzed data; and M.M.-K. wrote
the paper.
The great diversity of anthracyclines is generated during tai-
loring steps in their biosynthesis, where glycosylations, hydrox-
ylations, decarboxylations, and methylations modify the common
7,8,9,10-tetrahydro-5,12-naphthacenoquinone aglycone chromo-
phore to obtain the biologically active molecules of the various
pathways (1–3). A detailed understanding of these biosynthetic
steps is essential for the rational design of improved bioactive
The authors declare no conflict of interest.
This article is a PNAS Direct Submission. F.M.R. is a guest editor invited by the Editorial Board.
Data deposition: The atomic coordinates and structure factors have been deposited in the
Protein Data Bank, www.pdb.org (PDB ID code 4WXH).
1To whom correspondence should be addressed. Email: mikko.mk@gmail.com.
This article contains supporting information online at www.pnas.org/lookup/suppl/doi:10.
1073/pnas.1501765112/-/DCSupplemental.
9866–9871
|
PNAS
|
August 11, 2015
|
vol. 112
|
no. 32
www.pnas.org/cgi/doi/10.1073/pnas.1501765112

pathways, respectively, that are homologous to classical S-adeno-
syl-L-methionine (SAM)-dependent methyltransferases.
demonstrate that the insertion of a single amino acid is sufficient
for introduction of the 10-hydroxylase activity into the DnrK
scaffold, thus providing an especially illustrative example of di-
vergent enzyme evolution. Protein crystallography and further
biochemical experiments were then conducted to reveal the
molecular basis for the altered enzymatic activity.
Methyltransferase DnrK is a true methyltransferase that cat-
alyzes the 4-O-methylation of carminomycin (1; Fig. 1) in one of
the final steps in the biosynthesis of the antitumor drug dauno-
rubicin (2; Fig. 1) in Streptomyces peucetius (20, 21). The enzyme
possesses rather relaxed substrate specificity in regard to modi-
fications in the polyaromatic anthracycline ring system, but it is
quite specific with respect to the length of the carbohydrate chain
at C-7, accepting only monoglycosides (22). In addition, DnrK has
even been shown to be able to methylate various flavonoids (23).
The crystal structure of the ternary complex of DnrK revealed that
the mechanism of methylation is consistent with canonical SN2-
type reactions of SAM-dependent methyltransferases, although
proximity and orientation effects rather than acid/base catalysis
appear to be more critical in the case of DnrK (22).
On the other hand, RdmB from the β-rhodomycin pathway in
Streptomyces purpurascens is a very unusual enzyme that is evo-
lutionarily closely related to DnrK with a sequence identity of
52% (24). Determination of the crystal structure of RdmB con-
firmed that the enzyme binds SAM (25) and that the ternary
complex is indeed very similar to that of DnrK with a root-mean-
square deviation of 1.14 Å for 335 equivalent Cα atoms (26).
Despite these features, RdmB completely lacks methyltransferase
activity (24, 27, 28); it is an anthracycline 10-hydroxylase, which
requires SAM, molecular oxygen, and a thiol reducing agent
for activity (26). Contrary to DnrK, RdmB is able to use both
monoglycosylated and triglycosylated anthracyclines as substrates.
Furthermore, both rhodomycin and daunorubicin gene clus-
ters encode homologous 15-methylesterases RdmC (28) and
DnrP (21), respectively. On the rhodomycin pathway, RdmB
works strictly in conjunction with RdmC, which first converts
e-rhodomycin (3; Fig. 1) into 15-demethoxy-e-rhodomycin (4; Fig.
1), whereas RdmB is responsible for conversion of the RdmC
reaction product into β-rhodomycin (5; Fig. 1) (28). This is in
contrast to the daunorubicin pathway, where the reaction order of
the two enzymes DnrP and DnrK is more relaxed, as DnrK is able
to methylate at least five biosynthetic intermediates (21).
Results and Discussion
Identification of Protein Regions for Chimeragenesis. Structural studies
of DnrK and RdmB (22, 25) have confirmed that both enzymes
consist of an N-terminal dimerization domain, an α-helical middle
domain, and a C-terminal catalytic domain that binds the cofactor
SAM (Fig. 1). The substrate binding site is formed between the
C-terminal domain, which has a Rossmann-like α/β-fold typical to
nucleotide-binding proteins, and the middle domain. As an initial
experiment, we fused the dimerization domain of DnrK onto the
catalytic domain of RdmB, generating the enzyme variant RdmB-CT.
Because the activity of RdmB-CT was not altered and the enzyme
catalyzed exclusively 10-hydroxylation (Fig. 2), we were able to
focus on amino acid residues residing in the catalytic subunit in
subsequent experiments. In the absence of a high-throughput
screening system, we decided to create chimeric enzymes by inter-
changing key subdomain regions ranging from 10 to 12 aa to probe
the functional differentiation of the enzyme pair.
Comparison of the two enzymes revealed that most of the amino
acid residues residing in the vicinity of the binding site for SAM
were highly conserved and that differences could be mainly attrib-
uted to three key regions around the putative entrance to the active
site cavity (Fig. 1). The first region (R1; Fig. S1) selected to the
study resides near the C-10 position of the polyaromatic anthracy-
cline ligand and consists of a loop region and the N-terminal half of
α16 from the middle domain (Fig. 1). This loop region is disordered
in the RdmB–SAM binary complex (25), whereas it is well de-
fined in the ternary complex containing 11-deoxy-β-rhodomycin
A (6; Fig. 1) (26). In addition, in comparison with DnrK (22), the
loop region has moved closer to the ligand in RdmB, suggesting
that R1 may be important for the 10-hydroxylation activity and/or
the entry of the ligand into the active site. Second, moderate
differences can be observed in the loop region between β8 and β9
(R2; Fig. S1), which folds over the ligands (Fig. 1), and as such
might be responsible for assuring the correct orientation of
the substrate. Finally, the region between α11 and α12 defines
a restricted space for binding of the carbohydrate unit of
To gain insight into how a methyltransferase has evolved into a
monooxygenase, we report here the generation of several chi-
meric DnrK/RdmB proteins that were designed based on the
available crystal structures. These initial experiments guided us
in further structure/function studies, where we were able to
Fig. 1. Similar structures, distinct functions. The structures and reactions catalyzed by (A) DnrK and (B) RdmB. In the overall structures, the N-terminal di-
merization, the middle, and the catalytic domains are shown in yellow, green, and red, respectively. In the Inset, the catalytic domains are color coded based
on dissimilar (black), similar (gray), and identical (white) residues. SAH and the ligands are shown in red, whereas the three regions selected for chimera-
genesis (R1, R2, and R3) are shown in orange. The R3 region of RdmB is disordered in the crystal structure and is shown as a dashed orange line in the Inset. In
the reaction scheme, “R” represents a thiol reducing agent such as DTT or glutathione.
Grocholski et al.
PNAS
|
August 11, 2015
|
vol. 112
|
no. 32
|
9867

4-methoxy-e-rhodomycin T (7; Fig. 1) in the DnrK structure (Fig. corresponding region in RdmB is rather different and the carbo-
1), which has been noted as the likely explanation for why the hydrate chain extends from the binding pocket into the bulk
solution in the RdmB ternary complex (26), which prompted
us to select this region (R3; Fig. S1) as the third segment for
chimeragenesis.
enzyme only accepts monoglycosylated substrates (22). The
Enzymatic Activities of the Native Enzymes. The activity assays were
performed in a coupled reaction together with the 15-methyl-
esterase RdmC using either the monoglycosylated aclacinomycin
T (8; Fig. 2) or the triglycosylated aclacinomycin A (9; Fig. 2) as
substrates. The reaction mixtures also included SAM, which is
essential as a cosubstrate for DnrK (22) and as a cofactor for
RdmB (26), and a reducing agent (DTT) that is required solely
by RdmB (26). The reaction with native DnrK resulted in a
single main product (Fig. 2A), which was identified as the 4-O-
methylated and 10-decarboxylated aclacinomycin T (10; Figs. S2
and S3, and Table S1), when 8 was used as a substrate, whereas
no methylation activity could be observed with the triglycosylated
9. In agreement with previous experiments (27), addition of na-
tive RdmB to the assays with 8 and 9 led to the accumulation of
the expected 10-hydroxylated products 11-deoxy-β-rhodomycin T
(11; Fig. 2B) and 6 (Fig. 2D), respectively.
Discovery of a Moonlighting 10-Decarboxylation Activity of DnrK.
The RdmC and DnrP reaction products with a free 10-carboxyl
group (compounds such as 4 in Fig. 1) have been described as
unstable and undergo spontaneous 10-decarboxylation on the
rhodomycin and daunorubicin pathways, respectively (21, 29).
However, indirect evidence has suggested that the decarboxyl-
ation might also be enzymatically catalyzed by DnrP (21). During
our assays with 8, we noted that the RdmC reaction product, 15-
demethoxy-aclacinomycin T (15; Fig. 2) is stable if the enzymatic
reactions are conducted in the dark, but exposure to light is
sufficient to promote the 10-decarboxylation reaction to proceed
to near completion in 60 min resulting in 10-decarboxymethyl-
aclacinomycin T (13; Fig. S3). To our surprise, we next observed
that if 15 is isolated and incubated in the dark together with
DnrK, both 4-O-methylation and 10-decarboxylation reactions
occur, because 10 is observed as the sole product (Fig. S3). The
result therefore demonstrated directly that it is DnrK instead of
DnrP that harbors the moonlighting 10-decarboxylation activity,
which in our view has important evolutionary implications for the
emergence of 10-hydroxylation activity (see below).
Enzymatic Activities of the Chimeric Proteins. The single chimeric
DnrK enzymes, where the amino acid residues in the R1, R2, or R3
regions were changed to correspond to RdmB, behaved much like
native DnrK in the activity measurements and led to the accu-
mulation of 10 (Fig. 2A) as the main product. However, the re-
action mixture of DnrK R1 contained a small amount of the
expected RdmB reaction product, 11 (Fig. 2B), and significant
quantities of a compound that was identified as the double DnrK/
RdmB reaction product, 4-O-methyl-11-deoxy-β-rhodomycin T
(12; Fig. 2C). Neither DnrK nor RdmB alone produced 12, but
when both native enzymes were incubated together with RdmC,
the same metabolite was observed (Fig. 2C). The double and triple
chimeric enzymes reinforced our initial results; chimeric proteins
that contained the R1 region were able to catalyze 10-hydroxyl-
ation, whereas the R2 + R3 chimera was only able to perform 4-O-
methylation. These multiple exchanges enhanced the RdmB-like
activity in comparison with the single R1 chimera, with the R3
region having a slightly greater effect than the R2 region (Fig. 2B).
A similar trend was observed when 9 was used as a substrate
and the 10-hydroxylated reaction product 6 was observed only in
chimeras containing the R1 region from RdmB. It is noteworthy
that the increased rate of 10-hydroxylation correlated exceed-
ingly well with the number of amino acid exchanges from the
RdmB template (Fig. 2D), which highlights the importance of
Fig. 2. Relative activities of native and engineered enzymes in a coupled re-
action with RdmC. The reactions with 8 lead to the accumulation of three
products (A–C), whereas in the case of 9 only one product was observed (D). The
columns present the formation of (A) 4-O-methyl-15-decarboxyaclacinomycin
T (10), (B) 11-deoxy-β-rhodomycin T (11), (C) 4-O-methyl-11-deoxy-β-rhodomycin
T (12), and (D) 11-deoxy-β-rhodomycin A (6) by the enzymes. The overall per-
centages may not add up to 100% in all samples because some unreacted
substrate 15 was left with chimeras harboring poor activity (e.g., DnrK R2 + R3).
The excess substrate was converted to the corresponding 10-decarboxylated
compounds by exposure to light (overall conversion of 8 to 13 and 9 to 14; Fig.
S2) to facilitate the HPLC analyses.
9868
|
www.pnas.org/cgi/doi/10.1073/pnas.1501765112
Grocholski et al.

Fig. 3. Cartoon (Top) and surface (Bottom) repre-
sentations of the active sites of DnrK in complex with
4-methoxy-e-rhodomycin T (7) and SAH, DnrK-Ser with
bound 8 and SAH, and the ternary complex of RdmB,
11-deoxy-β-rhodomycin (6), and SAM. Insertion of
S297 in DnrK-Ser rotates F296 toward the ligand,
which reshapes the DnrK active site toward the RdmB
architecture. In particular, DnrK-Ser F296 blocks a
channel to the surface of the protein that is open in
native DnrK, but closed in RdmB. A conserved arginine
residue that is important for both methylation and
hydroxylation activities is also shown.
the R2 and R3 regions in correct alignment of the triglycosidic 9 exception of two amino acids residing in loop regions (D85 and
substrate in the active site. Previous studies have shown that the
R3 region of DnrK folds around the amino sugar moiety (22),
and it was surprising to find that exchange of the R3 region was
not essential for the gain of the 10-hydroxylation activity. Despite
these changes, no 4-O-methylation activity was detected with this
substrate in any of the chimeric enzymes, which would indicate
that the aglycone is still not correctly positioned in terms of
geometry in the active-site cavity. The lack of methylation ac-
tivity in native RdmB has been suggested to occur due to wrong
alignment of the substrate for a methyl transfer reaction using a
SN2 mechanism, where the methyl group has to be in line with
the oxygen of the substrate as well as the sulfur atom of SAM for
the reaction to proceed (26).
H293) in chain B, where the density in general was less well
defined than in chain A. Based on electron density, one molecule
of 8 and S-adenosyl-L-homocysteine (SAH) was bound to each
subunit (Fig. S4).
The overall structure of DnrK-Ser was exceedingly similar to
native DnrK with an r.m.s.d. of 0.90 Å over 327 residues, and the
main differences were mostly restricted to the R1 region. In
short, introduction of the additional S297 residue in R1.2 ap-
peared to trigger a transition of all preceding residues in helix
α16 toward the surface of the protein, which altered the structure
of the R1.1 loop region. Unexpectedly, the side chain of the
inserted serine was found pointing away from the active site, but
at the same time the preceding phenylalanine (F296) had rotated
inward adjacent to the ligand near the site of monooxygenation
(Fig. 3). Comparison of the structures of DnrK-Ser and RdmB
reveals that both enzymes have a phenylalanine, F296 and F300,
respectively, next to the anthracycline ligand, which is in contrast
to native DnrK, which contains glutamine Q295 at the equivalent
position (Fig. 3). The implication of these changes is that the
glutamine in DnrK allows bulk solvent to access the active site
cavity via an open channel, whereas the bulkier hydrophobic
phenylalanine residues block this route to the surface of the
protein in DnrK-Ser and RdmB (Fig. 3). The structural analysis
therefore suggested that critical to the shift in activity is rotation
of F296 and that the amino acid inserted at position 297 would
not necessarily have to be serine. Consistent with this hypothesis,
reintroduction of the original phenylalanine observed in this
structural position in native DnrK led to hydroxylation of both
substrates by DnrK R1.2 S297F (Fig. 2 B and D).
In contrast to the engineered DnrK enzymes, when the equiv-
alent changes were made to the RdmB template, no methylation
activity could be detected. The chimeric RdmB enzymes either
behaved in a manner similar to the native enzyme or simply lost all
enzymatic activity.
Mapping the Minimal Region Required for Gain of 10-Hydroxylation
Activity in DnrK. In an attempt to further pinpoint the origin of the
10-hydroxylation activity in the DnrK R1 chimera, we divided the
R1 region into two segments corresponding to the loop region
and the following α16 helix, which resulted in two additional
mutants denoted as DnrK R1.1 and DnrK R1.2, respectively.
The 10-hydroxylation activity could be attributed solely to the
α16 helix (Fig. 2), because the activity of DnrK R1.2 was similar
to that of DnrK R1 (i.e., both methylation and hydroxylation of 8
and hydroxylation of 9), whereas DnrK R1.1 behaved like native
DnrK (i.e., methylation of 8 and no activity with 9). Inspection of
the amino acid sequences of the R1.2 region (Fig. S1) revealed
that the RdmB sequence contains an additional serine insertion
in this area in comparison with DnrK. Remarkably, introduction
of this single S297 residue to the DnrK scaffold was sufficient to
persuade the 10-hydroxylation activity to emerge as seen from
the activity measurements of the DnrK-Ser variant (Fig. 2).
Mechanistic Divergence of the DnrK and RdmB Reactions. The chi-
meragenesis studies of DnrK and RdmB permit us to propose
the following mechanistic proposal for the divergent reactions
catalyzed by the two enzymes (Fig. 4). It is tempting to suggest
that a common initial step is the decarboxylation of the substrate
15 to generate a carbanion intermediate. The more open active
site of DnrK allows facile protonation of the carbanion by sol-
vent molecules and results in the formation of a neutral 10-
decarboxylated and 4-O-methylated anthracycline 10. However,
such compounds are no longer suitable as substrates for RdmB
Structural Basis for the Altered Enzymatic Activity of the DnrK-Ser
Mutant. The DnrK-Ser mutant crystallized in a different space
group than native DnrK, and consequently the diffraction data
could be refined to an improved resolution of 1.9 Å (Table S2). (26), which avoids the protonation step by closing the channel to
Electron density could be traced for most of the polypeptide
chain in the two subunits found in the asymmetric unit with the
the surface of the protein and excluding solvent ions from the
active site. The carbanion may be stabilized through distribution
Grocholski et al.
PNAS
|
August 11, 2015
|
vol. 112
|
no. 32
|
9869

hydroxylation may be observed from analysis of eight gene
clusters putatively involved in anthracycline biosynthesis (Fig.
S5). In all cases, the gene order is conserved, indicating possible
coevolution of the proteins. In our view, the ancestral config-
uration may have been a daunorubicin-type system where one
enzyme catalyzes anthracycline 15-demethylation, whereas an-
other protein promotes 4-O-methylation and 10-decarboxylation
under light-deprived conditions in the soil. The 10-hydroxylation
activity may then have evolved from the moonlighting 10-
decarboxylation activity upon changes in the R1 region leading
to generation of more diversity in anthracycline scaffolds.
Further changes in other parts of the protein have then led to
full conversion of the enzymatic activity.
Our hypothesis is supported by sequence analysis of the R1
area, which indicates that the region is an evolutionary hot spot
in these enzymes (Fig. 5). Three sequences, including RdmB, are
rhodomycin-type with a serine insertion, whereas DnrK appears
to be the sole daunorubicin-type sequence. Interestingly, three
of the proteins of unknown function contain an insertion of
“TSDLY” where the tyrosine may be positioned in an equivalent
manner to the phenylalanine in DnrK-Ser, and these proteins
may represent an alternative evolutionary path to 10-hydroxylated
anthracyclines. The final sequence is even more distantly related
with an insertion of “TADLH” with a histidine at the critical
position, making prediction of the activity of this protein chal-
lenging in silico. Investigations into this previously unidentified
group of putative anthracycline methyltransferases are currently
in progress in our laboratory.
Fig. 4. Mechanistic proposal for the reactions catalyzed by DnrK and RdmB.
Both DnrK and RdmB are postulated to catalyze the decarboxylation of 15 as a
first step. In DnrK, the open active site would allow protonation of the
resulting carbanion by solvent molecules and 4-O-methylation to generate 10.
In contrast, RdmB stabilizes the carbanion and catalyzes the formation of an
O2–anthracycline caged radical pair in the closed active site. Consequent for-
mation of a peroxyl intermediate and subsequent reduction of the peroxide by
intracellular thiols possibly outside the active site of the enzyme yields the 10-
hydroxylated product 11. The reaction sequence leading to the double-prod-
uct 12 is only possible if the RdmB reaction happens first. R1, L-rhodosamine.
Materials and Methods
General DNA Techniques. Plasmids containing native dnrK and rdmB (22, 25)
were used as templates for all of the PCRs. DnrK R1.2 S297F, R303Q, and
R303K mutants were ordered as synthetic genes from GeneArt (Strings DNA
Fragments). The remaining DnrK mutants and RdmB-CT were generated
using the four-primer overhang extension method (32). Details regarding
the cloning experiments, including generation of the RdmC expression
construct, are presented in SI Materials and Methods. Oligonucleotides used
are listed in Table S3. The gene encoding RdmB-CT was cloned into the
pBAD/His B vector (Invitrogen), whereas all other enzymes were expressed
from the modified pBHBΔ vector (33). All DNA sequences were confirmed by
sequencing before protein expression.
of the negative charge throughout the polyphenolic ring system
and the adjacent positive charge of SAM, which is the reason
why the enzyme requires the cofactor for catalysis (26). The
delocalization of electrons enables substrate-assisted activation
of molecular oxygen to overcome the spin-barrier of dioxygen
reactivity in manner similar to other cofactor-independent
monooxygenases (12–14, 30, 31). The exclusion of solvent ions
from the active site may also be essential for this step. Once the
carbanion has reacted with dioxygen, the 10-peroxy anthracycline
formed is reduced by a thiol reducing reagent possibly outside
the active-site cavity to generate the 10-hydroxylated end prod-
uct 11 (26). It is important to note that formation of the DnrK/
RdmB double-reaction product 12 must proceed through initial
10-hydroxylation followed by 4-O-methylation (Fig. 4), because
neutral anthracyclines are suitable substrates only for DnrK (21).
It has been proposed that arginine R307 near the R1 region in
RdmB (Fig. 3), which is also conserved in DnrK (R302), has an
important role for the initial decarboxylation of the substrate (26).
The discovery of the 10-decarboxylation activity for DnrK prompted
us to study the importance of this residue for the activity of DnrK
R1.2. Expectedly, the mutants DnrK R1.2 R303K and R303Q
displayed greatly reduced activities for both 4-O-methylation and
10-hydroxylation. The 10-hydroxylation activity was completely lost
in the R303Q mutant, whereas only trace activities remained in the
R303K mutant when 8 was used as a substrate (Fig. 2). The 4-O-
methylation activity was also affected, although both mutants still
harbored ∼10% of their activities. In addition, changing the argi-
nine R302 to isoleucine in DnrK has been described to greatly re-
duce the methylation activity of flavonoids (23), confirming the
importance of this residue for activity.
Protein Expression and Purification. Escherichia coli cell strain and expression
conditions were performed as previously described (34), except LB media
was used and cells were collected 5 h after induction. Purification, yields, and
storage of all enzymes are described in SI Materials and Methods.
Enzyme Activity Measurements. All activity measurements were performed in
100 mM Tris·HCl (pH 7.5), 10 mM DTT, and 350 μM SAM with 140 μM 8 or 9.
The concentration of RdmC was set to 100 nM in the reactions, whereas all
other enzymes were used at 5.8 μM. Further details are described in SI
Materials and Methods.
Fig. 5. A phylogenetic tree of eight putative SAM-dependent methyl-
transferases residing in anthracycline biosynthesis gene clusters and multiple
sequence alignment of the R1 area. Amino acids marked with an arrow in
the blue column are predicted to be situated at the critical position near C-10
of the anthracycline ligands. Further details regarding these unknown gene
clusters and their accession numbers are shown in Fig. S5.
Evolutionary Implications. The intimate connection between
the 15-methylesterase reaction to both 4-O-methylation and 10-
9870
|
www.pnas.org/cgi/doi/10.1073/pnas.1501765112
Grocholski et al.

Analysis of Metabolites. The various reaction products were analyzed and
identified as described in SI Materials and Methods, Fig. S2, and Tables S1 and S4.
All reactions were monitored by HPLC using a SCL-10Avp/SpdM10Avp system
with a diode array detector (Shimadzu). High-resolution MS (micrOTOF-Q;
Bruker Daltonics; linked to an Agilent Technologies 1200 series HPLC system)
was used for confirmation of the molecular formulae.
Synchrotron Radiation Facility (ESRF) (Grenoble, France) at beamline ID23-1.
Details of the data collection and refinement statistics are listed in Table S2.
Figures depicting protein structures were prepared using PyMOL (The
PyMOL Molecular Graphics System, version 1.3; Schrödinger, LLC).
ACKNOWLEDGMENTS. We thank Elina Taipalus and Jonna Pörsti for assis-
tance with protein crystallization, and Vilja Siitonen, MSc, for the high-resolution
mass spectrometry measurements. We acknowledge access to synchrotron radi-
ation at ESRF (Grenoble, France). This study was supported by Academy of Fin-
land Grant 136060 (to M.M.-K.).
Crystallization, Data Collection, and Structure Determination of DnrK-Ser.
Crystallization and structure determination of DnrK-Ser is described in SI
Materials and Methods. Diffraction data were collected at the European
1. Metsä-Ketelä M, Niemi J, Mäntsälä P, Schneider G (2008) Anthracycline biosynthesis:
Genes, enzymes and mechanisms. Anthracycline Chemistry and Biology I: Biological
Occurence and Biosynthesis, Synthesis and Chemistry, ed Krohn K (Springer, Berlin),
pp 101–140.
2. Hertweck C, Luzhetskyy A, Rebets Y, Bechthold A (2007) Type II polyketide synthases:
Gaining a deeper insight into enzymatic teamwork. Nat Prod Rep 24(1):162–190.
3. Hutchinson CR (1997) Biosynthetic studies of daunorubicin and tetracenomycin C.
Chem Rev 97(7):2525–2536.
4. Pang B, et al. (2013) Drug-induced histone eviction from open chromatin contributes
to the chemotherapeutic effects of doxorubicin. Nat Commun 4:1908.
5. Weiss RB (1992) The anthracyclines: Will we ever find a better doxorubicin? Semin
Oncol 19(6):670–686.
26. Jansson A, et al. (2005) Aclacinomycin 10-hydroxylase is a novel substrate-assisted
hydroxylase requiring S-adenosyl-L-methionine as cofactor.
3636–3644.
J Biol Chem 280(5):
27. Wang Y, et al. (2000) Modifications of aclacinomycin T by aclacinomycin methyl es-
terase (RdmC) and aclacinomycin-10-hydroxylase (RdmB) from Streptomyces purpur-
ascens. Biochim Biophys Acta 1480(1-2):191–200.
28. Wang Y, Niemi J, Mäntsälä P (2002) Modification of aklavinone and aclacinomycins
in vitro and in vivo by rhodomycin biosynthesis gene products. FEMS Microbiol Lett
208(1):117–122.
29. Jansson A, Niemi J, Mäntsälä P, Schneider G (2003) Crystal structure of aclacinomycin
methylesterase with bound product analogues: Implications for anthracycline rec-
ognition and mechanism. J Biol Chem 278(40):39006–39013.
6. Wei G, et al. (2011) A meta-analysis of CAG (cytarabine, aclarubicin, G-CSF) regimen
for the treatment of 1029 patients with acute myeloid leukemia and myelodysplastic
syndrome. J Hematol Oncol 4:46.
7. Nitiss JL (2009) Targeting DNA topoisomerase II in cancer chemotherapy. Nat Rev
Cancer 9(5):338–350.
8. Singal PK, Iliskovic N (1998) Doxorubicin-induced cardiomyopathy. N Engl J Med
339(13):900–905.
9. Kaye S, Merry S (1985) Tumour cell resistance to anthracyclines—a review. Cancer
Chemother Pharmacol 14(2):96–103.
30. Thierbach S, et al. (2014) Substrate-assisted O2 activation in a cofactor-independent
dioxygenase. Chem Biol 21(2):217–225.
31. Fetzner S, Steiner RA (2010) Cofactor-independent oxidases and oxygenases. Appl
Microbiol Biotechnol 86(3):791–804.
32. Ho SN, Hunt HD, Horton RM, Pullen JK, Pease LR (1989) Site-directed mutagenesis by
overlap extension using the polymerase chain reaction. Gene 77(1):51–59.
33. Kallio P, Sultana A, Niemi J, Mäntsälä P, Schneider G (2006) Crystal structure of the
polyketide cyclase AknH with bound substrate and product analogue: Implications for
catalytic mechanism and product stereoselectivity. J Mol Biol 357(1):210–220.
34. Koskiniemi H, Grocholski T, Schneider G, Niemi J (2009) Expression, purification and
crystallization of the cofactor-independent monooxygenase SnoaB from the nogala-
mycin biosynthetic pathway. Acta Crystallogr Sect F Struct Biol Cryst Commun 65(Pt 3):
256–259.
35. Ylihonko K, Hakala J, Niemi J, Lundell J, Mäntsälä P (1994) Isolation and character-
ization of aclacinomycin A-non-producing Streptomyces galilaeus (ATCC 31615) mu-
tants. Microbiology 140(Pt 6):1359–1365.
36. Niemi J, et al. (1994) Hybrid anthracycline antibiotics: Production of new anthracy-
clines by cloned genes from Streptomyces purpurascens in Streptomyces galilaeus.
Microbiology 140(Pt 6):1351–1358.
37. Kunnari T, Niemi J, Ylihonko K, Mäntsälä P, Hakala J (1997) Hybrid anthracyclines by
a genetically engineered Streptomyces galilaeus mutant. Bioorg Med Chem Lett 7(6):
725–726.
10. Mitchell W (2011) Natural products from synthetic biology. Curr Opin Chem Biol 15(4):
505–515.
11. Metsä-Ketelä M, Oja T, Taguchi T, Okamoto S, Ichinose K (2013) Biosynthesis of
pyranonaphthoquinone polyketides reveals diverse strategies for enzymatic carbon-
carbon bond formation. Curr Opin Chem Biol 17(4):562–570.
12. Oja T, et al. (2012) Biosynthetic pathway toward carbohydrate-like moieties of al-
numycins contains unusual steps for C-C bond formation and cleavage. Proc Natl Acad
Sci USA 109(16):6024–6029.
13. Grocholski T, et al. (2010) Crystal structure of the cofactor-independent mono-
oxygenase SnoaB from Streptomyces nogalater: Implications for the reaction mech-
anism. Biochemistry 49(5):934–944.
14. Siitonen V, Blauenburg B, Kallio P, Mäntsälä P, Metsä-Ketelä M (2012) Discovery of
a
two-component monooxygenase SnoaW/SnoaL2 involved in nogalamycin bio-
synthesis. Chem Biol 19(5):638–646.
38. Gasteiger E, et al. (2005) Protein identification and analysis tools on the ExPASy
server. The Proteomics Protocols Handbook, ed Walker JM (Humana, Totowa, NJ), pp
571–607.
39. Kabsch W (2010) Integration, scaling, space-group assignment and post-refinement.
Acta Crystallogr D Biol Crystallogr 66(Pt 2):133–144.
40. Collaborative Computational Project, Number 4 (1994) The CCP4 suite: Programs for
protein crystallography. Acta Crystallogr D Biol Crystallogr 50(Pt 5):760–763.
41. Long F, Vagin AA, Young P, Murshudov GN (2008) BALBES: A molecular-replacement
pipeline. Acta Crystallogr D Biol Crystallogr 64(Pt 1):125–132.
42. Langer G, Cohen SX, Lamzin VS, Perrakis A (2008) Automated macromolecular model
building for X-ray crystallography using ARP/wARP version 7. Nat Protoc 3(7):
1171–1179.
43. Murshudov GN, Vagin AA, Dodson EJ (1997) Refinement of macromolecular struc-
tures by the maximum-likelihood method. Acta Crystallogr D Biol Crystallogr 53(Pt 3):
240–255.
44. Schüttelkopf AW, van Aalten DMF (2004) PRODRG: A tool for high-throughput crys-
tallography of protein-ligand complexes. Acta Crystallogr D Biol Crystallogr 60(Pt 8):
1355–1363.
15. Patrikainen P, et al. (2012) Tailoring enzymes involved in the biosynthesis of angu-
cyclines contain latent context-dependent catalytic activities. Chem Biol 19(5):
647–655.
16. Williams GJ, Gantt RW, Thorson JS (2008) The impact of enzyme engineering upon
natural product glycodiversification. Curr Opin Chem Biol 12(5):556–564.
17. Zabala AO, Cacho RA, Tang Y (2012) Protein engineering towards natural product
synthesis and diversification. J Ind Microbiol Biotechnol 39(2):227–241.
18. Firn RD, Jones CG (2000) The evolution of secondary metabolism—a unifying model.
Mol Microbiol 37(5):989–994.
19. Bar-Even A, Salah Tawfik D (2013) Engineering specialized metabolic pathways—is
there a room for enzyme improvements? Curr Opin Biotechnol 24(2):310–319.
20. Madduri K, Torti F, Colombo AL, Hutchinson CR (1993) Cloning and sequencing of a
gene encoding carminomycin 4-O-methyltransferase from Streptomyces peucetius
and its expression in Escherichia coli. J Bacteriol 175(12):3900–3904.
21. Dickens ML, Priestley ND, Strohl WR (1997) In vivo and in vitro bioconversion of
e-rhodomycinone glycoside to doxorubicin: Functions of DauP, DauK, and DoxA.
J Bacteriol 179(8):2641–2650.
22. Jansson A, Koskiniemi H, Mäntsälä P, Niemi J, Schneider G (2004) Crystal structure of a
ternary complex of DnrK, a methyltransferase in daunorubicin biosynthesis, with
bound products. J Biol Chem 279(39):41149–41156.
23. Kim NY, et al. (2007) O-Methylation of flavonoids using DnrK based on molecular
docking. J Microbiol Biotechnol 17(12):1991–1995.
45. Emsley P, Lohkamp B, Scott WG, Cowtan K (2010) Features and development of Coot.
Acta Crystallogr D Biol Crystallogr 66(Pt 4):486–501.
46. Afonine PV, et al. (2012) Towards automated crystallographic structure refinement
with phenix.refine. Acta Crystallogr D Biol Crystallogr 68(Pt 4):352–367.
47. Chen VB, et al. (2010) MolProbity: All-atom structure validation for macromolecular
crystallography. Acta Crystallogr D Biol Crystallogr 66(Pt 1):12–21.
48. Sultana A, et al. (2004) Structure of the polyketide cyclase SnoaL reveals a novel
mechanism for enzymatic aldol condensation. EMBO J 23(9):1911–1921.
49. Hutchinson CR, Colombo AL (1999) Genetic engineering of doxorubicin production in
Streptomyces peucetius: A review. J Ind Microbiol Biotechnol 23(1):647–652.
50. Sullivan MJ, Petty NK, Beatson SA (2011) Easyfig: A genome comparison visualizer.
Bioinformatics 27(7):1009–1010.
24. Niemi J, Mäntsälä
P (1995) Nucleotide sequences and expression of genes from
Streptomyces purpurascens that cause the production of new anthracyclines in
Streptomyces galilaeus. J Bacteriol 177(10):2942–2945.
25. Jansson A, Niemi J, Lindqvist Y, Mäntsälä P, Schneider G (2003) Crystal structure of
aclacinomycin-10-hydroxylase, a S-adenosyl-L-methionine-dependent methyltransferase
homolog involved in anthracycline biosynthesis in Streptomyces purpurascens. J Mol Biol
334(2):269–280.
Grocholski et al.
PNAS
|
August 11, 2015
|
vol. 112
|
no. 32
|
9871