You are on page 1of 10

NATURE CHEMICAL BIOLOGY | VOL 7 | NOVEMBER 2011 | www.nature.

com/naturechemicalbiology 843
ARTICLE
PUBLISHED ONLINE: 9 OCTOBER 2011 | DOI: 10.1038/NCHEMBIO.671
A
minoglycosides act by binding to the bacterial ribosome,
thereby inhibiting protein synthesis and generating errors in
the translation of the genetic code
1
. This class of antibiotics
has been widely used to treat severe bacterial infections for decades
2
,
beginning with the use of the first effective antituberculosis agent,
streptomycin
3
. Aminoglycosides are defined by two structural
classes: kanamycins (Fig. 1), isolated from Streptomyces kanamy-
ceticus
4
, are representative 4,6-disubstituted 2-deoxystreptamine
(1)-containing aminoglycoside antibiotics along with the genta-
micins
5
and tobramycin
6
, whereas the neomycins
7
and butirosins
8

exemplify the 4,5-disubstituted aminoglycosides (Supplementary
Results, Supplementary Fig. 1). As with other antibiotics, the
emergence of aminoglycoside-resistant pathogens raises serious
problems
1,9
. The issue of resistance has been addressed by the semi-
synthetic modification of natural aminoglycosides such as amikacin,
dibekacin or arbekacin (Supplementary Fig. 1)
10
. A detailed under-
standing of the biosynthetic pathway of aminoglycosides is a pre-
requisite for not only the generation of more robust antibiotic agents
but also the direct fermentative production of clinically important
antibiotics. However, although the biochemistry and genetics
underlying the biosynthesis of 4,5-disubstituted aminoglycosides
have been relatively well characterized
11
, those of 4,6- disubstituted
aminoglycosides, including kanamycin, still remain unclear, mainly
because of difficulties in genetically manipulating the producing
actinomycetes and obtaining soluble functional enzymes.
The core moiety 1 and the pseudodisaccharides 2-N-
acetylparomamine (2), paromamine (3) and neamine (4) are com-
mon biosynthetic intermediates of most aminoglycosides, including
kanamycin. The biosynthetic route to 4 via 1 and 3 from D-glucose-6-
phosphate has been elucidated in the context of butirosin and neomycin
biosynthesis using purified recombinant enzymes, with the exception
of the glycosylation step
11
, which has been partially characterized via
either in vivo gene disruption or the study of cell-free extracts (CFEs)
(Fig. 1). In particular, 1 is biosynthesized from D-glucose-6-phosphate
by 2-deoxy-scyllo-inosose (2-DOI) synthase
12
, a dual-functional
L-glutamine:2-DOI aminotransferase
13
, and 2-deoxy-scyllo-inosamine
(2-DOIA) dehydrogenase
14
. Next, N-acetylglucosaminyltransferase
transfers D-2-N-acetylglucosamine (GlcNAc) onto 1 to give 2
(refs. 1517), which is deacetylated to 3 by 2-N-acetylparomamine
deacetylase
18
. We have also shown that the recombinant KanA acts as
a 2-DOI synthase
19
, that kanB encodes a dual-functional aminotrans-
ferase by interspecies complementation in the neomycin producer
Streptomyces fradiae
20
, and that kanA-kanB-kanK-kanF and kacA
constitute a minimal gene set for the biosynthesis of 3 by heterolo-
gous expression in Streptomyces lividans
21
. Most recently, tests of the
CFE of Escherichia coli expressing kanE
22
indicated that this glyco-
syltransferase (otherwise known as KanM2) catalyzes the glucosyla-
tion of 3 to give 3-deamino-3-hydroxykanamycin C (5), suggesting
that 5 is a precursor for kanamycin C (6). Other steps in the pathway
have been predicted totally on the basis of the metabolites isolated
from culture extracts and isotope-labeled precursor incorporation
studies
23,24
. For example, 6 and kanamycin B (7) are supposed to
be formed by the transfer of a kanosamine (D-3-glucosamine, Kns)
moiety onto 3 and 4, respectively, by the glycosyltransferase KanE
25
.
Additionally, kanamycin A (8) has been long believed to be pro-
duced by the deamination of the 2 position of 7 (refs. 2527) (Fig. 1).
However, as this proposed biosynthetic route lacks substantial genetic
and biochemical evidence
15,2527
, the true biosynthetic pathway to
kanamycin products still awaits discovery.
Recently, we isolated the kanamycin biosynthetic gene cluster
from S. kanamyceticus
28
. The production of 8 by expression of this
entire cluster in a nonaminoglycoside-producing S. venezuelae
indicated that it contains all of the genes sufficient for the biosynthe-
sis of the kanamycin complex
29
. We now report further exploration
of this gene cluster, which allowed us to completely decipher the
biosynthetic pathway of the kanamycins, revealing both previously
undescribed kanamycin biosynthetic intermediates and a remark-
able parallel pathway leading to kanamycins A and B independently.
1
Department of Chemistry and Nanoscience, Ewha Womans University, Seoul, Republic of Korea.
2
Department of Pharmaceutical Engineering, Institute of
Biomolecule Reconstruction, Sun Moon University, Chungnam, Republic of Korea.
3
Interdisciplinary Program of Biochemical Engineering and Biotechnology,
Seoul National University, Seoul, Republic of Korea.
4
Systems Microbiology Research Center, Korea Research Institute of Bioscience and Biotechnology,
Daejeon, Republic of Korea.
5
These authors contributed equally to this work. *e-mail: sohng@sunmoon.ac.kr or joonyoon@ewha.ac.kr.
Discovery of parallel pathways of kanamycin
biosynthesis allows antibiotic manipulation
Je Won Park
1,2,5
, Sung Ryeol Park
1,5
, Keshav Kumar Nepal
2
, Ah Reum Han
3
, Yeon Hee Ban
1
,
Young Ji Yoo
1
, Eun Ji Kim
1
, Eui Min Kim
2
, Dooil Kim
4
, Jae Kyung Sohng
2
* & Yeo Joon Yoon
1
*
Kanamycin is one of the most widely used antibiotics, yet its biosynthetic pathway remains unclear. Current proposals suggest
that the kanamycin biosynthetic products are linearly related via single enzymatic transformations. To explore this system, we
have reconstructed the entire biosynthetic pathway through the heterologous expression of combinations of putative biosyn-
thetic genes from Streptomyces kanamyceticus in the nonaminoglycoside-producing Streptomyces venezuelae. Unexpectedly,
we discovered that the biosynthetic pathway contains an early branch point, governed by the substrate promiscuity of a
glycosyltransferase, that leads to the formation of two parallel pathways in which early intermediates are further modified.
Glycosyltransferase exchange can alter flux through these two parallel pathways, and the addition of other biosynthetic
enzymes can be used to synthesize known and new highly active antibiotics. These results complete our understanding of
kanamycin biosynthesis and demonstrate the potential of pathway engineering for direct in vivo production of clinically useful
antibiotics and more robust aminoglycosides.


2
0
1
1

N
a
t
u
r
e

A
m
e
r
i
c
a
,

I
n
c
.


A
l
l

r
i
g
h
t
s

r
e
s
e
r
v
e
d
.
844 NATURE CHEMICAL BIOLOGY | VOL 7 | NOVEMBER 2011 | www.nature.com/naturechemicalbiology
ARTICLE
NATURE CHEMICAL BIOLOGY DOI: 10.1038/NCHEMBIO.671
With knowledge of the biosynthetic steps in hand, we engineered
this pathway to modulate kanamycin biosynthetic flux toward 7
instead of 8 as well as to biosynthesize 3-deoxykanamycins includ-
ing tobramycin and 1-N-[S-4-amino-2-hydroxybutyric acid]
(AHBA)-conjugated kanamycins such as amikacin and the new
1-N-AHBA-kanamycin X, which shows greater antibacterial activ-
ity than amikacin.
RESULTS
Discovery of a key pseudodisaccharide intermediate
We reconstructed the kanamycin pathway stepwise in a mutant
strain of S. venezuelae lacking the genes for biosynthesis of the native
deoxysugar
17
by transplanting plasmids harboring different combi-
nations of putative genes from S. kanamyceticus in S. venezuelae
(Supplementary Methods; Supplementary Tables 1 and 2) and
then carrying out biochemical analysis and chemical characterization
(Figs. 2,3 and Supplementary Results). As anticipated, a recombinant
host expressing kanA-kanB-kanK (strain DOS) produced 1 (21.4 M)
(Fig. 2a). To confer resistance to kanamycin in S. venezuelae without
modifying the aminoglycoside itself, three resistance genes from the
gentamicin gene cluster
17,28
, gtmF-gtmK-gtmL, were introduced into
the engineered strain. Transformants expressing these genes were
resistant to >10.0 mg ml
1
of commercial kanamycin, as opposed
to the control strains, which were only resistant up to 1.0 g ml
1
.
These resistance determinants were expressed along with the kana-
mycin biosynthetic genes in the heterologous host in subsequent
experiments. Next, a recombinant (strain DOSf) expressing genes
for the biosynthesis of 1 plus kanF, which encodes the first glycosyl-
transferase, produced 2 (1.3 M) as expected. Notably, however, a
greater amount of 2-deamino-2-hydroxyparomamine (9; 7.1 M)
was produced (Fig. 2a,e). 9 has previously been synthesized to
examine the structure-activity relationship (SAR) of kanamycin
derivatives
30
, but its occurrence as a natural biosynthetic inter-
mediate in S. kanamyceticus has never been reported. The CFE of
S. venezuelae expressing kanF supplemented with 1 as a glycosyl
acceptor and uridine 5-diphospho-D-glucose (UDP-Glc) or UDP-2-
N-acetyl-D-glucosamine (UDP-GlcNAc) as a glycosyl donor produced
nine-fold greater 9 than 2 (Fig. 2b). An in vitro experiment using the
same CFE supplemented with UDP-D-2-glucosamine (UDP-GlcN)
did not produce the pseudodisaccharide 3 as is the case with GtmG,
which is involved in gentamicin biosynthesis
17
. Confirmation of the
structure of 9 produced by the recombinant host (Supplementary
Fig. 10 and Supplementary Table 11) indicates that KanF is a glu-
cosyltransferase as well as an N-acetylglucosaminyltransferase.
The CFE of the wild-type S. venezuelae supplemented with 1
and UDP-Glc was not able to synthesize 9, whereas the CFE of
S. kanamyceticus incubated with 1 and UDP-Glc produced 9, results
that exclude the possibility that an unknown glycosyltransferase of
S. venezuelae glucosylates 1 (Supplementary Fig. 22). Previously
we reported the role of KacA as a 2-N-acetylparomamine deacety-
lase
17
; we confirmed this finding here, as a recombinant (strain PAR)
expressing kacA along with the genes for biosynthesis of 2 and 9
produced 3 (2.2 M) (Fig. 2a).
Kinetic analysis of the first glycosyltransferase KanF
To learn more about the dual function of KanF, we investigated
its substrate preference through kinetic analysis. Soluble recom-
binant histidine-tagged KanF was obtained by expression in
S. venezuelae because KanF could not be functionally expressed in
E. coli (Supplementary Methods). The thorough kinetic analysis
revealed that addition of UDP-GlcNAc to 1 can be catalyzed by
KanF, but with a catalytic efficiency (k
cat
/K
m
(donor)) only one-fifth
O
O
OH
HO
HO
HO
H
2
N
NH
2
OH
NH
2
NH
2
HO
HO
H
2
N
OH
O
O
OH
HO
HO
HO
H
2
N
NH
2
OH
NH
O
2-N-acetylparomamine (2) Paromamine (3) Neamine (4)
UDP-GlcNAc
Kanamycin C (6) Kanamycin B (7) Kanamycin A (8)
UDP-Glc
O
OH
HO
HO
OPO
3
2-
OH
OH
HO
HO
O
OH
OH
HO
HO
H
2
N
OH
O
HO
HO
H
2
N
OH
KanA/
BtrC/
NeoC
KanB/
BtrS/
NeoS
KanK/
BtrE/
NeoE
KanB/
BtrS/
NeoS
D-Glucose-6-phosphate Keto-2-DOIA 2-DOI 2-DOIA 2-DOS (1)
KanF/
BtrM/
NeoM
KacA/
BtrD/
NeoD
BtrQ-BtrB/
NeoQ-NeoB
KanE ?
? ? ?
?
O
O
OH
HO
HO
HO
H
2
N
NH
2
NH
2
NH
2
O
O
O
O
HO
HO
HO
H
2
N
NH
2
OH
NH
2
OH
HO
NH
2
OH
O
O
O
O
HO
HO
HO
H
2
N
NH
2
OH
NH
2
OH
HO
NH
2
NH
2
O
O
O
O
HO
HO
HO
H
2
N
NH
2
OH
NH
2
OH
HO
OH
NH
2
O
O
O
O
HO
HO
HO
H
2
N
NH
2
OH
OH
OH
HO
OH
NH
2
3-Deamino-
3-hydroxykanamycin C (5)
Figure 1 | The previously proposed biosynthetic pathway of kanamycin. Steps to the formation of 3-deamino-3-hydroxykanamycin C (5) have been
characterized (solid lines). Kanamycin biosynthetic enzymes involved in the above steps as well as their homologous enzymes from butirosin and neomycin
biosynthetic gene clusters are listed. However, the complete biosynthetic route to the final product, kanamycin A (8), has yet to be discovered (dashed lines).
2-DOI: 2-deoxy-scyllo-inosose; 2-DOIA: 2-deoxy-scyllo-inosamine; UDP-GlcNAc: uridine 5-diphospho-D-2-N-acetylglucosamine; UDP-Glc; UDP-D-glucose.


2
0
1
1

N
a
t
u
r
e

A
m
e
r
i
c
a
,

I
n
c
.


A
l
l

r
i
g
h
t
s

r
e
s
e
r
v
e
d
.
NATURE CHEMICAL BIOLOGY | VOL 7 | NOVEMBER 2011 | www.nature.com/naturechemicalbiology 845
ARTICLE
NATURE CHEMICAL BIOLOGY DOI: 10.1038/NCHEMBIO.671
as high as for the addition of UDP-Glc to the same acceptor 1
(Table 1; Supplementary Fig. 23). Therefore, it is evident that KanF
accepts both UDP-Glc and UDP-GlcNAc as cosubstrates but prefer-
entially transfers Glc to 1.
The parallel pseudodisaccharide C6 amination pathways
The conversion of paromamine (3) to neamine (4) may be
required for kanamycin biosynthesis as it is in the case of neo-
mycin biosynthesis
31
. Based on the sequence analysis of the
1-containing aminoglycoside gene clusters, there are two sets of
genes predicted to be involved in the amination of the hydroxyl
groups of kanamycins. kanI and kacL are suggested to encode a
6-paromamine dehydrogenase and a 6-oxoparomamine amin-
otransferase, respectively
25
, which are involved in the step from
3 to 4. On the other hand, kanC and kanD products are proposed
to function in the C3 amination of UDP-Glc, yielding UDP-Kns
25
.
Alternatively, KanC and KanD could be responsible for the forma-
tion of 6 from 5 (refs. 2527). To assign functional roles to the
gene product pairs proposed, we first expressed kanI-kacL in the
strain PAR, which normally produces 3 and 9; the resulting strain
PARil produced 4 (2.2 M) and 2-deamino-2-hydroxyneamine
(10; 4.3 M) along with 3 and 9 (Fig. 2c). 10 has been known to
be a natural kanamycin intermediate for over 40 years
32
, but its
precursor and the gene products involved in its formation have
never been characterized. Moreover, when the CFE of the recom-
binant strain expressing kanI-kacL was incubated in vitro with 3
and 9, the above 6-aminated pseudodisaccharides 4 and 10 were
produced, respectively (Fig. 2d), establishing their roles in the
C6 amination of both 3 and 9. The kanI-kacL products seem to
prefer 3 as a substrate over 9, based on the conversion yields of
the reactions. However, that expression of kanC-kanD in the strain
PAR (strain PARcd) produced only 3 and 9 (Fig. 2c) indicates that
KanC-KanD does not participate in amination of the pseudodi-
saccharides and instead may be involved in the subsequent steps
in pseudotrisaccharide biosynthesis.
The second glycosyltransferase adds UDP-Kns and UDP-Glc
A second glycosyltransferase activity is required for the biosynthe-
sis of kanamycin complex. Expression of kanE encoding a second
glycosyltransferase along with the genes for biosynthesis of 3 and 9
b 1
1
1
1
9 2
100
0 5 10 15
+ 1 + UDP-GlcN KanF
+ 1 + UDP-GlcNAc KanF
+ 1 + UDP-Glc KanF
+ 1 KanF
e
d
3
3
4
100
10 15
+ 3 Kanl-KacL
+ 3 Kanl-KacL
9
9
10 + 9 Kanl-KacL
+ 9 Kanl-KacL
100
5 10
c
9
3
9
1
10
3
4
100
0 5 10 15 20
PARil
kanA-B-K kanF kacA kanl-kacL
PARcd
kanA-B-K kanF kacA kanC-kanD
1
1
1
2
9
9
3
100
0 5
R
e
l
a
t
i
v
e
a
b
u
n
d
a
n
c
e
R
e
l
a
t
i
v
e
a
b
u
n
d
a
n
c
e
R
e
l
a
t
i
v
e
a
b
u
n
d
a
n
c
e
R
e
l
a
t
i
v
e
a
b
u
n
d
a
n
c
e
R
e
l
a
t
i
v
e
a
b
u
n
d
a
n
c
e
10
T
ret
(min)
T
ret
(min)
T
ret
(min)
T
ret
(min) T
ret
(min)
15 20
PAR
kanA-B-K kanF kacA
DOSf
kanA-B-K kanF
DOS
kanA-B-K
a
O
O
OH
HO
HO
H
2
N
NH
2
OH
OH
HO
2-Deamino-2-hydroxyparomamine (9)
O
O
OH
HO
HO
H
2
N
NH
2
NH
2
OH
HO
2-Deamino-2-hydroxyneamine (10)
6 6
Figure 2 | HPLC-ESI-MS/MS analysis of kanamycin pseudodisaccharide intermediates obtained from the in vivo expression of candidate kanamycin
gene sets in recombinant S. venezuelae hosts and in vitro reactions using cell-free extracts of recombinants. (a) Chromatograms of culture extracts
from recombinants (DOS, DOSf and PAR) expressing kanA-kanB-kanK, kanA-kanB-kanK-kanF (the latter gene encoding the first glycosyltransferase)
and kanA-kanB-kanK-kanF-kacA (the latter gene encoding deacetylase). Each colored rectangle represents the gene(s) expressed in the corresponding
recombinant strains. (b) Chromatograms of the first glycosyltransferase KanF-catalyzed production of 2-N-acetylparomamine (2) and 2-deamino-2-
hydroxyparomamine (9) supplemented with 1 and three different uridine 5-diphospho-(UDP) sugars: UDP-glucose (Glc), UDP-2-N-acetylglucosamine
(GlcNAc) and UDP-2-glucosamine (GlcN). Filled boxes represent the CFEs of recombinant host expressing the corresponding gene(s) used for in vitro
reactions, and black-framed boxes indicate boiled CFEs used as controls. (c) Chromatograms of culture extracts from recombinants (PARcd and PARil)
expressing kanA-kanB-kanK-kanF-kacA together with kanC-kanD and kanA-kanB-kanK-kanF-kacA together with kanI-kacL. (d) Chromatograms of the
KanI-KacLcatalyzed production of neamine (4) and 2-deamino-2-hydroxyneamine (10) supplemented with paromamine (3) and 9, respectively.
(e) Structures of kanamycin pseudodisaccharide intermediates. T
ret
, retention time.
Table 1 | Kinetic parameters for KanF and KanE with donors and acceptors
Enzyme Acceptor (A) Donor (D) k
cat
(min
1
) K
m
(A) (mM) K
m
(D) (mM)
k
cat
/K
m
(D)
(min
1
mM
1
)
KanF 1 UDP-Glc 17.41 2.40 0.28 0.03 0.44 0.06 39.54
UDP-GlcNAc 19.82 2.42 0.32 0.04 2.54 0.24 7.79
KanE 3 UDP-Glc 3.18 0.37 0.25 0.04 0.26 0.05 12.23
UDP-Kns 3.36 0.39 0.05 0.01 0.03 0.01 112.02
Shown are kinetic parameters for KanF with two donors (UDP-Glc and UDP-GlcNAc) and an acceptor 2-deoxystreptamine (1) and for KanE with two donors (UDP-Glc and UDP-Kns) and an acceptor
paromamine (3). All kinetic data represent mean s.d. (n = 3), derived from the Michaelis-Menten equation using GraphPad Prism (GraphPad, version 5).


2
0
1
1

N
a
t
u
r
e

A
m
e
r
i
c
a
,

I
n
c
.


A
l
l

r
i
g
h
t
s

r
e
s
e
r
v
e
d
.
846 NATURE CHEMICAL BIOLOGY | VOL 7 | NOVEMBER 2011 | www.nature.com/naturechemicalbiology
ARTICLE
NATURE CHEMICAL BIOLOGY DOI: 10.1038/NCHEMBIO.671
(strain KCXcd) produced 5 (3.2 M), a finding that is consis-
tent with a previous study using CFEs of E. coli expressing kanE
22

and the hitherto unknown kanamycin congener 3-deamino-3-
hydroxykanamycin X (11; 4.5 M) (Fig. 3a,e), which, like 9, has
also been synthesized to investigate SAR
30
but has not been reported
to be a natural kanamycin congener. Additional expression of
kanC-kanD in the strain KCXcd (stain KCX) further produced 6
(1.6 M) and kanamycin X (12; 6.0 M), indicating the coopera-
tive function of KanC-KanD in C3 amination. Coexpression of
kanI-kacL in the strain KCXcd (strain KABcd) generated
3-deamino-3-hydroxykanamycin B (13; 2.0 M) and kanamycin D
(14; 3.2 M). Finally, expression of kanC-kanD in the strain
KABcd (strain KAB) produced 1.0 mg l
1
(2.1 M) 7 and
3.1 mg l
1
(6.3 M) 8 (Fig. 3a,e).
To determine whether free UDP-Kns synthesized from UDP-Glc
is directly incorporated into a pseudodisaccharide or whether a Glc
moiety attached at the C6 of pseudodisaccharide is converted to
Kns, CFEs were prepared from S. venezuelae recombinants express-
ing kanE or kanC-kanD. Incubation of 3 and 4 with UDP-Glc plus
both CFEs produced 5 and 6 as well as 13 and 7, respectively, thereby
verifying the functionality of the CFE systems (Fig. 3b,c). In subse-
quent experiments, the CFEs were used sequentially. Thus, when
CFE containing KanE was incubated with UDP-Glc plus 3 or 4, only
5 or 13 was produced. When these reactions were quenched and
CFE containing KanC-KanD was added, 5 and 13 remained, and
no further conversion was detected. However, when 3 and 4 were
first incubated with UDP-Glc plus CFE containing KanC-KanD and
then, after quenching, with CFE containing KanE, the products were
e
O
O
O
O
HO
HO
HO
H
2
N
NH
2
OH
OH
OH
HO
OH
OH
3-Deamino-3-hydroxykanamycin X (11)
6
2
3
2
O
O
O
O
HO
HO
HO
H
2
N
NH
OH
NH
2
OH
HO
OH
OH
Kanamycin X (12)
6
2
3
O
O
O
O
HO
HO
HO
H
2
N
NH
2
OH
OH
OH
HO
NH
2
OH
Kanamycin D (14)
6
2
3
O
O
O
O
HO
HO
HO
H
2
N
NH
2
OH
OH
OH
HO
NH
2
NH
2
3-Deamino-3-hydroxykanamycin B (13)
6
2
3
a
100
5 10 15 20 25 30
12
14
4
4
6
9
10
11
11
3
5
5
6
9
8
7
14
13
12
KCX
kanA-B-K kanC-kanD kanF kacA kanE
KAB
KABcd
KCXcd
kanA-B-K kanC-kanD kanF kacA kanE kanl-kacL
kanA-B-K kanF kacA kanE kanl-kacL
kanA-B-K kanF kacA kanE
T
ret
(min)
R
e
l
a
t
i
v
e
a
b
u
n
d
a
n
c
e
T
ret
(min)
b 3
3
3
3
5
3
6
6
5
5
3
5
100
0 5 10 15 20
+ 3 + UDP-Glc KanC-KanD KanE
+ 3 + UDP-Glc KanC-KanD
+ 3 + UDP-Glc KanE
+ 3 + UDP-Glc KanE KanC-KanD
+ KanE KanC-KanD
+ KanE KanC-KanD
+ 3 + UDP-Glc
+ 3 + UDP-Glc R
e
l
a
t
i
v
e
a
b
u
n
d
a
n
c
e
T
ret
(min)
c
7
7
4
4
4
13
13
13
13
4
4
4
+ 4 + UDP-Glc KanC-KanD KanE
+ 4 + UDP-Glc KanC-KanD
+ 4 + UDP-Glc KanE
+ 4 + UDP-Glc KanE KanC-KanD
+ KanE KanC-KanD
+ KanE KanC-KanD
100
10 15 20 25 30
+ 4 + UDP-Glc
+ 4 + UDP-Glc R
e
l
a
t
i
v
e
a
b
u
n
d
a
n
c
e
T
ret
(min)
d
7
6
6
100
20 25
+ 6 Kanl-KacL
+ 6 Kanl-KacL
12 100
10 15
+ 12 Kanl-KacL
12
8
+ 12 Kanl-KacL
T
ret
(min)
R
e
l
a
t
i
v
e
a
b
u
n
d
a
n
c
e
R
e
l
a
t
i
v
e
a
b
u
n
d
a
n
c
e
Figure 3 | HPLC-ESI-MS/MS analysis of kanamycin pseudotrisaccharides obtained from the in vivo expression of candidate kanamycin gene sets in
recombinant S. venezuelae hosts and in vitro reactions using cell-free extracts of recombinants. (a) Chromatograms of culture extracts from recombinants
(KCXcd, KCX, KABcd and KAB) expressing kanA-kanB-kanK-kanF-kacA together with kanE (encoding a second glycosyltransferase), kanA-kanB-kanK-
kanF-kacA-kanE-kanC-kanD, kanA-kanB-kanK-kanF-kacA-kanE-kanl-kacL and kanA-kanB-kanK-kanF-kacA-kanE-kanC-kanD-kanl-kacL. Each colored rectangle
represents the gene(s) expressed in the corresponding recombinant strains. (b) Chromatograms of the second glycosyltransferase KanE-catalyzed production
of 3-deamino-3-hydroxykanamycin C (5) and kanamycin C (6), supplemented with paromamine (3) and uridine 5-diphosphoglucose (UDP-Glc). Filled boxes
represent the CFEs of recombinant host expressing the corresponding gene(s) used for in vitro reactions, and black-framed boxes indicate boiled CFEs used as
controls. Arrows represent quenching of sequential reactions. (c) Chromatograms of the KanE-catalyzed production of kanamycin B (7) and 3-deamino-
3-hydroxykanamycin B (13) supplemented with neamine (4) and UDP-Glc. (d) Chromatograms of the KanI-KacLcatalyzed production of 7 and kanamycin A
(8) supplemented with 6 and kanamycin X (12), respectively. (e) Structures of kanamycin pseudotrisaccharide congeners.


2
0
1
1

N
a
t
u
r
e

A
m
e
r
i
c
a
,

I
n
c
.


A
l
l

r
i
g
h
t
s

r
e
s
e
r
v
e
d
.
NATURE CHEMICAL BIOLOGY | VOL 7 | NOVEMBER 2011 | www.nature.com/naturechemicalbiology 847
ARTICLE
NATURE CHEMICAL BIOLOGY DOI: 10.1038/NCHEMBIO.671
6 and 7, respectively (Fig. 3b,c). No intermediate was produced, as
expected, in the reaction of 3 or 4 with the CFE containing KanC-
KanD and UDP-Glc only (Fig. 3b,c). That equivalent results were
observed using 9 and 10 as substrates (Supplementary Fig. 22)
demonstrates that the activity of KanC-KanD precedes that of KanE.
We were also able to synthesize UDP-Kns enzymatically from
UDP-Glc using the CFE containing KanC-KanD (Supplementary
Methods). These results prove that this enzyme pair acts only on the
nucleotide-activated sugar, as in the biosynthesis of free kanosamine
in rifamycin-producing Amycolatopsis mediterranei
33
, and that the
previously proposed biosynthetic route to 6 from 5 (ref. 22) does
not exist. The amounts of 5 and 13 produced using S. venezuelae
CFEs were relatively low compared to those of 6 and 7, respectively,
(Fig. 3b,c), results implying that UDP-Kns is a better glycosyl donor
for KanE than UDP-Glc. Kinetic analysis using the recombinant
histidine-tagged KanE expressed in S. venezuelae and the enzymati-
cally synthesized UDP-Kns (Supplementary Methods) confirmed
that KanEs preferred sugar donor is UDP-Kns as KanE transfers
Kns to 3 with nine-fold higher catalytic efficiency than Glc (Table 1
and Supplementary Fig. 24).
Finally, we examined whether the KanI-KacL pair was able to
catalyze C6 amination of 6 and 12 to generate 7 and 8, respectively,
using CFEs (Fig. 3d). In the case of pseudotrisaccharides, the KanI-
KacL pair seems to have a preference for 12 over 6. Given our earlier
demonstration (Fig. 2d) of their ability to generate 4 and 10 from 3
and 9, respectively, these results reproduce the 6-amination func-
tion of KanI-KacL and reveal their substrate flexibility embracing
both pseudodisaccharides and pseudotrisaccharides.
Kanamycin synthesis follows independent parallel pathways
These combined results support the development of a new bio-
synthetic proposal for the production of kanamycin, constructed
on the basis of several lines of evidence. First, identification of
the hitherto unknown kanamycin biosynthetic intermediate 9
and the pseudotrisaccharide congener 11 allows us to envision
new pathways of molecular transformations that could yield the
final products. Second, our discovery that the glycosyltrans-
ferase KanE is able to use four different pseudodisaccharides as
glycosyl acceptors and two different glycosyl donors highlights
an early branch point in the synthetic route and points to an
array of pseudotrisaccharides that must be considered in the
final mechanistic proposal. Of the pseudotrisaccharide prod-
ucts, 12 and 14 were previously discovered from the kanamycin
producer
34
and the mother liquors of kanamycins
35
, respectively;
13, also known as nebramycin factor 3, has been isolated in
tobramycin-producing Streptomyces tenebrarius
36
. However, the
interrelationships between even the known biosynthetic inter-
mediates and congeners in the kanamycin biosynthesis, such as
10, 12, 13 or 14, and the roles of gene products involved in the
each biosynthetic step have never been elucidated, to our knowl-
edge, until this study. Our stepwise heterologous reconstruction
of the kanamycin pathway clearly establishes the position of each
biosynthetic intermediate and congener in the kanamycin bio-
synthetic network as well as the function of each gene product.
The new biosynthetic proposal uses the unexpected architec-
ture of two parallel pathways governed by the substrate-flexible
KanF and KanE glycosyltransferases as well as the ability of
KanI-KacL to accept both pseudodisaccharides and pseudotri-
saccharides (Fig. 4). This newly established pathway is entirely
contradictory to all previous proposals that 7 is the direct precur-
sor of 8 (refs. 2527). Indeed, S. kanamyceticus CFE converted
exogenous 6 and 12 to 7 and 8, respectively, whereas 7 was not
transformed into 8, confirming that 7 is not the precursor of 8
(Supplementary Fig. 25).
O
OH
HO
HO
OPO3
2-
OH
NH2
HO
HO
H2N
OH
D-Glucose-6-phosphate 2-DOS (1)
O
O
OH
HO
HO
HO
H2N
NH2
OH
NH
O
2-N-acetylparomamine (2)
2-Deamino-2-hydroxyneamine (10)
O
O
O
O
HO
HO
HO
H2N
NH2
OH
OH
OH
HO
NH2
NH2
Kanamycin D (14)
Kanamycin B (7) Kanamycin A (8)
UDP-Glc
O
O
OH
HO
HO
HO
H2N
NH2
OH
OH
2-Deamino-2-hydroxyparomamine (9)
Neamine (4)
3-Deamino-
3-hydroxykanamycin C (5)
3-Deamino-
3-hydroxykanamycin X (11)
Kanamycin X (12) Kanamycin C (6)
O
O
OH
HO
HO
HO
H2N
NH2
OH
NH2
Paromamine (3)
UDP-GlcNAc
UDP-Kns
UDP-Glc
UDP-Kns
UDP-Glc
UDP-Glc UDP-Glc
kanA-B-K
kacA
kanF
kanI-kacL kanI-kacL
kanE kanE
kanE
kanE
kanC-kanD kanC-kanD
kanI-kacL
kanI-kacL
O
UDP
HO
H2N
OH
OH
O
UDP
HO
H2N
OH
OH
O
O
OH
HO
HO
HO
H2N
NH2
NH2
NH2
O
O
OH
HO
HO
HO
H2N
NH2
NH2
OH
O
O
O
O
HO
HO
HO
H2N
NH2
OH
OH
OH
HO
OH
NH2
O
O
O
O
HO
HO
HO
H2N
NH2
OH
OH
OH
HO
OH
OH
O
O
O
O
HO
HO
HO
H2N
NH2
OH
NH2
OH
HO
NH2
OH
O
O
O
O
HO
HO
HO
H2N
NH2
OH
NH2
OH
HO
OH
OH
O
O
O
O
HO
HO
HO
H2N
NH2
OH
NH2
OH
HO
OH
NH2
O
O
O
O
HO
HO
HO
H2N
NH2
OH
NH2
OH
HO
NH2
NH2
UDP-Glc UDP-Glc
kanE kanE
3-Deamino-
3-hydroxykanamycin B (13)
O
O
O
O
HO
HO
HO
H2N
NH2
OH
OH
OH
HO
NH2
OH
Figure 4 | Biosynthetic pathway of kanamycin complex. Each colored rectangle represents an annotated gene, the product of which has been proven
to catalyze a step in the biosynthesis of a kanamycin intermediate. Each colored set of atoms represents the functional group formed by the product of
the annotated gene indicated by the rectangle of the same color. Color codes used for arrows: blue, previously reported biosynthetic steps; black, the
characterized steps in this study. 2-DOS, 2-deoxystreptamine; UDP-GlcNAc, uridine 5-diphospho-D-2-N-acetylglucosamine; UDP-Glc, UDP-D-glucose;
UDP-Kns, UDP-kanosamine.


2
0
1
1

N
a
t
u
r
e

A
m
e
r
i
c
a
,

I
n
c
.


A
l
l

r
i
g
h
t
s

r
e
s
e
r
v
e
d
.
848 NATURE CHEMICAL BIOLOGY | VOL 7 | NOVEMBER 2011 | www.nature.com/naturechemicalbiology
ARTICLE
NATURE CHEMICAL BIOLOGY DOI: 10.1038/NCHEMBIO.671
Modulation of kanamycin biosynthetic flux to kanamycin B
Kanamycin A (8) is the major fermentation product in both the
wild-type kanamycin producer and the heterologous host because
KanF glycosyltransferase and KanI-KacL prefer UDP-Glc over
UDP-GlcNAc and 12 over 6, respectively (Table 1 and Fig. 3d).
With the knowledge that the biosynthetic routes to kanamycins A
and B are separated by the initial KanF reaction, we were curious to
see whether we could bias the flux to produce kanamycin B (7) as
a main product as it is a more valuable congener for the chemical
synthesis of dibekacin and arbekacin (Supplementary Fig. 1).
Initially, this was addressed by separately substituting kanF in the
DOSf strain with other glycosyltransferase-encoding genes includ-
ing nemD (otherwise known as neoM or neo8), tobM1 and gtmG
from the neomycin
16,37
, tobramycin
38
and gentamicin
17
biosynthetic
clusters, respectively. A control strain DOSf expressing kanF gener-
ated 2 (1.4 M) and 9 (7.1 M), amounts corresponding to 16%
and 84% of the total pseudodisaccharides produced, respectively. In
contrast, the production of 2 (4.4 M; 57%) and 9 (3.4 M; 43%)
by the recombinant expressing nemD (strain DOSn) indicates that
NemD is flexible toward the UDP sugars but seems to use UDP-
GlcNAc preferentially over UDP-Glc in contrast to KanF. It was
previously reported that the CFE from E. coli expressing nemD cat-
alyzed the glycosylation of 1 with UDP-GlcNAc to give 2, whereas
UDP-Glc was not accepted as a glycosyl donor
37
. However, in this
study, NemD was found to be able to transfer Glc to 1 in addition to
GlcNAc, which was still the more favorable substrate. When tobM1
and gtmG replaced kanF (DOSt
1
and DOSg, respectively), there was
no considerable change in the ratio of 2 to 9 produced (Fig. 5a).
In additional studies, these three glycosyltransferase-encoding
genes were separately substituted for kanF in the strain KCX, which
normally produces 1.6 M 6 and 6.0 M 12 as 16% and 60% of total
pseudotrisaccharides, respectively. However, in the recombinant
expressing nemD (strain KCXfn), the ratio of pseudotrisaccharide
yields was inverted (6: 5.0 M, 65%; 12: 0.4 M, 7%). Recombinants
expressing either tobM1 (KCXft
1
) or gtmG (KCXfg) had unal-
tered ratios of pseudotrisaccharide products (Fig. 5b). Finally, the
yields of 7 and 8 generated by the strain KAB were 21% (2.3 M)
and 61% (6.8 M) of the total amount of pseudotrisaccharides,
respectively. Substitution of kanF with nemD in the strain KAB
(KABfn) resulted in an increase of the yield of 7 to 46% (3.9 M)
(Fig. 5c), a result demonstrating the feasibility of biasing biosyn-
thetic flux toward the kanamycin B (7) and C (6) pathways at the
expense of kanamycins A (8) and X (12). The yield of 7 produced in
the strain KABfn was increased by approximately two-fold, which
was less than expected considering the production of 6-hydroxy
kanamycins (Fig. 5) and was probably caused by the preference of
KanI and KacL for 12 over 6 (Fig. 3d).
A noticeable difference in the substrate preferences of the glyco-
syltransferases KanF and NemD prompted us to examine molecular
models of the glycosyltransferase-substrate complex (Supplementary
Methods). Quantification of hydrogen bonds between glycosyl
donors and glycosyltransferases and differences in binding free ener-
gies of the various complexes supported the notion that the KanF
UDP-Glc and NemDUDP-GlcNAc complexes are preferred over
the KanFUDP-GlcNAc and NemDUDP-Glc complexes, respec-
tively (Supplementary Figs. 26, 27; Supplementary Table 19).
Direct fermentative production of modified kanamycins
We attempted to engineer the kanamycin pathway further for the direct
in vivo production of 1-N-acylated kanamycins containing AHBA such
as amikacin (15) and 3-deoxykanamycins such as tobramycin (16)
(Fig. 6). The AHBA pharmacophore was originally observed in the
natural product butirosin
8
. Recently, it was revealed, through the use
of in vitro chemoenzymatic reactions, that a set of seven genes (btrG-
btrH-btrI-btrJ-btrK-btrO-btrV) is responsible for the biosynthesis
and incorporation of the AHBA side chain into the amino group at
the C1 of 1 in butirosin
39
. Introduction of these genes into the strain
KCX (creating strain KCXb) led to the in vivo generation of 0.6 mg l
1

of a new aminoglycoside, 1-N-AHBA-kanamycin X (17; 1.0 M).
Furthermore, coexpression of the btr gene set in the stain KAB (strain
KABb) led to the successful production 0.5 mg l
1
of 1-N-AHBA-
kanamycin A, known as amikacin (15; 0.8 M) (Fig. 6a,c).
Tobramycin (16) from S. tenebrarius
6
is known to be less prone
to deactivation by aminoglycoside-modifying enzymes because it
lacks a susceptible 3-hydroxy group
10
. Based on the bioinformatics
100
0 5 10 15 20
9
2
2
2
2
1
1
1
1
9
9
9
2 4 6 8
DOSf
DOSn
2
2
2
2
9
9
9
9
kanF
nemD
tobM1
DOSg
gtmG
DOSt1
Tret (min)
Tret (min)
Tret (min)
a
kanA-B-K
b
12
12
12
12
6
6
6
11
11
11
11
5
5
5
5
6
100
5 10 15 20 25
2 4 6 8
KCX
KCXfn
KABfn
KCXfg
KCXft1
5
5
5
5
6
6
6
6
11
11
11
11
12
12
12
12
kanF
nemD
tobM1
gtmG
kanA-B-K kacA kanE kanC-kanD
c
8
12
12
14
14
4
4
6
6
8
7
7
100
10 15 20 25
Yield (M)
6
6
7
7
8
8
12
12
2 4 6 8
KAB
kanF
nemD
kanA-B-K kacA kanE kanC-kanD kanl-kacL
R
e
l
a
t
i
v
e
a
b
u
n
d
a
n
c
e
R
e
l
a
t
i
v
e
a
b
u
n
d
a
n
c
e
R
e
l
a
t
i
v
e
a
b
u
n
d
a
n
c
e
Figure 5 | HPLC-ESI-MS/MS analysis of kanamycin biosynthetic
intermediates from recombinant S. venezuelae hosts in which the
first glycosyltransferase-encoding gene has been swapped.
(a) Chromatograms of culture extracts from recombinants (DOSf, DOSn,
DOSt
1
and DOSg) expressing kanA-kanB-kanK together with kanF from
S. kanamyceticus, nemD from neomycin-producing S. fradiae, tobM1 from
tobramycin-producing S. tenebrarius or gtmG from gentamicin-producing
Micromonospora echinospora. Each colored rectangle represents the gene(s)
expressed in the corresponding recombinant strains. The bar graph
at right indicates the yield of 2-N-acetylparomamine (2) and
2-deamino-2-hydroxyparomamine (9) produced by each recombinant
host. (b) Chromatograms of culture extracts from recombinants
(KCX, KCXfn, KCXft
1
and KCXfg) expressing kanA-kanB-kanK-kacA-
kanE-kanC-kanD together with kanF, nemD, tobM1 or gtmG. The bar
graph summarizes the yield of kanamycin pseudotrisaccharides
such as 3-deamino-3-hydroxykanamycin C (5), kanamycin C (6),
3-deamino-3-hydroxykanamycin X (11) or kanamycin X (12), produced
by each recombinant host. (c) Chromatograms of culture extracts from
recombinants (KAB and KABfn) expressing kanA-kanB-kanK-kacA-kanE-
kanC-kanD-kanI-kacL together with kanF or nemD. The bar graph indicates
the yield of kanamycins such as 6, kanamycin B (7), kanamycin A (8) and
12, produced by each recombinant host. Data represent mean s.d. (n = 3).


2
0
1
1

N
a
t
u
r
e

A
m
e
r
i
c
a
,

I
n
c
.


A
l
l

r
i
g
h
t
s

r
e
s
e
r
v
e
d
.
NATURE CHEMICAL BIOLOGY | VOL 7 | NOVEMBER 2011 | www.nature.com/naturechemicalbiology 849
ARTICLE
NATURE CHEMICAL BIOLOGY DOI: 10.1038/NCHEMBIO.671
analysis of various aminoglycoside gene clusters, it seems that two
putative apramycin genes, aprD3 and aprD4 from S. tenebrarius, which
produces a small amount of 16, are responsible for C3 deoxygenation
of pseudodisaccharides, pseudotrisaccharides or both, and their
gene products were predicted to be a radical S-adenosylmethionine
oxidoreductase and a dehydrogenase, respectively
26,27
. Both genes
were introduced into the strain KABfn, which was engineered
for the increased production of 7. The resulting strain (KABfna)
generated nebramine
36
(18) as a major product as well as
3-deoxykanamycin C (19) and 3-deoxykanamycin A (20), which
both have been previously chemically synthesized from 6 (refs. 40,41),
unlike 3-deoxykanamycin B (tobramycin; 16) (Fig. 6b,c). The struc-
tures of 18, 19 and 20 were proposed by comparing their retention
behaviors on HPLC-ESI-MS/MS captured by multiple-reaction mon-
itoring (307 > 163, 469 > 163, and 469 > 163, respectively) and ESI-
MS/MS spectra with those of 18 isolated from wild-type S. tenebrarius
ATCC 17920 and authentic 16, which differs from 19 and 20 only
by having an amine instead of a hydroxyl group at C2 and C6,
respectively. However, these results showed that the joint activity
of AprD3 and AprD4 eliminates a 3-hydroxyl group in 4 as well as
T
ret
(min)
T
ret
(min)
100
R
e
l
a
t
i
v
e

a
b
u
n
d
a
n
c
e
R
e
l
a
t
i
v
e
a
b
u
n
d
a
n
c
e
10 15 20 25 30
100
5 10 15 20 25
8
4
6
7
15
17
12
12
5
6
11
14
KABb
kanA-B-K kanC-kanD kanF kanl-kacL
btrl-J-K-O-V-G-H +
KCXb
kanA-B-K kanC-kanD kanF kacA kanE
btrl-J-K-O-V-G-H +
100
10 15 20 25 30
16
12
12
8
8
20
20
6
6
19
19
7
7
16
18
18
KABfnet
2
a
KABfna
kanA-B-K nemD
aprD3-aprD4 +
kanC-kanD
Authentic
kanA-B-K nemD
aprD3-aprD4 +
a
b
c
kacA kanE
kacA
kacA
tobM2
kanE
kanl-kacL
kanC-kanD kanl-kacL
Neamine (4)
Amikacin (15)
Tobramycin (16)
1-N-AHBA-kanamycin X (17) 3-Deoxykanamycin C (19)
aprD3-aprD4 tobM2
Nebramine (18)
O
O
O
O
HO
HO
H
2
N
NH
2
OH
NH
2
OH
HO
OH
NH
2
6
3
1
O
O
O
O
HO
HO
H
2
N
NH
2
OH
NH
2
OH
HO
NH
2
OH
6
3
1
3-Deoxykanamycin A (20)
O
O
O
O
HO
HO
H
2
N
NH
2
OH
NH
2
OH
HO
NH
2
NH
2
6
3
1
O
O
OH
HO
HO
H
2
N
NH
2
NH
2
NH
2
6
3
1
O
O
OH
HO
HO
H
2
N
NH
2
NH
2
NH
2
6
3
1
HO
O
O
O
O
HO
HO
HO
H
2
N
OH
NH
2
OH
HO
NH
2
OH
6
3
1
O
O
O
O
HO
HO
HO
H
2
N
OH
NH
2
OH
HO
OH
OH
6
3
1
H
N
O
OH
H
N
O
OH
NH
2
NH
2
Figure 6 | HPLC-ESI-MS/MS analysis of 1-N-AHBA-kanamycins and 3-deoxykanamycins produced by recombinant S. venezuelae hosts.
(a) Chromatograms of culture extracts from recombinants (KCXb and KABb) coexpressing a total of seven butirosin biosynthetic genes (btrG-btrH-btrI-
btrJ-btrK-btrO-btrV) from butirosin-producing Bacillus circulans, together with the gene set for the biosynthesis of kanamycin X (12) (kanA-kanB-kanK-
kanF-kacA-kanE-kanC-kanD) or kanamycin A (8) (kanA-kanB-kanK-kanF-kacA-kanE-kanC-kanD-kanI-kacL). Each colored rectangle represents the gene(s)
expressed in the corresponding recombinant strains. 15 is the AHBA-conjugated kanamycin derivative amikacin, and 17 is the new AHBA-conjugated
kanamycin derivative 1-N-AHBA-kanamycin X. (b) Chromatograms of culture extracts from recombinants (KABfna and KABfnet
2
a) coexpressing
aprD3-D4 from apramycin-producing S. tenebrarius, together with the biosynthetic gene set for 8, in which kanF or kanF-kanE was replaced by nemD
or nemD-tobM2 (kanA-kanB-kanK-nemD-kacA-kanE-kanC-kanD-kanI-kacL or kanA-kanB-kanK-nemD-kacA-tobM2-kanC-kanD-kanI-kacL). 16 and 18 are
tobramycin and the tobramycin biosynthetic intermediate nebramine, respectively. (c) Structures of kanamycin analogs produced by combinatorial
biosynthesis and the engineered biosynthetic route to tobramycin (16) from neamine (4) using combinatorial expression of aprD3-aprD4 and tobM2. Each
colored atom represents the functional group formed by the product of the annotated gene indicated by the rectangle of the same color.


2
0
1
1

N
a
t
u
r
e

A
m
e
r
i
c
a
,

I
n
c
.


A
l
l

r
i
g
h
t
s

r
e
s
e
r
v
e
d
.
850 NATURE CHEMICAL BIOLOGY | VOL 7 | NOVEMBER 2011 | www.nature.com/naturechemicalbiology
ARTICLE
NATURE CHEMICAL BIOLOGY DOI: 10.1038/NCHEMBIO.671
in 3 and 10. Besides, the second glycosyltransferase, KanE, seems
to selectively transfer Kns to 3 and 10 but not to 18. To choose
a glycosyltransferase that can transfer Kns to 18 to yield 16,
CFEs from S. kanamyceticus and S. tenebrarius producing 16 were
prepared and incubated with 18. The CFE from S. tenebrarius
with tobM2 as a second glycosyltransferase-encoding gene con-
verted 18 into 16, whereas the CFE from S. kanamyceticus did
not (Supplementary Fig. 28). Therefore, we constructed a new
strain (KABfnet
2
a) expressing aprD3-aprD4, nemD and tobM2
encoding glycosyltransferases and genes for 7 biosynthesis to
produce 16 (0.4 mg l
1
; 0.8 M), thus demonstrating that TobM2
is able to use 18 as a glycosyl acceptor together with the 3-deoxy
analogs of 3 and 10 (Fig. 6b,c). The CFE containing AprD3-AprD4
converted 4 to 18 but did not transform 7 into 16, a result indi-
cating that these enzymes are active only for pseudodisac charides
(Supplementary Fig. 29).
Improved antibacterial activity of 1-N-AHBA-kanamycin X
The antibiotic activity of kanamycin biosynthetic intermediates
and derivatives obtained in this study were tested against two
Gram-negative bacteria (E. coli and Pseudomonas aeruginosa)
(Supplementary Table 20). The pseudodisaccharides were not
as active as the major pseudotrisaccharide kanamycin congeners
6, 7 and 8 against kanamycin-sensitive (Kan
S
) test strains. Of the
pseudodisaccharides, the 6-amino compounds 4 and 10 were more
active than the 6-hydroxy derivatives 3 and 9 against Kan
S
E. coli
strains. The 6-amino counterparts 7 and 8 of the 6-hydroxy pseudo-
trisaccharide derivatives 6 and 12 were slightly more active against
Kan
S
test strains. In addition, the 3-amino derivatives 6, 7, 8 and 12
had increased activity against the test strains compared with the cor-
responding 3-hydroxy derivatives 5, 13, 14 and 11. Therefore, ami-
nation of the C6- and/or C3-hydroxyl groups of kanamycins seems
to increase their antibacterial activity, in agreement with the results
of an earlier study
32
. All kanamycin biosynthetic intermediates were
inactive (minimal inhibitory concentration (MIC) >128 g ml
1
)
against kanamycin-resistant (Kan
R
) strains. When tested against an
amikacin-sensitive (Amk
S
) P. aeruginosa clinical isolate, most kana-
mycin intermediates had very low activity, with the exception of the
main kanamycin congeners 6, 7, 8 and 12. Remarkably, compared
with amikacin (15), the new compound 17 had enhanced activity
against all the test strains and 17, in particular, retained potency
(MIC ~64 g ml
1
) against an amikacin-resistant (Amk
R
) clinical
isolate of P. aeruginosa that was insensitive to 15.
DISCUSSION
It has long been believed that 6 and 7 would be the direct biosyn-
thetic precursors of 8 and that all kanamycin congeners might be
derived from 3 (refs. 2527), predictions based on metabolic profiles
from the culture extracts and the precursor feeding study
23
, which
demonstrated the exclusive incorporation of [1-
14
C]GlcN into the
GlcN moiety of 3 and the 6-amino-6-deoxy Glc moiety of 8 during
fermentation of S. kanamyceticus. However, our results clearly dem-
onstrate that there are two independent biosynthetic routes to 7 and
8, and we show that 3 is a precursor of 6 and 7, whereas 9 and 12
are the true precursors of 8. The kanamycin biosynthetic pathway
revealed herein is consistent with the results of another previous
precursor feeding study
24
in which radiolabeled 3 was not incorpo-
rated into 8, but it is contradictory to all previous suggestions about
the nature of the pathway.
Chemoenzymatic attachment of AHBA side chain onto several
1-containing aminoglycosides, including 8, was reported
39
.
However, such synthesis requires two steps in vitro even when
using a synthetic substrate: the acyltransferase BtrH first transfers
a synthetic acyl-N-acetylcysteamine thioester (acyl-SNAC) sub-
strate, -L-glutamyl-AHBA-SNAC, onto the aminoglycosides, and
the resulting acyl-aminoglycoside products are treated with the
-L-glutamylcyclotransferase BtrG to give AHBA-aminoglycoside
species. In contrast, we were able to produce AHBA-conjugated kana-
mycins 15 and 17 directly in recombinants. To our knowledge, this is
the first report of the direct fermentative production of the semisyn-
thetic amikacin (15). The new aminoglycoside 17 has more potent anti-
bacterial activity than 15 against both the Kan
R
and Amk
R
test strains
(Supplementary Table 20). The sole difference in structure between
15 and 17 is in the functional group attached to the C6 (Fig. 5c),
indicating that the removal of the 6-amino group in 15, which serves
as a target for aminoglycoside 6-acetyltransferase
10
, confers an effect
against a 15-resistant pathogen. Another expected advantage of 17
would be an improved toxicity profile because the side effects of
aminoglycoside therapeutics decrease with a decreasing number of
attached amino groups
42
. It is quite likely that this molecule can be
a candidate for further development of next- generation aminogly-
cosides on the basis of its structural resemblance to amikacin and
reduced number of amino groups. In addition, the direct biosyn-
thesis of tobramycin (16) via an engineered pathway would be more
economical than conventional procedures, which currently involve
hydrolysis of 6-O-carbamoyltobramycin that constitutes only 9% of
the total aminoglycosides produced by wild-type S. tenebrarius
43
.
Although the results presented herein were obtained in a hetero-
logous host, application of the same strategy in industrial strains should
enable the practical mass production of clinically valuable antibiotics
directly by fermentation. In addition, a detailed understanding of the
biosynthetic pathways of kanamycins is sure to open new opportuni-
ties to assist in comprehending the biosynthesis of other related amin-
oglycosides and to form the basis for pathway engineering toward the
next generation of antibiotics (exemplified by 17).
METHODS
General information. Bacterial strains, culture conditions, expression and purifica-
tion of proteins, and materials are described in Supplementary Methods.
Construction of expression plasmids and mutants. Details regarding DNA
isolation and manipulation as well as the construction of expression plasmids
(Supplementary Tables 1,2) and mutants are described in Supplementary Methods.
Genes used in this study are from kanamycin (AJ582817), neomycin (AJ786317 and
AJ629247), gentamicin (AJ575934), tobramycin (AJ810851), butirosin (AJ494863
and AB097196) and apramycin (AJ629123) biosynthetic gene clusters.
Analysis of kanamycin biosynthetic intermediates and their analogs. Kanamycin
biosynthetic intermediates and their analogs produced by recombinant strains of
S. venezuelae harboring diverse combinations of aminoglycoside biosynthetic gene
sets were extracted from the fermentation broth using the OASIS MCX (Waters)
solid-phase extraction cleanup procedure (described in Supplementary Methods)
and were then analyzed by HPLC-ESI-MS/MS; samples were separated on an XTerra
MS C
18
column (50 2.1 mm, 3.5 m, Waters) interfaced with a Waters/Micromass
Quattro micro/MS instrument tracing by MS/MS using a gradient of acetonitrile and
10 mM heptafluorobutyric acid (Fluka) over 45 min. Tracing was done by MS/MS
operated in multiple-reaction monitoring mode. Quantification was conducted by
choosing mass pairs specific for the selected analytes to detect the transition from
parent ion to product ion: 163 > 84 for 1; 366 > 163 for 2; 324 > 163 for 3 and 10;
323 > 163 for 4; 486 > 163 for 5, 12 and 14; 485 > 163 for 6, 8, and 13; 484 > 163 for
7; 325 > 163 for 9; 487 > 163 for 11; 586 > 264 for 15; 468 > 163 for 16; 587 > 264 for
17; 307 > 163 for 18; and 469 > 163 for 19 and 20. Three separate cultivations and
independent extractions were performed.
Isolation and structural identification of kanamycin biosynthetic intermediates and
their analogs. Details regarding the isolation and characterization of products obtained
from in vivo samples are described in Supplementary Methods. The retention behav-
iors of products on HPLC-ESI-MS/MS and data including MS/MS and NMR spectra
are described in Supplementary Figures 221 and Supplementary Tables 318.
In vitro reactions using cell-free extracts. CFEs of S. venezuelae were prepared by
glass-bead homogenization (described in Supplementary Methods). The resulting
CFEs were suspended in Tris buffer (pH 7.5) containing 100 mM Tris-HCl (pH 7.6),
10 mM MgCl
2
, 6 mM 2-mercaptoethanol and 1 mM phenylmethylsulfonyl fluoride
(Sigma), and their protein concentrations were corrected. KanF glycosyltransfer
reactions were initiated by adding 100 M 1 to the CFE obtained from a recom-
binant host expressing only kanF and 200 M UDP-Glc (Sigma), UDP-GlcNAc
(GeneChem) or UDP-GlcN (described in Supplementary Methods) as the cosub-
strate. Reactions to determine the activity of KanI-KacL on 6-amination were carried
out by supplementing CFE from the recombinant host expressing kanI-kacL with
100 M of the 6-hydroxy pseudodisaccharides or pseudotrisaccharides 3, 6, 9 or 12.


2
0
1
1

N
a
t
u
r
e

A
m
e
r
i
c
a
,

I
n
c
.


A
l
l

r
i
g
h
t
s

r
e
s
e
r
v
e
d
.
NATURE CHEMICAL BIOLOGY | VOL 7 | NOVEMBER 2011 | www.nature.com/naturechemicalbiology 851
ARTICLE
NATURE CHEMICAL BIOLOGY DOI: 10.1038/NCHEMBIO.671
After incubation for 2 h at 30 C, the reaction was quenched with ice-cold phenol/
chloroform/isoamyl alcohol (25:24:1; Sigma) and centrifuged at 18,000g for 5 min.
The supernatant containing the product of interest was extracted using OASIS MCX
SPE cleanup (described in Supplementary Methods), reconstituted with 100 l of
water, and was then analyzed by HPLC-ESI-MS/MS as described above.
Reactions to examine the activity of KanC-KanD were sequentially conducted
using CFE from two different recombinant strains of S. venezuelae harboring either
kanE or kanC-kanD. CFE containing KanE (or KanC-KanD) was incubated with
200 M UDP-Glc plus 100 M 3 or 4 under the same conditions described above,
and then the reaction was quenched. Supernatants containing products obtained from
the first reaction were incubated for 2 h at 30 C with CFE containing KanC-KanD
(or KanE), and then the reaction was quenched again. The resulting supernatant was
subjected to the same cleanup procedure described above and was then analyzed by
HPLC-ESI-MS. Sequential reactions using both CFEs were also carried out using the
pseudodisaccharides 9 and 10. Independent experiments were performed in duplicate.
Kinetic analysis of KanF and KanE. KanF (6 M) was incubated with 1 and UDP-
Glc or UDP-GlcNAc, respectively, in 100 l reaction buffer (75 mM Tris-HCl,
10 mM MgCl
2
and 1 mg ml
1
BSA) at 37 C for 1 h. For the measurement of the
K
m
value for 1, the concentration of UDP-Glc and UDP-GlcNAc was fixed at 3 mM,
while the concentration of 1 was varied from 0.05 mM to 1.5 mM. In addition, to
determine the kinetic parameters of UDP-Glc or UDP-GlcNAc, the concentration
of 1 was maintained at 1.5 mM, while the concentration of UDP-Glc and UDP-
GlcNAc was varied from 1.0 mM to 6.0 mM. On the other hand, KanE (30 M)
was incubated with 3 and UDP-Glc or UDP-Kns (described in Supplementary
Methods), respectively, in the same buffer as described above. For the determina-
tion of the K
m
for 3, the concentration of UDP-Glc and UDP-Kns was kept at
1.0 mM, while the concentration of 3 was varied from 0.2 mM to 1.0 mM. The
parameters for UDP-Glc or UDP-Kns were determined by varying their concentra-
tions from 0.2 to 2.0 mM, while the concentration of 3 was fixed at 1.0 mM. All
reactions were quenched, extracted using SPE and finally subjected to HPLC-ESI-MS
analysis as described above. Independent experiments were performed in triplicate.
Measurement of MIC of kanamycin biosynthetic intermediates and their
analogs. MICs of various kanamycin-related pseudodisaccharides and pseudo-
trisaccharides as well as AHBA-conjugated analogs (except 16, 18, 19 and 20)
were determined using the microdilution method of the Clinical and Laboratory
Standard Institute
44
. Gram-negative E. coli and P. aeruginosa type strains and clini-
cal isolates were grown at 30 C in Mueller-Hinton broth. Serial two-fold dilutions
of aminoglycosides were carried out to give final concentrations between 0.25 g ml
1

and 128 g ml
1
, and an aliquot of water was used as a negative control. The growth
of test strains was monitored at 600 nm using a Labsystems Bioscreen C reader,
and the MIC was determined as the lowest concentration of the aminoglycoside
diluted in broth medium that inhibited the growth of the test bacterium.
Computational methods. Details regarding homology modeling, docking and
molecular dynamics simulation are described in Supplementary Methods.
Received 24 June 2010; accepted 28 July 2011;
published online 9 October 2011
References
1. Magnet, S. & Blanchard, J.S. Molecular insights into aminoglycoside action
and resistance. Chem. Rev. 105, 477498 (2005).
2. Nagabhushan, T.L., Miller, G.H. & Weinstein, M.J. Structure-activity
relationships in aminoglycoside-aminocyclitol antibiotics. in Te
Aminoglycosides: Microbiology, Clinical Use, and Toxicology (eds. Whelton, A. &
Neu, H.C.) 327 (Marcel Dekker, New York, 1982).
3. Jones, D., Metzger, H.J., Schatz, A. & Waksman, S.A. Control of Gram-
negative bacteria in experimental animals by streptomycin. Science 100,
103105 (1944).
4. Umezawa, H. et al. Production and isolation of a new antibiotic: kanamycin.
J. Antibiot. (Tokyo) 10, 181188 (1957).
5. Weinstein, M.J. et al. Gentamicin, a new antibiotic complex from
Micromonospora. J. Med. Chem. 6, 463464 (1963).
6. Stark, W.M., Hoehn, M.M. & Knox, N.G. Nebramycin, a new broad-spectrum
antibiotic complex. I. Detection and biosynthesis. Antimicrob. Agents
Chemother. (Bethesda) 7, 314323 (1967).
7. Waksman, S.A. & Lechevalier, H.A. Neomycin, a new antibiotic active against
streptomycin-resistant bacteria, including tuberculosis organisms. Science 109,
305307 (1949).
8. Howells, J.D. et al. Butirosin, a new aminoglycosidic antibiotic complex:
bacterial origin and some microbiological studies. Antimicrob. Agents
Chemother. 2, 7983 (1972).
9. Fischbach, M.A. & Walsh, C.T. Antibiotics for emerging pathogens. Science 325,
10891093 (2009).
10. Kondo, S. & Hotta, K. Semisynthetic aminoglycoside antibiotics: development
and enzymatic modifcations. J. Infect. Chemother. 5, 19 (1999).
11. Kudo, F. & Eguchi, T. Biosynthetic enzymes for the aminoglycosides butirosin
and neomycin. Methods Enzymol. 459, 493519 (2009).
12. Kudo, F. et al. Molecular cloning of the gene for the key carbohydrate-
forming enzyme in the biosynthesis of 2-deoxystreptamine-containing
aminocyclitol antibiotics and its comparison with dehydroquinate synthase.
J. Antibiot. (Tokyo) 52, 559571 (1999).
13. Huang, F. et al. Te neomycin biosynthetic gene cluster of Streptomyces
fradiae NCIMB 8233: characterisation of an aminotransferase involved
in the formation of 2-deoxystreptamine. Org. Biomol. Chem. 3, 14101418
(2005).
14. Kudo, F. et al. Biosynthesis of 2-deoxystreptamine by three crucial enzymes in
Streptomyces fradiae NBRC 12773. J. Antibiot. (Tokyo) 58, 766774 (2005).
15. Wehmeier, U.F. & Piepersberg, W. Enzymology of aminoglycoside biosynthesis-
deduction from gene clusters. Methods Enzymol. 459, 459491 (2009).
16. Fan, Q., Huang, F., Leadlay, P.F. & Spencer, J.B. Te neomycin biosynthetic
gene cluster of Streptomyces fradiae NCIMB 8233: genetic and biochemical
evidence for the roles of two glycosyltransferases and a deacetylase.
Org. Biomol. Chem. 6, 33063314 (2008).
17. Park, J.W. et al. Genetic dissection of the biosynthetic route to gentamicin A
2

by heterologous expression of its minimal gene set. Proc. Natl. Acad. Sci. USA
105, 83998404 (2008).
18. Truman, A.W., Huang, F., Llewellyn, N.M. & Spencer, J.B. Characterization of
the enzyme BtrD from Bacillus circulans and revision of its functional
assignment in the biosynthesis of butirosin. Angew. Chem. Int. Edn Engl. 46,
14621464 (2007).
19. Tuy, M.L. et al. Expression of 2-deoxy-scyllo-inosose synthase (kanA) from
kanamycin gene cluster in Streptomyces lividans. Biotechnol. Lett. 27, 465470
(2005).
20. Jnawali, H.N., Subba, B., Liou, K. & Sohng, J.K. Functional characterization of
kanB by complementing in engineered Streptomyces fradiae neo6:tsr.
Biotechnol. Lett. 31, 869875 (2009).
21. Nepal, K.K., Oh, T.J. & Sohng, J.K. Heterologous production of paromamine
in Streptomyces lividans TK24 using kanamycin biosynthetic genes from
Streptomyces kanamyceticus ATCC12853. Mol. Cells 27, 601608 (2009).
22. Kudo, F., Sucipto, H. & Eguchi, T. Enzymatic activity of a glycosyltransferase
KanM2 encoded in the kanamycin biosynthetic gene cluster. J. Antibiot.
(Tokyo) 62, 707710 (2009).
23. Kojima, M., Yamada, Y. & Umezawa, H. Studies on the biosynthesis of
kanamycins. Part I. Incorporation of
14
C-glucose or
14
C-glucosamine into
kanamycins and kanamycin-related compounds. Agric. Biol. Chem. 32,
467473 (1968).
24. Kojima, M., Yamada, Y. & Umezawa, H. Studies on the biosynthesis of
kanamycins. Part II. Incorporation of the radioactive degradation products of
kanamycin A or related metabolites into kanamycin A. Agric. Biol. Chem. 33,
11811185 (1969).
25. Llewellyn, N.M. & Spencer, J.B. Biosynthesis of 2-deoxystreptamine-
containing aminoglycoside antibiotics. Nat. Prod. Rep. 23, 864874
(2006).
26. Piepersberg, W., Aboshanab, K.M., Schmidt-Beiner, H. & Wehmeier, U.F.
Te biochemistry and genetics of aminoglycoside producers in
Aminoglycoside Antibiotics: From Chemical Biology to Drug Discovery
(ed. Arya, D.P.) 15118 (John Wiley and Sons, 2007).
27. Kudo, F. & Eguchi, T. Biosynthetic genes for aminoglycoside antibiotics.
J. Antibiot. (Tokyo) 62, 471481 (2009).
28. Kharel, M.K. et al. A gene cluster for biosynthesis of kanamycin from
Streptomyces kanamyceticus: comparison with gentamicin biosynthetic gene
cluster. Arch. Biochem. Biophys. 429, 204214 (2004).
29. Tapa, L.P. et al. Heterologous expression of the kanamycin biosynthetic gene
cluster (pSKC2) in Streptomyces venezuelae YJ003. Appl. Microbiol. Biotechnol.
76, 13571364 (2007).
30. Lemieux, R.U., Nagabhushan, T.L., Clemetson, K.J. & Tucker, L.C.N. Te
synthesis of kanamycin analogs. I. -d-glucopyranosyl derivatives of
deoxystreptamine. Can. J. Chem. 51, 5366 (1973).
31. Huang, F. et al. Elaboration of neosamine rings in the biosynthesis of
neomycin and butirosin. ChemBioChem 8, 283288 (2007).
32. Yamada, Y., Kojima, M. & Umezawa, H. Studies on structure and activity of
kanamycins and kanamycin-related compounds. Jpn. J. Med. Sci. Biol. 21,
223224 (1968).
33. Guo, J. & Frost, J.W. Kanosamine biosynthesis: a likely source of the
aminoshikimate pathways nitrogen atom. J. Am. Chem. Soc. 124, 1064210643
(2002).
34. Chang, L.T., Behr, D.A. & Elander, D.A. Te efects of intercalating agents on
kanamycin production and other phenotypic characteristics in Streptomyces
kanamyceticus. Dev. Ind. Microbiol. Series 21, 233243 (1980).
35. Adams, E., De Bie, E., Roets, E. & Hoogmartens, J. Isolation and
identifcation of a new kanamycin component. Eur. J. Pharm. Sci. 4 (suppl. 1),
S110 (1996).
36. Koch, K.F. et al. Structures of some of the minor aminoglycoside factors of
the nebramycin fermentation. J. Org. Chem. 43, 14301434 (1978).
37. Yokoyama, K., Yamamoto, Y., Kudo, F. & Eguchi, T. Involvement of two
distinct N-acetylglucosaminyltransferases and a dual-function deacetylase in
neomycin biosynthesis. ChemBioChem 9, 865869 (2008).


2
0
1
1

N
a
t
u
r
e

A
m
e
r
i
c
a
,

I
n
c
.


A
l
l

r
i
g
h
t
s

r
e
s
e
r
v
e
d
.
852 NATURE CHEMICAL BIOLOGY | VOL 7 | NOVEMBER 2011 | www.nature.com/naturechemicalbiology
ARTICLE
NATURE CHEMICAL BIOLOGY DOI: 10.1038/NCHEMBIO.671
(R0A-2008-000-20030-0) and Global Frontier Program for Intelligent Synthetic
Biology through the National Research Foundation of Korea (NRF), NRF grants
(2010-0001487 and 2010-0028193) funded by the Ministry of Education, Science &
Technology and the Marine and Extreme Genome Research Center Program of the
Ministry of Land, Transportation and Maritime Affairs, Republic of Korea. J.W.P.
gratefully acknowledges a grant (20100623) from the Technology Development
Program for Agriculture and Forestry, Ministry for Food, Agriculture and Fisheries,
Republic of Korea.
Author contributions
J.W.P., S.R.P., J.K.S. and Y.J.Y. designed research and wrote the paper; J.W.P., S.R.P.,
K.K.N., A.R.H., Y.H.B., Y.J.Y., E.J.K., E.M.K. and D.K. performed research; and J.W.P.,
S.R.P. and Y.J.Y. analyzed data.
Competing financial interests
The authors declare competing financial interests: details accompany the full-text HTML
version of the paper at http://www.nature.com/naturechemicalbiology/.
Additional information
Supplementary information and chemical compound information is available online at
http://www.nature.com/naturechemicalbiology/. Reprints and permissions information
is available online at http://www.nature.com/reprints/index.html. Correspondence and
requests for materials should be addressed to J.K.S. or Y.J.Y.
38. Kharel, M.K. et al. Isolation and characterization of the tobramycin
biosynthetic gene cluster from Streptomyces tenebrarius. FEMS Microbiol. Lett.
230, 185190 (2004).
39. Llewellyn, N.M. & Spencer, J.B. Chemoenzymatic acylation of aminoglycoside
antibiotics. Chem. Commun. (Camb.) 32, 37863788 (2008).
40. Kondo, S. et al. Synthesis and properties of kanamycin C derivatives active
against resistant bacteria. J. Antibiot. (Tokyo) 30, 11501152 (1977).
41. Umezawa, S. et al. Synthesis of 3-deoxykanamycin efective against
kanamycin-resistant Escherichia coli and Pseudomonas aeruginosa. J. Antibiot.
(Tokyo) 24, 274276 (1971).
42. Hainrichson, M., Nudelman, I. & Baasov, T. Designer aminoglycosides: the
race to develop improved antibiotics and compounds for the treatment of
human genetic diseases. Org. Biomol. Chem. 6, 227239 (2008).
43. Park, J.W. et al. Te nebramycin aminoglycoside profles of Streptomyces
tenebrarius and their characterization using an integrated liquid
chromatography-electrospray ionization-tandem mass spectrometric analysis.
Anal. Chim. Acta 661, 7684 (2010).
44. Clinical and Laboratory Standard Institute. Methods for Dilution Antimicrobial
Susceptibility Tests for Bacteria that Grow Aerobically: Approved Standard:
M07A8 (National Committee for Clinical Laboratory Institute, Wayne,
Pennsylvania, USA, 2009).
Acknowledgments
We thank E. Cundliffe for discussions and for critically reading this manuscript.
This work was supported by the National Research Laboratory (NRL) program


2
0
1
1

N
a
t
u
r
e

A
m
e
r
i
c
a
,

I
n
c
.


A
l
l

r
i
g
h
t
s

r
e
s
e
r
v
e
d
.

You might also like