The neomycin biosynthetic gene cluster of Streptomyces fradiae NCIMB 8233: Genetic and biochemical evidence for the roles of two glycosyltransferases and a deacetylase

Article abstract of DOI:10.1039/b808734b

An efficient protocol has been developed for the genetic manipulation of Streptomyces fradiae NCIMB 8233, which produces the 2-deoxystreptamine (2-DOS)-containing aminoglycoside antibiotic neomycin. This has allowed the in vivo analysis of the respective roles of the glycosyltransferases Neo8 and Neo15, and of the deacetylase Neo16 in neomycin biosynthesis. Specific deletion of each of the neo8, neo15 and neo16 genes confirmed that they are all essential for neomycin biosynthesis. The pattern of metabolites produced by feeding putative pathway intermediates to these mutants provided unambiguous support for a scheme in which Neo8 and Neo15, whose three-dimensional structures are predicted to be highly similar, have distinct roles: Neo8 catalyses transfer of N-acetylglucosamine to 2-DOS early in the pathway, while Neo15 catalyses transfer of the same aminosugar to ribostamycin later in the pathway. The in vitro substrate specificity of Neo15, purified from recombinant Escherichia coli, was fully consistent with these findings. The in vitro activity of Neo16, the only deacetylase so far recognised in the neo gene cluster, showed that it is capable of acting in tandem with both Neo8 and Neo15 as previously proposed. However, the deacetylation of N-acetylglucosaminylribostamycin was still observed in a strain deleted of the neo16 gene and fed with suitable pathway precursors, providing evidence for the existence of a second enzyme in S. fradiae with this activity.

Full text of DOI:10.1039/b808734b

View Article Online / Journal Homepage / Table of Contents for this issue
PAPER
www.rsc.org/obc | Organic & Biomolecular Chemistry
The neomycin biosynthetic gene cluster of Streptomyces fradiae NCIMB
8233: genetic and biochemical evidence for the roles of two
glycosyltransferases and a deacetylase†
Qingzhi Fan,a,b Fanglu Huang,b Peter F. Leadlay*a and Jonathan B. Spencer‡b
Received 23rd May 2008, Accepted 20th June 2008
First published as an Advance Article on the web 25th July 2008
DOI: 10.1039/b808734b
An efficient protocol has been developed for the genetic manipulation of Streptomyces fradiae NCIMB
8233, which produces the 2-deoxystreptamine (2-DOS)-containing aminoglycoside antibiotic
neomycin. This has allowed the in vivo analysis of the respective roles of the glycosyltransferases Neo8
and Neo15, and of the deacetylase Neo16 in neomycin biosynthesis. Specific deletion of each of the
neo8, neo15 and neo16 genes confirmed that they are all essential for neomycin biosynthesis. The
pattern of metabolites produced by feeding putative pathway intermediates to these mutants provided
unambiguous support for a scheme in which Neo8 and Neo15, whose three-dimensional structures are
predicted to be highly similar, have distinct roles: Neo8 catalyses transfer of N-acetylglucosamine to
2-DOS early in the pathway, while Neo15 catalyses transfer of the same aminosugar to ribostamycin
later in the pathway. The in vitro substrate specificity of Neo15, purified from recombinant Escherichia
coli, was fully consistent with these findings. The in vitro activity of Neo16, the only deacetylase so far
recognised in the neo gene cluster, showed that it is capable of acting in tandem with both Neo8 and
Neo15 as previously proposed. However, the deacetylation of N-acetylglucosaminylribostamycin was
still observed in a strain deleted of the neo16 gene and fed with suitable pathway precursors, providing
evidence for the existence of a second enzyme in S. fradiae with this activity.
modification of the structure has inseveralcases produced valuable
Introduction
semisynthetic aminoglycosides such as amikacin9 that circumvent
2-Deoxystreptamine (2-DOS)-containing antibiotics1,2 comprise
such resistance mechanisms. This has prompted increasing interest
the largest subgroup of the aminoglycoside class of antimicrobial
in the possibility of producing novel aminoglycoside antibiotics
compounds, which includes the clinically-useful gentamicin and
by engineering of the biosynthetic pathway. For this approach
neomycin. Neomycin, first isolated from Streptomyces fradiae and
to be applied effectively, we will require a much more detailed
understanding of the genes and enzymes involved.
Streptomyces albogriseus in 1949,3 is a broad-spectrum antibiotic
active against both Gram-positive and Gram-negative bacteria.
Neomycin obtained from fermentation comprises a mixture
Like most aminoglycosides, it targets the bacterial 30S ribosomal
of three structurally-related compounds: neomycin A (usu-
subunit and inhibits protein synthesis by inducing codon mis-
ally referred to as neamine), neomycin B and neomycin C
reading and interfering with initiation and translocation.4,5 The
(Fig. 1). Neamine consists of 2,6-diamino-2,6-dideoxy-D-glucose
aminoglycosides are significantly nephrotoxic and ototoxic and are
(neosamine C, ring II) attached by a glycosidic linkage to
therefore normally reserved for treatment of deadly infections such
as those caused by Mycobacterium tuberculosis and methicillin-
resistant Staphylococcus aureus (MRSA). More recent work has
indicated the potential of neomycin and its analogues as a new
class of anti-HIV agents,6,7 and the demonstration of their in vitro
anti-angiogenesis activity has led to their evaluation in anticancer
drug development.8
the aminocyclitol aglycone 2-deoxystreptamine (2-DOS, ring I).
Neomycin B has a neamine core to which 2,6-diamino-2,6-
dideoxy-L-idose (neosamine B, ring IV) and D-ribose (ring III)
are attached via ring I. Neomycin C and neomycin B are identical
except for the different configuration of the aminomethyl group
at C5ꢀꢀꢀ in ring IV. The paromomycins, which are produced by
As with other broad-spectrum antibiotics, the emergence among
human pathogens of antibiotic resistance to aminoglycosides
has blunted their clinical effectiveness. Encouragingly, chemical
aDepartment of Biochemistry, University of Cambridge, 80 Tennis Court
Road, Cambridge, CB2 1GA, United Kingdom. E-mail: pfl10@mole.bio.
cam.ac.uk; Fax: +44 1223 766002; Tel: +44 1223 766041
bUniversity Chemical Laboratory, University Chemical Laboratory, Lens-
field Road, Cambridge, CB2 1EW, United Kingdom
† Electronic supplementary information (ESI) available: Fig. S1–S4; Table
S1. See DOI: 10.1039/b808734b
‡ Deceased 6th April 2008.
Fig. 1 Structure and ring numbering of neomycins and paromomycins.
3306 | Org. Biomol. Chem., 2008, 6, 3306–3314
This journal is
The Royal Society of Chemistry 2008
©

View Article Online
Streptomyces rimosus,10 have been reported as minor components
of the fermentation in S. fradiae.11
acting to remove the acetyl group on N-acetylglucosamine moi-
eties of aminoglycoside intermediates, based on its significant
sequence similarity to BtrD which has also been recently un-
veiled as a deacetylase.16 Given this insight, it would appear
that neomycin biosynthesis from 2-DOS requires two distinct
cycles of transfer of an N-acetyl-D-glucosamine (GlcNAc) from
UDP-N-acetylglucosamine (UDP-GlcNAc) (to 2-DOS in cycle
1 and to ribostamycin in cycle 2) followed by deacetylation,
and C6 modification of the newly-introduced sugar (Scheme 1).
This raises the intriguing question of whether different enzymes
carry out the comparable step in each cycle, or whether a
single enzyme has the requisite dual specificity. Recombinant
Neo8 and Neo15 were recently reported to catalyse in vitro
the specific transfer of GlcNAc to 2-DOS and to ribostamycin
respectively, while Neo16 in vitro was reported to be active
in deacetylation of both N-acetylparomamine and 2ꢀꢀꢀ-N-acetyl-
6ꢀꢀꢀ-hydroxyneomycin (Nꢀ-acetylglucosaminylribostamycin).17 We
report here our own independent findings on the specificity
of Neo15 and Neo16 in vitro, as well as the development of
an efficient method for genetic manipulation of the neomycin-
producing S. fradiae. This has allowed us to demonstrate by an in
vivo approach that the two glycosyltransferases Neo8 and Neo15
must act specifically and in that order, at different points in the
pathway. In contrast, Neo16 is sufficient in vitro to accomplish
deacetylation in both cycles, but it is not essential in vivo
for the second cycle of deacetylation. This provides the first
Extensive early studies on the biosynthesis of the 2-DOS-
derived aminoglycosides, including the isolation of compounds
from blocked mutants of S. fradiae, and the feeding of labelled
putative precursors to 2-DOS idiotrophs, have strongly implicated
paromamine, neamine and ribostamycin as intermediates in the
formation of neomycin, and provided the first evidence for
the order of assembly.1,2 More recently, the biosynthetic gene
cluster for neomycin has been cloned, by four research groups
independently, from S. fradiae NCIMB 8233 (or the equivalent
strain deposited elsewhere under other reference numbers) and
sequenced (accession numbers: AJ843080,12 AJ629247, AJ786317
and AB21195913). This information, together with comparison
of the gene organisation with that of biosynthetic gene clusters
for other aminoglycosides, has allowed a detailed biosynthetic
pathway to be proposed.12,13 The enzymes responsible for for-
mation of the aglycone 2-DOS, in four steps from D-glucose-6-
phosphate, have been fully characterised in vitro.12–14 In contrast,
many of the details of the proposed pathway from 2-DOS to
neomycin (Scheme 1) remain to be clarified. For example, Neo16
(NeoL), which shows significant sequence similarity to BtrD
in the butirosin gene cluster, was initially suggested to provide
NDP-glucose/NDP-glucosamine sugar donors for glycosylation
of 2-DOS to form paromamine.15 However, as summarised in
Scheme 1, Neo16 is now thought instead to act as a deacetylase
Scheme 1 Proposed biosynthetic pathway of neomycin C. PRPP: 5-phosphoribosyl-1-diphosphate.
The Royal Society of Chemistry 2008 Org. Biomol. Chem., 2008, 6, 3306–3314 | 3307
This journal is
©

View Article Online
evidence for overlapping enzymatic activities in aminoglycoside
biosynthesis.
expressed using the vector pET28a(+) in E. coli and with plasmid
pCJW9320 in Streptomyces lividans. The protein was expressed
poorly, and was found to be insoluble under a wide range of
conditions tried. While this manuscript was in preparation, it was
reported that Neo8 (NeoD) could be successfully co-expressed
with E. coli chaperone proteins GroES and GroEL.17 Although
the enzyme could not be purified, its activity in the E. coli lysate
revealed an apparent specificity for 2-DOS as the acceptor of a
GlcNAc group, while ribostamycin, neamine and paromamine
were not substrates.17 Thus, in vitro and in vivo data are in
agreement in suggesting that a distinct, second glycosyltransferase
is required to transfer a GlcNAc group to ribostamycin.
Results and discussion
Attachment of N-acetyl-D-glucosamine to 2-deoxystreptamine
Since the early observation18 that a mutant of S. fradiae could be
isolated which only produced ribostamycin, it has been considered
likely that the two glycosylation steps in neomycin biosynthesis
require two independent glycosyltransferases. However, initial
inspection of the proteins encoded by the neomycin gene cluster
only readily identified Neo8 (NeoD) as a candidate glycosyl-
transferase catalysing one or both of the required transfers of an
aminohexose in neomycin biosynthesis.12,13 Neo8 has homologues
in many aminoglycoside biosynthetic gene clusters that share
high percentage amino acid sequence identities: most strikingly
RibM (ribostamycin, 90%); but also LivM (lividomycin, 72%),
ParM (paromomycin, 69%), TobM1 (tobramycin, 64%), KanM1
(kanamycin, 63%), AprM (apramycin, 61%) and GenM1 (gen-
tamicin, 54%). All of these aminoglycoside pathways require
attachment of at least one hexosamine sugar.
The neo8 gene of S. fradiae was knocked out by targeted in-
frame deletion, which removed 840 bp of the gene (out of a total
of 1185 bp). The identity of the mutant was confirmed by PCR
screening, followed by sequencing of the PCR products (Fig. S1-b).
The mutant, Dneo8, failed to produce neomycin. However, Dneo8
did produce neomycin when the medium was supplemented with
neamine (0.1 mg mL−1) or ribostamycin (0.2 mg mL−1) (Fig. S1-d
and Table S1) but not when the medium was supplemented with
2-DOS (0.1 mg mL−1). This provides direct evidence that Neo8
is essential for the first transfer of UDP-GlcNAc but not for the
second. Feeding of paromamine (0.2 mg mL−1) led, interestingly, to
the production of paromomycin in good yield (Fig. S1-e and Table
S1), accompanied by minor amounts of neomycin. This unusual
production of paromomycin likely reflects a broad substrate
specificity of the downstream enzymes, especially those involved in
ribosylation. In agreement with this, it has previously been shown
that BtrL, the homologue of Neo17 in butirosin biosynthesis,
acts as a phosphoribose transferase in vitro on both neamine
and paromamine substrates.19 In the S. fradiae Dneo8 mutant,
Neo17 apparently competes effectively for the added paromamine
against the dehydrogenase and aminotransferase enzymes Neo11
and Neo18, which normally convert paromamine to neamine. As a
control, ribostamycin and paromamine were also fed to wild-type
S. fradiae, and in each case there was an approximately three-fold
decrease in neomycin production. Production of paromomycin
was again observed with the addition of paromamine.
Attachment of N-acetyl-D-glucosamine to ribostamycin
Further, extensive BLAST searches against the open reading
frames in the neomycin biosynthetic cluster12 revealed Neo15
as a second candidate hexosaminyltransferase. Its closest three
potential homologues in the available gene clusters of aminoglyco-
sides are RibF (ribostamycin, 77%), LivF (lividomycin, 65%) and
ParF (paromomycin, 60%). Both lividomycin and paromomycin
possess the same ring IV as neomycin, a further hint that Neo15
could be the glycosyltransferase responsible for its addition. The
neo15 gene was knocked out by an in-frame deletion, which
removed 759 bp of the gene (out of a total of 1101 bp). The
mutation was confirmed by PCR screening and by sequencing
of the PCR products (Fig. S2-c). Neomycin production was
abolished in Dneo15. Instead, ribostamycin was detected in the
culture supernatant (Fig. S2-a), indicating that the biosynthetic
pathway has been disrupted at the stage of the addition of the
last sugar (ring IV). This result agrees with the glycosyltransferase
activity proposed for Neo15. Complementation of the Dneo15
mutant was carried out in three alternative ways: first, by using
a plasmid (pQZ54) containing neo15 under the control of the
ermE promoter; secondly, using a plasmid (pQZ61) housing
neo15 under the control of its natural promoter together with
the other genes in the operon (neo12–neo16); or, finally, using
a plasmid (pQZ62) in which only neo15 is placed downstream
of its natural promoter. In each case, neomycin production
was restored (Fig. S2-b), consistent with Neo15 catalysing the
specific transfer of the GlcNAc moiety from UDP-GlcNAc to
ribostamycin.
To confirm the role of the neo15 gene product in neomycin
biosynthesis, the Neo15 protein was expressed with an N-terminal
His6-tag in E. coli BL21 (DE3) and purified by nickel chelate-
affinity chromatography in a yield of 0.9 mg L−1. The mass of
the purified protein was determined to be 41 295 Da (Fig. S2-d)
(including the His6-tag, and with loss of N-terminal methionine)
by LC–ESI-MS (calculated, 41 289 Da). A peak (41 466 Da)
corresponding to the a-N-gluconoylated21 Neo15 (calculated,
41 467 Da) was also observed. The purified His6-tagged Neo15
was tested for its glycosyltransferase activity in vitro. Incubation
of Neo15 with UDP-GlcNAc and ribostamycin overnight at 30 ◦C
converted all ribostamycin to N-acetylglucosaminylribostamycin
([M + H]+ = 658) as observed by LC–ESI-MS (Fig. 2). The
same conversion was detected using a cell-free extract of E. coli
carrying the Neo15 expression plasmid (pQZ39), but not with
an extract of E. coli containing an empty pET28a(+) vector.
Neo15 was found to be relatively stable, retaining significant
The Dneo8 mutant was complemented by introduction of a
plasmid (pQZ53) in which neo8 is under the control of the
constitutively-expressed ermE promoter; or alternatively by intro-
duction of a plasmid (pQZ60) in which neo8 is under the control of
its natural promoter. In either case, production of neomycin was
restored (Fig. S1-c), albeit at a reduced level compared to wild-
type, confirming that the loss of neomycin production in Dneo8
was not caused by an unintended polar effect on the expression of
any gene downstream of neo8.
Heterologous expression of the Neo8 protein was also at-
tempted, in order to test its substrate specificity in vitro. Neo8 was
3308 | Org. Biomol. Chem., 2008, 6, 3306–3314
This journal is
The Royal Society of Chemistry 2008
©

View Article Online
Fig. 2 LC–MS analysis of Neo15 glycosyltransferase activity in vitro. (a) Control with the substrate ribostamycin but without Neo15. (b) Conversion of
ribostamycin to N-acetylglucosaminylribostamycin by Neo15.
◦
enzymatic activity in Tris-HCl buffer, pH 8.4 at 4 C even after
37 days. The in vitro identification of Neo15 (NeoK) as the
glycosyltransferase required to transfer GlcNAc to ribostamycin
has also recently been independently reported by Eguchi and
colleagues.17
family (7868 entries at time of writing), all of which are retaining
glycosyltransferases and have the GT-B fold.
Both Neo8 and Neo15 show significant sequence matches
to four GT-B glycosyltransferases from Family 4 whose crystal
structures have been determined. WaaG, also known as RfaG, is
an a-1,3-glucosyltransferase responsible for transferring glucose
from UDP-D-glucose onto L-glycerol-D-manno-heptose II in the
biosynthesis of the core structure of lipopolysaccharide (LPS).23
AviGT4 is one of several glycosyltransferases required for the
biosynthesis of the polyketide antibiotic avilamycin A. WaaG and
AviGT4 display only 16% sequence identity, but their GT-B fold
(two distinct domains each with a “Rossmann-like” (b/a/b) fold)
and their conserved catalytic residues reveal their close structural
resemblance.23 PimA, the third Family 4 glycosyltransferase,
is an essential phosphatidylinositol mannosyltransferase from
Mycobacterium tuberculosis24 and MshA, the fourth, is involved
in mycothiol biosynthesis in Corynebacterium glutamicum.25 Al-
though they all utilise different NDP-hexose substrates, the C-
terminal domain housing the donor binding site is particularly
highly conserved, and contains a signature sequence which is
universally found in the retaining glycosyltransferases of CAZy
Family 4.26 This sequence contains two conserved glutamate
residues (Fig. 3), spaced eight residues apart, and it is also present
in the C-terminal region of both Neo8 and Neo15. Interestingly,
MshA and the following enzyme in the mycothiol biosynthetic
pathway (MshB) together also catalyse transfer of a GlcNAc group
from UDP-GlcNAc to an acceptor, followed by deacetylation of
the product.27
Neo15 appears to have a narrow specificity for its ribostamycin
substrate. Incubation of Neo15 with 2-DOS under the same
conditions did not produce any N-acetylparomamine, confirming
that Neo15 and Neo8 act independently. Replacing UDP-GlcNAc
with UDP-D-glucosamine (UDP-GlcN) or TDP-glucosamine
(TDP-GlcN) as alternative sugar donors did not yield any
product. Neo15 also failed to catalyse glycosylation of butirosin,
which differs from ribostamycin by having an (S)-4-amino-2-
hydroxybutyrate substituent at the C1 amine of 2-DOS (data not
shown).
Structural comparisons between glycosyltransferases Neo8 and
Neo15
Neo8 and Neo15 share only modest amino acid sequence identity
(23.5%) but their three-dimensional structures are likely to be
more similar than this might suggest. Glycosyltransferases are
currently classified into 91 distinct sequence families22 (CAZy;
carbohydrate-active enzymes database at http://www.cazy.org/)
but almost all display only two different three-dimensional folds,
referred to as GT-A and GT-B.23 Both Neo8 and Neo15 are
classified by CAZy in Family 4, the second largest described
This journal is
The Royal Society of Chemistry 2008
Org. Biomol. Chem., 2008, 6, 3306–3314 | 3309
©

View Article Online
Fig. 3 Conserved active site motif in Neo8, Neo15 and closely-related glycosyltransferases of known structure. Fully conserved residues are in white text
with a red background. Similar (at least 70%) residues are in red text and framed in blue. The conserved glutamate residues are marked with wedges (ꢀ).
Neo16 as a dual-function N-deacetylase in neomycin biosynthesis
at a unique BsaXI site internal to the gene. An in-frame deletion
was not required because the location of neo16 means that there
is no possibility of an unwanted polar effect on a downstream
gene. The mutant was verified by PCR screening and Southern
blot (Fig. S3-a & c). Neomycin production was abolished in
the Dneo16 mutant and no neomycin was detected when 2-DOS
was fed to the mutant. However when neamine (0.2 mg mL−1)
was added to the medium, neomycin production was restored
(Fig. S3-d and Table S1). Paromomycin was again detected
in the culture supplemented with paromamine (0.2 mg mL−1)
(Fig. S3-e). Ribostamycin feeding (0.2 mg mL−1) led to only
small amounts of neomycin (Fig. S3-f), which is surprising when
contrasted with the ability of ribostamycin to complement the
Dneo8 mutant with high efficiency. However, if Neo16 were the
only enzyme competent to carry out both of the deacetylation steps
in neomycin biosynthesis then none of these feeding experiments
should have given rise to neomycin or paromomycin. These
data taken together strongly suggest the presence of a second
deacetylase in S. fradiae capable of the deacetylation of N-
acetylglucosaminylribostamycin. Therefore, even though Neo16
shows the requisite broad specificity in vitro, and therefore is likely
to have a dual specificity in vivo too, another deacetylase enzyme,
perhaps located outside the boundaries of the currently-recognised
neo gene cluster, cannot be excluded from a role in neomycin
biosynthesis.
Since both Neo8 and Neo15 use the same sugar donor, UDP-
GlcNAc, deacetylation of the glucosamine moiety must take
place twice during neomycin biosynthesis. We have previously
characterised BtrD as a deacetylase responsible for converting N-
acetylparomamine to paromamine inbutirosin biosynthesis, where
only one deacetylation reaction is required.16 Therefore Neo16
(NeoL), a potential homologue of BtrD (19% identity) in the neo
cluster, was a candidate to fulfil at least one, and possibly both, of
these roles.
The Neo16 protein was expressed with an N-terminal His6-
tag in E. coli BL21 (DE3) and purified by nickel-chelate affinity
chromatography in a yield of 0.7 mg L−1 of culture. The mass of
the purified protein was determined to be 32 138 Da (Fig. S3-b)
(including the His6-tag, and with loss of N-terminal methionine)
by LC–ESI-MS (calculated, 32 141 Da). A peak (32 313 Da)
corresponding to the a-N-gluconoylated21 Neo16 (calculated,
32 319 Da) was also observed. Purified Neo16 was found to lose
its activity within a few days, so fresh samples of Neo16 were
used for all in vitro assays. Neo16 was found to function as a
deacetylase in vitro with broad substrate specificity. Incubation
of Neo16 with N-acetylparomamine overnight at 30 ◦C led to
the formation of paromamine ([M + H]+ = 324) (Fig. 4b) as
observed by LC–ESI-MS. Under the same conditions, incubation
of Neo16 together with Neo15, ribostamycin and UDP-GlcNAc
led to the formation of glucosaminylribostamycin ([M + H]+
=
616) (Fig. 4c). The outcome was the same if Neo16 was added after
the completion of the reaction of Neo15 and removal of Neo15
protein by chloroform precipitation. A competitive assay between
N-acetylparomamine and N-acetylglucosaminylribostamycin re-
vealed that Neo16 does not have a clear preference for either
substrate (data not shown). These results demonstrate that Neo16
is a deacetylase fully capable in vitro of accepting either N-
acetylparomamine or N-acetylglucosaminylribostamycin, as also
shown independently by Eguchi and colleagues.17
In view of these results, we tested the ability of purified
BtrD to deacetylate N-acetylglucosaminylribostamycin. Given
its normal activity in the formation of butirosin,16 BtrD was
expected to show selectivity for N-acetylparomamine as a sub-
strate. Surprisingly, it acted on N-acetylglucosaminylribostamycin
as effectively as did Neo16 to form glucosaminylribostamycin
(Fig. S4). Moreover, like Neo16 it showed no substrate preference
in a competitive assay between N-acetylparomamine and N-
acetylglucosaminylribostamycin (data not shown). This suggests
that BtrD and Neo16 have probably evolved to accommodate
multiple substrates and their actions in vivo are controlled by other
factors such as substrate availability or upstream regulators.
To complement the in vitro experiments, the neo16 gene was
disrupted by the insertion of a thiostrepton resistance gene (tsr)
Conclusion
Evidence has been obtained for the respective roles of two
specific glycosyltransferases and a deacetylase in the biosynthetic
pathway to neomycin from S. fradiae. A convenient system for
genetic manipulation of the neomycin producer was developed
and this allowed the role of Neo8 to be studied in vivo. A
specific Dneo8 mutant was created and found to incorporate
paromamine and later intermediates but not 2-DOS. These results
complement the in vitro data17 that suggest Neo8 specifically
adds N-acetylglucosamine to the aglycone 2-DOS. Meanwhile,
a Dneo15 mutant accumulated ribostamycin in its supernatant,
and recombinant Neo15 was purified and shown specifically to
transfer GlcNAc from UDP-GlcNAc to ribostamycin to form N-
acetylglucosaminylribostamycin. Both Neo16, the sole deacety-
lase to be recognised in the gene cluster, and BtrD, its counterpart
in butirosin biosynthesis, show broad deacetylase activity for
neomycin biosynthetic intermediates in vitro, consistent with a
dual specificity for Neo16. Nevertheless, such dual specificity is
not essential, for a second enzyme with the ability to substitute
for Neo16 in deacetylation of N-acetylglucosaminylribostamycin
appears to be present in S. fradiae. There is a precedent for this
finding in mycothiol biosynthesis, where the deacetylase mshB was
3310 | Org. Biomol. Chem., 2008, 6, 3306–3314
This journal is
The Royal Society of Chemistry 2008
©

View Article Online
Fig. 4 LC–MS analysis of Neo16 deacetylase activity in vitro. (a) Control assay with the substrate N-acetylparomamine but without Neo16.
(b) Conversion of N-acetylparomamine to paromamine by Neo16. (c) Conversion of N-acetylglucosaminylribostamycin to glucosaminylribostamycin by
Neo16.
found not to be essential, because the mshB mutant still produced
in a flask (250 mL) containing a spring (to aid dispersal of mycelial
growth) at 30 ◦C in a rotary shaker (200 rpm) for seven days. For
feeding studies, filter-sterilised compounds were added to TSB
just before inoculation. Escherichia coli DH10B strain was used
for general cloning; ET12567/pUZ8002 strain,30 as a donor strain
for conjugation with S fradiae; BL21 (DE3) strain, as a host for
protein expression.
approximately 20% of the wild-type level of mycothiol.28
Experimental
Bacterial strains and culture conditions
S. fradiae NCIMB 8233 was maintained on SFM solid medium29
or in tryptic soy broth (TSB) (Difco). The Dneo16 mutant was
maintained on SFM plates containing thiostrepton (25 lg mL−1),
and the complemented Dneo8 and Dneo15 mutants on plates
containing apramycin (25 lg mL−1). For fermentation, S. fradiae
and its mutants were grown in TSB (50 mL) without any antibiotics
Construction and verification of mutants
The neo8 deletion plasmid, pQZ27, was constructed based on an
E. coli–Streptomyces shuttle vector, pQZ21, derived from pYH731
by HindIII and MscI digestion followed by blunting the ends and
This journal is
The Royal Society of Chemistry 2008
Org. Biomol. Chem., 2008, 6, 3306–3314 | 3311
©

View Article Online
self-ligation of an 8615 bp fragment. A SapI (3613 bp) fragment
from the cosmid FR3F712 of the neomycin gene cluster, containing
neo8, was first cloned into the SmaI site of pUC18 vector. The
resulting plasmid, pQZ25, was digested with EcoRI and BamHI to
recover a 3637 bp fragment, which was further digested by BspHI
and NcoI. Two fragments, EcoRI-BspHI (1243 bp) and NcoI-
BamHI (1554 bp), were ligated into EcoRI and BamHI digested
pUC18 to form pQZ26. Finally, a blunt-ended EcoRI-BamHI
fragment (2797 bp) from pQZ26 was cloned into the blunt-ended
BamHI site of pQZ21 to form pQZ27.
The neo15 deletion plasmid, pQZ37, was also constructed based
on pQZ21. An FspI fragment (4592 bp) from the cosmid FR3F7,
containing neo15, was first cloned into the SmaI site of pUC18.
The resulting plasmid, pQZ35, was linearised by AfeI digestion be-
fore it was used as a PCR template for the construction of pQZ36.
The following primers were used: 5ꢀ-GACGCCCTGGTGGACC-
3ꢀ and 5ꢀ-GGACCGGTAGAACTCCG-3ꢀ. The PCR reaction was
performed using Phusion High-Fidelity Master Mix with GC
buffer (New England Biolabs) under the following conditions:
98 ◦C, 2 min; 35 cycles of 10 s at 98 ◦C, 20 s at 60 ◦C and
3 min at 72 ◦C, and a final extension at 72 ◦C for 5 min. A
typical reaction mixture contained AfeI-digested pQZ35 (1 lL),
primers (1 lL each, stock concentration at 50 lM), DMSO (3 lL),
MgCl2 (1.5 lL, 50 mM), the master mix (25 lL) in a final volume
of 50 lL. The product (6519 bp) was phosphorylated by T4
polynucleotide kinase (New England Biolabs) before self-ligation
to form pQZ36. Finally, a blunt-ended KpnI–PvuII fragment (3465
bp) from pQZ36 was cloned into the blunt-ended BamHI site of
pQZ21 to form pQZ37. During this work, a mistake was noted
in the publicly deposited neo15 sequence (AJ843080) at nucleotide
520, which should be G not C. This correction changes an aspartic
acid into a histidine residue.
The neo16 deletion plasmid, pQZ22, was also based on pQZ21.
A SacI fragment (4242 bp) from the cosmid FR3F7, containing
neo16, was cloned into the SacI site of pUC18 to form pQZ15. This
plasmid was modified to remove the BsaXI site by TfiI digestion,
followed by recovery of a 6788 bp fragment which then self-ligated
to form pQZ15ꢀ. Digestion of pQZ15ꢀ with BsaXI and blunting of
the ends allowed the insertion of a thiostrepton resistance marker
(tsr), an SfoI fragment (1461 bp) from pJTU41232 to form pQZ16.
A blunt-ended SacI fragment (5670 bp) was then ligated into the
blunt-ended BamHI site of pQZ21 to form pQZ22.
The plasmids pQZ27 (neo8), pQZ37 (neo15) and pQZ22 (neo16)
were introduced into S. fradiae through conjugation from the
E. coli ET12567/pUZ8002 strain transformed with the relevant
plasmids. Conjugation SFM plates (20 mL) were incubated at
30 ◦C for 16 hours before being overlaid with water (1 mL each)
containing nalidixic acid (500 lg), and additionally apramycin
(500 lg) for all except for the pQZ22 plates for which thiostrepton
(500 lg) was added. Exconjugants (usually >102 per plate) were
picked after four days and were verified on corresponding SFM
plates with nalidixic acid (25 lg mL−1), and additionally apramycin
(25 lg mL−1) or thiostrepton (25 lg mL−1). Confirmed colonies
were then propagated on SFM plates without antibiotics. Single
colonies of the putative Dneo8 and Dneo15 mutants from the
non-selective plates were patched onto paired SFM plates, one
of them containing apramycin (25 lg mL−1). Patches showing
sensitivity to apramycin were screened by PCR. Single colonies
of the putative Dneo16 mutant from the non-selective plates were
patched onto paired SFM plates, one with apramycin (25 lg mL−1)
and the other with thiostrepton (25 lg mL−1). The patches
with resistance to thiostrepton and sensitivity to apramycin were
screened by PCR. The following primers were used for PCR
mutant screening: (Dneo8) 5ꢀ-TCTCCCCGTGGAGTCCCC-3ꢀ
and 5ꢀ-CCACGTTCACCACCAGATAGATG-3ꢀ; (Dneo15) 5ꢀ-
TCGTCAACATCCTCAACCG-3ꢀ and 5ꢀ-TCACCCACCGTG-
CTCCT-3ꢀ; (Dneo16) 5ꢀ-TCCTCGTAGAACAGGGTCAGC-3ꢀ
and 5ꢀ-GGACGGGAGGAGGAGCAC-3ꢀ. PCR reactions were
performed using BioMixTM Red (Bioline) under the following
conditions: 94 ◦C, 5 min; 35 cycles of 30 s at 94 ◦C, 30 s at
55 ◦C and 1 min at 72 ◦C, and a final extension at 72 ◦C for
7 min. A typical reaction mixture contained genomic or plasmid
DNA template (1 lL), primers (1 lL each, stock concentration at
50 lM), DMSO (1.5 lL), BioMixTM Red (15 lL) in a final volume
of 30 lL. An 865 bp PCR product (Dneo8) (Fig. S1-b) and a
486 bp PCR product (Dneo15) (Fig. S2-c) were recovered from
0.7% agarose gel for sequencing to confirm the in-frame deletion.
The Dneo16 mutant was also checked by Southern blot using AgeI-
digested pQZ16 (2606 bp and 2040 bp fragments) as a probe (Fig.
S3-c). The genomic DNA of the mutants and the wild-type, and the
plasmid controls were all digested with AgeI and SacI separately.
Metabolite purification and identification
After seven days of fermentation in TSB, the culture was harvested
by centrifugation. The supernatant was collected and its pH was
adjusted to 6 with HCl (2 M). The supernatant was then filtered
through a GF/A glass fibre filter before being loaded onto a
R
column (2 g) of DOWEXꢀ 50 WX8–200 ion-exchange resins, that
had been pre-washed with acetonitrile (50 mL) followed by water
(50 mL), and equilibrated with sodium phosphate buffer (50 mM,
pH 5.4). The column was washed with sodium phosphate buffer
(60 mL, 50 mM) and then with ammonium hydroxide solution
(1 mL, 1 N). Target compounds were eluted with ammonium
hydroxide solution (10 mL, 1 N). The eluate was freeze-dried
before being redissolved in water (1 mL). It was washed twice with
an equal volume of chloroform. The aqueous layer was analysed
by LC–ESI-MS on a Hewlett-Packard HPLC 1100 series system
(Agilent), coupled to a Finnigan LCQ (Thermo Finnigan) mass
spectrometer. The sample was eluted from a 2.0 × 250 mm Gemini
5 l C18 column (Phenomenex) using a linear gradient of 15% to
100% acetonitrile (+ 0.1% pentafluoropropionic acid) in water (+
0.1% pentafluoropropionic acid) over 30 minutes, at a flow rate of
0.15 mL min−1. MS/MS was carried out with 20% relative collision
energy (helium as collision gas). Fragmentation patterns of the
products were compared to those of commercial or synthesised
standards when available.
Complementation of the Dneo8 and Dneo15 mutants
a) Plasmid pIB139-based complementation. The neo8 gene
was amplified using the following primers: 5ꢀ-CGTCGGGCA-
¯ ¯
TATGCAGGTGCAGATCCT-3ꢀ (NdeI) and 5ꢀ-ACAGCGCC-
¯ ¯ ¯ ¯
GCAGCGAATTCACGG-3ꢀ (EcoRI). The neo15 gene was
¯ ¯ ¯ ¯ ¯ ¯
amplified using the primers for its protein expression. The products
were digested with NdeI and EcoRI, before being ligated into NdeI-
and EcoRI-digested integrative vector pIB13933 to give pQZ53
(neo8) and pQZ54 (neo15). These complementation plasmids were
3312 | Org. Biomol. Chem., 2008, 6, 3306–3314
This journal is
The Royal Society of Chemistry 2008
©

View Article Online
introduced into the Dneo8 and Dneo15 mutants respectively by
conjugation from E. coli ET12567/pUZ8002. As a control, the
empty vector pIB139 was used for conjugation with the mutants.
Exconjugants were verified based on their apramycin resistance.
Complemented mutants were confirmed by PCR from their
genomic DNA using the original mutants as a negative control,
and the wild-type DNA and the plasmids as positive controls.
into a storage buffer (100 mM NaCl, 20 mM Tris-HCl, pH 7.5)
by ultrafiltration. The sample was applied on a 2.0 × 250 mm
Jupiter 5 l C4 column (Phenomenex). The method used a linear
gradient of 35% to 75% acetonitrile (+ 0.1% trifluoroacetic acid)
in water (+ 0.1% trifluoroacetic acid) over 25 min with a flow rate
of 0.3 mL min−1.
Protein expression of BtrD and Providencia stuartii 2ꢀ-N-
acetyltransferase (AAC(2ꢀ)-Ia)35 was carried out as described
previously.16
b) Plasmid pSET152-based complementation. The neo8 gene
was excised from pQZ25 with FspI. The 2890 bp fragment
was inserted into PvuII-digested pSET15234 to form pQZ60.
A 4914 bp fragment containing the genes neo12 to neo16 was
cut out from pQZ35 with PvuII, and then ligated into PvuII-
digested pSET152 to form pQZ61. This plasmid was digested
with AgeI and a 9839 bp fragment was recovered to be used
as a template for a PCR reaction, which amplified the 6537 bp
vector flanking the genes neo12 to neo16. The PCR reaction was
carried out with 3 min extension times using the following primers:
5ꢀ-CCCGGCCTCGTGGAGGCACGGC-3ꢀ and 5ꢀ-GGCGTGT-
CTCCTCGTTCGGGACGCTG-3ꢀ. Meanwhile, the neo15 gene
was amplified using the following primers: 5ꢀ-GTGGCTGAGG-
CGCCTGCCGGG-3ꢀ and 5ꢀ-TCACCCACCGTGCTCCTCC-
TCCCGT-3ꢀ. The neo15 product was phosphorylated before being
ligated into the 6537 bp vector to form pQZ62 (forward insertion
selected only). The sequence of the plasmid was checked by DNA
sequencing. The complementation plasmids were again introduced
into the Dneo8 (pQZ60) and Dneo15 (pQZ61 and pQZ62) mutants
by conjugation from E. coli ET12567/pUZ8002. As a control, the
empty vector pSET152 was used. The complemented mutants were
again checked by PCR from their genomic DNA.
Chemicals
Paromamine and neamine were synthesised by acid methanolysis
of paromomycin and neomycin (Sigma) using the method of
Dutcher and Donin.36 N-Acetylparomamine was synthesised
with AAC(2ꢀ)-Ia.16 UDP-GlcN and TDP-GlcN were made as
previously described.37 Ribostamycin and UDP-GlcNAc were
purchased from Sigma.
Enzymatic assays
A typical reaction mixture for assay of glycosyltransferase activity
(50 lL) contained Tris-HCl (25 mM, pH 8.4), ribostamycin
(1 mM) or butirosin (1 mM) or 2-DOS (1 mM), UDP-GlcNAc
◦
(2 mM) and Neo15 (0.15 mg mL−1) and was incubated at 30 C
overnight. For assay of deacetylase activity, Neo16 (1 mg mL−1)
or BtrD (1 mg mL−1) was included in the reaction mixture.
The assay for the first deacetylation reaction contained Tris-
HCl (25 mM, pH 8.4) and N-acetylparomamine (0.15 mM).
The competitive assays included N-acetylparomamine (0.5 mM)
and N-acetylglucosaminylribostamycin (0.5 mM) as substrates,
which were obtained from quenched completed reactions of
Neo15 and AAC(2ꢀ)-Ia. The reaction goes to completion overnight
with ribostamycin concentration up to 10 mM for Neo15 and
paromamine concentration up to 5 mM for AAC(2ꢀ)-Ia. All
control reactions omitted either the substrate(s) or the protein(s).
All reactions were quenched with chloroform (50 lL), followed
by vortexing and centrifugation. The aqueous layer was analysed
by LC–ESI-MS using the same method as for the knockout
experiments.
Cloning, overexpression and purification of proteins
The neo15 gene was amplified by PCR from the cosmid FR3F7
using the following primers: 5ꢀ-CCGCCGGAGGTGCCATAT-
¯ ¯ ¯ ¯ ¯
GGCTGAG-3ꢀ (NdeI) and 5ꢀ-TCCGCCGCCTCCGAATTC-
¯
¯ ¯ ¯ ¯ ¯ ¯
GGCTC-3ꢀ (EcoRI). The neo16 gene was amplified with
the primers: 5ꢀ-CGGGAGGAGGAGCCATATGGGTGAGCC-
3ꢀ (NdeI) and 5ꢀ-ACGAGGCGAATTCACCGGGCACCC-3ꢀ
(EcoRI). The PCR reactions were performed under the same
¯ ¯ ¯ ¯ ¯ ¯
¯ ¯ ¯ ¯ ¯ ¯
conditions as for the construction of pQZ36 except that 1 min
extension time at 72 ◦C was used instead of 3 min. The prod-
ucts were separately digested and ligated into the appropriate
restriction sites of pET28a(+) (Novagen) expression vector to form
plasmids pQZ39 (Neo15) and pQZ30 (Neo16), which were used
to transform E. coli BL21 (DE3) competent cells. The sequences
of neo15 in pQZ39 and neo16 in pQZ30 were verified by DNA
sequencing.
The E. coli cells harbouring the recombinant plasmids were
grown in LB medium (1 L) containing kanamycin (50 lg mL−1)
with shaking (250 rpm) at 30 ◦C until the A600 reached 0.6 to 0.9.
Protein overexpression was induced by the addition of isopropyl b-
D-thiogalactopyranoside (0.2 mM), and the culture was incubated
further with shaking (200 rpm) at 18 ◦C overnight. The cells were
harvested by centrifugation and the pellet was resuspended in
binding buffer (30 mL, containing 5 mM imidazole, 0.5 M NaCl,
20 mM Tris-HCl, pH 7.9). The cells were lysed by sonication,
followed by centrifugation. His6-Neo15 and His6-Neo◦16 were
purified on a Ni2+-NTA affinity column (Novagen) at 4 C (Fig.
S2-d and S3-b). After elution, the protein solution was exchanged
Acknowledgements
We are grateful to Andrew W. Truman for the synthesis of UDP-
GlcN and TDP-GlcN, and Dr Nicholas M. Llewellyn for the
synthesis of paromamine, as well as for their helpful advice on
preparation of this manuscript. This work was supported by a
project grant from the BBSRC to J.B.S. and F.H. Q.F. thanks
Trinity College, the Cambridge Overseas Trust and the ORS
scheme for studentship support.
References
1 C. J. R. Pearce and K. L. Rinehart, Jr., in Antibiotics, vol. IV, ed
J. W. Corcoran, Springer-Verlag, Berlin/Heidelberg/New York, 1981,
pp. 74–100.
2 N. M. Llewellyn and J. B. Spencer, Nat. Prod. Rep., 2006, 23, 864–874.
3 S. A. Waksman and H. A. Lechevalier, Science, 1949, 109, 305–307.
4 P. Edelmann and J. Gallant, Cell (Cambridge, Mass.), 1977, 10, 131–
137.
5 A. P. Carter, W. M. Clemons, D. E. Brodersen, R. J. Morgan-Warren,
B. T. Wimberly and V. Ramakrishnan, Nature, 2000, 407, 340–348.
This journal is
The Royal Society of Chemistry 2008
Org. Biomol. Chem., 2008, 6, 3306–3314 | 3313
©

View Article Online
6 J. B. H. Tok, L. J. Dunn and R. C. Des Jean, Bioorg. Med. Chem. Lett.,
2001, 11, 1127–1132.
7 A. Litovchick, A. Lapidot, M. Eisenstein, A. Kalinkovich and G.
Borkow, Biochemistry, 2001, 40, 15612–15623.
22 P. M. Coutinho, E. DeLeury, G. J. Davies and B. Henrissat, J. Mol.
Biol., 2003, 328, 307–317.
23 Y. Bourne and B. Henrissat, Curr. Opin. Struct. Biol., 2001, 11, 593–
600.
8 S. Hirukawa, K. A. Olson, T. Tsuji and G. F. Hu, Clin. Cancer Res.,
2005, 11, 8745–8752.
24 M. Guerin, J. Kordulakova, F. Schaeffer, A. Svetlikova, A. Buschiazzo,
D. Giganti, B. Gicquel, K. Mikusova, M. Jackson and P. M. Alzari,
J. Biol. Chem., 2007, 282, 20705–20714.
25 M. W. Vetting, P. A. Frantom and J. S. Blanchard, J. Biol. Chem., 2008,
283, 15834–15844.
26 R. A. Geremia, E. A. Petroni, L. Ielpi and B. Henrissat, Biochem. J.,
1996, 318, 133–138.
9 M. J. Cron, J. G. Keil, M. V. Ruggeri and D. Walker, Chem. Commun.,
1979, 266–267.
10 G. L. Coffey, L. E. Anderson, M. W. Fisher, M. M. Galbraith, A. B.
Hillegas, D. L. Kohberger, P. E. Thompson, J. Weston and K. Ehrlich,
Antibiot. Chemother., 1959, 9, 730–738.
11 E. J. Hessler, H. K. Jahnke, J. H. Robertson, K. Tsuji, K. J. Rinehart
Jr. and W. T. Shier, J. Antibiot. (Tokyo), 1970, 23, 464–466.
12 F. Huang, S. F. Haydock, T. Mironenko, D. Spiteller, Y. Li and J. B.
Spencer, Org. Biomol. Chem., 2005, 3, 1410–1418.
13 F. Kudo, Y. Yamamoto, K. Yokoyama, T. Eguchi and K. Kakinuma,
J. Antibiot., 2005, 58, 766–774.
27 G. L. Newton, P. Ta, K. P. Bzymek and R. C. Fahey, J. Biol. Chem.,
2006, 281, 33910–33920.
28 N. A. Buchmeier, G. L. Newton, T. Koledin and R. C. Fahey, Mol.
Microbiol., 2003, 47, 1723–1732.
29 T. Kieser, M. J. Bibb, M. J. Buttner, K. F. Chater and D. A. Hopwood,
Practical Streptomyces Genetics, John Innes Foundation, Norwich,
2000.
30 D. J. MacNeil, J. L. Occi, K. M. Gewain, T. MacNeil, P. H. Gibbons,
C. L. Ruby and S. J. Danis, Gene, 1992, 115, 119–125.
31 Y. Sun, M. Samborskyy, T. Mironenko, P. F. Leadlay and S. F. Haydock,
Microbiology, 2006, 152, 3507–3515.
32 Y. Sun, X. Zhou, J. Liu, K. Bao, G. Zhang, G. Tu, T. Kieser and Z.
Deng, Microbiology, 2002, 148, 361–371.
33 C. J. Rowe, I. U. Bo¨hm, B. Wilkinson, B. A. M. Rudd, G. Foster, A. P.
Blackaby, P. J. Sidebottom, Y. Roddis, A. D. Buss, J. Staunton and P. F.
Leadlay, Chem. Biol., 2001, 8, 475–485.
14 F. Huang, Y. Li, J. Yu and J. B. Spencer, Chem. Commun., 2002, 2860–
2861.
15 F. Kudo, K. Kawabe, H. Kuriki, T. Eguchi and K. Kakinuma, J. Am.
Chem. Soc., 2005, 127, 1711–1718.
16 A. W. Truman, F. Huang, N. M. Llewellyn and J. B. Spencer, Angew.
Chem., Int. Ed. Engl., 2007, 46, 1462–1464.
17 K. Yokoyama, Y. Yamamoto, F. Kudo and T. Eguchi, ChemBioChem,
2008, 9, 865–869.
18 H. Baud, A. Betencourt, M. Peyre and L. Penasse, J. Antibiot., 1977,
30, 720–723.
19 F. Kudo, T. Fujii, S. Kinoshita and T. Eguchi, Bioorg. Med. Chem.,
2007, 15, 4360–4368.
20 C. J. Rowe, J. Corte´s, S. Gaisser, J. Staunton and P. F. Leadlay, Gene,
1998, 216, 215–223.
21 K. F. Geoghegan, H. B. F. Dixon, P. J. Rosner, L. R. Hoth, A. J.
Lanzetti, K. A. Borzilleri, E. S. Marr, L. H. Pezzullo, L. B. Martin,
P. K. LeMotte, A. S. McColl, A. V. Karnath and J. G. Stroh, Anal.
Biochem., 1999, 267, 169–184.
34 M. Bierman, R. Logan, K. O’Brien, E. T. Seno, R. N. Rao and B. E.
Schoner, Gene, 1992, 116, 43–49.
35 K. Franklin and A. J. Clarke, Antimicrob. Agents Chemother., 2001, 45,
2238–2244.
36 J. D. Dutcher and M. N. Donin, J. Am. Chem. Soc., 1952, 74, 3420–
3422.
37 A. W. Truman, L. Robinson and J. B. Spencer, ChemBioChem, 2006, 7,
1670–1675.
3314 | Org. Biomol. Chem., 2008, 6, 3306–3314
This journal is
The Royal Society of Chemistry 2008
©