Arabinose 5-phosphate isomerase as a target for antibacterial design: Studies with substrate analogues and inhibitors

Article abstract of DOI:10.1016/j.bmc.2013.08.012

Structural requirements of d-arabinose 5-phosphate isomerase (KdsD, E.C. 5.3.1.13) from Pseudomonas aeruginosa were analysed in detail using advanced NMR techniques. We performed epitope mapping studies of the binding between the enzyme and the most poten

Full text of DOI:10.1016/j.bmc.2013.08.012

Bioorganic & Medicinal Chemistry 22 (2014) 2576–2583
Contents lists available at ScienceDirect
Bioorganic & Medicinal Chemistry
journal homepage: www.elsevier.com/locate/bmc
Arabinose 5-phosphate isomerase as a target for antibacterial design:
Studies with substrate analogues and inhibitors
Luca Gabrielli a, , Silvia Merlo a, , Cristina Airoldi a, Paola Sperandeo a, Serena Gianera a,
Alessandra Polissi a, Francesco Nicotra a, Tod P. Holler b, Ronald W. Woodard b, , Laura Cipolla a,
⇑
⇑
a Department of Biotechnology and Biosciences, University of Milano-Bicocca, Piazza della Scienza 2, 20126 Milano, Italy
b Department of Medicinal Chemistry, University of Michigan, 428 Church Street, Ann Arbor, MI 48109-1065, USA
a r t i c l e i n f o
a b s t r a c t
Article history:
Structural requirements of
D-arabinose 5-phosphate isomerase (KdsD, E.C. 5.3.1.13) from Pseudomonas
Received 30 May 2013
Revised 18 July 2013
Accepted 5 August 2013
Available online 11 August 2013
aeruginosa were analysed in detail using advanced NMR techniques. We performed epitope mapping
studies of the binding between the enzyme and the most potent KdsD inhibitors found to date, together
with studies of a set of newly synthesised arabinose 5-phosphate (A5P) mimetics. We report here the first
experimental evidence that KdsD may bind the furanose form of A5P, suggesting that catalysis of ring
opening may be an important part of KdsD catalysis.
Keywords:
Ó 2014 Published by Elsevier Ltd.
D-Arabinose 5-phosphate isomerase
NMR binding studies
Enzyme inhibitors
Molecular recognition
Pseudomonas aeruginosa
Escherichia coli
1. Introduction
Two paralogous genes in Escherichia coli, kdsD and gutQ, possess
API activity.9 Indeed, API is an essential function, as inactivation of
An effective response to the emergence of multidrug resistance
among an increasing number of pathogenic bacteria will require
the rapid discovery of new antimicrobial drug scaffolds.1,2 The dis-
covery of these new scaffolds may result from targeting novel or
underexploited essential microbial biosynthetic pathways. Lipo-
polysaccharide is an essential glycolipid and a major virulence fac-
tor located in the outer membrane of Gram-negative bacteria3 that
has no human counterpart. Therefore, its biosynthesis represents
an excellent target for novel antimicrobial discovery.4,5 LPS is com-
posed of three distinct regions: lipid A, core oligosaccharide and
both kdsD and gutQ leads to non-viable cells.10 The KdsD homo-
logue in the opportunistic pathogen Pseudomonas aeruginosa
shares significant sequence similarity (57% identity and 75% simi-
larity) to its counterpart in E. coli and, because P. aeruginosa lacks
gutQ, it is essential for P. aeruginosa cell survival.11
Despite the relevance of API activity as an antibacterial target
and the availability of structural information on E. coli KdsD,12,13
few potent inhibitors have been found.14 In fact, the fundamental
requirements for potent inhibitory activity are still poorly under-
stood.11,15 To better characterise KdsD as a molecular target and
gain further insight into the enzyme–inhibitor interaction required
for the design of potent inhibitors, we undertook a molecular rec-
ognition study, based on advanced NMR techniques,16,17 using two
of the most potent inhibitors found to date, compounds 3 and 4
(Fig. 1). Compounds 3 and 4 have been shown to inhibit Francisella
O-antigen.6 The carbohydrate 3-deoxy-
D-manno-oct-2-ulosonic
acid (Kdo) is a conserved essential residue found in the core oligo-
saccharide region of all bacterial LPS analysed so far.7 The biosyn-
thesis and activation of Kdo involves four sequentially acting
enzymes. The first committed step in Kdo biosynthesis is catalysed
by a specific
D
-arabinose 5-phosphate isomerase (API) activity,
tularensis KdsD with IC50s of 7 and 10 l
M, respectively.14 These
which converts
D
-ribulose 5-phosphate (Ru5P) into
D
-arabinose
compounds are both characterised by a nitrogen-containing func-
tionality in place of carbon 1 of the enediol intermediate (Fig. 1),
while a carbonyl group is coincident with C-2 and identical to
the carbonyl group of the starting material, Ru5P. The resonance
structures in which the nitrogen lone pair is delocalized to form
a C–N double bond mimic the C–C double bond found in the 1,2-
enediol intermediate 5 (Fig. 1). The remaining inhibitor positions
5-phosphate (A5P, Scheme 1).8
⇑
Corresponding authors. Tel.: +1 7347647366; fax: +1 7346478430 (R.W.W.);
tel.: +39 0264483460; fax: +39 0264483565 (L.C.).
E-mail addresses: rww@umich.edu (R.W. Woodard), laura.cipolla@unimib.itUSA
(L. Cipolla).
These authors contributed equally to this work.
http://dx.doi.org/10.1016/j.bmc.2013.08.012
0968-0896/Ó 2014 Published by Elsevier Ltd.

L. Gabrielli et al. / Bioorg. Med. Chem. 22 (2014) 2576–2583
NHOH
2577
CH2OH
O
NH2
OPO3=
O
O
OAc
OAc
O
OH
OH
API
OH
H
H
H
H
OH
H
H
OH
HO
OH
CH2OPO3=
CH2OPO3=
Ru5P (1)
CH2OPO3=
A5P (2)
3
4
Scheme 1. Isomerization of ribulose 5-phosphate (Ru5P, 1) to arabinose 5-
phosphate (A5P, 2) catalyzed by -arabinose 5-phosphate isomerase (API).
OPO3=
OPO3=
O
O
D
OCH3
OH
OCH3
OH
HO
O-
HO
O-
(C2–C4) are identical, in terms both of stereochemistry and substit-
uents, to positions 3, 4 and 5 of the substrates and the enediol
intermediate, except that inhibitor 3 has the hydroxyl groups
esterified as acetates. Despite the presence of the acetate groups,
compound 4 showed the greatest inhibition of KdsD.14 This level
of activity is far from the potency likely to be required for clinical
utility, so significant structural changes must be made to this
scaffold.
6
7
O
O
OCH3
OCH3
OH
O
O
HO
OH
HO
9
8
The absence of typical optimization tools, like structural data
from enzyme–inhibitor co-crystals, and the limited structural data
available for KdsD suggest that NMR studies of the interaction of
these compounds with the isolated enzyme are the best optimisa-
tion tool available. Given the limited size and complexity of com-
pounds 3 and 4, we decided to synthesize different arabinose
derivatives to gain additional information that would be useful
for the design of more potent inhibitors.
CHO
H
OH
OH
CH2OPO3=
CH2OH
O
OH
OH
CH2OPO3=
CHOH
OH
OH
OH
CH2OPO3=
HO
H
H
H
H
H
H
2
1,2-enediol intermediate (5) A5P ( )
1
Ru 5P ( )
API catalyses the reversible interconversion of Ru5P, an open-
chain ketose, to A5P, an aldose that can assume a furanose form
(Fig. 2, inset). It is not yet clear if the enzyme binds the furanose
form, which must be converted to the open-chain aldose to under-
go enzymatic isomerisation. In order to shed light on the possibil-
ity that the enzyme might accommodate the substrate A5P in
either, or both of its cyclic hemiacetal forms, we synthesised
methyl arabinoside 5-phosphates 6 and 7, which are fixed in their
furanose forms by the presence of the methyl group. The methyl
group was chosen for its relatively low steric hindrance. Previous
studies have shown that the acidic phosphate group is necessary
for recognition and binding to the enzyme active site.15 We synthe-
sized compounds 8 and 9 (Fig. 2) in order to determine if a carbox-
ylate is an adequate mimetic of the natural phosphate. In addition
to using compounds 6–9 in NMR studies, we also tested their bio-
logical activity of a panel of Gram-negative bacterial strains.
=O3PO
O
OH
OH
A5P (furanose form)
HO
Figure 2. Structures of the compounds used in the present study.
cally more stable methyl pyranosides. The
a and b anomers (a/b
1:0.4) were separated by flash chromatography and used as com-
mon intermediates for the synthesis of the phosphates 6 and 7,
and carboxylate analogues 8 and 9. In order to introduce the phos-
phate group, both the
a and b methyl D-arabinofuranosides (com-
pounds 11 and 12) were phosphorylated with diphenyl
chlorophosphate in dry pyridine to derivatives 13 and 14 (around
80% yield). The phenyl protective groups were removed by hydrog-
enolysis (PtO2 catalyst), followed by neutralisation with sodium
hydroxide, affording the desired cyclic phosphates 6 and 7 as so-
dium salts in quantitative yields.
The carboxylate analogues 8 and 9 were synthesised by direct
oxidation of methyl D-arabinofuranosides 11 and 12 with sodium
hypochlorite in the presence of catalytic amounts of the oxidant
TEMPO (63% yields).
2. Results and discussion
2.1. Chemical synthesis
The target arabinosides 6–9 were synthesised from D-arabinose
(10) (Scheme 2). As Dangerfield et al. previously observed,18 reac-
tion of 10 with a solution of AcCl in dry MeOH for 18 h at room
temperature led to the formation of the desired methyl furanosides
11 and 12 in 89% yield, instead of the undesired, thermodynami-
1
2
1
2
1
NH2
CHOH
OH
NH2
O
OH
OH
CH2OPO3=
NHOH
O
OAc
OAc
CH2OPO3=
NHOH
2
O
O
OAc
OAc
3
3
4
3
H
H
OH
OH
H
H
OH
H
H
H
H
H
H
4
4
OH
5
5
5
CH2OPO3=
CH2OPO3=
CH2OPO3=
1,2-enediol
3
4
intermediate (5)
Figure 1. Similarities between the enediol intermediate 5 and the most potent KdsD inhibitors 3 and 4 found to date (for sake of clarity IUPAC numbering has been ignored
and the inhibitors have been numbered starting at the nitrogen atom).

2578
L. Gabrielli et al. / Bioorg. Med. Chem. 22 (2014) 2576–2583
+
O-NH4
OH
OH
OH
TEMPO,
NaOCl,
KBr
MeOH,
AcCl
O
O
O
O
O
OH
OMe
OMe
OMe
+
89 %,
α:β 1:0.4
HO
HO
NaHCO3 (aq)
63%
HO
HO
OH
10
OH
OH
OH
12
D-Ara (
)
11
8 α-methyl
β-methyl
PO(OPh)2Cl
9
Py dry
80%
(PhO)2OP
O
(NaO)2OP
O
1. H2, PtO2, EtOH
O
O
OMe
OMe
2. 5 % aq. NaOH
Quant yields
HO
HO
OH
OH
13
α-methyl
14 β-methyl
6
α-methyl
7 β-methyl
Scheme 2. Synthetic scheme for compounds 6–9.
Figure 3. 1H NMR spectra of inhibitor 3 (PBS, pH 7.0, 25 °C) recorded after dissolution (A), after 1 day (B) and after 2 days (C).
1
1
2
1
1
NHOH
NHOH
O
NHOH
NHOH
2
2
2
O
O
O
OH
OH
3
4
3
3
4
3
4
H
H
OAc
OAc
H
H
OAc
H
H
OH
OAc
H
H
4
OH
5
5
5
5
CH2OPO3=
CH2OPO3=
CH2OPO3=
CH2OPO3=
3
3a (MW=272.13)
3b
3c (MW=230.09)
Figure 4. Structures of the de-acetylated compounds (3a–c) derived from inhibitor 3.

L. Gabrielli et al. / Bioorg. Med. Chem. 22 (2014) 2576–2583
2579
2.2. NMR molecular recognition studies
Despite the presence of species with different acetylation pat-
terns (compounds 3 and 3a–c), STD NMR (Saturation-Transfer Dif-
We began by studying compound 3 in the absence of the
enzyme. A 1H NMR analysis performed on a solution of compound
3 in PBS (Fig. 3) revealed that compound 3 slowly hydrolyses to the
corresponding fully de-acetylated product (3c, Fig. 4), through the
partially esterified derivatives 3a and 3b.
ference NMR) experiments,19,20 performed according to
a
previously described procedure,11 allowed the identification the
derivatives best able to bind to P. aeruginosa KdsD (Pa-KdsD). The
STD spectra were acquired on samples of compound 3 dissolved
in PBS, pH 7.0, at 25 °C, and prepared by adding the enzyme to dif-
ferent aliquots of a solution of inhibitor (5 mM) after three incuba-
tion periods (0, 2 and 3 days). The working pH and temperature
values were slightly different from physiological (pH 7.4, 37 °C)
in order to slow down the interconversion process. Since the rela-
tive amounts of the species derived from inhibitor 3 present in the
analyses conducted at these different times (t = 0, 2 days and
3 days), were different, all the compounds present in the mixture
could be identified and screened for their ability to act as KdsD
ligands (Fig. 5).
Careful analysis of the 1H, 1H COSY, 1H, 1H TOCSY and 1H, 13C
HSQC spectra acquired at different times allowed the identifica-
tion of the components, which differed in the number and posi-
tion of acetyl groups. The nature of the different compounds
derived from inhibitor 3 was also supported by mass spectro-
scopic analysis that gave mass peaks at m/e 314.1, 272.0 and
230.0, corresponding to diacetylated (3), monoacetylated (3a
and 3b) and fully deacetylated (3c) derivatives, respectively. Com-
pound 3 thus undergoes spontaneous deacetylation in aqueous
solution, probably through acetate migration from position 4 to
the vicinal phosphate, affording an unstable mixed anhydride that
immediately hydrolyzes to the 4-deacetylated product 3a. The
acetate at C-3 now can migrate to C-4, affording the 3-deacety-
lated product 3b. Finally, the 4-O-acetate in compound 3b
migrates to the phosphate group and is immediately hydrolyzed
to the fully deacetylated product 3c. Migration of acetates is a
common reaction in carbohydrate chemistry, thus this spontane-
ous deacetylation can be expected.
NMR data clearly showed that resonances belonging to the 4-O-
acetylated derivative 3b and the fully deacetylated derivative 3c
appeared in the STD spectra, demonstrating that these derivatives
interact with the isomerase. In contrast, signals from the fully acet-
ylated compound 3 and 3-O-acetylated 3a were not present in the
STD spectrum. Therefore, these derivatives did not bind to KdsD
under the conditions of this experiment. The ability of compound
3 to inhibit F. tularensis KdsD with the relatively potent IC50 of
7 lM might be explained by the hydrolytic lability of compound
Figure 5. (A, C and E) 1H NMR spectra of a mixture containing compound 3 (5 mM) in PBS, pH 7.0, 25 °C, at to (A), after 2 days (C) and after 3 days (E). (B, D and F) STD NMR
spectra of compound 3 (5 mM) in PBS, pH 7.0, 25 °C in the presence of KdsD enzyme (30 lM) at to (B), after 2 days (D) and after 3 days (F). STD spectra were acquired with a
saturation time of 3 s and a saturation frequency of 7.9 ppm.

2580
L. Gabrielli et al. / Bioorg. Med. Chem. 22 (2014) 2576–2583
3, which was certainly a combination of di-, mono-, and non-acet-
ylated compounds during the experiment. This suggests that one of
the monoacetylated compounds (compound 3b) binds to F. tular-
ensis somewhat more tightly than the non-acetylated form, but
does not rule out other explanations. Initial experiments with com-
pound 4 proved it to be stable in aqueous solution, so we pro-
ceeded to test its ability to bind to KdsD. STD experiments
showed that compound 4 was also recognized by KdsD, however,
no further details on the mode of interaction could be deduced
due to the high degree of resonance overlap (see Supplementary,
Fig. S2).
The results with compound 3, its derivatives 3a–3c and com-
pound 4 strongly suggest that acetylation of the hydroxyl group
in position 3 of the inhibitor is detrimental for enzyme binding,
despite the correct stereochemistry of the stereocenter. This result
is in perfect agreement with our previous observations,11,15 which
demonstrated that KdsD is sensitive to steric hindrance at carbon
3. In fact, inversion of the stereochemistry at C-3 (to form
concerning catalysis of ring opening, STD NMR studies can shed
light on the binding of the furanose ring to the enzyme.
We studied the binding of compounds 6 and 7 to Pa-KdsD using
a mixture containing both compounds at a 1:1 molar ratio (Fig. 6),
in order to directly compare their relative affinity for KdsD. The 1H
NMR spectrum of a mixture containing the
a and b anomers (com-
pounds 6 and 7, respectively) and the enzyme dissolved in PBS, pH
7.0, 37 °C, is depicted in Figure 6C. This spectrum was acquired
about 3 h after incubation of 6 and 7 with the protein and, as
expected, neither compound isomerized to the corresponding
ketose. The NMR data indicated that both of the glycosides are KdsD
ligands. However, the b anomer has lower affinity for the protein.
A competitive STD with the natural substrates A5P and Ru5P
was also performed (Supplementary data, Fig. S2). These data indi-
cate that the arabinosides were displaced by A5P and Ru5P, thus
demonstrating that compounds 6 and 7 bind to the enzyme to a
lesser extent than the natural substrates. It is worth noting that
1H NMR spectra of A5P in aqueous solution show that at least
99% of the compound is present in its cyclic hemiacetalic form.
Moreover, a very similar enzyme, glucose 6-phosphate isomerase,
has been reported to bind its carbohydrate substrate in the cyclic
form, and to catalyse ring opening during the isomerisation pro-
cess.21 Taken together, these data suggest that API recognises
A5P in its furanose form, and catalyses ring opening.
D-xylulose 5-phosphate) causes complete loss of binding, despite
all the other functional groups remaining exactly as in the natural
substrates. The active site apparently also accepts 3-deoxyarabi-
nose 5-phosphate as a substrate, although enzymatic interconver-
sion occurs 10-fold slower than the natural substrate.15 The factors
that dictate the binding requirements at position 3 may be
explained in the light of the following considerations.
Finally, compounds 8 and 9 were tested as API ligands in order
to elucidate if the carboxylate may be a good phosphate mimic,
according to the same experimental approach. STD data indicated
that these compounds are not recognized by the enzyme (see Sup-
plementary, Fig. S3). Thus, the carboxylate group is not an accept-
able mimic of the phosphate group present in the natural
substrates.
ꢀ The isomerisation reaction involves the sp2 M sp3 interconver-
sion of C-1 and C-2, which requires the active site to be flexible
enough to accept interconverting tetrahedral M planar geome-
tries around the C-1/C-2 bond. At the same time, the hydrogen
atoms at C-1 and C-2 that are added/removed during isomerisa-
tion must be located within a suitable distance from the cata-
lytic residues.
2.3. Biological assays
ꢀ The carboxylate group is not an acceptable mimic of the phos-
phate group (vide infra).
To more thoroughly investigate the properties of compounds
6–9, we tested their ability to inhibit the growth of P. aeruginosa
in liquid culture. As Pa-KdsD shares a high level of sequence simi-
larity with its E. coli counterpart, we tested an E. coli mutant strain
(AS19) that is more permeable to many antibiotics22 in order to
address the extent to which the Gram-negative bacteria cell barrier
limits the action of the compounds. We also assayed the effect of
the compounds on growth of wild type strain of E. coli
(BW25113) in the presence or absence of glucose-6-phosphate
(G6P). It has been shown that G6P induces the hexose phosphate
transport system (uhp) in E. coli, and that A5P is a high-affinity,
though non-inducing, substrate of this system.9 As shown in
Table 1, none of the compounds inhibited the growth of wild-type
strains of P. aeruginosa or E. coli strain at the concentrations tested.
Addition of G6P did not result in the gain of inhibitory potency,
suggesting that compounds 6–9 may not be substrates of the uhp
system. Interestingly, we did observe slight growth-inhibitory
activity against the permeable mutant AS19, suggesting that the
lack of effect on the wild-type cultures may have been related to
an inability of our compounds to effectively penetrate the wild-
type membrane. Surprisingly, compounds 8 and 9, which were
not recognized by the enzyme in STD NMR experiments, had MICs
against the permeable mutant AS19 cultures that were identical to
those of compounds 6 and 7, suggesting that the inhibitory effect is
unrelated to inhibition of KdsD. We cannot, however, rule out the
possibility that these compounds bind to, and inhibit, enzymes
downstream of KdsD, including KDO8P synthase, KDO8P phospha-
tase or CMP-Kdo synthetase. Regardless of the mechanisms
involved, these MICs confirm our original assertion that a signifi-
cant amount of optimization will be required to convert these
simple compounds into clinically useful inhibitors.
ꢀ The phosphate group is a fundamental requirement for enzyme
recognition and binding.
It seems reasonable to suggest that the enzyme uses the hydro-
xyl group at position 3 and the phosphate at position 5 in order to
lock the substrate into the correct orientation for catalysis, with
appropriate distances between the hydrogen atoms at C1/C2 and
key catalytic residues, and appropriately flexible structure sur-
rounding carbons 1 and 2.
An open question concerns the role of substituents at position 4.
Our STD NMR results indicate that acetylation of the hydroxyl
group at position 4 does not abrogate binding. This suggests there
is either enough flexibility at this position to accommodate an
acetyl group, or the acetyl group is solvent-exposed. The hydroxyl
group at position 4 is responsible for the equilibrium between the
open-chain and furanose forms of A5P. In addition to geometrical
changes between the two forms, there is also a change in the phys-
icochemical properties of the group at C-4. The hydroxyl group at
C4 in the open chain form is a polar protic group that can act as
both hydrogen bond donor and acceptor. In the furanose form,
the polar aprotic furanose oxygen is only able to act as hydrogen
bond acceptor. Previous studies of this, and similar isomerases
have shown that the substrate must be in the open-chain form in
order for the isomerisation to occur. For example, we have shown
that 4-deoxyarabinose 5-phosphate, which is locked in the open-
chain form because it lacks the hydroxyl group at C4, is recognised
by KdsD and isomerised to its ketose isomer at higher rates than
A5P. The question remains, however, whether KdsD, or any API,
is able to bind the furanose form of A5P, and catalyse ring opening
and closing. While our compounds 6–9 cannot answer questions

L. Gabrielli et al. / Bioorg. Med. Chem. 22 (2014) 2576–2583
2581
Figure 6. (A) 1H NMR spectrum of a solution of compound 6 (3 mM); (B) 1H NMR spectrum of a solution of compound 7 (3 mM); (C) 1H NMR spectrum of a mixture
containing of compound 6 (3 mM), compound 7 (3 mM) and API (30
API (30
l
M); (D–I) STD NMR spectra of the a mixture containing of compound 6 (3 mM), compound 7 (3 mM) and
l
M) acquired with different saturation times (D, 3.0 s; E, 2.0 s; F, 1.5 s; G, 1.0 s; H, 0.6 s; I, 0.3 s) and a saturation frequency of 7.9 ppm. All samples were dissolved in
PBS, pH 7.4, 37 °C.
Table 1
at C4 and perhaps at C1–C2. The binding of catalytically inert fura-
nose derivatives to the active site suggests future efforts should in-
clude the design of this type of inhibitor.
MIC determination of compounds 6–9
Strain
MIC (mM)
Compound
4. Experimental
6
7
8
9
P. aeruginosa PA01
E. coli BW25113
E. coli BW25113+G6P
E. coli AS19
>2
>2
>2
0.25
>2
>2
>2
0.25
>2
>2
>2
0.25
>2
>2
>2
0.25
4.1. General methods
When dry conditions were required, solvents were dried over
molecular sieves for at least 24 h prior to use and the reaction
was performed under Ar or N2 atmosphere. Thin-layer chromatog-
raphy (TLC) was performed on silica gel 60F254-coated glass plates
(Merck) with UV detection when possible, or by charring with a
concd H2SO4/EtOH/H2O solution (10:45:45 v/v/v), or through a
solution of (NH4)6Mo7O24 (21 g), Ce(SO4)2 (1 g), and concd H2SO4
(31 mL) in water (500 mL) followed by heating to 110 °C for
5 min. Flash column chromatography was performed on silica gel
230–400 mesh (Merck). Platinum oxide was filtered with an Acro-
3. Conclusions
These results confirm previous reports that the hydroxyl group
at C3 and the phosphate group at C5 of the KdsD substrate are
important for substrate recognition by KdsD. Our results suggest
however, that there is some flexibility at C4 that may be related
to the catalysis of furanose ring opening. We reported the first
experimental evidence that KdsD may bind the furanose form of
A5P, suggesting that the catalysis of ring opening may be an impor-
tant part of KdsD catalysis.
It is also clear from our results that considerable optimisation
will be required to convert the compounds studied here into clin-
ically useful inhibitors. The lack of flexibility at C3 and C5 suggest
that future efforts to optimize open-chain derivatives be directed
disÒ Premium 25 mm Syringe Filter, with 0.45
lm Nylon mem-
brane. Routine 1H and 13C NMR spectra were recorded at
400 MHz (1H) and at 100.57 MHz (13C) on a Varian Mercury instru-
ment. Chemical shifts are reported in parts per million downfield
from TMS as an internal standard; J values are given in Hz. Mass
spectra were recorded on a Applied Biosystems MDS SCIEX instru-
ment (Q TRAP, LC/MS/MS, turbon ion spray).

2582
L. Gabrielli et al. / Bioorg. Med. Chem. 22 (2014) 2576–2583
4.2. Synthesis of methyl
D
-arabinofuranosides (11/12)
at 0 °C, and then the reaction was stirred at room temperature
for 2 h. The reaction was diluted with EtOH, and then the sol-
vents were evaporated. The resulting oil was purified by flash
chromatography (2:8?1:9 PE:EtOAc) to give pure diphenyl
phosphates.
To a solution of
(30 mL), AcCl (100
room temperature for 18 h. The reaction was quenched by the addi-
tion of triethylamine and concentrated to dryness under vacuum.
The resulting crude mixture was purified by flash chromatography
(9.5:0.5 EtOAc:MeOH) and the two anomers were separated to give
D
-arabinose (10) (1 g, 6.67 mmol) in dry MeOH
lL) was added and the reaction was stirred at
4.4.1.
Following the general procedure,
(11) (126 mg, 0.768 mmol) was converted to methyl D
a-Methyl
D
-arabinofuranoside-5-diphenyl phosphate (13)
-methyl -arabinofuranoside
-arabinofu-
a
D
the pure methyl furanosides (970 mg, 89%,
a
/b 1:0.39).
ranoside-5-diphenyl phosphate (13) (colourless oil, 250 mg, 82%).
1H NMR (400 MHz, CDCl3) d 7.56–6.87 (m, 10H, HPh), 4.82 (s, 1H,
H1), 4.46–4.31 (m, 1H, H5), 4.15 (q, J4,3 = 4.1 Hz, 1H, H4), 4.03 (d,
J2,1 = 1.2 Hz, 1H, H2), 3.93 (dd, J3,4 = 4.1, 2.5 Hz, 1H, H3), 3.45 (br
s, 2H, OH), 3.33 (s, 3H, OCH3). 13C NMR (101 MHz, CDCl3) d
150.44 (d, JCq,P = 3.4 Hz, CquatPh), 150.37 (d, JCq,P = 3.4 Hz, CquatPh),
130.01 (C-Ph), 125.73 (C-Ph), 120.21 (C-Ph), 108.97 (C-1), 83.21
(d, JC4,P = 6.8 Hz, C-4), 80.85 (C-2), 77.52 (C-3), 68.59 (d, JC5,P = 6.6 -
Hz, C-5), 55.27 (OCH3). 31P NMR (162 MHz; CDCl3) d À11.67 (1P).
MS: m/z = 419.1 [M+Na], 815.3 [2M+Na].
4.2.1. -Methyl -arabinofuranoside 11
a
D
1H NMR (400 MHz, CD3OD) d 4.68 (d, J1,2 = 1.4 Hz, 1H, H1), 3.85
(dd, J2,3 = 3.5, J2,1 = 1.4 Hz, 1H, H2), 3.84–3.80 (m, 1H, H4), 3.75 (dd,
J3,4 = 6.2, J3,2 = 3.5 Hz, 1H, H3), 3.66 (dd, J5a,5b = 11.9, J5a,4 = 3.2 Hz,
1H, H5a), 3.55 (dd, J5b,5a = 11.9, J5b,4 = 5.3 Hz, 1H, H5b), 3.28 (s,
3H, OCH3). 13C NMR (101 MHz, CD3OD) d 111.25 (C-1), 86.18 (C-
4), 84.07 (C-2), 79.40 (C-3), 63.77 (C-5), 53.9 (OCH3). MS: m/
z = 187.1 [M+Na].
4.2.2. b-Methyl D-arabinofuranoside 12
1H NMR (400 MHz, CD3OD) d 4.63 (d, J1,2 = 4.4 Hz, 1H, H1), 3.85
(dd, J2,3 = 7.6, J2,1 = 4.4 Hz, 1H, H2), 3.79 (t, J3,4 = 7.1 Hz, 1H, H3),
3.94–3.63 (m, 1H, H4), 3.57 (dd, J5a,5b = 11.6, J5a,4 = 3.6 Hz, 1H,
H5a), 3.44 (dd, J5b,5a = 11.6, J5b,4 = 7.2 Hz, 1H, H5b), 3.30 (s, 3H,
OCH3). 13C NMR (101 MHz, CD3OD) d 104.44 (C-1), 84.87 (C-4),
79.45 (C-3), 77.25 (C-2), 65.99 (C-5), 55.96 (OCH3). MS: m/
z = 187.1 [M+Na].
4.4.2. b-Methyl
The b-methyl
was converted to the corresponding methyl
5-diphenyl phosphate 14 (colourless oil, 228 mg, 75%) following
the general procedure. 1H NMR (400 MHz, CDCl3) d 7.36–7.11 (m,
10H, HPh), 4.75 (d, J1,2 = 3.7 Hz, 1H, H1), 4.37–4.21 (m, 2H, H5a,
H5b), 4.07–3.97 (m, 3H, H2, H4, H3), 3.71 (br s, 2H, OH), 3.33 (s,
3H, OCH3). 13C NMR (101 MHz, CDCl3) d 150.44 (d, JCq,P = 3.4 Hz,
D
-arabinofuranoside-5-diphenyl phosphate (14)
-arabinofuranoside 12 (126 mg, 0.768 mmol)
-arabinofuranoside-
D
D
4.3. General procedure for the synthesis of methyl
arabinuronic acid salts (8/9)
D
-
CquatPh), 150.37 (d, JCq,P = 3.4 Hz, CquatPh), 129.95 (C-Ph), 125.65
(C-Ph), 120.16 (C-Ph), 102.25 (C-Ph), 80.24 (d, JC4,P = 7.4 Hz, 1C,
C4), 77.81 (C-3), 75.82 (C-2), 69.74 (d, JC5,P = 5.5 Hz, C-5), 55.37
(OCH3). 31P NMR (162 MHz; CDCl3) d À11.94 (1 P). MS: m/
z = 419.1 [M+Na], 815.3 [2M+Na].
Methyl D-arabinofuranosides (11/12) were dissolved in satu-
rated aqueous NaHCO3. A 0.5 M aqueous solution of potassium
bromide (0.3 equiv) and a 0.1 M acetonitrile solution of TEMPO
(0.5 equiv) were added under stirring. A 0.35 M NaOCl solution
(1.5 equiv) was added slowly and the reaction was stirred for 2 h
at room temperature. The solvent was evaporated under vacuum
and the resulting solids were purified by flash chromatography
(2-propanol:NH4OH 33%, 8:2), affording the corresponding ammo-
nium carboxylates.
4.5. General procedure for the synthesis of methyl
arabinofuranoside-5-phosphate sodium salt (6/7)
D-
Methyl
D-arabinofuranoside-5-diphenyl phosphates (13/14)
were dissolved in EtOH and PtO2 was added. The mixture was then
stirred under a H2 atmosphere for 15 h. The catalyst was filtered off
and the solvent evaporated. The solids were re-suspended in water
and the pH was adjusted to pH 7 by addition of 5% aq NaOH. The
solution was then freeze-dried, affording the phosphate disodium
salt.
4.3.1.
a-Methyl D-arabinuronic acid ammonium salt (8)
a
-Methyl D-arabinofuranoside (11) (235 mg, 1.433 mmol) was
oxidised following the general procedure; compound 8 was ob-
tained as a yellowish solid (180 mg, 64%). 1H NMR (400 MHz,
D2O) d 4.96 (d, J1,2 = 1.0 Hz, 1H, H1), 4.26 (d, J4,3 = 4.3 Hz, 1H, H4),
4.09 (dd, J3,4 = 4.3, J3,2 = 2.3 Hz, 1H, H3), 3.96 (dd, J2,3 = 2.3,
J2,1 = 1.0 Hz, 1H, H2), 3.37 (s, 3H, OCH3). 13C NMR (101 MHz, D2O)
d 180.92 (C-5), 107.27 (C-1), 82.43 (C-4), 78.43 (C-2), 77.85 (C-3),
53.47 (OCH3). MS: m/z = 177.1 [MÀ1].
4.5.1.
(6)
a
-Methyl
D
-arabinofuranoside-5-phosphate disodium salt
-methyl -arabinofuranoside-5-diphenyl
Hydrogenolysis of
a
D
phosphate (13) using the general procedure described above affor-
ded compound 6 (colourless oil, 112 mg, 99%). 1H NMR (400 MHz,
D2O) d 4.74 (d, J1,2 = 1.3 Hz, 1H), 3.99–3.91 (m, 2H, H4, H5a), 3.90–
3.79 (m, 3H, H2, H3, H5b), 3.21 (s, 3H, OCH3). 13C NMR (101 MHz,
D2O) d 108.12 (C-1), 82.09 (d, JC4,P = 8.2 Hz, C4), 80.28 (C-2), 75.82
(C-3), 64.89 (d, JC5,P = 5.1 Hz), 54.79 (OCH3). 31P NMR (162 MHz;
D2O) d 6.21 (1P). MS: m/z = 243.1 [MÀ1].
4.3.2. b-Methyl
Following the general procedure, b-methyl
(12) (156 mg, 0.951 mmol) was converted to methyl
D
-arabinuronic acid ammonium salt (9)
-arabinofuranoside
-arabinuron-
D
D
ic acid ammonium salt 9 (yellowish solid, 115 mg, 62%). 1H NMR
(400 MHz, D2O) d 4.88 (d, J = 4.5 Hz, 1H, H1), 4.26–4.19 (m, 1H,
H3), 4.10 (d, J = 6.5 Hz, 1H, H4), 4.04 (dd, J = 7.5, 4.5 Hz, 1H, H2),
3.43 (s, 3H, OCH3). 13C NMR (101 MHz, D2O) d 177.18 (C-5),
102.54 (C-1), 80.15 (C-4), 76.95 (C-3), 76.18 (C-2), 55.42 (OCH3).
MS: m/z = 177.1 [MÀ1].
4.5.2. b-Methyl D-arabinofuranoside-5-phosphate sodium salt
(7)
The b-methyl
D-arabinofuranoside-5-diphenyl phosphate (14)
(81 mg, 0.204 mmol) was deprotected to the corresponding methyl
D
-arabinofuranoside-5-phosphate disodium salt 7 (colourless oil,
4.4. General procedure for the synthesis of methyl
arabinofuranoside-5-diphenyl phosphates (13/14)
D
-
57 mg, 97%) using the general procedure above. 1H NMR
(400 MHz, D2O) d 4.76 (d, J1,2 = 4.4 Hz, 1H, H1), 4.01 (dd, J2,3 = 8.0,
J2,1 = 4.4 Hz, 1H, H2), 3.96–3.77 (m, 4H, H3, H4, H5a, H5b), 3.27
(s, 3H, OCH3). 13C NMR (101 MHz, D2O) d 101.87 (C-1), 80.09 (d,
JC4,P = 8.5 Hz, C4), 75.92 (C-3), 74.00 (C-2), 66.47 (d, JC5,P = 5.1 Hz,
Methyl
D-arabinofuranosides (11/12) (1 equiv) were dissolved
in dry pyridine. chlorodiphenylphosphate (1.1 equiv) was added

L. Gabrielli et al. / Bioorg. Med. Chem. 22 (2014) 2576–2583
2583
C5), 54.78 (OCH3). 31P NMR (162 MHz; D2O) d 6.10 (1P). MS: m/
4.7.2. MIC determination
z = 243.1 [MÀ1].
Minimal inhibitory concentrations (MIC) were determined
following standard protocols for testing susceptibility to antibiotic
agents.24 The bacterial strains used were Pseudomonas aeruginosa
PAO1 (ATCC 15692), Escherichia coli K12 wild type strain
BW25113,25 and the permeable E. coli mutant strain AS19.21 The
procedure for a broth microdilution test was followed. Briefly,
96-well microtiter plates were prepared with compound concen-
4.6. NMR spectroscopy epitope mapping studies
All of the experiments were recorded with a Varian Mercury
(400 MHz) or a Bruker Advance III (600 MHz) instrument equipped
with a cryoprobe. The ligand resonances were assigned by 1H, 1H
COSY and 1H, 13C HSQC NMR spectroscopy. A basic 1D-STD
sequence was used with the on-resonance frequency of 7.9 ppm
trations ranging from 2 mM to 2
BW25113, the procedure was performed in the presence or
absence of 10 M glucose 6-phosphate. The assayed strains were
lM, in triplicate. For E. coli
and the off-resonance frequency of 40 ppm. A train of GAUSSIAN
-
l
shaped pulses of 50 ms each was employed, with total saturation
times of the protein envelope ranging from 5.0 to 0.3 s. The total
saturation time was adjusted by the number of shaped pulses. A
grown on Mueller–Hinton broth (MHB) and inoculums of approx-
imately 5 Â 105 cfu mLÀ1 were added to each well. The microtiter
plates were incubated for 18–20 h at 37 °C and the MIC of each
compound determined. MIC was defined as the lowest concentra-
tion inhibiting bacterial growth.
T1 filter of 2 ms was employed to eliminate the background sig-
q
nals from the protein. All of the samples were dissolved in deuter-
ated PBS (pH 7.4) at 25 or 37 °C. Total sample volumes were
550 mL. The on- and off-resonance spectra were acquired simulta-
neously, with the same number of scans. The STD spectrum was
obtained by subtraction of the on-resonance spectrum from the
off-resonance spectrum. The subtraction was performed by phase
cycling to minimize artefacts arising from magnet and temperature
instabilities. Reference experiments with samples containing only
the free compounds tested were performed under the same exper-
imental conditions to verify true ligand binding. The effects
observed in the presence of the protein were due to true saturation
transfer, because no signals were present in the STD spectra
obtained in the reference experiments, except for residues from
HDO, indicating that artefacts from the subtraction of compound
signals were negligible.
Acknowledgments
This project was funded by Fondazione Cariplo (Grant No. 2010-
0653), Regione Lombardia ‘Cooperazione scientifica e tecnologica
internazionale’ (Grant 16876 SAL-18), MIUR-Regione Lombardia
(ID 30190679), Fondazione Banca del Monte di Lombardia, and
MIUR under project PRIN2008/24M2HX.
Supplementary data
Supplementary data associated with this article can be found, in
the online version, at http://dx.doi.org/10.1016/j.bmc.2013.08.012.
References and notes
4.7. Biological assays
1. Fischbach, M. A.; Walsh, C. T. Science 2009, 325, 1089.
2. Brötz-Oesterhelt, H.; Sass, P. Future Microbiol. 2010, 5, 1553.
3. Nikaido, H. Microbiol. Mol. Biol. Rev. 2003, 67, 593.
4. Gabrielli, L.; Capitoli, A.; Bini, D.; Taraballi, F.; Lupo, C.; Russo, L.; Cipolla, L. Curr.
Drug Targets 2012, 13, 1458.
5. Cipolla, L.; Polissi, A.; Airoldi, C.; Gabrielli, L.; Merlo, S.; Nicotra, F. Curr. Med.
Chem. 2011, 18, 830.
6. Raetz, C. R.; Whitfield, C. Annu. Rev. Biochem. 2002, 71, 635.
7. Holst, O. FEMS Microbiol. Lett. 2007, 271, 3.
8. Meredith, T. C.; Woodard, R. W. J. Biol. Chem. 2003, 278, 32771.
9. Meredith, T. C.; Woodard, R. W. J. Bacteriol. 2005, 187, 6936.
10. Sperandeo, P.; Pozzi, C.; Dehò, G.; Polissi, A. Res. Microbiol. 2006, 157, 547.
11. Airoldi, C.; Sommaruga, S.; Merlo, S.; Sperandeo, P.; Cipolla, L.; Polissi, A.;
Nicotra, F. ChemBioChem 2011, 12, 719.
12. Sommaruga, S.; De Gioia, L.; Tortora, P.; Polissi, A. Biochem. Biophys. Res.
Commun. 2009, 388, 222.
13. Gourlay, L. J.; Sommaruga, S.; Nardini, M.; Sperandeo, P.; Dehò, G.; Polissi, A.;
Bolognesi, M. Protein Sci. 2010, 19, 2430.
14. Yep, A.; Sorenson, R. J.; Wilson, M. R.; Showalter, H. D. H.; Larsen, S. D.; Keller, P.
R.; Woodard, R. W. Bioorg. Med. Chem. Lett. 2011, 21, 2679.
15. Airoldi, C.; Sommaruga, S.; Merlo, S.; Sperandeo, P.; Cipolla, L.; Polissi, A.;
Nicotra, F. Chem. Eur. J. 2010, 16, 1897.
16. Mayer, M.; Meyer, B. Angew. Chem., Int. Ed. 1999, 38, 1784.
17. Caraballo, R.; Dong, H.; Ribeiro, J. P.; Jiménez-Barbero, J.; Ramstrom, O. Angew.
Chem., Int. Ed. 2010, 49, 589.
18. Dangerfield, E. M.; Gulab, S. A.; Plunkett, C. H.; Timmer, M. S. M.; Stocker, B. L.
Carbohydr. Res. 2010, 345, 1360.
19. Haselhorst, T.; Fiebig, T.; Dyason, J. C.; Fleming, F. E.; Blanchard, H.; Coulson, B.
S.; von Itzstein, M. Angew. Chem., Int. Ed. 2011, 5, 1055.
4.7.1. Expression and purification of Pa-KdsD
Pa-KdsD (PA4457) was expressed from pQE31/Pa-KdsD, a plas-
mid that encodes Pa-KdsD with an N-terminal hexahistidine tag.11
Cultures of E. coli M15/pREP4 harbouring the pQE31/Pa-KdsD plas-
mid were grown overnight in LD broth23 supplemented with ampi-
cillin (100
fresh medium (1 L) and grown to mid-logarithmic phase (OD600
l l
g mLÀ1) and kanamycin (25 g mLÀ1), diluted 1:100 in
,
0.6) at 37 °C. Induction of Pa-KdsD was then carried out for 4 h post
addition of IPTG (0.5 mM). Cells were harvested, washed with
sodium phosphate (20 mM, pH 7.0) and re-suspended in buffer
containing sodium phosphate (50 mM, pH 8.0), NaCl (300 mM),
imidazole (10 mM) supplemented with DNase (0.1 mg mLÀ1) and
phenylmethylsulfonyl fluoride (1 mM). After treatment with lyso-
zyme (1 mg mLÀ1) for 30 min, the cells were disrupted by a single
cycle through a Cell Disruptor (One Shot Model by Constant
Systems LTD) at 25,000 psi, and the unbroken cells were removed
by centrifugation at 39,000g for 30 min. The supernatant was
loaded at a flow rate of 0.5 mL minÀ1 onto a Ni-NTA Agarose
(Qiagen) column (bed volume 5 mL) pre-equilibrated with sodium
phosphate (50 mM, pH 8.0) and NaCl (300 mM). The column was
washed with 10 volumes of sodium phosphate (50 mM, pH 8.0),
NaCl (300 mM) and imidazole (20 mM) and the enzyme was eluted
with a 5 mL step-wise imidazole gradient (100 to 300 mM) in the
same buffer. Fractions (5 mL) were collected and those displaying
the highest purity (>95%, as assessed by SDS–PAGE analysis) were
dialyzed in sodium phosphate (50 mM, pH 8.0). The isomerase
activity of the protein was assessed as described previously.11 Pro-
20. Airoldi, C.; Giovannardi, S.; La Ferla, B.; Jiménez-Barbero, J.; Nicotra, F. Chem.
Eur. J. 2011, 17, 13395.
21. Jeffery, C. J.; Hardre, R.; Salmon, L. Biochemistry 2001, 40, 1560.
22. Sekiguchi, M.; Iida, S. Proc. Natl. Acad. Sci. U.S.A. 1967, 58, 2315.
23. Sabbattini, P.; Forti, F.; Ghisotti, D.; Dehò, G. J. Bacteriol. 1995, 177, 1425.
24. Wiegand, I.; Hilpert, K.; Hancock, R. E. Nat. Protoc. 2008, 3, 163.
25. Datsenko, K. A.; Warren, B. L. Proc. Natl. Acad. Sci. U.S.A. 2000, 97, 6640.
tein aliquots (600
l
g each) were lyophilized and stored at À80 °C.