Structure-activity relationships of galabioside derivatives as inhibitors of E. coli and S. suis adhesins: Nanomolar inhibitors of S. suis adhesins

Article abstract of DOI:10.1039/b416878j

Four collections of Galα1-4Gal derivatives were synthesised and evaluated as inhibitors of the PapG class II adhesin of uropathogenic Escherichia coli and of the P_N and P_O adhesins of Streptococcus suis strains. Galabiosides carrying

Full text of DOI:10.1039/b416878j

View Article Online / Journal Homepage / Table of Contents for this issue
A R T I C L E
Structure–activity relationships of galabioside derivatives as
inhibitors of E. coli and S. suis adhesins: nanomolar inhibitors of
S. suis adhesins†
Jo¨rgen Ohlsson,a Andreas Larsson,b Sauli Haataja,c Jenny Alaja¨a¨ski,c Peter Stenlund,d
Jerome S. Pinkner,e Scott J. Hultgren,e Jukka Finne,*c Jan Kihlberg*b and Ulf J. Nilsson*a
a Organic Chemistry, Lund University, P.O. Box 124, SE-221 00 Lund, Sweden.
E-mail: ulf.nilsson@organic.lu.se; Fax: (+46) 46 222 82 09; Tel: (+46) 46 222 82 18
b Organic Chemistry, Department of Chemistry, Umea˚ University, SE-901 87 Umea˚, Sweden
c Department of Medical Biochemistry and Molecular Biology, University of Turku, FI-20520
Turku, Finland
d Biochemistry, Department of Chemistry, Umea˚ Universitet, SE-901 87 Umea˚, Sweden
e Department of Molecular Biology, Washington University School of Medicine, St. Louis,
Missouri 63110, USA
Received 9th November 2004, Accepted 10th January 2005
First published as an Advance Article on the web 4th February 2005
Four collections of Gala1-4Gal derivatives were synthesised and evaluated as inhibitors of the PapG class II adhesin
of uropathogenic Escherichia coli and of the PN and PO adhesins of Streptococcus suis strains. Galabiosides carrying
aromatic structures at C1, methoxyphenyl O-galabiosides in particular, were identified as potent inhibitors of the
PapG adhesin. Phenylurea derivatisation at C3ꢀ and methoxymethylation at O2ꢀ of galabiose provided inhibitors of
the S. suis strains type PN adhesin with remarkably high affinities (30 and 50 nM, respectively). In addition,
quantitative structure–activity relationship models for E. coli PapG adhesin and S. suis adhesin type PO were
developed using multivariate data analysis. The inhibitory lead structures constitute an advancement towards
high-affinity inhibitors as potential anti-adhesion therapeutic agents targeting bacterial infections.
colonizer of the pig respiratory tract. Gala1-4Gal terminating
oligosaccharides have been shown to be optimal receptors for S.
suis. Systematic competitive inhibition studies characterized the
Introduction
The alarming increase in the resistance of bacteria to traditional
antibiotics1–3 makes it imperative to develop alternative ways of
key hydroxyl groups that are required for binding to Gala1-4Gal
treating bacterial infections. The majority of infectious diseases
and also classified the adhesion activities into two types, PN and
are initiated by adhesion of pathogenic organisms to host tissue
PO.16
and in many cases glycoconjugates present on the mammalian
Most natural carbohydrate ligands bind lectins with low
cell surface, e.g. glycoproteins and glycolipids, act as receptors
affinity (Kd normally in the 0.1–1 mM range). One attractive
for a wide variety of extracellular bacterial proteins termed
strategy to overcome this problem is to use a small key saccharide
adhesins.4,5 This carbohydrate–protein interaction is often a pre-
as the core structure and attach substituents that interact
requisite for the later stages of bacterial infection and inhibitors
with the lectin in a favourable manner. It has been shown
of this recognition process are potential pharmaceutical agents.
that for the PapG class II adhesin the galabiose disaccharide
Bacterial resistance towards such anti-adhesive drugs is believed
is such a key structure17 necessary for recognition and that
to evolve slowly because the infecting bacteria are not killed and
galabioside derivatives substituted at C1 and C3ꢀ often display
are consequently not under selection pressure.
enhanced affinity for the PapG adhesins.18 Furthermore, the
Two well-known examples of pathogenic bacteria adhering to
crystal structure of the class II PapG adhesin in complex with the
glycoconjugates are uropathogenic Escherichia coli, which is the
main cause of urinary tract infections, and Streptococcus suis,
which causes meningitis in pig and man. The majority of E.
coli bacteria causing pyelonephritis (kidney infection) adhere
globotetraose tetrasaccharide has been solved19 and it confirmed
that the galabiose disaccharide unit is the most critical structural
element for the formation of the complex. The complex structure
revealed an extended surface composed of H1, H2 and H6 of
via proteinaceous appendices, termed P-pili. These pili are
Gala, H1, H3, H4, H5 and H6 of Galb, and H2 and H4 of
terminated with an adhesin, PapG, that binds to the Gala1-4Gal
Glc, that make hydrophobic contact primarily with the Trp107
(galabiose)6–8 moiety present in the globoseries of glycolipids on
sidechain. The adhesin additionally forms four hydrogen bonds
uroepithelial cells and erythrocytes. Three different classes of
to HO4, O5 and HO6 of the non-reducing GalNAc residue,
seven to HO2, O3, HO4 and HO6 of the Gala residue, four to
HO3, O5 and HO6 of the Galb, and three to HO2 and HO3 of
the Glc residue. Apparently, the two residues flanking galabiose,
GalNAc and Glc, are involved via hydrophobic contacts and
hydrogen bonding (e.g. Lys172 to GalNAc and Arg170 and
the PapG adhesin (classes I–III)9,10 have been identified based
on different erythrocyte agglutination patterns. Pyelonephritis
in both children and adult women is associated with PapG
class II,11,12 while class III is associated with cystitis in adult
women.13 In addition to anchoring the bacteria to the host cell,
the adhesion of PapG induces the release of ceramides14 that
Trp107 to Glc), but to a lesser extent.
are important second messenger molecules, and results in up-
The galabiose disaccharide has also been shown to be the
regulation of, and eventual secretion of, several immunoregula-
key recognition element for adhesins from S. suis strains of
both type PN and PO. The two adhesins from S. suis, however,
tory cytokines from host cells.15 Streptococcus suis is a frequent
display differences in the sub-molecular details of galabiose
recognition.16 Multivalent galabiose derivatives have been re-
† Jo¨rgen Ohlsson, Andreas Larsson and Sauli Haataja contributed
ported to display greatly enhanced affinity.20 However, although
equally.
8 8 6
O r g . B i o m o l . C h e m . , 2 0 0 5 , 3 , 8 8 6 – 9 0 0
T h i s j o u r n a l i s
T h e R o y a l S o c i e t y o f C h e m i s t r y 2 0 0 5
©

View Article Online
multivalent galabiosides provide potent inhibitors, they all share
the disadvantages of large size and high polarity resulting in poor
bioavailability.
are probably due to the inherent properties of galactose (or
galabiose).
Collection II galabiose derivatives were prepared from ei-
ther the galabiosyl a-trichloroacetimidate 2 (compounds 11–
18) or from the galabiosyl bromide 1 (compounds 19–22).
Trimethylsilyl trifluoromethanesulfonate (TMSOTf)-promoted
glycosylation of aromatic and aliphatic alcohols with the
trichloroacetimidate 2, followed by deacylation in methanolic
sodium methoxide, furnished compounds 11–18 in 44–83%
yield (b/a ∼15 : 1). Under these conditions, phenols car-
rying electron withdrawing groups typically gave low yields
and poor a/b-selectivities. Instead, nucleophilic displacement
of the galabiosyl bromide 1 with the corresponding sodium
phenolates fortunately afforded compounds 19–22 in 36–70%
yield after deacylation. Complete b-selectivity was observed
for compounds 20–22, while compound 19 had a b/a-ratio of
about 10 : 1. Compound 23 was prepared from 22 by catalytic
hydrogenation followed by acylation under conditions similar to
those described for the preparation of collection I.
Galabioside collection III was prepared by nucleophilic
displacement of the galabiosyl bromide 1 with thiophenolates,
followed by deacylation, to furnish b-galabiosides 24–27 in 76–
88% yield.
Synthesis of collection IV required the introduction of a han-
dle (amine) at C3ꢀ of p-methoxyphenyl galabioside (Scheme 2).
Henceforth, the known galactoside 2825 was deacylated and
benzylated to give the galactosyl donor 29. a-Galactosylation
of the acceptor 30,22 using N-iodosuccinimide-trimethylsilyl
trifluoromethanesulfonate as promoter,26,27 gave the protected 3ꢀ-
azido galabioside 31 in 93% yield. Debenzoylation in methanolic
sodium methoxide gave 32 in 80% yield, the key starting material
for the synthesis of galabioside collection IV.
The present paper describes an attempt to improve the affinity
of inhibitors for the PapG class II adhesins and the two S.
suis adhesins by the synthesis of four collections of galabiosides
modified at C1 and C3ꢀ in four different ways. In order to position
the substituents at C1 as close as possible to the galabiose
core structure, either amide formation (galabioside collection I),
glycosylation of alcohols (collection II), or of thiols (collection
III) were employed. At C3ꢀ, amide formation was used to
create structural diversity on a p-methoxyphenyl galabioside
core structure (collection IV). Recently reported synthesis and
evaluation of earlier generations of galabiose derivatives has
shown that the p-methoxyphenyl galabioside itself is an excellent
inhibitor of the PapG adhesins.18,21 The further evaluation of
these earlier generations of galabiose derivatives against the two
S. suis adhesins is also reported herein. In addition, quantitative
structure–activity relationship models for E. coli PapG adhesin
and S. suis adhesin type PO were developed using multivariate
data analysis.
Results and discussion
Synthesis of galabiose collections I–IV (Table 1)
For the synthesis of the collection I galabiosides, the galabiosyl
bromide 122 was converted to the b-azide 4 (Scheme 1) via
treatment with trimethylsilyl azide and tetrabutylammonium
fluoride,23 (3, 92%) followed by conventional deacylation with
methanolic sodium methoxide to give 4 in 96% yield.
Scheme 2 a) iNaOMe, MeOH, iiNaH, BnBr, DMF, 86%. b) NIS,
TMSOTf, CH2Cl2–Et2O (1 : 2), −50 ◦C, 93%. c) NaOMe, MeOH, 80%.
d) See Table 1.
Scheme 1 a) TMSN3, Bu4NF, THF, 92%. b) NaOMe, MeOH, 96%.
The azido group of galabioside 32 was reduced to the amine
33 under reaction conditions similar to those described for the
preparation of collection I, with the exception that 2 equivalents
of HCl were added to ensure complete hydrogenolysis of the
benzyl groups. The 3ꢀ-amino galabioside 33 was treated with
sodium carbonate and either an acyl chloride (34–35 and 37–
41), acid anhydride (36 and 42), or isocyanate (43–44) to give
galabiosides 34–44 in 49–83% yield.
For c) and d) see Table 1.
Hydrogenation of 4 over Pd/C in methanol gave an inter-
mediate amine, which was immediately converted to amides by
treatment with five different acid chlorides in the presence of
sodium carbonate in THF, affording compounds 5–9 in 48–90%
overall yields and b/aratios of ∼15 : 1. No product was observed
under these reaction conditions with a hindered acid chloride
(i.e. 10). However, acylation with the acid chloride, pyridine, and
a catalytic amount of DMAP, gave compound 10 in a moderate
28% yield. Attempts to increase the yields for compounds
6–9 with these latter reaction conditions were unsuccessful.
Furthermore, different reaction conditions were unsuccessfully
evaluated in attempts to improve the b-selectivities (i.e. PtO2 as a
hydrogenation catalyst, one-pot azide reduction and acylation,
hydrogenation in the presence of HCl, and the use of different
solvents). It is likely that the anomerisation of the amine
occurs faster than the acylation, because similar a/b ratios were
obtained under all conditions tried. No a-anomer was detected
when fully acylated lactosyl azide24 was reacted under the
same reaction conditions, suggesting that the resulting anomeric
mixtures obtained in the case of the galabiose derivatives 5–10
Binding of the E. coli PapG class II adhesin to galabioside
derivative collections I–IV
Binding of galabiose derivative collections I–IV by an N-
terminal 196 amino acid truncate of the class II PapG adhesin19
was determined by surface plasmon resonance as recently
described.21 The binding data were analysed using the software
Scrubber28 (Fig. 1, Table 1).
The presence of amides at C1 of galabiose (collection I) turned
out to be detrimental to binding. The benzamido derivatives (5–
8, and 10) displayed Kd of 1.0–2.3 mM, which is much worse
than that of the known reference compound p-methoxyphenyl
galabioside 4518 (Kd 140 lM21). Virtually no interaction was seen
with an aliphatic amide at C1 (9). The amide-functionality at
O r g . B i o m o l . C h e m . , 2 0 0 5 , 3 , 8 8 6 – 9 0 0
8 8 7

View Article Online
8 8 8
O r g . B i o m o l . C h e m . , 2 0 0 5 , 3 , 8 8 6 – 9 0 0

View Article Online
O r g . B i o m o l . C h e m . , 2 0 0 5 , 3 , 8 8 6 – 9 0 0
8 8 9

View Article Online
8 9 0
O r g . B i o m o l . C h e m . , 2 0 0 5 , 3 , 8 8 6 – 9 0 0

View Article Online
O r g . B i o m o l . C h e m . , 2 0 0 5 , 3 , 8 8 6 – 9 0 0
8 9 1

View Article Online
8 9 2
O r g . B i o m o l . C h e m . , 2 0 0 5 , 3 , 8 8 6 – 9 0 0

View Article Online
Fig. 1 Equilibrium isotherms fit to a 1 : 1 interaction model for 45 (ꢁ),
25 (ꢀ), 46 (ꢁ) and 36 (᭺) binding to PapGII immobilized to a CM5
surface plasmon resonance biosensor surface. The binding responses
at equilibrium were normalised against maximum binding (Rmax). In
cases where the affinity of the galabioside was too low (36) to saturate an
expected binding isotherm, the Rmax was determined using a reference
substance and the galabioside’s molecular weight.
C1 probably positions the aromatic rings of 5–8 and 10 in non-
favourable positions relative to the side chains of Trp107 and
Arg170 of PapGII. In contrast, the p-methoxyphenyl glycoside
of galabiose 45 positions the aromatic ring to interact favourably
with Trp107 and Arg170, resulting in a Kd as low as 140 lM.21
Collection II (O-galabiosides 11–23) turned out to be more
successful in providing ligands for the class II PapG ad-
hesin. All compounds showed higher affinities for the adhesin
than those observed for aliphatic galabiosides (i.e. the 2-
(trimethylsilyl)ethyl galabioside 4618). The positions of methoxy
groups on the aromatic aglycons had some impact on the affinity.
p-Methoxyphenyl galabioside 45 and m-methoxyphenyl galabio-
side 11 had virtually the same Kd. However, the Kd (176 lM) for
o-methoxyphenyl galabioside 12 was somewhat higher than 45
and close to that of the phenyl galabioside 13 (170 lM). Thus,
the methoxyphenyl group interacts favourably with the adhesin
when positioned in the m- or p-position. Exchange of the p-
methoxy for a p-methyl group (i.e. 14) resulted in an increase
in Kd in the same range as removal of the p-methoxy group
did. This suggests that the oxygen atom of the p-methoxy group
is important for affinity to the adhesin. Introduction of other
groups at the p-position of a phenyl aglycon (i.e. methylester 20,
nitro 22, or acetamido 23) resulted in galabiosides with the same
Kd as the p-methylphenyl galabioside 14, as did introduction of
a methylester at the o-position (21). In contrast, the introduction
of fluorine (19) resulted in a large increase in Kd to 373 lM, i.e.
2.5 times of that observed for phenyl galabioside 13. Increasing
the size of the aromatic substituent (i.e. naphthyl 15 or indolyl
16) or moving the aromatic substituent away from the galabiose
C1 (i.e. benzyl galabioside 17), resulted in lowered affinity for
the adhesin compared to phenyl galabioside 13, further demon-
strating the sensitivity towards the exact position of the aromatic
substituents at C1. Furthermore, cyclohexyl galabioside 18 (Kd
269 lM) was a less potent inhibitor than phenyl galabioside 13,
suggesting that aromatic aglycons are beneficial. Most likely,
phenyl aglycons of galabiosides stabilise complex formation via
interactions with the aromatic side chain of Trp107 and with the
guanidino group of Arg170 in the PapGII adhesin.21
Exchange of the anomeric oxygen atom for sulfur results in
more hydrolytically stable galabiosides. However, the phenyl
thio-galabiosides 24–27 (collection III) had Kd values in the
320–600 lM range, i.e. about one third of the affinity of
their O-glycosidic counterparts. An altered conformation of
phenyl thio-galabiosides, compared to the corresponding O-
glycosides,29 most likely explains the higher Kd values observed
for these compounds. The aromatic aglycons of 24–26 are folded
O r g . B i o m o l . C h e m . , 2 0 0 5 , 3 , 8 8 6 – 9 0 0
8 9 3

View Article Online
Table 2 IC50 values for the 10 best and 8 poorest inhibitors of the
adhesins from S. suis strains type PN and type PO
back onto the galabiose disaccharide moiety, which causes a
conformational change in the a(1–4) disaccharide linkage of
galabiose. Hence, the phenyl thio-galabiosides 24–26 must adopt
a high-energy conformation in order to be recognised by the class
II PapG adhesin.
IC50/lM S. suis PN Range
IC50/lM S. suis PO Range
5
5.2
3.9–7.8
2.0–7.8
Exchange of a hydroxyl for an amino group at C3ꢀ (33)
turned out to be detrimental for binding; the Kd value for 33
was 29 times higher than that of p-methoxyphenyl galabioside
45. Functionalisation of the amine at C3ꢀ (34–44) resulted in
even worse inhibitors (Kd 10–500 mM). The reason for the low
affinity could be that the interaction between the Lys172 and O3ꢀ
seen in the crystal structure between the adhesin and the Gb4
tetrasaccharide is lost when an amine, amide or urea replaces
the hydroxyl at C3ꢀ of the galabiose disaccharide.
12 4.4
13
0.42
0.31–0.63
15 0.20
19
0.16–0.31
0.31
0.42
0.31
0.42
0.21
0.31
0.31
0.31–0.63
0.31
0.31–0.63
0.16–0.31
0.31
22
23
24
25
26
33 15.6
35
15.6
54.7
52.1
20.8
31.3–62.5
31.3–62.5
15.6–31.3
37
38
Binding of the S. suis adhesins type PN and type PO to
galabioside derivative collections I–IV (5–44) and to known
galabioside derivatives 45–64
39 0.16
40
41 0.32
42
43 0.18
44 0.03
45 0.18
47 0.63
49
0.08–0.31
0.16–0.63
62.5
41.7
62.5
62.5
62.5
31.3–62.5
62.5
Binding of galabiose derivatives 5–64 by S. suis types PN and
PO was determined by hemagglutination inhibition essentially
as previously described.30 The galabiose substituents of 5–64
clearly interact with the two S. suis adhesins as the structures
of the substituents exert large influences on their inhibitory
powers (Table 1). Furthermore, the two adhesins display dif-
ferent preferences with regard to the substituents’ position
and structure, which is consistent with earlier observations.16
The recognition patterns of the type PN and PO adhesins are
also markedly different to that of the E. coli class II adhesin.
For S. suis, galabiosyl amides 5–10 (collection I) were poor
inhibitors suggesting that O-galabiosides are preferred, possibly
because the adhesins donate hydrogen bonds to galabiose O1.
Hydrophobic (aromatic) galabioside aglycons (11–27, collection
II) had, in general, a minor influence on the inhibitory powers
compared to the reference methyl galabioside 47, except for
phenyl aglycons carrying an ortho-substituent (12 and 21) or
an m-methoxy group (11) which were detrimental to inhibition.
Introducing amides at galabiose C3ꢀ (34–42) was also detri-
mental to inhibition of the type PO adhesin, while it was well
tolerated by the type PN adhesin. Ureas at C3ꢀ (43–44) were, not
surprisingly, poor inhibitors of the type PO adhesin. However,
the ureas 43–44 were potent inhibitors of the type PN adhesin.
This correlates with the previously suggested Gal/GalNAc
binding pocket at this site, which differentiates the PN adhesin
from the PO adhesin.16 Galabiose derivatives modified at O2ꢀ
with alkyl groups (48–51) appeared to be accepted by the type
PN adhesin. Interestingly, O2ꢀ-methoxymethyl substitution (51)
provided one of the two best inhibitors against this adhesin,
indicating the proximity of a hydrogen bond donor in the
adhesin. Galabiosides carrying alkoxy-substituents at C3ꢀ (52–
64) were all relatively poor inhibitors of both S. suis adhesins.
The screening experiments against the two S. suis adhesins
were further confirmed by selecting the ten best inhibitors
and eight poorest inhibitors against each S. suis adhesin for
refined evaluations in triplicate (Table 2). The refined evaluation
established the IC50 values of the two best inhibitors, the C3ꢀ-
phenylurea 44 and the O2ꢀ-methoxymethyl 51, against the type
PN adhesin to be 30 and 50 nM, respectively, which is up to
one order of magnitude better than the parent unsubstituted
p-methoxyphenyl galabioside 45 (IC50 310 nM) and significantly
better than the previously reported best small-molecule inhibitor
against this adhesin, the natural globotriose trisaccharide (IC50
190 nM).16 The high affinities of 44 and 51 are extraordinary
within the field of small-molecule inhibition of lectins. The
synthesis of further galabiose collections modified with O-alkyl
or alkoxymethyl substituents at O2ꢀ or with ureas at C3ꢀ thus
emerges as an attractive route towards improved inhibitors. An
obvious extension of this result would also be to combine the
substituents of 44 and 51 into one single novel inhibitor, which
would be significantly more potent provided that the affinity-
0.08–0.31
0.02–0.04
0.08–0.31
0.63
62.5
0.31
0.63
31.3
0.31
0.63
31.3
50 0.18
51 0.05
56 6.5
57 10.4
58 15.6
61 7.8
64 10.4
0.08–0.31
0.04–0.08
3.9–7.8
7.8–15.6
15.6
7.8
7.8–15.6
enhancing effects of each substituent are additive. Furthermore,
displaying 44 or 51, or a combination of these two inhibitors, on
a multivalent scaffold would most likely result in more powerful
inhibitors as multivalent inhibitors are known to be particularly
efficient against the S. suis adhesins.20
The results with the type PO adhesin were less impressive as
only marginal affinity enhancements, compared to the known
references 46–47, were obtained. Clearly, other strategies have
to be considered for the development of inhibitors against this
adhesin. From a drug development perspective, it would of
course be desirable to find one inhibitor with high affinity
against both S. suis adhesins. However, this appears to be a
formidable challenge in light of the results reported herein.
Possibly, chemical modifications at positions other than C1 and
C3ꢀ of the galabiose disaccharide will be required to find an
efficient inhibitor against both adhesins.
Structure–activity relationships for the E. coli class II PapG
adhesin and the adhesin from S. suis type PO
In order to develop a quantitative structure–activity relation-
ship using multivariate data analysis, the four carbohydrate
collections were characterised with molecular descriptors using
the MOE software.31 The 2D molecular descriptors included
described properties such as size, lipophilicity, flexibility and
hydrogen-bonding capabilities (see Table 3). The affinity of the
carbohydrates as measured by log Kd and log IC50 for the E. coli
adhesin PapGII and the S. suis type PO adhesin, respectively,
were related to the various molecular descriptors by means of
partial least-squares projections to latent structures (PLS).32
A PLS model was calculated including compounds 5, 8, 10–
17, 19–20, 22–24, 26–27, 33–35, 39–48, 51 and 53–54 using all
descriptors and log Kd for the PapGII adhesin as the response.
After variable selection leaving 34 molecular descriptors as
predictor variables (X), the model explained 78% (R2Y = 0.78)
of the total variation in the response data (Y) and was able to
predict 68% (Q2 = 0.68) of the response variation according to
cross-validation. The predictive properties of the model were
8 9 4
O r g . B i o m o l . C h e m . , 2 0 0 5 , 3 , 8 8 6 – 9 0 0

View Article Online
Table 3 List of structural descriptors used for characterisation of galabiosides and regression coefficients for important factors in local PLS models
for galabioside substituents in C1 and C3ꢀ when biologically evaluated against Escherichia coli adhesin PapG type II and Streptococcus suis adhesin
type PO
S. suis type PO
C1
C3ꢀ
E. coli PapG
C1
C3ꢀ
Abbreviation
Descriptors
1
2
3
4
5
6
7
diameter
radius
VDistEq
VDistMa
weinerPath
weinerPol
a_aro
Molecular diameter
Molecular radius
Vertex distance equation
Vertex distance magnitude
Weiner path number
Weiner polarity number
Number of aromatic atoms
0.030
0.016
0.013
−0.029
0.052
0.053
8
b_ar
Number of aromatic bonds
9
b_rotN
b_rotR
chi0v
chi0v_C
chi1v
chi1v_C
Weight
chi0
chi0_C
chi1
Number of rotatable bonds
Fraction of rotatable bonds
Atomic valence connectivity index
Carbon valence connectivity index
Atomic valence connectivity index
Carbon valence connectivity index
Molecular weight
Atomic connectivity index
Carbon connectivity index
Atomic connectivity index
Carbon connectivity index
Sum of formal charges
Vertex adjacency equation
Vertex adjacency magnitude
Zagreb index
Balaban connectivity index
Total positive partial charge
Total negative partial charge
Relative positive partial charge
Relative negative partial charge
Fractional hydrophobic van der Waals surface area
Fractional negative van der Waals surface area
Fractional polar negative van der Waals surface area
Fractional polar van der Waals surface area
Fractional positive van der Waals surface area
Fractional polar positive van der Waals surface area
Total hydrophobic van der Waals surface area
Total negative van der Waals surface area
Total polar negative van der Waals surface area
Total polar van der Waals surface area
Total positive van der Waals surface area
Total polar positive van der Waals surface area
Kappa shape index
Kappa shape index
Kappa shape index
Alpha modified shape index
Alpha modified shape index
Alpha modified shape index
Flexibility index
Atomic polarizabilities
Atomic polarizabilities
−0.058
−0.050
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
0.082
−0.061
0.092
−0.033
−0.033
chi1_C
0.053
FCharge
VAdjEq
VAdjMa
zagreb
balabanJ
Q_PC+
Q_PC-
Q_RPC+
Q_RPC-
Q_VSA_FHYD
Q_VSA_FNEG
Q_VSA_FPNE
Q_VSA_FPOL
Q_VSA_FPOS
Q_VSA_FPPO
Q_VSA_HYD
Q_VSA_NEG
Q_VSA_PNEG
Q_VSA_POL
Q_VSA_POS
Q_VSA_PPOS
Kier1
Kier2
Kier3
KierA1
KierA2
KierA3
KierFlex
apol
bpol
mr
a_acc
a_acid
a_base
a_don
a_hyd
vsa_acc
vsa_acid
vsa_base
vsa_don
vsa_hyd
vsa_other
vsa_pol
SlogP
SMR
TPSA
density
0.045
−0.078
0.058
−0.062
0.029
−0.029
−0.030
0.029
−0.010
−0.059
0.021
−0.008
0.008
0.062
−0.002
−0.008
0.064
0.008
−0.062
−0.071
−0.042
−0.016
−0.096
−0.038
−0.104
0.039
−0.052
−0.099
−0,039
−0.043
−0.018
−0.188
−0.200
−0.054
−0.052
−0.018
0.007
Molecular refractivity
Number of hydrogen bond acceptors
Number of acidic atoms
−0.051
0.116
−0.008
0.044
0.058
−0.074
Number of basic atoms
Number of hydrogen bond donors
Number of hydrophobic atoms
van der Waals surface areas of hydrogen bond acceptors
van der Waals surface areas of acidic atoms
van der Waals surface areas of basic atoms
van der Waals surface areas of hydrogen bond donors
van der Waals surface areas of hydrophobic atoms
van der Waals Surface Areas of other atoms
van der Waals surface areas of polar atoms
Log of the octanol/water partition coefficient
Molecular refractivity
−0.109
−0.044
−0.356
−0.140
−0.197
−0.006
−0.008
−0.074
−0.105
−0.058
0.109
−0.171
−0.217
0.008
−0.140
−0.174
0.025
−0.210
0.100
0.116
Total polar surface area
Molecular mass density
van der Waals surface area
van der Waals volume
−0.080
−0.029
0.074
−0.052
−0.050
−0.049
−0.096
vdw_area
vdw_vol
logP(o/w)
Log of the octanol/water partition coefficient
0.009
0.084
0.085
O r g . B i o m o l . C h e m . , 2 0 0 5 , 3 , 8 8 6 – 9 0 0
8 9 5

View Article Online
further validated using an independent test set which included
three carbohydrates from collection I (6, 7, 9), three from
collection II (18, 21, 23), one from collection III (25), three
from collection IV (36–38) and one C2ꢀ substituted compound
50 (Fig. 2a). The model was able to predict the affinity of the
test set compounds in an excellent way with a root mean square
prediction error (RMSEP) of 0.49. Only one compound (23)
was poorly predicted by the model. It was predicted to have a
rather low affinity with a Kd of 350 lM, while the experimentally
determined value was 150 lM.
the C2ꢀ position were too few (48, 50–51) to be able to create a
representative model. Variables not related to the response were
removed by means of filtering. Two models with 20 and 23 im-
portant factors were retrieved for the anomeric and the C3ꢀ posi-
tion, respectively (see Table 3). For the anomeric position it could
be further verified that aromatic substituents are important for
affinity to the PapGII adhesin since regression coefficients for
variables describing aromaticity and lipophilicity (a_aro, b_ar,
SlogP and logP(o/w)) were positively correlated with the re-
sponse. Coefficients related to flexibility (KierFlex, b_rotR) were
negatively correlated indicating that groups with a high degree of
freedom are unfavourable for binding (cf. 9 and 46). In addition,
it could be seen that the presence of either hydrogen bond
donors or acceptors on the anomeric substituent was strongly
correlated with the affinity. This could be seen in the poor
affinity for inhibitors containing an amide functionality with
hydrogen bond donation capacities adjacent to C1 (collection I)
in comparison with the relatively high affinity for inhibitors with
only hydrogen bond accepting properties at the same position
(collection II and III). For the C3ꢀ position it could be confirmed
that replacing the ether functionality with a hydrogen bond do-
nating amide is detrimental to binding ability since the presence
of hydrogen bond acceptors was positively correlated and the
presence of donors was negatively correlated with the response.
Furthermore, positively correlated coefficients for flexibility
indicate that a higher degree of freedom might be necessary
in order to achieve the correct positioning of the substituent.
A schematic summary of the structure–activity relationships
for the PapGII adhesin is shown in Fig. 3a. The shallow pocket
formed by Arg170, Trp107, and Asp108, which is seen in the
To further evaluate two of the positions that were varied, C1
and C3ꢀ, two local PLS-models were created using compounds
5–27 and 45–47 for the anomeric position and 33–44 and 53–54
for the C3ꢀ position. The number of substances with variation in
Fig. 2 Calculated response values for galabiosides (ꢀ) using the two
different QSAR versus the experimental values for a) the binding affinity
to E. coli adhesin PapG type II expressed as −log Kd (R2Y = 0.78, Q2 =
0.68) and b) the inhibition of S. suis adhesin type PO expressed as −log
IC50 (R2Y = 0.89, Q2 = 0.75). Both models were validated with an
independent set of diverse galabiosides (♦); for chemical structures see
Table 1.
Fig. 3 a) Graphic summary of the structure–activity relationship of
galabioses 5, 8, 10–17, 19–20, 22–24, 26–27, 33–35, 39–48, 51 and
53–54 in binding to the PapGII adhesin. Amino acids shown are in the
proximity of the galabiose substituents according to the crystal structure
of the adhesin–globotetraose complex.19 b) Graphic summary of the
structure–activity relationship of galabiose inhibitors 5–9, 11, 14–17,
19, 21–26, 34–36, 38–48, 50–53, 56 and 59 of the S. suis PO adhesin.
8 9 6
O r g . B i o m o l . C h e m . , 2 0 0 5 , 3 , 8 8 6 – 9 0 0

View Article Online
crystal structure of PapGII together with a tetrasaccharide,19
could explain the increase in affinity provided by C1 substituents
with low flexibility. The preference for aromatic groups in
the same pocket could derive from P-stacking or cation-P
interactions from Trp107 and Arg170. The hydrogen bonding
properties seen at both O1 and O3ꢀ indicate the presence
of important hydrogen bonds from Lys172 to O3ꢀ, as seen
in the crystal structure, and from either residue Arg170 or
Trp107 to the neighbouring O1. Flexibility in inhibitors is not
normally beneficial for entropic reasons and the prediction by
the model that the inhibitors should have flexible substituents at
C3ꢀ probably reflects that the geometric requirements on rigid
substituents are higher, as rigid substituents are less adaptable to
the steric requirements of the protein binding site. Presumably,
a rigid substituent properly designed to sterically match the
binding site of the PapGII adhesin would improve the affinity.
Relating the aforementioned molecular descriptors for com-
pounds 5–9, 11, 14–17, 19, 21–26, 34–36, 38–48, 50–53, 56
and 59 to IC50 values (from the refined measurements in
Table 2 when applicable) for S. suis adhesins PO and PN by
using PLS gave, after variable selection, a prediction model for
inhibition of the adhesin type PO with 38 variables describing
89% of the total variation in the response and Q2 = 0.75. The
predictive properties of the model were further validated using
an independent test set, which included one compound from
collection I (10), four from collection II (12–13, 18, 20), one from
collection III (27), two from collection IV (33 and 37), one C2ꢀ
substituted compound (49) and two known galabiosides (54 and
55), Fig. 2b. The prediction of the test set gave good results, with
the exception of compound 33, with an RMSEP of 0.66. The
galabioside 33, however, is the only basic amine in the collection
tested, which could explain the prediction difficulties for this
compound. Excluding that object from the test set gave excellent
prediction results, RMSEP = 0.44. No significant model for the
adhesin type PN could be retrieved, maybe due to insufficient
variation in the response. The galabiosides 57–58 and 60–64
were excluded, since the long and flexible side-chains in both of
the positions varied made them difficult to model with the 2D
descriptors that were used.
and the S. suis adhesin type PO in an excellent fashion. In
addition, local models for each position varied and provided
quantitative structure–activity relationships for both adhesins.
These relationships may be used to optimise the substituents
further and constitute a base for future designed libraries where
all positions will be varied at the same time in order to reveal
interaction effects between the different substituents.
Conclusions
Four collections of galabiosides derivatised at C1 and C3ꢀ have
been synthesised aiming at enhanced affinity, compared to the
parent galabioside disaccharide, for the E. coli class II PapG
adhesin and the type PN and type PO adhesins from S. suis. The
present study clearly shows that the E. coli class II PapG adhesin
recognizes m- or p-methoxyphenyl galabiosides 11 and 45
with high affinity and specificity. Although various approaches
towards galabiose derivatives carrying aromatic structures at C1
were evaluated (i.e. replacing the anomeric oxygen with an amido
group or a sulfur atom or modifying the aromatic structure)
no improvements in affinity for the adhesin were revealed.
Furthermore, structural modifications at C3ꢀ did not provide any
significant affinity enhancements for the PapGII adhesin. Hence,
the m- or p-methoxyphenyl galabiosides 11 and 45 constitute
lead structures in future work towards agents targeting urinary
tract infections. Remarkably potent inhibitors were discovered
against one of the S. suis adhesins (type PN, IC50 down to 30 nM),
which provide an ideal starting point for the further development
of mono-, as well as multivalent, inhibitors of this adhesin. The
PapGII preferentially bound to substituted phenyl galabiosides,
while the PO preferentially bound p-methoxyphenyl galabiosides
substituted with hydrogen bond accepting alkyl groups at O2ꢀ
or phenyl ureas at C3ꢀ. PLS models could accurately predict
affinity-enhancing effects of substituents in binding to the E.
coli PapGII and S. suis type PO adhesins, which is of value for
the design of further improved inhibitors.
Experimental
Synthesis
Local PLS models on galabiosides 5–27, 45–47 and 33–44,
52–56 for the anomeric and C3ꢀ positions, respectively, resulted
in two models with 22 important variables. The model for the
anomeric position could further verify that substituents with
low flexibility and high lipophilicity such as aromatic rings
are beneficial for affinity. Both hydrogen bond accepting and
donating capabilities are negatively correlated to the response,
donors to a larger extent. The negatively correlated term for
hydrogen bond donors is probably related to the C1 linkage
position since good activity can be observed in all cases when
donor substituents are present elsewhere (16, 23). The opposite
can be said of the negatively correlated term for hydrogen
bond acceptors where it seems that acceptors on positions
other than the C1 linkage are unfavourable, as can be seen
in the galabiosides with methoxy substituted aromatic rings
(e.g. 11, 12). The C3ꢀ position model clarifies the importance
of having the ether linkage intact. Amides with hydrogen bond
donor coefficients are strongly negatively correlated, whereas
hydrogen bond acceptor terms are positively correlated with
affinity (cf. 52–55 with collection IV). A schematic summary
of the structure–activity relationships is shown in Fig. 3b. This
proposes that a good inhibitor of the S. suis adhesin type PO
should be a galabioside with large aromatic and highly lipophilic
aglycons, e.g. naphthyl galabiosides. In addition, a highly flexible
group should be attached to a hydrogen bond acceptor at C3ꢀ.
However, the prediction that a flexible group at C3ꢀ is beneficial
could be misleading, as discussed for the PapGII adhesin above,
and a rigid substituent properly designed to sterically match the
binding site of the PO adhesin could improve the affinity.
General methods. All non-aqueous reactions were run in
septum-capped, oven-dried flasks under Ar (1 atm). CH2Cl2 was
dried by distillation from CaH2 and Et2O was distilled from Na.
Organic solutions were concentrated by rotary evaporation with
a bath temperature at or below 40 ◦C. Flash chromatography
was performed on Grace Amicon Silica gel 60 (35–70 lm) and
TLC was performed on Kieselgel 60 F254 plates (Merck). C18
cartridges were from IST Ltd, UK. NMR spectra were recorded
with Bruker DRX-400 or ARX-300 instruments. Residual
CHCl3 or CD2HOD were used as internal references at 7.27
and 3.31 ppm, respectively. 1H-NMR spectral assignments were
made based on COSY spectra. J values are given in Hz.
2,3,4,6-Tetra-O-acetyl-a-D-galactopyranosyl-(1→4)-2,3,6-tri-
O-benzoyl-b-D-galactopyranosyl azide (3). To a solution of
2,3,4,6-tetra-O-acetyl-a-D-galactopyranosyl-(1→4)-2,3,6-tri-O-
benzoyl-a-D-galactopyranosyl bromide 122 (2.58 g, 2.96 mmol)
in THF (100 mL) were added trimethylsilyl azide (0.55 mL,
4.15 mmol) and tetrabutylammonium fluoride (3.55 mL,
3.55 mmol, 1 M in THF) and the mixture was stirred for 24 h.
The mixture was filtered through a silica column (2 : 1→2 : 3,
heptane–EtOAc gradient) to give 3 (2.31 g, 92%); [a]2D3 +91 (c
1.0 in CDCl3); dH(300 MHz; CDCl3) 8.10–7.95 (m, 6H, Ar–H),
7.65–7.36 (m, 9H, Ar–H), 5.71 (dd, 1H, J 10.6, 12.5, H-2), 5.50
(m, 2H, H-3ꢀ, H-4ꢀ), 5.42 (dd, 1H, J 2.8, 10.5, H-3), 5.30–5.23
(m, 2H, H-1ꢀ, H-2ꢀ), 4.92 (d, 1H, J 8.6, H-1), 4.77 (dd, 1H, J
6.9, 11.5, H-6), 4.58–4.51 (m, 2H, H-5ꢀ, H-6), 4.49 (d, 1H, J
2.6, H-4), 4.23 (t, 1H, J 6.6, H-5), 3.83 (dd, 1H, J 7.5, 11.0,
H-6ꢀ), 3.69 (dd, 1H, J 6.4, 11.0, H-ꢀ), 2.18, 2.11, 2.04, 1.87 (s,
3H each, OAc); dC(75 MHz; CDCl3) 170.6, 170.2, 170.1, 169.8,
Two PLS models were obtained with the ability to predict
the affinity of new galabiosides for the E. coli adhesin PapGII
O r g . B i o m o l . C h e m . , 2 0 0 5 , 3 , 8 8 6 – 9 0 0
8 9 7

View Article Online
166.0, 165.9, 165.0, 133.8, 133.6, 133.5, 129.82, 129.75, 129.1,
128.8, 128.7, 128.6, 128.4, 98.0, 88.5, 74.8, 74.6, 73.1, 68.7, 68.2,
67.7, 67.35, 67.25, 62.1, 60.8, 20.8, 20.7, 20.6, 20.5; m/z (FAB)
870.2336 (M+ + Na. C41H41N3O17Na requires 870.2334).
and flash chromatographed (SiO2, 2 : 1→2 : 3, heptane–EtOAc
gradient) to give 29 (1.19 g, 86%); [a]2D3 −42 (c 0.9 in CHCl3);
dH(300 MHz; CDCl3); 7.48 (m, 2H, Ar–H), 7.37–7.28 (m, 13H,
Ar–H), 4.97/4.78 (ABq, 2H, J 9.9, CH2), 4.88/4.60 (ABq, 2H,
J 11.4, CH2), 4.47/4.44 (ABq, 2H, J 11.7, CH2), 4.37 (d, 1H, J
9.4, H-1), 3.93 (d, 1H, J 2.7, H-4), 3.80 (t, 1H, J 9.6, H-2), 3.60
(m, 3H, H-5, 2 × H-6), 3.54 (dd, 1H, J 3.0, 9.7, H-3), 2.25 (s, 3H,
CH3); dC(75 MHz; CDCl3) 138.08, 138.07, 137.7, 137.4, 128.69,
128.67, 128.64, 128.57, 128.49, 128.46, 128.44, 128.40, 128.35,
128.29, 128.26, 128.2, 128.1, 128.04, 128.01, 127.98, 127.89,
127.87, 127.86, 127.70, 127.68, 85.9, 76.6, 75.2, 75.09, 75.06,
73.53, 73.52, 68.2, 67.0, 12.8; m/z (FAB) 528.1940 (M+ + Na.
C28H31N3O4SNa requires 528.1933).
a-D-Galactopyranosyl-(1→4)-b-D-galactopyranosyl azide (4).
To a solution of 3 (2.00 g, 2.36 mmol) in MeOH (150 mL) was
added NaOMe (0.05 mL, 0.05 mmol, 1 M in MeOH) and the
mixture was stirred for 18 h then 10% methanolic acetic acid was
added until a neutral reaction was obtained on moist pH-paper.
Concentration and flash chromatography (SiO2, 66 : 33 : 4→50 :
50 : 4, CH2Cl2–MeOH–H2O gradient) gave 4 (832 mg, 96%); [a]D23
+72 (c 1.0 in MeOH); dH(300 MHz; CD3OD); 4.99 (d, 1H, J 3.6,
H-1ꢀ), 4.56 (d, 1H, J 8.3, H-1), 4.22 (m, 1H, H-5ꢀ), 4.03 (d, J 2.4,
H-4ꢀ), 3.93 (dd, 1H, J 1.0, 2.9, H-4), 3.87–3.71 (m, 7H), 3.57 (dd,
1H, J 3.0, 10.0), 3.45 (dd, 1H, J 8.3, 10.0); dC(75 MHz; CD3OD)
103.5, 93.5, 80.5, 78.9, 75.5, 73.8, 73.2, 72.0, 71.8, 71.3, 63.5,
62.0; m/z (FAB) 390.1127 (M+ + Na. C12H21N3O10Na requires
390.1125).
4-Methoxyphenyl (3-azido-3-deoxy-2,4,6-tri-O-benzyl-a-D-
galactopyranosyl)-(1→4)-2,3,6-tri-O-benzoyl-b-D-galactopyra-
noside (31). To a mixture of 29 (55 mg, 0.11 mmol), 3022
(50 mg, 84 lmol) and N-iodosuccinimide (48 mg, 0.22 mmol)
were added CH2Cl2 (1.0 mL) and Et2O (2.0 mL) and the solution
was cooled to −50 ◦C. TMSOTf (3 lL, 17 lmol) was added
and the mixture was stirred for 1 h. Triethylamine (0.5 mL) was
added and the mixture was stirred for another hour at −50 ◦C,
then allowed to reach ambient temperature, diluted with
CH2Cl2 (20 mL), washed with 10% Na2S2O3(aq.) (10 mL) and
sat’d NaHCO3(aq.) (10 mL), dried (MgSO4), and concentrated.
Flash chromatography (SiO2, 3 : 1 heptane–EtOAc) gave 31
(82 mg, 93%); [a]2D3 +75 (c 1.0 in CHCl3); dH(300 MHz; CDCl3);
8.07–7.95 (m, 6H, Ar–H), 7.66–7.20 (m, 24H, Ar–H), 6.97 (m,
2H, OPhOMe), 6.69 (m, 2H, OPhOMe), 6.00 (dd, 1H, J 7.7,
10.5, H-2), 5.29 (dd, 1H, J 2.8, 10.5, H-3), 5.17 (d, 1H, J 7.7,
H-1), 4.97 (d, 1H, J 3.2, H-1ꢀ), 4.86–4.69 (m, 5H, 2 × H-6),
4.55–4.43 (m, 3H, H-4, H-5ꢀ), 4.22–4.15 (m, 4H, H-3ꢀ, H-5),
4.07 (m, 1H, H-4ꢀ), 4.04 (dd, 1H, J 3.2, 10.7, H-2ꢀ), 3.74 (s, 3H,
OMe), 3.44 (t, 1H, J 8.7, H-6ꢀ), 3.10 (dd, 1H, J 4.8, 8.5, H-6ꢀ);
dC(100 MHz; CDCl3) 166.9, 166.5, 165.8, 156.0, 151.6, 138.64,
138.62, 137.8, 133.9, 133.71, 133.69, 130.4, 130.20, 130.17,
130.1, 129.9, 129.4, 128.94, 128.91, 128.86, 128.78, 128.76,
128.7, 128.6, 128.4, 128.1, 128.0, 119.2, 114.8, 101.3, 100.3,
76.4, 76.2, 75.9, 74.6, 74.0, 73.4, 69.91, 69.86, 67.7, 63.2, 61.7,
56.0; m/z (FAB) 1078.3743 (M+ + Na. C61H57N3O14Na requires
1078.3738).
General procedure for the synthesis of compounds 5–10.
A
solution of 4 (15 mg, 0.041 mmol) in MeOH (0.7 mL) was
hydrogenated (H2, 1 atm, Pd/C 10%, cat) for 30 min. Na2CO3
(30 mg, 0.27 mmol) was added, followed by dropwise addition of
a solution of the acid chloride or anhydride (0.41 mmol, 10 eq.) in
THF (0.7 mL). After 30 min, the solution was filtered through
Celite, concentrated and dissolved in water. The solution was
applied to a C-18 cartridge (2 g), which was washed with water
and eluted with MeOH–H2O (2 : 1). Concentration and flash
1
chromatography (SiO2, CH2Cl2–MeOH–H2O) gave 5–10. H-
NMR and FAB-HRMS data are listed in Table 4.
General procedure for the synthesis of compounds 11–19. To
a solution of 222 (40 mg, 42 lmol) in dry CH2Cl2 (1.2 mL) at
◦
−20 C was added the alcohol (84 lmol, 2 eq.) and TMSOTf
(2 lL, 13 lmol). The reaction mixture was stirred for 30 min and
Et3N (0.1 mL) was added. The mixture was concentrated and
flash chromatographed (SiO2, heptane–EtOAc). Deacylation
(2.0 mL, 0.01 M NaOMe in MeOH) overnight, followed by
addition of 10% methanolic acetic acid until a neutral reaction
was obtained on moist pH-paper, concentration, and flash
chromatography (SiO2, CH2Cl2–MeOH) gave 11–19. 1H-NMR
and FAB-HRMS are listed in Table 4.
4-Methoxyphenyl (3-azido-3-deoxy-2,4,6-tri-O-benzyl-a-D-
galactopyranosyl)-(1→4)-b-D-galactopyranoside (32). To
a
General procedure for the synthesis of compounds 20–22 and
24–27. To a solution of the alcohol (69 lmol, 1.5 eq.) in
DMF (0.5 mL) was added NaH (2.9 mg, 74 lmol, 60% in
mineral oil) and the mixture was stirred for 15 min. The
resulting solution was added dropwise to a solution of 1 (40 mg,
46 lmol) in DMF. After 45 min, the reaction mixture was
diluted with CH2Cl2 (5 mL), washed with sat’d NaHCO3(aq.)
(2 mL), dried (Na2SO4), and flash chromatographed (SiO2,
heptane–EtOAc). Deacylation (2.0 mL, 0.01 M NaOMe in
MeOH) overnight was followed by addition of 10% methanolic
acetic acid until a neutral reaction was obtained on moist pH-
paper. Concentration and flash chromatography (SiO2, CH2Cl2–
MeOH) gave 20–22 and 24–27. 1H-NMR and FAB-HRMS are
listed in Table 4.
solution of 31 (318 mg, 0.43 mmol) in MeOH (10 mL) was
added NaOMe (0.10 mL, 1 M in MeOH) and the reaction
mixture was stirred overnight, then 10% methanolic acetic
acid was added until a neutral reaction was obtained on moist
pH-paper. Concentration and flash chromatography (SiO2,
toluene–acetone, 4 : 1) gave 32 (201 mg, 89%); [a]2D5 −8 (c 1.0 in
CHCl3); dH(400 MHz; CDCl3); 7.40–7.31 (m, 15H, Ar–H), 6.99
(m, 2H, OPhOMe), 6.81 (m, 2H, OPhOMe), 4.89/4.64 (ABq,
2H, J 11.6, CH2), 4.89/4.53 (ABq, 2H, J 11.4, CH2), 4.89 (m,
1H, H-1ꢀ), 4.71 (d, 1H, J 7.5, H-1), 4.46/4.40 (ABq, 2H, J 11.4,
CH2), 4.20 (m, 1H, H-5ꢀ), 4.04 (dd, 1H, J 3.4, 10.6, H-2ꢀ), 3.98
(dd, 1H, J 2.8, 10.6, H-3ꢀ), 3.94 (m, 1H, H-4), 3.88 (d, 1H, J
11.9, OH), 3.83 (m, 1H, H-4ꢀ), 3.77 (m, 5H, 2 × H-6, OMe),
3.70 (m, 2H, H-2, H-5), 3.53 (t, 1H, J 9.5, H-6ꢀ), 3.45 (m, 1H,
H-3), 3.25 (dd, 1H, J 3.7, 9.6, H-6ꢀ), 2.94 (t, 1H, J 6.7, OH),
2.15 (s, 1H, OH); dC(100 MHz; CDCl3) 155.8, 151.6, 137.79,
137.78, 136.9, 129.3, 129.2, 129.0, 128.91, 128.87, 128.64,
128.61, 128.57, 119.0, 114.9, 102.8, 100.0, 81.0, 76.07, 76.05,
75.7, 74.9, 74.5, 74.3, 74.2, 72.3, 71.6, 69.7, 62.1, 61.0, 56.1; m/z
(FAB) 766.2956 (M+ + Na. C40H45N3O11Na requires 766.2952).
Methyl 3-azido-3-deoxy-2,4,6-tri-O-benzyl-1-thio-b-D-galac-
topyranoside (29). To a solution of 2825 (990 mg, 2.74 mmol) in
MeOH (50 mL) was added NaOMe (0.1 mL, 1 M in methanol)
and the solution was stirred overnight. Methanolic acetic acid
(10%) was added until a neutral reaction was obtained on
moist pH-paper and the mixture was concentrated. The residue
was dissolved in DMF (10 mL), NaH (390 mg, 9.9 mmol,
60% in mineral oil) was added, and the mixture was stirred
for 15 min, then cooled to 0 ◦C. Benzyl bromide (1.2 mL,
9.9 mmol) was added dropwise and the mixture was allowed
to reach ambient temperature overnight. MeOH (5 mL) was
added, the mixture was diluted with CH2Cl2 (100 mL), washed
with sat’d NaHCO3(aq.) (50 mL), dried (MgSO4), concentrated
4-Methoxyphenyl (3-amino-3-deoxy-a-D-galactopyranosyl)-
(1→4)-b-D-galactopyranoside (33). A solution of 32 (30 mg,
40 lmol), HCl(aq.) (5 lL, conc.), and Pd/C (10%, 10 mg) in
MeOH (0.7 mL) was hydrogenated (H2, 1 atm) for 30 min. The
solution was filtered through Celite, concentrated, re-dissolved
in H2O, and applied to a C-18 cartridge (2 g). The cartridge
8 9 8
O r g . B i o m o l . C h e m . , 2 0 0 5 , 3 , 8 8 6 – 9 0 0

View Article Online
Table 4 1H NMR and FAB-HRMS data of compounds 5–27 and 33–44
dH(300 MHz; CD3OD)
m/z (FAB) (M+ + Na) required/found
5
6
7.91 (m, 2H, Ar), 7.58–7.45 (m, 3H, Ar), 5.13 (d, 1H, J 8.9, H-1), 5.03 (d, 1H, J 3.6, H-1ꢀ),
4.31 (t, 1H, J 6.2, H-5ꢀ)
468.1482/468.1482
498.1587/498.1583
7.89 (m, 2H, Ar), 7.00 (m, 2H, Ar), 5.11 (d, 1H, J 8.8, H-1), 5.03 (d, 1H, J 3.4, H-1ꢀ), 4.31 (t,
1H, J 6.8, H-5ꢀ)
7
8
7.55 (m, 1H, Ar), 5.08 (ma, 1H, H-1), 5.02 (d, 1H, J 3.7, H-1ꢀ), 4.23 (t, 1H, J 6.0, H-5ꢀ)
7.53 (m, 2H, Ar), 7.18 (m, 1H, Ar), 5.10 (d, 1H, J 8.9, H-1), 5.03 (d, 1H, J 3.8, H-1ꢀ), 4.29 (t,
1H, J 6.8, H-5ꢀ)
540.1105/540.1107
504.1293/504.1300
9
5.01 (d, 1H, J 3.8, H-1ꢀ), 4.88 (d, 1H, J 8.3, H-1), 4.26 (t, 1H, J 5.6, H-5ꢀ), 2.29 (m, 2H, CH2),
1.14 (t, 3H, J 7.6, CH3)
420.1481/420.1501
528.1693/528.1697
471.1478/471.1471
471.1478/471.1482
441.1373/441.1359
455.1529/455.1520
491.1529/491.1525
480.1482/480.1490
455.1529/455.1526
447.1842/447.1849
10
11
12
13
14
15
16
17
18
8.01 (m, 1H, Ar), 6.67 (m, 2H, Ar), 5.12 (ma, 1H, H-1), 5.02 (d, 1H, J 2.6, H-1ꢀ), 4.34 (t, 1H,
J 6.7, H-5ꢀ)
7.17 (m, 1H, Ar), 6.69 (m, 2H, Ar), 6.60 (m, 1H, Ar), 5.02 (ma, H-1ꢀ), 4.93 (d, 1H, J 7.4,
H-1), 4.33 (t, 1H, J 5.9, H-5ꢀ)
7.16 (m, 1H, Ar), 7.02 (m, 2H, Ar), 6.90 (m, 1H, Ar), 5.02 (d, J 1.7, H-1ꢀ), 4.93 (d, 1H, J 7.6,
H-1), 4.34 (t, 1H, J 6.4, H-5ꢀ)
7.29 (m, 2H, Ar), 7.10 (m, 2H, Ar), 7.02 (m, 1H, Ar), 5.01 (ma, 1H, H-1ꢀ), 4.96 (d, 1H, J 7.4,
H-1), 4.33 (t, 1H, J 5.8, H-5ꢀ)
7.08 (m, 2H, Ar), 6.99 (m, 2H, Ar), 5.01 (ma, 1H, H-1ꢀ), 4.90 (d, 1H, J 7.5, H-1), 4.32 (t, 1H,
J 5.8, H-5ꢀ)
7.78 (m, 3H, Ar), 7.49–7.28 (m, 4H, Ar), 5.12 (d, 1H, J 7.5, H-1), 5.03a (m, 1H, H-1ꢀ), 4.34 (t,
1H, J 5.6, H-5ꢀ)
7.34–7.21 (m, 3H, Ar), 6.97 (m, 1H, Ar), 6.37 (m, 1H, Ar), 5.02 (d, 1H, J 2.8, H-1ꢀ), 4.87 (d,
1H, J 7.5, H-1), 4.36 (t, 1H, J 6.6, H-5ꢀ)
7.45–7.26 (5, 2H, Ar), 4.97 (d, 1H, J 1.7, H-1ꢀ), 4.91 (AB, 1H, J 11.7, CH2), 4.69 (AB, 1H, J
11.7, CH2), 4.40 (ma, 1H, H-1), 4.31 (t, 1H, J 6.5, H-5ꢀ)
4.97 (ma, 1H, H-1ꢀ), 4.40 (d, 1H, J 7.5, H-1), 4.34 (t, 1H, J 6.6, H-5ꢀ), 1.96 (m, 2H, CH2), 1.77
(m, 2H, CH2), 1.55 (m, 1H, CH2), 1.47–1.20 (m, 5H, CH2)
19
20
5.01 (d, 1H, J 2.8, H-1ꢀ), 4.93 (d, 1H, J 7.5, H-1), 4.29 (t, 1H, J 6.4, H-5ꢀ)
7.97 (m, 2H, Ar), 7.16 (m, 2H, Ar), 5.06 (d, 1H, J 7.5, H-1), 5.01 (d, 1H, J 3.0, H-1ꢀ), 4.31 (t,
1H, J 5.9, H-5ꢀ)
531.0901/531.0901
499.1428/499.1432
21
22
23
7.78 (m, 1H, Ar), 7.54 (m, 1H, Ar), 7.37 (m, 1H, Ar), 7.14 (m, 1H, Ar), 5.01 (d, 1H, J 1.8,
499.1428/499.1426
486.1224/486.1222
498.1587/498.1584
H-1ꢀ), 4.92 (d, 1H, J 7.5, H-1), 4.32 (t, 1H, J 6.6, H-5ꢀ)
8.22 (m, 2H, Ar), 7.25 (m, 2H, Ar), 5.11 (d, 1H, J 7.5, H-1), 5.01 (d, 1H, J 3.4, H-1ꢀ), 4.30 (t,
1H, J 6.1, H-5ꢀ)
7.45 (m, 2H, Ar), 7.06 (m, 2H, Ar), 5.01 (d, 1H, J 1.8, H-1ꢀ), 4.90 (d, 1H, J 7.5, H-1), 4.32 (t,
1H, J 5.9, H-5ꢀ)
24
25
26
27
7.55 (m, 2H, Ar), 6.90 (m, 2H, Ar), 4.87 (d, 1H, J 3.8, H-1ꢀ), 4.38 (d, 1H, J 9.3, H-1)
7.59 (m, 2H, Ar), 7.30 (m, 3H, Ar), 4.93 (d, 1H, J 3.8, H-1ꢀ), 4.58 (d, 1H, J 9.0, H-1)
7.49 (m, 2H, Ar), 7.15 (m, 2H, Ar), 4.90 (d, 1H, J 3.8, H-1ꢀ), 4.49 (d, 1H, J 9.2, H-1)
7.91 (m, 1H, Ar), 7.74 (m, 1H, Ar), 7.59 (m, 1H, Ar), 7.42 (m, 1H, Ar), 4.95 (d, 1H, J 2.9,
H-1ꢀ), 4.89 (d, 1H, J 9.6, H-1), 4.11 (t, 1H, J 6.5)
487.1250/487/1251
457.1144/457.1137
471.1301/471.1294
515.1199/515.1208
33
34
35
36
37
38
39
40
41
5.06 (d, 1H, J 3.6, H-1ꢀ), 4.85 (d, 1H, J 7.1, H-1), 4.43 (t, 1H, J 6.3, H-5ꢀ), 3.53 (dd, 1H, J
1.7, 10.8, H-3ꢀ)
470.1638/470.1642
512.1744/512.1744
526.1900/526.1893
570.1799/570.1786
574.1900/574.1895
604.2006/604.2021
634.2112/634.2101
610.1712/610.1699
619.1751/619.1774
5.03 (d, 1H, J 3.8, H-1ꢀ), 4.83 (d, 1H, J 7.6, H-1), 4.44 (t, 1H, J 6.5, H-5ꢀ), 4.23 (dd, 1H, J
2.9, 11.4, H-3ꢀ), 2.02 (s, 3H, NHAc)
5.08 (d, 1H, J 3.8, H-1ꢀ), 4.86 (d, 1H, J 7.6, H-1), 4.44 (t, 1H, J 6.1, H-5ꢀ), 4.23 (dd, 1H, J
2.9, 11.4, H-3), 2.29 (q, 2H, J 7.7, CH2), 1.15 (t, 3H, J 7.6, CH3)
5.04 (d, 1H, J 3.7, H-1ꢀ), 4.82 (d, 1H, J 7.6, H-1), 4.43 (t, 1H, J 6.8, H-5ꢀ), 4.23 (dd, 1H, J
2.9, 11.4, H-3), 2.60 (m, 4H, J 7.7, CH2)
7.89 (m, 2H, Ar–H), 7.50 (m, 3H, Ar–H), 5.10 (d, 1H, J 3.7, H-1ꢀ), 4.84 (d, 1H, J 7.7, H-1),
4.49 (m, 2H, H-3ꢀ, H-5ꢀ)
7.86 (m, 2H, Ar–H), 6.98 (m, 2H, Ar–H), 5.09 (d, 1H, J 3.7, H-1ꢀ), 4.84 (d, 1H, J 7.6, H-1),
4.51 (t, 1H, J 5.8, H-5ꢀ), 4.45 (dd, 1H, J 2.9, 11.4, H-3ꢀ), 3.85 (s, 3H, OMe)
7.06 (m, 2H, Ar–H), 6.64 (m, 1H, Ar–H), 5.09 (d, 1H, J 3.7, H-1ꢀ), 4.84 (d, 1H, J 7.6, H-1),
4.52 (t, 1H, J 6.1, H-5ꢀ), 4.44 (dd, 1H, J 2.9, 11.4, H-3ꢀ)
7.54 (m, 2H, Ar–H), 7.15 (m, 1H, Ar–H), 5.09 (d, 1H, J 3.8, H-1ꢀ), 4.84 (d, 1H, J 7.6, H-1),
4.52 (t, 1H, J 5.9, H-5ꢀ), 4.43 (dd, 1H, J 2.9, 11.4, H-3ꢀ)
8.78 (m, 1H, Ar–H), 8.40 (m, 1H, Ar–H), 8.28 (m, 1H, Ar–H), 7.73 (m, 1H, Ar–H), 5.10 (d,
1H, J 3.8, H-1ꢀ), 4.84 (d, 1H, J 7.7, H-1), 4.53 (t, 1H, J 6.7, H-5ꢀ), 4.49 (dd, 1H, J 2.9, 11.4,
H-3ꢀ)
42
43
44
7.89 (m, 1H, Ar–H), 7.52 (m, 3H, Ar–H), 5.06 (d, 1H, J 3.7, H-1ꢀ), 4.83 (d, 1H, J 7.6, H-1),
4.50 (t, 1H, J 6.1, H-5ꢀ), 4.38 (dd, 1H, J 2.8, 11.3, H-3ꢀ)
618.1799/618.1791
541.2009/541.2011
589.2009/589.1995
5.02 (d, 1H, J 3.7, H-1ꢀ), 4.81 (d, 1H, J 7.5, H-1), 4.40 (t, 1H, J 5.7, H-5ꢀ), 4.10 (dd, 1H, J
3.0, 11.2, H-3ꢀ), 3.15 (q, 2H, J 7.2, CH2), 1.09 (t, 3H, J 7.2, CH2CH3)
7.35 (m, 2H, Ar–H), 7.23 (m, 2H, Ar–H), 6.95 (m, 1H, Ar–H), 5.06 (d, 1H, J 3.8, H-1ꢀ), 4.83
(d, 1H, J 7.6, H-1), 4.45 (t, 1H, J 5.6, H-5ꢀ), 4.49 (dd, 1H, J 3.0, 11.2, H-3ꢀ)
a Virtual long-range couplings.
was washed with water, eluted with a 10–50% MeOH gradient,
and concentrated to give 33 (18 mg, 93%). 1H-NMR and
FAB-HRMS are listed in Table 4.
Pd/C (10%, 10 mg) in MeOH (0.7 mL) was hydrogenated (H2,
1 atm) for 30 min. Na2CO3 (30 mg, 0.27 mmol) was added,
followed by a solution of an acid chloride, acid anhydride,
or isocyanate (0.41 mmol, 10 eq.) in THF (0.7 mL). After
30 min, the solution was filtered through Celite, concentrated,
dissolved in 10% MeOH (9 : 1), and applied to a C-18 cartridge
General procedure for the synthesis of 34–44. A solution of
32 (30 mg, 40 lmol) in MeOH (0.7 mL) HCl(aq.) (5 lL, conc.), and
O r g . B i o m o l . C h e m . , 2 0 0 5 , 3 , 8 8 6 – 9 0 0
8 9 9

View Article Online
(2 g). The cartridge was washed with water, then eluted with
50% MeOH. Concentration and flash chromatography (SiO2,
CH2Cl2–MeOH–H2O) gave 34–44. 1H-NMR and FAB-HRMS
are listed in Table 4.
Systems” sponsored by the Swedish Foundation for Strategic
Research, and from the Finnish Academy.
References
1 V. Morell, Science, 1997, 278, 575–576.
Surface plasmon resonance experiments
2 R. J. Williams and D. L. Heymann, Science, 1998, 279, 1153–1154.
3 C. Walsh, Nature, 2000, 406, 775–781.
Surface plasmon resonance studies of binding of compounds
5–46 to the class II PapG adhesin were performed as described
earlier.21 The binding data were analysed using the software
Scrubber.28
4 D. Mirelman and I. Ofek, Microbial Lectins and Agglutinins, ed.
D. Mirelman, Wiley, New York, 1986, p. 1–19.
5 K.-A. Karlsson, Curr. Opin. Struct. Biol., 1995, 5, 622–635.
6 H. Leffler and C. Svanborg-Ede´n, FEMS Lett., 1980, 8, 127–134.
7 G. Ka¨llenius, R. Mo¨llby, S. B. Svensson, J. Winberg, A. Lundblad,
S. Svensson and B. Cedergren, FEMS Lett., 1980, 7, 297–302.
8 K. Bock, M. E. Breimer, A. Brignole, G. C. Hansson, K.-A. Karlsson,
G. Larsson, H. Leffler, B. E. Samuelsson, N. Stro¨mberg, C. Svanborg-
Hemagglutination inhibition experiments
◦
Bacteria were grown overnight at 37 C at 5% CO2-incubator.
The bacteria were harvested by centrifugation, 5000 × g, 15 min,
+4 ◦C, and washed twice with PBS. The hemagglutination
activities were titrated and the lowest bacterial densities causing
agglutination were used for the inhibition studies. Twofold
dilutions of the inhibitors were tested and the results were
observed after incubation on ice for two hours. A first screen was
done once with strain 628 (type PN) and strain 836 (type PO).
The hemagglutination type was verified by using galactose and
N-acetylgalactosamine as inhibitors. Ten strong inhibitors and
eight poor inhibitors were selected for refined evaluation in trip-
licate. The results are presented as the average of three individual
determinations and the range of the inhibitory values is shown.
Eden and J. Thurin, J. Biol. Chem., 1985, 260, 8545–8551.
9 N. Stro¨mberg, B.-I. Marklund, B. Lund, D. Ilver, A. Hamers, W.
Gaastra, K.-A. Karlsson and S. Normark, EMBO J., 1990, 9, 2001–
2010.
´
10 N. Stro¨mberg, P.-G. Nyholm, I. Pascher and S. Normark, Proc. Natl.
Acad. Sci. U. S. A., 1991, 88, 9340–9344.
11 I. M. Johansson, K. Plos, B.-I. Marklund and C. Svanborg, Microb.
Pathog., 1993, 15, 121–129.
12 G. Otto, T. Sandberg, B.-I. Marklund, P. Ulleryd and C. Svanborg,
Clin. Infect. Dis., 1993, 17, 448–456.
13 J. R. Johnson, T. A. Ruso, J. J. Brown and A. Stapleton, J. Infect.
Dis., 1998, 177, 97–101.
14 M. Hedlund, M. Svensson, A. Nilsson, R. D. Duan and C. Svanborg,
J. Exp. Med., 1996, 1843, 1037–1044.
15 M. Hedlund, C. Wachtler, E. Johansson, L. Hang, J. E. Somerville,
R. P. Darveeau and C. Svanborg, Mol. Microbiol., 1999, 33, 693–703.
16 S. Haataja, K. Tikkanen, U. Nilsson, G. Magnusson, K.-A. Karlsson
and J. Finne, J. Biol. Chem., 1994, 269, 27466–27472.
17 R. Striker, U. Nilsson, A. Stonecipher, G. Magnusson and S. J.
Hultgren, Mol. Microbiol., 1995, 16, 1021–1029.
18 J. Ohlsson, J. Jass, B. E. Uhlin, J. Kihlberg and U. J. Nilsson,
ChemBioChem, 2002, 3, 772–779.
19 K. W. Dodson, J. S. Pinkner, T. Rose, G. Magnusson, S. J. Hultgren
and G. Waksman, Cell, 2001, 105, 733–743.
20 (a) H. C. Hansen, S. Haataja, J. Finne and G. Magnusson, J. Am.
Chem. Soc., 1997, 119, 6974–6979; (b) during the preparation of this
paper dendritic galabiose compounds with low nanomolar affinity
for Streptococcus suis adhesins were reported:J. A. F. Joosten, V.
Loimaranta, C. C. M. Appeldoorn, S. Haataja, F. A. E. Maate,
R. M. J. Liskamp, J. Finne and R. J. Pieters, J. Med. Chem., 2004,
47, 6499–6508.
21 A. Larsson, J. Ohlsson, K. W. Dodson, S. J. Hultgren, U. J. Nilsson
and J. Kihlberg, Bioorg. Med. Chem., 2003, 11, 2255–2261.
22 J. Ohlsson and G. Magnusson, Carbohydr. Res., 2000, 329, 49–55.
23 E. D. Soli and P. DeShong, J. Org. Chem., 1999, 64, 9724–9726.
24 C. Peto, G. Batta, Z. Gyorgydeak and F. Sztariskai, Liebigs Ann.
Chem., 1991, 505–507.
Computational methods
Characterisation of oligosaccharides. The structures were
generated using the MOE software carbohydrate and molecule
builder interface and energy minimized with the MMFF94
merck and PEF95SAC carbohydrate force fields and an implicit
solvent electrostatic correction model as implemented in MOE.31
Characterisation of the carbohydrates was done by the MOE
software and the molecular descriptors include properties of
size, lipophilicity, polarizability, charge, flexibility, rigidity and
hydrogen-bonding capacities (Table 3).
Data analysis methods. The binding affinities of the
oligosaccharides as measured by −log Kd and −log IC50 for
PapGII and S. suis adhesin type PO respectively were related
to the molecular descriptors by means of partial least squares
projection to latent structures (PLS)32 using the statistical
software Simca.33 The number of significant components was
decided by cross-validation using the default set up.33 Generally,
variable selection was accomplished by excluding all variables
with a variable importance in the projection value (VIP) below 1,
hence keeping the variables inducing an increase in the predictive
power of the model. The VIP value is a weighted sum of squares
of the PLS weights, w, taking into account the amount of
explained Y variance of each PLS dimension.34
The two prediction models were validated using an inde-
pendent test set. The local PLS models were validated using
a permutation test where the order of the response (Y) was
randomly permutated 30 times. By plotting the explanatory
power (R2) and the predictive power (Q2) of the mutated models
as a function of the correlation coefficient between the original
and predicted values, the degree to which these values rely on
chance is reflected by the intercept with the y-axis. A model is
generally considered valid if the intercept is negative for Q2 and
below 0.3 for R2.34
25 P. So¨rme, Y. Qian, P.-G. Nyholm, H. Leffler and U. J. Nilsson,
ChemBioChem, 2002, 3, 183–189.
26 P. Konradsson, U. E. Udodong and B. Fraser-Reid, Tetrahedron
Lett., 1990, 31, 4313–4316.
27 G. H. Veeneman, S. H. van Leeuwen and J. H. van Boom, Tetrahedron
Lett., 1990, 31, 1331–1334.
28 Scrubber 1.1f, 2000, BioLogic Software Pty Ltd, Unit 30 116 Blamey
Crescent, Campell, ACT 2612, Australia.
29 J. Ohlsson, A. Sundin and U. J. Nilsson, Chem. Commun., 2003,
384–385.
30 S. Haataja, K. Tikkanen, J. Liukkonen, C. Franc¸ois-Gerard and J.
Finne, J. Biol. Chem., 1993, 268, 4311–4317.
31 MOE 2003.02, 2003, Chemical Computing Group Inc., 1010 Sher-
brooke St. West Suite 910 Montreal, Canada H3A 2R7.
32 S. Wold, in Chemometric methods in molecular design, ed. H. van de
Waterbeemd, Wiley, Weinheim, 1995, p. 195–218.
33 Simca-P 10.02ꢀ, 2002, Umetrics, Box 7960, S-907 19 Umea˚, Sweden.
34 L. Eriksson, E. Johansson and S. Wold, in Quantitative Structure-
Activity Relationships in Environmental Sciences-VII, ed. F. Chen
and G. Schu¨u¨rmann, 1997, SETAC Press, Pensacola, FL, p. 381–
397.
Acknowledgements
This work was supported by grants from the Swedish Research
Council, from the program “Glycoconjugates in Biological
35 J. Ohlsson and U. J. Nilsson, Tetrahedron Lett., 2003, 44, 2785–2787.
9 0 0
O r g . B i o m o l . C h e m . , 2 0 0 5 , 3 , 8 8 6 – 9 0 0