Novel glycodendrimers self-assemble to nanoparticles which function as polyvalent ligands in vitro and in vivo

Article abstract of DOI:10.1002/1521-3773(20020902)41:17<3195::AID-ANIE3195>3.0.CO;2-X

Size is important: Noncovalent nanoparticles are formed by the self-assembly of glycodendrimers containing both a ligand and a self-assembling moiety. Particle size reaches an optimum at the second and third generation, larger dendritic species show less

Full text of DOI:10.1002/1521-3773(20020902)41:17<3195::AID-ANIE3195>3.0.CO;2-X

Evaluation Only. Created with Aspose.PDF. Copyright 2002-2021 Aspose Pty Ltd.
COMMUNICATIONS
Novel Glycodendrimers Self-Assemble to
Nanoparticles which Function as Polyvalent
Ligands In Vitro and In Vivo**
Gebhard Thoma,* Andreas G. Katopodis,
Nicolas Voelcker, Rudolf O. Duthaler, and
Markus B. Streiff
The recognition of oligosaccharides by proteins represents
the basis of many biologically important events.[1] Individual
protein carbohydrate interactions are generally weak (KD ¼
10À3 10À4 mÀ1).[2] To overcome this, such processes often
involve polyvalent binding, which is characterized by the
simultaneous contact of multiple ligands (oligosaccharides)
on one biological entity to multiple receptors (proteins) on
another.[3] Polyvalent carbohydrate protein interactions oc-
cur frequently in recognition events on cellular membranes.
Collectively, they can be much stronger than corresponding
monovalent interactions rendering it difficult to control them
with individual small molecules.[4] Therefore, complex macro-
molecules have been used as polyvalent antagonists, however,
both characterization and preparation of these nonuniform
entities is demanding.[4] Here we present an alternative
concept for the polyvalent presentation of ligands based on
the supramolecular chemistry[5] of small molecules that fulfil
single-molecule entity criteria (Figure 1). Novel dendrons
capped with carbohydrate ligands (glycodendrimers[6]) were
found to self-assemble to noncovalent nanoparticles which
function as polyvalent ligands. We demonstrate that these
particles–not the individual molecules–efficiently inhibit
polyvalent interactions, such as IgM binding (IgM ¼ immu-
noglobulin), to the aGal-epitope[7] (a-d-Gal-(1!3)-b-d-Gal-
(1!4)-d-GlcNAc), both in vitro and in vivo. As self-assembly
is dynamic, optimization of size and shape of the polyvalent
ligand could occur utilizing the receptor as a template.
Dendrimer cores were prepared by a convergent ™outside-
in∫ approach[8] based on a single building block 1a which was
obtained from methyl 3,5-diaminobenzoate and 4-(tert-butoxy-
carbonylaminomethyl)benzoic acid (Scheme 1a). Selective
deprotection furnished 1b and 1c (first-generation dendrimer
core, two end-groups). A one-pot procedure comprising
coupling of 1c (1 equiv) and 1b (0.5 equiv) followed by
methyl ester cleavage gave 2c (second-generation dendrimer
core, four end groups).[9] The third-generation dendrimer 3c
(eight end groups) was obtained from 2c (1 equiv) and 1b
(0.5 equiv).[9] Applying the same procedure repetitively gave
Figure 1. Individual molecules, comprised of both a ligand and a self-
assembling moiety, form noncovalent nanoparticles which function as
multivalent ligands. If self-assembly is a dynamic process natural polyvalent
receptors could serve as templates optimizing size and shape of their own
polyvalent inhibitors.
dendrimers with up to 64 end groups (4c, fourth generation,
16 end groups; 5c, fifth generation, 32 end groups; 6c, sixth
generation, 64 end groups). Dendrimers 1c 6c were depro-
tected (!1d 6d) and transformed into their chloroacetamide
derivatives (1e 6e) to allow subsequent introduction of
thiolated oligosaccharides such as aGal-SH[10] and Lac-SH
(Figure 2b) furnishing water-soluble glycodendrimers 1 f 6 f
and 3g which were purified by ultrafiltration. Compound 7,
which is similar to 2 f but contains butylene chains instead of
the disubstituted aromatic rings, was also prepared (Sche-
me 1a). The integrity of all compounds was established by 1H
NMR spectroscopy. Accordingly, the first- to third-generation
dendrimers exist as single molecules (purity > 95%). The
fourth- to sixth-generation dendrimers possibly contain
minute quantities of smaller fragments. The 500 MHz 1H
NMR spectra of compound 3 f in [D6]DMSO demonstrates
the remarkable purity of these compounds (Figure 2).
The first indication that our glycodendrimers were aggre-
1
gating in water came from H NMR spectroscopy of 2 f in
D2O. At ambient temperature, we observed very broad signals
which sharpened at elevated temperatures. The aggregation
was quantified using multiangle light scattering (MALS;
Table 1). The first-generation dendrimer 1 f forms small
aggregates (50 kDa) whereas 2 f forms large particles of
7000 kDa (more than 1500 individual molecules per particle).
Interestingly, the particle weight obtained for 3 f 6 f drops
(2200 to 200 kDa) with increasing mass of the individual
molecule. The root-mean-square radii of the particles formed
by 2 f 6 f showed the same trend (for 2 f, 3 f, and 4 f 49, 34, and
12 nm, respectively; for 5 f and 6 f the radii were below the
detection limit of 10 nm). Core-modified second-generation
glycodendrimer 7 (4 î aGal), which is of comparable size and
lipophilicity as 2 f but contains butylene chains instead of the
disubstituted aromatic rings, does not form aggregates. The
third-generation compounds 3g (8 î Lac; 1900 kDa) and 3 f
[*] Dr. G. Thoma, Dr. A. G. Katopodis, Dr. R. O. Duthaler,
Dr. M. B. Streiff
Novartis Pharma AG
P.O. Box, 4002 Basel (Switzerland)
Fax : (þ 41) 61 324 6735
E-mail: gebhard.thoma@pharma.novartis.com
Dr. N. Voelckerþ
The Scripps Research Institute
Beckman Centre for Chemical Sciences
10550, N. Torrey Pines Rd., La Jolla, CA 92034 (USA)
[**] G.T. thanks Prof. M. Reza Ghadiri for the opportunity of a sabbatical
stay in his laboratory at the Scripps Research Institute. We also would
like to thank Mr. Klaus Menninger for his assistance in the monkey
experiment.
Angew. Chem. Int. Ed. 2002, 41, No. 17
¹ 2002 WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim
0044-8249/02/4117-3195 $ 20.00+.50/0
3195

Evaluation Only. Created with Aspose.PDF. Copyright 2002-2021 Aspose Pty Ltd.
COMMUNICATIONS
Scheme 1. A) Reaction conditions: a) TFA, Et3SiH, H2O (95:2.5:2.5), 3 h, RT (> 80%); b) 1n NaOH/dioxane (1:1), 16 h, RT (82%); c) 1) 1a (0.5 equiv),
HBTU (1.0 equiv), DMF/iPr2NEt (4:1), 6 16 h, RT; 2) LiOH, H2O (> 75%); d) chloroacetic acid anhydride (3.0 equiv), DMF/2,6-lutidine (3:1), 3 h, RT (50
95%); e) aGal-SH or Lac-SH (1.5 equiv per end group), DMF, DBU, 1 h, RT; B) structures of aGal-SH and Lac-SH. TFA ¼ trifluoroacetic acid, HBTU ¼ O-
(benzotriazol-1-yl)-N,N,N’,N’-tetramethyluronium hexafluorophosphate, DMF ¼ N,N-dimethylformamide
Table 1. Properties of Dendrimers.
Compound Mr[a]
(individualMr[b]
(aggrega-IgM[c] Hemolysis[c]
(end groups) molecule) [kDa] te) [kDa]
IC50 [mm]IC50 [mm]
1 f (2)
2 f (4)
3 f (8)
4 f (16)
5 f (32)
6 f (64)
3g (8)
7 (4)
1.908
4.198
8.779
17.941
36.264
72.911
7.154
50 ꢀ 5
> 10
0.025
0.010
0.019
> 1
> 1
> 100
> 100
> 100
0.035
0.010
0.180
0.550
2.240
n.d.
7100 ꢀ 250
2200 ꢀ 70
1200 ꢀ 50
270 ꢀ 10
200 ꢀ 10
1900 ꢀ 100
7 ꢀ 1
4.078
> 100
[a] Satisfactory MALDI mass spectra were obtained for all compounds
with the exception of 5 f and 6 f. Interestingly, intense fragmentation was
observed for all compounds, even for the first-generation glycodendrimer
1 f. [b] All determinations of the weight-average molar mass (Mr) were
performed by Wyatt Technology Deutschland GmbH by multi-angle light
scattering (MALS) using a DAWN EOS and
a Wyatt Optilab 903
refractometer; at 208C; concentration 0.3 mgmLÀ1 (0.06 mgmLÀ1 for 2 f
as the signal intensity was too high at 0.3 mgmLÀ1 because of high-
molecular-weight particles); refractive index increment (dn/dc) values were
between 0.18 and 0.22 mLgÀ1. [c] Average of at least two experiments
(variations within a factor of 2.5). For assay conditions see bioassays.
Concentrations refer to equivalent concentration of trisaccharide but not to
the concentration of oligovalent glycodendrimer.
Figure 2. 500 MHz 1H NMR spectrum of 3 f (third generation, eight end
groups) recorded in [D6]DMSO/D2O (5:1) at 758C. The spectrum
demonstrates the remarkable purity (> 95%) of this compound. The
indicated signals represent 8 î H-1 of the terminal a-linked galactose
residue (see Scheme 1b, red) and 6 î Ar-CH2-NHCO-Ar of the dendrimer
core (see Scheme 1a, green). Their relative intensities (8.0:12.2; expected
8.0:12.0) proves the quantitative functionalization with aGal. The small
signals (indicated by ≤) at d ¼ 8.05 ppm (intensity 1H) and d ¼ 8.38 ppm
(intensity 2H) can be assigned to the protons of the terminal trisubstituted
ring of the dendrimer core structure, bearing the carboxylic acid
functionality. The signal at d ¼ 8.26 ppm (intensity 6H, indicated by #)
(8 î aGal; 2200 kDa), with identical backbones but different
carbohydrate end groups, form aggregates of very similar size.
Thus, the particle weight strongly depends on the dendrimer
core structure but less on the size of the hydrophilic end
groups. Intermolecular hydrogen bonds seem not to be
significantly involved in self-assembly since particle weights
reflects the protons between the
trisubstituted rings.
N substituents of the six internal
3196
¹ 2002 WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim
0044-8249/02/4117-3196 $ 20.00+.50/0
Angew. Chem. Int. Ed. 2002, 41, No. 17

Evaluation Only. Created with Aspose.PDF. Copyright 2002-2021 Aspose Pty Ltd.
COMMUNICATIONS
are not reduced by the H-bond-disrupting reagent guanidi-
nium hydrochloride (1.5m), as determined by light scattering.
Most probably, the highly aromatic core induces aggregation
as a result of p stacking and/or rigidity. The core of 1 f seems
to be too small for efficient core core interactions whereas
the second-generation core of 2 f is optimal for self-assembly.
The decreasing particle weight obtained for higher generation
glycodendrimers (3 f 6 f) could be explained by the increasing
number of large end groups per individual molecule inducing
a more globular shape. As a consequence, the core is more
efficiently shielded by carbohydrates, which renders intermo-
lecular core core contact more difficult. The particle weight
does not substantially depend on the concentration, as shown
for dendrimer 2 f (7000 kDa at 0.06 mgmLÀ1, 6000 kDa at
0.03 mgmLÀ1, 3700 kDa at 0.01 mgmLÀ1), but is significantly
affected by temperature (Table 2); it drops ten fold with an
increase in temperature to 708C, but increases as the temper-
Figure 3. AFM images of particles formed by compounds 2 f 6 f. The size
of the particles decreases with increasing molecular weight of the individual
glycodendrimer. Aqueous solutions (10 mL; 2 mgmLÀ1 for 2 f 5 f,
0.2 mmLÀ1 for 6 f) were spotted on freshly cleaved mica. The solvent was
allowed to evaporate for at least 120 min. A Nanoscope III microscope
(Digital Instruments) was used to generate the images. The microscope was
operated in tapping mode in air with Olympus etched silicon probes
(cantilever length 120 mm, tip height 10 mm). The particles exhibit disklike
morphologies. Their flatness (1 5 nm) might be caused by interactions with
the mica surface. The diameters (1) were determined by measuring over
30 randomly selected particles.
Table 2. Effect of temperature on the weight of nanoparticles formed by
2 f.
T
Heating period[a]
Cooling period[b]
Mr [kDa]
Mr [kDa]
308C
508C
708C
5000 ꢀ 500
2200 ꢀ 300
550 ꢀ 100
4600 ꢀ 500
2600 ꢀ 500
350 ꢀ 50
been demonstrated that polymers providing multiple copies
of the aGal epitope can inhibit IgM binding in vitro.[10,11] To
assess our compounds as polyvalent IgM ligands we employed
two in vitro assays in which the inhibition of both the anti-
aGal IgM binding to the xenoantigen[12a] and the aGal-
mediated lysis of pig erythrocytes[13a] were measured (Ta-
ble 1). In both assays monomeric aGal was inactive at 100 mm,
which is in agreement with the finding that individual
carbohydrate protein interactions are often weak.[2] Divalent
first-generation dendrimer 1 f which forms small aggregates
(50 kDa) also showed no effect in the binding assay. The
second- and third-generation dendrimers 2 f and 3 f, which
form large nanoparticles, were highly potent in both assays
(0.025, 0.035 mm and 0.010, 0.010 mm, respectively).[14] Com-
pound 4 f also showed high potency in the binding assay
(0.019 mm) but was significantly less potent in the hemolysis
assay (0.18 mm). Potency dropped even more for the larger
dendrimers 5 f and 6 f which do not aggregate. Thus, especially
in the more relevant hemolysis assay, potency clearly corre-
lates with the size of the aggregates but not with the size of the
individual molecules. The inhibition data suggest that the
optimal particle weight and size for IgM inhibition are
obtained with 3 f. The particles formed by 2 f might be larger
than the optimum, but still displayed high potency. The
aggregates formed by 1 f, 4 f, 5 f, and 6 f are apparently too
small to accomplish polyvalent amplification of IgM inhib-
ition. Further evidence for the involvement of polyvalent
aggregates (but not oligovalent dendrimers) in antibody
binding stems from the observation that backbone-modified
tetravalent aGal dendrimer 7 (size and lipophilicity compa-
rable to 2 f), which does not aggregate, was found to be
inactive whereas 2 f was highly potent.[15] Nonspecific inter-
actions of the dendrimer backbone can be excluded because
dendrimer 3g (identical backbone and similar aggregation
[a] A cold solution of 2 f (0.03 mgmLÀ1) was slowly heated to the indicated
temperature and after 15 min equilibration time was injected into a
thermostat-controlled cell. [b] A hot solution of 2 f (0.03 mgmLÀ1) was
slowly cooled down to the indicated temperature and after 15 min
equilibration time was injected into a thermostat-controlled cell.
ature is lowered. Comparable particle weights were deter-
mined at identical temperatures for both the heating and the
cooling period which indicates there is a dynamic self-
assembly process which rapidly reaches thermodynamic
equilibrium. Both the shape and size of the formed nano-
particles was investigated by atomic force microscopy (AFM).
Aqueous solutions of 2 f 6 f were applied to mica surfaces and
the solvent evaporated. The observed particles have a disklike
morphology (Figure 3), and their average diameters decrease
with increasing mass of the individual molecules (61 nm, 2 f;
13 nm, 6 f), thus showing the same trend as the molecular
weights in solution. The size distribution for the individual
generations was found to be remarkably homogenous (within
a factor of two).
To exemplify the use of our glycodendrimers as noncova-
lent multivalent ligands we selected the inhibition of deca-
valent IgMs directed against the aGal xenoantigen.[7] This
epitope is expressed on the cells of mammals, with the
exception of humans and Old World monkeys. Anti-aGal
antibodies are the most abundant natural human antibodies
and cause the hyperacute rejection of pig organs transplanted
into primates. They include IgG, IgM, and IgA classes with the
IgM class being dominant in the immediate, complement-
dependent, hyperacute destruction of the xenograft. Current-
ly there are no pharmacological agents that affect the
production of this class of antibodies and research efforts
have focused on blocking or removing anti-aGal antibodies
prior to pig-to-primate xenotransplantation. Recently, it has
Angew. Chem. Int. Ed. 2002, 41, No. 17
¹ 2002 WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim
0044-8249/02/4117-3197 $ 20.00+.50/0
3197

Evaluation Only. Created with Aspose.PDF. Copyright 2002-2021 Aspose Pty Ltd.
COMMUNICATIONS
[7] a) Subcellular Biochemistry, Vol. 32 (Eds.: U. Galili, J. L. Avila),
Kluwer Academic/Plenum, New York, 1999; b) M. S. Sandrin, I. F. C.
McKenzie, Immunol. Rev. 1994, 141, 169.
[8] a) J. M. J. Frÿchet, Science 1994, 263, 1710; b) G. R. Newkome, C. N.
Moorefield, F. Vˆgtle, Dendritic Molecules: Concepts, Syntheses,
Perspectives VCH, Weinheim, 1996.
properties as 3 f but Lac instead of aGal) did not inhibit
binding. The most potent compound 3 f was selected for
in vivo profiling in cynomolgus monkeys (1 mgkgÀ1, intra-
venous, n ¼ 3; Figure 4). Within five minutes after injection
the anti-aGal IgMs, as detected by enzyme-linked immuno
[9] Ethyl diisopropylamine (3 mL) was added to
(0.50 equiv), 1c 5c (1.00 g; 1.00 equiv), and O-(benzotriazol-1-yl)-
N,N,N’,N’-tetramethyluronium hexafluorophosphate (HBTU)
a mixture of 1b
(1.00 equiv) in dry DMF (10 mL). The solution was stirred for 16 h
at room temperature. DMF (5 mL), water (3 mL, dropwise), and
LiOH monohydrate (approximately 50 equiv) were added and the
mixture was stirred for an additional 16 h at room temperature. The
solution was added dropwise to a mixture of acetone and 0.2n HCl
(250 mL, 1:3). The precipitate formed was filtered off and washed with
water, acetone, and dichloromethane. Products 2c 6c (all yields
greater than 80%) were isolated as beige solids.
[10] R. Duthaler, A. Katopodis, W. Kinzy, R. Ohrlein, G. Thoma,
WO 9847915 1998; [ Chem. Abstr. 1998, 129, 331058].
[11] J.-Q. Wang, X. Chen, W. Zhang, S. Zacharek, Y. Chen, P. G. Wang, J.
Am. Chem. Soc. 1999, 121, 8174.
[12] ELISAs: a) Human serum: Wells on plate A were coated with 0.5 mg
of aGal-HSA (HSA ¼ human serum albumin) conjugate in NaHCO3
(100 mL, 100 mm; pH 9.6) for 2 h at 378C. Blocking with phosphate
buffer saline (PBS, 250 mL) containing 0.5% Tween20 for 2 h at
ambient temperature was followed by removal of the liquid. On
plate Bpooled human serum was diluted (1:20) with PBS containing
10% Seablock and 0.2% Tween20 and incubated for 30 min with
serial dilutions of dendrimers or other inhibitors. Samples (100 mL)
from each well on plate Bwere transferred to plate A, incubated for
1 h at ambient temperature, and washed three times with PBS
containing 0.15% Tween20. They were then incubated with PBS
containing anti-IgM secondary antibody peroxidase conjugate (1/2000
diluted, 100 mL), 10% Blocker-BSA, and 0.2% Tween20. After
washing with PBS containing 0.1% Tween20 (6 ꢁ 300 mL), color
development was initiated by adding TMBsubstrate solution (100 mL,
a mixture of 3,3’,5,5’-tetramethylbenzidin, H2O2, and stabilizers) and
stopped after 5 min with 1m H2SO4 (50 mL). The absorbance at 450nm
(A450) was measured; b) Cynomolgus monkey serum: wells were
coated with aGal-polymer[10] (0.5 mg) in PBS (100 mL) overnight at
48C. Blocking with PBS containing 10% Seablock (300 mL) for 2 h at
ambient temperature was followed by removal of the liquid. The wells
were incubated for 30 min with 100 mL of serially diluted serum
sample in PBS containing 10% Seablock and 0.2% Tween20 and,
subsequently, washed three times with PBS containing 0.15%
Tween20. Incubation with a secondary antibody and color develop-
ment were performed as described above.
[13] Hemolysis assays. a) Human serum: pig erythrocytes were obtained
from heparinized pig blood, washed three times with CFD buffer
(3.1 mm diethyl barbituric acid, 0.9 mm sodium barbitone, 145 mm
NaCl, 0.83 mm MgCl2, 0.25 mm CaCl2; pH 7.4), and suspended in
CFD at a concentration of 1 î 109 mLÀ1. Seven serial dilutions of
inhibitor were prepared on low-bind U-shaped plates in CFD (50 mL).
Human serum was serially diluted in CFD (50 mL, 10 dilutions assayed
for each inhibitor concentration). Then, pig erythrocyte solution
(50 mL) was added to each well and the plate incubated for 60 min at
378C with mild shaking. Plates were centrifuged for 10 min at 2000g to
precipitate the unlyzed erythrocytes, the supernatant was transferred
to a flat-bottomed plate, and the released hemoglobin (A420) was
measured; b) Cynomolgus monkey serum: the assay was carried out
as above, but with the addition of baby rabbit complement (50 mL,
heterologous complement source at a final dilution of 1:10) in CFD
and pig erythrocyte solution (50 mL) to each well.
Figure 4. Effects of 3 f on anti-aGal IgM levels and cytolytic activity of the
sera of cynomolgus monkeys (n ¼ 3; 1 mgkgÀ1
; intravenous; plasma
concentration approximately 10 mm assuming all of 3 f is present). Prior
to injection (t ¼ 0) the IgM 100% level (l) was 2.3 times higher than the
IgM level of pooled human sera.
sorbent assay (ELISA), were reduced to 20% of the initial
value and remained at low levels for more than 4 h.[12b] Most
importantly, anti-aGal antibody-mediated haemolytic activity
was completely eliminated.[13b]
In conclusion, we have described novel, self-assembling
glycodendrimers which form noncovalent nanoparticles in
water. The particle sizes are remarkably homogenous and can
be varied over a broad range by the appropriate choice of the
dendrimer generation. The particles can be deposited on
surfaces by evaporation of the solvent. In solution, they
function as noncovalent polyvalent inhibitors both in vitro
and in vivo. The self-assembly of the glycodendrimers is a
dynamic equilibrium process. Thus, it is conceivable that
noncovalent polyvalent ligands are optimized with respect to
size and shape in the presence of natural polyvalent receptors.
The control of a broad variety of physiologically relevant
polyvalent interactions should be possible.
Received: April 16, 2002 [Z19101]
[1] a) A. Varki, Glycobiology 1993, 3, 97; b) R. A. Dwek, Chem. Rev.
1996, 96, 622.
[2] Y. C. Lee, R. T. Lee, Acc. Chem. Res. 1995, 28, 321.
[3] a) L. L. Kiessling, N. L. Pohl, Chem. Biol. 1996, 3, 71; b) R. Roy,
Trends Glycosci. Glycotechnol. 1996, 8, 79.
[14] Polyvalent amplification was found to be less pronounced for the
inhibition of divalent IgG binding. Compounds 1 f 6 f showed very
similar IC50 values in the range of 0.45 0.85 mm. Compound 3g gave
no inhibition at 10 mm whereas 7 showed an IC50 value of 1.80 mm.
[15] It also has to be considered that the individual molecules 2 f (Mr
4 kDa) and 3 f (9 kDa) are too small to allow for simultaneous contact
of several oligosaccharide ligands with several binding sites of an
individual IgM molecule (900 kDa).
[4] M. Mammen, S.-K. Choi, G. M. Whitesides, Angew. Chem. 1998, 110,
2908; Angew. Chem. Int. Ed. 1998, 37, 2754.
[5] J. M. Lehn, Supramolecular Chemistry, VCH, Weinheim, 1995.
[6] Glycodendrimers have been described earlier. See, for example, a) N.
Rockendorf, T. K. Lindhorst, Top. Curr. Chem. 2001, 217, 201; b) R.
Roy, M.-G. Baek, K. Rittenhouse-Olson, J. Am. Chem. Soc. 2001, 123,
1809; c) B. Colonna, V. D. Harding, S. A. Nepogodiev, F. M. Raymo,
N. Spencer, J. F. Stoddart, Chem. Eur. J. 1998, 4, 1244.
3198
¹ 2002 WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim
0044-8249/02/4117-3198 $ 20.00+.50/0
Angew. Chem. Int. Ed. 2002, 41, No. 17