Design, synthesis, and biological evaluation of human sialidase inhibitors. Part 1: Selective inhibitors of lysosomal sialidase (NEU1)

Article abstract of DOI:10.1016/j.bmcl.2007.11.084

We here report the design and synthesis of selective human lysosomal sialidase (NEU1) inhibitors. A series of amide-linked C9 modified DANA (2-deoxy-2,3-dehydro-N-acetylneuraminic acid) analogues were synthesized and their inhibitory activities against al

Full text of DOI:10.1016/j.bmcl.2007.11.084

Available online at www.sciencedirect.com
Bioorganic & Medicinal Chemistry Letters 18 (2008) 532–537
Design, synthesis, and biological evaluation of human
sialidase inhibitors. Part 1: Selective inhibitors
of lysosomal sialidase (NEU1)
Sadagopan Magesh,a,* Setsuko Moriya,b Tohru Suzuki,c
Taeko Miyagi,b,d Hideharu Ishidaa and Makoto Kisoa,d,
*
aDepartment of Applied Bioorganic Chemistry, Gifu University, Gifu 501-1193, Japan
bDivision of Biochemistry, Miyagi Cancer Center Research Institute, Natori, Miyagi 981-1293, Japan
cLife Science Research Center, Gifu University, Gifu 501-1193, Japan
dCREST, Japan Science and Technology Corporation (JST), Tokyo, Japan
Received 25 September 2007; revised 12 November 2007; accepted 21 November 2007
Available online 28 November 2007
Abstract—We here report the design and synthesis of selective human lysosomal sialidase (NEU1) inhibitors. A series of amide-
linked C9 modified DANA (2-deoxy-2,3-dehydro-N-acetylneuraminic acid) analogues were synthesized and their inhibitory activ-
ities against all four human sialidases (NEU1–NEU4) were determined. Structure-based approach was used to investigate the basis
of selectivity of the compounds with experimentally observed activity. Results from the present study are found to be informative in
a qualitative manner for the further design of isoform selective human sialidase inhibitors for therapeutic value.
Ó 2007 Elsevier Ltd. All rights reserved.
Sialidases (Neuraminidases) contribute to the regulation
of many cellular activities by catalyzing the hydrolysis of
terminal sialic acid moiety from various oligosaccha-
rides, glycolipids, and glycoproteins. They are wide-
spread in nature, from viruses, microorganisms like
bacteria, fungi, and protozoa to higher animals includ-
ing humans.1 They are thought to be involved in several
biological events such as molecular transport, antigen
masking, proliferation, differentiation, transformation,
and signal transduction.2
four types of sialidases are known and have been classified
based on their subcellular localization, namely the intra
lysosomal sialidase NEU1, the cytosolic sialidase NEU2,
the plasma membrane-associated sialidase NEU3, and
the lysosomal or mitochondrial membrane-associated
sialidase NEU4. These isoforms differ in their substrate
specificities and physiological functions.6
Aberrant expression of human sialidases has been found
to be associated with the development of various patho-
logical conditions. In a recent study with NEU1 knock-
out mice, reduced IDL and LDL cholesterol levels and
decreased atherosclerotic lesions were observed.7
NEU3 was found to be significantly up-regulated in var-
ious human cancers and its up-regulation was correlated
with the apoptosis suppression and the promotion of
motility in cancer cells.8 Apart from this, NEU3 overex-
pression in mice developed a diabetic phenotype associ-
ated with hyperinsulinemia, islet hyperplasia, and
increased b-cell mass.9 These studies suggest that they
might play important etiological and pathogenic roles
in cancer, arteriosclerosis, and diabetes. Furthermore,
NEU2 expression was found to induce the apoptosis
and was inversely correlated with the metastasis, inva-
sion, and motility.10 NEU4 has proved to have an
important role in the maintenance of normal mucus
Sialidase is a virulence factor for many pathogens like
virus,bacteria,andprotozoa.3Todate,sialidasefromviral
origin has been most extensively studied, which led to
structure-based drug discovery of a new class of antiviral
agentsthatspecificallytargettheinfluenzavirus.4Sialidase
from the bacteria Vibrio cholerae and the trans-sialidase
from the protozoa Trypanosoma cruzi are also considered
as potential drug targets for the therapeutic agents against
cholera and chagas’s disease, respectively.5 In humans,
Keywords: Sialidase; Inhibitor design; DANA; Selectivity; Homology
models; Docking.
*
Corresponding authors. Tel.: +81 58 293 2918; fax: +81 58 293
2840; e-mail addresses: sadmag77@gmail.com; kiso@gifu-u.ac.jp
0960-894X/$ - see front matter Ó 2007 Elsevier Ltd. All rights reserved.
doi:10.1016/j.bmcl.2007.11.084

S. Magesh et al. / Bioorg. Med. Chem. Lett. 18 (2008) 532–537
533
and its down-regulation may contribute to invasive
properties of colon cancers, and also found to be associ-
ated with mitochondrial apoptosis pathway.11
homology modeling techniques to construct a three-
dimensional model structure of sialidases NEU1,
NEU3, and NEU4 using the crystal structure of
NEU2.12 The overlay of crystal structure of NEU2–
DANA complex with predicted models, and DANA
(1) and its four constituent groups required for binding
are given in Figure 1. Despite the fact that human NEU
isoforms share a high-level amino acid conservation in
the active site and its vicinity, they seem to have some
striking differences, particularly in the DANA’s glycerol
binding group, which could be exploited to design the
selective inhibitors. Accordingly, we designed some of
the hydrophobic derivatives of DANA modified at glyc-
erol side chain. We began our investigation with some
varying hydrophobic groups linked to N-amide at C9
position of DANA, thus allowing utilization of steric
fit/hindrance principles to probe the selectivity among
human isoforms.
The differences in the localization and the physiological
function of the various isoforms of this enzyme
prompted us to use these differences to design inhibitors
that could be specific to each of these isoforms. The
main objective of our work is ‘Design in’ over NEU1
or NEU3 and ‘Design out’ of its paralogs to minimize
the cross reactivity or potential side effects. In any case,
identification of selective inhibitors for individual iso-
forms can be used as molecular probes for specific func-
tions of human sialidases.
In our previous study we used both sequence and struc-
tural information to predict the selectivity-related resi-
dues using computational methods. We employed
A
B
6-glycerol
1
OH
9
H
O
2
1COOH
8
2-carboxyl
HO
7
6
OH
3
AcHN
5
4
OH
5-N-acetyl
4-hydroxyl
Figure 1. (A) Overlay of the crystal structure of NEU2–DANA complex with predicted models. The active site residues of NEU1, NEU2, NEU3,
and NEU4 are colored green, red, pink, and blue, respectively. (B) DANA (1) and its four constituent groups required for binding.

534
S. Magesh et al. / Bioorg. Med. Chem. Lett. 18 (2008) 532–537
OAc
OH
OH
H
Cl
H
OH
H
OH
O
COOEt
O
O
c
COOH
a
b
COOEt
AcO
HO
HO
OAc
OH
OH
AcHN
AcHN
AcHN
OAc
OH
OH
3
4
2
OAc
OH
OH
H
H
H
O
COOEt
COOEt
O
COOEt
O
COOEt
g
d, e
f
AcO
TsO
N3
OAc
OH
AcHN
OH
AcHN
AcHN
OAc
OH
OH
7
5
6
OAc
O
OH
O
OAc
H
H
H
O
O
COOH
O
COOEt
h
i
N3
R
N
H
R
N
H
OAc
OH
OAc
AcHN
AcHN
AcHN
OAc
OH
10a-j
OAc
9a-j
8
Scheme 1. Reagents and conditions: (a) H+ resin, EtOH, 50 °C, 1 d; (b) AcCl, rt, 3 d; (c) Et3N, DCM, rt, o/n (62% over three steps); (d) MeONa,
EtOH, rt, o/n (75%); (e) TsCl, pyridine, 0 °C, 1 d (71%); (f) NaN3, acetone–H2O, 70 °C, 2 d (80%); (g) Ac2O, pyridine, rt, o/n (76%); (h) RCOCl,
DPPE, THF, rt, 2–24 h (45–70%); (i) 1—LiOH, MeOH–H2O–THF, 0 °C, 6 h; 2—H+ resin (>70%).
Synthetic route to the present series of C9 amide deriv-
atives of DANA is outlined in Scheme 1. Starting mate-
rial, N-acetylneuraminic acid (Neu5Ac) 2, was esterified
with ethanol in the presence of an acidic ion exchange
resin to give an ester 3. This ester was treated with acetyl
chloride under neat conditions to afford 4. Compound 5
was prepared by b-elimination of 4 in the presence of tri-
ethylamine as a base. Deprotection of the O-acetyl
groups in 5 with CH3ONa, followed by selective tosyla-
tion of the primary hydroxyl group using p-toluenesul-
fonyl chloride, provided 9-O-tosyl derivative 6, which
was converted to the corresponding azide 7 by the treat-
ment with sodium azide in aqueous acetone. Key inter-
mediate 8 was obtained from 7 in a high yield using
excess of acetic anhydride in pyridine. A Staudinger-
type reductive acylation of compound 8 with various
acid chlorides and DPPE gave the compounds 9a–j in
moderate yields. A small library of these compounds
was prepared using a simple parallel synthesizer. Subse-
quent deprotection of derivatives 9a–j with LiOH/H2O/
MeOH gave the target C9 amides of DANA 10a–j in
good yields.13
low lM to >1000 lM) in their binding affinities in com-
parison to the reference compound, DANA. The change
of linear alkyl groups from methyl 10a, propyl 10e to
butyl 10h (58 lM, 32 lM to 10 lM) has steadily in-
creased the potency of NEU1 though the order of mag-
nitude of improvement is low. Replacement of linear
chain with cycloalkyl like cyclopropyl 10c (680 lM) de-
creased the affinity toward NEU1, whereas cyclopentyl
10d (135 lM) showed nearly similar activity as DANA.
In the case of NEU4, 10a showed to some extent im-
proved affinity in comparison to other analogues in this
series. The replacement of methyl group in 10a with iso-
propyl 10g (>1000 lM) and a-methyl substitution of 10e
to 2-methylpropyl in 10f (565 lM) also reduced the
NEU1 inhibition. Surprisingly, phenyl ring-containing
derivative 10b (13 lM) inhibited NEU1 comparably to
10h, and showed slightly higher affinity toward human
sialidase NEU3 (320 lM) when compared with other
derivatives. Sterically bulky and branched alkyl groups
like tert-butyl 10i and 2-ethylpropyl 10j severely dimin-
ished the NEU1 inhibitory activity to over 1000 lM
range. These data indicate that n-butyl analogue 10h
has improved binding affinity (10 lM) to NEU1 with
ꢀ15-fold improvement over DANA (143 lM), whereas
10h has decreased binding affinity over a 100-fold to-
ward NEU2, NEU3, and NEU4 (>1000 lM) compared
to that of DANA. Taken together, more than a 200-fold
selectivity for NEU1 was achieved with 10h over NEU2,
NEU3, and NEU4 with respect to DANA.
Inhibitory activities of the C9 DANA amides 10a–j
against all four human sialidases were investigated.14
The IC50 values of compounds 10a–j and DANA (1)
are given in Table 1. In this series it has been found that
the size and the geometry of the alkyl side chains signif-
icantly influenced the potency and the selectivity of the
compounds toward the different NEU isoforms. The re-
sults obtained suggest that the hydrophobic amide
derivatives at C9 might be more selective for NEU1 as
the other isoforms show a significant reduction (from
In order to gain further insight into the structural basis
of inhibition, we performed the docking studies on the
basis of our homology models and the crystal structure

S. Magesh et al. / Bioorg. Med. Chem. Lett. 18 (2008) 532–537
535
Table 1. Chemical structures of compounds 10a–j and their inhibitory activities against human sialidases: NEU1, NEU2, NEU3, and NEU4
O
OH
H
O
COOH
R
N
H
OH
AcHN
OH
Compound
R
Sialidase inhibition (IC50 lM)
NEU1
NEU2
NEU3
NEU4
1 (DANA)
10a
10b
143
58
43
>1000
865
61
>1000
320
74
580
Methyl
Phenyl
13
135
810
10c
10d
Cyclopentyl
Cyclopropyl
Propyl
>1000
>1000
>1000
>1000
>1000
>1000
>1000
>1000
>1000
>1000
>1000
>1000
>1000
>1000
>1000
>1000
>1000
825
923
680
32
10e
10f
2-Methylpropyl
Isopropyl
n-Butyl
565
>1000
10
>1000
>1000
>1000
>1000
>1000
10g
10h
10i
10j
tert-Butyl
2-Ethylpropyl
>1000
>1000
of NEU2 (PDB entry 1VCU)15 using DS LigandFit.
LigandFit gives the best poses at the binding site by a
stochastic conformational search and by evaluation of
the energy of the ligand–protein complex.16 The com-
pounds 10a–j were docked into the active site of human
sialidase isoforms using DANA as a control ligand.17
The binding modes of compounds 10a, 10b, 10e, and
10h are convincing with NEU1, but are apparently in
an irrational mode with other isoforms active sites. In
our earlier study,12 we reported that the active site of
NEU1 is relatively larger and possesses a different shape
and charge distribution as compared to other isoforms,
which differ from each other by only two amino acids
(Leu217 and Gln270 in NEU2, Val224 and His277 in
NEU3, Gly221 and Trp274 in NEU4) proximal to the
glycerol side chain of DANA. Even though glycerol
binding subsite of NEU1 is majorly composed of acidic
amino acids (Asp226, Asn261, Asp263, and Gln282)
similar to viral sialidase, the hydrophobic faces of the
these amino acid side chains and Leu313 might form a
hydrophobic channel in the active site. The docking
observations suggest that linear alkyl and phenyl side
chains could sterically fit into this channel by forming
favorable van der Waals interactions which may ac-
count for their improved binding affinities. The binding
pose of C9 substituent of model compounds 10h and 10b
in NEU1 is shown in Figure 2. It appears that the com-
mon Tyr residue (181 in NEU2 and NEU3, 179 in
NEU4) is present in the edge of the active site of
NEU2–NEU4, sterically preventing the linear, branched
or cycloalkyl chains from the binding and this steric
clash might force the inhibitors to adopt irrational bind-
ing modes. As a result alkyl groups are exposed out into
the solvent space, and may contribute to the decrease in
their binding affinities towards NEU2, NEU3, and
NEU4, and the increased NEU1 selectivity. Interest-
ingly, the binding pose of phenyl analogue 10b in iso-
forms other than NEU1 shows p–p stacking against
the phenyl ring of the same Tyr residue, which can con-
tribute to the slight improvement in the observed activ-
ity as compared to other analogues. Additionally, the
docked pose of analogue 10b in NEU3 shows an extra
H-bond with His277 which can explain its increased
affinity to NEU3 as compared to NEU2 and NEU4.
The docking pose of simple methyl analogue 10a in
NEU4 clearly suggests that the loss of hydrophobic face
due to the replacement of Leu217 in NEU2 and Val224
in NEU3 by Gly221 leads to less steric hindrance at the
Figure 2. The binding pose of C9 substituent of model compounds 10h (A) and 10b (B) in the active site of NEU1 model. Green line represents the
hydrogen bonding.

536
S. Magesh et al. / Bioorg. Med. Chem. Lett. 18 (2008) 532–537
Figure 3. The binding pose of C9 substituent of compound 10a in NEU4 model (A) and compound 10b in NEU3 model (B). Green line represents
the hydrogen bonding.
Figure 4. The binding pose of C9 substituent of a branched chain compound 10g (A) and a cycloalkyl analogue 10c (B) in the active site of NEU1
Model. Green and pink lines represent the hydrogen bonding and steric clash, respectively.
binding site of NEU4. This can explain the slightly high-
er affinity of 10a to NEU4 over NEU2 and NEU3.
However, this compound (10a) is 10-fold less potent
than parent compound DANA 1 for NEU4. The bind-
ing pose of C9 substituent of compound 10a in NEU4
(A) and compound 10b in NEU3 (B) is shown in Figure
3. These observations strongly imply that C9 substitu-
ents of any size are too bulky and inflexible to fit into
the active site of NEU2, NEU3, and NEU4 and confer
an opportunity for NEU1 selectivity. The binding
modes of the branched chain analogues 10f, 10g, 10i,
and 10j suggest that the steric clashing of inhibitors with
the side chain of Asp226 in NEU1 can abolish their
binding affinity toward NEU1. Although the interacting
poses of cycloalkyl analogues 10c and 10d in NEU1 are
seemingly favorable, probably their conformational
rigidity could be a reason for the reduced binding affin-
ity. The binding pose of C9 substituent of isopropyl ana-
logue 10g and cyclopropyl analogue 10c in NEU1 is
shown in Figure 4. On the whole, our docking studies
clearly reveals a reasonable correlation between the
experimental IC50 values of the inhibitors with their pre-
dicted binding poses and further suggest that differences
in the active site topology of human sialidases might im-
part the selectivity to the inhibitors.
our knowledge, this is the first report of structure-based
design of selective human sialidase inhibitors. The ef-
forts to integrate structural information with inhibitor
activity were actively pursued. The results obtained in
the present study are thought to be important for our
continuing research efforts in the design and synthesis
of isoform selective human sialidase inhibitors for ther-
apeutic value. The synthesis and optimization of molec-
ular probes in this regard are ongoing, and the results
will be reported in due course.
Acknowledgments
The first author (S.M.) is grateful to Ministry of
Education, Culture, Sports, Science and Technology,
Government of Japan (MEXT), for the financial sup-
port, No. 042271. This work was supported in part by
Grants-in-Aid (No. 17101007) for Scientific Research
from MEXT and by CREST of JST (Japan Science
and Technology Corporation) to M.K. The authors
are thankful to Ms. Subashree and Ms. Savita for many
valuable discussions during preparation of the manuscript.
References and notes
In summary, we report here some C9 amide-linked
hydrophobic derivatives of DANA as selective human
lysosomal sialidase (NEU1) inhibitors. To the best of
1. Monti, E.; Preti, A.; Venerando, B.; Borsani, G. Neuro-
chem. Res. 2002, 27, 649.

S. Magesh et al. / Bioorg. Med. Chem. Lett. 18 (2008) 532–537
537
2. Achyuthan, K. E.; Achyuthan, A. M. Comp. Biochem.
Physiol. B: Biochem. Mol. Biol. 2001, 129, 29.
T.; Yamaguchi, K.; Uemura, T.; Kakugawa, Y.; Hujiya,
T.; Miyagi, T. Cancer Sci. 2007, 98, 299.
3. Rosenberg, A.; Schengrund, C.-L. In Biological Roles of
Sialic Acid; Rosenberg, A., Schengrund, C.-L., Eds.;
Plenum: New York, 1976; pp 295–359.
4. von Itzstein, M.; Wu, W. Y.; Kok, G. B.; Pegg, M. S.;
Dyason, J. C.; Jin, B.; Phan, T. V.; Smythe, M. L.;
White, H. F.; Oliver, S. W.; Colman, P. M.; Varghese,
J. N.; Ryan, D. M.; Woods, J. M.; Bethell, R. C.;
Hotham, V. J.; Cameron, J. M.; Penn, C. R. Nature
1993, 363, 418.
5. (a) Mann, M. C.; Thomson, R. J.; Dyason, J. C.;
McAtamney, S.; von Itzstein, M. Bioorg. Med. Chem.
2006, 14, 1518; (b) Neres, J.; Bonnet, P.; Edwards, P. N.;
Kotian, P. L.; Buschiazzo, A.; Alzari, P. M.; Bryce, R. A.;
Douglas, K. T. Bioorg. Med. Chem. 2007, 15, 2106.
6. Miyagi, T.; Kato, K.; Ueno, S.; Wada, T. Trends Glycosci.
Glycotechnol. 2004, 16, 371.
7. (a) Yang, A. E.; Tedesco, V.; Kufaishi, H. F.; Trigatti, B.
L.; Igdoura, S. A. Atheroscler. Suppl. 2006, 7, 488; (b)
Igdoura, S. A.; Yang, A.; DeSA, D.; Trigatti, B. L.
Atheroscler. Suppl. 2007, 8, 187; (c) Igdoura, S. A.; Trigatti
B. L. U.S. Patent Application 20070074300, 2007.
8. (a) Kakugawa, Y.; Wada, T.; Yamaguchi, K.; Yamanami,
H.; Ouchi, K.; Sato, I.; Miyagi, T. Proc. Natl. Acad. Sci.
U.S.A. 2002, 99, 10718; (b) Nomura, H.; Tamada, Y.;
Miyagi, T.; Suzuki, A.; Taira, M.; Suzuki, N.; Susumu, N.;
Irimura, T.; Aoki, D. Oncol. Res. 2006, 16, 289; (c) Ueno,
S.; Saito, S.; Wada, T.; Yamaguchi, K.; Satoh, M.; Arai,
Y.; Miyagi, T. J. Biol. Chem. 2006, 281, 7756; (d) Wada,
T.; Hata, K.; Yamaguchi, K.; Shiozaki, K.; Koseki, K.;
Moriya, S.; Miyagi, T. Oncogene 2007, 26, 2483.
12. Magesh, S.; Suzuki, T.; Miyagi, T.; Ishida, H.; Kiso, M.
J. Mol. Graph. Model. 2006, 25, 196.
13. All new compounds gave the satisfactory spectroscopic
and analytical data.
14. Sialidase inhibition assay: Inhibitory activities were eval-
uated using the homogenates obtained from HEK-293
cells transiently transfected with the expression plasmid
containing full-length human sialidase cDNA. For NEU1
expression, a protective protein cDNA was co-transfected
with NEU1 cDNA. Assays were performed using 4MU-
NeuAc (NEU1, NEU2, and NEU4) or mixed gangliosides
(NEU3) as a substrate. One unit (nmol/h) of the enzyme
was incubated for 30–60 min for each assay. Each com-
pound was tested at four different mM concentrations
(0.05, 0.1, 0.5, and 1) and IC50 values from concentration-
inhibition curves were calculated by means of non-linear
regression analysis using Microsoft Excel.
15. Chavas, L. M.; Tringali, C.; Fusi, P.; Venerando, B.;
Tettamanti, G.; Kato, R.; Monti, E.; Wakatsuki, S.
J. Biol. Chem. 2005, 280, 469.
16. Venkatachalam, C. M.; Jiang, X.; Oldfield, T.; Waldman,
M. J. Mol. Graph. Model. 2003, 21, 289.
17. Docking analysis: All docking calculations were performed
using LigandFit algorithm, which is implemented in DS
modeling 1.5, Accelrys Inc., San Diego, USA (www.
accelrys.com). Docking was performed in two-step pro-
cess. First, the active site was defined as a subset that
˚
contains residues in which any atom lies within 8.0 A from
the DANA. The defined active site was expanded to
accommodate the shapes of test compounds by adding
extra grid points. Then compounds 10a–j were flexibly
docked into active site of human sialidase models and into
the crystal structures of NEU2. Test compounds were
built and processed before docking by defining bond
orders and charges correctly. In the docking preferences, the
number of Monte Carlo trials was set to a fixed value of 5000
and the maximum number of saved poses to 10. Dreiding
force field was selected for the grid energy preferences and
conjugate gradient method was used for energy minimiza-
tion. The docked poses were manually checked for their
geometrical and electronic matching quality.
9. Sasaki, A.; Hata, K.; Suzuki, S.; Sawada, M.; Wada, T.;
Yamaguchi, K.; Obinata, M.; Tateno, H.; Suzuki, H.;
Miyagi, T. J. Biol. Chem. 2003, 278, 27896.
10. (a) Miyagi, T.; Wada, T.; Yamaguchi, K.; Hata, K.
Glycoconj. J. 2004, 20, 189; (b) Tringali, C.; Lupo, B.;
Anastasia, L.; Papini, N.; Monti, E.; Bresciani, R.;
Tettamanti, G.; Venerando, B. J. Biol. Chem. 2007, 282,
14364.
11. (a) Yamaguchi, K.; Hata, K.; Koseki, K.; Shiozaki, K.;
Akita, H.; Wada, T.; Moriya, S.; Miyagi, T. Biochem.
J. 2005, 390, 85; (b) Yamanami, H.; Shiozaki, K.; Wada,