Structural investigation of the binding of 5-substituted swainsonine analogues to golgi α-mannosidase II

Article abstract of DOI:10.1002/cbic.200900750

Golgi α-mannosidase II (GMII) is a key enzyme in the N-glycosylation pathway and is a potential target for cancer chemotherapy. The natural product swainsonine is a potent inhibitor of GMII. In this paper we characterize the binding of 5α-substituted swai

Full text of DOI:10.1002/cbic.200900750

DOI: 10.1002/cbic.200900750
Structural Investigation of the Binding of 5-Substituted
Swainsonine Analogues to Golgi a-Mannosidase II
Douglas A. Kuntz,[a] Shinichi Nakayama,[b] Kayla Shea,[a] Hitoshi Hori,[b] Yoshihiro Uto,[b]
Hideko Nagasawa,*[b, c] and David. R. Rose*[a, d]
Dedicated to the memory of Warren Delano (1972–2009).
Golgi a-mannosidase II (GMII) is a key enzyme in the N-glyco-
sylation pathway and is a potential target for cancer chemo-
therapy. The natural product swainsonine is a potent inhibitor
of GMII. In this paper we characterize the binding of 5a-substi-
tuted swainsonine analogues to the soluble catalytic domain
of Drosophila GMII by X-ray crystallography. These inhibitors
enjoy an advantage over previously reported GMII inhibitors in
that they did not significantly decrease the inhibitory potential
of the swainsonine head-group. The phenyl groups of these
analogues occupy a portion of the binding site not previously
seen to be populated with either substrate analogues or other
inhibitors and they form novel hydrophobic interactions. They
displace a well-organized water cluster, but the presence of a
C(10) carbonyl allows the reestablishment of important hydro-
gen bonds. Already approximately tenfold more active against
the Golgi enzyme than the lysosomal enzyme, these inhibitors
offer the potential of being extended into the N-acetylglucosa-
mine binding site of GMII for the creation of even more potent
and selective GMII inhibitors.
Introduction
The glycosylation profile on the surface of a tumor cell corre-
lates with its metastatic potential,[1] and control of aberrant
glycosylation offers potential for chemotherapy.[2] The enzyme
Golgi a-mannosidase II (GMII, E.C.3.2.1.114) is a promising
target for intervention in the glycosylation process. Swainso-
nine, one of the most potent inhibitors of GMII yet discovered,
and related pyrrolidines offer promising antitumor activity;[3]
however, unwanted swainsonine co-inhibition of the lysosomal
a-mannosidase (LM, E.C.3.2.1.24) has resulted in a search for
highly potent inhibitors that are more selective for GMII.
The a(1–3)/a(1–6)-mannosidases, including GMII and LM are
members of the family 38 glycoside hydrolases in the CAZy
classification system (http://www.cazy.org). These enzymes
cleave the bond between two mannose residues, by sequential
cleavage in a single active site.[4] The cleavage mechanism in-
volves formation of a covalent glycosyl-enzyme intermediate
and results in net retention of configuration.[5]
Branched complex-type carbohydrates have been observed
to increase following malignant transformation and appear to
play a role in metastasis. Treating murine tumor cells with the
plant derived indolizidine alkaloid swainsonine (Table 1) results
in cells that are less metastatic.[8] At the same time swainsonine
has positive effects on cellular immunity, and alleviates tumor-
associated immune suppression and increases bone marrow
cell proliferation.[8] Despite these positive aspects, the potential
for neurological damage resulting from lysosomal mannosi-
dase inhibition has spurred the search for new, more GMII-spe-
cific inhibitors.
Placing substituent groups on swainsonine at all positions
save the C(3) and C(5) positions (numbering as indicated in
[a] Dr. D. A. Kuntz, K. Shea, Prof. Dr. D. R. Rose
Ontario Cancer Institute and Department of Medical Biophysics
University of Toronto
101 College Street, Toronto, Ontario, M5G 1L7 (Canada)
Fax: (+1)519-746-0614
E-mail: drrose@uwaterloo.ca
GMII is involved in the creation of glycoproteins that contain
complex carbohydrates. It is responsible for the formation of
the “core” trimannose structure to which all complex carbohy-
drates are appended. It catalyses the hydrolysis of an a(1–6)-
and an a(1–3)-linked mannose from GlcNAc-Man5-GlcNAc2
(which is N-linked to an asparagine residue) to form GlcNAc-
Man3-GlcNAc2-Asn-X.[6]
[b] S. Nakayama, Prof. Dr. H. Hori, Dr. Y. Uto, Prof. Dr. H. Nagasawa
Department of Life System, Institute of Technology and Science
Graduate School, The University of Tokushima
Minamijosanjimacho-2, Tokushima 770-8506 (Japan)
[c] Prof. Dr. H. Nagasawa
Current address: Laboratory of Medicinal and Pharmaceutical Chemistry
Gifu Pharmaceutical University
1-25-4 Daigaku-nishi, Gifu 501-1196 (Japan)
E-mail: hnagasawa@gifu-pu.ac.jp
Lysosomal a-mannosidases, in contrast, are involved in the
degradation of complex sugars derived from glycoproteins. If
LM activity is impaired, either chemically or through mutation,
there is buildup of partially processed oligosaccharides in large
vacuolar structures within cells; this can have severe neurologi-
cal consequences.[7]
[d] Prof. Dr. D. R. Rose
Current address: Department of Biology, University of Waterloo
Waterloo, Ontario N2L 3G1 (Canada)
Supporting information for this article is available on the WWW under
http://dx.doi.org/10.1002/cbic.200900750.
ChemBioChem 2010, 11, 673 – 680
ꢀ 2010 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim
673

H. Nagasawa, D. R. Rose et al.
cluster of water molecules is dis-
placed by the substituents and a
water-free region close to the
backbone around Arg876 is cre-
ated.
Table 1. IC50 and Ki values for swainsonine derivatives in both lysosomal and Golgi mannosidase II. Numbering
of the swainsonine compound is shown below. The arrow indicates the position where the substituents are at-
tached. Compounds 2–4 and 6 are referred to as a-configured, while 5 and 7 are b-configured. The carbonyl
group in 2–5 is at C(10).
Lysosomal
Golgi
(Jack Bean)
(Drosophila)
(Drosophila)
(Drosophila)
Results and Discussion
C(5a)-substituted analogues
are as potent as swainsonine
Substituent at C(5)
H
IC50 [nm]
IC50 [nm]
IC50 [nm]
Ki [nm]
swainsonine
250
220
37
3.0
A
series of C(5)-substituted
2
510
250
n.d.
310
30
29
2.8
2.7
swainsonine analogues was syn-
thesized from amine acetal 1,
which was obtained from d-(À)-
isoascorbic acid through the
Mannich reaction[10] as shown in
Scheme 1. Their inhibitory po-
tency (IC50) was tested on jack-
bean a-mannosidase, Drosophila
lysosomal mannosidase, and the
recombinant catalytic domain of
Golgi a-mannosidase II from
Drosophila (dGMII) (Table 1). For
swainsonine and the (5a)-car-
bonyl-containing analogues 2–4,
the IC50 values measured under
standard conditions against
dGMII were virtually identical,
ranging from 30–40 nm. These
compounds are all poorer inhibi-
tors of the lysosomal enzyme of
Drosophila or jack bean with IC50
3
4
220
340
29
2.7
5
2500
2550
250
n.d.
6
7
420
930
n.d.
44
n.d.
n.d.
3700
250
Table 1) results in compounds with a dramatic loss of inhibitory
activity. The C(3)-substituted swainsonine analogues[9] were re-
ported to be more active than the parent compound against
jack-bean mannosidase (an enzyme with lysosomal-like proper-
ties). However, their potency was seen to be considerably
worse than that of swainsonine towards GMII from Drosophila
melanogaster (dGMII) (D.A.K., L. Guo, W. Pearson, D.R.R., un-
published observation). We have previously reported on C(5)-
substituted analogues of swainsonine and showed that the in-
hibitory activity of the (5a)-substituted swainsonine analogues
was similar to the activity of swainsonine towards jack-bean
mannosidase.[10] As reported here, the (5a)-substituted swain-
sonine analogues are almost indistinguishable from swainso-
nine in their ability to inhibit dGMII, and so provide a promis-
ing starting point from which to extend swainsonine to exploit
other binding groups within the GMII active site. Knowledge of
the spatial positioning of and interactions formed with the 5-
substituted swainsonine analogues is invaluable in the design
of future analogues. We have utilized X-ray crystallography to
determine the high-resolution structures of dGMII with both
(5a)- and (5b)-substituted swainsonine analogues. The aromat-
ic substituents are found in a region of the active site pocket
not occupied in previous dGMII structures. Both the (5a) and
(5b) analogues are found in a similar position. A well-organized
values ranging from 220–~500 nm. Under these conditions the
selectivity for the Golgi enzyme is thus ~9–17-fold; this selec-
tivity is encouraging but inadequate for the prevention of
long-term side effects. The C(5b) analogues were six- to eight-
fold worse than the corresponding C(5a) analogues but had
similar selectivity (10–15ꢁ) in comparison to the jack-bean
enzyme.
The standard conditions of the assay to assess IC50 were
identical to those previously used and utilize commercially
available para-nitrophenyl a-d-mannopyranoside (pNP-man-
nose) as the substrate with an enzyme concentration of 40 nm
for 45 min. At these concentrations the IC50 of the swainsonine
analogues is similar to the enzyme concentration and it is
likely that a considerable proportion of the inhibitor will be
bound to the enzyme. Thus, to determine Ki values, reactions
were carried out at 258C with 80-fold less enzyme (0.5 nm) in
the presence of the stabilizing agents, 50 mgmLÀ1 bovine
serum albumin, 0.1% Triton X-100 and 10 mm ZnSO4. However,
the stabilizing agents were insufficient in preventing loss of
activity during the long time course of the reaction that was
necessitated by the slow turnover rate with PNP-mannose. 2,4-
dinitrophenyl a-d-mannopyranoside (DNP-mannose) proved to
be a much better substrate and allowed us to measure the re-
action directly within an hour. We used DNP-mannose to de-
674
www.chembiochem.org
ꢀ 2010 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim
ChemBioChem 2010, 11, 673 – 680

5-Substituted Swainsonine Analogues Bound to GMII
lowest Ki and the best-defined
electron density. It is therefore
the best starting point for the
design of future compounds.
The dGMII:complex structures
were superimposed with the
high resolution (1.3 ꢂ) structure
of dGMII bound to swainsonine
[PDB ID: 3BLB]. The active site
amino acids overlap very closely
(not shown) as do the bound in-
hibitors (especially the C(10) car-
bonyl-containing inhibitors 2–4;
Figure 2A). The compounds lack-
ing the C(10)-carbonyl substitu-
ent (6 and 7) have the largest
difference (Figure 2B and C). The
presence of the C(10) carbonyl
in the b-oriented inhibitor 5 re-
sults in an orientation quite simi-
lar to that of 4 (Figure 2C). Dis-
Scheme 1. Synthesis of 5-substituted swainsonine analogues.
tances for the interactions made
by the compounds are listed in
termine the Ki of the dGMII with the three best inhibitors (2–4)
and with swainsonine. Again, for all these compounds, almost
identical inhibition constants were obtained and were in the
range of 2.5 to 3 nm (Table 1), the lowest values so far reported
with GMII.
Table S2 and they are all very comparable with the interactions
made by swainsonine. The swainsonine moieties of the C(5b)-
substituted compounds are displaced slightly from the parent
The aromatic C(5) substituents occupy similar positions
The binding of the C(5)-substituted compounds to the active
site of dGMII was studied by using X-ray diffraction data col-
lected at the CHESS synchrotron. High-resolution data from
cocrystals of 2–7 with dGMII were obtained, which allowed the
visualization of the electron density for the bound compounds.
Statistics for data collection and refinement are shown in
Table S1 in the Supporting Information. The electron density
for the bound compounds is shown in Figures 1 and S1. High-
quality, unambiguous density can be seen for compounds 2–4,
which are C(5a) substituted and contain a carbonyl group at
C(10). For these compounds the Fo–Fc omit map can be visual-
ized clearly when the contour level is set to 6s. For com-
pounds 5–7 the electron density for the complete molecules
only becomes clear when the Fo–Fc omit maps are plotted at
3s. The electron density for the phenyl ring of 6, which lacks
the C(10) carbonyl, is much more poorly resolved than that of
4, and this indicates the importance of the C(10) carbonyl in
anchoring the aromatic moiety. The average temperature fac-
tors (B-factors) are also shown for both the swainsonine
moiety and the C(5) substituent. The B-factors are an indication
of molecular movement, with lower B-factors denoting more
rigid atoms. It can be seen in all cases that the B-factors of the
multiply hydrogen-bonded swainsonine moiety are lower than
those of the aromatic ring, and the aromatic rings of 6 and the
C5b analogues have the highest B factors. In the following dis-
cussion we will focus on the dGMII:4 because it had the
Figure 1. Electron density in the active site of the dGMII complexes. Simulat-
ed annealing (to 3000 K) Fo-Fc omit maps were calculated after removal of
the compound and zinc from the final model. Electron density is shown con-
toured at 6s for A–C, or at 3s for the poorly bound compounds (D–F). Car-
bons are colored white, oxygens grey, and nitrogens black. Zinc is shown as
a large sphere. Average temperature factors for the swainsonine moiety and
the phenyl moiety are indicated to the left of each compound. A) 2; B) 3;
C) 4; D) 5; E) 6; and F) 7.
ChemBioChem 2010, 11, 673 – 680
ꢀ 2010 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim
www.chembiochem.org
675

H. Nagasawa, D. R. Rose et al.
These previously investigated compounds include benzyl-man-
nostatin (SK2, from PDB ID: 2F7P),[11] an N-aryl carbamate de-
rivative of gluco-hydroxyiminolactam (GHR, from PDB ID:
3D52),[12] a benzyl 5-thio-d-mannopyranosylamidinium deriva-
tive (LKS, from PDB ID: 1R34),[13] and
a
phenyl-
(ethyl)aminomethyl pyrrolidone derivative (GB6, from PDB ID:
3DDF).[14] In the case of SK2, GHR, and LKS the addition of the
phenyl moiety was markedly detrimental and reduced the in-
hibitory potency 7, 7.4, and 13-fold respectively in comparison
to the parental compounds.[11–13] (There is no unsubstituted
“parent” for GB6). This was in contrast with the C(5a)-substitut-
ed swainsonine analogues studied here, which had similar in-
hibitory potencies as the parent compound.
The overlay of the four previously solved structures with one
of the substituted swainsonine analogue complexes, (dGMII:4)
is shown in Figure 2D and Figure S5. It can be seen that the
phenyl groups of the other compounds are found in a position
completely different from what is observed in dGMII:4. The
aromatic moiety of GHR was found in two different locations,
one of which more closely resembles that of 4, but its very
weak electron density was in marked contrast to the well-de-
fined position of 4.
The dGMII:4 structure was also superposed with solved
structures containing bound sugars (Figures 2E and S6). These
included synthetic GlcNAc-Man4 (from PDB ID: 3BVX),[15] and
natural derived GlcNAc-Man5-GlcNAc (from PDB ID: 3CZN).[4] It
is clear that the sugars in the +1 position do not match the
position of the phenyl group attached to C(5) of swainsonine.
In contrast, the other previously investigated phenyl-contain-
ing inhibitors reside in a region of the active site that is much
closer to that where the +1 sugar is found, as can be seen
from Figure 2D.
Figure 2. Comparison of bound 4 with new and previously solved bound in-
hibitors and substrates. Coordinates of the dGMII complexes were superim-
posed on the dGMII:swainsonine structure. The carbon atoms for 4 are col-
ored green, and the active site zinc is shown as a cyan sphere. A) The bind-
ing of 4 is compared with swainsonine (yellow), 2 (magenta), and 3
(orange). B) Comparison of the binding of C(5a) compounds with (4) or
without (6, light blue) the C(10) carbonyl. C) Comparison of C(5a)- (4) and
C(5b)-containing compounds (5, salmon) and (7, brown). D) Comparison
with benzyl-substituted inhibitors, benzyl-mannostatin (SK2, grey), GHR
(plum, both conformers of the aromatic substituent are shown), LKS
(yellow), and GB6 (sky-blue). E) Comparison of 4 with substrate analogues;
GlcNAcMan5 (tan), GlcNAcMan4 (dark blue). The sugars are labeled in terms
of the various sites to which they bind on GMII; À1 is the cleavage site; the
+2’ site or “holding site” contains the a(1–3) linked mannose, which will be
cleaved in the second reaction after rotation of the +1 “swivel” sugar; the
+4 “anchor” site binds GlcNAc and is absent in lysosomal mannosidases.
In the study of other glycosyl hydrolase inhibitors addition
of an aromatic group often leads to dramatically improved
affinity, as stacking interactions are important at the +1 (agly-
cone) site (some examples are given in ref. [16]). In contrast,
from the results presented here and previously, it seems that
in the case of GMII the +1 site is not a good target. We have
previously shown that the +1 sugar is not tightly bound.[15]
This is a reflection of the mechanism of GMII cleavage in which
the a(1–3) linked mannose has to swing around the +1 sugar
(the “swivel” position) in order to reposition itself in the cleav-
compound; this might contribute to the lowered inhibitory po-
tential of these compounds.
The C(5) substituents occupy
a position not previously ob-
served to be occupied in other
complexes
We next compared the position
of the bound 5-substituted
swainsonine analogues with
other aromatic-substituted inhib-
itors of dGMII whose binding
was characterized by X-ray crys-
tallography.[11–14] Structures and
inhibitory activities of these in-
Figure 3. Previously studied compounds discussed in this paper. Compounds are identified by their PDB ligand
hibitors are shown in Figure 3. code, and their inhibitory activity against dGMII is indicated.
676
www.chembiochem.org
ꢀ 2010 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim
ChemBioChem 2010, 11, 673 – 680

5-Substituted Swainsonine Analogues Bound to GMII
age pocket following the initial cleavage of the a(1–6)-linked
mannose.[4,15,17] Thus, lack of interacting residues at the +1
position supports the observation that aromatic groups that
target this region do not demonstrate improved inhibitory
properties. The synthetic C(5) analogues produced here target
a unique region of GMII that would not be readily obvious
from modeling studies based on previous inhibitor or substrate
structures and appear to derive binding energy from previous-
ly unobserved interactions.
tates (Asp92, Asp204, Asp472), and the hydrogen-bonding net-
work at O8 that involves Trp95 NE1, Asp472 OD1, and Try727
OH. A new interaction not found in swainsonine is between
the C(10) carbonyl oxygen and Arg228, both directly and
through an intermediate water. Arg228 in turn interacts with
the catalytic nucleophile Asp204. The OD1 of Asp204 is 2.9 ꢂ
from both the NH2 and NE atoms of Arg228. Table S2 com-
pares the polar interaction distances observed in the com-
plexes of the six compounds presented here with those seen
in the dGMII:swainsonine structure. No significant differences
are seen except in the O10:Arg228 interaction.
Both polar and hydrophobic interactions mediate binding
Hydrophobic interactions seem to play an important role in
the binding of the swainsonine analogues, as there are numer-
ous close contacts (<4 ꢂ) between aromatic groups in the
active site of dGMII and the rings of the inhibitor (Figures 4B
and S3). The close contacts (3.5–3.6 ꢂ) with Phe206 should be
particularly strong. Of particular note is a new stacking interac-
tion between the phenyl ring of 4 with the backbone of the
protein at Arg876 with distances of 3.3–4 ꢂ (see below).
Figure 4A and B illustrate the principal polar and hydrophobic
interactions, respectively, made between dGMII and 4. Stereo-
views of the interactions are given in Figures S2 and S3. Of par-
ticular importance for binding of the swainsonine moiety are
the polar interaction between the positively charged nitrogen
(N4) and the OD2 oxygen of the catalytic nucleophile Asp204,
interactions of the O1 and O2 to the active site zinc and aspar-
Figure 4. Binding of 4 in the active site of dGMII. A) Polar interactions. Distances [ꢂ] are indicated in the top left corner. B) Hydrophobic interactions. C) and
D). A water cluster that occurs in the dGMII:swainsonine structure (PDB ID: 3BLB) is disrupted on binding of 4 . Waters are shown as small spheres. The aster-
isk denotes the position of a water molecule missing in the complex with 5–7 (see text).
ChemBioChem 2010, 11, 673 – 680
ꢀ 2010 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim
www.chembiochem.org
677

H. Nagasawa, D. R. Rose et al.
The C(5) substituents displace a well defined water cluster
Implications for inhibitor design: Bridging the gap between
the C5 substituents and the GlcNAc binding site
Figure 4C and D provide a side-by-side comparison of swainso-
nine and the C(5a) analogue 4 bound in the active site of
dGMII. A well defined water structure that occurs in the
dGMII:swainsonine structure (PDB ID: 3BLB), bridging the back-
bone carbonyl of Arg876 (Arg876O) and Arg228 (Figures 4C
and S4A), is displaced by the binding of the phenyl ring. Five
water molecules are displaced, including a central low B factor
water at the center of this cluster. Importantly, in the com-
plexes formed with the tight inhibitors (2, 3 and 4) this key
central water is replaced by the C(10)-carbonyl group, which
re-establishes hydrogen bonds with Arg228 and results in vir-
tually unchanged geometry in this region (shown with dGMII:4
in Figures 4D and S4B).
We have previously shown[18] that removal of another con-
served water molecule in the active site (indicated by an aster-
isk in Figure 4C and D) caused by the displacement of Arg228
results in considerable loss of inhibitor potency (PDB ID:
2OW6). In that case addition of a single hydroxyl group to 8-
epi-thiolentiginosine changed the IC50 from 14 mm to
2000 mm.[18] We had not previously observed this Arg228-medi-
ated loss of water in other dGMII:inhibitor complexes, and this
water is present in the complexes with 2–4. Interestingly, in
the complexes of dGMII with the three least potent inhibitors
(5–7) Arg228 is displaced to a significant extent and the critical
water is missing in the electron density (results not shown). In
the case of the b-substituted noncarbonyl-containing 7 the
movement of Arg228 is even greater (by 0.5 ꢂ) than was ob-
served in PDB: 2OW6.[18]
An important consideration in the creation of new Golgi a-
mannosidase inhibitors is the selectivity that they show for the
Golgi enzyme in comparison to their ability to inhibit the lyso-
somal enzyme. The effective small molecule inhibitors such as
swainsonine and mannostatin, as well as the C(5)-substituted
swainsonine analogues studied here, bind to residues in the
À1 site that are highly conserved between the lysosomal and
Golgi enzymes, and thus the limited amount of selectivity (ca.
tenfold) exhibited by these compounds is not surprising. GMII
is dependent on the prior action of N-acetylglucosaminyltrans-
ferase I (GnT1), which adds GlcNAc to Man5GlcNAc2 to create
GlcNAcMan5GlcNAc2 the substrate of GMII.[6] Lysosomal manno-
sidase, in contrast, cleaves the Man5 substrate but not the
GlcNAc-modified one.[15,20] Our previous structural work[4,15] has
shown the position of the GlcNAc binding site in GMII (the +4
site) is located within a loop of the protein that is absent in
the structure of the lysosomal enzyme.[21] In GMII GlcNAc forms
an important stacking interaction with Tyr267 and makes other
polar interactions. GlcNAc bound to the +4 site is quite well-
ordered in the electron density, and its presence on the man-
nose-containing substrates completely alters their binding pat-
tern in the active site.[4,15] The GlcNAc binding site thus pres-
ents an attractive region to target for developing highly selec-
tive inhibitors.
Ideally the distance between the À1 site and GlcNAc site
could be spanned without compromising the binding affinity
to the À1 site. The C(5a)-substituted analogues presented
here are a big step towards this goal as they extend the inhibi-
tor from the À1 site without significant loss of activity. In the
superposed structures, the terminal methyl group of 4 is only
0.7 ꢂ away from C2-OH of the mannose bound in the +2 site
of native-sugar complex (PDB:3CZN). A superposition of 4 with
the terminal residues GlcNAc-Man4 (from PDB:3BVX) is shown
in Figures 5 and S7. In this case the branching carbon (C17) is
0.9 ꢂ away from the O5 of the +2 sugar, while the terminal
methyl (C20) is 1.5 ꢂ from C1. It is easy to imagine extension
of the C5 analogues by addition of a short linker, a mannose
and GlcNAc to create an inhibitor with the desired properties
of potency and selectivity. We have noted previously that the
+2 sugar-binding site is quite weak and does not seem to con-
tribute significantly to the substrate binding. The replacement
of the +2 mannose by a very short linker to the +3 mannose
should have no detrimental effect on the binding. The +3
mannose is slightly better defined in the electron density but
does not appear to make many contacts with the enzyme. It
should also be possible to replace this with an extended acyl
chain in a novel multisite inhibitor.
The C(5)a substituents form new hydrophobic interactions
with Arg876
The backbone carbonyl of Arg876 (Arg876O) has been shown
to form hydrogen bonds with the C(6)OH of the natural sub-
strate as well as many inhibitors including LKS, GB6, and
GHR.[4,13–15,18] Further, in a study of mannostatin analogues,
interaction with the Arg876O appeared to be a key compo-
nent.[19] In that study, removal of the thiomethyl group of man-
nostatin A (or a hydroxyl group in an analogue) to which
Arg876O formed hydrogen bonds, resulted in a greater than
1000-fold reduction in inhibitor potency even though other
interactions remained the same.[19] In contrast, Arg876O does
not seem to contribute greatly to the potency of swainsonine,
the binding of which appears to be facilitated by the consider-
able number of hydrophobic interactions with the region of
the cleavage pocket formed by Trp95, Phe206, Trp415, and
Tyr727. In the binding of 4, the displacement of the water clus-
ter by the C(5) substituent results in the formation of a new
hydrophobic region, and the phenyl ring stacks against the
backbone of the protein chain around Arg876O, creating new,
additional, hydrophobic interactions (Figures 4B, D and S3).
Conclusions
We have determined the mode of binding of a new class of
substituted swainsonine analogues, the phenyl substituents of
which displace a well-defined water cluster and form new
stacking interactions with the backbone of GMII. This previous-
678
www.chembiochem.org
ꢀ 2010 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim
ChemBioChem 2010, 11, 673 – 680

5-Substituted Swainsonine Analogues Bound to GMII
and lyophilized. The residue was purified by preparative reversed-
phase HPLC (C18, acetonitrile/water gradient elution system). The
solution of 4 (29.7 mg, 0.086 mmol) in MeOH (6 mL) was stirred at
room temperature for 3 h and then evaporated. The residue was
purified by reversed-phase HPLC (C18, acetonitrile/water gradient
elution system) to afford 5 (21.4 mg, 72%).
(1S,2R,5S,8R,8aR)-5-(2-(4-tert-Butylphenyl)ethyl]octahydroindoli-
zine-1,2,8-triol [(5S)-5-(4-tert-butylphenethyl)-swainsonine] (6):
10% Pd-C (10 mg, 26 mol%) was added to a solution of 4
(24.2 mg, 0.070 mmol) in ethanolic HCl (2.6 mL of 50 mm). The
flask was evacuated by aspirator and purged with hydrogen three
times. The reaction mixture was heated at 508C. After 22 h, the hy-
drogen was evacuated and the mixture was filtered through Celite
and rinsed with MeOH. The solution was then concentrated. The
residue was applied to an ion exchange column (Dowex 1X8 200
OHÀ), which was eluted with water. The solution was freeze-dried.
The residue was purified by using reversed-phase HPLC (C18, ace-
tonitrile/water gradient elution system) to afford 6 (8.1 mg, 33%)
Figure 5. Extension of 4 should allow spanning of the catalytic site and the
specificity determining GlcNAc-binding site. The dGMII:4 (light grey) struc-
ture was superposed with the dGMII:GlcNAcMan4 structure from PDB: 3BVX.
Only the terminal sugars of GlcNAcMan4 are shown (dark grey). Tyr267
forms an important stacking interaction with GlcNAc.
1
as a white solid.: H NMR (400 MHz, CD3OD): d=7.93 (m, 2H), 7.56
(m, 2H), 4.17 (m, 2H), 3.80 (m, 1H), 3.60 (m, 1H), 3.04 (dd, J=8.2,
15.5 Hz, 1H), 2.87 (m, 1H), 2.75 (dd, J=3.5, 10.4 Hz, 1H), 2.45 (m,
1H), 1.79 (m, 2H), 1.52 (m, 2H), 1.34 (s, 9H), 1.34 (m, 1H); IR (KBr):
3415, 2950, 2864, 1513, 1459, 1130, 1087, 1032, 838 cmÀ1; EI-HRMS
m/z calcd for C20H31NO3 [M]+ 333.2304, found 333.2307.
ly unobserved binding mode provides a direct pathway be-
tween the cleavage site and specificity determining GlcNAc-
binding site, and could lead to the synthesis of selective GMII
inhibitors.
(1S,2R,5R,8R,8aR)-5-(2-(4-tert-Butylphenyl)ethyl]octahydroindoli-
zine-1,2,8-triol [(5R)-5-(4-tert-butylphenethyl)-swainsonine] (7)):
10% Pd-C (9.0 mg, 26 mol%) was added to a solution of 5
(21.4 mg, 0.062 mmol) in ethanolic HCl (50 mm, 3.8 mL). Similarly,
in a hydrogen atmosphere, the reaction mixture was heated at
508C for 9 h. After the reaction, the reaction solution was treated
in the same way as mentioned above to afford 7 (7.1 mg, 38%) as
Experimental Section
Substrates and inhibitors: PNP-mannose was purchased from
Sigma, and DNP-mannose was a kind gift of Steve Withers (Univer-
sity of British Columbia). Swainsonine was purchased from Toronto
Research Chemicals (Toronto, ON). Stock concentrations of inhibi-
tors were made up in DMSO and stored at À708C.
1
a white solid. H NMR (400 MHz, CD3OD): d=7.29 (d, J=8.3, 2H),
7.09 (d, J=8.3 Hz, 2H), 4.18 (m, 2H), 3.79 (m, 1H), 3.05 (dd, J=1.7,
10.2 Hz, 1H), 2.69 (m, 1H), 248 (m, 1H), 2.38 (m, 1H), 2.06 (m, 1H),
1.89 (m, 3H), 1.80 (m, 1H), 1.57 (m, 1H), 1.29 (s, 9H), 1.27 (m, 2H);
IR (KBr): 3369, 2939, 2864, 2818, 2776, 1518, 1457, 1368, 1131,
1082, 1040, 823 cmÀ1 ; EI-HRMS m/z calcd for C20H31NO3 [M]+
333.2304, found 333.2308.
Chemistry: All reactions were conducted under an atmosphere of
dry nitrogen. Thin-layer chromatography (TLC) was carried out on
Merck silica gel 60 F254 plates and detected by UV light, iodine
vapor, and phosphomolybdic acid solution. Flash column chroma-
tography separations were carried out by using Merck Kieselgel 60
silica gel (0.063–0.200 mm) or Kanto Chemical silica gel 60 N (40–
50 mm, spherical, neutral). Reagents and solvents were purchased
Enzymes: Jack-bean a-mannosidase was purchased from Sigma.
Recombinant Drosophila lysosomal a-mannosidase (CG6206 gene
product) was purified as described previously.[12] dGMII purification
was performed essentially as described previously[17] with an added
anion-exchange step to further purify the enzyme. Briefly, the
Golgi lumenal located catalytic domain of Drosophila GMII (resi-
dues 76–1108) containing an N-terminal His6 tag was expressed in
a secreted form in Drosophila S2 cells. The medium was batch
bound to Blue F3GA agarose (Sigma). dGMII was eluted with NaCl
(0.35m) and directly batch bound to Ni-NTA agarose (Qiagen) from
which it was eluted with imidazole (30 mm). After dialysis the pro-
tein was further purified by salt elution from a MonoQ anion-
exchange column (GE Biosciences), prior to buffer exchange, con-
centration, and freezing in liquid N2.
1
from common suppliers. H NMR and 13C NMR spectra were record-
ed on a JNM-EX-400 (400 MHz) spectrometer. Chemical shifts (d,
ppm) were determined by using TMS as internal standard. MS
spectra were recorded on JEOL JMS-AX-500 and JEOL JMS-DX-303
mass spectrometers. Elemental analysis for C, H, and N were per-
formed on a Yanaco CHN CORDER MT-5 analyzer (Kyoto, Japan).
High-performance liquid chromatography (HPLC) was performed
on Shiseido fine chemicals C18 column (Tokyo, Japan) by using a
JASCO PU-1586 pump equipped with a UV/Vis detector JASCO UV-
1570.
General synthetic procedure for swainsonine analogues via the
Mannich reaction: The 5-substituted swainsonine analogues 2–4
are synthesized by the Mannich reaction of amine acetal 1 with
the corresponding ketone according to the method reported previ-
Inhibition assays: Inhibition assays with jack-bean mannosidase
were performed at pH 5 in acetate buffer essentially as outlined by
Dennis et al.,[8] while assays on Drosophila lysosomal mannosidase
were carried out at pH 4.5. Inhibition of dGMII was measured at
pH 5.75.[13] Determination of the IC50 values (concentrations of in-
hibitor at which 50% of activity remains) was carried out with pNP-
mannose (4 mm) and enzyme (40 nm) as described previously.[22]
ously.[10] Briefly,
a solution of the amine acetal 1 (10 mg,
0.04 mmol) in EtOH/acetonitrile (1:4; 500 mL) was added to a solu-
tion of EtOH/acetonitrile (1:4, 500 mL) containing excess ketone
(250–500 mL) and a few drops of conc. HCl. The mixture was stirred
for 3 h under reflux. After the reaction, the mixture was evaporated
and dissolved in water. The aqueous phase was washed with Et2O
The Ki values for the tight binding inhibitors (swainsonine and 2–4)
were assessed in microtiter plates with dGMII (0.5 nm) in a buffer
ChemBioChem 2010, 11, 673 – 680
ꢀ 2010 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim
www.chembiochem.org
679

H. Nagasawa, D. R. Rose et al.
Gerber-Lemaire, L. Juillerat-Jeanneret, J. Med. Chem. 2005, 48, 4237–
4246; c) S. Favre, H. Fiaux, C. Schuetz, P. Vogel, L. Juillerat-Jeanneret, S.
Gerber-Lemaire, Heterocycles 2006, 69, 179–192.
containing MES (40 mm, pH 5.75), bovine serum albumin
(50 mgmLÀ1), zinc (10 mm), Triton X100 (0.1%), phosphate (1 mm),
NaCl (10 mm), and DMSO (10%). Final reaction volume was 80 mL
and the reaction was carried out at 258C. Due to the low enzyme
concentration, DNP-mannose was used as a substrate because it
has a much faster turnover rate than PNP-mannose. The reaction
was followed directly by monitoring the change in absorption at
[4] N. Shah, D. A. Kuntz, D. R. Rose, Proc. Natl. Acad. Sci. USA 2008, 105,
9570–9575.
[5] S. Numao, D. A. Kuntz, S. G. Withers, D. R. Rose, J. Biol. Chem. 2003, 278,
48074–48083.
[6] N. Harpaz, H. Schachter, J. Biol. Chem. 1980, 255, 4894–4902.
[7] a) H. P. Broquist, Annu. Rev. Nutr. 1985, 5, 391–409; b) D. R. Tulsiani, H. P.
Broquist, L. F. James, O. Touster, Arch. Biochem. Biophys. 1988, 264, 607–
617; c) A. C. Crawley, S. U. Walkley, J. Neuropathol. Exp. Neurol. 2007, 66,
687–697.
[8] J. W. Dennis, S. L. White, A. M. Freer, D. Dime, Biochem. Pharmacol.
1993, 46, 1459–1466.
[9] W. H. Pearson, L. Guo, Tetrahedron Lett. 2001, 42, 8267–8271.
[10] T. Fujita, H. Nagaswa, Y. Uto, T. Hashimoto, Y. Asakawa, H. Hori, Org. Lett.
2004, 6, 827–830.
[11] S. P. Kawatkar, D. A. Kuntz, R. J. Woods, D. R. Rose, G. J. Boons, J. Am.
Chem. Soc. 2006, 128, 8310–8319.
[12] D. A. Kuntz, C. A. Tarling, S. G. Withers, D. R. Rose, Biochemistry 2008, 47,
10058–10068.
[13] L. M. Kavlekar, D. A. Kuntz, X. Wen, B. D. Johnston, B. Svensson, D. R.
Rose, B. M. Pinto, Tetrahedron: Asymmetry 2005, 16, 1035–1046.
[14] H. Fiaux, D. A. Kuntz, R. C. Janzer, S. Gerber-Lemaire, D. R. Rose, L. Juil-
lerat-Jeannere, Bioorg. Med. Chem. 2008, 16, 7337–7346.
[15] W. Zhong, D. A. Kuntz, B. Ember, H. Singh, K. W. Moremen, D. R. Rose,
G.-J. Boons, J. Am. Chem. Soc. 2008, 130, 8975–8983.
[16] a) T. Granier, N. Panday, A. Vasella, Helv. Chim. Acta 1997, 80, 979–987;
b) L. Gartenmann Dickson, L. Emmanuel, J.-L. Reymond, Org. Biomol.
Chem. 2004, 2, 1217–1226; c) T. M. Gloster, J. M. Macdonald, C. A. Tarl-
ing, R. V. Stick, S. G. Withers, G. J. Davies, J. Biol. Chem. 2004, 279,
49236–49242; d) H. C. Dorfmueller, V. S. Borodkin, M. Schimpl, S. M.
Shepherd, N. A. Shpiro, D. M. van Aalten, J. Am. Chem. Soc. 2006, 128,
16484–16485.
405 nm in
a microplate reader. DNP-mannose concentrations
ranged from 2–30 mm, and inhibitor concentrations ranged from
1–15 nm. The program GraFit 4.0.21 (Erithacus Software, Horley,
UK) was used to calculate apparent Ki and standard deviations.
This program employs nonlinear fitting of the data for each inhibi-
tor concentration to the Michaelis–Menten equation.
Structure determination: Crystallization, data collection, and re-
finement of the complexes were essentially performed as outlined
in detail previously.[14,15] All crystals were prepared as cocrystals
and all data was collected on beamlines A1 and F1 at the Cornell
High Energy Synchrotron Source (CHESS, Ithaca, NY, USA). Model
building was carried out in Coot.[23] Final refinement was with
Refmac5[24] as implemented in CCP4.[25] Simulated annealing omit
maps were generated in CNS.[26] Both zinc and the inhibitor were
removed from the final model, which was subject to simulated an-
nealing to 3000 K. Graphics were generated with Molscript[27] and
Conscript[28] and rendered with Pymol.[29]
For the comparisons of substrate-binding positions, each of the in-
dependently refined structures was superposed to the 1.3 ꢂ struc-
ture of swainsonine bound to dGMII (PDB ID: 3BLB; the reference
structure). To prevent bias, an overlay of the Ca backbone of the
entire protein was used as the fitting criteria. The superposition
was carried out by using the SSM subroutine[30] of Coot and the co-
ordinates of the inhibitors extracted.
[17] J. M. H. van den Elsen, D. A. Kuntz, D. R. Rose, EMBO J. 2001, 20, 3008-
3017,.
[18] N. S. Kumar, D. A. Kuntz, X. Wen, B. M. Pinto, D. R. Rose, Proteins Struct.
Funct. Bioinf. 2008, 71, 1484–1496.
[19] D. A. Kuntz, W. Zhong, J. Guo, D. R. Rose, G.-J. Boons, ChemBioChem
2009, 10, 268–277.
Structures and structure factors were deposited in the Protein Data
Bank with the accession numbers 3EJP, 3EJQ, 3EJR, 3EJT, 3EJU, 3EJS
for the complexes of dGMII with 2–7, respectively.
[20] C. Park, L. Meng, L. H. Stanton, R. E. Collins, S. W. Mast, Yi, H. Strachan,
K. W. Moremen, J. Biol. Chem. 2005, 280, 37204–37216.
[21] P. Heikinheimo, R. Helland, H.-K. S. Leiros, I. Leiros, S. Karlsen, G. Evjen,
R. Ravelli, G. Schoehn, R. Ruigrok, O. K. Tollersrud, S. McSweeney, E.
Hough, J. Mol. Biol. 2003, 327, 631–644.
[22] D. A. Kuntz, A. Ghavami, B. D. Johnston, B. M. Pinto, D. R. Rose, Tetrahe-
dron: Asymmetry 2005, 16, 25–32.
[23] P. Emsley, K. Cowtan, Acta Crystallogr. Sect. D Biol. Crystallogr. 2004, 60,
2126–2132.
[24] G. N. Murshudov, A. A. Vagin, E. J. Dodson, Acta Crystallogr. Sect. D Biol.
Crystallogr. 1997, 53, 240–255.
[25] Collaborative Computational Project, Number 4, Acta Crystallogr. Sect. D
Biol. Crystallogr. 1994, 50, 760–763.
[26] A. T. Brꢃnger, P. D. Adams, G. M. Clore, W. L. DeLano, P. Gros, R. W.
Grosse-Kunstleve, J.-S. Jiang, J. Kuszewski, M. Nilges, N. S. Pannu, R. J.
Read, L. M. Rice, T. Simonson, G. L. Warren, Acta Crystallogr. Sect. D Biol.
Crystallogr. 1998, 54, 905–921.
Acknowledgements
We thank Hongming Chen (Steve Withers’ laboratory) for provid-
ing DNP-mannose. This work is based upon research conducted
at the Cornell High Energy Synchrotron Source (CHESS), which is
supported by the National Science Foundation under award
DMR-0225180, using the Macromolecular Diffraction at CHESS
(MacCHESS) facility, which is supported by award RR-01646 from
the National Institutes of Health, through its National Center for
Research Resources. D.R.R. holds funding from the Canadian Insti-
tutes for Health Research (MOP79312) and the Mizutani Founda-
tion (080032).
[27] P. J. Kraulis, J. Appl. Crystallogr. 1991, 24, 946–960.
[28] M. C. Lawrence, P. Bouke, J. Appl. Crystallogr. 2000, 33, 990–991.
[29] W. L. DeLano, The PyMol Molecular Graphics System, DeLano Scientific,
Palo Alto, CA, 2008.
Keywords: antitumor agents · aza sugar · glycosidases ·
inhibitors · X-ray diffraction
[30] E. Krissinel, K. Henrick, Acta Crystallogr. Sect. D Biol. Crystallogr. 2004, 60,
2256–2268.
[1] a) R. J. Bernacki, M. J. Niedbala, W. Korytnyk, Cancer Metastasis Rev.
1985, 4, 81–101; b) J. W. Dennis, M. Granovsky, C. E. Warren, Biochim.
Biophys. Acta Gen. Subj. 1999, 1473, 21–34; c) S. Hakomori, Proc. Natl.
Acad. Sci. USA 2002, 99, 10231–10233.
[2] D. H. Dube, C. R. Bertozzi, Nat. Rev. Drug Discovery 2005, 4, 477–488.
[3] a) P. E. Goss, C. L. Reid, D. Bailey, J. W. Dennis, Clin. Cancer Res. 1997, 3,
1077–1086; b) H. Fiaux, F. Popowycz, S. Favre, C. Schuetz, P. Vogel, S.
Received: December 9, 2009
Published online on March 5, 2010
680
www.chembiochem.org
ꢀ 2010 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim
ChemBioChem 2010, 11, 673 – 680