Enantiospecific synthesis of 1-azafagomine

Article abstract of DOI:10.1002/(SICI)1521-3765(20000117)6:2<278::AID-CHEM278>3.0.CO;2-6

For the first time the two enantiomeric forms of the glycosidase inhibitor 1-azafagomine have been synthesised starting from D- and L-xylose. D-Xylose was converted to the 2,3,5-tribenzylfuranose, which upon reductive amination with tert-butyl carbazate gave the protected 1-hydrazino-1- deoxypentitol in high yield. N-acetylation, mesylation of the 4-OH, removal of the Boc group, cyclisation and deprotection gave (+)-1-azafagomine ((+)- 1). By a similar sequence of reactions, L-xylose was converted to (-)-1- azafagomine ((-)1). Enzymatic and other routes to optically pure 1- azafagomine were also studied. Compound (-)-1 is a potent competitive glycosidase inhibitor, while (+)-1 has no biological activity. The inhibition of almond β-glucosidase by (-)-1 was found to be slow owing to a slow binding step of inhibitor to enzyme, with no subsequent conformational rearrangement. The rate constants for binding and release were found to be 3.3 x 10⁴ M^-1 s^-1 and 0.011 s^-1, respectively, yielding K(i) = 0.33 μM.

Full text of DOI:10.1002/(SICI)1521-3765(20000117)6:2<278::AID-CHEM278>3.0.CO;2-6

FULL PAPER
1-Azafagomine, Part 2[=]
Enantiospecific Synthesis of 1-Azafagomine
Bettina V. Ernholt,[a] Ib B. Thomsen,[a] Anders Lohse,[a] Igor W. Plesner,[a] Kenneth B.
Jensen,[b] Rita G. Hazell,[a] Xifu Liang,[a] Astrid Jakobsen,[a] and Mikael Bols*[a]
È
Dedicated to Professor Pierre Sinay on the occasion of his 60th birthday
Abstract: For the first time the two
enantiomeric forms of the glycosidase
inhibitor 1-azafagomine have been syn-
thesised starting from d- and l-xylose.
d-Xylose was converted to the 2,3,5-
tribenzylfuranose, which upon reductive
amination with tert-butyl carbazate gave
the protected 1-hydrazino-1-deoxypen-
titol in high yield. N-acetylation, mesy-
lation of the 4-OH, removal of the Boc
group, cyclisation and deprotection gave
(‡)-1-azafagomine ((‡)-1). By a similar
competitive glycosidase inhibitor, while
(‡)-1 has no biological activity. The
inhibition of almond b-glucosidase by
( )-1 was found to be slow owing to a
slow binding step of inhibitor to enzyme,
with no subsequent conformational re-
arrangement. The rate constants for
binding and release were found to be
sequence of reactions, l-xylose was
converted to ( )-1-azafagomine (( )-
1). Enzymatic and other routes to opti-
cally pure 1-azafagomine were also
studied. Compound ( )-1 is a potent
Keywords: azasugars ´ enzyme in-
hibitors ´ glycosidases ´ iminosugars
´ lipases
1
3.3 Â 104 m 1 s 1 and 0.011 s , respective-
ly, yielding Ki ˆ 0.33 mm.
For some time it has been known that replacing the ring
oxygen of a saccharide with nitrogen creates a glycosidase
inhibitor.[5] Subsequently it was found that replacing the
anomeric carbon of a saccharide with nitrogen also creates a
powerful glycosidase inhibitor.[6] The latter type of inhibitor,
isofagomines, were predominantly b-glycosidase inhibitors,[7]
while the former type, nojirimycins, more predominantly
inhibited a-glycosidases. An explanation for this has been
suggested.[8]
It was recently found that the hybrid compound (Æ)-1-
azafagomine ((Æ)-1) is a strong inhibitor of both b-glucosi-
dase and a-glucosidase (Figure 1).[8] It was anticipated that
the active enantiomer was the one resembling d-glucose and
this then supported the hypothesis given in reference[8]. It
may be argued, however, that since l-nojirimycin is an
inhibitor[9] of glycosidases the racemic 1 gives an unclear
picture in such a study, because some of the activity may be
Introduction
A great number of biochemical processes in human biology
involve glycoside-cleaving enzymes.[1] There is therefore a
considerable drug potential in selective inhibitors of glyco-
sidases and similar enzymes. So far glycosidase inhibitors have
been investigated as antidiabetic, antiviral or antimetastatic
agents, work which has resulted in the antidiabetes drug
acarbose[2] and the new antiinfluenza drug Zanamivir.[3] There
is, however, a far greater potential in inhibition of these
enzymes, provided the required potent, selective inhibitors
can be designed and synthesised. An exciting development is
the increased understanding of the elements required for
glycosidase inhibition that is now being obtained,[4] which may
allow future chemists to predict inhibitor structure to a
greater extent.
[a] Dr. M. Bols, B. V. Ernholt, A. Jakobsen, I. B. Thomsen, A. Lohse,
X. Liang, Dr. I. Plesner, Dr. R. G. Hazell
Department of Chemistry, Aarhus University
DK-8000 Aarhus C (Denmark)
OH
OH
OH
H
H
OH
O
N
N
O
OH
OH
OH
OH
NH
NH
OH
OH
Fax : (‡ 45)8619-6199
HO
HO
HO
HO
E-mail: mb@kemi.aau.dk
O
HO
OH
OH
OH
OH
[b] Dr. K. B. Jensen
D-glucose
(-)-1
D-xylose
(+)-1
Department of Chemistry, Odense University
DK-5230 Odense M (Denmark)
[=] Part 1: Ref. [8].
Figure 1. The stereochemical resemblance of 1-azafagomine (1) to glucose
and xylose.
278
ꢀ WILEY-VCH Verlag GmbH, D-69451 Weinheim, 2000
0947-6539/00/0602-0278 $ 17.50+.50/0
Chem. Eur. J. 2000, 6, No. 2

278±287
caused by the enantiomer resembling l-nojirimycin. It is
therefore important to investigate the individual activities of
the compounds, and in the present paper we describe the
synthesis of both enantiomers of 1 and the investigation of
their biological activities.
selectivity with this enzyme varied widely with solvent. The
highest E value (E ˆ enantiomeric ratio; that is the enzymeꢁs
selectivity towards one enantiomer over the other), 3.8 in
favor of conversion of the R enantiomer, was obtained in
dimethoxymethane, while in benzene and acetonitrile no
selectivity was observed. In chlorinated solvents the selectiv-
ity even reversed so that the S enantiomer was preferentially
converted. The most selective enzyme was the Humicola
Thermo lipase (HTL) which gave a selectivity of 5.6. The
Candida Antarctica (CAL, Novozym) and the Humicola
Langinosa (HLL) lipases were somewhat less efficient.
Aspergillus Niger (ANL) and Pseudomonas Cepacia (PCL)
lipases could be used to hydrolyse the octanoyl ester 4 to (R)-
2, the latter with a selectivity comparable to the mucor-
catalysed reaction. To summarise, the highest selectivity
obtained was 5.6 for the R enantiomer, and this was not
enough for the reaction to be useful for synthesis of the pure
enantiomers of 1.
Results and Discussion
Synthesis of optically pure 1 was attempted by three different
routes, with the third ultimately being successful. The first
approach took advantage of the attractive short synthesis of
racemic 1.[8] A key intermediate in this synthesis was the
alcohol (Æ)-2, which seemed an obvious candidate for
enzymatic resolution using lipases. This could be performed
in two ways, either by enzyme-catalysed esterification of 2 or
by enzymatic hydrolysis of an ester of 2 (Scheme 1). Half-a-
dozen enzymes were investigated for catalysis of both
reactions. In general the esterification process gave the better
conversion especially when carried out with vinyl acetate in
organic solvent. The hydrolysis was extremely slow when the
acetate 3[8] was the substrate, but somewhat better when the
octanoate 4 was used.
During the chromatographic purification of enantiomeri-
cally enriched batches of both 2 and 3, the phenomenon of
enantiomeric enrichment on an achiral column[10] was ob-
served. On chromatography of a 51% ee batch of acetate (R)-
3, the early fractions of the column contained a higher content
of R enantiomer (59% ee) and the late fractions had an ee of
38%.
O
R
The absolute configuration of the main product (R)-3 was
determined by converting a batch of 3, obtained by Mucor-
catalysed acetylation and having an enantiomeric excess of
50%, into 1-azafagomine (1) by deacetylation with NaOMe in
MeOH to 2 followed by the previously published three-step
procedure[8] of epoxidation, hydrolysis and hydrazinolysis.
The rotation of this product [a]2D5 was measured as 5.4
corresponding to 1 with an enantiomeric excess of ( )-1 of
54%. The rotation and configuration of pure ( )-1 was
determined by synthesis from l-xylose, as described later in
this paper.
O
HO
O
O
O
O
lipase
N
N
N
N
N
Ph
N
Ph
O
lipase
buffer
O
2
3 R = Me
4 R = C7H15
Scheme 1. Enzymatic resolution of (Æ)-2.
The results of the experiments yielding significant con-
version are shown in Table 1. The enantiomeric excess of the
products was determined by HPLC on an amylose acetate
column. The immobilised Mucor Mihei lipase (MML) from
NOVO (Bagsvaerd, Denmark; Lipozyme) gave some of the
best selectivity in catalysing esterification. Interestingly the
Since enzymatic resolution of 2 was too inefficient we
turned our attention to total synthesis of 1 from a carbohy-
drate. Inspection of the structure of d-xylose (Figure 1)
revealed that a nucleophilic substitution of the 4-OH with
Table 1. Enantioselectivity of the enzyme-catalysed acylation/deacylation reactions.
Enzyme[a]
Substrate
Major product
Solvent
Reactant
Conversion [%]
ee [%]
E
ˆ
CAL
CAL
CAL
CAL
MML
MML
MML
MML
MML
MML
MML
HTL
HLL
ANL
PCL
(Æ)-2
(Æ)-2
(Æ)-2
(Æ)-4
(Æ)-2
(Æ)-2
(Æ)-2
(Æ)-2
(Æ)-2
(Æ)-2
(Æ)-2
(Æ)-2
(Æ)-2
(Æ)-4
(Æ)-2
(Æ)-4
(R)-3
(R)-3
(R)-4
(R)-2
(R)-3
(R)-3
(R)-3
(S)-3
(S)-3
(Æ)-3
(Æ)-3
(R)-3
(R)-3
(R)-2
(S)-3
(R)-2
dioxane
AcOCH CH2
31
36
17
14
30
46
12
55
28
50
25
26
35
10
28
41
41
39
41
30
51
37
27
29
22
0
2.8
2.8
2.6
2.0
3.8
2.9
1.8
2.5
1.7
1
ˆ
ˆ
AcOCH CH2
dioxane
dioxane
DME
dioxane
MIBK
TCE
CHCl3
C6H6
MeCN
AcOCH CH2
C7H15COOMe
MeOH
ˆ
AcOCH CH2
ˆ
AcOCH CH2
ˆ
AcOCH CH2
ˆ
AcOCH CH2
ˆ
AcOCH CH2
ˆ
AcOCH CH2
ˆ
AcOCH CH2
0
1
ˆ
ˆ
AcOCH CH2
AcOCH CH2
aq. CHCl3
AcOCH CH2
AcOCH CH2
48
61
20
21
47
3.4
5.6
1.5
1.7
3.8
ˆ
ˆ
AcOCH CH2
H2O
ˆ
ˆ
AcOCH CH2
PCL
aq. CHCl3
H2O
[a] CAL: Candida Antarctica lipase; MML: Mucor Mihei lipase; HTL: Humicola Thermo lipase; HLL: Humicola Langinosa lipase; ANL: Aspergillus
Niger lipase; PCL: Pseudomonas Cepacia lipase.
Chem. Eur. J. 2000, 6, No. 2
ꢀ WILEY-VCH Verlag GmbH, D-69451 Weinheim, 2000
0947-6539/00/0602-0279 $ 17.50+.50/0
279

M. Bols et al.
FULL PAPER
hydrazine with inversion of stereochemistry followed by
reductive ring-closure would give (‡)-1. As (‡)-1 was
expected to be the less active enantiomer, preparation and
investigation of that stereomer would in all likelihood give the
most information about the inhibitory potency of the enan-
tiomers. Since d-xylose is also readily available, this plan was
attractive.
Our first plan for synthesis of (‡)-1 from d-xylose followed
this scheme: Reductive amination of the aldehyde of open-
chain xylose with a hydrazine derivative would lead to a
1-hydrazinopentitol. By tosylation of the primary alcohol
followed by base treatment a 4,5-epoxide would be formed
that upon nucleophilic ring-opening would give 1 with the
desired stereochemistry. Thus d-xylose was treated with
benzoylhydrazide in aqueous EtOH to give an adduct in
80% yield that turned out to be the cyclic hydrazide 5
(Scheme 2). The structure was confirmed by X-ray crystal
H
N
H2
O
O
OH
OH
BzNHNH2
Pt/C
NHBz
H2O/EtOH
80%
HO
OH
HO
H2O
80%
OH
OH
D-xylose
5
OH
OH OH
Ac2O
MeOH
NHBz
NHBz
N
N
H
HO
Ac
OH OH
OH OH
100%
Figure 2. X-ray structures of 5 and 6.
6
7
O
O
revealed that tetrols with xylo stereochemistry on monotosy-
lation of the primary alcohol undergo very rapid attack by the
d-OH to form a tetrahydrofuran.[12] Interestingly other polyol
tosylates with different stereochemistry such as the corre-
sponding arabino compounds do not undergo this cyclisation
as readily, and the monotosylates can be isolated without
problems.[13] Compounds with xylo stereochemistry have a
pronounced tendency to undergo this cyclisation. In this light
the problems with monosulfonylation of 7 are quite under-
standable. The 2-OH group could attack the initially formed
sulfonate ester, and indeed 13C NMR of the crude reaction
product showed signals at low field (d ˆ 80 ± 85) suggesting
that tetrahydrofurans had been formed.
NHBz
NHBz
AcOH
Me2C(OMe)2
TsOH
N
O
N
HO
Ac
Ac
O
O
H2O
55%
OH O
(2 steps)
9
8
Scheme 2. Synthesis based on direct functionalisation of xylose.
structure determination (Figure 2). The structure of 5 con-
tains two independent molecules, but they are practically
identical, with only small differences in torsion angles. All
bond lengths and angles are as could be expected. Compound
5 is probably formed by attack by the 4-OH on an initially
formed hydrazone. The formation of cyclic hydrazide 5 was
unexpected, but not unprecedented.[11] It did not impede the
synthesis, however, because 5 could be hydrogenated with
platinum catalyst to the acyclic hydrazine 6 in 80% yield. The
structure of 6 was also confirmed by X-ray crystallographic
structure analysis (Figure 2). Treatment of 6 with acetic
anhydride in methanol gave a quantitative yield of the N-
acetate 7. The structure of 7 was proposed based on fitting the
molecular ion in the MS spectrum and the appearance of an
acetyl group in the NMR spectrum.
According to the plan mentioned above, an epoxide had to
be introduced at C-4 and C-5 through the selective conversion
of the primary alcohol to a leaving group. Therefore selective
tosylation, mesylation and similar reactions were studied on
N-acetate 7, but in no case was it possible to isolate the desired
monosulfonylated product. In general the reactions seemed to
give several products, according to TLC and NMR, which
initially seemed surprising. However a search of the literature
To avoid tetrahydrofuran formation, we decided to protect
the 2-OH. This was done by treatment of 7 with dimethoxy-
propane and TsOH to give the diacetonide 8 followed by
selective hydrolysis of the primary acetonide with aqueous
acetic acid to form 9 in 55% yield from 7 (Scheme 2). The
structure of 8 was determined on the basis of a mass spectrum
showing the expected molecular weight, and four peaks from
d ˆ 25 ± 30 in the 13C NMR spectrum showing the presence of
two isopropylidene groups. The structure of 9 was proposed
based on the mass spectrum showing the expected molecular
1
weight, the H NMR spectrum showing only six protons at
d ˆ 1.3 and thus implying the existence of only one isopropy-
lidine group, and the fact that 9 was formed by selective
hydrolysis of 8. It may be argued that the isopropylidene
group could be located in a different position, but its position
was confirmed by subsequent transformations. The primary
alcohol was now converted to a leaving group by treatment of
280
ꢀ WILEY-VCH Verlag GmbH, D-69451 Weinheim, 2000
0947-6539/00/0602-0280 $ 17.50+.50/0
Chem. Eur. J. 2000, 6, No. 2

1-Azafagomine
278±287
9 with mesitylenesulfonyl chloride in pyridine, which gave 10
in 73% yield (Scheme 3). The presence of the sulfonate was
clearly seen from the NMR spectrum by the presence of a new
aryl group. When 10 was treated with base under relatively
mild conditions, K2CO3 in acetone or NaOMe in MeOH, with
O
O
NHBz TBDMSCl
NHBz
N
HO
N
Me SiO
tBu
2
Ac
Ac
imidazole, DMF
98%
HO
O
HO
O
9
13
Swern
97%
MsCl
pyridine
63%
O
O
O
NHBz
N
NHBz
ArSO2Cl
O
NHBz
N
Ac
Me SiO
tBu
N
2
tBuMe2SiO
OH
O
9
Ac
SO2
Ac
O
O
MsO
O
pyridine
25oC, 6 h
73%
10
15
14
Scheme 4. Functionalisation of diol 9.
Bz
H
N
N
NAc
NH
OH
HCl
NaOMe
77%
HO
HO
H2O
97%
ison of these results with the facile formation of 11 suggests
that the trans acetonide makes six-membered ring formation
very difficult. Nevertheless, formation of a six-membered ring
can occur when a trans-fused acetonide is present, but in those
cases the substitution occurred at a primary center and not a
to the acetonide.[14]
O
O
HO
11
Scheme 3. Synthesis of diazepine 12.
12
As an alternative, a reductive amidation ring-closure was
attempted.[15],[16] Alcohol 13 was converted to the ketone 15 in
97% yield by a Swern oxidation. Treatment of 15 with
hydrogen and palladium on carbon under high pressure and
temperature did not give conversion, however.
It was now clear that a major problem in the synthesis was
the competing attack from the 2-OH. It was therefore
necessary to protect this alcohol, but the presence of a trans
2,3-acetonide impeded the formation of a six-membered ring.
It was consequently necessary to develop a strategy where
noncyclic protection groups could be used. Our attention was
called to the benzylated xylose derivatives 16a and 17, both
available from d-xylose (Scheme 5).[17] In these molecules the
2-, 3- and 5-positions are already substituted with benzyl
groups.
Obviously direct disubstitution of the dimesylate 17 was
tempting, though five-membered ring formation could be
expected to be favored. Indeed, treatment of 17 with neat
hydrazine hydrate at 1008C led to a single product 18 that was
isolated in 59% yield (Scheme 5). The downfield 13C chemical
shifts of the ring carbon atoms clearly showed that a
pyrrolidine and not a pyridazine had been formed. Partic-
ularly the peaks at d ˆ 84.9 and 80.3 were inconsistent with a
six-membered ring. The MS spectrum showed a peak at 418
consistent with the proposed formula. To avoid pyrrolidine
formation it was necessary to control the protection of the two
hydrazine nitrogen atoms in a manner similar to the strategy
of Schemes 2 and 5. So in a similar way 16a was subjected to
reductive amination with tert-butyl carbazate in the presence
of NaCNBH3 and acetic acid (Scheme 6). This gave the
hydrazide 19a in 84% yield; the structural assignment relied
the intention of forming an epoxide, the diazepine 11 was
formed in good yield. Only one product was observed. The
NMR spectrum of 11 showed the presence of 3 ± 4 different
amide rotamers and were very difficult to interprete. The
structure determination therefore had to rely on the mass
spectrum and subsequent conversion of 11 into 12 by
hydrolysis with aqueous hydrochloric acid. Diazapine 12,
formed in 97% yield, was easily identified by its symmetry.
The formation of compound 11 could occur by direct
substitution of the mesitylenesulfonate by the amide anion.
However, it is likely that the intended epoxide intermediate is
formed in this reaction, but opens rapidly and selectively by
attack in the 5-position, for the following reasons: The 4,5-di-
O-mesylate of 9 (data not shown) was rather unreactive
towards base and no diazepine was formed when that reaction
was pushed towards completion; only elimination occurred.
The endo ring-opening of the epoxide is supposedly not so
favorable, but in its favor is the fact that the 5-position is
primary and that exo attack would lead to a six-membered
ring with a trans-fused acetonide. The diazapine 11 is probably
less strained.
To force the formation of the six-membered ring, we
protected the 5-OH of 9 with a tert-butyldimethylsilyl
(TBDMS) group by treatment of 9 with TBDMSCl and
imidazole in DMF (Scheme 4). This gave the silyl ether 13 in
98% yield, as witnessed by the appearance of a tertiary butyl
signal in the 1H NMR spectrum. This compound was
mesylated with mesyl chloride in pyridine to give the mesylate
14 in 63% yield. Structural assignment of 14 relied on the
1
appearance of a H NMR signal at d ˆ 3.2 corresponding to
the mesyl group, and the downfield shift of a single proton
from d ˆ 4.25 to 4.65, corre-
OBn
OMs OBn
OBn OBn OMs
17
NH2
N
OBn
sponding to the deshielding ef-
O
OH
NH2NH2
O
3 steps
2 steps
ref. [16]
OH
OBn
fect of the mesyl group on H-4.
However, treatment of 14 with
various bases (LDA, NaH,
KOtBu) did not lead to the
desired pyridazine. A compar-
100oC
59 %
ref. [16]
HO
OH
BnO
BnO
OBn
OH
16a
D-xylose
Scheme 5. Synthesis of pyrrolidine 18.
18
Chem. Eur. J. 2000, 6, No. 2
ꢀ WILEY-VCH Verlag GmbH, D-69451 Weinheim, 2000
0947-6539/00/0602-0281 $ 17.50+.50/0
281

M. Bols et al.
FULL PAPER
OH OBn
NaCNBH3
OBn
BnO
OH OBn Ac
N
H
N
Ac2O
O
OH
OH
1) NaCNBH3
O
3 steps
ref. [16]
OH
OBn
16b
16a
BocNHNH2
NHBoc
1) TFA
NHBoc
BocNHNH2
MeOH
OBn OBn
AcOH
OBn OBn
HO
2) Ac2O
3) MsCl
OH
19a
84 %
20a
L-xylose
OBn
2) EtNiPr2
MeNO2
H
N
OMs OBn Ac
N
Ac
MsCl
pyridine
92%
OH
N
H
NHBoc
1) TFA
2) EtNiPr2
OMs OBn Ac
N
N
60 oC,18 h
75%
NH
BnO
OBn OBn
NHBoc
OBn
HO
3) H2, Pd/C
4) HCl
OBn OBn
21a
OH
22a
OH
(-)-1
OH
21b
Scheme 7. Synthesis of ( )-1.
H
H
HCl, 100 oC
H2, Pd/C
N
Ac
N
N
NH
HCl, EtOH
18h, 64% (2 steps)
HO
HO
OH
OH
(+)-1
Table 2. Ki values [mm] for the inhibition of glycosidase by 1 and 12.
23a
Scheme 6. Synthesis of (‡)-1.
Enzyme
(‡)-1
(Æ)-1[8]
(
)-1
12
almond b-glucosidase
yeast a-glucosidase
isomaltase
> 10000
2900
0.65
3.9
1.06
0.32
6.9
0.27
63.7
42
31
on the disappearance, in the 13C NMR spectrum, of an
anomeric carbon and the appearance of two protons at d ˆ 3.2
and 3.1 in the 1H NMR spectrum, corresponding to the
reduced C-1. The secondary amine was now protected as the
N-acetate with acetic anhydride in methanol to give 20a, as
seen from the appearance of an N-acetyl group in the NMR
spectrum. This compound was treated with methanesulfonyl
chloride in pyridine to give 21a in 92% yield (from 19a). The
presence of the mesyl group was clearly seen by the
appearance of a mesyl signal at d ˆ 2.95 in the 1H NMR
spectrum and the downfield shift of H-4 to d ˆ 5.0.
871
done in the usual way, by adding the compound to the
enzyme-catalysed hydrolysis of 4-nitrophenyl glycosides and
following the formation of nitrophenol spectrophotometri-
cally. The (‡) form was found to be a very poor inhibitor. It
follows that the activity of the racemic form[8] had to be caused
by the ( ) form; this was confirmed. Thus ( )-1 competitively
inhibited almond b-glucosidase with a Ki of 0.33 mm, which is
about half of the value of 0.65 mm measured for (Æ)-1.
Similarly a-glucosidase and isomaltase were inhibited com-
petitively and strongly by the ( )-1 form.
The diazepine 12 was found to inhibit b-glucosidase with a
Ki of 64 mm, while a-glucosidase and isomaltase were inhib-
ited at 30 ± 40 mm. Though this is not weak inhibition, it is
much weaker than azafagomine. It suggests that 12 may fit
into the active site and bind to some of the functional groups
that azafagomine binds to, but that some interactions are
prevented.
These results show a remarkably large difference in bio-
logical activity between two enantiomers. Against b-glucosi-
dase, the inhibition differed by a factor of more than 30000.
This shows that close resemblance to the substrate or
transition state is essential for binding, and suggests that
( )-1 binds in a mode resembling either one.
Close inspection of the progress curves for the b-glucosi-
dase-catalysed formation of 4-nitrophenol from 4-nitrophenyl
b-glucoside in the presence of ( )-1 revealed that the curves
at the beginning of the reaction were slightly convex and not
linear as one would expect (progress curves with and without
the inhibitor ( )-1 are shown in Figure 3). Linearity was,
however, reached after a few minutes of reaction. Even more
pronounced when ( )-1 was allowed to stand with the enzyme
before addition of 4-nitrophenyl b-glucoside, a clearly con-
cave progress curve was seen. Again a linear curve was
reached in a few minutes.
Treatment of 21a with trifluoroacetic acid removed the Boc
group. The resulting amine did not undergo spontaneous
intramolecular substitution when treated with base, but
overnight treatment with ethyldiisopropylamine in nitrome-
thane at 608C gave the desired pyridazine 22a in 75% overall
yield. The structure determination of this compound relied on
a molecular ion obtained from mass spectrometry and the
presence of two signals at d ˆ 44.7 and 60.4 in the 13C NMR
spectrum, corresponding to the two nitrogen-substituted
carbons C-3 and C-6. Note (see Experimental Section) that
the 13C NMR chemical shift values in general are much lower
than those of 18; this shows that this compound is a six-
membered ring. After hydrogenolysis in hydrochlorous etha-
nol to 23a and acidic hydrolysis with refluxing 6n hydro-
chloric acid, 1-azafagomine was obtained in 64% yield from
22a. The product, (‡)-1, had all the physical and spectroscopic
characteristics of the racemic compound, except a specific
rotation [a]2D5 of ‡10.[8] On the basis of this rotation the
configurational preference in the enzymatic reaction of 2
could now be determined.
With the synthesis of (‡)-1 secured it was now an easy task
to synthesise the enantiomer ( )-1 from l-xylose (Scheme 7).
Though l-xylose is an unnatural sugar, it is relatively
inexpensive. Thus l-xylose was converted to 16b, the enan-
tiomer of 16a, by the same sequence of reactions.[17] From 16b
reductive amination, acetylation and mesylation gave 21b.
TFA treatment, base-promoted cyclisation and deprotection
gave ( )-1, which had a specific rotation of 9.8.
These observations mean that inhibition of the enzyme is a
relatively slow process that requires measurable time to run to
completion. This is remarkable because normally host ± guest
binding is a very fast, almost diffusion-controlled process. The
The two enantiomeric forms of 1-azafagomine were then
investigated for glycosidase inhibition (Table 2). This was
282
ꢀ WILEY-VCH Verlag GmbH, D-69451 Weinheim, 2000
0947-6539/00/0602-0282 $ 17.50+.50/0
Chem. Eur. J. 2000, 6, No. 2

1-Azafagomine
278±287
because it is linked to a reasonable explanation for the
phenomenon: The slow transformation of EI to EI* may
involve a large conformational change of the enzyme, which
could well be a time-consuming process.
To get a more detailed insight into the binding mode of 1 we
used a kinetic analysis of the progress curves to distinguish
between the two mechanisms A and B. The two models have
been analysed before.[18, 24±26] The progress curve for product
formation for both models may be written as in Equation (1),
where VSS is the steady state rate attained as t ! 1 , A and C
are constants, the latter of which depends on the initial
conditions, and b is the observed rate constant. VSS, A and b
are functions of the substrate and inhibitor concentrations.
P(t) ˆ VSSt ‡ AC(exp( bt) 1)/b
(1)
A study of b as a function of inhibitor concentration allows
us to distinguish between models A and B. For mechanism A
we have b ˆ ki[I]Km/(Km ‡ [S]) ‡ k , that is, b is a linear
i
function of the inhibitor concentration. In contrast, for
mechanism B, b ˆ kr[I]Km/(KmKi ‡ Ki[S] ‡ Km[I]) ‡ k , which
r
describes a hyperbola. In both cases a numerical analysis of b
allows determination of the relevant kinetic constants. There-
fore we determined b for almond b-glucosidase at a number of
different concentrations of ( )-1 by fitting an exponential
function to the corresponding progress curves using the curve-
fitting software package Simfit.[27] In Figure 4, b is plotted
against [I]. It is seen that the data are consistent with a straight
line and not with a hyperbolic relationship. It must therefore
be concluded that the binding follows mechanism A.
Figure 3. Control experiment demonstrating that the time-dependent
hydrolysis rate observed is due solely to the presence of inhibitor. The
enzyme was preincubated without inhibitor for 45 min, and the reaction
was then initiated by the addition of substrate and buffer (upper curve)
and, in addition, the inhibitor ( )-1 (1.5 mm). The substrate concentration
was 8 mm, corresponding to 2 Km. The continuous curves are fits of
Equation (1) to the data by nonlinear regression. The temperature was
258C.
phenomenon is called slow inhibition. The occurrence of slow-
binding enzyme inhibitors is known[18] for a number of
different enzymes including glycosidases;[19±23] however, little
is known about when or why it occurs. Two possible
mechanisms may be considered for the slow inhibition
phenomenon (Scheme 8). Binding between enzyme and
Mechanism A:
Mechanism B:
k2
k2
ES
ES
P
P
k1S
k1S
k-1
k-1
E
E
kiI
kiI
kr
k-i
k-i
EI*
EI
EI
k-r
Scheme 8. Alternative kinetic schemes for the inhibition of b-glucosidase
by ( )-1.
inhibitor may either involve the usual single step, but having
very slow on and off rates (kiI and k i being small, Scheme 8,
mechanism A). Alternatively, an initial fast binding step is
followed by a slow reversible transformation of the enzyme ±
inhibitor complex (EI) into a more stable complex (EI*,
Scheme 8, mechanism B).[18] The latter more ªcomplexº
model B seems to have become the accepted model, probably
Figure 4. Plots of the observed rate constant b at 258C vs. concentration of
inhibitor ( )-1. Each point is a mean of four seperate determinations.
Chem. Eur. J. 2000, 6, No. 2
ꢀ WILEY-VCH Verlag GmbH, D-69451 Weinheim, 2000
0947-6539/00/0602-0283 $ 17.50+.50/0
283

M. Bols et al.
FULL PAPER
Rf(3): 0.63) to give 2 and 3. The ee of 2 and 3 was determined by HPLC
(column: Daicel Chiralcel AD; UV: 220 nm; flow rate: 0.5 mLmin 1; (S)-2:
53 min and (R)-2: 58 min, (R)-3: 39 min and (S)-3: 58 min, n-hexane/
isopropanol 9:1).
From this plot the kinetic binding constants of the inhibitor
1
could be determined as ki ˆ (3.3 Æ 0.2)  104 m 1 s , k
ˆ
i
(1.1 Æ 0.2)  10 2 s and Ki ˆ 0.32 Æ 0.02 mm. The association
constant ki is relatively small, showing that the rate of binding
of the inhibitor is much smaller than diffusion-controlled
1
1-(2-Benzoyl)hydrazino-1-deoxy-a-d-xylopyranoside (5): d-xylose (4.53 g,
30 mmol) was dissolved in water (10 mL), and a solution of benzoyl
hydrazide (4.08 g, 30 mmol) in ethanol (10 mL) was added. The solution
was refluxed for 18 h, and then concentrated to a volume of 10 mL that
gave white crystals of 5 (5.85 g) upon cooling. Further concentration and
precipitation from the mother liquor gave an additional 0.58 g for a total
binding (108 ± 109 m 1 s ). The reason why binding is so slow is
1
not clear. It may be that the entrance and arrangement of the
inhibitor in the active site is a difficult and time-consuming
process. Given the high degree of resemblance of inhibitor
and substrate one would, however, then also expect the
substrate and enzyme to associate at an equally slow rate.
From the present set of experiments it is not possible to see
whether they do, because the substrate concentration is so
high that binding on the time scale of the experiment will be
fast in any case.
Another explanation is that the inhibitor and the substrate
bind to the enzyme in different fashions. Given the fact that a
small deviation from glucose stereochemistry in the pattern
and stereochemistry of the hydroxyl groups on the inhibitor
results in a large decrease in inhibition, it seems likely that this
different binding mode is a mode resembling the transition
state.
yield of 6.43 g (80%). M.p. 170 ± 1718C. [a]D25
ˆ
40.3 (c ˆ 0.6, H2O);
MS(EI): m/z: 269 [M‡1]; MS(FAB): m/z: 269.11; calcd for C12H16N2O5 ‡
H: 269.11; 13C NMR ([D6]DMSO, 50 MHz): d ˆ 166.2 (COPh), 132.8, 131.8,
128.6, 127.6 (Ph), 92.2 (C-1), 76.8, 71.6, 70.1 (C-2, C-3, C-4), 67.2 (C-5);
1H NMR ([D6]DMSO, 200 MHz): d ˆ 10.19 (d, 1H, BzNH, JNH,NH
ˆ
5.5 Hz), 7.8 (dd, 2H, PhH, Jo,m ˆ 8.1, Jo,p ˆ 1.5 Hz), 7.43 ± 7.55 (m, 3H,
PhH), 5.94 (dd, 1H, NH, JNH,NH ˆ 5.5, JNH,1 ˆ 3.0 Hz), 5.24 (d, 1H, OH, J ˆ
4.3 Hz), 5.04 (d, 1H, OH, J ˆ 4.3 Hz), 4.99 (d, 1H, OH, J ˆ 4.8 Hz), 3.86
(dd, 1H, H-1, J1,2 ˆ 8.4, JNH,1 ˆ 3.0 Hz), 3.69 ± 3.75 (dd, 1H, H-5a, J5a,5b
ˆ
11.0, J4,5a ˆ 6.2 Hz), 2.97 ± 3.37 (m, 4H, H-2, H-3, H-4, H-5b).
1-(2-Benzoyl)hydrazino-1-deoxy-d-xylitol (6): As above, d-xylose (4.53 g,
30 mmol) was dissolved in water (10 mL), and a solution of benzoylhy-
drazide (4.08 g, 30 mmol) in ethanol (10 mL) was added. The solution was
refluxed for 18 h. Water (30 mL) and platinum on carbon (600 mg, 5%)
were added, and the mixture was hydrogenated for 18 h at 40 atm and
358C. The mixture was diluted to 200 mL with warm water and filtered
through Celite. The filtrate was concentrated to a crystalline residue that
was recrystallised from methanol. Yield: 5.2 g (64%). M.p. 161 ± 1628C.
[a]D
ˆ
17.9 (c ˆ 0.6, H2O); MS(EI): m/z: 271 [M‡1], 149 [M C4H9O4],
Conclusion
121 [M NHNHBz], 105 (Bz); MS(FAB): m/z: 271.13; calcd for
C12H18N2O5‡H: 271.13; 13C NMR ([D6]DMSO, 50 MHz): d ˆ 165.3
(COPh), 133.3, 131.5, 128.6, 127.2 (Ph), 72.2, 71.8, 69.8 (C-2, C-3, C-4),
63.0 (C-5), 54.6 (C-1); 1H NMR ([D6]DMSO, 200 MHz): d ˆ 10.07 (d, 1H,
BzNH, JNH,NH ˆ 6.2 Hz), 7.82 (dd, 2H, PhH, Jo,m ˆ 6.0, Jo,p ˆ 1.8 Hz), 7.42 ±
7.50 (m, 3H, PhH), 5.29 (dd, 1H, NH, JNH,NH ˆ 6.2, JNH,1 ˆ 5.6 Hz), 4.70 (d,
1H, OH, J ˆ 4.8 Hz) 4.42 ± 4.51 (m, 2H, OH), 4.27 (d, 1H, OH, J ˆ 6.2 Hz),
3.67 (dt, 1H), 3.55 (m, 1H), 3.43 (m, 1H, H-2, H-3, H-4), 2.78 ± 2.91 (m, 2H,
H-5a, H-5b), 2.50 (m, 2H, H-1a, H-1b).
In this study we have synthesised the two enantiomers of
glycosidase inhibitor 1-azafagomine (1) and found that only
the stereoform resembling d-glucose causes significant inhib-
ition. This supports the belief that the inhibitor binds
glycosidases in a mode resembling the substrate or the
transition state.[28]
1-(1-Acetyl-2-benzoyl)hydrazino-1-deoxy-d-xylitol (7): The hydrazide 6
(5.2 g, 19.3 mmol) was dissolved in methanol (100 mL), and acetic
anhydride (9.82 g, 5 mol equiv) was added. The solution was kept for 1 h
at 258C and then concentrated. The residue was coevaporated several
Experimental Section
times with water and toluene. Yield: 6.0 g (100%). [a]D
ˆ
17.1 (c ˆ 0.2,
General: 13C NMR and 1H NMR spectra were recorded on a Varian
Gemini 200 instrument with DHO (1H NMR: d ˆ 4.7) or acetone
(1H NMR: d ˆ 2.05; 13C NMR: d ˆ 29.8) as reference. FAB mass spectra
were obtained on a Kratos MS50RF mass spectrometer with EB geometry.
The resolving power of the spectrometer was set to 3000 (10% valley
definition). The accelerating voltage was 8 kV, and the post-accelerating
detector potential was 9 kV. Ions were generated by fast atom bombard-
ment (FAB) of the compounds dissolved in 3-nitrobenzylalcohol, glycerol
or sulfolane matrices. The FAB gun employed xenon gas, 99.99%, and was
operated at 9 kV. The instrument mass scale was externally calibrated using
CsI, and the mass-to-charge ratio (m/z) of observed signals was corrected
using the coobserved signals of known glycerol cluster ions, whereby an
estimated mass accuracy of 100 ppm was achieved. The electrospray mass
spectra (ESMS) were obtained by means of a Finnigan TSQ 700 triple
quadropole instrument equipped with a nanospray source (Protana A/S,
Odense, DK). Source conditions were: spray potential 800 V; heated
(1508C) capillary: 20 V; tube lens: 70 V. Water solutions were approx-
imately 1 mm in the analyte and 1 mm in glycerol. Again the mass-to-charge
ratios of the observed M‡1 signals were corrected using the coobserved
signals of known cluster ions.
H2O); MS(FAB): m/z: 313.13; calcd for C14H20N2O6‡H: 313.14; 13C NMR
(D2O, 50 MHz): d ˆ 175.6 (COMe), 167.8 (COPh), 133.3, 131.7, 129.3, 128.2
(Ph), 73.3, 71.9, 70.1 (C-2, C-3, C-4), 63.9 (C-5), 52.1 (C-1), 21.0 (Ac);
1H NMR (D2O, 200 MHz): d ˆ 10.37 (brs, 1H, BzNH), 7.82 (dd, 2H, PhH),
7.30 ± 7.50 (m, 3H, PhH), 3.3 ± 4.2 (m, 7H), 2.05 (s, 3H, CH3).
‡1-(1-acetyl-2-benzoyl)hydrazino-1-deoxy-2,3-O-isopropylidene-d-xylitol
(9): The N-acetate 7 (6.0 g, 19.2 mmol) was dissolved in acetone (130 mL)
and 2,2-dimethoxypropane (26 mL). p-Toluenesulfonic acid (660 mg,
0.2 mol equiv) was added. The reaction mixture was stirred for 2 h, after
which time TLC (Rf ˆ 0.85; EtOAc/EtOH 9:1) showed completion of the
reaction. The mixture was neutralised with saturated NaHCO3 solution and
then concentrated. The residue was redissolved in CH2Cl2 (200 mL) and
washed with water (3 Â 200 mL). The organic phase was dried with MgSO4
and concentrated to give the diacetonide 8 as a colourless syrup: MS(FAB):
m/z ˆ 393.20; calcd for C12H18N2O5 ‡ H: 393.20; 13C NMR (CDCl3,
200 MHz): d ˆ 173.3 (COMe), 166.3 (COPh), 132.6, 131.7, 128.6, 127.2
(Ph), 109.6 (2C, C(CH3)2), 76.8, 74.7, 69.5 (C-2, C-3, C-4), 65.5 (C-5), 53.9
(C-1), 29.2, 27.1, 26.0, 25.3 (C(CH3)2), 20.3 (CH3); 1H NMR (CDCl3,
200 MHz): d ˆ 9.1 (s, 1H, BzNH), 7.8 (dd, 2H, PhH), 7.40 ± 7.60 (m, 3H,
PhH), 3.3 ± 5.5 (m, 7H), 2.0 (s, 3H, Ac), 1.3 (2s, 12H, CH3ꢁs). This syrup was
dissolved in 80% acetic acid (30 mL) and stirred at 408C for 2 h, after
which TLC (Rf ˆ 0.47 EtOAc/EtOH 9:1) showed one principal product.
The solution was concentrated, and the residue subjected to flash
Lipase-catalysed transesterification of (Æ)-2: These reactions were per-
formed in two ways. Either a mixture of (Æ)-2 (20 mg) and a lipase (2 mg)
in vinyl acetate (2 mL) was stirred at 258C for several days, or (Æ)-2
(20 mg) and the lipase (10 mg) was reacted with vinyl acetate (0.08 mL) in a
solvent (0.2 ± 10 mL, Table 1) at 408C for a number of hours. In both cases
the reaction was stopped when TLC showed a desirable conversion. For
work-up, the lipase was filtered off and the solution concentrated in vacuo.
The residue was separated by flash chromatography (EtOAc, Rf(2): 0.31,
chromatography in EtOAc to give 9 as a syrup (3.72 g, 55%). [a]D
ˆ
19.3 (c ˆ 0.4, CHCl3); MS(FAB): m/z: 353.17; calcd for C12H18N2O5‡H:
353.17; 1H NMR (CDCl3, 200 MHz): d ˆ 9.3 (brs, 1H, BzNH), 7.85 (dd,
2H, PhH), 7.40 ± 7.60 (m, 3H, PhH), 4.25 (dt, 1H), 3.9 (m, 3H), 3.7 (m, 3H),
2.1 (s, 3H, Ac), 1.3 (s, 6H, CH3ꢁs).
284
ꢀ WILEY-VCH Verlag GmbH, D-69451 Weinheim, 2000
0947-6539/00/0602-0284 $ 17.50+.50/0
Chem. Eur. J. 2000, 6, No. 2

1-Azafagomine
278±287
1-(1-Acetyl-2-benzoyl)hydrazino-1-deoxy-2,3-O-isopropylidene-5-O-
over 10 min. The reaction mixture was allowed to warm to 508C, and
Et3N (2.5 mL, 18.1 mmol) was added. After 30 min in which the mixture
was allowed to reach 258C, the reaction was complete, and one product was
observed on TLC (Rf ˆ 0.78, pentane/EtOAc 1:1) The solution was diluted
with CH2Cl2 (200 mL), and washed with water (2 Â 200 mL) and saturated
NaCl solution (200 mL). The organic phase was dried (MgSO4) and
concentrated to give 15 as a colourless syrup (1.26 g, 97%). 1H NMR
(CDCl3, 200 MHz): d ˆ 7.9 (dd, 2H, PhH), 7.40 ± 7.60 (m, 3H, PhH), 4.3 ±
4.7 (m, 5H), 3.1 (q, 1H), 2.1 (s, 3H, Ac), 1.4 (d, 6H, CH3ꢁs), 0.9 (s, 9H, tBu),
0.1 (s, 6H, SiCH3).
(2,4,6-trimethylbenzene)sulfonyl-d-xylitol (10): The diol
9 (352 mg,
1 mmol) was dissolved in pyridine (2 mL), and 2,4,6-trimethylbenzenesul-
fonyl chloride (240 mg, 2.2 equiv) and 4-N,N-dimethylaminopyridine
(7 mg) were added. The reaction mixture was stirred for 18 h, and the
mixture was separated between EtOAc (10 mL) and H2O (10 mL). The
organic layer was washed twice with water (10 mL) and then dried
(Na2SO4) and concentrated. The residue (870 mg) was subjected to flash
chromatography (EtOAc/pentane 1:1), which gave first 1-(1-acetyl-2-
benzoyl)hydrazino-1-deoxy-4,5-di-O-(2,4,6-trimethylbenzene)sulfonyl-2,3-
O-isopropylidene-d-xylitol (170 mg, 24%) and then 10 (388 mg 73%).
(2S,3S,4S)-1-N-Amino-2',3,4-tri-O-benzyl-3,4-dihydroxy-2-hydroxyme-
thylpyrrolidine (18): The dimesylate 17 (106 mg, 0.18 mmol) was dissolved
in hydrazine hydrate (1 mL) and heated to 1008C for 18 h. The solution was
concentrated, and the residue was subjected to flash chromatography
[a]D
ˆ
12.7 (c ˆ 2.2, CHCl3); MS(FAB): m/z: 535.208; calcd for
C26H34N2O8S‡H: 535.211; 13C NMR (CDCl3, 50 MHz): d ˆ 143.6, 140.0,
132.8, 131.8, 128.9, 127.3 (Phꢁs), 78.7, 75.0, 70.4, 67.2 (C-2, C-3, C-4, C-5),
49.0 (C-1), 27.2, 26.7 ((CH3)2C), 22.6 (ArCH3), 21.0 (Ac); 1H NMR (CDCl3,
200 MHz): d ˆ 9.0 (brs, 1H, BzNH), 7.8 (d, 2H, PhH, Jo,m ˆ 8.1 Hz), 7.5 (t,
1H, PhH), 7.4 (t, 2H, PhH), 6.9 (s, 2H, PhH), 4.2 (m, 1H), 3.8 ± 4.0 (m, 3H),
3.75 (d, 1H), 2.9 (m, 2H, H-5a, H-5b), 2.5 (s, 6H, ArCH3ꢁs), 2.25 (s, 3H,
ArCH3), 2.05 (s, 3H, Ac), 1.3 (s, 6H, (CH3)2C).
‡
(EtOAc) to give 18 as a clear syrup (45 mg, 59%). MS(EI): m/z: 418 [M ];
13C NMR (CDCl3): d ˆ 138.2, 137.9, 127.6 ± 128.4 (Ph), 84.9, 80.3 (C-2, C-3),
74.1, 73.4, 71.6, 71.1, 70.7 (C-4, C-5, 3Bn), 62.9 (C-1).
1-(2-tert-Butyloxycarbonyl)hydrazino-1-deoxy-2,3,5-tri-O-benzyl-d-xylitol
(19a): A solution of 16a (5.9 g, 14 mmol) in methanol (70 mL) was treated
with tert-butylcarbazate (3.7 g, 28 mmol), acetic acid (1.47 g, 24.5 mmol)
and NaCNBH3 (3.33 g, 53.4 mmol) at 608C for 18 h. NaHCO3 solution
(satd, 30 mL) was added, then water (100 mL), and the mixture was
extracted with CH2Cl2 (3 Â 50 mL). The organic layer was dried, concen-
trated and subjected to flash chromatography (EtOAc/CH2Cl2, 1:4). This
gave 19a (6.4 g, 84%). MS(FAB): m/z: 537.296; calcd for C31H40N2O6‡H:
537.297; 13C NMR (CDCl3): d ˆ 138.1, 127.6 ± 128.4 (Ph), 74.2, 73.2, 72.1,
(4S,5S,6R)-2-N-Acetyl-1-N-benzoyl-4,5-O-isopropylidene-4,5,6-trihy-
droxy-1,2-diazapine (11): Compound 10 (245 mg, 0.46 mmol) was dissolved
in a solution of Na (35 mg, 1.46 mmol) in MeOH (1 mL) and kept at 258C
for 30 min. A small lump of solid CO2 was added to neutralise excess base.
The solution was concentrated and subjected to flash chromatography in
EtOAc. This gave 11 as a syrup (118 mg, 77%). [a]D ˆ ‡16.1 (c ˆ 0.5,
CHCl3); MS(FAB): m/z: 335.17; calcd for C17H22N2O5‡H: 335.16; 13C NMR
ˆ
(CDCl3) at least two rotamers: d ˆ 171.6, 171.5 (C O), 131.7 ± 133.2, 126.4 ±
1
71.1, 68.2 (C-4, C-5, 3Bn), 28.3 ((CH3)3C); H NMR (CDCl3): d ˆ 7.3 (m,
129.1 (Ph), 111.0, 110.8 (C(CH3)2), 84.8, 83.7 (C-5), 73.0, 72.5 (C-4), 70.8,
67.8 (C-6), 55.3, 52.6 (C-3), 48.2, 46.7 (C-7), 27.0, 26.7 (C(CH3)2), 20.3, 20.2
(Ac).
15H, PhH), 6.2 (brs, 1H, BocNH), 4.4 ± 4.7 (m, 6H, Bnꢁs), 4.3 (brs, 2H,
OH, NH), 4.0 (t, 1H, J ˆ 6 Hz, H-3), 3.7 (m, 2H), 3.55 (dd, 1H, J ˆ 8.3 Hz,
J ˆ 6 Hz, H-5a), 3.45 (dd, 1H, J ˆ 8.3 Hz, J ˆ 7 Hz, H-5b), 3.2 (dd, 1H, J ˆ
12 Hz, J ˆ 4.3 Hz, H-1a), 3.1 (dd, 1H, J ˆ 12 Hz, J ˆ 3.2 Hz, H-1b), 1.4 (s,
9H, CH3ꢁs).
(4,5-trans-5,6-trans)-4,5,6-Trihydroxy-1,2-diazapine hydrochloride (12):
Compound 11 was dissolved in aqueous hydrochloric acid (4m, 12 mL)
and refluxed for 18 h. The solution was concentrated to syrup containing 12
and some benzoic acid. The latter was removed by extracting the residue
with diethyl ether (4 Â 20 mL). Very pure 12 remained as the hydro-
1-(2-tert-Butyloxycarbonyl)hydrazino-1-deoxy-2,3,5-tri-O-benzyl-l-xylitol
(19b): Synthesis carried out as for 19a above, but starting from 16b. [a]D
‡3.7 (c ˆ 1.0, CHCl3).
ˆ
‡
chloride. Yield: 141 mg (97%); MS(EI): m/z: 148 [M ]; 13C NMR (D2O):
d ˆ 75.7 (C-5), 68.5 (C-4, C-6), 46.5 (C-3, C-7); 1H NMR (D2O): d ˆ 3.82
(dt, 2H, J3b,4 ˆ J6,7b ˆ 6.3, J4,5 ˆ J5,6 ˆ 6.3, J3a,4 ˆ J6,7a ˆ 4.0 Hz, H-4, H-6), 3.61
(t, 1H, J4,5 ˆ J5,6 ˆ 6.3 Hz, H-5), 3.24 (d, 2H, J3a,4 ˆ J6,7a ˆ 4.0 Hz, H-3a,
H-7a), 3.23 (d, 2H, J3b,4 ˆ J6,7b ˆ 6.3 Hz, H-3b, H-7b).
1-(1-Acetyl-2-tert-butyloxycarbonyl)hydrazino-1-deoxy-2,3,5-tri-O-ben-
zyl-4-O-methanesulfonyl-d-xylitol (21a): Compound 19a (203 mg,
0.38 mmol) was dissolved in methanol (20 mL) and acetic anhydride
(2 mL) was added. The mixture was kept for 18 h and then concentrated.
The residue was identified by NMR as the acetate 20a (1H NMR (CDCl3):
d ˆ 7.2 (brs, 15H, PhH), 6.85 (brs, 1H, NH), 4.4 ± 4.7 (m, 7H), 3.9 (m, 3H),
3.4 ± 3.8 (m, 3H), 1.95 (s, 3H, Ac), 1.4 (s, 9H, ((CH3)3C)). This compound
was dissolved in pyridine (10 mL) and cooled to 08C. Methanesulfonyl
chloride (50 mg, 0.44 mmol) was added, and the mixture was allowed to
reach room temperature over 1 h. Water (10 mL) was added, and the
mixture was extracted with CH2Cl2 (2 Â 10 mL). Finally the organic layer
was concentrated and coevaporated with toluene. The residue was
subjected to column chromatography (EtOAc/CH2Cl2 0:1 to 1:4) to give
21a (229 mg, 92%). MS(FAB): m/z: 657.285; calcd for C34H44N2O9S‡H:
657.285; 1H NMR (CDCl3): d ˆ 7.3 (brs, 15H, PhH), 6.85 (brs, 1H, NH), 5.0
(m, 1H), 4.4 ± 4.7 (m, 7H), 3.9 (m, 1H), 3.5 ± 3.8 (m, 3H), 3.25 (m, 1H), 2.95
(s, 3H, Ms), 1.95 (s, 3H, Ac), 1.45 (s, 9H, ((CH3)3C).
1-(1-Acetyl-2-benzoyl)hydrazino-5-O-tert-butyldimethylsilyl-1-deoxy-2,3-
O-isopropylidene-d-xylitol (13): Imidazole (510 mg, 2.5 mol equiv) and
tert-butyldimethylsilyl chloride (652 mg, 1.2 mol equiv) were added to a
solution of 9 (1.06 g, 3 mmol) in DMF (2.2 mL) at 08C. The mixture was
stirred until TLC showed the disappearance of starting material (20 min)
and formation of one product (Rf ˆ 0.46, EtOAc/pentane 1:1). Water
(100 mL) was added, and the mixture was extracted with ether (3 Â
100 mL). The combined organic phase was washed with water (200 mL),
dried (MgSO4) and concentrated to give 13 as a syrup (1.38 g, 98%). [a]D
ˆ
‡
‡3.1 (c ˆ 0.3, CHCl3); MS(EI): m/z: 467 [M ]; 1H NMR (CDCl3,
200 MHz): d ˆ 8.9 (s, 1H, BzNH), 7.85 (dd, 2H, PhH), 7.40 ± 7.60 (m, 3H,
PhH), 4.25 (dt, 1H), 3.6 ± 3.9 (m, 6H), 2.1 (s, 3H, Ac), 1.4 (s, 6H, CH3ꢁs), 0.9
(s, 9H, tBu), 0.05 (s, 6H, SiCH3).
1-(1-Acetyl-2-tert-butyloxycarbonyl)hydrazino-1-deoxy-2,3,5-tri-O-ben-
zyl-4-O-methanesulfonyl-l-xylitol (21b): Synthesis carried out as for 21a
1-(1-Acetyl-2-benzoyl)hydrazino-5-O-tert-butyldimethylsilyl-1-deoxy-4-
O-methanesulfonyl-2,3-O-isopropylidene-d-xylitol (14): The alcohol 13
(245 mg, 0.5 mmol) was dissolved in pyridine (1.7 mL) at 08C and
methanesulfonyl chloride (61 mL, 0.79 mmol, 1.5 equiv) was added. After
2.5 h the reaction was complete, and a small amount of water (0.2 mL) was
added and allowed to react for 5 min. The mixture was diluted with CH2Cl2
(10 mL) and washed with HCl (10 mL), NaHCO3 (10 mL) and water
(10 mL). The organic phase was dried (MgSO4) and concentrated to give
the mesylate 14 as a colourless syrup (171 mg, 63%). 1H NMR (CDCl3,
200 MHz): d ˆ 9.1 (brs, 1H, BzNH), 7.85 (dd, 2H, PhH), 7.40 ± 7.60 (m, 3H,
PhH), 4.65 (m, 1H), 3.7 ± 4.3 (m, 6H), 3.2 (s, 3H, CH3SO2), 2.1 (s, 3H, Ac),
1.4 (s, 6H, CH3ꢁs), 0.9 (s, 9H, tBu), 0.05 (s, 6H, SiCH3).
above, but starting from 19b. [a]D
ˆ
4.8 (c ˆ 1.0, CHCl3).
(3S,4S,5S)-1-N-Acetyl-3',4,5-tri-O-benzyl-4,5-dihydroxy-3-hydroxymethyl-
hexahydropyridazine (22a): The mesylate 21a (220 mg, 0.33 mmol) was
dissolved in TFA (2 mL) and kept for 1 h at room temperature. The TFA
was removed by evaporation and NaHCO3 solution (satd, 10 mL) was
added to the residue. This mixture was extracted with CH2Cl2 (2 Â 10 mL)
and the combined organic layers were dried (MgSO4) and concentrated.
This residue was dissolved in nitromethane (2 mL), EtNiPr2 (200 mL,
148 mg, 1.15 mmol) was added, and the mixture was heated at 608C for
18 h. The solvent was removed by concentration, and the residue was
subjected to chromatography (EtOAc/CH2Cl2, 0:1 to 1:4) to give 22a
5-(1-Acetyl-2-benzoyl)hydrazino-1-O-tert-butyldimethylsilyl-5-deoxy-3,4-
O-isopropylidene-l-thre-2-ulose (15): To a solution of oxalyl chloride
(0.36 mL, 4.18 mmol) in dry CH2Cl2 (6 mL) at 658C was added a solution
of DMSO (0.695 mL, 9.76 mmol) in dry CH2Cl2 (3 mL). After 5 min at that
temperature a solution of 13 (1.3 g, 2.79 mmol) in CH2Cl2 (3 mL) was added
‡
(115 mg, 75%). MS(EI): m/z: 460 [M ], 339 [M CH2OBn]; MS(FAB):
m/z: 461.238; calcd for C28H32N2O4‡H: 461.244; 13C NMR (CDCl3): d ˆ
172.8, 138.3, 138.0, 137.7, 127.8 ± 128.5 (Ph), 77.8, 74.7, 73.4 (2C), 72.0, 67.0
(3Bn, C-3, C.4, C-5), 60.4 (C-6), 44.7 (C-2), 20.6 (Ac).
Chem. Eur. J. 2000, 6, No. 2
ꢀ WILEY-VCH Verlag GmbH, D-69451 Weinheim, 2000
0947-6539/00/0602-0285 $ 17.50+.50/0
285

M. Bols et al.
FULL PAPER
(3R,4R,5R)-1-N-Acetyl-3',4,5-tri-O-benzyl-4,5-dihydroxy-3-hydroxyme-
thylhexahydropyridazine (22b): Synthesis carried out as for 22a above, but
starting from 21b. [a]D ˆ ‡8.8 (c ˆ 1.1, CHCl3).
Table 3. X-ray data for the structures of 5 and 6.
5
6
formula
formula weight
crystal system
space group
a [Š]
b [Š]
c [Š]
b [8]
Z
C12H16N2O5
268.28
monoclinic
C2
20.003(1)
8.3239(4)
16.9126(9)
112.359(3)
8
C12H19N2O5
271.30
monoclinic
P21
9.318(2)
5.228(1)
13.469(2)
89.58(1)
2
0.047
0.046
1.36
880
(3S,4S,5S)-4,5-Dihydroxy-3-hydroxymethylhexahydropyridazine, (‡)-1-
azafagomine ((‡)-1): To a solution of 22a (140 mg, 0.30 mmol) in EtOH
(10 mL) was added HCl (1m, 2 mL) and Pd/C (50 mg), and the solution was
hydrogenated at 1 atm H2 until no more hydrogen was absorbed (4 ± 6 h).
The mixture was filtered through Celite and concentrated to give a residue
containing (3S,4S,5S)-1-N-acetyl-4,5-dihydroxy-3-hydroxymethylhexahy-
dropyridazine (23a). This was dissolved in HCl (6m, 8 mL) and heated to
1008C for 18 h. The solution was concentrated and subjected to ion-
‡
exchange chromatography using Amberlite IR-120 resin (H , 10 mL) and
R
0.058
eluting with dilute NH4OH solution (1m, 100 mL). Concentration of the
basic eluate gave (‡)-1 (29 mg, 64%): [a]D ˆ ‡10.0 (c ˆ 0.3, H2O). The
NMR spectra were identical to those of (Æ)-1.[8]
Rw
0.065
GoF
2.19
reflections used
parameters
2404
342
(3R,4R,5R)-4,5-Dihydroxy-3-hydroxymethylhexahydropyridazine
( )-1-
172
azafagomine (( )-1): Synthesis carried out as for (‡)-1, but starting from
22b: m.p.: 157 ± 1678C (decomp); [a]D
ˆ
9.8 (c ˆ 0.85, H2O). The NMR
spectra were identical to those of (Æ)-1.[8]
Enzymatic assays
[1] T. D. Heightman, A. T. Vasella, Angew. Chem. 1999, 111, 794 ± 815;
Angew. Chem. Int. Ed. 1999, 38, 750 ± 770.
[2] E. Truscheit, W. Frommer, B. Junge, L. Müller, D. Schmidt, Angew.
Chem. 1981, 93, 738 ± 755; Angew. Chem. Int. Ed. Engl. 1981, 20, 744 ±
761.
a- and b-glucosidase: 4-Nitrophenyl-a- and -b-d-glucopyranoside, a-
glucosidase (yeast, Sigma G5003, 4.5 sigma unitsmg 1), isomaltase (yeast,
Sigma I-1256, 47 sigma unitsmg 1) and b-glucosidase (sweet almonds,
Sigma G4511, 21 sigma unitsmg 1) were purchased from Sigma Chemical.
The b-glucosidase is a purified enzyme that showed only a single band on
an SDS electrophoresis gel at 62 kDa. 4-Nitrophenyl-a- or -b-d-glucopyr-
anoside was used as substrate and a- or b-glucosidase as catalyst in a
sodium phosphate buffer (0.05m, pH 6.8) at 258C. Formation of the
product, 4-nitrophenol, was measured continually at 400 nm by means of a
Milton Roy Genesys 5 spectrometer. In all kinetic runs less than 1% of the
initial substrate was consumed, assuring the constancy of the substrate
concentration.
[3] a) M. von Itzstein, W.-Y. Wu, G. B. Kok, M. S. Pegg, J. C. Dyason, B.
Jin, T. V. Phan, M. L. Smythe, H. F. White, S. W. Oliver, P. M. Colman,
J. N. Varghese, D. M. Ryan, J. M. Woods, V. J. Hotham, J. M.
Cameron, C. R. Pean, Nature 1993, 363, 418 ± 423; b) W. Lew, H.
Wu, D. B. Mendel, P. A. Escarpe, X. Chen, W. G. Laver, B. J. Graves,
C. U. Kim, Bioorg. Med. Chem. Lett. 1998, 8, 3321 ± 3324.
[4] a) B. Ganem, Acc. Chem. Res. 1996, 29, 340 ± 347; b) K. Kraehen-
buehl, S. Picasso, P. Vogel, Bioorg. Med. Chem. Lett. 1997, 7, 893 ± 896;
c) I. B. Parr, B. A. Horenstein, J. Org. Chem. 1997, 62, 7489 ± 7494;
d) N. Panday, T. Granier, A. Vasella, Helv. Chim. Acta 1998, 81, 475 ±
490; e) B. A. Johns, C. R. Johnson, Tetrahedron Lett. 1998, 39, 749 ±
752; f) U. Katzmaier, C. Schneider, Tetrahedron Lett. 1998, 39, 817 ±
818; g) D. A. Berges, M. D. Ridges, N. K. Dalley, J. Org. Chem. 1998,
63, 391 ± 392.
Ki determinations were performed as follows: Two thermostatted solutions
of a) 1 mL of 0.1m buffer, 800 mL substrate in varied concentration ([S] ˆ
0.2 Km to 5 Km) and 100 mL water, and b) 100 mL enzyme were mixed, and
the reaction was immediately monitored. From eight experiments with
varied substrate concentration, initial reaction rates were calculated from
the slope of the first-order plot of product absorption vs. reaction time. KM
and Vmax were calculated from a Hanes plot. From four experiments KM'
was calculated from the Hanes plot, and from KM and K'M, Ki was
calculated.
[5] a) M. L. Sinott, Chem. Rev. 1990, 90, 1171 ± 1202; b) G. Legler, Adv.
Carbohydr. Chem. Biochem. 1990, 48, 319 ± 384; c) B. Winchester,
G. W. J. Fleet, Glycobiology 1992, 2, 199 ± 210.
[6] a) M. Bols, Acc. Chem. Res. 1998, 31, 1 ± 8; b) Y. Ichikawa, Y. Igarashi,
M. Ichikawa, Y. Suhura, J. Am. Chem. Soc. 1998, 120, 5854.
[7] a) T. M. Jespersen, W. Dong, T. Skrydstrup, M. R. Sierks, I. Lundt, M.
Bols, Angew. Chem. 1994, 106, 1858 ± 1860; Angew. Chem. Int. Ed.
Engl. 1994, 33, 1778 ± 1779; b) Y. Ichikawa, Y. Igarashi, Tetrahedron
Lett. 1995, 36, 4585 ± 4586.
[8] M. Bols, R. G. Hazell, I. Thomsen, Chem. Eur. J. 1997, 3, 940 ± 947.
[9] N. Chida, Y. Furuno, S. Ogawa, J. Chem. Soc. Chem. Commun. 1989,
1230 ± 1231.
[10] P. Diter, S. Taudin, O. Samuel, H. B. Kagan, J. Org. Chem. 1994, 59,
370 ± 373.
[11] H.-H. Stroh, A. Arnold, H.-G. Scharnow, Chem. Ber. 1964, 98, 1404 ±
1410.
Slow inhibition experiments were carried out with and without preincu-
bation of enzyme and inhibitor. Experiments without preincubation were
carried out as follows: Two thermostatted solutions of a) 1 mL of 0.1m
buffer, 800 mL substrate (40 mm) and 100 mL inhibitor in varied concen-
trations ([I] ˆ 5 ± 200 Ki) and b) 100 mL enzyme (1 mgmL 1) were mixed,
and the reaction was immediately monitored spectrophotometrically.
Experiments with preincubation were carried out as follows: Solutions of
25 mL inhibitor ([I] ˆ 20 Ki) and 25 mL enzyme (2 mgmL 1) were mixed
and kept at 258C for 60 min. A thermostatted mixture of 1 mL of 0.1m
buffer, 800 mL substrate (40 mm) and 150 mL of inhibitor in varied
concentrations ([I] ˆ 0 ± 130 Ki) was added, and the reaction was immedi-
ately monitored spectrophotometrically.
Crystallographic data: Both 5 and 6 grew as thin, colourless needles. X-ray
data for 6 were measured on a Huber four-circle diffractometer in the q ± 2q
step scan mode, whereas those for 5 were collected on a Siemens SMART
area detector diffractometer, collecting narrow-frame w scans. In both
cases, MoKa radiation was used at room temperature. Both structures were
solved by means of the SIR92 direct methods package and refined by
the full-matrix least-squares method, allowing anisotropic displacement
parameters for all but the hydrogen atoms; hydrogen atoms on carbon
were kept in calculated positions. Details of the analysis are given in
Table 3.
[12] H. Zinner, H. Brandhoff, H. Schmandke, H. Kristen, R. Huun, Chem.
Ber. 1959, 92, 3151 ± 3155.
[13] H. Zinner, K. Wessely, H. Kristen, Chem. Ber. 1959, 92, 1618 ± 1625.
[14] A. Dureault, I. Tranchepain, J.-C. Depezay, J. Org. Chem. 1989, 54,
5324 ± 5330.
[15] I. B. Parr, B. A. Horenstein, J. Org. Chem. 1997, 7, 7489 ± 7494.
[16] F. Fache, L. Jaquot, M. Lemaire, Tetrahedron Lett. 1994, 35, 3313 ±
3314.
[17] G. W. J. Fleet, S. J. Nicholas, P. W. Smith, S. V. Evans, L. E. Fellows,
R. J. Nash, Tetrahedron Lett. 1985, 26, 3127 ± 3130.
[18] J. F. Morrison, C. T. Walsh, Adv. Enzymol. Relat. Areas Mol. Biol.
1988, 61, 201 ± 301.
[19] A. K. Grover, R. J. Cushley, Biochim. Biophys. Acta 1977, 482, 109 ±
124.
[20] C. Danzin, A. Ehrhard, Arch. Biochem. Biophys. 1987, 257, 472 ± 475.
[21] G. Hanozet, H.-P. Pircher, P. Vanni, B. Oesch, G. Semenza, J. Biol.
Chem. 1981, 256, 3703 ± 3711.
Acknowledgments
We thank the Danish Natural Science Research Council for financial
support, and the Carlsberg foundation for the SMART diffractometer.
286
ꢀ WILEY-VCH Verlag GmbH, D-69451 Weinheim, 2000
0947-6539/00/0602-0286 $ 17.50+.50/0
Chem. Eur. J. 2000, 6, No. 2

1-Azafagomine
278±287
[22] G. Legler, E. Julich, Carbohydr. Res. 1984, 128, 61 ± 72.
[23] M. Kang, A. Elbein, Plant Physiol. 1983, 71, 551 ± 554.
[24] R. G. Duggleby, P. V. Atwood, J. C. Wallace, D. B. Keech, Biochem-
istry 1982, 21, 3364 ± 3370.
[27] W. G. Bardsley, University of Manchester, England, personal com-
munication.
[28] A preliminary report describing some of this work has appeared: B. V.
Ernholt, I. Thomsen, M. Bols, Synlett 1999, 6, 701 ± 704.
[25] J.-M. Zhou, C. Liu, C. L. Tsou, Biochemistry 1989, 28, 1070 ± 1076.
[26] M. J. Sculley, J. F. Morrison, W. W. Cleland, Biochem. Biophys. Acta
1996, 1298, 78 ± 96.
Received: April 26, 1999 [F1749]
Chem. Eur. J. 2000, 6, No. 2
ꢀ WILEY-VCH Verlag GmbH, D-69451 Weinheim, 2000
0947-6539/00/0602-0287 $ 17.50+.50/0
287