An isofagomine analogue with an amidine at the pseudoanomeric position

Article abstract of DOI:10.1021/ol200942g

(3R,4R,5R)-2-Imino-3,4-dihydroxy-5-hydroxymethylpiperidine hydrocloride or isofagomidine was synthesized from d-arabinose in 12 steps and an overall yield of 9.9%. The synthesis proceeded by introduction of an aminomethyl group in the 4-position of d-arabinose and conversion of C-1 into a nitrile. The key step in the synthesis was a copper-catalyzed cyclization of aminonitrile to amidine. Isofagomidine was a potent α-mannosidase inhibitor (K_i = 0.75 μM).

Full text of DOI:10.1021/ol200942g

ORGANIC
LETTERS
2011
Vol. 13, No. 11
2908–2911
An Isofagomine Analogue with an Amidine
at the Pseudoanomeric Position
†
‡
‡
†
ꢀ
€
ꢀ
ꢀ
ꢀ
~
Emil Lindback, Oscar Lopez, Jose G. Fernandez-Bolanos, Stephan P. A. Sauer,
and Mikael Bols*,†
Department of Chemistry, University of Copenhagen, Universitetsparken 5, 2100 Kbh Ø,
Denmark, and Department of Organic Chemistry, University of Seville,
´
ꢀ
Profesor Garcıa Gonzalez 1, 41012 Seville, Spain
bols@chem.ku.dk
Received April 11, 2011
ABSTRACT
(3R,4R,5R)-2-Imino-3,4-dihydroxy-5-hydroxymethylpiperidine hydrocloride or isofagomidine was synthesized from D-arabinose in 12 steps and an
overall yield of 9.9%. The synthesis proceeded by introduction of an aminomethyl group in the 4-position of D-arabinose and conversion of C-1 into
a nitrile. The key step in the synthesis was a copper-catalyzed cyclization of aminonitrile to amidine. Isofagomidine was a potent R-mannosidase
inhibitor (Ki = 0.75 μM).
Iminosugar glycosidase inhibitors of the 1-deoxyno-
jirimycin type (1) are natural products that resemble
monosaccharides;1 in fact, the conjugate acids of these
amines resemble the transition state of glycoside clea-
vage, which is the likely basis of their biological activity
(Figure 1).2 These compounds have therapeutic poten-
tial, which has been exploited in antidiabetes treatment,3
and are also being evaluated for treatment of diseases
ranging from cancer and AIDS to lysosomal storage
disorders and hepatitis infections.4 A number of inter-
esting analogues of 1 have been made over the years.
Particularly noteworthy is the glucoamidine 2 that was
designed by Ganem and found to be a broad-spectrum
glycosidase inhibitor.5 Amidine 2 is obviously a much
stronger base than 1, and the ionic interaction between
enzyme and inhibitor should be more profound.
A series of glycosidase inhibitors related to 1 are 1-aza-
sugars, such as isofagomine 3,6 which in protonated form
resemble the alternate resonance form of the oxocarbenium
ion transition state (Figure 1). These compounds are
potent β-glycosidase inhibitors presumably because they
form a strong salt bridge with the nucleophilic carboxylate
in these enzymes.7 X-ray structures of the inhibitorꢀenzyme
complex provide evidence of this, showing a protonated
isofagomine.8
On the basis of the observation that ionic interaction is
so important for the binding of 3, the idea emerges that
application of Ganem’s idea of introducing an amidine
into the isofagomine structure may lead to interesting
inhibitors. The structure 4 (Figure 1), which has an ami-
dine placed at the 1,2-positions, has the advantage that it
has a polar group at the 2-position where NHAc9 and OH
† University of Copenhagen.
‡ University of Seville.
(6) Jespersen, T. M.; Dong, W.; Sierks, M. R.; Skrydstrup, T.; Lundt,
I.; Bols, M. Angew. Chem., Int. Ed. 1994, 33, 1778–1779.
(7) Williams, S. J.; Hoos, R.; Withers, S. G. J. Am. Chem. Soc. 2000,
122, 2223–2235.
€
(1) Stutz, A. Iminosugars as Glycosidase Inhibitors; Wiley-VCH:
Weinheim, 1999.
(2) Bols, M. Acc. Chem. Res. 1998, 31, 1–8.
(3) Somsak, L.; Nagy, V.; Hadady, Z.; Docsa, T.; Gergely, P. Curr.
Pharm. Des. 2003, 9, 1177–1189.
(8) Varrot, A.; Tarling, C. A.; Macdonald, J. M.; Stick, R. V.; Zechel,
D. L.; Withers, S. G.; Davies, G. J. J. Am. Chem. Soc. 2003, 125, 7496–
7497.
(4) Asano, N. Glycobiology 2003, 13, 93R–104R.
(5) (a) Tong, M. K.; Papandreou., G.; Ganem, B. J. Am. Chem. Soc.
1990, 112, 6137–6139. (b) Papandreou, G.; Tong, M. K.; Ganem, B.
J. Am. Chem. Soc. 1993, 115, 11682–11690.
(9) Nishimura, Y.; Shitara, E.; Kojima, F.; Takeuchi, T. Bioorg.
Med. Chem. 1999, 7, 1241–1246.
€
(10) Liu, H.; Liang, X.; Søhoel, H.; Bulow, A.; Bols, M. J. Am. Chem.
Soc. 2001, 123, 5116–5117.
r
10.1021/ol200942g
Published on Web 05/06/2011
2011 American Chemical Society

Scheme 2. Elimination Reactions from the Henry Adductsa
Figure 1. Nitrogen-containing glycosidase inhibitors 1-deoxyno-
jirimycin (1), glucoamidine (2), isofagomine (3), and isofagomidine
(4) and their similarity with the oxocarbenium ion intermediate.
a Key: (a) Ac2O/TsOH; (b) silica; (c) MeNO2/Et3N; (d) Et3N.
came in contact with silica gel during chromatography to
give the E-alkene 8 in 55% yield (Scheme 2). If a more
conventional elimination with triethylamine was chosen
the Z-alkene 9 was obtained in a more favorable yield
(Scheme 2). However, only reduction of the former led to
the desired product: when 9 was treated with lithium
aluminum hydride conjugate reduction was not observed,
but double bond migration (Scheme 3).
groups are acceptable; the latter group has been shown to
improve inhibition.10 In this paper, we report the synthesis
of 4 and some of its properties as a glycosidase inhibitor.
Scheme 1. Synthesis of Henry Adducts 7a and 7ba
Scheme 3. Reduction of Nitroalkenes 8 and 9a
a Key: (a) DessꢀMartin periodinane; (b) MeNO2, Et3N.
Synthesis of 4 started with known building block 5 that
can be made from inexpensive D-arabinose in two steps.11
Alcohol 5 was oxidized to the ketone 6 with DessꢀMartin
periodinane giving 84% yield. The ketone 6 was then
reacted with Et3N in nitromethane giving the two diaster-
eomeric Henry adducts 7a and 7b in a ratio of 6:5 and in a
combined yield of 90% (Scheme 1).
The subsequent elimination/reduction was a little more
tricky than anticipated. When the mixture of 7a/7b was
acetylated under acidic conditions (using Ac2O andTsOH),
spontaneous elimination occurred when the compound
a Key: (a) H2, Pd/C, Boc2O; (b) LiAlH4; (c) Boc2O; (d) Ac2O, TsOH;
(e) NaBH4; (f) Boc2O, Et3N.
(11) Shing, T. K. M.; Wong, A. W. F.; Ikeno, T.; Yamada, T. J. Org.
Chem. 2006, 71, 3253–3263.
Org. Lett., Vol. 13, No. 11, 2011
2909

The nitro group was reduced, and upon Boc-protection
the derivative 10 was isolated in 47% yield; reduction
under hydrogenation conditions furnished 11 (37% yield).
The reduction of 8, on the other hand, gave the desired
reaction and stereoisomer. When lithium aluminum hy-
dride was used, a 53% yield of 12 was obtained after Boc
protection of the amine. Yet since the silica-promoted
elimination reaction also gave a modest yield, the synthesis
of 12 from 7a/7b by this route had on overall yield below
30%. A problem may be that LiAlH4 can reduce the nitro
group before conjugate reduction has occurred. Therefore,
experimentation with sodium borohydride was also done
with the crucial difference here being that NaBH4 does
promote elimination of the acetate and subsequent con-
jugate addition but does not reduce the nitro group.
Treatment of the acetylated 7a/7b (Scheme 3) with NaBH4
led to three compounds: the epimeric pair of saturated
nitro derivatives 13 and 14 was obtained in a 5:1 ratio and a
30% combined yield, while the alkene 8 was obtained in
38% yield.
stereochemistry. With the rather high selectivity toward
13, a one-pot procedure was devised from 7a/7b consisting
of acetylation, NaBH4 treatment, LiAlH4 reduction, and
Boc protection (Scheme 3). This gave reproducibly 12 in
43% yield, which is significantly better than via 8.
From 12, hydrogenolysis with palladium on carbon
gave the hemiaminal 15 with the Boc group on nitrogen
and a yield of 95% (Scheme 4). The released C-5 hydro-
xyl could now be silylated with tert-butyldiphenylsilyl
chloride and imidazole in DMF, giving silyl ether 16 in
85% yield. Oxidation of the aminal to lactam 17 was
carried out with TEMPO, m-chloroperbenzoic acid, and
tetrabutylammonium bromide in dichloromethane to
give the product in 81% yield. This lactam was reacted
with Lawesson’s reagent, which led to formation of
thiolactam with loss of the Boc group. Yields in this
reaction were variable, giving from 40 to 70% of 18.
Disappointingly, we were not able to convert 18 to
amidine in a manner similar to Papandreau and Ganem.5
Reaction of 18 with ammonia under a variation of condi-
tions, including benzylamine, hydrazine, or hydroxyla-
mine, led to complex mixtures apparently caused by
partial opening of the lactam ring and possibly epimeriza-
tion as well. Eventually, the strategy using the thiolactam
18 had to be abandoned, which was not a big loss since its
synthesis was not very reliable.
Compound 13 could be converted to 12 by reduction
with LiAlH4 and Boc protection, confirming its
Scheme 4. Synthesis of Isofagomidine 4a
Fortunately, it was found that the lactam 17 quite easily
reacts with ammonia with ring-opening and formation of a
primary amide. Reaction of 17 with a mixture of aqueous
ammoniaꢀdioxane gave 19 in 82% yield (Scheme 4).
Dehydration of 19 using Swern conditions led to a 91%
yield of nitrile 20. Now removal of the Boc group with
trifluoroacetic acid led to the aminonitrile that with Cu-
catalysis cyclized to the amidine. The copper salt com-
plexes the amidine quite powerfully but with a subsequent
treatment with H2S gas the copper is precipitated as the
sulfide and removed conveniently so that the amidine
product 21 can be obtained pure in 77% yield. Finally,
deprotection was carried out using concentrated hydro-
chloric acid, which led to 4 (as the hydrochloride) in 81%
yield (Scheme 4).
Partial atomic and group charges of selected atoms in
isofagomine (3), isofagomidine (4), and the lactam 22 were
calculated with the CHelpG procedure of the Gaussian 03
program, which fits atomic charges to the molecular electro-
static potential with constraints on the electric dipole
Table 1. Atomic and Group Charges Calculated with the CHelpG
Procedure at the DFT-B3LYP/6-31G Level on DFT-B3LYP/6-31G-
Optimized Geometries
structure
N-1 NH2-1 C-2 CH2-2
N-2
NH2-2
isofagomine (3)
ꢀ0.2
0.4 ꢀ0.2
0.1
isofagomidine (4) ꢀ0.5 ꢀ0.1
0.5
0.6
ꢀ0.8
0.1
ꢀ0.6a
a Key: (a) H2, Pd/C; (b) TBDMSCl, imidazole, DMF; (c) MCPBA,
TEMPO; (d) Lawesson’s reagent; (e) NH3; (f) (COCl)2, DMSO, Et3N;
(g) TFA; (h) CuCl; (i) H2S; (j) HCl, 60 °C, 24 h.
lactam 22
ꢀ0.7 ꢀ0.3
a Partial change of the carbonyl oxygen in the 2-position.
2910
Org. Lett., Vol. 13, No. 11, 2011

Compound 4 was tested for inhibition of seven commer-
cially available glycosidases (Table 2), and the values were
compared with structurally related isofagomine 3 and
isofagomine lactam 22. Although compound 4 quite re-
sembles the structure of both 3 and 22, it is clear from the
data inhibitory properties of 4 is very different: Isofago-
midine 4 is basically a weaker inhibitor than 3 and 22 of all
the tested glycosidases except for R-mannosidase, where it
is much stronger (Ki = 0.75 μM). Change in pH and
incubation time (to see the role of slow onset inhibition12)
had no major effect (Table 2).
Table 2. Inhibition Constants (Ki, μM) of 3, 22, and 4
The cause of the remarkable differences in binding
between 3, 4, and 22 seems to be caused by the very
different charge distribution in the three compounds
(Table 1). The slightly positive group charge on the
exocyclic amino group in 4 is the most profound difference
in the charge distributions and is therefore likely to play a
role.
In summary, this paper shows a successful synthesis of
novel mannosidase inhibitor isofagomidine 4. Sugar ami-
dines of this type are rare, and our synthesis presents a new
way of reaching such compounds. As an inhibitor 4 is
strong versus R-mannosidase and in other ways is different
from the structurally similar 3 and 22.
a From almonds. b From baker’s yeast. c From green coffee beans.
d From Aspergillus oryzae. e From E. coli. f From jack beans. g From
Helix pomatia. h Enzyme and inhibitor preincubated for 30 min.
i Enzyme and inhibitor preincubated for 20 h. j At pH 4.0.
ꢀ
Acknowledgment. We thank FNU, DCSC, Direccion
General de Investigacion of Spain (CTQ2008 02813), and
ꢀ
Junta de Andalucıa (FQM 134) for support. We also thank
Prof. Jesper Bendix (KU) for helpful discussions.
moment. Atomic and group charges for the atoms in the
1- and 2-positions (see Figure 1) are shown in Table 1.
The calculations show that the more stable isomer of
isofagomidine (4) has atom and group charges at N-1
that are more negative than 3 and more similar to those
of the lactam 22. The partial charge at C-2 is also similar
to the lactam, but 4 has a slightly positive group charge
(þ0.1) for NH2-2 contrary to that of O-2 of 22. Thus, the
compound resembles neither 3 nor 22 completely, but is
more akin to the latter.
Supporting Information Available. Experimental pro-
cedures and copies of NMR spectra of new compounds.
This material is available free of charge via the Internet at
http://pubs.acs.org.
(12) (a) Lohse, A.; Hardlei, T.; Jensen, A.; Plesner, I. W.; Bols, M.
€
Biochem. J. 2000, 349, 211–215. (b) Bulow, A.; Plesner, I. W.; Bols, M.
Biochim. Biophys. Acta 2001, 1545, 207–215.
Org. Lett., Vol. 13, No. 11, 2011
2911