Active site protonation of 1-azafagomine in glucosidases studied by solid-state NMR spectroscopy

Article abstract of DOI:10.1002/ejoc.200600816

Protonation of an azasugar inhibitor inside the active site of a glycosidase has been studied by solid-state NMR spectroscopy. By measuring ¹³C chemical shifts of azafagomine bound to yeast α-glucosidase, almond β-glucosidase, and Aspergillus n

Full text of DOI:10.1002/ejoc.200600816

FULL PAPER
DOI: 10.1002/ejoc.200600816
Active Site Protonation of 1-Azafagomine in Glucosidases Studied by Solid-
State NMR Spectroscopy
Astrid C. Sivertsen,[a] Magdalena Gasior,[a] Morten Bjerring,[a] Steen U. Hansen,[a]
Oscar Lopez Lopez,[a] Niels C. Nielsen,*[a] and Mikael Bols*[a]
Keywords: Azasugars / Carbohydrates / Inhibitors / Enzymes / Molecular recognition
Protonation of an azasugar inhibitor inside the active site of
a glycosidase has been studied by solid-state NMR spec-
troscopy. By measuring 13C chemical shifts of azafagomine
bound to yeast α-glucosidase, almond β-glucosidase, and
Aspergillus niger glucoamylase, and 15N chemical shift of
azafagomine bound to β-glucosidase, we find evidence for
an N1-protonation of azafagomine inside β-glucosidase. For
α-glucosidase and glucoamylase the corresponding chemical
shifts are similar to those of the non-protonated inhibitor,
which shows that charge stabilization at anomeric position is
not occurring in these enzymes.
(© Wiley-VCH Verlag GmbH & Co. KGaA, 69451 Weinheim,
Germany, 2007)
protonated form mimic the charge of oxocarbenium ion or
carbocation transition states.
Introduction
Carbohydrates linked to proteins and lipids are ubiqui-
tous in biology. For example, they fine-tune the processes
of self and non-self distinction, cancer metastasis, inflam-
matory and immune response such as diabetes, and the abil-
ity of HIV, Hepatitis and other vira to assemble and bud
from host cells. The emerging field of glycobiology uses
carbohydrate-structures as vaccines against carbohydrate-
binding proteins and carbohydrate-processing enzymes.[1]
There is a great interest in elucidating the regulatory mecha-
nism of such enzymes, which typically involve specific li-
gand binding. Understanding what happens to inhibitors
upon active-site binding in terms of structural and confor-
mational changes, charge dislocation, and protonation is re-
garded an important step towards understanding enzymatic
reaction mechanisms. With the general belief that a central
element in enzymatic catalysis is the ability of the enzyme
to lower the energy of the transition state for the catalyzed
reaction,[2] much effort has been devoted to design of com-
pounds that resemble the expected transition state of glyco-
side cleavage in terms of polarity and shape to create potent
enzyme inhibitors.
However, despite the fact that the protonation states are
critical to the interpretation of inhibitor strength very little
is known about them. Inhibition profiles as a function of
pH are ambiguous because they show a combination of the
protonation behaviour of free enzyme, inhibitor and com-
plexЈ pKa values. Direct observation of protonation states
is therefore necessary. This is, however, not trivial. Davies
and collaborators have used atomic-resolution X-ray crys-
tallography to study complexes of several basic inhibitors
in glycosidases.[7] They showed that 4-O-(β--glu)-isofago-
mine, an 1-azasugar, was protonated in the active site of
endocellulase Cel5A and bound to the dicarboxylate form
of the enzyme (E).[8] This result, protonated inhibitor (IH)
binding to E, was consistent with the pH/Ki profile of the
inhibitor. On this basis Davies concluded that this isofago-
mine could not be a transition state analogue since it did
not bind to the catalytic form of the enzyme.[8] On the other
hand, in two other complexes between xylanase Xyn10A
and 4-O-(β--xyl)--xylo-1-deoxynojirimycin or 4-O-(β--
xyl)--xylo-isofagomine protonation states could not be de-
termined despite atomic resolution.[9] At present the pro-
tonation states of iminosugars have not been reported.
13C or 15N cross-polarization (CP) magic-angle-spinning
(MAS) solid state NMR of inhibitor complexes using 13C
or 15N labelled inhibitors are methods that may be used
to study protonation states of bound glycosidase inhibitors.
Although several isotope labelled glycosidase inhibitors
(Figure 1) have been reported no such study has apparently
been made. In this area of research, Vasella and collabora-
tors prepared a 15N labelled glucoimidrazol and used it to
distinguish tautomeric forms,[10] and Stütz et al. have pre-
pared a 13C-labelled 1-deoxynojirimycin.[11]
1-Azasugars, characterized by having a nitrogen in place
of anomeric carbon, and iminosugars, characterized by hav-
ing a nitrogen in place of ring-oxygen, range among the
most potent inhibitors of glycosidases and related en-
zymes.[3,4] These inhibitors have often been considered tran-
sition-state analogues of glycoside cleavage[5,6] since they in
[a] Center for Insoluble Protein Structures (inSPIN), Interdisci-
plinary Nanoscience Center (iNANO) and Department of
Chemistry, University of Aarhus,
8000 Aarhus, Denmark
Supporting information for this article is available on the
WWW under http://www.eurjoc.org or from the author.
Eur. J. Org. Chem. 2007, 1735–1742
© 2007 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim
1735

N. C. Nielsen, M. Bols et al.
FULL PAPER
chose in this study the glycosidase inhibitor, 1-azafagomine
(1) for several reasons: 1) It is a potent inhibitor (Ki 0.1–
10 µ) of both α- and β-glycosidases.[12,13] 2) It has a low
pKa value making it largely unprotonated at assay condi-
tions. This has the advantage that protonation under these
conditions is unlikely unless being induced specifically by
the enzyme. 3) It has two protonation sites, either at a
pseudo-anomeric position or at a pseudo-ring oxygen posi-
tion. This means that this compound may reveal infor-
mation about an enzymeЈs preference for protonation. It
also means that this compound potentially can act as a
“stand-in” for both 1-azasugar and iminosugar glycosidase
inhibitors.
In the present work we report the synthesis of 1b, a 15N-
labelled version of 1, and the results of 13C- and 15N (CP/
MAS) solid-state NMR with this and 3-13C-labelled 1a
bound to some glycosidases. The results of the study are
not conclusive, but suggest that 1a is protonated in the
active site of β-glucosidase and neutrally bound in α-gluco-
sidase and glucoamylase.
Figure 1. Glycoside hydrolysis transition states and glycosidase in-
hibitors.
Results and Discussion
Synthesis. Compounds 1 and (Ϯ)-1a were prepared as
In the present work, it was the intention to see whether previously reported.[13,15] Compound 1b, a 2-(15N)-labelled
NMR could be used to give insight into this problem and analogue of 1, was prepared as outlined in Scheme 1. This
reveal protonation states of nitrogen containing glycosidase synthesis is related to our original synthesis of 1, but with
inhibitors. These protonation states would have an influ- the important difference that the N–N bond has to be syn-
ence on the transition state mimicry of such inhibitors. We thesised to ensure control over the position of 15N. So -
Scheme 1. Synthesis of 15N-labelled azafagomine 1b.
1736
www.eurjoc.org
© 2007 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim
Eur. J. Org. Chem. 2007, 1735–1742

Study of Active Site Protonation of 1-Azafagomine in Glucosidases
FULL PAPER
xylose was converted into the tribenzylxylofuranose 2 as Ki(real) = E*1H+/EI becomes 2.1 µ, while the latter case
previously, but then oxidised to the lactone 3 using PCC. Ki(real) = EH*1/EI becomes 84 nM. The curve is not com-
The lactone was treated with ammonia in methanol to give patible with 1H binding EH or 1 binding EH2.
the amide 4 in 99% yield, which was reduced with LAH to
give the amine 5 in 86% yield. This two-step protocol for
reaching 5 from 3 is much better than reductive amination
of 3, which if carried out with ammonia leads to a signifi-
cant amount of overalkylation, and when done with pri-
mary amines, like allylamine and p-methoxybenzylamine,
necessitates some troublesome deprotection chemistry at N.
Compound 5 was acetylated with acetic anhydride in meth-
anol to give N-acetamide 6 in 90% yield.
Now introduction of 15N was carried out by nitrosation
of the amide 6 using Na15NO2 in a mixture of acetic acid
and acetic anhydride giving the N-nitroso compound 7 in
77% yield. The unprotected alcohol was mesylated with
mesyl chloride in pyridine giving 8 in 89% yield. Selective
reduction of the N-nitroso function to a hydrazine, a rather
1 as a function of pH.
tricky reaction, was realised with TiCl3/Mg to give a prod-
Figure 3. Curve of 1/Ki for inhibition of almond β-glucosidase by
uct that was immediately cyclised with base to give 9 in
For the β-glucosidase a similar relationship is found:
53% yield over the two steps. The final two deprotection Here the enzyme pH optimum was 5.75 with acidic and
steps, hydrogenolysis and acidic hydrolysis, followed the nucleophilic pKa values of 4.8 and 6.7. Azafagomine (1)
previous synthesis of 1 and gave a 59% yield of 1b, having inhibition has optimum at pH 6.0 and the observed Ki at
the expected mass and a 15N NMR chemical shift of this pH is 0.2 µ. Again, both the case of 1 binding to EH
83.1 ppm.
or 1H+ binding to E will theoretically fit this curve, when
Inhibition profile of 1 as a function of pH. At a series of the above pKa values and the pKa of 5.3 for 1 is used in the
different pH, we measured the kcat/Km and the Ki values for calculation, while other possibilities, such as 1H binding
the inhibition by 1 of yeast α-glucosidase (see Figure 2 and EH or 1 binding EH2 do not fit. In the case of 1 binding
S1 in the supporting information) and almond β-gluco- to EH the real Ki(real) = E*1H+/EI becomes 0.13 µ, while
sidase (Figure 3 and S2) obtaining relationships that reflect for 1H+ binding to E the Ki(real) = EH*1/EI becomes
protonation changes in inhibitor and enzyme active site. 5.2 nM.
From the kcat/Km curve (see supporting information) it is
Study of binding of 1a. Firstly, we studied the binding of
seen that the α-enzyme has a pH optimum of 6.1, and that the inhibitor (Ϯ)-3-(13C)-4,5-dihydroxy-3-(hydroxymethyl)-
the pKa values of the protonating and nucleophilic carbox- hexahydropyridazine[15] (1a) to three different sugar-pro-
ylate residues in its active site are 5.5 and 6.7, respectively. cessing enzymes: yeast α-glucosidase, almond β-glucosidase,
The inhibition profile for 1 with the α-glucosidase is sig- and Aspergillus niger glucoamylase.[14] These enzymes were
moid shaped with a pH optimum at 6.3, and an observed chosen because iminosugars are potent inhibitors of α-glu-
optimum Ki of 3.6 µ. A simulation using the pKa values cosidase, while 1-azasugars are potent inhibitors of β-gluco-
of 5.5 and 6.7 for the enzyme and 5.3 for 1 shows that sidase. It has been proposed that the charge development
this inhibition behaviour fits well with either 1 binding to occurring in the α- and β-glucosidase transition states dif-
monoprotonated enzyme (EH) or protonated 1 (1H+) bind- fers, with the former having oxocarbenium ion character
ing to unprotonated enzyme (E). In the former case the real and the latter carbocation character.[6] As mentioned above,
1 inhibits all three enzymes and this may be due to the
compoundЈs ability to mimic several transition states by
having two possible protonated forms. The large enzyme-
inhibitor complexes tumble too slowly for detection of en-
zyme-bound 1a using traditional liquid-state NMR experi-
ments,[15] (For some examples of protein-carbohydrate com-
plexes that can be studied by NMR in solution, see
ref.[16,17]) and the exchange rate of the inhibitor between
the free and enzyme-bound state is too slow for indirect
observation of the bound inhibitor by means of transferred
NOE experiments (data not shown). Solid-state NMR is
not dependent on fast molecular tumbling, and is therefore
applicable in investigation of these systems.
By appropriate tuning of the dipolar-coupling mediated
1
1H to 13C (or H to 15N) polarization transfer conditions
Figure 2. Curve of 1/Ki for inhibition of yeast α-glucosidase by 1
as a function of pH.
relative to the molecular motion, CP/MAS solid-state
Eur. J. Org. Chem. 2007, 1735–1742
© 2007 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim
www.eurjoc.org
1737

N. C. Nielsen, M. Bols et al.
FULL PAPER
NMR offers the possibility to distinguish ligands in the
binding site from those non-specifically bound to the en-
zyme or free in the solvent. In the present case, such condi-
tions were met by freezing an aqueous solution of the en-
zyme with different amounts of ligand to –42 °C and using
CP/MAS with 3.37 kHz spinning, 1 ms contact time, and 4
s relaxation delay. Specifically, the discrimination of signals
from frozen non-bound ligands and enzyme-bound ligands
was established by carefully optimizing the CP/MAS condi-
tions at different temperatures for samples with and without
enzyme present to make preference to signals from bound
ligands. The discrimination of signals from bound and non-
bound ligands was subsequently tested by titrating in ex-
cessive amounts of unlabelled ligand, where attenuation/eli-
mination of the signal from the 13C-labelled ligand in agree-
ment with the number of binding sites verifies that it is not
(or much less) observable under displaced (unbound) condi-
tions.
Figure 4. A series of 13C CP/MAS spectra of α-glucosidase, β-glu-
cosidase, and glucoamylase with different amounts of 1a[18,19] and
1[13] titrated into the solution prior to freezing. For each enzyme
the lower row contains the relevant aliphatic region of the 13C CP/
MAS spectrum for the enzyme, while the four upwards following
rows contain spectra for the enzyme upon addition of approx. 1,
2, and 3 equiv. of 1a, and finally further 40–80 equiv. of 1. To dis-
criminate the ligand 13C resonance from the large natural abun-
dance 13C background of the enzyme, we have for each enzyme
included difference plots obtained by subtracting the pure enzyme
spectrum (lower row) from the ligand-titrated spectra (left) to the
right of each spectrum in Figure 4.
In the 13C CP/MAS spectra for β-glucosidase, the ligand
displays a quite narrow resonance at δ = 60.2 ppm after
addition of 1 equiv. of 1a (Table 1). Upon further addition
of 1 and 2 eqs. of ligand the overall intensity of the peak
increases, as manifested by an increase in the peak height
for the narrow resonance as well as an increase in the line
width near the baseline. The latter contribution is attributed
to 1a non-specifically bound to β-glucosidase, while the
narrow component is assigned to Figure 4. 13C CP/MAS
Study of binding of 1b. The chemical shifts of the two
nitrogen sites in 1b were measured in neutral and acidic
solution (see Table 2).
spectra of (a) yeast α-glucosidase, (b) almond β-glucosidase, Table 2. 15N chemical shifts (relative to external liq. NH3) of 1b in
and (c) Aspergillus niger glucoamylase with different
neutral and acidic solution.
amounts of 1a and 1. From bottom to top the spectra repre-
sent the enzyme (148, 126, 151 nmol in 60 µL H2O) without
1a and 1, with ca. 1 equiv. 1a (195, 105, 152 nmol), with ca.
2 equiv. 1a (371, 268, – nmol), with ca. 3 equiv. 1a (541, Acidic
473, 456 nmol), and ca. 3 equiv. 1a plus 40–80 equiv. (7.4,
8.4, 12.2 mmol) of 1. Left: the aliphatic region of the 13C
Chemical shift (ppm)
1-15
N
2-15
N
Neutral
77.2
73.8
85.3
84.9
The binding of the inhibitor 2-(15N)-4,5-dihydroxy-3-(hy-
CP/MAS spectrum (5000–15000 scans). Right: difference droxymethyl)hexahydropyridazine (1b) to β-glucosidase was
between the spectrum to the left and the bottom left spec- studied by various solid-state NMR methods. Figure 5
trum.
shows a 15N CP/MAS NMR spectrum of a 1b-β-gluco-
sidase complex at –21 °C, displaying a broad signal at 110–
130 ppm which is assigned to natural abundance backbone
amides of the enzyme, and a single signal at δ = 82.6 ppm,
which is assigned to isotope labeled 2-15N site from specifi-
cally bound 1b. It is noted that the two signals have a 3:1
relationship in intensities, as would be expected based on a
67 kDa molecular weight of β-glucosidase, combined with
0.37% natural abundance of 15N, and a 1:0.9 molar ratio
Table 1. 3-13C chemical shifts (relative to external TMS) and bind-
ing constants for 1a specifically bound to the three enzymes (see
text). The Ki values are in µ at 25 °C, pH 6.8 and were taken from
ref.[12,13]
β-Glucosidase
α-Glucosidase Glucoamylase
Ki (µM)
Chemical shift (ppm) 60.2
0.33
6
10
62.7
63.0
1a in the inhibitor binding site. This interpretation is sup- of enzyme and inhibitor in the sample.
ported by further addition of 65 eqs. of 1, which eliminates
To further explore the protonation properties of the ni-
the narrow resonance by dilution of the 13C labels in the trogens, 15N cross-polarization depolarization (CPD) ex-
binding site. Under the present experimental conditions 13
C
periments[20] were carried out on 15N-tBoc-valine (included
signals from 1a in the frozen solvent are excited much less as a control as it is known to contain an NH group), free
efficiently than those for specific enzyme bound 1a (and the 1b, and the 1b-β-glucosidase complex, shown in Figure 6.
non-bound are thereby invisible). Similarly, the spectra of The data were fit to Equation (1) for an NH group and
1a bound to α-glucosidase and glucoamylase reveals 13C Equation (2) for an NH2 group, where Td is a time constant
resonances at δ = 62.7 and 63.0 ppm, respectively, for spe- that describes magnetization decay caused by spin diffusion
cific bound 1a (Table 1). For comparison the 13C chemical between directly bonded and more distant protons, T2 de-
1
shift of 1a was 63.5 ppm in neutral solution and 61.7 ppm scribes decay caused by direct dipolar coupling between H
in acidic solution.
and 15N, and τ2 is depolarization period of the CPD experi-
1738
www.eurjoc.org
© 2007 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim
Eur. J. Org. Chem. 2007, 1735–1742

Study of Active Site Protonation of 1-Azafagomine in Glucosidases
FULL PAPER
Symmetry-based R187 and R185 pulse sequences were
1
2
also considered in order to determine the protonation ex-
1
tend of the nitrogens. An R18 sequence applied on the H
decoupling channel during acquisition on the 15N channel
allows recoupling of the heteronuclear 1H-15N dipolar
coupling combined with homonuclear H decoupling. The
Figure 5. 15N CP/MAS spectrum of 572 nmol almond β-gluco-
sidase in complex with 0.9 equiv. 1b. The spectrum was recorded
using 12000 scans, 5 kHz spinning, 1 ms CP contact time, and a 4
s relaxation delay.
1
15N spectrum is therefore highly dependent on the proton-
ation, and large splittings of the signals are observed. How-
ever, conclusions on the protonation could not be drawn
from the obtained spectra of the 1b–glucosidase complex
due to large overlaps between the broad – and now also
split – backbone signals and the signal of the isotope lab-
elled 2-15N site. Low signal-to-noise ratio in the spectra also
hampered these types of experiments.
ment.[18] Solid and dashed lines in Figure 6 represents fits
of the experimental data (+) to NH and NH2 spin systems,
respectively. For 15N-tBoc-valine, 1b, and protein backbone,
fits to an NH spin system (rmsd values of 0.11, 0.15, and
0.21) are clearly better than fits to an NH2 spin system
(rmsd values of 0.99, 0.50, and 0.36), consistent with the
known structure for all three compounds. The shown NH
curves correspond to TIS values of 46, 64, and 54 us and
Td values of 787, 535, and 1032 us. We note that these val-
ues are in the same range for the three NH compounds. For
the 1b-β-glucosidase complex, iterative fitting to the NH
and NH2 models result in relatively higher rmsd values of
0.47 and 0.44, with optimized TIS values of 48 and 60 µs
and Td values of 0.998 and 99.2 ms, respectively.
Discussion. The inhibition profiles at variant pH show
not only that the inhibition of α- and β-glucosidase by 1 is
dependent on pH, but also that protonation of the inhibitor
is important for the inhibition, since if 1 and 1H+ bound
equally well, the inhibition curve would be identical to the
kcat/Km curve. For both enzymes, the data are consistent
with 1 binding EH or 1H+ binding E (or a combination)
and since the latter possibility requires that the inhibitor is
considerably stronger than the observed Ki (because little E
and 1H+ are present at optimum pH) it may appear less
likely. In any case these results show that the bound inhibi-
tor enzyme complex contains only one proton, the position
of which, on the inhibitor or on the enzyme, can be a result
of proton exchange upon binding.
The 13C chemical shifts of 1a bound to α-glucosidase and
glucoamylase are remarkably similar, while the β-gluco-
sidase signal is shifted significantly. This strongly suggests
a different mode of binding in the two cases. Furthermore,
the two former shifts are very similar to the chemical shift
of 1a in neutral solution (δ = 63.5 ppm), indicating that
the bound form of 1a in α-glucosidase and glucoamylase is
uncharged 1a. On the other hand, the relatively low chemi-
cal shift of 1a bound to β-glucosidase is much closer to that
of 1a in acidic solution (δ = 61.7 ppm) suggesting that the
inhibitor is protonated in β-glucosidase. Since 1 in acidic
solution is a 2:1 mixture of the N1- and N2-protonated spe-
cies so that the chemical shift of 61.7 ppm is an average of
the two forms, and since protonation of amines mainly give
an upfield shift in the β-position,[22] the low value observed
in β-glucosidase (δ = 60.2 ppm) is strongly indicative of ex-
clusive protonation at N1. A disruptive effect could, how-
ever, be caused by an enzyme specific through-space influ-
ence on the chemical shift, e.g., through ring current effect
from an adjacent aromatic structural element.
Figure 6. 15N CPD/MAS depolarization data points (+) for a) 15N-
tBoc-valine, b) 1b, c) natural abundance protein backbone, and d)
1b-β-glucosidase complex. Data in a) were acquired at room tem-
perature, while those of b)–d) were recorded at –21 °C. The experi-
ments used 5 kHz spinning, 1 ms CP contact time, and 80 kHz SPI-
NAL64[21] decoupling. Solid and dashed lines illustrate best fits to
NH and NH2 groups, respectively.
Eur. J. Org. Chem. 2007, 1735–1742
© 2007 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim
www.eurjoc.org
1739

N. C. Nielsen, M. Bols et al.
FULL PAPER
From the 15N chemical shift data in Table 2, calculated
chemical shifts of hypothetical pure 1- and 2-protonated
species are 72.1 and 84.1 ppm (taking the 2:1 1- to 2-pro-
tonation rate into account). 84.1 ppm is close to both the
value of non-protonated 1b, 85.3 ppm, and to the measured
shift for β-glucosidase-bound 1b, 82.6 ppm. It is evident
that the 2-15N chemical shift does not change significantly
either upon protonation, or enzyme binding (where proton-
ation is not expected). The latter fact suggests indirectly
that the 13C chemical shift changes upon binding are not
governed by ring current effect of aromatic side chains, a
possible concern for an enzyme with unknown structure,
and therefore unknown side chain composition in the active
site.
Experimental Section
General: Solvents were distilled under anhydrous conditions. All
reagents were used as purchased without further purification.
Evaporation was carried out on a rotary evaporator with the tem-
perature kept below 40 °C. Glassware used for water-free reactions
was dried min. 2 h at 130 °C before use. Columns were packed with
silica gel 60 (230–400 mesh) as the stationary phase. TLC plates
(Merck, 60, F254) were visualized by spraying with cerium sulfate
(1%) and molybdic acid (1.5%) in 10% H2SO4 and heating till
coloured spots appeared. 1H-NMR, 13C-NMR and COSY were
carried out on a Varian Mercury 400 instrument. Chemical shifts
are given in ppm and referenced to internal SiMe4 (δH, δC = 0.00).
J values are given in Hz. Mass spectra were obtained on a Micro-
mass LCT-QTOF instrument. 13C CP/MAS solid-state NMR spec-
troscopic data were recorded on a Varian Inova 400 MHz spec-
The 15N CPD experiments were anticipated to be suitable
for distinguishing between different protonation states of
enzyme-bound inhibitors. It turns out, however, that by
using current models for depolarization in IS and I2S spin
systems (here NH and NH2 groups), it is not possible to
unambiguously distinguish the two protonation models, al-
though the expected NH model seems to give a slightly bet-
ter fit than NH2, and is described by more realistic spin
diffusion constants. The lack of conclusive information may
partly be due to a relatively low signal-to-noise ratio for
15N labelled azafagomine in the enzyme binding site.
Nevertheless, the chemical shift measurements are al-
together consistent with protonation occurring at N1 in β-
glucosidase, while a different situation, either no proton-
ation, or protonation at N2, exists in α-glucosidase. Since 1
is unprotonated at the pH of the NMR experiment it is
clear that β-glucosidase induces protonation in the active
site. This also means that the enzyme stabilizes the charged
form of the inhibitor more than the neutral form. Such sta-
bilizing interactions may also be used by the enzyme, in this
case almond β-glucosidase, to stabilize a charged transition
state relative to a neutral ground state. The results indicate
that the bound form of 1a is protonated at N1, which sug-
gest that this enzyme stabilizes charge at the anomeric posi-
tion.
trometer using a 5 mm double-resonance VT MAS probe. The 15
CP/MAS and CPD/MAS experiments were carried out using a
Bruker 700 MHz Avance NMR spectrometer using a 4 mm triple-
resonance MAS probe.
N
2,3,5-Tri-O-benzyl-L-xylono-1,4-lactone (3): To the solution of 2
(328 mg, 0.78 mmol) in dichloromethane (2 mL) was added solu-
tion of PCC (288 mg, 1.30 mmol) and Celite (288 mg) in dichloro-
methane (5 mL). This mixture was stirred overnight at room tem-
perature. The solution was filtered through Celite, concentrated
and subjected to flash chromatography (pure pentane to pen-
tane:Et2O, 7:3) to give 3 (308 mg, 95%), as a colorless oil. MS (ES):
m/z 441.1667 [M + Na]+. C26H26O5Na m/z 441.1672. 1H NMR
(CDCl3): δ = 7.42–7.21 (m, 15 H, CHPh), 5.05 (d, 1 H, Jgem
=
11.2 Hz, H-CHPh), 4.68 (d, 1 H, Jgem = 11.2 Hz, H-CHPh), 4.65
(d, 1 H, Jgem = 11.6 Hz, H-CHPh), 4.63–4.52 (m, 5 H, CH2Ph, 4-
H, 2-H), 4.37 (t, 1 H, J = 7.2 Hz, 3-H), 3.76 (dd, 1 H, J = 2.8 Hz,
Jgem = 10.8 Hz, 5a-H), 3.72 (dd, 1 H, J = 3.6 Hz, Jgem = 10.8 Hz,
5b-H).
2,3,5-Tri-O-benzyl-L-xylonic Amide (4): Compound 3 (309 mg,
0.74 mmol) was dissolved in saturated solution of ammonia in
methanol (8 mL). After stirring overnight, under N2, the reaction
mixture was concentrated to give 4 (320 mg, 99%), as a white solid.
MS (ES): m/z: 458.1952 [M + Na]+. C26H29O5NNa m/z 458.1943.
13C NMR (CDCl3): δ = 174.1 (C=O), 138.2, 138.1, 136.9 (Cipso),
128.9–128.0 (CHPh), 81.1, 80.4, 77.6, 75.6, 73.9, 73.6, 70.8, 70.7.
1H NMR (CDCl3): δ = 7.32–7.15 (m, 15 H, CHPh), 4.81–4.32 (m,
6 H, CH2Ph), 4.12 (d, 1 H, J = 4.4 Hz, 2-H), 4.05–4.02 (m, 1 H,
4-H), 3.93 (t, 1 H, J = 4.4 Hz, 3-H), 3.47 (dd, 1 H, J = 5.6 Hz,
Jgem = 9.6 Hz, 5a-H), 3.42 (dd, 1 H, J = 6 Hz, Jgem = 9.6 Hz, 5b-
H), 3.21 (br. d, 1 H, OH).
1-Amino-1-deoxy-2,3,5-tri-O-benzyl-L-xylitol (5): Compound 4
Conclusion
(344 mg, 0.79 mmol) was dissolved in THF (10 mL) under N2 at
room temperature. Lithium aluminum hydride (150 mg, 3.96 mmol)
was then cautiously added. The reaction mixture was refluxing for
next 7 h. The mixture was quenched by slowly addition of ethyl
The solid-state 13C NMR results show that the inhibitor
1a changes the chemical shifts in β-glucosidase, while it
does not when bound to α-glucosidase and glucoamylase. acetate, water (0.5 mL), 15% NaOH (0.5 mL) and water (1.3 mL).
Since 15N NMR results reveal that no ring-current or sim-
The resulting homogeneous solution was filtered and the remaining
ilar effects from the protein are influencing the 15N chemi-
organic layer was diluted with water and extracted with diethyl
cal shift significantly in β-glucosidase the 13C chemical shift
ether (3ϫ20 mL). Combined organic layers were dried (MgSO4),
and concentrated. The residue was subjected to flash chromatog-
change strongly suggest that the inhibitor is protonated at
raphy (pure EtOAc to EtOAc:MeOH, 1:1) to give 5 (290 mg, 86%),
N1 in β-glucosidase but not in the other enzymes. This me-
as a colorless oil. MS (ES): m/z 422.2159 [M + H]+. C26H32NO4
ans that the β-glucosidase active site stabilizes the proton-
ated form of the inhibitor and consequently that it must
also stabilize formation of positive charge in the substrate;
this appears to be at the anomeric position (N1) rather than
at the ring-oxygen position (N2).
m/z 421.2331. 13C NMR (CDCl3): δ = 138.4, 138.1, 137.9 (Cipso),
128.8–127.8 (CHPh), 77.0, 76.6, 73.7, 73.2, 72.7, 71.1, 67.1, 39.4 (C-
NH2). 1H NMR (CDCl3): δ = 7.45–7.18 (m, 15 H, CHPh), 4.58–
4.44 (m, 6 H, CH2Ph), 4.06 (dt, 1 H, J = 1.2 Hz, J = 6.8 Hz, 4-H),
3.74 (dd, 1 H, J = 1.2 Hz, J = 6.0 Hz, 5a-H), 3.64–3.51 (m, 3 H,
1740
www.eurjoc.org
© 2007 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim
Eur. J. Org. Chem. 2007, 1735–1742

Study of Active Site Protonation of 1-Azafagomine in Glucosidases
FULL PAPER
2-H, 3-H, 5b-H), 3.25 (dd, 1 H, J = 4.8 Hz, Jgem = 12.4 Hz, 1a-H),
3.00 (dd, 1 H, J = 2.8 Hz, Jgem = 12.4 Hz, 1b-H).
diethyl ether (4:1, 10 mL). Magnesium turnings (34.8 mg,
1.45 mmol) were added and the reaction mixture was stirred at
room temperature for 3 h. After cooling to –5 °C ethereal solution
of nitroso compound 8 (208 mg, 0.36 mmol) was added to the black
suspension and stirred for 10 min. Hydrochloric acid (0.3 ,
0.4 mL) was added, and reaction was stirred for another 25 min
at room temperature. The reaction mixture was made alkaline by
addition of diluted NaOH, filtered through Celite and extracted
with dichloromethane (100 mL). The organic phase was washed
with brine, dried (MgSO4) and concentrated. This residue was dis-
solved in nitromethane (10 mL), EtN(iPr)2 (1.1 mL, 6.32 mmol)
was added, and the mixture was heated at 60 °C for 18 h. The sol-
vent was evaporated, and the residue was subjected to flash
chromatography (pure CH2Cl2 to CH2Cl2:EtOAc, 1:4) to give 9
(347 mg, 53%) as a colorless oil. MS (ES): m/z 462.2591 [M +
H]+. C28H3315NNO4 m/z 462.2409. 13C NMR (CDCl3): δ = 173.0
(C=O), 138.6, 138.2, 137.9 (Cipso), 128.7–127.9 (CHPh), 78.0, 74.9,
1-(Acetylamino)-1-deoxy-2,3,5-tri-O-benzyl-L-xylitol (6): Com-
pound 5 (400 mg, 0.95 mmol) was dissolved in methanol (25 mL)
and acetic anhydride (6 mL) was added. The mixture was kept for
18 h at room temperature, then concentrated and subjected to flash
chromatography (EtOAc) to give 6 (393 mg, 90%), as a slowly crys-
tallizing, white solid. MS (ES): m/z 486.2256 [M
+
Na]+.
C28H33NO5Na m/z 486.2256. 13C NMR (CDCl3): δ = 170.6 (C=O),
138.2, 138.1, 138.0 (Cipso), 129.3–128.0 (CHPh), 78.1, 76.8, 74.4,
73.5, 72.6, 71.5, 69.4, 39.2, 23.4 (CH3). 1H NMR (CDCl3): δ =
7.28–7.18 (m, 15 H, CHPh), 5.85 (br. s, 1 H, NH), 4.70–4.44 (m, 6
H, CH2Ph), 4.05 (ddd, 1 H, J = 2.6 Hz, J = 3.6 Hz, J = 6.0 Hz, 4-
H), 3.76–3.74 (m, 1 H), 3.62 (dd, 1 H, J = 3.6 Hz, Jgem = 9.2 Hz,
5a-H), 3.55 (dd, 1 H, J = 2.6 Hz, Jgem = 9.2 Hz, 5b-H), 3.48 (dd,
1 H, J = 6.4 Hz, Jgem = 9.6 Hz, 1a-H), 3.43 (dd, 1 H, J = 6.0 Hz,
Jgem = 9.6 Hz, 1b-H), 3.35–3.29 (m, 1 H), 2.69 (br. s, 1 H, OH),
1.82 (s, 3 H, CH3).
1
73.6 (2C), 72.2, 67.2, 60.6, 45.0, 20.8 (CH3); H NMR (CDCl3): δ
= 7.29–7.12 (m, 15 H, CHPh), 4.79 (d, 1 H, Jgem = 11.2 Hz, H-
CHPh), 4.66 (d, 1 H, Jgem = 11.2 Hz, H-CHPh), 4.72–4.65 (m, 1
1-(Acetyl-nitroso-amino)-1-deoxy-2,3,5-tri-O-benzyl-
L-xylitol
(7):
Compound 6 (1.0 g, 2.17 mmol) was dissolved in glacial acetic acid
(1 mL) and acetic anhydride (5 mL) and cooled to –3 °C. Under
vigorously stirring 15N-sodium nitrite (379 mg, 5.43 mmol) was
slowly added. Then the reaction mixture was stirred for 10 h at
room temperature and subsequently poured over ice-water bath
and extracted with diethyl ether (4ϫ50 mL). Combined organic
layers were washed with 5% NaHCO3 several times until all acetic
acid was removed, then washed with water, dried (MgSO4) and
concentrated. The residue was subjected to flash chromatography
(pentane:Et2O, 4:6) to give 7 (826 mg, 77%) as a yellow oil. MS
(ES): m/z 516.2117 [M + Na]+. C28H3215NNO6Na m/z 516.2127.
13C NMR (CDCl3): δ = 174.8 (C=O), 138.3, 137.9, 137.8 (Cipso),
128.7–127.8 (CHPh), 77.6, 74.7, 74.2, 73.5, 73.0, 71.5, 68.7, 39.3
H), 4.53 (d, 1 H, Jgem = 11.2 Hz, H-CHPh), 4.47 (d, 1 H, Jgem
=
11.2 Hz, H-CHPh), 4.39 (d, 1 H, Jgem = 11.2 Hz, H-CHPh), 4.29
(d, 1 H, Jgem = 11.2 Hz, H-CHPh), 3.73 (ddd, 1 H, J = 3.2 Hz, J
= 5.6 Hz, J = 9.0 Hz), 3.68 (t, 1 H, J = 8.2 Hz), 3.63–3.57 (m, 2
H), 2.83–2.67 (m, 2 H), 2.08 (s, 3 H, CH3).
2-15N-(3R,4R,5R)-4,5-Dihydroxy-3-(hydroxymethyl)hexahydropyri-
dazine [(–)-2-15N-labelled-azafagomine, 1b]: To a solution of 9
(290 mg, 0.63 mmol) in EtOH (15 mL) was added HCl (1 ,
4.2 mL) and Pd/C (126 mg) and the solution was hydrogenated at
1 atm H2 until all starting material disappeared (4–5 h). The mix-
ture was filtered through Celite and concentrated to give a residue
containing (3R,4R,5R)-1-acetyl-4,5-dihydroxy-3-(hydroxymethyl)-
hexahydropyridazine. This was dissolved in HCl (6 , 15 mL) and
heated to 100 °C for 18 h. The solution was concentrated and sub-
jected to ion-exchange chromatography using Amberlite IR-120
resin (H+, 10 mL) and eluting with diluted NH4OH solution
(150 mL). Concentration of alkaline eluate gave 1b (55 mg, 59%)
as a solid. MS (ES): m/z 150.0849 [M + H]+. C5H1315NNO3 m/z
150.0895. 13C NMR (D2O): δ = 72.0, 71.6, 63.1, 59.7, 51.8. 1H
NMR (D2O): δ = 3.71 (ddd, 1 H, J = 5.6 Hz, J = 9.6 Hz, J =
10.8 Hz, 2-H), 3.56 (dd, 1 H, J = 3.2 Hz, Jgem = 10.4 Hz, 5a-H),
3.47 (dd, 1 H, J = 6 Hz, Jgem = 10.4 Hz, 5b-H), 3.21 (t, 1 H, J =
9.6 Hz, 3-H), 3.06 (dt, 1 H, J = 5.6 Hz, Jgem = 12.8 Hz, 1a-H), 2.55
[dddd, 1 H, J = 1.1 Hz (15N-1H), J = 3.2 Hz, J = 6.0 Hz, J =
9.6 Hz, 4-H], 2.45 (dd, 1 H, J = 10.8 Hz, Jgem = 12.8 Hz, 1b-H);
15N NMR (D2O): δ = 83.1 ppm.
1
(CH2-N), 22.6 (CH3). H NMR (CDCl3): δ = 7.41–7.03 (m, 15 H,
CHPh), 4.65–4.62, 4.47–4.36, 4.16–4.13 (m, 6 H, CH2Ph), 4.23 (dd,
1 H, J = 8.8 Hz, Jgem = 14.0 Hz, 5a-H), 3.99 (ddd, 1 H, J = 2.8 Hz,
J = 6.0 Hz, J = 6.4 Hz, 2-H), 3.78 (dd, 1 H, J = 3.4 Hz, Jgem
=
14.0 Hz, 5b-H), 3.63 (ddd, 1 H, J = 3.4 Hz, J = 5.2 Hz, J = 8.8 Hz,
4-H), 3.61 (dd, 1 H, J = 6.4 Hz, Jgem = 9.6 Hz, 1a-H), 3.51 (dd, 1
H, J = 2.8 Hz, J = 5.2 Hz, 3-H), 3.42 (dd, 1 H, J = 6.0 Hz, Jgem
=
9.6 Hz, 1b-H), 2.51 (s, 3 H, CH3).
1-(Acetyl-nitroso-amino)-2,3,5-tri-O-benzyl-4-O-methylsulfonyl-L-
xylitol (8): Compound 7 (820 mg, 1.66 mmol) was dissolved in pyri-
dine (40 mL) and cooled to 0 °C. Methanesulfonyl chloride
(162 µL, 2.09 mmol) was added and the reaction mixture was al-
lowed to reach room temperature over 1.5 h. Water (20 mL) was
added and the mixture was extracted with CH2Cl2 (2ϫ20 mL). Fi-
nally the combined organic layers were concentrated by coevapora-
tion with toluene. The residue was subjected to flash chromatog-
raphy (pentane:Et2O, 1:1) to give 8 (844 mg, 89%) as a yellow oil.
MS (ES): m/z 594.1901 [M + Na]+. C29H3415NNO8SNa m/z
594.1903. 13C NMR (CDCl3): δ = 174.9 (C=O), 137.7, 137.6, 137.5
(Cipso), 128.7–128.0 (CHPh), 80.7, 77.8, 74.9, 74.3, 73.5, 73.1, 67.0,
38.7 (CH2-N), 38.4 (CH3), 22.7 (CH3). 1H NMR (CDCl3): δ =
7.38–7.01 (m, 15 H, CHPh), 4.84–4.81 (m, 1 H, 4-H), 4.67–4.30 (m,
6 H, CH2Ph), 4.03 (dd, 1 H, J = 6.8 Hz, Jgem = 14.2 Hz, 1a-H),
3.92 (dd, 1 H, J = 4.4 Hz, Jgem = 14.2 Hz, 1b-H), 3.66 (dd, 1 H, J
= 3.6 Hz, Jgem = 11.2 Hz, 5a-H), 3.59–3.54 (m, 2 H, H-3, 2-H),
3.51 (dd, 1 H, J = 6.8 Hz, Jgem = 11.2 Hz, 5b-H), 2.95 (s, 3 H,
CH3), 2.53 (s, 3 H, CH3).
Supporting Information (see also the footnote on the first page of
this article): Figure S1 and S2 showing curves of kcat/Km for α- and
β-glucosidases (2 pages) are available.
Acknowledgments
This work has been supported by the Danish Biotechnological In-
strument Center, Carlsberg Fonden, the Lundbeck Foundation,
Novo Nordisk fonden, the Danish Natural Science Foundation and
the Danish National Research Foundation. We are grateful to
Anne Bülow for assistance with sample preparation.
[1] J. Alper, Science 2001, 291, 2338–2343.
[2] L. Pauling, Am. Sci. 1948, 36, 51–58.
[3] N. Asano, R. J. Nash, R. J. Molyneux, G. W. J. Fleet, Tetrahe-
dron: Asymmetry 2000, 11, 1645–1680.
(3R,4R,5R)-1-Acetyl-4,5-bis(benzyloxy)-3-(benzyloxymethyl)hexa-
hydropyridazine (9): Titanium(IV) chloride (159 µL, 1.45 mmol)
was slowly added under N2 to the solution of dichloromethane and
Eur. J. Org. Chem. 2007, 1735–1742
© 2007 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim
www.eurjoc.org
1741

N. C. Nielsen, M. Bols et al.
FULL PAPER
[4]
which we have calculated a purity of 75, 95, and 67%, respec-
tively, by assuming one active site.
V. Lillelund, X. Liang, H. H. Jensen, M. Bols, Chem. Rev. 2002,
102, 515–553.
G. W. J. Fleet, Tetrahedron Lett. 1985, 26, 5073–5076.
M. Bols, Acc. Chem. Res. 1998, 31, 1–8.
A. Varrot, M. Schuelein, M. Pipelier, A. Vasella, G. J. Davies,
J. Am. Chem. Soc. 1999, 121, 2621–2622.
A. Varrot, C. A. Tarling, J. M. MacDonald, R. V. Stick, D. L.
Zechel, S. G. Withers, G. J. Davies, J. Am. Chem. Soc. 2003,
125, 7496–7497.
[15] S. U. Hansen, I. W. Plesner, M. Bols, ChemBioChem 2000, 1,
[5]
[6]
[7]
177–180.
[16] A. García-Herrero, E. Montero, J. L. Muñoz, J. F. Espinosa,
A. Vián, J. L. García, J. L. Asensio, F. J. Cañada, J. Jiménez-
Barbero, J. Am. Chem. Soc. 2002, 124, 4804–4810.
[17] J. F. Espinosa, E. Montero, A. Vian, J. L. García, H. Dietrich,
R. R. Schmidt, M. Martín-Lomas, A. Imberty, F. J. Cañada, J.
Jiménez-Barbero, J. Am. Chem. Soc. 1998, 120, 1309–1318.
[18] In this study (Ϯ)-1a was used, which practically means (–)-1a
because the corresponding (+)-form is a very weak inhibitor
(Ki ϾϾ1 m) of these enzymes.
[8]
[9] T. M. Gloster, S. J. Williams, S. Roberts, C. A. Tarling, J. Wicki,
S. G. Withers, G. J. Davies, Chem. Commun. 2004, 1794–1795.
[10] R. Hoos, A. B. Naughton, W. Thiel, A. Vasella, K. Rupitz,
S. G. Withers, Helv. Chim. Acta 1993, 76, 2666–2686.
[11] A. Berger, M. Ebner, A. E. Stuetz, Tetrahedron Lett. 1995, 36,
4989–4990.
[19] S. U. Hansen, M. Bols, J. Chem. Soc., Perkin Trans. 1 1999,
3323–3325.
[20] R. Sangill, N. Rastrup-Andersen, H. Bildsøe, H. J. Jacobsen,
N. C. Nielsen, J. Magn. Reson. Ser. A 1994, 107, 67–78.
[21] B. M. Fung, A. K. Khitin, K. Ermolaev, J. Magn. Reson. 2000,
142, 97–101.
[22] E. Pretch, W. Simon, J. Seibl, T. Clerc, Tables of Spectral Data
for Structure Determination of Organic Compounds, Springer,
Berlin, 1989.
[12] M. Bols, R. G. Hazell, I. B. Thomsen, Chem. Eur. J. 1997, 3,
940–947.
[13] B. V. Ernholt, I. B. Thomsen, A. Lohse, I. Plesner, K. B. Jen-
sen, R. G. Hazell, X. Liang, A. Jacobsen, M. Bols, Chem. Eur.
J. 2000, 6, 278–287.
[14] The enzymes were yeast α-glucosidase (Fluka, 63 kDa, 66 U/
mg), almond β-glucosidase (Sigma, 135 kDa, 25 U/mg) and
Aspergillus niger glucoamylase (Sigma, 80 kDa, 40 U/mg) for
Received: September 19, 2006
Published Online: December 12, 2006
1742
www.eurjoc.org
© 2007 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim
Eur. J. Org. Chem. 2007, 1735–1742