Inhibition of recombinant human maltase glucoamylase
by salacinol and derivatives
Elena J. Rossi
1,2,
*, Lyann Sim
1,2,
*, Douglas A. Kuntz
2
, Dagmar Hahn
3
, Blair D. Johnston
4
,
Ahmad Ghavami
4
, Monica G. Szczepina
4
, Nag S. Kumar
4
, Erwin E. Sterchi
3
, Buford L. Nichols
5
,
B. M. Pinto
4
and David R. Rose
1,2
1 Department of Medical Biophysics, University of Toronto, Canada
2 Division of Cancer Genomics and Proteomics, Ontario Cancer Institute, Canada
3 Institute of Biochemistry and Molecular Medicine, University of Berne, Switzerland

2006, accepted 13 April 2006)
doi:10.1111/j.1742-4658.2006.05283.x
Inhibitors targeting pancreatic a-amylase and intestinal a-glucosidases
delay glucose production following digestion and are currently used in the
treatment of Type II diabetes. Maltase-glucoamylase (MGA), a family 31
glycoside hydrolase, is an a-glucosidase anchored in the membrane of small
intestinal epithelial cells responsible for the final step of mammalian starch
digestion leading to the release of glucose. This paper reports the produc-
tion and purification of active human recombinant MGA amino terminal
catalytic domain (MGAnt) from two different eukaryotic cell culture sys-
tems. MGAnt overexpressed in Drosophila cells was of quality and quantity
suitable for kinetic and inhibition studies as well as future structural stud-
ies. Inhibition of MGAnt was tested with a group of prospective a-glucosi-
dase inhibitors modeled after salacinol, a naturally occurring a-glucosidase
inhibitor, and acarbose, a currently prescribed antidiabetic agent. Four
synthetic inhibitors that bind and inhibit MGAnt activity better than acar-
bose, and at comparable levels to salacinol, were found. The inhibitors are
derivatives of salacinol that contain either a selenium atom in place of
sulfur in the five-membered ring, or a longer polyhydroxylated, sulfated
chain than salacinol. Six-membered ring derivatives of salacinol and
compounds modeled after miglitol were much less effective as MGAnt
inhibitors. These results provide information on the inhibitory profile
of MGAnt that will guide the development of new compounds having
antidiabetic activity.
Abbreviations
HPA, human pancreatic a-amylase; MGA, maltase glucoamylase; MGAnt, maltase glucoamylase N-terminal catalytic domain; pNP, para-
nitrophenyl; SIM, sucrase isomaltase.
FEBS Journal 273 (2006) 2673–2683 ª 2006 The Authors Journal compilation ª 2006 FEBS 2673
postprandial hyperglycemia [3,4] (Fig. 1). Because
these inhibitors decrease both hyperglycemia and

2674 FEBS Journal 273 (2006) 2673–2683 ª 2006 The Authors Journal compilation ª 2006 FEBS
starch digestion. SIM accounts for almost all sucrase
activity, all isomaltase activity, and 80% of the maltase
activity, while MGA accounts for all glucoamylase
activity, 20% of the maltase activity, and 1% of the
sucrase activity [10]. Together, these two enzymes form
a complex in the epithelial cells of the small intestine
and complete the hydrolysis of oligosaccharide chains
in starch digestion.
Human MGA encoded by the gene MGAM [8,11] is
an a-glucosidase responsible for hydrolysis of a-1,4-
linkages from the nonreducing end of maltose oligo-
saccharides [9] and belongs to glycoside hydrolase
family 31. It is type II membrane protein 1857 amino
acids in length anchored in the brush border epithelial
cells of the small intestine. MGA contains five distinct
protein domains: a small cytosolic domain (26 amino
acids) a transmembrane domain (20 amino acids), an
O-glycosylated linker (or stalk) (55 amino acids), and
two homologous (family GH31) catalytic domains
(each 900 amino acids) [9]. The domain organization
is illustrated schematically in Fig. 2. Each MGA cata-
lytic domain contains a putative catalytic site made up
of the amino acid sequence tryptophan-X-aspartate-
methionine-asparagine-glutamate (WXDMNE), where
X indicates a variable amino acid. This catalytic site is
conserved in other human a-glucosidases and family
31 enzymes including SIM [12]. Human SIM is the clo-
sest known homologue of hMGA, sharing 59% amino
acid sequence identity, and is responsible for the hydro-

logue, blintol 5 (Fig. 1), have been shown to be very
effective in controlling blood glucose levels in rats
after a carbohydrate meal, thus providing lead candi-
dates for the treatment of Type 2 diabetes [23]. In
order to examine the mechanism of action of this
class of inhibitors, 3, 5, the stereoisomers of salacinol
6, 7 [25], and the six-membered ring analogues of
salacinol (8, 9) (N. S. Kumar and B. M. Pinto,
unpublished results) were synthesized along with
analogues of miglitol (10, 11) [22]. In view of the
reported antiglucosidase activity of 4 [14], we also
synthesized chain-extended analogues ( 12 –15) (Fig. 1)
[26], whose acyclic, polyhydroxylated, sulfated chains
varied between the four-carbon chain of salacinol and
the seven-carbon chain of kotalanol.
Fig. 2. Schematic diagram of MGA protein
organization and expression construct.
Amino acid boundaries of each of the
domains comprising the full size protein,
and the region inserted into the Drosophila
expression plasmid, are indicated.
E. J. Rossi et al. Inhibition of human maltase glucoamylase
FEBS Journal 273 (2006) 2673–2683 ª 2006 The Authors Journal compilation ª 2006 FEBS 2675
It has been difficult previously to carry out extensive
studies on the inhibitor profiles of these compounds
due to the lack of large amounts of active enzyme.
Here we report heterologous overexpression of recom-
binant DNA encoding the MGA amino terminal cata-
lytic domain (MGAnt) in Drosophila S2 cells in order
to overcome this difficulty. The purified recombinant

Whereas salacinol at 5 lm concentration inhibited 60–
70% of the breakdown of maltose, 5 lm acarbose only
inhibited 4% of the activity. Thus, it would appear that
acarbose acts mainly by inhibiting human pancreatic
a-amylase (HPA) and the breakdown of starch, and
possibly other intestinal glucosidases but not MGA.
The synthetic analogues of salacinol appeared to be
slightly more active than the parent compound in these
crude extracts. At 0.2 lm, blintol 5 inhibited 50% of
MGA activity, and the chain extended analogues
(13–15) inhibited 88%, 91%, and 90% of MGA activ-
ity, respectively, when tested at 5 lm.
Expression of active recombinant MGAnt
in Drosophila S2 cells
Due to limited expression levels in COS-1 cells and dif-
ficulties in purification of the resultant membrane
anchored protein we decided to express the catalytic
domain as a secreted protein in Drosophila melanoga-
ster cells (DES system, Invitrogen). We designed a
construct that lacked the cytosolic, transmembrane
region and most of the O-glycosylated stalk region that
occurs at the amino terminus (Fig. 2). The N-terminal
catalytic domain of MGA, starting at Ser87 and end-
ing at amino acid 955, was fused to a C-terminal hexa-
histidine tag. This domain was placed downstream of
a metallothionein promoter and behind a Bip secretion
signal. Correctness of the construct was determined by
sequencing in each direction. An active protein was
successfully expressed in Drosophila S2 cells. Secreted
protein was isolated from the cell media using chelat-

was used in a double reciprocal Lineweaver–Burk plot
(1 ⁄ velocity versus 1 ⁄ substrate) in order to calculate
the V
max
and K
M
of the reaction (Fig. 3). The V
max
was determined to be 32.6 ± 1.4 Units ⁄ mg enzyme
and the K
M
4.6 ± 0.5 mm maltose. This differs some-
what from the previously published results for purified
murine MGA (34.7 UÆmg
)1
, 1.24 mm, respectively) [27]
but it must be pointed out that the purified rodent
enzyme was almost twice the size of full size human
MGA and was composed of a number of disulfide-
linked proteolytic fragments [27]. The K
M
is close to
the 3.4 mm measured for human MGA immunoprecip-
itated from pooled clinical homogenates (B. Nichols,
unpublished results).
Inhibition analysis
The availability of larger amounts of recombinant
enzyme permitted a more thorough analysis of the
inhibitor activities than was possible with the COS-1
homogenates. The effectiveness of a-glucosidase inhibi-

acarbose, salacinol and its synthetic analogues (5, 13–
15) are listed (Table 1) and the Dixon plot visualiza-
tion given in Fig. 4. Salacinol and 15 showed the best
inhibition against MGA (K
i
¼ 0.2 lm) while acarbose
showed the worst inhibition (K
i
¼ 62 lm). These val-
ues are comparable to the preliminary data described
above in COS-1 cells, despite the differences in assays
and source of enzyme.
Discussion
Initial expression of active MGAnt protein in COS-1
cells demonstrated the validity of the cDNA clones,
but suffered from low yields and the difficulty in isola-
ting large quantities for physico-chemical studies. The
Drosophila S2 cell expression system proved to be a
successful method for the production of MGAnt in
substantial quantities. The N-terminal catalytic domain
was expressed and secreted into the medium, from
which it was purified with sufficient purity (> 95%)
and yield (> 40 mg ⁄ 3 L) for use in kinetic and inhibi-
tion analysis as well as future use in structural studies.
Kinetic analysis confirmed the enzyme activity of the
recombinant protein, and inhibition analysis confirmed
classic competitive inhibition by a-glucosidase inhibi-
tors. Salacinol with a K
i
of 0.2 lm was the best inhib-

Acarbose (1) 62.0 ± 13
Salacinol (3) 0.19 ± 0.02
5 0.49 ± 0.05
13 0.26 ± 0.02
14 0.25 ± 0.02
15 0.17 ± 0.03
E. J. Rossi et al. Inhibition of human maltase glucoamylase
FEBS Journal 273 (2006) 2673–2683 ª 2006 The Authors Journal compilation ª 2006 FEBS 2677
structure that is thought to mimic the oxocarbenium
ion intermediates in glycoside hydrolysis reactions [20].
There is a current debate as to whether carbohydrate
mimics containing sulfonium ions and ammonium ions
are effective inhibitors because of their ability to mimic
the shape and charge of the presumed transition state,
or because they bind with high affinity due to electro-
static interactions with a carboxylate residue in the
enzyme active site [16,18]. If electrostatic stabilization
is the key to enzyme affinity, inhibitors bearing a
permanent positive charge should function as well or
better than current glycosidase inhibitors, as proven by
the effectiveness of salacinol [16,18].
Inhibitors modeled after salacinol, all contain either
a sulfur or a selenium atom resulting in a permanent
positive charge in the five-membered ring. The differ-
ences between these seven salacinol analogues involve
the stereochemistry at the stereogenic centers in the
polyhydroxylated, sulfated chain, as well as the num-
ber of carbons in the acyclic chain linked to the
6
4

6
4
2
0
-3 -2 -1 0 1 2 3
1/A450
1/A450
-3 -2
-1
0
1
2
3
[14] (µ
M)
[15] (µ
M)
12
10
10
8
6
4
2
0
8
6
4
2
0

(Fig. 1), suggests that the positively charged five-mem-
bered ring is a better transition-state mimic because of
its ring shape [29,30].
A preliminary inhibition screen showed four com-
pounds of the group of salacinol analogues that
were the most potent inhibitors of MGAnt activity (5,
13–15) (Fig. 4). The common element of these four
derivatives is the identical stereochemistry at the carbon
centers in the heteroalditol ring to that of salacinol.
Inhibitor 5 is most similar to salacinol in that the only
alteration is the replacement of the ring sulfur atom by
selenium. Inhibitors 6 and 7, which were not effective
as inhibitors of MGAnt, differ from salacinol (3)in
stereochemistry at the carbon centers in the ring. These
results suggest that the stereochemistry at these centers
is critical for effective inhibition, the OH groups at
C-2 and C-3 interacting with complementary groups in
the enzyme active site. The five-membered carbon ring
is likely the portion of the molecule that is most
important in conferring affinity for the enzyme active
site. This conclusion is reinforced by the observation
that the four best inhibitors share three different
carbon chain lengths linked to the ring heteroatom,
suggesting that the chain length does not play a pre-
dominant role in the binding or effectiveness of the
inhibitors. Unfortunately, kotalanol, with the longest
chain length, was not available for this study. The ana-
lysis is clearly an oversimplification, since compound 12
was proven to be ineffective although it shares the same
ring stereochemistry as salacinol and compounds 5, 13,

i
of 15 nm [33].
However the method of action of acarbose is quite
complex and it appears to be acting as a type of sui-
cide inhibitor of a-amylase in a mechanism whereby
the acarbose is rearranged into an active entity by the
a-amylase [33]. Thus the acarbose itself is not the act-
ive inhibitor. The active rearranged entity may be
inhibitory to MGA and could be generated in the
intestinal scrapings by a-amylase present in the hetero-
geneous sample or by activity in the C-terminal
domain of the full-size protein, thereby accounting for
the inhibition by ‘acarbose’ reported previously [28].
Our previous studies of the inhibitory effect of salac-
inol and its derivatives against human a-amylase and
fungal glucoamylase, rather than MGA, report the
effectiveness of salacinol to be in the millimolar range
[16,18,20]. In addition the analogues 5 and 13–15 did
not inhibit human pancreatic a-amylase (S. G. Withers
and B. M. Pinto, unpublished results). The present
study reports activities of salacinol and synthetic deriv-
atives, against active human recombinant MGAnt. By
confirming the higher potency of salacinol and its
derivatives against human MGAnt as compared with
a-amylase and fungal glucoamylase, our results suggest
that the inhibitors show specificity towards different
a-glucosidases. This observation is important clinically
because the design of a-glucosidase inhibitors for the
treatment of Type II diabetes might require specificity
for enzymes later in the starch digestion pathway in

ity and affinity of these compounds towards their
potential development as antidiabetics.
Further confirmation of the importance of inhibitor
stereochemistry and how it affects binding in the active
site will only be possible with an analysis of the atomic
structure of the MGA binding site in both the presence
and absence of bound inhibitor. Determination of this
structural information, building on the groundwork
reported in this study, will be a valuable tool in future
design and synthesis of a-glucosidase inhibitors effect-
ive against and specific to MGA. These inhibitors
should be promising lead candidates as oral agents for
the treatment and prevention of Type II diabetes.
Experimental procedures
Intestinal maltase assay
Recombinant expression of C-terminally truncated human
MGA in COS-1 cells has been published [8]. The COS cells
transiently transfected with MGA-P1A2 were scraped off
the tissue culture plates in 150 mm KCl. Aliquots were so-
nicated and assayed for hydrolysis of 2% maltose for 2 h
at 37 °C by the Dahlqvist method [34]. The reaction was
stopped by boiling and glucose production was measured
by the glucose oxidase assay (below). Protein was measured
with a Bio-Rad (Hercules, CA, USA) protein assay kit.
Recombinant MGAnt in Drosophila S2 cells
The N-terminal catalytic domain of human MGA was
expressed in Drosophila cells. The coding sequence was
isolated from MGA-P1A2, which lacks the 903 amino acid
C-terminal domain [8]. We also chose to delete the base
pairs coding for the N-terminal cytosolic domain, the trans-

recombinant MGAnt. The S2 cells were maintained at
25 °C as a semiadherent monolayer in Schneider’s Insect
Medium (Sigma, St Louis, MO, USA) enriched with 10%
heat-inactivated fetal bovine serum (FBS). The cells were
split with enriched media at a ratio of 1 : 4 every 3–4 days
until transfection. The recombinant MGAnt vector was
transfected, in combination with the pCoBLAST selection
vector, which contains a blasticidin resistance cassette under
the control of the Drosophila copia promoter, into S2 cells
using the calcium phosphate procedure. Cells at a concen-
tration of 3 · 10
6
cellsÆmL
)1
were transfected with 19 lg
of expression vector and 1 lg of selection vector. The
procedure is carried out in duplicate to allow for one tran-
siently and a second stably transfected cell line. Transfected
cells were washed the next day with enriched medium to
remove the calcium phosphate solution. Two days later, the
transiently transfected cells were induced with 10 lm CdCl
2
and after a further three days, the cell medium was assayed
for protein expression by SDS ⁄ PAGE and immunoblotting
Inhibition of human maltase glucoamylase E. J. Rossi et al.
2680 FEBS Journal 273 (2006) 2673–2683 ª 2006 The Authors Journal compilation ª 2006 FEBS
with antipentaHis antibody (Qiagen, Montreal, Canada). In
order to obtain stably transfected cells, transfectants were
passaged for one month in selective medium [enriched med-
ium containing 16 lgÆmL

SDS ⁄ PAGE and the pNP-glucose activity assay to identify
fractions containing active MGAnt. These fractions were
pooled, concentrated, and dialyzed against 100 mm NaCl,
20 mm Tris pH 8.5 to remove residual copper and imidaz-
ole, and to lower the salt concentration of the sample in
preparation for ion exchange chromatography.
A BioCAD Poros-HQ anion exchange column (PerSep-
tive Biosystems, Framingham, MA, USA) was used to
further purify the MGAnt. The column was washed and
equilibrated with starting buffer, 100 mm Bis-Tris Propane
pH 7. Sample was diluted by half with 100 mm Bis-Tris
Propane pH 7 then was loaded on column and washed with
starting buffer. Sample was eluted over a linear gradient of
0–1 m NaCl. Eluate was collected in 3 mL fractions and
assayed for active MGA using SDS ⁄ PAGE and pNP-glu-
cose assay. Fractions containing pure, active MGAnt were
pooled and concentrated to 23 mgÆmL
)1
.
Inhibitors
Acarbose 1, salacinol 3 and synthesized derivatives were
analyzed as inhibitors for recombinant human MGAnt
using the glucose oxidase enzyme activity assays described
below. The inhibitors were dissolved in water as 50 mm
stock solutions and stored at )20 °C.
Enzyme activity assay
Two methods were used to assess MGAnt activity. For
rapid measurements of cell supernatants and assessment of
column fractions the pNP-glucose assay was used. For
detailed kinetic analysis the glucose-oxidase assay was used.

data to the Michaelis-Menten equation and estimate the
kinetic parameters, K
m
and V
max
, of the enzyme. K
i
values
for each inhibitor were determined by measuring the rate of
maltose hydrolysis by MGAnt at varying inhibitor concen-
trations. Data were plotted in Lineweaver-Burk plots
(1 ⁄ rate versus1 ⁄ [substrate]) and K
i
values for the compet-
itive inhibition were determined by the equation K
i
¼
K
m
[I] ⁄ (V
max
)s ) K
m
, where ‘s’ is the slope of the line. The
K
i
reported for each inhibitor was estimated by averaging
the K
i
values obtained from each of the different inhibitor

A & Laakso M (2002) Acarbose for prevention of type
2 diabetes mellitus: the STOP-NIDDM randomised
trial. Lancet 359, 2072–2077.
7 Scheen AJ (2003) Is there a role for alpha-glucosidase
inhibitors in the prevention of type 2 diabetes mellitus?
Drugs 63, 933–951.
8 Nichols BL, Avery S, Sen P, Swallow DM, Hahn D &
Sterchi E (2003) The maltase-glucoamylase gene: com-
mon ancestry to sucrase-isomaltase with complementary
starch digestion activities. Proc Natl Acad Sci USA 100,
1432–1437.
9 Nichols BL, Eldering J, Avery S, Hahn D, Quaroni A
& Sterchi E (1998) Human small intestinal maltase-glu-
coamylase cDNA cloning: homology to sucrase-isomal-
tase. J Biol Chem 273, 3076–3081.
10 Semenza G & Auricchio S (1989) The Metabolic Basis of
Inherited Disease II (Scriver, C R, Beaudet, A L, Sly, W S,
Valle, D, Stanbury, J B, Wyngaarden, J B & Fredrickson,
D S, eds), pp. 2975–2997. McGraw-Hill Inc., New York.
11 Roy R, Gautier M, Hayes H, Laurent P, Zaragoza P,
Eggen A & Rodellar C (2002) Assignment of maltase
glucoamylase (MGAM) gene to bovine chromosome
4q34 by in situ hybridization and confirmation by
radiation hybrid mapping. Cytogenet Genome Res 98,
311C.
12 Lee SS, He S & Withers SG (2001) Identification of the
catalytic nucleophile of the family 31 alpha-glucosidase
from Aspergillus niger via trapping of a 5-fluoroglyco-
syl-enzyme intermediate. Biochem J 15, 381–386.
13 Yoshikawa M, Murakami T, Shimada H, Matsuda H,

Synlett 9, 1259–1262.
20 Johnston BD, Ghavami A, Jensen MT, Svensson B &
Pinto BM (2002) Synthesis of selenium analogues of the
naturally occurring glycosidase inhibitor salacinol and
their evaluation as glycosidase inhibitors. J Am Chem
Soc 124, 8245–8250.
21 Liu H & Pinto BM (2005) Efficient synthesis of the glu-
cosidase inhibitor blintol, the selenium analogue of the
naturally occurring glycosidase inhibitor salacinol.
J Org Chem 70, 753–755.
22 Szczepina MG, Johnston BD, Yuan Y, Svensson B &
Pinto BM (2004) Synthesis of alkylated deoxynojirimy-
cin and 1,5-dideoxy-1,5-iminoxylitol analogues: polar
side-chain modification, sulfonium and selenonium het-
eroatom variants, conformational analysis, and evalua-
tion as glycosidase inhibitors. J Am Chem Soc 126,
12458–12469.
23 Pinto BM, Johnston BD, Ghavami A, Szczepina MG,
Liu H & Sadalapure K (2004) Glycosidase inhibitors
Inhibition of human maltase glucoamylase E. J. Rossi et al.
2682 FEBS Journal 273 (2006) 2673–2683 ª 2006 The Authors Journal compilation ª 2006 FEBS
and methods of synthesizing same. US patent
10/877490.
24 Pinto BM, Johnston BD, Ghavami A, Szczepina MG,
Liu H, Sadalapure K, Jensen HH, Kumar NS & Nasi R
(2006) Glycosidase inhibitors and methods of synthesiz-
ing same. US patent 11/368014.
25 Kumar NS & Pinto BM (2005) Synthesis of d-lyxitol and
d-ribitol analogues of the naturally occurring glycosidase
inhibitor salacinol. Carbohydr Res 340, 2612–2619.

M (2001) Synthesis of a nitrogen analogue of salacinol
and its alpha-glucosidase inhibitory activity. Chem
Pharm Bull (Tokyo) 49, 1503–1505.
33 Li C, Begum A, Numao S, Park KH, Withers SG &
Brayer GD (2005) Acarbose rearrangement mechanism
implied by the kinetic and structural analysis of human
pancreatic alpha-amylase in complex with analogues
and their elongated counterparts. Biochemistry 44,
3347–3357.
34 Dahlqvist A (1964) Method for assay of intestinal disac-
charidases. Anal Biochem 7, 18–25.
E. J. Rossi et al. Inhibition of human maltase glucoamylase
FEBS Journal 273 (2006) 2673–2683 ª 2006 The Authors Journal compilation ª 2006 FEBS 2683