Structural Basis for Substrate Selectivity in Human Maltase-Glucoamylase and Sucrase-Isomaltase N-terminal Domains*

Human maltase-glucoamylase (MGAM) and sucrase-isomaltase (SI) are small intestinal enzymes that work concurrently to hydrolyze the mixture of linear α-1,4- and branched α-1,6-oligosaccharide substrates that typically make up terminal starch digestion products. MGAM and SI are each composed of duplicated catalytic domains, N- and C-terminal, which display overlapping substrate specificities. The N-terminal catalytic domain of human MGAM (ntMGAM) has a preference for short linear α-1,4-oligosaccharides, whereas N-terminal SI (ntSI) has a broader specificity for both α-1,4- and α-1,6-oligosaccharides. Here we present the crystal structure of the human ntSI, in apo form to 3.2 Å and in complex with the inhibitor kotalanol to 2.15 Å resolution. Structural comparison with the previously solved structure of ntMGAM reveals key active site differences in ntSI, including a narrow hydrophobic +1 subsite, which may account for its additional substrate specificity for α-1,6 substrates.

anchored to the small intestinal brush-border membrane via an O-glycosylated stalk stemming from the N-terminal domain. Given that MGAM and SI genes arose from duplication and divergence of an ancestral gene, which itself has undergone tandem duplication (3), the N-terminal domains of MGAM and SI are more similar to one another in sequence, as are the C-terminal domains (ϳ60% sequence identity), than are the N-and C-terminal domains associated with the same enzyme (ϳ40% sequence identity).
Within the carbohydrate-active enzymes (CAZY) classification system (36), which groups enzymes based on sequence similarity and reflects the functional and structural similarities of family members, N-and C-terminal MGAM and SI domains are members of the glycoside hydrolase 31 family (GH31). The four domains exhibit exo-glucosidase activities against ␣-1,4linked maltose substrates ( Fig. 1A) but display different specificities for malto-oligosaccharides of various lengths (4 -6). ntSI and ctSI subunits have additional activity for the ␣-1,6 linkages of starch branch points (and isomaltose substrates; Fig.  1A) and the ␣-1,2 linkage of sucrose, respectively (7), and are historically referred to as isomaltase and sucrase.
As they are involved in the breakdown of dietary sugars and starches, MGAM and SI are attractive targets for inhibition by ␣-glucosidase inhibitors as a means of controlling blood glucose levels in individuals with type 2 diabetes (8). Acarbose (Fig.  1B) is the most widely used ␣-glucosidase inhibitor currently on the market and has been shown to be an efficient inhibitor of ␣-amylase (9) and of the C-terminal domains of MGAM and SI (4), but a weaker inhibitor of the N-terminal domains (4,10). More recently, studies on a new class of ␣-glucosidase inhibitors based on the natural extracts of the plant Salacia reticulata, including the active compounds salacinol, kotalanol (Fig. 1B), and de-O-sulfonated kotalanol, have been shown to be stronger inhibitors of ntMGAM compared with acarbose (10 -13).
A thorough understanding of the structural basis for substrate specificity in these intestinal ␣-glucosidases requires the analysis and comparison of crystal structures of the individual catalytic domains. Structural details of the active sites will give insight into how the MGAM and SI domains have evolved their substrate binding sites to accommodate different sugar and starch linkages.
We have previously solved the crystal structure of ntMGAM and analyzed co-crystal structures in complex with ␣-glucosi-dase inhibitors acarbose and salacinol-based compounds (6,13). From structural studies of the ntMGAM-acarbose complex, the active site of ntMGAM was found to be composed of a shallow substrate-binding pocket comprising Ϫ1 and ϩ1 subsites (6). Substrates bind to the pocket via their non-reducing end, with the non-reducing sugar ring interacting with the buried Ϫ1 subsite and the reducing ring interacting with the surface-exposed ϩ1 subsite. Substrate cleavage occurs between Ϫ1 and ϩ1 subsites and follows a catalytic mechanism, resulting in a net retention of configuration at the anomeric center.
Here we present the crystal structure of the N-terminal domain of human SI in apo form to 3.2 Å and in complex with kotalanol to 2.15 Å resolution and compare its active site and kinetic features with those of ntMGAM. This analysis has identified the active site similarities in the Ϫ1 subsite and differences in the ϩ1 subsite that may account for the additional activity of ntSI for ␣-1,6-linkages.
ntSI was expressed in Drosophila S2 cells using the Drosophila Expression System (Invitrogen), following a similar expression protocol as for ntMGAM (10). Stable cell lines were obtained by co-transfecting S2 cells with pMT-TEVA-ntSI and pCopBlast, a blasticidine resistance plasmid. Transfected cells underwent single cell selection, and the best expressing single cell clones were adapted to serum-free medium, scaled up to Fernbach shaker flasks (800 ml/flask), and induced with 4 M CdCl 2 , and the medium was harvested after 72 h. The pMT-TEVA-ntSI construct allowed for secretion of ntSI into the culture medium under the control of a metallothionein promoter with an N-terminal hexahistidine tail engineered with a tobacco etch virus (TEV) cut site.
Protein Purification-The secreted ntSI protein was purified from the medium using nickel-Sepharose resin (GE Healthcare) (9 ml of resin/liter of medium). Nickel affinity-purified ntSI ran as a single band (ϳ100 kDa) on a reducing SDS-polyacrylamide gel but as three distinct bands on a non-reducing gel. Further purification using size affinity chromatography (S200) revealed three peaks, each corresponding to a band on the non-reducing gel. Fractions were tested for the ability to cleave p-nitrophenyl ␣-D-glucopyranoside (pNPG), and the third peak was found to be the only active peak. Fractions in this peak were pooled and concentrated. The total yield of pure ntSI from expression in Drosophila S2 cells was ϳ10 mg/liter of medium.
Crystallization, Data Collection, and Processing-ntSI was crystallized using the hanging drop vapor diffusion method using 1 l of protein (ϳ7.5-22 mg/ml) plus 1 l of reservoir buffer (0.5 M NaCl, 0.1 M Bistris propane, pH 7.0, and 18% polyethylene glycol 4000). These conditions produced clusters of thin crystal plates, which were then used to streak seed equilibrated drops in order to grow single crystal plates. Equilibrated drops were composed of 1.5 l of protein and 3 l of reservoir buffer (0.1 M MgCl 2 , 0.1 M Bistris propane, pH 7.0, 15% polyethylene glycol 4000). For easier crystal handling, crystals were cross-linked with glutaraldehyde (16) before being drawn into a 4:1 ratio of paratone/mineral oil. Crystal complexes were obtained by soaking cross-linked crystals in mother liquor supplemented with 200 M kotalanol for ϳ2 h.
Data for the apo-ntSI and ntSI-kotalanol crystals were collected on an ADSC Quantum-270 CCD detector at beamline F1 at the Cornell High Energy Synchrotron Source and were processed with HKL2000 (17). Data collection statistics are given in Table 1.
Phasing and Refinement-apo-ntSI phases were solved by molecular replacement (Phaser (18)), using a trimmed structure of ntMGAM (Protein Data Bank accession code 2QLY) as a model. Four monomers were found in the asymmetric unit, corresponding to a crystal solvent content of 52%. Loop and side-chain rebuilding was carried out in Coot (19) and was alternated with cycles of non-crystallographic symmetry restrained refinement in Refmac (20). The electron density allowed for the building of a Man 3 GlcNAc 2 chain in two of the monomer active sites.
The apo-ntSI structure was then used as a starting model in the refinement of the ntSI-kotalanol structure (2.15 Å). Rigid body refinement was first carried out in Refmac (20), followed by restrained refinement (without non-crystallographic symmetry restraints). The higher resolution data of ntSI-kotalanol allowed for rebuilding of missing side-chain residues in Coot (19) and the addition of solvent molecules using ARP/wARP (21). The electron density in the active site allowed for the modeling of kotalanol molecules in two of the monomer active sites and of Man 2 GlcNAc 2 glycans and Tris molecules in the remaining two active sites.
B-factor analysis revealed that monomer D had significantly higher overall B-factor (48 Å 2 ) compared with monomers A, B, and C (22, 25, and 27 Å 2 ). TLS refinement (22) was then carried out for 15 cycles using each monomer as a TLS group, followed by 15 cycles of restrained refinement. After TLS refinement, the R/R free dropped from 0.198/0.253 to 0.178/0.223 and resulted in a slight improvement in electron density.
A summary of the refinement statistics is given in Table 1. Inhibitor topologies and restraints were generated using the PRODRG server (23), and protein stereochemistry was validated using the Molprobity server (24). Graphics were generated using PyMOL (available on the World Wide Web).
Enzyme Assay-ntSI and ntMGAM activities against maltose and isomaltose substrates were determined using the glucose oxidase assay (25). Assays were carried out in 96-well microtiter plates containing substrate (2-40 mM maltose or 2.5-400 mM isomaltose), 100 mM MES buffer, pH 6.5, and enzyme (4 nM ntMGAM or 2 nM ntSI in maltose assays and 20 nM ntMGAM or 2 nM ntSI in isomaltose assays). Assay reactions were linear over the entire reaction period; however, ntSI required supplementation of 0.05% Tween 20 to maintain enzyme stability.
Reactions were incubated at 37°C for 30 min and terminated with Tris-HCl (pH 7) to a final concentration of 1 M. Aliquots of glucose oxidase/peroxidase reagent (125 l) were added to each well and left to develop at 37°C for 30 min. Absorbance was measured at 450 nm and compared with a standard glucose curve to determine the amount of glucose released. Each molecule of maltose or isomaltose releases two molecules of glucose upon cleavage. All reactions were performed in triplicate, and absorbance measurements were averaged to give a final result. Inhibition assays of ntMGAM and ntSI with kotalanol and acarbose followed protocols that have been described previously (10,13).
Kinetic parameters of ntMGAM and ntSI were also determined using pNPG as a substrate. Reactions were carried out in microtiter plates and consisted of 1.5-40 mM pNPG, 0.1 M MES, pH 6.5, 20 nM ntMGAM or 10 nM ntSI supplemented with 0.05% Tween 20. Plates were incubated at 37°C for 45 min before being quenched with 0.5 M sodium carbonate, and absorbance was measured at 405 nm.

RESULTS
Substrate Specificity-The substrate specificities of recombinant human ntMGAM and ntSI for maltose, isomaltose, and the general glucosidase substrate pNPG (Fig. 1A) were examined, and kinetic parameters were calculated ( Table 2). For ntSI, specificity constants (k cat /K m ) were calculated to be 19 Ϯ 4, 9 Ϯ 2, and 13 Ϯ 1 s Ϫ1 mM Ϫ1 , for maltose, isomaltose, and pNPG substrates, respectively. The constants are on the same order of magnitude, suggesting that ntSI shows no overt preference for one substrate over another. In contrast, ntMGAM displays a clear preference for maltose substrates because its k cat /K m for maltose (26 Ϯ 8 s Ϫ1 mM Ϫ1 ) is ϳ20 times greater than for pNPG (1.4 Ϯ 0.1 s Ϫ1 mM Ϫ1 ) and ϳ450 times greater than for isomaltose (0.06 Ϯ 0.01 s Ϫ1 mM Ϫ1 ).
Regarding their substrate specificities against natural substrates, both ntMGAM and ntSI have the ability to hydrolyze the ␣-1,4 bond of maltose, but only ntSI is able to efficiently hydrolyze the ␣-1,6 bond of isomaltose. Kinetic results also reveal that pNPG is more efficiently cleaved and tightly bound to ntSI because its K m is ϳ10-fold less (1.3 Ϯ 0.1 mM) than that of ntMGAM (12.1 Ϯ 1.0 mM).
Overall Structure of ntSI-The crystal structures of recombinant human ntSI in apo form and in complex with the inhibitor kotalanol were solved to 3.2 and 2.15 Å, respectively. Phases were solved by molecular replacement using the structure of ntMGAM (ϳ60% sequence identity; Protein Data Bank code 2QLY) as a model. Due to its higher resolution, the ntSI-kotalanol structure will form the basis of the initial discussion. However, several insights from comparison with the apo-ntSI structure will also be addressed below. Crystal packing analysis of ntSI-kotalanol reveals that each asymmetric unit contains four ntSI molecules (monomers A, B, C, and D, each ϳ102 kDa) ( Fig.  2A) with an average root mean square (r.m.s.) deviation of 0.29 Ϯ 0.07 Å between monomer subunits. In all monomers, the N-terminal residues (aa 1-27 (monomer A), 1-29 (mono- mer B), 1-27 (monomer C), and 1-33 (monomer D)), including the His 6 tag and engineered TEV cleavage site are disordered. It is interesting to note that the removal of the disordered N-terminal region by TEV cleavage resulted in poorly diffracting crystals (Ͼ10 Å resolution).
Oligomeric Assembly-In studies with full-length SI (encompassing the transmembrane domain, O-glycosylation stalk, ntSI, and ctSI), homodimers have been observed in electron microscopy and sedimentation studies and are assembled with N-terminal domains interacting and C-terminal domains interacting with one another (28,29). However, the observance of SI dimers seems to be highly dependent on experimental conditions because other cross-linking studies have suggested that SI exists as a monomer (30). In our studies, ntSI was found to elute as an active monomer during gel filtration chromatography (data not shown). However, an interesting interaction between monomers B and C was observed in the crystal-packing lattice (Fig. 2B). These monomers are related by a non-crystallographic 2-fold axis of symmetry and share an interface composed of residues from the trefoil subdomain and catalytic insert 2. This interaction is also observed between monomer A and a symmetry-related copy of monomer D. Analysis using the Protein Interfaces, Surfaces, and Assemblies (PISA) server (31) reveals that this interface has a buried surface area of ϳ800 Å 2 /interface and contains extensive hydrophobic interactions and 10 -12 hydrogen bonds. Although the dimer interface area is small for a protein of this size (32), it might represent a low affinity physiological interaction, as will be discussed below (see "Discussion").
ntSI Active Site-Prior to freezing and data collection, ntSI crystals were soaked in mother liquor supplemented with a 200 M concentration of the inhibitor kotalanol. F o Ϫ F c maps clearly reveal the presence of the kotalanol inhibitor bound to active sites of monomers B and D (Fig. 2C). In comparison, a Tris molecule (from the buffer used during purification) and a Man 2 GlcNAc 2 glycan chain (originating from glycosylation site Asn 822 of a crystal symmetry-related molecule) simultaneously occupy the active sites of monomers A and C (Fig. 2, C and D). It was noted that active sites of monomers A, B, C, and D are structurally identical to one another and are not affected by the kotalanol, Tris, or Man 2 GlcNAc 2 binding. In addition, the Tris and glycan chains and the kotalanol molecules bind identically in the A and C monomers and the B and D monomers, respectively (data not shown).
The Man 2 GlcNAc 2 glycan interacts with the ntSI active site primarily through the terminal M4 ring and a to a lesser extent with the NAG2 ring. These interactions include hydrogen bonding between the side chain hydroxyls of Asp 231 with the C3-OH and C4-OH groups of M4 and between the side chain of Asp 571 and the C4-OH M4 group via a water molecule (supplemental Fig. 3). Additional stabilization of the Man 2 GlcNAc 2 may result from the hydrophobic interactions with Leu 233 , Trp 327 , Trp 435 , Phe 479 , Val 605 , and Tyr 634 , which line the ntSI active site.
In a 3.2 Å resolution structure of apo-ntSI, a Man 3 GlcNAc 2 chain attached to a symmetry-related Asn 822 residue is observed in the active sites of monomers A and C, whereas monomers B and D remain empty. The electron density of the apo-ntSI structure was clear enough to position the five Man 3 GlcNAc 2 rings within the density (Fig. 2C) but not detailed enough to assign hydrogen bonding interactions. The exact location of glycan binding, however, is not fixed and is dependent on the arrangement of monomers in the asymmetric unit. The structure of apo-ntSI has slightly different unit cell dimensions (76.34 ϫ 172.59 ϫ 343.87 Å) compared with the ntSI-kotalanol structure (69.4 ϫ 165.8 ϫ 341.4 Å) ( Table 1), which results in a slight rearrangement of ntSI monomers in the asymmetric unit. Comparing the Man 2 GlcNAc 2 binding in the 2.15 Å structure with the Man 3 GlcNAc 2 binding in the 3.2 Å structure, the terminal mannose residues (M4) overlap, but the rest of the chain diverges from there, resulting in an ϳ9-Å separation between the N-linked NAG1 residues (Fig. 2D). Further detail about the Man 2 GlcNAc 2 glycan, including its location in relations to the substrate binding subsites, will be discussed below in context with active site comparison with ntMGAM.
Comparison of Kotalanol Binding in ntSI and ntMGAM-To characterize the ntSI active site architecture relative to that of ntMGAM and to determine the extent of conservation in the Ϫ1 and ϩ1 subsites, the active site of ntSI with bound kotalanol was superimposed with the recently solved structure of ntMGAM-kotalanol (Protein Data Bank code 3L4V). Although monomers B and D both contain kotalanol in their active sites, monomer B is of higher structural quality (lower B factors) and was used as the basis for structural comparison with ntMGAM.
Structural comparison of the ntMGAM and ntSI complexed structures reveals identical binding of the kotalanol ring group in the Ϫ1 subsite and the conservation of most of the Ϫ1 subsite residues (Fig. 3A). These residues include catalytic nucleophiles Asp 443 and Asp 472 and acid/base catalysts Asp 542 and Asp 571 in ntMGAM and ntSI, respectively. Substrate binding residues His 600 , Asp 327 , Arg 526 , Asp 203 , Trp 406 , and Phe 450 in ntMGAM are also structurally conserved with ntSI residues His 629 , Asp 355 , Arg 555 , Asp 231 , Trp 435 , and Phe 479 . The most obvious structural change in the Ϫ1 site is attributed to the substitution of ntMGAM Tyr 299 with ntSI Trp 327 , which results in the extension of the Trp side chain toward the ϩ1 subsite (Fig. 3A).
In the ϩ1 subsite, there are notable structural differences between ntMGAM and ntSI. These include the substitution of ntMGAM Thr 205 with ntSI Leu 233 , which positions the Leu side chain closer to the ϩ1 subsite and the substitution of ntMGAM residue Ala 570 by ntSI residue Val 605 (Fig. 3A). There is also a 5.3-Å shift of the ntSI Lys 509 side chain toward the ϩ1 subsite compared with the side chain of ntMGAM Lys 480 . These structural differences seem to affect the binding of the polyhydroxylated chain of kotalanol in the ϩ1 subsite. The sulfate group in the ntSI-bound kotalanol is displaced by ϳ1-1.5 Å from its ntMGAM-bound position to accommodate the bulky Trp 327 ntSI group. Additionally, the hydrogen bonding interaction between Thr 205 and the C5Ј-OH group (via a water molecule) in the ntMGAM-bound kotalanol structure is lost in the ntSIkotalanol structure. Instead, hydrogen bonding interactions appear between Lys 509 and C6Ј-OH and between Gln 232 and C5Ј-OH (via a water molecule) (Fig. 3A).
Comparison of Man 2 GlcNAc 2 and Acarbose Binding in ntSI and ntMGAM-As was described above, Man 2 GlcNAc 2 and Man 3 GlcNAc 2 Asn 822 -linked glycans were found in the monomer A and C active sites in the 2.15 and 3.2 Å ntSI structures, respectively. Because the terminal M4 residues of both glycans overlap (Fig. 2D) and were found to interact with acid/base catalyst Asp 571 and ϩ1 subsite residue Asp 231 (supplemental Fig. 3), we wanted to determine whether the M4 binding could represent a biologically relevant interaction occurring between substrate and subsite. Although there is no precedence for mannose binding in ntSI, it is a C2 epimer of glucose and thus very similar in structure. Because an ntSI crystal structure complexed with a substrate analogue or inhibitor, such as acarbose, has not been determined, we compared the ntSI structure with that of ntMGAM-acarbose (Protein Data Bank code 2QMJ) in order to map the location of the Man 2 GlcNAc 2 binding in relation to the proposed Ϫ1 and ϩ1 subsites.
Comparison of the structures reveals that acarbose and Man 2 GlcNAc 2 groups bind differently from one another with essentially no overlap between rings (Fig.  3B). The ntMGAM Ϫ1 subsite is occupied by the non-reducing ring A of acarbose, whereas in ntSI, it is occupied by a Tris molecule. The binding of the terminal M4 group in Man 2 GlcNAc 2 reveals that the ring is positioned ϳ4 Å away from the ϩ1 subsite delineated by the B ring of acarbose and that its C6-OH arm is oriented toward the Ϫ1 subsite (Fig. 3B). We considered the possibility that the binding site occupied by M4 marks the ϩ1 subsite for the reducing ring of isomaltose substrates. However, the distance between its C6-OH arm and the C1 group in acarbose ring A is too large (ϳ7 Å) to accept an ␣-1,6 bond without a significant movement in M4.

DISCUSSION
In this study, we report the crystal structure of human ntSI. Its overall architectural fold is very similar to that of the paralogous enzyme ntMGAM (r.m.s. deviation ϭ 0.5 Å). From crystal packing analysis, the largest interface between two ntSI monomers suggests a possible mode of dimerization mediated through the ntSI trefoil and catalytic insert 2 subdomains. This arrangement is consistent with the assembly of full-length SI, with O-glycosylation stalks stemming from ntSI toward the membrane and the continuation of the ctSI domains on the distal side away from the membrane attachment (Fig.  2B), similar to the assembly observed in electron microscopy studies (28). However, it is important to realize that this interaction occurs under the high protein concentrations required Kotalanol and ntMGAM/ntSI residues are represented as thick sticks and thin sticks, respectively. Selected bound waters in the ntMGAM and ntSI structures are shown in yellow and green, respectively. ntMGAM residues are labeled with an asterisk. B, superposition of Man 2 GlcNAc 2 (yellow) and Tris (orange) from ntSI-kotalanol with acarbose (green) from ntMGAM-acarbose in the ntSI active site (blue surface representation). A doubleheaded arrow indicates the distance (ϳ7 Å) between the C6-OH of the M4 ring and C1 of acarbose ring A. C, surface representation of the ntSI (top) and ntMGAM (bottom) active sites with non-structurally conserved residues displayed as blue and orange sticks, respectively. One possible conformation of isomaltose (pink, surface-rendered) is modeled into the Ϫ1 and ϩ1 subsites.
for crystallization and that ntSI is observed only in the monomeric state during our purification studies. Nevertheless, because full-length SI dimers have been observed in other sedimentation and electron microscopy studies, it is possible that interactions between ntSI monomers are weak, at least under our experimental conditions and in the absence of membrane, O-glycosylation stalk, and C-terminal SI domain. The physiological relevance of a putative dimerization is still not clear, and further biochemical tests are needed to investigate the presence of this interface. However, many brush-border enzymes exist as dimers, and dimerization is believed to be relevant for intracellular trafficking (29). In any case, because we have observed active monomers in our kinetic studies, dimerization is not crucial for catalytic activity. Moreover, because the active site pockets on the ntSI monomers are exposed and are not composed of elements from adjacent monomers (Fig. 2B), we predict that they act independently of one another.
From kinetic studies, it was observed that maltose could be hydrolyzed by both ntMGAM and ntSI but that isomaltose could only be hydrolyzed efficiently by ntSI ( Table 2). The differential substrate specificities of ntMGAM and ntSI seem consistent with their roles in terminal starch digestion; the redundancy in ␣-1,4 activity seen in both enzymes (also in ctMGAM and ctSI) is beneficial due the higher distribution of ␣-1,4 linkages (95%) in starch molecules. In contrast, ␣-1,6 linkages make up a small fraction of the linkages (5%) in starch molecules. Thus, it is logical that only one subunit (ntSI) is responsible for cleavage of this particular bond. To investigate the structural basis for differential substrate specificity observed between ntMGAM and ntSI, their active sites were compared.
Structural superposition of the ntMGAM and ntSI kotalanol bound structures revealed strong structural conservation of Ϫ1 subsite residues (Fig. 3A). This is quite expected because the Ϫ1 subsite binds the non-reducing ring of substrates, and this ring is identical in maltose and isomaltose (Fig. 1A). Substrate discrimination is most likely mediated by the ϩ1 subsite, and there are accordingly structural differences observed between ntSI and ntMGAM in this region. A significant difference between ntMGAM and ntSI subsite architecture is that the ntMGAM ϩ1 subsite is wide and open compared with the narrow groove observed in the ntSI ϩ1 subsite (Fig. 3C). This narrowing results from the presence at analogous positions of smaller residues in ntMGAM (Tyr 299 , Ala 576 , Thr 205 , and Thr 204 ) as compared with larger residues in ntSI (Trp 327 , Val 605 , Leu 233 , and Gln 232 ) and the displacement of side chain Lys 509 closer to the ϩ1 subsite of ntSI. Sequence comparison of the four MGAM and SI domains reveals that Trp 327 in ntSI may be important in conferring ␣-1,6 specificity because it is conserved as Tyr in the other ntMGAM, ctMGAM, and ctSI domains (6). However, it seems counterintuitive that the narrow ϩ1 subsite of ntSI would accommodate both its ␣-1,4 and ␣-1,6 substrates, whereas a wider ϩ1 subsite in ntMGAM would account for specificity for the ␣-1,4 substrate.
Unfortunately, there is currently no crystal structure of ntSI or ntMGAM in complex with an ␣-1,6-linked substrate or inhibitor analogue that can be used to predict isomaltose binding in ntSI. Instead, isomaltose was modeled by hand into the Ϫ1 subsites of ntMGAM and ntSI by aligning the non-reducing glucose ring of isomaltose to that of acarbose (using the ntMGAM-acarbose structure superposed on ntSI as a reference) (Fig. 3C). The non-reducing ring of isomaltose was then fixed in place while rotating torsion angles of the ␣-1,6 linkage to avoid steric clashes between the reducing ring and the ϩ1 subsite residues. From this modeling exercise, it was evident that the number of conformations that the reducing isomaltose ring can adopt in the ntSI structure is limited due to narrowness of the ϩ1 subsite. In comparison, the ntMGAM ϩ1 architecture is wide and open, and supports the modeling of multiple conformations at the reducing isomaltose ring. Moreover, conformational studies have indicated that isomaltose is a more flexible disaccharide compared with maltose (33,34). This is due to the three-bond ␣-1,6 linkage in isomaltose, in comparison with the two-bond ␣-1,4 linkage in maltose, which results in less interaction between reducing and non-reducing rings. We speculate that the narrow ntSI ϩ1 subsite may be necessary for spatially constraining the flexible ␣-1,6 bond in place during substrate binding. Moreover, the presence of large hydrophobic residues Trp 327 , Leu 233 , and Val 605 , in the ϩ1 subsite, which are absent in ntMGAM, may help greatly in binding by interacting with hydrophobic patches on the reducing ring of isomaltose.
To gauge the hydrophobic environment of the ϩ1 subsite, ntSI and ntMGAM were assayed with pNPG (Fig. 1A), a general glucosidase substrate that lacks hydrogen-bonding capability in its leaving group. With the assumption that the glucose moiety of pNPG binds identically in both ntMGAM and ntSI Ϫ1 subsites, any interactions with the ϩ1 leaving group should be mostly hydrophobic. From these studies, pNPG was found to bind ϳ10 times tighter to ntSI compared with ntMGAM (Table  2), supporting the idea of a more hydrophobic ϩ1 ntSI subsite. We propose that because the ϩ1 subsite is spatially constricted and probably limits the variety of hydrogen bonding possibilities, nonspecific hydrophobic interactions contribute greatly toward binding of the reducing ring of substrates and broader substrate specificity of ntSI.
Moreover, although the binding of the Man 2 GlcNAc 2 glycan to the ntSI active site (Figs. 2D and 3B) is not biologically relevant, the binding of the M4 mannose moiety close to the ϩ1 subsite is a further indication of the nonspecificity of the ntSI substrate-binding site. However, for further validation of the hydrophobic environment of the ϩ1 ntSI subsite and the entropic changes upon isomaltose binding, additional calorimetric studies as described in Ref. 35 will be pursued in future studies.
Inhibition studies with ␣-glucosidase inhibitors revealed that in comparison with acarbose, kotalanol is a ϳ20 times and ϳ300 times better inhibitor of ntSI and ntMGAM, respectively (Table 3). Structural comparison of kotalanol in the ntSI and ntMGAM active sites revealed similarity in binding, especially in the Ϫ1 subsite. Although there are differences in the binding of the polyhydroxylated chain of kotalanol in the ϩ1 subsite, this does not seem to affect significantly its inhibition properties toward ntSI (K i ϭ 0.60 Ϯ 0.06 M) and ntMGAM (K i ϭ 0.19 Ϯ 0.03 M). This is most likely due to the structural conservation of the C2Ј and C4Ј chain hydroxyls (Fig. 3A). From previous inhibitor and structural studies, it was determined that stereochemistry of the C2Ј and C4Ј substituents is crucial for the activity of kotalanol because important hydrogen bonds are formed with the active site residues (11,13).
The structural comparison between ntMGAM and ntSI provides preliminary insights into the structural basis for their complementary activities toward ␣-1,4and ␣-1,6-linked substrates. The crystal structures will also be helpful in the design of specific inhibitors that exploit differences between MGAM and SI domains. Kotalanol was shown to be a specific inhibitor of both ntMGAM and ntSI, with no selectivity for either. However, removal of the essential hydrogen bonding interactions required for its specificity (through the removal of key kotalanol hydroxyls C2Ј-OH or C4Ј-OH) and the addition of a hydrophobic extension that could be stabilized by the narrow hydrophobic ϩ1 ntSI subsite might result in a modified kotalanol inhibitor that is more specific for ntSI. The design of domain-specific inhibitors will greatly aid in the characterization of these domains and is of interest as a means of addressing ntSI and ntMGAM activities in vivo, in order to better define the complementary roles of MGAM and SI in the process of terminal starch digestion.