Mutational Analysis of Target Enzyme Recognition of the β-Trefoil Fold Barley α-Amylase/Subtilisin Inhibitor*

The barley α-amylase/subtilisin inhibitor (BASI) inhibits α-amylase 2 (AMY2) with subnanomolar affinity. The contribution of selected side chains of BASI to this high affinity is discerned in this study, and binding to other targets is investigated. Seven BASI residues along the AMY2-BASI interface and four residues in the putative protease-binding loop on the opposite side of the inhibitor were mutated. A total of 15 variants were compared with the wild type by monitoring the α-amylase and protease inhibitory activities using Blue Starch and azoalbumin, respectively, and the kinetics of binding to target enzymes by surface plasmon resonance. Generally, the mutations had little effect on kon, whereas the koff values were increased up to 67-fold. The effects on the inhibitory activity, however, were far more pronounced, and the Ki values of some mutants on the AMY2-binding side increased 2–3 orders of magnitude, whereas mutations on the other side of the inhibitor had virtually no effect. The mutants K140L, D150N, and E168T lost inhibitory activity, revealing the pivotal role of charge interactions for BASI activity on AMY2. A fully hydrated Ca2+ at the AMY2-BASI interface mediates contacts to the catalytic residues of AMY2. Mutations involving residues contacting the solvent ligands of this Ca2+ had weaker affinity for AMY2 and reduced sensitivity to the Ca2+ modulation of the affinity. These results suggest that the Ca2+ and its solvation sphere are integral components of the AMY2-BASI complex, thus illuminating a novel mode of inhibition and a novel role for calcium in relation to glycoside hydrolases.

The double-headed barley ␣-amylase/subtilisin inhibitor (BASI) 1 of the Kunitz soybean trypsin inhibitor family acts on proteases of the subtilisin family and the endogenous high pI ␣-amylase (AMY2) but has no effect on the minor isozyme AMY1 that shares 80% sequence identity with AMY2 (1)(2)(3)(4)(5)(6)(7)(8). Under favorable conditions, BASI inhibits the AMY2-catalyzed hydrolysis of starch with a K i ϳ0.1 nM (2,4). This K i value is in excellent agreement with the K D of 0.07 nM estimated by equilibrium fluorescence titration and stopped-flow kinetics according to a fast 1:1 two-step tight binding reaction (9). Surface plasmon resonance (SPR) analysis gave weaker affinities (K D ϳ1 nM) presumably due to mass transfer limitations characteristic of the two-phase system and chip surface heterogeneity (4).
In vivo, plausible functions of BASI might be to inhibit AMY2 emerging during premature seed sprouting (1,10) or the inhibition of proteases from pathogens, e.g. fungi belonging to the genus Fusarium (11). The in vitro demonstration of a ternary complex of AMY2, BASI, and subtilisin (1) is consistent with this latter function. BASI and wheat ␣-amylase/subtilisin inhibitor (WASI) are highly homologous (92% sequence identity), whereas rice ␣-amylase/subtilisin inhibitor (RASI), acting on insect ␣-amylase, shares only 58% sequence identity with BASI (12,13). BASI is assigned to the soybean trypsin inhibitor-like superfamily of ␤-trefoil fold proteins implicated in various protein-protein interaction roles and shares 20 -30% sequence identity with Kunitz-type trypsin inhibitors (6,14). Six different loops and three ␤-strands on one side of BASI present residues that interact with both the A and B domains of AMY2 via several hydrogen bonds and salt bridges, resulting in a relatively large protein interface of 2355 Å 2 (3) (Fig. 1). Distinct differences in the corresponding AMY1 regions provide a structural rationale for the strict specificity of BASI for AMY2. Most interestingly, a novel feature of the AMY2-BASI complex is the presence of a fully hydrated Ca 2ϩ (Ca 503 ) embedded at the complex interface (3). This ion seems to mediate binding between inhibitor residues and the catalytic groups in the enzyme via an extended hydrogen bonding network ( Figs. 1 and 2B). In addition to Ca 503 , AMY2 contains three calcium ions bound to domain B (15,16). In the ␣-amylase family, or glycoside hydrolase family 13 (17), 2 the active site is formed between domain B (18) and the ␤ 3 ␣ loops from domain A, rendering this part of the protein crucial for substrate binding (15,19).
The solved structures of complexes between ␣-amylases and their inhibitors show great structural diversity suggesting important variations in the mode of action of the different inhibitors (for review see Refs. 6 and 20). The majority of these inhibitors, however, elicit inhibition by directly binding to key catalytic groups (21)(22)(23) in the target enzyme or by docking on key aromatic groups in proximal subsites (24). In this context, the AMY2-BASI system presents a different example of how inhibition can be attained even though no direct bonds between any of the catalytic groups of the enzyme and the inhibitor are present. Although some binding kinetics and mutagenesis were reported for lectin-like (25)(26)(27), cereal-type (28,29), and tendamistat inhibitors (30), thorough analysis is lacking for most systems. An exception is AMY2-BASI, which has been the subject of extensive studies with respect to binding mechanisms and structure/function relationships (2,5,6,9). Recently data from mutants of AMY2 residues binding BASI (7), SPR, and isothermal titration calorimetry of AMY2-BASI wild-type protein interactions have demonstrated the effects of ionic strength, pH, and Ca 2ϩ on kinetics and thermodynamics of the complex formation (5,8), but many important questions are still unanswered. The newly established heterologous expression system in Escherichia coli (4) prompted dissection of the contribution of individual regions and side chains to the overall inhibitory activity of BASI. Another key question addressed is whether the embedded Ca 503 and its solvent coordination sphere, as well as other buried solvent molecules, play a significant role in the complex formation and dissociation. This first elaborate mutational analysis of a bifunctional Kunitztype inhibitor offers insight into the determinants of affinity and the driving forces for complex formation with target enzymes, which paves the way for rational design of other ␤-trefoil fold family members.
Protein Characterization-SDS-PAGE (4 -12%) and isoelectric focusing, pH 3-10 (NOVEX and Invitrogen), were performed as recommended by the manufacturers. Protein concentrations were calculated from amino acid contents of hydrolysates (0.2-1.0 nmol; 6 M HCl, 110°C, 22 h) determined on a Biochrom 20 analyzer (Amersham Biosciences). Circular dichroism was measured with a Jasco J-810 spectropolarimeter dichrograph at 25°C. Ten spectra were scanned at 250 to 190 nm by using a quartz cell (path length 1.0 mm) and scaled in the protein concentration range 1.5-5 M.
Calculation of K i -Determination of the intrinsic K i ϭ K i,app ϫ (1 ϩ S/K m ) Ϫ1 (36) assumed equilibrium between AMY2, BASI, substrate (Blue Starch), and the AMY2-BASI complex. Inhibition curves showed % inhibition, 100 ϫ (1 Ϫ(Act i /Act 0 )) versus BASI:〈⌴Y2 (molar ratio); Act 0 and Act i are activities in the absence and presence of BASI. K i,app was obtained from the slope of 1/Act i versus [BASI 0 ]/(1 Ϫ Act i ), using the intercept on the y axis indicated the enzyme concentration (according to Fig. 5 in Ref. 36). The K i value for inhibition of savinase was determined by using the above equations with the appropriate K m values and substrate (azoalbumin).
Surface Plasmon Resonance-Binding kinetics of BASI and either AMY2 or savinase was determined by using a BIAcore 3000 (Biosensor, Amersham Biosciences) (4,8). The sensor chips were charged by 300 -1000 response units of either biotinylated AMY2 (streptavidin-sensor chips), BASI, or savinase bound by using the amine coupling procedure (CM5 sensor chips) (8). Sensorgrams (response units versus time) in duplicate were recorded at a flow rate of 30 l ϫ min Ϫ1 at 25°C, using five analyte concentrations (20 -300 nM) in 10 mM Hepes, pH 8.0, 5 mM CaCl 2 , and 0.005% surfactant P20. The effect of Ca 2ϩ was measured in 10 mM Mes, pH 6.5, 0.005% surfactant P20. The association and dissociation were monitored for 4 and 15 min, respectively, and the chip was regenerated by 10 mM sodium acetate, pH 5.0 (AMY2), or by 5 min of prolonged dissociation (savinase). Sensorgrams were analyzed using BIAevaluation version 3.1 software applying a single site 1:1 (Langmuir) binding model: A ϩ B^AB) and deriving k off , k on , and K D values (8). Binding energy differences were calculated using ⌬⌬G ϭ ϪRT ln(K D,mut /K D,wt ) (37).

Choice and Production of BASI Mutants-
The structure of AMY2-BASI (3), sequence alignment of BASI, WASI, and RASI ( Fig. 3), and the structure of proteinase K/WASI (40) guided mutations in four different regions of BASI (Figs. 1 and 2). Three of the four regions were located at the AMY2-BASI interface, and the fourth was on the opposite side of the inhibitor, which is suggested to be involved in protease inhibition. In the case of the AMY2-BASI interface, mutations were focused on two regions of BASI, the first of which is in contact with AMY2 domain B, and the other overlooks the active site of the enzyme (Fig. 2, A and B). In addition, one charge-reversed mutation was introduced in a third region of BASI in contact with AMY2 domain A but fairly distant from the active site (Figs. 1 and 2C). In the first region of BASI, Ser 77 , Tyr 131 , Lys 140 , and Asp 150 , which form various hydrogen bonds to the two previously mutated AMY2 residues Arg 128 and Asp 142 (Table I; Figs. 2, A and B and 3) and also to Gly 144 , were replaced mostly by sterically similar side chains (41). Glu 168 and Tyr 170 were the targets of mutations in the second binding region located at the center of the binding interface. These residues are involved in an extended hydrogen bonding network comprising the hydration shell of the embedded Ca 503 and the three AMY2 catalytic groups Asp 179 , Glu 204 , and Asp 289 (Table I; Fig.  2B). The only AMY2-BASI direct bond involving these residues is between Glu 168 in BASI and Lys 182 in AMY2 that in turn forms a hydrogen bond to a substrate glucosyl residue at subsite ϩ2 (15,19). Moreover, Glu 168 and Tyr 170 correspond to glutamine and proline in the rice homologue RASI (Fig. 3) that inhibits insect ␣-amylase (13), and the RASI mimics E168Q, Y170P, and E168Q/Y170P were designed to assess if they could confer inhibitory activity on insect ␣-amylase. In the third region Asp 156 (Fig. 2C), located at the periphery of the complex and having only a water-mediated bond to AMY2, was chargereversed to probe how changes in charge density in that region would affect the binding kinetics of BASI to AMY2. Another interesting feature of this region is the presence of a network of buried solvent molecules mediating several AMY2-BASI interactions. Finally, although coordinates for the complex of the wheat homologue WASI and proteinase K were not deposited, a molecular graphics representation (40) showed that the segment Ala 85 -Thr 88 (BASI Ala 86 -Thr 89 ) of the loop connecting ␤-strands 5 and 6 may participate in protease binding. This was tested by the Y87A and T89A as well as two adjacent BASI mutants S93A and E95Q.
Initially the mutants S77A, K140L/K140N, E168Q/E168T, and Y170F/Y170P were made using the autocleavable intein-CBD tag, but poor yields motivated shifting to the His 6 tag system. The mutations did not adversely alter the expression levels indicating that proper global folding of the recombinant proteins and SDS-PAGE confirmed the molecular size and the purity (not shown). As expected, neutral mutations caused no change in pI (Fig. 4, lanes 3, 8, 9, 12, and 14 -16 compared with lanes 2 and 11), whereas charge elimination mutations decreased or increased the pI depending on the change in net charge (Fig. 4, lanes 4 -7, 13, and 17). K140N had lower pI than K140L, probably due to an intramolecular hydrogen bond (3.3 Å) formed between Asn 140 N␦2 and Asp 150 O␦1, which stabilizes the negative charge of Asp 150 . The differences in activity of the various recombinant forms of wild-type BASI were negligible (Table III). The inhibitory activity of factor Xa-cleaved mutants was indistinguishable from those with a His 6 tag (4); therefore, the cumbersome and inefficient factor Xa cleavage step was omitted. The CD spectra of selected mutants exhibiting varied affinities and located on different parts of BASI (S77A, Y87A, S93A, Y131F, D150N, E168T, and E168Q) were identical to that of wild-type BASI (not shown), indicating correct folding. The unaffected K D value for savinase of BASI mutants confirmed maintenance of global structural integrity (Table II)  with the exception of E168T, for which decreased k on caused slightly increased K D values.
AMY2 Inhibition by BASI Mutants-The effect of mutations in BASI on the inhibitory activity of on AMY2 was assayed on Blue Starch under optimal and suboptimal conditions with respect to pH and ionic strength. The K i values as a result of the mutations in the three AMY2 binding regions increased from a few to several hundred times or were out of range (at 400-fold BASI molar excess) ( Fig. 5; Table III). The four mutants, Y87A, T89A, S93A, and E95Q, located on the opposite site of the AMY2-BASI interface and wild-type BASI showed identical K i values. Thus, these mutants served as positive controls, and the effects on inhibition observed for the rest of the mutants could be considered as the result of local changes and not due to unspecific long range structural perturbation.
The mutant S77A showed only 3-fold reduced affinity. This mutation, however, was accompanied by remarkably high sensitivity to electrostatic screening of the inhibition of AMY2. The K i value of S77A BASI thus increased 160-fold at pH 6.8 and 0.2 M NaCl relative to the optimal binding at pH 8.0, whereas the K i value of wild-type increased 11-fold (Table III). Apart from S77A, the rest of the mutants in this region were characterized by a substantial loss of inhibition especially in the case  of charge-altered mutants. This is evident by the total apparent loss of inhibition observed for K140L and D150N (Table III). In the second binding region, all the mutants showed a large decease in inhibitory activity varying from 10-fold for Y170F to a total loss of inhibition in the case of E168T. The double mutant E168Q/Y170P resulted in an additive loss of affinity as compared with the single mutants E168Q and Y170P as judged by the ⌬⌬G values (Table IV). Finally, the charge-reversed mutant D156K led to a 30-fold increase in K i values.
SPR Analysis of BASI Mutants Binding with AMY2-The K D values increased 2.5-200-fold for the BASI mutants. The real time kinetics (Table IV) showed that changes in affinity were predominantly the result of increased k off , which is manifested by a relatively more rapid decline in the response signal during the dissociation phase (Fig. 6). The k on values showed only marginal variation, especially in the case of mutations involving charge invariant and neutral side chains. The charge-reversed mutant D156K, however, showed a 6.5-fold decrease in k on and 4.3-fold increase in k off , which contributed almost equally to the resulting loss in affinity (Table IV). The Y131F, K140N, E168Q, and Y170F/Y170P mutations affected K D values somewhat less than K i values (Tables III and IV), and generally charge-altering replacements had the largest effect (Table IV). The mutants Y87A, T89A, S93A, and E95Q BASI at the putative protease-binding site were bound to AMY2 with essentially the same kinetics as wild type (Table IV), confirming the global conformational integrity of the mutants.
Calcium Effect on AMY2-BASI-The effect of the mutations overlooking the active site of AMY2 on the calcium dependence of inhibition was explored by SPR and by Blue Starch inhibition assay. Identical SPR conditions were chosen to reproduce the calcium dependence pattern reported earlier for wild-type BASI (8). Regarding the inhibition of catalysis, optimal conditions were chosen to perform the assay. Increasing [Ca 2ϩ ] from the low micromolar range (when no CaCl 2 was added) to 20 mM resulted in a 29-and 26-fold decrease in K i and K D values, respectively, for wild-type BASI (Table V). In contrast, the calcium dependence of binding was significantly reduced for the Y170F mutant, as the same addition resulted in 13-and 15-fold decreases of K i and K D values, respectively. Finally, E168Q BASI was rendered virtually insensitive to calcium as judged by the modest change in both its K D and K i values in the tested [Ca 2ϩ ] range (Table V).
Inhibition of TMA-The RASI mimics E168Q, Y170P, and E168Q/Y170P BASI (Fig. 3) failed to inhibit TMA at a 2000-fold molar excess (data not shown), suggesting that additional elements besides these two residues are required for the molecular recognition of TMA by RASI. However, a marked decrease in affinity and inhibition potency of these mutants was observed on AMY2 (Tables III and IV). The affinity of these three RASI mimics to savinase, however, was unaffected (Table II), thus ruling out that the lack of TMA inhibition stemmed from misfolding of the mutants.
Mutants in the Putative Savinase-binding Site-Y87A BASI increased K i and K D 3.5-and 11-fold, respectively, whereas T89A, S93A, and E95Q, located in the same region of BASI (Fig. 1), showed little change in affinity as judged from both the K i and K D values (Table II). The residue Tyr 87 thus may be involved, albeit not critically, in inhibition of savinase. These four mutants had similar affinity for AMY2 as wild-type BASI (Tables III and IV), discounting the possibility of major structural changes as a result of the mutations. DISCUSSION Proteinaceous inhibitors of ␣-amylases play a role in human and animal nutrition (42,43), plant defense against pests and pathogens (6,44), and in control of endogenous ␣-amylase activity (1,10,45). BASI, the sole example of an ␣-amylase inhibitor being active on an endogenous enzyme (AMY2), is a relevant target for counteraction of pre-harvest sprouting and for enhancing cereal defense systems against pathogens. Currently, BASI is one of the best characterized ␣-amylase inhibitors, as a complex structure, and kinetic and thermodynamic data on the interaction of BASI to AMY2 have been reported. Yet it remains unclear how individual regions at the AMY2-BASI interface contribute to the observed high affinity binding. This mutational study was undertaken to identify pivotal sites along this large interface. The effects of the different mutations were evaluated by SPR and inhibition of enzyme activity assays. Each of these techniques presents a number of advantages and drawbacks, but together, they offer a powerful tool for the evaluation of the designed mutations. The enzyme activity inhibition assay is based on steady state binding kinetics in solution, whereas SPR monitors real time association and dissociation of an analyte in solution (e.g. BASI) to its ligand    immobilized on the sensor chip (e.g. AMY2 or savinase) (46,47). The conditions of the two assays were chosen based on earlier studies for comparison with the validated published data (4,8). Indeed, the obtained K D and K i values, which differ roughly by a factor of 10, are in very good agreement with those reported earlier (4). Discrepancy between SPR and data derived in solution is not uncommon, and despite the increased usage of SPR for analyzing interactions between biomolecules, there is an ongoing debate regarding the consistency and validity of such data (48 -52). Mass transport limitations, rebinding to the chip surface, more complex modes of binding and multivalency, surface inhomogeneities, and other factors have been reported to set off the absolute values of the binding constants orders of magnitude. Despite these concerns, SPR analysis helps with discerning the kinetics of interaction and requires low amounts of protein. In general, the same conclusions could be drawn based on the relative increase of K D values obtained by SPR or by K i values obtained from the AMY2 inhibition assay. However, the relative affinity changes caused by most mutations reported by SPR were smaller. Thus, binding parameters could be obtained for K140L, E168T, and D150N, whereas inhibitory activities of these mutants were out of range of the inhibition assay. BASI Contacts to AMY2 Domain B-In the BASI region in contact with domain B of AMY2, the impact of S77A, Y131F, K140L, K140N, and D150N varied dramatically. The affinity of S77A decreased marginally, whereas mutations of the charged residues Lys 140 and Asp 150 severely impaired inhibition of AMY2. Ser 77 hydrogen-bonds with Glu 129 ⑀O2 BASI and Arg 128 N⑀ AMY2 (Figs. 1 and 2A). The side chains of the two latter residues are 5.3 Å apart, making a weak salt bridge consistent with the marginal loss of affinity and the increased sensitivity to electrostatic screening for S77A (Table III). The isosteric Y131F mutation is very unlikely to cause large changes in that region. Therefore, the intermediate decrease in affinity observed can be best rationalized by the loss of a solvent-mediated network of bonds involving the side chain OH group of Tyr 131 BASI , Lys 182 N AMY2 , and Tyr 170 N BASI (Table I; Fig. 2C). Abolishing the charged side chain of either Lys 140 or Asp 150 resulted in a drastic drop in affinity underscoring the critical role of these residues in complex formation. Lys 140 N and Asp 150 O␦1 in BASI are at a distance of 2.7 Å, making a strong ionic bond and bridging the two ␤-strands carrying these resi-dues. In addition, both residues are a part of a consortium of bonds involving Lys 140 N BASI and Asp 142 O␦2 AMY2 2.86 Å apart (resulting in an ionic network between Asp 150 BASI , Lys 140 BASI , and Asp 142 AMY2 ), as well as Gly 144 AMY2 , Tyr 170 BASI , and Wat 564 ( Fig. 2A, Table I). It is likely that eliminating the charge of either Lys 140 or Asp 150 would result in increased local flexibility and perturbation of the interactions with domain B. The enthalpic loss caused by disruption of ionic bonds in addition to the high entropic penalty because of increased flexibility is consistent with the severe loss of inhibitory activity of these mutants (Table III). It has been reported earlier that the BASI-AMY2 interaction is mainly enthalpically driven and that temperature had little effect (Ͻ2.9 kcal/mol) on the enthalpy of binding in the interval 25°C-37°C (8). Assuming little or no change in ⌬H, increased conformational flexibility of BASI, equivalent to an increased entropic penalty of binding to AMY2 (⌬⌬S Ͻ 0), would result in a large loss of binding at 37°C as compared with 25°C. This provides a possible explanation for the apparent difference between the SPR data showing less relative decrease in inhibitory activity for these mutants as compared with the inhibition kinetics data (Tables III and IV). Other factors, however, inherent with the SPR analysis such as rebinding on the chip surface cannot be ruled out. Similar arguments apply for the mutant D150N.
BASI Contacts to Domain A-Asp 156 is situated in a positively charged environment about 5.5 Å from Lys 158 and making an intramolecular salt bridge with Arg 106 presented by an adjacent loop. In addition, this residue makes a water-mediated hydrogen bond to AMY2 Pro 298 O (Fig. 2C). The D156K mutant results in the loss of these interactions, and it changes the charge density in that region. Bearing in mind the drastic nature of such a mutation, this residue is unlikely to be critical for inhibition as judged by the limited, albeit significant, decrease in affinity concomitant with this mutation. A previous study has shown that Arg 155 BASI is critical for inhibition of AMY2 (7). This residue makes several intramolecular interactions, as well as a direct hydrogen bond to Ser 208 AMY2 . Therefore it seems that the interactions conferred by Arg 155 grant cardinal importance to this residue as compared with Asp 156 .
AMY2-BASI Contacts via the Fully Hydrated Ca 503 -The mutants designed in the BASI region facing the active site of AMY2 and involving Glu 168 and Tyr 170 in various combinations shed light on an intriguing facet of BASI inhibitory action on The effect of calcium on the AMY2/BASI formation Wild-type BASI was purified from barley seeds. The mutants were made using the intein-CBD tag expression system. The rate constants represent mean of determinations using five AMY2 concentrations (61-245 nM). Measurements were performed at 25°C in 10 mM Mes, pH 6.5, 0.005% surfactant P20, and the indicated ͓CaCl 2 ͔. K i values were calculated as described under "Experimental Procedures." Assay conditions are as follows: 40 mM Tris, pH 8.0, 0.05% bovine serum albumin including the indicated ͓Ca 2ϩ ͔. Standard deviations are shown in parentheses.  (Fig. 2B). The Y170F mutant probably maintains the backbone hydrogen bond to Glu 168 , but it will lose contact with the solvent/Ca 2ϩ system. Because these contacts would be retained by Glu 168 , the effect of Y170F is modest. By contrast, Y170P caused a 200-fold increase in K i . The double mutant E168Q/Y170P reiterates the deleterious effect of introducing the proline side chain as it results in a more than 3 orders of magnitude decrease of inhibitory activity as judged by K i , whereas E168Q shows a modest 25-fold increase of K i . The conservative mutation E168Q could compensate for the loss of the charged interaction by alternative hydrogen bonds either to Trp 206 O or to Lys 182 N in AMY2. In contrast, no inhibitory activity was detected for E168T, and this variant showed a substantial increase in k off , manifesting the loss of important short range interactions.
A previous study has demonstrated that Ca 2ϩ modulates the AMY2-BASI affinity mainly by lowering k off (8). It was not clear if Ca 503 was responsible for this effect or if it was caused by Ca 2ϩ bound to domain B. The present data clearly implicate Ca 503 in the dependence of affinity on Ca 2ϩ as judged by comparison of wild type, Y170F, and E168Q (Table V). By assuming that Ca 503 is associated with the relative affinity change (mainly due to altered k off ), this ion is likely to be bound with an affinity in the lower millimolar range. Insight into the affinity of Ca 503 may be instrumental for assessing the biological significance of such a Ca 2ϩ -modulated inhibition, but an accurate affinity estimate is not warranted by current data. Most interestingly, one Ca 2ϩ was recently identified at the same position in the solved crystal structure of AMY1 in complex with thiomethyl maltotetraoside, excluding this substrate from the active site (53).
Relation to Other ␣-Amylase/Inhibitor Systems and the Energetics of Binding-The determination of several complex structures between ␣-amylases and their inhibitors has unveiled the vast diversity of structural motifs and binding modes characterizing these systems (6). While providing valuable insight, structural data alone fall short of reliably pinpointing energetically important side chains along the interface of a protein complex. It is not uncommon that protein-protein interactions are dominated by a few high energy interactions, so-called "hot spots," with most crystallographically observed contacts making limited contribution to the binding energy (54,55). Typically, a hot spot is defined as a residue that when mutated to an alanine gives rise to a distinct drop of binding affinity of 2 kcal/mol or higher (54,56,57). In BASI, Asp 150 , Glu 168 , and Lys 140 match this definition. The limitations of the used inhibition assay and protein availability precluded an estimate of K i and thus ⌬⌬G of the above-mentioned mutants. Nonetheless, it is evident that the affinity of these mutants decreased several orders of magnitude corresponding to ⌬⌬G values in the 4 -9 kcal/mol range attesting to their thermodynamic contribution. Most interestingly, these residues are clustered in one patch at the center of the inhibitor, whereas the peripheral residue Ser 77 seems less significant for the binding. This is in conformity with reports suggesting that hot spots of binding energy are surrounded by a shell of energetically insignificant interactions referred to as an O-ring for excluding bulk solvent from hot spots (57). The involvement of a divalent ion and solvent in crucial interactions seems to be a unique feature of BASI, which docks on two regions of AMY2 separated by the deep catalytic groove and another minor solvent-filled cavity. The solvent exclusion shell is much more obvious in the complex structure of the lectin type ␣-A1 inhibitor with the mammalian PPA ␣-amylase, where the central region of this complex is protected from bulk solvent (22). By contrast to the hydrophilic nature of the BASI surface interacting with AMY2, ␣-AI, which inserts two loops in the active site cleft of PPA, makes substantial hydrophobic interactions in its substrate mimetic action. Similarly, tendamistat inhibits PPA by inserting a segment presenting its characteristic Trp-Arg-Tyr. A comparison of the energetic contribution of various contact points of these inhibitors with their targets awaits detailed mutational studies. For instance, it has been reported on the basis of homology modeling that mutating Arg 74 , Trp 188 , and Tyr 190 abolished the inhibitory activity of ␣-AI on PPA (26). Inspection of the complex structure, however, shows that only Tyr 190 is in contact with PPA, implying that the loss of activity observed for the two other residues is probably caused by the perturbation of intramolecular interactions rather than loss of crucial contacts to PPA.
In conclusion, this is the first mutational study that addresses the interactions of a Kunitz-type inhibitor with target enzymes. The data presented illuminate the primary role of charge interaction in the context of BASI inhibitory potency on AMY2. Hence, the hot spots identified were charged residues making a range of intra-and intermolecular interactions. Moreover, the data strongly suggest that the fully hydrated Ca 503 at the AMY2-BASI interface modulates the activity of BASI, revealing a novel facet of ␣-amylase inhibition. The involvement of divalent ions and buried solvent in the steric and electrostatic complementarity of BASI and AMY2 affords additional attention and offers an engineering tool for up-or down-regulation of inhibition of this and related systems.