Determinants of Binding Affinity and Specificity for the Interaction of TEM-1 and SME-1 β-Lactamase with β-Lactamase Inhibitory Protein

The hydrolysis of β-lactam antibiotics by class A β-lactamases is a common cause of bacterial resistance to these agents. The β-lactamase inhibitory protein (BLIP) is able to bind and inhibit several class A β-lactamases, including TEM-1 β-lactamase and SME-1 β-lactamase. Although the TEM-1 and SME-1 enzymes share 33% amino acid sequence identity and a similar fold, they differ substantially in surface electrostatic properties and the conformation of a loop-helix region that BLIP binds. Alanine-scanning mutagenesis was performed to identify the residues on BLIP that contribute to its binding affinity for each of these enzymes. The results indicate that the sequence requirements for binding are similar for both enzymes with most of the binding free energy provided by two patches of aromatic residues on the surface of BLIP. Polar residues such as several serines in the interface do not make significant contributions to affinity for either enzyme. In addition, the specificity of binding is significantly altered by mutation of two charged residues, Glu73 and Lys74, that are buried in the structure of the TEM-1·BLIP complex as well as by residues located on two loops that insert into the active site pocket. Based on the results, a E73A/Y50A double mutant was constructed that exhibited a 220,000-fold change in binding specificity for the TEM-1 versus SME-1 enzymes.

Protein-protein interactions play a central role in many cellular processes, such as signal transduction, the immune response, and biochemical regulation of enzyme function. Engineering of protein-protein interactions or the rational design of drugs that disrupt interactions requires an understanding of the physical basis of affinity and specificity within an interface (1). Protein-protein interfaces are commonly large (Ͼ900 Å 2 ) and exhibit good shape and electrostatic complementarity (2)(3)(4)(5). Hydrophobic patches on the surface of one protein commonly pack against hydrophobic patches on the other protein, whereas charged residues also commonly match across an interface. Alanine-scanning mutagenesis has been used to experimentally test the contributions of individual residues to protein interactions (6,7). Studies of several interfaces suggest that functional epitopes determined by mutagenesis are more restricted than structural epitopes revealed by x-ray crystallography (8). The presence of a small subset of "hotspot" residues that contribute significantly to the binding free energy of a complex may be a general feature of protein-protein interfaces (3). Based on current understanding, however, it is not possible to reliably predict the energetic contributions of individual residues based on the structure of a protein-protein interface (5). Therefore, additional structure-function studies are required to generate a detailed understanding of the critical components of protein-protein interfaces.
␤-Lactam antibiotics such as the penicillins and cephalosporins are among the most frequently used antimicrobial agents. Resistance to these drugs is most commonly due to the action of ␤-lactamase enzymes. ␤-Lactamases catalyze the hydrolysis of the amide bond in the ␤-lactam ring to create an ineffective antimicrobial (9). There are four classes (A-D) of ␤-lactamases based on primary sequence homology (10). TEM-1 and SME-1 are class A ␤-lactamases that are found in Gram-negative bacteria. These enzymes are clinically significant because of their ability to confer ␤-lactam antibiotic resistance, which has become a serious threat for human health. Both enzymes are able to hydrolyze most of the penicillins and early cephalosporins but not third generation cephalosporins. SME-1 has a wider substrate spectrum than TEM-1 in that it catalyzes the hydrolysis of carbapenem antibiotics (11). The TEM-1 and SME-1 enzymes are 33% identical at the amino acid sequence level. The x-ray structure of the TEM-1 and SME-1 ␤-lactamases have been solved and indicate the enzymes possess a similar fold (12,13). The charge characteristics of TEM-1 and SME-1 are substantially different, however, with pI values of 5.4 and 9.5, respectively (11).
The ␤-lactamase inhibitory protein (BLIP) 1 is a 165-amino acid protein produced by Streptomyces clavuligerus that has been shown to be a potent inhibitor of class A ␤-lactamases including TEM-1 (14 -16). The x-ray structure of BLIP reveals that it is a flat molecule composed of a tandem repeat of a 76-amino acid domain (16). The two domains form a concave surface that is largely lined with uncharged, polar residues such as serine and tyrosine. However, there are also three tryptophan and two phenylalanine residues that contribute to two hydrophobic patches on the concave surface. The co-crystal structure of BLIP with TEM-1 ␤-lactamase indicates that BLIP uses the large (2636-Å 2 ) concave surface to clamp over a protruding loop and helix region of TEM-1 (residues 99 -114) (17). In addition, at the periphery of the interface, two loops from BLIP insert into the active pocket of TEM-1 to block substrate binding (17). BLIP also contains three charged residues on its surface, Asp 49 , Glu 73 , and Lys 74 , which become buried on complex formation with TEM-1 (17). Asp 49 is on a loop inserted into the TEM-1 active site pocket and has previously been shown to contribute to binding (14,15). The role of residues 73 and 74 has not yet been assessed. However, Albeck and Schreiber (14) have demonstrated that electrostatic interactions between BLIP and TEM-1 strongly influence the association kinetics for complex formation. At neutral pH, both TEM-1 and BLIP carry a net negative charge, and mutations at the periphery of the interface that decrease the negative charge on either molecule increase the association rate (14,18).
In this study, alanine-scanning mutagenesis has been performed to determine the functional epitope of BLIP for binding two class A ␤-lactamases from Gram-negative bacteria. Alanine-scanning mutagenesis is a powerful tool for understanding the contributions of the side chains of individual residues to binding affinity (6,7,19,20). The apparent binding constants (K i ) of wild type BLIP and 23 alanine mutants for the interaction with TEM-1 and SME-1 were determined using a kinetic assay of ␤-lactamase inhibition. Several previous studies have focused on the TEM-1⅐BLIP complex (14,15,17,18,(21)(22)(23). The use of SME-1 in this study allowed an additional investigation into the determinants of binding specificity. It was found that the specificity of binding is significantly altered by mutation of two charged residues, Glu 73 and Lys 74 , that are buried in the structure of the TEM-1⅐BLIP complex as well as by residues located on two loops that insert into the active site pocket. This information was used to engineer a BLIP molecule with a 220,000-fold change in binding specificity relative to wild type.

EXPERIMENTAL PROCEDURES
Materials-All enzymes were purchased from New England Biolabs except for Pfu polymerase, which was purchased from Stratagene. Oligonucleotide primers were purchased from Integrated DNA Technologies. Talon cobalt resin was purchased from Clontech. Cephalosporin C was purchased from Sigma. Cation exchange columns (SP Fast Flow) were purchased from Amersham Biosciences.
BLIP Cloning and PCR Mutagenesis-Wild type, D49A, and F142A BLIPs with N-terminal His tags were constructed previously (15). All other BLIP mutants were constructed by overlapping PCR (24) and cloned into pGR32 with an N-terminal His tag as previously described (15). The external PCR primers used to amplify BLIP were as follows: PD-bla1 (top strand, N-terminal), 5Ј-CGGGGAGCTCGTTTCTTA-GACGTCAGGTGGC-3Ј; MALBLI-2 (bottom strand C-terminal), 5Ј-GG-GAAATCTAGATTATACAAGGTCCCACTGCCG-3Ј. A SacI site in PD-bla1 and an XbaI site in MALBLI-2 allowed the PCR product to be digested with SacI and XbaI and cloned into SacI and XbaI-digested pTP123 following treatment of the vector with calf intestinal alkaline phosphatase (15). The internal top and bottom primers used for constructing the BLIP mutants are shown in Table I. The double mutant E73A/Y50A was constructed by overlap extension PCR using the BLIP Y50A mutant as the template and E73A top and bottom as inside primers. The coding sequences of the R160A and W162A mutants overlap with the MALBLI-2 primer, and therefore the inside primers for the two mutants were not useful, and the external MALBLI-2 was modified to contain the appropriately mutated codon. GCC was chosen as the alanine codon for the mutants because of its abundant usage in Escherichia coli (25). The DNA sequence of each mutant was confirmed by the dideoxy chain termination method using an ABI 3100 capillary DNA sequencer.
BLIP Expression and Purification-Wild type BLIP and the BLIP mutants were expressed and purified using the protocol published previously with minor changes (15). The plasmids containing either wild type BLIP or the mutants were electrotransformed into E. coli RB791 (equivalent to strain W3110 lacI Q L8) (26). An overnight culture of each mutant was grown at 37°C with shaking in 10 ml of LB medium in the presence of 12.5 g/ml chloramphenicol. 1.5 liters of LB medium containing 12.5 g/ml chloramphenicol were then inoculated using the 10-ml overnight culture. The culture was then grown shaking at 37°C until A 600 ϭ 1.2. For induction of BLIP, 3 mM isopropyl-1-thio-␤-Dgalactopyranoside was added to each culture, and the cultures were allowed to grow overnight at 25°C. The cells were pelleted and frozen at Ϫ80°C for at least 1 h. The frozen cells were then thawed and resuspended in 40 ml of B-Per (Pierce). The cell debris was removed by centrifugation, and the soluble protein in the supernatant was purified over a 4-ml TALON cobalt resin column (Clontech) according to the manufacturer's instructions. A 10 mM imidizole wash step was utilized to remove protein from the column that bound less tightly than the His-tagged BLIP. BLIP was eluted using a buffer consisting of 50 mM imidizole in washing buffer (20 mM Tris-HCl and 500 mM NaCl, pH 8.0).  a The coding sequences for these mutants overlap with MALbli-2 primer. Therefore, these constructions were performed by one-step PCR with the PDbla-1 primer and modified Malbli-2 primers as described in Experimental Procedures.
Fractions were examined by SDS-PAGE to estimate purity and yield. Fractions with more than 90% purity were pooled, concentrated, and then changed into 50 mM phosphate, pH 7.0, buffer using Amicon Ultra-4 centrifugal filter devices (Millipore Corp.). SDS-PAGE was used to determine the purity of the BLIP preparations ( Fig. 1). BLIP concentrations were determined by Bradford assay with a standard curve calibrated by quantitative amino acid analysis of purified, wild type BLIP (27).
Kinetic Inhibition Assay-The kinetic assay was performed as described previously with minor changes (15). The assays were performed in a 96-well quartz plate. 12 reactions were monitored simultaneously in a Tecan ultraviolet spectrophotometer controlled by Magellan (Phenix) software. 2 nM TEM-1 ␤-lactamase or 4 nM SME-1 ␤-lactamase was incubated with 12 different concentrations of BLIP for 1 h at 25°C in 50 mM phosphate buffer (pH 7.0) containing 1 mg/ml bovine serum albumin. 100 M cephalosporin C (K m is 700 M for TEM-1 and 1300 M for SME-1) was added to the ␤-lactamase/BLIP incubation buffer with a 12-channel pipetter. The final reaction volume was 0.3 ml, and hydrolysis of cephalosporin C was monitored at A 280 . Plots of the concentration of free ␤-lactamase versus inhibitor concentration were fit by nonlinear regression analysis to Equation 1, where [E free ] is the concentration of active ␤-lactamase calculated from the measured velocity and the activity and concentration of uninhibited ␤-lactamase, [E O ] is the total ␤-lactamase concentration, and [I O ] is the total inhibitor concentration (30). From the equation, apparent equilibrium dissociation constants (K i *) were determined.

RESULTS
Determination of the Functional Epitope for BLIP Binding to TEM-1 ␤-Lactamase-Alanine-scanning mutagenesis of BLIP was performed to identify the determinants critical for binding TEM-1 ␤-lactamase. Amino acid residues in BLIP within 4 Å from TEM-1 in the co-crystal structure were chosen for mutagenesis (17). Wild type BLIP and the 23 BLIP mutants were expressed in E. coli and purified (Fig. 1), and apparent binding constant data were obtained using a kinetic assay of inhibition of ␤-lactam hydrolysis (Fig. 2, Table II). Residues that result in a Ͼ10-fold decrease in binding affinity when mutated are considered components of the functional epitope. By this definition, 11 residues including Phe 36 , His 41 , Asp 49 , Tyr 53 , Lys 74 , Trp 112 , Phe 142 , His 148 , Trp 150 , Arg 160 , and Trp 162 (Fig. 3) comprise the functional epitope. The critical residues are largely located in two patches near the center of the interface as well in the two loops that insert in the active site pocket of the enzyme (Fig. 3).
As described above, two loops on BLIP (loop 1, residues 46 -51; loop 2, residues 136 -144) insert into the active site of TEM-1 ␤-lactamase (17). Within loop 1, Asp 49 forms two salt bridges and two hydrogen bonds with four conserved TEM-1 residues that are essential for enzyme activity: Ser 130 , Lys 234 , Ser 235 , and Arg 244 (17). The carboxylate of Asp 49 assumes the position of the carboxylate of penicillin G in the acyl-enzyme complex of TEM-1 with penicillin G (17). Phe 142 accounts for most of the van der Waals interactions in loop 2, and it mimics the benzyl group of penicillin G for binding TEM-1 (17). As expected based on these interactions and consistent with previous results, mutation of these residues to alanine results in Ͼ20-fold loss of binding affinity for TEM-1 ␤-lactamase (Table  II). Other loop residues do not play a significant role in binding with the exception of Tyr 50 . Surprisingly, mutation of this residue to alanine results in a ϳ50-fold increase in binding affinity. Tyr 50 forms extensive van der Waals interactions interactions with Pro 107 , Val 216 , and Met 129 of TEM-1. In addition, Tyr 50 forms an aromatic patch on BLIP together with residues Tyr 51 , Phe 36 , His 41 , and Tyr 53 . The aromatic ring of Tyr 50 is located between the hydrophobic rings of Pro 107 of TEM-1 and Tyr 51 of BLIP. Perhaps the Tyr 50 side chain hinders Asp 49 or other residues in loop 1 from assuming optimal interactions in the active site of TEM-1 ␤-lactamase. For example, removal of the bulky Tyr 50 side chain could result in better contacts between BLIP residues His 41 and Tyr 51 and the TEM-1 enzyme.
The x-ray structure of the BLIP⅐TEM-1 complex reveals several ordered water molecules at the interface between BLIP and the Gln 99 -His 112 region of ␤-lactamase (17). The water molecules are clustered near polar residues such as several serines on the concave surface of BLIP. Mutation of Ser 35 , Ser 39 , Ser 71 , and Ser 113 at the interface had no effect on binding (Table II). It has been proposed that the intervening waters at the BLIP-TEM-1 interface may form an adaptable interface that allows BLIP to bind several different ␤-lactamase sequences (17). These results suggest that the BLIP side of this interface is also adaptable in that the serine residues do not contribute to binding affinity.
Two negatively charged BLIP residues in the interface, Glu 31 and Glu 73 , do not contribute to binding affinity (Table II). There is a salt bridge between Glu 31 and Lys 215 of TEM-1 that Analysis of ␤-Lactamase-BLIP Interactions is partially exposed to the solvent (17). The solvation effect may limit the contribution of Glu 31 to binding affinity. Glu 73 is fully buried inside the interface and makes hydrogen bonds with Glu 104 , Tyr 105 , and Ser 106 of TEM-1. Removal of the Glu 73 side chain, however, does not affect binding affinity to TEM-1 (Table II).
In contrast, mutation of several positively charged residues on the binding surface of BLIP significantly increased K i values (Table II). The mutated residues include Lys 74 , Arg 160 , His 41 , and His 148 . The significance of the histidine residues may arise from their ring structure rather than the partial positive charge, which will be discussed below. Lys 74 is fully buried at the interface and makes a salt bridge with Glu 104 from TEM-1 and neutralizes a negative region formed by Glu 73 of BLIP and Glu 104 of TEM-1 (17). Schreiber and colleagues (14,18) have shown that electrostatic effects strongly influence the associa-tion rate and thereby the equilibrium constant for the BLIP-TEM-1 interaction. It was observed that mutations that reduce the negative charge on either molecule increase the association rate. Since removal of the Lys 74 side chain increases the negative charge of BLIP, the mutation may act by slowing the association rate. A similar explanation may apply to Arg 160 , which is near the periphery of the interaction surface. However, the aliphatic portion of the side chain also makes packing interactions with His 148 , Trp 150 , and Trp 162 in BLIP that, as described below, form an aromatic patch on BLIP that is critical for binding affinity. The only positively charged residue that did not affect binding when mutated to alanine was

FIG. 3. Dissection of the BLIP-␤-lactamase interaction interface.
A, space fill model of the BLIP-␤-lactamase structure determined by Strynadka et al. (17). TEM-1 ␤-lactamase is colored white, and BLIP is colored yellow. B, same view as in A, except the TEM-1 molecule has been deleted from the figure. Residues identified as critical (Ͼ10-fold increase in K i ) in the alanine-scanning experiments are colored red. Residues that had no effect on binding when changed to alanine are colored green, whereas residues that influence the substrate specificity of binding are colored cyan. Residues that were not mutated are colored yellow. The aromatic patch from domain one includes residues Phe 36 , His 41 , Tyr 50 , and Tyr 53 , whereas the patch from domain two includes residues His 148 , Trp 150 , and Trp 162 . The figure was prepared with Py-MOL (W. L. DeLano; available on the World Wide Web at www.pymol.org).
Arg 144 . This residue is at the edge of loop 2 on the periphery of the binding surface and remains largely exposed on complex formation. The lack of effect is not surprising from the standpoint of its limited direct interactions with ␤-lactamase, but it might have been expected to decrease affinity based on the change in electrostatics (14,18).
In contrast to the diverse effect of mutating hydrophilic residues, the majority of the hydrophobic residues tested contribute significantly to the BLIP⅐TEM-1 interaction (Table II). The mutation of these residues to alanine increases K i values at least 20-fold. It should be noted that all hydrophobic residues mutated in this study are also aromatic residues. Two patches of aromatic residues account for most of the binding affinity, with His 148 , Trp 150 , and Trp 162 forming one patch and Phe 36 , His 41 , Tyr 50 , and Tyr 53 near loop 1 forming the other (Fig. 3). Mutation of Trp 150 results in the largest increase (370-fold) in the K i value among all mutants tested (Table II). The patch centered at His 148 makes extensive interactions with Gln 99 -Asn 100 and Leu 102 -Val 103 in TEM-1 ␤-lactamase, whereas the patch near loop 1 makes contacts with Ser 106 -Pro 107 -Val 108 of TEM-1. Leu 102 -Val 103 and Pro 107 -Val 108 are conserved positions in the loop-helix domain among several class A ␤-lactamases (10). In addition, a previous study from this laboratory demonstrated that residues Leu 102 , Val 103 , Ser 106 , Pro 107 , and Val 108 in TEM-1 are important for binding to BLIP (23). Therefore, the residues in the critical aromatic patches in BLIP largely interact with TEM-1 residues that are also important for binding. This suggests that the residues in the aromatic patches make precise interactions with the TEM-1 residues in these regions.
Determination of the Functional Epitope for BLIP Binding to SME-1 ␤-Lactamase-SME-1 ␤-lactamase is 33% identical to TEM-1 but differs significantly from TEM-1 in both surface charge and shape in the loop-helix region that BLIP binds (Fig.  4). The inhibition assay indicated that wild type BLIP inhibits SME-1 with a K i of 2.4 nM, which is ϳ5-fold weaker than the interaction with TEM-1 (Fig. 2, Table II). Clearly, the difference in both surface charge and shape of SME-1 does not abolish the protein-protein interaction. The differences between the TEM-1 and SME-1 ␤-lactamases provide an interesting system with which to investigate the determinants of binding specificity for BLIP.
The functional epitope of BLIP for binding SME-1 (Ͼ10-fold increase in K i ) consists of 12 residues: Phe 36 (Table II). The sequence requirements for the SME-1-BLIP interaction are, in general, similar to those for the TEM-1-BLIP interaction. For example, the two aromatic patches critical for the TEM-1-BLIP interaction are also critical for the SME-1-BLIP interaction. The loop 1 residue Asp 49 that makes multiple interactions in the active site of TEM-1 is also impor-tant for binding to SME-1, as is the positively charged residue Arg 160 . Finally, the multiple serine residues are not essential for the interaction of BLIP with either enzyme.
There are, however, two significant differences between the functional epitopes of BLIP for TEM-1 and SME-1. First, two charged residues buried at the interface strongly influence binding specificity. Glu 73 is very important for binding to SME-1 ␤-lactamase but does not contribute to binding TEM-1 ␤-lactamase. In contrast, Lys 74 is important for binding TEM-1, but the K74A substitution actually binds SME-1 10fold tighter than wild type BLIP. As stated above, Lys 74 of BLIP interacts with Glu 73 of BLIP and Glu 104 of TEM-1 ␤-lactamase. Based on a sequence alignment between TEM-1 and SME-1, the equivalent residue to Glu 104 of TEM-1 is Tyr 104 of SME-1, which would interact differently with Lys 74 . In addition, the topology of the loop-helix region is somewhat different between TEM-1 and SME-1 in this region (Fig. 4), which could alter multiple interactions. Finally, the mutation of Lys 74 increases the negative charge on BLIP, which may augment electrostatic interactions with the positively charged SME-1 (14). It should be noted, however, that mutation of Arg 144 and Arg 160 would have a similar effect on electrostatics, but these mutations either do not affect or are detrimental for binding to SME-1 (Table II).
The second significant difference in sequence requirements for binding the enzymes is that mutation of loop 1 residues Tyr 50 and Tyr 51 either improves or does not affect binding of BLIP to TEM-1 ␤-lactamase but significantly reduces affinity for SME-1. In contrast, mutation of loop 2 residue Phe 142 significantly reduces binding to TEM-1 but has a relatively minor effect on binding affinity for SME-1. In fact, none of the loop 2 residues examined displayed significantly altered K i values for SME-1 when mutated to alanine. Taken together, these data suggest that loop 1 of BLIP does not make optimal interactions with the TEM-1 active site, whereas loop 2 of BLIP does not interact strongly with the SME-1 active site. Subtle differences in the active sites of the enzymes may influence the loop interactions. Alternatively, differences in the interactions of BLIP with the loop-helix region of the ␤-lactamases may alter the positioning of the loops with respect to the enzymes. Although the exact mechanisms for changes in binding affinity are not clear, it is apparent that a relatively small number of residues, Glu 73 , Lys 74 , Tyr 50 , Tyr 51 , and Phe 142 , are important determinants of the binding specificity of BLIP.
Engineering the Binding Specificity of BLIP-The observation that single alanine substitutions result in large changes in the binding specificity of BLIP suggested that the binding specificity could be drastically altered with just a few changes in sequence (Fig. 5). For example, the BLIP mutant containing the Y50A substitution binds TEM-1 ␤-lactamase 50-fold tighter than wild type but binds SME-1 13-fold weaker. In addition, the E73A mutant binds TEM-1 with the same affinity as wild type BLIP but binds SME-1 1,100-fold weaker than wild type (Fig. 5, Table II). Because the Tyr 50 and Glu 73 residues are not in direct contact in the BLIP structure and mutations of residues that are not in contact often act additively when combined (31), it was reasoned that the Y50A/E73A double mutant would possess the combined effects of the mutations and thereby exhibit a vast difference in binding specificity for TEM-1 versus SME-1 ␤-lactamase. This was found to be the case, since the Y50A/E73A double mutant binds TEM-1 ␤-lactamase ϳ10-fold tighter than wild type BLIP, whereas it binds SME-1 18,300fold weaker than wild type (Fig. 5, Table II). As a result, wild type BLIP exhibits a 5-fold preference for binding TEM-1 versus SME-1 ␤-lactamase, whereas the double mutant displays a 1.1 million-fold preference for binding TEM-1 versus SME-1 FIG. 4. Electrostatic surface representation of TEM-1 and SME-1 ␤-lactamase. Regions of negative charge are shown in red, and positive charge is shown in blue. The boxed region of each enzyme is bound by BLIP. The yellow circle denotes the active site pocket. The figure was generated using the GRASP program (38) and the PDB coordinates 1BTL (TEM-1) (39) and 1DY6 (SME-1) (12).
␤-lactamase (Fig. 5, Table II). Therefore, the binding specificity of BLIP can be drastically altered with relatively few amino acid substitutions. DISCUSSION The common functional epitope of BLIP for binding both TEM-1 and SME-1 ␤-lactamase consists of 9 of the 23 residues mutagenized and includes positions Phe 36 , His 41 , Asp 49 , Tyr 53 , Trp 112 , His 148 , Trp 150 , Arg 160 , and Trp 162 (Fig. 3). These residues represent two aromatic patches in addition to Asp 49 from loop 1. It is interesting that each domain of BLIP contributes an aromatic patch (Fig. 3). The locations of the patches with respect to the binding surfaces, however, are somewhat different. For example, the patch from domain one includes residues Phe 36 , His 41 , and Tyr 53 and is located in the center of the binding interface provided by domain one. The periphery of this patch contains the polar residues Ser 35 , Ser 39 , and Ser 71 , which do not contribute substantially to binding of either ␤-lactamase. This type of organization has been noted for several protein-protein interaction interfaces, and it has been proposed that the major role of the residues on the periphery of the critical residues is to serve as an "O-ring" to occlude bulk solvent from the hotspot (8).
The aromatic patch from domain two includes residues His 148 , Trp 150 , and Trp 162 . Although not an aromatic residue, Arg 160 is critical for binding and also packs into this hotspot (Fig. 3). In contrast to the patch from domain one, this patch is located near the periphery of the binding interface (Fig. 3). For example, His 148 , Trp 150 , Arg 160 , and Trp 162 retain 1.1, 28.6, 71.2, and 26.0 Å 2 of accessible surface area in the complex, respectively, due to their position near the edge of the interface. In contrast, Phe 36 , His 41 , and Tyr 53 from patch one exhibit only 0.3, 0.0, and 0.46 Å 2 of accessible surface area in the complex, respectively, due to their more centered position in the domain one binding interface. Therefore, it appears that the aromatic patch in domain two does not require a ring of surrounding residues to protect the hotspot (8). Finally, it is of interest that Trp 112 is critical for binding but does not belong to either of the aromatic patches. Rather, Trp 112 serves as a bridge between the aromatic patches from domains one and two (Fig. 3B).
Analysis of an extensive alanine-scanning database has shown that the amino acid composition of hotspots is enriched for tryptophan, arginine, tyrosine, aspartate, isoleucine, and histidine residues (8,32). With the exception of isoleucine, these residues are also enriched on the binding surface of BLIP (17). Of the 23 BLIP residues within 4 Å of the TEM-1 ␤-lactamase surface, 13 are within this class. In general, these residues were also found to be important for the ␤-lactamase-BLIP interaction (Table II). For example, all three of the tryptophan residues, both of the histidine residues, and the loop 1 aspartate residue at the interface are important for binding to both TEM-1 and SME-1 ␤-lactamase. In contrast, only one of the four tyrosine residues at the interface contributes significantly to binding TEM-1 ␤-lactamase based on the mutation results. However, two additional tyrosines on loop 1, Tyr 50 and Tyr 51 , are important for binding SME-1 ␤-lactamase. Phenylalanine is not enriched at hotspots in the alaninescanning database (32). Nevertheless, it is important at the BLIP interface in that Phe 36 is a critical residue from one of the aromatic patches and Phe 142 on loop 2 is critical for binding TEM-1 ␤-lactamase. These results are consistent with an analysis of structurally conserved residues and binding energy hotspots that showed tryptophan and phenylalanine are conserved at interaction surfaces but not on the exposed surface of proteins (33). In addition, analysis of the alanine-scanning database as well as structurally conserved residues in binding sites indicates that serine residues are very rarely found in energy hotspots (8,33). This is true for the BLIP interface as well in that none of the four serines at the interaction interface make a significant contribution to the binding of either ␤-lactamase. A similar result was reported for the peripheral polar residues in the human growth hormone and hormone receptor complex (20). The serines on the periphery of BLIP interact with water molecules that bridge the interface with ␤-lactamase. The intervening water molecules may contribute to the adaptability of the periphery of the interface.
BLIP binds a number of ␤-lactamases from both Gram-negative and Gram-positive bacteria (16). Examination of the binding of the alanine scan mutants to TEM-1 and SME-1 ␤-lactamase indicates that most of the positions, including the two aromatic patches, make similar contributions to binding each enzyme. Only a relatively small number of residues influence binding specificity. Two fully buried charged residues in BLIP, Glu 73 and Lys 74 , contribute to the specificity of binding, as do Tyr 50 and Tyr 51 located on loop 1 and Phe 142 on loop 2. Schreiber and colleagues (14,18) have shown that alteration of charged residues on the periphery of the BLIP-TEM-1 ␤-lactamase interface can strongly influence the association rate and thereby the equilibrium binding constant. Although Glu 73 and Lys 74 are not in the periphery but rather fully buried in the binding interface, it is possible that they could act on the association rate. It has previously been suggested that buried electrostatic interactions may contribute to specificity (34). Further studies are required, however, to determine the mechanism for the specificity change associated with Glu 73 and Lys 74 .
The roles of Tyr 50 , Tyr 51 , and Phe 142 in binding specificity may be due to altered interactions between loop 1 and loop 2 with the TEM-1 and SME-1 active sites. Removal of the Tyr 50 side chain significantly increases binding affinity of BLIP with TEM-1 but not SME-1 ␤-lactamase (Table II). The side chain of Tyr 50 fits into a pocket formed by TEM-1 residues Pro 107 , Met 129 , and Val 216 (17). Although the catalytic machinery of the SME-1 active site is similar to TEM-1, there are several differences that could prevent similar interactions from occurring. For example, the equivalent Tyr 50 -interacting residues in SME-1 are Pro 107 , Tyr 129 , and Thr 216 that form a somewhat FIG. 5. Comparison of ⌬⌬G values for binding to TEM-1 (white bars) and SME-1 (black bars) for BLIP alanine mutants. Relative binding free energy is the difference between binding free energy of wild-type BLIP versus the alanine mutants: ⌬⌬G ϭ ϪRTln(K i w.t./ K i mut), where R is the gas constant and T stands for absolute temperature. The units of ⌬⌬G are kcal/mol. altered binding pocket. In contrast, Tyr 51 packs onto Tyr 50 but does not directly contact ␤-lactamase. Changes at Tyr 51 , however, could alter the position of Tyr 50 or the conformation of loop 1 and thereby affect binding.
Replacement of the Phe 142 side chain with alanine significantly reduces binding affinity of BLIP with TEM-1 but not SME-1 ␤-lactamase. The Phe 142 side chain interacts with several TEM-1 active site residues including Glu 104 , Tyr 105 , Asn 170 , Gly 238 , and Glu 240 . The equivalent residues in SME-1 are Tyr 104 , His 105 , Asn 170 , Cys 238 , and Ala 240 . These differences may result in a poor interaction between loop 2 and SME-1 and therefore little contribution from Phe 142 .
The ability to engineer protein-protein interactions or design binding sites could facilitate the design of new therapeutics. One interesting challenge is to engineer the specificity of protein interactions. Wild type BLIP exhibits a 5-fold binding affinity preference for TEM-1 versus SME-1 ␤-lactamase. The results of the mutational analysis, however, suggested that the binding specificity of BLIP could be tailored to interact strongly with one substrate but not another. For example, the K74A mutant exhibits a significant change in binding specificity in that it binds SME-1 460-fold tighter than TEM-1 rather than 5-fold weaker as seen with wild type BLIP. Thus, a single amino acid substitution in BLIP resulted in a 2300-fold change in specificity versus wild type.
The observation that the E73A mutation greatly reduced BLIP binding to SME-1 while not affecting binding to TEM-1 as well as the finding that the Y50A mutant bound more tightly to TEM-1 but not SME-1 suggested that the Y50A/E73A double mutant would exhibit greatly altered specificity. This was the case, since the double mutant exhibited a ϳ1.1 million-fold preference for binding TEM-1 versus SME-1 ␤-lactamase. Because wild type BLIP exhibits a 5-fold preference for TEM-1 the double mutant exhibits a 220,000-fold change in binding specificity.
One reason for the changes in specificity observed here is that the Tyr 50 side chain not only does not favor binding to TEM-1 ␤-lactamase; it actually decreases affinity by an order of magnitude. Similarly, Lys 74 side chain decreases affinity of BLIP for SME-1 by more than an order of magnitude. A large increase in affinity upon mutation of a residue has not been frequently observed in other protein-protein interfaces (6,7,35,36). Nevertheless, it is clear from these results that the binding affinity of BLIP for either TEM-1 or SME-1 ␤-lactamase has not been optimized by evolution. This is not surprising in that it is very unlikely that BLIP evolved to bind either TEM-1 or SME-1 ␤-lactamase, since BLIP is produced by a Gram-positive soil bacterium, and TEM-1 and SME-1 ␤-lactamase are found in Gram-negative enteric bacteria (37). The role of BLIP in the biology of S. clavuligerus is not known, so it is difficult to speculate on the selective pressures that did shape BLIP function.
Although BLIP binds to several class A ␤-lactamases, it does not efficiently bind other classes of ␤-lactamase or penicillinbinding proteins (16). The results presented here suggest that BLIP can be engineered to serve as an inhibitor of a number of clinically important ␤-lactamases and penicillin-binding proteins. More generally, the results suggest that the specificity of protein-protein interactions can be modified with limited amino acid substitutions.