We performed a scanning mutagenesis study of heavy chain complementarity-determining region (CDR) residues to identify how mutations affected binding of the anti-carcinoma mAb BR96 to Ag, Lewis Y, and to an anti-Id Ab (anti-Id). By ELISA, we demonstrated that the anti-Id bound close to the Ag binding site of BR96, but the anti-Id and Ag sites were not identical. Immunoblot analysis and screening of light and heavy chain CDR libraries with multiple mutations in each CDR suggested that the heavy chain had greater involvement in anti-Id binding. We then analyzed contributions of individual residues in the heavy chain CDRs to binding of Ag and anti-Id. In a filamentous phage vector containing BR96 V region sequences, mutations were introduced by codon-based mutagenesis at single positions within the three heavy chain CDRs. The resulting libraries of Fab fragments had all amino acids represented at a CDR position. We evaluated the expressed Fabs for binding to Ag and anti-Id by plaque lift assay. We identified the positions with mutations that had the greatest negative effect on binding to the anti-Id and to Ag and analyzed them on the basis of the BR96 x-ray structure. The residues most important for binding to the anti-Id were located in heavy chain CDR1 and CDR2 and were peripheral to the residues within the Lewis Y binding pocket.

Anti-Id Abs recognize antigenic determinants in the V region of other Igs. Anti-Id reagents are used frequently in research and in support of clinical studies with therapeutic Abs. As therapeutic agents in themselves, anti-Ids have been investigated since Jerne’s proposal of the Ab/anti-idiotypic network (1). This led to the concept that, if an anti-Id resembled or mimicked the Ag recognized by the Ab, the anti-Id might substitute as Ag to stimulate an immune response, essentially acting as a vaccine (2, 3, 4, 5, 6, 7). An anti-Id would be a reasonable alternative to an Ag-based vaccine, particularly if the Ag were difficult to produce and purify in large quantities. The potential utility of anti-Ids for this purpose has prompted studies of the structural relationship between the Ag binding site of an Ab and the sites to which anti-Ids bind (8, 9, 10).

The three-dimensional structure of Abs alone or complexed with Ag has been defined for numerous mAbs by x-ray crystallography. Few studies have examined mAbs complexed with an anti-Id (9, 11), or have compared the detailed interaction of an Ab to Ag and anti-Id (8, 10, 12). The ultimate demonstration that an anti-Id represents a good mimic of Ag is empirical, i.e., to immunize an animal with the anti-Id and show that the animal’s immune response includes Abs that bind to Ag. This approach is laborious and not necessarily predictive of the human immune response. A more direct structural and functional comparison of Ab binding to Ag and anti-Ids is possible in view of the advances in molecular biology techniques for manipulating and expressing functionally active Ab fragments (13, 14, 15, 16, 17).

The use of filamentous phage vectors for the expression of Fv, sFv, or Fab fragments in bacteria allows rapid examination of multiple clones with mutations in the V region sequence. We prepared libraries of Fab fragments with mutations in the CDRs4 of BR96, a mAb that binds the Ley carbohydrate tumor Ag (18, 19). These libraries were originally designed and created for the isolation of chimeric BR96 (chiBR96) (20) Fab fragments with increased affinity for tumor Ag (21). In this study, we probed the libraries with synthetic Ley (sLey) (21) Ag and with an anti-Id to identify how mutations at a CDR position affect binding to each ligand. By identifying the residues critical for binding, the method provides a structural footprint that compares the binding sites of an anti-Id and Ag at a functional level. The identified residues were compared on the basis of x-ray coordinates of chiBR96 Fab complexed with Ley (22). Despite the proximity of the anti-Id epitope to the Ag binding site, this analysis clearly reveals several Ab residues that influence Ag binding differently from anti-Id binding.

BALB/c mice were immunized repeatedly with murine BR96 IgG3 and the isotype switch variants IgG1 and IgG2a (23). Fusions were performed with splenic cells from the animals and the P3×63-AG-8.653 myeloma line (24). One cell line produced Ab that bound to BR96, but not to isotype-matched nonspecific Abs. The anti-Id is called 757.

By ELISA, we assessed binding of 757 to chiBR96 once chiBR96 was bound to Ag. The assay was performed essentially as described (21, 25). Briefly, microtiter plates (Immulon II; Dynatech Labs., Chantilly, VA) were coated with sLey conjugated to human serum albumin (sLey-HSA) (Alberta Research Council (ARC), Edmonton, Alberta, Canada) or with H3396 tumor cell membranes overnight at 4°C. Binding of chiBR96 and chiBR96 Fab to Ag was detected with biotinylated 757, followed by avidin conjugated to horseradish peroxidase (HRP) (Vector Labs., Burlingame, CA). ChiBR96 Fab was derived proteolytically by papain digestion of chiBR96 Ig.

ChiBR96 (20 ng) and chiBR96 Fab (20 ng) were applied to a 4 to 20% Tris-glycine gel (Novex, San Diego, CA) in sample buffer under nonreducing and reducing conditions. After electrophoresis, the Igs were transferred electrophoretically to a nitrocellulose membrane at 200 mA for 1 h in 0.1 M Tris-glycine, pH 8.3, 20% methanol buffer (26). Immunoblotting was performed essentially as described (27). After the blocking step, 757 (1 μg/ml) was applied to the nitrocellulose membrane overnight at 4°C. Binding of 757 to the Ig chains was detected with goat anti-murine IgG (Fc specific) conjugated to alkaline phosphatase (Jackson ImmunoResearch Laboratories, West Grove, PA).

Molecular modeling of BR96 heavy and light chain V regions (28) showed the Ag binding site was better defined by canonical structure (29) than by sequence-defined hypervariable regions (30). Based upon the canonical definition, the following light chain regions were chosen for mutagenesis: Val30-Tyr37 in CDR1 (L1); Tyr54-Ser57 in CDR2 (L2); and Gly96-Phe101 in CDR3 (L3) (31). The residues in the heavy chain V region targeted for amino acid substitutions were Gly26-Asp31 in CDR1 (H1); Tyr50-Asp55 in CDR2 (H2); and Gly95-Trp100a in CDR3 (H3) (21, 31). The sequence-numbering system conforms to Kabat et al. (30).

The M13 phage BR96 Fab expression vector, IXBR96, was constructed as described (21). BR96 VL and VH sequences precede the human constant κ sequence and human IgG1 CH1 plus five residues of the IgG1 hinge, respectively. A decapeptide sequence (21) is appended to the carboxyl terminus of the heavy chain C region. The light chain and heavy chain are directed to the periplasmic space by bacterial leader peptides, pelB and phoA, respectively, to the periplasmic space, where the chains assemble into Fabs.

Construction of large multiresidue libraries by codon-based mutagenesis was performed as discussed previously (21). Briefly, chiBR96 vectors were modified to include a short deletion, followed by a stop codon within each of the light and heavy chain CDRs, creating six vectors, each with one modified CDR. These modifications prevented wild-type expression of chiBR96 Fab from the template, and therefore, ensured that expression of Fab from a library would be the result of correctly incorporated oligonucleotides. The oligonucleotides were synthesized with a 50% bias for the parent sequence and encompassed the CDR residues per hypervariable region listed above (21). Libraries contained 106 to 108 members, predominantly single, double, and triple mutations (21).

Single residue libraries for the heavy chain CDRs were also prepared. The single residue libraries had a membership of 32, with one mutation per phage, representing all 20 amino acids and one translation termination codon. The residues for individual CDR residue libraries were H1 positions 26–33; H2 positions 50, 52–56, and 58; and H3 positions 95–99 and 100a. Libraries for positions 51, 57, and 100 were technically difficult to make, and because we were unsure they contained all mutant codons, we did not include them in this study.

Oligonucleotides were synthesized by β-cyanoethyl phosphoramidite chemistry on an ABI 394 DNA synthesizer (Applied Biosystems, Foster City, CA) and purified as described (21). Mutagenic oligonucleotides encoding random amino acid residues at specified positions were synthesized by incorporating the trinucleotide C/ANN (n = A, C, G, T). The oligonucleotides were constructed with 15 to 20 bases from the chiBR96 vector flanking the mutagenic site. Each set of oligonucleotide mixtures was phosphorylated and annealed to the appropriate CDR-modified IXBR96 templates. Uridinylated ssDNA IXBR96 templates were prepared in the Escherichia coli dut ung strain CJ236 (Bio-Rad, Hercules, CA), and site-directed mutagenesis reactions were performed according to the method of Kunkel (32), as described by the reagent manufacturer (Bio-Rad). A portion of the extended, ligated mutagenesis reaction was electroporated into E. coli strain DH10B (Life Technologies, Gaithersburg, MD) and titered onto a lawn of E. coli strain XL-1 Blue (Stratagene, San Diego, CA). Each library was amplified by infecting a 50-ml liquid culture of XL-1 Blue in 2×YT broth + 10 μg/ml tetracycline and then growing the infected strain for 5 h at 37°C. Bacteria were removed by centrifugation and the libraries were stored at 4°C. T4 polynucleotide kinase, T4 DNA ligase, and T4 DNA polymerase were purchased from Boehringer Mannheim Corp. (Indianapolis, IN).

Filter lifts from plated phage libraries were prepared (33) and analyzed as follows. Plaque lift assays probed with 2 μg/ml sLey hydrazide conjugated to horseradish peroxidase (sLey-HRP) were performed as described (21). Additional probes used included: goat anti-human κ light chain conjugated to alkaline phosphatase (Southern Biotechnology, Birmingham, AL); anti-decapeptide mAb 52.4 (1 μg/ml) (34); and purified 757 (1 μg/ml). The murine mAbs 52.4 and 757 were detected with a goat anti-murine IgG (Fc-specific) reagent conjugated to alkaline phosphatase (Jackson ImmunoResearch Laboratories). Bromochloroindolyl phosphate-nitro blue tetrazolium was the chromogen used for the alkaline phosphatase reactions (35). In this study, we did not distinguish between weak and strong reactions when enumerating plaques. If a plaque stained with a probe, the reaction was considered positive. Although weak staining may reflect poorer interaction with Ag or anti-Id, expression levels can also vary from assay to assay. A more objective assessment of the data was achieved by counting plaques that were positive independent of the intensity of the signal.

A representative subset of BR96 mutant clones from one single residue library was selected for DNA sequence analysis. Phage ssDNA was prepared, and the heavy chain CDR region of interest was sequenced by the dideoxynucleotide termination method (Sequenase Version 2; United States Biochemical Corp., Cleveland, OH).

Computer graphic analysis was completed with the Insight II program (MSI, San Diego, CA) using the x-ray coordinates of chiBR96 Fab in complex with the nonoate methyl ester derivative of Lewis Y (nLey) (pdb code “1CLY”) (22). Color figures were produced with Insight II.

The 757 anti-Id Ab recognizing BR96 was raised in syngeneic mice immunized multiple times with the original isotype of BR96, IgG3, and the IgG1 and IgG2a isotype switch variants. Because the syngeneic immune response was weak, 757 was the only BR96-specific anti-Id isolated. We have subsequently immunized BALB/c mice with a humanized form of BR96 and generated many anti-Id Abs with a variety of binding characteristics, including several that recognized chiBR96, although with lower affinity than 757, showing that in a different framework context the BR96 CDR regions are immunogenic.

The BR96 anti-Id was generated as a reagent for pharmacokinetic assays and immunohistochemistry to support clinical trials with chiBR96. Detection of a chimeric or human mAb in human sera and tissues requires a specific reagent, since commercially available reagents cannot distinguish between the therapeutic Ab and endogenous Abs. To demonstrate specificity of an anti-Id, a binding assay is often performed in which mAb and anti-Id are preincubated; then the solution is added to Ag, and binding of the mAb is measured (36). When this experiment was performed in an ELISA format with 757, the anti-Id did block binding of chiBR96 to Ag. Therefore, we knew that the anti-Id was not recognizing framework residues distal to the Ag binding site.

An anti-Id may block binding of the mAb to Ag, but still detect the mAb if the mAb is allowed to bind Ag first (36). To investigate further how the anti-Id and Ag binding sites were related, we performed an ELISA in which chiBR96 was first bound to Ag, followed by incubation with biotinylated 757. For this experiment, both sLey-HSA and H3396 tumor cell membranes were used as the source of Ag. We also performed this assay with chiBR96 Fab to rule out the possibility that only one binding site of the IgG bound to Ag, leaving the other binding site accessible to the anti-Id. The anti-Id recognized chiBR96 whole Ab and the Fab fragment when they bound to either the synthetic Ag or to tumor cell membranes (Fig. 1). The Fab fragment has 50 to 100 times less binding activity on both Ags due to loss of avidity (18, 21). The data showed that the Ag and anti-Id binding sites were not identical. This did not exclude the possibility that the binding sites have some structural similarity and may share certain CDR residues of BR96.

FIGURE 1.

Binding of chiBR96 and chiBR96 Fab to Ag detected with biotinylated 757. An ELISA was performed to assess whether the anti-Id 757 would bind to chiBR96 once the mAb was bound to Ag. ChiBR96 and its proteolytically derived Fab were incubated on sLey-HSA and H3396 tumor cell membranes, followed by incubation with biotinylated 757.

FIGURE 1.

Binding of chiBR96 and chiBR96 Fab to Ag detected with biotinylated 757. An ELISA was performed to assess whether the anti-Id 757 would bind to chiBR96 once the mAb was bound to Ag. ChiBR96 and its proteolytically derived Fab were incubated on sLey-HSA and H3396 tumor cell membranes, followed by incubation with biotinylated 757.

Close modal

Heavy chain CDR residues generally contribute more to Ag binding than do light chains (37, 38, 39). For BR96, greater than 60% of the contacts with sLey are formed to heavy chain CDR residues (22). To investigate whether we could detect any interaction of 757 with heavy or light chain by immunoblot, we performed an experiment with nonreduced and reduced chiBR96 whole Ab and Fab.

Detection by 757 was greatest with nonreduced Ab and Fab fragment (Fig. 2, lanes 1 and 2). When the whole Ab was reduced (Fig. 2, lane 4), no staining of the light chain was observed, and staining of the heavy chain was significantly less than observed with nonreduced Ab. Since the light chain of the reduced whole Ab was not stained by 757 (lane 4), we assume that the band visible in the reduced Fab fragment (lane 3) is also heavy chain. The two chains of the Fab fragment comigrate such that the two cannot be distinguished by immunoblot with specific reagents. While the anti-Id did not bind the κ light chain alone, the immunoblot does not eliminate the possibility that the interaction of heavy and light chain together is important for optimal binding, and that conformational epitopes are recognized. The reduced heavy chain was recognized by the anti-Id, indicating that either the heavy chain has greater involvement with 757 or that the heavy chain epitope has greater linear sequence preserved in an immunoblot.

FIGURE 2.

Immunoblot analysis of 757 binding to chiBR96 heavy and light chains. Twenty nanograms each of chiBR96 Fab (lanes 1 and 3) and chiBR96 whole Ig (lanes 2 and 4) were separated by SDS-PAGE under nonreducing (lanes 1 and 2) and reducing (lanes 3 and 4) conditions. After electrophoresis, the proteins were transferred to nitrocellulose and analyzed by immunoblot with the anti-Id 757 (2 μg/ml). Molecular weights of marker proteins are indicated at the left (K = thousands).

FIGURE 2.

Immunoblot analysis of 757 binding to chiBR96 heavy and light chains. Twenty nanograms each of chiBR96 Fab (lanes 1 and 3) and chiBR96 whole Ig (lanes 2 and 4) were separated by SDS-PAGE under nonreducing (lanes 1 and 2) and reducing (lanes 3 and 4) conditions. After electrophoresis, the proteins were transferred to nitrocellulose and analyzed by immunoblot with the anti-Id 757 (2 μg/ml). Molecular weights of marker proteins are indicated at the left (K = thousands).

Close modal

Because BR96 has a relatively low affinity for tumor Ag (Ka = 107 M−1) (18), we created filamentous phage libraries consisting of mutations within each of the BR96 CDRs to identify mutants with higher affinity (21). Each library consisted of 106 to 108 members, depending on the length of the CDR. Separate libraries of mutations were created for each H and L chain CDR and contained predominantly single, double, and triple mutations (21). For this study, we probed the CDR libraries separately to determine how mutations in each CDR of each chain affected binding of 757. If a CDR were particularly important for anti-Id binding, we expected most mutations to be detrimental so that a large proportion of the library would not bind to 757. In contrast, for CDRs that did not contribute significantly to binding, mutations in the CDR should have relatively little effect on binding to the anti-Id.

Each of the six CDR libraries was examined in three to four assays (1000–2000 plaques per assay) with an anti-human κ reagent and 757. The number of clones examined represents only a fraction of the total membership of each library, but it does provide a measure for the interaction of each CDR with the anti-Id. We found that mutations in light chain CDRs did not reduce the number of clones binding the anti-Id compared with the number binding the anti-human κ reagent. The percentage of plaques in the three light chain libraries, L1, L2, and L3, that bound the anti-Id was slightly greater than the percentage that bound to the anti-human κ reagent (Table I). The higher percentage resulted, in part, because of a qualitative difference between the staining patterns provided by each reagent. The intensity of the staining reaction varied markedly for the plaque lifts probed with 757, which decreased the accuracy of enumeration. This variation suggests that light chain mutations may have affected the expression of correctly assembled Fab. In addition, mutations introducing stop codons in L3 eliminate the expression of Cκ, but may not affect assembly of the light chain V region with the heavy chain.

Table I.

Percent clones in BR96 CDR libraries binding anti-Id 757

Library% κ% 757
L1 66 81 
L2 70 86 
L3 57 81 
H1 95 
H2 94 
H3 90 25 
Library% κ% 757
L1 66 81 
L2 70 86 
L3 57 81 
H1 95 
H2 94 
H3 90 25 

The heavy chain CDRs were more critical to anti-Id binding (Table I). Binding of 757 to the H1 and H2 libraries was almost entirely lost, with only 1 to 2% of plaques providing a positive signal. The H3 library had a greater percentage of plaques (25%) binding the anti-Id than observed with H1 and H2 (Table I).

Large libraries permit only a statistical view of how a CDR contributes to binding, since all mutations cannot be assessed. For the affinity maturation study of BR96 (21), we also prepared codon-based mutagenesis libraries for each individual residue within the three heavy chain CDRs. The smaller size of these libraries, 32 members representing all 20 amino acids at a given residue, allows complete screening of the library multiple times to ensure that all members have been examined. Because the inferences we draw from mAb reactivity for each library are statistical, it is important that, on the whole, the libraries contain the complete complement of 32 members.

To best demonstrate a library has the diversity of membership is to sequence clones and show that all of the mutagenic codons were incorporated. The purpose of this study was to compare and map residues that are critical to binding to Ag or anti-Id. Even though the library membership is only 32, to sequence the requisite number of clones to ensure all codons are represented is a very laborious effort that undermines the essential simplicity of this method, which we view as its major advantage. Showing with 90% confidence that all codons are present in one library requires sequencing 73 clones, and 95% confidence requires sequencing 95 clones per library (40, 41, 42). Rather than sequence more than 2000 clones, we eliminated libraries from our study that did not perform as expected (see Materials and Methods). We examined one library in detail by sequencing clones of the position 54 library. Forty-three clones were sequenced and we found 17 of 32 codons representing 14 of the 20 amino acids and the stop codon (TGA). This analysis represents 75% confidence that all residues are represented and is consistent with our statistical expectations.

From our experience with the libraries, we believe that they represent the theoretical membership. Growth, infection, and immunochemical characteristics serve as indicators that libraries are constructed correctly and have all or nearly all mutant codons represented. We performed plaque lifts with reagents that would bind to each of the C regions to show that V region genes had been inserted in frame at expected frequencies, and we were careful not to overinterpret the data quantitatively.

Each library was probed with sLey-HRP and the anti-Id. We performed two to three plaque lift assays with 300 to 500 plaques per lift per CDR residue library. In addition, we probed each library with an anti-human κ reagent. Because clones making only light chain are detectable with the anti-human κ antisera, we included a second control probe, mAb 52.4 (34), that recognizes a decapeptide tag placed at the carboxyl end of CH1. Positive reactions with 52.4 indicate that the clones are expressing assembled Fab. Since the mutations are in the heavy chain CDRs, mutations that either introduce a stop codon in one of the CDRs or prevent appropriate assembly will not be detected by 52.4, and consequently are not included in the experiment. Therefore, the percentage of positive plaques for sLey or 757 was calculated as follows:

\[\frac{(No.\ sLe\ ^{y}\ or\ No.\ 757\ positive\ plaques)}{No.\ 52.4\ positive\ plaques}\ {\times}\ 100\]

As suggested from data with the larger, more complex CDR libraries, when residues in H1 and H2 were substituted, binding to the anti-Id was most affected. Less than 20% of the clones in CDR H1 libraries, residues 26, 28, and 31, and in H2 CDR libraries, residues 52a, 53, 54, and 55, bound 757 (Fig. 3). Of these residues, H2 positions 52a, 54, and to a lesser extent 55 were also nonpermissive to mutation for binding to sLey. In addition, mutations of H1 residue 29 and H2 residues, 50 and 52, resulted in a loss of sLey binding. Binding of sLey was more affected than anti-Id binding by mutations in H3, primarily at residues 95 and 96, which had only about 30% of the clones binding to Ag.

FIGURE 3.

Percentage of clones in single residue heavy chain BR96 CDR libraries binding to 757 and sLey. Libraries containing all amino acid substitutions at single BR96 heavy chain CDR residues were prepared in an M13 filamentous phage vector. The libraries were analyzed by plaque lift assay to show the percentage of positive for binding to anti-Id 757 (▦) and to sLey (▪). The percentage represents the number of clones binding to each ligand compared with the number of clones expressing assembled Fab (see Results).

FIGURE 3.

Percentage of clones in single residue heavy chain BR96 CDR libraries binding to 757 and sLey. Libraries containing all amino acid substitutions at single BR96 heavy chain CDR residues were prepared in an M13 filamentous phage vector. The libraries were analyzed by plaque lift assay to show the percentage of positive for binding to anti-Id 757 (▦) and to sLey (▪). The percentage represents the number of clones binding to each ligand compared with the number of clones expressing assembled Fab (see Results).

Close modal

To validate our scanning mutagenesis approach as a method to study the relationship between the binding sites for Ag and an anti-Id, we compared residues least permissive for anti-Id and for sLey binding on the BR96 x-ray structure (Fig. 4). Residues were considered least permissive to change if less than 50% of the tested mutant clones bound to either ligand. The Ag binding site in BR96 has the shape of a deep pocket (22). The residue positions that most affected Ag binding, Tyr33, Tyr50, Gly95, and Leu96 (coded red in Fig. 4), are located deepest within the pocket of the Ag binding site. Those that impacted binding of the anti-Id only, Gly26, Phe27, and Gly53 (coded gold in Fig. 4), are located at the periphery of the pocket. Residues Phe29, Asp31, Tyr32, Ser52, Gln52A, Gly54, and Asp55 affected binding to both ligands (coded purple in Fig. 4). Gly26, Phe27, Phe29, which are critical determinants of the H1 canonical loop conformation, and Gly54, critical to the H2 loop conformation (29), were important to anti-Id binding. Mutations at two determinants, Phe29 and Gly54, were detrimental to Ag binding as well. Residues Asp31 and Gln52A are located at the interface between H1 and H2 and are important for anti-Id binding. In summary, mutations of residues in the H1 and H2 loops had greater impact on anti-Id binding, and those in loops H2 and H3 had greater effect on sLey binding.

FIGURE 4.

Mapping of BR96 heavy chain residues important for sLey and/or anti-Id binding. Residues were analyzed on the basis of the chiBR96 Fab-nLey crystal structure (pdb code “1CLY”) (22). In A, the complex is shown in space-filling representation. For clarity, the BR96 C regions are omitted. The variable heavy and light chains are colored light blue and dark blue, respectively. The carbohydrate ligand nLey is shown in green. Labels indicate the approximate positions of heavy chain CDR loops H1, H2, and H3. Residues for which no more than 50% of the tested mutant clones displayed binding are color coded (red, important for sLey binding only; gold, important for anti-Id binding only; purple, important for both sLey and anti-Id binding). In B, these residues are shown relative to the position of nLey.

FIGURE 4.

Mapping of BR96 heavy chain residues important for sLey and/or anti-Id binding. Residues were analyzed on the basis of the chiBR96 Fab-nLey crystal structure (pdb code “1CLY”) (22). In A, the complex is shown in space-filling representation. For clarity, the BR96 C regions are omitted. The variable heavy and light chains are colored light blue and dark blue, respectively. The carbohydrate ligand nLey is shown in green. Labels indicate the approximate positions of heavy chain CDR loops H1, H2, and H3. Residues for which no more than 50% of the tested mutant clones displayed binding are color coded (red, important for sLey binding only; gold, important for anti-Id binding only; purple, important for both sLey and anti-Id binding). In B, these residues are shown relative to the position of nLey.

Close modal

The results of our study and those of Evans et al. (9) suggest that it is unlikely that an anti-Id to an anti-carbohydrate mAb would mimic Ag. Evans and colleagues solved the crystal structures for an anti-Brucella abortus LPS Fab complexed to Ag and to its anti-Id Fab. They found that adequate contact was not made by the anti-Id in the deep and narrow anti-LPS binding site. BR96 represents a similar example.

The best example to date of an anti-Id that, to some extent, structurally mimics Ag is the anti-Id E5.2 to the anti-hen egg-white lysozyme mAb D1.3 (10). Furthermore, E5.2 induced an immune response in animals that included Abs to Ag (10). The complexity of the structure-function relationship between D1.3 and its ligands was illustrated in an alanine-scanning mutagenesis study (43). The contributing bond energies of the 13 amino acids common to binding of both the anti-Id E5.2 and hen egg white lysozyme Ag were measured (43). Dall’Acqua and colleagues (43) found that the functional epitopes (44) were formed by a subset of structural contact residues. Mapping by binding provides a different perspective than does a strictly structural analysis of Ag or anti-Id contact residues.

In this study, we used codon-based scanning mutagensis to identify BR96 heavy chain CDR residues important for binding to Ag and anti-Id. Screening the BR96 CDR single residue libraries by binding indicated that the binding sites for sLey and anti-Id partially overlap. Residues least permissive to mutation for anti-Id binding were found in CDRs H1 and H2, while those nonpermissive for Ag binding were found in CDRs H2 and H3. Residues important for Ag binding map to the deep groove that forms the sLey binding site (22). In contrast, residues important for anti-Id binding are more exposed and accessible.

Mapping of BR96 residues nonpermissive to change illustrated that mutations disrupted both direct interactions, that is, contact with Ag, and indirect interactions. The four residues important only for Ag binding, Tyr33, Tyr50, Gly95, and Leu96, all contact nLey in the x-ray structure (22). Both tyrosine residues, 33 and 50, participate in the formation of the deep pocket binding site of BR96 (22). Indirect interactions include loop conformations, the interface between loops, and more distant interactions in which a change in one residue affected another’s contact with Ag. Residues Phe29 and Gly54 are critical determinants of the canonical loop conformations (29) of H1 and H2, respectively. Mutations of these residues are expected to significantly alter these loop conformations, and they did affect both sLey and anti-Id binding. Residues Gly26 and Phe27, also determinants of the H1 conformation, significantly reduced only anti-Id and not sLey binding. The importance of the H1 and H2 loops to anti-Id binding was also demonstrated by the sensitivity to changes at residues located at the interface of the two loops (Asp31 and Gln52A). Residues more distant from the Ag binding site that affected anti-Id as well as Ag binding included Asp55, which is greater than 10 Å of nLey. A second residue, Ser52, packs against Tyr33, which does closely contact the bound Ag.

Results of this study demonstrated the importance of individual residues to binding of chiBR96 to its ligands. We observed that when either residue Gln52A or Gly54 in the H2 loop was mutated, binding to both sLey and anti-Id was lost. Mutations of residue Gly53 in the H2 loop had quite different effects on binding to each of the ligands. Approximately 10% of clones in the H53 library bound to the anti-Id. In contrast, virtually all mutations at this position had no effect on Ag binding. Furthermore, at position 53, a change from glycine to aspartic acid resulted in an increase in affinity to tumor Ag (21). We noted then that this amino acid change caused complete loss of anti-Id binding (unpublished observation).

Evaluating single residue libraries can be accomplished more rapidly than examining larger libraries, and the results may present a clearer indication of which residues should be the focus of further mutagenesis. Furthermore, we gain a more comprehensive view of the interactions between a mAb and its ligands by substituting residues with all possible amino acids, rather than a single amino acid as for alanine scanning. While a mutation of a single residue may eliminate all binding activity to a ligand, multiple mutations in CDRs may be necessary to enhance binding. We and others (20, 45) have reported that mutations improving Ab affinity for Ag are frequently additive. Single residue libraries offer a distinct advantage compared with larger, multimillion member libraries, because the smaller membership can be completely screened. After screening single residue libraries to identify mutations that affect binding, the individual mutations can often be combined into a molecule and the function reassessed.

Just as one might alter the binding or specificity of a mAb, the strategy described in this study could also be applied to engineering the properties of an anti-Id. One application would be to distinguish an anti-Id with a desired property among a panel of anti-Ids. Similarity in the profile of mAb residues affecting binding of an anti-Id and Ag would indicate which anti-Id would most likely elicit an immune response similar to Ag in vivo. Those with overlapping or independent binding sites would be more useful for analytic reagents for clinical trials, particularly if free Ag exists in the blood. Understanding the contribution of CDR residues to ligand binding in detail is often critical to attempts to improve Ab function. The combination of structural analysis by crystallographic approaches and functional analysis by molecular genetics provides a comprehensive body of information to guide Ab engineering.

We thank Dr. David Stock for his analysis of statistical methods to estimate the complexity of bacteriophage libraries.

4

Current address: MDS Panlabs, Bothell, WA 98011.

5

Current address: Amylin Pharmaceuticals, San Diego, CA 92121.

1
Jerne, N. K..
1974
. Towards a network theory of the immune response.
Ann. Immunol. (Paris)
125C
:
373
2
McNamara, M. K., R. E. Ward, H. Kohler.
1985
. Monoclonal idiotype vaccine against Streptococcus pneumoniae infection.
Science
226
:
1325
3
Gaulton, G. N., A. H. Sharpe, D. W. Chang, B. N. Fields, M. I. Green.
1986
. Syngeneic monoclonal internal image antiidiotypes as prophylactic vaccines.
J. Immunol.
137
:
2930
4
Raychandhuri, S., Y. Saeki, J. J. Chen, H. Iribe, H. Fuji, H. Kohler.
1987
. Tumor-specific idiotype vaccine.
J. Immunol.
139
:
271
5
Buattacharya-Chatterjee, M., S. K. Chatterjee, S. Vasile, B. K. Seon, H. Kohler.
1988
. Idiotype vaccines against human T cell leukemia.
J. Immunol.
141
:
1398
6
Schreiber, J. R., K. L. Nixon, M. F. Tosi, G. B. Pier, M. B. Patawaran.
1991
. Antiidiotype-induced, lipopolysaccharide-specific antibody response to Pseudomonas aeruginosa. II. Isotype and functional activity of the anti-idiotype-induced antibodies.
J. Immunol.
146
:
188
7
Su, S., M. N. Ward, M. A. Apicella, R. E. Ward.
1992
. A nontoxic, idiotope vaccine against Gram-negative bacterial infections.
J. Immunol.
148
:
234
8
Bentley, G. A., G. Boulot, M. M. Riottot, R. J. Poljak.
1990
. Three-dimensional structure of an idiotope-anti-idiotope complex.
Nature
348
:
254
9
Evans, S. V., D. R. Rose, R. To, N. M. Young, D. R. Bundle.
1994
. Exploring the mimicry of polysaccharide antigens by anti-idiotypic antibodies: the crystallization, molecular replacement, and refinement to 2.8 Å resolution of an idiotope-anti-idiotope Fab complex and of the unliganded anti-idiotope Fab.
J. Mol. Biol.
241
:
691
10
Fields, B. A., F. A. Goldbaum, X. Ysern, R. J. Poljak, R. A. Mariuzza.
1995
. Molecular basis of antigen mimicry of an anti-idiotope.
Nature
374
:
739
11
Ban, N., C. Escobar, R. Garcia, K. Hasel, J. Day, A. Greenwood, A. McPherson.
1994
. Crystal structure of an idiotype-anti-idiotype Fab complex.
Proc. Natl. Acad. Sci. USA
91
:
1604
12
Braden, B. C., B. A. Fields, X. Ysern, W. Dall’Acqua, F. A. Goldbaum, R. J. Poljak, R. A. Mariuzza.
1996
. Crystal structure of an Fv-Fv idiotope complex at 1.9 Å resolution.
J. Mol. Biol.
264
:
137
13
Boss, M. A., J. H. Kenten, C. R. Wood, J. S. Emtage.
1984
. Assembly for functional antibodies from immunoglobulin heavy and light chains synthesized in E. coli.
Nucleic Acids Res.
12
:
3791
14
Cabilly, S., A. D. Riggs, H. Ande, J. E. Shively, W. E. Holmes, M. Rey, L. J. Perry, R. Wetzel, H. L. Heyneker.
1984
. Generation of antibody activity from immunoglobulin chains produced in Escherichia coli.
Proc. Natl. Acad. Sci. USA
81
:
3273
15
Skerra, A., A. Plückthun.
1988
. Assembly of a functional immunoglobulin Fv fragment in Escherichia coli.
Science
240
:
1038
16
Better, M., C. P. Chang, R. R. Robinson, A. H. Horwitz.
1988
. Escherichia coli secretion of an active chimeric antibody fragment.
Science
240
:
1041
17
Ward, E. S., D. Güssow, A. D. Griffiths, P. T. Jones, G. Winter.
1989
. Binding activities of a repertoire of single immunoglobulin variable domains secreted from Escherichia coli.
Nature
341
:
544
18
Hellström, I., H. J. Garrigues, U. Garrigues, K. E. Hellström.
1990
. Highly tumor-reactive, internalizing mouse monoclonal antibodies to Ley-related cell surface antigen.
Cancer Res.
50
:
2183
19
Nakasaki, H., T. Mitomi, T. Noto, K. Ogoshi, H. Hanaue, Y. Tanaka, H. Mkuuchi, H. Clausen, S. Hakomori.
1989
. Mosaicism in the expression of tumor-associated carbohydrate antigens in human colonic and gastric cancers.
Cancer Res.
49
:
3662
20
Fell, H. P., S. Yarnold, I. Hellström, K. E. Hellström, K. R. Folger.
1989
. Homologous recombination in hybridoma cells: heavy chain chimeric antibody produced by gene targeting.
Proc. Natl. Acad. Sci. USA
86
:
8507
21
Yelton, D. E., M. J. Rosok, G. Cruz, W. L. Cosand, J. Bajorath, I. Hellström, K. E. Hellström, W. D. Huse, S. M. Glaser.
1995
. Affinity maturation of the BR96 anti-carcinoma antibody by codon-based mutagenesis.
J. Immunol.
155
:
1994
22
Jeffrey, P. D., J. Bajorath, C. Y. Chang, D. Yelton, I. Hellström, K. E. Hellström, S. Sheriff.
1995
. The X-ray structure of an anti-tumour antibody in complex with antigen.
Nat. Struct. Biol.
2
:
466
23
Schreiber, G. J., K. E. Hellström, I. Hellström.
1992
. An unmodified anticarcinoma antibody, BR96, localizes to and inhibits the outgrowth of human tumors in nude mice.
Cancer Res.
52
:
3262
24
Yeh, M. Y., I. Hellström, J. P. Brown, G. A. Warner, J. A. Hansen, K. E. Hellström.
1979
. Cell surface antigens of human melanoma identified by monoclonal antibody.
Proc. Natl. Acad. Sci. USA
76
:
2927
25
Stenzel-Johnson, P., D. Yelton, J. Bajorath.
1994
. Identification of residues in the monoclonal anti-tumor antibody L6 important for binding to its tumor antigen.
Biochemistry
33
:
14400
26
Towbin, H., T. Staehelin, J. Gordon.
1979
. Electrophoretic transfer of proteins from polyacrylamide gels to nitrocellulose sheets: procedure and some applications.
Proc. Natl. Acad. Sci. USA
76
:
4350
27
Rosok, M. J., M. R. Stebbins, K. Connelly, M. E. Lostrom, A. W. Siadak.
1990
. Generation and characterization of murine antiflagellum monoclonal antibodies that are protective against lethal challenge with Pseudomonas aeruginosa.
Infect. Immun.
58
:
3819
28
Bajorath, J..
1994
. Three-dimensional model of the BR96 monoclonal antibody variable fragment.
Bioconjug. Chem.
5
:
213
29
Chothia, C., A. M. Lesk, A. Tramontano, M. Levitt, S. J. Smith-Gill, G. Air, S. Sheriff, E. A. Padlan, D. Davies, W. R. Tulip, P. M. Colman, S. Spinelli, P. M. Alzari, R. J. Poljak.
1989
. Conformations of immunoglobulin hypervariable regions.
Nature
342
:
877
30
Kabat, E. A., T. T. Wu, M. Reid-Miller, H. M. Perry, K. S. Gottesman.
1987
.
Sequences of Proteins of Immunological Interest
U.S. Department of Health and Human Services, Public Health Service, National Institutes of Health, Bethesda, MD.
31
Rosok, M. J., D. E. Yelton, L. J. Harris, J. Bajorath, K. E. Hellström, I. Hellström, G. A. Cruz, K. Kristensson, H. Lin, W. D. Huse, S. M. Glaser.
1996
. A combinatorial library strategy for the rapid humanization of anticarcinoma BR96 Fab.
J. Biol. Chem.
271
:
22611
32
Kunkel, T. A., J. D. Roberts, R. A. Zakour.
1987
. Rapid and efficient site-specific mutagenesis without phenotypic selection.
Methods Enzymol.
154
:
367
33
Huse, W. D., T. J. Stinchcombe, S. M. Glaser, L. Starr, M. MacLean, K. E. Hellström, I. Hellström, D. E. Yelton.
1992
. Application of a filamentous phage pVIII fusion protein system suitable for efficient production, screening, and mutagenesis of Fab antibody fragments.
J. Immunol.
149
:
3914
34
Gilliland, L. K., N. A. Norris, H. Marquardt, T. T. Tsu, M. S. Hayden, M. G. Neubauer, D. E. Yelton, R. S. Mittler, J. A. Ledbetter.
1996
. Rapid and reliable cloning of antibody variable regions and generation of recombinant single chain antibody fragments.
Tissue Antigens
47
:
1
35
Leary, J. J., D. J. Brigati, D. C. Ward.
1983
. Rapid and sensitive colorimetric method for visualizing biotin-labeled DNA probes hybridized to DNA or RNA immobilized on nitrocellulose: bio-blots.
Proc. Natl. Acad. Sci. USA
80
:
4045
36
Hellström, K. E., D. E. Yelton, H. P. Fell, D. Beaton, M. Gayle, M. MacLean, M. Kahn, I. Hellström.
1990
. Epitope mapping and use of anti-idiotypic antibodies to the L6 monoclonal anticarcinoma antibody.
Cancer Res.
50
:
2449
37
Poljak, R. J..
1991
. Structure of antibodies and their complexes with antigens.
Mol. Immunol.
28
:
1341
38
Mian, I. S., A. R. Bradwell, A. J. Olson.
1991
. Structure, function and properties of antibody binding sites.
J. Mol. Biol.
217
:
133
39
Glaser, S. M., M. Vasquez, P. W. Payne, W. P. Schneider.
1992
. Dissection of the combining site in a humanized anti-tac antibody.
J. Immunol.
149
:
2607
40
Leontin, R. C., T. Prout.
1956
. Estimation of the number of different classes in a population.
Biometrics
12
:
211
41
Chao, A., S. Lee.
1992
. Estimating the number of classes via sample coverage.
J. Am. Stat. Assoc.
87
:
210
42
Bunge, J., M. Fitzpatrick.
1993
. Estimating the number of species: a review.
J. Am. Stat. Assoc.
88
:
364
43
Dall’Acqua, W., E. R. Goldman, E. Eisenstein, R. A. Mariuzza.
1996
. A mutational analysis of the binding of two different proteins to the same antibody.
Biochemistry
35
:
9667
44
Cunningham, B. C., J. A. Wells.
1989
. High-resolution epitope mapping of hGH-receptor interactions by alanine-scanning mutagenesis.
Science
244
:
1081
45
Sharon, J..
1990
. Structural correlates of high antibody affinity: three engineered amino acid substitutions can increase the affinity of an anti-ρ-azophenylarsonate antibody 200-fold.
Proc. Natl. Acad. Sci. USA
87
:
4814