VL:VH domain rotations in engineered antibodies: Crystal structures of the Fab fragments from two murine antitumor antibodies and their engineered human constructs

The crystal structures of two pairs of Fab fragments have been determined. The pairs comprise both a murine and an engineered human form, each derived from the antitumor antibodies A5B7 and CTM01. Although antigen specificity is maintained within the pairs, antigen affinity varies. A comparison of the hypervariable loops for each pair of antibodies shows their structure has been well maintained in grafting, supporting the canonical loop model. Detailed structural analysis of the binding sites and domain arrangements for these antibodies suggests the differences in antigen affinity observed are likely to be due to inherent flexibility of the hypervariable loops and movements at the V_L:V_H domain interface. The four structures provide the first opportunity to study in detail the effects of protein engineering on specific antibodies. Proteins 29:161–171, 1997. © 1997 Wiley-Liss, Inc.     

Structural repertoire of the human VH segments.

The VH gene segments produce the part of the VH domains of antibodies that contains the first two hypervariable regions. The sequences of 83 human VH segments with open reading frames, from several individuals, are currently known. It has been shown that these sequences are likely to form a high proportion of the total human repertoire and that an individual's gene repertoire produces about 50 VH segments with different protein sequences. In this paper we present a structural analysis of the amino acid sequences produced by the 83 segments. Particular residue patterns in the sequences of V domains imply particular main-chain conformations, canonical structures, for the hypervariable regions. We show that, in almost all cases, the residue patterns in the VH segments imply that the first hypervariable regions have one of three different canonical structures and that the second hypervariable regions have one of five different canonical structures. The different observed combinations of the canonical structures in the first and second regions means that almost all sequences have one of seven main-chain folds. We describe, in outline, structures of the antigen binding site loops produced by nearly all the VH segments. The exact specificity of the loops is produced by (1) sequence differences in their surface residues, particularly at sites near the centre of the combining site, and (2) sequence differences in the hypervariable and framework regions that modulate the relative positions of the loops.     

Canonical antigen-binding loop structures in immunoglobulins: more structures, more canonical classes?

Grafting the antigen-binding loops onto a human antibody scaffold is a widely used technique to humanise murine antibodies. The success of this approach depends largely on the observation that the antigen-binding loops adopt only a limited number of canonical structures. Identification of the correct canonical structure is therefore essential. Algorithms that predict the main-chain conformation of the hypervariable loops using only the amino acid sequence often provide this information. Here, we describe new canonical loop conformations for the hypervariable regions H1 and H2 as found in single-domain antibody fragments of dromedaries or llama. Although the occurrence of these new loop conformations was not predicted by the algorithms used, it seems that they could occur in human or mouse antigen-binding loops. Their discovery indicates that the currently used set of canonical structures is incomplete and that the prediction algorithms should be extended to include these new structures.     

Ab initio structure prediction of the antibody hypervariable H3 loop

Antibodies have the capability of binding a wide range of antigens due to the diversity of the six loops constituting the complementarity determining region (CDR). Among the six loops, the H3 loop is the most diverse in structure, length, and sequence identity. Prediction of the three‐dimensional structures of antibodies, especially the CDR loops, is an important step in the computational design and engineering of novel antibodies for improved affinity and specificity. Although it has been demonstrated that the conformation of the five non‐H3 loops can be accurately predicted by comparing their sequences against databases of canonical loop conformations, no such connection has been established for H3 loops. In this work, we present the results for ab initio structure prediction of the H3 loop using conformational sampling and energy calculations with the program Prime on a dataset of 53 loops ranging in length from 4 to 22 residues. When the prediction is performed in the crystal environment and including symmetry mates, the median backbone root mean square deviation (RMSD) is 0.5 Å to the crystal structure, with 91% of cases having an RMSD of less than 2.0 Å. When the prediction is performed in a noncrystallographic environment, where the scaffold is constructed by swapping the H3 loops between homologous antibodies, 70% of cases have an RMSD below 2.0 Å. These results show promise for ab initio loop predictions applied to modeling of antibodies. © 2012 Wiley Periodicals, Inc.     

Modeling the antigen combining site of an anti-dinitrophenyl antibody, ANO2.

A model structure has been constructed for a monoclonal anti-dinitrophenyl antibody. The antibody, ANO2, has been sequenced and cloned (Anglister, J., Frey, T., & McConnell, H.M., 1984, Biochemistry 23, 1138-1142). Its amino acid sequence shows striking homology with the anti-lysozyme Fab fragments HyHel5 (83%) and HyHel10 (73%). Based on this homology, a model for the ANO2 variable heavy and variable light chain framework was constructed using a hybrid of the HyHel5 light chain and the HyHel10 heavy chain backbone, omitting the hypervariable loops. These coordinates were used as scaffolds for the model building of ANO2. The CONGEN conformational sampling algorithm (Bruccoleri, R.E. & Karplus, M., 1987, Biopolymers 26, 127-196) was used to model the six hypervariable loops that contain the antigen-combining site. All the possible conformations of the loop backbones were constructed and the best loop structures were selected using a combination of the CHARMM potential energy function and evaluation of the solvent-accessible surface area of the conformers. The order in which the loops were searched was carried out based on the relative locations of the loops with reference to the framework of the beta-barrel, namely, L2-H1-L3-H2-H3-L1. The model structures thus obtained were compared to the high resolution X-ray structure (Brünger, A.T., Leahy, D.J., Hynes, T.R., & Fox, R.O., 1991, J. Mol. Biol. 221, 239-256).     

Cloning, characterization, and modeling of a monoclonal anti-human transferrin antibody that competes with the transferrin receptor.

In this report we describe the isolation and characterization of a monoclonal antibody against human serum transferrin (Tf) and the cloning and sequencing of its cDNA. The antibody competes with the transferrin receptor (TR) for binding to human Tf and is therefore expected to bind at or very close to a region of interaction between Tf and its receptor. From the deduced amino acid sequence, we constructed a 3-dimensional model of the variable domains of the antibody based on the canonical structure model for the hypervariable loops. The proposed structure of the antibody is a first step toward a more detailed characterization of the antibody-Tf complex and possibly toward a better understanding of the Tf interaction with its receptor. The model might prove useful in guiding site-directed mutagenesis studies, simplifying the experimental elucidation of the antibody structure, and in the use of automatic procedures to dock the interacting molecules as soon as structural information about the structure of the human Tf molecule will be available.     

Antibody modeling assessment

A blinded study to assess the state of the art in three‐dimensional structure modeling of the variable region (Fv) of antibodies was conducted. Nine unpublished high‐resolution x‐ray Fab crystal structures covering a wide range of antigen‐binding site conformations were used as benchmark to compare Fv models generated by four structure prediction methodologies. The methodologies included two homology modeling strategies independently developed by CCG (Chemical Computer Group) and Accerlys Inc, and two fully automated antibody modeling servers: PIGS (Prediction of ImmunoGlobulin Structure), based on the canonical structure model, and Rosetta Antibody Modeling, based on homology modeling and Rosetta structure prediction methodology. The benchmark structure sequences were submitted to Accelrys and CCG and a set of models for each of the nine antibody structures were generated. PIGS and Rosetta models were obtained using the default parameters of the servers. In most cases, we found good agreement between the models and x‐ray structures. The average rmsd (root mean square deviation) values calculated over the backbone atoms between the models and structures were fairly consistent, around 1.2 Å. Average rmsd values of the framework and hypervariable loops with canonical structures (L1, L2, L3, H1, and H2) were close to 1.0 Å. H3 prediction yielded rmsd values around 3.0 Å for most of the models. Quality assessment of the models and the relative strengths and weaknesses of the methods are discussed. We hope this initiative will serve as a model of scientific partnership and look forward to future antibody modeling assessments. Proteins 2011; © 2011 Wiley‐Liss, Inc.     

A single-domain antibody fragment in complex with RNase A: non-canonical loop structures and nanomolar affinity using two CDR loops

BACKGROUND: Camelid serum contains a large fraction of functional heavy-chain antibodies - homodimers of heavy chains without light chains. The variable domains of these heavy-chain antibodies (VHH) have a long complementarity determining region 3 (CDR3) loop that compensates for the absence of the antigen-binding loops of the variable light chains (VL). In the case of the VHH fragment cAb-Lys3, part of the 24 amino acid long CDR3 loop protrudes from the antigen-binding surface and inserts into the active-site cleft of its antigen, rendering cAb-Lys3 a competitive enzyme inhibitor. RESULTS: A dromedary VHH with specificity for bovine RNase A, cAb-RN05, has a short CDR3 loop of 12 amino acids and is not a competitive enzyme inhibitor. The structure of the cAb-RN05-RNase A complex has been solved at 2.8 A. The VHH scaffold architecture is close to that of a human VH (variable heavy chain). The structure of the antigen-binding hypervariable 1 loop (H1) of both cAb-RN05 and cAb-Lys3 differ from the known canonical structures; in addition these H1 loops resemble each other. The CDR3 provides an antigen-binding surface and shields the face of the domain that interacts with VL in conventional antibodies. CONCLUSIONS: VHHs adopt the common immunoglobulin fold of variable domains, but the antigen-binding loops deviate from the predicted canonical structure. We define a new canonical structure for the H1 loop of immunoglobulins, with cAb-RN05 and cAb-Lys3 as reference structures. This new loop structure might also occur in human or mouse VH domains. Surprisingly, only two loops are involved in antigen recognition; the CDR2 does not participate. Nevertheless, the antigen binding occurs with nanomolar affinities because of a preferential usage of mainchain atoms for antigen interaction.     

Systematic classification of CDR-L3 in antibodies: implications of the light chain subtypes and the VL-VH interface

Antibody modeling is widely used for the analysis of antibody-antigen interactions and for the design of potent antibody drugs. The antibody combining site is composed of six complementarity determining regions (CDRs). The CDRs, except for CDR-H3, which is the most diverse CDR, form limited numbers of canonical structures, which can be identified from the amino acid sequences. A method to classify the CDR-H3 structure from its amino acid sequence was previously proposed. However, since those CDR structures were classified, many more antibody crystal structures have been determined. We performed systematic analyses of the CDR-L3 structures and found novel canonical structures, and we also classified a previously identified canonical structure into two subtypes. In addition, two differently defined canonical structures in the kappa and lambda subtypes were classified into the same canonical structure. We also identified a key residue in CDR-L3, which determines the conformation of CDR-H3. Several analyses of CDR-L3 loops longer than nine residues were performed. These new findings should be useful for structural modeling and are eventually expected to accelerate the design of antibody drugs.     

Synthetic antibody libraries focused towards peptide ligands

Synthetic antibody libraries have proven immensely useful for the de novo isolation of antibodies without the need for animal immunization. Recently, focused libraries designed to recognize particular classes of ligands, such as haptens or proteins, have been employed to facilitate the selection of high-affinity antibodies. Focused libraries are built using V regions encoding combinations of canonical structures that resemble the structural features of antibodies that bind the desired class of ligands and sequence diversity is introduced at residues typically involved in recognition. Here we describe the generation and experimental validation of two different single-chain antibody variable fragment libraries that efficiently generate binders to peptides, a class of molecules that has proven to be a difficult target for antibody generation. First, a human anti-peptide library was constructed by diversifying a scaffold: the human variable heavy chain (V_H) germ line gene 3-23, which was fused to a variant of the human variable light chain (V_L) germ line gene A27, in which L1 was modified to encode the canonical structure found in anti-peptide antibodies. The sequence diversity was introduced into 3-23 (V_H) only, targeting for diversification residues commonly found in contact with protein and peptide antigens. Second, a murine library was generated using the antibody 26-10, which was initially isolated based on its affinity to the hapten digoxin, but also binds peptides and exhibits a canonical structure pattern typical of anti-peptide antibodies. Diversity was introduced in the V_H only using the profile of amino acids found at positions that frequently contact peptide antigens. Both libraries yielded binders to two model peptides, angiotensin and neuropeptide Y, following screening by solution phage panning. The mouse library yielded antibodies with affinities below 20 nM to both targets, although only the V_H had been subjected to diversification.     

Thermodynamics of three-way multibranch loops in RNA

RNA multibranch loops (junctions) are loops from which three or more helices exit. They are nearly ubiquitous in RNA secondary structures determined by comparative sequence analysis. In this study, systems in which two strands combine to form three-way junctions were used to measure the stabilities of RNA multibranch loops by UV optical melting and isothermal titration calorimetry (ITC). These data were used to calculate the free energy increment for initiation of a three-way junction on the basis of a nearest neighbor model for secondary structure stability. Imino proton NMR spectra were also measured for two systems and are consistent with the hypothesized helical structures. Incorporation of the experimental data into the mfold and RNA structure computer programs has contributed to an improvement in prediction of RNA secondary structure from sequence.     

Functional consequences of insertions and deletions in the complementarity-determining regions of human antibodies

Insertions and deletions of nucleotides in the genes encoding the variable domains of antibodies are natural components of the hypermutation process, which may expand the available repertoire of hypervariable loop lengths and conformations. Although insertion of amino acids has also been utilized in antibody engineering, little is known about the functional consequences of such modifications. To investigate this further, we have introduced single-codon insertions and deletions as well as more complex modifications in the complementarity-determining regions of human antibody fragments with different specificities. Our results demonstrate that single amino acid insertions and deletions are generally well tolerated and permit production of stably folded proteins, often with retained antigen recognition, despite the fact that the thus modified loops carry amino acids that are disallowed at key residue positions in canonical loops of the corresponding length or are of a length not associated with a known canonical structure. We have thus shown that single-codon insertions and deletions can efficiently be utilized to expand structure and sequence space of the antigen-binding site beyond what is encoded by the germline gene repertoire.     

The (1)H-NMR solution structure of the antitryptic core peptide of Bowman-Birk inhibitor proteins: a minimal canonical loop

Bowman-Birk inhibitor (BBI) proteins contain an inhibitory motif comprising a disulfide-bonded sequence that interacts with serine proteinases. Recently, a small 14-residue peptide from sunflowers (SFTI-1), which has potent anti-trypsin activity, has been found to have the same motif. However, this peptide also has an unusual head-to-tail cyclisation. To address the role of the core inhibitory sequence itself, we have solved the (1)H-NMR solution structure of an antitryptic 11-residue cyclic peptide that corresponds to the core reactive site loops of both SFTI-1 and Bowman-Birk inhibitor proteins. A comparison is made between the secondary chemical shifts found in this family and the canonical regions of several other inhibitors, giving some insight into relative flexibility and hydrogen bonding patterns in these inhibitors. The solution structure of the core peptide in isolation is found to retain essentially the same three-dimensional arrangement of both backbone and side chains as observed in larger antitryptic BBI and SFTI-1 fragments as well as in the complete proteins. The retention of the canonical conformation in the core peptide explains the peptids inhibitory potency. It therefore represents a minimization of both the BBI and SFTI-1 sequences. We conclude that the core peptide is a conformationally defined, canonical scaffold, which can serve as a minimal platform for the engineering of biological activity.     

The structural repertoire of the human V kappa domain.

In humans, the gene for the V kappa domain is produced by the recombination of one of 40 functional V kappa segments and one of five functional J kappa segments. We have analysed the sequences of these germline segments and of 736 rearranged V kappa genes to determine the repertoire of main chain conformations, or canonical structures, they encode. Over 96% of the sequences correspond to one of four canonical structures for the first antigen binding loop (L1) and one canonical structure for the second antigen binding loop (L2). Junctional diversity produces some variation in the length of the third antigen binding loop (L3) and in the identity of residues at the V kappa-J kappa join. However, this is limited and 70% of the rearranged sequences correspond to one of three known canonical structures for the L3 region. Furthermore, we show that the canonical structures selected during the primary response are conserved during affinity maturation: the key residues that determine the conformations of the antigen binding loops are unmutated or undergo conservative mutation. The implications of these results for immune recognition are discussed.     

Structure refinement of the OpcA adhesin using molecular dynamics

OpcA from Neisseria meningitidis, the causative agent of meningococcal meningitis and septicemia, is an integral outer membrane protein that facilitates meningococcal adhesion through binding the proteoglycan receptors of susceptible cells. Two structures of OpcA have been determined by x-ray diffraction to 2 A resolution, revealing dramatically different conformations in the extracellular loops--the protein domain implicated in proteoglycan binding. In the first structure, a positively charged crevice formed by loops 1 and 2 was identified as the site for binding proteoglycans, whereas in the second structure the crevice was not evident as loops 1 and 2 adopted different conformations. To reconcile these results, molecular-dynamics simulations were carried out on both structures embedded in a solvated lipid bilayer membrane. Free of crystal contacts and crystallization agents, the loops were observed to undergo large structural transformations, suggesting that the conformation of the loops in either x-ray structure is affected by crystallization. Subsequent simulations of both structures in their crystal lattices confirmed this conclusion. Based on our molecular-dynamics trajectories, we propose a model for OpcA that combines stable structural features of the available x-ray structures. In this model, all five extracellular loops of OpcA have stable secondary structures. The loops form a funnel that leads to the base of the beta-barrel and that includes Tyr-169 on its exposed surface, which has been implicated in proteoglycan binding.     

Identification of a stem-loop structure important for polyadenylation at the murine IgM secretory poly(A) site.

We have previously shown that a distal GU-rich downstream element of the mouse IgM secretory poly(A) site is important for polyadenylation in vivo and for polyadenylation specific complex formation in vitro. This element can be predicted to form a stem-loop structure with two asymmetric internal loops. As stem-loop structures commonly define protein RNA binding sites, we have probed the biological activity of the secondary structure of this element. We show that mutations affecting the stem of the structure abolish the biological activity of this element in vivo and in vitro at the level of cleavage and polyadenylation specificity factor/cleavage stimulation factor complex formation and that both internal loops contribute to the enhancing effect of the sequence in vivo. Lead (II) cleavage patterns and RNase H probing of the sequence element in vitro are consistent with the predicted secondary structure. Furthermore, mobility on native PAGE suggests a bent structure. We propose that the secondary structure of this downstream element optimizes its interaction with components of the polyadenylation complex.     

A new apo-caspase-6 crystal form reveals the active conformation of the apoenzyme

Caspase-6 has been identified as a key component in the pathway of neurodegenerative diseases such as Alzheimer's disease and Huntington's disease. It has been the focus of drug development for some time, but only recently have structural data become available. The first study identified a novel noncanonical conformation of apo-caspase-6 contrasting with the typical caspase conformation. Then, the structures of both caspase-6 zymogen and the Ac-VEID-CHO peptide inhibitor complex described caspase-6 in the canonical conformation, raising the question of why the intermediate between these two structures (mature apo-caspase-6) would adopt the noncanonical conformation. In this study, we present a new crystal form of the apoenzyme in the canonical conformation by identifying the previous apostructure as a pH-inactivated form of caspase-6. Our new apostructure is further compared to the Ac-VEID-CHO caspase-6 inhibitor complex. The structural comparison allows us to visualize the organization of loops L2, L3, and L4 upon ligand binding and how the catalytic groove forms to accommodate the inhibitor.     

Antibody-antigen recognition: A canonical structure paradigm

The antibodies of known three-dimensional structure exhibit a definite number of conformations (canonical structures) for five of six hypervariable loops. In the present study it was found that approximately 85% of the immunoglobulin sequences analyzed fall into a small number of canonical structure combinations, representing only 3% of the total possible. These structures were classified into six distinct groups, depending on the type of antigen with which they interact.Within each loop, the positions responsible for maintaining these canonical structures show a use frequency of amino acids that fits an inverse power law, whereas the use frequency of the amino acids responsible for the detailed antigenic specificity follows an exponential distribution. We propose an evolutionary interpretation that connects these data, using the fact that the inverse power law is generated by statistical processes of the type that yield a wealth curve and the fact that expoential distribution is generated by processes that are not biased by past history.