H-2Dd exploits a four residue peptide binding motif.

We have characterized the amino acid sequences of over 20 endogenous peptides bound by a soluble analog of H-2Dd, H-2Dds. Synthetic analogs corresponding to self, viral, tumor, or motif peptides were then tested for their ability to bind to H-2Dd by serologic epitope induction assays using both purified soluble protein and cell surface H-2Dd. The dominant primary sequence motif included glycine at position 2, proline at position 3, and a hydrophobic COOH terminus: leucine, isoleucine, or phenylalanine at position 9 or 10. Ancillary support for high affinity binding was contributed by a positively charged residue at position 5. Three-dimensional computer models of H-2Dds/peptide complexes, based on the crystallographic structure of the human HLA-B27/peptide complex, showed that the basic residue at position 5 was in position to form a salt bridge with aspartic acid at position 156, a polymorphic residue of the H-2Dd heavy (H) chain. Analysis of 28 such models, including 17 based on nonamer self-peptides, revealed considerable variation in the structure of the major histocompatibility complex (MHC) surrounding peptide residue 1, depending on the size and charge of the side chain. Interactions between the side chains of peptide residues 5 and 7, and 6 and 8 commonly occurred. Those peptide positions with limited sequence variability and least solvent accessibility may satisfy structural requirements for high affinity binding of the peptide to the MHC class I H chain, whereas the highly variable positions of the peptide (such as positions 4, 6, and 8) may contribute more to the T cell epitopes.

A T cell receptor V alpha domain expressed in bacteria: does it dimerize in solution?

To evaluate the potential for dimerization through a particular T cell receptor (TCR) domain, we have cloned the cDNA encoding a TCR V alpha from a hybridoma with specificity for the human immunodeficiency virus (HIV) envelope glycoprotein 120-derived peptide P18-110 (RGPGRAFVTI) bound to the murine major histocompatibility complex (MHC) class I molecule, H-2Dd. This cDNA was then expressed in a bacterial vector, and protein, as inclusion bodies, was solubilized, refolded, and purified to homogeneity. Yield of the refolded material was from 10 to 50 mg per liter of bacterial culture, the protein was soluble at concentrations as high as 25 mg/ml, and it retained a high level of reactivity with an anti-V alpha 2 monoclonal antibody. This domain was monomeric both by size exclusion gel chromatography and by sodium dodecyl sulfate polyacrylamide gel electrophoresis. Circular dichroism spectra indicated that the folded V alpha domain had secondary structure similar to that of single immunoglobulin or TCR domains, consisting largely of beta sheet. Conditions for crystallization were established, and at least two crystal geometries were observed: hexagonal bipyramids that failed to diffract beyond approximately 6 A, and orthorhombic crystals that diffracted to 2.5 A. The dimerization of the V alpha domain was investigated further by solution nuclear magnetic resonance spectroscopy, which indicated that dimeric and monomeric forms of the protein were about equally populated at a concentration of 1 mM. Thus, models of TCR-mediated T cell activation that invoke TCR dimerization must consider that some V alpha domains have little tendency to form homodimers or multimers.

Endogenous peptides of a soluble major histocompatibility complex class I molecule, H-2Lds: sequence motif, quantitative binding, and molecular modeling of the complex.

To gain insight into the rules that govern the binding of endogenous and viral peptides to a given major histocompatibility complex (MHC) class I molecule, we characterized the amino acid sequences of a set of self peptides bound by a soluble analogue of murine H-2Ld, H-2Lds. We tested corresponding synthetic peptides quantitatively for binding in several different assays, and built three-dimensional computer models of eight peptide/H-2Lds complexes, based on the crystallographic structure of the human HLA-B27/peptide complex. Comparison of primary and tertiary structures of bound self and antigenic peptides revealed that residues 2 and 9 were not only restricted in sequence and tolerant of conservative substitutions, but were spatially constrained in the three-dimensional models. The degree of sequence variability of specific residues in MHC-restricted peptides reflected the lack of structural constraint on those amino acids. Thus, amino acid residues that define a peptide motif represent side chains required or preferred for a close fit with the MHC class I heavy chain.

T-cell receptors: feeling out the complex.

Recent crystallographic studies show that the T-cell receptor has a largely immunoglobulin-like structure and binds to MHC-peptide complexes through loops from paired Valpha and Vbeta domains that focus on the central amino acids of the MHC-bound peptide, and to bacterial superantigens via peripheral aspects of the Vbeta domain.

Development of more efficacious antibodies for medical therapy and diagnosis.

Two procedures for improving the efficacy of medically important antibodies are described. The first procedure is designed to reduce the immunogenicity of nonhuman antibodies to the barest minimum--the "humanization" is accomplished by transplanting only the specificity-determining residues of the nonhuman antibody onto a human antibody template. The second procedure is designed to permit the easy production of multispecific/multivalent antibodies via heterodimer formation of electrostatically complementary Fc regions.

Phosphocholine binding immunoglobulin Fab McPC603. An X-ray diffraction study at 2.7 A.

The crystal structure of the Fab of McPC603, a phosphocholine-binding mouse myeloma protein, has been refined at 2.7 A resolution by a combination of restrained least-squares refinement and molecular modeling. The overall structure remains as previously reported, with an elbow bend angle between the variable and constant modules of 133 degrees. Some adjustments have been made in the structure of the loops as a result of the refinement. The hypervariable loops are all visible in the electron density map with the exception of three residues in the first hypervariable loop of the light chain. A sulfate ion occupies the site of binding of the phosphate moiety of phosphocholine.

Glycera dibranchiata hemoglobin. X-ray structure of carbonmonoxide hemoglobin at 1.5 A resolution.

The structure of carbonmonoxide Glycera hemoglobin has been determined to 1.5 A resolution by X-ray diffraction. The model, including ordered solvent, has been refined by the method of restrained least-squares to an R-value of 0.146. The positions of 1104 protein atoms and the oxygens of 155 water molecules have been determined with an estimated r.m.s. error of 0.10 to 0.13 A. The r.m.s. errors in protein geometry are 0.027 A for bond distances, 0.038 A for angle distances and 0.012 A for deviations of planar groups from their least-squares planes. The iron lies exactly in the plane of the heme nitrogens and the heme is very slightly domed toward the proximal side. The carbon-oxygen bond in the carbon monoxide ligand is bent 7.9 degrees away from the normal to the plane of the heme nitrogens. The ligand is in close contact with, and slightly removed from the heme normal by distal residues Leu 58(E7) and Val62(E11). Comparison of the CO structure with the 1.5 A deoxy structure shows that the majority of the rather small structural changes occurring upon ligation are mediated by movement of the heme due to shortening of the five iron to nitrogen bonds. There is very little empty space inside the molecule, and no direct channel from the solvent into the heme pocket; however, rotation of the side-chain of the distal leucine residue Leu 58(E6) could provide a ligand pathway.

Crystalline monoclonal antibody Fabs complexed to hen egg white lysozyme.

The Fab of a monoclonal anti-lysozyme antibody (HyHEL-10) has been crystallized as the free Fab and as the Fab-antigen complex. Crystals have also been grown of the antigen complex of the Fab of another monoclonal anti-lysozyme antibody (HyHEL-9), which recognizes a different binding surface of lysozyme. All three crystals diffract to at least 3 A resolution and are suitable for X-ray diffraction studies.

The X-ray crystal structure of a Valpha2.6Jalpha38 mouse T cell receptor domain at 2.5 A resolution: alternate modes of dimerization and crystal packing.

We describe here the structure of a murine T cell receptor (TCR) Valpha2.6Jalpha38 (TCRAV2S6J38) domain, derived from a T cell hybridoma with specificity for the H-2Ddmajor histocompatibility complex class I molecule bound to a decamer peptide, P18-I10, from the HIV envelope glycoprotein gp120, determined by X-ray crystallography at 2.5 A resolution. Unlike other TCR Valpha domains that have been studied in isolation, this one does not dimerize in solution at concentrations below 1 mM, and the crystal fails to show dimer contacts that are likely to be physiological. In comparison to other Valpha domains, this Valpha2.6 shows great similarity in the packing of its core residues, and exhibits the same immunoglobulin-like fold characteristic of other TCR Valpha domains. There is good electron density in all three complementarity-determining regions (CDRs), where the differences between this Valpha domain and others are most pronounced, in particular in CDR3. Examination of crystal contacts reveals an association of Valpha domains distinct from those previously seen. Comparison with other Valpha domain structures reveals variability in all loop regions, as well as in the first beta strand where placement and configuration of a proline residue at position 6, 7, 8, or 9 affects the backbone structure. The great variation in CDR3 conformations among TCR structures is consistent with an evolving view that CDR3 of TCR plays a plastic role in the interaction of the TCR with the MHC/peptide complex as well as with CDR3 of the paired TCR chain.

Structure and refinement at 1.8 A resolution of the aspartic proteinase from Rhizopus chinensis.

The structure of rhizopuspepsin (EC 3.4.23.6), the aspartic proteinase from Rhizopus chinensis, has been refined to a crystallographic R-factor of 0.143 at 1.8 A resolution. The positions of 2417 protein atoms have been determined with a root-mean-square (r.m.s.) error of 0.12 A. In the final model, the r.m.s. deviation from ideality for bond distances is 0.010 A, and for angle distances it is 0.034 A. During the course of the refinement, a calcium ion and 373 water molecules, of which 17 are internal, have been located. The active aspartate residues, Asp35 and Asp218, are involved in similar hydrogen-bonding interactions with neighboring residues and with several water molecules. One water molecule is located between the two carboxyl groups of the catalytic aspartate residues in a tightly hydrogen-bonded position. The refinement resulted in an unambiguous interpretation of the highly mobile "flap", a beta-hairpin loop region that projects over the binding pocket. Large solvent channels are formed when the molecules pack in the crystal, exposing the binding pocket and making it easily accessible. Intermolecular contacts involve mainly solvent molecules and a few protein atoms. The three-dimensional structure of rhizopuspepsin closely resembles other aspartic proteinase structures. A detailed comparison with the structure of penicillopepsin showed striking similarities as well as subtle differences in the active site geometry and molecular packing.

Generation and characterization of a single gene-encoded single-chain-tetravalent antitumor antibody.

Monoclonal antibody (mAb) CC49, a murine IgG1, reacts with the tumor-associated glycoprotein-72 expressed on a variety of carcinomas. In clinical trials, radiolabeled CC49 has shown excellent tumor localization to a variety of carcinomas. To minimize the immunogenicity of CC49 mAb in patients, a humanized CC49 (HuCC49) was generated by complementarity-determining region (CDR) grafting. The relative affinity of HuCC49 was 2-3-fold less than that of the murine mAb. With the aim of improving tumor targeting, attempts have been made to enhance the avidity of the HuCC49 mAb. Previous research has yielded a single gene-encoded immunoglobulin, SCIgcCC49deltaCH1, which is a dimer of a single chain consisting of CC49 single-chain Fv linked to the NH2 terminus of the human gamma1 Fc through the hinge region. This molecule is comparable to the mouse-human chimeric CC49 in terms of in vitro antigen binding properties, cytolytic activity, and rate of plasma clearance in athymic mice bearing human tumor xenografts. Recently, a single gene encoding a single-chain immunoglobulin consisting of a HuCC49 diabody attached to human gamma1 Fc via the hinge region was constructed. The diabody, a bivalent antigen-binding structure, is made up of variable heavy (V(H))/variable light (V(L)) domains and V(L)/V(H) domains. In each of the variable domain pairs, the V(H) and V(L) domains are linked through a short linker peptide. Meanwhile, the two pairs are linked via a 30-residue Gly-Ser linker peptide to yield two antigen-binding sites by lateral and noncovalent association of the V(L) of one pair with the V(H) of the other. Transfectomas expressing the single-gene immunoglobulin secrete a homodimer of about Mr 160,000 that reacts to tumor-associated glycoprotein-72. This tetravalent humanized antitumor immunoglobulin molecule may potentially be an efficacious therapeutic and diagnostic reagent against a wide range of human carcinomas.

The codon for the methionine at position 129 (M129) in the human prion protein provides an alternative initiation site for translation and renders individuals homozygous for M129 more susceptible to prion disease.

Single amino-acid substitutions in the prion protein have been found to lead to resistance or susceptibility to amyloid fibril formation. In humans, the presence of methionine at position 129 in the prion protein results in increased susceptibility to prion disease, while the presence of valine at that position appears to be protective. It is hypothesized that the codon for M129 is an alternative initiation site for translation, which results in a truncated molecule that is missing the first 128 amino acids, including the signal peptide. This N-terminal truncated form of the prion molecule will not be transported to the extracellular space and thus will accumulate in the cytosol where it is more susceptible to fibril formation and aggregation; this aggregation could hinder normal degradation processes and cause disease. The results of experimental studies on truncated prion molecules support this hypothesis. To test the hypothesis, a gene segment, which when transcribed would result in a prion molecule starting at methionine 129, could be introduced into a convenient experimental animal to see if there is increased incidence of prion disease. Or, fibrils from the brains of affected M129/M129 homozygous individuals could be isolated and the molecules in the fibrils analyzed to determine the identity of the N-terminal amino acid(s). We predict that those isolates will have a preponderance of molecules that start with the methionine at position 129 in the intact protein.

Why don't humans get scrapie from eating sheep? A possible explanation based on secondary structure predictions.

In an effort to find a structural explanation for the lack of direct transmission of scrapie from sheep to humans, secondary structure predictions are used to locate the segments of the prion sequence which may be involved in the transformation from the normal form of the prion protein, which has high helix content, to the pathogenic form, which has high beta-sheet content. The Chou-Fasman algorithm, which calculates propensities for both helix and sheet formation, was used to predict the secondary structures of the scrapie-resistant and the scrapie-susceptible variants of the ovine prion protein. The scrapie-susceptible variant, which has a glutamine at residue position 168 (human prion protein numbering), is predicted to have a propensity for sheet formation in that region of the molecule, while the scrapie-resistant variant, which has an arginine at position 168, does not. The valine at position 133, additionally present in the ovine variant which is the most susceptible to scrapie, is predicted to result in even more sheet formation. When the predicted secondary structure of the human prion protein is compared to those of the ovine prion protein variants, the human protein is found to be most similar to the scrapie-resistant variant. This result is proposed to provide a possible explanation for the observation that scrapie is not directly transmitted from sheep to humans.

Does base composition help predispose the complementarity-determining regions of antibodies to hypermutation?

A survey of the base usage in genes coding for human antibodies reveals that more (A+T) and less (C+G) are found in the segments coding for the complementarity-determining regions, while the opposite is true for the segments coding for framework and constant regions. The possibility that this bias in base usage may contribute to hypermutation is explored.

Quantitation of the immunogenic potential of protein antigens.

A procedure to quantitate the immunogenic potential of protein antigens is presented. It is hypothesized that the intrinsic immunogenicity of an accessible region on the protein arises from the overall structural effect of the presence of the particular assemblage of amino acid residues in the given region. Structural parameters previously derived by Grantham [Science 185, 862-864 (1974)] to differentiate the various amino acids are assigned to each residue position. At each point chosen on the molecule, an av. value is computed due to all residues within 8.5 A of this point; it is proposed that this local average is proportional to the immunogenic potential of the region centered at this point. The method can be used to locate the immunodominant regions of a molecule and to compare the antigenicity of related molecules. Test calculations on hen egg-white lysozyme, sperm whale myoglobin and horse cytochrome c show that segments in these molecules, that have been shown in immunochemical studies to possess antigenic activity, are predicted by this method to be immunodominant.

A possible procedure for reducing the immunogenicity of antibody variable domains while preserving their ligand-binding properties.

It is proposed to reduce the immunogenicity of allogeneic antibody variable domains, while preserving ligand-binding properties, by reducing their antigenicity through replacement of the exposed residues in the framework regions which differ from those usually found in host antibodies. The results of a comparison of representative murine antibody sequences with those of human origin suggest that the number of residues that need to be replaced to "humanize" those antibodies could be small.

Anatomy of the antibody molecule.

The structures of the various regions of an antibody molecule are analysed and correlated with biological function. The structural features which relate to potential applications are detailed.

A model of the Fc of immunoglobulin E.

A model of the Fc of human IgE was constructed using the known three-dimensional structure of IgG Fc. On the basis of amino acid sequence homology, the CE2, CE3 and CE4 domains were modelled after CG3, CG2 and CG3, respectively. The mode of association of the CG2 and of the CG3 pairs of domains was assumed for the CE3 and the CE4 pairs, respectively; the CE2 pair of domains was assembled such that they interact like the CG3 domains. An asymmetric linkage is favored for the two inter-epsilon chain disulfide bridges, i.e. the bonds were assumed to be between non-homologous cysteines. The atomic interactions in the interface of the CE2:CE2 and CE4:CE4 domain pairs were computed, and predictions are made on the solvent accessibility of the individual residues in the fragment. The model can be useful in the study of regions that may be involved in the interaction of IgE with Fc(epsilon) receptors.

Antibody Fab assembly: the interface residues between CH1 and CL.

The effective assembly of an antibody molecule requires the proper association of the light and heavy chains, namely the tight, canonical association of VH with VL, and of CH1 with CL. In this paper the interaction of CH1 is examined by looking at the degree of conservation of residues in the interface between CH1 and CL, where CH1 can belong to any of the heavy chain classes, and CL can be either lambda or kappa. The three-dimensional structures of four antibody Fabs have been examined to see which are the significant interacting residues and to see whether they also correspond to the conserved residues in the different classes. It was found that there are a few hydrophobic residues buried in the interface which make numerous contacts with residues of the other chain and which remain invariant, or else are highly conserved. Around the periphery of the interface there are numerous interacting residues that have appreciable variability. Within the interface there is a cavity, the function of which may be to permit some changes in the central interface residues while still preserving the same relative orientation of CH1 and CL.

Localization of rabbit kappa light chain allotypic determinants.

The amino acid sequences of the constant regions of rabbit kappa light chains (C kappa) are remarkably divergent. The K1 allotypes differ at 47 of 106 positions; the K1 and K2 isotypes differ at three additional positions. Variability and structural dissimilarity plots reveal that most of these differences occur in clusters. Major hydrophilic areas are also found near some of these clusters. The structures of rabbit C kappa are modeled using the known alpha-carbon backbone structure of the Fab fragment of mouse myeloma protein McPC603. The effect of sequence variations upon the hypothetical three-dimensional structures was assessed and immunogenic determinants predicted and located. It was found that predicted determinants were external and located in or near loops. Two clusters of potentially interacting regions were predicted. Within each there could be several "topographical" and overlapping sets of epitopes that are recognized by different antibody-combining sites. One of the predicted immunogenic sites clearly interacts with the CH1 domain of the heavy chain. A heavy-chain dependent serological determinant has been correlated with amino acid differences in this region (kappa chain positions 121 and 124).