(three-dimensional structure comparisons/structure-seq antibody diversity/evolutionary relationships)

A comparison of various variable immuno- globulin domains demonstrates that structural variability parallels the observed frequency of amino-acid substitu- tions. The similarity of the structural frameworks is in good agreement with the evolutionary relationships of the chains as derived from sequence analysis. Statistical analyses of the frequency of amino-acid substitu- tions in immunoglobulins have shown that some regions are distinctly more variable than others (1-3). Three such hyper- variable regions were found in light (L) chains (1) and four were found in heavy (H) chains (2, 3). Recent crystallographic studies on immunoglobulin Fabs and Bence-Jones dimers (4- 10) have shown that the domains of these fragments have similar overall folding, as predicted earlier on the basis of sequence data (11-13). The variable (V) domains resemble each other closely, as do the constant (C) domains. Larger differences were found between V and C domains than be- tween any two V domains or any two C domains (14). The crystallographic results also revealed that the polypeptide segments in the hypervariable regions are quite dissimilar in their folding. These regions, which together form the antigen- binding surface, are in the form of loops; some loops are longer or shorter, depending on insertions or deletions; some have extended configurations, while some are helical. Thus, the hypervariable regions provide the molecule with a combining surface which, because of the hypermutability of its con- stituents, can account for the diversity of the antibody re- sponse as well as its specificity. The nonhypervariable seg- ments, on the other hand, apparently provide the rigid frame- work necessary for maintaining the structural integrity of the molecule. Using the method of diagonal plots (15-17), we have compared the domains of the murine McPC603 (a, K) protein Fab (4-7) and the VL domain of the kappa-type Bence-Jones protein REI (10). With such plots, similarities in folding of molecules can be assessed without an actual superposition of the structures. We present a statistical com- parison that has enabled us to obtain a measure of structural homology and variability in immunoglobulin domains. The results of our structural comparisons will be correlated with the results of the statistical analyses of the sequences.