STRUCTURE PREDICTION OF E5.2 (ANTI-IDIOTYPE ANTIBODY TO ANTIBODY D1.3) 1. SEQUENCE ALIGNMENT AND ANALYSIS
The amino acid sequences of the variable region light and heavy chains of E5.2 were the starting points for the structure prediction. These sequences were compared with the Fab fragment amino acid sequences of known immunoglobulins. Sequence homology was checked using a multiple sequence alignment routine from the program PCGENE (Intelligenetics, San Jose, CA). Framework and hypervariable regions were determined by comparing E5.2 to known structures. The best X-ray structure for homology modeling was determined by comparing E5.2 to known canonical structures. The best starting structure for comparative homology modeling was determined to be mouse IgG Fab fragment R19.9 ( anti-arsenate, 1FAI ). The two protein sequences, E5.2 and R19.9, show a considerable degree of homology, 93.5% for light chain, and 58.7% for heavy chain. The canonical structure groups (Chothia et al., Nature, 342:877, 1989) for the six hypervariable regions were: L1 Group 2, L2 Group 1, L3 Group 1, H1 Group 1, H2 Group 2, H3 Not available.
2. COMPARATIVE HOMOLOGY MODELING OF E5.2
All framework residues were substituted without modification of 1FAI protein backbone (Bolger et al. Methods in Enzymol., 203:21 1991). Likewise, hypervariable regions (L1, L2, L3, H1, and H2) from E5.2 were equal in length and canonical structure to 1FAI backbone and were not modified. Hypervariable region H3 was 2 residues shorter for E5.2 than for 1FAI. Consequently, the backbone for H3 of 1FAI was shortened by removing Glycine #102 H and Tyrosine #109 H using the program Hyperchem (Hypercube, Inc., Canada). A new bond was formed and the hypervariable loop alone was subjected to energy minimization to adjust the bond lengths. The whole structure of E5.2 was subjected to energy minimization using the program AMBER 4.0 (UCSF). The energy minimized structure was converted to PDB format and submitted.
3. DOCKING OF E5.2 ANTI-IDIOTYPE WITH D1.3 ( ANTI-LYSOZYME )
Several modes of interaction are possible for binding of E5.2 to D1.3.
We assumed that E5.2 hypervariable loops could recognize public or
private epitopes of D1.3. Therefore, we attempted to use DOCK (Kuntz,
et al., J. Mol. Biol. 161:269, 1982) to produce a family of
geometrically plausible binding interactions. First, solvent excluded
molecular surfaces were created for just the hypervariable region of
E5.2 ("receptor") and the entire D1.3 Fv region ("ligand"). Then
spheres of both receptor and ligand were generated using SPHGEN. The
resulting set of dock spheres were too large for the program DOCK 2.0
to handle. Consequently, the spheres from SPHGEN were visualized using
MIDAS (UCSF) and edited to suit DOCK 2.0. Also, the entire D1.3 Fv
region was too large to handle with DOCK 2.0. Therefore, we attempted
to "dock" the hypervariable regions. The resulting "docked"
macromolecules were less than satisfactory. Only a single chain of
D1.3 made contact with E5.2 (see poster at meeting for details).
Reasons for the failure of DOCK 2.0 to work in our hands may be due to
a lack of familiarity with the details of the software or it could be
related to the fact that DOCK was designed to handle small molecule
docking rather than macromolecular surfaces.
Structural Biology home page Asilomar Conference home page