Cytotoxic T lymphocytes (CTL) play a pivotal role in preventing persistent viral infections and aborting acute infections. H-2D b-restricted CTL optimally recognize a specific peptide of 9 to 11 amino acids (aa) derived from a viral protein and held in place (restricted) by a MHC class I glycoprotein on the surfaces of infected cells. Only three peptide sequences with the appropriate D b motif from lymphocytic choriomeningitis virus Armstrong strain (LCMV) are known to be presented to CTL by H-2D b molecules; they are from the glycoproteins (GP), residues 33-41 KAVYN FATC (GP1) and 276-286 SGVENPGGYCL (GP2), and the nucleoprotein (NP), 396-404 FQPQNGQFI. Incubation of virally infected H-2 b cells with CTL clones that recognize only GP1, GP2, or NP leads to the selection of viral variants which upon infecting cells hearing H-2 b molecules, escape recognition by CTL of the appropriate specificity. Nucleic acid sequencing showed a single mutation in GP1 (aa 38 F→L), GP2 (aa 282 G→D), or NP (aa 403 F→L) in the variant viruses. When wild-type (wt) LCMV peptides and the three variant peptides (GP1, GP2, NP) were synthesized and subjected to a competitive inhibition binding assay, no differences in binding affinity for H-2D b were found between the wt and variant peptides. Uninfected cells coated with the wt peptide were recognized and lysed by the appropriate CTL clone or by in vivo-primed bulk CTL, but similar targets coated with the GP1, GP2, or NP variant peptides were not. This result, coupled with computer graphic analysis of these variant peptides with the recently solved three-dimensional structure for the D b MHC class I molecule, placed the side chain of the mutated residues on the outer surface of the MHC-peptide complex and accessible to the T cell receptor, Ala substitution at GP residue 38 or 282 or at NP 403 also abrogated CTL recognition and lysis. Inoculation of any one of the mutated viral variants into mice produced an effective CTL response to the other two nonmutated GP or NP peptides, suggesting that production of biologically relevant CTL escape virus variants in vivo requires selection of mutations in more than one and likely all the CTL epitopes, a low probability event.