SummaryTo better understand the biological implications of the association of ligand with major histocompatibility complex class I molecules, we have studied the Ld molecule of the mouse. The culturing of various nonselected cell lines with three different known Ld peptide ligands resulted in a two-to fourfold specific increase in surface Ld expression as detected by 10 of 11 different monoclonal antibodies (mAbs) recognizing Ld epitopes . These findings suggest that Ld molecules are not saturated with endogenous peptide ligands and thus have accessible binding sites . Exploiting this feature of Ld we demonstrate that the physical association of Ld with ligand is exquisitely specific, indicating that they function in determinant selection . In addition, a non-peptide-bound antigenic variant of Ld was specifically detected with an exceptional mAb designated 64-3-7. In comparison with other Ld molecules, 64-3-7+ Ld molecules are not peptide ligand inducible, are more susceptible to proteolysis, lack 02 microglobulin association, and display a slower rate of oligosaccharide maturation . In spite of their deficiencies, the non-ligandassociated 64-3-7 Ld molecules were detected on the surface of all cell types tested; however, they appear not to be recognized by alloreactive cytotoxic T lyphocytes .Class I MHC molecules are highly polymorphic 45-kD membrane glycoproteins that associate noncovalently with a-2 microglobulin (02m), 1 a non-MHC-encoded, non-membrane-bound 11-kD polypeptide . Although each MHC haplotype of the mouse contains approximately 40 class I genes, only a few have known functions . For example, the H-2d haplotype, represented in the BALB/c inbred strain, expresses three class I molecules designated Kd, Dd, and Ld that function as classical transplantation antigens. The K, D, and L molecules on virus-infected or allogeneic cells function as recognition structures for CTL . Crystallographic studies revealed that the highly polymorphic ul and cr2 domains of the class I molecule combine in an intricate folding pattern to form a single potential binding site (1-3) . In fact the putative ligand binding site of the crystallized class I molecules was found to contain heterogeneous material estimated to be til-2 kD. In other studies using functional assays, virusspecific CTL were found to recognize a processed virus-derived peptide ligand in the context of a self class I molecule (4) . There are now several examples where the specific virus-derived peptide has been identified for a given CTL clone (e .g., refer-'Abbreviations used in this paper. Ag, antigen ; 02m, /3-2 microglobulin ; BEA, brefeldin A . ences 5, 6). Furthermore, these peptide ligands have been found to be between 5 and 20 amino acids in length . Thus, there is complete concordance between the crystallographic and functional studies. In spite of this knowledge, direct evidence of the binding of peptide to class I molecules has been very difficult to demonstrate using in vitro binding assays analogous to ones previously used to show class II...