Structures of antibody/MHC-I complexes reveal deficiencies in structure prediction programs

Abstract The remarkable success of structural prediction programs such as AlphaFold2 and RosettaFold have revolutionized the contemporary approach to protein structure determination. However, intrinsically disordered proteins (IDPs) and protein/protein complexes provide a greater challenge to such algorithms. In particular, there has been limited success in the structural prediction of antibody/protein antigen complexes and TCR/MHC complexes, in part because the database of such structures is limited. Here, we present several high-resolution X-ray structures of anti-MHC/MHC-I complexes consistent with previous exon-shuffling and epitope mapping data that contradict computational predictions. Specifically, two mAb previously mapped to the alpha3 domain of H2-D(d) or H2-L(d)/D(b) respectively are shown to bind in distinct orientations, either of which may sterically block CD8 binding. In addition, a peptide dependent, but peptide non-specific anti-H2-D(d) mAb binds to loops of the alpha2 domain. These results substantiate the need to determine experimental structures of antibody/protein complexes as we proceed to develop avenues for the de novo design of specific antibodies. Supported by the Intramural Research Program of the NIAID, NIH Supported by the Intramural Research Program of the NIAID, NIH