To study pathogenic T helper cells in myasthenia gravis (MG) reacting against the acetylcholine receptor (AChR), we have previously selected five CD4+ T cell lines/clones from MG patients (or healthy controls) against full-length recombinant human AChR alpha subunit (alpha 1-437); these can all recognize AChR solubilized from human muscle. Recently, T cells selected with pooled AChR subunit synthetic peptides have shown greater heterogeneity than above. Hoping to validate that, we have characterized three MG and six control T cell lines selected with pooled peptides (averaging 33 residues long) covering the alpha subunit sequence; recurring responses to three particular peptides each showed preferred HLA class II restrictions--p75-115/DR4, p138-167/DR4, and p309-344/DR3 (or DR52a). However, none of three lines from MG patients recognized p138-167--even one from a previous responder to this epitope in full-length alpha 1-437; otherwise they resembled those from controls. Moreover, no peptide-selected line responded significantly to whole AChR, alpha 1-437, or even to shorter polypeptides sharing one terminus with the peptide, suggesting specificity for epitopes not naturally processed by APCs from blood. Of 20 sublines maintained with individual peptides, at least 10 responded to independently synthesized overlapping sequences, but four others depended on contaminants in the original peptides. A single line did recognize one longer polypeptide, but only after tryptic digestion; the processing of this cryptic epitope was evidently the limiting factor here rather than its concentration or the T cell sensitivity. Therefore, while synthetic peptides are essential for mapping epitopes, assessment of the pathogenic MG T cell repertoire requires full-length Ag processed naturally.