The immune system is normally tolerant to mitochondrial self-antigens, but responsive against bacteria. Low-titre anti-mitochondrial antibodies (AMA) might be involved in this discrimination. Tolerance is broken in diseases characterised by high titre AMA. Some of these AMA, against cardiolipin, cross-react with DNA. The best studied AMA are those characterising primary biliary cirrhosis (PBC). These are directed against E2 subunits of the oxo-acid dehydrogenase complexes, and also against subunits E1 alpha, E1 beta and X of the pyruvate dehydrogenase complex. AMA of PBC patients also react with bacterial E2s. Reactivities are primarily peptide-specific but with cross-reactivity between mitochondrial and microbial antigens and between E2s of respective complexes. Immunodominant epitopes, for anti E2 AMA, include the conserved sequence flanking the site of lipoyl attachment. It is proposed that the initial stimulus for antibody production is chronic urinary tract infection. AMA themselves are not pathogenic, but CD4+ T-cells would be primed, recognising the lipoyl domain epitope in association with class II HLA. Inappropriate expression of class II antigens on bile duct epithelia, (as found in PBC), might lead to presentation of a particular fragment of HLA-DR alpha, known to be a major MHC presented self-peptide in the mouse. That sequence strongly mimics the lipoyl domain and might be recognised by primed T-cells, initiating the autoimmune cascade. In the mouse, a peptide of ND1 of Complex I is presented in association with class I MHC. Cells exhibiting somatic mutation of such a peptide might thus be subject to attack by CD8+ T-cells. If such peptides were presented by class II HLA, autoimmune diseases might arise, related to mimicry between such peptides and microbial sequences and/or self-antigens. These considerations might apply in Leber's disease and in age-related pathology.