Altered peptide ligands (APL) are highly effective in inhibiting experimental autoimmune encephalomyelitis (EAE) in rodents although clinical trials in multiple sclerosis reveal severe limitations probably due to the diverse and differential effects of APL in vivo compared to in vitro. Myelin oligodendrocyte glycoprotein (MOG 8-21) induces relapsing EAE in ABH (A(g7)) mice associated with broadening of the autoimmune repertoire thus providing a dynamic system to examine the efficacy of peptide analogues. Subtle changes in MOG 8-21 dramatically influenced disease susceptibility and T cell responses in vitro. Non-encephalitogenic APL that induce production of the 'regulatory' cytokines IL-10 and/or TGFbeta and concomitant low levels of the 'proinflammatory' cytokines IFNgamma and TNFalpha modulated relapsing EAE but were far less effective than the 'proinflammatory' wild-type MOG 8-21 peptide. These data reveal that APL differ greatly in their ability to activate encephalitogenic T cells. The extensive heterogeneity of responses of APL in vitro suggests that selection of APL on this criteria is highly unpredictable and probably less effective for therapy than selecting the dominant wild-type epitope and delivering it using a tolerogenic route.