in binding lipids whereas CDld (the only isotype expressed by mouse and rats) binds peptides? The conservation of the 0.1 and 0.2 domains of different CDI isotypes among species (3)(see figure) certainly argues for unique properties of at least two subfamilies of CDI molecules-CDld and CDla, b, c, and e. However, further biochemical studies and determination of the crystal structures of both mouse CDI and human CDlb are needed to clarify these issues. One may wonder whether the raison d'etre of CDI is solely to allow T cells to recognize lipids or peptides with particular motifs. Although targeting of bacterial cell wall lipids and glycolipids makes evolutionary sense-they are conserved components of mycobacteria and other prokaryotic pathogens, which are nowhere to be found in eukaryotes (7)-recognition of mycobacterial peptides is already performed by the classical MHC molecules. They can be presented by conventional MHC class II molecules to elicit responses by classical CD4+ helper cells. Can the additional and redundant recognition of bacterial lipids impart enough evolutionary advantage to justify the conservation of CDl?

There are indeed several hints that the unique function of CDI may also be related to the special properties of the T cells that interact with it. Sieling et aI.(5) report that CDlb is abundantly expressed in the lesions of patients with the self-healing form of leprosy [where interferon-y (lFN-y)-secreting T helper 1 (THI) CD4+ cells are recruited] but not with the lepromatous form (where TH2 CD4+ cells predominate), and suggest that CDlb expression may determine the THl {fH2 set of specialized cytokines produced in response to M. leprae. The CDl-specific cells identified so far in humans consist mainly of cells with the unusual CD4-8-phenotype (2, 8). In addition, many of the human CDI-restricted cell lines seem to be autoreactive, in that they recognize CDI+ cells without foreign antigen. This autoreactive aspect of CDl-specific cells has been reported mainly but not exclUSively for CDla, c, and d (2, 8, 9). This is where the mouse system may provide more information than humans. CDlspecific T cells have been identified in normal unimmunized mice (10, 11) and appear also to be autoreactive. Although they include cells with the unusual CD4-8-phenotype, a majority are CD4+ cells. The absence of CD8+ cells is explained by the fact that persistence of the CD8 coreceptor imparts negative selection (I2), presumably because of the high avidity of their T cell receptors (TCRs) for CDl. Indeed, a promi-