There are two main reasons why natural killer T (NKT) cells fascinate immunologists. First, they exhibit a unique mélange of properties found in innate immune cells as well as the properties of conventional T cells, part of adaptive immunity (1–4). Second, NKT cells exercise a determining influence on a variety of immune responses in mice, ranging from autoimmunity to their response to tumors and infections (1–4). Because NKT cells have a limited diversity of their T cell antigen receptor (TCR) chains, they are often called invariant (i) NKT cells (3, 5). iNKT cells recognize glycolipid antigens from certain bacteria that are presented by CD1d, a nonpolymorphic antigen-presenting molecule (4), but this is not the entire story. Although students are taught that lymphocytes must be tolerant of self, this is not the case for iNKT cells, because there is evidence that iNKT cells are activated as a result of contact with self-antigens in the thymus (1–3, 6). Self-reactivity may allow iNKT cells to respond rapidly, not by specifically recognizing a panoply of microbial antigens but by responding to a secondary signal of infection with a very limited TCR diversity. It is widely believed that the same self-antigens expressed in the thymus activate the mature cells. These mature cells are in contrast to conventional T cells, which are positively selected by weak TCR interactions and would be negatively selected by TCR agonists. Therefore, although iNKT cells obey the ancient aphorism ‘‘to know thyself,’’the chemical definition of the self-antigen has been elusive. It has been proposed that a single glycosphingolipid (GSL), isoglobotrihexosylceramide (iGb3), is required for iNKT cell differentiation in the thymus (7) and activation in the periphery (8, 9). GSLs are glycolipids that contain a ceramide lipid linked to one or more sugars (Fig. 1). Two papers in this issue of PNAS (10, 11) now cast doubt on the hypothesis that iGb3 is the unique self-antigen required for iNKT cells. In a biochemical analysis, Speak et al.(10) report that they could not detect iGb3 in the thymus or in dendritic cells (DCs) from mice and humans, whereas Porubsky et al.(11) find that mice lacking iGb3 synthase have normal iNKT cell number and function. Together, these studies suggest that there must be other self-antigens for iNKT cells.

The identification of iGb3 as a critical self-antigen was not based on biochemistry, but instead used genetics, immunology, and a series of deductions so incisive that the paper ‘‘unfolded like a detective story’’(12). Because earlier work showed that cells that cannot produce-glucosyl ceramide (see Fig. 1) cannot stimulate iNKT cell autoreactivity (13), Bendelac and coworkers (7) reasoned that the selfantigen was likely to be a GSL; therefore, they screened mice defective for enzymes involved in GSL metabolism. Mice lacking the gene encoding the-subunit of lysosomal hexosaminidase A and B enzymes (hexbJ/J mice) were found to be deficient for iNKT cells (7). GSLs with terminal N-acetyl galactosamine (GalNAc) sugars (shown in green in Fig. 1) therefore were thought to be potential precursors for the iNKT cell ligand because they are cleaved by-hexosaminidases. The ganglio series was eliminated, because mice deficient for the GalNAc transferase (Galgt1J/J) or GM3 synthase (Siat9J/J) had normal iNKT cells (7). Among the remaining candidates, only iGb3 proved to be an agonist for iNKT cells and was the likely self-antigen if one assumed that the selecting thymic ligand was a TCR agonist. In a subsequent study (8), it was shown that DCs activated by