Structurally modified forms of αGalCer have been explored in recent years as a strategy for more selectively activating specific functions of iNKT cells to exploit their therapeutic potential (Miyamoto et al., 2001; Schmieg et al., 2003; Miyake and Yamamura, 2005; Yu et al., 2005; Forestier et al., 2007
). The current study addresses the fundamental mechanisms that cause certain structural analogs of αGalCer to activate the strong, preferential production of Th2 cell-type cytokines by iNKT cells that is associated with anti-inflammatory or tolerizing effects. Our studies extend other work centered mainly on the truncated sphingosine analog known as OCH, which emphasized the extremely low affinity or avidity of this ligand for iNKT cell TCRs (Oki et al., 2004
). The weak binding of OCH-loaded CD1d molecules to the iNKT cell TCR, confirmed in the current study, suggested that an affinity threshold model could account for the Th2 cell-type cytokine bias and overall anti-inflammatory effects that are observed after activation of iNKT cells with this αGalCer variant (McCarthy et al., 2007
However, our findings argue against an affinity threshold model as the principle mechanism accounting for the properties of many or most Th2 cell-type cytokine-biasing iNKT cell agonists. As demonstrated in an earlier study (Forestier et al., 2007
) and extended here, these analogs showed a remarkable range of avidities for mouse or human iNKT cell TCRs. Structural studies of CD1d-glycolipid complexes also suggest that altered functional outcomes of iNKT cells to N-acyl variants are unlikely to be due purely to direct effects on TCR affinity (Zajonc et al., 2005; Koch et al., 2005; Borg et al., 2007
). These studies reinforce the view that alterations in the sphingoid base or acyl chain of αGalCer are mostly hidden from the TCR contact surface by their positioning within the A' and F' pockets of the CD1d lipid-binding groove. Although modeling suggests that a structural rearrangement of the TCR contact surface may be a result of incomplete filling of the A' pocket by the truncated sphingoid base of OCH (McCarthy et al., 2007
), the high avidity of at least some Th2 cell-type biasing N-acyl variants of αGalCer for iNKT cell TCRs argues against such effects as a universal explanation for the altered cytokine responses. Thus, mechanisms other than or in addition to variations in signaling based simply on TCR affinity must underlie the altered cytokine responses stimulated by certain αGalCer analogs.
An alternate mechanism supported by our data relates to how these compounds become associated with CD1d in APCs. Our findings did not strongly support the previously proposed hypothesis that presentation by different APC types accounts for qualitative differences in iNKT cell responses (Bezbradica et al., 2005; Im et al., 2006
). Instead, what emerged strongly from our studies was a consistent tendency for Th2 cell-type-biasing glycolipids to be presented with rapid kinetics and without a requirement for intracellular loading of CD1d, apparently as a result of their ability to rapidly associate with CD1d molecules directly at the cell surface. This was in marked contrast to αGalCer-C26:0, which underwent intracellular loading and was presented more slowly. It is noteworthy that a similar finding has been reported recently for αGalCer-C20:2 in studies that used novel phage display-derived antibodies with TCR-like specificity against the hCD1d/αGalCer complexes (Denkberg et al., 2008
), suggesting the potential relevance of our observations to human iNKT cell responses.
The lack of a requirement for detergent or for a lipid transfer protein (e.g., saposin B) as an accessory factor for glycolipid loading was also consistently observed for Th2 cell-type-biasing glycolipids, and we speculate that this was related directly to their ability to rapidly load surface-expressed CD1d molecules. The chemical structures of αGalCer analogs with pronounced Th2 cell-type cytokine-biasing effects are variable, but all of these are characterized by substantial shortening of their aliphatic chains or the inclusion of polar substitutions such as double bonds or oxygen atoms. These modifications increase the overall polarity and reduce the hydrophobicity of these glycolipids relative to compounds giving mixed cytokine responses like αGalCer-C26:0 and -C24:0, or the even more hydrophobic Th1 cell-type cytokine-biasing C-glycoside variant of αGalCer-C26:0 (Schmieg et al., 2003; Fujii et al., 2006
Our demonstration that αGalCer analogs can differ markedly with regard to the extent to which they are presented by lipid-raft-associated CD1d molecules provides an attractive mechanism to explain the stimulation of biased cytokine responses by particular glycolipids. It is likely that the forced intracellular loading of a compound such as αGalCer-C26:0 leads to the organized transport of glycolipid-CD1d complexes into cholesterol-rich lipid rafts, which also are known to facilitate recruitment into immunological synapses along with a variety of other molecules involved in T cell activation (Gombos et al., 2004b
). Conversely, direct loading of CD1d molecules on the surface of the cell results in exclusion from such lipid rafts and therefore promotes a very different type of iNKT cell activation. Such lipid-raft-dependent alternative activation, although novel for iNKT cells and CD1d-dependent antigen recognition, has previously been described in studies of MHC class II presentation that have remarkable parallels with our observations on CD1d. For example, it has been shown that a substantial fraction of plasma membrane MHC class II molecules are concentrated in cholesterol-rich lipid rafts and tetraspannin microdomains (Poloso and Roche, 2004
). Analogous to our findings with CD1d-presented glycolipids, peptide antigens loaded onto MHC class II molecules in intracellular compartments are presented preferentially in rafts and favor the stimulation of Th1 cell-type responses by conventional CD4+
T cells. In contrast, peptides loaded exogenously at the cell surface are presented mainly by molecules outside of rafts and show a tendency to stimulate Th2 cell-type responses (Buatois et al., 2003
Although our experiments used synthetic lipid agonists to show how ligand structure may regulate the quality of iNKT cell responses, the results may provide insight into how natural endogenous and foreign lipid ligands are presented and sensed by the CD1d-iNKT cell axis. Thus, the preferential loading of glycolipid antigens with relatively short or polyunsaturated alkyl tails onto cell-surface CD1d proteins may represent a constitutive process of normal self -recognition that biases the immune system toward tolerance or Th2 cell-type cytokine responses in the absence of an infectious agent. Interestingly, other recent data indicate that glycolipids with this type of structure are not only more efficient at loading directly onto surface CD1d but are also actively excluded from forming stable complexes with endosomal CD1d at low pH (A. Bendelac, personal communication). Such findings are mirrored by our observation that recombinant saposins have the ability to unload αGalCer-C10:0 from CD1d at pH 5.0, but not at neutral pH. Taken together, these studies are evidence for a finely tuned antigen-presenting system in which the sorting of structurally divergent glycolipids onto different cohorts of CD1d molecules is a key mechanism for regulating the balance between tolerogenic and proinflammatory functions of iNKT cells.