Studies investigating insulitis in non-obese diabetic (NOD) mice were amongst the first to elucidate the regulatory activity of iNKT cells, with early work revealing a critical link to IL-4 and IL-10 production (18–21), and further analysis showing that they promote the differentiation of tolerogenic APCs that limit the activation of autoreactive T cells (22–25). A similar axis has been observed in murine models of diet-induced obesity, where adipose-resident iNKT cells play a powerful role in glucose tolerance by promoting macrophage polarization into a non-inflammatory M2 phenotype through secretion of IL-4 and IL-10 (26, 27), and by transactivating regulatory T cells via secretion of IL-2 (28). iNKT cells also contribute to the resolution phase of sterile inflammation in the liver by promoting monocyte transition into an anti-inflammatory phenotype through secretion of IL-4 (29, 30). In murine models of allogeneic hematopoietic transplantation, iNKT cells protect against GVHD through IL-4 dependent mechanisms (31–33), and by promoting the regulatory functions of myeloid-derived suppressor cells (MDSCs) while driving T_reg expansion via secretion of IL-2 (34–36).

Analyses of human iNKT cells have suggested that they may participate in similar regulatory processes. IL-10 producing iNKT cells recently identified in the intestinal lamina propria of Crohn’s Disease patients showed suppressive activity towards pathogenic CD4⁺ T cells, and the frequency of IL-10 producing iNKT cells in colon tissue of these patients correlated inversely with T_H1 and T_H17 cell frequency, and was associated with reduced disease severity, higher TGFB gene expression, and lower levels of inflammatory proteins (37). Moreover, co-culture of human T_regs with iNKT cells led to increased T_reg FOXP3 expression, enhanced IL-10 secretion, and more profound inhibition of conventional T cell proliferation (38).

Human iNKT cells can also mediate potent suppression of T cell IFN-γ production by modulating the functions of monocytic cells. Our research group showed that GM-CSF and IL-13 secretion by human iNKT cells induced monocytes to differentiate into tolerogenic APCs that produced high levels of IL-10, expressed the checkpoint inhibitors PDL-1 and PDL-2, and potently suppressed T cell proliferation and IFN-γ secretion (39, 40). The regulatory phenotype of the APCs was due to iNKT cell release of extracellular ATP, which signaled through the P2X7 receptor on the monocytes to induce upregulation of PD-L1 and PD-L2 (41). This iNKT-monocyte interaction resembles a pathway observed in a murine model in which IL-13 secreted by CD1d-restricted T cells promoted monocyte expression of TGFβ, which led to suppression of T cell effector responses (42, 43), although the role of TGFβ in the human iNKT-monocyte pathway remains unclear.

We also used a xenotransplantation model of hematopoietic engraftment to investigate the impact of the human iNKT-monocyte pathway in vivo. The addition of allogeneic adult iNKT cells to human cord blood mononuclear cell grafts resulted in dramatically improved engraftment, which was due to iNKT cells inducing cord blood monocytes to secrete prostaglandin E₂, which potently suppressed T cell IFN-γ production (44). Since hematopoietic engraftment is suppressed by excessive IFN-γ (45), this analysis shows that human iNKT cells can engage powerful regulatory pathways that limit adverse effects of human T_H1 activation in vivo.

A number of studies have suggested that iNKT cells may also control inflammation by eliminating pro-inflammatory APCs through a mechanism involving CD1d-dependent activation of the iNKT cells and lysis of APCs by cytotoxic granule deposition (46–50). Human iNKT cells were found to kill monocyte-derived DCs and blood DCs, but did not kill monocytes or plasmacytoid DCs, suggesting they specifically target certain types of APCs (46, 49). In another analysis, human iNKT cells preferentially eliminated monocyte-derived DCs that produced high levels of IL-12 while those that produced mainly IL-10 were spared, resulting in a DC population that limited T_H1 activation (48). Together these studies suggest that this cytolytic pathway selectively targets pro-inflammatory APCs, and might thereby limit pathological inflammation. Consistent with this, in mice infected with a highly pathogenic strain of influenza A virus, iNKT cells were associated with reduced accumulation of inflammatory monocytes in the lungs (50). iNKT cell activity in this model was associated with reduced levels of MCP-1 (a chemokine that recruits monocytes and CD4⁺ T cells), reduced damage to lung tissue, and improved survival even though viral loads were not affected (50). The effect of iNKT cells was thought to be due to their cytolytic activity against influenza-infected monocytes, suggesting that iNKT cells may limit pathological inflammation during viral infections by eliminating inflammatory APCs. However, an important note is that in all of these studies the iNKT cells were experimentally exposed to strong TCR stimulation prior to analysis of their cytolytic activity. Therefore, the physiological conditions that might lead to APC-targeted cytolytic activity by iNKT cells remain unclear.

Due to their shared use of a canonical TCRα chain, all iNKT cells recognize an unusual type of glycolipid in which the sugar head group is present in an α-anomeric configuration. Certain microbes produce glycolipids of this type that are potent antigens for iNKT cells (reviewed in (8)). Recent studies indicate that bacterial species that can be found within the normal gut microbiota can produce similar antigenic lipids (58, 59), although these may be counter-regulated by related forms produced by other bacteria that are antagonists (60). These studies suggest that, particularly at mucosal sites, TCR-dependent activation of iNKT cells may fluctuate according to the composition of the microbial community.

iNKT cells can also recognize self-lipids as antigens. Mammalian cells do not directly synthesize the α-linked glycolipids recognized by iNKT cells, but the β-linked forms they produce may be converted at low frequencies to α-linked forms that are strongly antigenic (61, 62). Additionally, iNKT cells can recognize mammalian β-linked glycolipids as weak agonists (63). Some antigenic self-lipids, including lysophospholipids, glycosylated sphingolipids, and neutral lipids, are specifically upregulated during inflammation or cellular stress (64–70). Conversely, some non-antigenic self-lipids, such as sphingomyelin, can inhibit presentation of antigenic species (66). Together, the available data suggest that antigenic self-lipids are constitutively present, but are maintained in a manner that is only weakly agonistic for iNKT cells, and that during inflammation or cellular stress the abundance or nature of the antigenic self-lipids changes in a way that provides stronger TCR signals to iNKT cells. Additionally, as discussed below, activation by self antigens can be markedly enhanced by TCR-independent signals (71, 72).

In contrast to conventional T cells that become polarized into different effector phenotypes by priming in the periphery, iNKT cells are already cytokine competent as they exit the thymus (74). Murine iNKT cells are segregated into four functionally distinct subsets based on their expression of master-regulator transcription factors that govern cytokine production (Tbet, GATA3, RORγT, E4BP4) and on differences in expression levels of PLZF (promyelocytic leukemia zinc finger), a transcription factor that promotes cellular characteristics associated with innate lymphocytes (28, 75–78). NKT1 cells have a T_H1 cytokine profile, often express a cytotoxic effector program, and are PLZF^loTbet^hi; NKT2 cells are characterized by high levels of IL-4 secretion and are PLZF^hiGATA3⁺; NKT17 cells produce IL-17 and express RORγT with intermediate levels of PLZF; NKT10 cells produce IL-10, are preferentially found within adipose tissues, and are negative for PLZF but express E4BP4 ( Figure 2A ). NKT1, NKT2, and NKT17 lineages are generated during thymic selection, and are thought to home to distinct tissues (79). In contrast, NKT10 cells may originate from other subsets and differentiate into a regulatory phenotype as a result of exposure to factors in adipose tissues (27). The identification of these iNKT sub-lineages has led to the paradigm that the anti-inflammatory effects of iNKT cells are due to NKT2 or NKT10 cells, which become activated in different situations than NKT1 and NKT17 subsets as a result of differences in tissue localization.

In contrast, it has thus far not been straightforward to categorize human iNKT cells into NKT1, NKT2, and NKT17 lineages matching those in mice. Similar to their murine counterparts, most human iNKT cells express PLZF (80–82), and are characterized by an innate-like transcriptional profile that results in a “poised-effector” status allowing them to rapidly mediate functional responses (83). Multi-parameter flow cytometric analyses and gene expression studies have revealed human iNKT cells to express a diverse selection of cytokines and chemokines (84–88). Human iNKT cells can be segregated into two major subsets according to CD4 expression (84, 85). Those that express CD4 often appear to co-produce GM-CSF, IL-13, TNF-α, IFN-γ, IL-4, and IL-2, while those lacking CD4 appear more specialized for cytolysis ( Figure 2C ). These two major populations are sub-divided into further subsets characterized by additional markers (e.g. CD8α, CD161, CD62L) with distinctions in functional characteristics, but it is not clear that these subsets equate to the NKT1, NKT2, or NKT17 lineages observed in mice (89, 90). It is also not clear whether anti-inflammatory activity segregates according to CD4⁺ or CD4^- status of human iNKT cells, although CD4⁺ iNKT cells are the ones that have been found to induce regulatory functions in monocytic cells, and the CD4^- subset has appeared more likely to kill DCs.

The prototypical iNKT antigen is called α-galactosylceramide (α-GalCer) (91), and synthetic forms of this lipid have proved extremely valuable as pharmacological agents that activate iNKT cells in a highly specific manner (92). Observations that structural variants of α-GalCer can produce substantially different immunological outcomes in vivo have led to interest in using these agents to selectively tune iNKT responses towards pro- or anti-inflammatory functions (92). Administration of α-GalCer to mice potently stimulates iNKT cells, and induces a mixed response where T_H1, T_H2, and regulatory cytokines are all produced, although with different kinetics (93). In contrast, certain analogues of α-GalCer have been shown to produce a T_H2-biased cytokine response (94, 95), while other variants produce a highly T_H1-biased response (96) ( Figure 2B ). The mechanisms underlying these differential responses appear complex. One component may be that certain variants induce biased cytokine production from iNKT cells themselves (97), while another important element likely relates to whether or not antigen-driven interactions between iNKT cells and APCs result in release of cytokines (e.g. IL-12) that activate a secondary IFN-γ response by NK cells (96, 98). A key factor may be the relative duration of antigen presentation by CD1d molecules, with more durable antigens being associated with T_H1-biased responses (99). Additionally, T_H1-biasing forms of α-GalCer may be selectively presented by APCs that produce IL-12, whereas T_H2-biasing forms may be more promiscuously presented and thus less likely to produce a secondary wave of IFN-γ production by NK cells (100). It is not clear whether antigen variants selectively activate different iNKT cell subsets, or bias the cytokine profile produced within a given subset (for example, by inducing higher IL-4 production by NKT1 cells, or increased IFN-γ by NKT2 or NKT17 cells), or whether any structural variants selectively promote IL-10 production. Interestingly, repeated administration of α-GalCer results in selective loss of its T_H1-promoting features, but under such “anergizing” conditions α-GalCer retains the ability to induce IL-4 secretion and to promote control of EAE pathology (101).

Whether human iNKT cell responses can be modulated similarly using α-GalCer structural variants remains an open question. It has become clear that TCR differences between murine and human iNKT cells result in significant discrepancies in TCR-signaling strength induced by lipid variants (102). Perhaps more importantly, polyfunctional human iNKT cells show a hierarchy of cytokine production in response to TCR stimulation that does not neatly segregate into clear T_H1 or T_H2 patterns. Weak TCR stimulation of human iNKT cells preferentially induces production of IL-3, GM-CSF, and IL-13, with increasing stimulation leading first to IFN-γ, then IL-4, then IL-2 (44, 103, 104) ( Figure 2D ). Secondary induction of NK cell IFN-γ secretion was associated with activation of human iNKT cells by strong TCR agonists (104). It is therefore not clear that it will be feasible to selectively polarize human immune responses towards IL-4 production through the use of specific lipid antigen variants, although it may be possible to drive IL-13 production through administration of weak agonists.