Strategy of lipid recognition by invariant natural killer T cells: ‘one for all and all for one’


T. Mallevaey, Department of Immunology, University of Toronto, Medical Sciences Building Room 7308, 1 King’s College Circle, Toronto, ON M5S 1A8, Canada. Email:
Senior author: Thierry Mallevaey


Invariant natural killer T (iNKT) cells are evolutionarily conserved lipid-reactive T cells that bridge innate and adaptive immune responses. Despite a relatively restricted T-cell receptor (TCR) diversity, these cells respond to a variety of structurally distinct foreign (i.e. microbial or synthetic) as well as host-derived (self-) lipid antigens presented by the CD1d molecule. These multi-tasking lymphocytes are among the first responders in immunity, and produce an impressive array of cytokines and chemokines that can tailor the ensuing immune response. Accordingly, iNKT cells play important functions in autoimmune diseases, cancer, infection and inflammation. These properties make iNKT cells appealing targets in immune-based therapies. Yet, much has to be learned on the mechanisms that allow iNKT cells to produce polarized responses. Responses of iNKT cells are influenced by the direct signals perceived by the cells through their TCRs, as well as by indirect co-stimulatory (and potentially co-inhibitory) cues that they receive from antigen-presenting cells or the local milieu. A decade ago, biochemists and immunologists have started to describe synthetic lipid agonists with cytokine skewing potential, paving a new research avenue in the iNKT cell field. Yet how iNKT cells translate various antigenic signals into distinct functional responses has remained obscure. Recent findings have revealed a unique and innate mode of lipid recognition by iNKT cells, and suggest that both the lipid antigen presented and the diversity of the TCR modulate the strength of CD1d-iNKT TCR interactions. In this review, we focus on novel discoveries on lipid recognition by iNKT cells, and how these findings may help us to design effective strategies to steer iNKT cell responses for immune intervention.


Natural Killer T (NKT) cells were originally identified as T cells co-expressing bona fide natural killer (NK) cell receptors, such as NK1.1 (CD161).1 Only later were these cells shown to respond to lipid antigens presented by the MHC class Ib molecule CD1d. The use of NK1.1 to define NKT cells is inaccurate because this receptor is not uniformly expressed by all NKT cells, and its expression is also regulated during ontogeny and upon activation. Hence, NKT cells can be broadly defined as T cells that respond to lipid antigens presented by CD1d.1 The most widely studied NKT cells are referred to as type 1, or invariant NKT (iNKT) cells. These cells express αβ T-cell receptors (TCR-αβ) with restricted diversity. Mouse iNKT TCRs are composed of an invariant TCR-α chain formed by the canonical rearrangement of the Vα14 to Jα18 gene segments. This TCR-α chain is associated with TCR-β chains limited in their Vβ usage (Vβ8, Vβ7 and Vβ2) but with extensive CDR3β junctional diversity. Human iNKT TCRs are formed by a canonical Vα24-Jα18 TCRα chain associated with Vβ11. Strikingly, the high degree of conservation of iNKT TCRs and CD1d molecules between humans and mice allows for inter-species reactivity. This feature seems to be a landmark of MHC class Ib molecules, such as CD1d, MR1 and Qa-1b, and highlights the importance of MHC class Ib-restricted innate-like T cells in the immune system.2 Almost all iNKT cells recognize the prototypical glycolipid α-galactosylceramide (αGalCer) presented by CD1d, and can be stained with CD1d tetramers loaded with this lipid antigen.3,4 Of note, CD1d-restricted αGalCer-responsive T cells expressing TCRs that differ from the above-described Vα24-containing and Vβ11-containing iNKT TCRs have been identified in humans.5–7 In addition, the group of Godfrey recently identified a population of αGalCer-reactive NKT cells in mice that express another canonical TCR-α chain, formed by the rearrangement of Vα10 to Jα50 gene segments, and paired with a limited set of Vβ chains.8 These cells, named Vα10 NKT cells, appear reminiscent of iNKT cells in their phenotype and function. Beyond the classification of these cells under the type 1 iNKT cell umbrella or a distinct category, the important question in the future relates to the functions that these cells play in immune responses.

In addition to iNKT cells, mice and humans have other populations of NKT cells that have been named type 2 NKT cells. These cells are CD1d-restricted, are considered to have broader TCR diversity, and usually express NK receptors. These cells are commonly believed to be more heterogeneous in their antigenic specificities, and recognize lipid antigens that are presumably distinct from type 1 NKT cell antigens. Indeed, type 2 NKT cells with limited TCR usage have been shown to respond to sulfatide antigens.9 The study of type 2 NKT cells is arduous, mainly because of the lack of specific markers. One way to study the functions of type 2 NKT cells is to dissect differences between Jα18−/− mice, which specifically lack iNKT cells, and CD1d−/− mice that lack all CD1d-restricted T cells, including type 1 and type 2 NKT cells. The discovery of Vα10 NKT cells complicates the interpretation of these studies, because these cells are present in Jα18−/− mice.8

In addition to CD1d, humans but not mice, express other isoforms of the CD1 molecule, namely CD1a, CD1b, CD1c and CD1e. Whereas CD1e is not directly involved in lipid presentation to the TCR,10 the other CD1 molecules present a variety of self and microbe-derived (mainly derived from mycobacteria) lipids to T cells. Recent findings have revealed that these cells, sometimes referred to as NKT-like cells, constitute a sizeable fraction of peripheral T cells,11,12 and probably play important functions in antimicrobial immunity. In summary, mammals are equipped with an army of lipid-reactive T-cell populations that have important functions in immune responses. Although type 2 NKT cells and NKT-like cells are important components of the immune system, this article focuses on iNKT cells.

Development and function of iNKT cells

Recognition of CD1d by newly-rearranged iNKT TCRs is a key event that drives iNKT cell selection from the pool of developing double-positive thymocytes.13 Accordingly, mice that lack CD1d or the Jα18 gene segment are devoid of iNKT cells. The nature of the positively selecting self-lipid(s) remains obscure, and may include the glycosphingolipid (GSL) isoglobotrihexosylceramide (iGb3).14 Although iGb3 is an agonist antigen for mouse and human iNKT cells, a series of recent observations has challenged its role in iNKT cell development.15–17 In addition to CD1d–TCR interactions, two members of the signalling lymphocytic-activation molecule (SLAM) family provide important second signals for the development of iNKT cells.18 As homotypic SLAM–SLAM interactions are uniquely provided by thymocyte–thymocyte contacts, it explains previous findings that CD1d-expressing thymocytes, not thymic epithelial cells, were necessary to support iNKT cell development.19,20 Finally, intrinsic expression of the promyelocytic leukaemia zinc finger (PLZF) transcription factor is essential to their ontogeny and drives their innate phenotype.21,22 Newly selected iNKT cells undergo expansion and a highly orchestrated maturation process, along which they acquire expression of NK receptors and, importantly, the ability to produce interleukin-4 (IL-4) and then interferon-γ (IFN-γ). Most mature iNKT cells have an activated/memory phenotype at steady state, characterized notably by high expression levels of CD44 and CD69 and low levels of CD62L. This peculiar phenotype suggests that iNKT cells may be the product of agonist selection in the thymus. In agreement with this, a recent study suggested that developing iNKT cells receive stronger TCR stimuli than conventional T cells upon selection.23

A signature feature of iNKT cells is their ability to rapidly produce copious amounts of a large array of cytokines and chemokines upon activation. This includes IFN-γ, IL-4, IL-17, IL-10 and transforming growth factor-β (TGF-β), which allow these cells to go ‘multiple ways’ and tailor ensuing immune responses, notably by recruiting or influencing other innate and adaptive immune cells, such as conventional CD4+ and CD8+ T cells, B cells, NK cells, neutrophils and dendritic cells.24 The breadth and kinetics of cytokine production by iNKT cells differ from those of conventional T cells. First, whereas it usually takes days for conventional T cells to mount a sizeable cytokine response, cytokines derived from iNKT cells are detectable in the serum minutes to hours following activation. Second, individual iNKT cells have the ability to secrete cytokines that antagonize each other, such as IFN-γ and IL-4, and that are normally produced by discrete T-cell functional lineages. As such, iNKT cell responses have sometimes been labelled ‘Th0’. Despite this, examination of iNKT cell functions in numerous diseases suggests that these cells are highly versatile in vivo. iNKT cell-derived T helper type 2 (Th2) -related cytokines trigger the onset of asthma,25,26 but may participate in iNKT cell beneficial functions in several autoimmune diseases, such as diabetes.27,28 Similarly, their production of Th1-related cytokines favours the development of atherosclerosis29 and sickle cell disease,30 while protecting from tumour growth.31

This high degree of functional plasticity makes iNKT cells appealing targets for immune-based therapies. Yet, much has to be learned on the mechanisms controlling iNKT cell responses, specifically regarding cytokine secretion. Differences in phenotype, maturation state, tissue of origin, as well as environmental cues (cytokines, chemokines, co-stimulatory molecules) have been shown to influence iNKT cell cytokine secretion.24,32 In addition, distinct lipid antigens may also influence iNKT cell responses. In agreement with this, promising and exciting studies have identified iNKT cell lipid antigens with polarizing potential33–35 (see below).

iNKT cell lipid antigens

iNKT cells recognize a variety of microbe-derived as well as mammalian lipid antigens.36 Most iNKT cell antigens identified so far share an overall common architecture, i.e. a fatty acid moiety, usually a ceramide or diacyglycerol that is buried into the CD1d binding pockets, and a hydrophylic glycan head group that protrudes out of CD1d, available for TCR recognition. The main difference between microbial and mammalian (self-) lipid antigens lies in the linkage between these two components. Microbes produce α-linked glycolipids, which can be considered as a microbial signature, whereas mammals are enzymatically equipped to generate only β-linked anomers. Microbial iNKT cell antigens include α-glycuronosylceramides expressed by Sphingomonas and Ehrlichia bacteria,37,38 diacylglycerol-containing glycolipids from Borrellia burgdorferi39 and Streptococcus species,40 a cholesteryl α-glucoside from Helicobacter pylori,41 the tetramannosylated phosphatidylinositol PIM4 from Mycobacterium bovis BCG42 and a lipophosphoglycan from the parasite Leishmania donovani.43 Whereas some of these compounds only activate small fractions of iNKT cells, most of them have strong adjuvant properties and potently activate iNKT cells. In addition, many synthetic derivatives of α-linked glycolipids have been studied,36 some of which hold great promise in immunotherapies or vaccination, such as α-C-GalCer,34 C:2035 and plakoside A analogues44 (a naturally-occurring GSL with similarities to αGalCer) that produce Th1-biased responses, and OCH933 that is more pro-Th2.

Self-reactivity (or autoreactivity) is at the core of iNKT cell biology and led to the discovery of their restriction element CD1d. In many instances, iNKT cells display some level of activation towards CD1d-expressing antigen-presenting cells in the absence of exogenous antigen. This is reminiscent of their memory-like phenotype and their constitutive expression of cytokine mRNAs.45 On the other hand, some studies reported that peripheral iNKT cells do not perceive ‘tonic’ TCR signals in quiescent conditions,23 and that peripheral expression of CD1d is not required for their proliferation and maintenance.46,47 Nevertheless, iNKT cells have been shown to respond to several self-lipid antigens, including β-linked GSLs such as iGb3,14β-glycosylceramides such as βGalCer and βGlcCer,48,49 and the ganglioside GD3.50 In addition, recent findings suggest that iNKT cell self-lipid antigens may not be restricted to GSLs.51 In agreement with this, the phospholipid lysophosphatidylcholine has been shown to activate human52 but not mouse53 iNKT cells. The discovery of self-lipid antigens is impaired by their overall lower potency compared with α-linked species. This weak antigenicity may result from differential cellular uptake, transport and loading, degradation, as well as physical constraints because of the bulging β-linked sugar head (see below). Identifying agonist self-lipid antigens is fundamental to our understanding of iNKT cell biology.53 Self-lipid antigens presumably drive iNKT cell development in the thymus, although their precise nature remains elusive. Moreover, the same, or similar, self-lipids are believed to trigger their peripheral activation following exposure of APCs to conserved microbial molecules that activate innate receptors, such as Toll-like receptors38,54–58 or the C-type lectin Dectin-1.59 In a recent study, Brigl et al.55 demonstrated that iNKT cells sense a wide array of bacteria in an indirect fashion that probably involves self-lipid antigens, even if these bacteria produce α-linked iNKT cell antigens. Furthermore, self-lipid antigens probably underpin iNKT cell activation in a broad range of non-infectious diseases, such as cancer or autoimmunity.

Structural insight into lipid recognition by iNKT TCRs: the rules of engagement

Interaction between TCR-αβ and peptide–MHC (pMHC) complexes is achieved through six complementarity-determining region (CDR) loops. Whereas the CDR1 and CDR2 loops are germline-encoded by the variable (V) α and β gene segments, the CDR3α/β regions encompass Vα-Jα and Vβ-Dβ-Jβ joints, and are therefore subject to tremendous diversity. Generally, restricted Vα and Vβ usage, as well as particular CDR3 motifs, characterize the specificity of T-cell responses. Whether TCR-αβ interact with pMHC complexes in a versatile versus conserved manner has not reached a consensus among immunologists and structural biologists.60,61 Nonetheless, TCR-αβ usually dock onto pMHC using a broadly conserved solution, whereby the TCR sits perpendicular, in a diagonal (defined by the centre of mass of the TCR) and polarized orientation, relative to MHC, and the CDR1/2 and CDR3 loops position over the MHC α1 and α2 helices (or α1 and β1 for MHC class II) and the peptide, respectively. Within this docking topology, positioning of the TCR on pMHC and CDR residue requirements can vary significantly between complexes. Nevertheless, evidence of conserved residues within the TCR V segments that mediate interaction with MHC molecules starts to emerge. As an example, two tyrosine residues within the Vβ8 CDR2 loop mediate generic interaction of Vβ8-containing TCRs with various MHC alleles.62–64 These tyrosine residues are shared between many Vβ segments in humans, mice, as well as evolutionarily distant organisms such as fish and amphibians.63 However, the generalization of this concept is premature because of the large number of Vα and Vβ pairs (up to approximately 3500) and the high polymorphism of MHC molecules.60

From a structural standpoint, the rules of CD1–lipid recognition by TCR-αβ may appear easier to crack, a priori, because of the quasi-monomorphic nature of CD1 molecules and the overall limited TCR diversity of lipid-reactive T-cell populations. As the crystallized iNKT TCR showed the common architecture of previously reported TCR-αβ, and considering the structural similarity between CD1d and MHC class I, original models predicted a docking solution on CD1d–lipid complexes similar to that of pMHC–TCR interactions.7 The seminal study from Borg et al.65 revealed that lipid recognition by iNKT TCRs occurs in a radically different way. This study demonstrated that the human iNKT TCR docked in a tilted and parallel orientation, at the very edge of the CD1d-binding cleft that contained αGalCer. Specifically, the iNKT TCR used exclusively germline-encoded residues within CDR1α, CDR3α and CDR2β to mediate interaction. The subsequent resolution of over 20 additional ternary complexes, and extensive mutational analysis, revealed that this docking solution is remarkably conserved regardless of (i) the species (mouse versus human), (ii) TCR-β chain composition (Vβ usage or CDR3β composition) or (iii) the lipid antigen associated with CD1d (as discussed later). Hence, iNKT TCRs seem to use a simple one-size-fits-all mode of recognition, reminiscent of a pattern recognition receptor.66,67

Although the similarity of interaction of mouse and human iNKT TCRs could have perhaps been anticipated from the high sequence homology between the invariant TCR-α chains, and from early work that revealed iNKT cell cross-species reactivity in the context of αGalCer, the finding that neither the TCR-β chain nor the lipid antigen involved substantially influenced the positioning of iNKT TCRs onto CD1d–lipid ligands was unexpected. In addition, Uldrich et al.8 demonstrated that TCRs containing the canonical Vα10-Jα50 TCR-α chain also docked similarly on CD1d-αGalCer, despite very little sequence homology with the Vα14-Jα18 TCR-α chain. Together, these findings demonstrate that diverse TCR-αβ can engage CD1d molecules that bear lipid antigens in a strikingly conserved fashion. This suggests that CD1d itself, rather than the TCR or the lipid antigen, may govern this ‘immutable’68 recognition strategy. Whether type 2 NKT TCRs, and other CD1-restricted TCRs are forced by their respective CD1 molecule to adopt a similar position remains an open question.

Role of the TCR-β chain

The basis for differential Vβ usage and CDR3β diversity within the iNKT repertoire is a confounding matter. Although it was suspected to confer iNKT cells with the ability to recognize structurally distinct antigens, it was also well established that most iNKT cell hybridomas (e.g. DN32.D3), which were widely used in the quest for iNKT cell antigens, had a broad spectrum of antigen recognition (including α-linked and β-linked antigens). In the crystal structures, the invariant TCR-α chain dominates the interaction with CD1d and the lipid antigen (65% of the buried surface area in the original structure65), through contacts mediated by CDR1α and CDR3α residues. Most of the TCR Vβ chain contribution comes from the CDR2β loop that solely interacts with the α1 helix of CD1d. Specifically, the two tyrosines and a glutamic acid within CDR2β, that are conserved between Vβ11 and most mouse Vβ8 segments (and somehow conserved in Vβ7), were shown to be critical to the interaction. The Vβ2-containing iNKT TCR, that lacks the conserved tyrosines, docks similarly and still relies on CDR2β residues (the conserved glutamic acid and one arginine) to interact with CD1d.69 In this complex, ‘compensatory’ interactions involving CDR1β and CDR3β residues have been observed. Of note, differential Vβ usage seems to induce slight differences in the Vα-Vβ juxtapositioning, which can lead to distinct interactions between the TCR and the CD1d–lipid complex.69,70 With respect to CDR3β, the structural data available show that this loop does not contact the lipid antigen, and instead positions over, and sometimes contacts, the α2 helix of CD1d. It is therefore unclear how the TCR-β chain (Vβ usage and CDR3β composition) could directly favour preferential antigen recognition. In summary, it appears that despite differences in sequence, iNKT TCRs find ‘ways’ to interact with CD1d in a strikingly conserved fashion. As discussed below, the functional consequences of the fine differences observed within this innate mode of interaction remain unclear.

If the TCR-β chain does not influence the docking topology, and stays away from the antigen, what could then be the rationale for different TCR-β usage? Previous affinity measurements revealed that differing TCR-β chain composition greatly affects CD1d-αGalCer binding affinities and kinetics (recently reviewed in ref. 71). Accordingly, a series of mutational studies of the CDR2β72,73 and CDR3β49,72–74 regions demonstrated a functional cross-talk between these two CDR loops that ultimately influence the avidity of interaction with CD1d tetramers as well as the breadth of antigen recognition. In this scenario, high-affinity iNKT TCRs are able to recognize a large array of lipid antigens, whereas low-affinity iNKT TCRs only respond to the most potent lipids, such as αGalCer. As such, whereas conventional pMHC-reactive αβ T cells display strong antigenic specificity and clonality within the repertoire, iNKT cells appear to gradually broaden their antigen-recognition capability, and gain the ability to recognize weaker lipid antigens as the affinity of their TCR increases (Fig. 1). In agreement with this, engineered as well as naturally occurring high-affinity iNKT TCRs demonstrated the ability to interact with CD1d loaded with a variety of naturally occurring self-lipids, including iGb3 and some phospholipids, while retaining recognition of more potent α-linked GSLs such as αGalCer, OCH and GSL-1’.49,72

Figure 1.

 T-cell receptor-β (TCR-β) chain, TCR affinity and modulation of invariant natural killer T (iNKT) cell responses. TCR-β chain amino acid composition has been shown to modulate the affinity/avidity of CD1d–lipid recognition (depicted is αGalCer). As TCR affinity increases (from the right, lighter green shade, to the left, darker green shade), individual iNKT cell clones gain the ability to recognize and respond to lower potency lipid antigens, such as iGb3, potentially other glycosphingolipids (GSLs) and phospholipids, while retaining the ability to respond to high-potency lipids such as αGalCer or OCH9. Whether the strength of TCR signalling influences the magnitude (upward-pointing arrows) and/or quality (depicted as differing IFN-γ/IL-4 ratios) of cytokines secreted by individual clones, and how this affects the overall iNKT cell response, remain open questions. Note that the ordering of the lipids depicted here constitutes a tentative summary of their relative potencies, measured by different methods, such as immunization, antigen presentation assay, tetramer staining or affinity measurements, based on the available data.

Influence of the lipid on TCR binding

How discrete lipids ultimately influence iNKT cell responses is not understood, and probably relies on complex mechanisms that include differences in half-life (such as differences in β-linked versus α-linked glycolipid degradation), transport and cellular uptake, type of antigen-presenting cell involved, intracellular trafficking, loading efficiency into CD1d, stability within the CD1d-binding cleft and micro-clustering of CD1d within the membrane, such as within lipid rafts. Once exposed at the cell surface, individual lipids can stabilize or reshape the CD1d landscape, and influence TCR binding. Also, persisting CD1d–lipid complexes have the potential to stimulate TCRs repeatedly, while fast decaying antigenic complexes may result in single ‘kiss and run’ events, which may contribute to cytokine skewing.75 Ultimately, how iNKT TCRs read CD1d–lipid complexes may influence iNKT cell activation and cytokine secretion. In just a few years, impressive crystallographic success has widened our knowledge of lipid recognition to a large panel of synthetic or microbe-derived α-linked antigens, including αGalCer65,69,70 and variants thereof,76,77 the C-glycoside α-C-GalCer78 and plakoside A44 that both produce Th1-biased responses, OCH9 and the di-unsaturated (C20:2) N-acyl αGalCer analogue that both produce Th2-biased responses,77αGlcCer77 and the diacylglycerol-containing glycolipids from Borrelia burgdorferi66 and Streptococcus pneumoniae.79 The conclusions of these studies is that the lipid does not influence the positioning of the TCR, emphasizing the pattern-recognition receptor properties of iNKT TCRs, as originally proposed by mutational studies.67 One key finding came for the resolution of the Borrelia burgdorferi diacylglycerol–CD1d structure. In this complex, the TCR repositioned the galactosyl head group in an orientation that resembles αGalCer, to reach the conserved docking mode.66

Self-lipid recognition constitutes a central aspect of iNKT cell biology. Not only do self-lipids drive iNKT cell ontogeny in the thymus, they also probably underpin their peripheral activation in many infectious and non-infectious diseases. Self-lipids include the β-linked GSLs iGb3, βGlcCer and βGalCer. Previous work demonstrated that β-linked head groups adopt a perpendicular and substantially protruding orientation,80–83 especially iGb3 that comprise a tri-hexosyl head group.14 How such drastic structural differences were dealt with to maintain iNKT cell response was unclear. Two recent studies demonstrated that iNKT TCRs achieved recognition of this antigen by remodelling the CD1d–lipid landscape.84,85 In the published structures, the iGb3 trisaccharide head group was pushed away by the TCR and laid flat against the α2 helix of CD1d, so that the TCR could reach CD1d in the conserved docking topology. Moreover, the use of engineered high-affinity TCRs revealed that a similar remodelling of the CD1d–lipid landscape occurs for complexes involving βGalCer, as well as βLacCer and Gb3, two lipids that are not antigenic for iNKT cells.84 In the CD1d-βGalCer-TCR structure, the outward-pointing galactose of βGalCer was bent in a position similar to that of αGalCer.84 Gb3 differs from iGb3 in the linkage of its terminal galactose (α1–4 versus α1–3 for iGb3) and βLacCer lacks the third distal galactose. In the three structures, polar and van der Waals interactions between the head groups and residues of the CD1d α2 helix were observed. The antigenicity of iGb3 could be explained by additional interactions between its most distal galactose and residues of the α2 helix of CD1d that ‘locked’ the complex, interactions that were reduced or impossible for Gb3 and βLacCer, respectively.84 This is in agreement with a previous report that showed that substitutions on the terminal sugar of iGb3 affected its antigenicity.74 In summary, these studies demonstrated that iNKT cell reactivity to self-lipids might arise from an induced-fit mechanism, whereby ligation of the TCR repositions the self-lipids to mimic microbe-derived α-linked lipid antigens. Importantly, this structural gymnastics comes with a substantial energetic cost that explains, in part, the lower potency of β-linked antigens, and is likely to have important functional consequences on iNKT cell functions. This also suggests that some self-lipids (and potentially microbe-derived lipids) with bulky head groups, such as gangliosides, can decrease or prevent interactions between the TCR and CD1d (Fig. 2), which may regulate iNKT cell activation. This is supported by a recent study showing that engineered iNKT TCRs with high built-in affinities for CD1d molecules only weakly bound CD1d tetramers loaded with gangliosides.72 Mouse and human CD1d can be associated with a large array of self-lipids, including phospholipids, glycerophospholipids, lysophospholipids, sphingolipids and GSLs.86–89 Recent evidence suggests that immune challenges (e.g. Toll-like receptor stimulation) most likely induce changes in CD1d lipid repertoires.88,90 The precise regulation of these repertoires in disease, the identification of the lipids involved, and how this relates to iNKT cell activation and function constitute a new and exciting avenue of research.

Figure 2.

 Potential consequences of the ‘one-size-fits-all’ docking topology. Invariant natural killer (iNKT) T-cell receptors (TCRs) recognize CD1d presenting structurally distinct lipids in a conserved manner. Within this conserved docking framework, fine interactions between the antigen and the TCR-α chain can differ (depicted as pink dots), but their functional consequences remain unclear. Although distinct lipids trigger iNKT cell responses of differing magnitudes (αGalCer is a strong agonist, iGb3 a weaker antigen), consensus information linking iNKT cell polarization and antigenic stimulus is lacking. iNKT TCRs have been shown to reshape the CD1d–lipid landscape to reach the conserved interaction mode. As such, the glycan head group of iGb3 (depicted) is pushed away and squashed flat over CD1d. It is possible that for glycosphingolipids (GSLs) (or other glycolipids) with more protruding head groups, this conformational gymnastics is impossible and the TCR cannot reach CD1d (far right, depicted is GT1b). The presentation of such compounds by CD1d molecules may constitute a way to regulate iNKT cell activation.

Do iNKT TCR binding properties modulate iNKT cell responses?

Given that all CD1d–lipid–TCR interactions seemingly occur in an almost identical structural pattern, how can this be reconciled with findings suggesting that iNKT cells can produce distinct quantitative and/or qualitative responses according to the stimuli they perceive? When studying iNKT cell activation, one has to consider cytokine production by individual iNKT cell clones, the entire iNKT cell population, and other immune cells indirectly activated by iNKT cells. All these factors are only imperfectly understood. Less than a decade ago, two models proposed that either the iNKT cell repertoire was composed of discrete populations with differing cytokine secretion capabilities (subset model), or that environmental cues, including the strength of TCR stimulation, could modulate iNKT cell responses (environmental model).91 The high degree of conservation of CD1d–lipid–TCR interactions as well as findings from functional studies49,67,72,73 argue against the subset model. However, it is important to keep in mind that careful examination of these interactions has revealed subtle variations within and around the common docking theme, specifically in the network of interactions between the three partners. As such, outside the core TCR residues necessary for interaction, different TCR-β chains can make additional differing contacts with CD1d69 or influence TCR-α-mediated interactions.70 Also, discrete lipids can modulate the conformation of CD1d and therefore indirectly influence the stability of the complex and TCR binding properties.71 Lipid antigens can also be directly contacted by distinct TCR-α amino acids within or outside the core residues (see Fig. 2). For example, the galactose moiety of αGalCer is contacted by CDR1/3α, the inositol ring of phosphatidylinositol is contacted by CDR1/2α but not CDR3α, and the trisaccharide head group of iGb3 makes interactions with the three CDRα loops. Although these subtle differences do not significantly impact on the positioning of the TCR, they could influence antigen recognition by particular iNKT cell clones, and consequently alter iNKT cell responses. In agreement with this, several studies found that iNKT cells expressing Vβ7-containing TCRs responded better to iGb3 than the ones expressing Vβ8-containing TCRs.92,93 On the other hand, in vitro studies with large libraries of unselected iNKT TCRs containing Vβ8.2 or Vβ7 found opposite results.73 In summary, although the iNKT cell repertoire appears largely devoid of antigen specificity, some iNKT cell clones may be able to bend the rules, and preferentially respond to particular lipid antigens. How this may affect the response of the entire iNKT cell repertoire, such as during infection, remains unknown.

In line with the environmental model, distinct lipids or TCR composition may fine-tune the biochemistry of CD1d–lipid–TCR interactions and consequently influence the magnitude and/or quality of iNKT cell responses (Figs 1 and 2). Although the final outcome of iNKT cell stimulation (e.g. cytokine secretion) is probably influenced by many factors, what happens at the CD1d–lipid–TCR interface is most certainly a contributing factor.32 Interaction between a TCR and its MHC ligand can be described as the rates at which the TCR associates with (ka), and dissociates from (kd), the antigen–MHC complex. These parameters can be accurately measured by surface plasmon resonance (and to some extent by tetramer staining), and TCR affinity can be determined in equilibrium conditions. In pMHC immunity, although cytokines are a major contributor of functional differentiation, the strength of pMHC–TCR interactions seems to correlate with T-cell responses,94 and several lines of evidence suggest that low-affinity interactions (or with fast dissociation rates) favour Th2 differentiation and IL-4 secretion. As iNKT cells acquire the ability to secrete cytokines during their development, and because their activation is generally less dependent on environmental cytokines and co-signals, the biochemistry of CD1d–lipid–TCR interactions may play a more prominent role in the functional response of these cells. As Th1-skewing and Th2-skewing iNKT cell antigens became available, their binding kinetics have been compared.71,77 The pro-Th2 C:20 and OCH9 analogues interacted with the iNKT TCR with similar and lower affinity compared with αGalCer, respectively, yet they both had faster dissociation rates than αGalCer. α-C-GalCer, which produces Th1-biased responses, also displayed lower affinity and shorter dwelling time, compared with αGalCer. Hence, it seems difficult to correlate the strength of CD1d–lipid–TCR interactions with the functional response. However, several issues have to be considered. First, although it is clearly established that these lipid analogues produce skewed cytokine responses following immunization, consensus evidence that immediate iNKT cell responses are polarized is lacking.75 Long-term (over 2 hr) iNKT cell responses are extremely difficult to study because these cells down-regulate their surface receptors rapidly following activation, and can no longer be tracked. Second, because the TCR-β chain modulates the affinity or avidity of interaction with CD1d–lipid complexes, accurate comparison of binding properties can only be made using identical TCRs. Third, solution-based affinity measurements may not fully reflect the in vivo situation. Indeed, it was recently shown that the interaction of membrane-bound TCR and MHC complexes is significantly different from pMHC–TCR interactions in solution.95,96 Finally, the influence of the strength of TCR stimulation on iNKT cell responses has never been tested from the perspective of the TCR, and whether iNKT cell clones expressing TCRs of differing affinities/avidities produce distinct responses upon stimulation remains unknown (Fig. 1). As the cytokine secretion programme of iNKT cells is at least partly imprinted during their thymic development,45 such a polarization could occur during ontogeny, when the TCR is tested during selection events, and/or upon activation in the periphery. New complementary approaches are clearly required to address these fundamental questions of iNKT cell biology.


Lipid antigen recognition by iNKT TCRs is at the forefront of the singular development and functions of these fascinating cells. The recent years have seen important strides in the field of lipid recognition by iNKT cells. Studies have revealed that lipid–CD1d recognition not only differs greatly from pMHC recognition, but is also surprisingly conserved across complexes and through evolution. Within this relatively simple mode of recognition, the lipid involved and the TCR diversity seem to modulate the magnitude of interaction between iNKT TCRs and CD1d. The functional consequences of this invariant mode of recognition prove challenging to comprehend at first glance, and this subject is far from being exhausted. Whether the diversity of iNKT cell responses to distinct antigenic challenges reflects the biochemistry of CD1d–lipid–TCR interactions remains obscure. Understanding iNKT cell responses to distinct antigens, both at the clonal and population levels, constitute important questions to be addressed in the future.


The authors are grateful to Ms Melanie Burger for preparing the figures and to Dr Michele Anderson (Sunnybrook Research Institute and Department of Immunology, University of Toronto) for critical feedback during the preparation of the manuscript. We apologize to those whose works were not cited because of space constraints or omission. This work was supported by a Canadian Institutes of Health Research (CIHR) operating grant (MOP 114911).


The authors declare no conflict of financial interests.