How a T cell sees sugar

Article metrics

T cells of the immune system recognize lipids, as well as peptides, extending our ideas about such target antigens. A crystal structure of a T-cell receptor docked to antigen shows how a sugar controls lipid recognition.

Most cellular receptors coevolve with their target ligands through repeated interactions over millennia. The immune system has no such luxury. When new infectious agents such as HIV or avian influenza strike a population, immune receptors must reliably respond to foreign molecules — known as antigens — on their first encounter. For the immune cells known as T cells, this remarkable feat is accomplished by scrambling and recombining the peptide sequences that comprise the α- and β-chains of the T-cell receptor (TCR). These chains pair in millions of combinations, generating αβ-TCR heterodimers that can respond rapidly to almost any infecting microbe.

Immunologists used to believe that all T cells recognize peptide antigens that have become bound to proteins of the major histocompatibility complex (MHC I and II). But it is now known that T cells also respond to lipid antigens displayed by CD1 proteins1,2,3. On page 44 of this issue, Borg et al.4 report the crystal structure of an αβ-TCR bound to a glycolipid antigen that is in a complex with a CD1 proteinFootnote 1. These images show at the atomic level how a TCR recognizes the glycolipid's sugar, and provide surprises that confound expert prediction.

The activation of T cells by CD1 and lipids has captured the imagination of immunologists (and physicians) for two main reasons. First, nearly all existing vaccines, adjuvants and immunodiagnostic tests involving T cells are designed to stimulate MHC function and track responses to peptide antigens. But if human T cells commonly recognize CD1 and lipids in disease states, then current immunomodulation strategies are missing many therapeutic targets.

The second, conceptual issue is that T cells responsive to CD1 proteins operate in a grey zone between the two major modes of immune recognition: innate and acquired immunity. CD1-reactive T cells might harness the power of rearranged TCRs — containing many different types of TCR α- and β-chain sequences5 — to recognize almost any lipid molecule bound to CD1, and so adapt the T-cell population after encountering lipid antigens. However, certain T cells, known as invariant NK T cells, are distinguished by their expression of nearly invariant TCR α-chains that recognize glycolipid antigens bound to CD1d (a member of the CD1 family). By sharing a limited receptor repertoire, many NK T cells can be activated in unison, eliciting particularly strong responses that have been shown to alter the outcomes of infectious, autoimmune, allergic, atherosclerotic and neoplastic6 diseases in mouse models.

When lipid antigens for T cells were discovered, one commentator lamented that immunologists, who had only recently become comfortable with peptide chemistry, would next have to master lipids. But the situation was even worse than had been feared, as CD1-presented lipids are composed of lipid anchors conjugated to chemically diverse structures such as sugars, phosphates and peptides. The lipid chains of these antigens are inserted in a hydrophobic groove in CD1, allowing the more hydrophilic conjugated groups to protrude for recognition7. Glycolipids have emerged as major antigens, forcing biologists to consider how T cells recognize carbohydrates — structurally rigid rings of atoms connected by α- or β-linkages that define the orientations of the rings to each other.

The existence of glycolipid antigens makes immunological sense, because microbes synthesize lipids, sugars and peptides that differ markedly from those of vertebrates. Any of their characteristic chemical structures could be used by an organism to discriminate between its own and foreign matter. For example, NK T cells recognize CD1d bound to α-galactosyl ceramide (αGalCer), a glycolipid in which the sugar is in α-linkage with the lipid8. This antigen stimulates a potent immune response in mammals, but is considered 'artificial' because it is found in marine sponges that pose no threat to terrestrial vertebrates. However, certain bacteria also produce α-linked glycolipids9,10, whereas the analogous mammalian glycolipids have sugars attached with β-linkages. Therefore, T cells might recognize α-linkages as markers of infection.

The structures of NK TCRs are known11,12, so a sceptic might wonder what can be learnt from seeing how a known structure docks to CD1d. But Borg and colleagues' report4 provides a wealth of fresh information. For example, the TCR forms contacts with CD1d at certain amino-acid residues that are shared with mouse CD1d, but are absent in human CD1a, CD1b and CD1c proteins. This explains the observation that NK T cells are specific for CD1d and why trans-species crossreactions have been observed.

Antigen motifs recognized by T cells have previously been described according to the position of a given amino-acid residue in the antigen's linear peptide chain. Borg et al. consider instead the point of attachment of groups to a sugar ring, and from this perspective speculate on what kinds of glycolipid other than αGalCer might promote the best TCR–CD1 interaction. Their structure also shows how the α-linked sugar lies down to allow a tight CD1–TCR interface, whereas an upright β-linked sugar might block the approach of the TCR to CD1d.

The TCR footprint on CD1 is perhaps the most pleasing aspect of this study, because it provides a simple explanation for the observation that nearly all NK T cells require the same kind of TCR α-chain to function6. Imagine that the α- and β-chains of TCRs are legs. MHC-reactive TCRs generally adopt a confident, two-footed stance, with both chains contacting their peptide antigen near the centre of the MHC platform. With some variation13, this mode of binding holds true for 15 TCR–MHC structures14 and was predicted to apply also to CD1. But the NK TCR is a ballerina in a pirouette: it leans sideways and has spun clockwise, contacting its antigen only with the α-chain, while using the β-chain to graze CD1 at its edge (Fig. 1). Whether other TCRs dance on one foot is not yet settled, as T cells responding to CD1a, CD1b and CD1c do not seem to share the same type of α-chain for antigen recognition5.

Figure 1: Footprints of the T-cell receptor.

T-cell receptors (TCRs) are composed of α- and β-chains that bind to a composite surface formed by an antigenic molecule (red) in complex with antigen-presenting proteins from the MHC or CD1 families. a, In the classical mechanism of antigen recognition, both the α- and β-chains make contact with the antigen, so that their footprint covers the centre of the MHC platform. b, Borg and colleagues' crystal structure4 shows that a CD1-reactive TCR is rotated clockwise from the classical position and pushed laterally so that only its α-chain makes contact with the antigen. This footprint shows how a TCR can recognize the carbohydrate structure in a glycolipid antigen.

Another notable feature is the displaced location of the TCR β-chain — it is pushed to an extreme lateral position so that it hangs over the edge of CD1d. This lateral approach is reminiscent of a TCR found on γδ T cells15. Perhaps these two unusual footprints will serve as counterpoints to help explain how most MHC proteins guide TCRs to their centres14.

During the long wait for the first TCR–antigen–CD1 crystal, specialists guessed that the TCR would align near the centre of CD1, as occurs with the MHC. This reasonable — but apparently mistaken — prediction did not arise from lack of consistency in MHC footprints, but rather from the faulty assumption that the behaviour of the MHC could be used to predict that of CD1. Such thinking has a chequered past, and dates back to the description of CD1 proteins as MHC-like molecules. This oft-used comparison provides a good introduction to CD1, but is superficial and even misleading. Modern evidence shows that these systems differ in their levels of polymorphism, mutability of antigens, loading mechanisms, trafficking pathways and other features. CD1 is more than just the MHC for lipids, and its footprint provides another reminder that we should think about CD1 biology in its own terms.


  1. 1.

    This article and the paper concerned4 were published online on 20 June 2007.


  1. 1

    Calabi, F. & Milstein, C. Nature 323, 540–543 (1986).

  2. 2

    Porcelli, S., Morita, C. T. & Brenner, M. B. Nature 360, 593–597 (1992).

  3. 3

    Beckman, E. M. et al. Nature 372, 691–694 (1994).

  4. 4

    Borg, N. A. et al. Nature 448, 44–49 (2007).

  5. 5

    Grant, E. P. et al. J. Exp. Med. 189, 195–205 (1999).

  6. 6

    Cui, J. et al. Science 278, 1623–1626 (1997).

  7. 7

    Moody, D. B., Zajonc, D. M. & Wilson, I. A. Nature Rev. Immunol. 5, 387–399 (2005).

  8. 8

    Kawano, T. et al. Science 278, 1626–1629 (1997).

  9. 9

    Mattner, J. et al. Nature 434, 525–529 (2005).

  10. 10

    Kinjo, Y. et al. Nature Immunol. 7, 978–986 (2006).

  11. 11

    Gadola, S. D. et al. J. Exp. Med. 203, 699–710 (2006).

  12. 12

    Kjer-Nielsen, L. et al. J. Exp. Med. 203, 661–673 (2006).

  13. 13

    Hahn, M., Nicholson, M. J., Pyrdol, J. & Wucherpfennig, K. W. Nature Immunol. 6, 490–496 (2005).

  14. 14

    Rudolph, M. G., Stanfield, R. L. & Wilson, I. A. Annu. Rev. Immunol. 24, 419–466 (2006).

  15. 15

    Adams, E. J., Chien, Y. H. & Garcia, K. C. Science 308, 227–231 (2005).

Download references

Author information

Rights and permissions

Reprints and Permissions

About this article

Cite this article

Moody, D. How a T cell sees sugar. Nature 448, 36–37 (2007) doi:10.1038/nature05890

Download citation

Further reading


By submitting a comment you agree to abide by our Terms and Community Guidelines. If you find something abusive or that does not comply with our terms or guidelines please flag it as inappropriate.