News & Views | Published:


A false sense of non-self

Nature volume 486, pages 479481 (28 June 2012) | Download Citation

A drug used for HIV treatment can alter the set of antigens that activates T cells of the immune system, thereby triggering life-threatening reactions against the body's own proteins. See Letter p.554

The drug abacavir can be an effective treatment for HIV but it is also associated with an autoimmune response in 5–10% of patients1. These hypersensitivity reactions, which include fever, rash and gastrointestinal problems, mandate immediate drug cessation to prevent the emergence of life-threatening symptoms. It has previously been reported2,3 that abacavir hypersensitivity syndrome occurs in patients who express one particular major histocompatibility complex (MHC) class I protein, and genetic screening for this protein variant is an effective way to avoid this adverse drug reaction4. In an elegant study on page 554 of this issue, Illing et al.5 reveal the molecular basis of this response, showing that abacavir alters the repertoire of antigens that activate the immune system.

T cells are immune cells that allow the body to distinguish abnormal or foreign cells from normal cells. The ability of T cells to discriminate between 'self' and 'non-self' is endowed by surface-bound T-cell receptors (TCRs)6. In any mammal, there are millions to billions of T cells, each with a slightly different TCR structure that confers a unique antigen-binding specificity. However, TCRs recognize antigens only when these are bound to MHC receptors on the surface of other cells. These MHC molecules display an array of antigen peptides that provides a snapshot of the cell's internal composition. Aberrant cellular processes, such as viral infection, are reflected by alterations in the antigens that are displayed. When a T cell recognizes a variant peptide (one derived from a viral protein, for example), cellular signalling pathways are initiated that cause the T cell to proliferate and to differentiate. When activated in this manner, T cells called CD8+ T cells eliminate the cells expressing these abnormal antigens7 (Fig. 1a). T cells are able to detect a variant peptide even if just a few copies are present among the hundreds of thousands of normal self peptides that are displayed by the cell-surface MHC receptors8.

Figure 1: Mistaken identity.
Figure 1

a, In the cytoplasm of most human cells, self proteins and foreign proteins, such as those from a virus that has infected a cell, are cleaved into multiple peptides by a complex called the proteasome. A subset of these peptides is then carried by TAP proteins to an organelle called the endoplasmic reticulum. Here, an even more limited set of peptides are loaded onto HLA molecules, which are transported to the cell surface. These surface-displayed HLA–peptide complexes are recognized by T-cell receptors (TCRs) on the surface of immune cells called T cells. Any T cells with TCRs that recognize HLA–self-peptide complexes are destroyed during T-cell development, such that only T cells with TCRs specific for foreign antigens survive. b, Illing et al.5 show that the anti-HIV drug abacavir binds to a specific HLA molecule and alters its peptide-binding characteristics, causing the HLA molecule to load and display a new range of peptides on the cell surface. Because this repertoire includes self peptides that were not displayed during T-cell development, the immune system contains some T cells that can recognize these antigens and launch an immune attack against the cell.

In humans, the MHC molecules are called HLA proteins, and different people express versions encoded by different gene variants (alleles). In general, peptides associate with HLA molecules by inserting parts of their amino-acid residues into a set of six binding pockets (termed A–F) in the HLA9. The structure of these pockets is highly allele-specific, thereby dictating peptide-binding preferences for each HLA molecule. Illing et al. show that abacavir binds directly to the F pocket of one of these proteins, HLA-B*57:01, but that the drug does not bind to the closely related molecule HLA-B*57:03, which is not linked to abacavir hypersensitivity.

Illing and colleagues also reveal that the binding of abacavir to HLA-B*57:01 alters the F pocket's binding preference for side chains of the amino acids tryptophan or phenylalanine, so as to include those of the amino acids isoleucine and leucine. By determining the amino-acid sequences of almost 2,500 HLA-B*57:01-bound peptides, the researchers demonstrated an enormous shift in peptide presentation: 20–25% of the peptides bound by the HLA molecule following drug treatment differed from those that bound beforehand. The authors observed no such drug-induced change in the peptides that bound to HLA-B*57:03.

These data imply that HLA-B*57:01 and abacavir form a complex before the HLA molecules are loaded with peptides inside the cell, thereby altering the pool of self peptides that are bound to the HLA and displayed on the cell surface for CD8+ T-cell recognition. This shift in the specific HLA-associated cell-surface-peptide display leads to the activation of different CD8+ T cells (Fig. 1b). However, the authors detected no obvious skewing towards a single CD8+ T-cell population, or clone, being activated by the altered range of peptides. Instead, they observed responses by many different CD8+ T cells (polyclonality). This is in keeping with a previous study10 that revealed broad-scale activation of CD8+ T cells as the cellular basis of abacavir hypersensitivity reactions.

To examine the structural cause of the change in bound peptides, the authors re-folded the HLA-B*57:01 protein in the presence of abacavir and a self peptide that binds to the HLA molecule only after drug treatment. They then compared the X-ray crystallographic structure of the resulting HLA-B*57:01–self-peptide complex with that of a different HLA-B*57:01–self-peptide complex that forms in the absence of abacavir. The researchers found that part of the drug, the cyclopropyl moiety, protrudes into the HLA molecule's F pocket, reducing the pocket's size, which accounts for its preferred binding of smaller amino acids following drug exposure. The structures also revealed that abacavir binds to the two amino-acid residues that are unique to the HLA-B*57:01 molecule, thereby explaining the drug's allele specificity.

This shift in the bound-peptide repertoire is a plausible explanation for drug-induced hypersensitivity. As T cells develop in the thymus, TCRs are randomly generated by a process of gene rearrangement. A mechanism referred to as negative selection11 ensures that any developing T cell expressing a receptor that strongly recognizes self peptides bound to self-MHC molecules is destroyed, whereas T cells that lack self-reactivity but that have the potential to recognize foreign antigens survive.

Some self peptides are never encountered during T-cell development but can be exposed during pathological conditions10 — when this occurs, a situation of 'mistaken identity' can arise, in which the self peptides are perceived as foreign by the immune system. For example, the eye is an immune privileged site, such that some ocular proteins are hidden from the immune system. However, if these proteins are released into the body — as occurs following a traumatic eye injury — T cells are activated, leading to inflammation in both the injured and uninjured eyes. This condition, referred to as sympathetic ophthalmia, can result in catastrophic blindness. It seems that the autoimmune response associated with abacavir treatment stems from a similar effect — the exposure of self peptides attached to HLA-B*57:01 that are not normally associated with this molecule leads T cells to perceive these self peptides as foreign.

This false sense of non-self is not restricted to abacavir. Illing et al. show that the anti-convulsive drug carbamazepine, which is associated with hypersensitivity responses in patients with the HLA-B*15:02 allele, also binds to that HLA molecule and alters its set of associated peptides. The awareness engendered by this study is likely to result in many more examples of known drug–HLA associations12 being linked to similar autoimmune mechanisms. If so, it may become desirable to screen patients for their HLA alleles before choosing drugs for treatment. By extension, it is possible that organ-specific drug-metabolism processes might limit the altered HLA–peptide display to specific cell types, and thus limit the resulting immune reaction to certain organs, such as the liver, kidneys or bone marrow.

Finally, it is worth considering that environmental toxins and other chemicals could also interact with HLA molecules. The genes that encode HLA are extremely variable — at the last count, there were 4,269 different HLA (-A,-B and -C) variants referred to as MHC class I molecules13. This diversity is believed to have evolved owing to the selection pressures imposed by infectious pathogens14, but it also seems to afford ample opportunity for HLA interactions with drugs and other molecules. Similar drug interactions undoubtedly exist for the other class of HLA molecules (MHC class II, with -DR, -DP and -DQ types) that binds peptides recognized by CD4+ T cells. The field of pharmacogenetics is going to expand and, in so doing, redefine our understanding of the basis of many forms of autoimmunity.


  1. 1.

    et al. Clin. Ther. 23, 1603–1614 (2001).

  2. 2.

    et al. Lancet 359, 727–732 (2002).

  3. 3.

    et al. Lancet 359, 1121–1122 (2002).

  4. 4.

    et al. N. Engl. J. Med. 358, 568–579 (2008).

  5. 5.

    et al. Nature 486, 554–558 (2012).

  6. 6.

    et al. Front. Immunol. 3, 76 (2012).

  7. 7.

    Science 240, 1169–1176 (1988).

  8. 8.

    , , , & Immunity 4, 565–571 (1996).

  9. 9.

    , , & Nature 353, 321–325 (1991).

  10. 10.

    et al. Immunity 28, 822–832 (2008).

  11. 11.

    Immunol. Today 13, 454–458 (1992).

  12. 12.

    & Pharmaceuticals 4, 69–90 (2011).

  13. 13.

    , , , , & Nucleic Acids Res. 39, D1171–D1176 (2011).

  14. 14.

    & Lancet 305, 1406–1409 (1975).

Download references

Author information


  1. Ellis L. Reinherz is in the Harvard Medical School and Dana-Farber Cancer Institute, Boston, Massachusetts 02115, USA.

    • Ellis L. Reinherz


  1. Search for Ellis L. Reinherz in:

Corresponding author

Correspondence to Ellis L. Reinherz.

About this article

Publication history




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.

Newsletter Get the most important science stories of the day, free in your inbox. Sign up for Nature Briefing