Humans—even immunologists—like to make sense of the world's complexity by establishing simple, and if possible, binary categories. For example, the discovery that the processing of extracellular antigens for presentation by major histocompatibility class II (MHC II) molecules takes place in endosomal compartments1, whereas antigens presented by MHC I molecules are produced by the cells themselves and processed in the cytosol2,3, led to the establishment of a simple, now 30-year-old paradigm: the antigens presented by MHC I are endogenous (synthesized by the cells performing the presentation) and those presented by MHC II are exogenous (synthesized by other cells). This is a rule stated in most immunology textbooks and frequently reiterated in the introductions of primary research papers on antigen presentation. But it is a gross simplification. Cross-presentation—the process whereby exogenous antigens are presented via MHC I—has long been known to represent an important departure from the paradigm4, but it took three decades for its physiological role to be widely accepted5. In this issue of Nature Medicine, Miller et al. present data indicating that, at least in the mouse influenza model studied, induction of robust CD4+ T cell responses requires MHC II presentation of endogenous antigens produced by infected APCs6. This paradigm shift has important implications for our understanding of antiviral immunity and vaccine development.

The confusion about the origins of antigens presented via MHC I as compared to II is partly semantic. The most accurate criterion to categorize the type of antigens presented by each molecule is the site of peptide production: the cytosol for MHC I and endosomal compartments for MHC II. The problem is that the term 'cytosolic' is often used interchangeably with 'endogenous' and the term 'endosomal' as synonymous with 'exogenous'. It is true that, except in the few cell types capable of cross-presentation, the vast majority of cytosolic proteins are endogenous. However, the content of endosomal compartments is always both endogenous (membrane proteins, endosomal components, cytosolic proteins delivered to endosomes by autophagy) and exogenous (endocytosed from the extracellular environment). Because the proteases that produce MHC II ligands cannot distinguish exogenous from endogenous proteins, MHC II molecules end up presenting peptides derived from both7. However, this conflation of endosomal with exogenous helped cement a common misconception in the field: that the APCs that activate CD4+ T cells by presenting antigens via MHC II—namely, dendritic cells (DCs)—must have obtained this antigen from an exogenous source. In the case of viral infection, this exogenous source would correspond to virions or to cells infected with the virus (Fig. 1). In this scheme, the location of the viral antigen within the infected cell is irrelevant; what matters is that the antigen has been produced by a different cell from the one performing the presentation.

Figure 1: Influenza A virus (IAV) infects lung epithelial cells, which release virions and/or die, generating cell fragments that harbor virus antigens.
figure 1

(a) Dendritic cells (DCs) can capture these exogenous antigens by endocytosis (left). The DCs process the antigens in endosomal compartments and present them via MHC II to CD4+ T cells. The repertoire of viral peptides presented, and the magnitude of presentation, are relatively small, so the antiviral CD4+ T cell response elicited is limited. (b) IAV can also infect DCs, which produce viral proteins. These endogenous antigens access endosomal compartments by endocytosis (membrane proteins) or autophagy (cytosolic proteins) and are also presented via MHC II to CD4+ T cells. The viral peptides presented by this pathway are more diverse and their presentation is more efficient, leading to more vigorous antiviral CD4+ T cell responses. MHC II presentation of endogenous viral antigens by the infected DCs may involve accessory molecules that are not used by DCs performing presentation of exogenous antigens.

Miller et al. put this assumption to the test using mouse models in which the influenza A virus could infect either only non-DCs or both DCs and non-DCs. Their conclusion is that only when DCs are infected, enabling presentation of viral antigens produced by the DCs themselves, does an efficient CD4+ T cell response develop6. Next they explored the role of known components of the MHC II presentation machinery8 in the presentation of endogenous viral antigens in DCs. Regardless of the mechanism used to access endosomes, it is assumed that the conversion of proteins into peptides capable of binding to MHC II will be carried out by proteases and other enzymes located in the endocytic route. For example, the endosomal reductase GILT would be expected to assist proteolysis by breaking disulfide bonds contained in the antigen. Binding of the peptide antigens to MHC II molecules should be dependent on the endosomal chaperone H-2M. However, Miller et al. found that the presentation of most viral peptides derived from endogenous sources did not require H-2M or GILT. Even more surprisingly, presentation of some peptides required the proteasome and the transporter TAP, molecules involved in MHC I, not MHC II, presentation. A role for MHC I presentation machinery in MHC II presentation is not without precedent9, but its contribution in the setting of infection is not clear; additional work is needed to clarify the precise role of MHC I machinery in this process.

The conclusions of this paper raise important considerations for future studies. For example, when assessing the role of different APCs in priming CD4+ T cells against viruses, it will be important to take into account the susceptibility of each APC subtype to infection. Specializations amongst APC may also be based on differential capacity to present endogenous viral antigens via MHC II10. These new results also point to potential strategies that might be exploited by viruses to impair the induction of neutralizing antibody responses, which require B cells to present antigens to the same helper CD4+ T cells that were primed by DCs11. If the viral antigens are presented only by the endogenous route, and the offending virus infects DCs but not B cells, the helper CD4+ T cells and B cells would not interact productively, and antibody production would be blunted.

Finally, there is a practical and rather unsettling implication of the conclusions of Miller et al.6: 'dead' or inactivated viruses used in vaccine preparations may not be capable of inciting robust and protective virus-specific CD4+ T cell responses because, unlike live viruses, inactivated viruses will not productively infect DCs and produce endogenous antigens. To what extent are these implications correct? To how many and which viruses do they apply? These are questions that viral immunologists will now have to address by seeking the truth not only outside but also inside the APCs.