No one is naïve when it comes to infections. The immune system of any individual is shaped by lifelong and continuous exposure to antigens derived from foreign organisms. Those who use experimental models of viral infection tend to ignore this fact, although all experimentalists know that the hygiene of the mouse house can influence challenge experiments dramatically. In this issue of Nature Immunology Chen et al.1 investigated whether infection with one virus can modify the way in which an unrelated virus is subsequently handled by the immune system. The results were unexpected.

Chen et al. first exposed mice intranasally to lymphocytic choriomeningitis virus (LCMV), an RNA virus that naturally infects the mouse. Infection with LCMV results in a vigorous CD8+ T lymphocyte response, which is responsible for early antiviral control through perforin-mediated lysis of infected cells and release of interferon-γ (IFN-γ). After the virus is cleared, large populations of virus-specific T cells persist; these lymphocytes are rapidly effective at controlling a second challenge by the same virus. Once the LCMV-treated mice had cleared the virus, Chen et al. challenged them intranasally with vaccinia virus (VV), a presumably unrelated poxvirus. About 95% of these mice survived the challenge infection with VV, compared to only 40% of naïve control mice. Titers of VV in lung, mediastinal lymph nodes (MLN) and spleen were 50% lower at day 3 and 90–97% lower at day 6 after challenge in LCMV-immune mice compared to naïve controls, particularly if natural killer (NK) cells had been depleted before the challenge. In addition, adoptive transfer of splenocytes from LCMV-immune mice conferred a similar level of protection against intranasal VV challenge. This implies that there is also active recruitment of transferred cells to the site of VV replication. Notably, mice that showed a reduced VV titer had altered lung pathology. Whereas most naïve mice had severe alveolar edema and a mixed infiltration of polymorphonuclear and mononuclear cells, in LCMV-immune mice the lung infiltrate was largely dominated by lymphocytes and macrophages and there was much less alveolar edema.

Chen et al. showed that LCMV-specific memory cytotoxic T lymphocytes (CTLs) convey protection against heterologous virus challenge. Without any antigenic stimulation in vitro, 20% of LCMV-specific CTLs within the lung infiltrates produced IFN-γ 3 days after VV challenge, but these cells had weak LCMV-specific ex vivo cytotoxic activity when tested 6 days after VV challenge. More notably, the protective effects of previous LCMV infection against VV challenge were abolished by the in vivo neutralization of IFN-γ. This IFN-γ secretion was not, however, confined to LCMV-specific cells, because somewhere between 40% and 75% of the activated CD8+ T cells in the lung were non-LCMV–specific. In the range of 26–40% of CD8+ T cells present in the lung were LCMV-specific when the VV challenge was given, and the proportion of LCMV-specific T cells did not change drastically upon VV challenge. It is thus not clear that LCMV-specific T cells were vital to the mechanism with which the immune system modulated the challenge. By analyzing this IFN-γ response for a series of epitopes, Chen et al. showed a shift in the immunodominant specificity of the responding LCMV-specific CTL populations after heterologous VV challenge, indicating that these antiviral cells did have some involvement in the enhanced protection that was seen.

How might exposure to one pathogen lead to enhanced resistance against another, unrelated, infection (Fig. 1)? A simple explanation is that T cell receptors (TCRs) of the first response recognize epitopes from both viruses. Chen et al. suggest that CTL epitopes of VV may cross-react with particular epitopes from LCMV (notably NP205), although the protection was not dependent on a particular peptide. To investigate this, it would be useful to establish whether the cross-reactive peptides act as full, partial or weak agonists in this system and whether a mutant LCMV lacking the NP205 epitope would confer the same level of protection against VV.

Figure 1: Memory CD8+ T cells in heterogolous immunity.
figure 1

Memory CD8+ T cells specific for virus X are confronted with infection by virus Y. Infection with heterologous virus Y triggers cytokine secretion (via target cell infection and activation of innate immune response). Bystander activation mediates memory T cell proliferation via the effector cytokine IL-15 and/or crossreactivity with virus X at the level of TCR recognition. A fraction of anti-X–specific memory CD8+ T cells are activated, proliferate, redistribute between different organs and acquire direct ex vivo effector function, which can be asymmetric for different epitope specificities.

Other possible explanations relate to some form of 'bystander' effect. Non-TCR–mediated bystander activation and proliferation as well as maintenance of memory CTL frequencies are dependent on the effector cytokine interleukin 15 (IL-15)2,3. Two distinct pathways involved in bystander activation can induce IL-15: type I IFN, and IFN-γ in combination with IL-12 and IL-184. Given that many pathogens (including VV) lead to substantial production of IFNs, it seems likely that these infections will induce bystander proliferation of memory CTLs. Where memory CTLs specific for a previous infection are present at very high frequencies (in LCMV, these comprise 20–30% of CD8+ T cells in spleen and lung), either low-level cross-reactivity or cytokine-mediated bystander activation of even a small fraction of these cells will be detectable after heterologous challenge. Such high frequencies of memory CTLs are usually observed in persistent viral infections, which also influence the protective capacity of these cells5. Despite the presence of large numbers of antigen-specific CD8+ T cells in—for example, human cytomegalovirus infection—there is no evidence that these cells protect against new infections. Although heterologous challenge leads to detectable functional activation of a fraction of some antigen-specific T cells, the effect is relatively small compared to that of a challenge with the cognate antigen.

How general is this phenomenon likely to be? LCMV is unusual among viruses in that it induces very strong CD8+ T cell responses that persist at high levels. This is why it is such a successful experimental model. The model is complicated, however. With some virus-host combinations, there is a tendency to persistence6 and the immune response continues to evolve over a period of several weeks: for example, neutralizing antibodies take 6 weeks to emerge. During the memory phase, the overall proportion of CD8+ T cells may increase by about 50% in the spleen and by two- to threefold in peripheral organs, such as the liver (U. Karrer and P. Klenerman, unpublished data). This suggests a profound long-term influence on the entire immune equilibrium of the mouse. It will be revealing to see to what extent the cross-protection shown by Chen et al. is seen in other viral systems, such as influenza, where there is no evidence of viral persistence.

How far can these animal experiments be extrapolated to human disease? The hepatitis G virus (GBV-C) is a persistent flavivirus that is common in patients with HIV but does not cause any significant disease. Unexpectedly, GBV-C infection appears to slow the progression of HIV7,8. Given that HIV is controlled by CD8+ T cells, it is tempting to attribute this to the sort of immunological cross-protection explored by Chen et al., although there are other plausible explanations, including virological interference. Notably, that finding contrasts with the accelerated HIV progression in those who also harbor hepatitis C virus. Indeed, when HIV patients are exposed to superinfecting pathogens, the outcome is often fatal (with the exception of scrub typhus which, oddly, is beneficial)9.

Can bystander activation of T cells be potent enough to activate pre-existing autoreactive T cells and so cause tissue damage and clinical disease? This old concept has been addressed in the past with an experimental transgenic mouse model also involving LCMV and VV. High frequencies of CTLs (90% of CD8+ T cells) specific for a neo-antigen expressed in pancreatic islets were insufficient to cause diabetes when they were activated in a “bystander-like” manner, despite substantial cytotoxic activity ex vivo10. These results suggest that unless bystander-induced CTL activation exceeds a critical threshold, which may be very high, organ damage does not occur.

As the host matures, diverse pathogenic challenges to the immune system could help to shape the composition of the T cell memory pool by causing proliferation1 and “attrition”11. In all these experiments, it is also important to recognize that CTLs can be preferentially sequestered or “compartmentalized”, confounding attempts to measure the total memory CTL pool12.

Overall, the responses reported by Chen et al. are examples of the “fuzziness” evident in the logic of the immune system. Although the results these researchers found in manipulating the immune response do not represent classical memory, they resemble “memory-like effects”, which may become apparent when multiple antigens are encountered sequentially. Perhaps what these antiviral T cells experience is not so much total recall as a case of déjà vu.