Vaccination is the cheapest and most effective means of protecting against certain infectious diseases. Part of this protection comes from a numbers game. In response to disease, the immune system's T lymphocytes are drafted into service: they detect and kill infected cells, and send signals that the corpses need to be removed. But killer T cells that recognize the structures (antigens) characteristic of a given infectious organism are rare in people who have not encountered that pathogen previously. Vaccination can cause the numbers of those T lymphocytes to increase many-fold, and also leads to the formation of a population of long-lived memory T cells, ensuring a rapid response to future infection. A better understanding of this process should aid vaccine development. Writing last week in Cell, Ahmed and colleagues1 proposed that the production of memory killer cells is a gradual process that may take weeks to accomplish, and requires the elimination of antigen from the body.

Killer, or cytotoxic, T lymphocytes are produced from bone-marrow-derived stem cells, which mature into precursor killer cells in the thymus and are then released into the bloodstream and lymph vessels. Each killer T cell carries the CD8 molecule on its surface, and also expresses a receptor that detects foreign antigens. These receptors vary greatly within the cell population, such that perhaps only ten cells in a million have the same antigen specificity.

When 'naive' killer T-cell precursors encounter their cognate antigen — in the form of a pathogen or a vaccine — they multiply rapidly (Fig. 1a). As they do, they acquire the ability to kill antigen-expressing target cells and to secrete cytokine and chemokine proteins that are important in combating infection. They also change their migratory behaviour, allowing them to get out of the blood and lymphoid tissue and into any tissue or organ that may harbour the infection. At the peak of the response, in certain infections, activated killer T cells (effector cells) specific for the pathogen can comprise as much as 50% of the entire CD8-expressing T-cell population.

Figure 1: The T-cell response to a virus that is rapidly cleared from the body.
figure 1

a, The number of T lymphocytes that react with viral antigens increases dramatically for 8 days following infection, producing larger numbers of effector lymphocytes that kill infected cells. T-cell numbers decrease after the effector phase, leaving a stable population of memory cells. b, c, Models for how memory cells might develop. b, A few effector cells might be randomly selected for survival and differentiation into memory cells. c, Alternatively, the differentiated precursors of memory cells may already exist at the effector phase. Ahmed and colleagues' finding1 that a few effector cells gradually convert to long-lived memory cells, even as most effector cells die, supports proposition b.

Full size image

If the infection is cleared, the number of effector cells declines dramatically, and a stable population of memory T cells is left behind, often making up 1% of the CD8-expressing population. These memory cells are more common than are the naive precursors, and can also respond more quickly following re-exposure to the antigen2. The mystery is, how does the immune system 'know' to leave behind an expanded pool of long-lived memory cells? Is it that not all effector cells die when antigen is cleared (and if so, why)? Or do all effector cells in fact die off, with memory cells forming by a separate pathway that splits off early from the path leading to full-blown effectors? Ahmed and co-workers1 aimed to find out which of these models (Fig. 1b, c) is correct.

The authors started by studying the response of mice to a short-lived (acute) infection with lymphocytic choriomeningitis virus, isolating virus-specific killer T lymphocytes at 8 days after infection (effector cells) and at later time points as memory cells developed. They probed these cell populations by analysing gene expression and function, searching for clues to when memory cells appear.

They found that the expression profiles of effector and the later, memory cells overlap, suggesting that memory cells derive from effector cells and are not a separate lineage. The results also support the interpretation that memory cells do not exist at the peak of the effector response, but appear gradually over three weeks after the antigen has gone. The results of the functional and expression analyses — carried out at various time points between the peak of the effector stage and over 40 days after the infection was cleared — imply that the properties of effector cells fade gradually as the hallmark properties of memory cells are acquired. This slow conversion occurs even as the number of antigen-specific cells declines.

Ahmed and colleagues1 also tested the ability of effector and later populations to handle a second encounter with the same pathogen: only the late populations did this effectively. The explanation for the superiority of late memory cells may be provided by the authors' demonstration that only these cells can undergo a second explosive population expansion in response to re-encounter with the antigen. Clearance of the antigen after the first acute infection, resulting in the cessation of signalling from the T-cell antigen receptor, apparently allows a fraction of effector cells to 'reload' to respond to re-infection.

This type of immunological memory, however, probably protects only against acute infections, such as that studied by Ahmed and colleagues, and not against chronic infections by pathogens that find ways to coexist with the immune system. In diseases such as AIDS, tuberculosis and leprosy, the differentiation of this kind of memory population may not be permitted, because antigen is not removed. Control of these infections probably depends instead on the continuous presence of effector cells3.

Recently, it has become apparent that memory T cells exist in two flavours. First, there are central memory cells that circulate through the lymphoid system just like naive T cells specific for the antigen, but in greater numbers and with a more rapid response rate. And second, there are 'effector' or 'tissue' memory T cells that can leave the blood and access all tissues where they might meet antigen (mucosal tissue, for example), before the pathogen has a chance to disseminate4. It remains to be seen whether these memory T cells represent distinct pathways of differentiation from effector cells, or different stages along a single pathway — although another recent report5 suggests the latter possibility, that effector memory T cells convert into central memory cells. And it is still a mystery as to how a fraction of effector cells 'decides' not to die but to become memory cells. What is clear is that the more we learn about the progression of the immune response, the better able we will be to design strategies for effective vaccination. Judging by the present results1, it is clear that 'booster shots' for vaccinations should be timed for the period when the first antigen has been cleared, and a stable population of memory cells has developed.