In elderly people, the ability to respond in a versatile way to pathogens, and potentially to cancer, is reduced due to dysregulated immunity, a state known as immunosenescence. This is characterized firstly by low numbers of naïve T cells in the peripheral blood 7 and probably also in other immune organs such as the lymph node 8, which is robustly observed in different human populations 9. There is a corresponding reduction in the diversity of the naïve T-cell receptor (TCR) repertoire, which helps to explain the decreased ability of the elderly to resist infections to which they were not previously exposed, or to respond to the de novo appearance of tumor antigens. Second, although the number of memory T cells increases, there is a decrease in the diversity and functional integrity of both the CD4⁺ and CD8⁺ T-cell subsets 1011 which contributes to the decreased ability of the elderly to respond adequately to re-infection or to retain memory for cancer antigens expressed by relapsing tumors. Moreover, the ability to respond appropriately to persisting infection or persistent exposure to other antigen sources, such as cancer antigens, may also be compromised, a process sometimes termed "immune exhaustion" 12.

The focus of much research on immunosenescence has been T cells, mainly because antigen presentation by dendritic cells (DCs) generally seems well-retained in the elderly 13, and because appropriate B cell 14 function depends on helper T cells. This is not to imply that non-clonotypic immunity is entirely resistant to age-related changes, or unimportant, but the alterations to natural killer (NK) cells and other innate immune cells 13 seem to be less marked than those seen with T cells. Nonetheless, it has been reported that NK cell status can predict morbidity and mortality in the elderly, emphasizing the importance of innate as well as adaptive immunity in ensuring healthy longevity and possibly cancer resistance 15. In addition, and at least partly because of innate immune activity, the levels of pro-inflammatory factors are commonly increased in the elderly and are associated with negative effects on health 16. Together, these changes contribute to morbidity and mortality, and are associated with poor responses to vaccination against pathogens such as those causing influenza and pneumonia 17, leading to decreased quality of life and increased health-care burdens in later life.

Studying immune cells directly ex vivo, as well as in culture models of ageing in vitro, has contributed to our understanding of age-associated changes in human T-cell-mediated immunity. However, exactly which factors are most important in immunosenescence and what causes these age-associated changes remain largely unclear. Accumulating data initially unexpectedly implicated infection with persistent herpesviruses, particularly cytomegalovirus (CMV), as contributing to the age-associated changes observed in many studies of human immune ageing 1819202122. By analogy, long-term exposure to other persistent stimulating agents, possibly parasite antigens rather than CMV, especially in developing countries, as well as dysregulated responses to autoantigens and cancer antigens, may yield similar effects. Clearly, more detailed knowledge of the reasons for immune dysfunction at advanced age and development of strategies to prevent or reverse this should contribute to maximizing healthy lifespan for the ever-growing numbers of elderly people in industrialized and, increasingly, in industrializing countries.

The cells of the immune system turn over rapidly, so compromised production or function of haematopoietic stem cells (HSCs) would affect all aspects of immunity downstream 23. There is evidence for an association between HSC decline and longevity 24 probably due to differences in DNA damage repair 2526. Purified HSCs from old mice have less ability to reconstitute haematopoiesis 27. It has been recently suggested that the observed age-associated skewing of hematopoiesis away from lymphocyte generation may be one of the factors contributing to increased cancer occurrence with age 28. There is also a gradual decline in the ability of mouse HSCs to progress through the various B-cell differentiation stages, partly reflecting microenvironmental changes. These may involve altered production of interleukin-7 (IL-7) by stromal cells 29. Changes at the level of T-cell progenitors also contribute to age-associated deficits in mice 30. In any event, age is likely to impact both on HSCs and the niche environment that determines their fate (Figure 1).

Although even the very elderly can retain some degree of thymic function 31, thymic involution which represents the next hurdle that a T-cell progenitor must overcome in order to develop into a mature naïve T cell. That this could be clinically important in cancer is demonstrated in a study of age-dependent glioma-associated mortality, in which the numbers of CD8⁺ recent thymic emigrants accounted for the effect of age on clinical outcome 32. However, assessing immune reconstitution after HSC transplantation has demonstrated exceedingly low levels of thymic activity after the age of around 40 years 33. Exactly why thymic involution occurs remains unclear. Because the most marked (but not the only 34 changes to thymic size and structure occur around puberty, it has been suggested that increased levels of sex hormones contribute to causing thymic involution; thymic output of naïve T cells can increase in both mice and humans following androgen ablation 3536. Nevertheless intrinsic defects in the commitment of haematopoietic stem cells to the lymphoid lineage 37 and the migration of T-cell progenitors into the thymus have also been implicated in this process 38. Whatever the reason, it is clear that major age-associated changes to the thymus and greatly decreased output of naïve T cells are observed in both mice and humans (Figure 1). The elderly must therefore increasingly rely on naïve T cells already previously produced and the small numbers possibly still being produced. The presence of cancer, or particularly the treatments applied as cancer therapy, are also likely to further compromise any residual thymic function 39.

In the next stage in T-cell-dependent immunity, naïve T cells must be activated by appropriate contact with antigen-presenting cells (APCs), such as DCs (Figure 2). Although DCs seem to be only subtly different in the elderly in many respects 40, there are quantitative differences, with less peripheral blood 41 and follicular DCs 42. Chemotaxis and phagocytosis may be impaired in DCs, as well as neutrophils 13, from the elderly 43. DCs from young and elderly people are reported to stimulate naïve CD8⁺ T cells equally well, but those from the elderly may fail to stimulate naïve CD4⁺ T cells properly 43, perhaps due to altered signal transduction pathways involving phosphoinositide 3-kinase signalling 44. In cancer, the effect of the tumor on the APC may enhance these deficits 45. Many active immunotherapy protocols for cancer patients rely largely on DCs to be recruited to the site of vaccination and take up the vaccine antigen. Reduced DC numbers as well as their chemotactic and phagocytic activity in the elderly might represent another hurdle to be overcome in developing successful vaccination strategies for elderly cancer patients.

One very good example of the interplay between innate and adaptive immunity is illustrated by γδ-T cells 46, which also show age-associated alterations. These are able to recognize phosphoantigens present on cancer cells and have anti-tumor activities, as well as other functions, such as anti-CMV activity 47. Their numbers rise from birth to puberty and gradually decrease again beyond 30 years of age, and they show functional alterations 4849. A functional analysis in young/adult and middle aged donors revealed that effector/memory γ9/δ2 CD27- T cells are increased after in vitro culture in the presence of specifically stimulatory isopentenylpyrophosphate and IL-2 for 10 days. In contrast, this did not occur when using cells from old subjects, confirming a lack of naive and central memory cells responding to IL-2 5051. In these respects, γδ T cells behave quite similarly to conventional αβ T cells, as discussed in the next section.

Despite decreased thymic output of T cells, in healthy people, the overall number of T cells in the periphery remains remarkably constant at different ages, although there may be an association between slightly lower levels of both CD4⁺ and CD8⁺ cells and mortality at 10-year follow-up 73. Until recently it was impossible to determine which cell subsets were proliferating and whether the rate of expansion was different in people of different ages. Pioneering work in which deuterated glucose is incorporated into dividing lymphocytes to distinguish the fraction of the pool that proliferates during the labelling time is technically challenging but very informative in humans and has now been applied to T cells, B cells and NK cells. In naïve, memory and regulatory T-cell subsets, label uptake and loss occurs rapidly but at different rates in each subset, thus indicating divergent rates of cell division, with regulatory T cells proliferating most rapidly 74. However, the only substantial difference between young and old donors is that the CD8⁺ memory T cells have a much longer half-life than any other subset in the elderly, but not in the young 75. These results are consistent with an accumulation of CD8⁺ T memory cells in the elderly. In B cells, there are also only slight differences between young and old individuals, with memory B cells turning over more rapidly than naïve B cells 76, whereas in NK cells, the production and proliferation rates were lower in the elderly than in the young 77. Nonetheless, B cell memory may be maintained for a very long time; thus, it has been recently shown that individuals who survived the Spanish flu (H1N1) still have specific antibodies 90 years after the epidemic. These donors still possess circulating B cells that secrete antibodies that binding the respective recombinant haemagglutinin protein 78.

Given the high measured levels of peripheral T-cell turnover, it is not surprising that their proliferative potential eventually becomes exhausted; these exhausted cells are characterized by a lack of CD27, CD28 and telomerase expression 79, and by having short telomeres 80. Memory T-cell turnover may not only result in an eventual loss of the cells, but also in an active acquisition of suppressive activity as demonstrated in vitro 81 and more recently in vivo, suggesting that development of regulatory T cells (Tregs) from terminally-differentiated CD4⁺ cells can take place 74. Tumor antigens as the driving force in this process might account for the commonly-observed higher levels of Tregs in cancer patients, which are believed to be a bad prognostic sign, and to interfere with immunotherapy 82. Mechanisms therefor may include competition for antigen on the DC surface, or altered cytokine secretion patterns 59. At least for pathogens establishing only acute infection, antigen seems not to be required for the constant turnover of specific memory T cells. In mice after clearance of acute viral infection, memory T cells persist for the lifetime of the animal and also undergo clonal expansions and accumulate even in the absence of the original antigen 83. Thus, T-cell clonal exhaustion may be an intrinsic characteristic of adaptive immunity, which still occurs even in the absence of chronic infection 84. However, the biological norm is likely to be chronic infection; in humans, mostly herpes viruses, especially CMV and in CMV-negative elderly individuals, Epstein-Barr virus (EBV), may be the main agents driving CD8⁺ T cells to exhaustion 85. In different environments, possibly where EBV and CMV are not so prevalent, other pathogens may play a similar role. Analogous phenomena of clonal exhaustion and senescence may also occur in HIV disease 86 and autoimmunity 8788, and, as we argue here, cancer. Although the proliferative reserve of T cells is high, at least until telomere erosion reaches its critical limit (termed the Hayflick limit), being approximately 70 population doublings for CD4⁺ T cells 70, the old idea that T-cell clonal exhaustion leads to clonal attrition and has clinical consequences has stood the test of time. The clinical consequences may be broad and impact not only on those very advanced in years, as discussed in the next section.

Establishing reliable biomarkers identifying T-cell subsets and determining which are the most relevant to healthy ageing and adequate responses to vaccination requires longitudinal studies following the same individuals over time and correlating test parameters with clinical outcome (such as mortality, pathology and vaccination responses). Cross-sectional studies have the disadvantage that the young and old populations compared will have differed tremendously in their previous exposure to environmental factors of all kinds, and probably, also genetic constitution. However, so far, most human studies have been cross-sectional, and because of logistical constraints we cannot even be sure whether acute infectious complications are really more severe in the elderly due to impaired innate and adaptive immunity or due to other factors. Apparently simple questions such as the correlation of immune status with outcome of vaccination in the elderly cannot yet be answered definitively. Correlations between the presence of certain T-cell subpopulations and poor responses to vaccination against influenza virus have, however, been reported. Excessive CD8⁺CD28^- T cells correlate with poor responsiveness, whereas, counter-intuitively, frequencies of CD4⁺ memory T cells, regardless of CD28 expression, do not 89. In mouse models, the virus is cleared from the lungs more slowly in old than young mice, correlating with a delayed and decreased peak of cytotoxic T-cell production. Thus, cellular responses are crucial for controlling the virus but do not function adequately in old animals; this appears to be likely in humans too 90. In mice, cell transfer experiments demonstrated that it was both the old environment and the old cells that contributed to the problem – that is, even young functional T cells do not respond properly when transferred to an old environment (and neither do old T cells when transferred to a young environment) 91.

The apoptosis-resistant cells that accumulate in both old mice and humans may be dysfunctional in several ways. In young mice, the number of T cells staining with soluble MHC/peptide multimers loaded with influenza-virus epitopes is very similar to the number of T cells producing interferon-γ (IFNγ) on antigen stimulation, indicating that essentially every antigen-specific cell present is able to respond to antigen. However, in old mice, the number of tetramer-positive cells exceeds the number of IFNγ-producers, indicating that a proportion of the antigen-specific cells fails to respond. This is quite similar to the situation in very elderly individuals who accumulate large clonal expansions of CMV-specific T cells, the majority of which are dysfunctional (or anergic). Longitudinal studies of elderly Swedes have established that individuals with the largest accumulations of clonally expanded CD8⁺ T cells, most of which are CMV-specific and have a mature phenotype (CD27^-CD28^-CD57⁺KLRG1⁺) 9293, have a shorter remaining survival time than people of the same age with less of these cells at baseline. In CMV-positive subjects, the number of different clonal T-cell expansions increases from middle age to old age and decreases in the terminal phase of life, resulting in a highly significant correlation between the number of T-cell clones and survival of the individual 11. This finding suggests that eventual loss of control of CMV infection after a lifetime of immunosurveillance may indeed be a cause of mortality in this elderly population (Figure 4). Whether this will be a general finding in other populations remains to be seen. Clonal expansions of CD8⁺ T cells are also found in old mice; whether this is antigen independent (random TCR usage; homeostatic proliferation) or antigen driven may depend on the pathogen environment 94. Thus, although antigen driven, especially CMV-stimulated, T-cell clones may be the type most frequently found in aged humans, this finding suggests that there are multiple, independent mechanisms that can contribute to age-altered CD8⁺ T-cell homeostasis. All these findings suggest that dysfunctional, end-stage differentiated cells may accumulate in various physiological situations 95. To what extent this process is amplified in CMV carriers who are also cancer patients, and not only for CD8⁺ but also CD4⁺ T cells as well as possibly components of innate immunity 1396 must be further investigated. This notion is consistent with the hypothesis that such CD8⁺ cells fill the "immunological space", thereby decreasing immune competence and raising the possibility of intervening to restore appropriate immunological function 9798. As we learn more about the dysfunctional status of these cells, in terms of basic biochemical functions 99100 and receptor signalling 101 we may begin to identify biomarkers that can be used in monitoring therapy success 102103. There is some evidence for the existence of such "exhausted" cells in cancer patients 104105. Whether the accumulation of dysfunctional cells can influence response to therapy, in particular to immunotherapy in cancer patients, remains to be explored; very few approaches to this question have been made so far, but a limited literature exists in animal models.

The extent to which tumors are controlled by immunity is controversial. Nonetheless, there is clear evidence for the immunogenicity of cancer cells and large numbers of cancer immunotherapy trials have shown beyond reasonable doubt that T cells recognizing tumor antigens can be amplified in many patients. The calculated frequency of cells responding to a vaccine antigen (MAGE-3) in melanoma suggests that responding T cells must divide a large number of times to generate the observed number of responding cells and to be effective in patients 106. This extreme requirement for clonal expansion might well be partially responsible for the lack of success of many cancer vaccination trials, due to the phenomenon of replicative senescence in T cells 70.

Limited studies in preclinical animal models have indicated age-associated alterations in tumor immunity and therapeutic responses. Thus, in a mouse breast cancer model, there are effective anti-cancer immune responses in young animals which are mostly mediated by T cells, whereas the (less effective) responses in old animals primarily rely on innate immunity 107. This could be a general finding, because adaptive immunity is highly susceptible to deleterious changes with age, but innate responses tend to be better-preserved 108. In fact, retention of strong inflammatory responses with age, but now in the absence of the counterbalancing and beneficial effects of the adaptive responses that they normally amplify at younger age may even enhance immunopathology and carcinogenesis. This might also help to explain the genetic component of certain cancers in which inflammation is well-established as an amplificatory factor 109. On the other hand, tumors may be less aggressive in the elderly, as seen in breast, lung, colon and other sites in very old people and centenarians, partly because of the lower inflammatory status of "successfully aged" people, which also decreases the ability of the stromal tissue to support tumour growth and angiogenesis 110.

According to these considerations, young and old individuals will most likely require different treatment approaches, which could solve some of the problems of treating geriatric cancer patients with other, more traditional, modalities. Cancer could be less responsive to chemo- and radiotherapy in the elderly, and this might be because the immune system is less effective in the elderly, following the notion that even responses to chemo- and radiotherapy require an intact immune system 111. In the few animal models investigated thus far, cancer immunity successful in young individuals is mostly less effective in older ones. Young mice injected with highly immunogenic tumors such as chemical- or radiation-induced sarcomas controlled the tumor better compared to old mice. In contrast, tumors with weak immunogenicity, like B16 melanoma, were more aggressive and grew better in young animals compared to old (reviewed in 112). Sometimes, decreased efficacy in old mice can be compensated for by increasing costimulation 113114, which is focussing efforts on improving adjuvants for the purpose of better vaccination results in the elderly in general. This may be particularly important given the possible synergistic effects of ageing and cancer on DC function, such that the relative contribution of DC dysfunction may be greater in cancer than in healthy ageing 115. However, as with other aspects of the complex interactions between cancer and ageing, it will be difficult to generalize from one system to another. Thus, it has already been demonstrated that some immunotherapies are actually effective only in old but not young animals 116, implying that patient-individualized therapies will have to take the immunological age of the individual closely into account. For this, it will clearly be necessary to differentiate between chronological age and the "functional" age of immunity, as could be based on the constellations of immune markers making up the "immune risk profile" or the modified set of markers investigated by Hirokawa et al. mentioned above 105.

As discussed above, immunosenescence results from multifactorial processes that act on all components of the immune system and some approaches to reconstitute appropriate function have already been mentioned. Recent results from studies on long-lived non-human primates maintained on a calorie-reduced diet showing improved maintenance and/or production of naïve T cells, better T-cell function and reduced pro-inflammatory status 117, and the possible beneficial effects of moderate exercise 118, serve to emphasize the continued relevance of common-sense medical advice in "anti-ageing" medicine as applied to immunosenescence. Although as ageing proceeds, the T- and to some extent B-cell compartments seem to be particularly susceptible to senescence, NK cells and other components of innate immunity are also affected. As mentioned above, some limited data do point to a crucial impact of NK-cell status on resistance to infectious disease and hence healthy longevity 15, as well as being implicated in the outcome of influenza vaccination 119. It is likely that NK cells also play a role in determining the outcome of cancer immunotherapy 103. Clearly, it is not only the T-cell status and/or CMV status of elderly subjects that affects the outcome of vaccination 120, but also the pre- (and post-) vaccination NK-cell status 119 (although how much the latter is affected by CMV is not yet clear). It is therefore imperative that these studies are extended and repeated in different populations. Nonetheless, the most severe clinical impact is likely to remain primarily the result of loss of TCR and BCR repertoire diversity due to clonal deletions, decreased thymic and bone marrow output, and accumulations of dysfunctional cells. In addition to avoiding over-eating and ensuring that sufficient exercise is taken, what else can be done to reconstitute immunity at old age?