Our data demonstrate for the first time in a mouse model that long-term latent MCMV-infection interferes with the induction of protective immunity in aged mice. Latent MCMV-infection additively impaired the poor control of LCMV-infection in old mice and this loss of antiviral protection was linked to reduced expansion of LCMV-specific CD8+
T cells. Moreover, latent MCMV-infection was associated with decreased CD8+
T cell expansions in old mice after infection with VACV but not after VLP-immunisation. Nevertheless, protective efficacy of VLP-immunisation against LCMV-challenge was impaired in old Tx mice with latent MCMV-infection. These data from a variety of infectious and non-infectious experimental models establish a causal role of latent MCMV-infection as a propagating factor for poor immunity in old age contributing to T cell based immune senescence. Our data are strongly corroborated by similar results from Cicin-Sain et al., who found reduced CD8+
T cell responses after superinfection with Influenza virus, West Nile virus (WNV) and Herpes simplex virus Type 1 in old mice with latent MCMV-infection 
At first glance our findings seem to contradict an earlier study demonstrating partial protection against bacterial challenge in mice recently infected with MCMV or murine herpesvirus-68 by activated innate immune responses 
. Although our study primarily focused on the impact of long-lasting latent MCMV-infection on T cell immunity in old mice, where we demonstrate impaired T cell mediated antiviral protection in latently infected mice, we have neither seen any protective effect of MCMV-infection against LCMV or VACV in young mice (; VACV not shown). These discrepancies are most likely explained by differences in challenge systems: while Barton et al. used bacterial pathogens (i.e. Listeria and Yersinia) 
we and Cicin-Sain have applied viral challenges 
. Alternatively, partial protection by activated innate immune mechanisms may have been counteracted by impaired adaptive immunity in young mice with latent MCMV-infection.
Our findings in this experimental mouse model strongly support previous observational human studies suggesting that HCMV-infection may be associated with accelerated immune senescence 
. This notion was largely based on longitudinal cohort studies and cross-sectional epidemiological studies demonstrating a significant association of HCMV-seropositivity with decreased survival of very elderly 
. Moreover, HCMV-infection was associated with reduced immunogenicity of influenza vaccination in elderly 
although this was not confirmed more recently 
. Interestingly, immunogenicity of CpG-adjuvanted VLPs was sufficient to largely overcome the age- and MCMV-associated immune attenuation in our experimental model (). Only very old Tx mice with particularly low naïve T cell counts before VLP-immunisation were not protected against a LCMV-challenge () This may indicate that more immunogenic vaccines than those currently available for influenza could help to improve protective immunity in the subset of elderly with a minimally maintained naïve T cell repertoire.
All of the above mentioned human studies about the impact of naturally acquired HCMV-infection on waning immunity in the elderly are limited by the fact that a variety of factors may strongly influence any significant association. Potential confounders include 1) individual characteristics of HCMV-infection like virus isolate, dose, route, time point and severity of primary infection and the frequency of re-infection or re-activation, 2) genetic variability of the host directing the antiviral immune response, 3) general health and socio-economic situation of the host both presently and in the past (co-morbidities, medication, nutrition, living conditions, physical activity, health seeking behaviour a.s.o.), and 4) the host's past infectious history apart from HCMV. For obvious reasons, it is impossible to control for most of these parameters in human studies. In contrast, we were able to exclude or balance these confounders in our mouse model. Therefore, our data provide the most direct evidence that CMV-infection itself is the driving force for the association between CMV-infection and impaired immunity in aged hosts.
Concerning the major mechanism of CMV-enhanced immune senescence, most experts in the field favour the concept that CMV-driven memory T cell accumulations progressively restrict the size and the diversity of the naïve T cell pool, thus exacerbating the age associated alterations of the T cell compartment 
. Studies in old mice and macaques have clearly established that both size and diversity of the naïve T cell pool are crucial for maintenance of immune protection 
. Indeed, several human studies have shown that naïve CD8+
T cell frequencies in the blood were significantly reduced in HCMV-infected elderly or thymectomised young adults 
but only few studies have reported significant reductions of absolute naïve CD8+
T cell numbers 
. Our data demonstrate that total numbers of naïve CD4+
T cells were not significantly reduced by MCMV-infection in blood, spleen, lymph nodes and lung. This argues against the currently favoured hypothesis that CMV-infection reduces the number of available naïve T cells. At present, we cannot formally exclude the possibility of MCMV-dependent qualitative alterations of the naïve T cell pool, especially concerning repertoire diversity.
Since adoptively transferred CD4+
T cells displayed comparable immune responses in all recipients irrespective of ageing, thymectomy and MCMV-infection, our results are best explained by a deficient activation and/or expansion of the endogenous T cell population (). Indeed, Cicin-Sain et al were able to demonstrate an inhibitory effect of latent MCMV-infection on T cell recruitment and/or activation exclusively in draining lymph nodes after intranasal Influenza infection 
. Our data further suggest that MHC class I and class II dependent antigen presentation was not substantially altered by ageing, thymectomy or latent MCMV-infection to interfere with activation and expansion of tg CD4+
T cells after VACV-infection. This is in line with similar antigen presentation in young and old monkeys after Vaccinia immunisation 
. Although productive MCMV-infection has been shown to severely impair DC function 
our results argue against a major interference of latent MCMV-infection with overall functionality of DCs as an explanation for reduced heterologous antiviral immunity: 1) Latent MCMV-infection of young mice was not associated with any measurable reduction of T cell immunity. 2) Activation, expansion and functionality of adoptively transferred naïve transgenic CD4+
T cells were comparable in all hosts tested, irrespective of MCMV-infection, age and Tx (). 3) Latent MCMV-infection did not significantly impair the primary CD8+
T cell response after VLP-GP33 immunisation in old mice () although protective efficacy was eventially reduced in MCMV-infected old Tx mice (). Moreover, Cicin-Sain et al show that latent MCMV-infection with a mutant virus deficient for critical immune evasive genes was still associated with poor WNV-specific T cell immunity 
. Lastly, it is unlikely that limiting T cell help was a major factor for MCMV-infection associated immune failure in our experimental setup. Provision of additional T cell help by adoptive transfer of tg CD4+
T cells was not sufficient to rescue the poor endogenous CD8+
T cell response in old, Tx and MCMV-infected mice although CD8+
T cell expansion after VACV-GP infection is known to be help dependent 
As opposed to the naïve T cell compartment, there is compelling evidence in mice, macaques and humans that CMV-infection drastically increases the memory T cell pool, particularly regarding CD8+
. Interestingly, those human studies indicating an association of HCMV-infection with poor vaccine immunogenicity or with reduced survival of very elderly, have also reported an association with CD8+
Tem expansions and not with a reduction of naïve T cells 
. Similarly, the correlation of poor T cell expansion after heterologous WNV infection was stronger for increased Tem than for reduced naïve T cells in the MCMV-model 
. Therefore, data from the MCMV-model support the human findings and demonstrate a vigorous expansion of Tem after MCMV-infection which was maintained into old age by memory inflation. MCMV-dependent Tem accumulation caused a significant and long lasting expansion of the total CD8+
T cell pool without restricting the available immunologic ‘space’ for naïve T cells explaining the discrepancy between reduced frequencies but maintained total numbers of naïve T cells after MCMV-infection (, S3A
). This highlights the fact that the CD8+
T cell compartment has considerable plasticity and can grow in size according to the cumulative antigen experience of the host 
Based on these findings, we postulate that CMV-dependent CD8+
Tem expansion is more important for CMV-enhanced immune senescence than naïve T cell suppression. Moreover, we would like to propose a model of enhanced competition for immunologic niches and survival factors between pre-existing CMV-specific memory T cells and newly generated effector T cells after heterologous infection or immunisation (). CMV maintains memory T cells at very high frequencies well into old age, preferentially in the CD8+
T cell compartment as Tem, which persistently occupy critical niches both within secondary lymphoid (i.e. spleen) and peripheral organs. At first glance, the absence of Tem accumulations in lymph nodes of MCMV-infected mice () may contradict a competitive model. However, Torti et al have recently shown that MCMV-specific inflationary T cells maintain a Tcm phenotype in lymph nodes where they interact with antigen presenting cells of non-hematopoietic origin and then egress the lymph nodes probably as Tem 
. Therefore, competition of expanded MCMV-specific memory T cell populations with newly generated effector cells after heterologous infection or immunisation may still take place within, on the way out or even outside lymph nodes. Of note, we find the most robust differences in T cell immunity after systemic LCMV-infection (), where T cell competition may preferentially occur in the spleen and not in lymph nodes. Moreover, LCMV-infection induces a massive inflammatory response which has the potential to trigger MCMV-reactivation. If this occurs expanded CMV-specific memory T cell populations are not only induced by bystander activation but also directly by TCR-mediated signals leading to further memory inflation and potentially to enhanced competition with heterologous T cell responses. However, our hypothesis of enhanced T cell competition needs to be tested in future studies together with a detailed analysis of the naïve T cell repertoire in old mice with and without latent MCMV-infection, since the absolute number of naïve precursor T cells is most likely to be CMV-independent but nevertheless crucial for maintained immunity in aged hosts. This is suggested by our findings of comparable T cell immunity in latently infected young mice, where the T cell repertoire was not yet affected by ageing or Tx, and supported by the comparable expansion of tg T cells irrespective of age, MCMV-infection or Tx, where we artificially increased and normalized the naïve precursor frequency.
Integrative model of the influence of MCMV-infection, ageing and thymectomy on CD8+ T cell populations and protective immunity.
In summary, our results demonstrate that MCMV-infection has a moderate but significant negative influence on antiviral immune responses and protective immunity in old mice. However, Tx at young age and ageing itself seem to restrict immunity more profoundly than MCMV-infection. The involved mechanisms are likely to be different since ageing and Tx operate via the restriction of the naïve T cell pool while latent MCMV infection leaves the naïve T cell pool untouched but mainly propagates memory and total T cell expansions (). Our model of latent MCMV-infection with or without Tx seems to be well suited to further investigate the underlying mechanisms and to develop and test preventive strategies. Finally, our results strongly support previous human studies that have found an association of CMV-infection with declining immunity in old age 
. Since we and others were able to demonstrate a similar association in a very well controlled mouse model 
there is now solid evidence that latent CMV-infection is a propagating factor for immune senescence in mice and in humans.