Major features of immunesenescence, including reduced thymic output, are ameliorated by high levels of physical activity in adulthood

Summary It is widely accepted that aging is accompanied by remodelling of the immune system including thymic atrophy and increased frequency of senescent T cells, leading to immune compromise. However, physical activity, which influences immunity but declines dramatically with age, is not considered in this literature. We assessed immune profiles in 125 adults (55–79 years) who had maintained a high level of physical activity (cycling) for much of their adult lives, 75 age‐matched older adults and 55 young adults not involved in regular exercise. The frequency of naïve T cells and recent thymic emigrants (RTE) were both higher in cyclists compared with inactive elders, and RTE frequency in cyclists was no different to young adults. Compared with their less active counterparts, the cyclists had significantly higher serum levels of the thymoprotective cytokine IL‐7 and lower IL‐6, which promotes thymic atrophy. Cyclists also showed additional evidence of reduced immunesenescence, namely lower Th17 polarization and higher B regulatory cell frequency than inactive elders. Physical activity did not protect against all aspects of immunesenescence: CD28−ve CD57+ve senescent CD8 T‐cell frequency did not differ between cyclists and inactive elders. We conclude that many features of immunesenescence may be driven by reduced physical activity with age.

proliferation and retention in secondary lymphoid tissues (Thome et al., 2016). Other hallmarks of T-cell immunesenescence include the following: accumulation of CD28 Àve CD57 +ve T cells with shortened telomeres and reduced proliferative capacity (Di Mitri et al., 2011), which also acquire NK cell receptors such as KLRG1 (Weng, Akbar & Goronzy, 2009) increasing the risk of autoimmune responses, skewing of T-cell responses towards Th17 cell differentiation (Ouyang et al., 2011). Although less well evidenced, some studies also suggest altered regulatory capacity with age with older adults showing increased T reg frequency (reviewed in Jagger, Shimojima, Goronzy & Weyand, 2014) but lower frequency and IL-10 production by CD19 +ve CD24 hi CD38 hi suppressive B cells (Duggal, Upton, Phillips, Sapey & Lord, 2013). Another feature of human aging is an increase in circulating levels of pro-inflammatory cytokines (IL-1b, IL-6, TNFa,) termed Inflammaging, which several population-level studies have associated with increased risk of age-related disease and mortality (Franceschi et al., 2007).
What confounds these human studies is that physical activity is not taken into account in either cross-sectional or longitudinal studies of immune aging. The majority of older adults are largely sedentary and fail to meet the recommended guidelines for physical activity of 150 min of aerobic exercise per week. Regular physical activity in older adults has been associated with lower levels of proinflammatory cytokines such as IL-6, TNFa (Gleeson et al., 2011), improved neutrophil chemotaxis (Bartlett et al., 2016) and NK cell cytotoxicity (Woods et al., 1999), increased T-cell proliferation (Woods et al., 1999) and improved vaccination responses (Kohut et al., 2005). Thus, the current literature on immunesenescence is not able to determine which aspects of age-related immune change are driven by extrinsic factors and which may be the consequence of a constitutive aging programme.
Here, we studied several aspects of the adaptive immune system in highly physically active, nonelite older individuals (master cyclists) in which we have shown the maintenance of a range of physiological functions previously reported to decline with age (Pollock et al., 2015). We show that compared with more sedentary older adults, the cyclists show reduced evidence of a decline in thymic output, inflammaging and increased Th17 cell responses, although accumulation of senescent T cells still occurred. We reveal high serum levels of IL-7 and IL-15 and low IL-6, which would together provide a thymoprotective environment (Lynch et al., 2009) and also help to maintain na€ ıve T cells in the periphery (Wallace et al., 2006). We conclude that maintained physical activity into middle and old age protects against many aspects of immune aging which are in large part lifestyle driven.

| RESULTS
Immune cell phenotype was determined in peripheral blood mononuclear cells (PBMC). The sample size in each analysis varies slightly due to cell availability varying between donors and does not reflect the removal of any outlier data.
2.1 | The effect of long-term physical activity on

T-cell subset distribution
On comparing total T-cell frequency in the PBMC fraction between healthy young donors, healthy old sedentary donors and old master cyclists, significant differences were observed, F(2, 244) = 11.64, p < .001, b = .08. CD3 +ve , T-cell frequency was lower in healthy old sedentary adults compared with young donors, p < .005. This decline was not seen in the master cyclists as their T-cell frequency was higher than the inactive elders, p = .0003 ( Figure 1a) and not different from the young adults.
The frequency of na€ ıve and memory CD8 T cells between the three groups also differed. For na€ ıve CD8 T cells, F(2, 244) = 91.48, p = .0002, b = .42, we found a lower frequency in healthy old sedentary adults compared with young donors (p = .0001) indicative of the well-documented reduction in thymic output with age (Mitchell et al., 2010;Palmer, 2013). However, na€ ıve CD8 T cell frequency did not decline to the same extent in master cyclists being significantly higher than healthy old sedentary adults (p = .001), but lower than that seen in young adults, p = .001 ( Figure 1b). The frequency of memory CD8 T cells was also different between the groups F(2, 244) = 90.65, p < .001, b = .42, due to a higher frequency in healthy old sedentary adults in comparison with young adults, p < .0001 and master cyclists, p = .0002 ( Figure 1c). However, the cyclists did have a higher frequency of these cells than young subjects, p < .0001, suggesting that maintained physical activity did not entirely prevent the expansion of memory cells.
We similarly compared the distribution of na€ ıve and memory CD4 T cells and memory cell subsets between the groups, and here, the differences were more marked than for CD8 cells. The frequency of na€ ıve CD4 T cells varied across the groups, F(2, 244) = 16.43, p < .0001, b = .11, with lower values in healthy old sedentary adults in comparison with young donors, p = .0001. This did not occur in the cyclists who had a higher frequency of na€ ıve cells than the sedentary older adults, p = .01, but a lower frequency than young subjects, p = .005 ( Figure 1d). There was a higher frequency of memory CD4 T cells across the groups, F(2, 244) = 15.69, p < .0001, b = .11, with frequency higher in old sedentary adults compared with young donors, p = .0001 ( Figure 1e). The cyclists had a higher frequency of CD4 memory T cells than the young subjects, p < .05, but the frequency was lower than for the sedentary older adults, p = .003.
Further analysis of the distribution of memory T-cell phenotypes revealed that central memory CD8 T cells were at a higher frequency in old sedentary adults compared with young adults, p = .0001 only, with no significant differences with the master cyclists, p = .29. The frequency in the cyclists was also higher than the young adults, p < .001 (Figure 2a). For effector memory CD8 T cells, there was a significant increase in frequency in both old sedentary adults and cyclists in comparison with young donors, p = .0001 and p = .0001, respectively ( Figure 2b). The differences seen between the groups in frequency of EMRA CD8 T cells were driven by a higher frequency in old sedentary adults in comparison with young donors, p = .0001 and master cyclists p = .0001, although EMRA frequency was raised in cyclists compared with young subjects, p < .05 ( Figure 2c). There were no differences in the CD4 effector ( Figure 2d) and central memory ( Figure 2e) subsets between the groups although CD4 EMRA cells were raised in the sedentary old group compared with young adults, p < .001 and the cyclists, p < .005 (Figure 2f).

| Physical activity and senescent T cells
Despite expressing telomerase to stem telomere attrition and support extensive proliferation during antigen challenge, T cells have a finite replicative potential and repeated cell division results in shortened telomeres and cell senescence (Fletcher et al., 2005).
We therefore determined whether a physically active lifestyle could modify the accumulation of senescent T cells with age.
CD28 and CD57 are among the cell surface markers that used to identify senescent T cells in humans (Onyema et al., 2012). On examining CD28 Àve CD57 +ve CD4 T cells, a significantly higher frequency was observed in old sedentary adults compared with young adults, p = .01, although cyclists did not show a raised level of these cells, p = .47 (Figure 2g). A much higher frequency of CD28 Àve CD57 +ve CD8 T cells was observed compared to senescent CD4 cells overall, and in both old sedentary adults and master cyclists, the frequency of senescent cells was higher compared with young adults and master cyclists, p = .005 and p = .0001, respectively ( Figure 2h).
2.3 | The effect of long-term physical activity on B-cell subset distribution B cells can be divided into four subsets on the basis of CD27 and IgD expression (Figure 3a): na€ ıve (CD27 Àve IgD +ve ); switched memory (CD27 +ve IgD Àve ) and unswitched memory (CD27 +ve IgD +ve ) B cells (Shi, Agematsu, Ochs & Sugane, 2003). The CD27 Àve IgD Àve subset is also a memory cell population (Wei et al., 2007), although much less is known of its functional significance. Comparing the peripheral B-cell frequency between these groups, significant differ- F I G U R E 1 The impact of maintained physical activity on T-cell subset distribution. Immunostaining of PBMCs shows (a) the percentage of CD3 +ve T cells. Further analysis of the CD3 +ve T cell population shows the frequency of (b) na€ ıve CD8 +ve CD45RA +ve T cells and (c) memory CD8 +ve CD45RA Àve T cells in the CD8 T cell pool, (d) na€ ıve CD4 +ve CD45RA +ve T cells and (e) memory CD4 +ve CD45RA Àve T cells in the CD4 pool. The solid bar represents the mean value. *p < .05, **p < .005, ***p < .001 F I G U R E 2 The impact of maintained physical activity on memory T-cell subsets. Immunostaining of PBMCs shows (a) CD8 central memory (CM) characterized as CD45RA Àve CCR7 +ve , (b) CD8 effector memory (EM) as CD45RA Àve CCR7 Àve , (c) CD8 EMRA as CD45RA +ve CCR7 -ve , (d) CD4 central memory (CM) characterized as CD45RA Àve CCR7 +ve , (e) CD4 effector memory (EM) as CD45RA Àve CCR7 Àve , (f) CD4 EMRA as CD45RA +ve CCR7 -ve , (g) CD28 Àve CD57 +ve CD4 T cells in the CD4 pool, (h) CD28 Àve CD57 +ve CD8 T cells in the CD8 pool, in healthy young donors (n = 55), healthy sedentary old donors (n = 75) and master cyclists (n = 118). The solid bar represents the mean value. *p < .05, **p < .005, ***p < .001 reveal for the first time that the effect of age on na€ ıve T-cell populations was partially prevented in those older adults who had maintained high levels of physical activity through adult life.

| Maintained physical activity in adulthood and thymic output
The higher frequency of naive T cells in the master cyclists suggests that thymic output may be better preserved in these adults than their Immunostaining of PBMCs shows the frequency of CD19 +ve B cells. Further analysis of the B-cell population shows the frequency of (c) na€ ıve IgD +ve CD27 Àve CD19 +ve B cells, (d) switched IgD Àve CD27 +ve CD19 +ve B cells, (e) unswitched IgD +ve CD27 +ve CD19 +ve B cells and (f) IgD Àve CD27 Àve CD19 +ve B cells in the B-cell pool, in healthy young donors (n = 55), sedentary healthy old donors (n = 75) and master cyclists (n = 108). The solid bar represents the mean value. *p < .05, **p < .005, ***p < .001 Douek, Koup & Picker, 2000). The frequency of CD103 +ve na€ ıve CD8 T cells differed between our groups, F(2, 240) = 18.55, p = .0001, b = .15, with a higher frequency in master cyclists in comparison with old sedentary adults, p = .0001, but lower than young adults, p = .0017 ( Figure 4b).
We next wanted to determine whether the positive effects of maintained physical activity on thymic output were due to an effect on progenitor cells entering the thymus. CD3 Àve CD19 Àve CD14 Àve CD34 +ve CD10 +ve cells have been identified as thymic immigrants (Bender et al., 1991), and their frequency differed between our groups, F(2, 235) = 5.25, p = .006, b = .04. This difference was driven by a higher frequency in master cyclists compared with old sedentary adults, p = .01, although the frequency in cyclists was lower than in young subjects, p = .01 ( Figure 4c).
These data suggest a beneficial effect of long-term physical activity on thymic output. We thus explored factors that might be contributing towards a maintained thymic output in the cyclist cohort. On measuring serum levels of a thymosuppressive cytokine IL-6 (Sempowski et al., 2000), significant differences were seen between the three groups, F(2, 224) = 12.24, p = .0001, b = .09, with higher IL-6 levels in old sedentary adults compared with both young donors, p = .0001 and old cyclists, p = .0001. IL-6 levels in the cyclists were only slightly higher than in young subjects, p = .04 ( Figure 5a). Cortisol also affects thymic cellularity negatively (Roggero et al., 2006). Analysis of serum cortisol levels in the three groups revealed significant differences F(2, 241) = 12.79, p = .0001, b = .09, driven by higher levels in sedentary older adults in comparison with young donors, p = .002.
However, the master cyclists also had higher cortisol levels than the young adults, p = .0001 ( Figure 5b).
Keratinocyte growth factor (KGF) is another protein important in thymus organogenesis and maintenance, and KGF administration in aged mice increases thymic size and thymocyte production (Alpdogan et al., 2006). No significant differences were seen in serum KGF levels between the groups, F(2, 244) = .80, p = .44, b = .007; (

| Regulatory T and B cells
Foxp3-expressing CD25 +ve CD4 T cells have been classified as regulatory T cells, T regs (Hori & Sakaguchi, 2003). The frequency of these cells differed between the three groups F(2, 241) = 11.93, p < .001,

| DISCUSSION
Remodelling of the immune system with age in humans has been assumed to be an inherent process characterized by immunodeficiency, low-grade chronic systemic inflammation and increased risk of autoimmunity. However, the studies that have generated this assumption are confounded by not considering the impact of physical inactivity, which influences immunity and increases dramatically with age in humans. Even studies in animal models such as mice will be influenced, unless the mice are provided with the option to exercise, as laboratory-housed animals are likely more sedentary than in the wild. This study aimed at determining to what extent the phenomenon of immunesenescence may be a consequence of our modern sedentary lifestyles and is an extrinsically driven process rather than an endogenous developmental programme.
We show here that the effect of maintained high-level physical activity on T-cell subset distribution is profound. One of the most striking and unexpected findings was the high frequency of CD4 and CD8 RTEs in the master cyclists, being higher than in the healthy but more sedentary older adults who had not been involved in a high level of physical activity in adulthood and equivalent to that seen in our young cohort in the case of CD4 RTEs. The thymus has been shown to be responsive to acute exercise interventions previously, with acute exercise inducing an increase in the na€ ıve: memory T-cell ratio in both aged mice (Woods, Ceddia, Zack, Lowder & Lu, 2003) and humans (Simpson, 2011). However, these studies did not consider RTEs and it is possible that these changes, at least in the humans, were driven by a reduction in lymphocyte margination.
Here, the subjects had not exercised in the previous 24 hr, and thus, shown to enhance thymopoiesis and increase export of RTEs. Importantly, recent studies have shown that IL-7 is produced by human skeletal muscle (Haugen et al., 2010), providing a possible explanation for raised levels in the cyclists who did not lose muscle mass with age (Pollock et al., 2015). Thymosuppressive factors, such as leukaemia inhibitory factor, oncostatin M and IL-6, have been reported to be elevated in the aged thymus (Sempowski et al., 2000). Here, we found raised levels of serum IL-7 and lower levels of IL-6 in our cyclists, which would all support thymic maintenance.
Furthermore, IL-15 plays a role in regulating immune homeostasis, acting as a lymphocyte survival factor especially for na€ ıve T cells (Wallace et al., 2006) and is also an exercise responsive hormone (Pedersen & Febbraio, 2012). Our cohort of cyclists had higher IL-15 levels than sedentary old participants and young adults, and we observed a significant although modest positive correlation between serum IL-15 and peripheral na€ ıve CD4 T cells, suggesting that IL-15 is another contributing factor towards the maintenance of the na€ ıve cell pool in the cyclists.
Hematopoietic stem cells show an age-associated skewed differentiation towards myeloid progenitors (Chambers, 2007) and studies in aged mice have reported fewer numbers of thymocyte progenitor cells compared to young mice (Min, Montecino-Rodriguez & Dorshkind, 2004). Although not as well defined as in mice a population of potential thymic progenitors has been identified in humans (Bender et al., 1991), and we show here an age-associated reduction in their frequency in the circulation, which can also be prevented with long-term physical activity. The high level of RTEs in our cyclists is thus likely to be a result of both an increased number of thymocyte progenitor cells entering the thymus as well as an improved thymic microenvironment to promote their development and release.
Aging is also accompanied by an increase in the frequency of memory T cells and accumulation of senescent CD28 Àve CD57 +ve T cells, with the effects most marked in the CD8 pool (Fletcher et al., 2005). In our study, long-term physical activity in adulthood, with accompanying maintained thymic output, was not able to prevent the increase in the frequency of CD8 memory cells, although the increase was lower than for sedentary adults, or the accumulation of senescent CD28 Àve , CD57 +ve T cells which was the same in the master cyclists and inactive elders. One cross-sectional study has reported lower senescent T cells in physically active elders, but in that study, 52-to 61-year-olds were considered as old (Spielmann et al., 2011) which might partially explain the difference in our findings. As the accumulation of memory cells and differentiation of T cells towards the CD28 Àve CD57 +ve cells are driven largely by antigen exposure, including to latent viruses such as cytomegalovirus (Fletcher et al., 2005), it is perhaps not surprising that these immune cell changes were not counteracted by the active lifestyle of the cyclists. Indeed, analysis of CMV seropositivity in our cyclists found that 51% were CMV-positive, which is only slightly lower than the incidence reported for adults in this age range in the United Kingdom (Vyse, Hesketh & Pebody, 2009). The analysis of the memory cell subsets revealed that the large increase in CD8 EMRA cell frequency seen in the sedentary older adults was less marked in the cyclists, despite CD28 Àve CD57 +ve cells being a component of this subset. Further analysis will be required to identify other elements of the EMRA subset which are increased in the sedentary elders but not the cyclists.
Aging is also associated with an increased potential to differentiate to a Th17 phenotype (Ouyang et al., 2011), but little is known of the effects of physical activity on Th17 cell differentiation. We found a significant decline in Th17 cells in the master cyclists and as IL-6 plays an important role in inducing Th17 cells (Weaver, Hatton, Mangan & Harrington, 2007), the reduced serum IL-6 in these adults may underlie the reduced Th17 skewing. Regulatory T cells play a central role in maintaining the delicate balance between pathogenic and protective immune responses. IL-10 secretion is one mechanism used by T regs to mediate immune suppression. An age-associated increase in numbers of T regs has been reported previously (Gregg et al., 2005) and was observed here in the sedentary elders, but was ameliorated in the cyclists. Wilson et al. have reported an increase in circulating T regs after an acute bout of intense exercise in athletes (Wilson, Zaldivar, Schwindt, Wang-Rodriguez & Cooper, 2009); however, this could have been mediated by reduced margination in response to adrenaline. It has been proposed that the increased number of T regs with age will favour development of Th17 responses because regulatory T cells consume IL-2 that inhibits Th17 development (O'Garra & Vieira, 2004). In our master cyclists, we saw lower peripheral T regs that might in turn contribute towards the reduced Th17 polarization. The accumulation of senescent T cells, Th17 polarization, elevated levels of pro-inflammatory cytokines and impairments in B regs have all been identified as factors that might be contributing towards the age-associated increase in risk of inflammatory autoimmune disease (Duggal et al., 2013;Goronzy & Weyand, 2012), we would thus predict reduced risk of such conditions with habitual physical activity.
The study has some limitations. The immune phenotyping data are for cell frequency only, and as lymphocyte numbers fall with age, it is possible that some of the differences between the young subjects and cyclists may be less marked for cell numbers. However, total T-cell frequency did not decline in the cyclists suggesting this shift away from lymphoid cell generation with age is also ameliorated by physical activity. If lymphocyte numbers overall did decline in both older subject groups, this would not affect the differences between the cyclists and their inactive peers and these differences could be even more marked. Also, we do not have detailed historical physical activity data for the inactive older cohort to complement the cyclists data (Pollock et al., 2015), but we do know that they have not been involved in a high level of sports such as cycling through their adult lives and thus are a valid comparator group.
In conclusion, aging is a complex process involving the interaction of a number of factors, including genetics, environment and lifestyle. Our findings highlight that physical inactivity with age may be a profound driver of several aspects of immunesenescence, most notably reduced thymic output, changes to regulatory cell frequency and function and T-cell polarization. Our future studies in this cohort will aim to test immune function, notably the response to vaccina-  ing strategy used to identify T-cell subsets (Duggal et al., 2014) and Bregs (Duggal et al., 2013) has been reported previously. Data analysis was carried out using SUMMIT software. Spectral overlap when using more than one colour was corrected via compensation. Appropriate isotype controls were used for setting gates.

| Serum cytokine and hormone-level assays
A Bio-plex cytokine assay for IL-6 (Bio-Rad Laboratories, Munich, Germany) was used for cytokine analysis using Luminex technology and performed according to manufacturer's instructions. Data acquisition and analysis were carried out using BIO-PLEX MANAGER software version 6.0. Serum IL-7, KGF and IL-15 levels were measured by Duo Set ELISA (R and D Systems, UK) according to manufacturer's instructions. Serum cortisol levels were measured by ELISA using a commercial kit (IBL International, Germany) and according to manufacturer's instructions.

| Stimulation of PBMCs to induce cytokine production by CD4 T cells
PBMCs were stimulated with PMA and ionomycin (both from Sigma-Aldrich) in the presence of brefeldin A (Sigma-Aldrich) at 37°C for 4 hr to determine the frequency of Th1 IFNc +ve , Th2 IL-4 +ve and Th17 IL-17 +ve CD4 T cells as described previously (Duggal et al., 2014). 4.6 | CD3 stimulation of PBMCs to induce IL-10 production by CD19 +ve CD24 hi CD38 hi B cells PBMCs were cultured in anti-CD3-coated wells for 72 hr and IL-10 expression measured by immunostaining of CD19 +ve CD24 hi CD38 hi B cells as described previously (Duggal et al., 2013).

| Statistical analysis
Statistical analysis was performed using the IBM SPSS STATISTICS 20 software (IBM software, UK). Univariate ANOVA with least significant difference post hoc tests was used to assess differences between the three groups. A Bonferroni correction was performed to adjust for multiple comparisons. For normally distributed data, a Student's t test analysis was performed to assess differences between two conditions. Statistical significance was determined at p < .05.

ACKNOWLEDG MENTS
The project was supported by a grant from the BUPA Foundation and the Glenn Foundation. JML is supported by the NIHR Birmingham Biomedical Research Centre in Inflammation, University Hospital Birmingham, UK.

CONFLI CT OF INTEREST
The authors have no conflict of interest to declare.

AUTHOR CONTRI BUTIONS
NAD and RDP carried out the experimental work; NAD performed the data analysis and prepared the manuscript; and NL, SH and JML conceived the study and contributed to writing the manuscript.