Asymmetric cell division shapes naive and virtual memory T-cell immunity during ageing

Efficient immune responses rely on heterogeneity, which in CD8+ T cells, amongst other mechanisms, is achieved by asymmetric cell division (ACD). Here we find that ageing, known to negatively impact immune responses, impairs ACD in murine CD8+ T cells, and that this phenotype can be rescued by transient mTOR inhibition. Increased ACD rates in mitotic cells from aged mice restore the expansion and memory potential of their cellular progenies. Further characterization of the composition of CD8+ T cells reveals that virtual memory cells (TVM cells), which accumulate during ageing, have a unique proliferation and metabolic profile, and retain their ability to divide asymmetrically, which correlates with increased memory potential. The opposite is observed for naive CD8+ T cells from aged mice. Our data provide evidence on how ACD modulation contributes to long-term survival and function of T cells during ageing, offering new insights into how the immune system adapts to ageing.

Efficient T-cell responses rely on heterogeneity. In the context of CD8+ T cells, this is achieved both by a diverse naive T-cell repertoire and by the ability of single cells to generate effector and memory progenies, responsible for both clearance of intracellular pathogens and long-term protection against reinfections. Asymmetric cell division (ACD) is an important mechanism to generate T-cell diversity, and its loss contributes to impaired immune responses, particularly with regard to the formation and maintenance of a memory T-cell pool1,2,3,4. Upon ACD, a single T cell can give rise to daughter cells inheriting distinct contents of fate determinants through the polarized segregation of cell cargoes, which culminates in different potential fates. We reported previously a correlation between stemness and the ability of a CD8+ T cell to undergo ACD by showing that short-lived effector and exhausted CD8+ T cells divide mainly symmetrically, while naive and central memory pluripotent CD8+ T cells are capable to divide asymmetrically3.

The ageing of the immune system, a multifaceted phenomenon also known as immunosenescence, threatens T cell diversity, as a result of thymic involution and antigen exposure history. This culminates in both inefficient immune responses and increased susceptibility to autoimmunity5,6,7. In other cell types, such as haematopoietic stem cells (HSCs), the loss of diversity observed during ageing has been linked to their impaired ability to undergo ACD. In this scenario, the loss of asymmetric fates in HSCs is clearly detrimental for haematopoiesis and maintenance of stemness8.

The decline of thymic functions observed in aged individuals results in a substantial change in T cell composition as fewer naive T cells emigrate to the periphery, and memory-phenotype CD8+ T cells start to play a more dominant role as they accumulate due to cytokine-driven homoeostatic proliferation9,10,11. Not being previously exposed to foreign antigen, memory-phenotype CD8+ T cells are usually termed virtual memory T cells (TVMcells)12,13,14,15,16,17,18,19. How TVM cells are generated has been a long-standing question, especially concerning whether this is a TCR-independent or TCR-instructed process. While TVM cells can be found in TCR-transgenic Rag deficient hosts20 and in both pathogen-free and germ-free mice21, a recent report identified a group of thymic precursors with self-reactive TCRs that give rise to memory-phenotype CD8+ T cells22. In contrast to cognate antigen-experienced memory T cells, TVM cells show consistent low expression of the α4 integrin CD49d. Therefore, CD49d can be used as a suitable marker to discriminate between true and virtual memory lymphocytes15,23,24.

TVM cells show paradoxical characteristics: their increased expression of CD122, a feature of memory cells, endows them with a competitive advantage in sensing IL-15, hence facilitating their long-term survival14,25, but they are also efficient producers of IFNγ, a feature of effector cells, resulting in their major protective role against infections in aged individuals26despite their lower proliferative potential27. Indeed, TVM cells can readily produce cytokines in the absence of antigenic exposure upon stimulation with IL-12, IL-15 and IL-1813,14,19,27, and constitutively express T-bet (a transcription factor that promotes the expression of effector genes) as well as effector molecules such as granzyme B and NKG2D (usually found in NK cells and activated T lymphocytes), features that facilitate their effector and bystander function28.

As ageing drives a profound change in the composition of T cells, particularly due to the drop of naive T cells with a concommitant rise of TVM cells, it remains to be addressed whether the ability of CD8+ T cells to undergo ACD is impacted by ageing. In this study, we found that ageing leads to an overall decline in the ability of CD8+ T cells to undergo ACD. This phenomenon was due to the lower ACD rates found in naive CD8+ T cells from aged animals, as CD8+ TVM cells had this feature unaltered by ageing. As opposed to truly naive CD8+ T cells from aged mice, TVM cells were unresponsive to mTOR inhibition-mediated enforcement of ACD, a strategy shown to promote ACD and memory potential in naive and memory CD8+ T cells from young mice3. Functionally, CD8+ TVM showed better re-expansion potential in adoptive transfer experiments compared to naive CD8+ T cells from aged mice, providing new evidence of a correlation between the ability to divide asymmetrically and memory potential. Our data suggest that ACD might play an important role to counteract immunosenescence, in that it is retained in TVM cells but lost in aged naive cells, thereby preserving the ability to adopt differential fates in TVM cells.

CD8+ T cells from aged mice are less able to undergo ACD

Considering the reduced capability of naive T cells to form memory populations during immunoscenescence, we investigated whether the ability of CD8+ T cells to undergo ACD is affected by ageing. To this end, we used bulk CD8+ T cells isolated from naive young (8–16 weeks old), middle-aged (>40 weeks old) and old (>70 weeks old) P14 mice (carrying a transgenic T-cell receptor specific for the lymphocytic choriomeningitis virus (LCMV) glycoprotein residues 33-41 (GP33) presented on H-2Db). The CD8+ T cells were stimulated on plates coated with human Fc-ICAM-1, α-CD3 and α-CD28 for 36 h, allowing them to enter the first round of cell division, identifiable as different stages of mitosis by confocal microscopy. CD8 polarization on mitotic cells was used as a reliable readout of ACD4. Comparison of the ACD rates from naive CD8+ T cells isolated from young or middle-aged/old P14 mice revealed that ageing correlated with an impaired ability to divide asymmetrically. We have shown previously that transient inhibition of the mTOR pathway by rapamycin is an efficient strategy to increase or restore the ability of CD8+ T cells to divide asymmetrically3. Interestingly, ACD rates, comparable to young CD8+ T cells, were restored when middle-aged/aged CD8+ T cells were submitted to transient mTOR inhibition (Fig. 1). We observed a similar trend when measuring the polarized inheritance of both T-bet and phosphorylated S629,30,31 (Supplementary Fig. 1). Together, our data provide evidence that ageing impairs the ability of CD8+ T cells to undergo ACD, and that transient mTOR inhibition can restore ACD in cells from aged mice. As ACD is correlated with stemness and memory potential3,4,30,31, these results reveal a possible strategy to reinvigorate CD8+memory T-cell responses in the elderly.

A Confocal images of murine CD8+ T cells fixed 36–40 h after in vitro stimulation on α-CD3, α-CD28 and human Fc-ICAM-1 coated wells (young, n = 59; young R, n = 49; middle-aged/old, n = 171; middle-aged/old R, n = 74). Transient mTOR inhibition was achieved by treatment with rapamycin (R) (20 nM). Cells were exposed to drug treatment from 12 h post activation until fixation for confocal microscopy analysis. Scale bar = 10 µm. B ACD rates observed in CD8+ T cells from middle-aged/aged and young mice upon transient mTOR inhibition with rapamycin. Data are represented as mean ± SEM. C Pie charts indicating frequencies of asymmetrically dividing cells from B. Level of CD8 asymmetric inheritance is depicted in black and level of CD8 symmetric inheritance in grey. Pooled data from three independent experiments. Age of the animals: experiments 1 and 2 (40–43 weeks) and experiment 3 (71 weeks). Statistical analysis was performed using the unpaired two-tailed Student’s t test. Exact P values are depicted in the figure (see also Supplementary Fig. 1).

Full size image

Re-establishment of ACD restores memory potential

An increased ability to divide asymmetrically has been previously reported to promote memory potential of emerging CD8+ T cell progeny3. Thus, we decided to test whether this would also hold true for cells from aged mice in which transient mTOR inhibition increased ACD rates in vitro. CD8+ T cells isolated from aged P14 mice (>70 weeks) were stimulated under ACD conditions (+ICAM) in presence or absence of transient rapamycin treatment, and their progenies were adoptively transferred into naive and young wild-type recipient mice (Fig. 2A). Recipients were then acutely infected with 200 ffu LCMV-WE > 30 days after adoptive transfer, favoring maintenance of long-lived cells that can survive in absence of antigen, a feature of memory cells. We monitored the expansion of the adoptively transferred P14 cells in the blood, spleen and lymph nodes (LNs) of recipient mice (Fig. 2B, C, F). We observed higher frequencies and numbers of progenies derived from stimulation conditions in which mTOR was transiently inhibited with rapamycin. Phenotypic changes were only observed in P14 cells in the LNs, where transient rapamycin treatment during in vitro stimulation gave rise to progenies with higher frequencies and numbers of cells committed to a memory fate, defined as KLRG1lo IL-7Rαhi expressing cells (Fig. 2D, G). Furthermore, progenies originating from higher asymmetry rate conditions, imposed by transient mTOR inhibition, exhibited increased cytokine production (TNF; Fig. 2E). As total P14 cell numbers were higher in progenies of cells transiently treated with rapamycin, we could also observe a significantly higher number of both IFNγ and TNF producing cells in this group. Together, these data provide further evidence regarding a correlation between increased ACD rates and progenies with increased memory potential, as determined by their enhanced survival and recall response upon re-challenge.

A Experimental setup (illustration by MB). Naive P14 cells from aged mice were stimulated in vitro on α-CD3, α-CD28 and human Fc-ICAM-1 coated wells in presence or absence of transient rapamycin (R) treatment, harvested after 36 h, and adoptively transferred into young and naive wild-type recipients. Each recipient mouse received 1 × 104 cells and was infected intravenously with LCMV-WE > 30 days later. B Frequencies of P14 cells within the CD8+ T-cell population in the blood. CFrequency and numbers of adoptively transferred P14 cells within CD8+ T cells in the spleens of recipient mice. D Frequencies and numbers of KLRG1hi IL-7Rαlo or KLRG1lo IL-7Rαhi P14 cells in the spleens of recipient mice. E Frequencies and numbers of splenic IFNγ and TNF P14 producing cells. F Frequencies and numbers of adoptively transferred P14 cells within CD8+ T cells in the inguinal lymph nodes of recipient mice. G Representative flow cytometry plots of KLRG1 and IL-7Rα expression in P14 cells in the lymph nodes of recipient mice 30 days after LCMV-WE challenge (left panel); percentages and numbers of KLRG1hi IL-7Rαlo and KLRG1lo IL-7Rαhi P14 cells in the lymph nodes of recipient mice (right panel). Representative data (n = 5 per group) from one out of two experiments. Aged P14 animals used for adoptive cell transfer: 95 weeks (experiment 1); 71 weeks (experiment 2). Data are depicted as mean + SEM. Statistical analysis was performed using the unpaired two-tailed Student’s t test (B–E, G, week 3) or when data did not pass the normality test, the unpaired two-tailed Mann–Whitney U test (F, G, #KLRG1hi IL-7Rαlo). Exact P values are depicted in the figure.

Full size image

TVM cells have a unique proliferative and metabolic phenotype

Our analysis of CD8+ T cells isolated from young and old P14 mice revealed an impaired ability of CD8+ T cells from aged mice to undergo ACD. In view of the heterogeneity of CD8+T cells present in “naive” mice, we decided to assess phenotypical differences between cells derived from young or old mice. To that end, we analyzed the expression of CD44 in CD8+ T cells from the blood of young (<16 weeks old) and aged (>70 weeks) P14 mice, as previous work has shown that ageing leads to an increased CD44hi/CD44lo T-cell ratio32,33,34. The frequencies of CD44hi cells were significantly higher within the CD8+ T cell pool in aged mice (Fig. 3A). A previous study provided evidence of a correlation between a higher expression of CD44 and an exhausted signature during ageing32. Considering the lower asymmetry rates observed in CD8+ T cells from aged mice and the negative effect of T cell exhaustion on the ability of CD8+ T cells to divide asymmetrically3, we characterized whether CD44hi cells would exhibit an exhausted phenotype. We stained peripheral blood CD8+ T cells from young and old P14 mice for the expression of both the co-inhibitory molecule PD-135,36,37, and the exhaustion marker CD3938. In agreement with more recent studies27, despite a trend towards an increase in PD-1hi and CD39+ cells amongst CD44hi CD8+ T cells in aged animals, these cells were still a minority in the CD44hi compartment (Supplementary Fig. 2). These data exclude exhaustion as a direct or major cause for the lower ACD rates observed in CD8+T cells from aged mice.

A Representative flow cytometry plots showing CD122, CXCR3, NKG2D and CD49d expression on CD8+ T cells in the blood of young and old P14 mice. Frequencies of CD44lo and CD44hi CD8+ T cells expressing high CD122, CXCR3 and NKG2D levels and low CD49d levels are depicted in the bar graphs. Pooled data from two independent experiments (young, n = 4: old, n = 5; aged mice: 70, 90 or 100 weeks old). B Histograms depicting proliferation of purified naive or TVM CD8+ T cells from young or aged mice (CTV-labelled, stimulated on α-CD3, α-CD28 and human Fc-ICAM-1 coated wells). After 42 h, the proliferation profile was analyzed by CTV dilution (left panel) or cells were transferred to uncoated wells for further 30 h in medium containing IL-2, IL-7 and IL-15, until their proliferation kinetics was assessed 72 h post stimulation (right panel). Representative data from one out of two experiments (aged mice: 70 or 82 weeks old). C Oxygen consumption rate (OCR) (upper panel) and extracellular acidification rate (ECAR) (lower panel) of activated CD8+ T cells was measured under basal conditions and in response to indicated drugs. Representative data from one out of two experiments, where data points are technical replicates (young naive, n = 10; old naive, n= 5, old TVM, n = 4; aged mice > 100 weeks old in all experiments). Data are depicted as mean ± SEM. Statistical analysis was performed using the unpaired two-tailed Student’s t test. Exact P values are depicted in the figure.

Full size image

CD44hi CD8+ T cells are known to accumulate with ageing39, and several reports have identified a large proportion of these cells as virtual memory cells (TVM)14,18,19,26,27. To investigate whether the CD44hi cells that were found at higher frequencies in aged P14 mice were indeed TVM cells, we first performed a phenotypic characterization16. We collected blood from both young and old unimmunized P14 mice and analyzed the expression of CD44, CD122, CXCR3, NKG2D and CD49d in CD8+ T cells. Confirming the TVM cell phenotype, CD44 expression on CD8+ T cells was accompanied by high expression of CD122, CXCR3 and NKG2D, and low expression of CD49d (Fig. 3A). Furthermore, as the CD44hi CD8+T cells were not CD49dhi in either young or aged unimmunized P14 mice, we excluded them being antigen-driven memory cells. Other studies have shown an accumulation of true memory CD8+ T cells upon ageing in wild-type animals with a polyclonal lymphocyte repertoire (Quinn et al.27). The absence of this population in naive P14 mice, independent of their age, corroborated that TVM cells derive from naive cells in a mechanism distinct from the differentiation of “true” memory cells by cognate antigen recognition.

To further characterize the TVM cell population present in the blood of aged P14 animals, we separately analyzed the proliferation profiles of naive and TVM splenic CD8+ T cells isolated from young and aged mice 42 or 72 h post activation (Fig. 3B). At an early time point (42 h post activation), TVM cells from both young and aged animals had undergone fewer divisions in comparison to their naive counterparts. However, 72 h post activation, TVM cells isolated from young hosts exhibited the highest proliferative capacity among all tested cell types, while TVM cells from aged hosts were the ones showing the lowest proliferation speed, but still similar to the profile seen in naive CD8+ T cells from young or aged animals. Truly naive CD8+ T cells from young and aged animals showed very similar proliferation profiles at both timepoints analyzed. These data corroborate previous observations, showing that TVMcells from young hosts exhibit faster cell cycling in comparison to naive cells27, but also suggest a unique cell division kinetics of TVM cells, as the time for the first division seemed to be extended, especially in the ones originating from an ageing host. Cell cycling speed can be a result of distinct metabolic activities and plays a role in T cell differentiation40,41,42. It has been reported that resting TVM CD8+ T cells have a different metabolism in comparison to naive or true memory cells, exhibiting elevated spare respiratory capacity (SRC)25. We therefore investigated whether activated naive and TVM CD8+ T cells exhibited metabolic differences. Because frequencies of TVM cells were particularly rare in young P14 mice, these cells were not used for metabolic assays. We assessed the metabolic performance of naive CD8+ T cells from young and aged animals, and TVM cells from aged animals 36 h post activation (when cells either did not divide or underwent their first round of mitosis) using Seahorse assays. Both naive and TVM CD8+ T cells from aged animals outperformed naive cells from young individuals concerning their SRC (Fig. 3C, upper panel). TVM cells from aged individuals showed a similar glycolytic performance as young naive cells, both with respect to their glycolytic capacity and glycolytic reserve, features for which naive cells from aged hosts exhibited impaired function (Fig. 3C, lower panel). This could indicate that activated TVM cells from aged hosts are metabolically adapted to swiftly respond to environmental changes through their ability to use both glycolysis and oxidative phosphorylation (OXPHOS) in order to sustain their metabolic demands, while simultaneously being “slow-dividers”, a feature previously reported to benefit ACD3 and memory formation42.

TVM CD8+ T cells show intrinsically high ACD rates

The observation of phenotypic, functional and metabolic differences between naive and TVM CD8+ T cells led us to investigate whether there would also be differences in their ability to divide asymmetrically. To that end, we sorted CD8+ T cells isolated from spleens and LNs of young and aged P14 mice into naive (CD44lo) and TVM (CD44hi, CD49dlo) populations. Cells were then stimulated in vitro under ACD conditions and submitted to transient mTOR inhibition with rapamycin or Akt-kinase inhibitor, known to increase ACD rates in naive and memory CD8+ T cells from young individuals3. The separation into naive and TVM cells prior to stimulation allowed us to directly compare their abilities to undergo ACD and to discriminate which cells were responsive to mTOR inhibition, evoking higher ACD rates and exhibiting improved memory potential after adoptive transfer (Figs. 1 and 2). TVM CD8+ T cells from both young and aged mice showed intrinsically high ACD rates, based on the levels of CD8 polarization (Fig. 4B, C). By contrast, naive CD8+ T cells from old mice had a markedly impaired capacity to undergo ACD in comparison to naive CD8+ T cells from young mice (Fig. 4A, C). As TVM cells take longer to undergo their first mitosis, it is likely that the frequencies of dividing TVM cells analysed amongst the bulk CD8+ T-cell population at 36 h post stimulation (Fig. 1) are under-represented. Furthermore, we cannot exclude that the analysed mitotic cells might represent a pool of TVM cells that retains their proliferation potential. These results, however, suggest that naive CD8+ T cells are responsible for the overall lower ACD rates observed in total CD8+ T cells isolated from aged P14 mice in comparison to their young counterparts (Fig. 1). The intrinsically high ACD rates in TVMCD8+ T cells could not be further increased by both rapamycin and Akt-kinase inhibitor treatment, while CD44lo naive CD8+ T cells from both young and old mice showed increased ACD rates under the same conditions (Fig. 4C). The responsiveness of naive CD8+ T cells from old mice to rapamycin treatment explains the re-established asymmetry observed when total CD8+ T cells from old P14 mice were exposed to rapamycin treatment (Fig. 1). We speculate that the failure in modulating ACD in TVM CD8+ T cells might result from both their higher mTOR activity in comparison to the their naive counterparts, and the inefficacy of the inhibitors, at least in the concentrations that they were administered, in downregulating this activity, especially what concerns downstream targets of mTORC2 (Supplementary Fig. 3). As ACD relies on the formation of a stable immune synapse, we analyzed the expression of CD3, CD28 and LFA-1 in naive and TVM CD8+ T cells, and observed that the latter exhibited significantly higher levels of LFA-1 on the surface, perhaps contributing to their intrinsically high ACD rates (Supplementary Fig. 4). We obtained similar results when performing experiments with polyclonal CD8+ T cells from wild-type C57BL/6 mice (Supplementary Fig. 5). Considering the superior memory potential exhibited by progenies of cells able to divide more asymmetrically, these results suggest that impaired immune responses in the elderly can be partially rejuvenated by re-establishing their ability to undergo ACD in naive CD8+ T cells through transient mTOR inhibition. In addition, as ACD proved to be beneficial for establishment of immune memory, the intrinsic ability of TVMCD8+ T cells to divide asymmetrically might be a compensatory mechanism to sustain efficient T-cell responses upon ageing.

A Confocal images from murine naive CD8+ T cells. B Confocal images from murine TVM CD8+ T cells. Cells were fixed 36–40 h after in vitro stimulation on α-CD3, α-CD28 and human Fc-ICAM-1 coated wells. Transient mTOR inhibition was achieved by treatment with rapamycin (20 nM) or Akt-kinase inhibitor (5 µM). Cells were exposed to drug treatment from 12 h post activation until fixation for confocal microscopy analysis. Scale bar = 5 µm. C CD8 asymmetry rates in naive and TVM CD8+ T cells isolated from young or old mice (young naive C, n = 28; young naive Akt i, n = 30; young naive R, n = 38; young TVM C, n = 23; young TVM Akt i, n = 19; young TVM R, n = 20; old naive C, n = 51; old naive Akt i, n = 47; old naive R, n = 42; old TVM C, n = 37; old TVM Akt i, n = 51; old TVM R, n = 52). Frequencies of asymmetrically dividing cells are depicted in pie charts on the right side of each graph. D Comparison of CD8 polarization amongst young naive, old naive, young TVM and old TVMCD8+ T cells (young naive, n = 28; old naive, n = 51; young TVM, n = 23; old TVM, n = 37). Data are shown as mean ± SEM. Pooled data from five independent experiments (aged mice: 71, 95, 97, 102 or 108 weeks old). For each experiment sorted cells from at least two animals were pooled. Statistical analysis was performed using the unpaired two-tailed Student’s t test. Exact P values are depicted in the figure.

Full size image

ACD modulation targets naive CD8+ T cells from aged mice

Within CD8+ T cells from old P14 mice, the naive subset showed strongly impaired ability to undergo ACD, but simultaneously the best responsiveness to transient mTOR inhibition as a strategy to re-establish ACD. To investigate a potential improvement of immune responses by modulating ACD in these naive CD8+ T cells, we adoptively transferred progenies of stimulated naive and TVM P14 cells from young and old mice into naive wild-type recipients (Fig. 5A). Stimulation was performed under ACD conditions, with or without transient mTOR inhibition with rapamycin. After 30 days, recipients were infected with LCMV-WE and numbers of P14 cells were evaluated in the spleen 30 days post viral challenge (Fig. 5B, C). Progenies of rapamycin treated naive P14 cells from both young and old P14 mice were present in higher numbers than their untreated counterparts (Fig. 5B). Progenies of CD8+TVM cells from both young and aged mice expanded to comparable frequencies or total numbers in presence or absence of mTOR inhibition during in vitro stimulation. As mTOR inhibition could not enhance their ability to undergo ACD, these data strengthen the correlation between the ability to undergo ACD and the potential to form a pool of memory cells. Rapamycin treatment did not lead to significant changes in the phenotypes of the resulting progenies, however, at least for naive cells from young mice, there was a trend towards higher frequencies of KLRG1lo IL-7Rαhi and lower frequencies of KLRG1hi IL-7Rαlocells of the rapamycin treated group (Fig. 5C), as previously reported by us3. Interestingly, progenies of aged CD8+ TVM cells exhibited significantly lower frequencies of KLRG1lo IL-7Rαhi cells, with concomitant emergence of a double-positive population for KLRG1 and IL-7Rα, known to be effector-like memory cells that exhibit enhanced cytotoxic activity43 (Fig. 5C). Functionally, frequencies of IFNγ and TNF producing cells were not affected by cell type or ageing (Fig. 5D). As total P14 cell numbers were higher in progenies of naive CD8+ T cells transiently treated with rapamycin and in progenies of TVM cells (independent of previous rapamycin treatment), we also observed significantly elevated numbers of both IFNγ and TNF producing cells in these groups. Together, these results emphasize a link between ACD, memory potential and efficient immune responses, and provide new insights into how the ability to undergo ACD is affected by ageing. We propose that CD8+ TVM cells might represent an adaptation to immunosenescence and compensate for the loss of ACD potential in naive cells. Furthermore, as the memory potential of naive CD8+ T cells from aged hosts could be increased by transient mTOR inhibition and increased ACD rates, these data reinforce the potential of this intervention for translational purposes in regenerative medicine.

A Experimental setup (illustration by MB). Naive and TVM CD8+ T cells were sorted from young and old P14 mice spleens, stimulated under different conditions, harvested after 36 h and adoptively transferred into wild-type recipients. Each recipient mouse received 1 × 104 cells and was infected intravenously with LCMV-WE (200 ffu) 30 days later. B Numbers of adoptively transferred P14 cells within CD8+ T cells in the spleens of recipient mice. C Percentages and numbers of KLRG1hi IL-7Rαlo, KLRG1lo IL-7Rαhi and KLRG1hi IL-7Rαhi P14 cells in the spleens of recipient mice (left panel). Representative plots showing gating strategy used to access expression of KLRG1 and IL-7Rα (right panel). D Frequencies and numbers of IFNγ, TNF and IL-2 producing cells in the spleens of recipient mice. Pooled data from two independent experiments (young naive, n = 8; young naive R, n = 8; young TVM, n = 3; young TVM R, n = 3; old naive, n = 6; old naive R, n = 6; old TVM, n = 7; old TVM R, n = 8). Aged P14 animals used for adoptive cell transfer: 92 or 96 weeks (experiment 1); 70 or 75 weeks (experiment 2). Data are depicted as mean + SEM. Statistical analysis was performed using the unpaired two-tailed Student’s t test. Exact P values are depicted in the figure.

Full size image

Ageing is associated with a progressive decline in the efficiency of mounting effective and long-lasting immune responses and results in higher vulnerability to infections in both mice and humans44. Virtually all aspects of immunity are affected by ageing, from the initial contact with a pathogen to its clearance or coexistence with a persisting pathogen7. Concerning T cell responses, the most profound change observable upon ageing is thymic involution10,45, which leads to marked changes in peripheral T cell maintenance mechanisms. With the decrease of thymic output, T cells that are already in the periphery go through more rounds of homoeostatic proliferation. Amongst possible other factors yet to be discovered, this contributes to the naive-to-memory conversion of their phenotype, which is independent of cognate antigen encounter, hence these cells are referred to as virtual memory cells13,14,15,17,18,19,27. TVM CD8+ T cells accumulate with ageing and depend on homoeostatic cytokines for maintenance. Besides activation through the T-cell receptor to exert effector functions, TVM CD8+ T are also responsive to cytokine activation14,19,28.

In the present study, our aim was to investigate a new aspect of the ageing immune system and characterize the ability of CD8+ T cells to undergo ACD. Our results on bulk CD8+ T cells show that total naive cells from aged mice divide less asymmetrically than their young counterparts. We have previously developed a strategy to increase ACD rates by transient mTOR inhibition with either rapamycin or Akt-kinase inhibitor, which was based on the reported benefits of mTOR inhibition on memory formation46,47. This transient mTOR inhibition during in vitro stimulation proved to be efficient in enhancing memory potential of progenies of naive, memory and PD-1int Tcf1+ exhausted CD8+ T cells3. Of relevance, mTOR is a key regulator of ageing in different organisms, from yeast to mammals48, and its inhibition has been linked to extended life span and improved immune functions in old mice and humans49,50. Furthermore, the inhibition of the homologous pathway TOR in yeast contributes to rejuvenation of yeast cells through higher ACD rates51. Transient rapamycin treatment of CD8+ T cells from aged mice indeed led to increased ACD rates, confirming our initial hypothesis that ACD can be modulated in these cells. This remarkable result proved to be of functional relevance, as progenies originating from enforced ACD conditions exhibited improved memory potential in a viral challenge model.

In line with previous reports, we observed a marked increase in the CD44hi/CD44lo ratio with ageing32,33,34, and identified the emerging CD44hi population as TVM cells. Besides phenotypic differences with naive T cells in surface marker expression, TVM cells also exhibited different proliferation rates and metabolism. After stimulation in vitro, TVM cells from both young and aged animals required a longer time to undergo their first rounds of division in comparison to their naive counterparts. However, the opposite was observed at later timepoints, when progenies of TVM cells showed similar or higher proliferation rates. Activated TVM cells/progenies acquired a unique effector-memory mixed metabolic profile, being capable to rely on both glycolysis and OXPHOS.

The differences in observed phenotype, proliferation kinetics and metabolism between naive and TVM CD8+ T cells led us to investigate which cell type is responsible for the lower ACD rates found in total CD8+ T cells originating from aged mice. Surprisingly, the ability of CD44lo naive CD8+ T cells to undergo ACD was strongly affected by ageing, as CD8+ T cells from old mice exhibited low ACD rates compared to their young counterparts. This defect could be reversed by transient mTOR inhibition. As expected, increased ACD rates observed in vitro led to improved function and correlated with a higher re-expansion potential upon antigenic challenge in vivo. By contrast, TVM cells exhibited intrinsically high ACD rates, which could not be further increased by transient mTOR inhibition. As ACD is thought to be beneficial for diversified T-cell responses, in particular to impart memory potential, we propose that TVM cells are an adaptation of the immune system to immunesenescence. Indeed, our results might explain how TVM cells end up comprising the majority of the CD44hiCD62Lhi TCM cell population in LCMV infected mice25, or how these cells are able to compose the majority of the central memory CD8+ T cell (TCM) compartment after LM-OVA antigenic challenge, while also being able to generate effector responses19.

The better expansion potential of TVM CD8+ T-cell progenies in comparison to their naive counterparts in adoptive transfer experiments followed by antigenic challenge contradicts previous findings, where CD44lo naive CD8+ T cells were described as superior in that context27,32. However, the divergent results may be explained by the different experimental setups. In our study, adoptively transferred cells were progenies of TVM cells that underwent ACD and were required to survive for over 30 days before re-challenge. In previous studies27,32, recipient mice received unstimulated cells and were either immediately immunized with cognate antigen or left untreated to assess the differential survival of naive CD8+ T cells and TVM CD8+ T cells.

Taken together, our results suggest that the extended time taken by TVM cells to undergo their first rounds of division are beneficial for ACD, and thus for the establishment of asymmetric fates, which potentially also explains their exclusive proliferation and mixed metabolic phenotypes. ACD contributes to the simultaneous rise of (1) a pool of memory cells reliant on OXPHOS and exhibiting elevated SRC, and (2) a pool of effector cells that relies on glycolysis, and undergoes rapid proliferation at later timepoints42. We observed that TVM cells from aged animals, which exhibit higher expression of CD122 than their naive counterparts, are better survivors under limiting IL-15 concentrations in vitro, highlighting a potential role of survival in the observed increased recovery of adoptively transferred cells upon re-challenge/re-expansion (Supplementary Fig. 6). Thus, the virtual memory pool of CD8+ T cells might present an adaptation of the ageing immune system to maintain a memory-like pool of cells in absence of previous antigen encounter without compromising effector responses.

A question that remains unresolved is the reason for the refractoriness of TVM CD8+ T cells to mTOR inhibition when measured by modulation of ACD rates. CD44 engagement is known to lead to downstream phosphorylation of Akt, resulting in overall higher mTOR activity52. When we assessed mTORC1 and mTORC2 activities, measured by phosphorylation levels of S6 and NFκB53, our data indicated that TVM cells have higher basal mTOR activity in comparison to their naive CD44lo counterparts and are refractory to mTORC2 inhibition (Supplementary Fig. 3). As there is evidence on the role of mTORC2 inhibition as an efficient strategy for ACD modulation3, these data suggest that constitutive mTORC2 activity might be the reason for both the lack of ACD modulation in TVM cells and their efficient glycolytic metabolism. Furthermore, the inability to modulate ACD in TVM CD8+ T cells, by either rapamycin or Akt-kinase inhibitor treatment, could be caused by the altered role of mTOR during CD8+ T cell senescence, which potentially leads to differential activation of its downstream targets54,55,56. However, the reason for their higher intrinsic ability to divide asymmetrically remains unresolved and needs to be further investigated.

Transient mTOR inhibition efficiently restored the ability of naive CD8+ T cells from aged mice to undergo ACD. This correlated with an increased re-expansion of transferred cells upon LCMV challenge, which is evidence of their better survival capacity and memory potential. Re-establishment of ACD might therefore represent a strategy for rejuvenation of senescent naive CD8+ T cells found in aged individuals. Conversely, TVM cells exhibited intrinsically high ACD rates which could not be further increased by transient mTOR inhibition. Functionally, these properties culminated in increased memory potential shown by adoptive transfer experiments, providing a further link between ACD and efficient immunity. The accumulation of TVM cells with ageing might therefore, at least to some extent, compensate for the diminished memory potential of naive cells. Reinforcing ACD in these naive cells in aged individuals might offer new perspectives for improvement of immune functions in the elderly, and brings new insights on how the immune system adapts to ageing.