Autophagy in T cells from aged donors is maintained by spermidine, and correlates with function and vaccine responses

Older adults are at high risk for infectious diseases such as the recent COVID-19 and vaccination seems to be the only long-term solution to the pandemic. While most vaccines are less efficacious in older adults, little is known about the molecular mechanisms that underpin this. Autophagy, a major degradation pathway and one of the few processes known to prevent aging, is critical for the maintenance of immune memory in mice. Here, we show induction of autophagy is specifically induced in human vaccine-induced antigen-specific T cells in vivo. Reduced IFNγ secretion by vaccine-induced T cells in older vaccinees correlates with low autophagy. We demonstrate in human cohorts that levels of the endogenous autophagy-inducing metabolite spermidine, fall with age and supplementing it in vitro recovers autophagy and T cell function. Finally, our data show that endogenous spermidine maintains autophagy via the translation factor eIF5A and transcription factor TFEB. With these findings we have uncovered novel targets and biomarkers for the development of anti-aging drugs for human T cells, providing evidence for the use of spermidine in improving vaccine immunogenicity in the aged human population.


Introduction
The outbreak of coronavirus disease 2019 (COVID-19) caused a great threat to world-wide public health in 2020 with the majority of deaths occurring in older adults. The development of effective treatments and vaccines against COVID-19 is now more than ever becoming a pressing and urgent challenge to overcome 1,2 .
However the successful vaccination of the elderly against pathogens is considered one of the big challenges in our society 3,4 . Immunosenescence, which is characterized by poor induction and recall of B and T memory responses upon exposure to new antigens, can lead to reduced immune responses following immunization of older adults. While most vaccines are less immunogenic and effective in the older population 3 , little is known about the molecular mechanisms that underpin immune senescence. Autophagy is thought to be one of the few cellular processes that underlie many facets of cellular ageing including immune senescence 5 . By delivering unwanted cytoplasmic material to the lysosomes for degradation, autophagy limits mitochondrial dysfunction and accumulation of reactive oxygen species (ROS) 6 . Autophagy degrades protein aggregates that accumulate with age and its age-related decline could contribute to "inflammaging" 7 , the age-related increase in inflammatory cytokines in in blood and tissue.
Loss of autophagy strongly promotes production of the inflammatory cytokines TNFa, IL-6 and IL1-b 8,9 . We previously found autophagy levels decline with age in human peripheral CD8 + T cells 10 . Deletion of key autophagy genes leads to a prematurely aged immune phenotype, with loss of function in mouse memory CD8 + T cells 11 12 , hematopoietic stem cells 13 , and macrophages 9 with a myeloid bias 13 . In addition, we find in autophagy-deficient immune cells the same cellular phenotype that cells display in older organisms; they accumulate ROS and damaged mitochondria 9,12 .
Importantly, we can improve CD8 + T memory responses from aged mice with spermidine 12 , an endogenous metabolite synthesized from arginine. It was shown in yeast and other model organisms that spermidine extends life-span via increased autophagy 14 . Several downstream mechanisms of spermidine-induced autophagy have been described in mice, including the inhibition of histone deacetylases 14 . Recently we uncovered a novel pathway in which spermidine lends a residue for the hypusination of the translation factor eIF5a, which is necessary for the translation of a three proline motif present in the master transcription factor of autophagy and lysosomal biogenesis, called TFEB 15 . We demonstrated this pathway operates in activated B cells, which upon activation have an unusually high protein synthesis rate, owing to the high production of immunoglobulins. It is likely that B cells may be particularly reliant on the unfolded protein response, the proteasome, and autophagy, to cope with this high rate of protein synthesis. B cells may have evolved special coping strategies including the translational signalling for autophagy via eIF5A and TFEB. We therefore sought to extend our findings to another immune cell type, CD8+ T cells, to investigate whether this pathway may be conserved in a related adaptive immune subset and possibly broadly applicable.
Here we show for the first time that autophagy is indeed highly active in human CD8+ T cells after the in vivo encounter of antigens in donors from two different experimental vaccination trials. Our data show that polyamine levels fall with age in peripheral mononuclear cells. When supplemented with spermidine, the dysfunctional autophagic flux can be rejuvenated in CD8+ T cells from old donors, and levels of the important effector molecules IFN and perforin are enhanced as a consequence. Moreover, autophagy and effector function are maintained by spermidine in T cells from young donors. Lastly, in human CD8+ T cells we show that spermidine signals via eIF5A and TFEB to maintain autophagy levels. This study demonstrates that the function of human CD8+ T cells can be improved with spermidine. Taken together with our previous work on B cells, this leads us to the hypothesis that both T and B cell responses to infections and vaccinations are exquisitely reliant on sufficient autophagy levels, which are maintained by intracellular spermidine. This work highlights the potential of spermidine as a vaccine adjuvant in the older adults.

Results
First we optimised a flow cytometry-based assay to reliably and reproducibly measure autophagy, before applying it to measure autophagy after in vivo antigen stimulated T cells post-vaccination. To inhibit the autophagic flux and thereby degradation of LC3-II, the lysosomal inhibitor bafilomycin A was added to the culture for 2 hrs before washing out non-membrane bound LC3-I and staining. In    show a significant increase in autophagic flux that is not observed in HCV nonspecific CD8+ T cells (Fig 1c). In antigen-specific T cells, autophagic flux is highest shortly after vaccination but had declined to levels equivalent to antigennon-specific T cells cells by the end of the study (week 36 or 52). Together, these data show that antigen exposure induces autophagic flux in CD8+ T cells in humans in vivo. In previous work we found autophagic flux was reduced in CD8+ T cells from 24 month old mice 12 . To test whether this is true in human

CD8+ T cells from human individuals, and whether this correlates with vaccine
immunogenicity, we measured autophagic flux in vaccinees of various ages that were given an experimental RSV vaccine (Fig1a and Supplementary Fig S3b).   As older adults are particularly susceptible to severe disease from RSV infection, the vaccine was given to two age groups (18-50 years and 60-75 years of age).

Figure 1 Autophagy is induced by vaccination in antigen-specific T cells and correlates with donor age. Peripheral blood mononuclear cells were isolated from blood samples of vaccinated healthy donors. LC3-II was measured in CD8+ cells using flow cytometry after
As expected, the naturally-derived population of RSV-specific IFNγ producing CD4+ and CD8+ T cells in peripheral circulation in response to the infection declines with age (Fig. S4).

Figure S4
Correlation of age with total and peptide-pool specific T cell IFNg response to RSV exposure measured by ELISpot in CD8 + cells, donors as in Fig 1E.    No MHC class I pentamers were available for RSV to identify antigen-specific T cells, however, bulk T cells of the >60year vaccinees showed significantly lower basal autophagic flux (Fig. 1d). We correlated autophagy levels in T cells with IFNγ ELISpot responses in individual vaccinees and found a strong inverse correlation in the aged group (Fig. 1e) between autophagy and IFNγ responses, but not in the young group (Fig. 1f). Taken together these data suggest that reduced T cell autophagy in aged T cells may underpin reduced T cell responses to vaccination.
We have recently shown that treating old but not young mice with the metabolite spermidine improves autophagy levels in B lymphocytes due to an age-related decline of endogenous spermidine 15 . Here we sought to confirm this in human lymphocytes. Firstly, we determined spermidine and putrescine levels in PBMC by gas chromatography-mass spectrometry (GC-MS), and found an inverse correlation between age and spermidine but not with putrescine (Fig. 2a). We hypothesized that low levels of spermidine are responsible for low levels of autophagy and poor T cell function in PBMC from old donors. We therefore tested whether supplementation with spermidine recovers T cell autophagy and function.
As activation of PBMC with anti-CD3/CD28 optimally induces autophagy levels on day 4 18 , we activated PBMC from old donors in the presence of spermidine for 4 days and tested their autophagic flux and function by flow cytometry. Both autophagic flux and the secretion of IFNγ measured by ELISA was improved significantly in T cells from older vaccinees ( Fig. 2b and c), Similarly, increased IFNγ can be detected after spermidine treatment by intracellular staining for flow cytometry (Fig. 2d). Interestingly, spermidine supplementation also increases the expression of Perforin (Fig. 2e) but not of Granzyme B (Fig. 2f).  (Fig. 3a). DFMO also partially blocks IFN and Perforin expression in anti-CD3/CD28 activated CD8+ T cells, which can also be rescued by spermidine (Fig. 3b and c). As before, Granzyme B was not affected by DFMO + spermidine (Fig 3d). T cells from young donors, with their high endogenous spermidine levels, do not respond to spermidine supplementation (Fig. 3e, f and g). We previously found that spermidine maintains autophagic flux via hypusination of eIF5A and TFEB 15 , and we sought to test this pathway in human T cells.  4a) and also confirmed that it decreases LC3-II expression in a dose-dependent manner (Fig. 4b). GC7 reduces TFEB and hypusinated eIF5A in CD8+ T cells activated for 4 days (Fig 4c). It also diminishes the autophagic flux in activated CD8+ T cells over a time course of 7 days (Fig. 4d). We then tested whether

Figure 4 Spermidine's mode of action is via eIF5A and TFEB in human CD8+ T cells (a) Human T cell line Jurkat was cultured for 24h with 100 µM GC7, then eIF5A and hypusinated eIF5A were measured by WB. (b) Jurkat cell line was stimulated with increasing concentrations of GC7 and cell lysates blotted for LC3B. (C, D) PBMC from young human donors were cultured with anti-CD3/CD28 for 7 days and treated with GC7. The protein levels of TFEB and eIF5A hypusination were measured in CD8+ cells by Western blot on day 4 (c)
and autophagic flux was determined as in Fig 1 (d)  response. In addition, to be active, TFEB needs to be dephosphorylated for its translocation into the nucleus, which mTOR inhibition facilitates [32][33][34] . Therefore, spermidine and an mTOR inhibitor may have to be combined to optimally restore immune responses in the older adults.
We find that spermidine levels decline in PBMCs in the older humans. This confirms earlier study in plasma in which spermidine levels were found to be low in the >65 age group and rising again in centenarians 35 . Studies of this kind usually indicate that the phenotype is maladaptive with age as centenarians do not display many of the aging features of the age group below. However, the reasons for the age-related decline in spermidine in blood cells are not clear, and currently under investigation.
Overall it is evident that endogenous spermidine maintains autophagy in human T cells, a novel metabolic pathway that has not been investigated before.
Spermidine has recently been administered to humans in a small experimental trial with a beneficial effect on cognitive function without adverse effects 36,37 . It remains to be shown whether at such low doses it has an effect on autophagy. Our study strongly suggests that a small experimental trial should be conducted to test whether spermidine can be used to improve vaccination efficiency in older adults -eIF5A, TFEB and autophagy in PBMCs could be used as biomarkers to determine dose, duration and biomarkers for the anti-immune senescence effect.
In conclusion we validated a novel anti-immune senescence pathway in humans with druggable targets and biomarkers, which could also be used for other broader antiaging drug trials.

Human Samples Human peripheral blood mononuclear cells (PBMC) were
obtained under the ethics reference NRES Berkshire 13/SC/0023, from phase I clinical trials of novel viral-vectored vaccines for hepatitis-C virus (HCV; NCT01070407 and NCT01296451) or respiratory syncytial virus (RSV), described in more detail elsewhere 16,17 38 39 40 . Volunteers were self-selected adults who provided written informed consent and who were carefully screened for being healthy before vaccination. The vaccine schedules are described in Diagrams ( Fig   S4 A and B). Blood samples were collected in heparinised tubes for assays that required PBMC. PBMC were isolated within 6 h of sample collection.  Table 2.
Following incubation cells were washed with FACS buffer and immediately analysed on a four-laser LSR Fortessa X-20 flow cytometer. Acquired data were analyzed using FlowJo 10.2.

MHC class I pentamer staining to identify antigen-specific T cells
An HCV-specific HLA-A*02-restricted pentamer, peptide sequence KLSGLGINAV (Proimmune) was used to identify HCV-specific CD8+ T cells ex vivo. PBMC were washed in PBS and were stained with pentamers at room temperature (20mins) in PBS, washed twice in PBS before further mAb staining as described above.
During analysis, stringent gating criteria were applied with doublet and dead cell exclusion to minimise nonspecific binding contamination. and blocked with 5% skimmed milk-TBST. Membranes were incubated with primary antibodies dissolved in 1% milk overnight and secondary antibodies dissolved in 1% milk with 0.01% SDS for imaging using the Odyssey CLx Imaging System. Data were analyzed using Image Studio Lite.

Spermidine measurement in cell lysates by GC-MS
This protocol was used as published previously 41 . Briefly, cells were washed with PBS and the pellet resuspended in lysis buffer (80% methanol + 5% Trifluoroacetic acid) spiked with 2.5 µM 1,7-diaminoheptane (Sigma). The cell suspension, together with acid-washed glass beads (G8772, Sigma), was transferred to a bead beater tube and homogenized in a bead beater (Precellys 24, Bertin Technologies) for four cycles (6500 Hz, 45 s) with 1 minute of ice incubation between each cycle. The homogenized samples were centrifuged at 13,000 g for 20 minutes at 4°C. The supernatant was collected and dried overnight. For chemical derivatization, 200 µL trifluoroacetic anhydride was added to the dried pellet and incubated at 60°C for 1 hour, shaking at 1200 rpm. The derivatization product was dried, re-suspended in 30 µL isopropanol and transferred to glass loading-vials. The samples were analyzed using a GCxGC-MS system as described 41 . The following parameters were used for quantification of the 1D-GC-qMS data: Slope: 1000/min, width: 0.04 s, drift 0/min and T. DBL: 1000 min without any smoothing methods used.

Statistical analyses
Prism software (GraphPad) was used for statistical analyses. Data are represented as mean ± SEM. All comparative statistics were post-hoc analyses.
Paired or unpaired two-tailed Student's t-test was used for comparisons between two normally distributed data sets with equal variances. Linear regression with a 95% confidence interval was used to assess the relationships between age and the expression of target proteins or spermidine levels, in which R 2 was used to assess the goodness of fit and the P value calculated from F test was used to assess if the slope was significantly non-zero. P values were used to quantify the statistical significance of the null hypothesis testing. *P≤0.05, **P≤0.01, ***P≤0.001, ****P≤0.0001, ns, not significant.