Our clinical protocol was approved by the NRES Committee North East (16/NE/0073) and all subjects provided written informed consent. We recruited healthy “young” volunteers (Age range: 20–40 yr, n = 18), healthy “old” volunteers (Age range: 52–75 yr, n = 30) and people aged >18 yr living with T2D (Age range: 50–77 yr, n = 52), identified through the Diabetes Alliance for Research in England (DARE) database, with preferential sampling of people aged 50–80 yr. Exclusion criteria for all participants were: inability to provide written informed consent, infection or immunisation in the month prior to blood collection, any known immunodeficiencies or a history of chemotherapy/radiotherapy, on any immunosuppressive medications within the last 6 mo, significant comorbidity or had a history of neoplasm in the last 10 yr. Peripheral blood was obtained using heparinised tubes, and peripheral blood mononuclear cells (PBMCs) were isolated using Ficoll hypaque (Amersham Biosciences). In addition to a blood sample, the participants age, gender, length of time since T2D diagnosis and medications were documented. However height and weight was not collected (Table 1).

Flow cytometric analysis was performed using the following antibodies: anti-glut1 PE (202915) from R&D Systems, anti-KLRG1 PE (MAFA) from Miltenyi Biotec, anti-CD8 PerCP (SK1), anti-CD45RA BV605 (HI100), anti-CD45RA APC (HI1000), anti-CD27 BV421 (O323), anti-CD27 FITC (O323), anti-CD28 BV785 (CD28.2), and anti-CCR7 PECy7 (G043H7) from BioLegend. anti-Ki67 PE (B56; BD Bioscience), anti-p-p53 AF674 (Ser15, 16G8; Cell Signaling) and anti-pAMPK (T172, 40H9; Cell Signaling) were used for intracellular staining using solution AB (ThermoFisher). All samples were analysed using a LSR Fortessa (BD Biosciences) and the resulting data examined using FlowJo software (BD Bioscience). UMAP analysis was performed using FlowJo on down sampled CD8⁺ T cell populations (5,000 events), with standard parameters (nearest neighbours = 15, minimum distance = 0.5) and clustered via flowSOM.

Magnetic beads were used to isolate CD8⁺ T cells by positive selection according to the manufacturer’s instructions (Miltenyi Biotec). The purity of T cell subsets was assessed by flow cytometry.

Mitochondrial mass was assessed by incubating antibody labelled PBMCs with 100 nM of MitoTracker Green FM (ThermoFisher) for 30 min at 37°C, 5% CO₂. Mitochondrial ROS was measured in labelled PBMCs using 2 µM MitoSOX (ThermoFisher) incubated for 20 min at 37°C, 5% CO₂. Unfixed samples were immediately analysed using a LSR Fortessa (BD Bioscience).

CD27/CD45RA defined CD8⁺ EMRA T cell subsets were isolated and fixed using 2% paraformaldehyde, 1.5% glutaraldehyde in 0.1 m phosphate buffer at pH 7.3. The EMRA T cells were processed as previously described (Henson et al., 2014) and were examined using a Jeol 1010 transmission electron microscope (Jeol) with image capture using a Gatan Orius CCD camera (Gatan).

Oxygen consumption rates (OCR) and extracellular acidification rates (ECAR) were measured in bead sorted CD8⁺ T cells following 15 min stimulation with 1 μg/ml anti-CD3 and 5 ng/ml IL-2. The assay was performed using RPMI without phenol red and carbonate buffer (Sigma) supplemented with 25 mM glucose, 2 nM L-glutamine and 1 mM pyruvate. The metabolic stress test was performed using 1 µM oligomycin, 1.5 µM fluorocarbonyl cyanide phenylhydrazone (FCCP), 100 nM rotenone and 1 µM antimycin A (Sigma) using a XF-96 Extracellular Flux Analyzer (Agilent).

Fatty acid oxidation (FAO) was performed in substrate limited media: DMEM (Gibco), 0.5 mM glucose, 1 mM L-glutamine, 0.5 mM carnitine, and 1% FCS). 45 min prior to the assay cells were transferred to warm Krebs-Henseleit KH Buffer (KHB) (111 mM NaCL, 4.7 mM KCL, 1.25 mM CaCl2, 2 mM MgSO4, 1.2 mM NaH2PO4) supplemented with 2.5 mM glucose, 0.5 mM carnitine, 5 mM HEPES, pH 7.4. Subsequently, 15 min prior to the assay, 40 μM Etoximir was added to selected groups. Finally, the cells were provided with Palmitate:BSA (Aglient) as a substrate and subjected to a Mitochondrial Stress Test as described above.

Lipid uptake was measured in PBMCs that had been stimulated for 1 h at 37°C with 1 μg/μl of anti-CD3 in complete RPMI-1640 medium. 1 nM of the fluorescent palmitate analogue BODIPY FL C16 (Invitrogen) was then added and the cells incubated in the dark for 10 min. Cells were subjected to surface staining prior to flow cytometric analysis.

PBMCs isolated from a young donor were cultured in either low (5 mM) or high glucose (25 mM) RPMI-1640 medium with or without 10% sera from either T2D patients or healthy age-matched control participants. Antibody staining was carried out on samples at day 0, 3 and 7 and the cells analysed by flow cytometry in order to phenotype the samples.

The Proteome Profiler Human Cytokine Array kit (R&D Systems) was used to quantitate the cytokines present in sera taken from people living with T2D or age-matched individuals according to the manufacturer’s instructions.

To assess mitochondrial morphology, CD8⁺ T cells were incubated in either high (25 mM), low glucose (5 mM) or with 200 μM palmitate overnight prior to plating onto 12 well microscope slides at a density of 0.05 × 10⁶ per well. Cells were fixed in 2% PFA and 100 nM of MitoTracker Green FM (ThermoFisher) added for a duration of 30 min at 37°C, 5% CO₂. 4ʹ,6-diamidino-2-phenylindole (DAPI) (ThermoFisher Scientific) was then added for 10 min at room temperature in the dark. Samples were imaged on a Zeiss LSM 880 confocal microscope with a ×63 oil immersion objective lens. Excitation was at 488 nm from an argon-ion laser. Fluorescence detection was in the green, 488 nm and UV, 405 nm channel.

Unstimulated CD8⁺ EMRA T cells from T2D and healthy age-matched control participants were isolated using MACS sorting for CD8⁺ T cells and then FACS sorted using anti-CD27/anti-CD45RA for EMRA isolation. RNA was extracted using the RNAeasy Micro Kit (Qiagen) and pre-amplified using the RT2 PreAMP cDNA Synthesis Kit (Qiagen) and RT2 PreAMP Pathway Primer Mix (Qiagen). Resulting gene-specific cDNA was then analysed using the Cellular Senescence RT2 Profiler PCR Arrays according to the manufactures instructions (Qiagen).

RNA from sorted CD27/CD45RA defined CD8⁺ EMRA T cells was isolated using the RNeasy kit (Qiagen) according the manufactures instructions. Transcripts were quantified using the High-Capacity cDNA Reverse Transcription Kit (Applied Biosystems) and the SsoAdvanced Universal SYBR Green Supermix (Bio-Rd Laboratories) according the manufactures instructions. The following primers were purchased from IDT. Fis1 F: GAT GAC ATC CGT AAA GGC ATC G; Fis 1 R: AGA AGA CGT AAT CCC GCT GTT, Mff F: ACT GAA GGC ATT AGT CAG CGA; Mff R: TCC TGC TAC AAC AAT CCT CTC C, DNM1L F: GGT GAA CCC GTG GAT GAT AAA; DNM1L R: CCT CAG GCA CAA ATA AAG CAG, BActin F: CAC CAT TGG CAA TGA GCG GTT C; BActin R: AGG TCT TTG CGG ATG TCC ACG T.

GraphPad Prism was used to perform statistical analysis. Statistical significance was evaluated using the paired Student t-test or a two-way ANOVA with Bonferroni correction used for post-hoc testing. Data was expressed as mean ± SEM and p values are represented using the following notation: p* = <0.05, p** = <0.01, p*** = <0.005 and p**** = < 0.001.

The cell surface markers CD45RA, CD28, CD27, CCR7 and KLRG1 were used to quantify the frequency of CD8⁺ T cell subsets in people living with T2D compared with healthy age-matched controls and young participants. Four populations of T cells can be distinguished using these markers, Naïve T cells (N: CD45RA⁺CD28⁺CD27⁺CCR7⁺KLRG1^low), central memory cells (CM: CD45RA⁻CD28⁺CD27⁺CCR7⁺KLRG1^low), effector memory cells (EM: CD45RA⁻CD28⁻CD27⁻CCR7⁻KLRG1^high) and effector memory CD45RA re-expressing T cell (EMRA: CD45RA⁺CD28⁻CD27⁻CCR7⁻KLRG1^high). UMAP (Uniform manifold Approximation and Projection) analysis was subsequently performed (Figure 1A). UMAP is an unbiased manifold learning technique for dimension reduction. The plot represents each CD8⁺ T cell subset as a point, which when combined with heat map expression data allows the properties of each cluster to be determined (Figure 1B). UMAP analysis indicate that older people living with T2D have an elevated expression of the senescent EMRA T cell subset. This was confirmed when the data was stratified into people aged 55–64 and those over 65 yr together with disease status. T2D participants were found to have an increased percentage of CD8⁺ EMRA T cells compared to age matched individuals, demonstrating that people living with T2D have a prematurely aged T cell phenotype (Figure 1C, Supplementary Figure S1A). However we did not measure CMV status of the participants, it has been reported that CMV seropositive oldest old (>85 yr) individuals were more likely to have T2D (Chen et al., 2012a). Therefore, any effect CMV may be having on T cell subset profile cannot be excluded.

All people living with T2D were on medication, with the majority taking three to six different types of medication to control their condition (Table 1). To assess the potential impact of medications on CD8⁺ EMRA numbers, we compared the percentages of the T cells with the total number of drugs used, however no relationship was found (Figure 1D). This led us to conclude that the rise in CD8⁺ EMRA T cells was a consequence of T2D and not the medications taken by the participants.

Mitochondrial health and cellular metabolism can influence the onset of senescence in T cells (Callender et al., 2020). CD8⁺ EMRA T cells exhibit mitochondrial dysfunction and impaired OXPHOS when compared to the other memory subsets (Henson et al., 2014). However, the metabolic properties of CD8⁺ EMRA T cells from people living with T2D have not been assessed. The mitochondrial mass of CD8⁺ T cell subsets was determined using Mitotracker green staining. While T2D CD8⁺ EMRA T cells showed a tendency towards a decreased mitochondrial content it was not significantly different from the age-matched controls, however there was a loss of mitochondria when T2D EMRAs were compared to those from young individuals (Figure 2A). When the mitochondria were observed by electron microscopy, mitochondria from T2D CD8⁺ EMRA cells were found to exhibit a different morphology compared to participants without T2D. The mitochondria from participants with T2D were typically smaller, rounder with little or no cristae (Figure 2B). Mitochondria frequently change shape when stressed and fragmented mitochondria can increase reactive oxygen species (ROS) production leading to further damage (Yu et al., 2006).

Next, we went on to assess whether mitochondria from T2D EMRAs had a propensity to produce more ROS than EMRA T cells from healthy individuals. Data obtained from a cellular senescence RT2 profiler array indicated that NADPH oxidases 4 (NOX4) was greatly upregulated in the T2D CD8⁺ EMRA T cells compared with the age-matched control EMRAs (Figure 2C). NOX enzymes, including NOX4, are major contributors to cellular ROS (Chen et al., 2012b). The mitochondrial gene; superoxide dismutase 2 (SOD2), which converts superoxide by-products of OXPHOS into hydrogen peroxide and diatomic oxygen (Velarde et al., 2012), was also elevated. This is likely an attempt to reduce the damaging effects of elevated oxidative stress. Despite the elevated SOD2 expression, examination of mtROS levels using the mitochondrial superoxide indicator MitoSOX revealed that mtROS levels were highest in T2D EMRA subsets compared with the age-matched controls or EMRAs isolated from younger individuals (Figure 2D). These data suggest that the elevated ROS content potentially has a damaging effect of the EMRA T cells from people living with T2D.

The consequence of a reduced mitochondrial content in CD8⁺ EMRA T cells isolated from people living with T2D was investigated further. Due to the limited cell numbers obtained from T2D blood donations we used whole CD8⁺ T cells to investigate metabolic changes. We believe this to reflect the EMRA population as more than 50% of the total CD8⁺ population isolated from people living with T2D are EMRAs (Figure 1C, Supplementary Figure S1A). Furthermore, mitotracker and mitoSOX data show very similar staining patterns when the EMRA subset from young, old and old T2D individuals were compared to the total CD8⁺ fraction (Supplementary Figure S1B). Therefore, we compared the bioenergetic profiles of whole CD8⁺ T cells from young, healthy age-matched controls and T2D participants (Figure 3A). Interestingly, CD8⁺ T cells from people living with T2D had an elevated basal and spare respiratory capacity (Figure 3B) compared to age-matched or younger individuals. We do not believe this to be a consequence of metformin usage, which has been demonstrated to increase oxidative capacity in mitochondria (Wang et al., 2019), as we observed similar flux profiles when we separated the data into those who were and were not treated with metformin (Supplementary Figure S1C). However, despite this apparent increase in respiratory function, CD8⁺ T cells from people living with T2D displayed a far higher proton leak indicative of mitochondrial damage (Figure 3B).

The energy source fuelling the increased oxidative respiration was then investigated. We have shown previously that CD8⁺ EMRA T cells taken from people living with T2D displayed a lack of glucose uptake (Lau et al., 2019). This was confirmed when we investigated the expression of the glucose transporter Glut1. CD8⁺ T cells from old and T2D individuals were found to display a recycling defect, while this prolonged surface duration of the Glut1 transporter we failed to observe an increase in glucose uptake and glycolysis. The rate of nutrient uptake is largely determined by the permanence of transporters at the cell surface, controlled by the balance of endocytosis and recycling (Antonescu et al., 2014). We have shown previously that CD8⁺ EMRAs exhibit defective autophagy (Henson et al., 2014), which promotes Glut1 trafficking (Roy et al., 2017). Together with a lack AKT and mTOR phosphorylation (Henson et al., 2014; Plunkett et al., 2007), both of which control Glut1 recycling. EMRA T cells also show defective endocytosis of Glut1, demonstrated by an increased expression of thioredoxin-interacting protein (TXNIP), which has an inhibitory effect on Glut1 internalisation (Supplementary Figure S1C) (Antonescu et al., 2014). These alterations to receptor recycling resulted in differing levels of surface and intracellular Glut1 expression following an 8 h stimulation with anti-CD3 (Figure 3C). This led to CD8⁺ T cell subsets isolated from young individuals having a surface/intracellular ratio of less than one. Whereas the ratio of EMRA subset from old individuals together with all subsets from people living with T2D was above one, suggesting that Glut1 was retained on the cell surface (Figure 3D). However, despite this we did not find an increase in glycolysis, shown by a drop in hexokinase two gene expression from CD8⁺ T cells isolated from old individuals (Figure 3E). Furthermore, CD8⁺ T cells isolated from old and T2D individuals failed to switch to glycolysis, the extracellular acidification rate (ECAR) with the addition of oligomycin did not increase as seen in the young cohort (Figure 3F).

In contrast to glucose uptake, CD8⁺ EMRA T cells from T2D participants significantly increased uptake of the fatty acid palmitate (Figure 4A). Consequently, we used the Seahorse XF Palmitate assay to investigate fatty acid oxidation (FAO). A significant decrease in FAO was observed in CD8⁺ T cells taken from people living with T2D (Figure 4B). As palmitate was not being utilised, we subsequently used Nile Red to assess the accumulation of lipid droplets. This revealed that CD8⁺ EMRA T cells from T2D participants were indeed accumulating lipid (Figure 4C, Supplementary Figure S1E). Unfortunately, these experiments were unable to determine what energy source was being utilised for mitochondrial respiration by the EMRA T cells from people living with T2D (Figure 3A).

The observed alterations in nutrient usage in T2D CD8⁺ EMRA T cells led us to examine the levels of phosphorylated AMP-activated protein kinase (pAMPK), a master regulator of both glucose and lipid metabolism (Hardie et al., 2012). pAMPK was found to be reduced in the T2D CD8⁺ EMRA T cells compared to EMRAs from people without a diagnosis of T2D (Figure 4D, Supplementary Figure S1E). This finding supports our observation that the data is not an artefact of metformin usage, which activates AMPK by increasing its phosphorylation at Thr-172 (Lin and Hardie, 2018). These data suggest that the change in nutrient uptake caused by a failure to phosphorylate AMPK drives metabolic collapse in CD8⁺ EMRA T cells.

Finally, we assessed whether the senescent phenotype was induced by high concentrations of glucose or by inflammatory mediators present in the sera of those with T2D. To mimic the environment of T2D we first cultured young PBMCs in either low (5 mM) or high glucose (25 mM), however we found no difference in the percentage of CD8⁺ EMRA T cells observed between the two conditions (Figure 5A). Then we cultured young PBMCs with either 10% T2D or age-matched control sera. We found no difference in CD8⁺ T cell phenotype between the two conditions at day 3. However, by day 7 the T2D sera resulted in an increase in the CD8⁺ EMRA T cell population (Figure 5B). We then went on to determine the factors present in T2D sera that could be causing the increased number of CD8⁺ EMRA T cells (Figure 5C). While all inflammatory factors investigated were elevated, only C5, CXCL1, IL-6, IL-8, IL-16 and SerpinE1 were significantly upregulated. Both IL-6 and complement factor C5 are known to induce proliferation, expansion, and differentiation of T cells (Lalli et al., 2008; Li et al., 2018). Interestingly, C5, IL-8 and SerpinE1 have been demonstrated to be senescence inducing factors in human plasma irrespective of the senescent stimuli (Basisty et al., 2020). This suggests that T2D sera contains factors capable of driving premature senescence.

While we found no change in T cell phenotype by increasing the concentrations of glucose and palmitate, we did observe more fragmentation of mitochondria (Figure 6A). These observed changes in mitochondrial morphology resemble the mitochondrial dysfunction seen via electron microscopy in CD8⁺ EMRA T cells taken from people living with T2D (Figure 2B). Indeed, when we incubated young PBMCs with low (5 mM) or high (25 mM) glucose we observe a shift to a lower mitochondrial mass in the CD8⁺ EMRA subset (Figure 6B). A lower mitochondrial mass can indicated the presence of smaller more fragmented mitochondria (Siska et al., 2017), and this was confirmed by qPCR. Transcription of genes associated with fission, DNM1L, Fis1 and Mff, were found to be upregulated upon incubation with high glucose (Figure 6C). Furthermore, the changes observed using high glucose were associated with the induction of senescence, demonstrated by an increased expression of p-p53 in the CD8⁺ EMRA subset from young individuals (Figure 6D, Supplementary Figure S1E). This finding was mirrored in people living with T2D, where we observed a higher level of p-p53 expression compared to old age-matched controls (Figure 6E). Again this p-p53 data are suggestive of premature senescence in people living with T2D.

Taken together these data imply that the inflammatory serum factors observed in people living with T2D drives the differentiation of CD8⁺ T cells, the build-up of senescent CD8⁺ EMRA T cells leads to changes in nutrient uptake and usage promoting the accumulation of dysfunctional mitochondria. These poorly functioning mitochondria generate high levels of mtROS, further adding to the inflammatory burden in a detrimental feed-forward loop.