The study complied with the principles of the Declaration of Helsinki, and the use of human blood samples in this study was approved by the St Vincent’s Hospital Melbourne Human Research Ethics Committee (HREC 130/22). Informed consent was obtained from all human participants involved in this study prior to their participation. Confidentiality and privacy were ensured throughout the study, and all data collected were handled and stored in accordance with the ethics approval.

Blood samples were centrifuged at 1,000 g for 10 min in a 50 mL conical tube. The plasma was discarded before carefully pouring the buffy coat layer into a new 50 mL tube. The buffy coat was diluted 1:2 with PBS with 2% foetal bovine serum (FBS) (Thermo Fisher Scientific 26,140-079). All samples that were directly compared were extracted using the same FBS batch. For fresh samples, the entire buffy coat was used. For remnant samples, the volume exceeded what could be processed in one batch so 14 mL of buffy coat was used. PBMCs were isolated using SepMate-50 PBMC isolation tubes (Stem Cell Technologies 85,450) filled with 15 mL of Ficoll-Paque (Sigma Aldrich GE17-5442-02) according to the manufacturer’s instructions. The serum layer was aspirated and discarded without disrupting the PBMC layer. The PBMCs were rapidly poured into a new 50 mL tube. PBMCs were washed by resuspending the pellet in 40 mL of PBS with 2% FBS, centrifuging at 200 g for 8 min and discarding the supernatant. For remnant samples only, erythrocyte lysis was performed as described below. Both fresh and remnant PBMCs were then washed once more with 40 mL of PBS with 2% FBS.

Cells were resuspended in 2 mL (fresh samples) or 20 mL (remnant samples) of RPMI complete medium (RPMI 1640 with GlutaMAX and HEPES (Thermo Fisher Scientific 72,400,047), 10% FBS, 1 × MEM non-essential amino acids (Sigma M7145), 1 mM Sodium pyruvate (Thermo Fisher Scientific 11,360,070)). For the fresh samples, cells were counted after staining with trypan blue (Sigma Aldrich T8154) using the LUNA-II cell counter (Logos Biosystems). For the remnant samples, cells were stained with acridine orange and propidium iodide (Logos Biosystems F23001) and counted using the LUNA-FX7 cell counter (Logos Biosystems).

For remnant samples, cells were frozen in CryoStor CS10 (Stem Cell Technologies 07,930) according to the manufacturer’s instructions.

For DNA methylation arrays, blood was collected into Tempus Blood RNA Tubes (Thermo Fisher Scientific 4342792). These tubes contain stabilization buffer which immediately lyses the cells. RNA was pelleted by centrifugation according to the manufacturer’s instructions. Genomic DNA remains in the supernatant. Previously, Ferrante et al. (2018) demonstrated the genomic DNA can be recovered from the supernatant using phenol:chloroform:isoamyl alcohol. Based on this, we modified the protocol for DNA extraction from whole blood using the QIAamp DNA Micro Kit (Qiagen 56304). This kit recommends using 100 μL of whole blood as input which is equivalent to 400 μL of Tempus RNA blood tube supernatant (the supernatant contains 1 part blood, 2 parts stabilization reagent and 1 part PBS). We added 40 μL of proteinase K and 400 μL of buffer AL to 400 μL of Tempus RNA blood tube supernatant. Samples were pulse vortexed for 15 s then incubated at 56°C for 10 min with shaking at 1500 RPM. After adding 200 µL of ethanol and pulse vortexing for 15 s, samples were incubated at room temperature for 3 min before loading 700 µL into a QIAamp MinElute column. Samples were centrifuge at 6,000 g for 1 min and the flow through discarded. The remaining sample was loaded on the same column and the column was centrifuged again at 6,000 g for 1 min. The wash and elution steps were performed according to the manufacturer’s instructions. Samples were eluted in a volume of 50 μL and typically had concentrations of 75–125 ng/μL (3,750–6,250 ng total recovery) as measured using the Qubit 1x broad range dsDNA kit (Thermo Fisher Scientific Q33266).

Illumina DNA MethylationEPIC arrays were performed by the Australian Genome Research Facility (AGRF). The sample positions on the arrays were randomised.

DNA methylation data was pre-processed and Noob normalised with Minfi version 1.46.0 (Aryee et al., 2014). Epigenetic age was determined from the resulting beta values using PC-Clocks (Higgins-Chen et al., 2022).

Immunophenotyping of PBMCs was performed using the following antibodies: CD3-AF700 (BD biosciences 557,943), CD8-BV605 (BD biosciences 564,115), CD56-PE (Miltenyi 130-113-312), CD19-APC-Vio770 (Miltenyi 130-113-643), CD14-FITC (Miltenyi 130-113-146), CD95-APC (Miltenyi 130-113-005). IgG matched controls were run in parallel: IgG2a kappa-APC (Miltenyi 130-113-269), REA IgG1 APC (Miltenyi 130-113-446), IgG1 k APC (Thermo scientific 17-4714-82).

FACS buffer was cooled on ice before use. The PBMCs were aliquoted at 100,000 cells per well of a 96-well U bottom plate then pelleted by centrifugation at 300 g for 8 min. The cells were resuspended in 50 µL of FACS buffer containing 1:5 diluted FcR blocking reagent (Miltenyi Biotec 130-059-901) then incubated for 10 min at 4°C. To each sample, 50 μL of antibody cocktail in FACS buffer was added and samples were incubated for 20 min at 4°C, protected from light. Cells were washed once with 100 μL of FACS buffer then resuspended in 90 μL of FACS buffer. Immediately before flow cytometry, 10 μL of 1 μg/mL DAPI (Sigma Aldrich D9542) was added to the sample. Data were acquired on a Beckman Coulter CytoFLEX flow cytometer using CytExpert (version 2.4) and analysed using FlowJo software (BD Biosciences).

A compensation matrix was applied to all samples prior to the flow acquisition to correct for fluorescent spill over. The matrix was generated using compensation beads (Thermo Scientific 01-3,333-42) individually stained with each of the 6 antibodies, and PBMCs stained with DAPI. Gating was manually performed for all antibodies except CD95-APC populations. A CD95-APC FMO control was used to set gates for all CD95 populations except for CD3⁺CD8⁺ cells, where two distinct CD95 populations were discernible. Data analysis was performed by an experienced investigator independent to the study and blinded to study objectives. An example of the gating strategy is provided in Supplementary Figure S1.

20S chymotrypsin-like proteasomal activity was measured using the Promega Chromotrypsin-like Cell-Based Proteasome-Glo Assay (G8660), following the manufacturer’s instructions. For the fresh cells, 20,000 PBMCs in 100 µL of complete RPMI were seeded in a white 96 well plate and incubated at room temperature for 2 hours. The assay was miniaturised for the remnant samples and 5,000 PBMCs were seeded in 20 µL of complete RPMI in a black 384 well plate. Both cell dilutions were confirmed to fall within the linear range of the assay (data not shown). The Proteasome-Glo substrate was thawed and equilibrated at room temperature for at least 30 min before adding 20 µL to each well. The plates were briefly centrifuged and then mixed on a plate shaker at 700 RPM for 2 min. Samples were incubated at room temperature for 10 min and luminescence was measured using a Perkin Elmer Victor Nivo plate reader. To normalize the luminescence between plates, a PBMC standard was included on each plate. A proteasome inhibitor control (20 µL lactacystin) was included in initial experiments to confirm the detection of non-specific protease activity was negligible (data not shown).

PBMCs at 3 × 10⁶ cells/mL were seeded with 500 μL per well in 24 well non-TC treated plates (Thermo Scientific 144530). Cells were stimulated by adding 2 μL of lipopolysacharise (LPS, InvivoGen tlrl-b5lps) and phytohemagglutinin (PHA, InvivoGen inh-phap) solution or endotoxin free water as a control. The final concentrations of LPS and PHA were 0.1 μg/mL and 10 μg/mL respectively. Cells were incubated at 37°C, 5% CO₂ for 4 (for CCL3 and IL-6 measurement) or 18 (for IFNγ and IL-2) hours. At the end of the incubation, the supernatant was collected, centrifuged at 300 g for 10 min then transferred to a new tube to remove any cell contamination. Samples were snap frozen on dry ice and stored at −80°C until ELISAs were performed.

Secreted cytokines were quantified by ELISA according to the manufacturer’s instructions (Thermo Scientific, CCL3: 88-7035-86, IL-6: 88-7066-86, IFNγ: 88731686, IL-2: 88-7025-86).

Sample size calculations and data analysis was performed using R statistical software (version 4.1.3).

Sample size calculations were performed using pwr version 1.3–0. For Pearson’s correlation a minimum sample size of 29 was determined using the following parameters: r = 0.5, power = 0.8, alpha = 0.05. For t-tests of means a minimum sample size of 6 per group was determined using the following parameters: power = 0.8, alpha = 0.05, effect size estimated from past studies (Supplementary Table S1). To account for uncertainty in effect size estimates, the minimum sample size was increased by 50%. All comparisons between young and aged donors were performed with at least 6 samples per group.

Pearson’s correlation coefficients were calculated between donor age and tested variables. To compare the mean values between young and aged groups, two-sample t-tests were performed. The young and aged groups were defined as 20–30 and 65–75 years respectively. Benjamini–Hochberg multiple testing correction was applied to adjust p-values for multiple comparisons (Benjamini et al., 2009).

We evaluated the performance of a first- and second-generation DNA methylation clock on fresh whole-blood samples. The Horvath 2018 is a first-generation chronological age predictor trained on skin and blood cells (Horvath et al., 2018). PhenoAge is a second-generation clock trained on phenotypic age estimated from blood-based biomarkers and is a predictor of mortality risk (Levine et al., 2018). We applied the updated principal component (PC) versions of these clocks which aggregate information from many CpGs to reduce the influence of technical noise at any one CpG (Higgins-Chen et al., 2022). Both the Horvath 2018 and PhenoAge clocks accurately predict the chronological age of donors (Figure 1). The Horvath 2018 clock predicted the chronological age with a mean absolute deviation (MAD) of 6.4 years while the PhenoAge clock had a MAD of 4.2 years. Consistent with previous reports, we also detected a modest decline in global DNA methylation levels with age (Supplementary Figure S2).

Surface expression of the transmembrane glycoprotein CD95 is reportedly upregulated in some immune cells with age (Shinohara et al., 1995; Phelouzat et al., 1997; Aggarwal and Gupta, 1998; Jelaska and Korn, 1998; Potestio et al., 1998; Schindowski et al., 2000; Li et al., 2020). Interaction of CD95 with its ligand (CD95L) typically induces apoptosis although activation of other signalling pathways has also been reported (Seyrek et al., 2022). CD95 induced apoptosis plays a role in the elimination of self-reactive B and T lymphocytes during development and the elimination of activated immune cells after an infection has been resolved (Seyrek et al., 2022).

We measured both the proportion of PBMCs expressing CD95 and intensity of the CD95 signal on CD95⁺ cells (median fluorescence intensity, MFI) for PBMCs isolates from fresh blood samples. The proportion of CD95⁺ PBMCs decreased slightly with age although this was not significant after applying a multiple testing correction (Figure 2; Table 2). There was also no significant difference between the young and aged groups.

The proportion of CD3⁺CD8⁺ T cells expressing CD95 increased robustly with donor age, in agreement with previous studies (Shinohara et al., 1995; Phelouzat et al., 1997; Aggarwal and Gupta, 1998). Both the correlation and difference between young and aged groups were significant (Figure 2; Table 2). The proportion of B cells that were CD95⁺ was unchanged with age. However, there was an increase in MFI for the CD95⁺ cells suggesting the level of receptor expression increases with age (Figure 2; Table 2).

In young donors, near 100% of natural killer (NK) T cells expressed CD95. There was a significant decline in CD95⁺ NK T cells with age although the effect size was very small and the difference between young and aged groups was not significant (Figure 2; Table 2). The CD95 MFI trended down with age for NK T cells and the signal was significantly lower for aged donors compared to young donors. This suggests CD95 expression may decline with age in NK T cells, although this requires further validation, and the effect size is likely to be very small.

We found no relationship between the proportion of any cell type and age (Supplementary Figure S2; Supplementary Table S2).

The age dependent CD95 expression by CD3⁺CD8⁺ T cells is an interesting area for future research. There is a potential for more extensive apoptosis after T cell activation. However, activation CD95 receptors only leads to apoptosis in some contexts (Zhang et al., 1996; Varadhachary et al., 1997). Direct measurements of apoptosis and cell survival after T cell activation are needed to determine if this contributes to the decline in immune function reported with age.

The proteasome is a crucial cellular machinery which degrades damaged proteins, eliminates transient proteins, and produces peptides for antigen presentation on MHC class II molecules (Kumar Deshmukh et al., 2019). Proteasomal activity is reported to decline with age in several cell types (Hwang et al., 2007; Papanagnou et al., 2018; Sarkar et al., 2020). This may contribute to aging related pathologies such as cancer, cardiovascular disease, and neurodegenerative diseases which are associated with proteasomal impairment (Chondrogianni et al., 2014; Opoku-Nsiah and Gestwicki, 2018).

We measured chymotrypsin-like proteasomal activity in PBMCs which will capture the cumulative activity of both the proteasome and the immunoproteasome (Johnston-Carey et al., 2016; Murata et al., 2018; Kumar Deshmukh et al., 2019). For PBMCs isolated from fresh blood, chymotrypsin-like proteasomal activity did not depend on donor age (Table 2; Figure 3). For the buffy coat remnant samples, mean proteasomal activity was higher for donors aged 65–75 compared to donors aged 20–30 (Table 2; Figure 3). This is in contrast to the age-dependent decline in proteasomal activity previously reported for PBMCs (Papanagnou et al., 2018) and is not reflected in the correlation (Table 2; Figure 3).

Cytokines are signalling proteins that play a crucial role in inflammation and the immune response. Cytokines are dysregulated with age and this may contribute to chronic inflammatory diseases (Rea et al., 2018).

We stimulated PBMCs with LPS and PHA and measured secreted pro-inflammatory cytokines IL-6, CCL3, IL-2 and IFNγ. IL-6, IL-2 and IFNγ were selected based on previous reports that secretion of these cytokines correlated with donor age (Supplementary Table S3). CCL3 was included as it is secreted by many types of PBMCs and has been implicated in cancer and other diseases (Bhavsar et al., 2015).

All cytokines were measured for PBMCs isolated from fresh samples and IL-6 and CCL3 were validated in PBMCs from remnant buffy coats. Secreted levels of IL-2 were below the limit of detection at 18 h (data not shown). Unstimulated cells secreted extremely low levels of CCL3, IL-6 and IFNγ (Supplementary Figure S4). There was no correlation with age for any of the cytokines tested (Figure 4; Table 2).

There are obvious visual and functional differences between young and aged individuals at the whole organism level. However, recapitulation of these differences in cell culture has proven difficult. Potential reasons for this are discussed below.

Cell culture systems offer the advantage of low cost and high throughput screening for potential aging interventions. However, to conduct meaningful screens, it is crucial to have a model in which age reversal can be measured. The accurate chronological age predictions of epigenetic clocks present an attractive tool for such screens. However, confidence in the use of epigenetic clocks is still lacking due to the unclear relationship between methylation age and clinical outcomes.

Here we have shown that some reported age-related changes in cells may not reproduce on different populations. Our goal was to perform all assays on the same blood samples so we could only test a small subset of the measures reported in the literature. Further investigation is needed to determine if other measures may be more reliable.

At present, there is a lack of reliable measures of aging in cell culture. Due to the complex, multifaceted nature of aging, it is plausible that few robust signatures of aging are present in cells in culture. Instead aging may be better studied using whole organism models such as Drosophila melanogaster, Caenorhabditis elegans or Mus musculus. The potential to perform high throughput screens for aging interventions is an exciting area of research. However, the success of these programs is critically dependent on high-quality models of aging.