Impact of cryopreservation on immune cell metabolism as measured by SCENITH

Abstract

The dynamic functioning of immune cells is regulated by cellular metabolic processes, and there is growing interest in the study of immunometabolic correlates of dysfunctional immune responses. SCENITH is a novel flow cytometry-based technique that allows for ex vivo metabolic profiling of immune cells within heterogeneous samples. Cryopreservation of clinical samples is frequently undertaken to facilitate high throughput processing and longitudinal analyses of immune responses, but is thought to lead to cellular metabolic dysfunction. We aimed to investigate the impact of cryopreservation on immune cell metabolism, harnessing SCENITH’s unique ability to describe the divergent bioenergetic characteristics of distinct immune cell subsets. We demonstrate that upon activation, T cells are unable to sufficiently/readily undergo metabolic reprogramming. Additionally, we find that cryopreservation introduces a time-dependent metabolic artefact that favours glycolysis and impairs oxidative phosphorylation, suggesting that cryopreservation results in mitochondrial dysfunction. Despite this artefact, SCENITH was still able to reveal the distinct bioenergetic profiles of contrasting immune cells populations following cryopreservation. Whilst SCENITH can provide valuable information about immune cell metabolism even in cryopreserved samples, our findings have important implications for the design of future studies. Investigators should carefully consider how to process and store clinical samples to ensure that cryopreservation does not confound analyses, particularly where longitudinal sampling is required.

Introduction

Immune cell behaviour is underpinned by tightly regulated cellular metabolic processes, whereby metabolic pathway intermediaries are directly implicated in modulating the epigenetic landscape necessary for the dynamic functioning of immune cells [1]. Naïve T cells maintain low levels of transcription and translation, largely meeting their limited bioenergetic requirements by catabolising glucose and amino acids via oxidative phosphorylation [2]. Conversely, T cell activation, differentiation, and effector functions are associated with higher overall cellular energetic expenditure, which is achieved by broadly upregulating cellular metabolism, and aerobic glycolysis in particular [3–5]. A deeper understanding of the immunometabolic correlates of physiological and dysfunctional adaptive immune responses in health and disease offers the potential to identify novel immunomodulatory therapeutic targets in the treatment of infection, auto-immune disease, and cancer.

Single Cell ENergetIc metabolism by profiling Translation inHibition (SCENITH) is a novel functional assay which can be used to metabolically profile individual immune cells using flow cytometry. Compared to bulk analysis techniques such as extracellular flux analysis (Seahorse), SCENITH can facilitate a comparison between distinct cell populations in mixed samples (e.g. whole blood), including those that occur at a relatively low frequency. Furthermore, SCENITH does not require purification of immune cells for metabolic profiling, which is highly advantageous when analysing clinical samples which may be of a limited volume. Protein translation is an energetically expensive process, the rate of which is closely correlated with the overall rate of adenosine triphosphate (ATP) synthesis in immune cells [6]. In SCENITH, cells are treated with puromycin, which inhibits protein synthesis by becoming incorporated into—and thus causing the premature termination of—nascent polypeptide chains [7]. The rate of puromycin incorporation is quantified by staining cells with anti-puromycin antibody before measuring its mean fluorescence intensity (MFI). Anti-puromycin MFI acts as the assay’s readout and acts as a proxy measure for the cell’s overall energetic expenditure [6].

Additionally, SCENITH takes advantage of selective inhibition of specific metabolic pathways that contribute to cellular biosynthesis of ATP (Fig. 1). 2-deoxy-D-glucose (2DG) is a glucose analogue that inhibits hexokinase (HK) and thus glycolysis [8]. Oligomycin-A (OMA) is an ATP synthase (ATPS) inhibitor which inhibits oxidative phosphorylation [9]. Comparing the anti-puromycin MFI of cells treated in parallel with 2DG, OMA, or both (‘DGO’) versus an uninhibited positive control (‘P-Co’) facilitates the calculation of a cell’s reliance upon glycolysis, oxidative phosphorylation, or other fuels (the oxidation of fatty acids and amino acids, Fig. 2).

Argüello et al. originally describe performing SCENITH on whole blood, and it has subsequently been applied to peripheral blood mononuclear cell (PBMC) samples [10, 11], including samples which have been cryopreserved [12]. Cryopreservation of PBMCs prior to their use in phenotypic and functional assays allows for the biobanking of samples, and for cheaper, more practical, and higher-throughput processing of multiple samples in parallel. However, environmental changes directly influence cellular bioenergetics [13], meaning that cryopreservation may introduce a systematic bias that compromises the reliability and reproducibility of metabolic assays.

The impact of cryopreservation on cellular bioenergetics measured via the Seahorse assay has previously revealed progressive mitochondrial dysfunction and a shift towards dependence upon glycolysis [14, 15]. However, these studies analysed mixed PBMC samples using a bulk analysis assay. Given that distinct immune cell populations have divergent bioenergetic characteristics, the conclusions that can be drawn from the existing literature are therefore inherently limited. In this study, we aim to investigate the impact of cryopreservation on immune cell metabolism as measured by SCENITH, with a particular focus on T cells.

Materials and Methods

Study design and participant characteristics

This study was conducted using peripheral blood mononuclear cells (PBMCs) from five consenting healthy participants (n = 5). Participants were Caucasian, aged 35.2 ± 8.7, and had a body mass index (BMI) within the healthy range. Two participants were male and three were female.

Timepoints of 15, 35 and 70 days were selected pragmatically to provide an indication of metabolic changes induced by an increasing duration of cryopreservation.

PBMC isolation and cryopreservation

Venous blood was collected from each participant into EDTA vacutainers (BD). This was diluted 1 : 1 in PBS and then applied to Ficoll-Paque PLUS (Cytiva) at a 3 : 1 concentration, and centrifuged at 1256 g for 25 min, with minimum acceleration and no braking. PBMCs were extracted from the interphase and washed with PBS 1% foetal bovine serum (FBS, Gibco) twice. The total number of cells prior to cryopreservation were counted diluted 1 : 1 in Trypan Blue dye (Lonza), using a haemocytometer. Cells were suspended into 1 ml of freezing media containing 90% FBS and 10% Dimethyl sulfoxide (DMSO) (Sigma-Aldrich) and immediately stored at − 80°C for 24 h in a MrFrosty container (Nalgene) containing propan-1-ol. The cryovials were then transferred to liquid nitrogen (−196°C) for storage.

Thawing of PBMCs and assessment of cell recovery and viability

PBMCs were thawed at the appropriate time point by incubating the cryovial in a water bath for 2 min at 37°C. The thawed or fresh (on day 0 before cryopreservation) PBMCs were transferred into a fresh falcon tube and gently resuspended in 14 ml of human plasma-like media (HPLM) (Gibco) containing 10% dialysed FBS (dFBS) (Gibco). Cells were counted using a haemocytometer, diluted 1 : 1 with Trypan Blue dye, to determine total number of cells and total number of viable cells post-thawing, from which the cell recovery and viability was calculated.

SCENITH assay, flow cytometry staining and data analysis

SCENITH was performed as outlined by Argüello et al. [6]. PBMCs were plated at between 0.5–10 x10⁶ cells per well into 96-well plates. Six conditions—differing in T cell activation status, addition of metabolic inhibitors and addition of puromycin—were plated simultaneously (Table 1). PBMCs were stimulated using TransAct (Miltenyi Biotec) at 1 : 300 titre or left to rest in HPLM for 2 h at 37°C, 5% CO₂. PBMCs were treated with DMSO, 100 mM 2-Deoxy-D-Glucose (Merck), 1 μM Oligomycin A (Sigma-Aldrich) or a combination of both inhibitors, for 15 min at 37°C, 5% CO₂. 11 μg/ml puromycin (Sigma-Aldrich) was added to the relevant wells and incubated for 30 min. PBMCs were washed with cold PBS and incubated with viability dye Zombie Aqua (Biolegend) for 10 min at room temperature. PBMCs were washed with PBS 1% Bovine Serum Albumin (BSA, Sigma Aldrich) and incubated for 30 min at 4°C with antibodies diluted in PBS 1% BSA. PBMCs were fixed and permeabilized with Foxp3 kit (eBioscience), as per manufacturer's instructions. PBMCs were incubated for 60 min at 4°C with antibodies for intracellular markers diluted in Perm-wash (eBioscience), including anti-puromycin-AF488 antibody (see Supplementary Table 1 for a full list of antibodies used). PBMCs were analysed using a LSRFortessa™ X-20 (BD) within 24 h of processing. Metabolic capacities and dependencies were calculated by comparing the anti-puromycin AF488 GeoMFI of different conditions as outlined in Fig. 2.

Statistical analysis

Flow cytometry data was analysed using FlowJo v10.8.1 and plug-ins FlowSOM v.3.0.18 [16] and ClusterExplorer v.1.7.4. Statistical analyses were performed using GraphPad Prism v.9.5.1. Comparisons between timepoints were made using the Friedman test and Dunn’s multiple comparisons test (fresh versus day 15, day 35 and day 70), or using the Wilcoxon matched-pairs signed rank test and two-stage false discovery rate correction (fresh versus day 70 only). Comparisons between immune cell subsets at the same point were made using the Kruskal Wallis test and Dunn’s multiple comparisons test (for 3 or more subsets) or using the Mann-Whitney test (for 2 subsets, e.g. CD69⁺ versus CD69⁻). In text, averages are expressed as mean ± standard error of the mean. In figures, error bars indicate the standard error of the mean. For derived calculations (dependencies and capacities), occasionally values of <0% or >100% are obtained, conceptually equivalent to values of 0 and 100% respectively. Where this has occurred, raw data has been adjusted accordingly.

Results

The effect of cryopreservation on immune cell recovery, immune cell viability and sample composition

As expected, the cellular recovery (Fig. 3A) and viability (Fig. 3B) of PBMCs decreased after cryopreservation and thawing. After 35 days, there was a significant decrease in the percentage recovery (P = .0197) and viability (P = .0270) when compared with fresh PBMCs, a trend which continued at 70 days. The lowest mean recovery was 60.24 ± 0.89% on day 70 (Fig. 3A) whilst the lowest mean viability was 82.90 ± 1.60% on day 35 (Fig. 3B).

The impact of cryopreservation on PBMC cellular composition was assessed by analysing cell viability on manually-gated subpopulations, specifically: classical, intermediate and non-classical monocytes (respectively CD14⁺ CD16⁻, CD14⁺ CD16⁺ and CD14⁻ CD16⁺; Fig. 3C), CD56^Bright and CD56^Dim NK cells (Fig. 3D), CD4⁺ T cell subsets (Fig. 3E) and CD8⁺ T cell subsets (Fig. 3F). A full gating strategy is outlined in Supplementary Fig. 1. Additionally, the FlowSOM clustering algorithm was used to identify 15 distinct populations within PBMC samples based on cellular expression of lineage and T cell differentiation markers (CCR7, CD45RA) within our panel (Supplementary Fig. 2). Manually gated (Fig. 3G) and FlowSOM (Fig. 3H) populations were visualized by superimposing them onto a map created using the unsupervised nonlinear dimensionality reduction algorithm t-Distributed Stochastic Neighbour Embedding (tSNE) [16]. There is a strong visual correspondence between FlowSOM clusters and manually gated populations, and between tSNE and FlowSOM clustering (Figs 3G–H and 4). There was no significant difference in the frequency in either manually gated (Fig. 3I) or FlowSOM populations (Fig. 3J) at fresh and day 70, indicating that cryopreservation universally decreased PBMC viability but did not impact upon PBMC composition in this study.

To minimize duplication, we present data for the most abundant T cell memory subsets only (i.e. CD4⁺ TEM, CD8⁺ TCM), rather than for all T cell memory subsets defined above.

The effect of cryopreservation on the T cell translational response to activation

A general trend of increased non-specific binding was demonstrated at day 70 when compared to fresh (Fig. 5A), but this was only statistically significant within the CD4⁺ naïve (P = .0099) and classical monocyte (P = .0099) populations (Fig. 5A). The non-specific binding was significantly different between NK CD56^Dim and CD8⁺ Naïve (P = .0021), NK CD56^Dim and classical monocytes (P = .0003) and CD8⁺ TEM and classical monocytes (P = .0161) at fresh, however these significant differences were not found at day 70 (Fig. 5A). Together, these results indicate that the degree of non-specific binding varies according to cell type and may be influenced by factors such as cryopreservation, highlighting the value of including a no-puromycin negative control (N-Co) when performing SCENITH.

At fresh, resting and activated CD8⁺ T cell subsets undertook more protein translation than CD4⁺ T cell subsets. In resting CD4⁺ and CD8⁺ cells (UA-Co), naïve subsets performed lower levels of translation than memory subsets. After activation (P-Co), translation levels performed by all subsets increased, with CD4⁺ and CD8⁺ naïve subsets undertaking more translation than their respective memory subsets, suggesting that despite their lower baseline translational activity, protein translation of naïve cells is more rapidly upregulated than in memory cells following 2 h of activation with an anti-CD3/CD28 agonist (Fig. 5B).

There was a general decrease in the percentage translational response after cryopreservation for all T cell subpopulations (Fig. 5C–F). Compared to fresh, there was a significant decrease in the percentage increase of translation after activation within the CD4⁺ Naïve T cells at day 70 (P = .0099) (Fig. 5C), and CD8⁺ Naïve T cells at day 15 (P = .0429) and day 70 (P = .0212) (Fig. 5E).

Metabolic characteristics of immune cell subsets, and the effect of cryopreservation on immune cell inhibition by 2-deoxy-D-glucose and oligomycin a

At fresh, classical monocytes were the immune cell subset which undertook the highest level of protein translation, undertaking approximately twice as much translation as non-classical monocytes (mean anti-puromycin GeoMFI of 14075 versus 6969, respectively). Lymphocyte subsets had a lower level of translation, potentially reflecting their smaller size in comparison to monocytes (mean anti-puromycin GeoMFI ranging from 3262 (CD4⁺ TCM) to 4987 (CD8⁺ naïve), Fig. 6A).

Both classical and non-classical monocyte populations had the highest mean levels of glucose dependence (75 and 86%, respectively), whereas lymphocyte subsets had mean glucose dependencies of under 20% (Fig. 6B). When considering mitochondrial dependence, further differences between classical and non-classical monocytes became apparent, with classical monocytes having a significantly lower mean mitochondrial dependence (45 versus 99% respectively, P = .0091, Fig. 6C). Compared to other lymphocyte subsets, NK CD56^Dim had a higher mitochondrial dependence (<75 versus 92%), potentially reflecting the activation and subsequent switch to aerobic glycolysis for non-NK cell lymphocyte subsets following the addition of TransAct.

When considered together, these three different indicators of immune cell metabolic activity can clearly delineate between the manually gated subsets specified by our panel.

Firstly, there was no statistically significant difference in the inhibitory activity of the DGO combination at any timepoint and for any cell subset (Fig. 6D–L), indicating that the basal level of inhibition was unchanged by cryopreservation.

OMA demonstrated a trend towards decreased inhibitory activity in all cell subsets as the duration of cryopreservation increased (Fig. 6D–L). Compared to fresh, the inhibitory activity of OMA was significantly reduced at day 70 in non-classical monocytes (P = .0018) (Fig. 6F), NK CD56^Bright (P = .0212) (Fig. 6G), NK CD56^Dim (P = .0018) (Fig. 6H), CD4⁺ TCM (P = .0429) (Fig. 6J) and CD8⁺ TEM (P = .0044) (Fig. 6L) cells.

2DG demonstrated a trend towards increased inhibitory activity in the lymphocyte populations (Fig. 6G–L), while maintaining a consistent percentage inhibition in the monocyte populations (Fig. 6D–F). In lymphocytes, 2DG inhibition was maximal at day 35 (Fig. 6G–L). Compared to fresh, the inhibitory activity of 2DG was significantly increased at day 35 in NK CD56^Bright (P = .0099) (Fig. 6G), NK CD56^Dim (P = .0099) (Fig. 6H), CD4⁺ Naïve (P = .0212) (Fig. 6I), CD4⁺ TCM (P = .0099) (Fig. 6J), CD8⁺ Naïve (P = .0099) (Fig. 6K) and CD8⁺ TEM (P = .0429) (Fig. 6L) cells.

Despite the artefact introduced by cryopreservation, the metabolic profiles of classical versus non-classical monocytes remained distinct following cryopreservation and followed the same pattern as when analysed fresh, with classical monocytes having higher levels of protein translation (anti-puromycin GeoMFI 14897 versus 10569, P = .055556), similar levels of glucose dependence (87 versus 92%, P = .746032), and lower levels of mitochondrial dependence (7 versus 81%, P = .007937, data not shown).

The metabolic differences observed between CD69⁺ and CD69⁻ T cells are preserved following cryopreservation

As a cytometry-based technique, a core strength of SCENITH compared to bulk analysis assays is the ability to compare between cells subsets within heterogeneous samples (e.g. PBMCs) based upon cellular expression of phenotypic markers. CD69 is an early activation marker expressed by T cells and is associated with divergent immune cell metabolic signatures when measured via other single-cell techniques [17]. We therefore asked whether CD69 expression was associated with divergent T cell metabolic profiles when measured via SCENITH, and whether this was still evident following cryopreservation.

CD69⁺ and CD69⁻ populations of the naïve and most abundant memory subset of CD4⁺ and CD8⁺ T cells are compared at fresh. Within the naïve CD4⁺ and CD8⁺ populations, CD69 expression was associated with a significant increase in protein translation, indicating increased metabolic activity (anti-puromycin GeoMFI of 1932 versus 3550 and 3205 versus 5234, P = .0159 and 0.0317 respectively, Fig. 7B). Within memory CD4⁺ and CD8⁺ populations, CD69 expression was also associated with a trend towards increased protein translation, but this difference was not statistically significant (2620 versus 3459 and 3353 versus 3718, P = .3095 and .8413 respectively, Fig. 7B). CD69 expression was not associated with a particular pattern of glucose dependence (Fig. 7C). Conversely, there is a clear trend towards increased glycolytic capacity associated with T cell CD69 expression, including in CD4⁺ naïve (3.38 versus 38.55%, P = .0079), CD4⁺ TCM (5.65 versus 37.50%, P = .0317), CD8⁺ naïve (3.80 versus 34.05%, P = .0159) and CD8⁺ TEM (10.83 versus 44.11%, P = .0317) populations (Fig. 7D). Together, these results demonstrate that lymphocyte CD69 expression is associated with increased cellular metabolic activity, met by an upregulation of glycolytic capacity. Although CD69 expression was not associated with higher glucose dependence in this study, a longer duration of activation with an anti-CD3/CD28 agonist may produce different results.

Next, we considered the impact of cryopreservation on the metabolic characteristics of CD69⁺ and CD69⁻ T cell subpopulations, and we observe patterns consistent with those seen for T cell subsets as a whole (Fig. 6G–L). Following cryopreservation, CD69⁺ and CD69⁻ subpopulations also show a trend towards increased 2DG inhibition (Fig. 7E–H) and decreased OMA inhibition (Fig. 7I–L). However, there was no significant change found in the difference in glucose dependency (Fig. 7M–P) and glycolytic capacity (Fig. 7Q–T) between CD69⁻ and CD69⁺ cells in any CD4⁺ and CD8⁺ subsets across the timepoints investigated in this study, demonstrating that the metabolic differences observed between CD69⁺ and CD69⁻ T cells are preserved following cryopreservation.

Discussion

SCENITH is a novel functional assay which can be used to metabolically profile immune cells using flow cytometry. In this study, we demonstrate that although naïve T cells are less metabolically active than memory T cells at rest, they are more readily activated following the addition of an anti-CD3/28 agonist (Fig. 5). Secondly, we demonstrate that classical and non-classical monocytes have divergent metabolic characteristics, with non-classical monocytes being less metabolically active and having greater mitochondrial dependence (Fig. 6), consistent with prior transcriptomic comparisons of these subsets [18]. Additionally, we show that within a mixed PBMC sample treated with an anti-CD3/28 agonist, CD69 expression is associated with increased metabolic activity, increased glycolytic capacity and diminished mitochondrial dependence (Fig. 7). This demonstrates the ability of SCENITH to reveal the divergent metabolic characteristics of cell subsets within a mixed sample without the need for purification.

Specifically, this study aimed to investigate the impact of cryopreservation on immune cell metabolism as measured by SCENITH, and we present a number of key findings. Firstly, we show that T cells undergoing activation with a CD3/CD28 stimulus are less readily metabolically reprogrammed following cryopreservation (Fig. 5). Secondly, we find that cryopreservation introduces a metabolic artefact that favours glycolysis and impairs oxidative phosphorylation, as indicated by the increasing inhibitory potential of 2DG and diminishing inhibitory potential of OMA following cryopreservation, with our results suggesting that this artefact becomes more pronounced as the duration of cryopreservation increases (Fig. 6). This was observed in all cell types defined by our multiparametric panel and was not explained by a cryopreservation-induced change in sample cellular composition (Fig. 3). Despite this artefact, SCENITH was still able to reveal the distinct bioenergetic profiles of contrasting immune cells populations following cryopreservation, such as classical and non-classical monocytes (Fig. 6), or T cell subsets stratified by CD69 expression (Fig. 7). As such, we believe that SCENITH can provide valuable insights into how cellular metabolism influences immune cell function even after cryopreservation.

This is the first study to investigate the impact of cryopreservation on immune cell bioenergetics using SCENITH, but our findings are in line with previous studies. Utilising the bulk analysis assay Seahorse, other groups have demonstrated that cryopreservation causes a time-dependent impairment of immune cell mitochondrial function, alongside an enhanced glycolytic response [14, 15]. In Seahorse, spare respiratory capacity is a derived measurement that quantifies the ability of cells to generate additional ATP following an increase in energetic demand [19]. This metric becomes impaired following cryopreservation [15], analogous to our finding of diminished T cell translational responses to CD3/CD28 activation (Fig. 5). Currently, robust evidence as to the mechanistic basis for mitochondrial dysfunction following cryopreservation in immune cells is lacking, but the literature focussing on other cell types (e.g. reproductive cells or placental tissue) highlights the possible role of oxidative stress caused by excessive reactive oxygen species generation, or osmotic stress that compromises the integrity of the mitochondrial membrane [20–23].

Our findings have implications for the design of future studies. Upon study conception, investigators should consider whether the metabolic artefact introduced by cryopreservation will impact upon their ability to satisfactorily address their research question. Performing SCENITH on cryopreserved samples may not be appropriate for studies where participants are sampled longitudinally and biobanking and batch analysis is planned. In these circumstances, cryopreservation is likely to confound results and result in spurious findings. Conversely, intra-donor comparisons between conditions from the same cryopreserved sample is likely to result in valid findings. For example, researchers may wish to investigate the metabolic differences between drug-tested versus control cells, which should not be problematic if cells have been processed and stored equally.

There are several important limitations associated with this study. Firstly, we sampled a modest number of healthy donors, limiting the generalisability of our findings to other settings. For example, in the context of acute severe infection, we might expect a greater degree of upregulation of glycolysis in certain T cell subsets than was possible to achieve via our protocol with 2 h of activation with a CD3/CD28 agonist. Applying this study design to disease states such as acute severe infection may therefore yield different results, with metabolically divergent subsets variably susceptible to artefacts induced by cryopreservation. Secondly, for this study we followed cryopreservation and sample storage protocols that we routinely use for our clinical samples. It is indeed possible that differences in cryopreservation methods may impact upon cellular bioenergetics [15]. The findings that we present are therefore most relevant for the protocols we describe in our methods section. Thirdly, the process of incubating cells with culture medium to achieve T cell activation may have introduced a metabolic bias compared to analysis directly after thawing. Fourthly, the addition of an anti-CD3/CD28 agonist will have directly activated T cells, but not NK cells or monocytes. It is plausible that the metabolic characteristics of NK cells and monocytes may have been altered by this step. Finally, we were only able to study four time-points, limiting our ability to predict the impact of cryopreservation beyond 70 days.

Acknowledgements

The authors wish to acknowledge the assistance of Dr Andrew Herman and Poppy Miller, and the University of Bristol Faculty of Biomedical Sciences Flow Cytometry Facility.

Author contributions

Curtis Luscombe (Conceptualization [equal], Formal analysis [equal], Investigation [equal], Methodology [equal]), Eben Jones (Conceptualization [equal], Formal analysis [equal], Investigation [equal], Methodology [equal]), Michaela Gregorova (Conceptualization [supporting], Formal analysis [supporting], Methodology [supporting], Supervision [supporting]), Nicholas Jones (Methodology [supporting]), and Laura Rivino (Conceptualization [equal], Formal analysis [supporting], Methodology [supporting], Supervision [lead])

Supplementary data

Supplementary data is available at OXFIMM Journal online.

Conflict of interest: L.R. holds the position of Editorial Board Member for Oxford Open Immunology and is involved neither in the review process nor in any decision-making on the manuscript.

Funding

This research was supported by the Elizabeth Blackwell Institute, University of Bristol, and funded in whole, or in part, by the Wellcome Trust [Grant number—The WT ISSF3 grant number is 204813/Z/16/Z] and by grants from the Royal Society (RGS\R1\221078) and the Academy of Medical Sciences (Springboard Award SBF007\100173) to LR. NJ is funded by an MRC New Investigator Research Grant (MR/X000095/1). For the purpose of Open Access, the author has applied a CC BY public copyright licence to any Author Accepted Manuscript version arising from this submission.

Data availability

Access to raw FCS files will be facilitated upon written request to the corresponding author.

Ethics approval

Samples were obtained and used under research ethics approval of the Bristol Biobank (NHS Research Ethics Committee Ref 20/WA/0273, use application U0046) and handled in line with the Human Tissue Act.

References

Author notes