Effector memory-type regulatory T cells display phenotypic and functional instability

Abstract

Regulatory T cells (T_reg cells) hold promise for sustainable therapy of immune disorders. Recent advancements in chimeric antigen receptor development and genome editing aim to enhance the specificity and function of T_reg cells. However, impurities and functional instability pose challenges for the development of safe gene-edited T_reg cell products. Here, we examined different T_reg cell subsets regarding their fate, epigenomic stability, transcriptomes, T cell receptor repertoires, and function ex vivo and after manufacturing. Each T_reg cell subset displayed distinct features, including lineage stability, epigenomics, surface markers, T cell receptor diversity, and transcriptomics. Earlier-differentiated memory T_reg cell populations, including a hitherto unidentified naïve-like memory T_reg cell subset, outperformed late-differentiated effector memory-like T_reg cells in regulatory function, proliferative capacity, and epigenomic stability. High yields of stable, functional T_reg cell products could be achieved by depleting the small effector memory-like T_reg cell subset before manufacturing. Considering T_reg cell subset composition appears critical to maintain lineage stability in the final cell product.

Fig. 1.. Composition of subsets defining T cell differentiation stages suggests coherence in memory formation between bulk CD4⁺ T cells and T_reg cells.

(A) Gating strategy for defining conventional CD4⁺ T cells and CD4⁺CD25⁺FoxP3⁺ T_reg cells from PBMCs. Illustrated are density plots of flow cytometry data of one representative donor [fluorescence minus one (FMO) controls shown in fig. S1]. SSC-A, Side Scatter Area; FSC-A, Forward Scatter Area; FSC-H, Forward Scatter Height. (B) Summary of ex vivo investigation of defining conventional CD4⁺ T cells and T_reg cell subset distribution of freshly isolated PBMCs based on flow cytometry (A). Statistical analysis for differences between cell subsets by two-way analysis of variance (ANOVA) with Holm-Sidak testing for multiple comparisons is shown. n = 53. Black lines in violin plots show the median. *P ≤ 0.05; ****P ≤ 0.0001. ns, not significant. Scatter plot diagrams show Pearson correlation analysis between (C) bulk conventional CD4⁺ T cells and their subsets, as well as (D) bulk T_reg cells and T_reg cell subsets with age. Age (in years) and frequencies of T_reg cell subsets are shown on the x and y axes, respectively. (E) Scatter plot diagrams show Pearson correlation analysis between frequencies of T_reg cell and conventional CD4⁺ T cell subsets. Frequencies of conventional CD4⁺ T cells and T_reg cell subsets are shown on the x and y axes, respectively. n = 53. The respective Pearson correlation coefficients (r) are reported in each plot. *P < 0.05 and ****P < 0.00005.

Fig. 2.. Phenotypical and functional ex vivo characterization of CD4⁺ T_CONV cell and T_reg cell subsets.

(A) MFI of CD25 and FoxP3 of CD4⁺CD25⁺FoxP3⁺ T_reg cells. n = 12. Black lines indicate the mean with SEM. Ex vivo isolated and unstimulated PBMCs were analyzed by flow cytometry on the basis of extracellular, intracellular, and intranuclear proteins labeled with fluorochrome-conjugated monoclonal antibodies. n = 12. Frequencies within T_reg cell and conventional CD4⁺ T cell subsets of CD31 (B), Ki-67 (C), and CD134 (D). n = 6. Results are presented as means ± SEM. Normalized (E) CD45RO, (F) CD45RA, (G) GZMA, (H) ceramide synthase 6 (CERS6), and (I) KLRG mRNA expression in ex vivo fluorescence-activated cell sorted (FACSorted) CD4⁺CD25⁺FoxP3⁺ T_reg cell subsets. Depicted are floating bars indicating minimum and maximum values and a line indicating the mean. n = 3. (J) Gating strategy for analyzing ex vivo proinflammatory cytokine production by conventional CD4⁺ T cells and CD4⁺CD25⁺FoxP3⁺ T_reg cells. PBMCs were stimulated with phorbol 12-myristate 13-acetate (PMA)/ionomycin for 6 hours, permeabilized, and intracellularly stained for IFN-γ (K) and IL-2 (L). Illustrated are density plots of flow cytometry data of one representative donor. n = 32 for IFN-γ and n = 24 for IL-2 expression, respectively. Black lines in violin plots show the median. In (A) to (L), statistical significance for differences between cell subsets was determined by two-way ANOVA with Tukey multiple comparison correction *P ≤ 0.05; **P ≤ 0.01; ***P ≤ 0.001; ****P ≤ 0.0001. (M) Suppression capacity of ex vivo isolated T_reg cell subsets. Freshly isolated PBMCs were FACSorted for T_reg cell subsets and CD4⁺CD25⁻ T_CONV cells serving as T_RESP cells and to control for T_reg cell–specific proliferation suppression. Carboxyfluorescein diacetate succinimidyl ester (CFSE)–labeled T_CONV cells were cocultured with increasing numbers of T_reg cells and stimulated with anti-CD3/CD28–coupled microbeads. After 96 hours of culture, proliferation was assessed on the basis of CFSE signal. n = 3. Results are presented as means ± SEM.

Fig. 3.. TCR diversity and FoxP3-locus methylation analyses of the distinct T_reg cell memory differentiation subsets.

(A) TCRβ repertoire analysis of ex vivo FACSorted T_reg cell subsets. n = 2. Mean data are shown. (B) Intuitive view depicting the relative proportions of the top 10 clones of each T_reg cell subset of one donor. (C) Ex vivo FACSorted T_reg cell subsets were analyzed by quantitative polymerase chain reaction (PCR) to define the percentage of TSDR demethylation. n = 3. Results are presented as means ± SEM. (D) Weighted average DNA methylation in partially methylated domains (PMDs) and highly methylated domains (HMDs) based on reduced representation bisulfite sequencing (RRBS) data. Weighted average methylation in PMD and HMD: RRBS data to determine the weighted average DNA methylation across defined DNA segments, divided into PMD and HMD. A loss of methylation is observed in the order: T_regN > T_regNLM > T_regCM > T_regEM. The loss of methylation is more prominent in PMDs. n = 6. Statistical significance for differences between T_N, T_SCM, T_CM, and T_EM cells but not bulk conventional CD4⁺ T cells was determined by t test.

Fig. 4.. CITE-seq between CD4⁺ T cells and T_reg cells.

(A to F) UMAP representation of FACSorted T_reg cells (n = 7) and CD4⁺ T cells (n = 4) based on top 1000 highly variable genes. (A) Color-encoded cluster membership of unsupervised graph-based clustering using the Louvain community detection method. (B) Distribution of manually gated subsets is shown. Subsets were defined by applying the flow cytometry gating strategy on CITE-seq surface epitope expression levels. (C) Distribution of external cell type identities. Single cells were automatically labeled leveraging two publicly available datasets from bulk RNA-sequenced FACSorted established T_H and T_reg cell populations. T_FH, T follicular helper. (D to F) Gene expression of canonical T_reg cell markers. Blue to red rainbow color scale denotes low to high expression. (G) Heatmap representation of average expression of top differentially expressed genes between manually gated subsets.

Fig. 5.. Trajectory analysis confirms T cell differentiation into distinct T_reg and T_H cell lineages.

(A) UMAP representation based on top 1000 highly variable genes (same as Fig. 4) color-encoded for pseudo-time and UMAP-embedded, higher-dimensional differentiation trajectories for T_reg cell (left) and T_CONV cell (right) lineages. (B) Distribution of cell identity labels and clonal expansion along the lineage trajectories. Pseudo-time values were quantized and binned into uniform cellular distribution for each lineage. Bottom: Average pseudo-time values per bin. Top: Distribution charts for labels assigned by graph-based clustering, manual gating annotation, and reference-based annotation with the same color-coded as in Fig. 4 (A to C). Cells that could not be assigned a label (none) or a clonotype are shown in light-gray. (C) Heatmaps depict the expression of surface epitope markers (top) and expression of differentially regulated genes over pseudo-time and between lineages (bottom). Expression levels were smoothed over pseudo-time trajectories depicted in (A) and (B), respectively. Hallmark differentiation genes, T_reg cell functional genes, and master transcription factors are highlighted. ADT, antibody derived tag.

Fig. 6.. FACS strategy for isolating T_reg cell subsets—expansion capacity, assessment of T_reg cell subset–specific phenotypic, and functional characteristics during expansion.

(A) FACS strategy for isolating T_reg cell subsets. The lymphocyte population was gated for singlets, and CD4⁺ T cells and bulk T_reg cells were further defined as CD25^highCD127^low. From bulk T_reg cells, T_regCM and T_regEM cells were defined as CD45RA⁻CCR7⁺ and CD45RA⁻CCR7⁻, respectively. CD45RO⁺ cells were excluded to further identify and isolate CD95⁺CCR7^low T_regNLM cells and CD95⁻CCR7⁺ T_regN cells. All T_reg cell populations sorted for expansion are highlighted. FSC-W, Forward Scatter Width; SSC-H, Side Scatter Height; SSC-W, Side Scatter Width. (B) T_reg cell population purity was assessed after FACS. (C) FACSorted T_reg cell populations were stimulated with anti-CD3/CD28–coated microbeads on day 1 and weeks 1, 2, 4, and 6 (indicated by dots) in the presence of rhIL-2 and rapamycin. Phenotypic and proinflammatory cytokine analysis was performed after 3, 5, and 7 weeks of expansion (indicated by triangles). (D and E) Fold expansion of T_reg cell subset–derived cells from a close-up view of weeks 2 to 3 (D) and weeks 2 to 7 (E). n = 6. Results are presented as means ± SEM. (F to I) Examination of the T_reg cell subsets phenotypically and functionally during in vitro expansion. (F) Change of T_reg cell lineage markers CD25 and FoxP3 of T_reg cell subset–derived cells over the entire expansion period of 7 weeks. n = 6. Results are presented as means ± SEM. (G) Changes in the initial naïve/memory phenotype during 7 weeks in culture using their CD45RA and CCR7 expression profile. (H and I) Proinflammatory cytokine production of cultured T_reg cell–derived subsets. T_reg cell subset–derived cells were stimulated at indicated time points with PMA/ionomycin for 6 hours and stained for IFN-γ (H) and IL-2 (I). For (F) to (I), n = 6. Results are presented as means ± SEM. For (G), statistical significance for differences between cell subsets was determined by two-way ANOVA with Tukey multiple comparison correction. *P ≤ 0.05; **P ≤ 0.01; ****P ≤ 0.0001.

Fig. 7.. Dynamics of transcription factor expression and T_reg cell subset–specific functional and epigenetic characteristics during expansion.

(A) Expression patterns of transcription factors GATA3, RORγt, and Tbet versus FoxP3 in expanded bulk T_reg cells and T_reg cell subsets. FACSorted T_reg cell populations were cultured according to strategy presented in Fig. 6 (A and B). Cells were stained intracellularly for CD3 and respective transcription factors. Illustrated are density plots of flow cytometry data of one representative donor at day 35 after culture initiation. (B) Bulk T_reg cells and T_reg cell subsets were analyzed for the transcription factors GATA3, RORγt, Tbet, and FoxP3 after 21, 28, 35, and 42 days of expansion. Quantified data from three donors are presented. Results are presented as connected data points (days) and mean. (C) Cytokine production profiles of expanded T_reg cell subsets: IL-4, IL-5, IL-10, IL-13, IL-17, GM-CSF, IFN-γ, and TNFα cytokine production was evaluated in supernatants derived from bulk T_reg cell–and T_reg cell subset–derived cultures at indicated time points (days 21, 28, and 35) upon 24 hours of αCD3/28 stimulation (Meso Scale Diagnostics assay). n = 3. (D and E) Suppression capacity of expanded T_reg cell subset–derived cells. In vitro expanded T_reg cell subset–derived populations were cocultured with freshly isolated autologous CFSE⁺CD3⁺ T_RESP cells and stimulated with αCD3/CD28-coated microbeads for 96 hours. Proliferation of CD4⁺ and CD8⁺ T_RESP cells was analyzed by CFSE dilution. Percentage suppression of CD4⁺ and CD8⁺ T_RESP cell proliferation at different T_RESP:T_reg cell ratios after 3 weeks of expansion. n = 6. Results are presented as means ± SEM. (F) Expanded T_reg cell subset–derived cells were analyzed by bisulfite amplicon sequencing of the TSDR to define the percentage of TSDR demethylation (weeks 3, 5, and 7). n = 6. Results are presented as means ± SEM. For (F) to (H), statistical significance for differences between cell subsets was determined by two-way ANOVA with Tukey multiple comparison correction. *P ≤ 0.05; **P ≤ 0.01; ***P ≤ 0.001; ****P ≤ 0.0001.

Fig. 8.. Improved T_reg cell products by prior depletion of EM-like T_reg cells.

(A) Analyses of the effect of depleting the EM-like T_reg cell subset before manufacturing the T_reg cell product. We used a FACS to deplete EM cells (CD45RA⁻CCR7⁻) from the initial T_reg cell population (T_reg − EM). n = 6. Illustrated are density plots of flow cytometry data derived from the FACS of one representative donor. (B) Fold expansion of bulk T_reg cell–and T_reg − EM–derived cells of days 7 to 21. n = 6. (C) T_reg cell lineage markers CD25 and FoxP3 of bulk T_reg cell–and T_reg − EM–derived cells at day 21. n = 6. Black lines in violin plots show the median. (D) Memory phenotypes of bulk T_reg cell–and T_reg − EM–derived cells at day 21. Black lines in violin plots show the median. n = 6. (E) Expanded bulk T_reg cell–and T_reg − EM–derived cells were analyzed to define the percentage of TSDR demethylation at day 21. Black lines in violin plots show the median. n = 6. (F to H) Impact of T_regEM cell subset depletion on effector cytokine production. IL-4, IL-5, IL-13, GM-CSF (F), IL-17, IL-10 (G), IFN-γ, and TNFα (H) cytokine production was evaluated in supernatants derived from bulk T_reg cells and T_reg − EM at day 21 upon 24 hours of αCD3/28 stimulation. n = 3. (I and J) To evaluate the propensity of T_reg cells to dedifferentiate into T_H17 cells, a T_H17 cell differentiation assay was performed. T_reg cells were isolated and either left as bulk T_reg cells or depleted of T_regEM cells. T_reg and T_reg − EM cells were expanded for 14 days and then exposed to inflammatory conditions for another 7 days prior. Cells were restimulated with PMA/ionomycin and stained intracellularly for IL-17A. (I) Representative flow cytometry density plots and quantification (J) of IL-17A production of bulk T_reg cell (left) and T_reg − EM cultures, IL-6Rα knockout (KO), gp130 KO and wild-type (WT) T_reg cells, respectively. n = 5. Statistical significance for differences between cells was determined by paired t test, two-tailed. *P ≤ 0.05.