Memory Stem T Cells in Autoimmune Disease: High Frequency of Circulating CD8⁺ Memory Stem Cells in Acquired Aplastic Anemia Free

Aplastic anemia (AA), the prototypical bone marrow (BM) failure syndrome, is caused by immune-mediated destruction of hematopoietic stem/progenitor cells (HSPCs) (1). BM transplantation and immunosuppressive therapy (IST) are effective for the treatment of AA (2). Although no Ag has been convincingly demonstrated in AA, autoimmunity to HSPCs has been considered etiologic, mediated by activated CTLs recognizing an HSPC-restricted Ag through their class I or II HLA molecules (3, 4). CD8⁺ CTLs with restricted TCR diversity (oligoclonal T cells) are expanded in AA, leading to the production of proinflammatory cytokines (e.g., IFN-γ) that induces apoptosis of CD34⁺ cells (5). Detection of similar clonotypes, as shown by significantly skewed CDR3 size distribution, suggests the presence of common Ag-driven T cell expansion (6). In AA, there is overrepresentation of HLA-DR2 and class I HLA-A*02:01, -A*02:06, -A*31:01, and -B*40:02 (7, 8). Hematologic recovery after IST with anti-thymocyte globulin (ATG) and cyclosporine (CsA) occurs in 60–75% of AA patients (2, 9), which correlates with diminution of expanded T cell clones. Relapse appears to be associated with re-emergence of the original oligoclonal T cells and, sometimes, new clones (6).

Conventionally, memory T cells are divided into central memory T cell (TCM) and effector memory T cell (TEM) subsets that home to secondary lymphoid and peripheral tissues, respectively (10). The percentages of CD4⁺ and CD8⁺ TEM subsets are increased in peripheral blood (PB) and BM of AA patients (11). Elevated TEM cells with potent effector capacity may relate to abnormal immunity in AA. Recent studies identified a new subset of memory stem T cells (TSCMs) (12, 13), which are the least differentiated cells of all distinct memory populations. These TSCMs express multiple naive markers, as well as the memory Ag CD95. Functionally, they possess an enhanced capacity for self-renewal, can generate multiple memory T cell populations, and likely play important roles in controlling immunity (12). In addition, these TSCMs are endowed with superior immune reconstitution potential in immunodeficient hosts and can mediate antitumor immunity in a humanized mouse model (12).

In autoimmune diseases, there is abnormal CD4⁺ and CD8⁺ T cell activation. We hypothesized dysregulation of the TSCM compartment in autoimmunity. In this study, we evaluated TSCM frequency in AA and its association with severity, treatment response, relapse, and changes after IST. Further, to evaluate TSCMs in other autoimmune diseases, we examined CD4⁺ and CD8⁺ TSCM frequencies in uveitis, systemic lupus erythematosus (SLE), and sickle cell disease (SCD) compared with healthy controls. Our results suggest that TSCMs may contribute to the pathophysiology in cancer immunology, as well as in autoimmune diseases. Specifically, CD8⁺ TSCMs may be a biomarker and a therapeutic target in AA.

PB specimens were collected from 55 AA samples and 41 age-matched healthy donor samples after informed consent was obtained in accordance with the Declaration of Helsinki. Of the 55 AA samples, 21 were analyzed at diagnosis, and 34 were analyzed after IST. Standard criteria were used for diagnosis and disease severity of AA (14, 15). Serial samples were collected before and after IST from 13 patients. The median age of the AA patients was 29 y (range: 13–69 y). All AA patients received horse ATG + CsA + eltrombopag in a clinical research protocol (clinicaltrials.gov, #NCT01623167). For comparison, blood samples were obtained from 34 uveitis patients (27 inactive and 7 active cases), 43 SLE patients who met the American College of Rheumatology criteria for the disease (16, 17) (19 with inactive SLE [SLE disease activity index-2K (SLEDAI-2K) score ≤ 3] and 24 with active SLE [SLEDAI-2K score > 3]), and 5 SCD patients who were receiving frequent transfusions. Demographic and clinical characteristics of patients and healthy controls are summarized in Table I. There were no significant differences in the median age between each patient group and healthy controls (p > 0.05, Supplemental Fig. 1A). All human subjects were enrolled in clinical protocols approved by the National Heart, Lung, and Blood Institute, National Eye Institute, and National Institute of Arthritis and Musculoskeletal and Skin Diseases Institutional Review Boards.

PBMCs were separated from PB samples using Lymphocyte Separation Medium (MP Biomedicals, Solon, OH) and cryopreserved in RPMI 1640 (Life Technologies, Gaithersburg, MD) supplemented with 20% heat-inactivated FBS (Sigma-Aldrich, St Louis, MO) and 10% DMSO, according to the standard protocol, until use.

The following fluorochrome-conjugated mAbs were purchased from commercial vendors and used for surface staining: anti-CD4–V500, anti-CD8–allophycocyanin–H7, anti-CD45RA–PE–Cy7, anti-CD45RO–allophycocyanin, anti-CCR7–Alexa Fluor 700, anti-CD95–PE, anti-CD160–FITC, anti-CD244–FITC, and anti-PD-1–FITC (all from BD Biosciences, San Jose, CA); anti-CD3–BV605 (BioLegend, San Diego, CA); anti-CD14–Pacific Blue and anti-CD19–Pacific Blue (Life Technologies, Carlsbad, CA); and anti-CD27–PC5 (Beckman Coulter, Indianapolis, IN). The fixable violet amine reactive dye (ViViD; Invitrogen/Molecular Probes, Eugene, OR) was used to eliminate dead cells by flow cytometry. For intracellular cytokine staining, the following mAbs were used: anti-granzyme B (GZMB)-FITC, anti-IL-2–FITC, and anti-IFN-γ–FITC (BD Biosciences).

The gating strategy for TSCMs and experimental protocols were adapted from a previous article (18). Briefly, cryopreserved PBMCs were thawed and subjected to surface staining as follows. PBMCs were incubated with a viability marker at room temperature for 20 min, washed, and incubated with anti-human CCR7 at 37°C for 20 min. After washing, cells were stained with a mixture of Abs against CD3, CD4, CD8, CD45RA, CD45RO, CD27, CD95, CD14, and CD19, with or without exhaustion markers (anti–PD-1 Ab, anti-CD160 Ab, and anti-CD244 Ab, respectively), for 30 min on ice. Subsequently, cells were washed, and ≥150,000 events gated on CD3⁺ T cells were acquired with a Fortessa flow cytometer (BD Biosciences) to analyze the frequency of each T cell subset. The gating strategy for T cell subsets is summarized in Fig. 1A. Lymphocytes were gated based on their scatter characteristics, and a forward scatter-area versus forward scatter-height profile was used to exclude cell aggregates and to obtain single lymphocytes. Live T cells were separated from dead cells, monocytes, and B cells in a CD3 versus ViViD/CD14/CD19 bivariate plot. CD4⁺ and CD8⁺ T cells were then gated based on characteristic expression patterns of CCR7 and CD45RO, followed by gating based on CD27/CD45RA and CCR7/CD95 expression. Each T cell subset was defined as follows: TCMs: ViViD⁻ CD3⁺ CD4 (CD8)⁺ CD45RO⁺ CCR7⁺; TEMs: ViViD⁻ CD3⁺ CD4 (CD8)⁺ CD45RO⁺ CCR7⁻; terminally differentiated TEMs (TEs): ViViD⁻ CD3⁺ CD4 (CD8)⁺ CD45RO⁻ CD45RA⁺ CCR7⁻ CD27⁻; naive T cells (TNs): CD3⁺ CD4 (CD8)⁺ CD45RO⁻ CD45RA⁺ CCR7⁺ CD27⁺ CD95⁻; and TSCMs, CD3⁺ CD4 (CD8)⁺ CD45RO⁻ CD45RA⁺ CCR7⁺ CD27⁺ CD95⁺. Data were analyzed using FlowJo software version 9.6 (TreeStar, Ashland, OR). The quantification of inhibitory receptor expression in T cell subsets was described in detail (19).

Expression levels of GZMB, IL-2, and IFN-γ in CD4⁺ and CD8⁺ T cell subsets were analyzed by intracellular cytokine staining 6 h poststimulation. Briefly, cells were stimulated by addition of Dynabeads Human T-Activator CD3/CD28 and then 2 h later by the addition of Golgi transport inhibitor (GolgiPlug; BD Biosciences). After another 4 h of culture, cells were incubated with the cell surface–staining Ab mixture, as described in Immunostaining for surface Ags, and were fixed/permeabilized using Cytofix/Cytoperm Fixation and Permeabilization Solution (BD Biosciences), according to the manufacturer’s protocol. Subsequently, intracellular cytokine staining was performed using anti-GZMB–FITC, anti-IL-2–FITC, and anti-IFN-γ–FITC at 4°C for 30 min.

All statistical analyses were performed using GraphPad PRISM version 6.0 (GraphPad, La Jolla, CA). Data are shown as mean ± SEM. A Student t test was used to calculate statistical significance between two groups. A statistical analysis was performed using one-way or two-way ANOVA with the post hoc Tukey or Dunnett test for multiple comparisons, when appropriate. The Spearman rank test with linear regression was used for correlation analysis. A two-tailed p value < 0.05 was considered statistically significant.

First, we measured five T cell subsets (TN, TSCM, TCM, TEM, and TE) in AA and healthy controls. Within the CD4⁺ or CD8⁺ T cell compartments, AA patients showed decreased CD4⁺ or CD8⁺ TN frequency (p < 0.05, Fig. 1B), compared with controls, consistent with previous reports (11). CD4⁺ TE frequency was very low in the CD4⁺ T cell compartment in both AA and controls, but CD8⁺ TE frequency was higher among CD8⁺ T cells in both. In healthy controls, TSCMs represented a relatively small percentage of circulating CD4⁺ or CD8⁺ T cells (median 2.4% CD4⁺ TSCMs and 2.1% CD8⁺ TSCMs), confirming the findings of Gattinoni et al. (12). Samples collected from the same healthy donors but on different dates showed similar results, confirming technical and biological reproducibility (Supplemental Fig. 2). A significantly higher CD8⁺ TSCM frequency was detected in AA patients (4.2% versus 2.1%, p < 0.05), whereas there was no difference in CD4⁺ TSCM frequency (p > 0.05) compared with controls (Fig. 1C, 1D). Within the AA group, CD8⁺ TSCMs (4.2%) were more frequent than CD4⁺ TSCMs (2.1%) (p < 0.05, Fig. 1C), whereas the frequencies of CD4⁺ and CD8⁺ TSCMs within the control group showed no differences.

We assessed TSCM subset correlations with clinical manifestations and treatment responses in the AA cohort. CD4⁺ and CD8⁺ TSCM populations were evaluated in AA patients by clinical parameters, including age, sex, severity, and responses to IST. Responses to IST were defined according to established criteria (20). In AA (n = 21), CD8⁺ TSCM frequency was measured at diagnosis, and response was assessed at 3 mo post-IST (Fig. 2A). In AA, a high CD8⁺ TSCM frequency at diagnosis correlated with a complete response (CR) or partial response (PR) to IST (5.0% in CR and PR versus 2.8% in nonresponders [NRs], p < 0.05) (Fig. 2A). In AA patients prior to IST (n = 21), CD8⁺ TSCM frequency did not correlate with age, sex, absolute neutrophil count, platelet count, time from diagnosis to therapy, or serum ferritin levels (p > 0.05, Supplemental Fig. 3A). To elucidate the effects of IST on the TSCM populations in AA patients, we next compared CD4⁺ or CD8⁺ TSCM frequency among three groups: 21 AA subjects without IST, 34 AA subjects with IST (3 mo to 2 y post-IST), and 41 healthy controls. CD8⁺ TSCMs were significantly increased in the two AA cohorts (with or without IST) relative to controls (p < 0.05), but CD4⁺ TSCM frequency was not statistically different among the three groups (p > 0.05) (Fig. 2B). We further evaluated CD4⁺ or CD8⁺ TSCM frequency among three cohorts (27 responders [CR or PR], 7 NRs or relapsed cases after IST, and 41 healthy donors). CD4⁺ TSCM frequency was not significantly different, but a higher CD8⁺ TSCM frequency was observed in the 7 NRs or relapsed cases after IST compared with the 27 responders (CR or PR) and with the 41 controls. A higher CD8⁺ TSCM frequency after IST was associated with treatment failure (3.5% in responders versus 5.5% in NRs or relapse, p < 0.05) (Fig. 2C). In serially collected samples from 13 AA patients, CD8⁺ TSCM frequency tended to decrease or remain the same after IST, with the exception of two patients (relapse at 9 mo and NR at 3 mo) who showed marked expanded CD8⁺ TSCMs after IST (Fig. 2D).

In AA patients, a high CD8⁺ TSCM frequency at diagnosis correlated with better response to IST, whereas an increased CD8⁺ TSCM frequency was observed in NRs or relapsed cases after IST, suggesting TSCM as a potential biomarker. Memory T cell subsets often are described by their effector functions: TCMs are distinguished by greater proliferative and IL-2–producing capacities, and TEMs are characterized by increased secretion of effector cytokines, such as IFN-γ and GZMB. Stimulation with anti-CD3/CD28 beads successfully induced cytokine production by CD4⁺ and CD8⁺ T cells from AA and healthy controls (Fig. 3A). Elevated IFN-γ and IL-2 levels were observed in both total CD4⁺ and CD8⁺ T cell populations in AA patients compared with healthy controls (Fig. 3B, 3C). We next investigated which CD4⁺ or CD8⁺ T cell subsets (TNs, TSCMs, TCMs, TEMs, and TEs) were responsible for cytokine (GZMB, IFN-γ, and IL-2) production (Fig. 3B, 3C). Consistent with observations in total CD4⁺ and CD8⁺ T cell populations, no significant differences were detected in the frequencies of GZMB-producing CD4⁺ and CD8⁺ T cell subsets in AA patients compared with normal controls (Fig. 3B, 3C). In the CD4⁺ T cell compartment, IFN-γ and IL-2 levels were significantly elevated in the TSCM, TCM, and TEM subsets in AA. T cell subsets in the CD8⁺ T cell compartment showed greatly increased IFN-γ levels in CD8⁺ TSCM, TCM, TEM, and TE subsets and moderately elevated IL-2 levels in the CD8⁺ TSCM and TCM subsets. Taken together, elevated IFN-γ and IL-2 levels were seen in CD4⁺ and CD8⁺ TSCMs in AA compared with healthy controls (Fig. 3B, 3C).

We next compared CD4⁺ or CD8⁺ TSCM frequency between each patient group (AA, uveitis, SLE, or SCD) and a healthy control group (Table I). We chose uveitis as an organ-specific immune-mediated disorder characterized by inflammatory ocular lesions, SLE as a systemic autoimmune disease, and SCD as controls for transfusions. Among the four patient groups, the uveitis group displayed a reduction in CD4⁺ TSCM frequency (1.8%) relative to the healthy controls (2.4%, p < 0.05). An elevated CD8⁺ TSCM frequency was observed in AA (4.2%), uveitis (3.6%), and SCD (4.3%), but not in SLE, compared with controls (2.1%, p < 0.05) (Fig. 4A). CD4⁺ and CD8⁺ TSCM frequencies were positively correlated in patients with AA (r = +0.45, p = 0.0010), uveitis (r = +0.67, p < 0.0001), or SLE (r = +0.47, p = 0.0024) but not in healthy controls (r = +0.14, p = 0.42) (Fig. 4B). CD4⁺ TSCM frequency was inversely correlated with TEM frequency in AA (r = −0.38, p = 0.0045), uveitis (r = −0.58, p = 0.0003), and SLE (r = −0.49, p = 0.0009) (Supplemental Fig. 4A). For CD8⁺ TSCM and TE frequencies, the same inverse correlation was also observed in AA (r = −0.39, p = 0.0044) and uveitis (r = −0.55, p = 0.0010) (Supplemental Fig. 4B). There was no correlation between TSCM frequency and age (r = +0.03, p = 0.70 for CD4⁺ TSCMs; r = +0.06, p = 0.41 for CD8⁺ TSCMs, Supplemental Fig. 1B, 1C).

Aberrant overexpression of PD-1 (a cell surface inhibitory receptor, also known as an exhaustion marker of T cells) is associated with persistent activation of self-reactive T cells in autoimmune diseases (21–23). We measured surface PD-1 expression on individual CD4⁺ or CD8⁺ T subsets (focusing on TSCMs) in four groups (AA, uveitis, SLE, and healthy control) (Fig. 5A). In CD4⁺ T cell subsets, PD-1 expression was lower on TSCMs than on TCMs, TEMs, and TEs within clinical groups (AA, uveitis, SLE, and healthy controls, p < 0.05), identifying TSCMs as the least exhausted population among CD4⁺ T cells. When PD-1 expression levels on individual CD4⁺ T cell subsets were compared between each patient group and controls, significantly higher PD-1 expression levels were seen on TEs in uveitis and on TCMs, TEMs, and TEs in SLE (p < 0.05, Fig. 5B). In CD8⁺ T cell subsets, PD-1 expression also was significantly lower on TSCMs than on TCMs and TEMs in all four groups (p < 0.05) (Fig. 5B). Significantly higher PD-1 expression levels were seen on CD8⁺ TEMs in AA. These results were consistent with previous reports demonstrating that PD-1 expression correlated with the differentiation of CD4⁺ and CD8⁺ T cells (24, 25). When PD-1 expression on CD4⁺ or CD8⁺ TSCMs was compared among AA, uveitis, SLE, and controls, it was elevated on CD4⁺ TSCMs in SLE and on CD8⁺ TSCMs in AA and uveitis (p < 0.05, Fig. 5C).

To verify the results for PD-1 expression in T cell subsets, the expression of other inhibitory receptors (CD160 and CD244) was examined in T cell subsets from various autoimmune diseases (19, 26, 27). CD160 and CD244 were primarily expressed in CD8⁺ T cell subsets, among which the CD8⁺ TSCM population was less exhausted compared with the other T cell subsets (Fig. 5A, 5B). CD160, as well as PD-1, was highly expressed in CD8⁺ TSCMs of AA compared with healthy donors (Fig. 5C).

Finally, we assessed TSCM subset correlations with clinical manifestations and treatment responses in the uveitis and SLE cohorts. CD4⁺ and CD8⁺ TSCM populations were evaluated in patients with or without some clinical parameters, including age, sex, severity, and immune therapies specific for individual diseases. In uveitis, patients (n = 34) were classified into two groups, with or without current IST, and CD8⁺ TSCM frequency was compared between the two groups. The group receiving therapy included patients treated with prednisolone (PSL), mycophenolate mofetil (MMF), CsA, methotrexate (MTX), anti-TNF therapy, or a combination. A higher CD8⁺ TSCM frequency (4.8%) was observed in uveitis patients without any systemic immune therapy compared with those receiving immune therapies (3.2%) (p < 0.05, Fig. 6A). A higher CD8⁺ TSCM frequency was observed in uveitis patients not receiving PSL (4.1%) than in those receiving PSL (2.8%), as well as in those patients not receiving the anti-TNF agents infliximab and adalimumab (4.0%) than in patients receiving these drugs (2.4%) (p < 0.05, Fig. 6A). However, there were no statistically significant differences in patients with or without CsA, MTX, or MMF (p > 0.05, Supplemental Fig. 3B). CD8⁺ TSCM frequency did not correlate with age, sex, diagnosis of underlying disease (sarcoidosis versus Vogt–Koyanagi–Harada disease versus Birdshot), localization of disease, or disease activity (p > 0.05) (Supplemental Fig. 3B).

Because most SLE patients had received IST before or at the time of sampling, we could not examine therapy effects on the TSCM population. There was a higher frequency of CD4⁺ TSCMs in patients not receiving hydroxychloroquine (HCQ; 3.3%) than in those receiving HCQ (2.3%) (p < 0.05, Fig. 6B), but no relationship was seen for any other drug (Supplemental Fig. 3C). CD4⁺ TSCM frequency was not correlated with age, disease activity (p > 0.05; Supplemental Fig. 3C), or the presence of specific disease associations (including secondary Sjögren’s syndrome, antiphospholipid syndrome, or immune thrombocytopenic purpura) (data not shown).

Collectively, immune therapies appeared to have negative effects on the TSCM population in the uveitis and SLE patient cohorts, as well as in AA.

In our flow cytometry–based study, we retrospectively analyzed CD4⁺ and CD8⁺ TSCM populations and their relationship to disease activity, treatment response, and disease severity by characterizing these subsets in a total of 178 subjects (55 AA, 34 autoimmune uveitis, 43 SLE, 5 SCD, and 41 age-matched healthy controls). An increased CD8⁺ TSCM population was observed in AA, uveitis, and SCD, and a higher tendency of CD4⁺ TSCM population was seen in SLE. Activation of CD4⁺ and CD8⁺ T cells is critical in diseases of uncontrolled inflammation (5, 28, 29), but we believe the current study to be the first report of CD4⁺ and CD8⁺ TSCM populations in autoimmune diseases. Consistent with previous studies, we observed ∼2–4% TSCMs in the total CD4⁺ or CD8⁺ T cell population in healthy controls (12, 18), and there was no correlation between CD4⁺ or CD8⁺ TSCM frequency and age in any group (30).

TSCMs have been examined in infectious diseases (25, 31, 32), in the development of immunotherapies and vaccines for cancer (33, 34), in allogeneic hematopoietic stem cell transplantation (35, 36), and in adult T cell leukemia (37). TSCMs represent the earliest and longest-lasting developmental stage of memory T cells and exhibit a gene expression profile between TNs and TCMs (12, 18). One mechanism for the maintenance of long-term T cell memory may be the unique homeostatic properties of TSCMs (38). In this context, we observed an inverse correlation between the CD4⁺ or CD8⁺ TSCM population and the CD4⁺ TEM or CD8⁺ TE population, respectively, indicating that TSCM differentiation leads to a shift in these cell proportions as a possible mechanism to re-establish the homeostasis of the immune system. TSCMs may be progenitors for TEMs or TEs, because these two proportions were inversely correlated. It is unknown whether immune activation is attributable to increased TSCMs or the increase in this subset is the result of aberrant immune responses. Because of their capacity to generate all memory and effector T cell subsets, we hypothesized that the increased frequency of TSCMs contributed to the progression of autoimmune diseases.

Indeed, a higher CD8⁺ TSCM frequency at diagnosis or after IST was associated with a better response to IST or treatment failure, respectively, in AA, suggesting CD8⁺ TSCMs as a potential biomarker. Longitudinal analysis of 13 AA patients also suggested a potential use for TSCMs as a disease biomarker. In the uveitis cohort, the CD8⁺ TSCM population was lower in patients who received PSL and anti-TNF therapy, indicating a potential negative role for IST in this population, although the analysis of serial samples would be required to support this conclusion. Further, SLE patients receiving HCQ therapy displayed a reduced CD4⁺ TSCM frequency, also suggesting that IST affected this population. HCQ is known to correlate negatively with the percentage of CD45RO⁺ cells among CD4⁺ T cells (39). Because most SLE samples were collected after IST, IST might have masked the TSCM population. Additional studies in large patient cohorts are necessary to define whether TSCMs are useful as general autoimmune disease biomarkers.

There are similarities and differences in T cell autoimmunity in AA, uveitis, and SLE. Abnormalities of regulatory T cells and Th17 cells are reported in patients with AA (40–42), uveitis (43, 44), or SLE (45–47). CD8⁺ CTLs with restricted TCR diversity are expanded in AA and secrete proinflammatory cytokines, which induce apoptosis of CD34⁺ cells (6). A recent study suggested that the clonally restricted expansion of Th1 cells is most likely Ag-driven and induces an inflammatory environment, exacerbating the functional impairment of regulatory T cells (40). In uveitis, a CD4⁺ T cell–driven disease is the dominant paradigm in animal models (28). However, Ag-specific CD8⁺ T cells can mediate autoimmunity within the eye (48), and CD8⁺ T cells increase in experimental uveitis (49). SLE is characterized by the production of a wide array of autoantibodies and, thus, has traditionally been classified as a B cell disease. However, evidence indicated that the assistance of Th cells might be required to produce SLE-related inflammation (50).

Upon TCR stimulation in healthy controls, TSCMs exhibit effector activity, including TNF-α, IFN-γ, and IL-2 secretion, whereas TNs remain relatively quiescent (12). In AA patients, CD4⁺ and CD8⁺ TSCMs produced more IFN-γ and IL-2 upon TCR stimulation compared with healthy controls, showing the effector functions and greater IL-2–producing capacity of these subsets. CD8⁺ CTLs are expanded in AA and produce proinflammatory cytokines, such as IFN-γ, which induce apoptosis of CD34⁺ cells (5). In this study, we identified TSCMs as another source of IFN-γ production by CD4⁺ and CD8⁺ T cells in AA.

PD-1 is one of the immune checkpoint molecules expressed on activated and exhausted T cells (51). PD-1 overexpression is detected in CD4⁺ and CD8⁺ T cells in AA patients (21). It was demonstrated in SLE patients that PD-1 expression is upregulated in CD4⁺ T cells, and that PD-1 expression in CD4⁺ T cells is associated with IFN-γ expression in CD3⁺ T cells (52). In this study, PD-1 expression in TSCMs identified this subset as the least exhausted population relative to other memory T cell subsets. Previous studies demonstrated PD-1 expression on tumor-infiltrating T cells (TILs) and its correlation with prognosis (53, 54). The PD-1 pathway exerts inhibitory functions in chronic viral infections and tumors, with special relevance to autoimmunity (23). In addition to well-documented negative-regulatory roles, PD-1 expression on CD8⁺ TILs accurately identified the repertoire of clonally expanded tumor-reactive cells (55). The higher expression of PD-1 in CD8⁺ TSCMs in AA suggests this population as self-reactive T cells. Recent studies suggested that PD-1 expression in healthy donors does not correlate with lower functionality but rather with differentiation to an effector memory phenotype (19, 26, 56). We interpret increased expression of PD-1 and increased IFN-γ production in CD8⁺ TSCMs of AA patients as evidence of clonal expansion, as occurs for CD8⁺ TILs (55) and as we reported previously for CD8⁺ effector cells in AA (6, 57). Enhanced cytolytic effector activity in CD8⁺ TSCMs of AA was also supported by the observation of higher CD160 expression of this population.

Our study has a number of limitations. We were unable to demonstrate causality. However, it seems likely that an increase in TSCMs may be etiologically associated with immune response to autoantigens and that the mechanism may be related to homeostatic maintenance of other memory T cell subsets. It is also plausible that increased TSCMs are a consequence of nonspecific inflammation induced by environmental factors, because CD8⁺ TSCMs were also increased in SCD, which is not typically considered an autoimmune disease. Transfusion is chronic immune stimulation, which may drive SCD to an active inflammatory state (58, 59). Transfusions administered to AA and SCD patients may affect the CD8⁺ TSCM population as a result of allogeneic stimulation. However, this seems unlikely because CD8⁺ TSCM frequency did not correlate with serum ferritin levels in AA (Supplemental Fig. 3A). Although we were unable to characterize the Ag-specific TSCM population because of the limited cell numbers, a nonhuman primate model of SIV infection demonstrated that SIV-specific TSCMs preferentially survive after Ag elimination compared with other memory subsets and are fully functional, even in chronic infection (38). Thus, Ag-specific TSCMs are presumably able to contribute directly to disease progression.

In conclusion, we provide evidence for increased circulating CD8⁺ TSCMs in AA, underscoring the importance of this novel subset in the regulation of immune responses and pathogenesis of autoimmunity. Our work described previously unknown potential roles for TSCMs in AA, such as cytokine secretion correlated with effector functions. Understanding the CD8⁺ TSCM population may offer new therapeutic strategies and novel mechanistic insights into various autoimmune diseases. Longitudinal analysis of AA patients suggests the potential use of TSCMs as disease biomarkers. Additional studies in large patient cohorts are required to validate our data, which were obtained from small numbers of patients with autoimmune diseases.

We thank Marie Desierto, Susan Wong, and Pilar Fernandez for technical assistance; Olga Rios, Kinneret Broder, and Carolyne Smith for assistance with obtaining patient and healthy volunteer samples; and Barbara Weinstein, Sarfaraz Hasni, and Zerai Manna for obtaining patient clinical information.

The authors have no financial conflicts of interest.