Deficit of circulating CD19+ CD24hi CD38hi regulatory B cells in severe aplastic anaemia.

Immune aplastic anaemia (AA) is caused by cytotoxic T lymphocytes (CTLs) that destroy haematopoietic stem and progenitor cells. Enhanced type 1 T helper (Th1) responses and reduced regulatory T cells (Tregs) are involved in the immune pathophysiology. CD24hi CD38hi regulatory B cells (Bregs) suppress CTLs and Th1 responses, and induce Tregs via interleukin 10 (IL-10). We investigated circulating B-cell subpopulations, including CD24hi CD38hi Bregs, as well as total B cells, CD4+ T cells, CD8+ T cells and natural killer cells in 104 untreated patients with severe and very severe AA, aged ≥18 years. All patients were treated with standard immunosuppressive therapy (IST) plus eltrombopag. CD24hi CD38hi Bregs were markedly reduced in patients with AA compared to healthy individuals, especially in very severe AA, but residual Bregs remained functional, capable of producing IL-10; total B-cell counts and the other B-cell subpopulations were similar to those of healthy individuals. CD24hi CD38hi Bregs did not correlate with responses to IST, and they recovered to levels present in healthy individuals after therapy. Mature naïve B-cell counts were unexpectedly associated with IST response. Markedly reduced CD24hi CD38hi Bregs, especially in very severe AA, with recovery after IST suggest Breg deficits may contribute to the pathophysiology of immune AA. Published 2020. This article is a U.S. Government work and is in the public domain in the USA. British Journal of Haematology published by British Society of Haematology and John Wiley & Sons Ltd.

Introduction

Immune aplastic anaemia (AA), caused by T cells that destroy haematopoietic stem and progenitor cells (HSPCs), is successfully treated with immunosuppressive therapy (IST). Cytotoxic T‐lymphocytes (CTLs) , , , and type 1 T helper (Th1) cytokines, , , , , such as interferon‐γ, are implicated in pathophysiology of immune AA. Frequent human leucocyte antigen (HLA) class I allele loss in HSPCs of patients with AA are evidence of involvement of CTLs. , , ,

In addition to CTL activation, IFN‐γ directly inhibits HSPCs, , , , , , and blocks haematopoietic signals of thrombopoietin. Eltrombopag, a thrombopoietin receptor agonist, can stimulate HSPCs in the presence of IFN‐γ, which may in part contribute to its efficacy in AA. ,

Regulatory T cells (Tregs) that suppress effector T cells are reduced in AA and their numbers increase in response to IST, suggesting a role of Tregs in the immune pathophysiology. , A Treg subpopulation correlates with response to IST. Recent studies have revealed immune suppressive functions in regulatory B cells (Bregs). Multiple suppressive B‐cell subpopulations with different target cells and functions are collectively referred to as Bregs. Bregs maintain tolerance and immune homeostasis via interleukin 10 (IL‐10). Human CD19+CD24hiCD38hi B cells are a well‐studied Breg population; they suppress CTLs and Th1 responses, and convert CD4+ T cells into Tregs via IL‐10, programmed death receptor‐ligand 1, CD80, CD86 and CD1d. , , , The quantity and function of CD24hiCD38hi Bregs are impaired in various immune disorders, such as systemic lupus erythematosus, rheumatoid arthritis, immune thrombocytopenia and chronic graft‐versus‐host disease.

B cells and their subpopulations are not well characterised in AA. Peripheral blood CD19+ B‐cell frequencies are higher in AA compared to myelodysplastic syndrome. IL‐10 producing B cells in bone marrow are reduced in AA, suggesting that Breg defects might contribute to the pathophysiology. In the present study, we evaluated peripheral blood B‐cell subpopulations, including CD24hiCD38hi Bregs, as well as total B cells, CD4+ T cells, CD8+ T cells and natural killer (NK) cells, at diagnosis and 6 months after institution of IST plus eltrombopag, and further tested their correlations with clinical presentations and outcomes after therapy.

Patients and methods

Patients

A total of 125 previously untreated patients with AA, aged ≥18 years, were included in this retrospective study. All patients were treated with standard IST (horse anti‐thymocyte globulin [ATG] and cyclosporine) plus eltrombopag in the National Institutes of Health (NIH) Clinical Center, Bethesda, between July 2012 and October 2018, under protocols approved by the Institutional Review Board of the National Heart, Lung, and Blood Institute (clinicaltrials.gov, NCT01623167). Treatment regimens have been described: all patients continued cyclosporine at least until 6 months after institution of IST, eltrombopag was discontinued at 3 months in 25 patients (20%) or continued until 6 months in the remaining patients.

Definitions

Severe AA (SAA) was diagnosed when at least two of the following three criteria were met: an absolute neutrophil count (ANC) of <0·5 × 109/l, an absolute reticulocyte count (ARC) of <60 × 109/l and a platelet count of <20 × 109/l. Very severe AA (VSAA) was defined as an ANC of <0·2 × 109/l in addition to the criteria for SAA. Haematological response to IST was evaluated at 6 months after the institution of ATG, consistent with our previous studies: , response was defined as blood counts not meeting criteria for SAA, and non‐response as blood counts that continued to meet the SAA criteria.

Samples

Blood was sampled within the 4 weeks preceding IST and at 6 months after institution of ATG, after obtaining written informed consent in accordance with the Declaration of Helsinki. Fresh blood anti‐coagulated with ethylene diamine tetra‐acetic acid was used to quantify absolute numbers of B cells, CD4+ T cells, CD8+ T cells and NK cells (TBNK subsets). For B‐cell subpopulation analysis, peripheral blood mononuclear cells (PBMCs) were separated from heparinised blood using Lymphocyte Separation Medium (MP Biomedicals, Santa Ana, CA, USA) and were cryopreserved in Roswell Park Memorial Institute (RPMI)‐1640 medium (Life Technologies, Gaithersburg, MD, USA) with 20% heat‐inactivated fetal bovine serum (FBS; Sigma‐Aldrich, St. Louis, MO, USA) and 10% dimethyl sulphoxide, according to the standard protocol until use.

Immunostaining of lymphocytes

Absolute numbers of B cells, CD4+ T cells, CD8+ T cells and NK cells were analysed with BD Multitest™ 6‐color TBNK reagent (BD Biosciences, San Diego, CA, USA). The following cell surface markers were used to define B cells, CD45+CD3−CD19+CD16−CD56−; CD4+ T cells, CD45+CD3+CD4+CD8−; CD8+ T cells, CD45+CD3+CD4–CD8+; and NK cells, CD45+CD3–CD19−(CD16/CD56)+ (Figure S1A).

B‐cell subpopulations in cryopreserved PBMCs were studied as previously described with modifications. , Cell counts and viabilities of thawed PBMCs were assessed by automated cell counter with trypan blue, and 1 × 106 live PBMCs were stained for flow cytometry analysis. The CD3cCD19+CD33–glycosylphosphatidylinositol (GPI)‐anchor+ live B cells were classified based on their expressions of CD24 and CD38: CD24hiCD38hi Bregs, CD24lowCD38low mature naïve B cells, CD24hiCD38low memory B cells and CD24lowCD38hi plasma cells/plasmablasts (Fig 1A, Figure S1B). GPI‐anchor deficient B cells were excluded to avoid underestimation of CD24hiCD38hi Bregs using fluorescent‐labelled aerolysin variant (FLAER) because CD24 is a GPI‐anchored protein. Absolute numbers of B‐cell subpopulations were estimated using their frequencies and mononuclear cell counts in the clinical record. To assess IL‐10 production from B cells, 1 × 106 PBMCs were cultured in the presence of 10 µg/ml of CpG (ODN 2006, Thermo Fisher Scientific, Grand Island, NY, USA) and 1 µg/ml of recombinant human CD40 ligand (CD40L; R&D, Minneapolis, MN, USA) in 200 µl of RPMI‐1640 supplemented with 10% FBS, 1% penicillin/streptomycin and 1% glutamate for 24 h, as previously published, and a cell activation cocktail, including phorbol‐12‐myristate 13‐acetate, ionomycin and brefeldin A (BioLegend, San Diego, CA, USA), was added during the last 6 h of culture, but brefeldin A was not used for negative controls. Intracellular IL‐10+ cell frequencies were quantified in CD3–CD19+ live B cells and their subpopulations, CD24hiCD38hi Bregs and non‐Bregs (Fig 3A, Figure S1C). The following reagents and fluorochrome‐conjugated monoclonal antibodies were used to stain B cells: live/dead fixable blue dead cell stain kit for UV excitation (Thermo Fisher Scientific), anti‐CD19‐BV421 (BioLegend), anti‐CD3‐V500, anti‐CD24‐BV605, anti‐CD33‐PE, anti‐CD38‐APC (all from BD Biosciences), FLAER‐Alexa Fluor 488 (Cedarlane, Burlington, NC, USA), anti‐IL‐10‐PE‐Cy7 (eBioscience, Vienna, Austria) and fixation/permeabilisation solution (BD Bioscience).

Fig 1

B‐cell subpopulations. (A) Representative dot plots in a healthy individual and a patient with AA. CD3–CD19+CD33–GPI‐anchor+ live B cells were classified based on their expressions of CD24 and CD38: CD24hiCD38hi Bregs, CD24lowCD38low mature naïve B cells, CD24hiCD38low memory B cells, and CD24lowCD38hi plasma cells/plasmablasts. (B) Comparisons of CD24hiCD38hi Breg frequencies among healthy individuals (n = 29), patients with SAA (n = 37) and VSAA (n = 23). [Colour figure can be viewed at wileyonlinelibrary.com]

Fig 3

IL‐10 production from B cells. (A) Representative dot plots of intracellular IL‐10 staining in CD24hiCD38hi Bregs and non‐Breg B cells. The control was B cells cultured without brefeldin A. (B) IL‐10+ B‐cell frequencies in total B cells were reduced in patients with AA (n = 33) compared to healthy individuals (n = 12). (C) IL‐10+ cell frequencies in B‐cell subpopulations in healthy individuals (n = 12) and patients with AA (n = 10). Intracellular IL‐10+ B cells were significantly enriched in CD24hiCD38hi Bregs in both groups. [Colour figure can be viewed at wileyonlinelibrary.com]

Statistics

All statistical analyses were performed using the EZR software package (version 1·38), a graphical user interface for R (version 3·5·2). Data were shown as an n (%) or a median (interquartile range, IQR), unless otherwise specified. Fisher’s exact test and Mann–Whitney U‐test were used to compare categorical variables and numerical variables, respectively. Spearman’s rank correlation test was used to assess correlations among numerical variables.

Results

A total of 125 consecutive treatment‐naïve patients with AA, aged 18–82 years (median, 40 years), were treated with ATG, cyclosporine and eltrombopag from July 2012 to October 2018. Baseline TBNK lymphocyte subsets were evaluated in 104 of the 125 patients within 4 weeks prior to the therapy, and B‐cell subpopulations were analysed in 60 patients whose cryopreserved PBMCs were available. A healthy control dataset of TBNK lymphocyte subsets was obtained from 40 healthy individuals, aged 18–55 years. Cryopreserved PBMCs from 29 healthy individuals, aged 18–72 years, were used as controls of B‐cell subpopulations. The patients’ characteristics are summarised in Table I and Table SI.

Table I

Clinical characteristics of the 104 patients

Characteristics	Value
Number of patients	104
Age, years, median (IQR)	39 (25–56)
Sex, n (%)
Male	52 (50)
Female	52 (50)
Neutrophil, ×109/l, median (IQR)	0·30 (0·085–0·51)
Reticulocyte, ×109/l, median (IQR)	20 (9·4–33)
Platelet, ×109/l, median (IQR)	10 (6·0–15)
Severity, n (%)
Severe	60 (58)
Very severe	44 (42)
GPI‐anchor deficient clone, n (%)
<1%	68 (65)
≥1%	36 (35)
Clone size, %, median (IQR)	7·4 (2·5–28)
IST response at 6 months, n (%)
Responder	91 (87)
Non‐responder	13 (13)

GPI, glycosylphosphatidylinositol; IST, immunosuppressive therapy; clone size is a summary of 36 patients with ≥1% GPI‐anchor deficient clone.

Reduced CD24hiCD38hi Bregs in patients with AA

Peripheral blood CD24hiCD38hi Breg frequencies were markedly reduced in patients with AA compared to healthy individuals, at a median (IQR) of 0·31 (0·14–0·85)% vs. 1·9 (1·3–3·6)% (P < 0·0001; Table II, Fig 1), although neither total CD19+ B‐cell counts nor other B‐cell subpopulation frequencies were significantly different between patients with AA and healthy individuals.

Table II

Age‐matched comparisons of lymphocyte subsets between patients with aplastic anaemia (AA) and healthy individuals.

Variable, median (IQR)	Healthy individuals		Patients with AA		P
	n	Summary measure	n	Summary measure
B cell, /μl	40	163 (106–242)	71	143 (73–225)	0·29
CD24hiCD38hi regulatory B cell, %	29	1·9 (1·3–3·6)	60	0·31 (0·14–0·85)	<0·0001
CD24lowCD38low mature naïve B cell, %	29	71 (61–75)	60	65 (52–76)	0·12
CD24hiCD38low memory B cell, %	29	24 (17–31)	60	27 (13–38)	0·56
CD24lowCD38hi plasma cell/plasmablast, %	29	1·7 (1·1–3·0)	60	2·1 (1·2–6·1)	0·33
CD4+ T cell, /μl	40	798 (625–1021)	71	698 (452–963)	0·054
CD8+ T cell, /μl	40	400 (296–616)	71	411 (282–537)	0·57
Natural killer cell, /μl	40	322 (202–386)	71	79 (50–112)	<0·0001

NK cells were lower in patients with AA (P < 0·0001; Table II) as previously reported. CD4+ T cells and CD8+ T cells were also reduced in AA (P = 0·0094 and P = 0·022, respectively), but the difference was due to a negative correlation of CD8+ T cell counts with age (correlation efficiency [r], −0·42, P < 0·0001): age‐matched comparisons of 71 patients with AA and 40 healthy individuals, aged 18–55 years, showed that CD8+ T cell counts in patients with AA were similar to those in healthy individuals (P = 0·57; Table II); CD4+ T cell counts tended lower in AA (P = 0·054); and total B‐cell counts and NK cell counts of patients with AA remained non‐significant (P = 0·29) and significant (P < 0·0001), respectively, compared to healthy individuals.

Reduced CD24hiCD38hi Bregs correlate with severity of AA

An analysis of lymphocyte subsets and baseline clinical parameters, including age, ANC, ARC and platelet counts revealed that only ANC significantly correlated with CD24hiCD38hi Bregs, both in frequency (r = 0·33, P = 0·0098) and absolute number (r = 0·31, P = 0·016). CD24hiCD38hi Bregs were significantly lower in 23 patients with VSAA compared to 37 with SAA (median [IQR] 0·18 [0·11–0·34]% vs. 0·50 [0·17–1·4]%, P = 0·017; P = 0·044 in absolute numbers; Fig 1B, Table SII, SIV). In contrast, CD24lowCD38low mature naïve B‐cell frequencies negatively correlated with ARC (r = −0·38, P = 0·0024) and with ANC (r = −0·33, P = 0·0087), and were significantly higher in VSAA (P = 0·024); but negative correlations with blood counts and with severity were not observed in absolute numbers (Table SIV). CD24hiCD38hi Breg and mature naïve B‐cell frequencies did not show a correlation (P = 0·58). NK cell and CD8+ T‐cell counts were also reduced in VSAA compared to SAA (P = 0·0084 and P = 0·026, respectively). Age, as described above, significantly correlated with CD8+ T cells (r = −0·42, P < 0·0001), and mature naïve B‐cell frequencies (r = 0·29, P = 0·023), but did not show a correlation with CD24hiCD38hi Bregs or with other lymphocyte subsets studied.

Responses to IST plus eltrombopag

Initial CD24hiCD38hi Breg displayed no difference between IST‐responders and non‐responders (Fig 2A, Table III). Among the eight lymphocyte subpopulations studied before therapy, only CD24lowCD38low mature naïve B cells correlated with IST responses: 47 responders showed lower initial mature naïve B cells compared to 13 non‐responders, both in frequency (P = 0·014, Fig 2B, Table III) and absolute number (P = 0·036, Table SIV). Responders exhibited a significantly higher ARC compared to non‐responders (P = 0·027), consistent with a previous report. Clonal evolution was documented in 13 of 104 patients after IST, which did not correlate with Bregs or the other subsets.

Fig 2

B‐cell subpopulation frequencies and response to IST. (A) Initial CD24hiCD38hi Bregs did not correlate with IST responses. (B) Among eight lymphocyte subsets studied, only mature naïve B cells correlated with haematological responses. (C) CD24hiCD38hi Bregs recovered in 6 months after institution of ATG in both responders (blue dot lines, n = 34) and non‐responders (red solid lines, n = 5). R, responders; NR, non‐responders. [Colour figure can be viewed at wileyonlinelibrary.com]

Table III

Lymphocyte subsets in IST‐responders and ‐non‐responders.

Variable,	IST non‐responder		IST responder		P
	n	Summary measure	n	Summary measure
Age, years, median (IQR)	13	47 (30–56)	91	38 (25–56)	0·25
Male sex, n (%)	13	8 (62%)	91	44 (48%)	0·56
Median (IQR):
B cell (/μl)	13	142 (84–218)	91	136 (73–210)	0·93
CD24hiCD38hi regulatory B cell, %	13	0·37 (0·10–1·9)	47	0·30 (0·15–0·82)	0·88
CD24lowCD38low mature naïve B cell, %	13	73 (58–88)	47	64 (42–71)	0·014
CD24hiCD38low memory B cell, %	13	13 (9–33)	47	29 (19–39)	0·076
CD24lowCD38hi plasma cell/plasmablast, %	13	2·5 (1·3–3·8)	47	1·9 (1·2–6·3)	0·98
CD4+ T cell, /μl	13	535 (280–712)	91	673 (445–917)	0·24
CD8+ T cell. /μl	13	197 (107–359)	91	334 (218–502)	0·15
Natural killer cell, /μl	13	64 (45–79)	91	82 (49–113)	0·12

Recovery of CD24hiCD38hi Bregs after IST

TBNK lymphocyte subsets and B‐cell subpopulations were re‐evaluated at 6 months after institution of ATG in 94 and 39 patients, respectively. Total B‐cell counts were lower after IST (median [IQR] 74 [46-107]/μl from 137 [73-212]/μl, P < 0·0001; Table SIII), but CD24hiCD38hi Bregs increased to levels present in healthy individuals (median [IQR] 2·3 [0·98–4·8]% from 0·31 [0·14–0·85]%, P < 0·0001; Fig 2C), in both responders (n = 34) and non‐responders (n = 5). CD24lowCD38hi plasma cells/plasmablasts were significantly reduced after IST (P < 0·0001). Memory and mature naïve B‐cell frequencies did not change after IST, but their absolute numbers were reduced significantly (P = 0·00084 and P = 0·029, respectively; Table SIV). None of the lymphocyte subsets at 6 months significantly correlated with IST responses, but mature naïve B‐cell frequencies showed a higher trend in non‐responders at 6 months (P = 0·068). Patients who stopped eltrombopag at 3 months and those who continued until 6 months exhibited no difference in their lymphocyte subsets at 6 months (TBNK analysis, n = 17 vs. n = 77; B‐cell subpopulation analysis, n = 8 vs. n = 31; P > 0·34 in all comparisons).

IL‐10 production from CD24hiCD38hi Bregs

B cells that produce IL‐10 in response to CpG and CD40L stimulation were also significantly reduced in 33 patients with AA compared to 12 healthy individuals (median [IQR] 1·7 [1·1–2·4]% vs. 3·0 [2·4–4·4]%, P = 0·0017; Fig 3B), as observed by others, and they showed a positive correlation with CD24hiCD38hi Bregs (r = 0·37, P = 0·012). IL‐10 production from CD24hiCD38hi Bregs was analysable only in 10 of the 33 patients with AA, in which CD24hiCD38hi Bregs were retained but still these were significantly reduced compared to the 12 healthy individuals (median [IQR] 0·76 [0·31–2·2]% vs. 2·8 [2·0–4·1]%, P = 0·0056). Intracellular IL‐10+ cells were significantly enriched in CD24hiCD38hi Bregs both in patients with AA and healthy individuals, and their percentages in CD24hiCD38hi Bregs were comparable between patients with AA and healthy individuals (Fig 3C). These findings indicate that residual CD24hiCD38hi Bregs of patients with AA remain functional.

Bone marrow CD24hiCD38hi Bregs

CD24hiCD38hi Bregs were studied in cryopreserved bone marrow samples of three healthy individuals and four patients with AA at diagnosis and 6 months after institution of IST (Figure SII). Percentages of CD24hiCD38hi Bregs in the three healthy bone marrow samples were 20%, 19% and 13%, respectively, which were much higher than those we have observed in peripheral blood. In contrast, CD24hiCD38hi Bregs in bone marrow of untreated patients with AA were markedly reduced and correlated with those in peripheral blood, and they recovered to even higher levels at 6 months than did cells in peripheral blood.

Discussion

Functions, subsets and repertoires of T cells have been characterised in patients with immune AA, but little has been reported on B cells and their subpopulations. We found that CD24hiCD38hi Bregs were markedly reduced in untreated patients with AA, without significant changes in total B‐cell counts and other B‐cell subpopulations. CD24hiCD38hi Bregs were especially reduced in VSAA and recovered after IST. Residual CD24hiCD38hi Bregs were functional, capable of producing IL‐10. A numerical deficit of CD24hiCD38hi Bregs, that suppress CTLs and Th1 responses and induce Tregs, is consistent with immune pathophysiology of AA, in which autoreactive CD8+ CTLs and Th1 responses, in the absence of Treg suppression, are responsible for destruction of HSPCs.

Initial CD24hiCD38hi Breg deficit and recovery of this cell population after IST did not correlate with response to the therapy. Tregs are also reduced in AA and have not correlated with IST response, although a subpopulation of Tregs may be predictive. Other factors, such as residual haematopoiesis and elimination of autoreactive T cells, may be more important. We unexpectedly observed higher CD24lowCD38low mature naïve B cells to associate with a lower probability of response to IST. A negative correlation of mature naïve B‐cell frequencies with baseline ARC, the strongest predictor of a response to ATG‐based IST, may explain this correlation, but it is unclear why mature naïve B cells had a negative correlation with ARC.

Reduced CD24hiCD38hi Bregs in AA, especially in VSAA, may be in part a consequence of marrow failure, as this population represents an immature subset among circulating B cells. But CD24hiCD38hi Bregs recovered in response to IST even in non‐responders, suggesting that CD24hiCD38hi Bregs are not just reflecting haematopoiesis.

Reduction of B cells and changes in their subpopulations after IST should be attributable to therapy. Cyclosporine suppresses human B‐cell proliferation and reduces Bregs. However, CD24hiCD38hi Bregs in AA remarkably increased after IST on cyclosporine, implying that immune reconstruction by ATG associated with the CD24hiCD38hi Bregs recovery, while eltrombopag appeared not to affect Bregs or the other subpopulations because earlier discontinuation of eltrombopag did make a difference in the lymphocyte subsets studied.

IL‐10‐producing B cells in bone marrow have been reported to be reduced in patients with AA compared to healthy individuals; we confirmed this finding in peripheral blood, because marrow samples are diluted by peripheral blood, perhaps more so with hypocellular samples. Further the reduction of IL‐10+ B cells associated with the quantitative deficit of CD24hiCD38hi Bregs, but residual CD24hiCD38hi Bregs in AA were functional, as they produced IL‐10 in response to CpG and CD40L.

In addition to B‐cell subpopulations, we assessed clinical correlations of conventional lymphocyte subsets, but only NK cells were found to be reduced in AA as previously reported. A distinct lineage origin of NK cells and their short lifespan may associate with a selective defect of NK cells in AA. None of lymphocyte subsets correlated with responses to IST plus eltrombopag.

CD16 and CD24 are GPI‐anchored proteins and their cell surface expressions are reduced in GPI‐anchor deficient cells. Therefore, we excluded GPI‐anchor deficient B cells to avoid underestimation of CD24hiCD38hi Bregs, but NK cell counts were potentially underestimated in the present study because GPI‐anchor expressions were not assessed in the TBNK subset analysis. However, the presence of GPI‐anchor deficient clones cannot explain the reduced NK cells in AA: CD16+CD56– NK cells that become undetectable by GPI‐anchor defects are a minor subpopulation of NK cells, and NK cell counts of patients with AA who did not have GPI‐anchor deficient clones remained markedly reduced compared to healthy individuals (P < 0·0001).

While paediatric patients were not enrolled in this retrospective study due to differences in normal lymphocyte counts between children and adults, we did study B‐cell subpopulations in three patients aged 4–14 years: CD24hiCD38hi Bregs were also low in these three children (0·09%, 0·20% and 0·32%), suggesting CD24hiCD38hi Bregs to be reduced in both adult and paediatric AA.

In summary, CD24hiCD38hi Bregs that potentially suppress autoreactive CTLs and Th1 responses were markedly reduced in AA, especially in VSAA, and recovered in response to IST plus eltrombopag. A quantitative deficit in functional CD24hiCD38hi Breg may contribute to pathophysiology in immune AA. We unexpectedly observed higher CD24lowCD38low mature naïve B‐cell numbers to be associated with a lower probability of responses to IST. B‐cell phenotype analysis may be useful for monitoring of AA and predicting outcomes of therapy.

Conflict of interest

The authors have no conflict of interest to declare.

Author contributions

Yoshitaka Zaimoku, Sachiko Kajigaya and Neal S. Young designed the study and wrote the first draft of the manuscript. Bhavisha A. Patel, Emma M. Groarke and Neal S. Young participated in the patient care. Yoshitaka Zaimoku and Xingmin Feng performed the flow cytometry. Lemlem Alemu and Diego Quinones Raffo managed the patient samples. All authors critically reviewed the draft and approved the final version for publication.

Fig S1. Representative gating strategies. (A) TBNK lymphocyte subset analysis. (B) B‐cell subpopulation analysis. (C) Intracellular IL‐10 expressions in CD24hiCD38hi Bregs and non‐Bregs.

Fig S2. Bone marrow CD24hiCD38hi Bregs. (A) Representative dot plots of a healthy individual and a patient with AA at diagnosis and at 6 months after institution of IST plus eltrombopag. (B) CD24hiCD38hi Breg frequencies in peripheral and bone marrow B cells in patients with AA before IST (black; n = 4) and at 6 months after therapy (red; n = 4); and those in healthy bone marrow (green; n = 3).

Table SI. Clinical characteristics of 60 patients in the B‐cell subpopulation analysis.

Table SII. Lymphocyte subsets in SAA and VSAA.

Table SIII. Lymphocyte subsets before and after IST.

Table SIV. B‐cell subpopulation comparisons in percentages and absolute numbers.

Click here for additional data file.