Human B-cell subset identification and changes in inflammatory diseases

Summary

Our understanding of the B-cell subsets found in human blood and their functional significance has advanced greatly in the past decade. This has been aided by the evolution of high dimensional phenotypic tools such as mass cytometry and single-cell RNA sequencing which have revealed heterogeneity in populations that were previously considered homogenous. Despite this, there is still uncertainty and variation between studies as to how B-cell subsets are identified and named. This review will focus on the most commonly encountered subsets of B cells in human blood and will describe gating strategies for their identification by flow and mass cytometry. Important changes to population frequencies and function in common inflammatory and autoimmune diseases will also be described.

This review describes gating strategies to identify B-cell subsets in blood and the changes that occur in these subsets with inflammatory and autoimmune diseases.

Graphical Abstract

Graphical Abstract.

Introduction

A unified and consistent approach to the identification and description of B-cell subsets is essential to ensure reproducibility of scientific studies and that analogous conclusions can be drawn from them. Although many recent advances have been made in B-cell biology, significant areas of uncertainty remain regarding the nature of certain B-cell subsets and their role in homeostasis and inflammatory disease. This is likely to have arisen due to a number of reasons. Firstly, whilst murine models are a valuable tool to understand B-cell biology, there are significant differences between murine and human B-cell subsets which mean that it can be difficult to draw analogous conclusions from them [1–4]. Secondly, the evolution of high dimensional phenotypic techniques has identified many novel subsets but further work is needed to expand on their functional and pathogenic roles [5–8]. Thirdly, B cells in blood often represent a developmental continuum and overlap phenotypically without sharply demarcated gates to differentiate them [5, 9–11]. And finally, there remain significant gaps in our understanding of human B-cell maturation and therefore the developmental trajectories that link human B-cell subsets to one another [12].

B cells are known to play important roles in a variety of inflammatory diseases via the production of antibodies and cytokines as well as antigen presentation [13–20]. Their pathogenic role is supported by the response of a variety of inflammatory conditions to B-cell directed therapies such as rituximab and B-cell activating factor (BAFF) antagonists [21–25]. Immune mediated inflammatory diseases such as systemic lupus erythematosus (SLE), Sjogren’s syndrome (SS), rheumatoid arthritis (RA), and COVID-19 are associated with significant changes in B-cell subset frequency which is likely driven by inflammatory cytokines and pro-survival factors as well as changes in B-cell differentiation and homing [26–28].

In this review we will describe the main subsets of B cells in peripheral blood and how these can be identified using flow and mass cytometry, as well as important changes in B-cell subsets in inflammatory immune mediated conditions. B1 cells will not be discussed in this review due to uncertainties regarding their status and because they represent a rare subset in adult human blood [29–31].

Identification of B cells in peripheral blood

The surface expression of both CD19 and CD20 are induced early in B-cell development in the bone marrow [11, 32]. They are both expressed on B-cell subsets in blood but CD20 expression is lost during terminal B-cell differentiation into antibody secreting cells (ASCs) [33, 34]. CD19 therefore allows identification of both B cells and ASCs. A gating strategy using CD19 for the identification of both B cells and ASCs in blood by flow cytometry is illustrated in Fig. 1A. An antibody panel of 9 ‘key’ markers can be used to identify the main B-cell subsets in blood and these are summarized in Table 1. Additional markers are also given in this table which can aid gating and the identification of further subsets.

Figure 1:

identification of B cells in peripheral blood. (A) Flow cytometry plots showing a gating strategy used to identify CD19⁺ B cells and CD19⁺CD27^hiCD38^hi ASCs. Data are from PBMCs from a representative healthy control donor (HCD). (B) A gating strategy of mass cytometry data to identify live CD19⁺ B cells. Antibodies to CD3 and CD14 were conjugated to the same metal and therefore created a ‘dump’ channel which allows exclusion of T cells and cells of myeloid origin. Data are from a female representative healthy control donor. (C) ViSNE plots generated using the following markers; CD10, CD24, CD38, CD27, CD45RB IgD, IgM, IgA, and IgG. The plots are from a female HCD and depict expression of lineage markers used to identify B-cell subsets. (D) SPADE on viSNE plots demonstrating expression of B-cell lineage markers used to classify CD19⁺ B cells into transitional (TS), naïve, activated naive (aNAV), marginal zone B cell (MZB), and their precursors (MZP) and IgG⁺ and IgA⁺ memory subsets. Data adapted from Tull TJ et al [9]

Table 1:

Lineage markers to define B-cell subsets with expected frequencies in healthy peripheral blood and studies reporting altered frequency or function in autoimmune rheumatic diseases and COVID-19

Subset	Key markers	Additional markers and subsets	Frequency (% CD19+)	Disease associations
T1	CD27−IgD+IgM+CD10+ CD21loCD24+++CD38+++ [35–37]	β1hiβ7loCCR7loCD62Llo [9, 37]	1−4 [37, 38]	SLE [38, 39] SSc [40] COVID [41, 42]
T2	CD27−IgD+IgM+CD10+ CD21hiCD24++CD38++ [35–37]	β1hi/intCD62Lhi/int IgMhiT2: IgMhiCCR7hiβ7hi [9] IgMloT2: IgMloCCR7loβ7lo [9]	2−10 [28, 38]	SLE [9, 28, 38, 39] SSc [40] COVID [41, 42]
T3	CD27−IgD+IgM+CD10−CD24+CD38+RHhi/MTGhi [43–45]	IgMhiT3: IgMhiCD1chiβ7hiCD45RBint/hi [9, 46] IgMloT3: IgMloCD1cloβ7loCD45RBlo/int [9]	9−13 [9, 45]	SLE [9, 44]
Naive	CD27−IgD+IgM+CD10−CD24+CD38+RHlo/MTGlo [43, 44]		40−60 [9, 45]	SLE [47, 48] SSc [49]
Naïve MZP	CD27−IgD+IgMhiCD10−CD24+CD38+RHlo/MTGloCD45RBhi [46, 50]	CD1chi [9, 46]	0.5−3 [9, 46]	SLE [9]
Activated naive	CD27−IgD+CD10−CD21lo CD24loCD38lo [26, 27]	FCRL5+, CXCR5+, T-bet+ [26]	0.1−3 [9]	SLE [26, 27, 51] COVID [42]
MZB	CD27+IgD+IgM+ [10]	CD1chi [52] MZB1: β7hiCCR7hi [6] MZB2: β7loCCR7lo [6]	2−30 [53, 54]	SLE [54, 55] SS [56, 57] SSc [58] RA [59] COVID [42, 60]
IgM only	CD27+IgD−IgM+ [61, 62]	CD45RB+ [5]	10 [63]
CSM	CD27+IgD−IgM− [5, 61]	CD45RB+, IgG, IgA or IgE+ [64]	5−40 [26, 53]	SLE [54, 65] SS [66] RA [67, 68] ANCA vasc [69]
DN	CD27−IgD− [70, 71]	DN1: CD21int/hi, CXCR5+ [26, 41] DN2: CD21lo, FCRL5+, CXCR5−, T-bet+, CD11chi [26, 41] DN3: CD21−CD11c−CXCR5− [72] DN4: CD21+CD11c+ [73]	1−10 [26, 71]	SLE [26, 27, 74] RA [75–77] SSc [49] COVID [41, 42, 78, 79]

ANCA vasc = ANCA-associated vasculitis; B1 = B1 integrin; B7 = B7 integrin; CSM = class switched memory; DN = double negative; MZP = marginal zone precursor; SLE = systemic lupus erythematosus; SS = Sjorgen’s syndrome; SSc = Systemic sclerosis;.

Mass cytometry utilizes the ionic signal from antibodies tagged to rare earth metals and therefore is not limited by the spectral overlap from fluorochrome-conjugated antibodies that are used for flow cytometry [80]. Mass cytometry therefore allows single cells to be labelled with ~40 metal tagged antibodies that can be directed to cell surface or intracellular targets. A gating strategy to identify live single B cells by mass cytometry is displayed in Fig. 1B. Further identification of B-cell subsets can then be performed by biaxial gating or dimensional reduction and clustering methods. Biaxial gating of mass cytometry data can be challenging as the expression of B-cell subset markers often lack distinct pinch point gates and metal minus one control samples that guide the accurate placement of gates are impractical due to cost. Dimensionality reduction and clustering techniques that identify subsets based on the expression of multiple markers can therefore be advantageous over biaxial gating. ViSNE, a visualization tool for high-dimensional single-cell data based on the t-Distributed Stochastic Neighbour Embedding algorithm and uniform manifold approximation and projection (UMAP) are common dimensionality reduction tools used to visualize mass cytometry data [81, 82]. These algorithms can be run using all panel markers or select B-cell subset markers and allow the two-dimensional projection of B cells arranged in multidimensional space. Representative viSNE plots of CD19⁺ B cells from a healthy control donor (HCD) generated by running the algorithm on lineage markers are displayed in Fig. 1C. In these viSNE plots, memory B cells appear as distinct islands but significant continuity exists between CD10⁺CD24^hiCD38^hi transitional and CD10⁻CD24^intCD38^int naïve B-cell subsets and manually gating such plots can therefore be challenging (Fig. 1C). For this reason, spanning tree progression of density normalized events (SPADE) run on ViSNE coordinates can be a useful tool to separate subpopulations existing in phenotypic continuity (Fig. 1D) [9, 10]. With SPADE, B-cell populations that are phenotypically similar to one another are displayed as nodes which are interlinked by branches to form trees. Nodes representing B-cell subsets can then be manually grouped together to form bubbles representing that population (Fig. 1D). Whilst this approach is in part directed it has the advantage over biaxial gating in that multiple markers can be used simultaneously to define B-cell subsets.

Transitional B cells

Transitional B cells are defined as recent emigrant B cells from the bone marrow and represent the most immature B-cell subset in peripheral blood [12]. Studies of B-cell repopulation following bone marrow stem cell transplantation and B-cell depletion with rituximab have identified sequential maturation of transitional B cells through T1, T2, and T3 phases [35, 36, 45, 83, 84]. This maturational sequence is associated with the acquisition and loss of various markers. Notably, CD24, CD38, and CD10 are downregulated and these are commonly utilized markers to distinguish transitional B-cell subsets. T1 and T2 B cells share expression of CD10 with immature B cells in the bone marrow, but this is not expressed by T3 cells [37, 45]. T1 and T2 cells are often distinguished from one another based on high and intermediate expression of CD24 and CD38 (Fig. 2A). This approach is subjective as the definition of high and intermediate expression can differ between individuals and experiments given technical factors such as fluorescent signal intensity. T1 cells are known to have low expression of CD21 and the use of this marker in conjunction with CD24 and CD38 can be useful to guide the T1 gate (Fig. 2A) [35].

Figure 2:

transitional B-cell subsets and phenotype. (A) Flow cytometry plots showing a gating strategy used for identification of T1 cells (CD27⁻IgD⁺CD10⁺CD24⁺⁺⁺CD38⁺⁺⁺) and T2 cells (CD27⁻IgD⁺CD10⁺CD24⁺⁺CD38⁺⁺). Data are from PBMCs from a representative HCD. (B) Schema highlighting the transitional B-cell population from the SPADE on viSNE plot in Fig. 1D. (C) SPADE on viSNE plots of TS B cells exported from TS SPADE bubble from Fig 1D demonstrating marker expression and classification of transitional B-cell populations. T1 cells were defined as CD10⁺CD24⁺⁺⁺CD38⁺⁺⁺CD21^lo, T2 as CD10⁺CD24⁺⁺CD38⁺⁺CD21^hi, and T3 as CD10⁻CD24⁺CD38⁺CD21^hi. Data are from a representative female HCD. (D) SPADE on viSNE plot displaying IgM expression. IgM^hi T2 and T3 populations are shown by dotted red and light blue lines, respectively. IgM^lo T2 and T2 populations are shown by dotted green and purple lines, respectively. (E) Flow cytometry plot showing gating strategy of T2 B cells into T2 IgM^hi and T2 IgM^lo subpopulations and β7 integrin median fluorescence intensity (MFI) in these populations. Data are from six HCD

The use of additional markers can also help to accurately identify T1, T2, and T3 subsets and are represented in Fig. 2B and C as SPADE trees. T1 cells have relatively high expression of β1 integrin, which is likely to be important for interaction with bone marrow stromal cells [85]. T1 cells also lack expression of CD62L and CCR7 (Fig. 2C), which is likely to result in a lack of homing capability as seen in studies of both murine and human transitional B cells [9, 86, 87]. IgM was traditionally thought to be downregulated through T1–T3 development but T2 and T3 populations are both heterogeneous in their expression of this marker (Fig. 2D). IgM^hi T2 cells have high expression of β7 integrin and are enriched in gut associated lymphoid tissue (Fig. 2E) [9]. Furthermore, IgM^hi T2 cells align with cells undergoing marginal zone B (MZB) cell development in terms of surface and transcriptomic features [9]. This suggests that a gut homing T2 MZP population exists in humans as has been demonstrated in mice [88–90].

T3 cells are a late transitional B-cell subset and are thought to represent an intermediate stage of development between T2 and naïve B cells. In keeping with this, T3 cells appear in the blood after T2 cells during B-cell repopulation post B-cell ablation therapy and differentiate into naïve B cells in vitro [44, 45]. Transitional B cells can be distinguished from naïve B cells as they lack expression of the ABCB1 cotransporter, which as a result prevents the extrusion of dyes such as rhodamine (RH) or mitotracker (MTG) [43]. T3 cells can therefore be gated as CD27⁻IgD⁺CD10⁻RH^hi/MTG^hi (Fig. 3A and B). The distinction between RH low and high populations can be guided by the signal obtained from CD27⁺ memory B cells which also lack to ABCB1 cotransporter and therefore are RH^hi (Fig. 3C). The expression of CD45RB in T3 cell populations is heterogeneous (Fig. 3D) and CD45RB^hi cells also have higher expression of IgM and CD1c and align with cells undergoing MZB cell differentiation (Fig. 3E) [9, 46]. The use of dye extrusion is essential for the distinction of T3 cells from naïve B cells as they otherwise share an overlapping surface phenotype (Fig. 3F).

Figure 3:

T3 and naïve B-cell subsets. (A) Flow cytometry plots showing a gating strategy used to identify CD27⁻IgD⁺CD10⁻ B cells. (B) Flow cytometry plots demonstrating gating strategies to identify naïve (CD27⁻IgD⁺CD10⁻RH123^lo) and T3 (CD27⁻IgD⁺CD10⁻RH123^hi) populations. (C) Flow cytometry plot demonstrating the placement of the RH123^hi gate from the signal from CD27⁺ memory B cells. Data are from a representative HCD. (D) Flow cytometry plot depicting a gating strategy used to identify variants of T3 cells based on CD45RB. (E) Comparative histograms taken from flow cytometry data displaying IgM and CD1c MFI in T3 CD45RB^lo T3 CD45RB^hi populations. (F) Comparative histograms taken from flow cytometry data displaying MFI of CD24, CD21, CD38, CD5, IgM, and IgD between T3 and naïve B cells. (G) Comparative histograms taken from flow cytometry data displaying MFI of IgM and CD1c between MZP and naive B cells. (H) Gating strategy of flow cytometry data to identify T3 and aNAV cells based on their expression of CD21, CD24, and CD38

Multiple abnormalities in transitional B-cell subsets have been reported in inflammatory diseases. Transitional B cells harbour a high proportion of autoreactive cells which are removed through sequential maturation into naïve B cells in healthy individuals and this checkpoint is defective in SLE [91]. The mechanisms that underlie this checkpoint failure are not known but are may be secondary to aberrant transitional B-cell maturation or migration into lymphoid tissue. Transitional B cells from patients with SLE have been reported to have lower expression of the gut homing receptor β7 integrin suggestive of defective gut homing capacity and β7^hi T2 cells are depleted in lupus nephritis [9, 39]. Transitional B cells have been shown to be increased in frequency in SLE and this has been correlated with serum BAFF levels [28, 38, 92]. The important role of BAFF as a survival factor is also supported by the increased frequency of transitional B cells in patients with multiple sclerosis that are treated with fingolimod and interferon-β which are both known to increase serum BAFF levels [93]. Transitional B cells in SLE have also been shown to be hyperresponsive to B-cell receptor (BCR) signalling which is likely secondary to type I interferon [94, 95]. Increased frequencies of transitional B cells have also been reported in a number of other autoimmune diseases such as juvenile dermatomyositis, primary SS (pSS), and systemic sclerosis (SSc) [40, 96, 97]. Altered T1:T2 cell ratios have also been identified in patients with renal transplant rejection and are postulated to be an important predictor of graft loss [98, 99]. Finally, a reduction in transitional B cells has been seen in COVID-19 infection which correlates with infection severity [41, 42].

Transitional B cells are enriched in B regulatory cells (Bregs) that are best described as B cells that can suppress T-cell responses mainly via the production of IL-10 [100, 101]. Breg dysfunction has been identified in a number of autoimmune diseases. CD24^hiCD38^hi B cells in SLE have less capacity to inhibit CD4 T-cell cytokine production [100]. Induction of Bregs by plasmacytoid dendritic cells has also been reported to be defective in SLE which is thought to be mediated by type I interferon [102]. Breg dysfunction in SLE may also be mediated by defective B cell homing or maturation in GALT, which is known to be important for Breg induction [9, 39, 103]. Reduced numbers of CD24^hiCD38^hi B cells have been identified in patients with RA and they failed to repress Th17 responses but maintained their ability to suppress Th1 responses [104]. Furthermore, treatment of RA with tumour necrosis factor alpha (TNF-α) antagonists has been demonstrated to increase the number of CD27⁺ Bregs [105]. Primary SS has been associated with increased frequency of CD24^hiCD38^hi B cells but these cells fail to suppress T cell TNF-α or IFN-γ secretion [97]. Furthermore, adoptive transfer of CD24^hiCD38^hi B cells from patients with pSS into mice with experimental SS demonstrated failure of suppression of T follicular cell expansion [106]. Reduced Breg frequency or defective Breg function has also been reported in MS, SSc, pemphigus, and myasthenia gravis [107–111]. Breg dysfunction has therefore been identified in a number of autoimmune diseases but further work is required to identify the pathogenic significance and the mechanisms that underlie this.

Naïve B cells

Naïve B cells represent B cells that have not encountered antigen and have not undergone germinal centre maturation. They therefore lack CD27 expression and have a low level of somatic mutations within their immunoglobulin variable (IgV) genes [63]. Naïve B cells represent 40–60% of peripheral blood B cells and as discussed above can only be distinguished from T3 cells by their ability to extrude RH or MTG dyes due to their expression of the ABCB1 cotransporter [43]. Naïve B cells can therefore be gated as CD27⁻IgD⁺CD10⁻RH^loCD45RB^lo (Fig. 3A). Distinction by dye extrusion is not possible using mass cytometry although a metal tagged antibody to ABCB1 would theoretically allow the distinction between T3 and naïve B cell, to our knowledge this is not currently commercially available.

Naïve B cells are a heterogeneous population and a number of naïve subsets have been identified including marginal zone precursor cells (MZP), activated naïve (aNAV), and an IgM^lo anergic subset [26, 46, 47]. Naïve MZP cells are characterized by high expression of CD45RB and also highly express IgM and CD1c have been shown to undergo differentiation into MZB like B cells following NOTCH ligation (Fig. 3G) [46]. A T3 population with a similar surface phenotype has also been identified and is thought to be in developmental continuity with naïve MZP cells [9, 30]. Naïve B cells with high-surface expression of CD45RB has been described as an early memory population and found to have more somatic mutations than CD45RB⁻ naïve B cells [5]. The same study commented on an anergic profile of naïve B cells as evidenced by lower expression of intracellular transport proteins suggesting that this subset would be less responsive to stimulation. Other functional studies have, however, demonstrated robust proliferative responses of naïve B cells to IgM crosslinking as well as CD40 and TLR9 agonists [37, 44, 45, 112].

Naïve B cells with low IgM expression have been reported as having anergic properties as evidence by lower intracellular calcium influxes and expression of activation markers following IgM crosslinking [47]. This subset was also found to express higher levels of the inhibitory receptor CD22. The identity of anergic B cells as IgM^loIgD⁺ is supported by a number of murine studies using the hen egg lysozyme model in which the majority of B cells recognize this epitope. When HEL is expression was induced endogenously the majority of B cells adopted this phenotype, suggesting anergy can be induced, and is a key tolerance mechanism [113–115].

ANAV are a subset of CD27⁻IgD⁺ naïve cells with low expression of CD21, CD23, CD24, and CD38 but high expression of FCRL5, CD11c, and T-bet [26, 27] (Fig. 3H). This subset also does not express the ABCB1 cotransporter and is therefore RH or MTG high following dye extrusion assays [116]. ANAV were first identified in SLE as being clonally related to ASC populations and are hyperresponsive to TLR7 stimulation [26, 27]. The aNAV subset is therefore thought to undergo extrafollicular maturation into ASC in SLE and are primed by interferon and TLR7 agonists.

Inflammatory disorders in which extrafollicular B-cell maturation is thought to occur include severe SLE and COVID-19 and are associated with an increase in aNAV frequency alongside increased frequencies of ASCs [26, 41, 51]. Conversely, MZP and naïve B cells are reduced in severe SLE [9]. Furthermore, inhibition of BAFF with belimumab results in a lower frequency of naïve B cells, suggesting the importance of BAFF for their survival [117, 118]. Of note, numerous studies of B-cell subset abnormalities in inflammatory disease have gated naïve B cells as solely CD27⁻ or CD27⁻IgD⁺ and therefore included double negative (DN) and transitional B cells, respectively in these gates [119–122].

MZB cells

The splenic marginal zone refers to a microanatomical area on the periphery of B-cell follicles in the white pulp [123]. MZB cells are termed as such as they share similar surface phenotype to the subset of cells found in the splenic marginal zone that express low levels of IgD and high levels of CD1c [52, 124]. In mice, MZB cells are confined to the spleen and do not have IgV gene mutations and therefore arise independently of germinal centre responses. In humans, MZB cells circulate in blood and lymphoid tissue and have IgV gene mutations, although to a lower degree when compared to class-switched memory (CSM) B cells [10, 52, 63, 125]. The finding that MZB have IgV mutations in patients with CD40L deficiency who lack germinal centres suggest they may mature independent of germinal centre responses as in mice [64, 125]. This is supported by studies that demonstrate little overlap between MZB and CSM B-cell clones, suggesting that MZB represent a distinct entity rather than just a precursor to CSM [10, 61]. Germinal centre independent maturation of MZB is, however, challenged by the finding that 20% of blood MZB carry BCL6 gene mutations and that human MZB have been reported to share clones with GC delivered IgG memory B cells [126–128]. Similarly, whilst a specific MZB maturational pathway exists in mice, the developmental origins of human MZB remain incompletely understood. As described above, human transitional and naïve MZP have an IgM^hi gut homing phenotype suggestive of a role of GALT in their induction [9]. Furthermore, MZB are observed to occupy a peri-germinal centre niche and to diversify their BCR within GALT [10]. Functionally, MZB undergo T cell independent responses and are important for immunity against encapsulated bacteria [52, 129].

Uncertainty regarding the developmental origin and functional status of MZB cells is reflected by the plethora of names assigned to this subset in the literature. These include unswitched memory [56, 130, 131], pre-switched memory [58, 119, 132], non-switched [133, 134], natural effector [64, 135], and IgM memory [136, 137] B cells. MZB have a CD27⁺IgD⁺IgM⁺ surface phenotype and have high expression of CD1c, CD21, and CD45RB (Fig. 4A) [63]. MZB have also been reported to have low levels of CD23 and high expression of the inhibitory receptor CD32b [138, 139]. Recently, two distinct populations of MZB have been described that differ in cell surface properties, transcriptional factors, and distribution in lymphoid tissues [6]. These have been termed MZB1 and MZB2, MZB1 having higher expression of β7 integrin, CCR7, CD24, and CD27 than MZB2 (Fig. 4B and C). MZB1 and MZB2 were not found to be clonally related and NOTCH induced genes were only upregulated in MZB2 suggesting independent maturational pathways.

Figure 4:

human MZB cells. (A) Flow cytometry plot showing a gating strategy used to identify MZB cells as CD27⁺IgD⁺ and histograms showing MFI of markers that demonstrate that MZB cells are IgM⁺CD1c⁺CD45RB⁺CD21⁺. (B) Flow cytometry plots demonstrating the identification of two marginal zone subsets, which can be distinguished based on expression of CCR7 and β7 integrin. MZB1 are CD27⁺IgD⁺IgM⁺CD1c⁺CCR7⁺β7^hi and MZB2 are CD27⁺IgD⁺IgM⁺CD1c+CCR7^-β7^lo. (C) SPADE on viSNE plot showing events exported from the MZB SPADE bubble in Fig. 1D. Higher expression of CCR7 and β7 integrin is seen in MZB1 and low expression in MZB2. (D) Comparative flow cytometry plots showing the CD19⁺ compartment in a HCD and a patient with SLE. Plots show reduction in CD27⁺IgD⁺ MZB cells in SLE compared to HCD [9]. (E) Quantification of flow cytometry data demonstrating a significant reduction in CD27⁺IgD⁺ cells in SLE compared to HCD [9]

Significant changes in MZB frequency have been reported in association with inflammatory diseases. As seen with MZP subsets, severe depletion of MZB is observed in severe SLE and this predominantly affects the MZB1 subset (Fig. 4D) [6, 9]. As well as SLE, reduced frequencies of MZB have also been reported in SS, SSc, and COVID-19 [9, 55, 56, 58–60, 135]. The factors that result in this depletion in blood are not understood and could represent reduced cell survival or genesis, altered tissue distribution or increased differentiation into ASCs. The depletion of MZB alongside their putative precursors in severe SLE suggests defective MZB differentiation. The reciprocal reduction of MZB cells and the increase of ASCs in SLE could also support their differentiation into short-lived ASCs that are characteristic of SLE disease flares [27]. Also, as MZB depletion in COVID and SLE has been shown to be reversible following resolution of the infection and with treatment respectively, inflammatory factors can result in significant plasticity in this population [55, 135]. The functional significance of depletion in MZB is unknown but could underly the increased susceptibility of patients to infection with encapsulated bacteria or defective B regulatory responses [140, 141].

CD27⁺ memory and IgM only B cells

Memory B cells are formed from mature naïve B cells that have encountered antigen and undergone T dependent immune responses and germinal centre maturation [2, 142]. They can be distinguished from naïve B cells by a much higher degree of IgV gene mutations, the gain of expression of CD27, and the loss of IgD expression [1, 63, 128, 143, 144]. Rounds of proliferation and diversification within germinal centres create clones of memory B cells that produce high-affinity immunoglobulin and differentiate into long-lived ASCs [145–147]. Functionally, memory B cells proliferate rapidly in response to stimulation with CD40L and interleukin (IL) 2 and IL21 [148, 149].

Distinct subsets of CD27⁺IgD⁻ memory B cells can be identified by the expression of IgM, IgA, or IgG (Fig. 5A). CD27⁺ memory cells also share high expression of CD24 and CD45RB and relatively low expression of CD38 (Fig. 5B). CD27 and CD38 are also important in distinguishing memory B cells from ASCs (Fig. 5C). IgM only B cells are CD27⁺IgD⁻IgM⁺ and have been identified to have a higher frequency of IgV gene mutations than MZB cells and to share clones with CSM B cells, suggesting that these cells are GC derived [10, 61]. The functional differences of human IgM only and CSM B cells in terms of recall responses and longevity are not well understood in humans and other than IgM expression they share a similar surface phenotype (Fig. 5D) [150]. Of interest is a recent study of BK polyoma virus responses demonstrated different antiviral specificities of IgM and IgG memory B cells which suggests a distinct function of IgM memory B cells [151]. CSM B cells consist of IgA⁺, IgG⁺, and IgE⁺ subsets although the latter is a rare subset in peripheral blood. CSM have a high frequency of IgV gene mutations and differentiate into ASCs that produce high-affinity immunoglobulin that are essential for humoral immunity and the basis of effective vaccine responses [127]. Heterogeneity within memory B-cell populations has been identified through differential expression of CD73 and CD95. Memory B cells can be CD73 positive or negative, its expression is inversely correlated with IgM and is mainly seen in CSM B cells [5]. CD95 or Fas is a pro-apoptotic receptor which is mainly expressed by CSM, ASCs, and CD27⁻ memory B cells which is elevated in SLE and RA and thought to indicate an activated memory B-cell phenotype [5, 152–154].

Figure 5:

CD27⁺ memory B cells and IgM only B cells. (A) Flow cytometry plots showing a gating strategy used to identify memory (CD27⁺IgD⁻IgM⁻ and IgG⁺/IgA⁺) and IgM only (CD27⁺IgD⁻IgM⁺) B cells. Data are from a representative healthy control donor. (B) Comparative histogram plots from flow cytometry data showing differences in expression of CD45RB, CD27, and CD38 in CD27⁺IgD⁻ cells compared to naïve (CD27⁻IgD⁺) cells. (C) Histogram plots from flow cytometry data showing lower expression of CD38 and CD27 and higher CD24 in CSM compared to ASCs. (D) Heatmap from mass cytometry data from Fig. 1D showing median expression of lineage markers on naïve, CSM and IgM only B cells

Memory B cells play a critical role in autoimmune, inflammatory and allergic diseases as they can differentiate into ASCs that produce pathogenic autoantibodies. Autoreactive memory B-cell clones arise due to checkpoint failure, where autoreactive naïve B cells are allowed to undergo germinal centre or extrafollicular responses. This checkpoint has been shown to be defective in a variety of diseases including SLE and SS, although the mechanisms that result in this are still unclear [57, 155]. B cells in SLE are known to express greater levels of CD40 and blockade of germinal centre responses with CD40L (CD154) antagonists has therapeutic promise [156–158]. Variable frequencies of CSM B cells have been observed in SLE which is likely to represent heterogeneity in disease activity, duration, and treatment [54, 65, 66]. Interestingly, unlike naïve and MZB subsets, CSM frequency is not affected by BAFF inhibition with belimumab [118]. Variable frequencies of CSM B cells have also been observed in RA with a lower proportion of CXCR3⁺CD27⁺ cells which correlated inversely with disease activity [67, 68]. Increased frequency of switched memory B cells have also been observed in other autoimmune conditions including ANCA vasculitis [69]. B-cell depleting agents such as rituximab have long-lasting effects of memory B cells, which may underlie their efficacy in treating autoimmune disease [159].

DN memory B cells

CD27⁻ memory B cells were first identified as IgG⁺ cells in blood with a lower level of somatic IgV gene mutations than CSM [70]. DN B cells expressing IgM, IgA, and IgA were then identified and found to be expanded in SLE [71]. They are termed DN B cells as they lack both CD27 and IgD expression. Their accumulation in chronic infections, autoimmune diseases and with advanced age has also led to them being termed exhausted memory B cells and age associated B cells (ABC) [160–162]. ABC share features of DN cells such as upregulation of T-bet and low expression of CD21 but have been described using a variety of markers and likely represent a heterogeneous population which are not entirely synonymous with DN cells [34, 162].

Since their first description a number of DN B-cell subsets have been identified. Initially DN B cells were split into DN1 (CD21⁺CXCR5⁺) and DN2 B cells (CD21⁻CXCR5⁻) [26]. Similar to aNAV, DN2 also have low expression of CD24 (Fig. 6A). DN2 cells were found to be a rare subset in health but are found in much higher frequencies in SLE [26]. DN2 B cells also highly express CD11c as well as the transcription factor T-bet highly express CD1c and FCRL5 as well as the transcription factor T-bet (Fig. 6B and C) [26, 116]. Functionally, DN2 cells are hyperresponsive to TLR7 signalling and can undergo rapid ASC differentiation in response to IL21 and IFN-γ and are therefore felt to be pivotal in extrafollicular B-cell responses [116, 163]. DN2 cells have been identified to be clonally related to aNAV B cells and ASC with a lower level of IgV gene mutations and are enriched in autoreactive cells [27]. More recently DN3 and DN4 subsets have been reported, the DN3 cluster being CD11c⁻CXCR5⁻ and the DN4 subset being enriched in IgE expressing cells and to highly express transcripts for IL4R and CD40 [72, 73].

Figure 6.

DN B cells. (A) Flow cytometry plots demonstrating a gating strategy used to identify DN1 and DN2 DN B cells. DN1 cells are CD27⁻IgD⁻CD21⁺CD24⁺ and DN2 cells are CD27⁻IgD⁻CD21⁻CD24⁻. Data are from a representative HCD. (B) SPADE on viSNE plot from mass cytometry data showing expression of CD21, CD11c, and CXCR5 used to identify DN1 (CD21⁺CD11c⁻CXCR5⁺), DN2 (CD21⁻CD11c⁺CXCR5⁻), DN3 (CD21⁻CD11c⁻CXCR5⁻), and DN4 (CD21⁺CD11c⁺CXCR5⁺). (C) Heatmap demonstrating marker expression by DN B-cell subsets from mass cytometry data. (D) Comparative SPADE on viSNE plots demonstrating expansion of DN2 cells in SLE compared to health (Tull TJ et al. [9])

SLE disease flares are associated with dramatic increases in ASCs that secrete pathogenic autoantibodies that are derived from extrafollicular B-cell responses [27]. This is associated with expansion of aNAV and DN2 B-cell populations (Fig. 6D) which are proposed to differentiate into ASCs via extrafollicular B-cell responses [116]. DN2 B cells are more frequent in SLE patients of African–American ethnicity and with lupus nephritis in whom they have been identified within the nephrotic kidney [26, 163]. Extrafollicular B-cell maturation has been correlated with the severity of COVID-19 and similar to severe SLE, increased frequencies of aNAV, DN2 cells are observed alongside depletion of MZB cells [9, 41, 60, 78, 116]. The correlation between the severity of infection and B-cell subset abnormalities suggests a pivotal role of cytokines such as Type II interferon. CD11c positive B cells have been reported in patients with MS, RA, and SSc and to accumulate with age in female patients with RA [164]. Autoreactive CD27⁻CD19^hiCD21^lo B cells akin to DN2 cells have also been reported in SS [165]. Increased frequency of DN B cells has also been correlated with improved clinical responses to rituximab of patients with RA whilst reduced numbers have been reported to predict response to IL-6 inhibition with tocilizumab [77, 166]. Further work is needed to investigate the clinical significance of these findings. DN cells may be of pathogenic significance through the production of inflammatory cytokines or differentiation into ASCs. Alternatively, altered DN cell frequency may be indicative of certain inflammatory environments that affect treatment outcomes.

Conclusion

The evolution of high dimensional phenotypic tools has greatly advanced our understanding of B-cell subsets in blood and the surface markers and transcriptomic properties that define them. However, significant gaps remain in our understanding of the functional roles of these subsets, their developmental origins and pathogenic roles in disease. By addressing these areas of uncertainty, we will advance our knowledge of human B-cell development and the role these cells play in normal homeostasis and in the pathogenesis of autoimmune and inflammatory disease.

Materials and methods for data used in figures and figure generation

Experimental subjects

Blood samples were donated by adult HCD and patients with SLE with regional ethics committee (REC) approval and informed consent (REC reference 11/LO/1433).

Isolation and storage of PBMCs

PBMCs were obtained from whole blood by Ficoll-Hypaque density gradient centrifugation as previously described [9]. Isolated PBMCs were then centrifuged and cryopreserved in ice-cold freezing medium (FCS with 10% DMSO) to give a cell concentration of 5–10 × 10⁶ cells/ml.

Mass cytometry staining and analysis

The staining protocol and antibody panel for the mass cytometry data displayed in Figs. 1 and 2 is described by Tull TJ et al. [9]. FCS files were normalized using Nolan Laboratory Software (https://github.com/nolanlab/beadnormalization). Single live B cells from 10 HCD (5 male and 5 female, mean age 31.8 years) are gated as displayed in Fig. 1B. ViSNE plots displayed in Fig. 1C are generated using Cytobank software (https://mrc.cytobank.org/) using equal numbers of B cells (n = 35 000) from 10 HCD. SPADE was then run on the viSNE coordinates and nodes corresponding to B-cell subsets are grouped as displayed in Fig. 1D. Transitional B cells were identified as CD27⁻IgD+CD24^+++/++CD38^+++/++. Events within this transitional cell bubble were then exported and a further viSNE run using equal events (n = 3535) and all panel markers except CD45, CD3, CD14, and class-switched isotypes IgA and IgG, which are not expressed by transitional B cells. The SPADE plots in Fig. 1C are generated from these viSNE coordinates. SPADE on viSNE plots in Fig. 4C are generated by exporting events within the MZB SPADE bubble in Fig. 1D and re-clustered using CD45RB, IgD, CD21, integrin beta 7, CD27, CD24, IgM, HLA-DR, and CCR7. The DN SPADE on viSNE was generated using data from 8 HCD and 8 patients with active severe SLE as described by Tull TJ et al. [9]. CD19⁺ cells were gated and SPADE on viSNE plots created as in Fig. 1D. Events within the DN SPADE bubble were then exported and DN nodes were re-clustered using CD11c, CD21, CXCR5, and CD24 to identify DN subsets.

Flow cytometry staining and analysis

Cryopreserved PBMCs were stained with the fluorochrome-conjugated antibodies as previously described [9]. The following antibodies clones were used: CD27 APC (M-T271), IgD APC-Cy7 (IA6-2), CD19 PerCP-Cy5.5 (HIB19), CD10 BV421 (HI10a), CD24 BV605 (ML5), IgM BV711 (MHM-88), CD1c PE (L161), and CD45RB PE-TexasRed (MEM-55). CD21 BUV737 (HB7) and CD38 BUV395 (B-ly4). DAPI was used for live dead staining. Flow cytometry data were obtained by running samples on a Fortessa flow cytometer. FCS files obtained were analysed using FlowJo software.

Glossary

Abbreviations

ABC

age associated B cell

aNAV

activated naïve

ASC

antibody secreting cell

BAFF

B cell activating factor

BCR

B cell receptor

Breg

B regulatory cell

CSM

class switched memory

DN

double negative

IgV

immunoglobulin variable

HCD

healthy control donor

GALT

gut associated lymphoid tissue

IFN

interferon

MFI

median fluorescence intensity

MS

multiple sclerosis

MTG

mitotracker

MZB

marginal zone B cell

MZP

marginal zone precursor

RA

rheumatoid arthritis

RH

rhodamine

SLE

systemic lupus erythematosus

SPADE

spanning tree progression of density normalized events

SS

Sjogren’s syndrome

SSc

systemic sclerosis

TLR

toll like receptor

TNF

tumour necrosis factor

Conflict of interest

None declared.

Funding

This work was funded by the Medical Research Council of Great Britain (MR/R000964/1) and the St Thomas’ Lupus Trust.

Data availability

Raw mass cytometry data can be requested from the corresponding author.

Author contributions

T.J.T.: Manuscript generation and figure review. R.L.V.: Figure generation and manuscript review.

Permission to reproduce

Granted within terms set out by the licensing agreement.

Clinical trial registration

Not applicable.

The animal research adheres to the ARRIVE guidelines

Not applicable.