Polychromatic flow cytometry in evaluating rheumatic disease patients

As extensively reviewed elsewhere [3,4], the customarily used IgD/CD27 scheme classifies human peripheral blood CD19⁺ B cells into four core subsets: naïve IgD⁺CD27⁻ B cells, unswitched memory (UM) IgD⁺CD27⁺ B cells, switched memory (SM) IgD⁻CD27⁺ B cells and double-negative (DN) IgD⁻CD27⁻ switched B cells (refer to Table 1 for definitions). Plasmablasts are a rare population in steady-state healthy subjects and can be better discriminated as CD27⁺⁺CD38⁺⁺ cells within the IgD⁻ fraction. It should be noted that, in addition to mature naïve B cells, the IgD⁺CD27⁻ compartment also harbors transitional B cells. Although the fraction of transitional B cells in this compartment is fairly small in healthy subjects, it can be quite prominent in patients with autoimmune diseases such as systemic lupus erythematosus (SLE) either in untreated disease [5] or after B-cell depletion therapy [6]. Transitional B cells have traditionally been identified as CD24⁺⁺CD38⁺⁺ cells, and they can be distinguished from naive B cells in the IgD⁺CD27⁻ compartment by their lack of expression of the ABCB1 transporter and the resulting retention of dyes such as Rhodamine 123 and MitoTracker Green [7].

Substantial phenotypic heterogeneity has been recognized among human memory B cells (as defined by the expression of CD27), although their functional heterogeneity is less well understood [8]. Approximately one-half of all human CD27⁺ memory B cells have undergone isotype switch (IgG⁺ and IgA⁺), and the rest express surface IgM with or without the concomitant expression of surface IgD [8,9]. While the classical switched memory (SM) cells are generated from the germinal center reaction, IgD⁺IgM⁺CD27⁺ unswitched memory (UM) cells have been proposed to represent circulating marginal zone B cells, which are critical for protection against infections with encapsulated bacteria [10]. A variable fraction of CD27⁺ memory cells express only surface IgM (IgM-only memory) and may represent pre-SM cells that will eventually join the pool of isotype SM cells after participating in subsequent germinal center reactions [9,11].

Also well established is the existence of a subset of isotype switched B cells lacking expression of CD27, an antigen widely considered a universal marker of human memory cells [12,13]. These cells are comparable with conventional CD27⁺ SM cells in that they are class switched and somatically mutated and they experience much greater proliferative responses than naïve B cells after TLR9 stimulation through CpG DNA in the absence of simultaneous B-cell receptor engagement. IgD⁻CD27⁻ double-negative (DN) switched cells undergo substantial expansion in SLE patients, and the degree of expansion correlates well with disease activity [12]. Similar to CD27⁺ SM cells, the IgD⁻CD27⁻ DN compartment contains a fraction of IgM-only cells as well as class switched IgG and IgA cells [12]. Overall, the origin and role of IgD⁻CD27⁻ DN B cells remain to be understood although a derivation from initial germinal center reactions has been suggested in some studies [14]. DN B cells resemble a tissue-based memory population phenotypically, but they do not express the characteristic FcRL4 cell surface marker in the peripheral blood of both healthy and SLE subjects [12]. However, FcRL4 expression in these cells was observed in HIV-infected viremic individuals and those chronically infected with malaria [15,16]. It has been suggested that, at least in chronic infections such as HIV and malaria, DN cells may represent prematurely exhausted cells owing to the influence of FcRL4 and possibly other inhibitory receptors [15-17]. Nonetheless, other studies have suggested their active participation in the generation of anti-malaria antibodies [18].

The concept of effector and regulatory functions of B cells in cellular immune responses has received great attention in recent years. Hence, it is important to discuss the different phenotypes proposed for regulatory B cells (Bregs), a population with protective effects in autoimmune conditions [19] and whose preservation or enhancement should be an important consideration in the design of B-cell targeting therapies. Bregs suppress inflammation and autoimmunity through the production of cytokine interleukin (IL)-10. In the mouse, Breg function has been ascribed to different cell types including B1 cells [20], marginal zone B cells [21], B10 cells with a CD1d^hiCD5⁺ phenotype [22] and transitional cells [23]. Similarly, Breg function has been proposed in the human for naïve B cells (IgD⁺CD27⁻) [24], transitional B cells (CD24^hiCD38^hi) [25], B10 cells (CD24^hiCD27⁺) [26] and orchestrator B1 cells (B1orc) [27]. Mouse B cells of the B1 lineage have been known to be a rich source of IL-10 [20], but the identity of the human B1 counterpart remains elusive.

Recently, a population of human peripheral blood B cells with a CD20⁺IgD⁺CD27⁺CD43⁺CD70⁻ phenotype has been shown to exhibit the functional hallmarks of the mouse B1 cells [28], although its actual significance and magnitude remain to be further explored. Of note, the CD11b⁺ fraction of this human B1 population, termed B1orc, spontaneously secretes IL-10 and suppresses T-cell activation [27]. Mouse B10 cells, designated to represent splenic IL-10-producing CD1d^hiCD5⁺ B cells, share some phenotypic markers with other IL-10-producing cells including B1 and marginal zone B cells [22]. Human B10 cells, on the other hand, are predominantly found within the CD24^hiCD27⁺ compartment [26].

Given the diverse cell types capable of exerting regulatory function, there is no definitive cell surface marker(s) that can serve as a surrogate for IL-10 production. Furthermore, there might be subtle differences in the regulatory capacity among the different Breg subsets. For instance, suppression of CD4⁺ T-cell proinflammatory cytokine tumor necrosis factor alpha (TNFα) production by human transitional B cells is dependent on IL-10 [25]. In contrast, human B10 cells do not seem to regulate CD4⁺ T-cell TNFα expression, although they do suppress TNFα production by monocytes [26].

Other unique populations of B cells can be relevant for specific autoimmune diseases. For instance, a group of B cells recognized by an anti-human idiotype 9G4 antibody represents a highly informative experimental model to understand the breakdown of B-cell tolerance in SLE. In healthy subjects, effective tolerance ensures that 9G4 responses are restricted to acute infections with mycoplasma and Epstein–Barr virus, and that they do not persist in the long-lived IgG memory and plasma cell compartments [29]. In contrast, we have shown that 9G4⁺ B cells are substantially expanded in the SLE IgG memory B-cell compartment, and 9G4⁺ antibodies contribute disproportionally to circulating IgG levels owing to defective germinal center censoring [30,31]. Among other autoreactivities, 9G4⁺ antibodies have been shown to constitute a major species of anti-apoptotic cell antibodies in SLE serum [32]. CD19^hi memory B cells are enriched in anti-Sm B cells in SLE, and the degree of enrichment correlates with the level of serum anti-Sm antibodies as well as with adverse outcome and poor response to ritxumibab in small studies [33].

Another B-cell subset, which lacks the expression of RP105, is increased in the peripheral blood of patients with SLE, Sjogren’s syndrome and dermatomyositis [34]. Particularly in SLE, the increase in this B-cell population, which has been shown to produce anti-double-stranded DNA antibodies, appears to correlate with disease activity [34]. Several lines of evidence, including surface phenotype (CD20⁻CD38^hiCD138^dull) and spontaneous production of antibodies in vitro, suggest that these RP105⁻ B cells consist of antibody secreting cells (ASC) [34]. Hence, the observed increase of RP105⁻ B cells is consistent with the expansion of ASC in SLE patients with active disease (discussed in the next section).

B cells play a critical role in the pathogenesis of autoimmune diseases, so B-cell-targeted therapies have become an attractive treatment modality. Depending on the mechanisms of action, B-cell-targeting agents can be categorized into those that directly kill most B cells and those that compromise the survival, differentiation and activation of B cells. The latter class of agents tends to target discreet B-cell subsets. The efficacy of these more selective B-cell targeting agents will thus depend on their effect on specific B-cell subsets and the contribution of the affected subsets to regulatory or pathogenic functions. A precise understanding of the phenotype and function of different B-cell subsets is therefore of the essence for a rational design of B-cell-targeted therapies.

Rituximab, a chimeric anti-CD20 monoclonal antibody, was the first B-cell-targeting biologic agent to receive US Food and Drug Administration approval for the treatment of autoimmune rheumatic diseases. Rituximab induces universal depletion of all B cells except those that lack the expression of CD20, such as pro-B cells and plasma cells. Nonetheless, small numbers of residual memory B cells as well as plasmablasts can be detected in the peripheral blood at the point of maximal depletion even in patients with effective B-cell depletion [76]. Studies in SLE patients treated with rituximab show that different patterns of B-cell reconstitution would emerge that correlate well with the clinical outcomes of the treatment. Short-term responders are characterized by rapid accumulation of memory B cells and plasmablasts [76,77], most probably resulting from the preferential homeostatic proliferation and expansion of these residual cells. On the contrary, long-term responders demonstrate a delayed memory B-cell recovery and a prolonged expansion of transitional B cells [6,77]. Given the success of treating SLE with rituximab in many open studies, the failure of two recent randomized, placebo-controlled trials (EXPLORER and LUNAR) to show added values from rituximab over conventional therapy was quite unexpected [78,79]. Aside from the other plausible explanations for the failure, the degree of initial B-cell depletion might have impacted the efficacy [80], as demonstrated by the recent studies in which a deeper B-cell depletion increases the efficacy of rituximab treatment in RA [81,82].

In contrast to direct killing of pan B cells by rituximab, other B-cell-targeted therapies induce quite different B-cell changes reflecting different mechanisms of action. For instance, belimumab, a monoclonal antibody that blocks BAFF binding to its receptors, preferentially inhibits the survival and hence decreases the numbers of transitional and activated naïve B cells [83,84]. The numbers of CD27⁺ SM cells and plasma cells are not affected, indicating these subsets are independent of BAFF for survival. However, a subset of IgD⁻CD27⁻ DN switched cells undergoes significant and sustained reduction [83]. Reflecting both the impact of sample size and duration of follow-up as well as the impact of different definitions and measurement of seemingly similar cell types, discordant results have been reported regarding the impact of belimumab on plasmablasts and other ASC [83,84]. Epratuzumab, another monoclonal antibody, exerts an agonistic effect on the inhibitor receptor CD22, further dampening B-cell activation. In contrast to rituximab, epratuzumab does not drastically deplete circulating B cells, but induces an average reduction of peripheral B cells by 30% mainly in the CD27⁻ compartment [85], which includes transitional cells, naïve cells as well as IgD⁻CD27⁻ switched cells. Attenuation of the B-cell receptor signaling pathway can also be mediated through the inhibition of tyrosine kinases such as Syk and Btk by small molecules that were developed initially to treat B-cell lymphomas [86]. Although the effects of these inhibitors on nonmalignant B cells are largely unknown, a recent study shows that short-term use of the Syk inhibitor fostamatinib in lymphoma patients impairs B-cell development at the transitional stage without affecting mature B-cell populations [87]. Even though two recent phase 3 clinical trials of fostamatinib in RA were a disappointment, other agents that target the B-cell receptor signaling pathway hold significant promises in treating autoimmune diseases [88].

Our knowledge of human B-cell subsets and of the perturbation of their homeostasis in disease could be exploited to apply B-cell profiling as a means to optimize disease diagnosis, prognosis and treatment. To achieve this goal, a comprehensive B-cell phenotyping is of the essence. Hence, we have developed several 12-color panels for in-depth characterization of memory cells, naïve/transitional cells and ASC [2-4] (Table 2). These panels share seven anchor markers, a B-cell lineage and two exclusion markers (CD19, CD3 and Live/Dead) as well as four developmental markers (IgD, CD27, CD38 and CD24), which enable the precise identification of the same core human B-cell subsets across panels (Figure 1A). Also common to all three panels is an anti-idiotype 9G4 antibody, which provides a useful measure of autoreactivity through the identification of B cells expressing autoantibodies encoded by the VH4-34 variable region gene [30]. Panel-specific markers then allow in-depth characterization of these core B-cell subsets and aid the identification of potentially novel subsets. The incorporation of CD21, CD95 and CXCR3 in the memory panel thus provides information regarding the activation status and homing potential of the memory B cells. The addition of MitoTracker Green in the transitional panel further segregates the late transitional (T3) cells from the resting naïve population. CD138 and Ki-67 in the plasma cell panel provide additional information on the subsets and proliferation status of plasmablasts/plasma cells (Figure 1).

Increasingly complex high-dimensional PFC data create new challenges for data mining and interpretation. Just as challenging is the difficulty in the level of standardization required for large datasets and multicenter studies typical of large clinical trials [89]. These challenges are being proactively addressed by many groups, often on a collaborative basis, to develop various clustering algorithms that can identify discrete cell populations based on simultaneous assessment of multiple parameters, and hold significant promise for the automated analysis of PFC data [90,91]. To overcome the time-consuming and variable nature of manual gating, a normalization algorithm has been developed that, when integrated into the manual template gating procedure, is capable of mitigating the sample-to-sample variation and allows for high-throughput processing of large PFC datasets [89]. Likewise, recently developed software called AutoGate is promising to become an automated tool for processing and analyzing PFC data [92].

Traditionally, the frequency (or absolute number) of each B-cell subset derived from flow cytometry analysis is presented independent of that of other subsets, largely as part of univariate analysis. Perturbation of B-cell homeostasis in a disease state is often described separately for each affected subset as discussed in the previous section. However, univariate approaches on individual subsets fail to reveal how collections of subsets and their relative distributions might contribute to patient groupings. Thus, we have applied a global B-cell profiling approach, in which all of the B-cell subset data are considered simultaneously to obtain a system-wide view of B-cell populations [4,51,93]. In this manner, patient-specific complex B-cell fingerprints are generated that can be directly compared with the profile of other patients. An unsupervised hierarchical clustering analysis can then divide patients into groups based on their B-cell profiles (Figure 2), and stringent correlations of B-cell fingerprints with clinical, immunological and other emergent features can be identified. Our results from a large multicenter study [93] (and manuscript in preparation) provide a proof of concept that, when combined with other informative clinical parameters, B-cell profiling offers a systems biology approach to identifying potential biomarkers for the diagnosis, prognosis and treatment monitoring of lupus disease.