Characterization of Lymphocyte Subsets in Patients with Common Variable Immunodeficiency Reveals Subsets of Naive Human B Cells Marked by CD24 Expression

Increased proportions of naive B cell subset and B cells defined as CD27negCD21negCD38neg are frequently found in patients with common variable immunodeficiency (CVID) syndrome. Current methods of polychromatic flow cytometry and PCR-based detection of κ deletion excision circles allow for fine definitions and replication history mapping of infrequent B cell subsets. We have analyzed B cells from 48 patients with CVID and 49 healthy controls to examine phenotype, frequency, and proliferation history of naive B cell subsets. Consistent with previous studies, we have described two groups of patients with normal (CVID-21norm) or increased (CVID-21lo) proportions of CD27negCD21negCD38neg B cells. Upon further analyses, we found two discrete subpopulations of this subset based on the expression of CD24. The B cell subsets showed a markedly increased proliferation in CVID-21lo patients as compared with healthy controls, suggesting developmental arrest rather than increased bone marrow output. Furthermore, when we analyzed CD21pos naive B cells, we found two different subpopulations based on IgM and CD24 expression. They correspond to follicular (FO) I and FO II cells previously described in mice. FO I subset is significantly underrepresented in CVID-21lo patients. A comparison of the replication history of naive B cell subsets in CVID patients and healthy controls implies refined naive B cell developmental scheme, in which human transitional B cells develop into FO II and FO I. We propose that the CD27negCD21negCD38neg B cells increased in some of the CVID patients originate from the two FO subsets after loss of CD21 expression.

C ommon variable immunodeficiency (CVID) is a clinically heterogeneous group of primary Ab immunodeficiency diseases with an estimated prevalence of 1 in 30,000 with unclear etiology (1). It is characterized by decreased serum levels of IgG, IgA, and occasionally IgM and impaired Ab response to Ag stimulation and/or vaccination as well; it is usually accompanied by recurrent bacterial infections. Splenomegaly, granuloma formation, autoimmune disorders, and increased risk of malignancy are common complications (1,2).
Previous studies have shown abnormalities in B cell subset composition -a diminution in memory B cells and increase in naive follicular (FO) B cells (5). Naive FO B cells IgM pos CD27 neg (17,18) constitute a major B cell subset in the peripheral blood of patients with CVID. In some CVID patients, increased numbers of transitional B cells and CD21 neg B cells were observed (11,19,20). The CD21 neg B cell subpopulation was originally described as CD19 high CD21 neg B cells and has since developed into the present characterization CD21 neg CD38 neg CD19 high CD27 neg to distinguish this population from other CD21 neg B cell populations: CD21 neg plasmablast, CD21 neg transitional B cells. Expansion of the CD21 neg B cell subset has been found in patients suffering from systemic lupus erythematosus or HIV infection (21,22). In patients with systemic lupus erythematosus, the CD21 neg subpopulation represents activated B cells, as these cells express activation markers CD86 and CD95 (21,23). In CVID patients, CD21 neg B cells were shown to have increased expression of these activation markers as well as low expression of CD24. Although CD21 neg B cells were found among CD27 neg naive B cells in CVID patients, recent results showed that CD21 neg B cells in CVID patients were preactivated, polyclonal, partially autoreactive B cells that home to peripheral tissues (23).
The heterogeneity of naive CD27 neg B cell subpopulations has not been thoroughly studied. The CD21 neg B cell subpopulation forms ,10% of all B cells in healthy controls (5,20). The CD21 pos naive CD27 neg B cell subset can be further divided into two groups, as follows: 1) transitional immature B cells (IgM high CD24 high CD38 high ) making up ∼3% of B cells (24), and 2) the remaining CD27 neg B cells that are comprised of mature IgM pos CD24 pos CD38 int CD27 neg FO B cells, which reside in the spleen and lymph nodes and also circulate through the body. FO B cells are primed by T cells to generate class-switched memory and plasma cells producing high-affinity Abs. Two FO B cell recirculating popu-FIGURE 1. Scheme for gating B cell subsets. A, Gating of B cells is shown for a representative healthy donor. Lymphocytes were gated on forward and side light scatter and then on CD19 positivity, after which the remaining cells were divided into CD21 pos and CD21 neg subsets. Furthermore, naive B cells were defined as IgM pos and CD27 neg . The CD21 pos IgM pos CD27 neg cells were subdivided into transitional and naive subsets. The CD21 neg IgM pos CD27 neg subset was increased in some CVID patients. The expression of IgM and CD24 defined two subsets of naive CD21 pos (B) and CD21 neg (C) B cells. Both subsets were observed in a CVID patient and in a healthy control. Overlays of CD19 histograms show differences between CD21 neg CD24 neg (black) and CD21 neg CD24 pos (gray) cells, whereas no difference was found between CD21 pos CD24 neg (black) and CD21 pos CD24 pos (gray) cells. D, Bar graph presenting the frequency of CD21 neg CD24 pos /CD21 neg CD24 neg cells in individual patients and controls. lations, the IgD high IgM low CD21 int CD24 int and the IgD high IgM high CD21 int CD24 high , designated FO I and FO II B cells, respectively, were recently described in mice (25). FO I B cells require Agderived signals and make up the majority of cells in the recirculating FO B cell pool. FO II B cells can develop in the absence of Ag into marginal zone B cells (26). To date, no corresponding B cell subpopulations have been described in humans.
We focused on a detailed analysis of naive CD27 neg B cells using polychromatic flow cytometry (FC) in healthy human donors and in CVID patients. Consistent with previous studies, we describe two groups of patients with normal (CVID-21norm) or increased (CVID-21lo) proportions of CD27 neg CD21 neg CD38 neg B cells. The CVID-21lo group contains both the smB 2 21 low (decreased number of CD27 pos IgD neg IgM neg switched memory B cells and increased number of CD21 neg B cells) and smB + 21 low (normal number of switched memory B cells and increased number of CD21 neg B cells) patients, as defined by Wehr et al. (20), whereas the CVID-21norm group comprises smB 2 21 norm (decreased number of switched memory B cells and normal number of CD21 neg B cells) and smB + 21 norm (normal number of switched memory B cells and CD21 neg B cells) patients. To assess the developmental relationships between subsets identified by FC, we performed a replication history analysis in these cells by means of k-deleting recombination excision circles (KRECs) (27). We describe two phenotypically distinct subpopulations of CD21 neg B cells that are expanded in CVID patients, but can be also identified in healthy controls. Furthermore, we identified cells in human peripheral blood that correspond to mouse FO I and FO II B cells. The frequencies, replication history, and mutual proportions of these populations differ between healthy controls and CVID patients. Based on these findings, we propose a revised developmental scheme of human naive B cell development.

Patients and healthy donors
We examined 48 CVID patients (25 females, 23 males; mean of age 43.4 6 15.3 y, range 12-74 y) and 49 healthy donors (26 females, 23 males; mean of age 42.6 6 13.6 y, range 20-78 y). In all CVID patients, B cells constituted .1% lymphocytes in the peripheral blood. The presence of polymorphisms in the TNFR family member transmembrane activator and cyclophilin ligand interactor gene (TNFRSF13B) was analyzed in all patients. The p.C104R polymorphism was found in two patients, and the p.I87N polymorphism was found in one patient; all three of the identified polymorphic genes were heterozygous. In 15 patients, ICOS gene mutations were analyzed with negative results. The patients were on regular i.v. or s.c. Ig substitution. For patients who received i.v. Ig treatment, blood samples were always collected before the treatment. Informed consent was approved by the Medical Ethics Committee of St. Anne's Faculty Hospital and was obtained before blood sampling.
CVID patients were divided into two groups based on the expression patterns of CD21 and CD38 on the patients' B cells. Twenty-four patients fell into the CVID-21lo group, which was characterized by an increased proportion of B cells with low expression of CD21 and CD38 (.10%). The remaining 24 patients were placed in the CVID-21norm group, which was characterized by a normal frequency (,10%) of CD21 low CD38 low B cells.
The raw FC data from these patients and the control cohort were used in a previously published technical study describing probability binning algorithms for the evaluation of multicolor FC data (28). FC data analysis was performed using the FlowJo 7.5 or 8.8.7 software (Tree Star, Ashland, OR).

Investigation of B cell replication history
B cell subsets were purified from the blood samples of 16 CVID patients (8 CVID-21lo, 8 CVID-21norm). To ensure that we had sufficient numbers of cells belonging to subpopulations that were not expanded in healthy controls, we used cells that were sorted from 11 buffy coats (instead of peripheral blood) as control subpopulations. The cells were sorted using a FACSAria flow cytometer (BD Biosciences) using the same set of mAbs that were described above. DNA from the sorted cell populations was extracted using the QIAamp DNA Blood Micro Kit (Qiagen, Hilden, Germany). We used the KREC detection method that was described previously (27) with in-house modifications, as follows. Unlike the method that was described by van Zelm et al. (27), we used a d cycle threshold method with a calibration sample consisting of sorted transitional B cells from a healthy donor that was split into two quantitative PCR reactions. The normalized d cycle threshold (KREC) values indicated the number of cell divisions in a given subpopulation in relation to the calibration sample. If the KREC amplification produced no signal within the sensitivity limit of the quantitative PCR reaction (e.g., the KRECs in a given sample were too dilute), we used the lowest possible number of cell divisions that was calculated using the number of intronRSS-KDE rearrangements in the sample.

Statistical analysis
The numerical differences between the subgroups were evaluated using the nonparametric Mann-Whitney U test and Wilcoxon test. In all statistical analyses, p , 0.05 was considered statistically significant. Student's t test and Fisher's exact test were used when appropriate. Statistical software STATISTICA (StatSoft, Tulsa, OK) version 7 was used.
The CD27 neg CD21 neg CD38 neg B cell subset is composed of two cell types that are defined by CD24 expression For cytometric analysis, all B cells were divided into naive and memory B cell populations, based on the expression of CD27 (Fig.  1A). We confirmed the previously described increased frequency of naive IgM pos CD27 neg B cells that was observed in subgroups of CVID patients (Table I). In CVID-21lo patients, the total B cell population contained an increased percentage of CD21 neg CD38 pos B cells, which was expected, given that this patient group was defined by this characteristic (Table I).
Unlike other groups (23), we found that the CD27 neg CD21 neg CD38 neg B cell subset is composed of two cell types that are defined by CD24 expression: CD27 neg CD21 neg CD38 neg IgM pos CD24 neg cells (referred to in this study as CD21 neg 24 neg cells) and CD27 neg CD21 neg ICD38 neg gM pos CD24 pos cells (referred to in this study as CD21 neg 24 pos cells) (Fig. 1C). Both of these cell types contributed to the increased frequency of CD27 neg CD21 neg CD38 neg B cells, but their relative contributions varied among patients (Fig. 1D).
Previous studies found elevated expression levels of CD19 in the CD21 neg naive B cell subset, compared with total B cells (11,19,20,23). Our findings show that this elevation was restricted to the CD21 neg CD24 neg subset in both healthy controls and CVID patients, although the elevation was more prominent in CVID patients (Fig. 1C, Table I).
Although CVID-21lo patients displayed relatively increased frequencies of naive IgM pos CD27 neg B cells, the increase was caused primarily by higher numbers of CD21 neg CD38 neg B cells. The frequency of naive B cells with the typical CD21 pos phenotype was actually decreased in CVID-21lo patients compared with the frequency of naive CD21 pos B cells in healthy controls, whereas naive CD21 pos B cells were significantly increased in CVID-21norm patients ( Table I). The proportion of transitional B cells was higher in both groups of CVID patients compared with healthy controls (Fig. 2A), as already reported (20).

Replication history of the naive B cell subset (excluding transitional B cells) showed a higher number of cell divisions in CVID patients
The analysis of KRECs in transitional B cells showed that the number of cell divisions in both of the CVID subgroups was similar to that in healthy donors. However, in CVID-21lo patients, the CD21 neg CD24 neg subset was found to undergo a higher number of divisions than in controls (Fig. 2B). The same pattern (although not significant) was found for the CD21 neg CD24 pos B cell subset.

Subsets of naive B cells corresponding to mouse FO I and FO II were found in humans and were underrepresented in CVID-21lo patients
Two FO B cell populations, CD21 int IgM low IgD high CD24 pos and CD21 int IgM high IgD high CD24 high , that were designated FO I and FO II, respectively, were recently described in mice. Cariappa et al. (25) distinguished mouse FO II cells from FO I cells according to the enhanced expression of IgM in FO II B cells. Both subpopulations showed increased expression of CD24. When we analyzed CD21 pos CD27 neg CD38 int B cells, we found two subpopulations of cells: IgM high CD21 pos CD27 neg CD38 int CD24 pos and IgM int CD21 pos CD27 neg CD38 int CD24 int . We observed both subpopulations in human healthy donors and in CVID patients. The FO I/FO II difference was marked by enhanced CD24 expression on FO II-like cells in CVID patients, whereas in healthy individuals, the CD24 expression was only moderately increased on FO II cells (Fig. 1B). In mice, the distinction between FO subsets from marginal zone B cells is possible because of the enhanced expression of CD21 on marginal zone B cell subsets. Both subpopulations that we found in human peripheral blood had identical CD21 expression levels. The expression of CD21 was higher than it was in marginal zone-like B cells (IgM pos CD27 pos CD38 neg CD24 high ) (Supplemental Fig. 1). Because we observed two subpopulations with uniform expression of CD21, we assumed that the IgM high CD27 neg CD38 int CD24 pos subset corresponds to FO II B cells and that the IgM int CD27 neg CD38 int CD24 int subset corresponds to FO I B cells.

Different responses to activation stimuli by FO I and FO II cells in human B cells
The mouse model that was reported by Cariappa et al. (25,26) suggests that FO I B cells develop from FO II B cells after strong BCR stimulation, and that FO II B cells can give rise to MZ-like cells. To support the phenotype distinction between FO I and FO II B cells in human B cells using functional data, we stimulated human B cells from healthy donors in vitro using a TLR9 agonist and a BCR agonist. We chose the Peptostreptococcus magnus protein L superantigen that binds to conserved k L chains (29) as the BCR agonist because anti-IgM stimulation does not allow for a surface IgM-based definition of the responding B cells. The FO I and FO II definition was based solely on surface IgM in this experiment because CD24 is rapidly downregulated after activation by BCRs (30).
Both agonists triggered cell activation leading to CD69 upregulation and cell proliferation (Fig. 3A, 3B), whereas apoptosis was always below 9% and did not differ between stimulated or control cells at 2 h, 24 h, 48 h, or 3 d after stimulation (Supplemental Fig. 2). In line with the data for the mouse model, the BCR agonist led to the activation of ∼55% of the naive B cells (corresponding to the proportion of k L chain-positive B cells) that had acquired a predominantly FO I phenotype (Fig. 3A, 3C). In contrast, the TLR9 agonist activated all naive B cells and forced ∼40% of cells to proliferate by day 3 after stimulation. Of note, all proliferating B cells displayed the FO II phenotype (Fig. 3B), which is in line with the lack of cell divisions that was documented in the FO I compartment ex vivo (Fig. 3C).

The FO I compartment is disrupted in CVID patients
The naive CD21 pos B cell compartment is composed of FO I, FO II, and transitional B cells. The CVID-21lo patients had an increased percentage of CD21 neg B cells by definition, and thus had reciprocally decreased percentages of CD21 pos B cells. We also found an elevated frequency of transitional B cells, but the frequency of FO II cells was similar to that in healthy controls. Fig. 4A shows that the FO I subset of cells formed a significantly lower proportion of B cells in CVID-21lo patients than in healthy donors. Thus, the decrease of FO I B cells accounts for the diminution of CD21 pos B cells in CVID-21lo patients.
CVID-21norm patients presented with a higher frequency of CD21 pos B cells, and both of the FO subsets were relatively increased (Fig. 4A).
The number of cell divisions for FO I and FO II cells in healthy controls was similar to those of transitional B cells (Fig. 4B). However, in CVID-21norm patients, the FO I and FO II cells divided more often than the FO B cells from healthy controls. FO II B cells also divided more often in CVID-21lo patients.

Discussion
Based on the combined analyses of healthy donors and CVID patients, we identified new subsets of human naive B cells that are defined by IgM and CD24 expression. First, we found subsets corresponding to mouse FO I and FO II cells (25,31) in both healthy donors and CVID patients. Second, when analyzing the expanded CD27 neg CD21 neg CD38 neg B cell subset in CVID-21lo patients, we were able to distinguish between two distinct populations, based on CD24 expression. A detailed analysis of peripheral blood using sensitive six-color FC also revealed that both CD21 neg CD24 pos and CD21 neg CD24 neg subsets were present in healthy donors, although they were present in low numbers. This finding was in contrast to a previous study by Rakhmanov et al. (23), which showed that the entire CD21 neg population was also CD24 neg . This was presumably caused by the fact that in that study, CD21 neg cells were gated as CD19 high , whereas we observed that the expression of CD19 in CD21 neg CD24 pos B cells was similar to that of naive and memory B cells, that is, lower than in CD21 neg CD24 neg B cells. CD24 represents the first pan-B cell molecule, and this GPI-anchored protein belongs to the heat-stable protein family. CD24 is downregulated after BCR activation, but its complete physiological function in B cells is unknown (32). Further studies will be needed to determine whether this molecule plays an essential role in naive B cell homeostasis. Currently, CD24 is used in FC as a marker of the different stages of naive B cell development.
Low numbers of cell divisions in the FO II (IgM high CD24 pos ) and FO I (IgM int CD24 int ) subpopulations from healthy control individuals indicate that these cells mature from transitional B cells in the absence of proliferation, whereas the CD21 neg CD24 pos and CD21 neg CD24 neg subpopulations showed a higher number of cell divisions (Fig. 5). The median numbers of divisions in the latter subsets correspond to the number of divisions in the population of naive B cells, defined as CD19 pos IgD pos CD27 neg by van Zelm et al. (27). In contrast to healthy donors, the CVID-21lo and CVID-21norm patients displayed not only an altered distribution of naive B cells, but also elevated numbers of cell divisions in this compartment (Figs. 2, 5). We also found that the percentage of FO I B cells was strongly reduced in CVID-21lo patients. This may be explained by deficient responses to BCR signaling.
A marked decrease or absence of switched memory B cells and an increased frequency of naive B cells (e.g., CD21 neg CD38 neg in a subgroup of patients), together with an increased division rate, suggests a developmental arrest in normal bone marrow migration. Recent studies of a large cohort of patients showed that there was an increase in autoimmune phenomena in the CVID-21lo patient group (20). This may be a consequence of an oligoclonal expan-sion in the naive B cell compartment occurring during prolonged homeostatic proliferation or during proliferation in response to non-BCR stimuli, such as the TLR9 agonist that is exemplified in Fig. 3.
In contrast to the proportions of the examined B cell subsets in CVID-21lo patients, the proportions in CVID-21norm patients were inflated in all naive CD21 pos B cell compartments. FO I and FO II B cell subsets were found to undergo a higher number of divisions than controls. This result may be due to prolonged homeostatic proliferation, as described above.
Of note, both groups of CVID patients had an increased proportion of transitional B cells compared with healthy donors (20). However, these cells did not show significantly increased numbers of cell divisions compared with control transitional B cells. Thus, we speculate that increased bone marrow migration represents an attempt to compensate for the reduced numbers of mature B cells in the periphery of these patients.
Replication history data from healthy donors, together with the findings from Cariappa et al. (25), point to a sequence of subset development from transitional B cells to FO II (IgM high CD24 pos ) cells, and later, to FO I (IgM int CD24 int ) cells in the absence of cell division. The two later stage cell types can presumably give rise to the CD21 neg CD24 pos and CD21 neg CD24 neg subsets, respectively, as indicated by the fact that there are more cell divisions at these stages (Fig. 2B). Naive B cells in CVID patients fail to undergo the germinal center (GC) reaction and fail to enter the switched memory pool. These naive B cells were, however, shown to react to stimulation with anti-IgM plus IL-2, as demonstrated by the downregulation of CD24 (33), but the upregulation of the costimulatory molecules CD70 and CD86 was impaired. It is reasonable to suggest that the physiological (and likely repeated) stimulation of naive B cells by Ag and/or proinflammatory signals in CVID patients may lead to proliferation, after which these cells lose CD21 and become CD21 neg CD24 pos or CD21 neg CD24 neg cells (with added cell divisions) instead of undergoing the GC reaction. CD21 neg cells cannot participate in the GC reaction because CD21 expression is essential for GC formation, B cell FO dendritic cell contact, and B cell survival during the GC reaction (34).
In conclusion, we used IgM and CD24 surface expression patterns to refine our knowledge of naive peripheral B cell development, which is disrupted in CVID patients. We propose that cell proliferation without developmental progression is responsible for the accumulation of CD21 neg CD38 neg cells in a subset of CVID patients.

Disclosures
The authors have no financial conflicts of interest.