Germinal Center Founder Cells Display Propensity for Apoptosis before Onset of Somatic Mutation

Antibodies (clone number, isotype, and source) used for phenotyping and immunomagnetic bead depletion are listed in Table 1.

Tonsillar B cells were prepared as previously described (25). Briefly, tonsils taken from patients during routine tonsillectomy were finely minced and the resulting cell suspension was subjected to two rounds of depletion of non-B cells: (a) T cells were depleted by rosetting with sheep red blood cells, (b) residual non-B cells were depleted by T cell specific antibodies (CD2, CD3, and CD4), and then by magnetic beads coupled with anti–mouse IgG (Dynabeads; Dynal, Oslo, Norway). The resulting cells from all the experiments contained >98% CD19 positive B cells.

Total tonsillar B cells were incubated with mouse anti–human CD38-PE, goat anti–human IgD-biotin, and a set of FITC-conjugated mouse anti–human IgM, CD23, CD44, CD10, CD71, CD77, and Fas/CD95 for 20 min at 4°C. After washing twice with PBS containing 2% BSA, cells were incubated with streptavidin-tricolor for 30 min and analyzed with a FACScan^® flow cytometer. For intracellular Bcl-2 and nuclear Ki67 staining, cells were permeabilized by incubation with 3 g/100 ml saponin for 15 min at 4°C.

Total tonsillar B cells were incubated with mouse anti–human CD38-PE and goat anti–human IgD-biotin for 20 min at 4°C. After washing twice with PBS containing 2% BSA, the cells were incubated with mouse anti–human IgM-FITC and streptavidin-tricolor for 20 min. Then cells were washed twice and suspended in PBS at a concentration of 3 × 10⁶/ml. IgD⁺CD38⁺ B cells were separated into two subsets according to the expression of sIgM on a FACStar plus^®. IgM⁺IgD⁺CD38⁺ B cells from one tonsil were further size fractionated according to their forward scatter parameters.

10⁵ cells from each of the purified B cell subsets were cytocentrifuged for 5 min at 500 rpm on a microscope slide. Slides were fixed in methanol for 5 min and then stained with Giemsa staining solution (BDH Chemicals Ltd., Poole, England) diluted one in five with distilled water. Some slides were fixed in cold acetone at 4°C for 10 min for immunocytology. Slides were washed in PBS for 5 min and incubated with mouse mAb against Ki67 antigen (DAKO, Glostrup, Denmark) for 45 min. The slides were washed twice in PBS and then incubated with sheep anti–mouse Ig (The Binding Site, Birmingham, England). After 45 min, slides were washed and incubated with mouse mAb against alkaline phosphatase and alkaline phosphatase complexes (APAAP) (DAKO). After an additional 45 min, the slides were washed three times in PBS and the enzyme activity was developed by the FAST RED substrate (DAKO).

Cells were cultured in RPMI 1640 medium containing 10% heat inactivated fetal calf serum, 80 μg/ml gentamicin, and 2 mM glutamine (all from Flow Laboratories, Inc., MacLean, VA) at 37°C. Cells (2.5 × 10⁵/ml) were cultured for 5 d in one of the following conditions: IL-2 (10 U/ml), IL-4 (50 U/ml), or IL-10 (100 ng/ml). 2.5 × 10⁴ CD40-ligand transfected L cells and 2.5 × 10³ human fibroblasts from rheumatoid synovium (irradiated with 75Gy) were used for the cultures. DNA synthesis was assessed by an 8 h pulse with 1 μCi [³H]TdR before cell harvesting.

Portions of tonsils were snap frozen in liquid nitrogen and stored at −70°C. 5 μm frozen sections were cut and mounted on glass slides. They were thoroughly dried at room temperature for ≥1 h and were fixed in acetone at 4°C for 15 min. Sections were stained by double immunoenzyme technique using biotin-avidin-peroxidase system and alkaline phosphataseanti-alkaline phosphatase system (APAAP technique). Briefly, sections were washed in PBS for 5 min. Then sections were incubated with goat anti–human IgD-biotin and mouse anti–human IgM (IgG1 isotype). After washing for 5 min in PBS, the sections were incubated with streptavidine-peroxidase and sheep anti– mouse IgG1 for 30 min, and then incubated with alkaline phosphatase coupled to mouse antibodies specific for alkaline phosphatase (APAAP complexes). After a final wash, peroxidase was developed by 3-amino-9-ethylcarbazole which gives a red color, and alkaline phosphatase was developed by Fast blue substrate which gives a blue color (27).

mRNA was extracted from 25 × 10³ B cells (11). cDNA was obtained by reverse transcription using the Superscript Reverse Transcriptase Kit (GIBCO BRL, Gaithersburg, MD), with oligo dT_{12 –18} primers (Pharmacia, Upsalla, Sweden). Full length VH5 transcripts were amplified with L-VH5 primer (5′CCCGAATTCATGGGGTCAACCGCCATCCT3′) with 3′ primer CHμ (TGGGGCGGATGCACTCCC) with Taq polymerase (Perkin-Elmer Corp., Norwalk, CT) using the reaction buffer provided by the manufacturers and a DNA thermal cycler (Perkin-Elmer Corp.) with 35 cycles of 1 min denaturation at 94°C, 2 min of primer annealing at 60°C, and 3 min extension at 72°C. After the last cycle, the reaction mixtures were incubated for 10 min at 72°C to ensure complete extension of all products. The PCR products were cloned in PCR™II vector, using the TA Cloning Kit (Invitrogen, San Diego, CA). Both DNA strands of plasmids extracted from individual bacterial colonies were sequenced on an automated DNA sequencer (Applied Biosystems Inc., Foster City, CA). Sequencing was done with the −21M13 and M13RP primers flanking the plasmid cloning sites, and with the CHμ primer annealing with the 3′ end of CH1-μ.

Human tonsillar B cells that coexpress sIgD and CD38 have been identified, which represent 5–15% of total human tonsillar B cells (Fig. 1,A). Three-color flow cytometry shows that sIgD⁺CD38⁺ B cells express low levels of the naive B cell markers CD23, CD44, and IgM, while they express high levels of the germinal center markers CD71, CD10, and CD77 (Fig. 1,B). Unlike the sIgD⁺CD38− naive B cells, they express low levels of Bcl-2, high levels of Fas, and many of them express the proliferation associated nuclear antigen Ki67 (Fig. 1,B). The sIgD⁺CD38⁺ B cells were sorted into a sIgM⁺ subset (3–9% of total tonsillar B cells) and a sIgM⁻ subset (2–6% of total tonsillar B cells) (Fig. 1 C). Our recent study has shown that sIgM⁻IgD⁺CD38⁺ B cells represent a peculiar population of highly mutated GC centroblasts (28). We have now further characterized the sIgM⁺IgD⁺CD38⁺ B cell subset which displays a phenotype intermediate between follicular mantle naive B cells and GC B cells.

Unlike small dense sIgM⁺sIgD⁺CD38⁻ naive B cells, sIgM⁺IgD⁺CD38⁺ B cells are large- or medium-sized lymphocytes, similar to IgD⁻CD38⁺ GC B cells (12, 23). Large IgM⁺IgD⁺CD38⁺ B cells, which account for 20– 30% of the IgM⁺IgD⁺CD38⁺ B cells, express the proliferation associated nuclear antigen Ki67, while medium-sized IgM⁺IgD⁺CD38⁺ B cells are Ki67⁻ (Fig. 2,B). Like IgD⁻CD38⁺ GC B cells, >90% of IgM⁺IgD⁺CD38⁺ B cells display apoptotic figures after 16 h of culture (Fig. 2,C and Table 2). The survival and proliferation of IgM⁺ IgD⁺CD38⁺ B cells in vitro depend on the presence of CD40-ligand and T cell cytokines such as IL-2, -4, and -10 (Table 3). The level of DNA synthesis observed under identical culture conditions is always lower in IgM⁺IgD⁺CD38⁺ B cells than in IgD⁺CD38⁻ naive B cells, but higher than in IgD⁻CD38⁺ GC B cells (Table 3).

VH5-μ sequence analysis has been successfully used to trace the progression of human tonsillar B cell subsets from the virgin to the memory compartment (11). Therefore, 32 VH5-μ sequences from sIgM⁺IgD⁺ CD38⁺ B cells were compared to 30 VH5-μ sequences from sIgM⁺IgD⁺CD38⁻ naive B cells of three tonsil samples (Fig. 3, A and B). In agreement with our previous reports, all 30 sequences from naive B cells contain 0–2 mutations per sequence, with an overall mutation frequency of 2 × 10⁻³ bp (20 mutations/9,600 bp) that is barely distinguishable from the PCR Taq-error rate (1 × 10⁻³ bp). Within 32 sequences from sIgM⁺IgD⁺CD38⁺ B cells, while 17 might be considered germline (0–2 mutations/sequence), 15 have accumulated from 3–13 mutations, with an average mutation frequency of 20 × 10⁻³ bp (88 mutations/4,480 bp). The replacement mutations in 12 less mutated sequences (3.2, 3.4, 3.10, 3.14, 3b.5, 3b.6, 3b.9, 3b.13, 3c.5, 3c.7, 3c.9, and 3c.12) are randomly distributed within the CDRs and FWs, an indication of the absence of antigen-driven selection. In contrast, sequences 3.11, 3b.7, and 3c.6, which have accumulated 10–13 mutations, showed replacement mutations mainly focused in the CDRs, suggesting that these cells have undergone antigen-driven selection. Unlike the VH5-γ sequences from sIgD⁻CD38⁺ GC B cells and VH5-δ sequences from sIgM⁻IgD⁺CD38⁺ reported previously (11, 28), no clonal relatedness could be found among the VH5-μ transcripts from tonsillar sIgM⁺ IgD⁺CD38⁺ B cells. To establish whether the mediumsized nonproliferating cells from this population are those with unmutated V regions, sIgM⁺IgD⁺CD38⁺ B cells were isolated from a fourth tonsil and fractionated according to their size. The subset of medium-sized sIgM⁺IgD⁺CD38⁺ B cells was enriched for unmutated B cells (Fig. 3, C and D) as five out of nine sequences from the medium-sized B cells displayed less than two mutations and were therefore considered as nonmutated. In contrast, only one out of nine sequences from the large B cells was unmutated. In those two subsets, the most mutated sequences show features of antigen-driven selection.

It has been previously reported that a small proportion of human GC contain IgD⁺ B cells (29, 30). Indeed, by double immunohistological staining, we have recently confirmed these observations and further shown that IgD⁺ B cells within GC display CD38, the majority of them expressing the proliferation associated nuclear antigen Ki67 (28). To distinguish IgM⁺IgD⁺CD38⁺ GC founder cells from highly mutated IgM⁻IgD⁺CD38⁺ GC B cells in situ (26), double staining with anti-IgM and anti-IgD were performed on tonsil sections. While numerous large IgM⁻IgD⁺ B blasts were found to be packed within occasional GC dark zones (26), only scattered IgM⁺IgD⁺ B cells were found throughout a few GCs (Fig. 4).

During T cell–dependent primary immune responses, antigen-specific naive B lymphocytes undergo either plasma cell reaction in the T cell zones or enter into the B cell follicles to initiate GC reaction. One of the questions regarding GC development has been the identity of GC precursor cells (16, 31, 32). Progress has been made by analyzing the capacity of B cell subsets to form GC in the host animals after adoptive cell transfer and immunization. SCID mice immunized after repopulation with carrier-primed CD4⁺ T cells and J11D^low splenic B cells developed many GC in their spleens. In contrast, when J11D^high cells or CD5⁺ peritoneal B cells were transferred, few, if any, GC were generated (33). However, thoracic duct B cell transfer experiments in rats have generated contradictory conclusions as to whether GC precursor cells are sIgD⁺ or sIgD⁻ (34, 35). The present identification of tonsillar sIgM⁺IgD⁺ CD38⁺ B cells represents a parallel attempt in the human system to characterize GC founder cells.

The formal identification of GC founder cells would require the demonstration of a precursor–progeny relationship between such cells and GC mutated B cells. However, we could not find clonally related unmutated sIgM⁺sIgD⁺ CD38⁺ B cells and mutated sIgD⁻CD38⁺ B cells from the same tonsil within a limited number of sequences (data not shown). In agreement with our results, when GC B cells were individually picked from the same GC on human lymph node tissue sections, no clonal relatedness could be observed between the germline VH/Vκ sequences and the mutated sequences (8). It suggests that it might be difficult to directly illustrate such a precursor-progeny relationship in humans, since kinetic analysis is not easy to perform. However, several important features suggest that mediumsized nonproliferating sIgM⁺IgD⁺CD38⁺ B cells might be enriched for GC founder cells. First, these sIgM⁺IgD⁺ B cells coexpress, albeit at a reduced level, markers of naive B cells such as CD23 and CD44, as well as markers of bona fide GC B cells such as CD10, CD38, and CD71. As sIgD⁺ CD38⁺ B cells are found within GC exclusively, their intermediate phenotype is suggestive of a transitional stage of maturation between follicular mantle naive B cells and early GC B cells. Second, 29 out of 32 IgVH clonally independent sequences are either in germline configurations (17 sequences) or are poorly mutated (12 sequences) and show no clear evidence for antigen-driven selection. Moreover, when sorted according to their size, >50% of the medium-sized sIgM⁺IgD⁺CD38⁺ B cells are germline or low mutated, a characteristic expected for early GC B cells. Third, in contrast to typical GC centroblasts which give rise, after clonal expansion, to centrocytes, medium-sized sIgM⁺IgD⁺CD38⁺ B cells are not actively dividing as they are Ki67⁻. This suggests that these cells have not yet undergone clonal expansion, a conclusion that is supported by the lack of clonal relatedness between the IgVH sequences of sIgM⁺IgD⁺CD38⁺ B cells isolated from the same tonsil. In conclusion, medium-sized, nonproliferating sIgM⁺IgD⁺ CD38⁺ B cells are enriched for early GC B cells.

An important finding of the present study is the early triggering of apoptosis program in sIgM⁺IgD⁺CD38⁺ GC founder cells, which happens before the onset of somatic mutation. This early propensity to undergo apoptosis may provide a mechanism for selection of cells bearing high affinity unmutated antigen receptors within GC. Consequently, only the cells bearing antigen receptors with better affinity will have the opportunity to undergo affinity maturation. This hypothesis is supported by the finding that many PNA⁺ mouse GC B cells, appearing during the first 10 d of primary response to NP (4-hydroxy-3-nitrophenyl) contained selected IgV genes in germ line or low mutated configurations (7, 36). In addition, this early propensity to undergo apoptosis may also allow the immediate selection of mutating cells (37). Consequently, the B cells entering into GC have to mutate rapidly and efficiently in order to survive.

The demonstration of sIgD on naive B cell derived GC founder cells supports the hypothesis of Thorbecke et al. (16) and Roes and Rajewsky (38) suggesting an auxiliary receptor function for sIgD in antigen-mediated recruitment of B cells into GC reaction and memory B cell formation. This hypothesis was based on the following observations. (a) sIgD are expressed on naive B cells 10 times more efficiently than slgM (39) and bind antigen better than sIgM due to their structural flexibility (40). These two characteristics of sIgD may help naive B cells to colonize the network of follicular dendritic cells when antigen become limiting during GC development (38). (b) A subpopulation of helper T cells bearing receptors for sIgD was found to play an important role in enhancing T cell–dependent humoral immune responses (41, 42). (c) In vivo injection of anti– mouse IgD dramatically increases, within 6 d, the volume of GC in the spleen (43). (d) IgD-knock out mice displayed a delayed affinity maturation of T cell–dependent antibody responses (38) and a higher sensitivity to tolerance induction (44).

The present demonstration of sIgD on GC founder cells also sets a limit to the concept of sIgD being an absolute marker for naive resting B cells. This concept was essentially based on the rapid loss of sIgD on naive B cells after activation in vitro and in vivo (45, 46), and the apparent lack of IgD expression in GC (21). However, consistent with our present identification of sIgD⁺ proliferating GC founder cells in vivo, sIgD⁺ proliferating B cells were observed in long-term culture of CD40 activated human naive B cells (47). Furthermore, the three VH-5μ sequences with more than 10 mutations showed evidence for antigendriven selection. These three mutated sIgM⁺IgD⁺CD38⁺ B cells may either represent GC founder cells derived from recirculating sIgM⁺IgD⁺ memory B cells or correspond to GC centrocytes currently differentiating into sIgM⁺IgD⁺ memory B cells. Therefore, our present findings further support the existence of sIgD⁺ memory B cells, demonstrated in mouse adoptive transfer experiments a decade ago (48, 49).

In conclusion, human GC founder cells have been isolated as medium-sized, nonproliferating sIgM⁺IgD⁺CD38⁺ B cells. Further study of this subpopulation of B cells should lead to better understanding of the early events governing GC development.

We thank Drs. F. Brière and P. Garrone for critical reading of the manuscript, Dr. J. Chiller for support, Mrs. Bonnet-Arnaud and Mrs. Vatan for editorial assistance, and Ms. I. Durand for cell sorting.