V(D)J Recombinase Activity in a Subset of Germinal Center B Lymphocytes

See allHide authors and affiliations

Science  10 Oct 1997:
Vol. 278, Issue 5336, pp. 301-305
DOI: 10.1126/science.278.5336.301


Reexpression of the V(D)J recombinase-activating genesRAG1 and RAG2 in germinal center B cells creates the potential for immunoglobulin gene rearrangement and the generation of new antigen receptor specificities. Intermediate products of V(D)J recombination are abundant in a subset of germinal center B cells, demonstrating that the κ immunoglobulin light-chain locus becomes a substrate for renewed V(D)J recombinase activity. This recombinationally active cell compartment contains many heavy-chain VDJ rearrangements that encode low-affinity or nonfunctional antibody. In germinal centers, secondary V(D)J recombination may be induced by diminished binding to antigen ligands, thereby limiting abrupt changes in receptor specificity to B cells that are usually eliminated from the germinal center reaction. This restriction preserves efficient antigen-driven selection in germinal centers while allowing for saltations in the somatic evolution of B cells.

The primary antibody repertoire is created during the pre-B and immature stages of B lymphogenesis by V-to-J recombination in the κ and λ light-chain (L-chain) loci (1) and by tolerance mechanisms that either purge newly generated, autoreactive B cells by apoptosis (2) or render them harmless by a process of secondary V(D)J rearrangements known as receptor editing (3). Remarkably, during the immune response known as the germinal center (GC) reaction, some B cells begin to transcribe λ5, a component of the pre-B cell receptor complex (4, 5), express the RAG1 and RAG2 V(D)J recombinase proteins (5), and become highly sensitive to receptor-induced apoptosis (6). This broad recapitulation of lymphocyte development and the availability of V(D)J recombinase suggests that at least some GC B cells might also undergo secondary immunoglobulin (Ig) gene rearrangements.

GCs are established by the focal proliferation of mature, antigen-reactive lymphocytes in the B cell zones of secondary lymphoid tissues (7). They are crucial for immunological memory in the B cell compartment and are primary sites for the V(D)J hypermutation and cellular selection necessary for affinity maturation of antibody responses (8). B cells that enter nascent GCs bear IgM and IgD on their surface and express the low-affinity IgE Fc receptor, CD23 (CD23+). With time, GC B cells cease to express membrane IgD and CD23 and acquire a characteristic phenotype that includes avid binding of peanut agglutinin and the GL-7 antibody (PNAhi, GL-7+), upregulation of the Fas death trigger (CD95+), and increased expression of CD24 (CD24hi) (7, 9). This core phenotype remains stable during the primary GC reaction.

Although the majority of splenic GC B cells (GL-7+, CD95+, CD24hi) express amounts of B220 (CD45R) typical of mature B lymphocytes (B220hi), a population of GC B lymphocytes that expresses B220 in amounts similar to pre-B cells (B220lo) appears by day 10 to 12 after immunization (10) and expands, eventually constituting as much as half (25 to 53%) of all GL-7+ B lymphocytes (Fig.1). Proliferation, as measured by incorporation of the nucleotide analog, 5-bromo-2′-deoxyuridine (BrdU), is equivalent in B220lo and B220hi GC cells and consistent with the rapid proliferation of GC B lymphocytes (7,11).

Figure 1

Flow cytometric analysis of splenic germinal center (GC) B cells from C57BL/6 mice at 8 and 16 days after immunization with CGG. Dissociated splenocytes were stained with anti-B220–FITC and GL-7–biotin followed by streptavidin-PE, as described (5). Two subpopulations, B220lo and B220hi, of GC cells are present in late phase (day 16) GCs but not at early time points (day 8) after immunization.

The delayed appearance of the B220lopopulation of GC B cells is similar to the kinetics of local RAG expression (5). To compare the distribution of recombinase and λ5 expression in B220lo and B220hi GC cell populations, we purified 1 × 104 follicular (GL-7, B220+), GC (GL-7+, B220+), and B220lo and B220hi GC (GL-7+) cells from the splenocytes of immunized C57BL/6 mice by fluorescence-activated cell sorting (FACS) (12). A reverse transcriptase–dependent polymerase chain reaction (RT-PCR) was then used to determine the presence of RAG1, RAG2, λ5, and hypoxanthine-guanine phosphoribosyl transferase (HPRT) mRNA in aliquots of each sorted population (13). Whereas equivalent amounts of HPRT mRNA were present in all cell samples (Fig.2D), recombinase and λ5 message (Fig.2, A to C) were present in B220+ GC cells but could not be detected in follicular B lymphocytes. Among GC cells, RAG and λ5 transcripts were largely confined to the B220locompartment. Transcription of the RAG andλ5 genes is correlated with the B220loGC cell phenotype.

Figure 2

RT-PCR assays to detect the expression ofRAG1, RAG2, and λ5 in GC B cell populations. Splenic cells recovered from immunized C57BL/6 mice at day 16 were stained with anti-B220–FITC and GL-7–biotin (11). B220lo/GL-7+, B220hi/GL-7+, and B220+/GL-7 B cells were sorted into TRIzol, and RT-PCR was performed as described (12) to determine the presence of mRNA for RAG1 (A), RAG2 (B), and λ5 (C) in different cell populations. HPRT (D) was used as a control to ensure the amounts of mRNA recovered in each sample were equivalent. In addition, RNA was purified from bone marrow (BM) of young adult C57BL/6 mice and used as a control. After 35 amplification cycles, PCR products were electrophoresed on agarose gels and detected by staining with ethidium bromide. Lane 1, molecular size markers; lane 2, control (no reverse transcriptase, 1 × 104 BM cells); lane 3, 1 × 104 BM cells; lane 4, 1 × 104GL-7B220+ B cells; lane 5, 1 × 104 B220hi GL-7+ GC cells; lane 6, 1 × 104 B220lo GL-7+ GC cells; lane 7, molecular size markers. The open circle (O) indicates a molecular size of 600 bp.

To determine if V(D)J recombination was active in GC B cells, we designed a locus-specific, ligation-mediated PCR (LM-PCR) assay to detect double-stranded (ds) recombination signal sequence (RSS) breaks in the Ig κ locus (Fig. 3A). These blunt, 5′-phosphorylated, ds breaks at RSSs are intermediates of V(D)J recombination and depend on the unique enzymatic activity of RAG1 and RAG2 (14). Purified DNA samples were ligated to an unphosphorylated linker oligonucleotide (BW) containing one blunt and one overhanging end. RSS breaks 3′ to a prior V-to-Jκ rearrangement were then amplified by PCR, using a linker-specific primer (BW-H) and a primer, VκB or VκS, that hybridizes to most Vκ gene segments (15, 16). A fraction of this initial PCR product was then amplified in a second round of PCR using BW-H and primers annealing 5′ of Jκ2, Jκ4, or Jκ5. Because the distance from unrearranged Vκ gene segments to the Jκ cluster is too great for amplification, this assay preferentially detects RSS breaks associated with replacement rearrangements.

Figure 3

LM-PCR assay to detect secondary V(D)J recombination in GC B cells. (A) Diagram of the LM-PCR assay designed to detect DNA breaks associated with replacement rearrangements at the Ig κ locus. The κ locus is shown with a primary Vκ-Jκ1 rearrangement and, as an example, an attempted secondary rearrangement to Jκ2. The PCR primers used to detect RSS breaks at Jκ2, Jκ4, and Jκ5 (Jκ3 is a nonfunctional pseudogene) are also depicted (Jκ910F, Jκ1474F, Jκ1847F, respectively). Hypothetical broken DNA molecules corresponding to signal ends—with the nonamers (N) and heptamers (H) depicted as hatched boxes—are shown alongside the BW linker, an unphosphorylated blunt-ended 25-mer with a 14-nucleotide 5′ overhang. The BW linker ligates to RAG-dependent, double-stranded signal sequence breaks, which are blunt and 5′-phosphorylated (15). BW linker-ligated DNA is subjected to two rounds of PCR amplification, as described (18). PCR products were electrophoresed on 2% agarose gels, blot-transferred to nylon membranes, probed with a radiolabeled oligonucleotide (Jκ1-2, Jκ1568F, or Jκ2000F for breaks at Jκ2, Jκ4, or Jκ5, respectively) internal to the PCR primers as shown, and analyzed on a phosphorimager. (B) RSS breaks in the Ig κ locus are present in DNA from sorted GC B cells. Seven-month-old C57BL/6 mice were immunized with CGG in alum. At day 16, splenocytes isolated from these mice were sorted to collect GL-7+ GC B cells, GL-7 follicular B cells, and B220loand B220hi GC populations. Two-month-old C57BL/6 mice were also used to obtain CD43+/GL-7 plasmacytes or CD43/GL-7+ GC cells at day 8 after immunization. Analyses after sorting revealed that the purity of the GC cells was >90%, whereas the follicular cells were >95% pure. DNA from sorted populations was purified, dissolved in a small volume of H2O, then ligated overnight to the BW linker (18). An aliquot of each sample (∼20 ng) was tested in a control PCR (18) using primers for the nonrearranging CD14 locus, to ensure that a similar amount of DNA was present in each sample (upper panel). LM-PCR using primers designed to detect RSS breaks upstream of Jκ2, Jκ4, and Jκ5 was performed in the absence of DNA template (lane 1), or with ∼20 ng of DNA isolated from the following cells: NIH 3T3 fibroblasts (lane 2); thymocytes (lane 3); day-3 newborn spleen (lane 4); unimmunized adult spleen (lane 5); day-8 sorted CD43GL-7+ GC B cells (lanes 6 and 7), CD43hiGL-7 plasma cells (lane 8), and CD43loGL-7 B cells (lane 9); and day-16 sorted B220+GL-7 follicular B cells (lanes 10 and 11), B220+GL-7+ GC B cells (B220lo and B220hi, lane 12), B220loGL-7+ (lane 13), B220hiGL-7+ (lane 14), and stained but unsorted splenocytes (lanes 15 to 17).

To search for active, secondary Ig κ rearrangement in GC B cells, we immunized mice with chicken γ-globulin (CGG) (10). At 8 and 16 days after immunization, specific B cell populations were purified by cell sorting (12), and DNA from sorted lymphocytes and control cells was subjected to the LM-PCR assay (17). Although amplification of a nonrearranging control gene, CD14, was comparable in all samples (Fig. 3B, upper panel) and dsDNA breaks at Jκ RSSs were absent in 3T3 fibroblasts and thymocytes, distinctive patterns of RSS breaks appeared in the sorted cell populations (Fig. 3B, lower panel). Breaks were abundant in the splenocytes of 3-day-old mice because of the high frequency of immature B cells present in the neonatal spleen (16), but were undetectable in spleens from unimmunized aged mice. In immunized, young adult mice, we found Jκ2 RSS breaks in both resting and antigen-activated cell populations, but Jκ4 and Jκ5 RSS breaks were present only in cell samples containing activated lymphocytes. Jκ2 RSS breaks in resting cells probably derive from maturing B cells (IgDlo, CD23, CD24hi) present in splenic T cell areas (18). In contrast, Jκ4 or Jκ5 RSS breaks were present in a single sample of GC (CD43GL-7+) B cells recovered 8 days after immunization and in all GL-7+ GC cells taken 16 days after immunization. These Jκ4 and Jκ5 RSS breaks were present largely, if not exclusively, within B220lo GC B cells. Hence, the GC B cells that contain abundant amounts of RAG1 and RAG2 mRNA (Fig. 2) also contain the intermediate DNA products of V(D)J recombination. Few or no RSS breaks were detected in B220hiGC cells or CD43hi plasmacytes, even though Jκ2 RSS breaks were present in unactivated, follicular B cells (B220+, GL7). The dearth of RSS breaks in B220hi GC cells and plasmacytes is most likely the result of antigen-driven cellular proliferation, because RSS breaks are repaired upon reentry into the cell cycle (19). Thus, RSS breaks in the rapidly proliferating B220lo GC population indicates ongoing V(D)J recombination. Amplification of RSS breaks (17) in decreasing amounts of DNA from B-lineage (CD19+) bone marrow cells, B220lo and B220hi GC cells, and the temperature-sensitive pre-B cell line 103-bcl2 (20) indicated that replacement rearrangements are abundant in the B220lo GC compartment. Secondary Vκ-to-Jκ2 rearrangements could be detected in as little as 0.8 ng of DNA from bone marrow and B220loGC cells, whereas 5- to 25-fold more template was required to detect RSS breaks in induced 103-bcl2 cells and B220hi GC cells, respectively.

To determine if B220lo and B220hi GC cells express comparable antigen receptors, mice were immunized with CGG substituted with the (4-hydroxy-3-nitrophenyl)acetyl (NP) hapten (10). Initially, NP activates B cells bearing the λ1 L-chain and heavy-chains (H-chains) encoded by the V23, CH10, C1H4, or V186.2 VH gene segments (21, 22); later, the response becomes dominated by cells that express the higher affinity V186.2 rearrangement (22). NP-reactive B cells that contain VH V186.2 rearrangements with a tyrosine-rich motif (YYYGS) (23) in the third complementary-determining region (CDR3), and mutations that yield tryptophan → leucine (W→L) replacements at position 33 strongly bind the NP hapten and are considered highly selected (24).

Sequence analysis of randomly chosen VDJ fragments amplified (25) from B220lo (n = 20) and B220hi (n = 20) GC cells showed that these cell populations express very different collections of antigen receptors (Table 1). The B220hi population is enriched for VH gene rearrangements and mutations consistent with avid NP-binding (26). In contrast, VDJ fragments amplified from the B220lo population of GC cells exhibited few characteristics of high-affinity, NP-specific antibodies. Nearly one-third (30%) of VDJ rearrangements from B220lo GC cells contained VH gene segments (V23, C1H4, CH10, 24.8, and 593.3) that are common in early, NP-reactive GCs but encode antibodies with low affinities for this hapten (22, 26). These data suggest that although B220hi GC cells strongly bind antigen, the B220lo population consists of lower affinity B lymphocytes that no longer compete effectively for antigen (7). This conclusion was tested by transfecting 14 productive and representative VDJ rearrangements from B220hi (n = 8) and B220lo (n = 6) GC cells into the J558L cell line (27). The association constant (K a) of each transfectoma antibody for the NP hapten was then measured by fluorescence quenching (26). As expected, the average K a (2.4 × 106 M 1) for the B220hi group was significantly greater than that of B220lo cells (8.0 × 104M 1).

Table 1

Comparison of VDJ rearrangements from B220hi and B220lo GC (GL-7+) cells. B220hi and B220lo GL-7+ cells were isolated by FACS as described (11).

View this table:

Germinal centers support a stage in B cell development where renewed V(D)J recombinase expression in B220lo cells (Fig. 1) drives replacement rearrangements in the Ig L-chain loci (Fig. 3). Secondary V(D)J rearrangement and revision of receptor specificity also occurs in some autoreactive, immature B cells in the bone marrow (3). Although GC B cells express many properties of pre-B and immature B cells, the Vκ-to-Jκrecombination we observed is not ectopic lymphogenesis. GCs were founded by antigen-reactive B cells bearing high amounts of IgD and CD23 on their surface (9), and mutated H-chain (Table 1) and λ1 L-chain gene rearrangements were present in both B220hi and B220lo GC cells. In addition, the signals that drive secondary Ig gene rearrangements in bone marrow and GC B cells appear to differ. Receptor editing in immature B cells is believed to be in response to self-reactivity (3). In GCs, experimental models of autoreactivity induce rapid apoptosis (6) without superinduction of RAG1 protein (28). Instead, patterns of VH gene segment usage, the dearth of affinity-enhancing mutations in B220lo GC cells (Table 1), and affinity measurements of transfectant antibodies imply that receptor revision in GCs is induced by diminished or lost antigen-binding. V(D)J recombination in the B220losubset of GC B cells limits L-chain replacement and the consequent saltations in receptor specificity to a cell compartment that is normally eliminated from GCs (7). The improbable genetic experiment of receptor revision is confined to B cells that are destined for programmed cell death.

However improbable, we believe that replacement rearrangements in GC B cells must confer significant benefit to immune function, because recombinase activity carries appreciable risks, including malignant transformation by illegitimate recombination (29). Replacement rearrangements may rescue failing GC B cells by revising low-affinity receptors or those debilitated by mutation. Indeed, pairings of H- and L-chains consistent with receptor revision in GCs have been observed in the response of BALB/c mice to the 2-phenyl oxazolone hapten (30). More speculatively, saltations in receptor specificity may provide new avenues for clonal evolution, leading to affinities that could not be achieved by mutation alone.

In response to exogenous antigen, the GC microenvironment allows B cells to reactivate the fundamental event of lymphocyte development, V(D)J recombination. Immunoglobulin gene rearrangements in GC B cells that restore or enhance antigen-binding would be analogous to those in immature B lymphocytes that abolish autoreactivity (3): in each case, the recombinase machinery is reactivated to preserve cells that would otherwise be lost to the immune response.

  • * These authors contributed equally to this work.

  • To whom correspondence should be addressed. E-mail: gkelsoe{at}


Stay Connected to Science

Navigate This Article