Relaxed Negative Selection in Germinal Centers and Impaired Affinity Maturation in bcl-x_L Transgenic Mice

The succinic anhydride ester of NP was reacted with chicken γ-globulin (CG; Sigma Chemical Co.) or BSA (U.S. Biochemical Corp.) as described 32. The coupling ratio of each conjugate was determined spectrophotometrically. Antibodies specific for IgM^b (AF6-78) and mouse λ1 L chain (Ls136) were purified over protein G–Sepharose (Amersham Pharmacia Biotech) from culture supernatants. Horseradish peroxidase (HRP)-conjugated goat anti–mouse IgG₁ and biotinylated anti-IgD antibodies were purchased from Southern Biotechnology Associates. Anti-FcγRI/RII (2.4G2), FITC-labeled GL-7, PE-conjugated anti-B220, biotinylated anti–Mac-1, –Gr-1, -Thy1.2, -CD4, -CD8, and -Ter119, and PE-conjugated anti-CD138 (syndecan) antibodies were purchased from PharMingen. Anti–Bcl-x and anti–Bcl-2 mAbs were purchased from Transduction Laboratories.

bcl-x_L transgenic mice were generated as described previously 24 and backcrossed with the C57BL/6 strain (The Jackson Laboratory). Igh^b progeny were used in all experiments. bcl-x_L transgenic or transgene-negative littermate control mice were immunized intraperitoneally with 50 μg of an NP₂₀-CG conjugate precipitated in alum.

The frequencies of NP-specific AFCs in spleen and BM were estimated by enzyme-linked immunospot (ELISPOT) assay using two different coupling ratios of NP-BSA 11. In brief, splenocytes (10⁵ cells/well) or BM cells (5 × 10⁵ cells/well) were incubated on nitrocellulose filters coated with NP₂-BSA, NP₂₃-BSA, or BSA alone at 37°C, 5% CO₂ for 2 h. After washing, filters were stained with HRP-conjugated anti-IgG₁ antibodies, and HRP activities were visualized using 3-aminoethyl carbazole. The frequencies of high-affinity and total AFCs were determined from NP₂-BSA– and NP₂₃-BSA–coated filters after background on BSA-coated filters was subtracted.

IgG₁ specific for the NP hapten was detected by ELISA using two different coupling ratios of NP-BSA as described 11. In brief, serially diluted sera were added to plates coated with NP₂-BSA or NP₂₃-BSA and incubated at 4°C overnight. After washing, HRP-conjugated goat anti–mouse IgG₁ was added, and HRP activity was visualized using a TMB peroxidase substrate kit (Bio-Rad Laboratories). The concentrations of anti-NP IgG₁ antibodies were estimated by comparison with standard curves created from the H33Lγ1/λ1 control antibody on each plate 12. To estimate the affinity of NP-binding antibody in the sera, ratios of NP₂-binding antibody to NP₂₃-binding antibody were calculated.

All histological procedures were conducted as described previously 32. The number of λ1⁺ GCs was determined by staining sections with biotinylated Ls136, followed by alkaline phosphatase–conjugated streptavidin (Southern Biotechnology Associates) and HRP-conjugated peanut agglutinin (PNA). Apoptotic cell death in GCs was estimated by terminal deoxynucleotidyl transferase–mediated dUTP-biotin nick end labeling (TUNEL) as described 17. TUNEL⁺ cells in GCs were counted at 200× magnification by systematic scanning. Cell proliferation in GCs was determined by the incorporation of 2-bromodeoxyuridine (BrdU) as described 33. In brief, 10 d after immunization, bcl-x_L transgenic and control mice were given 1.0 mg BrdU by intraperitoneal injection; 2 h later, the mice were killed and their spleens prepared for histology. Proliferation indices were determined by microscopic inspection as the fraction (%) of PNA-binding (PNA⁺) cells that exhibited nuclear BrdU incorporation.

λ1⁺ GC cells were microdissected from day 12 spleen sections of bcl-x_L transgenic mice and control mice. VDJ DNA was amplified by PCR and cloned into Bluescript plasmid 34. The frequency of VDJ genes using the V186.2 gene segment was estimated by colony hybridization using oligonucleotides specific for V186.2 and one specific for the framework 3 region of the mouse V_H V186.2 and V3 subfamilies of the J558 group 32,35. Plasmid DNA was extracted from ≤3 bacterial colonies from each GC, and V_H gene sequences were determined by automated sequencing.

Single cell suspensions of splenocytes and BM cells were prepared as described 11. Cells were then washed in PBS (pH 7.4) containing 2% FCS and 0.08% sodium azide at 4°C for cytometric analysis, or washed with deficient RPMI 1640 (Irvine Scientific) containing 2% FCS for sorting. The enumeration of GC B cells and sorting of BM AFCs were carried out as described 11.

To collect GC B cells, splenocytes pooled from four mice were blocked with anti-FcγRI/RII and then stained with biotinylated anti-IgD, –Mac-I, –Gr-1, -Thy1.2, -CD4, -CD8, and -Ter119 antibodies for 30 min. After three washes, cells were incubated with streptavidin-conjugated microbeads (Miltenyi Biotec) for 15 min. Cells attached to microbeads were depleted by passage through a CS column (Miltenyi Biotec) in a magnetic field based on the manufacturer's protocol. Recovered cells were stained with FITC-labeled GL-7, PE-conjugated anti-B220, Tricolor-conjugated streptavidin (Caltag Laboratories), and 7-amino-actinomycin D (7-AAD). Finally, GL-7⁺B220⁺ cells within the Tricolor⁻7-AAD⁻ fraction were sorted into biotin-deficient RPMI containing 2% FCS using a FACStar^Plus™ (Becton Dickinson). GL-7⁻B220⁺ cells were sorted from the same pooled splenocytes without preenrichment by depletion with magnetic beads. To collect B220⁺ cells and CD3⁺ cells, splenocytes from naive bcl-x_L transgenic and control mice were blocked and stained with FITC-labeled anti-CD3 and PE-conjugated anti-B220 antibody, and then single positive cells were sorted by the same procedure. Sorting routinely yielded populations of >97% purity.

Cells were lysed in a buffer containing 137 mM NaCl, 1 mM MgCl₂, 0.1 mM CaCl₂, 20 mM Tris-HCl (pH 9.0), 1% NP-40, and 10% glycerol. After homogenization, cells were centrifuged and supernatants were recovered. The cell lysates were loaded (10 μg protein) onto 12% SDS polyacrylamide gel and resolved by electrophoresis. The proteins were then transferred by electrophoresis onto polyvinylidene fluoride membranes. After blocking with TBS (Tris-buffered saline) containing 5% nonfat dry milk, the membranes were incubated with 1:500 diluted anti–Bcl-x or anti–Bcl-2 mAbs, then washed five times with TBS containing 0.1% Tween-20. After incubation with 1:20,000 diluted HRP-conjugated goat anti–mouse Ig antibodies (Amersham Pharmacia Biotech), the reaction was developed by enhanced chemiluminescence using the ECL kit (Amersham Pharmacia Biotech) and detected by exposure to X-ray film.

Spleens of mice carrying the bcl-x_L transgene are ∼50% larger and contain 30% more mononuclear cells than those of nontransgenic littermates. This increased cellularity is due to a near doubling in the number of mature IgM⁺ B220⁺ cells (Fig. 1 A). Flow cytometric analyses of splenic B cells from transgenic mice revealed that expression of IgM, IgD, CD19, CD21, CD22, CD23, and CD24 was identical to that of control littermates (data not shown). Despite the increased numbers of peripheral B lymphocytes, transgenic animals displayed normal levels of serum IgM (1,242 ± 351 vs. 1,295 ± 379 μg/ml) and IgG (1,341 ± 101 vs. 1,777 ± 379 μg/ml) as measured by specific ELISA. As expected 24, the thymic and peripheral T cell compartments of transgenic mice were normal in size and cellular composition (Fig. 1 A, and data not shown).

Initial assays were performed to assess the ability of transgene-bearing B cells to survive in culture medium containing little FCS. Purified splenic B cells from transgenic mice showed a significant survival advantage over control cells when cultured in medium containing 1% (Fig. 1 B) or 0.1% serum (not shown), indicating their strong resistance to the effects of serum starvation.

Despite their resistance to serum starvation, transgenic B cells displayed no evidence for increased proliferation in response to CD40 cross-linking or T cell help in vitro (Fig. 1 C). In addition, proliferative responses and antibody production in cultures containing LPS were the same for both transgenic and control splenocytes.

The product of the bcl-x_L transgene carries a short epitope tag at its NH₂ terminus and migrates more slowly in SDS-PAGE gels than endogenous Bcl-x_L, which runs as a doublet at ≈31 and 32 kD. Fig. 2 A illustrates that transgenic Bcl-x_L is expressed almost exclusively in the B220⁺ fraction of splenic lymphocytes. To define transgene expression in B cells participating in an immune response, C57BL/6 mice were immunized with NP-CG, and 12 d later splenic B220⁺ cells were sorted into GL-7⁻ and GL-7⁺ fractions to identify follicular and GC B cells, respectively 31. The expression of endogenous Bcl-2 and of endogenous and transgenic Bcl-x_L in these populations were then compared (Fig. 2 B).

Follicular B (GL-7⁻B220⁺) cells from wild-type mice are positive for Bcl-2 but express little Bcl-x_L. Both proteins are abundant in the follicular population of mice carrying the bcl-x_L transgene. In contrast, GC B cells (GL-7⁺ B220⁺) isolated from both wild-type and transgenic mice abundantly express Bcl-x_L but little or no Bcl-2. This observation is consistent with studies of GCs in humans 18,30. Thus, the reciprocal expression of Bcl-x_L and Bcl-2 observed in pre-B cells holds for GC B cells 24,25. Interestingly, although transgenic Bcl-x_L is strongly expressed in follicular B cells, only modest amounts of tagged Bcl-x_L could be demonstrated in GC B cells. Reverse transcription PCR studies confirm lower steady state levels of transgenic Bcl-x_L message in the GL-7⁺ B cell population (data not shown). This biased expression of transgenic Bcl-x_L may represent distinct Eμ activity in each B cell compartment (the density of mIg on GC B cells is ≤10% of that found on follicular B cells [29]) or downregulation of the transgene's herpes TK promoter in activated cells 24.

To assess the effects of the bcl-x_L transgene on GC development, we compared the GC reaction of transgenic and control mice at day 12 after immunization with NP-CG. This antigen elicits a characteristic hapten-specific antibody that bears a λ1 L chain and an H chain encoded by a canonical VDJ gene rearrangement 32. We identified λ1⁺ GCs in spleen sections by labeling with PNA and anti-λ1 antibody 32 and determined the average number of λ1⁺ GCs per section from groups of transgenic and control mice (Fig. 3 A). Differences in the number or size (data not shown) of GCs were not observed between the groups, nor did the mean frequency of λ1⁺ GCs significantly differ between transgenic (35.4%) and wild-type mice (41.7%). When transgenic mice were immunized with carrier protein alone, the average frequency of λ1⁺ GCs was 7.6%. Thus, frequent λ1⁺ GCs in transgenic mice result from immunization with NP rather than altered λ1 L chain expression.

The frequencies of splenic GC B cells (GL-7⁺B220⁺ fraction) in transgenic and control mice were also determined by flow cytometry. Both groups supported equivalent and typical GC responses (Fig. 3 B): in transgenics, the frequency of GC B cells peaked at an average of 2.46% of splenocytes compared with 2.39% in controls at day 12 after immunization. Proliferative activity in the GC compartments of both transgenics and controls was also equivalent; 10 d after immunization, 21 vs. 24% of PNA⁺ GC cells were labeled by a 2-h pulse of BrdU (not shown).

T cell–dependent antigens induce two distinct populations of AFC, a short-lived splenic population that generates the earliest primary antibody and a long-lived set in the BM that maintains the serum response 10,11. Frequencies of NP-specific, IgG₁ AFCs in the spleen and BM of transgenic and control mice were determined by ELISPOT assay 12, 35, and 69 d after immunization (Fig. 3 C). The kinetics of AFC production were virtually identical in both groups of mice, but splenic AFCs were threefold more abundant in transgenic mice than in controls. This increase may reflect the approximately twofold increase in the number of B cells in the spleens of transgenic mice (Fig. 1 A). Despite their greater numbers, splenic AFCs in transgenic mice were lost at the same rate as in control animals. This rapid decline contrasts with bcl-2 transgenic mice, which support higher numbers and longer-lived splenic AFCs 22. Frequencies and kinetics of specific BM AFCs were indistinguishable between transgenic and control mice (Fig. 3 C).

The expanded splenic AFC pool in transgenic mice resulted in a minor increase in serum antibody titers on day 12, but later levels of antibody did not differ significantly between transgenic and control mice. In both groups, antibody concentrations were at maximal levels on day 12 and then slowly declined to about one third of this peak by day 69 (Fig. 3 D). Thus, overexpression of Bcl-x_L modestly expands recruitment into the splenic AFC pool but does not change cellular recruitment into GCs, entry into the BM AFC pool, or maintenance of long-lasting serum antibody.

GCs contain more apoptotic lymphocytes as determined by TUNEL than other regions of spleen 17. These TUNEL⁺ cells are thought to represent lymphocytes that have been negatively selected during the GC response. We performed TUNEL assays on spleen sections from transgenic and control mice to determine if the small addition of transgenic Bcl-x_L expressed in GC B cells was sufficient to reduce programmed cell death. TUNEL⁺ cells in GCs from both groups were counted by microscopic examination, and the frequency of TUNEL⁺ cells per unit area was calculated. These frequencies were subdivided into 12 categories, and the distribution histogram for each category was plotted (Fig. 4). GCs from bcl-x_L transgenic mice contained fewer TUNEL⁺ cells per unit area (P < 0.01) than those from control mice (Fig. 4). The most common apoptotic index in wild-type animals was 2.0–2.5 TUNEL⁺ cells/unit area but only 1.0–1.5 in the bcl-x_L transgenics. Perhaps more significantly, >20% of GCs in control mice contained >3 TUNEL⁺ cells/unit area, whereas only 5% of GCs in bcl-x_L transgenic animals held 3.0–4.0 apoptotic cells/unit area with no GCs in the 4.5–6.0 categories. Thus, a modest addition of Bcl-x_L in transgene-bearing GC B cells leads to a readily detectable decrease of TUNEL⁺ cells.

Initially, immunization with NP conjugates stimulates a broad population of splenic B cells that bear the λ1 L chain and H chain genes made from the V186.2 and V3 subgroups of the J558 V_H gene family 34. Until day 6–7 of the primary response, many GC B cells express V_H gene segments that closely resemble the canonical V186.2 element but encode lower-affinity NP-binding antibodies (12,34,36,37; and Shimoda, M., and G. Kelsoe, unpublished data). By day 10, the majority of B cells bearing these noncanonical V_H gene rearrangements are replaced by higher-affinity cells bearing V186.2/DFL16.1 rearrangements and a tyrosine-rich junctional motif, YYGS 12,34. Thus, after day 10 the efficiency of affinity-based competition is estimated by the ratio of GC B cells bearing V186.2 versus noncanonical V_H rearrangements. Cells were microdissected from a total of 27 λ1⁺ GCs from 4 transgenic and 4 control mice 12 d after immunization with NP-CG. VDJ rearrangements from both groups of mice were amplified by PCR and cloned into bacteria, and >600 bacterial colonies were then subjected to hybridization using V186.2-specific and subgroup-specific primers. About 80% (64/81) of VDJ rearrangements recovered from control mice selectively hybridized to the V186.2 V_H gene segment, consistent with previous studies 12,34, whereas fewer than half (261/554) of the VDJ fragments amplified from bcl-x_L transgenic mice exhibited preference for V186.2.

These hybridization data were confirmed by sequencing VDJ inserts from representative bacterial colonies (Table). From normal C57BL/6 control mice, 79% (11/14) of sequenced VDJ rearrangements contained the V186.2 gene segment, confirming our hybridization analysis and prior sequence studies 33. In contrast, only 47% (17/36) of sequenced VDJ fragments from bcl-x_L transgenics carried the canonical V186.2 element (Table). Noncanonical rearrangements from both control and transgenic mice contained other V_H genes from the V186.2 and V3 subfamilies 35 commonly recovered in early primary anti-NP responses 12,34.

The use of noncanonical VDJ rearrangements by immunized transgenic mice was not due to altered usage of VH gene segments in naive λ1⁺ B cells. We recovered splenic λ1⁺B220⁺ cells from unimmunized, transgenic, and wild-type mice by fluorescence-activated cell sorting, amplified their VDJ rearrangements with the PCR primers used to study GC populations, and determined the ratios of VDJ rearrangements containing V186.2 versus related V_H genes. There was no significant difference in the percentage of V186.2 genes used by naive Bcl-x_L transgenic mice (14%; 12/88) and naive controls (16%; 12/74). These observations suggest that even slight overexpression of Bcl-x_L in GC B cells leads to lower numbers of TUNEL⁺ GC cells and the persistence of clones bearing noncanonical VDJ rearrangements that commonly encode low-affinity antibodies present early in the primary response to NP.

The abundance of noncanonical VDJ rearrangements was not associated with impaired positive selection. Patterns of mutation in VDJ rearrangements containing the canonical V186.2 gene segment were similar in transgenic and control mice, with no significant difference (P > 0.05) in the ratios of replacement versus silent mutations (R/S ratios) in CDRs (Table). Other characteristics indicative of high-affinity, NP-specific B cells, e.g., the fraction of rearrangements containing DFL16.1 and the YYGS CDR3 motif, were also similar in both groups. Thus, cellular recruitment, V(D)J hypermutation, and positive selection in GCs are unaffected by the bcl-x_L transgene.

Serum antibody is maintained by long-lived BM AFCs that depend on the GC differentiation pathway 10,11,38,39. Affinity maturation of serum antibody and the BM AFC compartment can be monitored by differential binding to ELISA or ELISPOT substrates with sparse (NP₂) or dense (NP₂₃) hapten coatings 10,11. High-affinity antibody from serum and AFCs binds equally well to both hapten densities, whereas low-affinity binding is evident only on the NP₂₃ substrate.

The high-affinity compartment of BM AFCs in wild-type mice rapidly increased between days 12 (30.3%) and 35 (75.6%) of the response, with a more gradual increase up to day 69 (88.4%) (Fig. 5 A). This kinetic is typical of normal responses 11. At day 12, high-affinity AFCs were as common in the BM of transgenic mice (34.3%) as in controls. However, this population expanded more slowly in animals with the bcl-x_L transgene, reaching only 57.5 and 60.6% by days 35 and 69, respectively (Fig. 5 A). Remarkably, at day 69 of the response three transgenic mice had smaller high-affinity AFC compartments than were present at day 35, indicating little or no proliferation/survival advantage for high-affinity cells even when antigen concentrations should be minimal.

The average affinity of NP-specific serum antibody was determined for the same mice by ELISA (Fig. 5 B). In wild-type controls, early (day 12) serum antibody contained little or no high-affinity component; by day 35 roughly half of the serum antibody displayed high-affinity binding, and by day 69 this value increased to >90% (Fig. 5 B). The average affinity of serum antibody in transgenic mice also increased from day 12 to day 69, but again the extent of affinity maturation was only ∼60% of controls. Overexpression of Bcl-x_L led to diminished affinity maturation in both BM AFCs and the serum antibody.

To determine the cause of decreased affinity in the BM AFCs of transgenic mice, we recovered the λ1⁺ BM AFC populations from immunized wild-type (n = 5) and transgenic (n = 5) mice by cell sorting 11. Typically, at day 69 after immunization >50% of sorted cells from both groups of mice secreted IgG₁ antibody specific for NP. Enriched BM AFC populations were subjected to a reverse transcription PCR that preferentially amplifies cDNA representing rearrangements of the V186.2 and V3 subfamilies of V_H gene segments joined to Cγ1 11. Amplified VDJ rearrangements were cloned and sequenced to identify the V_H and D gene segments used and any mutations present. Table summarizes this work and shows that only half (11/21) of the VDJ sequences recovered from bcl-x_L transgenic mice used the V186.2 gene segment. In contrast, nearly all (16/17) VDJ rearrangements from wild-type mice contained the V186.2 gene segment. Thus, the high frequency of B cells bearing noncanonical VDJ rearrangements present in day 12 GCs (47%; Table) was maintained in the day 69 BM AFC population (53%) of mice with the bcl-x_L transgene. The reduced average affinity of BM AFCs in bcl-x_L transgenic mice results from the retention of B cells bearing noncanonical VDJ rearrangements. Interclonal competition in both the GC and BM AFC compartments of transgenic mice is relaxed, even when the amounts of residual antigen are thought to be limiting.

Comparison of the frequency and pattern of mutations in rearranged V186.2 gene segments from transgenic and control mice suggests that intraclonal selection may also be weakened by increased Bcl-x_L expression. The average frequency of mutations in V186.2 V_H gene segments from the BM AFCs of bcl-x_L transgenic mice was about twice as high as that from control mice (Table). B cells in transgenic mice might better survive the GC environment than B cells in control animals, allowing them to accumulate more mutations before entering the long-lived AFC compartment 11. Furthermore, R/S ratios in both CDR1 and CDR2 of these genes were lower in transgenic mice than in controls. Higher R/S ratios in CDRs arise in part as a consequence of antigenic selection. Other characteristics indicative of high-affinity, NP-specific B cells, such as the percentage of genes bearing DFL16.1, the YYGS CDR3 motif, and a Trp→Leu (W→L) mutation at position 33, were also less common in transgenic mice than in controls. Together, these data show that the bcl-x_L transgene enhances the survival of low-affinity GC B cells bearing noncanonical VDJ rearrangements, increases their accumulation of VDJ mutations, and allows low-affinity GC cells to remain in the BM AFC compartment. Persistence of this low-affinity component does not come at the expense of positive selection for higher-affinity clones but by diminished apoptosis in low-affinity B cells.

After day 35, the population of high-affinity BM AFCs grew twice as fast in wild-type mice as in transgenics (Fig. 5) even though both groups supported equivalent numbers of NP-specific, IgG₁ AFCs (Fig. 3). Administration of anti-CD154 antibody to abrogate the GC reaction and reduce the genetic diversity present in the BM AFC compartment 11,40 enhanced this difference. Control mice that received the CD154-specific antibody, MR1, on days 6, 8, and 10 after immunization had a high-affinity BM AFC compartment of 15.1% at day 22 of the response that grew to 55.8% by day 69 (data not shown). Immunized transgenic mice that received MR1 antibody began with a comparable high-affinity AFC compartment, 12.0% at day 22, but this population grew to only 22.0% by day 69. Thus, overexpression of Bcl-x_L also retards affinity maturation outside the GC microenvironment.

In this report, we have demonstrated that a bcl-x_L transgene reduces apoptosis in the GC reaction and impairs affinity maturation by sparing cells normally lost from the primary response. This transgene also enhances the survival of peripheral B cells in response to serum starvation in vitro and rescues developing B lymphocytes with aberrant VDJ rearrangements. These effects represent supplementation of endogenous Bcl-x_L activity; although Bcl-x_L is abundant in the GL-7⁺B220⁺ GC cells of wild-type mice, Bcl-2 is not (Fig. 2 B). Similar observations have been reported for human GC cells where Bcl-x_L rather than Bcl-2 mediates the CD40-dependent survival of centrocytes ex vivo 30.

This result contrasts with that of Bcl-2 overexpression, which does not interfere with affinity maturation 22 but permits the survival of mature autoreactive B cells in the periphery 41. The bcl-x and bcl-2 transgenes also act differently during negative selection in immature B cells, as transgenic Bcl-x_L has the ability to block negative selection and promote developmental maturation, whereas autoreactive cells transgenic for bcl-2 remain arrested in development 42,43. Given the similar reciprocal expression of bcl-2 and bcl-x in GC B cells and pre-B cells, bcl-x may have a distinct role in regulating the survival of B cells undergoing selection via mIg or the pre-B cell antigen receptor (BCR) 24,30. Bcl-x_L becomes abundant in B cells after cross-linking mIg or CD40 25,28, and the fate of GC B cells is controlled by these same signals 40. We speculate that the degree or quality of mIg signaling in low-affinity B cells does not induce Bcl-x_L expression as effectively as in high-affinity cells, and that this deficit leads to apoptosis. That even a slight addition of transgenic Bcl-x_L to the higher levels of the endogenous protein in GC B cells leads to significant effects on cell death and affinity maturation indicates that GC B cells are quite sensitive to small changes in levels of this death antagonist. The fate of lower-affinity GC B cells appears to be determined by a regulatory threshold of Bcl-x_L.

Relaxed negative selection and the retention of low-affinity B cells in transgenic mice did not alter the duration or magnitude of the GC response in bcl-x_L transgenic mice (Fig. 3). At 35 d after immunization, the splenic GC reaction had ended both in transgenic (0.33% GL-7⁺B220⁺ spleen cells) and control (0.37%) animals. This is the earliest time after immunization that the numbers of splenic GL-7⁺ B220⁺ cells return to preimmune levels in normal mice 11. Thus, the GC response appears to be regulated by factors beyond affinity-driven competition and selective apoptosis. The rise and fall of GCs depend on the presence of antigen, sustained cell–cell interactions, and cues for cellular location 40,44,45,46,47. It is not surprising that this important immunological response is controlled by finer means than that afforded by Darwinian competition alone.

Nie et al. 37 have reported that immunization of C57BL/6 mice with complexes of antibody and antigen elicits lower-affinity serum antibody and a genetically diverse GC reaction similar to that we observe in bcl-x_L transgenic mice. These authors hypothesize that immune complexes decorated with C3d efficiently recruit the CD21/CD19/CD81 coreceptor to antigen-binding BCRs to reduce the threshold of B cell activation. Lowered activation thresholds would result in reduced levels of affinity-driven selection. Although Nie and colleagues describe a phenotype similar to that of bcl-x_L transgenic mice, we think it is unlikely that Bcl-x_L reduces selection intensity by enhancing BCR signals. Background levels of IgM and IgG are similar in the serum of transgenic and control mice, and both contain only trace amounts of NP-binding IgG before immunization. Thus, immune complexes are no more likely to form in bcl-x_L transgenics than in wild-type controls after primary immunization. Our data do support the notion that enhanced BCR signals, perhaps mediated by coreceptor recruitment, result in reduced apoptosis; immunization with immune complexes may facilitate Bcl-x_L expression in responding B lymphocytes.

A possible mechanism for continued selection by apoptosis outside of GCs is competition among memory B cells for restimulation 48,49. However, memory B cells are thought to regain Bcl-2 expression, and it would be surprising if their survival depended also on Bcl-x_L 18. Alternatively, selective competition among BM AFCs for antigen might drive sustained affinity maturation. BM AFCs express low levels of mIg 11 and could interact with antigen depots in a Bcl-x_L–dependent fashion. Indeed, plasmacytomas express Bcl-x_L 50, and human plasma cells exhibit high levels of Bcl-x_L but low levels of Bcl-2 51, although it is unclear if these cells represent long- or short-lived AFCs. It will be important to learn how BM AFCs integrate the usually antagonistic processes of differentiation to antibody secretion and cellular longevity so as to maintain protective levels of serum antibody over long time periods.

Our data provide strong evidence of a continuing role for antigen in the maintenance of the long-lived AFC pool. However, Manz et al. 38 have reported that the transfer of BM AFCs into unimmunized recipients reconstitutes long-term serum antibody and conclude that antigen is unnecessary for the survival of these cells. Such experiments are complicated by the possibility of coincidental transfer of residual antigen 6,7,48, but we cannot exclude the possibility that post-GC selection acts on precursors of the long-lived AFC pool. In this case, the characteristic somatic genetic changes observed in BM AFCs (11; Table and Table) would first occur in the precursor population. Such selection would be antigen dependent and affinity driven. Recent work on the longevity and affinity of BM AFCs and serum antibody (our unpublished studies) support the importance of antigen retention and/or BCR signaling in shaping the long-lived AFC population. What remains unchallenged is that affinity maturation of serum antibody continues for months after primary immunization (1,2,11; Fig. 5). Although this progressive increase in affinity could be programmed in the early phase of the response, we suggest that in some way antigen continues to exert selection on the responding B cells.

We gratefully acknowledge the assistance of M. Gendelman, and the additional flow cytometric and BrdU-labeling data provided by Dr. B. Zheng. We thank T.F. Tedder, S. Foster, and M. Davila for help with the manuscript.

This work was supported in part by U.S. Public Health Service grants AI24335, AG10207, AG13789 (to G. Kelsoe), CA63111 (to R. Freund), and AR43805 (to T.W. Behrens).