Osteoprotegerin, a Crucial Regulator of Bone Metabolism, Also Regulates B Cell Development and Function¹

Previously, we characterized a molecule from a human follicular dendritic cell (DC) line called follicular DC-derived receptor-1 (1), which is identical with osteoprotegerin (OPG)³ (2). OPG is a member of an emerging subgroup of the TNFR family that can function as soluble decoy receptors (DcR). Examples of other members of this subgroup are DcR-1, -2, and -3 (3). A major function of these soluble inhibitors is to compete with membrane receptors, thereby inhibiting apoptosis induction by their ligands. For example, the recently described CD95 DcR3 may inhibit Fas ligand- and LIGHT-induced apoptosis of certain tumors (4, 5), and it is possible that amplification of DcR3 by tumor cells may select for apoptosis-resistant cells.

Bone metabolism is one of the in vivo processes regulated by OPG. OPG has a dramatic effect on both osteoclast differentiation and activation (2, 6, 7). Addition of rOPG to osteoclast cultures inhibits the differentiation of precursors to mature, multinucleated osteoclasts (2). OPG can also directly inhibit the function of mature osteoclasts in bone slice cultures (7). Furthermore, opg transgenic mice develop osteopetrosis (2). Clearly, this molecule regulates osteoclastogenesis in the bone marrow.

OPG has two known TNF family ligands: receptor activator of NF-κB ligand (RANKL) and TNF-related apoptosis-inducing ligand (8, 9, 10). RANKL, like OPG, regulates final stages of osteoclast differentiation (8, 9, 11, 12). RANKL is primarily expressed by T cells and bone marrow cells (8, 9, 13, 14), implying that it functions within the immune system in addition to bone metabolism. TNF-related apoptosis-inducing ligand is known to cause apoptosis of a variety of cell lines by ligating one of its several death receptors (15), yet its in vivo role still remains to be elucidated (15, 16, 17).

The RANKL-RANK system influences processes in the immune system. Both RANKL- and RANK-deficient mice have defects in T and B cell development and lymphorganogenesis (11, 18), implicating these molecules in lymphocyte and lymph node (LN) development. Also, in vitro studies suggest that RANKL and RANK may play a role in DC survival and function (13, 14, 19). Indeed, both RANK and RANKL were originally described in studies of T cell and DC activation (13, 14, 20). The expression pattern of these molecules also suggests they are involved in the immune system. While RANK and OPG have been detected on a variety of cells, they are also expressed on B lymphocytes and DCs (1, 13, 21). Furthermore, the expression of OPG is up-regulated in DC and primary B cells by activation through CD40 (1), a receptor required for germinal center (GC) formation (22, 23). Thus, OPG may be involved in regulating B cell or DC functions.

To address the in vivo role of OPG in the immune system, we generated OPG-deficient mice using targeted genetic recombination in embryonic stem (ES) cells. Mice homozygous for the disrupted opg allele are viable and progressively developed severe osteoporosis as they age. Further analysis revealed perturbations in central and peripheral B compartments, including an accumulation of type 1 transitional (T1) B cells (24). Ex vivo, pro-B cells and DCs were hyperresponsive in functional assays. These immune phenotypes are converse to that found in RANKL^−/− mice, thus providing a genetic link between these molecules in the immune system. When challenged with a T-dependent (TD) Ag, the opg^−/− mice were less effective in their ability to isotype switch. We propose that these differences are attributable to dysregulation of RANK stimulation of both DC and B cells and discuss possible mechanisms that can account for these observations.

Mouse opg cDNA was isolated using RT-PCR using the following primers: 5′-GAGGTTTCTCGAGGACCACAATGAACAA-3′ (upstream) and 5′-GGCCCATCTAGAAGAAACAGCCCAGTG-3′ (downstream). Using the mouse opg cDNA fragment as a probe, we screened filters representing clones from a mouse genomic bacterial artificial chromosome (BAC) library (strain 129S6/SvEvTac; RPCI-22; Research Genetics, Huntsville, AL) following the manufacturer’s protocol. We identified a BAC clone that was positive for the mouse opg genomic locus. From this clone, we subcloned two XbaI-SmaI fragments that were ∼2.5 kb and 9 kb into pBluescript II (SK⁺; Stratagene, La Jolla, CA) and obtained partial sequences of both. Next, we subcloned a 2-kb SnaBI-EcoRI fragment into the targeting vector pKMCS, upstream of the PGK-neo^r gene. For the long arm, we subcloned a 6-kb SmaI-StuI fragment downstream of PGK-neo^r and upstream of PGK-dta. The construct was linearized by digestion with XhoI. As previously described (25), AK7 ES cells were transfected with 20 μg of the linearized targeting construct. Transfected ES cell clones that had integrated the targeting construct were selected by their resistance to neomycin. Cells that had randomly integrated the construct were negatively selected by their expression of the PGK-dta gene. Over 300 clones were screened and one ES cell clone had a successfully targeted disruption of an opg allele, which was confirmed by southern blotting of ES cell DNA.

Briefly, DNA was extracted from spleen cells. Purified genomic DNA (10 μg) was completely digested with XbaI. Fragments were resolved by electrophoresis through a 0.7% agarose gel in TAE buffer, then transferred to GeneScreenPlus membrane (NEN Life Sciences Products, Boston, MA) as previously described (25). The DNA was hybridized to a ³²P-labeled probe generated from random hexamer primer extension of a genomic opg fragment. The probe was prepared using a random primers DNA labeling system (Life Technologies, Gaithersburg, MD) following the manufacturer’s protocol.

PCR was used to genotype mice harboring the disrupted allele. An opg locus-specific primer (5′-GGTCCTCCTTGATTTTTCTATGCC-3′) was used in combination with upstream primers specific for neo^r (5′-TGACCGCTTCCTCGTGCTTTAC-3′) or opg (5′-TGCCCTGACCACTCTTATACGGAC-3′). Approximately 2 μg of genomic DNA was used as a template in a 50-μl reaction composed of 2.5 U Taq polymerase (Promega, Madison, WI), 10× Mg²⁺-free PCR buffer (Promega), and 1.5 mM MgCl₂.

Livers from mice were harvested and quick-frozen in liquid nitrogen. The frozen livers were ground using a mortar and pestle, then lysed in TRIzol (Life Technologies). Total liver RNA was extracted following the manufacturer’s protocol. RNA was resolved on a 1.5% agarose gel in MOPS buffer, then transferred to GeneScreenPlus, following the manufacturer’s protocol. The RNA was hybridized by standard methods using ³²P-labeled probe generated from random hexamer primer extension of the specified opg cDNA fragments.

A total of 13 mice were used to examine the effects of OPG on the morphology and patterns of growth/remodeling of the skeleton. The mice ranged in age from 1 to 9 mo and consisted of littermate groups of opg^+/+, opg^+/−, and opg^−/− animals. Ten days before sacrifice, the mice were injected i.p. with a solution of 30 mg/kg calcein (Sigma, St. Louis, MO) that was neutralized with a 1-N solution of NaOH and forced through a 0.22-μm syringe filter. A second solution of alizarin complexone (Sigma) was administered in a similar manner 7 days later. The animals were euthanized 3 days following the last injection. Both calcein and alizarin complexone bind with calcium ions on the surfaces of newly forming apatite crystals, thus labeling the bone undergoing mineralization during the period of exposure (26). Following euthanasia the skeletons were stripped of soft tissue, and the individual elements were radiographed and photographed before being embedded in methyl methacrylate. The plastic blocks were sectioned to a thickness of 30–40 μm with a Leica (Deerfield, IL) SP1600 saw microtome and the mounted sections were viewed using the fluorescent mode of a Nikon (Melville, NY) Eclipse E400 microscope.

For flow cytometry analyses, cell suspensions were prepared from the specified organs of 1.5- to 4-mo-old mice from opg^+/+, opg^+/−, and opg^−/− littermates. Inguinal and/or cervical LNs were collected. RBCs from spleen and bone marrow cells were lysed using ACK (0.155 M NH₄Cl, 0.1 mM EDTA, 0.01 M KHCO₃). Flow cytometric analysis was performed as previously described (27). The mAbs used for the experiments reported are as follows: anti-B220 (PE-, FITC-, biotin-conjugated RA3-6B2; PharMingen, San Diego, CA), anti-CD43 (PE-conjugated S7; PharMingen), anti-IgM (FITC-, biotin-conjugated Bet-2, PE-conjugated R6-60.2; PharMingen), anti-CD3 (biotin-conjugated 145-2C11; PharMingen), anti-CD8 (PE-conjugated 53-6.7; PharMingen), anti-CD4 (FITC-conjugated H129.19; PharMingen), anti-CD19 (FITC-conjugated ID3; PharMingen), anti-CD21 (FITC-conjugated 796; PharMingen), anti-IgD (biotin-conjugated JA12.5), anti-CD11c (PE-conjugated HL3; PharMingen), anti-CD86 (biotin-conjugated GL1; PharMingen), anti-IA^b (biotin-conjugated Y3P, a generous gift from A. Rudensky; biotin-conjugated AF6-120.1; PharMingen), and anti-CD40 (biotin-conjugated 1C10). Staining by biotinylated mAbs was visualized by streptavidin conjugated to PerCP (Becton Dickinson, Mountain View, CA). Appropriate FITC, PE, and biotinylated isotype controls were run in each experiment to determine gates.

Pro-B cell proliferation assays were performed as previously described (28). Using FACS, pro-B cells (B220⁺CD43⁺) cells were collected from opg^+/+, opg^+/−, and opg^−/− littermates. For the experiments reported, 7500 cells were aliquoted into each well and incubated with 10 ng/ml of rIL-7 (R&D Systems, Minneapolis, MN) for 3 or 4 days. On the specified days, the cells were pulsed with 1 μCi/well of [³H]thymidine. Cells were harvested and [³H]thymidine incorporation was measured.

DCs were prepared from mouse bone marrow as previously described (29), with some modifications. Briefly, bone marrow from opg^+/+, opg^+/−, and opg^−/− littermates was flushed from femurs and tibia, adherent cells were removed, and nonadherent cells were cultured in medium in the presence of GM-CSF (∼40 ng/ml). Mouse GM-CSF was produced by infection of NIH-3T3 cells with Psi2-pM5DGM#6 (M. Mohaupt, manuscript in preparation) The concentration of GM-CSF in the supernatant was determined by ELISA (Dianova, Hamburg, Germany) to be 80 ng/ml.

On day 1 and 3 of culture, the nonadherent cells were discarded and the remaining adherent cells were washed with warm RPMI 1640, than fed with fresh medium and GM-CSF containing supernatant every second day. The DCs were used at days 8–9 for additional experiments. For induction of maturation, DC were incubated for 24 h (days 8–9) with 10 μg/ml LPS (Sigma-Aldrich, Deisenhofen, Germany).

T cell hybridoma assays were performed as previously described (30, 31). Triplicate cultures of 1 × 10³ cells/well of LPS-activated opg^−/− or opg^+/− bone marrow-derived DCs were incubated with variable concentrations of peptide (Eα(52–68) or hClip), and 1 × 10⁵ cells/well of T cell hybrids specific for Eα(52–68):I-A^b or hClip:I-A^b. After 24 h of culture, the supernatants were assayed for production of IL-2 by HT-2 proliferation. The degree of proliferation was assessed by the Alamar blue colorimetric assay. The results are expressed in arbitrary OD units (A550-A600 nm).

For the MLR, the day 9 bone marrow-derived DCs were further purified using FACS. DCs were washed, then incubated with dialyzed anti-CD11c⁺. Cells were initially seeded at 10,000 cells/well, then serially diluted. The opg^−/− and control mice were H-2^b. T cells were isolated from LN and spleens of BALB/c mice (H-2^d). For the spleen, RBCs were lysed using Gey’s hemolysis solution. CD4⁺ T cells were enriched by complement lysis. Cells were incubated with 10 μg/ml of Bet-2 (anti-IgM) and 3.155 (anti-CD8) (at 1:200 dilution) for 30 min at 4°C. Cells were then incubated with 20 μg/ml of MAR185 (Rat anti-mouse Ig), for 30 min at 4°C. Cells were then treated with guinea pig Low-Tox complement (Cedarlane Laboratories, Westbury, NY) for 45 min at 37°C. The cells were washed, then overlayed onto a Ficoll-Hypaque cushion (Pharmacia, Piscataway, NJ). The cells at the interface were collected and washed. The enriched CD4⁺ T cells were incubated with 0.5 μg/ml 2C11 (anti-CD3) for 18 h. The cells were then washed, then ∼10,000 cells were seeded into wells containing DCs. The reaction was pulsed with 0.5 μCi/well of [³H]thymidine after 3 days. After 18 h, cells were harvested and [³H]thymidine incorporation was measured.

Immunization with 100 μg of DNP-keyhole limpet hemocyanin (KLH; Calbiochem, La Jolla, CA), and measurement of serum DNP-specific Ab isotypes were performed as previously described (27). Briefly, 100 μg alum-precipitated DNP-KLH was injected i.p. opg^−/− and control littermate mice of 4–6 wk of age were chosen, so that the last time point would be before the onset of severe osteoporosis. Serum was collected and was assayed for DNP-specific Ab by ELISA. Nunc-Immuno Plate MaxiSorp 96-well plates (Van Waters and Rogers, Bisbane, CA) were coated with 4 μg of DNP-BSA. Plates were washed, then blocked with 1% BSA and 0.05% Tween 20. Serial dilutions of sera were prepared and added in triplicate to plates. The specified isotypes were detected with 450 ng/ml sheep antisera specific for the indicated mouse isotype (The Binding Site, Birmingham, U.K.) followed by 375 ng/ml HRP-conjugated donkey anti-sheep serum. Titers of DNP-specific isotypes were revealed by peroxidase reaction using o-phenylenediamine as a substrate. Variation among plates was controlled by pooling anti-DNP antiserum from immunized mice, preparing bulk serial dilutions as in the test samples, and using it as a relative standard on each plate. Relative isotype-specific Ab response to DNP of each test sample was calculated by normalizing the value at half-maximal OD of the test sample against the value at half-maximal OD of the standard dilutions.

Using mouse opg cDNA as a probe, we isolated a mouse genomic BAC clone including the mouse opg locus. The targeting construct was designed so that a targeted recombination event would eliminate the coding sequence for the first two Cys-rich repeats of the mature protein. In addition, 8 of 21 residues in the signal peptide and the first half of the third Cys-rich region would be deleted (Fig. 1 A).

The recombination of the targeting construct with the opg locus introduces a XbaI site that could distinguish the disrupted allele from the wild-type (WT) allele, using a probe specific for a 3-kb region 3′ of the 6-kb SmaI StuI long arm. In XbaI-digested genomic DNA, the mutant band is ∼4 kb smaller than the WT band, as predicted (Fig. 1,B). A PCR-based screen was developed to facilitate genotyping (Fig. 1 B).

To demonstrate that the intact OPG product was not expressed, we performed Northern blot analysis of total RNA from the liver isolated from opg^−/− or opg^+/− mice. Using a probe specific for the region deleted (Fig. 1,C), we could not detect a transcript from the opg^−/− mice (Fig. 1,D). When the entire cDNA was used as a probe, we detected a smaller transcribed product (Fig. 1 D). This indicates that the disrupted allele is transcribed, and suggests that the PGK-neo^r gene is eliminated by splicing. Based on the genomic sequence of opg (32), the spliced product would no longer be in frame, and thus would produce a nonfunctional protein.

Because OPG plays a critical role in regulating osteoclast formation, opg^−/− mice were expected to develop severe osteoporosis (12, 33). To test whether the disrupted allele was indeed a functional null, we radiographed opg^+/− and opg^−/− littermates (Fig. 2,A). The bone density of the 3-mo-old opg^−/− male mouse is dramatically decreased compared with the control. Older opg^−/− mice often exhibit fractures in long bone diaphyses, particularly the humerus, femur, and fibula (K.L.R., S.W.H., and T.J.Y., unpublished observations). Animals as young as 2 mo showed shortened femoral necks, suggesting common traumatic fracture at this location (Fig. 2,B). Surprisingly, the outer diameters of the long bone diaphyses and the mandibular corpus are greater in opg^−/− mice compared with their opg^+/− littermates (Fig. 2, B and C). To follow up on the radiographic results, mice were injected with fluorochromes, which label newly mineralized bone (26). Histologically, the cortical bone of opg^−/− mice is porous compared with that of opg^+/− littermates and shows a greater quantity of fluorescing label (Fig. 2 C). This finding suggests that increased osteogenesis accompanies the increased osteoclast activity in opg^−/− mice.

In long bones, the marrow cavity is the site of hematopoiesis and B lymphopoiesis. In opg^−/− mice, the dysregulation of osteoclast production leads to drastic changes in the bone architecture. Furthermore RANKL-deficient mice have a defect in B cell development from the pro-B to pre-B cell transition (11), suggesting that RANKL/RANK/OPG may play a role in regulating B cell development.

To test whether OPG can regulate the pro-B cell population, we compared the ability of ex vivo opg^−/−, opg^+/−, and opg^+/+ pro-B cells to respond to rIL-7. Pro-B cells are dependent on the presence of IL-7 for survival and proliferation (28, 34, 35). We isolated and cultured pro-B cells from opg^−/− or opg^+/− controls in the presence of IL-7 for 3 or 4 days, then measured differences in proliferation (Fig. 3). opg^−/− pro-B cells gave a greater proliferative response to IL-7 than either opg^+/+ or opg^+/− controls. Interestingly, a gene dosage effect was evident because cells with one copy of the intact locus had an intermediate response to IL-7. Generally, opg^−/− pro-B cells had a 1.7- to 2-fold increase in IL-7 responsiveness compared with opg^+/+ pro-B cells.

Because pro-B cells from the opg^−/− bone marrow had an enhanced proliferative response to IL-7 in vitro, we next questioned whether this population would be increased in vivo. Using flow cytometry we observed no significant differences in the numbers of nonlymphoid lineage (B220⁻) and lymphoid lineage (B220⁺) cells in opg^−/− and control mice (Table I). However, B220^lowCD43⁺IgM⁻ pro-B cells (36), were increased in opg^−/− mice compared with littermate controls. Similar results were obtained using CD25 to distinguish pro- and pre-B populations (data not shown). Because the mean percentages of other B220⁺ populations were similar, these data show that there is an increase in pro-B cells in the opg^−/− bone marrow. The observed increasein the pro-B cell compartment is unlikely to be due to the expansion of the marrow cavity in opg^−/− osteoporotic mice because the absolute number of lymphoid lineage cells is not increased. Together with the enhanced proliferative response of opg^−/− pro-B cells in vitro, this result suggests that OPG negatively regulates expansion of the pro-B cell pool.

Because both RANK and OPG are expressed in B lineage cells (1, 13) and both are up-regulated by CD40 cross-linking, we next analyzed peripheral B cell populations. We consistently observed a greater percentage of peripheral B cells (B220⁺CD19⁺) in opg^−/− mice vs controls. This difference was evident in the absolute numbers of splenic and LN B cells in opg^−/− mice vs controls (Table II). In contrast, there were no significant differences in mean numbers of LN or splenic CD8⁺ and CD4⁺ T cells.

Based on levels of membrane IgM and IgD expression, the majority of splenic B cells can be subdivided into IgM^highIgD^low, IgM^highIgD^high, or IgM^lowIgD^high (24). These three populations have been classified as T1 plus marginal zone (MZ), type 2 transitional (T2), and mature B cells, respectively (24). Using IgM and IgD expression levels, we found that the mean percentage and absolute numbers of the IgM^highIgD^low B cells were significantly increased in opg^−/− vs controls (Table II). On average, there were 53% more of this population in opg^−/− mice than controls. There was also a modest increase in the IgM^highIgD^high B cell population, but no significant difference in the mature splenic B cell population.

The IgM^highIgD^low population includes the T1 and MZ B cells (24). These two subpopulations can be distinguished by CD21 expression; T1 B cells are CD21^low and MZ B cells are CD21^high (24, 37). Based on the expression of IgM and IgD on mature B cells in LN and immature B cells in bone marrow, we determined the gates for the IgM^highIgD^low, IgM^highIgD^high, or IgM^lowIgD^high populations in the spleen (Fig. 4,A). In Fig. 4,B, the expression level of CD21 was measured in the IgM^highIgD^low population. The percentage of IgM^highIgD^lowCD21^low population was consistently increased in the opg^−/− mice, compared with opg^+/+ mice. This increase in the percentage of T1 B cells was reflected by a significant increase in the absolute numbers (Fig. 4,C). We also observed a modest increase in numbers of T2 and mature B cells in opg^−/− mice and also to some extent in opg^+/− mice, compared with opg^+/+ mice, again suggesting a gene dosage effect. In contrast, there was no discernable difference in the numbers of MZ B cells or T cells (Fig. 4 C).

Given these lymphoid developmental perturbations in opg^−/− mice, we next investigated the effect of OPG deficiency on immune responses. Both RANK and OPG are up-regulated by CD40 ligation in DCs (1, 13), which suggests that during T cell activation by DCs, OPG may be expressed as a negative regulator to modulate the T cell responses (38). According to this model, we hypothesized that in the absence of OPG, dysregulated RANK signaling could alter stimulatory capabilities of opg^−/− DCs.

Bone marrow cells from opg^−/− or opg^+/− littermates were cultured in the presence of GM-CSF. After 7 days, an enriched population of functional DCs differentiated from monocytic precursors. Flow cytometric analysis of these cells revealed coexpression of CD11c, MHC class II, and CD86 (data not shown). Maturation of these cultured DCs was induced by incubation with LPS, as measured by up-regulation of costimulatory molecules (39). The levels of expression of MHC class II or CD86 were not different on opg^−/− and opg^+/− CD11c⁺ cells, both on stimulated and unstimulated cells (Fig. 5,A). Also, when tested for their ability to present exogenous peptide to T cell hybridomas expressing a peptide-Ia^b complex-specific TCR, we observed no difference in this early presentation event (Fig. 5,B). However, in 3-day MLRs, we found that DCs from opg^−/− mice consistently had a 2- to 5-fold enhanced ability to stimulate allogeneic T cell proliferation (Fig. 5 C) compared with control opg^+/− DCs.

Next, we measured Ab response to TD Ag mounted by opg^−/− mice. We immunized mice with the TD Ag, DNP-KLH, and measured Ag-specific serum Ig levels of several isotypes from opg^−/− and control mice (opg^+/+ and opg^± ) on days 0, 7, 14, and 21. On day 21, the mice were boosted with a second challenge Ag and the secondary Ab response was measured on days 28 and 35.

The Ab responses of each isotype in the opg^+/+ control mice were very similar; however, the immune responses in opg^−/− mice were highly variable (Fig. 6). We observed a significant difference in the amount of anti-DNP IgG3 produced by the opg^−/− mice on days 14, 21, and 35, as compared with the opg^+/+ control mice. When averaged together, the ability of the opg^−/− mice to mount an IgM, IgG1, IgG2a, and IgG2b response to DNP was comparable to the response by opg^+/+ control mice. However, three of the eight opg^−/− mice assayed had deficient primary IgG2a, IgG2b, and/or IgG3 responses. One of the four mice failed to isotype switch to IgG2a, and another mouse failed to isotype switch to IgG3. This failure to isotype switch was not observed in any of the 11 opg^+/+ or opg^+/− control mice. Thus OPG may influence the ability of mice to sustain an IgG3 response to a TD Ag.

Previous studies of opg-transgenic and opg^−/− mice have revealed that OPG plays a vital role in regulating osteoclast production in the bone marrow (2, 12, 33). In this study, we have shown that OPG also regulates B lymphopoiesis, most notably at the pro-B cell and transitional B cell stages.

Clues as to how OPG may regulate B lymphopoiesis may come from examining OPG’s role in regulating bone metabolism. In the absence of OPG’s normal “braking” of osteoclast maturation, the opg^−/− mice exhibited drastic increases in cortical bone porosity that was accompanied by a striking acceleration of new apposition on periosteal and intracortical surfaces. Preliminary data indicate most pronounced increases in mineralization in those parts of the skeleton under greatest mechanical load, the long bones and the mandibular body, leading to bone hypertrophy on a gross morphological level. We hypothesize that because the quality and, therefore, strength of bone is compromised by increased bone resorption in the opg^−/− mice, the compensatory response of increased apposition is most pronounced in biomechanically critical regions. This “coupling” of resorption and formation is a well known component in the pathophysiology of osteoporosis, although the underlying biochemical mechanisms are not fully understood (40). Physiologically, the defects in bone metabolism are the most striking phenotype of OPG deficiency; however, multiple processes in the immune system are dysregulated as well.

Could the same OPG dependent regulatory mechanism affect the production of pro-B or transitional B cells in opg^−/− mice? Despite drastic alterations in medullary bone architecture in these mice (12, 33), the overall composition and proportions of lymphoid and nonlymphoid cells is not changed in the absence of OPG. However, the pro-B cells and transitional B cells recently arising from the bone marrow are disrupted (Tables I and II, and Fig. 4).

A likely explanation of our results is that the B lymphoid compartment itself is directly regulated by OPG. Not only are pro-B cells and transitional T1 B cells expanded in opg^−/− mice, but opg^−/− pro-B cells have a greater in vitro proliferative response to IL-7 than heterozygous or WT pro-B cells (Fig. 3). Our results suggest that pro-B cells isolated from opg^−/− mice are intrinsically different in their proliferative response to IL-7 because pro-B cells were cultured in the absence of stromal cells or other growth factors, and because addition of rOPG had no effect on pro-B cell proliferation (data not shown). Also, although the proliferative response of opg^−/− pro-B cells may be affected, the ability of these cells to differentiate into pre-B cells does not appear to be affected (A.J.M. and T.J.Y., unpublished observation). The expansion of peripheral B cells and pro-B cells in our opg^−/− mice is the opposite to that observed in the RANKL^−/− mice (11) or RANK^−/− mice (18). Both of these mutant mice have fewer mature B cells and RANKL^−/− mice have an arrest in the pro-B to pre-B transition. Collectively, these results suggest that RANKL/RANK is involved in the proliferation or proliferative expansion of pro-B cells, and that OPG regulates this process.

There is a striking negative regulation of B lymphopoiesis by sex steroids, particularly by estrogen (41). This observation was noted by Kincade et al. (41) while studying B lymphopoiesis during pregnancy. Smithson et al. (42) demonstrated that estrogen stimulates bone marrow stroma to secrete factors that negatively regulate B lymphopoiesis. Interestingly, opg expression is positively regulated by estrogen (43). Also, we have observed by RT-PCR analysis that human stromal cells (kindly provided by T. LeBien, University of Minnesota, Minneapolis, MN) and mouse stromal cell lines (S10 and S12), which are capable of supporting B lymphopoiesis, can express opg (data not shown). Finally, Medina et al. (44) demonstrated that estrogen affects proliferation of developing B cells at an early stage in vivo, likely at the pro-B cell stage or at a discrete stage before pro-B cells. It is tempting to speculate that estrogen stimulates stromal elements to produce OPG that negatively regulates B lymphopoiesis.

In the spleen, the numbers of T1 B cells, which represent the newly produced B cells from the bone marrow (45), are selectively increased in opg^−/− mice. The T1 B cells are thought to be dependent on B cell receptor-induced signals for entry into primary follicles where they receive additional maturation signals. T1 B cells differentiate into T2 B cells and up-regulate the expression of IgD and other surface receptors (24). Additional signals are then required for the transition of T2 B cells into mature B cells. Because the T2 and mature B cell populations appeared marginally affected by the lack of OPG, OPG likely regulates the T1 B cell pool.

The increase in LN B cells and splenic T1 B cells may represent an accumulation of cells, as opposed to proliferation, because we could not detect differences in activation markers expressed by splenic or LN B cells from opg^−/− and control mice. We favor the possibility that due to the increase in the proliferative response of pro-B cells, there is an increase in the production of immature B cells in vivo. Another explanation is that there is more efficient homing of B cells to these peripheral sites, analogous to the mechanism that OPG regulates mature osteoclast migration into bone matrix (9, 11, 18). Finally, because T1 B cells that do not receive maturation signals likely undergo apoptosis (46), another possibility is that there is less apoptosis of B cells in opg^−/− mice. RANK mediated Bcl-X_L induction (14) may also occur in B cells so that absence of OPG indirectly augments Bcl-X_L expression and prolongs T1 B cell survival.

The fact that opg^−/− mice develop osteoporosis (Fig. 2 and Refs. 12 and 33) and that both RANKL^−/− and RANK^−/− mice develop osteopetrosis has led to the proposal that OPG functions as a molecular brake during osteoclastogenesis (11, 12, 18, 33). OPG apparently has a similar function during some cellular interactions in the immune system. On a per cell basis, we consistently observed that opg^−/− DCs were 2- to 5-fold more effective at stimulating T cells (Fig. 5 C) than opg^+/− DCs. This result is the converse to that observed in RANKL^−/− mice in that RANKL^−/− T cells are impaired such that they require more allogeneic DCs to stimulate IL-2 production than RANKL^+/− T cells. Because RANK stimulation of DCs by RANKL increases their ability to stimulate T cells (13, 14), collectively, these results suggest that OPG normally functions as a molecular brake on the RANKL/RANK pathway during T cell-DC interactions.

OPG most likely modulates the ability of DCs to stimulate T cells after Ag processing and later during presentation, because there is no difference in the ability of opg^−/− and opg^+/− DC to present exogenous peptides to Ag-specific T hybridomas. Also, early costimulatory events are probably not affected by the absence of OPG because the levels of class II and CD86 were identical between mature opg^−/− and opg^+/− DCs. Using RNase protection assays, we observed no difference in a selected set of early T cell cytokine gene induction between opg^+/− and opg^−/− DCs, including IL-2, IL-4, IL-13, and IFN-γ (data not shown). What is more likely is that in the absence of OPG, the quality and duration of RANKL/RANK interactions are altered. Because RANK stimulation can up-regulate Bcl-X_L expression (14) in the absence of OPG, survival of DC may be increased, thereby promoting more effective T cell proliferation.

Our data indicate that OPG deficiency affects two cell types that are important mediators of the immune response to a TD-Ag; DCs, which initiate the response, appear to have increased ability to stimulate T cells, and peripheral B cells, whose populations are elevated in spleen and LN. When immunized with a TD-Ag, the lack of OPG affects the ability of B cells to sustain an IgG3 response. Interestingly, we observed a lack of IgG2a response in one mouse, and a lack of IgG3 in another, suggesting that the absence of OPG may affect IgG isotype switch. Perhaps OPG may be modulating B cell response by altering cytokine patterns or expression of membrane costimulatory molecules. Another possibility is that the accumulation of T1 B cells in opg^−/− mice indirectly influence GC formation. The possible effects of OPG on GC formation will be the subject of future investigation.

The evidence in this report genetically establishes the link between RANKL, RANK, and OPG in the immune system. opg^−/− mice exhibit a contrasting phenotype to RANKL^−/− and RANK^−/− mice in terms of osteoclast activity and numbers of peripheral B cells. Also, RANKL^−/− B cell development and RANKL^−/− T cell–DC interactions, were disrupted. In cells isolated from opg^−/− mice, we see the converse effect of these cell types in similar functional assays. Physiologically, OPG deficiency during a TD-Ag immune response was demonstrated to result in an inefficient ability to switch to particular isotypes, implicating OPG’s role in stochastic processes that influence Ab isotype switch. We conclude that OPG is a normal regulator during B cell development and during DC and B cell function in an immune response.

We thank D. Magaletti for excellent technical assistance; members of the Clark laboratory, especially K. L. Otipoby, for intellectual discussion; Dr. P. Soriano for valuable intellectual input, technical guidance and critical review of the manuscript; P. Wong, Dr. T. Nakagawa, and Dr. A. Rudensky for intellectual discussion, reagents, and technical assistance; and Dr. W. C. Dougall and Dr. J. J. Peschon for intellectual discussion and sharing unpublished results.