RANKL (Receptor Activator of NFκB Ligand) Produced by Osteocytes Is Required for the Increase in B Cells and Bone Loss Caused by Estrogen Deficiency in Mice*

The cytokine receptor activator of NFκB ligand (RANKL) produced by osteocytes is essential for osteoclast formation in cancellous bone under physiological conditions, and RANKL production by B lymphocytes is required for the bone loss caused by estrogen deficiency. Here, we examined whether RANKL produced by osteocytes is also required for the bone loss caused by estrogen deficiency. Mice lacking RANKL in osteocytes were protected from the increase in osteoclast number and the bone loss caused by ovariectomy. Moreover, these mice did not exhibit the increase in bone marrow B lymphocytes caused by ovariectomy that occurred in control littermates. Deletion of estrogen receptor α from B cells did not alter B cell number or bone mass and did not alter the response to ovariectomy. In addition, lineage-tracing studies demonstrated that B cells do not act as osteoclast progenitors in estrogen-replete or estrogen-deficient mice. Taken together, these results demonstrate that RANKL expressed by osteocytes is required for the bone loss as well as the increase in B cell number caused by estrogen deficiency. Moreover, they suggest that estrogen control of B cell number is indirect via osteocytes and that the increase in bone marrow B cells may be a necessary component of the cascade of events that lead to cancellous bone loss during estrogen deficiency. However, the role of B cells is not to act as osteoclast progenitors but may be to act as osteoclast support cells.

Osteoclasts are bone resorbing cells derived from myeloid progenitors in the bone marrow (1). The cytokine receptor activator of NFB ligand (RANKL), 2 encoded by the Tnfsf11 gene, is essential for osteoclast formation but plays important roles in other processes such as mammary gland and lymphocyte development (2,3). Consistent with this, RANKL is produced by a variety of different cell types and in response to many different stimuli (4). Osteocytes are cells that live in mineralized bone and are derived from osteoblasts, which produce bone matrix (5). Gene deletion studies in mice have demonstrated that osteocytes are an essential source of the RANKL involved in osteoclast formation under physiological conditions as well as in response to biomechanical unloading and dietary calcium deficiency (6 -8).
Estrogen deficiency in mice increases osteoclast number on cancellous and cortical bone and causes bone loss in both compartments (9). Estrogen deficiency also causes a striking increase in B lymphocyte number in the bone marrow (10,11). Moreover, deletion of the Tnfsf11 gene from B cells prevents both the increase in B cell number and the increase in cancellous osteoclast number caused by ovariectomy (12). These findings suggest that estrogen may suppress osteoclast number in part by suppressing B cell number in the bone marrow.
How B cells might contribute to osteoclast formation during estrogen deficiency is unclear. On the one hand, RANKL produced by B cells may directly interact with its receptor RANK on osteoclast progenitors and thereby stimulate osteoclast formation. On the other hand, several independent studies have demonstrated that purified populations of B cells can be induced to differentiate into osteoclasts when exposed to recombinant RANKL in vitro (13)(14)(15)(16)(17). Thus, B cells may act as a source of osteoclast progenitors, at least under some conditions. However, there has been no evidence that this phenomenon occurs in vivo either in estrogen-replete or estrogen-deficient conditions.
The goal of the current study was to determine whether RANKL produced by osteocytes contributes to the elevated osteoclast formation and bone loss caused by estrogen deficiency. We found that this is the case but that deletion of the Tnfsf11 gene from osteocytes also prevented the increase in B cell production caused by estrogen deficiency, suggesting that estrogen controls B cell number indirectly. Consistent with this, we found that deletion of estrogen receptor ␣ (ER␣), encoded by the Esr1 gene, from B cells had no effect on B cell number. Lastly, we used in vivo lineage-tracing studies to investigate the possibility that cells committed to the B cell lineage can act as osteoclast progenitors and found that this was not the case.

Osteocyte RANKL Is Required for Ovariectomy-induced Bone
Loss-To determine whether RANKL production by osteocytes is required for the bone loss caused by estrogen deficiency, adult female mice lacking the Tnfsf11 gene in osteocytes (hereafter referred to as Tnfsf11 ⌬Ot ) and their control littermates (hereafter referred to as Tnfsf11 f/f ) underwent either a sham operation or ovariectomy. Six weeks after the operations, ovariectomized mice had lower uterine weight than sham-operated mice, confirming estrogen deficiency (Fig. 1A). Quantification of the Tnfsf11 locus in genomic DNA from tissues harvested FIGURE 1. Deletion of Tnfsf11 from osteocytes prevents ovariectomy-induced bone loss. 6-Month-old female Tnfsf11 f/f and Tnfsf11 ⌬Ot mice were either sham-operated (white bars) or ovariectomized (OVX) (dark gray bars) and killed 6 weeks later. A, uterine weight (n ϭ 10 -12 animals per group). B, real time PCR quantification of Tnfsf11 genomic DNA in femoral cortical bone, CD19ϩ bone marrow cells, CD19Ϫ bone marrow cells, spleen, kidney, liver, and muscle (n ϭ 3-12). C, CT images of vertebral cancellous bone and femoral cortical bone. Bar ϭ 500 m. D, bone volume over tissue volume (BV/TV) of cancellous bone measured in lumbar vertebra by CT (n ϭ 10 -12). E, cortical thickness of the femoral midshaft measured by CT (n ϭ 10 -12). F and G, osteoclast surface per bone perimeter (Oc.S/B.Pm) and osteoclast number per bone perimeter (Oc.N/B.Pm) measured by histomorphometric analysis of lumbar vertebra (n ϭ 6 -10). H and I, real time RT-PCR of Tnfsf11 and Tnfrsf11b mRNA in tibial cortical bone (n ϭ 10 -12). J, Tnfsf11 mRNA expression in CD19ϩ bone marrow cells (n ϭ 3-5). *, p Ͻ 0.05. from the sham-operated mice confirmed deletion of the gene in osteocyte-enriched bones but also revealed a small but significant deletion in muscle tissue (Fig. 1B). However, RANKL mRNA levels were Ͼ100-fold lower in muscle compared with bone tissue and were not altered by ovariectomy, arguing against any functional contribution of muscle-derived RANKL to bone resorption (supplemental Fig. 1).
Ovariectomy led to reduced vertebral cancellous bone volume and femoral cortical thickness in Tnfsf11 f/f mice but not Tnfsf11 ⌬Ot mice (Fig. 1, C-E). However, as noted in previous studies, both of these parameters were elevated in sham-operated Tnfsf11 ⌬Ot mice compared with sham-operated Tnfsf11 f/f littermates. Histological analysis of vertebral cancellous bone revealed that ovariectomy increased osteoclast number and surface in Tnfsf11 f/f mice, but this did not occur in Tnfsf11 ⌬Ot mice, which had fewer osteoclasts than Tnfsf11 f/f mice, when compared with the sham-operated groups (Fig. 1, F-G). As expected from Tnfsf11 gene deletion, RANKL mRNA levels were lower in cortical bone from Tnfsf11 ⌬Ot mice but were not changed by ovariectomy in either these mice or control littermates (Fig. 1H). Osteoprotegerin mRNA levels were also lower in Tnfsf11 ⌬Ot mice but were unchanged by ovariectomy (Fig.  1I). In addition and consistent with our earlier report, ovariectomy did not change RANKL mRNA in B lymphocytes in Tnfsf11 f/f or Tnfsf11 ⌬Ot mice (Fig. 1J). These results demonstrate that RANKL produced by osteocytes is essential for the increase in osteoclast number and the bone loss caused by estrogen deficiency but that this occurs without significant changes in RANKL or osteoprotegerin production.
Osteocyte RANKL Is Required for the Increase in B Cells after Ovariectomy-We have shown previously that deletion of RANKL from B lymphocytes prevents the increase in bone marrow B cells and cancellous osteoclasts after ovariectomy (12). This observation suggested a possible link between the two phenomena. To determine if this relationship was present in mice lacking RANKL in osteocytes, we used flow cytometry to measure bone marrow B lymphocytes in the sham-operated and ovariectomized Tnfsf11 ⌬Ot and Tnfsf11 f/f mice. Although the expected increase in B cells occurred in Tnfsf11 f/f mice, it did not occur in Tnfsf11 ⌬Ot mice ( Fig. 2A). T lymphocyte and erythrocyte progenitor percentages were not altered by ovariectomy in either genotype, but the percentage of monocytes was slightly reduced only in Tnfsf11 f/f mice (Fig. 2, B-D). The effect on monocytes has been observed previously and is likely a consequence of the increase in the percentage of B cells (11). Supporting this idea, the total number of B cells isolated by magnetic bead separation was elevated in ovariectomized Tnfsf11 f/f but not Tnfsf11 ⌬Ot mice (Fig. 2E). These results also demonstrated a trend toward reduced B lymphocyte number under estrogen-replete conditions, but the differences were not significant possibly due to the small effect size and limited number of samples. Therefore, we measured the percentage of bone marrow B cells in a larger number of intact (no operation) Tnfsf11 f/f and Tnfsf11 ⌬Ot mice and found that the percentage of these cells was indeed lower in Tnfsf11 ⌬Ot mice (Fig. 2F).
Osteocyte RANKL may promote B cell number in the bone marrow by binding to RANK on B cells to promote their proliferation or survival. However, deletion of RANK from B cells does not alter B cell number indicating that RANK signaling in another cell type mediates the effects of RANKL on B cells (19). Our observation that Tnfsf11 mRNA in osteocyte-enriched bone was not altered by ovariectomy also argues against such a scenario. Alternatively, osteocyte RANKL could indirectly control B cell number by controlling the number of osteoblast-lineage cells. Osteoblasts increase after ovariectomy in response to increased bone resorption, a phenomenon known as coupling (20). Moreover, osteoblasts provide factors such as IL-7 and CXCL-12 that support B cell development (21)(22)(23)(24). Consistent with the idea that the failure to increase B cells was due to reduced support by osteoblast-lineage cells, IL-7 mRNA in bone was increased by ovariectomy in Tnfsf11 f/f but not Tnfsf11 ⌬Ot mice (Fig. 2G). A similar trend was observed for CXCL-12 mRNA (Fig. 2H).
Estrogen Indirectly Suppresses B Cell Number-The requirement of osteocyte RANKL for the increase in B cells by estrogen deficiency suggests that estrogen controls B cell number indirectly. However, it is possible that a direct effect is also required. To address this question, we deleted ER␣, encoded by the Esr1 gene, from B lymphocytes using CD19-Cre mice, which delete target genes early in the B cell lineage (25). Deletion of the conditional Esr1 allele occurred specifically in bone marrow CD19ϩ cells (Fig. 3A). Moreover, estrogen receptor ␤ mRNA, encoded by Esr2, was undetectable in this cell population and not up-regulated by Esr1 deletion (Fig. 3B).
We then compared the effect of Esr1 deletion on the skeleton and bone marrow B cells in sham-operated and ovariectomized adult mice. Reduced uterine weight confirmed estrogen deficiency in the ovariectomized groups of both Esr1 f/f and Esr1 ⌬B mice (Fig. 3C). Deletion of Esr1 from B cells did not alter bone mass or the loss of bone caused by ovariectomy (Fig. 3, D-F). Importantly, the percentage of CD19ϩ B cells in the bone marrow was not affected by Esr1 deletion, nor was the increase in B cells caused by ovariectomy (Fig. 3G). Moreover, quantification of several stages of B lymphocyte development also showed no difference between genotypes (Fig. 3, H-K). Taken together, these results argue against a direct effect of estrogen on B cell number in the bone marrow in either estrogen-replete or estrogen-deficient mice.
B Cells Do Not Act as Osteoclast Progenitors-Our findings that Tnfsf11 deletion from either B cells or osteocytes prevents the increase in both B cells and osteoclasts caused by ovariectomy supports the idea that the increase in these two cell types may be functionally linked. Many previous studies have shown that isolated B cells can act as osteoclast progenitors when cultured in vitro with M-CSF and RANKL. However, the ability of B cells to act as osteoclast progenitors in vivo has not been demonstrated. Therefore, we used lineagetracing studies to determine whether cells committed to the B lymphocyte lineage could be redirected to the myeloidosteoclast lineage either in the estrogen-replete or estrogendeficient state.
First we confirmed that preparations of CD19ϩ B cells isolated from mouse bone marrow by magnetic beads could be induced to form osteoclasts in vitro after exposure to M-CSF and RANKL (Fig. 4A). We found that CD19ϩ cells isolated to 96% purity formed osteoclasts under these conditions but that this required a higher concentration of cells and longer incubation time than CD19Ϫ cell preparations (Fig. 4, A-C). The requirement for different conditions could be due to the need for B cells to undergo a lineage switch. Alternatively, the oste-oclasts in these cultures may form from contaminating myeloid-lineage cells, and the higher cell number and culture period may be required for these cells to expand to sufficient numbers for osteoclast formation.

Osteocyte RANKL, B Cells, Estrogen, and Bone Loss
To determine if the osteoclasts in these cultures differentiated from cells of the B lymphocyte lineage, we crossed CD19-Cre mice, which cause recombination in all B cells, with tdTomato Cre-reporter mice. In the offspring of this cross, bone marrow B cells were efficiently and specifically labeled with the tdTomato fluorescent protein (supplemental Fig. 2). We performed ovariectomy or sham operations on these mice and isolated CD19-positive cells from bone marrow by magnetic beads (supplemental Fig. 2). Freshly isolated B cells from either sham-operated or ovariectomized mice displayed robust tdTomato fluorescence (Fig. 5A). However, the osteoclasts formed from these same preparations did not display tdTomato fluorescence (Fig. 5B), suggesting that the osteoclasts formed from contaminating myeloid-lineage cells. It is also possible that B cells and contaminating myeloid cells both contributed to osteoclast formation but that the fluorescence from the B cells was diluted to the point of being undetectable. To identify the lower limit of detection from such a contribution, we marked bone marrow macrophages (BMMs) using LysM-Cre;tdTomato mice and mixed these with increasing amounts of unmarked BMMs and found that we could still detect tdTomato-positive osteoclasts even when only 2% of the cells in the culture were from LysMCre;tdTomato mice ( Fig. 5C). Thus, the majority of osteoclasts formed in cultures from purified B cells are not derived from B cells but are likely derived from contaminating myeloid cells. More importantly, all osteoclasts on the cancellous bone surface of LysM-Cre;tdTomato mice were tdTomato-positive but none was positive in either sham-operated or ovariectomized CD19-Cre;tdTomato mice ( Fig. 6 and supplemental Fig. 3). Based on these results we conclude that B cells do not differentiate into osteoclasts under normal physiological conditions or during estrogen deficiency.

Discussion
We have shown previously that RANKL produced by osteocytes is required for the increase in bone resorption caused by either hind limb unloading or dietary calcium deficiency (6,8). Herein we show that osteocyte RANKL is also required for the increase in resorption caused by estrogen deficiency. One important difference between these earlier studies and the present work is that both unloading and calcium deficiency increase levels of Tnfsf11 mRNA in osteocyte-enriched bone, but this did not occur with estrogen deficiency, nor were changes in Tnfrsf11b mRNA observed. Thus RANKL expression by osteocytes is permissive for the increase in osteoclast number and did not appear to be a direct target of estrogen. Consistent with this finding, deletion of ER␣ from osteocytes has no effect on osteoclast number (26,27). We and others have shown that the amount of soluble RANKL protein in bone marrow supernatants does increase with either estrogen or androgen deficiency (12,28,29). However, whether this increase in soluble RANKL contributes to the increase in bone resorption caused by these conditions is unknown.
The Dmp1-Cre transgene used in this study also causes recombination in osteoblasts (6,30). We have shown previously that deletion of Tnfsf11 using a Sost-Cre transgene that does not delete in osteoblasts or lining cells leads to the same reduction in osteoclasts as deletion using the Dmp1-Cre transgene (30). Based on this, we concluded that osteocytes, not osteoblasts or lining cells, are the major source of RANKL involved in oste-oclast formation. This finding together with our observation that Tnfsf11 mRNA levels were unaffected by ovariectomy supports the conclusion that RANKL produced by osteocytes is responsible for supporting osteoclast formation in estrogenreplete and estrogen-deficient mice. It has been reported recently that the Dmp1-Cre transgene targets a subset of stromal cells designated Cxcl12-abundant reticular (CAR) cells (31). CAR cells are also targeted by an Osx1-Cre transgene (24), and we have shown that deletion of Tnfsf11 using this same transgene beginning at 4 months of age does not alter osteoclast number (6), indicating that CAR cells do not supply RANKL for osteoclast formation.
We also found that osteocyte RANKL contributes to the number of B lymphocytes in the bone marrow of estrogen-replete mice and is required for the increase caused by estrogen deficiency. The latter finding was unexpected given that RANKL produced by B lymphocytes themselves is also required for this phenomenon (12). The mechanisms by which osteocyte RANKL controls B cell number are unclear but appear to be indirect as deletion of the receptor for RANKL from B cells does not alter B cell number (19). One possibility is that the low level of bone remodeling in Tnfsf11 ⌬Ot mice may have reduced B cell number by reducing the amount of support factors supplied by osteoblast-lineage cells.
Consistent with this idea, multiple lines of evidence suggest that there is a linkage between the number of osteoblasts on cancellous bone and the number of B lymphocytes in the bone marrow. For example, ablation of osteoblasts reduces B cell number before reductions in other hematopoietic lineages (32). In addition, suppression of PTH receptor signaling specifically in osteoblasts reduces B cell number (22). Perhaps most relevant to the current work, suppression of bone remodeling by administration of bisphosphonates is sufficient to reduce the number of osteoblasts on the bone surface and the number of B cells in the bone marrow of mice (23,33).
Many of these previous studies include evidence that osteoblast-lineage cells support B cell development by expressing IL-7 and CXCL-12 (22,23,32). We found that IL-7 and CXCL-12 mRNAs were elevated by ovariectomy in control mice but not in Tnfsf11 ⌬Ot mice. Consistent with this, others have shown that IL-7 protein in bone marrow fluid is elevated at 7 or 14 days after ovariectomy (34,35). These results together with evidence that osteoblast-lineage cells are an essential source of the IL-7 and CXCL-12 involved in B cell production (21,24) support the idea that the failure to increase B cells in Tnfsf11 ⌬Ot mice is a consequence of a failure to increase osteoblast number. It is also possible that cells of the osteoblast lineage are direct targets of estrogen, as in vitro studies have shown that estrogen suppresses B cell growth factors in bone marrow stromal cell cultures, which contain osteoblast progenitors (36,37). It is also possible that the failure of these cytokines to increase in Tnfsf11 ⌬Ot mice is due to direct regulation of IL-7 and CXCL-12 genes by RANK signaling. However, we did not observe any change in the levels of either IL-7 or CXCL-12 mRNA in bone marrow osteoblasts treated with 100 ng/ml RANKL for 24 h (data not shown), arguing against this scenario.
An increase in B cell number is one of the most consistently observed changes in the bone marrow of estrogen-deficient mice (10,11,39). Because this increase is coincident with bone loss and because of the juxtaposition of the bone marrow with bone matrix, Suda and co-workers (40) proposed a causal link between the increase in B cell number and the elevated bone resorption caused by estrogen deficiency. Consistent with their idea, administration of IL-7 to estrogen-replete mice was sufficient to increase not only B cell number but also osteoclast number and bone resorption (40). Mature B cells, however, do not appear to be required as similar amounts of bone were lost after ovariectomy in mice with and without mature B cells (41).
B cells may also contribute to the bone loss that occurs in estrogen-deficient humans. Comparison of bone marrow cells from pre-menopausal, post-menopausal, and estrogen-treated post-menopausal women revealed higher levels of RANKL on the surface of B cells in post-menopausal women not receiving estradiol (42). Although an increase in total B cells was not noted in this study, a second study reported a decrease in the number of bone marrow cells expressing RANKL in post-menopausal women after 3 weeks of estradiol treatment (43). Although the identity of the RANKL-expressing cells in this latter study was not determined, the cells were in the flowcytometric fraction that contains lymphocytes. Consistent with the idea that B cells may increase in estrogen-deficient women, administration of estradiol or raloxifene to post-menopausal women reduces circulating levels of IL-7 (44,45).
If the increase in B cell number contributes to the increase in osteoclast number, our current study demonstrates that they do this by some mechanism other than by acting as osteoclast progenitors. Although multiple studies utilizing purified B cells have demonstrated that these preparations can produce osteoclasts in vitro, our lineage-tracing studies demonstrate that B cells cannot act as osteoclast progenitors in vivo. Based on this and on our in vitro lineage-tracing studies we conclude that the osteoclasts formed after culture of isolated B cells must be derived from contaminating myeloid cells. A similar conclusion was reached by Aguila and co-workers (46), who demonstrated that more highly purified B cell preparations failed to generate osteoclasts in vitro.
The current results together with earlier findings regarding RANKL produced by B cells and ER␣ produced by osteoclasts suggest the following model to explain the cancellous bone loss caused by estrogen deficiency (Fig. 7). First, loss of estrogen action directly on osteoclasts allows their number to increase, which in turn elevates osteoblast number. RANKL produced by osteocytes is permissive for this increase. The increase in remodeling increases the number of B lymphocytes in the bone marrow due to elevated production of factors such as IL-7 and CXCL-12 by osteoblast-lineage cells. The increase in these factors may be due to an increased number of osteoblasts as well as to loss of estrogen inhibition of their expression in stromal cells. The increase in B cell number then introduces additional RANKL into the system, which promotes even more osteoclast formation, resulting in a feed-forward loop and bone loss. The ability of B cell RANKL to promote bone resorption is highlighted by the recent finding that activation of mTOR in B cells is sufficient to increase RANKL production in these cells and stimulate osteoclast formation and bone loss (47). Overall, our results support the contention that bone loss due to estrogen deficiency in mice results from dysregulation of a complex set of interactions between the skeletal and hematopoietic systems.

Experimental Procedures
Animals-To disrupt Tnfsf11 in mature osteoblasts and osteocytes, Tnfsf11 f/f mice (6) were crossed with mice harboring a Dmp1-Cre transgene (48). Mice used in the ovariectomy experiments were littermates generated by crossing Tnfsf11 f/f with Dmp1-Cre;Tnfsf11 f/f mice. To disrupt Esr1 in B lymphocytes, Esr1 f/f mice (49) were crossed with CD19-Cre mice (25). Mice used in the ovariectomy experiments were littermates generated by crossing Esr1 f/f mice with CD19-Cre;Esr1 f/f mice. Lineage-tracing studies were performed in tdTomato Cre-reporter mice (line Ai9) (50) after crossing with either CD19-Cre or LysM-Cre (51) mice. All lines used in this study were backcrossed into the C57BL/6J genetic background for Ͼ10 generations before the crosses described here. Offspring were genotyped by PCR using the following primer sequences: Cre-for, 5Ј-GCGGTCTGGCAGTAAAAACTATC-3Ј; Cre-rev, 5Ј-GTGAAACAGCATTGCTGTCACTT-3Ј (product size 102 bp); RANKLflox-for, 5Ј-CTGGGAGCGCAGGTTAAATA-3Ј; RANKLflox-rev, 5Ј-GCCAATAATTAAAATACTGCAGG-AAA-3Ј (product size 108 bp (wild type) and 251 bp (floxed allele)); ER␣ f/f -for, 5Ј-TCGTTTTGAATTAATTATGAAT-GTCTG-3Ј, ER␣ f/f -rev, 5Ј-TTCATGTGTTGTGCAAAT-AGC-3Ј (product size 647 bp (wild type) and 933 bp (floxed allele)); oIMR9020, 5Ј-AAGGGAGCTGCAGTGGAGTA-3Ј; oIMR9021, 5Ј-CCGAAAATCTGTGGGAAGTC-3Ј; oIMR9103, 5Ј-GGCATTAAAGCAGCGTATCC-3Ј; oIMR9105, 5Ј-CTG-TTCCTGTACGGCATGG-3Ј (product size 297 bp (wild type) and 196 bp (tdTomato)). Sham and ovariectomy operations were performed at 6 -7 months of age for gene deletion studies and at 3 months of age for lineage-tracing studies. Mice were assigned to ovariectomy or sham groups based on bone mineral density of the lumbar spine. Specifically, mice were rank-ordered by bone mineral density and then assigned the number 1 or 2, successively. Animals with the same number were assigned to the same operation group to give identical group means. Mice were euthanized 6 weeks after the operations. All studies involving mice were approved by the Institutional Animal Care and Use Committees of the University of Arkansas for Medical Sciences and the Central Arkansas Veterans Healthcare System.
Microcomputed Tomography (CT)-Cortical and trabecular architecture was measured by CT of the fourth lumbar vertebra and femur. Soft tissue was removed from L4 vertebra or femurs, which were then fixed in Millonig's 10% formalin for 24 h and transferred gradually from 70% to 100% ethanol. Bones were loaded into a 12.3-mm-diameter scanning tube and imaged using a CT (model CT40, Scanco Biomedical, Bruttisellen, Switzerland). Scans were integrated into three-dimensional voxel images (1024 ϫ 1024 pixels) and a Gaussian filter (sigma ϭ 0.8, support ϭ 1) was used to reduce signal noise. A threshold of 200 was applied to all scans at medium resolution (E ϭ 55 peak kilovoltage, I ϭ 145 A, integration time ϭ 200 ms). The entire vertebral body was scanned with a transverse orientation. The cortical bone and the primary spongiosa were manually excluded from the analyses. All trabecular measurements were made by drawing contours every 10 -20 slices and using voxel counting for bone volume per tissue volume and sphere-filling distance transformation indices without presumptions about the bone shape as a rod or plate for trabecular microarchitecture. Cortical thickness was measured at the femoral mid-diaphysis. Calibration and quality control were performed weekly using five density standards, and spatial resolution was verified monthly using a tungsten wire rod. Beam hardening correction was based on the calibration records.
RNA Purification and Gene Expression-Tibal cortical bone was prepared by removing the distal and proximal ends and removing bone marrow cells by centrifugation at 13,000 ϫ g for 2 min. Cortical bone was then stored at Ϫ80°C before RNA extraction. Total RNA was purified from bone or isolated cells using TRIzol Reagent (ThermoFisher Scientific) according to the manufacturer's instructions. RNA was quantified using a Nanodrop instrument (ThermoFisher Scientific), and RNA integrity was verified by resolution on 0.8% agarose gels. 500 ng of RNA was then used to synthesize cDNA using the High-Capacity cDNA Reverse Transcription kit (ThermoFisher Scientific) according to the manufacturer's directions. Transcript abundance in the cDNA was measured by quantitative PCR using TaqMan Universal PCR Master Mix (ThermoFisher Scientific) (52). The following Taqman assays were used: Tnfsf11 (Mm0041908_m1); Tnfrsf11b (Mm00435452_m1); IL-7 (Mm01295803_m1); CXCL12 (Mm00445553_m1); the house-keeping gene Mrps2 (forward, 5Ј-CCCAGGATGG-CGACGAT-3Ј; reverse, 5Ј-CCGAATGCTGTAATGGCGTAT-3Ј; probe 5Ј-FAM-TCCAGAGCAGGATCC-NFQ-3Ј). Gene expression was calculated using the ⌬Ct method relative to Mrps2 levels (53).
Genomic DNA Isolation-The distal and proximal ends of the left femur were removed, and bone marrow cells were flushed out completely with PBS. The surfaces of the bone shafts were scraped with a scalpel to remove the periosteum and then cut into a few small pieces. Bone pieces were decalcified in 14% EDTA for 1 week. Soft tissues were dissected from animals, frozen immediately in liquid nitrogen, and stored at Ϫ80°C. Decalcified bone and soft tissues were digested with proteinase K (0.5 mg/ml in 10 mM Tris, pH 8.0, 100 mM NaCl, 20 mM EDTA, and 1% SDS) at 55°C overnight. Genomic DNA was then isolated by phenol/chloroform extraction and ethanol precipitation. Two custom Taqman assays were designed for quantifying Tnfsf11 gene deletion efficiency: one specific for sequences between the loxP sites (forward, 5Ј-GCCAGTGGA-CTTACTCAAACCTT-3Ј; reverse, 5Ј-GGTAGGGTTCAAC-TGAAGGGTTTA-3Ј; probe, 5Ј-FAM-CCTCCTCCTCATG-  and the other specific for sequences downstream from the 3Ј loxP site (forward, 5Ј-GGTGCCGTG-CATTATCCTAGAC-3Ј; reverse, 5Ј-AAGTAATGTGACCC-TTGGAGAACTG-3Ј; probe, 5Ј-FAM-CTAGCACACGTG-CCTGCTNFQ-3Ј).

Osteocyte RANKL, B Cells, Estrogen, and Bone Loss
Histology-Sections for histomorphometry were obtained from lumbar vertebrae (L1-L3) by fixing them for 24 h in Millonig's 10% formalin followed by decalcification in 14% EDTA, pH 7.1, for 1 week, dehydration, paraffin-embedding, and cutting 5-m longitudinal sections. After removal of paraffin and rehydration, the sections were stained for tartrate-resistant acid phosphatase (TRAP) activity and counterstained with methyl green. The number and the surface of TRAP-positive cells on the cancellous perimeter (osteoclast number and surface) were measured using an Olympus BX53 microscope and Olympus DP73 camera (Olympus Corp.) interfaced with a digitizer tablet with Osteomeasure TM software version 4.1.0.2 (OsteoMetrics Inc). Histomorphometry measurements were made in a blinded fashion. The terminology used in this study has been recommended by the Histomorphometry Nomenclature Committee of the American Society for Bone and Mineral Research (54). Sections for lineage-tracing studies were obtained from lumbar vertebrae by fixing overnight in 4% paraformaldehyde at 4°C followed by decalcification in 14% EDTA for 1 week and immersion in 30% sucrose for a minimum of 1 day. Bones were then embedded in Cryo-Gel (Electron Microscopy Sciences, Hatfield, PA) for frozen sectioning. 4-m sections were obtained using a Leica cryostat with tape transfer system (Leica Microsystems LM3050S, Buffalo Grove, IL). The sections were imaged on the Olympus BX53 microscope using a DAPI/FITC/Texas Red triple filter cube. The same sections were subsequently stained for TRAP activity and counterstained with methyl green, after which brightfield images were obtained.
Isolation of B Cells Using Magnetic Beads-B cells were isolated with a magnetic cell isolation system (MACS; Miltenyi Biotec, Bergisch Gladbach, Germany). After the column was flushed with ice-cold PBS, 1% BSA, cells labeled with anti-CD19 antibody-conjugated magnetic beads were applied to the column. Unbound cells were washed out with 3 column volumes of PBS, 1% BSA. The column was further washed with 3 volumes of PBS, 1% BSA. After the steel wool column was removed from the external magnetic field, bound cells were eluted and analyzed by flow cytometry.
Osteoclast Culture-BMMs were prepared as described previously (18). Briefly, bone marrow cells were isolated from tibia and femur by flushing with PBS. Bone marrow cells were plated in ␣-10 medium (␣-minimum Eagle's medium, 10% heat-inactivated FBS, 1% penicillin-streptomycin-L-glutamine solution), and 1 ⁄ 10 volume of CMG 14 -12 (conditioned medium supernatant containing recombinant M-CSF at 1 mg/ml) (38) in Petri dishes. Cells were cultured for 4 -5 days with fresh media, and CMG 14 -12 supernatant was replaced every other day. To generate osteoclasts, BMMs were harvested and re-plated in medium containing 1 ⁄ 100 volume of CMG 14 -12 supernatant and 100 ng/ml concentrations of recombinant RANKL and cultured until large multinucleated cells formed (4 -6 days). For osteoclastogenesis using CD19ϩ cell preparations, CD19ϩ and CD19Ϫ cells were isolated using magnetic beads then cultured in the presence of 1 ⁄ 100 volume CMG 14 -12 supernatant and 100 ng/ml RANKL. Osteoclast formation was detected after 5 or 8 days.
Statistics-Data were analyzed using SigmaStat (SPSS Science, Chicago, IL). Two-way analysis of variance was used to detect statistically significant treatment effects after determining that the data were normally distributed and exhibited equivalent variances. In some cases log or ranks transformations were used to obtain normally distributed data and equal variance. This was followed by all pairwise comparisons using Tukey's procedure. For experiments involving comparison of only two groups, Student's t test or two-sample test with multiple comparisons was used. p values less than 0.05 were considered as significant. All quantitative results are presented as box and whisker plots in which the boundary of the box closest to zero indicates the 25th percentile, the line within the box marks the median, and the boundary of the box farthest from zero indicates the 75th percentile. Whiskers (error bars) above and below the box indicate the 90th and 10th percentiles, and values outside the 90th and 10th percentiles are shown as individual points.