Physiological Functions of Osteoblast Lineage and T Cell–Derived RANKL in Bone Homeostasis

Authors


ABSTRACT

The cytokine RANKL is essential for osteoclast development in bone. The cellular sources of RANKL for support of osteoclast generation under various pathophysiological conditions have remained unclear, however. Here we show that inactivation of Rankl specifically in osteoblast lineage cells of mice with the use of an Osterix-Cre transgene results in typical osteopetrosis in the trabecular compartment of the tibia, with the phenotype being progressively less marked in the femur and vertebrae. In contrast to its effects on trabecular bone, RANKL deficiency in osteoblast lineage resulted in thinning of the femoral cortex in association with suppression of bone formation during the modeling process. Ablation of RANKL specifically in T cells resulted in a moderate but significant increase in tibial trabecular bone. Mice with RANKL deficiency in osteoblast lineage were protected from bone loss induced by ovariectomy as well as from joint destruction associated with arthritis, whereas loss of RANKL in T cells did not confer such protection. Finally, inducible deletion of Rankl selectively in the osteoblasts from 6 to 12 weeks of age resulted in an increase in bone mass in association with reduced bone resorption and formation. Our results thus suggest that RANKL produced by osteoblasts contributes to osteoclast development in vivo. © 2014 American Society for Bone and Mineral Research.

Introduction

Osteoclast-mediated bone resorption serves important physiological functions in skeletal and calcium homeostasis as well as in hematopoiesis.[1] Although the origin of osteoclasts was long obscure, they are now thought to arise from hematopoietic precursors.[2, 3] Establishment of a coculture system for osteoclast formation ex vivo led to the notion that direct contact with stromal or osteoblastic cells is required for hematopoietic cells of the monocyte-macrophage lineage to differentiate into osteoclasts, with the hypothetical molecule (or molecules) provided by the former support cells being termed osteoclast differentiation factor (ODF).[4] Receptor activator of NF-κB ligand (RANKL) was subsequently identified as ODF, along with a counteracting inhibitory molecule termed osteoprotegerin (OPG).[5]

RANKL is a type II transmembrane protein that belongs to the TNF family and performs various physiological functions related to such diverse processes as immunity, hematopoiesis, lymphoid organ and mammary gland development, and control of body temperature.[6] Mice with targeted disruption of the RANKL or RANK genes, as well as patients with mutations in RANKL, manifest pronounced osteopetrosis and a complete absence of osteoclasts,[7-9] demonstrating definitively that the RANKL-RANK system is essential for osteoclast formation in vivo.

Although RANKL has been established as an essential cytokine for osteoclast development and has emerged as a key target for pharmacological intervention in bone disorders,[10, 11] its cellular sources in the bone microenvironment and their relative contributions to osteoclastogenesis and skeletal homeostasis in vivo have remained unclear.[12] Two recent studies suggested that matrix-embedded osteocytes are the major source of RANKL for regulation of osteoclastogenesis.[13, 14] To address this important issue, we have generated a mouse model that allows somatic inactivation of Rankl in distinct cell types and examined the pathophysiological functions of osteoblast lineage RANKL in bone remodeling and modeling.

Materials and Methods

Mice

A BAC clone (RP22-429A3) containing Rankl was obtained from a 129S6/SvEvTac mouse genomic library. The LoxP-FRT-PGKneo-FRT cassette and a LoxP site were introduced into introns 2 and 4, respectively, of Rankl in RP22-429A3 with the use of Red/ET recombination technology in Escherichia coli.[15] The linearized targeting vector was introduced into 129Sv/EvTac ES cells by electroporation, and recombinant colonies were selected with G418. Targeted ES clones were screened by Southern blot analysis of Bst1107I- or KpnI-digested DNA with the use of 5' and 3' probes, respectively, and were then injected into C57BL/6 mouse blastocysts. F1 heterozygotes were crossed with CAG-FLPe (flippase, or FRTase) transgenic mice (RIKEN Bioresource Center, Tsukuba, Japan),[16] and the Ranklflox mice were backcrossed onto the C57BL/6J background for eight generations. Mice hemizygous for the Osterix-GFP::Cre transgene were obtained from The Jackson Laboratory (Bar Harbor, ME, USA),[17] CAG-Cre mice[18] from RIKEN Bioresource Center, and CD4-Cre mice from Taconic (Hudson River Valley, NY, USA).[19] KRN TCR transgenic mice[20] were obtained from K. Omura (Kyoto University, Japan) with permission from D Mathis and C Benoist (Harvard Medical School, Boston, MA, USA), and NOD mice were obtained from Clea Japan (Tokyo, Japan).

Mice were raised under standard laboratory conditions at 24°C ± 2°C and 50% to 60% humidity, and they were allowed free access to tap water and standard rodent chow (CE-2, Clea Japan) containing 1.20% calcium, 1.08% phosphate, and vitamin D3 (240 IU per 100 g). Global and osteoblast lineage–specific Rankl knockout mice, which developed osteopetrosis with impaired tooth eruption, were fed with powdered chow. OVX or sham surgery was performed at 12 weeks of age. Experiments were performed with male mice unless indicated otherwise. All experiments were performed in accordance with NCGG ethical guidelines for animal care, and the experimental protocols were approved by the animal care committee.

Inducible deletion of Rankl

To suppress expression of the Cre transgene in Ranklflox/flox;Osterix-Cre offspring, we provided dams with chow (Labo MR stock; Nosan Corporation, Yokohama, Japan) containing doxycycline hyclate (LKT Laboratories, St. Paul, MN, USA) at 1 g/kg as well as drinking water containing Dox (2 g/L) and sucrose (5%) beginning at least 3 days before breeding. For deletion of Rankl postnatally, chow and drinking water were switched to those without Dox within 24 hours after birth of the offspring, or those of the offspring were so switched at 6 weeks of age.

K/BxN serum transfer model

KRN TCR transgenic mice were mated with NOD mice to obtain K/BxN mice. Transgene-negative littermates served as controls. Serum from mice of both genotypes was obtained at 9 weeks of age and pooled separately. Arthritis was induced by ip injection of serum (200 µL) on days 0, 2, 7, and 14 beginning at 6 weeks of age, and recipients were killed at 21 days according to the reported method.[21] Clinical signs of arthritis were scored at all four paws on a scale of 0 to 3 according to a standard protocol.[22] Ankle thickness was measured with the use of a Digimatic Indicator (Mitutoyo, Kawasaki, Japan). For histological analysis, ankle joints were fixed in 4% PFA for 12 hours and then decalcified in 10% EDTA for 2 to 4 weeks. Serial sagittal sections (5 µm) were cut and stained with H&E, toluidine blue, or TRAP.

Cell isolation and culture

Osteoblasts were isolated from newborn mouse calvaria as described previously,[23] and adherent bone marrow macrophages (BMMs) were used as osteoclast precursors, as previously described.[24] Coculture of calvaria-derived osteoblasts with BMMs was performed in the presence of 10 nM 1α,25-(OH)2D3 (Nacalai Tesque, Kyoto, Japan), according to the established technique.[25] CD4-positive cells were purified with the use of MACS Micro Beads (Miltenyi Biotec K.K., Tokyo, Japan).

Osteoblast- and osteocyte-rich fractions were isolated from the tibia and femur. After removal of surrounding soft tissue, both ends of each bone were cut and BM cells were flushed out with PBS. The remaining diaphysis was cut longitudinally, and the cells on the endocortical surface were collected as an osteoblast-rich fraction by brushing with a toothbrush in the presence of the TRIzol reagent (Invitrogen, San Diego, CA, USA) and were then incubated with rotation for 1 hour at room temperature in the presence of TRIzol for extraction of total RNA. The residual bone pieces were crushed in liquid nitrogen to yield an osteocyte-rich fraction, from which total RNA was isolated with the use of TRIzol.

Gene expression analysis

Total RNA was isolated from cells and tissues with the use of the TRIzol reagent, purified with an RNeasy Mini Kit (Qiagen, Valencia, CA, USA), and subjected to RT with a high-capacity cDNA RT kit (Applied Biosystems, Carlsbad, CA, USA). The resulting cDNA was subjected to quantitative PCR analysis with the use of PowerSYBR Green PCR master mix and an ABI7300 real-time PCR system (Applied Biosystems). The primers used for PCR are summarized in Supplemental Table S1.

Blood biochemistry

Serum osteocalcin (Ocn) and carboxyl-terminal collagen crosslinks (CTX) concentrations were determined with the use of a mouse Ocn enzyme immunoassay kit (Biomedical Technologies, Stoughton, MA, USA) and a RatLaps enzyme immunoassay kit (Immunodiagnostic Systems, Fountain Hills, AZ, USA), respectively.

Bone analysis

Micro-computed tomography (CT) scanning was performed with the use of a µCT-40 device (SCANCO Medical, Bassersdorf, Switzerland) at a resolution of 12 µm, and 3D microstructural parameters were calculated as described previously.[26] Bone samples were fixed in 70% ethanol for determination of histomorphometric parameters with nondecalcified sections at the Ito Bone Science Institute (Niigata, Japan); double labeling with calcein and tetracycline was performed by consecutive administration of the labels sc with a 2-day interval. Frozen sections (10 µm) of the femur were fixed with 2% PFA for staining with a TRAP stain kit (Wako, Osaka, Japan) and counterstaining with methyl green (Muto Pure Chemicals, Tokyo, Japan). The nomenclature for micro-CT and histomorphometry followed the recommendations of the published guidelines.[27, 28]

Statistical analysis

Quantitative data are presented as means ± SD and were analyzed with Student's t test. A p value <0.05 was considered statistically significant.

Results

Generation of mice with a floxed Rankl allele

To achieve cell type–specific deletion of the RANKL gene in mice, we introduced LoxP sites flanking exons 3 and 4 of the gene (Supplemental Fig. S1A). Southern blot hybridization and PCR analysis of genomic DNA confirmed successful recombination with integration of a neo cassette (Supplemental Fig. S1B), deletion of the neo cassette and generation of the floxed Rankl allele after crossing of mice harboring the neo allele with flippase deleter mice (Supplemental Fig. S1C), deletion of Rankl exons 3 and 4 after crossing of Ranklflox/+ mice with CAG-Cre mice (Supplemental Fig. S1D), which express Cre recombinase systemically,[18] and generation of the +/+, +/−, and −/− genotypes of Rankl (Supplemental Fig. S1E).

The resultant global knockout (Rankl−/−, hereafter designated gKO) mice exhibited a reduced body size (Supplemental Fig. S1F), typical osteopetrosis including the absence of tooth eruption (Supplemental Fig. S1G), BM cavities, and osteoclasts (Supplemental Fig. S1H, data not shown), as well as a lack of lymph nodes (Supplemental Fig. S1I). These characteristics are identical to those described for mice with germline inactivation of Rankl,[7] indicating that the floxed allele functions as designed. We further confirmed that primary osteoblasts isolated from the calvaria of Rankl gKO mice failed to support the formation of tartrate-resistant acid phosphatase (TRAP)–positive osteoclasts in ex vivo cocultures with bone marrow macrophages (BMMs) from WT mice in the presence of 1α,25-dihydroxyvitamin D3 [1α,25-(OH)2D3] (Supplemental Fig. S1J).

Osteoblast lineage–specific deletion of Rankl

To achieve osteoblast lineage–specific inactivation of Rankl, we crossed Ranklflox/+ mice with Osterix-Cre mice, in which Cre recombinase is expressed specifically in osteoblasts.[17] We then intercrossed the resulting progeny to yield Ranklflox/flox;Osterix-Cre (hereafter designated ΔOb/ΔOb) mice. The ΔOb/ΔOb animals showed a markedly reduced level of Rankl mRNA in bone but not in other tissues including spleen, thymus, and lymph nodes (Fig. 1A). Very few TRAP-positive cells were detected on the surface of either trabecular or cortical bone in the mutant mice, whereas TRAP staining in the primary spongiosa area immediately above the growth plate was similar to that in control mice (Fig. 1B). The serum concentrations both of CTX, a marker of bone resorption, and of Ocn, a product of osteoblasts, were markedly reduced in ΔOb/ΔOb mice compared with those in control mice (Fig. 1C). These observations thus suggested that deletion of Rankl in osteoblast lineage cells results in a pronounced decrease both in Rankl expression in bone and in the number of TRAP-positive osteoclasts as well as in reduced bone turnover.

Figure 1.

Osteopetrosis in osteoblast lineage–specific Rankl knockout (ΔOb/ΔOb) mice. (A) Quantitative RT-PCR analysis of Rankl mRNA in various tissues of control (Ranklflox/flox or Ranklflox/+), ΔOb/+, and ΔOb/ΔOb mice at 8 weeks of age. LN = lymph nodes. Data are expressed relative to the corresponding value for control mice and are means ± SD for 5 mice of each group. **p < 0.01. (B) Representative TRAP staining of sections obtained from the femur of 8-week-old control (Ranklflox/flox) and ΔOb/ΔOb mice. Scale bar = 1 mm (middle panels). Closed arrowheads indicate TRAP-positive cells, which are apparent on the surface of both trabecular and cortical bone in the control mouse and immediately above the growth plate in both genotypes; open arrowheads indicate the absence of such cells at the bone surface in the mutant. (C) Serum concentrations of CTX and Ocn in 8-week-old mice of the indicated genotypes. Data are means ± SD for 7 mice of each group. ***p < 0.001, **p < 0.01. (D) Representative micro-CT images of the tibia, femur, and third lumbar vertebra of control (Ranklflox/flox), ΔOb/+, ΔOb/ΔOb, Rankl+/−, and Rankl−/− mice at 8 weeks of age. (E) Quantitation of the 3D bone volume fraction (BV/TV) by micro-CT imaging as in (D). cKO = conditional knockout; gKO = global knockout. Data are means ± SD for 5 mice of each group. ***p < 0.001, **p < 0.01, *p < 0.05. NS: not significant.

Skeletal site–specific requirement for RANKL produced by the osteoblast lineage

To characterize further the skeletal phenotype of Rankl ΔOb/ΔOb mice, we analyzed trabecular structure as well as bone volume in the tibia, femur, and third lumbar vertebra by micro-CT scanning in comparison with control and gKO mice. Three-dimensional micro-CT images (Fig. 1D) as well as quantitation of the bone volume fraction (bone volume/tissue volume, or BV/TV) (Fig. 1E) indicated that the osteopetrotic phenotype of ΔOb/ΔOb mice was similar to that of gKO mice for the tibia but was less pronounced for the femur and even milder for the spine. The extent of the reduction in Rankl expression in ΔOb/ΔOb mice did not differ among these three skeletal sites, however (Fig. 1A). These data suggested that, although RANKL produced by osteoblast lineage cells plays an important physiological role in osteoclast development, its requirement differs among skeletal sites. Osteoblast lineage cells are thus an essential source of RANKL for osteoclast formation in appendicular bone such as the tibia, whereas RANKL produced by these cells is largely dispensable in vertebrae, with other cell types possibly compensating for RANKL deficiency in osteoblasts. The bone volume fraction for the tibia and femur was significantly increased in global Rankl heterozygous (Rankl+/−) mice as well as in the tibia of ΔOb/+ mice (Fig. 1E), indicative of a gene-dosage effect.

Bone histomorphometry at the tibial metaphysis of 8-week-old ΔOb/ΔOb mice with normalization of parameters by bone surface (BS) revealed that osteoclast surface (Oc.S) and number (N.Oc) as well as eroded surface (ES) were significantly decreased, and that osteoid surface (OS), osteoblast surface (Ob.S) and number (N.Ob), and bone formation rate (BFR), all of which are histological indices of bone formation, were also markedly decreased, compared with control mice (Fig. 2A). These data suggested that osteoclastic bone resorption is suppressed in ΔOb/ΔOb mice, and that, given that bone resorption is normally coupled to subsequent bone formation in each remodeling cycle, bone formation is also suppressed in ΔOb/ΔOb mice as a result of the suppression of preceding bone resorption. However, the parameters of bone formation in ΔOb/+ mice were significantly reduced to a level intermediate between those of control and ΔOb/ΔOb mice, whereas the parameters of bone resorption were not significantly altered in the heterozygotes (Fig. 2A).

Figure 2.

Reduced bone resorption and formation in osteoblast lineage–specific Rankl knockout (ΔOb/ΔOb) mice. The tibial metaphysis (A) and second lumbar vertebra (B) of 8-week-old mice of the indicated Rankl genotypes were subjected to histomorphometric analysis. Osteoclast surface (Oc.S), number of osteoclasts (N.Oc), eroded surface (ES), osteoid surface (OS), osteoblast surface (Ob.S), number of osteoblasts (N.Ob), and bone formation rate (BFR) were corrected for bone surface (BS). Data are means ± SD for 4 mice of each genotype. ***p < 0.001, **p < 0.01, *p < 0.05.

Furthermore, whereas the reductions in bone resorption indices for the second lumbar vertebra of ΔOb/ΔOb mice were similar to or less pronounced than those for the tibia, the suppression of bone formation at this site was marked and even greater than that apparent in the tibia (Fig. 2B). Again, significant suppression of bone formation was also apparent in the lumbar vertebra of ΔOb/+ mice, whereas the histological indices of osteoclastic bone resorption were little changed (Fig. 2B). These results thus suggested that RANKL in osteoblast lineage cells may play a role in the coupling of bone formation to preceding bone resorption, in addition to its function in osteoclastic bone resorption per se.

Role of RANKL expressed by the osteoblast lineage in cortical bone modeling

In contrast to the markedly increased bone mass in the trabecular compartment, analysis of femoral cortical bone by micro-CT scanning revealed that the cortex was thinner in ΔOb/ΔOb mice than in control mice (Fig. 3A). Cross-sectional views of the femoral diaphysis showed the contrast between the thin cortex and increased trabecular bone in ΔOb/ΔOb mice (Fig. 3B). Cortical thickness and cortical bone volume, as determined by micro-CT scanning, were significantly decreased in ΔOb/ΔOb mice (Fig. 3C).

Figure 3.

Role of RANKL produced by the osteoblast lineage in cortical bone modeling. (A) Representative micro-CT images of the femur of control (Ranklflox/flox) and Rankl ΔOb/ΔOb mice at 8 weeks of age. Note the thin cortex of the femur in the mutant. (B) Cross-sectional micro-CT images of the femoral diaphysis. Note the thin cortex and markedly increased trabecular bone in the ΔOb/ΔOb mouse. (C) Quantitation of cortical thickness (C.Th) and the cortical bone volume fraction (C.BV/TV) by micro-CT. Data are means ± SD for 5 mice of each group. ***p < 0.001, **p < 0.01, *p < 0.05. (D) Orientation of the femur for histomorphometric analysis of cortical bone. Note the labeling on the periosteal surface (posterior side) and endocortical surface (anterior side) at the midshaft, indicating that the bone is drifting posteriorly. The boxed regions in the left panel are shown at higher magnification in the right panels. Arrows show that the gap between the calcein and tetracycline labels is narrowed in the mutant. (E) Histomorphometric parameters for cortical bone of the femur. Data are means ± SD for four mice of each group. *p < 0.05.

To delineate the physiological function of RANKL produced by osteoblast lineage cells in cortical bone modeling, where either bone formation or resorption predominates at opposite bone surfaces, we performed detailed histomorphometric analysis at the periosteal and endocortical surfaces of the diaphyseal femoral cortex (Fig. 3D). In control mice, bone labeling was apparent on the periosteal surface of the posterior midshaft and on the endocortical surface of the anterior midshaft, indicating that the femoral bone was drifting posteriorly at the midshaft (Fig. 3D). In contrast, such labeling was weak in the cortical bone of ΔOb/ΔOb mice, and the gap between the calcein and tetracycline labels was narrow (Fig. 3D), suggesting that bone apposition is suppressed in the absence of RANKL produced by the osteoblast lineage.

These results were further substantiated by quantitation of histomorphometric parameters at the femoral cortex (Fig. 3E). The periosteal surface of the posterior midshaft and the endocortical surface of the anterior midshaft were exclusively in the bone formation phase, as reflected by a value of almost 100% for OS/BS and 0% for ES/BS, in both control and ΔOb/ΔOb mice. However, BFR/BS was reduced significantly in both of these bone regions in the mutant. The fact that the loss of osteoblast lineage RANKL resulted in suppression of bone formation on the bone surface, where normally only bone apposition takes place without preceding bone resorption, suggests that RANKL produced by osteoblast lineage cells determines the direction and extent of cortical bone modeling through dynamic regulation of the balance between bone resorption and formation on opposite sides, or that RANKL somehow regulates the bone-forming activity of osteoblasts independently of its effect on osteoclastogenesis.

Role of RANKL produced by T cells

It has been shown that activated T cells express RANKL and not only induce osteoclast formation in vitro but also promote systemic bone loss in association with an increase in osteoclast number and local joint destruction in vivo.[29] To achieve T-cell–specific inactivation of Rankl and further study its physiologic functions, we crossed mice harboring the floxed Rankl allele with CD4-Cre mice, in which Cre recombinase is specifically expressed in CD4+ T cells.[19] The resulting homozygous mutant (ΔT/ΔT) mice manifested a markedly reduced abundance of Rankl mRNA in the thymus and spleen, but not in bone or lymph nodes, compared with control mice (Supplemental Fig. S2A). Furthermore, the amount of Rankl mRNA was significantly reduced in CD4+ cells, but not in CD4 cells, isolated from the thymus or spleen of ΔT/ΔT mice (Supplemental Fig. S2B). Micro-CT imaging revealed that young (5 to 8 weeks of age) ΔT/ΔT male mice manifested a moderate but significant increase in 3D bone volume and trabecular connectivity at the proximal tibia (Supplemental Fig. S2C, E; data not shown) but not at vertebrae (Supplemental Fig. S2D, F). Young (5-week-old) female ΔT/ΔT mice also exhibited a significant increase in bone mass at the proximal tibia, although the phenotype gradually diminished with advancing age (16 weeks of age, data not shown) in both males and females. These results thus suggested that RANKL produced by T cells plays a physiological role in basal bone metabolism at least in mice at younger ages.

Role of RANKL in estrogen deficiency–associated bone loss and in RA

To determine the relative contributions of RANKL expression in the osteoblast lineage and in T cells to bone loss induced by estrogen deficiency, we subjected Rankl ΔObOb and ΔTT female mice to OVX. Whereas control mice manifested a significant decrease in trabecular bone volume at the tibial metaphysis after OVX, ΔObOb mice were resistant to such bone loss (Fig. 4A, B). The serum concentration of CTX was also increased significantly after OVX in control mice but not in ΔObOb mice (Fig. 4C). In contrast, the extent of bone loss and the serum concentration of CTX after OVX did not differ significantly between ΔTT and control mice (Fig. 4D–F). These results indicated that RANKL produced by osteoblast lineage cells, but not that produced by T cells, plays a major role in the development of osteopenia associated with estrogen deficiency.

Figure 4.

Mice with osteoblast lineage–specific deficiency of RANKL are resistant to bone loss after OVX. (A–C) Representative micro-CT images (A) and the bone volume fraction (B) for the proximal tibia as well as the serum CTX concentration (C) for control (Ranklflox/flox or Ranklflox/+) and Rankl ΔObOb mice at 16 weeks of age after OVX or sham surgery at 12 weeks of age. Data in (B) and (C) are means ± SD for 12 and 10 mice of each group, respectively. **p < 0.01, *p < 0.05. (D–F) Representative micro-CT images (D) and the bone volume fraction (E) for the proximal tibia as well as the serum CTX concentration (F) for control (Ranklflox/flox) and Rankl ΔTT mice at 16 weeks of age after OVX or sham surgery at 12 weeks of age. Data in (E) and (F) are means ± SD for 12 and 10 mice of each group, respectively. ***p < 0.001, **p < 0.01.

We next examined the role of RANKL produced by the osteoblast lineage in RA with the use of the K/BxN serum transfer model.[30] Deletion of Rankl in the osteoblast lineage had no discernible effect on ankle swelling, the clinical score of arthritis, or the loss of articular cartilage (Fig. 5A–C) in this model, but it partially ameliorated bone erosion (Fig. 5D), development of TRAP-positive osteoclasts (Supplemental Fig. S3), and bone loss (Fig. 5E, F) in the periarticular region. In contrast, ΔTT and control mice showed no difference not only in ankle swelling, the clinical score of arthritis, and destruction of articular cartilage (Supplemental Fig. S4A–C) but also in bone erosion and periarticular osteopenia (Supplemental Fig. S4D, E; data not shown). These results thus indicated that RANKL produced by osteoblast lineage cells, but not that expressed in T cells, contributes to the pathogenesis of inflammation-induced joint destruction.

Figure 5.

Mice with osteoblast lineage–specific deficiency of RANKL are partially protected from joint destruction and periarticular bone loss associated with RA. (A, B) Representative images of the left hind paw at 21 days (A) as well as the arthritis score and hind paw thickness at various times (B) after injection of control (Ranklflox/flox or Ranklflox/+) or Rankl ΔObOb mice with control or arthrogenic K/BxN mouse serum beginning at 6 weeks of age. Data in (B) are means or means ± SD for 5 mice of each group. (C–F) Representative toluidine blue staining (C), hematoxylin and eosin staining (D), and micro-CT images (E) as well as the bone volume fraction (F) for the distal tibia of mice treated as in (A) at 21 days after serum injection. Arrows in (D) indicate bone erosion in the control mouse injected with arthrogenic serum and the corresponding protected region of the mutant, whereas those in (C) indicate no difference in the loss of articular cartilage. Scale bars in (C) and (D) = 100 and 500 µm, respectively. Data in (F) are means ± SD for 5 mice in each group. *p < 0.05.

Inducible ablation of RANKL in adult bone

Finally, taking advantage of the tetracycline sensitivity of the Osterix-Cre construct,[17] we examined the role of RANKL produced by the osteoblast lineage in the remodeling of adult bone by inducing the inactivation of Rankl postnatally. To suppress Cre-mediated recombination at the Rankl locus, we administered doxycycline (Dox) in drinking water and chow both to dams beginning before mating and then to their offspring after weaning. Ranklflox/flox;Osterix-Cre mice born to dams fed Dox manifested normal tooth eruption and no sign of osteopetrosis on X-rays (data not shown), and the abundance of Rankl mRNA in bone was similar to that for control mice (Fig. 6A). Micro-CT analysis of the tibial metaphysis at 12 weeks of age also revealed that Dox treatment suppressed the development of osteopetrosis in Ranklflox/flox;Osterix-Cre mice, although the bone volume fraction of these animals was slightly greater than that of control Ranklflox/flox mice supplied with Dox (Fig. 6B, C), possibly as a result of “leaky” expression of Cre even in the presence of Dox. These data indicated that the expression of Cre, and therefore that of Rankl, in osteoblast lineage cells can be controlled through Dox administration.

Figure 6.

Inducible inactivation of Rankl in the osteoblast lineage of adult mice. (A) Quantitative RT-PCR analysis of Rankl mRNA in the tibia of 12-week-old Ranklflox/flox;Osterix-Cre (F/F;Cre) mice subjected to continuous Dox administration (Dox+) or to discontinuation of Dox treatment at 6 weeks of age (off [6–12 weeks]) or at birth (off [0–12 wk]). Ranklflox/flox mice subjected to continuous Dox administration served as controls. Data are expressed relative to the value for control mice and are means ± SD for 4 mice of each group. **p < 0.01, *p < 0.05. (B) Representative micro-CT images of the tibial metaphysis for mice as in (A). (C) Micro-CT analysis of the bone volume fraction and 3D trabecular structure for mice as in (A). Conn-Dens = connectivity density; SMI = structure model index; Tb.N, Tb.Th, and Tb.Sp = trabecular number, thickness, and separation, respectively. Data are means ± SD for 4 mice of each group. ***p < 0.001, **p < 0.01, *p < 0.05. (D) Serum concentrations of CTX and Ocn for mice as in (A). Data are means ± SD for 5 mice of each group. ***p < 0.001, **p < 0.01. (E) Quantitative RT-PCR analysis of Rankl and Opg mRNAs in osteocyte and osteoblast fractions isolated from the tibia and femur of C57BL/6J mice at the indicated ages. Data are expressed relative to the corresponding value for the osteocyte fraction from 2-month-old mice and are means ± SD for three mice of each group. **p < 0.01, *p < 0.05. (F) Quantitative RT-PCR analysis of Rankl mRNA in osteocyte and osteoblast fractions isolated from the tibia and femur of 12-week-old Ranklflox/flox;Osterix-Cre (F/F;Cre) mice subjected to continuous Dox administration (Dox+) or to discontinuation of Dox treatment at 6 weeks of age (off [6–12 wk]) as in (A). Data are expressed relative to the value for the osteocyte fraction in control mice and are means ± SD for 5 mice of each group. **p < 0.01, *p < 0.05.

To examine the effect of RANKL depletion in adult bone, we terminated Dox administration at 6 weeks of age and maintained the mice without Dox for the subsequent 6 weeks. The amount of Rankl mRNA in bone of such Ranklflox/flox;Osterix-Cre mice at 12 weeks of age had decreased significantly compared with that for control animals maintained with continuous Dox administration (Fig. 6A). We reasoned that, within a certain period of time, Rankl inactivation would result in preferential depletion of RANKL in osteoblasts on the bone surface, whereas Rankl expression would be preserved in matrix-embedded osteocytes. Indeed, the bone volume fraction for the tibial metaphysis at 12 weeks of age in the mice subjected to inducible Rankl inactivation at 6 weeks was significantly greater than that in those subjected to continuous Dox administration (Fig. 6B, C). This increased bone mass was associated with an altered trabecular microstructure, as revealed by a decreased structure model index (reflecting an increased prevalence of platelike structure relative to rodlike structure) and increased trabecular thickness (Fig. 6C). Inactivation of Rankl immediately after birth by discontinuation of Dox administration resulted in an even greater increase in the bone volume fraction at 12 weeks of age compared with that apparent after discontinuation of Dox treatment at 6 weeks of age (Fig. 6B, C). However, the shorter period of RANKL depletion was sufficient to suppress biochemical parameters of bone resorption and formation to an extent similar to that apparent for mice subjected to Dox termination for 12 weeks (Fig. 6D). Collectively, these data suggested that timed inhibition of RANKL expression in osteoblasts of mice between 6 and 12 weeks of age results in increased bone mass in association with suppression of bone resorption and turnover, and that RANKL produced by osteoblasts is also responsible for bone homeostasis.

To assess the relative levels of Rankl expression within the osteoblast lineage, we processed bone specimens obtained from WT mice at 2, 4, or 6 months of age to obtain fractions consisting mostly of osteoblasts from the bone surface or of osteocytes that were embedded within the bone. The osteoblast-rich fraction was found to preferentially express osteoblastic marker genes such as Runx2, Sp7 (Osterix), Col1a1, and Alp, whereas the osteocyte-rich fraction preferentially expressed osteocyte marker genes including Sost, Mepe, Fgf23, and Dmp1 (Supplemental Fig. S5). The expression level of Dmp1 was actually similar in the two fractions from mice at 2 months of age, but it was higher in the osteocyte-rich fraction at 4 and 6 months of age. In contrast to the previously described preferential expression of Rankl in osteocytes,[14] we found that the amount of Rankl mRNA was substantially greater in the osteoblast-rich fraction than in the osteocyte-rich fraction of mice at 2, 4, and 6 months of age (Fig. 6E). Although the level of Rankl expression declined with age, the osteoblast/osteocyte ratio for expression of this gene remained relatively constant. In contrast, the amount of Opg mRNA was greater in the osteocyte-rich fraction than in the osteoblast-rich fraction, especially for mice at 4 and 6 months of age (Fig. 6E), consistent with the notion that OPG is highly expressed in osteocytes as a result of β-catenin signaling and contributes to the regulation of osteoclastogenesis and bone mass in vivo.[31] Evidently, Dox discontinuation from 6 through 12 weeks caused a significant reduction in Rankl expression in the osteoblast fraction, whereas that in the osteocyte fraction remained unaltered (Fig. 6F), which further supports our contention that osteoblast-derived RANKL also contributes to osteoclastogenesis and bone homeostasis.

Discussion

RANKL has been thought to be produced by and presented at the surface of marrow stromal or osteoblastic cells, with its expression being regulated by osteotropic hormones that stimulate bone resorption, such as 1α,25(OH)2D3 and parathyroid hormone (PTH).[12] The key cellular sources of RANKL for support of osteoclastogenesis under various pathophysiological conditions in vivo have remained unclear, however. Indeed, a recent study suggested that commitment to the osteoblastic lineage is not required for RANKL gene expression, on the basis of the findings that Rankl expression in bone was not affected by osteoblast depletion and was independent of Runx2, an essential gene for osteoblast differentiation.[32] Furthermore, two recent studies have suggested that matrix-embedded osteocytes, not osteoblasts, are the major and essential source of RANKL in bone, with depletion of RANKL specifically in osteocytes resulting in an osteopetrotic phenotype.[13, 14] Our present finding that deletion of Rankl in osteoblast lineage cells with the use of the Osterix-Cre transgene results in osteopetrosis most prominently in appendicular bone such as the tibia is largely consistent with the results of one of these previous studies.[13]

Our findings that even heterozygous deletion of Rankl has a significant impact on bone mass in vivo suggest that osteoclastic bone resorption is sensitive to the level of RANKL expression. Our data also show that the requirement for osteoblast lineage–derived RANKL in osteoclast development varies according to skeletal site. Vertebral bone mass in Rankl ΔOb/ΔOb mice thus did not differ to a great extent from that in control mice, in contrast to the marked osteopetrotic phenotype for vertebrae of Rankl gKO mice. In axial bone, other cell types, such as osterix-negative marrow stromal cells or immune cells other than T cells, may therefore compensate for the lack of osteoblast lineage RANKL by interacting with hematopoietic osteoclast precursor cells.

Osteocytes were recently shown to produce RANKL at higher levels than osteoblasts as well as to induce osteoclast formation ex vivo, most prominently in response to treatment with 1α,25(OH)2D3.[14] In addition, deletion of Rankl predominantly in osteocytes with the use of the Dmp1-Cre deleter mouse was found to induce osteopetrosis.[13, 14] However, the Dmp1-Cre construct also appears to confer Cre expression in osteoblasts at the bone surface, and definitive evidence for a reduction in the level of Rankl expression in bone was not obtained.[13] Moreover, in one of these studies,[14] deletion of Rankl in osteocytes was induced on a germline heterozygous Rankl knockout background, and we show in the present study that Rankl heterozygous knockout (+/−) mice manifest a significant increase in bone volume. The impact of Rankl deletion in osteocytes per se is therefore not clear in a model that already lacks one allele of Rankl globally.[14]

Transient deletion of Rankl in osteoblasts of adult mice by Dox deprivation for 2 months beginning at 4 months of age also failed to reduce osteoclast number and to induce an osteopetrotic phenotype, despite a significant reduction in Rankl expression, leading the authors to conclude that osteoblasts and their progenitors do not contribute to osteoclastic bone resorption.[13] In contrast, with a similar experimental design, we found that deletion of Rankl in the osteoblasts of adult mice for an even shorter period (6 weeks), beginning at 6 weeks of age, resulted in a significant reduction in bone resorption and an increase in bone mass at 3 months of age, consistent with the notion that RANKL provided mainly by osteoblasts contributes to osteoclast development in vivo, at least at a young age. The importance of osteoblasts as a source of RANKL is also supported by our previous observation that ablation of osteocytes is associated with upregulation of Rankl expression in bone and increased bone resorption.[23]

A recent study showed that deletion of the vitamin D receptor (VDR) gene specifically in osteoblasts (not in osteocytes) with the use of a deleter mouse harboring a Cre transgene under the control of a collagen gene promoter resulted in an increase in bone mass together with a reduced number of osteoclasts and level of Rankl expression in bone.[33] These findings suggested that VDR in osteoblasts contributes to osteoclastogenesis in vivo by increasing the production of RANKL in response to 1α,25-(OH)2D3, consistent with our conclusion that osteoblasts are a key source of RANKL in vivo. Furthermore, another study found that mice in which the VDR gene was deleted in osteocytes, with the use of the same Dmp1-Cre construct as used for the deletion of Rankl,[34] manifested normal bone metabolism, did not develop osteopetrosis or a reduced level of Rankl expression in bone, and exhibited a robust response to 1α,25-(OH)2D3 administration with an increase in the number of osteoclasts.[35] Together, these findings suggest that, at least with regard to the VDR-RANKL axis, RANKL provided by osteoblasts, rather than that produced by osteocytes, plays a central role in osteoclastogenesis.

In bone remodeling cycles, bone resorption is tightly coupled to subsequent bone formation, with the result that inhibition of osteoclastic bone resorption is almost always accompanied by suppression of bone formation.[36] Histomorphometric analysis of ΔOb/ΔOb mice in the present study also revealed that suppression of bone formation accompanied inhibition of bone resorption in the remodeling compartment of trabecular bone. However, the suppression of bone formation induced by ablation of RANKL in the osteoblast lineage was more pronounced than was that of bone resorption. This difference was most evident in vertebrae, in which the dependence of osteoclastic bone resorption on osteoblastic RANKL appeared to be lowest and in which substantial suppression of bone formation was observed in association with only a relatively small impact on osteoclastic activity. Suppression of bone formation was also evident in mice subjected to transient (6-week) depletion of RANKL in the osteoblasts, as reflected by a substantial decrease in the serum Ocn concentration. RANKL produced by osteoblasts may thus contribute to the coupling of bone formation to preceding resorption, in addition to its role in the development of osteoclasts per se.

In cortical bone of the femur, RANKL depletion in osteoblast lineage cells resulted in a reduction in the level of bone-forming activity on the bone surface, where bone formation proceeds with no preceding osteoclast activity. Several possibilities might account for this observation. First, expansion of trabecular bone inside the femur of ΔOb/ΔOb mice may result in changes to the metabolic as well as mechanical environment of the organ as a whole, leading to impairment of periosteal bone apposition (resulting, for example, from a lack of calcium available for bone formation) and cortical thinning. Second, inhibition of osteoclastic bone resorption on the endocortical surface might influence bone-forming activity on the periosteal surface, possibly through the intracortical osteocyte network. It will be of interest to determine whether cortical thinning is similarly observed in association with osteopetrosis in osteocyte-specific Rankl knockout models.[13, 14]

In the present study, we found that deletion of Rankl in T cells with the use of the CD4-Cre deleter mouse resulted in a small but significant increase in bone volume in young male and female mice, in contrast to previous results showing that deletion of Rankl with the use of the Lck-Cre construct generated no obvious phenotype.[14, 37] This apparent discrepancy might be because of a difference in excision efficiency between the two Cre transgenes.[38] However, we did not detect any effect of T cell–specific Rankl deletion on bone loss induced by OVX or on joint destruction in a model of RA, making it unlikely that T cell–derived RANKL plays essential roles in these pathologies. It is worth pursuing the role of T cell–derived RANKL using other mouse models of RA that require T cells, such as CIA.[39]

RANKL produced by osteoblast lineage cells thus appears to play pathogenic roles not only in osteopenia associated with estrogen deficiency but also in joint destruction associated with RA. Whether the estrogen receptor in osteoblasts or osteocytes functions to regulate RANKL expression remains to be determined, as does the relative contribution of osterix-negative cells, specifically synovial fibroblasts, to joint destruction in RA.

In conclusion, our results have revealed a key contribution of RANKL produced by the osteoblast lineage to bone loss associated with estrogen deficiency and RA. Furthermore, in addition to matrix-embedded osteocytes, osteoblasts on the bone surface contribute to osteoclast development, at least in mice at younger ages. The Ranklflox/flox mouse model should provide a tool for further elucidation of the diverse physiological as well as pathological roles of RANKL at the organismal level.

Disclosures

All authors state that they have no conflicts of interest.

Acknowledgments

We thank D Mathis and C Benoist for permission to use the K/BxN mouse, K Omura for the transfer of KRN TCR transgenic mice and for suggestions on the K/BxN model, I Taniuchi and Y Naoe for suggestions on the use of CD4-Cre, M Okabe for permission to use CAG-Cre mice, M Suzuki for technical assistance, and A Ito for suggestions on bone histomorphometry.

This study was supported by JSPS KAKENHI for Scientific Research B (22390064 to KI) and for Young Investigator (21790374 to TF) and by MEXT KAKENHI for Scientific Research on Innovative Areas (22118007 to KI) from the Ministry of Education, Culture, Sports, Science, and Technology of Japan; by a grant from the Promotion of Fundamental Studies in Health Sciences program of the National Institute of Biomedical Innovation (NIBIO) of Japan (06-31 to KI and MI); and by a grant for Longevity Sciences from the Ministry of Health, Labor, and Welfare of Japan (H23-12 to KI and MI).

Authors' roles: Study design: TF and KI. Data acquisition: TF, ST, and MI. Data analysis and interpretation: FT, ST, MI, and KI. Drafting manuscript: FT and KI. Approving final version of manuscript: TF, ST, MI, and KI. KI takes responsibility for the integrity of the data analysis.

Ancillary