Systemic tumor necrosis factor α mediates an increase in peripheral CD11bhigh osteoclast precursors in tumor necrosis factor α–transgenic mice

Authors


Abstract

Objective

To investigate the mechanisms whereby tumor necrosis factor α (TNFα) increases osteoclastogenesis in vivo.

Methods

TNFα-transgenic (TNF-Tg) and wild-type mice injected with TNFα were studied. In vitro osteoclastogenesis assays, monocyte colony-forming assays, and fluorescence-activated cell sorting were performed using splenocytes, peripheral blood mononuclear cells (PBMCs), and bone marrow cells to quantify and characterize osteoclast precursors (OCPs). Etanercept, a TNFα antagonist, was used to block TNFα activity in vivo. The effects of TNFα on proliferation, apoptosis, and differentiation of OCPs were assessed using 5-bromo-2′-deoxyuridine labeling, annexin V staining, and reverse transcriptase–polymerase chain reaction.

Results

OCP numbers were increased 4–7-fold in PBMCs and spleen, but not in bone marrow of TNF-Tg mice. The OCPs in spleen were in the CD11bhigh population and contained both c-Fms− and c-Fms+ cells. The increased number of OCPs correlated with the initiation of detectable TNFα in serum and the onset of inflammatory arthritis in TNF-Tg mice. Etanercept eliminated the increase in peripheral OCPs. TNFα did not affect proliferation, survival, or differentiation of CD11bhigh splenocytes in vivo or in vitro, but caused a rapid increase in CD11b+ cells in blood within 4 hours of a single injection and an accumulation of CD11bhigh OCPs in spleen after 3 days of multiple injections.

Conclusion

Systemic TNFα induces a marked increase in circulating OCPs that is reversible by anti-TNF therapy and may result from their mobilization from bone marrow. Our findings provide a new mechanism whereby TNFα stimulates osteoclastogenesis in patients with inflammatory arthritis, suggesting that CD11b+ PBMCs could be used to evaluate a patient's potential for erosive disease and the efficacy of anti-TNF therapy.

Chronic inflammatory bone diseases, such as rheumatoid arthritis (RA), are accompanied by bone loss around affected joints due to increased osteoclastic resorption. This process is mediated largely by increased local production of proinflammatory cytokines (1, 2). These cytokines may act directly on cells in the osteoclast lineage or indirectly by affecting the production of the essential osteoclast differentiation factor, receptor activator of NF-κB ligand (RANKL), and/or its soluble decoy receptor, osteoprotegerin, by osteoblast/stromal cells (3). TNFα is a major mediator of inflammation; the importance of TNFα in the pathogenesis of various forms of bone loss is supported by several lines of experimental and clinical evidence (4). However, TNFα is not essential for osteoclastogenesis (5), erosive arthritis (6), or osteolysis (7), because these can occur in the absence of TNFα. The critical question of how TNFα increases osteoclastogenesis in vivo remains to be answered.

Osteoclasts are multinucleated cells formed by fusion of mononuclear precursors in the monocyte/macrophage lineage (colony-forming unit–macrophage [CFU-M]). Cell culture techniques (8) and studies of transgenic and knockout mice (9) have advanced our understanding of osteoclastogenesis and established that macrophage colony-stimulating factor (M-CSF) and RANKL are required for osteoclastogenesis (10–15). Osteoclastogenesis is also dependent on intracellular signaling molecules, including the adapter protein TNF receptor–associated factor 6 (16, 17), and the transcription factors activator protein 1 (18, 19) and NF-κB (20, 21), which are involved in mediating the M-CSF and RANKL signals (22–24).

TNFα stimulates RANKL production by stromal cells (25), T lymphocytes (26), B lymphocytes (27), and endothelial cells (28), and stimulates M-CSF production by murine or human stromal cells (29). In addition to this indirect mechanism, TNFα also can stimulate osteoclastogenesis directly (30) and strongly synergizes with interleukin-1 and RANKL to promote osteoclast differentiation and activation (31–33). However, the systemic effects of chronic TNFα exposure on osteoclastogenesis remain unknown. Recently, we reported that patients with active erosive arthritis have a marked increase in the frequency of osteoclast precursors (OCPs) in their peripheral blood mononuclear cell (PBMC) population (34). Remarkably, this increase was corrected after 12 weeks of anti-TNF therapy, and this correlated with a dramatic improvement in clinical signs and symptoms. This suggests that a central mechanism of TNFα-mediated resorption is the regulation of OCPs in the systemic circulation.

To analyze this phenomenon in greater detail and to elucidate the mechanisms involved, we examined the impact of TNFα on OCPs, using TNFα-transgenic (TNF-Tg) mice (the 3647 line) and mice treated with TNFα. TNFα significantly enhanced osteoclastogenesis by increasing the numbers of CD11bhigh OCPs in peripheral blood or tissue. It did not affect proliferation, differentiation, or survival of spleen CD11bhigh cells, but strongly stimulated the mobilization of CD11b+ cells from the bone marrow compartment. Thus, one of the mechanisms by which TNFα promotes osteoclastic resorption in chronic inflammatory bone diseases is to increase the number of CD11bhigh OCPs available for osteoclastogenesis.

MATERIALS AND METHODS

Reagents.

Human RANKL and etanercept were provided by Dr. W. Dougall (Amgen, Seattle, WA), murine TNFα was provided by Dr. C. R. Dunstan (Amgen, Thousand Oaks, CA), and fluorescein-conjugated RANKL was provided by Dr. M. Tondravi (American Red Cross, Rockville, MD). Recombinant human M-CSF was purchased from R&D Systems (Minneapolis, MN). Anti-murine CD11b (M1/70), c-Fms (AFS98), CD3 (145-2C11), B220 (RA3-6B2), and isotype controls were purchased from eBioscience (San Diego, CA). Anti-murine CD16/32 (FcγIII/II), c-Kit (2B8), Gr-1 (1A8), and isotype controls were from PharMingen (San Diego, CA). Anti-murine F4/80 (A3-1) and isotype control were from Serotec (Oxford, UK).

Animals.

TNF-Tg mice in a CBA × C57BL/6 background (3647 TNF-Tg line) (35) were obtained from Dr. G. Kollias. The Institutional Animal Care and Use Committee approved all animal studies.

Generation of osteoclasts.

Splenocytes, PBMCs, and bone marrow cells from TNF-Tg mice and their wild-type littermates were used to generate osteoclasts in the absence of osteoblast/stromal cells, as described previously (31). These cells were cultured in α-modified essential medium (Gibco BRL, Grand Island, NY) with 10% fetal calf serum (Hyclone, Logan, UT), RANKL (100 ng/ml), and M-CSF (10 ng/ml) for 5 days. Cells were fixed and stained for tartrate-resistant acid phosphatase (TRAP) using the Diagnostics Acid Phosphatase Kit (Sigma, St. Louis, MO) to identify osteoclasts. TRAP-positive cells containing ≥3 nuclei were counted as mature osteoclasts. For the functional study, splenocytes were cultured on bone slices for 10 days under the same conditions as described above. Osteoclasts were then removed, and the pits were visualized with 0.1% toluidine blue. The area of pits was quantified, and the data were expressed as the mean ± SEM area of pits (mm2)/osteoclast, as described previously (36).

CFU-M colony assay.

The in vitro colony-forming assay was performed as described previously (37). Freshly isolated spleen cells from wild-type or TNF-Tg mice were plated at a density of 105 cells/ml in a 35-mm dish. Cells were cultured in methylcellulose-based medium (Stemcell Technologies, Vancouver, British Columbia, Canada) supplemented with 30 ng/ml of M-CSF for 10 days. Colonies composed of >40 cells were counted under an inverted microscope.

Fluorescence-activated cell sorting (FACS) analysis and cell sorting.

Surface protein staining was performed on freshly isolated splenocytes, blood cells, and bone marrow cells. After red blood cell lysis, a single-cell suspension was incubated with anti-murine CD16/32 to block Fc receptor–mediated antibody binding. Cells were then labeled with fluorescent probes, as described previously (38). Data were acquired using a FACSCalibur instrument (Becton Dickinson, Bedford, MA) and analyzed with CellQuest software version 3.1 (Becton Dickinson). Pooled splenocytes from TNF-Tg or wild-type mice were labeled with anti-murine CD11b or double-stained with anti-murine CD11b and c-Fms antibodies and sorted on a FACSCalibur instrument. CD11bhigh, CD11blow, and CD11b− or CD11bhigh/c-Fms+ and CD11bhigh/c-Fms− cells were collected separately, reanalyzed to assure their purity (≥98%), and used for osteoclastogenesis assays, as described above.

Enzyme-linked immunosorbent assay (ELISA) for human TNFα in mouse serum.

Blood was drawn from TNF-Tg mice by cardiopuncture, and the serum was collected by centrifugation. The levels of human TNFα were detected according to the manufacturer's instructions (R&D Systems). The entire procedure was performed at room temperature. Briefly, 96-well plates were coated with 4 μg/ml of capture antibody (MAB610) overnight and blocked with phosphate buffered saline (PBS) containing 1% bovine serum albumin, 5% sucrose, and 0.05% NaN3 for 2 hours. Serum samples and standards were added and incubated for 2 hours. The plates were incubated with 200 ng/ml of biotinylated detection antibody (BAF210) for 1 hour, and then with streptavidin–horseradish peroxidase (DY998) for 20 minutes. The color reaction was developed by adding substrate solutions to the plates, and the optical density was read at 450 nm.

In vivo 5-bromo-2′-deoxyuridine (BrdU) labeling.

TNF-Tg mice and their wild-type littermates were given intraperitoneal injections of 1 mg BrdU (Sigma) 3 times over the course of 1 day at 8-hour intervals. Spleens were collected 8 hours after the last injection. BrdU staining was performed using standard immunohistochemistry (39). Briefly, spleen cell suspensions were surface-labeled with anti-CD11b antibody as described above, then fixed and permeabilized in PBS containing 1% paraformaldehyde plus 0.05% Tween 20 for 48–72 hours at 4°C. The cells were treated with 250 units/ml of DNase I (Sigma) for 60 minutes at 37°C, and BrdU incorporation was revealed with anti-BrdU antibody (PharMingen).

Quantitative real-time PCR.

RNA from TNFα-treated and untreated whole spleen cells, sorted CD11b− cells, and CD11blow cells was extracted using the RNeasy kit and the QIAshredder (Qiagen, Valencia, CA). Complementary DNA synthesis was performed as described previously (37). Quantitative PCR amplification was performed with gene-specific primers using a Rotor-Gene 2000 real-time amplification operator (Corbett Research, Mortlake, Australia). The primer sequences were as follows: CD11b 5′-ACAGACAAACAGCCCAAACC-3′ and 5′-GCCTCACCCATCAGTTGTTT-3′, actin 5′-AGATGTGGATCAGCAAGCAG-3′ and 5′-GCGCAAGTTAGGTTTTGTCA-3′. The quantity of CD11b messenger RNA (mRNA) in each sample was normalized using the threshold cycle value obtained for the actin RNA amplifications run in the same plate.

In vivo blockade with etanercept.

TNF-Tg mice (age 5 months, 3 mice per group) were given intraperitoneal injections of etanercept (10 mg/kg) or PBS twice weekly for 2 weeks. The mice were killed 3 days after the last injection, and spleen cells were subjected to FACS analysis, osteoclastogenesis assays, and CFU-M colony assays.

In vivo administration of TNFα.

Two protocols were used in this study. In the first protocol, 8-week-old CBA × C57BL/6 mice were first given intraperitoneal injections of BrdU (1 mg/mouse), 3 times daily for 3 days to obtain maximal labeling of bone marrow CD11b+ cells (≥96%). The mice were then challenged with a single intraperitoneal injection of murine TNFα (1 μg/mouse). After 4 hours, bone marrow, spleen, and blood cells were collected for FACS analysis with antibodies to CD11b and BrdU. In the second protocol, TNFα (1 μg/mouse) was injected into wild-type mice 4 times daily for 3 days, as described previously (40). Two hours after the last injection, spleens were obtained for FACS analysis, osteoclastogenesis assays, and CFU-M colony assays. Blood was collected for FACS analysis.

Statistical analysis.

All results are reported as the mean ± SEM. Comparisons were made by analysis of variance and Student's t-test for unpaired data. P values less than 0.05 were considered significant.

RESULTS

Osteoclastogenic potential of splenocytes and PBMCs increased by systemic TNFα.

The TNF-Tg mice (3647 line) used in this study (5) have 1 copy of a modified human TNFα transgene in which the AU-rich sequence–containing 3′-untranslated region (3′-UTR) was replaced with the 3′-UTR from the β-globin gene. This mutation increases the stability and translational efficiency of TNFα mRNA and results in persistent TNFα overexpression.

To determine whether exposure to persistent low levels of TNFα increases osteoclast formation, we performed in vitro osteoclastogenesis assays using splenocytes, PBMCs, and bone marrow cells from TNF-Tg mice and their wild-type littermates. Splenocytes and blood cells from TNF-Tg mice cultured with 100 ng/ml of RANKL and 10 ng/ml of M-CSF for 5 days formed more mature osteoclasts than did those from wild-type cells (Figure 1A). However, no significant differences were observed in the bone marrow cultures. In vitro TNFα blockade with the TNFα antagonist etanercept (105-fold over the media concentration of TNFα [<10 pg/ml]) had no effect on the enhanced osteoclast formation in TNF-Tg cultures (Figure 1B). Furthermore, splenocytes from TNF-Tg mice formed 2–3-fold more CFU-M colonies than did those from wild-type mice (Figure 1C). Taken together, these findings suggest that TNFα may stimulate osteoclastogenesis by increasing OCP numbers, but that it does not directly affect this process beyond the precursor stage.

Figure 1.

Increased frequency of osteoclast precursors in spleen and blood caused by persistent exposure to tumor necrosis factor α (TNFα). Splenocytes, blood cells, and adherent bone marrow cells from wild-type (WT) or TNFα-transgenic (TNF-Tg) mice were cultured with macrophage colony-stimulating factor (M-CSF) and receptor activator of NF-κB ligand (RANKL). A and B, A tartrate-resistant acid phosphatase assay was used to determine the osteoclastogenic potential of cells. C, M-CSF in methylcellulose was used to determine the colony-forming unit–macrophage (CFU-M) colony-forming potential of cells. Values are the mean and SEM of 4 replicate samples from a representative mouse. Similar results were obtained with 2 additional pairs of WT and TNF-Tg mice. ∗ = P < 0.005.

To examine whether TNFα overexpression affects mature osteoclast function ex vivo, we cultured splenocytes from TNF-Tg and wild-type mice on bone slices under osteoclastogenic conditions for 10 days, and measured the area of the resorption pits excavated by the mature osteoclasts. We observed no difference in the resorptive activity of osteoclasts from the 2 types of mice (pit area/osteoclast 0.0044 ± 0.0005 mm2 in wild-type mice, 0.00525 ± 0.0007 mm2 in TNF-Tg mice).

Increased late-stage osteoclast precursors in spleens of TNF-Tg mice.

Cell surface markers have been used to characterize OCPs at various stages of differentiation (41). The earliest OCP, which differentiates from the pluripotent hematopoietic stem cell, is c-Kit+/c-Fms−/CD11b−/RANK−. This cell differentiates into the c-Kit+/c-Fms+/CD11b−/RANK− early-stage precursor and, following M-CSF stimulation, proceeds to the c-Kit−/c-Fms+/CD11b+/RANK+ late-stage precursor, which differentiates fully in response to RANKL (10, 11, 24, 41). FACS characterization of splenocytes from TNF-Tg mice showed a 4–7-fold increase in the CD11b+ population compared with wild-type mouse cells (Figure 2) and a consistent increase in the c-Fms+ population. According to their expression levels of CD11b, the CD11b+ splenocytes can be further divided into CD11bhigh and CD11blow cells. Only the CD11bhigh population was significantly increased in TNF-Tg mice compared with that in wild-type mice (Figure 3A). To functionally characterize this CD11bhigh population, we sorted CD11bhigh, CD11blow, and CD11b splenocytes and cultured them with M-CSF and RANKL. TRAP+ osteoclasts formed only from the CD11bhigh population (Figures 3B and C). Thus, all OCPs capable of forming mature osteoclasts in the culture are in the CD11bhigh population.

Figure 2.

Markedly increased CD11b+ population in the spleens of TNF-Tg mice. Splenocytes from WT or TNF-Tg mice were stained with fluorescent probes specific for c-Kit, c-Fms, CD11b, or RANK and analyzed by fluorescence-activated cell sorting, as described in Materials and Methods. Values are the percentage of cells that were positive for the indicated surface protein from representative animals. A total of 5 pairs of WT and TNF-Tg mice were examined, with similar results. See Figure 1 for definitions.

Figure 3.

All splenocytes with osteoclastogenic potential are in the CD11bhigh population. A, Splenocytes from WT and TNF-Tg mice were stained with fluorescent antibodies specific for CD11b. B, The CD11bhigh (R1), CD11blow (R2), and CD11b− (R3) populations were sorted and cultured with M-CSF and RANKL, and their osteoclastogenic potential was determined by tartrate-resistant acid phosphatase (TRAP) assay. Micrographs show that TRAP+ osteoclasts formed only from CD11bhigh (R1) cells. Values are representative of 3 independent experiments. C, Statistical analysis of the distribution of CD11bhigh splenocytes from 16 pairs of mice, showing a significant increase in the percentage of splenocytes from TNF-Tg mice versus controls. See Figure 1 for other definitions.

To further characterize the increased CD11bhigh cell population, we double-stained them with antibodies to CD11b and markers for other cell lineages, including CD3 (T cells), B220 (B cells), F4/80 (mature macrophages), and Gr-1 (granulocytes). We also investigated markers for OCP, including c-Kit, c-Fms, and RANK. Representative histograms from experiments in which we gated on the CD11bhigh splenocytes are shown in Figure 4. Because this population contains both c-Fms+ and c-Fms− cells, we sorted the CD11bhigh/c-Fms+ and CD11bhigh/c-Fms− subpopulations and performed the osteoclastogenesis assay. Both subpopulations had osteoclastogenic potential, and the CD11bhigh/c-Fms+ cells formed more osteoclasts than did the CD11bhigh/c-Fms− cells (Figure 5). Thus, CD11bhigh alone can be used as a representative marker for OCPs in the spleen.

Figure 4.

Phenotypic characterization of the CD11bhigh splenocyte population. Splenocytes from tumor necrosis factor α–transgenic mice were double-stained with anti-CD11b antibody in combination with probes for CD3, B220, F4/80, Gr-1, c-Kit, c-Fms, and receptor activator of NF-κB (RANK). Live CD11bhigh cells were gated. The histograms are representative of 3 independent experiments.

Figure 5.

Osteoclast formation by both CD11bhigh/c-Fms− and CD11bhigh/c-Fms+ splenocytes. Splenocytes from tumor necrosis factor α–transgenic mice were double-stained with antibodies to CD11b and c-Fms. CD11bhigh/c-Fms− and CD11bhigh/c-Fms+ populations were sorted (A) and cultured as described in Figure 3, and the number of tartrate-resistant acid phosphatase–positive osteoclasts was counted (B). Values are representative of 2 independent experiments. Bars show the mean and SEM. ∗ = P < 0.01.

To determine whether there is a correlation between the increased frequency of CD11bhigh OCPs and the blood concentration of human TNFα in the transgenic mice, we collected splenocytes and blood from wild-type and TNF-Tg mice at different ages, corresponding to various stages of development and progression of inflammatory arthritis: prior to onset (1 month), onset (2–3 months), and advanced stage (4 months). The frequency of CD11bhigh OCPs in the spleen was determined by FACS analysis (Figure 6A), and the concentration of human TNFα in serum was measured by ELISA (Figure 6B). Increased numbers of CD11bhigh OCPs were first observed in TNF-Tg mice at age 2 months, corresponding to the time of initial detection of human TNFα in serum and the development of swollen ankles, the first macroscopic sign of inflammatory arthritis. After the onset of arthritis, the frequency of CD11bhigh OCPs and the concentrations of human TNFα serum remained elevated and did not increase further with progression of erosive arthritis.

Figure 6.

CD11bhigh osteoclast precursor (OCP) frequency and serum human tumor necrosis factor α (TNFα) concentrations in TNFα- transgenic (TNF-Tg) mice at different stages of development. Splenocytes and blood were collected from TNF-Tg mice and their wild-type (WT) littermates at 1, 2, 3, and 4 months of age. A, The frequency of CD11bhigh OCPs in the spleen was determined by fluorescence-activated cell sorting. Representative results are from ≥3 animals per group. B, The concentration of human TNFα in serum was analyzed by enzyme-linked immunosorbent assay. No statistically significant differences were found in the frequency of CD11bhigh OCPs and the serum levels of human TNFα between 2-, 3-, and 4-month-old TNF-Tg mice. Bars show the mean.

Increased frequency of CD11bhigh osteoclast precursors and enhanced osteoclastogenesis in TNF-Tg mice are reversible by etanercept treatment in vivo.

To investigate whether TNFα blockade in vivo prevents the increases in CD11bhigh OCPs and osteoclast formation, we administered etanercept (10 mg/kg) or placebo intraperitoneally into TNF-Tg mice with established joint disease, twice a week for 2 weeks. Etanercept reduced the numbers of CD11bhigh splenocytes (Figure 7A) and the osteoclastogenic and CFU-M colony-forming potential (Figures 7B and C) of these cells to the levels observed in wild-type mice. Thus, the TNFα-mediated increase in OCP is reversible with anti-TNF therapy, which is consistent with our clinical findings (34).

Figure 7.

Reversal of increased frequency of CD11bhigh osteoclast precursors in TNF-Tg mice, with anti-TNF therapy. TNF-Tg mice were treated with intraperitoneal injections of etanercept (10 mg/kg) or placebo twice weekly for 2 weeks. Splenocytes from these mice were isolated and analyzed for CD11b expression by fluorescence-activated cell sorting (A), for osteoclastogenic potential by tartrate-resistant acid phosphatase assay (B), and for colony-forming unit–macrophage (CFU-M)–forming potential by colony assay (C). Values in A are from 1 representative animal; values in B and C are the mean and SEM of 4 replicate samples from a representative mouse (n = 3 per group). Similar results were obtained from 2 additional pairs of WT and TNF-Tg mice. Bars show the mean and SEM. ∗ = P < 0.005 versus WT; ∗∗ = P < 0.01 versus phosphate buffered saline (PBS) treatment. See Figure 6 for other definitions.

Systemic TNFα does not affect the proliferation rate, differentiation, or apoptosis of CD11bhigh osteoclast precursors in the periphery, but may mobilize them from the bone marrow into the circulation.

There are 4 fundamental mechanisms by which TNFα may increase the number of CD11bhigh OCPs in the periphery: proliferation, survival, differentiation, and redistribution from the bone marrow. In proliferation assays, cells from TNF-Tg and wild-type mice were labeled with BrdU for 24 hours, and spleen cells were stained with antibodies against CD11b and BrdU. TNF-Tg mice had the expected increase in CD11bhigh cells (Figure 8A), but no increase in the percentage of BrdU+ CD11bhigh cells (Figures 8B and C). In survival assays, freshly isolated spleen cells from TNF-Tg and wild-type mice were analyzed by FACS using antibodies against CD11b, fluorescently labeled annexin V, and 7-aminoactinomycin D (7-AAD). In the CD11bhigh population, the percentage of annexin V+/7-AAD− cells (apoptotic cells) was similar in TNF-Tg (9.5%) and wild-type (9.6%) mice. In differentiation assays, wild-type splenocytes were cultured with TNFα (10 ng/ml), and the percentage of CD11bhigh OCPs after 24 hours was similar in control and TNF-treated cultures (Figure 9A), as were the levels of CD11b mRNA analyzed by quantitative real-time PCR after 1, 4, and 24 hours of TNF treatment (Figure 9B). Furthermore, CD11b mRNA expression assessed by real-time PCR in CD11b− and CD11blow splenocytes sorted by FACS, as described in Figure 3, and cultured in the presence of TNFα (10 ng/ml) for 12 hours, was not detectable above background levels (Figure 9C).

Figure 8.

Lack of effect of systemic administration of TNFα on proliferation of CD11bhigh splenocytes in vivo. Cells from TNF-Tg and WT mice were labeled with 5-bromo-2′-deoxyuridine (BrdU) for 24 hours, as described in Materials and Methods. Splenocytes were isolated, double-stained with antibodies to CD11b and BrdU, and analyzed by fluorescence-activated cell sorting. A, Percentage of CD11bhigh/BrdU+ cells. B, Percentage of BrdU+ cells in this population with gating of live CD11bhigh cells. C, Statistical analysis of 10 pairs of mice demonstrated no significant difference in the percentage of BrdU+ cells in the CD11bhigh population of TNF-Tg and WT mice. Bars show the mean. See Figure 6 for other definitions.

Figure 9.

Tumor necrosis factor α (TNFα) does not induce differentiation of CD11b−/low to CD11bhigh splenocytes in vitro. Spleen cells from wild-type mice were treated with TNFα (10 ng/ml) for 24 hours and subjected to fluorescence-activated cell sorting analysis using anti-CD11b antibody. A, Percentage of CD11bhigh cells. B, Total wild-type splenocytes and C, sorted CD11b− and CD11blow splenocytes were treated with TNFα (10 ng/ml). Total RNA was extracted at various time points and analyzed by real-time polymerase chain reaction using CD11b-specific primers. Values are representative of 2 independent experiments. Bars show the mean and SEM.

To examine whether accumulation of CD11bhigh cells in the periphery is attributable to TNFα-induced mobilization of precursors from bone marrow, we labeled wild-type mice with BrdU for 3 days, then challenged them with 1 injection of TNFα (1 μg intraperitoneally) or PBS. After 4 hours, bone marrow, spleen, and blood cells were collected and analyzed by FACS, using antibodies to CD11b and BrdU. We observed no change in the percentage of CD11b+ cells in bone marrow (Figure 10A) or spleen (data not shown), but we did observe a 4-fold increase in the percentage of CD11b+/BrdU+ cells in blood (Figure 10B), suggesting that TNFα rapidly mobilizes a small fraction of OCPs from the bone marrow to produce a marked increase in the blood. However, further distribution of these cells into peripheral tissues such as spleen may require more time. To test this possibility, wild-type mice were given TNFα injections for 3 days (1 μg/injection, intraperitoneally, 4 times daily) and killed 2 hours after the last injection. The percentage of CD11bhigh splenocytes (Figure 10C) and PBMCs (data not shown) was increased significantly in the TNFα-treated mice compared with controls, similar to that observed in untreated TNF-Tg mice (data not shown). Correspondingly, this treatment caused an increase in the osteoclastogenic and CFU-M colony-forming potential of the splenocytes from these mice (Figure 10D).

Figure 10.

Tumor necrosis factor α (TNFα)–stimulated release of CD11b+ osteoclast precursors from the bone marrow to the periphery. A and B, Cells from 8-week-old wild-type mice were labeled with 5-bromo-2′-deoxyuridine (BrdU) in vivo for 3 days and then challenged with TNFα (1 μg/mouse, intraperitoneally) for 4 hours. Bone marrow cells (A) and blood cells (B) were collected, stained with antibodies to BrdU and CD11b, and analyzed by fluorescence-activated cell sorting (FACS). The percentage of CD11b+/BrdU+ cells is shown in the upper-right corners, and that of CD11b+/BrdU− cells is in the upper-left corners. C and D, Wild-type mice were given intraperitoneal injections of phosphate buffered saline (PBS) or TNFα (1 μg/mouse), 4 times daily for 3 days. Splenocytes were analyzed for CD11b expression by FACS (C), and were assessed for their osteoclastogenic potential by tartrate-resistant acid phosphatase assay and colony-forming unit–macrophage (CFU-M)–forming potential by colony assay (D). Values are the mean and SEM of 4 replicate samples from a representative mouse. Similar data were obtained from studies of 2 additional mice. Bars show the mean and SEM. ∗ = P < 0.005 versus control.

DISCUSSION

A relationship between TNFα and osteoclastic resorption is firmly established in diseases associated with erosive bone loss, such as RA (4, 42). TNFα increases osteoclast formation in vitro (30–32), and directly affects OCPs in vivo in normal mice (33). In these studies, however, very large amounts of TNFα were used in vitro and in vivo, which may not mimic the disease states. Thus, it is important to determine whether long-term exposure to a relevant concentration of TNFα affects osteoclast formation, considering the number of RA patients receiving anti-TNF therapy (43, 44) and the potential new indications for this therapy in other inflammatory bone diseases (45). Although it is clear that anti-TNF therapy is efficacious for RA, clinical studies to evaluate its effects on bone resorption have commenced only recently (46, 47), and a unifying hypothesis as to how anti-TNF therapy inhibits bone erosion in patients is warranted.

In TNF-Tg mice (3647 line), there is persistent low-level expression of TNFα. As a result, these mice develop an erosive arthritis with features similar to those seen in human RA (35), including focal erosions affecting the immediate subchondral bone and bone at the joint margins. Therefore, use of this model is appropriate for studying the mechanisms of TNFα-mediated osteoclast formation in inflammatory arthritis. As expected, splenocytes from TNF-Tg mice had enhanced osteoclastogenic and CFU-M colony-forming potential compared with those from wild-type mice (Figure 1). This enhanced osteoclastogenesis was not inhibited by TNFα blockade in vitro (Figure 1B) but could be recapitulated in cultures of splenocytes from wild-type mice injected with TNFα (Figure 10). From these data, we conclude that TNFα has a priming effect on OCPs in vivo, thereby increasing the number of preosteoclasts in the periphery outside the bone environment. However, these TNFα-induced preosteoclasts do not have increased bone-resorbing capacity, indicating that the systemic TNFα effect is restricted to increasing the number of these cells rather than increasing their function.

If our interpretation is correct, i.e., that systemic TNFα increases the number of preosteoclasts in peripheral tissues such as spleen and blood, then these cells should be identifiable by phenotypic surface markers. Indeed, splenocytes from TNF-Tg mice had 4–7-fold more CD11bhigh cells compared with those from wild-type mice. It is important to note that the characterization of the majority of these cells as RANK− by FACS (81% by double-staining) does not preclude them from being the RANKL-responsive preosteoclasts that ultimately fuse to form the bone-resorbing cell. These cells could express low levels of functional RANK or up-regulate surface RANK expression shortly after stimulation with M-CSF (41). The CD11bhigh cells do not express common markers for T or B lymphocytes or mature macrophages, but they do express osteoclast precursor markers such as c-Fms (Figure 4). Furthermore, all of the OCPs in spleen are CD11bhigh, because CD11blow and CD11b− cells do not form osteoclasts. Both CD11bhigh/c-Fms+ and CD11bhigh/c-Fms− populations have the potential to form osteoclasts. Thus, in our subsequent studies, we used CD11b as a single surface marker to identify the TNFα-induced preosteoclasts.

The increase in the number of OPCs in the blood was observed at 2–3 months of age in the transgenic mice, at the same time that blood concentrations of TNFα increased. At 1 month of age, TNF-Tg mice had undetectable blood concentrations of human TNF and normal OPC numbers (Figure 6). This increase in the number of CD11bhigh splenocytes and the osteoclastogenic and CFU-M colony-forming potential to levels observed in wild-type mice was reversible by in vivo TNF blockade with etanercept (Figure 7). Based on these 2 findings, we conclude that the increase in the number of peripheral OCPs is directly attributable to persistent TNF exposure. Consistent with our finding in the TNF-Tg mice, we recently demonstrated that patients with psoriatic arthritis have a marked increase in the number of preosteoclasts in their PBMC population compared with normal and osteoarthritic controls (34). This increase also appears to be reversible with anti-TNF therapy and may be a dominant mechanism by which this treatment inhibits erosions.

Last, we examined the cellular mechanisms of the increased frequency of TNFα-mediated CD11bhigh OCPs in the periphery. We observed that accumulation of these cells was not attributable to an alteration in differentiation, proliferation, or survival of the cells (Figures 8 and 9), and that another mechanism must be involved. In adult life, hematopoietic precursors that give rise to preosteoclasts are derived mainly from the differentiation of stem cells in the bone marrow compartment. Our finding that injection of TNFα into wild-type mice induced a remarkable change in the tissue distribution of CD11bhigh cells, similar to that observed in the blood and spleen of TNF-Tg mice, suggests that redistribution may be a mechanism responsible for the TNFα-mediated increase in peripheral OCP frequency (Figure 10). However, we were unable to detect a significant decrease in CD11b+ cells in the bone marrow. Our explanation for this discrepancy is that bone marrow contains a very large pool of CD11b+ cells (∼70% of bone marrow monocytes are CD11b+, and adult mice have ∼3.5 × 107 CD11b+ bone marrow cells), while blood contains very few (∼10% of PBMCs are CD11b+, and adult mice have ∼2 × 105 CD11b+ PBMCs). A 4-fold increase in CD11b+ PBMCs induced by TNFα would correspond to the release of only 2% of the total bone marrow pool, which is too small to be detected as a significant change in bone marrow. However, continuous release of these cells would lead to a significant increase in OCP numbers in the periphery, which we believe results in a dramatic change in the number of osteoclasts, as seen in the joints of TNF-Tg mice (35) and patients with aggressive psoriatic arthritis (34).

Taken together, our findings provide a new mechanism for TNFα-induced erosive arthritis in which a significant component of its effect could be attributable to the mobilization of CD11bhigh OCPs from the bone marrow, thereby increasing their numbers in the circulation. This is consistent with the clinical studies that demonstrated alterations in the commitment of precursor cells in the peripheral blood of patients with RA (51), and our finding of a strong correlation between the level of circulating OCPs and erosive arthritis. Based on the results of these studies, we propose that determining the frequency of CD11bhigh PBMCs could be used as a diagnostic approach to identifying patients with active disease or flares that might lead to further bone erosion. Furthermore, patients whose disease will be responsive or refractory to anti-TNF therapy might be identified by changes in the OCP population following therapy.

Acknowledgements

The etanercept and RANK:Fc used in this study were provided by Immunex Inc. The murine TNFα was provided by Amgen Inc. The authors would like to thank W. Dougall for critical advice and J. Harvey for technical assistance with the histologic analysis.

Ancillary