Dual Modulation of Osteoclast Differentiation by Lipopolysaccharide


  • Wei Zou,

    1. The H. Hubert Humphrey Center for Experimental Medicine and Cancer Research, The Hebrew University Faculty of Medicine, Jerusalem, Israel
    Search for more papers by this author
  • Zvi Bar-Shavit Ph.D.

    Corresponding author
    1. The H. Hubert Humphrey Center for Experimental Medicine and Cancer Research, The Hebrew University Faculty of Medicine, Jerusalem, Israel
    • The H. Hubert Humphrey Center for Experimental, Medicine and Cancer Research, The Hebrew University Faculty of Medicine, P.O. Box 12272, Jerusalem 91120, Israel
    Search for more papers by this author

  • The authors have no conflict of interest.


Lipopolysaccharide (LPS) modulates bone resorption by augmentation of osteoclastogenesis. It increases in osteoblasts the production of RANKL, interleukin (IL)-1, prostaglandin E2 (PGE2), and TNF-α, each known to induce osteoclast activity, viability, and differentiation. We examined the role of direct interactions of LPS with osteoclast precursors in promoting their differentiation. To this end, we have used bone marrow mononuclear cell preparations in the absence of osteoblasts or stromal cells. We found that LPS does not induce osteoclast differentiation in these cells. Moreover, the inclusion of LPS blocked the osteoclastogenic activity of RANKL. However, LPS is a potent inducer of osteoclastogenesis in RANKL-pretreated cells, even if present in the absence of exogenous RANKL. Osteoprotegerin (OPG) does not affect the stimulatory phase of LPS modulation of osteoclastogenesis, ruling out involvement of endogenous RANKL. LPS induces the expression of TNF-α and IL-1β in osteoclast precursors, regardless if they were or were not pretreated with RANKL. These two cytokines induced osteoclast differentiation in RANKL-pretreated cells. To examine if these cytokines mediate LPS effect in an autocrine mechanism, we measured the effect of their neutralization on LPS osteoclastogenic activity. Although neutralization of IL-1β did not affect LPS activity, a marked inhibition was observed when TNF-α was neutralized. However, TNF-α expression was increased also in conditions in which LPS inhibited RANKL osteoclastogenic activity. We found that LPS reduces the expression of RANK and macrophage colony—stimulating factor (M-CSF) receptor. In summary, LPS impacts on osteoclastogenesis also via its interactions with the precursor cells. LPS inhibits RANKL activity by reducing the expression of RANK and M-CSF receptor and stimulates osteoclastogenesis in RANKL-pretreated cells via TNF-α.


THE NUCLEAR translocation of NF-κB in response to the interaction of RANKL with its surface receptor RANK leads to osteoclast differentiation from its precursor cell, member of the monocyte-macrophage lineage.(1–3) Numerous studies, including the use of RANK- and RANKL-deficient mice, suggest that this pathway is physiologically important.(4–9) In vitro and in vivo studies(10–13) also established an essential role in osteoclastic differentiation to macrophage colony-stimulating factor (M-CSF) interaction with its surface receptor.

The involvement of a variety of cytokines, among them TNF-α and interleukin (IL)-1β, in this differentiation is well documented.(14–17) These two cytokines are capable of promoting osteoclast differentiation, although less efficiently than RANKL.(17,18) Recently, we have shown that TNF-α mediates, at least in part, RANKL-osteoclastogenic activity in an autocrine mode of action.(18) Moreover, reexamination of the ability of TNF-α to induce osteoclastic differentiation(19) suggests that it depends on prior exposure of the precursor cell to RANKL. Thus, TNF-α is an efficient osteoclastogenic factor in cells that already received signals to undergo osteoclastic differentiation. TNF-α and other cytokines are thought to play a role in postmenopausal osteoporosis; their synthesis and circulatory levels are increased as a result of reduced female sex hormones levels.(20,21)

The bacterial-derived lipopolysaccharide (LPS) is responsible for inflammation and osteolysis because of its ability to augment cytokine synthesis and release.(22–31) For example, a recent report(30) suggests that LPS adherent to the wear particles causes aseptic loosening of orthopedic implants. Several years ago, Abu-Amer and colleagues(31) showed that TNF-α, through interaction with its P55 receptor, mediates LPS osteoclastogenic activity.

In spite of the information from the literature on LPS modulation of osteoclast differentiation,(24–31) the contribution of direct interactions between LPS and the osteoclast precursor in the absence of osteoblasts or stromal cells was not addressed. In this study we show that the result of LPS interactions with osteoclast precursors depends on the stage of cell differentiation and can result in inhibition or enhancement of osteoclastogenesis.



Seven- to 9-week-old male BALB/c mice were obtained from Harlan Laboratories Limited (Jerusalem, Israel). These studies were approved by the institutional animal care committee.


Glutathione S-transferase (GST)-RANKL (residues 158-316) was prepared(32) from a plasmid kindly provided by Dr. Ross and Dr. Teitelbaum (Washington University, St. Louis, MO, USA). Each preparation was tested for lack of detectable LPS by limulus amoebocyte lysate assay (BioWhittaker, Walkersville, MD, USA) following the manufacturer's instructions and for a complete inhibition of activity by osteoprotegerin (OPG). Rat monoclonal anti-mouse TNF-α and hamster monoclonal anti-mouse p55 (TNF-α receptor-1 [TNFR-1]) neutralizing antibodies were purchased from Pharmingen (San Diego, CA, USA); rabbit anti-mouse M-CSF receptor (C-Fms) antibody was purchased from Santa Cruz Biotechnology, Inc. (Santa Cruz, CA, USA). Recombinant mouse TNF-α, OPG-Fc (OPG/Fc chimera containing OPG amino acids 1-398 residues), and M-CSF were purchased from R & D Systems, Inc. (Minneapolis, MN, USA). Recombinant mouse IL-1β and IL-1 receptor antagonist were kindly provided by Dr. Charles Dinarello (University of Colorado, Denver, CO, USA). Endotoxin testing confirmed the absence of LPS contamination in our reagents. Purified Escherichia coli 055:B5 LPS was purchased from Sigma (St. Louis, MO, USA). Media and sera were purchased from Biological Industries (Beth Haemek, Israel). All chemicals and reagents were of analytical grade.

In vitro osteoclast formation assay

Primary bone marrow monocytes (BMMs) were prepared as described.(18) Femora and tibia of 7- to 9-week-old mice were aseptically removed. The bone ends were cut and marrow cells were obtained by flushing the bones with α-minimal essential medium (α-MEM). Mononuclear cells were isolated by Ficoll-Hypaque gradient centrifugation (500g, 20 minutes). Cells were plated in 96-well plates at a density of 1.3 × 105 cells/well in 0.2 ml of α-MEM containing 10% FCS in the presence of mouse M-CSF (30 ng/ml) and an indicated dose of RANKL, TNF-α, or IL-β. On day 3, medium was changed and on day 4 or day 5, osteoclast formation was evaluated. Similar results were obtained if cells were grown first for 72 h in the presence of M-CSF (30 ng/ml), and the actual experiment starts at this point after monolayer rinsing. Absence of stromal cells was confirmed by alkaline phosphatase cytochemical assay (<0.5% positive cells), and the absence of T cells was confirmed by fluorescence-activated cell sorting (FACS) analysis (0.6% CD3e+ cells). Furthermore, no message for RANKL could be detected in the cultures.(18)

TRAP staining

A commercial kit (catalogue no. 387-A; Sigma) was used according to the manufacturer's instructions, omitting counterstain in hematoxylin. TRAP+ cells containing three or more nuclei were scored as osteoclasts.

Western blot analysis

Cells were solubilized in lysis buffer containing 10 mM of Tris-HCl, pH 7.4, 150 mM of NaCl, 1 mM of EDTA, 1% Triton X-100, and protease inhibitors [10 ng/ml of aprotinin (Sigma), 10 ng/ml of leupeptin (Sigma), and 50 mM of 4(2-aminoethyl)benzensulfonyl fluoride (AEBSF; Sigma)] at 4°C for 20 minutes. The cell lysates were subjected to centrifugation at 15000g at 4°C for 20 minutes. The supernatants were saved and their protein contents were determined using Coomassie Plus Protein Assay Reagent (Pierce, Rockford, IL, USA). Fifty micrograms of lysates were loaded onto 10% SDS-polyacrylamide gels. After the proteins were separated, they were transferred to an immunobilon-P membrane (Millipore, Bedford, MA, USA). Membranes were blocked for 1 h with 5% nonfat milk in Tris-buffered saline containing 0.05% Tween 20 (TBST). Membranes were probed with relevant antibodies (c-fms, 1:300; β-actin, 1:10000) in 5% nonfat milk in TBST. After washes with TBST, the blots were incubated with secondary antibodies conjugated to horseradish-peroxidase. Immunoreactive bands were detected by the enhanced chemiluminescence (ECL) reagent (Pierce).

Methylene blue uptake

The cell number was estimated by the methylene blue uptake assay using a Dynatech plate reader (Dynatech Laboratories, Inc., Chantilly, VA, USA).(33)

Northern blot analysis

Total cellular RNA was extracted using TRI reagent (Sigma), fractionated by electrophoresis on 1.2% agarose-formaldehyde gels (10 μg/lane), and transferred to nylon membranes (Hybond; Amersham, Little Chalfont, UK). The membranes were hybridized for 16-18 h at 42°C with radioactive cDNA probes for mouse TNF-α (1.1 kb), c-fms (752 bp), IL-1β (671 bp), RANK (511 bp), and mouse ribosomal protein L32 cDNA (1 kb) or β-actin (1 kb) as an internal housekeeping gene control. cDNA probes of IL-1 β, c-fms, and RANK were generated by reverse-transcription polymerase chain reaction (RT-PCR) with a total RNA from mouse macrophages. The primers for IL-1β were sense 5′-AATCTCACAGCAGCACATCAA-3′ and antisense 5′-AGCCCATA CTTTAGGAAGACA-3′; for c-fms, sense 5′- AACAAGTTCTACAAACTGGTGAAGG-3′ and antisense 5′-GAAGCCTGTAGTCTAAGCATCTGTC-3′; and for RANK, sense 5′-CTCTGCGT GCTGCTCGTTCC-3′ and antisense 5′ -TTGTCCCCTGGTGTGCTTCT-3′. Then, the hybridized membrane was subjected to autoradiography and the density of each TNF-α, c-Fms, RANK, IL-1β, and L32 or β-actin mRNA bands were quantified using Fluor-S MultiImager and Multi-Analyst/PC software (Bio-Rad Laboratories, Hercules, CA, USA).

Statistical analysis

Student's t-test was used to determine significance of difference. Values are presented as mean ± SD (n = 4-6 of a representative experiment of four to eight experiments).


Incubation of BMMs with soluble murine RANKL (20 ng/ml) and M-CSF (30 ng/ml) results in numerous multinucleated, TRAP+ osteoclasts. LPS, in the absence of RANKL, does not induce osteoclast formation (not shown). Moreover, addition of LPS to BMM cultures for 5 days inhibits the osteoclastic effect of RANKL in a dose-dependent manner (Fig. 1). Already, at 0.2 ng/ml of LPS, inhibition of 30% (p < 0.01) was observed, and almost 90% inhibition (p < 0.005) was observed at 20 ng/ml. Measurement of methylene blue uptake shows that the inhibition of osteoclastogenesis was accompanied by increased cellular content. Thus, the osteoclastogenesis inhibition by LPS is specific, and not caused by a toxic effect or a general inhibition of proliferation. We next examined whether LPS modulates RANKL-induced osteoclastogenesis when added at different time points after addition of RANKL. LPS (20 ng/ml) or the same volume of medium were added to the BMMs (incubated with M-CSF [30 ng/ml] and RANKL [5 ng/ml]) at the indicated time after the start of the BMM cultures. We see in Fig. 2A that when LPS is included for the whole experiment (102 h), almost 90% (p < 0.001) inhibition of osteoclast formation is observed, consistent with the data shown in Fig. 1. The inhibition becomes less pronounced when LPS is added at later time points. Addition of LPS 10 h and 24 h after the start of the experiment results in inhibition of 83% (p < 0.002) and 67% (p < 0.002), respectively. No significant inhibition is observed when LPS is added 48 h after the start of the culture. In contrast, addition of LPS 72 h after the start of the experiment results in a remarkable enhancement of RANKL osteoclastogenic activity (more than threefold; p < 0.001). Medium addition at the various time points does not affect the activity of RANKL. In Fig. 2B we show the dose response of the enhancement observed when LPS is present at the last 30 h of the experiment; the enhancement is evident already at 1 ng/ml (98%; p < 0.002). Moreover, LPS induces osteoclastogenesis even in the absence of RANKL during the last 30 h. In Fig. 2C we see that LPS increases the cellular content of wells when present together with RANKL during the last 30 h of the experiment; addition of 20 ng/ml of LPS results in more than a twofold increase (p < 0.005). On the other hand, in the absence of RANKL during the last 30 h, LPS does not affect cell content of wells.

Figure FIG. 1.

Modulation of RANKL-induced osteoclastogenesis by LPS. BMMs were grown with M-CSF (30 ng/ml) and RANKL (20 ng/ml) for 5 days in the presence or absence of LPS (0.2-20 ng/ml). Number of TRAP+ cells containing more than two nuclei (open bars) and methylene blue staining value (hatched bars) are presented (*p < 0.01; **p < 0.005 [comparing LPS-treated with -untreated cells]).

Figure FIG. 2.

Modulation of RANKL-induced osteoclastogenesis by LPS. Dependence on the timing of LPS addition. BMMs were grown with M-CSF (30 ng/ml) and RANKL (5 ng/ml) for 102 h. (A) At the indicated time point 20 ng/ml of either LPS or the same volume of medium was added (*p < 0.01; **p < 0.001 [comparing LPS-treated with -untreated]). (B) Different concentrations of LPS were added for the last 30 h of the culture (in the presence of M-CSF, with or without RANKL). Number of TRAP+ cells containing more than two nuclei are presented in panels A and B (*p < 0.05; **p < 0.001 [comparing LPS-treated with -untreated]). (C) Methylene blue staining of the experiment presented in panel B (*p < 0.05; **p < 0.001 [comparing LPS-treated with -untreated]).

Next, we examined if RANKL is involved in mediating the osteoclastogenic effect of LPS. As expected, we see in Fig. 3 that OPG, the decoy receptor of RANKL, completely blocked its osteoclastogenic effect. In contrast, OPG did not affect LPS-stimulated osteoclastogenesis, indicating that this LPS activity is RANKL independent.

Figure FIG. 3.

LPS osteoclastogenic activity is RANKL independent. BMMs were grown with M-CSF (30 ng/ml) and RANKL (5 ng/ml) for 102 h. LPS (20 ng/ml) was added for the last 30 h in the absence of RANKL, with or without OPG. Alternatively, RANKL (5 ng/ml, without LPS) was present in the last 30 h with or without OPG. Data are number of TRAP+ cells containing more than two nuclei.

LPS is a potent activator of mononuclear phagocytes, known to induce the production of proinflammatory cytokines, including TNF-α and IL-1β, which are capable of stimulating osteoclastogenesis. Therefore, we examined the possibility that these cytokines participate in mediating LPS effect. Using Northern blot analysis we confirmed that LPS increases mRNA levels of both cytokines in BMMs, regardless of if they were or were not pretreated with RANKL (Fig. 4). Increased TNF-α mRNA levels are detected within 30 minutes after LPS addition. Maximal effect is observed after 1-2 h. Increased IL-1β mRNA levels are detected within 1 h after LPS addition, reaching a maximum after 4 h.

Figure FIG. 4.

Modulation of TNF-α and IL-1β mRNA abundance by LPS. BMMs were grown (72 h) in the presence of M-CSF with or without RANKL (5 ng/ml). Monolayers then were washed and incubated with 10 ng/ml of LPS in the presence of M-CSF to the indicated time. RNA then was prepared and examined using Northern blot analyses for the abundance of TNF-α and IL-1β transcripts. β-Actin serves as a housekeeping gene control. Numbers indicate TNF-α or IL-1β/β-actin ratio as compared with control (normalized to 1).

If indeed the stimulation of osteoclast formation by LPS is mediated via TNF-α and IL-1β, then antibodies to TNF-α and to its type 1 receptors and IL-1β receptor antagonist (IL-1ra) are expected to block the LPS osteoclastogenic effect. In Fig. 5A we confirm that the cytokines induce osteoclastogenesis in RANKL-pretreated BMMs, which is inhibited by their respective inhibitors. In Fig. 5B, we see that antibodies to TNF-α and to its type 1 receptor, but not IL-1ra, significantly inhibit the LPS osteoclastogenic effect. Control immunoglobulin G (IgG) preparations (at 10 μg/ml, similar to the antibody concentrations in this experiment) are inactive (not shown). It is of note that LPS treatment increases the abundance of TNF-α transcripts also under conditions in which LPS inhibits osteoclastogenesis. RANKL-RANK and M-CSF-M-CSF-receptor (c-fms) interactions are key essential events to osteoclast differentiation. Therefore, we examined the effects of LPS on the abundance of RANK and c-fms mRNAs in BMMs. Northern blot analysis (Fig. 6) shows that LPS reduced the abundance of both transcripts. Dramatic decreases in mRNA abundance of RANK (∼65-80%) and c-fms (∼40-60%) are obtained after 72 h. The kinetics (Fig. 6A) and dose response (Fig. 6B) of the LPS effect are presented. Western blot analysis (Fig. 6C) shows reduction of c-fms protein by LPS.

Figure FIG. 5.

Modulation of LPS osteoclastogenic activity by anti-TNF-α and anti-TNFR-1 antibodies and by IL-1ra. BMMs were grown for 72 h in the presence of M-CSF (30 ng/ml) and RANKL (5 ng/ml). (A) Monolayers then were washed and incubated for additional 30 h with either TNF-α or IL-1β in the presence or absence of the corresponding antibodies or receptor antagonist (*p < 0.01 comparing to IL-1 treated; **p < 0.001 comparing to TNF-α treated). (B) Alternatively, for the additional 30 h, cells were incubated with 20 ng/ml of LPS in the presence or absence of anti-TNF-α or anti-TNFR-1 antibodies, or IL-1ra. Data are number of TRAP+ cells containing more than two nuclei (*p < 0.01 comparing with LPS-treated).

Figure FIG. 6.

Modulation of M-CSF receptor transcript and protein abundance and RANK transcript by LPS. BMMs were grown for 72 h in the presence of M-CSF (30 ng/ml) and RANKL (5 ng/ml). LPS was included for either the whole experiment or the last 48 h or 24 h as indicated. (A) Northern and (C) Western analyses were performed. (B) Alternatively, BMMs were grown in the presence of M-CSF and RANKL (20 ng/ml). LPS at different doses were included for the whole experiment. Northern analyses were performed. The number on the top of the M-CSFR and RANK bands indicate M-CSF/β-actin and RANK/β-actin ratios. These ratios, in the absence of LPS are given arbitrarily the value 1.


LPS is a well-known player in regulating inflammatory reactions.(34,35) For example, it has been identified as an important factor in the pathogenesis of periodontitis, a chronic inflammatory disease characterized by gingival inflammation and alveolar bone resorption.(36) LPS stimulates osteoclastic bone resorption in vivo(25,26) and in vitro in organ culture(27,28) and promotes osteoclast differentiation in whole bone marrow cell culture.(29) Osteoclastogenesis induced by LPS is mediated by TNF-α via its p55 receptor.(31) LPS induces RANKL expression in osteoblasts(24) and stimulates these cells to secrete IL-1, prostaglandin E2 (PGE2), and TNF-α, each of which seems to be involved in LPS-mediated bone resorption.(23) It also was shown that TNF-α activity accounts for calvarial bone resorption induced by local injection of LPS.(26) In all of these studies, mixed cell populations were used. The discovery of RANKL and specifically its ability to promote osteoclast differentiation in the absence of osteoblasts or stromal cells enables us to examine direct interactions of LPS with osteoclasts and their precursors.

Here, we show that LPS does not exhibit osteoclastogenic activity and, in fact, inhibits RANKL-induced osteoclast differentiation in the absence of stromal cells or osteoblasts. However, LPS promoted osteoclastogenesis in RANKL primed BMMs. This enhancement is RANKL independent, because OPG does not affect this LPS activity. Thus, depending on the experimental design, LPS inhibits, induces, or does not effect osteoclast differentiation. LPS administration into an animal results in an increased osteoclastogenesis, followed by a decrease to the level of the untreated animal. The latter phase was not investigated in details, but an inhibitory effect by LPS on osteoclast precursor cells could contribute to the reduction in osteoclastogenesis. In vivo, other cell lineages, especially stromal cells and osteoblasts, participate in mediating the LPS osteoclastogenic effect. This raises a question regarding the relative significance of the direct interactions of LPS with osteoclast lineage cells. LPS administration is an experimental manipulation mimicking the effect of bacterial infection. Bacteria, and especially in immune subjects, are targeted preferentially to cells of the mononuclear phagocytes lineage. Thus, it is expected that at least for some time, the local concentration of LPS in the vicinity of mononuclear phagocytes will be higher than in the vicinity of the osteoblast, enabling this cell to play a major role in mediating the LPS osteoclastogenic effect. It seems that LPS has multiple effects on osteoclast differentiation and that the relative importance of the different effects depends on their relative magnitudes in various physiological and/or pathological situations. Because the induction of osteoclastogenesis by LPS in BMMs requires prior priming of the cells (with RANKL), we postulate that LPS is capable of promoting this differentiation only in cells that already received a signal to differentiate toward osteoclastic phenotype. The cytokines TNF-α and IL-1β have been shown to play a role in osteoclastogenesis and bone resorption.(15–17,26,37) We have shown recently that TNF-α mediates at least in part the osteoclastogenic effect of RANKL.(18) We and others also showed that TNF-α is a potent activator of osteoclast differentiation in RANKL-primed cells.(18,19) These, together with the documented LPS increase of cytokine expression, including TNF-α and IL-1β,(26,31,38,39) prompted us to examine the involvement of these cytokines in the induction of differentiation by LPS. We found that although IL-1ra does not affect LPS osteoclastogenic activity, antibodies to TNF-α and to its type 1 receptors block this activity. Together with the known capability of TNF-α to induce differentiation in RANKL-primed cells(18,19) and the induction of the cytokine expression by LPS,(38,39) our data suggest that TNF-α mediates the osteoclastogenic activity of LPS. This conclusion is consistent with previous findings.(31) Unlike the previous studies, in this work we examine the effect of LPS on osteoclastogenesis in the absence of stromal cells, which are known to be modulated by LPS and to mediate at least in part LPS osteoclastogenic effect. Thus, here, we show that LPS induces osteoclast precursor to secrete TNF-α that promotes the differentiation of these cells in an autocrine mode of action. Despite the induction of TNF-α, LPS inhibits osteoclastogenic activity of RANKL if added together from the beginning of the BMM culture. Because the interactions of M-CSF and RANKL with their respective receptors M-CSF receptor and RANK are essential for osteoclastogenesis,(4–13) we examined the effect of LPS on the expression of these receptors. Indeed, we found that LPS decreases the levels of both M-CSF-receptor and RANK. It is of note that LPS has been shown previously to down-regulate M-CSF receptor expression in murine macrophage.(40,41) In addition, TGF-β stimulates osteoclastogenesis via up-regulation of RANK.(42) Thus, regulation of RANK and/or M-CSF-receptor is a useful mechanism to regulate osteoclast differentiation. Additional studies are required to reveal if LPS uses also other mechanisms to exert its “antiosteoclastogenic” effect. For example, LPS could modulate signal transduction pathways essential to osteoclastogenesis and/or stimulate release of unidentified yet inhibitory factors from the bone marrow osteoclast precursor cell. Recently, we have shown that oligodeoxynucleotides containing unmethylated CG dinucleotides in specific sequence contexts (CpG ODNs), representing bacterial DNA, exhibit similar opposite effects on osteoclastogenesis.(43) The two bacterial-derived proinflammatory factors seem to use a similar mechanism of stimulating osteoclastogenesis in RANKL-primed cells via TNF-α. However, their inhibition of RANKL-osteoclastogenic activity seems to differ, because CpG ODN does not affect RANK, but similarly to LPS reduces M-CSF receptors. The ability of bacterial-derived inflammatory factors such as LPS or CpG ODN to induce osteoclast-like cell formation through a direct effect on osteoclast precursors primed by RANKL might be a mechanism by which inflammation leads to osteolysis in diseases such as periodontitis and rheumatoid arthritis.


We thank F. Ross, Ph.D., and S. Teitelbaum, M.D. (Washington University), for the plasmid containing GST-RANKL, and C. Dinarello, M.D. (University of Colorado), for IL-1β and IL-1ra. This research was supported by a grant from The German-Israeli Foundation (GIF) for Scientific Research And Development. W.Z. is the recipient of a Kraut fellowship for graduate students.