TNF‐α has both stimulatory and inhibitory effects on mouse monocyte‐derived osteoclastogenesis

Phenotypically different osteoclasts may be generated from different subsets of precursors. To what extent the formation of these osteoclasts is influenced or mediated by the inflammatory cytokine TNF‐α, is unknown and was investigated in this study. The osteoclast precursors early blasts (CD31hiLy‐6C−), myeloid blasts (CD31+Ly‐6C+), and monocytes (CD31−Ly‐6Chi) were sorted from mouse bone marrow using flow cytometry and cultured with M‐CSF and RANKL, with or without TNF‐α. Surprisingly, TNF‐α prevented the differentiation of TRAcP+ osteoclasts generated from monocytes on plastic; an effect not seen with early blasts and myeloid blasts. This inhibitory effect could not be prevented by other cytokines such as IL‐1β or IL‐6. When monocytes were pre‐cultured with M‐CSF and RANKL followed by exposure to TNF‐α, a stimulatory effect was found. TNF‐α also stimulated monocytes’ osteoclastogenesis when the cells were seeded on bone. Gene expression analysis showed that when TNF‐α was added to monocytes cultured on plastic, RANK, NFATc1, and TRAcP were significantly down‐regulated while TNF‐αR1 and TNF‐αR2 were up‐regulated. FACS analysis showed a decreased uptake of fluorescently labeled RANKL in monocyte cultures in the presence of TNF‐α, indicating an altered ratio of bound‐RANK/unbound‐RANK. Our findings suggest a diverse role of TNF‐α on monocytes’ osteoclastogenesis: it affects the RANK‐signaling pathway therefore inhibits osteoclastogenesis when added at the onset of monocyte culturing. This can be prevented when monocytes were pre‐cultured with M‐CSF and RANKL, which ensures the binding of RANKL to RANK. This could be a mechanism to prevent unfavorable monocyte‐derived osteoclast formation away from the bone.

Osteoclasts, the multinucleated bone-resorbing cells, are crucial in bone diseases with excessive bone loss. Severe bone destruction occurs when the equilibrium of osteoclast and osteoblast activity is disturbed. Osteoclasts arise from monocytic precursors under the influence of M-CSF and RANKL. Next to these cytokines, TNF-α has been shown to stimulate osteoclast generation and bone resorption both in vitro (Thomson, Mundy, & Chambers, 1987) and in vivo (König, Mühlbauer, & Fleisch, 1988).
Recently, studies have reported that these three precursor subsets respond differently to the growth factor M-CSF (De Vries et al., 2015) and the cytokine IL-1β .
Myeloid blasts were found to respond the fastest to M-CSF and RANKL  and early blasts was the only population that proliferated under the influence of IL-1β . How these three myeloid lineage-osteoclast precursor subsets respond to another crucial inflammatory cytokine, TNF-α, is unknown and is the topic of the present study. We investigated the effect of TNF-α on the osteoclastogenic potential of different mouse bone marrow-derived osteoclast precursors.
Cells were cultured on plastic or on 650 μm-thick bone slices and culture media were refreshed every 3 days. After 5 days the cultures on plastic plates were stopped by fixation in 4% PBSbuffered formaldehyde for TRAcP analysis, or lysed in RNA lysis buffer (Qiagen, Hilden, Germany) at day 0, 3, and 5 for RNA isolation. Bone slices were stored in water after 6 days of culture and culture media was collected for CTX assay.
The TRAcP + multinucleated cells were categorized as cells containing 3-10, and >10 nuclei, and the number of cells in each category was counted using a combination of light and fluorescence microscopy (Leica DFC320; Leica Microsystems, Wetzlar, Germany).

| Bone resorption
In order to evaluate bone resorption, formation of resorption pits on bone slices was analyzed by Coomassie brilliant blue staining.
After 6 days of culture, bone slices were washed with water and the cells were removed by sonication for 30 min in 10% ammonia (Merck, Darmstadt, Germany) on ice. To reduce staining background the bone surface was incubated in saturated alum for 10 min and subsequently washed with water. Resorption pits were stained with Coomassie brilliant blue (Pharmacia, Uppsala, Sweden) and visualized by light microscopy (Leica DFC320). In addition, CTX concentrations of the culture supernatants were determined using CrossLaps® for culture ELISA (Immunodiagnostic Systems Limited, Frankfurt am Main, Germany). Culture supernatants were collected after 6 days and CTX assay was conducted following the manufacturer's instructions.

| Quantitative RT-PCR
The procedure of RT-PCR was described in detail in a previous paper . Quantitative expression of TRAcP, NFATc1, RANK, TNF-αR1, and TNF-αR2 were determined. Primers were designed using Primer Express 3.0. TCTggACCATCTTCTTgCTgA. The relative expression of each gene was calculated as 2 −ΔCt , ΔCt = C gene of interest −Ct PBGD .
The results were shown as fold increased, normalized by the same gene expression at day 0.

FIGURE 2
Effect of TNF-α on osteoclastogenesis by different osteoclast precursors on plastic. (a-c) Early blasts (1st row), myeloid blasts (2nd row), and monocytes (3rd row) were cultured under three different culture conditions: (a) condition 1 (control), with 30 ng/ml M-CSF and 20 ng/ml RANKL for 5 days; (b) condition 2, with 30 ng/ml M-CSF, 20 ng/ml RANKL and 10 ng/ml TNF-α for 5 days; (c) condition 3, first cultured with 30 ng/ml M-CSF and 20 ng/ml RANKL for 3 days, then cells were further cultured with 30 ng/ml M-CSF, 20 ng/ml RANKL and 10 ng/ml TNF-α for 2 days. Cells were stained for TRAcP activity and nuclei were counterstained by DAPI. Osteoclasts were recognized as TRAcP + multinucleated cells (purple) and nuclei were visualized as blue. Note the absence of any multinucleated cells in monocyte cultures in condition 2. Scale bar = 100 μm.

| Statistical analysis
All data were analyzed from six mice (n = 6) and GraphPad Prism (version 6.00; GraphPad Software, LaJolla, CA) was used for the statistical analysis. One-way ANOVA followed by Tukey-Kramer's multiple comparison was used for multiple comparisons, and t-test was used for two comparisons. Data are displayed as mean ± SD and p < 0.05 was considered as significant difference. . As a control, we also cultured whole bone marrow cells with or without TNF-α, and such inhibitory effect was not seen (data not shown). There was no significant difference between the two concentrations (10 and 100 ng/ml) of TNF-α (Figure S1a-c), therefore we chose 10 ng/ml TNF-α for the subsequent experiments. Lam et al. (2000) reported a similar TNF-α-induced inhibitory effect on osteoclast formation by purified murine myeloid cells and they found that the inhibitory effect could be overcome by RANKL exposure. In their study, TNF-α induced osteoclast formation was maximal when the cells were exposed to RANKL for 2-4 days before adding TNF-α (Lam et al., 2000). Therefore, in the present study we cultured the precursor cells with M-CSF and RANKL for 3 days before adding TNF-α (condition 3) to find out if this condition could alter the response to the cytokine.  and no TRAcP + cells were seen in monocyte cultures ( Figure S1d). This indicates that RANKL is essential for the TNF-α induced osteoclast formation by monocytes.
3.2 | IL-1β and IL-6 were not able to prevent the inhibitory effect of TNF-α on monocytes' osteoclastogenesis Next to TNF-α also two other inflammatory cytokines, IL-1β and IL-6, have been shown to stimulate osteoclastogenesis. We wondered whether IL-1β or IL-6 might influence the effect of TNF-α as shown above. In the presence of IL-1β or IL-6, high numbers of TRAcP + cells  (Figure 3b). Monocytes proved to be insensitive to IL-1β: no increased osteoclastogenesis was observed here; a finding in line with previous findings . However, whenever TNF-α was added at the onset of the monocyte cultures, osteoclastogenesis was completely inhibited (see conditions 2-6 in Figure 3). This was apparent both in the absence and in the presence of IL-1β and/or IL-6, indicating that the latter two cytokines were not able to prevent the inhibitory effect of TNF-α. In the presence of TNF-α most cells proved to be TRAcP-negative (Figure 3a, conditions 2-6). This inhibitory effect was not found with early blasts and myeloid blasts ( Figure S2). Most of these cells became TRAcP + multinucleated cells under all conditions ( Figure S2a,c). With these two cell lineages, no inhibitory effect was found in any of the conditions ( Figure S2b,d).

| The inhibitory effect of TNF-α did not occur when monocytes were seeded on bone
In addition to seeding the precursors on plastic, we also seeded them

| TNF-α stimulated bone resorption
To determine the level of bone resorption, the bone slices were stained with Coomassie brilliant blue to visualize pit formation, and the culture media were analyzed for the level of CTX (

| TNF-α down-regulated osteoclast-related genes and up-regulated TNF-α receptors by monocytes on plastic
In an attempt to offer an explanation for the inhibitory effect of TNF-α on osteoclastogenesis of monocytes, we analyzed gene expression of osteoclast-related genes, TRAcP, NFATc1, and RANK as well as the TNF-α receptors, TNF-αR1, and TNF-αR2. The expression of TRAcP The expression of TRAcP, NFATc1, and RANK by myeloid blasts and early blasts was not affected by TNF-α ( Figure S3a-c, f-h). No significant differences were seen for any of the tested genes in early blast cultures ( Figure S3a-e). In myeloid blasts an up-regulation of TNF-αR2 was seen ( Figure S3j). At the onset of the culture period monocytes showed a significantly higher expression of TNF-αR1 as well as TNF-αR2 than the other two subsets (Figure S3l-m). All three subsets expressed comparable RANK at the start of the culture period ( Figure S3k).

| TNF-α diminished the uptake of fluorescently labeled RANKL in monocyte cultures
In order to assess whether TNF-α could interfere with binding of RANKL to RANK, FACS analysis was used to investigate the interaction of RANKL with the membrane-bound receptor RANK. Monocytes and | 3281 myeloid blasts were cultured for 3 days with M-CSF and RANKL, with or without TNF-α before labeling. The expression of RANK did not show a significant difference with or without TNF-α in any of the cultures (data not shown). By adding TAMRA-conjugated RANKL, we were able to analyze the binding of RANKL to its receptor RANK on the plasma membrane. TAMRA-conjugated RANKL labeling of myeloid blasts was not influenced by the presence of TNF-α ( Figure 7A). In monocyte cultures, we found two populations with a different binding of labeled RANKL: one with a low binding (RANKL low ) showed on the left hand side of Figure 7b, and one with a high binding (RANKL high ), seen on the right hand side of the graph (Figure 7b). In the control group, the RANKL high sub-population was almost two times higher (66 ± 6%) than the RANKL low sub-population (34 ± 6%). In the group with TNF-α, this pattern was reversed. The percentage of RANKL high sub-population was much lower (37 ± 4%) than that of RANKL low subpopulation (63 ± 4%). Thus, TNF-α decreased the binding of fluorescently labeled RANKL; an effect particularly apparent in monocytes.

| DISCUSSION
In the present study we found that TNF-α had the capacity to prevent osteoclastogenesis of one specific precursor subset: bone marrow derived monocytes (CD31 − Ly-6C hi ). Our results indicate that TNF-α interferes with the binding of fluorescently labeled RANKL in monocytes. This likely results in the impairment of the RANK-induced signaling pathway for osteoclastogenesis. However, this inhibitory effect can be prevented in different ways.
First, by the presence of M-CSF and RANKL prior to the addition of TNF-α.
Second, by seeding monocytes on bone. These phenomena were not found with early blast cultures nor with myeloid blast cultures. Our findings indicate a multifunctional role of TNF-α on osteoclastogenesis of a particular subset of osteoclast precursors, the monocytes.

| A model for TNF-α on monocytes' osteoclastogenesis
The most intriguing finding of this study was the inhibitory effect of TNF-α on osteoclastogenesis of monocytes when cultured on plastic, whereas it increased osteoclastogenesis when the cells were first exposed to M-CSF and RANKL or, alternatively, seeded on bone. Under the former (on plastic) condition we found a decreased mRNA level of RANK by the monocytes, and an increased population of monocytes with a lower level of fluorescently labeled RANKL. The protein level of RANK proved to be unaffected. Thus, it is likely that the RANKL/RANK signaling pathway was affected. Our findings indicate that TNF-α can express both inhibitory and stimulatory effects on monocytes' osteoclastogenesis. A hypothetical model that offers an explanation for our findings is shown in Figure 8.
When monocytes are cultured with M-CSF and RANKL, RANKL will bind to RANK expressed on the membrane of monocytes, leading to upregulation of osteoclast-related genes, such as NFATc1. This will result in osteoclast differentiation (Figure 8a). However, when TNF-α is present from the onset together with M-CSF and RANKL, it appears that RANK signaling is not well triggered (Figure 8b). This results in a down-regulated gene

| Relationship between TNF-α and RANK/RANKL
Clearly the relationship between TNF-α and RANKL in osteoclastogenesis is complex. Kobayashi et al. (2000) showed that the stimulatory effect of TNF-α on osteoclastogenesis is M-CSF-dependent rather than RANK/RANKL-dependent. Studies by Kim et al. (2005) demonstrated that TNF-α together with TGF-β stimulates osteoclastogenesis independent of the RANKL-RANK axis. However, Zhang et al. (2001) showed that TNF-α synergistically cooperates with RANK/RANKL-induced osteoclastogenesis and an overlapping signaling pathway of RANKL and TNF-α was proposed. Others stated that RANKL-and TNF-induced osteoclastogenesis share a similar intracellular signaling pathway, including c-fos and NFATc1 (Yamashita et al., 2007). Lam et al. (2000) showed that a basal level of RANKL is necessary for TNF-α induced osteoclast formation: TNF-α alone, or with M-CSF, at any concentration, failed to stimulate osteoclast differentiation. Depending on the time point, TNF-α proved to inhibit or potentiate RANKL-mediated osteoclastogenesis. TNF-α inhibited osteoclastogenesis by purified myeloid cell cultures when it was present with RANKL at the onset of the culture, but the cytokine stimulated osteoclastogenesis when it was added 2-4 days after RANKL priming (Lam et al., 2000). This coincides with the present study which shows adding TNF-α at a later time point after RANKL and M-CSF priming stimulates monocyte-derived osteoclast formation.
In order to prove that TNF-α is closely associated with RANK/RANKL interaction, we cultured the three osteoclast precursor subsets with only M-CSF for 3 days before adding TNF-α together with M-CSF and RANKL. Under these conditions only very few TRAcP + multinucleated cells were formed ( Figure S1d). This indicates that M-CSF alone is not sufficient; a RANKL exposure is needed for the TNF-α-induced osteoclastogenesis. To assess if TNF-α can substitute RANKL in osteoclastogenesis, we also cultured the three osteoclast precursor subsets with M-CSF and TNF-α ( Figure S1e,f). Interestingly, high number of TRAcP + mononuclear cells were formed in early blast and myeloid blast cultures ( Figure S1e), but only a few of them were multinucleated ( Figure S1f). Almost no TRAcP + cells were present in monocyte cultures ( Figure S1e,f). This shows that TNF-α can partly substitute RANKL in early blasts' and myeloid blasts' osteoclastogenesis, but not with monocytes.

| Bone's stimulatory effect on osteoclastogenesis
The inhibitory effect of TNF-α on osteoclastogenesis by monocytes was prevented not only with an M-CSF and RANKL pre-incubation, but also by seeding monocytes on bone slices. Under these conditions even higher numbers of osteoclasts were formed with TNF-α. The data indicate that attachment of the monocytes to the bone surface changes the sensitivity of these cells to TNF-α. Although we do not have an explanation for such an effect, several authors have shown that an interaction of osteoclast precursors with bone greatly affects release of compounds like IL-1 (Yao, Xing, Qin, Schwarz, & Boyce, 2008). An increased level of IL-1 has been shown to stimulate TNF-induced osteoclastogenesis (Wei, Kitaura, Zhou, Ross, & Teitelbaum, 2005). In our study, exposure to IL-1β could not overcome the inhibitory effect found in monocytes on plastic ( Figure 3).
Therefore, key molecules of bone, or released by osteoclasts that were grown on bone, remain to be elucidated.
When we compare the number of osteoclasts per surface area when cells were cultured either on bone slices or on plastic plates, the number of osteoclasts formed on the bone slices were shown to be significantly higher than the number formed on plastic ( Figure S1g). This is in line with our previous study in which we showed higher numbers of TRAcP + multinucleated cells on bone than on plastic . Of considerable interest was the finding that cells present on plastic next to the bone slices did form osteoclasts. This counts also for monocyte cultures (Figure 4f in monocyte panel). Since monocytes seeded on plastic in the absence of bone were inhibited in their capacity to generate osteoclasts, this finding suggests the presence of one or more compounds prevented the TNF-α induced inhibition. Such compounds could be released either by the monocytes due to their interaction with the bone surface or by the bone slice itself. Bone tissue, as a natural matrix for osteoclast formation, was shown to overcome osteoclastogenesis-insensitivity of certain osteoclast precursor subsets (De Vries et al., 2015). It was demonstrated that expression of NFATc1, being a key molecule in the regulation of osteoclastogenesis, was stimulated on bone. Osteoclast precursors are likely to interact with a number of stimulatory components elicited FIGURE 9 Physiological explanation of the effect of TNF-α on monocytes' osteoclastogenesis in vivo. TNF-α inhibits monocytes' osteoclastogenesis at sites at a distance from bone. When monocytes interact with the bone surface, TNF-α stimulates RANKL-induced osteoclast formation from the bone matrix, for example transforming growth factor-β (Mundy & Bonewald, 1990;Zwerina et al., 2004) and proteins such as osteopontin and bone sialoprotein (Fisher, Torchia, Fohr, Young, & Fedarko, 2001;Qin, Baba, & Butler, 2004, Yao et al., 2008. Molecules such as osteoclast-associated receptor (OSCAR) expressed by preosteoclasts, were shown as a collagen receptor that stimulates osteoclastogenesis (Barrow et al., 2011). All of these molecules have the potential to stimulate osteoclast generation. These findings suggest that the release of factors by osteoclast precursors or by the bone matrix may overcome the TNF-α inhibitory effect. Further investigations are needed to explore the nature of the molecule(s) released.

| Monocytes: a distinct osteoclast precursor cell lineage?
This study showed that from all the different bone marrow derived osteoclast precursors, only the monocyte-derived osteoclast formation could be abolished by TNF-α, a phenomenon that does not apply to early blast-or myeloid blast-derived osteoclast formation. A previous study already addressed the differences among the three subsets in osteoclast differentiation ) and showed myeloid blasts being the most responsive cells to M-CSF and RANKL ). Both myeloid blasts and early blasts but not monocytes were shown to be sensitive to IL-1β . In line with the inhibitory effect of TNF-α which was only found in monocyte cultures in this study, Hayashi et al. (2003) showed that splenic osteoclastogenesis was completely inhibited by TNF-α; an effect not found with bone marrow cells. Since the spleen stores about half of the number of monocytes present in the body (Swirski et al., 2009), our finding of the inhibitory effect of TNF-α on bone marrow monocytes suggests a similarity between spleen monocytes and those present in the marrow. Comparing the three myeloid precursor subsets, monocytes showed a different distribution of the receptors of TNF-α: monocytes have a significantly higher mRNA level of TNF-αR1 and TNF-αR2 on day 0 compared to myeloid blasts and early blasts ( Figure S3l,m). This suggests that monocytes are the cell type more sensitive to TNF-α, and this sensitivity has a close relationship to RANKL exposure.
Is there a physiological explanation for a cell type that becomes osteoclastogenesis-insensitive to a cytokine like TNF-α? Under pathological conditions the cytokine is present throughout the body and this could in theory result in the formation of osteoclasts anywhere in the body. After all TNF-α as well as the monocytes are present at many different sites, and RANKL is also present in serum (Findlay & Atkins, 2011). We propose that monocytes, not in the vicinity of bone, are insensitive to the cytokine, therefore, osteoclast formation does not occur at these sites away from the bone (Figure 9). When these monocytes attach to the bone surface, they become sensitive to TNF-α and osteoclastogenesis is stimulated. We hypothesize that such a versatile role of the cytokine is meaningful in modulating the process of osteoclastogenesis.

CONFLICTS OF INTEREST
The authors declare no conflict of interest.