LDL uptake-dependent phosphatidylethanolamine translocation to the cell surface promotes fusion of osteoclast-like cells

Multinucleated osteoclasts are the cells responsible for physiological and pathological bone resorption and belong to the monocyte/macrophage cell lineage. Although many systemic hormones and local cytokines regulate osteoclast differentiation (Teitelbaum, 2007), the receptor activator of nuclear factor κB ligand (RANKL, also known as TNFSF11) and macrophage colony-stimulating factor (M-CSF, also known as CSF1) are essential cytokines in osteoclastogenesis (Yasuda et al., 1998). RANKL signals produced by bone marrow stromal cells or osteoblasts are introduced into osteoclast precursors via a receptor of RANKL (RANK, also known as TNFRSF11A) on the plasma membrane of osteoclast lineage cells, leading to the activation of nuclear factor κB (NF-κB) and MAPKs (Asagiri and Takayanagi, 2007; Mizukami et al., 2002). Activation of the signaling molecules then induces the expression of NFATc1 (Takayanagi et al., 2002), a master transcription factor for osteoclast differentiation. Therefore, the expression of osteoclast function-related molecules, including tartrate-resistant acid phosphatase (TRAP), cathepsin K and fusion-related proteins such as ATP6v0d2 and DC-STAMP, is induced (Inaoka et al., 1995; Lee et al., 2006; Kim et al., 2010; Yagi et al., 2005; Mensah et al., 2010).

Accumulating epidemiological evidence has demonstrated that dyslipidemia is a risk factor not only for atherosclerosis and vascular calcification but also for osteoporosis, indicating a tight correlation between skeletal homeostasis and lipid metabolism (Pinals and Jabbs, 1972; Banks et al., 1994; Yamaguchi et al., 2002; Alagiakrishnan et al., 2003). We have thus far demonstrated that differentiation of osteoclast-like cells (OCLs) is impacted by the status of intracellular and extracellular low-density lipoprotein (LDL) or cholesterol (Hada et al., 2012; Okayasu et al., 2012). Caveolin-1, a principal scaffolding protein of cholesterol and sphingolipid-enriched flask-shaped caveolae and lipid rafts in the plasma membrane, expression in OCLs was induced by RANKL signaling during osteoclastogenesis, and the removal of cholesterol from the plasma membrane caused abnormal signaling for osteoclastogenesis (Hada et al., 2012). Furthermore, the abrogation of LDL receptor (LDLR) impaired OCL formation because of defects in cell–cell fusion of OCLs; thereby, the trabecular bone mass was increased, indicating that uptake of LDL is required for osteoclastogenesis (Okayasu et al., 2012). On the other hand, lectin-like oxidized LDL receptor-1 (LOX-1, also known as OLR1) recognizes extracellular oxidized LDL and is crucially involved in the pathogenesis of atherosclerosis, a multifactorial inflammatory disease of the vessel wall that can cause heart disease and stroke (Yoshimoto et al., 2011). The deletion of LOX-1 promoted OCL formation as a result of increased cell–cell fusion, thereby decreasing bone volume (Nakayachi et al., 2015). Although LDLR and LOX-1 oppositely affect OCL cell–cell fusion, neither receptor affects the expression of osteoclast differentiation molecules, such as NFATc1 and cathepsin K (Nakayachi et al., 2015). These results indicate that LDLR and LOX-1 modulate OCL cell–cell fusion through a cascade other than the well-known osteoclast differentiation signaling pathway via the RANKL/RANK complex. However, the molecular mechanism underlying the actions of LDLR and LOX-1 on cell–cell fusion remains unknown.

It has been recently demonstrated that dynamic translocation of phospholipids, principally phosphatidylethanolamine (PE) and phosphatidylserine (PS), from the inner to outer leaflet of the plasma membrane plays an important role in cell–cell fusion during OCL maturation (Irie et al., 2017; Verma et al., 2018). Cell–cell fusion is a fundamental physiological event for egg–sperm fusion for fertilization and for the generation of multinucleated placental syncytiotrophoblasts, myogenesis and osteoclastogenesis (Oren-Suissa and Podbilewicz, 2007; Xing et al., 2012); moreover, cell–cell fusion is involved in the pathological formation of macrophage foreign body giant cells (Xing et al., 2012). Cell–cell fusion occurs in several steps. First, cells recognize partner cells and then the outer leaflets of the plasma membrane of the two cells come into contact with each other. This is followed by formation of a hemifusion stalk and a hemifusion diaphragm. Finally, a fusion pore is generated to complete cell–cell fusion (Chernomordik and Kozlov, 2008). During the cell–cell fusion process, dynamic changes in the distribution of phospholipids, the principal components of membrane bilayers, occur between the outer and inner leaflets of the membrane. The kinetic alteration (flip-flop) is mediated by three different types of enzymes (Montigny et al., 2016): flippases that mediate the transport of phospholipids from the outer to inner leaflet of the plasma membrane, floppases that arbitrate migration from the inner to outer leaflet and scramblases that catalyze the bidirectional distribution change of phospholipids. Flippases, floppases and scramblases include the P-type ATPase superfamily (Zhou and Graham, 2009), ATP binding cassette (ABC) transporters (Hankins et al., 2015) and anoctamin/TMEM16 family (Suzuki et al., 2010), respectively.

To explore the molecular mechanism underlying the effects of LDLR and LOX-1 on cell–cell fusion of OCLs, we used LDLR single KO (sKO), LOX-1 sKO and LDLR/LOX-1 double KO (dKO) mice in this study. We found that translocation of PE to the outer leaflet of the plasma membrane from the inner leaflet is pivotally dependent on LDL uptake through LDLR and that the translocation mediated by the ATP binding cassette G1 (ABCG1) transporter is required for the cell–cell fusion of OCLs. These results provide novel insight into the mechanism of osteoclast cell–cell fusion.

We generated LDLR/LOX-1 dKO mice by cross-mating LDLR sKO and LOX-1 sKO mice. Using LDLR/LOX-1 dKO mice, we first examined the effect of abrogation of both LDLR and LOX-1 genes on in vitro OCL formation in response to M-CSF and sRANKL (Fig. 1A). The number of TRAP-positive multinucleated OCLs with more than three nuclei in the culture of LDLR/LOX-1 dKO monocytes treated with M-CSF and sRANKL was equivalent to that in the culture of wild-type (WT) cells (Fig. 1B, left panel). However, the area of LDLR/LOX-1 dKO multinucleated OCLs was decreased by approximately 20% of the area of WT osteoclasts (Fig. 1B, middle panel). Consistent with the decreased area of OCLs, the fusion index, which was expressed as the mean number of nuclei per multinucleated OCL, in the cultures of LDLR/LOX-1 dKO monocytes was also reduced to approximately 25% of that of WT cells (Fig. 1B, right panel). Because the fusion index indicates the frequency of cell fusion to form multinucleated OCLs, the decreased fusion index indicated that the cell–cell fusion of mononuclear preosteoclasts (pre-OCLs) into mature multinucleated OCLs was impaired by the deletion of both LDLR and LOX-1 genes. The reduction in cell–cell fusion is consistent with the decrease in cell–cell fusion as a result of LDLR single deletion (Okayasu et al., 2012); this finding suggests that LDLR is required for cell–cell fusion of OCLs. However, the mRNA levels of osteoclast differentiation-related molecules, such as c-Fos, NFATc1, TRAP and cathepsin K, and fusion-related proteins, including DC-STAMP, OC-STAMP and ATP6v0d2, in LDLR/LOX-1 dKO osteoclastic cells were not altered compared with the mRNA levels in WT OCLs (Fig. S1).

To address the role of LDLR in osteoclastogenesis, we next compared the uptake of LDL into the cells using four different genotypes of OCLs: WT, LDLR/LOX-1 dKO, LOX-1 sKO and LDLR sKO (Fig. S2A). The OCLs were incubated with pHrodo Red-labeled LDL following starvation of extracellular and intracellular LDL in the serum. An obvious red-color signal indicating the uptake of exogenous LDL was observed in WT OCLs. Although LDL uptake into LOX-1 sKO OCLs was lower than that of WT OCLs, significant uptake was confirmed. In contrast, the uptake of LDL into LDLR sKO and LDLR/LOX-1 dKO OCLs was small, indicating that the LDL uptake of OCLs is dependent on the presence of LDLR.

Furthermore, neutral lipids, including cholesterol, that are stained with Oil Red O solution were evidently accumulated in WT and LOX-1 sKO OCLs (Fig. S2B). In contrast, such accumulation was not observed in LDLR/LOX-1 dKO and LDLR sKO OCLs.

Next, we determined the amount of cell surface cholesterol using cell membrane-impermeable θ-toxin-derived domain 4 (D4), which specifically binds cholesterol (Shimada et al., 2002). Cholesterol was abundant on the cell surface of WT and LOX-1 sKO OCLs, and the amount of cholesterol was abolished by pretreatment with methyl-β-cyclodextrin (MβCD), which binds and removes cholesterol from the outer leaflet of the plasma membrane (Fig. 2A,B). The cell surface cholesterol level of LDLR/LOX-1 dKO and LDLR sKO OCLs was faint, and the decreased amount was approximately 20% of that of WT and LOX-1 sKO cells (Fig. 2B), suggesting that the status of intracellular cholesterol in OCLs is crucially determined by the level of LDLR.

It has been recently reported that PE dynamics of the plasma membrane are required for cell–cell fusion of OCLs (Irie et al., 2017). Therefore, we next determined the PE distribution on the cell surface of WT and LDLR/LOX-1 dKO OCLs using SA–biotin-conjugated Ro09-0198 (SA–Bio-Ro), which specifically binds PE (Choung et al., 1988). Because SA–Bio-Ro is cell membrane impermeable, the probe has been used for the detection of PE residing in the outer, but not inner, leaflet of the plasma membrane. However, in a variety of cells, PE is abundantly distributed in the inner leaflet of the plasma membrane (van Meer et al., 2008). Mononuclear WT pre-OCLs prior to maturation possessed numerous filopodia structures, which were connected to the filopodia of adjacent cells. Furthermore, PE was concentrated and restricted in the filopodia (Fig. S3A,B), indicating that PE accumulates on the surface of filopodia of pre-OCLs as a result of translocation of PE from the inner leaflet. In contrast, the distribution of cholesterol was ubiquitous on the whole cell surface.

PE translocation into the outer leaflet of the plasma membrane of WT OCLs occurred from the early stage of osteoclast differentiation, and as osteoclast multinucleation proceeded, the cell surface PE level gradually decreased (Fig. 3A-C). In all cases, the cell surface PE level of LDLR/LOX-1 dKO OCLs was significantly lower than that of WT OCLs (Fig. 3A-C).

The movement of lipid substrates, including various phospholipids and cholesterol from the inner to outer leaflets of the plasma membrane, is mediated by a variety of ABC transporters (Coleman et al., 2013). Irie et al. (2017) reported that ABCB4 and ABCG1 are required for cell–cell fusion of OCLs to the cell surface through PE dynamics. Thus, we examined the mRNA expression profile of the ABC transporters in four different types of genotypic OCLs. On day 1 after sRANKL addition, the level of Abcg1 mRNA in WT OCLs was attenuated to approximately 70% of the level before addition of sRANKL; however, the reduced level was maintained at approximately 50-70% of the initial level during OCL formation (Fig. 4A). In LDLR/LOX-1 dKO OCLs, the initial (on day 0) expression level of Abcg1 mRNA was approximately 60% lower than that of WT OCL precursors (Fig. 4A). The expression level was greatly reduced to approximately 1% of the initial level by day 1 of incubation with sRANKL, and the lower level was sustained throughout OCL formation (Fig. 4A). The difference in Abcg1 mRNA levels between WT and LDLR/LOX-1 dKO OCLs was significant during OCL differentiation. Although the Abcg1 mRNA expression profile in LOX-1 sKO OCLs was comparable to that of WT cells, the Abcg1 mRNA level in LDLR sKO OCLs was severely decreased by sRANKL stimulation, consistent with the reduction in LDLR/LOX-1 dKO OCLs (Fig. 4B). On the other hand, the Abcb4 mRNA level was elevated with OCL differentiation, and the mRNA expression level in LDLR/LOX-1 dKO OCLs was equivalent to that in WT OCLs (Fig. 4C). In addition, we recognized no significant difference in the Abcb4 mRNA expression levels among WT, LOX-1 sKO and LDLR sKO OCLs (Fig. 4D). In addition to ABCG1 and ABCB4, we also determined the mRNA levels of some other ABC transporters. During OCL differentiation, there was no significant difference in the mRNA expression profiles of ABCA1, ABCA2, ABCA3 and ABCB1β among the four different genotypic OCLs (Fig. S4). The expression of Abca4 and Abcg4 mRNAs was not detected in the OCLs. These results indicate that the large reduction in Abcg1 mRNA expression in LDLR/LOX-1 dKO and LDLR sKO OCLs coincides with the decrease in the cell–cell fusion of OCLs. Therefore, we again determined the PE level in the outer leaflet of the plasma membrane of LDLR sKO OCLs. We found that the PE level on the cell surface of LDLR sKO OCLs was markedly lower than that of WT cells, similar to the reduction in LDLR/LOX-1 dKO cells (Fig. 5).

Based on the above findings, we next examined the effect of Abcg1 and Abcb4 gene knockdown on OCL formation and cell–cell fusion by means of siRNA transduction. Electroporation of Abcg1 siRNA decreased the expression level of Abcg1 mRNA in WT OCLs to less than 10% compared with scramble RNA transduction, whereas the mRNA levels of Abcb4 were not altered (Fig. 6A). In contrast, Abcb4 siRNA transduction caused a significant reduction in Abcb4, but not Abcg1, mRNA expression (Fig. 6A). We found that the number, cell size and cell–cell fusion index of the OCLs formed in culture were attenuated by knockdown of Abcg1 or Abcb4 mRNAs in WT OCLs (Fig. 6B,C), whereas the mRNA expression of osteoclast differentiation-related molecules was not changed (Fig. S5).

Furthermore, we confirmed the role of ABCG1 in cell–cell fusion of OCLs using Abcg1 sKO mice. As shown in Fig. 7A, the number of OCLs formed in the culture of Abcg1 sKO OCL precursors in response to M-CSF and sRANKL decreased to 75% of that of WT OCLs. The average area and the fusion index were also reduced by approximately 50% of those in WT OCLs. Consistent with the decrease in OCL formation, PE translocation to the outer leaflet of the plasma membrane was also diminished by Abcg1 gene deletion, indicating that ABCG1 abrogation caused impaired cell–cell fusion (Fig. 7B). The mRNA levels of osteoclast differentiation-related molecules in the Abcg1 sKO osteoclastic cells were equivalent to the levels in WT cells during OCL formation (Fig. S6).

Abcg1 is well known as a target gene of nuclear liver X receptor (LXR) (Sabol et al., 2005). Thus, we examined the effects of the LXR antagonist GSK 2033 (Griffett and Burris, 2016) and LXR agonist GW 3965 (Joseph et al., 2002) on OCL formation and on the mRNA levels of Abcg1 in WT OCL culture in the presence of M-CSF and sRANKL. LXRα and LXRβ mRNAs were constantly expressed during OCL differentiation from WT OCL precursors (Fig. 8A). The addition of GSK 2033 to the culture of WT OCLs greatly attenuated OCL formation and decreased expression of Abcg1 mRNA (Fig. 8B-D). In contrast, the addition of GW 3965 to the culture of LDLR sKO OCLs restored the expression of Abcg1 mRNA up to approximately 140% of the expression level of WT OCLs. At the same time, the OCL formation also augmented and the increased number of OCLs became equivalent to the level of WT OCL formation (Fig. 8E-G). These results suggest the involvement of LXR in the cell–cell fusion of OCLs associated with PE translocation to the outer leaflet of the plasma membrane.

In this study, we demonstrated using LDLR sKO, LDLR/LOX-1 dKO and Abcg1 sKO mice that the uptake of LDL into OCL precursors via LDLR is pivotal for PE translocation from the inner to outer leaflet of the plasma membrane, in particular, of the filopodia; thereby, the multinucleation of OCLs through promoting cell–cell fusion proceeds. The LDLR sKO and LDLR/LOX-1 dKO mice consistently exhibited reduced influx of LDL into the cells and the intracellular depletion of neutral lipids, including cholesterol, suggesting impaired intracellular de novo cholesterol biosynthesis. Consistent with the depleted status of intracellular cholesterol, the translocation of PE on the cell surface was impaired, and cell–cell fusion was subsequently reduced. The mRNA expression of ABCG1, a floppase that transfers phospholipids from the inner to outer leaflet of the plasma membrane, was greatly downregulated in response to sRANKL in the LDLR sKO and LDLR/LOX-1 dKO OCL precursors, which coincided with reduced cell–cell fusion. In addition, Abcg1 sKO mice showed reduced PE translocation on the cell surface and decreased cell–cell fusion of OCLs, similar to the LDLR sKO and LDLR/LOX-1 dKO OCL precursors. Taken together, this study for the first time indicates that a novel cascade (LDLR∼ABCG1∼PE translocation to cell surface∼cell–cell fusion) plays an important role in OCL maturation, as described in Fig. S7.

So far, several studies have reported a close relationship between bone and lipid metabolisms. Both elevated levels of LDL and decreased levels of high-density lipoprotein in plasma are correlated with low bone mass, and high triglyceride levels are associated with a decreased incidence of vertebral fractures in postmenopausal women (Yamaguchi et al., 2002). Patients with low bone densities and osteoporosis are characterized by elevated plasma lipid levels, more severe coronary atherosclerosis and an increased risk of stroke (Barengolts et al., 1998). Furthermore, an in vivo study demonstrated that a mouse strain with susceptibility to atherosclerosis (C57BL/6) lost a greater amount of bone mass in response to a high-fat diet than atherosclerosis-resistant strains (Parhami et al., 2001). Regarding osteoclastogenesis, we previously reported that deletion of LDLR causes reduced OCL formation as a result of impairment of cell–cell fusion of OCLs (Okayasu et al., 2012), whereby trabecular bone mass is increased due to the decrease in bone resorption. In contrast, the abrogation of LOX-1 promoted OCL formation following augmentation of cell–cell fusion, resulting in decreased bone volume (Nakayachi et al., 2015). Thus, the actions of LDLR and LOX-1 on osteoclastogenesis are opposing; that is, LDLR and LOX-1 are positive and negative regulators for bone resorption, respectively. However, the detailed molecular mechanism underlying the actions of LDLR and LOX-1 remained obscure. Initially, we presumed that the LDLR/LOX-1 dKO mouse could be a control animal suitable for elucidating the mechanisms of both receptors and, therefore, we generated dKO mice. However, the phenotype of the LDLR/LOX-1 dKO for OCL formation was similar to that of LDLR sKO, resulting in reduced osteoclast cell–cell fusion associated with decreased PE translocation on the cell surface. As a result, the impact of LDLR deletion exceeded that of LOX-1 abrogation regarding cell–cell fusion of OCLs.

Osteoclasts are a unique cell in the body and are responsible for bone resorption; mononuclear preosteoclasts are required for cell–cell fusion to acquire full osteoclast functions. Extensive studies have thus far revealed that several proteins, such as CD9, ATP6v0d2, DC-STAMP, OC-STAMP and dynamin, are involved in osteoclast cell–cell fusion during osteoclast differentiation in response to RANKL (Lee et al., 2006; Kim et al., 2010; Yagi et al., 2005; Ishii et al., 2006; Yang et al., 2008; Shin et al., 2014). In particular, in vitro overexpression of ATP6v0d2 and DC-STAMP augmented the formation of multinucleated OCLs (Kim et al., 2008). In addition, ATP6v0d2 and DC-STAMP knockout mice exhibited an increase in trabecular bone volume in parallel with a decrease in osteoclast formation (Lee et al., 2006; Yagi et al., 2005). However, the molecular mechanism of the actions of osteoclast fusion-related proteins has not yet been fully elucidated; for example, partner proteins, which interact with the fusion-related proteins, have not been identified. Cell–cell fusion is simply the fusion of the plasma membrane of a donor cell with the plasma membrane of a neighboring cell; that is, cell–cell fusion occurs through transfer of the plasma membrane phospholipid bilayer of two cells. Thus, cell–cell fusion cannot be interpreted only from the perspective of protein interactions but should also be considered from the viewpoint of the status change of the lipid bilayer of the plasma membrane. The interaction of phospholipids with osteoclast fusion-related proteins could also be important for understanding osteoclast cell–cell fusion.

The state of the plasma membrane lipid bilayer varies according to the species of phospholipids present in the outer and inner leaflets of the plasma membrane, depending on the size and hydrophilic properties of the head group of the phospholipids. When phosphatidylcholine with a relatively large polar head and sphingomyelin gather, the structure of the membrane becomes convex and tends to exist in the outer leaflet of the plasma membrane of spherical cells (Zimmerberg and Kozlov, 2006). On the other hand, PE, which has the smallest head and weak polarity, preferentially forms a concave structure. PE, along with PS, tends to be distributed in the inner leaflet of the plasma membrane (Zachowski, 1993). Thus, the phospholipid bilayers of the outer and inner leaflets exhibit asymmetric and different distributions of phospholipids, whereas the asymmetric structure is easily changed depending on cell function, motility and shape. Recently, Irie et al. (2017) demonstrated the relationship between changes in the PE distribution on the cell membrane and the cell–cell fusion of OCLs. The amount of PE in the cell surface increased with the differentiation of OCLs. In addition, gene knockdown of acyl-CoA:lysophosphatidylethanolamine acyltransferase 2 (catalyzing de novo PE biosynthesis), ABCB4 and ABCG1 reduced the amount of PE in the outer leaflet of the plasma membrane followed by a decrease in OCL formation, indicating the involvement of these enzymes in osteoclastogenesis. The results of this study also indicate that cell–cell fusion of OCLs and PE translocation to the outer leaflet of the cell membrane are associated with ABCG1. However, the expression profile of ABCG1 is controversial. Irie's group showed that the expression of mRNA and the cell surface PE level were increased during OCL differentiation in response to RANKL signaling (Irie et al. 2017). In contrast, our study indicated that ABCG1 expression was not elevated by RANKL addition, but rather, the RANKL stimulus reduced Abcg1 mRNA expression to 50-70% of the level of untreated cells, suggesting that the RANKL signal is intrinsically suppressive for ABCG1 expression. However, the 50-70% reduction in Abcg1 mRNA levels was maintained without further reduction. Additionally, mononuclear OCL precursors and pre-OCLs already possessed a high level of PE in the outer leaflet of the plasma membrane, and the level was reduced at the stage of multinucleated OCLs. On the other hand, the expression of Abcg1 mRNA was markedly decreased in LDLR sKO and LDLR/LOX-1 dKO OCLs compared with WT osteoclasts and, in parallel with the decrease, the PE distribution on the cell surface was also reduced. These results show that ABCG1 expression is decreased by RANKL stimulation, as described above, but that the decrease is ceased and maintained by the uptake of extracellular LDL via LDLR. Alternatively, a certain amount of ABCG1 could be maintained by increasing ABCG1 expression, dependent on extracellular LDL uptake.

Recently, Verma et al. (2018) reported that translocation of PS to the outer leaflet of the plasma membrane of OCLs is involved in cell–cell fusion. They treated mononuclear pre-OCLs with lysothosphatidylcholine to inhibit cell–cell fusion. After the removal of lysothosphatidylcholine, the authors evaluated the change in PS distribution on the cell surface and demonstrated that the interaction of cell surface PS with annexin A2 and annexin A5, which recognize PS on the cell surface, annexin-binding protein S100A4 and fusogenic protein syncyin 1 generates the driving force for cell–cell fusion. However, in this study, we could not detect PS translocation to the cell surface (data not shown), as reported by Irie et al. (2017). It is widely known that apoptotic cells externalize PS on the cell surface, producing an ‘eat me’ signal (Suzuki and Nagata, 2014). Annexin located on the cell surface of macrophages recognizes the cell surface PS; thereby, macrophages phagocytize apoptotic cells (Irie et al., 2017; Roelofs et al., 2010). Furthermore, PE is less hydrated than PS and fusion occurs easily when it comes into contact with the PE layer of an adjacent cell (Lis et al., 1982; Scheule, 1987). In contrast to PE, a highly hydrated PS layer repels the PS layer from each of the other cells. Therefore, PE could play a more important role than PS in efficient cell fusion. However, additional studies are needed to clarify the physiological involvement of both phospholipids in the OCL fusion process.

The translocation of phospholipids, sphingomyelin and cholesterol from the inner to outer leaflet of the plasma membrane is mediated by floppases, which include a variety of ABC transporters such as ABCAs, ABCBs, ABCCs and ABCGs (Coleman et al., 2013). In this study, we examined the expression of ABC transporters during OCL formation. Among the transporters, the expression of ABCB4 and ABCA2 was upregulated with OCL differentiation in response to RANKL signaling, whereas no difference in the expression profiles of the transporters was detected in the four OCL genotypes (WT, LDLR/LOX-1 dKO, LOX-1 sKO and LDLR sKO). Similarly, the expression patterns of ABCA1, ABCA3 and ABCG1 were not altered among the genotypes. Only the expression profile of ABCG1 was consistent with osteoclast formation; that is, although the expression levels in WT and LOX-1 OCLs were relatively maintained, the expression levels in LDLR/LOX-1 dKO and LDLR sKO OCLs were markedly decreased, indicating correlation of ABCG1 with PE translocation and OCL formation. In addition, the expression of ABCG1 depends on LDL uptake. Although ABCG1 also mediates the efflux of cholesterol into high-density lipoprotein and regulates lipid homeostasis (Schmitz et al., 2001), it remains controversial whether the transporter is involved in the transfer of PE to the cell surface. In this study, Abcg1 sKO OCLs exhibited less translocation of PE to the outer leaflet of the plasma membrane and impaired cell–cell fusion. Thus, ABCG1 could mediate the efflux of various phospholipids as well as cholesterol. However, the distribution of PE and cholesterol on the osteoclastic cell surface differed; although cholesterol was ubiquitously distributed on the cell surface, the PE distribution was restricted to the filopodia. The results indicate that PE translocation depends on LDL uptake but that it differs from cholesterol translocation, suggesting cooperation of ABCG1 with another unknown translocation system.

RANKL/RANK signaling facilitates osteoclastogenesis via the activation of downstream signal transduction molecules such as c-Fos, NF-κB and NFATc1 (Asagiri and Takayanagi, 2007; Mizukami et al., 2002; Takayanagi et al., 2002). This study showed that RANKL/RANK signaling intrinsically represses ABCG1 expression. Indeed, it has been reported that the activation of NF-κB downregulates the expression of ABCG1 in RAW cells (Suzuki et al., 2017). Our results using LDLR/LOX-1 dKO and LDLR sKO OCLs indicate that the downregulation becomes more pronounced by deletion of the LDLR gene; that is, the decrease in ABCG1 expression by the activation of NF-κB is reduced by LDL uptake through LDLR. The alleviation by LDL uptake may be attributed to the inhibition of downstream signaling pathways from the RANKL/RANK complex. Another mechanism for the reduction could be upregulation of ABCG1 expression by LDL incorporation into the cells. It has been reported that the expression of Abcg1 is promoted by the nuclear transfer of the complex of LXR with retinoid X receptor or peroxisome proliferator-activated receptor-γ into the nucleus (Sabol et al., 2005; Uehara et al., 2007; Ozasa et al., 2011). The activation of LXR is provoked by intracellular oxysterol derived from cholesterol (Lehmann et al., 1997). As shown in this study, LXR blockage by the antagonist GSK 2033 reduced the expression of Abcg1 mRNA followed by decreased osteoclast formation. On the other hand, LXR activation by the agonist GW 3965 restored the expression of Abcg1 mRNA and the OCL formation. Together, these results suggest the involvement of LXR in osteoclast cell–cell fusion through the floppase activity of ABCG1.

In conclusion, PE translocation to the outer leaflet of the plasma membrane, which is mediated by ABCG1, is strictly dependent on LDL uptake via LDLR. PE translocation on the surface of filopodia is required for cell–cell fusion of OCLs. Our findings provide novel knowledge pertaining to the mechanism of cell–cell fusion of OCLs.

Recombinant human macrophage colony-stimulating factor (M-CSF) was kindly provided by Dr M. Yamada of the Morinaga Milk Industry Co. (Tokyo, Japan) and Prof. S. Suzu (Joint Research Center for Human Retrovirus Infection, Kumamoto University, Kumamoto, Japan). Recombinant murine soluble receptor activator of NF-κB ligand (sRANKL) was obtained from R&D Systems (Minneapolis, MN). LXR antagonist GSK 2033 and LXR agonist GW 3965 were purchased from Tocris Bioscience (Bristol, UK).

LOX-1^−/− and Abcg1^−/− mice have previously been described (Mehta et al., 2007; Draper et al., 2012). LDLR-deleted (LDLR^−/−) mice were obtained from The Jackson Laboratory (Bar Harbor, ME). The mice were maintained under specific pathogen-free conditions and were backcrossed into a C57BL/6J (CLEA Japan, Tokyo, Japan) background for a minimum of six generations. The experimental animal procedures were reviewed and approved by the Animal Care Committee of Meikai University School of Dentistry and conformed to relevant guidelines and laws. For the generation of LDLR^−/−LOX-1^−/− dKO mice, male LOX-1^−/− mice were first mated with female LDLR^−/− mice to create LDLR^−/+LOX-1^−/+ mice. Then, the LDLR^−/+LOX-1^−/+ mice were mated with LDLR^−/− mice to obtain LDLR^−/−LOX-1^−/+ mice. Finally, male LDLR^−/−LOX-1^−/+ mice were mated with female LDLR^−/−LOX-1^−/+ mice to generate LDLR^−/−LOX-1^−/− mice. To determine the genotypes of WT and mutated LOX-1 and LDLR genes, we used the primers described in Table S1. The resulting LDLR^−/−LOX-1^−/− dKO mice were bred with each other and maintained under specific pathogen-free conditions on 12 h dark/12 h light cycles. The primer sequences used for the genotyping of Abcg1^−/− and WT mice are indicated in Table S1.

In vitro OCL formation has been previously described (Okayasu et al., 2012). Briefly, the femora and tibiae were obtained from 4- to 8-week-old male mice and the soft connective tissues removed from the bones. The bone marrow cells were flushed from the bone marrow cavity and cultured for 3 days in α-modified Eagle’s minimum essential medium (α-MEM; Thermo Fisher Scientific, Rockford, IL) supplemented with 10% fetal bovine serum (FBS; Sigma-Aldrich, St Louis, MO), M-CSF (100 ng/ml) and 100 U/ml penicillin in Petri dishes in a humidified atmosphere of 5% CO₂. We used Petri dishes instead of tissue culture dishes to avoid the adhesion of bone marrow stromal cells on the dishes. Following the removal of the nonadherent cells and the small population of stromal cells by washing the dishes with PBS and subsequent incubation for 3 min in PBS containing 0.25% trypsin and 0.05% EDTA, the adherent monocytes were harvested for use as OCL precursors by vigorous pipetting in α-MEM containing 10% FBS. In the population of isolated OCL precursors, contaminating stromal cells represented less than 0.01% of the total cells, as described previously (Kaneda et al., 2000). The harvested OCL precursors were seeded in various tissue culture dishes and plates at an initial density of 2.5×10⁴/cm² and cultured in α-MEM containing 10% FBS, M-CSF (20 ng/ml) and sRANKL (10 ng/ml). The culture medium was exchanged with fresh medium every 2 days. After culturing for the desired time, the cells were fixed in 10% formalin and stained for TRAP activity with a leukocyte acid phosphatase kit (Sigma-Aldrich). TRAP-positive multinucleated cells (MNCs) with more than three nuclei, which were considered OCLs, were counted under a microscope. The area of the multinucleated TRAP-positive MNCs was measured using the image analysis software Sensiv MEASURE (Mitani, Tokyo, Japan).

The uptake of LDL by osteoclast precursors was measured using the pHrod Red LDL uptake kit (Thermo Fisher Scientific) according to the manufacturer's instructions. Briefly, OCL precursors obtained from M-CSF-pretreated bone marrow cells were seeded at 1.6×10⁵ cells/well and cultured with M-CSF (20 ng/ml) and sRANKL (10 ng/ml) in α-MEM containing 10% FBS for 6 h in each well of a polylysine-coated multiwell glass bottom culture dish (Matsunami, Osaka, Japan). After washing the cells once with PBS, the culture medium was replaced with LDL-starvation medium composed of FluoroBrite DMA (Thermo Fisher Scientific) plus 10% LDL-deficient serum (Biomedical Technologies, Stoughton, MA). The cells were cultured overnight with M-CSF and sRANKL to starve the cells of LDL. Thereafter, pHrode Red LDL (final concentration 10 µg/ml) was added to the culture and the cells incubated for 3 h at 37°C. For the negative control, cells were pretreated with unlabeled LDL (25 µg/ml) for 45 min prior to pHrode Red LDL labeling. For the last 30 min of labeling, the nuclei of the cells were stained with NucBlue live ReadyProbe reagent. Fluorescence images were obtained using a confocal laser microscope (LSM800, Carl Zeiss, Stuttgart, Germany).

OCL precursors were cultured in the presence of M-CSF and sRANKL in α-MEM containing 10% FBS for one day. Then, the cells were fixed with 3.7% formaldehyde in PBS for 15 min at room temperature, washed twice with PBS and then incubated in 60% 2-propanol for 1 min. Then, the cells were stained with Oil Red O solution (1.8 mg/ml of 60% 2-propanol) for 15 min. After staining, the cells were washed once with 60% 2-propanol and twice with PBS.

To stain the cell surface cholesterol of OCLs, we employed θ-toxin-derived domain 4 (D4), which specifically binds cholesterol (Shimada et al., 2002). The pET28/His6-mCherry-D4 plasmid was obtained from Dr T. Kobayashi, RIKEN (Tsukuba, Japan). Escherichia coli strain BL21 (DE) was used for the overexpression of His-tag-mCherry D4 proteins. After induction by isopropyl β-D-1-thiogalactopyranoside, E. coli cells were harvested and lysed in HisTALON xTractor Buffer (Clontech Laboratories, Mountain View, CA) containing 10 units/ml DNase I and lysozyme solution. The overexpressed His-tag-mCherry-D4 proteins in the cell lysate were partially purified by TALON metal affinity resins in accordance with the manufacturer's instructions. Furthermore, the fractions eluted from the TALON metal affinity resins were applied to a butyl-agarose (Bio-Rad, Hercules, CA) column equilibrated with 20 mM Tris-HCl (pH 7.5) containing 0.8 M (NH₄)₂SO₄. The cleaved His-tag-mCherry fragments did not bind to the hydrophobic resins. The bound His-tag-mCherry-D4 was eluted with 0.2 M (NH₄)₂SO₄, dialyzed against 20 mM Tris-HCl (pH 7.5) and concentrated using Amicon Ultra centrifugal filters (MWCO 10 kDa; Merck, Darmstadt, Germany). Finally, an equal volume of glycerol was added to the solution of His-tag-mCherry-D4. The protein concentration of the solution was measured and the solution stored at −20°C until use. The purity of His-tag-mCherry-D4 was confirmed to be more than 80% by SDS-PAGE analysis.

OCL precursors were cultured with M-CSF and sRANKL on polylysine-coated glass bottom culture dishes (Matsunami) for the indicated periods. Thereafter, the cells were treated with or without MβCD (10 mM) in BSA/HEPES/α-MEM [α-MEM containing 0.1% fatty acid-free BSA (Sigma-Aldrich) and 20 mM HEPES-NaOH, pH 7.4] for 30 min at 37°C. After washing once with BSA/HEPES/α-MEM, the cells were incubated in BSA/HEPES/α-MEM containing His-tag-mCherry-D4 (10 µg/ml) for 30 min. Then, the cells were quickly washed once with PBS and fixed with 4% paraformaldehyde in PBS for 15 min at room temperature. After washing with PBS, the nuclei of the cells were counterstained with 4′,6-diamidino-2-phenylindole (DAPI; Thermo Fisher Scientific). Fluorescence images were obtained using a confocal laser microscope (LSM800, Carl Zeiss).

To stain PE on the outer leaflet of the plasma membrane, we used Ro09-0198, which specifically recognizes and binds PE (Choung et al., 1988). Streptavidin (SA; Nacalai Tescue, Kyoto, Japan) was conjugated with biotinylated Ro09-0198 (Bio-Ro; provided by Dr M. Umeda, Kyoto University, Japan) and was dialyzed against PBS using Slide-A Lyzer (20,000 molecular weight cut off; Thermo Fisher Scientific) to separate free Bio-Ro in the solution.

OCL precursors were cultured with M-CSF and sRANKL on polylysine-coated glass bottom culture dishes (Matsunami) for the indicated periods. After washing once with BSA/HEPES/α-MEM prewarmed at 37°C, the cells were incubated with SA–Bio-Ro (5 µg/ml) in BSA/HEPES/α-MEM for 30 min at room temperature. Then, after washing once with BSA/HEPES/α-MEM, the cells were fixed with 3% paraformaldehyde in PBS for 30 min at room temperature and washed three times with PBS followed by 5 min of permeabilization with 0.1% TritonX-100 in PBS. Then, the cells were blocked with 3% BSA (Fraction V, Sigma-Aldrich) in PBS for 1 h prior to 1 h incubation with FITC-conjugated anti-SA-antibody (400-fold dilution; Vector Laboratories, Burlingame, CA) at room temperature. After incubation, the nuclei of cells were counterstained with a DAPI solution. We captured fluorescence images using a confocal laser microscope (LSM800, Carl Zeiss). The relative immunofluorescence intensity of SA–Bio-Ro on the cell surface was calculated as per the intensity of DAPI in the cells using ZEN2.1 (blue edition) software (Carl Zeiss).

The total RNA (100 ng) from cultured cells was used as the template for cDNA synthesis using a high-capacity RNA to cDNA kit (Thermo Fisher Scientific). Quantitative real-time PCR was performed using SsoAdvancedTM Universal Probes Supermix PCR (Bio-Rad) on a CFX Connect Real-Time PCR Detection System (Bio-Rad). TaqMan probes for the indicated genes were obtained from Thermo Fisher Scientific and are listed in Table S2. The relative expression of target mRNAs was quantified and normalized to the amount of 18S rRNA.

The data are presented as the mean±standard deviation (s.d.) or standard error of the mean (s.e.m.). The mean values of the groups were compared by unpaired Student's t-test or by one-way and two-way ANOVAs; the significance of the observed differences was subsequently determined by post hoc testing using Tukey's method and Bonferroni's method, respectively. P<0.05 was considered significant.

We thank Dr M. Yamada (Morinaga Milk Industry Co.) and Prof. S. Suzu (Joint Research Center for Human Retrovirus Infection, Kumamoto University) for providing the recombinant human M-CSF, and we thank Prof. M. Umeda (Kyoto University) for providing the biotin-conjugated Ro09-0198. We also appreciate Dr T. Kobayashi (RIKEN, and Directeur de recherche INSERM) for his help obtaining the pET28/His6-mCherry-D4 plasmid.

The peer review history is available online at https://jcs.biologists.org/lookup/doi/10.1242/jcs.243840.reviewer-comments.pdf