During bone resorption, osteoclasts firmly adhere to bone tissue which is mediated by specialized actin-based membrane extensions, called podosomes (Touaitahuata et al., 2013). Each podosome is composed of a central core of a dense F-actin network and actin polymerization activators (Jurdic et al., 2006). The core is surrounded by a loose F-actin network interspersed with regulatory proteins and adaptor proteins linking the podosome structure with integrins, referred as the podosome cloud (Jurdic et al., 2006; Georgess et al., 2014). At the onset of bone resorption, podosomes are organized into an extensive circular pattern: the sealing zone (Georgess et al., 2014), which is of great importance and essential for bone resorption (Jurdic et al., 2006). It defines an isolated extracellular compartment, the resorption lacunae, where bone tissue is degraded (Helming and Gordon, 2009). Therefore, the resorption lacunae could be seen as an extracellular lysosome, able to destroy bone matrix (Vignery, 2005a; Helming and Gordon, 2009). Since bone is too large to internalize, the formation of this extracellular lysosome is a crucial step in bone resorption (Cappariello et al., 2014).

Active bone-resorbing osteoclasts are characterized by a ruffled border facing the bone surface (Novack and Teitelbaum, 2008; Brooks et al., 2019). By contrast, LGCs and FBGCs do not have a ruffled border and are not considered to resorb bone, illustrating the importance of the ruffled border in bone resorption. This border originates from fusion of lysosomes or storage of granule-like structures with the plasma membrane, enabling the insertion of lysosomal membrane proteins into the plasma membrane and the release of lysosomal enzymes into the resorption lacunae (Stenbeck, 2002; Na et al., 2020).

As mentioned before, osteoclasts are able to internalize bone remnants for intracellular degradation (Stenbeck, 2002; Mulari et al., 2003; Madel et al., 2019). Lysosomal enzymes are secreted at the periphery of the ruffled border, whereas uptake of degradation products takes place at the center (Mulari et al., 2003). The internalization is coordinated by the regular clathrin-mediated endocytosis machinery (Mulari et al., 2003) and results in the formation of intracellular vesicles filled with bone remnants: transcytotic vesicles (Madel et al., 2019). By transcytotic trafficking, vesicles are transported to the functional secretory zone (i.e., the membrane in opposite of the ruffled border) where the content is released into the extracellular environment.

Osteoclasts also stimulate bone formation directly through the release of various soluble osteogenic factors, called clastokines (Lotinun et al., 2013). Amongst others, sphingosine-1 phosphate (SP-1 P), BMP6, TRAP, platelet-derived growth factor (PDGF), and hepatocyte growth factor (HGF) are well-described clastokines (DiGiovanni et al., 2012; Teti, 2013; Cappariello et al., 2014). SP-1 P is a signaling sphingolipid that enhances bone regeneration through multiple mechanisms (Higashi et al., 2016; Meshcheryakova et al., 2017). SP-1 P recruits osteoblast precursors (Meshcheryakova et al., 2017), promotes osteoblast proliferation, differentiation, activity, and survival (Higashi et al., 2016; Meshcheryakova et al., 2017). BMP6 induces osteoblast differentiation (Demirtaş et al., 2016) and bone mineralization (Luo et al., 2011; Zhu et al., 2012). In addition to its role in bone resorption, TRAP acts as a clastokine by facilitating osteoblast proliferation and differentiation (Gradin et al., 2012). PDGF acts as a chemotactic and mitogenic factor for osteoblasts (DiGiovanni et al., 2012), whereas HGF facilitates osteoblast proliferation and differentiation (Hossain et al., 2005; Frisch et al., 2016). Together, these factors facilitate the migration of osteoblast precursors, osteoblast proliferation, differentiation, and/or activity (Gradin et al., 2012; Cappariello et al., 2014).

Osteoclasts also promote bone formation through direct cell contact with osteoblasts, mediated by members of the Eph receptor family and their ephrin ligands (Zhao et al., 2006; Edwards and Mundy, 2008; Matsuo and Otaki, 2012; Tamma and Zallone, 2012). Eph receptors are tyrosine kinase receptors that are involved in multiple biological processes, including neuronal development (Kania and Klein, 2016). Interaction between Eph receptors and ephrin ligands results into bi-directional signaling, meaning that signaling pathways are activated in both receptor-expressing and ligand-expressing cells (Edwards and Mundy, 2008; Kania and Klein, 2016). Signal transduction from ephrin ligands to Eph receptors is referred as forward signaling, whereas reverse signaling points to signal transduction from Eph receptors to ephrin ligands (Kania and Klein, 2016). Several reports indicate that ephrinB2 ligand and EphB4 receptor are involved in osteoclast-osteoblast intercommunication (Zhao et al., 2006; Edwards and Mundy, 2008; Matsuo and Otaki, 2012; Tamma and Zallone, 2012). Differentiating and mature osteoclasts express ephrinB2, whereas osteoblasts express EphB4 (Zhao et al., 2006; Edwards and Mundy, 2008; Matsuo and Otaki, 2012). Forward signaling stimulates osteoblast differentiation (Zhao et al., 2006; Edwards and Mundy, 2008; Matsuo and Otaki, 2012), facilitating the formation of new bone tissue, whereas reverse signaling inhibits osteoclast maturation (Zhao et al., 2006). The combined action of both signaling cascades ensures efficient bone replacement by mediating simultaneously cessation of bone resorption and generation of new bone tissue (Edwards and Mundy, 2008).

Phagocytosis is the internalization of particles larger than 0.5 µm (Hirayama et al., 2001) in size by plasma membrane-derived vesicles (Rosales and Uribe-Querol, 2017). Professional phagocytes, including macrophages and DCs, internalize foreign particles for degradation purposes or in advance of antigen preparation and presentation (Madel et al., 2019). Osteoclasts engulf bone remnants for the purpose of bone degradation through clathrin-mediated endocytosis (Mulari et al., 2003). Other than bone remnants, osteoclasts have been reported to internalize many other particles through phagocytosis, including latex, polymethylmethacrylate, and titanium (Wang et al., 1997; Meng et al., 2010; Madel et al., 2019). Whether phagocytosis of particles other than bone remnants influences bone resorption remains unclear as conflicting results have been reported (Wang et al., 1997; Meng et al., 2010). Wang et al. showed that osteoclasts still exert bone resorption after they have internalized latex or polymethylmethacrylate particles, whereas Meng et al. suggested that bone resorption decreases upon phagocytosis of titanium particles (Meng et al., 2010). Moreover, several reports indicate that osteoclasts are able to phagocytose apoptotic cells (Taniwaki and Katchburian, 1998; Bronckers et al., 2000; Boabaid et al., 2001; Cerri et al., 2003; Harre et al., 2012). Apoptosis is a sterile form of programmed cell death, such that it does not elicit an inflammatory response (Xu et al., 2019). During apoptosis, various molecules are exposed to the cell surface that label the cell as being apoptotic (Savill, 1997). These molecules are recognized by phagocytic receptors on phagocytes followed by the internalization of the apoptotic cell (Savill, 1997). Macrophages and DCs are well-equipped to clear apoptotic cells through phagocytosis (Harre et al., 2012). Moreover, Harre et al. reported that osteoclasts express multiple proteins that are involved in the engulfment of apoptotic cells, to a similar or even higher extend than macrophages and DCs (Harre et al., 2012). This is reflected by the high phagocytic capacity of osteoclasts for murine apoptotic thymocytes in vitro. In contrast to macrophages, osteoclasts internalized the cells independently from serum, indicating that they do not use opsonization or that they make the opsins themselves. In vivo evidence has also been provided that osteoclasts perform phagocytosis under physiological conditions (Taniwaki and Katchburian, 1998; Bronckers et al., 2000; Boabaid et al., 2001; Cerri et al., 2003). Osteoclasts are able to engulf apoptotic bone cells, including chondrocytes, osteocytes, and osteoblasts (Taniwaki and Katchburian, 1998; Bronckers et al., 2000; Boabaid et al., 2001; Cerri et al., 2003). Osteocytes and chondrocytes are tightly surrounded by bone and cartilage matrices respectively, making them difficult to reach by classical phagocytes, such as macrophages (Harre et al., 2012). Therefore, osteoclasts are crucial to remove dying bone cells to prevent chronic inflammation or autoimmunity. Not only the accessibility renders osteoclasts more suitable for clearance of dying bone cells, also the fact that they do not rely on serum-derived opsins is important since bone tissue is poorly vascularized.

Growing evidence indicates that osteoclasts interact with T cells. In bone marrow, T cells are in close proximity to osteoclasts (Grassi et al., 2011), suggesting that osteoclasts attract T cells. This hypothesis is supported by the finding that osteoclasts secrete T cell-attracting chemokines, including CC chemokine ligand (CCL)2, CCL3, and CXCL10 (Kiesel et al., 2009; Grassi et al., 2011). Adhesion assays showed that osteoclasts are able to recruit and retain T cells (Grassi et al., 2011). Next to mediating chemotaxis, osteoclasts can also regulate T cell activity and differentiation. In order to become activated, naïve T cells need to be primed with antigens presented on the surface of antigen presenting cells (APCs) (Sprent, 2005). Several reports demonstrated that osteoclasts can act as APCs. Indeed, osteoclasts cross present exogenous antigens in MHC-I molecules to CD8⁺ T cells (Buchwald et al., 2013; Le Goff et al., 2013). Characterization of these CD8⁺ T cells revealed that osteoclasts induce a regulatory T cell (Treg) phenotype, thereby inducing an immune modulatory effect (Buchwald et al., 2013; Le Goff et al., 2013; Shashkova et al., 2016). In addition, osteoclasts have been reported to present antigens in MHC-II molecules to CD4⁺ T cells, which in turn differentiate into Th1 cells (Le Goff et al., 2013; Ibáñez et al., 2016) (Figure 2).

It should be noted that not all studies that have examined the APC function of osteoclasts reached the same conclusion. Kiesel et al. reported that osteoclasts cannot present antigens to CD4⁺ T cells since they do not express MHC-II molecules (Kiesel et al., 2009). By contrast, other reports demonstrated that osteoclasts do express MHC-II molecules (Li et al., 2010, 2014; Le Goff et al., 2013; Ibáñez et al., 2016) and that they are able to present antigens to CD4⁺ T cells (Li et al., 2010, 2014; Le Goff et al., 2013). Furthermore, conflicting results about the expression of co-stimulatory molecules have been reported (Kiesel et al., 2009; Li et al., 2010, 2014; Ibáñez et al., 2016). Kiesel et al. reported that osteoclasts only express CD80 but not CD86 (Kiesel et al., 2009), whereas other groups found that osteoclasts express both molecules (Li et al., 2010, 2014; Ibáñez et al., 2016). These discrepancies could be explained by the immune environment in which osteoclast-T cell interactions take place (Le Goff et al., 2013; Ibáñez et al., 2016). Under physiological conditions, osteoclasts exhibit an immune-modulatory phenotype, characterized by the secretion of anti-inflammatory cytokines, such as IL-10 (Ibáñez et al., 2016). These osteoclasts induce Tregs by antigen cross presentation to CD8⁺ T cells (Buchwald et al., 2013; Le Goff et al., 2013; Shashkova et al., 2016). Additionally, they secrete soluble factors that inhibit T cell proliferation, inflammatory cytokine production, and activation-induced apoptosis of T cells (Grassi et al., 2011). It is assumed that these immunomodulatory functions of osteoclasts are crucial to maintain immune tolerance to bone remnants in order to prevent autoinflammation (Ibáñez et al., 2016). Conversely, under inflammatory conditions, osteoclasts may exacerbate the ongoing inflammation. Upon stimulation with LPS or IFN-γ, MHC-II expression is upregulated in osteoclasts (Grassi et al., 2011; Le Goff et al., 2013), which leads to increased Th1 cell induction. Furthermore, osteoclasts merely produce pro-inflammatory cytokines (TNF-α, IL-1β, IL-6 and IL-23) and chemokines (CCL2, CCL5 and CCL7) in an inflammatory environment (Ibáñez et al., 2016).

The differences in osteoclast activity might be also explained by the different origin of osteoclasts in various milieus (Ibáñez et al., 2016). Under physiological conditions, osteoclasts are considered to be derived from monocytes, whereas under inflammatory conditions they may originate from immature DCs. Therefore, the inflammatory osteoclast can be seen as an intermediate between the monocyte-derived osteoclast and the DC and probably plays a role in many immune diseases, including RA (Narisawa et al., 2021).

Osteoclasts may also indirectly regulate the activity of immune cells (van Niekerk et al., 2018). Bone tissue forms a reservoir for calcium, phosphate and magnesium ions, which are released during bone resorption. These ions are crucial for many immune related processes, such as immune cell proliferation, chemotaxis, and signaling.

Taken together, osteoclasts exert multiple functions within the human body of which bone resorption is the best recognized (Drissi and Sanjay, 2016). Throughout the last decades, increasing evidence was found that osteoclasts cannot longer be regarded as solely bone-resorbing cells and that they contribute to multiple processes, including osteoblast stimulation, vasculogenesis and immune regulation (Le Goff et al., 2013; Cappariello et al., 2014; Drissi and Sanjay, 2016). Compared to osteoclasts, the functionality of FBGCs and LGCs is less well established, and will be discussed in the next two sections.

One possible mechanism by which FBGCs may destroy their substrate is phagocytosis and intracellular degradation (Lee et al., 1981; Milde et al., 2015). Several researchers demonstrated that FBGCs are able to engulf particles that single macrophages cannot (Moreno et al., 2007; Milde et al., 2015). Milde et al. showed that FBGCs are more efficient in phagocytosing beads of 20 µm compared to macrophages and that beads of 45 µm are exclusively taken up by FBGCs (Milde et al., 2015). In addition, Moreno et al. showed that fusion of macrophages into FBGCs is crucial to be able to phagocytose 25 µm beads while IL-4 stimulated macrophages failed to phagocytose these beads. (Moreno et al., 2007). Inhalation experiments in rats, hamsters and guinea pigs provided in vivo evidence that alveolar macrophages preferably internalize small dust fibers (i.e. particles smaller than 5 µm in size), whereas particles larger than 10 µm are phagocytosed by FBGCs (Lee et al., 1981).

Protein adsorption on biomaterial surfaces plays probably an important role in the phagocytic properties of FBGCs. Milde et al. found that FBGCs are highly phagocytic for complement-opsonized particles (Milde et al., 2015). Indeed, the excessive phagocytic capacity was diminished when medium was depleted of complement component 3 (C3) or when FBGCs did not express the complement receptor 3 subunit CD11b. Complement-mediated phagocytosis is very efficient in FBGCs since the complement receptor 3 (CR3) localizes into membrane rumples, which provide excess cellular membrane to engulf large particles. Complement opsonization may be of great importance in phagocytosis of foreign bodies, such as polymer or metal particles, that do not have any ligands for phagocytic receptors. Since the nature of the biomaterial highly influences protein adsorption (Anderson et al., 2008), it can be speculated that it also interferes with the phagocytic capacity of FBGCs. Further research is required to investigate whether other phagocytic receptors are involved in the internalization of foreign bodies and whether the contribution of the receptors is altered by the nature of the biomaterial.

Next to phagocytosis, FBGCs destroy foreign bodies through the secretion of ROS and matrix metalloproteases (MMPs) (Jones et al., 2007). Release of these deleterious substances was found to lead to the failure of medical implants (Anderson et al., 2008), stressing the need for the development of resistant biomaterials (Ebert et al., 2005). Microscopic analyses revealed that FBGCs firmly adhere to their substrate thereby creating an isolated extracellular environment between the FBGC plasma membrane and the particle surface, called the sealing zone (ten Harkel et al., 2015; ten Harkel et al., 2016). The sealing zone can be regarded as an extracellular lysosome in which harmful substances are released in order to degrade the biomaterial (Vignery, 2005a). Evidence has been provided that FBGCs degrade collagen, which is often used as biomaterial for medical devices (van Wachem et al., 1991; Ye et al., 2011; ten Harkel et al., 2016). At sites where FBGCs are in close proximity to collagen bundles, the collagen structure loosens and fibrils look thinner, pointing to extracellular degradation (ten Harkel et al., 2016). FBGCs are also able to break down gelatin (i.e. denatured collagen) (Ye et al., 2011). The degradation of collagen and gelatin both rely on MMP13, an active collagenase and gelatinase. Furthermore, collagen and gelatin can be phagocytosed, suggesting that they undergo additional intracellular degradation.

Whether FBGCs preferably degrade their substrate internally (i.e. through phagocytosis) or externally (i.e. through secretion of ROS and MMPs) depends on the size of the particle (Sheikh et al., 2015). Existing data suggest that FBGCs efficiently phagocytose particles up to 100 µm in diameter. Though the phagocytic efficiency decreases with increased particle size and if needed, FBGCs could switch to external degradation.

LGCs are most studied within the context of tuberculosis (Figure 4). Although LGCs are inefficient in destroying M. tuberculosis, either through phagocytosis or through production of microbicidal oxidants (Lay et al., 2007; Yasui et al., 2011), they still may exert protective functions. Within the granulomatous structure, LGCs form a barrier that shields M. tuberculosis from the rest of the body, and prevent cell-to-cell spread and growth of bacteria (Byrd, 1998; Brooks et al., 2019). Furthermore, LGCs express high amounts of MHC-II molecules, enabling them to function as antigen presenting cells, capable to prime the adaptive immune response (Lay et al., 2007). Conversely, tuberculosis-associated LGCs produce MMPs thereby inducing tissue destruction (Helming et al., 2009). LGCs may be involved in the inflammatory reaction of the granuloma through the production of cytokines and chemokines. In a murine model for pulmonary tuberculosis, immunostaining for TNF-α and IL-1α, two cytokines that have been reported to be crucial for granuloma formation (Huaux et al., 2015; Pereira et al., 2018), was strongly positive in LGCs (Hernandez-Pando et al., 1997). In addition to initiation and maintenance of granulomas, LGCs may also be involved in resolution of inflammation and development of fibrosis. As for FBGCs, LGCs undergo a phenotypic switch with time, characterized by reduced expression of TNF-α and IL-1β, and simultaneous upregulation of TGF-β, a major regulator of tissue repair and fibrosis (Hernandez-Pando et al., 1997; Wynn and Ramalingam, 2012).

Although LGCs are a major morphological characteristic of infectious granulomas, their specific biological activity and function still needs further elucidation.

DNAX-activating protein of 12 kDa (DAP12) is a transmembrane adaptor protein (Paloneva et al., 2003) predominantly expressed on NK cells (Takaki et al., 2006; Tessarz and Cerwenka, 2008) and on cells of the myeloid lineage (Takaki et al., 2006), including monocytes (Tessarz and Cerwenka, 2008), macrophages, neutrophils (Tessarz and Cerwenka, 2008), and DCs (Humphrey et al., 2004). DAP12 is expressed as a homodimer at the cell surface, where it associates with DAP12-associated receptors to exert its function (Humphrey et al., 2004). In DAP12-deficient mice, bone mass is increased (Humphrey et al., 2004), suggesting a role for DAP12 in osteoclast formation and/or function. Osteoclast fusion is inefficient in monocytes derived from patients with loss-of-function mutations in DAP12 (Paloneva et al., 2003) and in bone marrow macrophages derived from DAP12-deficient mice (Humphrey et al., 2004). By contrast, stimulation of DAP12 in RAW264.7 cells (a monocyte/macrophage cell line derived from mice) enhances fusion, resulting in more multinucleated osteoclasts with increased cell size and number of nuclei (Humphrey et al., 2004). A role for DAP12 in fusion is not exclusive to osteoclasts as Helming et al. demonstrated that IL-4 induced fusion is also impaired in human macrophages silenced for DAP12 (Helming et al., 2008). Triggering receptor expressed on myeloid cells 2 (TREM2) functions as a DAP12-associated receptor in osteoclasts (Takaki et al., 2006; Peng et al., 2010; Zou and Teitelbaum, 2015) and has also been reported to mediate osteoclast multinucleation (Paloneva et al., 2003). It is hypothesized that macrophages themselves express a TREM2 ligand and thus that TREM2/DAP12-mediated signaling relies on the interaction between macrophages (Peng et al., 2010). During osteoclastogenesis, TREM2 is upregulated and its expression positively correlates with the number of nuclei in mature osteoclasts (Humphrey et al., 2006). Stimulation of TREM2 enhances multinucleation (Humphrey et al., 2004), whereas silencing or blockade of TREM2 inhibits osteoclast multinucleation (Humphrey et al., 2006; Peng et al., 2010). Finally, human PBMCs derived from TREM2-deficient patients, show impaired fusion when differentiated into osteoclasts (Paloneva et al., 2003). Together, these results demonstrate that combined action of DAP12 and TREM2 is important to macrophage fusion. After recognition of its unknown ligand, TREM2 associates with DAP12 leading to the phosphorylation of the immunoreceptor tyrosine-based activation motif (ITAM) in the cytoplasmic domain of DAP12 (Paloneva et al., 2003; Pereira et al., 2018). Once phosphorylated, DAP12 interacts with the cytoplasmic protein tyrosine kinases SYK and ZAP70, thereby triggering several downstream pathways (Paloneva et al., 2003). Although both FBGCs and osteoclasts rely on DAP12/TREM2-signaling, distinct downstream pathways are involved to acquire a fusion competent state (Pereira et al., 2018). In FBGC precursors, cooperation of IL-4 and ITAM signaling results in the activation of transcription factor STAT6 which in turn induces the expression of essential fusogens, including E-cadherin, DC-STAMP, and MMP9 (Moreno et al., 2007; Helming and Gordon, 2009; Van den Bossche et al., 2012; Pereira et al., 2018). In osteoclast precursors, recruitment of SYK to DAP12 leads to Ca²⁺ signaling downstream of PLCγ (Koga et al., 2004). Ca²⁺ signaling enhances the induction of nuclear factor of activated T cells c1 (NFATc1), the master transcription factor of osteoclastogenesis, through an autoamplification mechanism (Koga et al., 2004; Kameda et al., 2013).

It is worth mentioning that DAP12-depleted mice still form multinucleated osteoclasts in vivo, whereas multinucleation is severely inhibited in vitro (Humphrey et al., 2004). Therefore, other mechanisms compensate for fusion in case of DAP12 deficiency in vivo. Zou et al. demonstrated that FcRγ, another ITAM-containing signaling adaptor that activates SYK signaling, can rescue osteoclast multinucleation in absence of DAP12 (Zou and Teitelbaum, 2015).

P2RX7 is an ATP-gated ion channel that belongs to the family of P2X purinergic receptors (Di Virgilio et al., 1999; Gartland et al., 2003; Lemaire et al., 2006). Although P2RX7 functions as a cation-selective channel under transient stimulation with ATP, the receptor transforms into a nonselective pore for hydrophilic molecules of a molecular mass up to 900 Da upon repetitive ATP stimulation (Di Virgilio et al., 1999; Falzoni et al., 2000). During macrophage fusion, P2RX7 localizes at sites of cell-cell contact in podosomes (Falzoni et al., 2000; Lemaire et al., 2012), indicating that the receptor plays a role in macrophage fusion. The involvement of P2RX7 in fusion is further supported by the finding that mouse J774 macrophages expressing high levels of P2RX7 spontaneously form MGCs (Chiozzi et al., 1997; Di Virgilio et al., 1999; Falzoni et al., 2000). Previously, it was assumed that P2RX7 drives fusion through the formation of a “fusion pore”, thereby connecting the cytoplasm of two neighboring cells (Falzoni et al., 2000; Lemaire et al., 2012). This hypothesis was supported by the finding that P2RX7 pore-forming activity is required for macrophage fusion (Lemaire et al., 2012). Blockade of the pore activity with P2RX7 neutralizing antibodies or with the receptor antagonist oxidized ATP, impairs fusion of human and murine macrophages (Chiozzi et al., 1997; Di Virgilio et al., 1999; Falzoni et al., 2000; Gartland et al., 2003; Steinberg and Hiken, 2007). Additionally, polymyxin B, a natural cationic peptide that potentiates pore activity in response to ATP, facilitates fusion of HEK293 cells transfected with P2RX7 (Lemaire et al., 2006). By contrast, HEK293 cells transfected with a truncated form of P2RX7 that lacks ATP-induced pore-forming activity, show impaired fusion even in presence of polymyxin B. Interestingly, hydrolyzation of extracellular ATP, by hexokinase or apyrase, accelerates MGC formation, which conflicts with the idea of ATP-induced pore formation during fusion (Di Virgilio et al., 1999; Pellegatti et al., 2011). These observations may be explained by receptor desensitization when stimulated with high ATP concentrations as degradation of ATP by hexokinase or apyrase might restore the responsiveness of P2RX7 (Falzoni et al., 2000). More recently, Pellegatti et al. proposed a new concept in which P2RX7 solely functions as a receptor for local ATP release to provide extracellular ATP for biosynthesis of adenosine, the actual fusogen (Pellegatti et al., 2011). Indeed, small amounts of extracellular ATP restore fusion in presence of anti-P2RX7 antibodies. Moreover, extracellular degradation of adenosine abolishes fusion, whereas addition of adenosine has the opposite effect. During macrophage fusion, adenosine likely acts at the adenosine receptor A2A as pharmacological inhibition of the A2A receptor reduces fusion, whereas A2A receptor agonists increase fusion. The finding that combined action of a P2RX7 inhibitor and A2A receptor agonists allows fusion, indicates that proper adenosine supply is required for MGC formation.

E-cadherin belongs to the family of the cadherins, which are transmembrane or membrane-associated glycoproteins that mediate calcium-dependent cell-cell adhesions (Mbalaviele et al., 1995; van Roy and Berx, 2008). During FBGC formation, E-cadherin is upregulated at the macrophage cell surface downstream of IL-4 or IL-13 signaling (Moreno et al., 2007; Van den Bossche et al., 2009, 2012; Wanat et al., 2014; Fiorino and Harrison, 2016), suggesting for a role in macrophage fusion. Moreover, E-cadherin neutralizing antibodies impair IL-4 induced macrophage fusion, especially in early stages of differentiation (Moreno et al., 2007), and fusion is diminished in macrophages derived from E-cadherin deficient mice (Van de Bossche et al., 2009; Van de Bossche et al., 2012). Fusion-efficient macrophages are characterized by membrane extension, called podosomes, which form zipper-like structures at cell fusion sites (Balabiyev et al., 2020). E-cadherin is probably implicated in the generation of these zipper-like structures as neutralizing antibodies disrupt formation of these structures and subsequent fusion (Balabiyev et al., 2020). E-cadherin is also upregulated during osteoclastogenesis, especially at early differentiation stages, suggesting that E-cadherin is involved in early osteoclastogenesis (Fiorino and Harrison, 2016). Blocking of E-cadherin, either by neutralizing antibodies (Mbalaviele et al., 1995; Fiorino and Harrison, 2016; Sun et al., 2020) or synthetic peptides containing the cell adhesion recognition sequence of cadherins (Mbalaviele et al., 1995), diminishes osteoclast fusion (Mbalaviele et al., 1995; Fiorino and Harrison, 2016), whereas E-cadherin overexpression transiently accelerates fusion (Fiorino and Harrison, 2016). Although E-cadherin is upregulated at membrane extensions and cell-cell contact sites, the protein is not detected at fusion sites, indicating that it does not make part of the fusion machinery (Fiorino and Harrison, 2016). Instead, E-cadherin neutralizing antibodies impact migration and expression of fusogens, including DC-STAMP, suggesting that cell-cell contact mediated through E-cadherin alters macrophage proliferation and induce the transition to cell migration and fusion.

Integrins are a superfamily of heterodimeric transmembrane proteins comprised of α and β subunits, which mediate both cell-extracellular matrix and cell-cell adhesions (Campbell and Humphries, 2011; Aghbali et al., 2017). The integrin subunits β1, β2, and β5 have been reported to contribute to FBGC formation. Anti-β1 and anti-β2 antibodies impair the adhesion of human monocyte/macrophage to biomaterials, an essential prerequisite for FBGC formation (McNally and Anderson, 2002). Although blockage of β1 dramatically reduces the number of adherent monocyte/macrophages, some FBGCs are still formed, indicating that fusion competency is remained. By contrast, β2 neutralizing antibodies also abolishes fusion, suggesting that β2 mediates both intercellular and cell-biomaterial interactions during FBGC formation. αMβ2 is involved in IL-4 induced fusion of murine peritoneal macrophages both in vivo and in vitro. Indeed, blocking of αMβ2 results in less FBGCs formation and the remaining FBGCs are smaller in size and contain less nuclei (Podolnikova et al., 2016). αDβ2 might also contribute to IL-4 induced fusion of murine macrophages, however to a lower extend as αMβ2. Finally, IL-4 has been reported to increase the expression of β5 on human monocytes, suggesting a role for this integrin in FBGC fusion (Aghbali et al., 2017). Integrins are also involved in the formation of multinucleated osteoclasts (Rao et al., 2006). It has been found that α9β1 contributes to osteoclast multinucleation, reflected by reduced osteoclast formation in human bone marrow cultures treated with anti-α9 antibodies and smaller osteoclast size in α9-deficient mice.

Macrophage fusion receptor (MFR), also called signal regulatory protein α (SIRPα), and CD47 are transmembrane proteins that belong to the superfamily of immunoglobulins (Igs) (Han et al., 2000; Gautam and Acharya, 2014; Zhang et al., 2020). It was demonstrated that MFR and CD47 are induced at the onset of fusion both in vivo and in vitro, suggesting that these proteins contribute to macrophage fusion (Saginario et al., 1995; Saginario et al., 1998; Han et al., 2000; Podolnikova et al., 2019). Moreover, blocking of MFR or CD47, either by monoclonal antibodies or engineered proteins containing the extracellular domain of MFR or CD47, impairs fusion. Finally, IL-4 induced fusion is abrogated in RAW264.7 cells depleted for MRF by short hairpin (sh)RNA (Podolnikova et al., 2019). Han et al. proposed a model for MFR and CD47 mediated fusion. In order to guarantee cell-cell attachment, CD47 may initially interact with a “long” form of MFR on the neighboring macrophage, consisting of an extracellular immunoglobulin V domain and two adjacent immunoglobulin constant domains (Han et al., 2000; Vignery, 2005a). Afterwards, CD47 switches to a poorly expressed “short” form of MFR, only containing the immunoglobulin V domain. This homotypic interaction brings the plasma membranes of the two cells to a distance of 5–10 nm, allowing cell fusion (Han et al., 2000; Vignery, 2000, 2005a). MFR can also interact with the integrin CD11b contributing to fusion (Podolnikova et al., 2019). In macrophages, MFR and CD11b are detected at sites of cell-cell contact. Furthermore, IL-4 induced fusion has been observed in co-cultures of MFR and CD11b expressing HEK293 cells. Despite the findings favoring a role for MFR and CD47 in macrophage fusion, van Beek et al. reported that the number of nuclei remains unaltered in mice lacking the cytoplasmic signaling region of MFR, indicating that MFR or at least MFR signaling is dispensable for osteoclast fusion (van Beek et al., 2009). Additionally, osteoclasts derived from bone marrow cells of CD47 KO mice or differentiated in presence of MFR neutralizing antibodies do not differ in size or number of nuclei (Lundberg et al., 2007). Together, these findings suggest that MFR and CD47 may be involved in macrophage fusion other that osteoclast formation.

Dendritic cell-specific transmembrane protein (DC-STAMP) is a seven-transmembrane protein (Chiu et al., 2012) that is essential for macrophage fusion in human and mice (Yagi et al., 2005, 2006; Kim et al., 2008; Miyamoto, 2011; Zeng et al., 2013; Møller et al., 2020). The expression of DC-STAMP in human osteoclast precursors is positively correlated with the number of nuclei in mature osteoclasts (Møller et al., 2020). Moreover, suppression of DC-STAMP in human monocytes through lentivirus-mediated RNA interference impairs the formation of multinucleated osteoclasts (Zeng et al., 2013). Depletion of DC-STAMP in mice abrogates fusion of preosteoclasts into multinucleated osteoclasts both in vivo and in vitro (Yagi et al., 2005, 2006; Kim et al., 2008; Miyamoto, 2011) and reintroduction of DC-STAMP in osteoclast precursors derived from DC-STAMP-deficient mice rescues multinucleation (Yagi et al., 2005, 2006). Although osteoclast precursors isolated from DC-STAMP-deficient mice are unable to fuse, the expression of osteoclast markers, including TRAP (Yagi et al., 2005, 2006) and NFATc1, is not impaired (Yagi et al., 2006). Furthermore, ruffled borders and actin rings, typical morphological features of osteoclasts, are observed and cells exhibit minor bone-resorbing capacities (Yagi et al., 2005, 2006). These findings suggest that DC-STAMP is required for macrophage fusion without mediating osteoclast differentiation. Next to osteoclast multinucleation, DC-STAMP is indispensable for the formation of multinucleated FBGCs and LGCs (Yagi et al., 2005, 2006; Kim et al., 2008; Miyamoto, 2011; Sakai et al., 2012). Generation of multinucleated FBGCs is abolished in mice deficient for DC-STAMP in vivo and in vitro (Yagi et al., 2005, 2006; Kim et al., 2008; Miyamoto, 2011). Additionally, DC-STAMP is upregulated during LGC formation and siRNA-mediated knock-down of DC-STAMP inhibits fusion of human monocytes into LGCs (Sakai et al., 2012). Taken together, DC-STAMP is considered a common fusion mediator, yet different signaling mechanisms are responsible for induction of DC-STAMP in osteoclasts, FBGCs, and LGCs. During osteoclastogenesis, DC-STAMP is induced downstream of c-FOS and NFATc1 signaling (Yagi et al., 2007; Kim et al., 2008; Miyamoto, 2011; Sakai et al., 2012; Zhang et al., 2014), whereas PU.1 and NF-κB mediate the expression of DC-STAMP during FBGC formation (Yagi et al., 2007; Sakai et al., 2012), and probably NF-κB and MAP kinases during LGC formation (Sakai et al., 2012). It is speculated that DC-STAMP promotes macrophage fusion in a receptor-ligand fashion (Yagi et al., 2005), but the ligand for DC-STAMP required for macrophage fusion remains currently unknown (Quinn and Schepetkin, 2009). Since DC-STAMP shows structural similarities with chemokine receptors, it is speculated that a certain chemokine, such as CCL2, serves as ligand for DC-STAMP (Yagi et al., 2007; Quinn and Schepetkin, 2009). Other possible candidates include MFR and CD47. The exact mechanism by which DC-STAMP mediates macrophage fusion remains speculative as well. It is hypothesized that DC-STAMP induces the expression of fusogens (Zhang et al.), including MFR, and therefore indirectly mediates macrophage fusion (Vignery, 2005b).

Tetraspanins are four-span transmembrane proteins (Sangsri et al., 2020) that bind to one another and to a variety of other transmembrane proteins (Parthasarathy et al., 2009). Because of their ability to interact with several molecules, tetraspanins act as membrane organizers, mediating the generation of functional protein clusters in tetraspanin-enriched microdomains (Parthasarathy et al., 2009) of which some may modulate macrophage fusion. CD9 and CD81 are the best studied tetraspanins in the context of macrophage fusion, but whether they regulate fusion in a positive or negative manner remains controversial. For example, CD81 neutralizing antibodies support fusion of mouse macrophages infected with B. thailandensis, whereas inhibition of CD81 diminishes fusion upon infection with B. pseudomallei (Elgawidi et al., 2020; Sangsri et al., 2020). Additionally, CD9 neutralizing antibodies promote the formation of multinucleated osteoclast from murine bone marrow cells (Takeda et al., 2003), whereas blocking of CD9, either by neutralizing antibodies or small interfering RNA (siRNA), impairs the formation of multinucleated osteoclasts in RAW264.7 cells (Ishii et al., 2006). Although the effect of CD9 and CD81 on macrophage fusion remains unclear, CD63 seems to be a positive regulator (Takeda et al., 2003; Parthasarathy et al., 2009). CD63 is upregulated in human monocytes stimulated with Concanavalin A, and monoclonal antibodies directed against CD63 abrogate MGC formation. Tspan-5 and NET-6 have also been suggested to influence macrophage fusion as Tspan-5 is upregulated during osteoclastogenesis and NET-6 is downregulated (Iwai et al., 2007). Furthermore, silencing of Tspan-5 impairs fusion of RAW264.7 into multinucleated osteoclasts, whereas the opposite is true for NET-6.

Matrix metalloproteases (MMPs) compromise a family of endopeptidases known to degrade various extracellular matrix components (EMCs) (Gonzalo et al., 2010; Kim and Lee, 2020). MMPs may modulate macrophage fusion by degrading EMCs and/or cell-surface bound molecules involved in cell fusion (Jones et al., 2008). In vitro and in vivo fusion of macrophages into osteoclasts and FBGCs is abolished in mice deficient for MT1-MMP, a membrane-anchored MMP (Gonzalo and Arroyo, 2010; Gonzalo et al., 2010). Depletion of MT1-MMP affects macrophage spreading, formation of lamellipodia, and Rac1 activity, indicating that impaired migration lies at the basis of reduced fusion events in these cells (Gonzalo and Arroyo, 2010). MT1-MMP probably promotes fusion through its interaction with adaptor protein p130Crk-associated substrate (CAS), which in turn triggers Rac1 activation (Gonzalo and Arroyo, 2010; Gonzalo et al., 2010). Although MT1-MPP positively regulates macrophage fusion (Gonzalo and Arroyo, 2010; Gonzalo et al., 2010), MMP8 and MMP13 have been described to negatively impact fusion (Kim and Lee, 2020). RANKL decreases the expression of MMP8 and MMP13 in murine bone marrow macrophages, indicating that their down-regulation is required for proper osteoclastogenesis. Targeting these MMPs, either through siRNA or pharmacological inhibition, facilitates osteoclast fusion without altering differentiation (Kim and Lee, 2020). These findings are in contrast to data from Fu et al. who demonstrated that MMP13 promotes osteoclast formation starting from mouse bone marrow macrophages and that fusion is abrogated in macrophages derived from MMP13-deficient mice (Fu et al., 2016). Several reports indicate that MMPs also regulate FBGC formation. Pharmacological inhibition experiments suggested that MMP1, -8, -13, and -18 act as positive regulators of IL-4 induced fusion in human monocytes, whereas MMP2, -3, -9, and -12 do not influence fusion (Jones et al., 2008). Notwithstanding the observation that MMP9 is not implicated in human macrophage fusion, MMP9 neutralizing antibodies reduce IL-4 induced fusion of mouse macrophages and MMP9-deficient mice display reduced FBGC formation after implantation of biomaterials (Maclauchlan et al., 2009). The impaired fusion observed in MMP9-deficient mouse macrophages may result from defective cytoskeletal dynamics prior to cell fusion. The discrepancy on the effect of MMP-9 on macrophage fusion between human and mice may be explained by suboptimal specificity of pharmacological MMP inhibitors used in human experiments.