Interaction between the integrin Mac-1 and signal regulatory protein α (SIRPα) mediates fusion in heterologous cells

Macrophage fusion leading to the formation of multinucleated giant cells is a hallmark of chronic inflammation. Several membrane proteins have been implicated in mediating cell–cell attachment during fusion, but their binding partners remain unknown. Recently, we demonstrated that interleukin-4 (IL-4)-induced fusion of mouse macrophages depends on the integrin macrophage antigen 1 (Mac-1). Surprisingly, the genetic deficiency of intercellular adhesion molecule 1 (ICAM-1), an established ligand of Mac-1, did not impair macrophage fusion, suggesting the involvement of other counter-receptors. Here, using various approaches, including signal regulatory protein α (SIRPα) knockdown, recombinant proteins, adhesion and fusion assays, biolayer interferometry, and peptide libraries, we show that SIRPα, which, similar to ICAM-1, belongs to the Ig superfamily and has previously been implicated in cell fusion, interacts with Mac-1. The following results support the conclusion that SIRPα is a ligand of Mac-1: (a) recombinant ectodomain of SIRPα supports adhesion of Mac-1–expressing cells; (b) Mac-1–SIRPα interaction is mediated through the ligand-binding αMI-domain of Mac-1; (c) recognition of SIRPα by the αMI-domain conforms to general principles governing binding of Mac-1 to many of its ligands; (d) SIRPα reportedly binds CD47; however, anti-CD47 function-blocking mAb produced only a limited inhibition of macrophage adhesion to SIRPα; and (e) co-culturing of SIRPα- and Mac-1–expressing HEK293 cells resulted in the formation of multinucleated cells. Taken together, these results identify SIRPα as a counter-receptor for Mac-1 and suggest that the Mac-1–SIRPα interaction may be involved in macrophage fusion.

Cell-cell fusion is a fundamental biological process that is required for development and homeostasis (1)(2)(3). Cellular fusion leading to the formation of multinucleated cells is essential for the formation of skeletal muscles (myoblast fusion), placental morphogenesis (fusion of trophoblast cells), fertilization (fusion of sperm and egg), and bone resorption (fusion of cells of the monocyte/macrophage lineage resulting in osteoclast generation). Cell fusion may also play a role in a number of pathological conditions. In particular, in contrast to most fusing cell types, which undergo fusion as a part of their normal development program, fusion of macrophages leading to the formation of multinucleated giant cells (MGCs) 4 is a common feature of many granulomatous infections (4). It also accompanies chronic inflammatory conditions, including foreign body reactions to implanted biomaterials, rheumatoid diseases, giant cell arteritis, and others (4,5). The functional role of macrophage fusion in these diseases remains unclear, but recent studies favor the idea that the increased membrane area arising from multinucleation facilitates phagocytosis of large particles and thus the removal of debris from tissues (6).
Although the molecular mechanisms of intracellular membrane fusion mediated by SNAREs and virus-host cell fusion have been established, the processes that mediate fusion of the plasma membranes, including those of macrophages, remain poorly understood (7). Consistent with the requirement for cell-cell contacts as an obligatory step in fusion, several receptors on the surface of macrophages, including SIRP␣ (MFR), CD47, CD44, E-cadherin, tetraspanins, DC-STAMP, and others, have been shown to participate in this process (2). However, the counter-receptors for these molecules, perhaps with the exception of E-cadherin, which mediates homophilic interactions, are poorly characterized. Previous studies demonstrated that integrin Mac-1 (␣ M ␤ 2 , CD11b/CD18, and CR3), a multiligand receptor abundantly expressed on the surface of macrophages, is essential for MGC formation in vitro inasmuch as IL-4 -induced fusion of Mac-1-deficient macrophages was reduced (8,9). The examination of adhesive reactions known to be required for fusion showed that only macrophage spreading, but not adhesion to Permanox plastic, a surface permissive for fusion, was reduced in Mac-1-deficient cells (9). Furthermore, migration of IL-4 -induced WT and Mac-1-deficient macro-phages was similar (9). Although Mac-1-initiated signaling leading to cytoskeletal rearrangements and cell spreading may be critical early events during macrophage fusion, Mac-1 may fulfill other functions.
Macrophage fusion requires bringing two plasma membranes together and may involve the interaction of Mac-1 with its counter-receptor(s) on opposing cells. In addition to its role in cell adhesion to the extracellular matrix, Mac-1 interacts with several counter-receptors on other cells, including ICAM-1 (10). ICAM-1 is expressed on the surface of fusing macrophages (11,12). However, our investigations using ICAM-1-deficient murine macrophages did not support the essential involvement of this molecule in fusion (9), suggesting that Mac-1 can interact with other counter-receptor(s).
It is widely accepted that molecules containing Ig-like domains are involved in fusion. For example, recognition and adhesion between Drosophila myoblasts are mediated by Ig-domain-containing transmembrane proteins (13,14). We have tested the hypothesis that signal regulatory protein ␣ (SIRP␣), which, similar to ICAM-1, belongs to the Ig superfamily, interacts with Mac-1. SIRP␣ (also known as a macrophage fusion receptor, MFR) was one of the first discovered molecules implicated in macrophage fusion (15). The experiments in this study describe the utilization of a variety of cell biology and biochemistry techniques to show that SIRP␣ is a ligand for Mac-1. We also provide evidence of direct interaction between the ␣ M I-domain, a ligand-binding region of Mac-1, and the extracellular domain of SIRP␣. Furthermore, we established a cell-fusion system with HEK293 cells transfected separately with Mac-1 and SIRP␣ to show that co-culturing these cells in the presence of IL-4 results in cell fusion.

SIRP␣ is critical for macrophage fusion
Previous studies using mAbs raised against SIRP␣ expressed in rat alveolar macrophages demonstrated that SIRP␣ is induced by ϳ1.5-2-fold at the onset of fusion (15,16) and that the recombinant ectodomain of SIRP␣ inhibited fusion (15), suggesting the role for this receptor in macrophage fusion. We showed that SIRP␣ is expressed in mouse thioglycollate-elicited peritoneal macrophages, and its expression is increased by ϳ1.4-fold after 6 h in culture in the presence of fusion-promoting cytokine IL-4 and is then gradually elevated (ϳ1.7-fold) until 48 h (Fig. S1, A and B). To substantiate the role of SIRP␣ in fusion, we performed knockdown (KD) of SIRP␣ in a RAW264.7 macrophage cell line using shRNA. Stable clones expressing shRNAs were selected, expanded, and sorted to obtain cells with no SIRP␣ expression (Fig. 1A). Macrophage fusion was induced by IL-4 on fusion-permissive Permanox dishes. As shown in Fig. 1, B and C, fusion of selected SIRP␣-KD cells was efficiently ablated compared with wildtype (WT) RAW264.7 cells and cells transduced with control GFP-expressing viruses. Macrophage fusion is known to involve adhesion of cells to the surface (2). Consequently, we determined whether the procedure affected the adhesive properties of cells. We showed that SIRP␣-KD cells adhered to Permanox to a similar extent as control cell (Fig. 1D). Because macrophages move during fusion in vitro (2,9), we also examined whether SIRP␣-KD cells have different migratory behavior during IL-4induced fusion. Using live-cell microscopy, we found no difference in the rate of migration of control and SIRP␣-KD cells (Fig.  1E). We recently showed that integrin Mac-1 is required for macrophage fusion induced by IL-4 (9). To determine whether knockdown of SIRP␣ affected expression of Mac-1 on the surface of RAW264.7, we performed FACS analyses using mAb 44a against the ␣ M integrin subunit. As shown in Fig. 1F, similar mean fluorescence intensity signals were obtained using selected SIRP␣-KD and lentivirus control cells (1520 Ϯ 40 versus 1640 Ϯ 130, respectively). Furthermore, the activation state of Mac-1 probed with an activation-dependent mAb CBRM1/5 was similar in both cell lines (Fig. 1G). These results indicate that SIRP␣ is a critical mediator of macrophage fusion. Furthermore, because the absence of SIRP␣ on the surface of macrophages does not alter the adhesive and migratory properties of macrophages, the data suggest that during fusion SIRP␣ may perform other functions, potentially serving as a binding partner for cell-surface molecules.

Integrin Mac-1 mediates cell adhesion to SIRP␣ and its individual Ig domains
To assess the possibility that SIRP␣ might serve as a counterreceptor for Mac-1, we expressed recombinant fragments corresponding to the entire ectodomains of homologous mouse (m) and human (h) SIRP␣. In addition, we prepared three individual mSIRP␣ Ig-like domains or their combinations ( Fig. 2A). After isolation by affinity chromatography, the fragments were characterized by size-exclusion chromatography to confirm their monomeric state (Fig. 2B, left panel). In addition, SDS-PAGE analyses provided additional verification of the proteins' homogeneous monomeric state (Fig. 2B, right panel). When immobilized on microtiter plates, both mouse and human SIRP␣ fragments (termed Ig1-2-3) supported efficient adhesion of Mac-1-expressing HEK293 cells (Mac-1-HEK293), a cell line often used for investigations of the capacity of proteins to serve as Mac-1 ligands (Fig. 3, A and C) (17,18). By contrast, no adhesion of WT HEK293 cells (WT HEK293) was observed (Fig. 3, A and C). Adhesion of Mac-1-HEK293 cells to SIRP␣ was similar to that of fibrinogen, a well-characterized Mac-1 ligand (Fig. S2) (19,20). In agreement with the involvement of Mac-1 in binding to SIRP␣, function-blocking mAb 44a against the ␣ M integrin subunit of Mac-1, but not an isotype control antibody, inhibited adhesion of Mac-1-HEK293 cells to both mouse and human proteins (Fig. 3, B and C). The specificity of the interaction was further confirmed using HEK293 cells expressing the related monospecific ␤2 integrin ␣ L ␤ 2 (LFA-1); no cell adhesion to SIRP␣ was observed (Fig. 3C). The interaction between Mac-1 and hSIRP␣ was mediated by the ␣ M Idomain of Mac-1 inasmuch as cells expressing the "I-less" form of the receptor were incapable of supporting adhesion (Fig. 3C). The lack of adhesion of LFA-1-and I-less-Mac-1-HEK293expressing cells to SIRP␣ was not due to the different density of receptors because, as assessed by flow cytometry, both cell lines expressed the same levels of the ␤2 integrin subunit as Mac-1-HEK293 cells (Fig. S3). The ability of individual Ig domains of mSIRP␣ (Ig1, Ig2, and Ig3) or their combinations (Ig1-2 and SIRP␣ is a counter-receptor for integrin Mac-1  to support adhesion of Mac-1-HEK293 cells was also examined. As shown in Fig. 3D, all fragments were capable of supporting concentration-dependent cell adhesion, albeit to a different extent. Because adsorption of proteins on plastic leads to their unfolding, which may result in the exposure of cryptic Mac-1binding sites, we examined whether soluble SIRP␣-derived fragments can interact with Mac-1. Mac-1-HEK293 cells were preincubated with 1.5 M of each soluble fragment and then added to the wells coated with mSIRP␣ Ig1-2-3. All of the tested fragments significantly decreased cell adhesion (Fig. 3E), with Ig2-3 and Ig3 being most active (92 Ϯ 2 and 84 Ϯ 3% of inhibition, respectively). It was shown that SIRP␣ can dimerize in cis on the cell surface with all three Ig extracellular domains being implicated in the complex formation and with N-linked protein glycosylation playing some role in dimerization (21). Although our recombinant fragments were produced in Escherichia coli cells that lack this modification, we nevertheless examined whether the inhibitory effect of individual domains on cell adhesion might have arisen from complex formation between the immobilized mSIRP␣ Ig1-2-3 and soluble fragments, resulting in masking of the Mac-1-binding site(s). In these experiments, the Ig1-2-3 protein immobilized onto microtiter wells was first preincubated with 1.5 M of each fragment, after which cells were added. In contrast to the experimental format in which cells were initially preincubated with the fragments (Fig. 3E), little inhibition was observed (Fig. S4), suggesting that the observed blocking of adhesion was mainly due to the interaction between Mac-1 and soluble SIRP␣ fragments.
We next assessed the ability of the mouse macrophage cell line RAW264.7 and mouse peritoneal inflammatory macrophages naturally expressing Mac-1 to interact with mSIRP␣. As

SIRP␣ is a counter-receptor for integrin Mac-1
shown in Fig. 4A, PMA-stimulated RAW264.7 macrophages and peritoneal macrophages adhered to mSIRP␣ Ig1-2-3 in a concentration-dependent manner. The mAb M1/70 directed to the mouse ␣ M integrin subunit inhibited adhesion of RAW264.7 cells by 80 Ϯ 2%, suggesting that Mac-1 is involved in adhesion to mSIRPa (Fig. 4B). The specificity of M1/70's effect was established using rat IgG2b, an isotype control for mAb M1/70. Preincubation of RAW264.7 cells with this IgG did not inhibit cell adhesion (Fig. 4B). Partial inhibition of adhesion by mAb M1/70 suggested that in addition to Mac-1, other structures on the surface of macrophages may be involved in binding to mSIRP␣. To investigate this possibility, we examined whether heparan sulfate and chondroitin sulfate proteoglycans, which commonly cooperate with Mac-1 in adhesion to its ligands (22)(23)(24)(25), are involved in adhesion to SIRP␣. Whereas preincubation of RAW264.7 cells with 10 g/ml of each heparin, chondroitin sulfate A, and chondroitin sulfate B partially inhibited adhesion (35 Ϯ 1, 47 Ϯ 8, and 28 Ϯ 3%, respectively), their combination with mAb M1/70 did not potentiate the inhibitory effect of the mAb (Fig. S5). Some involvement of CD47, a known ligand for SIRP␣ (26), was noted as mAb miap301 against mouse CD47 inhibited adhesion by 29 Ϯ 5%, whereas an isotype control IgG2a inhibited adhesion by ϳ5% (Fig. 4B). However, the combination of anti-CD47 and M1/70 did not produce inhibition greater than that caused by mAb M1/70 alone. Similar to RAW264.7 cells, anti-Mac-1 and anti-CD47 mAbs partially inhibited adhesion of mouse peritoneal macrophages (by 56 Ϯ 2 and 46 Ϯ 4%, respectively) ( Fig.  4C). However, in these cells, the effect of anti-Mac-1 was less pronounced, whereas anti-CD47 inhibited adhesion to a greater extent than in RAW264.7 cells. Furthermore, the combined effect of two antibodies was slightly, albeit significantly, greater (66 Ϯ 2%) than that caused by anti-Mac-1 mAb M1/70 alone, suggesting that both receptors may contribute to adhesion. Together, the cell adhesion data identify the Mac-1 integrin on the surface of macrophages as a receptor for SIRP␣. They also suggest that Mac-1 may cooperate with CD47.

Mac-1 and CD47 form a complex on the surface of Mac-1-HEK and RAW264.7 cells
Because anti-CD47 mAb partially decreased adhesion of RAW264.7 cells to mSIRP␣ (Fig. 4C), we examined the contribution of CD47 to adhesion of Mac-1-HEK293 cells, which also express this cell-surface protein (Fig. 5A, left panel). In these experiments, we used mAb B6H12, which reportedly inhibited adhesion of CD47 on erythrocytes to the immobilized recombinant ectodomain of hSIRP␣ (27). Although inhibition of cell adhesion to hSIRP␣ by mAb B6H12 was detected, the effect appeared to be nonspecific because an isotype IgG1 control decreased adhesion to a similar extent (46 Ϯ 6 versus 42 Ϯ 5%) (Fig. 5B). Furthermore, consistent with the lack of specific inhibition, B6H12 also decreased cell adhesion to mSIRP␣, which

SIRP␣ is a counter-receptor for integrin Mac-1
does not interact with human CD47 (Fig. 5B) (28). These data suggested that CD47 expressed on Mac-1-HEK293 cells was not involved in the interaction with hSIRP␣ during adhesion. The lack of interaction between CD47 and hSIRP␣ is in agreement with complete inhibition of adhesion of these cells by anti-Mac-1 mAb 44a and consistent with the inability of SIRP␣ to support adhesion of WT HEK293 cells (Fig. 3, A and B).
Nevertheless, the specific effect of anti-CD47 mAbs on adhesion of RAW264.7 macrophages (Fig. 4C) suggested that Mac-1 and CD47, which are present on these cells (Fig. 5A, right panel), may separately interact with different SIRP␣ molecules. Alternatively, Mac-1 can form a lateral complex with CD47, which may engage SIRP␣. CD47 was shown to interact with ␤1 and ␤3 integrins (29) and proposed to interact with ␤ 2 integrins SIRP␣ is a counter-receptor for integrin Mac-1 (30). However, the latter interaction was not documented. To investigate the latter possibility, we immunoprecipitated Mac-1 from RAW264.7 and Mac-1-HEK293 cells using mAbs against the corresponding ␣ M integrin subunits and analyzed immune complexes for the presence of CD47. As shown in Fig. 5C, mAbs 44a and M1/70 precipitated the ␣ M and ␤ 2 integrin subunits from Mac-1-HEK293 and RAW264.7 cells, respectively, as well as a protein with a molecular mass of 47 kDa. Western blotting analyses confirmed the identity of this protein as CD47 in the precipitates from both types of cells (Fig. 5D). Conversely, anti-CD47 mAbs immunoprecipitated ␣ M and ␤ 2 integrin subunits (Fig. 5C). Control immunoprecipitations performed with isotype-specific IgGs for mAbs against human and mouse antigens apparently did not pull down proteins (Fig. 5, C and D). These experiments suggest that Mac-1 forms a lateral complex with CD47, which potentially may engage the SIRP␣ molecule(s) on the opposite cell.

Binding of the ␣ M I-domain to the ectodomain of SIRP␣
The adhesion data in Fig. 3, B and C, suggested that the ␣ M Idomain of Mac-1 mediates SIRP␣ binding. To corroborate these data, we analyzed the binding parameters of the interaction between the ␣ M I-domain and hSIRP␣ by biolayer interferometry (BLI). Because ␣ M I-domain can exist in two different states, active and inactive, with the length of the C-terminal ␣-helix regulating its activation state (31-33), we examined both conformers. Soluble Ig1-2-3 ectodomain of hSIRP␣ was coupled to the matrix on the biosensor, and the binding of ␣ M Idomains was measured in buffer containing 1 mM MgCl 2 . Fig.  6A shows that active ␣ M I-domain bound to SIRP␣ in a dose-dependent manner. The K D value for the binding was found to be 1.4 M. As a specificity control, the interaction was inhibited by mAb 44a directed to the ␣ M I-domain (Fig. 6B). The inactive form of the ␣ M I-domain minimally bound SIRP␣ (Fig. 6B). In contrast to the active ␣ M I-domain, the active form of ␣ L I-domain derived from the monospecific integrin ␣ L ␤ 2 (LFA-1) did not interact with SIRP␣. These results indicate that similar to many other ligands, the binding of Mac-1 to SIRP␣ is mediated by ␣ M I-domain, and this interaction requires the active state of ␣ M I-domain.

Identification of ␣ M I-domain-binding sequences in the extracellular Ig domains of SIRP␣
As shown previously, within its ligands the ␣ M I-domain can bind short sequences enriched in basic and hydrophobic residues (34). To examine whether the binding specificity of Mac-1 toward SIRP␣ conforms to this recognition principle, we analyzed the sequences of the Ig domains of hSIRP␣ by the computer program that determines the capacity of 9-mer peptides spanning the sequences of Mac-1 ligands to interact with the ␣ M I-domain (34). It assigns each peptide the energy value that serves as a measure of probability for the ␣ M I-domain that binds this sequence: the lower the energy, the higher the likeli-  To demonstrate the importance of the identified SIRP␣ sequences for interaction with Mac-1, we examined the ability of hSIRP␣-derived peptides to support Mac-1-mediated cell adhesion. In these experiments, we synthesized selected peptides that correspond to the ␣ M I-domain-binding sequences in each Ig domain and that are fully exposed on the surface of the protein (Fig. 7C): 36 IQWFRGAGP 44 (termed PD1), 149 ITLKWFKNG 157 (termed PD2), and 245 RKFYPQRLQ 253 (termed PD3). As shown in Fig. 7D, all three peptides supported efficient adhesion of Mac-1-HEK293 cells, whereas the control peptide 161 SDFQTNVDP 169 (13.2 kJ/mol) was negative. We further tested the ability of peptides to block cell adhesion to hSIRP␣. Preincubation of cells with soluble peptides resulted in a dose-dependent inhibition of adhesion (data not shown). At 100 g/ml (maximal testable concentration), all peptides decreased adhesion by ϳ50% (Fig. 7E), whereas the control peptide had no effect. These findings indicate that similar to other Mac-1 ligands, the ␣ M I-domain recognizes the SIRP␣ peptides enriched in basic and hydrophobic residues and suggests that all three Ig domains contain putative ␣ M I-domainbinding sites.

Reconstitution of cell-cell fusion in the mixture of Mac-1-and SIRP␣-expressing HEK293 cells
Previous studies demonstrated that HEK293 cells stably transfected with the P2X 7 receptor cDNA can fuse forming multinucleated cells (35). We used HEK293 cells to examine the hypothesis that the interaction between Mac-1 and SIRP␣ can promote fusion. The hSIRP␣ was stably expressed in HEK293 cells (Fig. 8A), and these cells were mixed with Mac-1-HEK293 cells at 1:1, 1:2, and 1:4 ratios. Cells were plated on Permanox dishes, and IL-4 was added to the mixture for 72 h. As shown in Fig. 8B, cell fusion was detected (three representative images are shown). Although these events were rare (ϳ3-4%, n ϭ 4) at a 1:1 SIRP␣-HEK293/Mac-1-HEK293 cell ratio, the fraction of fused multinucleated cells increased to 15 Ϯ 3% (n ϭ 3) at a 1:4 ratio. In contrast, fused cells were never observed in the populations of IL-4 -treated WT HEK293 cells or among cells expressing either Mac-1 or SIRP␣ alone (Fig. 8C). It is unlikely that multinucleated cells in the mixture of Mac-1-and SIRP␣expressing cells originated from karyokinesis in the absence of cytokinesis, as some cells contained 6 -20 nuclei. Furthermore, confocal images of multinucleated cells stained with anti-␣ M mAb 44a showed the presence of two labels, Mac-1 (red) and endogenous GFP (co-expressed with SIRP␣), which is only possible if SIRP␣-expressing cells fused with Mac-1-expressing cells (Fig. 8D). In contrast, individual Mac-1-HEK293 and SIRP␣-HEK293 cells (not treated with IL-4) were seen as small mononuclear cells stained with either red or green labels (Fig.  8D, bottom panel). In addition, the presence of two labels was
Having established conditions for the induction of HEK293 cell fusions, we examined the fusion-promoting capacity of individual SIRP␣ Ig domains. The SIRP␣ proteins lacking the Ig1 (termed hSIRP␣ 2-3) and Ig1-2 (termed hSIRP␣ 3) extracellular domains were expressed on the surface of HEK293 cells (Fig. 8A), and their fusion was compared with that of cells expressing the entire SIRP␣ ectodomain. The HEK293 cells expressing WT and SIRP␣ mutants were mixed with Mac-1-HEK293 cells at a 1:4 ratio and allowed to fuse for 72 h. Fig.  8F shows that the fusion index of cells expressing hSIRP␣-Ig2-3 was lower than that of WT protein. In contrast, cells expressing hSIRP␣-Ig3 fused somewhat better than WT cells. Although the trend in both cases was clearly seen, the difference between WT and mutant cells did not reach a statistical significance. Nevertheless, the fusion rate of cells expressing only Ig3 was significantly higher than that of cells expressing Ig2-3. Although the interpretation of these results is difficult, the higher fusion-promoting capacity of Ig3 is in agreement with its higher activity in blocking the interaction of Mac-1-HEK293 cells with SIRP␣ in adhesion assays (Fig. 3E). Furthermore, the ability of Ig3 alone to support fusion further suggests that on the surface of Mac-1-HEK293 cells, CD47 is not involved in the interaction with SIRP␣ because it reportedly interacts with Ig1 of SIRP␣ (16,36,37).

Discussion
In a previous in vitro study, we established that integrin Mac-1 plays an essential role in macrophage fusion (9). We also demonstrated that macrophages lacking ICAM-1, a counterreceptor for Mac-1, undergo fusion to a similar extent as WT cells implicating an unknown molecule(s) on the surface of fusing macrophages as a counter-receptor for this integrin. The major finding of this study is that SIRP␣ serves as a heretofore unrecognized ligand for Mac-1. The following data support this conclusion. First, the recombinant ectodomain of SIRP␣ encompassing three Ig-like domains (Ig1-2-3) supports adhesion of Mac-1-expressing HEK293 cells as well as natural macrophages. Second, the functional region of Mac-1, which mediates the interaction with SIRP␣, is the ␣ M I-domain. Third, recognition specificity of the ␣ M I-domain toward SIRP␣ conforms to the expected pattern exhibited by many Mac-1 ligands. Fourth, co-culturing of HEK293 cells separately expressing Mac-1 and SIRP␣ results in cell fusion and formation of multinucleated cells, whereas no fusion is observed in cell populations expressing only one of these receptors. Fifth, although CD47 was previously reported to bind SIRP␣, no significant involvement of CD47 in the binding of Mac-1-expressing cells to SIRP␣ was detected: instead, CD47 was found to form a lateral complex with Mac-1.
Integrin Mac-1 is a major adhesion receptor on the surface of myeloid cells. This receptor exhibits multiligand binding properties enabling it to bind a variety of proteins in the extracellular matrix as well as numerous cationic proteins released from stimulated neutrophils and damaged cells during the inflammatory response (18,22,34). In addition, Mac-1 can bind coun-ter-receptors expressed on a number of cells, including ICAM-1, GPIb-IX, and JAM-3 (10,38,39). Two of these molecules, ICAM-1 and JAM-3, contain multiple extracellular Iglike domains. It has been noted that the molecules containing Ig-like domains widely participate in fusion events mediating cell-cell tethering (14). For example, recognition and adhesion between Drosophila melanogaster myoblasts are mediated by Ig-like domain-containing transmembrane proteins DUF, RST, and SNS (1,14). Similar to ICAM-1, SIRP␣ is a transmembrane protein that contains three extracellular Ig domains (the N-terminal IgV followed by two IgC1 domains). A shorter variant of SIRP␣ lacks the C1 domains and contains only the IgV domain. SIRP␣, also known as MFR, was identified initially by antibodies that blocked fusion of rat alveolar macrophages and was shown to be up-regulated after 24 h in culture under fusogenic conditions (40). We found that SIRP␣ is also up-regulated on the surface of IL-4 -treated mouse peritoneal macrophages (Fig.  S1). To test Mac-1-SIRP␣ binding, we initially characterized the interaction between the Mac-1-expressing cells and the ectodomain of SIRP␣ (Ig1-2-3) using adhesion assays. Both human and mouse Mac-1-expressing cells supported efficient Mac-1-dependent adhesion to immobilized SIRP␣. We found that all three Ig-like domains of SIRP␣ have the capacity to interact with Mac-1, with Ig3 being the most active. However, although Mac-1 appears to be the main receptor for SIRP␣ on the surface of Mac-1-expressing HEK293 cells (Fig. 3), CD47 on the surface of RAW264.7 cells and mouse peritoneal macrophages may contribute to adhesion (Fig. 4, B and C).
Because immobilization of proteins on plastic may result in their partial denaturation resulting in exposure of hidden interior sequences that are favored by Mac-1 (41), we have also tested the interaction of Mac-1-expressing cells with soluble SIRP␣ fragments. The SIRP␣ ectodomain, as well as recombinant fragments duplicating its constituent Ig-like domains, blocked cell adhesion (Fig. 3E), suggesting that Mac-1 may bind intact SIRP␣ expressed on the cell surface. The activity of soluble fragments in inhibition adhesion assay appears to closely correspond to their capacity to support Mac-1-mediated cell adhesion, with Ig3 again being most potent. Furthermore, the interaction between SIRP␣ and the recombinant ligand-binding ␣ M I-domain of Mac-1 has been detected using BLI. In this system, SIRP␣ Ig1-2-3 was immobilized to a dextran-coated surface by chemical cross-linking, a procedure that largely preserves the native protein conformation. Based on these results, we propose that SIRP␣ is a novel counter-receptor of Mac-1 on the surface of macrophages involved in cell-cell interactions.
The characteristic feature of some Mac-1 ligands is the presence of short sequences containing basic residues surrounded by hydrophobic residues, which form ␣ M I-domain recognition motifs (34). Screening of the peptide library spanning the sequence of hSIRP␣ Ig1-2-3 revealed that it contains several typical ␣ M I-domain recognition motifs in all three Ig domains. The ability of selected SIRP␣-derived peptides to interact with Mac-1 in adhesion and inhibition adhesion assays (Fig. 7) recapitulates the behavior of other well-characterized Mac-1 peptide ligands (23,24,42,43). Consistent with the presence of multiple recognition sites in Mac-1 ligands (34, 44), the ␣ M Idomain-binding sequences have been found in all three Ig

SIRP␣ is a counter-receptor for integrin Mac-1
domains of SIRP␣ (Fig. 7C). The greater activity of SIRP␣ Ig3 in adhesion (Fig. 3, D and E) and fusion assays (Fig. 8F) seems to implicate this domain as a major site in Mac-1 binding, which may be attributable to the presence of an extended cluster of the ␣ M Idomain-binding sequences containing 245 RKFYPQRLQ 253 , the most active peptide in the SIRP␣-derived peptide library (Fig. 7A). Previous studies showed that CD47, a widely distributed plasma membrane protein, is a ligand for SIRP␣ (26,29). Furthermore, the recombinant soluble ectodomain of CD47 blocked fusion of rat alveolar macrophages in culture, suggesting the involvement of CD47 in macrophage fusion (16). CD47, which also belongs to the Ig superfamily, contains one Nterminal extracellular variable Ig domain (IgV) followed by five transmembrane segments. Because the recombinant IgV domain of CD47 binds both forms of SIRP␣, long and short, it has been proposed that the IgV domain of CD47 interacts with the IgV domain of SIRP␣ (16). The interaction between the IgV domains of these two molecules has been confirmed by solving the three-dimensional structure of their complex (37). The lack of adhesion of HEK293 cells, which express CD47, to the ectodomain of SIRP␣ in our experiments and the finding that adhesion of Mac-1-expressing HEK293 cells was completely inhibited by anti-Mac-1 mAb, but not by anti-CD47 blocking antibody, suggest that in these cells CD47 is not involved in SIRP␣ binding. The SIRP␣-CD47 interactions exhibit little cross-reactivity across species (27). In this regard, human SIRP␣ does not bind mouse CD47 (27), and mouse SIRP␣ does not significantly interact with human CD47 (28). Therefore, the finding that both mouse and human SIRP␣ supported adhesion of Mac-1-HEK293 cells (Fig. 3) seems to further support the CD47-independent binding of these cells to SIRP␣. Nevertheless, on the surface of RAW264.7 and mouse peritoneal macrophages, CD47 may potentially contribute to binding to SIRP␣, because the antibody specific for mouse CD47 (miap301) partially blocked cell adhesion to mouse SIRP␣ (Fig. 4, B and C). The inhibitory effect of both anti-Mac-1 and anti-CD47 mAbs suggests a functional relationship between Mac-1 and CD47.
CD47 has been shown to interact with ␤1 and ␤3 integrins, including ␣ v ␤ 3 , ␣ IIb ␤ 3 , and ␣ 2 ␤ 1 , and proposed to interact with ␤ 2 integrins (30). However, the latter interaction has not been documented. We show for the first time that anti-Mac-1 mAbs immunoprecipitated CD47 from Mac-1-expressing HEK293 cells and RAW264.7 macrophages and, conversely, the mAbs directed to CD47-immunoprecipitated Mac-1 (Fig. 5). This indicates that similar to other integrins, Mac-1 interacts with CD47 in cis (schematically shown in Fig. 9). This also suggests that this association may influence the function of integrin. Our recent studies showing that IL-4 -induced macrophage fusion is most efficient when Mac-1 is present on each fusion partner may provide indirect support for this idea (9). In particular, we showed that the ability of WT macrophages to fuse with Mac-1-deficient counterparts was strongly impaired, and fusion in the mixture containing WT and Mac-1-deficient macrophages occurred mainly between WT cells. Because CD47 and SIRP␣ were both present on WT and Mac-1-deficient cells, where they could have mediated cell-cell interaction, these data suggest that the CD47-SIRP␣ interaction alone on Mac-1deficient macrophages does not support fusion. At present, the ways by which CD47 can modify Mac-1 functions remain to be defined.
It is unclear why the interaction between Mac-1 and its counter-receptor ICAM-1 is not involved in macrophage fusion (9). ICAM-1 is expressed on the surface of mouse blood monocytes and macrophages, although its expression does not correlate with multinucleation (45). One possibility for the lack of involvement of ICAM-1 may be differential localization of Mac-1 and ICAM-1 on fusing macrophages. Indeed, we noted that Mac-1 and SIRPa were present at the sites of cell-cell contact between macrophages (Fig. S6A), whereas ICAM-1 was rarely observed (Fig. S6B). Furthermore, both Mac-1 and SIRP␣ were detected on filopodia, which may facilitate cell-cell recognition and adhesion. Macrophages express several other members of the Ig superfamily, including CD4 and CD200 and the role of CD200 in the fusion of osteoclasts, which originate from fusion of receptor activator of NF-B ligand (RANKL)treated macrophages as reported previously (46). Based on the resemblance of SIRP␣ and CD200 structures, it will be interesting to examine whether CD200 is induced in macrophages exposed to IL-4 and whether Mac-1 is capable of engaging this molecule during fusion.
Although fusion of normally nonfusing HEK293 cells after transfection with P2X 7 has been shown (35), the finding of fusion of HEK293 cells after transfection with Mac-1 and SIRP␣ was still surprising. It has been proposed that proteins that fuse cells should fulfill several "gold standards," including the ability to fuse heterologous cells in culture and be expressed at the time and place of fusion (47). Indeed, transfection of heterologous nonfusing insect cells with Caenorhabditis elegans fusogenic protein EFF-1 reconstituted cell fusion (7,48). Although SIRP␣ is expressed at the time and place of fusion, thus obeying one of the rules, transfection of HEK293 cells with SIRP␣ alone was not sufficient to induce fusion (Fig. 8). It was only after

SIRP␣ is a counter-receptor for integrin Mac-1
mixing SIRP␣-with Mac-1-expressing HEK293 cells that fusion was detected. Yet, it is unlikely that SIRP␣ and Mac-1 are authentic fusogenic proteins, i.e. the proteins that mediate fusion of plasma membranes. Rather, these molecules are more likely required to bring membranes into close proximity before fusion and thus participate in the cell-cell adhesion step. If this is the case, then other undefined molecules on the surface of HEK293 cells may serve as fusogens.
In summary, we have identified SIRP␣ as a novel counterreceptor for Mac-1 and showed that the interaction of these molecules is mediated by the ␣ M I-domain of Mac-1. It is interesting to discover that two molecules, Mac-1, and SIRP␣ that have been independently implicated in fusion are, in fact, binding partners. Based on the ability of Mac-1 and SIRP␣ to mediate fusion in heterologous cells, this study also proposes that the interaction between Mac-1 and SIRP␣ may facilitate cellcell recognition and adhesion during macrophage fusion. Because Mac-1 forms lateral complexes with CD47, it appears that these complexes rather than individual Mac-1 molecules bind SIRP␣ expressed on neighboring macrophages. Furthermore, discerning the role of the Mac-1-SIRP␣-CD47 complex may provide new insights into the mechanisms of macrophage fusion.

Cells
The HEK293 cells stably expressing human integrins Mac-1, ␣ L ␤ 2 , and the "I-less form" of Mac-1 were previously described (18,50,51). To prepare HEK293 cells expressing human SIRP␣, the full-length cDNA of human SIRP␣ was cloned from the Mammalian Gene Collection cDNA library (Thermo Fisher Scientific) into pAcGFP1-N1-vector (Clontech). HEK293 cells were stably transfected with the pAcGFP1-N1-SIRP␣ plasmid using Lipofectamine 2000 reagent (Invitrogen) according to the manufacturer's protocol. After 48 h at 37°C in 5% CO 2 , cells were harvested and cultured in medium with 500 g/ml G418 (Invitrogen). After 14 days, the surviving cells were collected, sorted, and analyzed using flow cytometry. The truncated SIRP␣ mutants with the first (SIRP␣2-3) and both first and second (SIRP␣3) domains deleted were produced by introducing restriction sites within the SIRP␣ sequence cloned into pAcGFP1-N1. The accuracy of the obtained constructs was verified by sequencing. The HEK293 cells expressing truncated SIRP␣ receptors were prepared as described above for the fulllength SIRP␣.
The mouse macrophage RAW264.7 cell line was obtained from the ATCC, and the cells were cultured in DMEM/F-12 SIRP␣ is a counter-receptor for integrin Mac-1 medium supplemented with 10% fetal bovine serum and antibiotics (0.1 mg/ml streptomycin and 0.1 unit/ml penicillin). Lentiviral particles SIRP␣ shRNA ((m) sc-36493) to knock down SIRP␣ expression in RAW264.7 cells were obtained from Santa Cruz Biotechnology. Viral particles contained three target-specific constructs that encode 19 -25 nucleotides (plus hairpin) of shRNA. Stable clones expressing shRNAs were selected using puromycin dihydrochloride (2.5 g/ml) for 14 days. Four clones were expanded and subsequently tested for SIRP␣ expression by FACS analysis. Thirty to 40% of the cells in each clone had SIRP␣ knocked down. Cells in one selected clone were sorted to obtain cells showing no SIRP␣ expression. As a control, shRNA lentiviral particles (sc-108080; Santa Cruz Biotechnology) containing an shRNA construct encoding a scrambled sequence were used.
Inflammatory macrophages were isolated 3 days after thioglycollate injection into the peritoneum of 8 -16-week-old WT C57BL/6 and Mac-1-deficient mice (The Jackson Laboratory, Bar Harbor, ME) by lavage using cold PBS containing 5 mM EDTA (9). All procedures were performed in accordance with the animal protocols approved by the Institutional Animal Care and Use Committee at the Arizona State University.

Synthesis of cellulose-bound peptide libraries
The SIRP␣-derived peptide library assembled on a single cellulose membrane support was prepared by parallel spot synthesis (52, 53). The membrane-bound peptides were tested for their ability to bind the ␣ M I-domain according to a previously described procedure (44). In brief, the membrane was blocked with 1% BSA and then incubated with 10 g/ml of 125 I-labeled ␣ M I-domain in TBS containing 1 mM MgCl 2 . After washing, the membrane was dried, and the ␣ M I-domain binding was visualized by autoradiography.

Biolayer interferometry
The binding parameters of the interaction between the ␣ M Idomain and SIRP␣ were determined using an Octet K2 instrument (FortéBIO, Pall Corp.). The purified extracellular domain of human SIRP␣ was immobilized on the Amine Reactive Second-generation (AR2G) biosensor using the amine coupling kit according to the manufacturer's protocol. Different concentrations of the active and inactive forms of ␣ M I-domain and active ␣ L I-domain were applied in the mobile phase, and the association between the immobilized and flowing proteins was detected. Experiments were performed in 20 mM HEPES, 150 mM NaCl, 1 mM MgCl 2 , 0.05% (v/v) Tween 20, pH 7.5. The SIRP␣-coated surface was regenerated with 25 mM NaOH. Analyses of the binding kinetics were performed using FortéBio Data Analysis 9.0 software. The value of the equilibrium dissociation constant (K D ) was obtained by fitting a plot of response at equilibrium (R eq ) against the concentration.

Adhesion assays
Adhesion assays were performed as described previously (18,50). Briefly, the wells of 96-well Immulon 4HBX polystyrene microtiter plates (Dynex Technologies, Chantilly, VA) were coated with mSIRP␣ or hSIRP␣ overnight at 4°C. The coated wells were post-coated with 1.0% PVP for 1 h at 22°C. Mac-1-expressing HEK293 cells and mouse macrophages were labeled with 10 M calcein (Molecular Probes, Eugene, OR) for 30 min and then washed twice. RAW264.7 cells were activated with 100 nM PMA for 30 min at 37°C at the time of labeling with calcein. Aliquots (0.1 ml) of labeled cells (5 ϫ 10 5 /ml of HEK293 cells, 10 6 /ml of RAW 264.7 cells, and 7.5 ϫ 10 5 /ml of mouse macrophages) in HBSS supplemented with 1 mM Ca 2ϩ , 1 mM Mg 2ϩ , and 0.1% BSA were added to each well. For inhibition experiments, cells were mixed with either antibodies or the mSIRP␣ recombinant proteins or peptides and incubated for 15 min at 22°C before they were added to the coated wells. After 30 min of incubation at 37°C, nonadherent cells were removed by two washes with PBS. Fluorescence was measured in a Cyto-FluorII fluorescence plate reader (PerSeptive Biosystems, Framingham, MA), and the number of adherent cells was determined by using the fluorescence of 100-l aliquots with a known number of labeled cells.

Cell fusion
RAW264.7-, Mac-1-, and SIRP␣-expressing HEK293 cells were washed and diluted at 10 5 -2.5 ϫ 10 5 /ml in DMEM/F-12 medium, and 0.5 ml of cell suspensions were loaded into the center of 6-cm Permanox dishes (Nalge Nunc International, Rochester, NY). After 30 min of incubation, 5 ml of OptiMEM medium (Invitrogen) containing antibiotics (0.1 mg/ml streptomycin and 0.1 unit/ml penicillin) was added, and the cells were incubated at 37°C in 5% CO 2 for 2 h. Following this, fusion was induced by adding 10 ng/ml IL-4. After incubation for 24 -72 h, the dishes were washed with PBS, and cells were fixed with 3.7% paraformaldehyde followed by staining with Wright stain or incubation with Alexa Fluor 546 -conjugated phalloidin. Images of representative fields were obtained using a Leica DM4000B (Leica Microsystems, Buffalo Grove, IL) microscope, and the numbers of nuclei in MGCs (Ն3) and mononuclear cells were counted. The extent of MGC formation was evaluated by determining the fusion index, which is defined as a fraction of nuclei within the MGCs expressed as a percentage of total nuclei counted (9). The number of visible nuclei per MGC was also counted. A total of 3-5 low-power fields (ϫ20) containing ϳ100 cells was analyzed for each experimental condition.

Immunoprecipitation
Cells (5 ϫ 10 6 ) were labeled with 100 g of Immunopure Sulfo-NHS-LC-Biotin (Pierce) in 200 l of PBS for 30 min at 22°C. Cells were solubilized with a lysis buffer (20 mM Tris-HCl, pH 7.4, 150 mM NaCl, 1% Triton X-100, 1 mM CaCl 2 , 1 mM phenylmethylsulfonyl fluoride, 100 g/ml leupeptin, 10 mM benzamidine) for 30 min at 22°C. The lysates were incubated with 10 g of normal mouse IgG (Sigma) and 50 l of Zysorbin-G (Zymed Laboratories Inc.) for 2 h at 4°C. After centrifugation, the supernatant was incubated with 10 g of mAb M1/70, 44a, or CD47 for 2 h at 4°C. The immune complexes were captured by incubating with 50 l of protein A-Sepharose (GenScript, Piscataway, NJ) for 16 h at 4°C. The immunoprecipitated proteins were eluted with SDS-PAGE loading buffer and analyzed by Western blotting. The Immobilon-P membranes (Millipore, New Bedford, MA) were incu-SIRP␣ is a counter-receptor for integrin Mac-1 bated with streptavidin conjugated to horseradish peroxidase and developed using SuperSignal West Pico Chemiluminescent Substrate (Thermo Fisher Scientific).

Flow cytometry
FACS analyses were performed to assess the expression of Mac-1 and SIRP␣ on the surface of transfected HEK293 cells. Cells were incubated with mAb 44a (anti-␣ M ) followed by Alexa Fluor 488 -conjugated secondary antibody or SE7C2 (anti-hSIRP␣) followed by Alexa Fluor 633-conjugated secondary antibody and analyzed using a FACScan (BD Biosciences). Populations of cells expressing similar amounts of Mac-1 and SIRP␣ were selected by FACS using a FACS Vantage instrument (BD Biosciences). To assess expression of Mac-1, CD47, and SIRP␣ on the surface of RAW264.7 cells, cells were harvested, and 10 6 cells were incubated in 100 l of 3% normal goat serum in 1% BSA/HBSS solution for 20 min at 4°C. After blocking, cells were incubated in 100 l of HBSS solution containing 10 g/ml of respective primary antibody (anti-␣ M mAb M1/70, anti-CD47 mAb miap301, and anti-SIRP␣ polyclonal H-300 antibody) for 30 min at 4°C. Cells were then washed and incubated with 5 g/ml Alexa Fluor 488 -conjugated secondary antibody for an additional 30 min at 4°C. Finally, cells were washed and analyzed as described above.

Immunofluorescence
Mouse macrophages and HEK293 cells expressing Mac-1 or SIRP␣ were incubated on glass slides or in Permanox chambers with 10 ng/ml IL-4 for 2-48 h and fixed in 3.7% paraformaldehyde for 20 min. Cells were incubated in a blocking buffer containing 3% normal goat serum and 1% BSA for 1 h at 22°C. Cells were then incubated with the primary anti-mouse SIRP␣ mAb P84 (10 g/ml) for 2 h at 22°C prior to exposure to secondary goat anti-rat antibodies conjugated to Alexa Fluor 488 (5 g/ml) or Alexa Fluor 633 (5 g/ml). The second primary antibody M1/70 conjugated with FITC (10 g/ml) was then applied in the same buffer. The rat IgG2b isotype control antibody conjugated with FITC was used as an isotype control for M1/70. Cells were mounted using Vectashield with DAPI. Confocal images were obtained using a Leica TCS SP5 AOBS Spectral Confocal Microscope (Exton, PA) housed in the WM Keck Bioimaging Facility at Arizona State University.

Time-lapse microscopy and image processing
Migration of SIRP␣-KD macrophages and RAW264.7 cells transduced with control lentivirus was visualized with the EVOS Live Cell imaging system (Thermo Fisher Scientific) equipped with the onstage incubator that enabled control of temperature, humidity, and carbon dioxide. Macrophages were applied on Permanox slides assembled in the 4-well chamber (Nunc, Rochester, NY), and IL-4 was added after 2 h of incubation. The chambers were transferred from the CO 2 incubator to an onstage incubator 2 h after addition of IL-4, and the recording continued for 24 h. Phase-contrast images were acquired every 30 s with the use of a ϫ10 objective (NA 0.25). To measure the speed of migration, 10 randomly selected cells from each cell line were tracked using ImageJ TrackMate software (National Institutes of Health, Bethesda, MD).

Statistical analyses
Data are presented as a mean Ϯ S.E. The statistical comparisons between treatment groups were made using one-way analysis of variance, followed by a post hoc test (Bonferroni's or Dunnett's) appropriate to the analyzed data set. Statistical analyses were performed using GraphPad Prism 5 software (La Jolla, CA). A difference of p Ͻ 0.05 was considered statistically significant.