Dual Regulation of SIRPα Phosphorylation by Integrins and CD47*

Signal regulatory protein α (SIRPα, SHPS-1) is a plasma membrane receptor for CD47 and a key regulator of phagocytosis, growth factor signaling, and migration. Phosphorylation of immunoreceptor tyrosine-based inhibition motifs in its cytoplasmic tail is essential for the functional effects of SIRPα, at least in part, because the phosphorylated immunoreceptor tyrosine-based inhibition motifs recruit Src homology 2 domain-containing tyrosine phosphatases. Ligation by CD47 and integrin engagement both have been thought to regulate SIRPα phosphorylation. However, their distinct contributions have not been distinguished. Here, we show that the importance of CD47 varies with cell type, since ligation of CD47 is not necessary for SIRPα phosphorylation in myeloid cells, whereas it is required in endothelial cells. In contrast, integrin-mediated adhesion is required for SIRPα phosphorylation in both cell types. This shows that SIRPα phosphorylation is dually regulated and demonstrates a new mechanism for functional cooperation between integrins and the integrin-associated protein CD47.

Cell adhesion is an important modulator of phenotype, as contact with extracellular matrix and with adjacent cells modulates responses to proliferative, apoptotic, migratory, phagocytic, and other fundamental cues. Often, these adhesion signals result from ligation of the heterodimeric integrin family of cell surface receptors (1,2). The integrin contribution to signaling frequently involves activation or amplification of tyrosine kinase cascades that may integrate with other signaling pathways (1,3,4). In some cell types, such as endothelial cells, signals from adhesion to extracellular matrix are tightly integrated with cell-cell adhesion signals mediated by members of the cadherin or immunoglobulin superfamilies (5,6).
Integrin-mediated adhesion has been shown to promote SIRP␣ phosphorylation (12,23,24), as has transcellular interaction with CD47 (16). One major caveat to these studies is that CD47 and integrins regulate each other's signaling (25)(26)(27). Since CD47 is ubiquitously expressed and was present on all cells used in the previous studies, it is not possible to distinguish the specific contribution of integrin signaling from integrin-CD47 cooperative signaling or even to determine whether integrin ligation simply led to SIRP␣-CD47 interactions in these previous reports. Indeed, other studies with antibodies and in mice and cells lacking CD47 have led to the hypothesis that interaction with CD47 is required for SIRP␣ phosphorylation and recruitment of SHP-1 and SHP-2 phosphatases (9,16).
In this work, we have used cells genetically deficient in CD47 to dissect the contributions of integrin ligation and interaction with CD47 for SIRP␣ phosphorylation in leukocytes and endothelial cells. We show that in leukocytes, integrin ligation is necessary and sufficient for SIRP␣ phosphorylation, which is substantial, even in the complete absence of CD47 expression. CD47 ligation of SIRP␣ amplifies the signal, but is not required. In endothelial cells, integrin-mediated adhesion also is necessary, but amplification by CD47-SIRP␣ interaction has a much more dramatic effect on SIRP␣ phosphorylation and phosphatase recruitment than in leukocytes. Indeed, in these cells, in contrast to leukocytes, SIRP␣ phosphorylation is almost undetectable without ligation by CD47. Thus, SIRP␣ phosphorylation requires integrin signaling in all cell types, whereas CD47 appears to be a costimulator of SIRP␣ phosphorylation whose importance depends upon cell type.

EXPERIMENTAL PROCEDURES
Reagents and Antibodies-Human fibronectin was obtained from Invitrogen, human vitronectin was from EMD Biosciences, Inc. (La Jolla, CA), collagen I ("Purecol") was from Inamed Biomaterials (Fremont, CA), and the poly(RGD) substrate was obtained from BIOSOURCE (Camarillo, CA) ("ProNectin F") or from Sigma ("fibronectin-like engineered protein polymer"). Fatty-acid free bovine serum albumin and gelatin were also obtained from Sigma.
Antibodies to CD47 (clones miap301 and miap420), the antibody to SIRP␣ (P84), and a control antibody (KLHR2A) were purified from hybridoma supernatants using conventional methods and have been described previously (28 -30). For integrin blocking experiments, anti-␤1 (clone Ha2/5), anti-␣v (clone H9.2B8), and appropriate isotype controls were purchased from BD Biosciences (San Jose, CA). Anti-␤2 (clone 5C6) was purchased from AbD Serotec (Raleigh, NC). Rat antibodies to intracellular cell adhesion molecule 2 (clone 3C4(IC2/ 4)), platelet-endothelial cell adhesion molecule (clones MEC13.3 and 390), and vascular endothelial cadherin (VE-cadherin) (clone 11D4.1) were purchased from BD Biosciences. Anti-platelet-endothelial cell adhesion molecule clone 390 also was the generous gift of S. M. Albelda (University of Pennsylvania Medical Center). Hybridomas for the antibodies MECA32 and endoglin were purchased from the Developmental Studies Hybridoma Bank, University of Iowa (Iowa City, IA). The antibody to JAM-A (clone BV11) was the kind gift of E. Dejana (Milan, Italy). Antibodies used for Western blotting included anti-phosphotyrosine (clone 4G10) and a rabbit polyclonal antibody to the cytoplasmic tail of SIRP␣, both obtained from Millipore (Billerica, MA), as well as mouse anti-SHP-2 (clone 79) purchased from BD Biosciences. Antibodies to ␣-, ␤-, and ␥-catenin were purchased from Zymed Laboratories Inc. (South San Francisco, CA), anti ␤-actin was from Cell Signaling Technologies (Danvers, MA), and anti-vinculin was from Sigma. Secondary antibodies used for blotting were obtained from Jackson ImmunoResearch Laboratories, Inc. (West Grove, PA) and Caltag Laboratories (Burlingame, CA). The appropriate anti-rat, anti-rabbit, and anti-mouse Alexa-488-and Alexa-594-coupled secondary antibodies along with Alexa-647-coupled phalloidin were purchased from Invitrogen for immunofluorescence microscopy experiments. The inhibitory cyclic peptide cyclo(Arg-Gly-Asp-D-Phe-Val) along with a control peptide, cyclo(Arg-Ala-Asp-D-Phe-Val), were purchased from Peptides International (Louisville, KY) and used at 100 M to assess ␣v integrin function in some experiments.
Plating onto Mouse CD47-Fc-The mouse CD47-Fc chimera was made using conventional methods by replacing the human IgV domain of CD47 in plasmid pIAP412 (31) with a PCR fragment containing the murine IgV domain (piap369). The resulting construct contains the human leader sequence and the first 13 amino acids of the mature human CD47 protein, the Cys at position 14 mutated to Ser, followed by 102 amino acids in the murine IgV domain and the hinge and constant regions of human IgG1 heavy chain. This soluble protein was purified with Gammabind G-Sepharose (GE Healthcare, Chalfont St. Giles, UK) from the supernatant of 293 cells transiently transfected using Lipofectamine 2000 (Invitrogen) according to the manufacturer's directions. The concentration of mouse CD47-Fc was determined using a Pierce BCA TM protein assay. Murine CD47-Fc bound to murine SIRP␣ on BMDM, as demonstrated by flow cytometry and shown previously by others (32).
Tissue culture dishes were incubated with poly-L-lysine (Sigma) and coated with mouse CD47-Fc or human IgG essentially as described (33), following which they were incubated with mouse lung endothelial cells (MLEC) in Iscove's modified Dulbecco's medium, 0.1% fatty acid-free bovine serum albumin for various times, and SIRP␣ phosphorylation was quantitated as described below.
Macrophages-Macrophages were derived from the bone marrow of CD47 ϩ/Ϫ and CD47 Ϫ/Ϫ mice (31) as well as mice lacking the SIRP␣ cytoplasmic domain (11) (SIRP CT Ϫ/Ϫ ) as previously described (34). All mice were bred for at least 10 generations onto the C57Bl/6 background. SIRP CT Ϫ/Ϫ bones were the kind gift of Mary Nakamura (University of California, San Francisco, CA). Macrophages were differentiated and maintained in macrophage complete medium (Dulbecco's modified medium, 10% fetal bovine serum (FBS, Hyclone, South Logan, UT), 10 mM HEPES with 10% conditioned medium from CMG cells (35) as a source of M-CSF) and grown on bacteriologic plastic (Kord-Valmark, Brampton, Ontario, Canada). For experiments to determine the effect of cell density on SIRP␣ phosphorylation, macrophages were plated at 3 ϫ 10 6 cells/10-cm dish ("high density"), 10 6 cells/10-cm dish ("medium density"), or 0.3 ϫ 10 6 cells/10-cm dish ("low density") in complete medium on bacteriologic plastic and allowed to grow for 3 days before doing the experiment. Macrophages were plated at 3 ϫ 10 6 cells/10-cm dish for all other experiments except those with anti-integrin antibodies, where they were plated at 2 ϫ 10 6 cells/dish. For some experiments, complete medium was removed, the cells were rinsed with PBS, and serum-free medium (Dulbecco's modified medium, 10 mM HEPES, 0.11 mg/ml sodium pyruvate, and 0.1% fatty acid-free bovine serum albumin) was added back for the indicated amount of time. For suspension and readhesion experiments, macrophages were harvested with 10 mM EDTA/PBS, washed once with serum-free or complete medium, and then resuspended and replated in the same medium. For experiments to examine integrins, macrophages were starved in serum-free medium overnight. As indicated, cells were harvested, pelleted, and incubated at 37°C for 1 h 30 min before adding the indicated antibodies at a final concentration of 20 g/ml, resuspending, and plating onto bacteriologic plastic or onto bacteriologic plastic coated with 10 g/ml FN. Cells were allowed to adhere for 1 or 2 h, as indicated, at 37°C before processing for immunoprecipitation. For these experiments, cells treated with anti-␤2 (5C6) did not spread on the bacteriologic dishes regardless of genotype, and cells treated with anti-␤1 (Ha2/5) did not attach to FN-coated dishes.
Bone Marrow Neutrophils-Neutrophils were isolated from the bone marrow of CD47 ϩ/Ϫ and CD47 Ϫ/Ϫ C57Bl/6 mice exactly as described in Ref. 36. The neutrophils were resuspended in HBSS containing 20 mM HEPES, 0.5% FBS, 0.5 mM CaCl 2 , 1 mM MgCl 2 , pH 7.4 ("HBSS ϩϩ ") to a concentration of 10 ϫ 10 6 cells/ml. To assess basal levels of SIRP␣ phosphorylation in nonadherent, nonactivated neutrophils, 1 ml of CD47 ϩ or CD47 Ϫ neutrophils (10 ϫ 10 6 cells) was rotated at 6 rpm at room temperature for 30 min before processing for immunoprecipitation as described below. For adhesion experiments, 1 ml of each genotype of neutrophils was added to poly(RGD)coated dishes and allowed to settle onto the surface for 10 min before the addition of 10 M formylmethionylleucylphenylalanine (Sigma) in HBSS ϩϩ , following which neutrophils were allowed to adhere for an additional 20 min prior to processing for immunoprecipitation. Without additional stimulation, the neutrophils remained round and nonadherent on this surface.
MLEC Line Isolation and Culture-An immortalizing transgene carried by the ImmortoMouse (Charles River Laboratories, Wilmington, MA) was bred into Sv129 CD47 ϩ/ϩ and CD47 Ϫ/Ϫ mice to facilitate establishment of endothelial cell lines. This immortalizing transgene consists of a temperaturesensitive SV40 large T antigen under the control of the major histocompatibility complex class I H-2K b promoter (38). Endothelial cells were isolated from the lungs of transgene-positive CD47 ϩ and CD47 Ϫ mice using a protocol generously provided by J. Lively and R. O. Hynes (Massachusetts Institute of Technology, Cambridge, MA) (39). Polyclonal cultures of CD47 ϩ and CD47 Ϫ MLEC were sorted for high and equivalent expression of intercellular cell adhesion molecule 2. Cell lines were maintained at 32.5°C in "Lively medium" composed of Dulbecco's modified medium/Ham's F-12 medium (Invitrogen), 20% heat-inactivated FBS, 50 g/ml endothelial mitogen (Biomedical Technologies, Stoughton, MA), 0.1 mg/ml heparin (Sigma), 20 units/ml interferon-␥ (R&D Systems, Minneapolis, MN), and 50 g/ml gentamicin (Sigma). Cells were passaged every 2-3 days using trypsin/EDTA and always plated onto tissue culture plastic that had been coated with 0.1% gelatin/PBS for at least 30 min at 37°C and rinsed once with PBS. High and consistent expression of endothelial markers and morphology is maintained for several months under these conditions. Expression of six antigens found predominantly on endothelial cells, including intercellular cell adhesion molecule 2, platelet-endothelial cell adhesion molecule 1, VE-cadherin, JAM-A, MECA32, and endoglin, was equivalent on CD47 ϩ and CD47 Ϫ MLEC (supplemental Fig. 1). Both CD47 ϩ and CD47 Ϫ MLEC have polygonal cell bodies and form cobblestone-like monolay-ers characteristic of endothelial cells (data not shown). In addition, MLEC of both genotypes were contact-inhibited and able to maintain confluent monolayers for several days (data not shown).
For all experiments, cells were plated at 0.3 ϫ 10 6 cells/10-cm tissue culture dish (low density), 10 6 cells/10-cm dish (medium density), or 3 ϫ 10 6 cells/10-cm dish (high density) in Lively medium. Two days after plating the cells, dishes were rinsed with PBS and Iscove's modified Dulbecco's medium (Invitrogen), 0.1% fatty acid-free bovine serum albumin (Sigma) (Assay Medium) without any FBS or antibiotics was added back to each. Cells were returned to 32.5°C and allowed to grow for an additional 48 h prior to the experiment. For some experiments, various antibodies, at a final concentration of 10 g/ml, were added to cells and incubated for 15 min at room temperature with occasional mixing by inversion prior to plating. Additional antibodies were added on each subsequent day to ensure that the concentration of antibody was maintained at 10 g/ml. In preparation for experiments to assess the rate of induction of SIRP␣ phosphorylation, cells were plated at low density. On the day of the experiment, cells were harvested with trypsin/EDTA, followed by neutralization with a serum-free trypsin inhibitor (Sigma), and pelleted (233 ϫ g, 5 min, room temperature) before resuspending in assay medium to the appropriate volume. Samples were rotated gently at 6 rpm with or without the indicated antibodies, cyclic peptides, 3 mM EDTA, or 1 mM MnCl 2 for 15 min at room temperature. Following the rotation, three dishes of cells initially plated at low density were replated into one dish and placed back into a 32.5°C tissue culture incubator for 2 h. For these experiments, tissue culture dishes were coated with 0.1% gelatin, collagen I at 20 g/ml, or vitronectin at 10 g/ml, as indicated. For experiments to examine the effects of cell-cell contact in the absence of adhesion to a substratum by pelleting the cells, MLEC were plated at low density, harvested, and combined as described above. Some cells were then replated at higher density, whereas others were pelleted (233 ϫ g, 5 min, room temperature). As indicated, some pelleted cells were lysed immediately, whereas others were incubated for an additional 60 min at 32.5°C in the presence or absence of EDTA or MnCl 2 before lysis and processing for immunoprecipitation. For experiments in which CD47 Ϫ MLEC were mixed with SIRP CT Ϫ/Ϫ BMDM, CD47 Ϫ MLEC were initially plated at low density and subsequently incubated with SIRP CT Ϫ/Ϫ BMDM to achieve a total cell density equal to the MLEC plated at high density, as described above. As controls, MLEC of both genotypes were harvested and replated at high density, as described above. Cells were allowed to readhere for 2 h before lysis and processing for immunoprecipitation. To examine components of the adherens junction (AJ), cells were plated at the indicated densities. Twenty minutes prior to lysis, cells were treated with 0.5 M pervanadate to preserve phosphorylation levels of VE-cadherin and the ␣-, ␤-, and ␥catenins (40).
Flow Cytometry-The expression of a variety of cell surface molecules on CD47 ϩ and CD47 Ϫ MLEC, BMDM, or polymorphonuclear leukocytes was determined using a Coulter Epics XL flow cytometer (Coulter Corp., Hialeah, FL). For each marker assessed, 2 ϫ 10 5 MLEC or 1 ϫ 10 5 BMDM were incu-bated with primary antibody at a final concentration of 10 g/ml (for purified antibodies) or in hybridoma supernatant for 30 -60 min on ice, washed, and analyzed. For BMDM, 1 mg/ml human IgG was included in the incubation buffer to block Fc␥R binding of the monoclonal antibodies. Incubation with a rat monoclonal antibody anti-keyhole limpet hemocyanin, KLHR2A, was used to assess nonspecific binding.
Binding of mouse CD47-Fc to BMDM was assessed by flow cytometry as described above. 20 g/ml anti-Fc␥RII/III 2.4G2 was used to inhibit Fc␥R binding of the construct, and nonspecific binding was assessed using human IgG rather than CD47-Fc.
Immunoprecipitations-For all immunoprecipitations, antibodies (5-10 g/sample) were precoupled to resin (20 -30 l/sample of 50% slurry) by incubating with Gammabind G-Sepharose (GE Healthcare) overnight at 4°C with rotation in 20 mM HEPES, pH 7.4, 150 mM NaCl, followed by cross-linking with 10 mM dimethyl pimelimidate (Pierce). For all endothelial cell experiments, MLEC were transferred to medium lacking serum and added growth factors for 48 h prior to lysis. Cells were solubilized in ice-cold lysis buffer (20 mM HEPES, pH 7.4, 150 mM NaCl, 1 mM EGTA, 1% Nonidet P-40 (Pierce), 1 mM NaF, 10 g/ml leupeptin (Roche Applied Science), 10 g/ml aprotinin (Roche Applied Science), 2 mM diisopropylfluorophosphate (Sigma) and 0.5 M pervanadate). After removing nuclei and debris by centrifugation, the protein concentration of each sample was assessed using a Pierce BCA TM protein assay. For each experiment, equal amounts of protein in each lysate were used for each immunoprecipitation. After incubating at 4°C with rotation overnight, the immunoprecipitates were washed three times with 0.75 ml of wash buffer (20 mM HEPES, pH 7.4, 150 mM NaCl, 1 mM EGTA, 0.1% Nonidet P-40, 1 mM NaF, 10 g/ml leupeptin, 10 g/ml aprotinin, and 0.5 M pervanadate) before adding reducing sample buffer and boiling for 5 min.
SDS-PAGE and Western blotting were performed according to standard procedures. Western blot band densities were quantified using CCD detection of emitted light during blot development with chemiluminescent reagents (Pierce) using the Fluorochem8000 (Alpha Innotech, San Leandro, CA). The extent of SIRP␣ tyrosine phosphorylation in each sample was determined by dividing the intensity of the SIRP␣ phosphotyrosine signal by the intensity of the band detected with an antibody recognizing SIRP␣ in a separate aliquot of the same sample. Values were expressed as percentages of the indicated reference sample. Graphs of relative phosphotyrosine content represent at least three independent experiments. One-sample t tests were used for statistical analysis where indicated.
Microscopy-To visualize CD47 and SIRP␣ expressed on the surface of MLEC, cells were plated onto gelatin-coated coverslips and then stained with a control antibody (KLHRA2), anti-CD47 (miap301), or anti-SIRP␣ (P84) before fixation with 3.7% paraformaldehyde. Following incubation with an Alexa-488coupled anti-rat secondary antibody, cells were visualized with a Zeiss Axiovert 100TV microscope. VAYTEK Microtome software was used to collect images in the z-plane, and VAYTEK Haze Buster software was used for deconvolution. For these experiments, fixation occurred after incubation with the pri-mary antibodies, since miap301 and P84 did not recognize their respective antigens when MLEC were fixed.
Confocal microscopy images to visualize SIRP␣ in conjunction with the cytoskeleton were obtained using a Zeiss LSM510 microscope. BMDM and MLEC were plated onto FN-coated coverslips. Prior to staining, cells were fixed with 3.7% paraformaldehyde and permeabilized with 1% Triton X-100, 1% bovine serum albumin, PBS. Primary antibodies used for staining included mouse monoclonal antibody to vinculin and rabbit polyclonal antibody to the SIRP␣ cytoplasmic tail. After incubation with Alexa-594-coupled anti-mouse and Alexa-488-coupled anti-rabbit secondary antibodies along with Alexa-647-coupled phalloidin, coverslips were mounted in Prolong (Invitrogen) and inspected.

Macrophage SIRP␣ Phosphorylation Requires Adhesion and
Is Enhanced by CD47-Previous work has demonstrated that macrophage SIRP␣ phosphorylation increases upon exposure to CD47 ϩ erythrocytes of the same species, inhibiting phagocytosis of these red blood cells (16). However, regulation of basal SIRP␣ phosphorylation by CD47 was not investigated. CD47 ϩ and CD47 Ϫ BMDM express equal amounts of SIRP␣ (supplemental Fig. 2), and there is substantial SIRP␣ phosphorylation in adherent CD47 Ϫ BMDM, whether or not they are in medium containing M-CSF and serum (Fig. 1A). SIRP␣ phosphorylation was greater in CD47 ϩ BMDM; the extent of SIRP␣ phosphorylation in confluent CD47 Ϫ BMDM cultured in complete medium was 61 Ϯ 9% of that in confluent CD47 ϩ BMDM (quantitated as described under "Experimental Procedures") ( Fig. 1A). In the absence of M-CSF and serum, SIRP␣ phosphorylation in CD47 ϩ decreased by about 15%, whereas it remained virtually unchanged in CD47 Ϫ BMDM, suggesting that CD47 expression enhances the effects of M-CSF and/or serum-derived growth factors that lead to SIRP␣ phosphorylation. When BMDM were put into suspension for 1 h, SIRP␣ phosphorylation decreased dramatically, regardless of whether the macrophages expressed CD47 or were exposed to M-CSF and serum (Fig. 1A). After 1 h in suspension, the amount of phosphorylated SIRP␣ in both CD47 ϩ and CD47 Ϫ BMDM decreased more than 80% compared with adherent cells in both complete and serum-free medium. These data indicated that adhesion was a major stimulus for SIRP␣ phosphorylation. CD47 expression enhanced adhesion-dependent phosphorylation by about 1.6-fold in macrophages in complete medium and by about 1.3-fold in the absence of serum or additional growth factors. The addition of M-CSF and other growth factors had only a small effect on the extent of SIRP␣ phosphorylation in either adherent or nonadherent BMDM.
Unlike BMDM, neutrophils are not spontaneously adherent, but they do express high levels of SIRP␣. In both CD47 ϩ and CD47 Ϫ bone marrow-derived neutrophils, SIRP␣ phosphorylation in suspended cells was not detectable (Fig. 1B). Significant SIRP␣ phosphorylation was induced when formylmethionylleucylphenylalanine-stimulated neutrophils were allowed to adhere to a surface coated with the synthetic integrin ligand poly(RGD) (Fig. 1B). As in macrophages, SIRP␣ phosphorylation did not require, but was enhanced by, CD47 expression.
To examine the role of CD47 further, CD47 ϩ and CD47 Ϫ BMDM were plated at different densities. Cells plated at low density grew as single cells without touching other cells. Cells plated at medium density were ϳ50% confluent. Cells at high density appeared as a confluent layer of macrophages with all cells in contact with their neighbors. As shown in Fig. 2A, significant levels of phospho-SIRP␣ were found in both CD47 ϩ and CD47 Ϫ macrophages at all three cell densities. SIRP␣ phosphorylation was enhanced with increasing cell density in CD47 ϩ but not CD47 Ϫ cells, suggesting that CD47 interactions with SIRP␣ on adjacent cells increased SIRP␣ phosphorylation, as suggested by previous work with macrophage-erythrocyte interactions (16,17). However, cell-cell contact alone was insufficient to induce SIRP␣ phosphorylation, because when either CD47 ϩ or CD47 Ϫ BMDM were centrifuged and incubated in a pellet for 1.5 h, no increase in SIRP␣ phosphorylation was observed (Fig. 2B). In fact, SIRP␣ phosphorylation decreased by about 50% in each cell type without adhesion, despite the close cell-cell contact induced by centrifugation. Together, these data show that BMDM SIRP␣ phosphorylation requires adhesion and can be enhanced by CD47, probably because of intercellular interactions between CD47 and SIRP␣.
To determine the kinetics of SIRP␣ phosphorylation in response to adhesion, CD47 ϩ and CD47 Ϫ macrophages were placed into suspension for 1 h and replated, and SIRP␣ phosphorylation was quantified at various times. At 30 min after replating, most cells were round and loosely tethered to the dish, and very few had spread; by 60 min, cells were more tightly tethered and had begun to spread. These kinetics of adhesion and spreading correlated with SIRP␣ phosphorylation, with detectably increased SIRP␣ phosphorylation by 30 min and maximal phosphorylation in both CD47 ϩ and CD47 Ϫ BMDM by 60 min (Fig. 3A). CD47 expression did not affect the kinetics of SIRP␣ phosphorylation in response to adhesion.
To characterize the role of integrins in the induction of SIRP␣ phosphorylation in BMDM, CD47 ϩ and CD47 Ϫ BMDM from serum-starved cultures were harvested and incubated in a pellet for 1.5 h before replating onto bacteriologic plastic for 2 h in the presence of antibodies to several different integrins. Adhesion and spreading on bacteriologic plastic is known to require␤2integrins (41).Anti-␤2antibodiesinhibitedadhesiondependent induction of SIRP␣ phosphorylation in both CD47 ϩ and CD47 Ϫ BMDM (Fig. 3B). Antibodies to the ␣v and ␤1 inte- incubated either with MCSF-and serum-containing medium (CM) or medium lacking both MCSF and serum (serum-free medium). Wild type macrophages (CD47 ϩ ) are shown in gray; CD47-deficient macrophages (CD47 Ϫ ) are shown in the checkered columns. The extent of phosphorylation was determined by quantitating the signal from anti-phosphotyrosine (pY) and anti-SIRP␣ Western blots (WB) of each sample, as described under "Experimental Procedures." SIRP␣ phosphorylation in experimental samples was compared with that in adherent CD47 ϩ macrophages, which was set at 100%. Data were obtained from 3-8 independent experiments. *, p Ͻ 0.05. Bottom, representative Western blots from SIRP␣ immunoprecipitations (IP). The position of a 116-kDa molecular mass marker is indicated. B, SIRP␣ phosphorylation was determined in suspended and adherent CD47 ϩ and CD47 Ϫ neutrophils. The position of a 116-kDa marker is indicated. grins, which are also well expressed on BMDM, had no effect on their own (not shown) but slightly lowered the amount of phospho-SIRP␣ further when combined with anti-␤2, although these differences were not significantly different from treatment with anti-␤2 alone (Fig. 3B). As expected, neither CD47 ϩ nor CD47 Ϫ BMDM spread on bacteriologic plastic in the presence of anti-␤2. Plating CD47 ϩ and CD47 Ϫ BMDM onto FN also led to adhesion-dependent induction of SIRP␣ phosphorylation in both cell types (Fig. 3C). Induction of SIRP␣ phosphorylation on FN was inhibited in both genotypes when BMDM were plated in the presence of antibodies to the ␤1 integrin, demonstrating the importance of integrin-mediated adhesion independent of the expression of CD47. The addition of anti-␤2 or cyclic RGD peptides did not have any additional inhibitory effect when combined with anti-␤1, probably because they did not have any further inhibitory effects on adhesion to the FN-coated surface under the conditions used (data not shown). Thus, induction of SIRP␣ phosphorylation in BMDM requires integrin-mediated adhesion but not CD47.
Endothelial SIRP␣ Phosphorylation Requires CD47-Although SIRP␣ expression and function has been studied primarily in myeloid cells and neurons, it also is expressed in endothelial cells. CD47 ϩ and CD47 Ϫ murine lung endothelial cell lines expressed similar amounts of SIRP␣ on their surfaces (supplemental Fig. 3). Although both CD47 ϩ and CD47 Ϫ MLEC grew as adherent cell lines, phosphorylated SIRP␣ was detected only in the CD47 ϩ MLEC (Fig. 4A). In contrast to BMDM, where CD47 expression increased SIRP␣ phosphorylation by less than 2-fold (Fig. 1A), CD47 expression increased SIRP␣ phosphorylation by more than 30-fold. SHP-2, a cytosolic phosphatase known to be recruited in a phosphorylation-dependent manner to the cytoplasmic ITIM motifs in the cytoplasmic domain of SIRP␣ (12), was detected in anti-SIRP␣ immunoprecipitates only from CD47 ϩ lysates (Fig. 4A). SHP-1, also known to interact with the SIRP␣ ITIMs, was not detected in MLEC (data not shown). To determine whether CD47-SIRP␣ interaction was required for MLEC SIRP␣ phosphorylation, CD47 ϩ and CD47 Ϫ MLEC were grown in the presence of antibodies that block CD47-SIRP␣ interaction (30). Four days after plating, the amount of phosphorylated SIRP␣ was assessed. As shown in Fig. 4B, the anti-CD47 antibodies miap301 and miap420 significantly inhibited SIRP␣ phos-FIGURE 3. BMDM SIRP␣ phosphorylation requires integrin-mediated adhesion. A, top, SIRP␣ phosphorylation was quantitated in CD47 ϩ (solid bars) and CD47 Ϫ (checkered bars) BMDM continuously adherent (Adh), after 1 h in suspension (Sus, 60Ј), and after various times of readhesion. Phosphorylation in continuously adherent CD47 ϩ BMDM was set to 100%. Data were obtained from 3-5 independent experiments (*, p Ͻ 0.05; **, p Ͻ 0.001). Bottom, representative Western blots (WB) from SIRP␣ immunoprecipitations (IP). The position of a 116-kDa molecular mass marker is indicated. B, SIRP␣ phosphorylation was quantitated in CD47 ϩ (solid bars) and CD47 Ϫ (checkered bars) BMDM that were continuously adherent (Adh), pelleted to maintain cell-cell contact (pellet), or removed from the pellet and allowed to readhere to bacteriologic plastic for 2 h with antibodies to the indicated integrin chains or control antibody. Statistical comparisons were made between BMDM allowed to readhere with control antibodies or anti-integrin antibodies for each genotype (*, p Ͻ 0.05; #, p Ͻ 0.01). C, SIRP␣ phosphorylation was quantitated in CD47 ϩ (solid bars) and CD47 Ϫ (checkered bars) BMDM that were continuously adherent (Adh), pelleted to maintain cell-cell contact (pellet), or removed from the pellet and allowed to readhere to bacteriologic plastic (bact. plastic) or to FN-coated plastic for 1 h with a control or anti-␤1 integrin antibody. Statistical comparisons were made between BMDM allowed to readhere with control antibody or anti-␤1 for each genotype (*, p Ͻ 0.05). For B and C, data shown are the summary of 3-6 independent experiments. pY, phosphotyrosine.
To examine this requirement further, CD47 ϩ and CD47 Ϫ MLEC were plated at three different densities. For these exper-iments, cells plated at low density were just reaching confluence on the day of the experiment, cells plated at medium density reached confluence 1-2 days prior to the day of the experiment, and cells plated at high density were confluent for 3-4 days prior to the day of the experiment. Identical amounts of lysate protein were used for each immunoprecipitation. Although plating density did not alter the surface expression levels of SIRP␣ (data not shown), it had a significant effect on the extent of phosphorylated SIRP␣ detected in the CD47 ϩ cells (Fig. 4C). Phosphorylated SIRP␣ was not detected in CD47 Ϫ MLEC at low or medium density and was minimally detectable in high density CD47 Ϫ MLEC (Fig.  4C). Increasing cell density also increased the extent of SIRP␣ phosphorylation in the brain-derived CD47 ϩ endothelial cell line bEND.3 (Fig. 4D). Because increased phosphorylation occurred with increased cell density, it is likely that these CD47-SIRP␣ interactions occurred at intercellular contacts.
To determine whether intercellular CD47-SIRP␣ interactions were required for SIRP␣ phosphorylation in adherent MLEC, CD47 Ϫ MLEC were cultured for 2 h with BMDM from CD47 ϩ , SIRP CT Ϫ/Ϫ macrophages. Because these BMDM express SIRP␣ lacking its cytoplasmic tyrosine phosphorylation sites, SIRP␣ immunoprecipitated from these cells does not contribute to the phosphorylation signal. CD47 Ϫ MLEC cultured at either low or high density had barely detectable phospho-SIRP␣, whereas there was density-dependent SIRP␣ phosphorylation in CD47 ϩ MLEC (Fig. 4E). When CD47 Ϫ MLEC were coincubated with SIRP CT Ϫ/Ϫ BMDM, the extent of SIRP␣ phosphorylation was equivalent to that in CD47ϩ MLEC (Fig. 4E). Since CD47 was expressed only on BMDM and SIRP␣ could only be phosphorylated on MLEC, this experiment demonstrates that intercellular interactions between CD47 and SIRP␣ induce SIRP␣ phosphorylation in adherent endothelial cells. Consistent with a primary role for intercellular interaction in MLEC SIRP␣ This figure is representative of more than five independent experiments. B, CD47 ϩ and CD47 Ϫ MLEC were plated in the presence of antibodies (10 g/ml) to a control antigen (ctrl) or to CD47 (301 and 420), following which SIRP␣ was immunoprecipitated and phosphorylation was detected by Western blot. C, MLEC were plated at increasing cell density as described under "Experimental Procedures," following which SIRP␣ was immunoprecipitated, and phosphorylation was detected by Western blot. B and C are representative of at least three independent experiments. D, SIRP␣ phosphorylation was quantitated as described under "Experimental Procedures" in the CD47 ϩ , brain-derived endothelial cell line bEND.3 as a function of increasing density. Data shown are the summary of three independent experiments. E, CD47 ϩ and CD47 Ϫ MLEC were plated at low density. Some of each were left adherent (Low Dens), whereas other cells were harvested and replated on gelatin-coated tissue culture plastic at high cell density, either three low density dishes combined into one (Replate High) or with sufficient SIRP CT Ϫ/Ϫ BMDM to achieve the same total cell density as in those replated at high density. SIRP␣ phosphorylation was quantitated as described. Data shown are the summary of three independent experiments. *, p Ͻ 0.05; **, p Ͻ 0.005; ***, p Ͻ 0.001. Inset, lysate samples from CD47 Ϫ , CD47 ϩ , and SIRP CT macrophages blotted with antibody recognizing the cytoplasmic tail of SIRP␣ and with ␤-actin as a loading control. pY, phosphotyrosine. phosphorylation, both CD47 and SIRP␣ concentrated at cellcell junctions in nonpermeabilized MLEC (supplemental Fig.  4). Thus, intercellular SIRP␣-CD47 interactions have a much more significant role in regulation of SIRP␣ phosphorylation in MLEC than in BMDM or neutrophils.
Kinetics of SIRP␣ Phosphorylation in MLEC-Because the requirements for CD47 in SIRP␣ phosphorylation in MLEC and leukocytes were distinct, we examined the kinetics of SIRP␣ phosphorylation in MLEC. MLEC plated at low density, which show minimal SIRP␣ phosphorylation, were harvested and then replated at higher density for various times. When replated onto surfaces coated with fibronectin and gelatin or gelatin alone (data not shown), robust SIRP␣ phosphorylation was detected by 60 min in CD47 ϩ MLEC (Fig. 5A). At this time, about 90% of both the CD47 ϩ and CD47 Ϫ MLEC had attached and spread (data not shown). This rapid density-dependent augmentation of SIRP␣ phosphorylation was inhibited by antibody to CD47 both for MLEC (Fig. 5B) and bEND.3 cells (Fig.  5C). There was a less robust increase in SIRP␣ phosphorylation in bEND.3 cells than MLEC after 1 h at high density (Fig. 5C), probably because bEND.3 did not attach or spread as quickly as MLEC.
Adhesion to a CD47-coated Surface Is Not Sufficient for Endothelial SIRP␣ Phosphorylation-Since SIRP␣ phosphorylation in endothelial cells, unlike BMDM or polymorphonuclear leukocytes, required CD47, we tested whether MLEC adhesion to CD47 was sufficient to induce this signaling. MLEC, grown at low density to minimize SIRP␣ phosphorylation, were replated at higher density onto tissue culture dishes coated with mouse CD47-Fc or human IgG as a control (Fig. 6A). Murine CD47-Fc is a soluble ligand competent to bind cell-expressed SIRP␣, as shown previously by others (32). At no time did surface-bound mouse CD47-Fc induce SIRP␣ phosphorylation in CD47-MLEC (Fig. 6A). Increasing SIRP␣ phosphorylation occurred over time in CD47 ϩ MLEC but was equivalent in cells plated onto mouse CD47-Fc and nonspecific human IgG. Over this extended time course, the MLEC adhered and spread on the protein-coated surface. Since SIRP␣ phosphorylation occurred only in CD47 ϩ MLEC and was equivalent on mouse CD47-Fc and control surfaces, these data suggest that adhesion to the tissue culture plate together with intercellular interactions was responsible for SIRP␣ phosphorylation rather than direct ligation of MLEC SIRP␣ by the surface-bound CD47-Fc. As a further test for the sufficiency of CD47-SIRP␣ interaction to induce SIRP␣ phosphorylation, CD47 ϩ MLEC grown at low density were suspended and then pelleted by centrifugation to maximize cell-cell contact (Fig. 6B). Without cell adhesion to a substratum, SIRP␣ phosphorylation was not activated in the pelleted cells. Equivalent data were obtained with bEND.3 cells (Fig. 6C). Together, these experiments demonstrate that CD47-SIRP␣ interaction is not sufficient to induce SIRP␣ phosphorylation.
Integrin Ligation Is Required for Induction of SIRP␣ Phosphorylation in MLEC-Several experiments suggested a role for integrins in the induction of MLEC SIRP␣ phosphorylation. The addition of EDTA, which blocks integrin recognition of ligands, prevented induction of SIRP␣ phosphorylation in MLEC (Fig. 5B). Plating MLEC, even at high density, on sur-faces lacking integrin ligands did not induce SIRP␣ phosphorylation (Fig. 6, A and B), and pelleting MLEC in a buffer containing Mn 2ϩ , which increases integrin affinity, enhanced SIRP␣ phosphorylation compared with cells in buffer lacking divalent cations (Fig. 6B).
To test formally the role for integrins in induction of SIRP␣ phosphorylation in CD47ϩ MLEC, MLEC were plated onto various surfaces. First, cells grown at low density to minimize SIRP␣ phosphorylation were replated on gelatin in the presence of antibodies to ␤1 integrins with and without ␣v integrin antibodies. Antibodies to ␤1 integrins completely inhibited the increased SIRP␣ phosphorylation resulting from replating A, CD47 ϩ and CD47 Ϫ MLEC were plated at low density, harvested, and replated on FN/gelatin-coated tissue culture plastic in the absence of serum or growth factors for the indicated periods of time. SIRP␣ was immunoprecipitated (IP), and its phosphorylation was detected by Western blot (WB). The position of a 116-kDa molecular mass marker is indicated. Results are representative of at least three independent experiments. B, CD47 ϩ MLEC were grown at low density and then replated at high density for 2 h in the presence of control antibody, anti-CD47 antibody miap420, or 3 mM EDTA. SIRP␣ phosphorylation was quantitated as described under "Experimental Procedures." Data are the summary of three experiments (*, p Ͻ 0.05 compared with the low density cells). C, CD47 ϩ bEND.3 cells were grown and replated as described in B, except that the anti-CD47 antibody used was miap301. SIRP␣ phosphorylation was quantitated as previously described (n ϭ 3). pY, phosphotyrosine.
CD47 ϩ MLEC at high density (Fig. 7A). On their own, antibodies to the ␣v integrin slightly decreased the amount of induced SIRP␣ phosphorylation (data not shown) but did not augment the effect of the anti-␤1 antibodies when combined with anti-␤1. To test the requirement for integrins further, CD47 ϩ FIGURE 6. CD47 ligation is not sufficient to induce SIRP␣ phosphorylation in endothelial cells. A, CD47 ϩ and CD47 Ϫ MLEC were plated onto mouse CD47-Fc-or human IgG-coated surfaces, and SIRP␣ phosphorylation was detected as described in previous figure legends. B, CD47 ϩ MLEC plated at low density were left adherent or harvested for replating at high density on FN or uncoated tissue culture plastic or were pelleted after deadhesion. Some MLEC were pelleted in the presence of 3 mM EDTA or 1 mM MnCl 2 . SIRP␣ phosphorylation was quantitated as previously described, and phosphorylation in MLEC kept at low density was set at 100%. Statistical comparisons were made between other groups and MLEC at low density (*, p Ͻ 0.05). Data were from three independent experiments. C, CD47 ϩ bEND.3 plated at low density were left adherent or harvested and pelleted as described above for MLEC. SIRP␣ phosphorylation was quantitated, and the phosphorylation of low density bEND.3 was set at 100%. Statistical comparisons of SIRP␣ phosphorylation in adherent cultures and pelleted cells were performed (**, p Ͻ 0.01). Data were from three independent experiments. FN/gel, fibronectin/gelatincoated tissue culture plastic; TC, uncoated tissue culture plastic; pellet 0, pelleted cells lysed immediately after centrifugation; pellet 60, pelleted cells lysed after 60 min; pellet ϩ EDTA, 60, cells pelleted and maintained in the presence of 3 mM EDTA for 60 min before lysis; pellet ϩ MnCl 2 , 60, cells pelleted and maintained in the presence of 1 mM MnCl 2 for 60 min before lysis. IP, immunoprecipitation; WB, Western blot; pY, phosphotyrosine. FIGURE 7. Endothelial SIRP␣ phosphorylation requires integrin-mediated adhesion and may influence phosphorylation of adherens junction components. A, CD47 ϩ MLEC plated at low density were left adherent (Low dens) or harvested and replated at high density for 2 h on gelatin-coated tissue culture plastic in the presence or absence of 10 g/ml antibodies to the integrins ␤1 and/or ␣v. SIRP␣ phosphorylation was quantitated as previously described, and phosphorylation in the initial cell cultures was set to 100%. Data were from three independent experiments. A t test was used to compare replating in the presence of antibodies to replating under control conditions (**, p Ͻ 0.01). B, as in A, CD47 ϩ MLEC plated at low density were left adherent (Low dens) or replated for 2 h at high density on gelatin-coated tissue culture plastic (Gel), collagen I (20 g/ml) (COL I), or vitronectin (10 g/ml) (VN). Cells were plated on collagen I in the presence of control or anti-␤1 integrin antibodies (20 g/ml) or on VN in the presence of control peptide, cRADfV, or an inhibitory peptide, cRGDfV, at 100 M. SIRP␣ phosphorylation was quantitated as previously described, and phosphorylation in the initial cell cultures was set to 100%. Data were from three independent experiments. A t test was used to compare replating in the presence of antibodies with replating under control conditions (*, p Ͻ 0.05). C, VE-cadherin was immunoprecipitated from CD47 ϩ and CD47 Ϫ MLEC that had been plated at low, medium, and high density. Phosphorylation of VE-cadherin and co-immunoprecipitated ␣-, ␤-, and ␥-catenin was detected by Western blot. Results are representative of at least five independent experiments. IP, immunoprecipitation; WB, Western blot; pY, phosphotyrosine.
MLEC were replated at high density onto collagen I and vitronectin, two well defined integrin ligands. Replating CD47 ϩ MLEC at higher density onto these defined integrin ligands also induced SIRP␣ phosphorylation (Fig. 7B). SIRP␣ phosphorylation was inhibited by antibody to ␤1 integrin when cells were plated on collagen and by a cyclic RGD peptide when cells were plated on vitronectin but not by control antibodies or peptides (Fig. 7B). We conclude that integrin ligation and CD47-SIRP␣ interactions are both required for induction of SIRP␣ phosphorylation in MLEC and that both ␣v and ␤1 integrins can provide the required adhesive signal.
Adherens Junction Components Are More Highly Phosphorylated in CD47 Ϫ MLEC-Since CD47-SIRP␣ interactions appeared to occur at cell-cell junctions in MLEC and also led to SHP-2 recruitment, we tested whether tyrosine phosphorylation of other junctional components was altered in CD47-deficient MLEC. Indeed, tyrosine phosphorylation of VE-cadherin and associated ␣-, ␤-, and ␥-catenins increased with increasing cell density in CD47 Ϫ MLEC but not in CD47 ϩ cells (Fig. 7C). Thus, it is possible that CD47-SIRP␣ interactions at cell-cell borders in endothelial cells contrib-ute to the recruitment of tyrosine phosphatase activity required to regulate endothelial adherens junctions. Although phosphorylation levels of these junctional components were different, CD47 ϩ and CD47 Ϫ MLEC monolayers appear virtually identical to one another with no apparent differences in morphology or monolayer integrity (data not shown) under basal conditions. SIRP␣ Localization Differs in BMDM and MLEC-Because the dependence on CD47 for induction of SIRP␣ phosphorylation differed between BMDM and MLEC, we examined SIRP␣ localization in the two cell types. The predominant location for both CD47 and SIRP␣ in intact CD47 ϩ MLEC was at cellcell junctions, and SIRP␣ localized to cell-cell junctions less well in CD47 Ϫ MLEC (supplemental Fig.  4). In contrast, SIRP␣ significantly localized to the adhesive surface of BMDM and was prominent at the membrane periphery in both CD47 ϩ and CD47 Ϫ cells ( Fig. 8 and supplemental Fig. 5). To investigate whether SIRP␣ localization correlated with known cytoskeletal structures, CD47 ϩ and CD47 Ϫ BMDM and MLEC were stained for vinculin, a marker of focal adhesions, focal complexes, and podosomes, and for polymerized actin structures. As determined by confocal microscopy, SIRP␣ showed considerable colocalization with polymerized actin and vinculin in both CD47 ϩ and CD47 Ϫ BMDM ( Fig. 8 and supplemental Fig. 5). The colocalization of SIRP␣, vinculin, and phalloidin typically appeared strongest at the periphery of both CD47 ϩ and CD47 Ϫ BMDM, coinciding with the greatest concentrations of SIRP␣, although more internal adhesive structures also showed some SIRP␣ colocalization (Fig. 8). Phalloidin staining showed a similar pattern of colocalization with SIRP␣, indicating a close association between SIRP␣ and the actin cytoskeleton in these cells. In contrast, SIRP␣ staining showed little colocalization with either vinculin or actin in permeabilized CD47 ϩ and CD47 Ϫ MLEC ( Fig. 8 and supplemental Fig. 5). There was no colocalization of SIRP␣ staining with focal adhesions demonstrated by vinculin staining or with phalloidin staining (Fig. 8). These data demonstrate that there is more SIRP␣ association with sites of substrate adhesion in BMDM than in MLEC. This association with sites of integrin-mediated adhesion may account for SIRP␣ phosphorylation independent of CD47 in this cell type.

DISCUSSION
SIRP␣ phosphorylation is a critical regulatory step in a variety of signaling pathways and cell types, including macrophages, neutrophils, dendritic cells, smooth muscle cells, and fibroblasts. Although growth factors, cell adhesion, and CD47 ligation are known to modulate SIRP␣ phosphorylation in fibroblasts, macrophages, and smooth muscle cells (7,9,12,14,16,23), all of these signals interact, and the specific requirements for SIRP␣ phosphorylation have not been carefully dissected. In this work, a genetic approach was taken in three different cell types to isolate the effects of cell adhesion and intercellular interactions in regulating SIRP␣ phosphorylation. In macrophages and neutrophils, SIRP␣ phosphorylation was highly dependent on integrin-mediated adhesion. Expression of CD47 enhanced this phosphorylation but clearly was not required, since abundant phospho-SIRP␣ was detected in adherent CD47 Ϫ/Ϫ BMDM and neutrophils. In contrast, basal SIRP␣ phosphorylation in endothelial cells required both expression of CD47 and integrin-mediated adhesion. Increased phosphorylation of SIRP␣ was detected in CD47 ϩ MLEC and bEND.3 cells plated at increasing cell density, consistent with an intercellular interaction between CD47 and SIRP␣. Antibodies to CD47 inhibited SIRP␣ phosphorylation, confirming the requirement of CD47-SIRP␣ interactions in this phosphorylation. Furthermore, SIRP␣ expressed by CD47 Ϫ MLEC mixed with CD47 ϩ , SIRP CT Ϫ/Ϫ BMDM was phosphorylated, demonstrating the importance of CD47 ligation in trans in MLEC. As in BMDM and neutrophils, phosphorylation of SIRP␣ in endothelial cells required adhesion, since CD47 ϩ MLEC incubated in a pellet, in which cell-cell contact was maximized, but not adherent to a surface, had reduced levels of SIRP␣ phosphorylation, and even contact with CD47 on a surface could not induce SIRP␣ phosphorylation in the absence of integrin-mediated adhesion. Although previous reports have shown that integrin-mediated adhesion could induce SIRP␣ phosphorylation, these earlier studies all utilized CD47 ϩ cells. Thus, these earlier reports could not differentiate between the roles of CD47 in modulating integrin activity, in CD47-integrin cooperation, or in direct CD47-SIRP␣ interactions. Our genetic data demonstrate that in both myeloid and endothelial cells, integrin-mediated adhesion is a requirement for SIRP␣ phosphorylation and also illuminate a fundamental distinction in regulation of this phosphorylation. In myeloid cells, CD47 expression increased levels of SIRP␣ phosphorylation by ϳ1.5fold, with abundant phospho-SIRP␣ detected in CD47 Ϫ cells, whereas in endothelial cells, CD47 expression increased SIRP␣ phosphorylation more than 30-fold, and phospho-SIRP␣ was essentially undetectable in CD47 Ϫ MLEC. Thus, adhesion is sufficient in the myeloid cells, whereas adhesion and ligation of SIRP␣ with CD47 both are required in endothelia.
The nature of the adhesive signal required for SIRP␣ phosphorylation in leukocytes and endothelial cells is consistent with activation of Src family kinases by integrin signaling, since this family has been shown to phosphorylate SIRP␣ ITIMs (12,13,42). Indeed, the Src family kinase inhibitors PP1 and PP2 blocked adhesion-dependent SIRP␣ phosphorylation in both fibroblasts (12) and endothelial cells. 3 Our data suggest that the difference in requirement for CD47 for SIRP␣ phosphorylation in leukocytes and endothelial cells is due to a difference in the association of SIRP␣ with the actin cytoskeleton in the two cell types. In BMDM, SIRP␣ is tightly associated with actin, whereas in MLEC, we found no such colocalization. These data suggest that the close proximity of SIRP␣ in myeloid cells to focal complexes and podosomes where integrin ligation recruits activated Src family kinases presumably enhances the kinetics and extent of phosphorylation. In endothelial cells, the majority of integrin-containing focal adhesions (and associated Src family kinase activity) are located on the basal surface of the cells, and actin fibers are organized around the perimeter of the cells, where they connect to junctional complexes between adjacent cells. Src family kinases are recruited to these intercellular contacts, where they are involved in the regulation of VEcadherin phosphorylation and endothelial permeability (6,(43)(44)(45)(46)(47). We hypothesize that ligation of SIRP␣ with CD47 at cell-cell junctions in endothelial cells facilitates its phosphorylation by Src family kinases recruited to nearby adherens junctions.
Several previous studies have suggested that macrophage SIRP␣ phosphorylation induced by interaction with target CD47 negatively regulates Fc␥Rand CR3-mediated phagocytosis (15)(16)(17)(18)48). The presence of considerable basal, CD47independent SIRP␣ phosphorylation in macrophages potentially complicates these models. In order for CD47-mediated SIRP␣ phosphorylation to inhibit phagocytosis in the presence of a considerable background of CD47-independent phosphorylation, the cell probably must be able to decipher spatial cues as well. In other words, the adherent macrophages would have to distinguish phosphorylated SIRP␣ associated with adhesion from those molecules phosphorylated in response to CD47 on a phagocytic target. One possibility is that the macrophage is able to differentiate between SIRP␣ phosphorylation occurring at its apical surface from SIRP␣ phosphorylation occurring at its lateral edges or basal surface. Another possibility is that other receptor-ligand interactions, in addition to CD47-SIRP␣, are required for the macrophage to see SIRP␣ phosphorylation as a "don't eat me signal" as opposed to a steady state adherence signal.
Our experiments suggest that in endothelial cells, CD47 and SIRP␣ interactions occur at cell-cell contacts. The increasing amounts of phosphorylated SIRP␣ that occur with increasing cell density in endothelia are reminiscent, although opposite in intensity, of phosphorylation changes that occur with the adherens junction components VE-cadherin and associated catenins in maturing endothelial monolayers (40,47). It has been suggested that decreased cadherin and catenin phosphorylation results from recruitment of SHP-2 to the adherens junction, leading to strengthening of the association of the adherens junction with the actin cytoskeleton (40,47,49). We detected increased phosphorylation of VE-cadherin and associated catenins in high density CD47 Ϫ MLEC monolayers relative to CD47 ϩ monolayers (Fig. 7B), although there appear to be similar amounts of ␣-catenin associated with VE-cadherin in both the CD47 ϩ and CD47 Ϫ MLEC. 3 Thus, CD47-SIRP␣ interactions in endothelial cells may be a mechanism for recruitment of SHP-2 to the adherens junction with consequences for junctional maturation.