Endothelial CD47 interaction with SIRPγ is required for human T-cell transendothelial migration under shear flow conditions in vitro

Recruitment of leukocytes from the bloodstream into tissues during an inflammatory response involves a multistep adhesive and signaling cascade composed of selectin-mediated rolling and initial attachment, subsequent integrin-mediated arrest and migration on the apical surface, followed by transendothelial migration (TEM; diapedesis).¹  TEM involves multiple adhesion molecule pathways such as intracellular adhesion molecule (ICAM)-1–CD18, vascular cell adhesion molecule (VCAM)-1–VLA-4, CD47, platelet endothelial cell adhesion molecule (PECAM)-1, CD99, endothelial cell adhesion molecule (ESAM), junctional adhesion molecules (JAMs), and DNAX accessory molecule 1 [CD226]–polio virus receptor [CD155] (DNAM-1–PVR), although the cellular mechanisms controlling TEM are incompletely understood.² 

CD47 or integrin-associated protein (IAP) belongs to the immunoglobulin superfamily (IgSF) and is highly expressed in most cell types. It is a 50-kDa transmembrane protein that consists of an extracellular amino-terminal Ig domain, 5 highly hydrophobic putative membrane-spanning segments, and a short cytoplasmic tail.³  CD47 has been shown to associate with integrins at the cell surface and signal through G-coupled proteins.⁴ 

Important cellular ligands for CD47 are the signal regulatory proteins (SIRPs) SIRPα and SIRPγ. The SIRPs comprise a family of several transmembrane glycoproteins that belong to the IgSF and are present in the immune and central nervous system.⁵  Each molecule contains 3 homologous extracellular Ig-like domains with distinct transmembrane and cytoplasmic domains. SIRPα was the first member to be identified and was detected on hematopoietic progenitors, on myeloid cells such as granulocytes and macrophages, and on dendritic cells and neurons.⁵  The extracellular Ig domain of SIRPα binds to CD47 and transmits intracellular signals through its cytoplasmic domain that contains 4 immunoreceptor tyrosine-based inhibitory motifs (ITIMs). CD47 association with SIRPα has been reported to induce phosphorylation of ITIMs which, in turn, recruits Shp-1 and −2^6,7  and induces a negative signal. SIRPα has been linked to cell adhesion and migration of neutrophils.^5,8,9  CD47 interactions with SIRPα mediate, in part, neutrophil transepithelial migration in vitro.^5,8-10  CD47 was also shown to regulate PMN transepithelial migration through both SIRPα-dependent and SIRPα-independent mechanisms.⁸  In addition, monocyte SIRPα interacting with endothelial cell CD47 was shown to mediate, in part, TEM across rat cerebral endothelium under flow conditions in vitro.¹¹  A second member is SIRPβ1, and it is also expressed in myeloid cells. SIRPβ1 generates a positive signal by intracellular signaling of its cytoplasmic tail through its association with a transmembrane protein called DNAX activation protein 12 or DAP12. The cytoplasmic tail of DAP12 possesses immunoreceptor tyrosine-based activation motifs (ITAMs) that link SIRPβ1 to activation machinery.^12,13  Although SIRPβ1 shares significant amino acid sequence homology with SIRPα, SIRPβ1 is not a ligand for CD47.⁵  A third member of the SIRP family is SIRPγ (also termed SIRPβ2 or CD172g).¹⁴  Although SIRPγ, SIRPα, and SIRPβ1 are highly homologous in the extracellular Ig domains, the cytoplasmic tail of SIRPγ is distinct. It consists of 4 amino acids and lacks the basic lysine residue required for association to the adaptor protein DAP12; thus, it has been suggested to act as a decoy receptor.⁵  Unlike the other SIRP proteins, SIRPγ is highly expressed in peripheral T cells and in 10% to 20% of B cells. SIRPγ was also shown to bind to CD47 but with a lower affinity (K_d ≈ 23 μM) than SIRPα (K_d ≈ 1.5 μM).^15,16 

Recent reports indicate that CD47 contributes to T-cell arrest on inflammatory vascular endothelium.¹⁷  The contribution of endothelial cell CD47 to T-cell TEM under flow conditions, however, has not been addressed. In this study, we assessed the role of endothelial cell CD47 and T-cell SIRPγ in T-cell TEM under shear flow conditions using blocking mAbs and human umbilical vein and murine heart microvascular endothelial cells derived from strain matched wild-type (WT) and CD47^−/− mice. Our results indicate that endothelial cell CD47 engagement by its T-cell ligand SIRPγ, or other as yet identified molecules in the mouse, play a dominant role in T-cell transendothelial migration under shear flow in vitro. Because SIRPγ lacks obvious signaling potential, the data suggest engagement of endothelial CD47 triggers downstream signaling in endothelium and that these signals, either direct, or indirect by interactions with endothelial cell integrins or G_i-coupled receptors,⁴  are essential for T-cell transmigration.

DPBS (Dulbecco phosphate-buffered saline), with Ca²⁺ and Mg²⁺ (DPBS⁺) and without (DPBS⁻), and M199 were obtained from ThermoFisher (Waltham, MA). Human recombinant TNF-α and SDF-1α (stromal cell-derived factor-1α, CXCL12) were from PeproTech (Rocky Hill, NJ). Murine recombinant TNF-α was from BD PharMingen (San Diego, CA) and was used to activate endothelial cells at 125 ng/mL.

The following mAbs have been reported previously and were used as purified IgG: C5D5¹⁸  and B6H12 (ATCC, Manassas, VA) are function blocking mAbs to human CD47; a nonblocking mAb against CD47, 2D3 was obtained from Dr Eric Brown (Genentech, South San Francisco, CA); blocking anti-SIRPα (SE7C2)¹⁹  and anti-SIRPγ (LSB2.20)¹⁵  mAbs were from Santa Cruz Biotechnology (Santa Cruz, CA); rabbit polyclonal antibody to the cytoplasmic tail of SIRPα (catalog no. 1125) was obtained from ProSci (Poway, CA); Hec-1 mAb²⁰  recognizes human VE-cadherin and was obtained from Dr William Muller (Northwestern University, Evanston, IL); anti-MHC class I (W6/32) mAb was from ATCC²¹ ; and anti–ICAM-1 function blocking (HU5/3) mAb. The anti–alkaline phosphatase mAb was from Sigma-Aldrich (St Louis, MO). Alexa 488–conjugated goat anti–mouse F(ab′)2 was obtained from Invitrogen (Carlsbad, CA). Rat mAbs were as follows: anti–mouse IL-4 (clone BVD4-1D11), anti-CD3 (clone 145-2C11), and anti-CD28 (clone 37.51) were from BD PharMingen. PE-conjugated rat anti–mouse ICAM-1 (YN1/1.7.4) and FITC-conjugated rat anti–mouse VCAM-1 (clone 429) mAbs were from eBioscience (San Diego, CA). PE-conjugated rat anti–mouse ICAM-2 (3C4) and anti–mouse E-selectin (clone 10E9.6) mAbs were purchased from BD PharMingen. Rat anti–mouse CD47 (Miap301) and anti–mouse VE-cadherin (11D4.1) mAbs were from BD PharMingen. Secondary antibodies coupled to alkaline phosphatase were from Promega (Madison, WI).

Human umbilical vein endothelial cells (HUVECs) were isolated, pooled, and cultured as described.²²  Endothelial cells were grown to confluence on 25-mm-diameter circular glass coverslips precoated with fibronectin (5 μg/mL; BD Biosciences). HUVEC monolayers were stimulated with TNF-α (25 ng/mL for 4 hours) and then treated with SDF-1α (50 ng/mL; CXCL12) for 15 minutes before insertion into the flow chamber for TEM studies as previously described.²²  This step promotes T-cell TEM.²³  Human CD3⁺ T cells (> 90% pure) were isolated by negative selection from anticoagulated whole blood of healthy volunteers as previously described.²²  Blood was drawn and handled according to protocols for protection of human subjects approved by the Brigham and Women's Hospital Institutional Review Board, and all volunteer subjects gave informed consent, in accordance with the Declaration of Helsinki. The isolation and culture of murine heart endothelial cells (MHECs) from 7- to 9-day-old pups was performed as described.²¹  Confluent monolayers of MHECs were used in flow studies between passages 2 and 4.

Lymph nodes and spleens were removed from C57BL/6 mice, and CD4⁺ T cells were isolated by positive selection as described previously.²⁴  Overall, 95% (± 11%; n = 2) of the cells were CD4⁺ as assessed by flow cytometry. The cells were suspended in RPMI 1640 medium supplemented with 10% fetal calf serum, 2 mM l-glutamine, 10 mM HEPES, 100 U/mL penicillin, and 100 U/mL streptomycin, sodium pyruvate, sodium bicarbonate, and nonessential amino acids (Invitrogen). The cell suspension was incubated overnight in polystyrene culture wells coated with anti-CD3 mAb (5 μg/mL). Subsequently, CD4⁺ T cells were incubated under T helper type 1 (Th1) polarizing conditions (anti-CD28; [1 μg/mL], recombinant IL-12 [10 ng/mL], and neutralizing anti–IL-4 [0.5 μg/mL]). After 3 days the cultures were supplemented with murine IL-2 (25 U/mL). T cells attained a Th1 phenotype after 5 to 6 days in culture and thus were used in flow studies.²⁴  The Harvard Medical School animal resources program is accredited by the American Association for the Accreditation of Laboratory Animal Care and meets National Institutes of Health standards as set forth in the Guide for the Care and Use of Laboratory Animals (DHHS Publication No. [NIH]85-23, revised 1985).

Confluent human HUVEC monolayers were incubated in culture medium alone or medium containing 25 ng/mL TNF-α, 10 U/mL IL-1β, or IL-1β plus IFN-γ (100 U/mL). Human CD47 was detected with B6H12 mAb. MHECs were stimulated for 4 to 6 hours with 125 ng/mL murine TNF-α and then stained with directly labeled ICAM-1, E-selectin, ICAM-2, or VCAM-1 mAb for 30 minutes. Rat anti–mouse CD47 (miap301) and VE-cadherin (11D4.1) mAbs were also used, and their binding was detected FITC-labeled secondary mAb and analyzed on a fluorescence-activated cell scanner (FACScan) instrument (BD Biosciences).

Culture supernatant containing secreted recombinant CD47-alkaline phosphatase (CD47-AP) fusion protein was prepared by stable transfection of Chinese hamster ovary K1 (CHO-K1) cells as described earlier.⁸  Human CD3⁺ T-cell adhesion to immobilized soluble CD47-AP fusion protein was performed as described²⁵  with the following modifications. Briefly, an anti-AP antibody (8 μg/mL) was immobilized (4°C overnight) in microtiter wells, blocked with 3% BSA in PBS, and then incubated with culture supernatant containing soluble recombinant CD47-AP protein at 4°C for 30 minutes. The wells were washed with TBS containing 0.1% BSA and then pretreated with appropriate mAb. CD3⁺ T cells (10⁵/well) suspended in RPMI 1640 with 1 mM MgCl₂ and 0.1% BSA were added to wells for 30 minutes at 37°C. Nonadherent cells were removed by 4 gentle washes, and adherent cells were counted in 4 fields in triplicate wells. The results of mAb inhibition were expressed as a relative percentage of attached cells, whereby 100% is the number of cells bound in medium alone.

Confluent HUVEC monolayers (subculture 2) on fibronectin-coated glass coverslips were treated with media containing TNF-α (10 ng/mL) or medium alone for 4 hours, washed with DPBS⁺, and fixed at RT for 8 minutes in 2% paraformaldehyde (Sigma-Aldrich). Fixed HUVEC coverslips were blocked with DPBS⁺ containing 1% goat serum (Vector Laboratories, Burlingame, CA) at 4°C for 1 hour. Monolayers were incubated with primary mAb (B6H12 and Hec-1; each at 10 μg/mL) at 25°C for 1 hour, washed with DPBS⁺, and incubated with goat F(ab′)2 anti–mouse IgG Alexa-488–conjugate (1:400 dilution) in block buffer at 25°C for 30 minutes. Monolayers were then washed and mounted with FluorSave Reagent (Calbiochem, San Diego, CA). Cellular distribution of CD47 and VE-cadherin was examined by immunofluorescence microscopy using a 60×/1.4 NA oil objective, and images were captured with a digital imaging system coupled to a Nikon TE2000 inverted microscope (Melville, NY) as previously described.²⁶  Images were analyzed and processed using MetaMorph software and Image J v1.31 (http://rsb.info.nih.gov/ij). Images presented are representative of 4 experiments with HUVECs from different preparations.

Purified human CD3⁺ T cells and PMNs from single donors were lysed with 1% Triton X-100 in PBS supplemented with a protease and phosphatase inhibitor cocktail (Sigma-Aldrich). Equal aliquots of samples (30 μg protein) were loaded on a 10% sodium dodecyl sulfate–polyacrylamide gel electrophoresis (SDS-PAGE) gels (Bio-Rad, Hercules, CA), separated by electrophoresis under reducing conditions and then transferred to nitrocellulose membranes (Whatman Schleicher and Schuell, Dasel, Germany).²⁵  The membranes were incubated first with anti-SIRPα (SE7C2; 10 μg/mL) at 37°C for 30 minutes, followed by an HRP-conjugated goat anti–mouse IgG (1/5000 dilution), and the blot was developed by Enhanced ChemiLuminescence (ECL) System (Pierce Biotechnology, Rockford, IL). The membrane was stripped and reblotted with a SIRPγ (LSB2.20) mAb (10 μg/mL) under identical conditions.

Confluent 4-hour TNF-α–activated human monolayers on glass coverslips were inserted into the flow chamber for adhesion and transmigration assays at 37°C. T cells (10⁶ cells in 100 μL of DPBS⁺ containing 0.1% human serum albumin [flow buffer]) were drawn through the chamber as a bolus at 1 dyne/cm² until 3 to 5 cells per field attached. The shear flow was then reduced to 0.2 dyne/cm² for 1 minute to allow T-cell accumulation. The shear flow was then increased to 0.75 dyne/cm² for a 10-minute period. In murine migration experiments, MHECs were stimulated with TNF-α for 6 hours, and murine Th1 effector cells were drawn though the chamber as a bolus of 2 × 10⁶ cells in 200 μL as above. Live cell imaging was performed using MetaMorph, v5.0 software (Molecular Devices, Downingtown, PA) to control a digital imaging system coupled to an inverted microscope (model TE2000; Nikon, Melville, NY) equipped with a PlanApo 20× differential interference contrast (DIC) objective. A high-sensitivity cooled CCD camera (ORCA-ER; Hamamatsu, Bridgewater, NJ) acquired images via the MetaMorph software. Images were analyzed and processed using MetaMorph and ImageJ v1.31 (http://rsb.info.nih.gov/ij) software. Selected images are representative of multiple experiments on different days with leukocytes isolated from different donors. The percentage of TEM was calculated as follows: total transmigrated T cells/[total adhered + transmigrated T cells] × 100. T-cell migration velocity was analyzed by Image J software (National Institutes of Health, Bethesda, MD).

Data were compared using one-way analysis of variance for multiple groups. Data were considered statistically significant if P was less than .05.

Flow cytometric experiments were performed to assess the surface expression of CD47 in confluent HUVEC monolayers (Figure 1A). The data show that resting, nonstimulated HUVECs express modest levels of CD47 (0 hours) and that the total CD47 expression level was slightly increased in HUVECs stimulated for 4 or 24 hours with TNF-α, IL-1β alone, or IL-1β plus IFN-γ. Treatment of HUVECs with phorbol esters, or Gram-negative lipopolysaccharide alone, did not significantly alter expression of CD47 (data not shown). These data indicate that CD47 is constitutively expressed in HUVECs and is slightly increased after cytokine treatments. To study the cell-surface distribution of CD47, confluent HUVEC monolayers were stimulated with TNF-α or left untreated, followed by staining with anti-CD47 (B6H12) or VE-cadherin (Hec 1) mAbs and examined by immunofluorescence microscopy (Figure 1B). The staining pattern of CD47 on the apical surface is heterogeneous with enrichment at cell-cell junctions, which is consistent with a recent report.²⁷  For comparison, VE-cadherin staining is used as a marker of endothelial cell junctions. HUVECs treated for 4 hours with TNF-α showed the same pattern of CD47 and VE-cadherin staining, indicating that CD47 surface expression and spatial localization is not significantly altered by cytokines in vitro (data not shown).

T-cell trafficking across HUVEC monolayers was assessed by live cell, time-lapse digital differential interference contrast (DIC) microscopy. In this model purified CD3⁺ T cells arrest uniformly on 4 hours of TNF-α activated monolayers and in the presence of exogenously added apical chemokine SDF-1α (CXCL12), polarize, migrate to endothelial cell-cell junctions, and ultimately many transmigrate.²²  Pretreatment of HUVECs with anti-CD47 mAb B6H12 or mAb that recognize ICAM-1 (HU5/3) or class I (W6/32) did not alter T-cell accumulation on TNF-α–activated endothelium (Figure 2A). However, anti-CD47 mAb significantly inhibited T-cell transmigration (70% ± 6%; P = .001; Figure 2B) and caused a 50% drop in T-cell apical migration velocity (P = .001; Figure 2D). A similar level of inhibition was obtained with another anti-CD47 mAb C5D5 (data not shown). Treatment of HUVECs with a blocking mAb against ICAM-1 (HU5/3) also significantly reduced TEM but not to the extent of anti-CD47 mAb. Thus, interfering with either the CD47-SIRPγ or ICAM-1–LFA-1 interaction seemed to have a similar inhibitory effect on migration, suggesting a role of both pathways in T-cell TEM under flow conditions. Interestingly, live cell microscopy of T cells treated with anti-CD47 mAb showed that many T cells (identified by arrowheads in Figure 2C middle; Video S1, available on the Blood website; see the Supplemental Materials link at the top of the online article) had a rounded-up morphology and failed to transmigrate. In contrast, medium alone (Figure 2C left, no Ab) or nonblocking class I mAb (Figure 2C right; Video S2) treated T cells polarized and transmigrated (identified by arrows). The blocking ICAM-1 mAb did not alter T-cell morphology compared with anti-CD47 mAb (data not shown).

Recent studies using protein-protein binding assay have reported that SIRPγ is a ligand for CD47 and is selectively expressed by CD3⁺ T cells and 10% to 20% of CD19⁺ B cells.¹⁵  On the basis of these data, we tested commercial mAbs to SIRPγ (LSB2.20) and SIRPα (SE7C2) for staining of human CD3⁺ T cells and PMNs. We confirmed these reports and show that SIRPγ is expressed uniformly by CD3⁺ T cells but not PMNs by fluorescence-activated cell sorting (FACS) analysis (Figure 3A). In contrast, SIRPα is abundant on PMNs but not detected on CD3⁺ T cells. For biochemical analysis, T cells and PMNs were lysed, and identical amounts of lysate protein were subjected to SDS-PAGE and Western blot. The mAb to SIRPγ strongly reacted with a protein of 45 to 47 kDa, whereas the mAb to SIRPα was not reactive to T-cell lysates under reduced or nonreduced conditions (Figure 3B top), consistent with recent work by Brooke et al.¹⁵  In addition, a polyclonal antibody to the carboxy terminus of SIRPα, which in principal can detect all SIRPα species including the recently described polymorphic variants,²⁸  also failed to react with T-cell lysates, whereas this Ab reacted strongly with a 80- to 110-kDa band in PMN lysates (Figure 3B lower). In contrast, PMNs expressed SIRPα but not SIRPγ. Thus, CD3⁺ T cells express SIRPγ but not SIRPα. Interestingly, FACS analysis data showed that HUVECs express SIRPα (Figure 3C left) but not SIRPγ (Figure 3C right) under resting conditions and that this pattern was not changed after 4 hours of TNF-α treatment. This was confirmed by Western blot (data not shown).

Because SIRPγ mAb LSB2.20 was reported to block the SIRPγ-CD47 interactions between T cells and dendritic cells,²⁹  we tested this mAb in an in vitro adhesion assay using immobilized CD47-AP fusion protein under static conditions (Figure 4). Resting CD3⁺ T cells adhered to immobilized CD47. Microtiter wells coated with anti-AP and BSA alone did not support CD3⁺ T-cell adhesion (data not shown). T-cell adherence was efficiently inhibited by anti-SIRPγ (LSB2.20; 20 μg/mL) when preincubated with T cells and by both anti-CD47 mAbs B6H12 or C5D5, indicating a SIRPγ protein-directed binding. Nonblocking anti-CD47 mAb 2D3 and a control Class I mAb W6/32 had no effect on T-cell adhesion under the same conditions. Thus, even though the affinity of SIRPγ for CD47 is 10-fold less than the SIRPα-CD47 interaction, the affinity is clearly adequate for cell adhesion.

In a transmigration assay under flow conditions, pretreatment of T cells with anti-SIRPγ mAb LSB2.20 had no effect on T-cell accumulation (Figure 5A). However, the same mAb caused a dramatic block in T-cell transmigration of TNF-α–activated HUVECs under the same conditions (Figure 5B). Control mAb to class I (Figure 5B,C left, W6/32 mAb) and media alone had no effect on T-cell TEM under the same conditions. In the presence of SIRPγ mAb (LSB2.20 mAb; Figure 5C middle), T cells showed defects in cell polarization (cells identified by arrowheads; also see Video S3) and eventually, failed to transmigrate (75% ± 2% inhibition). The SIRPγ mAb also reduced T-cell apical migration velocity by 45% (6.2 μm/min class I mAb vs 3.4 μm/min CD47 mAb; n = 30 cells, n = 3 experiments; Figure 5D), which was similar to the effect observed with mAb blocking of endothelial CD47.

SIRPα expression and function have been mostly studied in myeloid cells and neurons.⁵  In a recent study, it was reported that SIRPα is constitutively expressed in murine endothelial cells.²⁷  We showed that HUVECs also express low levels of SIRPα but not SIRPγ as assessed by both FACS (Figure 3C) and Western blot (not shown) under resting conditions, and this pattern was not changed after 4 or 24 hours of TNF-α treatment.

To determine whether the endothelial SIRPα plays a role in T-cell migration, human endothelial cell monolayers were treated with anti-SIRPα blocking mAb. Although the anti-CD47 mAb, B6H12 significantly inhibited T-cell TEM, anti-SIRPα mAb (SE7C2) showed no effect. Taken together, these data indicate that the endothelial cell CD47–T-cell SIRPγ interactions are critical for T-cell transmigration in vitro and that disruption of endothelial cell SIRPα-CD47 cis interactions by blocking mAb to SIRPγ do not affect T-cell TEM.

To further corroborate the importance of endothelial CD47 in T-cell transmigration, we isolated and cultured MHECs from CD47^−/− and WT (C57/BL6) mice as described earlier.³⁰  As predicted, flow cytometric studies showed no expression of CD47 in the CD47^−/− endothelial cells compared with WT MHECs, either with (Figure 6A right) or without (Figure 6A left) TNF-α treatment. Surprisingly, WT in vitro differentiated murine Th1 cells showed a significant increase in adhesion to CD47^−/− MHEC monolayers compared with WT (Figure 6B). Despite this increased adhesion, however, there was a dramatic decrease (75% ± 2.4%) in T-cell migration across TNF-α–treated CD47^−/− MHECs compared with WT (Figure 6C). This result correlates with the mAb blocking experiments in the human cell studies in Figure 2. Interestingly, Th1 cells interacting with TNF-α–treated CD47^−/− MHEC monolayers exhibited a rounded-up morphology and had a significantly lower rate of TEM (Figure 6D). This behavior is similar to the phenotype observed in the human mAb blocking studies as shown in Video S1. In contrast, WT MHECs supported significant Th1 cell adhesion and transmigration under identical conditions.

To investigate the increased adhesion phenotype observed with CD47^−/− MHECs, we tested whether there were differences in the expression of cytokine-inducible endothelial adhesion molecules. Interestingly, CD47^−/− MHECs showed significantly higher ICAM-1 expression compared with WT MHECs in both TNF-α–treated and sham-treated MHECs (Figure 6E), which would explain the increased Th1 cell adhesion to CD47^−/− MHECs. This effect was specific to ICAM-1, because there was no difference in WT versus CD47^−/− MHEC expression of ICAM-2, VCAM-1, E-selectin, or VE-cadherin, with or without TNF-α (Figure 6F). Taken together, these data clearly indicate that endothelial CD47 plays a crucial role in T-cell transendothelial migration under flow conditions in human and murine in vitro cell systems.

CD47 is ubiquitously expressed and known to modulate several cellular processes on hematopoietic cells, including phagocytosis, cytokine production, T-cell responsiveness, and leukocyte adhesion and transendothelial and transepithelial migration in vitro.^4,8,29,31  Anti-CD47 mAbs, such as B6H12 and C5D5 were previously shown to block human neutrophil transendothelial and transepithelial migration under static conditions, respectively.^10,32  CD47 interacting with SIRPα was recently reported to participate in rat monocyte TEM of rat cerebral endothelium in vitro under static conditions, and this process was dependent on the activation of G_i protein, a known signaling pathway of CD47.¹¹  However, the contribution of endothelial CD47 in T-cell adhesion and transendothelial migration has not been explored in vitro. In this study, we present evidence that CD47 is enriched in cell-cell junctions in endothelial cells and that CD47 interacting with human T-cell SIRPγ or a comparable molecule(s) in murine T cells plays an important role during T-cell transmigration in vitro under shear flow conditions.

Flow cytometric analysis showed moderate surface expression of CD47 in HUVECs under basal and TNF-α–activated conditions, and its expression pattern was found to be enriched at cell-cell junctions, suggesting a role for this molecule in paracellular T-cell transendothelial migration. Strikingly, despite the lack of effect on T-cell adhesion, the blocking mAbs to CD47 inhibited 70% of T-cell transmigration across TNF-activated endothelium in vitro, and adherent T cells showed a rounded-up morphology. Most cells were unable to polarize and flatten on CD47 mAb-treated endothelial monolayers, which correlated with a significant drop in apical migration velocity of T cells. These data suggest endothelial cell CD47 provides essential spatial orientation signals to T cells during transendothelial migration (see Videos S1Video 2. Adhesion and migration of CD3+ T cells of control Class I mAb treated TNF-α activated endothelium under flow (MOV, 406 KB)–S3). In this regard, the CD47-SIRPγ pathway seems to behave similarly to Mac-1 on neutrophils during transendothelial migration in vivo³³  and in monocytes during transmigration in vitro.³⁴  Future studies are necessary to understand the mechanisms underlying the CD47 and SIRPγ pathway signaling.

In T cells, CD47 has been shown to interact in cis with α₄β₁ and α₅β₁ integrins³⁵  and to act as a receptor for thrombospondin-1 (TSP-1) and as well SIRPα and SIRPγ.^4,15  To date, several SIRPs have been identified in human cells, and each has different signaling properties and different affinities for CD47. SIRPγ was identified as a CD47 ligand by protein-protein binding studies.²⁹  In accordance with previous studies,²⁹  we show that SIRPγ is expressed in human CD3⁺ T cells, whereas SIRPα is not as assessed by flow cytometry or Western blot. A previous study reported that soluble SIRPγ binds to CD47 with a lower affinity compared with SIRPα using BIAcore analysis.^15,16,29  However, it has not been established what role SIRPγ plays in T-cell recruitment. Here, we show that the recombinant CD47-AP fusion protein supports robust CD3⁺ T-cell adhesion despite the low binding affinity reported.^15,16,29  The specificity of these interactions was clearly shown by the inhibition of T-cell adhesion to immobilized CD47 by mAb to CD47 and SIRPγ. Recent studies showed that integrin-specific blocking mAbs had no effect on T-cell SIRPγ-CD47 complex formation and by extension, on cell adhesion, strongly suggesting this interaction is independent of leukocyte β₂ integrins.²⁹  In contrast to SIRPα, SIRPγ contains a short cytoplasmic tail, which is considered not to signal and rather may act as a decoy receptor.²⁹  This raises the intriguing question of how this pathway signals. Unlike the bidirectional signaling that occurs after CD47-SIRPα interactions, our data suggest that CD47 engagement by SIRPγ activates CD47-induced unidirectional signaling pathways in the endothelial cell. These signals may facilitate TEM, either directly or indirectly, by G_i-coupled signaling or through association with αvβ3 or α2β1integrins expressed in endothelial cells. Future studies are necessary to address the identity of these signals. SIRPγ association with CD47 was capable of promoting T-cell TEM despite its 10-fold lower affinity compared with the interaction between SIRPα-CD47 that promotes myeloid cell TEM (Figure 7). As previously reported for the CD47-SIRPα complex, it is also possible that SIRPγ ligation to CD47 plays a role in other functions of T cells, including T-cell activation and migration to peripheral tissues and lymph nodes during inflammation,^4,8,38  as well as a role in organ graft rejection.²⁸ 

CD47 interactions with their natural ligands SIRPs are known to occur in both human and mouse cells, but they also are likely to have functional differences such as cell-surface mobility, kinetics, and clustering across species. Interestingly, there was no evidence for SIRPγ in mice and rats, as shown in searches for genomic and EST sequence databases.¹⁵  However, there was no biochemical, immunohistochemical, or flow cytometric data corroborating this conclusion. Nevertheless, these data suggest that an as yet unidentified SIRP family member or a new ligand for CD47 is expressed by murine T cells.

To date, there are limited reports examining the role of CD47 in models of inflammation in vivo. CD47 knockout mice develop normally and show normal appearance, fertility, and organ histology. However, these mice show reduced neutrophil recruitment at early time points, fail to kill ingested bacteria in a model of peritonitis, and develop mild anemia and thrombocytopenia.^39,40  In these studies, no quantitation of CD3⁺ T-cell recruitment to the inflammatory sites was included. A second report addressed the role of CD47, TSP-1, and TSP-2 in a murine model of delayed-type hypersensitivity (DTH).⁴¹  Interestingly, in the absence of CD47, TSP-1, or TSP-2, no defect in T-cell recruitment was observed to dermal DTH sites in vivo. It is clear that these 2 studies offer conflicting results. In CD47^−/− mice, CD47 is absent in both endothelial cells and leukocytes. To begin to address the role of endothelial CD47 alone in T-cell TEM and to corroborate to the human endothelial CD47 mAb blocking data, we isolated endothelial cells from hearts of 8-day-old WT and CD47^−/− mice. Th1 effector cells prepared from WT mice transmigrated poorly across TNF-α–treated CD47^−/− MHECs, and T cells lacked the ability to polarize compared with WT MHECs. These data clearly show that endothelial CD47 is essential for T-cell transendothelial migration. Surprisingly, T-cell adhesion was shown to be increased by 2-fold in CD47^−/− MHECs. The increased T-cell adhesion strongly correlated with a greater than expected increase in ICAM-1 expression in CD47^−/− MHECs compared with WT. This effect seemed to be specific to ICAM-1 because no change in the level of other adhesion molecules was detected. Previously, we have shown that enhanced expression of ICAM-1 in iHUVEC transfectants resulted in a variable increase in leukocyte adhesion on TNF-α–stimulated iHUVECs but had no effect on the total number of adherent PMNs that transmigrated.⁴²  Therefore, we suggest it is unlikely that the T-cell migration inhibitory effect observed is due to the increased ICAM-1 levels present in CD47^−/− MHEC monolayers. These results indicate that CD47 may not only be important during the transmigration step but also that it could act as a regulator of either expression or turnover of ICAM-1 in mouse heart endothelial cells in vitro.

We thank Drs Eric Brown and Mette Johansen (Genentech Inc and University of California at San Francisco) for providing the CD47^−/− mice and Deanna Lamont and Pilar Alcaide for technical assistance.

This work was supported by National Institutes of Health (Bethesda, MD) grants HL36028 (F.W.L.), DK79392 (C.A.P.), and DK72564 (C.A.P.). M.S. was supported by a fellowship award from the American Heart Association (0725989T).

National Institutes of Health

Contribution: M.S. designed the research, performed all the experimental work, analyzed the data, and wrote the paper; W.Y.L. and C.A.P. provided key reagents; G.N. provided critical technical assistance, and F.W.L. supervised the study.

Conflict-of-interest disclosure: The authors declare no competing financial interests.

Correspondence: Francis W. Luscinskas, Brigham and Women's Hospital, 77 Avenue Louis Pasteur, NRB-752P, Boston, MA 02115; e-mail: fluscinskas@rics.bwh.harvard.edu.