CD47 Signaling Regulates the Immunosuppressive Activity of VEGF in T Cells

Vascular endothelial growth factor (VEGF)-A is a critical growth factor for endothelial cells, and even a moderate decrease in VEGF gene dosage is lethal for embryonic vascular development (1). Conversely, many cancer patients have increased levels of circulating VEGF, and VEGF is a major driver of tumor neovascularization (2). In addition to stimulating angiogenesis, some tumor cells express the VEGF tyrosine kinase receptors VEGF receptor-1 (VEGFR1) and VEGFR2, and VEGF can be an autocrine growth and motility factor for these cancers (3–8). Based on these functions in tumor growth, several drugs targeting VEGF or the kinase activity of VEGFR2 have proved effective for controlling tumor angiogenesis and growth (9). However, many cancers develop resistance to VEGF antagonists, and tumor vascular responses to treatment may not correlate with improved survival (10).

The immune system is emerging as another important target of VEGF (11, 12). High levels of VEGF result in immunosuppression by inhibiting dendritic cell functions (13–17). Human CD4⁺CD45RO⁺ T cells and Jurkat T lymphoma cells express VEGFR1 and VEGFR2 at the mRNA and protein levels (18). Expression of VEGF and its receptors VEGFR1 and VEGFR2 is induced in T lymphocytes activated by anti-CD3 and anti-CD28 (19, 20). VEGF induces AKT and ERK phosphorylation in T cells, which can be inhibited by VEGFR2 small interfering RNA. VEGF signaling via VEGFR2 inhibits proliferation of T lymphocytes derived from ovarian cancer patients and healthy volunteers (19). Despite some inconsistencies, these reports demonstrate that VEGF and its receptors are expressed and functional in T cells, and therapeutic VEGF antagonists may consequently have effects on tumor immunity that could be exploited to improve their efficacy.

Further insights into this process may come from studies of endogenous antagonists of VEGF signaling such as thrombospondin-1 (TSP1). TSP1 inhibits tumor angiogenesis and blocks endothelial cell proliferation and chemotaxis by engaging its receptors CD36 and CD47 (21). Binding of the C-terminal domain of TSP1 to CD47 (also known as integrin-associated protein) redundantly inhibits eNOS/NO/cGMP signaling in endothelial cells (22). We recently demonstrated that TSP1 signaling through CD47 also inhibits VEGFR2 phosphorylation at Y¹¹⁷⁵ in human umbilical vein and dermal microvascular endothelial cells (23). However, TSP1 is unable to inhibit VEGFR2 phosphorylation in CD47-null endothelial cells. TSP1 signaling via CD47 suppresses VEGF-induced VEGFR2 phosphorylation without inhibiting VEGF binding. Based on fluorescence resonance energy transfer (FRET) data and coimmunoprecipitation, CD47 laterally associates with VEGFR2 in the absence of their respective ligands. TSP1 binding to CD47 dissociates it from VEGFR2, inhibiting downstream AKT activation and functional responses of endothelial cells to VEGF.

Initial reports that certain immobilized CD47 Abs enhance T cell activation suggested that CD47 is a costimulatory receptor (24, 25). However, microarray data and signal transduction studies revealed that TSP1 globally inhibits TCR signaling induced by anti-CD3 (26), and this inhibitory activity requires CD47 (27–29). A prolonged T cell inflammatory response in CD47 and TSP1-null mice, and in vitro studies using human T cells further indicate that TSP1 signaling through CD47 directly and indirectly limits T cell activation (30, 31).

The findings that VEGF and TSP1 signaling via CD47 separately inhibit TCR-mediated T cell activation combined with our recent identification of the TSP1 receptor CD47 as a VEGFR2 regulator in endothelial cells suggested that CD47 may also modulate the immunoinhibitory activity of VEGF. However, the outcome of this interaction was difficult to predict given that TSP1 and VEGF elicit opposing signals through the CD47/VEGFR2 complex in endothelial cells. We demonstrate that ligation of CD47 by TSP1 inhibits VEGFR2 signaling in T cells in a similar manner as in endothelial cells. VEGF signaling through VEGFR2 inhibits TCR-dependent activation in Jurkat and primary murine T cells but, remarkably, not in human or mouse T cells lacking CD47. Conversely, CD47 signaling regulates the expression of VEGF and VEGF receptors by T cells, and we show that CD47 signaling determines whether VEGF is an autocrine inhibitor or enhancer of T cell activation.

Jurkat T lymphoma cells (American Type Culture Collection) and CD47-deficient JinB8 Jurkat cells (32) were cultured at 37°C with 5% CO₂ in RPMI 1640 medium. Wild-type (WT) and CD47-null C57BL/6J mice (The Jackson Laboratory) were housed in a specific pathogen-free environment and had ad libitum access to standard rat chow and water. Care and handling of animals and all experimental procedures used in this study were approved under protocol LP-012 and in compliance with standards established by the Animal Care and Use Committee of the National Cancer Institute. Pan anti-CD3⁺, CD4⁺, CD8⁺, CD11b⁺, and CD19⁺ cell isolation kits were used to isolate the respective cells from mouse WT and CD47-null spleens according to the manufacturer’s instructions (Miltenyi Biotec).

The TSP1-derived, CD47-binding peptide 7N3 (FIRVVMYEGKK) and inactive control peptide 604 (FIRGGMYEGKK) were synthesized by Peptides International (Louisville, KY) (33). Human TSP1 was purified from the supernatant of thrombin-activated platelets obtained from the National Institutes of Health Blood Bank as described previously (34). VEGF-165 was obtained from R&D Biosystems (Minneapolis, MN) or generously provided by Frank Cuttitta (Advanced Technology Center, National Cancer Institute). Alexa Fluor 594 donkey anti-rabbit IgG and Alexa Fluor488 goat anti-mouse IgG Abs were purchased from Invitrogen (Frederick, MD). VEGFR2, Y¹¹⁷⁵-P, Src, and Src Y⁴¹⁶-P Abs were obtained from Cell Signaling (Danvers, MA). Anti-CD69 was obtained from BD Biosciences. Anti-human CD47 Ab B6H12 was obtained from Abcam, biotinylated anti-human CD47 Ab was from Biolegend, and CD47-FITC was from BD Biosciences. Human VEGFR PE MAb Sampler (R&D Systems) containing PE-conjugated anti-VEGFR1 (clone 49560), VEGFR2 (clone 89106), and VEGFR3 (clone 54733) was used for flow cytometry. VEGF blocking Ab was from Sigma.

The Y¹¹⁷⁵F mutation in the VEGFR2-mCherry fusion vector Ex-I0114-M56 (GeneCopoeia) was generated using PCR primer 5′-AGCAGGATGGCAAAGACTTCATTGTTCTTCCGATATC-3′ and its reverse complement with the QuikChange Lightning kit (Stratagene Agilent Technologies) using the protocol supplied by the manufacturer. This modified nt 3523–3525 from TAC to TTC (underlined) to generate the Y¹¹⁷⁵F codon change in the VEGFR2 coding sequence.

The VEGFR2-mCherry fusion vector Ex-I0114-M56 and CD47-GFP (Dr. Dennis E. Discher) plasmid DNA either together or alone were transfected into Jurkat and JinB8 T cells using FuGENE HD Transfection Reagent (Promega, Madison, WI). At 24 h posttransfection, the GFP- and mCHERRY-positive cells were sorted via flow and used for FRET experiments.

Similarly, Jurkat T cells were transfected with VEGFR2-mCherry (WT) and VEGFR2-mCherry (mutant) plasmid DNA using FuGENE HD Transfection Reagent. The Jurkat T cells were labeled using CellTrace CFSE Cell Proliferation Kit for Flow Cytometry according to the manufacturer’s instructions.

Jurkat WT and JinB8 T cells were grown overnight and serum starved for 2 h in RPMI 1640 basal medium containing 0.1% BSA. Jurkat T cells were plated in Labtek II chambers (Rochester) coated with poly-d-lysine. The cells were washed gently three times with cold PBS. The cells were fixed with 4% paraformaldehyde for 15 min at room temperature (RT) and washed with PBS three times for 5-min intervals. Cells were permeabilized with 0.1% Triton X100 for 2 min and washed immediately with PBS 5 min each. Cells were blocked with 1% BSA in PBS-0.1% Tween 20 for 1 h at RT. The cells were incubated in the mixture of two primary Abs (1:300), mouse anti-human CD47 (B6H12.2) and rabbit against anti-human VEGFR2 (1:500; Cell Signaling), overnight at 4°C. The cells were washed with PBS three times for 5 min each and incubated with the mixture of two secondary Abs (Alexa 488 and Alexa 594) in 1% BSA for 1 h at RT in the dark. The cells were washed with PBS three times for 5 min each and mounted by using Vectashield HardSet with DAPI (Vector Laboratories).

Confocal images were sequentially acquired with Zeiss Zen software on a Zeiss LSM 780 Confocal system (Carl Zeiss) with a Zeiss Observer Z1 inverted microscope and a 25-mW Argon visible laser tuned to 488 nm and a 15-mW laser tuned to 594 nm. A 40× Plan-Apochromat 1.4 NA oil immersion objective was used at various digital zoom settings. Emission signals after sequential excitation were collected, from 490 to 553 nm or from 598 to 696 nm, respectively, using individual photomultipliers. For sensitized emission FRET, an additional FRET channel was collected with 488 nm excitation and 598–696 nm emissions. The same confocal parameters (laser intensity, detector gain) were used for all FRET samples. mCherry alone and GFP alone samples served as FRET negative controls. FRET images were analyzed with the Zeiss Zen FRET macro.

Jurkat and primary T cells were plated overnight in six-well plates with complete RPMI medium at 37°C with 5% CO₂. The next day, the cells were serum starved with 0.1% BSA for 2 h, pretreated with TSP1 (2 nM) for 15 min, and incubated in the presence or absence of VEGF (30 ng/ml) for 5 min. The cells were centrifuged, washed one time with cold PBS, and lysed using Nonidet P-40 (NP-40) buffer. The cells were kept on ice for 5 min and centrifuged at 15,000 rpm for 15 min. The cell supernatant (300 μl) of each lysate was using 5 μl beads to preclear the lysate. A 20-μl volume of Dynabeads was used for immunoprecipitation of each sample. Immunoprecipitation was performed using 2 μl VEGFR2 and Src Abs with 300 μl lysate at 4°C overnight. The beads were washed with NP-40 lysis buffer three times and boiled with 1× SDS buffer for 5 min at 95°C. Total protein was quantified by using the bicinchoninic acid method. Western blot was performed using Bis-Tris 4–12% gels with NuPAGE buffer at 150 V for 45 min. The gels were transferred to polyvinylidene difluoride membranes at 45 V for 1 h. The membrane was blocked with 3% BSA in Tris-buffered saline with Tween 20 (TBSTw) for at least 1 h. Human anti-CD47 (B6H12), extracellular domain of CD47 (R529), VEGFR2, or SRC Abs (1:1000) were incubated overnight in TBSTw containing 3% BSA. Membranes were washed two times with 1× TBSTw for 10 min each. Anti-rabbit or anti-mouse secondary Abs conjugated to HRP were used diluted 1:5000. Membranes were washed three to five times with TBSTw for 15 min each, and bands were detected using chemiluminescence. Protein loading was normalized using β-actin Ab or immunoblotted for total VEGFR2, human CD47 B6H12, extracellular CD47 cleavage R259, or Src Abs as indicated.

Total RNA was prepared using whole spleen from WT and CD47-null mice or purified subsets of splenocytes using the TRIzol method. Jurkat cells were plated on anti-CD3 (1 μg/ml)–coated plates and treated with TSP1 in the presence or absence of VEGF for 3–6 h in 2% serum. cDNA was made using the Superscript III kit from Invitrogen, and real-time PCR was performed using HPRT1 and B2M primers for normalization. The following primers were used: murine VEGF (5′-GCGGATCAAACCTCACCAAA-3′; 5′-TTCACATCGGCTGTGCTGTAGGA-3′), murine VEGFR2 (5′-TCTGTCGCCTCGGAAGGCTCC-3′; 5′-TCGTCGCTGGAGTACACGGTG-3′), human TNF-α (5′-AGAAGGGTGACCGACTCAGC-3′; 5′-TCCCAAAGTAGACCTGCCCAG-3′), human CD69 (5′-CTGGTCACCCATGGAAGTGG-3′; 5′-ACTTTCCATGCTGCTGACCT-3′), human CD40 (5′-GGCAACAATCCATTCACTTGGGAGG-3′; 5′-CCAAAGGACGTGAAGCCAGTGCCA-3′), human TOLLIP (5′-TGTTCCCCAACATGGACCAGGAGGT-3′; 5′-GCAGGGAGTTGATGGCGGCA-3′), human IL-10A (5′-AGGTGGTCTCCTGGGCAGCT-3′; 5′-CTGCAGGCTAGAGATGAGGGGCA-3′), human CCL3 (5′-TGTCATCTTCCTAACCAAGCGAAG-3′; 5′-AGGCACTCAGCTCCAGGTCG-3′), human IL-2 (5′-CCTGCCACAATGTACAGGATGCA-3′; 5′-GGTGCACTGTTTGTGACAAGTGC-3′), murine IL-2 (5′-TTGGACCTCTGCGGCATGTTC-3′; 5′- CAAATGTGTTGTCAGAGCCCT-3′), murine CD69 (5′-ACGCTCTTGTTCTGAAGATGCTGC-3′; 5′-TCCAATGTTCCAGTTCACCA -3′), and murine VEGFR1 (5′-TGAGGAGCTTTCACCGAACTC-3′; 5′-TGGTCTCAGTCCAGGTGAACC-3′). Real-time PCR was performed using SYBR Green (Thermo Scientific) on an MJ Research Opticon I instrument (Bio-Rad) with the following amplification program: 95°C for 15 min, followed by 40 cycles of 95°C for 15 s, 58°C for 20 s, 72°C for 25 s, and 72°C for 1 min. Melting curves were performed for each product from 30 to 95°C, reading every 0.5°C with a 6-s dwell time. Fold change in mRNA expression was calculated by normalizing to HPRT1 or B2M mRNA levels, which differed in WT and CD47-null samples with equal input of RNA to generate the first-strand template.

Cell proliferation assay was performed using CellTiter 96 AQueous Non-Radioactive Cell Proliferation Assay (MTS) according to the manufacturer’s instructions (Promega). Jurkat T and JinB8 cells (5 × 10⁴) were plated in triplicate using 96-well plates. The cells were treated either with VEGF (30 ng/ml) or TSP1 (2 nM) or in combination. In other experiments, Jurkat and JinB8 cells were treated with VEGF (30 ng/ml) or VEGF blocking Ab (10 ng/ml) alone or in combination. The cells were cultured for 72 h. Signal at 0 h was subtracted from the 72-h signal to calculate net proliferation, and results are expressed as a percentage of untreated control cell proliferation.

Pan T cell, CD4⁺, and CD8⁺ isolation kits were used to isolate T cells from WT and CD47^−/− spleen according to the manufacturer’s instructions (Miltenyi Biotech). The primary cells were prestimulated using either anti-CD3 (1 μg/ml) or in combination of anti-CD3 (2 μg/ml) and anti-CD28 (5 μg/ml) as indicated. Equal numbers (∼2.5 × 10⁴) of primary WT and CD47^−/− cells were plated on uncoated 96-well plates in triplicate. Cell proliferation was assessed using the MTS assay after treatment with TSP1 (1 μg/ml), VEGF (30 ng/ml), or SU1498 (5 μg/ml) for 72 h as described earlier.

Jurkat and JinB8 T cells were plated onto anti-CD28– (3 μg/ml) and anti-CD3 (1 μg/ml)–coated plates and treated with VEGF blocking Ab or TSP1 in the presence or absence of VEGF for 24 h in 2% serum. Samples of the conditioned media were analyzed for secreted IL-2, TNF-α, and MIPα concentrations (NCI-Leidos Biomedical Research/FNLCR). Duplicate experimental samples were analyzed in duplicate.

Tissue from WT or CD47^−/− mice was harvested and paraffin embedded. Slides were deparaffinized in Histoclear (National Diagnostics) three times for 5 min and rehydrated in graded alcohol (100, 95, and 70%). Ag retrieval was performed using a Coplin jar filled with Target Retrieval Solution, pH 6.10 (Dako Corporation) in a microwave for 5 min at high power, 15 min at low power, and cool down for 20 min. After retrieval, slides were washed in PBS twice for 5 min. Endogenous peroxidase activity was quenched by 0.3% H₂O₂ treatment (Dako Corporation). Nonspecific binding was reduced using Protein Block Serum-Free (Dako Corporation) for 10 min. The slides were incubated with VEGFR2 Ab (Cell Signaling Technologies) 1:100 overnight at 4°C. Slides were then incubated with Streptavidin-biotin (Dako LSBA+ Kit, HRP). 3, 3-Diaminobenzidine solution (Dako Corporation) was used as chromogen for 5 min, and hematoxylin was used for counterstaining. Negative control slides omitted the primary Ab. Images were evaluated using a Nikon Eclipse E1000 microscope equipped with a microcolor camera (RGB-MS-C). A computer-assisted counting technique with a grid filter to select cells was used to quantify the immunohistochemical staining of VEGFR2 using ImageJ (v. 1.42q National Institutes of Health).

Jurkat and JinB8 T cells were plated on anti-CD3 (1 μg/ml)–coated plates for 6 h or on anti-CD3 + anti-CD28 for 24 h, and harvested by centrifugation. For analysis of cell-surface proteins, cells were stained with fluorophore-conjugated anti-CD69, anti-VEGFR1, anti-VEGFR2, anti-VEGFR3, and anti-CD47 in separate tubes for 30 min at room temperature in FACS buffer (HBSS containing 1% FBS), washed twice with same buffer, and fixed in 1% paraformaldehyde. Isotype control Abs were used to validate flow cytometric results. Flow cytometric acquisition was performed on an LSR II cytometer with FACSDiva software (BD Biosciences). Flow cytometry data were analyzed by using FlowJo software (Tree Star).

The p values for VEGFR2 and Src phosphorylation, cell proliferation, flow cytometry, and cell migration assays were calculated using t tests. For all real-time PCR experiments, two-factor ANOVA with replicates was used. A p value <0.05 was considered significant.

To investigate the potential interaction between endogenous CD47 and VEGFR2 in T cells, we first performed confocal immunofluorescence analysis using Jurkat T lymphoma cells. Endogenous CD47 (red) and VEGFR2 (green) were partially colocalized in untreated Jurkat T cells, suggesting a physical interaction (Fig. 1A). Interaction between CD47 and VEGFR2 was confirmed by coimmunoprecipitation and FRET analysis (Fig. 1B–D). TSP1 disrupts FRET between VEGFR2 and CD47 in endothelial cells (23), and a corresponding decrease in FRET efficiency was observed in Jurkat T cells treated with TSP1 but did not reach significance (Fig. 1B).

To further examine this interaction, we lysed WT and a CD47-deficient Jurkat T cell mutant (JinB8) using NP-40 and immunoprecipitated them using a VEGFR2 Ab. CD47 was detected by Western blotting using two Abs specific for the extracellular domain of CD47 (B6H12 and R529; Fig. 1C). Both Abs detected the 50- to 55-kDa N-glycosylated and >200-kDa proteoglycan isoforms of CD47 and a previously described 36-kDa proteolytic fragment of CD47 (28, 35) in the immunoprecipitates from WT Jurkat cells that were absent in the corresponding immunoprecipitates from JinB8 cells. The CD47 Ab B6H12 detected several nonspecific bands in VEGFR2 immunoprecipitates from JinB8 cells that were not detected by the CD47 Ab R529. Treatment with VEGF did not significantly alter coimmunoprecipitation of CD47 and VEGFR2, but treatment with TSP1 in the presence of VEGF selectively inhibited association of the two smaller isoforms, but not the proteoglycan isoform of CD47 with VEGFR2 (Fig. 1D). This is consistent with disruption of the VEGFR2/CD47 complex by TSP1 plus VEGF in endothelial cells but suggests that the proteoglycan isoform of CD47 remains associated with VEGFR2 in T cells (23). VEGFR2 can bind directly to heparan sulfate (36), and we propose that the proteoglycan isoform of CD47 expressed by T cells binds with high affinity to VEGFR2 and is therefore more resistant to TSP1 inhibition than VEGFR2 binding to CD47 in endothelial cells.

Reverse immunoprecipitation using CD47 Ab and detection with anti-VEGFR2 confirmed that VEGFR2 interacts with CD47 (Fig. 1E, upper panels). VEGFR2 signaling in endothelial cells recruits and activates Src (37), and coimmunoprecipitation demonstrated that VEGFR2 is associated with Src in Jurkat T cells (Supplemental Fig. 1A). Src was also detected in CD47 immunoprecipitates from WT but not CD47-deficient JinB8 cells, indicating that Src is a constitutive component of this membrane complex (Fig. 1E, lower panels).

TSP1 binding to CD47 suppresses VEGF-induced VEGFR2 phosphorylation in endothelial cells (23). Similar results were obtained in Jurkat T cells treated with 2 nM TSP1 (Fig. 2A, 2C). Addition of VEGF induced VEGFR2 phosphorylation at Y¹¹⁷⁵ in Jurkat cells (p ≤ 0.0001), and pretreatment with 0.6 or 2 nM TSP1 for 15 min prevented this response to VEGF (p ≤ 0.04 and 0.006, respectively). The data in Fig. 2A normalized to actin was confirmed by normalizing to total VEGFR2 (Supplemental Fig. 1B), indicating that TSP1 specifically inhibits VEGFR2 phosphorylation rather than merely reducing VEGFR2 expression. In contrast, TSP1 did not inhibit tyrosine phosphorylation of VEGFR2 at Y⁹⁶¹ (Supplemental Fig. 1C). The ability of TSP1 to modulate VEGFR2 phosphorylation required CD47 because parallel experiments performed using CD47-deficient JinB8 cells showed higher basal phosphorylation of VEGFR2 and no significant effects of VEGF or TSP1 on this signal (Fig. 2C, 2D). Notably, levels of total VEGFR2 appeared to be higher in the JinB8 cells (Fig. 2C).

Primary CD3⁺ T cells isolated from spleens of WT and CD47-null mice confirmed the Jurkat T cell data. VEGF induced VEGFR2 Y¹¹⁷⁵ phosphorylation, and TSP1 inhibited this response in WT but not in CD47-null murine T cells (Fig. 2E, 2F). An elevation in total VEGFR2 was also observed in the CD47-null murine T cells.

As shown in Fig. 1E, Src interacts with the CD47/VEGFR2 complex in T cells, and VEGF-VEGFR2 signaling induces Src Y⁴¹⁶ phosphorylation in endothelial cells (37). Treatment with TSP1 inhibited VEGF-induced Src Y⁴¹⁶ phosphorylation in primary murine T cells and Jurkat T cells (Fig. 2B, 2F). At a concentration of 0.2 nM, which approximates circulating TSP1 levels in plasma, TSP1 significantly inhibited VEGF-induced Src phosphorylation at Y⁴¹⁶ (p ≤ 0.02). Notably, higher concentrations of TSP1 tended to show less Src inhibition, which may result from positive effects of TSP1 signaling through other receptors (p ≤ 0.07; Fig. 2B).

To exclude a requirement for additional TSP1 receptors in regulating VEGFR2, we examined whether a known CD47 binding peptide derived from TSP1 (7N3) is sufficient to inhibit VEGFR2 Y¹¹⁷⁵ phosphorylation. Treatment of Jurkat T cells with 0.2 μM 7N3 inhibited VEGF-induced VEGFR2 phosphorylation, but the control peptide 604 had no effect (Supplemental Fig. 1D, 1E), demonstrating that engaging CD47 is sufficient for this response.

The elevated VEGFR2 protein expression that we observed in CD47-deficient Jurkat T cells correlated with elevated VEGFR2 mRNA levels in these cells (Fig. 3A). However, flow cytometry showed only a modest increase in cell-surface levels of VEGFR2 in JinB8 relative to WT Jurkat cells (Fig. 3B, 3C), which was statistically significant. VEGFR1 was also expressed and showed no difference in surface expression between WT and JinB8 cells (Fig. 3D). Surface expression of VEGFR3 was slightly elevated in JinB8 cells (Fig. 3E). Because a large fraction of the VEGFR2 in endothelial cells is intracellular and recycles to the cell surface (38), we also compared flow cytometry analysis using permeabilized and nonpermeabilized cells (Fig. 3F–I). Consistent with the Western blots, JinB8 cells showed higher total VEGFR2 expression than WT cells.

To determine whether CD47 regulates VEGF and VEGFR2 expression in vivo, we first analyzed mRNAs isolated from WT and CD47-null mouse spleens immediately after euthanasia (Fig. 4A). VEGF and VEGFR2 mRNA levels were ∼13- and 7-fold higher, respectively, in CD47-null spleens compared with WT spleens. In contrast, VEGFR1 mRNA was lower in CD47-null spleens relative to WT spleens. To examine the distribution of the enhanced VEGFR2 expression in the spleen, we performed immunohistochemical analysis of WT and CD47-null spleens. In WT spleen, aside from highlighting the vasculature (weak membrane staining), VEGFR2 was strongly expressed (membrane and cytoplasmic staining) in some of the periarteriolar lymphoid sheaths, a T cell–rich area. In addition, occasional positive cells were distributed through the red pulp. In contrast, CD47-null spleen exhibited a more diffuse pattern of VEGFR2⁺ cells within the periarteriolar lymphoid sheath, as well as a greater number of VEGFR2⁺ cells within follicles and red pulp (Fig. 4B). Quantitative analysis of tissue sections (n = 6) revealed that CD47-null mouse spleens have significantly more VEGFR2⁺ cells as compared with WT (p ≤ 0.0003; Fig. 4C). Thus, CD47 regulates VEGF production and potentially the responsiveness of splenic T cells to VEGF mediated by VEGFR2.

To define which splenic cell types exhibit CD47-dependent VEGF and VEGFR2 expression, we examined mRNA levels in several purified populations isolated from WT and CD47-null spleens. Consistent with the Jurkat cell data, CD4⁺ T cells from CD47-null mice had increased mRNA levels for VEGF and VEGFR2 (Fig. 4D, 4E). Interestingly, this was not the case for purified CD8⁺ T cells, because VEGF and VEGFR2 were expressed 13- and 8-fold less in CD47-null CD8⁺ T cells, respectively. CD19⁺ B cells from CD47-null spleen also had higher VEGFR2 expression, but CD47-null CD11b⁺ monocytes had lower VEGFR2 expression than WT monocytes (Fig. 4F). The differential regulation of VEGF by CD47 in CD4⁺ versus CD8⁺ cells may be c-Myc dependent because we recently showed differences in c-Myc expression in WT versus CD47-null CD4⁺ and CD8⁺ T cells (39), and VEGF is a known target of c-Myc (40, 41).

We examined several functional responses of T cells to VEGF to study cross talk between TSP1/CD47 and VEGF signaling. Preliminary experiments using 1–10% FBS and 5–50 ng/ml VEGF established that 1% FBS with 30 ng/ml VEGF gave optimal stimulation of Jurkat T cell migration in a modified Boyden Chamber assay. VEGF treatment caused a significant 2.5-fold increase in cell migration as compared with a control using 1% BSA in the lower chamber (p < 0.01). TSP1 added to the upper chamber at 1 or 2 nM significantly inhibited VEGF-induced cell migration induced by VEGF (p ≤ 0.007 and p ≤ 0.006, respectively; Fig. 5A).

Proliferation of Jurkat cells in RPMI medium containing 2% FBS was assessed at 72 h. VEGF treatment inhibited proliferation of WT but stimulated proliferation of CD47-deficient cells (Fig. 5B). Both activities of VEGF were blocked by a VEGF function-blocking Ab. Therefore, CD47 is necessary for VEGF to inhibit T cell proliferation, but VEGF can also stimulate T cell proliferation via a CD47-independent mechanism. To confirm this result in primary T cells, we assessed proliferation of CD4⁺ T cells isolated from WT and CD47-null mice in the presence of 10% FBS. The CD47-null T cells proliferated significantly more than WT CD4⁺ T cells at 6 d (p ≤ 0.001; Fig. 5C). Because CD47-null CD4⁺ T cells express more VEGFR2, we tested its function in regulating proliferation using the VEGFR2 kinase inhibitor SU1498. We observed a significant decrease in CD47-null cell proliferation (p ≤ 0.00001), whereas WT CD4⁺ T cells at day 6 showed only marginal inhibition (p ≥ 0.08; Fig. 5D). Therefore, autocrine production of VEGF can stimulate CD4⁺ T cell proliferation in a VEGFR2-dependent but CD47-independent manner.

To further confirm that inhibitory VEGFR2 signaling for proliferation is CD47 dependent, Jurkat T cells were transiently transfected with VEGFR2-mCherry (WT VEGFR2) or VEGFR2-mCherry mutated to prevent phosphorylation at residue Y¹¹⁷⁵ (VEGFR2-Y^1175F), and proliferation was assessed by CSFE dilution. At days 1 and 3, proliferation of cells overexpressing WT VEGFR2 was inhibited relative to cells overexpressing the phosphorylation-deficient mutant. This occurred despite more persistent expression for the mutant VEGFR2 construct relative to that of the WT construct (Fig. 6A, 6B). Therefore, expression of functional VEGFR2 specifically limits proliferation of T cells that express CD47.

TSP1 signaling via CD47 inhibits TCR signaling, including induction of the early activation marker CD69 (28, 33, 42). Consistent with the previous report that VEGF inhibits T cell activation (19), VEGF treatment resulted in a 4-fold reduction in CD69 mRNA expression induced by activation of WT Jurkat T cells in the presence of anti-CD3+CD28 for 24 h (Fig. 7A). Addition of TSP1 reversed the inhibition of CD69 mRNA induction by VEGF in cells, as did the VEGF blocking Ab, which increased CD69 mRNA induction 3-fold. The latter result suggested that endogenous VEGF production limits activation.

CD47-deficient JinB8 cells had substantially lower basal CD69 mRNA expression, but anti-CD3+CD28 Abs induced an 18-fold increase in CD69 mRNA (Fig. 7A). The weaker induction of CD69 in CD47-deficient cells has been reported previously and is associated with defective integrin activation (43, 44). In contrast with WT cells, addition of VEGF did not inhibit this induction. Addition of the VEGF blocking Ab resulted in a 2-fold decrease, suggesting that VEGF has a positive effect on T cell activation in the absence of CD47.

The differential effects of VEGF on CD69 induction in the presence and absence of CD47 were confirmed at the level of cell-surface protein expression (Fig. 7B, 7C). VEGF inhibited both the percentage of CD69⁺ WT Jurkat T cells induced by anti-CD3+CD28 treatment for 24 h and their mean level of CD69 expression. In contrast, CD47-deficient cells were resistant to VEGF inhibition of CD69 induction by anti-CD3+CD28 (Fig. 7C). To confirm the role for endogenous VEGF suggested by the VEGF Ab blocking data, we examined the effect of a VEGF blocking Ab on CD69 induction by anti-CD3+CD28 at 24 h. Consistent with the mRNA data in Fig. 7A, the VEGF blocking Ab neutralized the inhibitory activity of VEGF in WT cells but did not alter CD69 induction in CD47-deficient cells (Fig. 7D, 7E).

A similar differential sensitivity to VEGF inhibition was seen in the absence of CD28 costimulation. Induction of CD69 mRNA by anti-CD3 alone was partially inhibited in WT but not in CD47-deficient Jurkat T cells (Supplemental Fig. 2A–C). Under these conditions, TSP1 did not significantly reverse the inhibition by VEGF. Cell-surface CD69 protein induction by anti-CD3 paralleled the mRNA changes and the results obtained with CD28 costimulation (Supplemental Fig. 2C, 2D). Individually, VEGF and TSP1 inhibited anti-CD3–stimulated CD69 induction. However, combining TSP1 and VEGF yielded no inhibition, suggesting antagonism between TSP1 and VEGF consistent with that observed at the mRNA level. As expected, anti-CD3 stimulation resulted in a weaker induction of CD69 protein expression in CD47-deficient cells, and VEGF and TSP1 alone or in combination did not inhibit this induction. Therefore, the presence of CD47 plays a positive role in T cell activation as reported previously and presumably mediated by integrin activation (44), but the inhibitory effects of VEGF and TSP1 on T cell activation also require CD47.

A similar partial inhibition by VEGF of anti-CD3–stimulated CD69 mRNA induction was observed using WT primary murine T cells (Supplemental Fig. 2E). CD47-null murine T cells were more resistant to anti-CD3 for inducing CD69 mRNA under the same conditions, but addition of VEGF in the presence of anti-CD3 increased CD69 mRNA expression (Supplemental Fig. 2F), further supporting a positive effect of VEGF on activation of T cells lacking CD47.

CD47-dependent effects of VEGF on T cell activation were not limited to CD69 induction. A similar dependence on CD47 was observed for modulation by VEGF of additional activation-dependent cytokines and a chemokine. CD28 costimulation was required for induction of IL-2 in CD47-null murine T cells, so we used these conditions to compare with WT T cell responses (Fig. 7F). Anti-CD3⁺CD28 induced a 14-fold increase in IL-2 mRNA, and VEGF inhibited this induction to baseline levels. The VEGFR2 inhibitor SU1498 blocked inhibition by VEGF, confirming that this inhibition is mediated by VEGFR2. Although TSP1 is known to inhibit IL-2 induction (43), addition of TSP1 in the presence of VEGF resulted in a 3-fold increase in IL-2 mRNA, providing further evidence of antagonism between TSP1 and VEGF to regulate T cell activation responses. Induction of IL-2 mRNA in CD47-null T cells was not inhibited by exogenous VEGF, but SU1498 reduced induction 2-fold, which is again consistent with endogenous VEGF being an autocrine activator in the absence of CD47. As expected, TSP1 did not alter IL-2 induction in the CD47-null cells. Thus, VEGF signaling via VEGFR2 negatively regulates IL-2 induction in primary murine T cells in a CD47-dependent manner, but autocrine production of VEGF signaling positively regulates IL-2 induction in a CD47-independent manner.

VEGF partially inhibited the induction of TNF-α mRNA stimulated by anti-CD3 in Jurkat T cells (Supplemental Fig. 3A). Addition of TSP1 prevented the inhibition of TNF-α induction by VEGF, providing further evidence of functional antagonism between TSP1 and VEGF. TNF-α was not induced above basal levels by anti-CD3 in the CD47-deficient T cell line, but VEGF added alone or in the presence of anti-CD3 increased TNF-α mRNA expression. Notably, TSP1 partially inhibited the positive effect of VEGF on TNF-α mRNA in the CD47-deficient cells, indicating that antagonism between TSP1 and VEGF can be CD47 independent.

Induction of the TCR-dependent chemokine IL-10A, CD40, and Toll interacting protein (TOLLIP) (42) mRNAs by anti-CD3 was also inhibited by VEGF in Jurkat cells (Supplemental Fig. 3B–D). Consistent with the TNF-α data, addition of TSP1 in the presence of VEGF restored induction of IL-10A and TOLLIP mRNAs in WT Jurkat T cells. Anti-CD3 failed to induce the expression of CD40, IL-10A, and TOLLIP mRNAs in CD47-deficient cells. However, VEGF induced expression of these genes in a CD47-independent manner.

We further investigated VEGF regulation of mRNA and protein expression induced with CD28 costimulation for 24 h in WT and CD47-deficient Jurkat T cells. IL-2, TNF-α, and CCL3 (MIP-1α) mRNAs were strongly induced in WT cells under these conditions, and coincubation with VEGF in each case significantly inhibited induction (Fig. 8A, 8C, 8E). Specificity was confirmed by reversal of the VEGF inhibition by a VEGF blocking Ab. ELISA assays for TNF-α and IL-2 secretion at 24 h showed parallel VEGF inhibition of activation-induced secretion and reversal by the VEGF blocking Ab. Although induction of CCL3 mRNA expression was inhibited by VEGF, no significant inhibition of CCL3 protein induction by VEGF was observed.

The CD47-deficient JinB8 T cells showed less induction of TNF-α mRNA than in WT, but TNF-α protein secretion was strongly induced, and the mRNA and protein responses were not altered by VEGF (Fig. 8A, 8B). IL-2 induction was confirmed at the mRNA and protein levels, and was similarly resistant to VEGF inhibition (Fig. 8C, 8D). CCL3 (MIP-1α) showed a moderate induction of its mRNA and strong induction of CCL3 protein secretion that was similarly resistant to VEGF inhibition (Fig. 8E, 8F). Therefore, CD47 is necessary for VEGF to inhibit T cell activation assessed by several methods in both human Jurkat and murine primary T cells.

This study demonstrates that the colocalization and functional cross talk between VEGFR2 and CD47 that was first identified in endothelial cells (23) also occurs in human and murine T cells. VEGF and the CD47 ligand TSP1 individually inhibit TCR signaling, but CD47-dependent cross talk between TSP1 and VEGF can override some of these direct inhibitory signals to result in increased activation responses (Supplemental Fig. 4). This cross talk between CD47 and VEGFR2 has functional consequences in the regulation of T cell migration and proliferation, and controls the induction of TCR activation-dependent receptors, cytokines, and chemokines including CD69, IL-2, CCL3, and TNF-α. We also found that expression of VEGF and VEGFR2 by CD4⁺ T cells is limited by inhibitory CD47 signaling, and CD47 thereby controls an autocrine feedback loop in T cells involving VEGF and VEGFR2. In WT cells that express CD47, this feedback by VEGF is inhibitory. However, in human Jurkat and murine CD4⁺ T cells lacking CD47, most inhibitory effects of VEGFR2 signaling are lost, and VEGF becomes a positive regulator of proliferation and some TCR signaling responses including the induction of IL-2. TSP1, which is known to be produced in a regulated fashion by T cells (45, 46), further modulates VEGF signaling by its interactions with CD47.

We describe a novel role of TSP1 and its receptor CD47 in VEGF signaling in T cells that provides some insight into some of the discrepancies related to the role of VEGF as an immunomodulator (18, 19). In primary murine CD4⁺ T cells and human Jurkat T cells, CD47 functions as a negative regulator of the VEGF-VEGFR2 pathway at the level of receptor phosphorylation and downstream phosphorylation of Src. Binding of TSP1 to CD47 further inhibits VEGFR2 signaling in these cells. Although TSP1 is known to modulate distinct T cell functions by engaging α4β1 and CD47 in T cells (27), this is the first report, to our knowledge, to implicate VEGFR2 in the ability of TSP1 to modulate anti-CD3–mediated T cell activation. The fact that CD47 differentially regulates VEGF and VEGFR2 expression in CD4⁺ T versus CD8⁺ T cells may also account for some conflicting published evidence regarding whether VEGF is an inhibitor or stimulator of T cell function.

CD47 signaling modulates VEGFR2 and Src phosphorylation in T cells. Treatment of CD47-expressing human Jurkat and primary murine T cells with TSP1 inhibits VEGFR2 phosphorylation at Y¹¹⁷⁵ and Src at Y⁴¹⁶ in a CD47-dependent manner. This is consistent with an earlier report where cross-linking CD47 on T cells using two Abs stimulated tyrosine phosphorylation of proteins including the Src family protein tyrosine kinase p56(Lck) (47). Because CD47 does not have intrinsic tyrosine kinase activity, these downstream phosphorylation events may be secondary to perturbing the lateral association of CD47 with VEGFR2, which may prevent recruitment of Src kinases including Lck, thereby ablating the signal from VEGF binding to VEGFR2. Rapid phosphorylation of VEGFR2 at Y¹¹⁷⁵ mediates proliferative signals in vascular cells, but our findings support the conclusion of Ziogas et al. (19) that VEGFR2 phosphorylation inhibits T cell proliferation. We demonstrated by site-directed mutagenesis that phosphorylation of VEGFR2 at Y¹¹⁷⁵ is required for this antiproliferative activity. VEGF also stimulates chemotaxis of activated T cells (43), and TSP1 signaling via CD47 inhibits VEGF-stimulated T cell migration, as we previously reported for migration of vascular cells from muscle explants stimulated by VEGF or its signaling mediator NO (48).

We further demonstrate that CD47 signaling negatively regulates T cell expression of VEGF and VEGFR2. Thus, CD47-deficient murine B cells and CD4⁺ T cells, but not monocytes and CD8⁺ T cells, exhibit elevated expression of VEGF and VEGFR2. Expression of VEGF by T cells antagonizes TCR signaling, and thus inhibits T cell activation in a CD47-dependent manner as demonstrated by the expression of the TCR activation marker CD69 and several cytokines. Recent findings also revealed that VEGF expression and VEGFR2 activation suppresses the host immune response by inhibiting CD3-mediated activation of T cells, thereby protecting tumor cells (19). In addition to the autocrine role of VEGF, CD47-null T and B cells may stimulate angiogenesis by releasing VEGF. These data may also provide insights for understanding the recruitment and regulation of tumor-associated/infiltrating lymphocytes by TSP1 in the tumor microenvironment.

Current clinical antiangiogenic therapy uses the VEGF Ab bevacizumab to inhibit VEGF binding to VEGFR2 or VEGFR2 kinase inhibitors to block its signaling. In addition to the expected inhibition of tumor angiogenesis, blocking VEGF signaling is known to cause negative side effects including hypertension and thrombosis (22). The data presented in this article and by others demonstrate that VEGFR2 antagonists can also modulate T cell immunity in the tumor microenvironment, and that T cell proliferation, migration, and activation are directly modulated by VEGF. Thus, we predict that bevacizumab and VEGFR2 inhibitory drugs can also act as enhancers of antitumor T cell immunity. Based on our finding that disrupting the cross talk between CD47 with VEGFR2 increases the ability of T cells to proliferate and be activated in response to TCR stimulation, we propose that therapeutic targeting of CD47 on T cells could enhance antitumor T cell immunity.

We thank Dr. Silvio Gutkind for providing SU1498 and Dr. Laura Maile for the R529 Ab.

The authors have no financial conflicts of interest.