Wild-type (WT) mice (C57BL/6 from Harlan Olac Ltd Bicester, UK), IFNγ^-/- mice (kind gift from Simon Clare, Wellcome Trust Sanger Institute, Cambridge, UK), ROSA-enhanced yellow fluorescent protein (EYFP) mice (8) and heterozygous mice expressing a human tissue factor pathway inhibitor (TFPI) fusion protein (18) under control of a CD31 promoter (13), were bred and maintained at King’s College London. All genetically modified animals have been maintained for more than 10 generations on a WT background. All procedures were approved by the UK Home Office.

Vessels were embedded in optimum cutting temperature compound (OCT) (VWR International, Dorset, UK) before cross-sectioning and staining with Accustain Elastin Stain kit (Sigma). Morphometric analysis was done on an Olympus U-ULH microscope (Olympus Optical Co Ltd, Tokyo, Japan). Areas were determined with Image-Pro Plus TM software version 4.0 (Media Cybernetics, Silver Spring, MD, USA). Three random sections from each of six wire-injured arteries were examined by an investigator blinded to the identity of the sections. The mean value from each vessel was used to prepare figures and statistics.

For immunofluorescence (IF) analysis, OCT-embedded vessels were cut into 5 µm sections prior to fixing in methanol for 60 minutes at –20°C. After immersion for 30 minutes in 1% BSA (Sigma)-PBS, frozen sections were incubated overnight at 4°C with combinations of the following antibodies: rabbit polyclonal collagen 1, rat anti-mouse F4/80, rabbit polyclonal F4/80, rat anti-mouse collagen 1 and hamster anti-mouse IFNγR (all from Abcam), mouse anti-human αSMA (Sigma), rat anti-mouse monoclonal IFNγ (Invitrogen), goat anti-angiopoietin-2 (Santa Cruz biotechnology Inc.), and rabbit anti-mouse TF (American Diagnostica.inc. USA). All stained sections were mounted in Vectashield medium with DAPI (Vector Laboratories). Sections were analysed by a Leica DM-IRBE confocal microscope (Leica, Wetzlar, Germany) equipped with Leica digital camera AG and a confocal laser scanning system with excitation lines at 405, 488, 543, and 560 nm at magnifications 10x/0.40CS and 20x/0.70IMM (Leica, Planapo, Wetzlar, Germany). Images were processed using the Leica-TCS-NT software associated with the Leica confocal microscope. Three random sections from each of six wire-injured arteries were examined by an investigator blinded to the identity of the sections. The mean value from each vessel was used to prepare figures and statistics. Isotype-matched antibodies were used in initial sections to confirm the specificity of staining of individual antibodies (data not shown).

CD34+ cells (2x10⁴ cells per well in a 96-well plate) were washed and suspended in 200μl Dulbecco modified Eagle medium (DMEM; Sigma-Aldrich MO, USA) containing 0-320 nM human Factor X (FX) and 0- 20 nM human factor VIIa (FVIIa) at 37°C. In some experiments, cells were first incubated with 100μg/ml of rat anti-mouse TF (20). After 20 minutes, aliquots or the reaction mixture were transferred into Tris-EDTA buffer with the chromogenic substrate S-2222 (Chromagenix, Milan, Italy) to assess FXa generation. Absorbance at 405 nm was converted to give the FXa concentration after comparison to defined standard controls. A similar assay was used to assess thrombin generation, using either 10 nM FX and 6 nM FVIIa, OR 10nM FXa, with added pre-prepared 6 nM human factor Va (FVa) and 0-10nM human prothrombin (all from Enzyme Research Laboratories) added in HEPES-buffered saline (Life Technologies, Grand Island, NY). At defined times, aliquots of the reaction mixture were transferred into Tris-EDTA buffer with the chromogenic substrate S-2238 (Chromagenix, Milan, Italy) to assess thrombin generation as for FXa, using absorbance at 405 nm and standard controls to calculate thrombin concentration.

For IFNγ expression, 2x10⁴ CD34⁺ cells were seeded on a round glass coverslip in a 24-well plate and serum starved for 24 hours, before addition of varying concentrations of either FX and FVII, FXa, FVIIa, FX and prothrombin or thrombin (0-100nM) (Enzyme Research Laboratories). The cells were incubated for 5 days in Iscove modified Dulbecco medium (IMDM; Sigma-Aldrich, MO, USA) supplemented with 2% FCS (StemCell Technology, Grenoble, France). IFNγ secretion was measured by ELISA (following manufacturer’s instructions (R&D Systems, Abingdon, UK)). The same kit was used to measure plasma concentrations.

PAR-induced cell signalling was measured in serum starved cells treated for 30 minutes with 0 - 80 µM of either PAR-1 (FLLRN), PAR-2 (FSLLRY-amide) or PAR-4 (Trans-cinnamoyl YPGKF-NH2) antagonists (Peptides International, Louisville, KY) before the cells were stimulated by FVIIa+FX or FVIIa+FX+FII at indicated doses. Alternatively, cells were stimulated with 0 - 100 µM of PAR-1 (SFLLR-amide), PAR-2 (2-Furoyl-LIGRLO-Amide) or PAR-4 (GYPGKF) agonists (Peptides International) at the indicated doses. In some assays, cells were first treated with siRNA for angiopoietin 2 or the fluorescein conjugated control siRNA, using 2 x 10⁵ CD34+ cells per well were seeded in a six well tissue culture plate (see below).

50 pmols siRNA for angiopoietin 2 (sc-39294) and fluorescein conjugated controls (sc-36869) (Santa Cruz biotechnology Inc. Texas 75220, USA) were used to transiently transfect 2 x 10⁵ CD34+ cells per well in a six-well plate using a siRNA transfection system supplied by Santa Cruz Biotechnology Inc. 24 hours later, cells were washed twice in serum-free medium and some were then incubated with 10nM thrombin for further incubation with the culture medium containing 2% FCS. Analysis of cell phenotype or supernatants was performed 48 hours later. For in vivo use, 5 wells of transfected cells were combined to treat a single mouse.

CD34+ cells spread onto a glass coverslip (VWR International, Leuven, Belgium) were fixed for 10 minutes with methanol at –20°C, before incubation at RT for 60 minutes with one or more of the following antibodies: mouse anti–human α-SMA conjugated with Cy3 (Sigma-Aldrich, St Louis, MO), rat anti-mouse monoclonal IFNγ (Invitrogen), rat anti-mouse CD31 (BD) and rabbit anti- collagen-1, rabbit anti-Angiopoietin-2, and rabbit anti-TIE-2 (all from Abcam). Second layer staining was with a goat anti-rabbit IgG-FITC, goat anti-rat IgG-FITC or a rabbit anti-goat IgG-FITC (all from Sigma-Aldrich). Stained cells were mounted in Vectashield medium with DAPI (Vector Laboratories), and analysed by a Leica DMIRBE confocal microscope (Leica, Wetzlar, Germany) as above.

Statistical analysis was performed with GraphPad Prism software. Mann-Whitney or T test was used for comparison of 2 groups and the Kruskal-Wallis test for ≥3 groups. All data were reported as mean ± SEM. A P value of <0.05 was considered statistically significant.

Wire induced injury is characterised by immediate loss of endothelium (21), followed by platelet deposition (21, 22), non-occlusive luminal thrombosis (23) and the rapid induction of TF expression by medial VSMC (24). Compared to WT BL/6 mice, IFNγ^-/- mice developed little IH when examined 4 weeks post endoluminal injury ( Figure 1A ).

Reciprocal BM reconstitution showed that IFNγ^-/- mice reconstituted with WT BM developed IH to same degree as WT, whereas in WT mice reconstituted with IFNγ^–/- BM the IH that developed was significantly reduced compared to WT controls ( Figures 1A, B ). Plasma levels of IFNγ reflected the source of BM cells ( Table 1 ). These data suggest that IFNγ production is by BM-derived cells and is both necessary and sufficient to mediate IH.

We have previously defined and validated an adoptive transfer model, in which CD34 + cells injected intravenously at the time of vascular injury outcompete endogenously circulating CD34+ cells for uptake at the site of injury (14). The CD34+ cells are obtained from mice that undergo wire-injury 2-4 days earlier, so reflect cells mobilised into the circulation by the injury. Transfer of IFNγ^-/- CD34+ cells into WT mice significantly impaired the development of IH, whereas transfer of WT cells into injured IFNγ^-/- mice was associated with development of IH, although not to the extent see in WT mice ( Figures 1C, D ). As above, plasma levels of IFNγ reflected the source of CD34+ cells ( Table 1 ). These data suggest that CD34+ cells capable of making IFNγ and recruited early after injury contribute to the development of IH.

Immunofluorescence of sectioned arteries from both BM reconstitution and CD34+ transfer experiments confirmed that when intimal cells came from WT, a substantial proportion of them made IFNγ, whereas when intimal cells came from IFNγ-/- mice, the proportion expressing IFNγ was substantially diminished ( Figures 1E, F ). Staining for IFNγ seen in the IFNγ-/- mice was regarded as artefact. All together, these data imply that IFNγ expression by BM-derived, CD34+ cells recruited to the site of injury makes a significant contribution to the subsequent development of IH.

We examined the phenotype of neointimal cells expressing IFNγ on day 5 post-injury, by which time the acute thrombotic changes associated with injury have cleared (9). Approximately 35% of the intimal area was occupied by cells expressing IFNγ ( Figures 2A–D ). Collagen type 1+ and F4/80+ cells each occupied 20-30% of intimal area ( Figures 2A–H ). Equivalent proportions of the intimal area were occupied by cells co-expressing IFNγ with collagen-1 (median 14% [IQR 12.6-18], Figures 2A, B ), and cells co-expressing IFNγ with F4/80 (median 15% [IQR 13.6-18], Figures 2C, D ). Since collagen-1 and F4/80 appeared to be co-expressed by cells occupying approximately 5% of the intimal area ( Supplementary Figure S1 ), we conclude that 25% of the intimal area was occupied by myeloid cells expressing IFNγ and that >70% of the cells expressing IFNγ were therefore myeloid lineage cells.

We have previously shown that development of IH in this model is TF dependent. We confirmed that virtually all collagen-1+ and all F4/80+ cells co-expressed TF ( Figures 2E–H ). In WT controls treated with an isotype matched antibody at the time of injury ( Figures 2I, J ), by day 5 a median of 36% [IQR 33-38] of intimal cells expressed IFNγ. In contrast, in mice treated with an anti-TF antibody ( Figures 2K, L ) this was significantly reduced (median 0.4% [IQR 0.38-0.7] p=0.004) and plasma IFNγ levels were significantly reduced in these mice ( Table 1 ). These data suggested that both IFNγ expression by intimal cells and plasma levels of IFNγ were dependent on TF expression.

To interrogate these data in more detail, we performed several rounds of additional experiments. First, we purified CD34+ cells from the blood of WT mice 2-4 days post-injury. We confirmed these cells expressed TF, as they could convert FX to FXa in a dose-dependent way when FVIIa was present ( Figure 3A ), and prothrombin to thrombin when factors FVIIa, FX and FV were present ( Figure 3B ) and both these were inhibited by an anti-TF antibody ( Figures 3C, D ).

After incubation with FXa +/-prothrombin, IFNγ was released into the medium ( Figures 3E, F ) dose-dependently. IFNγ secretion was partially inhibited by titration of a PAR-2 antagonist, and almost completely inhibited by a PAR-1 (but not PAR-4) antagonist ( Figures 4A, B ), suggesting that both FXa and thrombin acted primarily through PAR-1. In support of this a PAR-1 agonist induced the secretion of twice the amount of IFNγ in a dose and time dependent manner compared to a PAR-2 agonist ( Figures 4C, D ). A PAR-4 agonist induced no IFNγ secretion ( Figure 4E ). At baseline, none of the CD34+ cells expressed IFNγ ( Figures 4F, G ), but it was expressed by 50-60% after incubation with either FXa or thrombin. In addition, in this in vitro system, these manipulations initiated expansion of both F4/80+ cells ( Figure 4H ), from 12 to 40% of the CD34+ cells and collagen-1+ cells ( Figure 4I ), from 35 to 60%. However, the vast majority of IFNγ+ cells were collagen-1+, with <10% co-expressing F4/80. These data indicate that TF on the surface of CD34+ cells, via FXa and thrombin generation, induces IFNγ expression and secretion primarily from collagen-1+ cells, via activation of PAR-1 and to a lesser extent PAR-2.

Next we studied CD34+ cells from CD31-TFPI-Tg mice (13). In the peripheral blood of these mice, circulating CD34+ cells expressing CD31+ co-express a TFPI fusion protein (9, 11, 14). Compared to cells from WT mice, cells from CD31-TFPI-Tg were unable to convert FX to FXa in a dose-dependent manner when FVIIa was present ( Figure 3A ), unless an inhibitory anti-TFPI antibody was titrated in ( Figure 5A ). In accordance, they were unable to convert prothrombin to thrombin when factors FVIIa, FX and FV were present ( Figure 3B ), unless the inhibitory anti-TFPI antibody was present ( Figure 5B ). Incubation with FXa with or without prothrombin failed to induce IFNγ secretion at the same concentrations that worked in WT ( Figures 3E, F ) except when the anti-TFPI antibody was present ( Figures 5C, D ). These data confirm that inhibition of TF on only the fraction of CD31+ myeloid cells, which are predominantly fibrocytes (9, 11, 14), by the transgenic fusion TFPI fusion protein inhibited IFNγ secretion.

Finally we looked at the early intima in wire-injured CD31-TFPI-Tg mice which, as we have previously described, fail to develop progressive IH. There were few detectable IFNγ+ cells in the day 5 intima ( Figures 5E, F ) of mice treated with an isotype control antibody (median 0.6% [IQR 0.5-0.9] and circulating levels of IFNγ were low ( Table 1 ). However, if these animals were treated with the inhibitory anti-hTFPI antibody at the time of injury ( Figures 5G, H ), the proportion of IFNγ+ cells increased significantly (median 23% [IQR 21-27] p=0.004), all of which were TF+ and plasma IFNγ levels approached those seen in WT animals ( Table 1 ).

All these data suggest that intimal IFNγ expressed by newly recruited CD34+ cells is TF-dependent. Moreover, the TF expressed by fibrocytes is primarily responsible for FXa and thrombin generation and these two enzymes stimulate IFNγ release by these cells predominantly through activation of PAR-1, with PAR-2 playing a lesser role.

We have previously shown that angiopoietin-2 secretion by recruited myofibrocytes is necessary for the development of IH, by inducing myofibrocyte proliferation through TIE-2, and promoting CXCL-12 secretion to mediate continuous recruitment of new myofibrocytes from the circulation as well as through induction of angiopoietin-2 expression by non-myofibrocytes (14). In day 5 sections from WT mice, more than 90% of the cells expressing IFNγ co-expressed angiopoietin-2 ( Figures 6A, B ) although only approximately half of the angiopoietin-2 positive cells were IFNγ+.

We adoptively transferred CD34+ cells from ROSA-EYFP mice, which express yellow fluorescent protein in all cells. By day 5, 35% of the cells in the new intima expressed IFNγ and all IFNγ+ cells were EYFP+ ( Figures 6C, D ), indicating that in the adoptive transfer model, all of the early intimal IFNγ+ cells come from cells injected into the circulation. Incubation with either an inhibitory anti-TIE-2 antibody prior to transfer ( Figures 6C–E ) or with siRNA targeted to angiopoietin-2 ( Figures 6F–H ), significantly reduced the intimal area occupied by IFNγ+ cells, compared to controls (in both cases from approx. 20% to <5%, in both cases p=0.004), without influencing the overall recruitment of EYFP+ cells, and substantially reduced levels of plasma IFNγ ( Table 1 ). Thus, inhibition of angiopoietin-2 production by, or signalling through TIE-2 on the recruited CD34+ cells, significantly inhibits the expression of IFNγ by these early intimal cells.

Experiments in vitro with WT CD34+ cells suggested that although approximately 30% of the CD34+ cells isolated on day 3 post injury expressed angiopoietin-2+, none expressed IFNγ ( Figure 6I ). However, after stimulation with thrombin, approximately 60% of cells were IFNγ + ( Figure 6I ), almost all of which co-expressed angiopoietin-2. This was associated with detectable IFNγ in the supernatant ( Figure 6K ). Incubation with an anti-TIE-2 antibody significantly reduced the proportion of IFNγ+ cells ( Figure 6I ) and inhibited IFNγ secretion ( Figure 6K ) without altering the expression of angiopoietin-2. Similar results were seen when the cells were incubated with siRNA against angiopoietin-2 ( Figures 6J, K ), though this was achieved with significant reduction in the expression of angiopoietin-2 ( Figure 6J ). In contrast, inclusion of an inhibitory antibody against IFNγR had no impact on the expression of either angiopoietin-2 or IFNγ ( Figure 6I ). Thus, thrombin induced angiopoietin-2 expression induces IFNγ expression and secretion.

Finally, we confirmed that in day 5 sections from injured IFNγ^-/- mice approximately 25% of the intima was occupied by angiopoietin-2+ cells, almost all of which were TF+. Thus, expression of angiopoietin-2 is upstream of IFNγ ( Figures 6L, M ).

All these data suggest that the angiopoietin-2 secretion induced in CD34+ cells circulating post injury by TF-mediated FXa and thrombin is directly involved in the induction of IFNγ by intimal myeloid cells and that this is critical for the development of IH.