Interaction of fibrin with VE-cadherin and anti-inflammatory effect of fibrin-derived fragments

Authors


Leonid Medved, University of Maryland School of Medicine, Center for Vascular and Inflammatory Diseases, 800 West Baltimore Street, Baltimore, MD 21201, USA.
Tel.: +1 410 706 8065; fax: +1 410 706 8121.
E-mail: Lmedved@som.umaryland.edu; and
Li Zhang, University of Maryland School of Medicine, Center for Vascular and Inflammatory Diseases, 800 West Baltimore Street, Baltimore, MD 21201, USA.
Tel.: +1 410 706 8065; fax: +1 410 706 8121.
E-mail: Lizhang@som.umaryland.edu

Abstract

Summary. Background: The interaction of the fibrin βN-domain with VE-cadherin on endothelial cells is implicated in transendothelial migration of leukocytes, and the β15–42 fragment representing part of this domain has been shown to inhibit this process. However, our previous study revealed that only a dimeric (β15–66)2 fragment, corresponding to the full-length βN-domain and mimicking its dimeric arrangement in fibrin, bound to VE-cadherin. Objective: To test our hypothesis that dimerization of β15–42-containing fragments increases their affinity for VE-cadherin and ability to inhibit transendothelial migration of leukocytes. Methods: Interaction of β15–42-containing fragments with VE-cadherin was characterized by ELISA and surface plasmon resonance. The inhibitory effect of such fragments was tested in vitro with a leukocyte transendothelial migration assay and in vivo with mouse models of peritonitis and myocardial ischemia–reperfusion injury. Results: First, we prepared the monomeric β15–42 and β15–64 fragments and their dimeric forms, (β15–44)2 and (β15–66)2, and studied their interaction with the fibrin-binding domain of VE-cadherin, VE-cad(3). The experiments revealed that both dimeric fragments bound to VE-cad(3) with high affinity, whereas the affinities of β15–42 and β15–64 were significantly lower. Next, we tested the ability of these fragments to inhibit leukocyte transmigration in vitro and infiltration into the inflamed peritoneum in vivo, and found that the inhibitory effects of the dimers on these processes were also superior. Furthermore, (β15–44)2 significantly reduced myocardial injury induced by ischemia–reperfusion. Conclusion: The results confirm our hypotheses and indicate that (β15-66)2 and (β15-44)2, which exhibited much higher affinity for VE-cadherin, are highly effective in suppressing inflammation by inhibiting leukocyte transmigration.

Introduction

Fibrinogen is a hemostatic plasma protein whose major function is to form fibrin clots that prevent the loss of blood upon vascular injury. In addition, fibrin(ogen) participates in a number of other physiologic and pathologic processes through the interaction with various proteins and cell types. Among these processes are inflammation and angiogenesis, which take place during normal wound healing or in pathologic states. It has been shown that interaction of fibrin with endothelial cells promotes the formation of capillary tubes, i.e. angiogenesis [1,2]. Numerous studies have also implicated interactions of fibrin(ogen) with endothelial cells and leukocytes in promoting transendothelial migration of leukocytes and thereby inflammation [3–6].

Fibrinogen is a chemical dimer consisting of two identical disulfide-linked subunits, each formed by three non-identical polypeptide chains, Aα, Bβ, and γ [7]. These chains are folded into a number of domains that contain numerous binding sites for interaction with various proteins and cellular receptors. The central region of the fibrinogen molecule is formed by N-terminal portions of all six chains linked together by 11 disulfide bonds, and is often called the N-terminal disulfide knot (NDSK) [8]. Digestion of fibrinogen with CNBr results in an NDSK fragment corresponding to this region. This fragment retains some binding sites, and is therefore often used in functional studies. Namely, it contains two pairs of fibrin polymerization sites [7] and sites involved in the interaction with the endothelial cell receptor VE-cadherin [2,9,10], cell proteoglycans [11,12], and integrin receptor αXβ2 [13]. The VE-cadherin-binding and proteoglycan-binding sites are located in the N-terminal portions of two fibrin β-chains, each forming a βN-domain [10–12], whereas αXβ2-binding sites are in the N-terminal portion of the fibrin α-chain [13]. These sites are cryptic in fibrinogen and become exposed in fibrin after thrombin-mediated removal of fibrinopeptides A and B from the N-termini of the Aα-chain and Bβ-chain, respectively.

The ability of fibrin(ogen) to interact with different cellular receptors and its dimeric nature make this protein well suited for bridging different cell types. It was suggested more than a decade ago that fibrin(ogen) binding to a variety of vascular cell receptors mediates a specific pathway of cell–cell adhesion by bridging together leukocytes and endothelial cells [3]. Furthermore, it was proposed that such bridging may occur through the interaction of fibrin(ogen) with the leukocyte receptor Mac-1 and endothelial cell receptor ICAM-1, and may contribute to transendothelial migration of leukocytes to injured tissues [3,4]. More recent studies have revealed that a fibrin-derived NDSK fragment devoid of fibrinopeptides A and B (NDSK-II) promotes transendothelial migration of leukocytes [5]. It was suggested that NDSK-II induces leukocyte transmigration by a similar mechanism, namely, by bridging inflammatory cells to the endothelium through its interaction with leukocyte integrin αXβ2 and endothelial cell VE-cadherin [5,6]. Furthermore, it was shown that NDSK-II-induced leukocyte transmigration can be inhibited by a β15–42 peptide [5], which represents part of the fibrin βN-domain and can be easily prepared by plasmin digestion of fibrin or synthesized. The same study [5] also demonstrated that, in rat and mouse models of myocardial ischemia–reperfusion injury, β15–42, by inhibiting leukocyte transmigration, substantially reduced myocardial inflammation and infarct size. This peptide was proposed to be a potential drug candidate for myocardial reperfusion therapy in humans [5,6].

Our studies on the interaction of fibrin with VE-cadherin resulted in identification in the fibrin β-chain of amino acid residues critical for this interaction and localization of the fibrin-binding site in the third domain of VE-cadherin [10,14]. They also revealed that a recombinant dimeric (β15–66)2 fragment, which corresponds to the full-length βN-domain and mimics its dimeric arrangement in fibrin, bound to VE-cadherin with high affinity, whereas the recombinant monomeric β15–64 fragment and the proteolytically prepared β15–42 peptide failed to bind in the studied concentration range [10]. On the basis of these findings, we hypothesized that dimerization is important for the high-affinity interaction of β15–42-containing fragments with VE-cadherin, and that dimeric β15–42-containing fragments should be more potent inhibitors of leukocyte transmigration and thereby inflammation than monomeric β15–42. The major goal of the present study was to test this hypothesis.

Materials and methods

Reagents

Human fibrinogen (depleted of plasminogen, von Willebrand factor, and fibronectin) was from Enzyme Research Laboratories (South Bend, IN, USA); phorbol 12-myristate 13-acetate (PMA) was from Promega (Madison, WI, USA); N-formyl-Met-Leu-Phe (fMLP), extravidin-alkaline phosphatase, dimethylsulfoxide (DMSO) and triphenyltetrazolium were from Sigma (St Louis, MO, USA); and Calcein AM fluorescent dye was from BD Biosciences (Franklin Lakes, NJ, USA).

Mice

C57BL/6 mice of 8–12 weeks old (20–22 g) were from the Jackson Laboratory (Bar Harbor, ME, USA). All mice were housed in a pathogen-free facility, and all procedures were performed with University of Maryland Institutional Animal Care and Use Committee approval.

Preparation of fibrin-binding VE-cadherin fragment

The recombinant fibrin-binding VE-cad(3) fragment (Phe210–Phe325), corresponding to the third domain of the extracellular portion of VE-cadherin, tagged with six histidine residues at the C-terminus, was expressed in Escherichia coli and purified and refolded as described previously [14]. VE-cad(3) was biotinylated with EZ-Link Sulfo-NHS-LC-LC-Biotin (Thermo Scientific, Rockford, IL, USA) as recommended by the manufacturer.

Preparation of fibrin(ogen) fragments

The NDSK fragment corresponding to the central region of fibrinogen was prepared by digestion of human fibrinogen with CNBr [9]. The recombinant Bβ1–64 fragment corresponding to the fibrinogen βN-domain and its dimeric version, (Bβ1–66)2, were produced in E.  coli and purified as described previously [10].

The thrombin-treated NDSK fragment lacking fibrinopeptides A and B (NDSK-II) was prepared by incubation of NDSK with thrombin–agarose from the Thrombin CleanCleave Kit (Sigma) for 2 h at room temperature and subsequent purification by size-exclusion chromatography on Superdex 75. The β15–64 and (β15–66)2 fragments were also prepared by incubation of Bβ1–64 and (Bβ1–66)2 with thrombin–agarose, as above; however, the final purification from contaminating fibrinopeptides was performed on a Superdex Peptide column (GE Healthcare, Piscataway, NJ, USA).

The β15–42 fragment representing a part of the fibrin βN-domain was synthesized by Bachem (Torrance, CA, USA). Its dimeric version, (β15–44)2, was made by addition of two more amino acid residues, Cys43 and Gly44, to the C-terminus of β15–42 and dimerization of the resulting β15–44 fragment through Cys43 by incubation in 0.05 m ammonium acetate buffer (pH 8.0). The dimeric (β15–44)2 fragment with the scrambled β15–42 sequence and Cys43-Gly44 at the C-terminus (scrambled [β15–44]2) was prepared similarly. All fragments used in the inhibition studies were additionally purified on a Detoxi-Gel column (Thermo Scientific) to remove endotoxin contamination.

Solid-phase binding assay

Wells of Immulon 2HB microtiter plates were coated overnight at 4 °C with 2 μg mL–1 (β15–66)2, (β15–44)2, β15–64 or β15–42 in 0.1 m Na2CO3 (pH 9.5) (coating buffer). The wells were then blocked with SuperBlock blocking buffer in Tris-buffered saline (TBS) (20 mm Tris, pH 7.4, 150 mm NaCl) (Thermo Scientific) for 1 h at room temperature. Following washing of the wells with TBS containing 0.05% Tween-20 and 1 mm CaCl2 (binding buffer), the indicated concentrations of biotinylated VE-cad(3) in the binding buffer were added to the wells and also to control wells coated with SuperBlock only, and incubated for 1 h at 37 °C. Bound VE-cad(3) was detected with extravidin conjugated to alkaline phosphatase. The alkaline phosphatase substrate 1-Step p-nitrophenyl phosphate (Thermo Scientific) was added to the wells, and the amount of bound ligand was measured spectrophotometrically at 405 nm. Data were analyzed by non-linear regression analysis, with the equation:

image(1)

where A represents absorbance of the oxidized substrate, which is assumed to be proportional to the amount of ligand bound, Amax is the absorption at saturation, [L] is the molar concentration of ligand, and Kd is the dissociation constant.

Surface plasmon resonance (SPR) analysis

The interaction of β15–42-containing fragments with immobilized VE-cad(3) was studied by SPR with the BIAcore 3000 biosensor (BIAcore AB, Uppsala, Sweden), essentially as described previously [14]. Briefly, immobilization of VE-cad(3) on the surface of the CM5 sensor chip was performed with the amine coupling kit. Binding experiments were performed at a flow rate of 20 μL min–1 in binding buffer HBS-P (GE Healthcare) containing 1 mm CaCl2. Fibrin fragments were injected at increasing concentrations, and the association/dissociation between them and immobilized VE-cad(3) were monitored by the change in the SPR response. Experimental data were analyzed with BIAevaluation 4.1 software (BIAcore AB, Uppsala, Sweden). The dissociation equilibrium constant, Kd, was calculated as Kd = kdiss/kass, where kass and kdiss represent kinetic constants that were estimated by global analysis of the association/dissociation data with the 1 : 1 Langmurian interaction model (kinetic analysis). To confirm the kinetic analysis, Kd was also estimated by analysis of the association data with the steady-state affinity model (equilibrium analysis).

Cell culture and treatments

Human umbilical vein endothelial cells (HUVECs) were purchased from Lonza (Walkersville, MD, USA), grown in EGM-2 complete medium, and used at passage 4–6. The HL-60 promyelocytic cell line (American Type Culture Collection, Manassas, VA, USA) was cultured in Iscove’s modified Dulbecco’s medium (IMDM) (Invitrogen, Carlsbad, CA, USA) supplemented with 10% fetal bovine serum (Atlanta Biologicals, Lawrenceville, GA, USA) and antibiotics. Differentiation of HL-60 cells to a neutrophil-like lineage was achieved by adding DMSO to a final concentration of 1.3% and incubating for 5 days [15]. To assess the morphology of DMSO-induced HL-60 cells, samples containing 100 μL of non-induced or induced cells at a concentration of 0.5 × 106 cells mL–1 were spun at 500 r.p.m. for 5 min on a cytospin slide; the slide with cells was dried and stained with Mayer’s hematoxylin solution (Sigma). Differentiated HL-60 cells were labeled with Calcein AM fluorescent dye (BD Biosciences) in serum-free IMDM, as recommended by the manufacturer. All cell lines were cultured or incubated at 37 °C and 5% CO2.

Leukocyte transendothelial migration assay

Transendothelial migration experiments were performed with 24-well plates containing 8-μm pore size PET membrane inserts (BD Biosciences). HUVECs (1 × 105 in 300 μL) were seeded onto the insert membrane precoated with 0.1% gelatin (Millipore, Billerica, MA, USA) and grown to confluence for 3 days without medium in the lower chamber. The integrity of confluent endothelial monolayers was confirmed by staining the membrane with fixed cells with Coomassie dye. HUVECs were washed twice with IMDM and serum-starved for 2 h before experiments. Calcein AM-labeled differentiated HL-60 cells were stimulated with 100 nm PMA for 30 min, and washed twice with IMDM, and 5 × 105 stimulated cells in 300 μL of IMDM containing 1.5 μm NDSK-II without or with 10 μmβ15–42, (β15–44)2, or 10 μm (β15–66)2 were added on top of the HUVEC monolayer. The inserts were placed into the wells, each of which contained 700 μL of 100 nm chemoattractant fMLP in IMDM. Transmigration proceeded for 4 h, and was stopped by removing the inserts from the wells. Migrated HL-60 cells were recovered from the bottom of the wells, and counted with a fluorescence plate reader at 480 nm/530 nm.

Mouse model of peritonitis

Mice (five per group) were injected intraperitoneally with 3.85% Bacto Fluid thioglycollate (1 mL per mouse) to induce leukocyte infiltration into the peritoneum. To compare the effects of β15–42, (β15–44)2 and (β15–66)2 on leukocyte infiltration, mice received an intravenous injection (via the tail vein) of the fragments, all at the indicated amounts, in 200 μL of phosphate-buffered saline (PBS) (Lonza), prior to intraperitoneal injections of thioglycollate. Mice in control groups received an intravenous injection of the same volume of PBS or the scrambled peptide in PBS. Four hours after the injections, each group of mice were killed by CO2 inhalation followed by cervical dislocation, and injected intraperitoneally with 3 mL of ice-cold PBS; total lavage fluid was withdrawn after massage of their abdomens. The total cell number in lavage fluid was determined with a hemocytometer, and the percentage of neutrophils (∼ 90%) was determined by cytospin, as previously described [16].

Mouse model of myocardial ischemia–reperfusion injury

C57BL/6J mice (8–12 weeks of age) were anesthetized initially with 4.5% isoflurane, and then maintained on 2% isoflurane via a face mask. An ocular lubricant (Paralube) was applied to the animals’ eyes to prevent corneal desiccation. A 1-cm incision on the ventral surface of the neck over the trachea was made to expose the trachea for visualization during orotracheal intubation with a 20G catheter (0.9-mm outside diameter). The mouse was connected to a Harvard Rodent Ventilator, which is supplied with room air supplemented with oxygen at a rate of 105 breaths min–1 and with a tidal volume of 10–15 mL kg−1 body weight. The left jugular vein was isolated and ligated with 6-0 silk suture. A saline-filled PE 10 tube was cannulated into the vein, and positioned to the superior vena cava. A midline thoracotomy was then made between the third and fourth ribs. The left coronary artery near the atrial appendage was ligated with a 8-0 silk suture, and ischemia was maintained for 20 min, after which the suture was removed to initiate blood flow into the ischemic myocardium. The (β15–44)2 fragment at 320 μm in 50 μL of PBS (total dose of 16 nmol) or PBS (50 μL) was bolus injected via the jugular vein catheter into the mice twice, 1 min prior to reperfusion, and 30 min after reperfusion. After 2 h of reperfusion, mice were killed and the hearts were retrieved. For identification of infarcted areas, the hearts were perfused with 1% triphenyltetrazolium, and cut into 2 mm-thick sections with a standard heart matrix (Roboz, Gaithersburg, MD, USA), and the size of infarcted areas, which appear pale in color, was estimated with ImageJ (NIH, Bethesda, MD, USA).

Statistical analysis

Statistical analysis was performed with Student’s t-test, with a P-value of < 0.05 being considered significant. All statistical analyses were performed in SigmaPlot 8.0 (Systat Software, San Jose, CA, USA).

Results

Interaction of β15–42-containing fibrin fragments with VE-cadherin

The previous finding that (β15–66)2 has high affinity for VE-cadherin (Kd = 80 nm) but β15–64 and its truncated variant β15–42 fail to bind even at high concentrations (up to 400 nm) [10] raises two questions. The first concerns the effect of dimerization of the βN-domain (β15–64) on its interaction with VE-cadherin, and the second concerns the contribution of the β43–64 region of this domain, which is missing in β15–42, to this interaction. To address these questions, we prepared recombinant β15–64 and (β15–66)2, as well as synthetic β15–42 and (β15-44)2, and measured their affinity for VE-cad(3), a recombinant fragment corresponding to the fibrin-binding domain of VE-cadherin [14].

To measure the affinity of this interaction by ELISA, microtiter wells were coated with the above-mentioned fibrin fragments and then incubated with increasing concentrations of VE-cad(3) (up to 500 nm). VE-cad(3) exhibited prominent binding to both (β15–66)2 and (β15–44)2, whereas no binding to β15–64 and β15–42 was observed in the tested concentration range (Fig. 1). The Kd value of 18 nm determined for the interaction of VE-cad(3) with (β15–66)2 was comparable with that determined earlier using the same method [14], whereas the interaction of VE-cad(3) with (β15–44)2 was weaker, with a Kd value of 146 nm (Table 1). We were not able to test whether the monomeric fragments exhibit any binding at higher concentrations of VE-cad(3), as further increases in its concentration resulted in the formation of aggregates that prevented reliable measurements. To overcome this problem, we used SPR.

Figure 1.

 Analysis of the interaction between β15–42-containing fibrin fragments and VE-cadherin by ELISA. Increasing concentrations of the VE-cad(3) fragment, representing the fibrin-binding domain of VE-cadherin, were added to immobilized (β15–66)2, (β15–44)2, β15–64, or β15–42 (filled circles, empty circles, filled triangles, and empty triangles, respectively), and bound VE-cad(3) was detected as described in Materials and methods. The curves represent the best fit of the data to Eqn 1. All results are means ± standard deviations of triplicate determinations.

Table 1.   Equilibrium dissociation constants for the interaction between the β15–42-containing fibrin fragments and the VE-cad(3) fragment determined by ELISA and surface plasmon resonance (SPR)
  1. NB, No binding observed up to 500 nm. *Values are means ± standard deviations (SDs) of three independent experiments. Values are means ± SDs of at least three independent experiments.

FragmentsKd (ELISA)*Kd (SPR)
(β15–66)218 ± 6 nm93 ± 16 nm
(β15–44)2146 ± 37 nm715 ± 61 nm
β15–64NB33 ± 5 μm
β15–42NB267 ± 42 μm

In SPR experiments, VE-cad(3) was immobilized on the surface of a sensor chip, and the β15–42-containing fragments were added at increasing concentrations. The (β15–66)2 fragment exhibited prominent binding (not shown), with Kd = 93 nm (Table 1), which is essentially the same value that was obtained by this method earlier [10,14]. The binding of (β15–44)2 was observed at higher fragment concentrations, up to 1 μm (Fig. 2A). The Kd value for this binding was found to be 715 nm (Table 1), i.e. almost one order of magnitude higher than that for (β15–66)2. Binding of β15–42 and β15–64 was observed only at much higher concentrations, up to 100 μm (Fig. 2B,C), and the determined Kd values were 267 and 33 μm, respectively (Table 1). Again, as in the case of the dimeric fragments, the Kd value for β15–42 was almost one order of magnitude higher than that for β15–64. These results suggest that the β43–64 portion of the full-length βN-domain contributes to its interaction with VE-cadherin. At the same time, dimerization of this domain seems to make the major contribution to this interaction, as the Kd values for (β15–66)2 and (β15–44)2 were much lower than those for β15–42 and β15–64.

Figure 2.

 Analysis of the interaction between β15–42-containing fibrin fragments and VE-cadherin by surface plasmon resonance. The (β15–44)2 (A), β15–42 (B) or β15–64 (C) fragments, each at increasing concentrations ([β15–44]2 at 10, 25, 50, 100, 250, 500 and 1000 nm; β15–42 at 5, 10, 25, 50 and 100 μm; β15–64 at 0.25, 0.5, 1, 2.5, 5, 10 and 25 μm), were added to immobilized VE-cad(3), and their association and dissociation, represented by solid curves, were monitored in real time; the dotted curves represent the best fit of the binding data using global fitting analysis, the results of which are presented in Table 1. RU, response units.

In vitro study of the inhibitory effect of β15–42-containing fragments

The fact that (β15–66)2 and (β15–44)2 have much higher affinities for VE-cadherin than their monomeric analogs suggests that both of these dimers should be more potent inhibitors of leukocyte transmigration than the β15–42 used in previous studies [5,17,18]. To test this suggestion, we first compared the inhibitory effect of (β15–66)2, (β15–44)2, and β15–42, using an in vitro transendothelial leukocyte migration assay. For this assay, HL-60 cells were induced with DMSO to differentiate them into neutrophil-like cells [19] (Fig. 3A,B), which were labeled with Calcein AM and stimulated with PMA. It was shown earlier that such cells constitute a valid model system for the analysis of human neutrophil migration [19]. The cells were then added to the upper chamber, which was separated from the lower chamber by the insert membrane coated with a confluent endothelial cell (HUVEC) monolayer, and their transmigration to the lower chamber was stimulated by fMLP (control) and by fMLP in the presence of NDSK-II or NDSK-II with the β15–42-containing fragments. As shown in Fig. 3C, NDSK-II added to the upper chambers stimulated transendothelial migration of DMSO-differentiated HL-60 cells, as expected on the basis of the previous observation [5]. When NDSK-II was added in the presence of (β15–66)2, (β15–44)2, or β15–42, all at 10 μm, the transmigration was reduced, indicating that all three fragments inhibited NDSK-II-induced transmigration of these cells. However, the inhibitory effect of (β15–66)2 and (β15–44)2 was much more pronounced than that of β15–42. Even at a three-fold higher concentration, 30 μm, the effect of β15–42 was lower and did not exceed that of (β15–44)2 at 5 μm, as revealed in subsequent dose-dependence experiments (Fig. 3C, inset).

Figure 3.

 Inhibitory effect of β15–42-containing fibrin fragments on N-terminal disulfide knot (NDSK)-II-dependent transendothelial migration of neutrophils in vitro. Human umbilical vein endothelial cells (HUVECs) were grown to confluency on gelatin-coated cell culture inserts. Calcein AM-labeled HL-60 cells (A) were differentiated into neutrophil-like cells (B) and added to the upper chambers (inserts) on top of the HUVEC monolayer in the presence of 1.5 μm NDSK-II without or with 10 μmβ15–42, (β15–44)2, or (β15–66)2. The cells were allowed to migrate into the lower chambers for 4 h at 37 °C, collected, and measured by fluorescence at 530 nm (C). The results in (C) are expressed as percentage of NDSK-II-dependent cell migration, calculated by subtracting migrated cells in the absence of NDSK-II from total migrated cells. The graph shows combined data obtained from four independent experiments; bars denote means ± standard errors (n = 15–20 wells for each condition); ***P < 0.001. The inset in (C) shows the dose dependence of the inhibitory effect of β15–42 and (β15–44)2 (empty and filled circles, respectively) on NDSK-II-dependent neutrophil transmigration.

In vivo study of the anti-inflammatory effect of β15–42-containing fragments

To compare the effects of (β15–66)2, (β15–44)2 and β15–42 on leukocyte transmigration in vivo, we used a peritonitis mouse model. In this model, leukocyte migration from the circulation into the peritoneum is stimulated by intraperitoneal injection of thioglycollate, and leukocyte (neutrophil) accumulation is evaluated after 4 h by counting the cells in the peritoneal lavage. In our experiments, each mouse was injected intravenously with 200 μL of either β15–42, (β15–44)2, or (β15–66)2, all at 80 μm (total dose per mouse of 16 nmol), prior to intrapeitoneal injection of thioglycollate; control mice were injected intravenously either with the same amount of the scrambled fragment (β15–44)2 or PBS. The experiments revealed that the number of neutrophils accumulating in the peritoneum of fragment-treated mice was significantly lower than that in control mice (Fig. 4). No statistically significant difference was observed between accumulation of neutrophils in control mice injected with the scrambled fragment and PBS. This finding indicates that, in this model, all three fragments inhibited leukocyte transmigration and thereby inflammation. However, the extent of inhibition by (β15–66)2 and (β15–44)2 in the selected experimental conditions was almost twice as high as that of β15–42.

Figure 4.

 Inhibitory effect of β15–42-containing fibrin fragments on neutrophil infiltration in vivo in a mouse model of peritonitis. The indicated fragments were injected intravenously, each in 200 μL of phosphate-buffered saline (PBS) at 80 μm (total dose per mouse was 16 nmol); control mice were injected intravenously with 200 μL of PBS or scrambled (β15–44)2 fragment at 80 μm in 200 μL of PBS. The graph shows combined data from four independent experiments, and the results are means ± standard errors (the combined number of mice in each group was as follows: injected with PBS or the scrambled fragment, n = 25 and n = 9, respectively; injected with β15–42, [β15–44]2, or [β15–66]2, n = 15, n = 25, and n = 10, respectively). **P < 0.01; ***P < 0.001.

Next, using the same peritonitis model, we performed a systematic study of the anti-inflammatory effects of (β15–66)2, (β15–44)2 and β15–42 to identify the optimal physiologically active amounts of these fragments for further in vivo experiments. The results of the study, in which 200 μL of each fragment at different concentrations (total dose per mouse of 4, 8 and 16 nmol) were injected intravenously, revealed that both (β15–66)2 and (β15–44)2 exhibited similar inhibitory effects on transendothelial migration of leukocytes, and were about twice as potent in inhibiting leukocyte transmigration as β15–42 (Fig. 5). These results further reinforce the above conclusion that the dimeric fragments are more potent inhibitors of leukocyte transmigration than β15–42. They also suggest that the maximal inhibitory effect can be achieved at a fragment amount of 8 nmol per mouse and above. This was taken into account in our further in vivo experiments.

Figure 5.

 Dose dependence of the inhibitory effect of β15–42-containing fibrin fragments on neutrophil infiltration in vivo in a mouse model of peritonitis. β15–42, (β15–44)2 and (β15–66)2 (circles, squares, and triangles, respectively) were injected intravenously in the indicated amounts (each in 200 μL at 20, 40 and 80 μm; total dose per mouse was 4, 8 and 16 nmol, respectively). The results are expressed as percentage of control groups, and are means ± standard errors (n = 5).

Testing the cardioprotective effect of (ββ15–44)2

By inhibiting leukocyte transmigration, β15–42 was shown to reduce myocardial inflammation and infarct size in animal models of myocardial ischemia followed by reperfusion [5,17,18]. To evaluate the potential of dimeric β15–42-containing fragments for myocardial infarction therapy, we used a mouse model of myocardial ischemia–reperfusion injury. In this model, ischemia is achieved by ligation of the left coronary artery with a suture; after 20 min, the suture is cut to initiate blood flow into the ischemic myocardium (reperfusion), and the size of infarcted area is evaluated after 2 h of reperfusion. Using this model, we tested the cardioprotective effect of (β15–44)2, which, together with (β15–66)2, demonstrated the highest inhibitory effect on leukocyte transmigration in the experiments described above. This fragment was injected into the mice twice via the jugular vein (16 nmol per mouse in each injection), 1 min prior to and 30 min after reperfusion; control mice were injected with PBS. The results presented in Fig. 6 indicate that infarct size in mice treated with (β15–44)2 was reduced more than two-fold in comparison with that in control mice. Thus, in this in vivo model, (β15–44)2 exhibited a significant cardioprotective effect.

Figure 6.

 Cardioprotective effect of (β15–44)2 during myocardial reperfusion injury. Representative mouse heart slices after myocardial ischemia–reperfusion in mice treated with phosphate-buffered saline (PBS) (A) or (β15–44)2 (B). The size of infarcted areas, which appear pale in color (contoured by broken lines in [A] and [B]), was determined with ImageJ (NIH), and the results are presented in (C) as percentage of total area of the slices. The results are means ± standard errors (n = 5). ***P < 0.001.

Discussion

Previous studies revealed that the interaction of fibrin with VE-cadherin on endothelial cells promotes angiogenesis, and that this interaction occurs through the βN-domain of fibrin and is inhibited by β15–42, representing part of this domain [1,2,9]. A subsequent study [5] demonstrated that this interaction also promotes transendothelial migration of leukocytes, and thus inflammation. Furthermore, it has been shown that β15–42 inhibits VE-cadherin-dependent leukocyte transmigration, thereby reducing reperfusion-induced inflammation and infarct size in animal models of ischemia–reperfusion [5]. At the same time, our study [10] revealed that (β15–66)2, which corresponds to the full-length βN-domain and mimics its dimeric arrangement in fibrin, has much higher affinity for VE-cadherin than β15–42, suggesting that this dimer may be a more potent inhibitor of leukocyte transmigration. The present study clarified the structural basis for the increased affinity of (β15–66)2 for VE-cadherin, and confirmed its superior inhibitory activity over β15–42. In addition, this study revealed that a truncated variant of this dimer, (β15–44)2, also has high affinity for VE-cadherin and inhibitory activity towards leukocyte transmigration.

The major difference between (β15–66)2 and β15–42 is the dimeric nature of the former and the absence of the β43–64 region in the later. Thus, to check the contribution of this region as well as that of the dimerization to the interaction with VE-cadherin, we prepared β15–64 and β15–42 and their dimeric variants (β15–66)2 and (β15–44)2, and studied their interaction with VE-cadherin by ELISA and SPR. The recombinant third domain of VE-cadherin, VE-cad(3), was used as a simple model of VE-cadherin, as its affinity for fibrin is comparable to that of the extracellular portion of VE-cadherin [10,14]. The Kd values for the interaction of (β15–66)2 and (β15–44)2 with VE-cad(3) determined by ELISA were found to be lower than those determined by SPR. A similar difference for the interaction of VE-cad(3) with (β15–66)2 was observed in our previous study [14]. The lower apparent Kd values determined by ELISA may be connected with the observed tendency of VE-cad(3) to form aggregates, whose binding to immobilized (β15–66)2 or (β15–44)2 may increase the apparent concentration of bound VE-cad(3), thus decreasing Kd. This possible problem was absent in SPR experiments, in which VE-cad(3) was immobilized on the surface of a sensor chip, and fibrin fragments were kept in solution. Thus, the data obtained by SPR should provide more accurate estimates of the Kd values than those obtained by ELISA.

Analysis of the SPR data revealed substantial differences in the affinities for VE-cadherin between the full-length and truncated fibrin fragments, both monomeric and dimeric. Namely, the Kd values for the interaction of (β15–66)2 and β15–64 with VE-cad(3) were found to be eight-fold lower than those determined for their truncated variants, (β15–44)2 and β15–42 (Table 1). This indicates that removal of the β43–64 region from both the dimeric and monomeric full-length fragments substantially decreases their affinity for VE-cad(3) suggesting the involvement of this region in the interaction with VE-cadherin. Our previous study localized the VE-cadherin-binding sites to the N-terminal portions of the fibrin β-chains (βN-domains), and identified two amino acid residues, His16 and Arg17, that are critical for the binding [10]. The results of the present study suggest that the VE-cadherin-binding site is more complex, and may also involve amino acid residues from the β43–64 region.

Our previous study also revealed that dimerization of the βN-domain is important for the interaction with VE-cadherin, as (β15–66)2 bound to VE-cadherin with high affinity, whereas β15–64 and β15–42 failed to bind [10]. However, the highest concentration of β15–64 and β15–42 tested in that study was 400 nm. In the present study, we performed binding experiments at higher concentrations of these fragments, and found that they do bind to VE-cadherin; however, their affinities were much lower than those determined for their dimeric variants. Namely, the Kd values determined for the interaction of (β15–66)2 and (β15–44)2 with VE-cad(3) were found to be 355-fold and 373-fold lower than those determined for β15–64 and β15–42, respectively (Table 1). This difference is much greater than that between the full-length and truncated variants. Thus, this finding indicates that, although the β43–64 region contributes to the interaction with VE-cadherin, dimerization of these fragments plays the major role in the high-affinity interaction. This finding also raises the question of whether shorter variants of (β15–44)2 will retain high affinity for VE-cadherin.

Our finding that (β15–66)2 and (β15–44)2 have much higher affinity for VE-cadherin than β15–42 suggests that they may both be more potent inhibitors of transendothelial migration of leukocytes than β15–42. To test this suggestion, we compared the inhibitory effects of all three fragments on leukocyte transmigration in vitro and in vivo. In the in vitro transmigration assay, in which we used HL-60 cells differentiated into neutrophil-like cells, both dimeric fragments exhibited similar inhibitory effects on NDSK-II-induced migration of these cells through a HUVEC monolayer, whereas the effect of β15–42 was less pronounced. This difference was also observed in the in vivo experiments with a peritonitis model, a well-established mouse model for leukocyte migration into sites of acute inflammation [16,20–22]. In these experiments, the inhibitory effect of the dimeric fragments exceeded that of β15–42 more than two-fold (Figs 4 and 5). Altogether, these experiments confirmed the hypothesized superior anti-inflammatory properties of (β15–66)2 and (β15–44)2. It remains to be established whether shorter variants of (β15–44)2, which may retain high affinity for VE-cadherin, also exhibit superior anti-inflammatory effects.

It should be noted that, in spite of the very low affinity of β15–42 for VE-cadherin, its inhibitory effect on leukocyte transmigration in our in vitro and in vivo experiments was fairly high, about half of that exhibited by (β15–66)2. This discrepancy can be explained by the fact that its affinity for the endothelial cell surface (Kd = 0.18 μm), which was determined earlier [2], is much higher than that for VE-cadherin determined in this study (Kd = 267 μm; Table 1). This suggests that β15–42 may interact with other endothelial cell receptor(s). Indeed, as fibrin and its β15–42-containing fragments interact with heparin [11,12], they should also interact with cell surface proteoglycans, and this interaction may contribute to their increased affinity for the endothelial cell surface. Thus, this interaction may represent an additional mechanism contributing to fibrin-dependent leukocyte transmigration; the existence of other mechanisms cannot be excluded.

As the difference between the anti-inflammatory effects of (β15–66)2 and (β15–44)2 was statistically insignificant, we tested the effect of the smaller fragment, (β15–44)2, on infarct size in a mouse model of myocardial ischemia–reperfusion injury. The advantage of this short dimer is that its monomeric unit contains only 30 amino acids and can be easily synthesized and dimerized through Cys43. The experiments revealed that treatment with this fragment of mice subjected to ischemia–reperfusion reduced myocardial infarct size more than two-fold, indicating that (β15–44)2 has a significant cardioprotective effect in this model. The question of whether this dimeric fragment, as well as (β15–66)2, is a more efficient inhibitor of ischemia–reperfusion-induced myocardial inflammation than β15–42 remains to be addressed.

A number of recent studies with β15–42 further confirmed its cardioprotective effect in rodent and pig models of myocardial ischemia–reperfusion injury [17,18], and revealed its organ-protective effect in pig models of hemorrhagic shock and reperfusion [23,24]. Furthermore, the results of a recent trial on the efficacy of β15–42 in the prevention of myocardial reperfusion injury revealed the potential of this fragment to protect during myocardial infarction in humans [25]. At the same time, they also suggest that more potent fibrin-derived inhibitors of inflammation may be required for more efficient treatment of reperfusion injury in human patients. In this connection, the dimeric fragments described in the present study, (β15–66)2 and (β15–44)2, which exhibited a much higher anti-inflammatory effect than β15–42 in animal models, may also have a higher cardioprotective effect in humans.

In summary, the present study has revealed that dimerization of β15–42-containing fragments, giving (β15–66)2 and (β15–44)2, plays a major role in increasing their affinity for VE-cadherin, and that such dimeric fragments are more potent inhibitors of leukocyte transmigration than monomeric β15–42 in the in vitro transendothelial migration assay and the in vivo peritonitis model. The study has also demonstrated a significant cardioprotective effect of (β15–44)2 in a mouse model of myocardial ischemia–reperfusion injury. Whether dimeric β15–42-containing fragments and their truncated variants are more potent cardioprotective agents in animals and humans remains to be tested.

Acknowledgements

We thank D. Ehirchiou for technical assistance.

Disclosure of Conflict of Interests

This work was supported by a Maryland Industrial Partnerships Program grant (MIPS Contract Agreement no. 4013), RegeneRx Biopharmaceuticals Inc., Rockville, MD, and National Institutes of Health grants HL-056051 (L. Medved), HL-054710 (D. K. Strickland & L. Zhang), and HL-072929 (D. K. Strickland). The other authors state that they have no conflict of interest.

Ancillary