Mechanisms underlying FeCl3-induced arterial thrombosis


  • This work has been presented as an oral communication at the XXIst Conference of the International Society on Thrombosis and Haemostasis in Boston, MA, USA (July 2009).

Christian Gachet, UMR-S949 INSERM-Université de Strasbourg, Etablissement Français du Sang-Alsace (EFS-Alsace), 10, rue Spielmann, B.P. N 36, 67065 Strasbourg Cedex, France.
Tel.: +33 388 21 25 25; fax: +33 388 21 25 21.


See also Brill A. A ride with ferric chloride. This issue, pp 776–8.

Summary. Background: The FeCl3-induced vascular injury model is widely used to study thrombogenesis in vivo, but the processes leading to vascular injury and thrombosis are poorly defined. Objectives: The aim of our study was to better characterize the mechanisms of FeCl3-induced vascular injury and thrombus formation, in order to evaluate the pathophysiological relevance of this model. Methods: FeCl3 was applied at different concentrations (from 7.5% to 20%) and for different time periods (up to 5 min) to mouse carotid or mesenteric arteries. Results: Under all the conditions tested, ultrastructural analysis revealed that FeCl3 diffused through the vessel wall, resulting in endothelial cell denudation without exposure of the inner layers. Hence, only the basement membrane components were exposed to circulating blood cells and might have contributed to thrombus formation. Shortly after FeCl3 application, numerous ferric ion-filled spherical bodies appeared on the endothelial cells. Interestingly, platelets could adhere to these spheres and form aggregates. Immunogold labeling revealed important amounts of tissue factor at their surface, suggesting that these spheres may play a role in thrombin generation. Invitro experiments indicated that FeCl3 altered the ability of adhesive proteins, including collagen, fibrinogen and von Willebrand factor, to support platelet adhesion. Finally, real-time intravital microscopy showed no protection against thrombosis in GPVI-immunodepleted and β1−/− mice, suggesting that GPVI and β1 integrins, known to be involved in initial platelet adhesion and activation, do not play a critical role in FeCl3-induced thrombus formation. Conclusion: This model should be used cautiously, in particular to study the earliest stage of thrombus formation.


Increasing knowledge of the mechanisms of thrombus formation has been gained from experimental models that enable real-time in vivo observation of thrombus formation in mice. One of the most widely used procedures employs topical application of ferric chloride (FeCl3) to an artery [1–4]. This is a simple and well-established model known to be sensitive to both anticoagulant and antiplatelet drugs. It is generally admitted that FeCl3 causes major oxidative stress with the generation of free radicals, which leads to lipid peroxidation and destruction of endothelial cells and results in occlusive thrombus formation. However, this model remains poorly characterized. Tseng et al. [5] reported that FeCl3 crossed the endothelium in small vesicles by an endocytic-exocytic pathway, thereby causing complete endothelial denudation. While this study focused on the effects of FeCl3 on the intima, little is known about the consequences for underlying vascular layers (media, adventitia). Moreover, whether FeCl3 induces chemical modification of adhesive proteins has not been investigated and how it might influence their function is likewise not known. Another question concerns the role of circulating blood cells, which have been proposed to mediate FeCl3-induced vascular injury. One ex vivo study reported a direct hemolytic effect of FeCl3 on erythrocytes, the subsequent hemoglobin oxidation participating in endothelial denudation [6]. The importance of this mechanism nevertheless remains to be confirmed in vivo.

A hallmark of the FeCl3-injury model is its sensitivity to thrombin inhibitors [7–10]. How thrombin is generated in this model is, however, an open question. Tissue factor expressed on leukocytes or circulating microparticles or derived from the arterial wall has been proposed to contribute to thrombus growth by supporting thrombin generation, but the relative importance of blood-borne and vessel wall tissue factor (TF) is not clear [11–14].

The role of collagen-dependent platelet adhesion via glycoprotein VI (GPVI) and integrin α2β1 in this model remains controversial [7,15–18]. Whereas some groups reported severe inhibition of FeCl3-induced thrombus formation in GPVI-immunodepleted or FcRγ−/− mice, which lack GPVI expression on the platelet surface, others concluded that GPVI deficiency in itself did not impair thrombosis. These divergent results are difficult to interpret as the vascular lesions were not examined in these studies and the discrepancies could arise from differences in experimental conditions, such as the FeCl3 concentration (7–20%), the method (filter paper or drop) and time (2–5 min) of FeCl3 application, or the vessel bed targeted (carotid or mesenteric artery). Such differences in experimental conditions can cause variations in the depth and extent of the vascular lesions, which consequently may modify the exposure of specific proteins to blood flow.

Overall, these questions require clarification and point to the need for a detailed characterization of the ferric chloride model of arterial thrombosis. The aims of the present work were (i) to examine in vivo the structure of the thrombus with regard to the activation of platelets and coagulation, and to investigate in this context the mechanism of thrombin generation by determining the contribution of tissue factor; (ii) to characterize in vivo the damage to the vessel wall with respect to the modification and/or exposure of adhesive proteins; (iii) to study in vitro the reactivity of adhesive proteins following FeCl3 treatment; and (iv) to investigate the potential involvement of circulating blood cells in this model.

Materials and methods


Xylasine (Rompun®) and ketamine (Imalgen®) were from Bayer (Puteaux, France) and Merial (Lyon, France), respectively. JAQ1, a monoclonal anti-mouse GPVI antibody, was provided by EMFRET Analytics GmbH & Co. KG (Würzburg, Germany) and recombinant hirudin by Transgene SA (Strasbourg, France). Botrocetin was kindly provided by Dr S.P. Jackson (Melbourne, Australia). Fibrinogen and human von Willebrand factor (VWF) were from EFS-Alsace (Strasbourg, France). A sheep antibody against rabbit tissue factor was from American Diagnostica (Stamford, CT, USA) and a rabbit anti-sheep antibody from Jackson ImmunoResearch (Baltimore, MA, USA). A rat anti-mouse PECAM-1 antibody was from BD Pharmingen (San Diego, CA, USA). Clopidogrel was a generous gift from Sanofi-Aventis (Toulouse, France) and equine collagen (Kollagenreagent Horm) was purchased from Hormon Chemie (Munich, Germany).

Animals and treatment

C57BL6/J (Charles River, Arbresle, France), αIIb-deficient (αIIb−/− provided by W. Vainchenker, INSERM UMR 1009, Paris, France), β1-deficient (β1−/− provided by R. Fässler, Max Planck Institute, Martinsried, Germany) and VWF-deficient (VWF−/−provided by C. Denis, INSERM UMR S770, Paris, France) mice were maintained and experiments were performed in the Animal Facilities of EFS-Alsace. All procedures for animal experiments were performed in accordance with the Guide for Care and Use of Laboratory Animals as defined by European laws.

JAQ1 was given i.p. at 50 μg per mouse, a dose resulting in full inhibition of ex vivo platelet aggregation in response to collagen (50 μg mL−1) 4 days post-injection, due to GPVI removal [19]. Hirudin (20 mg kg−1) was administered s.c., 30 min before vessel injury, a dose giving an APTT (activated partial thrombin time) of > 200 s. Clopidogrel treatment was performed by oral administration (50 mg kg−1) the day before and 2 h before the experiment.

FeCl3 injury thrombosis models

Mice (8–10 weeks old) were anesthetized with an i.p. injection of ketamine (100 mg kg−1) and xylazine (20 mg kg−1) and injected with a fluorescent dye, DIOC6 (5 μL of a 100 μM solution g of body weight−1), to label platelets and assist thrombus visualization. The common carotid arteries were exposed and vascular injuries were generated by applying a Whatmann filter paper saturated with (i) 7.5% FeCl3 on top of the left carotid artery for 2 min (1 mm diameter patch of 1M Whatmann paper) or (ii) 20% FeCl3 laterally to the vessel for 5 min (3MM Whatmann filter, 3 mm length and 1 mm width). The latter method is used to allow better visualization of thrombus growth. For mesenteric arterioles, a filter paper (1 mm diameter patch of 1M Whatmann paper) saturated with 10% FeCl3 was applied topically for 30 s. Thrombus formation was monitored in real time under a fluorescent microscope (Leica Microsystems SA, Westlar, Germany) coupled to a CCD (charge-coupled device) camera (CoolSNAP HQ2; Photometrics, Tucson, AZ, USA). Analyses were performed with Metamorph software (Molecular Devices, Sunnyvale, CA, USA). To determine the time to occlusion, a miniature Doppler flowprobe (Model 0.5 PSB; Transonic Systems, Ithaca, NY, USA) was placed at the surface of the carotid artery and blood flow was recorded using a Transonic Model TS420 flowmeter (Transonic Systems).

Histology and electron microscopy of FeCl3-induced injury of mouse carotid arteries

The arteries were fixed with 2.5% glutaraldehyde by transcardiac perfusion at various times after injury: 2 min corresponding to the time from removal of the filter until the end of fixative perfusion (earlier time points could not be explored), 5 min corresponding to the growth phase of the thrombus and 8 min corresponding to the maximal thrombus size (Fig. 2). The injured artery and control contralateral artery were then excised and post-fixed in 2.5% glutaraldehyde overnight. To investigate the effect of FeCl3 exclusively on the vessel wall, circulating blood cells were removed by in vivo whole-body perfusion with phosphate-buffered saline (PBS) through the left ventricle, after which 7.5% FeCl3 was applied to the carotid for 2 min. The vessels were fixed immediately after removal of the filter and processed for transmission electron microscopy (TEM) or scanning electron microscopy (SEM).

Figure 2.

 Identification of the spherical bodies observed 2 min after FeCl3 treatment. The carotid arteries were fixed 2, 5 or 8 min after removal of the FeCl3-soaked (7.5% FeCl3) filter. (A) Scanning electron microscopy images of the luminal side of longitudinally cut mouse carotid arteries. Lower panels are close-up views of the frames in the upper panels and the arrows indicate blood flow. (a) At 2 min, discoid platelets (P), elongated erythrocytes (Er) and a few fibrin strands were observed, together with numerous bodies (*) of unusual appearance (white insert). (b) At 5 min, a parietal thrombus composed of highly contracted platelets and covered with numerous fibrin fibers (F) was observed. (c) At 8 min, an extensive thrombus composed of closely packed platelets had formed. Platelets close to the vessel wall were contracted (insert 1) while those at the periphery of the thrombus were more discoid (insert 2). Scale bars: 2 μm. (B) (a) Closer examination of the bodies by scanning electron microscopy showed that they were spherical, about 3 μm in diameter and presented some rough material at their surface. Scale bar: 1 μm. (b) Transmission electron microscopy revealed that the bodies were surrounded by a bilayer (magnification in the insert) with characteristics of a plasma membrane. Arrows indicate the position of small electron dense precipitates. Scale bar: 1 μm. E, endothelium. (c) Immunogold labeling showed PECAM-110-positive staining of these membranes (see Materials and methods for details). (C) Chemical composition of the spheres by energy dispersive X-ray analysis. Left: scanning electron microscopy area analyzed showing the position (*) of some spheres. Right: energy dispersive X-ray cartography of ferric ions in the same area. The distribution of ferric ions is superimposable on that of the spheres. Scale bars: 2 μm.

The isolated vessel segments were sectioned in the middle and processed separately for TEM and SEM. For TEM, one half-segment was post-fixed in 1% osmium tetroxide for 1 h, washed in cacodylate buffer and dehydrated in graded ethanol solutions. The specimens were embedded in Epon, 100 nm thick cross-sections were generated, stained with lead citrate and uranyl acetate, and examined under a CM120 TEM at 120 kV (FEI, Eindhoven, the Netherlands). To identify endothelial cells, the vessels were embedded in Lowicryl and the endothelial plasma membrane was stained with a rat anti-mouse PECAM-1 antibody (dilution 1/50) followed by protein A conjugated to 10 nm gold particles. All incubation and blocking steps were carried out as described previously [20]. Sections incubated with irrelevant IgG served as negative controls.

For SEM, the other half-segment was dehydrated in increasing concentrations of ethanol followed by hexamethyldisilazine and glued onto a cover slip with the lumen uppermost. The cover slips were sputtered with gold prior to observation at 5 kV under an FEG Sirion SEM (FEI). The chemical distribution of elements was studied by energy dispersive X-ray emission microanalysis (EDX) of selected sections using an ESEM (FEI) at 20 kV, spot size 4.

For histological studies, the isolated vessel segments were processed in 70% ethanol and xylene, embedded in paraffin blocks and sectioned at 5-μm thickness. Prussian blue staining performed according to the recommendations of the manufacturer (Polysciences, Eppelheim, Germany) was used to localize iron in the vessels.

In vitro effect of FeCl3 on the reactivity of matrix proteins

For static platelet adhesion studies, glass coverslips were coated with 2.5 mg mL−1 collagen or 50 μg mL−1 fibrinogen for 90 min or with 10 μg mL−1 human VWF for 2 h at room temperature. After coating, the coverslips were blocked with 0.1% human albumin in PBS for 1 h at room temperature, and then treated for 2 min with FeCl3 (0.0002–0.4%) or Tyrode’s buffer (control). Use of a higher concentration of FeCl3 was not possible due to its marked precipitating effect on proteins. After rinsing, the coverslips were incubated with washed platelets prepared as described [21] (300 μL per well) in Tyrode’s albumin buffer (collagen and fibrinogen studies) or the same buffer containing 1 μg mL−1 botrocetin (VWF studies) at room temperature for 30 min. Non-adherent platelets were removed by washing the coverslips three times with PBS. Adherent platelets were then fixed with 2.5% glutaraldehyde for 1 h, washed with PBS and prepared for SEM. Platelets were counted in five random fields (magnification × 2000, 1110 μm2 per field). Statistical analyses were performed by anova, a P value < 0.05 being considered significant.

For platelet adhesion under flow conditions, glass capillaries were coated with fibrinogen (300 μg mL−1), human VWF (10 μg mL−1) or collagen (2.5 mg mL−1) for 2 h and blocked with PBS (1% HSA) for 30 min at room temperature. The capillaries were then treated with FeCl3 (0.1% or 0.4%) or Tyrode’s for 2 min before being washed with PBS. Hirudinated (100 U mL−1) whole blood was perfused at 300 s−1 over fibrinogen and at 1500 s−1 over VWF and platelet adhesion was observed with differential interference contrast (DIC) microscopy (Leica, Mannheim, Germany). For flows over collagen, blood was preincubated with DIOC6 (1 μM) to label platelets before perfusion at 1500 s−1 and platelet adhesion was observed with epifluorescence microscopy.

Immunogold labeling of tissue factor

The method was adapted from an intravital immunolabeling procedure [22] to detect tissue factor by electron microscopy. A sheep antibody against rabbit tissue factor (8 μg g−1) or control sheep IgG (8 μg g−1) was pre-incubated for 20 min with rabbit anti-sheep IgG (4 μg g−1) before infusion into mice. FeCl3 (7.5% for 2 min) was applied to the carotid artery and fixation was performed immediately after removal of the filter. The artery was cut open longitudinally and directly incubated for 20 min with protein A conjugated with 10 nm gold particles, prior to embedding the artery in resin. Using this pre-embedding approach, the immunostaining is carried out directly on the artery and the gold conjugates stain only the TF accessible on external surfaces.


Evaluation of thrombosis induced by various FeCl3 concentrations and sensitivity to anti-platelet and anticoagulant agents

Thrombus formation was evaluated using increasing concentrations of FeCl3 (2.5–20%). Thrombus formation increased dose-dependently from 2.5% to 10% FeCl3, applied on top of the carotid artery (Fig. 1Aa). Higher concentrations (20% FeCl3) led to vessel opacification, impairing thrombus visualization. Therefore, 20% FeCl3 was applied laterally, which resulted in a large thrombus inducing occlusion in six out of 10 mice (Fig. 1B). FeCl3 applied topically at 7.5% generated a reproducible (Fig. 1Ab) (mean maximal size of the thrombus, 4.3 ± 0.02 × 105 μm2, with a standard error of the mean of 4.7% as acceptable variability, n = 34) subocclusive thrombosis (Fig. 1Ac). Most of the histological studies have been performed using this experimental condition (7.5% FeCl3 for 2 min). Thrombus size was strongly inhibited in αIIb−/− mice and in animals treated with hirudin (20 mg kg−1), which blocks thrombin, or with clopidogrel (50 mg kg−1), which antagonizes the platelet P2Y12 receptor (Fig. 1C). These data confirmed the importance of platelet aggregation and coagulation in this model.

Figure 1.

 Carotid artery thrombosis induced by FeCl3. (A) The common carotid arteries were exposed and a filter paper saturated with FeCl3 was placed on top of the left vessel for 2 min. (Aa) Tracings representing the mean surface area (± standard error of the mean, SEM, n = 5) of developing thrombi after application of different concentrations of FeCl3. (Ab) Left: individual data for the peak thrombus area of 34 injuries induced by 7.5% FeCl3. Right: mean ± SEM of the peak thrombus area (4.3 ± 0.02 × 105 μm2, n = 34). (Ac) Representative photograph of the carotid artery cross-section showing the thrombus at maximal size following 7.5% FeCl3 injury. Bar: 50 μm. (B) FeCl3 was applied laterally to the carotid artery for 5 min. (Ba) Tracings representing the mean surface area (± SEM, n = 5) of developing thrombi after application of 7.5% (2.5 min) or 20% (5 min) FeCl3. (Bb) Left panel: for each thrombus formed, the time to occlusion was evaluated. Right panel: representative images of the thrombus formed 15 min after application. (C) FeCl3-induced thrombosis in αIIb-deficient mice (αIIb−/−, n = 6), in mice treated with 50 mg kg−1 clopidogrel per os (n = 8) or injected with 20 mg kg−1 hirudin (n = 5) and in control mice (wild-type, WT). Panels depict the mean thrombus surface area as a function of time and the dotted lines represent the SEM at each time point (1-s intervals).

Platelet activation and fibrin formation after exposure to FeCl3

The structure of the thrombus was analyzed by SEM at different time points with regard to platelet activation and the presence of fibrin (Fig. 2A). At 2 min, examination of the luminal side of the vessel showed the presence of discoid platelets, a few scattered erythrocytes displaying shapes elongated in the flow direction and some long fibrin strands (Fig. 2Aa). At 5 min, a parietal thrombus was observed. Platelets close to the vessel wall were highly contracted, tightly packed and enmeshed in a fibrin network providing evidence of a thrombin-rich environment (Fig. 2Ab). At 8 min, a compact and subocclusive thrombus had formed (Fig. 2Ac). Platelets visible at the periphery of the thrombus were less contracted (insert 2) than those close to the site of FeCl3 application (insert 1), indicating a gradation in platelet activation. Interestingly, at all time points the blood cells present in the vessel displayed a well-preserved morphology, suggesting that under these conditions FeCl3 induced no significant blood cell damage in vivo. Of further interest was the observation at 2 min of numerous, unidentified spherical bodies approximately 3–5 μm in diameter at the luminal surface of the vessel (see insert in Fig. 2Aa). These spheres were analyzed in more detail.

Formation of iron-filled spherical bodies originating from the endothelium

SEM revealed that the spherical bodies displayed a smooth surface with some rough material (Fig. 2Ba). TEM showed these spheres to be electron dense round bodies budding off from endothelial cells and surrounded with a bilayer presenting characteristics of a plasma membrane (Fig. 2Bb). Immunogold labeling revealed PECAM-1-positive staining, strongly suggesting that the spheres were surrounded by membranes originating from endothelial cells (Fig. 2Bc). The chemical nature of the content of the spheres was analyzed by energy dispersive X-ray (EDX) SEM, which allows one to determine the elemental composition (Fig. 2C). Ferric ions were identified with this technique and EDX cartography showed their superposition on the spheres, indicating that the latter contained ferric ions. These results suggest that ferric chloride diffuses through the vessel wall and forms large spherical bodies budding off from the endothelium into the vessel lumen. It is noteworthy that activated platelets were observed on top of and in contact with the spheres, some platelets being in direct contact with the rough material of the spheres (Fig. 3A–C). In contrast, at early time points, no platelet attachment was detected directly on the endothelial cell surface, that is, in the absence of the spheres (Fig. 3A,B).

Figure 3.

 Platelet attachment to the spherical bodies. (A) Scanning electron microscopy and (B) transmission electron microscopy observation showed that platelets could adhere to the spheres (*). Scale bars: 5 μm. (C) Close-up transmission electron microscopy analysis revealed that activated platelets were in close apposition with the dense material (arrow). Bar: 1 μm. P, platelets; IEL, internal elastic lamina.

TF positive staining of the spherical bodies

An important fibrin network was detected around the thrombus 5 min after arterial injury, indicating a thrombin-rich environment (Fig. 2Ab). We therefore looked at how thrombin was generated in this model and whether the iron-containing spherical bodies were involved in this process. Strong TF staining was observed inside and at the surface of the spheres (Fig. 4Aa and insert) as well as on the surface of activated platelets (Fig. 4Ab and insert), whereas no gold particles were found on endothelial cells or in the control IgG-treated artery (data not shown). Interestingly, fibrin fibers could be detected in direct contact with the FeCl3-loaded spheres already 2 min after injury, at a time point when only a few platelets adhered to the site of injury, suggesting that thrombin could participate in the initiation of thrombosis (Fig. 4B). These results showed that application of FeCl3 to the carotid artery led to the exposure of TF at the surface of FeCl3-containing spheres, which could thus support thrombin generation and trigger in turn platelet adhesion, activation and coagulation.

Figure 4.

 Exposure of tissue factor after FeCl3 injury of mouse carotid arteries. (A) Tissue factor (TF) was detected ex vivo by i.v. injection of a mixture of sheep antibody against rabbit tissue factor (8 μg g−1) and rabbit anti-sheep antibody (4 μg g−1). 7.5% FeCl3 was applied to the carotid artery and fixation was performed 2 min later, immediately after removal of the filter. The artery was then incubated for 20 min with protein A conjugated with 10 nm gold particles, prior to embedding it in resin. The immunogold staining was visible (arrows and insert) on the spheres (a) and at the surface of activated platelets (b). No gold particles were detected in the artery when control IgG was used instead of the anti-TF antibody. P, platelet. Bars: left, 1 μm; right, 200 nm. (B) Transmission electron microscopy observation of the carotid arteries showed that some spheres filled with FeCl3 (*) were covered with fibrin fibers (arrows) already 2 min after FeCl3 injury. Bar: 500 nm.

FeCl3 causes endothelial denudation without exposure of the inner layers of the vessel

On cross-sections, FeCl3 was found to diffuse through the entire vessel wall at the site of filter application, as evaluated by Prussian blue staining (Fig. 5A). The classical ultrastructure of a control artery (Fig. 5Ba) displays a luminal monolayer of endothelial cells, a media containing smooth muscle cells and an adventitia rich in collagen fibers with a typical periodicity of 67 nm. FeCl3 treatment caused a loss of undulation and severe damage in all layers of the vessel wall (Fig. 5Bb). The endothelial cells appeared empty with a fragmented plasma membrane and loss of organelles, smooth muscle cells in the media were flat and unconnected and collagen fibers in the adventitia exhibited an alteration of the striated pattern (Fig. 5B insert). It is noteworthy that the internal elastic lamina (IEL) was intact, suggesting that the components of the inner layers of the vessel had not been exposed to circulating blood. Higher concentrations of FeCl3 (up to 20%) or longer application times (up to 5 min) did not induce IEL rupture or exposure of the media or adventitia (Fig. S1). These results indicate that after FeCl3 injury only the basement membrane components are exposed to flowing blood and are therefore susceptible to contributing to platelet adhesion and thrombus growth. The loss of periodicity of the collagen fibers in the deep layers upon FeCl3 treatment raises the question of the reactivity of known platelet adhesive proteins exposed to FeCl3.

Figure 5.

 Histological and ultrastructural analyses of the vascular wall following exposure to FeCl3. Carotid artery injury was induced in mice as indicated in Fig. 1 and fixation with 2.5% glutaraldehyde was performed 5 min after removal of the filter. (A) Prussian blue-stained cross-sections of the carotid artery treated (+FeCl3) or not (control) with FeCl3 showed diffusion of iron throughout the vessel wall with loss of the inner vessel undulation. An arrow indicates the site of filter application. Scale bars: 50 μm. (B) Transmission electron microscopy cross-section images of FeCl3-injured carotid arteries. (a) In an untreated artery (control), a normal structure was observed with a monolayer of endothelial cells and a media containing smooth muscle cells. The insert shows the presence of collagen fibers in the adventitia (box) with a typical periodicity of 67 nm. (b) FeCl3 treatment (+FeCl3) resulted in a loss of undulation and damage to the vessel wall layers, although the internal elastic lamina remained intact. The insert shows alterations in the striated pattern of collagen fibers in the adventitia. Bars: 2 μm. E, endothelial cells; SMC, smooth muscle cells; IEL, internal elastic lamina.

Effect of FeCl3 on the function of adhesive proteins

We evaluated whether application of 0.0002–0.4% FeCl3 for 2 min could affect the function of several adhesive proteins in both static and flow-based assays. In agreement with in vivo observations, FeCl3 treatment resulted in loss of periodicity, destructuration and shrinkage of the collagen fibers at 0.1% and higher concentrations (Fig. 6A inserts). In the static assay, the number of platelets adhering to 0.1% FeCl3-treated collagen, fibrinogen and VWF was reduced by 59%, 81% and 55%, respectively (Fig. 6A,B). Similarly, 0.1% FeCl3 resulted in an 82% and 38% reduction in platelet adhesion under flow to fibrinogen at 300 s−1 and VWF at 1500 s−1, respectively, whereas thrombus formation on collagen was abolished (Fig. 6C). These results indicate that FeCl3 strongly alters the function of these adhesive proteins and their capacity to recruit platelets.

Figure 6.

 Adhesion of washed platelets to adhesive proteins pretreated with FeCl3. (A–B) Platelet adhesion to FeCl3-treated surfaces under static condition. Coverslips were coated with collagen (2.5 mg mL−1), fibrinogen (50 μg mL−1) or human von Willebrand factor (VWF) (10 μg mL−1) with botrocetin (1 μg mL−1), washed in PBS and blocked with 0.1% human albumin, before being treated for 2 min with FeCl3 (0.002–0.1%). Washed platelets were then allowed to adhere to the coverslips for 30 min at 37 °C. After fixation, the samples were prepared for scanning electron microscopy and the number of adherent platelets was determined in five random fields. (A) Images are representative of platelet adhesion to collagen in the presence or absence of FeCl3. The inserts illustrate the structural features of collagen. (B) Quantification of platelet adhesion to FeCl3-treated collagen, fibrinogen or VWF. (C) Platelet adhesion to FeCl3-treated surfaces under flow conditions. Hirudinated whole blood was perfused through the coated capillaries at 300 s−1 over fibrinogen and at 1500 s−1 over VWF and collagen. For flows over collagen, DIOC6-labeled platelet adhesion was observed by epifluorescent microscopy in the absence (black line) or presence of 0.1% and 0.4% FeCl3 (dark gray and light gray lines, respectively). Tracings represent the mean thrombus area (μm2) measured at each time. The dotted lines represent the SEM at each time point. For flows over fibrinogen and VWF, platelet adhesion was observed with differential interference contrast (DIC) microscopy. The line graphs represent the total number of adherent platelets at different time points. Platelet adhesion to collagen, fibrinogen and VWF was strongly inhibited by 0.1% or 0.4% FeCl3 treatment as compared with untreated adhesive proteins. Results are the mean values ± SEM from three separate experiments.

Impact of GPVI depletion and β1 integrin and VWF deficiency on FeCl3-induced thrombosis

As FeCl3 affects the adhesive properties of collagen and VWF and the periodicity of collagen, we wondered whether these adhesive proteins and their receptors significantly contributed to thrombosis in the FeCl3 model. Using 7.5% FeCl3 injury of the carotid artery, real-time video-microscopic observation revealed that GPVI-immunodepleted (JAQ1, 50 μg per mouse) and β1−/− mice presented no significant reduction in thrombus size as compared with the controls (Fig. 7A). At higher FeCl3 concentration (20%), thrombus formation and time to occlusion were also normal in the carotid artery of GPVI-immunodepleted mice (Fig. 7B) and thrombus formation and growth unaltered in β1−/− mice (Fig. 7C). Similarly, 10% FeCl3 applied to mesenteric arterioles resulted in a normal thrombus formation in JAQ1-treated mice (Fig. S2A). Examination of the corresponding injuries showed an intact internal elastic lamina (Figs S1 and S2B). Altogether, these results indicate that collagen-dependent platelet adhesion and activation via GPVI and β1 integrins are not critical for FeCl3-induced thrombosis under the various experimental conditions tested. In contrast, VWF−/− mice displayed a 50% reduction in thrombus size with a normal onset time, indicating that this glycoprotein plays a significant role in thrombus growth after a FeCl3-induced injury (Fig. 7A).

Figure 7.

 Role of platelet glycoprotein VI (GPVI), integrin β1 and von Willebrand factor (VWF) in carotid artery thrombosis induced by FeCl3. (A) FeCl3 was applied at 7.5% for 2 min on top of the carotid artery in control (wild-type, WT) or GPVI-immunodepleted (JAQ1-treated) mice (n = 5), β1−/− (n = 5) or VWF−/− mice (n = 10). (B) FeCl3 was applied at 20% for 5 min laterally to the vessel in control (n = 10) or GPVI-immunodepleted mice (JAQ1 treated) (n = 9). For each thrombus formed, both the thrombus surface area (a) and the time to occlusion (b) were evaluated. Tracings represent the mean surface area ± SEM of developing thrombi. (C) FeCl3 was applied at 20% for 5 min laterally to the carotid artery in wild-type (n = 5) or β1−/− mice (n = 5). Panels depict the mean thrombus surface area as a function of time. The dotted lines correspond to the SEM at each point.

Potential contribution of circulating blood cells to vascular injury

It has been suggested from ex vivo experiments that FeCl3-induced vessel damage occurs indirectly upon erythrocyte hemolysis [6]. To evaluate the direct effect of FeCl3 on the vessel wall independently of the blood components, carotid artery injury was induced in mice by application of 7.5% FeCl3 following whole-body saline perfusion and the vessel was fixed immediately after removal of the filter. Under these conditions, FeCl3 altered the vascular wall to the same extent as in the presence of circulating blood cells (Figs 8A and 5Bb). Interestingly, the spherical bodies were also present, although no rough material was detected at their surface (Fig. 8B). This was probably due to removal of blood plasma by PBS perfusion, suggesting that the rough material could be precipitated plasma proteins, cell debris or microparticles released from damaged cells. These results indicate that in vivo, blood cells probably do not play an important role in FeCl3-induced vascular damage.

Figure 8.

 Direct effect of FeCl3 on the vessel wall. Carotid artery injury was induced in mice following in vivo whole-body saline perfusion and fixation was performed immediately after removal of the filter. (A) In the absence of circulating blood cells, the vessel wall was altered to the same extent as in Fig. 5(Bb). The insert shows alterations in the striated pattern of collagen fibers similar to the images in Fig. 5(Bb). Bar: 2 μm. (B) The iron-filled spheres were likewise present, although no rough material was detected at their surface. Scale bars: upper panel, 5 μm; lower panel, 1 μm.


The present study provides new insights into the mechanism by which FeCl3 induces vascular injury and thrombus formation. Using an ultrastructural approach, we observed that shortly after topical application of FeCl3, numerous spherical bodies appeared to be budding off the endothelium into the vessel lumen. These bodies contained ferric ions and may participate in thrombus formation by supporting platelet adhesion and exposing large amounts of tissue factor, which can generate thrombin. FeCl3 treatment caused severe damage in all layers of the vessel wall but without rupturing the internal elastic lamina. Hence only the basement membrane components were exposed to circulating blood cells. In vitro experiments indicated that FeCl3 induced significant protein damage, resulting in a decreased reactivity of the subendothelial proteins.

Although the FeCl3 injury model is a well-accepted experimental system to study thrombosis, little is known about the sequential vascular lesions and the determinants initiating thrombus formation. Electron microscopy revealed a time-dependent gradation in the degree of FeCl3-induced endothelial cell lesion, from minor damage at 2 min to complete denudation at 8 min. This latter observation is in agreement with previous studies showing that FeCl3 injury results in endothelial cell removal [5,23]. Recently, Tseng et al. [5] described, 10 and 30 min after FeCl3 injury, the presence in endothelial cells of small (100 nm diameter) endocytic-exocytic vesicles containing ferric ions. We observed similar small structures 10 min after vascular injury (data not shown). Moreover, at earlier stages (2 min) we detected the presence of much larger spheres (3 μm diameter), which also contained ferric ions. As these spherical bodies were not entirely membrane enclosed, they differed from the previously described endocytic-exocytic vesicles and rather appeared to be formed by ferric ion concentrates diffusing through the vessel wall, before pushing the luminal endothelial cell membrane to form a FeCl3-filled bud. Platelets appeared to adhere to the rough material present on the endothelial-derived spheres, suggesting that they could participate in the initiation of thrombus formation. While the precise nature of this surface remains to be determined, immunogold labeling showed very strong staining of TF at the surface of the bodies, suggesting that important amounts of thrombin could be generated. This finding is consistent with the dramatic inhibition of FeCl3-induced thrombus formation observed in the presence of thrombin inhibitors such as hirudin. In addition, fibrin fibers could be detected on these spheres already 2 min after arterial injury, at a time point when platelets adhered to the site of injury. Considering the platelet receptors and ligands mediating the initial platelet recruitment in this model, GPIbα has been reported to be crucial [24]. We hypothesize that the fibrin network also participates in platelet recruitment. This could occur via the αIIbβ3 integrin and our studies on αIIb−/− mice showed a strong decrease in thrombus growth as well as in the number of platelets that attached at early time points. By comparison, thrombus formation in VWF−/− mice was reduced but not absent, suggesting that VWF was not absolutely required to initiate platelet adhesion in this model, in agreement with previous reports [25]. Whether endothelial-derived membranes covering the bodies provide other adhesive receptors able to recruit platelets, for example of the CAM family, or whether local release of proteins from damaged endothelial cells contributed to platelet adhesion in this model, remains to be explored [26].

One important finding of this work was that FeCl3 injury left the internal elastic lamina intact and continuous, which suggests that only the basement membrane components were exposed to blood and could contribute to thrombus formation. This observation was not restricted to one particular experimental condition (7.5% FeCl3 for 2 min), because we could show that the carotid internal elastic lamina was still not ruptured at much higher FeCl3 concentrations (up to 20%), for longer application times (up to 5 min) and upon injury of mesenteric arterioles. These findings suggest that the adhesive proteins present underneath the internal elastic lamina, such as type I collagen, were not exposed to flowing blood and therefore did not participate in thrombus formation. Moreover, in vitro studies demonstrated that FeCl3 affected the structure and function of several major adhesive proteins, including fibrinogen, VWF and collagen. It was not possible to estimate in vivo the concentration of FeCl3 to which the subendothelial proteins were exposed, but similar destructuration of collagen fibers was observed in vitro and in vivo, suggesting that the concentration of FeCl3 present at the site of the lesion would be sufficient to affect the structure and the reactivity of collagens exposed after endothelial cell removal. Real-time intravital microscopy experiments showed that mice lacking GPVI and β1 integrins did not present a significant decrease in thrombus formation in this model under all the conditions tested (7.5–20% FeCl3, 2–5 min application, in carotid or mesenteric arteries), suggesting that these receptors do not play a critical role in this model. These results are in apparent contradiction to those reported by Dubois et al. [7] and Massberg et al. [18] showing a role for GPVI in this FeCl3 model. The reasons for this discrepancy are not clear. One possibility could be that the degree of lesion might differ despite use of similar conditions, but this point is difficult to settle in the absence of comparable histological evaluation. Other differences might be related to the method used to quantify thrombus formation, because Dubois et al. and Massberg et al. have evaluated the time to occlusion. However, we could not measure any difference in carotid artery occlusion time between GPVI-immunodepleted and control mice, which appears to rule out differences due to quantification procedures. Whether differences in genetic background could provide another line of explanation is also possible, as suggested by the work of Kunicki and colleagues [15], who found that modifier genes could influence the extent to which thrombosis is affected by the absence of GPVI. Together these results suggest that GPVI and β1 integrins that participate in platelet adhesion in several thrombosis models [18,22,27] do not play a critical role in FeCl3-induced thrombus formation.

It has been recently proposed that FeCl3-induced vessel injury is indirect and occurs as a consequence of erythrocyte hemolysis [6]. This conclusion contrasts with our findings that FeCl3 injuries in exsanguinated mice perfused with saline were identical to those in control mice, indicating that FeCl3 itself causes major vascular lesions. The reason for this discrepancy is unclear but could be linked to the model employed because Woollard et al. made their observations in an ex vivo perfusion chamber, whereas we used an in vivo model. These authors described the presence of a high proportion of erythrocytes in the thrombus and the occurrence of thrombus formation even in the presence of hirudin. Such results are surprising because we and others have observed that FeCl3 injury led to the formation of platelet-rich thrombi in vivo, which were highly sensitive to thrombin blockers [10,11].

In conclusion, FeCl3-induced arterial injury recapitulates the following features in (i) exposure of the basement membrane constituents but not the inner layers of the vessel wall, (ii) significant damage to and loss of in vitro function of subendothelial proteins and (iii) attachment of platelets to bodies containing ferric ions. The present study also suggests that this model should be used cautiously and combined with alternate models, in particular to study the earliest stage of thrombus formation, as it may not be sufficient to evaluate the role of subendothelial adhesive proteins and their platelet receptors in thrombus formation. In contrast, the ferric chloride model is very useful to evaluate the importance of soluble agonists and plasma proteins in thrombus growth. It remains a valuable tool to screen inhibitors of platelet activation and aggregation or antagonists of platelet receptors and thereby determine their importance in thrombogenesis.


A. Eckly, B. Hechler, P. Mangin, M. Freund and M. Zerr conceived and designed the research, acquired the data, analyzed and interpreted the data and wrote the manuscript. J.-P. Cazenave and F. Lanza handled funding and carried out critical revision of the manuscript. C. Gachet conceived and designed the research, wrote the manuscript and handled funding and supervision.


The authors wish to thank J.-Y. Rinckel, F. Proamer, S. Magnenat, C. Bourdon, C. Ziessel and S. Roux (INSA, Strasbourg) for technical assistance. This work was supported by INSERM, EFS-Alsace, ARMESA (Association de Recherche et Développement en Médecine et Santé Publique), Fondation de France (Grant 2007001964) and the Agence Nationale de la Recherche (Grant ANR-06-PHYSIO-036-01). B. Hechler was the recipient of a ‘contrat d’interface’ between EFS and INSERM.

Disclosure of Conflict of Interests

The authors state that they have no conflict of interest.