von Willebrand factor (VWF), a multimeric protein with a central role in hemostasis, has been shown to interact with complement components. However, results are contrasting and inconclusive. By studying 20 patients with congenital thrombotic thrombocytopenic purpura (cTTP) who cannot cleave VWF multimers because of genetic ADAMTS13 deficiency, we investigated the mechanism through which VWF modulates complement and its pathophysiological implications for human diseases. Using assays of ex vivo serum-induced C3 and C5b-9 deposits on endothelial cells, we documented that in cTTP, complement is activated via the alternative pathway (AP) on the cell surface. This abnormality was corrected by restoring ADAMTS13 activity in cTTP serum, which prevented VWF multimer accumulation on endothelial cells, or by an anti-VWF Ab. In mechanistic studies we found that VWF interacts with C3b through its three type A domains and initiates AP activation, although assembly of active C5 convertase and formation of the terminal complement products C5a and C5b-9 occur only on the VWF-A2 domain. Finally, we documented that in the condition of ADAMTS13 deficiency, VWF-mediated formation of terminal complement products, particularly C5a, alters the endothelial antithrombogenic properties and induces microvascular thrombosis in a perfusion system. Altogether, the results demonstrated that VWF provides a platform for the activation of the AP of complement, which profoundly alters the phenotype of microvascular endothelial cells. These findings link hemostasis-thrombosis with the AP of complement and open new therapeutic perspectives in cTTP and in general in thrombotic and inflammatory disorders associated with endothelium perturbation, VWF release, and complement activation.
Evidence is accumulating that the coagulation system interacts with complement, a complex network of proteins that is part of the innate immunity (1). Thrombin and plasmin cleave C3 and C5, whereas the endothelial glycoprotein thrombomodulin (THBD), which inhibits thrombin, also modulates complement by favoring C3b inactivation and the degradation of the C3a and C5a anaphylatoxins (2–4). Moreover, recent evidence indicated that multimeric von Willebrand factor (VWF), which initiates hemostasis by promoting platelet adhesion, interacts at various levels with complement components (5, 6). Other studies suggested that VWF may favor C3b degradation by factor I (FI) (7). However, the precise molecular mechanism through which VWF modulates complement and its pathophysiological implications for human diseases have not been clarified yet.
Much of our knowledge about the interaction between the coagulation and complement systems derives from studies in patients with atypical hemolytic uremic syndrome (aHUS) and thrombotic thrombocytopenic purpura (TTP), two rare diseases characterized by thrombocytopenia, microangiopathic hemolytic anemia, and widespread microvascular thrombi that result in multiorgan dysfunction (8, 9).
aHUS is associated with genetic defects of regulatory proteins of the complement alternative pathway (AP), factor H (FH), FI, membrane-cofactor protein or THBD, anti-FH autoantibodies, or gain-of-function mutations in genes encoding the components of the AP C3 convertase, C3, and factor B (FB) (8, 10–13). These defects result in cell surface–restricted complement activation, leading to microvascular endothelial injury and activation of platelets and coagulation (14, 15). The role of complement in aHUS was supported fundamentally by clinical trials documenting that the anti-C5 mAb eculizumab prevented terminal complement pathway activation and protected from microvascular thrombosis, dramatically improving patient outcomes (16).
In contrast, TTP is associated with severe deficiency in ADAMTS13, the plasma metalloprotease that cleaves VWF multimers (17, 18). Severe ADAMTS13 deficiency in TTP is mostly caused by autoantibodies (19). In 5–10% of cases, the deficiency is caused by mutations in the ADAMTS13 gene (congenital TTP [cTTP]) (20). The earlier described defects result in little or no cleavage of endothelial cell–secreted/anchored ultra-large m.w. VWF (ULVWF) multimers (5). These hyperadhesive ULVWF multimers may spontaneously aggregate platelets, leading to the formation of microvascular thrombi in several organs. Several publications have highlighted the fact that complement activation via the AP may occur in TTP patients (21–27). Most data relate to patients with acquired ADAMTS13 deficiency; however, a role of complement has been suggested in cTTP as well (21, 28). It is less clear whether complement activation in TTP is a pathogenetic factor or an epiphenomenon.
Using an assay of serum-induced complement activation on endothelial cells that we recently developed in patients with aHUS (15), we provide in this article evidence that in cTTP, complement is activated via the AP at endothelial cell level. This abnormality was fully corrected by the replacement of ADAMTS13 activity in cTTP serum, which prevented the accumulation of VWF multimers on endothelium. This finding indicated a direct interaction between VWF and the complement system, and prompted us to investigate the underlying molecular mechanisms. We found that VWF interacts with C3b through A1, A2, and A3 domains and provides a nucleus for the activation of the AP. Assembly of active C5 convertase and formation of the terminal complement products C5a and C5b-9 occurs on VWF-A2, but not on VWF-A1 and VWF-A3 domains. Finally, we document that VWF-mediated activation of the complement terminal pathway alters endothelial antithrombogenic properties and initiates microvascular thrombosis.
Materials and Methods
TTP was diagnosed in all cases reported to have one or more episodes of microangiopathic hemolytic anemia and thrombocytopenia defined on the basis of hematocrit <30%, hemoglobin <10 g/dl, platelet count <150,000/μl, serum lactate dehydrogenase >460 IU/l, undetectable haptoglobin, and the presence of fragmented erythrocytes in the peripheral blood smear, with or without neurologic symptoms and with or without acute renal impairment during bouts. cTTP was diagnosed in patients with a severe deficiency of ADAMTS13 activity [<10% as measured by collagen binding assay (CBA) (29) or by fluorescence resonance energy transfer, FRETS-rVWF73 (30)] in remission and absence of inhibitory anti-ADAMTS13 autoantibodies as tested by CBA (31). Acquired TTP (aTTP) was defined in patients with <10% ADAMTS13 activity with inhibitory Abs.
Controls were 50 healthy subjects. The protocol was approved by the Ethical Committee of the Azienda Sanitaria Locale Bergamo, Italy (U0148858/III). Participants or their legal guardians provided written informed consent.
Serum and plasma complement profile
Complement C3 and C4 levels in serum were measured by nephelometry. SC5b-9 levels were evaluated in plasma by MicroVue SC5b-9 Plus EIA (SC5b-9 Plus; Quidel). For the latter test, blood was collected in ice-cold EDTA tubes and immediately centrifuged at 4°C to avoid ex vivo complement activation. Plasma was quickly separated and frozen at −80°C until assay. Levels below or above the limits of normal ranges (defined as mean ± 2 SD of the laboratories of the “Azienda Ospedaliera Papa Giovanni XXIII,” Bergamo, Italy, for C3 and C4 [C3, 83–180 mg/dl; C4, 15–40 mg/dl] and as mean ± 2 SD of 50 healthy control subjects tested at least twice for SC5b-9 [127–400 ng/ml]) were considered abnormal.
Ex vivo studies with HMEC-1 and TTP serum
The ex vivo studies used in this research were described in detail previously (15). The human microvascular endothelial cell line of dermal origin (HMEC-1, a kind gift from Dr. E. Ades and F.J. Candal of the Centers for Disease Control and Prevention and Dr. T. Lawley of Emory University, Atlanta, GA) was cultured as described previously (15). For the experiments, HMEC-1 were plated on glass slides and used when confluent. Cells were resting or activated with 10 μM ADP (Sigma) for 10 min and then were incubated for 4 h with serum diluted 1:2 with test medium (HBSS with 0.5% BSA). At the end of the incubation, HMEC-1 were fixed in 3% paraformaldehyde and stained with the following specific Abs: FITC-conjugated rabbit anti-human C3c-complement (Dako), rabbit anti-human complement C5b-9 complex (Calbiochem) followed by FITC-conjugated secondary Ab (Jackson ImmunoResearch Laboratories), goat anti-human C4α (Santa Cruz Biotechnology) followed by Cy3-conjugated secondary Ab (Jackson ImmunoResearch Laboratories), rabbit polyclonal anti-human VWF (Dako) followed by Cy3-conjugated secondary Ab (Jackson ImmunoResearch Laboratories), and mouse anti-human THBD (R&D Systems) followed by Cy3-conjugated secondary Ab (Jackson ImmunoResearch Laboratories). Isotype-matched irrelevant Abs were used as negative controls. In each experiment, serum from a healthy control subject was tested in parallel with TTP serum.
VWF and C3 costaining was evaluated by a rabbit polyclonal anti-human VWF Ab (Dako) followed by a Cy3-conjugated secondary Ab and by a mouse anti-human C3 mAb (Abbiotec) followed by an FITC-conjugated secondary Ab.
Expression of receptors for the C5 cleavage product C5a (C5aR) on HMEC-1, either resting or activated with 10 μM ADP for 10 min, was analyzed using a mouse anti-human CD88 (anti-C5aR; AbD Serotec) followed by a Cy3-conjugated secondary Ab.
The effect of C5a on VWF and THBD staining was evaluated by challenging HMEC-1 with 200 ng/ml C5a (Complement Technology) for 10 min. The cells were then treated with a rabbit polyclonal anti-human VWF Ab (Dako) or mouse anti-human THBD (R&D Systems) followed by the specific Cy3-conjugated secondary Abs.
A confocal inverted laser microscope (LSM 510 Meta; Zeiss) was used for the acquisition of the fluorescent staining on the endothelial cell surface. Fifteen fields, systematically digitized along the surface, were acquired using a computer-based image analysis system. The area occupied by the fluorescent staining was evaluated by automatic edge detection using built-in specific functions of the software ImageJ (National Institutes of Health [NIH], Bethesda, MD) and expressed as pixel2 per field analyzed. For each sample the mean of the 15 fields was calculated.
Recombinant VWF-A1A2A3 and FB cleavage assays
Recombinant VWF-A1A2A3 (rVWF-A1A2A3; 10 ng/μl) was mixed with 0.17 nM rADAMTS13 (R&D Systems) for 2 h at 37°C in 5 mM Tris-HCl buffer containing 1.5 M urea and 3 mM BaCl2 [pH 8]. The reaction was stopped by adding Na2SO4 (60 mM), and the proteins were subjected to 6% SDS-PAGE and transferred by electroblotting to polyvinylidene difluoride (PVDF) membrane (Bio-Rad). rVWF-A1A2A3 was visualized by Western blot (WB) with a mouse mAb directed against the C-terminal of the rVWF-A1A2A3 (kindly provided by Prof. Z.M. Ruggeri, The Scripps Research Institute, La Jolla, CA) followed by an HRP-conjugated anti-mouse secondary Ab. In these conditions, the intact rVWF-A1A2A3 of ∼80 kDa and the C-terminal fragment of ∼30 kDa generated by cleavage of rVWF-A1A2A3 are revealed (32).
A mixture of 500 ng/ml FB (Complement Technology) was incubated in the presence of two concentrations of rADAMTS13 (0.56 nM, 100 ng/ml and 0.17 nM, 30 ng/ml) for 2 h at 37°C in 5 mM Tris-HCl buffer containing 1.5 M urea and 3 mM BaCl2 [pH 8]. The reactions were stopped by adding Na2SO4 (60 mM), and the proteins were subjected to 6% SDS-PAGE and transferred by electroblotting to PVDF membrane (Bio-Rad). ADAMTS13 and FB were visualized by WB with goat anti-ADAMTS13 Ab (Bethyl Laboratories) followed by HRP-conjugated anti-goat Ab (Sigma) and rabbit anti-FB Ab (Atlas) followed by HRP-conjugated anti-rabbit Ab (Vector Laboratories), respectively. Signals were developed using ECL chemiluminescence detection system (Amersham).
C3 proconvertase and C3 convertase formation
A combined microplate and WB technique was used to generate and detect C3bB and C3bBb complexes (33). To specifically generate either C3bB or C3bBb complexes, we exploited the selective stabilization ability of Mn2+ (34) and Mg2+ (35), respectively (Supplemental Fig. 1A–D). Microtiter wells were coated with 3 or 5 μg/ml C3b (Complement Technology) in PBS by overnight incubation at 4°C, blocked with 1% BSA, 0.1% Tween 20 in PBS for 1 h at 37°C, and washed with a wash buffer (8.1 mM Na2HPO4, 1.8 mM NaH2PO4, 0.1% Tween 20, and 25 mM NaCl) supplemented with 2 mM MnCl2 or 10 mM MgCl2 for C3bB or C3bBb, respectively. C3bB(Mn2+) complexes were formed by incubating C3b-coated wells at 37°C for 2 h with FB (1000 ng/ml, 10.8 nM), in the absence or in the presence of rADAMTS13 (0.04 nM, 6.6 ng/ml; 0.11 nM, 19.8 ng/ml; 0.55 nM, 99 ng/ml), diluted in the assay buffer (8.1 mM Na2HPO4, 1.8 mM NaH2PO4, 0.1% Tween 20, and 75 mM NaCl) supplemented with 4% BSA and 2 mM MnCl2. C3bBb(Mg2+) complexes were formed by incubating C3b-coated wells at 25°C for 12 min with FB (1000 ng/ml, 10.8 nM) and FD (5 ng/ml, 0.22 nM; Complement Technology), in the absence or in the presence of rADAMTS13 (0.04 nM, 6.6 ng/ml; 0.11 nM, 19.8 ng/ml; 0.55 nM, 99 ng/ml), both diluted in assay buffer supplemented with 0.5% BSA and 10 mM MgCl2. After washes, the protein complexes were detached from microtiter wells with EDTA 10 mM and SDS 1%, subjected to 10% SDS-PAGE, and transferred by electroblotting to PVDF membrane (Bio-Rad). Proteins were detected with a rabbit anti-FB Ab (Atlas) or a goat anti-FB Ab (Quidel) followed by HRP-conjugated anti-rabbit or anti-goat Abs (Vector Laboratories) and the ECL system (Amersham). C3-proconvertase and C3-convertase formation were evaluated by the visualization by WB of the FB band (93 kDa) or the Bb band (60 kDa), respectively.
In additional experiments, C3bBb(Mg2+) complexes were formed as earlier in the presence or in the absence of rVWF-A1A2A3 (14.08 or 1408 ng/ml).
FH cofactor activity for FI-mediated proteolysis of fluid-phase C3b
The fluid-phase cofactor activities of FH and VWF were determined in a C3b proteolytic assay using purified proteins. C3b and FI were from Complement Technology, and FH was purchased from Merck. rVWF was a gift from Baxter. LFB-VWF was from LFB Biotechnologies. Emoclot-VWF was from Kedrion, and rVWF-A2 was from R&D Systems.
In brief, C3b, FH, VWF, and FI were mixed in 10 mM HEPES (pH 7.5), 150 mM NaCl, and 0.02% Tween 20. Final concentrations were: C3b 0.42 μM, FH 31.3 nM, FI 43 nM, and VWF 0.313 nM (1×, physiological molar ratio with FH) or 3.13 nM (10×) or 156.5 nM (500×). Molarities were calculated using the following masses: C3b, 180 kDa; FH, 150 kDa; FI, 88 kDa; VWF, 250 kDa; VWF-A2, 18 kDa. Mixtures were incubated at 37°C in a water bath for 10 or 1 min in the absence or in the presence of FH, respectively. The reactions were stopped by the addition of 4× Laemmli protein sample buffer for SDS-PAGE (Bio-Rad) supplemented with 10% 2-ME (Sigma), and the samples were analyzed in 10% SDS-PAGE. Gels were stained with ProteoSilver Stain Kit (Sigma), and proteolysis of C3b was determined by analyzing the disappearance of the α′-chain (110-kDa) band and the appearance of the α65- and α43-kDa bands. To check for FH contamination, we analyzed Emoclot-VWF, LFB-VWF, and rVWF on WB with a goat anti-FH Ab (Calbiochem) followed by HRP-conjugated anti-goat Ab (Sigma) and the ECL chemiluminescence detection system (Amersham).
C3b-VWF interaction by ELISA
Microtiter wells were coated with 5 μg/ml rVWF either multimeric or monomeric (reduced with DTT 30 mM), or plasma-purified LFB-VWF, purified rVWF-A1A2A3 domains, rVWF-A1 (U-Protein Express BV), rVWF-A2 (R&D Systems), or rVWF-A3 (U-Protein Express BV) in PBS by overnight incubation at 4°C. Then coated wells were blocked with 1% BSA for 1 h at room temperature (R.T.) and washed with PBS supplemented with 0.1% Tween 20 and 10 mM MgCl2. Different amounts of C3b ranging from 0 to 50 μg/ml were incubated with coated wells for 2 h at 37°C. In additional rVWF-A2–coated wells, C3b (5 μg/ml, molecular mass 180 kDa) was mixed with an equimolar concentration of rVWF-A2 (0.5 μg/ml, molecular mass 18 kDa). Protein complexes were detected by a mouse anti-human C3 Ab (Abbiotec) followed by HRP-conjugated anti-mouse Ab (Zymed). In additional experiments, the binding of C3b in HBSS supplemented with 1.9 mM CaCl2 and 0.8 mM MgSO4 on rVWF-A2–coated wells was evaluated.
C4b was added at increasing concentrations (0–10 μg/ml) to rVWF-A1 or rVWF-A2 or rVWF-A3 immobilized on plastic wells. Binding was detected with a goat anti-human C4α followed by an HRP-conjugated anti-goat Ab (Sigma).
In additional experiments, microtiter wells were coated with 5 μg/ml C3b (Calbiochem) in PBS by overnight incubation at 4°C, blocked with 1% BSA for 1 h at R.T., and washed with PBS supplemented with 0.1% Tween 20 and 10 mM MgCl2. Different amounts of purified rVWF-A1A2A3 or rVWF-A2 domains of VWF ranging from 50 to 0 μg/ml or from 5 to 0 μg/ml (dilution 1:2 in PBS), respectively, were incubated with C3b-coated wells for 2 h at 37°C. Protein complexes were detected by HRP-conjugated anti-human VWF Ab (Dako).
Color was developed using tetramethylbenzidine (TMB) substrate (Bethyl Laboratories), stopped with H2SO4 2 M, and absorbance was measured at 450 nm. BSA was considered as an irrelevant negative control.
Surface plasmon resonance
Surface plasmon resonance (SPR) analyses were carried out on a Biacore X100 (GE Healthcare). Purified plasma-derived VWF (LFB) or rVWF-A2 were immobilized (5100 or 2600 resonance units, respectively) on a CM5 chip by amine coupling protocol according to the standard procedure. Purified C3b (500 nM or 1 μM) was flowed for 120 s in HBSS (pH 7.4), supplemented with MgCl2 or CaCl2 at a flow rate of 10 μl/min. Chip surface was regenerated using 100 mM sodium acetate, 500 mM NaCl (pH 4), and in each cycle the signal was subtracted from reference cell with no immobilized protein.
C3 was purified as previously described (36) and cleaved to C3b by C3-convertase cleavage.
AP C3 and C5 convertase assembly, C3 convertase decay, and C5b-9 formation in the presence of normal human serum
To obtain C3 convertase formation by using a physiological mixture of complement components, we coated microtiter wells with 5 μg/ml purified rVWF-A1A2A3, monomeric rVWF, properdin (Complement Technology), C3b, fibrinogen (MyBioSource), rVWF-A1, or with 1, 2.5, or 5 μg/ml rVWF-A2 or rVWF-A3 in PBS, or PBS alone by overnight incubation at 4°C, blocked with 0.5% BSA in PBS for 1 h at R.T., and washed with wash buffer supplemented with 5 mM MgCl2. C3bBb(Mg2+) complexes were formed by incubating coated wells at 37°C for 30 min with 20% normal human serum (NHS; a pool obtained from 15 healthy volunteers) diluted in PBS containing 5 mM MgCl2 in the presence or absence of 1 mM EGTA or 5 mM EDTA, 150 μg/ml soluble complement receptor 1 (sCR1), or 100 and 200 μg/ml FH. After washes, formed C3bBb(Mg2+) complexes and bound FH and complement components were detached from wells by incubation with 10 mM EDTA and 1% SDS for 1 h at R.T., subjected to 10% SDS-PAGE under reducing conditions, and transferred by electroblotting to a PVDF membrane (Bio-Rad). Proteins were detected with polyclonal rabbit anti-human FB (Atlas) or goat anti-human FB (Quidel), rabbit anti-C3c (Dako), or goat anti-FH (Calbiochem) Abs, followed by HRP-conjugated anti-rabbit (Vector Laboratories) or anti-goat (Sigma) Abs and the ECL chemiluminescence detection system (Amersham). C3bBb formation was evaluated by the visualization by WB of the Bb band (60 kDa) and C3b β-chain (76 kDa). In addition, C5b formation was evaluated by using a polyclonal rabbit anti-human C5/C5b Ab (Abcam), followed by HRP-conjugated anti-rabbit Ab (Vector Laboratories). C5b was visualized as α′-chain C5b (104 kDa) band. C6 (105 kDa) and C9 (66 kDa) were detected with rabbit polyclonal anti-C6 (Abgent) or goat-polyclonal anti-C9 (Origene) Abs, respectively, followed by HRP-conjugated secondary Abs. In additional experiments, C3bBb complexes were formed on rVWF-A2 or properdin as described earlier, then were washed and allowed to decay in PBS with 5 mM MgCl2 with or without 60 μg/ml FH for 15, 30, and 60 min at 37°C. The remaining complexes were detached and detected as described earlier. Densitometric analysis of Bb band was performed using the software ImageJ (NIH), and results are expressed as pixel2.
In further experiments, properdin or C3b or rVWF-A2–coated wells were incubated with NHS in the absence or presence of 1 mM EGTA or 150 μg/ml sCR1 or with PBS only. C5b-9 complexes were detected by HRP-conjugated anti–C5b-9 Ab (Quidel) followed by TMB substrate (Bethyl Laboratories). Color was stopped with H2SO4 2 M, and absorbance was measured at 450 nm. Each reaction was performed in triplicate and the OD values averaged.
AP C3 convertase assembly with purified proteins
Microtiter wells were coated with 3 μg/ml C3b or 5 μg/ml rVWF-A1, rVWF-A2, rVWF-A3, or fibrinogen in PBS. After washing, coated wells were incubated with 50 μg/ml C3b in PBS with 5 mM MgCl2 with or without 5 mM CaCl2 for 2 h at 37°C, followed by washing. C3bBb(Mg2+) complexes were formed by incubating wells at 25°C for 12 min with FB (1000 ng/ml) and FD (5 ng/ml), with or without properdin (250 ng/ml), EDTA (5 mM), FH (2640 ng/ml) diluted in assay buffer with 0.5% BSA and 10 mM MgCl2 with or without 5 mM CaCl2. The protein complexes were detached and detected as described earlier.
Detection of rVWF-A2 coated on plastic wells
The amount of rVWF-A2 that coated plastic wells after incubation with PBS in the absence or presence of 1 mM EGTA was detected by ELISA using a rabbit anti-VWF Ab (Dako), followed by HRP-conjugated anti-rabbit Ab (Vector Laboratories) and the TMB substrate (Bethyl Laboratories). Color was stopped with H2SO4 2 M, and absorbance was measured at 450 nm. Each reaction was performed in triplicate and the OD values averaged.
On-Cell Western assay
Quantitative measurements of the level of surface expression of VWF and THBD on HMEC-1 were performed by On-Cell Western using the Odyssey Infrared Imaging System CLx (LI-COR Biosciences) (37).
HMEC-1 were seeded in 96-well microplates and used 96 h after seeding. HMEC-1 monolayers were activated with 10 μM ADP for 10 min, washed, and incubated for 4 h with serum from control or from patients with cTTP diluted 1:2 with test medium. Selected samples were incubated in the presence or absence of sCR1 (150 μg/ml) or eculizumab (100 μg/ml).
Each sample was tested in triplicate. HMEC-1 were washed, fixed in 3% paraformaldehyde, and blocked for 1 h (Odyssey blocking buffer; LI-COR Biosciences). The cells were stained with rabbit polyclonal anti-human VWF (Dako) followed by the secondary Ab IRDye 800 CW goat anti-rabbit IgG (H + L) (LI-COR Biosciences), or with mouse anti-human THBD (R&D Systems) followed by the secondary Ab IRDye 800 CW goat anti-mouse IgG (H + L) (LI-COR Biosciences). Background was determined on cells incubated with control serum without primary and secondary Abs. Background was subtracted and the fluorescence intensity was normalized with the cell number: after the acquisition of IRDye 800 staining, cells were permeabilized with PBS supplemented with 0.1% Triton X-100 (Sigma) and then challenged with CellTag 700 Stain (LI-COR Biosciences), a near-infrared fluorescent nonspecific cell staining.
The levels of soluble THBD (sTHBD) were evaluated in plasma collected on EDTA using a commercial kit (R&D Systems).
Thrombus formation under flow condition
The assay was performed as previously described (38). HMEC-1 were treated with ADP and exposed for 4 h to serum from patients or controls diluted 1:2 with test medium in static conditions. Perfusion of heparinized whole blood (heparin 10 U/ml) from healthy subjects (added with the fluorescent dye mepacrine 10 μM, to label platelets) was then performed in a thermostatic flow chamber (37°C) in which one surface of the perfusion channel was a glass slide seeded with a monolayer of endothelial cells at a constant flow rate of 1500 s−1 (60 dynes/cm2). After 3 min, perfusion was stopped, and the slide with the endothelial cell monolayer was dehydrated and fixed in acetone for 20 min. In parallel experiments, sCR1 was added to serum to inhibit complement.
Slides were examined under confocal inverted laser microscopy. Fifteen fields for each slide were systematically digitized along the surface, and the area covered by thrombi was quantified by ImageJ (NIH) and expressed as pixel2 per field analyzed. In each experiment, serum from one healthy control subject was tested in parallel to cTTP serum. For each sample, the mean of 15 fields (excluding the lowest and the highest values) was calculated.
The perfusion of whole blood was done after complement activation had already occurred on HMEC-1 during incubation with control or TTP serum; thus, it is unlikely that heparin present in whole blood could have affected complement deposits. To further exclude this possibility, we evaluated the effect of addition of 10 U/ml heparin to aHUS sera (as positive controls) on C5b-9 deposits on ADP-activated HMEC-1 and found no effect (control serum: 1944 ± 168, aHUS serum: 4038 ± 316, aHUS serum + heparin: 3738 ± 284 pixel2 per field; mean ± SE, n = 2, 15 fields each).
Results are expressed as mean ± SE. Data were analyzed by ANOVA (two-tailed) or by Kruskal–Wallis tests, as appropriate. Statistical significance was set at p < 0.05.
All patients with TTP referred to the International Registry of HUS/TTP (http://www.marionegri.it/it_IT/home/medico/ricerca_clinica/registri_patologia) were evaluated for inclusion in this study. Inclusion criteria were: 1) diagnosis of cTTP defined as one or more episodes of microangiopathic hemolytic anemia and thrombocytopenia in patients with a severe ADAMTS13 deficiency in remission and absence of anti-ADAMTS13 Abs (see 2Materials and Methods) (31); 2) homozygous or compound heterozygous ADAMTS13 mutations; and 3) availability of plasma and/or serum taken either during the acute phase (before plasma treatment) and/or in remission, at least 30 d after the last plasma treatment. Patients with renal involvement were screened for common aHUS-associated complement genes, which allowed us to exclude the presence of genetic abnormalities predisposing to complement dysregulation (28, 39, 40). Details of the ADAMTS13 mutations (most of them have been functionally characterized previously) (32, 39, 41) and the clinical parameters of the 20 cTTP patients are reported in Table I. Patients with aTTP (in hematological remission but with still severe ADAMTS13 deficiency and presence of ADAMTS13 inhibitors, n = 4) were also recruited (Supplemental Fig. 2A).
|Patient’s Code .||Sex .||Age of Onset (y) .||ADAMTS13 Mutations .||ADAMTS13 Activity (%) .||Disease Phase .||Platelets (150–400 × 103/μlb) .||LDH (266–500 IU/lb) .||Hb (14–18 g/dlb) .||Renal Impairmentc .||Serum C3 (83–180 mg/dlb) .||Plasma SC5b-9 (127–400 ng/mlb) .||Endothelial C3 Deposits (% of ctr)d .||Endothelial C5b-9 Deposits (% of ctr)d .|
|c.1786+1G > A he|
|Patient’s Code .||Sex .||Age of Onset (y) .||ADAMTS13 Mutations .||ADAMTS13 Activity (%) .||Disease Phase .||Platelets (150–400 × 103/μlb) .||LDH (266–500 IU/lb) .||Hb (14–18 g/dlb) .||Renal Impairmentc .||Serum C3 (83–180 mg/dlb) .||Plasma SC5b-9 (127–400 ng/mlb) .||Endothelial C3 Deposits (% of ctr)d .||Endothelial C5b-9 Deposits (% of ctr)d .|
|c.1786+1G > A he|
At the time of sampling.
Limits of normal ranges.
Renal impairment at onset or during relapses: acute kidney injury, need for dialysis, and/or urinary abnormalities (hematuria and/or proteinuria) in patients without previous renal involvement.
On ADP-activated HMEC-1.
p < 0.05 versus control (statistical comparisons were made for each patient by comparing deposits in pixel2 recorded in the 15 fields analyzed for the patient and for the corresponding control run in parallel).
This patient was not studied in remission because he was treated with eculizumab [Pecoraro et al. (28)].
Mutation not functionally characterized yet.
ac., acute phase; ctr, control; F, female; Hb, hemoglobin; he, heterozygous mutation; ho, homozygous mutation; LDH, lactate dehydrogenase; M, male; n.a., sample not available; n.d., not done; rem., remission phase.
Circulating complement profile in cTTP
We found lower than normal serum C3 levels in only 1 of 9 (11%) and 2 of 17 (12%) cTTP patients tested during the acute phase and at remission, respectively (Table I). C4 levels were normal in all patients.
Plasma levels of the cytolytically inactive terminal-complement complex (SC5b-9) were higher than normal in 57% (4 of 7) and 43% (6 of 14) of patients during the acute phase and at remission, respectively (Table I). Mean values of plasma SC5b-9 in cTTP patients were significantly higher than in healthy control subjects and comparable with values recorded in aHUS patients (Fig. 1).
TTP serum induced C3 and C5b-9 deposition on microvascular endothelial cells
To evaluate whether complement is activated at endothelial cell level in cTTP, we incubated resting or ADP-activated HMEC-1 with serum from cTTP patients or controls (15). ADP activation was used to mimic an activated/perturbed endothelium, resulting in exocytosis of P-selectin and VWF before incubation with serum (Fig. 2A) (15).
Sera from all cTTP patients studied during the acute phase (n = 4) induced more C3 and C5b-9 deposits (Fig. 2B–E, Table I) than control sera both on resting and on ADP-activated HMEC-1. On resting HMEC-1, C3 and C5b-9 deposits were comparable after exposure to serum from cTTP patients in remission (n = 4) or from controls (Fig. 2B, 2D), whereas on activated HMEC-1, cTTP sera taken both during the acute phase (n = 4) and in remission (n = 11) induced more C3 and C5b-9 deposits than control sera (Fig. 2C, 2E, Table I). Notably, the intensities of complement deposits on HMEC-1 exposed to cTTP sera were comparable with those on cells incubated with aHUS sera taken as positive controls (15) (Fig. 2B–E). HMEC-1 exposed to ADP alone without serum showed no C3 staining (Fig. 2A). Higher than normal C3 and C5b-9 deposits on ADP-activated HMEC-1 were also observed with sera from patients (n = 4) with aTTP (Supplemental Fig. 2).
C3 and C5b-9 deposits were blocked by adding the pan complement inhibitor sCR1 to cTTP serum (Fig. 2B–E). In addition, the AP inhibitors CR2FH (42) and FH [230 μg/ml, which approximately doubled the FH serum concentration (15)] significantly reduced cTTP serum-induced C3 deposits (n = 4; Fig. 2F). Sera from cTTP patients and controls induced comparable C4 staining on ADP-activated HMEC-1 (n = 3; Fig. 2G), confirming selective activation of the complement AP.
Restoration of ADAMTS13 activity prevented cTTP serum-induced C3 and C5b-9 deposition on microvascular endothelial cells
To investigate whether cTTP serum-induced complement activation was related to ADAMTS13 deficiency, we added recombinant ADAMTS13 (rADAMTS13; at the concentration of 4 μg/ml, which restored 100% ADAMTS13 activity in cTTP serum, as measured by CBA) to cTTP serum taken in remission. rADAMTS13 fully normalized cTTP serum-induced C3 and C5b-9 endothelial deposits (n = 3; Fig. 2H).
ADAMTS13 neither cleaves FB nor inhibits the formation of the C3 convertase
We next investigated the mechanism(s) through which ADAMTS13 deficiency in cTTP may affect complement AP activation.
Based on structural homology between the VWF type A (VWA) domain of FB and the A2 domain of VWF that contains the binding sites through which ADAMTS13 recognizes and cleaves VWF (http://www.rcsb.org/pdb) (43, 44), and the ∼19% amino acid sequence identity between the two domains, we investigated whether rADAMTS13 proteolyses FB, thus reducing substrate availability for formation of the AP C3 convertase C3bBb. In the presence of urea and BaCl2, rADAMTS13 efficiently cleaved rVWF-A1A2A3 fragment but had no effect on FB (Fig. 3A, 3B).
We also evaluated whether rADAMTS13 inhibits the formation of the C3bB proconvertase and C3bBb convertase. By using a microplate/WB assay (Fig. 3C) (33), we found no effect of rADAMTS13, added either at physiological or at molar excess ratios with FB and FD, on the formation of C3bB and C3bBb (Fig. 3D, 3E).
Neither ULVWF multimers nor normal plasma VWF multimers inactivate C3b
We then verified whether VWF had any effect on FI-mediated C3b degradation, as suggested by previous studies (7, 45). Neither normal plasma-derived VWF from two different commercial sources [LFB-VWF concentrate that has a multimeric pattern similar to normal plasma VWF (46); Emoclot-VWF that exhibits a lower content of high m.w. VWF multimers than normal plasma VWF (47)], rVWF [which includes ULVWF multimers, thus mimicking the multimeric pattern in TTP plasma (48)], nor the rVWF-A2 domain, either alone or in combination with FH, had any effect on C3b cleavage when tested both at physiological molar ratios with FH and FI and in large excess (Fig. 4A–C). Faint bands of iC3b products were only observed with 500 M excess of plasma VWF (Emoclot-VWF; Fig. 4C), which, however, also contained small amounts of contaminating FH (Fig. 4D).
VWF accumulation on ADP-activated microvascular endothelial cells exposed to cTTP serum favors C3 deposition
ULVWF multimeric strings secreted by, and anchored to, activated endothelial cells were found to bind C3b, other components of the AP (FB, FD, properdin), and C5 (6). Based on the earlier findings, we investigated whether the severe ADAMTS13 deficiency in cTTP patients leads to VWF accumulation on endothelial cell surface and whether VWF deposits may act as nuclei for C3 deposition.
On ADP-activated HMEC-1 incubated with cTTP serum taken during remission, we found intense VWF staining that partially colocalized with C3 (Fig. 5A), suggesting active interaction between VWF and C3 activation products. The addition of an anti-VWF Ab to cTTP serum during incubation with HMEC-1 prevented the excessive C3 deposition on ADP-activated HMEC-1, whereas an irrelevant Ab had no effect (n = 3; Fig. 5B).
These results indicate that in cTTP, VWF strings accumulating on endothelial cells because of severe ADAMTS13 deficiency participate in complement activation.
VWF interacts with C3b through the three type A domains and favors the formation of the AP C3 convertase
We then investigated the molecular mechanisms underlying VWF-C3b interaction and its effect on complement activation.
We first studied VWF-C3b interaction in static conditions by ELISA. Recombinant multimeric VWF and plasma-derived VWF poorly interacted with C3b (Fig. 5C), possibly because of the nonpermissive close conformation in static conditions, whereas we found a dose–response interaction of C3b with coated VWF monomer (Fig. 5C). We then examined whether VWF binds C3b through its A1A2A3 region. rVWF-A1A2A3 and C3b dose-dependently interacted with each other (Fig. 5D, 5E), and the same was observed between rVWF-A2 and C3b (Fig. 5F–I). In competition studies, addition of rVWF-A2 (at equimolar concentration with C3b) to the C3b solution inhibited C3b binding to coated rVWF-A2, confirming the specificity of the observed interaction (Fig. 5G). The presence of Ca2+ and Mg2+ in the ELISA buffer only modestly increased the C3b binding to coated rVWF-A2 (Fig. 5H). Recombinant VWF-A1 (rVWF-A1) and rVWF-A3 domains bound C3b, although to a lower degree than rVWF-A2 (Fig. 5I). Whether the rVWF-A1 and rVWF-A3 domains coated on wells were in a proper conformation cannot be established on the basis of present data.
No binding was found between the rVWF-A1, -A2, and -A3 domains and C4b (Fig. 5J).Through SPR we found that C3b binds specifically to plasma-derived VWF in flow condition, confirming previous data by Feng et al. (7). The interaction was dose dependent, rapidly reversible, and not affected by the presence of Ca2+ (Supplemental Fig. 3A, 3B). We also found a specific interaction between C3b and coated rVWF-A2 (Supplemental Fig. 3C, 3D).
We then studied whether the interaction between VWF and C3b results in the formation of the C3bBb AP C3 convertase. rVWF-A1A2A3–coated wells were incubated with 20% NHS as a source of complement. The products were detached from wells and analyzed by WB (Fig. 6A). C3bBb formation occurred on rVWF-A1A2A3–coated wells, as documented by specific C3b β-chain and Bb bands, respectively (Fig. 6B). C3bBb formation was observed in positive control wells coated with properdin (Fig. 6B) (49). No C3 convertase formed on rVWF-A1A2A3– or properdin-coated wells incubated with NHS treated with EDTA or sCR1 (Fig. 6B).
The C3 convertase assay was repeated on rVWF-A2–coated wells. Results showed specific bands of C3b and Bb that were fully abolished by EDTA (Fig. 6C). Similar results were observed when the experiments were repeated on rVWF-A1– or rVWF-A3–coated wells; however, the intensities of C3b and Bb bands were lower than those of the products from coated rVWF-A2 (Supplemental Fig. 4A–C). A Bb band could also be detected after the incubation of NHS in wells coated with rVWF monomer (Fig. 6D), whereas no products of C3 convertases could be recovered in wells coated with fibrinogen, a protein that does not bind C3b (50) (Fig. 6D, Supplemental Fig. 4).
The addition of EGTA to NHS reduced the C3b and Bb bands from rVWF-A2 (Fig. 6C, 6E), without altering the amount of rVWF-A2 coated on wells (Fig. 6F). EGTA also reduced C3 convertase formation on rVWF-A1 and rVWF-A3 (Supplemental Fig. 4B, 4C), although it did not affect the Bb band recovered from properdin-coated wells (Fig. 6C). However, FH added in excess to NHS dose-dependently reduced the amount of C3bBb (as Bb band) recovered from rVWF-A2–coated wells (Fig. 6E) and also accelerated its decay (Fig. 6G), similarly to its effect on C3bBb formed on properdin (Fig. 6H), confirming that rVWF-A domains favored the assembly of the AP C3 convertase.
We then repeated the C3 convertase experiments in a clean condition using purified C3b, FB, and FD proteins (Fig. 7A), and found specific AP C3 convertase formed on rVWF-A2, but not on fibrinogen (Fig. 7B). Interestingly, addition of properdin, which enhances the stability of AP C3 convertase (51), to the reaction mixture increased the amount of C3 convertase formed on rVWF-A2 (Fig. 7B). Consistent with the ELISA data, the presence of Ca2+ in the buffer during the first step of C3b binding to coated rVWF-A2 did not affect the amount of C3 convertase formed (Fig. 7C). C3 convertase formation on rVWF-A2 with purified complement components was increased when Ca2+ was present in both steps (C3b binding and C3 convertase formation; Fig. 7D). Either EDTA or FH fully prevented C3 convertase formation on rVWF-A2 (Fig. 7D). C3 convertase formed also on rVWF-A1 and rVWF-A3, although to lesser degree than on rVWF-A2 (Fig. 7B).
To evaluate whether circulating VWF may favor C3 convertase formation, we added rVWF-A1A2A3 with FB and FD on C3b-coated wells. The intensity of the Bb band of the reaction product was not affected by rVWF-A1A2A3 (Fig. 7E).
Altogether, these results indicate that surface-bound but not fluid-phase rVWF type A domains may act as a platform for the formation of the AP C3 convertase, with the A2 domain showing a higher activity than the A1 and A3 domains.
VWF-C3b interaction forms the AP C5 convertase and activates the terminal pathway
In preliminary experiments we documented that immobilized properdin or C3b in the microplate/WB assay formed an active C5 convertase in the presence of NHS (Supplemental Fig. 1E, 1F). To evaluate the impact of VWF-C3b interaction on the terminal complement pathway, we analyzed the products recovered from rVWF-A2–coated wells after incubation with NHS on WB with anti-C5, anti-C6, or anti-C9 Abs (Fig. 8A). Bands corresponding to the C5b α′-chain, C6, and C9 were observed in samples recovered from wells coated with rVWF-A2 or with properdin (Fig. 8B). No C5b, C6, or C9 bands could be recovered from rVWF-A2–coated wells exposed to EGTA-treated NHS (Fig. 8B). No activation products of C3 and C5 convertases could be recovered after the incubation of NHS in wells coated with fibrinogen (Fig. 8B), supporting the specificity of our findings. Finally, labeling the products on rVWF-A2–coated wells, after incubation with NHS, with an anti–C5b-9 Ab in an ELISA assay showed the formation of C5b-9 complexes (Fig. 8C). Notably, the absorbance values of C5b-9 formed in rVWF-A2–coated wells were comparable with absorbance values in C3b- or properdin-coated wells (Fig. 8C). The addition of sCR1 reduced the absorbance values in all three conditions, whereas EGTA reduced only the C5b-9 signal in rVWF-A2–coated wells (Fig. 8C). At variance, no activation products of C5 convertase could be recovered from wells coated with rVWF-A1 or rVWF-A3 (Supplemental Fig. 4D, 4E).
These results indicate that the rVWF-A2 domain, but not rVWF-A1 or rVWF-A3, favors the assembly of an active C5 convertase, the formation of C5b, and the activation of the terminal complex (Fig. 8D).
cTTP serum exerts a prothrombogenic effect on endothelial cells mediated by VWF and complement terminal pathway products
To investigate whether in the presence of ADAMTS13 deficiency VWF-induced complement activation perturbed the thromboresistant endothelial phenotype, we analyzed the effect of cTTP serum on VWF staining and THBD expression on ADP-activated HMEC-1, in the absence and presence of complement inhibitors.
HMEC-1 incubated with cTTP serum (taken in remission, n = 5) showed more intense VWF staining than cells incubated with control serum (Fig. 9A). The VWF staining intensity was normalized through supplementation of cTTP serum with ADAMTS13 (4 μg/ml corresponding to 100% protease activity) that cleaved VWF strings on the cell surface (Fig. 9A). Blocking either the entire complement cascade with sCR1 or the terminal pathway alone with eculizumab significantly limited the excessive VWF staining induced by cTTP serum (Fig. 9A). After exposure of ADP-activated HMEC-1 to cTTP serum, the intensity of THBD staining was significantly reduced relative to HMEC-1 exposed to control serum (n = 5; Fig. 9B). Addition of a serine protease inhibitor, aprotinin (38, 52), to cTTP serum significantly limited the loss of THBD staining, suggesting that cTTP serum induced THBD shedding from HMEC-1 (Fig. 9B). Also, eculizumab antagonized the effect of cTTP serum on THBD loss (Fig. 9B). cTTP serum-induced VWF increase and THBD loss on the HMEC-1 cell surface, as well as the role of complement activation products in these abnormalities, were confirmed by On-Cell Western assay, a quantitative immunofluorescent test for accurate measurement of protein levels in fixed cells (37) (Fig. 9C, 9D). In addition, plasma levels of sTHBD in acute cTTP patients (n = 8) were higher than in control plasma (n = 5). sTHBD levels decreased in remission (n = 13) but remained higher than in controls (Fig. 9E). Interestingly, in patient F1689#922, eculizumab treatment (28) reduced plasma sTHBD levels (acute pre-eculizumab: 16.28 ng/ml; after eculizumab: 6.24 ng/ml).
Together, the earlier results indicated that C5 activation products have a role in cTTP serum-induced VWF accumulation on HMEC-1 and in THBD loss from the HMEC-1 cell surface.
HUVECs and human microvascular endothelial cells have been reported to express the C5aR (53, 54). In this study, we found that HMEC-1 also express C5aR, and C5aR expression levels increased after stimulation with ADP (Fig. 9F).
To investigate whether C5a mimicked the effects of cTTP serum on HMEC-1, we exposed HMEC-1 to C5a in test medium, in the absence of serum (Fig. 9G, 9H). HMEC-1 incubation with C5a resulted in intense VWF staining compared with cells incubated with test medium alone. C5a also caused loss of THBD staining on the HMEC-1 cell surface (Fig. 9G, 9H).
Finally, to evaluate whether the earlier abnormalities contributed to microvascular thrombosis of cTTP, we preincubated ADP-activated HMEC-1 with cTTP serum (n = 11, in remission) and then perfused with normal heparinized whole blood (added with mepacrine; Fig. 10A) at a shear stress encountered into the microcirculation (60 dynes/cm2). In these experimental conditions blood did not clot and platelet thrombi could be evaluated as green fluorescence signal on HMEC-1. A much wider surface area covered by thrombi was observed upon blood perfusion on HMEC-1 pretreated with cTTP serum, compared with cells exposed to control serum (Fig. 10B, 10C).
The thrombus area was completely normalized by sCR1, FH, or eculizumab (Fig. 10B, 10C). A C5aR antagonist significantly reduced but did not fully normalize the thrombus area (Fig. 10B, 10C), indicating that both terminal complement products, C5a and C5b-9, have a role in the prothrombogenic effect of cTTP serum (Fig. 10). cTTP serum-induced thrombus formation was also inhibited by rADAMTS13 (Fig. 10B, 10C), which prevented VWF accumulation on HMEC-1 (Fig. 9A), and by an anti-VWF Ab (Fig. 10B, 10C).
Together, these data indicate that VWF and complement cooperate to induce microvascular thrombosis in conditions of ADAMTS13 deficiency (Fig. 11).
We demonstrated that severe ADAMTS13 deficiency, which results in the accumulation of ULVWF, is associated with activation of the complement AP until the terminal pathway on the endothelial cell surface.
This is supported by the following findings: 1) sera from all patients with cTTP and ADAMTS13 mutations induced abnormal C3 and C5b-9 deposits on ADP-activated endothelial cells, which were prevented by selective AP inhibitors (similar results were obtained with sera from patients with the acquired form of TTP and undetectable ADAMTS13 activity); 2) abnormal C3 deposits induced by cTTP serum colocalized with VWF; 3) the effect of cTTP serum on C3 and C5b-9 deposition was completely prevented by rADAMTS13 supplementation or by treatment with an anti-VWF Ab; and 4) VWF interacted with C3b and provided a platform for the formation of active C3 and C5 convertases of the AP. Notably, circulating C3 and SC5b-9 levels were altered in only a fraction of patients, indicating that complement activation in cTTP occurs at cell surface rather than in the fluid phase. Altogether, these results match the condition of endothelial-restricted complement activation of patients with aHUS (11, 15, 55) and support the existence of a cross-talk between ADAMTS13/VWF and the complement system.
Previous studies suggested that complement activation may occur in cTTP, as documented by complement-induced hemolysis of sheep erythrocytes in the presence of cTTP plasma (26), and by C3 and C5b-9 deposition on endothelial cells after exposure to cTTP sera (23). Furthermore, exposure of histamine-stimulated glomerular endothelial cells to cTTP platelet-rich plasma caused deposition of C3 on VWF-platelet strings (21). In the earlier reports (23, 26) the diagnosis of cTTP could not be confirmed because no genetic data were provided; in addition, samples were taken during an acute episode (23, 26), so the possibility that complement activation was a secondary phenomenon triggered by extensive platelet activation and ischemic tissue injury could not be ruled out.
Unlike in the earlier studies, Gavriilaki et al. (56) failed to find increased C5b-9 deposits and complement-mediated killing in cells exposed to TTP serum; however, the tests were done on cell lines (human endothelial hybrid and human erythroblast cell lines) made artificially deficient for GPI-anchored complement regulatory proteins (CD55 and CD59), an experimental setting that was far from reproducing the conditions of human vascular endothelium. In this regard, and unlike in other studies (15), the earlier tests could not distinguish aHUS patients with active disease from aHUS patients in remission under terminal complement blockade through eculizumab treatment, because both groups caused the same complement-mediated cell killing.
In this article, we wanted to clarify whether a direct link exists between impaired VWF cleavage caused by ADAMTS13 deficiency and complement activation. For this purpose, we studied cTTP patients with identified ADAMTS13 mutations and without mutations in complement genes. It is relevant that all but two ADAMTS13 mutations have been functionally characterized previously and cause a severe reduction in protein secretion and/or activity or result in protein interruption (28, 32, 39, 41). In addition, to separate the relative contribution of ongoing thrombotic microangiopathy from that of ADAMTS13 deficiency per se to complement activation, we studied cTTP patients both during acute episodes and at remission. Serum from acute cTTP caused C3 and C5b-9 deposits on resting endothelial cells, which could reflect in vivo complement activation triggered by widespread platelet thrombi and coagulation (2, 3, 57). In contrast, cTTP serum taken in remission induced excessive complement deposition only on activated endothelial cells, and the same effect was observed with serum from patients with aTTP and severe ADAMTS13 deficiency studied in remission. These data fit with the observation that TTP onset or relapses may occur in concomitance with a triggering event (infections, drugs, pregnancy) that perturbs microvascular endothelium (17), and suggest that ADAMTS13 deficiency predisposes to AP dysregulation on the cell surface. The observation that supplementing cTTP serum with rADAMTS13 fully prevented C3 and C5b-9 deposits supported the earlier hypothesis. However, mechanistic studies did not reveal any direct regulatory activity of ADAMTS13 on the complement AP. Indeed, rADAMTS13 failed to proteolyze FB and did not alter the assembly of the AP C3 proconvertase and convertase. Some structural homology exists between the A2 domain of VWF that contains the ADAMTS13 cleavage site (43, 44) and the VWA domain of FB that is involved in its binding with C3b (58, 59). However, tyrosine1605 in the ADAMTS13 cleavage site is substituted by leucine in the VWA FB domain, which could explain failure of FB proteolysis by ADAMTS13. One can speculate that ADAMTS13 interaction with FB is too weak to efficiently antagonize the binding of FB with C3b in the C3 proconvertase and convertase complexes formation.
Once established that ADAMTS13 does not directly modulate complement, we investigated whether C3b interacted with VWF. We found a specific interaction between C3b and VWF, and provided the evidence suggesting that C3b binds to the three VWA domains and that the A2 domain represents the main binding site for C3b. We could not observe any interaction of C3b with multimeric rVWF and plasma-derived VWF in ELISA, likely because of closed conformation of VWF multimers in static condition (60). At variance, C3b interacted with plasma-derived VWF under flow in SPR. Altogether, these results are in line with published data showing high-affinity binding between VWF and C3b in SPR, and the colocalization of C3/C3b with ULVWF on histamine-stimulated HUVECs (6, 7). Notably, we could confirm C3b-VWF interaction at the endothelial cell level in static condition, by colocalization of multimeric VWF and C3 deposits on HMEC-1 exposed to cTTP serum. It has been previously documented that secretion of VWF from Weibel-Palade bodies is accompanied by prodomain dissociation, dimeric bouquet unzipping, and VWF expansion to an irregular conformation (61). We speculate that in our assay, VWF multimers released and bound to the endothelial cell surface had a permissive conformation that favored their interaction with C3b and complement activation.
We found no C3 staining on ADP-activated HMEC-1 before exposure to serum, despite the evidence of secreted/attached VWF on the cell surface, indicating that the C3 deposited on HMEC-1 after incubation with cTTP serum mostly originated from serum.
How the interaction between VWF and C3b impacts on the activation of the complement AP is a matter of debate. VWF has been shown to favor C3b inactivation by enhancing FH cofactor activity (45). Another study proposed that normal plasma VWF multimers alone may exert cofactor activity, whereas ULVWF multimers lacked cofactor activity and did not inhibit the generation of C3b from C3 (7). In this study, we failed to find any cofactor activity, either with plasma-derived VWF or with rVWF that includes ULVWF multimers, either in the absence or in the presence of FH. Discrepancies between the present and published results could be because of the largely supraphysiological molar ratios of VWF used in published studies, compared with FI and FH (7, 45). Rather, this study’s finding that the binding of C3b to rVWF monomer, rVWF-A1A2A3, or rVWF type A domains results in the formation of the C3 convertase indicates that VWF may act as an initiator of the complement AP. This possibility is confirmed by data that FH, which dissociates the AP C3 convertase (62), inhibited C3 convertase formation and accelerated its decay on rVWF-A2, and normalized C3 deposits induced on HMEC-1 by cTTP serum. The current dogma states that, although Ca2+ is required for the initiation of the complement classical pathway (35, 63), the AP is Ca2+ independent. A major finding of this study is the identification of a Ca2+-dependent mechanism of AP activation that is mediated by VWF, as supported by data that EGTA reduced the formation of C3bBb on rVWF monomer and on the rVWF type A domains. SPR and ELISA studies showed that the interaction between C3b and VWF was not appreciably affected by the presence of Ca2+. On the other hand, finding that more C3 convertase formed when Ca2+ was present during its assembly on rVWF-A2 would suggest a possible role of calcium in favoring C3bBb complex formation on VWF.
In contrast with our results, Noone et al. (64) recently suggested that VWF has a role in protecting endothelial cells from complement, based on the finding of increased C3c deposition on cultured blood outgrowth endothelial cells from patients with type 3 VWF disease that have severe deficiency of VWF. However, in this study, cells were treated with Abs that artificially blocked all membrane-anchored complement regulators and induced complement fixation via the classical pathway. Furthermore, the in vivo relevance of the earlier data is debatable because, to the best of our knowledge, there is no published evidence indicating complement hyperactivation in patients with VWF disease.
Another relevant finding of our study is the formation of C5b and C5b-9 on rVWF-A2 domain, but not on rVWF-A1 and rVWF-A3, indicating that the AP C3 convertase assembled on VWF-A2 formed the C5 convertase that cleaved C5 to C5a and C5b and initiated the terminal pathway. Altogether, these results provide a mechanistic explanation of the formation of C5b-9 deposits on cTTP serum-exposed HMEC-1.
C5a and C5b-9 cause profound perturbations of the physiologically thromboresistant endothelial phenotype, including upregulation of tissue factor, loss of THBD and proteoglycans, and exocytosis of P-selectin and ULVWF (65–69). This study’s finding that treatment with eculizumab prevented the loss of THBD and the VWF increase on endothelial cells exposed to cTTP serum supports a role of terminal complement components in the loss of endothelial thromboresistant phenotype. The in vivo relevance of these results is supported by published (69) and present data showing higher than normal plasma sTHBD levels in cTTP patients and reduction of plasma sTHBD in a patient after eculizumab treatment. The evidence that treating HMEC-1 with C5a caused VWF exocytosis and THBD shedding points to C5a as a plausible major player in the cTTP serum-induced prothrombogenic endothelial abnormalities. The pathophysiological relevance of these data to microvascular thrombosis is highlighted by evidence that blocking complement with either sCR1, FH, eculizumab, or a C5aR antagonist consistently reduced the prothrombogenic effect of cTTP serum on microvascular endothelium.
In summary, the results presented in this article indicate that VWF multimer accumulation on endothelium, as occurs in cTTP, promotes the activation of the complement AP that proceeds until C5 cleavage with the formation of C5a and the terminal C5b-9 complex. In turn, the terminal complement components cause the loss of the antithrombotic THBD and the increase in VWF multimers secreted/anchored on endothelial cells, thus creating a positive amplification loop that results in microvascular thrombosis (Fig. 11). These findings link hemostasis-thrombosis with the complement AP and open potential therapeutic perspectives of complement-inhibitory drugs in cTTP and in general in thrombotic and inflammatory disorders, such as diabetes, autoimmune disorders, and cardiovascular diseases, associated with endothelium perturbation, VWF release, and complement activation (70–73). The observation that eculizumab given as sole treatment to a boy with cTTP resulted in prompt disease remission is in keeping with such a possibility (28).
We are deeply grateful to Fabio Sangalli for assistance with confocal microscopy, Anna Pezzota for technical assistance in cell culture, Kerstin Mierke for editing the manuscript, and Manuela Passera for secretarial assistance. We are deeply grateful to Dr. Santiago Rodriguez de Cordoba for precious help in designing the SPR experiments and interpreting the results. We also thank Dr. Ariela Benigni and Dr. Marina Morigi for useful discussions, and Dr. Andrea Remuzzi for setting up the perfusion chamber.
This work was supported by the Fondazione ART per la Ricerca sui Trapianti ART ONLUS (Milan, Italy), a grant from the European Union Seventh Framework Programme FP7-EURenOmics (Project 305608), a fellowship from the Fondazione Aiuti per la Ricerca sulle Malattie Rare ARMR ONLUS (to S.B.), grants from Schweizerischer Nationalfond (to G.S.), and grants from Novartis (to G.S.). The funding sources had no role in study design; in collection, analysis, and interpretation of data; or in the writing of the report and the decision to submit the paper for publication.
The online version of this article contains supplemental material.
Abbreviations used in this article:
atypical hemolytic uremic syndrome
receptors for the C5 cleavage product C5a
collagen binding assay
human microvascular endothelial cell line of dermal origin
normal human serum
National Institutes of Health
soluble complement receptor 1
surface plasmon resonance
thrombotic thrombocytopenic purpura
ultra-large m.w. VWF
VWF type A
von Willebrand factor
M.N. has received honoraria from Alexion Pharmaceuticals for giving lectures and participating in advisory boards and has received research grants from Omeros and Chemocentryx. None of these activities has had any influence on the results or interpretation in this article. G.R. has consultancy agreements with AbbVie, Alexion Pharmaceuticals, Bayer Healthcare, Reata Pharmaceuticals, Novartis Pharma, AstraZeneca, Otsuka Pharmaceutical Europe, and Concert Pharmaceuticals. No personal remuneration is accepted; compensations are paid to his institution for research and educational activities. The other authors have no financial conflicts of interest.