Proteolytic cleavage of von Willebrand factor by ADAMTS-13 prevents uninvited clumping of blood platelets


Miha Furlan, Liebeggweg 7, CH-3006 Bern, Switzerland.
Tel.: + 41 31 3524487; e-mail:

‘Speak the truth, but leave immediately after’ (Slovenian proverb)

Two types of proteins play the role of main players in blood coagulation and hemostasis. They are either ‘sticky’, preferably multimeric structures and possess affinity for several ligands, or they function as proteolytic enzymes and show high specificity with regard to their substrates. These proteins have represented my research interests for the last 30 years. That I came into blood coagulation was a casual incident, as are most other crucial events. My original college education was in chemistry at the University of Ljubljana, Slovenia. In 1957, the last year before graduation, I went to Bonn, Germany, to work in a brick factory for 6 months with the main objective of going home on a motorcycle. Upon my return to the University I realized that someone else had completed the experimental work on pigments of secale cornutum that I had carried out as a graduate student. Accidentally, I found a job as laboratory technician in the Radiobiology Section of the Jozef Stefan Nuclear Institute in Ljubljana and thus switched from organic chemistry to biochemistry.

In the 1950s, we did not know yet how proteins were synthesized. I learned from Joseph S. Futon's textbook of biochemistry [1] that protein synthesis might represent a reversal of proteolysis catalyzed by proteolytic enzymes. The cathepsins, the only known proteases of animal tissues at that time, were thought to be responsible for hydrolysis of all intracellular proteins. While their pH optimum for proteolytic degradation was about 5, it was suggested that at physiological pH their major function was to catalyze condensation of amino acids or peptides.

After graduation I spent a year with Peter Alexander at the Royal Cancer Hospital, now Royal Marsden Hospital, in London. He had proposed that ionizing irradiation leads to release of hydrolytic enzymes, including intracellular proteases, from the lysosomes representing the ‘suicide bags’. I loved the scientific atmosphere, although the laboratories were extremely small and crowded. What I disliked in my project was decapitating live rats with a hatchet.

During my postgraduate work at the Nuclear Institute Jozef Stefan in Ljubljana my research was supported by the Atomic Energy Agency of Yugoslavia. I was interested in biochemistry of intracellular protein breakdown, but to make my grant application acceptable, I also had to propose some experiments with irradiation. My thesis of 1964 had, as a consequence, the following title: ‘Studies on activity of cathepsins in some organs after γ-irradiation’. As an aside, I could publish other work in which ionizing radiation had been omitted.

Gastrointestinal proteases were reasonably well known in the early 1960s. It was recognized that these proteolytic enzymes were specific with regard to the cleavable peptide bond. Thus, pepsin and chymotrypsin required an amino acid with an aromatic side chain, e.g. tyrosine or phenylalanine, whereas trypsin preferred peptide bonds on the carboxy-terminal side of an arginine or lysine. Gastrointestinal proteases did not seem to be very selective with regard to the protein substrate that was available for digestion. It was thought that the intracellular proteases were also highly specific for the cleavable peptide bond and capable of digesting any cellular proteins. The conventional activity assay was then based on cleavage of denatured hemoglobin. In 1965 I fractionated proteolytic activities of extracts from rat spleen and calf thymus and showed that they were specific in their action on different protein substrates. This work, published in Enzymologia[2], remained unnoticed and Enzymologia passed away a few years later. In subsequent work, we isolated from calf thymus nuclei two proteases and demonstrated that they were responsible for digestion of nucleoproteins [3]. I fell in love with the proteases.

In 1967 I met Harold Moroson at the International Congress on Radiation Research in Cortina d'Ampezzo. He invited me to join him at the Sloan Kettering Institute for Cancer Research in New York to study X-ray-induced DNA breaks. My first task in America was earning a driver's license and getting a car – something I had never owned before. In the laboratory, I evaluated the effects of X-irradiation on survival of mice bearing ascites tumor cells. Little by little, I got the impression that cancer cells, chemotherapy drugs and X-rays required too many animal lives and after 2 years I decided to quit cancer research and returned to Ljubljana. Apparently I was not adequately compliant at the Institute Jozef Stefan and had to look for another job. I applied for a position available at the Central Hematology Laboratory, Inselspital/University of Bern, where Eugen Beck, discoverer of the abnormal fibrinogen variant Baltimore, needed a biochemist to investigate structure and function of human fibrinogen. By then, I had heard almost nothing about coagulation, not to mention the Swiss-German language, but it seems that this was not too obvious during the interview and I got the position. Fibrinogen structure and function were my main subjects but plasmin, thrombin and proteases from snake venoms were the tools. Before 1970 we only knew that fibrinogen was composed of three pairs of unequal polypeptide chains and that removal of fibrinopeptides led to polymerization of fibrin monomers. We used plasmin digestion and SDS-PAGE for structural characterization of normal human fibrinogen and of congenital, functionally abnormal fibrinogen variants. In 1984 I had an opportunity to spend 2 weeks at the Max-Planck Institute, Martinsried, Germany, and to learn from Agnes Henschen and Maria Kehl how to separate fibrinogen chains and their fragments by HPLC. At that time, PCR was not yet around and we had to do classical protein biochemistry: cyanogen bromide cleavage, proteolytic digestion, and amino acid sequence analysis of separated peptides. We isolated, functionally characterized and identified mutations in 19 fibrinogen variants from plasmas collected in a number of institutions, nine of them from Milan.

We were fortunate that the Central Laboratory, Blood Transfusion Service of the Swiss Red Cross, Bern, started manufacturing its own factor VIII (FVIII) concentrate and it goes without saying that we eagerly offered our assistance in characterizing their product. In the early 1970s we still believed that FVIII carries two biological activities, one associated with clotting and another inducing platelet aggregation. It was not yet decided whether these two activities resided on one or two different molecules. Therefore, the neutral term ‘FVIII-related protein’ was used for description of the protein moiety of FVIII. We demonstrated that proteolytic degradation was responsible for low recoveries of FVIII and von Willebrand factor (vWF) activity. SDS-PAGE and two-dimensional Laurell electrophoresis indicated that FVIII-related protein was not a homogenous protein but rather a mixture of multimeric molecules. Unfortunately, only the smallest of these multimers would enter 3% polyacrylamide, an extremely soft gel. To increase its rigidity, we reproduced a gel system, containing 2% acrylamide and 0.5% agarose [4], designed originally for analysis of high-molecular-weight RNAs with Mr values of up to 108. Our study revealed that the electrophoretic bands represented disulfide-linked multimers composed of dimers of the basic polypeptide chain [4]. We reported on size-heterogeneity of purified FVIII-related protein in 1% agarose at an ETRO Conference on FVIII in London in 1977. After the meeting I received a letter from Zaverio Ruggeri, then at the Institute of Internal Medicine, University of Milan, asking me whether our SDS–agarose method might also be used for multimer analysis in plasma. Because Coomassie staining was not sensitive enough, he greatly increased the detection limit by using radioiodinated antibodies to FVIII-related protein.

In 1981 we showed that mild disulfide reduction of vWF resulted in progressively decreasing molecular size and ristocetin cofactor activity. Reduced or native small molecular forms of vWF were adsorbed onto colloidal gold granules and thereby generated von Willebrand activity, suggesting that vWF oligomers possess potential binding affinity for platelets which is only manifested in large vWF aggregates [5]. Today, it is generally accepted that the binding affinity of vWF for glycoprotein receptors on platelet membranes and for collagen is primarily associated with the high-molecular-weight multimers. Zimmerman et al.[6] showed that vWF was degraded to smaller multimeric forms in plasma of patients with type 2 von Willebrand disease (vWD). This degradation went on in vitro unless blood was collected in the presence of protease inhibitors [7]. Minor amounts of proteolytically degraded vWF were also noted in normal plasma, indicating that even normal vWF is slowly degraded in the circulation. Proteolytic degradation is responsible for appearance of triplet bands of vWF oligomers in SDS–agarose electrophoresis. Analysis of these satellite bands led to conclusion that low-molecular-weight forms of vWF were either partially [8] or predominantly [9] derived from larger multimers by proteolytic degradation. Several proteases, such as elastase or cathepsin G from leukocytes, a platelet-aggregating cysteine protease, and calpain from platelets, were proposed as candidates for vWF cleavage, but they degraded vWF to fragments that were different from those found in normal human plasma or in patients with vWD type 2A. An excellent study by Dent et al.[10] indicated that the peptide bond between amino acid residues 842Tyr and 843Met was cleaved in the vWF subunit of patients with vWD type 2A. This finding now permitted authenticating the protease responsible for physiologic vWF degradation.

Normal vWF is slowly but continuously degraded in the circulation but it is quite stable in citrated plasma in a test tube. In 1994 Tsai et al.[11] made the important observation that vWF was readily cleaved when normal human plasma was perfused through long capillary tubing at a high shear rate. They proposed that shear stress caused conformational change(s) in vWF, rendering the cleavage site more accessible to protease(s). The model of Tsai was met with some doubt because no-one had ever before heard of a proteolytic process that would be enhanced under increased shear stress; it was also conceivable that the largest, most ‘sticky’ vWF multimers had disappeared from the plasma by simple adsorption onto the inner surface of the capillary tubing. However, he was correct.

Already in 1987 we had tested binding of vWF to latex particles using a cone-and-plate viscometer and had found that the rate of vWF binding to polystyrene beads was enhanced at increased fluid shear stress [12]. The largest vWF multimers were preferentially adsorbed to latex beads. We suggested a tentative model for stretching of vWF under strong tensile forces, compatible with observations of Slayter et al. who had previously shown that in a resting fluid only 13% of vWF molecules are unfolded, while 87% of them are coiled upon themselves in the ‘ball of yarn’ form [13]. Because the hydrodynamic strain generated on a macromolecular filament is approximately proportional to the square of its length, the longest vWF multimers apparently undergo the maximum uncoiling under elevated shear stress and thus expose multiple binding sites for collagen and platelets.

Resistance of vWF to proteolytic attack in resting citrated plasma hindered development of an assay for the protease involved in physiologic vWF cleavage. In the search of adequate experimental conditions I recollected two reports of van Mourik, Bouma et al.[14,15] who 20 years earlier had observed a decrease in the size of FVIII, its procoagulant and platelet-agglutinating activity after dialysis against buffers of low ionic strength. They proposed that the cleavage of FVIII at low ionic strength was non-proteolytic in nature. It must be recalled that in 1976 we were still speaking of FVIII as a non-covalent complex of two functionally dissimilar components, FVIII and FVIII-related protein, that had not yet been well purified and characterized. I repeated dialysis experiments and observed that vWF was indeed digested at low ionic strength by a protease present in normal human plasma. vWF degradation was further enhanced in the presence of 1 mol L−1 urea solution. Gel filtration on large-pore agarose partially resolved vWF and vWF-cleaving protease. The multimeric pattern of degraded vWF substrate was estimated by SDS-agarose electrophoresis and immunoblotting. With an operational activity assay in our hands, we set out to purify and characterize the vWF-cleaving protease (vWF-cp). Only a minor fraction of protease activity was recovered in the cryoprecipitate and the enzyme was absent in human platelets. Proteolytic activity was inhibited by EDTA and the protease cleaved the same bond in vWF that had been shown to be cleaved in vivo[16]. In the same issue of Blood, Tsai described independently another method for partial purification and characterization of vWF-cp [17]. In his assay he used a vWF substrate that had been denatured by guanidinium chloride, and he deduced the extent of proteolytic digestion from the rate of formation of the carboxy-terminal dimeric vWF fragment of 350 kDa. It is apparent that the hypotonic salt concentration, urea, guanidinium chloride or shear stress induce a conformational change in vWF molecule resulting in exposure of the cleavage site. Recombinant vWF lacking the A2 domain was resistant to proteolytic degradation whereas wild-type vWF was readily digested at low ionic strength or in 1 mol L−1 urea solution [18]. Improved purification procedures yielded sufficient protease, free of other plasma proteins, for amino-terminal sequence analysis [19], and permitted identification of the gene coding for vWF-cp. It took plenty of overtime and some tears from Helen Gerritsen, a talented postgraduate student and a gifted violin player, to eliminate the last impurities. It is not easy to hope every morning that the day will bring a break-through unearthing something important that nobody has noted before, and to put up with the disappointment of going home late at night after the experiment has failed. Helen also developed a faster and simpler assay for vWF-cp than immunoblotting, based on size-dependent binding of degraded vWF to collagen [20].

We were curious as to what might be the physiological importance of vWF-cp that generates functionally incompetent low-molecular-weight multimers of vWF. From the storage organelles in the activated endothelial cells, vWF is secreted in the form of ultra-large vWF multimers. The length of these unusually large polymeric filaments, if stretched, would surpass the diameter of an intact platelet (Fig. 1). These ultra-large vWF multimers, released from the endothelial cells, are even more effective than the largest vWF forms, circulating in normal plasma, in inducing platelet agglutination under conditions of high fluid shear [21]. Moake et al.[22] observed unusually large vWF multimers in plasma from patients with a chronic relapsing form of thrombotic thrombocytopenic purpura (TTP) and suggested that the presence of these ultra-large vWF multimers in plasma of TTP patients might be due to an excessive release of vWF from endothelial cells and/or to an impaired degradation of the highly multimeric forms of vWF by a ‘depolymerase’.

Figure 1.

Model for cleavage of ultra-large vWF by ADAMTS-13. Under high fluid shear stress, the unusually large vWF multimers stretch and link a pair of platelets. ADAMTS-13 prevents spontaneous agglutination of platelets by decreasing the length of the multimeric strand of VFW.

I shall never forget our excitement when we detected no vWF-cp activity in two brothers with chronic relapsing TTP who showed the presence in plasma of unusually large vWF multimers [23]. These ultra-large vWF multimers were absent in plasma samples from their asymptomatic parents who had about 50% protease activity. We found no inhibitor of vWF-cp after incubation of normal plasma with an equal volume of patient plasma. The manuscript on constitutional protease deficiency in TTP was published in Blood after rejection by the New England Journal of Medicine. We were surprised to learn that the in vivo recovery of vWF-cp following plasma exchange was about 100% and its biological half-life was 2–4 days [24]. These results suggest that in patients with congenital deficiency of vWF-cp, plasma exchange or even plasma infusion alone may be effective in suppressing acute TTP episodes and in preventing TTP relapses.

The lack of vWF-cp activity in another patient with recurrent severe episodes of TTP was found to be due to an acquired inhibitor of the protease activity [25]. We showed that the inhibitor was an IgG, apparently an autoantibody. Treatment with plasma, corticosteroids and vincristine led to transient disappearance of the autoantibody and appearance of vWF-cp activity. Three months after remission from the initial TTP event, the protease inhibitor returned, vWF-cp activity disappeared, and the platelet count gradually decreased. Severe relapses of TTP occurred 7 and 11 months after the first acute TTP event. Splenectomy performed 1 year after the first TTP event resulted in disappearance of the autoantibody and normalization of protease activity and of the platelet count. No clinical relapse occurred and no protease inhibitor was detected as of the recent visit, more than 6 years following splenectomy.

We subsequently confirmed an association of TTP with constitutional as well as with acquired deficiency of vWF-cp in a retrospective study on the prevalence of severe protease deficiency in patients with familial and acquired thrombotic microangiopathy [26]. Lacking or markedly decreased protease activity was found in all 30 examined patients during an acute TTP event (six familial and 24 non-familial TTP cases). An inhibitor of vWF-cp was established in 20 of 24 patients with non-familial TTP but in none with familial TTP. The protease inhibitor was again an IgG autoantibody that usually disappeared, at least temporarily, in remission. However, we observed normal protease activity in 21 of 23 patients with an initial diagnosis of idiopathic hemolytic uremic syndrome (HUS) (10 familial and 13 non-familial cases) [26] and in 120 healthy individuals. We proposed that the difference in vWF-cp activity between TTP and HUS permits a differential diagnosis of these two similar disorders that are often difficult to distinguish clinically.

The manuscript was initially not appreciated by the reviewers of the New England Journal of Medicine but we obtained help from Tsai and Lian who submitted independently to the same journal another study with similar findings [27]. They found no vWF-cp activity in plasma samples of 37 patients during the acute TTP episode. IgG antibodies with protease inhibitory activity were detected in two-thirds of samples collected during the acute event. Normal activity of vWF-cp was found in 16 plasma samples obtained during the remission of TTP, and in 74 plasma samples from healthy subjects or hospitalized patients with hemolysis, autoimmune disorders, thrombocytopenia or thrombosis from other causes [27].

A couple of years later, Levy et al. used a very elegant approach to identify the gene coding for vWF-cleaving protease, using the genetic linkage analysis in families with one or more members afflicted with congenital TTP [28]. In this study, they identified 12 mutations in the same gene that accounted for 14 of the 15 disease alleles studied. The above studies indicated that vWF-cp is one of 19 members of the ADAMTS family of metalloproteases and it was designated as ADAMTS-13. The ADAMTS-13 gene contains 29 exons. The translation product consists of 1427 amino acids and is composed of a signal peptide and a propeptide, a metalloprotease domain, 8 thrombospondin-1 (TSP-1) repeats, a cysteine-rich domain, an ADAMTS spacer and 2 CUB domains. The predicted molecular mass of non-glycosylated ADAMTS-13 is 154 kDa [28]. Forty of 63 mutations in the ADAMTS-13 gene, so far reported, are missense mutations, in addition to eight nonsense mutations, nine frame shifts, and six splice mutations. All these mutations are spread along the entire length of the ADAMTS-13 gene rather than cluster in one region.

Dong et al.[29] have recently shown that ADAMTS-13 docks to and then cleaves unusually large vWF multimers on the surface of endothelial cells. These observations support the newly proposed ingenious model of in vivo vWF degradation, suggested by Moake [30]. According to this model, the ultra-large vWF multimers, secreted from stimulated endothelial cells, bind to the endothelial cell surface under conditions of fluid shear stress and induce adhesion of platelets from the flowing blood. The string-like structures of adhering platelets align in the direction of fluid flow and resist detachment unless the size of vWF multimers is decreased due to proteolytic cleavage by ADAMTS-13.

In the first 4 years after discovery of vWF-cp deficiency we examined plasma samples from 239 patients with suspected TTP/HUS sent to our laboratory from 85 clinics: in 23 patients from 16 families we confirmed severe constitutional deficiency of vWF-cp whereas 72 patients had an acquired protease deficiency caused by inhibiting autoantibodies. In these studies I enjoyed a productive collaboration with a laboratory technician from the Philippines, Rudy Robles. Apart from being an expert in pouring reproducible agarose gels, he was a bowling champion and a great karaoke singer, especially when impersonating Frank Sinatra's ‘I did it my way’.

I am highly indebted to the Swiss National Science Foundation and to the Central Laboratory of the Swiss Red Cross for funding our research on vWF and fibrinogen during three decades. The Swiss Red Cross was also our virtually unlimited source of outdated frozen plasma. I was surprised and overjoyed in 1997 when I had a telephone call from Adelaide L. Lewis, the Executive Director of the Malcolm Hewitt Wiener Foundation, who offered to support our project. I am very obliged to the Foundation for financial assistance during the last 3 years of our active research.

In Switzerland, there is a strict rule to retire from a public office at the age of 65. I left the laboratory bench before answering why a short exposure of vWF to 1 mmol L−1 mercuric chloride, an activator of matrix metalloproteases, or to pH 11, leads to complete smearing of the vWF multimeric pattern in SDS–agarose electrophoresis (unpublished observations). The extreme longevity of vWF-cleaving protease in vitro[19] (half-life > 1 week) and in vivo[24] (half-life about 3 days) remains mysterious since most proteases are quite short-lived in the plasma milieu. I look forward to reading future reports on why some children with congenital ADAMTS-13 deficiency experience an acute episode immediately after birth whereas other are asymptomatic for years, but after an initial TTP event their acute episodes may relapse every 3 weeks [31] if not treated with FFP. Levels of 5% ADAMTS-13 or even less, following plasma infusion in patients with congenital deficiency normalize the platelet count within a couple of days [31]. But what is it that raises the platelet count even after the protease has virtually disappeared from the circulating plasma? Is complete deficiency of ADAMTS-13 indeed incompatible with life? What is the function of the spleen in the development of autoantibodies directed against ADAMTS-13? I believe that answering these questions will help to improve the prophylaxis and therapy of patients with TTP.

During my 30 years at the University Hospital/Inselspital, I greatly appreciated and enjoyed working as a biochemist with two hematologists, Eugen Beck and Bernhard Lämmle. They both saw patients and had an outstanding interest in basic and clinical research. They shared with me the joy of coagulation in the test tube. I suspect they envied me for my exploration of validity of new ideas at the laboratory bench. I envied them for their excellence in medicine. Furthermore, I admired Eugen who was an exceptionally gifted piano player, and Bernhard who was the boldest and fastest skier on the snow-covered slopes of Bernese Alps.