A journey from sickle cell anemia to ADAMTS13


Dr H.-M. Tsai, Division of Hematology, Montefiore Medical Center, 111 East 210th Street, Bronx, NY 10467, USA.
Tel.: +1 718 920 4410; fax: +1 718 881 7108; e-mail: htsai@montefiore.org

When I started bench research after finishing the clinical training, I had two great mentors, Ronald L. Nagel, a renowned leader in the study of red blood cells, and Ira I. Sussman, an expert in the physiology of von Willebrand factor (VWF). They created an intellectually challenging environment. Around that time, one study had suggested that VWF mediated the adhesion of sickle erythrocytes to vascular endothelium under flowing conditions [1]. The implications of that study were attractive. Firstly, knowledge of the molecular mechanisms of the adhesion might be applicable clinically to ameliorate the severity of vasoocclusion among patients with sickle cell anemia. Secondly, it involved VWF and sickle cell anemia (Fig. 1).

Figure 1.

A detour from sickle cell anemia to von Willebrand factor, thrombotic thrombocytopenic purpura, and ADAMTS13.

Sickle cell anemia and VWF

My work started in two directions. One was to help create an ex vivo vascular model to investigate sickle cell adhesion. Working with Dhananjaya K. Kaul, rat mesocecum preparations were treated with desmopressin. Desmopressin not only augmented the adhesion of sickle erythrocytes, but also promoted adhesion of normal erythrocytes to the endothelium. Anti-VWF antibodies abolished the adhesion, suggesting that VWF might mediate both normal and sickle erythrocyte–endothelium interaction [2,3].

At the other front, the immediate aim was to isolate endothelial VWF. Victor B. Hatcher generously provided access to his laboratory for umbilical vein endothelial cells, which were noted for long-lasting viability. The analysis revealed that VWF was secreted from these endothelial cells as an ultra-large band instead of a series of multimers [4]. Previously Moake et al. had demonstrated that endothelial VWF contained ultra-large as well as large multimers normally found in the plasma [5]. Nevertheless, the prevailing view was that ultra-large multimers were secreted only upon endothelial stimulation or injury. In contrast, the new data indicated that, irrespective of perturbation, endothelial cells secreted only a high molecular weight VWF polymer, not a series of multimers. Since difference in cell culture conditions might have caused the variability in the composition of endothelial VWF, I turned to Kaul's ex vivo vascular model to tackle the discrepancy. As in cell cultures, the vascular perfusates contained an ultra-large polymer, not a series of multimers [3]. Based on these results, it was concluded that the multimers found in normal plasma were not directly released from endothelial cells. Both ultra-large and normal multimers were generated upon or subsequent to its release from endothelial cells. Although the issue in question did not appear to have immediate clinical relevance, it had irresistible attraction. With the unwavering support of Ronal Nagel I continued on the project.

VWF multimers and proteolysis

How did endothelial VWF become plasma multimers? There were two plausible mechanisms: reduction of the interchain disulfide bonds linking the individual VWF polypeptide; and cleavage of a peptide bond located outside the intrachain disulfide bonds. Because treatment of endothelial VWF with a reducing agent produced a series of multimers that did not exhibit the complex composition of plasma VWF, a disulfide bond reductase alone would not have been sufficient to generate the multimers. Therefore the attention was focused on proteases.

To determine the location of the protease, the intravascular components, including plasma, platelets, erythrocytes, leukocytes and endothelial cells, were investigated. Initially only leukocytes were found to contain VWF cleaving activity, which was mediated by cathepsin G, elastase, and proteinase 3. These proteases converted the endothelial VWF to a series of multimers that were indistinguishable from plasma multimers as analyzed by high-resolution agarose gel electrophoresis [6]. This observation supported the concept that endothelial VWF could be converted to a series of multimers by proteases. However, these leukocyte proteases were unlikely to be responsible for multimer generation because inhibitors in normal plasma immediately suppressed their activities. Furthermore, I did not detect evidence of defective VWF proteolysis among patients with leukopenia. Phillip et al. reported a similar observation [7].

During these studies, the limitation of agarose gel electrophoresis became apparent – even the high-resolution version of the technique could not distinguish the changes mediated by different enzymes. Thus, a new assay was needed. This presented a dilemma: how could one design an assay if the nature of the protease was completely unknown?

Previous studies had observed that plasma VWF contained proteolytic fragments and had identified the main cleavage site [8,9]. Therefore I assumed that the protease responsible for the cleavage was also involved in multimer generation. In preliminary analysis, the conventional SDS PAGE technique for detection of VWF fragments did not appear to be suitable for quantitative analysis. Furthermore, since the antibodies used for detecting VWF were raised against VWF multimers, they reacted inadequately with disulfide-reduced VWF. A new analysis was needed.

The order of disulfide bonding in the formation of VWF polymers had been delineated [10]. Based on that knowledge, I sketched a multimer generation scheme, which predicted that SDS PAGE under non-reducing conditions would best separate the smallest proteolytic products generated by the protease. Through a series of gel electrophoresis, elution, and immunoblotting by using peptide specific antibodies kindly provided by Zaverio M. Ruggeri, an assay was developed to detect the protease activity. This assay demonstrated that the fragments produced by plasmin or leukocyte serine proteases differed from the physiologic fragments [11].

An assay in search of its target

As the new assay was developed, the whereabouts of the putative protease remained elusive. Conditioning of endothelial cells with VWF-deficient whole blood or plasma resulted in the release of the ultra large VWF polymer in the plasma. On some occasions, the high molecular weight VWF polymer appeared to become slightly smaller, indicating that the protease might exist in plasma. Nevertheless the change was insufficient to produce multimers (Fig. 2), suggesting that a cofactor was missing. In the search for that missing factor, I focused on the shear stress created by blood flow.

Figure 2.

It needs a kick. This 1% SDS agarose gel demonstrates that in the culture medium (Medium) endothelial-secreted von Willebrand factor (VWF) consisted of a high molecular weight (HMW) polymer, not a series of multimers (shown at 2 and 24 h). When endothelial cells were conditioned with a plasma sample from a patient with type 2 A von Willebrand's disease (Plasma, 0), some of the secreted VWF became smaller at 2 h. At 24 h, the HMW VWF was completely replaced by smaller forms. Nevertheless, the change stopped short of generating multimers. In order for ADAMTS13 to cleave VWF efficiently, the substrate first has to be unfolded by shear stress. (The approximate migration of IgM in the gel is indicated.)

A simple device consisted of a syringe pump attached to a long capillary tubing was used to apply shear stress on VWF; amazingly it worked. After shear stress, the large multimers of plasma VWF decreased and the decrease was accompanied by an increase in the proteolytic fragments [11]. Similarly, when endothelial-secreted high molecular weight VWF was exposed to shear stress and followed by incubation with plasma, it was converted to a series of multimers [12]. Finally, the secret began to unravel – the protease had hitherto defied detection because its interaction with VWF required shear stress.

Von Willebrand factor and shear stress

How did shear stress promote the cleavage of VWF? To determine if shear stress increased proteolysis by affecting the conformation of VWF, two approaches were contemplated. One was to directly demonstrate that shear stress changed the conformation of VWF. The other was to use chaotropic agents to chemically alter the conformation of VWF.

Before imaging studies of VWF were initiated, the answer to the first question appeared in Blood[13]. In an elegant study using atomic force microscopy, Siedlecki in the laboratory of Roger E. Marchant noted that shear stress unfolded the three-dimensional structure of VWF from a globular to elongated configuration. Interestingly the authors of that study were attracted to shear stress for different reasons. Their finding proved critical in the development of the emerging model.

The efforts to chemically unfold VWF also yielded positive results. As guanidine HCl unfolded the conformation of VWF, it also increased the susceptibility of VWF to cleavage by the plasma protease [12]. Because guanidine HCl was easier to apply compared to shear stress, it replaced shear stress in subsequent assays. Had we continued using the shear stress device, the work would have progressed at a much slower pace.

The ability to detect the VWF cleaving protease in plasma allowed us to revisit some of the mysteries associated with VWF. Previous studies had observed that in a subset of type 2 A von Willebrand's disease (VWD), the large multimers could be preserved, at least partially, by adding EDTA to the blood samples [14,15]. These observations suggested that some loss of large VWF multimers observed in type 2 A patients occured in vitro. Separately Lyons in David Ginsburg's laboratory observed that type 2 A VWF consisted of two subtypes: some mutant VWF failed to form large polymers, while others polymerized normally during biosynthesis [16]. Together these observations suggested that type 2 A VWD contained a subtype in which the mutant VWF polymerized normally but was more susceptible to proteolysis. Using the recombinant VWF proteins provided by David Ginsburg and Hanneke Lankhof, I observed that some type 2 A mutant VWF was susceptible to cleavage even in the absence of shear stress or guanidine HCl. Thus some type 2 A mutations presumably altered the conformation of VWF, making it constitutively susceptible to cleavage [17].

The data so far suggested that VWF formed a very large polymer in endothelial cells. In the circulation, it was cleaved by a plasma protease in a shear-dependent manner. Why did the VWF evolve to become such a large molecule, only to be cleaved in the circulation to smaller forms? What was the purpose of this circuitous process? Size per se would not be a driving force of evolution, unless it was linked to function. The puzzle led to the subject of the next exploration: size and adhesive function of VWF.

Clinical and laboratory observations had indicated that large VWF multimers were more effective than small multimers in hemostasis. However, the molecular basis of this difference according to size was unknown. Separately, what were the characteristics of VWF that enabled it to function under high shear stress conditions? To both questions, conformational flexibility of VWF could provide a unified answer: conformational unfolding by shear stress allowed VWF to perform adhesive function at site of injury under high shear stress conditions, and large multimers were hemostatically more effective because they were more flexible and therefore more responsive to shear stress. According to this scheme, small multimers were hemostatically ineffective because their conformation was not responsive to shear stress. This hypothesis led to a series of experiments comparing the adhesive activity of large and small multimers before and after shear stress. The results supported the premise that conformational responsiveness to shear stress accounts for the peculiarity of VWF [18].

The experience of these shear stress experiments indicated that the activity of VWF, not its size, was the main target of regulation by VWF-cleaving protease. This view raised the next question, what would happen if the protease were absent? If unfolded VWF was more active in supporting platelet aggregation, a deficiency of the protease should result in microvascular platelet aggregation and thrombosis, as observed in thrombotic thrombocytopenic purpura (TTP).

VWF-cleaving protease and TTP

TTP is a serious disease characterized by the development of thrombosis in the arterioles and capillaries [19]. It almost invariably ended with a fatal outcome, resulting from complications of cerebral or myocardial thrombosis. The introduction of plasma therapy in the 1970s improved the prognosis of the disease, yet the etiology and pathogenesis of the disease remained largely unknown [20,21]. A study by Asada et al. demonstrated that the thrombi in TTP consisted of VWF and platelets, suggesting that VWF-platelet aggregation caused thrombosis in TTP [22]. Thus, TTP was chosen as the target of investigation.

Initially, nine samples of TTP patients were tested, in none of whom was VWF cleaving activity detectable. This result was astounding but preliminary.

As TTP was relatively uncommon, it would have been impossible to recruit a large number of cases within a reasonable period of time. Fortunately, Eric Lian, a famous investigator of TTP, kindly provided his samples for the study. Initially the assay results were puzzling: many of his samples did not have deficiency of the protease. It tuned out that Eric had included samples from normal subjects and patients with other conditions. This unsuspected blinding scheme enhanced the confidence level of the assay. Furthermore, since some of the control samples had been stored for as long as or longer than the TTP samples, it alleviated the concern that protease activity was lost during sample storage.

Next, it was essential to identify the cause of protease deficiency. Only then could we be certain that the deficiency data were valid. Soon the studies yielded positive results: IgG molecules isolated from the TTP samples suppressed the protease activity in normal plasma [23].

With the collaboration of Ravindra Sarode, Gian Visentin, Wayne Chandler, Phil Tarr and many others, I investigated samples from patients with various conditions including heparin-induced thrombocytopenia and E. coli O157:H7-associated hemolytic uremic syndrome (HUS) [23,24]. These studies demonstrated that deficiency of VWF cleaving protease was specific for TTP. Joel Moake also kindly provided his samples. Three of his cases had developed TTP following the use of ticlopidine, an antiplatelet agent that was commonly used in patients with cardiovascular disorders at that time. His cases and those referred by Ravindra Sarode demonstrated that inhibitors of the VWF cleaving protease caused ticlopidine-associated TTP [25].

By analyzing the VWF-cleaving activity levels, we now know that the syndrome of thrombocytopenia and microangiopathic hemolysis consists of several disease entities. These diseases differ in pathogenesis but share the common feature of widespread microvascular thrombosis. Presumably microvascular thrombosis creates abnormally high levels of shear stress, leading to the fragmentation of red blood cells. TTP represents the subset of patients that develop thrombosis as a consequence of VWF cleaving protease deficiency. In the HUS and other disorders of microangiopathic hemolysis, thrombosis is caused by different mechanisms. In these disorders, as in TTP, thrombosis increased the level of shear stress in the microcirculation; consequently, the red blood cells are also fragmented. Interestingly, in patients without VWF cleaving protease deficiency, microangiopathic hemolysis is often accompanied by evidence of increased VWF proteolysis. Presumably the shear stress causing red blood cell fragmentation also increases the proteolysis of VWF by the protease.

Molecular cloning of ADAMTS13 and its association with Schulman–Upshaw syndrome

By using the newly developed assay, the nature of the protease was delineated. The data indicated that it was a metalloprotease that was similar in many aspects to members of the matrix metalloprotease family [17]. However, other features indicated that the protease did not belong to the family of matrix metalloprotease. One interesting common feature of VWF cleaving protease and MMP was the susceptibility to inhibition by tetracyclines. Based on this observation, a tetracycline affinity column was prepared for isolation of the protease. Also, the IgG antibodies from TTP patients provided another step of purification. Together these two steps greatly improved the purity of the isolated protease. Unfortunately isolated protein consisted of multiple bands in SDS gels and was met with skepticism when it was submitted for sequence analysis.

Fortunately, the TTP connection provided another route for identifying the protease. By analyzing pedigrees with a congenital deficiency of the protease, it was hoped that positional cloning could help identify the protease. This strategy, if successful, would not only clone the gene of the protease, but would also identify the mutations of the protease; thereby providing a direct proof that deficiency of VWF cleaving protease caused TTP.

Initially no hereditary cases were available for study, and this aim had to be put on hold. During a discussion with Eric Lian about the plan, he recalled two possible cases of congenital TTP among his previous referrals. With the help of Ralph A. Gruppo and his staff in Cincinnati, the patients and their family members generously donated their blood samples for the study. Analysis of these two pedigrees identified three cases of congenital deficiency of the protease. Furthermore the parents and some of their siblings were partially deficient.

A critical factor for the success of positional cloning was sample size; there might not have sufficient number of cases available to reach the necessary statistical power. The experience from these two pedigrees pointed to an alternative direction. If the assay could distinguish the carriers from normal subjects, a large pedigree of hereditary deficiency might suffice.

One day, Visalam Chandrasekaran, director of the blood bank at Long Island Jewish Medical Center, referred a TTP case for protease analysis. By the way, she mentioned that the patient had two cousins whom she had treated for a similar disorder. After confirming that the patient had severe protease deficiency and her mother was partially deficient, I wanted to track down the cousins. Since the family had moved to Florida and the siblings were not regularly followed up with a physician, Eric Lian and I decided to make a house visit. This turned out to be most productive: both cousins were indeed deficient of the protease. Furthermore, the father had four siblings and the mother had eight, most of whom had children. Although they lived in different parts of the world, those living in the USA agreed without hesitation to participate in the study. The endeavor to obtain blood samples from the scattered family members was not straightforward and would not have been accomplished without the assistance of my wife, Tai-Ping Lee, also a physician, who has provided invaluable supports over the years.

By then, it was time to begin seriously considering genome-wide scan and linkage analysis. A pretest statistical analysis by Jurg Ott (Rockefeller University) showed that the sample size was sufficient to justify a project of genomic scan. For genotyping and cloning we needed a collaborator with expertise in the field. Fortunately David Ginsburg, a renowned geneticist in hemostasis, agreed to undertake this project. Gallia G. Levy, an outstanding MD PhD student in his laboratory, accepted the task. Within a few months, Gallia and David had pinpointed the defect to the long arm of chromosome 9. Making use of the newly published human genomic database, they quickly identified mutations in a candidate gene, C9ORF8. This gene encoded a product that consisted of a disintegrin domain and a thrombospondin type-1 motif domain, suggesting that it belonged to a newly recognized zinc metalloprotease family, ADAMTS (a disintegrin and metalloprotease with thrombospondin type 1 motif). Oddly, C9ORF8 did not contain a metalloprotease domain, raising the speculation that it was not an active protease. After further analysis of human cDNA library and genomic sequences, Gallia and David determined that C9ORF8 represented only part of the cDNA. They went on to determine the entire sequence of the full-length ADAMTS13 cDNA and identified mutations in the ADAMTS13 gene in 14 of 15 disease alleles [26]. These results show that ADAMTS13 is the VWF cleaving protease and confirm that ADAMTS13 deficiency causes TTP.

During that period, Jefferson D. Upshaw, Jr. referred a case that he had recently encounted. Previously Upshaw had reported a similar case, in which he demonstrated that plasma infusion was followed by correction of hemolysis. Based on that observation he had proposed that normal plasma contained a factor deficiency of which led to hemolysis and thrombocytopenia in his patient [27]. His theory coincides with our current view of the function of ADAMTS13 but precedes the cloning of ADAMTS13 by more than 20 years.

Most discoveries in medicine have evolved from advances made by numerous investigators, and the discovery of ADAMTS13 was no exception to this rule. During a course spanning 1988 to 2001, it stumbled on difficulties, dead-ends and surprises, which together made the experience anything but boring. In retrospect, I was fortunate to have associated with people of visions. The back-to-back publications by Furlan et al. and our group contributed to the pace of progress [12,23,28,29]. Eventually, within a period of several months, four different groups published isolation or cloning of the protease [26,30–32].


The works cited in the article were supported in part by grants (RO162136 and RO172876) from the National Heart, Lung and Blood Institute of the National Institutes of Health.

The author wish to thank the people, too numerous to list, whose contribution made the works described here possible.