A cell-based high-throughput screen identifies drugs that cause bleeding disorders by off-targeting the vitamin K cycle.

Drug-induced bleeding disorders contribute to substantial morbidity and mortality. Antithrombotic agents that cause unintended bleeding of obvious cause are relatively easy to control. However, the mechanisms of most drug-induced bleeding disorders are poorly understood, which makes intervention more difficult. As most bleeding disorders are associated with the dysfunction of coagulation factors, we adapted our recently established cell-based assay to identify drugs that affect the biosynthesis of active vitamin K-dependent (VKD) coagulation factors with possible adverse off-target results. The National Institutes of Health (NIH) Clinical Collection (NCC) library containing 727 drugs was screened, and 9 drugs were identified, including the most commonly prescribed anticoagulant warfarin. Bleeding complications associated with most of these drugs have been clinically reported, but the pathogenic mechanisms remain unclear. Further characterization of the 9 top-hit drugs on the inhibition of VKD carboxylation suggests that warfarin, lansoprazole, and nitazoxanide mainly target vitamin K epoxide reductase (VKOR), whereas idebenone, clofazimine, and AM404 mainly target vitamin K reductase (VKR) in vitamin K redox cycling. The other 3 drugs mainly affect vitamin K availability within the cells. The molecular mechanisms underlying the inactivation of VKOR and VKR by these drugs are clarified. Results from both cell-based and animal model studies suggest that the anticoagulation effect of drugs that target VKOR, but not VKR, can be rescued by the administration of vitamin K. These findings provide insights into the prevention and management of drug-induced bleeding disorders. The established cell-based, high-throughput screening approach provides a powerful tool for identifying new vitamin K antagonists that function as anticoagulants.

Evaluation of oral anticoagulants with vitamin K epoxide reductase in its native milieu.

Warfarin, acenocoumarol, phenprocoumon, and fluindione are commonly prescribed oral anticoagulants for the prevention and treatment of thromboembolic disorders. These anticoagulants function by impairing the biosynthesis of active vitamin K-dependent coagulation factors through the inhibition of vitamin K epoxide reductase (VKOR). Genetic variations in VKOR have been closely associated with the resistant phenotype of oral anticoagulation therapy. However, the relative efficacy of these anticoagulants, their mechanisms of action, and their resistance variations among naturally occurring VKOR mutations remain elusive. Here, we explored these questions using our recently established cell-based VKOR activity assay with the endogenous VKOR function ablated. Our results show that the efficacy of these anticoagulants on VKOR inactivation, from most to least, is: acenocoumarol > phenprocoumon > warfarin > fluindione. This is consistent with their effective clinical dosages for stable anticoagulation control. Cell-based functional studies of how each of the 27 naturally occurring VKOR mutations responds to these 4 oral anticoagulants indicate that phenprocoumon has the largest resistance variation (up to 199-fold), whereas the resistance of acenocoumarol varies the least (<14-fold). Cell-based kinetics studies show that fluindione appears to be a competitive inhibitor of VKOR, whereas warfarin is likely to be a mixed-type inhibitor of VKOR. The anticoagulation effect of these oral anticoagulants can be reversed by the administration of a high dose of vitamin K, apparently due to the existence of a different enzyme that can directly reduce vitamin K. These findings provide new insights into the selection of oral anticoagulants, their effective dosage management, and their mechanisms of anticoagulation.

Employing a gain-of-function factor IX variant R338L to advance the efficacy and safety of hemophilia B human gene therapy: preclinical evaluation supporting an ongoing adeno-associated virus clinical trial.

Vector capsid dose-dependent inflammation of transduced liver has limited the ability of adeno-associated virus (AAV) factor IX (FIX) gene therapy vectors to reliably convert severe to mild hemophilia B in human clinical trials. These trials also identified the need to understand AAV neutralizing antibodies and empty AAV capsids regarding their impact on clinical success. To address these safety concerns, we have used a scalable manufacturing process to produce GMP-grade AAV8 expressing the FIXR338L gain-of-function variant with minimal (<10%) empty capsid and have performed comprehensive dose-response, biodistribution, and safety evaluations in clinically relevant hemophilia models. The scAAV8.FIXR338L vector produced greater than 6-fold increased FIX specific activity compared with wild-type FIX and demonstrated linear dose responses from doses that produced 2-500% FIX activity, associated with dose-dependent hemostasis in a tail transection bleeding challenge. More importantly, using a bleeding model that closely mimics the clinical morbidity of hemophilic arthropathy, mice that received the scAAV8.FIXR338L vector developed minimal histopathological findings of synovitis after hemarthrosis, when compared with mice that received identical doses of wild-type FIX vector. Hemostatically normal mice (n=20) and hemophilic mice (n=88) developed no FIX antibodies after peripheral intravenous vector delivery. No CD8(+) T cell liver infiltrates were observed, despite the marked tropism of scAAV8.FIXR338L for the liver in a comprehensive biodistribution evaluation (n=60 animals). With respect to the role of empty capsids, we demonstrated that in vivo FIXR338L expression was not influenced by the presence of empty AAV particles, either in the presence or absence of various titers of AAV8-neutralizing antibodies. Necropsy of FIX(-/-) mice 8-10 months after vector delivery revealed no microvascular or macrovascular thrombosis in mice expressing FIXR338L (plasma FIX activity, 100-500%). These preclinical studies demonstrate a safety:efficacy profile supporting an ongoing phase 1/2 human clinical trial of the scAAV8.FIXR338L vector (designated BAX335).

A conserved region of human vitamin K-dependent carboxylase between residues 393 and 404 is important for its interaction with the glutamate substrate.

Certain individuals with combined deficiencies of vitamin K-dependent proteins have a mutation, L394R, in their gamma-glutamyl carboxylase causing impaired glutamate binding. The sequence surrounding Leu394 is similar in all known carboxylases, suggesting that the region is functionally important. To test this hypothesis we made the following mutant enzymes: W390A, Y395A, S398A, W399A, and H404A. We purified the enzymes and corrected the activity measurements for active enzyme concentration. Carboxylases W390A, S398A, and H404A had activities similar to that of wild type; however, Y395A and W399A had lower activities than did wild type. In the following descriptions we include our previously reported results for L394R. Kinetic studies with the substrate FLEEL, revealed Km values of 0.5 (wild type), 6.5 (L394R), 15 (Y395A), and 24 (W399A) mm. The kcat values relative to wild type were 51% (L394R), 1% (Y395A), and 2% (W399A). The kcat/Km values were 24-fold (L394R) and >2000-fold lower for Y395A and W399A than for wild-type carboxylase. Inhibition of FLEEL carboxylation by the competitive inhibitor, Boc-mEEV, gave Ki values of 0.013 (wild type), 1.4 (L394R), 2.1 (Y395A), and >5 (W399A) mm. The Y395A propeptide affinity was similar to that of wild type, but those of L394R and W399A were 16-22-fold less than that of wild type. Results of kinetic studies with a propeptide-containing substrate were consistent with results of propeptide binding and FLEEL kinetics. Although propeptide and vitamin K binding in some mutants were affected, our data provide compelling evidence that glutamate recognition is the primary function of the conserved region around Leu394.

The putative vitamin K-dependent gamma-glutamyl carboxylase internal propeptide appears to be the propeptide binding site.

The vitamin K-dependent gamma-glutamyl carboxylase binds an 18-amino acid sequence usually attached as a propeptide to its substrates. Price and Williamson (Protein Sci. (1993) 2, 1997-1998) noticed that residues 495-513 of the carboxylase shares similarity with the propeptide. They suggested that this internal propeptide could bind intramolecularly to the propeptide binding site of carboxylase, thereby preventing carboxylation of substrates lacking a propeptide recognition sequence. To test Price's hypothesis, we created nine mutant enzyme species that have single or double mutations within this putative internal propeptide. The apparent K(d) values of these mutant enzymes for human factor IX propeptide varied from 0.5- to 287-fold when compared with that of wild type enzyme. These results are consistent with the internal propeptide hypothesis but could also be explained by these residues participating in propeptide binding site per se. To distinguish between the two alternative hypotheses, we measured the dissociation rates of propeptides from each of the mutant enzymes. Changes in an internal propeptide should not affect the dissociation rates, but changes to a propeptide binding site may affect the dissociation rate. We found that dissociation rates varied in a manner consistent with the apparent K(d) values measured above. Furthermore, kinetic studies using propeptide-containing substrates demonstrated a correlation between the affinity for propeptide and V(max). Taken together, our results indicated that these mutations affected the propeptide binding site rather than a competitive inhibitory internal propeptide sequence. These results agree with our previous observations, indicating that residues in this region are involved in propeptide binding.