Watzka et al.  have recently reported the phenotype of patients with therapeutic resistance to 3-substituted-4-hydroxycoumarin oral anticoagulants (OACs) and with point mutations in VKORC1, which encodes vitamin K epoxide reductase subunit 1 (VKORC1). This study extends the repertoire of naturally occurring VKORC1 variants and provides valuable confirmation that VKORC1 is a molecular target of OACs. Watzka et al. also present a new topology model of VKORC1 and demonstrate that most mutations associated with OAC resistance predict substitutions within, or adjacent to, the VKORC1 endoplasmic reticulum (ER) loop (residues 30–79). This model also identifies the VKORC1 active site (CXXY; residues 132–135) that mediates de-epoxidation of vitamin K 2,3-epoxide (KO) to vitamin K (K) and the reduction of K to K hydroquinone (KH2) and a putative 4-hydroxycoumarin binding motif (TYA; residues 138–140).
Several groups have proposed that OACs inhibit VKORC1 reductase function by binding irreversibly at the VKORC1 active site to prevent the formation of transition state complexes that are formed normally during reduction of KO to K . It is a plausible extension of this model to suggest that in OAC resistance, structural disruption of VKORC1 diminishes the affinity of OAC binding and thereby reduces this inhibition. Watzka et al. argue further that structural disruptions of VKORC1 sufficient to reduce OAC binding are also likely to reduce the binding of the KO transition state complexes and thereby, impair VKORC1 function. This is supported by previous data from ex vivo expression studies that indicate that most OAC-resistant VKORC1 variants showed a markedly impaired KO de-epoxidase activity compared with wild-type controls .
However, some OAC-resistant ER-loop variants showed similar, or even increased, KO de-epoxidase function compared with wild-type controls . Mutagenesis of the conserved Cys residues at 51 and 43 in VKORC1 did not diminish dithiothreitol-driven KO de-epoxidase and K reductase activities , although there was marked loss of both activities compared with wild-type VKORC1 when reduction was supported by thioredoxin oxidoreductase . These data highlight continued uncertainty about the mechanism of OAC resistance and how structural variation in VKORC1 contributes to substrate and OAC interactions.
We contribute to this debate by presenting phenotypic data from a subject with a previously unreported Ala34Pro substitution and two subjects with the Val66Met substitution, which are both conserved residues within the VKORC1 ER loop. Both substitutions were associated with OAC resistance yet did not diminish VKDCF carboxylation in vivo in the absence of OAC therapy.
Subject 1 was a 74-year-old man with atrial fibrillation who was loaded with warfarin using a standardized regime. There was no increase in INR by day 28 of therapy and warfarin dose escalation to 27 mg day−1 was required to achieve a therapeutic INR (Fig. 1A). The serum warfarin concentration was 5.9, 5.8 and 4.2 mg L−1 on three occasions spanning 7 months of stable OAC treatment (Fig. 1B). These concentrations are markedly higher than our reference range of 0.7–2.3 mg L−1 determined from 137 individuals stably anticoagulated with INR 2–4. The reduced plasma VKDCF activities and increased serum KO and under-carboxylated prothrombin (PIVKA-II) concentrations were consistent with OAC therapy. The serum K concentration was normal (Fig. 1B). These findings indicate pharmacodynamic OAC resistance not caused by increased K intake. Consistent with this phenotype, subject 1 harboured a heterozygous VKORC1 c.100G>C transversion that predicted an Ala34Pro substitution in the VKORC1 ER loop. Ala34Pro has not been associated previously with OAC resistance although substitutions affecting the adjacent Arg33 and Arg35 have been identified in OAC-resistant rodents. Two weeks after stopping warfarin following cardioversion, the serum warfarin concentration fell to a subtherapeutic level, the VKDCF activities were within our normal reference ranges and there was no detectible PIVKA-II or KO (Fig. 1B).
Subjects 2 and 3 were a 35-year-old women and an 18-year-old man who both harboured a heterozygous VKORC1 c.196G>A transition predictive of a Val66Met substitution in the VKORC1 ER loop. These subjects were relatives of a proband with the Val66Met substitution who required 32 mg day−1 of warfarin for therapeutic anticoagulation, resulting in a serum warfarin concentration of 6.3 mg L−1. Both subjects were asymptomatic and were not receiving OACs. The VKDCF activities were within our reference ranges and there was no detectible PIVKA-II or KO (Fig. 1B).
These observations extend those of Watzka et al. by providing an insight into the function of OAC-resistant VKORC1 variants in the absence of OAC therapy. Our PIVKA-II assay allows detection of under-carboxylated prothrombin at a lower threshold of 200 ng mL−1. This corresponds to 0.2% of total prothrombin and provides a highly sensitive marker of VKDCF carboxylation. The absence of circulating PIVKA-II in the study subjects therefore indicates that the Ala34Pro and Val66Met substitutions in the VKORC1 ER loop were insufficient to impair VKCDF carboxylation in vivo, suggesting no major reduction in hepatic KH2 availability. We also showed that there was no detectable accumulation of KO, which is a direct marker of VKORC1 KO de-epoxidase function. These findings must be interpreted carefully in the light of the model proposed by Watzka et al., which predicts marked loss of VKORC1 KO de-epoxidase and K reductase function for most OAC-resistant variants.
One explanation for our results is that the VKORC1 ER loop substitutions markedly diminish binding of OAC but that the disruption to substrate interactions at the VKORC1 active site predicted by Watzka et al. is less significant. While it is generally accepted that the 4-hydroxycoumarin ring in OACs interacts directly with the VKORC1 active site, there is evidence that the substituent at the 3-position also contributes significantly to the affinity of OAC binding through an accessory site. This is illustrated by the ability of VKORC1 to harbour the bulky and lipophilic 3-substituents of the highly potent superwarfarin rodenticides . An accessory OAC binding site in VKORC1 that can influence the reduction of KO at the catalytic site was also previously proposed as an explanation for the non-competitive nature of warfarin and KO binding . The distribution of VKORC1 OAC resistance mutations observed in the Watzka et al. study identifies the ER loop as a possible accessory binding site that would be disrupted by amino-acid substitutions in the variant VKORC1 proteins and could therefore reduce the affinity of OAC binding. However, without OAC therapy, any effect of ER loop mutations on substrate binding at the VKORC1 active site would be comparatively small and may still enable synthesis of sufficient KH2 to support VKDCF carboxylation.
It is also possible for our observations in OAC-resistant patients to fit with a more marked loss of VKORC1 function as suggested by Watzka et al. However, in this circumstance, our demonstration of absent circulating PIKVA-II is hard to explain without postulating an alternative means of generating KH2 that is independent of VKORC1. The presence of several reductase enzymes in mammalian cells has been suspected for many years and there is now indirect evidence of an OAC-sensitive K reductase pathway with very low specificity for KO and which generates KH2 at low efficiency compared with VKORC1 . At normal dietary K levels, a pathway with these characteristics may supply sufficient KH2 to support full VKDCF carboxylation even in the presence of VKORC1 ER loop substitutions that markedly disrupt the VKORC1 active site. As the postulated alternative pathway only offers partial rescue of KH2 synthesis, and is OAC sensitive, therapeutic anticoagulation would still be possible in these individuals, albeit at high OAC doses. Against this model, we were unable to demonstrate elevated KO levels in our study subjects. However, de-epoxidation of KO to K by VKORC1 is considerably faster than reduction of K to KH2 in the absence of OACs . Therefore, a subtle defect in VKORC1 KO de-epoxidase activity conferred by the ER loop substitutions may not have caused sufficient accumulation of KO to be identified by our assay, which has a lower detection threshold of 0.12 μg L−1.
The data presented in this study provide an insight into KO metabolic pathways in vivo in subjects with dysfunctional VKORC1 variants. Controversy remains over the structural determinants of VKORC1 that mediate OAC drug and substrate binding, which will require further structural studies to resolve.