Warfarin Sensitivity Related to CYP2C9, CYP3A5, ABCB1 (MDR1) and Other Factors

Warfarin sensitivity related to CYP2C9, CYP3A5, ABCB1 (MDR1) and other factors

M Wadelius1,4 ABSTRACT 1 The required dose of the oral anticoagulant warfarin varies greatly, and KSo¨rlin overdosing often leads to bleeding. Warfarin is metabolised by cytochrome 2 O Wallerman P450 enzymes CYP2C9, CYP1A2 and CYP3A. The target cell level of warfarin J Karlsson1 may be dependent on the efflux pump P-glycoprotein, encoded by the Q-Y Yue3 adenosine triphosphate-binding cassette gene ABCB1 (multidrug resistance PKE Magnusson2 gene 1). Genetic variability in CYP2C9, CYP3A5 and ABCB1 was analysed in 2 201 stable warfarin-treated patients using solid-phase minisequencing, C Wadelius pyrosequencing and SNaPshot. CYP2C9 variants, age, weight, concurrent H Melhus1 drug treatment and indication for treatment significantly influenced warfarin dosing in these patients, explaining 29% of the variation in dose. CYP3A5 did 1Department of Medical Sciences, Clinical not affect warfarin dosing. An ABCB1 haplotype containing the exon 26 Pharmacology, University Hospital, Uppsala, Sweden; 2Department of Genetics and Pathology, 3435T variant was over-represented among low-dose patients. Thirty-six Medical Genetics, Rudbeck Laboratory, Uppsala, patients with serious bleeding complications had higher prothrombin time Sweden; 3Medical Products Agency, Uppsala, international normalised ratios than 189 warfarin-treated patients without Sweden serious bleeding, but there were no significant differences in CYP2C9, CYP3A5 or ABCB1 genotypes and allelic variants. Correspondence: The Pharmacogenomics Journal (2004) 4, 40–48. doi:10.1038/sj.tpj.6500220 M Wadelius, Department of Medical Published online 16 December 2003 Sciences, Clinical Pharmacology, University Hospital, SE-751 85 Uppsala, Sweden. Keywords: warfarin; bleeding; CYP2C9; CYP3A5; ABCB1; MDR1 Tel: þ 46 18 611 49 45 Fax: þ 46 18 51 92 37 E-mail: [email protected]

4Temporary address: The Wellcome Trust Sanger Institute, Wellcome Trust Genome INTRODUCTION Campus, Hinxton, Cambridge CB10 1SA, UK. Nearly 1% of the Swedish population is treated with the oral anticoagulant Tel: þ 44 1223 49 47 09. warfarin. The indications for treatment include atrial fibrillation, heart valve prosthesis, recurrent stroke, deep vein thrombosis and pulmonary embolism.1 Interindividual variation in the response to warfarin and a narrow therapeutic range are two of the factors that make warfarin therapy difficult to handle. In all, bleeding complication during treatment with warfarin is the most common adverse drug reaction with a fatal or disabling outcome.2,3 Warfarin is a coumarin derivative that interferes with the recycling of vitamin K in the liver.2 Vitamin K is involved in the carboxylation of the precursor proteins for the coagulation factors II, VII, IX and X. In the presence of warfarin, the activity of these components in the blood is lowered by around 10–40% and coagulation is inhibited.3,4 The effect is delayed until dysfunctional coagulation factors have been synthesised, which takes 4–5 days.4 Patients treated with warfarin are monitored by prothrombin time (PT) expressed as an international normalised ratio (INR). PT INR measures the sum of activity of the vitamin K- Received: 09 July 2003 dependent coagulation factors II, VII and X. For most indications, the ideal PT Revised: 03 October 2003 4–6 Accepted: 14 October 2003 INR value lies between 2.0 and 3.0. One exception is heart valve prosthesis, Published online 16 December 2003 which requires a higher PT INR. Warfarin related to CYP2C9, CYP3A5 and ABCB1 (MDR1) M Wadelius et al 41

The dose requirement of warfarin varies more than 10-fold hepatocytes, which indicates that warfarin is a P-GP between patients.7 In general, women need lower doses than substrate in the liver.27 In theory, low P-GP activity in men.4 A high dietary intake of vitamin K leads to a higher hepatocytes may lead to a high level of warfarin in the target dose requirement, while old age is associated with lower cell, causing warfarin sensitivity in some people. doses.8–11 Diseases that increase the sensitivity to warfarin The ABCB1 gene, also called the multidrug resistance gene include liver dysfunction impairing the synthesis of coagu- 1 (MDR1), encodes P-GP. There are many known poly- lation proteins, hypermetabolic states that increase the morphisms in ABCB1: for example, À12T4C in exon 1 in clearance of coagulation factors and congestive heart failure the 50 untranslated region, À1G4A at the initiation of that alters the distribution of warfarin.3,6,11 Warfarin is translation in exon 2, 1199G4A in exon 11 leading to highly bound to serum albumin, which in combination amino-acid change S400N, 1236C4T in exon 12 at a wobble with its extensive cytochrome P450 (CYP) metabolism and position G412G, 2677G4TorG4A in exon 21 leading to narrow therapeutic range makes warfarin susceptible to drug A893S or T, and in exon 26 two polymorphisms at wobble interactions.12 positions, 3396C4T A1132A and 3435C4T I1145I.28,30 An asymmetric carbon in the warfarin molecule gives rise Polymorphisms of ABCB1 are suggested to be important to the enantiomeric forms S- and R-warfarin.13 Both for variability in drug bioavailability, but the pharmacolo- enantiomers are eliminated extensively via hepatic metabo- gical implication of these polymorphisms has not been fully lism with a low extraction ratio. The S-form is a 3–5 times established.30 more potent vitamin K antagonist than the R-form.6,7 Haemorrhage is the most common adverse reaction to CYP2C9 is the main enzyme responsible for metabolism of coumarin anticoagulants, and a great under-reporting of the active S-warfarin,7,13 which renders the genetic variation these events is believed to exist.3,31 Swedish studies have of this metabolic pathway extremely important.14 In shown that 4.5% of warfarin-treated patients experience addition to the CYP2C9*1 wild-type allele, at least 11 major bleeding and 0.5% suffer fatal complications.31 The variant alleles are known to exist (http://www.imm.ki.se/ risk of bleeding is 10 times higher during the first month cypalleles/ in September 2003). The most common variants compared to after the first year.5 There is a clear relationship CYP2C9*2 (430C4T, R144C) and CYP2C9*3 (1075A4C, between haemorrhage and the intensity of treatment, with I359L) generate enzymes with impaired hydroxylation of PT INR elevation being a strong predictor.3,5,10,32 Age, S-warfarin due to amino-acid changes, and several studies concurrent medication, specific comorbid conditions, espe- have shown that these variants have an effect on warfarin cially cerebrovascular, kidney, heart and liver disease as well dose requirement.7,11,15–24 as prosthetic heart valves, are independent risk factors.5,32 R-warfarin is mainly metabolised by the CYP enzymes Several studies suggest that patients with CYP2C9 variant CYP1A2 and CYP3A, with CYP3A likely to be the most alleles have a higher incidence of bleeding complications important.13 CYP3A activity is derived from CYP3A4 and than carriers of the wild-type genotype.15,16,20,22 The risk of CYP3A5 that have similar substrate specificities. CYP3A5 is haemorrhage must always be weighed against the preven- considered to be 17–50% of the total CYP3A content in the tion of thromboembolism, and in most patients the livers of people that express it.25,26 There is considerable preventive effect outweighs the risk of bleeding.4,32 intra- and interindividual variation in CYP3A4 enzyme The aim of the study was to identify factors that influence activity, and even though numerous polymorphisms have the effect of warfarin and the required dose. Warfarin dose been found, the molecular basis of this is not yet under- requirement was compared to genetic variation of CYP2C9, stood. On the other hand, CYP3A5 exists in distinct fast and CYP3A5 and ABCB1, age, gender, weight, smoking, indica- slow variants,25 and is therefore worth studying even if it is tion for treatment, and concurrent medication. In the case– less abundant than CYP3A4. CYP3A5*1 encodes an active control part, we studied these factors in relation to warfarin enzyme, whereas CYP3A5*3 (6986A4G) and CYP3A5*6 bleeding. (14690G4A) are inactive due to splicing defects. The CYP3A5*3 allele is the most common cause of low CYP3A5 functionality in Caucasians, while CYP3A5*6 is rarely found. RESULTS There is preliminary evidence that P-glycoprotein (P-GP) Dose Requirement contributes to warfarin disposition.27 P-GP belongs to a class The first aim was to identify factors that influence the of adenosine triphosphate-binding cassette (ABC) transport required dose of warfarin. The patients were recruited in proteins.28 It serves as a potent efflux pump for a wide 2000, when they were 28–88 years old (Table 1). They had variety of lipophilic compounds that are frequently CYP3A stable PT values, and had been treated with warfarin for substrates.29 P-GP is expressed in tissues such as the between 2.4 months and 26 years (median 2 years). intestine, liver, kidney, blood–brain barrier and placenta.28 Individual warfarin dose requirement varied between 4.5 Warfarin is well absorbed from the intestine with over 90% and 77.25 mg per week. The most frequent indications for bioavailability, making it unlikely that intestinal P-GP anticoagulation were atrial fibrillation, heart valve prosthe- affects the absorption of warfarin to a great extent.1 After sis and deep vein thrombosis/pulmonary embolism. The absorption, the target of warfarin action is the liver, where it most common concomitant diseases were hypertension, inhibits vitamin K epoxide reductase.2 Warfarin has been heart failure, angina pectoris and type II diabetes mellitus. shown to be a moderate inhibitor of P-GP activity in human The patients had a total of 107 concurrent medications

www.nature.com/tpj Warfarin related to CYP2C9, CYP3A5 and ABCB1 (MDR1) M Wadelius et al 42

Table 1 Characteristics of patients in the dose requirement study (n ¼ 201)

Subjects All (%) CYP2C9*1/*1 (%) Others (%)

Indication Atrial fibrillation 113 (56.2) 73 (54.5) 40 (59.7) Heart valve prosthesis 49 (24.4) 35 (26.1) 14 (20.9) Deep vein thrombosis/pulmonary embolus 9 (4.5) 7 (5.2) 2 (3.0) Cardiomyopathy 8 (4.0) 5 (3.7) 3 (4.5) Transischaemic attack 5 (2.5) 5 (3.7) —

Other diseases Hypertension 78 (38.8) 51 (38.1) 27 (40.3) Heart failure 51 (25.4) 29 (21.6) 22 (32.8) Angina pectoris 35 (17.4) 25 (18.7) 10 (14.9) Type II diabetes mellitus 18 (9.0) 13 (9.7) 5 (7.5)

Interacting medication Simvastatin s 25 (12.4) 17 (12.7) 8 (11.9) Aspirin s 21 (10.4) 13 (9.7) 8 (11.9) Paracetamol s 18 (9.0) 14 (10.4) 4 (6.0) Amiodarone s 9 (4.5) 6 (4.5) 3 (4.5) Disopyramide s 7 (3.5) 5 (3.7) 2 (3.0) Dextropropoxyphene s 7 (3.5) 4 (3.0) 3 (4.5) Propafenone s 2 (1.0) 2 (1.5) — Carbamazepine r 3 (1.5) 2 (1.5) 1 (1.5) Nonsteroidal anti-inflammatory drug s 2 (1.0) 1 (0.7) 1 (1.5) Phenytoin r 1 (0.5) 1 (0.5) —

Gender Men 135 (67.2) 89 (66.4) 46 (68.7) Women 66 (32.8) 45 (33.6) 21 (31.3)

Age Mean years (range) 66.9 (28–88) 66.5 (28–88) 67.6 (46–86)

Comparisons between patients with the CYP2C9*1/*1 genotype (n ¼ 134) and other CYP2C9 genotypes (n ¼ 67). A patient can have more than one indication, concurrent disease and interacting medication, and only the most common ones are shown. Drugs potentiating the effect of warfarin are indicated by s and drugs decreasing its effect by r

known to influence warfarin, the most common being Analysis of covariance was performed to see which of the simvastatin, aspirin, paracetamol, amiodarone, disopyra- factors, CYP2C9, CYP3A5, ABCB1, age, gender, bodyweight mide and dextropropoxyphene. Four patients were treated (BW), smoking habits, mean PT INR value, concurrent drug with medicines that lower the effect of warfarin by inducing treatment and indication for treatment, have a significant its metabolism, while 74 were treated with medications that effect on mean warfarin dose and dose/BW. The following potentiate its effect. When the medical records were factors were significantly related to warfarin dose and dose/ examined in 2002, we found that 12 of the patients (6.0%) BW according to the generalised linear models (GLM) had suffered serious warfarin bleeding since the start of procedure in SAS: CYP2C9 genotype, age, weight, concur- therapy, and 41 (20.4%) had reported minor bleeding, rent drug treatment and indication for treatment (Table 3). mainly nose bleeding (19.5%). These factors together explain approximately 29% of the Two-thirds of the 201 patients carried the CYP2C9*1/*1 variation in warfarin doses (R2 E0.29). Least-squares means genotype. Patient characteristics were similar in carriers of (LS means) for the effects in the model with adjustment for the CYP2C9*1/*1 genotype and carriers of other CYP2C9 multiple comparisons were computed to describe how the genotypes (Table 1). CYP2C9, CYP3A5 and ABCB1 allele and variables affect dose (Table 4). Patients with the CYP2C9*1/ genotype frequencies were calculated (Table 2). In all, 2% of *1 genotype had the highest mean weekly dose of warfarin the individuals were predicted to be CYP2C9 poor meta- (Figure 1). This group had significantly higher doses than bolisers, that is, were genotyped CYP2C9*2/*2, *2/*3 or patients with the genotypes *1/*3 and *2/*2, *2/*3 or *3/*3. *3/*3. According to genotyping, 12% had an active CYP3A5 Patients carrying CYP2C9*1/*2 also differed significantly enzyme, that is, were genotyped CYP3A5*1/*1 or *1/*3. No from predicted poor metabolisers. The findings were similar CYP3A5*6 alleles were detected. for warfarin dose/BW (Tables 3 and 4). Patients treated with

The Pharmacogenomics Journal Warfarin related to CYP2C9, CYP3A5 and ABCB1 (MDR1) M Wadelius et al 43

Table 2 CYP2C9, CYP3A5 and ABCB1 allele and genotype frequencies in the 201 patients (402 alleles) in the warfarin dose study

CYP2C9 alleles % CYP2C9 genotypes %

*1 331 (82.3) *1/*1 134 (66.7) *2 45 (11.2) *1/*2 or *1/*3 63 (31.3) *3 26 (6.5) *2/*2, *2/*3 or *3/*3 4 (2.0)

CYP 3A5 allelesa CYP 3A5 genotypesa

*1 26 (6.5) *1/*1 2 (1.0) *3 372 (93.5) *1/*3 22 (11.1) *6 — *3/*3 175 (87.9)

ABCB1 allelesa ABCB1 genotypesa

Exon 1 (À12T4C) T 397 (99.7) Exon 1 (À12T4C) T/T 198 (99.5) C 1 (0.3) C/T 1 (0.5) Exon 2 (À1G4A) G 357 (88.8) Exon 2 (À1G4A) G/G 159 (79.1) A 45 (11.2) A/G 39 (19.4) A/A 3 (1.5) Exon 11 (1199G4A) G 388 (96.5) Exon 11 (1199G4A) G/G 188 (93.5) A 14 (3.5) A/G 12 (6.0) A/A 1 (0.5) Exon 12 (1236C4T) C 230 (57.2) Exon 12 (1236C4T) C/C 66 (32.8) T 172 (42.8) C/T 98 (48.8) T/T 37 (18.4) Exon 21 (2677G4T/A) G 214 (53.2) Exon 21 (2677G4T/A) G/G 57 (28.4) T 174 (43.3) G/T 91 (45.3) A 14 (3.5) T/T 40 (19.9) A/G 9 (4.5) A/T 3 (1.5) A/A 1 (0.5) Exon 26 (3396C4T) C 402 (100) Exon 26 (3396C4T) C/C 201 (100) Exon 26 (3435C4T) C 162 (40.3) Exon 26 (3435C4T) C/C 31 (15.4) T 240 (59.7) C/T 100 (49.8) T/T 70 (34.8) aGenotyping of CYP3A5 and ABCB1 Exon 1 failed in two patients.

Table 3 The factors CYP2C9, CYP3A5, ABCB1, age, gender, that induce warfarin metabolism, and younger patients weight, smoking, mean PT value, concurrent drug treatment tended to have higher doses than older patients. and indication for treatment were tested for covariance with Patients were divided into three equally sized dose/BW mean warfarin dose and dose/BW using GLMs procedure in SAS groups (33% in each group) to test for an association with CYP2C9 and CYP3A5 genotypes. The distribution over low, GLM of warfarin Dose Dose/BW medium and high dose/BW differed markedly between patients with the CYP2C91*/1* genotype: 27.6% low, CYP2C9 Po0.0001 Po0.0001 29.8% medium and 42.5% high, compared with patients Age P ¼ 0.0002 Po0.0001 Weight Po0.0001 P ¼ 0.0127 with other CYP2C9 genotypes: 44.8% low, 40.3% medium 2 PT P ¼ 0.0624 P ¼ 0.1216 and 14.9% high (w , 2 df, P ¼ 0.0004). No association Interaction P ¼ 0.0271 P ¼ 0.0213 between CYP3A5 allelic variants and dose/BW was found. Indication P ¼ 0.0344 P ¼ 0.0230 ABCB1 haplotypes were predicted based on seven variable sites and their frequencies calculated (Table 5). The ABCB1 D Factors clearly not affecting dose and dose/BW were removed from the model: CYP3A5 genotype, ABCB1 genotype, gender and smoking. haplotype occurred significantly more often than expected in the low-dose/BW group: 60% low, 25% medium and 15% high (w2, 2 df, P ¼ 0.0242, Table 5). There were no patients homozygous for the D haplotype. The 20 patients, who were warfarin because of heart valve prosthesis had significantly heterozygous for the D haplotype, had on an average 24% higher doses and doses/BW than other patients. Warfarin lower warfarin dose and dose/BW compared to the 181 doses were significantly increased by interaction with drugs patients who did not carry the D haplotype.

www.nature.com/tpj Warfarin related to CYP2C9, CYP3A5 and ABCB1 (MDR1) M Wadelius et al 44

Table 4 LS means were computed to describe the effect of studied factors on mean warfarin dose and dose/BW

LS means of warfarin Dose (mg), P-value Dose/BW (mg/kg), P-value

CYP2C9*1/*1 vs *1/*3 36.06 vs 24.11, P ¼ 0.0007 0.46 vs 0.31, P ¼ 0.0018 CYP2C9*1/*1 vs 2/*2, *2/*3 or *3/*3 36.06 vs 11.67, P ¼ 0.0083 0.46 vs 0.12, P ¼ 0.0041 CYP2C9*1/*2 vs 2/*2, *2/*3 or *3/*3 32.08 vs 11.67, P ¼ 0.0449 0.41 vs 0.12, P ¼ 0.0242 Heart valve prosthesis vs other indications 37.42 vs 32.29, P ¼ 0.0344 0.48 vs 0.41, P ¼ 0.0230 Enzyme-inducing drugs vs noninteracting drugs 52.01 vs 33.83, P ¼ 0.0309 0.67 vs 0.43, P ¼ 0.0278 Enzyme-inducing drugs vs interaction with warfarin-potentiating drugs 52.01 vs 31.93, P ¼ 0.0147 0.67 vs 0.41, P ¼ 0.0129

Multiple comparisons were adjusted by Tukey–Kramer’s method.

Table 5 Alignment and frequencies of predicted ABCB1 haplotypes in the dose requirement study

ABCB1 haplotypes A B C D E F G Others

Variable sites Exon 1 (À12T4C50untranslated region) T T T T T T T Exon 2 (À1G4A translation initiation) G G A G G G G Exon 11 (1199G4A S400N) G G G G A G G Exon 12 (1236C4T G412G) T C C C C C T Exon 21 (2677G4T/A A893S/T) T G G G G A T Exon 26 (3396C4T A1132A) C C C C C C C Exon 26 (3435C4T I1145I) T C T T C T C

Frequencies in dose study (201 patients)a 40.8% 33.3% 10.0% 5.0% 3.5% 2.5% 1.0% p1.0% each

Distribution of haplotypes in three dose groups Low dose/BW (o0.33 mg/kg) 33.3% 34.8% 24.3% 60.0% 30.8% 40.0% 25.0% Medium dose/BW (0.33-0.46 mg/kg) 35.7% 30.4% 29.7% 25.0% 23.1% 20.0% 25.0% High dose/BW (40.46 mg/kg) 31.0% 34.8% 46.0% 15.0% 46.2% 40.0% 50.0% P w2 0.5575 0.6016 0.1786 0.0242 0.5623 0.6564 0.7749 P Fischer’s exact 0.5769 0.6369 0.1821 0.0348 0.6778 0.7767 1.00 P trend 0.5888 1.00 0.0745 0.0094 0.4824 1.00 0.5362 aTwo out of 201 were not genotyped for Exon 1 (À12T4C), and their haplotypes are estimated as B/B and A/F, respectively, according to the results from polymorphisms in Exon s 2, 11, 12, 21, 26 and 26. The patients were divided into three equally large-dose/BW groups, and the distribution of haplotypes was compared between low-, medium- and high-dose patients. The D haplotype was significantly more common among low-dose patients and less common among high-dose patients than other haplotypes

In conclusion, CYP2C9 genotype, age, weight, concurrent bleeding using contingency tables and logistic regression. drug treatment and indication for treatment all significantly Prothrombin time was the only factor that differed sig- influenced warfarin dosing. One ABCB1 haplotype was nificantly between cases and controls (mean PT INR 3.56 associated with a low warfarin dose requirement. and 2.58, respectively); logistic regression, Po0.05. The frequency of the indication heart valve prosthesis, which Bleeding Complications requires a higher target PT INR, was similar among cases The second aim was to identify risk factors for serious (19.4%) and controls (24.3%). warfarin bleeding requiring hospital care. Cases had been treated with warfarin for 1 week up to 10 years (median: 6.5 months) prior to the bleeding episode. The most common DISCUSSION indications for treatment were atrial fibrillation, heart valve Genetic variation of CYP2C9 and its effect on warfarin prosthesis and deep vein thrombosis/pulmonary embolism. dosing has rendered great interest over the last few years.14 In all, 16 patients had suffered cerebral bleeding, six patients CYP2C9 is responsible for the metabolism of the active S- severe nose bleeding, two patients rectal bleeding and the warfarin. Our results corroborate previous studies by finding remaining 12 patients had other serious bleeding. a significant relationship between the number of functional None of the factors CYP2C9, CYP3A5 or ABCB1 (Table 6), CYP2C9 alleles and the required maintenance dose of age, gender, indication for treatment or interacting drugs warfarin. The CYP2C9*3 variant inactivates the enzyme could be identified as a significant risk factor for warfarin more than the CYP2C9*2 variant, which is clearly illustrated

The Pharmacogenomics Journal Warfarin related to CYP2C9, CYP3A5 and ABCB1 (MDR1) M Wadelius et al 45

Table 6 Results of CYP2C9, CYP3A5 and ABCB1 genotyping in the warfarin case–control study

CYP2C9 alleles Cases (%) Controls (%) CYP2C9 genotypes Cases (%) Controls (%)

*1 60 (83.3) 310 (82.0) *1/*1 124 (65.6) 25 (69.4) *2 9 (12.5) 42 (11.1) *1/*2 or *1/*3 62 (32.8) 10 (27.8) *3 3 (4.2) 26 (6.9) *2/*2, *2/*3 or *3/*3 3 (1.6) 1 (2.8)

CYP3A5 allelesa CYP3A5 genotypesa

*1 4 (5.9) 27 (7.2) *1/*1 0 3 (1.6) *3 64 (94.1) 347 (92.8) *1/*3 21 (11.2) 4 (11.8) *6 — — *3/*3 163 (87.2) 30 (88.2)

ABCB1 allelesb ABCB1 genotypesb

Exon 1 (À12T4C) T 66 (100) 373 (99.7) Exon 1 (À12T4C) T/T 186 (99.5) 33 (100) C 1 (0.3) C/T — 1 (0.5) Exon 2 (À1G4A) G 66 (91.7) 335 (88.6) Exon 2 (À1G4A) G/G 149 (78.8) 31 (86.1) A 6 (8.3) 43 (11.4) A/G 37 (19.6) 4 (11.1) A/A 3 (1.6) 1 (2.8) Exon 11 (1199G4A) G 70 (100) 364 (96.3) Exon 11 (1199G4A) G/G 176 (93.1) 35 (100) A — 14 (3.7) A/G — 12 (6.3) A/A — 1 (0.5) Exon 12 (1236C4T) C 36 (51.4) 217 (57.4) Exon 12 (1236C4T) C/C 64 (33.9) 7 (20.0) T 34 (48.6) 161 (42.6) C/T 89 (47.1) 22 (62.9) T/T 36 (19.0) 6 (17.1) Exon 21 (2677G4T/A) G 36 (50.0) 201 (53.2) Exon 21 (2677G4T/A) G/G 55 (29.1) 7 (19.4) T 36 (50.0) 163 (43.1) G/T 82 (43.4) 22 (61.1) A — 14 (3.7) T/T 39 (20.6) 7 (19.4) A/G — 9 (4.8) A/T — 3 (1.6) A/A — 1 (0.5) Exon 26 (3396C4T) C 72 (100) 378 (100) Exon 26 (3396C4T) C/C 189 (100) 36 (100) Exon 26 (3435C4T) C 29 (40.3) 153 (40.5) Exon 26 (3435C4T) C/C 30 (15.9) 4 (11.1) T 43 (59.7) 225 (59.5) C/T 93 (49.2) 21 (58.3) T/T 66 (34.9) 11 (30.6)

ABCB1 haplotypesb A B C D E F G Others

Cases 43.1% 34.7% 6.9% 5.6% 0.0% 0.0% 2.8% p2.8% each Controls 40.5% 33.1% 10.1% 4.8% 3.7% 2.6% 1.1% p1.1% each All 40.9% 33.3% 9.6% 4.9% 3.1% 2.2% 1.3% p1.1% each aTwo out of 36 cases and two out of 201 controls were not genotyped for CYP3A5 due to technical problems. bGenotyping of one case failed for ABCB1 polymorphisms in Exon s 11 and 12, and of three cases for Exon 1. Genotyping of two controls failed for ABCB1 polymorphism in Exon 1. There are 36 cases (72 alleles and haplotypes) and 189 controls (378 alleles and haplotypes). Cases have experienced serious warfarin bleeding, while controls have been treated without this side effect.

www.nature.com/tpj Warfarin related to CYP2C9, CYP3A5 and ABCB1 (MDR1) M Wadelius et al 46

CYP2C9 genotype vs. warfarin dose because of the risk of interactions, and care should be taken 50 when changing medication during warfarin therapy. When a new potentially interacting medicine is given to a warfarin 40 patient, the risk of bleeding is well recognised. The high- 30 dose requirement in patients on enzyme-inducing drugs is a reminder of the fact that it in some cases, it may be just as 20 important to increase warfarin doses to avoid thrombosis. mg/week Previous studies have shown that about 5% of warfarin- 10 treated patients experience major bleeding.31 Of the 201 patients recruited from the Uppsala University Hospital 0 CYP2C9*1/*1 *1/*2 *1/*3 *2/*2, *2/*3, *3/*3 anticoagulation clinic, 12 (6%) had at some time experi- (n=134) (n=42) (n=21) (n=4) enced a serious bleeding requiring hospital care. Thus, the Figure 1 Average weekly warfarin doses with error bars indicating incidence of major bleeding in our study is in good 95% confidence intervals within different CYP2C9 genotypes. agreement with previously published studies. Significantly Maintenance dose of warfarin was significantly related to the higher PT INR values were found in connection with the number of functional CYP2C9 genes (Table 4). bleeding. The frequency of heart valve prosthesis, which requires a higher target PT INR, was similar among cases and controls, indicating that the high INR found among cases by Figure 1. CYP3A enzymes, with activity derived from was not intentional, and may have been caused by both CYP3A5 and CYP3A4, are involved in the metabolism accidental overdosing. CYP3A5 and ABCB1 were not of R-warfarin.33 Only 12% of the individuals in this study identified as risk factors for bleeding complications and in were predicted to express CYP3A5, and there was no contrast to other studies, neither were CYP2C9 genotypes or indication that CYP3A5 influenced warfarin dosing. This alleles.15,16,20,22 However, the number of cases in this study pathway is probably less important for warfarin dosing is small and the results are not conclusive. Owing to the because R-warfarin is not as potent as S-warfarin, and sampling procedure, mainly patients who survived bleeding furthermore CYP3A5 hydroxylates R-warfarin at a lower complications were included in the study, which means that rate than CYP3A4.33 an important group of patients with fatal bleeding was The hitherto most studied ABCB1 polymorphism missed. 3435C4T in exon 26 is in itself silent, but there is evidence In summary, our results confirm that CYP2C9 genotyping of linkage disequilibrium extending from at least exons 26 can identify individuals who are poor metabolisers of to 12, indicating that the 3435C4T polymorphism may be a warfarin, and hence need lower doses of the drug. Owing marker for other polymorphisms within this block.34 to warfarin’s narrow therapeutic range and severe side Individuals homozygous for the 3435T polymorphism have effects, it would be advantageous if individuals requiring low levels of duodenal P-GP expression and increased extremely low doses of warfarin could be identified before bioavailability and higher plasma concentrations of digox- initiating therapy. These patients could be monitored more in.28,30 There are conflicting results from other studies, closely or in the future be offered a novel oral anticoagulant possibly explained by their use of different substrates and that does not require coagulation monitoring or dose ethnic groups.35,36 Our ABCB1 haplotype frequencies corre- adjustment.38 Age, weight, indication for treatment and late very well with those found in Caucasians in a recent concurrent medication are other important factors that study.37 We found that patients heterozygous for the ABCB1 should be taken into consideration when initiating warfarin haplotype D have a lower warfarin dose requirement than therapy. The effect of ABCB1 haplotype D on warfarin others. This haplotype contains the 3435C4T nucleotide dosing needs to be further evaluated in a larger study. exchange, and no change at the other studied positions. Low P-GP efflux activity in the hepatocyte leading to high levels of warfarin in the target cell could be an explanation MATERIALS AND METHODS as to why patients with the ABCB1 D haplotype appear to be Patients sensitive to warfarin. Owing to the limited number of For the warfarin dose study, 201 predominantly Caucasian individuals studied, this could be a chance observation, and patients were recruited at the Uppsala University Hospital a much larger study would be needed to confirm the results. anticoagulation clinic in 2000. They had been treated with A patient’s target PT INR is determined by the indication warfarin for a minimum of 2 months, and their PT INR for warfarin treatment. Patients with heart valve prosthesis values had been stabilised. At the five following visits, their have a higher target PT INR, and it was clearly shown that weekly warfarin doses and corresponding PT INR values they are given higher warfarin doses. BW is another factor were registered. Information about age, gender, BW (missing that self-evidently is taken into consideration when pre- in seven patients), smoking, treatment indication and scribing warfarin. Reflecting a decline in bodily functions, duration, other diseases and concurrent medication was elderly patients often need lower doses of medicines, and it taken from the patients’ medical records. Other drugs were was shown that warfarin dose requirement decreased with classified as interacting, if they had moderate or major age. Concurrent medications are extremely important interactions with warfarin according to the database

The Pharmacogenomics Journal Warfarin related to CYP2C9, CYP3A5 and ABCB1 (MDR1) M Wadelius et al 47

MICROMEDEXs Healthcare Series (http://www.microme explanatory variables have a significant effect on mean dex.com/ in May 2002). warfarin dose and mean warfarin dose per BW. To describe To find cases who had experienced serious bleeding how the measured factors affect dose, means were adjusted episodes requiring hospital care, the 201 patients’ medical for an imbalance in other variables by LS means. Adjustment records were gone through from start of therapy and up to for multiple comparisons was performed by the Tukey– 2002. The exclusion criteria were thrombolysis, surgery or Kramer method. The significance level was set at 0.05. trauma immediately preceding the bleeding. Of the 201 ABCB1 haplotypes were estimated according to Phase, patients, 12 classified as cases. In addition, 24 other patients which is a statistical method for haplotype reconstruction with serious warfarin bleeding were recruited; 20 through from population data.44 The association between the the Swedish spontaneous reporting of adverse drug reactions presence of allelic variants of the studied genes and warfarin and four from an ongoing national study on cerebral dose/BW was investigated by dividing subjects into three bleeding and warfarin.39 In total, 36 cases with serious equally sized groups. The distribution of allele-positive vs warfarin bleeding requiring hospital care were included, allele-negative subjects was compared over the three groups: while the remaining 189 patients at the anticoagulation low (o0.33 mg/kg), medium (0.33–0.46 mg/kg) and high clinic without serious bleeding were used as controls. The (40.46 mg/kg) dose categories. Deviation from expected Local Research Ethics Committee approved the study. distribution was evaluated by w2, Fisher’s exact and Trend tests. Genotyping In the case–control study, logistic regression and con- Deoxyribonucleic acid (DNA) was extracted from whole tingency tables were used to investigate whether CYP2C9, blood using standard procedures. Genotyping was per- CYP3A5 and ABCB1, age, gender, interactions, indications formed by three different methods based on polymerase or PT INR values were risk factors for bleeding. chain reaction (PCR). The variants CYP2C9*2 and *3 were identified using solid- ACKNOWLEDGEMENTS 40 phase minisequencing. In short, a DNA fragment spanning We thank all participating patients, nurses and doctors, especially the variable nucleotide position is amplified using PCR with RN Ulla Prose´n and RN Marine Sundin. We are indebted to Drs one biotinylated and one nonbiotinylated PCR primer. The Hans-Go¨ran Ha˚rdemark and Anders Carlsson for helping to hunt biotinylated fragment is captured on a streptavidin-coated down the cases. Dr Pa¨r Hallberg is thanked for interesting ideas and microtiter plate and rendered single stranded. The two alleles discussions. BSc Gunilla Frenne, Dr Sherwan Zindrou and Mr Lars are detected in two separate reactions using a single- So¨derlund have given invaluable technical assistance. Pyrosequen- cing AB provided reagents for CYP3A5 genotyping. Ms Kristi Kuljus nucleotide primer extension with radioactively labelled has provided statistical expertise. The Swedish Foundation for deoxynucleotide triphosphate (dNTP) for the variable base. Strategic Research, the Federation of County Councils, the Tore CYP3A5 was genotyped with pyrosequencing, which is a Nilson Foundation for Medical Research, the Royal Scientific method based on the quantification of pyrophosphate Society’s Research Foundation to the memory of Margit Ba¨xell, release during synthesis.41 Nucleotides are added sequen- and the foundations Gustaf Adolf Johansson and Mary, A˚ ke and tially, and the pyrophosphate released from incorporated Hans La¨ndell and the clinical research support (ALF) at Uppsala nucleotides triggers an enzymatic reaction yielding light. As University funded this study. the process continues, the complementary DNA strand is built up and the nucleotide sequences differentiatng CYP3A5*3 and CYP3A5*6 from CYP3A5*1 are determined DUALITY OF INTEREST from the signal peaks in the pyrogram. None declared. ABCB1 was genotyped by SNaPshot multiplex SNP detection.42,43 Multiplex PCR is used to amplify fragments for the following seven exonic SNPs: exon 1 (À12T4C), exon 2 ABBREVIATIONS (À1G4A), exon 11 (1199G4A), exon 12 (1236C4T), exon ABC adenosine triphosphate-binding cassette BW bodyweight 21 (2677G4T/A), exon 26 (3396C4T) and the hitherto CYP cytochrome P450 most studied polymorphism exon 26 (3435C4T). The PCR dNTP deoxynucleotide triphospate product is purified by enzymatic degradation of dNTPs and GLM generalised linear models procedure in SAS primers (1 U SAP, 2 U ExoI/reaction). Single base extension of LS means least-squares means MDR multidrug resistance multiple primers is performed by a cycle sequencing P-GP P-glycoprotein reaction with fluorescently labelled dideoxynucleotide PT INR prothrombin time international normalised ratio triphosphates. A tail of (GACT)n is added to the primers to avoid overlap between the final SNaPshot products that are detected with an ABI 3700 capillary electrophoresis instru- ment, and the results are analysed using ABI Genotyper 3.7. REFERENCES 1 Hardman JG, Limbird LE, Goodman Gilman A (eds). Goodman and Statistical Analysis Gilman’s. The Pharmacological Basis of Therapeutics, 10th edn McGraw- Hill Professional: New York 2001. Analysis of covariance was performed using the SAS software 2 Linder MW. Gentic mechanisms for hypersensitivity and resistance to GLM. The aim was to analyse which of the studied the anticoagulant warfarin. Clin Chim Acta 2001; 308: 9–15.

www.nature.com/tpj Warfarin related to CYP2C9, CYP3A5 and ABCB1 (MDR1) M Wadelius et al 48

3 Penning van Beest FJA, van Meegen E, Rosendaal FR, Stricker BHC. 26 Westlind-Johnsson A, Malmebo S, Johansson A, Otter C, Andersson TB, Characteristics of anticoagulant therapy and comorbidity related to Johansson I et al. Comparative analysis of CYP3A expression in human overanticoagulation. Thromb Haemost 2001; 86: 569–574. liver suggests only a minor role for CYP3A5 in drug metabolism. Drug 4 Horton JD, Bushwick BM. Warfarin therapy: evolving strategies in Metab Dispos 2003; 31: 755–761. anticoagulation. Am Fam Physician 1999; 3: 635–646. 27 Sussman NL, Waltershied M, Butler T, Cali JJ, Riss T, Kelly JH. The 5 Landefeld CS, Beyth RJ. Anticoagulant-related bleeding: clinical epide- predictive nature of high throughput toxicity screening using a human miology, prediction and prevention. Am J Med 1993; 95: 315–328. hepatocyte cell line. Cell Notes 2002; 3: 7–10, http://www.promega.- 6 Hirsh J, Dalen JE, Anderson DR, Poller L, Bussey H, Ansell J et al. Oral com/cnotes/cn003/cn003_07.htm. anticoagulants: mechanism of action, clinical effectiveness, and optimal 28 Brinkmann U, Eichelbaum M. Polymorphisms in the ABC drug therapeutic range. Chest 1998; 114: 445S–469S. transporter gene MDR1. Pharmacogenomics J 2001; 1: 59–64. 7 Takahashi H, Echizen H. Pharmacogenetics of CYP2C9 and interindivi- 29 Wandel C, Kim RB, Kajiji S, Guengerich P, Wilkinson GR, Wood AJ. dual variability in anticoagulant response to warfarin. P-glycoprotein and cytochrome P-450 3A inhibition: dissociation of Pharmacogenomics J 2003; 3: 202–214. inhibitory potencies. Cancer Res 1999; 59: 3944–3948. 8 Lubetsky A, Dekel-Stern E, Chetrit A, Lubin F, Halkin H. Vitamin K intake 30 Hoffmeyer S, Burk O, von Richter O, Arnold HP, Brockmo¨ller J, Johne A and sensitivity to warfarin in patients consuming regular diets. Thromb et al. Functional polymorphisms of the human multidrug-resistance Haemost 1999; 81: 396–399. gene: multiple sequence variations and correlation of one allele with 9 Wells PS, Holbrook AM, Crowther NR, Hirsh J. Interactions of warfarin P-glycoprotein expression and activity in vivo. Proc Natl Acad Sci USA with drugs and food. Ann Intern Med 1994; 121: 676–683. 2000; 97: 3473–3478. 10 Fihn SD, Callahan CM, Martin DC, McDonell MB, Henikoff JG, White 31 Taghavi A, Jonson T, Stockelberg D. Survey of complications following RH. The risk for and severity of bleeding complications in elderly treatment with anticoagulants. A computerized search for hemorrhagic patients treated with warfarin. Ann Intern Med 1996; 124: 970–979. complications completes manual reporting.. La¨kartidningen 1999; 96: 11 Loebstein R, Yonath H, Peleg D, Almog S, Rotenberg M, Lubetsky A et 3421–3424. al. Interindividual variability in sensitivity to warfarin—nature or 32 Levine MN, Raskob G, Landefeld S, Kearon C. Hemorrhagic complica- nurture? Clin Pharmacol Ther 2001; 70: 159–164. tions of anticoagulant treatment. Chest 1998; 114: 511S–523S. 12 Harder S, Thurmann P. Clinically important drug interactions with 33 Rettie AE, Korzekwa KR, Kunze KL, Lawrence RF, Eddy AC, Aoyama T et anticoagulants. An update. Clin Pharmacokinet 1996; 30: 416–444. al. Hydroxylation of warfarin by human cDNA-expressed cytochrome 13 Kaminsky LS, Zhang Z. Human P450 metabolism of warfarin. Pharmacol P-450: a role for P-450 in the etiology of (S)-warfarin–drug interactions. Ther 1997; 73: 67–74. Chem Res Toxicol 1992; 5: 54–59. 14 Daly AK, King BP. Pharmacogenetics of oral anticoagulants. Pharmaco- 34 Siddiqui A, Kerb R, Wheale ME, Brinkmann U, Smith A, Goldstein DB genetics 2003; 13: 247–252. et al. Association of multidrug resistance in epilepsy with a polymorph- 15 Aithal GP, Day CP, Kesteven PJ, Daly AK. Association of polymorphisms ism in the drug-transporter gene ABCB1. N Engl J Med 2003; 348: in the cytochrome P450 CYP2C9 with warfarin dose requirement and 1442–1448. risk of bleeding complications. Lancet 1999; 353: 717–719. 35 Kim RB, Leake BF, Choo EF, Dresser GK, Kubba SV, Schwarz UI et al. 16 Higashi MK, Veenstra DL, Kondo LM, Wittkowsky AK, Srinouanpra- Identification of functionally variant MDR1 alleles among European chanh SL, Farin FM et al. Association between CYP2C9 genetic variants Americans and African Americans. Clin Pharmacol Ther 2001; 70: and anticoagulation-related outcomes during warfarin therapy. JAm 189–199. Med Assoc 2002; 287: 1690–1698. 36 Drescher S, Schaeffeler E, Hitzl M, Hofmann U, Schwab M, Brinkmann 17 Sullivan-Klose TH, Ghanayem BI, Bell DA, Zhang ZY, Kaminsky LS, U et al. MDR1 gene polymorphisms and disposition of the Shenfield GM et al. The role of the CYP2C9-Leu 359 allelic variant in the P-glycoprotein substrate fexofenadine. Br J Clin Pharmacol 2002; 53: tolbutamide polymorphism. Pharmacogenetics 1996; 6: 341–349. 526–534. 18 Furuya H, Fernandez Salguero P, Gregory W, Taber H, Steward A, 37 Kroetz DL, Pauli-Magnus C, Hodges LM, Huang CC, Kawamoto M, Gonzalez FJ et al. Genetic polymorphism of CYP2C9 and its effect on Johns SJ et al. Sequence diversity and haplotype structure in the human warfarin maintenance dose requirement in patients undergoing antic- ABCB1 (MDR1, multidrug resistance transporter) gene. Pharmacoge- oagulation therapy. Pharmacogenetics 1995; 5: 389–392. netics 2003; 13: 481–494. 19 Steward DJ, Haining RL, Henne KR, Davis G, Rushmore TH, Trager WF et 38 Iqbal O, Aziz S, Hoppensteadt DA, Ahmad S, Walenga JM, Bakhos M et al. Genetic association between sensitivity to warfarin and expression of al. Emerging anticoagulant and thrombolytic drugs. Exp Opin Emerg CYP2C9*3. Pharmacogenetics 1997; 7: 361–367. Drugs 2001; 6: 111–135. 20 Ogg MS, Brennan P, Meade T, Humphries SE. CYP2C9*3 allelic variant 39 Sjo¨blom L, Ha˚rdemark HG, Lindgren A, Norrving B, Fahlen M, and bleeding complications [Comment]. Lancet 1999; 354: 1124. Samuelsson M et al. Management and prognostic features of 21 Taube J, Halsall D, Baglin T. Influence of cytochrome P-450CYP2C9 intracerebral hemorrhage during anticoagulant therapy: a Swedish polymorphisms on warfarin sensitivity and risk of over-anticoagulation multicenter study. Stroke 2001; 32: 2567–2574. in patients on long-term treatment. Blood 2000; 96: 1816–1819. 40 Syva¨nen AC. Solid phase mini-sequencing. In: Taylor GR (ed). 22 Margaglione M, Colaizzo D, D’Andrea G, Brancaccio V, Ciampa A, Laboratory Methods for the Detection of Mutations and Polymorphism. Grandone E et al. Genetic modulation of oral anticoagulation with CRC Press: Boca Raton, FL 1997: 53–64. warfarin. Thromb Haemost 2000; 84: 775–778. 41 Ronaghi M, Uhle´n M, Nyre´n P. A sequencing method based on real- 23 Tabrizi AR, Zehnbauer BA, Borecki IB, McGrath SD, Buchman TG, time pyrophosphate. Science 1998; 281: 363–365. Freeman BD. The frequency and effects of cytochrome P450 (CYP) 2C9 42 Lindblad-Toh K, Winchester E, Daly MJ, Wang DG, Hirschhorn polymorphisms in patients receiving warfarin. J Am Coll Surg 2002; 194: JN, Laviolette JP et al. Large scale discovery and genotyping of 267–273. single-nucleotide polymorphisms in the mouse. Nat Genet 2000; 24: 24 Scordo MG, Pengo V, Spina E, Dahl ML, Gusella M, Padrini R. Influence 381–386. of CYP2C9 and CYP2C19 genetic polymorphisms on warfarin main- 43 Makridakis NM, Reichardt JK. Multiplex automated primer extension tenance dose and metabolic clearance. Clin Pharmacol Ther 2002; 72: analysis: simultaneous genotyping of several polymorphisms. Biotechni- 702–710. ques 2001; 31: 1374–1380. 25 Kuehl P, Zhang J, Lin Y, Lamba J, Assem M, Schuetz J et al. Sequence 44 Stephens M, Smith NJ, Donnelly P. A new statistical method for diversity in CYP 3A promoters and characterization of the genetic basis haplotype reconstruction from population data. Am J Hum Genet 2001; of polymorphic CYP 3A5 expression. Nat Genet 2001; 27: 383–391. 68: 978–989.

The Pharmacogenomics Journal