Association of CYP2C19*2 and associated haplotypes with lower norendoxifen concentrations in tamoxifen‐treated Asian breast cancer patients

What is Already Known about this Subject

• Most studies have investigated the associations between CYP2D6 variants and tamoxifen pharmacokinetics, but the impact of CYP2C19 genetic variants on tamoxifen disposition has been less extensively studied.

• Recent findings suggest that a tertiary metabolite of tamoxifen, 4‐hydroxy‐N,N‐didesmethyltamoxifen or norendoxifen (NorEND), exhibits potent aromatase inhibitory activity and estrogen receptor antagonism.

• To date, the influence of CYP2C19 polymorphisms and haplotypes on the pharmacokinetics of NorEND is not known.

What this Study Adds

• Of the 66 CYP2C19 polymorphisms identified in healthy Asians, 14 were found to be tightly linked with CYP2C19*2, CYP2C19*3 and CYP2C19*17.

• No significant associations were observed between these CYP2C19 functional polymorphisms and their linked SNPs with the steady‐state concentrations of tamoxifen, N‐desmethyltamoxifen (NDM), (Z)‐4‐hydroxytamoxifen and (Z)‐endoxifen.

• However, two other potentially important metabolites, N,N‐didesmethyltamoxifen (NDDM) and NorEND were genetically associated. The CYP2C19 H2 haplotype, which included CYP2C19*2, was significantly correlated with lower plasma concentrations of NorEND, and metabolic ratios MR_NorEND/NDDM (P < 0.0001) and MR_{NorEND/(Z)‐END} (P = 0.001), indicating significantly lower formation rates of NorEND.

Introduction

The selective estrogen receptor modulator tamoxifen is frequently employed in the treatment of hormone receptor (HR) positive breast cancer in both pre‐ and post‐menopausal women 1. Despite clear evidence demonstrating its efficacy, treatment resistance and relapse have been observed in approximately 30% of all tamoxifen‐treated patients 2, 3. Wide variations in the plasma concentrations of tamoxifen and its metabolites have been reported 3, 4, and altered rates of formation of its active metabolites are believed to contribute at least partially to the differential treatment response.

Tamoxifen undergoes extensive metabolism by CYP3A4/5, CYP2D6 and other cytochrome P450 (CYP) enzymes to form primary metabolites N‐desmethyltamoxifen (NDM) and 4‐hydroxytamoxifen (4‐OHT). Both metabolites are further converted to endoxifen 5, which exhibits an anti‐estrogenic potency similar to that of 4‐OHT but is present at 5–10 times higher plasma concentrations than 4‐OHT in tamoxifen‐treated patients 6, 7. Other CYP enzymes, including CYP1A2, CYP2B6, CYP2C9 and CYP2C19, have also been shown to be involved in catalyzing the 4‐hydroxylated metabolite formation 8, 9, 10. In addition, recent findings suggest that a tertiary metabolite of tamoxifen, 4‐hydroxy‐N,N‐didesmethyltamoxifen or norendoxifen (NorEND), exhibits a potent aromatase inhibitory activity comparable with that of letrozole, a clinically used aromatase inhibitor 11. In addition to estrogen receptor antagonism by tamoxifen and endoxifen, aromatase inhibition by NorEND may contribute to the overall clinical efficacy of tamoxifen by lowering the peripheral conversion of androgens to estrogen. Lu et al. 11 also suggested that its dual mode of action as an anti‐estrogen and aromatase inhibitor highlights its potential importance as a critical active moiety of tamoxifen, higher concentrations of which may lead to improved treatment outcomes. Although the exact metabolic routes are still unknown, previous studies have suggested that NorEND may be formed via N‐demethylation of endoxifen or 4‐hydroxylation of N‐didesmethyltamoxifen (NDDM) 11, 12.

Thus far, most studies have examined the impact of CYP2D6 variants on tamoxifen pharmacokinetics and treatment outcomes 13, 14, 15, 16, 17, 18. Among the other CYP enzymes involved in the biochemical pathway, the influence of CYP2C19 activity on the disposition of tamoxifen and its metabolites has recently generated considerable interests 4, 15, 19, 20, 21, 22, 23. Several studies have suggested a more prominent role of CYP2C19 in influencing tamoxifen metabolism or outcome 19, 23. Moreover, recent findings by Saladores et al. 24 showed that CYP2C19 loss‐of‐function polymorphisms correlated with the metabolic ratio (MR) of NDDM/NorEND, suggesting that the formation of norendoxifen from NDDM may be influenced by the differential activities of CYP2C19 polymorphic variants. Therefore, genetic variation in CYP2C19 likely affects the metabolic conversions of tamoxifen metabolite precursors to NorEND, which may in turn contribute to the variability in the pharmacokinetics and treatment outcomes of tamoxifen.

CYP2C19 activity exhibits wide inter‐individual and inter‐ethnic variabilities 25, 26. This has been mainly attributed to the presence of deficiency alleles, such as CYP2C19*2 (618G > A; rs4244285) and *3 (636G > A; rs4986893) 25, 27. Of note, approximately 20% of Asians are CYP2C19 poor metabolizers (PM) carrying two copies of defective CYP2C19 alleles. In contrast, only 5% of the Caucasian and African populations exhibit the PM phenotype 28, 29. The higher frequencies of CYP2C19 polymorphisms, particularly CYP2C19*2 and CYP2C19*3, in Asians underscore the potential relevance of CYP2C19 pharmacogenetics on the disposition of tamoxifen and its metabolites in this patient population. Moreover, previous studies have only considered the effects of known CYP2C19 functional polymorphisms, such as CYP2C19*2, CYP2C19*3 and CYP2C19*17 [−3402C > T (rs11188072) and ‐806C > T (rs12248560)] 15, 19, 20 and very few studies have comprehensively screened the CYP2C19 gene for genetic variations which could potentially influence its activity 30, 31, 32, 33.

In view of the lack of data on the impact of CYP2C19 variants on tamoxifen pharmacogenetics in Oriental Asian breast cancer patients, this study was undertaken with a two‐fold objective, firstly, to identify comprehensively genetic variants in the CYP2C19 gene in each of the three Asian ethnicities, Chinese, Malays and Indians and secondly, to investigate the influence of CYP2C19 functional polymorphisms (CYP2C19*2, *3 and *17) and their associated haplotypes on the disposition of tamoxifen and its metabolites, particularly NorEND, in Asian breast cancer patients receiving adjuvant tamoxifen therapy.

Methods

Study subjects

The healthy subjects comprised of three Asian populations in Singapore, namely Chinese, Malays and Indians (n = 80 each). Ethnicities were verified against their National Registry Identification Cards (NRICs), which would be the most accurate way of assigning ethnicites since information on the ethnicities of their parents are also verified at the time of registration. A total of 201 patients who were histologically diagnosed with HR‐positive breast tumours and receiving tamoxifen 20 mg daily for at least 8 weeks were prospectively recruited, as described previously 4. Menopausal status was determined by recruiting clinicians at the point of recruitment by confirming that menstrual periods had stopped for at least 12 full months in post‐menopausal patients. In cases of uncertainty, follicle‐stimulating hormone (FSH) and estradiol concentrations were determined. Compliance was secured by measuring tamoxifen concentrations at the time of blood sampling. Patients who had received CYP2D6 or CYP2C19 modulators, including rifampicin, phenytoin, omeprazole, cimetidine as well as anti‐depressants such as selective serotonin re‐uptake inhibitors, within 4 weeks prior to enrolment were excluded. Recruitment of healthy subjects was in accordance with the guidelines of the ethics review committee of the National University Hospital, Singapore. The study was approved by the ethics review committee of the National Cancer Centre, Singapore (Reference number: 2008/402/B). All healthy subjects and patients provided written consent. The study is registered with the Health Sciences Authority Singapore (Registration number: HPRG/CTB 78:10/08–053).

Pharmacogenetic analyses of CYP2C19

Genomic DNA was extracted from peripheral blood mononuclear cells of healthy subjects and tamoxifen‐treated breast cancer patients using ethanol precipitation. The exons as well as intron‐exon boundaries of the CYP2C19 gene were comprehensively screened in 240 healthy subjects for the presence of genetic variations. Fifteen pairs of primers were designed using Primer 3 software 34 to cover all nine exons (including exon‐intron boundaries) as well as the 5′ upstream (−3.5 kb from translational start site) and 3′ downstream (2.5 kb) regions of the CYP2C19 gene (UCSC RefSeq: NM_000769.1) and submitted to NCBI Primer‐BLAST (http://www.ncbi.nlm.nih.gov/tools/primer‐blast/) to ensure binding specificities. PCR amplicons were treated with shrimp alkaline phosphatase and exonuclease I, followed by sequencing on an ABI 3730 DNA Analyzer (Applied Biosystems Inc., CA, USA). The electropherograms were aligned against the reference sequence (UCSC RefSeq: NM_000769.1) with SeqScape^® v2.5 (Applied Biosystems Inc., CA, USA). Pairwise linkage disequilibrium (LD) analyses were subsequently performed on the polymorphisms found in each of the three ethnic groups. Seventeen CYP2C19 variants, which included three functional SNPs (CYP2C19*2, *3 and *17) and 14 SNPs tightly linked to these three SNPs, were subsequently selected for sequencing in breast cancer patients treated with tamoxifen.

Pharmacogenetic analyses of CYP2D6

Due to the prominent role of CYP2D6 in tamoxifen metabolism, screening of 15 CYP2D6 polymorphic variants in 201 Asian breast cancer patients was performed using the INFINITI™ CYP450 2D6I assay (AutoGenomics Inc., CA, USA) as described previously 4. The classification of CYP2D6 allelic and phenotypic groups is provided in the Supplementary Information. Five patients were excluded from the analysis, one due to poor DNA quality and four due to uncertain genotype‐predicted phenotypic effects of multiple alleles. The CYP2D6 genotypes of these four patients are listed in Supplementary Table S1.

LC–MS/MS analyses of tamoxifen and its metabolites

Steady state blood samples (3 ml) were drawn from recruited patients at ≥8 weeks after the start of tamoxifen therapy. Within 30 min of venipuncture, plasma separation was carried out by centrifugation under light protection. Plasma samples were stored at −80°C until analysis. Steady‐state plasma concentrations of tamoxifen and 29 of its metabolites were quantified as described previously 15, via LC–MS/MS multiple reaction monitoring mode, performed on an Agilent 1200 Series Rapid Resolution LC System coupled to a 6460 triple quadrapole mass spectrometer with a Jet Stream electrospray source (Agilent Technologies Inc., CA, USA). Briefly, 100 μl of 1% of acetic acid in acetonitrile containing the mixture of deuterated internal standards was added to 50 μl of plasma. After centrifugation, the clear supernatant was diluted with 0.1% acetic acid and chromatographic separation of the isomers of tamoxifen metabolites was achieved on an Eclipse Plus C18 rapid resolution HPLC column with a gradient of acetonitrile in 0.1% acetic acid using an Agilent 1200 Series Rapid Resolution LC System. A 6460 triple quadrupole mass spectrometer with a Jet Stream electrospray source (Agilent Technologies Inc., CA, USA) was used in multiple reaction monitoring mode to quantify the analytes. The calibration ranges and validation data of the LC–MS/MS assay for the quantification of plasma concentrations of tamoxifen and its metabolites are tabulated in Supplementary Table S2. Plasma concentrations of analytes below the lower limit of quantification (LLOQ) were taken to be half the LLOQ 35.

Due to the involvement of multiple drug metabolizing enzymes in the biochemical pathway of tamoxifen, the plasma concentrations of individual metabolites can be influenced by multiple enzymes catalyzing its formation and elimination. Therefore, metabolic ratios (MR) were used in addition to plasma concentrations which directly indicate the conversion of metabolites of interest 13, 36. MRs were calculated as plasma concentrations (C) of derived compound relative to parental compound as follows: MR_NDM/TAM = C _NDM/C _TAM, MR_{(Z)‐4‐OHT/TAM} = C _{(Z)‐4‐OHT}/C _TAM (×100), MR_{(Z)‐END/NDM} = C _(Z)‐END/C _NDM (×100), MR_{(Z)‐END/(Z)‐4‐OHT} = C _(Z)‐END/C _{(Z)‐4‐OHT} and MR_NDDM/NDM = C _NDDM/C _NDM (×10). Since previous studies have suggested that NorEND may be formed via the N‐demethylation of endoxifen or 4‐hydroxylation of NDDM 11, we further evaluated MR_{NorEND/(Z)‐END} = C _NorEND/C _(Z)‐END (×100) and MR_NorEND/NDDM = C _NorEND/C _NDDM (×100). The (E)‐norendoxifen isomer was not analyzed due to low plasma concentrations.

Statistical analyses

Since this study is exploratory with respect to the discovery of novel CYP2C19 polymorphisms and unknown effect sizes of known functional variants on tamoxifen metabolite formation, no sample size/power analysis was carried out a priori. Associations between CYP2C19 polymorphisms and plasma concentrations of analytes and metabolic ratios were determined using Kruskal–Wallis and Mann–Whitney U‐tests in SPSS v14 (IBM, IL, USA). To assign the haplotype status for each patient, haplotype phasing was conducted using PLINK v1.02 37. The effects of CYP2C19*2 and haplotype of the LD blocks within the CYP2C19 gene were examined using the haplotype‐specific generalized linear model (haplo.glm) under the haplostats package in R‐software. For this purpose, all parameters were natural log‐transformed (ln‐transformed), with the exception of plasma concentrations of tamoxifen, (Z)‐4‐OHT and NorEND as well as MR_NorEND/NDDM, which were square root‐transformed to ensure conformity with normal distribution. Statistically significant associations were further adjusted for significant clinical and genetic covariates [race, age, body mass index (BMI), menopausal status, prior radiotherapy, prior chemotherapy and CYP2D6 metabolizer phenotype], which were identified with the univariate General Linear Model (GLM) procedure in SPSS v14 (IBM, IL, USA). Jonckheere–Terpstra tests were conducted in SPSS v14 (IBM, IL, USA) to assess for statistically significant trends between haplotype copy number and plasma NorEND concentrations as well as metabolic ratios. All tests were two‐sided and the level of significance was set at P < 0.05.

Results

Patient demographics

The median age of the patients was 49 years (range 31–74 years), with nearly 80% being premenopausal. The median height and weight were 156 cm (range 134–172 cm) and 58.0 kg (range 39.0–91.7 kg), respectively. The majority of patients were of Chinese descent (83.1%), followed by Indians (9.4%) and Malays (7.5%).

CYP2C19 genetic variations in Asians

A total of 66 genetic variants, of which 25 were novel, were identified from the comprehensive screening of all nine exons (including exon‐intron boundaries) as well as the upstream and downstream regions in 240 healthy subjects from all ethnic groups (Supplementary Table S3). In silico analyses using FastSNP indicated that 15 of these polymorphisms were located in putative binding sites for transcription factors which could potentially alter the binding of regulatory proteins and lead to variable CYP2C19 gene expression (Supplementary Table S4). Significant inter‐ethnic variability in the genotypic frequencies of eight polymorphisms were observed after Bonferroni adjustment for multiple comparisons [−3402C > T (*17; rs11188072), −806C > T (*17; rs12248560), IVS1 + 359 A > T (rs7918461), IVS1‐231G > A (rs7916649), 518C > T (rs61311738), IVS7‐106 T > C (rs4917623), IVS8‐119C > T (rs12268020) and * + 1189_* + 1191delCCA] (P < 0.00076, Supplementary Table S3).

Linkage disequilibrium analysis in healthy Orientals

Pairwise LD analyses were performed on the polymorphisms found in each of the three ethnic groups, Chinese, Malays and Indians (Supplementary Figures S1 A, B, and C, respectively). Of the 66 genetic variants identified, 17 CYP2C19 variants, which included three functional SNPs (CYP2C19*2, *3 and *17) and 14 SNPs tightly linked to these three SNPs, were selected for genotyping in patients. Of note, the functional polymorphism 636G > A (*3; rs4986893) was in high LD with –3332 A > C (rs7101258), IVS5‐154delG (rs56007608), IVS1‐47G > A (rs17878649), 1251 A > C (rs17886522) and *1264C > G (│D’│ = 1, r ² > 0.88). Similarly, 681G > A (*2; rs4244285) was found to be in tight linkage with seven other polymorphic variants [−3266G > T (rs4532967), IVS2‐23 A > G (rs12769205), IVS5 + 228 A > G (rs12571421), IVS5‐51C > G (rs4417205), IVS6‐196 T > A, 990C > T and * + 2054G > T (│D’│ > 0.95, r ² > 0.89)]. Consistent with previous findings in Caucasians 38, –806C > T (*17; rs12248560) was observed to be in complete linkage with ‐3402C > T (*17; rs11188072) in all three ethnic groups (D’ = 1, r ² = 1). These two CYP2C19*17 polymorphisms were also strongly linked to IVS8‐119C > T (rs12268020) and *1189_*1191DelCCA (D’ = 1, r ² > 0.88).

Effects of CYP2C19 polymorphisms on the disposition of tamoxifen and its metabolites

No significant associations were observed between any of the CYP2C19 functional polymorphisms or their 14 linked SNPs with plasma concentrations of tamoxifen and its three main metabolites (NDM, (Z)‐4‐OHT, (Z)‐endoxifen or their respective MRs, MR_NDM/TAM, MR_{(Z)‐END/NDM}, MR_{(Z)‐4‐OHT/TAM}, MR_{(Z)‐END/(Z)‐4‐OHT} and MR_NDDM/NDM (Supplementary Table S5).

However, CYP2C19*2 and each of its seven linked SNPs were independently associated with significantly lower plasma concentrations of NorEND, MR_{NorEND/(Z)‐END} and MR_NorEND/NDDM (Supplementary Table S5). Patients with homozygous variant alleles at the CYP2C19*2 locus had significantly lower NorEND levels [median (range), GG vs. GA vs. AA: 1.51 (0.38–3.28) vs. 1.28 (0.30–3.36) vs. 1.15 ng ml⁻¹ (0.26–2.45, P = 0.010)] as well as significantly lower MR_{NorEND/(Z)‐END} [median (range), GG vs. GA vs. AA: 9.40 (3.27–28.35) vs. 8.15 (2.67–18.9) vs. 6.06 (4.47–14.6), P < 0.0001] and MR_{NorEND /NDDM} [median (range), GG vs. GA vs. AA: 2.75 (0.62–6.26) vs. 2.43 (0.96–4.18) vs. 1.75 (1.10–2.49), P < 0.00001], indicating significantly lower formation rates of NorEND in these patients (Supplementary Table S5). Strong associations were observed between CYP2C19*2 and NorEND levels (P < 0.001), even after adjusting for age, BMI and CYP2D6 metabolizer status (Table 1). Similarly, strong associations were also observed between CYP2C19*2 and MR_{NorEND/(Z)‐END} (P < 0.001) as well as MR_NorEND/NDDM (P < 0.001), even after adjusting for age and CYP2D6 metabolizer status (Table 1).

Table 1.

Effect of CYP2C19*2 on (A) plasma concentration of NorEND, (B) MR_NorEND/NDDM, (C) MR_{NorEND/(Z)‐END} and (D) MR_NDDM/NDM

CYP2C19*2 and covariates	Coefficient	Standard error	t statistic	P value
(A) Plasma concentration of NorEND
Intercept	1.459	0.132	11.076	‐
CYP2C19*2	−0.075	0.021	−3.508	<0.001
Age	0.006	0.002	2.854	0.005
BMI	‐0.020	0.004	‐5.138	0.006
CYP2D6 hetEM	0.006	0.046	0.134	0.893
CYP2D6 IM	‐0.101	0.043	‐2.357	0.019
CYP2D6 UM	0.123	0.091	1.357	0.176
(B) MR NorEND/NDDM
Intercept	2.414	0.156	15.497	‐
CYP2C19*2	−0.169	0.0262	−6.477	<0.001
Age	‐0.005	0.003	‐2.115	0.036
BMI	‐0.019	0.005	‐4.056	0.049
(C) MR NorEND/(Z)‐END
Intercept	2.335	0.168	13.875	‐
CYP2C19*2	−0.131	0.034	−3.893	<0.001
Age	‐0.008	0.003	‐2.369	0.019
CYP2D6 hetEM	0.119	0.072	1.649	0.101
CYP2D6 IM	0.550	0.067	8.175	<0.00001
CYP2D6 UM	0.199	0.142	1.408	0.161
(D) MR NDDM/NDM
Intercept	1.170	0.076	15.385	‐
CYP2C19*2	0.057	0.015	3.759	<0.001
Age	0.004	0.001	2.611	0.010
CYP2D6 hetEM	‐0.100	0.033	‐3.069	0.002
CYP2D6 IM	‐0.199	0.030	‐6.564	<0.001
CYP2D6 UM	‐0.086	0.064	‐1.345	0.180

CYP2D6 EM was the reference in the model. Data in bold font pertain to CYP2C19*2 as well as significant P values (<0.05).

Additionally, a sub‐group analysis performed in Indian breast cancer patients with highest frequencies of both CYP2C19*17 polymorphisms among the three ethnicities revealed a significant association between the gain‐of‐function CYP2C19*17 polymorphisms (−3402 and –806C > T) and median MR_{NorEND/(Z)‐END}, CC vs. CT genotype of 5.99 (3.85–7.90) vs. 7.22 (6.27–13.9), P = 0.01. No further ethnic specific conclusions could be drawn with respect to CYP2C19*17 effects on tamoxifen metabolism because there were no patients with the homozygous variant TT genotype in Indians and CYP2C19*17 genotype frequencies in Chinese patients were too low. The frequencies of CYP2C19*3 genotypes were also too low to perform genotype–phenotype association analyses.

Effects of CYP2C19 haplotypes on the disposition of tamoxifen and its metabolites

Since haplotype analyses have been suggested to be more robust than single marker analysis 39, linkage disequilibrium and haplotype phase were explored to evaluate multi‐locus genotype–phenotype associations. As the frequencies of the haplotypes containing the CYP2C19*3 and *17 SNPs were too low (<5%) to permit further investigation, subsequent association analyses only included the CYP2C19*2 containing haplotype.

Haplotype assessment of the seven SNPs in LD with CYP2C19*2 led to the inference of two major haplotypes, H1 reference haplotype (frequency 64%) and H2 haplotype (frequency 33%, Table 2). After adjusting for age, BMI and CYP2D6 metabolizer phenotypes, the H2 haplotype was significantly associated with lower plasma NorEND concentrations (P = 0.001, Table 3). Patients homozygous for the H2 haplotype had 22.5% and 12.0% lower median plasma NorEND concentrations compared with those homozygous for haplotype H1 and those with one copy of each haplotype (H1/H2), respectively (P = 0.007 and P = 0.07, respectively). This was reflected by a gradual decrease of median NorEND concentration as per H2 copy numbers: H1/H1 1.51 ng ml⁻¹ (0.38–3.28), H1/H2 1.33 ng ml⁻¹ (0.30–3.36), H2/H2 1.17 ng ml⁻¹ (0.26–2.45) (Figure 1A). The Jonckheere–Terpstra test for ordered alternatives showed that there was a statistically significant trend of lower median plasma NorEND concentrations with increasing copy number of haplotype H2 (P = 0.020).

Table 2.

Frequencies of CYP2C19 associated haplotypes in Asian breast cancer patients

CYP2C19 haplotypes	Haplotype frequencies	−3266G > T	IVS2‐23 A > G	681G > A (*2)	IVS5 + 228 A > G	IVS5‐51C > G	IVS6‐196 T > A	990C > T	*2054G > T
Reference haplotype (Haplotype 1)	0.6343	G	A	G	A	C	T	C	G
Haplotype 2	0.3383	T	G	A	G	G	A	T	T
Haplotype 3 **	0.0050	G	A	G	A	C	A	C	G
Haplotype 4 **	0.0050	T	G	A	G	G	A	C	G
Haplotype 5 **	0.0025	G	A	G	A	C	T	T	G
Haplotype 6 **	0.0025	G	A	G	A	G	T	C	G
Haplotype 7 **	0.0025	T	A	G	A	C	T	C	G
Haplotype 8 **	0.0025	T	A	A	G	G	A	T	T
Haplotype 9 **	0.0025	T	G	A	A	C	A	T	T
Haplotype 10 **	0.0025	T	G	A	A	G	A	T	T
Haplotype 11 **	0.0025	T	G	A	G	G	A	T	G

^** Rare haplotypes with cumulative frequency of 0.0274

Table 3.

Effect of haplotype H2 on (A) plasma concentration of NorEND, (B) MR_NorEND/NDDM, (C) MR_{NorEND/(Z)‐END} and (D) MR_NDDM/NDM

Haplotypes and covariates	Coefficient	Standard error	t statistic	P value
(A) Plasma concentration of NorEND
Intercept	1.454	0.134	10.877	‐
Haplotype 2 (TGAGGATT)	−0.076	0.022	−3.453	0.001
Age	0.006	0.002	2.836	0.005
BMI	‐0.020	0.004	‐4.897	<0.0001
CYP2D6 hetEM	0.008	0.047	0.161	0.872
CYP2D6 IM	‐0.097	0.044	‐2.225	0.027
CYP2D6 UM	0.122	0.091	1.346	0.18
(B) MR NorEND/NDDM
Intercept	2.400	0.156	15.398	‐
Haplotype 2 (TGAGGATT)	−0.159	0.026	−6.080	<0.0001
Age	‐0.005	0.003	‐1.982	0.049
BMI	‐0.019	0.005	‐3.933	<0.0001
(C) MR NorEND/(Z)‐END
Intercept	2.325	0.172	13.550	‐
Haplotype 2 (TGAGGATT)	−0.122	0.035	−3.521	0.001
Age	‐0.007	0.003	‐2.298	0.023
CYP2D6 hetEM	0.120	0.074	1.635	0.104
CYP2D6 IM	0.558	0.069	8.090	<0.0001
CYP2D6 UM	0.198	0.142	1.390	0.166
(D) MR NDDM/NDM
Intercept	1.177	0.078	15.120	‐
Haplotype 2 (TGAGGATT)	0.049	0.016	3.110	0.002
Age	0.004	0.001	2.531	0.012
CYP2D6 hetEM	‐0.100	0.033	‐3.000	0.003
CYP2D6 IM	‐0.203	0.031	‐6.497	<0.0001
CYP2D6 UM	‐0.085	0.065	‐1.321	0.188

Reference haplotype 1 (GAGACTCG, wild type) and CYP2D6 EM were references in the model. Data in bold font pertain to H2 haplotype as well as significant P values (<0.05).

Figure 1.

Associations of CYP2C19 H2 haplotype with (A) plasma concentrations of NorEND (ng ml⁻¹), (B) MR_NorEND/NDDM, (C) MR_{NorEND/(Z)‐END} and (D) MR_NDDM/NDM (n= 191). Statistical significance was determined using the Jonckheere–Terpstra test

Strong associations were observed between haplotype H2 and the metabolic ratios, MR_{NorEND/(Z)‐END} and MR_NorEND/NDDM after adjusting for age and CYP2D6 metabolizer status (Table 3). Specifically, patients harbouring H2/H2 were found to have 34.1% and 27.0% lower MR_NorEND/NDDM compared with those harbouring H1/H1 and H1/H2, respectively [MR_NorEND/NDDM, median (range) across H2 copy number(s), 0 vs. 1 vs. 2: 2.75 (0.62–6.26) vs. 2.48 (1.19–4.18) vs. 1.81 (1.22–2.49); Figure 1B]. Similarily, H2/H2 carriers had 35.5% and 24.4% lower MR_{NorEND/(Z)‐END} compared with those carrying H1/H1 and H1/H2, respectively [MR_{NorEND/(Z)‐END}, median (range) across H2 copy number(s), 0 vs. 1 vs. 2: 9.40 (3.27–28.35) vs. 8.02 (2.67–18.90) vs. 6.06 (4.47–14.60), Figure 1C]. Significant trends of lower MR_NorEND/NDDM and MR_{NorEND/(Z)‐END} with increasing copy number of H2 haplotype were also observed with the Jonckheere–Terpstra test (P < 0.0001 for both MR_NorEND/NDDM and MR_{NorEND/(Z)‐END}).

Because NDDM has been shown to be associated with the CYP2C19 haplotype H2, we also tested the metabolic ratio with respect to its precursor NDM. As with the conversion from NDM to NDDM, an inverse relationship was observed for MR_NDDM/NDM (P = 0.002, Table 3), with patients harbouring H2/H2 having 30% and 15% higher MR_NDDM/NDM compared with those carrying H1/H1 and H1/H2 respectively [MR_NDDM/NDM, median (range) across H2 copy number, 0 vs. 1 vs. 2: 1.42 (0.76–2.91) vs. 1.60 (0.99–3.09) vs. 1.84 (1.19–2.81), Figure 1D]. The Jonckheere–Terpstra test showed a significant trend of higher MR_NDDM/NDM with increasing copy number of haplotype H2 (P < 0.0001). No significant associations were evident between CYP2C19 haplotypes with plasma (Z)‐endoxifen concentrations or MR_{(Z)‐END/NDM} (data not shown).

Comparison between haplotype H2 and CYP2C19*2 in predicting norendoxifen disposition

To determine whether haplotype H2, which comprises CYP2C19*2 and seven other linked SNPs, is more informative than CYP2C19*2 alone in predicting the plasma concentration of NorEND, MR_NorEND/NDDM, MR_{NorEND/(Z)‐END} and MR_NDDM/NDM, we compared the goodness of fit of the regression models for CYP2C19*2 itself and haplotype H2 (Table 4). For NorEND concentration, MR_{NorEND/(Z)‐END} and MR_NDDM/NDM, the CYP2C19*2 regression models gave better (smaller) Akaike Information Criterion (AIC). However, the r ² differed by less than 1% from the haplotype H2 regression models. The standard deviations (SDs) of NorEND concentration, MR_{NorEND/(Z)‐END} and MR_NDDM/NDM were 0.243, 0.431 and 0.178, respectively. The differences in mean absolute error between the two sets of analysis for these variables were tiny in comparison with the SDs of these variables. For MR_NorEND/NDDM, the haplotype H2 regression model gave better AIC and r ² than the model with CYP2C19*2 alone. The SD of MR_NorEND/NDDM was 0.307. The difference in mean absolute error between the two sets of analysis was also tiny in comparison with the SD of this variable. Overall, haplotype H2 was not more informative than the CYP2C19*2 alone in predicting the parameters with respect to NorEND.

Table 4.

Goodness‐of‐fit analyses of regression models based on haplotype H2 comprising CYP2C19*2 and seven other SNPs and CYP2C19*2 itself

Parameters	Models	AIC	r 2	Mean absolute error
(A) Plasma concentration of NorEND	Haplotype H2	‐33.882	23.4%	0.164
	CYP2C19*2	‐34.069	22.7%	0.165
(B) MR NorEND/NDDM	Haplotype H2	38.949	28.5%	0.193
	CYP2C19*2	44.209	25.8%	0.197
(C) MR NorEND/(Z)‐END	Haplotype H2	146.540	39.1%	0.256
	CYP2C19*2	145.57	39.1%	0.254
(D) MR NDDM/NDM	Haplotype H2	‐169.44	26.4%	0.120
	CYP2C19*2	‐171.32	27.1%	0.119

Discussion

This study aimed to address the knowledge gap in the role of CYP2C19 polymorphisms and their associated haplotypes in influencing the disposition of tamoxifen and its metabolites in Asian breast cancer patients, particularly NorEND, which has remained largely unknown thus far.

Prior to investigating the impact of CYP2C19 genetic variations on tamoxifen metabolism, a comprehensive genetic screening of CYP2C19 was conducted to explore the genetic diversity of CYP2C19 in healthy Asian populations, which revealed the highly polymorphic nature of CYP2C19. Thus far, 41 polymorphisms have been reported in the coding regions of CYP2C19 albeit with variable functional consequences 40, 41. The majority of these variants are ethnic specific and were not observed in this study with the exception of 99C > T (rs17885098), 518C > T (rs61311738), 636G > A (*3; rs4986893), 681G > A (*2; rs4244285), 990C > T (rs3758580), 991 A > G (rs3758581) and 1251 A > C (rs17886522).

Subsequent genotypic–phenotypic analyses did not reveal any significant associations between CYP2C19 genetic variations and the plasma concentrations as well as metabolic ratios of tamoxifen and its three main metabolites NDM, (Z)‐4‐OHT and (Z)‐endoxifen. This is in line with previous findings by Murdter et al. 15 Importantly, however, the CYP2C19*2 (681G > A; rs4244285) loss‐of‐function polymorphism was associated with significantly lower steady‐state concentrations of NorEND. Further haplotypic analysis revealed that the H2 haplotype, which consists of seven SNPs in linkage with the defective CYP2C19*2 allele, was also associated with decreased plasma NorEND concentrations as well as lower metabolic ratios, MR_NorEND/NDDM and MR_{NorEND/(Z)‐END}, thus suggesting that the extent of N‐demethylation of endoxifen and 4‐hydroxylation of NDDM to NorEND was lower in patients harbouring the H2 haplotype. However, relative goodness‐of‐fit of regression models based on the H2 haplotype and the CYP2C19*2 alone showed that using the H2 haplotype did not provide any practical significant advantage over using the CYP2C19*2 alone in predicting these parameters. It also has to be noted that such associations were not investigated for another known loss‐of‐function polymorphism CYP2C19*3, due to the small number of patients carrying the *3 allele.

Although NorEND has long been known to be a metabolite of tamoxifen, its activity has not been studied until recently 11. Unlike other well‐characterized active metabolites of tamoxifen, 4‐OHT and endoxifen, which preferentially act as an antiestrogen, a recent study by Lu et al. showed that NorEND inhibited recombinant human aromatase competitively, with a K _i of 35 nm, and these effects were shown to be comparable with that of the commonly used aromatase inhibitor letrozole 11. In addition, Liu et al. 42 also demonstrated the high selectivity of NorEND toward aromatase (CYP19) among other CYP450 enzymes, including CYP2B6, CYP2C9, CYP2C19, CYP2D6 and CYP3A 42, therefore suggesting that NorEND is a potent and selective aromatase inhibitor. NorEND has been previously shown to antagonize the activity of estrogen receptors in breast tissues, although its antagonism is reportedly weaker than those observed with (Z)‐4‐OHT and endoxifen 12. Taken together, these findings suggest that tamoxifen‐treated patients who harbour the CYP2C19*2 polymorphism and the CYP2C19 H2 haplotype may have significantly lower concentrations of NorEND.

Notably, the results of clinical studies examining the impact of CYP2C19 functional polymorphisms on tamoxifen treatment outcomes have been controversial. van Schaik et al. 23 found that patients carrying at least one CYP2C19*2 allele had significantly longer time to tamoxifen treatment failure than those carrying wild‐type alleles. However, Schroth et al. 19 reported more favourable breast cancer outcomes among patients who were carriers of the gain‐of‐function CYP2C19*17 compared with those with normal or null alleles. Contrary to these findings, Okishiro et al. 20 subsequently reported a null association between CYP2C19 polymorphisms and tamoxifen treatment outcomes. The reason for such inconsistencies may be due to the presence of yet unknown contributing factors. Murdter et al. 15 suggested that the factors affecting tamoxifen metabolism beyond the known active metabolites 4‐OHT and endoxifen may involve other metabolites for which there may be pharmacogenomic relevance. Our data suggest that one of these factors may be the variations in the plasma concentrations of NorEND. Thus, our study may aid in providing hypotheses for possible cofactors that may explain the lack of a consistent relationship between CYP2D6 genotype and tamoxifen treatment outcomes, which have so far only looked at associations with known active tamoxifen metabolites. The present study thus underscores the need to take into consideration CYP2C19 pharmacogenetics and NorEND concentrations in future studies evaluating the clinical benefits of tamoxifen.

In conclusion, the CYP2C19 loss‐of‐function polymorphism, CYP2C19*2, as well as the CYP2C19 H2 haplotype were found to be significantly associated with lower plasma concentrations of NorEND and lower formation rates of NorEND, although the H2 haplotype was not more informative than CYP2C19*2 alone. As NorEND is an active metabolite of tamoxifen that inhibits both aromatase and estrogen receptors, variability in its plasma concentration can potentially influence the therapeutic outcomes of tamoxifen therapy. These data thus suggest that CYP2C19 may potentially serve as a complementary biomarker for the identification of patients who may or may not benefit from tamoxifen treatment. The impact of CYP2C19 polymorphisms and their associated haplotypes on the treatment outcomes of tamoxifen should be further evaluated.

Competing Interests

The authors have completed the Unified Competing Interest form at www.icmje.org/coi_disclosure.pdf (available on request from the corresponding author) and declare no support from any organization for the submitted work, no financial relationships with any organizations that might have an interest in the submitted work in the previous 3 years and no other relationships or activities that could appear to have influenced the submitted work.

This study was supported by the National Medical Research Council (NMRC/1159/2008 and NMRCB1011) Singapore, Robert Bosch Stiftung, Stuttgart, Germany and in part by the Deutsche Foschungsgemeinschaft (SCHR 1323/2–1 and MU 1727/2–1) grants.

Supporting information

Figure S1 Pairwise linkage disequilibrium (LD) of CYP2C19 polymorphisms in healthy (A) Chinese, (B) Malays and (C) Indians (n = 80 in each ethnic group). LD values expressed as │D’│×100

Table S1 CYP2D6 genotypes of patients with uncertain genotype‐predicted phenotypic effects due to presence of multiple CYP2D6 alleles

Table S2 Validation data for quantification of plasma concentrations of tamoxifen and its metabolites

Table S3 Genotypic and allelic frequencies of CYP2C19 polymorphisms in Asian healthy subjects (Chinese, Malay, Indians, n = 80 each)

Table S4 In silico predictions of functional effects of CYP2C19 polymorphisms

Table S5 Impact of CYP2C19 polymorphisms on pharmacokinetics of tamoxifen and its metabolites in Asian breast cancer patients (n = 201)

Supporting info item

Supporting info item

Supporting info item

Supporting info item

Supporting info item

Supporting info item

Lim, J. S. L. , Sutiman, N. , Muerdter, T. E. , Singh, O. , Cheung, Y. B. , Ng, R. C. H. , Yap, Y. S. , Wong, N. S. , Ang, P. C. S. , Dent, R. , Schroth, W. , Schwab, M. , and Chowbay, B. (2016) Association of CYP2C19*2 and associated haplotypes with lower norendoxifen concentrations in tamoxifen‐treated Asian breast cancer patients. Br J Clin Pharmacol, 81: 1142–1152. doi: 10.1111/bcp.12886.