Characterization of Raloxifene Glucuronidation: Potential Role of UGT1A8 Genotype on Raloxifene Metabolism In Vivo

There was an estimated 226,870 subjects who developed breast cancer and 39,510 deaths arising from this disease in the United States in 2012 (1). Much of the treatment for estrogen receptor–positive (ER+) breast cancer in postmenopausal women has targeted the blocking of the ER-binding activity of estrogen or reducing estrogen synthesis. Tamoxifen, a first-generation selective estrogen receptor modulator (SERM), has been used for treatment and chemoprevention of breast cancer for over 25 years, but its long-term use is associated with rare but serious adverse effects (2). Raloxifene is a second-generation SERM approved by the U.S. Food and Drug Administration for the treatment and prevention of osteoporosis and the chemoprevention of invasive breast cancer in postmenopausal women (3). Raloxifene acts as an estrogen agonist in bone and liver to increase bone mineral density and decreases low-density lipoprotein (LDL) cholesterol (4) and exhibits strong anti-estrogen effects in breast and uterus (5). Recent clinical trials showed that raloxifene significantly reduced the incidence of breast cancer in high-risk women although not as effectively as tamoxifen (38% vs. 50%, respectively; ref. 6). However, in contrast to tamoxifen, raloxifene does not cause endometrial proliferation (7). Although not as serious as those associated with tamoxifen, adverse effects associated with raloxifene include hot flashes, vaginal dryness, and leg cramps and thromboembolic events such as deep venous thrombosis, pulmonary emboli, and retinal vein thrombosis (8).

Up to 60% of the raloxifene dose is absorbed rapidly after oral administration (8–11), but there is less than 2% bioavailability due mainly to extensive in vivo glucuronidation (8–11). Raloxifene is primarily excreted in feces, with less than 0.2% excreted as unchanged raloxifene and less than 6% eliminated as urinary glucuronide conjugates. In addition to the hepatic metabolism of raloxifene, several studies suggest that the intestine may play an important role in raloxifene metabolism (12–16). Previous studies have shown the presence of two raloxifene glucuronides in the plasma of women taking raloxifene, raloxifene-6-β-glucuronide (ral-6-Gluc) and raloxifene-4′-β-glucuronide (ral-4′-Gluc), with a plasma ral-4′-Gluc:ral-6-Gluc ratio of ∼about 8:1. Unconjugated raloxifene comprises less than 1% in human plasma (9–11, 17).

Previous studies characterizing the family 1A UDP-glucuronosyltransferase (UGT) enzymes involved in the glucuronidation of raloxifene showed that the hepatic UGTs 1A1 and 1A9 and the extrahepatic UGTs 1A8 and 1A10 were active against raloxifene (14). A recent study suggested that the UGT1A1*28 allelic variant, which contains an A(TA)₇TAA in the TATAA box of the UGT1A1 transcriptional promoter and is associated with decreased expression of the UGT1A1 gene (18), is associated with altered raloxifene pharmacokinetics (19). No studies have as yet been conducted examining the role of genotypes in other active UGTs on raloxifene glucuronidation phenotype. The goal of the present study was to fully characterize the glucuronidating activity of individual UGT1A and UGT2B enzymes against raloxifene and to compare the overall glucuronidating activity of variant active UGTs versus their wild-type counterparts both in vitro and in vivo.

Raloxifene, UDP-glucuronic acid (UDPGA), alamethicin, β-glucuronidase, β-actin, and bovine serum albumin were purchased from Sigma-Aldrich. Ral-6-Gluc, ral-4′-Gluc, raloxifene-d4, ral-6-Gluc-d4, and ral-4′-Gluc-d4 were purchased from Toronto Research Chemical. Dulbecco's modified Eagle's medium (DMEM), Dulbecco's PBS (minus calcium chloride and magnesium chloride), FBS, penicillin/streptomycin, and Geneticin (G-418) were purchased from Invitrogen. The BCA protein assay kit was purchased from Pierce Chemical.

A description of the normal human liver tissue specimens used for these studies and the methods used for liver microsomal preparation and protein quantification was provided previously (20). Normal jejunum tissues (n = 46) were purchased from Sun Health Research Institute (Sun City, AZ) and were obtained from non-cancer subjects between 2.5 and 4 hours postmortem and flash-frozen at −70°C. Jejunum homogenates were prepared by tissue homogenization in Tris-buffered saline at 4°C and stored in 100 μL aliquot (10–20 mg protein/mL) at −80°C until use. gDNA was extracted from liver and jejunum specimens using a Qiagen DNeasy Blood & Tissue extraction kit. All protocols involving the collection and analysis of tissue specimens were approved by the Institutional Review Board at Penn State University and were in accordance with assurances filed with and approved by the United States Department of Health and Human Services.

HEK293 cells were originally obtained from American Type Culture Collection and were stably transfected with individual wild-type and variant UGTs as has been previously described (21–26). To authenticate the cell lines, overexpressed UGTs were directly sequenced using intron-spanning primers, and all of the UGT-overexpressing cell lines and homogenates used in these studies were tested for glucuronidation activity by glucuronidation assays/ultra-pressure liquid chromatography (UPLC) detection against known test substrates within 3 months before a given experiment as previously described (21–24).

Plasma samples were obtained from subjects entered into a clinical trial conducted at Penn State University College of Medicine examining the combined effects of raloxifene on biomarkers of hormone-independent breast cancer (27). All subjects were postmenopausal women with a breast density in excess of 25% and without a history of thromboembolic disorders and cardiovascular disease. Blood samples were collected from subjects within the 2 raloxifene-only treatment groups (30 or 60 mg daily doses) immediately before commencement of raloxifene treatment (time “0”) and at one or more time points (6, 12, 18, and 24 months) after the commencement of treatment. All subjects provided written consent and agreed to their tissues being used for genetic studies. Bloods were fractionated by centrifugation at 1,200 × g at 4°C for 5 minutes, and plasma and lymphocyte fractions were stored in 1 mL aliquots at −80°C until analysis or gDNA extraction as described above.

Glucuronidation activity assays were conducted essentially as previously described (24) for human liver microsomes (HLM; 15 μg), human jejunum homogenates (HJH; 10 μg), or homogenates from human UGT1A and UGT2B-overexpressing cells (2–100 μg). Glucuronidation rate assays were conducted in duplicate for HLM (n = 105) and HJH (n = 46) using 1 to 2 μmol/L raloxifene. For kinetic analysis, 0.0625 to 256 μmol/L raloxifene were used for assays with UGT-overexpressing cell homogenates, 3 randomly chosen HLM, and 3 randomly chosen HJH and was conducted in triplicate in independent assays. All kinetic data for the analysis of UGT-overexpressing cell homogenates were analyzed after normalizing relative to UGT protein levels expressed in each of the overexpressing cell lines, conducted by Western blot analysis as described previously (24, 28).

Raloxifene glucuronidation was analyzed using a Waters ACQUITY ultrapressure liquid chromatography UV detector (UPLC/UV) system with a 1.7-μm ACQUITY UPLC BEH C18 analytical column (2.1 × 50 mm², Waters, Ireland) in series with a 0.2-μm Waters assay frit filter (2.1 mm). The gradient elution was conducted using a flow rate of 0.5 mL/min with 5% acetonitrile and 95% buffer A (5 mmol/L ammonium acetate, pH 5.0) for 1 minute, a subsequent linear gradient to 100% acetonitrile over 5 minutes, and then maintained at 100% acetonitrile for 2 minutes. The wavelength for determination of raloxifene and its glucuronides was 274 nm. Raloxifene-glucuronides (ral-6-Gluc and ral-4′-Gluc) were confirmed by their stability in 1 mol/L NaOH and sensitivity to the treatment of β-glucuronidase. In addition, confirmation of raloxifene glucuronide formation was conducted by loading up to 5 μL of incubation product onto an UPLC identical to that described above in tandem with a Waters TQD triple quadrupole MS system. By using a positive mode, the parent compound [M + H]⁺ peak and their corresponding glucuronide [M-Gluc. + H]⁺ peaks were characterized.

Stock solutions of raloxifene, ral-6-Gluc, ral-4′-Gluc, and their deuterated internal standards were prepared in dimethyl sulfoxide (DMSO). Raloxifene, ral-6-Gluc, and ral-4′-Gluc were combined into a standard stock solution and used to make a standard working solution from 25 ng/mL to 25 μg/mL for raloxifene and 100 ng/mL to 100 μg/mL for ral-6-Gluc and ral-4′-Gluc. Deuterated internal standards were combined with final concentrations of 5 μg/mL for raloxifene-d4, 20 μg/mL for ral-6-Gluc-d4, and 20 μg/mL for ral-4′-Gluc-d4 and were kept at −20°C before use.

Standard curves were constructed by plotting the ratio of analyte peak area to peak area of the corresponding internal standard versus analyte concentration for at least 7 analyte concentrations. The standard working solution and deuterated internal standard were spiked into plasma from untreated women and mixed (125 μL) with extraction solution (375 μL, containing 49.9:49.9:0.2 methanol:acetonitrile:formic acid) to precipitate out proteins and extract raloxifene metabolite standards. After vortexing and subsequent centrifugation at 13,000 × g for 20 minutes at 4°C, the supernatant was dried and the residue was reconstituted in 125 μL of reconstitution solution (50.0:49.9:0.1 acetonitrile:H₂O:formic acid) to make a final concentration of 0.32 to 320 ng/mL for raloxifene, 1.28 to 1,280 ng/mL for both ral-6-Gluc and ral-4′-Gluc, 60 ng/mL for raloxifene-d4, and 240 ng/mL for both ral-6-Gluc-d4 and ral-4′-Gluc-d4.

The plasma from raloxifene-treated subjects was preconcentrated 3.2-fold before loading onto the UPLC/tandem mass spectrometry (MS-MS) system. After spiking of deuterated internal standards into plasma (320 μL) from raloxifene-treated subjects, samples were mixed with extraction solution (960 μL) and processed as described above for plasma specimens from untreated subjects using 100 μL of reconstitution solution to keep analyte concentrations within the range of the calibration curve. The calculated concentrations from standard curves were divided by 3.2 to reflect the final raloxifene metabolites levels in plasma from raloxifene-treated subjects.

Calibration standards as well as the plasma sample extracts from each subject were analyzed by UPLC/MS-MS. Quantification of raloxifene, ral-6-Gluc, and ral-4′-Gluc was conducted using MRM of the transitions of m/z 474.2 → 112.2 for raloxifene, m/z 650.5 → 474.3 for ral-6-Gluc and ral-4′-Gluc, m/z 478.2 →116.2 for raloxifene-d4, and m/z 654.5 → 478.3 for ral-6-Gluc-d4 and ral-4′-Gluc-d4. The ral-6-Gluc and ral-4′-Gluc were distinguished by matching the retention time with the commercial standards in the chromatograms. The optimized MS conditions were positive ionization mode, capillary voltage 3.0 kV, cone voltage 30 V, collision voltage 30 V, source temperature 150°C, and desolvation temperature 350°C. Nitrogen was used as the desolvation and cone gas with the flow rate at 760 and 50 L/h, respectively. Argon was used as the collision gas at a flow rate of 0.1 L/h. The dwell time for each ion was 0.01 second. All data were quantified by MassLynx NT 4.1 software with QuanLynx program (Waters Corp.).

gDNA from human liver, human jejunum, and human lymphocytes was used to genotype the UGT1A1*28 allele and the UGT1A8 codons 173 (Ala > Gly; rs1042597) and 277 (Cys > Tyr; rs17863762) SNPs. For the UGT1A1 TATAA box polymorphism genotype, DNA was PCR-amplified as previously described (29) using sense and antisense primers: 5′-GAGTATGAAATTCCAGCCAGTTCAAC-3′ and 5′-TCCACTGGGATCAACAGTATCTT-3′ (corresponding to −224 to −198 and +107 to +85 relative to the UGT1A1 ATG translation start site, respectively), resulting in an amplicon of 331 bp with the TATAA box polymorphism near the middle of the amplicon. After running on a 1.0% agarose gel and extraction using a Qiagen gel extraction kit, purified PCR products were sequenced using an ABI Hitachi 3730XL DNA Analyzer, with sequencing confirmed using both the forward and reverse amplification primers described above. Sequencing results were confirmed by visual inspection of the TATAA box chromatogram peaks. The UGT1A1*28 allele was in Hardy–Weinberg equilibrium and indicated an allelic frequency of 35% in subjects from whom HLM samples were obtained and 34% in subjects taking raloxifene from whom plasma raloxifene metabolites were analyzed.

The UGT1A8 coding single-nucleotide polymorphisms (SNP) were genotyped by real-time PCR using ABI TaqMan Drug Metabolism Genotyping Assays (C__11742072_10 for rs1042597 and C__34418788_20 for rs17863762) according to manufacturer's protocols. The three UGT1A8 alleles were in Hardy–Weinberg equilibrium with allelic frequencies of 76.2% for UGT1A8*1, 20.2% for UGT1A8*2, and 3.6% for UGT1A8*3 in subjects from whom HJH samples were obtained and 71.3%, 26.4%, and 2.3%, respectively, in subjects taking raloxifene from whom plasma raloxifene metabolites were analyzed.

RNA was extracted from all 46 jejunum specimens using the Qiagen RNeasy Mini Kit according to the manufacturer's protocol. After digestion with DNAse I digestion, RNA concentrations were determined using a Nanodrop ND-1000 spectrophotometer. RNA purity was assessed by absorbance ratios A₂₆₀/A₂₈₀ (>1.9) and A₂₆₀/A₂₃₀ (>1.8). RNA integrity was determined using an Agilent 2100 Bioanalyzer with Agilent RNA 6000 Nano chips, and all 46 jejunum samples used in this study had an RIN > 4.0 with clearly visible 28S and 18S rRNA bands. Reverse transcription (RT) real-time PCR was carried out for 5 randomly selected jejunum RNA specimens as previously described (30) to assess the relative expression levels of UGTs 1A1, 1A8, 1A9, and 1A10. Real-time PCR was carried out using a 25 ng RNA equivalent of cDNA, and expression levels were normalized to the expression of the glyceraldehyde-3-phosphate dehydrogenase (GAPDH) gene. Quadruplicate real-time PCRs were carried out for each cDNA sample analyzed using a 10 μL final reaction volume according to manufacturer's protocols (assay IDs: UGT1A1, Hs02511055_s1; UGT1A8, Hs01592482_m1; UGT1A10, Hs02516990_s1; UGT1A9, Hs02516855_sH; GAPDH, Hs99999905_m1). Reactions were conducted in a 384-well plate using the ABI 7900 HT Sequence Detection System under the following conditions: 1 cycle at 50°C for 2 minutes, 1 cycle at 95°C for 10 minutes, and 40 cycles of 95°C for 15 seconds and 60°C for 1 minute. Relative quantification (RQ) of UGT1A expression was calculated using the ΔΔC_t method as previously described (30).

Competitive binding assays of raloxifene, ral-6-Gluc, and ral-4′-Gluc with the estrogen receptor (ER) were conducted essentially as previously described (31) by incubating the cytosolic fraction of MCF-7 cells (500 μg total protein) with 10⁻⁹ mol/L ³H-labeled-estradiol (E2) and between 10⁻¹¹ and 10⁻⁶ mol/L of competitor (raloxifene, ral-6-Gluc, or ral-4′-Gluc). Data were expressed as the percentage of specific binding of ³H-E2 for the ER when competitor was not present. The relative binding affinity (RBA) for each test compound was calculated as IC₅₀ which was normalized to that of E2.

The Student t test (2-sided) was used for comparing kinetic values of glucuronidation formation for UGT wild-type versus variant overexpressing cell lines and for comparing raloxifene Gluc formation rates in HLM and HJH between 2 different genotypes. The ANOVA trend test was used to examine the overall effect of UGT genotypes on raloxifene glucuronide formation in HLM and HJH, whereas the Jonckheere–Terpstra trend test was used to examine the overall effect of UGT genotypes on raloxifene glucuronide levels in human plasma. Kinetic constants were determined using the Michaelis–Menten Model in GraphPad Prism 5 software.

Previous studies have shown that HLM catalyzes the formation of two glucuronides of raloxifene, raloxifene-6-Gluc and raloxifene-4′-Gluc (14, 32). Similar to that observed in previous studies (14, 32), 2 major peaks with retention times of 2.70 and 2.97 minutes were observed by UPLC/MS-MS in in vitro raloxifene glucuronidation assays with HLM (Fig. 1A) and HJH (Fig. 1B). The retention time of the 2 peaks was the same as that of purchased ral-6-Gluc (peak 1) and ral-4′-Gluc (peak 2) standards (Fig. 1C) and were confirmed to be O-glucuronides of raloxifene by their insensitivity to alkali but sensitivity to β-glucuronidase treatment (results not shown). Ral-6-Gluc and ral-4′-Gluc were also observed in plasma samples from subjects treated with raloxifene (Fig. 1D), exhibiting the same retention times as the plasma-spiked ral-6-Gluc-d4 and ral-4′-Gluc-d4 internal standards (results not shown), the same retention times and MS/MS spectrum as the ral-6-Gluc and ral-4′-Gluc standards, and the in vitro assays with HLM and HJH described above. A third peak detected in plasma from raloxifene-treated subjects exhibited a retention time of 3.61 minutes (Fig. 1E), which was the same as that observed for the raloxifene standard (peak 3; Fig. 1F). Similar to previous studies (19, 32, 33), a ral-6,4′-diGluc was not observed in the plasma of subjects treated with raloxifene or in vitro with HLM or HJH. Using MS/MS daughter scan mode, the mass spectrum of both peaks showed a [M + H⁺] peak at m/z 650 for raloxifene-O-glucuronide, a [M + H]⁺ peak at m/z 474 for raloxifene after loss of the glucuronide acid moiety (molecular weight = 176 g/mol; Fig. 1G), and a [M + H]⁺ peak at m/z 474 for raloxifene and a major daughter fragment at m/z 112.2 (Fig. 1H).

UGTs 1A1, 1A3, 1A7, 1A8, and 1A9 catalyzed both ral-6-Gluc and ral-4′-Gluc formation, whereas UGT1A10 specifically catalyzed the formation of ral-4′-Gluc (Table 1). Representative kinetic analysis curves for ral-6-Gluc and ral-4′-Gluc formation are shown for UGTs 1A1 (Fig. 2A) and 1A8 (Fig. 2B) against raloxifene. After normalizing for UGT1A protein expression as determined by Western blot analysis (34), the order of ral-6-Gluc formation based on V_max/K_M was UGT 1A8 > 1A1 > 1A7 ≈ 1A9 > 1A3, whereas the order of ral-4′-Gluc formation based on V_max/K_M was UGT 1A10 > 1A8 > 1A9 > 1A1 > 1A7 > 1A3. In addition to UGTs 1A4 and 1A6, none of the UGT2B enzymes screened in this analysis exhibited detectable levels of raloxifene glucuronide formation.

Of the hepatic UGT enzymes active against raloxifene, UGT1A1 was the most active UGT for ral-6-Gluc formation and was the second-most active UGT for ral-4′-Gluc formation (Table 1). The UGT1A1*28 allele is a common variant (∼30% frequency in Caucasians) that encodes an A(TA)₇TAA repeat in the TATAA box of the UGT1A1 promoter region instead of the more common A(TA)₆TAA repeat encoded by the wild-type UGT1A1*1 allele, leading to lower UGT1A1 expression (18) and a decreased glucuronidation phenotype against a variety of endogenous and exogenous compounds (29, 35, 36). To investigate the possible relationship between raloxifene glucuronidation phenotype and this polymorphism, a series of 105 HLMs were examined in vitro. Using a concentration of raloxifene (2 μmol/L) that was close to the K_M of 8 μmol/L for ral-6-Gluc formation and 1.5 μmol/L for ral-4′-Gluc formation for 3 randomly chosen HLMs (data not shown), there was a 16- and 43-fold range in formation observed for ral-6-Gluc and ral-4′-Gluc, respectively, in the 105 HLM specimens. When stratifying the HLM by UGT1A1 genotype, ral-6-Gluc formation was significantly (P_trend = 0.005) decreased with increasing numbers of the UGT1A1*28 allele (Fig. 3A), with significant decreases observed in UGT1A1 (*1/*28) (26%; P = 0.004; n = 49) and (*28/*28) (39%; P = 0.01; n = 11) HLMs as compared with HLM with the UGT1A1 (*1/*1) genotype (n = 45). Significant differences in the levels of ral-4′-Gluc formation were not observed for HLM after stratification by UGT1A1 genotype (Fig. 3A). No significant differences in K_M for the formation of ral-4′-Gluc were observed in HLM with the UGT1A1 (*28/*28) or (*1/*28) genotypes versus UGT1A1 (*1/*1) HLM (3 HLM examined per genotype group; results not shown).

Of the extrahepatic UGT enzymes shown to be active against raloxifene in vitro, UGTs 1A8 and 1A7 exhibit missense SNPs with prevalences of more than 3% in the population. Two coding region SNPs resulting in Ala to Gly at codon 173 (encoded by the UGT1A8*2 allele) and Cys to Tyr at codon 277 (encoded by the UGT1A8*3 allele) are present in the UGT1A8 gene [prevalences of 0.24 and 0.036, respectively, in Caucasians according to HapMap (37)]. In an in vitro analysis of UGT1A8-overexpressing HEK293 cell homogenates, the UGT1A8^{173Gly/277Cys} variant (encoded by the UGT1A8*2 allele) exhibited a significantly lower K_M (P < 0.005) and higher overall activity as determined by V_max/K_M (P < 0.01) for ral-6-Gluc formation, as compared with wild-type UGT1A8^{173Ala/277Cys} (Table 1). While this variant also exhibited a significantly lower K_M (P < 0.005) for ral-4′-Gluc formation, a similar V_max/K_M was observed. No detectable glucuronidation activity was observed for the UGT1A8^{173Ala/277Tyr} variant (encoded by the UGT1A8*3 allele) against raloxifene in vitro. No difference in glucuronidation activity was observed against raloxifene for any of the UGT1A7 variants [UGT1A7*1, UGT1A7*2, UGT1A7*3, and UGT1A7*4 (refs. 26, 38)] in an in vitro analysis of UGT1A7-overexpressing HEK293 cell homogenates (results not shown).

Previous reports have shown that small intestine is an important contributor to raloxifene glucuronidation and clearance in vivo (13, 14, 39) and that several UGTs including UGT1A8 are well-expressed in tissues of the digestive tract (40). However, previous studies examining the expression of intestinal UGTs have been relatively nonquantitative (41). Using real-time PCR, UGT1A1 was shown to be expressed at the highest levels in jejunum (Fig. 3B). While UGT1A8 was expressed in jejunum at levels that were about 2-fold lower than UGT1A1, UGT1A8 was expressed at levels that were 10-fold higher than UGT1A10. The mRNA expression level of UGT1A9 was not quantifiable in all 5 jejunum specimens analyzed.

As UGT1A8 was among the 2 most active UGTs against raloxifene and was shown to exhibit high levels of relative expression in human jejunum, a potential role for UGT1A8 genotypes on raloxifene glucuronidation phenotype was examined using a series of 46 HJH specimens. A concentration of 0.8 μmol/L raloxifene was used for HJH glucuronidation activity assays, which approximated the K_M of 0.75 μmol/L for the major raloxifene metabolite in HJH, ral-4′-Gluc (as measured by kinetic analysis of 3 randomly chosen HJH; data not shown). There was a 64-fold range in formation observed for total raloxifene glucuronide formation in the 46 HJH specimens. When stratifying the HJH specimens by UGT1A8 codon 173 genotype, a significant (P = 0.018) 1.8-fold increase in total raloxifene glucuronide formation was observed in HJH from subjects with the UGT1A8 (*2/*2) genotype (n = 3) as compared with subjects with at least one UGT1A8*1 allele (n = 40; Fig. 3C). No difference in total raloxifene glucuronide formation was observed in HJH specimens that exhibited the UGT1A8 (*1/*3) genotype [n = 3; there were no specimens with the UGT1A8 (*3/*3) genotype]. There was a near-significant (P = 0.058) decrease in K_M for ral-4′-Gluc formation in HJH with the UGT1A8 (*2/*2) genotype (0.46 ± 0.11 μmol/L) than UGT1A8 (*1/*1) HJH (0.75 ± 0.16 μmol/L); the K_M of UGT1A8 (*1/*3) HJH (0.80 ± 0.39 μmol/L) was similar to that observed for HJH with the UGT1A8 (*1/*1) genotype.

The levels of raloxifene and its glucuronides were determined simultaneously in the plasma of subjects treated with either 30 or 60 mg/d raloxifene. Validation of the UPLC/MS-MS analytical method used for this analysis showed high assay recovery of a range of plasma raloxifene/raloxifene glucuronide levels (88%–110%). The quantification limit (signal/noise > 10) was 0.08 ng/mL for raloxifene, 0.625 ng/mL for ral-6-Gluc, and 0.78 ng/mL for ral-4′-Gluc. The intra- and interday precision (CV), respectively, for raloxifene were 9.4% and 12.3% at 0.32 ng/mL, 4.1% and 6.5% at 2.5 ng/mL, and 0.6% and 1.4% at 160 ng/mL. For ral-6-Gluc, the CV were 8.0% and 9.6% at 1.28 ng/mL, 6.2% and 7.0% at 20 ng/mL, and 1.6% and 4.1% at 640 ng/mL. For ral-4′-Gluc, the CV were 3.8% and 8.8% at 1.28 ng/mL, 2.7% and 4.0% at 20 ng/mL, and 1.3% and 2.3% at 640 ng/mL. The levels of plasma raloxifene and its metabolites showed extensive variability in bloods drawn from patients coming into the breast cancer clinic for all time points examined (6-, 12-, 18-, and 24-month blood draws following initiation of the trial). However, the range of raloxifene metabolite levels was similar for all time points. For example, the range of the plasma ral-4′-Gluc for subjects from the 30 mg daily treatment group was 2.7 to 95 ng/mL for the month 6 visit (n = 35), 3.3 to 96 ng/mL for the month 12 visit (n = 27), 5.1 to 66 ng/mL for the month 18 visit (n = 14), and 13 to 89 ng/mL for the month 24 visit (n = 3). A similar pattern was observed for ral-4′-Gluc for the 60-mg daily treatment group as well as ral-6-Gluc for both treatment groups (results not shown).

Using bloods drawn from the first available visiting time for each raloxifene-treated subject, there was extensive metabolism to ral-6-Gluc and ral-4′-Gluc, with unchanged raloxifene comprising only 0.98% and 0.88% of the total plasma raloxifene metabolite profile (raloxifene + ral-6-Gluc + ral-4′-Gluc) in subjects taking either 30 or 60 mg/d raloxifene, respectively (Table 2). The level of ral-4′-Gluc was about 3- to 4-fold higher than that of ral-6-Gluc in both groups. There were 26- and 23-fold differences in the level of plasma ral-6-Gluc, 35- and 32-fold differences in the level of plasma ral-4′-Gluc, and 4.8- and 13-fold differences in the level of plasma raloxifene between subjects from the 30 and 60 mg/d treatment groups, respectively. While the mean ratios of both plasma ral-4′-Gluc/ral-6-Gluc and total plasma ral-Gluc/raloxifene were similar between groups, the levels of ral-6-Gluc, ral-4′-Gluc, and raloxifene increased by 2.4-, 2.3-, and 2.1-fold, respectively, in subjects from the 60 mg/d group as compared with subjects from the 30 mg/d group. Similar raloxifene metabolite profiles were observed when examining bloods drawn from subjects at other time points or when using an average of all time points (results not shown).

On the basis of the cell line data obtained in this study, it was predicted that individuals with a UGT1A8*3 allele would exhibit lower raloxifene glucuronidation capacities, whereas subjects with a UGT1A8*2 allele would exhibit higher raloxifene glucuronidation capacities. When stratifying by UGT1A8 genotype, we found essentially no difference in the levels of plasma ral-6-Gluc, ral-4′-Gluc, or raloxifene in subjects who were either UGT1A8 (*1/*2) versus UGT1A8 (*1/*1) (results not shown), so they were combined into one group. To best compare the levels of plasma raloxifene and its glucuronides in the two treatment groups, the ratios of ral-6-Gluc/raloxifene, ral-4′-Gluc/raloxifene, and total raloxifene glucuronide/raloxifene were examined in subjects after stratifying by UGT1A8 genotype. As the two raloxifene glucuronides comprise more than 99% of the total plasma raloxifene metabolites in subjects talking raloxifene, other ratios including total ral-Gluc/(total ral-Gluc + raloxifene) were not informative. As shown in Table 3, increases in the ratios of ral-6-Gluc/raloxifene, ral-4′-Gluc/raloxifene, and total ral-Gluc/raloxifene were observed for plasma specimens from subjects who were UGT1A8 (*1/*3) versus subjects who were either UGT1A8 (*1/*1) or UGT1A8 (*1/*2). A similar pattern was observed when comparing plasma ral-6-Gluc/raloxifene, ral-4′-Gluc/raloxifene, and total ral-Gluc/raloxifene ratios from subjects who were either UGT1A8 (*1/*1) or UGT1A8 (*1/*2) versus subjects who were UGT1A8 (*2/*2). This trend was significant (P_trend = 0.020 for ral-6-Gluc/raloxifene, P_trend = 0.003 for ral-4′-Gluc/raloxifene, and P_trend = 0.005 for total ral-Gluc/raloxifene) when subjects from both treatment groups were combined. The levels of dose-adjusted plasma ral-6-Gluc, ral-4′-Gluc, and total ral-Gluc in the combined group were significantly (P_trend = 0.0025, 0.001, and 0.001, respectively) increased in predicted slow metabolizers [UGT1A8 (*1/*3)] versus intermediate metabolizers [UGT1A8 (*1/*1) or UGT1A8 (*1/*2)] versus fast metabolizers [UGT1A8 (*2/*2); Fig. 3D]. No difference in the levels of ral-6-Gluc or ral-4′-Gluc were observed in plasma from subjects with either the UGT1A1 (*1/*1) (n = 34), UGT1A1 (*1/*28) (n = 39), or UGT1A1 (*28/*28) (n = 9) genotypes (results not shown). Unfortunately, there was insufficient power to examine combined UGT1A1/UGT1A8 genotypes versus plasma raloxifene metabolites in this study.

The mechanism of raloxifene action for the prevention of breast cancer is to compete with estrogen for binding to the ER to prevent the stimulation of proliferation of breast cancer cells. To examine the relative binding affinity to the ER of the two raloxifene glucuronides versus raloxifene, cytosolic fractions of MCF-7 cells were used as an ER source as previously described (31). The IC₅₀ for raloxifene, ral-6-Gluc, and ral-4′-Gluc was (4.0 ± 3.5) × 10⁻¹⁰ mol/L, (2.9 ± 0.8) × 10⁻⁷ mol/L, and (3.7 ± 1.9) × 10⁻⁸ mol/L, respectively.

While raloxifene does not undergo significant P450-dependent oxidation (10), it is extensively glucuronidated by first-pass metabolism. Similar to that observed in previous studies (14), several UGT1A enzymes were found to exhibit raloxifene glucuronidating activity in the current study, with UGTs 1A1 and 1A9 the most active hepatic UGTs, and the extrahepatic UGTs 1A8 and 1A10 exhibiting the highest levels of activity of any UGT screened in this study. UGT1A8 exhibited the highest overall activity for ral-6-Gluc formation and the second highest activity for ral-4′-Gluc formation, and UGT1A10 exhibited the lowest K_M and highest overall activity for ral-4′-Gluc formation. The K_M values reported in the current study are 25- and 19-fold lower for ral-6-Gluc and ral-4′-Gluc formation by UGT1A8 and 23-fold lower for ral-4′-Gluc formation by UGT1A10, compared with previous studies (14), discrepancies that are likely due to differences in assay conditions. The K_M values for UGT1A1 against raloxifene were about 10 μmol/L for both ral-6-Gluc and ral-4′-Gluc formation in our study, whereas the K_M for UGT1A1 could not be determined in previous studies due to solubility limitations as indicated in that study (14). UGTs 1A3 and 1A7 were also shown to be active in the present study. While this previous study did not test the activity of UGT1A3, no glucuronidation activity was previously observed for UGT1A7. This is likely due to the fact that UGT-overexpressing baculosomes, which have been found to exhibit significant differences in substrate specificities as compared with UGT-overexpressing human cell homogenates (22), were used in this previous study. While the present study is the first to examine UGT2B enzyme activities against raloxifene, none were found to be active.

Previous studies have shown that the level of UGT1A1 in human liver is 2-fold higher than that of UGT1A9 (30). There was only a ∼2.5-fold difference in the K_M values for ral-6-Gluc and ral-4′-Gluc formation between enzymes, suggesting that both UGTs 1A1 and 1A9 may be important in the hepatic glucuronidation of raloxifene. However, while UGT1A1 exhibits 2- and 10-fold higher levels of expression in jejunum than UGTs 1A8 and 1A10, respectively, it exhibits a ∼29-fold higher K_M and a ∼10-fold lower V_max/K_M for ral-6-Gluc formation as compared with UGT1A8, and a ∼5- and 57-fold higher K_M and a ∼8- and 80-fold lower V_max/K_M for ral-4′-Gluc formation activity as compared with UGTs 1A8 and 1A10, respectively, suggesting a lesser role for UGT1A1 in jejunum raloxifene glucuronidation activity. The barely detectable level of expression for UGT1A9 in the small intestine in the present study is consistent with that observed previously (42) and suggests a minimal role for UGT1A9 in raloxifene glucuronidation in this tissue. Given the very low activity of UGTs 1A3 and 1A7 against raloxifene, it is likely that these UGTs play only a marginal role in raloxifene glucuronidation in either liver or jejunum.

Previous studies have shown that the UGT1A1*28 allele is associated with altered glucuronidation activity against a variety of endogenous and exogenous substrates (29, 35, 36). The association observed between UGT1A1 genotype and HLM raloxifene glucuronidation in the present study is consistent with the likely importance of UGT1A1 in overall hepatic raloxifene glucuronidation activity. The fact that this was observed specifically for the formation of ral-6-Gluc is consistent with UGT1A1 cell homogenates exhibiting the highest overall activity of any hepatic UGT for this metabolite. The fact that this pattern was not observed for hepatic ral-4′-Gluc formation is consistent with UGT1A9 playing a more important role in the formation of this metabolite given the higher V_max/K_M exhibited by UGT1A9 versus UGT1A1 in vitro. The fact that no differences in raloxifene glucuronide K_M values were observed in HLM stratified by UGT1A1 genotypes is consistent with UGTs 1A9 and 1A1 exhibiting similar K_M values for both raloxifene metabolites in vitro. The fact that no association was observed between HJH raloxifene glucuronidation activities and UGT1A1 genotype is consistent with UGT1A1 playing a more minor role in jejunum raloxifene glucuronidation. As high-prevalence coding SNPs are not observed for UGT1A9 (43), a similar hepatic phenotype–genotype study was not conducted for this enzyme.

The vast majority of circulating raloxifene in the plasma of subjects treated with raloxifene was in the form of a glucuronide conjugate, with unchanged raloxifene comprising approximately 1% of total plasma raloxifene in subjects treated with either 30 or 60 mg raloxifene/d. Considerable variation in raloxifene glucuronide levels were observed in plasma between individuals taking raloxifene. Previous studies focusing on the role of the UGT1A1*28 allele on raloxifene glucuronidation gave conflicting results (19, 44, 45). The fact that UGT1A1 genotype did not contribute to variation in plasma raloxifene glucuronide levels in vivo in the present study suggests that, despite it contributing to ral-6-Gluc formation variation in HLM, potential effects by the UGT1A1*28 allele were overcome in vivo by glucuronidation contributions of other UGTs in both liver (UGT1A9) and jejunum (UGTs 1A8 and 1A10).

In previous studies, the polymorphic variants of UGT1A8 at codons 173 and 277 have been associated with altered glucuronidation activity (46) and cancer risk (47). A functional effect by these variants was also observed in the present study. UGT1A8 genotype was significantly correlated with raloxifene glucuronide formation in HJH in vitro and plasma raloxifene glucuronide levels in subjects treated with raloxifene, with the plasma levels of both ral-6-Gluc and ral-4′-Gluc significantly lower in subjects with the predicted UGT1A8 slow metabolizer genotype as compared with subjects with intermediate or fast metabolizer genotypes. These data were also consistent with the results from cell lines overexpressing UGT1A8 variants showing functional effects by UGT1A8 variants on raloxifene glucuronidation capacity. The fact that no association was observed between HJH raloxifene glucuronidation and the UGT1A8*3 allele may have been due to low power due to low UGT1A8*3 prevalence (allelic frequency = 2.2%) and the fact that all of the HJH specimens with an UGT1A8*3 allele examined in this study were from subjects who were heterozygous (*1/*3) for that allele. As high-prevalence coding SNPs are not observed for UGTs 1A9 (43) or 1A10 (48), a similar phenotype–genotype study in jejunum was not conducted for these enzymes. Therefore, in addition to suggesting that UGT1A8 coding SNPs may play an important role in the glucuronidation of raloxifene in vivo, these data support an important role for jejunum in overall raloxifene metabolism.

Ral-4′-Gluc was the major form of raloxifene metabolite observed in the plasma of subjects treated with raloxifene. The ER-binding affinity of ral-4′-Gluc was shown to exhibit an IC₅₀ value that was approximately 100-fold less than raloxifene itself. However, as more than 70% of total circulating plasma raloxifene is in the form of ral-4′-Gluc and because the levels of ral-4′-Gluc are more than 70-fold higher than parent unconjugated raloxifene, variations in the level of ral-4′-Gluc could potentially have an important effect on overall patient response to raloxifene. Therefore, UGT1A8 genotype could potentially impact overall patient response to raloxifene by altering circulating levels of raloxifene glucuronides, particularly the levels of ral-4′-Gluc. Clinical studies with a larger population size will be required to further examine the role of UGT1A8 genotype on the preventive effect of raloxifene for breast cancer.

No potential conflicts of interest were disclosed.

Conception and design: P. Lazarus

Development of methodology: D. Sun, N.R. Jones, P. Lazarus

Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): N.R. Jones, A. Manni, P. Lazarus

Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): D. Sun, N.R. Jones, P. Lazarus

Writing, review, and/or revision of the manuscript: D. Sun, N.R. Jones, A. Manni, P. Lazarus

Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): D. Sun

Study supervision: A. Manni, P. Lazarus

The authors thank the Molecular Genetics Core facility at Penn State University College of Medicine for conducting genotyping and DNA sequencing and Dr. Gang Chen for helpful discussions.

These studies were supported by a Penn State Center for Pharmacogenetics award to D. Sun, a Public Health Service (PHS) R01-CA164366 (National Cancer Institute) from the NIH, Department of Health and Human Services to P. Lazarus, and a Promise Grant KG081632 from Susan G. Komen for the Cure to A. Manni.

The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.