|Home | About | Journals | Submit | Contact Us | Français|
We aim to identify genetic variation, in addition to the UGT1A1*28 polymorphism, that can explain the variability in irinotecan (CPT-11) pharmacokinetics and neutropenia in cancer patients.
Pharmacokinetic, genetic, and clinical data were obtained from 85 advanced cancer patients treated with single-agent CPT-11 every 3 weeks at doses of 300 mg/m2 (n = 20) and 350 mg/m2 (n = 65). Forty-two common variants were genotyped in 12 candidate genes of the CPT-11 pathway using several methodologies. Univariate and multivariate models of absolute neutrophil count (ANC) nadir and pharmacokinetic parameters were evaluated.
Almost 50% of the variation in ANC nadir is explained by UGT1A1*93, ABCC1 IVS11 –48C>T, SLCO1B1*1b, ANC baseline levels, sex, and race (P < .0001). More than 40% of the variation in CPT-11 area under the curve (AUC) is explained by ABCC2 –24C>T, SLCO1B1*5, HNF1A 79A>C, age, and CPT-11 dose (P < .0001). Almost 30% of the variability in SN-38 (the active metabolite of CPT-11) AUC is explained by ABCC1 1684T>C, ABCB1 IVS9 –44A>G, and UGT1A1*93 (P = .004). Other models explained 17%, 23%, and 27% of the variation in APC (a metabolite of CPT-11), SN-38 glucuronide (SN-38G), and SN-38G/SN-38 AUCs, respectively. When tested in univariate models, pretreatment total bilirubin was able to modify the existing associations between genotypes and phenotypes.
On the basis of this exploratory analysis, common polymorphisms in genes encoding for ABC and SLC transporters may have a significant impact on the pharmacokinetics and pharmacodynamics of CPT-11. Confirmatory studies are required.
Irinotecan (CPT-11) is a topoisomerase I inhibitor approved worldwide for the treatment of metastatic colorectal cancer. A genetic polymorphism in UGT1A1 increases the risk of CPT-11 toxicity, particularly when administered as a single agent.1,2 The pathophysiology of this susceptibility resides in the decreased glucuronidation of SN-38, the active metabolite of CPT-11.3
Several groups, including ours, have established that white patients who are homozygous for the UGT1A1*28 allele are at the highest risk of developing severe toxicity of CPT-11, whereas heterozygous patients seem at intermediate risk (for a review, see Kim and Innocenti1). The UGT1A1*28 allele seems to confer reduced gene expression compared with the reference UGT1A1*1 allele,4,5 leading to increased exposure of patients to the cytotoxic metabolite SN-38.6 The accumulated evidence prompted the US Food and Drug Administration and the pharmaceutical sponsor to revise the CPT-11 label in June 2005. The label now includes homozygosity for the UGT1A1*28 genotype as one of the risk factors for severe neutropenia.7,8 A US Food and Drug Administration–approved UGT1A1*28 genotyping method is also commercially available.9
Over the last two decades, there have been many studies of the pharmacology of CPT-11 identifying a series of genes to be investigated for their possible contribution to the variability in CPT-11 pharmacokinetics and adverse effects, as shown in the pathway of CPT-11 disposition in the liver (Fig 1). In addition to the conversion to SN-38 by carboxylesterases, CPT-11 undergoes oxidation to the metabolites APC and NPC by the CYP3A4/5 enzymes. Moreover, UGT1A7 and UGT1A9 are also involved in the inactivation of SN-38. Whereas UGT1A9 is highly expressed in human liver, UGT1A7 is expressed in extrapatic tissues and is potentially relevant to the enterohepatic circulation of SN-38.10,11 Membrane transporters are responsible for the uptake of SN-38 from plasma into the hepatocytes (ie, SLCO1B1) and the elimination of CPT-11 and its metabolites into the bile (ie, ABCB1, ABCC2, and ABCG2). The ABCC1 transporter is responsible for the efflux of SN-38 from the hepatocyte into the interstitial space (Fig 1).
Despite the fact that the UGT1A1*28 polymorphism represents one of the few validated and widely accepted pharmacogenetic markers to predict drug toxicity in oncology, the performance of this test is far from optimal.7,8 Drug toxicity and disposition in patients are complex phenotypes. Their variability is a result of environmental and polygenic factors. It is likely that, in addition to UGT1A1*28, other genetic variants contribute to the toxicity phenotype. For example, it has been suggested that variation in genes encoding for SLCO1B1, ABC transporters, and other UGT1A enzymes is associated with the pharmacokinetics and pharmacodynamics of CPT-11.12–17 In the present study, we provide the first assessment of the association between genetic variants in the vast majority of the known genes in the CPT-11 pathway and the pharmacokinetic parameters of CPT-11, SN-38, SN-38 glucuronide (SN-38G), and APC in patients previously treated with single-agent CPT-11.6,18 We also investigated the polygenic basis of neutropenia, a common adverse effect of CPT-11. To our knowledge, a comprehensive investigation like this one has never been conducted, and this study proposes new gene variants as candidates with predictive significance.
Human investigations were performed after review and approval by the Biological Sciences Division/University of Chicago Hospitals Institutional Review Board and in accordance with Federalwide Assurance for the protection of human subjects. Patient characteristics are listed in Appendix Table A1 (online only).
We analyzed associations of gene variants with phenotypes using linear regression. The phenotypes of interest included neutropenia (expressed as absolute neutrophil count [ANC] nadir) and the area under the concentration-time curves (AUC) of CPT-11, SN-38, SN-38G, APC, and the SN-38G/SN-38 AUC ratio. These data were log10 transformed. Although we report the frequency data of genotypes in all races, we performed the regression analyses in patients of white and African American race only because there were too few patients in the other racial categories (8% of the total population of patients).
For the univariate analysis, we first carried out simple regressions on each variant separately. We kept for further analysis those variants with a significance level less than .15. Forward selection of this subset of phenotypes led to the final multivariable regression models, in which we required a P < .05 (two sided) to call an association nominally significant. We adjusted all analyses (univariate and multivariate) for sex, race (white or African American), age (as a continuous variable), and CPT-11 dose (either 300 or 350 mg/m2) to avoid potential problems with population stratification. Body-surface area (BSA) was not included to calculate actual dose (based on dose · BSA) because it generally did not perform better than using dose as either 300 or 350 mg/m2 in each model (data not shown). In addition, the logANC nadir values were adjusted for baseline logANC.
In the univariate analyses, we considered that pretreatment bilirubin levels might reflect underlying genetic variation that is relevant to glucuronidation. If variation in pretreatment bilirubin corresponds to genetic variability, pretreatment bilirubin may serve as a surrogate biomarker for the genotypes of one or several single nucleotide polymorphisms (SNPs). We examined this possibility by fitting the univariate models with and without pretreatment bilirubin and comparing the estimated SNP effects in both cases. For example, suppose that SNP1 is already in the model and its association with the phenotype arises because it is in linkage disequilibrium (LD) with another SNP (SNP2), which is not in the model. If both SNPs are in a gene that alters total bilirubin, then total bilirubin may explain a greater percentage of the phenotype variation in our study population. If it does explain a greater percentage of the phenotype's variation, then the regression estimate of SNP1 will get smaller (ie, go to zero) if bilirubin is in the model. We describe situations where this phenomenon occurred. Additional statistical evaluations are described in the Appendix.
Table 3 summarizes the univariate regression analysis for each parameter. UGT1A1*28 and *93 were associated with decreased ANC nadir and SN-38G/SN-38 AUC, as well as increased SN-38 AUC. UGT1A1*60 was also associated with decreased SN-38G/SN-38 AUC, as well as decreased ANC nadirs. UGT1A7 387G>T was associated with increased APC AUC (similar to UGT1A9*1b), and UGT1A7*4 was associated with decreased SN-38G/SN-38 AUC.
Concerning the CPT-11 transporters, ABCC1 1684T>C and IVS18-30C>G were associated with variation in APC AUC and SN-38G/SN-38 AUC, respectively, and ABCC1 1684T>C was associated with increased SN-38 AUC. ABCC1 IVS11 −48C>T was associated with decreased ANC nadir. SLCO1B1*5 was associated with increased CPT-11 AUC, and SLCO1B1*1b was associated with increased ANC nadir. Several ABCC2 variants were associated with variability in CPT-11, SN-38, SN-38G, and APC AUC, without any notable effect on ANC nadir. Two ABCB1 variants were associated with variability in SN-38 AUC (Table 3).
When pretreatment bilirubin was included in the ANC nadir model, the estimates of the UGT1A1 *28/*28 and *93/*93 genotypes retained significance but were reduced by approximately 30% compared with the unadjusted values. For CPT-11 AUC, bilirubin was a significant covariate (P = .02 to .05) for the four independent variables but had a minimal effect on the estimate (< 8%).
For SN-38 AUC, pretreatment bilirubin was a significant covariate (P < .0001) for each independent variable, the most notable change being on UGT1A1 *93/*93 and *28/*28, where the regression estimates became smaller (−54% and −49% reduction, respectively) in the presence of bilirubin. The smaller effect was also evident in the nonsignificant P value for these polymorphisms with bilirubin in the model, even though the SEs did not change; for UGT1A1 *93/*93, the change in P value went from .002 (unadjusted) to .14 (adjusted), and for UGT1A1 *28/*28, the P value increased from .002 to .09 after adding bilirubin to the model.
For SN-38G AUC, bilirubin was a marginally significant covariate (P = .04), with no notable effect on the estimate (2.9%). For APC AUC, including bilirubin did not significantly change the estimated effect of each polymorphism (≤ 2.5%). It was never a significant covariate (P ≥ .5).
For SN-38G/SN-38 AUC ratio, bilirubin was a significant covariate in the models with ABCC1 (1684T>C and IVS18-30G>C) and UGT1A7*4. Including bilirubin in the model led to the largest change for the effect of UGT1A1*60 (−17%). It had the smallest change on the effect of ABCC1 1684 T>C (3%).
The model for ANC nadir explains approximately 50% of the variation in neutropenia. ANC nadirs are negatively affected by the UGT1A1*93 genotype according to an additive model in which the trend is for lower nadirs as one goes from *1/*1 to *93/*93 genotypes. The C allele of ABCC1 IVS11 −48C>T appears dominant, with the TT genotype having lower nadirs. A smaller effect (compared with UGT1A1 and ABCC1) is observed for the SLCO1B1*1b variant (Fig 2). Significant adjusting covariates in this model are sex (females associated with significantly reduced ANC nadir, P = .009; 4.9% of the total percent variation), dose (P = .015, 2.5%), and baseline logANC (P = .005, 7%).
The CPT-11 AUC model explains approximately 40% of the variation in CPT-11 exposure. ABCC2 −24C>T and SLCO1B1*5 increase the exposure of patients to CPT-11 according to a dominant model; in addition, HNF1A 79A>C has an additive effect. Significant adjusting covariates in this model are age (P = .002) and CPT-11 dose (P < .0001).
The SN-38 AUC model explains close to 30% of the variation in SN-38 exposure. ABCC1 1684T>C has a dominant effect, increasing the exposure of patients to SN-38, whereas ABCB1 IVS9 −44A>G has a similar dominant effect but decreases the exposure of patients to SN-38. UGT1A1*93 is associated with increased SN-38 AUC, according to an additive model. No significant adjusting covariates were found.
The SN-38G AUC model explains just 17% of the variation in SN-38G exposure. ABCC2 3972C>T is associated with increased exposure to SN-38G. The only significant adjusting covariate in this model is sex (females associated with higher SN-38G AUC, P = .05).
The APC AUC model explains 23% of the total variation in APC exposure. The only significant genotype in this model is ABCC2 3972C>T, which increases APC exposure. The only significant adjusting covariate in this model is CPT-11 dose (P = .012).
The SN-38G/SN-38 AUC model explains 27% of the total variation in this parameter. UGT1A1*28 (as an additive model) and ABCC1 1684T>C (as a dominant model) are associated with reduced ratios. No covariates are significant in this model.
This study provides the most comprehensive survey of genetic variation of known drug-metabolizing enzyme and transporter genes of CPT-11 and has identified potential new genetic determinants of its neutropenic effects. The availability of detailed pharmacokinetics has been pivotal to interpreting the possible pharmacologic basis of the observed associations and selecting gene variants of potential interest, as well as to excluding others from further evaluation.
Our multivariate genetic model of neutropenia can explain almost half of the observed variation. Three SNPs explain 28% of ANC nadir variation, which is a significant contribution to the overall model, considering that five other nongenetic covariates contribute to the remaining 19% of the variation (Table 4). Among all the variables in the model, UGT1A1*93 seems to have the strongest effect. We first discovered this variant during a resequencing study of the region 5′ to the UGT1A exon 119 and hypothesized that UGT1A1*93 was a better predictor of neutropenia than UGT1A1*28.6 This variant has unknown function at the molecular level and, in our study, is associated with increased exposure of patients to SN-38. Despite its high LD with the UGT1A1*28 allele, our data are consistent with the recent results of a study identifying UGT1A1*93 as the only predictor of severe hematologic toxicity in colorectal cancer patients receiving fluorouracil, leucovorin, and CPT-11.20
In addition to UGT1A1, our data suggest that ABCC transporter genes play a major role in the pharmacology of CPT-11. Different variants of ABCC1 (IVS11 −48C>T and 1684T>C) are associated with ANC nadir, SN-38 AUC, and SN-38G/SN-38 AUC (Table 4). Two ABCC1 intronic variants that were not genotyped in the present study were previously shown to have no effect on the pharmacokinetics of CPT-11.16 The clinical relevance of ABCC1 variation is relatively unexplored compared with that of ABCB1, and functional studies would be important to validate our preliminary findings. In an analysis of resequencing data from 24 SLC and ABC membrane transporters, ABCC1 was found to have one of the lowest values for heterozygosity at nonsynonymous sites, indicating very little genetic variation that results in changes in the MRP1 amino acid sequence.21 Several rare variants of ABCC1 do affect transport function, but their low allele frequency precludes a major role in human drug disposition.22–25 Similarly, there is no evidence that ABCC1 polymorphisms are associated with mRNA levels in lymphocytes or duodenal enterocytes.26,27 The lack of association of the synonymous 1684T>C variant with lymphocyte ABCC1 mRNA levels provides no support for our observed association of this polymorphism with SN-38 AUC and the SN-38G/SN-38 AUC ratio. The positive association between ABCC1 1684T>C and SN-38 AUC is consistent with increased transport of SN-38 from the hepatocyte into the plasma. The functional basis for this association requires further investigation.
Recent data point toward ABCC2 variation as a determinant of CPT-11 diarrhea.12,15 The 3972C>T variant (which we typed in this study) and a haplotype of the six variants typed in this study were associated with severe diarrhea in both Korean and European patients.12,15 In the present study, we could not ascertain the effect of ABCC2 variation on severe diarrhea because of the low incidence of this event on this schedule. However, we observed associations between –24C>T and higher CPT-11 AUC and between 3972C>T and higher SN-38G and APC AUCs. The –24C>T variant is in high LD with 3972C>T,28 has been found to reduce promoter activity in an in vitro assay, and is associated with lower ABCC2 mRNA levels in human kidney.28,29 The observed increases in exposure of patients to CPT-11 and its metabolites might be a result of impaired excretion as a result of a genetically determined reduction of ABCC2 expression.
The variability in SN-38 AUC was a strong determinant of neutropenia,6,18 but only one third of its variability can be accounted for by the covariates we analyzed, although genetic variants predict the majority of such variability (Table 4). In addition to UGT1A1*93 and ABCC1 1684T>C, ABCB1 IVS9 −44A>G is associated with reduced SN-38 AUC. There is no evidence of function for this intronic variant, but it is possible that it is in LD with a causative SNP. ABCB1 1236C>T has been previously associated with increased SN-38 AUC,16 but this was not replicated in the current study. There is also no indication that variation in the CES gene controlling SN-38 formation is a major determinant of SN-38 exposure. Clearly, more research is required to discover genetic and environmental determinants of exposure of patients to SN-38.
In addition to the ABC transporters, our study highlights the importance of SLCO1B1 for the exposure of patients to parent drug and its neutropenic effects. The SLCO1B1*5 allele increases the exposure of patients to CPT-11, similar to studies with statins, where the deficient *5 allele confers impaired drug uptake by the liver after dosing.30,31 Although increased exposure to statins clearly results in increased toxicity risk of myopathy,32 we did not observe similar increased neutropenia in *5 carriers because CPT-11 AUC is not associated with increased toxicity in our study. However, the SLCO1B1*5 allele was associated with increased risk of severe neutropenia and increased SN-38 AUC (and no effect on CPT-11 AUC) after administration of CPT-11 in Asian patients.12 Our results also show that the SLCO1B1*1b allele seems to be protective of neutropenia. This result is difficult to interpret because the *1b allele showed no neutropenic effect in Asians patients.12 A different SLCO1B1 haplotype structure has been recently reported in Japanese compared with Europeans,33 and caution should be used when extrapolating interpretation from studies conducted in different ethnicities. More importantly, our findings on SLCO1B1, similar to all the other associations, should be taken with caution. Because of the retrospective and exploratory nature of this analysis, they require independent confirmation.
In this study, we also have clarified the significance of nongenetic factors. Our data suggest that females tend to be more prone to neutropenia, a finding we observed in our previous publication when only UGT1A1 variants were surveyed6 and in another recent preliminary report.34 This effect does not seem to be mediated by differences in BSA because BSA did not affect the pharmacokinetics of CPT-11 in our and other studies.35–37 Finally, the effect of pretreatment total bilirubin on the association for each gene variant suggests that total bilirubin not only carries information on the genes we surveyed (ie, UGT1A1, ABCC2, SLCO1B1) but might also be a proxy for other hitherto unknown genes and/or for the overall excretory capability of the liver. Total bilirubin remains to be validated as a marker to predict the risk of severe neutropenia in CPT-11 patients in the absence of UGT1A1 genetic information, although some studies are suggestive of its association with severe neutropenia.6,38–40
The development of pharmacogenetic algorithms to predict the risk/benefit ratio of patients will become increasingly available to practicing physicians; warfarin algorithms are currently applied in the clinic.41 In the present study, we provide information on genetic variation that is likely to affect the neutropenic effects of CPT-11. If confirmed in independent studies, these findings may have direct clinical application because they could help establish methods for assessing the toxicity risk and help physicians individualize therapy for colorectal cancer patients. A composite pharmacogenetic test is likely to be more predictive than the UGT1A1*28 test alone.
We acknowledge the contributions of R. Michael Baldwin in sample genotyping.
CES2-363C>G, IVS1 1361A>G, and 108C>G were genotyped by polymerase chain reaction (PCR) and single base extension (SBE) and denaturing high-performance liquid chromatography (DHPLC). The single nucleotide polymorphisms (SNPs) 5′UTR -363 and 3′UTR +108 were genotyped by a duplex PCR and a duplex SBE reaction. Then, the SBE products were mixed with intron 1 +1361 SBE products and run together in a triplex format on DHPLC. The genotyping primers are as follows: for –363, 5′-CTT CgC ATT TCT CCA TCT GG-3′ (forward PCR primer), 5′-GGC TGT GCC ATT CCT GCA-3′ (reverse PCR primer), 5′-GCC CGA TGA GCa CGC TGa GG-3′ (downstream extension primer); for 108, 5′-CGA CCA GGA GGA GCA ATA CC-3′ (forward PCR primer), 5′-GTC TCA AAC TCC TGT CCT CA-3′ (reverse PCR primer), 5′-ACG GAA GTA TGA ATG AAT GGC GAA-3′ (downstream extension primer); and for 1361, 5′-CAC TAC TGC ACT CCA GCC T-3′ (forward PCR primer), 5′-GAT GGC ACA CAT CAC CTG TA-3′ (reverse PCR primer), 5′-ATC CTC CCG CCA CAG tCT CC-3′ (downstream extension primer). The modified bases are indicated in lowercase. The duplex PCRs were set up in a 15-μL volume containing 140 nM of each 5′UTR –363 primer and 125 nM of each 3′UTR +108 primer (the ratio of the two sets of PCR primers was adjusted to 1.12:1 to achieve even amplification of the two PCR fragments 187 base pairs [bp] and 672 bp), 30 ng of genomic DNA, 1.5 mM of MgCl2, 100 μm of each deoxynucleotide triphosphate (dNTP), and 0.375 U of AmpliTaq Gold polymerase (Applied Biosystems, Foster City, CA) in the buffer provided by the manufacturer. Amplification was performed in a GeneAmp PCR System 9600 thermal cycler (Applied Biosystems) with an initial denaturation step of 15 minutes at 95°C followed by 40 cycles of 95°C for 15 seconds, 60°C for 15 seconds, and 72°C for 45 seconds, and then an extension step of 72°C for 10 minutes. The PCR fragment for intron 1 +1361 is 789 bp, and the PCR was performed with touch down cycling conditions. The PCR reactions were set up in a 15-μL volume containing 125 nM of each primer, 1.5 mM of MgCl2, 100 μm of each dNTP, 0.375 unit of AmpliTaq Gold, and 30 ng of genomic DNA. Amplification was performed in a 9600 thermal cycler with an initial denaturing step at 95°C for 15 minutes followed by 14 cycles of 95°C for 30 seconds, 71°C for 30 seconds (touch down, 0.5°C per cycle), and 72°C for 1 minutes; then 25 cycles of 95°C for 30 seconds, 64°C for 30 seconds, and 72°C for 1 minute; and a final extension step at 72°C for 10 minutes. Before SBE reaction, PCR products were purified by treatment with shrimp alkaline phosphatase (Roche, Indianapolis, IN) and exonuclease I (USB Corporation, Cleveland, OH) at 37°C for 45 minutes. SBE reactions were carried out in a 10-μL volume containing 1 μM of SBE primer, 250 μM each of four dideoxynucleotide triphosphates (ddNTPs), 1.25 U of ThermoSequenase (GE Healthcare, Piscataway, NJ), and 6 μL of purified PCR products. For the duplex SBE reaction, 1 μL of each extension primer was added to the total volume. Reactions were run in a 9600 thermal cycler under the following conditions: 96°C for 2 minutes, followed by 60 cycles of 96°C for 30 seconds, 55°C for 30 seconds, and 60°C for 30 seconds. Samples were denatured at 96°C for 4 minutes and held at 4°C before separation of the SBE products on a WAVE 3500HT DHPLC system (Transgenomic Inc, Omaha, NE). Mixed SBE products (16 μL per sample) were injected onto the DHPLC for analysis. Samples were run on a high throughput (HT) column (Transgenomic Inc) at 70°C oven temperature using a start gradient of 23% B for 2.5 minutes (slope at 5% B per minute). The extended products were eluted in the order of C<G<T<A dependent on the hydrophobicity differences of the four bases. Three extension primers for a triplex format on DHPLC were designed to be eluted at different times and to be able to separate each extended product. The known genotype controls were included in the run of the patient samples.
The CYP3A4*1B and CYP3A5*3 variants were typed according to Blanco et al (Blanco JG, Edick MJ, Hancock ML, et al. Pharmacogenetics 12:605-611, 2002). HNF1A 79A>C genotyping was performed by SBE with DHLPC as previously described (Ramírez J, Mirkov S, Zhang W, et al. Pharmacogenomics J 8:152-161, 2008). The UGT1A1*28 allele was previously genotyped by both Iyer et al (Iyer L, Das S, Janisch L, et al. Pharmacogenomics J 2:43-47, 2002) and Innocenti et al (Innocenti F, Undevia SD, Iyer L, et al. J Clin Oncol 22:1382-1388, 2004). The UGT1A1*6, *27, *60, and *93 variants were previously genotyped only in patients treated at 350 mg/m2 (Innocenti F, Undevia SD, Iyer L, et al. J Clin Oncol 22:1382-1388, 2004) and, in the present study, have been genotyped in patients treated at 300 mg/m2 (Iyer L, Das S, Janisch L, et al. Pharmacogenomics J 2:43-47, 2002), following the procedure described in Innocenti et al (2004). The UGT1A9 variants were genotyped as previously described (Ramirez J, Liu W, Mirkov S, et al, Drug Metab Dispos 35:2149-2153, 2007).
The UGT1A7 387T>G and 391C>A variants were genotyped by sequencing. SNPs covered were 387T>G, 391C>A, 392G>A, and 622T>C. PCR amplification and sequencing were performed using one set of primers (UGT1A7* forward: 5′-TTGCCTATGCTCGCTGGAC-3′; UGT1A7* reverse: 5′-TTTCAGGGGCTATTTCTAAGA-3′) to generate a 423-bp fragment covering all SNPs for both variants. PCR reactions were set up in a 25-μL volume containing 2.5 mM of MgCl2, 200 μM each of dNTP, 500 nM of forward and reverse primers, 0.5 U of AmpliTaq Gold (Applied Biosystems), and 75 ng of DNA. The PCR reaction was cycled for 38 cycles at 94°C for 45 seconds, 59°C for 30 seconds, and 72°C for 1 minute. In preparation for sequencing, PCR products were purified using QIAquick PCR Purification Kit (Qiagen Inc, Valencia, CA). Purified products were eluted in 35 μL of elution buffer. DNA cycle sequencing reactions were carried out in 10-μL reactions using 2 μL of purified PCR product, 400 nM of forward or reverse primer, and BigDye Terminator Version 3.1 Cycle Sequencing Kit (Applied Biosystems). Cycle sequencing was performed using standard sequencing conditions, and reactions were run on a 3100 DNA Sequencer (Applied Biosystems).
The ABCC1, ABCC2, and ABCB1 variants were genotyped by direct sequencing according to the methods described in Leabman et al (>ebhidden Leabman MK, Huang CC, DeYoung J, et al. Proc Natl Acad Sci USA 100:5896-5901, 2003). SLCO1B1*1b and *5 were genotyped by an SBE method. The SLCO1B1 genomic sequence from GenBank AC022335.8 was used for designing the primers. The PCR and SBE primers used for 388A>G were as follows: 5′-TTC AGT AcA TAA GCA AAA TGT T-3′ (forward PCR primer), 5′-CAC AAC AAgTcT TAG AGA TG-3′ (reverse PCR primer), and 5′-AGG TAT TCT AAA GcA ACT AAT ATC-3′ (upstream extension primer). The modified bases in the 388A>G primers are indicated in lowercase. The PCR and extension primers used for 521T>C were as follows: 5′-TGA AAC ACT CTC TTA TCT AC-3′ (forward PCR primer), 5′-TTA CCT AAA TAC AAA GAA GAA T-3′ (reverse PCR primer), and 5′-CGA AGC ATA TTA CCC ATG AAC-3′ (downstream extension primer). The PCR reactions for 388A>G (239-bp fragment) and 521T>C (175-bp fragment) were performed separately in a 15-μL volume containing 125 nM of each primer, 2.5 mM of MgCl2, 50 μM of each dNTP, 0.375 U of AmpliTaq Gold polymerase (Applied Biosystems), and 30 ng of genomic DNA. Amplification was performed in a GeneAmp PCR System 9600 thermal cycler (Applied Biosystems) with an initial denaturation step of 15 minutes at 95°C followed by 42 cycles of 95°C for 15 seconds, annealing temperature for 15 seconds, and 72°C for 30 seconds, and a final extension step of 72°C for 10 minutes. The annealing temperatures were 54.5°C and 54°C for 521T>C and 388A>G, respectively. PCR products were purified by treatment with shrimp alkaline phosphatase and exonuclease I at 37°C for 45 minutes before the SBE reaction. The SBE reactions were carried out separately for 388A>G and 521T>C in a 12-μL volume containing 1 μM of SBE primer, 250 μM each of four ddNTPs, 1.5 U of ThermoSequenase (GE Healthcare, Piscataway, NJ), and 7.2 μL of purified PCR products. Reactions were run in a 9600 thermal cycler under the following conditions: 96°C for 2 minutes, followed by 60 cycles of 96°C for 30 seconds, 55°C for 30 seconds, and 60°C for 30 seconds. The reactions were denatured at 96°C for 4 minutes and held at 4°C before separation of the SBE products on a WAVE 3500HT DHPLC system (Transgenomic Inc). Each SBE product for 388A>G and 521T>C was pooled, and 18 μL was injected onto the WAVE 3500HT system for analysis. An oven temperature of 70°C was used, and samples were run on an HT column (Transgenomic Inc) with a buffer B (Transgenomic Inc) gradient range of 27.1% to 39.6% over 2.5 minutes (buffer B contains 25% acetonitrile). The extended products were eluted in the order of C<G<T<A dependent on the hydrophobicity differences of the four bases. Both SBE products and unextended primers were eluted at different times and could be distinguished. Positive controls with known genotypes at the 388A>G and 521T>C loci were included in each run.
The ABCG2 34G>A and 421C>A variants were genotyped by duplex PCR and duplex SBE analyzed on DHPLC. The PCR and SBE primers used for 34G>A were as follows: 5′-GCA ATC TCA TTT ATC TGG ACT A-3′ (forward PCR primer), 5′-AAT AGC CAA AAC CTG TGA GG-3′ (reverse PCR primer), and 5′-CCA TTG GTG aTT CCT aGT GAC A-3′ (downstream extension primer). The modified bases are indicated in lowercase to reduce 3′ end dimmers and hairpins. The PCR and SBE primers used for 421C>A were as follows: 5′-ACT AAA CAG TCA TGG TCT TAG A-3′ (forward PCR primer), 5′-ATC AGA GTC ATT TTA TCC ACA C-3′ (reverse PCR primer), and 5′-CCG AAG AGC TGC TGA GAA CT-3′ (downstream extension primer). The amplified PCR fragments are 344 and 296 bp in size, respectively. Duplex PCRs were performed in a 15-μL volume containing 125 nM of each primer for the two PCR amplicons, 30 ng of genomic DNA, 2.5 mM of MgCl2, 100 μm of each dNTP, and 0.375 U of AmpliTag Gold polymerase (Applied Biosystems) in the buffer provided by the manufacturer. Amplification was performed in a GeneAmp PCR System 9600 thermal cycler (Applied Biosystems) with an initial denaturation step of 15 minutes at 95°C followed by 40 cycles of 95°C for 15 seconds, 56°C for 15 seconds, and 72°C for 45 seconds, and then an extension step of 72°C for 10 minutes. PCR products were purified by treatment with shrimp alkaline phosphatase (Roche) and exonuclease I (USB Corporation) at 37°C for 45 minutes before the SBE reaction. Duplex SBE reactions were carried out in 12.6 μL containing 1 μM of each SBE primer, 250 μM each of four ddNTPs, 7.2 μL of purified PCR product, and 1.5 U of ThermoSequenase (GE Healthcare) in 1× Reaction buffer provided by the manufacturer. Reactions were run in a 9600 thermal cycler (Applied Biosystems) under the following conditions: 96°C for 2 minutes, followed by 60 cycles of 96°C for 30 seconds, 55°C for 30 seconds, and 60°C for 30 seconds. A Wave 3500HT DHPLC system (Transgenomic Inc) was used for separating SBE products. Before run on DHPLC, the samples were denatured at 96°C for 4 minutes and held at 4°C. For analyzing SBE on Wave DHPLC, 8 μL of SBE products of each sample was injected. We used Mutation Detection application type as a template, selected Normal clean as clean type (ie, 100% buffer B clean off step after each injection), and manually set the following variables for this application: the flow rate was at 1.5 mL/min by using an HT column, oven temperature was set at 70°C, and the gradient used for elution of the SBE products was from 24% to 36.5% buffer B over 2.5 minutes (buffer B contains 25% acetonitrile). The extended products were eluted in the order of C<G<T<A dependent on the hydrophobicity differences of the four bases. In duplex SBE, two extension primers were designed to separate unextended primers and each extended product. The known genotype controls were included in each run.
In the univariate and multivariate analyses, when a particular variant included very few observations (n ≤ 3) in the homozygous genotype and was significant, these genotypes were grouped with the heterozygous genotype group; the variant was kept in the model if still significant after combining. We also combined common (n > 3) heterozygous with homozygous genotypes when they had similar effects on phenotype (ie, the same positive or negative sign of the estimate and a similar magnitude of the estimate). When the heterozygous group had a different effect from that of the two homozygous genotypes (both having a similar effect), it was combined with either homozygote genotype, and the most significant combined data were kept in the model. When the heterozygous genotype had an intermediate effect between the two homozygous genotypes, an additive model (ie, 1<2<3) was chosen and included if significant. However, when the model with combined genotypes fitted better than the additive model, the model with combined genotypes was used, and vice versa. Hardy-Weinberg equilibrium was also evaluated for all the variants tested (P < .01 for significance).
|Characteristic||No. of Patients (N = 85)||%|
|No. of prior chemotherapy regimens|
NOTE. All patients had a Karnofsky performance status ≥ 70%.
Abbreviation: BSA, body-surface area.
Supported by the Pharmacogenetics of Anticancer Agents Research Group (National Institutes of Health [NIH]/National Institutes of General Medical Sciences [NIGMS] Grant No. U01 GM61393) and the Pharmacogenetics of Membrane Transporters Research Group (NIH/NIGMS Grant No. U01 GM61390). Data will be deposited into PharmGKB (supported by NIH/NIGMS Grant No. U01GM61374). Also supported by Grant No. UO1 CA069852 and the Pharmacology Core of the University of Chicago Cancer Research Center (NIH Grant No. P30 CA14599).
Presented in part at the 40th Annual Meeting of the American Society of Clinical Oncology, June 5-8, 2004, New Orleans, LA; and 41st Annual Meeting of the American Society of Clinical Oncology, May 13-17, 2005, Orlando, FL.
Authors' disclosures of potential conflicts of interest and author contributions are found at the end of this article.
Although all authors completed the disclosure declaration, the following author(s) indicated a financial or other interest that is relevant to the subject matter under consideration in this article. Certain relationships marked with a “U” are those for which no compensation was received; those relationships marked with a “C” were compensated. For a detailed description of the disclosure categories, or for more information about ASCO's conflict of interest policy, please refer to the Author Disclosure Declaration and the Disclosures of Potential Conflicts of Interest section in Information for Contributors.
Employment or Leadership Position: None Consultant or Advisory Role: Mark J. Ratain, Genzyme (C) Stock Ownership: Gary L. Rosner, Pfizer Inc Honoraria: None Research Funding: None Expert Testimony: None Other Remuneration: Federico Innocenti, University of Chicago (royalties); Mark J. Ratain, University of Chicago (royalties)
Conception and design: Federico Innocenti, Deanna L. Kroetz, M. Eileen Dolan, Mark J. Ratain
Financial support: Deanna L. Kroetz, Mark J. Ratain
Administrative support: Mark J. Ratain
Provision of study materials or patients: Mark J. Ratain
Collection and assembly of data: Federico Innocenti, Deanna L. Kroetz, Erin Schuetz, Jacqueline Ramírez, Mary Relling, Peixian Chen, Soma Das, Mark J. Ratain
Data analysis and interpretation: Federico Innocenti, Deanna L. Kroetz, Jacqueline Ramírez, Peixian Chen, Gary L. Rosner, Mark J. Ratain
Manuscript writing: Federico Innocenti, Deanna L. Kroetz, Jacqueline Ramírez, Peixian Chen, Gary L. Rosner, Mark J. Ratain
Final approval of manuscript: Federico Innocenti, Deanna L. Kroetz, Erin Schuetz, M. Eileen Dolan, Jacqueline Ramírez, Mary Relling, Peixian Chen, Soma Das, Gary L. Rosner, Mark J. Ratain