UGT1A7 and UGT1A9 Polymorphisms Predict Response and Toxicity in Colorectal Cancer Patients Treated with Capecitabine/Irinotecan

INTRODUCTION

Colorectal cancer (CRC) is the third most common cancer in both men and women and the third most prevalent cause of cancer-related death (1). A common approach to the systemic management of metastatic colorectal cancer is a combination of fluoropyrimidine (e.g., 5-fluorouracil or capecitabine) and irinotecan (reviewed in ref. 2). Fluoropyrimidines act primarily through the inhibition of thymidylate synthase (TS), resulting in impaired DNA synthesis and cell death. Irinotecan is a camptothecin compound that inhibits DNA topoisomerase I, resulting in an accumulation of DNA damage and cell death. Irinotecan is a prodrug that undergoes conversion by liver carboxylesterases to form the active compound SN-38, 7-ethyl-10-hydroxy-camptothecin (2, 3). SN-38 is largely metabolized to the inactive glucuronide, SN-38G, via the action of several UDP-glucuronosyltransferase (UGT) including hepatic UGT1A1, UGT1A6, and UGT1A9 and extrahepatic UGT1A7 (4–7).

The existence of significant patient variability in response to fluoropyrimidines and irinotecan, coupled with knowledge of common genetic polymorphisms within genes important to the pharmacology of these drugs, suggests that pharmacogenetic studies could identify individuals likely to benefit from treatment or develop severe toxicity. Genetic variation within the promoter and the 3′-untranslated region (UTR) of the human TS gene has been previously linked to the efficacy of 5-fluorouracil-related drugs (reviewed in ref. 8). The polymorphic UGT1A family members may govern variability in patient response to irinotecan (reviewed in ref. 9). A common TA repeat within the promoter of the human UGT1A1 gene has been associated with CPT-11-related toxicity, most predominantly neutropenia (10–12). Functionally significant genetic variation has also been described for human UGT1A6, UGT1A7, and UGT1A9 (6, 13–17). In this study, we correlate the efficacy and toxicity of combined capecitabine/CPT-11 therapy with genetic variation in genes important in the metabolism of CPT-11 (UGT1A1, UGT1A6, UGT1A7, and UGT1A9). Given the narrow therapeutic index, frequent resistance, and expanding options for treatment of CRC, the need for predictive markers of response and toxicity has assumed increased importance.

MATERIALS AND METHODS

Patient Eligibility. This pharmacogenetic analysis was a secondary objective of a multicenter phase II trial of capecitabine/CPT-11 combination therapy in patients with metastatic CRC. The primary clinical objective was determination of objective response. Eligible patients were at least 18 years old with histologically confirmed metastatic colorectal adenocarcinoma. Previous cytotoxic chemotherapy was not permitted except for neoadjuvant or adjuvant treatment completed 12 months before study enrollment. Patients were required to be ambulatory with a Karnofsky performance status of ≥70%. Required laboratory values for inclusion included neutrophil count ≥1.5 × 10⁹, platelet count ≥100 × 10⁹/L, serum creatinine ≤1.5 × upper limit normal, estimated creatinine clearance ≥50 ml/min, serum bilirubin ≤1.25 × upper limit normal, ALAT and ASAT ≤2.5 × upper limit normal (<5 with liver metastasis or <10 with bone metastasis), or alkaline phosphatase ≤2.5 × upper normal limit (<5 with liver metastasis or <10 with bone metastasis). Exclusion criteria included known Gilbert's disease, pregnancy, central nervous system metastasis, active cardiac disease, or myocardial infarction within the previous 12 months, active infections, physical disorders of the gastrointestinal tract, or problems with malabsorption. Written informed consent was required, and the study was approved by the institutional review boards at all of the participating sites in accord with an assurance filed with and approved by the U.S. Department of Health and Human Services.

Drug Administration. Capecitabine (Xeloda, Roche Laboratories, Inc., Nutley, NJ) was given orally at a dosage of 1,000 mg/m² (cohort 1, 15 patients) or 900 mg/m² (cohort 2, 52 patients) twice daily on days 2 through 15 of 3-week cycles. Irinotecan (Camptosar, Pfizer, Inc., New York, NY) was given at a dosage of 125 mg/m² (cohort 1, 15 patients) or 100 mg/m² (cohort 2, 52 patients) as a 90-minute i.v. infusion on days 1 and 8 of each cycle. The doses of capecitabine and irinotecan were decreased as described for cohort 2 after the first 15 patients (cohort 1) because of unacceptable toxicity. The maximum number of cycles given was 12.

Evaluation of Response and Toxicities. Baseline tumor measurements were obtained within 21 days before initial treatment. Response assessments were obtained every two cycles (6 weeks) with confirmation of response after 4 weeks according to RECIST criteria (18). Toxicity was assessed weekly during the first two cycles of treatment according to the National Cancer Institute Common Toxicity Criteria, version 2.0 (National Cancer Institute Common Toxicity Criteria, http://ctep.cancer.gov). All toxicities evaluated in this study occurred during the first two cycles of treatment.

Genomic DNA Preparation and Genotyping Assays. Blood samples were collected for isolation of genomic DNA during venipuncture for other diagnostic labs at least 1 week before starting treatment. Genomic DNA was prepared from peripheral WBCs and suspended in 10 mmol/L Tris with 50 mmol/L EDTA (TE; Covance, Princeton, NJ). DNA was isolated from 66 of the 67 subjects and was used to identify UGT1A1, UGT1A6, UGT1A7, UGT1A9, and thymidylate synthase polymorphisms. For each batch of assays, appropriate positive and negative controls of established genotype were assayed.

UGT1A1. The UGT1A1 variable length (TA)_n repeat polymorphism (n = 5-8) was evaluated using Genescan technology (Applied Biosystems, Foster City, CA; refs. 19, 20). A 254-bp region of the UGT1A1 gene was amplified using a hexachloro-6-carboxyfluorescein fluorescently tagged antisense primer (5′-ATCAACAGTATCTTCCCAG-3′) and a nonfluorescent sense primer (5′-TATCTCTGAAAGTGAACTC-3′). The PCR reaction mixture included 20 mmol/L Tris-HCl (pH 8.3), 50 mmol/L KCl, 1.5 mmol/L MgCl₂, 0.2 mmol/L each deoxynucleotide triphosphate (dNTP), 0.2 μmol/L each primer, 1.25 units Platinum Taq polymerase (Invitrogen, Carlsbad, CA), and 20 ng genomic DNA in a 50 μL reaction volume. The PCR profile included 1-minute denaturation at 94°C, 34 cycles of 30 seconds at 94°C, 30 seconds at 52°C, and 30 seconds at 72°C, with a final 5-minute extension at 72°C in a Perkin-Elmer 9700 Thermal Cycler (Perkin-Elmer, Boston, MA). Fluorescently labeled products were electrophoretically separated on a 4% polyacrylamide gel on an ABI 377 DNA Sequencer. Fluorescent bands were analyzed using GENESCAN 2.1 software to determine fragment length (Applied Biosystems). Genotypes were assigned as UGT1A1*1 (reference), UGT1A1*36, UGT1A7*28, and UGT1A7*37 for TA repeat numbers of 6, 5, 7, and 8, respectively.

UGT1A6.UGT1A6 genotype was determined with a PCR-based assay that uses Pyrosequencing technology (Pyrosequencing, Westborough, MA). UGT1A6 alleles are defined by permutations of three single-nucleotide polymorphisms (SNPs) that alter the encoded amino acid sequence at nucleotide positions 19T > G, 541A > G, and 552A > C (for amino acids S7A, T181A, and R184S, respectively; ref. 15). For detection of the 19T > G SNP, a 238-bp fragment was amplified using primer F-53 (5′-GATTTGGAGAGTGAAAACTCTTT-3′) and R184 (5′-biotin-CAGGCACCACCACTA-CAATCTC-3′). For the remaining two SNPs, a 215-bp fragment was amplified using primer F414 (5′-biotin-CTTTAAGGAGAGCAAGTTTGATG-3′) and R628 (5′-CCACTCGTTG-GGAAAAAGTC-3′). Approximately 25 to 50 ng of genomic DNA was amplified in a reaction mixture containing 20 mmol/L Tris-HCl (pH 8.4), 50 mmol/L KCl, 1.5 mmol/L MgCl₂, 0.2 mmol/L dNTPs, 0.2 μmol/L primers, and 2.5 units of Platinum Taq polymerase in a 50 μL volume. Reactions were performed in a Perkin-Elmer 9700 Thermal Cycler with 45-second denaturation at 94°C, followed by 35 cycles of 94°C for 30 seconds, 56°C for 30 seconds, and 72°C for 50 seconds. A final 3-minute extension at 72°C completed the amplification.

Amplicons were prepared for automatic Pyrosequencing SNP analysis on the PSQ 96 system using reagents from the PSQ96 SNP Reagent Kit (Pyrosequencing). A 25-μL volume of double-stranded biotinylated amplicon was incubated with 100 μg of streptavidin-coated M280 DynaBeads (Dynal, Brown Deer, WI) in binding buffer [5 mmol/L Tris (pH 7.6), 1 mol/L NaCl, 0.5 mmol/L EDTA, 0.05% Tween 20] and incubated at 65°C for 15 minute followed by denaturation in 0.5 mol/L NaOH. Single-stranded biotinylated DNA was transferred to annealing buffer [20 mmol/L Tris Acetate (pH 7.6) and 5 mmol/L Mg(OAc)₂] for 1 minute then transferred to sequencing primer solution containing annealing buffer and 10 pmol of the appropriate sequencing primer. The 19T > G primer was 5′-GATGGCCTGCCTCCTT-3′, whereas the 541A > G and 552A > C primer was 5′-AGGACACAGGGTCTG-3′. The nucleotide dispensation sequence for the 19T > G SNP was TCGCTGACA with the underlined nucleotide representing the negative control and those in bold representing the polymorphic site. The nucleotide dispensation sequence for the remaining SNPs was CGCTGACTGCTGATGTCAT. Genotypes were assigned as UGT1A6*1 (S7, T181, and R184 reference), UGT1A6*2 (A7, A181, and S184), UGT1A6*3 (A7) and UGT1A6*4 (A7 and S184).

UGT1A7.UGT1A7 genotype was determined by direct sequencing of a PCR product that spans all of the polymorphic sites. The UGT1A7 alleles are defined by permutations of six SNPs (342G > A, 387T > G, 391C > A and 392G > A, 417G > C, and 622T > C) that alter the encoded amino acid sequences (G115S, N129K, R131K, E139D, and W208R, respectively; ref. 14). A 415-bp fragment from UGT1A7 was amplified using primer F277 (5′-TTTGCCGATGCTCGCTGGACG-3′) and R692 (5′-GCTAT-TTCTAAGACATTTTTGAAAAAATAGGG-3′). The PCR reaction mixture included JumpStart REDTaq ReadyMix PCR reaction mix (Sigma, St. Louis, MO) containing 10 mmol/L Tris-HCl (pH 8.3), 50 mmol/L KCl, 2 mmol/L MgCl₂, 0.2 mmol/L each dNTP, and 0.03 units/Taq polymerase with the addition of 0.2 μmol/L each primer and 40 ng genomic DNA in a 50-μL reaction volume. The PCR conditions included 1-minute denaturation at 94°C, 35 cycles of 30 seconds at 94°C, 30 seconds at 56°C, and 30 seconds at 72°C, with a final 5-minute extension at 72°C in a Perkin-Elmer 9700 Thermal Cycler. Amplified products were sequenced using an ABI Prism 377 DNA Sequencer (Applied Biosystems) and primers F292 (5′-TGGACGGCACCATTG-3′) and R675 (5′-TTTGAAAAAATAGGGGCAA-3′). Genotypes were assigned as UGT1A7*1 (G115, N129, R131, and W208 reference), UGT1A7*2 (K129 and K131), UGT1A7*3 (K129, K131, and R208), UGT1A7*4 (R208).

UGT1A9. Five UGT1A9 polymorphisms were evaluated by direct DNA sequencing of a PCR amplicon spanning all of the polymorphic sites. Polymorphisms of interest included a −118 (dT)_n repeat (n = 9 or 10), C3Y (8G > A), M33T (98T > C), Y242X (726T > G), and D256N (766G > A; refs.14, 21, 22). We also detected the following promoter: and intronic SNPs: −87G > A, and intron 1 I143C > T, I152G > A, I201A > C, I219T > A, and I313A > C. The −118 (dT)_9/10 polymorphism has been previously localized to nucleotides −98 through −109 (16, 17). However, we believe it is most accurate to assign the localization of this poly d(T) tract as spanning nucleotides −118 through −129 (based on assigning the “A” in the “ATG” translation start codon as +1) and we refer to the polymorphisms as −118 (dT)₉ or −118 (dT)₁₀.

A 1.4-kb fragment from UGT1A9 was amplified using primer F-273 (5′-AAACTTAACATTGCAGCACAGGGC-3′) and RI337 (intron 1, 5′-CTAAGACCATTT-CCTCTGGGGC-3′). The PCR reaction mixture included 10 mmol/L Tris-HCl (pH 8.3), 50 mmol/L KCl, 2 mmol/L MgCl₂, 0.2 mmol/L each dNTP, 0.2 μmol/L each primer, 5% DMSO, 1.5 units Platinum Taq polymerase (Invitrogen), and 40 ng genomic DNA in a 50-μL reaction volume. The PCR conditions included 1-minute denaturation at 94°C, 35 cycles of 94°C for 30 seconds, 55°C for 30 seconds, and 72°C for 1.5 minutes, with a final 5-minute extension at 72°C in a Perkin-Elmer 9700 Thermal Cycler. Polymorphisms were detected by direct sequencing (ABI 377 DNA Sequencer, Applied Biosystems) of amplified sequences using primers F-273, R361 (5′-AAAAATAAGTCAAAAATGTC-ATTGT-3′), Fsp (5′-GGAGGAACATTTATTATGCCACCG-3′), and RI337 primers.

Observed haplotypes were designated UGT1A9HI (10GCGAC), UGT1A9HII (9GCGTA), UGT1A9HIII (9GCATA), UGT1A9HIII (9ACGTA), and UGT1A9HIV (9GTGTA) for the following polymorphisms detected in 66 subjects: −118 (dT)_n repeat, −87G > A, and the intron 1 SNPs I143C > T, I152G > A, I219A > T, and I313C > A.

Thymidylate Synthase Promoter.TS genotypes were determined using a combination of gel electrophoresis–based and direct DNA sequencing techniques. Common functional polymorphisms in the TS gene include a 28-bp variable tandem repeat (n = 2, 3, 4, 5, or 9) within an enhancer of the promoter and a SNP within the 5′-UTR of the three repeat (3R) allele (−58G > C; refs.23, 24). The variable tandem repeat was detected by PCR amplification with F-220 (5′-GTGGCTCCTGCGTTTCCCCC-3′) and R + 1 primers (5′-TCCGAGCCGGCCACAGGCAT-3′) that flank the repeat region. The PCR reaction mixture included 10 mmol/L Tris-HCl (pH 8.3), 50 mmol/L KCl, 2 mmol/L MgCl₂, 5% DMSO, 0.2 mmol/L each dNTP, 0.2 μmol/L each primer, 1.5 units Platinum Taq polymerase (Invitrogen), and 40 ng genomic DNA in a 50-μL reaction volume (25). The PCR conditions included 1-minute denaturation at 94°C, 34 cycles of 94°C for 40 seconds, 70°C for 40 seconds and 72°C for 40 seconds, with a final 5-minute extension at 72°C in a Perkin-Elmer 9700 Thermal Cycler. After separating the gene fragments via gel electrophoresis, fragment lengths were determined relative to known molecular markers. Direct sequencing of the amplified products allowed identification of the 5′-UTR SNP. Genotypes were assigned according to repeat lengths and the existence of the SNP: one repeat (1R), two repeats (2R reference), three repeats (3RC or 3RG), four repeats (4R), five repeats (5R), or nine repeats (9R).

Thymidylate Synthase 3′-UTR. The presence or absence of the 6-bp sequence TTAAAG at position 1494 of the TS mRNA was detected by PCR amplification with F3′utrTS (5′-CAAATCTGAGGGAGCTGAGT-3′) and R3′utrTS (5′-CAGATAAGTGGCAGTACAGA-3′) to produce either 148- or 142-bp fragments (26). The PCR reaction mixture included 10 mmol/L Tris-HCl (pH 8.3), 50 mmol/L KCl, 2.5 mmol/L MgCl₂, 0.2 mmol/L each dNTP, 0.2 μmol/L each primer, 1.5 units Platinum Taq polymerase (Invitrogen), and 40 ng genomic DNA in a 50 μL reaction volume (26, 27). The PCR profile included 1-minute denaturation at 94°C, 35 cycles at 94°C for 30 seconds, 58°C for 45 seconds, and 72°C for 45 seconds, and a final 5-minute extension at 72°C in a Perkin-Elmer 9700 Thermal Cycler. Amplified products were separated via gel electrophoresis on 3.5% Metaphor agarose and fragment lengths were determined relative to known molecular markers. Genotypes were assigned according to the presence or absence of the 6-bp sequence as determined by fragment lengths: 6 bp (148 bp, reference) or 0 bp (142 bp, deletion).

Statistical Methods. The sample size for this trial (ultimately cohort 2) was governed by the primary objective of response rate (complete or partial) and powered to reject the null hypothesis of 30% with a one-sided significance level of 5% (α = 0.05). This provided at least 80% power to reject the null hypothesis at a response rate of 50% with a sample of 45 patients. The overall significance level was 0.028 and the overall power was 0.814. A total of 50 patients allowed for an approximate dropout or nonevaluable rate of 10%.

Two-sided tests of statistical significance were used to determine statistically significant P values (P < 0.05) using SAS software. A permutation procedure based on an exact test was applied to check Hardy-Weinberg equilibrium of each marker, and P < 0.05 indicated a lack of agreement with Hardy-Weinberg equilibrium (28). Haplotype frequencies and their corresponding variance were estimated by expectation-maximization algorithm and jackknife method, respectively. Toxicity was dichotomized with toxicity grades 3 or 4 as one category and grade ≤2 as another. Statistical associations between genotypes and drug response or toxicity were examined using a two-sided Fisher's exact test for each individual marker. The exact test of Cochran-Armitage trend test was done to detect the direction of the relationship between toxicity or response rate and genotype activity bins. Associations between genotype groups and clinical measurements were examined using the nonparametric Kruskal-Wallis test and the Jonckheere-Terpstra test.

For the purposes of genotype/response analyses, we binned genotypes into functional categories based on published descriptions of the alleles or genotypes of interest. UGT1A1 genotypes were grouped into high (UGT1A1 5/6 and UGT1A1 6/6), moderate (UGT1A1 6/7), and low (UGT1A1 7/7 and UGT1A1 7/8) activities (19, 20). UGT1A6 was defined by high (UGT1A6*2/*2), moderate (UGT1A6*1/*1), low (UGT1A6*1/*2), and undefined/unknown activities (UGT1A6*1/*3 and UGT1A6*1/*4; 15). UGT1A7 genotypes were categorized into high (UGT1A7*1/*1), moderate (UGT1A7*1/*2 and UGT1A7*1/*3), low (UGT1A7*2/*2 and UGT1A7*3/*3), and undefined/unknown (UGT1A7*2/*3) activity (14, 29). For UGT1A9, genotypes and polymorphisms were analyzed individually for associations with toxicity and response due to the lack of functional data regarding these variations. However, a recent publication reported 2.6-fold higher transcriptional activity associated with the −118 (dT)₁₀ allele compared with the −118 (dT)₉ allele based on in vitro transcriptional reporter assays (16). The TS promoter activities were defined as high (3G/3G), moderate (2/3G and 3G/3C), and low activity (2/2, 2/3C, and 3C/3C). A separate analysis of high (3G/3G, 2/3G, and 3G/3C) and low activity (2/2, 2/3C, and 3C/3C) was also completed due to discrepancies in the literature (23, 24), . Finally, the 3′-UTR TS binning included high (6/6), moderate (0/6), and low (0/0) activities (26, 30).

RESULTS

A total of 67 patients with metastatic colorectal cancer were enrolled in this prospective phase II trial, and germline DNA was available for 66 of these patients. Of those 66 subjects, response and toxicity data were collected from 56 and 66 subjects, respectively. Fifty-five subjects (83%) were Caucasian, nine (14%) were African American, one was Samoan, and one was Hispanic (Table 1). The overall objective response rate (partial or complete response) was 45% (30 of 67 patients). The incidence of severe or life-threatening diarrhea or neutropenia (grade 3 or 4, National Cancer Institute Common Toxicity Criteria version 2.0) was 33% (22 of 66 subjects). Nineteen of 66 subjects (29%) developed grade 3 or 4 diarrhea alone, one subject (1%) exhibited grade 3 or 4 neutropenia alone, and two subjects (3%) experienced both diarrhea and neutropenia within the first two cycles of treatment.

Allele and Genotype Frequencies and Hardy-Weinberg Equilibrium. We evaluated the predictive value of genotypes within the UDP-glucuronosyltransferases (UGT1A1, UGT1A6, UGT1A7, and UGT1A9) as well thymidylate synthase. Table 2 lists the observed allele frequencies, whereas Table 3 lists genotype frequencies. There were no significant differences in allele frequencies between cohorts 1 and 2. Therefore, these cohorts were combined for genotype/phenotype analyses. With the exception of UGT1A9 and TS promoter polymorphisms, all genotypes were observed in Hardy-Weinberg equilibrium (Table 2). The fairly rare UGT1A1 TA₅ and TA₈ alleles were observed only in African American patients at frequencies consistent with published values (19, 31). The genotype frequencies observed for UGT1A1 6/6, UGT1A1 6/7, and UGT1A1 7/7 (Table 3) agreed with that reported in a random population (13, 20) despite the exclusion of patients with elevated bilirubin, a phenotype associated with the UGT1A1 TA₇ allele (10, 32). Hence, the distribution of UGT1A1 alleles followed Hardy-Weinberg equilibrium (Table 2). Allele and genotype frequencies for UGT1A6 (Tables 2 and 3) agreed with our previous study (15). UGT1A7 allele and genotype frequencies were as expected from the literature (13, 14), although we did not observe UGT1A7*4 (Table 3).

We did not detect the C3Y, M33T, Y242X, or D256N polymorphisms described for UGT1A9 (14, 21, 22). However, we did identify the −118 (dT)_n polymorphism as well as four SNPs within intron sequences of the UGT1A9 gene (Tables 2 and 3) that defined a total of five haplotypes (Table 4). The −118 (dT)_n repeat and intron 1 SNPs I219 and I313 were completely linked in this study population. UGT1A9 genotypes were in agreement with Hardy-Weinberg equilibrium (Table 4), although the haplotype frequencies were not (Table 2). Although this study was completed in a diseased population, haplotype frequencies closely resembled those observed within a healthy population (n = 53) of mixed ethnicity, mainly comprised of Caucasians (59.4% for n = 9 and 40.6% for n = 10 repeats).⁴

We identified two novel alleles within the TS promoter, defined by one repeat (1R) from a Caucasian subject and a 2R allele with a novel SNP (−30C > T) in the second repeat from a patient of undefined ethnicity (Tables 2 and 3). The rare 4R and 9R alleles were detected in two subjects of non-Caucasian ethnicity (31). Allele (Table 2) and genotype (Table 3) frequencies for the TS 3′-UTR polymorphism were within the ranges reported by previous studies (26, 27). However, the linkage disequilibrium reported between the 3R TS promoter and 0 bp 3′-UTR polymorphisms (26, 30) was not observed in this study.

Genotype Correlations with Tumor Response. We observed an association between the low activity UGT1A7*2/*2 and UGT1A7*3/*3 genotypes and the UGT1A9 (dT)₉ polymorphism and increased tumor response (complete or partial, Tables 5 and 6). Among individuals with low activity UGT1A7*2/*2 and UGT1A7*3/*3 genotypes, the response rate was 85% (P = 0.013, 11 of 13 patients responded compared with 44% or 19 of 43 patients with other genotypes, Table 5). Furthermore, a statistically significant trend was observed for increased response across UGT1A7 genotypes (p for trend = 0.017; Table 5). A statistically significant trend was also observed between the UGT1A9 −118 (dT)_n repeat and tumor response (p for trend = 0.033, Table 6). The −118 (dT)_9/9 genotype was significantly associated with efficacious tumor response when compared with all other genotypes (p = 0.047; 74% or 14 of 19 patients with −118 (dT)_9/9 versus 43% or 16 of 37 patients with other genotypes).

We found no significant association between UGT1A1 genotype and tumor response. Although not statistically significant, subjects with low UGT1A1 enzyme activity responded better to treatment (83% for UGT1A1 7/7 or UGT1A1 7/8 genotype relative to 46% for high activity UGT1A1 5/6 or UGT1A1 6/6, Table 7). No significant associations or trends were observed between UGT1A6 genotype (Table 8), TS polymorphisms, or overall UGT1A9 haplotypes and tumor response.

Genotype Correlation with Toxicity. Twenty-two subjects developed grade 3 or 4 diarrhea or neutropenia (33%, Table 1). The predominant toxicity observed was diarrhea (21 subjects) as opposed to neutropenia (three subjects). We observed a striking correlation between the low activity UGT1A7 genotypes (UGT1A7*2/*2 and UGT1A7*3/*3) and lack of drug toxicity (p = 0.003, Table 5). The UGT1A9 −118 (dT)_9/9 genotype was also associated with a lower incidence of toxicity (p = 0.002, Table 6). We found no statistically significant association between UGT1A1 genotype and toxic events. However, none of the six subjects with low activity UGT1A1 genotypes (UGT1A1 7/7 or UGT1A1 7/8) experienced toxicity (Table 7). There was no significant association between UGT1A6 genotype and incidence of toxicity, but none of the five patients with high activity UGT1A6*2/*2 genotype experienced toxicity (Table 8). There was no association between TS polymorphisms and incidence of toxicity.

Because the UGT1A1 promoter polymorphism has been associated with neutropenia, it is of interest to report the genotypes specifically of the individuals that experienced neutropenia. Three subjects (4.5%) developed grade 3 neutropenia. Two subjects possessed the UGT1A1 6/7 genotype coupled with either the UGT1A7*1/*2 or UGT1A7*1/*3 genotype, whereas the other subject possessed UGT1A1 6/8 and UGT1A7*1/*2. All three were heterozygous for the UGT1A9 −118 (dT)_n repeat. Two of the subjects with neutropenia also exhibited grade 3 diarrhea; both patients possessed the UGT1A7*1/*2 genotype.

Linkage Disequilibrium/Haplotyping Analysis. We observed UGT1A haplotype frequencies similar to those obtained by Kohle et al. (13) for UGT1A1, UGT1A6, and UGT1A7 (Table 9). Additionally, we expanded our haplotype analysis to include the UGT1A9 −118 (dT)_n repeat. We considered only this polymorphism for UGT1A9 because the observed UGT1A9 haplotypes were not in Hardy-Weinberg equilibrium (Tables 2 and 4). We discovered that the low activity UGT1A7*2 and UGT1A7*3 alleles were completely associated with the UGT1A9 −118 (dT)₉ allele, whereas the high activity UGT1A7*1 allele was linked with the UGT1A9 −118 (dT)₁₀ allele (Table 9, likelihood ratio test of association p < 0.0001). Such clearly defined associations across haplotypes were not observed for UGT1A1 or UGT1A6. Finally, a statistically significant association was identified between haplotype I and patient response and toxicity (p = 0.04 and p = 0.035, respectively, Table 9).

DISCUSSION

This study was significant for two major observations. First, genetic variation in the human UGT1A7 and UGT1A9 genes predicted for tumor response as well as development of diarrhea during combination irinotecan/capecitabine therapy and second, low activity UGT1A1 alleles did not predict for the development of diarrhea. We observed associations between UGT1A7 and UGT1A9 genotypes with both tumor response and toxic events (Tables 5 and 6). We also found that the UGT1A9 −118 (dT)_9/9 genotype was predictive for efficacious tumor response with lower incidence of toxicity (Table 6), whereas the UGT1A9 −118 (dT)_10/10 genotype predicted for poor tumor response. The functional consequences of the UGT1A9 promoter polymorphism are not well established. However, the −118 (dT)₁₀ allele has recently been associated with 2.6-fold greater transcriptional activity than the −118 (dT)₉ allele (16). If these in vitro studies are predictive of in vivo function (which remains to be proven), then we would predict that alleles containing the UGT1A9 (dT)₉ polymorphism might confer low activity relative to alleles defined by the UGT1A9 (dT)₁₀ polymorphism.

Only one previous study has examined the association of UGT1A7 pharmacogenetics in patients receiving irinotecan and those investigators reported no significant association between UGT1A7 genotype and the occurrence of toxicity (33). However, it is important to note that the frequencies of the UGT1A7*2 and UGT1A7*3 alleles are significantly lower in the Japanese population, so that the power to detect associations in that study may have been compromised. To our knowledge, no previous studies have addressed the association between UGT1A9 polymorphisms and the incidence of irinotecan toxicity or tumor response.

The present study allowed us to evaluate UGT1A haplotypes (Table 9). This is an important component of this study because the human UGT1A genes comprise a unique nested gene structure on human chromosome 2 allowing for linkage disequilibrium among these genes (13). This raises the possibility that association between UGT1A alleles and specific phenotypes may be due to linkage disequilibrium. We observed complete linkage between the UGT1A7 low activity alleles (UGT1A7*2 and UGT1A7*3) and the UGT1A9 (dT)₉ polymorphism (Table 9). Both enzymes have been reported to catalyze the glucuronidation of SN-38 with significant efficiency (6, 7, 34). UGT1A9 is expressed in the colon and liver and might be predicted to affect plasma levels of SN-38 (35, 36). UGT1A7 is not expressed in the liver, but is expressed in the proximal gastrointestinal tract and may influence the disposition of SN-38 within the gut (35, 36). Based on our current knowledge of tissue expression patterns, one might speculate that UGT1A9 is more likely to affect irinotecan disposition than is UGT1A7; however, the UGT1A7 alleles are clearly functionally significant and it is possible that UGT1A7 is expressed (either constitutively or inducibly) in as yet undetermined tissues that might influence SN-38 disposition. Therefore, it is premature at this time to predict which of the two genes is more likely the biological connection to the observed responses to irinotecan. Overall, we determined that low activity UGT genotypes were associated with better tumor response (Tables 5 and 6). This is consistent with the notion that low UGT activity might predict for higher concentrations of plasma SN-38 and increased tumor concentration of SN-38.

Previous studies have described an association between the UGT1A1 TA₇ promoter polymorphism and the occurrence of irinotecan-induced toxicity (10–12). We were surprised that we did not observe the expected association between UGT1A1 low activity genotypes (UGT1A1 7/7 or UGT1A1 7/8) and increased incidence of toxicity (Table 7). Indeed, although not statistically significant, we observed the opposite trend. Of the six subjects with low activity genotypes, none experienced toxicity. The distribution of UGT1A1 genotypes in this study was reflective of a random population although patients were excluded with high serum bilirubin levels. A careful study of the primary literature regarding UGT1A1 pharmacogenetics suggests a clear link between low activity alleles and plasma ratio of SN-38/SN-38G as well as the development of neutropenia (10–12, 33), but the association between these alleles and the development of diarrhea has been much weaker. That is true, in part, because neutropenia was the predominant dose-limiting toxicity observed in these studies.

We suggest that different dosing regimens of irinotecan predispose patients to different toxicities. Our study differed from the previous studies in that subjects received combination irinotecan (100 or 125 mg/m² on days 1 and 8) with oral capecitabine therapy. With this regimen, we observed diarrhea to be the predominant toxicity (22 subjects with grade 3 or 4 toxicity: 19 with diarrhea alone, two with diarrhea and neutropenia, and one with neutropenia alone). This provided us with a unique opportunity to study the association between UGT alleles and specifically irinotecan-induced diarrhea. We observed that low activity alleles protected subjects from developing diarrhea, rather than predisposing for toxicity as has been observed for irinotecan-associated neutropenia.

This apparent dichotomous set of observations is plausible when one considers the complex pharmacology of irinotecan. Glucuronidation represents an elimination pathway from the plasma. Hence, low capacity to glucuronidate results in increased plasma levels of SN-38 (11) and increased susceptibility to systemic toxicities such as neutropenia (10–12, 33). However, glucuronidation also represents a route of delivery of SN-38 to the gastrointestinal tract. In humans, renal excretion of irinotecan and its metabolites accounts for <20% of the dose with most of the remaining elimination, particularly for SN-38G, occurring via the biliary route (37–39). It is well documented that bacterial glucuronidases within the gut catalyze the deglucuronidation of many drugs, including SN-38G (reviewed in ref. 40). Therefore, glucuronidation of SN-38 directs its delivery to the gut where the toxic SN-38 molecule can be regenerated to promote lesions within the mucosa of the gastrointestinal tract leading to diarrhea. Thus, glucuronidation detoxifies SN-38 in the liver, contributing to lower systemic circulation of SN-38 and protects against systemic toxicity including neutropenia; but the glucuronidation pathway, in directing SN-38G to the gut, may predispose for irinotecan-induced diarrhea.

Clinical and animal studies seem to support this hypothesis. At least two clinical pharmacokinetic studies suggested that neutropenia, but not diarrhea, was significantly associated with the AUC for CPT-11 and/or SN-38 (11, 41). In rats, the activity of bacterial β-glucuronidase, the enzyme that generates SN-38 from SN-38G in the gut, correlated with irinotecan-induced cecal damage and was attenuated by administration of antibiotics (42) or of a specific β-glucuronidase inhibitor (43). Furthermore, human studies have shown that administration of antibiotics before irinotecan therapy significantly diminishes the incidence of severe diarrhea (44). Thus, glucuronidation may well promote one toxicity (diarrhea) yet protect for the other (neutropenia). Hence, genetic variation in specific UGT genes may predict for either high incidence of toxicity or low incidence depending on the dosing regimen.

These data have important public health implications. It has been suggested that clinical testing for the UGT1A1 promoter polymorphism should be implemented as a predictor of toxicity in patients receiving irinotecan (45, 46). It possible that for dosing regimens in which neutropenia is clearly the dose-limiting toxicity this practice might benefit patients. However, we observed that the dose limiting toxicity of combination capecitabine/irinotecan was diarrhea and that low capacity to glucuronidate may be protective for (rather than a risk factor for) occurrence of diarrhea and also predicts for efficacious tumor response. Although our proposed model regarding the role of glucuronidation in predicting different toxicities remains to be proven, we suggest that caution be applied in implementing clinical and regulatory precedence regarding low activity UGT alleles as general predictors for toxicity.