Gilbert's Syndrome and Irinotecan Toxicity: Combination with UDP-Glucuronosyltransferase 1A7 Variants Increases Risk

Irinotecan (camptothecin) represents a standard treatment option for metastatic colorectal cancer and other solid tumors (1, 2). The prodrug irinotecan is converted to 7-ethyl-10-hydroxycamptothecin (SN-38) by ubiquitously expressed carboxylesterases. SN-38 has a 100-fold higher antitumor activity (Fig. 1A; ref. 3). SN-38 is mainly inactivated by UDP-glucuronosyltransferases (UGT), including UGT1A1 and UGT1A7, which generate SN-38 glucuronides that are eliminated via bile and urine (4, 5). Irinotecan has a narrow therapeutic range and leads to side effects, such as myelosuppression (mainly leukopenia and thrombocytopenia) and diarrhea (6), observed in 29% to 44% of patients (1) frequently resulting in the necessity to discontinue or lower the dose of the drug. Based on irinotecan detoxification by UGTs, these side effects have been linked to genetic variations of UGT activity present in Gilbert's syndrome (UGT1A1*28; refs. 5, 7-9).

The UGT family of enzymes represents a critical metabolic pathway for the glucuronidation of hydrophobic endobiotics, such as bilirubin, bile acids, and steroid hormones as well as xenobiotics, which include many therapeutic drugs and environmental carcinogens (10). An increasing number of functional single nucleotide polymorphisms (SNP) have been identified at the human UGT1A gene locus with a potential relevance for drug therapy (Fig. 1B; refs. 9, 11-15). Most prominently, Gilbert's syndrome has been characterized to be associated with the UGT1A1*28 allele, which harbors a TA insertion into the A(TA)₆TAA promoter reducing transcription to 30% of wild-type (11).

However, regarding the predisposition for irinotecan toxicity by UGT1A1*28, conflicting results have been reported. In one study with 66 patients, UGT1A1*28 variants were not found to predict all cases of irinotecan-associated neutropenia (9). Rouits et al. showed in a cohort of 75 patients that UGT1A1*28 had an effect on irinotecan-induced neutropenia but not on gastrointestinal toxicity, whereas Marcuello et al. observed in 95 irinotecan-treated patients a significant association between UGT1A1*28 and diarrhea but not neutropenia (16, 17). Moreover, Carlini et al. failed to observe an association between irinotecan-induced side effects and UGT1A1*28 and conversely reported an opposing trend based on six patients with Gilbert's syndrome without any observed adverse event (18). In combination, these data raise the question whether the genetic disposition for irinotecan-induced toxicity may be more complex than believed previously.

The UGT1A7 protein tract exhibits a 5-fold higher specific activity with SN-38 than does UGT1A1 (4, 14). SNPs at the UGT1A7 gene locus alter enzyme activity and/or transcription (12, 14, 15). Recently, it was shown in Gunn rats that intestinal UGTs rather than hepatic UGTs appear to be critical for irinotecan-induced toxicity (19). Reported analyses of the UGT1A7 gene sequence have identified three SNPs in the coding region at base pair positions 129, 131 (UGT1A7 N129K/R131K), and 208 (W208R), which result in functionally altered UGT1A7 protein variants with a reduced glucuronidation activity (Fig. 1B; refs. 12, 15). In addition, UGT1A7-57T/G, a SNP in the TATA box of the UGT1A7 gene, leads to a 70% reduction of transcriptional activity (14). UGT1A7-57T/G was found to occur in linkage disequilibrium with W208R, indicating that patients exhibiting the UGT1A7-57G variant not only have a reduced transcriptional activity but also a UGT1A7 enzyme with a reduced catalytic function. Moreover, SNP association studies showed the presence of homozygous UGT1A7-57G variants in up to 75% of patients with Gilbert's syndrome (14).

However, UGT1A7 gene polymorphisms in the UGT1A7 first exon alone were not found to be associated with irinotecan toxicity in Japanese patients (20). In view of these findings, the contribution of SNPs, which not only reduce transcription but also alter catalytic activity of the expressed human UGT1A7 protein, was analyzed together with UGT1A1*28 in irinotecan-treated patients with metastatic colorectal cancer to improve the prediction of a predisposition to irinotecan toxicity.

The reported analysis represents an ancillary assessment of 105 blood samples of patients that had been enrolled in a phase III treatment trial (Fluorouracil-Folic Acid-Irinotecan-Eloxatin trial, conducted by A.S. and V.H. in Munich and T.Z. in Augsburg, Germany) evaluating treatment with irinotecan (CPT-11) plus oxaliplatin versus CPT-11 plus 5-fluorouracil and folic acid. The aim of this study was to assess adverse events, dose reductions, and clinical outcome. In both treatment arms, 80 mg/m² CPT-11 was administered i.v. over 30 min on days 1, 8, 15, 22, 29, and 36, and the cycle repeated after day 50. In arm A of the Fluorouracil-Folic Acid-Irinotecan-Eloxatin study, folic acid (500 mg/m², 120-min infusion) and 5-fluorouracil (2,000 mg/m², 24-h infusion) were coadministered on days 1, 8, 15, 22, 29, and 36; in arm B, oxaliplatin (85 mg/m², 120 min infusion) were coadministered on days 1, 15, and 29.

At the beginning of each treatment cycle, adverse events and clinical symptoms were documented. Patients with known Gilbert's syndrome were excluded from this study. The following adverse events were analyzed: diarrhea within 24 h and later than 24 h after administration of CPT-11, loss of body weight of >5%, anemia, thrombocytopenia, and leukopenia. The intensity of adverse events was classified according to the WHO Adverse Reaction Terminology with scores from 0 (“no adverse events”) to 4 (“life-threatening adverse events”). If the CPT-11 dose had to be reduced to <80%, dose reduction was documented at each cycle.

Dose reduction criteria for both treatment arms were assessed according to the National Cancer Institute Common Toxicity Criteria classification, including dose reduction in the presence of severe leukopenia (grade 4), thrombocytopenia (grade ≥3), diarrhea (grade ≥3), mucositis (grade ≥3), or constipation of >96 h (grade ≥3). Tumor response was evaluated according to the UICC classification. Complete remission was defined as the disappearance of all measurable tumor manifestations for a minimum period of at least 8 weeks. Partial remission was defined as a decrease of ≥50% decrease in the sum of the products of the largest perpendicular diameters of measurable lesions, provided that other new lesions had not appeared. No change (stable disease) was documented if no change in tumor dimensions were detected. Progressive disease was defined as a 25% increase in the size of any measurable lesions or appearance of new lesions.

Genomic DNA was isolated from whole blood of patients (15) and genotyped after written informed consent was obtained. The study was approved by the local ethics committees.

All genotyping studies were done in a blinded fashion (T.O.L. and C.P.S.) without prior knowledge of the clinical data of the analyzed patient. Genomic DNA (∼10 ng) was used as a template in TaqMan 5′-nuclease assays. Primers and probes specific for each SNP were designed with Primer Express software (Applied Biosystems) and labeled with either 6-FAM or VIC as reporter dyes and MGB-NFQ (Applied Biosystems) as a quencher as described previously (14). The TaqMan assays were done using 600 nmol/L primer concentrations and 200 nmol/L probe concentrations (Applied Biosystems) and qPCR MasterMix Plus (Eurogentec). The run consisted of a hot start at 95°C for 10 min and 35 cycles of 94°C for 15 s and 61°C for 1 min. All assays were done in 25 μL reactions in 96-well trays using an ABI 7000 instrument (Applied Biosystems). UGT1A1*28 was determined by Oncoscreen (P. Häussler and W. Schwabe) and as reported previously (13).

Patients were classified according to the number of SNPs determined at the UGT1A1*28, UGT1A7-57T/G, and UGT1A7 N129K/R131K gene loci. In the absence of allelic variants, patients were classified as “low risk,” whereas high-risk patients were those individuals with at least one allele of the allelic variant at each analyzed gene locus. If only one or two allelic variants were detected at the UGT1A1 or UGT1A7 gene loci, patients were classified as “intermediate risk.” In nine patients, the UGT1A1*28 genotyping was not possible for lack of sufficient available DNA. Homozygous UGT1A1*28 variants were rare, and heterozygous as well as homozygous UGT1A1*28 variants were summarized in a single group. The risk groups were compared by multivariate analysis (nonparametric Weil-Lachin procedure) for differences in intensity of five adverse events simultaneously.

The Mann-Whitney test was used to quantify the difference between the risk groups. Frequency and intensity of the adverse events were also compared between the risk groups defined above by Cochran-Mantel-Haenszel statistics using modified ridit scores and adjusted for the randomized underlying chemotherapy (CPT-11 + 5-fluorouracil/folic acid versus CPT-11 + oxaliplatin). To evaluate whether classification of variants in UGT1A1 or UGT1A7 genes represented an independent risk factor for the occurrence of adverse events, logistic regression was calculated which included the underlying chemotherapy and other risk factors present at baseline-adjuvant pretreatment, performance status, and lactate dehydrogenase. Odds ratios and 95% confidence intervals are depicted. Association between genotype variants and overall survival time as well as time to tumor progression were calculated using a Cox proportional hazards model, including the same risk factors at baseline as in the logistic regression analysis. The analysis of incidence and intensity of adverse events was based on the documented 297 treatment cycles involving all 105 patients. The χ² test was used for statistical analysis of CPT-11 dose reductions. P values were not adjusted for multiple testing and therefore do not reflect results of hypothesis testing but show the magnitude of associations. All statistical analyses were calculated by an independent statistician who was not involved in patient treatment or genotyping (Estimate GmbH; T.Z.).

First, UGT1A1*28 polymorphisms were analyzed in all 105 irinotecan-treated patients and correlated with the presence of adverse events including diarrhea, anemia, thrombocytopenia, or leukopenia. Patients with UGT1A1*28 wild-type status were compared with heterozygous or homozygous UGT1A1*28 carriers. This analysis detected no significant association between patients with UGT1A1*28 alleles and irinotecan-induced toxicity based on 273 treatment cycles (Table 1). Furthermore, dose reductions to <80% were not associated with the individual presence of UGT1A1*28 alleles in irinotecan-treated patients.

The UGT1A7 enzyme exhibits a high specific activity for SN-38. We therefore did genotyping analysis by TaqMan allelic discrimination for UGT1A7 N129K/R131K and UGT1A7-57T/G. Again, patients with UGT1A7 wild-type variants were compared with patients carrying at least one allelic UGT1A7 variant. The individual presence of the UGT1A7 N129K/R131K allelic variant was associated with thrombocytopenia and diarrhea (<24 h) under irinotecan chemotherapy as well as with CPT-11 dose reductions. Moreover, UGT1A7-57 variants were associated with anemia under irinotecan chemotherapy.

In a second analysis, the study cohort was grouped according to the genotyping data into low-risk, intermediate-risk, and high-risk groups according to the presence of the three different UGT1A1 and UGT1A7 gene SNPs (Table 2A) to allow for a genotype-based comparison of different risk groups. This procedure identified 54 (51.4%) patients with a high-risk genotype combining variants of the UGT1A1 and UGT1A7 genes, 35 (33.3%) patients with an intermediate-risk type, and 16 (15.2%) patients with a low-risk type, respectively. The clinical characteristics of the different risk groups of this study are given in Table 2B, and the association of adverse events is shown in Table 3. When the high-risk group was studied in comparison with those patients with a low-risk genotype, the overall incidence of adverse events was significantly higher (P = 0.0035, Mann-Whitney test: odds ratio, 0.5511; 95% confidence interval, 0.5169-0.5854). Specifically, an association of thrombocytopenia and leukopenia with the high-risk genotype group was observed (Table 4), whereas the prediction of early and late diarrhea, weight loss, and anemia did not reach significance. In the low-risk genotype group, WHO grade 1 thrombocytopenia (Table 3) was the highest grade of this adverse event observed. Conversely, grade 2 and 3 thrombocytopenia was detected only in those patients exhibiting the intermediate-risk and high-risk genotypes. Similarly, severe leukopenia of WHO grade 3 and 4 was only observed in the intermediate-risk and high-risk groups, indicating that leukopenia as well as thrombocytopenia are more prevalent in irinotecan-treated patients with variants of both UGT1A1 and UGT1A7 (Table 3).

In addition, dose reductions to <80% had to be implemented more often in individuals exhibiting a high-risk UGT genotype when compared with patients with the low-risk group (based on 252 cycles; Tables 3 and 4). Furthermore, the analysis of risk factors by logistic regression to evaluate the odds ratios and 95% confidence intervals showed that the high-risk group was found to represent a significant independent predictor for early-onset diarrhea (Table 5). The tumor response was examined in all genotyping risks groups, but no association between time to tumor progression or overall survival and risk groups was detected (data not shown).

Gilbert's syndrome is a well-recognized risk factor for the development of irinotecan-associated drug toxicity, which is supported by the fact that UGT1A1 is capable of SN-38 glucuronidation and thus detoxification (9). However, reports do not uniformly find a significant association of the UGT1A1*28 variant responsible for Gilbert's syndrome with irinotecan side effects such as anemia, thrombocytopenia, leukopenia, and diarrhea (16, 17, 20).

The cohort of 105 irinotecan-treated patients in this study was therefore first analyzed for the presence of UGT1A1*28 alleles and their association with adverse drug side effects. This analysis failed to identify a significant association in these patients. Therefore, our data derived from one of the largest pharmacogenomic study cohorts of irinotecan-treated individuals to date corroborate data from different studies that have failed to find hematologic or gastrointestinal drug toxicity in patients carrying the UGT1A1*28 allele and suggest that additional risk factors may play a permissive role. Thus, in an attempt to expand the risk assessment strategy, a second analysis was done aimed at genetic variants of the UGT1A7 gene, which has been shown to exhibit a high specific activity with the active irinotecan metabolite SN-38 exceeding that of UGT1A1 (4, 14). UGT1A7 genotyping included a coding exon 1 variant (N129K/R131K; ref. 15) and a functional promoter variant (-57T/G) of the UGT1A7 gene (14), both of which have been shown previously to lead to a reduced function of UGT1A7 protein, or a 70% reduction of UGT1A7 gene transcription, respectively.

In contrast to the analysis of UGT1A1*28, an association of adverse side affects was already detected with the UGT1A7 gene markers alone, indicating that the UGT1A7 enzyme may play an important role in the detoxification of SN-38, possibly superior to that of UGT1A1. To further elucidate whether combined markers may additionally influence the prediction of irinotecan toxicity, we then stratified the patients into three genetically defined risk groups based on the presence of UGT1A1 as well as UGT1A7 SNPs. At first sight, the resulting high number of variants leading to 51.4% of patients in the high-risk group appears unexpected. However, previous reports have documented an association of the UGT1A7*3 genotype, which encompasses N129K/R131K, with colorectal cancer (15, 21), and the genotyping results found here are in agreement with an association of UGT1A7 SNPs with this disease. The classification of these genetically defined risk groups revealed patients at greater risk for irinotecan-induced toxicity. The overall incidence of adverse events was observed to be significantly higher in high-risk group patients compared with those in the low-risk group.

Primarily, leukopenia and thrombocytopenia are predicted by the combination of UGT1A1 and UGT1A7 polymorphisms. As would be clinically expected, the rate of dose reductions to <80% was higher in individuals characterized by the high-risk genotype, which shows that the risk associated with a combined genotype involving both UGT1A1 and UGT1A7 was associated with drug toxicity leading to significant treatment consequences in the affected individuals. Furthermore, the high-risk group was a significant independent predictor for the early-onset diarrhea, which is cholinergic in nature and may be life threatening in some patients. However, the analyses presented in Tables 1 and 3 regarding diarrhea cannot rule out the contribution of additional mechanisms outside reduced SN-38 glucuronidation.

From the perspective of drug metabolism, the pharmacogenetic association of more than one genetic UGT variant with irinotecan toxicity elucidated in this study is plausible. SN-38 undergoes glucuronidation catalyzed by several UGT1A proteins, specifically by UGT1A1 and UGT1A7 (4, 14). By determining the three markers UGT1A1*28, UGT1A7 N129K/R131K, and UGT1A7-57T/G in this study, the group of Gilbert's syndrome patients (UGT1A1*28) that has been recognized previously as a major risk group for irinotecan toxicity was more precisely defined as evidenced by the lack of association of UGT1A1*28 with irinotecan toxicity in this study cohort. Based on our data, the contribution of UGT1A7 variants to irinotecan toxicity is greater.

A recent analysis has found that 75% of patients homozygous for UGT1A1*28 additionally exhibit UGT1A7 variants (14). A positive association with UGT1A1*28 is therefore likely to be detected in individual cohorts because a majority of Gilbert's syndrome patients with UGT1A1*28 also exhibit UGT1A7 N129K/R131K and UGT1A7-57T/G. Our detailed analysis further indicates that UGT1A7 N129K/R131K and UGT1A7-57T/G are differentially associated with individual components of toxicity (diarrhea, anemia, and thrombocytopenia; see Table 1). These data may help to explain the controversial results obtained in other studies analyzing UGT1A1*28 alone and emphasize that the combination of markers shown here significantly improves the prediction of irinotecan-associated drug toxicity based on the role of UGT1A7 variants. As UGT1A1 is involved in the glucuronidation of bilirubin and estradiol as endogenous substrates and xenobiotics including buprenorphine, N-hydroxy PhIP, hydroxybenzo(a)pyrenes, simvastatin, gemfibrozil, and valproate, and UGT1A7 glucuronidates mutagenic polyaromatic hydrocarbons, patients in the high-risk group are likely to be at an increased risk for side effects for these UGT1A1- and UGT1A7-specific compounds.

In the present study, colorectal cancer patients with known Gilbert's syndrome had initially been excluded from the protocol. Our data nevertheless identified among the included patients previously undiagnosed cases of Gilbert's syndrome, which further emphasizes the utility of pharmacogenetic testing before irinotecan therapy. The inclusion of UGT1A7 markers refines the detection of potential irinotecan-associated risks and appears to be superior to determination of UGT1A1 variants alone.