CANCER PHARMACOGENETICS: POLYMORPHISMS, PATHWAYS AND BEYOND

Transcription

1 CANCER PHARMACOGENETICS: POLYMORPHISMS, PATHWAYS AND BEYOND Cornelia M. Ulrich*, Kim Robien* and Howard L. McLeod Inherited genetic variations can affect a patient s response to chemotherapeutic agents given for cancer. Pharmacogenetics aims to use knowledge of these variations to tailor therapy for improved response and reduced toxicity. Most research so far has focused on single polymorphisms. A more comprehensive approach to predict treatment response will be to consider genetic variation in entire biological and pharmacological pathways. Of particular relevance to cancer chemotherapy is folate metabolism, which is the target of methotrexate and 5-fluorouracil. Furthermore, efforts have begun to construct pathways of genes that have pharmacological relevance for individual chemotherapeutic agents. Together, these pathway strategies offer a higher likelihood of achieving the promise of genetically guided cancer therapy. THERAPEUTIC INDEX The ratio of the median lethal dose to the median effective dose for a given medication. Used to describe the dose range over which a drug has a therapeutic effect without unacceptable toxicity. POLYMORPHISM Variation within a gene (often at a single nucleotide) where two or more alleles exist at a frequency of at least 1% in the general population. * Fred Hutchinson Cancer Research Center, Cancer Prevention Research Program, Seattle, Washington 98109, USA. Washington University, Department of Medicine, St Louis, Missouri 63130, USA. Correspondence to C.M.U. nulrich@fhcrc.org doi: /nrc1233 With increasingly comprehensive information available from the Human Genome Project, pharmacogenetics the science of incorporating information on inherited genetic variability into predicting treatment response is flourishing. A better understanding of pharmacogenetic associations is especially important in cancer chemotherapy, as many chemotherapeutic agents given for cancer are characterized by a narrow THERAPEUTIC INDEX 1. POLYMORPHISMS in both the individual s genome, as well as the tumour genome, will affect drug response 1 tumours are expected to be of the same genetic makeup with respect to specific polymorphic sites as somatic tissue, unless new mutations have occurred or the site is subject to chromosomal loss. Yet, drug-related toxicity almost exclusively depends on the genotype of nontumour tissue 1. So, inherited polymorphisms will have a key role with respect to toxicity, a crucial dose-limiting factor in most cancer chemotherapy regimens. Research that has used single polymorphisms as markers of variation in treatment response has been promising, but has lead to some conflicting results. Many of the variant forms of drug-metabolizing enzymes show only a small deviation from wild-type enzyme activity, and are part of polygenic metabolic or pharmacological pathways. It is quite rare that a single polymorphism results in significant changes in the ability to metabolize drugs. However, it is possible that the combination of several polymorphisms in components of a biological pathway or pharmacological pathway might significantly influence therapeutic response. FOLATE metabolism has been the target for several key chemotherapeutic agents, and is an example of a biological pathway with pharmacogenetic relevance 2. This review will illustrate the relative contribution of genetic variation in individual candidate genes within a biological pathway to patient toxicity and therapeutic outcome, and show how genetic variability in both biological and pharmacological pathways can affect treatment response. In addition, novel strategies for pharmacogenetic discovery will be described as we try to move beyond the limitations of our current understanding of the genetic mechanisms that regulate drug activity. Single-gene/single-variant pharmacogenetics The classical approach to identifying a genetic basis for extreme drug toxicity or aberrant drug effect has served us well during the initial phase of pharmacogenetic research; a clinical phenotype such as altered pharmacokinetics or extreme toxicity has been used to identify PROBANDS, followed by evaluation of genetic 912 DECEMBER 2003 VOLUME 3

2 Summary The field of pharmacogenetics attempts to use genetic information to predict an individual s drug response. It is especially important in cancer chemotherapy given the narrow therapeutic index of these drugs. So far, pharmacogenetic research has largely focused on the effect of single candidate polymorphisms. However, many of the genetic variants that are associated with extreme drug toxicity are rare and explain only a small portion of the variation seen in drug response. Understanding the interactions of genetic variants within a biological or pharmacological pathway will allow for an improved ability to predict drug response. Folate metabolism a target of antifolate chemotherapeutic agents and thymidylate-synthase inhibitors is a biological pathway of substantial interest to pharmacogenetic researchers. Pharmacological pathways are being constructed for the systematic evaluation of the genes that regulate variation in the toxicity and efficacy of anticancer agents. Mouse models show promise in identifying key enzymes in pharmacogenetic pathways and will allow study of genetic variation in these pathways. FOLATE One of the B vitamins. The primary role of this B vitamin is as a carrier of methyl groups, especially for purine, pyrimidine and methionine synthesis. PROBAND An individual with the condition of interest, who serves as the starting point for exploration of a family pedigree for the genes that are responsible for the condition. CYTOCHROME-P450 ENZYMES A family of haem-containing intracellular oxidizing enzymes that are responsible for the first phase of metabolism of many drugs and other ingested toxins. THYMIDYLATE-SYNTHASE INHIBITORS A category of chemotherapeutic agents, including 5-fluorouracil, which tightly bind and inhibit the activity of thymidylate synthase, and therefore prevent DNA replication. polymorphisms, for example in the drug-metabolizing enzyme that is responsible for the degradation of the drug. These genetic polymorphisms include nucleotide repeats, deletions, insertions and mutations that influence gene expression and/or function 3. Genetic variants within a specific candidate gene provide the mechanistic basis for many of the early examples in pharmacogenetics. For instance, mutations or deletions in the CYTOCHROME P450 enzyme CYP2D6 occur in the general population and about 10% have been associated with a poor-metabolizer phenotype 4. This is important for codeine and related pain medications, as the activation of codeine to morphine is dependent on the catalytic function of CYP2D6 (REF. 5). Patients with low CYP2D6 activity do not receive pain relief from codeine or other analogues, because they are not able to form morphine 5. As another example, polymorphisms in thiopurine methyltransferase and dihydropyrimidine dehydrogenase have been associated with altered drug metabolism and increased risk of severe toxicity from the anticancer agents 6-mercaptopurine and 5-fluorouracil, respectively In addition, a variant number of dinucleotide-repeat sequences (5 8 repeats) in the promoter for uridine 5 -diphosphateglucuronosyltransferase 1A1 (UGT1A1) has an influence on in vitro and in vivo glucuronidation of SN-38, the active metabolite of irinotecan 11. Patients with seven repeat sequences have a fourfold relative risk of experiencing severe toxicity after treatment with irinotecan, including grade III/IV diarrhoea and neutropaenia, compared with patients with six repeat sequences 12,13. These examples of altered drug metabolism from genetic polymorphisms have recently been complemented with studies that associate tumour response or patient survival with polymorphisms in membrane transporters (such as ABCB1 in patients with acute myelogenous leukaemia 14 ) and excision repair enzymes (such as ERCC2 or XRCC1 in patients with colorectal cancer 15,16 ). However, many of the genetic variants that are associated with extreme drug toxicity are uncommon, and therefore explain only a small proportion of the population variance that is seen in drug response. In addition, it is recognized that most drug effects are polygenic in nature. Genetic polymorphisms have been identified in 93% of all known genes, with two coding-region single-nucleotide polymorphisms (SNPs) observed in most genes that have been evaluated so far 17,18. The initial SNP map from the Human Genome Project discovered 1.42 million variants, with the public databases, which are growing rapidly, now containing over 5 million human SNPs. Research on the impact of common genetic variants on drug response is just beginning, and will constitute a considerable research effort over the next years. The biological-pathway approach Most enzymes function in complex networks with several regulatory mechanisms. So, it is unlikely that any one variant with modest effects on enzyme function will affect disease or treatment outcomes, whereas the combination of several variants within the same pathway might result in significant disturbances. Alternatively, several genetic variants might cancel each other out and diminish differences in the response that is associated with a single polymorphic allele. An example of a biological pathway with relevance to cancer chemotherapy is folate metabolism. Folate metabolism and antifolates. Folate acts as a donor for methyl groups, such as in the conversion of homocysteine to methionine a precursor for the universal methyl donor S-adenosyl-methionine and in the synthesis of purines and pyrimidines 19 (FIG. 1). Therefore, folate has a key role in normal cell growth and replication and has been an attractive target for chemotherapeutic agents. Antifolate agents are drugs that target specific enzymes in folate metabolism (for example, dihydrofolate reductase) and are used for the treatment of various cancers, including haematological, colorectal, breast and pancreatic malignancies. Methotrexate and 5-fluorouracil are folate-pathway inhibitors that have been available for 50 years and continue to be used in various treatment regimens for cancer and autoimmune diseases. Their specific mechanisms of action have been reviewed previously 20,21. Folate analogues are structurally similar to folate but are able to inhibit the action of various enzymes in folate metabolism (in the case of methotrexate, primarily dihydrofolate reductase) Fluorouracil is a fluoropyrimidine that, on activation to the nucleotide form, develops a stable complex with thymidylate synthase, which inhibits the activity of the enzyme 20,25. Capecitabine (Xeloda ) is an orally administered fluoropyrimidine that is subsequently activated to fluorouracil 26. The therapeutic effectiveness and cytotoxicity of these drugs are largely attributed to the role of folate in nucleotide synthesis both folate antagonists and NATURE REVIEWS CANCER VOLUME 3 DECEMBER

3 Cell membrane Regulatory enzyme C677T: TT genotype associated with greater toxicity of MTX and possibly 5-FU A1298C: Common polymorphism affecting protein function MTHFR Pyrimidine synthesis 5-FU dump hfr Serum folate 5-Methyl THF 5, 10-Methylene THF Thymidylate synthase dtmp RFC Drug transporter G80A: AA genotype associated with higher MTX serum levels, and worse prognosis among MTX-treated children Methionine synthase SHMT THF 10-Formyl THF Drug target TSER polymorphism: the 2rpt/2rpt genotype responds better to 5-FU treatment, but with greater toxicity. Possible association with survival among MTX-treated patients Homocysteine Methionine GART AICARFT GAR AICAR DHF CBS Purine synthesis Cystathionine SAH SAM DHFR Cysteine CH 3 X X DNA methylation Methotrexate Figure 1 Folate metabolism and related pathways. This simplified figure illustrates the interconnectedness of folate metabolism and proteins for which functional polymorphisms have been identified. Polymorphisms have been found that are associated with pharmacogenetic outcomes in three key proteins in these pathways: the drug transporter protein reduced folate carrier (RFC); the regulatory enzyme 5,10-methylenetetrahydrofolate reductase (MTHFR); and the drug target thymidylate synthase. Key enzymes are denoted as ovals, substrates as rectangles. Red ovals denote enzymes with genetic polymorphisms that have been investigated in pharmacogenetic studies. Orange ovals denote enzymes for which functional genetic polymorphisms have been described. 5-FU, 5-fluorouracil; AICAR, 5-aminoimidazole-4-carboxamine ribonucleotide; AICARFT, AICAR formyltransferase; CBS, cystathionine-β-synthase; DHF, dihydrofolate; DHFR, DHF reductase; dtmp, deoxythymidine monophosphate; dump, deoxyuridine monophosphate; GAR, glycinamide ribonucleotide; GART, phosphoribosylglycinamide formyltransferase; hfr, human folate receptor; MTX, methotrexate; SAH, S-adenosylhomocysteine; SAM, S-adenosylmethionine; SHMT, serine hydroxymethyltransferase; THF, tetrahydrofolate; X, various substrates for methylation. MUCOSITIS Inflammation, irritation and ulceration of the mucosal membranes, especially of the oral cavity and gastrointestinal tract, that occurs as a result of cytotoxic chemotherapy or radiation therapy. MYELOSUPPRESSION Suppression of blood-cell formation in the bone marrow. TRANSCRIPTIONAL-ENHANCER ELEMENT A region of DNA that might be several thousand base pairs upstream from a given gene s promoter region, but can bind with gene-regulatory proteins and increase the rate of transcription of that gene. THYMIDYLATE-SYNTHASE INHIBITORS reduce proliferation of neoplastic cells. In situations of folate deficiency, insufficient conversion of deoxyuridine monophosphate to deoxythymidine monophosphate can result in uracil misincorporation into DNA strands during replication, subsequently resulting in double-strand breaks Activated metabolites of fluoropyrimidines inhibit the activity of thymidylate synthase and are also incorporated into both DNA and RNA, which results in further DNA damage and interference with RNA synthesis and function 20. Since folate is needed for DNA synthesis and repair, rapidly dividing normal tissues such as the gastrointestinal tract and haematopoietic cells are the main sites of toxicity, with common side effects including oral and intestinal MUCOSITIS and MYELOSUPPRESSION 20,21,31,32. Folate metabolism and genetic variants. Recently, it has become evident that there is large inter-individual variability in the mechanisms to maintain homeostasis within folate metabolism because of genetic polymorphisms in key enzymes (reviewed in REF. 2). Genetic variants in thymidylate synthase, 5,10-methylenetetrahydrofolate reductase (MTHFR) and the reduced folate carrier (RFC) have been investigated with respect to clinical outcomes (FIG. 1; TABLE 1). The 5 -untranslated region of the thymidylatesynthase gene functions as a cis-acting TRANSCRIPTIONAL- ENHANCER ELEMENT and contains polymorphic 28-basepair tandem repeats 33,34. The presence of a triple repeat (3rpt) results in approximately 2 4-fold greater gene expression compared with a double repeat (2rpt) 35,36. The triple-repeat homozygote (3rpt/3rpt) genotype is 914 DECEMBER 2003 VOLUME 3

4 associated with lower plasma folate concentrations and, among individuals with low dietary folate intake, increased plasma homocysteine concentrations 37.The 3rpt allele might protect against the development of acute lymphocytic leukaemia 38, and is associated with a reduced risk of colorectal adenomas in the presence Table 1 Polymorphisms in folate metabolism and response to cancer drugs Polymorphism Study population Findings References TSER 28-base-pair repeats Tumour tissue from 50 patients with 2rpt/2rpt individuals (n = 11) had 36 metastatic colorectal cancer receiving significantly greater toxicity 5-fluorouracil 65 patients with rectal cancer 3rpt/3rpt individuals (n = 27) had a 77 receiving pre-operative 5-fluorouracil lower probability of tumour downstaging compared with 2rpt/2rpt individuals (n = 13) Tumour tissue from 221 patients with No association between TSER polymorphism 59 Dukes C colorectal cancer receiving and survival. Similar effects of 5-fluorouracil either surgery alone or surgery and treatment seen among each of the TSER 5-fluorouracil genotypes 24 patients with metastatic colorectal 3rpt/3rpt homozygous individuals had 58 cancer receiving 5-fluorouracil decreased median survival 205 children with acute lymphoblastic 3rpt/3rpt homozygous children (n = 50) had 68 leukaemia treated with methotrexate significantly shorter event-free survival times compared with those with the other TSER genotypes 24 patients with metastatic colorectal Individuals with the 2rpt/2rpt genotype (n = 4) 78 cancer receiving capecitabine had a greater response rate compared with individuals with the 2rpt/3rpt (n = 12) or homozygous 3rpt/3rpt genotypes (n = 8) MTHFR C677T 51 patients with stage III colon No differences in survival based on genotype 61 cancer receiving 5-fluorouracil and 677TT (n = 4) leucovorin 6 patients with early breast cancer 5 of the 6 patients who experienced grade IV 66 receiving cyclophosphamide, toxicity (NCI-CTC) had the 677TT genotype methotrexate and 5-fluorouracil 220 patients with CML following Patients with 677TT genotype (n = 36) 62 haematopoietic-cell transplant receiving methotrexate experienced increased oral mucositis and somewhat slower platelet recovery 61 patients with acute leukaemia Increased methotrexate toxicity among 64 receiving maintenance chemotherapy patients with the 677TT genotype (n = 15) including methotrexate 6 children with acute lymphoblastic Lymphocytes from the two children with the 67 leukemia who had had methotrexate- 677TT genotype (n = 2) exhibited greater related neurotoxicity in vitro methotrexate sensitivity compared with those with the 677CC (n = 2) or 677CT genotypes (n = 2) 43 patients with ovarian cancer 677TT individuals (n = 13) had a significantly 65 receiving either methotrexate alone higher risk of experiencing grades III IV toxicity or in combination with carboplatin (WHO criteria), and had significantly higher plasma homocysteine levels after methotrexate treatment 43 patients with metastatic colorectal Individuals with the 677TT genotype (n = 17) 70 cancer receiving fluoropyrimidine were significantly more likely to respond to chemotherapy treatment than with those with the 677CT (n = 21) or 677CC genotypes (n = 5) Reduced folate carrier G80A 204 children with acute lymphoblastic Patients with the 80AA genotype (n = 45) 69 leukaemia treated with methotrexate had significantly shorter event-free survival compared with those with the 80GG genotype (n = 61); patients with the 80AA genotype had higher plasma methotrexate levels than those with other genotypes, suggesting decreased cellular uptake of methotrexate CML, chronic myeloid leukaemia; MTHFR, 5,10-methylenetetrahydrofolate reductase; NCI-CTC, National Cancer Institute Common Toxicity Criteria; rpt, repeat; TSER, thymidylate-synthase enhancer region; WHO, World Health Organization. NATURE REVIEWS CANCER VOLUME 3 DECEMBER

5 LOSS OF HETEROZYGOSITY The loss of one allele in a tumour cell for a gene in which the individual is normally heterozygous. HAZARD RATIO The comparison of the risk of experiencing an event at any given point in time (hazard) among those with a particular risk factor with those without the risk factor. GRAFT-VERSUS-HOST DISEASE A potentially life-threatening condition, which might occur following transplantation of solid organs or haematopoietic cells, in which donor T lymphocytes recognize host cells as foreign and attack host tissues. of high folate intake or a MTHFR 677TT genotype (see below) 39. A second common polymorphism in thymidylate synthase a 6bp deletion at base pair 1494 in the 3 -untranslated region 40 seems to be associated with decreased messenger RNA levels of the enzyme in colorectal tumours 41. MTHFR is a central regulatory enzyme in folate metabolism that diverts folate metabolites from pyrimidine synthesis towards methionine synthesis. A common functional polymorphism of the gene encoding this enzyme occurs at C677T (alanine222valine) 42. The 677TT genotype produces an enzyme with only 30% of the activity of the wild-type (677CC) enzyme in vitro and is associated with increased plasma concentrations of homocysteine. Several diseases have been associated with the MTHFR C677T polymorphism, including acute lymphocytic leukaemia 43,44, colorectal polyps and colon cancer 45 49, neural-tube defects 50,51 and, possibly, cardiovascular disease 52 ; increased risks are usually seen among individuals with a low folate status. A second common polymorphism in the MTHFR gene (occuring at A1298C; Glut429Ala) 53,54 also results in somewhat decreased MTHFR activity and has been associated with a reduced risk of certain forms of acute adult and paediatric leukaemias 43,44 and colon cancer 55. Furthermore, a polymorphism in the RFC gene (G80A, argenine27histidine) seems to be associated with a higher affinity for folate 56. Among 169 healthy individuals who were stratified by MTHFR C677T genotype, the variant A allele in RFC was consistently and linearly associated with higher plasma folate concentrations 56. Response to antifolates and thymidylate-synthase inhibitors. The variability in receptors, drug targets and other key regulatory enzymes within folate metabolism that is described above seems to trigger a differential response to antifolate chemotherapeutic agents and thymidylate-synthase inhibitors. Several recent studies show how polymorphisms in the folate pathway affect the treatment response to drugs that target this biological pathway (TABLE 1). The thymidylate-synthase promoter-enhancer region (TSER) polymorphism has been the most commonly investigated, usually in studies that evaluate response to or toxicity from 5-fluorouracil treatment. Among 50 patients with metastatic colorectal cancer who underwent therapy with 5-fluorouracil as a component of several different regimens, patients who were homozygous for the TSER-double-repeat genotype (2rpt/2rpt; associated with lower thymidylatesynthase expression) were more likely to respond to 5-fluorouracil therapy, but were also more likely to experience significant toxicity 36. It is of concern that genotyping in this study was undertaken with DNA that had been collected from tumour tissue rather than from somatic cells, which are the sites for drug toxicity. LOSS OF HETEROZYGOSITY can occur in tumour cells on chromosome 18 where TSER is found and the TSER-genotype frequencies among the patients who were evaluated for clinical outcomes were dissimilar to the distributions that were seen in other patients at the centre where the study was carried out or to those seen in other studies 36,57. Our group studied 24 patients who were treated with 5-fluorouracil and leucovorin chemotherapy for metastatic colorectal cancer as part of a pilot study 58. Tumour response, as measured after every four cycles of chemotherapy by computed tomography, was more likely among those who had the TSER 2rpt/2rpt genotype. Further investigations that are based on this pilot study are underway. Another study compared Australian patients with colorectal cancer who did not receive chemotherapy with patients treated with a cycle of 5-fluorouracilbased adjuvant chemotherapy once a month for six months 59. The survival benefits that were associated with 5-fluorouracil-based chemotherapy were similar among patients with the TSER 3rpt/3rpt genotype (HAZARD RATIO (HR) of the effect of chemotherapy on survival = 0.62) compared with those with the 2rpt/2rpt or 3rpt/2rpt genotype (HR = 0.52) The analyses did not account for other factors that might affect survival, such as the age of the patients or the treatment regimens that were used. Overall, findings regarding the TSER variant indicate that genotypes that are associated with lower thymidylate-synthase expression confer an increased sensitivity towards 5-fluorouracil therapy, although possibly at the expense of greater toxicity. However, most of these studies had limited sample sizes, which needs to be taken into consideration when interpreting the results. Another commonly investigated genetic variant is the C677T polymorphism in MTHFR (discussed above), which alters the activity of the key regulatory MTHFR enzyme in folate metabolism 60. A study among 51 patients with stage III colon cancer who were treated with 5-fluorouracil did not show a difference in survival by MTHFR genotype 61. However, the study included only four individuals with the MTHFR 677TT genotypes and lacked adjustment for other factors that are potentially related to survival. Our group investigated the toxicity of methotrexate in 220 patients with chronic myelogenous leukaemia who were undergoing haematopoietic-cell transplantation 62. These patients all received methotrexate as part of a standardized regimen with cyclosporine for the prevention of GRAFT-VERSUS-HOST DISEASE 63.The MTHFR 677TT genotype which corresponds to lower enzyme activity was associated with 36% higher oral mucositis scores compared with the 677CC genotype; patients with 677CT showed intermediately increased mucositis, consistent with the residual in vitro enzyme activity of the heterozygous genotype. Furthermore, recovery of platelet counts a sign of haematopoietic-cell engraftment after transplantation was 20 40% slower in patients with the MTHFR 677TT genotype 62. Two studies from Italy support the concept of increased methotrexate toxicity among patients with a MTHFR 677TT genotype. Among 61 patients with acute leukaemia who received maintenance chemotherapy 916 DECEMBER 2003 VOLUME 3

6 CPT-11 CES1 CES2 SN-38 SN-38 ABCB1 ABCG2 ABCC1 ABCC2 ADPRT CPT-11 CPT-11 SN-38 TOPI TDP1 ABCB1 CES1 CES2 XRCC1 CDC45L 3435C>T associated with higher drug clearance CYP3A4 UGT1A1 NFκB1 CYP3A5 Cell death APC NPC UGT1A1*28 allele associated with fourfold risk of severe toxicity SN-38G The interconnectedness of folate metabolism and the need to consider the entire pathway rather than individual polymorphisms is further illustrated by three recent investigations The findings of these studies strongly indicate that polymorphisms in proteins that are not direct drug targets of the chemotherapeutic agent used such as the MTHFR protein can affect treatment response to other drugs that target folate metabolism. Cohen et al. reported that the MTHFR genotype seems to alter response to 5-fluorouracil therapy; individuals with the MTHFR 677TT genotype (n = 5) were significantly more likely to respond to treatment than those with 677CT or 677CC genotypes 70. Furthermore, the TSER genotype seems to affect response among children who are treated with methotrexate 68. Among 205 children with acute lymphoblastic leukaemia who were treated with methotrexate induction and maintenance chemotherapy, the risk of having an event (such as relapse or death due to disease) varied by TSER genotype. Compared with children with a TSER 2rpt/2rpt genotype, those with a 3rpt/3rpt genotype were at increased risk (multivariate ODDS RATIO (OR) = 4.8). Within the same patient population, children with the RFC 80AA genotype had a significantly shorter eventfree survival than those with the 80GG genotype, indicating that the cellular uptake of methotrexate might have a role in the response 69. These studies point towards the need to consider the entire folate metabolism as a biological pathway when evaluating associations between genetic polymorphisms and treatment outcomes, rather then the current focus on individual components in isolation. Figure 2 Possible pathways of irinotecan metabolism. Irinotecan (CPT-11) can be converted into the active metabolite SN-38 by carboxylesterases (CES) outside or inside the cell. CPT-11 and SN-38 are both substrates for the ATP-binding cassette (ABC) transport proteins P-glycoprotein (ABCB), ABCC and ABCG which transport the drug out of the cell. Alternatively, CPT-11 and SN-38 can be inactivated by cytochrome P450 enzymes (CYP) or uridine diphosphate glycosyltransferase (UGT), respectively. If SN-38 persists, it binds to its target topoisomerase I (TOPI), interfering with DNA synthesis and repair processes, culminating in cell death. ADPRT, ADP-ribosyltransferase; APC, inactive metabolite of SN-38; CDC45L, cell-division cycle 45L; NPC, inactive metabolite of SN-38; SN-38G, SN-38 glucuronide; TDP, tyrosyl-dna phosphodiesterase; XRCC1, X-ray-repair cross-complementing defective-1. ODDS RATIO A comparison of the risk of experiencing a particular outcome (such as a disease or adverse drug reaction) between two groups of people one group with a particular risk factor (such as a genotype or a nutrient), and the other group without the risk factor being studied. TOPOISOMERASE I An enzyme that creates a transient single-strand break in the phosphodiester backbone of DNA, allowing the strand of DNA to uncoil and rotate freely during replication. including methotrexate, increased methotrexate intolerance was observed among patients with the 677TT genotype 64. Similarly, patients with ovarian cancer and the 677TT genotype who received methotrexate experienced significantly higher plasma homocysteine levels than those with other genotypes, as well as greater toxicity 65. Lastly, two case reports 66,67 provide further indication for greater methotrexate sensitivity and, possibly, significant toxic responses among those with a MTHFR 677TT genotype. Most of these studies are based on small sample sizes, which precludes robust conclusions and might result in publication bias. Nevertheless, the consistent findings of these studies regarding a link between decreased MTHFR activity (677TT genotype) and increased toxicity are compelling. The pharmacological-pathway approach A second approach to understanding pathways that are key to pharmacogenetics research involves the construction of pharmacological pathways. It is clear that although the single-candidate-gene approach has provided the foundation for cancer pharmacogenetics, evaluation of a gene in isolation as part of its normal biological pathway does not often provide sufficient information about variation in expected therapeutic response to justify dose modification. A logical next step is the construction of pathways of genes that are specifically of importance for the regulation of drug activity, with the use of a composite view of these pathways as a tool for individualized therapy. For drugs that mimic natural molecules, such as the antifolates, antipyrimidines, anti-oestrogens and others, there are pre-existing biological pathways that define the disposition and activity of the agents. However, many anticancer agents do not follow an existing biological pathway, but, rather, interact with a pharmacological pathway that consists of a diverse array of genes that would not have direct interplay in normal cell biology. The construction of these pharmacological pathways entails the use of knowledge regarding drug absorption, excretion, activation and other metabolic functions. This is illustrated in FIG. 2, for the TOPOISOMERASE I NATURE REVIEWS CANCER VOLUME 3 DECEMBER

7 Table 2 Various approaches to analyzing pharmacogenetic variation Approach Advantages Limitations Candidate polymorphism Quick and relatively inexpensive Only helpful in the clinical setting if the analysis Uses knowledge regarding drug polymorphic genotype corresponds to a pharmacodynamics and distinct phenotype (high sensitivity/specificity) pharmacokinetics Assumes extensive previous knowledge Hypothesis driven regarding what gene is likely to be important such as one that determines the excreted proportion of a drug Very limited approach Pathway approach Accounts for associations between Requires a solid biochemical understanding proteins in a metabolic pathway pharmacological or biological of the Uses knowledge regarding drug respective pathway metabolism and other meaningful Need to understand how the drug is biological associations (such as metabolized and its mechanisms of action regulatory function and homeostasis) Data analysis is more complex than evaluation More likely to explain inter-individual Requires large study sizes variation in drug response Hypothesis driven Genomics or proteomics Does not require a hypothesis Does not use information on known biological (microarray technology) Provides a complete gene- or protein- or pharmacological associations expression profile (tumour or individual) Optimal data management and data analysis Provides information on associations techniques are not well defined at this point that have not previously been suspected Expensive (expression arrays), not fully Provides a large amount of data developed (proteomics) or not currently Probably useful for prediction of practical in a clinical setting tumour response Not clear how useful for prediction of toxicity inhibitor irinotecan. This figure shows the complexity of factors that influence an individual drug, with a need for activation of irinotecan to SN-38; transporters causing efflux of both parent and metabolite; P450-mediated inactivation of parent drug; glucuronidation of the active metabolite; variation in the cellular therapeutic target; and then a series of death genes that mediate the ultimate fate of the cell. It is no surprise that this level of complexity dilutes the contribution of any one component of the pathway. The key issue now is the comprehensive evaluation of pharmacological pathways to identify the components that are rate-limiting to toxicity and antitumour effect. This will allow more focused polygenic studies to be conducted in clinical trials to provide validation of their worth in clinical decision-making. Pharmacogenetic-discovery strategies The recognition that drug effect is under the control of several genes or defined drug pathways has led to a need for pharmacogenetic-discovery experiments without the constraints of predetermined candidate genes. Pharmacogenetic discovery in the field of cancer chemotherapy is severely limited by the inability to perform classical human genetic screens. The use of familial genetic strategies to identify putative genes for predicting toxicity or drug response is not possible, because members of a given family do not acquire the same malignancy in the same window of time. It is also unethical to give cytotoxic chemotherapy to normal volunteer individuals, which prevents the acquisition of large data sets of unrelated individuals on which to perform genome-wide screens. Microarray studies will reveal some of the genes that are associated with drug effect, but these studies often include an insufficient number of samples to be definitive. One under-used tool is mouse genetics, in which ex vivo and in vivo strategies can be performed using data analysis approaches to discover genes that are associated with drug activity 71. The laboratory mouse and humans share a high degree of similarity in anatomy, physiology and genome content. By using well-defined inbred mouse strains, breeding strategies have been used to identify regions of the genome that are associated with a particular phenotype such as drug effect to localize the gene that contributes to the phenotypic variation and to discover the genetic variants that provide the mechanistic basis for the effect. A recent example of the power of mouse genomics in the context of cancer therapy was performed to identify the gene(s) that are associated with susceptibility to pulmonary fibrosis after bleomycin chemotherapy. By conducting a genome-wide scan in offspring that were produced from breeding a sensitive mouse strain with a resistant strain, two regions of the genome were associated with bleomycin-related lung fibrosis 72. The first locus was on mouse chromosome 17 and is a region that has been associated with pulmonary fibrosis from other causes (for example, radiation, ozone and particle exposure) and contains the candidate gene tumour-necrosis factor-α. The second locus on mouse chromosome 11 contains bleomycin hydrolase, an enzyme that detoxifies bleomycin, and a glycoprotein Sparc, which showed expression differences between susceptible and resistant mouse strains 72. Lessons learned from the past 25 years of mouse genetics have provided a rich resource of genomic information from the commonly available inbred strains and the knowledge of statistical genetics to rapidly apply this approach to pharmacogenetic questions. These pharmacogenetic-discovery strategies can be used to explore genetic variability in relevant 918 DECEMBER 2003 VOLUME 3

8 pathways. Concerted efforts are ongoing to take advantage of the power of mouse genetics, including knockout and transgenic techniques, to identify pharmacological pathways of relevance to the prediction of drug effect in individuals with cancer. Outlook Considering the narrow therapeutic index for many cancer chemotherapeutic agents, the ability to better predict response and potentially life-threatening toxic effects is highly important to the clinician. We expect that, in the future, pharmacogenetic research using the pharmacological-pathway approach and research using the genomic or proteomic approaches will complement each other (TABLE 2). The use of single candidate genes has been useful as part of initial investigations but it will most likely never provide the sensitivity and specificity that is needed for tailored treatment decisions. Recently developed -omic approaches, which use microarray technology to describe gene or protein expression in its entirety will certainly be useful in refining cancer diagnoses and, in turn, predicting tumour response to specific drugs However, these methods might be less helpful for predicting potential toxicity, since gene and protein expression in susceptible tissues are likely to differ once a drug is administered. Furthermore, genomic or proteomic methods do not take variations in protein function or stability into account, which, in the case of folate metabolism, have been shown to be associated with clinical outcome. Knowledge regarding existing pathways, both biological and pharmacological, has been assembled over several decades and can be used for targeted pathway approaches. We propose that future pharmacogenetic research considers entire pathways, their regulatory mechanisms and homeostasis, and evaluates genetic variability in a given pathway comprehensively. 1. Relling, M. V. & Dervieux, T. Pharmacogenetics and cancer therapy. Nature Rev. Cancer 1, (2001). 2. Ulrich, C. M., Robien, K. & Sparks, R. Pharmacogenetics and folate metabolism: a promising direction. Pharmacogenomics 3, (2002). 3. Watters, J. W. & McLeod, H. L. Cancer pharmacogenomics: current and future applications. Biochim.Biophys. Acta 2, (2003). 4. Rogers, J. F., Nafziger, A. N. & Bertino, J. S. Jr. Pharmacogenetics affects dosing, efficacy, and toxicity of cytochrome P450-metabolized drugs. Am. J. Med. 113, (2002). 5. Sindrup, S. H. & Brosen, K. The pharmacogenetics of codeine hypoalgesia. Pharmacogenetics 5, (1995). 6. Relling, M. V. et al. Mercaptopurine therapy intolerance and heterozygosity at the thiopurine S-methyltransferase gene locus. J. Natl Cancer Inst. 91, (1999). Lowering the dose of mercaptopurine allowed children with the thiopurine S-methyltransferase enzyme deficiency to complete the course of chemotherapy and maintain sufficient erythrocyte thioguanine nucleotide concentrations. 7. Evans, W. E. et al. Preponderance of thiopurine S-methyltransferase deficiency and heterozygosity among patients intolerant to mercaptopurine or azathioprine. J. Clin. Oncol. 19, (2001). 8. McLeod, H. L. & Siva, C. The thiopurine S-methyltransferase gene locus: implications for clinical pharmacogenomics. Pharmacogenomics. 3, (2002). 9. Milano, G. & McLeod, H. L. Can dihydropyrimidine dehydrogenase impact 5-fluorouracil-based treatment? Eur. J. Cancer 36, (2000). 10. Mattison, L. K., Soong, R. & Diasio, R. B. Implications of dihydropyrimidine dehydrogenase on 5-fluorouracil pharmacogenetics and pharmacogenomics. Pharmacogenomics 3, (2002). 11. Iyer, L. et al. Phenotype genotype correlation of in vitro SN-38 (active metabolite of irinotecan) and bilirubin glucuronidation in human liver tissue with UGT1A1 promoter polymorphism. Clin. Pharm. Ther. 65, (1999). Individuals who were homozygous for the UGT1A1 variant were found to have significantly lower levels of glucuronidation following irinotecan administration. grade III and grade IV diarrhoea and neutropaenia only occurred in patients who were UGT1A1 heterozygous or a homozygous variant. 12. Ando, Y. et al. Polymorphisms of UDP-glucuronosyl transferase gene and irinotecan toxicity: a pharmacogenetic analysis. Cancer Res. 60, (2000). 13. Iyer, L. et al. UGT1A1*28 polymorphism as a determinant of irinotecan disposition and toxicity. Pharmacogenomics J. 2, (2002). 14. Illmer, T. et al. MDR1 gene polymorphisms affect therapy outcome in acute myeloid leukemia patients. Cancer Res. 62, (2002). 15. Park, D. J. et al. A Xeroderma pigmentosum group D gene polymorphism predicts clinical outcome to platinum-based chemotherapy in patients with advanced colorectal cancer. Cancer Res. 61, (2001). 16. Stoehlmacher, J. et al. A polymorphism of the XRCC1 gene predicts for response to platinum based treatment in advanced colorectal cancer. Anticancer Res. 21, (2001). 17. Marsh, S., Kwok, P. & McLeod, H. L. SNP databases and pharmacogenetics: great start, but a long way to go. Hum. Mutat. 20, (2002). 18. Sachidanandam, R. et al. A map of human genome sequence variation containing 1.42 million single nucleotide polymorphisms. Nature. 409, (2001). 19. Wagner, C. Biochemical role of folate in cellular metabolism (ed. Bailey, L.) (Marcel Dekker, New York, 1995). 20. Longley, D. B., Harkin, D. P. & Johnston, P. G. 5-Fluorouracil: mechanisms of action and clinical strategies. Nature Rev. Cancer 3, (2003). 21. Chu, E. & Allegra, C. in Cancer Chemotherapy and Biotherapy (eds. Chabner, B. & Longo, D.) (Lippincott Raven Publishers, Philadelphia, 1996). 22. Allegra, C. J. et al. Enhanced inhibition of thymidylate synthase by methotrexate polyglutamates. J.Biol. Chem. 260, (1985). 23. Baggott, J. E., Vaughn, W. H. & Hudson, B. B. Inhibition of 5-aminoimidazole-4-carboxamide ribotide transformylase, adenosine deaminase and 5 -adenylate deaminase by polyglutamates of methotrexate and oxidized folates and by 5-aminoimidazole-4-carboxamide riboside and ribotide. Biochem. J. 236, (1986). 24. Chu, E., Drake, J. C., Boarman, D., Baram, J. & Allegra, C. J. Mechanism of thymidylate synthase inhibition by methotrexate in human neoplastic cell lines and normal human myeloid progenitor cells. J. Biol. Chem. 265, (1990). 25. Grem, J. in Cancer chemotherapy and biotherapy (eds. Chabner, B. & Longo, D.) (Lippincott Raven Publishers, Philadelphia, 1996). 26. McGavin, J. K. & Goa, K. L. Capecitabine: a review of its use in the treatment of advanced or metastatic colorectal cancer. Drugs 61, (2001). 27. Blount, B. C. et al. Folate deficiency causes uracil misincorporation into human DNA and chromosome breakage: implications for cancer and neuronal damage. Proc. Natl Acad.Sci. USA 94, (1997). 28. Duthie, S. J. & McMillan, P. Uracil misincorporation in human DNA detected using single cell gel electrophoresis. Carcinogenesis 18, (1997). 29. Duthie, S. J. & Hawdon, A. DNA instability (strand breakage, uracil misincorporation, and defective repair) is increased by folic acid depletion in human lymphocytes in vitro. FASEB J. 12, (1998). 30. Lorico, A. et al. Accumulation of DNA strand breaks in cells exposed to methotrexate or N10-propargyl-5,8-dideazafolic acid. Cancer Res. 48, (1988). 31. Chabner, B. A. & Young, R. C. Threshold methotrexate concentration for in vivo inhibition of DNA synthesis in normal and tumorous target tissues. J. Clin. Invest. 52, (1973). 32. Sirotnak, F. M. & Moccio, D. M. Pharmacokinetic basis for differences in methotrexate sensitivity of normal proliferative tissues in the mouse. Cancer Res. 40, (1980). 33. Kaneda, S. et al. Role in translation of a triple tandemly repeated sequence in the 5 -untranslated region of human thymidylate synthase mrna. Nucleic Acids Res. 15, (1987). 34. Marsh, S. et al. Novel thymidylate synthase enhancer region alleles in African populations. Hum. Mutat. 16, 528 (2000). 35. Horie, N., Aiba, H., Oguro, K., Hojo, H. & Takeishi, K. Functional analysis and DNA polymorphism of the tandemly repeated sequences in the 5 -terminal regulatory region of the human gene for thymidylate synthase. Cell Struct. Funct. 20, (1995). 36. Pullarkat, S. T. et al. Thymidylate synthase gene polymorphism determines response and toxicity of 5-FU chemotherapy. Pharmacogenomics J. 1, (2001). 37. Trinh, B. N., Ong, C.-N., Coetzee, G. A., Yu, M. C. & Laird, P. W. Thymidylate synthase: a novel genetic determinant of plasma homocysteine and folate levels. Hum. Genet. 111, (2002). 38. Skibola, C. F. et al. Polymorphisms in the thymidylate synthase and serine hydroxymethyltransferase genes and risk of adult acute lymphocytic leukemia. Blood 99, (2002). 39. Ulrich, C. M., Bigler, J., Bostick, R., Fosdick, L. & Potter, J. D. Thymidylate synthase promoter polymorphism, interaction with folate intake, and risk of colorectal adenomas. Cancer Res. 62, (2002). 40. Ulrich, C. et al. Searching expressed sequence tag databases: discovery and confirmation of a common polymorphism in the thymidylate synthase gene. Cancer Epidemiol. Biomarkers Prev. 9, (2000). 41. Lenz, H.-J. et al. A 6 base-pair deletion in the 3 UTR of the thymidylate synthase (TS) gene predicts TS mrna expression in colorectal tumors. A possible candidate gene for colorectal cancer risk. Proc. Am. Assoc. Cancer Res. 43, 660 (2002). 42. Frosst, P. et al. A candidate genetic risk factor for vascular disease: a common mutation in methylenetetrahydrofolate reductase. Nature Genet. 10, (1995). 43. Skibola, C. F. et al. Polymorphisms in the methylenetetrahydrofolate reductase gene are associated with susceptibility to acute leukemia in adults. Proc. Natl Acad. Sci. USA 96, (1999). 44. Wiemels, J. L. et al. Methylenetetrahydrofolate reductase (MTHFR) polymorphisms and risk of molecularly defined subtypes of childhood acute leukemia. Proc. Natl Acad. Sci. USA 98, (2001). 45. Ma, J. et al. Methylenetetrahydrofolate reductase polymorphism, dietary interactions, and risk of colorectal cancer. Cancer Res. 57, (1997). 46. Chen, J. et al. A methylenetetrahydrofolate reductase polymorphism and the risk of colorectal cancer. Cancer Res. 56, (1996). NATURE REVIEWS CANCER VOLUME 3 DECEMBER

9 47. Ulrich, C. M. et al. Colorectal adenomas and the C677T MTHFR polymorphism: evidence for gene environment interaction? Cancer Epidemiol. Biomarkers Prev. 8, (1999). 48. Slattery, M. L., Potter, J. D., Samowitz, W., Schaffer, D. & Leppert, M. Methylenetetrahydrofolate reductase, diet, and risk of colon cancer. Cancer Epidemiol. Biomarkers Prev. 8, (1999). 49. Le Marchand, L. et al. B-vitamin intake, metabolic genes, and colorectal cancer risk (United States). Cancer Causes Control 13, (2002). 50. Shaw, G. M., Rozen, R., Finnell, R. H., Wasserman, C. R. & Lammer, E. J. Maternal vitamin use, genetic variation of infant methylenetetrahydrofolate reductase, and risk for spina bifida. Am. J. Epidemiol. 148, (1998). 51. Ou, C. Y. et al. 5,10 Methylenetetrahydrofolate reductase genetic polymorphism as a risk factor for neural tube defects. Am. J. Med. Genet. 63, (1996). 52. Kluijtmans, L. A. et al. Thermolabile methylenetetrahydrofolate reductase in coronary artery disease. Circulation 96, (1997). 53. van der Put, N. M. et al. A second common mutation in the methylenetetrahydrofolate reductase gene: an additional risk factor for neural-tube defects? Am. J. Hum. Genet. 62, (1998). 54. Weisberg, I., Tran, P., Christensen, B., Sibani, S. & Rozen, R. A second genetic polymorphism in methylenetetrahydrofolate reductase (MTHFR) associated with decreased enzyme activity. Mol. Genet. Metab. 64, (1998). 55. Keku, T. et al. 5,10-Methylenetetrahydrofolate reductase codon 677 and 1298 polymorphisms and colon cancer in African Americans and whites. Cancer Epidemiol. Biomarkers Prev. 11, (2002). 56. Chango, A. et al. A polymorphism (80G->A) in the reduced folate carrier gene and its associations with folate status and homocysteinemia. Mol. Genet. Metab. 70, (2000). 57. Marsh, S., Collie-Duguid, E. S., Li, T., Liu, X. & McLeod, H. L. Ethnic variation in the thymidylate synthase enhancer region polymorphism among Caucasian and Asian populations. Genomics 58, (1999). 58. Marsh, S., McKay, J. A., Cassidy, J. & McLeod, H. L. Polymorphism in the thymidylate synthase promoter enhancer region in colorectal cancer. Int. J. Oncol. 19, (2001). 59. Iacopetta, B., Grieu, F., Joseph, D. & Elsaleh, H. A polymorphism in the enhancer region of the thymidylate synthase promoter influences the survival of colorectal cancer patients treated with 5-fluorouracil. Br. J. Cancer 85, (2001). 60. Ueland, P. M., Hustad, S., Schneede, J., Refsum, H. & Vollset, S. E. Biological and clinical implications of the MTHFR C677T polymorphism. Trends Pharmacol. Sci. 22, (2001). 61. Wisotzkey, J. K., Toman, J., Bell, T., Monk, J. S. & Jones, D. MTHFR (C677T) polymorphisms and stage III colon cancer: response to therapy. Mol. Diagn. 4, (1999). 62. Ulrich, C. M. et al. Pharmacogenetics of methotrexate: toxicity among marrow transplantation patients varies with the methylenetetrahydrofolate reductase polymorphism. Blood 98, (2001). Increased severity of oral mucositis and slower platelet recovery was observed among individuals with the homozygous variant MTHFR 677TT genotype compared with individuals with the wildtype (MTHFR 677CC) or heterozygous (MTHFR 677CT) genotype following haematopoietic-cell transplantation for chronic myelogenous leukaemia. 63. Storb, R. et al. Methotrexate and cyclosporine compared with cyclosporine alone for prophylaxis of acute graft versus host disease after marrow transplantation for leukemia. N. Engl. J. Med. 314, (1986). 64. Chiusolo, P. et al. Preponderance of methylenetetrahydrofolate reductase C677T homozygosity among leukemia patients intolerant to methotrexate. Ann. Oncol. 13, (2002). 65. Toffoli, G. et al. Effect of methylenetetrahydrofolate reductase 677C T polymorphism on toxicity and homocysteine plasma level after chronic methotrexate treatment of ovarian cancer patients. Int. J. Cancer 103, (2003). 66. Toffoli, G., Veronesi, A., Boiocchi, M. & Crivellari, D. MTHFR gene polymorphism and severe toxicity during adjuvant treatment of early breast cancer with cyclophosphamide, methotrexate, and fluorouracil (CMF). Ann. Oncol. 11, (2000). 67. Taub, J. W. et al. Polymorphisms in methylenetetrahydrofolate reductase and methotrexate sensitivity in childhood acute lymphoblastic leukemia. Leukemia 16, (2002). 68. Krajinovic, M., Costea, I. & Chiasson, S. Polymorphism of the thymidylate synthase gene and outcome of acute lymphoblastic leukaemia. Lancet 359, (2002). 69. Laverdiere, C., Chiasson, S., Costea, I., Moghrabi, A. & Krajinovic, M. Polymorphism G80A in the reduced folate carrier gene and its relationship to methotrexate plasma levels and outcome of childhood acute lymphoblastic leukemia. Blood 100, (2002). 70. Cohen, V. et al. Methylenetetrahydrofolate reductase polymorphism in advanced colorectal cancer: a novel genomic predictor of clinical response to fluoropyrimidinebased chemotherapy. Clin. Cancer Res. 9, (2003). 71. Watters, J. W. & McLeod, H. L. Using genome-wide mapping in the mouse to identify genes that influence drug response. Trends Pharmacol. Sci. 24, (2003). 72. Haston, C. K. et al. Bleomycin hydrolase and a genetic locus within the MHC affect risk for pulmonary fibrosis in mice. Hum. Mol. Genet. 11, (2002). This report outlines how mouse models were used to conduct a genome-wide screen for identification of susceptibility loci for bleomycin-induced pulmonary fibrosis. 73. Marton, M. J. et al. Drug target validation and identification of secondary drug target effects using DNA microarrays. Nature Med. 4, (1998). 74. van de Vijver, M. J. et al. A gene-expression signature as a predictor of survival in breast cancer. N. Engl. J. Med. 347, (2002). This report outlines the use of microarray analysis to classify patients with breast cancer as having a geneexpression profile that is consistent with either poor or good prognosis. Multivariate Cox regression analysis found the gene-expression profile to be a strong predictor of disease outcome. 75. van t Veer, L. J. et al. Gene expression profiling predicts clinical outcome of breast cancer. Nature 415, (2002). 76. Rothenberg, M. L., Carbone, D. P. & Johnson, D. H. Improving the evaluation of new cancer treatments: challenges and opportunities. Nature Rev. Cancer 3, (2003). 77. Villafranca, E. et al. Polymorphisms of the repeated sequences in the enhancer region of the thymidylate synthase gene promoter may predict downstaging after preoperative chemoradiation in rectal cancer. J. Clin. Oncol. 19, (2001). Patients with rectal cancer receiving pre-operative 5-fluorouracil who were homozygous for the triple 28bp enhancer-region repeat of the thymidylatesynthase gene were found to have a lower probability of tumour downstaging and lower disease-free survival rates when compared with patients who were homozygous for the double repeat or heterozygous 2/3 repeat. 78. Park, D. J. et al. Thymidylate synthase gene polymorphism predicts response to capecitabine in advanced colorectal cancer. Int. J. Colorectal Dis. 17, (2002). Acknowledgements Support for K. R. was provided by a training grant from the National Cancer Institute. H. M. is supported in part by the National Institute of Health Pharmacogenetics Research Network. Competing interests statement The authors declare that they have no competing financial interests. Online links DATABASES The following terms in this article are linked online to: Cancer.gov: acute lymphocytic leukaemia acute myelogenous leukaemia colorectal cancer LocusLink: ABCB1 CYP2D6 ERCC2 MTHFR UGT1A1 XRCC1 FURTHER INFORMATION Pharmacogenetics and Pharmacogenomics Knowledge Base: Pharmacogenetics Research Network, Washington University: Kyoto Encyclopedia of Genes and Genomes: Pharmacogenomics glossary, Cambridge Healthtech Institute: pharmacogenomics.asp The Mouse Models of Human Cancers Consortium: Access to this interactive links box is free online 920 DECEMBER 2003 VOLUME 3