• ABCB1;
  • GST;
  • MTHFR;
  • DHFR;
  • DPYD;
  • TYMS;
  • TP53;
  • 5-FU;
  • doxorubicin;
  • breast cancer


  1. Top of page
  2. Abstract
  3. Material and methods
  4. Results
  5. Discussion
  6. References
  7. Supporting Information

Genetic factors are thought to play a role in resistance towards chemotherapeutic agents such as 5-fluorouracil (5-FU). Approximately 30 genes are directly or indirectly involved in 5-FU metabolism, and genetic variation in any of these may contribute to anti-tumor response. Polymorphisms in these genes were analyzed in relation to tumoral mRNA levels of 5-FU metabolizing genes, response to chemotherapy and survival. A total of 21 genetic variants were studied in 35 breast cancer patients treated with 5-FluoroUracil, mitomycin (FUMI) and in a similar cohort of 90 doxorubicin treated breast cancer patients. Genotype distributions were compared using 109 healthy controls. No significant association was found between any polymorphisms and response to chemotherapy as measured by shrinkage of tumor. However, carriers of 3 copies of the TYMS 5′UTR repeat had shorter survival than noncarriers (p = .048) in the FUMI treatment group, but not in the doxorubicin treated group. Carriers of 3 copies of the repeat were also more frequently observed in both patients groups than in healthy controls (p = .034). Several highly significant associations were observed between genotypes and expression levels of 5-FU metabolizing genes. A SNP in codon 72 of TP53 was revealed to be a key regulator of 5-FU metabolizing genes such as DHFR and MTHFR, constituting 50% of all significant associations observed after FUMI therapy. These data suggest that 3 copies of the TYMS 5′UTR repeat may give a treatment specific reduced survival in breast cancer patients, and that TP53 may have a direct, allele specific, role in 5-FU mediated response. © 2008 Wiley-Liss, Inc.

5-Fluorouracil (5-FU) is an uracil analogue widely used in the treatment of solid tumors, including colorectal and breast cancers. The mechanism of 5-FU cytotoxicity is ascribed to the incorporation of fluoronucleotides into RNA and DNA, and blocking of the active site in the nucleotide synthetic enzyme thymidylate synthase (TS), resulting in activation of TP531 and executing of apoptosis. In vitro studies have shown that loss of TP53 function reduces cellular sensitivity to 5-FU.2

Approximately 30 genes are involved in the metabolism and action of 5-FU (Fig. 1). Dihydropyrimidine dehydrogenase (DPD protein, DPYD gene) is the rate-limiting enzyme in 5-FU catabolism, responsible for the degradation of ∼80% of the drug to an inactive metabolite, DHFU.3 Several different low frequency polymorphisms affecting DPD enzyme activity have been described, the most frequent being a G to A transition in the splice donor site flanking exon 14 (rs3918290), resulting in deletion of exon 14.4, 5 This leads to decreased clearance and increased half-life of 5-FU, leaving the cancer hypersensitive to 5-FU treatment. Expression of DPYD mRNA has been reported to be inversely correlated with response to 5-FU.6

thumbnail image

Figure 1. Metabolism and action of 5-FU. Dihydropyrimidine dehydrogenase (DPYD)-mediated conversion of 5-FU to dihydrofluorouracil (DHFU) is the rate-limiting step of 5-FU catabolism in normal and tumor cells. Up to 80% of administered 5-FU is broken down by DPYD in the liver. 5-FU is converted to 3 main active metabolites, fluorodeoxyuridine monophosphate (FdUMP), fluorodeoxyuridine triphosphate (FdUTP) and fluorouridine triphosphate (FUTP), marked as 1, 2 and 3, respectively. The main mechanism of 5-FU activation is conversion to fluorouridine monophosphate (FUMP), either directly by uridine monophosphate synthase (UMPS), or indirectly via fluorouridine (FUR) through the sequential action of uridine phosphorylase (UPP1 & UPP2) and uridine-cytidine kinase (UCK1 and NT5C). FUMP is then phosphorylated to fluorouridine diphosphate (FUDP), which can be either further phosphorylated to the active metabolite fluorouridine triphosphate (FUTP), or converted to fluorodeoxyuridine diphosphate (FdUDP) by ribonucleotide reductase (RRM1 & RRM2). In turn, FdUDP can either be phosphorylated or dephosphorylated to generate the active metabolites FdUTP and FdUMP, respectively. An alternative activation pathway involves the thymidine phosphorylase catalysed conversion of 5-FU to fluorodeoxyuridine (FUDR), which is then phosphorylated by thymidine kinase (TK1) to FdUMP. The enzymes involved in the pathway have red color (dark grey ellipse), and those that were genotyped for 1 or more polymorphisms are shown with a black, thick border. Metabolites are shown in yellow color (light grey ellipse). Levels of mRNA expression were obtained from previous microarray analyses for all genes expect TK1 which is marked with an asterisk.

Download figure to PowerPoint

The main target of 5-FU is TS, and polymorphisms in TYMS may influence response to treatment. Variability in TYMS expression levels is in part determined by a 28 bp repeat polymorphism in the 5′UTR of the gene. The number of repeats varies from 2 to 9, with 2 and 3 being predominant. Increased number of repeats is associated with increased mRNA and protein levels of TYMS in vitro,7, 8 suggesting that the particular alleles have the potential to be expressed differentially under the experimental conditions. Colorectal cancer patients carrying 3 copies of the repeat have been reported to have reduced response to 5-FU treatment9 and reduced survival compared to individuals with 2 copies,10 but the data remains inconclusive,11, 12 and have not always been confirmed for other types of cancers.13, 14 A second polymorphism in the gene, a 6 bp insertion/deletion located in the 3′UTR of TYMS (rs28365050) has been reported to be associated with reduced mRNA stability.15, 16 This polymorphism is in linkage disequilibrium (LD) with the repeat polymorphism in the 5′UTR,16 and the haplotype composed of 2 repeats and the 6 bp insertion has been reported to be associated with increased risk of severe side effects of 5-FU treatment in colorectal cancer patients.17

Methylenetetrahydrofolate reductase (MTHFR) controls the intracellular levels of CH2FH4 by irreversibly converting it to FH2 (Fig. 1).18 Two common MTHFR SNPs, 677C>T in exon 4 (rs1801133) and 1298A>C (rs1801131) in exon 7, have been associated with lower enzyme activity,19, 20 which should in theory improve response to 5-FU. Several studies have correlated the wt, CC, genotype in MTHFR677 with reduced 5-FU response in colorectal cancer,21–24 while the evidence for the effect of MTHFR1298 SNP seems less conclusive.

Dihydrofolate reductase (DHFR) converts FH2 into FH4, a process required for the de novo synthesis of purines, thymidine monophosphate, and certain amino acids. Goto et al. described a SNP in the 3′UTR of DHFR (829C > T, rs1677669) where the TT genotype was significantly associated with higher expression of DHFR.25

The motivation of our study was to create an array with a panel of all SNPs, which have been previously functionally characterized and shown to have an effect on the response to 1 or another cancer related drug. The ultimate goal of this is to create a pharmacogenetic profile by genotyping a single sample of a patient prior to decision of which treatment would be chosen if all other criteria seem the same. A pathway based analysis, rather than single-gene studies, may enable us to better detect the combined effects of these polymorphisms on 5-FU treatment in breast cancer. In our study we have analyzed SNPs and repeat polymorphisms in the best characterized genes of the metabolism and action of 5-FU (DPYD, TYMS, DHFR, MTHFR and TP53) and additional SNPs in genes with relevance to several other related pharmacological pathways (TPMT, UGT1A1, ABCB1, GSTM, GSTT and GSTP), using multiplex genotyping assays on the Nanogen platform. The GSTs and UGT1A1 are of major importance in the conjugation and subsequent elimination of potentially toxic xenobiotics and endogenous compounds, while TPMT catalyzes the S-methylation of thiopurine drugs such as 6-mercaptopurine. The polymorphisms and intratumoral gene expression levels of selected genes were analyzed and compared to TP53 mutation status, antitumoral response and survival in locally advanced breast cancer patients undergoing primary chemotherapy with FUMI26 or doxorubicin.27 The doxorubicin treated cohort was included in our study to serve as a control group in the investigation of treatment specific associations.

Material and methods

  1. Top of page
  2. Abstract
  3. Material and methods
  4. Results
  5. Discussion
  6. References
  7. Supporting Information

Study populations

Patient material

Tumor DNA was collected from locally advanced breast cancer patients, treated with FUMI; 5-fluoro-uracil 100 mg/m2 day 1 and 2; mitomycin 6 mg/m2 day 2 for 3-weekly cycles (n = 35) and doxorubicin; 14 mg/m2 on a weekly schedule (n = 90), enrolled in an IRB approved protocol evaluating the drug response in a neoadjuvant setting.26–28 Response was classified as progressive disease, PD (increase in tumor diameter by ≥25%), partial response, PR (reduction ≥50%), or stable disease SD (anything between PR and PD). The SD and PR groups were combined since no difference in response or survival was observed between patients (explained in detail elsewhere28, 29). In the FUMI treated cohort, the distribution of patients was as follows: PD = 9, SD(15) + PR(10) = 25 (1 sample could not be evaluated with regards to response), and for the doxorubicin treated patients PD = 9, SD + PR = 81. Tumor category was graded according to the classification of Page et al.30 The median age at diagnosis for the FUMI cohort was 67 years (range, 37–82 years), and 64 years for the doxorubicin cohort (range, 32–88 years).

Tumor samples were collected by incisional biopsies and snap-frozen (liquid nitrogen) in the theatre as previously outlined.26, 27 A second sample was collected from the tumor specimen at surgery and snap-frozen according to the same procedures.

Blood-tumor DNA pairs were available for 24 and 49 patients from the FUMI and doxorubicin cohort, respectively. These pairs were genotyped for the 6 ABCB1 SNPs, the MTHFR677 and TP53 codon 72 SNP. No difference between the blood and tumor genotype for the ABCB1 SNPs was observed in either material. One discrepancy between blood and tumor genotype was observed in the FUMI treated material for the MTHFR677SNP. For the TP53 codon 72 SNP, 5 of the doxorubicin treated patients showed a discrepancy between blood and tumor genotype, all displayed an LOH (Loss Of Heterozygosity) event with loss of the C allele. Although LOH of codon 72 in TP53 was not observed in the FUMI treated samples (number of informative markers = 9), we chose to analyze the blood and tumor derived genotypes of this SNP separately when considering the association to case-control status or various clinical parameters.

Control material

Leukocyte DNA was isolated from blood obtained from 109 healthy women between 55 and 75 years of age participating in the national mammographic screening program and living in the same area as the cases. The women enrolled in the study had 2 consecutive negative mammograms in a period of 2 years and were not on any hormone replacement therapy (HRT).31

Sample preparation

Peripheral blood was collected in EDTA and stored at –40°C, and fresh-frozen tumor material was stored in liquid nitrogen. DNA was isolated using phenol/chloroform extraction and ethanol precipitation using standard protocols.

TP53 mutation analyses

Mutations in the TP53 gene in the tumor samples have previously been analyzed, and the data retrieved from Geisler et al.26, 27 Eighteen of the 35 patients in the FUMI cohort had tumors harboring mutations in the TP53 gene. Of the 18 mutations detected, 8 were missense (3 transversions and 5 transitions), 1 was nonsense, 2 were splice mutations, 4 were deletions, 2 were insertions resulting in frameshifts and 1 was an in-frame insertion.26 In the doxorubicin treated cohort, 26 out of the 90 patients harbored a TP53 mutation in their tumor. Of these, 16 were missense mutations, 4 were nonsense mutations, 2 had a frame shift mutation, 1 had an in frame mutation and 3 had splice mutations.27

SNP Genotyping

Selection of SNPs

The SNPs in our study were selected based on: (i) type of SNP (preferably cSNPs), (ii) reported frequency and (iii) previous associations to treatment response reported in the literature. The genes, the type of allele and the rs number of the selected SNPs are shown in Supplementary Table I.

Genotyping of SNPs on the nanogen

Fifteen SNPs were genotyped on the Nanogen Molecular Biology Workstation platform (Nanogen, San Diego, CA). The PCRs were performed using primers described in Supplementary Table II in a volume of 20 μL with 50 ng of genomic DNA, 20 pmol of each primer, 0.2 μL of dNTP [25 mM], 0.5 units of HotStartTaq DNA polymerase (Qiagen, Hilden, Germany), and 1x PCR buffer supplied with the polymerase. The PCR conditions were as follows: an initial denaturation of 96°C for 15 min; 35 cycles of 96°C for 1 min, X°C for 1 min (X, annealing temperature for each assay, shown in Supplementary Table II), 72°C for 1 min and a final extension of 72°C for 10 min on the MJ Research Tetrad Thermal Cycler. The PCRs were multiplexed as given by the PCR column in Supplementary Table II. The products were purified using a 96-well filter plate from Millipore, resuspended in 70 μL H20, and 30 μL of the purified PCR product was mixed with 30 μL of 100 mM Histidine before loading onto the microarray.

Genotyping on the Nanogen has been described elsewhere.32 Briefly, the Nanogen platform applies a positive current to attach the biotin labeled, PCR amplified, target DNA to a specific streptavidin coated pad33 (Supplementary Fig. 1). Subsequently, the addressed PCR product was denatured, incubated with High Salt Buffer (HSB), and a reporter mix was added and allowed to hybridize. The reporter mix constituted of stabilizers, wild type and mutant discriminators, and universal reporters mixed in HSB. The universal reporters are fluorescent labeled. The oldest assays are run without the universal reporter set up, and hence, the fluoressence is added to the wt and mt discriminator as indicated in Supplementary Table II. The microarray was placed in the reader and the temperature was set to ramp down (touch down protocol) from 56°C to a suitable discrimination temperature (temperature varies from assay to assay, listed in Supplementary Table II), at which mismatched discriminators were denatured and matched discriminators remained hybridized due to difference in base-stacking energies. Unbound reporters were washed away, the microarray cooled to 24°C, and pads scanned. Signals were normalized against the background represented by a pad addressed solely with 50 mM histidine. To determine the genotypes, the ratio of 2 fluorescent signals (red to green) in each pad was calculated. A homozygous (either wild type or mutant) genotype was read out when the ratio was higher than 5, and a heterozygote genotype was called when the ratio was less than 2. The genotypes were normalized to known heterozygote controls as determined by sequencing. The accuracy of genotyping on Nanogen has previously been evaluated, and, with regards to both success and error rate, found comparable to the other platforms assayed, such as TaqMan, SNPstream and MassArray.34 Further, 35 of the genotypes obtained with Nanogen were verified by sequencing and a perfect match (100%) to the genotype calls was observed.

Genotyping of repeat polymorphisms

PCR primers for the amplification of TYMS UTRs were designed using the Primer 3 software ( The UGT1A1 primers have been described previously by von Ahsen et al.35 The primers for both genes (shown in Supplementary Table II) were 5-carboxyfluorescein (FAM) labeled for fluorescent detection on AB310. PCR amplification was performed in 20 μL volume, and the reaction mixture consisted of 50 ng of DNA, 2.5 μL dNTP [2.5 mM], 20 pmol of each primer, 0. 5 U of HotStarTaq DNA polymerase (Qiagen) and 1x PCR buffer supplied with the polymerase. After initial denaturation at 96°C for 15 min, denaturation, annealing and extension were carried out at 96°C for 1 min, at X°C for 1 min (X, annealing temperature for each assay, given in Supplementary Table II), and at 72°C for 1 min, respectively, for 35 cycles in an MJ Research Tetrad Thermal Cycler followed by a 10-min extension at 72°C. One microliter of PCR product was mixed with 12 μL of formamide and 0.7 μL of the Genescan ROX500 size standard. The mixture was denatured at 95°C for 3 min and kept on ice. Electrophoretic analysis was performed with POP4 gel and an ABI 310 genetic analyzer (Applied Biosystems, Foster City, CA). The amplified products were analyzed by Genescan Analysis software (v3.1, Applied Biosystems). Accurate sizing of the alleles was determined by using the internal size marker, allowing relative allele peak areas to be calculated. The results were confirmed by sequencing of 12 control samples for each repeat. The PCR products were sequenced with their reverse primer, the BigDye Terminator v3.1 Cycle Sequencing Kit and the 377 Genetic Analyzer (Applied Biosystems).

Genotyping of the GSTs polymorphisms

The selected polymorphisms in the GSTs were genotyped as described by Kristensen et al.,36 and primer sequence are given in Supplementary Table II.

Gene expression

Gene expression profiles of the tumors from the breast cancer patients studied here have previously been published,37 and the expression data was downloaded from the publicly available Stanford database ( Tumor expression profiles before FUMI treatment existed for 29 patients of which 12 also had available expression data after treatment. For the doxorubicin treated patients, tumor expression profiles before treatment were available for 51 patients of which 37 also had after treatment data. The genes for which expression data was downloaded are shown with red ellipses in Figure 1. The raw gene expression ratios were log2-transformed and filtered based on their presence in at least 80% of the experiments with a signal intensity of at least 1.5 above background.

Statistical analysis

Associations between markers and disease phenotypes were analyzed using Pearson χ2 test (2-sided), and they were considered significant when the P-value was no higher than .05. Significant SNP-expression associations shown in the text or figures are given by their ANOVA P-value.

Survival analysis

Each polymorphism was represented as a 2-state (0/1) variable. Associations between survival (time to breast cancer specific death) and single polymorphisms were studied separately for the FUMI cohort and the doxorubicin cohort, using the survival library in the statistical software package R (v. 2.2.1). To test for differences in survival between genotypes, Cox proportional hazard models (method survdiff) were used with adjustment for age, and logrank tests were subsequently used if the Cox analysis showed no need for age adjustment. A Weibull model was used (using the method survreg) for further validation of some cases. The method cox.zph was used to test the proportional hazard assumption. Kaplan-Meier plots generated in SPSS (v13, SPSS, Chicago, IL) were used to visually compare survival between groups corresponding to the 2 states of a polymorphism.

Statistical analysis of SNP–mRNA expression associations

SNP-expression associations were studied before and after treatment for groups consisting of FUMI patients, doxorubicin patients and all patients combined. Initially, all SNP-expression associations in each of these groups were evaluated using two-tailed ANOVA. As a control to the ANOVA analysis, the procedure was repeated using Kendall's association test. To identify the specific SNP-expression associations, we used 2 complementary analyses, ANOVA and QMIS, as described by Tsalenko et al.38 ANOVA searches for a significant difference in distributions of expression values in each genotype class assuming normal distribution of expression values in each class. Kendall's association test measures the strength of dependence between 2 ordinal variables and assesses the significance of this correspondence. QMIS searches for a significant difference in the distribution of genotypes in the group of samples with high expression compared to the group of samples with low expression. A total of 18 polymorphisms and 21 transcripts were analyzed in the combined patient material before either treatment, and 16 variants in the post-treated samples. For the FUMI treated patients, the number of transcripts included in the analysis was 30, and the number of genotyped polymorphisms was 10. Three additional SNPs were analyzed only in the pretreatment SNP-expression analysis. For the doxorubicin samples, 15 different polymorphisms and the expression of 21 transcripts were included in the analysis.

Promoter analysis

Promoter analysis was performed on the genes whose expression after 5-FU treatment was associated with the TP53 codon 72 SNP. Sequences corresponding to approximately 600 bp upstream of the transcription initiation site were downloaded for each gene from the UCSC genome browser ( The sequences were selected to include as many hits as possible from the 5-way regulatory potential track in the browser. This track displays regulatory potential score computed from alignments of human, chimpanzee, mouse, rat and dog, and compares the frequencies of short alignment patterns between known regulatory elements and neutral DNA. The downloaded promoter sequences were analyzed for Transcription Factor Binding Sites (TFBS) using the web based tool Mapper.39 Mapper searches for potential TFBS by applying hidden Markov model profiles for each of the TFBS generated from the combined TRANSFAC and JASPAR databases.

Haplotype and linkage disequilibrium analysis

Genotype data was generated for each individual and haplotypes were inferred using PHASE 2.1,40 and the results were visualized in Excel (Microsoft). The PHASE algorithm is a Bayesian approach to haplotype inference with priors from population genetics. The overall haplotypes and haplotype blocks were calculated by and visualized in Haploview.41 Haploview estimates haplotypes using the accelerated EM algorithm. A haplotype block was defined as by Gabriel et al.,42 and is based on the 95% D′ confidence interval. Haploview was also used to estimate Hardy-Weinberg equilibrium for the genotypes, in order to control for technical genotyping errors and biases in the selection of study populations.


  1. Top of page
  2. Abstract
  3. Material and methods
  4. Results
  5. Discussion
  6. References
  7. Supporting Information

Overall distribution of genotypes in breast cancer patients and healthy controls

The genotype frequencies of the polymorphic variants of GSTM1, GSTT1, GSTP1, ABCB1, MTHFR, TPMT, DPYD, TP53, DHFR, TYMS and UGT1A1 in the case and control groups, as well as in the different treatment response groups of the breast cancer patients, are shown in Supplementary Table I, while the significant findings are listed in Table I.

Table I. SNPs Significantly Associated with Clinical Parameters and Expression
GeneAllelers IDTotalCase
Response to DoxorubicinResponse to 5-FUMITP53 mutation
ControlCasePR + SDPDPR + SDPDWtMt
  • PR, partly response; SD, stable disease; PD, progressive disease.

  • 1

    This SNP has for the cases been analyzed separately whether genotyped in tumor or blood DNA, marked with B and T, respectively.

  • 2

    Presence of 3 copies of the repeat. Note that response data was not available for all patients, and genotypes rates were not always with 100% success rate.

TP53 721 1042522 BT      
 GG 4643304961344230
 CC 5959140113
 GC 58211320184268
DHFR 3′ 1677666         
 CC 5971 498954525
 GG 99 702027
 CG 3738 221122308
         p = 0.005
TYMS 5′           
 3/3 3453 344863122
 2/2 2419 13132145
 2/3 5046 2741412916
   p = 0.0342 p = 0.066 
Samples 10990+9035125

The genotype frequencies for all polymorphisms, except those in TP53 codon 72 and MTHFR1298, did not deviate significantly from the predicted distribution (HWE; Hardy-Weinberg equilibrium) in neither cases nor controls. The SNP in MTHFR1298 was borderline deviating from HWE in the case population (p = 0.041), but not in the control population (p = 1), and was hence included in further analysis, while the TP53 codon 72 SNP deviated from HWE both in the case and control population (p = 0.009 and 0.019, respectively). Heterozygous individuals were underrepresented in the cases, and overrepresented in the controls. Under-representation of heterozygosity for the TP53 gene is expected in the cases when genotyping tumor DNA, and is explained by LOH. The SNP in intron 14 of DPYD, which has previously been reported to have a minor allele frequency of 0.91% in Caucasians,4 was found to be monomorphic in the 35 5-FU treated cases analyzed here and was therefore not included in further analyses. In comparison, HapMap ( reported the frequency of the major allele, G of DPYD to be equal to 1 in the 90 CEPH samples.

The only significant difference in frequency between cases and controls were observed for the TYMS 5′UTR repeat polymorphism, p = 0.034. The allele containing 3 copies of the 28 bp was overrepresented in the case group.

SNPs and TP53 mutation status

TP53 tumor mutation status was found to be significantly associated with a SNP in the 3′UTR of DHFR (rs1677666) (Table I) In the combined FUMI- and doxorubicin treated patient material, TP53 tumor mutations were observed more frequently in tumors from patients carrying the GG genotype than carriers of the CC or CG genotype (p = 0.005), with 7 of the 9 patients (78%) carrying the GG genotype harboring a TP53 mutation.

SNPs and treatment response

Although no significant association between any single SNP and treatment response to FUMI was seen, there was a trend for patients homozygous for 3 copies of the TYMS 5′UTR 28 bp repeat polymorphism to be overrepresented in the PD response group compared to the combined PR and SD groups (p = 0.066, Table I). No significant association between any single SNP and treatment response to doxorubicin was observed.

Polymorphism and survival

Polymorphisms were coded as 0/1 variables by combining the heterozygous genotype and the least frequent homozygous genotype in 1 category (presence of the rare allele). For each treatment (FUMI or doxorubicin) we fitted a Cox proportional hazard model for each polymorphism, using the patient's genotype (0/1) and age as predictors. No significance (not even borderline) was observed with respect to age. We omitted age from all further analysis and used the logrank test within each of the 2 treatment cohorts to test for each polymorphism whether breast cancer specific survival was associated with genotype. It was found that FUMI treated patients homozygous for 3 copies of the 28 bp TYMS 5′UTR repeat polymorphism had significantly reduced breast cancer specific survival compared to individuals carrying 1 or 2 alleles with 2 copies of the repeat (p = 0.048, Fig. 2a). A parametric Weibull model was then fitted to the survival times with the above polymorphism in the TYMS 5′UTR as a covariate. The polymorphic state was significant (p = 0.042) and also the overall model (p = 0.029). The proportional hazards assumption was tested for this model and was not found to be violated. No such association between the TYMS 5′UTR repeat and overall survival was seen for the doxorubicin treated cohort (p = 0.85, Fig. 2b).

thumbnail image

Figure 2. Association between the TYMS 5′UTR repeat and treatment specific survival. Kaplan-Meier plot for (a) patients in the FUMI cohort that are homozygous for 3 copies of 28 bp TYMS 5′UTR repeat (green curve) and patients in the same cohort carrying 2 copies of the repeat (blue curve) (p < 0.05); (b) patients in the doxorubicin cohort that are homozygous for 3 copies of 28 bp TYMS 5′UTR repeat (green curve) and patients in the same cohort carrying 2 copies of the repeat (blue curve) (p = 0.85).

Download figure to PowerPoint

SNP–mRNA expression associations

Table II shows all associations for which both ANOVA and QMIS gave a p-value of 0.05 or less. To account for the fact that many associations are tested in each group, only those groups for which the number of significant associations exceeds what is expected by chance are reported in the table. Results for the remaining groups are given in Supplementary Table III. Calculations on SNP-expression associations were performed on the combined patient material for robustness, and for each treatment regime separately, to investigate whether post-treatment specific SNP-mRNA expression associations may arise. When evaluating groups of patients for SNP-expression association, the group constituting of FUMI treated patients tested after therapy displayed a markedly higher number of significant associations than expected due to chance at the 1, 5 and 10% level, both using ANOVA and Kendall's association tests. Table II lists the variables that passed the procedure suggested by Tsalenko et al.,38 requiring significance both for ANOVA and QMIS tests. Note especially that the TP53 codon 72 SNP was not associated to a single transcript in the samples taken before FUMI treatment, while it constituted 50% of the associations observed after therapy (Table II, Figs. 3a3i). The codon 72 TP53 SNP was also associated to the expression level of key enzymes regulating effectiveness of 5-FU metabolism such as DHFR and MTHFR (Figs. 3a and 3c, respectively). Notably, none of the genes related to 5-FUMI metabolism had an expression pattern associated with mutation status of TP53 after treatment (data not shown). A similar increase in associations between the codon 72 SNP and expression after treatment was not observed in the patients given doxorubicin therapy. Promoter analysis of the genes whose expression was significantly associated with the TP53 codon 72 SNP revealed in silico binding sites for TP53 in the promoter of DHFR, ECGF1, MTHFR, NME1, RRM1, TPMT and UGT1A1, but not in UMPS and UPP2.

thumbnail image

Figure 3. Associations between genotypes and mRNA expression. The TP53 codon 72 SNP was found by ANOVA analysis significantly associated with tumoral mRNA expression levels of 9 different genes, and the box plots displays the genotypes associations to the expression levels of (a) DHFR, (b) ECGF1, (c) MTHFR, (d) NME1, (e) RRM1, (f) TPMT, (g) UGT1A1, (h) UMPS and (i) UPP2 after 5-FUMI. Box-plots with examples of SNP-transcript pairs with significant associations in the combined FUMI and doxorubicin treated patient material before treatment show the following: (j) deletion of GSTT1 significantly associated to reduced GSTT1 expression, p = 2.2 E −5 and (k) the 6-bp deletion in the 3′UTR of TYMS associated to increased expression of TYMS, p = 0.019. Each graph shows a box-plot of expression data grouped by sample genotype for each polymorphism. The boxes have lines at the lower quartile, median and upper quartile values, and the whiskers denote the highest and lowest values which are no farther than 1.5 times the interquartile range (IQR) from the median. Outliers are denoted with ° and extreme outliers with *. The mRNA expression values are log2 transformed. [Color figure can be viewed in the online issue, which is available at]

Download figure to PowerPoint

Table II. Significant SNP-Gene Expression Associations Before and after 5-FUMI Treatment, as Found by Both Anova and Qmis Analysis
Before 5′-FU TreatmentAfter 5-FU treatment
  1. Since no discrepancy was seen between the blood and tumor genotype for the 24 pairs of 5-FUMI treated samples, the patients were not divided into 2 subgroups when SNPs association to gene expression was assayed.

 rs1128503ABCB10.0020.009GSTM1 MTHFR0.0090.048
DHFRrs1677666UNG0.0250.044  TYMS0.0070.048
GSTM1 NT5C20.0040.014TP53rs1042522DHFR0.0270.030
GSTP1rs947894DPYD0.0210.041  ECGF10.0210.017
  RRM10.0250.030  MTHFR0.0060.018
MTHFRrs1801131ABCB10.0050.032  NME10.0180.030
TYMSrs28365050ECGF10.0140.043  RRM10.0520.030
  UCK10.0260.039  TPMT0.0030.042

The results for the SNP-expression associations in the combined material and the doxorubicin treated cohort both before and after therapy are given in Supplementary Table III. For the combined material before treatment, we observed more associations at the 1% significance level than expected by chance. The largest single association was seen for the GSTT1 gene where deletion of the gene was linked to change in mRNA expression of 6 different genes including reduction in GSTT1 mRNA (p < 0.0001, Fig. 3j). Also, for the combined material, we observed a significant, pretreatment specific association between TYMS expression levels in patients homozygous for the 6 bp deletion in the 3′UTR of TYMS (p = 0.019, Fig. 3k). No association was found between the TYMS 5′-UTR tandem repeat polymorphism and TS mRNA expression. The QMIS analysis in fact did find a significant association between the 28 bp repeat and expression of TYMS (p = 0.047) in the 35 FUMI treated samples after treatment, however, the ANOVA analysis did not confirm the same association (p-value = 0.367), and due to our strict filter criteria in order to reduce the number off false positives, this association is not reported as significant (data not shown).

The TPMT 460 and 719 SNPs showed a significant association to DPYD expression in the combined, pretreated samples (ANOVA; p = 0.007 and p = 0.011, respectively, Supplementary Table III). The TPMT SNPs were in complete LD (Supplementary Fig. 2d).

Haplotype structures and linkage disequilibrium in the ABCB1, TPMT, MTHFR, DHFR and TYMS genes

The 2 MTHFR SNPs were found to be in LD (D′ = –1) by both tools of analysis, PHASE and Haploview (Supplementary Fig. 2a). The haplotype composed of the 2 mutant alleles was never observed (Supplementary Fig. 2f). The 2 TYMS repeats were not found in linkage disequilibrium by either method, D′ = 0.36 (Supplementary Fig. 2b), and all 4 possible combinations of the alleles were observed (Supplementary Fig. 2g). The 2 SNPs in the 3′UTR of the DHFR gene were found to be in LD, D′ = 1 (Supplementary Fig. 2c), and the haplotype distribution is shown in Supplementary Figure 2h. The 3 SNPs in TPMT were in complete linkage disequilibrium, D′ = 1 (Supplementary Fig. 2d). However, the LOD score was low for some of the calculations, and the frequencies of the haplotypes composed of mutant alleles were low (Supplementary Fig. 2i). The LD pattern and haplotypes in ABCB1 have previously been reported for the doxorubicin treated patients and the controls,43 and is here shown for the combined FUMI and doxorubicin treated patients and the controls (Supplementary Figs. 2e and 2j).


  1. Top of page
  2. Abstract
  3. Material and methods
  4. Results
  5. Discussion
  6. References
  7. Supporting Information

Breast cancer is a complex disease, with genetic and environmental factors interacting with both tumor development and response to therapy. The aim of our study was to evaluate whether genetic markers selected from pharmacogenetic relevant pathways may function as predictive markers for response to FUMI treatment, in an attempt to develop new and simpler DNA based methods to predict the clinical outcome of treatment and improve the overall survival.

Our data suggest that 3 copies of a 28 bp repeat polymorphism in the 5′UTR of TYMS may be associated with shorter survival of breast cancer patients treated with FUMI, but not for those treated with doxorubicin. Notably, although the 2 treatment options were not administered in a randomized study, the patient population of the 2 studies was similar regarding ethnic background, age and stage of disease. The associated polymorphism resides in a gene that is a central player in the metabolism and action of 5-FU. The homozygous genotype with 3 copies of the 28-bp repeat polymorphism in the 5′UTR of TYMS was associated with short survival and was found more frequently in the PD response group. Iacopetta et al. showed that colorectal cancer patients homozygous for 3 copies of repeat and receiving 5-FU treatment had a significantly shorter survival.10 Inspired by this, Paré et al. studied the effect of the repeat on outcome of breast cancer patients treated with adjuvant therapy containing 5-FU, and observed no significant association.44 However, chemotherapy plays a more important role in the treatment of advanced diseases, suggesting that 3 copies of the repeat may have an effect on overall survival in a neoadjuvant, but not in an adjuvant setting. The 28 bp repeat resides in the promoter of TYMS and may function as an enhancer element. An E-box binding site for the upstream stimulatory factor USF has been found within the repeat allele, and experimental studies have shown that transcriptional regulation of TYMS is dependent on USF protein binding within the repeat.45 Increased number of repeats has previously been associated with higher mRNA and protein levels of TYMS.7, 8 However, we did not observe a significant association between the TYMS 5′ repeat and expression level of TYMS that passed our filter criteria in any of the cohorts. 5-FU exhibits its main function through the inhibition of TS, and increased intracellular levels of its protein in a cervical cancer cell line were reported to correlate with decreased sensitivity to 5-FU.46 Further, Yu et al. reported that 5-FU treated breast cancer patients positive for TS, as determined by immunhistochemistry, had an aggressive phenotype, and that TYMS expression was an independent prognostic marker for both disease-free and overall survival.47 These observations may offer an explanation to the reduced survival seen in our patients carrying 3 copies of the 5′UTR TYMS repeat.

For the combined material before treatment, we observed a marked higher number of significances at the 1% level than expected. This group holds the best power to detect the association of genotypes to tumoral mRNA expression. The analysis did not only confirm the previously reported association between the 6 bp deletion in the 3′UTR of TYMS and expression level of TYMS,16 but also revealed that deletion of GSTT1 affects the expression levels of a whole set of genes with pharmacological relevance. Further, the only mutation known to affect the enzymatic function of DPYD, the rate limiting enzyme in 5-FU catabolism, was found to be monomorphic in our population. However, we observed that the 2 TPMT SNPs in complete LD passing our frequency criteria were both associated with the expression of DPYD. Low levels of DPYD have previously been associated with severe 5-FU toxicity,48 so it is of interest to note that other SNPs present in the population at a higher frequency may also affect the expression level of DPYD.

In the 5-FU treatment cohort, we observed after therapy nearly twice as many significant associations than expected by chance at the 5% level, and report here for the first time the association of a polymorphism in TP53 codon 72 to the mRNA expression levels of a battery of enzymes directly involved in the metabolism of 5-FU. This finding in our view may suggest that 1. the expression of genes in the 5-FU pathway is changed after the administration of the drug (therefore difference in the association before and after treatment) and that 2. this drug dependent change of expression is also TP53 dependent (therefore more associations after than before). In silico binding sites for TP53 in the promoter of the majority of the genes whose expression was significantly associated with the TP53 codon 72 SNP suggest a potential for this TP53 dependent regulation. The thymidylate synthase inhibitor 5-FU is a constituent of many combinatorial chemotherapy regimens used for breast cancer. In theory, the observed effects could either be caused by 5-FU or Mitomycin C. Considering Mitomycin C to be an antibiotic executing its cytotoxic effects mainly through DNA alkylation, we find this rather unlikely. While our combined statistical analysis did not detect a single significant SNP-expression association between the TP53 codon 72 SNP and expression of genes directly involved in the metabolism and action of the 5-FU drug prior to treatment, the SNP constituted 50% of all significant associations observed after FUMI but fewer associations after doxorubicin treatment. This could reflect the fact that the transcripts included in this analysis were specifically selected to be coding for enzymes involved in 5-FU metabolism. This observation, together with our findings of TP53 binding sites in the promoters of these genes, suggests a participation of TP53 in their regulation and a more direct involvement of TP53 in the cellular response to 5-FU treatment. An allele-dosage response was observed with increased mRNA expression in DHFR, NME1, RRM1 and UMPS and decreased mRNA expression of MTHFR and UPP2 in individuals carrying CC (Arg) compared to CG and GG (Pro) in TP53 codon 72. For ECGF1, TPMT and UGT1A1, the heterozygous genotypes were associated with the lowest or highest expression levels. When the low frequent CC genotype was excluded from the analysis, its associations to DHFR and NME1 expression was further strengthened, while the finding for RRM1, TPMT, UGT1A1 and UPP2 became less significant, and 2 new significant associations were detected; the expression of RRM2 and UCK1, both containing a TP53 transcription factor binding site in their promoter. The codon 72 alleles encode an arginine amino acid with a positive-charged basic side chain and a proline residue with a nonpolar-aliphatic side chain. The polymorphism resides in a proline-rich region of the protein required for the growth suppression and apoptosis mediated by TP53, but not for cell cycle arrest (reviewed in49). Marin et al. suggested that the polymorphism influences mutant TP53s ability to form stable complexes with TP73 (a homologue of TP53), correlating with a loss of TP53–73 DNA binding capability.50 Boyer et al. observed that several of the transcripts found by whole genome expression microarray on MCF7 breast cancer cells induced by 5-FU treatment were abolished by the inactivation of wt TP53,51 suggesting that several of the genes activated by 5-FU treatment are dependent on the presence of wt TP53 for induction. However, the genes involved in 5-FU metabolism essayed in our study did not show an expression pattern associated with mutation status of TP53 after treatment. The increase in number of significant SNP-expression associations for the TP53 codon 72 SNP after 5-FUMI treatment described in our study may suggest a post-treatment specific role for TP53, in addition to that associated to its mutation status. This implies that there may be a direct regulatory role of TP53 on the 5-FU metabolites and may provide an alternative role of TP53 in treatment response in addition to the more established role as regulator of apoptosis during 5-FU treatment. The presence of TP53 binding sites in the promoter of many of these genes reported in the manuscript strengthens this hypothesis.

Our results are based on a small 5-FU treated sample set, and should therefore be confirmed in a larger, independent set, with a more quantitative method of mRNA detection. However, our results suggest that (i) the previously reported association between 3 copies of the 28 bp repeat in TYMS promoter and reduced survival seen in colorectal cancer also holds true for breast cancer patients treated with 5-FU in a neoadjuvant setting and (ii) TP53 is a central player of 5-FU metabolism as the different alleles have various effect on the regulation of the expression level of several of key players involved in the metabolism of 5-FU.


  1. Top of page
  2. Abstract
  3. Material and methods
  4. Results
  5. Discussion
  6. References
  7. Supporting Information
  • 1
    Longley DB,Harkin DP,Johnston PG. 5-fluorouracil: mechanisms of action and clinical strategies. Nat Rev Cancer 2003; 3: 3308.
  • 2
    Bunz F,Hwang PM,Torrance C,Waldman T,Zhang Y,Dillehay L,Williams J,Lengauer C,Kinzler KW,Vogelstein B. Disruption of p53 in human cancer cells alters the responses to therapeutic agents. JClin Invest 1999; 104: 2639.
  • 3
    Heggie GD,Sommadossi JP,Cross DS,Huster WJ,Diasio RB. Clinical pharmacokinetics of 5-fluorouracil and its metabolites in plasma, urine, and bile. Cancer Res 1987; 47: 22036.
  • 4
    van Kuilenburg AB,Muller EW,Haasjes J,Meinsma R,Zoetekouw L,Waterham HR,Baas F,Richel DJ,van Gennip AH. Lethal outcome of a patient with a complete dihydropyrimidine dehydrogenase (DPD) deficiency after administration of 5-fluorouracil: frequency of the common IVS14+1G>A mutation causing DPD deficiency. Clin Cancer Res 2001; 7: 114953.
  • 5
    van Kuilenburg AB,Vreken P,Beex LV,Meinsma R,Van LH,De Abreu RA,van Gennip AH. Heterozygosity for a point mutation in an invariant splice donor site of dihydropyrimidine dehydrogenase and severe 5-fluorouracil related toxicity. Eur J Cancer 1997; 33: 225864.
  • 6
    Scherf U,Ross DT,Waltham M,Smith LH,Lee JK,Tanabe L,Kohn KW,Rein hold WC,Myers TG,Andrews DT,Scudiero DA,Eisen MB, et al. A gene expression database for the molecular pharmacology of cancer. Nat Genet 2000; 24: 23644.
  • 7
    Kawakami K,Omura K,Kanehira E,Watanabe Y. Polymorphic tandem repeats in the thymidylate synthase gene is associated with its protein expression in human gastrointestinal cancers. Anticancer Res 1999; 19: 324952.
  • 8
    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 1995; 20: 1917.
  • 9
    Pullarkat ST,Stoehlmacher J,Ghaderi V,Xiong YP,Ingles SA,Sherrod A,Warren R,Tsao-Wei D,Groshen S,Lenz HJ. Thymidylate synthase gene polymorphism determines response and toxicity of 5-FU chemotherapy. Pharmacogenomics J 2001; 1: 6570.
  • 10
    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 2001; 85: 82730.
  • 11
    Lecomte T,Ferraz JM,Zinzindohoue F,Loriot MA,Tregouet DA,Landi B,Berger A,Cugnenc PH,Jian R,Beaune P,Laurent-Puig P. Thymidylate synthase gene polymorphism predicts toxicity in colorectal cancer patients receiving 5-fluorouracil-based chemotherapy. Clin Cancer Res 2004; 10: 58808.
  • 12
    Tsuji T,Hidaka S,Sawai T,Nakagoe T,Yano H,Haseba M,Komatsu H,Shindou H,Fukuoka H,Yoshinaga M,Shibasaki S,Nanashima A, et al. Polymorphism in the thymidylate synthase promoter enhancer region is not an efficacious marker for tumor sensitivity to 5-fluorouracil-based oral adjuvant chemotherapy in colorectal cancer. Clin Cancer Res 2003; 9: 37004.
  • 13
    Etienne MC,Ilc K,Formento JL,Laurent-Puig P,Formento P,Cheradame S,Fischel JL,Milano G. Thymidylate synthase and methylenetetrahydrofolate reductase gene polymorphisms: relationships with 5-fluorouracil sensitivity. Br J Cancer 2004; 90: 52634.
  • 14
    Ishida Y,Kawakami K,Tanaka Y,Kanehira E,Omura K,Watanabe G. Association of thymidylate synthase gene polymorphism with its mRNA and protein expression and with prognosis in gastric cancer. Anticancer Res 2002; 22: 28059.
  • 15
    Ulrich CM,Bigler J,Velicer CM,Greene EA,Farin FM,Potter JD. Searching expressed sequence tag databases: discovery and confirmation of a common polymorphism in the thymidylate synthase gene. Cancer Epidemiol Biomarkers Prev 2000; 9: 13815.
  • 16
    Mandola MV,Stoehlmacher J,Zhang W,Groshen S,Yu MC,Iqbal S,Lenz HJ,Ladner RD. A 6 bp polymorphism in the thymidylate synthase gene causes message instability and is associated with decreased intratumoral TS mRNA levels. Pharmacogenetics 2004; 14: 31927.
  • 17
    Lecomte T,Ferraz JM,Zinzindohoue F,Loriot MA,Tregouet DA,Landi B,Berger A,Cugnenc PH,Jian R,Beaune P,Laurent-Puig P. Thymidylate synthase gene polymorphism predicts toxicity in colorectal cancer patients receiving 5-fluorouracil-based chemotherapy. Clin Cancer Res 2004; 10: 58808.
  • 18
    Scott J,Weir D. Folate/vitamin B12 inter-relationships. Essays Biochem 1994; 28: 6372.
  • 19
    Frosst P,Blom HJ,Milos R,Goyette P,Sheppard CA,Matthews RG,Boers GJ,den Heijer M,Kluijtmans LA,van den Heuvel LP. A candidate genetic risk factor for vascular disease: a common mutation in methylenetetrahydrofolate reductase. Nat Genet 1995; 10: 1113.
  • 20
    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 1998; 64: 16972.
  • 21
    Cohen V,Panet-Raymond V,Sabbaghian N,Morin I,Batist G,Rozen R. Methylenetetrahydrofolate reductase polymorphism in advanced colorectal cancer: a novel genomic predictor of clinical response to fluoropyrimidine-based chemotherapy. Clin Cancer Res 2003; 9: 16115.
  • 22
    Etienne MC,Formento JL,Chazal M,Francoual M,Magne N,Formento P,Bourgeon A,Seitz JF,Delpero JR,Letoublon C,Pezet D,Milano G. Methylenetetrahydrofolate reductase gene polymorphisms and response to fluorouracil-based treatment in advanced colorectal cancer patients. Pharmacogenetics 2004; 14: 78592.
  • 23
    Jakobsen A,Nielsen JN,Gyldenkerne N,Lindeberg J. Thymidylate synthase and methylenetetrahydrofolate reductase gene polymorphism in normal tissue as predictors of fluorouracil sensitivity. J Clin Oncol 2005; 23: 13659.
  • 24
    Wisotzkey JD,Toman J,Bell T,Monk JS,Jones D. MTHFR (C677T) polymorphisms and stage III colon cancer: response to therapy. Mol Diagn 1999; 4: 959.
  • 25
    Goto Y,Yue L,Yokoi A,Nishimura R,Uehara T,Koizumi S,Saikawa Y. A novel single-nucleotide polymorphism in the 3′-untranslated region of the human dihydrofolate reductase gene with enhanced expression. Clin Cancer Res 2001; 7: 19526.
  • 26
    Geisler S,Borresen-Dale AL,Johnsen H,Aas T,Geisler J,Akslen LA,Anker G,Lonning PE. TP53 gene mutations predict the response to neoadjuvant treatment with 5-fluorouracil and mitomycin in locally advanced breast cancer. Clin Cancer Res 2003; 9: 55828.
  • 27
    Geisler S,Lonning PE,Aas T,Johnsen H,Fluge O,Haugen DF,Lillehaug JR,Akslen LA,Borresen-Dale AL. Influence of TP53 gene alterations and c-erbB-2 expression on the response to treatment with doxorubicin in locally advanced breast cancer. Cancer Res 2001; 61: 250512.
  • 28
    Aas T,Borresen AL,Geisler S,Smith-Sorensen B,Johnsen H,Varhaug JE,Akslen LA,Lonning PE. Specific P53 mutations are associated with de novo resistance to doxorubicin in breast cancer patients. Nat Med 1996; 2: 8114.
  • 29
    Lonning PE. Study of suboptimum treatment response: lessons from breast cancer. Lancet Oncol 2003; 4: 17785.
  • 30
    Page DL,Ellis IO,Elston CW. Histologic grading of breast cancer. Let's do it. Am J Clin Pathol 1995; 103: 1234.
  • 31
    Helle SI,Ekse D,Holly JM,Lonning PE. The IGF-system in healthy pre- and postmenopausal women: relations to demographic variables and sex-steroids. J Steroid Biochem Mol Biol 2002; 81: 95102.
  • 32
    Keen-Kim D,Grody WW,Richards CS. Microelectronic array system for molecular diagnostic genotyping: nanogen nanochip 400 and molecular biology workstation. Expert Rev Mol Diagn 2006; 6: 28794.
  • 33
    Gilles PN,Wu DJ,Foster CB,Dillon PJ,Chanock SJ. Single nucleotide polymorphic discrimination by an electronic dot blot assay on semiconductor microchips. Nat Biotechnol 1999; 17: 36570.
  • 34
    Lahermo P,Liljedahl U,Alnaes G,Axelsson T,Brookes AJ,Ellonen P,Groop PH,Hallden C,Holmberg D,Holmberg K,Keinanen M,Kepp K, et al. A quality assessment survey of SNP genotyping laboratories. Hum Mutat 2006; 27: 7114.
  • 35
    von Ahsen N,Oellerich M,Schutz E. DNA base bulge vs unmatched end formation in probe-based diagnostic insertion/deletion genotyping: genotyping the UGT1A1 (TA)(n) polymorphism by real-time fluorescence PCR. Clin Chem 2000; 46: 193945.
  • 36
    Kristensen V,Andersen TI,Erikstein B,Geitvik G,Skovlund E,Nesland JM,Borresen-Dale AL. Single tube multiplex polymerase chain reaction genotype analysis of GSTM1, GSTT1 and GSTP1: relation of genotypes to TP53 tumor status and clinicopathological variables in breast cancer patients. Pharmacogenetics 1998; 8: 4417.
  • 37
    Sorlie T,Tibshirani R,Parker J,Hastie T,Marron JS,Nobel A,Deng S,Johnsen H,Pesich R,Geisler S,Demeter J,Perou CM, et al. Repeated observation of breast tumor subtypes in independent gene expression data sets. Proc Natl Acad Sci USA 2003; 100: 841823.
  • 38
    Tsalenko A,Sharan R,Edvardsen H,Kristensen V,Borresen-Dale AL,Ben-Dor A,Yakhini Z. Analysis of SNP-expression association matrices. Proc IEEE Comput Syst Bioinform Conf 2005; 4: 13543.
  • 39
    Marinescu VD,Kohane IS,Riva A. MAPPER: a search engine for the computational identification of putative transcription factor binding sites in multiple genomes. BMC Bioinformatics 2005; 6: 79.
  • 40
    Stephens M,Smith NJ,Donnelly P. A new statistical method for haplotype reconstruction from population data. Am J Hum Genet 2001; 68: 97889.
  • 41
    Barrett JC,Fry B,Maller J,Daly MJ. Haploview: analysis and visualization of LD and haplotype maps. Bioinformatics 2005; 21: 2635.
  • 42
    Gabriel SB,Schaffner SF,Nguyen H,Moore JM,Roy J,Blumenstiel B,Higgins J,DeFelice M,Lochner A,Faggart M,Liu-Cordero SN,Rotimi C, et al. The structure of haplotype blocks in the human genome. Science 2002; 296: 22259.
  • 43
    Nordgard SH,Ritchie MD,Jensrud SD,Motsinger AA,Alnaes GI,Lemmon G,Berg M,Geisler S,Moore JH,Lonning PE,Borresen-Dale AL,Kristensen VN. ABCB1 and GST polymorphisms associated with TP53 status in breast cancer. Pharmacogenet Genomics 2007; 17: 12736.
  • 44
    Pare L,Altes A,Cajal TR,Del RE,Alonso C,Sedano L,Barnadas A,Baiget M. Influence of thymidylate synthase and methylenetetrahydrofolate reductase gene polymorphisms on the disease-free survival of breast cancer patients receiving adjuvant 5-fluorouracil/methotrexate-based therapy. Anticancer Drugs 2007; 18: 8215.
  • 45
    Mandola MV,Stoehlmacher J,Muller-Weeks S,Cesarone G,Yu MC,Lenz HJ,Ladner RD. A novel single nucleotide polymorphism within the 5′ tandem repeat polymorphism of the thymidylate synthase gene abolishes USF-1 binding and alters transcriptional activity. Cancer Res 2003; 63: 2898904.
  • 46
    Saga Y,Suzuki M,Mizukami H,Kohno T,Takei Y,Fukushima M,Ozawa K. Overexpression of thymidylate synthase mediates desensitization for 5-fluorouracil of tumor cells. Int J Cancer 2003; 106: 3246.
  • 47
    Yu Z,Sun J,Zhen J,Zhang Q,Yang Q. Thymidylate synthase predicts for clinical outcome in invasive breast cancer. Histol Histopathol 2005; 20: 8718.
  • 48
    Johnson MR,Hageboutros A,Wang K,High L,Smith JB,Diasio RB. Life-threatening toxicity in a dihydropyrimidine dehydrogenase-deficient patient after treatment with topical 5-fluorouracil. Clin Cancer Res 1999; 5: 200611.
  • 49
    Hainaut P,Hollstein M. p53 and human cancer: the first ten thousand mutations. Adv Cancer Res 2000; 77: 81137.
  • 50
    Marin MC,Jost CA,Brooks LA,Irwin MS,O'Nions J,Tidy JA,James N,McGregor JM,Harwood CA,Yulug IG,Vousden KH,Allday MJ, et al. A common polymorphism acts as an intragenic modifier of mutant p53 behaviour. Nat Genet 2000; 25: 4754.
  • 51
    Boyer J,Maxwell PJ,Longley DB,Johnston PG. 5-fluorouracil: identification of novel downstream mediators of tumour response. Anticancer Res 2004; 24: 41723.

Supporting Information

  1. Top of page
  2. Abstract
  3. Material and methods
  4. Results
  5. Discussion
  6. References
  7. Supporting Information

This article contains supplementary material available via the Internet at .

ijc23541-IJC-07-2447SupFig1Nanogen.tif3659KSupporting Information file ijc23541-IJC-07-2447SupFig1Nanogen.tif
ijc23541-IJC-07-2447SupFig2LDHaplotypes.tif6196KSupporting Information file ijc23541-IJC-07-2447SupFig2LDHaplotypes.tif
ijc23541-IJC-07-2447SupplementaryTableI.doc390KSupporting Information file ijc23541-IJC-07-2447SupplementaryTableI.doc
ijc23541-IJC-07-2447SupplementaryTableII.doc234KSupporting Information file ijc23541-IJC-07-2447SupplementaryTableII.doc
ijc23541-IJC-07-2447SupplementaryTableIII.doc144KSupporting Information file ijc23541-IJC-07-2447SupplementaryTableIII.doc

Please note: Wiley Blackwell is not responsible for the content or functionality of any supporting information supplied by the authors. Any queries (other than missing content) should be directed to the corresponding author for the article.