Mutation Research 602 (2006) 175–181

GSTM1 and GSTT1 polymorphisms as modifiers of age atdiagnosis of hereditary nonpolyposis colorectal cancer

(HNPCC) in a homogeneous cohort of individualscarrying a single predisposing mutation

Rebecca Felix a, Walter Bodmer b, Nicola S. Fearnhead b,Lize van der Merwe c, Paul Goldberg d, Rajkumar S. Ramesar a,∗

a MRC Research Unit for Human Genetics, Division of Human Genetics, Institute for Infectious Diseases and Molecular Medicine,Faculty of Health Sciences, University of Cape Town Medical School, Anzio Road, Observatory 7925, South Africa

b Cancer Research UK, Cancer Immunogenetics Laboratory, Institute of Molecular Medicine, John Radcliffe Hospital,Oxford OX3 9DS, United Kingdom

c Biostatistics Unit, Medical Research Council, Box 19070, Tygerberg 7505, South Africad Surgical Gastroenterology Unit, Groote Schuur Hospital, University of Cape Town, South Africa

Received 31 August 2005; received in revised form 8 September 2006; accepted 19 September 2006

Abstract

The variability in phenotype that occurs for so-called ‘single-gene disorders’ may be because of germline alterations in numerousprimary and “modifier” genes. Within HNPCC families harbouring the same primary predisposing mutation, differences exist inthe site of cancer, age of onset of disease symptoms and, consequently, survival until diagnosis of disease. The current studyinvestigated a cohort of 129 individuals, from 13 different families, who harbour the identical nonsense mutation (C1528T) in thehMLH1 gene, predisposing them primarily to Lynch I syndrome. This cohort was screened for previously described polymorphismsin the glutathione-S-transferase genes, viz. GSTT1 and GSTM1. Male null carriers for both GSTT1 and GSTM1 were approximatelythree times more at risk of developing cancer at an earlier age when compared to non-null males. This work, particularly because ofthe relatively large “homogeneous” primary mutation cohort, provides evidence that genotypic changes distinct from the primary‘HNPCC-causing’ mutation, influence the survival period until diagnosis of disease. It provides an impetus for expanding the studyto include a wider range of candidate modifier genes. Such work may potentially lead to the development of individualised interval

screening regimens for individuals with varying modifier genotypes—an attractive option in a resource-poor country.© 2006 Published by Elsevier B.V.

Keywords: Hereditary nonpolyposis colorectal cancer (HNPCC); Modifier ge

∗ Corresponding author. Tel.: +27 21 406 6297;fax: +27 21 406 6826.

E-mail address: rr@cormack.uct.ac.za (R.S. Ramesar).

0027-5107/$ – see front matter © 2006 Published by Elsevier B.V.doi:10.1016/j.mrfmmm.2006.09.004

nes; Glutathione-S-transferases (GSTT1, GSTM1)

1. Introduction

In the field of medical genetics, it has become increas-

ingly apparent that few, if any, human diseases arehomogeneous and solely the result of one mutation ina single-gene. Even when a Mendelian disorder has oneobvious predisposing genetic cause, the phenotype may

Resear

176 R. Felix et al. / Mutation

still be subject to wide variation. The lack of a clearobserved genotype/phenotype association even in suchsingle-gene disorders suggests that additional modifierfactors, including both environmental and genetic com-ponents, influence clinical phenotypes [1]. A “geneticmodifier” is an inherited genetic variation that leads toa qualitative or significant quantitative difference in theprimary disease phenotype [2]. These variations are dis-tinct from the disease locus and may play a role as lowpenetrance risk factors [3].

Hereditary nonpolyposis colorectal cancer (HNPCC)is an autosomal dominant disorder which results frommutations in certain DNA mismatch repair (MMR)genes. Individuals from families diagnosed with HNPCCexhibit phenotypic variation with respect to the age atdiagnosis of disease symptoms, the site of the cancerand the rate of recurrence of cancer after treatment, evenwhen carrying identical mutations in the MMR genes.A large proportion of the phenotypic variation observedbetween individuals who harbour the same predispos-ing mutation in these genes could be indicative of theinfluence of modifying genes.

A cohort of 129 individuals, all carrying the sameprimary disease-causing mutation, was investigated inthe present study. These individuals originate from 13independently recruited families, who harbour the iden-tical nonsense mutation (C1528T) in exon 13 of hMLH1.We have independently shown, using haplotype analy-sis that this mutation and its spread along the west coastof South Africa is due to a founder effect (submittedfor publication). The mutation has already been used toprovide predictive testing and influences clinical man-agement in South Africa, as previously described [4–6].The individuals in this cohort manifest a variable age atdiagnosis of the disease.

Variations in genes known to be involved in carcino-gen metabolism such as the glutathione-S-transferases(GSTM1 and GSTT1) may explain some of the differ-ences observed in disease expression within HNPCCfamilies. The GSTs play a role in the detoxification ofa wide variety of potentially toxic carcinogenic elec-trophiles by using reduced glutathione to conjugate orto reduce these compounds. This usually inactivates theelectrophiles and facilitates their excretion into urine orbile [7]. Most GST substrates are xenobiotics or prod-ucts of oxidative stress, including some environmentalcarcinogens [8]. In addition to their catalytic activity,GSTs engage in the intracellular transport of endoge-

nous metabolites and steroid hormones [9]. The commonvariant allele for both GSTM1 and GSTT1 is a deletionof the gene. Those individuals who are homozygous forthis deleted allele are said to possess the ‘null’ genotype

ch 602 (2006) 175–181

(0/0) and therefore are unable to produce the enzyme. Inthe present study, we tested the association between thenull genotypes of GSTM1 and GSTT1 and age at diagno-sis of colorectal cancer (CRC) in our HNPCC cohort byusing survival analysis methods. Although the effects ofgenetic modifiers have been reported in colorectal can-cers and even in HNPCC cohorts, we believe that themajor value of the current study is the fact that all ofthe subjects carry the exact same primary disease pre-disposing mutation, representing an ideal resource forexamining the effects of modifying factors.

2. Materials and methods

2.1. Subjects

This work is part of a large-scale molecular genetic andclinical screening project aimed at identifying individuals athigh risk for developing the HNPCC range of cancers (Univer-sity of Cape Town Research Ethics Review Reference number182/99). The cohort in the present study comprised of 129 indi-viduals harbouring the C1528T mutation in exon 13 of hMLH1.This previously described mutation [5,6], causes a truncationof the MLH1 protein. The C1528T-mutation positive researchcohort described here comprised 70 females and 59 males, whowere members of pedigrees derived through genealogical trac-ings from 13 probands. Twenty millilitre venous blood sampleswere collected from each of the individuals under sterile condi-tions in tubes containing the anti-coagulant EDTA. DNA wasprepared using the Genomix blood scale-up kit [Talent, Italy].

2.2. Polymerase chain reaction

The presence of the null genotype in both the GSTM1and GSTT1 genes was investigated in the C1528T researchcohort using the polymerase chain reaction and oligonucleotidesequences as previously described [10,11]. The presence orabsence of the GSTT1 and GSTM1 null alleles was deter-mined by co-amplifying with β-globin gene primers [10]. PCRamplification of GSTM1 and GSTT1 was performed in a totalvolume of 25 �l containing 100 ng genomic DNA, 20 �M ofeach oligonucleotide [10,11] (Table 1), PCR buffer contain-ing 50 mM KCl, 10 mM Tris–HCl (pH 8.3), 1.5 mM MgCl2,200 �M of each dNTP (Gibco, UK), 1 unit of Taq DNA Poly-merase (Gibco, UK). The reactions were overlayed with 20 �lof liquid paraffin and subjected to a denaturation step of 95 ◦Cfor 4 min followed by 30 cycles of 95 ◦C for 30 s, 50, 55 or60◦C for 30 s and 72◦C for 40 s, followed by a final extensionstep at 72 ◦C for 5 min.

2.3. Statistical methods

The dependent variable analysed was age at diagnosisof CRC, that is, the age at which the disease status isfirst ascertained in an individual. Since the C1528T researchcohort included individuals who were not yet diagnosed with

R. Felix et al. / Mutation Research 602 (2006) 175–181 177

Table 1Primer sequences for the PCR amplification of the GSTM1 and GSTT1 polymorphisms

Primer Sequence Size of PCR fragment (bp)

GSTT1F TTCCTTACTGGTCCTCACATCTC461GSTT1R TCACCGGATCATGGCCAGCA

GSTMG5 GAA CTC CCT GAA AAG CTA AAG C215GSTMG6 GTT GGG CTC AAA TAT ACG GTG G

� T TCA� CA GG

C(aedmnwtwwyfc[

cpfmaawstmn

-GlobinF CAA CTT CAT CCA CG-GlobinR GAA GAG CCA AGG A

RC, it was essential to use survival analysis techniquesKaplan–Meier product limit method, Cox proportional haz-rds analysis, Logrank tests). Survival analysis techniquesnable us to use the fact that certain individuals have beenisease-free up to the time of follow-up (censored). The esti-ated disease-free survival rate is the proportion of individuals

ot yet diagnosed at each age. The individuals in the cohortere identified retrospectively as those harbouring the muta-

ion. The date of birth of each individual is the time point fromhich disease-free survival is calculated. Those individualsho died before becoming affected and those who were notet affected on the follow-up date are classified as censoredor the survival analysis. We believe that the observed asso-iations are not spurious consequences of change over time12].

The Kaplan–Meier product-limit method was used to cal-ulate and graph the estimated disease-free survival rate (pro-ortion not yet diagnosed with CRC) at each age (in years),or various genotypes and subsets in the cohort. The estimatededian ages at diagnosis, with 95% confidence intervals are

lso presented. This statistic is a reliable indication of the aver-ge age at diagnosis. Logrank tests were used to determine

hether the observed differences in estimated disease-free

urvival, between groups with common modifiers, were sta-istically significant. The Logrank test was used as a global

easure to test simultaneously for the influence of a combi-ation of genotypes. Because Logrank tests are randomisa-

Fig. 1. Age distribution of the cohort, strati

CC268T AC

tion tests, their results are valid even for related individuals[13].

Cox (proportional hazards) regression was used in order toquantify the relative risk of onset of CRC in one group relativeto (the risk of onset of CRC) a reference group, at any age.The validity of Cox regression depends on the assumption ofproportional hazards in each group, which means that the RRis the same at all ages. In this study, the assumption seems tobe reasonable. We also used Cox regression to quantify the RRfor a specific comparison at any time after adjusting for othercovariates/confounders such as gender [13]. Statistical analy-sis was done in R, a language and environment for statisticalcomputing and graphics. (http://www.R-project.org).

3. Results

3.1. Subjects

The estimated median age at diagnosis of CRC forfemales was 59 years and 44 years for males. The over-all RR for males (versus females) developing diseaseat any specific age is 2.2 (p = 0.015). Fifteen of the 70

females (21%) were diagnosed with CRC, while the agerange for females still alive in this study was 17–78 years(Fig. 1). Twenty-six of the 59 male (44%) C1528T muta-tion carriers were diagnosed with CRC, while the age

fied by gender and affection status.

178 R. Felix et al. / Mutation Research 602 (2006) 175–181

Table 2Distribution of the GST alleles stratified by gender in the C1528Tcohort

Polymorphism Female Male

n f n f

GSTT1Null 32 46 29 49Positive 38 54 30 51

Total 70 100 59 100

GSTM1Null 8 12 10 17Positive 61 88 48 83

bined with GSTM1 was highly significant (p = 0.00068).More of the male null carriers are affected at every agethan any of the other groups. The average risk of develop-ing CRC for male null carriers compared to other carriers

Fig. 2. Kaplan–Meier curves showing the proportion of individualsin the C1528T research cohort without cancer at each age (in years)stratified by gender.

range for males recruited on this study was 22–71 years(Fig. 1).

The estimated proportion of individuals who aredisease-free at each age for the C1528T research cohortis shown in Fig. 2, stratified by gender. The curve showsthe proportion of individuals not yet diagnosed with CRC(y-axis) at each age (x-axis). There is a significant dif-ference between the two survival curves (Logrank test,p = 0.0125). The individuals shown in Fig. 2 representthe entire cohort of males and females (i.e. the affectedand unaffected individuals).

In order to minimise the confounding effect of thestrong gender bias, all analysis was adjusted for gender.

3.2. GSTT1 and GSTM1 and gender

The C1528T research cohort was stratified by genderto determine whether there was a difference in age atdiagnosis for the GSTT1 genotypes. Table 2 shows theGST polymorphism distributions in the C1528T cohort.The GSTT1 null genotype was observed in 29/59 (49%)of the males. In the female members of the cohort, theproportion of null GSTT1 was 32/70 (45%).

Fig. 3 shows the Kaplan–Meier disease-free survivalcurves, stratified by gender and GSTT1 genotype. Therelationship between age at diagnosis and gender com-bined with GSTT1 was highly significant (p = 0.00141).

More of the male null carriers are affected at every agethan any of the other groups. The estimated median ageof GSTT1 null males at diagnosis was 39 years, com-pared to 54 years for those who were positive for GSTT1

Total 69 100 58 100

Number of individuals (n) and percentage (f) given.

(Table 3). The estimated age at diagnosis is 59 for nullfemales. The affected females who were non-null forGSTT1 had an estimated median age at diagnosis of 69years. The average risk of developing CRC for malenull carriers compared to other carriers is summarisedin Table 4. No other risks were significant.

Fig. 4 shows the disease-free survival curves, strati-fied by gender and GSTM1 genotype. GSTM1 null wasfound in 10/58 (17.5%) of males, while 8/69 (10%) offemales under investigation were null for GSTM1. Therelationship between age at diagnosis and gender com-

Fig. 3. Kaplan–Meier curves showing the proportion of individualsin the C1528T research cohort without cancer at each age (in years)stratified by gender and GSTT1 genotype.

R. Felix et al. / Mutation Research 602 (2006) 175–181 179

Table 3Summary statistics for analyses of time to diagnosis, stratified by gender and genotypes

Gender n Affected % Median 0.95 LCL 0.95 UCL

F 70 15 21 59 51 InfM 59 26 44 44 39 Inf

Gender GSTT1 n Affected % Median 0.95 LCL 0.95 UCL

F N 32 7 22 59 47 InfF P 38 8 21 69 49 InfM N 29 16 55 39 37 InfM P 30 10 33 54 44 Inf

Gender GSTM1 n Affected % Median 0.95 LCL 0.95 UCL

F N 8 0 0 Inf Inf InfF P 61 15 25 59 49 InfM N 10 7 70 33 31 InfM P 48 19 40 45 42 Inf

Number in subgroup (n), number diagnosed (affected), percentage diagnosed (%) and estimated median age at diagnosis with 95% CI is presented;female (F) male (M) male, null genotype (N), positive for GSTs (P). “Inf” means that there was not enough information to calculate the estimate.

Table 4Estimated risk of male null GSTT1 carriers compared to other groups

GSTT1 Risk Lower 0.95 Upper 0.95 p-Value

FN 3.19 1.30 7.87 0.0110FM

iGia

Ftg

Table 5Estimated risk of male null GSTM1 carriers compared to other groups

GSTM1 Risk Lower 0.95 Upper 0.95 p-Value

FN Inf 0.00 Inf 1.0000

P 4.10 1.74 9.62 0.0012P 2.70 1.22 5.99 0.0150

s summarised in Table 5. The males who were null forSTM1 had almost three times (2.9) the risk of develop-

ng CRC at any age, when compared to non-null malesnd almost five times (4.89) the risk of developing CRC

ig. 4. Kaplan–Meier curves showing the proportion of individuals inhe C1528T cohort without cancer at each age (in years) stratified byender and GSTM1 genotype.

FP 4.89 1.97 12.14 0.0006MP 2.91 1.20 7.02 0.0180

at any age than female non-null carriers. These werethe only significant risks. The males who were null forGSTM1 had an estimated median age at diagnosis of 33years, 12 years earlier than those who harboured non-null GSTM1 alleles (45 years). Interestingly, there wereno affected females who were null for GSTM1 in theC1528T research cohort, so their estimated disease-freesurvival rate was 1 for ages up to 48, the age of the oldestnull female in the cohort. The median age at diagnosiscould not be estimated for them. The females who werepositive for GSTM1 had a median age at diagnosis of 59years.

4. Discussion

Following the identification of the C1528T mutationin the hMLH1 gene underlying the predisposition to CRC[4–6], it was recognised that initial symptoms presentedat a wide range of ages. The present study is an investiga-tion of specific previously reported polymorphisms in the

glutathione-S-transferase genes. The modifying effectsof the polymorphisms were assessed in subjects (n = 129)who all harbour the identical disease-causing geneticchange in the hMLH1 gene to determine whether these

Resear

180 R. Felix et al. / Mutation

potential modifiers had any effect on the age at diagno-sis. The C1528T cohort is predisposed predominantly tocolorectal disease rather than the Lynch 2 spectrum ofcancers.

Cells contain a variety of enzyme systems that pro-tect them from injury by toxic chemicals and carcino-gens. One such system is the family of glutathione S-transferases (GSTs). The GSTs are present in a numberof species and are also found in the epithelium of thehuman gastrointestinal tract [14]. The GSTs collectivelyplay a critical role in the detoxification of xenobiotics,thus providing a protective role for the cells in the gut[15]. Individuals who are null for GSTM1 and GSTT1have complete absence of activity of these genes [16].Deletions of one or more genes that play a role in detoxi-fication of harmful substances in the colon, together withinheriting the disease-causing mutation, could renderindividuals more susceptible to developing the diseaseat an earlier age.

The current study thus attempted to determinewhether the inheritance of a disease-predisposing muta-tion in a mismatch repair gene, in addition to the pos-session of the null alleles of either GSTT1 or GSTM1,predisposed individuals to develop CRC at an earlier agewhen compared to individuals of the same cohort whoare non-null for these two GST genes.

Various studies have examined the role of GSTs inCRC. The study by Zhong et al. [17] found an associa-tion with the null genotype in GSTM1 and CRC, with arelative risk of 1.8 [17]. Gene frequencies for the GSTM1null genotype range from 23 to 100%, in different eth-nic groups [18,19]. GSTM1 has primarily been studiedin bladder, breast, lung and colon cancers [17,20]. Astudy undertaken by Deakin et al. [21] found a posi-tive association between the GSTT1 null genotype andCRC, with an odds ratio of 1.88 and Katoh et al. [22]found an odds ratio of 1.2. Zhang et al. [23] found thatthe frequency of the GSTT1 null genotype was increasedin CRC patients compared to healthy controls and sug-gested that the absence of GSTT1 might be used as asusceptibility marker for CRC. Other researchers didnot find any significant changes in the frequencies ofGSTT1 or GSTM1 in patients with CRC when comparedwith control samples [22,24]. The strength of the cur-rent study lies in the relatively large cohort of individualswho harbour the same disease-associated mutation in thehMLH1 gene, compared to the above-mentioned studies.

In the current study, 44% of males developed the

disorder, with an estimated median age of diagnosisof 44 years (Table 3). Approximately 21% of femalesdeveloped CRC, with an estimated median age of diag-nosis of 59 years. Thus the percentage of affected males

ch 602 (2006) 175–181

was approximately twice that of affected females, witha younger median age of diagnosis. Undoubtedly animportant future development of this study should beto include the investigation of sex-linked or sex-limitedgenes that are known to play a role in the development ofcancers. These include the androgen-receptor genes [25],as well as the oestrogen receptor alpha gene which hasbeen shown to be a coactivator of the hMSH2 gene [26].

In the present study, the males who presented as nullfor GSTT1 were almost three times more at risk of devel-oping CRC at any age when compared to the males whowere positive. The estimated median age of diagnosis ofthe males who are GSTT1 null was 39 years, comparedto 54 years in the GSTT1-positive group (Table 5). Inaddition, males who were null for GSTM1 also had ayounger median age at diagnosis than those that werenon-null (33 years compared to 45 years, respectively),but this was not statistically significant. This could be dueto the small sample size of this group as only 10 of 59(16.9%) males were homozygous null for GSTM1, withonly seven of the 26 affected males in this group. It ispossible that the level of significance might improve withincreasing sample size as more individuals in this familyare recruited through the current clinical managementprogram. The age at diagnosis for female members ofthis cohort was not influenced by the polymorphisms ineither GSTT1 or GSTM1. None of the eight null femaleswere affected. Thus, as new individuals are added to thegrowing cohort, it would be advisable at least to genotypethe males for the polymorphisms in GSTT1 and GSTM1,given the association with a younger age at diagnosis forthe null genotypes.

The data on the modifiers we have studied, shouldbe considered as a pilot study. The results emphasise thepotential value of the C1528T cohort as a basis for testingfor a range of further potential genetic and environmen-tal modifiers. This information may then contribute toindividualised clinical surveillance and ultimately per-sonalised therapeutics. The study can be expanded by:(i) increasing the cohort size of C1528T carriers, whichis currently underway as families are extended and newfamilies are identified, (ii) investigating additional mod-ifiers and (iii) similar investigations in other cohorts inwhich mutations have been identified.

References

[1] K.M. Dipple, E.R. McCabe, Modifier genes convert “simple”

Mendelian disorders to complex traits, Mol. Genet. Metab. 71(1–2) (2000) 43–50.

[2] H.H. Muller, K. Heinimann, Z. Dobbie, Genetics of hereditarycolon cancer—a basis for prevention? Eur. J. Cancer 36 (10)(2000) 1215–1223.

Resear

[

[

[

[

[

[

[

[

[

[

[

[

[

[

[

[

R. Felix et al. / Mutation

[3] R.S. Houlston, I.P. Tomlinson, Polymorphisms and colorectaltumor risk, Gastroenterology 121 (2) (2001) 282–301.

[4] P.A. Goldberg, et al., In a resource-poor country, mutation identi-fication has the potential to reduce the cost of family managementfor hereditary nonpolyposis colorectal cancer, Dis. Colon. Rec-tum. 41 (10) (1998) 1250–1253 (discussion 1253–1255).

[5] R.S. Ramesar, et al., Molecular genetics improves the manage-ment of hereditary non-polyposis colorectal cancer, S. Afr. Med.J. 90 (7) (2000) 709–714.

[6] P.A. Goldberg, et al., Inherited colon cancers, S. Afr. Med. J. 90(7) (2000) 703–704.

[7] R. Whalen, T.D. Boyer, Human glutathione S-transferases, Semin.Liver. Dis. 18 (4) (1998) 345–358.

[8] J.D. Hayes, D.J. Pulford, The glutathione S-transferase supergenefamily: regulation of GST and the contribution of the isoen-zymes to cancer chemoprotection and drug resistance, Crit. Rev.Biochem. Mol. Biol. 30 (6) (1995) 445–600.

[9] A. Gsur, et al., Polymorphisms of glutathione-S-transferase genes(GSTP1, GSTM1 and GSTT1) and prostate-cancer risk, Int. J.Cancer 95 (3) (2001) 152–155.

10] D.A. Bell, et al., Genetic risk and carcinogen exposure: a commoninherited defect of the carcinogen-metabolism gene glutathione S-transferase M1 (GSTM1) that increases susceptibility to bladdercancer, J. Natl. Cancer Inst. 85 (14) (1993) 1159–1164.

11] S. Pemble, et al., Human glutathione S-transferase theta (GSTT1):cDNA cloning and the characterization of a genetic polymor-phism, Biochem. J. 300 (Pt 1) (1994) 271–276.

12] D.G. Altman, J.M. Bland, Time to event (survival) data, BMJ 317(7156) (1998) 468–469.

13] B. Dawson, R.G. T., Basic and Clinical Biostatistics, third ed.,

Lange Medical Books/McGraw-Hill, USA, 2001.

14] W.H. Peters, et al., Biotransformation enzymes in human intes-tine: critical low levels in the colon? GUT 32 (4) (1991) 408–412.

15] H. Nomani, et al., Glutathione S-transferases activity in patientswith colorectal cancer, Clin. Biochem. 38 (7) (2005) 621–624.

[

ch 602 (2006) 175–181 181

16] A. Beeghly, et al., Glutathione S-transferase polymorphisms andovarian cancer treatment and survival, Gynecol Oncol 100 (2)(2006) 330–337.

17] S. Zhong, et al., Relationship between the GSTM1 genetic poly-morphism and susceptibility to bladder, breast and colon cancer,Carcinogenesis 14 (9) (1993) 1821–1824.

18] S. Xu, et al., Characterization of the human class Mu glutathioneS-transferase gene cluster and the GSTM1 deletion, J. Biol. Chem.273 (6) (1998) 3517–3527.

19] S.C. Cotton, et al., Glutathione S-transferase polymorphisms andcolorectal cancer: a HuGE review, Am. J. Epidemiol. 151 (1)(2000) 7–32.

20] J. Seidegard, et al., A glutathione transferase in human leukocytesas a marker for the susceptibility to lung cancer, Carcinogenesis7 (5) (1986) 751–753.

21] M. Deakin, et al., Glutathione S-transferase GSTT1 genotypesand susceptibility to cancer: studies of interactions with GSTM1in lung, oral, gastric and colorectal cancers, Carcinogenesis 17(4) (1996) 881–884.

22] T. Katoh, et al., Glutathione S-transferase M1 (GSTM1) andT1 (GSTT1) genetic polymorphism and susceptibility to gas-tric and colorectal adenocarcinoma, Carcinogenesis 17 (9) (1996)1855–1859.

23] H. Zhang, et al., Glutathione S-transferase T1 and M1 genotypesin normal mucosa, transitional mucosa and colorectal adenocar-cinoma, Int. J. Cancer 84 (2) (1999) 135–138.

24] G. Chenevix-Trench, et al., Glutathione S-transferase M1 and T1polymorphisms: susceptibility to colon cancer and age of onset,Carcinogenesis 16 (7) (1995) 1655–1657.

25] E. Dietzsch, R. Laubscher, M.I. Parker, Esophageal cancer risk

in relation to GGC and CAG trinucleotide repeat lengths in theandrogen receptor gene, Int. J. Cancer 107 (1) (2003) 38–45.

26] O. Wada-Hiraike, et al., The DNA mismatch repair gene hMSH2is a potent coactivator of oestrogen receptor alpha, Br. J. Cancer92 (12) (2005) 2286–2291.