Role of DNA mismatch repair genetic polymorphisms in the risk of childhood acute lymphoblastic leukaemia


  • Géraldine Mathonnet,

    1. Service d'Hématologie-Oncologie, Centre de Cancérologie Charles-Bruneau, Centre de Recherche, Hôpital Sainte-Justine, and Département de Pédiatrie, Université de Montréal, Montréal, Québec, Canada
    Search for more papers by this author
  • Maja Krajinovic,

    1. Service d'Hématologie-Oncologie, Centre de Cancérologie Charles-Bruneau, Centre de Recherche, Hôpital Sainte-Justine, and Département de Pédiatrie, Université de Montréal, Montréal, Québec, Canada
    Search for more papers by this author
  • Damian Labuda,

    1. Service d'Hématologie-Oncologie, Centre de Cancérologie Charles-Bruneau, Centre de Recherche, Hôpital Sainte-Justine, and Département de Pédiatrie, Université de Montréal, Montréal, Québec, Canada
    Search for more papers by this author
  • Daniel Sinnett

    1. Service d'Hématologie-Oncologie, Centre de Cancérologie Charles-Bruneau, Centre de Recherche, Hôpital Sainte-Justine, and Département de Pédiatrie, Université de Montréal, Montréal, Québec, Canada
    Search for more papers by this author

Daniel Sinnett, Centre de Cancérologie Charles Bruneau, Centre de Recherche, Hôpital Sainte-Justine, 3175 Côte Sainte-Catherine, Montréal, Québec, H3T 1C5, Canada. E-mail:


Summary. Acute lymphoblastic leukaemia (ALL) is the most common childhood cancer. Genetic variants in the coding regions of the mismatch repair genes MLH1 (Ile-219Val) and MSH3 (Arg-940Glu and Thr-1036Ala) could contribute to an individual's susceptibility as modifiers in leukaemogenesis. To investigate this possibility, we conducted a case–control study on 287 children with ALL and 320 healthy controls both of French–Canadian origin. MLH1 and MSH3 variants, when taken independently, seemed to play little or no role in the aetiology of childhood ALL. However, when the MLH1 genotypes were combined with genotypes previously shown to influence ALL susceptibility, we found that the MLH1 variant Val-219 further increases the risk of GSTM1 null and CYP1A1*2A genotypes [combined odds ratio (OR) = 6·0, P = 0·002] as well as that of CYP2E1*5 (OR = 15·8, P = 0·001). No association was found with MSH3 variants. This study suggests an association of leukaemogenesis in children with both xenobiotic metabolism and DNA repair, and thus points to the effect of environmental exposure.

Acute lymphoblastic leukaemia (ALL) is the most common paediatric cancer. Its aetiology can be explained by a combination of genetic and environmental factors, especially the latter if active during vulnerable periods of development in fetal life and early infancy. Recently, we showed that certain variants in the carcinogen-metabolizing genes were significantly associated with the risk of ALL in children (Krajinovic et al, 1999, 2000, 2002). These results suggested that genetic factors, especially in the context of environmental exposure, might play an important role in leukaemogenesis.

The mismatch repair (MMR) system recognizes and repairs incorrectly paired nucleotides that arise in DNA as a result of physical damage, replication errors, deamination of cytosines or after heteroduplex formation for crossing over in meiosis (reviewed by Buermeyer et al, 1999; Hsieh, 2001). The importance of MMR in protecting against cell-transforming mutations has been clearly demonstrated in both hereditary and sporadic human cancers (e.g. Eshleman & Markowitz, 1996; Kolodner, 1996). Thus, polymorphisms in genes encoding components of DNA repair machinery might be relevant in determining susceptibility to cancer (e.g. Mohrenweiser & Jones, 1998; Berwick & Vineis, 2000; Miller et al, 2001).

Here, we report a case–control study assessing the relationship between variants in the MMR genes MLH1 and MSH3 (Benachenhou et al, 1998) and the susceptibility to childhood ALL in French–Canadians. In the context of DNA repair polymorphisms, we have re-examined the effect of variation in carcinogen-metabolizing genes by analysis of the combined multilocus genotypes.

Patients and methods

Subjects.  Consecutive incident cases of childhood ALL (n = 287, 167 males and 120 females with a median age of 5 years) were diagnosed in the Division of Hematology-Oncology of Ste-Justine Hospital, Montreal, Canada, between August 1988 and May 2001. The ALL subtypes, determined by immunophenotyping, were as follows: 234 pre-B, 24 T-cell ALL and 29 with undetermined lineage. Controls (n = 320) were carefully selected to be representative of the patients' ethnicity (French–Canadian) and geographical distribution (population served by the Ste-Justine Hospital). The institutional review board approved the research protocol, and informed consent was obtained from all participating individuals and/or their parents.

Genotyping.  DNA was isolated from mouth epithelial cells, peripheral blood or bone marrow in remission using standard procedures. We studied the MLH1 polymorphism Ile-219Val (isoleucine-219 to valine replacement resulting from A to G substitution at position 676; GenBank accession number HSMLH1S08) as well as the MSH3 polymorphism Arg-940Glu (G to A substitution at position 2835; GenBank accession number D61416) and MSH3 Thr-1036Ala polymorphism (A to G transition at position 3124; GenBank accession number D61418). The underlying DNA variants were genotyped by allele-specific oligonucleotide (ASO) hybridization assay as described by Labuda et al (1999). Similarly, all individuals had been genotyped for polymorphisms in the genes GSTM1, CYP1A1, CYP2E1, NQO1 and NAT2 in earlier studies (Krajinovic et al, 1999, 2000, 2002). The participating patients and controls could not be genotyped at all polymorphic loci because of sample limitation, which explains the variation in the total number of subjects in the tables.

Statistical analysis.  Chi-squared test was used to examine the differences in the distribution of genotypes between cases and controls. The level of significance was calculated by Fisher's exact test (two-sided). Odds ratios (OR) are given with 95% confidence intervals (CI). Unconditional multivariate logistic analysis was used to include potential confounding factors in the analysis (age and gender) as well as to compute the effects of combined genotypes and gene–gene interactions. In the latter, we considered the MSH3 and MLH1 polymorphisms together with variants in the GSTM1, CYP1A1, CYP2E1, NQO1 and NAT2 genes that were shown previously to influence the risk of ALL (Krajinovic et al, 1999, 2000, 2002). In the assessment of the combined effects of genotype, all variants were considered as dichotomous variables. The risk was assigned to GSTM1 null, NAT2 slow acetylator genotypes, carriers of CYP1A1*2A, CYP2E1*5, NQO1*2 or *3, MLH1 Ile-219/Ile-219 individuals and carriers of either the Glu-940 or Ala-1036 MSH3 variant. For two- or three-way gene–gene interactions, the full model comprising main effects and all interactions was considered. All analyses were performed using the Statistical Package for the Social Sciences (SPSS) software, version 7·5.


The frequencies of MLH1 and MSH3 alleles as well as the distribution of the corresponding genotypes in ALL patients (n = 287) and in healthy controls (n = 320) are given in Table I. No significant modification of risk was observed for any MLH1 and MSH3 variants, suggesting little or no role in the aetiology of childhood ALL, at least when considered independently. We have reported previously that children with the GSTM1 null genotype or carrying the CYP1A1*2A, CYP2E1*5, NQO1*2/*3 variants or the NAT2 slow acetylator genotypes were at increased risk of ALL (Krajinovic et al, 1999, 2000, 2002). Assuming that individuals who carry more than one risk-elevating genotype, because of a higher mutational burden, could be more susceptible to variations in DNA repair capacity, we investigated whether the associated risk was additionally modulated by MSH3 and MLH1 polymorphisms. The risk associated with CYP1A1 and GSTM1 genotypes was increased further in the presence of the MLH1 Ile-219/Ile-219 genotype (Table II), particularly among those carrying all three at-risk genotypes (OR = 6·0, 95% CI 1·9–18·9, P = 0·002). Similarly, we found an elevated risk in children carrying both the MLH1 Ile-219/Ile-219 genotype and at least one CYP2E1*5 allele (OR = 15·8, 95% CI 2·0–122·6, P < 0·001). The observed combined effect seems to arise from two- (MLH1 and CYP2E1) and three-way (MLH1, GSTM1 and CYP1A1) gene–gene interactions (P = 0·04 and 0·05 respectively). The use of the Bonferroni correction for multiple testing (k = 10, P < 0·005) did not change the interpretation of the results. No association was found with any of the MSH3 variants (data not shown). No other significant statistical interactions were observed between any other combinations of the variables studied here.

Table I.  Distribution of MLH1 and MSH3 genotypes in ALL cases and controls.
LocusGenotypeCasesControlsOR (95% CI)
n * No. (%)n * No. (%)
  • *

    Total number of individuals tested.

  • Number of individuals with a given genotype. Ref, reference value; ND, not determined because of insufficient data.

  • The genotyping assay used could not distinguish between the combined heterozygotes MSH3 Glu-940Thr-1036/Arg-940Ala-1036 and Arg-940Thr-1036/Glu-940-Ala-1036.

MLH1Ile-219/Ile-219 149 (51·9) 154 (48·1)1·0 (Ref.)
Ile-219/Val-219287112 (39·0)320132 (41·3)0·9 (0·6–1·2)
Val-219/Val-219 26 (9·1) 34 (10·6)0·8 (0·5–1·4)
MSH3Arg-940Thr-1036 133 (48·7) 139 (49·5)1·0 (Ref.)
Arg-940Thr-1036 5 (1·8) 6 (2·1)0·9 (0·3–2·9)
Arg-940Thr-1036 44 (16·1) 54 (19·2)0·9 (0·5–1·5)
Arg-940Thr-1036 61 (22·3) 49 (17·4)1·3 (0·8–2·0)
Glu-940Thr-10362732 (0·7)2810 (0)ND
Glu-940Thr-1036 5 (1·8) 4 (1·4)1·3 (0·3–0·5)
Arg-940Ala-1036 11 (4·0) 7 (2·5)1·6 (0·6–4·4)
Arg-940Ala-1036 6 (2·2) 13 (4·6)0·5 (0·2–1·3)
Glu-940Ala-1036 6 (2·2) 9 (3·2)0·7 (0·2–2·0)
Table II.  Combined effects of MLH1, GSTM1 and CYP1A1 genotypes in the risk of childhood ALL.
at risk
GSTM1CYP1A1 * 2A*Cases
(n = 254)
(n = 254)
OR (95%CI)
  • OR, crude odds ratio; CI, confidence interval.

  • *

    – / –, individuals without any copy of a given allele; –/+, +/+, individuals with one or two copies of the corresponding allele.

  • Null, homozygous deletion.

  • P  = 0·002. The reference group (OR = 1·0) was defined as children having ‘low-risk’ genotypes: presence of the MLH1 Val-219 variant and GSTM1, as well as the absence of the CYP1A1

  • *

    * 2A allele.

None – /+, +/+Present – / – 43571 (Ref.)
One – / – Present – / – 51591·1 (0·7–2·0)
– /+, +/+Present – /+, +/+8101·1 (0·4–2·9)
– /+, +/+Null – / – 67581·5 (0·9–2·6)
Two – / – Null – / – 53471·5 (0·9–2·6)
– / – Present – /+, +/+480·7 (0·2–2·3)
– /+, +/+Null – /+, +/+10111·2 (0·5–3·1)
Three – / – Null – /+, +/+1846·0 (1·9–18·9)


In this case–control study carried out in French–Canadians, we found that neither MLH1 nor MSH3 variants, when taken independently, appeared to play a significant role in the aetiology of childhood ALL. However, in a multilocus analysis, the combination of the MLH1 Ile-219/Ile-219 with either CYP2A1*2A and GSTM1 null genotypes or CYP2E1*5 carriers revealed a significant increase in risk of ALL among the carriers. These results support the notion that the aetiology of childhood ALL is more likely to depend upon interactions of many genes from different metabolic pathways, such as carcinogen metabolism and DNA repair. This is supported by a recent report indicating that mutagen sensitivity in lymphocytes of certain individuals could be associated with DNA variants in DNA repair and xenobiotic-metabolizing enzymes (Tuimala et al, 2002). Other mechanisms involving MMR proteins cannot be completely ruled out, because these proteins participate in different vital cellular processes, including homologous recombination (de Wind et al, 1995), G2 checkpoint control (Hawn et al, 1995), recognition of DNA damage and/or apoptosis (Kat et al, 1993), as well as in transcription-coupled repair (Mellon et al, 1996). The functional significance of genetic variants in MLH1 is still unknown, but a possible explanation of MLH1 implication in leukaemogenesis could be related to the overall mutation burden. DNA damage occurs as a result of normal cellular processes (e.g. replication, oxidative metabolism) and/or as a consequence of environmental exposure. The individuals carrying genotypes that permit higher accumulation of carcinogens (e.g. GSTM1 null genotype, CYP1A1*2A or CYP1E1*5 allele) should be more affected by genetic variants that lower DNA repair capacity, thus increasing the mutational burden. Therefore, an intrinsic individual risk of developing childhood leukaemia may depend on the equilibrium between its efficiency in metabolizing genotoxic carcinogens and its capacity to repair DNA damage.


We are grateful to all individuals who kindly consented to provide DNA for this study. D.S. and M.K. are scholars of the Fonds de la Recherche en Santé du Québec. G.M. was the recipient of a studentship from the Fondation Hôpital Ste-Justine/Power Corporation Inc. This work was supported by the Fondation Charles-Bruneau and the Canadian Genetic Diseases Network.