Genetic polymorphisms associated with increased risk of developing chronic myelogenous leukemia

Abstract

Little is known about inherited factors associated with the risk of developing chronic myelogenous leukemia (CML). We used a dedicated DNA chip containing 16 561 single nucleotide polymorphisms (SNPs) covering 1 916 candidate genes to analyze 437 CML patients and 1 144 healthy control individuals. Single SNP association analysis identified 139 SNPs that passed multiple comparisons (1% false discovery rate). The HDAC9, AVEN, SEMA3C, IKBKB, GSTA3, RIPK1 and FGF2 genes were each represented by three SNPs, the PSM family by four SNPs and the SLC15A1 gene by six. Haplotype analysis showed that certain combinations of rare alleles of these genes increased the risk of developing CML by more than two or three-fold. A classification tree model identified five SNPs belonging to the genes PSMB10, TNFRSF10D, PSMB2, PPARD and CYP26B1, which were associated with CML predisposition. A CML-risk-allele score was created using these five SNPs. This score was accurate for discriminating CML status (AUC: 0.61, 95%CI: 0.58-0.64). Interestingly, the score was associated with age at diagnosis and the average number of risk alleles was significantly higher in younger patients. The risk-allele score showed the same distribution in the general population (HapMap CEU samples) as in our control individuals and was associated with differential gene expression patterns of two genes (VAPA and TDRKH). In conclusion, we describe haplotypes and a genetic score that are significantly associated with a predisposition to develop CML. The SNPs identified will also serve to drive fundamental research on the putative role of these genes in CML development.

Trial registration: ClinicalTrials.gov NCT00219739.

Keywords: CML; SNPs; genetic predisposition; myeloid leukemia.

Conflict of interest statement

CONFLICTS OF INTEREST

All the authors read and approved the final manuscript and declare no conflicts of interest.

Figures

Figure 1. Manhattan Plot presenting p-values (−log10 scale) for the association between CML patients and controls for the 13 937 SNPs passing quality control criteria

The horizontal line indicates the 1% FDR significance threshold.

Figure 2. SNP selection procedure and prediction of the probability of developing CML

A)A classification tree was developed to discriminate CML versus control individuals, using the SNPs obtained from a stepwise multivariate logistic model from the 139 selected SNPs in the single analysis. As an example, individuals carrying two copies of a minor allele B of SNP rs376864 are more likely to develop CML than individuals carrying the normal allele and hence were classified as CML patients. For individuals carrying at least one normal allele, the SNP rs14178 was also taken into account; individuals having normal alleles for rs14178 and rs6668196 were classified as controls, however, individuals having a normal allele for rs14178 and two rare alleles for rs6668196 were classified as CML. B) Variable importance plot used to create the classification tree. The barplot shows a large decay of importance from the sixth SNP, so only the first five SNPs were used to classify individuals (i.e. in the classification tree in A). C) Distribution of the genetic score (sum of risk alleles) created using the five selected SNPs obtained from the classification tree methodology. D) ROC curve and its AUC value.

Figure 3. Average number of risk alleles as a function of age at diagnosis

Genetic risk alleles are more frequent in younger patients, with risk starting to increase again for older patients (>75 years). The X-axis depicts the number of risk alleles grouped in different categories (0, 1, 2, 3, +4 alleles). The right Y-axis represents the number of individuals (bars) while the left Y-axis shows the mean age (dots) for each risk score category.