Array-Based Karyotyping and Genotyping Demonstrates a Non Random Selection of Allelic Variants of Genes in Clones with 5q31 Deletion Mutants.

Interstitial deletion of chromosome 5 has been frequently reported in acute myeloid leukemia (AML) and myelodysplastic syndrome (MDS) pointing towards the pathogenic role of this region in phenotype and clonal evolution. To investigate the frequency of 5q LOH and better delineate the commonly deleted region (CDR) we applied metaphase cytogenetics (MC) and performed array-based interphase cytogenetics on DNA using high-density 250K SNP-array (SNP-A) in malignant myeloid disorders. SNP-A analysis was performed using Affymetrix 250K SNP-array on a subset of 250 patients and 118 controls. SNP-A data were analyzed using CNAG v.3.0. Using SNP-A-based karyotyping 5q abnormalities were identified in 42/250 patients vs. 29/250 by MC (17% vs. 11%, primarily as a result of resolving uninformative MC). By SNP-A, previously cryptic lesions other than 5q deletions were identified in 71% of the patients (30/42) and included e.g., upd(1)(p35.1pter), upd(17)(p13.1pter), del(7)(q21.3q36.2). In 2 patients somatic UPD5q was found. In 6 other patients 5q micro-deletions of chromosome 5 were detected likely representing CNVs. Subsequently, we undertook an SNP-A-based definition of the CDR covering approximately 1.92MB and positioned within previously reported regions. This region spans approximately from base pair 137,528,564-139,451,907 and several candidate tumor suppressor genes including those already reported e.g., CTNNA1 and EGR1 as well as CDC25, ETF1 to name a few. As mutations in any of the genes in CDR on 5q were not found, the malignant phenotype can result either from haploinsufficiency or hemizygozity for a pathogenic allele occurring otherwise in heterozygous or major allele in homozygous constellation. Previously, underexpression of RPS14 was demonstrated to be responsible for dysplastic features in 5q- syndrome, but haploinsufficiency of this gene does not explain the clonal evolution and malignant properties of clones with del5q. Utilizing another advantage of SNP-A, the ability to assign genotypes, we also performed analysis of the allele distribution within newly defined CRD on 5q31. We genotyped 987 SNPs located in the common area of LOH (5q31) to determine whether for certain loci allelic distribution in the hemizygous deletion mutant is skewed to one or the other allele (when the original diploid constellation is heterozygous). For the purpose of this study we hypothesized that informative alleles must display a very low frequency in homozygous constellation in controls. To test if the distribution of remaining alleles (and the inherent change of genotype) in the deleted region on 5q is random or skewed towards one allelic variant we applied a multinomial test statistics. Because of the discrete nature of the problem, p-values were calculated by selecting the minimum of the left or right tail of the event spectrum. This value was computed without approximation by using the exact form of the binomial distribution. We used the smaller of the two p-values from the two binomial models as our reported p-value. SNPs present in ARHGAP26 locus appeared to be most significantly skewed (p<.004 for 3 SNP loci within gene: rs2028268, rs3776227, rs17100139). For the alleles located in this gene one would expect that 14% of patient will be hemizygous for minor and 96% for major allele respectively. However, we have observed that 57% of clones showed minor constellation. ARHGAP26 and its function as a tumor suppressor may indicate involvement in the pathology and disease progression. In sum, we demonstrated that a combined genotypic and cytogenetic analysis is possible using SNP-A to investigate areas of LOH for the presence of genotypes which may convey selection advantage.