Inborn Genomic Variation as Predisposing Factors to Myeloid Malignancies.

Abstract 1590

Poster Board I-616

Single nucleotide polymorphism arrays (SNP-A) are increasingly accepted as a karyotyping tool, but with systematic application of this technology to study clonal somatic chromosomal lesion came a realization as to how widespread are copy number variations (CNV) and copy-number neutral runs of homozygosity (ROH) through the human genome. Recognition of recurrent CNVs and ROH is of importance for distinction of truly somatic lesions but clearly both ROH and CNVs themselves could constitute risk factors for development of MDS and AML prediction as sites frequently affected by these changes could constitute fragile sites for chromosomal breaks and recombination events. Alternatively these inherited genomic variants could affect expression of corresponding genes and alter regulatory elements. They can physically disrupt a gene, potentially creating new isoforms and can have a subtle effect on gene expression, either through variable penetrance of gene dosage effects or through interaction between the CNV/ROH and the genetic background. As all ROH is unlikely to be explained by autozygosity, meiotic errors or early embryonic mitotic events could be responsible. The role of CNVs in disease, in particular immune-related disorders, has become increasingly apparent, while acquired ROH is well-recognized as playing a role in carcinogenesis and malignant evolution. However, CNVs and inborn ROH as predisposing factors in myeloid malignancies has not been explored in depth. SNP-A (for example Affymetrix 6.0 arrays) offer an opportunity to systematically investigate not only somatic unbalanced translocations but also copy neutral loss of heterozygosity (CN-LOH), including ROH.

To determine how CNVs and inborn ROH may affect predisposition to myeloid malignancies, we first studied an large cohort of control individuals (N=995) from internal and publicly-available sources. We identified 261 CNVs distributed across the entire genome; 15 were unique to our cohort and had not been previously reported. The remaining CNVs were verified against the Database of Genomic Variants (http://projects.tcag.ca/variation/) and their frequency was established. We also identified 153 non-clonal regions of ROH in the normal cohort (9.8%), distributed across all chromosomes and mapped frequently occurring ROH. No correlation was found between chromosome size and size of ROH on that chromosome. ROH could be divided into two groups: interstitial (N=147) and telomeric (N=6). Interstitial ROH ranged in size from 0.29-64.9Mb with a median of 7.2Mb; the rare telomeric germline ROH was 2.5-13.2Mb with a median of 5.8Mb. There were 30 recurrent regions of ROH; the most frequent (at 0.4%) were at 4q13.1-q13.3 and 5q15-q21.1. We concluded that interstitial ROH<24.5Mb and telomeric ROH<13.2Mb most likely reflected a germline event.

Using this information as a reference we next investigated a large cohort of patients with MDS, MDS/MPN (N=395) and secondary AML (N=130). We identified several CNVs significantly overrepresented in each cohort. For MDS, gains at 3q26.1 (p<.0001), 4q13.2 (p<.0001), 12p11.1 (p<.0001), 15q11.2 (p=.015) 22q11.22 (p=.0008) and losses at 14q11.2 (p<.0001) and 1q21 (p<.0001) were significantly more frequent. In sAML, gain at 22q11.22 (p<.0001) and loss of 4q13.2, 7q34, 8p11.23 and 14q11.2 (all p<.0001) were more frequent than in normal controls. Regions of genomic structural instability, such as fragile sites or segmental duplications, are commonly associated with CNVs. The genomic architecture of each region was investigated with the UCSC Genome Browser (http://www.genome.ucsc.edu/cgi-bin/hgGateway). Overrepresented CNV were associated with regions harboring segmental duplications (1q21, 4q13.2, 12p11.1, 15q11.2, 22q11.22) or repetitive gene families (T cell receptor loci; 7p14.1, 7q34, 14q11.2). We also identified germline ROH in the patient cohorts. ROH at 11q14.1-q14.2 was more frequent in patients with MDS and sAML (6% MDS; 12% AML vs 0%, p=.02); these patients had high-risk disease and a normal karyotype or one chromosomal abnormality. In sAML, several regions were enriched in patients including 1q25.2-q25.3, 13q11q12.13 and 21q21.1 (all N=2; p=0.01).

We conclude, genomic variations once thought to be pathologically silent is now recognized as potential predisposing pathogenic factor in disease; we have demonstrate that specific CNVs and ROH can be associated with myeloid malignancies. It is possible that the presence of CNV or LOH predisposes to somatic chromosomal aberrations and points towards fragile sites.