Somatic and Germline Copy Neutral Loss of Heterozygosity Are Common in Chronic Lymphocytic Leukemia

Abstract 4567

Chronic lymphocytic leukemia (CLL) is the most common leukemia of adults, and prognosis is still difficult to predict, although cytogenetic abnormalities identified by FISH are most helpful. Isolated reports have suggested that copy neutral loss of heterozygosity (cnLOH) can involve 13q and 17p in CLL, but the extent and the impact on clinical outcome is not well established.

We therefore embarked upon characterization of cnLOH in a large set of 230 CLLs with matched normal DNA. The median age at diagnosis of CLL in this patient population was 54 (33–79). 87% of patients were Rai 0–1 at diagnosis, and 79% were chemotherapy naive at sampling. 121 of 230 patients were treated, with a median TTFT of 42 months. The median follow-up for surviving patients is 74 months. 44% of patients carried one somatic copy number abnormality (CNA) by SNP array, 20% two, 7% three, 5% four and 4% more than five.

cnLOH was called by the Affymetrix Genotyping Console Software, which evaluates each SNP for copy number and then subtracts the A allele value from the B allele value within an individual sample, thereby allowing independent evaluation of tumor (somatic) and normal (germline). All calls were manually reviewed. A size cut-off of 1.0 Mb was used to determine significant cnLOH events. In total, of 230 patients, we found 26 events of somatic cnLOH (11%) and 36 events of germline cnLOH (16%), affecting 56 separate patients (24%). This frequency of cnLOH was surprisingly high and suggested that cnLOH might be an alternative mechanism affecting known loci in CLL. This was the case, as the most common events overall involved 13q in 25 patients, the X chromosome in 9 patients, chromosomes 17 and 18 in five patients each, and chromosomes 9, 11 and 12 in four patients each.

Interestingly, germline events were quite common. Six patients had small regions of germline LOH with much more extensive adjacent somatic LOH, two on chr 13, one on chr 17, two on chr X and one on chr 20; these were coded as germline in the analysis. In addition, of the 25 patients with cnLOH on chromosome 13, 18 of these were in the germline and 7 were somatic. The region(s) of cnLOH were typically adjacent to a 13q deletion, and often involved the entire chromosome arm. Somatic cnLOH at 13q was associated with intermediate sized deletions including the RB gene (p=0.002). Of the 18 patients with germline cnLOH at 13q, 7 of them had no 13q deletion, while 7 had monoallelic deletion and 4 biallelic deletion. Thus 7 patients (3%) had cnLOH events at 13q, in the absence of 13q deletion, again suggesting an alternative mechanism affecting this locus.

Germline cnLOH was associated with treatment prior to sampling (44% vs 17%, p<0.001), possibly due to its association with unmutated IGHV(58% vs 32%, p=0.008), and ZAP70 positivity (59% vs 36%, p=0.024). Somatic cnLOH was not associated with any patient characteristics. Neither somatic nor germline cnLOH was associated with >= 1 somatic CNA, but an association between both LOH types and >= 2 somatic CNAs was observed (p=0.053 germline and p=0.030 somatic). TTFT was reduced in patients with either germline cnLOH (61 mos vs 103, p=0.004) or somatic cnLOH (53 mos vs 107, p=0.008). Presence of two or more CNAs was also associated with short TTFT (48 mos vs 115, p<0.001).

In order to assess the impact of cnLOH and CNAs on outcome independent of prior therapy, we evaluated TTFT in the 181 chemotherapy naive patients. In this subgroup, germline cnLOH was not associated with short TTFT, while somatic cnLOH (80 mos vs 125, p=0.018) and two or more somatic CNAs (80 mos vs 125, p=0.009) were. In multivariable Cox modeling including germline cnLOH, IGHV, and del 11q or 17p by FISH, the only significant predictor of TTFT was unmutated IGHV (hazard ratio (HR) 4.48, p<0.001). In multivariable Cox modeling including somatic cnLOH and the variables above, the only significant predictor of TTFT was again unmutated IGHV (HR 4.41, p<0.001). When the presence of two or more somatic CNAs was added to these models, this variable was significant along with IGHV (HR 2.04, p=0.009 in germline model; HR 1.84, p=0.033 in somatic model). We conclude that both somatic and germline cnLOH are common in CLL, affecting one quarter of patients in this dataset, and frequently involve chromosomal regions known to be important in CLL. cnLOH is associated with increased somatic CNAs and unmutated IGHV, and therefore poor prognosis.