Array-based karyotyping for prognostic assessment in chronic lymphocytic leukemia: performance comparison of Affymetrix 10K2.0, 250K Nsp, and SNP6.0 arrays

Abstract

Specific chromosomal alterations are recognized as important prognostic factors in chronic lymphocytic leukemia (CLL). Array-based karyotyping is gaining acceptance as an alternative to the standard fluorescence in situ hybridization (FISH) panel for detecting these aberrations. This study explores the optimum single nucleotide polymorphism (SNP) array probe density for routine clinical use, presents clinical validation results for the 250K Nsp Affymetrix SNP array, and highlights clinically actionable genetic lesions missed by FISH and conventional cytogenetics. CLL samples were processed on low (10K2.0), medium (250K Nsp), and high (SNP6.0) probe density Affymetrix SNP arrays. Break point definition and detection rates for clinically relevant genetic lesions were compared. The 250K Nsp array was subsequently validated for routine clinical use and demonstrated 98.5% concordance with the standard CLL FISH panel. SNP array karyotyping detected genomic complexity and/or acquired uniparental disomy not detected by the FISH panel. In particular, a region of acquired uniparental disomy on 17p was shown to harbor two mutated copies of TP53 that would have gone undetected by FISH, conventional cytogenetics, or array comparative genomic hybridization. SNP array karyotyping allows genome-wide, high resolution detection of copy number and uniparental disomy at genomic regions with established prognostic significance in CLL, detects lesions missed by FISH, and provides insight into gene dosage at these loci.

Figure 1

Acquired uniparental disomy. Acquisition of a second hit via deletion (top) versus acquired UPD (bottom). Blue cells represent the FISH results for each scenario. Both arrayCGH and SNP array karyotyping would generate the same blue copy number plot. Loss of heterozygous calls (green bars) and allele-specific analysis (red/green plots) can only be obtained using SNP-based array karyotyping.

Figure 2

Comparison of array density and software performance at a low percentage of CLL cells. CLL32 has a cytogenetically-confirmed trisomy 12 as the only genetic lesion (26% CLL cells by flow cytometry). Four karyotypes from different array densities and softwares are shown. GTC software cannot provide an equivalent whole genome image, so only chromosome 12 is shown (bottom plot). Graph Key: A: Linear log2ratio (copy number) smoothed over 10 SNPs is plotted in chromosomal order along the x axis (chromosome 1 on the left, chromosome 22 on the right). B: Heterozygous call bars indicating an AB genotyping call made in the tumor sample at that SNP. C: Hidden Markov Model (HMM) for copy number with dark blue = 0, aqua = 1, yellow = 2, pink = 3, pink-red = 4, red-pink = 5, and bright red >5. D: HMM for LOH likelihood with yellow = low LOH likelihood, blue = high LOH likelihood. E: Copy number HMM generated by GTC3.0.1. F: Segment report results obtained with the GTC3.0.1 segment reporting tool.

Figure 3

Whole genome view of CLL09 using 250K Nsp SNP array. This CLL sample did not show any copy number abnormalities by the standard CLL FISH panel. Graph key: Plots (A) and (B) are as described in Figure 2, with a log2ratio of zero = copy number of 2. C: Allele-specific copy number analysis in which each allele is plotted separately. In a normal/diploid region, both the red and the green plots will run along the zero line, indicating that there is one copy of each for an overall copy number of two. In regions of UPD, there are two copies of the red allele and zero copies of the green allele for an overall copy number of two. This is supported by the HMM plots and the lack of heterozygous call bars in the region (black arrowheads); D = copy number HMM; and E = LOH likelihood HMM.

Figure 4

Copy number and length heterogeneity at the 13q14 locus in CLL. Genomic position (A); ideogram (B); raw log2ratios for each probe (C); smoothing average of log2ratio over 10 probes (D); heterozygous call bars (E); allele-specific analysis (F) (AsCNAR algorithm), as in Figure 3; hidden Markov model for copy number (G), as in Figure 2; hidden Markov Model for LOH likelihood (H) as in Figure 2; and FISH results (I) for 13q14 locus. The percentage of CLL cells present in the sample as determined by flow cytometry is provided below the ideogram. CLL22: heterozygous deletion; CLL27: homozygous deletion with aUPD of 13q; CLL01, CLL03, CLL07: Examples of partially heterozygous and partially homozygous deletions.