Quantitative Ultra-Deep Sequencing of Non-Coding Regions to Characterize Clonal Evolution in Follicular Lymphoma

Abstract

Abstract 146 Background:Cancers are characterized by genomic instability, and the resulting intra-clonal diversity is thought to underlie the evolution of more aggressive clinical course behavior and therapy resistance. Intra-clonal heterogeneity in follicular lymphoma (FL) has been documented qualitatively by analysis of mutations induced by Activation Induced Deaminase (AID) in IGH coding regions. AID also aberrantly targets non-coding regions of the genome causing numerous “passenger” mutations. To obtain a quantitative measure of genetic variation and to characterize population dynamics in FL, we performed ultra-deep “next generation” sequencing of selected non-coding regions known to be aberrantly targeted by AID in FL. Herein, we quantify and characterize AID-induced genetic variation in FL and relate the extent of variation to grade. Method:An ultra-deep sequencing approach, able to identify sequences present at 0.3% frequency, was developed to quantify sub-clonal populations within a tumor so that the extent of intraclonal heterogeneity can be compared between tumors. Genomic DNA was obtained from fresh/frozen specimens (10 FL grade 1/2, 2 FL grade 3, 3 hyperplastic lymph nodes and 1 epithelial cell line). H&E sections and flow cytometry data were examined to corroborate the diagnosis and fraction of tumor (>50% in all cases). The ten sequenced (ABI-SoLID) regions were PCR amplicons of: 5’UTRs of BCL2, BCL6, MYC, PAX5, PIM1, RHOH, CD83 and SYK; IGH (mu enhancer-J6; clone-specific Vh-J6). Average mapped coverage was > 20,000-fold. To detect rare mutations while suppressing read noise, data were mapped using an algorithm (BFAST) designed to detect rare Single Nucleotide Variants (SNV); a novel recursive mapping pipeline enhanced detection of SNVs in heavily mutated regions. A novel filtering algorithm handled data as 34 base-pair “words” with confirmation based on partial assembly using independently obtained sequences from adjacent regions. Results:656 mutations were detected in non-coding regions of the 12 FL specimens. FL specimens had more mutations in all 5’ UTRs combined (range 14–193, median 47, n=12) than control specimens (range 2–7, n=4). The mutations were associated with AID-motifs (p<0.05). 8 regions (excluding SYK and MYC) had significantly more SNVs in FLs compared to controls. BCL2 had the largest number of SNVs and the number of SNVs in BCL2 predicted the sum of SNVs in the 9 other regions (r2=0.85). The specimens could be placed into 4 strata based on the number of BCL2 SNVs (Poisson ladder, p=0.005). No correlation was seen between the number of SNVs and grade, percent nodularity, percent tumor, or other histologic features. Of the 354 SNVs in BCL2, many were “rare”: 78 were present in less than 1% of alleles and 41 in less than 0.5%. Strikingly, all low frequency CNVs were found in cases of Grade 1/2 FL without any found in either of the Grade 3 FL specimens despite the abundance of higher frequency CNVs in these cases. Representing the data as evolutionary trees suggested that the grades have differing population dynamics: Grade 1/2 showed dominant populations from which similarly proliferative subpopulations were continuously derived while Grade 3 showed a dominant population with remnant evolutionarily antecedent cells but little evidence of continuing development of intraclonal variation. Conclusions:The number of SNVs in AID-targeted non-coding regions varied widely allowing stratification of FL cases into 4 groups. The number of SNVs in BCL2 non-coding regions correlated with the number of SNVs in the other 8 gene regions combined, suggesting that analysis of the BCL2 non-coding region is sufficient to characterize the genome-wide effects of aberrantly targeted AID. Furthermore, low frequency mutations (present in less than 10% of sequences) were found only in grade 1/2 but not in grade 3. Representing the data as evolutionary trees indicated that population dynamics also varied greatly, with evidence that in low grade FL there is active ongoing generation of diversity while in grade 3 FL the population is dominated by the most evolutionarily derived clone with relatively less ongoing diversification. In contrast to prior methods, this quantitative approach to tumor evolution may allow correlation of intraclonal genetic diversity with natural history, response to immune-chemotherapy and risk of transformation to diffuse large cell lymphoma. Disclosures:No relevant conflicts of interest to declare.