Whole-genome sequencing of chronic lymphocytic leukaemia reveals distinct differences in the mutational landscape between IgHVmut and IgHVunmut subgroups

Abstract

Chronic lymphocytic leukaemia (CLL) consists of two biologically and clinically distinct subtypes defined by the abundance of somatic hypermutation (SHM) affecting the Ig variable heavy-chain locus (IgHV). The molecular mechanisms underlying these subtypes are incompletely understood. Here, we present a comprehensive whole-genome sequencing analysis of somatically acquired genetic events from 46 CLL patients, including a systematic comparison of coding and non-coding single-nucleotide variants, copy number variants and structural variants, regions of kataegis and mutation signatures between IgHV^mut and IgHV^unmut subtypes. We demonstrate that one-quarter of non-coding mutations in regions of kataegis outside the Ig loci are located in genes relevant to CLL. We show that non-coding mutations in ATM may negatively impact on ATM expression and find non-coding and regulatory region mutations in TCL1A, and in IgHV^unmut CLL in IKZF3, SAMHD1,PAX5 and BIRC3. Finally, we show that IgHV^unmut CLL is dominated by coding mutations in driver genes and an aging signature, whereas IgHV^mut CLL has a high incidence of promoter and enhancer mutations caused by aberrant activation-induced cytidine deaminase activity. Taken together, our data support the hypothesis that differences in clinical outcome and biological characteristics between the two subgroups might reflect differences in mutation distribution, incidence and distinct underlying mutagenic mechanisms.

Figure 1

Overview of the mutational landscape in IgHV^mut and IgHV^unmut CLL. (a) Total number of SNVs and indels per sample and by IgHV mutation status. Asterisks denote cases with IgHV 3-21 rearrangement. (b) Dot plot of total somatic mutation count between IgHV subgroups. Patients with IgHV^mut CLL harbour higher total mutational load, but reduced driver coding (c) and higher promoter variant counts (d and e) than IgHV^unmut CLL. Error bars show±s.e.m.

Figure 2

Structural rearrangements in IgHV^unmut and IgHV^mut CLL. Circos plot depicting the frequency and locations of CNAs and translocations detected in 46 CLL genomes. The two outermost tracks represent CNAs: CN loss (red) and CN gain (green). Copy number data is displayed as the percentage of each IgHV subgroup affected. The centre plot shows translocations identified in the cohort. Red links indicate tier 1 events (gene:gene), green links indicate tier 2 events (gene:intergenic) and grey indicate tier 3 (intergenic:intergenic). The widths of the bands are indicative of the number of patients affected by the translocation.

Figure 3

Kataegic ATM mutations in CLL144. (a) Graphical representation of ATM showing the distribution of kataegic mutations in CLL144. Blue boxes represent exonic regions of the ATM transcript. (b) Dot plot comparing the ATM transcript expression levels of CLL cases with no 11q disruption (wild type), those with del(11q), mutations in ATM (ATM^mut) and CLL144. Expression levels are shown as reads per million aligned reads. Error bars show±s.e.m.

Figure 4

Non-coding mutations confirmed experimentally with ChIP-seq and ATAC-seq. Diagrams of non-coding mutations in three genes and the corresponding annotation data. All lower panels; correlation of mutation loci with H3K27Ac and ATAC-seq signal of CLL110 and ChromHMM annotation (all three layers obtained from experiments with primary CLL cells) and layered H3K27Ac data from the ENCODE database (GM12878). (a) IKZF3 locus and surrounding region on chromosome 17, with the position of all mutations detected in our cohort, (b) SAMHD1 locus and surrounding region on chromosome 20, with the position of all mutations detected in our cohort and (c) BIRC3 locus and surrounding region on chromosome 11, with the position of all mutations detected in our cohort.

Figure 5

Overview of the mutational landscape in IgHV^mut and IgHV^unmut CLL. Top panel—Type and number of IgHV subclones present in each patient. Where several clones are present, IgHV status is determined by the identity of the dominant clone. Samples where only Sanger sequencing data were available are labelled as ‘not known’. Middle panels—Presence of copy number aberrations within the cohort. Lower panel—Incidence of both coding and non-coding mutations within the cohort. Colour spectrum (light to dark blue or light to dark pink) corresponds with mutational or CNA load per patient, respectively. *Genes containing kataegic mutations. ^‡Genes with multiple mutations in regulatory regions and are involved in important B-cell pathways. ^§Genes with smaller mutational hotspots. Panel left: Number of mutations per gene across the entire cohort.

Figure 6

Mutation signature analysis. (a) Three mutation signatures identified across 46 patients using non-negative matrix factorisation (NMF). (b) Correlations between signatures identified in 46 patients and the corresponding age, or proportion of mutated canonical AID or APOBEC sites per subtracted somatic sample. Blue lines=regression lines, and P-values from the glm. (c) Proportions of mutation signatures from Alexandrov (2013),29,43 where Sig.1B corresponds to Alexandrov Signature 1B, and so on, and signatures found across 46 patients.