Cancer drivers and clonal dynamics in acute lymphoblastic leukaemia subtypes

Abstract

To obtain a comprehensive picture of composite genetic driver events and clonal dynamics in subtypes of paediatric acute lymphoblastic leukaemia (ALL) we analysed tumour-normal whole genome sequencing and expression data from 361 newly diagnosed patients. We report the identification of both structural drivers, as well as recurrent non-coding variation in promoters. Additionally we found the transcriptional profile of histone gene cluster 1 and CTCF altered tumours shared hallmarks of hyperdiploid ALL suggesting a 'hyperdiploid like' subtype. ALL subtypes are driven by distinct mutational processes with AID mutagenesis being confined to ETV6-RUNX1 tumours. Subclonality is a ubiquitous feature of ALL, consistent with Darwinian evolution driving selection and expansion of tumours. Driver mutations in B-cell developmental genes (IKZF1, PAX5, ZEB2) tend to be clonal and RAS/RTK mutations subclonal. In addition to identifying new avenues for therapeutic exploitation, this analysis highlights that targeted therapies should take into account composite mutational profile and clonality.

Fig. 1. Mutation burden by subtype.

Short somatic variants (SNVs and indels) were called in 361 matched normal/tumour whole genome sequencing samples. a Burden of SNVs and indels. Box and whiskers plot of mutation count per tumour. b Burden of structural variants (SVs). Box and whiskers plot of SV count, dots represent individual tumours. c Plot of the SV rate per chromosome retaining only intrachromsomal variants outside immunoglobulin loci. Y-axis; mean SVs rate per Mb. X- axis; chromosome.

Fig. 2. Driver gene analysis.

Oncoplot of somatic alterations for selected ALL driver genes (Supplementary Table 10), genes altered in > 20 tumours, compiled from SNV, indel, CNV (only focal events), SV, RNAseq, and LOH. Vertical lines represent one tumour. Coloured sections in grey grid denote alteration type, described in key - “Alteration type”. Short nucleotide variants span the entire row, other alterations span half the row width. The right bar plot shows the frequency of driver gene alteration, colour denotes alteration type, as prior. Deletions and amplifications derived from CNVs and SVs; disruption from RNAseq and SVs. Alterations are shown non-redundantly; tumours with multiple alterations in the same gene only counted once. Plot generated using Maftools [6] after the exclusion of genes in the psuedoautosomal regions.

Fig. 3. Recurrent copy number and structural variants.

Significantly amplified or deleted regions in CNV data, were filtered retaining only those with an enrichment of structural variants, based on a permutation test. Regional genetic plots showing recurrent deletions mapping to (a) HLA-DRB5 and (b) histone gene cluster 1. Upper panes shows gene position. Line plots show number of tumours with an overlapping variant; blue—CNVs, red—SVs, black—total count (tumours with both SVs and CNVs counted once). Central pane shows the individual variants. For convenience only variants starting or ending in the field of view are plotted. Vertical black lines denote region with the highest deletion frequency. Lower left pane, box plots of gene expression split by mutational status. Lower right pane, density plots of structural variant clonality; blue circles individual SVs. Genomic coordinates from GRCh38.

Fig. 4. Non-coding driver mutations.

Mutation burden within promoters and their transcriptional impact. Promoter mutations of (a) BTLA and (b) CHID1. Regional plot of mutations (coloured circles) relative to coding sequence (dark blue boxes) and promoter (yellow horizontal bar). Transcription factor binding sites (light blue horizontal line) overlapping mutations were extracted from Encode and ChIP atlas. Grey boxes correspond to transcriptional impact on respective gene. Box and whiskers plot, tumours are split by mutational status, dots represent individual tumours.

Fig. 5. Pathway analysis and signature analysis.

Radar plots showing the most frequently altered pathways for each subtype. Driver genes grouped according to the biological pathway. Somatic alterations for a selected ALL driver genes was compiled from SNV, indel, CNV, SV, RNAseq, and LOH data (CNVs include only focal events). Subtype defining events are excluded (e.g, disruption of ETV6 or RUNX1 in ETV6-RUNX1 positive tumours). The proportion of tumours with an alteration in any gene assigned to that pathway is plotted on the radial axis. Each axis is scaled separately. Gene—pathway assignments: RAS/RTK; NRAS, KRAS, PTPN11, FLT3, NF1, ABL1. B-cell development; PAX5, IKZF1, ETV6, ZEB2, RUNX1, TCF3, RAG1, RAG2, EBF1. Chromatin regulation; SETD2, HDAC7, NSD2, CTCF, KMT2A, STAG2, histone gene cluster 1. Cytokine signalling; JAK2, IL7R, CRLF2. Gene regulation; CREBBP, MLLT1, MLLT3, AFF1, BTG1, ERG, TCF4, NCOA6. Signal transduction; TBL1XR1, TBL1X, PBX1, PAG1. Cell cycle regulation; CDKN2A, CDKN2B, RB1. Immune regulation; BTLA, HLA-DRB5.

Fig. 6. Driver clonality and SV breakpoint enrichment.

a Driver gene mutation (SNV/indels) clonality. Box and whiskers plot showing the proportion clonal mutations for ALL driver genes. Each circle represents a mutation, coloured according to disease subtype, for tumours with multiple mutations in the same gene the variant with the highest clonal cell fraction is retained. b Structural variant motif enrichment. Bar chart showing motif enrichment at SV breakpoints. Two 100 bp of sequences flaking each breakpoint of an SV were extracted and analysed using HOMER. Y-axis; percent of extracted sequences containing motifs.

Fig. 7. Clonal architecture and evolution.

Variant cancer cell fractions (CCF) were calculated and variants clustered into clonal and subclonal populations. a Distribution of clones. Horizontal lines represent single tumours, circles represent clones; the size and colour of circles corresponding the proportion and number of variants assigned to each clone. Y-axis; clonal frequency (proportion of cell cells with a variant(s)). b Heterogeneity between subtypes. Box and whiskers plot of Simpson index (higher values indicative of increased heterogeneity). Each dot corresponds to a tumour.