Landscape of Somatic Mutations and Gene Expression Changes in Relapsed Infant MLL-Rearranged Acute Lymphoblastic Leukemia

Introduction

Relapse of infant ALL with MLL-rearrangement (MLL-r) is an unsolved clinical problem. Of the 80% of infants with ALL that harbor MLL-r, 60% relapse and die with current treatment. The typical pattern of failure is remission induction, followed by early relapse and very low second remission rate (<30%), suggesting rapid emergence of chemoresistant subclones. Previously, whole genome sequencing (WGS) of 22 MLL-r infant ALL cases showed only 1.3 non-silent mutations in the predominant clone at diagnosis and no significant increase of mutations at relapse in 2 diagnosis-relapse pairs (Andersson et al. (2015) Nat. Gen. 47(4):330-337).

Here, we examined the extent to which relapse in infant MLL-r ALL is driven by acquisition of new somatic mutations and/or gene expression changes. We assembled a large cohort of 14 paired diagnosis-relapse (DX-RL) infant MLL-r ALL cases treated on Children's Oncology Group (COG) trial AALL0631. MLL partners were 8 AF4, 5 ENL, and 1 unspecified. Paired remission bone marrow samples (RM) served as germline controls.

Methods

We performed WGS and RNA sequencing (RNA-seq) on 14 DX-RM-RL trios (N=42) using Illumina Hiseq 4000 and 2500. WGS samples were sequenced to a minimum of 90Gb and RNA-seq samples to an average of 8.8Gb. Alignment and variant calling were performed using BWA, GATK. Variants present in DX and/or RL and absent in paired RM samples were considered somatic. Filters included allelic depth threshold of 7 reads, minor allele frequency ≤0.1%, and ACMG category 1-3. RNA-seq gene expression was evaluated using GSNAP, Cufflinks, and Cuffdiff. Genes were ranked by average magnitude of fold change in expression between DX and RL for each case. We combined the gene ranks across cases to identify the 300 most highly activated and inactivated genes in RL, and evaluated gene set enrichment with MSigDB and Ingenuity software.

Results

In total, we found 148 non-synonymous somatic variants; 20 small (<40 bp) insertions/deletions (indels) and 128 single nucleotide variants (SNVs), of which 108 were missense, 10 nonsense, 8 splicing, and 2 mitochondrial somatic mutations. The mean variant allele frequency was 0.40 ± 0.14 (range 0.15-0.93). The mean number of non-synonymous mutations per case at DX was 1.71 ± 1.21 (range 0-4). Mutations in PIK3CD and NRAS were observed, each only in a single case. No gene was recurrently mutated across diagnostic samples. At RL, the number of mutations retained was 0-3 and the number lost was 0-4 per case. A large number of mutations (mean 8 ± 11.25, range 0-41) were acquired at RL. Notably, 5 of the 14 cases acquired ≥10 mutations at RL. The total number of variants in the cohort increased by 5-fold (24 to 124) from DX to RL. Acquired mutations at RL were observed in NRAS, KRAS, FLT3, BRAF, HIF1A, NOTCH3, and APC, among others. No RL samples shared a similar pattern of SNVs, suggesting the drivers of relapse and chemoresistance are diverse for this disease.

MSigDB gene set enrichment analysis showed that the 300 most highly inactivated genes at RL were enriched in gene sets that were targets of H3K27me3, EED, SUZ12, and Polycomb Repression Complex 2 (PRC2) (FDR q<9.30E-14), as well as in gene sets integral to the plasma membrane (q<8.20E-13). Additionally, these highly activated genes were enriched in sets encoding extracellular matrix-associated and other membrane-associated proteins (q<2.24E-14). Ingenuity pathway analysis revealed that the 300 most highly inactivated genes at RL were components of AMPK (p<0.04), BMP (p<0.04), NOTCH, WNT/β-catenin signaling, and B-cell development pathways, and the 300 most highly activated genes were components of IL17- (p<0.05), IL8-, and HIF1α- signaling, and granulocyte cell adhesion (p<0.01) pathways. This suggests that inactivating epigenetic events and dysregulation of cell membrane proteins are significant factors in the development of chemoresistance.

Conclusion

Relapsed infant MLL-r ALL is characterized by major gains in somatic variants and gene expression changes within cell signaling, adhesion, and B-cell development pathways. No single pathway or specific genomic variant was unifying across cases, but the appearance of mutations in many known cancer-associated genes and pathways suggests emergence of complex subclonal disease at relapse. Further analyses are ongoing to determine the contribution of germline mutations, structural variants, and DNA methylation changes in this cohort.