Genomic Analysis and Pathway Characterization of Genes with Somatic Variants in Infant Acute Lymphoblastic Leukemia

Introduction

Infant acute lymphoblastic leukemia (ALL) is an aggressive cancer with poor relapse-free and overall survival. Other than MLL ( KMT2A ) rearrangement ( MLL -r), infant ALL contains remarkably few genomic lesions. We previously reported results of whole genome sequencing (WGS) on 14 matched diagnosis-remission-relapse (DX-RM-RL) MLL -r infant ALL trios (cohort A) treated on Children's Oncology Group (COG) trial AALL0631 (Kostadinov, et al. Blood 2016, 128(22):1735).

Here, we examine the presence of somatic variants in both MLL -r (n=12) and non- MLL -r (n=15) infant ALL cases (COG AALL0631), which, to date, have not relapsed (cohorts B and C, respectively). For MLL -r cases, MLL partners were 6 AF4 ( AFF1 ), 5 ENL ( MLLT1 ), and 1 unspecified. For all cases, paired RM samples served as germline controls. In addition, we manually curated genes with somatic variants into functional pathway categories to determine whether any pathways were recurrently altered amongst all 3 cohorts.

Methods

We performed WGS on 27 paired DX-RM blood or bone marrow samples using an Illumina Hiseq 4000 or 2500 to a minimum depth of 90Gb. Alignment and variant calling were performed using BWA and GATK. Variants present in the DX sample and absent in the paired RM sample were considered somatic. Filters included an allelic depth threshold of 4 reads, minor allele frequency ≤0.1%, and a variant annotation of known pathogenic, likely pathogenic, or uncertain significance per RUNES (Saunders, et al. Sci Transl Med 2012, 4(154):154). Next, somatic variant genes (SVG) in these 2 cohorts, as well as cohort A (n=25 at DX and 156 at RL), were categorized into 1 or more of 41 functional pathways using publically available protein databases (UniProt, GenAtlas, BioGPS) and published literature. We then compared the prevalence of pathways affected across all 3 cohorts, specifically examining relapsed versus non-relapsed cases, MLL -r versus non- MLL -r, DX versus RL, and MLL - AF4 versus MLL - ENL , to identify any differences.

Results

We found 23 non-synonymous somatic variants (NSSV) across 22 genes in cohort B; 4 were small (<40 bp) insertions/deletions (indels) and 19, single nucleotide variants (SNVs), of which 17 were missense, 1 nonsense, and 1 splicing. The mean variant allele frequency (VAF) was 0.42 ± 0.12 (range (r) 0.14-0.58). The mean number of NSSV per case at DX was 1.92 ± 1.73 (r 0-6). Mutations in KRAS (n=2), PIK3R1 (n=1), and NRAS (n=1) were observed. In cohort C, 29 NSSV across 28 genes were identified; 3 were small indels and 26, SNVs: 23 missense, 1 nonsense, and 2 splicing. The mean VAF was 0.37 ± 0.13 (r 0.13-0.67). The mean number of NSSV per case at DX was 2.0 ± 1.73 (r 0-6). Mutations in KRAS (n=2) and JAK2 (n=1) were observed. Using the same analysis filters for cohort A, we found that the mean number of NSSV per case was 2.1 at DX, and 11.5 at RL. The most frequently mutated genes included NRAS and FLT3 (n=3 cases each) at DX and KRAS , NRAS , and TP53 ,at RL (n=2 cases each).

Pathway classification revealed that 2 categories-signal transduction and transcriptional regulation-contained the greatest number of SVG (n≥8) in each cohort at DX. Other categories with a high number of SVG (n≥4) across cohorts at DX included: cell cycle, metabolism, and developmental regulation. Within the signal transduction pathway group, the RAS/MAPK/ERK and PI3K/AKT/mTOR signaling pathways contained the greatest number of SVG; interestingly, these were more likely to be found in MLL-AF4 cases (n=9 and 10, respectively, of 14 total) than MLL - ENL (n=0 and 3, respectively, of 10 total). SVG involved in NF-kappa-B signaling were only found in MLL -r cases (n=2 of 26) versus non- MLL -r (n=0 of 15). SVG involved in WNT signaling and ion transport were more common at DX in cases that eventually relapsed (n=3 and 4, respectively, of 14 total) than those that did not (n=0 and 1, of 12 total). Of all the SVG gained at relapse (n=136), most fell into the signal transduction and transcriptional regulation categories.

Conclusions

Overall, the mean number of NSSV per case at DX for each cohort was low. Functional pathways containing the highest numbers of SVG were also largely similar across cohorts, though we have highlighted some potentially important differences in subsets of cases. Further correlation with the type of variant found, as well as the expression of SVG across infant ALL cases, is ongoing.