Mitochondrial DNA heteroplasmy and haplogroup variation in pediatric Acute Myeloid Leukemia Free

Introduction: While somatic mutations in pediatric AML risk stratification have been extensively characterized, the role of heteroplasmic missense single nucleotide variants in mitochondrial DNA (mtDNA) is undefined. These variants have been correlated with poorer cancer survival rates and mtDNA haplogroups have been shown to modify the risk of various diseases. In this context, we investigated the role of mtDNA heteroplasmic variants and haplogroup identity in pediatric AML survival outcomes.

Methods: The study cohort included 1097 eligible patients enrolled in the Children's Oncology Group (COG) clinical trial AAML1031 who had germline whole genome sequencing (WGS) data available. WGS and short read RNA-Seq data were generated through the COG TARGET AML initiative. The aligned CRAM files were harmonized by the Kids First Data Resource Center (DRC) and deposited on the CAVATICA platform. We used SAMtools view to extract mtDNA-mapped reads from WGS data. BCFtools was then used to generate mtDNA consensus FASTA files, which were used for haplogroup assignments with Haplogrep3. Patients were grouped into macrohaplogroups based on the mtDNA phylogenetic tree. For mtDNA variant identification, we used the GATK Best Practices workflow designed for MT short variant discovery. Mutations with variant allele frequency (VAF) between 0.03 and 0.95 were considered heteroplasmic, and those with VAF > 0.95 were deemed homoplasmic. We restricted our analyses to missense single nucleotide variants (SNVs) in coding regions and SNVs in tRNA and rRNA genes, excluding the control region. Gene expression was quantified by the Kids First DRC using RNA-Seq by Expectation-Maximization (RSEM). We extracted 37 MT genes' TPM values from the RSEM results and performed differential gene expression (DEG) analysis using Limma.

Results: For the 814 pediatric AML patients with available data on mtDNA variants, a total of 920 unique SNVs were identified. Of those, 257 (28%) were found at heteroplasmic levels. 171 patients (21%) were found to have at least one heteroplasmic SNV. Overall survival (OS) and event-free survival (EFS) probability were comparable in patients with heteroplasmy compared to those without. Moreover, when dividing patients by mtDNA haplogroup (HV, R, UK, N, L, and M), we found no significant differences in OS or EFS. We then examined patients' demographic and disease characteristics by heteroplasmy status. Notably, patients without a CBF-AML classification were significantly enriched in the group with heteroplasmy (p = 0.043), while patients with positive minimal residual disease (MRD) at end-of-induction I (p = 0.002) or belonging to a high-risk AML group (p < 0.001) were significantly over-represented among patients with heteroplasmy. When patient characteristics were analyzed across haplogroups, we found that the distribution of haplogroups varied by the presence of non-CNS extramedullary disease (EMD) (p = 0.014); patients with EMD were under-represented in haplogroup M and over-enriched in haplogroup N.

Next, to characterize the potential functional consequence of heteroplasmic missense mutations on respiratory chain proteins, we evaluated the predicted deleteriousness of such mutations using PolyPhen-2 (PMID: 23315928). Of the 158 heteroplasmic missense SNVs, 47% were predicted to be damaging, in contrast to 31% for the 450 homoplasmic missense SNVs (p < 0.001). OS and EFS were comparable in patients with at least one predicted damaging homo- or heteroplasmic SNV compared to those without.

Lastly, we examined the association of RNA transcript abundance of MT genes with heteroplasmy and haplogroup. Genes were considered significant if they had an adjusted p-value < 0.05, regardless of fold-change magnitude. Comparing heteroplasmy versus no heteroplasmy revealed no DEGs. In the haplogroup analysis (each haplogroup versus HV), no significant differences were observed for R vs HV or UK vs HV. Comparisons of L vs HV and N vs HV identified 12 genes with significantly decreased expression, with percent decreases ranging from 12% to 38%. The comparison of M vs HV identified 3 genes with significant decreased expression, with percent decreases up to 38%.Conclusions: Our results suggest an association between mtDNA heteroplasmy and high-risk pediatric AML phenotypes, but not differential OS or EFS. Future work on mitochondrial dysfunction in different AML disease statuses may clarify the role of mtDNA variants in risk stratification.