Leukemia Evolution and Resistance to Graft-Versus-Leukemia Responses Revealed By Mitochondrial DNA Mutations

Graft-versus-leukemia (GvL) responses enable successful allogeneic hematopoietic cell transplantation (HCT), a cornerstone treatment for high-risk blood cancers such as acute myeloid leukemia (AML) or advanced chronic lymphocytic leukemia (CLL). Yet, monitoring of GvL effects remains challenging, including early detection of emerging immune escape or tracking of clonal evolution during phases of incomplete leukemia control. We hypothesized that mitochondrial DNA (mtDNA) mutation analysis would be an effective tool for sensitive monitoring of clonal evolution and measurable residual disease (MRD) that could yield new insights into the dynamics and cellular phenotypes of GvL response or resistance. To establish mtDNA mutations as natural genetic barcodes capable of resolving leukemic evolution, we analyzed 126 serially collected bulk RNA and whole-exome sequencing (WES) datasets of CD19+ sorted CLL cells from 34 individuals (median interval first-to-last sample 2,184 days, range 280-4,402). We discerned high correlation between changes in heteroplasmy of mtDNA mutations (RNA-seq) and cancer cell fraction (CCF, calculated from WES data) (r=0.84; p<0.001). In particular, heteroplasmy and CCF changes were smaller in genetically stable (n=8) compared to evolving (n=18) CLL (coefficient of variation 24 vs 60 [heteroplasmy] and 31 vs 57 [CCF]; p<0.001). Following HCT (n=8), shorter vs longer days-to-relapse (median 310 vs 1,082) was associated with lesser vs more changes in heteroplasmy (p=0.034), suggesting the ability to measure cumulative GvL effects with mtDNA mutations. In 3 cases, we obtained single cell (sc) DNA-seq profiles (Tapestri) from circulating CLL before and after HCT and directly detected co-evolution of somatic nuclear with mtDNA mutations. Thus, mtDNA mutations track with leukemia clones and their dynamics are highly consistent with changes in somatic nuclear DNA mutations. To assess mtDNA mutations as sensitive markers of MRD, we determined their median limit-of-detection in bulk RNA-seq data as 0.3% (0.008-73%). This high sensitivity enabled tracking of resistant leukemia selected by GvL, such as in one CLL case, where a mtDNA mutation (6426G>A) expanded from 0.2% (pre-HCT) to 54% heteroplasmy (post-HCT) over a period of 6 years. By scDNA-seq, we validated that 6426G>A marked a CLL subclone represented by 3 of 2,416 cells (0.12%) pre-HCT that expanded in the post-HCT relapse. Donor/recipient deconvolution using mtDNA mutations is another approach for sensitively measuring post-HCT chimerism and MRD. We confirmed that mtDNA mutations could track rare cells at frequencies of 0.1-0.3% by single cell ATAC sequencing (scATAC-seq) in in-silico and in-vitro simulations of post-HCT MRD. In 10 primary human bone marrow samples, we defined residual leukemia cell phenotypes with scATAC-seq and capture of surface markers (ASAP-seq) at time of incipient AML relapse post-HCT. We also tracked residual donor hematopoiesis in 14 bone marrow (BM) samples of 4 frank post-HCT AML relapse cases with ASAP-seq, detectable in 0.1%-0.6% of cells and mostly of erythroid lineage. Together, mtDNA-based MRD tracking is highly sensitive, and when coupled with multimodal single cell sequencing, can define the phenotypes of residual leukemia. Finally, we charted clonal evolution in 21 BM samples of 5 AML cases with ASAP-seq, either at recurrence post-HCT (n=3) or during ipilimumab-based relapse treatment (n=5; ETCTN/CTEP 10026, NCT02890329). By mtDNA analysis, we defined a median of 7 AML subclones per case (range 5-12). In contrast to 2 instances with little evidence of clonal evolution after matched unrelated HCT, in 1 case, haploidentical transplantation led to pronounced subclonal shifts, revealing differential intensity of GvL. Throughout decitabine/ipilimumab treatment in post-HCT AML relapse, we found clonal and phenotypic evolution even in cases that stabilized only transiently, consistent with activated GvL countered by immune escape. In 3 cases, we validated the clonal shifts with scDNA-seq and directly demonstrated co-segregation and co-evolution of mitochondrial with somatic nuclear DNA mutations. In sum, mtDNA mutations resolve post-HCT immune pressures through dissection of clonal evolution. Together, mtDNA mutations enable the sensitive monitoring of the phenotypes and selection of residual leukemia cells after HCT and reveal differential activity of GvL.