Integrative Single-Cell Tracking of Clonal Evolution, Leukemia Phenotypes and Donor Engraftment Following Allogeneic Stem Cell Transplant Using Mitochondrial DNA Mutations

Mitochondrial DNA (mtDNA) mutations can distinguish cells from unrelated individuals and their spectrum per patient sample can evolve following therapeutic bottlenecks. As such, these natural barcodes can potentially enable integrated single cell tracking of chimeric cellular populations and clonal evolution following unrelated allogeneic hematopoietic stem cell transplantation (HSCT). However, the feasibility of mtDNA-based donor-recipient deconvolution or identification of leukemia-specific phenotypes and the extent of co-evolution with somatic mutations remains unknown. If confirmed, post-transplant monitoring using mtDNA mutations could enable detection of early post-HSCT relapse and provide qualitative information on leukemia phenotypes, clonal evolution, or residual donor engraftment via a combined single cell assay.

To investigate the extent of co-evolution of somatic nuclear mutations with mtDNA mutations, we generated 33,796 single cell DNA sequencing profiles (Tapestri) from 3 sample pairs of circulating chronic lymphocytic leukemia (CLL) collected before and after HSCT. The samples were characterized using a primer panel (n=67 nuclear and 21 mtDNA) targeting single nucleotide variants previously identified by whole-exome sequencing. We detected 8 mtDNA mutations that defined CLL subclones together with a median of 14 (8-19) somatic mutations from these cases. Clonal shifts defined by somatic mutations discernable pre- or post-HSCT tracked with analogous changes in mtDNA mutations and included clonal replacement, selection of a resistant subclone or skewing of CLL subclones. We thus provide clear evidence of co-segregation between somatic nuclear and mtDNA mutations.

To assess the sensitivity of mtDNA-based donor-recipient deconvolution, we performed an in-silico titration and spiked 1-1,000 single cell mtDNA profiles (mtscATAC-seq) from one CLL case (‘CLL4’) into 7,579 profiles from another (‘CLL5’). Overall, we recovered 98.3-100% of CLL4, identifying up to 1 in 7,579 cells (0.13%). Only 123 cells (1%) failed deconvolution, due to insufficient mtDNA coverage. In an in-vitro titration of expanded T cells (donor1) with peripheral blood mononuclear cells (donor2) at ratios of 1:3, 1:30 and 1:300, we recovered 1,449 of 5,345 (27%), 150 of 4,262 (3.5%) and 14 of 4,381 (0.3%) T cells of donor1. In sum, with current sequencing throughput, mtDNA-based donor-recipient deconvolution was assessed to have a sensitivity of ~0.1%.

To determine if mtDNA-based deconvolution can track early AML relapse and identify leukemia phenotypes, we analyzed peripheral blood and bone marrow of 4 patients with disease recurrence 87-185 days post-HSCT and subsequent reinstatement of remission 105-184 days after tapering of immunosuppression (IST). We obtained 19,143 ATAC with select antigen profiling by sequencing (ASAP-seq) profiles at incipient relapse (pre-IST) and following response to IST (post-IST). Donor-recipient deconvolution detected recipient-derived cells in progenitor, erythroid and monocytic populations pre-IST but no recipient-derived cells post-IST, consistent with elimination of AML. Overall, chromatin accessibility of donor- and recipient-derived cells was very similar, reflecting that AML recapitulates hematopoietic differentiation states. Nevertheless, we observed differences in accessibility of genes such as IL1B, CD36, chemokine or homeobox protein genes in recipient- compared to donor-derived cells. Together, mtDNA-based donor-recipient deconvolution reflected clinical responses and identified leukemia-specific chromatin accessibility profiles.

Finally, we tracked clonal evolution and donor-derived hematopoiesis in two non-responder cases of overt AML relapse following 3-10 cycles of decitabine and ipilimumab (NCT02890329). We defined 12 and 7 mtDNA-based AML subclones and observed contraction and expansion of subclones during treatment. Donor-derived non-T cells were exceedingly rare (12 of 10,708 [0.1%] and 34 of 5,863 [0.6%]), mainly of erythroid origin (12 and 23 cells) and disappeared during treatment.

Thus, mtDNA mutations can monitor donor- and recipient-derived hematopoiesis and resolve short-term changes in AML subpopulations during therapy. As sequencing throughput improves, we envision single cell-based post-transplant monitoring as a powerful approach for guiding clinical decision making.