AFFILIATIONS
1 Hans Messner Allogeneic Transplant Program, Princess Margaret Cancer Centre, University Health Network, Toronto, Ontario, Canada
2 Hans Messner Allogeneic Transplant Program, Princess Margaret Cancer Centre, University Health Network, Toronto, Ontario, Canada
3 Hans Messner Allogeneic Transplant Program, Princess Margaret Cancer Centre, University Health Network, Toronto, Ontario, Canada
4 Hans Messner Allogeneic Transplant Program, Princess Margaret Cancer Centre, University Health Network, Toronto, Ontario, Canada
5 Faculty of Medicine, University of Toronto, Toronto, Canada, Hans Messner Allogeneic Transplant Program, Princess Margaret Cancer Centre, University Health Network, Toronto, Ontario, Canada
6 Faculty of Medicine, University of Toronto, Toronto, Canada, Hans Messner Allogeneic Transplant Program, Princess Margaret Cancer Centre, University Health Network, Toronto, Ontario, Canada
7 Hans Messner Allogeneic Transplant Program, Princess Margaret Cancer Centre, University Health Network, Toronto, Ontario, Canada, Leukemia Program, Princess Margaret Cancer Centre, University Health Network, Toronto, Canada, Research Program, Princess Margaret Cancer Centre, University Health Network, Toronto, Canada
8 Tumor Immunotherapy Program, Princess Margaret Cancer Centre, Toronto, Ontario, Canada
9 Department of Molecular Genetics, University of Toronto, Toronto, Ontario, Canada, Research Program, Princess Margaret Cancer Centre, University Health Network, Toronto, Canada
10 Department of Molecular Genetics, University of Toronto, Toronto, Ontario, Canada, Research Program, Princess Margaret Cancer Centre, University Health Network, Toronto, Canada
11 Leukemia Tissue Bank, Princess Margaret Cancer Centre, University Health Network, Toronto, Ontario, Canada
12 Institute for Medical Science, University of Toronto, Toronto, Ontario, Canada, Leukemia Program, Princess Margaret Cancer Centre, University Health Network, Toronto, Canada, Leukemia Tissue Bank, Princess Margaret Cancer Centre, University Health Network, Toronto, Ontario, Canada, Research Program, Princess Margaret Cancer Centre, University Health Network, Toronto, Canada
13 Faculty of Medicine, University of Toronto, Toronto, Canada, Hans Messner Allogeneic Transplant Program, Princess Margaret Cancer Centre, University Health Network, Toronto, Ontario, Canada, Leukemia Tissue Bank, Princess Margaret Cancer Centre, University Health Network, Toronto, Ontario, Canada, Research Program, Princess Margaret Cancer Centre, University Health Network, Toronto, Canada
ABSTRACT
Background
Relapse is the leading cause of death after allogeneic transplantation (HCT) for acute myeloid leukemia (AML). While prior work mainly focused on loss of leukemic antigenicity in post-HCT relapse (PTR), we profiled paired leukemic blast and immune cell compartments at single-cell resolution to investigate whether Graft-vs-Leukemia failure arises from donor CD8 T-cell exhaustion acquired at PTR, or from blast-intrinsic shift to primitive, inhibitory-ligand-armed state at PTR.
Patients & Methods
We profiled 59 marrow and blood mononuclear samples from adult patients (diagnosis [Dx] n=16; post-HCT relapse [PTR] n=31; long-term remission [LTR, ≥2 years after HCT] n=12) by single-cell RNA-sequencing (scRNAseq; Parse Biosciences Evercode WT): 44,997 QC-passed cells (27,227 leukemic, 11,791 T cells). Leukemic comparisons were made on tissue-matched marrow (Dx vs PTR); LTR (~99% non-leukemic) served as a leukemia-free reference. Cells were annotated by consensus of BoneMarrowMap (Zeng 2025), CellTypist and scANVI, blasts placed on the leukemic stem cell hierarchy (Zeng 2022). To compare it to PTR, we adopted two post-chemotherapy relapse cohorts: Naldini et al. (2023; GSE185993; 6 samples, 5,977 blasts) and Ennis et al. (2023; GSE227903; 20 samples, 3,913 CD8 cells).
Results
At PTR, leukemic blasts shifted to a primitive, differentiation-arrested state: the LSPC-Primed fraction rose from 19% to 54%, and the primitive fraction from 0.22 to 0.55 (patient-clustered generalized estimating equation (GEE), OR 4.34 [2.0-9.6], p=0.0003), with higher stemness and a primitive-root pseudotime skew. Pathway enrichment (GSEA against MSigDB Hallmark gene sets), on the leukemic blasts, showed coordinated interferon-response program, with allograft-rejection and inflammatory programs and suppressed cell cycle (primitive blasts, NES: IFN-gamma-response +2.4, allograft rejection NES +2.5, E2F -2.8; FDR<0.01). These blasts up-regulated inhibitory checkpoint ligands, most sharply CD200 (delta +0.31; CD200+ cells GEE OR 11.0 [2.6-46]), with NECTIN2/PVR and LGALS9. This shift in ligand expression tracked stemness: CD200 correlated with stemness (rho 0.37) and fell with differentiation, so the primitive blast was also the most checkpoint-ligand-armed leukemic cells.
CD8 T-cell compartment was substantially exhausted across states (~31% terminally exhausted at Dx, 28% PTR, 23% LTR), spread along a progenitor-to-terminal trajectory, TOX highest at Dx. Terminal exhaustion of T-cells did not rise from diagnosis to relapse and did not differ across states (p=not-significant), consistent with a standing feature rather than one acquired at PTR. The blast's NECTIN2/PVR ligands met their CD8 receptors CD96 (72%) and TIGIT (15%; inferred by LIANA/CellPhoneDB). In contrast, the receptor for blast CD200, CD200R1, is a myeloid receptor near-absent from CD8 (1.5%), so CD200 acts on the myeloid niche rather than on CD8. For the shared NECTIN2/PVR ligands, inhibitory CD96/TIGIT exceeded activating DNAM-1/CD226 ~twofold by expression (CD96-dominant) in every state, a standing balance present from Dx, not acquired at PTR. Despite this, the T cells kept cytotoxic granzymes (GZMB, GNLY) and lost only GZMK (an earlier, less-differentiated CD8 marker).
Two external post-chemotherapy relapse cohorts were utilized to compare with PTR. In Naldini’s cohort, malignant stemness likewise rose at relapse (LSC17, GEE p<0.001), implying that the primitive shift is not just PTR specific, but shared with any relapse regardless of treatment modalities. However, CD8 T-cell exhaustion differed by time of origin: in Ennis’ cohort, it rose toward relapse after chemotherapy, whereas in our cohort it was already present at diagnosis and did not rise - a pre-existing rather than an acquired pattern; the cytotoxic granzyme program (GZMB retained, GZMK lost) was shared with Ennis’ cohort.
Conclusions
PTR AML behaves as an armed primitive blast meeting a pre-existing exhausted, but still-cytotoxic, donor CD8 T-cell compartment. The blast up-regulates inhibitory ligands (NECTIN2/PVR, CD200); its NECTIN2/PVR ligands are co-expressed with CD8 CD96/TIGIT, inhibition outweighing the activating DNAM-1 counterpart, a standing balance rather than a relapse-acquired switch. Because these T cells remain cytotoxic yet restrained, the CD8 checkpoint axis (TIGIT/CD96) is a rational target (with CD200 a myeloid/blast-directed candidate), and redirected T-cell approaches remain viable.