Clonal and Single Cell Dynamics of Resistance to Graft-Versus-Leukemia (GvL) in Chronic Lymphocytic Leukemia (CLL)

Although the GvL effect is the curative basis for allogeneic hematopoietic stem cell transplantation, the key cellular and molecular mechanisms driving GvL sensitivity and resistance remain incompletely understood. Using genomic approaches, we systematically characterized the changes in cellular composition and state of CLL and non-CLL cells in two settings of effective GvL: reduced intensity conditioning regimens (RIC) and donor lymphocyte infusion (DLI).

We identified 10 patients with CLL progression after RIC, 6 of whom had complete responses before relapse. To define the evolutionary trajectories of CLL cells after RIC, we generated paired whole-exome sequencing data from pre- and post-RIC CLL cells sorted from PBMCs. We used DNA from autologous CD4+ T cells for germline comparison and the algorithms MuTect2 and ABSOLUTE to identify somatic alterations with corresponding cancer cell fractions (CCFs). 5 of the 10 patients had clonal mutations in TP53 and/or SF3B1 pre-transplant. Mutation burden was higher at baseline than previously described for CLL (mean 29.7 vs 17.9 non-silent SNVs/exome, p<0.0001, Student's t-test) but not different from post-transplant. Neither mutations nor altered expression (based on RNA sequencing) of HLA class 1 or 2, or B2M were observed. 8 relapse pairs exhibited complex branched evolution involving CCF shifts of subclonal and clonal mutations whereas two relapse pairs showed CCF shifts only in subclonal mutations. Presence of clonal shifts associated with active immunity (off immune suppression or presence of chronic graft-vs-host disease; p=0.02, Fisher's exact test), longer time to relapse (>1 vs <1 year; p=0.02), and achievement of complete response (p=0.05). These data suggest that immune selective pressure by GvL can lead to gain of resistance capability, potentially facilitated by the replacement of dominant clones.

We likewise saw diverse clonal trajectories in 2 index cases of DLI response followed by relapse, either 11 [branched evolution] or 1.5 [linear] years after DLI. To deeply examine co-evolution of CLL and immune cells during DLI-relapse, we performed single cell RNA sequencing of both cell types collected from 4 paired PBMC samples representing either pre- or post-(relapsed)-DLI time points. Using the inDrop platform, we profiled a median of 11,686 (range: 9,101-16,756) cells per sample with a median of 5,363 CD19+ CD5+ expressing CLL cells (range: 3,622-9,463).

We first sought to define the transcriptional heterogeneity underlying CLL cells during DLI relapse. Data visualization using t-distributed stochastic neighbor embedding plots revealed broad transcriptional shifts in CLL clusters from pre- to post-DLI and also showed the complexity of transcriptional substructure to more closely relate to a patient's own genomic structure rather than a common CLL phenotype, in contrast to prior studies. DLI-relapsed CLL cells in both patients were marked by upregulation of CXCR4 and members of the RhoGTPase family, suggesting migration capacity and cytoskeletal remodeling to play a role in GvL escape. GO term enrichment analysis identified DLI sensitive CLL cells in these cases to associate with regulation of lipid and lipoprotein metabolism and interferon signaling.

We then determined parallel changes in PBMC immune states over time, which were subtle and not related to time point. To determine if the leukemic microenvironment can differentially affect immune states, we profiled, in total, 32,777 single bone marrow mononuclear cells (BMMC) from pre-DLI, during DLI response, and post-DLI relapse for one patient. Unlike PBMCs, BMMC-derived T cells clustered preferentially by time point, then state of differentiation. DLI response induced a pronounced shift in all T cell states, reflected by upregulation of NFKB and PI3K-AKT signaling; a dysfunctional state marked by metallothionein family expression, recently discovered in murine single cell studies, was unique to the post-DLI relapse timepoint in this patient.

Altogether, these data suggest that GvL selective pressure can shape genetic evolutionary trajectories; scRNA-seq analysis of the 2 informative DLI cases is consistent with the notion that the CLL microenvironment shapes immune states during GvL response and relapse. Ongoing studies will dissect the molecular pathways governing these trajectories to suggest therapeutic strategies for overcoming GvL resistance.