Graft Versus Leukaemia (GvL): Identification & Characterisation of Alloreactive Antigens and Cognate T Cell Responses in Acute Myeloid Leukemia

Allogeneic stem cell transplantation (allo-SCT) is the most established, curative T cell cancer immunotherapy. The curative effect of allo-SCT relies on a graft-versus-leukemia (GvL) effect; but this can be difficult to dissociate from graft-versus-host disease (GvHD). For decades, the key unsolved challenge has been to identify common alloreactive peptide-human leukocyte antigen (HLA) complexes targeted by T cells that mediate GvL but not GvHD. We developed a workflow for unbiased identification of alloreactive antigens and analysis of cognate donor alloreactive T cells and their T cell receptors (TCRs; Fig 1). We studied sequential bone marrow and peripheral blood samples, collected over 15 years, from 12 high-risk AML patients who remain in continuous complete remission following allo-SCT (all >2 years). All patients had relapsed at least once pre-allo-SCT and/or were AML flow cytometry minimal residual disease (MRD)+ pre-allo-SCT. 7 had received donor lymphocyte infusions. No patients developed >1 grade GvHD. We performed whole exome sequencing of DNA from isolated patient T cells (patient germline), AML blasts and donor DNA to identify: (i) AML-specific coding variants; and (ii) germline coding single nucleotide polymorphisms (SNPs), mismatched between patient and donor. 98.7% of the mismatched variants were germline. More variants were identified in unrelated compared with sibling donors (mean: 5744 vs 3253 variants). To narrow the list of variants taken forward for functional studies, we applied two sequential filtering steps: (i) we restricted analysis to variants expressed in the patients' AML cells (determined by RNA-Seq). 28.7% of all mismatched coding variants were expressed in AML cells; and (ii) we used in silico prediction of HLA-binding affinities of peptides spanning variant amino acids to patient-specific HLA-I and -II using NetMHCpan 4.0 and NetMHCIIpan 3.0. To functionally identify alloreactive antigens, we then screened ~200-900 peptides per patient-donor pair (total 5300 peptides) for recognition by T cells present in patient peripheral blood collected 2-7 years after allo-SCT. This enriches for the identification of long-lived anti-AML responses at a time when patients were both infection- and GvHD- free. Response screening, using IFNγ ELISPOT assays, identified T cell responses to 24 germline, mismatched epitopes. In all cases, responses were only observed when the amino acid encoded by the patient SNP, but not the donor allele, was present in the peptide. The phenotype of the epitope-specific T cells was established by intracellular cytokine staining (ICS) and flow cytometry analysis, and the HLA restriction of each T cell response was determined using HLA-I, -DR, -DP and -DQ blocking antibodies. Of the 24 alloreactive peptides,19 elicited CD4+ HLA class II restricted responses and 3 were targeted by CD8+ class I restricted T cells. Two remain to be characterised (Fig 2). The population allele frequency of the germline variant SNPs targeted varies from 0.5% to 72% in the Caucasian population. Likewise, some of the restricting HLA alleles are very common (e.g. a HLA-DP allele presenting 3 epitope peptides is present in ~66% of the Caucasian population). ICS showed that epitope-specific CD4+ T cells produce Th1 cytokines (IFNg, TNFa) and degranulate (express CD107a) in response to peptide stimulation (Fig 2). Next, we performed single cell 5'RNA-seq with TCR-seq of epitope-restricted T cells (isolated via IFN-g catch assay) for all CD4+ epitope-specific T cells. Alloreactive CD4+ T cells have cytokine-producing and cytotoxic effector potential, expressing MIP1a and b, IFNg, granzyme B and FasL (Fig 2). Finally, lentiviral-based gene transfer of one of the candidate alloreactive epitope-specific TCRs into third party primary T cells confers the capacity to generate IFNg and TNFa in response to peptide stimulation, and to recognise primary AML cell-derived antigen in the presence of monocyte-derived dendritic cells (Fig 2). In vitro and in vivo killing assays are in progress. The high proportion of CD4+ T Cell responses identified by our pipeline is concordant with downregulation of HLA II expression on AML blasts at relapse post-allo-SCT, which is a common mechanism of immune evasion, found in one-third of patients. Our approach dissects the biology of alloreactive anti-AML GvL and GvHD responses and to manipulate them for therapeutic purposes. Figure 1View largeDownload PPTFigure 1View largeDownload PPT Close modal