Local and Systemic Effects of Immune Checkpoint Blockade on Relapsed Myeloid Malignancies Following Allogeneic Hematopoietic Stem Cell Transplantation

Relapse after allogeneic hematopoietic stem cell transplantation (HSCT) remains an unmet medical need. The ETCTN 9204 study evaluated in 71 study subjects if immune checkpoint blockade with anti-CTLA-4 (Ipilimumab (Ipi), n = 43) or anti-PD-1 (Nivolumab (Nivo), n = 28) antibodies could stimulate graft-versus-leukemia (GvL) responses in this setting. We focused mainly on patients (pts) with relapsed AML/MDS, which constituted the majority of pts (n = 38; 54%). Ipi induced both long-term complete remissions (CR; n = 3) and transient CRs (TR; n = 3), while Nivo did not generate any CRs, but 4 patients demonstrated partial remissions (PR).

To define the molecular features associated with response to Ipi, we performed bulk transcriptomic analyses of formalin-fixed paraffin-embedded biopsies of sites of AML/MDS involvement (n = 35) collected before and after Ipi treatment from 13 pts. Our analysis of matched pre- and post-Ipi samples of patients with CRs identified 50 differentially expressed genes. By gene ontology analysis, these were enriched for signatures of 'lymphocyte activation' and 'antigen-receptor signaling'. Principal component analysis using these genes separated post-Ipi CR samples as a distinct cluster, transcriptionally apart from pre-Ipi CR samples and from non-responder (NR, n = 15) pre and post-Ipi samples. Of note, post-Ipi CR samples were similar to both pre- and post-Ipi TR samples as well as samples from GvHD/toxicity biopsies (n = 9). Consistent with these findings, we detected increased CD8⁺ T cell abundance post-Ipi in CR but not NR samples with CIBERSORTx (pre vs post, median score 0.043 vs. 0.56, p < 0.01). Likewise, reconstruction of T cell receptor (TCR) CDR3 sequences using the tool TRUST showed an increase in TCR clonotypes per million reads post-Ipi in CR samples (pre vs post, median 0.33 vs. 1.65, p < 0.05). In sum, response to CTLA-4 blockade is characterized by transcriptional evidence of T cell infiltration and activation within the tumor microenvironment, with similar gene programs observed in the setting of GvHD.

To determine if the changes within tissue sites following Ipi are also observed in peripheral blood (PB), we analyzed the immunophenotype and TCR repertoire of T cells from serially collected PB samples from pts with AML/MDS (n = 14) or non-myeloid malignancy (n = 6). In responders and non-responders, exposure to Ipi was associated with decreased naïve, increased effector memory, and increased expression of HLA-DR on central memory T cells (p < 0.05), consistent with T cell differentiation and activation. From 9 of 14 AML/MDS pts (CR = 4, NR = 5), bulk TCR sequencing before and after 1 cycle of Ipi yielded 572,017 PB TCR sequences. Only 776 clonotypes demonstrated significant change in frequency, and these TCRs were detected in greater proportion among responders (613/776 versus NR 163/776, p < 0.01). Thus, Ipi alters the differentiation states of circulating T cells irrespective of response and leads to higher TCR repertoire turnover in CR patients.

Of the AML patients achieving PR following Nivo, we also observed transcriptional evidence of CD8⁺ T cell infiltration in one patient. This signature, however, was not detected in a second such patient. In a separate instance, we had the opportunity to dissect the molecular features of a partial response in a patient with JAK2^V617F+ secondary AML following Nivo through single-cell RNA and ATAC-sequencing (scRNA-seq and scATAC-seq, 10x Genomics) of PB mononuclear cells collected serially across 6 timepoints. In total, we analyzed the transcriptomes and chromatin accessibility data from of 27,777 and 28,713 individual cells, respectively. At time of response, non-exhausted CD4⁺ T cells expanded while both exhausted CD8⁺ T cells and a malignant subpopulation with increased expression of PD-L1 and features of megakaryocytic differentiation preferentially contracted. The subsequent expansion of a PD-L1^- malignant population at progression suggests that decreased leukemic PD-L1 expression was associated with relapse.

Altogether, these data highlight the molecular and cellular features of successful reinvigoration of GvL using CTLA-4 blockade, from increased local T cell infiltration and activation in the leukemic microenvironment to peripheral T cell turnover. In addition, the selection of therapy-resistant subclones after PD-1 blockade underscores the need for further high-resolution studies of GvL responses.