Pembrolizumab and decitabine for refractory or relapsed acute myeloid leukemia

Abstract

Background: The powerful 'graft versus leukemia' effect thought partly responsible for the therapeutic effect of allogeneic hematopoietic cell transplantation in acute myeloid leukemia (AML) provides rationale for investigation of immune-based therapies in this high-risk blood cancer. There is considerable preclinical evidence for potential synergy between PD-1 immune checkpoint blockade and the hypomethylating agents already commonly used for this disease.

Methods: We report here the results of 17 H-0026 (PD-AML, NCT02996474), an investigator sponsored, single-institution, single-arm open-label 10-subject pilot study to test the feasibility of the first-in-human combination of pembrolizumab and decitabine in adult patients with refractory or relapsed AML (R-AML).

Results: In this cohort of previously treated patients, this novel combination of anti-PD-1 and hypomethylating therapy was feasible and associated with a best response of stable disease or better in 6 of 10 patients. Considerable immunological changes were identified using T cell receptor β sequencing as well as single-cell immunophenotypic and RNA expression analyses on sorted CD3+ T cells in patients who developed immune-related adverse events (irAEs) during treatment. Clonal T cell expansions occurred at irAE onset; single-cell sequencing demonstrated that these expanded clones were predominately CD8+ effector memory T cells with high cell surface PD-1 expression and transcriptional profiles indicative of activation and cytotoxicity. In contrast, no such distinctive immune changes were detectable in those experiencing a measurable antileukemic response during treatment.

Conclusion: Addition of pembrolizumab to 10-day decitabine therapy was clinically feasible in patients with R-AML, with immunological changes from PD-1 blockade observed in patients experiencing irAEs.

Keywords: adaptive immunity; immunotherapy; investigational; lymphocyte activation; therapies; translational medical research.

Conflict of interest statement

Competing interests: CSH receives research support from Merck Sharpe & Dohme and SELLAS Life Sciences Group AG. CL is on the Speakers’ Bureau for Astellas, Jazz Pharma and serves in a consulting or advisory role for Daiichi, Jazz Pharma, Amgen, Abbvie, Macrogenics, and Agios. The remaining authors declare no competing interests.

© Author(s) (or their employer(s)) 2022. Re-use permitted under CC BY. Published by BMJ.

Figures

Figure 1

PD-AML clinical trial schema, sampling intervals, and survival curve. (A) Pembrolizumab was administered on day 1 of every cycle for eight cycles, and decitabine was given on days 8–12 and 15–19 of every other cycle. Bone marrow aspirate and peripheral blood was collected before the start of treatment (day 0), on cycle 1 day 8 (C1D8) after one dose of pembrolizumab, and at the end of every other cycle (EOC2, EOC4, EOC6, EOC8); clinical responses were also assessed at these timepoints. (B) Summary of patients treated on NCT02996474 and responses. These patients have been reported previously. (C) Kaplan-Meier survival curve with 95% CI. (D) Summary of grades 3 and 4 adverse events and irAEs that occurred during treatment. AML, acute myeloid leukemia; EOC2, end of 2 cycles of treatment; EOC4, end of 4 cycles of treatment; EOC6, end of 6 cycles of treatment; EOC8, end of 8 cycles of treatment.

Figure 2

Development of irAEs is associated with clonal T cell expansion. The number of significantly expanded or contracted clones at timepoints sampled during treatment compared to baseline were calculated in (A) peripheral blood (PB) and (B) bone marrow (BM) for patients developing irAEs. Time of irAE indicated by black arrows. TSH levels and significantly expanded T cells at EOC2 and EOC4 in PB that were

Figure 3

Unique expression signatures in clonally expanded T cells at time of irAE. Cells harboring the unique TCRβ CDR3 from differentially abundant clones identified by bulk TCRβ-seq were grouped and differentially expressed genes (DEG) and cell surface proteins (DEP) were identified by supervised analyses compared with top 10 abundant T cell clones in the sample identified by bulk TCRβ-seq. (A) DEG and (B) DEP of expanded CD8+ T cell clones of interest in PD-AML two at hypothyroidism diagnosis vs other abundant CD8+ T cell clonotypes. (C) Frequencies from TCRβ-seq data of clonotypes coexpressing cell surface PD-1, TIM-3, and CD27 (four clones, top panel) and those not expressing this signature (two clones, bottom panel). (D) DEG and (E) DEP of expanded CD8+ T cell clones of interest in PD-AML three at time of diagnosis of central diabetes insipidus versus other abundant CD8+ T cell clonotypes. (F) Frequencies from TCRβ-seq data of clonotypes coexpressing cell surface PD-1, TIM-3, 4-1BB, HLA-DR, and CD27 (six clones, top panel) and those not expressing this signature (three clones, bottom panel). Heatmaps were created using the mean expression of all cells within each clone; differentially expressed markers were ordered by adjusted p values and fold changes. Adjusted p values are based on Wilcoxon rank sum test with Bonferroni corrections using all features in the dataset. ‘Difference **’ denotes adjusted p

Figure 4

Antileukemic response to combination immunotherapy was not associated with clonal T cell expansion. Number of significantly expanded or contracted clones at timepoints sampled during treatment compared with baseline were calculated in (A) PB and (B) BM for two patients achieving CR by eight cycles of treatment. Decreasing blast percentage in the BM and few to no significantly expanded clones at EOC2 or EOC4 in BM that were low in frequency (

Figure 5

Stable T cell immunophenotypes in patients with anti-leukemic responses. (A) Integrated UMAP visualization of T cell immunophenotyping based on cell surface expression CD4, CD8, CCR7, CD45RA, CD127 and CD25. (B) Quantification and proportion of CD4+ and CD8+ T cell subset frequencies, as a percentage of CD3+. (C) Differential gene expression analysis among different cell types with scaled log normalization. For panels A-C, T cell subsets indicated in legend. (D) Frequencies of CD4+ and CD8+ effector memory (EM) and terminal effector (TE) T cells at Day0, EOC2, and EOC4. (E) Per cent change in frequencies of CD4+ and CD8+ EM and TE cells at EOC2 and EOC4 compared with day 0. (F) T cell subset clonality in T cell subsets identified by immunophenotyping. Patients and time points indicated in key. (G) Cells expressing exhaustion (purple) and cytotoxicity (orange) predefined gene sets overlaid on integrated UMAP with exhaustion and cytotoxicity scores for CD4+ and CD8+ T cells. Scores were calculated for each cell based on mean expression of genes in the gene set and subtracting the background. For panels D–F, individual patients indicated in legend.

Figure 6

No signal of immune evasion in responders at relapse post-treatment. Through 3’v3 scRNA-seq on unsorted BMMCs, clustering was based on the cell surface protein expression and clusters were annotated by their top three or four most highly differentially expressed markers. Clusters were classified as putative leukemia using on protein expression of known leukemia-associated and myeloid markers based on the patients’ previous clinical flow cytometry records. Relative expression of HLA molecules and PD-L1 on leukemic blasts at baseline (top) and at relapse (bottom) in (A) PD-AML 1 and (B) PD-AML 5 visualized using ridge plots. Single-cell DNA and antibody-oligonucleotide sequencing of unsorted BMMCs in (C) PD-AML 1 and (D) PD-AML five shows the genomic and immunophenotypic features of the leukemic (AML, red line) and preceding/distinct CHIP (clonal hematopoiesis, gray lines) clones present at relapse as compared with day 0 (adapted from41). Left: Genomic subclones with wild type (WT), heterozygous (HET), present (Gain/+) or absent (−) features. Right: Cell surface protein expression for each subclone.