Immune Senescence and Exhaustion Correlate with Response to Flotetuzumab, an Investigational CD123×CD3 Bispecific Dart® Molecule, in Acute Myeloid Leukemia

We have recently shown that bone marrow (BM) RNA profiles stratify patients with acute myeloid leukemia (AML) into immune-infiltrated and immune-depleted subtypes and that type I/II interferon (IFN)-related gene signatures associate with complete response to flotetuzumab (FLZ), an investigational CD123×CD3 bispecific DART molecule. Within the AML tumor microenvironment CD8⁺ T cells exhibit features of immune exhaustion and senescence (IES). IES are dysfunctional states driven by metabolic alterations in the tumor microenvironment (TME) and emerging targets for cancer immunotherapy. The aim of the current study was to determine whether IES predicts response of relapsed-refractory (R/R) AML to FLZ in the CP-MGD006-01 clinical trial.

Based on prior knowledge and gene set enrichment analysis, we derived a 61-gene IES signature score from RNA-sequencing datasets (TCGA and Beat-AML Master Trial; 162 and 281 patients, respectively). The immunotherapy cohort included 139 BM samples from 71 patients with R/R AML treated with FLZ at the RP2D of 500 ng/kg/day (NCT02152956). BM samples were collected at time of study entry (n=71; n=66 with response data) and longitudinally post-cycle (PC)1 (n=40), PC2 (n=18), PC3 and 4 (n=4) and end of treatment (n=6). AML status at study entry was classified as primary induction failure (PIF, defined as lack of response to at least 2 induction treatment cycles), and early (ER) or late relapse (LR), defined as complete remission (CR) of <6-month or ≥6-month duration, respectively. Overall response rate (ORR), collectively complete response, was defined as <5% BM blasts (CR, CRh, CRi or MLFS), and partial response (PR) was defined as >50% decrease or decrease to 5-25% BM blasts. RNAs were profiled on the PanCancer IO 360™ gene expression panel on the nCounter^® platform. Formalin-fixed paraffin embedded BM biopsies were profiled using the human IO protein and RNA panels on the GeoMx^® digital spatial profiler (DSP).

The 61 genes in the IES signature included T/NK-cell markers (granzymes, CD8A, KLRD1, KLRK1), immune checkpoints (ICOS, CTLA4, EOMES), IFNG and IFN-stimulated genes (CXCR6, IFIH1, IL10RA, GBP1), and were enriched in KEGG pathways related to Th1/Th2 differentiation, TCR signaling, cytokine-cytokine receptor interaction, NK-mediated cytotoxicity and CD28 costimulation (false discovery rate<0.001 for all; Fig. 1A). Unsupervised hierarchical clustering of gene expression allowed the identification of BM samples with high, intermediate and low IES scores at time of study enrollment (Fig. 1B). Ninety-five percent (18/19) of patients in the IES^high cluster had PIF/ER AML, congruent with prior studies showing enhanced immune infiltration and IFN signaling in the TME of patients with PIF. Notably, ORR to FLZ (complete response, n=18 or PR, n=5) were documented in 11/19 (58%), 10/32 (31.2%) and 2/15 (13.3%) of patients in the IES^high, IES^int and IES^low cluster, respectively (Fig. 1B). The IES signature score was significantly higher at baseline in patients who responded to FLZ compared with non-responders (P=0.0052; Fig. 1C). High-dimensional flow cytometry of sequential BM samples collected at time of study entry and PC1 of FLZ showed the on-treatment upregulation on both CD4 and CD8 T cells of early activation markers CD69 and CD38 (but not the late activation marker HLA-DR), as well as immune checkpoints LAG3 and Tim-3, and proliferation marker Ki-67, indicating FLZ-mediated modulation of the immune TME. To determine the variation in co-expression of T-cell markers associated with FLZ treatment, we also measured lymphocytes obtained from 21 BM samples prior to and post-FLZ using an unsupervised multivariate analysis. Qualitative comparisons of the principal component analysis (PCA) showed distinct phenotypic changes in BM samples post-treatment (Fig. 1D). Characterization of BM biopsies using GeoMx DSP showed distinct T-cell clustering in responders (Fig. 1E). PCA showed enhanced CD45, CD3, CD4 and PDL1 in situ RNA/protein expression (fold change 1.96, 2.83, 3.32, 4.7, respectively, P<0.05 for all) at PC1 of FLZ in OR versus non-responders (Fig. 1F).

In conclusion, features of IES were associated with response to FLZ. T-cell functional rejuvenation by FLZ could benefit patients with R/R AML by counteracting pre-existing immune dysfunction.