Single-Cell Transcriptomic Analysis Reveals Loss of Activated Bone Marrow NK Cells in Multiple Myeloma Patients Which Associates with Disease Progression in Mice

Introduction

Multiple Myeloma (MM) disease progression and therapy response are the net result of tumor cell-intrinsic features and tumor cell-extrinsic cues from the bone marrow (BM) microenvironment. Natural killer (NK) cells are mediators of the cytotoxic immune response against MM and are important effector cells in antibody-based immune therapies, especially anti-CD38 monoclonal antibodies such as Daratumumab. Classically, NK cells are divided into a cytotoxic CD56 ^dim subset, important for antibody-dependent cellular cytotoxicity, and a cytokine-producing CD56 ^bright subset releasing inflammatory mediators such as IFNγ, TNFα and GM-CSF. However, accumulating evidence suggests greater heterogeneity in the NK cell compartment and modulation of these NK cell subsets could impact disease progression and response to NK cell-driven immunotherapies. Here, we combined the 5TGM1 murine model of MM with single-cell RNA sequencing of bone marrow (BM) NK cells of newly diagnosed MM patients to map NK cell heterogeneity and to investigate their role in MM progression.

Results

To gain insight in NK cell heterogeneity in MM disease we performed single-cell RNA sequencing on immune cells of viably frozen BM aspirates from 19 newly diagnosed MM patients and 5 non-cancer control patients. NK cells were identified in silico by transcription of KLRF1, KLRD1, GNLY and NKG7 resulting in a single-cell transcriptomic dataset of 30,373 NK cells from MM patients and 8,865 NK cells from control patients. Conventional CD56 ^bright and CD56 ^dim NK-cells were identified by increased transcription of GZMK or GZMB, respectively. The GZMK ⁺CD56 ^bright NK cells contained clusters of naïve and activated NK cells. The GZMB⁺CD56 ^dim NK cells consisted of 5 subclusters. To identify MM-induced alterations in NK cell subsets, we compared GZMK ⁺CD56 ^bright vs GZMB⁺CD56 ^dim cluster composition and distribution between controls and MM patients. Control BM was dominated by GZMB-transcribing cytotoxic CD56 ^dim NK cells, resulting in a low ratio of cytokine-producing GZMK ⁺CD56 ^bright vs cytotoxic GZMB⁺CD56 ^dim NK cells. In contrast, MM bone marrow was characterized by heterogeneity of this ratio with a subset of patients presenting with complete reversal of this ratio . In this subset of patients, the altered composition was due to a loss of cytotoxic GZMB ⁺CD56 ^dim NK cells, and more specifically a loss of NK cells with a transcriptome suggesting recent activation.

To better examine the significance of cytotoxic NK cells in MM disease course we utilized the well-established 5TGM1 mouse model. C57Bl/6 and KaLwRij mice both received 10 ⁶ 5TGM1-GFP cells intravenously. Three weeks after tumor injection all KaLwRij mice (18/18) developed MM, defined by >5% tumor cells in BM ("unrestrained tumor") and serum M-protein >2mg/ml. Interestingly, while 39% (7/18) of C57Bl/6 mice had no tumor, 44% (8/18) had low but detectable levels of MM cells (0.1-5% of BM cells, "restrained tumor") and 17% (3/18) presented with an unrestrained MM with BM tumor load similar to that seen in KaLwRij mice. With time the percentage of mice with unrestrained tumor increased (5/12, 42%) at the expense of restrained tumor (2/12, 16%). We hypothesized that C57Bl/6 mice with low tumor load could represent a model of immune-mediated tumor control. Detailed analysis of the NK cell compartment revealed an expansion of activated mature (CD69 ⁺ CD11b ⁺CD27 ⁺) NK cells in C57Bl/6 mice with restrained BM MM (p=0.0031). In contrast, high BM tumor burden in both genotypes was associated with a sharp decline in absolute numbers of activated NK cells.

Conclusion:

Through a combination of single-cell transcriptomic analyses of the BM immune microenvironment in MM patients and experimental mouse models we found a loss of activated NK cells in a subset of patients and mice. Our data suggests that loss of these activated NK cells is associated with MM progression in vivo. A subset of MM patients presented with a loss of activated cytotoxic GZMB ⁺CD56 ^dim NK cells in the BM, suggestive of reduced cytotoxic anti-tumor responses. Meanwhile, in vivo, high disease burden only occurred in mice with an absence of activated NK cells. Current analyses are focused on differences in human disease progression and efficacy of Daratumumab-based therapies in patients with various NK cell phenotypes.