Dynamic Analysis of Human Natural Killer Cell Response at Single-Cell Resolution in B-Cell Non-Hodgkin Lymphoma

Abstract

Natural killer (NK) cells are phenotypically and functionally diverse lymphocytes that recognize and kill cancer cells. The susceptibility of target cancer cells to NK cell-mediated cytotoxicity depends on the strength and balance of regulatory (activating/inhibitory) ligands expressed on target cell surface. We performed gene expression arrays to determine patterns of NK cell ligands associated with B-cell non-Hodgkin lymphoma (b-NHL). Microarray analyses revealed significant upregulation of a multitude of NK-activating and costimulatory ligands across varied b-NHL cell lines and primary lymphoma cells, including ULBP1, CD72, CD48, and SLAMF6. To correlate genetic signatures with functional anti-lymphoma activity, we developed a dynamic and quantitative cytotoxicity assay in an integrated microfluidic droplet generation and docking array. Individual NK cells and target lymphoma cells were co-encapsulated in picoliter-volume droplets to facilitate monitoring of transient cellular interactions and NK cell effector outcomes at single-cell level. We identified significant variability in NK-lymphoma cell contact duration, frequency, and subsequent cytolysis. Death of lymphoma cells undergoing single contact with NK cells occurred faster than cells that made multiple short contacts. NK cells also killed target cells in droplets via contact-independent mechanisms that partially relied on calcium-dependent processes and perforin secretion, but not on cytokines (interferon-γ or tumor necrosis factor-α). We extended this technique to characterize functional heterogeneity in cytolysis of primary cells from b-NHL patients. Tumor cells from two diffuse large B-cell lymphoma patients showed similar contact durations with NK cells; primary Burkitt lymphoma cells made longer contacts and were lysed at later times. We also tested the cytotoxic efficacy of NK-92, a continuously growing NK cell line being investigated as an antitumor therapy, using our droplet-based bioassay. NK-92 cells were found to be more efficient in killing b-NHL cells compared with primary NK cells, requiring shorter contacts for faster killing activity. Taken together, our combined genetic and microfluidic analysis demonstrate b-NHL cell sensitivity to NK cell-based cytotoxicity, which was associated with significant heterogeneity in the dynamic interaction at single-cell level.

Keywords: dynamic analysis; lymphoma; microfluidic system; natural killer cell cytotoxicity; non-Hodgkin; single cell.

Figure 1

Natural killer (NK) ligands in lymphoma and cytoxicity. (A) Heatmap represented by hierarchical clustering show signature of gene expression patterns related to NK cell regulatory ligands in lymphoma cells and primary tumors. (B) Bar graph represents cell-mediated cytotoxicity as percent target lysis in lymphoma target cells cocultured with increasing ratio of NK cells. Data represented as mean lysis ± SD of three replicates. Diffuse large B cell lymphoma (DLBCL).1 and DLBCL.2 represents primary DLBCL cells obtained from two different patients. BL indicates primary Burkitt lymphoma cells from one patient.

Figure 2

Natural killer (NK) cell dynamics in droplets. (A) Schematic of droplet microfluidic platform indicating NK cell and target cell channels and droplet docking array. (B) Contact between NK (unlabeled) and target SUDHL10 (Calcein AM labeled) cell in droplet. Loss of target cell viability was observed at 145 min. (C) Target cell detaches after contact with NK cell and remains viable. Scale bar: 50 µm. (D) Normalized fluorescent intensity (N.I) profile of live target cells (upper panel) and dying target cells (lower panel).

Figure 3

Quantitative analysis of NK cell interactions in droplets. Heat maps showing periods of contact (blue), non-contact (yellow), and death (red) during NK cell interaction with (A) SUDHL10 cells and (B) peripheral blood monocytes (PBMC). (C) Death of SUDHL10 (n = 265) and PBMC (n = 67) in droplets mediated in the presence or absence of NK cells. *indicates p < 0.005. Data show mean ± SD of two independent experiments. (D) Frequency of contacts made by SUDHL10 and PBMC with NK cells. (E) Distribution of contact duration between NK and target cells (SUDHL10 = 82; PBMC = 49) in droplets (* indicates p = 0.008). (F) SUDHL10 cell death due to contact-dependent and -independent mechanisms in droplets (p < 0.0005). Healthy NK cells were obtained from two different donors. Cell pairs assessed for Donor 1: 175; Donor 2: 90. The mean value of each distribution in (E,F) is indicated by the red line.

Figure 4

Effect of inhibitors on Natural killer (NK)-mediated cytotoxicity. (A) SUDHL10 cells were incubated in droplets with conditioned media (CM) from IL-2-treated (CM-IL-2; n = 199) or untreated (CM-No IL-2; n = 169) NK cells. Control SUDHL10 cells were incubated in growth media (n = 236). Target cell death was determined in the absence of NK cells over 4 h. Data show Mean ± SD of two independent experiments. (B) SUDHL10 cell death in droplets, with or without NK cells, in the presence of inhibitors Brefeldin (n = 144) and ethylene glycol tetraacetic acid (EGTA) (n = 148). Untreated: n = 175 cells. Data show mean ± SD of two independent experiments. (C) Time required for EGTA-induced SUDHL10 cell death compared to untreated cells (EGTA: n = 148; untreated: n = 175 cells). The mean value of each distribution is indicated by the red line. * indicates p < 0.0002.

Figure 5

Cytokine secretion by natural killer (NK) cells. (A) IFN-γ, (B) TNF-α, and (C) Perforin secretion by NK cells, cocultured with or without target SUDHL10 cells, in the presence of inhibitors Brefeldin and EGTA. Mean ± SD of three replicates.

Figure 6

Response of patient-derived lymphoma cells in droplets. (A) Death of primary Diffuse Large B cell Lymphoma (DLBCL) and Burkitts Lymphoma (BL) cells in droplets in the presence (with NK cells) or absence (w/o NK cells) of NK cells. DLBCL.1: n = 150; BL: n = 95. (B) Serial contacts made by NK cells with DLBCL.1, DLBCL.2, and BL cells in droplets. (C) Duration of contacts made by DLBCL and BL cells. Mean values are indicated in red. (D) Primary cancer cell death due to contact-dependent mechanisms in droplets. Mean values are indicated in red. Primary cells were obtained from one patient per disease subtype and data reported from one experiment per cell type. (A,B) % values shown for pooled data.

Figure 7

Comparison of killing efficiency of primary natural killer (NK) cells with NK92 cells. (A) Representative heat map showing periods of contact (blue), non-contact (yellow), and death (red) during NK92 cell interaction with SUDHL10 cells. (B) Schematic illustrating delay time (T_D) calculation and T_D detected for SUDHL10 cells from pooled data. (C) Distribution of contact duration between different types of NK cells and SUDHL10 target cells in droplets (* indicates p < 0.0008). (D) SUDHL10 cell death mediated by NK92 cell line and primary CD56⁺ NK cells. The mean value of each distribution in (C,D) is indicated by the red line. Cell pairs evaluated for NK92: 77; Primary NK: 76 from two independent experiments.