EBV⁺ LCLs were generated from 11 healthy donors by infection of freshly isolated PBMCs with the B95.8 laboratory strain of EBV, as previously described (33). LCL and the MHC-I^lo 721.221 cell line were maintained in RPMI (Corning) supplemented with 10% FBS (Serum Source International) and 1% penicillin/streptomycin (Corning) [complete RPMI (cRPMI)]. Primary NK cells or B cells were negatively selected from whole blood using the RosetteSep Human NK Enrichment Kit or Human B Cell Enrichment Kit, respectively (Stem Cell Technologies). Purity was routinely (90% (Figures S1A,B in Supplementary Material). Purified primary NK cells were cultured for 2 days in cRPMI supplemented with 300 U/mL IL-2 (NIH Reagent Program) prior to stimulation or coculture. This study was performed in accordance with the Declaration of Helsinki and approved by the Stanford University Institutional Review Board, and written informed consent was obtained from all participants.

Natural killer cell cytotoxicity was assayed by a modified ACT1 assay (Cell Technology). Briefly, target cells (721.221, primary B cells, autologous LCL) were incubated with 0.25 µM CFSE in PBS + 2.5% FBS for 5 min at room temperature, then washed twice with 10 volumes cRPMI. A total of 0.5 × 10⁵ target cells were cocultured for 4 h in a 37°C-5% CO₂ humidified incubator with 2 × 10⁵ NK cells, for a final ratio of 4 NK cells:1 target cell. Cocultures were pelleted, resuspended in 200 µL cRPMI, and incubated with 5 µL 7-aminoactinomycin D (7-AAD) for 15 min on ice. Unlabeled target cells served as a control for gating, while CFSE-labeled target cells treated with 1× final FACS Perm (BD Pharmingen) for the final 15 min of the coculture served as a positive control for maximum target cell death. Cells were analyzed on a FACScan flow cytometer with CellQuest Software, and dead target cells were identified as CFSE⁺7-AAD⁺. Specific killing was calculated as [(%CFSE⁺7-AAD⁺_NK:Target − % CFSE⁺7-AAD⁺_{Target Only})/(% CFSE⁺7-AAD⁺_{Permeabilized Target} − % CFSE⁺7-AAD⁺_{Target Only})] × 100.

For the cytotoxicity assay using purified NKG2A⁺ and NKG2A⁻ NK cells, NK cells were isolated using an NK Isolation Kit (Miltenyi) and cultured in RPMI complete with 300 U/ml IL-2 for 2 days. NK cells were then stained with NKG2A-PE (Beckman Coulter, clone Z199), CD3-APC (Biolegend), CD56-PE Cy7 (Biolegend), CD16-PerCP Cy5.5 (Biolegend), and Fixable Viability Dye eFluor 780 (eBioscience) and sorted for live CD3⁻CD56⁺CD16⁺ and NKG2A^± on a BD FACSAria. Sorted cells were rested in RPMI complete with 300 U/ml IL-2 for 2 h at 37°C before the start of the killing assay. LCLs were labeled with 0.5 µM CellTrace Violet (Invitrogen) for 20 min at 37°C, then washed twice with at least 10 volumes of RPMI complete. LCL target cells (2.5 × 10⁴) were cocultured with 1 × 10⁵ sorted NKG2A⁺ or NKG2A⁻ NK cells for 4 h in a 37°C incubator for a 4:1 E:T ratio. At the end of the coculture, cells were stained for viability using Fixable Viability Dye eFluor 780 (eBioscience) at the manufacturer’s recommended concentration, for 10 min on ice. Cells were then washed twice with PBS before analysis as above.

PBMCs were isolated from whole blood by Ficoll-Paque (GE Healthcare). PBMC and LCL were stained with the following antibodies or matched isotype controls: anti-CD48-PE, anti-CD19-FITC, anti-CD19-PE, anti-HLA-E-PE, and anti-HLA-A,B,C-FITC (BD Pharmingen) and analyzed by flow cytometry. All antibodies were titrated on positive control cell lines; these lines were stained as a positive control in each experiment. Stain indexes were used to compare NK ligand expression on CD19⁺-gated primary B cells and autologous LCL (34). Stain indexes were calculated using the mean fluorescence intensities (MFI) as (MFI_ligand − MFI_isotype)/(2 × SD_isotype).

IL-2 primed NK cells (0.25 × 10⁶ cells/condition) were cocultured with equal amounts of 721.221, primary B cells, or autologous LCL in 200 µL final volume (Figure S2A in Supplementary Material). NK cells were left unstimulated (NK only) as a negative control or were stimulated with 1× PMA/Ionomycin Cell Stimulation Cocktail (eBioscience) as a positive control. All cells were cultured in cRPMI supplemented with 300 U/mL IL-2, 1× final brefeldin A/monensin protein transport inhibitor (eBioscience), and anti-CD107a-Pacific blue (BioLegend). After 4 h incubation in a 37 C-5% CO₂ humidified incubator, 1.8 mM EDTA was added for 5 min to arrest stimulation. Cells were stained with the near IR Live/Dead stain (Life Technologies), washed, and then stained with anti-CD3-Alexa 700, anti-CD14-Alexa 700, anti-CD19-Alexa 700, anti-CD56-Brilliant Violet 605, anti-NKG2D-PE,Cy7, anti-2B4-PerCP,Cy5.5 (BioLegend), anti-CD16-V500 (BD Pharmingen), anti-NKG2C-PE (R&D Systems), and anti-NKG2A-FITC (Miltenyi). Cells were then fixed and permeabilized with FACS Lyse and FACS Perm II (BD Pharmingen), according to the manufacturer’s instructions. Cells were then stained with anti-IFNγ-Brilliant Violet 785. All antibodies were titrated on PBMC prior to use. Fluorescence minus one-stained PBMC were run for each experiment and used to set gates for each parameter. Data were collected on a four-laser LSRII with FACS DiVA software at the Stanford Shared FACS Facility.

Flow cytometry data were analyzed using FlowJo, version 9.7.6 (Tree Star). Statistical analysis was performed using Prism, version 6.0d (GraphPad Software). All paired data were compared using the Wilcoxon matched-pairs signed rank test. A Bonferroni correction was used to adjust for multiple comparisons.

Lymphoblastoid cell lines are CD19⁺ EBV-transformed B cell lines which display a viral latency type III pattern of gene expression and resemble EBV-LPDs like posttransplant lymphoproliferative disorder (35). We, therefore, analyzed expression of NKR ligands on LCL to evaluate how these cells, as a model of predominantly latent infection, stimulate NK cell recognition. We compared NK cell ligand staining on EBV⁺ LCL with primary autologous CD19⁺ B cells (Figure 1; Figure S1A in Supplementary Material). To account for the increased autofluorescence of EBV⁺ LCL, we used the stain index (34) to normalize expression levels to the corresponding isotype staining (Figures 1A–F). After normalization, EBV⁺ LCL (n = 6) display increased expression of the 2B4 ligand CD48 [Figures 1A,D (36)], lower expression of the inhibitory ligand MHC class I (HLA-A,B,C; Figures 1B,E), and no significant difference in expression of the NKG2A and NKG2C ligand HLA-E (Figures 1C,F) compared to autologous, primary CD19⁺ B cells.

We next tested the ability of purified NK cells to kill autologous EBV⁺ LCL using a flow cytometry-based cytotoxicity assay (Figures S1B,C in Supplementary Material; Figure 1G). NK cells robustly kill the MHC-I^lo positive control cell line 721.221 target (87.1 ± 3.7% SEM, Figure 1H). NK cells from healthy donors (n = 7) display significantly more cytotoxicity against autologous EBV⁺ LCL than against primary unstimulated B cells (8 ± 2.1% SEM versus 2.8 ± 0.6% SEM, p = 0.016; Figure 1H). The NK cytotoxicity against autologous EBV⁺ LCL is likely predominantly directed at latently infected LCL, because less than 2% of EBV⁺ LCL express the lytic antigen BZLF1 (Figure S3 in Supplementary Material) (6). Thus, transformed B cells harboring a latent EBV infection can elicit a cytotoxic response by autologous NK cells.

To evaluate the NK cell activity during the targeting of autologous EBV-infected LCL, cocultures were stained with a panel of antibodies to identify NK cell subsets and their function (Figure S2A in Supplementary Material). NK cells (defined as CD3⁺CD14⁺CD19⁺CD56⁺ live singlets, Figure S2B in Supplementary Material) capable of releasing cytotoxic granules and/or producing cytokines were identified by staining for CD107a mobilization and intracellular IFNγ, respectively. In response to the positive control cell line 721.221, 20 ± 2% SEM of NK cells produce IFNγ, 29 ± 2% SEM of NK cells degranulate, and 17 ± 2% SEM of NK cells both degranulate and produce IFNγ above NK only background (Figure 2). In response to the autologous EBV⁺ LCL, 4.1 ± 2% SEM of NK cells produce IFNγ, 13.5 ± 6% SEM degranulate, and 2.6 ± 1.2% SEM of NK cells both degranulate and produce IFNγ above NK only background levels in NK only conditions (Figure 2). Thus, a small population of NK cells can respond to autologous EBV⁺ LCL.

To identify markers that can distinguish NK cells that respond to EBV⁺ LCL, we examined expression of the activating receptors CD16, 2B4, NKG2C, and NKG2D, the inhibitory receptor NKG2A, and the maturity marker CD57 on NK cells that had produced IFNγ and/or degranulated after coculture with autologous EBV⁺ LCL. Within an individual, each marker is expressed on a subset of total NK cells (Figure S4A in Supplementary Material). Expression of NK cell markers also varied among individuals. For example, the frequency of NKG2A⁺ cells in the total NK population ranges from 23.6–76.3%, in line with other published data (19, 37). Responding (IFNγ⁺ and/or CD107a⁺) NK cells are far less likely to express CD16 than non-responding (IFNγ⁺ and/or CD107a⁻) NK cells (Figure 3), consistent with the fact that CD16 expression is downregulated upon NK cell activation (38, 39). There are no significant differences in the frequencies of responding versus non-responding NK cells that were 2B4⁺, CD57⁺, NKG2C⁺, or NKG2D⁺ (Figure 3). In contrast, the frequency of NK cells mounting a functional response to autologous EBV⁺ LCL is significantly enriched for expression of NKG2A compared to NK cells that do not mount a functional response (Figure 3).

To determine whether the population of NK cells that respond to EBV⁺ LCL is distinct from those that respond to uninfected B cells, we compared the expression of the markers CD16, CD57, 2B4, NKG2A, NKG2C, and NKG2D on NK cells responding to autologous EBV⁺ LCL versus primary B cells. There are no significant differences in the frequency of NK cells expressing CD16 in NK cells responding to primary B cells versus autologous EBV⁺ LCL (Figure 4). In contrast, the population of NK cells responding to autologous, EBV⁺ LCLs contains significantly higher frequencies of NKG2A⁺ cells than the NK cells responding to primary B cells (Figure 4). Moreover, the proportion of NKG2C⁺ cells in the subset of CD107a⁺IFNγ⁺ NK cells that respond to autologous EBV⁺ LCL is reduced (Figure 4C). There are no significant differences in the percentage of 2B4⁺, CD57⁺, or NKG2D⁺ NK cells responding to autologous EBV⁺ LCLs versus primary B cells (Figure S4 in Supplementary Material). Together, these data suggest that expression of NKG2A can distinguish a subset of NK cells that can specifically respond to B cells displaying latent EBV infection.

To further characterize the NK cell subsets responding to latent EBV⁺ LCL, we examined the NK cell response to EBV using combinations of the six NKRs 2B4, CD16, CD57, NKG2A, NKG2C, and NKG2D. Using a Boolean approach, we determined the frequency of each of the 64 potential NK cell subsets defined by expression of the six NK cell receptors on NK cells responding, on the basis of CD107a and/or IFNγ expression, to coculture with primary B cells or autologous LCL (shown for each donor in Figures 5A,B, respectively). To identify EBV-responsive NK cell subsets, we normalized the frequency of NK cells responding to autologous LCL to the frequency responding to primary B cells. The average normalized percentage of each of these subsets among the 10 donors is displayed in Figure 5C. This analysis suggests that the NKG2A⁺2B4⁺CD16⁻CD57⁻NKG2C⁻NKG2D⁺ NK cell population is the most responsive against EBV-infected LCL, as it was consistently more responsive to autologous LCL than to primary B cells, though this effect was not statistically significant after controlling for multiple comparisons (Figures 5D–N). Consistent with these observations, the diversity of the responding cells (CD107a⁺IFNγ⁺) was lower in 8 of 10 donors cocultured with autologous LCL compared to primary B cells (Figure S5 in Supplementary Material); this reduced diversity also suggests that the specific response to EBV infection may be driven by a subset of EBV-responsive NK cells. Together, these data suggest that a subset of NK cells, identified as NKG2A⁺2B4⁺CD16⁻CD57⁻NKG2C⁻NKG2D⁺, can respond specifically to EBV⁺ LCL.

To better define the NK cell subset responding to autologous LCL, we evaluated NK cell IFNγ secretion and CD107a activity in the presence of antibodies that were previously reported to block NKG2D, NKG2A, and 2B4. Blocking NKG2D led to a consistent decrease in the frequency of NK cells producing IFNγ and CD107a, but the effects of NKG2A and 2B4 blocking were less dramatic, in some cases enhancing activity (Figure S6 in Supplementary Material). This implies that NKG2D is involved in the recognition of autologous LCL, but cannot explain the entire effect. As NKG2A is an inhibitory receptor, it is likely that it may not be directly involved in killing but instead may mark a responsive subset. Thus, to confirm the role of NKG2A-expressing NK cells in the response to autologous EBV⁺ LCL, we compared the ability of NKG2A⁺ and NKG2A⁻ NK cells from the same donor for their ability to kill autologous EBV⁺ LCL. In all six donors, NKG2A⁺ NK cells display higher killing than did NKG2A⁻ NK cells (Figure 6; Figure S7 in Supplementary Material, p = 0.03), confirming that NK cell subsets that are functionally responsive to EBV⁺ LCL are enriched for NKG2A expression.