Human Natural Killer Cells Downregulate Zap70 and Syk in Response to Prolonged Activation or DNA Damage

Abstract

The extent of NK cell activity during the innate immune response affects downstream immune functions and, ultimately, the outcome of infectious or malignant disease. However, the mechanisms that terminate human NK cell responses have yet to be defined. When activation receptors expressed on NK cell surfaces bind to ligands on diseased cells, they initiate a signal that is propagated by a number of intracellular kinases, including Zap70 and Syk, leading eventually to NK cell activation. We assayed Zap70 and Syk content in NK cells from healthy human donors, and identified a subset of NK cells with unusually low levels of these two kinases. We found that this Zap70^lowSyk^low subset consisted of NK cells expressing a range of surface markers, including both CD56^hi and CD56^low NK cells. Upon in vitro stimulation with target cells, Zap70^lowSyk^low NK cells failed to produce IFNγ, and lysed target cells at one-third the capacity of Zap70^hiSyk^hi NK cells. We determined two independent in vitro conditions that induce the Zap70^lowSyk^low phenotype in NK cells: continuous stimulation with activation beads, and DNA damage. The expression of inhibitory receptors, including NKG2A and inhibitory killer immunoglobulin-like receptors (KIR), was negatively correlated with the Zap70^lowSyk^low phenotype. Moreover, expression of multiple KIR reduced the likelihood of Zap70 downregulation during continuous activation, regardless of whether NK cells had been educated through KIR-HLA interactions in vivo. Our findings show that human NK cells are able to terminate their functional activity without the aid of other immune cells through the downregulation of activation kinases.

INTRODUCTION

Natural Killer (NK) cells are lymphocytes of innate immunity that respond rapidly to infection. In this capacity, NK cells influence the outcome of acute infections, for example, infection with influenza virus (1). In chronic infections, for example infection with hepatitis C virus (HCV) (2, 3) or human immunodeficiency virus (HIV) (4), long-term disease progression is influenced by the NK cell response that occurred during the acute phase of infection.

In the early stages of the immune response, NK cells respond to virus-infected cells with proliferation, cytokine production and the killing of infected cells (3, 5, 6). Studies using mouse models of disease show how NK cells can steer the overall immune response to a pathogen by limiting antigen availability (7) and selectively killing subsets of dendritic cells (8) and T cells (9). The extent and duration of the NK response can also affect long-term immunity, such as T cell memory (10). Following their initial proliferation and population expansion, NK cells decline in number and their functional capacity is diminished (3, 5, 6).

This contraction and cessation of NK cell mediated immune responses is influenced by the cytokine environment (11). However, the contributions of other factors are unclear. For instance, a variety of inhibitory receptors help prevent NK cells from attacking healthy cells (12), but it is not known if these inhibitory signals also participate in attenuating the effector cytokine and cytolytic responses of human NK cell populations, once they are no longer needed. In contrast, the contraction and memory phases of B and T cell responses have been studied extensively (13), as has the contraction of mouse NK cell populations in response to murine cytomegalovirus (MCMV) (14, 15).

NK cells detect damaged, infected and malignant cells with an array of activating cell surface receptors (16–18). Upon engaging ligands expressed by target cells, NK cell surface receptors associate with transmembrane adaptor molecules, such as DAP-12 and CD3ζ, which in turn associate with Src kinases (19). These kinases initiate an intracellular phosphorylation cascade that leads to NK cell activation. Among the intracellular enzymes that propagate the cascade are other kinases, Zap70 and Syk (20, 21), as well as phosopholipase C-γ2 (PLC-γ2) (22).

Strongly influencing the development and response of human NK cells are inhibitory receptors that recognize HLA class I molecules. These comprise the conserved NKG2A receptor, which has a lectin-like binding site that recognizes conserved HLA-E (23), and diverse killer cell immunoglobulin-like receptors (KIR) that recognize polymorphic epitopes of HLA-A, -B and -C (24). During the development of an NK cell, interaction between self HLA class I and an inhibitory receptor educates the NK cell to be tolerant of healthy cells expressing a normal amount of HLA class I, but intolerant of unhealthy cells expressing abnormally low amounts of HLA class I (25).

When bound to HLA class I ligands on healthy cells, NKG2A and inhibitory KIR recruit tyrosine phosphatases, which inhibit signals coming from activating NK cell receptors (26), likely by direct de-phosphorylation of the intracellular proteins involved in the activating pathway (27). The reduced amount of HLA class I on unhealthy cells cannot block the activating signals, which instruct the NK cell to attack and kill the unhealthy target cell to which it is bound.

The signaling cascade of NK cell activation and its various components, are well studied for the initiation of an NK cell response (19). But the role of these mechanisms in winding down the response of an NK cell population, once it is no longer needed, is unknown. Likewise, although the contribution of inhibitory HLA class I receptors to NK cell activation is defined (28), their role, if any, in terminating NK cell responses is also unknown. These relationships were explored by the investigation we report here. In a study of human NK cells obtained from healthy, CMV-free donors, we identified a subset of NK cells that express unusually low amounts of the Zap70 and Syk kinases, which are key components in the signaling cascade of T cells and B cells (29, 30). We replicate this phenotype in vitro, and make connections between the emergence of these Zap70^lowSyk^low NK cells and the chronic stimulation of activating NK cell receptors.

MATERIALS AND METHODS

HLA and KIR genotyping

Whole human blood was purchased from the Stanford Blood Center (Stanford, CA). DNA was extracted from whole blood using the QIAamp DNA Blood Mini Kit (Qiagen, Valencia, CA) following the manufacturer’s instructions. Isolated DNA was HLA and KIR genotyped as described previously (31). In summary, genomic DNA was sheared and those fragments representing complete HLA-A, HLA-B, HLA-C and KIR genes were isolated using oligonucleotide probes (24). The DNA fragments were subjected to sequencing using an Illumina MiSeq machine, and v3 technology, with 300bp paired-end reads (Illumina Inc. San Diego, CA). KIR gene copy number and the identity of KIR and HLA-A, -B and -C alleles were defined as described (31).

Blood Acquisition and Processing

Leukoreduction and separation (LRS) chambers containing 10⁹ peripheral blood mononuclear cells (PBMCs) were obtained from healthy, cytomegalovirus-negative donors at the Stanford Blood Center (Stanford, CA). PBMCs were separated from other cells on a Ficoll-Paque gradient (GE Healthcare, Chicago, IL), pelleted and then suspended at 10⁷/ml in fetal bovine serum (FBS) (Gemini Bio-products, Sacramento, CA) containing 10% DMSO (EMD Millipore, Billerica, MA). Aliquots were frozen using a Mr. Frosty device (Thermo Fisher Scientific, Waltham, MA), then kept in a −80°C freezer for >24 hours, before being stored in liquid nitrogen.

For use in experiments, frozen aliquots of PBMC were thawed at 37°C in a water bath and suspended in 10ml of RPMI-1640 medium (Corning, Manassas, VA) containing 2mM L-glutamine (Thermo Fisher Scientific), 100U/ml of penicillin and streptomycin (Thermo Fisher Scientific), and 10% FBS (RPMI10%-C). DNAse I was added to a final concentration of 0.1mg/ml to prevent cells clumping (Sigma Aldrich, St. Louis, MO), and the cells were then incubated for 30 minutes at 37°C. To ensure that cell surface markers were expressed normally, the cells were washed and transferred to 12 well plates, at 1.0 × 10⁷cells/well, and kept in a 37°C incubator with 5% CO₂ for ~12 hours before any further manipulation was performed.

NK Isolation

NK cells were isolated from PBMCs using the Untouched NK Isolation Kit with LS columns as described by the manufacturer (Miltenyi, San Diego, CA). With this procedure other types of PBMC are depleted using specific antibodies. In summary, PBMCs were first resuspended in PBS with 0.5% bovine serum albumin and 2mM EDTA (MACS buffer). The Miltenyi cocktail of biotinylated antibodies was then added, allowing the antibodies to bind to their target antigens on PBMCs. On addition of paramagnetic streptavidin-coated beads, streptavidin on the beads bound to the biotin conjugated to the antibodies. The cell and bead mixture was then passed through a column in the presence of a magnet, which trapped all PBMCs, except NK cells, in the column. The flow-through, containing NK cells, was then centrifuged to pellet the cells, which were then washed and suspended in 1ml of RPMI10%-C.

We adapted the manufacturer’s protocol to isolate KIR⁻ NK cells. To do this, the mixture of MACS buffer and cocktail of biotinylated antibodies applied to PBMCs was supplemented with four additional biotinylated antibodies: anti-KIR2D (NKVFS1: Miltenyi); anti-KIR3DL1 (DX9: Biolegend; San Diego, CA); anti-KIR3DL (5.133: Miltenyi) and anti-KIR3DL2 (clone 539304: R&D Systems/Thermo Fisher Scientific). As this antibody mixture depleted all PBMCs except KIR⁻ NK cells (with the possible exception of those NK cells expressing only KIR2DL4 and, or, KIR2DL5 and no other KIR) it resulted in the purification of KIR3DL1⁻ and KIR3DL2⁻ NK cells. In making this modification, we found it was critical to keep the 10⁷ PBMC/50μl cell-per-volume ratio of the staining reaction prescribed in the Miltenyi protocol. Whereas this volume consists of 20% Miltenyi antibody cocktail and 80% MACS buffer in the Miltenyi protocol, in our modified protocol it consists of 20% Miltenyi antibody cocktail, 5% each of the four additional antibodies, and 60% MACS buffer. All other aspects of the protocol were unchanged.

NK Cell Culture and Functional Assays

In flat-bottom 96 well plates (BD Labware, Franklin Lakes, NJ) NK cells were activated using NK activation beads (Miltenyi). These beads are coated with streptavidin bound to biotinylated anti-CD2 and anti-NKP46 antibodies (Miltenyi). Either 100,000 or 200,000 NK cells were placed in each well with an equal number of NK activation beads and 200μl of RPMI10%-C medium containing 500U/ml of recombinant human IL-2. The IL-2 was obtained from Dr. Maurice Gately, (Hoffmann - La Roche Inc.) through the NIH AIDS Reagent Program, Division of AIDS, NIAID, NIH. For assays lasting >24 hours, 100μl of the culture medium were gently removed after 24 hours and replaced with 100μl of fresh RPMI10%-C medium containing 1000U/ml IL-2. In some experiments, cell division was tracked using 5-bromo-2′deoxyuridine (BrdU, Sigma Aldrich), which was included in the culture medium at a concentration of 10μM. BrdU was kept at 10μM in cultures over several days by replacing medium daily, as described above.

In experiments to study DNA damage and apoptosis in NK cells, VP-16 etoposide (Sigma Aldrich) and/or Z-VAD FMK (BD Biosciences) were included in the culture medium at concentrations of 0.25mg/ml and 15μM, respectively.

In assays measuring NK cytolysis, NK cells were first cultured overnight with 500U/ml IL-2 in the absence of target cells. K562 target cells were then added to NK cells at a 10:1 NK:target ratio. The mixture of NK cells and target cells was then transferred to a U-bottom 96 well plate, and pelleted at ~700 × g. To detect the increased expression of CD107a on the surface of NK cells that are killing targets, the cells were suspended in 200μl of fresh RPMI10%-C media containing 500U/ml IL-2 and 2μg/ml anti-CD107a antibody (Ebioscience, San Diego, CA). To facilitate contact between NK cells and K562 targets the co-cultures were centrifuged at 100 × g for 2 minutes. Co-cultures were then placed in a 37°C and 5% CO₂ incubator for 3.5 hours. Cytolysis was measured as the percent of NK cells that upregulated CD107a during the period of incubation (32).

Assay of IFNγ production was performed in the same way as the cytolytic assay, except that NK cells were co-cultured with K562 target cells for 16 hours in flat-bottom 96-well plates. During the final 6 hours of this assay, Brefeldin-A (Thermo Fisher Scientific) was added to a final concentration of 10μg/ml, in order to prevent NK cell secretion of IFNγ. In assays to test the effect of cytokines on Zap70 loss, NK cells were incubated in RPMI10%-C with 200ng/ml IL-12 (Ebioscience), 100ng/ml IL-18 (Thermo Fisher Scientific), or 200ng/ml IFNγ (Ebioscience). To induce KIR expression, KIR⁻ NK cells were incubated in RPMI10%-C with 10ng/ml IL-15 (Biolegend).

Antibodies

Fourteen fluorescently conjugated monoclonal antibodies specific for cell-surface proteins were used in the isolation and analysis of NK cells. These comprise: anti-CD3 (HIT3a, Biolegend); anti-CD3 (SP34-2, BD Horizon, BD Biosciences, San Jose, CA); anti-CD4 (A161A1, Biolegend); anti-CD8 (HIT3a, Biolegend); anti-CD14 (M5E2, BD Horizon); anti-CD16 (MHCD1617, Life Technologies); anti-CD19 (HIB19, Ebioscience); anti-CD27, (M-T271, Biolegend); anti-CD56 (CMSSB, Ebioscience); anti-CD-107a (eBioH4A3, Ebioscience); anti-KIR2DL1/S1 (A66899, Beckman Coulter, Brea, CA); anti-KIR3DL1 (DX9, Biolegend); anti-KIR3DL2, (clone 539304, R&D Systems); anti-NKG2A, (REA110, Miltenyi). In addition, unconjugated anti-KIR3DL2 (DX31, a gift from Lewis Lanier) was identified by fluorescently conjugated anti-mouse IgG (A-11001, ThermoFisher Scientific). Unconjugated anti-KIR3DL2 (MBB2-2994, Novus Biological, Littleton, CO) was detected with fluorescently conjugated anti-rabbit IgG (Poly4046, Biolegend).

Six fluorescently conjugated monoclonal antibodies were used to detect the following intracellular markers: anti-Zap70 (1E7.2 (Biolegend); anti-Syk (4D10.2, Biolegend); anti-PLC-γ2 (K86-1161, BD Phosflow); anti-IFNγ (45.B3, Biolegend); anti-BrdU (IC7225A, R&D Systems); anti-BrdU (51-23619L, BD Pharmingen). In addition, biotinylated anti-Zap70 (SBZAP, Abcam) was detected with fluorescently conjugated streptavidin (12-4317, Ebioscience).

Cell Staining, Fixation, and Flow Cytometry

Staining of PBMC, with antibodies specific for cell-surface proteins, was done in U-bottom, 96-well plates. Staining took place in a total volume of 50μl of PBS containing 5% FBS. The cells and fluorescently conjugated antibodies were incubated at 4°C in the dark for various times from 0.5–12 hours. The antibody panels comprised a maximum of 9 of the 16 antibodies described above. The amount of each antibody used for staining was determined empirically by titration. Typically, 2μl of the solution provided by the manufacturer were sufficient for one test. During subsequent steps, the reagents and plates were kept on ice. The stained cells in each well were washed 2 times with 200μl PBS. After each wash the cells were centrifuged at ~700 × g for 3 minutes at 4°C. In panels including a secondary anti-mouse IgG antibody, cells were stained in a three-step process to avoid off-target binding. Cells were first stained with the unconjugated primary antibody alone. After two washes, cells were then stained with the secondary antibody alone. Following three more washes, cells were stained with a cocktail of the other antibodies specific for cell-surface markers.

To identify and exclude dead cells from analysis, NK cells were stained with Live/Dead Yellow (Thermo Fisher Scientific) according to the manufacturer’s instructions. For each well, the cells were stained for 30 minutes in 50μl of a 500-fold dilution of Live/Dead Yellow in PBS. The cells were washed twice and then fixed with 100μl BD cytofix solution (BD Biosciences) for 10 minutes at 4°C. 100μl of perm buffer (1 × BD perm, BD Biosciences) were then added to the cells, while still in BD cytofix solution, in order to permeabilize them. The cells were then washed twice with 200μl of perm buffer. To ensure permeabilization of intracellular compartments, cells were then incubated with 200μl of perm buffer containing 10% DMSO, at 4°C for 20 minutes. Cells were then washed three times with perm buffer. To ensure fixation of intracellular compartments, cells were subjected to a second fixation with BD cytofix, followed by washing with permeabilization (perm) buffer, as cells were incubated with perm buffer containing 30μg/ml Dnase I (Sigma Aldrich) for 30 minutes at room temperature in the dark. For intracellular antibody staining, cells were stained in perm buffer in a 50μl volume for 30 minutes at 4°C, followed by two washes with perm buffer, and a final wash in PBS containing 5% FBS.

Intracellular staining of permeabilized cells was performed using the seven monoclonal antibodies specific for intracellular markers described above. Staining concentrations were determined empirically by titration, but typically 2μl/test was sufficient. Cells were analyzed using a LSR II flow cytometer (BD) at the Stanford Shared FACS Facility (Stanford, CA). The data obtained were analyzed using Flowjo Ver. 9.7.6 (Flowjo, LLC, Ashland, OR), and assessed for statistical significance using Prism Ver. 6 (Graphpad Software, La Jolla, CA).

Quantitative PCR analysis of NK cell mRNA

NK cells isolated from PBMCs were subjected to Live/Dead staining and then sorted on a BD Aria II instrument at the Stanford shared FACS Facility (Stanford, CA). NK cells from groups of 2-4 donors were pooled and sorted based on antibodies against KIR3DL1, pan-KIR2D, clone 539304, and/or NKG2A, into separate tubes containing RPMI 10%-C. The cells were centrifuged, the supernatant removed, and the cell pellets processed using the RNeasy Protect Kit (Qiagen), following manufacturer’s instructions. RNA was eluted in 30μl RNAse, DNAse-Free Ultrapure Distilled water (Invitrogen). 3μl RiboLock RNAse blocker (Thermo Fisher Scientific), 3μl 50mM MgCl₂ solution (Invitrogen), and 3 Units DNAse I (Thermo Fisher Scientific) were added to each sample. Incubation of these samples in a 37°C water bath for 30 minutes, digested any residual genomic DNA. To denature the DNAase, 3μl of 0.05M EDTA (Hoefer) were added to the samples, which were then incubated for 10 minutes in a 65°C water bath. RNA integrity was assessed using the RNA 6000 Pico Kit and an Agilent Bioanalyzer 2100 (Agilent Technologies, Santa Clara, CA) at the Stanford Protein and Nucleic Acid (PAN) Facility (Stanford, CA). cDNA was synthesized from mRNA using SuperScript VILO Mastermix (Thermo Fisher Scientific). cDNA was subjected to unbiased preamplification using PreAmp (Thermo Fisher Scientific), and then examined by qPCR using Taqman gene expression assays for Syk, Zap70, and Beta-Actin (Hs00895377_m1, Hs00896345_m1, and Hs99999903_m1 respectively, Thermo Fisher Scientific) using a StepOnePlus Real-Time PCR instrument (Applied Biosystems, Foster City, CA) at the Stanford PAN Facility. For each preparation of cDNA three replicate analyses were performed.

RESULTS

A subset of peripheral blood NK cells has low Zap70, low Syk, and high PLC-γ2

NK cells from different human donors exhibit considerable functional variation, which is not all due to HLA and KIR differences (33). One possible source of variation is the intracellular kinases that contribute to NK cell activation. To address this question we used flow cytometry to quantify Zap70 in the PBMC of healthy donors. The CD14⁻CD19⁻CD3⁻CD56^mid-hiCD16^mid-hi cells defined by our NK cell gate, exhibit bimodal expression of Zap70 (Fig. 1A). Averaging for 27 donors showed that 7.9±4.9% of NK cells had <10% of the mean Zap70 present in the other ~92% of NK cells (Fig. 1A and 1B). Less than 2% of NK cells expressed Zap70 at levels between those of the Zap70^low and Zap70^hi NK cell populations (Figure 1D). The fluorescent signal in the Zap70 channel observed for Zap70^low NK cells was higher than that with the isotype control, indicating that Zap70^low NK cells express some Zap70 (Fig. 1A and Supp. Fig. 1A).

Figure 1. A subset of peripheral blood NK cells has low Zap70, low Syk, and high PLC-γ2.

(A) Gating strategy to examine Zap70^low NK cells in human blood. From left to right: (top row) FSC-A and SSC-A define PBMCs, FSC-H and FSC-A define singlets, LIVE/DEAD stain and CD14^low define leukocytes, CD3^low and CD19^low exclude B and T cells, (continued on bottom row) CD16^mid-high and CD56^mid-high broadly define NK cells, KIR2DL1/S1 defines a subset of KIR⁺ NK cells, Zap70^low cells make up a small portion of KIR2DL1/S1⁺ NK cells. Plots represent the data obtained from 36 donors. (B) Percentage of Zap70^low NK cells from 27 donors, gated either on total NK cells, or NK cells positive for any KIR (pan-2D, 3DL1, or 3DL2). Results of paired T-test shown. (C) Percentage of Zap 70 deficient cells in NK cell subsets from 27 donors defined by various surface markers. Shown are the results of Tukey’s comparison of 1-way ANOVA. (D) (Left) Representative NK cell plot of Zap70 versus Syk. (Right) Comparison of Zap70 and Syk phenotypes in NK cell populations obtained from 21 donors. Results of paired t-tests are shown. (E) (Left) Representative NK cell plot of PLC-γ2 versus Syk. (Right) Comparison of Zap70 and PLC-γ2 expression in the NK cell populations of 32 donors. Results of paired t-tests are shown. All panels are representative of at least 5 independent experiments. SEM are shown in all panels. **** = p< 0.0001, *** = p< 0.001, ** = p<0.001, * = p<0.01, in all panels.

Because we defined NK cells by negative gating, it was possible that Zap70^low cells were not NK cells. As NK cells are distinguished by expression of killer immunoglobulin-like receptors (KIR) (34), we examined Zap70 expression by KIR⁺ NK cells. The proportion of KIR⁺ NK cells that were Zap70^low (4.8±3.4%) was comparable to the 7.9% Zap70^low cells present in total NK cells (Fig. 1B), consistent with most Zap70^low cells being NK cells.

A small subset of peripheral blood T cells with effector-memory phenotype expresses KIR (35). These cells might also exhibit low CD3 expression as a consequence of activation. To see if these cells corresponded to the Zap70^low NK cells, PBMCs were stained with T cell-specific antibodies. Those antibodies were then bound to paramagnetic beads, and T cells were removed (>98%) from other PBMCs by Magnetic Assisted Cell Sorting (MACS). The proportion of Zap70^low cells within the KIR⁺ NK cell gate was not reduced by T cell depletion (Supp. Fig. 1B), showing that KIR⁺Zap70^low NK cells were not KIR⁺ T cells.

Zap70 expression was independently assessed with two Zap70-specific monoclonal antibodies. Both antibodies identified a similar proportion of Zap70^low cells (7.1-7.6%) in isolated NK cell populations (Supp. Fig. 1C). For a subset of donors, PBMC samples were divided equally into one subsample that was analyzed immediately for Zap70^low NK cells, and another subsample that was frozen and thawed prior to Zap70 analysis. No significant difference was observed in the frequency of Zap70^low NK cells in these two preparations, although freeze-thawing lowered the variance between donors of Zap70^low NK cells (Supp. Fig. 1C). Zap70^low NK cells were present in all eight NK cell subsets defined by combinations of high and low expression of CD56, CD16, CD27, NKG2A, and LILRB1 (Fig. 1C, Supp. Fig. 1E). Of these markers, NKG2A expression was the best correlate of Zap70 expression. Among NKG2A^low NK cells, 8.2±5.4% were Zap70^low, whereas only 1.6±1.4% of NKG2A^hi NK cells, were Zap70^low .

NK cell expression of Syk and PLC-γ2, other enzymes that contribute to NK cell activation, were similarly examined. A strong correlation (p<0.0001) was observed between Syk and Zap70 expression. Thus, almost all Zap70^low NK cells were also Syk^low. Only small numbers of NK cells with the Syk^lowZap70^hi and Syk^hiZap70^low phenotypes were detected (Fig. 1D). In contrast, PLC-γ2 expression was not positively correlated with Zap70 expression (p<0.01) (Fig. 1D–1E).

Zap70^lowSyk^low NK cells have less functional potential than Zap70^hiSyk^hi NK cells

Subsets of NK cells from the same donor, and varying in expression of Zap70 and Syk, were compared for their capacity to respond to HLA class I deficient K562 cells with accumulation of CD107a, a surrogate measure of the degranulation that occurs during target cell killing (36). Zap70^hiSyk^hi NK cells made a strong response, accumulating CD107a in 31.2% of the cells (Fig. 2A). In contrast, Zap70^lowSyk^low NK cells from the same donors made a feeble response, in which only 2.4% of the cells accumulated CD107a. Thus, low levels of Zap70 and Syk correlated with severe functional impairment of NK cells. NK cells expressing only one of the two kinases had a functional capacity that was between that of the Zap70^lowSyk^low and the Zap70^hiSyk^hi NK cells. On average, 14.7% of Zap70^lowSyk^hi and 14.4% Zap70^hiSyk^low NK cells responded to K562 targets. In summary, high expression of both kinases correlated with strong cytotoxic potential, low expression of both kinases correlated with weak cytotoxic potential and high expression of one kinase correlated with intermediate cytotoxic potential.

Figure 2. Zap70^lowSyk^low NK cells have less functional potential than Zap70^hiSyk^hi NK cells.

(A) NK cells isolated from 11 donors were primed for 24 hours with IL-2, then co-cultured with K562 target cells, at an E:T ratio of 10:1, for 3.5 hours in the presence of anti-CD107a antibody and IL-2. (B) NK cells from the same 11 donors were co-cultured with K562 target cells at 10:1 E:T ratio for 8 hours in the presence of IL-2. For the last 6 hours, Brefeldin A was included in the culture to block secretion. Accumulation of IFNγ was then measured by flow cytometry (C) Isolated NK cells from 26 donors were primed for 24 hours with IL-2 and IL-12. NK cells were then co-cultured with K562 target cells at 10:1 E:T ratio for 16 hours. For the last 6 hours, Brefeldin A was included in the culture to block secretion. Accumulation of IFNγ was then measured by flow cytometry. Shown for each panel are the results of Tukey’s comparison from a paired ANOVA. All panels represent at least 3 independent experiments, each with different donors. SEM are shown in all panels. **** = p< 0.0001, ***=p<0.001, *=p<0.05 in all panels.

To compare the cytokine response of Zap70^hiSyk^hi and Zap70^lowSyk^low NK cells, unprimed NK cells were co-cultured with K562 targets for 8 hours in the presence of IL-2. Cellular accumulation of IFNγ was then measured. On average, 17% of Zap70^hiSyk^hi NK cells produced IFNγ compared to 9.7% of Zap70^lowSyk^low NK cells (Fig. 2B). In the assay of cytokine production, the difference between the Zap70^hiSyk^hi and Zap70^lowSyk^low NK cells was less than a factor of 2. In contrast, the difference between the cytotoxic responses was more than ten-fold. NK cells expressing only one of the two kinases had weaker responses than the Zap70^lowSyk^low NK cells: 4.8% of Zap70^lowSyk^hi NK cells and 2.1% of Zap70^hiSyk^low NK cells responded to K562 with the production of IFNγ (Fig. 2B–2C).

To determine if Zap70^lowSyk^low NK cells make IFNγ in response to inflammatory conditions, NK cells were cultured with IL-2 and IL-12 for 24 hours. Thus primed, the NK cells were then co-cultured with K562 targets for 16 hours. For flow cytometry, the gate was set for IFNγ⁺ NK cells based on control NK cells that were primed with cytokines but not exposed to K562 cells. Under these conditions, only 0.1% of the Zap70^lowSyk^low NK cells produced IFNγ. In contrast, an average of 23% of Zap70^hiSyk^hi NK cells made IFNγ in response to K562 cells (Fig. 2C). In this assay, the difference between the response of Zap70^lowSyk^low and Zap70^hiSyk^hi NK cells was more than 200 fold.

Zap70 and Syk loss is induced by extended NK cell activation

To see if NK cells downregulate Zap70 and Syk as a consequence of making an immune response, we co-cultured NK cells and K562 targets for 3.5 hours, a period sufficient for NK cell activation and target cell cytolysis. Ratios of 2, 5, 10 and 20 NK cells to one target cell were used. No significant differences in the frequency of Zap70^lowSyk^low NK cells were observed between cultures with different effector to target ratios, or with cultures of NK cells alone (Fig. 3A). Thus the loss of Zap70 and Syk was not induced by a short-term culture of NK cell activation and target cell cytolysis.

Figure 3. Zap70 and Syk loss is induced by extended NK cell activation.

(A) The frequency of Zap70^low/Syk^low NK cells among isolated NK cells that were co-cultured for 24 hours with K562 target cells at various NK: K562 ratios. Shown are the results from 6 donors. No significant differences were found as assessed by Tukey post-tests. Representative of 3 independent experiments. (B) Frequency of Zap70^low or Syk^low NK cells from 9 donors after isolated NK cells were co-cultured for 3 days with activation beads coated with either anti-CD2 and anti-NKp46 antibodies, or anti-DNAM1 and anti-NKG2D antibodies. Co-cultures were performed at a 2:1 bead:NK ratio in RPMI-10%C media with 500U/ml IL-2. Representative data are shown from at least 3 independent experiments. (C) Kinetics of Zap70 loss following incubation with anti-NKp46, anti-CD2 activation beads, as in B. NK cells and activation beads were co-cultured at a ratio of 2:1. Shown are the data combined from 4 experiments involving the NK cells of 24 donors. (D) Frequency of Zap70^low/Syk^low NK cells from 8 donors, following co-culture with, or without, K562 target cells at various T:E ratios. Co-cultures occurred for 1 or 3 days. Additional K562 cells were added on day 1 and 2 for the 3-day, 1:5 ratio group. Representative of 3 independent experiments. (E) Comparison of Zap70 loss following 24 hour incubation of isolated NK cells in RPMI-10%C media supplemented with various cytokine combinations. Shown are the combined data from 3 experiments, in which the NK cells of 16 donors were studied. (F) Comparison of the percentage of divided cells (BrdU⁺) in Zap70^low vs. Zap70^hi NK cells on day 3 of bead stimulation, as in C, for 6 donors. Representative data from at least 3 independent experiments. The result of a paired t-test is shown. (G) Correlation of the geometric mean of Zap 70 signal in KIR⁺ NK cells of 6 donors, as in C, versus the geometric mean of the BrdU signal in KIR⁺ NK cells that have divided, as in C. Representative of 3 independent experiments each with different donors. Statistical comparisons shown in all panels except F are the result of Tukey post-tests from paired 1-way or 2-way ANOVA, as appropriate. SEM are shown in all panels. ***=p<0.001, **=p<0.01 in all panels.

To assess the effect of long-term, chronic stimulation of NK cells, one set of beads was coated with anti-CD2 and anti-NKp46 antibodies and a second set of beads was coated with anti-DNAM1 and anti-NKG2D antibodies. Each bead has the potential to engage two activating NK cell receptors and thereby activate NK cells (37). Flow cytometry showed that isolated NK cell preparations in this experiment had an average of 3.0±4.0% Syk^low NK cells and 1.3±0.9% Zap70^low NK cells prior to any manipulation (Fig. 3B). NK cells were cultured with each bead set separately, for three days in the presence of IL-2. Of the NK cells stimulated through their CD2 and NKp46 receptors 49.3±15.6% were Zap70^low, and 45.8±15.7% were Syk^low. Of the NK cells stimulated through their DNAM1 and NKG2D receptors, 41.7±14.7% were Zap70^low, and 34.2±14.5% were Syk^low. Following culture with either set of activation beads, the frequencies of Zap70^low and Syk^low NK cells were significantly greater (p< 0.0001) than the frequencies observed before culture (Fig. 3B). Statistical comparison also showed that CD2/NKp46 stimulation resulted in significantly more Zap70 and Syk loss than DNAM1/NKG2D stimulation (Fig. 3B).

To compare the kinetics of Zap70 and Syk downregulation, NK cells were co-cultured with the two activation beads for different times (0, 1, 3, 5, and 7 days) and analyzed for Zap70 and Syk expression. By including bromodeoxyuridine (BrdU) in the culture medium, we could determine the extent of NK cell division that took place. In order to distinguish NK cells from beads, our analysis focused on KIR⁺ NK cells. During one day of co-culture the frequency of Syk^lowKIR⁺ NK cells increased from 3.5±3.4% to 26.6±4.7% (Fig. 3C). A similar downregulation of Zap70 required three days of co-culture, at which point 26.6±14.1% of KIR⁺ NK cells were Zap70^low. That Syk loss is faster than Zap70 loss raises the possibility that, in each cell, Syk is lost first and Zap70 second. After five days co-culture, the frequency of Syk^low NK cells had further increased to 35.4±7.4%, whereas the percentage of Zap70^low NK cells was 24.1±10.8%, not significantly different from that on day three (Fig. 3C).

To complement the above experiment, in which NK cells were stimulated with bead-bound antibodies, we stimulated NK cells with live K562 cells and assessed the loss of Syk and Zap70. NK cells were co-cultured with K562 target cells at a 5:1 or 1:1 ratio. During one day of co-culture, most target cells were killed, but no loss of Syk or Zap70 was observed in NK cells. To increase the stimulation, NK and K562 cells were co-cultured for three days. To maintain stimulation, fresh K562 target cells were added each day. Under these conditions, there was significant loss of Syk and Zap70. With an effector to target ratio of 5:1, the frequency of Zap70^low Syk^low NK cells increased from 2% to 13.8% (Fig. 3D). With an effector to target ratio of 1:1, the frequency of Zap70^low Syk^low NK cells further increased to 41.8±30.6%. In summary, chronic stimulation of NK cells, by human cells lacking HLA class I or by antibodies bound to activating NK cell receptors, induces significant loss of Syk and Zap70 in the NK cell population.

Because three days stimulation with bead-bound anti-receptor antibodies was necessary to achieve Zap70 loss, we explored the possibility that inflammatory cytokines produced by NK cells accumulate over several days in media and cause Zap70 and Syk loss. In examining the effect of such cytokines, IL-18 and IFNγ had no effect, and IL-12 slightly increased the frequency of Zap70^low NK cells from 4.4 to 6.4% (Fig. 3E). These results indicate that inflammatory cytokines are not a significant cause of NK cell loss of Zap70.

We co-cultured NK cells and beads coated with only one ligand, either anti-CD2 or anti-NKp46. This activation by single receptors did not increase the proportion of Zap70^lowSyk^low cells. Instead, it resulted in fewer Zap70^lowSyk^low cells; 10.2% from CD2-mediated activation and 20.7% from NKp46-mediated activation, compared to 23.7% from activation by both receptors (Supp. Fig. 2A). This result, together with the results shown in Figure 3B, support a model in which stronger forms of chronic activation produce higher frequencies of NK cells that are deficient for Zap70 and Syk.

To see if activated NK cells could induce Zap70 loss in unstimulated NK cells, we measured the frequency of Zap70^low NK cells after three days of stimulation by beads coated with anti-CD2 and anti-NKp46, among NK cells exclusively expressing the NKp44, NKp80 or NKp46 activation receptors. NKp46⁺ NK cells made up 23.93% of Zap70^low NK cells, compared to a combined average of 16.95% Zap70^low for NKp44⁺ and NKp80⁺ NK cells (Supp. Fig. 2B). These results point to Zap70 loss being a direct consequence of ligand engagement by the activation receptors, and not a bystander effect.

We assessed if Zap70 downregulation induced by the stimulating beads correlates with NK cell division and proliferation. An analysis of BrdU and Zap70 expression showed that the Zap70^low phenotype primarily occurred for NK cells that had divided during the period of stimulation (Fig. 3F). To see if the extent of Zap70 downregulation correlated with the number of NK cell divisions, we compared the geometric mean fluorescence intensity (gMFI) of BrdU in NK cells that had divided, to the Zap70 gMFI in all NK cells for each donor. Higher BrdU gMFI was correlated with lower Zap70 gMFI across donors (Fig. 3G), showing that Zap70 loss correlates with increasing proliferation.

The results of these various analyses point to chronic engagement of activation receptors as one cause of the Zap70^lowSyk^low phenotype in NK cells. Additionally, our results imply that the Syk^low phenotype occurs prior to the Zap70^low phenotype, and that both phenotypes can be passed from mother to daughter cells.

Zap70 and Syk loss is a consequence of DNA damage, but not of caspase activation

To see how DNA damage affects Zap70 and Syk expression, NK cells were treated with etoposide (VP-16). This drug causes DNA breaks, leading to caspase-mediated apoptosis (38). Treating NK cells with etoposide for 24 hours activated caspase 3, a component of apoptotic signaling, in 22.6% of the cells (histogram in Fig. 4A, averages in Fig. 4B). Etoposide also induced loss of Zap70 (Fig. 4A) and Syk (Fig. 4B); their expression was lost by 32.3% and 34.2% of the NK cells, respectively (Fig. 4B). That the frequency of Zap70^low and Syk^low NK cells was greater than that of cells activating caspase 3, suggested that the loss of Zap70 and Syk was not caused by apoptosis-induced proteolysis.

Figure 4. Zap70 and Syk loss is a consequence of DNA damage, but not of caspase activation.

(A) Representative flow cytometry plots showing the proportions of isolated NK cells positive for anti-cleaved Caspase 3 antibody (top) or NK cells that are Zap70^low (bottom). NK cells from a single donor, either treated with DMSO (left column, control) or with 0.25mg/mL VP-16 (right column) in DMSO, for 24 hours. Data are representative of 17 donors. (B) Percent of either Syk^low, Zap70^low or cleaved Caspase 3⁺ in isolated NK cells from 9 donors following 24 hour treatment with 0.25mg/mL VP-16 etoposide, or DMSO control. (C) Percent of cleaved Caspase 3⁺ NK cells from 6 donors following 24 hour treatment with: DMSO control, 15uM Z-VAD-FMK caspase inhibitor alone, 0.25mg/mL VP-16 etoposide, or Z-VAD-FMK inhibitor and VP16 together. All cultures contained 500U/ml IL-2. Shown is the result of Tukey’s post test of paired ANOVA. (D) Percent of Zap70^low NK cells following 24 hour treatment, as in C, for seven donors. All panels show data that are representative of at least 3 independent experiments using NK cells from different donors. Statistical comparisons shown in all panels are the result of Tukey post-tests from paired 1-way or 2-way ANOVA, as appropriate. SEM are shown in all panels. **** = p< 0.0001, ***=p<0.001, throughout *=p<0.05, NS = not significant in all panels.

As another test for involvement of apoptosis-induced proteolysis, NK cells were treated with Z-VAD FMK, a universal caspase inhibitor (39). In the context of etoposide treatment, Z-VAD FMK effectively blocked caspase cleavage (Fig. 4C). However, treating NK cells with Z-VAD FMK alone had no effect on the frequency of Zap70^low NK cells (Fig. 4D). Treating NK cells with etoposide and Z-VAD FMK, caused a more variable Zap70 loss than that caused by etoposide alone. But the frequency with which NK cells lost Zap70 was not significantly different for treatment with etoposide alone or combined with Z-VAD FMK (Fig 4D). Like Zap70, the frequency of Syk loss caused by etoposide was unaffected by caspase blockade (Supp. Fig. 2C). These results suggest that Zap70 and Syk loss is caused by the DNA damage response, but is independent of caspase-mediated apoptosis.

Inhibitory KIR and NKG2A protect NK cells from losing Zap70 and Syk

The frequency of Zap70^low cells is higher for KIR⁻ NK cells than KIR⁺ NK cells (Fig. 1B and 5A). Moreover, NK cells defined by the binding of anti-KIR3DL1 or anti-KIR2DL1/S1 have a higher frequency of Zap70^low NK cells than NK cells binding both antibodies. Thus, NK cells expressing KIR3DL1 and KIR2DL1/S1 are less likely to downregulate Zap70 than NK cells expressing just one of these receptors.

Figure 5. Inhibitory KIR and NKG2A protect NK cells from losing Zap70 and Syk.

(A) Comparison of the percentage of Zap70^low NK cells within various KIR⁺ subsets measured by flow cytometry following NK cell isolation in 11 donors. Representative of 3 independent experiments using NK cells from different donors. (B) Comparison of Zap70 expression in NK cells expressing KIR3DL1, KIR3DL2 or any KIR2D (KIR⁺), or expressing no inhibitory KIR (KIR⁻) in relation to Zap70 and NKG2A in isolated NK cells from 25 donors. Representative of 3 independent experiments using NK cells from different donors. (C) Relative transcript levels of Zap70 and Syk measured by qPCR, normalized to Beta-Actin transcript levels, derived from pools of sorted NK cells either NKG2A^hi/KIR⁺ or NKG2A^low/KIR⁻. Transcript amounts are represented as a percentage of the average normalized transcript amount present in the NKG2A^hi/KIR⁺ group. Representative data from 8 donors, acquired from 3 separate sorts and 3 technical replicates. Statistical comparisons shown in all panels are the result of Tukey post-tests from paired 1-way or 2-way ANOVA, as appropriate. SEM are shown in all panels. **** = p< 0.0001, ***=p<0.001, **=p<0.01 in all panels.

A similar hierarchy was observed for the pan-KIR2D antibody and KIR3DL1 antibodies. The frequency of Zap70^low NK cells was lowest in the subpopulation of NK cells binding both antibodies, highest in NK cells binding only anti-KIR3DL1 and intermediate in NK cells binding only anti-pan-KIR2D (Fig. 5A). These results initially suggested that educated NK cells were less likely to be Zap70^low than uneducated NK cells.

We similarly examined Zap70 and Syk loss in NK cells expressing NKG2A and/or KIR. Of NK cells lacking expression of inhibitory KIR and expressing low amounts of NKG2A, 36.6±20.8% were Zap70^lowSyk^low (Fig. 5B). The NKG2A^hiKIR⁻ subpopulation of NK cells contained 6.5±4.8% Zap70^lowSyk^low cells, a similar frequency to the 7.5±5.9% of Zap70^lowcells in the NKG2A^lowKIR⁺ subpopulation of NK cells (Fig. 5B). Thus, presence of either NKG2A or an inhibitory KIR appeared sufficient to retain Zap70 and Syk expression. Having the lowest frequency of Zap70^lowSyk^low cells, 3.5±3.1%, was the subpopulation of NK cells that co-express NKG2A with an inhibitory KIR (Fig. 5B).

One possible cause for the differences in kinase expression between Zap70^hiSyk^hi and Zap70^lowSyk^low NK cells is transcriptional regulation. We tested this hypothesis for NKG2A^low/KIR⁻ NK cells, because this subpopulation had the highest frequency of ex vivo Zap70^lowSyk^low cells (36.6±20.8%). NK cells were sorted into those co-expressing NKG2A and KIR, and those lacking KIR and having low level of NKG2A (Fig. 5B). These two pools were adjusted to contain the same number of NK cells, and RNA was extracted. RNA from several donors was combined and used to make cDNA, which was subjected to real-time quantitative PCR with primers amplifying Zap70, Syk, and Beta-Actin, a housekeeping gene for normalization.

The Zap70 transcript levels in NKG2A^lowKIR⁻ NK cells were 81% of those in NKG2A^hiKIR⁺ NK cells (Fig. 5C). Syk transcripts in NKG2A^lowKIR⁻ NK cells were 113% of those in NKG2A^hiKIR⁺ NK cells (Fig. 5C). If transcriptional regulation was the primary cause of Zap70 and Syk loss, we estimated that NKG2A^lowKIR⁻ NK cells should display ≤ 50% of the Zap70 and Syk transcripts of NKG2A^hiKIR⁺ NK cells. This conservative estimate was based on the ≥ 27-fold difference in the amounts of Zap70 and Syk proteins in Zap70^hiSyk^hi and Zap70^lowSyk^low cells (example histograms in Fig. 1A and 1D) and the 36.6% frequency of Zap70^lowSyk^low NK cells in the NKG2A^lowKIR⁻ NK cell population (Fig. 5B). Given that NKG2A^hiKIR⁺ NK cells have a lower level of Syk transcripts than NKG2A^lowKIR⁻ NK cells, and a level of Zap70 transcripts that is only 19% higher than that of the NKG2A^lowKIR⁻ NK cells, we conclude that the ex vivo differences in Zap70 and Syk protein levels between Zap70^hiSyk^hi and Zap70^lowSyk^low NK cells are not solely due to transcriptional regulation.

KIR expression prevents loss of Zap70 in a ligand independent manner

We studied education of NK cells expressing KIR3DL1, by the cognate Bw4 ligand, carried by subsets of HLA-A and –B allotypes (33). Analysis focused on the subpopulation of NK cells for which KIR3DL1 is the only inhibitory KIR. Between donors who have the Bw4 epitope and donors who lack the Bw4 epitope, we found no significant difference in the frequency of Zap70^low NK cells (Fig. 6A).

Figure 6. KIR expression prevents loss of Zap70 in a ligand independent manner.

(A) Comparison of the frequency of Zap70^low KIR3DL1⁺ single-positive NK cells from donors who either carry the Bw4 epitope (n=9), or who lack the Bw4 epitope (n=13). “Weak Bw4” denotes a donor with KIR3DL1*001, KIR3DL1*015, HLA-B*52:01 and HLA-A*24:02. “Strong Bw4” denotes a donor with two copies of KIR3DL1*001, HLA-B*51:01 and HLA-A*32:01. Results of student’s t-test are shown, and represent at least 3 replicate experiments. (B) Representative plots of NK cells prior to the MACS depletion of KIR⁺ NK cells using anti-KIR3DL1, anti-KIR3DL2, and anti-Pan-KIR2D antibodies (left), following the depletion of KIR⁺ NK cells (center), and following 24 hour treatment with IL-15 (right). Representative of the data from 18 donors and at least 3 replicate experiments. (C) Comparison of the frequency of Zap70^low cells from 8 donors within KIR3DL1 single-positive NKG2A⁻ NK cells and KIR⁻NKG2A⁻ NK cells, either prior to the MACS depletion of KIR⁺ NK cells (ex vivo), or following KIR induction by IL-15 on KIR⁻ NK cells. 2-way ANOVA of KIR3DL1 comparison = p<0.0001. Representative of 3 independent experiments, each with different donors. (D) Isolated NK cells from 4 donors were co-cultured with class I deficient 221 target cells or 221 cells expressing transfected HLA-B*58:01, in the presence of 500U/ml IL-2 at a 10:1 E:T ratio. Shown is the frequency of KIR3DL1⁺ NK cells that were CD107a⁺ (left) and the frequency that were Zap70^low (right) after 3.5 hours of co-culture. Representative of 3 independent experiments. (E) Kinetic of different KIR⁺ NK subsets during extended stimulation with activation beads, as in Figure 3C. Data were obtained from the NK cells of 17 donors. Representative data are from 3 combined experiments. Statistical comparisons in all panels except A are the result of Tukey’s post test from 2-way ANOVAs. SEM are shown in all panels. ***=p<0.001, NS = not significant in all panels.

Avidity of the interaction between KIR3DL1 and Bw4⁺ is modulated by the polymorphism of KIR3DL1, HLA-A and HLA-B and affects NK cell education (40). We assessed its effect on the population of Zap70^low NK cells, by determining high resolution HLA class I and KIR genotypes for our donor panel (Supp. Fig. 3–4), and comparing the frequency of Zap70^low NK cells in the two donors predicted to have strongest and weakest and interactions of Bw4 with KIR3DL1, as predicted from published data (40). The donor with weakest interactions had KIR3DL1*001, KIR3DL1*015, HLA-B*52:01 and HLA-A*24:02 and (Fig. 6A, weak Bw4), whereas the donor with strongest interactions had two copies of KIR3DL1*001, HLA-B*51:01 and HLA-A*32:01 and (Fig. 6A, strong Bw4). That the weakest interaction is associated with 1.6% Zap70^low NK cells and the strongest interaction with 1.7% Zap70^low NK cells, showed that strength of NK cell education does not contribute to the loss or retention of Zap70. This conclusion was consistent with our analysis of Bw4⁺ and Bw4⁻ donors as a whole, which did not show any significant difference in Zap70 levels due to the presence or absence of the Bw4 epitope. In contrast, the expression of KIR by NK cells does favor retention of Zap70, irrespective of its participation, or not, in NK cell education.

To test if KIR expression is sufficient to induce expression of Zap70, we isolated KIR-negative (KIR⁻) NK cells using a modified magnetic separation technique, and cultured them with IL-15 for 24 hours. This treatment induced KIR expression (41) (Fig. 6B). We then examined those NK cells with newly expressed KIR3DL1, but lacking other KIR and NKG2A. The proportion of Zap70^low cells in this subpopulation of NK cells was similar to that observed in the same subpopulation of ex vivo NK cells, but was significantly different from the proportion of Zap70^low NK cells lacking KIR3DL1, other KIR, and NKG2A following IL-15 treatment (Fig. 6C). Although it is possible that other factors, which influence Zap70 expression, were induced by IL-15, only the cells induced to express KIR showed increased Zap70 expression.

We examined if binding of KIR3DL1 to Bw4⁺HLA-B could increase NK cell expression of Zap70. Isolated NK cells were co-cultured for 3.5 hours with 221 target cells, which do not express class I HLA, or with transfected 221 cells expressing Bw4⁺HLA-B*58:01. Assaying for CD107a accumulation showed that presence of HLA-B*58:01 reduced the frequency of CD107a producing NK cells gated on KIR3DL1⁺ from 28.1% to 4.3% (Fig. 6D). This result shows that NK cell KIR3DL1 bound to HLA-B*58:01 on the target cells and generated inhibitory signals. Such signals were not produced in the NK cells cultured with 221 cells, which lack Bw4. In contrast, we observed no significant difference in the frequency of Zap70^low cells in KIR3DL1⁺ NK cells cultured with 221 cells or 221 cells expressing HLA-B*58:01 (Fig. 6D).

To test if KIR expression protected activated NK cells from loss of Zap70, we tracked Zap70 expression across a broad range of KIR⁺ NK cell subsets, during extended culture with activation beads. In these cultures, the relationships between specific KIR subsets and the extent to which they downregulated Zap70 did not fully replicate those seen for ex-vivo NK cells (compare Fig. 5A to Fig. 6E). However, co-expression of KIR3DL1 and a KIR2D by NK cells was significantly more protective against stimulation-induced Zap70 loss than expression of only KIR3DL1 or KIR2D (Fig. 6E). This protection occurred although NK cells co-expressing KIR3DL1 and KIR2D had undergone more division more than the cells expression KIR3DL1 or KIR2D (Supp. Fig. 2D). These results are consistent with KIR expression providing protection against Zap70 loss.

A KIR3DL2-specific antibody marks Zap70^lowSyk^low NK cells and T cells

Analysis of three anti-KIR3DL2 antibodies showed that mouse monoclonal antibody 539304 and a rabbit polyclonal antibody have an overlapping specificity, with 539304 binding to many fewer NK cells than the polyclonal antibody (Supp. Fig. 2E). The mouse monoclonal antibody DX31 binds to a different subpopulation of NK cells than the other two antibodies (Fig 7A).

Figure 7. The KIR3DL2 target of antibody 539304 marks Zap70^low NK cells and T cells.

(A) Frequency of NK cells, from 5 donors, that are recognized by the DX31 and 539304 monoclonal anti-KIR3DL2 antibodies, and a polyclonal anti-KIR3DL2 antibody. Representative data from at least 3 independent experiments, each with different groups of donors. (B) Percentage of Zap70^low NK cells in 5 donors identified by either DX31 or 539304, as in A. Representative of at least 3 independent experiments using different donors. (C) Comparison of the frequency of Zap70^low cells in the NK cell subsets identified by anti-KIR3DL1, anti-pan-KIR2D, or 539304 antibodies, for the NK cells of 34 donors. Representative of 3 independent experiments. (D) Frequency of NK cells binding the 539304 antibody, among various subpopulations of NK cells that express hi or low NKG2A. NK cells from 29 donors were studied. Representative of 3 independent experiments. (E) Binding of the 539304 antibody to the NKG2A⁺, pan-KIR2D⁺ and KIR3DL1⁺ NK cells obtained from the 29 donors. Donors are grouped into those that have the A3/11 epitope of HLA-A, and those that lack this epitope. Representative of 3 replicate experiments. (F) Comparison of the frequencies of Zap70^low NK cells expressing KIR3DL1 but not KIR3DL2, and of the NK cells that bind the 539304 antibody but do not express KIR3DL1, from 5 donors, either prior to MACS depletion of KIR⁺ NK cells (ex vivo), or following MACS depletion of KIR⁺ NK cells and subsequent KIR induction by IL-15. Representative of 3 independent experiments, each with different donors. (G) Comparison of the geometric mean fluorescence intensity of Zap70 in CD4⁺, CD8⁺ T cells, and NK cells from 6 donors expressing either KIR3DL1, or the KIR3DL2 epitope recognized by 539304 as the only KIR. Shown are the results of Sidak’s comparison of 2-way ANOVA. Result is representative of 3 independent experiments, each with different donors. ***=p<0.001. (H) Levels of Zap70 and Syk transcripts measured by qPCR, normalized to the transcript levels of Beta-Actin. Data were derived from pools of cells that were sorted to be either 539304⁺DX9⁻ or 539304⁻DX9⁺. Transcript amounts are represented as a percentage of the average normalized transcript in the 539304⁻DX9⁺ NK cells. Data are representative of those obtained from the NK cells of 12 donors from 3 independent sorts and 3 technical replicates. (I) For all NK cells studied, the geometric mean fluorescence intensity of Zap70 is correlated with the square root of the frequency of NK cells binding the anti-KIR3DL2 antibody 539304. n=34. Test of linear fit = p<0.0001. Data are representative of 3 technical replicates of subsets of donors. SEM are shown in all panels.

Nearly all NK cells bound by antibody 539304 are Zap70^low (Fig. 7B–7C). NK cell subpopulations expressing NKG2A or KIR3DL1, or the combination of KIR3DL1 and a KIR2D have higher frequencies of 539304⁺ NK cells than other NK cells (Fig. 7D). To determine if NK cell education affected the proportion of 539304⁺ NK cells, we divided the donor cohort into two groups: one comprising donors having the HLA-A3/11 epitope recognized by KIR3DL2, and the other comprising donors lacking the A3/11 epitope. Both groups of donors were seen to have similar frequencies (Fig. 7E), showing there is no affect of KIR3DL2-mediated education on expression of the epitope recognized by the 539304 antibody.

Culturing KIR⁻ NK cells with IL-15 induced de novo expression of KIR and of the epitope recognized by antibody 539304. The 539304⁺ NK cells were nearly all Zap70^low (Fig. 7F). Similar analysis of T cells (35) identified small, mutually exclusive subsets expressing either the KIR3DL2 epitope recognized by 539304 or KIR3DL1 recognized by the DX9 antibody (Supp. Fig. 2F). CD4 and CD8 T cells bound by 539304 were also seen to have lower Zap70 expression when analyzed for geometric mean fluorescence intensity (Fig. 7G).

Because the epitope recognized by 539304 appears to mark cells lacking Zap70, we investigated if 539304⁺ NK cells express Zap70 or Syk transcripts. Groups of 539304⁺DX9⁻ NK cells, and 539304⁻DX9⁺ NK cells were obtained from several donors and pooled. RNA was extracted from the two pools and reverse transcribed to give cDNA. The cDNA was analyzed by quantitative real-time PCR for Zap70 and Syk transcripts as described above. That 539304⁺DX9⁻ NK cells contained far fewer Zap70 and Syk transcripts than 539304⁻DX9⁺ NK cells (Fig. 7H) indicates that the Zap70 and Syk genes are not expressed in 539304⁺ NK cells, which in this regard are different from other ex-vivo Zap70^lowSyk^low NK cells (Compare Fig. 5C to 7H).

Expression of the 539304 epitope and Zap70 content were assessed in a cohort of 34 donors. The best linear fit for an x-y plot of these two parameters showed an inverse correlation between the square root of the percentage of NK cells bound by the 539304 antibody, and the median NK cell expression of Zap70 (Fig. 6F). Thus, NK cell populations with higher levels of Zap70 expression tend to have lower numbers of clone 539304⁺ NK cells.

DISCUSSION

We describe a subset of human NK cells having low levels of the Zap70 and Syk kinases, critical factors for NK cell activation. Consistent with this role, the Zap70^low/Syk^low NK cells are functionally deficient. Although Zap70^low/Syk^low cells comprise a small fraction of NK cells ex vivo, the Zap70^low/Syk^low phenotype can be induced in vitro by subjecting NK cells to prolonged stimulation from beads coated with antibodies specific for activating NK cell receptors. Characterization of Zap70^low/Syk^low NK cells was facilitated by a mouse monoclonal antibody 539304, which was raised against KIR3DL2 and binds specifically to the surface of Zap70^low/Syk^low NK cells and T cells.

Populations of ex vivo NK cells comprise cells expressing Zap70 and Syk at various levels. NK cells having both kinases at high level were strongly cytolytic, whereas NK cells having one kinase at high level achieved only half this response, and NK cells having neither kinase at high level were barely cytolytic. Zap70 and Syk are thus seen to make additive contributions to NK cell cytotoxicity. Downregulation of Zap70 and Syk is predominantly a property of NK cells for which an activating receptor has engaged its cognate ligand. Cell division is also highly associated with Zap70 loss, but it is unlikely to be its cause, because only some cells that lose Zap70 during prolonged stimulation have divided. Our observations favor a mechanism in which sustained activation of receptor signaling is the primary cause of Zap70 and Syk downregulation.

Paolini et al (42) demonstrated that human NK cells down-regulate Zap70 and Syk following CD16 engagement through ubiquitination and proteolysis (42). We now see that this is a more general phenomenon, induced by sustained engagement of multiple activating NK cell receptors. Syk loss by ex vivo NK cells caused by CD16 binding to Obinutuzumab, correlated with a functional shift toward IFNγ production (43). In our study, we found that loss of Syk and Zap70 in vitro correlates with reduction in IFNγ production.

Our observations suggest that the degree to which Zap70 and Syk are downregulated in an NK cell population, correlates directly with the extent of NK cell activation. Previous analysis of the reduction in Zap70 and Syk expression caused by CD16 engagement, fits with this trend (42), because CD16 is the one receptor that can activate NK cells without involvement of another activating receptor (37). Chronic stimulation of mouse NK cells by NKG2D causes down-regulation of the associated DAP-10 and DAP-12 signaling proteins, and impedes other, unrelated activation pathways (44, 45). Such studies point to chronic engagement affecting the abundance of proteins that contribute to various stages in the activation pathway. Our findings also suggest that Zap70 and Syk do not need to participate in signaling in order for their downregulation to occur following stimulation. This is evident by the fact that engagement of DNAM1 and NKG2D induced Zap70 and Syk loss, but neither activation receptor is known to propagate signals through Zap70 or Syk in human NK cells (46, 47). The connections between activating receptor signaling, kinase loss, and functional cessation, raise the possibility that these events combine to form a physiological mechanism for limiting the extent and duration of the NK cell response. Because an enhanced NK response can limit T cell memory (7), targets and mechanisms for limiting NK responses are of increasing relevance to vaccine design (48).

Our analysis shows that expression of an inhibitory NK cell receptor acts to prevent NK cells losing Zap70 and Syk, an effect observed for both ex vivo NK cells and NK cells stimulated in vitro. NK cells expressing NKG2A, inhibitory KIR or LILRB1 are all protected from losing Zap70 and Syk. Such protection does not require the presence of the inhibitory receptor’s cognate HLA class I ligand, and is thus independent of NK cell education. We also observed that NK cells expressing more types of inhibitory KIR had correspondingly increased retention of Zap70 and Syk, suggesting a direct and quantitative relationship between inhibitory receptor expression and Zap70/Syk retention. However, Zap70 and Syk are unlikely to interact directly with the ITIMs of these inhibitory receptors, particularly when those receptors are not engaged (49, 50). Further work will be necessary to determine how inhibitory receptors influence the recycling of proteins in the activation pathway.

Various models have been proposed to explain how NK cells become self-tolerant (51–53). That prolonged NK cell activation causes loss of Syk and Zap70, and that this can be prevented by the presence of an inhibitory receptor, seems compatible with the model in which NK cells lacking inhibitory receptors become “disarmed” as a consequence of chronic stimulation of the activating receptors (25). However, our finding that inhibitory receptor engagement does not affect Zap70 and Syk levels, does not support this model. Our observations suggest a new model in which human NK cells use dual mechanisms to achieve self-tolerance. In the first mechanism, NK cells that do not express sufficient inhibitory receptors have low amounts of Zap70 and Syk. In the second mechanism, NK cells that express an inhibitory receptor may be enhanced through education, which takes place when inhibitory receptors bind their cognate ligands. This educational effect likely occurs through a mechanism that is independent of Zap70 and Syk, but might involve the redistribution of activation receptors at micro-domains on the cell surface (54). Educated or not, mature peripheral NK cells increase their chances of losing Zap70 and Syk after sustained engagement with activation ligands.

In allogeneic NK cancer immunotherapies, the transfer of educated donor NK cells into a host with uneducated NK cells mediates a graft-versus-leukemia response, causing a significant reduction in mortality (55). Our findings suggest that selective transplantation of NK cells expressing more KIR, even ones irrelevant to the educational context, may help to retain NK cell functions over days of stimulation. When cancer patients are treated with radiotherapy, their NK cells are more likely than their T cells to die from radiation damage (56). Moreover, the population of NK cells that survive irradiation suffer dramatic loss of function (57). Results from our study raise the possibility that this functional deficiency is caused by NK cell downregulation of the Syk and Zap70.

Although it is well established that Zap70 and Syk are essential for T cell activation (30), their precise contributions to NK cell function are less clearly defined. Mutation in human Zap70 can cause a severe combined immunodeficiency disease (SCID) (30, 58) characterized by the absence of CD8 T cells, functionally defective CD4 T cells and normal numbers of NK cells (59, 60). Study of the NK cells from one such patient, showed they lyse K562 target cells effectively (60). The NK cells of mice deficient for both Zap70 and Syk can also lyse various tumor cells, despite their inability to signal through ITAM-bearing receptors (61). Several lines of evidence point to Zap70 and Syk having distinctive functions in humans and mice (58, 62). Here we have uncovered a natural and coordinated downregulation of Zap70 and Syk in human NK cells that profoundly affects their function.

Syk^low NK cells are associated with latent human cytomegalovirus (HCMV) infection (63, 64). In this context, NK cells with the Syk^low phenotype had normal Zap70 expression. Our analysis of HCMV⁻ donors shows that dynamic control of Syk and Zap70 is not limited to latent HCMV infection, but appears to be part of normal NK cell regulation.

Patients with hepatitis B (65), melanoma (66) and lung adenocarcinoma (67) have functionally deficient NK cells, characterized by high surface expression of Tim-3. In contrast, the functionally deficient Zap70^lowSyk^low NK cells that we describe have chronically low Tim-3 expression (Supp. Fig. 1F), as was also seen for Syk^low NK cells of HCMV-infected individuals (63, 64). Further investigation should define connections between NK cell exhaustion and the regulation of Zap70 and Syk in specific disease contexts.

KIR3DL2 is distinguished from other KIR in being expressed predominantly as a dimer at the cell surface (68). The observed differential reactivity of the anti-KIR3DL2 antibodies could be explained if antibody 539304 is specific for the monomeric form of KIR3DL2. One function of KIR3DL2 is to bind microbial CpG-oligodeoxynucleotides at the NK cell surface and transport them to endosomes, where they are recognized by Toll-like receptors (TLR) (69). If dimeric, but not monomeric, KIR3DL2 binds CpG-oligodeoxynucleotides, chronic activation of NK cells by CpG-oligodeoxynucleotides could cause endocytosis of dimeric KIR3DL2, as well as downregulation of Zap70 and Syk, leaving 539304-reactive monomers as the only form of KIR3DL2 at the NK cell surface.

Enhancement and inhibition of NK cell activity are increasingly likely to prove valuable immunotherapies in different disease contexts. Realizing these goals requires knowledge of mechanisms that activate NK cell populations in the immune response and of mechanisms that contract and terminate their activity when they are no longer needed. Here we have shown that NK cells can reduce their functional role through downregulation of Syk and Zap70 kinases.