Differential contribution of education through KIR2DL1, KIR2DL3, and KIR3DL1 to antibody‐dependent (AD) NK cell activation and ADCC

Abstract The engagement of activating NK receptors (aNKR) stimulates NK cell activity, provided that interactions between inhibitory NK receptors (iNKR) with their HLA ligands do not override them. Abs bound to target cells can also activate NK cells by engaging the CD16 aNKR. NK cell education status is an important factor for Ab‐dependent NK cell activation (ADNKA) of some NK cell subsets. However, whether NK cell education also influences Ab‐dependent cellular cytotoxicity (ADCC) levels is not fully known. ADCC‐GranToxiLux (GTL) assays measured ADCC activity as the frequency of granzyme B positive (%GzB+) target cells. Target cells were anti‐HIV Immunoglobulin G (HIVIG)‐opsonized CEM‐NKr.CCR5 (CEM) cells. Lymphocytes and sorted single positive (SP) NKG2A+, KIR2DL1+, KIR2DL3+, and KIR3DL1+ NK cells, to self‐ and nonself HLA, were used as effectors in ADCC‐GTL assays to examine how education status influenced ADCC activity. ADNKA activity was assessed by stimulating lymphocytes with HIVIG‐opsonized CEMs and measuring the frequency of NK cell populations defined by their expression of iNKRs, along with IFN‐γ, CCL4, and CD107a functions. ADCC: the %GzB+ CEM cells generated by self‐ versus nonself HLA‐specific SPiNKR did not differ. ADNKA: More NK cells educated through KIR2DL1 and KIR3DL1, but not KIR2DL3, responded to ADNKA than their uneducated counterparts. CD16 engagement induced ADCC and ADNKA activity. With the proviso that groups’ sizes were small, our results support the notion that NK cell education does not influence ADCC levels but does contribute to ADNKA activity.

protect humanized mice or rhesus macaques against challenge with HIV or simian/human immunodeficiency virus partially depends on their Fc region, which interacts with Fc receptors (FcR) on innate immune cells. [4][5][6][7][8] One of these FcRs, FcR IIIa (CD16), is found on NK cells, macrophages and monocyte subsets. 6,9,10 CD16 engagement leads to NK cell activation, as measured by secretion of chemokines, cytokines and the release of cytotoxic granules that can initiate the lysis of target cells recognized by Abs bridging effector and target cells. 11,12 The activation of NK cells by Ab-dependent stimuli and Ab-coated target cell lysis by NK cells (Ab bridging NK and target cells) are often measured separately, but are frequently referred to in the literature as Ab-dependent cellular cytotoxicity (ADCC). However, the former should be called Ab-dependent NK cell activation (ADNKA), whereas the latter is more correctly referred to as ADCC.
In a post hoc analysis of the results of the RV144 Thai HIV vaccine trial, which conferred modest protection against HIV infection, ADCC activity was shown to be a correlate of protection. 13,14 Follow-up analyses using systems serology that measured and correlated several HIV envelope (Env) specific Ab Fc-dependent functions and features revealed that NK cell activation by Ab-dependent stimulation that supported immune responses other than cellular cytotoxicity, were also correlated with protective responses. 15,16 Although this work underlines the value of Ab-mediated NK cell activation as a correlate of protection, the influence of NK cell education on the ADNKA and ADCC activities of these NK cells is not fully characterized. Here, we used a standardized source of target cells, that is, HIV Env gp120 coated CEM.NKr.CCR5 and polyclonal anti-Env Abs, to investigate the influence of NK cell education on ADNKA and ADCC.
Tolerance to self and the state of activation of NK cells is determined by an ontogenic process known as education, which requires the interaction of inhibitory NK receptors (iNKRs) with their cognate HLA ligands on neighboring cells. 17,18 Education is a complex process whereby functionality is tuned by the number of iNKRs engaged, the strength of interactions between iNKRs and their ligands and whether activating NK cell receptors (aNKRs) are also engaged. [19][20][21][22][23][24] NK cells lacking iNKRs for self-HLA ligands remain uneducated and hyporesponsive. 25 iNKRs include NKG2A and killer immunoglobulinlike receptor (KIR)3DL1 (hereafter 3DL1), KIR2DL1 (2DL1), KIR2DL2 (2DL2), and KIR2DL3 (2DL3). NKG2A, a C-type lectin receptor, forms a heterodimer with CD94 and interacts with nonclassical MHC class I (MHC-I) HLA-E molecules that present leader peptides from many MHC-I proteins. 26,27 NKG2A binds peptides from the signal sequence of HLA-A, -B, -C, and -G allotypes. 27,28 Recently, Horowitz et al. identified a dimorphism at position -21 of the HLA signal sequence that influences whether a nonamer peptide that includes this amino acid will preferentially bind HLA-E to form a ligand for NKG2A/CD94 or will provide ligands for iKIR. 29 The 3DL1 receptor interacts with a subset of HLA-A and -B antigens that belong to the Bw4 group. 30 Bw4 antigens differ from the remaining Bw6 variants at amino acids 77-83 of the HLA heavy chain. 31 Bw6 isoforms do not interact with 3DL1 receptors such that 3DL1 + NK cells from individuals carrying no Bw4 alle-les are not educated through this receptor. The 2DL3 receptor, which is encoded at the same locus as 2DL2, interacts with HLA-C group 1 (C1) variants that have an asparagine at position 80. [32][33][34] The remaining HLA-C variants, which belong to the C2 group, have a lysine at this position and are ligands for 2DL1 receptors on NK cells. 32 The 2DL3 receptor can also bind certain allelic variants of C2, though with lower affinity than either 2DL1 or 2DL2. 33,35 Therefore, 2DL3 + NK cells from individuals expressing the C1 ligand are educated but remain uneducated or modestly educated through this receptor in individuals who do not carry a C1 group ligand. In contrast, 2DL1 + NK cells require the expression of the C2 ligand on neighboring cells to be educated through this iKIR.
Genome-wide association studies (GWAS) found that genes influencing HIV viral load set point map to the MHC-I region on chromosome 6, which encodes MHC-I proteins that are also recognized by iKIR on NK cells. 36 22,39,40 An upstream region of HLA-C that plays a role in determining HLA-C expression levels was also associated with HIV control in individuals of European American origin in GWAS studies. 36 While the mechanisms underlying this association are related to HLA-C expression levels and the potency of CD8 + T cell recognition of HLA-C-HIV peptide complexes, the potential involvement of NK cells has not been excluded. 41 Our group previously showed that NK cells from carriers of the educating 3DL1-Bw4, compared to those from carriers of noneducating 3DL1-Bw6, KIR-HLA combinations generated similar levels of ADCC activity in target cells. 42

Study population and genotyping
Fifty-four healthy HIV-1-uninfected subjects were included in this study.

Cells and reagents
PBMCs were isolated from blood draws into vacutainers containing EDTA anti-coagulant or from leukophoresis samples by density gradient centrifugation, as previously described. 40

ADCC-GTL assay
PBMCs from 47 subjects were used as effector cells in ADCC-GTL assays. The HLA-A, -B, and -C allotypes of the 47 subjects has been previously reported. 42 The ADCC-GTL assays were performed as previously described. 42,46 Briefly, rgp120 coated CEM target cells were labeled with TFL4 and the viability marker NFL1 (both from OncoImmunin, Inc., Gaithersburg, MD, USA). Target cells (T) were plated at 10 4 cells/well in 96-well V-bottomed plates with effector cells at an E:T of 30:1 with the GzB substrate supplied by the manufacturer. Following incubation for 5 min at room temperature (RT), HIVIG was added (or not, as a control) at a concentration of 50 g/ml (a 1:1000 dilution).
The samples were then pelleted by centrifugation and incubated for 1 h at 37 • C in a humidified 5% CO 2 incubator. Cells were washed, resuspended in wash buffer (supplied by OncoImmunin, Inc., as part of the GTL kit) and acquired with a calibrated LSRFortessa flow cytometer using a high throughput system (BD Biosciences, San Jose, CA, USA).
Cells from 8 subjects were also used as effector cells in a modified ADCC-GTL assay. Table 1 . These populations were washed, counted and used as effector cells in ADCC-GTL assays as described in the previous section at E:Ts ranging from 10:1 to 1:1. PBMC from these subjects, cells were stained using the same cell surface specific Ab panel as described above, fixed with 2% paraformaldehyde, permeabilized with 0.125% saponin and stained intracellularly with anti-GzB-BV510 (GB11) and anti-granule associated perforin-AlexaFluor-647 ( G9, both from BD Biosciences). 47

ADNKA assay
Seventeen subjects were included in studies analyzing the role of NK cell education in ADNKA. Table 1 shows for each of these subjects their Bw4/Bw6 and C1/C2 generic genotypes, their HLA-A, -B and -C allotypes and which KIR genes whose gene products were stained for they carried. Stimulation of 2 × 10 6 PBMCs with rgp120-coated CEM cells were performed as previously described. 42 Briefly, PBMCs and rgp120-coated CEM cells prepared as above were cocultured at a 10:1

Flow cytometry analysis
For ADNKA experiments, between 4 × 10 5 and 1 × 10 6 total events were acquired for each sample using a calibrated LSRFortessa flow cytometer (BD Biosciences). Single stained control beads (CompBead; BD Biosciences) were used in every experiment to calculate compensation.
Boolean gating was used to identify the frequency of NK cells having surface markers defining NK cell populations with each of the 7 possible functional profiles: tri-functional (CD107a + IFN-+ CCL4 + ), bi-functional (any combination of 2 functions) and mono-functional.
Total responsiveness was defined as the sum of the frequencies of tri-, bi-, and mono-functional NK cells. Total CD107a, total IFN-, and total CCL4 were defined as the sum of all functional subsets that included these functions. Analysis was performed using FlowJo software V9.8 (TreeStar, Ashland, OR, USA). The data obtained were corrected for background using PBMCs stimulated with rgp120-coated CEM cells in the absence of HIVIG.

Statistical analysis
Analysis was performed using GraphPad Prism 6 software (GraphPad, La Jolla, CA, USA). Friedman tests with Dunn's post-tests were used to analyze the significance of differences between the frequencies of functional subsets among the NK cell populations gated on inclusively and to assess the significance of differences between the %GzB + CEM cells generated using SPiNKR or total NK cells as effector cells in ADCC-GTL assays when more than 2 matched groups were compared.
Wilcoxon matched pairs tests were used to assess the significance of comparisons for 2 matched data sets. Mann-Whitney tests were used to assess the significance of comparisons for 2 unmatched data sets. Results are displayed as median and interquartile range (IQR).
P-values <0.05 were considered significant.  Figure S1A shows the gating strategy used to assess the %GzB + CEM cells in ADCC-GTL assays. Figure   S1B shows the association between the %GzB + CEM cells generated in ADCC-GTL assays and the percentage loss of gp120 coated CEM cells in a 6 h ADCC assay using PBMCs as effector cells and HIVIG as a source of Ab. Figure S2 shows the gating strategy used to obtain SPiNKR by sorting and the staining of these SPiNKR cells for intracellular GzB and perforin. Figure S3 shows the %GzB + cells generated using NK cell effector cells (E) and opsonized gp120 coated CEM cells as target (T) cells at E:T ranging from 10:1 to 1.25:1. Figure S4 shows the strategy used to gate on NK cell populations expressing defined iNKR for the purpose of assessing the frequency of their functional subsets.

The effect of NK cell education through 2DL1 and 2DL2/2DL3 on the ADCC-GTL assay readout
An ADCC-GTL assay was used to measure %GzB + CEM target cells generated by PBMC effector cells in the presence of HIVIG. 46 Figure   S1A shows the gating strategy used to determine the %GzB + target cells. Figure S1B shows that the %GzB + CEM cells generated in an ADCC-GTL assay positively correlated with the percent loss of gp120 coated CEM target cells when PBMCs were used as effector cells and HIVIG as a source of anti-gp120 specific Ab. 48  genotypes of the effector cells showed no significant between-group differences (Fig. 1C). Thus, we found no evidence that effector cells from subjects with educating iKIR-HLA combinations were better

Populations of educated and uneducated SPiKIR NK effector cells generate similar %GzB + CEM cells in the ADCC-GTL assay
The absence of an effect of NK cell education in the ADCC-GTL assay when PBMCs are used as effector cells could have several   Figure S2 depicts the gating strategy used to isolate SPiNKR cells from total NK cells. Table 2  Due to the low frequency of SPiNKR populations we were unable to use the same E:T for all ADCC-GTL experiments for all study subjects. Table S2 shows for 6 of the study subjects, the frequency of the SPiNKR populations present in total NK cells and the frequency of each of these SPiNKR cells that was CD16 + . There was the possibility that the comparisons in ADCC-GTL results generated by self-and nonself-SPiKIR + NK cells were influenced by differences in the distribution of E:Ts used in the self-and nonself-SPiKIR groups. To control for this, we performed the ADCC-GTL assay using total NK cells as effectors at E:T ranging from 10:1 to 1.25:1. Following background correction there were no differences observed in the %GzB + CEM cells generated within this E:T range (Fig. S3). At lower E:Ts the %GzB + CEM cell declined in a dose-dependent manner. We found no correlation between the E:Ts used to test the effector activity of SPiKIR cells and the %GzB + CEM generated (not shown).
SPiNKR and total NK cells isolated from individual study subjects were tested at the same E:Ts. As another control, we examined %GzB + CEM generated by self-SPiKIR, SPNKG2A, and total NK cells at these E:T. Figure  were considered (Fig. 3). Levels of GzB and perforin in SPNKG2A + NK cells did not differ significantly from those in SPiKIR + NK cells (not shown).

Differential functional responses of NK cell populations to ADNKA
We next analyzed responses to ADNKA of 6 different NK cell popu- ADNKA did not differ significantly (Fig. 4A, P > 0.9, Dunn's post-test).

F I G U R E 2 The influence of NK cell education status of NK populations single positive for inhibitory NK receptors (SPiNKRs) on the frequency of granzyme B positive (%GzB + ) target cells generated in ADCC-GTL assays.
Panels A-C show the %GzB + CEM cells on the y-axis generated in an ADCC-GTL assay using SPiNKR to self-HLA (i.e., educated) and nonself-HLA (uneducated) as effector cells. SPiNKR effectors cells were KIR2DL1 + (2DL1) (A), KIR2DL3 + (2DL3) (B) and KIR3DL1 + (3DL1) (C). Bar heights and error bars represent the median and interquartile range %GzB + CEMs for each data set. In panels D-I, the y-axis shows the %GzB + CEM cells generated in an ADCC-GTL assay using effector cells that were self-SPiNKR + (D-F) or nonself-SPiNKR (G-I) compared to SPNKG2A + cells and total NK cells from the same individual tested at the same effector to target cell ratios. SPiNKR effector cells were 2DL1 + (D, G), 2DL3 + (E, H) and 3DL1 + (F, I

Impact of education through 3DL1 and 2DL1 on ADNKA
The differential functional responses induced by ADNKA prompted us to investigate the role of NK cell education on the frequency of cells responding to ADNKA. For these analyses, we examined total CCL4 responses to ADNKA was greater when from subjects who carried a C2 + than a C1/C1 hmz genotype (Fig. 5E-H). These differences achieved statistical significance for total IFN-and total CCL4 secretion (Fig. 5G, H, respectively, P ≤ 0.04 for both comparisons, Mann-Whitney tests). Thus, NK cell education through 3DL1 and 2DL1 resulted in a higher frequency of NK cells able to secrete IFNand CCL4 in response to ADNKA than NK cells not educated through these iKIR.

Similar frequencies of educated versus uneducated SP2DL3 + NK cells respond to ADNKA
On the other hand, comparisons of the frequency of SP2DL3 + NK cells from C1 versus C2/C2 carriers characterized by the sum of total IFNand total CCL4 secretion responses revealed no significant difference (Fig. 5I, K, L, P ≥ 0.1 for all, Mann-Whitney). Comparisons of the frequency of total CD107a + SP2DL3 + NK cells from subjects carrying an educating versus noneducating 2DL3-HLA-C combination trended towards being significantly lower (Fig. 5J, P = 0.06, Mann-Whitney).
Thus, there was no evidence that education through 2DL3 increased the frequency of SP2DL3 + NK cells to ADNKA.

DISCUSSION
In this study, we found no differences in ADCC activity as measured by the %GzB + targets generated in an ADCC-GTL assay between conditions where effector cells included NK cells from subjects educated or not through 2DL1 or 2DL2/2DL3, as was previously reported for education through 3DL1. 42 We showed that purified SPiKIR + cells originating from carriers of HLA allotypes that did, versus did not, support the education of the SPNK populations behaved similarly in terms of their ability to generate GzB + CEM cells. An When we extended these analyses to 3 SPiKIR + NK cell populations using an exclusive gating strategy we found that a higher frequency of educated, compared to uneducated 3DL1 + and 2DL1 + NK cells, responded to ADNKA by secreting IFN-and CCL4, but not by degranulation. In contrast, the frequency of educated, compared to uneducated, 2DL3 + NK cells stimulated by ADNKA did not differ significantly, with the exception of those expressing CD107a, where uneducated 2DL3 NK cells were more responsive than their educated counterparts.
We showed previously that the %GzB + CEM cells generated by effector cells did not differ based on whether these cells came from individuals carrying, or not, educating KIR-HLA combination for 3DL1 42 and confirmed here that this was also the case when results were stratified for iKIR-HLA 2DL1 and 2DL2/2DL3 pairs. We 2DL3. 42,53 If education were playing a role in ADCC activity it would be expected that educated self-SP2DL1 + NK cells would perceive CEM cells expressing no C2 ligands as missing self and respond to these targets more robustly than did uneducated nonself-SP2DL1 + NK cells.
Because both educated and uneducated SP2DL1 + cells generated frequencies of GzB + CEM that did not differ significantly from each other, this was not observed. There was also no difference in the ability of self-and nonself-SP2DL3 + NK cells to generate GzB + CEM cells. This was despite CEM cells expressing C1 alleles that can engage educated self-SP2DL3 + NK cells to inhibit their activity more effectively than they could inhibit the activity of nonself-SP3DL1 + NK cells. As CEM cells express no Bw4 alleles, they have the potential to be seen as missing self by self-SP3DL1 + NK cells, which should result in higher activation of self-than nonself-SP3DL1 + NK cells to generate GzB + CEM cells, a phenomenon that was also not observed. In sum, these analyses did not detect an effect of education on the ADCC-GTL read out. for their ability to support ADCC activity is valid, despite there being between-group differences in the distribution of E:Ts.
Comparisons of within individual self-SPiKIR + with their SPNKG2A + effector cells at the same E:Ts revealed no between group differences. The number of educated cells in these 2 groups would be expected to be similar. NKG2A + cells are educated through interactions with HLA-E, both of which exhibit limited variation and are expressed in most individuals. 26,27 It is important to note that education through NKG2A may differ from one person to another based on whether the HLA haplotype of the subject favors delivery of MHC-I signal sequence nonamer epitopes to ligands for NKG2A versus iKIR. 29 We did not detect a difference in the %GzB + CEM cells generated in our ADCC-GTL experimental set up by SPNKG2A + NK cells based on the T/M amino acid dimorphism at position -21 of the HLA signal sequence of the HLA-B alleles expressed by the donors whose cells were used in these experiments (not shown). However, the study population used for these analyses was small, which may have precluded detecting differences based on HLA genotype.
There is a report that educated NK cells are superior to uneducated NK cells in their ability to mediate ADCC. 54  Furthermore, interactions between 2DL3 and C1 are weaker and support more modest levels of education than do 2DL2-C1 or 2DL1-C2 combinations. 33,35,63 Overall, these results support the interpretation that NK cell education is playing a role in ADNKA by HIVIG opsonized gp120 coated CEM and that the pattern of activation of educated versus uneducated SP2DL1, SP3DL1 and SP2DL3, NK cells is consistent with missing-self recognition of CEM cells for the first 2 populations and inhibition through recognition of self-ligands for the 2DL3 + NK cells.
It is notable that the frequency of NK cells responding to opsonized gp120 coated CEM by CD107a expression did not differ based on education through 3DL1 or 2DL1. This is consistent with the absence of the effect of education through these receptors on the %GzB + cells generated in the ADCC-GTL assay. 42 The activation of SP3DL1 and SP2DL1 by ADNKA also requires signals through CD16 because the frequency of these cells stimulated by unopsonized CEM is lower than for opsonized gp120 coated CEM. The contribution of these signals to NK cell responsiveness would be expected to be similar for educated versus uneducated SPiKIR + NK cells as the frequency of CD16 + NK cells among these 2 groups of SP2DL1 + , SP2DL3 + and SP3DL1 + NK cells is not significantly different. 49 In summary, factors important in determining ADCC and ADNKA activity differ from each other. The effect of NK cell education in ADCC activity is limited. One explanation for this may be that once NK cells are activated, either by missing self or by Ab-dependent mechanisms they can form immune synapses, kill, disengage, and go on to kill other targets. 64 One NK cell has been estimated to kill a mean of 4 and up to 16 targets. [64][65][66][67] Given that most CD56 Dim NK cells are CD16 + , activation through CD16 would activate a higher frequency of NK cells than activation through missing-self recognition. Once activated, the lesser effect arising from educated NK cells may be overwhelmed by the greater frequency of NK cells activated through CD16. On the other hand, both CD16 engagement and missing-self recognition contribute to ADNKA. By virtue of the fact that individual NK cells acti- We are grateful to Ms. Rachel Bouchard for contacting study participants and coordinating their recruitment to this study, Ms.
Pascale Arlotto and Ms. Josée Girouard for expert nursing skills in obtaining leukophoresis samples, and to Ms. Tsoarello Mabanga and Ms. Xiaoyan Ni for expert technical assistance. We also acknowledge the contribution of the study participants.

DISCLOSURE
The authors declare no conflict of interest.