HLA-F on HLA-Null 721.221 Cells Activates Primary NK Cells Expressing the Activating Killer Ig-like Receptor KIR3DS1

Natural killer cells are effectors of the innate immune system that are particularly well adapted for antitumor and antiviral activity (1). They can control viral replication at early stages of infection by secreting cytokines and chemokines and by killing virally infected cells (1–4). NK cells can also link the innate and adaptive arms of the immune response, leading to NK cell activation of multiple functions, including cytotoxicity (5, 6)

The killer Ig-like receptors (KIRs) are a large family of receptors found on subsets of lymphocytes, including NK cells (7). They are encoded by genes within the KIR genetic region that maps to chromosome 19q 13.4, within the leukocyte receptor complex (8). The KIR region is polygenic, and the KIR genes within this region are polymorphic (9–13). KIR nomenclature is based on whether they have 2 extracellular Ig-like domains (2D) or 3 extracellular Ig-like domains (3D) and whether they have long cytoplasmic tails (L) or short cytoplasmic tails (S). The KIR-L are generally inhibitory, whereas KIR-S are activating receptors (14). NK cell activation status is determined by the integration of activating and inhibitory signals received from cell surface receptors binding their ligands on neighboring cells (15).

The highly polymorphic KIR3DL1/S1 locus is unique among KIR region genes in that it encodes both activating (KIR3DS1) and inhibitory (KIR3DL1) receptors. The 84 unique proteins encoded by this locus identified to date can be classified into four subgroups based on their cell surface density and sequence similarity: KIR3DL1 inhibitory receptors that are not expressed on the cell surface (KIR3DL1-null), those expressed at a low density (KIR3DL1-low), those expressed at a high density (KIR3DL1-high), and KIR3DS1 receptors (16–19). The ligand for KIR3DL1-low and KIR3DL1-high receptors are a subset of HLA-B and -A isotypes belonging to the HLA-Bw4 group, which is defined by amino acids present at positions 77–83 of the HLA H chain (20–22). A dimorphism present at position 80 of HLA-Bw4 H chains dichotomizes these isotypes into those with an isoleucine (*80I) or threonine (*80T) at this position. This dimorphism differentially affects receptor/ligand affinity, inhibitory strength, and NK cell education potency (23–26). The remaining HLA-B alleles belonging to the Bw6 group do not interact with KIR3DL1. KIR3DS1 and KIR3DL1-null receptors have not been reported to interact with HLA*Bw4 Ags (27, 28). However, the presence of two epitopes, one derived from HIV Pol and the other from HIV Nef, have been shown to enable interactions between KIR3DS1 and one of the HLA Bw4*80I Ags, HLA*B57 (29).

Epidemiological studies implicate a role for KIR3DS1 in HIV outcomes. HIV-seropositive individuals who are carriers of at least one copy of KIR3DS1 and HLA-Bw4*80I have a slower time to AIDS than those expressing the receptor or ligand alone or neither (30). NK cells from carriers of the KIR3DS1 plus Bw4*80I genotype also inhibit viral replication more potently than those from carriers of the receptor or ligand alone (31, 32). HIV-infected CD4⁺ T cells from carriers of this KIR/HLA genotype stimulate autologous KIR3DS1⁺ NK cells, but not those expressing inhibitory KIR3DL1, to degranulate (31). Furthermore, there is an association between carriage of the KIR3DS1 homozygous (hmz) genotype and protection from HIV infection because a higher frequency of HIV-exposed seronegative subjects than recently HIV-infected individuals are KIR3DS1 homozygotes, and HIV-exposed carriers of this genotype seroconvert more slowly than do carriers of other KIR3DL1/S1 genotypes (33, 34). KIR3DS1 has also been associated with outcomes of viral infections other than HIV (30, 35–37), cancer (38–40), transplantation (41, 42), and autoimmune diseases (43).

Garcia-Beltran et al. (28) reported that open conformers (OCs) of HLA-F are high-affinity ligands for KIR3DS1. They found that the interaction of KIR3DS1 with HLA-F was functional and led to activation of cell lines expressing KIR3DS1. In this study, we extended these findings by investigating the effects of HLA-F ligation expressed on the 721.221 (221) cell line with KIR3DS1 on the activation of primary KIR3DS1⁺ NK cells. We employed a comprehensive Ab panel able to detect several activating NK cell receptors (aNKR) and inhibitory NK cell receptors (iNKR) to gate on KIR3DS1⁺ NK cells that either coexpressed (inclusive gating) or not (exclusive gating) other aNKR and iNKR. The novel exclusive gating strategy enabled us to study the function of stimulated KIR3DS1⁺ NK cells without the contribution of signaling through other iNKR/aNKR. Abs able to detect CCL4/IFN-γ secretion and CD107a expression were included in the Ab panel to measure the frequency of functional NK cell in phenotypically defined NK cell populations. By comparing these two gating strategies and blocking the interaction of KIR3DS1 and HLA-F, we showed that KIR3DS1⁺ ligation by HLA-F on the 221 cell line activated primary NK cells expressing this receptor.

This study was conducted in accordance with the principles expressed in the Declaration of Helsinki. It was approved by the Institutional Review Boards of the Comité d’Éthique de la Recherche du Centre Hospitalier de l’Université de Montréal and the Research Ethics Committee of the McGill University Health Centre. All subjects provided written informed consent for the collection of samples and subsequent analysis.

The study included 10 KIR3DS1 hmz, 2 KIR3DL1 hmz, and 2 KIR3DL1/S1 heterozygous donors. Genomic DNA was extracted from PBMCs using QIAamp DNA Blood Mini Kits (Qiagen, Mississauga, ON, Canada) as per manufacturer’s directions. KIR3DL1/S1 generic genotyping was performed by PCR, as described previously (33).

PBMCs were isolated from leukapheresis samples or from blood draws into vacutainer tubes containing EDTA anticoagulant by density gradient centrifugation (Lymphocyte Separation Medium, Wisent Bioproducts, St-Bruno, QC, Canada) and cryopreserved in 10% dimethyl sulfoxide (Sigma-Aldrich, St. Louis, MO) and 90% FBS (Wisent).

The HLA-null cell line 221 was a kind gift from Dr. G. Alter (Ragon Institute, Harvard University, Cambridge, MA). Two hundred and twenty-one cells were cultured in RPMI 1640 medium supplemented with 10% FBS, 2 mM l-glutamine, 50 IU/ml penicillin, and 50 mg/ml streptomycin (R10; all from Wisent).

Two million 221 cells were resuspended and incubated in 1 ml R10 (pH 2.2) for 90 s. Acid pulsing was stopped by adding 13 ml of R10 (pH 7.4) to the cells (44). Cells were then washed and resuspended in R10 before use as acid-pulsed 221 cells.

To detect HLA-F on untreated and acid-pulsed 221 cells, we used the mAb 3D11(45–48). One million 221 cells were resuspended in 100 μl of PBS (Wisent) and 5% FBS (FACS buffer). 3D11 mAb or mouse IgG1 (IgG1) isotype control (MOPC-21; BioLegend, San Diego, CA) was added to the cell suspension for 40 min at 4°C. Primary Ab binding was detected by incubating washed cells with 1 μg/ml F(ab′)₂ goat anti-mouse IgG-eFluor 660 (eBioscience, San Diego, CA) secondary Ab for 25 min at 4°C. A recombinant KIR3DS1-Fc chimeric protein (R&D Systems, Minneapolis, MN) was also used to detect HLA-F on untreated and acid-pulsed 221 cells. One million 221 cells were resuspended in 100 μl of FACS buffer and incubated with KIR3DS1-Fc (25 μg/ml) or with human IgG1 (hIgG1) (ET901; BioLegend) as a negative isotype control for 40 min at 4°C. Binding of the KIR3DS1-Fc chimera protein and control were detected by adding goat anti-human IgG Fc-PE secondary Ab (eBioscience) for 25 min at 4°C. To verify that acid pulsing of 221 cells produced HLA cell surface OCs, the mAbs W6/32, 3D12, and 2M2 (BioLegend) were used to stain untreated and acid-pulsed 221 cells. Although 221 cells are HLA-A, -B, and -C null, they do express HLA-E, which is detected by the pan-HLA–specific mAb W6/32, the HLA-E–specific mAb 3D12, and the β₂M-specific mAb 2M2. These mAbs do not recognize OCs of HLA-E on acid-pulsed cells. One million 221 cells were resuspended in 100 μl of FACS buffer with W6/32, 3D12, 2M2, or a mouse IgG2a, κ isotype control for W6/32 (MOPC-173; BioLegend) or a mouse IgG1, κ isotype control for 3D12 and 2M2 (MOPC-21; BioLegend) for 30 min at 4°C. Primary Ab binding was detected using F(ab′)₂-goat anti-mouse IgG-PE secondary Ab (eBioscience). Cells were then washed twice and fixed in 2% paraformaldehyde (PFA; Santa Cruz Biotechnology, Santa Cruz, CA) until acquisition within 2 h.

Cryopreserved PBMCs were thawed and rested in R10 for 3 h in a humidified 5% CO₂ incubator. Rested PBMCs (effector) were cocultured with either untreated or acid-pulsed 221 cells (target) at an E:T ratio of 5:1 in 200 μl of R10 in U-bottom 96-well plates for 6 h at 37°C in a humidified 5% CO₂ incubator. PBMCs cultured alone in R10 served as an unstimulated negative control, and PBMCs were stimulated with 5.8 μg/ml PMA (Sigma-Aldrich); 1.2 μg/ml ionomycin (Sigma-Aldrich) in R10 served as a positive control to ensure that the NK cells being stimulated were viable and functional. Brefeldin A (24 μg/ml; Sigma-Aldrich) and monensin (3 μg/ml, GolgiStop; BD Biosciences, Mississauga, ON, Canada) were added 30 min after the initiation of the coculture. All cells responded to 5.8 μg/ml PMA, 1.2 μg/ml ionomycin stimulation.

After stimulation, cells were stained using a fluorochrome-conjugated Ab panel that was validated with single-stained control beads (CompBead; BD Biosciences). Cells were stained for viability using the UV Live/Dead staining kit (Invitrogen, Carlsbad, CA), as per manufacturer’s instructions. Cells were surface-stained with anti–CD3-BV785 (OKT3), anti–CD56-BV711 (HDC56), anti–KIR2DL1/S1/S3/S5-FITC (HP-MA4), and anti–CD107a-BV421 (H4A3) (all from BioLegend); anti–KIR3DL1/S1-PE (REA168) and anti–NKG2A-biotin (REA110) (both from Miltenyi Biotec, Auburn, CA); anti–KIR2DL2/L3/S2-FITC (CH-L; BD Biosciences), anti-KIR3DL2–allophycocyanin (539304), and anti-ILT-2–allophycocyanin (292305) (both from R&D Systems) for 30 min at 4°C. After washing, binding of anti-NKG2A was detected by incubating cells with Qdot 655 Streptavidin Conjugate (Life Technologies, Burlington, ON, Canada) for 20 min at 4°C. Cells were then washed and fixed with 2% PFA, permeabilized with Permeabilization Medium B (Invitrogen), and stained intracellularly with anti–CCL4-AF700 (D21-1351) and anti-IFN-γ–BV510 (B27) (both from BD Biosciences). Samples were then washed twice and fixed with 2% PFA until acquisition.

To investigate whether HLA-F on 221 cells can interact with ILT-2 on NK cells to mediate inhibitory signals affecting NK cell function, the anti-ILT-2–allophycocyanin Ab in the cell surface Ab panel described in the previous paragraph was changed to one conjugated with PE/Cy7 (clone GH1/75; BioLegend) so that ILT-2^+/− NK cells could be exclusively gated on and assessed for the frequency of functional cells. To verify that 221 stimulation activated NK cells, we measured the frequency of CD69⁺ NK cells in gated KIR3DS1⁺ NK cells that were unstimulated or stimulated exclusively by adding anti–CD69-BV605 (clone FN50; BioLegend) to the Ab mixture used for cell surface staining described in the previous paragraph.

For some experiments, PBMCs were stimulated with EBV-transformed B lymphoblastoid cell lines instead of 221 cells using the same conditions and Ab panels as described above. The EBV lines were HLA-F⁺ as determined by staining with 3D11 as described above for staining 221 cells.

For some experiments, the interaction between KIR3DS1 on NK cells and HLA-F on 221 cells was blocked by incubating 221 cells in R10 with 25 μg/ml KIR3DS1-Fc chimera protein (R&D Systems) or 25 μg/ml anti-HLA-F–specific mAbs 3D11, 4A11, or 6A4 for 50 min at 37°C in a humidified 5% CO₂ incubator. For these experiments, PBMCs were also incubated with anti-human CD16 (FcγRIII) F(ab′)₂ (3G8; Ancell Corporation, Bayport, MN) at 10 μg/ml final concentration for 45 min prior to coincubation with 221 cells to minimize the effect of signaling through this Fc receptor on NK cell activation. Fc receptor–blocked PBMCs were cocultured with 221 cells for 6 h at 37°C in a humidified 5% CO₂ incubator. The control conditions for these experiments were unstimulated PBMCs, PBMCs stimulated with 221 cells that were not preincubated with KIR3DS1-Fc chimera protein or anti-HLA-F–specific mAbs, and 221 cells preincubated with hIgG1 isotype control (ET901; BioLegend) at a concentration of 25 μg/ml as a control for KIR3DS1-Fc or with 25 μg/ml MOPC-21 as a control for anti-HLA-F–specific mAbs 3D11, 4A11, and 6A4. Following coculture, cells were stained for cell surface and intracellular markers as described above. The percentage of reduction in 221 KIR3DS1⁺ or KIR3DL1⁺ NK cell stimulation in the presence of blocking reagents was calculated using the following equation: ((frequency of functional cells in the 221-stimulated NK cell population in the presence of a blocking reagent − frequency of functional cells in the unstimulated NK cell population) / (frequency of functional cells in the 221-stimulated NK cell population − frequency of functional cells in the unstimulated NK cell population) × 100).

Between 7 × 10⁵ and 1.5 × 10⁶ total events were acquired for each sample using an LSR Fortessa X-20 flow cytometer (BD Biosciences). Results obtained from flow cytometry data were analyzed by FlowJo software (version 10.2; Tree Star, Ashland, OR). We measured the frequency of KIR3DS1^+/−CD3⁻CD56^dim NK cells exhibiting the seven possible combinations of CCL4, IFN-γ, and CD107a function (i.e., trifunctional, three combinations of bifunctional, and three combinations of monofunctional response patterns). We also assessed the sum of the frequencies of all functions tested (total) and the sums of the frequencies of functional subsets secreting CCL4 (total CCL4), secreting IFN-γ (total IFN-γ), and expressing CD107a (total CD107a). These functional subsets were assessed for KIR3DS1⁺ and KIR3DS1⁻ NK cell populations using gating strategies that either included or excluded NK cell populations coexpressing the NKRs KIR2DL1/L2/L3, KIR2DS1/S2/S3/S5, KIR3DL2, ILT-2, and NKG2A. The data presented are background subtracted using results for matched unstimulated control conditions.

GraphPad Prism 6 (GraphPad Software, La Jolla, CA) was used to perform data analysis and graphical presentation. Wilcoxon matched-pairs signed rank and Friedman tests with Dunn posttests were used to assess the significance of comparisons of two and more than two groups, respectively, within individual matched data sets. The p values <0.05 were considered significant.

Acid pulsing favors OCs of cell surface classical and nonclassical MHC class I (MHC-I) Ags, including HLA-F. HLA-F OCs are the preferred conformation for interactions with KIR3DS1 (28, 49). Fig. 1A and 1B show that 3D11 and KIR3DS1-Fc bound to untreated (top panel) and acid-pulsed (second panel) 221 cells with a similar mean fluorescence intensity (MFI). Isotype controls for these HLA-F–specific reagents bound untreated (third panel) and acid-pulsed (bottom panel) 221 cells with lower MFIs than did 3D11 and KIR3DS1-Fc and with MFIs that did not substantially differ from each other. To ensure that the acid-pulsing regimen produced MHC-I OCs, we stained untreated and acid-pulsed 221 cells with the mAbs W6/32, 3D12, and 2M2. W6/32 recognizes correctly folded classical MHC-1a and nonclassical MHC-1b HLA-E and -F Ags, but not their OC counterparts (28). 3D12 is specific for correctly folded HLA-E, which is expressed at a low intensity of 221 cells. mAb 2M2 is specific for β₂M, which is present on correctly folded HLA-E but not OCs on 221 cells. W6/32 bound untreated 221 cells (top panel) with a higher MFI than it did acid-pulsed 221 cells (second panel), which was in the range of the MFI with which the isotype control for this mAb bound untreated (third panel) and acid-pulsed (bottom panel) 221 cells (Fig. 1C). 3D12 bound untreated 221 cells (top panel) with a marginally higher MFI than it did acid-pulsed 221 cells (second panel). The MFI with which the isotype control for this mAb bound untreated (third panel) and acid-pulsed (bottom panel) 221 cells was similar (Fig. 1D). 2M2 bound untreated 221 cells (top panel) with a higher MFI than it did acid-pulsed 221 cells (second panel). The MFI with which the isotype control for this mAb bound untreated (third panel) and acid-pulsed (bottom panel) 221 cells was similar (Fig. 1E).

We next compared the flow cytometry plots generated for total CCL4 secretion, total IFN-γ secretion, and total CD107a expression for KIR3DS1⁺CD56^dim NK cells stimulated with untreated or acid-pulsed 221 cells and for KIR3DS1⁺CD56^dim NK cells left unstimulated. The strategy used to gate on functional KIR3DS1⁺CD56^dim NK cells is shown in Fig. 2A and 2C. The histograms for these three functional profiles overlapped in KIR3DS1⁺CD56^dim NK cells from a representative individual stimulated with untreated 221 cells (top panel) and acid-pulsed 221 cells (middle panel) and had a higher MFI than that seen in unstimulated cells (bottom panel) (Fig. 3A–C). The frequency of KIR3DS1⁺CD56^dim NK cells from 10 KIR3DS1 hmz subjects responding to untreated and acid-pulsed 221 cells was similar for the functional profiles characterized by the sum of all functions, total CCL4 secretion, total IFN-γ secretion, and total CD107a expression (Fig. 3D–G, respectively).

Collectively, these results show that acid pulsing 221 cells did not improve HLA-F recognition by either an HLA-F–specific mAb or the KIR3DS1-Fc chimera protein and did not enhance the ability of 221 cells to stimulate KIR3DS1⁺CD56^dim NK cells. Given these results, we used untreated 221 cells in subsequent experiments.

PBMCs were stimulated with 221 cells and then stained with the Ab panel described in the 2Materials and Methods that included fluorochrome-conjugated Abs able to recognize KIR2DL1/L2/L3/S1/S2/S3/S5, KIR3DL2, ILT-2, and NKG2A on KIR3DS1^+/−CD56^dim NK cells. As seen in Fig. 2A, live lymphocyte singlet CD3⁻CD56^dim NK cells that inclusively expressed, or not, KIR3DS1 were gated on. Boolean gating was used to assess the frequency of KIR3DS1⁺ and KIR3DS1⁻ cells that expressed the various combinations of the three functions tested. The frequency of cells exhibiting the sum of all functions tested, total CCL4 and total IFN-γ secretion, and total CD107a expression was higher among KIR3DS1⁺ than KIR3DS1⁻ NK cells (p < 0.0001 for all, Wilcoxon matched-pairs signed rank test) (Fig. 4A–D).

For 10 subjects, we assessed the frequency of trifunctional, three combinations of bifunctional, and three combinations of monofunctional KIR3DS1^+/− NK cells. A higher frequency of inclusively gated KIR3DS1⁺ than KIR3DS1⁻ NK cells secreted CCL4 only and both CCL4 and IFN-γ (p = 0.027 and p = 0.002, respectively, Wilcoxon test) (Supplemental Fig. 1A, 1B), whereas a lower frequency of KIR3DS1⁺ than KIR3DS1⁻ NK cells expressed CD107a only (p = 0.039) (Supplemental Fig. 1C).

Inclusively gated KIR3DS1⁺ and KIR3DS1⁻ NK cells can express various stochastic combinations of other inhibitory KIR (iKIR), activating KIR, and NKG2A depending on the immunogenetic profile of the study subject. NK cells bearing iKIR to self-HLA and NKG2A would be educated through the interaction of these receptors with their HLA ligands. This would make them responsive to stimulation by 221 cells. The level of stimulation resulting from missing-self recognition of HLA-null cells by educated NK cell subsets can differ from one person to another and may confound differences in the stimulation levels of inclusively gated KIR3DS1⁺ and KIR3DS1⁻ NK cells. This prompted us to minimize this possibility by using an exclusive gating strategy to gate out NK cell populations bearing KIR2DL1/L2/L3/S1/S2/S3/S5, KIR3DL2, ILT-2, and NKG2A from KIR3DS1^+/−CD56^dim NK cells. The gating strategy used for exclusive KIR3DS1⁺ and KIR3DS1⁻ NK cell gating is shown in Fig. 2B.

Supplemental Fig. 2 shows that 221 stimulation increased the frequency of CD69⁺ on exclusively gated KIR3DS1⁺ and KIR3DL1⁺ NK cells (p = 0.04 and p = 0.057, respectively, Wilcoxon test). Fig. 5 shows for the 10 KIR3DS1 hmz subjects that even after excluding NK cells expressing the NKR recognized by the Ab panel other than KIR3DS1, a higher frequency of KIR3DS1⁺ than KIR3DS1⁻ NK cells responded to 221 stimulation. This was the case when all functions tested were considered (p = 0.005, Wilcoxon test) as well as when all cells positive for CCL4, IFN-γ, and CD107a were considered (p < 0.04 for all, Wilcoxon test) (Fig. 5B–D). A higher frequency of exclusively gated KIR3DS1⁺ than KIR3DS1⁻ NK cells also secreted CCL4 only and both CCL4 and IFN-γ (p = 0.027 and p = 0.009, respectively, Wilcoxon test) (Supplemental Fig. 1D, 1E), whereas the frequency of KIR3DS1⁺ and KIR3DS1⁻ NK cells expressing CD107a only did not differ significantly (p > 0.05) (Supplemental Fig. 1F).

Dulberger et al. (50) recently showed that HLA-F–β₂M complexes loaded with peptide can interact with the iNKR ILT-2. We reasoned that if HLA-F–β₂M complexes on 221 cells were interacting with ILT-2 on NK cells, then a lower percentage of exclusively gated ILT-2⁺ than ILT-2⁻ NK cells would be functional. Supplemental Fig. 3 shows that there were no significant differences in the frequency of CCL4- and IFN-γ–secreting cells in the exclusively gated ILT-2⁺ versus ILT-2⁻ populations (p > 0.05, Wilcoxon test, Supplemental Fig. 3B, 3C). The frequency of ILT-2⁺ with total functionality and secreting CD107a was significantly lower than that of ILT-2⁻ cells (p = 0.004 and p = 0.04, Supplemental Fig. 3A, 3D).

To determine whether the responsiveness of KIR3DS1⁺ NK cells was due to the interaction of this receptor with HLA-F on 221 cells, we blocked this interaction by preincubating 221 cells with KIR3DS1-Fc chimera protein before using them to stimulate PBMCs. As shown in Fig. 6, this treatment diminished the frequency of exclusively gated KIR3DS1⁺ NK cells responding to 221 cells compared with that observed for responses to 221 cells that were not pretreated with KIR3DS1-Fc or were pretreated with control hIgG1. The frequency of functional KIR3DS1⁺ NK cells under KIR3DS1-Fc blocking conditions was significantly lower than that for either the no blocking or blocking with hIgG1 control conditions for total responsiveness, total CCL4 secretion, and total IFN-γ secretion (p < 0.05 for all, Friedman tests with Dunn posttests) (Fig. 6A–C). The frequency of functional KIR3DS1⁺ NK cells under KIR3DS1-Fc blocking conditions was also lower than that for the blocking with hIgG1 control conditions for total CD107a expression (p < 0.05, Dunn posttests) (Fig. 6D). By contrast, differences in the frequency of functional cells stimulated with KIR3DS1-Fc–blocked 221 cells was as low as that of unstimulated cells for each of these functional subsets (Fig. 6A–D).

Three HLA-F–specific mAbs were also used to block the interaction of HLA-F with KIR3DS1. Blocking with 3D11 was tested on five KIR3DS1 hmz subjects, whereas blocking with 4A11 and 6A4 was tested on three subjects each. Each of these mAbs reduced the frequency of functional cells compared with that seen for 221-stimulated cells in the presence or absence of hIgG1.The differences between stimulated cells with or without blocking with 3D11 all trended toward but did not achieve significance (p = values between 0.0625 and 0.12). Precoating 221 cells with these anti–HLA-F mAbs did not reduce the frequency of functional cells to the levels seen for unstimulated PBMCs (Fig. 6E–H).

Blocking experiments with KIR3DS1-Fc were also performed on exclusively gated KIR3DS1⁺ NK cells from two KIR3DL1/S1 heterozygotes (Supplemental Fig. 4) Blocking the KIR3DS1–HLA-F interaction reduced the frequency of functional cells to 37.8, 0, 16.4, and 74% of that of stimulated PBMCs for the sum of all and total CCL4, IFN-γ, and CD107a functions (Supplemental Fig. 4A–D). Garcia-Beltran et al. (28) reported that HLA-F is a ligand for KIR3DL1. Burian et al. (49) did not confirm this finding. KIR3DL1-Fc was used to block interactions between KIR3DL1 and HLA-F using PBMC responder cells from two KIR3DL1 hmz subjects and two KIR3DL1/S1 heterozygotes. The frequency of functional KIR3DL1⁺ NK cells averaged 74.3, 56.3, 46.3, and 99.1% of that of 221-stimulated PBMCs for the sum of all and for total CCL4, IFN-γ, and CD107a functions, respectively (Supplemental Fig. 4E–H). None of these differences achieved statistical significance, likely due to the small number of subjects tested.

Together, these results show that blocking the interaction of KIR3DS1 on NK cells bearing this receptor with HLA-F on 221 cells reduced the functionality of these cells to a level that was not significantly different from that of unstimulated KIR3DS1⁺ NK cells and that was significantly lower than that seen in conditions where this interaction was not blocked. Furthermore, the exclusive gating strategy used to detect KIR3DS1⁺ NK cells minimized the possibility that the activation of these cells was due to missing-self recognition of HLA-null cells lacking ligands for iNKR. Thus, the interaction of KIR3DS1 with HLA-F activated primary KIR3DS1⁺ NK cells.

EBV-transformed B cells also express HLA-F (51). We confirmed this expression by staining with 3D11 and used an HLA-F⁺ EBV–transformed cell line to stimulate PBMCs from five KIR3DS1 hmz subjects using the same conditions as described above. A higher frequency of exclusively gated KIR3DS1⁺ than KIR3DS1⁻ CD3⁻CD56^dim NK cells exhibited the sum of all functions tested, total CCL4 and IFN-γ secretion, and total CD107a expression following stimulation (Fig. 7A–D). However, the difference in the frequency of functional KIR3DS1⁺ versus KIR3DS1^– only trended toward statistical significance (p = 0.062 for total function, total CCL4- and total IFN-γ–secreting cells, and p = 0.12 for total CD107a-expressing cells, Wilcoxon tests). Blocking with KIR3DS1-Fc chimera protein reduced the frequency of EBV-transformed B cell–stimulated KIR3DS1⁺ cells exhibiting all functions, total CCL4 and IFN-γ secretion, and total CD107a expression to levels not significantly different from those seen in unstimulated PBMCs (Fig. 7E–H). Differences in the frequency of functional cells in the KIR3DS1-Fc blocking conditions compared with unblocked or control hIgG–blocked conditions were statistically significant for all functions except total CD107a expression when Wilcoxon tests were used, but not when Friedman tests were used, to assess the significance of between-group differences.

KIR3DS1 is an activating NKR whose presence is associated with beneficial HIV outcomes, such as slow time to AIDS and protection from HIV infection (30, 33, 52). The ligand for KIR3DS1 was recently shown to be the nonclassical MHC-Ib molecule HLA-F (28, 49). In this study, we confirmed that HLA-F is present on the surface of the HLA-null cell line 221. Coculture of 221 cells with PBMCs activates a higher frequency of CD3⁻CD56^dim KIR3DS1⁺ than KIR3DS1⁻ NK cells to express CD69, secrete CCL4 and IFN-γ, and express CD107a. We showed that the activation of KIR3DS1 by 221 cells occurred in the absence of interrupted signaling through iKIR because of missing-self recognition or by activation through several coexpressed aNKR by using a gating strategy that excluded KIR3DS1^+/− NK cells coexpressing a panel of iNKR and aNKR. We showed that the ligation of KIR3DS1 by HLA-F on 221 cells was responsible for the higher frequency of stimulated, exclusively gated CD56^dim KIR3DS1⁺ than KIR3DS1⁻ NK cells by demonstrating that blocking the interaction of KIR3DS1 with HLA-F reduced the frequency of 221-responsive KIR3DS1⁺ NK cells.

The nature of the ligand for KIR3DS1 has been elusive until recently. Epidemiological studies found that carriers of KIR3DS1 and HLA-Bw4*80I combinations progress to AIDS more slowly than carriers of the receptor or ligand alone (30). However, in the absence of HLA-Bw4-80I alleles, KIR3DS1 homozygosity is associated with more rapid progression to AIDS (30). Studies investigating NK cell–mediated inhibition of HIV replication found that effector cells from carriers of this KIR/HLA genotype combination suppressed HIV replication in autologous infected CD4⁺ T cells more effectively than carriers of the receptor or ligand alone (31, 32). Despite this, attempts to demonstrate a direct interaction between KIR3DS1 and HLA-Bw4*80I Ags have failed (27, 28). An exception to this was the observation that KIR3DS1 could bind the HLA-Bw4*80I Ag HLA*B57 if certain HIV-derived peptides were present to facilitate this interaction (29). Garcia-Beltran et al. used fusion proteins of the extracellular domain of several KIR gene products with the Fc portion of IgG1 to screen a panel of 100 recombinant MHC-I proteins. KIR3DS1-Fc bound no HLA-A, -B, or -C Ags. Rather, the preferred ligand for KIR3DS1 was HLA-F OC (28). Several other KIR3D receptors shared with KIR3DS1 the ability to bind HLA-F OCs (28, 53). Surface plasmon resonance, pull-down experiments, heterodimerization experiments, and HLA-F tetramers were used to show that KIR3DS1 binds to OCs of HLA-F (28, 49).

HLA-F differs from classical MHC-Ia proteins in several ways. MHC-Ia proteins are highly polymorphic, form complexes with β₂M and short peptides, and are expressed at the cell surface of most human cells (54). HLA-F is monomorphic, binds weakly with β₂M and peptide, is retained intracellularly in resting cells, and can be expressed on the cell surface of activated cells as an OC without peptide or β₂M, although its cell surface expression may increase as MHC-Ia is upregulated (44, 55–58). HLA-F is present on the surface of B lymphoblastoid cell lines such as 221, on EBV-transformed B cells, and on HIV-infected cells (28). Goodridge et al. (44) reported that HLA-F can be expressed independently of the endoplasmic reticulum peptide binding pathways because of its altered cytoplasmic domain. Based on differential Ab staining for HLA-F, they concluded that at least three different forms of HLA-F may be expressed over the course of lymphocyte activation. They confirmed that HLA-F is expressed on activated lymphocytes independently of TAP and tapasin. Thus, HLA-F can be spontaneously expressed as an OC (58).

Mild acid treatment of cells dissociates β₂M and peptide from HLA H chains producing OCs (51). We compared the ability of the HLA-F–specific mAb 3D11 and a KIR3DS1-Fc chimera protein to recognize HLA-F on untreated and acid-pulsed 221 cells and found that both reagents bound HLA-F on untreated and acid-pulsed 221 cells with a similar MFI. We also showed that untreated and acid-pulsed 221 cells stimulated a similar frequency of KIR3DS1⁺ NK cells to produce CCL4 and IFN-γ and express CD107a with similar intensities. We verified that the procedure used for acid pulsing produced OCs by showing that the mAbs W6/32 and 3D12 recognized HLA-E on untreated but not on acid-pulsed 221 cells. W6/32 and 3D12 only bind correctly folded HLA-E. Furthermore, the β₂M-specific 2M2 mAb served as a control for acid-pulsed stripping of β₂M from HLA-E on 221 cells because it only recognized unpulsed cells. These findings contrast with what has been shown by Garcia-Beltran et al. (28). They showed that Jurkat cells expressing KIR3DS1 were triggered more potently by acid-pulsed than untreated 221 cells. Despite this, the same group showed that cell-sized coated beads loaded with HLA-F activated KIR3DS1-expressing Jurkat cells to a similar extent, whether acid pulsed or not. Discrepant results may be due to disparate sensitivities of the cell types responding to stimulatory cells expressing HLA-F (primary NK cells versus KIR3DS1-transduced Jurkat cells) or to the nature of the stimulatory signal (untreated and acid-pulsed 221 cells versus HLA-F–coated beads). At least in our hands, the HLA-F present on the surface of 221 cells is in a conformation able to interact with KIR3DS1 present on primary NK cells isolated from KIR3DS1 hmz subjects, and acid pulsing improves neither HLA-F’s stimulatory capability nor its recognition by HLA-F–specific reagents.

The interaction of iNKR with their HLA ligands is necessary for NK cell education, which in turn confers NK cells with the functional potential to become activated if they encounter transformed, stressed, or virus-infected self-cells with downmodulated ligands for these iNKR (59, 60). Some activating KIRs may also participate in NK cell education, but these interactions should tune down NK cells’ responsiveness to altered self-cells (61, 62). HLA-null cells such as 221 are well known to stimulate educated NK cells because they lack the ligands for several iNKR, which transmit inhibitory signals (63–67). Reducing the effect of missing-self recognition on NK cell activation was the impetus for excluding KIR3DS1⁺ and KIR3DS1⁻ NK cells coexpressing a panel of iNKR and aNKR. Abs to KIR2DL5 were not included in this panel because we found that coexpression of KIR2DL5 with KIR3DS1 did not modulate the responsiveness of NK cells expressing KIR3DS1 without KIR2DL5 to 221 stimulation (68). Abs to KIR2DS4 were not included in the panel because none of the 10 KIR3DS1 hmz study subjects carried KIR2DS4, as the gene encoding this receptor is in negative linkage disequilibrium with KIR3DS1 frequently found on telomeric KIR region group B haplotypes (69, 70). The observation that exclusively gated KIR3DS1⁺ NK cells responded better to 221 stimulation than KIR3DS1⁻ NK cells narrowed the possibilities for the identity of the NKR responsible for 221-dependent KIR3DS1⁺ activation. The loss of exclusively gated KIR3DS1⁺ NK cell activation when the interaction between KIR3DS1 and HLA-F was blocked confirmed that the interaction between this receptor/ligand pair is functional and responsible for the activation of primary KIR3DS1⁺ NK cells. In contrast, mAbs to HLA-F were less effective at blocking the activation of KIR3DS1⁺ NK cells. The reason for this is unclear, as the concentrations of mAbs used were saturating in binding assays. It is possible that these mAbs have a higher off rate than the KIR3DS1-Fc chimera protein, exposing HLA-F to KIR3DS1. Alternately, this result may suggest that KIR3DS1 recognizes other ligands on 221 cells whose interaction with KIR3DS1 is not blocked by these Abs. Combining more than one mAb for blocking did not further reduce KIR3DS1⁺ NK cell activation. We cannot formally exclude the possibility that receptors other than KIR3DS1 are involved in the activation of KIR3DS1⁺ NK cells. However, the reduction in the frequency of activated KIR3DS1⁺ NK cells to levels at or below those seen for unstimulated PBMCs when the interaction between KIR3DS1 and HLA-F is blocked by KIR3DS1-Fc chimera protein suggests that the contribution of other NK receptors to activation is unlikely or minimal.

We found that 221 cells that were not acid pulsed did induce a significantly lower frequency of exclusively gated ILT-2⁺ than ILT-2⁻ NK cells exhibiting the sum of all functions and expressing CD107a. However, differences in the frequency of functional cells were small (i.e., 34 ± 12.4% versus 36.4 ± 12.3% for the sum of all functions for ILT-2⁺ versus ILT-2⁻ cells, respectively, and 13.0 ± 5.4% versus 14.6 ± 6.1% for total CD107a-expressing ILT-2⁺ versus ILT-2⁻ cells, respectively). Between-group differences were not significant for the frequency of total CCL4 and total IFN-γ secretion. Thus, there is some evidence for inhibition of the function of ILT-2⁺ cells by HLA-F on 221. The low level of inhibition and the absence of differences in 3D11 and KIR3DS1-Fc staining in acid-pulsed versus unpulsed cells would support the conclusion that much of the HLA-F on 221 cells are OCs and not in complexes with β₂M and peptide, which is the form that can interact with ILT-2 (50).

Activated KIR3DS1⁺ NK cells secrete CCL4 and IFN-γ and express CD107a. It is notable that CCL4 is a chemokine able to bind the CCR5 coreceptor for HIV entry, blocking HIV infection of new target cells (71). IFN-γ is a critical cytokine for innate and adaptive immune responses against viral infections (72), and CD107a is a marker of NK cell degranulation, a surrogate marker for target cell lysis (73). Production of these molecules by stimulated KIR3DS1⁺ NK cells may be a mechanism underlying the association of KIR3DS1 homozygosity with a reduced risk of HIV infection (33, 34, 52). Others have shown that HIV-infected cells express HLA-F and that KIR3DS1⁺ NK cells suppress HIV replication in autologous CD4⁺ T cells (28, 31, 32). The frequency of KIR3DS1⁺ cells in KIR3DS1 hmz subjects was reported to be nearly twice as high as that in KIR3DL1/S1 heterozygotes, with a median of 61% (range 41–81%) in a population of five KIR3DS1 hmz subjects (74). For the 10 KIR3DS1 hmz subjects we studied, the mean ± SD of KIR3DS1⁺CD56^dim NK cells was 46.0 ± 20.2% (range 7.5–75.9%). It is interesting to speculate that the high frequency of KIR3DS1⁺ cells in KIR3DS1 hmz donors may be a factor in the association of this genotype with protection from infection. Mature NK cells express prestored IFN-γ transcripts, granzyme, and perforin and are ready to lyse targets within minutes of activation, which would be a desirable characteristic of cells mediating early responses to HIV that may prevent the establishment of infection (72). Secretion of the chemokines CCL3, CCL4, and CCL5 by activated NK cells also mediate antiviral effects by blocking HIV entry (71). KIR3DS1 is also maintained on the surface of NK cells in HIV-infected individuals (74). The high frequency of NK cells bearing activating KIR3DS1 receptors in individuals carrying the gene encoding this receptor is consistent with the epidemiological data suggesting a critical role for this receptor in controlling HIV-1 pathogenesis and contributing to the prevention of HIV infection (28, 30–34, 52).

We acknowledge Xiaoyan Ni and Tsoarello Mabanga for expert technical assistance and Pascale Artotto and Josée Girouard for nursing support obtaining leukapheresis samples. We thank the study participants without whose generous donation of biological specimens this work would not have been possible.

The authors have no financial conflicts of interest.