Evidence That the Cellular Ligand for the Human NK Cell Activation Receptor NKp30 Is Not a Heparan Sulfate Glycosaminoglycan1

NKp30 (NCR3, CD337) is a natural cytotoxicity receptor, expressed on subsets of human peripheral blood NK cells, involved in NK cell killing of tumor cells and immature dendritic cells. The cellular ligand for NKp30 has remained elusive, although evidence that membrane-associated heparan sulfate (HS) proteoglycans are involved in the recognition of cellular targets by NKp30 was recently reported. The data presented in this report show conclusively that HS glycosaminoglycans (GAG) are not ligands for NKp30. We show that removing HS completely from the cell surface of human 293-EBNA cells with mammalian heparanase does not affect binding of rNKp30/human IgG1 Fc chimera complexes or binding of multimeric liposome-rNKp30 complexes. Removing HS from 293-EBNA cells, culture-generated DC, MM-170 malignant melanoma cells, or HeLa cells does not affect the NKp30-dependent killing of these cells by NK cells. We show further that the GAG-deficient hamster pgsA-745 cells that lack HS and the GAG-expressing parent CHO-K1 cells are both killed by NK cells, with killing of both cell lines inhibited to the same extent by anti-NKp30 mAb. From these results we conclude that HS GAG are not ligands for NKp30, leaving open the question as to the nature of the cellular ligand for this important NK cell activation receptor.

N atural killer cells are cells of the innate immune response essential for the early response to virally infected cells and tumor cells. NK cells interact with dendritic cells (DC) 3 to shape the adaptive immune response (1)(2)(3) and with CD4 T cells to facilitate immune responses to recall Ags (4,5). NK cell killing is dependent on activation receptors interacting with their relevant ligands on target cells, with this activity controlled through the class I HLA-binding inhibitory receptors namely, the killer Ig-like receptors (CD158a, b1, b2, f, e1, k, and z), CD94/ NKG2A, and CD85j (ILT2) (6). NK cell killing is permitted only with reduced levels of expression of class I HLA Ags on target cells that can occur as a result of viral infection and malignant transformation and is a feature of immature DC. Key receptors involved in triggering NK cell lysis against tumor cells are the NK cell-specific natural cytotoxicity receptors (NCR), NKp46 (NCR1, CD335), NKp44 (NCR2, CD336), and NKp30 (NCR3, CD337) (7). NKp46 and NKp30 are expressed on resting and activated NK cells, and NKp44 is expressed only on activated NK cells. NKp30 is the only NCR implicated in the cross-talk between NK cells and DC (8).
Despite the importance of NCR, there is limited information on their ligands. There is evidence that NKp46 and NKp44, but not NKp30, bind viral hemagglutinins (9 -11). However, the nature of the endogenous cellular ligands for the NCR remains unknown. Cellular cytotoxicity assays identify cells that are killed by NK cells through an NCR-dependent process (12,13). Binding studies with recombinant NCR fusion proteins (9,14) also have identified cells expressing ligands for these receptors. By these assays, ligands for the NKp46 and/or NKp30 are detected on human, simian, and hamster cells. Evidence that the sulfated glycosaminoglycan (GAG), heparan sulfate (HS), is involved in the recognition of cellular targets by NKp46 and NKp30 was recently reported (15). Such a widely distributed carbohydrate epitope expressed on a variety of protein and/or lipid structures would nicely explain the broad cellular distribution of the ligands for the NCR.
We show that HS expressed on 293-EBNA cells and other human cells, and on Chinese hamster ovary (CHO)-K1 cells, is not a ligand for NKp30. First, complete removal of HS by treatment with mammalian heparanase did not prevent binding of rNKp30/ human IgG1 Fc chimera complexes or liposome-rNKp30 complexes to 293-EBNA cells. Second, NK cell killing of 293-EBNA cells and other human cell lines was unaffected by heparanase treatment, and killing of untreated and treated cells was inhibited by an anti-NKp30 mAb. Third, the GAG-deficient hamster pgsA-745 cells, which lack cell surface HS, are killed by NK cells to the same extent as the GAG-expressing parent CHO-K1 cells, and killing of both cell lines is inhibited to the same extent by an anti-NKp30 mAb. These studies do not support the notion that HS is a ligand for NKp30. Polyclonal NK cells were generated from purified peripheral blood NK cells by culture with gamma-irradiated MM-170 malignant melanoma cells and rIL-2 (18). The cultured cells were entirely NK cells as they lacked cell surface CD3 and expressed CD16 and/or CD56 and/or CD94. DC were generated from plastic adherent PBMC by culture for 7 days with rGM-CSF and rIL-4 (19). The cultured cells lacked CD14, and expressed CD1c and high levels of HLA-DR, consistent with an immature DC phenotype. The procedures for obtaining peripheral blood were approved by the Human Ethics Committees of the Australian National University and the Australian Capital Territory Department of Health and Community Care.
Mammalian heparanase was prepared from human platelets as previously described (20). Cells, at a concentration of 5 ϫ 10 6 /ml, were incubated with heparanase, at a final concentration of 2 g/ml, in PBS/0.1% BSA for 90 min at 37°C. The cells were washed three times in PBS/0.1% BSA before subsequent assays. In some experiments bacterial heparinase-1 (cat. no. H-2519; Sigma-Aldrich) was used at a concentration of 3 U/ml.

Staining with mAb
Cells were incubated with mAb in PBS containing 0.1% BSA, for 30 min on ice. Cells were washed three times after incubation with mAb before adding the FITC-conjugated second reagent. Cells were fixed in 1% paraformaldehyde and analyzed on a FACScan (BD Biosciences), and the data were processed using WinMDI 2.8 software (͗facs.scripps.edu/͘). All reagents were titrated for optimum concentrations.

rNKp30/human IgG1 Fc/anti-human IgG complexes
Complexes were prepared in PBS/0.1% BSA by combining rNKp30/human IgG1 Fc chimera (final concentration 2.5 g/ml) or human IgG (final concentration 5 g/ml) with FITC-conjugated anti-human IgG (final concentration 20 g/ml). After 45 min on ice, the solution was diluted 4-fold and 25 l was added to 10 4 cells pelleted in V wells. This dilution of the preformed complex was predetermined as the optimum concentration for binding. Cells were incubated for 45 min on ice, washed in PBS/0.1% BSA, and fixed in 1% paraformaldehyde in PBS/0.05% BSA before analysis by flow cytometry. In some experiments the complexes were preformed using biotinylated anti-human IgG instead of FITC-conjugated antihuman IgG. In this case, after incubation with cells and washing, the bound complexes were detected by incubating the cells with streptavidin-PE for 30 min on ice. To ensure binding was specific, the preformed complexes were incubated with an anti-NKp30 mAb (10 g/ml) or a mouse isotype control (10 g/ml) before incubating with the cells.

Liposome-rNKp30 constructs
The extracellular domain of rNKp30 (1C7c) was prepared with a 6-His tag in a baculovirus expression system. Sequences corresponding to the extracellular domain of NKp30 were amplified from cDNAs prepared from mRNAs of NK cells. The PCR primers used for amplification were based on the cDNA sequence for NKp30 (21) (EMBL/GenBank/DDBJ accession number AJ223153). PCR primers also amplified a signal sequence at the 5Ј end and sequences for a 6-His tag at the 3Ј end of the cDNA. The PCR products were cloned into the pFASTbac vector and expressed in a baculovirus system using the Bac-to-Bac system (cat. no. 10359-016; Invitrogen Life Technologies). The recombinant protein was purified from insect culture supernatants by Ni-NTA affinity chromatography and Dynabeads TALON (Dynal Biotech) according to the manufacturer's instructions. Purity of the recombinant proteins was assessed by SDS-PAGE. Correct folding of the recombinant protein was shown by the ability of the protein to block binding of PE-conjugated anti-NKp30 mAb to NK cells. The 6-His rNKp30 protein was bound via the 6-His tag to nitrilotriacetic acid groups embedded in the lipid matrix of fluorescent liposomes (22) by incubation at room temperature for 1 h with mixing every 15 min (23). Control liposomes were prepared without 6-His rNKp30. The fluorescent liposome-rNKp30 complexes and control liposomes were added to cells prepared in RPMI 1640 containing 5 mM HEPES, 10% heat-inactivated FCS, and 5% BSA. The cells and liposomes were incubated at room temperature for 1 h with mixing every 15 min. Cells were washed in PBS containing 5% heatinactivated FCS, and analyzed by flow cytometry. To ensure binding was specific, the rNKp-30 liposome complexes were incubated with an anti-NKp30 mAb (10 g/ml) or a murine isotype control (10 g/ml) before incubating with the cells.

Cytotoxicity assays
Target cells were labeled with 51 Cr by incubation with Na 2 51 CrO 3 (25 Ci/0.5 ϫ 10 6 cells/0.1 ml volume) for 90 min, with and without heparanase, as previously detailed. After washing, 5000 51 Cr-labeled target cells were combined with NK-92 cells or polyclonal NK cells (at E:T cell ratios between 1:1 and 4:1) in a final volume of 0.1 ml of HBSS containing 10% heat-inactivated FCS. For assays containing polyclonal NK cells and human target cells, anti-class I HLA mAb was preincubated with the target cells for 15 min at room temperature before combining with NK cells, to prevent class I HLA binding the inhibitory class I HLA receptors. Also, before combining target cells with NK cells, anti-NKG2D mAb was preincubated with NK cells for 15 min at room temperature to prevent NKG2D ligands on target cells activating NK cell killing. Cultures were set up in triplicate using 96-round-well trays. After 4 h, an extra 0.1 ml of medium was added and the cells pelleted at 250 ϫ g for 2 min. Supernatant (100 l) was harvested and counted in a Packard gamma counter. Percentage of specific lysis was calculated after subtracting the amount of 51 Cr released spontaneously from target cells alone, compared with 51 Cr released from target cells in the presence of 1% Triton X-100. The spontaneous release from target cells was ϳ10% in different experiments.

293-EBNA cells express a ligand for NKp30, which is not HS
In initial experiments we showed that 293-EBNA cells bind the extracellular domain of NKp30 (Fig. 1). For these experiments two different rNKp30 preparations were used, a commercially prepared rNKp30/human IgG1 Fc chimera and a 6-His rNKp30 prepared in our laboratory. Preliminary studies showed that binding of rNKp30/human IgG1 Fc chimera to cells, detected with FITCconjugated anti-human IgG, is weak (2-fold above background). However when soluble complexes of the rNKp30/human IgG1 Fc chimera are prepared with FITC-conjugated anti-human IgG before incubation with the cells, the efficacy of binding is substantially increased, presumably due to a more multivalent structure. This binding is specific because preincubation of the soluble complexes with anti-NKp30 mAb before incubation with the cells inhibits binding by 95% (Fig. 1). In the case of the 6-His rNKp30, the protein was attached via nitrilotriacetic acid groups to fluorescent liposomes to achieve a multimeric complex (22,23). These liposome-rNKp30 complexes also bound specifically to 293-EBNA cells in that preincubation of the liposome-rNKp30 with anti-NKp30 mAb before incubation with cells abolished binding (Fig. 1). These experiments show that NKp30 ligands can be detected on 293-EBNA cells using multimeric complexes of NKp30.
To evaluate HS as a ligand for NKp30, 293-EBNA cells were treated with the mammalian HS degrading endoglycosidase, heparanase (20). The data in Fig. 2A show that treatment with heparanase totally removes cell surface HS detected with two anti-HS-specific mAbs, F58-10E, and HepSS-1. F58-10E reacts with an epitope present in many types of HS, which include N-sulfated glucosamine residues (24), whereas HepSS-1 recognizes a HSspecific epitope containing O-sulfated and N-acetylated glucosamine residues (25). Treatment of 293-EBNA cells with bacterial heparinase-1 only partially removed HS (data not shown), so this enzyme was not used in further studies. Having established the efficacy of mammalian heparanase treatment of 293-EBNA cells, the treated and untreated cells were compared for their ability to bind rNKp30. Data in Fig. 2B show that binding of the liposome-rNKp30 complexes and the rNKp30/human IgG1 Fc/anti-human IgG complexes to heparanase-treated 293-EBNA cells is equivalent to their binding to untreated cells. In all cases binding was inhibited by Ͼ95% by preincubation of the complexes with anti-NKp30 mAb (data not shown). These data show that NKp30 does not bind to HS on 293-EBNA cells.
To further substantiate that HS is not a ligand for NKp30, we tested the ability of NK cells to kill untreated and mammalian heparanase-treated 293-EBNA cells (Fig. 2C). These experiments  used the NK-92 cell line (16) and polyclonal NK cells established from human peripheral blood (18). The NK-92 cell line and the polyclonal NK cells are both effective at killing 293-EBNA cells. The NK-92 cells express 6-fold higher levels of NKp30 compared with the polyclonal NK cells as assessed by anti-NKp30 mAb staining (data not shown). NK-92 is a convenient cell line in cytotoxicity experiments with human target cells as it lacks inhibitory receptors for MHC class I (16) and has only weak expression of the NKG2D activation receptor (H. S. Warren, unpublished observation). By contrast the polyclonal NK cells express a range of class I HLA inhibitory receptors (KIR, CD94/NKG2A, ILT2) and a fully functional NKG2D activation receptor (data not shown). For polyclonal NK cells, blocking class I HLA on 293-EBNA cells with an anti-class I HLA mAb and blocking NKG2D on NK cells with an anti-NKG2D mAb was necessary to demonstrate NKp30dependent killing of 293-EBNA cells. The results in Fig. 2C show that NK-92 cells and polyclonal NK cells kill 293-EBNA cells and that killing is unaffected following treatment of 293-EBNA cells with mammalian heparanase. NK-92 killing of both untreated and mammalian heparanase-treated 293-EBNA cells was almost totally inhibited by anti-NKp30 mAb. In the case of polyclonal NK cells, killing was partially inhibited in the presence of anti-NKp30 mAb, with the extent of inhibition being similar for both the untreated and heparanase-treated cells. As a control in this experiment we showed that killing was not inhibited by a mAb to CD56, a receptor highly expressed by NK-92 cells. It is also of interest that both the NKp30-dependent and the residual NKp30-independent killing of 293-EBNA cells was not affected by removal of HS. Therefore ligands on 293-EBNA cells for other NK cell activation receptors on polyclonal NK cells are also not HS. In these experiments we established, by mAb staining, that HS was not regenerated on the cell surface during the 4 h incubation used for the cytotoxicity assay (data not shown). Therefore 293-EBNA cells are killed by NK cells through NKp30, and the NKp30 ligand is not HS.

Other human cells lines express a ligand for NKp30, which also is not HS
We next showed that HS is not a ligand for NKp30 on culturegenerated DC, MM-170 malignant melanoma cells, and HeLa cells (Fig. 3). All cells express HS, although at different levels, and treatment with mammalian heparanase removed HS from the cell surface (Fig. 3A). The data in Fig. 3B show that killing of DC, MM-170, and HeLa cells by polyclonal NK cells is unaffected by heparanase treatment, and that killing of both untreated and heparanase-treated cells is partially inhibited by NKp30 mAb. In these experiments we verified that HS was not regenerated on the heparanase-treated cells during the 4 h cytotoxicity assay (data not shown). Therefore a number of different human cell lines in addition to 293-EBNA are killed by NK cells through NKp30, and the NKp30 ligands on these cells are also not HS.

An NKp30 ligand is present on CHO-K1 cells and the GAGdeficient pgsA-745 cells
To definitively establish that HS is not a ligand for NKp30 we tested CHO-K1 cells and the mutant GAG-deficient pgsA-745 cells derived from the CHO-K1 cell line, for their ability to bind multimeric NKp30 complexes and to be killed by NK cells. We first established that CHO-K1 cells express cell surface HS detected with the F58-10E mAb (Fig. 4A) and HepSS-1 mAb (data not shown), and that the GAG-deficient mutant lacks cell surface HS (Fig. 4A). We next showed that both CHO-K1 and pgsA-745 cells bind NKp30 complexes (Fig. 4B). In this case we were unable to measure binding of the liposome-rNKp30 complexes as these cells exhibited very high background binding of liposomes lacking attached recombinant protein. However we were able to detect weak and specific binding of rNKp30/human IgG1 Fc/anti-human IgG complexes, provided that biotinylated anti-human IgG was used to make the preformed complexes. In this case bound complex was detected with streptavidin-PE. This detection system is more sensitive compared with complexes prepared with FITC-conjugated anti-human IgG. Both CHO-K1 and pgsA-745 cells bound the NKp30 complexes. We next compared the ability of CHO-K1 with pgsA-745 cells to be killed by NK92 cells and polyclonal NK cells. The extent of NK cell killing of CHO-K1 and pgsA-745 cells was similar and importantly, killing of both cell lines was inhibited to a similar extent by anti-NKp30 mAb and was not inhibited by the control anti-CD56 mAb.

Discussion
The conclusion from this study that HS is not a ligand for NKp30 is based on data using cells that were either replete or deficient in cell surface HS. In one series of experiments, human cell lines were treated with mammalian heparanase, which totally removes HS from the cell surface. In a second series of experiments CHO-K1 cells were compared with mutant pgsA-745 cells that are deficient in their ability to synthesize GAG including HS.
NKp30 complexes bound strongly to 293-EBNA cells irrespective of whether cell surface HS was present or absent. Similarly, NKp30 complexes bound equally well to GAG-expressing CHO-K1 cells and GAG-deficient pgsA-745 cells. In all cases binding was specific in that it was inhibited by an anti-NKp30 mAb. The NKp30 complexes were generated in two ways. The first was a complex prepared by preincubating rNKp30/human IgG1 Fc chimera with FITC-conjugated anti-human IgG, using commercially available reagents. The second was a fluorescent liposome-rNKp30 complex prepared using fluorescent liposomes (23) and a 6-His-tagged rNKp30 prepared in our laboratory. Both complexes gave equivalent and specific binding to 293-EBNA cells. In the case of the CHO-K1 and the pgsA-745 cells, only weak binding was detected and only with the rNKp30/human IgG1 Fc/anti-human IgG complexes.
NK cell killing of 293-EBNA cells, culture-generated DC, MM-170 cells, and HeLa cells was not dependent on the presence of cell surface HS. Similarly, killing of GAG-expressing CHO-K1 cells was equivalent to killing of the GAG-deficient pgsA-745 cells. In all cases killing was inhibited by an anti-NKp30 mAb, the extent of inhibition presumably related to the presence or absence of other activating NK cell receptor/ligand interactions involved in cytotoxicity. Killing of 293-EBNA cells by the NK-92 cells was almost entirely NKp30-dependent, whereas killing of the various human cell lines by polyclonal NK cells was only partly NKp30dependent. The fact that the HS-replete and HS-depleted human cell lines are killed to the same extent by polyclonal NK cells, and that the NKp30-dependent killing was similar shows that HS is not involved in other activating NK cell receptor/ligand interactions involved in NK cell cytotoxicity. Interestingly, killing of CHO-K1 and pgsA-745 cells is mostly NKp30-dependent, and this was the case using either NK-92 cells or polyclonal NK cells. These data show that a limited number of human NK cell activation receptors are involved in NK cell killing of these hamster cell lines. These studies show conclusively for a range of cell lines that the ligand for NKp30 is not HS.
In other studies (our unpublished observations) we have established that treatment of 293-EBNA cells with neuraminidase does not affect binding of NKp30 complexes or their ability to be killed by NK cells. However treatment of 293-EBNA cells with trypsin abolishes binding of NKp30 complexes. Therefore we conclude that the ligand for NKp30 is a protein, and that ligand binding is not dependent on sialic acid or HS moieties.
Our conclusion that HS is not a ligand for NKp30 contrasts with that of Bloushtain et al. (15). In studies with CHO-K1 cells and the mutant pgsA-745 cells, this group showed that binding of an rNKp30/human IgG1 Fc chimera was substantially less on the mutant cells. However the NKp30 specificity of this binding was not established. In our studies, binding of an rNKp30/human IgG1 Fc chimera was seen only when using a preformed complex with anti-human IgG and when the sensitivity of detection was increased using biotinylated anti-human IgG and streptavidin-PE. In other studies (data not shown) using 293-EBNA cells we observed that binding of the rNKp30/human IgG1 Fc chimera to the cells was in part nonspecific because binding was only partially inhibited by preincubating the rNKp30/human IgG1 Fc chimera with an anti-NKp30 mAb. Interestingly, nonspecific binding was eliminated by preforming rNKp30/human IgG1 Fc complexes with antihuman IgG before incubation with the cells. These experiments illustrate that specificity must be controlled in binding studies using recombinant protein/human IgG1 Fc chimeras. In fact, nonspecific binding of PECAM-1/Fc chimeras was noted in an early study by Sun et al. (26), and the potential for nonspecific binding of Fc chimeras to FcRs was noted in a recent study by Stark et al. (14).
The identity of the cellular ligands for NKp30 and other NK cell activation receptors has remained elusive. The identity is not revealed probably because of the low affinity of the receptor ligand interactions and difficulties with nonspecific binding of soluble recombinant protein/human IgG1 Fc chimeras. Enhanced avidity of In these experiments preformed complexes of rNKp30/ human IgG1 Fc-biotinylated anti-human IgG were incubated with the cells, and binding was detected with streptavidin-PE (FL2). Binding of the complexes after preincubation with control mouse IgG2a (filled histogram) and after their preincubation with anti-NKp30 mAb (dotted line histogram) are shown. The line histogram shows cells alone. C, Killing of CHO-K1 and mutant pgsA-745 cells by NK-92 cells and polyclonal NK cells, without a mAb (Medium), with a control anti-CD56 mAb, or with an anti-NKp30 mAb. The E:T cell ratio was 1:1 for NK-92 cells and 4:1 for polyclonal NK cells. The lower limit in the assays, which was 2 SD above the spontaneous release of 51 Cr from target cells in the absence of NK cells, was 0.6 (CHO-K1) and 1.3% (pgsA-745). FL1-H and FL2, Fluorescence intensity channels 1 and 2.
binding can be achieved using multimeric arrays of recombinant proteins, as shown with studies on the interaction of CD2 (27,28) and CD4 (23) with their low affinity ligands. The use of complexes of recombinant proteins as in this study, or the use of trimeric isoleucine zipper-fusion proteins described recently (14), should facilitate the isolation of ligands for NK cell activation receptors.