Preferential Interaction of a Novel Tumor Surface Protein (p38.5) with Naive Natural Killer Cells

N-hydroxy succinamide ester of biotin (biotin–NHS) was purchased from Calbiochem– Novabiochem Corp. (La Jolla, CA). Streptavidin alkaline phosphatase and all cell culture media and reagents were from GIBCO BRL (Gaithersburg, MD). Electrophoresis reagents and chemicals were purchased from Bio-Rad Laboratories (Melville, NY). All other chemicals used were from Sigma Chemical Co. (St. Louis, MO).

Mouse anti–rabbit IgG was purchased from Pierce (Rockford, IL), goat anti–rabbit alkaline phosphatase was obtained from Sigma Chemical Co. (St. Louis, MO). Goat anti–rabbit IgG conjugated to FITC was obtained from Tago, Inc. (Burlingame, CA). Antibody recognizing p38.5 was raised in a rabbit against an 11mer synthetic peptide derived from the amino acid sequence of an internal peptide of the molecule (see below). The synthetic peptide (250 μg) was mixed with Titermax (CytRx Co. Atlanta, GA) (1:1 vol/vol) and injected subcutaneously into a rabbit at four sites as instructed by the manufacturer. Subcutaneous booster injections (100 μg peptide in IFA) were given to the rabbit at bimonthly intervals. Serum was collected 1 wk after the third booster injection. Anti-p38.5 antibodies were affinity purified from sera using BSA, 11-mer peptide conjugate linked to CNBr– Sepharose 4B. Affinity-purified antibody had a titer of 1:2,000 and reacted with a single protein of 38.5 kD on Western blots (see results). Cell lines used in this study were obtained from the American Type Culture Collection (Rockville, MD) and were maintained in media according to instructions provided.

Human peripheral blood lymphocytes (HPBL)¹ were isolated as described previously (13). In brief, blood obtained from healthy volunteers was diluted 1:1 with HBSS, layered on Ficoll–Hypaque (Pharmacia, Piscataway, NJ) and centrifuged at 200 g for 20 min. The mononuclear leukocyte fraction (Ficoll–Hypaque interface) was passed through nylon wool columns to obtain HPBL depleted of B cells and monocytes.

Naive human NK cells are defined as freshly isolated CD3⁻, CD5⁻, CD16⁺ that exhibit limited target cell specificity, i.e., lytic activity against K562, but not A549 tumor cells. Activated NK cells, i.e., those that are stimulated with lymphokines such as Il-2 (LAK), are able to lyse NK-resistant tumor cells such as A549 (1, 4). Only those preparations of HPBL that lacked LAK cell activity were used for the isolation of naive NK cells and T lymphocytes. Isolation of naive NK cells and T lymphocytes was performed as follows. Freshly isolated HPBL were suspended in PBS containing 2% fetal bovine serum (FBS) at 20 × 10⁶ cells/ml and incubated (4°C for 30 min) with anti-CD5, 20 μl/1 × 10⁶ cells (Becton-Dickinson, San Jose, CA). Anti-CD5 reactive lymphocytes were captured by incubating the mixture with sheep anti–mouse IgG-coated magnetic beads, 30 μl/1 × 10⁶ cells (Dynal Inc., Lake Success, NY). The magnetic beads containing adherent CD5⁺ cells were washed, suspended in RPMI supplemented with 15% FBS, and incubated overnight at 37°C in 5% CO₂. Subsequently, CD5-enriched lymphocytes (T lymphocytes) were separated from the magnetic beads by vortexing 2 min and placing the mixture in a magnetic field. The supernatant containing CD5⁻ HPBL was incubated with anti-CD3, 20 μl/10⁶ cells (Becton-Dickinson, San Jose, CA) at 4°C for 30 min. Anti-CD3–reactive lymphocytes were separated using magnetic beads as described above. The supernatant was centrifuged to obtain cells that were depleted of CD3⁺ and CD5⁺ lymphocytes (i.e., CD 16⁺ enriched naive NK cells). The average phenotype of the T lymphocyte enriched fraction was <3% CD16⁺, 50% CD4⁺, and 93% CD3⁺. These cells did not exhibit NK or LAK cytolytic activity. The average phenotype of the negatively selected naive NK cell enriched fraction was 85% CD 16⁺, <3% CD3⁺. These cells exhibited increased NK cell lytic activity (60% cytotoxicity at an effector to target E/T ratios of 100:1) compared with unfractionated HPBL (32% cytotoxicity).

Plasma membrane proteins of viable tumor cells were labeled with biotin as described previously (13). In brief, cells ( >99% viable by trypan blue dye exclusion) were washed three times with a solution containing 10 mM Hepes, 145 mM NaCl, 4 mM KCl, 11 mM glucose, pH 8.0 (buffer A) and then incubated in the same buffer with biotin–NHS (1 mM biotin–NHS, 10 × 10⁶ cells/ml) for 1 h at 4°C. Cells were then washed three times with 20 vol of buffer A to remove unreacted biotin. The biotinylated cells were suspended in 250 mM sucrose with 10 mM Hepes (pH 8.0) and disrupted by N₂ cavitation (1,000 psi) in a Parr's chamber. The resulting suspension was centrifuged at 200 g for 5 min to remove undisrupted cells and the supernatant was centrifuged at 30,000 g for 20 min to obtain a crude membrane fraction. Membranes were washed once with 10 mM sodium borate, 10 mM benzamidine, 1 mM EDTA, 1 mM iodoacetamide, 1 mM PMSF, pH 8.0 (buffer B), and incubated in the same solution with 1% Triton X-100 at 4°C overnight. Solubilized membrane proteins were obtained as the 30,000 g supernatant of the detergent treated membranes and dialyzed extensively against buffer B containing 0.05% Triton X-100, as described previously (13).

Biotin-labeled solubilized tumor plasma membrane proteins were reacted with freshly isolated naive NK cells and T lymphocytes at a ratio of 2:1 (on the basis of cell numbers) in RPMI1640 supplemented with 15% FBS at 4°C for 2 h. The reacted cells were washed twice with media and three times with buffer A. Finally, cells were solubilized in Laemmli sample buffer and the solubilized proteins were subjected to SDS-PAGE and Western blotting. Biotinylated tumor membrane proteins that bound to lymphocytes were identified with a streptavidin biotin detection system as described previously (13).

Membrane proteins were purified as described previously (13). Approximately 10¹¹ cells of erythroleukemia cell line (K562) were washed three times with PBS and used to prepare plasma membrane proteins as described above. Solubilized membrane proteins were extensively dialyzed against buffer B and subjected to preparative SDS-PAGE. A vertical portion of the gel was sliced and stained with Coomassie blue to locate desired proteins. Protein bands of interest were cut from the gel, eluted, and further purified by SDS-PAGE. A portion of the purified protein was extensively dialyzed against 10 mM Tris–HCl, pH 8.0 and used for cytotoxicity inhibition studies or dialyzed against buffer B and labeled with biotin as described above for study of its binding properties to various lymphocyte subsets. A portion of the purified protein was also isolated on nitrocellulose paper after SDS-PAGE using the method described by Aebersold et al. (14) for the purpose of internal amino acid sequence analysis of the protein subsequent to in situ protease digestion (Harvard Microchemistry Laboratory, Cambridge, MA).

Immunoprecipitation of plasma membrane proteins was performed by initially incubating anti-p38.5 with anti–rabbit IgG coupled to CNBr– Sepharose 4B at 4°C for 90 min in PBS containing 1% BSA (13). The reacted beads were then washed three times with buffer and incubated at 4°C for 90 min with surface-biotinylated tumor plasma membrane proteins. Beads were then washed three times with PBS containing 1% BSA and two times with PBS alone, suspended in Laemmli SDS-PAGE sample buffer, and boiled for 10 min. The supernatant was subjected to SDS-PAGE and proteins transblotted to Immobilon-P membrane (Millipore Corp., Bedford, MA). Anti-p38.5 immunoprecipitated biotinylated tumor membrane proteins were identified on Western blots by reaction with streptavidin–biotin detection system.

Approximately, 10 ⁶ cells were washed three times with PBS containing 0.1% BSA (PBS– BSA) and incubated with appropriately diluted anti-p38.5 at 4°C for 30 min. Cells were then washed three times with PBS–BSA and reacted with FITC–F(ab′)₂ goat anti–rabbit IgG for 30 min at 4°C. Finally, cells were washed three times with PBS–BSA, resuspended to a concentration of 10⁶/ml, and analyzed on a FACSort^® Flow Cytometer (Becton-Dickinson, San Jose , CA).

Cytolysis of tumor cells by lymphocytes was measured as described previously (13, 15). In brief, 100 μl of ⁵¹Cr-labeled target cells (2 × 10⁴ cells/ml) was mixed with 100 μl of lymphocytes (2 × 10⁶ cells/ml) to yield serial twofold dilutions of E/T cell ratios of 100:1 to 6:1. Cells were incubated at 37°C for 3 h in 5% CO₂. Incubation of target cells without the effector cells and in media alone served as control for spontaneous release of ⁵¹Cr, and wells without effector cells but with 1% SDS provided the maximum amount of radioactivity present. Cytolysis was calculated as percent cytotoxicity = 100 × ([CPM_exp − CPM_spont.] / [CPM_max − CPM_spont.]).

A tagged ligand–cell adsorption technique (13) was used to identify membrane proteins from a susceptible tumor cell line that preferentially bind to NK cells. Solubilized membrane proteins from surface biotinlabeled K562 cells were reacted with immunomagnetic bead– enriched freshly isolated naive NK cells or T lymphocytes. Approximately 14 common bands were detected on Western blots of K562 plasma membrane proteins that bound to both NK cells and T lymphocytes (Fig. 1). However, a single band of 38.5 kD (p38.5) was detected from NK cell adsorbed preparations, but was not observed in T lymphocyte adsorbed preparations. This initial result suggested the possibility that p38.5 interacted with a receptor on the surface of naive NK cells that was not expressed on T lymphocytes. Accordingly, p38.5 was purified by preparative SDSPAGE to apparent homogeneity for additional studies. The purified protein resolved as a single band of 38.5 kD molecular mass upon SDS-PAGE (Fig 2, lane A). p38.5 was labeled with biotin for use in studying its binding property to lymphocyte subsets (in a similar manner as that described for proteins from the crude membrane preparation). Biotinylated p38.5 bound extensively to the NK cell–enriched fraction (Fig. 2, lane D) and not at all to the enriched T lymphocytes (Fig. 2, lane C). In this experiment, unfractionated cells bound a slight amount of p38.5 (Fig. 2, lane B), which is consistent with the low percentage of NK cells in peripheral blood lymphocyte specimens. These data demonstrate that SDS-PAGE–purified p38.5 retained its preferential binding property for naive NK cells.

Initial attempts to obtain a partial amino acid sequence of the purified p38.5 were unsuccessful because the NH₂ terminus of the molecule was blocked. Consequently, p38.5 was transferred onto nitrocellulose paper and subjected to in situ protease digestion (14). Amino acid sequence analysis of resultant peptides was performed at the Harvard Microchemistry Laboratory (Cambridge, MA). An 11-mer amino acid sequence of an internal peptide was obtained with a high degree of confidence (Table 1). Comparison of this sequence with known sequences of proteins in Gen Bank did not reveal significant homology. Treatment of p38.5 with sulfhydryl reducing agents (2-mercaptoethanol and dithiothreitol) or deglycosylating enzymes (O-glycosidase, N-acetylneuraminidase II and peptide-N-glycosidase F, deglycosylation kit; Glyko, Inc., Novato, CA) did not alter the molecular mass of the molecule, suggesting that the isolated protein is a monomer and nonglycosylated (Das, B., unpublished data).

Because p38.5 (derived from K562 cells) exhibited selective binding to naive NK cells, it was of interest to examine the expression of this molecule on the plasma membrane of NK-resistant cell lines as well as other NK-sensitive tumor cell lines. Affinity-purified antibody raised against the internal 11-mer synthetic peptide of p38.5 (anti-p38.5) was used for this purpose. Anti-p38.5 was reacted with three NK-sensitive and three NK-resistant tumor cell lines and analyzed separately by flow cytometry. As shown in Fig. 3,A, anti-p38.5 reacted with the surface of NK-sensitive tumor cell lines, K562 (erythroleukemia), Jurkat (T cell lymphoma), and Molt-4 (T cell leukemia). In contrast, anti-p38.5 did not bind to the three NK-resistant tumor cell lines A549 (lung adenocarcinoma), Raji (Burkitt lymphoma), and MDA–MB231 (breast carcinoma). Plasma membrane expression of p38.5 was also studied by anti-p38.5 immunoprecipitation of surface-labeled (biotin) membrane proteins of the above tumor cells. Western blot analysis of the immunoprecipitates showed the presence of a single 38.5-kD biotinylated protein from NK-sensitive cell lines, whereas biotinylated proteins were not observed when the NK-resistant cell lines were employed (Fig. 3 B). These results demonstrate an association between the expression of p38.5 on the surface of tumor cell lines and their susceptibility to NK cell–mediated lysis.

Though the expression of p38.5 is associated with susceptibility of well-established tumor cell lines to NK cell–mediated cytolysis, one must interpret such data with caution as the cell lines are derived from different tissues and are likely to have been transformed by different mechanisms. This issue could however be more readily addressed by examining the association of NK susceptibility and p38.5 expression in variants derived from wild-type tumor cells. After long-term culture of Jurkat and Molt-4 cell lines, we observed a substantial reduction in their susceptibility to NK cell–mediated cytotoxicity. These variant cell lines were then analyzed for surface expression of p38.5 by anti-p38.5 immunoprecipitation. As shown in Fig. 4, variant cell lines exhibited decreased susceptibility to lysis by NK cells and expressed significantly lower amounts of p38.5 on their surfaces as compared with the wild-type parental cell lines.

The preferential binding of p38.5 to NK cells suggested that the molecule may be involved as a ligand in the cytolytic process mediated by naive NK cells. To examine this possibility directly, a purified preparation of p38.5 was incubated with human lymphocytes and then tested in a standard cytotoxicity assay against ⁵¹Cr-labeled K562 target cells. Incubation of lymphocytes with the purified p38.5 preparation inhibited naive NK activity in a concentrationdependent manner (Fig. 5). A K562 membrane protein of 40 kD, purified by the same procedure as p38.5, did not affect NK activity. These data provide direct evidence that interaction of tumor surface protein p38.5 with its NK cell receptor is necessary for cytotoxicity, because binding of soluble ligand to the NK cells before contact with K562 cells inhibited their lysis.

To delineate further the preferential interaction of p38.5 with NK cells relative to T lymphocytes, we conducted investigations to identify an interactive molecular species on effector cells. In direct binding assays, biotin-labeled purified p38.5 was reacted with either NK cell or T lymphocyte membrane proteins that had been previously immobilized on Western blots. Labeled p38.5 was detected as a single band at 70 kD on blots of NK cell membrane proteins, whereas reactive bands were not detected on blots of T lymphocyte membrane proteins (Fig. 6). This result suggests that a protein of 70 kD from naive NK cells binds cell surface p38.5 from NK cell–susceptible tumors and that the 70-kD protein is either not present or is present in substantially lower amounts in T lymphocytes.

To establish the immunologic identity of the ligand bound to the NK cell receptor, anti-p38.5 serum was used. A Western blot of peripheral blood lymphocyte membrane proteins was initially incubated with a solubilized crude extract of K562 membrane proteins, washed, and then reacted with anti-p38.5. The antiserum reacted with a band at 70 kD rather than at 38.5 kD (Fig. 7, compare lanes C and D to lane A). Western blots of lymphocyte membrane proteins that were not preincubated with the tumor proteins did not react with anti-p38.5 (Fig. 7, lane B). These results suggest that the 38.5-kD surface molecule from K562 cells that binds to the 70-kD naive NK cell protein contains the 11-mer peptide to which antibody was raised. We conclude that the 11-mer epitope of p38.5 is probably not directly involved in the receptor–ligand interaction. (Presumably, if the p38.5–70-kD interaction included the 11-mer epitope of p38.5, antibody to the latter determinant could not bind to the complex).

The following experiment was performed to confirm that a 70-kD NK cell protein that specifically binds p38.5 is localized on the plasma membrane. Solubilized membrane proteins from surface-biotinylated naive NK cells were applied to a p38.5-Sepharose 4B column. After extensive washing with high salt buffer, specifically bound protein was extracted in denaturing buffer and subjected to SDS-PAGE and Western blot analysis. A single discreet biotinylated band of 70 kD from NK cells was detected on the blots suggesting that this p38.5 binding protein is located on the exterior of the plasma membrane (Fig. 8, lane B). One possibility is that the 70-kD, p38.5 binding protein is the inducible form of heat shock protein (Hsp 70). We performed flow cytometry studies using antibody to Hsp 70 (Stressgen Biotech. Corp., Victoria, BC Canada; Clone C92F3A-5) to determine surface expression of this moleule on T and NK cells. Such experiments indicated that neither NK nor T cells expressed Hsp70 on their plasma membranes in contrast with K562 cells that served as a positive control (Norin, A.J., unpublished data).

Previous attempts during the past two decades to identify recognition structures exclusive to NK cell–tumor interaction have been unsuccessful (reviewed in reference 8), though important components on both NK cells and on tumor cells that contribute to cellular adhesion and regulation of the cytolytic process have been revealed (5, 8–12). However, these receptor–ligand interactions do not appear to be unique to NK cells, because they also occur between T lymphocytes and respective target cells (10). The limited number of NK cells available for biochemical studies (<5% in peripheral blood) has undoubtedly contributed to the difficulty in analysis of NK cell–specific receptor–ligand interactions. Many investigators have addressed this issue by the expansion of freshly isolated NK cells in vitro using a variety of culture systems employing growth factors and/or feeder cells (11, 12, 16). Such techniques have provided sufficient numbers of cells for immunologic and biochemical studies but at the expense of altering the phenotype of the freshly isolated NK cells from a naive state to an activated one, as demonstrated by induction of the Lag 3 protein (17). It is possible that activation results in decreased expression of naive NK cell recognition structure(s), thereby preventing their detection. In the present study, several strategies were used to overcome the difficulties of heterogeneity and the low frequency of NK cell in HPBL preparations. Only freshly isolated NK cells were used. Cells from individuals that demonstrated activity against NK-resistant LAK-sensitive tumor cell lines were not used, because the activated NK cells in these fresh preparations might mask the detection of naive NK cell–specific ligands. Finally, a tagged ligand–cell adsorption technique (13) was used with enriched NK cell preparations to enhance the likelihood of detection of tumor proteins that selectively bound to the naive killer cells.

Using the above approaches and additional biochemical and immunologic techniques, we demonstrate a novel 38.5- kD protein on the plasma membrane of certain tumor cell lines that preferentially reacts with a surface component of naive human NK cells. The interaction appears to be unique to NK cells, because T lymphocytes did not bind p38.5. However, binding studies have not been conducted with B cells, monocytes, or polymorphonuclear leukocytes. In further experiments, a 70-kD protein on the plasma membrane of NK cells was identified as a p38.5 binding molecule. Consistent with p38.5 cellular binding studies, a 70-kD receptor was not detected on T lymphocytes. Addition of purified p38.5 to fresh lymphocyte preparations before incubation with K562 target cells substantially decreased NK cell lytic activity. Preferential interaction of p38.5 with NK cells and its blocking activity in functional assays are properties consistent with a role in an early recognition event in NK cell–mediated tumor cytolysis.

Additional evidence of a role for p38.5 as a target ligand in naive NK cell–mediated cytotoxicity is provided by data demonstrating an association of the expression of this molecule and susceptibility to cytolysis of different tumor cell lines. Flow cytometry and immunoprecipitation studies (of surface labeled cells) revealed that p38.5 is expressed on NK susceptible targets such as K562, MOLT-4, and Jurkat, whereas this molecule was not detected on the plasma membrane of NK-resistant LAK-sensitive targets such as Raji, A549, and MDA-MB-231, suggesting that p38.5 is not involved in LAK-mediated cytotoxicity. The functional role of p38.5 in NK cell–mediated cytolysis was also demonstrated in studies of p38.5 loss variants. After long-term culture of wild-type, NK-sensitive Jurkat and Molt-4 cell lines, variants were isolated that exhibited decreased levels of p38.5 and reduced susceptibility to lysis by NK cells. This property was not due to a phenotypic alteration in the cells as a result of culture conditions, because resistant clones were obtained at limiting dilution (Norin, A.J., unpublished data). These studies clearly establish a strong association between the expression of p38.5 on the tumor plasma membrane and susceptibility to NK cell–mediated cytolysis.

Recent studies suggest that cytolytic activity is affected by recognition of HLA class I polymorphisms on target cells by NK cell receptors of the C-lectin family (CD94) (18) or members of the immunoglobulin multi gene family (p50/ p58) (19–24). These receptors, which we note are detected on both T cells as well as NK cells, may downregulate or upregulate cytolytic activity (10, 20, 25) depending on the subtype of receptor and/or the presence of cytosol associated clonotypic signaling molecules (18, 21–24). Inhibitory receptors apparently interfere with proximal signaling events such as Ca²⁺ flux, phospholipase Cγ2 activity, and inhibition of specific tyrosine kinases activities (lck and ZAP-70), whereas the activity of other tyrosine kinases may remain unchanged (18, 26, 27). Experimental results in the current study are likely not influenced by the above families of molecules, because K562 cells do not express HLA class I or class II molecules (11, 12).

Current concepts regarding the mechanism of lymphocytemediated cytolysis have focused on three alternative pathways: (a) Fas–FasL-induced apoptosis of tumor cells (28–33), (b) extracellular ATP-mediated osmotic lysis (13, 34–36), and (c) granule exocytosis of effector molecules such as perforin and granzymes (29, 37–40). Clearly, NK cell–induced cytolysis is not mediated by Fas–FasL interaction, because K562 tumor cells do not express Fas and naive NK cells do not express FasL (41; Das, B., unpublished data). Previous reports by us and others have suggested that NK cells use the ATP–osmotic lysis pathway (13, 35) and the granule exocytosis pathway (28, 33). Further studies are likely to reveal which lytic pathway is mediated by the interaction of tumor surface ligand p38.5 with NK cell plasma membrane proteins.

We thank Dr. W. Solomon (State University New York Health Science Center, Brooklyn, NY) for his suggestions and critical reading of the manuscript. We are grateful to Dr. S. Kamholz (SUNY Health Science Center, Brooklyn, NY) for his continuous encouragement and support. The expert word processing by S. Lamy is appreciated.