MICA triggering signal for NK cell tumor lysis is counteracted by HLA-G1-mediated inhibitory signal

Authors

  • Catherine Menier,

    1. Service de Recherche en Hémato-Immunologie, Direction des Sciences du Vivant, Département de Recherche Médicale, Commissariat à l'Energie Atomique, Institut Universitaire d'Hématologie, Hôpital Saint Louis, Paris, France
    Search for more papers by this author
  • Béatrice Riteau,

    1. Service de Recherche en Hémato-Immunologie, Direction des Sciences du Vivant, Département de Recherche Médicale, Commissariat à l'Energie Atomique, Institut Universitaire d'Hématologie, Hôpital Saint Louis, Paris, France
    Search for more papers by this author
  • Edgardo D. Carosella,

    1. Service de Recherche en Hémato-Immunologie, Direction des Sciences du Vivant, Département de Recherche Médicale, Commissariat à l'Energie Atomique, Institut Universitaire d'Hématologie, Hôpital Saint Louis, Paris, France
    Search for more papers by this author
  • Nathalie Rouas-Freiss

    Corresponding author
    1. Service de Recherche en Hémato-Immunologie, Direction des Sciences du Vivant, Département de Recherche Médicale, Commissariat à l'Energie Atomique, Institut Universitaire d'Hématologie, Hôpital Saint Louis, Paris, France
    • Service de Recherche en Hémato-Immunologie, Commissariat à l'Energie Atomique-DSV-DRM, Hôpital Saint Louis, Institut Universitaire d'Hématologie, 1, avenue Claude Vellefaux, 75475 Paris Cedex 10, France
    Search for more papers by this author
    • Fax: +33-0-1-48-03-19-60


Abstract

MICA, a highly glycosylated membrane-anchored cell–surface MHC Class I-related chain, has recently been reported to activate NK cell cytolytic responses in epithelial tumors. Tumor cells may escape from NK lysis by counteracting NK cytotoxicity activating signals with inhibitory ones. Among the molecules that mediate an NK inhibitory signal, HLA-G1, a non-classical MHC Class I antigen, is of particular interest. HLA-G1 is ectopically expressed in various tumors, including melanoma and constitutes the major NK inhibitory ligand in the M8 melanoma cell line when coexpressed with HLA-A, -B, -C and -E molecules. We have evaluated the balance between 2 powerful signals that affect NK cell tumor lysis, one inhibitory and the other one activating, respectively HLA-G1 and MICA. For this purpose, we transfected the M8 melanoma cell line, which spontaneously expresses MICA, with HLA-G1 cDNA, using it as a target for the NKL effector. We carried out cytotoxicity assays, using antibodies that disrupt interactions between the MICA and HLA-G1 ligands and their respective NK effector counterparts, the NKG2D activating and ILT2 inhibitory receptors. Results showed that 1) MICA expressed in the M8 melanoma cell line triggered NK cell tumor lysis and 2) HLA-G1 coexpression mediated the inhibition of NK cytotoxicity by mitigating the MICA activating signal. HLA-G1 expression in a tumor cell line in which MICA is switched on would therefore appear to be a powerful way to turn off NK cells, supporting the emerging idea that the balance between positive and negative NK cytolysis signals critically influences tumor progression. © 2002 Wiley-Liss, Inc.

Abbreviations:

β2-m, β2-microglobulin; FCS, fetal calf serum; HLA, human leukocyte antigen; ILT/LIR; immunoglobulin-like transcript/leukocyte immunoglobulin-like receptor; KAR, killing activating receptor; KIR, killing immunoglobulin-like receptor; mAb, monoclonal antibody; MIC, MHC Class I-related chain; MHC, major histocompatibility complex; NCR, natural cytotoxicity receptor; NK, natural killer; PBS, phosphate-buffered saline.

Natural killer (NK) cells are able to rapidly detect and eliminate HLA Class I-deficient target cells, such as tumor cells, without prior sensitization. Multiple ligand-receptor interactions may be responsible for this triggering of natural cytotoxicity by NK cells.1 These triggering receptors are represented by: (i) HLA Class I-specific activating receptors, including the CD94/NKG2C heterodimer, which is specific for the non-classical MHC Class I HLA-E molecule2 and the killing activating receptor (KAR) family, such as the HLA-C-specific receptor (p50);3 (ii) the family of natural cytotoxicity receptors (NCRs), including NKp46,4, 5 for which hemagglutinin ligands have been recently identified on virus-infected cells,6 NKp30,7 NKp448 and NKp80,9 which recognize still undefined non-MHC ligands expressed by both normal and tumor cells; and (iii) NKG2D, which recognizes the MHC Class I homologs MICA10, 11 and members of a new ligand family, ULBP1-3 molecules that are glycosyl-phosphatidyl inositol-linked glycoproteins.12–15 MICA is a highly glycosylated membrane-anchored cell–surface protein not associated with β2-microglobulin (β2-m) that is conformationally stable in the absence of conventional MHC Class I peptide ligands.16 The solved crystal structure of a soluble fragment of human MICA has confirmed the general configuration of the MICA molecule within an MHC-I-fold.17 Further, recently, the crystal structure of the MICA-NKG2D complex reveals an NKG2D homodimer bound to a MICA monomer in an interaction that is similar to that of MHC Class I protein-T cell receptor complexes.18

MICA belongs to a gene family comprising 7 members, 2 of which encode functional glycoproteins (MICA and MICB), whereas MIC-C through -G are pseudogenes.19, 20 Unlike other non-classical HLA Class I genes, these MIC genes are highly polymorphic.21 MICA, which is of particular interest, originally identified as a stress-inducible molecule found in intestinal epithelium,16 is also selectively switched-on in many epithelial tumors, independent of their type, but not in the corresponding normal cells.22 MICA has been reported to activate NK tumor cytolysis that may potentially promote innate anti-tumor responses.10, 23

One way for tumors to escape from NK cytotoxicity is by delivering inhibitory signals to NK cells capable of counteracting triggering signals.24, 25 Indeed, inhibitory receptors have been described on NK cells, including: (i) killing immunoglobulin-like receptors (KIRs), which are specific for classical HLA-A, -B, -C or for non classical HLA-G molecules; (ii) the HLA-E-specific CD94/NKG2A (CD159a) heterodimer;26, 27 and (iii) ILT-2 (also called LIR-1 or CD85j), a member of the immunoglobulin-like transcript/leucocyte immunoglobulin-like receptor (ILT/LIR) family, which is characterized by broad specificity for HLA Class I molecules including HLA-G.28, 29 Of particular interest is the ectopic expression of HLA-G, which has been observed in various human tumor biopsies, such as, melanoma,30–33 renal carcinoma,34 primary cutaneous lymphoma,35 lung cancer36 and colorectal cancer.37 HLA-G may be distinguished from other HLA Class I molecules by its restricted distribution to immune privileged sites, such as the fetal-maternal interface38 and thymus39 and by the alternative splicing of its primary transcript, which results in at least 7 HLA-G isoforms, of which 4 are membrane-bound (HLA-G1 to -G4) and 3 are soluble proteins (HLA-G5 to -G7).33 Among these 7 HLA-G isoforms, only the full-length HLA-G1 membrane-bound protein is well documented in terms of both expression and function. Indeed, HLA-G1 inhibits the antigen-specific MHC-restricted cytotoxic T lymphocyte (CTL) response40 and the allogeneic proliferative T cell response41 and protects susceptible target cells from NK cytolysis by direct interaction with inhibitory receptors that belong to the KIR family (i.e., KIR2DL4 or CD158d42 and p4943) and to the ILT/LIR family, such as ILT-2.44 We reported recently that the other HLA-G2, -G3 and -G4 membrane-bound isoforms are expressed at the cell–surface as immature glycoproteins exhibiting immunotolerant properties similar to those of HLA-G1.45

Because HLA-G expression has been amply described in extravillous cytotrophoblast cells, where it may contribute to maternal-fetal tolerance,46 we first addressed the question of whether the MICA activating ligand is expressed on trophoblast cells. Second, because both MICA and HLA-G1 molecules have been reported to be switched on in tumors, we investigated the balance between the activating signal delivered by MICA and the inhibitory signal generated by HLA-G1 on NK cell tumor lysis. We therefore developed an HLA Class I-positive tumor cell line model (i.e., a melanoma cell line known as M8, which we have described previously as expressing HLA-A, -B, -C and -E, but not -G molecules47) in which MICA is spontaneously expressed, whereas HLA-G1 coexpression was obtained by HLA-G1 cDNA transfection (M8-HLA-G1). This experimental model allowed us to focus on these 2 important receptor/ligand pairs with respect to regulation of NK cells in the control of tumor growth.

MATERIAL AND METHODS

Cells

M8 is an HLA-A, -B, -C and -E-positive (HLA-A1, -A2, -B12 and -B40/male) but HLA-G-negative melanoma cell line.47 Stable transfectant cells, namely M8-HLA-G1 (transfected with HLA-G1 cDNA) and M8-pcDNA (transfected with the control vector alone), were obtained as described previously.48 Cells were maintained in RPMI 1640 medium supplemented with 10% heat-inactivated fetal calf serum (FCS) (Vysis, Voisins le Bretonneux, France), 2 mM L-glutamine, 1 μg/ml gentamicin and fungizone (Sigma, St. Quentin Fallavier, France) and selected with 100 μg/ml hygromycin B (Sigma). The JEG-3 cell line (American Type Culture Collection, Manassas, VA) was cultured in DMEM medium (Sigma) supplemented with 20% heat-inactivated FCS, 2 mM L-glutamine, 1 μg/ml gentamicin and fungizone. Human cervical tumor HeLa (American Type Culture Collection) was maintained in RPMI 1640 medium supplemented with 10% heat-inactivated FCS. The NK cell line NKL,49 kindly provided by E.H. Weiss (Department of Anthropology and Human Genetics, Munich, Germany), was maintained in RPMI 1640 medium, 10% heat-inactivated human AB serum, 2 mM L-glutamine, 1 μg/ml gentamicin, fungizone and 100 U/ml interleukin-2 (Sigma).

Mononuclear cytotrophoblast cells were isolated from trophoblast tissues obtained from first-trimester terminations of normal pregnancies at 6–12 weeks gestation (local ethics committee approval was obtained) by suction curettage, as previously described.50

Antibodies and flow cytometry analysis

The following monoclonal antibodies (mAbs) were used: 71 and 636, rat IgG anti-MICA culture supernatants (kindly provided by M. Colonna, The Basel Institute for Immunology, Basel, Switzerland);19 87G, mouse IgG2a anti-HLA-G1 (kindly provided by D. Geraghty, Fred Hutchinson Cancer Research, Seattle, WA);51 W6/32, IgG2a anti-HLA Class I heavy chain associated with β2-m (Sigma); Rabbit IgG antisera to the N-terminal region of KIR2DL4 (kindly provided by E. Long and S. Rajagopalan, NIH, Bethesda, MD);42 GHI/75, mouse IgG2b anti-ILT-2 (Immunotech, Marseille, France); Z199, mouse IgG2b anti-NKG2A (Immunotech); HP-3B1, mouse IgG2a anti-CD94 (Immunotech). Flow cytometry assays were carried out as described previously,52 except for the 71 and 636 MAbs, which were stained using a F(ab′)2 goat anti-rat IgG (H+L) conjugated with fluorescein (Beckman Coulter, Villepinte, France) and for the KIR2DL4 rabbit antiserum, which was stained using phycoerythrin-conjugated pig anti-rabbit. Experiments were analyzed using either the Epics XL4 flow cytometer (Beckman Coulter, Brea, CA) or the FACS Scan (Becton Dickinson, Franklin Lakes, NJ).

Cell–surface protein biotinylation, immunoprecipitation and SDS-PAGE

Washed cells (20 × 106) in phosphate-buffered saline (PBS) were biotinylated with sulfo-succinimidyl-6-(biotinamido) hexanoate (Pierce, Rockford, IL) (200 μg/ml) for 4 min at room temperature and the reactions quenched by addition of 50 mM glycine (Sigma) for 5 min. After washes in PBS, cells were incubated for 2 hr at 4°C with either the 71 or W6/32 mAb reacting, respectively with cell–surface MICA and HLA Class I biotinylated proteins. Cells were then lysed in 1 ml lysis buffer (50 mM Tris-HCl pH 7.4 and 0.5% Chaps [Sigma], Complete™ [Roche Diagnostics, Meylan, France]). Cell–surface MICA or HLA Class I biotinylated proteins that complexed with the specific mAb were then precipitated with protein G beads (Roche Diagnostics) at 4°C. After washing in 50 mM Tris-HCl pH 7.4, 150 mM NaCl, 5 mM EDTA and 0.05% Chaps and then in 10 mM Tris-HCl pH 7.4 and 0.05% Chaps, the immunocomplexes were or were not treated with N-glycanase (PNGase F; Roche Diagnostics) overnight at 37°C, as recommended by the manufacturer. Dissociated and dithiothreitol-reduced immune complexes were separated on an 8% polyacrylamide gel and electroblotted onto nitrocellulose membranes. Membranes were blocked in PBS 1× containing 10% non-fat dry milk and 0.05% Tween 20 for 30 min and repeatedly washed in PBS 1×, 0.3% Tween 20 (Sigma). Membranes were then incubated with streptavidin-horseradish peroxidase (Amersham Pharmacia Biotech, Les Ulis, France) diluted 1/1,500 in PBS 1×, 0.3% Tween 20 for 1 hr at 4°C. After washing, membranes were treated with enhanced chemiluminescence reagent (Amersham Pharmacia Biotech) and exposed to X-ray film.

Cytotoxicity assay

The cytolytic activity of the NKL cells used as effectors (E) was assessed in 4-hr 51Cr release assays in which effector cells were mixed with 5 × 10351Cr-labeled targets (T) (100 μCi 51Cr sodium chromate; 1 Ci = 37 Gbq; Amersham Pharmacia Biotech) at various E:T ratios, as previously described.53 The percentage of specific lysis was calculated as follows: % specific lysis = [(cpm experimental − cpm spontaneous release)/(cpm maximum release − cpm spontaneous release)] × 100. Spontaneous release was determined by incubation of labeled target cells in RPMI 1640 medium supplemented with 10% human AB serum. Maximum release was determined by solubilizing target cells in 0.5% Triton X-100. In all experiments, spontaneous release was lower than 10% of maximum release.

Ab blocking assay

In experiments in which mAb were used to block interactions between MICA or HLA-G1 molecules with their respective NK receptor counterpart, target cells were pre-incubated for 15 min at room temperature, respectively with the 71 or 87G mAb. The mAbs were present in the culture medium during the entire assay period. In experiments in which mAb were used to block NK inhibitory receptors, effector NKL cells were pre-incubated for 15 min at room temperature with either anti-ILT-2 or -CD94/NKG2A mAbs before NK cell cytotoxicity assay. Monoclonal Ab toxicity was checked in each assay and always found to be lower than 3%. The antibody-dependent cell cytotoxicity mechanism therefore might not be functioning in our experiments.

Statistical analysis

The statistical significance (p < 0.05) of cytotoxic assays was analyzed using Student's t-test to compare the percentage of specific lysis of M8-HLA-G1 to that of M8-pcDNA. Assays were conducted in triplicate for each experiment.

RESULTS

MICA is not expressed in freshly isolated cytotrophoblast cells and is expressed at a very low level in the JEG-3 placental tumor cell line

To determine whether an activating ligand such as MICA may be expressed in cytotrophoblast cells, flow cytometry experiments were carried out, using 2 different rat antibodies, 71 and 636, both of which specifically react with cell–surface MICA molecules.19 Although freshly isolated cytotrophoblast cells were positively stained by the HLA-G1-specific 87G mAb,51 demonstrating a high level of HLA-G1 cell–surface expression, no MICA molecule could be detected in them, as indicated by the fact that they were not stained by the 71 (Fig. 1) and 636 mAbs (data not shown). Furthermore, the JEG-3 placental tumor cell line, which constitutively expresses HLA-G1 molecules,54 as attested by positive staining with 87G mAb, exhibits only a very low level of MICA expression (Fig. 1). HeLa was used as a MICA-positive control cell line.55

Figure 1.

Cytofluorometry analysis of MICA and HLA-G1 expression on freshly isolated cytotrophoblast cells and tumor cell lines. Cytotrophoblasts cells, the JEG-3 placental tumor cell line and the cervical tumor HeLa cell line were labeled by indirect immunofluorescence with either the rat 71 anti-MICA or the murine 87G anti-HLA-G1 mAb (bold profiles). After washing, cells were stained with fluorescein-conjugated F(ab′)2 goat anti-rat IgG or phycoerythrin-conjugated F(ab′)2 goat anti-mouse IgG, respectively. Controls were the same cells stained with an isotype-matched control antibody (light profiles). The 71 mAb was used as the culture supernatant and the 87G mAb at 2 μg/ml. HeLa cells were used as an MICA-positive control. Freshly isolated cytotrophoblast cells from 4 distinct first-trimester pregnancy terminations were tested, yielding similar results with the 71 and 636 anti-MICA mAbs.

Characterization of MICA protein in a melanoma cCell line

Although we did not detect MICA at the cell surface of cytotrophoblast cells, this activating ligand has been selectively detected in various malignancies,22 which led us to examine the tumor context. Interestingly, MICA shares the tissue-distribution with HLA-G1, which is also expressed by several tumor cell types, such as melanoma.30 For this purpose, we studied the M8 melanoma cell line, described previously as NK target, even though it expresses HLA-A, -B, -C and -E molecules.47, 53 To determine whether MICA was expressed by M8 cells, we conducted cell–surface protein biotinylation, followed by immunoprecipitation with the 71 anti-MICA mAb and Western blot analysis, using streptavidin-peroxidase. Note that neither the 71 nor the 636 mAb works directly in Western blot (data not shown). As shown in Figure 2 (lane 1), a broad band ranging from 60–70 kDa was detected, corresponding to highly glycosylated MICA molecules, consistent with the eight potential N-glycosylation acceptor sites in the α1 α2 α3 domain sequences of MICA.20 To verify this data, we conducted a similar experiment in which N-glycosylations were deleted by N-glycanase (i.e., PNGase F) digestion. In this case, a single compact band of 43 kDa matching the predicted molecular mass of the MICA polypeptide was obtained (Fig. 2, lane 2). These specific bands were absent in the NKL cells used as a MICA-negative control cell line (Fig. 2, lane 5). To positively control these experiments, we immunoprecipitated the HLA Class I molecules present on the M8 cell line by using the pan Class I W6/32 mAb under similar conditions. The results showed bands at 45 and 43 kDa, which respectively correspond to glycosylated and deglycosylated HLA Class I molecules (Fig. 2, lanes 3, 4). Thus, its N-glycosylation pattern distinguished MICA from all conventional and non-classical HLA Class I molecules, which have a single glycosylation site at position 86 in their α1 domain that is not conserved in MICA.20, 56 Altogether, these results demonstrate that MICA molecules are expressed at the cell surface of the M8 melanoma cell line.

Figure 2.

Cell–surface expression of MICA on the M8 melanoma cell line. MICA from lysates of surface-biotinylated M8 cells were immunoprecipitated with the anti-MICA 71 mAb (lanes 1 and 2). Immunocomplexes were untreated (−) or treated (+) with N-glycanase (PNGase F) before SDS-PAGE. Labeled proteins on immunoblots were visualized with streptavidin-horseradish peroxidase. MICA protein is heavily glycosylated and has a polypeptide backbone of 43 kDa. Control lanes (3,4) show HLA Class I immunocomplexes from cell–surface biotinylated M8 cells with the monomorphic anti-HLA Class I W6/32 mAb, revealing bands at 45 kDa (lane 3, PNGase −) and 43 kDa (lane 4, PNGase +). NKL cells (lane 5) were used as MICA-negative control.

Having demonstrated that MICA is spontaneously expressed by the M8 melanoma cell line, our aim was to study the balance between MICA and HLA-G1 in tumor lysis by NK cells. The rare expression of HLA-G protein in tumor cell lines during long-term culture, however, led us to develop an in vitro model, in which the HLA-G1 cDNA or vector alone was transfected into the M8 melanoma cell line, giving rise to 2 stable transfectants, respectively, M8-HLA-G1 and M8-pcDNA. Although M8-pcDNA did not react with the HLA-G1-specific 87G mAb, M8-HLA-G1 was positively stained, showing a high level of HLA-G1 cell–surface expression (Fig. 3). To check MICA expression in both transfectants, flow cytometry experiments were carried out using both the 71 and 636 anti-MICA mAbs. The results showed that both the M8-pcDNA and M8-HLA-G1 transfectants were stained by these mAbs, indicating MICA expression on their cell–surface at similar levels (Fig. 3). The NK cell line, NKL, used as an HLA-G1- and MICA-negative control, was not stained by 87G or by either of the 2 anti-MICA mAbs (Fig. 3). The results of the NKL cell–surface protein biotinylation and MICA immunoprecipitation studies described above (Fig. 2) attest that the NKL cell line does not express MICA.

Figure 3.

Cell–surface expression of MICA and HLA-G1 molecules on M8 transfectants detected by cytofluorometry. M8 cells transfected with either the vector alone (M8-pcDNA) or with HLA-G1 cDNA (M8-HLA-G1) were labeled by indirect immunofluorescence, using the following primary mAb (bold profiles): 71 (rat anti-MICA), 636 (rat anti-MICA) and 87G (murine anti-HLA-G1). After washing, cells were respectively stained with fluorescein-conjugated F(ab′)2 goat anti-rat IgG or phycoerythrin-conjugated F(ab′)2 goat anti-mouse IgG. Controls were the same cells stained with an isotypic control antibody (light profiles). The NKL cell line was used as a MICA-negative control cell line. One of 10 representative experiments is shown.

MICA Expressed on the M8 melanoma cell line mediates a triggering signal for NK lysis

To determine whether MICA expressed by M8 cells influences lysis of this melanoma cell line by NK cells, cytotoxicity assays were carried out, using M8-pcDNA cells as the target in the presence of either the 71 or 636 anti-MICA mAb against the NKG2D-positive NKL line as the effector. Although both mAbs were able to disrupt MICA/NKG2D interaction, we present results here that were obtained with the 71 mAb, which more efficiently blocks such interaction in our target recognition experiments.

As shown in Figure 4, lysis of the M8-pcDNA melanoma cell line by NKL cells decreases significantly in the presence of the 71 anti-MICA mAb at each effector:target ratio tested (p < 0.05). In this regard, the percentage of MICA-mediated activation ranges between 37–46% on the M8 melanoma cell line. Under similar conditions, we checked that the 71 mAb did not modify NKL lysis of the U937 target cell line, which does not express MICA on its cell surface (data not shown).

Figure 4.

Effect of MICA expression on the cytolytic activity of the NKL line. M8 cells were used as targets against the NKL cell line used as the NK effector at the indicated effector:target ratios. This target was preincubated with either 71, or a rat control Ab. Results are expressed as the percentage lysis recorded in a 4-hr 51Cr-release assay. Values represent means of triplicates ± SD. One experiment representative of 4 is shown.

HLA-G1 coexpressed with MICA at the cell surface of M8-HLA-G1 mediates the NK lysis inhibitory signal, counteracting MICA-mediated activation

To analyze the balance between activation of NK lysis mediated by MICA and NK inhibition mediated by HLA-G1, we used M8-HLA-G1 cells as a target for the NKL effector cell line in cytotoxicity assays in which HLA-G1 or MICA were blocked by 87G or 71 mAb, respectively.

Because the NK cell line NKL expresses the HLA-E-specific CD94/NKG2A inhibitory receptor as well as the ILT-2 and KIR2DL4 inhibitory ones, both of which bind HLA-G1 (Fig. 5), we first determined the relative role exerted by HLA-G1 vs. HLA-E on the lytic activity of NKL cells. Ab blocking assays were carried out using anti-ILT-2 or anti-CD94/NKG2A. Unfortunately, the KIR2DL4 rabbit antiserum is not functional in cytotoxicity experiments. Therefore, we could not test the role of this receptor in NKL function. The results presented in Figure 6 showed that the lytic activity of NKL cells toward the HLA-A-, -B-, -C- and -E-positive M8-pcDNA cells is influenced by both the ILT-2 and CD94/NKG2A inhibitory receptor (percentage of inhibition of 21% and 30%, respectively). The NKL effect on M8-HLA-G1 cells is mainly influenced by ILT-2, whose blockage reversed HLA-G1-mediated inhibition. In contrast, the engagement of CD94/NKG2A with HLA-E expressed on M8-HLA-G1 cells, did not affect NKL lysis. This experiment above all shows the predominant inhibitory role played by HLA-G1 vs. HLA-E in our functional analysis.

Figure 5.

Cell–surface expression of KIR2DL4, ILT-2 and CD94/NKG2A on the NKL cells. Cells were labeled by indirect immunofluorescence with either rabbit anti-KIR2DL4, murine anti-ILT-2, or murine anti-CD94/NKG2A antibodies (bold profiles). After washing, cells were respectively stained with phycoerythrin-conjugated F(ab′)2 pig anti-rabbit or goat anti-mouse IgG.

Figure 6.

Influence of ILT-2 and CD94/NKG2A inhibitory receptors on the cytolytic activity of NKL cells. The NKL cells were used as effector cells at a 10:1 effector:target (E:T) ratio against M8 cells transfected with either the vector alone (M8-pcDNA, white) or HLA-G1 cDNA (M8-HLA-G1, black) used as targets. NKL cells were or were not preincubated with either anti-ILT-2 or anti-CD94/NKG2A (10 μg/ml). Results are expressed as the percentage lysis recorded in a 4-hr 51Cr-release assay. Values represent means of triplicates ± SD. Some bars are not visible because of the very small SD. One experiment representative of 4 is shown.

This allowed us to focus on the 2 antagonistic signals, activation by the interaction between NKG2D and MICA vs. inhibition by interactions between ILT-2 and HLA-G1. The Figure 7 shows results obtained from 3 separate experiments, in which: 1) HLA-G1 inhibits lysis of M8-HLA-G1 cell by NKL cells, compared to the M8-pcDNA control cell line. The percentage of HLA-G1-mediated inhibition ranges between 55–80%, 2) by masking HLA-G1 with 87G mAb, which prevents HLA-G1/KIR interaction, the HLA-G1-mediated NKL inhibition could be reversed and 3) more importantly, Ab blockage of MICA present on M8-HLA-G1 cells decreases lysis of these cells by NKL cells, showing that in the presence of HLA-G1, MICA is still capable of triggering NK cytotoxicity, but to a lesser extent than in the absence of HLA-G1 (see M8-pcDNA). Indeed, in 4 independent experiments, the mean value of the percentage of MICA-mediated activation is of 43% for M8-pcDNA, whereas it decreases significantly to 24% for M8-HLA-G1 (p < 0.004).

Figure 7.

Effects of MICA and HLA-G1 expressed on M8 transfectants on the lytic activity of the NKL line. M8 cells transfected with either the vector alone (M8-pcDNA, white) or HLA-G1 cDNA (M8-HLA-G1, black) were used as targets. These targets were or were not preincubated with either 71 or 87G. The NKL cell line was used as the NK effector at a 10:1 effector:target ratio. Results from 3 independent experiments are expressed as the percentage lysis recorded in a 4-hr 51Cr-release assay. Values represent means of triplicates ± SD.

DISCUSSION

In this work, we studied for the first time the balance between 2 powerful signals that affect NK lysis, one inhibitory and the other activating, respectively HLA-G1 and MICA. Particularly interesting is the fact that both molecules are expressed in various tumors, where they may play opposite roles in tumor lysis by NK cells.33, 22

Indeed, MICA, which is selectively expressed on diverse epithelial tumor cells, has been well-characterized as a ligand that triggers NK lysis.10, 22 The MICA receptor is NKG2D, a 42-kDa glycoprotein, whose surface expression requires the presence of a signal-transducing unit known as DAP10.11 The NKG2D receptor has been described in effector cells, such as T (CD8+ α/β and γ/δ) and NK cells.10 Engagement of NKG2D by MICA activates NK and γ/δ cell effector functions10, 23, 57 and strongly increases TCR-dependent cytolytic and cytokine responses by virus-specific CD28/-CD8+ α/β T cells.58

In our study, we used an in vitro target/effector model consisting of the M8 MICA-positive melanoma cell line target for the NK effector (the NKL cell line), which expresses NKG2D.10 Using a newly developed set of MICA-specific monoclonal rat antibodies, known as 71 and 636, we show MICA expression in a non-epithelial tumor (i.e., melanoma) in accordance with the recent study of Pende et al.23 Indeed, by carrying out flow cytometry and cell–surface biotinylation experiments, our results show that, in addition to its described previously presence in epithelial tumors,22 MICA is expressed in the M8 melanoma cell line as highly glycosylated membrane-anchored proteins.

We then studied the effect of MICA on NK function. For this purpose, we carried out cytotoxicity assays in which the MICA function was blocked by the 71 anti-MICA mAb. The used NK cell line NKL no longer expresses detectable levels of CD16 (FcR), due to prolonged in vitro culture (data not shown), as described previously by Robertson et al.49 The results showed that NKL lysis of M8 cells significantly decreased when MICA was masked, demonstrating its NK lysis triggering effects in our model. This data agrees with those of a recent study showing that NKG2D engagement activates the cytolytic responses of various effector cells, including the NKL line, against various carcinoma cell lines that express MICA.10 M8 cells are not fully protected against NKL lysis, however, when MICA activating signal is blocked. In this regard, NCRs, which may be expressed also on NKL cells, could be involved in cytotoxicity in synergy with NKG2D, as recently demonstrated by Pende et al.23 by using NK clones against tumor cell lines. Alternatively, other ligands belonging to the ULBP family may also contribute to mediate activating signals via NKG2D.12–15

The results obtained with the HLA-G1-transfected M8 melanoma cell line, showed that the inhibitory effect exerted by HLA-G1 on NK lysis overrules the activating signal mediated by MICA. Although Bauer et al.10 have described that the HLA-E Class I- mediated inhibition of NK lysis is overcome by the MICA activating signal, recently, Pende et al.23 have shown that the engagement of HLA-B, -C and -E-specific inhibitory receptors down-regulates NK cell triggering induced via NKG2D. Although the M8 melanoma cell line used in our study expresses HLA Class I molecules, such as HLA-E, which send an NK lysis inhibitory signal via its interaction with the CD94/NKG2A inhibitory receptor expressed in the NKL cell line,44 HLA-G1, coexpressed with the other MHC Class I molecules, constitutes the major NK inhibitory ligand, interacting principally with ILT-2 on NKL. This predominant role of HLA-G1 vs. HLA-E is similar to what we have previously described using polyclonal NK cells as effectors.53 Another HLA-G1-specific receptor, namely KIR2DL4, is also expressed by NKL cells. KIR2DL4 was first described as inhibitory receptor that specifically binds to HLA-G142 and has been reported recently to induce IFN-γ production, but not cytotoxicity, in resting NK cells.59 In our study, the function exerted by the engagement of KIR2DL4 on NKL cells remains to be determined.

These results have important implications in tumor immunosurveillance. Indeed, like MICA, which is expressed in tumors, HLA-G has been detected in various tumors, as reported on biopsies obtained from primary and metastatic melanoma,30–32 renal carcinoma,24 breast cancer, bladder cancer, primary cutaneous lymphoma,36 sarcoma and lung cancer patients.33, 36, 37 The known immunotolerant functions of HLA-G strongly suggest a particular role of HLA-G in allowing tumors to escape from immunosurveillance. Various strategies are employed by tumors to avoid immune destruction, such as totally or partially defective MHC Class I molecules,60, 61 absence of costimulation signals,62 secretion of immunosuppressive signals63 and Fas ligand expression.64 Our results indicate that HLA-G1, whose engagement generates an inhibitory signal capable of overcoming the MICA activating signal which mediates anti-tumor responses, may represent a novel way for tumors to escape from immunosurveillance.

The physiological expression of HLA-G in trophoblasts at the feto-maternal interface, which constitutes an immune privileged site,38 led us to investigate whether an activating ligand such as MICA was expressed in this tissue. Moreover, trophoblast cells are similar to tumor cells in terms of proliferative potential and host tissue invasion, specific expression of growth factors, cytokines, oncogenes and antigens.65 Our results show that no MICA expression could be observed on freshly isolated HLA-G-positive cytotrophoblast cells. In addition to the signal mediated by HLA-G, which may inhibit the lytic activity of decidual NK cells bearing HLA-G-specific KIRs such as p49,43 the failure to detect an NK activating signal such as MICA on cytotrophoblasts may contribute to immune privilege for the fetus. In this regard, Sivori et al.66 have shown that NKG2D is not involved in NK cell-mediated cytotoxicity against trophoblast tumor cell lines. Nevertheless, very low MICA expression was detected in the HLA-G-positive JEG-3 placental cell line, probably due the malignant status of this cell line (i.e., choriocarcinoma). This expression might be insufficient to engage NKG2D and thus, activate NK lysis.

Interestingly, in addition to the tumor context, HLA-G and MICA may be expressed during organ transplantation, in which MICA has been described as an allogeneic molecular target for cell-graft rejection, due its high polymorphism,67 whereas HLA-G expression has been associated with better graft acceptance in heart graft transplant patients.68 Another context in which both molecules may be expressed together is the skin of psoriatic patients. In this regard, MICA has been detected in the epidermis of psoriatic skin69 and the MICA 5.1 allele has been proposed as a susceptibility marker for psoriasis in the Korean population.70 It is of note that psoriasis is mediated by T cells71 whose functions may be activated by MICA engagement with NKG2D expressed on their cell surface. In this context, HLA-G expression has recently been described in macrophages infiltrating psoriatic skin, where it may send an inhibitory signal to T cells through ligation with ILT-2 expressed on their cell surface. Such a pathway may contribute to controlling the deleterious effects of T cell infiltrate in psoriasis.72 This may extend the potential implications of our findings to tissue graft tolerance and skin inflammatory diseases.

Finally, our study highlights the finding that HLA-G expression in tumor, even though it coexpresses a molecule that triggers NK lysis, such as MICA, would favor malignant progression by allowing such tumors to escape from NK immunosurveillance. In addition, the balance between positive MICA and negative HLA-G1 signals in tumors may also regulate anti-tumor T cell responses, because receptors for both MICA and HLA-G1 are present on T cells.

Acknowledgements

We are grateful to Dr. M. Colonna, Dr. D. Geraghty, Dr. E. Long and Dr. S. Rajagopalan for providing us with antibodies and to Dr. E. Weiss and Dr. M.J. Robertson, for the NKL cell line. We thank Mrs. S. Bruel for her technical assistance. We thank Dr. D. Jacob for providing us with first trimester pregnancy terminations samples. We also thank Dr. N. Hardy for editing the manuscript.

Ancillary