Switch of HLA-G alternative splicing in a melanoma cell line causes loss of HLA-G1 expression and sensitivity to NK lysis



Considerable information has been accumulated on HLA-G expression in tumor lesions in which HLA-G is viewed as a way to turn off anti-tumoral immunity. Nevertheless, there is little data concerning the mechanisms by which expression and function of HLA-G are regulated in malignant cells. Here, we have addressed these points by studying a melanoma cell line derived from a surgically-removed HLA-G-positive melanoma lesion. We show that HLA-G expression in melanoma cells can be regulated at the mRNA splicing level. Indeed, melanoma cells rapidly switched from cell-surface HLA-G1 to intra-cellular HLA-G2 expression. This mechanism restored tumor sensitivity to NK lysis. Moreover, switch from HLA-G1 to HLA-G2 was strong enough to prevent re-expression of immunoprotective HLA-G1 even following treatments with cytokines and DNA demethylating agent. Modulating HLA-G at the mRNA splicing level would be an efficient way of lifting in vivo HLA-G-mediated tumor immune escape. © 2005 Wiley-Liss, Inc.

Tumor cells frequently interfere with the development and function of immune responses allowing them to escape from destruction by immune system. One of these escape mechanisms is the expression of inhibitory ligands such as the nonclassical human leukocyte antigen (HLA) class I HLA-G. HLA-G inhibits immunocompetent cells through interaction with inhibitory receptors.1 HLA-G is normally absent on healthy tissues except trophoblast,2 thymus3 and cornea.4 Interestingly, both HLA-G transcription and protein expression are up-regulated on various tumors cells, as demonstrated in biopsies from patients with melanoma,5, 6, 7 breast cancer,8 renal carcinoma,9, 10 primary cutaneous lymphoma,11 lung cancer,12 glioma,13 epithelial cutaneous malignant lesions14 and colorectal cancer.15. Moreover, the immunotolerant role of HLA-G has been emphasized by its involvement in the resistance to IFN-based therapies that is observed in some melanoma patients.6, 7

In contrast to these data obtained with surgically removed tumor lesions, the detection of HLA-G on in vitro established tumor cell lines has been mainly unsuccessful (reviewed in 1). To date, only a few tumor cell lines, other than choriocarcinoma cell lines, have been found to express cell-surface HLA-G: 4 renal carcinoma cell lines,9, 16, 17 4 glioma cell lines13 and 9 short-term ovarian carcinoma cell lines.18 This observation may be due to the loss of HLA-G expression during long-term cell culture. Supporting this hypothesis, we show here that a melanoma cell line named Fon, which was derived from an HLA-G-positive surgically removed primitive melanoma lesion,19, 20 lost its constitutive HLA-G surface expression upon long-term in vitro propagation.

The primary transcript of HLA-G is alternatively spliced in 7 mRNA, leading to the expression of 7 HLA-G protein isoforms, 4 of which are membrane-bound (HLA-G1, -G2, -G3 and -G4) and 3 of which are soluble (HLA-G5, -G6 and -G7).21 The extracellular part of the HLA-G1 full-length protein is constituted by 3 domains: α1 (exon2), α2 (exon 3) and α3 (exon 4) noncovalently associated with β2-microglobulin (β2-m). The other membrane-bound HLA-G isoforms have lost 1 or 2 extracellular domains, HLA-G2 (loss of exon 3/α2 domain), HLA-G3 (loss of exons 3-4/α2 and α3 domains) and HLA-G4 (loss of exon 4/α3 domain). These HLA-G isoforms exhibit immunotolerant properties such as inhibition of natural killer (NK) cytolysis and cytotoxic T lymphocyte (CTL) responses.22, 23 The full-length HLA-G1 isoform binds directly to inhibitory receptors, namely immunoglobulin-like transcript (ILT)-2 (CD85j) expressed by lymphoid and myelomonocytic cells,24, 25, 26 and ILT-4 (CD85d) expressed by monocytes, macrophages and dendritic cells.27 Although both receptors have other HLA class I ligands, their highest affinity is for HLA-G.28 KIR2DL4/p49 (CD158d) was also described as an HLA-G-specific receptor expressed by all NK cells.29, 30

Following our previous studies that described HLA-G expression in tumor lesions, such as melanomas,5, 31, 32 we analyzed here the mechanisms by which both expression and function of HLA-G are regulated in malignant cells.


β2m, β2-microglobulin; CTL, cytotoxic T lymphocyte; HLA, human leukocyte antigen; ILT, immunoglobulin-like transcript; MAb, monoclonal antibody; MFI, mean fluorescence intensity; NK, natural killer; RT-PCR, reverse transcription-PCR.

Material and methods

Cell lines and antibodies

Fon (i.e., T1Fon, according to the initial description19) is an HLA-A, -B, -C, -E, -G1 positive melanoma cell line (HLA-A2, -A29, -B4402, -B4403, -Cw802 and -Cw1601/male) derived from a primary melanoma lesion, as previously described.20 M8 is an HLA-A, -B, -C, -E-positive, HLA-G-negative melanoma cell line (HLA-A1, -A2, -B12 and -B40/male), as previously described.31 The NK cell line NKL has been previously described.33 Cells were maintained in RPMI 1640 medium (Sigma-Aldrich, Saint Quentin Fallavier, France) containing 10% FCS (Biological Industries, Beit Haemek, Israel). The Fon−/β2m and Fon−/HLA-G1 cells were obtained after transfection into Fon cells of a vector containing, respectively, human β2m cDNA (kindly provided by E. Weiss, Universitat Munchen, Munich, Germany) or HLA-G1 cDNA, as previously described.22, 23 The Fon−/β2m cells were sorted according to their high β2m surface expression by using a cell sorter (FACS Vantage, Becton Dickinson, France). The M8-pcDNA cells (transfected with the control vector alone) and the M8-HLA-G1, -HLA-G2, -HLA-G3, -HLA-G4 and -HLA-G5 cells (transfected with the vector containing either the HLA-G1, HLA-G2, HLA-G3, HLA-G4 or HLA-G5 cDNA) were obtained as previously described and selected in media containing 100 μg/ml hygromicin B (Sigma Chemical Co.).23, 34 The cells used were routinely tested for and found to be free of mycoplasma.

The following MAbs (monoclonal antibodies) were used: MEM-G/9, a mouse IgG1 specific for native HLA-G1 and HLA-G5 isoforms (Exbio, Prague, Czech Republic);35 4H84, a mouse IgG1 anti-HLA-G free heavy chain (kindly provided by M. Mc Master, University of California, San Francisco, CA);34, 36 5A6G7, a mouse IgG1 specific for the intron 4-retaining part of both soluble HLA-G5 and HLA-G6;37 TP25.99, a mouse IgG1 anti-HLA-A, -B, -C and -E but not anti-HLA-G (kindly provided by S. Ferrone, Roswell Park Cancer Institute, Buffalo, NY);34, 38 B1G6, a mouse IgG2a, anti-β2 microglobulin (Immunotech, France); B8.12.2, a mouse IgG2b anti-HLA-DR (Immunotech); GHI/75, a mouse IgG2b anti-ILT-2 (Pharmingen, San Diego, CA); Z199, a mouse IgG2b anti-NKG2A (Immunotech) and B-5-1-2, a mouse IgG1 anti-α tubulin (Sigma Chemical Co.).

Cell treatments

Cytokine treatment was performed for 48 hr at 37°C using either 1,000 IU/ml recombinant human IFN-β (Tebu, France), 500 IU/ml recombinant human IFN-γ (Tebu), 50 ng/ml recombinant human IL-10 (Tebu), 10 ng/ml recombinant human TGF-β1 (TGF-β) (Tebu) or 2,500 IU/ml recombinant human leukemia inhibitory factor (LIF) (Sigma Chemical Co.). Demethylating treatment was carried out for 72 hr at 37°C with 5-aza-2′-deoxycytidine (5-Aza-dC) (Sigma Chemical Co.) at a final concentration of 5 μM.

Immunohistochemistry analysis

Deparaffinized tissue sections were subjected to epitope retrieval treatment by high temperature in 10 mM sodium citrate buffer (pH 6.0) using a commercial microwave to optimize immunoreactivity. Slides were then rehydrated for 5 min in PBS containing 0.05% saponin and 10 mM HEPES buffer. Endogenous peroxidase activity was quenched by treating sections for 5 min at room temperature with 3% hydrogen peroxide in water. Nonspecific binding was prevented by using 30% human serum for 20 min before staining with the primary MAb for 30 min at room temperature. An isotype-matched antibody was used under similar conditions to control nonspecific staining. Immunostaining was evaluated on tissues using the DAKO EnVision + System, Peroxidase (AEC) (Dako, France), as previously described.34

Western blot analysis

Cells were washed with PBS and treated with lysis buffer [50 mM Tris-HCl, pH 7.4, 0.5% Chaps (Sigma Chemical Co.), containing protease inhibitors (Complete™, Roche Diagnostics, Meylan, France)]. After centrifugation at 15,000g at 4°C for 20 min, supernatants were supplemented with 6× Laemli buffer. All samples were heated for 5 min at 95°C before loading on a 12% SDS-PAGE. Proteins were then electroblotted onto nitrocellulose membranes (Hybond, Amersham, UK) and the membranes blocked by incubation with PBS containing 0.2% Tween 20 and 5% nonfat dry milk. The membranes were then probed with the corresponding MAb overnight at 4°C and washed in PBS containing 0.2% Tween 20. The membranes were subsequently incubated for 30 min at room temperature with goat anti-mouse coupled to horseradish peroxidase (Amersham, Arlington Heights, IL) and washed thoroughly. Signals were detected using enhanced chemiluminescence reagent (ECL, Amersham).

Flow cytometry analysis

Cells were washed in PBS and stained with the corresponding primary MAb in PBS 2% heat-inactivated fetal calf serum for 30 min at 4°C. After wash, cells were subsequently labeled with a F(ab′)2 goat anti-mouse IgG antibody conjugated with phycoerythrin (PE) (Beckman Coulter, Villepinte, France) for 30 min at 4°C. For intracellular staining, cells were treated with 4% paraformaldehyde for 20 min and permeabilized with 0.1% saponin at 4°C. Control aliquots were stained with an isotype-matched antibody to evaluate nonspecific binding to target cells. Cells were analyzed on a flow cytometer EPICS XL (Beckman Coulter) using Expo-32 software (Beckman Coulter, Villepinte, France).

Isolation of RNA, reverse transcription, PCR amplification and Southern blot analysis

Total RNA were isolated from the indicated cell lines with the RNeasy® Mini total ARN kit (Qiagen S.A., Courtaboeuf, France) according to the manufacturer's recommendations. cDNA were prepared from 1–5 μg of total RNA using Ready To Go™ kit and oligo (dT)12–18, according to the manufacturer's recommendations. For HLA-G1 to -G5 transcript detection, PCR amplification and hybridization of the PCR products were performed as previously described.34, 39 The JEG-3 cell line was used as an HLA-G transcription positive control.

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.23 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% fetal calf serum. Maximum release was determined by solubilizing target cells in 0.1 M HCl. In all experiments, spontaneous release was lower than 10% of maximum release.

Blocking assay

In order to block interactions between HLA-G1 or HLA-E molecules with their respective NK receptors, namely, ILT-2 and CD94/NKG2A, respectively, effector NKL cells were preincubated for 15 min at room temperature with either anti-ILT-2 or anti-NKG2A MAb before NK cell cytotoxicity assay. The MAbs were present in the culture medium during the entire assay period. Monoclonal Ab toxicity was checked in each assay and always found to be lower than 3%.


The Fon melanoma cell line derived from HLA-G-positive primary melanoma lesion expresses cell-surface HLA-G1

HLA-G was found to be highly transcribed and translated into protein in the Fon melanoma cell line derived from a surgically-removed HLA-G-positive melanoma lesion (Fig. 1). Among the 7 HLA-G isoforms, only the full-length 39 kDa HLA-G1 protein was detected in the Fon cells by Western blot using the 4H84 MAb, specific for an epitope located in the α1 extracellular domain common to all HLA-G isoforms (Fig. 1c). The membrane-bound HLA-G1 isoform was highly expressed at the cell surface of Fon cells, as assessed by flow cytometry using the MEM-G/9 anti-HLA-G1 monoclonal antibody (Fig. 1d). No HLA-G5 soluble form was detected in Fon cell lysate by Western blot analysis and Fon culture supernatant by using HLA-G5-specific ELISA (data not shown). The JEG-3 HLA-G-positive choriocarcinoma cell line and the HLA-G-negative M8 melanoma cell line that had been transfected either with a mock vector (M8-pcDNA), or with a vector containing HLA-G1 cDNA (M8-HLA-G1), HLA-G2 cDNA (M8-HLA-G2) or HLA-G4 cDNA (M8-HLA-G4) were used as HLA-G-negative or -positive controls.

Figure 1.

Analysis of HLA-G transcription and protein expression in the Fon melanoma cell line derived from a primary melanoma lesion. (a) paraffin-embedded section of primary melanoma lesion from the Fon patient was stained with the anti-HLA-G 4H84 MAb. Original magnification: ×100. (b) RT-PCR detection of HLA-G alternative transcripts using pan HLA-G primers G.257F (exon2) and G.1004R (exon 5/exon 6): HLA-G5, 886 bp; HLA-G1, 764 bp; HLA-G2, 488 bp; HLA-G4, 491 bp; and HLA-G3, 215 bp. HLA-G transcripts were revealed by hybridization with exon 2-specific G.R probe. M8-HLA-G transfectants and JEG-3 cell line were included as HLA-G-positive controls. Absence of contaminant DNA was controlled by concomitant amplification of the PCR mixture without a template (H2O). (c) whole-cell protein extracts analyzed by Western blot using HLA-G-specific 4H84 MAb. The band corresponding to HLA-G1 (39 kDa) is indicated by an arrow. (d) flow cytometry analysis using MEM-G/9 anti-HLA-G1 MAb depicted as filled histogram (mean fluorescence intensity, MFI) and open histogram (mouse isotypic IgG1 control).

HLA-G1 expression is lost during in vitro long-term cell culture

In the course of the in vitro propagation of the Fon melanoma cell line, we observed a loss of HLA-G1 cell surface expression, as assessed by flow cytometry using the MEM-G/9 MAb. Indeed, the primary culture cell line exhibited a high level of HLA-G1 cell-surface expression, which was maintained until passage 40 (referred to as Fon+), and started to wane from passages 66 to 68 (referred to as Fon+/−) to rapidly become completely negative at passage 70 (referred to as Fon) (Fig. 2). These results clearly show that HLA-G1 surface expression is not progressively but suddenly lost: between passages 66 to 70, all HLA-G1-positive Fon cells divided into HLA-G1-negative cells, as shown by the presence of 2 very distinct populations of Fon cells during the 4 passages transition period. A bank of cryopreserved Fon+ cells was established at passage 40, which constitutes our reference of HLA-G1-positive expression.

Figure 2.

HLA-G1 cell-surface expression is lost during in vitro long-term cell culture. Cell surface expression of HLA-G1 by the Fon melanoma cell line after 40 (Fon+p40), 66 (Fon+/−p66), 68 (Fon+/−p68) and 70 (Fonp70) passages was analyzed by flow cytometry using the MEM-G/9 MAb.

HLA-G2 mRNA and protein expression is switched on in HLA-G1-negative Fon cells

In order to define whether loss of HLA-G1 protein expression by Fon cells was regulated at the transcriptional level, we studied the alternative splicing profile of HLA-G in Fon cells and compared it to that of Fon+ cells. Total RNA was extracted from both Fon+ and Fon cells, as well as from HLA-G-positive controls M8-HLA-G1, -G2, -G4 and JEG-3 cells. To visualize the alternative mRNA forms, HLA-G transcripts were analyzed by reverse transcription-PCR (RT-PCR) and Southern blot hybridization, using procedures validated at the 13th International HLA Workshop.34 As shown in Figure 3a, the HLA-G alternative splicing profile of Fon+ cells is similar to that of JEG-3 cells with a predominance of HLA-G1/G5 transcripts and lower levels of HLA-G2/G4 transcripts. Interestingly, Fon cells exhibited a totally different profile with an absence of HLA-G1 transcript and upregulation of HLA-G2/G4 transcripts. In order to precisely define the HLA-G transcript expressed by Fon cells, we used probes specific for HLA-G2 (i.e., G.647 exon 4-specific probe) and HLA-G4 (i.e., G.526 exon 3-specific probe) mRNA. Results showed that Fon cells expressed only HLA-G2 mRNA (without exon 3) and not HLA-G4 mRNA (Fig. 3a). By carrying out RT-PCR with HLA-G5-specific primers and probe, we found that both Fon+ and Fon cells expressed HLA-G5 mRNA (Fig. 3b).

Figure 3.

Analysis of HLA-G transcription in Fon+ and Fon cells. (a) RT-PCR detection of HLA-G alternative transcripts resulting from amplification with pan HLA-G primers G.257F (exon2) and G.1004R (exon 5/exon 6): HLA-G5, 886 bp; HLA-G1, 764 bp; HLA-G2, 488 bp; HLA-G4, 491 bp; and HLA-G3, 215 bp. HLA-G transcripts were revealed by hybridization with either the exon 2-specific G.R probe, or the exon 4-specific G.647 probe or the G.526 exon 3-specific probe. (b) specific amplification of HLA-G5 transcripts corresponding to the soluble HLA-G5 protein (489 bp) with primers G.526 (exon 3) and G.i4b (intron 4). HLA-G5 transcripts were revealed by hybridization with an intron 4-specific G.I4F probe. M8-HLA-G transfectants and JEG-3 cell line were included as HLA-G-positive controls. Absence of contaminant DNA was controlled by concomitant amplification of the PCR mixture without a template (H2O). β-actin PCR products coamplified in the same reaction are shown.

We investigated whether HLA-G2 and HLA-G5 transcripts were translated into proteins in Fon cells. Western blot analysis was performed using the 4H84 MAb (Fig. 4a), and the 5A6G7 MAb that is specific for the intron4-encoded part of soluble HLA-G5 and -G6 proteins (Fig. 4b).37 M8-HLA-G1, -G2, -G3, -G4 and -G5-transfected cells were used as positive controls for detection of corresponding HLA-G isoforms. Results show that in Fon+ cells at passage40, only the HLA-G1 full length membrane-bound isoform was detected but not any of the other HLA-G proteins (Fig. 4). In agreement with the HLA-G mRNA profile, Fon cells at passage p70 no longer expressed HLA-G1 but expressed HLA-G2 (Fig. 4a). When we analyzed Fon cells at the intermediate passage 64 (Fon+/−), both HLA-G1 and HLA-G2 proteins could be detected concomitantly (Fig. 4a). However, the HLA-G2 protein was not detected at the cell-surface of Fon cells as these cells were not stained with the 4H84 MAb in flow cytometry and in cell-surface protein biotinylation experiments (data not shown). No HLA-G5 protein was even detected in the Fon cells at any stage of culture, as assessed by Western blot analysis (Fig. 4b), or HLA-G5-specific ELISA (data not shown).

Figure 4.

Analysis of HLA-G protein expression in Fon+ and Fon cells. HLA-G isoforms were analyzed in Fon cells at the 40th passage (Fon+), the 64th passage (Fon+/−) and the 70th passage (Fon) in vitro. Same amounts of whole cell protein extracts were analyzed by Western blot using either pan HLA-G-specific 4H84 MAb (a) or HLA-G5-specific 5A6G7 MAb (b). M8-HLA-G1, M8-HLA-G2, M8-HLA-G3, M8-HLA-G4 and M8-HLA-G5 transfected cells were used as positive controls for the corresponding HLA-G isoform. Bands corresponding to HLA-G1, HLA-G2, HLA-G3, HLA-G4 and HLA-G5 are indicated by arrows.

HLA-G1 expression on melanoma cell confers them protection from NK lysis through interaction with ILT-2

We analyzed the biological relevance of HLA-G1 expression on Fon melanoma cells towards NK lysis. In addition to Fon+ cells, we used M8-pcDNA and M8-HLA-G1 melanoma cells as HLA-G1-negative and -positive control targets, respectively. The NKL cell line was used as NK effector expressing the HLA-G-binding inhibitory receptor ILT-228, 33 as well as the HLA-E-specific CD94/NKG2A inhibitory receptor. Since M8-pcDNA, M8-HLA-G1 and Fon+ cells express HLA-E (data not shown),40 we evaluated the relative impact of HLA-G1 and HLA-E on NK lytic activity. For this purpose, we carried out antibody blocking assays aimed at disrupting interactions of HLA-G1 and HLA-E with their corresponding receptors.

The results presented in Figure 5 show that the expression of HLA-G1 on Fon+ cells led to inhibition of NKL lysis through interaction with ILT-2. Indeed, masking ILT-2 reversed HLA-G1-mediated inhibition whereas engagement of CD94/NKG2A with HLA-E expressed on Fon+ cells, weakly affected NKL lysis. These data were supported by those obtained with the M8-HLA-G1 cells in which HLA-G1 expression protected them from NKL lysis (Fig. 5). By contrast, the lytic activity of NKL cells toward the M8-pCDNA cells was mainly influenced by engagement of the CD94/NKG2A inhibitory receptor with HLA-E molecules expressed by M8 cells, as we previously described.33

Figure 5.

HLA-G1 protects melanoma cells from NK lysis through interaction with ILT-2. The NKL cell line was used at a 10:1 effector:target (E:T) ratio against the Fon melanoma cell line and M8 cells transfected with either the vector alone (M8-pcDNA) or the HLA-G1 cDNA (M8-HLA-G1), used as targets. NKL cells were or were not (No MAb) preincubated with either anti-ILT-2 or anti-NKG2A (10 μg/ml). The HLA class I molecules expressed by the target cells are indicated at the bottom. Results are expressed as the percentage lysis recorded in a 4 hr 51Cr-release assay. Values represent means of triplicates. The standard deviation of the mean of the triplicates was below 5%. One experiment representative of 3 is shown.

HLA-G splicing switch rendered melanoma cells susceptible to NK lysis

We analyzed the impact of the HLA-G splicing switch on HLA-G-mediated protection from NK lysis. Results showed that while Fon+ cells were protected from NKL lysis, Fon cells, in which cell-surface HLA-G isoforms as well as HLA-A, -B,-C and -E molecules were absent, became very good NK targets (Fig. 6).

Figure 6.

Fon melanoma cells are no longer protected from NKL cell-mediated cytolysis. Both HLA-G1-positive (Fon+) and HLA-G1-negative (Fon) Fon cells, as well as Fon cells transfected with β2m cDNA (Fon/β2m), were used as targets for the NK effector NKL cell line, at various effector:target (E:T) ratio in a cytotoxicity assay. Results are expressed as the percentage lysis recorded in a 4 hr 51Cr-release assay. Values are means of triplicates. The standard deviation of the mean of the triplicates was below 5%. One experiment representative of 5 is shown.

Fon cells express HLA class I molecules intracellularly, as detected in Fon cell lysate by Western blot using anti-HLA-B, -C and anti-HLA-E antibodies (data not shown). However, no intracellular staining was observed using anti-β2m in Fon cells (Fig. 7a), which consequently lacked surface expression of all HLA class I molecules (Fig. 7c). We analyzed whether the reexpression of β2m and that of HLA-A, -B, -C, and -E molecules on the cell-surface of Fon cells may affect their sensitivity to NK lysis. For this purpose, β2m cDNA was transfected into the Fon cells giving rise to Fon−/β2m transfected cells. Results showed that i) β2m reexpression in Fon cells was sufficient to reinduce surface expression of HLA-A, -B, -C and -E molecules (Fig. 7c). As expected, HLA-G1 protein was not detected in Fon−/β2m (Fig. 7b,c) since HLA-G1 mRNA is no longer present into Fon cells; ii) re-expression of HLA-A, -B, -C and -E molecules did not fully protect Fon cells from NK lysis. Indeed, Fon−/β2m cells were less sensitive to NKL lysis compared to Fon cells, but they were not protected to the same extent as Fon+ cells (Fig. 6).

Figure 7.

Fon are β2m-deficient cells. (a) both Fon+ and Fon cells were analyzed by flow cytometry for β2-microglobulin expression. Cells were permeabilized with saponin for intracellular staining using either anti-β2m or anti-α tubulin (filled histograms). Open histograms correspond to the same cells stained with mouse isotype-matched control antibody. Tubulin was used as a positive protein control for intracellular staining. (b) Fon cells were transfected with either β2m cDNA or HLA-G1 cDNA, giving rise to the Fon−/β2m and Fon−/HLA-G1 transfected cells, respectively. These cells were analyzed by Western blot for HLA-G expression by using the 4H84 MAb. (c) cell-surface expression of HLA-G1, HLA-A, -B, -C, -E and β2-microglobulin was analyzed on Fon+, Fon, Fon−/β2m and Fon−/HLA-G1 cells. Antibodies used in flow cytometry were MEM-G/9 (anti-HLA-G1), TP 25.99 (anti-HLA-A, -B, -C and -E), B1G6 (anti-β2-microglobulin) and isotype-matched control Ab.

Moreover, HLA-G1 cDNA construct was transfected into Fon cells leading to the Fon−/HLA-G1 cells. These cells synthesized HLA-G1 protein (Fig. 7b) but were devoid of HLA-G1 and HLA-A, -B, -C and -E surface expression (Fig. 7c). This data highlights the requirement of β2m association with HLA class I heavy chain, including HLA-G1, to achieve stable surface expression.

HLA-G1 cell surface expression on Fon+ cells could be enhanced upon treatments known to up-regulate HLA-G

The expression of HLA-G in tumor cells is likely to be dependent upon microenvironmental factors that favor and maintain this expression in vivo. Among these factors, cytokines locally produced in situ, such as IFN, IL-10, TGF-β and LIF,41 have been described as enhancing HLA-G expression in various cell types.42, 43, 44 To investigate whether HLA-G1 expression could be up-regulated in Fon melanoma cells, Fon+ cells were treated with these cytokines. Cell-surface expression level of HLA-G1 molecules was then evaluated by flow cytometry. Concomitantly, cell-surface expression levels of the other HLA class I and class II molecules were analyzed. Results in Table I show that while IL-10, TGF-β and LIF had no effect on HLA-G1 cell-surface expression, IFN-β and IFN-γ led to a more than 2-fold increase. Notably, while IFN-γ enhanced the expression of HLA-A, -B, -C, -E and -G1 as well as it induced HLA class II expression, IFN-β had a selective effect on HLA-G1 expression.

Table 1. Enhancement of HLA-G1 Expression by IFN and DNA Demethylating Treatment in the FON Melanoma Cell Line
 Relative enhancement of HLA class I and HLA class II cell-surface expression (fold)1
  • 1

    HLA-G1, HLA-A, -B, -C, -E, and HLA-DR cell-surface expression was analyzed on Fon melanoma cells after treatment with the indicated cytokines and DNA demethylating agent by flow cytometry using MEM-G/9, TP25.99, and B8.12.2 MAbs, as indicated in Materials and Methods. The relative induction of HLA class I and class II cell-surface expression is presented as a ratio between mean fluorescence intensity (MFI) in treated cells and MFI in untreated cells.


In addition, we recently described that HLA-G expression may be upregulated through epigenetic mechanisms, such as DNA demethylation.39 Thus, we treated the Fon+ melanoma cells with the demethylating agent 5-Aza-dC. Results showed a very high enhancement of HLA-G1 cell-surface expression on melanoma cells that was specific for this HLA class I molecule since no effect was observed on the cell-surface expression levels of HLA, -A, -B, -C and -E molecules (Table I).

HLA-G1 mRNA and protein could not be reinduced in Fon cells upon treatment with IFN and 5-Aza-dC

We investigated whether Fon cells that had completely lost HLA-G1 expression may reexpress this molecule upon treatments with IFN-β, -γ and 5-Aza-dC, known to upregulate HLA-G1 expression in Fon+, as shown above (Table I). Results showed that none of these treatments could reinduce HLA-G1 mRNA (data not shown) and protein expression (Fig. 8). As exemplified with IFN-γ treatment, no effect was observed on Fon cells, in which HLA-G1 cell-surface expression could not be reinduced. By contrast, as mentioned above in Table I, IFN-γ treatment enhanced HLA-A, -B, -C, -E, -G1 and -DR cell-surface expression on Fon+ cells. Efficiency of IFN-γ treatment in Fon cells was attested by the upregulation of HLA class II (-DR) expression (Fig. 8).

Figure 8.

IFN-γ treatment of Fon+ and Fon cells. Cell-surface expression of HLA-G1, HLA-A, -B, -C, -E and HLA-DR was analyzed on Fon+ and Fon cells after (+) or in absence of (−) treatment with IFN-γ (500 IU/ml, 48 hr). Antibodies used in flow cytometry were MEM-G/9 (anti-HLA-G1), TP 25.99 (anti-HLA-A, -B, -C and -E), B8.12.2 (anti-HLA-DR) and isotype-matched control Ab.


More than 200 cell lines have been studied to date for their expression of HLA-G. These investigations mainly focused on HLA-G1 cell-surface expression by flow cytometry analysis. With the exception of choriocarcinoma cell lines, only a few of the cell lines analyzed within each type of tumor have been found to express HLA-G. These data are in opposition to those obtained ex vivo where HLA-G has been widely detected in surgically removed tumor lesions (reviewed in 1). We analyzed a melanoma cell line called Fon that was derived from an HLA-G-positive melanoma biopsy and that constitutively expresses cell-surface HLA-G1 molecules at a high level.

The discrepancy between the in vitro data and the results obtained by testing surgically removed lesions may reflect the lack in tissue culture media of factors which induce and maintain HLA-G expression in vivo in tumor microenvironment. Accordingly, we clearly showed in the present study that Fon melanoma cell line lost its cell-surface expression of HLA-G1 upon in vitro long-term culture. This data is in agreement with previous reports showing a downregulation of HLA-G1 cell-surface expression upon in vitro culture of renal carcinoma cell lines16 and short-term ovarian carcinoma cell lines.18

Interestingly, we found that loss of HLA-G1 mRNA in melanoma cells is associated with a selective switch in HLA-G alternative splicing toward HLA-G2 expression. This process is of particular interest since it abrogates the role of HLA-G1 in tumor escape from NK lysis. Although the molecular mechanisms underlying such a phenomenon have to be elucidated, the rapidity of the switch in few cell-culture passages is consistent with an active process involving specific factor(s). In this regard, steroid hormones have been previously described as controlling alternative premessenger-RNA splicing.45, 46 For instance, the alternatively spliced isoforms of the CD44 gene are regulated in breast tumor cells in which each isoform may play a particular role in time and space.47 We may hypothesize that such specific factor(s) affecting HLA-G splicing may be upregulated according to the physiopathological context and related microenvironment.

Factors involved in in vivo activation of HLA-G transcription and protein expression in tumor cells are not well defined. When malignant transformation occurs, tumor cells will be exposed to a variety of cytokines such as IFN, of stress factors such as hypoxia, heat shock, nutrient deprivation or increased acidity from the tumor microenvironment, and undergo epigenetic changes such as DNA hypomethylation and histone acetylation which activate the HLA-G gene.39 The appearance of HLA-G on tumor cells may then downregulate cellular immune responses and increase tumor survival. In agreement with these data, treatment of HLA-G-positive Fon melanoma cells with IFN (β and γ) or with 5-aza-dC demethylating agent enhances HLA-G1 cell-surface expression. Similar results were reported for other tumor cell lines.13, 39, 48, 49, 50, 51 Such an observation deserves particular attention in the context of IFN-based immunotherapy or anti-tumor chemotherapy using demethylating drugs such as decitabine in melanoma patients whose immune system may be inhibited by the boosted expression of HLA-G1.6, 7 Interestingly, in our study, the HLA-G switch from HLA-G1 to HLA-G2 was strong enough to prevent reexpression of HLA-G1 mRNA and protein even following treatment with cytokines and DNA demethylating agent.

The HLA-G switch process did not affect HLA class I, -A, -B, -C and -E expression in Fon melanoma cells. However, since Fon cells are β2m deficient, stable cell-surface expression of these HLA class I molecules could not be achieved. Similarly, HLA-G2 protein could not be detected at the cell surface of Fon cells. We may hypothesize that HLA-G2, whose extracellular structure consists of α1 linked to α3 domains, requires association with β2m for stable cell-surface expression, as do the other HLA class I molecules via the α3 domain. Alternatively, HLA-G2 may require association with folded HLA-A, -B, -C and/or -E/β2m molecules for transport to the cell surface, as previously described for other proteins such as the mouse nonclassical Qa-1 molecule, which forms heterodimers with mouse classical H-2Ld molecules.52 We have previously described HLA-G2 cell-surface expression on HLA-G2-transfected melanoma cells that properly expressed both β2-m and HLA class I molecules.23 Notwithstanding β2m deficiency, defects in the peptide loading machinery of HLA class I molecules may cause defective HLA class I expression in malignant cells (Seliger, 2002 number 2745)(Chang, 2004 number 4450). Such defects are due to abnormalities in the expression and/or function of different components of HLA class I processing and presentation pathway such as TAP or other components of the antigen processing machinery. In our study, β2m cDNA transfection into Fon cells was sufficient to restore HLA-A, -B, -C and -E surface expression, strongly suggesting that only β2m and no other molecules involved in HLA class I processing and transport to the cell surface, were deficient in Fon cells.

By studying tumor cells in which HLA-G is endogenously expressed, we confirm that HLA-G1, whose engagement generates inhibitory signals in various immune cells such as NK cells, represents a way for tumor cells to escape from immunosurveillance. When melanoma cells did no longer transcribe HLA-G1, due to the switch in HLA-G splicing, they became very sensitive to NK lysis. The involvement of HLA-G1 surface expression in efficient tumor escape from NK lysis was particularly reinforced by studying Fon−/β2m cells. Indeed, in absence of HLA-G1, Fon−/β2m cells were not fully protected from NKL lysis. Although HLA-A, -B, -C and -E molecules also inhibited NKL lysis, complete NK inhibition was only achieved when HLA-G1 was expressed at the cell surface. Finally, the reexpression of both β2-microglobulin and HLA-G1 into Fon cells would confirm the protective role of HLA-G1 surface expression from NK lysis. Unfortunately, despite several assays (i.e. cotransfection of β2m and HLA-G1 gene constructs together into Fon cells, cotransfection of HLA-G1 gene construct into Fon−/β2m transfected cells, cotransfection of β2m gene construct into Fon−/HLA-G1 transfected cells), double transfected cells expressing both β2-microglobulin and HLA-G1 could not be obtained. Of note, both Fon+ and Fon cells expressed the NK activating ligand MHC class I-like (MICA) at their cell surface at similar level (data not shown). We recently showed that MICA-triggering signal for NK cell tumor lysis can be counteracted by HLA-G1-mediated inhibitory signal, leading to the escape of HLA-G1+ MICA+ melanoma cells from NK lysis.33 Accordingly, the Fon+ melanoma cell line that evades from NK lysis, expressed both MICA and HLA-G1 molecules at levels comparable to that of M8-HLA-G1 cells (data not shown). The inhibitory role of HLA-G deserves particular attention in the progression and outcome of HLA-G-positive tumors as well as in the design of cancer immunotherapy and chemotherapy. It appears important to modulate the expression of HLA-G in such tumors in order to prevent its negative impact that paralyzes the patient's immunocompetent cells.


We are grateful to Dr. S. Ferrone and Dr. M. McMaster for providing us with antibodies and to Dr. E. Weiss and N. Kotzias for providing us with human β2m cDNA construct. We thank Dr. J. LeMaoult for helpful discussions.