Expression of adhesion molecules and ligands for activating and costimulatory receptors involved in cell-mediated cytotoxicity in a large panel of human melanoma cell lines

Abstract

Knowledge of the interactions between MHC-unrestricted cytotoxic effector cells and solid tumour cells is essential for introducing more effective NK cell-based immunotherapy protocols into clinical practise. Here, to begin to obtain an overview of the possible universe of molecules that could be involved in the interactions between immune effector cells and melanoma, we analyse the surface expression of adhesion and costimulatory molecules and of ligands for NK-activating receptors on a large panel of cell lines from the “European Searchable Tumour Cell Line and Data Bank” (ESTDAB, http://www.ebi.ac.uk/ipd/estdab/) and discuss their potential role in the immune response against this tumour. We show that most melanoma cell lines express not only adhesion molecules that are likely to favour their interaction with cells of the immune system, but also their interaction with endothelial cells potentially increasing their invasiveness and metastatic capacity. A high percentage of melanoma cell lines also express ligands for the NK-activating receptor NKG2D; whereas, the majority express MICA/B molecules, ULBP expression, however, was rarely found. In addition to these molecules, we also found that CD155 (poliovirus receptor, PVR) is expressed by the majority of melanoma cell lines, whereas CD112 (Nectin-2) expression was rare. These molecules are DNAM-1 ligands, a costimulatory molecule involved in NK cell-mediated cytotoxicity and cytokine production that also mediates costimulatory signals for triggering naïve T cell differentiation. The phenotypical characterisation of adhesion molecules and ligands for receptors involved in cell cytotoxicity on a large series of melanoma cell lines will contribute to the identification of markers useful for the development of new immunotherapy strategies.

Introduction

Melanoma is considered the most lethal form of skin cancer, because it tends to spread early and rapidly progresses to a disseminated metastatic stage. Because melanoma is generally poorly responsive to chemotherapy, research has focussed on developing new cell-based immunotherapies for this cancer. Thus, identification of parameters relevant for cell-mediated cytotoxicity against melanoma may bring us a step closer to an effective treatment of these patients.

NK cells and CD8 T cells are major players in cell-mediated cytotoxicity against tumours and frequently share adhesion and activating receptors. In the earlier phases of lymphocyte activation, the formation of a conjugate between cytotoxic cells and tumour cells requires direct cell-to-cell interaction and the formation of an immunological synapse providing a microenvironment for the release of cytotoxic granules [8, 13, 14]. During the effector phase, cytotoxicity can be modulated by inhibitory signals, such as those mediated by MHC class I-specific inhibitory receptors (e.g. KIR, ILT, CD94/NK2A) initially described in NK cells and afterwards found in subsets of T cells [22, 75, 79].

Tumours evade T cell recognition by many different mechanisms, in particular down-regulation of HLA expression [5, 51, 69], low expression of co-stimulatory molecules [30, 45] and loss of cell adhesion molecules. In addition, several types of dysfunctional antigen-specific T cells have been described both in vivo [4, 55] and after in vitro expansion of CD8 T cells [20]. In human melanoma cell lines, several alterations of HLA class I molecules have been described by Garrido et al. [5, 51, 69]. Total or partial loss of MHC class I molecules is a frequent event in human solid tumours of different origins, and constitutes an important hurdle for T cell based immunotherapy [5, 52]. In contrast to T cells, NK cells recognise melanoma cells with low MHC expression more efficiently [61], and their activation depends on a complex balance between inhibitory and activating signals [12, 75, 80]. Accumulating evidences support a crucial contribution of NK cells to the immunosurveillance of tumours (for review see Waldhauer and Steinle [87]). In the absence of inhibitory signals (e.g. in the case of MHC class I-negative cells), tumour cells can be susceptible to NK-mediated lysis by recognizing ligands for activating receptors [14, 15, 53].

Although MHC class I loss makes tumour cells more susceptible to NK cell-mediated cytotoxicity, they have also developed mechanisms to avoid such killing [12, 19, 53]. Thus, downregulation or shedding of ligands for NK cell-activating receptors has been reported in tumour cells [38, 71, 85, 86]. Therefore, the study of the basis of NK cell cytotoxicity of solid tumour cells is of interest for assessing the possibility of introducing NK cell-based immunotherapy protocols [50].

Here, we describe the analysis of adhesion and costimulatory molecules and the expression of ligands for NK-activating receptors on a large panel of melanoma cell lines and discuss their role in the immune response against melanoma. Although it can be argued whether cell lines faithfully reflect the properties of the cells of the originating tumour, nonetheless, regarding their expression of NK inhibiting and activating ligands they offer a range of phenotypes also found on tumour cells in situ and thus provide valuable models for detailed investigation of tumour cell–NK cell interactions not possible when using fresh tumour samples. The phenotype of each individual melanoma cell line was established as part of the European project “European Searchable Tumour Cell Line and Data Bank” (ESTDAB) and can be searched on-line (see http://www.ebi.ac.uk/ipd/estdab/). The 5th Framework Program project “Outcome and Impact of Specific Treatment in European Research on Melanoma” (OISTER) also aimed to characterise melanoma cell lines together with their relevant clinical information. The expression of HLA class I antigens, ligands for NK cell inhibitory receptors, on this panel of cell lines has been previously reported [51]. The analysis of the expression of ligands for adhesion and activating NK receptors involved in cell recognition and killing will help to clarify the mechanisms involved in melanoma susceptibility or resistance to NK cells.

Analysis of adhesion molecules expressed in melanoma cell lines

Intercellular interactions through cell surface molecules on effector cells and their ligands on tumour cells are required to form stable conjugates to construct the “immunological synapse” essential for cell activation. Normal melanocytes express few adhesion molecules, but melanoma cells show an increased expression of these molecules. As melanoma progresses the repertoire of adhesion molecules changes and the acquisition of cell adhesion molecules during the process of tumour progression is suggested to contribute to the development of metastasis in melanoma [35, 56]. Using the ESTDAB cell bank, we have analysed the expression of adhesion molecules of the immunoglobulin superfamily (CD54 and CD58) and integrin superfamily (CD49d and CD11b) and also CD56 and CD57 molecules that have previously been described as mediators of cellular adhesion.

As shown in Fig. 1, most melanoma cell lines (95%) expressed CD54 (ICAM-1) at different levels of intensity. The CD54/LFA-1 (CD11a) interaction has been shown to play a crucial role in enhancing anti-tumour immune responses by its participation in effector–target conjugate formation. The expression of CD54 is required for lysis of melanoma cell lines by specific CTLs [42] and CD54^high melanoma cells are more susceptible to NK cell-cytotoxicity than those with low CD54 expression [66], suggesting that CD54 expression on melanoma should associate with a better prognosis. However, high levels of CD54 are frequently found in metastatic or invasive melanoma cells [25] and CD54 expression correlates to a worse prognosis in primary melanomas [58]. Furthermore, it has been demonstrated that low expression of CD54 by melanoma cells associates with a longer overall survival following immunotherapy [67]. This apparent discrepancy may be explained by the fact that shedding of CD54 by melanoma cells has been described [35], and soluble CD54 can bind to its ligand on effector cells blocking effector–target interactions, contributing to tumour escape from immunity [2, 3, 10]. Furthermore, the analysis of CD54 polymorphism has shown an association of the R241 allele with high levels of soluble CD54 and with high relative risk of melanoma, supporting the possible significance of soluble CD54 in melanoma progression [83]. On the other hand, CD54 is also involved in the adhesion of tumour cells to the vascular epithelium, and it has been suggested that these interactions may promote the development of metastases [35].

Fig. 1.

Percentage of melanoma cell lines expressing the adhesion molecule indicated. Surface expression was determined by flow cytometry using a FACScalibur and specific mAbs. Intensity of staining of CD54, CD58, CD11b, CD49d, CD56 and CD57 (n = 124). Normalised scores were calculated by dividing the mean fluorescent channel (MFC) by its negative control. Negative <1.5; weakly positive (±) >1.5; positive (+) >2, strongly positive (++) >10 and very strong (+++) >100

CD58 (LFA-3) is a ligand of the cell surface receptor CD2, expressed on NK and T lymphocytes. In our study, all melanoma cell lines analysed expressed CD58 at variable intensity (Fig. 1). An increased CD58 expression on melanoma cells can modify their lysis susceptibility by melanoma-specific cytotoxic T lymphocyte (CTL) clones and co-stimulate cytokine production by these cells. It has been shown that a minimal expression of CD58 is required for optimal activation and full functionality of melanoma-specific CTL clones [42]. In particular, increasing CD58 density on melanoma cells enhances cytokine production by melanoma-specific CTL clones, indicating that CD58 expression is critical for the efficiency of specific immune reactions against melanoma cells [43]. In contrast to CD54, the expression of CD58 is not induced by cytokines such as TNF-α or IFN-γ [3].

As previously reported by Altomonte et al. [3], CD11a was not detected on the panel of melanoma cell lines analysed in this study (data not shown). However, our results showed that 37% of melanoma cell lines expressed detectable CD11b levels (Fig. 1). Although its role in melanoma cell interactions with lymphocytes or monocytes has not been defined so far, CD11b mediates cellular adhesion to endothelial cells through CD54 and its expression on melanoma cells may be relevant to their metastatic potential.

High levels of CD56 expression were observed on 71% of melanoma cell lines (Fig. 1). It is a NK-associated molecule commonly used for distinguishing cytotoxic or secretory NK subsets (CD56dim and CD56bright, respectively) [29]. It can be induced on CD8 T cells and it is a marker for T cells with a high cytotoxic potential [33]. Although CD56 is not a triggering molecule in NK cells, recent results suggest that it plays an important role in activating CD56⁺ CD8 CTL [44]. CD56 is also expressed on neural cells, where it is considered to be an adhesion molecule (neural cell adhesion molecule, N-CAM) playing a crucial role in neuronal development [27]. CD56 has been shown to be involved in homotypic adhesion between tumour cells and CD56⁺ cells of the immune system [36, 88]. It is expressed on a variety of tumours as Ewing’s sarcoma [60], small cell carcinoma [57], neuroblastoma [90], ovary tumours [84, 90] and melanoma [1, 31]. Although the functional implications of CD56 expression on tumour cells are not fully defined, the overexpression of CD56 by tumour cells leads to decreased NK cell adhesion and inhibition of lysis by NK cells [39]. CD56 is also involved in tumour cell binding to endothelial cells [92]. These interactions could be related to enhanced malignancy of CD56⁺ melanomas [1, 31].

Analysis of CD57 expression by flow cytometry showed that 12.9% melanoma cell lines were positive (Fig. 1). The CD57 molecule has been implicated in cell adhesion and cell migration [82]. It can be expressed in a wide range of tumours, including uveal and cutaneous melanoma and immunohistochemical studies have linked CD57 expression to their metastatic behaviour [78, 81]. In vitro assays using melanoma cell lines expressing CD57 have demonstrated an active role for this molecule in melanoma invasiveness and migration [21], supporting the functional relevance of CD57 expression in melanoma.

CD49d, the alpha chain of VLA-4 beta-1 integrin, was detected on 69% of melanoma cell lines (Fig. 1). This molecule plays a role in cell–cell interactions and cell adhesion to the extracellular matrix. It has been suggested that VLA-4 expressed on melanoma cells could allow the melanoma cells to migrate from the vascular system to tissues by enhancing melanoma cell tethering, adhesion to endothelial cells and establishing metastases [40, 46]. This supports a potential role of this integrin in the invasive and metastatic capacity of melanoma.

Altogether, it can be hypothesised that adhesion molecule expression by melanoma cells can act as a two-edged sword that could both favour recognition and elimination by immune effector cells and also their interaction with endothelial cells allowing tumour cells to cross the endothelial barrier.

Analysis of ligands for activating receptors involved in NK cell-mediated cytotoxicity

Ligands for NKG2D on melanoma cell lines

The important human NK-activating receptor NKG2D is a homodimeric C-type lectin-like receptor that has been shown to interact with several MHC class I-related molecules. Thus, human NKG2D ligands (NKG2DLs) include the stress-inducible surface glycoproteins MICA and MICB, and the UL16-binding proteins (ULBPs). The latter form a multigene family with at least six functional members. NKG2DL can be expressed on tumour and tumour cell lines of different origins and are also upregulated after viral infection. In humans, NKG2D is constitutively expressed on NK cells and CD8 T cells, as well as in gamma–delta T cells and NKT cells. Activation through NKG2D directly leads to triggering of NK cell-mediated cytotoxicity, whereas on T cells, it acts as a costimulatory molecule [9, 14, 59, 68, 93].

It has been shown that NKG2DL expression stimulates anti-tumour activity of NK cells [6, 7, 9, 16, 18, 26]. On the basis of these previous studies, a large panel of melanoma cell lines were analysed showing that 85% expressed at least one ligand for NKG2D (Fig. 2a). Detailed analysis revealed that most cell lines expressed MICA/B molecules (80%), but ULBP expression was less widespread (ULBP1, 15%; ULBP2, 25%; ULBP3, 20% of melanoma cell lines analysed).

Fig. 2.

Percentage of melanoma cell lines expressing the ligands for NKG2D and DNAM-1. a Analysis of NKG2D ligands (MICA/B and ULBP-1, 2 and 3). b Analysis of DNAM-1 ligands (CD155 and CD112). The study was performed by FACS and the scores normalised as indicated in Fig. 1 legend

The high level of expression of NKG2DL on melanoma cell lines suggests that NKG2D–NKG2DL interactions are likely to represent an important mechanism in NK cell recognition of melanoma cells. Therefore, we addressed the functional significance of NKG2DL expression on melanoma cell lines with regard to recognition by NK cells, by using the NK cell line NKL as model effectors. NKL expresses high levels of NKG2D, but only marginal levels of the NCR activating receptors (NKp30, NKp44, and NKp46). Analysis of melanoma susceptibility to NKL-mediated lysis using cell lines selected by their expression of NKG2DL, together with mAb directed against MICA/B or NKG2D, supports the role of NKG2D–NKG2DL on NK susceptibility of melanoma cell lines. As shown in Fig. 3a and b, NKL cytotoxicity towards the cell lines ESTDAB-075 and ESTDAB-081 (MICA⁺ ULBP⁺) was partially inhibited by anti-MICA/B and strongly by anti-NKG2D mAb. In a similar manner, cytotoxicity against ESTDAB-167 (MICA^negative ULBP2⁺) was not affected by anti-MICA/B, but the addition of anti-NKG2D mAb strongly blocked NKL-mediated lysis. This indicates that the recognition of ULBP2 was involved in the killing (Fig. 3c). In contrast, ESTDAB-067, expressing low levels of NKG2DL, was only marginally killed by NKL and this was not altered by the addition of mAb against MICA/B or NKG2D (Fig. 3d). As a control, we used the standard NK target, EBV transformed cell line 721.221, which does not express NKG2DL, but was highly susceptible to NKL lysis by an NKG2D independent mechanism. As expected, the addition of anti-MICA/B or anti-NKG2D mAb did not inhibit cytotoxicity (Fig. 3e) indicating that receptor–ligand interactions other than NKG2D–NKG2DL are involved in the lysis of this cell line. Killing of these cell lines was not affected by the addition of the appropriate isotype control (not shown). Together these results thus indicate that NKG2DL expression renders melanoma cells susceptible to NK cell-mediated cellular cytotoxicity and that NKL lysis of NKG2DL⁺ melanoma cell lines was critically dependent on MICA/B or ULBP interactions with NKG2D. This has been described previously for several cell lines of various tissue origins including melanoma [9, 63] and more extensively for leukaemia cells [70].

Fig. 3.

Analysis of melanoma susceptibility to NKG2D-dependent NKL cell cytotoxicity. Specific lysis of melanoma cell lines by NKL cell line was analysed by blocking experiments with mAb anti-MICA/B or anti-NKG2D mAb. The EBV transformed cell line 721.221, a standard NK target that does not express NKG2DLs, was used as control. Data shown are representative of three to five independent experiments at an E:T ratio of 20:1

It has been shown that cells expressing NKG2D infiltrate melanoma tumours and that a high percentage of metastatic melanoma lesions have lower levels of NKG2DL compared with the primary tumour that was positive for these molecules [48, 49]. These results suggest both a role for NKG2D-mediated activation in anti-melanoma response and that immunosurveillance by NKG2D positive NK and CD8 T cells can be subverted in vivo by the downregulation of MICA/B, although extensive correlative studies should be performed to confirm this possibility.

One of the mechanisms used by tumour cells to evade NKG2D-mediated recognition is the release of NKG2DL, which has been observed in a variety of human tumour entities and is thought to interfere with NKG2D-mediated tumour immunity in several ways (reviewed by Salih et al. [72]). Furthermore, increased levels of soluble NKG2DL, either MICA/B or ULBP2, have also been found in serum of patients with different malignant conditions [37, 38, 71, 86], indicating that established tumours can escape NKG2D-mediated tumour immunosurveillance by releasing NKG2DL. The shedding of both MICA and ULBP2 involves the activation of metalloproteases (MPs) [34, 70, 73, 86] likely belonging to the ADAM family of proteases [85].

Persistent NKG2DL expression induces a pronounced down-regulation of surface NKG2D on NK and activated CD8 T cells and a severe impairment of NKG2D-mediated cytotoxicity in vitro and in vivo [91]. Thus, the release of NKG2DL by tumour cells induces the downregulation of NKG2D on cytotoxic lymphocytes, thereby contributing to the escape of MICA/B or ULBP-positive tumour cells in vivo. The shedding of NKG2DL by tumour cells as a possible mechanism to escape immunosurveillance and the demonstration of the involvement of MPs in this process are of particular interest, since MP inhibitors are clinically available. Therefore, therapeutic blockade of MPs offers the possibility of interfering with this mechanism and subsequently increasing melanoma immunogenicity and susceptibility to NK and CD8 T cell cytotoxicity as suggested by Waldhauer et al. [85]. In a similar manner, several immunoregulatory molecules that can be released by tumour cells can also interfere with NKG2D–NKG2DL interactions. For example, it has been shown that TGF-β can transcriptionally inhibit the expression of NKG2D on effector cells [23] and also MICA, ULBP2 and ULBP4 expression [28]. In a recent study, it has been demonstrated that MIF can also contribute to the immune escape of ovarian carcinoma by transcriptionally down-regulating NKG2D in vitro and in vivo, thus impairing NK cell cytotoxicity towards the tumour cells [41].

Taken together, the results discussed earlier support the idea that the expression of NKG2DL on melanoma cells promotes anti-melanoma immunosurveillance by activating natural killer cells and, likely, by costimulating CD8 T cells, but NKG2DL shedding constitutes a major countermechanism of tumour cells to subvert NKG2D-mediated immunosurveillance. The possibility of interfering with NKG2DL shedding using MP inhibitors opens new therapeutic avenues to be explored.

Expression of ligands for DNAX accessory molecule-1 (DNAM-1)

DNAX accessory molecule-1 (DNAM-1) was originally identified as an adhesion molecule constitutively expressed on the majority of peripheral blood T lymphocytes [74]. DNAM-1 is also expressed by virtually all human NK cells and cross-linking DNAM-1 transduces activating signals resulting in enhancement of cytotoxicity and cytokine production by T and NK cells. It has recently been demonstrated that DNAM-1 induces NK cell activation through interactions with CD155 (poliovirus receptor, PVR) and CD112 (Nectin-2), two closely related molecules belonging to the Nectin family [11, 77]. Gilfillan et al. [ 32] show in mice lacking DNAM-1 that CD8 T cells require DNAM-1 for co-stimulation when recognizing antigen presented by nonprofessional antigen-presenting cells and that NK cells require DNAM-1-mediated signals for the elimination of tumour cells resistant to NK cell-mediated cytotoxicity due to low expression of other NK cell-activating ligands. The expression of DNAM-1 ligands in certain tumours is involved in cell-mediated cytotoxicity by NK and T cells [24, 62, 64, 76]. Here we show that DNAM-1 ligands are frequently expressed by melanoma cell lines (Fig. 2b). In particular, CD155 is expressed by the majority, but in contrast, CD112 was found on only 26% of the lines (Fig. 2b). These results suggest that these molecules may represent major ligands for triggering NK-mediated cytotoxicity and cytokine secretion against human tumour cells.

Expression of ligands for other activating NK receptors

The NK stimulating receptor 2B4 (CD244) interact with his ligand CD48, which is broadly expressed by cells of the haematological lineage [47]. Our results show that CD48 was not expressed by any of the melanoma cell lines studied (n = 75). Other NK-activating receptor is NKp80 that interacts with the C-type lectin-like receptors AICL, which is expressed on myeloid cells, such as monocytes and macrophages. NKp80–AICL interactions can promote lysis of a malignant myeloid cell line [89]. However, to our knowledge, the expression of this marker on solid tumours has not been studied.

A recent study has analysed the expression of NCR ligands on a small panel of tumour cell lines of different origins [17]. NCR are activating NK receptors that are expressed by resting (NKp30 and NKp46) or activated (NKp30, NKp46 and NKp44) NK cells [53]. Although the ligands for NCRs on tumour cells are still unknown, the use of NCR chimeric proteins has allowed the demonstration that NKp30 and NKp44 ligands, but not NKp46 ligands are expressed in a high percentage of cell lines from solid tumours, such as pancreatic and breast carcinoma, including a melanoma cell line [17]. Whereas NK cells use NCRs as activating receptors inducing killing of target cells [53, 54], it has also been shown that, on the contrary, tumour cells can induce apoptosis of NK cells by engaging the NCRs [65]. These findings indicate that NCRs are not only triggering molecules essential for antitumor activity, but also surface receptors that can be involved in NK cell death.

Concluding remarks

Melanoma cells can be recognized by NK cells through the NKG2D-activating receptor, but their lysis may require different thresholds of effector cell activation depending on the level and number of ligands for activating receptors that they express. Participation of different activating receptors may act in a synergistic manner favouring the elimination of melanoma cells. Thus, the redundancy of ligands for activating receptors will be advantageous for tumour immunosurveillance. It remains to be determined whether the activation threshold for NKG2D can be lowered by simultaneous engagement of other activating or costimulatory receptors within the immunological synapse.

Identification of ligands for NK-associated receptors and co-stimulatory molecules expressed on melanoma cells can be used as indicators of susceptibility to NK-mediated lysis. Further insights into the functional significance of the coexpression of different ligands for NKRs on melanoma cells and how different NKRs participate in melanoma recognition and lysis are required. For the use of NK cells in immunotherapy against melanoma, predicting NK cell-mediated therapeutic efficacy could be approached based on phenotypic analysis of ligand expression.