Melanoma cells become resistant to NK-cell-mediated killing when exposed to NK-cell numbers compatible with NK-cell infiltration in the tumor


Full correspondence: Dr. Massimo Vitale, IRCCS Azienda Ospedaliera Universitaria San Martino–IST, Istituto Nazionale per la Ricerca sul Cancro–Largo R. Benzi 10–16132 Genova, Italy

Fax: +39-010-354282


Additional correspondence: Prof. Lorenzo Moretta, Istituto Giannina Gaslini, Largo G. Gaslini, 5–16147 Genova, Italy

Fax: +39–010-3730671 e-mail:


During the past few years, a number of studies reported that different melanoma cell lines could be extensively lysed in vitro by IL-2-activated NK cells at appropriate effector/target ratios. Here, we show, by histological evaluation of different melanoma lesions, that NK/target-cell ratios compatible with those allowing efficient melanoma cell killing in vitro are hardly reached at the tumor site. We then investigated the outcome of cocultures established at low NK/melanoma cell ratios. After initial NK-mediated lysis, residual melanoma cells acquired resistance to IL-2-activated NK cells. This reflected primarily an increased expression, on melanoma cells, of classical and nonclassical HLA class I molecules, accompanied by a partial downregulation of NKG2D-ligands, and was dependent on NK-mediated IFN-γ release. Consistently, melanoma lesions showed a higher HLA class I expression on tumor cells that were proximal to infiltrating NK cells. In long-term cocultures, the “protective phenotype” acquired by melanoma cells was lost over time. However, this phenomenon was counteracted by downregulation of relevant activating receptors in cocultured NK cells. Analysis of different NK-cell-activating cytokines indicated that IL-15 can partially overcome this novel tumor escape mechanism suggesting that IL-15, rather than IL-2, may be eligible for NK-cell-based immunotherapy.


Natural killer (NK) cells represent one of the effectors involved in the host defense against tumors [1-6]. Thanks to their cyto-toxic activity and their capability of producing cytokines, NK cells play a role both in the direct clearance of tumor cells and in the regulation and polarization of downstream adaptive immune responses [7-10]. The susceptibility of tumor cells to NK-cell-mediated killing largely depends on the expression of ligands recognized by a complex array of either inhibitory or activating receptors and/or coreceptors [11-14]. These ligands include adhesion molecules (such as the ICAMs), the HLA class I (HLA-I) molecules (which are recognized by the main inhibitory NK receptors), and a heterogenous group of molecules recognized by different activating NK receptors. These include the NKG2D ligands MICA, MICB, and ULBP1–4 [15, 16]; the DNAM-1 ligands, NECL-5 (also termed PVR), and Nectin-2 [17, 18]; and the NKp30 ligands: B7H6 and BAT3 [19, 20]. In addition, although not yet characterized, further receptor–ligand pairs are involved in the regulation of NK-cell-mediated killing of tumor cells. For example, the engagement of NKp46- and NKp44-activating receptors by yet-undefined ligands on tumor cells [7, 11, 14, 21, 22] or the interaction of B7H3 expressed on some tumor cells with a putative inhibitory NK receptor influences tumor-cell susceptibility to NK-cell killing [23].

At the tumor site, the efficacy of NK cells in clearing tumor cells is influenced by a series of soluble factors and cellular interactions that determine the activation status of NK cells and their migratory capabilities. Thus, while certain cytokines including IL-2, IL-12, IL-15, IL-18, and IL-21 can enhance, in different ways, NK-cell function [24-28], other factors, such as IL-4, TGF-β, IDO, [29-31], or the interaction with tumor cells or even with tumor-associated stromal cells [32], have an opposite effect. Low NK/melanoma cell ratios at the tumor site can also limit the efficacy of NK cells in clearing tumor cells. At appropriate effector/target (E/T) ratios, NK cells can efficiently kill a large panel of tumor cell lines in short-term cytolytic assays [13, 16, 33, 34]. In addition, they eliminate tumors in vivo at their very early stages. However, in tumors that could increase their size, infiltrating NK cells may become insufficient to contain tumor burden. In these cases, the NK-tumor-cell cohabitation could either favor the selection of NK-resistant tumor-cell variants or induce phenotypic/functional modifications of NK or tumor cells. This could result in tumor escaping from immune-mediated control. Our study shows that melanoma cells can indeed acquire resistance to NK-cell killing when cocultured with NK cells at low E/T ratios. By a comparative analysis of different cytokines acting on NK cells, we also provide evidence that IL-15 may be effective in overcoming this newly defined mechanism of tumor escape and may offer a clue for improving the outcome of NK-cell-based immunotherapy.


Analysis of NK/melanoma cell cocultures at E/T ratios compatible with tumor NK-cell infiltration

The presence of NK cells at the tumor site was investigated by immunohistochemistry in 28 samples of infiltrating primary cutaneous melanomas (PCMs). In 18 of 28 cases (64%) NK-cell infiltrate was evident (Supporting Information Fig. 1A and B), while in the remaining ten samples (36%) NK cells were poorly represented or absent. NK cells were also detected in a group of in situ PCMs (three of eight, 37.5%) (Supporting Information Fig. 1C). Ten samples displaying maximal NK-cell infiltration were selected to evaluate the approximate NK/melanoma cells (NK/M) ratio. By cell counting of multiple tumor areas, the NK/M ratio ranged between 0.03/1 and 0.5/1 (Fig. 1A).

Figure 1.

Quantification of NK-cell infiltrate in melanoma lesions and analysis of NK/melanoma cell coculture at low E/T ratio. (A) The NK/M cell ratio in ten PCMs is reported. For each PCMs, the values from single areas (symbols) and the mean of values from the different areas (bars) are indicated. (B) MeCoP cells were cultured in 96-well flat-bottom microtiter plates (1 × 104 cells/well) with or without NK cells (ratios are indicated). At days 2 and 6, cocultures were photographed (Olympus IX70 microscope equipped with the Hamamatsu ORCA-ER digital camera, images analyzed by CellR 1.2 Olympus, Tokio, Japan). Original magnification: 50×. One representative experiment out of three performed is shown.

In view of these findings, we evaluated the effect of NK cells on tumor cell survival in long-term mixed cell cultures under low E/T ratios: namely, 0.2/1, 0.5/1, or 1/1. Freshly purified NK cells were pretreated overnight with IL-2 (in order to increase their killing capabilities) and then cocultured (in the presence of IL-2) with the melanoma cell line MeCoP (see the Materials and methods). At day 2 of culture, an evident reduction of melanoma cell number could be observed. This effect was proportional to the initial E/T ratio used in coculture (Fig. 1B). The decrease of melanoma cell numbers, however, did not proceed beyond day 2. Rather, in the following days, the remaining melanoma cells progressively proliferated and at day 6 reached the confluence (at least in the case of the 0.2/1 and 0.5/1 ratios) (Fig. 1B). Notably, melanoma cell proliferation occurred despite the presence of viable NK cells in culture at day 6.

NK cells induce functional and phenotypic changes on melanoma cells

A possible explanation for the above observations was that melanoma cells could have modified their susceptibility to NK cells during coculture. To assess this possibility, at day 2 of coculture, viable melanoma cells were isolated and analyzed in a cytolytic assay as a target for polyclonal NK-cell populations previously expanded in IL-2 for 2 weeks (IL-2-expanded NK cells) (see the Materials and methods). As shown in Fig. 2A (upper panels), melanoma cells cocultured with NK cells were less susceptible to lysis than those cultured alone. Remarkably, the degree of resistance to NK cell lysis was higher in melanoma cells recovered from cocultures at higher E/T ratios and was proportional to the efficacy of NK cells in inducing melanoma cell damage in the cocultures (as evaluated by the Annexin V/PI staining on the whole melanoma cell population, Fig. 2A, lower panels; gating strategies for all FACS analyses are shown in Supporting Information Fig. 2),

Figure 2.

Evaluation of functional and phenotypic changes induced on melanoma and NK cells upon coculture. (A) After 2 days of culture at the indicated NK/melanoma ratios, viable MeCoP cells were isolated and their susceptibility to NK cells was assessed in a cytolytic assay (top; white diamonds and black triangles: melanoma cells cultured alone with or without IL-2, respectively; black diamonds: melanoma cells from cocultures). From the same cultures, the whole melanoma cell population was stained with Annexin V/PI to quantify surviving cells (bottom). (B) The indicated melanoma cell lines were cultured as in (A) and then analyzed in cytolytic assays (left) or by flow cytometry (right). White diamonds/open profiles: melanoma cells alone; black diamonds/gray profiles: cocultured melanoma cells; black profiles indicate negative staining. Due to their variable intrinsic susceptibility to NK cells, each cell line was cocultured at a different NK/M ratio: 1/1 for MeCoP and MeMO, 0.5/1 for MeOV, and 0.2/1 for FO-1. (C, D) MeCoP cells cocultured or not with NK cells were analyzed in cytolytic assays either in the absence or in the presence of indicated blocking mAbs. E/T ratio of cytolytic assay in (D): 8:1. (E) MeCoP cells were cultured alone or with NK cells (open and gray-filled histograms) and analyzed by flow cytometry at day 2 (+NK day 2), 6 (+NK day 6), 15 (+NK day 15), or at day 15 after NK cells were washed out at day 2 (W day 15). (F) Surface expression of triggering receptors on NK cells cultured for 15 days in IL-2 either alone (open histograms) or with melanoma cells (gray-filled histograms). (A–D) Data are shown as mean + SD of six-eight samples pooled from three or four (panel B) experiments. Effectors used in cytolytic assays: IL-2-expanded NK cells. *p ≤ 0.05; **p ≤ 0.01; ***p ≤ 0.001 by Student's t-test.

We next analyzed, in analogous coculture experiments, different melanoma cell lines derived from metastatic specimens in our lab (MeCoP, MeOV, and MeMO), and the long-term established cell line FO-1 [35]. At day 2 of coculture, the different cell lines displayed variable percentages of viable cells (i.e. Annexin V/PI-negative cells, Supporting Information Table 1), possibly reflecting their different initial ability to interact with NK cells and trigger their effector functions. In all the cases, viable melanoma cells that were recovered from coculture displayed, at various extents, reduced sensitivity to lysis by IL-2-expanded NK cells (Fig. 2B). MeCoP and MeOV acquired strong resistance, while MeMO and FO-1 only slightly modified their susceptibility. Analysis of the surface phenotype revealed that, after coculture, MeCoP and MeOV significantly upregulated the surface expression of HLA-I molecules; MeMO (which already expressed high HLA-I levels and low NK-sensitivity before coculture) displayed only minimal HLA-I increase, and FO-1 (HLA-I because of the lacking β2-microglobulin) [35] remained negative (Fig. 2B). All melanoma cell lines showed a slight decrease in the expression of the NKG2D ligands (NKG2D-Ls) (MICA, ULBP2, and ULBP3) (Fig. 2B) but an increased expression of ICAM-1 (Supporting Information Fig. 3). The surface expression of Nectin-2 and NECL-5 (i.e. the ligands for the DNAM-1-activating NK receptor), and of B7H3 (whose expression on target cells may protect from NK-cell killing) [23] was not significantly modified after coculture (Supporting Information Fig. 3). In addition, exposure to NK cells induced decrease of GPR56 expression (Supporting Information Fig. 3). GPR56 is not apparently involved in the NK/target-cell interaction, while it would function as receptor for ECM components [36, 37]. Interestingly, GPR56 downregulation on melanoma cells has been proposed to favor their metastatizing properties [37].

Taken together, the above data suggest that acquisition of resistance to NK-cell-mediated killing may be primarily related to the increased expression of protective HLA-I molecules and, minimally, to the decreased surface density of the NKG2D-Ls.

This was confirmed by the fact that, in cytolytic assays, mAb-mediated masking of the HLA-I molecules on NK-conditioned MeCoP cells resulted in a significant reconstitution of their susceptibility to lysis (Fig. 2C). Similar results were obtained with MeMO (whose susceptibility was also increased in unconditioned cells), MeOV, and two additional melanoma cell lines (MeMI and MeBO) (Supporting Information Table 2).

The minor role exerted by NKG2D-Ls modulation on melanoma susceptibility is in line with the effects induced by the specific masking of different activating receptors on NK-mediated melanoma cell killing (Fig. 2D). These experiments indicate that (i) although reduced, on NK-conditioned melanoma cells the NKG2D-L expression levels may be sufficient to partially engage NKG2D; (ii) NK-conditioned and -unconditioned melanoma cells can engage multiple activating NK-receptors (see also Supporting Information Fig. 4).

Melanoma cells show transient increase of classical and nonclassical HLA-I molecules upon coculture

We next analyzed melanoma cells in long-term coculture experiments. At various time points, MeCoP cells were assessed for the expression of classical and nonclassical HLA-I molecules, and MICA. After 2 days of coculture, both classical HLA-I, and nonclassical HLA-G and -E molecules [38, 39] were upregulated or induced ex novo while MICA was partially downregulated (Fig. 2E). After an additional 4 days (i.e. day 6, Fig. 2E), increases of both classical and nonclassical HLA-I molecules were partially reverted while MICA was further downregulated. At day 15, the HLA-I expression returned to the initial levels. Melanoma cell susceptibility to NK-mediated lysis paralleled the variation of HLA-I expression, as it was initially reduced and then progressively recovered until day 15 (Supporting Information Fig. 5). The loss of the “resistant” phenotype in long-term cocultures was likely to be related to the dilution of NK cells primarily reflecting the progressive melanoma cell proliferation (see Fig. 1B). Consistently, when NK cells were washed out at day 2 of coculture, melanoma cells recovered their initial phenotype and susceptibility to NK cells (Fig. 2E and Supporting Information Fig. 5 — W day 15). Another conceivable explanation for the restoration of the initial melanoma cell phenotype after long-term coculture may be the partial impairment of NK-cell function (including their capability of affecting melanoma cells). As compared with NK cells cultured alone, cocultured NK cells indeed expressed lower levels of certain major triggering receptors including: NKp44, NKG2D, and DNAM-1 (Fig. 2F).

The gain of the NK-resistant phenotype on melanoma cells is dependent on IFN-γ released by NK cells

By the use of specific blocking mAbs and trans-well devices, we next investigated the mechanisms by which NK cells could induce the resistant phenotype on melanoma cells. In particular, we analyzed the role of NK-derived cytokines capable of inducing HLA-I modulation, that is IFN-γ and TNF-α. As shown in Fig. 3A (left) the addition of an anti-IFN-γ mAb in coculture experiments reconstituted the susceptibility to NK-lysis of MeCoP cells, while the anti-TNF-α mAb had only minimal effect. On the other hand, in cocultures carried under trans-well conditions (without mAbs), melanoma cells did not significantly modify their susceptibility (Fig. 3A, right). Under the various experimental conditions, susceptibility to lysis of melanoma cells (Fig. 3A) correlated with the degree of modulation of surface HLA-I and NKG2D-L expression (Fig. 3B). Altogether, these data indicate that induction of the resistant phenotype on melanoma cells is primarily sustained by IFN-γ released upon NK/melanoma cell interaction. As shown in Fig. 3C, IFN-γ could be detected in supernatants of melanoma/NK-cell cocultures. Roughly, the level of IFN-γ induced in cocultures correlated with the initial susceptibility to lysis of melanoma cells (i.e. their capabilities of triggering NK cells) and their ability to become NK-resistant (see also Fig. 2A and B). Importantly, IFN-γ concentrations corresponding to those detected in cocultures induced HLA-I upregulation on MeCoP and MeOV cell lines (on MeMO cells, which constitutively expressed HLA-I, at highest levels such increases were not detectable) (Supporting Information Fig. 6).

Figure 3.

IFN-γ is involved in the acquisition of an NK-cell-resistant phenotype by melanoma cells. (A, B) MeCoP cells were cultured alone (MeCoP), with NK cells (+NK), with NK cells in presence of the indicated anticytokine blocking mAbs, or in trans-well (+NK TW), and then analyzed in a cytolytic assay (using IL-2-expanded NK cells as effector), and by flow cytometry. Results in (B) are shown as: exposed melanoma cell MFI/control melanoma cell MFI (E/C MFI) ratio. (C) Supernatants from the indicated (co-)cultures (performed in the presence of IL-2) were analyzed for IFN-γ content. Data are shown as mean + SD of six samples pooled from three experiments.

Different cytokines can influence NK-cell effector functions and melanoma susceptibility in coculture

In these experiments, freshly purified NK cells and MeCoP cells were cultured (either alone or in combination) with different cytokines known to enhance the NK-cell functional capabilities including IL-2, IL-12, IL-15, IL-18, and IL-21. After 2 days, NK cells cultured alone in the presence of the different cytokines were used as effectors in cytolytic assays, to assess either their killing capability (against melanoma cells cultured alone), or the susceptibility of melanoma cells derived from cocultures in the corresponding cytokine. As shown in Fig. 4A (left panels, open diamonds), the different cytokines variably enhanced the NK-cell killing capability. In particular, IL-15 induced maximal effect, followed by IL-2, IL-12, and IL-21, respectively, while IL-18 had only marginal effect. The effect of the different cytokines on NK cells was also evident during coculture: where IL-15 induced maximal NK-mediated killing (see Annexin V/PI analysis in Fig. 4A). On the other hand, melanoma cells rescued from cocultures in the presence of IL-2, IL-15, IL-12, and IL-21 acquired (at various extents) resistance to NK-cell-mediated killing (Fig. 4A, black diamonds and Supporting Information Fig. 7). IL-15- and IL-2-cocultured MeCoP also showed major changes in the HLA-I and MICA expression levels, as compared with changes displayed by IL-12-, IL-18-, and IL-21-cocultured MeCoP cells (Fig. 4B). These phenotypic changes roughly correlated with the level of IFN-γ released under the different coculture conditions (Fig. 4C). Thus, the extent of effective killing of NK-conditioned melanoma cells appears to depend both on phenotypic changes in melanoma cells and on the level of NK-cell-activation induced by different cytokines. This would explain why IL-15-cocultured MeCoP cells, in spite of their highest HLA-I expression levels, could be partially killed by NK cells exposed to IL-15, which, indeed, induces maximal NK-cell activation. This is in line with the phenotypic profile of NK cells recovered from cocultures. As shown in Supporting Information Fig. 8, the presence of melanoma cells interfered with cytokine-induced upregulation of activating NK-receptors, nevertheless, among melanoma-conditioned NK cells, those exposed to IL-15 showed maximal expression of most receptors.

Figure 4.

Evaluation of the effect of various cytokines on NK/melanoma cell coculture. (A) MeCoP cells were cultured for 2 days, either alone or with NK cells, in the presence of the indicated cytokines. Live melanoma cells were isolated from cultures and tested in a cytolytic assay (left). For each test, effectors were represented by NK cells that had been cultured (without melanoma cells) for 2 days in the corresponding cytokine. White and black diamonds represent melanoma cells cultured alone or with NK cells, respectively. The whole melanoma cell population from each coculture was stained with Annexin V and PI and analyzed by flow cytometry (right). (B) After 2 days of culture, MeCoP cells were also analyzed by flow cytometry to assess the surface expression of the indicated molecules. Results are expressed as exposed melanoma cell MFI/control melanoma cell MFI ratio (E/C MFI ratio). (C) Supernatants from the indicated (co-)cultures were analyzed for IFN-γ content. Data are shown as mean + SD of six samples pooled from three experiments.

NK cells are detected in proximity of HLA-I+ melanoma cells in tumor lesions

As shown above, in the presence of NK cells, melanoma cells upregulate their HLA-I expression. Immunohistochemistry analysis on seven PCMs with NK-cell infiltrate showed a variable HLA-I expression on melanoma cells. In particular, in four of seven PCMs, this expression was heterogeneous (Fig. 5A), whereas in the remaining three cases there was lack of HLA-I expression, particularly at the center of the tumor area (Fig. 5B). Interestingly, melanoma cells that expressed higher HLA-I levels were in proximity of NK cells (Fig. 5C–F show images from four different PCMs). This suggests that NK cells may actually affect the HLA-I expression on the tumor cells in vivo.

Figure 5.

Immunohistochemical analysis of HLA-I expression on PCMs with evident NK-cell infiltrate. Sections from PCM cases were stained for HLA-I (blue), nuclear melanoma marker MITF (microftalmia-associated transcription factor) (brown; A, B) and CD56 (brown, C–F). Images were obtained by BX60 microscope equipped with a DP-70 digital camera and processed by Analysis Image Processing software (Olympus). Original magnification: 200× (A–E) and 400× (F). Data shown in panels (C–F) are representative of ten samples.


In this study, we show that, in the presence of low number of NK cells, NK-susceptible melanoma cells that are not rapidly eliminated increase their HLA-I expression, acquire resistance to NK-cell-mediated killing and proliferate in long-term cocultures. Analysis on PCMs specimens suggested that a comparable situation may occur at tumor sites characterized by NK-cell infiltration since: (i) the observed NK/melanoma cell ratios are compatible with those used in the coculture assays; (ii) melanoma cells displaying higher HLA-I expression are detected in proximity of infiltrating NK cells.

The acquisition of a “protective” phenotype by melanoma cells is essentially related to the upregulation of both classical and nonclassical HLA-I molecules. This suggests that the large majority of NK cells, expressing either killer-Ig-like receptors (KIRs) or CD94/NKG2A or immunoglobulin-like transcript (ILT-2) receptors (specific for HLA-A/B/C; HLA-E and HLA-G, respectively), may be eluded by this newly identified tumor escape mechanism [11, 38, 40]. In this context, high levels of nonclassical HLA-I molecules have been recently proposed to favor tumor resistance to NK-cell activity [41]. Remarkably the upregulation of HLA-E and -G molecules could also counterbalance the positive effect of the increased expression of classical HLA-I on CTL-mediated antitumor responses [38, 40, 42]. Indeed, HLA-E and HLA-G molecules are recognized by different inhibitory receptors, such as CD94/NKG2A and ILT-2 or ILT-4, respectively, which are also expressed by CTL and/or APC [38, 42, 43]. In this context, different studies have shown that the engagement of receptors specific for HLA-G could inhibit APC function and induce immunosuppressive T cells [38, 43]. In addition, induction of HLA-G expression has been correlated with survival advantages for different tumors [44, 45].

Inhibition of the NKG2D engagement, identified as an escape mechanism in different tumors [46, 47], only marginally affected susceptibility of the melanoma cells to NK-cell killing in our experimental setting. This may reflect the fact that different melanoma target cell lines are recognized by heterogeneous patterns of activating NK-receptors [4, 33, 34]. The melanoma cell lines analyzed in our study triggered NK cells also via activating receptors other than NKG2D (Fig. 2D and Supporting Information Fig. 4) [34]. Interestingly, this capacity was substantially maintained after coculture, since, upon HLA-I masking, the NK-conditioned melanoma cells became susceptible to NK-cell killing despite the NKG2D-ligands modulation.

IFN-γ released upon NK/melanoma cell interaction plays an important role in the acquisition of the protective phenotype by tumor cells. This is in line with its known capability of regulating the expression of MHC molecules, NKG2D-ligands, and adhesion molecules [42, 47]. In turn, induction of IFN-γ release by NK cells primarily depends on the melanoma cell capability of stimulating the NK cells and thus on their initial susceptibility to NK-cell-mediated lysis. For example, the MeMO cell line (Fig. 2B), which constitutively expresses high HLA-I levels, is poorly susceptible to NK-cell killing (see Annexin V/PI data in Supporting information Table 1), and displays only minimal phenotypic and functional changes upon coculture with NK cells. In contrast, MeCoP and MeOV, which are more susceptible to NK-cell-mediated killing, display major changes upon coculture with NK cells. Also the FO-1 cells are susceptible to NK cells, however, due to the lack of β2-microglobulin, they fail to express detectable surface HLA-I upon coculture (see Fig. 2B), and only partially decrease their susceptibility to NK cells. These observations suggest that this newly described escape strategy may not apply to all tumors, but it may depend on the phenotypic changes and the mutations occurring in transformed cells during tumor progression.

The correlation between the initial susceptibility to lysis and the acquisition of resistance to NK cells after coculture may suggest that cellular debris (produced in coculture) could have a role: for example, by inhibiting NK effectors in cytotoxicity assays. It should be noted, however, that in such assays, in the presence of anti-HLA-I blocking mAbs NK-conditioned and -unconditioned melanoma cells were killed at similar levels. In addition, IFN-γ blockade during coculture even increased melanoma cell killing (and cellular debris) (not shown), but allowed reconstitution of susceptibility to NK lysis in subsequent cytolytic assays.

To overcome this newly described tumor escape mechanism, possible factors capable of enhancing tumor susceptibility to NK cells (as proposed for the 1,25-dihydroxyvitamin D3 on melanoma cells [48]), or cytokines capable of activating NK function, may be considered. During the past few years, different cytokines have been used in an attempt to improve the host response against tumors [26, 49]. The efficiency of these therapeutic approaches, however, is still challenged by the insurgence of cytokine-related toxic effects, or the induction of tumor-escape mechanisms. In this context, our study shows that cytokines enhancing NK-cell effector functions may also favor the induction of a resistant phenotype in melanoma cells. A remarkable exception may be represented by IL-15. This cytokine, by inducing optimal NK-cell cytotoxicity, enables NK cells to partially overcome tumor resistance even at low E/T ratios. These data offer an important clue for the choice of IL-15 for NK-cell-based immunotherapy against tumors. Remarkably, the effect was obtained using soluble unconjugated IL-15, but its efficacy may be further enhanced when this cytokine is presented through the IL15-Rα [50]. In this context, great interest has been recently raised on the possible use of IL-15, and, in particular, of the IL15/IL-15Rα conjugate, in immunotherapy [49, 51].

Loss of susceptibility to NK-lysis acquired by melanoma cells was transient. This suggests that killing of target cells at the tumor site could be delayed but not definitely blocked. As shown in Fig. 2F, however, such tumor escape mechanism is complemented by the simultaneous downregulation of major triggering receptors in cocultured NK cells.

In conclusion, our present study, together with a number of previous reports [23, 30-32, 41, 42], helps to better delineate the complex and multifaceted strategy by which solid tumors may evade NK-cell-directed attack and provides new elements for an effective manipulation of NK cells for immunotherapeutic purposes.

Materials and methods


Tissue sections (4 μm) from formalin-fixed paraffin-embedded blocks (Department of Pathology, University Hospital of Brescia) were stained using primary antibodies to CD56 (Clone 123C3.D5, dilution 1:60, Thermo Scientific, Fremont, CA, USA), MHC-I/HC10 (dilution 1:400, kindly provided by Dr. Ferrone, Pittsburgh, PA, USA), MITF (clone D5, dilution 1:50, Dako Cytomation, Glostrup, Denmark). Reactivity was revealed using NovoLink Polymer Detection System (Novocastra Laboratories, Newcastle upon Tyne, UK) followed by DAB. For double immunohistochemistry, after the first staining, the second immune reaction was visualized using Mach 4 MR-AP, followed by Ferangi Blue (Biocare Medical, Concord, CA, USA) as chromogen. The NK/melanoma cell ratio was evaluated on sections stained for CD56 where melanoma cells were identified morphologically. Cell count was performed in multiple HPF (high-power field corresponding to 40×) per case.

NK cells

Purified NK cells were derived from peripheral blood of healthy donors by using the RosetteSepTM NK cell Enrichment kit (StemCell Technologies, Vancouver, Canada) [32]. IL-2-expanded NK cells were obtained by culturing purified NK cells on irradiated feeder cells, PHA, and 100 IU/mL rhIL-2 (Proleukin, Chiron, Emeryville, CA, USA).

Melanoma cell lines

Tumor samples were obtained in accordance with consent procedures approved by the Institutional Review Board for these studies. MeCoP, MeOV MeMO, MeMI, and MeBO cell lines were derived, as previously described [32], from metastatic melanoma resections obtained from different patients. The melanoma cell cultures were phenotypically characterized and assessed for purity by the analysis of informative markers including Mel-CAM/CD146, GD2, and HLA-I. All cell lines were tested in advance for their susceptibility to NK cell lysis and proliferation capability. For the initial coculture experiments (Fig. 1), a cell line (MeCoP) displaying intermediate proliferation rate and NK-sensitivity was chosen. Long-term established FO-1 cell line was from S. Ferrone (Pittsburgh, PA, USA).

Melanoma/NK-cell coculture

Melanoma cells were cultured in RPMI-1640 10% FCS, 50 IU/mL IL-2, in six-well flat bottom plates (2 × 105 cells/well) either in the absence or in the presence of NK cells (ratios and culture time are indicated in the text). For the evaluation of melanoma cell susceptibility, after coculture wells were extensively washed to eliminate dead cells and NK cells, then adherent, living melanoma cells were harvested and analyzed in cytotoxicity assays using as Effectors IL-2-expanded NK cells. For the evaluation of melanoma cell survival following culture with NK cells, after coculture, the whole melanoma cell population was harvested and analyzed by using the Annexin V/PI staining kit (Bender Systems, Wien, Austria). When specified, the experiments were done in the presence of 5 μg/mL anti-IFN-γ and/or 5 μg/mL anti-TNF-α mAbs (R&D System, Minneapolis, MN, USA). Where indicated (Fig. 4), (co-)cultures were done in the presence of the following cytokine concentrations: 50 IU/mL rhIL-2, 20 ng/mL rhIL-15, 2.5 ng/mL rhIL-12 (Peprotech, London, UK), 100 ng/mL rhIL-18 (MBL, Naka-Ku, Nagoya, Japan), 25 ng/mL rhIL-21 (ProSpec, Rehovot, Israel).

mAbs, flow cytofluorimetric analysis and functional assays

The mAbs 6A4 (IgG1 anti-HLA-A, B, C), A6/136 (IgM, anti-HLA-I), AZ20 (IgG1, anti-NKp30), BAB281, KL247 (IgG1 and IgM, respectively, anti-NKp46), Z231 (anti-NKp44), ECM217 (IgG2b, anti-NKG2D), GN18 (IgG3, anti-DNAM-1) were produced in our lab. The mAbs: 170,818, 165,903, 166,510 (IgG2a — anti-ULBP-1, anti-ULBP2, and anti-ULBP3, respectively, R&D System); 87G (IgG2a, anti-HLA-G) and 3D12 (IgG1 anti-HLA-E) (Biolegend, San Diego, CA, USA) were commercially available. BAM195 (IgG1, anti-MICA), F252 (IgM, anti-NKp30), and F5 (IgM, anti-DNAM-1) were from D. Pende (Genova, Italy). For cytofluorimetric analysis (FACSCalibur Becton Dickinson, Mountain View, CA, USA), cells were stained with appropriate unlabeled mAbs followed by PE-conjugated isotype-specific goat anti-mouse second reagent (Southern Biotechnology Associated, Birmingham, AL, USA).

Melanoma cell susceptibility and NK-cell killing capability were evaluated in 4 h 51Cr-release assays [32]. For IFN-γ quantification, NK and melanoma cells were cocultured for 16 h in 12-well micro-titer plates (1 × 105 melanoma cells/well — E/T ratios: 1/1 (MeCoP, MeMO), 0.5/1 (MeOV), 0.2/1 (FO-1)) and supernatants were analyzed by specific ELISA (BioSource International Camarillo, CA, USA).

Statistical analysis

Data regarding flow cytometry analysis (MFI and percentage of stained cells), percentage of cytotoxicity, and IFN-γ concentration are shown as the mean ± SD of six or eight samples/replicates pooled from three or four (Fig. 2B) independent experiments. *p ≤ 0.05; **p ≤ 0.01; ***p ≤ 0.001 by Student's t-test.


This work was supported by grants awarded by A.I.R.C.: IG projects no. 5282 (M.V.) no. 4725 (L.M.), MFAG project no. 6384 (G.P.), and special project 5 × 1000 no. 9962 (L.M.); Ministero della Salute: Ricerca Oncologica — Project of integrated program 2006–2008, agreements no. RO strategici 8/07 (M.C.M., G.P., M.V.) and 3/07 (L.M.); and Ricerca Finalizzata (2006, 2007); Istituto Superiore di Sanità; Ministero dell'Istruzione dell'Università e della Ricerca (M.I.U.R.-PRIN) 2006 and 2007. We thank C. Rossini (Brescia) for technical support, A. Poggi (Genova) for technical support for cocultures microphotographs, F. Cafiero and N. Solari (Genova) for biopsies.

Conflict of interest

The authors declare no financial or commercial conflict of interest.


Immunoglobulin-like transcript


MHC class I related chain A/B


UL16-binding protein