The engagement of CTLA-4 on primary melanoma cell lines induces antibody-dependent cellular cytotoxicity and TNF-α production

Primary melanoma cell lines were derived from tumor tissue samples of cutaneous melanoma patients, who underwent surgical resection of skin or lymph node metastases at the IRCCS AOU San Martino-IST (Genoa, Italy). This study was approved by the local Institutional Ethics Committee (n.OMA09.001) and patients gave written informed consent according to the Declaration of Helsinki.

Tissue specimens were processed for establishment of the primary cell lines as described [31].

Expression of Melan-A and GP100 melanocyte differentiation antigens (MDA), of CD133, CD117 and CD271 stem cell-related antigens (SCA), of nestin and CD56 neural crest antigens (NCA) was analyzed by immunofluorescence, as reported [32] and described in Additional file 1.

Among the established melanoma cell lines, C32 and MeWo were obtained from ECACC (Salisbury, UK) and FO-1 was kindly provided by S. Ferrone (New York Medical College, 1991), HLA typed by SSPO analysis [33] and authenticated in our lab by PCR-SSP. The human lymphoblastoid B cell line C1R-neo was obtained from ATCC (Manassas, USA, 2011) and validated according to its short tandem repeat. Last authentication was performed before using the cell lines for the present study.

Expression of surface and cytoplasmic CTLA-4 was analyzed by flow cytometry as reported [8] and described in Additional file 1. For CTLA-4 surface staining with Ipilimumab human antibody (Bristol-Myers-Squibb), indirect immunofluorescence was performed by incubating, for 30 min at 4°C, 2×10⁵ cells/sample with the mAb (20 μg/ml). CTLA-4 cytoplasmic staining with Ipilimumab was performed on fixed (2% paraformaldehyde) and permeabilized (0.1% saponin) 4×10⁵ cells/sample. Both stainings were followed by the addition of Alexafluor 647-conjugated goat anti-human IgG secondary antibody (Molecular Probes, Inc. Eugene, OR, USA). Negative controls included directly labelled and unlabeled isotype-matched irrelevant mAbs.

Results were expressed as mean ratio of relative fluorescence intensity (MRFI), calculated as follows: mean fluorescence intensity (MFI) of CTLA-4 staining/MFI of irrelevant isotype-matched mAb staining.

Analysis of CTLA-4 transcript variants by RT-PCR and quantitative RT-PCR (qRT-PCR) were performed as described in Additional file 1 and in the Table of Additional file 2.

Immunohistochemical (IHC) analysis of CTLA-4 expression was performed on formalin-fixed, paraffin-embedded (FFPE) tissues of cutaneous melanoma lesions by staining with either the anti-CTLA-4 14D3 mAb or Ipilimumab.

For reaction development, we used an Alkaline Phosphatase(AP)- Fast Red staining for 14D3 and a peroxidase-DAB staining for Ipilimumab. Both whole tissue slides and tissue microarray (TMA) were stained (see Additional file 1). Scores for percentage of stained cells were 0 (negative), 1 (1-29%), 2 (30-59%), 3 (60-100%). Scores for staining intensity were 0 (negative), 1+ (weak), 2+ (moderate) 3+ (strong). A final immunoreactive score (IRS) for CTLA-4 expression was obtained by multiplying both scores [34] resulting in the following IRS (values from 0 to 9): 0 (negative), 1–4 (low to intermediate) and ≥6 (high). Stained slides were analyzed by two independent observers under an optical microscope (Olympus BX41) using 10× ocular lens, 63× objectives. Image acquisition was performed with Leica (DMD1.08) microscope.

Peripheral blood mononuclear cells (PBMC) were obtained after Ficoll-Hypaque density centrifugation of blood samples derived from healthy volunteers. Highly purified preparations of NK cells and γδT cells were obtained from PBMC as described [35] and tested in a conventional 4h ADCC assay [36]. Production of TNF-α was determined by ELISA (see Additional file 1).

Non-obese diabetic/severe combined immunodeficiency (NOD/ SCID) mice were purchased from Harlan Laboratories (Udine, Italy) and housed according to the institutional animal care guidelines. All experiments were approved by the Ethics Committee for Animal Use in Cancer Research at our institute. All mice were approximately 7 weeks-old. Tumorigenicity assay of melanoma cell lines was performed as described in Additional file 1.

Different melanoma xenografts were prepared for subcutaneous (s.c.) injections into NOD/SCID mice. Briefly, freshly harvested MECO cells (2×10⁶) were washed twice and incubated with Ipilimumab or Rituximab (both at 20 μg/ml) at 4°C for 30 min. Treated and untreated MECO cells, either alone or mixed at 1:1 ratio with human NK cells isolated from the buffy coats of three different healthy donors, were injected s.c. (200 μl/mouse) into the mice (6 injections per each experimental condition). Tumor growth was evaluated as described in Additional file 1 starting from day 5 of melanoma and NK cell xenograft implantation.

In the present study, we analyzed CTLA-4 expression in 14 primary cell lines originating from metastatic lesions of cutaneous melanoma patients and in 3 long-term established melanoma cell lines.

We found that all the cell lines, except FO-1, expressed variable levels of surface and cytoplasmic CTLA-4 (Figure 1A). It is of note that FO-1 cell line appeared to be surface CTLA-4 negative but it was positive in the cytoplasm (Figure 1A).

We next investigated whether CTLA-4 expressed by melanoma cells was recognized by the therapeutic Ipilimumab antibody. Flow cytometric analysis showed that Ipilimumab reacted, with different intensity, at the cell surface of all, except FO-1, melanoma cell lines tested (Figure 1B). It is of note that Ipilimumab did react with FO-1 into the cytoplasm. Both polyclonal and Ipilimumab anti-CTLA-4 antibodies reacted with the established human melanoma cell lines C32 and MeWo (Figure 1A,B).

CTLA-4 expression in primary melanoma cell lines appeared to be independent from the stage of melanoma differentiation. Indeed, cytofluorimetric analysis did not point out any statistically significant Spearman’s correlation coefficient (not shown) for surface CTLA-4 expression, detected with either the polyclonal antiserum or Ipilimumab, and the expression pattern of MDA, as well as of SCA and NCA (Table 1). Flow cytometric analysis of melanoma cells, double-stained with Ipilimumab and an anti-CD56 (NCAM) Ab, showed that the cell lines expressed CTLA-4 and CD56 simultaneously (representative profiles are shown in Figure 2A), although at variable intensity (Table 1). The melanoma nature of our cell lines was further confirmed by IHC staining for S100 marker (representative experiments are shown in Figure 2B).

CTLA-4 expression was confirmed at transcriptional level by RT-PCR analysis identifying the two transcript variants CTLA-4TM and CTLA-4delTM/soluble [37].

Expression of CTLA-4TM transcript was found with variable intensity in all melanoma cell lines; CTLA-4delTM (sCTLA-4) transcript was expressed at lower levels in respect to CTLA-4TM, in all cell lines except in MECO (Figure 1C). The expression of CTLA-4delTM transcript confirmed the finding that the primary melanoma cell lines secreted detectable levels of sCTLA-4 (range: 0.1-1.05 ng/ml), as defined by ELISA (Figure 1D).

CTLA-4 protein expression was also found on FFPE tissue sections from 5 metastatic lesions (ML) used to originate the cell lines, and from additional melanoma lesions. IHC with a murine anti-CTLA-4 mAb (14D3) demonstrated a diffuse and strong positivity, uniformly spread throughout the tumor (representative staining in Figure 3A). A similar staining pattern was observed with Ipilimumab (representative staining in Figure 3B).

An anti-S100 mAb was used as positive control (representative staining in Figure 3C) whereas a murine anti-CD20 mAb and Rituximab were used as isotype-matched irrelevant mAbs (representative staining in Figure 3D,E).

Moreover, we analyzed 28 melanoma tissues by TMA immunostaining for CTLA-4 and we evaluated CTLA-4 expression through the immunoreactive score (IRS; 0 = negative, 1-4 = low to intermediate and ≥6 = high), an index which takes into account both parameters of percentage of positively stained cells and staining intensity, according to their individual scores (34), as described in the Methods. Twenty out of 28 (71.4%) tissues were found positive for CTLA-4 expression (IRS > 1), although with variable percentage of stained cells and intensity, whereas 8 (28.6%) were found CTLA-4-negative (IRS = 0) (Table 2). In particular, 9 out of 20 (45.0%) CTLA-4-positive tissues showed low to intermediate expression (IRS 1–3) and 11 (55.0%) showed high CTLA-4 expression (IRS ≥6).

We further confirmed the expression of CTLA-4 and reactivity of Ipilimumab by performing qRT-PCR in a CTLA-4⁺ melanoma tissue sample consisting of almost, if not all, melanoma cells. A strong expression of CTLA-4 transcript was detected in this tissue as compared to METR cell line and FO-1 used as control reference (see Figure of Additional file 3 and Additional file 4).

We further investigated whether Ipilimumab could trigger activation of NK cells to ADCC upon interaction with CTLA-4⁺ melanoma cells. To this aim, ex-vivo isolated NK cells were used in a conventional cytolytic assay using CTLA-4⁺ melanoma cell lines in the presence of Ipilimumab. The results showed that NK cells efficiently killed melanoma cells at high effector:target cell ratio (E:T) (40-60% of lysis at 40:1 E:T ratio) (Figure 4A). This lysis was barely detectable at very low E:T ratio of 1:1 (5-17% depending on cell line used), but it was significantly increased by the addition of Ipilimumab to the cytolytic assay. Indeed, the lysis of MECA, MECO, MEMO and METR cell lines at 1:1 E:T ratio (21, 7, 15 and 5% respectively in the absence of antibody) was enhanced in the presence of Ipilimumab (46, 35, 42 and 23% respectively) (Ipilimumab vs. control, P < 0.001; P = 0.023; P = 0.003; P = 0.023, respectively, referred to 2.0 μg/ml of Ipilimumab) (Figure 4A). Moreover, it appeared that this enhancement was superimposable at concentration of 20, 2.0 and 0.2 μg/ml of Ipilimumab.

The induction with Ipilimumab of NK cell-mediated ADCC was confirmed using the established melanoma cell lines C32 and MeWo (Figure 4A). It is of note that Ipilimumab did not trigger ADCC of the CTLA-4 negative cell line FO-1 (Figure 4A). Indeed, NK cell-mediated lysis of FO-1, in the presence of Ipilimumab, was superimposable to basal cell lysis observed in the absence of any antibody.

The triggering of ADCC was detectable only using Ipilimumab as the addition of the human anti-CD20 Rituximab antibody to the cytolytic assay did not affect basal lysis of the different melanoma cell lines tested (Figure 4A).

In addition, no lysis of melanoma cells was detected in the presence of Ipilumumab alone, avoiding a direct effect of Ipilimumab after a short time incubation (data not shown).

The triggering of ADCC detected using Ipilimumab with ex-vivo NK cells and CTLA-4⁺ melanoma targets was conceivably due to the binding of Ipilimumab to FcγRIIIA expressed on NK cells as the addition of an anti-FcγRIIIA antibody to the assay could almost block the ADCC (Figure 4B, P < 0.0001). The addition of an anti-NCAM antibody did not affect Ipilimumab-mediated ADCC (Figure 4B, P = 0.129 n.s.). Although not shown, Ipilimumab-triggered ADCC in the presence of human serum was superimposable to that observed in its absence. This indicates that human immunoglobulins do not compete with Ipilimumab bound to CTLA-4 expressed on melanoma cells for the binding with FcγRIIIA.

It is known that CTLA-4 is also expressed on activated T cells and this expression is maximal on day2 after stimulation with PHA [9, 38]. Thus, to analyze whether NK cells can kill T cells expressing CTLA-4 as well, we used ex-vivo isolated NK cells as effector cells and PHA-stimulated PBMC as target cells. No lysis of PBMC (Figure 4C, left panel) in the presence of Ipilimumab was detected although CTLA-4 was expressed on these cells (data not shown). Of note, no differences were observed using autologous or allogeneic NK cells. Rituximab used as control did not trigger lysis of PBMC as expected (being CD20 not expressed on PHA blasts), while it efficiently triggered lysis of CD20⁺ C1R-neo lymphoblastoid cell line (Figure 4C, right panel).

We next analyzed whether Ipilimumab could also trigger ex-vivo NK cells to produce anti-tumor cytokines, such as TNF-α, during co-cultures with CTLA-4⁺ melanoma cells. To this aim, TNF-α was measured in the SN obtained from co-cultures of NK cells isolated from 4 different donors and CTLA-4⁺ MECO cell line (1:1 E:T ratio), by ELISA (Figure 4D). The addition of Ipilimumab to NK-MECO cell co-cultures strongly enhanced TNF-α production as compared to NK-MECO co-cultures alone or in the presence of Rituximab used as control (Ipilimumab vs. Rituximab, P = 0.041).

We next analyzed whether ADCC can also be elicited in PBMC, IL-2 activated NK cells and γδT cell populations expressing FcγRIIIA. Furthermore, we analyzed ADCC using three different melanoma cell lines expressing different levels of reactivity with Ipilimumab. Indeed, as shown in Figure 5A, MECO cell line expressed high reactivity with Ipilimumab (Figure 5A, MRFI:6.1) while FO-1 melanoma cells did not react with Ipilimumab (Figure 5A, MRFI:1.4) and METR cells showed an intermediate reactivity with Ipilimumab (Figure 5A, MRFI:3.8).

PBMC can efficiently kill MECO at E:T ratio of 40:1 in the presence of Ipilimumab (35% vs. 1% of lysis in the absence of antibody, P < 0.001). It is of note that Ipilimumab-mediated ADCC was still evident at 20:1 and detectable at 10:1 (20% and 12% vs. 1% of lysis, respectively, in the absence of antibody, P = 0.002 and P = 0.003).

No difference between basal lysis and lysis in the presence of Ipilimumab was detectable using the CTLA-4 surface negative FO-1 cell line. Using METR as target cells, we detected, at the E:T target ratio of 40:1, an increment of lysis in the presence of Ipilimumab (from 5% to 20%, P = 0.004). This increase was lower at 20:1 (from 5% to 11%, P = 0.003) and almost undetectable at 10:1 (from 5% to 8%, P = 0.006) E:T ratios respectively. IL-2 activated NK cells (Figure 5B) killed more efficiently MECO than METR in the presence of Ipilimumab (at 5:1 E:T ratio from the basal lysis of 35% to 75% with Ipilimumab using MECO and from the basal lysis of 35% to 42% with Ipilimumab using METR as target, P < 0.001 and P = 0.228 n.s., respectively). No increment of lysis was detected with Ipilimumab using the CTLA-4 surface negative FO-1 cells. Finally, γδT cell populations expressing FcγRIIIA can kill very efficiently the MECO cell line in the presence of Ipilimumab (from 5% as basal level to 65% with Ipilimumab at 1:1 E:T ratio, P < 0.001). On the other hand, the lysis of METR was barely incremented in the presence of Ipilimumab while no effect was detected on the lysis of FO-1 (Figure 5C, middle and right panel). In no instance, Rituximab used as control antibody, could significantly enhance cytolysis of the melanoma cell lines using different effector cells (Figure 5A and not shown).

We next investigated in vivo, using a NOD/SCID murine xenograft model, whether NK cells can affect tumor growth in the presence of Ipilimumab. In vivo experiments were carried out with the MECO cells as this cell line efficiently triggered in vitro NK cell-mediated ADCC (Figure 4A).

Previous studies showed that human NK cells engraft and retain cytotoxic function when injected s.c. into SCID mice, along with allogeneic human tumor cells [39]. Thus, we injected s.c. a mixed cell suspension of ex-vivo isolated allogeneic human NK cells and MECO cell line incubated in medium alone (MECO/NK) or with Ipilimumab (IPI-MECO/NK) or Rituximab (used as control antibody) (RIT-MECO/NK) at 1:1 NK:MECO ratio. We choose this experimental setting as it more closely reflected the conditions used in the in vitro cytotoxicity assay (Figure 4A). Also, MECO cell line incubated in medium or with Ipilimumab (IPI-MECO) or Rituximab (RIT-MECO) was injected as additional control.

After a single inoculation of the six different MECO xenografts (IPI-MECO, IPI-MECO/NK, RIT-MECO, RIT-MECO/NK, MECO and MECO/NK), the tumor growth was monitored weekly for up to 30 days. In all the experimental groups, tumors were detected within 15 days (mean tumor volume: 56.93 ± 11.71 mm³) except in IPI-MECO/NK xenografts in which just palpable really small tumors started to appear at that time (mean tumor volume 3.00 ± 2.19 mm³, P = 0.042 vs. all other xenografts). In all mice, tumor growth progressively increased until day 30 (mean tumor volume: 307.11 ± 28.58 mm³) but, again, a significantly reduced tumor growth was observed in mice injected with IPI-MECO/NK xenografts (mean tumor volume: 163.15 ± 35.22, P = 0.024 vs. all other xenografts). Comparing the growth of IPI-MECO/NK xenografts to that of IPI-MECO control xenografts, we found a significant reduction of tumor growth at day 15 (P = 0.005), day 20 (P = 0.009) and day 30 (P = 0.028) (Figure 6A). Moreover, NOD/SCID mice engrafted with RIT-MECO/NK, used as control, did not show delay in tumor formation or inhibition of tumor growth compared with mice engrafted with RIT-MECO (P = 0.686) (Figure 6B). It is of note that MECO and MECO/NK xenografts gave rise to tumors of similar volume indicating that NK cells per se did not affect tumor cell growth (Figure 6C). The growth of IPI-MECO/NK xenografts was also significantly reduced in respect to the growth of MECO control xenograft as well as to the growth of all other MECO xenografts observed at day 30 (P = 0.018 and P = 0.042, respectively) (Figure 6A,B,C).

The in vivo reduced tumor growth observed with IPI-MECO/NK xenograft was found to correlate with the lytic activity observed in vitro when the human allogeneic NK cells used for the in vivo injections were co-cultured with MECO cells in the presence of Ipilimumab (data not shown).