Allogeneic SGMKCs were produced as previously described (34). Buffy coats of 35 healthy anonymous volunteer blood donors (purchased from the French National Blood Service following receipt of their written informed consent for research use of their blood donation) were used as a source of PBMCs. Buffy coats were obtained by centrifugation over a Ficoll layer (Invitrogen; Thermo Fisher Scientific, Inc., Waltham, MA, USA) at 360 × g for 20 min at 20°C. PMBCs were activated with 10 ng/ml CD3 monoclonal antibody (mAb) Orthoclone OKT3 (OKT3; Janssen-Cilag, Levallois-Perret, France) and 500 IU/ml human interleukin-2 (IL-2; Proleukin; Novartis Pharmaceuticals Canada, Inc., Dorval, QC, Canada), and transduced on day 3 with the MP71-T34FT retroviral vector (Eufets GmbH, Idar-Oberstein, Germany) encoding the truncated form of the human CD34 (selection marker) fused to the Herpes simplex virus thymidine kinase (HSV-tk) suicide gene (36) or the SFG.iCasp9.2A.ΔCD19 retroviral vector (provided by Professor M.K. Brenner, Center for Cell and Gene Therapy, Baylor College of Medicine, Houston, TX, USA) encoding the human CD19 (selection marker) and the inducible caspase-9 (iCasp9) suicide gene (26). Cells were then immunomagnetically selected on day 5 with an autoMACS Pro Separator (Miltenyi Biotec SAS, Paris, France), based on the expression of the CD34 or CD19 membrane markers, and expanded until day 14. Transduced and CD34 or CD19-selected cells were referred to as aSGMKCs, while non-transduced cells that expanded in parallel for 14 days were termed control cells. The transduction efficiency (percentage of CD34 or CD19-positive cells following transduction) was 8.6±1.0% (n=13) and 51.2±6.1% (n=3), and the purity of the immunomagnetically selected fraction was 92.4±1.1% (n=13) and 94.8±0.4% (n=3) for CD34/HSV-tk and CD19/iCasp9-transduced cells, respectively. Relative cell growth was calculated as the ratio of the number of cells (counted using a B-500 microscope; Optika, Ponteranica, Italy) of the indicated phenotype obtained at the end of the expansion at 37°C (day 14) to the input number of cells of the specific phenotype at initiation of the culture (day 0).

The cells were produced according to three previously published protocols: Protocol 1 (P1), PBMC activation with 100 U/ml interferon-γ (IFN-γ; PeproTech, Inc., Neuilly-sur-Seine, France) at day 0 and 10 ng/ml OKT3 + 500 U/ml IL-2 at day 1 (9,37). These cells were referred to as CIK P1. Protocol 2 (P2), PBMC activation with 100 U/ml IFN-γ at day 0 and 10 ng/ml OKT3 + 500 U/ml IL-2 + 100 U/ml IL-1 (PeproTech, Inc.) at D0 (11,15). These cells were referred to as CIK P2. Protocol 3 (P3): PBMCs activation with 10 ng/ml OKT3 + 500 U/ml IL-2 + 20 ng/ml IL-7 (PeproTech, Inc.) + 20 ng/ml IL-15 (PeproTech, Inc.) at day 0 (38). These cells were referred to as CIK P3. All cells were expanded until day 14 in culture medium consisting in RPMI 1640 medium (Invitrogen; Thermo Fisher Scientific, Inc.) supplemented with 10% v/v human serum and 500 U/ml IL-2.

Effector cells were analyzed for the relative repartition of CD3⁻/CD56⁺ NK, CD3⁺/CD56⁺ NK-like T and CD3⁺/CD56⁻ T cells following staining with Pacific Blue-conjugated CD3 mAb (BD Biosciences, San Diego, CA, USA) and phycoerythrincyanin (PE-Cy)5-conjugated CD56 mAb (Miltenyi Biotec GmbH, Bergisch Gladbach, Germany). The purity of aSGMKCs was evaluated by flow cytometry following staining with PE-Cy7-conjugated mAb specific for CD34 or CD19 (BD Biosciences). Multi-color samples were acquired on a LSRII special order research product flow cytometer (LSRII; BD Biosciences) calibrated using a Cytometer Setup & Tracking Beads kit (BD Biosciences) to ensure consistency of fluorescence intensity measurement throughout all experiments. Compensation was performed with a CompBeads kit (cat. no. 51-90-9001229; BD Biosciences). Cell debris and dead cells were excluded using forward scatter area and side scatter area and doublet cells were excluded using side scatter width and side scatter area. FACSDiva software (version 6.1.2; BD Biosciences) was used for the final analysis and graphical output.

Target cells (1,500 HeLa, Hu h-7 cel ls/wel l, 2, 50 0 SK-Hep1 cel ls/wel l, 20,000 PLC-PRF/5 cells/well; American Type Culture Collection, Molsheim, France) were co-cultured in RPMI 1640 medium (with 500 U/ml IL-2) in the presence or absence of graded quantities of effector cells for 6 days in flat bottom 96-well plates and, when indicated, in the presence of monoclonal mouse anti-human tumor necrosis factor-α (TNF-α; clone MAb1; cat. no. 16-7348), monoclonal mouse anti-human IFN-γ (clone NIB42; cat. no. 16-7318), monoclonal mouse anti-human Fas-ligand (Fas-L; clone NOK-1; cat. no. 16-9919) or monoclonal mouse anti-human TNF-related apoptosis inducing ligand (TRAIL; clone RIK-2; cat. no. 16-9927) mAbs (5 µg/ml; eBiosciences, Inc., San Diego, CA, USA; 1:100). Subsequently, non-adherent effector cells and dead target cells were removed by washing with phosphate-buffered saline (PBS), and the remaining viable, adherent target cells were stained with crystal violet (Sigma-Aldrich, St. Louis, MO, USA) at room temperature for 1 min. Absorbance was read at 560 nm on a Mithras LB940 microplate reader (Berthold Technologies GmbH & Co., Thoiry, France). Data are expressed as percentage killing, at the indicated effector:target (E:T) cell ratios as: [1 − (Abs._E:T / Abs.Ctrle)] × 100, where Abs.E:T indicates the absorbance at a given E:T ratio and Abs.Ctrle is the absorbance of target cells alone (E:T=0). Percentage killing is then expressed as lytic units 50% (LU50), calculated as the reverse of the number of cells per 10⁶ effector cells required to kill 50% of target cells (Fig. 1). LU50 of the experimental groups were normalized to the LU50 values of their control group, in order to normalize inter-experimental variations, and expressed as the mean ± standard error (20).

Lymphokine-activated killer (LAK) cells were generated by activating PBMCs with IL-2 (1,000 U/ml) for 4 days. The effector cells (aSGMKCs or LAK cells) and target cells (Huh-7 cells or K562 cells, used as positive control of lysis) were co-incubated at 37°C in culture medium at an E:T cell ratio of 4:1 in the presence of monoclonal mouse anti-human PE-CD107a (cat. no. 555801) or isotype control antibody (10 µl per 10⁶ effector cells; BD Biosciences; 1:10) to a total volume of 100 µl in a flat bottom 96-well plate. After 1 h, 50 µl culture medium containing brefeldin A (10 µg/ml) and GolgiStop (4 µl for 6×10⁶ cells; BD Biosciences) were added for a further 3 h incubation. The effector cells were then labeled in cytometry tubes with Pacific Blue-conjugated CD3 mAb and PE-Cy5-conjugated CD56 mAb, washed twice with 3 ml PBS (Invitrogen; Thermo Fisher Scientific, Inc.), fixed with PBS supplemented with 2% v/v formaldehyde (Merck Millipore, Darmstadt, Germany), and analyzed by flow cytometry (LSRII).

Animal experimentations were performed following approval by the local ethical committee (Comité Régional d'Ethique en Matière d'Expérimentation Animale de Strasbourg; approval no. AL/34/41/02/13) on November 7, 2012, according to European guidelines (directive 2010/63/UE) and relevant national rules. For ethical reasons, mice were not monitored until tumor-associated mortality but were sacrificed at the end of experiments by cervical dislocation following general anesthesia with 3% isoflurane.

In vivo cytotoxicity assays were performed as previously described (20). Luciferase-expressing Huh-7 target cells (1×10⁶ Huh-7-Luc cells) were subcutaneously co-injected into the right flank of 86 severe combined immunodeficient-beige (SCIB-bg) recipient mice (Taconic Biosciences, Inc., Ejby, Denmark) in the absence (PBS, n=13) or presence of 10×10⁶ iCasp9-expressing aSGMKCs effector cells (CIK P1, n=14; CIK P2, n=17; CIK P3, n=13; iCasp9-expressing aSGMKCs, n=17). Daily intraperitoneal injections of IL-2 (10⁶ IU/kg) were administered throughout the duration of the experiment. When indicated, an intraperitoneal injection of the iCasp9 prodrug [AP20187; also termed chemical inducer of death (CID); Clontech Laboratories, Saint Germain en Laye, France] at 2.5 mg/kg was performed together with the aSGMKCs injection (n=12). Eight-to-twelve week-old male and female mice were used and randomly distributed in experimental groups. The mice were housed under specific pathogen-free conditions with food and drinking water containing 1 mg/ml paracetamol (Doliprane 300; Sanofi-Aventis, Paris, France) ad libitum. Luciferase activity of Huh-7-Luc cells, determined by bioluminescence imaging (BLI) using an IVIS Lumina Series II camera (PerkinElmer, Roissy, France) following an intraperitoneal injection of 100 µl luciferin (20 mg/ml; Caliper Lifesciences), was determined at the time of target cell injection on days 0, 4, 7, 11 and 14. Tumor cell bioluminescence, analyzed with Living Image software (version 3.1; PerkinElmer), was quantified in a round-shape region of interest centered on the cell injection site and was expressed as photons/sec/cm²/steradian (p/sec/cm²/sr). Relative tumor growth was obtained by calculating the ratio of bioluminescence at the indicated day to the bioluminescence at day 0. A relative tumor growth value of ≤1 indicated tumor cell elimination and a value of >1 indicated an expansion of tumor cells.

Data that was not normally distributed was compared with Mann-Whitney U or Fisher's exact non-parametric tests; when data distribution was normal, Student's paired t-test was used. Data are expressed as the mean ± standard error. A χ² test was performed to compare the number of animals with and without relapse in the experimental groups compared with the control group. P<0.05 indicates a statistically significant difference.

As previously reported (20,35), aSGMKCs were predominantly constituted of CD3⁺/CD56⁻ T cells, then by CD3⁺/CD56⁺ NK-like T cells and to a lesser extent by CD3⁻/CD56⁺ NK cells. Other mononuclear subsets, including B cells or monocytes, were not detectable (Fig. 2A). This repartition was similar to the one observed in non-transduced control cells that were activated and expanded in parallel with aSGMKCs (Fig. 2A). This indicates that the cell culture process was responsible for these phenotypical modifications and that the gene transfer process did not lead to the preferential transduction or selection of a lymphocyte population. Furthermore, the ex vivo expansion was associated with a significantly higher relative expansion of NK-like T cells than that observed with NK or T cells (P<0.05; Student's paired t-test; Fig. 2B).

As previously reported, aSGMKCs may provide a potent cytotoxic activity toward various liver-derived cell lines, such as Huh-7, PLC-PRF5 (HCC cell lines) and SK-Hep1, and toward the cervical carcinoma cell line HeLa (20). The cytotoxic activity of aSGMKCs, evaluated at different E:T ratios, indicated that 50 and 90% target cell killing were obtained at mean E:T ratios of ~3:1 and ~13:1 to 40:1, respectively (Fig. 3A), excluding PLC-PRF5 cells, which were more sensitive to aSGMKC-mediated killing. It was previously reported that the cytotoxic activity of aSGMKCs, evaluated in more detail toward Huh-7 cells, was HLA class-I independent and was predominantly mediated by CD56⁺/CD3⁻ NK and CD56⁺/CD3⁺ NK-like T cells (20). In order to further characterize the mechanisms of target cell killing, the involvement of the granzyme/perforin pathway was evaluated by CD107a staining of aSGMKCs (and of LAK cells, used as a positive control of non-MHC-restricted cytotoxic cells) following incubation with Huh-7 cells (or K562 cells, used as a control of LAK-sensitive target cells). Following co-incubation with Huh-7 cells, the frequency of CD107a-positive cells, corresponding to effector cells with degranulated lytic granules, were higher in LAK cells than in aSGMKCs generated from the specific donors (Fig. 3B). A similar trend was observed, to a greater extent, when effector cells were incubated with K562 cells (Fig. 3B). Regardless of the E:T cell combinations, the frequency of CD107a-positive cells was higher in the NK cell subset than in the NK-like T cell subset, while no induction of CD107a expression was observed in T cells (Fig. 3C). Overall, considering that the frequency of NK and NK-like T cells was lower than that of T cells (Fig. 2A), the induction of CD107a expression in aSGMKCs following incubation with Huh-7 cells was minimal, suggesting that other killing mechanisms may be involved.

Thus, in vitro cytotoxicity assays were performed in the presence or absence of antibodies targeting IFN-γ, TNF-α, Fas-L or TRAIL. The cytotoxic activity of aSGMKCs was not affected by the addition of anti-IFN-γ or anti-TNF-α antibodies alone but a significant (P<0.05; Mann-Whitney U test; Fig. 3D) reduction of the cytotoxicity level was observed when both antibodies were added, suggesting a synergistic involvement of IFN-γ and TNF-α in target cell killing. TRAIL-mediated killing was also involved in aSGMKC-induced killing, as demonstrated by the significant (P<0.05; Mann-Whitney U test; Fig. 3D) inhibitory effect of blocking anti-TRAIL antibodies on the cytotoxicity level. Blocking anti-Fas-L antibodies had no significant inhibitory effect on sSGMKCs cytotoxicity but enhanced the reduction of cytotoxicity by anti-TRAIL antibodies (P<0.05; Fig. 3D).

It was previously reported, that aSGMKCs exhibited reduced in vitro and in vivo alloreactivity (20,34,35), in terms of proliferative response and potential of GvHD induction and that their cytotoxicity was non-MHC class I-restricted and predominantly mediated by NK and NK-like T cells (20). Together with the present observations indicating the involvement of TRAIL in the killing mechanism, this prompted the evaluation of similarities between aSGMKCs and CIK cells, another immunotherapy product identified to provide potent non-MHC class I-restricted anti-tumor activity toward hematologic and solid tumors. Indeed, CIK cells have also previously exhibited NK and NK-like T cell-mediated cytotoxicity and low potential of GvHD induction (39). Thus, the phenotypic and cytotoxic activity of aSGMKCs and CIK cells were compared, generated in parallel from the same healthy donors using three different cell expansion protocols for CIK cell production. There were no significant differences between aSGMKCs and CIK cells in terms of repartition of T, NK-like T and NK cell subsets (Fig. 4A). Furthermore, the cytotoxic activity of aSGMKCs, evaluated against Huh-7 cells (or HeLa cells, used as a positive control of target cell killing) was similar, or even greater than that of CIK cells, when evaluated following 1 or 6 days of E:T cell co-culture (Fig. 4B).

The data presented was generated using aSGMKCs transduced with an HSV-tk-expressing retroviral vector. However, a new generation of suicide genes, based on the inducible activation of caspase domains, has been recently developed, including the iCasp9 gene (40), whose efficacy has been demonstrated in clinical trials (26,27). This transgene induces rapid killing of aSGMKCs in the presence of their prodrug, CID, as >95% of cells are killed within 2 h, as determined by flow cytometry following Annexin V/propidium iodide staining (unpublished data). This is more rapid than the killing of HSV-tk-expressing aSGMKCs by its prodrug, ganciclovir (20,41). Therefore, iCasp9-expressing aSGMKCs were compared with CIK cells using an in vivo cytotoxicity assay. Immunodeficient SCID-bg mice were subcutaneously injected with Huh-7-Luc target cells, in the absence or presence of aSGMKCs or CIK cells, and monitored for tumor growth by BLI. Fig. 5 represents the data obtained for one representative mouse of each group.

When Huh-7-Luc cells were injected alone, the median relative tumor growth was >1 at all time points, indicating the expansion of tumor cells (Table I). When CIK cells were co-injected with target cells, the median relative tumor growth values were <1 (P<0.05, Mann-Whitney U test), indicating that CIK cells exhibited a significant antitumor effect. When aSGMKCs were co-injected with target cells, a similar anti-tumor effect was also observed (P<0.05, Mann-Whitney U test). Notably, the antitumor effect of aSGMKCs was reversed by CID administration, confirming that the effect was aSGMKC-mediated (Table I). However, tumor escape occurred in certain (12–38%) mice injected with aSGMKCs or CIK cells. Therefore, the frequency of mice with tumor regression (a relative tumor growth value <1) was also evaluated. As indicated by Table I, this frequency was 23.1% in the control group injected with Huh-7-Luc cells alone, indicating a lack of engraftment in these mice. However, this frequency significantly (P<0.05, χ² test) increased to 61.5–88.2% of mice injected with aSGMKCs or CIK cells, indicating tumor rejection in these groups. It is of note that the frequency reversed to 25.0% in mice receiving aSGMKCs when the prodrug CID was administered at the same time (Table I), indicating that aSGMKCs were rapidly killed, before having time to exert their anti-tumor effects.