In Vivo Proof of Concept of Adoptive Immunotherapy for Hepatocellular Carcinoma Using Allogeneic Suicide Gene-modified Killer Cells

Abstract

Cell therapy based on alloreactivity has completed clinical proof of concept against hematological malignancies. However, the efficacy of alloreactivity as a therapeutic approach to treat solid tumors is unknown. Using cell culture and animal models, we aimed to investigate the efficacy and safety of allogeneic suicide gene-modified killer cells as a cell-based therapy for hepatocellular carcinoma (HCC), for which treatment options are limited. Allogeneic killer cells from healthy donors were isolated, expanded, and phenotypically characterized. Antitumor cytotoxic activity and safety were studied using a panel of human or murine HCC cell lines engrafted in immunodeficient or immunocompetent mouse models. Human allogeneic suicide gene-modified killer cells (aSGMKCs) exhibit a high, rapid, interleukin-2–dependent, and non–major histocompatibility complex class I-restricted in vitro cytotoxicity toward human hepatoma cells, mainly mediated by natural killer (NK) and NK-like T cells. In vivo evaluation of this cell therapy product demonstrates a marked, rapid, and sustained regression of HCC. Preferential liver homing of effector cells contributed to its marked efficacy. Calcineurin inhibitors allowed preventing rejection of allogeneic lymphocytes by the host immune system without impairing their antitumor activity. Our results demonstrate proof of concept for aSGMKCs as immunotherapy for HCC and open perspectives for the clinical development of this approach.

Introduction

Hepatocellular carcinoma (HCC) is the third leading cause of cancer-related deaths worldwide. Although early-stage tumors can be curatively treated using surgical approaches, treatment options for advanced HCCs are limited.¹ Hepatic resection is the treatment of choice for HCC in non-cirrhotic patients, but this group accounts for a minority of patients.² Liver transplantation is the best therapeutic option for HCC as it simultaneously removes the tumor and the underlying cirrhosis, thereby minimizing the HCC recurrence risk. However, this option is limited by organ shortage and its restriction to patients with limited disease. Minimally invasive treatments such as percutaneous treatments or radiofrequency ablation are alternatives for early-stage HCC patients who are not eligible for surgical resection or transplantation. Transarterial chemoembolization may offer palliative benefits for patients with intermediate-stage HCC.³ Although the multikinase inhibitor, sorafenib, was the first agent to demonstrate a survival benefit for patients with locally advanced or metastatic HCC, treatment options for patients with advanced HCC remain limited.¹ Therefore, development of novel therapies for patients with advanced HCC or early-stage HCC not eligible for standard of care remains a major unmet medical need.

Cancer immunotherapy is emerging in preclinical and clinical evaluation.⁴ Allogeneic hematopoietic stem cell (HSC) transplantation has demonstrated that alloreactivity provides one of the most powerful antitumor effects in immunotherapy of hematological malignancies. Indeed, an HSC graft contains pluripotent HSCs (that will reconstitute complete long-term hematopoiesis) and immune cells, such as T cells. Up to 10% of these T cells are alloreactive and recognize differentially expressed human leukocyte antigen (HLA) determinants. Thus, they are responsible for a severe complication of HSC transplantation, graft-versus-host disease (GvHD), but also for graft rejection prevention and graft-versus-leukemia effect. In addition to T cells, natural killer (NK) cells also provide alloreactivity through the recognition of NK receptor ligands mismatches.⁵ Furthermore, alloreactivity may be efficient for the treatment of solid tumors.⁶ However, if allogeneic adoptive immunotherapy is highly efficient, it can elicit toxicities that need to be controlled. Ex vivo retroviral-mediated transfer of a “suicide” gene, such as the herpes simplex virus-thymidine kinase (HSV-tk) or the inducible caspase 9 (iCasp9) in donor lymphocytes allows their specific killing by their respective prodrug, ganciclovir (GCV)⁷ or chemical inducer of death (CID).⁸ Clinical trials demonstrated that allogeneic suicide gene-modified killer cell (aSGMKC)–induced GvHD can be controlled by their prodrugs^8,9 and that aSGMKCs are able to exert a strong graft-versus-leukemia effect.^9,10 However, the effect of this cell therapy product (CTP) against solid tumors is unknown. Using HCC as a model for solid tumors, here we show that aSGMKCs have a potent non–major histocompatibility complex (MHC) class I-restricted antitumor activity in vitro and in vivo and open a previously unknown perspective for the clinical treatment of HCC.

Results

Production and phenotypic analysis of aSGMKCs for HCC immunotherapy

Allogeneic SGMKCs from healthy blood donors were produced as shown in Supplementary Figure S1a. Prior to its evaluation in preclinical HCC models, we analyzed the phenotype of the CTP. As shown in Table 1, most aSGMKCs are CD3⁺CD56⁻ T cells, followed by CD3⁺CD56⁺ NK-like T cells. In contrast, CD3^–CD56⁺ NK cells make up the smallest proportion within aSGMKCs. This phenotype is similar to that of control cells and shows increased NK-like T cells frequencies, as compared with peripheral blood mononuclear cells (PBMCs) (Table 1).¹¹ While control cells were not GCV sensitive, growth of aSGMKCs was GCV inhibited (Supplementary Figure S2). Alloreactivity of aSGMKCs is impaired, as compared with PBMCs, in terms of proliferative response after allostimulation by PBMCs,^12,13 B-EBV cell lines^11,14,15 or HCC cell lines (Supplementary Figures S3 and S4) and in terms of allo- or xenogeneic GvHD induction in animal models.^14,16,17,18 The impairment of aSGMKCs' alloreactivity can be reversed by replacing CD3 activation with CD3/CD28 costimulation and/or by replacing interleukin (IL)-2 with IL-7.¹⁴ Thus, we evaluated the cytotoxicity against Huh7 cells or HeLa cells (as a positive control for target cell lysis) of aSGMKCs generated from PBMCs after CD3 versus CD3/CD28 activation and expansion with IL-2 versus IL-7 (Figure 1a). CD3+IL-2-activated cells exhibited the highest cytotoxic activity, expressed as lytic units 50% (Supplementary Figure S1b). Replacing CD3 activation by CD3/CD28 costimulation and/or IL-2 by IL-7 was associated with a lower cytotoxicity (Figure 1a). Therefore, only CD3⁺IL-2 activation was used to produce aSGMKCs in subsequent experiments.

Table 1. Relative distribution of T, NK-like T, and NK cells in aSGMKCs, as compared with PBMCs or control cells.

Figure 1.

In vitro cytotoxicity of allogeneic suicide gene-modified killer cells (aSGMKCs) against hepatoma cells. (a) Peripheral blood mononuclear cells (PBMCs) were expanded for 2 weeks after CD3 or CD3/CD28 monoclonal antibodies (mAbs) ± IL-2 or IL-7 activation before evaluation of their cytotoxic activity. Control groups: CD3 + IL-2: 653 ± 145 lytic units 50% ((LU50) (HeLa) and 224 ± 109 LU50 (Huh7); n = 4. (b) The cell lines Huh7, PLC-PRF5, and SK-Hep1 are sensitive to aSGMKCs-mediated killing. Control group: HeLa cells killing: 172 ± 22 LU50; n = 6. (c) The cytotoxic activity of effector cells (PBMCs, control cells or aSGMKCs from the same donors) was evaluated on D1, D3, or D6 of coculture with Huh7 cells. Control group: control (ctrl) cells on D6, 218 ± 64 LU50; n = 4. (d) aSGMKCs' cytotoxic activity was assessed without any cytokine (no CK) and in the presence of IL-2 or IL-7 during coculture with target cells. Control group: IL-2, 164.4 ± 46.5 LU50; n = 5. (e) aSGMKCs' cytotoxic activity is not inhibited by the addition of an anti–human leukocyte antigen (HLA) class I mAb during the coculture. Control group: culture with “no mAb”, 210 ± 98 LU50; n = 4. (f) Cytotoxic activity of immunomagnetically purified (see fraction purities in Supplementary Table S1) T, natural killer (NK)-like T, and NK cells was tested in parallel with unsorted aSGMKCs. Control group: unsorted aSGMKCs, 154 ± 33 LU50; n = 4. Data are expressed as mean ± SEM of the control group. In all panels, *P < 0.05 compared with the control group (Mann–Whitney test). IL, interleukin.

aSGMKCs demonstrate broad and sustained activity against hepatoma cells

Next, we studied aSGMKCs' cytotoxic activity against hepatoma cells. Allogeneic SGMKCs were tested against Huh7 and PLC-PRF5 hepatoma cells. SK-Hep1, an endothelial cell line, was also assessed in parallel. All target cells were sensitive to aSGMKCs killing (Figure 1b). Because Huh7 cells exhibited an intermediate sensitivity to killing, we chose this cell line to further characterize the aSGMKCs' cytotoxicity. PBMCs exhibited significant cytotoxic activity after 3 or 6 days of coculture with target cells, whereas aSGMKCs cytotoxic activity was maximal after only 1 day of coculture (Figure 1c), indicating that aSGMKCs are ready to exert their cytolytic potential. The gene transfer process did not affect aSGMKCs' cytotoxic activity, as shown by the similar level of cytotoxicity observed with control cells generated in parallel (Figure 1c).

Mechanism of action

Allogeneic SGMKCs' cytotoxic activity is IL-2 dependent, while IL-7 cannot replace IL-2 during the cytotoxicity assay (Figure 1d). Moreover, neither the W6/32 monoclonal antibody, a pan anti-HLA class I antibody inhibiting CD8⁺ T cells proliferative response in mixed lymphocyte reaction (data not shown), nor an isotype control monoclonal antibody significantly affected aSGMKCs' cytotoxicity (Figure 1e). These data demonstrated that the cytotoxic activity is non–MHC class I restricted. To go further, we analyzed the lytic potential of T, NK-like T, and NK cells sorted from aSGMKCs compared with that of unsorted aSGMKCs. Purified T cells had a significantly lower cytotoxicity than unsorted aSGMKCs, whereas purified NK-like T or NK cells displayed an increased cytotoxicity (Figure 1f). These data suggest that NK-like T and NK cells are mostly responsible for aSGMKCs' cytotoxic activity against hepatoma cells.

In vivo cytotoxicity of aSGMKCs in xenograft HCC models

In vivo cytotoxicity assays were performed using HeLa-Luc cells as a positive control for lysis sensitivity and Huh7-Luc cells as hepatoma target cells in xenograft HCC mouse models. When increasing aSGMKCs amounts were subcutaneously coinjected with fixed target cell numbers into immunodeficient mice, a dose-dependent reduction of the relative tumor growth was observed for HeLa-Luc (Figure 2a,c) and Huh7-Luc cells (Figure 2b,d), even at low effector:target ratios (3:1). However, tumor escape in some mice led to an increase of the mean relative Huh7 tumor growth on D7 or D14. Therefore, the frequency of mice with tumor progression was also evaluated and was shown to be dose-dependently reduced (Figure 2e,f). High effector:target ratios allowed for complete tumor eradication in most animals (Figure 2a,b), as demonstrated by the low percentage of animals with tumor progression (Figure 2e,f).

Figure 2.

In vivo cytotoxicity of allogeneic suicide gene-modified killer cells (aSGMKCs) against HeLa-Luc and Huh7-Luc cells. Mice were subcutaneously (s.c.) coinjected either with HeLa-Luc (a,c,e) or Huh7-Luc (b,d,f) cells and aSGMKCs. (a,b) The bioluminescence activity of target cells is shown at the time of injection (D0) and on D7 for representative mice s.c. injected with phosphate-buffered saline (PBS) or aSGMKCs at an effector:target (E:T) ratio of 30:1. Control (ctrl): negative control of bioluminescence (no target cell injection). (c,d) aSGMKCs dose-dependently delay or reduce (values <1) the relative tumor growth (mean ± SEM). (e,f) aSGMKCs dose-dependently reduce the proportion of mice with tumor progression (i.e., with a relative tumor growth value >1) of (e) HeLa-Luc and (f) Huh7-Luc cells. Data in c–f are pooled results of six mice from two independent experiments using three mice per group. (a–f) Daily intraperitoneal injections of interleukin-2 (IL-2) were performed in all mice. *P < 0.05 compared with control group without aSGMKCs injection (i.e., E:T ratio = 0:1), at the respective time points (c,d: Mann–Whitney test; e,f: Fisher exact test).

Antitumor activity of aSGMKCs against orthotopic HCC

In a next step, we investigated whether aSGMKCs exhibit antitumor activity in a more relevant model, using an orthotopic HCC model. In contrast to the classical xenograft models, Huh7 cells are injected intrasplenically, resulting in tumor growth in the liver. Intravenous (i.v.) aSGMKCs injection led to a rapid (Figure 3a and Supplementary Figure S5a) and preferential (Figure 3b) liver homing, which is more pronounced than after intraperitoneal (i.p.) injection (Supplementary Figure S5b,c). Twenty-four hours after their injection, aSGMKCs were found surrounding the tumor (Supplementary Figure S5d–f). These cells were CD3⁺, of which some were also CD56⁺. Huh7-Luc orthotopic tumor growth quantification was performed by in vivo bioluminescence imaging (BLI) before and after aSGMKCs i.v. injection (Figure 3c). The mean tumor growth, expressed either as absolute bioluminescent activity (Figure 3d) or relative bioluminescent activity (Figure 3e), and the frequency of tumor-progressing mice (Figure 3f) were markedly reduced on D3 and D7 after i.v. aSGMKCs injection, as compared with the nontreated control group. Allogeneic SGMKCs were recovered from the livers on D7 postinjection and analyzed by flow cytometry for the relative repartition of NK, NK-like T, and T cells. As compared with the infused product, the frequencies of NK and NK-like T cells were increased, whereas T cells were less frequent (Figure 3g), suggesting that NK and NK-like T cells may be more likely to be involved in the antitumor effect. In order to further demonstrate that the antitumor effect was performed by aSGMKCs, repeated administrations of GCV for 4 days were initiated at the time of aSGMKCs infusion. However, no effect on aSGMKCs-induced tumor regression was observed (data not shown), possibly because aSGMKCs-mediated killing of tumor cells could be more rapid than GCV-induced killing of aSGMKCs. Therefore, the antitumor effect of aSGMKCs modified to express iCasp9, a suicide gene known to eliminate the transduced cells more rapidly than the HSV-tk suicide gene (Supplementary Figure S6a,b)¹⁹ was evaluated in the absence or presence of its prodrug, the AP20187 CID. The antitumor effect was reversed by only one injection of CID at the time of iCasp9⁺ aSGMKCs infusion (Figure 3h and Supplementary Figure S6c). This suggests a superior efficacy of the iCasp9/CID system than the HSV-tk/GCV system in the current preclinical system and validates that the antitumor activity is in fact due to the aSGMKCs.

Figure 3.

Antitumor effect of allogeneic suicide gene-modified killer cells (aSGMKCs) in a humanized orthotopic model of hepatocellular carcinoma (HCC). (a,b) DiR-labeled aSGMKCs preferentially and rapidly migrate to the liver when intravenously injected to severe combined immunodeficiency–beige mice. (a) The in vivo fluorescence signal is shown for three mice representative of six. (b) Fluorescence quantification in the liver (Li), spleen (S), lungs (L), and intestine (I) 72 hours after aSGMKCs injection is reported in the right part of the figure (mean ± SEM, n = 4 mice). (c–f) aSGMKCs provide an antitumor effect in a humanized orthotopic model of HCC. (c) Representative experiment showing the Huh7-Luc bioluminescence in one mouse injected with phosphate-buffered saline (PBS) or aSGMKCs. The injection of aSGMKCs delays the tumor growth, (d) evaluated by the luciferase activity of Huh7-Luc cells or (e) expressed as relative tumor growth on D3 and D7. *P = 0.07 compared with control group with PBS injection (Mann–Whitney test). (f) The proportion of mice with tumor progression (i.e., with a relative tumor growth value >1) on D3 and D7 shows that aSGMKCs reduce the frequency of mice with Huh7-Luc tumor progression. Data in d–f expressed as mean ± SEM are pooled results of five and three independent experiments on D3 and D7, respectively, using the indicated total number of mice (n) per group. (g) The relative frequencies of T, natural killer (NK), and NK-like T cells harvested from the liver on D7 postinjection (“ex vivo”) are increased, as compared with the frequencies observed on D0 in the infused aSGMKCs product (“aSGMKCs”). Data are expressed as mean ± SEM of nine mice from three independent experiments. (h) The antitumor effect of CD19/iCasp9⁺ aSGMKCs in the orthotopic HCC model is reversed by only one administration of chemical inducer of death (CID). Representative experiment showing the Huh7-Luc bioluminescence in one mouse injected with or without 2.5 mg/kg CID at the time of intravenous injection of 150–200 × 10⁶ aSGMKCs. Interleukin-2 (IL-2) was intraperitoneally injected daily. (a–h) Daily intraperitoneal injections of IL-2 were performed in all mice.

Safety of aSGMKCs injection

No clinical side effects were observed in mice infused with aSGMKCs, suggesting that antitumor activity could be safely established in absence of GvHD induction. To further explore the toxicity of aSGMKCs, histological analyses were performed in the liver, intestine, kidney, lung, and spleen of tumor-bearing mice on D1 and D7 after aSGMKCs injection.

The livers presented focal zones of necrosis associated with massive inflammatory infiltrates of neutrophils at the periphery of the necrotic hepatocytes (Figure 4a). These zones of necrosis resulted from thrombosis (Figure 4a, insert) that were probably induced by the intrasplenic injection of Huh7 cells. Indeed, they were also observed in tumor-bearing control mice not injected with aSGMKCs, whereas they were not observed in non tumor-bearing control mice injected with aSGMKCs alone (data not shown). Multifocal inflammatory infiltrates of small size, found in the livers of all mice, were mostly CD3 positive (Figure 4b). No histopathological abnormalities were found in the intestines, kidney, and lung (Supplementary Figure S7a–c). As expected, the spleen displayed marked white pulp hypoplasia, but no histopathological abnormalities were observed (Supplementary Figure S7d), excepted the presence of neoplasia in some spleens, due to the intrasplenic injection of Huh7 cells. All together, these results suggest that aSGMKCs were safe to murine tissues and did not induce xenogeneic GvHD.

Figure 4.

Absence of significant cytotoxicity of allogeneic suicide gene-modified killer cells (aSGMKCs) toward primary murine and human hepatocytes. (a) Hematoxilin–eosin staining shows infiltrates of neutrophils (Inf) at the periphery of focal zones of necrosis (N) resulting from Huh7 cell–induced thrombus (insert). (b) Seven days after intravenous injection of 180 × 10.6 aSGMKCs, lymphocytes infiltrating the hepatic parenchyma do not induce histopathological abnormalities (CD3 immunostaining and eosin counterstaining). (c) Primary human hepatocytes revealed by human cytokeratin-18 immunohistochemistry in the liver of chimeric Alb-uPA/severe combined immunodeficiency–beige (SCID-bg) mice 2 weeks after aSGMKCs injection. (d) Magnification of the dashed area in c, showing infiltrates of proliferating lymphocytes (arrowhead) in the human areas. (e) Similar serum levels of human albumin in chimeric Alb-uPA/SCID-bg mice infused (white diamonds, n = 5) or not (white triangles, n = 12) with aSGMKCs (mean ± SEM: 72 ± 22 × 10⁶ cells/mouse). Data are expressed as percent (mean ± SEM) of initial human albumin concentration that were 377 ± 157 µg/ml in the control group and 220 ± 99 µg/ml in the aSGMKCs-infused group. Human albumin levels of one historical control chimeric mouse (black circles, dashed line) that spontaneously lost its graft are shown as example of decreased serum human albumin. PBS, phosphate-buffered saline.

The potential cytotoxicity of aSGMKCs against normal hepatocytes was further explored in human liver-chimeric mouse model, using albumin promoter-driven urokinase-like plasminogen activator transgenic severe combined immunodeficiency–beige (SCID-bg) mice whose livers were repopulated between 10 and 20% with primary human hepatocytes (Figure 4c). Two weeks after aSGMKCs injection, large areas of proliferative cellular infiltration were observed in perivascular areas and, more dispersed, in the hepatic parenchyma as well as in the human clusters (Figure 4d). Despite this infiltration, the level of liver repopulation by primary human hepatocytes was similar in aSGMKCs-injected mice compared with noninjected control mice, showing that there is no massive destruction of those cells after aSGMKCs injection. This is further substantiated by the observation that human albumin serum levels remain stable in both groups (Figure 4e). As reported above in tumor-bearing SCID-bg mice, no histopathological abnormalities were observed in the spleen, lung, kidney, and intestine of human liver-chimeric Alb-uPA/SCID-bg mice, injected or not with aSGMKCs (data not shown).

CsA can control the potential rejection of allogeneic lymphocytes

Allogeneic SGMKCs rejection by the recipient's immune system could be a clinical limitation to our approach. Allogeneic SGMKCs' cytotoxic activity is resistant to the immunosuppressive effect of the calcineurin inhibitor, cyclosporin A (CsA), both in vitro (Figure 5a) and in vivo (Figure 5b). Thus, we reasoned that CsA could prevent the rejection of aSGMKCs, without impairing their antitumor activity (Supplementary Figure S8). To address this question, we used a murine immunocompetent HCC mouse model. As a surrogate model of human aSGMKCs, we used allogeneic ex vivo-expanded splenocytes (aEESs), generated from spleen cells and known to be also resistant to CsA.¹⁸ This mouse model allowed us to assess the safety and efficacy of this approach in an immunocompetent host.

Figure 5.

Resistance of expanded lymphocytes to ciclosporine A (CsA) as a mean to prevent their allorejection. (a,b) Human allogeneic suicide gene-modified killer cells (aSGMKCs) are resistant to CsA in vitro and in vivo. (a) The in vitro cytotoxicity of peripheral blood mononuclear cells (PBMCs) (positive control for sensitivity of effector cells to the inhibitory effect of CsA) and aSGMKCs was assessed at an effector:target ratio of 40:1 in the presence of CsA (one experiment representative of four). *P < 0.05 compared with 0 ng/ml CsA. (b) In vivo aSGMKCs' cytotoxic activity against Huh7-Luc cells was assessed, as reported in Figure 2b,d, in the absence or presence of daily intraperitoneal injections of CsA (n = 3/group). (c–e) CsA prevents the rejection of allogeneic expanded ex vivo splenocytes (aEESs). (c) Luciferase-expressing aEESs were injected to Balb/c recipient mice (n = 3/group) in the presence of IL-2 ± CsA and their presence was monitored by in vivo bioluminescence imaging on D7 (one experiment representative of four). (d) Quantification of the luciferase activity in mice injected with aEESs + IL-2 in the absence or presence of CsA (mean ± SEM; n = 3/group). (e) The CsA does not prevent the homing of aEESs to the liver. DiR-labeled aEESs from FvB/N mice were injected to Balb/c recipient mice in the presence of IL-2 ± CsA and were monitored by in vivo fluorescence gated on the liver (mean ± SEM; n = 3/group). (f) The prevention of aEES rejection by administering CsA allows aEESs to induce a graft-versus-host disease, monitored by body weight loss (mean ± SEM; n = 3 mice without CsA and n = 9 mice with CsA). IL, interleukin.

To investigate and confirm that CsA was able to prevent the rejection of the CTP by the recipient's immune system, aEESs from luciferase-transgenic FvB/N mice (H-2^q) were infused into immunocompetent recipient Balb/c mice (H-2^d) in the presence of IL-2, with and without CsA. In the absence of CsA, aEESs were not detectable by BLI, indicating that they have been rejected by the host's immune system. By contrast, when CsA was administered, luciferase-transgenic aEESs became detectable on D11 (Figure 5c,d), suggesting that CsA prevents allogeneic CTP rejection in an immunocompetent host. At earlier time points postinjection, aEESs from FvB/N mice could be detected by in vivo fluorescence in the liver, to a higher level in the presence of CsA than in its absence (Figure 5e), indicating that CsA did not impair the preferential hepatic homing of aEESs to the liver. In the presence of CsA, aEESs induced a GvHD that results in weight loss (Figure 5f), prostration, and ruffled fur. Although some mice occasionally died, these clinical signs were usually weak, in accordance with histological analysis demonstrating mild lesions at the intestinal level and no lesions at the hepatic level (data not shown). This is in agreement with the observation that aEESs exhibited a decreased alloreactivity compared with fresh spleen cells,¹⁶ as evidenced in immunodeficient recipient Rag2^−/− IL-2Rγ^−/− Balb/c mice (Supplementary Figure S9a–d).

Allogeneic lymphocytes exhibit antitumor activity in immunocompetent mice

To investigate the antitumor activity of our approach in an immunocompetent host, recipient Balb/c mice were transplanted with syngeneic BNL1MEA.7R.1-Luc HCC murine cells. Assessment of the antitumor activity of aEESs in an orthotopic BNL1MEA.7R.1-Luc model was not possible because of a high death rate after BNL1MEA.7R.1-Luc transplantation. Therefore, in order to mimic HCC treatment in an immunocompetent host, established subcutaneous BNL1MEA.7R.1-Luc tumors were treated by repeated intratumoral injections of aEESs from FvB mice in the presence of IL-2 + CsA. As expected, tumors showed a minimal spontaneous regression on D18 in nontreated (Figure 6a) and IL-2–treated (data not shown) control groups. In the IL-2 + CsA control group, tumors ultimately expanded on D18, after a short period of reduced growth (data not shown). In sharp contrast, a continuous tumor size decrease was observed from D7 to 18 in recipient mice treated by aEESs and IL-2 + CsA (Figure 6a) and was associated with enhanced persistence of fluorescent aEESs in the tumor (Figure 6b). This tumor rejection occurred without inducing any clinically detectable GvHD. These data demonstrate proof of concept of this approach in an immunocompetent mouse model.

Figure 6.

Antitumor activity of allogeneic ex vivo-expanded murine splenocytes (aEESs), used as a surrogate cell therapy product in an immunocompetent hepatocellular carcinoma (HCC) mouse model. (a) In the presence of interleukin-2 (IL-2) and cyclosporine A (CsA), but not in the sole presence of IL-2, intratumoral injection of aEESs leads to the rejection of BNL1.MEA7.R1-Luc subcutaneous tumors. *P < 0.05 compared with the control group, injected with phosphate-buffered saline (PBS) alone (Mann–Whitney test; n = 5 mice/group). (b) CsA administration increases the amount of DiR-labeled aEESs present within the tumor, as monitored by in vivo fluorescence gated on the tumor. *P < 0.05 compared with the group injected with aEESs + IL-2 (Mann–Whitney test; n = 5 mice/group).

Discussion

Addressing an unmet medical need for the treatment of advanced HCC or early-stage HCC not eligible to standard of care, we have demonstrated in vivo proof of concept of aSGMKCs for HCC immunotherapy. We have uncovered for the first time that aSGMKCs exhibit a marked antitumor efficacy in state-of-the-art in vitro and in vivo models for HCC including an immunocompetent mouse model for HCC.

Addressing a future clinical development, we demonstrate that the aSGMKCs administration is safe in preclinical animal models. Because aSGMKCs are not tumor antigen specific, they could spread anywhere in the body and be harmful against any cell type. Suicide gene transfer is therefore necessary to control any unwanted side effects of aSGMKCs infusion. The ex vivo culture required for retroviral-mediated gene transfer is associated with a modification of lymphocytes' migratory properties, leading to a preferential hepatic tropism and thus limiting aSGMKCs spreading throughout the body. Furthermore, this hepatic tropism is also of advantage to target HCC and may be relevant for targeting intrahepatic metastasis of nonhepatic tumors such as colorectal cancer. With the aim to further increase the local concentration of aSGMKCs, we will consider the possibility to perform image-guided intratumor injections of aSGMKCs to improve their antitumor efficacy and limit their peripheral secondary effects. A second consequence of ex vivo culture is a decreased alloreactivity in terms of proliferation^12,13 and GvHD induction^14,16,17,18: this is of advantage to limit hepatic toxicity, because the liver is a preferential target organ of GvHD. It should be noted that no clinical signs of xenogeneic GvHD were observed in SCID-bg mice receiving aSGMKCs under conditions where an antitumor effect was observed. This suggests that the antitumor effect may be mediated by a different mechanism than GvHD and may relate to a differential involvement of NK and/or NK-like T cells versus T cells in the cytotoxic activity versus GvHD induction.²⁰ We hypothesize that NK and/or NK-like T cells present in aSGMKCs are already activated and ready to kill target cells, whereas alloreactive T cells may need more time to expand and induce GvHD. Therefore, a therapeutic window may exist, allowing NK and NK-like T cells to provide an antitumor effect before the induction of GvHD. Despite the reduced ability of aSGMKCs to induce GvHD, the suicide gene transfer as a safety control system remains mandatory, as severe GvHD may still occur in small fraction of patients.^9,10 Indeed, numerous in vitro and in vivo preclinical studies have proven the efficacy and safety of the HSV-tk gene as a suitable transgene allowing to control alloreactivity (for a review, see ref. ⁷) and GvHD after HSC transplantation.^9,10 Indeed, aSGMKCs-induced GvHD has been controlled by GCV in all clinical trials reported to date. Our results suggesting a superior efficacy of the iCasp9/CID system than the HSV-tk/GCV system will prompt us to further explore iCasp9⁺ aSGMKCs, especially using orthotopic models of HCC based on the use of primary tumor samples, which are more relevant than cell line-based models.²¹ The iCasp9/CID system has the interest of not requiring the proliferation of target cells to induce apoptosis, which is an advantage because aSGMKCs have a reduced proliferative capacity.

We also demonstrate that potential rejection of aSGMKCs can be controlled by CsA. Since patients are immunocompetent, immune responses against aSGMKCs will occur, leading to their rejection. This will limit the risk of long-term aSGMKCs toxicity, but also their short-term antitumor efficacy. Thus, effector cells must be cytotoxic enough to quickly eliminate the tumor before being rejected by patient's immune cells. In phase I/II clinical trials of administration of partially HLA-matched EBV-specific allogeneic T cells for treatment of EBV-induced lymphoma, complete responses were observed in 14 of 33 patients despite alloimmunization against this “ready-for-use” allogeneic CTP. This suggests that HLA mismatch may not represent a strong limitation to allogeneic adoptive immunotherapy if short-term antitumor effects are desired.^22,23 Similarly, in a phase I trial of infusion of “intentionally HLA-mismatched IL-2 activated killer” cells produced from normal donors, rejection of infused intentionally HLA-mismatched IL-2 activated killer cells was expected to occur within 1 week and was considered as a safety mechanism to avoid GvHD.²⁴ Infusion of intentionally HLA-mismatched IL-2 activated killer cells was safe and, among 35 patients with metastatic solid tumors, only one grade I GvHD was observed.²⁴ Of importance, our data in immunocompetent recipient mice demonstrate the feasibility of obtaining antitumor effects while preventing the rejection of effector cells through CsA administration. This result further strengthens the proof of concept that human aSGMKCs could lead to a solid tumor regression, provided that their rejection is prevented.

The production of a bank of ready-for-use allogeneic CTP has been successfully used to treat hematological immunotherapies in patients.^22,23,25 Compared with autologous immunotherapies, this approach allows an easier large-scale development. Indeed, CTP production is easier when using lymphocytes from normal donors, which usually expand more efficiently than those from heavily treated patients. Second, a ready-for-use allogeneic CTP can be rapidly infused to patients, without the waste of time required for the production and qualification of a patient-specific CTP (i.e., an autologous or patient-directed CTP). Finally, there is no risk of losing a clinical-grade CTP batch: indeed, if a frozen cell batch planned to be infused to a given patient cannot be finally administered, it can still be used later for another patient.

Cytokine-induced killer cells, namely ex vivo-activated autologous PBMCs expanded in similar conditions as our aSGMKCs but without gene transfer,^26,27 have been shown to exhibit potent non–MHC-restricted NK-like T cell–mediated cytotoxic activity especially against HCC cell lines^27,28 and low potential for GvHD induction,²⁹ similarly to our observations with aSGMKCs. Randomized clinical trials^30,31 evaluating cytokine-induced killer therapy associated with transarterial chemoembolization and/or radiofrequency have demonstrated their ability to reduce HCC recurrence and to increase median time to relapse, relapse-free survival, and overall survival in HCC patients. Compared with previously described autologous cytokine-induced killer immunotherapies, our innovative approach based on aSGMKCs alloreactivity presents significant advantages and progress by providing rapid availability, the possibility to control side effects through the introduction of a suicide gene and an additional mechanism for tumor recognition, leading to a more potent antitumor effect. This concept is supported by clinical studies of allogeneic HSC transplantation demonstrating that (i) leukemic patients were at lower risk of relapse when transplanted with allogeneic compared with syngeneic or autologous graft³² and (ii) delayed infusions of aSGMKCs³³ or allogeneic cytokine-induced killer cells³⁴ have antileukemic effects.

Collectively, our results provide a significant conceptual advancement in the immunotherapy of solid tumors by providing for the first time in vivo proof of concept for using aSGMKCs as immunotherapy for HCC. Furthermore, our results open the perspective for a clinical evaluation of aSGMKCs as an immunotherapy in patients with advanced HCC or early-stage HCC not eligible to standard of care.

Materials and Methods

Detailed Materials and Methods are available in Supplementary Materials.

Production of aSGMKCs. Allogeneic SGMKCs were produced as previously described¹⁴ and is detailed in Supplementary Figure S1a. PBMCs from 23 blood donors were activated with CD3 monoclonal antibody and human IL-2 and transduced on day (D) 3 with the MP71-T34FT retroviral vector encoding the human CD34 truncated form fused to the HSV-tk³⁵ or the SFG.iCasp9.2A.ΔCD19 retroviral vector (provided by M.K. Brenner, Center for Cell and Gene Therapy, Baylor College of Medicine, Houston, TX) encoding the human CD19 and iCasp9,⁸ immunomagnetically selected on D5 and expanded until D14. Transduced and CD34- or CD19-selected cells are referred as aSGMKCs, whereas nontransduced cells expanded in parallel for 14 days are referred as control cells. Effector cells (aSGMKCs or, when indicated, control cells or PBMCs) were then qualified for phenotype and GCV sensitivity and functionally characterized for in vitro and in vivo cytotoxicity, as outlined in Supplementary Figure S1a.

Cell lines. HeLa, Huh7, PLC-PRF5, and SK-Hep1 human cell lines were provided by the American Type Cell Culture (LGC Standards, Molsheim, France) and the Balb/c-derived HCC murine cell line BNL1MEA.7R.1³⁶ was kindly provided by Jean-François Dufour (University of Bern, Switzerland).

Animals. Eight- to 12-week-old C57BL/6 Rag2^−/−γ_c^−/− mice were provided by Naoko Uchiyama (Central Institute for Experimental Animals, Kawasaki, Japan), C.B-17 severe combined immunodeficiency–beige (SCID-bg) mice were obtained from Taconic (Ejby, Denmark), Balb/c and FvB/N mice were provided by Janvier (Le Genest Saint Isle, France), luciferase-transgenic FvB/N (FvB/N-Luc) mice were provided by Robert S. Negrin (Stanford University, CA), and uPA/SCID-bg mice were kindly provided by Maura Dandri (Hamburg University, Germany). Animal experimentations received the approval of the local ethical committee (Comité Régional d'Ethique en Matière d'Expérimentation Animale de Strasbourg, approval n°AL/34/41/02/13).

In vitro cytotoxicity assay. Target cells (HeLa, Huh7, PLC-PRF5, and SK-Hep1) and graded amounts of effector cells were cocultured for 1–6 days in the presence of IL-2 and, when indicated, of anti-HLA class I monoclonal antibody (5 µg/ml; clone W6/32; Tebu, Le-Perray-en-Yvelines, France). Subsequently, remaining viable, adherent target cells were stained with crystal violet. Data are expressed as lytic units 50% of the experimental group normalized to the lytic units 50% values of the control group (Supplementary Figure S1b).

In vivo cytotoxicity assay. In vivo cytotoxicity assays were performed as previously described.³⁷ Luciferase-expressing target cells (HeLa-Luc and Huh7-Luc) were subcutaneously injected into C57BL/6 Rag2^−/−γ_c^−/− immunodeficient recipient mice in the absence or presence of graded amounts of aSGMKCs, with daily i.p. IL-2 injection. Relative tumor growth was obtained by calculating the ratio of in vivo BLI on the indicated day to BLI on D0.

Orthotopic xenogeneic tumor model. SCID-bg mice with hepatic localization of intrasplenically injected Huh7-Luc cells were i.v. injected with aSGMKCs or vehicle and were daily i.p. injected with IL-2. Relative tumor growth was monitored by BLI. Homing of aSGMKCs was analyzed by in vivo fluorescence of DilC18(7) tricarbocyanine (DiR; Invitrogen, Cergy-Pontoise, France)-labeled aSGMKCs. CD19/iCasp9⁺ aSGMKCs were injected with or without 2.5 mg/kg CID (AP20187 or BB Homodimerizer; Ozyme, Montigny le Bretonneux, France).

Murine immunocompetent tumor model. Murine splenocytes were expanded as previously described.¹⁶ To evaluate the prevention of aEES rejection, Balb/c mice were i.v. injected with aEESs from FvB/N-Luc mice, daily i.p. injected with IL-2 ± CsA and monitored for the presence of aEESs by BLI. To evaluate the antitumor activity of aEESs, Balb/c mice were subcutaneously injected with BNL1.MEA7.R1-Luc cells. One week later, they received one intratumoral injection per day for 4 days of aEESs from FvB/N mice. Mice were daily i.p. injected with IL-2 + CsA and monitored by BLI for relative tumor growth. The GvHD was monitored clinically as described by Cooke et al.³⁸ and was quantified by body weight loss.

Statistical analysis. Data are expressed as mean ± SEM and were compared with the Mann–Whitney or Fisher exact nonparametric tests, excepted for Table 1, where normal data distribution allowed to perform a Student's paired t-test.

SUPPLEMENTARY MATERIAL Figure S1. Outline of experimental protocols for aSGMKCs production and determination of Lytic Units 50% (LU50). Figure S2. Allogeneic SGMKCs, but not control cells, are sensitive to the suicide gene prodrug. Figure S3. Hepatoma cells have a low allostimulating potential. Figure S4. Allogeneic SGMKCs have a decreased proliferative response to hepatoma cells, as compared to PBMCs. Figure S5. Intravenous injection is more efficient than i.p. injection for favoring hepatic homing of aSGMKCs. Figure S6. Efficacy of CD19/iCasp9⁺ aSGMKCs depletion by the prodrug CID. Figure S7. Administration of aSGMKCs does not induce histopathological abnormalities in intestines, kidney, lung and spleen. Figure S8. Experimental design for adoptive allogeneic immunotherapy in immunocompetent recipients. Figure S9. Allogeneic ex vivo-expanded splenocytes are less alloreactive than fresh SC. Table S1. Relative distribution of T, NK-like T, and NK cells before and after immunomagnetic sorting of T, NK-like T and NK cells from aSGMKCs.