Preclinical immunotherapy with Cytokine-Induced Killer lymphocytes against epithelial ovarian cancer

Despite improvements in surgery and medical treatments, epithelial ovarian cancer (EOC) remains the most lethal gynaecological malignancy. Aim of this study is to investigate the preclinical immunotherapy activity of cytokine-induced killer lymphocytes (CIK) against epithelial ovarian cancers, focusing on platinum-resistant settings. We generated CIK ex vivo starting from human peripheral blood samples (PBMCs) collected from EOC patients. Their antitumor activity was tested in vitro and in vivo against platinum-resistant patient-derived ovarian cancer cells (pdOVCs) and a Patient Derived Xenograft (PDX), respectively. CIK were efficiently generated (48 fold median ex vivo expansion) from EOC patients; pdOVCs lines (n = 9) were successfully generated from metastatic ascites; the expression of CIK target molecules by pdOVC confirmed pre and post treatment in vitro with carboplatin. The results indicate that patient-derived CIK effectively killed autologous pdOVCs in vitro. Such intense activity was maintained against a subset of pdOVC that survived in vitro treatment with carboplatin. Moreover, CIK antitumor activity and tumor homing was confirmed in vivo within an EOC PDX model. Our preliminary data suggest that CIK are active in platinum resistant ovarian cancer models and should be therefore further investigated as a new therapeutic option in this extremely challenging setting.

from peripheral blood mononuclear cells (PBMC), cord blood 37,38 , bone marrow 39 or other sources 40 , in presence of INF-ϒ, Ab-anti-CD3 and interleukin 2 (IL2) 41 . The cytotoxic activity is mostly mediated by the interaction of their NKG2D membrane receptor with several members of stress-inducible molecules expressed on tumors, such as UL-16-binding proteins (ULBPs) and MHC class I-related chain A and B (MIC A/B) 42,43 .
It has already been reported that MICA/B and ULBPs are expressed on EOC tumors and are associated with poor prognosis 44,45 .
Strong preclinical evidence [46][47][48][49] and early clinical trials with CIK have shown encouraging findings in challenging settings such as metastatic lung cancer, liver cancer, cervical cancer, gastrointestinal cancer, leukemia, soft tissue-sarcoma and melanoma. Moreover, some preclinical works underscore the killing capacity of CIK even against ovarian cancer cells in vitro 30,50,51 . We built on this evidence and focused our attention on the setting of platinum resistant ovarian cancer, exploring the antitumor activity of autologous CIK against chemoresistance patient-derived ovarian cancer cell lines (pdOVC).

Generation and characterization of patient-derived ovarian cancer cell lines (pdOVC). pdOVC
were successfully established from patients with advanced EOC in 4 to 12 weeks; characteristics of the 9 patients are shown in Table 1. Notably, 6 out of 9 pdOVC cultures were established from relapsed heavily treated chemotherapy resistant patients, while the remaining 3 were established from first relapse.
Interestingly, following the treatment in vitro with carboplatin, the residual pdOVC presented a lower intensity of HLA-I expression (Fig. 4C,D), providing further relevance to the potential benefit in clinical perspective from an HLA-independent approach with CIK immunotherapy.   www.nature.com/scientificreports www.nature.com/scientificreports/ In vivo activity of CIK against ovarian cancer. The antitumor activity of CIK was explored in vivo within a patient-derived high grade serous ovarian cancer xenograft model (PDX) obtained from an omental node of a chemo-naïve 75 years old patient 54 . ϒ null-NOD/SCID mice were subcutaneously injected with the PDX ovarian tumor sample (125 mm 3 , see methods). Starting 2 weeks after tumor implantation, mice (n = 6) were treated with 2 weekly intravenous infusions of mature CIK (1 × 10 7 ) for 5 weeks; mice injected with PBS alone were used as the untreated control (n = 6). Adoptive immunotherapy with CIK significantly reduced tumor cell viability, evaluated by tumor-glucose uptake in vivo, compared to untreated controls (p = 0.0022) (Fig. 5). Mice were sacrificed 10 days after the last CIK infusion observing i) higher rates of necrotic areas in treated mice compared with untreated controls (28% vs 14%, p < 0.0001, panel A and B, Fig. 6) and ii) detectable rates of human tumor infiltrating CIK (CD3+, 35% ± 5), mostly located near the necrotic tissue in treated tumors (panel C, Fig. 6).

Discussion
Our study reports the preclinical activity of HLA-independent CIK immunotherapy within patient-derived ovarian cancer models, including platinum resistant targets.
Novel surgical techniques and new drugs such as PARP inhibitors and antiangiogenics significantly improved EOC prognosis. Nevertheless, overall clinical outcome, especially after failure of platinum-based chemotherapy, is still severe and novel therapeutic strategies are largely awaited 3,55-57 . In recent years preclinical studies already suggested the efficacy of CIK against ovarian cancer cell lines in vitro and in vivo 51 , underscoring the role of NKG2D in tumor recognition and possible synergism with bispecific antibodies 58 .
We confirmed the importance of NDG2D ligands in this setting, reporting their high expression in pdOVC generated from metastatic tumors relapsed after platinum treatment.
It is worth to note however that blocking the NKG2D receptor on CIK reduced but not completely abrogated their killing activity, suggesting the activity of other molecules (e.g. DNAM, LFA-1) partially contributing to the observed antitumor effect. A deeper comprehension of the hierarchy and role of all CIK ligands is warranted as it could help the identification of predictors of response and, as a consequence, subsets of patients that could better benefit from CIK-based immunotherapy.
Early evidence from clinical trials support the feasibility and safety profile of CIK immunotherapy, with initial signs of activity and positive impact on survival outcomes in multiple solid tumors, including ovarian cancer either in advanced 31,34,38,42,[59][60][61] or even adjuvant or post-chemotherapy maintenance settings 62 .
A specific, clinical relevant, issue for translational researches is exploring the killing activity of CIK against relapsed chemo-resistant EOC targets.
Our preclinical model is mainly developed within this frame, with the intent of addressing the emerging need for reliable translational platforms in clinical perspective. The experiments with autologous tumor samples are important in the effort of recapitulating the intrinsic, and mostly unknown, biologic elements that regulate the interface between immune effectors and cancer targets. Patient-derived tumor targets in our work were obtained from metastatic ascites. This could be an important aspect, representative of a realistic clinical scenarios. Metastases may have indeed important biologic and immunogenic differences compared with the primary tumors and, in the hypothesis of clinical translations, CIK immunotherapy will certainly be tested in patients with advanced metastatic disease.
Circulating CIK precursors were obtained directly from PBMC of patients with high grade serous EOCs. Importantly, we observed an intense, ex vivo expansion rates that were not affected by previous or concomitant platinum based treatments. This supports the concept of a procedure that is clinically exploitable, especially in heavily pretreated patients with platinum resistant EOC. www.nature.com/scientificreports www.nature.com/scientificreports/ In a clinical perspective, the safe, simple and cost-effective protocol to generate CIK is a valuable issue; this makes CIK strategy compare favorably with other approaches currently under investigation, that include extensive lymphocyte manipulation or genetic engineering. CIK precursors for clinical use may be obtained by leukapheresis or even limited peripheral blood withdrawals, easily repeatable for patients with lower CIK expansion rates that might undergo multiple CIK ex vivo expansion processes. In preparation to clinical trials, the protocol to generate CIK has been recently validated by our and other groups in GMP conditions 63 .
The activity of patient-derived CIK was also confirmed in vivo within an ovarian cancer xenograft model 54 .
We performed multiple CIK infusions, hypothesizing that a similar schedule might be replicated also in a hypothetic clinical study. In a clinical prospective, considering the very favorable safety profile of CIK immunotherapy, multiple infusions could be pursued to provide a stronger effect. This animal model, however, was based on a PDX generated from a surgical biopsy collected before the chemotherapy treatment. A more realistic simulation of post-platinum treatment was instead simulated in vitro, observing the effective activity of patient-derived CIK versus pdOVC targets that survived a treatment with therapeutic doses (IC 50 ) with carboplatin. Of note, the stress-inducible NKG2D ligands trended to increase after platinum treatment, supporting the observed CIK cytotoxicity and providing rational to explore combinatorial/sequential approaches with chemotherapy. In conclusion, in the continuously evolving landscape of EOC, where targeted therapies such as PARP inhibitors are acquiring increasing relevance, there is still room for immunotherapy in multiple clinical settings. However, with the exceptions of limited subsets of tumor histotypes (i.e. clear cell OC), checkpoint inhibitors have not replicated the exciting results obtained in other cancers, framing the rationale to explore alternative or complementary strategies like adoptive immunotherapy.
Probably, considering general immunologic axioms suggesting that low tumor burden may elicit a better immunologic control, the optimal clinical setting to develop CIK immunotherapy is probably that of early disease (e.g. after primary or interval cytoreduction) or non-bulky disease relapse.
CIK activity is not impaired by a restrained neoantigen load, as it seems to be the case in most ovarian cancers, nor by intrinsic defects in antigen presentation, HLA integrity or interferon pathways all reported factors linked to failure of checkpoint inhibitors. Furthermore, CIK are ultimately activated T lymphocytes presenting variable expression of PD-1 and other checkpoints. It is conceivable, and currently under investigation in other malignancies, that their antitumor activity may synergize with checkpoint modulators antibodies.   . (A,B) The infusion of CIK lymphocytes determined higher rates of tumor necrosis compared to untreated controls. Representative images (squares indicate necrotic areas) and cumulative data reported (evaluation by color deconvolution method). (C) Representative images (human CD3 IHC staining, Magnification 10×) of CIK tumor infiltration in vivo.

Materials and Methods
Generation and ex-vivo expansion of CIK. Peripheral blood mononuclear cells (PBMC) were obtained, by density gradient centrifugation with Lymphoprep (Sentinel Diagnostic), from patients affected by EOC diagnosed at advanced stage at Candiolo Cancer Institute, Fondazione del Piemonte per l'Oncologia (FPO)-IRCCS (Candiolo, Torino, Italy). All individuals provided their written informed consent.
Phenotype of CIK was analyzed by standard flow cytometric assays. The following monoclonal antibodies We also tested commercially available ovarian cancer cell lines: OVCAR-3, OVCAR-5, IGROW-1, A2780 e OAW42 obtained from American Type Culture Collection (ATCC). All cell lines have been characterized by the provider and maintained as suggested. Phenotypical analysis was conducted as previously described.
In vitro assays of cytotoxicity of patient-derived CIK. The tumor-killing ability of patient-derived CIK was assessed in vitro against ovarian cancer cell lines (either autologous or allogeneic).
Whenever possible the antitumor activity of patient derived CIK was additionally confirmed by cytofluorimetric analysis. Target cells were stained in accordance with the manufacturer's protocol with the vital dye CFSE (5, 6-carboxyfluorescein diacetate succinimidyl ester; Molecular Probes) or PKH26 Red Fluorescent Cell Linker kit (Sigma Aldrich). The immune-mediated killing was determined evaluating cell viability by flow cytometry (Cyan ADP, Dako), after 24 hours incubation with expanded CIK cells, as previously described, according to the formula: experimental − spontaneous mortality/(100 − spontaneous mortality) × 100. Cytotoxicity was calculated through flow cytometry (Cyan ADP, Beckman Coulter s.r.l.) using DAPI permeability assay (Thermo Fisher Scientific) of target cells. As descriptive statistic al analysis, medians and ranges, mean ± SEM were used as appropriate; statistical significance was expressed as true P value. All P < 0.05 were considered statistically significant. Statistical analysis was conducted using software Graph Pad Prism 8.0.
In vivo antitumor activity of patient-derived CIK. EOC patient derived xenograft (PDX) model of was carried out as previously described 54 . Briefly, we implanted subcutaneously in right flank of severely immunocompromised NOD/Shi-scid/IL-2R γnull mice PDX line tumor samples of 125 mm 3 .Tumor size was evaluated twice-weekly with digital caliper and volume was calculated using the formula 4/3π*(d/2) 2 *D/2, where d is the minor tumor axis and D is the major tumor axis. All animal procedures were approved by the institutional competent Committee.
The analysis of tumor metabolic activity was performed using Fluorescent probe XenoLightRediJect 2-DeoxyGlucosone (DG)-750 (Caliper Life Sciences, USA). Fluorescent probe was injected through the mouse tail vein. The tumor location and glucose accumulation was acquired by IVIS ® SpectrumCT (Perkin Elmer, Waltham, MA, USA) 4 hours after injection of 2-DG750 and analyzed using Living Image Software (Perkin Elmer, Caliper Life Sciences). To calculate metabolic activity of tumor mass, we acquired 3D images and the ratio between total fluorescence emitted and total number of tumor mass voxels, was calculated. Also in this case, we represented the mean value ± SEM. Statistical analysis was performed by using T software Graph Pad Prism 8. The evaluation of necrotic versus alive cells area was conduct considering the color deconvolution method (acquisition of 10 field for each sample, 10X of magnitude).
Eosin/hematoxylin-stained cells were digitally separated using ImageJ software (version 1.46c; WS Rasband, National Institutes of Health, Bethesda, MD, USA,) and an ImageJ plugin for color deconvolution 64 . The hematoxylin deconvoluted image was subjected to histogram analysis to identify nuclei of alive cells. The total area of alive nuclei were subtracted from estimated total field area (measured and calculated as average in 4 field fully covered of alive cells and used ad reference data) and translate in percentage. Significant Statistical analysis was conducted using software Graph Pad Prism 8.The collection of clinical samples was undertaken with the understanding and written consent of each subject according to the PROFILING Protocol, conformed to the standards set by the Declaration of Helsinki and approved by the Regione Piemonte Ethical Committee (approval no 5141 on 9/3/2011) and then by the IRCCS Ethical Committee (approval no. 192/2016 on 19/7/2016).
All animal procedures were approved by the local Ethical Commission and by the Italian Ministry of Health in accordance with EU Directive 2010/63/EU for animal experiments; first authorization was sent on 12/7/2012 and, following subsequent regulations, approved on 14/01/2016 (no. 16/2016-PR) and extended for two additional years on 17/9/2018. immunohistochemistry. 5 μm in thickness FFPE tissue sections were utilized to visualize morphology of tumor tissues via H&E staining and to detect human lymphocytes with CD3 antibody (DAKO, Carpinteria, CA), for CD3 staining, deparaffinization, rehydration and target retrieval were performed following Dako instructions. To quantify data, at least 10 images of each sample were acquired by optical microscope (20×) connected with charge-coupled device (CCD) camera and analyzed by using automatic counter software (NIH ImageJ, W. Rasband, NIH) software. The percentage of necrosis was estimated calculating the difference between a "standard control area" covered by healthy cells and the treated and untreated groups. To define "standard control area" we considered 5H&E fields in which all the area was occupied by round and clearly healthy cells. To calculate the percentage of human lymphocytes in tumors, we compared in the number of all nuclei present in the field versus the number of CD3 positive cells (which are completely absent in non treated tumors).