Purpose: NKG2D ligands (NKG2DL) are expressed on various tumor types and immunosuppressive cells within tumor microenvironments, providing suitable targets for cancer therapy. Various immune cells express NKG2D receptors, including natural killer (NK) cells and CD8+ T cells. Interactions between NKG2DL and NKG2D receptors are essential for NK-cell elimination of osteosarcoma tumor-initiating cells. In this report, we used NKG2D–NKG2DL interactions to optimize an immunotherapeutic strategy against osteosarcoma. We evaluated in vitro and in vivo the safety and cytotoxic capacity against osteosarcoma cells of CD45RA− memory T cells expressing an NKG2D-4-1BB-CD3z chimeric antigen receptor (CAR).
Experimental Design: CD45RA− cells from healthy donors were transduced with NKG2D CARs containing 4-1BB and CD3z signaling domains. NKG2D CAR expression was analyzed by flow cytometry. In vitro cytotoxicity of NKG2D-CAR+ CD45RA− T cells against osteosarcoma was evaluated by performing conventional 4-hour europium-TDA release assays. For the in vivo orthotopic model, 531MII YFP-luc osteosarcoma cells were used as targets in NOD-scid IL2Rgnull mice.
Results: Lentiviral transduction of NKG2D-4-1BB-CD3z markedly increased NKG2D surface expression in CD45RA− cells. Genetic stability was preserved in transduced cells. In vitro, NKG2D-CAR+ memory T cells showed significantly increased cytolytic activity than untransduced cells against osteosarcoma cell lines, while preserving the integrity of healthy cells. NKG2D-CAR+ memory T cells had considerable antitumor activity in a mouse model of osteosarcoma, whereas untransduced T cells were ineffective.
Conclusions: Our results demonstrate NKG2D-4-1BB-CD3z CAR–redirected memory T cells target NKG2DL-expressing osteosarcoma cells in vivo and in vitro and could be a promising immunotherapeutic approach for patients with osteosarcoma. Clin Cancer Res; 23(19); 5824–35. ©2017 AACR.
Chimeric antigen receptors (CAR) are being increasingly considered as anticancer treatment. We previously demonstrated an essential role of NKG2D receptor and NKG2D ligand interactions for NK-cell elimination of osteosarcoma tumor-initiating cells. In an effort to develop new therapies for patients with osteosarcoma, we tested the cytotoxic ability of CD45RA− memory T cells expressing an NKG2D-4-1BB-CD3z CAR in vitro and in an orthotopic osteosarcoma murine model. We found that NKG2D-4-1BB-CD3z–expressing CD45RA− memory T cells had enhanced cytotoxicity against osteosarcoma cells in vitro compared with untransduced T cells. In vivo, mice receiving NKG2D CAR-redirected memory T cells showed limited tumor growth and prolonged survival. In addition, NKG2D CAR–redirected memory T cells showed no lytic activity against healthy cells, and no chromosomal aberrations due to lentiviral transduction. Hence, we provide evidence to support our hypothesis that NKG2D CAR–redirected memory T cells are effective and safe for osteosarcoma control.
Osteosarcoma is the most frequently observed malignant primary bone tumor, typically affecting children and adolescents (1). The long-term survival rate for the nonmetastatic type varies from 65% to 70%. However, current combined surgical and neoadjuvant chemotherapy are inefficient for the metastatic type, in which the 5-year survival rate is less than 20% (2). The rapid development of metastatic lesions and resistance to chemotherapy are the major causes of treatment failure and poor survival rates. Over the past few decades, efforts have been made to intensify dosing, vary timing, and to use multicombinational chemotherapy; however, no substantial improvement in survival has been achieved. In addition, several adverse effects accompany high-dose chemotherapy (3). Therefore, the development of alternative novel therapies is essential.
Immune system effector cells have the potential to attack and eliminate cancer cells. Primary and metastatic osteosarcoma cells heterogeneously express ligands for NKG2D receptor (4, 5). In a previous study, we demonstrated the efficacy of natural killer (NK) cell–based immunotherapies against osteosarcoma in preclinical studies (5). We observed that activated and expanded NK cells (NKAE) target osteosarcoma tumor-initiating cells (TIC) using NKG2D receptor and NKG2D ligand (NKG2DL) interactions, both in vitro and in an orthotopic osteosarcoma murine model. We reported that exposing osteosarcoma cells to NKAE cells in vitro decreased both their capacity to form sarcospheres and their TIC compartment. In an in vivo model, NKAE cells delayed tumor growth and inhibited development of lung metastases (5). The preclinical success of this immunotherapeutic approach, the importance of NKG2D-NKG2DL interactions and the clinical limitations of NK-based therapies that we and others have observed (the need to infuse large numbers of cells, dispersion of therapy, absence of memory, limited in vivo expansion; refs. 6–8) encouraged us to explore the efficacy of an NKG2D chimeric antigen receptor (CAR) to target osteosarcoma.
CARs are artificial receptors that combine the specific recognition of tumor-associated antigens (TAA) with T-cell–activating signals (9). CAR specificity is typically based on antibody single-chain fragment variable (scFv) regions or T-cell receptor (TCR)–binding domains; however, NK-cell receptors have also been used (10–15). The NKG2D receptor and its ligands have attracted considerable interest as a potential strategy to attack tumor cells as NKG2DL are expressed on most types of tumor cells and they demonstrate relative selectivity of ligand expression on tumor cells compared with healthy cells (16). Endogenous NKG2D receptors recognize NKG2DL expressed on healthy tissues without triggering an immune response.
CARs are usually expressed on autologous T cells; however, NK cells (12, 17) or different T-cell subsets (18, 19) can also be used. Most clinical trials use autologous T cells to express CARs; however, T cells from some patients fail to grow, rendering this approach useless for patients with rapid disease progression. Allogeneic T cells expressing CARs have also been infused into patients after a hematopoietic stem cell transplant (HSCT) from the same healthy donor (20, 21). However, the universal availability of large numbers of healthy donor T cells to express CARs and their infusion into patients without the requirement of a prior HSCT would be major challenges of CAR–T-cell immunotherapy. One potential complication of the use of allogeneic T cells is the undesirable graft versus host (GvH) reactions. To avoid GVH effects, T cells without a reactive T-cell receptor (TCR) can be used. One method to enrich for non-allogeneic T cells is by using antigen-experienced memory T cells for CAR transduction. One of the markers that can distinguish memory from naïve T cells is CD45RA. CD45RA is expressed on naïve T cells, whereas CD45RO is expressed on memory T cells (22). CD45RA+ naïve T cells have high potential for alloreactivity against recipient-specific antigens upon adoptive transfer, causing clinical graft versus host disease (GvHD; refs. 23, 24). In contrast, CD45RA−CD45RO+ T cells exert a memory response to prior pathogens or vaccines and can mediate graft-versus-tumor effects without inducing GvHD (25, 26).
Although CAR-T therapies have shown remarkable clinical responses, they can also over activate the immune system and cause harm. One life-threatening complications of CAR T-cell therapies is cytokine release syndrome (CRS) that is related to T-cell activation and expansion. To diminish the risk of CRS, we propose to use memory T cells to express an NKG2D CAR. This T-cell subpopulation has been demonstrated to maintain allogeneic T-cell immunity producing less CRS and to lack alloreactive response, mild “GvHD” like (27).
In this report, we provide preclinical data to support our hypothesis that CD45RA− memory T cells expressing an NKG2D-4-1BB-CD3z CAR could be effective and safe for the treatment of osteosarcoma. We found that NKG2D CAR-redirected CD45RA− cells lysed osteosarcoma cells in vitro, and they reduced tumor burden and prolonged survival in mouse models of metastatic osteosarcoma. Furthermore, they had no GvHD activity and showed no lytic activity against healthy tissues.
Materials and Methods
The 531MII primary cell line was provided by Dr. Patiño-García (University Clinic of Navarra, Spain) and was cultured in minimum essential medium (GIBCO, 22571-020) supplemented with 10% heat-inactivated FBS (GIBCO, 10270-098) and penicillin-streptomycin (P/S; GIBCO, 15140-122). The MG-63 (ATCC CRL-1427) and U-2 OS (ATCC HTB-96) osteosarcoma cell lines were purchased from the ATCC. The MG-63 cells were cultured in Eagle minimum essential medium (EMEM; ATCC, 30-2003) and U-2 OS cells were cultured in McCoy 5A Medium (ATCC, 30-2007), both supplemented with 10% FBS and P/S. For the in vivo experiments, 531MII cells were transduced with gammaretroviral vector containing MSCV-luciferase-IRES-YFP (Vector Laboratory). All cell lines were routinely tested for Mycoplasma by using the LookOut Mycoplasma PCR Detection Kit (MP0035-1KT) and Jump Start Taq Polymerase (D9307), both from Sigma-Aldrich, following the manufacturer's instructions.
Healthy primary cells and cell lines
Peripheral blood mononuclear cells (PBMC) and bone marrow samples were collected from healthy donors with approval from the St Jude Children's Research Hospital, Hospital La Paz and Hospital 12 de Octubre institutional review boards and written informed consent from the donors. Hematopoietic progenitors (CD34+ cells) were sorted by labeling of cells with anti-CD34 (Supplementary Table S1) and a FACS AriaFusion flow cytometer (BD Biosciences). Hs 1.Int (ATCC, CRL7820) normal fetal human intestine fibroblasts and CCD-18 Co (ATCC, CRL1459) normal human colon fibroblasts were purchased from ATCC. Hs 1.Int cells were cultured in EMEM and 10% FBS. CCD-18 Co cells were cultured in DMEM (GIBCO 61965-026) and 10% FBS. NuLi-1 (ATCC, CRL-4011) normal human lung epithelial cells were kindly provided by Dr. Paz-Ares (Hospital 12 de Octubre), and cultured in a serum-free medium: Airway Epithelial Cell Basal Medium (ATCC, PCS-300-030) supplemented with Bronchial Epithelial Cell Growth Kit (ATCC, PCS-300-040) additives.
Cell isolation and vector production
For the depletion of CD45RA+ cells, PBMCs were labeled with CD45RA microbeads (Miltenyi Biotec, 130-045-901) and depleted using CliniMACS or AutoMACs (both from Miltenyi Biotec) following the manufacturer's instructions. T cells were isolated by using the Pan T-cell Isolation Kit (Miltenyi Biotec, 130-095-535) following the manufacturer's instructions and AutoMACs device. The purity of the CD3-positive and CD45RA-negative fractions was always ≥99%, as determined by flow cytometry. The HL20i4r-MNDantiCD19bbz lentiviral vectors were derived from the clinical vector CL20i4r-EF1a-hgcOPT27 but expressed an NKG2D CAR. The anti–CD19-41BB-CD3z CAR designed by Imai and colleagues (28) was used as backbone to build the NKG2D CAR construct. It contained the extracellular domain of NKG2D (designed by Wai-Hang Leung and Wing Leung), the hinge region of CD8a and the signaling domains of 4-1BB and CD3Z. The cassette was driven by a prMND.
The viral supernatant was produced by transient transfection of HEK293T cells with the vector genome plasmid and lentiviral packaging helper plasmids pCAGG-HIVgpco, pCAGG-VSVG, and pCAG4-RTR2. The supernatant was concentrated by ultracentrifugation and titrated on HeLa cells by serial dilution followed by a quantitative PCR to determine vector genome copy number.
Cells were primed at 2 × 106 cells/mL overnight in X-VIVO-15 (Lonza, 04-418Q) supplemented with 250 IU/mL IL2 (130-097-746, Miltenyi Biotec), 10 ng/mL anti-CD3 (clone OKT3; 317303) and 10 ng/mL anti-CD28 antibody (clone 28.2; 302913), both from BioLegend. The following day, transduction was performed on RetroNectin (T100B; Takara Bio, Clontech Laboratories) coated plates using MOI = 2. The cells were harvested 4 days later for experiments.
Cell phenotypes were analyzed by flow cytometry (FCM). Details of the different fluorochrome-labeled mAbs are summarized in Supplementary Table S1. Analyses were performed with BD FACSCanto II or LSR Fortessa (BD Biosciences). The mean fluorescence intensity (MFI) ratio was defined as the MFI of the specific staining relative to the MFI of the appropriate isotype control staining. FlowJo vX.0.7 software (TreeStar) was used for data analysis.
The cytotoxicity of NKG2D CAR-expressing T cells against OS cells was evaluated at 20:1, 10:1, 5:1 and 2.5:1 effector-to-target (E:T) ratios by performing conventional 4-hour europium-TDA release assays (PerkinElmer; AD0116) as described previously (5, 29). In another set of experiments, the cytotoxicity of NKG2D CAR-expressing T cells was tested against PBMCs, bone marrow–derived mesenchymal stem cells (MSC) from healthy donors, Hs1.Int, CCD-18 Co and NuLi-1 healthy cell lines. To rule out NKG2D CAR-T cells could cause hematologic toxicity, colony-forming assays were performed. CD34+ cells were cocultured with CD45RA−NKG2D-CAR+ at 20:1, 10:1, 5:1, and 1:1 E:T ratios, or cultured alone (control) for 2 hours at 37°C and 5% CO2. Cells were then suspended in Methylcellulose (Methocult Express, StemCell Technologies, 04437) and seeded in triplicates, in a humid chamber to avoid methylcellulose dehydration. Plates were incubated at 37°C and 5% CO2 for 10 days, and colonies were counted. Representative images were acquired in a transiluminator G:BOX Chemi XX6 transiluminator (Syngene, Synoptics). GeneSys software was used for analysis (GeneSys image acquisition software, Syngene, Synoptics).
Allogeneic mixed leukocyte reaction
To compare the alloreactivity of CD45RA+ and CD45RA− cells, mixed leukocyte reactions (MLR) were performed. Responder cells were seeded at 1 × 105 cells per well into 96-well flat-bottom plates in triplicate. PBMCs from an unrelated third party were γ-irradiated with 3,000 rad and mixed with responder cells at a 10:1 responder/stimulator ratio. Wells containing responders or stimulators alone were used to set the background signals. Cell proliferation was measured by the incorporation of bromodeoxyuridine (BrdUrd) using the DELFIA Proliferation Kit (PerkinElmer; AD0200), according to the manufacturer's instructions. A 1420 VICTOR3 Multilabel counter plate (PerkinElmer) was used for analysis. The production of IFNγ in the supernatant of the MLRs was measured by the Human IFN-γ Flex set (CBA; BD Biosciences; 560111), following the manufacturer's instructions. The data were collected and analyzed by flow cytometry using a BD FACSCalibur flow cytometer (BD Biosciences).
In vivo studies
Two different in vivo assays were performed at St. Jude and the Instituto de Investigaciones Biomédicas Albert Sols (IIB). All the procedures were approved by the St. Jude Animal Care and Use Committee and by the ethics committee of the Consejo Superior de Investigaciones Científicas (CSIC) and the Comunidad de Madrid (CAM), respectively. In both experiments, 531MII YFP-luc cells (5 × 105) were engrafted by injection through the tibial plateau in the primary spongiosa of both tibias of 10- to 12-week-old NOD/scid IL2rgnull (NSG) mice with or without T cells (5 × 106 T cells/tibia). Human IL2 (100 IU/mouse, Miltenyi Biotec) was intraperitoneally injected starting one day after injection of effector cells and continued every other day for 3 weeks. In the first assay, we explored the efficacy of an NKG2D CAR–T-cell therapy against osteosarcoma. The mice were divided into 3 treatment groups: Group A (control), receiving 531MII YFP-luc osteosarcoma cells alone; Group B, receiving 531MII YFP-luc cells and 5 × 106 untransduced CD45RA− cells; and Group C, receiving 531MII YFP-luc cells and 5 × 106 NKG2D-CAR transduced CD45RA− cells. Tumor progression was regularly evaluated using a Xenogen IVIS-200 system (124262, PerkinElmer Life and Analytical Sciences), with imaging beginning 5 minutes after intraperitoneal injection of an aqueous solution of d-luciferin potassium salt (3 mg/mouse). Photons emitted from luciferase-expressing cells were quantified using the Living Image 3.0 software (Xenogen). At day 60 postinfusion, survivor mice were rechallenged with an intravenous injection of 1 × 105 of 531MII YFP-luc cells. The mice were euthanized when they displayed signs of significant osteosarcoma symptoms, including paralysis or tumor volumes larger than 1,000 mm3. In the second assay, we studied CAR–T-cell persistency and specific symptoms of NKG2D-CAR–related toxicity. The mice were divided into 3 treatment groups: group A (control), with no treatment; group B, receiving 5 × 106 CD45RA−NKG2D-CAR+ cells injected intratibially (IT), and group C, receiving 5 × 106 CD45RA−NKG2D-CAR+ cells injected intravenously. Weight loss was monitored every other day until day+21. At day 21 mice were sacrificed and samples were collected. To explore CAR–T cells persistence on treated mice, blood was analyzed by FCM at day+14 and +21. To evaluate hepatotoxicity in the serum of treated mice, the levels of albumin (ALB), alkaline aminotransferase (ALT), alkaline phosphatase (ALP), bile acids (BA), cholesterol (CHOL), gamma glutamyl transferase (GGT), total bilirubin (TBIL), and urea nitrogen (BUN) were measured by using VetScan Mammalian Liver profile and VetScan Chemistry Analyzer, both from ABAXIS (500–1,040) and (200–7,013). Serum from healthy and untreated NSG mice were used as controls.
Genomic stability assays
To test whether lentiviral transduction could cause any chromosomal abnormality, 5 million NKG2D-CAR–transduced memory T cells were harvested at days 14 and 30 after transduction for cytogenetic analysis in two separate experiments. Five million untransduced CD45RA− cells from the same donors were used as controls (20 metaphase cells of each type were examined per experiment). Comparative genomic hybridization (CGH) was performed to rule out chromosomal aberrations in the genomic DNA of NKG2D CAR and untransduced CD45RA− cells. DNA was isolated from cell cultures using the AllPrep DNA/RNA Micro Kit from Qiagen (80284) and frozen at −80°C until analysis. CGH microarray was performed using Agilent's eArray (Agilent Technologies) probe library in a custom high-resolution format of 8 × 60K that is enriched in 320 genes related with hematologic neoplasm. The average resolution of the probe coverage was 1 oligo every 175 kb to detect losses or gains of more than 400 kb. For target regions, 1 oligo every 9 kb to identify losses or gains more than 25 kb. Array experiments were performed as recommended by the manufacturer (5190-3399, G2534-60003, G5188-5222, all from Agilent Technologies). DNAs (500 ng) from the specimen and a sex-matched DNA control (G1521, Promega) were double-digested with RsaI and AluI for 2 hours at 37°C. After heat inactivation of the enzymes, samples were labeled by random priming using Genomic DNA enzymatic labelling kit (Agilent 5190-0449) for 2 hours using Cy5-dUTP for donor's DNAs and Cy3-dUTP for control DNA. Hybridization was performed at 65°C with rotation for 24 hours. After two washing steps, the array was analyzed with the DNA Micro array scanner with Surescan high-resolution technology (Agilent Technologies) using the Feature Extraction software (v9.1 Agilent Technologies) The data were analyzed with Aberration Detection Method 2 (ADM-2) algorithm and the minimum number of consecutive oligos has been established in 3 to consider a copy-number variation.
Results are shown as means ± SEM. Statistical difference between two groups was calculated using the Student t test. The Student paired t test was used when untransduced or transduced T cells from the same donor were compared. In the mouse model, survival was estimated by using the univariate Kaplan–Meier method and compared by using the log-rank test. Statistical significance was defined as P ≤ 0.05.
Osteosarcoma cells express NKG2DL
Cell surface expression of ligands for the NKG2D receptor on the 531MII, U-2 OS, and MG-63 osteosarcoma cell lines was assessed by flow cytometry. We found low to very high levels of expression of all NKG2DLs (MICA, MICB, or ULBP1, 2 or 3; Fig. 1). Other primary bone and metastatic osteosarcoma cell lines had been previously tested for NKG2DL expression in our laboratory, and showed similar expression patterns (5).
CD45RA− fraction contains memory T-cell subsets and shows minimal alloreactivity
The phenotype of cells contained in the CD45RA− fraction was analyzed by flow cytometry. We found CD4+ subset counted for 74.8%, whereas CD8+ subset counted for 14.48% of the total cells. Most CD4 and CD8 cells showed effector memory and central memory phenotypes. CD4- and CD8-naïve T cells were less than 0.2% of the total cells (Supplementary Table S2).
To test the alloreactivity of CD45RA+ and CD45RA− cells (responders), they were cocultured with irradiated PBMCs from the same unrelated donor (stimulator). The CD45RA+ cells proliferated much more than the CD45RA− counterparts in the MLR assays (Fig. 2A). Furthermore, only CD45RA+ cells exhibited high IFNγ production in the culture supernatant (Fig. 2B).
NKG2D CAR transduction and expression
Untransduced CD45RA− cells already express endogenous NKG2D receptor and were used as a control of baseline expression. After transduction, expression of the NKG2D receptor increased both in percentage of positive cells and in MFI. To ensure NKG2D CAR expression was stable, and that transduction did not affect cell survival, we monitored viability and NKG2D expression by flow cytometry at different time points after transduction. We found the transduced cells had similar viability as the untransduced cells, and they maintained a higher expression of NKG2D (both in percentage and in MFI) up to 20 days after transduction (Fig. 3).
Cytotoxicity of NKGD CAR-expressing CD45RA− cells
To test the cytotoxic capacity of transduced and untransduced cells against osteosarcoma, we used 531MII (n = 8), MG-63 (n = 5), and U-2 OS (n = 6) osteosarcoma cells as targets. The NKG2D CAR-expressing CD45RA− cells showed significantly higher cytotoxic capacity against osteosarcoma cells compared with the untransduced CD45RA− cells (P = 0.0009, P = 0.001, P = 0.007 and P = 0.007 for ratios 20:1, 10:1, 5:1 and 2.5:1, respectively, in 531MII cells; P = 0.0019 for ratio 20:1 in MG-63 cells, and P = 0.0034 and P = 0.044 for ratios 20:1 and 10:1 in U-2 OS cells, Fig. 4). We did not observe differences on cytolytic ability of different NKG2D-CAR T-cell subsets (Supplementary Table S3). There was no apparent relation between NKG2DL expression and sensitivity to NKG2D-CAR T cells' cytotoxicity.
Antitumoral effect of NKG2D-CAR+ CD45RA− cells in an orthotopic osteosarcoma murine model
Having shown the ability of NKG2D CAR+ CD45RA− cells to target osteosarcoma cells in vitro, we tested their in vivo efficacy in a xenograft orthotopic osteosarcoma murine model. Compared with the control groups corresponding to mice receiving osteosarcoma cells (531MII) alone or in combination with untransduced CD45RA− cells, the mice receiving NKG2D CAR+ CD45RA− cells and osteosarcoma cells (E:T ratio, 10:1) showed lower tumor burden, as seen with significantly lower average dorsal and ventral bioluminescent signals up to 56 days postinfusion (Fig. 5A and B). At day 60 postinfusion, we rechallenged all mice groups with an intravenous injection of osteosarcoma cells, and could observe a significant extended survival time only for those mice receiving NKG2D CAR+ cells (120 days). All other mice from the control groups died from tumor burden within 2 weeks of the rechallenge (70.4 days for the untreated control group and 83.8 days for the group receiving CD45RA− untransduced cells; Fig. 5A and C). 531MII YFP-luc cells were only found in the lungs of mice from control group or those treated with untransduced CD45RA− cells (Supplementary Fig. S1). We followed up these mice up to 120 days and did not detect treatment-related toxicity as CRS or any GVHD symptoms. Different parameters indicative of toxicity related to therapy and NKG2D-CAR T cells persistency were explored. Mice treated with NKG2D-CAR memory T cells showed similar weight curves than those in the control group (Fig. 5D). In addition, biochemical analysis of the serum showed no differences in the level of the studied liver enzymes between treated and control groups, indicating no liver toxicity is produced by the therapy either orthotopically or intravenously injected (Supplementary Table S4). Human CD45+NKG2D-CAR T cells were detectable in the peripheral blood of treated mice 14 and 21 days after the therapy was administered either locally or systemically (Supplementary Fig. S2).
Safety of NKG2D CAR–expressing memory T cells
Genome instability was studied by karyotype or by array-based CGH. Normal karyotype and no copy number alterations were found in NKG2D-CAR+ memory T cells compared with untransduced T cells up to 30 days after transduction (Fig. 6A).
To verify NKG2D CAR+ memory T cells showed no lytic activity against healthy tissues, PBMCs and MSCs from healthy donors, Hs 1. Int, CCD-18 Co and NuLi1 healthy cell lines were used as targets for in vitro cytotoxicity assays. No specific lytic activity was detected against healthy PBMCs, MSCs and CCD-18 Co. However, NKG2D CAR+ memory T cells showed high cytotoxic activity against Hs 1. Int. cells, and a low level of cytotoxicity against NuLi-1 cell line (Fig. 6B). To rule out NKG2D-CAR memory T cells could cause hematologic toxicity, colony-forming assays were performed after culture of healthy CD34+ progenitor cells alone (control) or at different E:T ratios. We observed the ability of CD34+ cells to form colonies was unaffected after coculture with NKG2D-CAR redirected memory T cells, indicating no hematologic toxicity was caused by the therapy (Fig. 6C). Expression of NKG2DL on healthy cells is shown in Supplementary Table S5.
Despite multimodal therapy, osteosarcoma outcomes have not improved over the past 3 decades, with patients with metastatic disease having an overall survival of only 20%. CAR–T-cell–based immunotherapy is a promising therapeutic approach that is currently being tested in several hematologic malignancies as well as in solid tumors, including osteosarcoma, opening a new possibility for those cancer patients resistant to traditional therapies.
The NKG2D cell receptor and its ligands are currently gaining interest as a potential strategy to target tumor cells. NKG2DL is expressed in the vast majority of tumor types, including osteosarcoma (4, 5), whereas its expression on healthy cells is rare (16). In fact, NKG2D receptor-NKG2DL interactions have been shown to be essential for NK-cell targeting of osteosarcoma TICs (5). In this study, we took advantage of NKG2D–NKG2DL interactions to explore the efficacy and safety of NKG2D–CAR–redirected memory T cells as immunotherapy for osteosarcoma treatment.
NKG2D CAR–redirected T and NK cells have shown potent anticancer effects against various tumors such as T-cell lymphoma (30), multiple myeloma (31) osteosarcoma, Ewing sarcoma, ovarian, and pancreatic carcinoma (11, 12, 32, 33). In addition, NKG2D CAR–redirected effector cells are able to target NKG2DL expressed not only on the tumor cells, but also in the immunosuppressive cells within the tumor microenvironment, inducing an immunologic response in the host (34). In the current report, we show for the first time that NKG2D-CAR–redirected memory T cells target osteosarcoma cells both in vitro and in vivo. One potential complication of the use of NKG2D CAR T-cell therapies could be the “on target–off tumor” toxicity, given NKG2DLs are not expressed exclusively on tumor cells. Recently, VanSeggelen and colleagues (35) described lethal toxicity in mice treated with murine NKG2D CAR T cells. This toxicity was both CAR construct and strain-dependent, and was exacerbated upon lymphodepleting conditioning with cyclophosphamide. In this study, T cells expressing a construct containing DAP10 showed the highest CAR expression and toxicity, followed by a second-generation CAR containing CD28 and CD3Z. A first-generation CAR with NKG2D and CD3Z showed the lowest toxicity. A construct similar to ours (NKG2D-4-1BB-CD3z) that lacks DAP10 and contains 4-1BB instead of CD28, was not tested in this report, and may behave differently. BALB/c T cells showed highest expression of NKG2D CAR, as well as cytokine production and cytotoxicity against NKG2DL+ targets, and also produced a more severe toxicity when injected into mice. These data indicate NKG2DL expression is variable among different mice strains, and will expectedly be different between mice and humans as NKG2DL genes differ significantly. However, the results showed in this study indicate NKG2DL-targeted therapies should be used with caution. In this study, we did not observe a correlation between NKG2DL expression and in vitro cytotoxicity of NKG2D–CAR–redirected T cells. This may indicate that other signals on the tumor cells, besides NKG2DL expression, are needed for triggering NKG2D-CAR T cells cytolytic activity. In addition, different NKG2DL bind NKG2D receptor with a wide range of affinities, and may vary in how they trigger or engage NKG2D, in such a way that different outcomes occur (36). This is in accordance with the lack of cytolytic activity of NKG2D CAR–redirected effector cells against various healthy tissues, that we and others have observed (12). We observed NKG2D–CAR T cells had cytotoxic activity against Hs 1.Int (normal fetal intestine) cells in vitro. The expression of NKG2DL on embryonic cells is well documented (16, 37). However, during pregnancy, placenta has an immunosuppressive role expressing and secreting NKG2DL that downmodulate NKG2D receptor expression (38). For this reason, the in vitro cytotoxicity against normal fetal intestine we have observed may not be extrapolated to their behavior in vivo. NKG2D–CAR T cells also showed a low level of cytotoxicity against NuLi-1 (normal epithelial lung) cells. NuLi-1 cell line is immortalized with HPV-16E6/E7-LXSN and some of the cells show chromosomal abnormalities, such as trisomy 5 and 20. Therefore, it cannot be considered as an entirely “normal” tissue. This could explain the cytotoxicity observed. The low level of cytotoxicity of NKG2D–CAR T cells observed against NuLi-1 cells (around 22%) contrasts with the high level of cytotoxicity observed against the fetal intestine cell line Hs 1.Int (around 86%), indicating adult healthy tissues could be much less susceptible to NKG2D–CAR T cells cytotoxicity. All these data taken together suggest that NKG2D CAR T-cell therapy might be safe in humans; however, more studies are needed to define the extent of ligand expression on normal tissues and to optimize the dose of therapeutic cells to rule out any toxicity. An ongoing phase I dose-escalation study is currently testing the safety of NKG2D CAR-T cells in patients with myeloid malignancies (ClinicalTrials.gov Identifier: NCT03018405). The results are encouraging. To date, 9 patients have been treated with a single dose of up to 3 × 107 of autologous NKG2D–CAR T cells without lymphodepleting conditioning. No cases of CRS, cell-related neurotoxicity, autoimmunity or CAR-T–related death have been recorded (39).
In a recent publication by Sentman and colleagues (40), they attribute the acute toxicity observed in mice treated with NKG2D-CAR-T cells by a CRS-like response similar to those observed in patients treated with CAR-T cells. We did not observe treatment-related toxicity in the treated mice. This can be explained by the incapability of murine NKG2DL to trigger a human NKG2D-CAR. In addition, xenograft models using immunocompromised mice are useful to evaluate the direct cytotoxic effect of CAR-redirected effector cells on cancer cells; however, they cannot be used to explore the adaptive immune response elicited by CAR-based immunotherapies. Because the animals are immunosuppressed, CAR therapy side-effects such as CRS or tumor lysis syndrome cannot be evaluated. The anticancer effects of an NKG2D–CAR-based cell immunotherapy have been previously reported using these animal models (12, 13). However, these models do not mirror the current human cancer scenario (41).
CRS is the most prevalent adverse effect following infusion of CAR-T cells and is correlated with expansion and immune activation of adoptively transferred cells, which provokes the elevation of inflammatory cytokines such as IFNγ, IL6, and GM-CSF. These inflammatory cytokines are mainly secreted upon activation, differentiation and clonal expansion of naïve T cells. To diminish the severity of CRS, we used differentiated memory T cells to express our CAR, as they show an impaired ability to clonally expand and to secrete inflammatory cytokines like IFNγ.
Other potential targets have been proposed for CAR–T-cell therapy against osteosarcoma. HER-2–modified T cells have been shown to target osteosarcoma TICs (42), however, HER-2 is expressed at low levels in the lung, and a patient died after infusion of HER-2–redirected CAR-T cells (43). IL11 Receptor α-chain (IL11Rα) has been proposed as target for CAR–T-cell therapy for osteosarcoma treatment. IL11Rα-CAR–redirected T cells have been shown to effectively kill osteosarcoma cells in vitro and to induce regression of established metastasis in vivo (44). A phase I clinical trial is currently testing the safety of a third-generation CAR directed against GD2 (ClinicalTrials.gov Identifier NCT: NCT02107963).
CARs can be inserted into effector cells by gammaretroviral, lentiviral, mRNA, transposon/transposase and Sleeping Beauty systems. The lentiviral vectors used in this report provided enhanced transduction efficiency and long-term NKG2D CAR expression. In addition, no chromosomal aberrations were found on NKG2D-CAR memory T cells, indicating the safety of lentiviral transduction. This observation is in accordance to a decade-long safety profile of the use of genetically modified T cells (45). However, in case of toxicity, temporal expression of the NKG2D-CAR provided by mRNA transfection might be desirable.
Although CARs can be expressed in autologous and allogeneic T cells, autologous infusions are not always an option due to low lymphocyte numbers or poor lymphocyte quality (and ex vivo expansion). Donor-derived CAR T cells are currently being infused into patients after allo-HSCT. However, having off-the-shelf cell products to potentially administer without a requirement for a prior HSCT from the same donor would be a major goal regarding CAR–T-cell therapies. One potential risk of the use of allogeneic T-cell–based therapies is the T-cell response against normal tissue: GvHD. To avoid undesirable GVH reactions, T-cell products lacking an alloreactive TCR are needed. Several methods have attempted to intensify GVT effects while minimizing GVH responses to lower toxicity and improve the outcome of treatment (19, 46–48). One approach to enrich nonallogeneic T cells is by using antigen-experienced memory T cells for CAR transduction (18, 19). Predictably, the vast majority of T cells with a memory phenotype are likely to have encountered antigens other than the allogeneic type. Thus, selection for memory phenotype cells should enrich for a nonalloreactive repertoire. Indeed, memory T cells showed less potential to generate GVHD in murine models (26, 49), in part due to nonalloreactive TCR enrichment along with the evidence that memory T cells are less likely to traffic to GVHD organs such as the gastrointestinal tract. One marker to distinguish naïve from memory T cells is CD45RA (22). Several preclinical models have used CD45RA memory T cells (19) or CD45RA− T (18) cells for CAR transduction. In our study, we used CD45RA− T cells because they have shown antipathogen memory response and antipotent leukemic activity without causing xenogeneic GvHD, circumventing a major survival limitation with the use of immunodeficient mice treated with human T cells (18). Although allogeneic memory T cells are nonalloreactive, they can be immunogenic, thus they could be eliminated by the host immune system once is recovered from a lymphodepleting conditioning. This could limit the efficacy of the therapy to a short period of time, so several infusions may be needed to maintain antitumor effects. However, a limited life span can be advantageous, so CAR-T cells would cause minimal or no toxicity. In this report, we described preclinical data to support the use of allogeneic NKG2D CAR–redirected CD45RA− cells for the treatment of osteosarcoma. NKG2D CAR-CD45RA− cells showed anti-osteosarcoma activity while showing no GVHD effects. NKG2D-CAR+ memory T cells failed to completely eradicate the tumor (as observed by the increase of the bioluminescence signal in treated mice after day 42), most likely because IL2 injections were limited to the first 3 weeks after T-cell injection, given it is an essential survival factor for T cells. In future experiments, repeated infusions of NKG2D-CAR memory T cells along with IL2 administration can be done to eradicate tumor progression.
In summary, our results show NKG2D CAR can be expressed in CD45RA− memory T cells with high transduction efficiency, the expression of CAR remains throughout, and it does not affect the viability or genetic stability of transduced cells. Most importantly, NKG2D–CAR+ CD45RA− cells had potent anti-osteosarcoma effects while remaining noncytotoxic against healthy tissues, as shown in in vitro Europium-TDA conventional assays and colony-forming assays. In vivo, mice treated with NKG2D CAR-expressing T cells showed less tumor burden and longer survival than those untreated or treated with control untransduced CD45RA− T cells without any toxicity related to CAR T-cell memory treatment. These data suggest that osteosarcoma patients could benefit from a new NKG2D CAR memory T-cell–based therapy.
Disclosure of Potential Conflicts of Interest
No potential conflicts of interest were disclosed.
Conception and design: L. Fernández, A. Patiño-García, J. Martínez, W. Leung, A. Pérez-Martínez
Development of methodology: L. Fernández, A. Escudero, M. Vela, A. Leivas, W. Leung
Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): L. Fernández, A. Escudero, M. Vela, J. Valentín, I. Vallcorba, A. Patiño-García, J. Martínez, W. Leung
Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): L. Fernández, J.-Y. Metais, J. Valentín, J. Torres, A. Pérez-Martínez
Writing, review, and/or revision of the manuscript: L. Fernández, J.-Y. Metais, A. Escudero, I. Vallcorba, J. Torres, A. Valeri, A. Patiño-García, J. Martínez, W. Leung, A. Pérez-Martínez
Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): L. Fernández
Study supervision: J. Martínez, A. Pérez-Martínez
The authors thank Jim Houston, Barbara Rooney, Martha Holladay, Patricia González, María Udriste, María Coronel, Manuel Pérez, Enrique Hernández, Victor Toledano, and Jesús Sánchez for technical support.
This work was supported in part by the National Health Service of Spain, Instituto de Salud Carlos III (ISCIII), a FONDOS FEDER grant (FIS) PI15/00973, Asociación Española Contra el Cáncer to Antonio Pérez Martínez and a CRIS Foundation to Beat Cancer grant to Lucía Fernández and Jaime Valentín. We are especially grateful to the family of 531MII, MEZ, whose generosity and capacity to see beyond, has allowed so many scientific achievements