CD171- and GD2-specific CAR-T cells potently target retinoblastoma cells in preclinical in vitro testing
Chimeric antigen receptor (CAR)-based T cell therapy is in early clinical trials to target the neuroectodermal tumor, neuroblastoma. No preclinical or clinical efficacy data are available for retinoblastoma to date. Whereas unilateral intraocular retinoblastoma is cured by enucleation of the eye, infiltration of the optic nerve indicates potential diffuse scattering and tumor spread leading to a major therapeutic challenge. CAR-T cell therapy could improve the currently limited therapeutic strategies for metastasized retinoblastoma by simultaneously killing both primary tumor and metastasizing malignant cells and by reducing chemotherapy-related late effects.
CD171 and GD2 expression was flow cytometrically analyzed in 11 retinoblastoma cell lines. CD171 expression and T cell infiltration (CD3+) was immunohistochemically assessed in retrospectively collected primary retinoblastomas. The efficacy of CAR-T cells targeting the CD171 and GD2 tumor-associated antigens was preclinically tested against three antigen-expressing retinoblastoma cell lines. CAR-T cell activation and exhaustion were assessed by cytokine release assays and flow cytometric detection of cell surface markers, and killing ability was assessed in cytotoxic assays. CAR constructs harboring different extracellular spacer lengths (short/long) and intracellular co-stimulatory domains (CD28/4-1BB) were compared to select the most potent constructs.
All retinoblastoma cell lines investigated expressed CD171 and GD2. CD171 was expressed in 15/30 primary retinoblastomas. Retinoblastoma cell encounter strongly activated both CD171-specific and GD2-specific CAR-T cells. Targeting either CD171 or GD2 effectively killed all retinoblastoma cell lines examined. Similar activation and killing ability for either target was achieved by all CAR constructs irrespective of the length of the extracellular spacers and the co-stimulatory domain. Cell lines differentially lost tumor antigen expression upon CAR-T cell encounter, with CD171 being completely lost by all tested cell lines and GD2 further down-regulated in cell lines expressing low GD2 levels before CAR-T cell challenge. Alternating the CAR-T cell target in sequential challenges enhanced retinoblastoma cell killing.
Both CD171 and GD2 are effective targets on human retinoblastoma cell lines, and CAR-T cell therapy is highly effective against retinoblastoma in vitro. Targeting of two different antigens by sequential CAR-T cell applications enhanced tumor cell killing and preempted tumor antigen loss in preclinical testing.
KeywordsAdoptive T-cell immunotherapy Retinoblastoma CD171 GD2 Antigen loss Sequential CAR-T cell therapy
Chimeric antigen receptor
Cluster of differentiation
Truncated epidermal growth factor receptor
Enzyme-linked immunosorbent assay
Fluorescence-activated cell sorting
Fetal calf serum
Green fluorescent protein
International classification of retinoblastoma
Lymphocyte-activation gene 3
Relative light unit
Single chain variable fragment
Trichocyst matrix protein T2A
T-cell immunoglobulin and mucin-domain containing-3
EBV-transformed lymphoblastoid cell line
Children suffering from unilateral intraocular retinoblastoma can be cured with the enucleation of the eye without any further treatment. However, 45% of all reported cases encompass the heritable form of retinoblastoma, a biallelic germline mutation of the retinoblastoma tumor suppressor gene RB1.  In 80% of children with heritable disease, retinoblastoma affects both eyes (bilateral) and 5% of the cases are associated with an intracranial tumor (trilateral). 
Saving life is the highest goal in retinoblastoma therapy followed by vision salvage. In order to salvage vision, if reasonable, the eye is preserved in case of localized tumors, which are treated with laser application cryo- or brachytherapy and/or local intra-arterial chemotherapy. In large tumors, initial reduction of the tumor size can be achieved by systemic chemotherapy, which enables subsequent local treatment options. High-dose systemic chemotherapy with stem cell rescue is reserved for non-responsive extraocular and/or metastastic disease. [3, 4] Overall survival is high in western countries (> 95%). However, due to a higher rate of secondary malignancies, long-term overall survival is reduced in children treated with eye preserving radio- and/or chemotherapy compared with enucleation alone. [5, 6]
Retinoblastoma can disseminate through the optic nerve into the central nervous system and through the sclera via lymphatic or blood circulation of the orbit bones to distant metastatic sites in the lymph nodes, bones, bone marrow and liver.  In these cases, salvage with high-dose chemotherapy is often not successful. In addition, high-dose chemotherapy is highly aggressive, and can create lifelong sequelae and morbidity for the patient. [4, 7, 8, 9] Therefore, the search for more efficient and better tolerated treatment options is warranted.
Adoptive T cell therapy might be a promising alternative. Adoptive T cell immunotherapy, in which T lymphocytes isolated from patients are engineered to express CD19-specific chimeric antigen receptors (CARs), has shown striking anti-tumor effects against acute B cell leukemia and non-Hodgkin lymphoma. [10, 11, 12, 13] CAR-T cells combine two striking characteristics of the immune system: the exquisite antigen-binding specificity of a monoclonal antibody and the potent toxicity of cytotoxic T lymphocytes. A spacer domain connects the antigen-binding domain, commonly a single-chain variable fragment (scFv) of a monoclonal antibody, to the transmembrane domain followed by a T cell signaling module.  Spacer length influences CAR-T cell function, as the distance between the CAR-T cell and tumor antigen epitope must be uniquely adjusted for optimal bridging. [15, 16] The signaling module incorporates the CD3-zeta domain and a co-stimulatory domain, commonly either 4-1BB or CD28, to provide signals necessary for full T cell activation. The co-stimulatory domain used can affect CAR-T cell functionality by triggering different signaling pathways. The 4-1BB domain has been associated with increased CAR-T cell persistence , but the CD28 domain has been demonstrated to enhance CAR-T cell cytotoxicity. 
GD2 and CD171 may present promising targets for CAR-T cell therapy of retinoblastoma. The GD2 ganglioside is expressed on the cell surface of several neuroectodermal tumors, including retinoblastoma. [19, 20, 21, 22] GD2 expression is highly restricted in nonmalignant tissue with only low-level expression on peripheral nerves, skin melanocytes, brain and osteoprogenitors. [23, 24] Anti-GD2 monoclonal antibodies have already proven safety and efficacy in clinical trials and are included in the standard treatment for children with high-risk neuroblastoma demonstrating its role as a target for immunotherapy. [25, 26, 27] CD171 (formerly L1CAM) plays a crucial role during nervous system development, including neuronal migration and axon guidance.  It was recently shown to be expressed in retinoblastomas, and expression in the Y79 and Rb1 cell lines correlated with increased in vitro proliferation and chemoresistance in a mouse model.  In most tumor entities CD171 expression is further described to be associated with poor prognosis making it a potential target for new treatment options like immunotherapy. [30, 31, 32] CD171 expression by normal tissue was examined by our group and a safety study in non-human primates revealed no on-target, off-tumor toxicity after infusion of up to 1 × 108/kg CD171-specific CAR-T cells in non-conditioned animals.  CAR-T-cell therapy could represent a new treatment option for extraocular and/or metastasized retinoblastoma. If successful, CAR-T cell therapy could also be integrated with vision-preserving therapies for children with bilateral retinoblastoma to reduce therapeutic toxicity and late treatment-related effects such as secondary cancers, while preserving vision in at least one eye.
As a first step towards a CAR-T-cell therapy targeting retinoblastoma, we here assessed CD171 expression in retrospective retinoblastoma samples and investigated the killing efficacy of CD171- and GD2-specific CAR-T cells harboring different spacer lengths (short versus long) and intracellular co-stimulatory domains (CD28 versus 4-1BB) in a panel of three different cell lines. Our aim was to assess functional differences of various CAR-T cell constructs to select the most efficient for further preclinical testing and entry into the clinic trials.
Retinoblastoma samples and cell lines
Tumor samples and medical records were retrospectively evaluated from 30 children diagnosed with retinoblastoma between January 2003 and August 2013 in the Department of Pediatric Oncology and Hematology, Charité - Universitätsmedizin Berlin. Cell lines RB247C3, RB355, RB383, RB522, RB1021 and RB3823 were kindly provided by Brenda Gallie (Department of Ophthalmology and Vision Sciences, Hospital for Sick Children, Toronto, Ontario, Canada).  RBL13, RBL15 and RBL30 cell lines were established at the University Hospital Essen and were obtained from the Institute of Cell Biology, University Hospital Essen.  Cell lines, Y79 (DSMZ-ACC 246), WERI-Rb1 (DSMZ-ACC 90) and NALM-6 (DSMZ-ACC 128), were purchased from the German collection of microorganisms and cell cultures. The WERI-Rb1, RBL15 and RB355 retinoblastoma cell lines and the NALM-6 B cell leukemia cell line were lentivirally transduced with green fluorescent protein (GFP)-firefly luciferase (ffLuc)_epHIV7, and GFP-expressing cells were selected using fluorescence-activated cell sorting (FACS). WERI-Rb1 and NALM-6 cells were maintained in RPMI 1640 (Gibco Life technologies) supplemented with 15% heat-inactivated fetal calf serum (FCS, Sigma-Aldrich). RBL15 and RB355 cell lines were maintained in DMEM (Gibco Life technologies) supplemented with 10% heat-inactivated FCS, 10 μg/ml insulin (Sigma-Aldrich) and 50 μM 2-mercaptoethanol (Sigma-Aldrich). The TMLCL EBV-transformed lymphoblastoid cell line was maintained in RPMI 1640 supplemented with 10% heat-inactivated FCS and 2 mM L-glutamine (Biochrom).
The CD171 and CD3 (T cell marker) were immunohistochemically detected in tumor sections. Primary antibodies were omitted as negative controls and kidney tissue (CD171) was used as positive control (see Additional file 1: Figure S1). T cell infiltration (number of CD3+ lymphocytes per cm2) in preserved retinoblastoma tissue sections was quantified using the Stereo Investigator® system (MBF Bioscience). Methods of immunohistochemistry are described in an additional file in more detail (see Additional file 2).
The CD171-specific CE7-CAR  and GD2-specific 14.2GA-CAR  were previously described, and were cloned into the SIN epHIV7 lentiviral vector plasmid. CAR lentiviruses were propagated in 293 T cells.  The scFvs in both CAR constructs were codon optimized and subsequently linked to a 12 (short) or 229 (long) amino acid spacer domains from the human IgG4 hinge. The long spacer domain was modified by substituting L235D and N297Q to reduce binding to the IgG Fc gamma receptor.  The spacer domain connects the antigen-binding domain to CD28 transmembrane domain followed by the signaling module containing the CD3zeta cytoplasmic domain and either 4-1BB or CD28. The CAR constructs were linked downstream to a T2A self-cleaving peptide and truncated epidermal growth factor receptor (EGFRt) allowing CAR-T cell detection and enrichment.
Generation and culture of CD171- and GD2-specific CAR-T cells
Apheresis products were obtained from healthy donors and peripheral blood mononuclear cells were isolated using Ficoll-Paque (GE Healthcare). CD8+ T cells were obtained by positive selection using immunomagnetic microbeads (Miltenyi Biotec), and activated with anti-CD3/CD28 beads (Life Technologies). On day three, activated CD8+ T cells were transduced with the CAR containing lentivirus. The EGFRt+ CAR-T cell subset was enriched by immunomagnetic selection with biotin-conjugated cetuximab (Bristol-Myers Squibb) and streptavidin microbeads (Miltenyi Biotec). T cells used as mock negative controls alongside CAR-T cells in experiments were not lentivirally transduced. CAR and mock control T cells were stimulated with irradiated peripheral blood mononuclear cells, irradiated TMLCL, and OKT3 (30 ng/mL, Miltenyi Biotec), expanded according to a rapid expansion protocol  and cryopreserved until further use. Cryopreserved cells were thawed, expanded as described above and functional in vitro assays were conducted between days 11 and 16 of culture.
For cytokine release assays, 2 × 105 T cells were seeded together with stimulator cells at a 2:1 effector:target ratio. All data points were performed as technical triplicates. After 24 h, supernatants were collected and stored at − 80 °C until analysis of IFNG and IL2 using the OptEIA™ Set (BD Biosciences) enzyme-linked immunosorbent assay (ELISA) kits in accordance with the manufacturer’s instructions. CAR-T cell-induced cytotoxicity was quantified in a biophotonic luciferase assay in which the retinoblastoma cells, stably transduced with a GFP-ffLuc_epHIV7 reporter, served as tumor target cells. Target cells were co-cultured in triplicate with mock-transduced or CAR-T cells. The maximal biophotonic luciferase signal was defined by target cells plated alone at the same densities (RLUmax, maximal relative light unit). After 24, 48 or 72 h, 0.14 mg D-luciferin (PerkinElmer Inc.)/ml medium was added to each well, and the biophotonic signal detected. Lysis was determined as [1-(RLUsample/RLUmax)]× 100 in relation to untreated cells. For sequential treatment, the additive amount of tumor lysis was calculated related to viable tumor cells at day 3.
Flow cytometric marker and antigen detection
Cell-surface expression of GD2 (cat#565991, BD Biosciences), CD8 (cat#301041, BioLegend) and CD171 (cat#130–100-691, Miltenyi Biotec) was detected by fluorophore-conjugated monoclonal antibodies. EGFRt expression was detected using biotinylated cetuximab (Bristol-Myers Squibb) and a phycoerythrin (PE)-conjugated streptavidin antibody (cat#12–4317-87, BioLegend). Activation and exhaustion were assessed by fluorophore-conjugated monoclonal antibodies detecting CD137 (cat#309819, BioLegend), CD25 (cat#302622, BioLegend), PD1 (also known as PDCD1 or CD279, cat#329922, BioLegend), TIM3 (cat#345006, Biolegend) and LAG3 (cat#565721, BD Biosciences). Flow cytometry was performed on a Fortessa X-20 (BD Biosciences) and data processed using FlowJo software (Tree Star Inc.). Dead cells were excluded from analyses using LIVE/DEAD™ Fixable Green Dead Cell Stain Kit (cat#L23101, Life Technologies). QuantiBRITE PE calibration beads (BD Biosciences) were used to determine GD2 and CD171 antigen density on retinoblastoma cells according to the manufacturer’s instructions.
To determine significance of differences in cytotoxic activity and cytokine release of GD2-and CD171-specific CAR-T cells compared with negative control mock T-cells, unpaired Student’s t-test was performed using GraphPad prism (GraphPad Software). P values < 0.05 were considered statistically significant.
CD171 expressed in primary retinoblastomas and retinoblastoma cell lines
Patient and tumor characteristics included in primary retinoblastoma cohort
Minimum age at diagnosis (years)
Maximum age at diagnosis (years)
Median age at diagnosis (years)
♀: 15 (50)
♂: 15 (50)
CD171-specific CAR-T cells target and kill retinoblastoma cells but induce escape mechanisms
Having demonstrated CAR-T cell activation after retinoblastoma encounter, we investigated the ability of CAR-T cells to kill retinoblastoma cells in a luciferase-based reporter assay. The WERI-Rb1, RBL15 and RB355 cell lines were transduced with a GFP-firefly luciferase reporter plasmid to support viable tumor cell quantification. T cells harboring the various CD171-specific constructs were co-cultured with retinoblastoma reporter cells in a 2:1 effector:target ratio. Cytotoxicity was assessed following 24 h, 48 h and 72 h of co-culture by measuring the biophotonic signal released by the remaining viable tumor cells. CD171-specific CAR-T cells killed significantly more retinoblastoma cells, regardless of cell type, than mock-transduced negative control T cells (Fig. 2e). There was no killing of the CD171 negative cell line, NALM-6 (Additional file 5: Figure S4D). Further, CD171-specific CAR-T cells showed a dose-dependent cytotoxicity of retinoblastoma cells. Cytotoxicity ranged from 97% with a 5:1 effector:target ratio to 76% (1:1 effector:target ratio) to no killing (1:10 effector:target ratio; Additional file 6: Figure S5). These experiments demonstrate that CAR-T cells targeting CD171 efficiently kill retinoblastoma cells in vitro.
Since it is known that different spacers and co-stimulatory domains have influence on the performance of CAR-T cells we compared the T cells harboring different CAR constructs in terms of cytokine release, activation, expression of inhibitory receptors and killing. T cells harboring the CD171 CAR construct with the long spacer and CD28 co-stimulatory domain (CD171-long-CD28) released the most IFNG and IL2 after encounters with any tested cell line (Fig. 2b), although differences did not reach any statistically significance. All CAR-T cells were comparably well activated following co-culture with WERI-Rb1, RBL15 or RB355 cell lines, except T cells harboring the CD171-short-4-1BB construct which displayed the lowest levels of both markers (Fig. 2c). Co-culture with any of the 3 retinoblastoma cell lines consistently induced TIM3 expression, regardless of the CD171-specific CAR construct type expressed (Fig. 2d). The CAR construct used did not have as large an impact on the CAR-T cell ability to kill retinoblastoma cells as it had on the cytokine release by the co-cultured CAR-T cells. Most CD171-specific CAR-T cells were able to kill nearly 100% (Fig. 2e) of the retinoblastoma cells underling the potency of CAR-T cell therapy against retinoblastoma. As already described by Long et al., the ability of CAR-T cells to produce polyfunctional cytokines in response to antigen encounter is a better predictor of CAR-T cell antitumor efficacy in vivo than cytolytic efficacy alone.  Since the CAR construct utilizing the long spacer and CD28 co-stimulatory domains produced slightly higher amounts of IFNG and IL2 in vitro, it was selected for the remaining experiments.
GD2 is a second promising target for CAR-T cell therapy in retinoblastoma
To test the ability of GD2-specific CAR-T cells to kill retinoblastoma cells in vitro, we assessed cytotoxic efficacy using the luciferase-based reporter assay. GD2-specific CAR-T cells were co-cultured in a 2:1 effector:target ratio with retinoblastoma reporter cells before assessing the biophotonic signal released by the remaining viable tumor cells after 24 h, 48 h and 72 h. GD2-specific CAR-T cells killed significantly more WERI-Rb1 and RBL15 cells than the mock-transduced negative control (Fig. 5d). Cytotoxicity against RB355 cells was weaker, in line with the lower target expression. The GD2-negative NALM-6 cell line was not killed by GD2-specific CAR-T cells (Additional file 5: Figure S4D). Dose-dependent cytotoxicity of GD2-specific CAR-T cells was demonstrated in co-culture with RBL15 cells in different effector:target ratios. Cytotoxicity ranged from 90.1% in a 5:1 effector:target ratio to 66% (1:1 effector:target ratio) to 1.5% (1:10 effector:target ratio; Additional file 6: Figure S5). These experiments demonstrate that CAR-T cells targeting GD2 efficiently kill retinoblastoma cells in vitro in an antigen density-dependent manner.
The impact of different spacer and co-stimulatory domains on cytokine release, activation, inhibitory receptor expression and killing ability of T cells targeting GD2 was also assessed. Similar to CD171-specific CAR-T cells, T cells harboring the long spacer and CD28 co-stimulatory domain released more IFNG and IL2 after encounters of any tested cell line than T cells harboring the short spacer and CD28 co-stimulatory domain (Fig. 5a), indicating superior effector functionality of T cells equipped with the GD2-long-CD28 construct. T cells harboring the long spacer with either co-stimulatory domain released comparable amounts of IL2 after encounters with WERI-Rb1 and RBL15. T cells harboring all constructs except the GD2-short-CD28 CAR construct were comparably well activated by WERI-Rb1 or RBL15 cell co-culture (Fig. 5b). Inhibitory receptor expression was comparably induced in T cells harboring all constructs, except that PD1 expression was stronger in T cells harboring the CD28 co-stimulatory domain (Fig. 5c). The CAR construct used did not have as large an impact on the CAR-T cell ability to kill retinoblastoma cells as it had on cytokine release by the T cells. Most GD2-specific CAR-T cells were able to kill nearly 90% (Fig. 5d) of the retinoblastoma cells underlining the potency of GD2-specific CAR-T cell therapy against retinoblastoma. Although the CAR construct had less impact on efficacy of GD2-specifc compared with CD171-specific CAR-T cells, the GD2-long-CD28 construct performed best in vitro and was selected for the remaining experiments.
Having verified cytokine release, activation and cytolytic capacity we also assessed tumor escape mechanisms of retinoblastoma cell lines following GD2-specific CAR-T cell encounter. As done before for the CD171-specific CAR-T cells, we investigated GD2 expression on WERI-Rb1 and RB355 cells via FACS analysis at several time points after transient co-culture with GD2-specific CAR-T cells (Fig. 5e, Additional file 8: Figure S7). Interestingly, while transient encounter with GD2-specific CAR-T cells downregulated the already low GD2 expression on RB355 cells by more than half when assessed on day 3 concomitantly with CAR-T cell removal, WERI-Rb1 cell lines did not change GD2 expression. RB355 cells regained GD2 expression by day 14, with antigen levels comparable to untreated cells. GD2 antigen expression may only be differentially regulated in cells with low antigen densities, reducing the risk of potential immune evasion when GD2 is targeted in retinoblastoma cells.
Sequential CAR-T cell treatment counteracts antigen loss and improves cytotoxicity against retinoblastoma cells
In parallel, we assessed cytotoxic efficacy of sequential CAR-T cell treatment targeting the same or 2 different antigens using the same luciferase-based reporter assay for tumor cell killing as in our monotreatment experiments. The same treatment regimens (Fig. 6a) were conducted with 1:5 effector:target ratios for CAR-T cells and WERI-Rb1 or RB355 cells (Fig. 6d). CD171- and GD2-specific CAR-T cells showed comparable killing efficacy of WERI-Rb1 cells during treatment 1 and 2. Switching from GD2- to CD171-specific CAR-T cells increased tumor cell lyses by 30% from day 3 to day 6 while a sequential treatment with GD2-specfifc CAR-T cells showed an increase of tumor lysis of only 19% indicating a benefit for administration of sequential CAR-T cell therapy targeting different antigens. Analyzing CD171- and GD2-specific CAR-T cell killing ability of RB355 cells revealed higher cytotoxicity of CD171- than of GD2-specific CAR-T cells during treatment 1 and 2. Treatment with GD2-specific CAR-T cells followed by subsequent CD171-specific CAR-T cell treatment displayed a significantly higher tumor lysis compared to a second GD2-specific CAR-T cell treatment. Our in vitro analyses support that sequentially targeting multiple antigens with CAR-T cells increases treatment efficacy, with the greatest benefit produced by targeting CD171 before GD2, even in cells with initial low-density GD2 expression.
We present preclinical evidence that CAR-T cell therapy targeting retinoblastoma-specific antigens could be a promising option to treat both primary tumors as well as metastasized or metastasizing malignant cells. CD171- and GD2-specific CAR-T cells were strongly activated by retinoblastoma cell encounter and demonstrated potent killing efficacy of retinoblastoma cells in vitro dependent on target antigen expression. Though all tested CAR constructs demonstrated equal activation and killing ability for either target, the CAR construct harboring the long extracellular spacer in combination with the CD28 co-stimulatory domain, tended to release higher amounts of effector cytokines, although this difference did not reach statistical significance While CD171 expression was completely lost on all cell lines following CAR-T cell treatment, GD2 was only down-regulated in cell lines initially harboring low GD2 levels. Here, we reveal enhanced retinoblastoma cell killing in alternating the CAR-T cell killing in sequential CAR-T cell challenges.
Watanabe et al. previously demonstrated for CD20-specific CAR-T cells that antigen density on acute lymphatic leukemia cell lines influenced CAR-T cell efficacy.  The same was shown by Fry et al. in patients with CD22-diminished or -negative leukemia blasts, who did not respond to CD22-specific CAR-T cell therapy.  Turrati et al. developed a cellular model where HER2 antigen expression could be increased, which caused enhanced CAR-T cell activity.  Here, we also show antigen expression level was correlated with cytotoxicity for GD2-specific CAR-T cells, since low GD2-expressing RB355 cells were only marginally killed, while WERI-Rb1 cells with higher GD2 expression elicited efficient killing.
The non-signaling extracellular spacer plays a pivotal role in tumor recognition: as each new scFv and target molecule on tumor membrane define a distinct biophysical synapse, a unique adjustment is needed for every target antigen. Hudecek et al. could previously demonstrate that T cells harboring a short spacer-containing CD19-CAR were able to eradicate tumor cells in vivo while the same dose of CD19-specific CAR-T cells with a long spacer failed to do so.  We previously showed that levels of activation and cytotoxicity for CD171-specific CARs varied depending on the extracellular spacer length when targeting neuroblastoma.  In this previous study, we compared CD171-specific CAR constructs that all harbored the 4-1BB co-stimulatory domain but different spacer length. The CAR construct with the long spacer domain performed best in in vitro assays, but, interestingly, the superiority of the long spacer CAR was reversed in vivo, most likely due to activation-induced cell death caused by recursive antigen exposure. The data we present here showed no clear superiority for any of the extracellular spacer length, irrespectively of target cell line or amount of antigen expression. However, investigation of the different CAR constructs in preclinical mouse models might reveal differences, since CAR-T cell performance has been shown to be influenced by additional factors such as CAR-T cell expansion and persistence in vivo.  The mechanisms by which different co-stimulatory domains influence T cell expansion, function and persistence are not yet fully understood, but the CD28 co-stimulatory domain is associated with enhanced anti-tumor efficacy and T cell functionality and enhanced T cell survival and persistence is attributed to the 4-1BB co-stimulatory domain.  Hudecek et al. revealed higher cytokine release for ROR1-specific CAR-T cells harboring a CD28 co-stimulatory domain compared to T cells with a 4-1BB co-stimulatory domain.  Long et al. previously showed that tonic signaling in CAR-T cells harboring a CD28 co-stimulatory domain led to T cell exhaustion and reduced functionality in vivo, which could be ameliorated by incorporation of 4-1BB into GD2-CAR-T cells.  In line with these findings, our data depicts increased PD1 expression on GD2-specific CAR-T cells harboring the CD28 co-stimulatory. However, from our functional analyses of GD2- and CD171-specific CAR-T cells, we conclude that all tested CD171- and GD2-specific CAR constructs, irrespective of the co-stimulatory domain, against retinoblastoma in vitro. Further testing in preclinical mouse models will reveal whether there is a ranking between CD171- and GD2-specific CARs equipped with 4-1BB or CD28 costimulatory domains against retinoblastoma.
Differences in CAR-T cell exhaustion of CAR-T cells with different scFvs were previously demonstrated by Yin et al., who analyzed EGFRvIII- and IL-13Rα2-targeting CAR-T cells against the same tumor model.  By using different scFvs, Yin et al. showed a higher expression of the immune checkpoint molecule CTLA-4 in higher stimulated CAR-T cells, while TIM-3 and PD1 are predominantly expressed on less activated CAR-T cells.  In our study, the differences between the expression of inhibitory receptors could also be explained by the different scFvs used. Further, the activation level is not significantly higher but a trend towards higher cytokine production and expression of activation markers can be observed after 24 h of co-culture and may reach significant difference after 48 h.
On the way to translate these findings into the clinic, in vivo experiments need to further explore potential toxicities, e.g. cytokine release syndrome, in context of an ocular tumor.
In general, tumor cells are capable of evading targeted immune therapy by downregulating the targeted antigen leading to reduced tumor recognition and rendering CAR-T cell therapy inefficient. [42, 52] Clinical application of sequential lymphoma-specific CAR-T cells was recently described by Shalabi et al.  Targeting B-cell malignancies with CD19-specific CAR-T cells resulted in outgrowth of antigen-negative variants. However, subsequent targeting with a second CAR-specificity could achieve cancer regression and prolonged survival of cancer patients, highlighting the importance of sequential tissue evaluation through the course of treatment. Fry et al. revealed CD22-specific CAR-T cell therapy is an option for patients relapsing after CD19-directed treatment.  Brown et al. demonstrated a response in a patient with recurrent multifocal glioblastoma using sequentially administered CAR-T cells having the same target specificity.  While loss of both antigens was observed in RB355 cell lines, the WERI-Rb1 cell line did not downregulate GD2 levels independently of CAR specificity. The different expression pattern could be related to differences in GD2 expression between the cell lines. As RB355 cells generally express low levels of GD2 and RB355 cells express fewer antigen molecules per cell, the downregulation may be detected earlier than in the high GD2 expression background on WERI-Rb1 cells. The subsequent GD2 upregulation after 10 days may be due to an indispensable GD2 function in the cell population. GD2 is associated with tumorigenesis as well as cancer cell proliferation and invasion.  Here we demonstrate that sequentially administering CAR-T cells targeting different antigens achieves higher CAR-T cell killing ability.
Our data revealed CD171 and GD2 as effective targets for CAR-T cell therapy for retinoblastoma in vitro. We demonstrated the risk potential of antigen loss on targeted tumor cells after CAR-T cell treatment, but also show that sequentially switching antigen specificity in the CAR-T cell therapy provides a striking benefit for retinoblastoma cell killing ability. This work provides the basis for in vivo testing to select the most beneficial regimens and target combinations on the way to establishing CAR-T cell therapy for retinoblastoma.
The authors thank Kathleen Anders and Kathy Astrahantseff for proofreading the manuscript, Benjamin Englert for his support in the analysis and quantification of histological retinoblastoma staining and Harald Stephan for his advice regarding the retinoblastoma cell lines. We acknowledge support from the German Research Foundation (DFG) and the Open Access Publication Fund of Charité – Universitätsmedizin Berlin.
LA performed the experiments. LA and AK2 were major contributors to study concepts, study design, and manuscript writing. JR and ES performed histological examination of primary retinoblastoma tissue sections. AJ provided the primary retinoblastoma tissue sections. MCJ provided the CD171-specific CE7-CAR construct, CR the GD2-specific 14.2GA-CAR construct. CF and FZ analyzed and interpreted the patient’s data regarding retinoblastoma disease and CD171 expression. LA, AK1, AK2, SS, AW, LG, FZ, CF, and AH contributed to conduction of experiments, analysis of data, quality control and interpretation of data. AE and JHS were involved in revising the article for important intellectual content. All authors approved the final manuscript.
JR, CF, FZ and AK are participants in the BIH-Charité Clinical Scientist Program funded by the Charité – Universitätsmedizin Berlin and the Berlin Institute of Health. The research reported in the article was financially supported by the German Eye Cancer Foundation. The funding bodies did not play a role in the design of the experiments, analysis or interpretation of data or writing of the manuscript.
Ethics approval and consent to participate
Patients or their guardian (for participants under the age of 16 years) had given written consent and research was conducted in accordance with the Declaration of Helsinki. Ethical approval was obtained from the University of Duisburg-Essen ethics committee (13–5405-BO).
Consent for publication
The authors have no financial or non-financial competing interests to declare.
- 1.Dimaras H, Corson TW, Cobrinik D, White A, Zhao J, Munier FL, Abramson DH, Shields CL, Chantada GL, Njuguna F, et al. Retinoblastoma. Nat Rev Dis Primers. 2015;1:1–22.Google Scholar
- 2.Soliman SE, Racher H, Zhang C, MacDonald H, Gallie BL. Genetics and molecular diagnostics in retinoblastoma--an update. Asia Pac J Ophthalmol (Phila). 2017;6(2):197–207.Google Scholar
- 3.Temming P, Arendt M, Viehmann A, Eisele L, Le Guin CHD, Schündeln MM, Biewald E, Astrahantseff K, Wieland R, Bornfeld N, et al. Incidence of second cancers after radiotherapy and systemic chemotherapy in heritable retinoblastoma survivors: a report from the German reference center. Pediatr Blood and Cancer. 2017;64:71–80.CrossRefGoogle Scholar
- 5.Temming P, Lohmann D, Bornfeld N, Sauerwein W, Goericke SL, Eggert A: Current concepts for diagnosis and treatment of retinoblastoma in Germany: aiming for safe tumor control and vision preservation TT - Aktuelle Konzepte zur diagnose und Behandlung des Retinoblastoms in Deutschland mit dem Ziel der sicheren Tumorkontrolle. Klin Padiatr 2012, 224:339–347.CrossRefGoogle Scholar
- 6.Shields CL, Ja S, Baez K, Cater JR, De Potter P. Choroidal invasion of retinoblastoma: metastatic potential and clinical risk factors. Brit J Opthalmol. 1993;73:692–8.Google Scholar
- 12.Kochenderfer JN, Dudley ME, Kassim SH, Somerville RPT, Carpenter RO, Maryalice SS, Yang JC, Phan GQ, Hughes MS, Sherry RM, et al. Chemotherapy-refractory diffuse large B-cell lymphoma and indolent B-cell malignancies can be effectively treated with autologous T cells expressing an anti-CD19 chimeric antigen receptor. J Clin Oncol. 2015;33:540–9.CrossRefGoogle Scholar
- 14.Guest RD, Hawkins RE, Kirillova N, Cheadle EJ, Arnold J, O'Neill A, Irlam J, Chester KA, Kemshead JT, Shaw DM, et al. The role of extracellular spacer regions in the optimal Design of Chimeric Immune Receptors: evaluation of four different scFvs and antigens. J Immunother. 2005;28(3):203–11.CrossRefGoogle Scholar
- 17.Milone MC, Fish JD, Carpenito C, Carroll RG, Binder GK, Teachey D, Samanta M, Lakhal M, Gloss B, Danet-desnoyers G, et al. Chimeric receptors containing CD137 signal transduction domains mediate enhanced survival of T cells and increased Antileukemic efficacy in vivo. Mol Ther. 2009;17:1453–64.CrossRefGoogle Scholar
- 19.Guillermo L. Chantada, Jorge Rossi, Fernando Casco, Fandin A, Marcelo Scopinaro, MariaT.G.de Da’vila, Abramson DH: An aggressive bone marrow evaluation including immunocytology with gd2 for advanced retinoblastoma. J Pediatr Hematol Oncol 2006, 28(6):369–373.
- 26.Cheung NK, Saarinen UM, Neely JE, Landmeier B, Donovan D, Coccia PF: Monoclonal antibodies to a glycolipid antigen on human neuroblastoma cells. Cancer Res 1985, 45:2642–2649 ST - Monoclonal antibodies to a glycolipid.Google Scholar
- 32.Inaguma S, Wang Z, Lasota JP, Miettinen MM. Expression of neural cell adhesion molecule L1 (CD171) in neuroectodermal and other tumors: an immunohistochemical study of 5155 tumors and critical evaluation of CD171 prognostic value in gastrointestinal stromal tumors. Oncotarget. 2016;7(34):55276–89.CrossRefGoogle Scholar
- 33.Künkele A, Taraseviciute A, Finn LS, Johnson AJ, Berger C, Finney O, Chang CA, Rolczynski LS, Brown C, Mgebroff S, et al. Preclinical assessment of CD171-directed CAR T-cell adoptive therapy for childhood neuroblastoma: CE7 epitope target safety and product manufacturing feasibility. Clin Cancer Res. 2017;23:466–77.CrossRefGoogle Scholar
- 39.Hombach A, Abken H. Adoptive immunotherapy with genetically engineered T cells: modification of the IgG1 fc spacer domain in the extracellular moiety of chimeric antigen receptors avoids off-target activation and unintended initiation of an innate immune response. Gene Ther. 2010;17:1206–13.CrossRefGoogle Scholar
- 42.Louis CU, Savoldo B, Dotti G, Ma P, Yvon E, Myers GD, Rossig C, Russell HV, Diouf O, Liu E, et al. Antitumor activity and long-term fate of chimeric antigen receptor – positive T cells in patients with neuroblastoma. Blood. 2011;14:1324–34.Google Scholar
- 43.Beatty GL, O'Hara MH, Lacey SF, Torigian DA, Nazimuddin F, Chen F, Kulikovskaya IM, Soulen MC, McGarvey M, Nelson AM, et al. Activity of Mesothelin-specific chimeric antigen receptor T cells against pancreatic carcinoma metastases in a phase 1 trial. Gastroenterology. 2018;155(1):29–32.CrossRefGoogle Scholar
- 50.Hudecek M, Lupo-Stanghellini M-T, Kosasih PL, Sommermeyer D, Jensen MC, Rader C, Riddell SR. Receptor affinity and extracellular domain modifications affect tumor recognition by ROR1-specific chimeric antigen receptor T cells. Clin Cancer Res. 2013.Google Scholar
- 53.Shalabi H, Kraft IL, Wang H-W, Yuan CM, Yates B, Delbrook C, Zimbelman JD, Giller R, Stetler-Stevenson M, Jaffe ES et al: Sequential loss of tumor surface antigens following chimeric antigen receptor T-cell therapies in diffuse large B-cell lymphoma. Haematologica 2018, 103:haematol.2017.183459.CrossRefGoogle Scholar
Open AccessThis article is distributed under the terms of the Creative Commons Attribution 4.0 International License (http://creativecommons.org/licenses/by/4.0/), which permits unrestricted use, distribution, and reproduction in any medium, provided you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons license, and indicate if changes were made. The Creative Commons Public Domain Dedication waiver (http://creativecommons.org/publicdomain/zero/1.0/) applies to the data made available in this article, unless otherwise stated.