The advances in genetic engineering have enabled us to confer T cells new desired functions or delete their specific undesired endogenous properties for improving their antitumor function. Due to their efficient gene delivery, viral vectors have been successfully used in T-cell engineering to provide gene transfer medicinal products for the treatment of human disease. One example is adoptive cell therapy with T cells that were genetically modified with gamma-retroviral and lentiviral (LV) delivery vectors to express a CD19-specific chimeric antigen receptor (CAR) for cancer treatment. This therapeutic approach has shown remarkable results against B-cell malignancies in pilot clinical trials. Consequently, there is a strong desire to make CAR T cell therapy scalable and globally available to patients. However, there are persistent concerns and limitations with the use of viral vectors for CAR T cell generation with regard to safety, cost and scale of vector production. In order to address these concerns, we aimed to improve non-viral gene transfer and genome editing tools as an effective, safe and broadly applicable alternative to viral delivery methods for T-cell engineering. In the first part of the study, we engineered CAR T cells through non-viral Sleeping Beauty (SB) transposition of CAR genes from minimalistic DNA vectors called minicircles rather than conventional SB plasmids. This novel approach dramatically increased stable gene transfer rate and cell viability and resulted in higher yield of CAR+ T cells without the need of long ex vivo expansion to generate therapeutic doses of CAR+ T cells. Importantly, CD19-CAR T cells modified by MC-based SB transposition were equally effective as LV transduced CD19-CAR T cells in vitro and in a murine xenograft model (NSG/Raji-ffLuc), where a single administration of CD8+ and CD4+ CAR T cells led to complete eradication of lymphoma and memory formation of CAR T cells after lymphoma clearance. To characterize the biosafety profile of the CAR T cell products, we did the most comprehensive genomic insertion site analysis performed so far in T cells modified with SB. The data showed a close-to-random integration profile of the SB transposon with a higher number of insertions in genomic safe harbors compared to LV integrants. We developed a droplet digital PCR assay that enables rapid determination of CAR copy numbers for clinical applications. In the second part of the study, we ablated expression of PD-1, a checkpoint and negative regulator of T cell function to improve the therapeutic index of CAR T cells. This was accomplished using non-viral CRISPR/Cas9 via pre-assemble Cas9 protein and in vitro-transcribed sgRNA (Cas9 RNP). Finally, we combined our developed Cas9 RNP tool with CAR transposition from MC vectors into a single-step protocol and successfully generated PD-1 knockout CAR+ T cells. Based on the promising results achieved from antibody-mediated PD-1 blockade in the treatment of hematological and solid tumors, we are confident that PD-1 knockout CAR T cells enhance the potency of CAR T cell therapies for treatment of cancers without the side effects of antibody-based therapies. In conclusion, we provide a novel platform for virus-free genetic engineering of CAR T cells that can be broadly applied in T-cell cancer therapy. The high level of gene transfer rate and efficient genome editing, superior safety profile as well as ease-of-handling and production of non-viral MC vectors and Cas9 RNP position our developed non-viral strategies to become preferred approaches in advanced cellular and gene-therapy.

The genetic modification of T cells for the expression a chimeric antigen receptor (CAR) endows them with a new specificity for an antigen. Adoptive immunotherapy with CD19-CAR T cells has achieved high rates of sustained complete remissions in B cell malignancies. However, the downregulation or loss of the targeted antigen after mono-specific CAR T cell therapy, e.g. against CD19 or CD22, has been reported. Targeting multiple antigens on tumour cells, sequentially or simultaneously, could overcome this limitation. Additionally, targeting multiple antigens with CAR T cells could drive the translation from hematologic malignancies to prevalent solid cancers, which often express tumour-associated antigens heterogeneously. We hypothesised that expression of a universal CAR, which can be programmed with hapten-like molecules, could endow T cells with specificities for multiple antigens. In this study we introduce a novel chemically programmable CAR (cpCAR) based on monoclonal antibody h38C2. Our data show, that cpCARs form a reversible chemical bond to molecules containing a diketone-group and therefore can be programmed to acquire multiple specificities. We programmed cpCAR T cells with hapten-like compounds against integrins αvβ3 and α4β1 as well as the folate receptor. We observed tumour cell lysis, IFN ɣ and IL-2 production and proliferation of programmed cpCAR T cells against tumour cells expressing the respective target antigen in vitro. As a reference to cpCARs programmed against αvβ3, we further introduced novel conventional αvβ3-CARs. These CARs, based on humanised variants of monoclonal antibody LM609 (hLM609), directly bind to integrin αvβ3 via their scFv. The four αvβ3-CAR constructs comprised either an scFv with higher affinity (hLM609v7) or lower affinity (hLM609v11) against αvβ3 integrin and either a long (IgG4 hinge, CH2, CH3) or short (IgG4 hinge) extracellular spacer. We selected the hLM609v7-CAR with short spacer, which showed potent anti-tumour reactivity both in vitro and in a murine xenograft model, for comparison with the cpCAR programmed against αvβ3. Our data show specific lysis of αvβ3-positive tumour cells, cytokine production and proliferation of both hLM609-CAR T cells and cpCAR T cells in vitro. However, conventional hLM609-CAR T cells mediated stronger anti-tumour effects compared to cpCAR T cells in the same amount of time. In line with the in vitro data, complete destruction of tumour lesions in a murine melanoma xenograft model was only observed for mice treated with conventional αvβ3-CAR T cells. Collectively, we introduce a cpCAR, which can be programmed against multiple tumour antigens, and hLM609-CARs specific for the integrin αvβ3. The cpCAR technology bears the potential to counteract current limitations, e.g. antigen loss, of current monospecific CAR T cell therapy. Targeting αvβ3 integrin with CAR T cells could have clinical applications in the treatment of solid malignancies, because αvβ3 is not only expressed on a variety of solid malignancies, but also on tumour-associated vasculature and fibroblast.

Adoptive immunotherapy using chimeric antigen receptor (CAR)-modified T cells is an effective treatment for hematological malignancies that are refractory to conventional chemotherapy. To address a wider variety of cancer entities, there is a need to identify and characterize additional target antigens for CAR-T cell therapy. The two members of the receptor tyrosine kinase-like orphan receptor family, ROR1 and ROR2, have been found to be overexpressed on cancer cells and to correlate with aggressive cancer phenotypes. Recently, ROR1-specific CAR-T cells have entered testing in phase I clinical trials, encouraging us to assess the suitability of ROR2 as a novel target for CAR-T cell therapy. To study the therapeutic potential of targeting ROR2 in solid and hematological malignancies, we selected two representative cancer entities with high unmet medical need: renal cell carcinoma and multiple myeloma. Our data show that ROR2 is commonly expressed on primary samples and cell lines of clear cell renal cell carcinoma and multiple myeloma. To study the efficacy of ROR2-specific CAR T cell therapy, we designed two CAR constructs with 10-fold binding affinity differences for the same epitope of ROR2. We found both cell products to exhibit antigen-specific anti-tumor reactivity in vitro, including tumor cell lysis, secretion of the effector cytokines interleukin-2 (IL-2) and interferon-gamma (IFNγ), and T cell proliferation. In vivo studies revealed ROR2 specific CAR-T cells to confer durable responses, significant survival benefits and long-term persistence of CAR-expressing T cells. Overall, there was a trend towards more potent anti-tumor efficacy upon treatment with T cells that expressed the CAR with higher affinity for ROR2, both in vitro and in vivo. We performed a preclinical safety and toxicology assessment comprising analyses of ROR2 expression in healthy human and murine tissues, cross-reactivity, and adoptive T cell transfer in immunodeficient mice. We found ROR2 expression to be conserved in mice, and low-level expression was detectable in the male and female reproductive system as well as parts of the gastrointestinal tract. CAR-T cells targeting human ROR2 were found to elicit similarly potent reactivity upon recognition of murine ROR2. In vivo analyses showed transient tissue-specific enrichment and activation of ROR2-specific CAR-T cells in organs with high blood circulation, such as lung, liver, or spleen, without evidence for clinical toxicity or tissue damage as determined by histological analyses. Furthermore, we humanized the CAR binding domain of ROR2-specific CAR-T cells to mitigate the risk of adverse immune reactions and concomitant CAR-T cell rejection. Functional analyses confirmed that humanized CARs retained their specificity and functionality against ROR2-positive tumor cells in vitro. In summary, we show that ROR2 is a prevalent target in RCC and MM, which can be addressed effectively with ROR2-specific CAR-T cells in preclinical models. Our preliminary toxicity studies suggest a favorable safety profile for ROR2-specific CAR-T cells. These findings support the potential to develop ROR2-specific CAR-T cells clinically to obtain cell products with broad utility.

Immunotherapy with engineered T cells expressing a tumor-specific chimeric antigen receptor (CAR) is under intense preclinical and clinical investigation. This involves a rapidly increasing portfolio of novel target antigens and CAR designs that need to be tested in time- and work-intensive screening campaigns in primary T cells. Therefore, we anticipated that a standardized screening platform, similar as in pharmaceutical small molecule and antibody discovery, would facilitate the analysis of CARs by pre-selecting lead candidates from a large pool of constructs that differ in their extracellular and intracellular modules. Because CARs integrate structural elements of the T cell receptor (TCR) complex and engage TCR-associated signaling molecules upon stimulation, we reasoned that the transcription factors nuclear factor-κB (NF-κB) and nuclear factor of activated T cells (NFAT) could serve as surrogate markers for primary T cell function. The nuclear translocation of both transcription factors in primary T cells, which we observed following CAR stimulation, supported our rationale to use NF-κB and NFAT as indicators of CAR-mediated activation in a screening platform. To enable standardized and convenient analyses, we have established a CAR-screening platform based on the human T cell lymphoma line Jurkat that has been modified to provide rapid detection of NF-κB and NFAT activation. For this purpose, Jurkat cells contained NF-κB and NFAT-inducible reporter genes that generate a duplex output of cyan fluorescent protein (CFP) and green fluorescent protein (GFP), respectively. Upon stimulation of NF-κB/NFAT reporter cells, the expression of both fluorophores could be readily quantified in high-throughput screening campaigns by flow cytometry. We modified the reporter cells with CD19-specific and ROR1-specific CARs, and we co-cultured them with antigen-positive stimulator cells to analyze NF-κB and NFAT activation. CAR-induced reporter signals could already be detected after 6 hours. The optimal readout window with high-level reporter activation was set to 24 hours, allowing the CAR-screening platform to deliver results in a rapid turnaround time. A reporter cell-screening campaign of a spacer library with CARs comprising a short, intermediate or long IgG4-Fc domain allowed distinguishing functional from non-functional constructs. Similarly, reporter cell-based analyses identified a ROR1-CAR with 4-1BB domain from a library with different intracellular signal modules due to its ability to confer high NF-κB activation, consistent with data from in vitro and in vivo studies with primary T cells. The results of both CAR screening campaigns were highly reproducible, and the time required for completing each testing campaign was substantially shorter with reporter cells (6 days) compared to primary T cells (21 days). We further challenged the reporter cells in a large-scale screening campaign with a ROR1 CAR library comprising mutations in the VH CDR3 sequence of the R11 scFv. This region is crucial for binding the R11 epitope of ROR1, and we anticipated that mutations here would cause a loss of specificity and affinity for most of the CAR variants. This provided the opportunity to determine whether the CAR screening platform was able to retrieve functional constructs from a large pool of CAR variants. Indeed, using a customized pre enrichment and screening strategy, the reporter cells identified a functional CAR variant that was present with a frequency of only 6 in 1.05x10^6. As our CAR-screening platform enabled the analysis of activating signal modules, it encouraged us to also evaluate inhibitory signal modules that change the CAR mode of action. Such an inhibitory CAR (iCAR) can be used in logic gates with an activating CAR to interfere with T cell stimulation. By selecting appropriate target antigens for iCAR and CAR, this novel application aims to improve the selectivity towards tumor cells, and it could readily be studied using our screening platform. Accordingly, we tested CD19-specific iCARs with inhibitory PD-1 signal module for their suppressive effect on reporter gene activation. In logic gates with CAR or TCR stimulation, a decrease of NF-κB and NFAT signals was only observed when activating and inhibitory receptors were forced into spatial proximity. These results were further verified by experiments with primary T cells. In conclusion, our reporter cell system is attractive as a platform technology because it is independent of testing in primary T cells, exportable between laboratories, and scalable to enable small- to large-scale screening campaigns of CAR libraries. The pre-selection of appropriate lead candidates with optimal extracellular and intracellular modules can reduce the number of CAR constructs to be investigated in further in vitro and in vivo studies with primary T cells. We are therefore confident that our CAR-screening platform based on NF-κB/NFAT reporter cells will be useful to accelerate translational research by facilitating the evaluation of CARs with novel design parameters.

Adoptive cellular immunotherapy with chimeric antigen receptor (CAR) T cells is highly effective in haematological malignancies. This success, however, has not been achieved in solid tumours so far. In contrast to hematologic malignancies, solid tumours include a hostile tumour microenvironment (TME), that poses additional challenges for curative effects and consistent therapeutic outcome. These challenges manifest in physical and immunological barriers that dampen efficacy of the CAR T cells. Preclinical testing of novel cellular immunotherapies is performed mainly in 2D cell culture and animal experiments. While 2D cell culture is an easy technique for efficacy analysis, animal studies reveal information about toxicity in vivo. However, 2D cell culture cannot fully reflect the complexity observed in vivo, because cells are cultured without anchorage to a matrix and only short-term periods are feasible. Animal studies provide a more complex tissue environment, but xenografts often lack human stroma and tumour inoculation occurs mostly ectopically. This emphasises the need for standardisable and scalable tumour models with incorporated TME-aspects, which enable preclinical testing with enhanced predictive value for the clinical outcome of immunotherapies. Therefore, microphysiologic 3D tumour models based on the biological SISmuc (Small Intestinal mucosa and Submucosa) matrix with preserved basement membrane were engaged and improved in this work to serve as a modular and versatile tumour model for efficacy testing of CAR T cells. In order to reflect a variety of cancer entities, TME-aspects, long-term stability and to enhance the read-out options they were further adapted to achieve scalable and standardisable defined microphysiologic 3D tumour models. In this work, novel culture modalities (semi-static, sandwich-culture) were characterised and established that led to an increased and organised tissue generation and long-term stability. Application of the SISmuc matrix was extended to sarcoma and melanoma models and serial bioluminescence intensity (BLI)-based in vivo imaging analysis was established in the microphysiologic 3D tumour models, which represents a time-efficient read-out method for quality evaluation of the models and treatment efficacy analysis, that is independent of the cell phenotype. Isolation of cancer-associated-fibroblasts (CAFs) from lung (tumour) tissue was demonstrated and CAF-implementation further led to stromal-enriched microphysiologic 3D tumour models with in vivo-comparable tissue-like architecture. Presence of CAFs was confirmed by CAF-associated markers (FAP, α-SMA, MMP-2/-9) and cytokines correlated with CAF phenotype, angiogenesis, invasion and immunomodulation. Additionally, an endothelial cell barrier was implemented for static and dynamic culture in a novel bioreactor set-up, which is of particular interest for the analysis of immune cell diapedesis. Studies in microphysiologic 3D Ewing’s sarcoma models indicated that sarcoma cells could be sensitised for GD2-targeting CAR T cells. After enhancing the scale of assessment of the microphysiologic 3D tumour models and improving them for CAR T cell testing, the tumour models were used to analyse their sensitivity towards differently designed receptor tyrosine kinase-like orphan receptor 1 (ROR1) CAR T cells and to study the effects of the incorporated TME-aspects on the CAR T cell treatment respectively. ROR1 has been described as a suitable target for several malignancies including triple negative breast cancer (TNBC), as well as lung cancer. Therefore, microphysiologic 3D TNBC and lung cancer models were established. Analysis of ROR1 CAR T cells that differed in costimulation, spacer length and targeting domain, revealed, that the microphysiologic 3D tumour models are highly sensitive and can distinguish optimal from sub-optimal CAR design. Here, higher affinity of the targeting domain induced stronger anti-tumour efficacy and anti-tumour function depended on spacer length, respectively. Long-term treatment for 14 days with ROR1 CAR T cells was demonstrated in dynamic microphysiologic 3D lung tumour models, which did not result in complete tumour cell removal, whereas direct injection of CAR T cells into TNBC and lung tumour models represented an alternative route of application in addition to administration via the medium flow, as it induced strong anti-tumour response. Influence of the incorporated TME-aspects on ROR1 CAR T cell therapy represented by CAF-incorporation and/or TGF-β supplementation was analysed. Presence of TGF-β revealed that the specific TGF-β receptor inhibitor SD-208 improves ROR1 CAR T cell function, because it effectively abrogated immunosuppressive effects of TGF-β in TNBC models. Implementation of CAFs should provide a physical and immunological barrier towards ROR1 CAR T cells, which, however, was not confirmed, as ROR1 CAR T cell function was retained in the presence of CAFs in stromal-enriched microphysiologic 3D lung tumour models. The absence of an effect of CAF enrichment on CAR T cell efficacy suggests a missing component for the development of an immunosuppressive TME, even though immunomodulatory cytokines were detected in co-culture models. Finally, improved gene-edited ROR1 CAR T cells lacking exhaustion-associated genes (PD-1, TGF-β-receptor or both) were challenged by the combination of CAF-enrichment and TGF-β in microphysiologic 3D TNBC models. Results indicated that the absence of PD-1 and TGF-β receptor leads to improved CAR T cells, that induce strong tumour cell lysis, and are protected against the hostile TME. Collectively, the microphysiologic 3D tumour models presented in this work reflect aspects of the hostile TME of solid tumours, engage BLI-based analysis and provide long-term tissue homeostasis. Therefore, they present a defined, scalable, reproducible, standardisable and exportable model for translational research with enhanced predictive value for efficacy testing and candidate selection of cellular immunotherapy, as exemplified by ROR1 CAR T cells.