Krotova, Karina Y.

Abstract Chimeric Antigen Receptor (CAR) T cell therapy is a transformative treatment for hematologic malignancies. However, ex vivo autologous CAR T cell manufacturing presents substantial logistical and clinical challenges. Direct in vivo CAR-T cell engineering offers a new paradigm for CAR T delivery. We recently developed the CD3-targeted lentiviral vector LV-169 which can deliver a CAR transgene to resting niche-resident T cells in lymphoid tissues. LV169 is pseudotyped with a mix of chimeric and nonchimeric VSV-G proteins detargeted (blinded) from the natural LDL receptor and re-targeted to T cells through the display of a CD3-specific scFv. Here we demonstrate that LV169 can specifically deliver a functional aBCMA-CAR transgene to resting human T cells in PBMC cultures from the healthy volunteers and from multiple myeloma patients. For clinical translation, we selected an αBCMA-CAR construct comprising a fully human single heavy chain nanobody linked to 41BB and CD3z cytoplasmic signaling domains. Human T cells expressing this CAR construct have low tonic signaling and efficiently kill BCMA-positive myeloma target cells with similar or lower associated cytokine release compared to a clinically relevant benchmark BCMA-CAR. To further improve LV169 and to minimize CAR protein incorporation in the LV envelope which may lead undesirably to myeloma cell transduction, we placed the CAR transgene under a T-cell specific Lck promoter in reverse orientation relative to the 5' LTR. The transcriptional activity of the Lck promoter was comparable to that of a constitutive EF1a promoter in primary T cells and, in contrast to EF1a had minimal activity in a variety of other tested cell types including the HEK293T producer cell line and the THP-1 monocytic cell line. Western blotting confirmed the absence of CAR protein in LV particles (LVP) generated using the Lck promoter-driven reverse orientation construct. In contrast, high levels of CAR protein were detected in LVP generated using the parental EF1a promoter-driven construct. The LV-incorporated CAR protein mediated off-target gene delivery to BCMA-positive myeloma cells. The therapeutic efficacy of intravenously administered LV169 LV particles carrying BCMA-CAR constructs was tested in the OPM-2 human multiple myeloma xenograft model. NSG mice with knockout of MHC-I and MHC-II, and with advanced disseminated OPM-2-Luc tumors were engrafted with human PBMC one week prior to intravenous LV169-BCMA-CAR administration. Mice were treated with single intravenous injection of LV169 at different doses 3 weeks after tumor engraftment. OPM-2-Luc tumor burden was monitored by regular luciferase imaging. All treated mice cleared their tumors within 28-days following LV administration and the rate of clearance was correlated to the LV particle dose. Even at the 4e8 dose level (by p24) tumors were completely cleared. Eighty four days post-treatment the treated mice remained tumor-free and resisted re-challenge with OPM-2 tumor cells. Circulating CAR-T cells, monitored by flow cytometry, became detectable by 14 dpi and peaked between 21-35 dpi. Following tumor clearance, the number of circulating CAR-T cells declined dramatically. The treatment was well tolerated and analysis of cytokines associated with CAR-T cell cytotoxicity confirmed the absence of a severe cytokine storm. IFNγ was the only cytokine which was elevated during the expansion period of CAR-T cells, while IL-2, IL-6, TNFα and GM-CSF were barely detectable. In summary the characteristics of our new developed LV169-based platform make it suitable for further clinical translation which is underway.

Abstract Direct in vivo T cell engineering offers a new paradigm for CAR T delivery. We have evaluated several approaches to engineer lentiviral (LV) vectors capable of delivering CAR transgenes directly and selectively to T cells in vivo, within the lymphoid niche. Our goal was to develop an LV delivery platform that following intravenous administration can selectively bind and activate T cells, deliver and integrate the (CAR) transgene into genomic DNA, and drive CAR expression on the T cell surface. Further requirements were that the retargeted LV should be associated with minimal off-target transduction of tumor cells and normal host tissues and should be manufacturable at titers sufficient for human application. Our strategy for retargeting has been to modify the trimeric VSV-G protein, which is well established for efficient LV pseudotyping. Of the approaches evaluated, three retargeting platforms have proven to be effective: direct covalent display, indirect covalent display, and G-capping. Direct covalent display. This (LV-169) approach uses chimeric VSV-G protein that contains a CD3-targeting ligand displayed as a N-terminal extension of VSV-G protein that has been mutated to ablate (blind) its interaction with the LDL receptor. Chimeric and nonchimeric blinded G proteins are co- incorporated into LV particles within mixed envelopes which bind to CD3. Upon binding to cell surface CD3, the engineered particles activate T-cells and enter acidifying endosomes, whereupon they trigger membrane fusion and deliver their payloads. LV-169 displaying an anti-human CD3 scFv selectively transduced resting human T cells in PBMC cultures, in fresh human whole blood and following intravenous administration to humanized mice. In immunocompetent Cre reporter mice, surrogate vectors displaying anti-murine CD3 scFv selectively transduced lymph node resident T cells, with off-target transduction being limited to tissue-resident macrophages. To eliminate the incorporation of CAR protein into the LV-169 envelope, which can lead to highly undesirable transduction of tumor cells expressing the CAR target, we placed the CAR transgene in reverse to 5'LTR orientation, downstream of a T-cell specific promoter/enhancer. In vivo efficacy of LV-169-BCMA-CAR was demonstrated in humanized mice bearing disseminated OPM-2 myeloma tumors. Clinical translation of this construct is now underway. Indirect covalent display. In this G-protein engineering approach, a short peptide (SPY-tag) is displayed as an N-terminal extension of the LDLR-blinded VSV-G protein. The SPY-tagged LV vectors are then covalently modified to display a CD3-targeting ligand by simply mixing them at room temperature with a bifunctional fusion protein consisting of the SPY-catcher and the CD3 ligand. Under these conditions, the SPY-catcher and the SPY-tag spontaneously form a covalent isopeptide bond. The CD3 ligand-modified LV particles were shown to specifically activate and transduce T cells ex vivo within human PBMC cultures. G-capping. In the final approach, the VSV-G protein is not genetically modified. Instead, a “capping” protein crosslinker that binds directly to the unmodified G protein on preformed LV particles is used to mask their LDLR binding sites and redirect them to interact with CD3. Mixing of the CD3-targeted capping protein with traditional G-pseudotyped LV particles leads to efficient binding of the cap to the trimeric G proteins. The cap-modified LV are unable to interact with LDLR but interact efficiently with CD3. CD3-capped LV particles were highly specific and efficient in human T cell activation and transduction in PBMC cultures. The association between the G protein and the bound cap is highly stable during freeze-thaw cycles and at room temperature, a critical attribute of this platform to facilitate the manufacture of a clinically and commercially viable LV product. In summary, we have developed three translationally relevant platform technologies for the generation of LV particles capable of delivering CAR transgenes specifically and efficiently to resting niche-resident T cells after intravenous administration. Manufacturing studies are underway and, in the case of the LV-169 platform, clinical translation is already in progress.

The high cost and complexity of CAR-T therapy has spurred efforts to generate CAR-T cells directly within patients via systemic infusion of serum-stable lentiviral vector particles that target T cells in situ for CAR transgene delivery. To ensure effective in vivo delivery, these vectors need to show resistance to neutralization by human serum, demonstrate transduction specificity for T cells, and support subsequent expansion of CAR-T cells. Lentiviral vectors (LV) pseudotyped with VSV G-glycoprotein (G-WT) transduce target cells via the LDL receptor (LDLR), which is ubiquitously expressed at variable levels and is upregulated on activated T cells. G-WT LVs are widely used to generate therapeutic CAR-T cells ex vivo, following purification and ex vivo activation of the T cells. To generate LVs capable of targeting circulating T cells in vivo, we utilized rational engineering of the VSV-G glycoprotein to design a G-CD3. This modified G protein avoids LDLR recognition, incorporates a CD3-specific single-chain antibody for targeted recognition of T cells, and minimizes LV particle inactivation upon exposure to human serum. G-CD3 LVs were shown to target and transduce resting T cells in peripheral blood mononuclear cell (PBMC) cultures with very high efficiency. The G-CD3 LVs powerfully activated T cells, evident from their increased CD25 expression, and the infection was not hindered in the presence of fresh, complement-active human serum (25%). Additionally, G-CD3 LVs efficiently transduced T cells in fresh human whole blood without apparent off-target transduction of non-T cell lineages. In vivo evaluation of G-CD3 LV efficiency was performed in humanized NSG mice by intravenous (IV) or intraperitoneal (IP) delivery of G-CD3 LVs encoding anti-CD19 CAR and GFP transgenes. Both IV and IP delivery of G-CD3 LV led to in vivo CAR T cell generation with peak levels of 17% of total circulating human CD45+ cells on day 21 post vector administration. The in vivo generated CD19-CAR-T cells were fully functional as evidenced by the disappearance of endogenous human CD19+ B cells from peripheral blood. Tissue analysis of spleen, bone marrow, and liver at the time of sacrifice (day 39 post LV injection), showed anti-CD19 CAR-T in all three tissues accompanied by an absence or dramatic reduction in tissue-resident human CD19+ B cells. In contrast, CAR-T cells were not detected at any time point in humanized NSG mice injected with G-WT LV, regardless of whether the virus was administered IV or IP. In summary, these data highlight the potential applicability of fully retargeted, serum stable LVs for in vivo gene delivery or genome editing, offering promising prospects for future human gene therapy applications.