FLT3 OR CD33 NOT EMCN Logic Gated CAR-NK Cell Therapy (SENTI-202) for Precise Targeting of AML

Background: While chimeric antigen receptor (CAR) cell therapies have provided extraordinary clinical responses in some hematological malignancies, developing effective CAR cell therapies for acute myeloid leukemia (AML) has been challenging due to: (a) the lack of a single target antigen robustly expressed across both AML leukemic stem cell (LSC) and immature leukemic blast cell subpopulations, and (b) the lack of truly AML-specific target antigens, since current targets are also expressed on healthy tissues and may result in off-tumor toxicity. Using logic gated gene circuits, we are engineering our SENTI-202 CAR-NK cell therapy to overcome these long-standing challenges to treating AML patients. Methods: To maximize clearance of AML tumor subpopulations and minimize off-tissue toxicities, we used a proprietary bioinformatics paired antigen discovery platform to identify the optimal combinations of AML tumor-associated and healthy tissue antigens to target using an OR and NOT logic gated CAR gene circuit approach. The SENTI-202 therapeutic candidate is a FLT3 OR CD33 NOT Endomucin (EMCN) gene circuit-enabled allogeneic CAR-NK cell, designed to broadly target FLT3 and/or CD33-expressing AML tumor cells (including both LSCs and blasts) but not healthy hematopoietic stem cells (HSCs). Results: First, for the OR GATE portion of the logic circuit we demonstrated that engineered primary human NK cells expressing activating CARs (aCARs) that recognize both FLT3 and CD33 outperformed more traditional single target CAR approaches with FLT3 (p<0.05) or CD33 (p<0.01), and exhibited >80% cytotoxicity and significant cytokine secretion (GrB, IFN-g, and TNF-a) against multiple leukemia cell lines in vitro, including MOLM13, THP1, and SEM. We successfully engineered FLT3 OR CD33 CAR-NK cells using both bicistronic and bivalent CAR configurations, where bicistronic CARs possess separate FLT3 and CD33 CARs linked via a 2A peptide, and bivalent CARs use a loop structure to connect FLT3 and CD33 scFvs within the same CAR. While both approaches demonstrated robust efficacy against AML cells, the bivalent approach enabled greater CAR expression and cytotoxicity (p<0.05). Importantly, our FLT3 OR CD33 CAR-NK cells demonstrated significant cytotoxicity against primary AML patient samples (p<0.01-0.001) and significantly reduced tumor burden and improved mouse survival in MOLM13 (p<0.05) and MV4-11 (p<0.01) xenograft AML models. We believe that our strategy of concurrently targeting FLT3 and CD33 will result in a more robust synergistic anti-tumor effect, leading to a more durable remission with decreased risk of relapse due to single antigen escape. Second, for the NOT GATE portion of the logic circuit to protect healthy HSCs, we developed NK and T cell inhibitory CARs (iCARs) consisting of an scFv against a healthy cell antigen, hinge and transmembrane domains, and functional intracellular domains derived from inhibitory co-receptors containing immunoreceptor tyrosine-based inhibitory motifs. In the case of SENTI-202, the iCAR scFv recognizes EMCN, a surface antigen expressed on up to 76% of healthy HSCs but not on AML cells. Using two different iCAR configurations, we demonstrated that FLT3 (CD28z) aCAR-NK cells engineered with an EMCN-specific iCAR protected up to 67% (iCAR#1, p<0.01) or 50% (iCAR#2, p<0.01) of FLT3+ EMCN+ cells from FLT3 aCAR-mediated cytotoxicity. Next, to replicate a clinical context more closely, we mixed FLT3+ EMCN- (AML-like) and FLT3+ EMCN+ (HSC-like) target cells and demonstrated that FLT3 NOT EMCN CAR-NK cells exhibit preferential killing of FLT3+ EMCN- target cells (p<0.0001), demonstrating that our NOT GATED gene circuit controls NK-mediated responses on a cell-by-cell basis. Conclusion: SENTI-202 is a novel NK cell product candidate to be engineered with both OR and NOT logic gated CAR gene circuits, wherein the OR gate is designed to increase AML LSC/blast tumor clearance (to prevent relapse), and the NOT gate is designed to protect healthy HSCs from off-tumor toxicity, enabling regeneration of a healthy hematopoietic system and mitigating the need for a bone marrow transplant. Beyond AML, OR and NOT logic gated CAR-NK cell therapy has applicability to other cancer-associated antigen targets that are potentially limited by antigen escape and/or off-tumor toxicity, increasing the potential for enhanced efficacy and reduced risk of undesirable side effects. Disclosures No relevant conflicts of interest to declare.

Gene-circuit therapy on the horizon: synthetic biology tools for engineered therapeutics

Therapeutic genome modification requires precise control over the introduced therapeutic functions. Current approaches of gene and cell therapy fail to deliver such command and rely on semi-quantitative methods with limited influence on timing, contextuality and levels of transgene expression, and hence on therapeutic function. Synthetic biology offers new opportunities for quantitative functionality in designing therapeutic systems and their components. Here, we discuss synthetic biology tools in their therapeutic context, with examples of proof-of-principle and clinical applications of engineered synthetic biomolecules and higher-order functional systems, i.e. gene circuits. We also present the prospects of future development towards advanced gene-circuit therapy.

Model-guided engineering of DNA sequences with predictable site-specific recombination rates

Site-specific recombination (SSR) is an important tool in genome editing and gene circuit design. However, its applications are limited by the inability to simply and predictably tune SSR reaction rates across orders of magnitude. Facile rate manipulation can in principle be achieved by modifying the nucleotide sequence of the DNA substrate of the recombinase, but the design principles for rationally doing so have not been elucidated. To enable predictable tuning of SSR reaction kinetics via DNA sequence, we developed an integrated experimental and computational method to parse individual nucleotide contributions to the overall reaction rate, which we used to analyze and engineer the DNA attachment sequence attP for the inversion reaction mediated by the serine recombinase Bxb1. A quantitative PCR method was developed to measure the Bxb1 reaction rate in vitro. Then, attP sequence libraries were designed, selected, and sequenced to inform a machine-learning model, which revealed that the Bxb1 reaction rate can be accurately represented assuming independent contributions of nucleotides at key positions. Next, we used the model to predict the performance of DNA site variants in reaction rate assays both in vitro and in Escherichia coli, with flipping rates ranging from 0.01- to 10-fold that of the wild-type attP sequence. Finally, we demonstrate that attP variants with predictable DNA recombination rates can be used in concert to achieve kinetic control in gene circuit design by coordinating the co-expression of two proteins in both their relative proportion and their total amount. Our high-throughput, data-driven method for rationally tuning SSR reaction rates through DNA sequence modification enhances our understanding of recombinase function and expands the synthetic biology toolbox.

Scalable recombinase-based gene expression cascades

Temporal modulation of the expression of multiple genes underlies complex complex biological phenomena. However, there are few scalable and generalizable gene circuit architectures for the programming of sequential genetic perturbations. Here, we describe a modular recombinase-based gene circuit architecture, comprising tandem gene perturbation cassettes (GPCs), that enables the sequential expression of multiple genes in a defined temporal order by alternating treatment with just two orthogonal ligands. We use tandem GPCs to sequentially express single-guide RNAs to encode transcriptional cascades that trigger the sequential accumulation of mutations. We build an all-in-one gene circuit that sequentially edits genomic loci, synchronizes cells at a specific stage within a gene expression cascade, and deletes itself for safety. Tandem GPCs offer a multi-tiered cellular programming tool for modeling multi-stage genetic changes, such as tumorigenesis and cellular differentiation.

A Synthetic, Closed-Looped Gene Circuit for the Autonomous Regulation of RUNX2 Activity during Chondrogenesis

The transcription factor RUNX2 is a key regulator of chondrocyte phenotype during development, making it an ideal target for prevention of undesirable chondrocyte maturation in cartilage tissue engineering strategies. Here, we engineered an autoregulatory gene circuit (cisCXp-shRunx2) that negatively controls RUNX2 activity in chondrogenic cells via RNA interference initiated by a tunable synthetic Col10a1-like promoter (cisCXp). The cisCXp-shRunx2 gene circuit is designed based on the observation that induced RUNX2 silencing after early chondrogenesis enhances the accumulation of cartilaginous matrix in 2D ATDC5 model. We show that the cisCXp-shRunx2 initiates RNAi of RUNX2 in maturing chondrocytes in response to the increasing intracellular RUNX2 activity without interfering with early chondrogenesis in ATDC5 cells. The induced loss of RUNX2 activity in turn negatively regulates the gene circuit itself. Furthermore, the efficacy of RUNX2 suppression from cisCXp-shRunx2 can be controlled by modifying the sensitivity of cisCXp promoter. Long-term 3D cultures of reprogrammed ATDC5 cells had increased matrix accumulation compared to naive cells. Overall, our results demonstrate that the negative modulation of Runx2 activity with our autoregulatory gene circuit can reduce the effects of RUNX2 activity and enhance matrix synthesis in chondroprogenitor cells.

Emergent Oscillation Induced by Nutrient-Modulating Growth Feedback

Growth feedback, the inherent coupling between the synthetic gene circuit and the host cell growth, could significantly change the circuit behaviors. Previously, a diverse array of emerged behaviors, such as growth bistability, enhanced ultrasensitivity, and topology-dependent memory loss, were reported to be induced by growth feedback. However, the influence of the growth feedback on the circuit functions remains underexplored. Here, we reported an unexpected oscillatory behavior of a self-activation gene circuit induced by nutrient-modulating growth feedback. Specifically, after dilution of the activated self-activation switch into the fresh medium with moderate nutrient, its gene expression first decreases as the cell grows and then shows a significant overshoot before it reaches the steady states, leading to oscillation dynamics. Fitting the data with a coarse-grained model suggests a nonmonotonic growth-rate regulation on gene production rate. The underlying mechanism of the oscillation was demonstrated by a molecular mathematical model, which includes the ribosome allocation towards gene production, cell growth, and cell maintenance. Interestingly, the model predicted a counterintuitive dependence of oscillation amplitude on the nutrition level, where the highest peak was found in the medium with a moderate nutrient but was not observed in rich nutrient. We experimentally verified this prediction by tuning the nutrient level in the culture medium. We did not observe significant oscillatory behavior for toggle switch, suggesting that the emergence of oscillatory behavior depends on circuit network topology. Our results demonstrated a new nonlinear emergent behavior mediated by growth feedback, which depends on the ribosome allocation between gene circuit and cell growth.

A glucose meter interface for point-of-care gene circuit-based diagnostics

Recent advances in cell-free synthetic biology have given rise to gene circuit-based sensors with the potential to provide decentralized and low-cost molecular diagnostics. However, it remains a challenge to deliver this sensing capacity into the hands of users in a practical manner. Here, we leverage the glucose meter, one of the most widely available point-of-care sensing devices, to serve as a universal reader for these decentralized diagnostics. We describe a molecular translator that can convert the activation of conventional gene circuit-based sensors into a glucose output that can be read by off-the-shelf glucose meters. We show the development of new glucogenic reporter systems, multiplexed reporter outputs and detection of nucleic acid targets down to the low attomolar range. Using this glucose-meter interface, we demonstrate the detection of a small-molecule analyte; sample-to-result diagnostics for typhoid, paratyphoid A/B; and show the potential for pandemic response with nucleic acid sensors for SARS-CoV-2.