Poster CC7 - Design of a Soluble Multivalent Notch Agonist for the Generation of T Cell from Pluripotent Stem Cells

Introduction: Human induced pluripotent stem cells (hiPSCs) provide an infinite source of starting cell material for in vitro production of different blood cell types including Hematopoietic stem cells (HSCs), T cells, and dendritic cells for a wide variety of applications in immunotherapy. Of these, iPSC-derived T cells have been further engineered as Chimeric Antigen Receptor (CAR) T cells and have gained much clinical attention for their ability to treat cancer, autoimmune diseases, and infection. However, the large-scale manufacturing of these cell products from hiPSCs in a bioreactor meets significant hurdles. The generation of T cells from iPSCs is achieved through a multi-step cellular differentiation protocol, in which various cytokines and growth factors are added sequentially. One such factor is the Notch ligand, an evolutionarily conserved signaling molecule that plays a key role in embryonic and adult development, which must be presented as a plate-bound protein factor, or from a separate sender cell adhered to the culture plate. These requirements hinder the manufacturing of T cells in a scalable three-dimensional (3D) suspension bioreactor. Therefore, a soluble Notch ligand in lieu of a 2D plate-bound or cell-presented ligand would significantly improve the bioreactor manufacturing of iPSC-derived T cells and is a critical unmet need in realizing the global clinical application of CAR T cells. Here we used computational protein design tools to develop a soluble multivalent Notch agonist and coupled it with in vitro differentiation protocol to generate T cells from hiPSCs in a 3D bioreactor manufacturing platform.

Materials and

Methods: Designing a soluble cytokine-like Notch agonist is especially challenging for two reasons: (i) Notch activation requires cell-cell interaction-mediated mechanical force, and (ii) free soluble Notch ligand is signal inhibitory. We used synthetic biology and computational protein design tools, i.e., Rosetta and RFdiffusion, to overcome these hurdles, and built multivalent oligomeric protein complexes with distinct geometries and numerical valencies (i.e., 2, 3, 5, 6, 8, 60, and 120); coupled the Notch ligand delta-like 4 (DLL4) to the oligomeric scaffolds to generate the multivalent Notch agonist. Among oligomers were designed were: homodimer (C2), homotrimer (C3), helical bundle homotrimer (C3-HB), homopentamer (C5), homohexamer (C6), hexavalent dihedral oligomer (D3), homo-octamer (C8), octavalent dihedral oligomer (D4), and icosahedral cage (I53 or ICOS). Further, these structures fall into two broad geometric groups: (i) cis-DLL4 where DLL4 ligands are conjugated to one side of the structure (i.e., C3-HB, C6, and C8), and (ii) trans-DLL4, where the ligands are affixed to opposite sides of the structure (i.e., C2, C3, D3, D4, and ICOS). We tested the Notch activation efficiency of these Notch agonists in adherent (U2OS) and suspension (K562) reporter cell lines with fluorescent readout. We investigated the structure-function relationship of these Notch agonists to identify the mode of activation, the kinetics of Notch activation, and the cell-cell interaction dynamics through Notch receptor clustering. Finally, we performed the T cell differentiation from both human cord blood hematopoietic progenitor cells and iPSC using these Notch agonists and used a 3D bioreactor for efficient T cell production.

Results, Conclusions, and Discussions: Results and conclusions. We hypothesized that multivalent trans-DLL4 can trigger cell-cell interaction to induce Notch activation through ligand-receptor pulling. Among oligomeric DLL4 complexes tested for Notch activation in reporter U2OS cells, trimeric C3-DLL4 demonstrated the highest activation. Other trans-DLL4 complexes, notably, C2-DLL4, D3-DLL4, D4-DLL4, ICOS-DLL4 exhibited medium activation, whereas, cis-DLL4 complexes exhibited lowest activation. Further, the comparison of Notch1 binding by DLL4 (monovalent), C2-DLL4, and C3-DLL4 revealed a valency-dependent binding and subsequent Notch activation, despite C2-DLL4 and C3-DLL4 having similar trans-DLL4 configuration. On the other hand, higher-order DLL4 complexes have unoccupied DLL4 (not bound to the Notch1 receptor), resulting in lower activation than C3-DLL4. Immunostaining and confocal microscopy studies revealed that Notch activation occurs through cell-cell interactions mediated by oligomeric-DLL4 complexes, where the Notch1 receptor is clustered at the cellular junctions forming a Notch synapse. Collectively, these data showed Notch activation by oligomeric DLL4 complexes is dependent on the geometry, valency, and binding affinity of the complexes, and the signal transduction occurs via receptor clustering and Notch synapse formation. We validated the C3-DLL4 mediated Notch activation in cells with endogenous Notch receptors, i.e., MDA-MB-231 (human breast cancer) and SVG-A (human astrocyte) cells: both these cells expressed robust Notch targeting genes (HES1, HEY1, HEYL, and NRARP) in response to C3-DLL4 treatment. Finally, C3-DLL4 promoted efficient production of CD4+/CD8+ double positive (DP) and CD3+TRCab+ T cells from both human cord blood hematopoietic progenitor cells and hiPSCs in 2D culture and in a 3D bioreactor platform, providing a means to large-scale commercial production of therapeutic T cells.
Discussions. Rationally designed protein agonists can significantly enhance the efficiency of endogenous signaling pathways, providing a powerful means to control cellular functions to develop disease therapeutics. Recent developments in artificial intelligence-based protein design tools facilitated the generation of de novo protein oligomers to engineer a soluble Notch agonist that activated signals through cell-cell interaction and was utilized to generate T cells in bioreactor cultures. Notch signaling is ubiquitously associated with many cellular differentiation processes in embryonic and adult development, and therefore our approach can be used in cost-effective in vitro differentiation of diverse therapeutic cell types.

Acknowledgements (Optional):