ML-Driven Design of Lipid Nanoparticles for In-Vivo T-Cell Delivery

T-cell-based immunotherapies like CAR-T have revolutionized cancer treatment, offering life-saving treatment for patients with refractory cancers. However, high manufacturing costs, challenging supply chains, labor-intensive ex-vivo manipulation, and prerequisite lymphodepletion limit scalability. In-vivo mRNA delivery to reprogram T-cells directly within the body offers a potentially transformative alternative by streamlining production, reducing costs, and broadening patient access. However, developing efficient delivery systems to target T-cells remains a critical challenge.

To address the complexity of lipid nanoparticle (LNP) development, Mana.bio has created a machine learning-Driven platform to optimize LNPs for T-cell transfection in-vivo and ex-vivo. Our platform integrates over 80,000 data points from proprietary experiments and 45,000 from public literature, using advanced ML algorithms to screen ~50 million formulations. The model demonstrates high performance with predicted T-cell transfection accuracy (MAE: 3.72, Pearson correlation: 0.83) and T-cell viability (MAE: 13.71, Pearson correlation: 0.72), effectively predicting lipids and LNP compositions based on key parameters. These algorithms rapidly identify LNP candidates for therapeutic development based on physicochemical properties, toxicity profiles, and the likelihood of successful T-cell transfection.

Initial validation using eGFP mRNA showed over 90% transfection efficiency in activated CD3+ T-cells from human peripheral blood mononuclear cells (hPBMCs) within 24 hours, with >80% cell viability, effectively outperforming an industry-standard immune cell-transfection reagent at the same dose. Notably, our LNPs also maintained high transfection efficiency (70%) in non-activated CD3+ hPBMC T cells. Mana’s formulations exhibited enhanced transfection efficacy compared to MC3, a clinically validated LNP, as well as toxicity profiles assessed by measuring TNF-α, IL-1β, and IL-6 levels in both activated and non-activated hPBMCs 24 hours post treatment.

In-vivo activity was validated in NSG mice injected with hPBMCs three days before LNP administration, and eGFP expression was evaluated in blood, spleen, lung, and liver after 24 hours. The results demonstrated successful transfection, with lead LNP candidates yielded ~25% transfection in CD3+ T-cells. Lead formulation preferentially transfected subsets of CD4+ and CD8+ T cells with our leading formulation, achieving 18% and 24% transfection, respectively. Spleen analysis exhibited a 10% transfection rate of CD3+ T-cells. Immuno-histological analysis of liver tissues revealed minimal to no eGFP expression, while flow cytometry analysis of lung sections showed only 1% of lung cells transfected with eGFP; thus demonstrating selective T-cell targeting in-vivo.