Instead of editing a T cell's genome ex vivo, an in situ CAR-T approach injects lipid nanoparticles (LNPs) carrying CAR-encoding mRNA directly into the bloodstream. Each LNP is coated with an anti-CD3/CD8 targeting ligand that lets it dock onto a passing T cell.
Per collision, binding succeeds with probability set by the targeting affinity. A bound LNP is then endocytosed; only a fraction — the endosomal escape efficiency — actually ruptures the endosome and releases its mRNA payload into the cytoplasm. Most LNPs are degraded in the lysosome before escaping, which is why real-world escape efficiency is usually well under 5% (this sim exaggerates it for visibility).
Inside the cell, mRNA is translated into CAR protein and both pools decay — this is a two-compartment linear ODE integrated every frame:
dM/dt = −(k_transl + k_Mdeg)·M (M = mRNA pool, + delivery pulses)
dC/dt = k_transl·M − k_Cdeg·C (C = CAR surface expression)
k_Cdeg = ln(2) / half-life
Because mRNA is never integrated into the genome, expression is transient: it rises, peaks, then decays with the CAR protein's own half-life — the slider you control. A T cell becomes CAR+ once C crosses an activation threshold; while active and within range of a tumor cell it engages and kills it. Lower the half-life and the kill window shortens even at maximum dose — the same dosing/frequency trade-off real in situ CAR-T programs face.