Predictive Modeling of Non-Viral Gene Transfer

In non-viral gene delivery, the variance of transgenic expression stems from the low number of plasmids successfully transferred. Here, we experimentally determine Lipofectamine- and PEI-mediated exogenous gene expression distributions from single cell time-lapse analysis. Broad Poisson-like distributions of steady state expression are observed for both transfection agents, when used with synchronized cell lines. At the same time, co-transfection analysis with YFP- and CFP-coding plasmids shows that multiple plasmids are simultaneously expressed, suggesting that plasmids are delivered in correlated units (complexes). We present a mathematical model of transfection, where a stochastic, two-step process is assumed, with the first being the low-probability entry step of complexes into the nucleus, followed by the subsequent release and activation of a small number of plasmids from a delivered complex. This conceptually simple model consistently predicts the observed fraction of transfected cells, the cotransfection ratio and the expression level distribution. It yields the number of efficient plasmids per complex and elucidates the origin of the associated noise, consequently providing a platform for evaluating and improving non-viral vectors.

💡 Research Summary

This paper addresses the pronounced cell‑to‑cell variability observed in non‑viral gene delivery, which stems from the fact that only a few plasmid molecules successfully reach the nucleus and become transcriptionally active. Using time‑lapse fluorescence microscopy, the authors quantified expression from Lipofectamine‑ and PEI‑mediated transfections in synchronized cell populations. Both reagents produced broad, Poisson‑like distributions of steady‑state fluorescence, indicating that the number of functional plasmids per cell is highly stochastic. A key observation came from cotransfection experiments with YFP‑ and CFP‑encoding plasmids: a large fraction of cells expressed both reporters simultaneously, far exceeding the probability expected if plasmids entered the cell independently. This result suggests that plasmids are delivered in correlated units, i.e., as complexes that either enter the nucleus together or not at all.

To explain these findings, the authors propose a simple two‑step stochastic model. The first step represents the rare event of a DNA‑lipid or DNA‑polymer complex crossing the nuclear envelope; the number of complexes that succeed follows a Poisson distribution with mean λ₁, which is on the order of 10⁻³ per cell. The second step describes the release of individual plasmids from each nuclear‑localized complex; the released plasmid count per complex is also Poisson‑distributed with mean λ₂ (≈1–2). Because the two steps are independent, the total number of active plasmids per cell follows a compound‑Poisson distribution. This formulation naturally reproduces the experimentally observed wide expression distribution and the elevated cotransfection ratio.

Parameter values were inferred by maximum‑likelihood fitting to three experimental observables: the fraction of transfected cells, the proportion of double‑positive cells in cotransfection assays, and the mean fluorescence intensity per cell. The fitted parameters indicate that, on average, only 0.1–0.2 % of complexes manage nuclear entry (λ₁≈0.001–0.002) and that each successful complex delivers roughly 1.5–2 functional plasmids (λ₂≈1.5–2). The model accurately predicts the transfection efficiency (≈12 % observed vs. 12 % predicted) and the cotransfection ratio (≈36 % observed vs. 38 % predicted), and its probability density matches the full fluorescence histogram, including the heavy tail.

Beyond fitting, the model provides mechanistic insight into the sources of noise in non‑viral delivery. The dominant contributor is the stochasticity of the nuclear entry step, while variability in plasmid release adds a secondary layer of noise. Consequently, strategies that increase λ₁ (e.g., nuclear‑targeting peptides, electroporation, or microfluidic confinement) or that fine‑tune λ₂ (optimizing plasmid loading per complex, using degradable polymers to control release kinetics) are predicted to reduce expression heterogeneity and raise overall efficiency. The authors discuss how such interventions could be systematically evaluated using the presented framework.

In summary, the study delivers a concise yet powerful quantitative description of non‑viral gene transfer, linking the physical process of complex trafficking to the statistical properties of gene expression. By capturing both the low‑probability entry event and the Poisson‑distributed plasmid release, the model explains the observed Poisson‑like expression distribution, the high cotransfection ratio, and the overall noise characteristics. This work therefore offers a valuable platform for rational design and optimization of non‑viral vectors, guiding future efforts to achieve more reliable and homogeneous gene delivery.