Week 1 Reading: Controlling Payload Heterogeneity in Lipid Nanoparticles for RNA-Based Therapeutics

Characterizing Payload Heterogeneity in Lipid Nanoparticles

• Lipid Nanoparticles (LNPs) represent the leading platform for nucleic acid delivery, specifically validated by FDA-approved vaccines using messenger RNA (mRNA).

• Payload heterogeneity is the uneven distribution of “cargo”, or therapeutic content within a vesicle such as an LPN.

• Conventional assembly by mixing lipids and RNA typically yields particles with a heterogeneous, bimodal payload distribution, often including "empty LNPs" (no RNA) and particles with multiple RNA copies.

• Research indicates that heterogeneous siRNA distribution significantly decreases gene knockdown efficiency.

• The study integrates coarse-grained molecular dynamics (MD), kinetic Monte Carlo (kMC) simulations, and single-particle characterization via Cylindrical Illumination Confocal Spectroscopy (CICS).

• The primary discovery is that payload heterogeneity is driven by the balance between RNA diffusion kinetics and lipid self-assembly dynamics, rather than thermodynamic phase separation.

• Actionable design principles proposed involve using turbulent mixing to minimize payload variance and adjusting salt and PEG-lipid content to tune RNA loading.

LNP Assembly Process and Mechanical Drivers

• LNP formation occurs through self-assembly when an aqueous RNA solution mixes with an alcoholic lipid solution (ionizable lipids, PEGylated lipids, cholesterol, and helper lipids).

• Interactions driving self-assembly:

  • The amphiphilic nature of lipids.

  • Electrostatic interactions between negatively charged RNAs and positively charged ionizable lipids.

• Bimodal RNA loading distribution: A common outcome where a population consists of both empty LNPs and LNPs with excessively high RNA loading.

• Therapeutic impacts of heterogeneity:

  • Compromised performance and consistency.

  • Linked to toxicity or reactogenicity; for example, empty LNPs containing YSK13 lipids are associated with liver toxicity.

  • Precise RNA distribution is necessary to minimize the overall administered lipid dose while maintaining potency.

• Previous findings show higher mRNA loading can reduce transfection potency due to deviations from optimal lipid/mRNA ratios and the formation of "bleb-like" structures.

Advanced Multi-Scale Modeling and Characterization Techniques

• Cylindrical Illumination Confocal Spectroscopy (CICS): A single-nanoparticle detection method that analyzes LNP characteristics. It differentiates populations into siRNA-encapsulated LNPs, empty LNPs, and free unencapsulated siRNAs.

• Molecular Dynamics (MD) Simulations: Coarse-grained (CG) MD is used to model the initial stage of LNP assembly. It captures lipids breaking into inter-shearing layers or turbulent eddies.

• Kinetic Monte Carlo (kMC) Simulations: Used to model later stages of LNP growth (timescales larger than the mixing time $\tau_M$). It incorporates charge regulation, Derjaguin–Landau–Verwey–Overbeek (DLVO) interactions, and PEG-PEG repulsion.

• Machine Learning Analysis: A random-forest model is used to compute feature importance to identify which parameters (mixing rate, salt, PEG) most influence size and payload distribution.

Findings from Molecular Dynamics: The Origin of Empty LNPs

• MD simulations reveal that during the mixing process, lipids aggregate rapidly to form small nanoparticles before they have significant interaction with siRNA molecules.

• This early lipid aggregation is identified as the primary cause of empty LNP formation.

• Mixing Length Scale ($l$): Defined as half the lipid layer thickness in the simulation setup. Thicker inter-shearing layers (slower mixing) lead to a higher proportion of empty LNPs.

• Mixing Timescale ($\tau_M$): Calculated as $\tau_M = \frac{l^2}{2D}$, where $D$ is the inter-diffusion constant (approximated as $D \approx D_{siRNA}$).

• Result: siRNA distribution on LNPs is governed by $\tau_M$. Data for different length scales collapse into a master curve when analyzed at $\tau_M$

Kinetic Monte Carlo Results and Mixing Flow Rates

• kMC simulations model the long-term growth dynamics of LNPs starting from an initial radius $R_0$ calculated via mean-field coalescence theory.

• Mixing Flow Rate ($Q$): In turbulent mixing regimes ($Q \geq 10\,mL/min$), the characteristic mixing time is $\tau_M = b_Q/1.75 \times Q^{-1.5}$, where $b_Q = 1.3 \times 10^3\,ms$.

• Impact on Size: Particles generally reach an average radius of $\approx 20\,nm$ (cryo-TEM) or $\approx 50\,nm$ (DLS) at $18\,h$ regardless of initial $R_0$. DLS-derived average radii are typically $\approx 2.8$ times larger than cryo-TEM measurements due to skewing by large particles.

• Impact on Payload: Rapid mixing (higher $Q$) improves homogeneity. Slower flow rates result in more empty LNPs and a longer distribution tail (LNPs with excessive siRNA).

• Verification: kMC results show excellent agreement with CICS experimental data, confirming that bimodal distribution results from kinetics rather than thermodynamics.

Biological Impact: In Vitro Transfection Efficiency

• Transfection assays used LNPs at flow rates of $1$, $10$, and $30\,mL/min$ to knock down GFP expression in $GFP^+$ cells.

• High dose ($100\,ng$): All formulations achieved near-complete knockdown; distribution heterogeneity was negligible at saturation.

• Low dose ($10\,ng$, $25\,ng$): Particles from the $30\,mL/min$ flow rate significantly outperformed others.

• Key Correlation: Improved knockdown efficiency at low doses correlates with more uniform siRNA payload distribution and fewer empty LNPs. Encapsulation efficiency remained > 90\% across all conditions, so performance differences are attributed to population distribution rather than total siRNA content.

Effects of PEGylation and Salt Concentration

• PEG Molecular Weight (MW): Investigated in the range of $1000$ to $5000\,Da$. PEG forms a hydration shell Preventing aggregation.

  • Increasing PEG MW slows LNP growth and decreases final LNP size.

  • Increasing PEG MW monotonically increases the ratio of empty LNPs and broadens the payload distribution.

  • Flory Radius ($R_F$): $R_F = an^{3/5}$, where $a = 0.37\,nm$ and $n$ is the degree of polymerization.

• Salt Concentration ($c_{salt}$): Relevant range of $10$ to $150\,mM$.

  • Increased $c_{salt}$ accelerates aggregation by reducing electrostatic repulsion between cationic LNPs.

  • Faster aggregation at higher $c_{salt}$ reduces the proportion of empty LNPs.

  • Beyond $50\,mM$, kinetics plateau due to charge neutralization.

Analysis of Polydispersity and Heterogeneity Metrics

• Aggregation Kernels: The probability of coalescence is influenced by PEG and DLVO interactions.

  • PEG fusion barrier: $E_{PEG} = C_{PEG} R_1 R_2$.

  • DLVO fusion barrier: $W = C_{DLVO} \frac{R_1 R_2}{R_1 + R_2}$.

• Size-selective merging: The probability of merging a small LNP with a large one is higher than merging two identical LNPs, which promotes a narrower size distribution. PEG Steric repulsion is the primary factor limiting size polydispersity.

• Volumetric Scaling: The ratio of empty LNPs ($\theta_{empty}$) is determined by the initial fraction ($\theta_0$) and the number of merging events ($m$):

  • $\theta_{empty}(R) = (\theta_0)^m$, where $m = \left(\frac{R}{R_0}\right)^3$.

• Coefficient of Variation (CV): Standard deviation divided by the mean. Payload heterogeneity (CV) increases with the fraction of empty LNPs. Inefficient mixing results in greater variability even among the loaded (non-empty) particles.

Machine Learning Design Rules and Engineering Principles

• Random forest feature importance analysis conducted across parameters: $R_0$ ($6-10\,nm$), PEG MW ($1000-5000\,Da$), PEG ratio ($1\%-3\%$), and $c_{salt}$ ($10-150\,mM$).

• Size Control: PEG MW and PEG ratio are the most significant factors determining final LNP size. The mixing flow rate has negligible effect on final size.

• Payload Control: Mixing flow rate (initial LNP size) is the most critical factor controlling the ratio of empty LNPs and payload distribution.

• Decoupled Control: Practitioners can independently tune payload distribution (via kinetic measures like flow rate) and LNP size (via energy barrier measures like PEG modification).

Implementation and Methodological Details

• Formulation: DLin-MC3-DMA, DSPC, cholesterol, and DMG-PEG2000 in a molar ratio of $50:10:38.5:1.5$.

• Dialysis: At $1\,h$ post-mixing, samples are dialyzed against $1x$ PBS buffer ($pH\,7.4$) for $12\,h$ at $4^\circ C$ (MWCO $3500\,Da$).

• Fusion Rate Equation: $k_{ij}(R_i, R_j, \Phi_i, \Phi_j) = \frac{K_{collide}(R_i, R_j) e^{-\frac{E_b}{k_B T}}}{V_{system}}$, with $E_b = E_{PEG} + W$.

• Charge Regulation Theory:

  • Potential $\psi$ inside the LNP is approximated as constant using the Donnan approximation.

  • Charge density: $\rho = (1-f_w) [\alpha_{MC3}\rho_{MC3}\phi_{MC3} + \alpha_{RNA}\rho_{RNA}\phi_{RNA}] + 2f_w c_{salt} N_A \sinh\left(\frac{e_0 \psi_0}{k_B T}\right)$.

  • Surface potential: $\frac{e_0 \psi_0}{k_B T} = \frac{4\pi R^2 \rho l_D}{3(1 + \frac{R}{l_D})}$.

• Code Availability: The "FormLNP" computational framework is provided for researchers to predict LNP size and RNA payload distribution.

Discussion and Methodological Caveats

• Higher molecular weight cargos (e.g., > 1\,kb mRNA) diffuse more slowly. Experiments with mRNA ($\approx 1000$ nucleotides) follow qualitatively similar principles, though the kinetic modulation of the empty fraction is less pronounced ($12\%$ variation for mRNA vs $21\%$ for siRNA).

• Inter-laboratory variability may arise from differences in fluorophore handling, optical calibration, or dialysis kinetics.

• Cross-platform validation via analytical ultracentrifugation, nanoparticle confinement microscopy, or cryo-electron tomography is encouraged for high-resolution measurement standards.

• Future research will focus on in vivo toxicity profiling of formulations with controlled payload heterogeneity, as empty LNPs may modulate biological responses and exhibit different safety profiles.