Insights into mRNA lipid nanoparticle polydispersity and shape of using quantitative solution biophysics

Introduction: Lipid nanoparticles (LNPs) are potent delivery vehicles that have accelerated the translation of RNA therapeutics and led to the FDA-approval of mRNA COVID-19 vaccines. However, one major barrier that has prevented researchers from unlocking the full potential of LNP-based therapies is the low sensitivity of current characterization techniques, which rely on bulk analysis of LNPs. These techniques cannot evaluate essential parameters, such as RNA loading distribution, particle morphology in ambient conditions, and the percentage of unloaded LNPs, which are critical features to monitor for pharmaceutical development. In this study, we combine standard characterization techniques with a multitude of solution-based biophysical tools, including sedimentation velocity analytical ultracentrifugation (SV-AUC), field flow fractionation-multi-angle light scattering (FFF-MALS), and size exclusion chromatography-small angle X-ray scattering (SEC-SAXS). We compare clinically relevant LNP formulations when prepared by either microfluidic rapid mixing or bulk mixing, which are two common methods for small-scale LNP production. With current characterization techniques, it is difficult to assess the quality of new formulation methods; however, these advanced characterization techniques were able to provide insight into how these preparation methods affects LNP structure and RNA loading. Additionally, by combining these techniques together, we demonstrate the effect of mRNA size on RNA loading distribution. Lastly, we perform structure-function analysis by correlating the biophysical results with transfection in human primary T cells. The characterization techniques employed here can enhance our understanding of LNP structure-property-function relationships and enable researchers to precisely define their RNA LNP products, which can improve LNP quality and potentially accelerate pharmaceutical development.

Materials and

Methods: Lipid excipients were obtained from Cayman Chemical and Avanti Polar Lipid. Firefly luciferase (FLuc) mRNA was purchased from TriLink BioTechnologies and contained 5-moU substitutions. The MC3, C12-200, SM-102, and ALC-0315 LNPs were formulated using their respective commercial recipes, either using our in-house microfluidic devices or by pipette mixing. The same LNPs were also prepared without mRNA (empty) using microfluidic devices only. The LNPs dialyzed in 20 kDa MWCO cassettes (Thermo Fisher Scientific) in 1X PBS for 2 h before being passed through either a 220 nm filter (microfluidic samples) or a 450 nm filter (pipette mixed samples). Traditional LNP characterization consisted of size and concentration via dynamic light scattering (DLS; Wyatt DynaPro Plate Reader III), ζ-potential (Malvern Zetasizer Ultra), encapsulation efficiency and loading (Thermo Fisher Scientific Quant-itTM RiboGreen Assay), relative pKA (6-(p-toluidino)-2-naphthalenesulfonic assay, Sigma-Aldrich), and cryogenic transmission electron microscopy (cryo-TEM, UPenn Beckman Center for Cryo-Electron Microscopy). SV-AUC was conducted with an XL-A analytical ultracentrifuge (Beckman-Coulter) using a TiAn60 rotor. FFF-MALS was operated utilizing a 1260 Infinity II HPLC (Agilent) with an Eclipse (Wyatt) system in-line with a VMD UV detector (Agilent), Optilab dRI detector (Wyatt), and DAWN MALS (Wyatt). SEC-SAXS was collected at the National Synchrotron Light Source II using the 16-ID (LiX) beamline with a Sepharose 4B sizing column.

Results, Conclusions, and Discussions: The traditional characterization methods revealed that the bulk LNPs are generally larger and have lower encapsulation than the microfluidically mixed LNPs, but the two formulation methods result in LNPs with similar ζ-potential and relative pKa values (Fig. 1A-D). Within the different LNPs, only marginal differences were observed, showcasing the poor resolution of these techniques among the LNP formulations. This is supported by the cryo-TEM images, which show wide polydispersity in both shape and size within each LNP sample (Fig. 1E). Utilizing SV-AUC, the LNPs are also revealed to have extraordinary polydispersity in RNA loading, with bulk mixed particles having a greater degree of loading polydispersity than the microfluidically mixed LNPs (Fig. 1F). These results correlate with other reports that estimate that over 50% of LNPs are unloaded. FFF-MALS provides a higher resolution snapshot of LNPs than DLS as the hydrodynamic radius, geometric radius, molecular weight and polydispersity (Fig. 1G), and RNA and lipid content can be obtained with a single run. To probe the possible sub-species as well as the inner dynamics of LNPs, we employed SEC-SAXS. Here, scattering information is obtained as a function of time, and by undergoing single value decomposition, we identified subspecies within each individual LNP formulation (Fig. 1H-K). Moreover, these species have unique RNA-lipid interactions, both in their intensity and interaction length (q = 0.10-0.20). The relative intensity profiles between microfluidic and bulk mixed formulations are completely predictive of human primary T cell mRNA transfection, where the LNP with the largest peak facilitates greater luminescence (Fig. 1L). Lastly, via DENSS, an ab initio electron reconstruction program, we discovered that LNPs are non-spherical and adopt an ellipsoidal shape (Fig. 1M). By combining solution-based biophysical analytical tools, the rules governing LNP formation and efficacy can begin to be untangled, allowing for the rationale design of next-generation particles.

Acknowledgements (Optional):