Toward a large-batch manufacturing process for silicon-stabilized lipid nanoparticles: A highly customizable RNA delivery platform

Abstract

While lipid nanoparticles (LNPs) are a key enabling technology for RNA-based therapeutics, some outstanding challenges hinder their wider clinical translation and use, particularly in terms of RNA stability and limited shelf life. In response to these limitations, we developed silicon-stabilized hybrid lipid nanoparticles (sshLNPs) as a next-generation nanocarrier with improved physical and temperature stability, as well as the highly advantageous capacity for "post-hoc loading" of RNA. Nevertheless, previously reported sshLNP formulations were produced using lipid thin film hydration, making scale-up impractical. To realize the potential of this emerging delivery platform, a manufacturing process enabling multikilogram batch sizes was required for successful clinical translation and deployment at scale. This was achieved by developing a revised protocol based on solvent injection mixing and incorporating other process adjustments to enable in-flow extrusion of multiliter volumes, while ensuring sshLNPs with the desired characteristics. Optimized procedures for nanoparticle formation, extrusion, and tangential flow filtration (to remove residual organic solvent) currently enable production of 2 kg finished batches. Importantly, sshLNPs produced via the modified large-scale workflow show equivalent physical and functional properties to those derived from the earlier small-scale methods, paving the way for GMP manufacturing protocols to enable vital translational clinical studies.

Keywords: Bio-Courier; RNA delivery; gene therapy; lipid nanoparticles; process development; silicon nanoparticles; sshLNP.

Graphical abstract

Figure 1

Schematic overview of conventional LNPs (left) and sshLNPs (right), highlighting key features of the latter that are responsible for their unique properties

Figure 2

Removing the first solvent evaporation step in production of sshLNPs (A) Workflow of the original lipid thin film hydration method (method 1) where organic solvent is evaporated from activated SiNPs, and the revised protocol (method 2) where the activated SiNP suspension is added directly to the aqueous phase. (B) More sedimentation was observed with method 2 (lower image) after lipid film hydration. (C) DLS results indicated lower PDI and higher zeta potential for sshLNPs produced using the revised method. Note: “DLS/Zeta” indicates points at which measurements were made of hydrodynamic size, polydispersity index (PDI), and zeta potential. Panel C report mean ± SD for samples analyzed in triplicates (n = 3).

Figure 3

Removing the second evaporation step and optimizing a solvent injection mixing method (A) Workflow of the initial direct injection mixing approach (method 3), including two in-process checks (IPCs). (B) Less sedimentation was seen compared with lipid thin film hydration. (C) Modified workflows incorporating an additional 0.8 μm extrusion step and other adjustments as indicated. (D) The additional extrusion step (method 4) led to smaller sshLNPs with similar PDI and zeta potential. (E) Prefiltration before extrusion (method 6, right) removed insoluble aggregates that would otherwise accumulate on the 0.8 μm extrusion membrane (method 4, left), although the 0.4 μm membrane appeared to remove additional aggregated material. (F) Introducing the prefiltration step did not impact the properties of the final sshLNPs. Graphs in panels B, D, and F report mean ± SD for Size, PDI and zeta potential for samples assessed in triplicates (n = 3).

Figure 4

Purification of sshLNPs by TFF (A) Arrangement of the experimental setup illustrating how a second pump (pump 2) was included to enable both ultrafiltration (UF) and diafiltration (DF) operations. (B) Schematic showing how the TFF cassette functions. (C) Representative result that illustrates successful removal of MeOH from sshLNPs produced via method 5, as judged by ¹H NMR. (D and E) The TFF process was found to slightly modify the DLS characteristics (D) and lipid content (E) of the particles, but these variations lie within relevant reference ranges and were therefore not considered problematic. Error bars within the graphs in panel D and E show mean values ± SD based on three independent measurements.

Figure 5

Initial large-scale (1 L) run to produce sshLNPs (A) Schematic of the workflow used; a scale-up of method 5. (B) Representative images of extrusion membranes and sample aliquots (lower-right quadrant), showing buildup of aggregates despite the use of larger membranes as well as some remaining insoluble material after the first extrusion step. (C) Accumulation of aggregates on the 0.8 and 0.4 μm membranes resulted in substantially higher operating pressure than on a 50 mL scale (method 4). (D) DLS characteristics of the sshLNPs were not adversely affected and were within reference ranges. Panel D error bars are represented as mean ± SD for Size, PDI and zeta potential for samples assessed in triplicates ( n = 3).

Figure 6

Production of a 2 L batch of sshLNPs (A) Schematic of the workflow used. Note that the TFF step contained two stages: UF to concentrate the sample from 4 to 2 L, followed by DF to remove MeOH and free lipids. (B) The extrusion steps were compromised by significant buildup of aggregates on the membranes, compromising their performance. (C) For the finished batch, average size and PDI were outside the reference ranges. (D) This was due to deterioration of the extrusion membranes with excessive accumulation of aggregates, indicating the need for a prefiltration step after activation of SiNPs when manufacturing sshLNPs at scale. Graphs in panel C and D were measured in triplicates (n = 3) with error bars represented as mean ± SD.

Figure 7

Prefiltering the activated SiNP suspension improves the process (A) Schematic of the modified workflow, designated method 7. (B) In this case, there was no significant accumulation of aggregates on the extrusion membranes. (C) This was reflected in considerably lower operating pressures during extrusion, compared with the protocol without prefiltration. (D) The modified procedure also led to sshLNPs with DLS properties that were well within reference ranges, both on a small scale and for a 1 L batch size. Panel D graphs report mean ± SD for samples analyzed in triplicates (n = 3).

Figure 8

Bio-Courier sshLNPs produced at small scale (using lipid thin film hydration, method 2) and large scale (using method 7) exhibited similar physical and functional properties (A) The large-scale process produced significantly smaller (p < 0.0001) final particles with comparable PDI. (B) Zeta potential was unaffected by scale-up. (C) Both methods gave sshLNPs with equivalent RNA encapsulation efficiency. (D) Large-scale manufacture also maintained transfection efficiency of HEK293 cells with an mRNA encoding firefly luciferase (fLuc). Luminescence was measured to determine fLuc expression levels at 24 h post-transfection. Lipofectamine 2000 was used as positive control and untreated cells as negative control. (A–C) Mean ± SD for samples analyzed in triplicate, while (D) shows mean ± SD for three independent biological replicates.