A detailed protocol for synthesizing lipid nanoparticles (LNPs) using confined impinging jet (CIJ) mixer technologies, including a two-jet CIJ and a four-jet multi-inlet vortex mixer (µMIVM), is demonstrated. The CIJ mixers generate reproducible, turbulent micro-mixing environments, resulting in the production of monodisperse LNPs.
Lipid nanoparticles (LNPs) have demonstrated their enormous potential as therapeutic delivery vehicles, as evidenced by the approval and global usage of two COVID-19 messenger RNA (mRNA) vaccines. On a small scale, LNPs are often made using microfluidics; however, the limitations of these devices preclude their use on a large scale. The COVID-19 vaccines are manufactured in large quantities using confined impinging jet (CIJ) turbulent mixers. CIJ technology enables production at a laboratory scale with the confidence that it can be scaled to production volumes. The key concepts in CIJ mixing are that the mixing length and time scale are determined by the turbulence intensity in the mixing cavity and that the nanoparticle formation occurs away from walls, eliminating the problem of deposition on surfaces and fouling. This work demonstrates the process of making LNPs using confined impinging jet mixer technology with two geometries: the two-jet CIJ and the four-jet multi-inlet vortex mixer (MIVM). The advantages and disadvantages of each mixing geometry are discussed. In these geometries, LNPs are formed by rapid mixing of an organic solvent stream (usually ethanol containing the ionizable lipids, co-lipids, and stabilizing PEG-lipids) with an aqueous anti-solvent stream (aqueous buffer containing RNA or DNA). The operating parameters for the CIJ and MIVM mixers are presented to prepare reproducible LNPs with controlled size, zeta potential, stability, and transfection effectiveness. The differences between LNPs made with poor mixing (pipetting solutions) compared to CIJ mixing are also presented.
mRNA-based therapeutics hold great potential for the treatment and prevention of a wide range of diseases, including infectious diseases, genetic disorders, and cancers1. Unlike small-molecule therapeutics, which can passively diffuse across the cell membrane, nucleic acids must be encapsulated for intracellular delivery2. Encapsulation provides both structure and stability to mRNA, facilitating their intracellular delivery via endocytic pathways as well as preventing degradation from intra- and extra-cellular components such as nucleases3. A host of materials and nanocarriers have been developed for the encapsulation and delivery of mRNA, including inorganic nanoparticles, polymers, lipids, and lipid-like materials1. Among these, LNPs have emerged as the most prominent delivery platform for mRNA-based therapeutics4.
LNPs are composed of four lipid components: ionizable lipid, cholesterol, zwitterionic lipid, and PEG-lipid stabilizer5. Ionizable lipids suitable for mRNA delivery exhibit a careful balance between the lipid hydrophobicity and the dissociation constant (pKa) of a ternary amine group6. The ionizable lipid pKa typically has a pH between 6.0 and 6.7, such as KC-2 (DLin-KC2-DMA), MC-3 (DLin-MC3-DMA), and ALC-03157. This pKa restriction on the ionizable lipid enables both the encapsulation of nucleic acid polymers as hydrophobic lipid salts and the intracellular delivery through an "endosome escape" process. LNPs enter a target cell through (various) endocytosis pathways that all involve acidification of the endosome from pH 7.4 to pH ~58. The ionizable lipid pKa ensures that LNPs have nearly neutral surfaces under physiological conditions but become cationic in an acidifying endosome9. This pH response enables selective disruption of only the endosomal membrane, the release of the encapsulated nucleic acid polymer, and preserves cell viability, unlike permanently cationic lipids used in transfection systems such as Lipofectamine. Cholesterol is a hydrophobic, interstitial molecule in the LNP structure that improves lipid fluidity. The zwitterionic lipid plays a structural role and forms a bilayer on the LNP surface. The poly(ethylene glycol)-lipid (PEG-lipid) is a colloidal stabilizer that enhances the LNP stability by imparting a polymeric steric stabilizer onto the LNP surface, which resists aggregation of LNPs. This stabilizes the LNP, particularly during changes in pH that regenerate the free base form of the ionizable lipid which behaves like a hydrophobic oil. The Onpattro (patisiran) recipe (hereafter, referred to as LNP formulation) is often used as a starting point for LNP formulation with ionizable lipid MC3, cholesterol, distearoylphosphatidylcholine (DSPC), and PEG2000-DMG dissolved in ethanol mixed against an aqueous solution of RNA10.
Several techniques can be used to manufacture LNPs encapsulating nucleic acid polymers, with most of them relying on a common theme of rapidly mixing an ethanol stream containing lipids with an aqueous stream incorporating the nucleic acid of interest (siRNA, mRNA, or DNA)9,11,12,13,14. In this regard, bulk mixing processes such as pipette mixing and vortex mixing offer a simple strategy to form LNPs that eliminates the need for using sophisticated instruments12. However, bulk mixing does not provide a homogeneous distribution of components, leading to a sub-optimal LNP size distribution along with significant batch-to-batch variability15.
Laboratories routinely use microfluidic mixing techniques to obtain reproducible LNPs by achieving more precise control over mixing conditions12,13,16. Yet, the laminar flow conditions in microfluidic devices, which are inherent due to the small length scales and low velocities in a microfluidics chamber, result in comparatively slow solvent/antisolvent mixing17. The small chamber dimensions severely limit the throughput and scalability needed for GMP production of LNPs, but researchers have parallelized microfluidic chambers to attempt to scale production volumes15. A parallelized microfluidics geometry does not eliminate the problem of lipid adsorption to surfaces during large volume processing, a problem commonly referred to as "fouling" of the mixing device, and there are problems with uniformity and stability of flows that make microfluidics scaleup challenging for industrial-scale production18,19. It is not surprising that pharmaceutical companies used turbulent impinging jet mixers to manufacture COVID-19 vaccination mRNA-LNPs20.
The production process of RNA-loaded LNPs entails the blending of an aqueous buffer stream containing the RNA payload with an ethanol stream containing the four distinct lipid components. These formulations utilize an acidic buffer with a pH of 4.0 or less, which charges the ionizable lipid as the aqueous and ethanolic streams mix. The positively charged ionizable lipids interact electrostatically with the negatively charged RNAs, forming a hydrophobic RNA-lipid salt. Hydrophobic lipid species, including the RNA-lipid salt, precipitate in the mixed solvents and form hydrophobic nuclei. These nuclei grow through the precipitation of zwitterionic lipid and cholesterol until reaching a critical point where sufficient pegylated lipid adsorbs on the surface of the LNPs, halting further growth -nucleation and growth mechanism21,22,23. The addition of aqueous buffer to the lipid solution, to the extent where lipids precipitate and LNPs form, hinges on two distinct time scales: the solvent-antisolvent mixing period, τmix, and the nuclei growth period, τagg. The dimensionless Damköhler number, defined as Da = τmix/τagg, captures the interplay between these time scales24. In instances of slow mixing (Da > 1), the final size of LNPs is transport-controlled and varies with mixing time. Conversely, during fast mixing (Da < 1), the fluid is fragmented into Kolmogrov-length striations or layers, whereby LNP formation is solely governed by the molecular diffusion of each constituent, resulting in homogeneous kinetics of LNP formation. Achieving the latter scenario demands that the lipid concentration exceeds a critical threshold, establishing a state of supersaturation conducive to uniform homogeneous nucleation.
It is estimated that τagg ranges from a few tens to a few hundreds of milliseconds25. In its most basic configuration, the two streams, one containing ethanol with lipids and the other containing an aqueous buffer with RNA cargo, are injected into a chamber known as a "confined impinging jet" (CIJ) mixer. Turbulent vortices produce solvent/antisolvent striation length scales of 1 µm within 1.5 ms when operated at appropriate velocities. The stream velocities and mixing geometry determine the conversion of linear momentum into turbulent vortices that mix the streams. This is parameterized by the dimensionless number, the Reynolds number (Re), that is linearly proportional to the flow velocities. Re is computed from Re = Σ (ViDi/vi), where Vi is the flow velocity in each steam, vi is the kinematic viscosity of each stream, and Di is the stream inlet diameter in 2-jet CIJ devices26 or the chamber diameter in 4-jet MIVMs27. Note: Some references for the CIJ use only a single jet diameter and velocity to define Re28. Re is in the range of 1-100 in a microfluidics device, whereas in the CIJ devices, Re of 125,000 can be achieved. In a CIJ mixer, streams with equal momentum collide, dissipating their momentum upon impact as turbulent mixing, which leads to efficient micromixing due to the small Kolmogorov microscales and small Damköhler number. Another type of mixer is the "multiple inlet vortex mixer" (MIVM), where four streams are directed into a central chamber. In this setup, continuous flows into the confined mixing chamber ensure a well-defined mixing time scale. All fluid elements pass through the high-energy mixing zone in both types of mixers. In contrast, simple mixing devices like T-junctions do not contain a chamber that provides a mixing zone, resulting in less mixing of the two streams due to the incoming stream momentum being largely deflected into the outlet direction rather than into turbulent vortex generation. Both CIJ and MIVM mixers can be operated in batch or continuous modes, offering flexibility for LNP production at various scales.
This protocol describes how optimal LNP formulations are made by employing two confined impinging jet technologies: 2-jet CIJ and the 4-jet MIVM mixers. The operation of CIJ and MIVM mixers has been previously demonstrated for the preparation of NPs with hydrophobic core materials29. That article and video should be consulted as an additional resource on the formation of NPs with these mixers. This update focuses on lipid-based NP formation. The ability to tune the size of LNPs by varying the micro-mixing conditions is demonstrated. Additionally, the utility of CIJ technologies in forming stable, monodisperse LNPs with improved in vitro transfection efficiencies in HeLa cells when compared to LNPs made using poor pipette mixing is shown. Furthermore, the advantages and disadvantages of each CIJ mixing geometry, along with appropriate conditions needed for the scaleup of these mixers, are discussed.
Synthesis of LNPs containing nucleic acid polymers using two confined impinging jet turbulent mixers has been presented. When conducted at appropriate velocities, CIJ turbulent mixers ensure the time scale of mixing is shorter than the LNP assembly time, producing homogenous supersaturation conditions for forming small LNPs with narrow size distributions21. Consequently, LNPs made with the same chemistry using different turbulent mixer geometries (the 2-jet CIJ and the 4-jet MIVM mixer) exhibit similar physicochemical properties and show good transfection efficiencies (Figure 5 and Figure 6). In contrast, LNPs made using pipetting that produces poorer mixing results in larger and more polydisperse LNPs (Figure 5A) with lower transfection efficiencies. It has long been understood that mixing and assembly kinetics play a significant role in LNP processing; Cullis et al. noted that rapid convective-diffusive mixing of ethanol and buffer leads to the formation of small particles with a narrow size distribution, whereas slow diffusive mixing leads to larger particles with broad size distributions9. The time scale of mixing in CIJ turbulent mixers decreases proportionately to the inlet flow rates of the streams to the mixer27. This is quantified by the dimensionless Reynolds number (Re), which measures the ratio between the inertial and viscous forces. The turbulence inside the mixing chambers of the CIJ and MIVM occurs at sufficiently high Re, such that the turbulent vortex stretching results in small scales of length that produce rapid solvent/antisolvent mixing by diffusion. The turbulent length scale depends on the Re and not the specific geometry of the mixing device. That is why either the CIJ or the MIVM makes the same LNP particles, and why various sizes of MIVM mixers make the same NP sizes27. At high Re, corresponding to high inlet velocities, LNPs can be made reproducibly without batch-to-batch variations (Figure 3B).
This protocol enables the formulation of a variety of mRNA, DNA, or siRNA LNPs with different physicochemical properties using turbulent CIJ mixers. In addition to allowing versatility in composition and concentrations, this technique provides a clear path to rapidly screen formulations at bench sale (a few milligrams) and scale up the lead formulations to larger industrial batch sizes at production rates of 5 L/min36. This has been a major hurdle for several other techniques, including bulk-mixing and microfluidics. For instance, bulk-processing techniques fail to consistently manufacture LNPs reproducibly, even at a few milliliter scales. Microfluidic techniques provide a significant improvement over bulk-mixing techniques to enable the production of uniform and reproducible LNPs; however, they are only in the milligram range29. As detailed in the introduction, parallelization of microfluidic devices provides an attempt at scaling to production scales but does not eliminate the issue of fouling, and it cannot be scaled as successfully as mixers based on confined impinging jet technology.
Apart from these advantages, CIJ mixers will be instrumental in manufacturing next-generation LNPs that exhibit targeting capabilities or perform gene editing. The current LNP formulations have lipids and nucleic acids that have similar diffusivities, and therefore, they can be made even with slightly poor mixing at bench-scale. However, gene-editing approaches may require the encapsulation of nucleic acid species with widely different molecular weights, such as small guide RNA molecules and large mRNA transcripts, to code a CAS9 protein37. The very different diffusion time scales of these different species make uniform encapsulation at stoichiometric ratios challenging. This problem of uniform encapsulation becomes more pronounced as the mixing efficiency becomes poorer. Likewise, targeting non-hepatic cells might need the incorporation of strongly bound slow-diffusing stabilizers (such as large molecular weight block copolymers with targeting ligands). Targeting ligands as large as 14 kDa can be conjugated to block copolymers prior to nanoparticle assembly, which enables their uniform incorporation into NPs using CIJ mixing38. CIJ turbulent mixers are useful tools for manufacturing LNPs made with components having different diffusivities.
While CIJ turbulent mixers demonstrate several advantages over other mixers for formulating LNPs, it is important to note the limitations associated with each geometry. The 2-jet CIJ mixer requires that both the inlet streams (ethanol and water) have equal momenta (within 10%-30%) to achieve uniform turbulent micromixing in the chamber. The fact that the exit stream comprises 50:50 solvent/antisolvent limits the level of supersaturation in the mixing cavity where precipitation occurs29. This drawback is addressed by the 4-jet MIVM mixer, as it can utilize four jets with unequal momenta to accomplish high supersaturation conditions in the mixing chamber. Additionally, both the mixers are required to be in the order of milligrams of total mass, making them a non-ideal choice for high-throughput screening of many different LNP formulations. For simple LNP formulations, screening may best be done with microfluidics or pipetting strategies at microgram scales and then transferred to the confined impinging jet technology when a few lead formulations have been identified. It is also crucial to consider the dead volumes in the mixers. In the CIJ, two jet mixers, the hold-up volumes are 50-100 microliters. This amount of material must be subtracted from the amount captured in the quench bath when calculating the recovery from the process. These losses are insignificant when operating on large scales but would account for 10% losses when total volumes of 5 mL are produced, as shown here. The impinging jet turbulent mixers are a valuable tool for producing LNPs at the GMP scale, as evidenced by the two FDA-approved COVID-19 vaccines.
The authors have nothing to disclose.
NSF Fellowship to BKW (DGA1148900), support from Tessera Therapeutics Inc., the Bill and Melinda Gates Foundation (BMGF, contract numbers OPP1150755 and INV-041182), and the FDA under award 75F40122C00186.
18:0 PC (DSPC) | Avanti Polar Lipids | 850365P | Helper lipid |
21 G x 1-1/2 in. BD PrecisionGlide Needle | BD | 305167 | |
96 Well Black Wall Black Bottom Plate | Fisher Scientific | 07-000-135 | |
96 Well White/Clear Bottom Plate, TC Surface | Thermo Fisher Scientific | 165306 | |
Acetic Acid, Glacial | Fisher Scientific | A38-212 | |
ALC-0315 | Avanti Polar Lipids | 890900 | Ionizable lipid |
Amicon Ultra Centrifugal Filter, 100 kDa MWCO, 15 mL | Millipore Sigma | UFC910024 | |
Amicon Ultra Centrifugal Filter, 100 kDa MWCO, 4 mL | Millipore Sigma | UFC810096 | |
Bright-Glo Luciferase Assay System | Promega | E2620 | |
Cholesterol | Millipore Sigma | C8667 | |
CleanCap FLuc mRNA (5 moU) | Trilink Biotechnologies | L-7202 | |
Confined Impinging Jets Mixer | Holland Applied Technologies, Helix Biotech, Diamond Tool and Die (DTD) | N/A | Contact Holland or DTD for custom orders and the Helix Biotech system is Nova BT. Review text for new mixer validation |
D-Lin-MC3-DMA | MedChemExpress | HY-112251 | Ionizable lipid |
DMEM, high glucose, pyruvate | Thermo Fisher Scientific | 11995065 | |
DMG-PEG 2000 | Avanti Polar Lipids | 880151P | PEG-lipid |
DODMA | Avanti Polar Lipids | 890899P | Ionizable lipid |
Ethanol 200 Proof | Decon Labs, Inc. | 2701 | |
Falcon 15 mL Conical Centrifuge Tubes | Fisher Scientific | 14-959-70C | |
Falcon 50 mL High Clarity Conical Centrifuge Tubes | Fisher Scientific | 14-959-49A | |
Fetal Bovine Serum, certified, United States | Thermo Fisher Scientific | 16000044 | |
HeLa | ATCC | CCL-2 | |
HEPES, free acid | IBI Scientific | IB01130 | |
HSW HENKE-JECT two-part 1 mL Luer | Henke Sass Wolf | 4010.200V0 | |
HSW HENKE-JECT two-part 5 mL (6 mL) Luer Lock | Henke Sass Wolf | 4050.X00V0 | |
Idex 1648 ETFE tubing ” OD 0.093” ID | Idex Health & Science | 1648 | |
Idex P-678 ¼”-28 to Luer fitting | Idex Health & Science | P-678 | |
Idex P-940 ferrule for ETFE tubing | Idex Health & Science | P-940 | |
Lipofectamine 3000 Transfection Reagent | Thermo Fisher Scientific | L3000001 | |
Luer fitting | Idex Health & Science | P-604 | Assemble on CIJ or MIVM mixer inlet with corresponding threads. Idex parts are also available through VWR and many other suppliers |
Mixer stand | Holland Applied Technologies | N/A | See Markwalter & Prud'homme for design.26 Contact Holland for Purchase |
Multi-Inlet Vortex Mixer | Holland Applied Technologies and Diamond Tool and Die (DTD) | N/A | Contact Holland or DTD for custom orders. Review text for new mixer validation |
O-ring (MIVM) | C.E. Conover | MM1.5 35.50 V75 | Order bulk – consumable part. Ensure solvent compatibility if using an alternative source. |
Outlet ferrule – CIJ | Idex Health & Science | P-200 | Assemble with outlet fitting (large end flush with tubing) |
Outlet fitting – CIJ | Idex Health & Science | P-205 | Assemble with ferrule and tubing on CIJ chamber outlet |
Outlet fitting – MIVM | Idex Health & Science | P-942 | Combination with ferrule |
Outlet tubing – CIJ | Idex Health & Science | 1517 | Use a tubing cutter for clean ends. Ensure extra tubing doesn't protrude into mixing chamber |
Outlet tubing – MIVM | N/A | N/A | Fit to ferrule ID. |
PBS – Phosphate-Buffered Saline (10x) pH 7.4, RNase-free | Thermo Fisher Scientific | AM9624 | |
Penicillin-Streptomycin (10,000 U/mL) | Thermo Fisher Scientific | 15140122 | |
PHD 2000 Programmable Syringe Pump | Harvard Apparatus | N/A | |
Plastic two-piece syringe 1 mL | Thermo Fisher Scientific | S7510-1 | |
Plug fitting | Idex Health & Science | P-309 | Assemble on CIJ mixer sides (seal access point from drilling) |
Quant-it RiboGreen RNA Assay Kit and RiboGreen RNA Reagent, RediPlate 96 RiboGreen RNA Quantitation Kit | Invitrogen by Thermo Fisher Scientific | R11491 | |
Resazurin, Sodium Salt | Thermo Fisher Scientific | R12204 | |
RNase AWAY Surface Decontaminant | Thermo Fisher Scientific | 7000TS1 | |
Scintillation vial | DWK Lifesciences | 74504-20 | |
SGE Gas Tight Syringes, Luer Lock, 100 mL | SGE | 100MR-LL-GT | |
SGE Gas Tight Syringes, Luer Lock, 50 mL | SGE | 50MR-LL-GT | |
Slide-A-Lyzer Dialysis Cassettes, 20 K MWCO | Thermo Fisher Scientific | 66012 | |
Sodium Acetate | Millipore Sigma | 32319-500G-R | |
Sodium Hydroxide | Fisher Scientific | S320-500 | |
Sucrose | Millipore Sigma | S7903-1KG | |
Syringe Filters, Sterile | Genesse Scientific | 25-243 | |
Triton X-100 | Millipore Sigma | 9036-19-5 | |
Trypsin-EDTA (0.25%), phenol red | Thermo Fisher Scientific | 25200056 | |
Water, Endotoxin Free | Quality Biological | 118-325-131 | RNAse and DNAse free |
Yeast RNA (10 mg/mL) | Thermo Fisher Scientific | AM7118 |
.