Leveraging Biological Buffers for Efficient Messenger RNA Delivery via Lipid Nanoparticles

Abstract

Lipid nanoparticles containing messenger RNA (mRNA-LNPs) have launched to the forefront of non-viral delivery systems with their realized potential during the COVID-19 pandemic. Here, we investigate the impact of commonly used biological buffers on performance and durability of mRNA-LNPs. We tested the compatibility of three common buffers – HEPES, Tris, and phosphate-buffered saline - with a DLin-MC3-DMA mRNA-LNP formulation before and after a single controlled freeze-thaw (FT) cycle. We hypothesized that buffer composition would affect lipid-aqueous phase separation. Indeed, the buffers imposed structural changes in LNP morphology as indicated by electron microscopy, differential scanning calorimetry, and membrane fluidity assays. We employed in vitro and in vivo models to measure mRNA transfection and found that Tris or HEPES-buffered LNPs yielded better cryoprotection and transfection efficiency compared to PBS. Understanding the effects of various buffers on LNP morphology and efficacy provides valuable insights in maintaining the stability of LNPs after long-term storage.

Graphical Abstract

Introduction

Recently, lipid nanoparticles (LNPs) have drawn much attention as non-viral nucleic acid delivery platforms for the treatment of various genetic diseases, vaccines, and protein replacement therapies.^1,2 The global success of COVID-19 mRNA-LNP vaccines has truly highlighted the potential of lipid-based nucleic acid delivery.^2,3 From a scientific perspective, the design of novel lipid-mRNA therapeutics is divided into a three-part problem. First is the mRNA sequence design, which includes creating a codon-optimized sequence of desired protein, chemical modification of nucleotides, along with selection of amenable untranslated regulatory regions and poly-adenine tail length.^4,5 Second, the molecular design of the lipids should take in to account properties such as biocompatibility, pH response, solubility, natural abundance or ease of synthesis, and interaction with other components.^6,7 Last, once hydrophobic lipids and hydrophilic mRNA cross paths, a challenge arises to engineer a stable LNP loaded with mRNA. Optimization of LNP formulation depends on the mixing method, rate, solvents used, pH neutralization, and purification process.^8,9

Although LNPs are normally perceived as condensed hydrophobic matter, they can entrap a significant amount of water depending on composition^10–12 and nucleic acid cargo.^13,14 LNPs containing messenger RNA can consist of up to 30% water,^11,15 which suggests that a seemingly benign variable in LNP formulation – choice of neutralization or storage buffer – could have dramatic consequences on LNP formation, properties, and stability, especially since mRNA-LNPs solutions require subzero temperatures for long-term storage.¹⁶ Crystallization of buffering solutions upon freezing can severely affect the properties of biologics, and can induce LNP rupture and aggregation as captured in recent reports.^17–19 Additionally, a peculiar property of buffers upon freezing is the induction of pH gradients.^20,21 For example, commonly used phosphate-buffered saline (PBS) can experience pH changes of as much as 4 units upon freezing.²² These effects may be suppressed, although not fully eliminated by common sugar cryoprotectants such as mannose, sucrose, or trehalose.^17,18,22 Phospholipid headgroups also reportedly interact with buffering agents, causing lipid membrane softening.²³ Thus, LNPs, typically comprised of an ionizable lipid, a phospholipid, a sterol, and a PEG-conjugated lipid, may be highly susceptible to dynamic changes in pH and membrane elasticity as buffering agent, ionic strength, or temperature are changed. Interestingly, LNP-based COVID-19 vaccines use different sucrose-supplemented buffers - Pfizer/BioNTech’s Comirnaty uses PBS and Moderna’s Spikevax uses Tris, suggesting that LNPs may require different excipients depending on the lipid composition.

In this work, we investigated the physicochemical properties of the MC3 LNPs stored in three different buffers - phosphate-buffered saline (PBS), Tris-buffered saline (TBS), or HEPES-buffered saline (HBS). These buffers are commonly used in cell culture and molecular biology experiments, and therefore may be suitable as storage solutions for LNPs. We evaluated LNPs before and after 3 weeks of storage at −20°C in respective buffers. We also controlled the freeze rate of LNP solutions by using a commercially available freeze cube (rated for 1°C/min cooling rate), assuming that controlled freeze would be less disruptive to lipid membranes and would help to distinguish the effects of the buffers without accounting for potential variabilities of the freezing process. We have compared LNP size, encapsulation efficiency, particle morphology, lipid membrane hydration and thickness, and the thermal behavior of LNP solutions. Overall, the properties of LNPs were nearly identical before freezing but were significantly affected by the freeze-thaw cycle. In vitro experiments also identified the variable extent of endosomal escape and transfection efficiency. Lastly, in vivo transfection was significantly affected by the buffer composition at the early timepoint (4 hours post-injection), with diminishing influence at a later timepoint (24 hours). Our findings emphasize the need to consider the properties of aqueous solutions utilized in LNP preparation to enhance the transfection efficiency, improve long-term durability, and generally suit the unique molecular profiles of lipid formulations.

Results and Discussion

LNP preparation and characterization.

Lipid nanoparticles containing firefly luciferase (FLuc) mRNA were prepared via a standard microfluidic mixing procedure. DLin-MC3-DMA, DSPC, cholesterol, and DMG-PEG-2000 lipids were mixed at a 50:10:38.5:1.5 molar ratio, 3:1 volumetric ratio between mRNA solution in pH 4 citrate buffer and ethanol lipid mix, respectively, with a 9 ml/min total flow rate. The resulting LNP suspension was then separated into three equal parts and dialyzed against phosphate-buffered saline (PBS), Tris-buffered saline (TBS), or HEPES-buffered saline (HBS) with the respective buffer composition shown in Figure 1a. After dialysis, LNPs were concentrated and analyzed by dynamic light scattering (DLS) and a modified RiboGreen assay. DLS analysis revealed that the particles possessed nearly identical size of ca. 70 nm with minor variations in polydispersity index (PDI of ca. 0.05; Figure 1b). Zeta potential values suggested a slightly negative surface charge for PBS and TBS formulations (ca. −3mV, see Figure 1d) and neutral charge for HBS formulation. Encapsulation efficiency was >92% for all formulations (Figure 1c; summary of all properties is also included in Table S1). Cryo-TEM micrographs of these LNPs (Figure 1e) did not reveal any significant morphological changes either, suggesting that the buffers have a negligible effect on LNP formation, morphology, or short-term stability. The LNPs were then transferred to a commercially available controlled-freeze cube graded for 1°C/min and stored at a temperature of −20°C.

Figure 1. Characterization of lipid nanoparticles.

(A) Buffer composition, room temperature pKa, and pH range for buffers used in this study. (B) DLS data for LNPs dialyzed in respective buffers, before and after a controlled freeze-thaw (FT) cycle (n=6). (C) Encapsulation efficiency and mRNA concentration in firefly luciferase mRNA-loaded LNPs (n=3). (D) Zeta-potential measurements for LNPs dialyzed in respective buffers, before and after a FT cycle (n=6). (E) Representative cryo-TEM micrographs of the dialyzed LNPs before and after controlled FT. Scale bars represent 100 nm. Statistical analyses were performed using Tukey’s multiple comparison test; ns: no significance; *: p < 0.05; ****: p < 0.0001.

After three weeks at −20°C, LNPs in respective buffer solutions were thawed in the controlled freeze cube at room temperature for 30 minutes and immediately used for the studies. Intriguingly, the mean particle size after FT were ca. 50% larger for samples stored in HBS and PBS, while LNPs stored in TBS decreased in size by ~10 nm (Figure 1b). PDI remained consistent after thawing for HBS and TBS LNPs, while PBS LNPs produced a much more heterogenous population as indicated by the 3-fold increase in PDI (Figure 1b). This observation was unusual since we expected to see significant LNP aggregation after FT for all samples. Zeta potential also marginally decreased after freeze-thaw (Figure 1d). Most importantly, the LNP morphology was severely affected by the FT. Cryo-TEM confirmed LNP populations for PBS LNPs were most diverse in size and showed highly heterogenous internal organization, as expected for particle aggregation. Most of these structures are likely a result of the separation between lipid and aqueous phases as suggested by the difference in electron density. Specifically, PBS LNP sample contained: typical dense LNPs under 100 nm in size, empty liposome-like structures of 100–150 nm, and a wide range of LNPs carrying aqueous pockets. Some of the aqueous pockets contained mRNA, as potentially concluded from observing the relatively electron-dense mass inside the said pockets. TBS LNPs formed fused nanostructures indicative of particle aggregation but, unlike PBS LNPs, the aqueous pockets of lower electron density were much smaller and did not contain mRNA. Additionally, some TBS LNP were significantly smaller in size (<50 nm). Despite all the observed morphological changes, the size and PDI did not change significantly according to DLS analysis, suggesting that TBS may impede particle aggregation to a much greater extent compared to PBS. On the other hand, HBS LNPs fused into unique hourglass-like structures of slightly larger size compared to fresh LNPs (Figure 1e). The aqueous blebs formed at the periphery of the LNPs and appear devoid of mRNA, similar to other reports discussing phase separation in cryoTEM images.^19,24 This may suggest that the buffer can affect the packing of the lipid around mRNA, potentially due to the changes in mRNA tertiary structure.²⁴ For example, HBS may promote stronger electrostatic interaction between lipids and mRNA - zeta potential of HBS LNPs is truly neutral while PBS and TBS LNPs have a slight negative charge of −2 mV to −5 mV, which could suggest a more effective neutralization of mRNA charge when HBS is present. Despite these pronounced changes, encapsulation efficiency determined by the modified RiboGreen assay was virtually unchanged (Figure 1c). Some mRNA was lost over time either through degradation of mRNA or upon particle fusion, although the loss was less pronounced for LNPs stored in TBS (ca. 10% vs 25% for PBS) (Figure 1c). Therefore, we have compelling evidence that buffers can affect lipid organization.

With these results in mind, we investigated how these buffers may influence lipid packing by utilizing a Laurdan hydration assay. Koitabashi et. al. previously used the fluorescent Laurdan dye to identify the hydration of LNPs formulated with ionizable lipid, KC2, by measuring generalized polarization (GP).²⁵ In a biological setting, LNPs are taken up into the cells by endocytosis, gradually acidify during endosome maturation, and initiate the release of encapsulated nucleic acid cargo by the binding of positively charged ionizable lipid (such as MC3) to a negatively charged endosomal membrane.²⁶ As the ionizable lipid transitions to a protonated state, it attracts more water molecules and destabilizes the lipid membrane. Therefore, a high GP value is indicative of low membrane hydration or denser packing of the phospholipid bilayer.^25,27 Results of the Laurdan assay comparing fresh LNPs to thawed are shown in Figure 2a,b. Overall, LNPs were agnostic towards the buffer in this assay. Fresh LNPs were not responsive to the changes of surrounding pH and maintained roughly the same value of GP (~0.6; Figure 2a). This was slightly unusual since we expected the GP to decrease at pH < 6 (pKa of MC3 LNPs is ca. 6.4)²⁸ but could indicate that the lipid membrane is devoid of ionizable lipid and the event of membrane perturbation upon acidification did not occur, or that the lipid membrane perturbations occur at the nanoscale which would be undetectable by the Laurdan assay.²⁷ On the other hand, FT LNPs were responsive to the environmental pH and yielded lower GP implying looser lipid packing at physiological pH (Figure 2b). At pH 7.4, the GP values were ~0.45 for all buffers while pH 4.5 caused GP to drop even further to ~0.2, suggesting significant membrane hydration. We then investigated the changes in the lipid membrane thickness based off the measurements from cryoTEM micrographs. On average, lipid membrane thickness increased by 15% (see Figure S2) upon FT, suggesting increased membrane hydration in agreement with the Laurdan assay. More specifically, lipid membrane thickness increased by ca. 0.8 nm for TBS LNPs, ca. 0.6 nm for PBS, and ca. 0.2 nm for HBS LNPs after freezing. Although it is not entirely clear whether these findings indicate changes in membrane composition or simply are the outcome of water incorporation, these results highlight the pronounced changes in lipid packing of thawed LNPs.

Figure 2. Lipid membrane hydration and thermal analysis.

Laurdan hydration assay of fresh LNPs before (A) and after (B) freeze-thaw (FT), where generalized polarization (GP) is indicative of changes in lipid packing (n=3). (C) Representative differential scanning calorimetry (DSC) thermograms for LNP solutions in respective buffers. The acquisition was repeated for two cycles; the cycles were reproducible. Replicate cycles are included in Figure S2. Arrows indicate the direction along the temperature axis.

Differential scanning calorimetry (DSC).

We employed DSC as another approach to investigate the properties of LNPs stored in respective buffers. DSC measurements were conducted over the range of −20 to +20°C to investigate the potential phase transitions of both the buffer and the LNPs. The ramping rate was set to 1°C/min to match the cooling rate of the freezing device. In this experiment, LNPs contained scramble mRNA of similar molecular weight to FLuc (~1.9 kb). The particle size, PDI, and encapsulation efficiency were the same as the FLuc particles described earlier. Each sample went through two cycles for both cooling and heating to assess reproducibility. Since buffering components may affect the solvation and the H-bonding of LNP components with excipients, we were expecting to see some differences in both the peak temperature and the enthalpy of fusion. Results are shown in Figure 2c and S2.

DSC thermograms produced a similar pattern for all the tested buffers. The cooling curves had no pronounced peaks but instead produced broad exothermic humps towards the end of each cooling cycle. This pattern is associated with the eutectic crystallization of sodium chloride from buffer solutions.²⁹ The heating curves, on the other hand, had well-defined endothermic peaks, likely associated with the melting of the buffering solutions since the peaks were reproducible over two cycles. The average enthalpies of fusion for the peaks were: 290.4 J/g for HBS, with peak temperature of −0.39°C; 294.4 J/g for PBS, with peak temperature of −0.40°C; 296.1 J/g for TBS, with peak temperature of −0.81°C. As expected, the freezing points were slightly depressed, and the enthalpies were lower than that of pure water (334 J/g) due to the presence of additional solutes. We were not able to resolve any additional peaks that could be associated with the lipid transitions in the tested temperature range. Lastly, we performed a scan in a much higher temperature range (0–100°C) to compare the thermal behavior of PBS LNPs in solutions to dried powder (Figure S2d). Our previous study found that dried PBS LNPs only displayed one defined transition likely associated with melting at 38°C.¹⁰ PBS LNP scan in solution revealed a broad unresolved peak at approximately the same temperature and a pronounced peak at 85°C. These peaks likely do not correspond to the same phase transition event (e.g., formation of hexagonal phase); however, this hypothesis should be tested using a method that allows direct identification of material phase at elevated temperatures, such as small-angle X-ray scattering (SAXS). These data call attention to the different thermodynamic properties of the buffer solutions and to the critical importance of water in LNP formation and integrity.

In vitro assessment.

We then assessed LNPs prepared in different buffers for transfection ability and the extent of endosomal escape. Two cell lines – HeLa and HEK293T/17 (the latter modified with Gal9-GFP reporter system³⁰) – were treated with respective LNPs encapsulating firefly luciferase mRNA and assayed after 24 hours. Cell viability was maintained at >80% regardless of the buffer used (Figure 3a,c). Evaluation of transfection presented that the buffers changed the efficiency depending on the cell line. In HeLa, LNPs transfection trend was HBS~PBS>TBS (Figure 3b), while HEK293T/17 cells showed a strong preference for TBS>HBS>PBS (Figure 3d). Although the lipid membrane reorganization was not obvious from cryoEM or GP measurements, the in vitro trends may indirectly indicate additional differences. LNPs can experience superficial reorganization upon exposure to proteins,¹⁵ and specific protein binding ultimately controls the cellular uptake and transfection. Since the trends in transfection were apparent both in fresh and thawed samples, we hypothesize that the lipid membrane composition may be similar in each respective buffer group. Generally, all formulations experienced diminished transfection efficacy after thawing. In HeLa, the most drastic loss in transfection efficiency was seen for PBS LNPs, with an average 4-fold drop across all treatments, while HBS held an average of a 2-fold drop across all treatments (Figure 3b). For HEK293T/17, both HBS and PBS LNPs showed no statistically significant decline while TBS LNPs at 100 ng and 200 ng led to a 2 to 2.5-fold drop respectively (Figure 3d). None of the treatments indicated a dose response; we have observed such behavior before and ascribed it to an overwhelming of cellular machinery.^10,31 Additionally, we observed an interesting trend for TBS LNPs in HEK293T/17 dataset that almost indicates an inverse dose response. We speculate that this may be due to the morphological differences in TBS LNP population – freeze-thaw produced a significant amount of smaller LNPs (<50 nm) in TBS according to the cryoEM, which may not carry any mRNA. Since the rate of cellular uptake is higher for smaller particles,³² they would likely outcompete the larger, loaded LNPs as the dose increases, which could result in an apparent drop in transfection efficacy as the dose increases.

Figure 3. In vitro transfection.

(A, C) Viability of (A) HeLa (n=6) and (C) Gal9-GFP HEK293T/17 cells treated with LNPs for 24 hr (n=4). (B, D) Luciferase expression in (B) HeLa (n=6) and (C) Gal9-GFP HEK293T/17 cells transfected with LNPs for 24 hr (n=4). Luminescence was normalized by the fluorescence values to account for variability in cell growth. Statistical analyses were performed using Tukey’s multiple comparison test (n = 6,4 respectively); ns: no significance; *: p < 0.05; **: p < 0.01; ***: p < 0.001; ****: p < 0.0001.

We then proceeded to evaluate the extent of endosomal escape for FLuc mRNA LNPs by utilizing the Galectin 9 (Gal9)-GFP modified HEK293T/17 cell line. Galectins constitute a family of sugar-binding lectins associated with pathogen invasion and immune sensing.^33,34 Galectins such as Gal 3,8, and 9 are redistributed from diffuse cytosolic expression to exposed glycosidic sites resulting from endosomal inner-leaflet exposure events, indicative of endosomal disruption.³⁵ Gal9 has been shown to induce the highest signal-to-noise ratio in the detection of endosomal escape rupture in reporter cells and therefore it made a perfect biological sensor for our purposes.^35,36 With LNPs utilizing these pathways we aimed in quantifying endosomal escape by GFP puncta as described in our previous work.^30,37 Notably, the magnitude of changes was dose-dependent, consistent with our prior observations.³⁰ Low dose treatments (50 ng) had little changes among all fresh LNPs, although HBS LNPs showed the highest Gal9 recruitment and the most significant change (2.6-fold) after FT (Figures 4a and S3). The overall trend between fresh LNPs persisted at a high dose (200 ng; Figure 4a,b), with TBS LNPs inducing highest Gal9 recruitment. However, the buffered LNP responded quite differently after FT. PBS LNPs had reduced Gal9 recruitment; TBS LNPs were virtually unchanged, and HBS LNPs once again had increased Gal9 recruitment (Figure 4b). The increase in endosomal rupture events for thawed HBS LNPs could be a result of morphological change seen in CryoTEM (Figure 1e). From our previous work, changes in LNP morphology had resulted in a boost of in vitro transfection³¹, so HBS may have promoted an advantageous LNP morphological change after thawing. These in vitro data further support the importance of LNPs buffer composition for transfection efficiency after cold storage and reveal the differences in cellular uptake are influenced by such changes.

Figure 4. Gal9 recruitment in modified HEK293T/17 cells treated with LNPs.

(A) Representative confocal microscopy images for the Gal9-GFP recruitment at 24 hr after 200 ng treatment with LNPs containing firefly luciferase mRNA. Images presented in maximum intensity projection, Gal9-GFP (green) and DAPI (blue). Scale bars represent 50 μm. (B) Normalized puncta counts indicating Gal9 recruitment after LNP treatment for 24 hr. Statistical analyses were performed using Tukey’s multiple comparison test (n = 6); ns: no significance; **: p < 0.01; ***: p < 0.001; ****: p < 0.0001.

In vivo imaging.

Encouraged by the in vitro results for buffered LNPs, we explored the mRNA expression in vivo. HBS, TBS, and PBS LNPs were administered intravenously to age-matched, female BALB/c mice to compare the levels of luciferase expression and organ specificity. Bioluminescent signals were measured at 4-hour and 24-hour time points post-injection for both fresh and thawed LNPs (Figures 5 and S4).

Figure 5. In vivo bioluminescence imaging of animals after administration of fresh and freeze-thawed LNPs.

(A) Representative IVIS images of mice 4 hrs after LNP treatment, note the scale for respective panels; (B) Quantification of luminescence signals from IVIS images. Statistical analyses were performed using Tukey’s multiple comparison test (n = 3); ns: no significance; *: p < 0.05; ***: p < 0.001; ****: p < 0.0001.

In all cases, the bioluminescence images indicated a strong pattern of liver transfection, as generally expected for LNP formulations administered intravenously (Figure 5a). Therefore, despite the lipid reorganization, LNPs likely do not differ in the composition of biomolecular corona and still utilize the ApoE-mediated pathway.³⁸ Overall, in vivo transfection trends were similar to that observed in vitro with the HEK293T/17 cell line (TBS>HBS>PBS). Cross-comparison between fresh LNPs concluded that the TBS group showed a 2-fold increase at 4 hours post-administration compared to other buffers (Figure 5b). This boost in efficacy could be the result of improved ApoE-mediated uptake due to water displacement from the lipid membrane, as indicated by reduced lipid membrane thickness (Figure S1). However, the buffers resulted in different magnitudes of transfection loss. For FT groups 4 hours post-administration, HBS showed a total flux decrease by ca. 2-fold, TBS showed ca. 3-fold decrease, and PBS, the most common buffer for LNP formulation, caused a decrease in total flux by almost 9-fold. Interestingly, the difference between buffered samples was less pronounced at the 24-hour timepoint, with no statistically significant differences between samples (Figures 5b and S4). This is likely the result of diminishing buffer impact as the LNPs circulate in the bloodstream. Overall, consistent with previous observations,^17,18,39 mRNA delivery was affected by the freeze-thaw process. However, the dramatic changes with in vivo delivery depending on the buffer suggest that leveraging the biological buffers in LNP preparation may offer a significant advantage to retain mRNA-LNP efficacy.

Conclusion

LNPs have proven themselves as potent non-viral delivery platforms for a wide range of nucleic acid therapies and vaccines.¹ Their preparation typically involves some form of rapid mixing, such as microfluidic, T-junction, or pipet mixing,⁴⁰ followed by pH neutralization and additional purification.⁷ Frequently, LNPs are neutralized by dialysis or buffer exchange with a physiological buffer such as PBS, which allows for direct administration to living organisms after purification. However, the role of the physiological buffer composition on nanoparticle structure, cargo retention, and transfection efficiency has been largely unexplored to date. The buffers may offer an additional advantage as cryoprotectants, alleviating the concerns with cold storage of LNP formulations.¹⁶ Notably, the FDA-approved COVID-19 vaccines Spikevax and Comirnaty use different buffers (Tris and phosphate-buffered saline, respectively, supplemented with sucrose cryoprotectant), suggesting that each LNP formulation may have specific molecular requirements from the storage buffer. In our work, we have explored this auxiliary formulation parameter using a “standard” DLin-MC3-DMA formulation before and after −20°C storage for 3 weeks.

Our results indicate that LNPs dialyzed in HEPES and Tris-buffered saline (HBS and TBS, respectively) were overall superior to the phosphate-buffered saline (PBS), as indicated by in vitro and in vivo transfection. HBS and TBS may act as cryoprotectants, preserving the LNPs and conserving mRNA delivery to a larger extent compared to PBS. However, LNPs stored in HBS have shown unique hourglasss-haped structures after the FT cycle, while TBS and PBS both produced highly aggregated structures. Therefore, HBS may offer better protection against a sharp drop in pH gradient, prevent aggregation, and ultimately control the phase separation process. On the other hand, use of TBS was particularly beneficial for transfection, suggesting that the role of excipients in LNP formulations may be vastly underestimated. To the best of our knowledge, the only reported discussion about LNP excipients included cryoprotectants such as sucrose and trehalose for the longer-term storage of nucleic acid therapeutics.^17,18 While additional studies are required to understand the influence that pH gradients and storage conditions have on LNP properties and nucleic acid delivery, our findings provide a novel angle on the seemingly minor parameters of the LNP formulation process.

Materials & Methods

Materials

Ionizable lipid [(6Z,9Z,28Z,31Z)-heptatriaconta-6,9,28,31-tetraen-19-yl] 4-(dimethylamino) butanoate (Dlin-MC3-DMA) was custom-synthesized by BioFine International (Blaine, WA). 1,2-dimyristoyl-rac-glycero-3-methoxy polyethylene glycol-2000 (DMG-PEG-2000) and cholesterol were purchased from Sigma Aldrich (St. Louis, MO). 1,2-distearoyl-sn-glycero-3-phosphocholine (DSPC) was purchased from Avanti Polar Lipids, Inc. (Alabaster, AL). CleanCap Firefly Luciferase (L-7202) mRNA was purchased from Trilink Biotechnologies. Nanoparticle formulations were prepared using a Nanoassemblr Benchtop microfluidic mixer (Precision Nanosystems, Vancouver, BC). Characterization for size, polydispersity, and zeta-potential was performed on Malvern Zetasizer ZS (Malvern Panalytical Inc., Westborough MA). Grids for cryo-TEM imaging were purchased from Electron Microscopy Sciences (Hatfield, PA). Laurdan dye, Fisherbrand controlled freeze cubes, HEPES, Tris, sodium chloride, potassium chloride, phosphate buffer saline, ethanol, and Slide-A-Lyzer^™ G2 Dialysis Cassettes were purchased from Fisher Scientific (Federal Way, WA). For transfection assays, Quanti-iT RiboGreen^™ RNA reagent, CellTiter-Fluor^™ Cell Viability and ONE-Glo^™ Luciferase assays, were also purchased through Fisher Scientific (Federal Way, WA). Fluorescence measurements were acquired using a Tecan Infinite M200 Pro multimode microplate reader (Tecan Trading AG, Switzerland).

Animals

Albino BALB/c mice (Cat #000651) mice were purchase from The Jackson Laboratory (Bar Harbor, ME, USA). All animals and experiments were performed at Oregon Health & Sciences University and abided by the Institutional Animal Care and Use Committee (IACUC, IP00001707).

LNP Formulation and Characterization

Firefly Luciferase mRNA was diluted in sterile 50 mM sodium citrate buffer (pH 4) and molecular grade water to achieve desired volume. DLin-MC3-DMA, DSPC, Cholesterol, and DMG-PEG-2000 were dissolved in pure ethanol (molar ratios of 50:10:38.5:1.5) with a total lipid concentration of 5.5 mM. Lipid nanoparticles (LNPs) were prepared by standard microfluidic mixing at a 3:1 volumetric ratio (aqueous : ethanol) and 9 ml/min total flow rate. HEPES, tris, and phosphate buffer saline for dialysis were each prepared at a salinity concentration of 150 mM and pH 7.4. Upon formulation, LNPs were immediately transferred to 10 kDa MWCO cassettes and dialyzed in respective buffers for 4 hours and overnight at a 1000-fold volume. Following dialysis, LNPs were concentrated by centrifuging at 3000g in Amicon^® Ultra centrifugal filter units with a 100 kDa molecular-weight cut-off (Millipore Sigma, Burlington, MA). LNPs were characterized for hydrodynamic size, polydispersity index, and surface zeta potential using a Zetasizer Nano ZSP (Malvern Panalytical Inc., Westborough MA). For Zetasizer analysis, LNPs were diluted tenfold in respective buffers and triplicate measurements were obtained. To quantify mRNA encapsulation efficiency and concentration, a modified Quanti-iT RiboGreen^™ RNA (Invitrogen) protocol was used.

LNP Freeze-thaw (FT)

LNPs after characterization were placed into a controlled freezer cube (ThermoFisher; graded for 1°C/min freezing rate) and kept at −20°C. Upon retrieval from the freezer 3 weeks later, the samples were allowed to thaw for 30 minutes at room temperature in the cube.

Cryo-electron microscopy

Cryo-transmission electron microscopy (cryo-TEM) acquisition was operated at 200 kV voltage using FEI Titan Glacios (ThermoFisher, Hillsboro, OR) equipped with K3 camera. 2 uL of LNP sample were dispensed onto a plasma cleaned grid (lacey carbon, 300 Cu mesh), rested for 10 seconds at 100% relative humidity, blotted for 1 second in a Vitrobot Mark III blotter, and plunged into liquid ethane. The frozen grids were then assembled into cassettes under liquid nitrogen and imaged. Lipid membrane measurements were conducted manually for at least 10 particles per group.

Membrane fluidity (Laurdan) assay

A Laurdan hydration assay was performed as described previously.^25,27 Laurdan dye was diluted in dimethylformamide (DMF) to reach a working concentration of 100uM. Two unique buffers (citrate and phosphate) were titrated to achieve the desired pH and prewarmed to 37°C. Subsequently, respective LNP solutions were diluted to reach a final lipid concentration of 50 uM (based on quantified mRNA concentration and assuming a w/w lipid-mRNA ratio of 20). LNP samples were carefully mixed with assay buffers and dye in a black 96-well plate and incubated for 30 minutes in the absence of light. Generalized polarization (GP) enables the measurement of the difference in emission shift from a packed (gel) phase (440 nm) to a disordered (liquid) phase (490 nm). The GP was calculated as

$$GP=\left(I_{440}-I_{490}\right)/\left(I_{440}+I_{490}\right),$$

where I – intensity at respective wavelengths. Each assay contained three technical replicates.

Thermal analysis

Differential scanning calorimetry (DSC) was performed using a DSC 2500 calorimeter (TA Instruments) in −20°C to +20°C range at 1°C/min. 2 cycles of heating and cooling were acquired for each MC3 formulation containing scramble (non-encoding) mRNA at ~1.9 kb prepared in the identical fashion as described above. Approximately 3 mg of total lipid were used for the DSC scans. The data was processed using TA Universal Analysis software. The data was presented without background subtraction; no peaks were observed upon background subtraction; therefore, the peaks were attributed to the respective buffer composition and properties.

Cell culture and in vitro transfection

HeLa and HEK293T/17 cell lines (ATCC, Manassas, VA) were cultured in Dulbecco’s modified Eagle’s medium supplemented with 10% heat-inactivated FBS (Hyclone Laboratories Inc., Logan, UT) and 1% penicillin/streptomycin (Thermo Fisher, Federal Way, WA). All cultures were grown in 37°C incubators supplemented with 5% CO₂ and maintained following suppliers’ instructions. Cells were plated in white, clear-bottom 96-well plates at 4,000 cells per well and allowed to adhere overnight. Then, cells were administered LNPs encapsulating firefly luciferase mRNA. Cell viability results (CellTiter-Fluor^™, Promega), and luciferase expression data (ONE-Glo^™ Luciferase Assay, Promega) were collected 24 hours post-treatment with a microplate reader. Luminescent readout (in relative luminescence units, RLU) was normalized by cell counts commensurate with fluorescence (relative fluorescence units, RFU). RFU values were also compared between treated and untreated wells to determine cell viability.

Development of Gal9-GFP reporter cell line and visualization of Gal9 recruitment in 293T/17 cells

To establish stable transfections of Gal9-GFP reporter cell line, Gal9 was cloned into plasmid PB-CMV-MCS-EF1α-GFP PiggyBac cDNA vector (PB511B-1). The Piggybac Gal9-GFP vector codes for Gal-9-GFP fusion protein and was co-transfected with Super Piggybac Transposase expression vector from System Biosciences (Palo Alto, CA) into HEK293T/17 cells at a ratio of 3:1 using Lipofectamine 3000 (Thermo Fisher, Federal Way, WA). To establish stable transfectants, fluorescence-activated cell sorting (FACS) was carried out using BD FACSAria^™ Fusion equipped with a 488 nm laser and BD FACS Diva v8.0.1 software. GFP positive cells were obtained through cell sorting at two distinct times (day 2 post-transfection and day 8 post-transfection), selecting for indicative stable genomic integrations at AATT sites by bright green fluorescence. Once successful stable transfection was established, passage number of 10–25 of reporter cells were ensured for all further LNP transfections and cell uptake experiments.

Ibidi 8-well chamber slides (Fitchburg, WI) were coated with poly-D lysine (Thermo Fisher, Federal Way, WA) and rinsed three times with PBS prior to seeding HEK293T/17 Gal9-GFP reporter cells at 60,000 cells per well in complete media. After overnight incubation, respective buffered LNPs were added dropwise onto the media at a concentration of 50 ng or 200 ng mRNA per well and incubated for 24 hours. After incubation, media was gently aspirated, cells were then washed twice with PBS and fixed with 4% paraformaldehyde in PBS for 10 min at room temperature. Once cells were fixed, wells were gently washed with PBS two more times and DAPI (Thermo Fisher, Federal Way, WA) was then added at 1:1000 in PBS for nuclear staining. Following DAPI staining, cells were washed again two times, aspirated, and mounted with coverslips using ProLong Diamond Antifade mountant (Thermo Fisher, Federal Way, WA). After the mountant had set, a thin layer of nail polish was utilized to prevent the drying of the sample. Reporter cells were imaged for GFP-positive puncta with a confocal Leica DMi8 microscope (Leica Microsystems) with an oil immersion objective at 40x to report maximum intensity projections.

Gal9-GFP puncta were quantified using a custom macro written in ImageJ. Confocal images were preprocessed using maximum intensity projection and Gaussian smoothing with a radius of 1. Next, the puncta were counted using Find Maxima with optimal prominence (here equal to 75 for all images). To normalize puncta counts, nuclei counts were obtained using standard segmentation procedures (automatic thresholding and watershed) and Gal9 recruitment was reported as the number of GFP puncta divided by the number of cells.

In vivo injections and imaging

Female BALB/c mice aged 5–8 weeks received 1 ug of FLuc mRNA encapsulated inside LNPs intravenously via retroorbital injection. For bioluminescence imaging, mice received D-luciferin substrate (150 mg/kg) intraperitoneally and were imaged according to the manufacturer’s protocol. Image acquisition and analysis were achieved using the IVIS^® Lumina XRMS and the manufacturer’s software (PerkinElmer).