Introduction

During the COVID-19 pandemic the mRNA-lipid nanoparticle (LNP) technology proved to be the ideal framework for rapid development of vaccines to control the global health crisis. The mRNA-LNP drug products including Comirnaty and Spikevax1,2,3 demonstrated their efficacy and safety, with formulations being scaled up efficiently. The stability of mRNA has been substantially investigated, showing impact of lipid adducts4, buffers5, multivalent ions6, RNases7,8, temperature and pH9. LNP stability and structure can have influence on payload delivery and consequential vaccine efficacy10,11,12.

One of the aspects related to mRNA-LNP stability is the impact of stress forces. Recent publications investigated the impact of liquid–air interfaces introduced by mixing, shaking, flicking and tapping of mRNA-LNP drug products13,14,15,16,17. It was shown that the introduced liquid–air interfaces had an impact on mRNA-LNP drug product quality attributes16,17. This impact translated into significant changes in classically measured quality attributes like particle size measured by dynamic light scattering (DLS) and encapsulation efficiency measured by a ribogreen fluorescence assay, as well as antigen expression by in-vitro expression assays (Flow-Cytometry and LC/MS/MS as shown in reference18). Additionally, reference17 showed that the liquid–air interface induced changes to particle size measured by DLS and encapsulation efficiency are highly correlated. A hypothesis that was put forward is that PEGylated lipids, mainly residing on the outside of the mRNA-LNPs, are dissociated from the surface leading to reduced steric hindrance resulting in fusion of mRNA-LNPs, increasing the particle size and potentially releasing the payload. While the latter hypothesis is plausible, a deeper mechanistic understanding of the impact of liquid–air interfaces on LNPs is still absent.

This manuscript presents state-of-the-art heightened characterization results of mechanically stressed mRNA-LNPs, elucidating the mechanics of liquid–air interfaces impacting mRNA-LNP morphology. Using the method outlined in reference17, mRNA-LNP formulations, were mechanically stressed by shaking for fixed times allowing prolonged exposure of LNPs to liquid–air interfaces. Consequently, samples were characterized through nanoparticle tracking analysis (NTA) to obtain particle size distributions and concentrations, nuclear magnetic resonance (NMR) measurements to probe changes in mRNA-LNP surface properties18 and cryogenic transmission electron microscopy (cryo-EM) to allow for detailed visualization of the samples. The approach used to evaluate the impact of mechanical stress presented in this manuscript allows for a deeper understanding of the structural impact to mRNA-LNPs when exposed to liquid–air interfaces introduced during shaking stress. By integrating the experimental findings, we construct a comprehensive depiction of the morphological changes in mRNA-LNPs as shaking duration increases.

Results

Sample and study overview

The focus of this manuscript is on the mechanistic understanding of the impact of mechanical stress induced by shaking on formulations representative of commercially available mRNA-LNP Drug Product. Heightened characterization results presented in this manuscript are an extension to the results presented in Reference17.

The mRNA-LNP samples used in this study contain four different lipids as described in references17,19: ionizable lipid ALC-0315, PEGylated lipid ALC-0159, phospholipid and cholesterol. All samples were prepared by dispensing mRNA in a low pH aqueous phase and the four lipids in an organic phase. The aqueous and organic phase were mixed using a T-mixer with an N:P ratio of 6 (N:P ratio represents the molar ratio of amine groups (N) in the ionizable lipid to phosphate groups (P) in the RNA within the LNP). All samples were formulated using the same methods outlined in reference17 with a Tris-Sucrose final buffer composition. To enable a deep understanding of the impact of liquid–air interfaces induced during mechanical stress, samples were shaken at 100 upward movements per minute using a single platform laboratory shaker for 30 min and 240 min, and samples were kept frozen prior to analysis as outlined in reference17. These conditions were selected based on previous studies (see reference17) to include structure-altering stress to the samples.

Particle size and concentration by Nanoparticle Tracking Analysis (NTA)

The mRNA-LNP particle concentrations and size distributions were measured by nanoparticle tracking analysis (NTA). NTA utilizes the properties of light scattering and Brownian motion in a liquid suspension to characterize the lipid nanoparticles. Compared to dynamic light scattering (DLS) measurements, NTA measurements provide a deeper understanding of the particle size distribution and concentration without interference of size effects on scattering. In Fig. 1, NTA results are presented for the mRNA-LNPs as a function of the stress exposure time (shaking time) with numerical results presented in Table 1, showing mean particle size, span (calculated based on the distribution of particle size and an indication of sample heterogeneity), and particle concentration. The NTA profile of samples shaken for 30 min is comparable to that of unshaken mRNA-LNPs. After 240 min of shaking, the particles had significantly increased in particle sizes and heterogeneity, as reflected by the mean and span values respectively, indicating a drastic impact of the shaking. The increase in particle size and span aligns well with the previous observations made in references16 and17. In addition, particle concentration decreased upon shaking with a four-fold change at 240 min. The combination of the decrease in particle concentration and increase in particle sizes indicates a disruption of particles and their rearrangement into larger particles.

Fig. 1
Fig. 1
Full size image

Particle size distributions measured through NTA with no shaking (blue), after 30 min of shaking (red) and after 240 min of shaking (green). Each plot was normalized to its highest particle concentration. The samples were run in triplicate and the averaged results were reported.

Table 1 Mean particle sizes, span, and particle concentrations of mRNA-LNP samples with no shaking, after 30 min of shaking, and after 240 min.
Full size table

Morphology of mRNA-LNPs by Cryo-EM imaging

To visualize morphological changes of the LNPs upon shaking, cryo-EM imaging was performed on the samples listed in Table 1. Representative cryo-EM images without staining are shown on the left panels in Fig. 2 for the unshaken, 30 min and 240 min shaken samples. Additionally, thionine staining was performed, as discussed in reference14, to make RNA more visible by increasing contrast in cryo-EM images. The results of stained mRNA-LNP samples are shown in the right panel of Fig. 2.

Fig. 2
Fig. 2
Full size image

Cryo-EM images of the no shaking (a and b), 30 min shaking (c and d) and 240 min shaking (e and f) samples. Left panels show unstained images. Right panels show thionine-stained images to enhance mRNA contrast. Colored arrows denote the following: Blue—electron dense, spherical mRNA-LNP, Red—bleb-like mRNA-LNP, Magenta—liposome-like mRNA-LNP, Green—encapsulated RNA, Cyan—unencapsulated RNA. Scale bars represent 100 nm.

The images of unstained samples showed no significant differences between the control and 30 min shaking. Most mRNA-LNPs are spherical and either electron dense or electron dense with a bleb like portion. This observation is consistent with the NTA measurements which only showed no significant differences after 30 min of shaking. In contrast, significant changes were observed after 240 min of shaking. The images of unstained samples revealed a population of much larger sized spherical mRNA-LNPs either being electron dense or liposome-like, which is consistent with the NTA measurements. No direct observations of possible unencapsulated mRNA could be made from the three images of unstained samples shown in the left panel of Fig. 2.

To enhance the contrast of mRNA present, thionine staining was performed (see right panels in Fig. 2). The images of the control and samples shaken for 30 min show no indications of unencapsulated mRNA and are similar to the unstained ones; however, the contrast of mRNA present in blebs is enhanced. Similar to the unstained samples, stained samples of mRNA-LNPs shaken for 240 min showed significant changes compared to the control and mRNA-LNPs shaken for 30 min. The staining shows the emergence of a significant fiber-like network present outside the LNPs. Since thionine enhances the contrast of mRNA, this fiber-like structure likely consists of unencapsulated mRNA exposed due to the impact of the liquid–air interfaces and the consequential structural rearrangement of the mRNA-LNPs.

To further confirm the presence of unencapsulated mRNA, additional experiments were performed using RNase A on the mRNA-LNPs shaken for 240 min and are presented in Fig. 3. Another image of the sample shaken for 240 min is shown in Fig. 3a, again showing the fiber-like structure as seen in the corresponding image in Fig. 2. Upon being treated with 1 ng/mL RNase A before staining and imaging, the fiber-like structure has completely disappeared supporting the hypothesis that these fiber-like structures are unencapsulated mRNA digested by RNase A prior to staining (Fig. 3b). As further confirmation, samples were pre-treated with 2 µL RNase Inhibitor and consequently treated with RNase A and stained (Fig. 3c). The fiber-like structure remains as the inhibitor blocks RNase A digestion. All the observations presented in Fig. 3 confirm that a large portion of the mRNA is liberated from the mRNA-LNPs interior after prolonged shaking and this can be clearly visualized by appropriate staining. From the images, it cannot be discriminated against whether the mRNA is truly free in solution or adsorbed on the surface of the mRNA-LNP. The location of RNA could merely be due to the cryo-EM grid prep process (grid charge, vitrification, etc.).

Fig. 3
Fig. 3
Full size image

Experiments to confirm the presence of unencapsulated mRNA in the 240 min shaking samples. (a) image of a stained sample before RNase A treatment (similar to the lower-right image in Fig. 2). (b) image of a stained sample after RNase A treatment. The fiber-structure completely disappears after RNase A treatment. (c) Image of a stained sample pre-treated with RNase inhibitor followed by RNase A treatment whereby the fiber-like structure remains. Scale bars represent 100 nm.

The surface structure of mRNA-LNPs by NMR

To investigate surface structure changes in mRNA-LNPs upon mechanical stress, a diffusion NMR experiment utilizing pulse gradient stimulated echo (PGSTE) with bipolar gradients was conducted. Wang et al.19 demonstrated that this method effectively suppresses excipient signals from the formulation buffer while maintaining sensitivity to signals originating from the mRNA-LNPs19. Figure 4 presents the NMR spectra of the mRNA-LNP samples listed in Table 1. The spectrum of control mRNA-LNPs included a prominent PEG-methylene signal at 3.71 ppm from the PEGylated lipid (ALC-0159), along with peaks attributed to the ionizable lipid (ALC-0315) as well as lipid methyl and methylene groups. These peaks are consistent with the previous report19 and annotated accordingly. Multiple new peaks appeared in the spectrum of the sample shaken for 30 min compared to the control sample, and the peaks became more pronounced after 240 min of shaking, as indicated by asterisks in Fig. 4a. These peaks correlate with sucrose, an excipient in the formulation buffer, as confirmed by a side-by-side comparison to a sucrose standard (Fig. S1 in the supplementary material). The diffusion NMR experiment should have completely excluded sucrose signals in the buffer; hence, it was hypothesized that sucrose must be encapsulated within the mRNA-LNPs upon shaking and, therefore, detected by the diffusion experiment. To confirm this hypothesis, the samples shaken for 240 min were buffer exchanged into tris(hydroxymethyl)aminomethane (Tris) buffer to eliminate sucrose from the original formulation. The NMR spectra of the buffer-exchanged sample exhibited similar LNP to sucrose peak area ratios (Table S1 in the supplementary material) to those in the sample not subjected to buffer exchange (Fig. S2 in the supplementary material), confirming that the sucrose signals observed in the experiment are not from the buffer and sucrose is encapsulated within mRNA-LNPs upon shaking.

Fig. 4
Fig. 4
Full size image

(a) NMR spectra for the control and shaken mRNA-LNP samples. Additional peaks from sucrose are indicated by asterisk symbols. The signal at approximately 4.78 ppm is a method artifact from water. (b) same as figure (a) with a focus on the region showing the decreasing ALC-0159 peaks and increasing ALC-0315 peaks.

Additional notable changes are also observed by NMR upon shaking. Figure 4b illustrates the comparison between the control sample and the sample shaken for 240 min, focusing on the regions related to the PEGylated lipid (ALC-0159) and the ionizable lipid (ALC-0315). The prominent PEG-methylene signal at 3.71 ppm, along with minor signals at 3.25 ppm for –CON–CH2– and 3.39 ppm for –CH3–O– associated with the PEGylated lipid (ALC-0159) decreased significantly, indicating a reduction in PEG mobility. Conversely, the amine signal at 2.41 ppm, HO–CH2– signal at 3.50 ppm, and –CH2–OCO– at 4.03 ppm of ionizable lipid (ALC-0315) increased, suggesting enhanced mobility. Table 2 contains the percent change of these lipid peaks in the 30 and 240 min NMR spectra compared to the control sample. These findings are consistent with the mechanism proposed in reference17, which suggests that liquid–air interfaces introduced by shaking can remove PEGylated lipids from the mRNA-LNP surface, thereby reducing lipid mobility and corresponding NMR signals. This hypothesis also aligns with the increased mobility and NMR signal observed for ALC-0315. The removal of the PEGylated lipid layer likely induces structural rearrangement of the mRNA-LNPs, leading to increased surface exposure of ALC-0315. An additional experiment was conducted on an mRNA-LNP formulation without sucrose (Fig. S4 in the supplementary material). The resulting NMR spectrum also demonstrated the same lipid peak changes upon air entrapment.

Table 2 Percent change in area of ALC-0159 and ALC-0135 peaks in the control sample compared to shaken samples at 30 and 240 min.
Full size table

Discussion

This study presented findings regarding the structural and morphological changes of mRNA-LNPs when subjected to mechanical stress induced by shaking. The key observations and their implications revolve around how liquid–air interfaces, created through shaking, impact the structural integrity and particle size distribution of mRNA-LNPs. Initial observations indicate that short-term exposure to the air–liquid interface (30 min) has a minor impact, while prolonged exposure (240 min) results in notable changes. A breakdown of the proposed dynamics that occur during air–liquid exposure for the mRNA-LNPs is shown in Fig. 5. This includes an increase in mean particle size and a broader particle size distribution as well as a reduction in particle concentration, as evidenced by nanoparticle tracking analysis (NTA). Cryogenic transmission electron microscopy (cryo-EM) provides visual confirmation of the morphological changes observed through NTA. The drastic increase in particle size after 240 min of exposure suggests significant particle merging and structural changes (Fig. 5), potentially compromising the uniformity and stability of mRNA-LNP formulations. Furthermore, thionine staining revealed the presence of unencapsulated mRNA in the 240 min shaking samples (Fig. 5, step 2), confirmed by additional experiments using RNase A and RNase inhibitor. These findings indicate that the encapsulation stability of mRNA within LNPs is compromised under prolonged extreme mechanical stress.

Fig. 5
Fig. 5
Full size image

Proposed degradation pathway of mRNA-LNP dynamics upon shaking: (1) air–liquid interfaces lead to structural changes of the mRNA-LNP surface decreasing the NMR PEGylated-signal (ALC-0159). Potential removal of PEGylated lipid by liquid air-interfaces is shown. (2) The mechanical stress leads to the presence of unencapsulated mRNA and incorporation of sucrose into the mRNA-LNPs interior. (3) The structural instability of the mRNA-LNPs is reflected by an increased particle size observed by NTA and CryoEM. The cationic lipid is the ionizable cationic lipid ALC-0315.

Nuclear magnetic resonance (NMR) experiments provided deeper insights into the surface structural changes of mRNA-LNPs s under prolonged exposure to air–liquid interfaces. The study revealed a decrease in signal for PEGylated lipids (ALC-0159) and an increase for ionizable lipids (ALC-0315), supporting a structural re-arrangement (Fig. 5, step 3). The emergence of additional peaks, corresponding to sucrose, suggests that mechanical stress allows sucrose to be incorporated in the newly formed mRNA-LNPs as they reach an equilibrium state (Fig. 5, step 2). These morphological changes could affect the protective barrier function of the PEGylated lipids (Fig. 5, step 1), rendering mRNA-LNPs more susceptible to aggregation and destabilization. While it is shown that sucrose gets incorporated within the LNPs, it is not expected to impact mRNA stability directly. To provide further evidence of the mechanism described above, the same shaking experiment was performed on mRNA-LNPs at a concentration of 0.06 mg/mL (Fig. S5 and Tables S2 and S3 in the supplementary material). The experiment gave similar results to the 0.12 mg/mL mRNA-LNPs supporting the observations made.

The advanced analytical techniques and findings presented in this study provide experimental evidence for the proposed degradation pathway of LNP dynamics upon prolonged air–liquid interface exposure as, illustrated in Fig. 5 and previously proposed in reference17. This work presents mechanistic insights into the morphological transformations of LNPs subjected to intense mechanical stress. Although the prolonged shaking conditions exceed those typically encountered during routine handling or shipment of lipid-based LNP formulations, the study further enhances the understanding of the impact of liquid–air interfaces induced during shaking and contributes to the orthogonal characterization of LNPs across formulation and processing stages, ultimately supporting the optimization of mRNA-LNP therapeutics. Nevertheless, the conditions presented in this manuscript could be applicable to real-world conditions in case of extensive atypical manual manipulation (excessive flicking, dropping or manual shaking) of product and/or severe deviations from controlled and validated shipments, especially in the presence of large headspace containing presentations. Although, two shaking conditions were evaluated in the manuscript the effects described in the manuscript are expected to be immediate and linear as supported by the results discussed in Reference17.

While this manuscript provides insights into the mechanics of mechanical stress induced by shaking, it’s main focus is a qualitative analysis using dedicated heightened characterization tools. The manuscript focuses on commercially representative formulations. For future work, it could be interesting to study the impact of different lipids, different N:P ratios, and final buffer compositions. In addition, further advances in NMR and cryo-EM instrumentation and experimentation could allow for more detailed insights into the intermediate steps 1 and 2 of the mechanism presented in Fig. 5.

Materials and methods

mRNA-LNP formulations

mRNA-LNP formulations were prepared by dispensing mRNA in a low pH aqueous phase and 4 different lipids in an organic phase. Both phases were mixed using a T-mixer with an N:P ratio of 6. All mRNA-LNP drug products within this manuscript were prepared using the same formulation parameters (mRNA and lipid concentrations, flow rates and ratios). For this reason, the mRNA concentration is directly related to the LNP and lipid concentration.

For all mRNA-LNP formulations the following lipids are used at fixed lipid to mRNA ratios, for 0.12 mg/mL mRNA being: (i) 1.72 mg/mL ALC-0315 (((4-hydroxybutyl)azanediyl)bis(hexane-6,1-diyl)bis(2-hexyldecanoate)), (ii) 0.22 mg/mL ALC-0159 (2-[(polyethylene glycol)-2000]-N,N-ditetradecylacetamide), (iii) 0.37 mg/mL DSPC (1,2-distearoyl-sn-glycero-3-phosphocholine), and (iv) 0.74 mg/mL cholesterol. The 0.06 mg/mL mRNA formulations were obtained by diluting 0.12 mg/mL mRNA drug product with final buffer. For all mRNA-LNP formulations the final buffer composition is 10 mM TRIS and 300 mM sucrose. Unless otherwise stated, a formulation of 0.12 mg/mL mRNA content was used. The mRNA used was approximately 2000 nt long. All samples were formulated using a small-scale set-up representative for commercial production.

Shaking experiments

Shaking experiments were performed using a single platform laboratory shaker (Model 55 12 × 16 from Reliable Scientific, MS, USA) with a vial holder attached to the shaking plate to put vials in a vertical position on the shaker. The shaking speed for the laboratory shaker was set at the maximum speed of 100 upward movements per minute, tilting up to 20° for the vertical positioned vials. Shaking experiments were performed at room temperature. All vials, also including ‘no shaking’ and ‘30 min shaking’, experienced the same room temperature exposure time of 240 min, after which they were frozen at −80 °C awaiting analysis. By freezing the samples, uncontrolled shaking stress during shipment is avoided. SCHOTT 2 mL vials were used with a fill volume of 0.68 mL allowing sufficient headspace.

A figure of the shaking set-up is shown in Fig. S3 of the supplementary material.

NTA measurements

To prepare for NTA, the mRNA-LNPs were diluted to 80,000X with 0.1 µm filtered PBS to achieve a final particle concentration of approximately 1 × 108 particles/mL. NTA experiments were performed using a Nanosight NS300 (Malvern Panalytical, MA, USA) instrument. The instrument was primed with filtered PBS until no air bubbles were present in the flow cell and then the mRNA-LNP sample was loaded in a 1 mL syringe and attached to the flow cell adapter. The samples were run in triplicate with 60 s capture time using the 532 nm laser. The triplicate results were averaged and reported.

NMR measurements

The mRNA-LNP samples for NMR were diluted tenfold in 0.2 × Dulbecco’s Phosphate-Buffered Saline (DPBS) with 10% D2O containing 0.05 wt.% 3-(trimethylsilyl) propionic-2,2,3,3-d4 acid, sodium salt (TSP). LNP NMR spectra were recorded at 25 °C (298 K) with a 5 mm proton-optimized triple resonance inverse (TCI) cryoprobe using a Bruker NEO 800 MHz spectrometer (Bruker BioSpin, MA, USA). A 1D 1H 90° with water suppression using excitation sculpting (zgesgp) pulse sequence was employed for chemical shift calibration of TSP at 0 ppm. For characterization of the surface structure of the LNP, a 1D 1H pulse field gradient stimulated echo (PGSTE, stebpesgp1s1d) experiment was performed. The diffusion time was 60 ms and the duration of gradients was 2 ms with the Z gradients applied at 56 G/cm. The relaxation delay and acquisition time were 2.5 s and 1.3 s, respectively.

Cryo-EM imaging

To prepare for Cryo-EM imaging, lacey carbon-coated 300 mesh gold TEM grids (Electron Microscopy Sciences) were glow discharged at 20 mA for 60 s with the PELCO easiGLOWTM glow discharge system (Ted Pella). The LNP samples (4 µL) were applied in triplicate directly onto the glow-discharged TEM grids. The grids were then plunged frozen in liquid ethane using a Vitrobot Mark IV system (Thermo Fisher Scientific, MA, USA), with the chamber conditions set to 5 °C and 100% humidity. The prepared Cryo-EM grids were stored at liquid nitrogen temperature and imaged using a Glacios-2 transmission electron microscope equipped with a Falcon4i camera (Thermo Fisher Scientific MA, USA). The electron accelerating voltage was set to 200 kV. Cryo-EM micrographs were captured at a nominal magnification of 57,000x. For mRNA thionine staining, 10 µL mRNA-LNP sample was applied on the carbon side of glow discharged lacey carbon-coated 300 mesh gold TEM grids for 2 min. The grid was then flipped on top of a 100 µL 0.1 mM thionine droplet for 10 min After incubation, the grid was blotted with filter paper to remove excess solution, then the grid was loaded to Vitrobot Mark IV. The Vitrobot was set to 5 °C and 100% humidity, with one wash step with ultra-pure H2O and was then plunged frozen in liquid ethane. The grid was then stored at liquid nitrogen temperature for imaging. To confirm the mRNA presence outside of the LNP, the LNPs were treated with 1 ng/mL RNase A (Thermo Fisher Scientific # EN0531) prior to the mRNA thionine staining. To further confirm the effect of RNase A, the LNP was pretreated with 2 U of RNase inhibitor (Thermo Fisher Scientific #AM2694) after which the RNase A treatment and mRNA thionine staining were performed.

Statistical information

For NTA, the average and standard deviation of 3 runs were reported in Table 1. The NMR spectra and Cryo-EM images form a qualitative and descriptive part of the study and consequently no formal statistical analyses were performed. Nevertheless, NMR and Cryo-EM results remained consistent when evaluating two different formulation concentrations.