Impact of mixing on mRNA-LNP characteristics
Mixing occurs at multiple stages of the mRNA-LNP manufacturing process, mainly to ensure a homogeneous temperature and mRNA concentration during formulation, and consequently good drug product quality. Therefore, the impact of excessive mixing is investigated using a 1600 L commercial manufacturing vessel. A comparison is made between a low fill volume with low mixing speed combination and a high fill volume with high mixing speed combination. Details on the set-up are provided in Figure S1 of the supplementary material.
Impact of low volume mixing
In a first experiment, the 1600 L vessel is filled with a low volume (135 L) of homogeneous mRNA-LNP drug product, with the magnetic mixer minimally submerged, having the impellers directly below the liquid surface (see figure S1 in the supplementary material). The drug product bulk is actively cooled at 2–8°C during 10 days of continuous mixing at 130 RPM.
A drastic and consistent linear decrease in mRNA encapsulation is observed with increasing mixing time (see Fig. 1). Regardless of this drop in mRNA encapsulation, the mRNA integrity is not impacted during the 10 days of continuous mixing. The latter can be explained by enhanced mRNA stability against hydrolysis due to the active cooling at 2–8°C and the use of a sterile vessel, DP bulk and sampling manifold. Visual inspection during mixing shows considerable disturbances of the liquid surface, inducing air entrainment, due to the low depth of the mixer in the liquid volume (see Figure S2(a) in the supplementary material).
The fast decrease in mRNA encapsulation points towards a potential structural re-arrangement of the LNPs, which is further evaluated with NanoFlowSizer measurements. The particle size distribution at the start of mixing is unimodal with an average LNP size of approximately 60 nm (see Fig. 2). This unimodal distribution transforms into a bimodal distribution after one day of mixing, and further on completely shifts to a wider unimodal distribution after 4 days of mixing with an average particle size around 200 nm. The simultaneous linear drop in mRNA encapsulation and structural transformation in particle size distribution has not previously been described in the literature to the best of our knowledge.
As mentioned earlier, the combination of low fill volume and 130 RPM mixing speed develops considerable liquid surface disturbances, resulting in air entrainment. To investigate a possible link between this air entrainment and the observed transformation in particle size distribution, the experiment is repeated with a mixing speed of 40 RPM, almost completely removing the interaction with air as the liquid surface remains stable (no visible formation of a turbulence/vortex/foam) (see Figure S2(b) in the supplementary material). In this case, the particle size distribution remains unimodal during the complete 10 days of mixing, where small variations in distribution profiles can possibly be linked to the sampling and/or measurement method (see Figure S3 in the supplementary material). This result suggests a correlation between air entrainment and the observed changes in particle size distribution.
A straightforward visual inspection can be used as supplementary technique to assess the impact of mixing on the mRNA-LNP drug product. A large difference is observed between DP bulk mixed for 10 days at 130 and 40 RPM, with the 130 RPM sample being very opaque compared to the transparent 40 RPM sample (see Fig. 3 bottom left). This observation can be explained by the transfer to a larger sized particle distribution at 130 RPM (see Fig. 3 bottom right), shifting the dominant scattering regime from Rayleigh to Mie scattering, resulting in a more wavelength independent scattering. When inspecting the samples in front of a white light source, the difference can be observed by a change in light color, or change in wavelengths, passing through the sample without scattering (see Fig. 3 top right). The 40 RPM sample allows for passage of almost all the light, giving a white color, while the 130 RPM sample has increased scattering of smaller wavelengths, letting through only higher wavelengths, and therefore transmitting a yellow light. As an additional control the 40 RPM sample is vigorously shaken for one minute resulting in even more scattering, for increasing wavelengths, changing the color of passing light to red. Again, the crossover from transparent to opaque, or from white to red (using a white light source), is a visual indication for changes in particle size distribution, possibly caused by air entrainment. Supplementary to more sensitive dynamic light scattering techniques, this visual inspection can be used as a quick and efficient check for possible air entrainment during formulation or after sample handling of mRNA-LNP drug products.
Impact of high volume mixing
To further investigate the influence of liquid surface disturbances and corresponding air entrainment, the mixing experiment is repeated with a high fill volume of 741L drug product and an increased mixing speed of 198 RPM (see Figure S1 in supplementary material). In this case, the mRNA encapsulation drop during 10 days of mixing is much smaller compared to the low volume experiment at 130 RPM (see Fig. 4 and Fig. 1). The mRNA integrity is not affected during the complete mixing experiment as observed and explained earlier for the low volume mixing. The reduced drop in mRNA encapsulation at a higher mixing speed of 198 RPM, instead of 130 RPM, clearly shows negligible effect of mixing shear in this respect, and on the other hand highlights the impact of fill volume and corresponding air entrainment. Increasing the fill volume expands the distance between the mixer impellers and the liquid surface, minimizing disturbance of the liquid surface. This result suggests a possible link between this air entrainment and the observed mRNA encapsulation drop.
The particle size distribution at the start of mixing is unimodal with an average LNP size of approximately 60 nm (see Fig. 5). This unimodal distribution slowly broadens, tailing to larger LNP sizes, and gradually transforms into an emerging bimodal distribution at the end of 10 days of continuous mixing. As with the mRNA encapsulation drop, also for the transition of particle size distribution a similar evolution is observed as with the small fill volume experiment, but at a much slower pace. The DP characteristics of these high fill volume samples at 10 days of mixing do not reach the situation of the low fill volume samples at 1 day of mixing. These results once again indicate a correlation between air entrainment and the simultaneous drop in mRNA encapsulation and transformation in particle size distribution.
In addition, samples from each mixing time point are put on stability at room temperature showing no additional impact on mRNA encapsulation nor particle size distribution, but only an expected mRNA integrity decrease due to temperature accelerated mRNA hydrolysis (see Figure S4 and Figure S5 in supplementary material).
Impact of shaking on mRNA-LNP characteristics
In the previous section it is shown that by minimizing air entrainment during mixing, through increase of the liquid volume, the corresponding impact on mRNA encapsulation and particle size distribution is drastically reduced, even with an increased mixing speed. This indicates that the interaction with air acts as main stress force, causing an LNP size increase and mRNA encapsulation drop. Based on these findings shaking experiments are performed on liquid mRNA-LNP drug product in vials, to further elucidate the stress sensitivity of mRNA-LNP drug product and to check for possible relationships with headspace volume (or fill volume) and LNP content (or mRNA concentration). In these experiments all vials are shaken vertically in a controlled manner using a single platform laboratory shaker as shown in Figure S6 in the supplementary material and described in the materials and methods section.
As an initial experiment, an mRNA-LNP drug product presentation with a low fill volume of 0.48 mL in a 2 mL vial is used, containing 0.12 mg/mL of mRNA. When shaking vials with such a large headspace volume a similar evolution of particle size distribution is observed as with the large-scale low volume mixing experiment (see Fig. 2 and Fig. 6). The particle size distribution transitions during shaking from narrow unimodal at small LNP size, to tailing (after 10 min), over bimodal (after 30 min), to wide unimodal at larger LNP size (after 240 min).
In a more elaborate set of shaking experiments, the effect of headspace volume and LNP content on the stress sensitivity is further investigated. An mRNA-LNP formulation batch is split up in three different mRNA concentrations (0.12 mg/mL; 0.06 mg/mL; 0.01 mg/mL). Each of these concentrations is filled in 2 mL glass vials, one part with large headspace volume (0.48 mL fill volume) and one part with small headspace volume (2.25 mL fill volume).
While the behavior of LNP size (measured with Zetasizer) as function of shaking time is similar for the six drug product presentations, a large impact is seen for both fill volume and mRNA concentration (see Fig. 7). The vials with high fill volume and high mRNA concentration are more resistant against LNP size increase, compared to the ones with low fill volume and low mRNA concentration. At the highest mRNA concentration of 0.12 mg/mL, the highest impact of headspace volume is observed, while at the lower mRNA concentrations of 0.01 mg/mL and 0.06 mg/mL, the mRNA concentration becomes the main separator, with a drastic increase in stress sensitivity for the 0.01 mg/mL mRNA presentations.
Another interesting observation is the evolution of LNP size and PDI with increasing shaking time. LNP size strongly increases the first 2 hours of shaking, after which the rate of size increase starts to reduce, with a stabilization after 4 hours of shaking (see Fig. 7). PDI also strongly increases during the first 2 to 4 hours of shaking, after which it starts to drop (see Fig. 8). These trends are in line with the previous observations with the NanoFlowSizer where shaking effectuates a transition of the particle size distribution from narrow unimodal over bimodal towards a broader unimodal larger sized LNP population (see Fig. 6).
An additional experiment with other mRNA concentrations (0.01, 0.1 and 0.5 mg/mL), focusing on the first 2 hours of shaking, gives similar conclusions for headspace volume and LNP content impact (see Figure S7 and S8 in supplementary material). In addition, it is shown that for the low fill volume (0.48 mL), low mRNA concentration (0.01 mg/mL), presentation, the PDI already starts to drop again after 1 hour of shaking, indicating the higher stress sensitivity with faster transition towards a larger sized LNP population.
When looking at the behavior of mRNA encapsulation as function of shaking time, an identical impact of fill volume and mRNA concentration is observed as shown earlier for LNP size (see Fig. 9 and Fig. 7), which is also confirmed in the additional shaking experiment with other mRNA concentrations (see Figure S9 in supplementary material). The vials with high fill volume and high mRNA concentration are more resistant against an mRNA encapsulation drop, compared to the ones with low mRNA concentration and low fill volume. The mRNA encapsulation strongly decreases the first 2 hours of shaking, after which the rate of encapsulation drop starts to reduce, ending with a stabilization after 4 hours of shaking for most presentations, except the 0.48 mL fill combined with 0.06 and 0.12 mg/mL mRNA. The reason for these two divergent values can either be linked with the Ribogreen analysis method, giving possible underestimation when measuring outside the calibrated range, or can either be linked with a mechanistic aspect of the imposed stress. Either way, during each shaking experiment a negative correlation between LNP size and mRNA encapsulation is present (see Fig. 10), which moves from a linear to a slightly quadratic correlation at low mRNA encapsulation values as indicated above.
This direct negative correlation suggests that a certain fraction of the mRNA is removed from the LNP when the LNP size increases. Based on the observations described above a possible mechanism is proposed and shown in Fig. 11. Shaking with air entrainment causes air-liquid interfacial forces which can subtract parts of the PEGylated lipids on the outer surface of the LNP. The corresponding lowering of steric hindrance between neighboring LNPs causes merging of these unstable particles. During this merging process the lipid layer surrounding the mRNA gets temporarily interrupted allowing some of the mRNA strains to be removed from the LNP (either completely or partially). In the end, stable LNPs with larger size, but decreased mRNA payload, are formed.
The slower increase in LNP size and drop in mRNA encapsulation for presentations with high fill volume and high mRNA concentration can also be explained by this proposed mechanism. A high fill volume or low headspace volume corresponds with less air bubbles per liquid volume, and therefore less interaction between air bubbles and a certain amount of LNPs. Following the same logic, a high mRNA concentration or high LNP content (due to constant N:P ratio during formulation) corresponds with a higher amount of LNPs for a certain amount of air bubbles, giving less interaction between air bubbles and LNPs. Interestingly, a similar concentration dependent impact of air entrainment is observed for monoclonal antibodies (mAbs), where hydrophobic interactions increase between neighboring proteins [14, 15], further supporting this proposed mechanism for LNPs.
The observation of the flattening of the LNP size and mRNA encapsulation curves, in Figs. 7 and 9 respectively, could indicate that the removal of PEGylated lipids leads to a certain equilibrium, which is determined by both the lipid (and therefore mRNA) concentration and the forces of the air-liquid interfacial stress from the air entrapment (and therefore headspace volume).
The reproducibility of the shaking method was also assessed through execution of 3 separate runs, using 4 different mRNA-LNP drug product presentations (see Figure S10 and S11 in supplementary material). For a short shaking time of 30 minutes, the variation introduced by the shaking method appears negligible compared to the inherent variation of the respective measurement methods, with a standard deviation of LNP size of 2.09 nm (Zetasizer measurement), and a standard deviation of mRNA encapsulation of 0.98% (Ribogreen measurement). For this short shaking time the impact of mRNA concentration on the stress sensitivity can clearly and repeatedly be visualized. For longer shaking times, the 95% confidence intervals of 0.03 and 0.1 mg/mL mRNA (long strain length) overlap, indicating possible difficulty in assessing the impact of mRNA concentration due to increased variation of the shaking method or the measurement method at these increased LNP sizes. To see minor differences in stress resistance of different mRNA presentations, in a consistent way, it can therefore be useful to take along at least 2 shaking timepoints and to restrict the maximum shaking time while still allowing sufficient differentiation between each presentation.