*3.2. DOE-Based Optimization of LNPs*

In the 1st DOE (screening A), 6 independent factors, including the molar percentage of CL (level: 40 to 60), PL (level: 10 to 40), PEG-lipid (level: 0.5 to 1.5), a type of CL (level: CL4H6 or CL7H6), PEG-lipid (level: PEG-DMG or PEG-DSG), and NaCl concentration (level: 0 to 400 mM), were systematically examined. Since TFR and FRR had only a limited impact on the size of LNPs in our previous study and have substantial issues (reduced reproducibility at lower levels and unintended dilution at higher levels, respectively), these parameters were fixed at 500 µL/min and 3, respectively. DSD was adopted in screening A to determine the significant contributing factor(s) for the responses (including physicochemical properties, cellular uptake, and gene expression) and to narrow down experimental space of the following 2nd DOE (screening B) (Table 1). The RNA-loaded LNPs were synthesized under 14 different formulation conditions (coded as A-1 to A-14) that were determined based on DSD. The diameter of the LNPs varied with a ζ-average from 88 to 754 nm (Table 1). The LNPs were typically uniform (PdI of 0.2 or less) except for A-4 and A-5. The encapsulation efficiency and ζ-potential of the LNPs were typically high (85% or over) and near neutral (within ± 5 mV), respectively, except for A-14 (72.4% encapsulation and −10.3 mV). The reproducibility of the synthesis of the LNPs between 2 technically independent experiments were confirmed (Figure S1). Statistical analysis revealed that an increase in NaCl concentration had the most significant effect on increasing the size of the LNP, an observation that is consistent with our previous study (Figure S2). A decrease in both %DOPE and %PEG also had a significant effect on the increase in size (Figure S2). This was also consistent with our previous study [39] and can be explained by the fact that the relatively bulky hydrophilic moieties of both DOPE and PEG-lipid resist the decrease in total surface area of the LNP with increasing particle size. Both CL4H6 and a higher %DOPE significantly contributed to a higher encapsulation efficiency (Figure S3). It is possible that an oxygen atom in the morpholino ring at the hydrophilic head of CL7H6 attracts electrons from the tertiary amine, thereby lowering the acid-dissociation constant and reducing the RNA encapsulation efficiency, especially under competitive conditions in the presence of NaCl. DOPE can facilitate encapsulation of RNA due to the fact that a primary amino group of DOPE, which is a proton donor, can form direct hydrogen bonds with phosphate groups of RNAs [30].

**Table 1.** Physicochemical properties of the LNPs examined in screening A.


Cellular uptake and Nluc activity in 3 types of splenic APCs and whole spleens for the 14 types of LNPs are summarized in Table S1. The reproducibility of Nluc activity in the whole spleen, cellular uptake and Nluc activity in splenic DCs were confirmed (Figure S4). Scatter plots revealed that the cellular uptake in the 3 types of splenic APCs was well correlated (Figure 2A,B), suggesting that factors and their levels examined in screening A have moderate effects on cell specificity between the 3 types of splenic APCs. On the other hand, the amount of cellular uptake differed significantly (1 to 2 orders of magnitude) among the LNPs, and the order was Mϕ > DCs >> B cells. Scatter plots of Nluc expression versus cellular uptake in DCs and Mϕ showed a trend toward higher activity in DCs (Figure 2C). An analysis of covariance was performed to reveal statistical differences in the Nluc expression level between DCs and Mϕ without the effect of cellular uptake level, and the findings revealed that Nluc expression in the DCs was significantly higher than that in Mϕ, which is consistent with observations obtained in a previous report [2]. The slope of Nluc activity to cellular uptake was obviously higher than 1 for DCs, suggesting that more LNPs are taken up per cell or LNPs with characteristics that allow them to be more easily taken up lead to higher transgene expression in DCs. The maximum Nluc expression in B cells was less than 100 RLU/3000 cells, which was over 100-fold lower than that in both DCs and Mϕ (Table S1). Scatter plots of Nluc expression in the whole spleen vs. splenic DCs showed a moderate correlation (R<sup>2</sup> = 0.3919) (Figure 2D). Although the transgene expression level in the whole spleen has been extensively measured in many studies on mRNA vaccine-oriented formulation development due to its simplicity [55–57], the data indicate that transgene expression level in the whole spleen would not be a good indicator of the corresponding process in DCs, even if the transgene expression level in DCs was high. This can be attributed to the fact that the population of splenic DCs is only ~1% of all splenocytes. Therefore, much lower Nluc signals derived from B cells (~50% of splenocytes) and other cell types that were not examined in this study would account for a non-negligible proportion.

**Figure 2.** Correlations between outputs of in vivo experiments in screening A. (**A**) A dot-plot of cellular uptake in Mϕ versus that in DCs. (**B**) A dot-plot of cellular uptake in B cells versus that in DCs. (**C**) A dot-plot of Nluc expression versus cellular uptake in both Mϕ and DCs. (**D**) A dot-plot of Nluc expression in spleen versus in DCs.

Concerning Nluc activity in splenic DCs, a well-fitted regression model was obtained by the effective design-based model selection for DSD process (Figure 3A). The Nluc expression (RLU/3000 cells) was converted to logarithms before regression because the range of the Nluc activity spans 3 orders of magnitude. A total of 5 out of all 6 factors that were examined in screening A were found to significantly contribute to Nluc expression in splenic DCs (Figure 3B). CL showed the highest impact on the Nluc expression level, and the use of CL4H6 was preferable to CL7H6. A higher %CL was also important in terms of improving the Nluc expression level. In addition, lower %PEG and higher NaCl concentration were also significant in improving the Nluc activity. The cellular uptake and Nluc expression in splenic DCs were plotted against ζ-average because the two factors are also significant determinants of LNP size. The plot clearly indicated that cellular uptake sharply increases by the ζ-average for sizes of up to ~200 nm (Figure 3C), but the upward trend was less pronounced in the range above 200 nm. On the other hand, the cellular uptake of extremely large LNPs (>700 nm) by DCs tended to be decreased. The plot for Nluc expression also showed a similar trend to cellular uptake (Figure 3D). Particles that are too large have the potential to induce toxicity and cause the obstruction of pulmonary microvessels. Therefore, an optimal LNP diameter for splenic DCs would be in the range of 200 to 500 nm. These observations suggest that the ζ-average is an intermediate factor that links the examined factors and cellular uptake or transgene expression level. To confirm the validity of the above regression model, the worst 5 and the top 5 LNPs were selected from 14 LNPs that were tested in screening A, and the occupancy of each level in each factor was then visualized. Nluc activity, cellular uptake, and ζ-average between the 2 groups were significantly different (Figure 3E–G). For %PEG, NaCl concentration, and CL, the levels of occupancy were clearly reversed between the two groups (Figure 3H–J). For example, 80% CL7H6 and 80% CL4H6 were occupied in the worst 5 and the top 5, respectively, for CL (Figure 3J). These results are consistent with the regression model. For the remaining factors, including %CL, %PL, and PEG, differences between the 2 groups were not clear, while both %CL and PEG were significant in the regression model (Figure 3K–M). Given the above findings, PEG was again examined in the next DOE. The level of %CL was increased from 40–60 to 50–60%. The level of %PL was narrowed down to 10–20% to reduce its impact on the size of LNPs. The level of %PEG was narrowed down from 0.5–1.5 to 0.75–1.5% to avoid the production of extremely large LNPs. CL was fixed with CL4H6. NaCl concentration was determined to be 300 mM for producing LNPs with an appropriate size based on simulations using the regression model.

In screening B, 8 different LNPs were synthesized based on the FFD (Table S2) and evaluated in vivo (Table S3). The diameters of the LNPs were in the range of approximately 110 to 420 nm with a narrow distribution (PdI < 0.15) and high encapsulation (>80%). In vivo screening indicated positive correlations between LNP size and cellular uptake or transgene expression in splenic DCs (R<sup>2</sup> = 0.8186 or 0.9173, respectively) (Figure S5A,B), a trend similar to that observed in screening A. Substantial correlations were observed when data from screening A and B were combined (Figure S5C,D). On the other hand, correlations between LNP size and transgene expression in the whole spleen were clearly lower (R<sup>2</sup> = 0.1148) (Figure S6A). The correlation between transgene activity in splenic DCs and the whole spleen was also low when data from screening A and B were combined (R <sup>2</sup> = 0.3582) (Figure S6B). These results suggest that transgene activity in the whole spleen is not a good indicator of the same process in splenic DCs, and that evaluation at the cell type level is essential for selecting the optimal formulation.

Based on the screenings, the A-11-LNPs showed the highest transgene activity in splenic DCs (Tables S1 and S3). The A-11-LNPs showed the most enhanced I-A/I-E expression in splenic DCs among all of the LNPs that were tested (Figure S7). The size and RNA encapsulation of the A-11-LNPs remained constant at 4 ◦C for at least 24 days after their preparation, indicating a stable formulation (Figure S8). Considering the above data, we conclude that the A-11-LNPs were the optimal formulation for targeting splenic DCs as mRNA vaccines.

**Figure 3.** Analysis of significant factors for outputs in in vivo experiment in screening A. (**A**) A plot of predicted and actual values for Nluc expression in splenic DCs obtained by a regression model. (**B**) Statistically significant main factors and interactions for Nluc expression in splenic DCs. (**C**,**D**) Plots of cellular uptake (**C**) or Nluc expression (**D**) in splenic DCs versus ζ-average. Two independent experiments with one mouse in each preparation were performed. The plots are expressed as the average of the two experiments. (**E**–**G**) Comparison of Nluc expression (**E**), cellular uptake (**F**), and ζ-average (**G**) between top 5 and worst 5 LNPs on Nluc expression in splenic DCs. *n* = 5, \* *p* < 0.05, \*\* *p* < 0.01. (**H**–**M**) Occupancy of each level of %PEG (**H**), NaCl conc. (**I**), CL (**J**), %CL (**K**), %DOPE (**L**), and PEG (**M**).
