Process Optimization of Charge-Reversible Lipid Nanoparticles for Cytosolic Protein Delivery Using the Design-of-Experiment Approach

Abstract

This study aimed to elucidate the manufacturing process parameters with optimal quality characteristics of protein-encapsulated dioleoylglycerophosphate–diethylenediamine (DOP-DEDA)-based lipid nanoparticles (LNPs) for intracellular protein drug delivery. DOP-DEDA is a pH-responsive and charge-reversible lipid for intracellular cargo delivery. In this study, bovine serum albumin (BSA) was used as a weakly acidic protein model, and LNPs were prepared using microfluidic technology, which has many advantages for practical applications. BSA-encapsulated DOP-DEDA-based LNPs showed pH-responsive charge reversibility and excellent quality characteristics for the intracellular delivery of proteins. A process optimization study was conducted by applying the Box–Behnken design in a design-of-experiment approach. The particle size, ζ-potential, and encapsulation efficiency were evaluated in response to the total flow rate, lipid concentration, and lipid solution ratio. The lipid solution ratio and total flow rate significantly affected the particle size and encapsulation efficiency, respectively. On the contrary, none of the process parameters affected the ζ-potential. Moreover, a map of the predicted values was constructed for the particle size and encapsulation efficiency using a multiple regression equation. In the predicted particle size range of 77–215 nm and encapsulation efficiency of 14–35%, the observed values were close to the predicted values, and 100-nm LNPs were reproduced with an encapsulation efficiency of 27%. Therefore, manufacturing process parameters were established to obtain protein-encapsulated DOP-DEDA-based LNPs with optimal quality characteristics.

INTRODUCTION

In recent years, biopharmaceuticals have become the center of drug development, and many biopharmaceuticals have been launched. Since the launch of the first recombinant protein therapeutic, namely, human insulin, in 1983, the number and frequency of use of protein therapeutics have increased drastically.^1,2)

However, many protein drugs have caused instability in blood, low tissue affinity, and difficulty in targeting, which hinder their clinical development.^1–3) Drug delivery systems (DDSs) for protein drugs have attracted considerable attention as a means to solve these problems.^3,4) For example, superoxide dismutase, which scavenges the superoxide anion of most damaging reactive oxygen species, has been bound to lecithin to improve its stability in blood and has been progressed to clinical trials.⁵⁾ Lipid nanoparticles (LNPs) are delivery systems used in nanomedicine to encapsulate and transport various small molecules, proteins, or genes to targeted cells or tissues. They consist of a core composed of a hydrophobic lipid surrounded by a layer of hydrophilic lipids.⁶⁾ This structure provides LNPs with a notable surface area-to-volume ratio, enhancing drug pharmacokinetics and biodistribution while reducing drug toxicity. In addition, LNPs enhance the solubility of many drugs and extend the shelf life and in vivo stability of peptides, proteins, and oligonucleotides. Other DDS approaches, such as polymer-based carriers, gold nanoparticles, and chemical modification, are also available. However, LNPs offer distinct advantages in terms of safety, including a lower likelihood of inducing immune responses.⁷⁾

Regarding the manufacturing of LNPs, batch-based methods such as reverse-phase evaporation, lipid hydration, freeze–thaw method, and extrusion have traditionally been used.⁸⁾ Recently, microfluidic devices and technologies have attracted attention for the manufacturing of LNPs. LNP production using microfluidic technology involves flowing an oil layer containing dissolved lipids and an aqueous layer containing dissolved drugs into two thin channels. Then, the two layers are mixed in a single channel to form LNPs. Microfluidics provides many advantages for LNP production, including precise LNP size controllability, high reproducibility, high-throughput optimization of LNP formulation, and continuous LNP production.⁹⁾ Particle size and encapsulation efficiency are important properties for the therapeutic efficacy, production efficiency, and pharmacokinetics of encapsulated drugs in LNPs.^10,11) Many studies have been conducted on various manufacturing processes that use microfluidic technology for small molecules and genes.^12–16) However, compared with other types of drug cargoes, fewer examples of various manufacturing processes of protein-based LNPs utilizing microfluidic techniques have been found in the literature.¹⁷⁾ For example, Anderluzzi et al. reported on the production of protein-encapsulated LNPs using a microfluidic device.¹⁸⁾ They established a process to encapsulate 100 µg/mL ovalbumin in LNPs with a particle size of 184 nm.

For the production of clinical trial materials and commercial products of protein-encapsulated LNPs, studying the manufacturing process from various perspectives is important. Accumulating considerable knowledge will help pharmaceutical companies to develop manufacturing processes efficiently. The International Conference on Harmonization Q8 guideline has recommended gaining information and knowledge of drug products from pharmaceutical development studies to support manufacturing controls.¹⁹⁾ In manufacturing protein-encapsulated LNPs with desirable properties, confirming the relationship between manufacturing conditions and drug product quality, as well as establishing the optimal manufacturing conditions, is necessary. The design-of-experiment (DoE) approach is used in many areas to optimize the manufacturing conditions.²⁰⁾ The response surface methodology is an effective technique to establish mathematical relationships between the predictors and the responses, aiming at process optimization. The central composite design and the Box–Behnken design are the main response surface methodologies used in the pharmaceutical field applied to nanotechnology.

The Box–Behnken design has several advantages, such as requiring fewer experiments, being efficient when dealing with many factors, reducing the influence of outliers, offering a good model fit, and simplifying the experimental process. It is also advantageous over the central composite design because it only requires three levels for each factor, whereas the central composite design requires five levels.²¹⁾

In the present study, microfluidic optimization of bovine serum albumin (BSA)-encapsulated dioleoylglycerophosphate–diethylenediamine (DOP-DEDA)-based LNPs (BSA-LNPs) was conducted using DoE approaches to efficiently accumulate knowledge on the manufacturing of protein delivery. DOP-DEDA is a pH-responsive and charge-reversible lipid developed by our laboratory for cytosolic gene delivery.²²⁾ Hirai et al. showed that DOP-DEDA-based LNPs can be applied to cytosolic protein delivery.²³⁾ BSA is a weakly acidic protein with a molecular weight of 66 kDa and an isoelectric point (pI) of 4.7.²⁴⁾ In this study, BSA is used in many biopharmaceutical studies²⁵⁾ and as a model protein. BSA is selected as a model protein because many approved biopharmaceuticals, such as cytokines and enzymes, have molecular weights of 10–100 kDa. For example, although pIs vary from acidic to basic, the pI of erythropoietin is approximately 3.5–4.0, indicating the presence of many weakly acidic proteins. BSA also has unique drug-binding characteristics using hydrophobic pockets where endogenous and exogenous hydrophobic substances such as fatty acids and drug molecules can bind. In addition, BSA is extensively studied as a carrier of poorly water-soluble compounds such as warfarin, diazepam, and paclitaxel.^26,27) One-way design experiments were used to independently determine the lipid/protein loading ratio and aqueous-phase pH of BSA-LNPs. These factors should be initially determined to prepare BSA-LNPs without aggregation. Then, a 2-(p-toluidinyl)naphthalene-6-sulfonic acid (TNS) assay should be performed to characterize the BSA-LNPs and evaluate the effect of surface charge by solution pH variation and stability in serum during storage. The Box–Behnken design was used as the DoE approach, and the particle size, ζ-potential, and encapsulation efficiency were evaluated. The total flow rate, lipid solution ratio, and lipid concentration are important parameters^9,13,28) that were selected for verification using the DoE approach. Conversely, the protein loading amount was fixed to allow particle formation because excessive loading can cause particle aggregation. The buffer pH was also fixed due to its notable effect on particle size. We excluded temperature from the verification parameters because a previous study¹³⁾ reported that temperature had no effect on particle size. A map of the predicted values for the particle size and encapsulation efficiency was created using the multiple regression equation.

MATERIALS AND METHODS

Materials

DOP-DEDA, dipalmitoylphosphatidylcholine (DPPC), and cholesterol were synthesized by Nippon Fine Chemical Co., Ltd. (Hyogo, Japan). BSA was purchased from Sigma-Aldrich (St. Louis, MO, U.S.A.). tert-Butanol was purchased from FUJIFILM Wako Pure Chemical Corporation (Osaka, Japan).

Preparation of BSA-LNPs for DoE Study

DPPC, DOP-DEDA, and cholesterol were dissolved in tert-butanol to a final lipid concentration of 10 mM (DPPC : DOP-DEDA : cholesterol 10 : 45 : 45). BSA was dissolved in 1 mM sodium citrate (BSA concentration of 10 mg/mL) and adjusted to pH 4.5. Lipid and BSA were fixed at a molar ratio of 150 : 1. (When the final lipid concentration was 1 mM, BSA was added. Thus, the final concentration was 6.67 µM [440 µg/mL assuming 66 kDa of BSA].) A microfluidic device (KC-M-S-SUS, YMC) and two syringe pumps (PUMP 11 ELITE, Harvard Apparatus) were placed in an incubator set at 45°C, and the lipid mixture and BSA solution were each pumped using syringe pumps, and BSA-LNPs were obtained. Then, BSA-LNPs were dialyzed against ultrapure water using a 300-kDa cut-off dialysis membrane (Spectrum Laboratories) to remove tert-butanol and free BSA.

Particle Size and ζ-Potential Measurement

In this study, 50 µL of BSA-LNPs was diluted with 900 µL of ultrapure water, and the particle size (Z-average) and polydispersity index (PDI) were measured using the Zetasizer Nano ZS (Malvern, Worcestershire, U.K.). After diluting 50 µL of BSA-LNPs with 900 µL of 10 mM Tris–HCl buffer (pH 7.4), ζ-potential was measured using the Zetasizer Nano ZS. Considering that DOP-DEDA-based LNPs exhibit a pH-dependent surface charge, 10 mM Tris–HCl buffer (pH 7.4) was used as the solvent to ensure consistent results. The ζ-potential was measured using automatic attenuator adjustment with up to 100 pulses to enhance the measurement accuracy. The quality of the results was confirmed by generating a quality report for each individual measurement.

Calculation of the BSA Encapsulation Efficiency in LNP Formulation

Twenty microliters of 25% octyl glycoside (Dojindo Laboratories; final concentration of 100 mg/mL) were added to 30 µL of BSA-LNPs to disrupt the LNPs. Subsequently, 200 µL of the bicinchoninic acid assay reaction solution (Thermo Fisher Scientific) was added and incubated at 37°C for 30 min. The absorbance at 562 nm was then measured using a microplate reader (TECAN Infinite 200) to calculate the BSA concentration. Standard solutions with concentrations ranging from 1 to 1000 µg/mL were used for calibration. The R² of the regression line was 0.998 ± 0.007 (n = 5).

The BSA encapsulation efficiency in BSA-LNPs was calculated as follows:

  

$$\begin{array}{c}\mathrm{\%}\text{of encapsulation efficiency of BSA in LNP}\\ \text{formulation = Amount of BSA in BSA-LNP (mg)/}\\ \text{Total amount of BSA added (mg)}\times 100\end{array}$$

Transmission Electron Microscopy (TEM)

To observe the morphology of the LNPs, TEM images were collected using a Hitachi transmission electron microscope (HT7700, Hitachi High-Tech Corporation). The LNPs were concentrated to a final lipid concentration of 5–10 mM. A small amount (5 µL) of LNPs was placed on a 200-mesh TEM copper grid covered with a formvar film. The excess solution on the grid was removed using filter paper, followed by negative staining using a 1% ammonium molybdate solution.

The sample was then transferred to a specimen stage of the transmission electron microscope operated at an accelerating voltage of 80 kV. TEM images were obtained at magnifications of ×12000–40000 at 22°C under vacuum conditions.

TNS Assay

The fluorescent probe TNS (Sigma-Aldrich), which exhibits increased fluorescence in a hydrophobic environment, can be used to assess the surface charge on lipid layers. A total of 32 µL of 1 mM BSA-LNPs was diluted in 760 µL of buffer containing 10 mM 4-(2-hydroxyethyl)piperazine-1-ethane-sulfonic acid (HEPES), 10 mM 2-(N-morpholino)-ethanesulfonic acid (MES), 10 mM ammonium acetate, and 130 mM sodium chloride. Then, 8 µL of 0.1 mM TNS was added. The fluorescence intensity of TNS was determined by using a Tecan Infinite M200 microplate reader in accordance with the manufacturer’s instructions (ex. 321 nm, em. 445 nm). In addition, pK_a was calculated as the pH at which the relative fluorescence intensity of 0.5 was shown, when the maximum fluorescence intensity of each BSA-LNP was set to a relative value of 1.0.

Hemolysis Assay

In preparing erythrocytes, 500 µL of caw blood was washed by gently vortexing with 1 mL of phosphate-buffered saline (PBS) and was centrifuged at 10000 × g for 10 min at 4°C. After washing with PBS 5 times, the pellet was suspended in PBS. BSA-LNPs diluted with 10 mM phosphate buffer (pH 7.4 or 5.5) containing 0.3 M sucrose were mixed with 20 µL of the erythrocytes (200 µM as amine moiety) and incubated at 37°C for 1 h in a shaking container. After centrifugation (10000 × g for 10 min at 4°C), the liberated hemoglobin was determined by colorimetric analysis of the supernatant at 405 nm using a Tecan Infinite M200 microplate reader. One hundred percent hemolysis was set from the erythrocytes treated with 10% Triton X-100. The hemolysis rate was calculated as follows: Hemolysis rate (%) = (absorbance of sample solution supernatant – absorbance of negative control solution supernatant)/absorbance of positive control solution supernatant – absorbance of negative control solution supernatant) × 100. The hemolysis grade was evaluated using the following index: 0 to less than 2 indicates no hemolysis, 2 to less than 5 indicates mild hemolysis, and 5 or more indicates hemolysis.²⁹⁾

Stability in Serum

BSA-LNPs containing 1 mM of lipid were mixed with 50% fetal bovine serum (FBS) in a 1 : 1 ratio (final concentration 25% FBS). After incubation at 37°C for 0, 2, and 4 h, the turbidity was measured by measuring the optical density at 600 nm using a Tecan Infinite M200 microplate reader. After incubation, the samples were diluted 20 times with ultrapure water, and the particle size was measured using the Zetasizer Nano ZS.

Long-term Stability

BSA-LNPs containing 1 mM of lipid were stored at 4°C for 0, 3, 6, and 15 weeks. In addition, the particle size and BSA concentration were measured.

DoE

A quadratic model was generated using a 3-factor Box–Behnken design. In this design, all elements have 3 levels: low, center, and high. Three central samples were included in this design and used as a source for error estimation. The quadratic model was calculated using multiple regression analysis, and the model was described as follows:

  

$$\begin{array}{c}\text{Y}={\text{a}}_0+{\text{a}}_1{\text{X}}_1+{\text{a}}_2{\text{X}}_2+{\text{a}}_3{\text{X}}_3+{\text{a}}_4{\text{X}}_1{\text{X}}_2+{\text{a}}_5{\text{X}}_1{\text{X}}_3\\ +{\text{a}}_6{\text{X}}_2{\text{X}}_3+{\text{a}}_7{\text{X}}_1^2+{\text{a}}_8{\text{X}}_2^2+{\text{a}}_9{\text{X}}_3^2\end{array}$$

where Y is the response, X is the variable, and a is the regression coefficient. The total flow rate (mL/min), lipid concentration (mM), and lipid solution ratio (%) were selected as variables in the microfluidic process. The lipid solution ratio refers to the liquid velocity of the lipid phase (mL/min)/the liquid velocity of the aqueous solution phase (mL/min) × 100. The statistically significant terms in the model were identified through ANOVA. The response surface plots were generated by using this model.

Statistical Analysis

JMP® (Version 15, JMP Statistical Discovery LLC., NC, U.S.A.) was used for multiple regression analysis, ANOVA, and response surface methodology. Multiple regression equations were calculated using the least squares method.

RESULTS

Preparation of BSA-LNPs

In this study, the basic preparation conditions for BSA-LNPs were investigated as a preliminary study. First, LNPs were prepared at various BSA/lipid ratios under provisional preparation conditions to determine the loading amount of BSA (Table 1). A transparent solution was obtained when a BSA/lipid ratio of 1 : 150 was reached, but a white precipitate was formed when the BSA/lipid ratio was higher than 1 : 50. During the formation of precipitates, the particle size was large.

Table 1. Physicochemical Characteristics of BSA-LNPs at Various Molar Ratios

BSA (µM)	Lipid (mM)	Molar ratio (BSA:lipid)	Particle size (nm)	PDI	Appearance
0	2.5	Not applicable	131.4	0.182	Clear
5	2.5	1 : 500	110.1	0.212	Clear
16.7	2.5	1 : 150	116.4	0.275	Clear
50	2.5	1 : 50	8092	0.181	Precipitated
167	2.5	1 : 15	3760	1.000	Precipitated

The lipid mixture was composed of DOP-DEDA, DPPC, and cholesterol at a 45 : 10 : 45 M ratio. The molar ratios denote BSA/total lipids. The microfluidic process was conducted under the following conditions: total flow rate of 3.3 mL/min, lipid concentration of 2.5 mM, lipid solution ratio of 24.1%, and pH 5.0. PDI = polydispersity index.

Next, LNPs were prepared by varying the pH of the BSA solution to determine the optimal pH. Using DOP-DEDA-based BSA-LNPs, the particle size ranged from 100 to 200 nm at pH 4.0 to 5.0, but a white precipitate was formed at pH 6.0–7.0. The particle size was also large (Fig. 1A). By contrast, DOPC-based BSA-LNPs, which are non-charge-reversible LNPs, were prepared, and no change in particle size was observed because of pH fluctuation. Empty DOP-DEDA-based LNPs (without BSA) did not aggregate between pH 4.0 and 7.0. DOP-DEDA-based BSA-LNPs showed higher BSA encapsulation efficiency of 20.6–25.7% than DOPC-based BSA-LNPs at pH 4.0–5.0 (16.6–17.5%), with pH 4.5 being the highest (Fig. 1B).

Fig. 1. Particle Size and BSA Encapsulation Efficiency of BSA-LNPs at Various pH Values

The DOP-DEDA-based BSA-LNPs and empty DOP-DEDA-based LNPs were prepared with a lipid mixture composed of DOP-DEDA, DPPC, and cholesterol at a molar ratio of 45 : 10 : 45. The DOPC-based BSA-LNPs were prepared with a lipid mixture composed of DOPC, DPPC, and cholesterol at a molar ratio of 45 : 10 : 45. The microfluidic process was conducted under the following conditions: total flow rate of 3.3 mL/min, lipid concentration of 1.9 mM, lipid solution ratio of 24.1%, and BSA : lipid with a molar ratio of 1 : 150. # indicates precipitation. The results are presented as the mean ± S.D. of more than two independent experiments. BSA = bovine serum albumin, DOP-DEDA = dioleoylglycerophosphate–diethylenediamine, DPPC = dipalmitoylphosphatidylcholine, DOPC = dioleoylphosphocholine.

The BSA concentrations of all samples ranged from 10 to 500 µg/mL. The BSA recovery rates were 103.8 ± 0.4% (n = 5) and 99.3 ± 7.0% (n = 5) at BSA concentrations of 10 and 500 µg/mL, respectively.

Characteristics of BSA-LNPs

The morphologies of BSA-LNPs and empty LNPs were analyzed by TEM (Figs. 2A, 2B). The TEM images revealed that the BSA-LNPs were spherical and covered with a thick membrane (Fig. 2A), whereas the empty LNPs were also spherical but had a thinner membrane (Fig. 2B). A TNS assay was performed to calculate the apparent pK_a value of the BSA-LNPs, which is approximately 6.0 (Fig. 2C). In observing the change in the surface charge of BSA-LNPs caused by changes in the external solution pH environment, the ζ-potential was measured when the external solution pH was changed to 10 mM Tris–HCl buffer. The results indicated that the surface charge changed from positive to negative as the pH changed from the acidic region to the neutral region and to a weak alkaline solution. BSA-LNPs showed a positive charge at pH 6.5 or less, a nearly neutral state at around pH 7.4, and a negative charge at pH 8.0 or more (Fig. 2D). In addition, a hemolysis test was performed to estimate the escape ability of BSA-LNPs from intracellular endosomes. DOP-DEDA-based BSA-LNPs did not show hemolysis at pH 7.4, but they showed hemolysis at pH 5.5. DOPC-based BSA-LNPs, which are non-charge-reversible lipids, did not show hemolysis at either pH 7.4 or pH 5.5 (Fig. 2E). In evaluating the stability of BSA-LNP in the blood, time-dependent changes were observed in BSA-LNPs in FBS. After 4 h in 25% FBS at 37°C, the particle size of BSA-LNPs decreased slightly, but no change in PDI, turbidity, or aggregation was observed (Figs. 2F, 2G). A stability test was also performed in ultrapure water at 4°C. No change was observed in the particle size, PDI, or BSA concentration during observation for 15 weeks (Figs. 2H, 2I).

Fig. 2. Characteristics of BSA-LNPs

(A) TEM image of BSA-LNPs. (B) TEM image of empty LNPs. Scale bars = 200 nm. (C) TNS assay of BSA-LNPs. (D) Effect of changes in solution pH on the ζ-potential of BSA-LNPs. A total of 10 mM Tris–HCl buffer was used as the solution buffer. (E) Hemolysis of BSA-LNPs at pH 5.5 and 7.4. DOP-DEDA-based BSA-LNPs were composed of DOP-DEDA, DPPC, and cholesterol at a 45 : 10 : 45 M ratio. DOPC-based BSA-LNPs were composed of DOPC, DPPC, and cholesterol at a 45 : 10 : 45 M ratio. Ten percent Triton X-100 was used as a positive control. (F) Time course of the particle size of DOP-DEDA-based BSA-LNPs in 25% fetal bovine serum (FBS) at 37°C. (G) Time course of the turbidity (optical density at 600 nm) of DOP-DEDA-based BSA-LNPs in 25% FBS at 37°C. The aggregated BSA-LNPs of the same composition were used as positive controls. (H) Stability in the particle size of DOP-DEDA-based BSA-LNPs in ultrapure water at 4°C. (I) Stability of the BSA concentration of DOP-DEDA-based BSA-LNPs in ultrapure water at 4°C. The relative changes from the initial values are shown. The results are presented as the mean ± S.D. of more than three independent experiments (C, E–G) and two independent experiments (D, H,I). n.s., not significant; **p < 0.01; ***p < 0.001 (unpaired t-test). BSA = bovine serum albumin, DOP-DEDA = dioleoylglycerophosphate–diethylenediamine, DPPC = dipalmitoylphosphatidylcholine, DOPC = dioleoylphosphocholine.

Calculation of the Model Formula by DoE

Fifteen experiments were planned using the Box–Behnken design. All BSA-LNPs prepared under the designed conditions were nanosized and uniform particles. The smallest and largest particle sizes were 98 and 315 nm, respectively. PDI ranged from 0.026 to 0.279. The smallest BSA encapsulation efficiency was 6.0%, and the highest was 44.0%. The ζ-potential ranged from –3.21 to +5.29 mV (Table 2). Multiple regression analysis of the particle size of BSA-LNPs was performed. The particle size increased significantly with the increase in lipid solution ratio (Table 3). The theoretical values calculated from this model were compared with the observed values, and a relatively high correlation (R² = 0.83) was observed (Figs. 3A, 3C, 3E). The factor effect and response surface plot on the particle size of BSA-LNPs showed a positive correlation between the particle size and the lipid solution ratio.

Table 2. Box–Behnken Design and Observed Results of the Optimization Study

No.	Microfluidic process			Particle size (nm)	PDI	ζ-Potential at pH 7.4 (mV)	Percentage of entrapped BSA (%)
	Total flow rate (mL/min)	Lipid concentration (mM)	Lipid solution ratio
1	3	0.5	30	171.0	0.109	0.481	16.8
2	3	0.5	20	94.3	0.077	2.34	26.1
3	3	1.5	30	199.5	0.026	0.730	29.2
4	3	1.5	20	98.3	0.161	1.04	22.5
5	2	1	20	112.4	0.150	–1.73	24.2
6	2	1	30	251.5	0.077	–1.50	38.3
7	2	0.5	25	226.5	0.135	2.53	26.9
8	2	1.5	25	262.5	0.135	–0.126	44.0
9	4	1	20	125.7	0.205	5.29	24.5
10	4	1	30	315.1	0.279	0.528	6.7
11	4	0.5	25	132.6	0.060	–1.92	6.0
12	4	1.5	25	198.5	0.117	0.963	12.2
13	3	1	25	146.4	0.034	–3.21	33.1
14	3	1	25	135.0	0.039	–2.46	33.2
15	3	1	25	149.3	0.043	–2.40	16.2

Evaluation of the effect of changing the total flow rate, lipid concentration, and lipid solution ratio on the particle size, ζ-potential, and BSA encapsulation efficiency. The lipid mixture was composed of DOP-DEDA, DPPC, and cholesterol at a 45 : 10 : 45 M ratio. The other conditions of the microfluidic process were fixed at pH 4.5, and the molar ratio of BSA : lipid was 1 : 150. PDI = polydispersity index.

Table 3. Results of ANOVA for the Particle Size of BSA-LNPs

Input variables	Sum of squares	Degree of freedom	p-Value
Total flow rate	820.125	1	0.5584
Lipid concentration	2257.248	1	0.3462
Lipid solution ratio	32050.056	1	0.0112*
Total flow rate × total flow rate	13704.938	1	0.0506
Total flow rate × lipid concentration	223.503	1	0.7568
Lipid concentration × lipid concentration	1.054	1	0.9830
Total flow rate × lipid solution ratio	632.523	1	0.6058
Lipid concentration ×lipid solution ratio	150.308	1	0.7992
Lipid solution ratio × lipid solution ratio	40.596	1	0.8946
Model	10330.323	3
Error	114.287	2
Adjusted total	10444.610	5

* p < 0.05.

Fig. 3. Statistical Analysis Results of BSA-LNPs Using the Box–Behnken Design

(A, B) Observed and predicted particle size, and the observed and predicted encapsulation efficiency of BSA-LNPs. The leverage plot includes a confidence curve for the regression line. This curve indicates whether the test is significant at the 5% level. The horizontal lines, which represent mean Y, were included among the curves, and the effects were not significant. (C, D) Factorial effect plot on the particle size and encapsulation efficiency of BSA-LNPs. This plot shows the mechanism by which the predicted value changes when changing the settings of each factor. The gray areas indicate the 95% confidence interval. (E, F) Response surface on the particle size and encapsulation efficiency. The response surfaces show the responses of the total flow rate and lipid solution ratio at a lipid concentration of 1.0 mM. Statistical analysis was performed using JMP^®. BSA = bovine serum albumin.

The results of multiple regression analysis on the encapsulation efficiency of BSA-LNPs indicated that the encapsulation efficiency was significantly improved by decreasing the total flow rate (Table 4). When the theoretical values calculated from the model were compared with the observed values, a relatively high correlation (R² = 0.84) was observed (Figs. 3B, 3D, 3F). The factor effect and response surface plot on the encapsulation efficiency of BSA-LNPs indicated that the BSA encapsulation efficiency was negatively correlated with the total flow rate. In addition, the results of multiple regression analysis of the ζ-potential of BSA-LNPs indicated that the process parameters of the total flow rate, lipid concentration, and lipid solution ratio had no effect on the ζ-potential (Table 5). This result indicates that the specific process parameters did not affect the ζ-potential of BSA-LNPs. The calculated model formula for BSA-LNPs on each subject is shown in Table 6.

Table 4. Results of ANOVA for the Encapsulation Efficiency of BSA-LNPs

Input variables	Sum of squares	Degree of freedom	p-Value
Total flow rate	882.00000	1	0.0103*
Lipid concentration	128.80125	1	0.1866
Lipid solution ratio	4.96125	1	0.7761
Total flow rate × total flow rate	27.41769	1	0.5118
Total flow rate × lipid concentration	29.70250	1	0.4956
Lipid concentration ×lipid concentration	23.07692	1	0.5459
Total flow rate × lipid solution ratio	254.40250	1	0.0843
Lipid concentration × lipid solution ratio	64.00000	1	0.3302
Lipid solution ratio × lipid solution ratio	6.72923	1	0.7408
Model	83.67750	3
Error	191.54000	2
Adjusted total	275.21750	5

* p < 0.05.

Table 5. Results of ANOVA for the ζ-Potential of BSA-LNPs

Input variables	Sum of squares	Degree of freedom	p-Value
Total flow rate	4.042746	1	0.3669
Lipid concentration	0.084872	1	0.8914
Lipid solution ratio	5.612925	1	0.2953
Total flow rate × total flow rate	6.007016	1	0.2808
Total flow rate × lipid concentration	7.670130	1	0.2302
Lipid concentration × lipid concentration	11.649467	1	0.1531
Total flow rate × lipid solution ratio	6.230016	1	0.2731
Lipid concentration × lipid solution ratio	0.599850	1	0.7182
Lipid solution ratio × lipid solution ratio	15.691504	1	0.1082
Model	20.150042	3
Error	0.407400	2
Adjusted total	20.557442	5

Table 6. Calculated Model Formula of BSA-LNPs for Each Subject

Subject	Calculated model formula
Particle size	Y1 = –10.125(X1 – 3) + 16.7975(X2 – 1)/0.5 + 12.659X3 + 7.475(X1 – 3)(X2 – 1)/0.5 + 2.515(X1 – 3)(X3 – 25) + 1.226(X2 – 1)(X3 – 25)/0.5 + 60.924166667(X1 – 3)(X1 – 3) + 0.5341666667(X2 – 1)(X2 – 1)/0.5/0.5 – 0.132633333(X3 – 25)(X3 – 25) – 172.9083333 (1)
ζ-Potential	Y2 = 0.710875(X1 – 3) – 0.103(X2 – 1)/0.5 – 0.167525X3 + 1.38475(X1 – 3)(X2 – 1)/0.5 – 0.2496(X1 – 3)(X3 – 25) + 0.07745(X2 – 1)(X3 – 25)/0.5 + 1.2755(X1 – 3)(X1 – 3) + 1.77625(X2 – 1)(X2 – 1)/0.5/0.5 + 0.08246(X3 – 25)(X3 – 25) + 1.498125 (2)
BSA encapsulation efficiency	Y3 = –10.5(X1 – 3) + 4.0125(X2 – 1)/0.5 – 0.1575X3 – 2.725(X1 – 3)(X2 – 1)/0.5 – 1.595(X1 – 3)(X3 – 25) + 0.8(X2 – 1)(X3 – 25)/0.5 – 0.054(X1 – 3)(X1 – 3) – 2.5(X2 – 1)(X2 – 1)/0.5/0.5 + 0.0906667(X3 – 25)(X3 – 25) + 31.4375 (3)

Y₁ is the particle size, Y₂ is the ζ-potential, Y₃ is the BSA encapsulation efficiency, X₁ is the total flow rate, X₂ is the lipid concentration, and X₃ is the lipid solution ratio.

Confirmation Study of the Model Formula

A confirmation study was conducted to verify the predictive ability of the model. The manufacturing parameters for the confirmation study were set to target the results within the particle size range of 77–215 nm or encapsulation efficiency of 14–35%, based on the map of the predicted values of particle size and encapsulation efficiency from each manufacturing condition (Fig. 4). The observed values were close to the predicted values (Nos. 1–4 in Table 7). Furthermore, under a particle size of 100 nm and a relatively high entrapment efficiency of 27%, BSA-LNPs were prepared three times in succession, and the expected results were obtained with small variability (Nos. 5–7 in Table 7). In addition, the PDI of Nos. 5–7 BSA-LNPs was 0.118 ± 0.050 (mean ± standard deviation (S.D.), n = 3), which is less than 0.2, indicating a small variability in the particle size distribution.

Fig. 4. Map of the Predicted Values for Particle Size and Encapsulation Efficiency from Each Manufacturing Condition

A total of 1331 (11³) manufacturing conditions were entered (11 conditions each with a total flow rate ranging from 2.0 to 4.0 mL/min, lipid concentration of 0.5–1.5 mM, and lipid solution ratio of 20–30%). The experiment numbers in the confirmation study were indicated by red dots. BSA = bovine serum albumin.

Table 7. Experimental Design and Observed Results of the Confirmation Study

No.	Microfluidic process			Particle size (nm)			Percentage of entrapped BSA (%)			PDI
	Total flow rate (mL/min)	Lipid concentration (mM)	Lipid solution ratio	Predicted	Observed		Predicted	Observed		Observed
1	3	1	20	77.0	63.2		26.9	24.8		0.358
2	3	1	30	203.5	189.2		25.4	24.7		0.024
3	2	1	25	214.6	183.9		35.3	31.2		0.082
4	4	1	25	194.4	189.7		14.3	12.7		0.059
5	3	1	22	104.4	100.1	Ave. 100.3 ± 0.2	27.5	26.4	Ave. 27.3 ± 2.8	0.189	Ave. 0.118 ± 0.050
6					100.6			31.0		0.082
7					100.2			24.4		0.084

Results of No. 5–7 are presented as mean ± S.D. PDI = polydispersity index.

DISCUSSION

Preparation of BSA-LNPs

First, the basic preparation method of BSA-LNPs was investigated. Aggregation occurred when the amount of loaded BSA exceeded that of the lipid. When the lipid formed LNPs in the microchannel, the BSA became an obstacle, hindering the formation of the lipid membrane and thus causing destabilization. Considering the production efficiency, the loading amount must be increased as much as possible, so a BSA/lipid ratio of 1 : 150 was adopted.

When the pH of the BSA solution during LNP preparation was changed, the particle size and BSA encapsulation efficiency remarkably changed in the charge-reversible DOP-DEDA-based LNPs but not in the non-charge-reversible DOPC-based LNPs. The pK_a value of each amine in DOP-DEDA is approximately 5.9 and 10; thus, the positively charged amine interacts intramolecularly with the negatively charged phosphate at pH 7.4, resulting in an almost neutral charge. At pH 6.0, DOP-DEDA has a positive charge because the number of charged amines increases, and the phosphates remain unchanged. Similarly, at pH 8.0, the number of charged amines decreases, resulting in a negative charge.²²⁾ In addition, the pI value of BSA is around 4.7, so the BSA solution is slightly positively charged at pH 4.0, neutral at pH 4.5 to 5.0, and negatively charged at pH 6.0 to 7.0. The positively charged lipids and negatively charged BSA were strongly attracted to each other because of electrostatic interactions; thus, DOP-DEDA-based BSA-LNPs were aggregated when prepared at pH 6.0–7.0, resulting in the failure to form particles. The fact that no aggregation was observed in empty DOP-DEDA-based LNPs without BSA at a pH range of 6.0–7.0 supports this conclusion. When the DOP-DEDA-based BSA-LNPs were prepared at pH 4.5–5.0, the positively charged lipids and the neutral BSA did not form strong electrostatic interactions; thus, particles with a size of 100–200 nm were successfully formed. At pH 4.0, the encapsulation efficiency was lower than that at pH 4.5. An electrostatic repulsion was observed between the positively charged lipids and the positively charged BSA. In addition, the encapsulation efficiency of the DOP-DEDA-based BSA-LNPs was higher at pH 4.5 compared to the non-charge-reversible DOPC-based BSA-LNPs. Moderate electrostatic interactions result in good encapsulation. The encapsulation efficiency of the DOP-DEDA-based BSA-LNPs is high compared with the encapsulation efficiency of 11.0% when Eltoukhy et al. encapsulated neutravidin (molecular weight 60 kDa and pI 6.3) in non-charge-reversible LNPs.³⁰⁾ Considering the balance between particle size and BSA encapsulation efficiency, pH 4.5 was used in subsequent experiments.

DOP-DEDA-based LNPs are positively charged at pH 6.0, neutral at pH 7.4, and negatively charged at pH 8.0. Therefore, if the pI value of a protein is lower than 6.0, the encapsulation efficiency may improve due to electrostatic interactions between the LNP positive charge and the protein negative charge. Conversely, if the pI value is greater than 8.0, then encapsulation efficiency may be improved by the electrostatic interaction between the LNP negative charge and the protein positive charge. The non-charge-reversible DOPC–based LNPs also encapsulated BSA with an encapsulation efficiency of 16.5–17.5%. Considering that DOPC is uncharged, this encapsulation likely occurs through hydrophilic/hydrophobic interactions between BSA and the lipids, along with stochastic incorporation of the protein into the LNPs, rather than through electrostatic interactions. The difference in BSA encapsulation efficiency between DOP-DEDA-based LNPs and DOPC-based LNPs (9.1% at pH 4.5) can therefore be attributed to the effect of electrostatic interactions.

Characteristics of BSA-LNPs

In this study, whether BSA-LNPs have desirable properties for intracellular protein delivery was evaluated. TEM images revealed that the BSA-LNPs were covered with a thicker membrane compared with empty LNPs. Hirai et al. obtained cryo-TEM images of LNPs with the same lipid composition as the BSA-LNPs and found that LNPs encapsulating supernegative green fluorescent protein exhibited a multilayer structure.²³⁾ In contrast, empty LNPs had a monolayer structure. Despite the differences between BSA (66 kDa, pI 4.7) and the supernegative green fluorescent protein (27 kDa, negative charge), the multilayer structure is likely formed by lipid–protein interactions. In determining the apparent pK_a value of BSA-LNPs, a TNS assay was performed, which obtained a pK_a value of 6.0. Jayaraman et al. investigated the relationship between pK_a and in vivo liver gene silencing activity and showed that the optimal pK_a value ranged from 6.2 to 6.5.³¹⁾ BSA-LNPs have a pK_a close to the optimal value, indicating their useful protein delivery functions. When the surface charge of the BSA-LNPs was measured, they showed a positive charge below pH 6.5, a nearly neutral charge around pH 7.4, and a negative charge above pH 8.0. DOP-DEDA has a pK_a value of approximately 5.9 and 10, which allows cationic amines to interact intramolecularly with anionic phosphates at pH 7.4, resulting in a nearly neutral total charge.²²⁾ On the contrary, at pH 6.5, DOP-DEDA has a total positive charge because the number of charged amines increases, and the number of phosphates remains unchanged. Similarly, at pH 8.0, the number of charged amines decreases, and the number of phosphates remains unchanged, resulting in a total negative charge. BSA-LNPs become cationic in a slightly acidic tumor microenvironment and interact with the anionic plasma membrane of tumor cells. This interaction allows the LNPs to be internalized into cancer cells by endocytosis. In addition, the cargo in the endosome is at a lower pH (pH 5.5) than the physiological pH (pH 7.4), and the cationic LNPs disrupt the endosomal membrane and release the protein from the LNPs into the cytoplasm. Therefore, BSA-LNPs are a useful protein delivery vector for cancer therapy.

In evaluating the endosomal escape ability of BSA-LNPs, the hemolysis of red blood cells was evaluated in an endosomal environment with pH 5.5, and hemolysis was observed. In contrast, hemolysis was not observed at physiological pH 7.4. Similarly, the same test was performed using non-charge-reversible DOPC-based BSA-LNPs, and hemolysis was not observed at pH 5.5. According to Alabi et al., endosomal escape hinders the efficient intracellular delivery of LNPs, and the relative capacity for LNP-mediated endosomal escape was simulated using a hemolysis assay.³²⁾ The hemolysis assay results indicate that DOP-DEDA-based BSA-LNPs have desirable properties for intracellular protein delivery because of their charge reversibility. Alabi et al. also noted that one of the early barriers LNPs encounter on the route to the target cell is serum protein binding and the potential for premature nanoparticle disassembly in the extracellular milieu.³²⁾ Using BSA-LNPs, no significant change in particle size and no aggregation were observed for at least 4 h in serum at 37°C. Therefore, BSA-LNPs stabilize when administered into the blood. Furthermore, while additives such as polyethylene glycol (PEG) and pH adjusters are generally added to LNP formulations,³³⁾ BSA-LNPs maintained a stable particle size and BSA concentration for at least 15 weeks in ultrapure water under refrigerated conditions, although they did not contain any PEG or pH adjusters. These data indicate the stability of BSA-LNPs.

Calculation of the Model Formula by DoE

A Box–Behnken design was used to optimize the microfluidic process, which allows the efficient estimation of a second-order model. The process parameters affected the particle size and encapsulation efficiency of BSA-LNPs. In contrast, no specific process parameters influenced the ζ-potential of BSA-LNPs. Particle size is an important property of the pharmacokinetics of the encapsulated drug in LNPs.¹⁰⁾ The particle size increased significantly with the increase in the lipid solution ratio in BSA-LNPs. In general, in microfluidic technology, when the total flow rate is high, the lipid concentration is low, or the flow ratio of the lipid phase is relatively lower than that of the aqueous phase. Moreover, the lipids become more stable spheres in a short time, resulting in small particles.^9,28,34,35) Conversely, when the total flow rate is low, the lipid concentration is high, or the flow ratio of the lipid phase is relatively lower than that of the aqueous phase. The lipids also become more stable spheres over a longer time, resulting in larger particles. Based on the result, the lipid solution ratio significantly affects the particle size. On the contrary, the total flow rate and lipid concentration,^28,35) which may affect the particle size, did not significantly affect the particle size in this study. Each level of the parameters set in this DoE was within the range where LNPs can be prepared with nanosized and uniform particles, referring to previous examinations and literature values. Making a detailed comparison is difficult because the setting range of the total flow rate, the setting range of the lipid concentration, the difference in lipid composition, the difference in microfluidic device, and the difference in the type of drug cargoes may have a complex effect. However, under the conditions set in this study, the lipid solution ratio dominated the particle size. This finding contrasts with studies on empty LNPs and nucleic acid-encapsulated LNPs.^28,35) For example, Maeki et al. reported that the total flow rate, lipid concentration, and lipid solution ratio all influenced particle size in empty LNPs, suggesting that LNP size can be tuned by optimizing residence time at critical concentrations within microfluidic devices. In this study, however, only the lipid solution ratio influenced particle size, likely due to the lipid–protein interactions that protect the lipids from water erosion.³⁶⁾ This finding was further supported by the observation that when proteins were encapsulated in LNPs, the particle size was larger than that of empty LNPs prepared under the identical conditions.

The ζ-potential also plays a role in the in vivo behavior of LNPs and their stability.³⁷⁾ The microfluidic process parameters did not affect the ζ-potential in this study. ζ-Potential measurements often exhibit high variability, especially near zero. In this study, all data were close to zero, indicating that the LNPs were neither significantly positively nor negatively charged. The lipid composition of LNPs is an important factor in the ζ-potential, and the amount of cationic and anionic lipids determines the ζ-potential of LNPs.³⁸⁾ Therefore, the microfluidic process may not be important for the ζ-potential in BSA-LNPs. The encapsulation efficiency is important with regard to therapeutic efficacy and production efficiency.¹¹⁾ The encapsulation efficiency significantly increased with the decrease in the total flow rate in BSA-LNPs.

According to Forbes et al., the encapsulation efficiency of ovalbumin (pI = 4.5) decreased in neutral and anionic LNPs because of the decrease in total flow rate. In addition, they stated that the encapsulation efficiency of proteins is substantially controlled by the electrostatic interactions between the proteins’ peripheral charge and the polar head group of phospholipids.³⁹⁾ In this study, the decrease in total flow rate increased the encapsulation efficiency of BSA by strengthening the electrostatic and hydrophilic/hydrophobic interactions between BSA and the cationically charged DOP-DEDA at pH 4.5²²⁾ during the microfluidic process. In addition, considering that BSA is an ellipsoid with a major axis of 14 nm and a minor axis of 4 nm in solution,⁴⁰⁾ it has a certain size compared with LNPs, which are approximately 100 nm. Therefore, reducing the total flow rate may help prevent BSA from leaking out of LNPs due to physical effects.

These findings indicate that the lipid solution ratio is the major factor influencing particle size and that lowering the total flow rate leads to improved encapsulation efficiency. These insights are expected to contribute to the development of LNP formulations encapsulating proteins.

Confirmation Study of the Model Formula

In this study, a prediction map was created to set the manufacturing conditions backward from the target property values. The predicted value was close to the observed value, indicating that the prediction ability of the model formula is high. When comparing samples 1, 5, 6, and 7, the lipid solution ratio differed by only 2%, yet a clear difference in particle size was observed. This difference suggests that the lipid solution ratio plays a critical role in controlling particle size.

The particle size was affected by the changes in the lipid solution ratio, and the encapsulation efficiency was affected by the changes in the total flow rate. This finding is important for setting up various experiments in the future, such as observing the pharmacokinetics of BSA-LNPs by varying the particle size. In contrast, under the manufacturing conditions that maximized the encapsulation efficiency (total flow rate of 2.0 mL/min, lipid concentration of 1.5 mM, lipid solution ratio of 30%, predicted particle size of 278 nm, and predicted encapsulation efficiency of 49.4%), BSA-LNPs were aggregated (>1000 nm). As shown in the prediction map, the particle size increases with the increase in encapsulation efficiency, and the possibility of aggregation may increase because of the destabilization of the lipid membrane.⁴¹⁾ Therefore, extreme conditions in the plot in the prediction map may not be favorable, and the total flow rate and lipid solution ratio must be set carefully.

As a DDS carrier, the particle size of LNPs must be adjusted to approximately 100 nm from the viewpoint of cell permeability and blood retention.⁴²⁾ The manufacturing conditions have been established for the BSA-LNP microfluidic process, which yields a particle size of 100 nm and a relatively high encapsulation efficiency. Although the variation in LNP quality among manufacturing batches is often an issue in the practical application of LNPs, the manufacturing conditions set in this study yielded results with little variation in three independent experiments. The basic concept of scale-up techniques in microfluidic technology is to arrange multiple identical channels without changing the flow channel conditions.³⁵⁾ Therefore, the manufacturing conditions obtained in this study are theoretically applicable to any scale. While the lipid composition was fixed in this study for intracellular protein delivery, there is potential for exploring alternative lipid compositions, incorporating helper lipids, and optimizing other additives to further improve protein encapsulation efficiency.

CONCLUSION

In this study, BSA-LNPs have shown favorable properties for intracellular protein delivery. Thus, the microfluidic process of BSA-LNPs was optimized and investigated using the DoE approach. The process parameters affected the particle size and encapsulation efficiency of BSA-LNPs. The derived multiple regression equations were used in the preparation of BSA-LNPs with desirable properties. The optimal lipid/protein loading ratio, aqueous-phase pH, total flow rate, lipid concentration, and lipid-phase/aqueous-phase flow rate ratio in a microfluidic device were also determined. Furthermore, conditions that can stably prepare BSA-LNPs with a particle size of 100 nm and with the most efficient encapsulation efficiency were established.

Acknowledgments

This research was supported by AMED under Grant Number: JP21ak0101171.

Conflict of Interest

The authors declare no conflict of interest.

REFERENCES

• 1)  Leader B, Baca QJ, Golan DE. Protein therapeutics: a summary and pharmacological classification. Nat. Rev. Drug Discov., 7, 21–39 (2008).
• 2)  Ebrahimi SB, Samanta D. Engineering protein-based therapeutics through structural and chemical design. Nat. Commun., 14, 2411 (2023).
• 3)  Wu J, Sahoo JK, Li Y, Xu Q, Kaplan DL. Challenges in delivering therapeutic peptides and proteins: a silk-based solution. J. Control. Release, 345, 176–189 (2022).
• 4)  Tabata Y. Regenerative inductive therapy based on DDS technology of protein and gene. J. Drug Target., 14, 483–495 (2006).
• 5)  Suzuki J, Broeyer F, Cohen A, Takebe M, Burggraaf J, Mizushima Y. Pharmacokinetics of PC-SOD, a lecithinized recombinant superoxide dismutase, after single- and multiple-dose administration to healthy Japanese and Caucasian volunteers. J. Clin. Pharmacol., 48, 184–192 (2008).
• 6)  Mohammadpour F, Kamali H, Gholami L, McCloskey AP, Kesharwani P, Sahebkar A. Solid lipid nanoparticles: a promising tool for insulin delivery. Expert Opin. Drug Deliv., 19, 1577–1595 (2022).
• 7)  Jena D, Srivastava N, Chauhan I, Verma M. Challenges and therapeutic approaches for the protein delivery system: a review. Pharm. Nanotechnol., 12, 391–411 (2024).
• 8)  Wagner A, Vorauer-Uhl K. Liposome technology for industrial purposes. J. Drug Deliv., 2011, 591325 (2011).
• 9)  Maeki M, Uno S, Niwa A, Okada Y, Tokeshi M. Microfluidic technologies and devices for lipid nanoparticle-based RNA delivery. J. Control. Release, 344, 80–96 (2022).
• 10)  Allen TM, Everest JM. Effect of liposome size and drug release properties on pharmacokinetics of encapsulated drug in rats. J. Pharmacol. Exp. Ther., 226, 539–544 (1983).
• 11)  Park K. Controlled drug delivery systems: past forward and future back. J. Control. Release, 190, 3–8 (2014).
• 12)  Ly HH, Daniel S, Soriano SKV, Kis Z, Blakney AK. Optimization of lipid nanoparticles for saRNA expression and cellular activation using a design-of-experiment approach. Mol. Pharm., 19, 1892–1905 (2022).
• 13)  Nakamura K, Aihara K, Ishida T. Importance of process parameters influencing the mean diameters of siRNA-containing lipid nanoparticles (LNPs) on the in vitro activity of prepared LNPs. Biol. Pharm. Bull., 45, 497–507 (2022).
• 14)  Belliveau NM, Huft J, Lin PJ, Chen S, Leung AK, Leaver TJ, Wild AW, Lee JB, Taylor RJ, Tam YK, Hansen CL, Cullis PR. Microfluidic synthesis of highly potent limit-size lipid nanoparticles for in vivo delivery of siRNA. Mol. Ther. Nucleic Acids, 1, e37 (2012).
• 15)  Sakurai Y, Hada T, Harashima H. Scalable preparation of poly(ethylene glycol)-grafted siRNA-loaded lipid nanoparticles using a commercially available fluidic device and tangential flow filtration. J. Biomater. Sci. Polym. Ed., 28, 1086–1096 (2017).
• 16)  Toyota H, Asai T, Oku N. Process optimization by use of design of experiments: Application for liposomalization of FK506. Eur. J. Pharm. Sci., 102, 196–202 (2017).
• 17)  Vogelaar A, Marcotte S, Cheng J, Oluoch B, Zaro J. Use of microfluidics to prepare lipid-based nanocarriers. Pharmaceutics, 15, 1053 (2023).
• 18)  Anderluzzi G, Lou G, Su Y, Perrie Y. Scalable manufacturing processes for solid lipid nanoparticles. Pharm. Nanotechnol., 7, 444–459 (2019).
• 19)  International Conference on Harmonization of Technical Requirements for Registration of Pharmaceuticals for Human Use, 2009. “ICH Harmonized Tripartite Guideline Pharmaceutical Development Q8 (R2).”: <https://www.ema.europa.eu/en/documents/scientific-guideline/international-conference-harmonisation-technical-requirements-registration-pharmaceuticals-human-use-considerations-ich-guideline-q8-r2-pharmaceutical-development-step-5_en.pdf>, accessed 1 October, 2024.
• 20)  Singh B, Dahiya M, Saharan V, Ahuja N. Optimizing drug delivery systems using systematic “design of experiments.” Part II: retrospect and prospects. Crit. Rev. Ther. Drug Carrier Syst., 22, 215–294 (2005).
• 21)  Tavares Luiz M, Santos Rosa Viegas J, Palma Abriata J, Viegas F, Testa Moura de Carvalho Vicentini F, Lopes Badra Bentley MV, Chorilli M, Maldonado Marchetti J, Tapia-Blácido DR. Design of experiments (DoE) to develop and to optimize nanoparticles as drug delivery systems. Eur. J. Pharm. Biopharm., 165, 127–148 (2021).
• 22)  Hirai Y, Saeki R, Song F, Koide H, Fukata N, Tomita K, Maeda N, Oku N, Asai T. Charge-reversible lipid derivative: A novel type of pH-responsive lipid for nanoparticle-mediated siRNA delivery. Int. J. Pharm., 585, 119479 (2020).
• 23)  Hirai Y, Hirose H, Imanishi M, Asai T, Futaki S. Cytosolic protein delivery using pH-responsive, charge-reversible lipid nanoparticles. Sci. Rep., 11, 19896 (2021).
• 24)  Medda L, Monduzzi M, Salis A. The molecular motion of bovine serum albumin under physiological conditions is ion specific. Chem. Commun. (Camb.), 51, 6663–6666 (2015).
• 25)  Wang J, Zhang B. Bovine serum albumin as a versatile platform for cancer imaging and therapy. Curr. Med. Chem., 25, 2938–2953 (2018).
• 26)  Okamoto Y, Taguchi K, Yamasaki K, Sakuragi M, Kuroda S, Otagiri M. Albumin-encapsulated liposomes: a novel drug delivery carrier with hydrophobic drugs encapsulated in the inner aqueous core. J. Pharm. Sci., 107, 436–445 (2018).
• 27)  Wang D, Liang N, Kawashima Y, Cui F, Yan P, Sun S. Biotin-modified bovine serum albumin nanoparticles as a potential drug delivery system for paclitaxel. J. Mater. Sci., 54, 8613–8626 (2019).
• 28)  Maeki M, Fujishima Y, Sato Y, Yasui T, Kaji N, Ishida A, Tani H, Baba Y, Harashima H, Tokeshi M. Understanding the formation mechanism of lipid nanoparticles in microfluidic devices with chaotic micromixers. PLOS ONE, 12, e0187962 (2017).
• 29)  “Notification No: 2020. 0106-1 of the Pharmaceuticals and Medical Devices Agency, Japan.”: <https://www.pmda.go.jp/files/000233545.pdf>, accessed 1 October, 2024.
• 30)  Eltoukhy AA, Chen D, Veiseh O, Pelet JM, Yin H, Dong Y, Anderson DG. Nucleic acid-mediated intracellular protein delivery by lipid-like nanoparticles. Biomaterials, 35, 6454–6461 (2014).
• 31)  Jayaraman M, Ansell SM, Mui BL, Tam YK, Chen J, Du X, Butler D, Eltepu L, Matsuda S, Narayanannair JK, Rajeev KG, Hafez IM, Akinc A, Maier MA, Tracy MA, Cullis PR, Madden TD, Manoharan M, Hope MJ. Maximizing the potency of siRNA lipid nanoparticles for hepatic gene silencing in vivo. Angew. Chem. Int. Ed., Engl., 51, 8529–8533 (2012).
• 32)  Alabi CA, Love KT, Sahay G, Yin H, Luly KM, Langer R, Anderson DG. Multiparametric approach for the evaluation of lipid nanoparticles for siRNA delivery. Proc. Natl. Acad. Sci. U.S.A., 110, 12881–12886 (2013).
• 33)  Schoenmaker L, Witzigmann D, Kulkarni JA, Verbeke R, Kersten G, Jiskoot W, Crommelin DJA. mRNA-lipid nanoparticle COVID-19 vaccines: structure and stability. Int. J. Pharm., 601, 120586 (2021).
• 34)  Maeki M, Saito T, Sato Y, Yasui T, Kaji N, Ishida A, Tani H, Baba Y, Harashima H, Tokeshi M. A strategy for synthesis of lipid nanoparticles using microfluidic devices with a mixer structure. RSC Adv., 5, 46181–46185 (2015).
• 35)  Roces CB, Lou G, Jain N, Abraham S, Thomas A, Halbert GW, Perrie Y. Manufacturing considerations for the development of lipid nanoparticles using microfluidics. Pharmaceutics, 12, 1095 (2020).
• 36)  Takeda K, Moriyama Y. Interaction of Surfactant with Protein. Oleoscience, 11, 3–10 (2011).
• 37)  Webb MS, Saxon D, Wong FM, Lim HJ, Wang Z, Bally MB, Choi LS, Cullis PR, Mayer LD. Comparison of different hydrophobic anchors conjugated to poly(ethylene glycol): effects on the pharmacokinetics of liposomal vincristine. Biochim. Biophys. Acta Biomembr., 1372, 272–282 (1998).
• 38)  Soema PC, Willems GJ, Jiskoot W, Amorij JP, Kersten GF. Predicting the influence of liposomal lipid composition on liposome size, zeta potential and liposome-induced dendritic cell maturation using a design of experiments approach. Eur. J. Pharm. Biopharm., 94, 427–435 (2015).
• 39)  Forbes N, Hussain MT, Briuglia ML, Edwards DP, Horst JHT, Szita N, Perrie Y. Rapid and scale-independent microfluidic manufacture of liposomes entrapping protein incorporating in-line purification and at-line size monitoring. Int. J. Pharm., 556, 68–81 (2019).
• 40)  Nakagawa T, Monkawa A, Kamiya N, Tsunomori F, Kagi H. Development of biosensing system using dynamic light scattering. Bulletin of Tokyo Metropolitan Industrial Technology Research Institute, 6, 108–109 (2011).
• 41)  Tamba Y, Tanaka T, Yahagi T, Yamashita Y, Yamazaki M. Stability of giant unilamellar vesicles and large unilamellar vesicles of liquid-ordered phase membranes in the presence of Triton X-100. Biochim. Biophys. Acta Biomembr., 1667, 1–6 (2004).
• 42)  Kikuchi H. Unlimited potential of liposomes and their application to medical field. Membrane, 34, 328–335 (2009).